diff --git a/Makefile b/Makefile
index a77825e8f9a..ab41116795d 100644
--- a/Makefile
+++ b/Makefile
@@ -181,7 +181,7 @@ ultrix-gcc:
 
 # Rules for making release tarballs
 
-VERSION=7.5-devel
+VERSION=7.6-devel
 DIRECTORY = Mesa-$(VERSION)
 LIB_NAME = MesaLib-$(VERSION)
 DEMO_NAME = MesaDemos-$(VERSION)
diff --git a/SConstruct b/SConstruct
index 1e5fd71adce..89144f7579d 100644
--- a/SConstruct
+++ b/SConstruct
@@ -41,16 +41,16 @@ else:
 	default_drivers = 'all'
 	default_winsys = 'all'
 
-opts = Options('config.py')
+opts = Variables('config.py')
 common.AddOptions(opts)
-opts.Add(ListOption('statetrackers', 'state trackers to build', default_statetrackers,
+opts.Add(ListVariable('statetrackers', 'state trackers to build', default_statetrackers,
                      ['mesa', 'python']))
-opts.Add(ListOption('drivers', 'pipe drivers to build', default_drivers,
+opts.Add(ListVariable('drivers', 'pipe drivers to build', default_drivers,
                      ['softpipe', 'failover', 'i915simple', 'i965simple', 'cell', 'trace', 'r300']))
-opts.Add(ListOption('winsys', 'winsys drivers to build', default_winsys,
+opts.Add(ListVariable('winsys', 'winsys drivers to build', default_winsys,
                      ['xlib', 'intel', 'gdi', 'radeon']))
 
-opts.Add(EnumOption('MSVS_VERSION', 'MS Visual C++ version', None, allowed_values=('7.1', '8.0', '9.0')))
+opts.Add(EnumVariable('MSVS_VERSION', 'MS Visual C++ version', None, allowed_values=('7.1', '8.0', '9.0')))
 
 env = Environment(
 	options = opts,
diff --git a/bin/mklib b/bin/mklib
index a3e826abac9..2fd95ba7751 100755
--- a/bin/mklib
+++ b/bin/mklib
@@ -176,6 +176,23 @@ if [ ${ARCH} = "auto" ] ; then
 fi
 
 
+if [ $STATIC = 1 ]; then
+    # filter out linker options inside object list
+    NEWOBJECTS=""
+    for OBJ in $OBJECTS ; do
+	case $OBJ in
+	    -Wl,*)
+		echo "mklib: warning: ignoring $OBJ for static library"
+		;;
+	    *)
+		NEWOBJECTS="$NEWOBJECTS $OBJ"
+		;;
+	esac
+    done
+    OBJECTS=$NEWOBJECTS
+fi
+
+
 #
 # Error checking
 #
@@ -264,18 +281,21 @@ case $ARCH in
 	    # expand any .a objects into constituent .o files.
 	    NEWOBJECTS=""
 	    DELETIA=""
-	    for OBJ in ${OBJECTS} ; do
-		if [ `expr match $OBJ '.*\.a'` -gt 0 ] ; then
-		    # extract the .o files from this .a archive
-		    FILES=`ar t $OBJ`
-		    ar x $OBJ
-		    NEWOBJECTS="$NEWOBJECTS $FILES"
-		    # keep track of temporary .o files and delete them below
-		    DELETIA="$DELETIA $FILES"
-		else
-		    # ordinary .o file
-		    NEWOBJECTS="$NEWOBJECTS $OBJ"
-		fi
+	    for OBJ in $OBJECTS ; do
+		case $OBJ in
+		    *.a)
+			# extract the .o files from this .a archive
+			FILES=`ar t $OBJ`
+			ar x $OBJ
+			NEWOBJECTS="$NEWOBJECTS $FILES"
+			# keep track of temporary .o files and delete them below
+			DELETIA="$DELETIA $FILES"
+			;;
+		    *)
+			# ordinary .o file
+			NEWOBJECTS="$NEWOBJECTS $OBJ"
+			;;
+		esac
 	    done
 
             # make lib
diff --git a/common.py b/common.py
index f1c6372abde..e57429eb75e 100644
--- a/common.py
+++ b/common.py
@@ -46,13 +46,13 @@ else:
 
 def AddOptions(opts):
 	try:
-		from SCons.Options.BoolOption import BoolOption
-	except ImportError:
 		from SCons.Variables.BoolVariable import BoolVariable as BoolOption
-	try:
-		from SCons.Options.EnumOption import EnumOption
 	except ImportError:
+		from SCons.Options.BoolOption import BoolOption
+	try:
 		from SCons.Variables.EnumVariable import EnumVariable as EnumOption
+	except ImportError:
+		from SCons.Options.EnumOption import EnumOption
 	opts.Add(BoolOption('debug', 'debug build', 'no'))
 	opts.Add(BoolOption('profile', 'profile build', 'no'))
 	#opts.Add(BoolOption('quiet', 'quiet command lines', 'no'))
diff --git a/configs/default b/configs/default
index eab36a36775..9c479fccdf9 100644
--- a/configs/default
+++ b/configs/default
@@ -9,7 +9,7 @@ CONFIG_NAME = default
 
 # Version info
 MESA_MAJOR=7
-MESA_MINOR=3
+MESA_MINOR=6
 MESA_TINY=0
 MESA_VERSION = $(MESA_MAJOR).$(MESA_MINOR).$(MESA_TINY)
 
diff --git a/configs/linux-static b/configs/linux-static
index 2fc39ff83ef..907904bda44 100644
--- a/configs/linux-static
+++ b/configs/linux-static
@@ -22,7 +22,5 @@ GLUT_LIB_DEPS =
 GLW_LIB_DEPS = 
 
 # Need to specify all libraries we may need
-APP_LIB_DEPS = -L$(TOP)/$(LIB_DIR) -l$(GLUT_LIB) -l$(GLU_LIB) -Wl,--start-group \
-	-l$(GL_LIB) $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a -Wl,--end-group -lm \
-	-L/usr/X11R6/lib/ -lX11 -lXext -lXmu -lXi -lpthread
-
+APP_LIB_DEPS = -L$(TOP)/$(LIB_DIR) -l$(GLUT_LIB) -l$(GLU_LIB) \
+	-l$(GL_LIB) -lm -L/usr/X11R6/lib/ -lX11 -lXext -lXmu -lXi -lpthread
diff --git a/configs/linux-x86-64-debug b/configs/linux-x86-64-debug
index 46e72c990af..6f631c0e879 100644
--- a/configs/linux-x86-64-debug
+++ b/configs/linux-x86-64-debug
@@ -4,4 +4,5 @@ include $(TOP)/configs/linux-x86-64
 
 CONFIG_NAME = linux-x86-64-debug
 
+OPT_FLAGS = -g
 DEFINES += -DDEBUG -DDEBUG_MATH
diff --git a/configs/linux-x86-static b/configs/linux-x86-static
index 65c92cf3524..16c8731c1d1 100644
--- a/configs/linux-x86-static
+++ b/configs/linux-x86-static
@@ -22,6 +22,5 @@ GLUT_LIB_DEPS =
 GLW_LIB_DEPS = 
 
 # Need to specify all libraries we may need
-APP_LIB_DEPS = -L$(TOP)/$(LIB_DIR) -l$(GLUT_LIB) -l$(GLU_LIB) -Wl,--start-group \
-               -l$(GL_LIB) $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a -Wl,--end-group \
-               $(EXTRA_LIB_PATH) -lX11 -lXext -lXmu -lXt -lXi -lpthread -lstdc++ -lm
+APP_LIB_DEPS = -L$(TOP)/$(LIB_DIR) -l$(GLUT_LIB) -l$(GLU_LIB) \
+	-l$(GL_LIB) $(EXTRA_LIB_PATH) -lX11 -lXext -lXmu -lXt -lXi -lpthread -lstdc++ -lm
diff --git a/configure.ac b/configure.ac
index 231b7e98a70..662a76edf8c 100644
--- a/configure.ac
+++ b/configure.ac
@@ -5,12 +5,8 @@ AC_PREREQ([2.59])
 dnl Versioning - scrape the version from configs/default
 m4_define([mesa_version],
     [m4_esyscmd([${MAKE-make} -s -f bin/version.mk version | tr -d '\n'])])
-m4_ifval(mesa_version,[],[
-    m4_errprint([Error: Failed to get the Mesa version from the output of
-       running `make -f bin/version.mk version'
-])
-    m4_exit([1])
-])
+m4_ifval(mesa_version,,
+    [m4_fatal([Failed to get the Mesa version from `make -f bin/version.mk version`])])
 
 dnl Tell the user about autoconf.html in the --help output
 m4_divert_once([HELP_END], [
@@ -58,15 +54,11 @@ fi
 AC_SUBST([MKDEP_OPTIONS])
 
 dnl Make sure the pkg-config macros are defined
-m4_ifdef([PKG_PROG_PKG_CONFIG],[],[
-    m4_errprint([Error: Could not locate the pkg-config autoconf macros.
-       These are usually located in /usr/share/aclocal/pkg.m4. If your
-       macros are in a different location, try setting the environment
-       variable ACLOCAL="aclocal -I/other/macro/dir" before running
-       autoreconf.
-])
-    m4_exit([1])
-])
+m4_ifndef([PKG_PROG_PKG_CONFIG],
+    [m4_fatal([Could not locate the pkg-config autoconf macros.
+  These are usually located in /usr/share/aclocal/pkg.m4. If your macros
+  are in a different location, try setting the environment variable
+  ACLOCAL="aclocal -I/other/macro/dir" before running autoreconf.])])
 PKG_PROG_PKG_CONFIG()
 
 dnl LIB_DIR - library basename
@@ -87,7 +79,7 @@ dnl Compiler macros
 DEFINES=""
 AC_SUBST([DEFINES])
 case "$host_os" in
-linux*|*-gnu*)
+linux*|*-gnu*|gnu*)
     DEFINES="$DEFINES -D_GNU_SOURCE -DPTHREADS"
     ;;
 solaris*)
@@ -411,7 +403,7 @@ esac
 dnl
 dnl Driver specific build directories
 dnl
-SRC_DIRS="mesa egl glew"
+SRC_DIRS="mesa glew"
 GLU_DIRS="sgi"
 WINDOW_SYSTEM=""
 GALLIUM_DIRS="auxiliary drivers state_trackers"
@@ -751,6 +743,10 @@ if test "$mesa_driver" = dri; then
                 unichrome savage sis swrast"
         fi
         ;;
+    gnu*)
+        DEFINES="$DEFINES -DUSE_EXTERNAL_DXTN_LIB=1 -DIN_DRI_DRIVER"
+        DEFINES="$DEFINES -DGLX_INDIRECT_RENDERING -DHAVE_ALIAS"
+	;;
     solaris*)
         DEFINES="$DEFINES -DUSE_EXTERNAL_DXTN_LIB=1 -DIN_DRI_DRIVER"
         DEFINES="$DEFINES -DGLX_INDIRECT_RENDERING"
@@ -870,14 +866,23 @@ AC_SUBST([OSMESA_PC_LIB_PRIV])
 dnl
 dnl EGL configuration
 dnl
-if test "$x11_pkgconfig" = yes; then
-    PKG_CHECK_MODULES([EGL],[x11])
-    EGL_LIB_DEPS="$EGL_LIBS"
-else
-    # should check these...
-    EGL_LIB_DEPS="$X_LIBS -lX11"
+AC_ARG_ENABLE([egl],
+    [AS_HELP_STRING([--disable-egl],
+        [disable EGL library @<:@default=enabled@:>@])],
+    [enable_egl="$enableval"],
+    [enable_egl=yes])
+if test "x$enable_egl" = xyes; then
+    SRC_DIRS="$SRC_DIRS egl"
+
+    if test "$x11_pkgconfig" = yes; then
+        PKG_CHECK_MODULES([EGL], [x11])
+        EGL_LIB_DEPS="$EGL_LIBS"
+    else
+        # should check these...
+        EGL_LIB_DEPS="$X_LIBS -lX11"
+    fi
+    EGL_LIB_DEPS="$EGL_LIB_DEPS $DLOPEN_LIBS"
 fi
-EGL_LIB_DEPS="$EGL_LIB_DEPS $DLOPEN_LIBS"
 AC_SUBST([EGL_LIB_DEPS])
 
 dnl
@@ -1121,7 +1126,7 @@ yes)
         GALLIUM_STATE_TRACKERS_DIRS=glx
         ;;
     dri)
-        GALLIUM_STATE_TRACKERS_DIRS=egl
+        test "x$enable_egl" = xyes && GALLIUM_STATE_TRACKERS_DIRS=egl
         ;;
     esac
     ;;
@@ -1131,6 +1136,10 @@ yes)
     for tracker in $state_trackers; do
         test -d "$srcdir/src/gallium/state_trackers/$tracker" || \
             AC_MSG_ERROR([state tracker '$tracker' doesn't exist])
+
+        if test "$tracker" = egl && test "x$enable_egl" != xyes; then
+            AC_MSG_ERROR([cannot build egl state tracker without EGL library])
+        fi
     done
     GALLIUM_STATE_TRACKERS_DIRS="$state_trackers"
     ;;
@@ -1246,6 +1255,7 @@ dnl Libraries
 echo ""
 echo "        Shared libs:     $enable_shared"
 echo "        Static libs:     $enable_static"
+echo "        EGL:             $enable_egl"
 echo "        GLU:             $enable_glu"
 echo "        GLw:             $enable_glw (Motif: $enable_motif)"
 echo "        glut:            $enable_glut"
diff --git a/docs/download.html b/docs/download.html
index 37b0b3da479..cf1cc21d374 100644
--- a/docs/download.html
+++ b/docs/download.html
@@ -8,12 +8,6 @@
 
 <H1>Downloading</H1>
 
-<p>
-Current development release: <b>7.3</b>
-<br>
-Last stable release: <b>7.2</b>
-</p>
-
 <p>
 Primary download site:
 <a href="http://sourceforge.net/project/showfiles.php?group_id=3"
diff --git a/docs/news.html b/docs/news.html
index b177f3bca25..98cee9b8b83 100644
--- a/docs/news.html
+++ b/docs/news.html
@@ -11,6 +11,26 @@
 <H1>News</H1>
 
 
+<h2>May tbd, 2009</h2>
+<p>
+<a href="relnotes-7.5.html">Mesa 7.5</a> is released.
+</p>
+
+
+<h2>April 18, 2009</h2>
+<p>
+<a href="relnotes-7.4.1.html">Mesa 7.4.1</a> is released.
+This is a stable release fixing bugs since the 7.4 release.
+</p>
+
+
+<h2>March 27, 2009</h2>
+<p>
+<a href="relnotes-7.4.html">Mesa 7.4</a> is released.
+This is a stable release fixing bugs since the 7.3 release.
+</p>
+
+
 <h2>January 22, 2009</h2>
 <p>
 <a href="relnotes-7.3.html">Mesa 7.3</a> is released.
diff --git a/docs/relnotes-7.4.1.html b/docs/relnotes-7.4.1.html
new file mode 100644
index 00000000000..40f4fb1722d
--- /dev/null
+++ b/docs/relnotes-7.4.1.html
@@ -0,0 +1,79 @@
+<HTML>
+
+<TITLE>Mesa Release Notes</TITLE>
+
+<head><link rel="stylesheet" type="text/css" href="mesa.css"></head>
+
+<BODY>
+
+<body bgcolor="#eeeeee">
+
+<H1>Mesa 7.4.1 Release Notes / 18 April 2009</H1>
+
+<p>
+Mesa 7.4.1 is a stable development release fixing bugs since the 7.4 release.
+</p>
+<p>
+Mesa 7.4.1 implements the OpenGL 2.1 API, but the version reported by
+glGetString(GL_VERSION) depends on the particular driver being used.
+Some drivers don't support all the features required in OpenGL 2.1.
+</p>
+<p>
+See the <a href="install.html">Compiling/Installing page</a> for prerequisites
+for DRI hardware acceleration.
+</p>
+
+
+<h2>MD5 checksums</h2>
+<pre>
+0c3a72f3295a53a134c04bd7d209ea62  MesaLib-7.4.1.tar.gz
+423260578b653818ba66c2fcbde6d7ad  MesaLib-7.4.1.tar.bz2
+84f78b154d4bd5c3ecc42eeff2e56676  MesaLib-7.4.1.zip
+aa0ad323e59d6d10ff33ac0dde462a60  MesaDemos-7.4.1.tar.gz
+1e169fb6abc2b45613f1c98a82dfe690  MesaDemos-7.4.1.tar.bz2
+294e42be2d74176596c994ec23322fcf  MesaDemos-7.4.1.zip
+92373bfa48e7b68dddf356e86b0e5699  MesaGLUT-7.4.1.tar.gz
+336f3824b578b072211e0beecf4f04f4  MesaGLUT-7.4.1.tar.bz2
+20751388d8ef16b42d25d9e3d705d101  MesaGLUT-7.4.1.zip
+</pre>
+
+
+<h2>Bug fixes</h2>
+<ul>
+<li>Fixed a two-sided lighting bug in fixed-function-to-GPU code generation
+<li>Fixed some Darwin issues (Jeremy Huddleston)
+<li>Indexing the GLSL gl_EyePlane[] or gl_ObjectPlane[] arrays with a variable
+    was broken, bug 20986
+<li>Fixed incorrect texture unit bias in TXB instruction
+<li>glTexParameter settings weren't always propogated to drivers
+<li>Assorted vertex/fragment program bug fixes
+<li>Fixed point rendering in software rasterizer
+<li>Fixed potential deadlock in object hash functions
+<li>Fix a couple bugs surrounding front-buffer rendering with DRI2, but this
+    is not quite complete.
+<li>Fixed glPopAttrib() bug when restoring user clip planes
+</ul>
+
+
+
+<h2>Driver Status</h2>
+
+<pre>
+Driver			Status
+----------------------	----------------------
+DRI drivers		varies with the driver
+XMesa/GLX (on Xlib)	implements OpenGL 2.1
+OSMesa (off-screen)	implements OpenGL 2.1
+Windows/Win32		implements OpenGL 2.1
+Glide (3dfx Voodoo1/2)	implements OpenGL 1.3
+SVGA			unsupported
+Wind River UGL		unsupported
+DJGPP			unsupported
+GGI			unsupported
+BeOS			unsupported
+Allegro			unsupported
+D3D			unsupported
+</pre>
+
+</body>
+</html>
diff --git a/docs/relnotes-7.4.html b/docs/relnotes-7.4.html
index 60ada854c16..55ba019b222 100644
--- a/docs/relnotes-7.4.html
+++ b/docs/relnotes-7.4.html
@@ -8,7 +8,7 @@
 
 <body bgcolor="#eeeeee">
 
-<H1>Mesa 7.4 Release Notes / date TBD</H1>
+<H1>Mesa 7.4 Release Notes / 27 March 2009</H1>
 
 <p>
 Mesa 7.4 is a stable development release fixing bugs since the 7.3 release.
@@ -20,28 +20,48 @@ Some drivers don't support all the features required in OpenGL 2.1.
 </p>
 <p>
 See the <a href="install.html">Compiling/Installing page</a> for prerequisites
-for DRI ardware acceleration.
+for DRI hardware acceleration.
 </p>
 
 
 <h2>MD5 checksums</h2>
 <pre>
-tbd
+ed6bd7437177307e51e16d0c7c381dfa  MesaLib-7.4.tar.gz
+7ecddb341a2691e0dfdb02f697109834  MesaLib-7.4.tar.bz2
+433e823f8245f9fd5f397e7b719a8e47  MesaLib-7.4.zip
+656eee6128016fb237e01aa8dabbc703  MesaDemos-7.4.tar.gz
+02816f10f30b1dc5e069e0f68c177c98  MesaDemos-7.4.tar.bz2
+44a70d6db4aa4c64ecc47871b6aceee8  MesaDemos-7.4.zip
+25f80db4f8645cd3e58e2c9af53ec341  MesaGLUT-7.4.tar.gz
+04ec01caebde44f5b0d619f00716b368  MesaGLUT-7.4.tar.bz2
+019dc213baecaa3cb1278847d41b8591  MesaGLUT-7.4.zip
 </pre>
 
 
 <h2>New features</h2>
 <ul>
+<li>Added MESA_GLX_FORCE_DIRECT env var for Xlib/software driver
+<li>GLSL version 1.20 is returnd by the GL_SHADING_LANGUAGE_VERSION query
 </ul>
 
 
 <h2>Bug fixes</h2>
 <ul>
+<li>glGetActiveUniform() returned wrong size for some array types
+<li>Fixed some error checking in glUniform()
+<li>Fixed a potential glTexImage('proxy target') segfault
+<li>Fixed bad reference counting for 1D/2D texture arrays
+<li>Fixed VBO + glPush/PopClientAttrib() bug #19835
 <li>Assorted i965 driver bug fixes
+<li>Fixed a Windows compilation failure in s_triangle.c
+<li>Fixed a GLSL array indexing bug
+<li>Fixes for building on Haiku
 </ul>
 
 <h2>Changes</h2>
 <ul>
+<li>Updated GL/glxext.h to version 48
+<li>Assorted updates for building on Solaris
 </ul>
 
 
diff --git a/docs/relnotes-7.5.html b/docs/relnotes-7.5.html
index bfeea3f433a..3de243d1b24 100644
--- a/docs/relnotes-7.5.html
+++ b/docs/relnotes-7.5.html
@@ -13,7 +13,7 @@
 <p>
 Mesa 7.5 is a new development release.
 People who are concerned with stability and reliability should stick
-with the 7.4.x branch or wait for Mesa 7.6.
+with the 7.4.x branch or wait for Mesa 7.5.1.
 </p>
 <p>
 The main new feature of Mesa 7.5 is the
@@ -27,7 +27,14 @@ Some drivers don't support all the features required in OpenGL 2.1.
 </p>
 <p>
 See the <a href="install.html">Compiling/Installing page</a> for prerequisites
-for DRI ardware acceleration.
+for DRI hardware acceleration.
+</p>
+<p>
+Note that the Mesa project is no longer using odd/even version numbers
+to indicate development/stable releases.
+The so-called development releases have been fairly stable.
+If you're especially concerned with stability you should probably look for
+"point" releases such as 7.5.1 which will be a bug-fix release.
 </p>
 
 
@@ -72,7 +79,7 @@ including GL_ATI_separate_stencil, GL_EXT_stencil_two_side and OpenGL 2.0
 <ul>
 <li>Remove support for GL_SGIX_shadow, GL_SGIX_shadow_ambient and
 GL_SGIX_depth_texture extensions.  Superseded by the ARB versions.
-<li>Omittd some old Mesa demos from the release tarballs, added some others.
+<li>Omitted some old Mesa demos from the release tarballs, added some others.
 </ul>
 
 </body>
diff --git a/docs/relnotes-7.6.html b/docs/relnotes-7.6.html
new file mode 100644
index 00000000000..77ce013d436
--- /dev/null
+++ b/docs/relnotes-7.6.html
@@ -0,0 +1,61 @@
+<HTML>
+
+<TITLE>Mesa Release Notes</TITLE>
+
+<head><link rel="stylesheet" type="text/css" href="mesa.css"></head>
+
+<BODY>
+
+<body bgcolor="#eeeeee">
+
+<H1>Mesa 7.6 Release Notes / date TBD</H1>
+
+<p>
+Mesa 7.6 is a new development release.
+People who are concerned with stability and reliability should stick
+with a previous release or wait for Mesa 7.6.1.
+</p>
+<p>
+Mesa 7.6 implements the OpenGL 2.1 API, but the version reported by
+glGetString(GL_VERSION) depends on the particular driver being used.
+Some drivers don't support all the features required in OpenGL 2.1.
+</p>
+<p>
+See the <a href="install.html">Compiling/Installing page</a> for prerequisites
+for DRI hardware acceleration.
+</p>
+<p>
+Note that the Mesa project is no longer using odd/even version numbers
+to indicate development/stable releases.
+The so-called development releases have been fairly stable.
+If you're especially concerned with stability you should probably look for
+"point" releases such as 7.5.1 which will be a bug-fix release.
+</p>
+
+
+<h2>MD5 checksums</h2>
+<pre>
+tbd
+</pre>
+
+
+<h2>New features</h2>
+<ul>
+<li><a href="openvg.html">OpenVG</a> front-end (state tracker for Gallium).
+This was written by Zack Rusin at Tungsten Graphics.
+</ul>
+
+
+<h2>Bug fixes</h2>
+<ul>
+<li>i965 DRI driver fixes, including support for "unlimited" size constant
+    buffers (GLSL uniforms)
+</ul>
+
+
+<h2>Changes</h2>
+<ul>
+</ul>
+
+</body>
+</html>
diff --git a/docs/relnotes.html b/docs/relnotes.html
index c5ed109390f..936f565236d 100644
--- a/docs/relnotes.html
+++ b/docs/relnotes.html
@@ -20,7 +20,9 @@ The release notes summarize what's new or changed in each Mesa release.
 </p>
 
 <UL>
+<LI><A HREF="relnotes-7.6.html">7.6 release notes</A>
 <LI><A HREF="relnotes-7.5.html">7.5 release notes</A>
+<LI><A HREF="relnotes-7.4.1.html">7.4.1 release notes</A>
 <LI><A HREF="relnotes-7.4.html">7.4 release notes</A>
 <LI><A HREF="relnotes-7.3.html">7.3 release notes</A>
 <LI><A HREF="relnotes-7.2.html">7.2 release notes</A>
diff --git a/include/GL/internal/dri_interface.h b/include/GL/internal/dri_interface.h
index 335bf62a801..910c9166b5e 100644
--- a/include/GL/internal/dri_interface.h
+++ b/include/GL/internal/dri_interface.h
@@ -41,7 +41,7 @@
 #define DRI_INTERFACE_H
 
 /* For archs with no drm.h */
-#if !defined(__APPLE__) && !defined(__CYGWIN__)
+#if !defined(__APPLE__) && !defined(__CYGWIN__) && !defined(__GNU__)
 #include <drm.h>
 #else
 typedef unsigned int drm_context_t;
@@ -649,6 +649,7 @@ struct __DRIswrastExtensionRec {
 #define __DRI_BUFFER_ACCUM		6
 #define __DRI_BUFFER_FAKE_FRONT_LEFT	7
 #define __DRI_BUFFER_FAKE_FRONT_RIGHT	8
+#define __DRI_BUFFER_DEPTH_STENCIL	9  /**< Only available with DRI2 1.1 */
 
 struct __DRIbufferRec {
     unsigned int attachment;
@@ -659,7 +660,7 @@ struct __DRIbufferRec {
 };
 
 #define __DRI_DRI2_LOADER "DRI_DRI2Loader"
-#define __DRI_DRI2_LOADER_VERSION 2
+#define __DRI_DRI2_LOADER_VERSION 3
 struct __DRIdri2LoaderExtensionRec {
     __DRIextension base;
 
@@ -680,6 +681,31 @@ struct __DRIdri2LoaderExtensionRec {
      *                       into __DRIdri2ExtensionRec::createNewDrawable
      */
     void (*flushFrontBuffer)(__DRIdrawable *driDrawable, void *loaderPrivate);
+
+
+    /**
+     * Get list of buffers from the server
+     *
+     * Gets a list of buffer for the specified set of attachments.  Unlike
+     * \c ::getBuffers, this function takes a list of attachments paired with
+     * opaque \c unsigned \c int value describing the format of the buffer.
+     * It is the responsibility of the caller to know what the service that
+     * allocates the buffers will expect to receive for the format.
+     *
+     * \param driDrawable    Drawable whose buffers are being queried.
+     * \param width          Output where the width of the buffers is stored.
+     * \param height         Output where the height of the buffers is stored.
+     * \param attachments    List of pairs of attachment ID and opaque format
+     *                       requested for the drawable.
+     * \param count          Number of attachment / format pairs stored in
+     *                       \c attachments.
+     * \param loaderPrivate  Loader's private data that was previously passed
+     *                       into __DRIdri2ExtensionRec::createNewDrawable.
+     */
+    __DRIbuffer *(*getBuffersWithFormat)(__DRIdrawable *driDrawable,
+					 int *width, int *height,
+					 unsigned int *attachments, int count,
+					 int *out_count, void *loaderPrivate);
 };
 
 /**
diff --git a/progs/demos/dinoshade.c b/progs/demos/dinoshade.c
index 451da2ec895..cbf8751e257 100644
--- a/progs/demos/dinoshade.c
+++ b/progs/demos/dinoshade.c
@@ -382,7 +382,7 @@ static GLfloat floorShadow[4][4];
 static void
 redraw(void)
 {
-  int start, end;
+  int start = 0, end = 0;
 
   if (reportSpeed) {
     start = glutGet(GLUT_ELAPSED_TIME);
@@ -624,6 +624,7 @@ redraw(void)
     glFinish();
     end = glutGet(GLUT_ELAPSED_TIME);
     printf("Speed %.3g frames/sec (%d ms)\n", 1000.0/(end-start), end-start);
+    fflush(stdout);
   }
 
   glutSwapBuffers();
@@ -878,6 +879,7 @@ main(int argc, char **argv)
       polygonOffsetVersion = MISSING;
       printf("\ndinoshine: Missing polygon offset.\n");
       printf("           Expect shadow depth aliasing artifacts.\n\n");
+      fflush(stdout);
     }
   }
 
diff --git a/progs/demos/ipers.c b/progs/demos/ipers.c
index 6e153c04e15..5d82b0dc924 100644
--- a/progs/demos/ipers.c
+++ b/progs/demos/ipers.c
@@ -236,11 +236,28 @@ special(int k, int x, int y)
    }
 }
 
+static void
+cleanup(void)
+{
+   int i;
+
+   glDeleteTextures(1, &t1id);
+   glDeleteTextures(1, &t2id);
+
+   glDeleteLists(skydlist, 1);
+   for (i = 0; i < MAX_LOD; i++) {
+      glDeleteLists(LODdlist[i], 1);
+      glDeleteLists(LODnumpoly[i], 1);
+   }
+}
+
+
 static void
 key(unsigned char k, int x, int y)
 {
    switch (k) {
    case 27:
+      cleanup();
       exit(0);
       break;
 
@@ -707,6 +724,7 @@ main(int ac, char **av)
    glutIdleFunc(draw);
 
    glutMainLoop();
+   cleanup();
 
    return 0;
 }
diff --git a/progs/demos/readpix.c b/progs/demos/readpix.c
index c0aac2272f7..bbb3a68effa 100644
--- a/progs/demos/readpix.c
+++ b/progs/demos/readpix.c
@@ -17,6 +17,7 @@
 #define IMAGE_FILE "../images/girl.rgb"
 
 static int ImgWidth, ImgHeight;
+static int WinWidth, WinHeight;
 static GLenum ImgFormat;
 static GLubyte *Image = NULL;
 
@@ -27,6 +28,7 @@ static int CPosX, CPosY;  /* copypixels */
 static GLboolean DrawFront = GL_FALSE;
 static GLboolean ScaleAndBias = GL_FALSE;
 static GLboolean Benchmark = GL_FALSE;
+static GLboolean Triangle = GL_FALSE;
 static GLubyte *TempImage = NULL;
 
 #define COMBO 1
@@ -150,11 +152,60 @@ Display( void )
    /* draw original image */
    glRasterPos2i(APosX, 5);
    PrintString("Original");
-   glRasterPos2i(APosX, APosY);
-   glEnable(GL_DITHER);
-   SetupPixelTransfer(GL_FALSE);
-   glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
-   glDrawPixels(ImgWidth, ImgHeight, ImgFormat, GL_UNSIGNED_BYTE, Image);
+   if (!Triangle) {
+      glRasterPos2i(APosX, APosY);
+      glEnable(GL_DITHER);
+      SetupPixelTransfer(GL_FALSE);
+      glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
+      glDrawPixels(ImgWidth, ImgHeight, ImgFormat, GL_UNSIGNED_BYTE, Image);
+   }
+   else {
+      float z = 0;
+
+      glViewport(APosX, APosY, ImgWidth, ImgHeight);
+      glMatrixMode( GL_PROJECTION );
+      glLoadIdentity();
+      glOrtho(-1.0, 1.0, -1.0, 1.0, -1.0, 1.0);
+      glDisable(GL_CULL_FACE);
+
+      /* Red should never be seen
+       */
+      glBegin(GL_POLYGON);
+      glColor3f(1,0,0); 
+      glVertex3f(-2, -2, z);
+      glVertex3f(-2, 2, z);
+      glVertex3f(2, 2, z);
+      glVertex3f(2, -2, z);
+      glEnd();
+
+      /* Blue background
+       */
+      glBegin(GL_POLYGON);
+      glColor3f(.5,.5,1); 
+      glVertex3f(-1, -1, z);
+      glVertex3f(-1, 1, z);
+      glVertex3f(1, 1, z);
+      glVertex3f(1, -1, z);
+      glEnd();
+
+      /* Triangle
+       */
+      glBegin(GL_TRIANGLES);
+      glColor3f(.8,0,0); 
+      glVertex3f(-0.9, -0.9, z);
+      glColor3f(0,.9,0); 
+      glVertex3f( 0.9, -0.9, z);
+      glColor3f(0,0,.7); 
+      glVertex3f( 0.0,  0.9, z);
+      glEnd();
+      
+      glColor3f(1,1,1);
+
+      glViewport( 0, 0, WinWidth, WinHeight );
+      glMatrixMode( GL_PROJECTION );
+      glLoadIdentity();
+      glOrtho( 0.0, WinWidth, 0.0, WinHeight, -1.0, 1.0 );
+   }
 
    /* might try alignment=4 here for testing */
    glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
@@ -218,6 +269,9 @@ Display( void )
 static void
 Reshape( int width, int height )
 {
+   WinWidth = width;
+   WinHeight = height;
+
    glViewport( 0, 0, width, height );
    glMatrixMode( GL_PROJECTION );
    glLoadIdentity();
@@ -236,6 +290,9 @@ Key( unsigned char key, int x, int y )
       case 'b':
          Benchmark = GL_TRUE;
          break;
+      case 't':
+         Triangle = !Triangle;
+         break;
       case 's':
          ScaleAndBias = !ScaleAndBias;
          break;
diff --git a/progs/demos/shadowtex.c b/progs/demos/shadowtex.c
index f10a01ec265..dc5a4bbc480 100644
--- a/progs/demos/shadowtex.c
+++ b/progs/demos/shadowtex.c
@@ -788,6 +788,7 @@ Key(unsigned char key, int x, int y)
          exit(0);
          break;
    }
+   fflush(stdout);
    glutPostRedisplay();
 }
 
@@ -1014,6 +1015,7 @@ PrintHelp(void)
    printf("  <shift> + cursor keys = rotate light source\n");
    if (HaveEXTshadowFuncs)
       printf("  o = cycle through comparison modes\n");
+   fflush(stdout);
 }
 
 
diff --git a/progs/demos/teapot.c b/progs/demos/teapot.c
index 38ede7ac3e1..6bf6e064095 100644
--- a/progs/demos/teapot.c
+++ b/progs/demos/teapot.c
@@ -173,10 +173,20 @@ static void special(int k, int x, int y)
   }
 }
 
+static void cleanup(void)
+{
+  glDeleteTextures(1, &t1id);
+  glDeleteTextures(1, &t2id);
+  glDeleteLists(teapotdlist, 1);
+  glDeleteLists(basedlist, 1);
+  glDeleteLists(lightdlist, 1);
+}
+
 static void key(unsigned char k, int x, int y)
 {
   switch(k) {
   case 27:
+    cleanup();
     exit(0);
     break;
     
@@ -670,6 +680,7 @@ int main(int ac, char **av)
   glutIdleFunc(draw);
 
   glutMainLoop();
+  cleanup();
 
   return 0;
 }
diff --git a/progs/demos/tunnel.c b/progs/demos/tunnel.c
index 6a240580e8a..6981da32988 100644
--- a/progs/demos/tunnel.c
+++ b/progs/demos/tunnel.c
@@ -202,11 +202,19 @@ special(int k, int x, int y)
    }
 }
 
+static void
+cleanup(void)
+{
+   glDeleteTextures(1, &t1id);
+   glDeleteTextures(1, &t2id);
+}
+
 static void
 key(unsigned char k, int x, int y)
 {
    switch (k) {
    case 27:
+      cleanup();
       exit(0);
       break;
 
@@ -531,5 +539,6 @@ main(int ac, char **av)
 
    glutMainLoop();
 
+   cleanup();
    return 0;
 }
diff --git a/progs/demos/tunnel2.c b/progs/demos/tunnel2.c
index f4171a8346d..0288ea0f8ce 100644
--- a/progs/demos/tunnel2.c
+++ b/progs/demos/tunnel2.c
@@ -200,6 +200,13 @@ special(int k, int x, int y)
    }
 }
 
+static void
+cleanup(void)
+{
+   glDeleteTextures(1, &t1id);
+   glDeleteTextures(1, &t2id);
+}
+
 static void
 key(unsigned char k, int x, int y)
 {
@@ -207,6 +214,7 @@ key(unsigned char k, int x, int y)
    case 27:
       glutDestroyWindow(channel[0]);
       glutDestroyWindow(channel[1]);
+      cleanup();
       exit(0);
       break;
 
@@ -602,6 +610,7 @@ main(int ac, char **av)
    calcposobs();
 
    glutMainLoop();
+   cleanup();
 
    return 0;
 }
diff --git a/progs/demos/vao_demo.c b/progs/demos/vao_demo.c
index ce416712fe2..206e06fc6c7 100644
--- a/progs/demos/vao_demo.c
+++ b/progs/demos/vao_demo.c
@@ -260,6 +260,8 @@ static void Key( unsigned char key, int x, int y )
    (void) y;
    switch (key) {
       case 27:
+         (*delete_vertex_arrays)( 1, & cube_array_obj );
+         (*delete_vertex_arrays)( 1, & oct_array_obj );
          glutDestroyWindow(Win);
          exit(0);
          break;
diff --git a/progs/redbook/polyoff.c b/progs/redbook/polyoff.c
index 2017b4d8eed..de34b2e7675 100644
--- a/progs/redbook/polyoff.c
+++ b/progs/redbook/polyoff.c
@@ -153,6 +153,7 @@ static void Benchmark( float xdiff, float ydiff )
    double seconds, fps;
 
    printf("Benchmarking...\n");
+   fflush(stdout);
 
    draws = 0;
    startTime = glutGet(GLUT_ELAPSED_TIME);
@@ -169,6 +170,7 @@ static void Benchmark( float xdiff, float ydiff )
    seconds = (double) (endTime - startTime) / 1000.0;
    fps = draws / seconds;
    printf("Result:  fps: %g\n", fps);
+   fflush(stdout);
 }
 
 
@@ -263,6 +265,7 @@ void keyboard (unsigned char key, int x, int y)
       default:
          break;
    }
+   fflush(stdout);
 }
 
 static void
diff --git a/progs/tests/.gitignore b/progs/tests/.gitignore
index e6369de3ab2..959ddcc51f6 100644
--- a/progs/tests/.gitignore
+++ b/progs/tests/.gitignore
@@ -16,22 +16,20 @@ blendminmax
 blendsquare
 blendxor
 bufferobj
-bumpmap
 bug_3050
 bug_3101
 bug_3195
 bug_texstore_i8
+bumpmap
 calibrate_rast
 copypixrate
 crossbar
 cva
-dinoshade
 drawbuffers
 extfuncs.h
 exactrast
 fbotest1
 fbotest2
-fbotexture
 fillrate
 floattex
 fog
@@ -40,6 +38,7 @@ fptest1
 fptexture
 getprocaddress
 getproclist.h
+glutfx
 interleave
 invert
 jkrahntest
@@ -49,6 +48,7 @@ mapbufrange
 mapvbo
 minmag
 mipgen
+mipmap_comp
 mipmap_limits
 mipmap_view
 multipal
@@ -56,7 +56,6 @@ no_s3tc
 packedpixels
 pbo
 prog_parameter
-projtex
 quads
 random
 readrate
@@ -64,22 +63,22 @@ readtex.c
 readtex.h
 rubberband
 seccolor
-sharedtex
 shader_api
 shaderutil.c
 shaderutil.h
+sharedtex
 stencil_twoside
-stencil_wrap
 stencilwrap
 stencil_wrap
+streaming_rect
 subtex
 subtexrate
 tex1d
-texcmp
 texcompress2
+texdown
 texfilt
-texgenmix
 texline
+texobj
 texobjshare
 texrect
 texwrap
diff --git a/progs/tests/Makefile b/progs/tests/Makefile
index f5fbf374f7a..628ca415354 100644
--- a/progs/tests/Makefile
+++ b/progs/tests/Makefile
@@ -58,6 +58,7 @@ SOURCES = \
 	mapvbo.c \
 	minmag.c \
 	mipgen.c \
+	mipmap_comp.c \
 	mipmap_limits.c \
 	mipmap_view.c \
 	multipal.c \
diff --git a/progs/tests/SConscript b/progs/tests/SConscript
index 4f22ca735c1..453fcecd9c1 100644
--- a/progs/tests/SConscript
+++ b/progs/tests/SConscript
@@ -81,6 +81,7 @@ progs = [
     'mapvbo',
     'minmag',
     'mipgen',
+    'mipmap_comp',
     'mipmap_limits',
     'mipmap_view',
     'multipal',
diff --git a/progs/tests/floattex.c b/progs/tests/floattex.c
index dd99d836c64..ad14cacdcbb 100644
--- a/progs/tests/floattex.c
+++ b/progs/tests/floattex.c
@@ -1,6 +1,5 @@
 /*
  * Test floating point textures.
- * No actual rendering, yet.
  */
 
 
@@ -103,7 +102,6 @@ Key(unsigned char key, int x, int y)
 }
 
 
-
 static void
 InitTexture(void)
 {
@@ -141,6 +139,8 @@ InitTexture(void)
                 GL_RGB, GL_FLOAT, ftex);
 
 
+   CheckError(__LINE__);
+
    /* sanity checks */
    glGetTexLevelParameteriv(GL_TEXTURE_2D, 0, GL_TEXTURE_RED_TYPE_ARB, &t);
    assert(t == GL_FLOAT);
@@ -152,32 +152,26 @@ InitTexture(void)
    assert(t == GL_FLOAT);
 
    free(image);
-   free(ftex);
-      
    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_REPEAT);
    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_REPEAT);
    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, filter);
    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, filter);
 
-#if 0
-   /* read back the texture and make sure values are correct */
-   glGetTexImage(GL_TEXTURE_2D, 0, GL_RGBA, GL_FLOAT, tex2);
-   CheckError(__LINE__);
-   for (i = 0; i < 16; i++) {
-      for (j = 0; j < 16; j++) {
-         if (tex[i][j][0] != tex2[i][j][0] ||
-             tex[i][j][1] != tex2[i][j][1] ||
-             tex[i][j][2] != tex2[i][j][2] ||
-             tex[i][j][3] != tex2[i][j][3]) {
-            printf("tex[%d][%d] %g %g %g %g != tex2[%d][%d] %g %g %g %g\n",
-                   i, j,
-                   tex[i][j][0], tex[i][j][1], tex[i][j][2], tex[i][j][3],
-                   i, j,
-                   tex2[i][j][0], tex2[i][j][1], tex2[i][j][2], tex2[i][j][3]);
+   if (1) {
+      /* read back the texture and make sure values are correct */
+      GLfloat *tex2 = (GLfloat *)
+         malloc(imgWidth * imgHeight * 4 * sizeof(GLfloat));
+      glGetTexImage(GL_TEXTURE_2D, 0, imgFormat, GL_FLOAT, tex2);
+      CheckError(__LINE__);
+      for (i = 0; i < imgWidth * imgHeight * 4; i++) {
+         if (ftex[i] != tex2[i]) {
+            printf("tex[%d] %g != tex2[%d] %g\n",
+                   i, ftex[i], i, tex2[i]);
          }
       }
    }
-#endif
+
+   free(ftex);
 }
 
 
@@ -193,7 +187,9 @@ CreateProgram(void)
 
    assert(program);
 
-   // InitUniforms(program, Uniforms);
+   glUseProgram_func(program);
+
+   InitUniforms(program, Uniforms);
 
    return program;
 }
@@ -211,8 +207,9 @@ Init(void)
       exit(1);
    }
 
-   if (!glutExtensionSupported("GL_MESAX_texture_float")) {
-      printf("Sorry, this test requires GL_MESAX_texture_float\n");
+   if (!glutExtensionSupported("GL_MESAX_texture_float") &&
+       !glutExtensionSupported("GL_ARB_texture_float")) {
+      printf("Sorry, this test requires GL_MESAX/ARB_texture_float\n");
       exit(1);
    }
 
diff --git a/progs/tests/mipmap_comp.c b/progs/tests/mipmap_comp.c
new file mode 100644
index 00000000000..5842e2b8805
--- /dev/null
+++ b/progs/tests/mipmap_comp.c
@@ -0,0 +1,295 @@
+/* Copyright (c) Mark J. Kilgard, 1994. */
+/*
+ * (c) Copyright 1993, Silicon Graphics, Inc.
+ * ALL RIGHTS RESERVED
+ * Permission to use, copy, modify, and distribute this software for
+ * any purpose and without fee is hereby granted, provided that the above
+ * copyright notice appear in all copies and that both the copyright notice
+ * and this permission notice appear in supporting documentation, and that
+ * the name of Silicon Graphics, Inc. not be used in advertising
+ * or publicity pertaining to distribution of the software without specific,
+ * written prior permission.
+ *
+ * THE MATERIAL EMBODIED ON THIS SOFTWARE IS PROVIDED TO YOU "AS-IS"
+ * AND WITHOUT WARRANTY OF ANY KIND, EXPRESS, IMPLIED OR OTHERWISE,
+ * INCLUDING WITHOUT LIMITATION, ANY WARRANTY OF MERCHANTABILITY OR
+ * FITNESS FOR A PARTICULAR PURPOSE.  IN NO EVENT SHALL SILICON
+ * GRAPHICS, INC.  BE LIABLE TO YOU OR ANYONE ELSE FOR ANY DIRECT,
+ * SPECIAL, INCIDENTAL, INDIRECT OR CONSEQUENTIAL DAMAGES OF ANY
+ * KIND, OR ANY DAMAGES WHATSOEVER, INCLUDING WITHOUT LIMITATION,
+ * LOSS OF PROFIT, LOSS OF USE, SAVINGS OR REVENUE, OR THE CLAIMS OF
+ * THIRD PARTIES, WHETHER OR NOT SILICON GRAPHICS, INC.  HAS BEEN
+ * ADVISED OF THE POSSIBILITY OF SUCH LOSS, HOWEVER CAUSED AND ON
+ * ANY THEORY OF LIABILITY, ARISING OUT OF OR IN CONNECTION WITH THE
+ * POSSESSION, USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ * US Government Users Restricted Rights
+ * Use, duplication, or disclosure by the Government is subject to
+ * restrictions set forth in FAR 52.227.19(c)(2) or subparagraph
+ * (c)(1)(ii) of the Rights in Technical Data and Computer Software
+ * clause at DFARS 252.227-7013 and/or in similar or successor
+ * clauses in the FAR or the DOD or NASA FAR Supplement.
+ * Unpublished-- rights reserved under the copyright laws of the
+ * United States.  Contractor/manufacturer is Silicon Graphics,
+ * Inc., 2011 N.  Shoreline Blvd., Mountain View, CA 94039-7311.
+ *
+ * OpenGL(TM) is a trademark of Silicon Graphics, Inc.
+ */
+
+/* mipmap_comp
+ * Test compressed texture mipmaps
+ *
+ * Based on mipmap_limits
+ */
+
+#include <string.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <GL/glew.h>
+#include <GL/glut.h>
+
+#include "readtex.h"
+
+#define SIZE 16 /* not larger then 16 */
+
+static GLint BaseLevel = 0, MaxLevel = 9;
+static GLfloat MinLod = -1, MaxLod = 9;
+static GLfloat LodBias = 0.0;
+static GLboolean NearestFilter = GL_TRUE;
+static GLuint texImage;
+
+
+static void
+initValues(void)
+{
+   BaseLevel = 0;
+   MaxLevel = 9;
+   MinLod = -1;
+   MaxLod = 2;
+   LodBias = 5.0;
+   NearestFilter = GL_TRUE;
+}
+
+
+static void
+makeImage(int level, int width, int height)
+{
+#if 0
+   GLubyte img[SIZE*SIZE*3];
+   int i, j;
+
+   (void)size;
+   for (i = 0; i < height; i++) {
+      for (j = 0; j < width; j++) {
+         int k = (i * width + j) * 3;
+         img[k + 0] = 255 * ((level + 1) % 2);
+         img[k + 1] = 255 * ((level + 1) % 2);
+         img[k + 2] = 255 * ((level + 1) % 2);
+      }
+   }
+
+   glTexImage2D(GL_TEXTURE_2D, level, GL_COMPRESSED_RGB_S3TC_DXT1_EXT, width, height, 0,
+                GL_RGB, GL_UNSIGNED_BYTE, img);
+#else
+   GLubyte img[128];
+   GLint size[] = {
+      128, /* 16x16 */
+      32,  /*  8x8  */
+      8,   /*  4x4  */
+      8,   /*  2x2  */
+      8,   /*  1x1  */
+   };
+   int i;
+   int value = ((level + 1) % 2) * 0xffffffff;
+   memset(img, 0, 128);
+
+   /* generate black and white mipmap levels */
+   if (value)
+      for (i = 0; i < size[level] / 4; i += 2)
+         ((int*)img)[i] = value;
+
+   glCompressedTexImage2D(GL_TEXTURE_2D, level,
+                          GL_COMPRESSED_RGB_S3TC_DXT1_EXT,
+                          width, height, 0,
+                          size[level], img);
+#endif
+}
+
+
+static void
+makeImages(void)
+{
+   int i, sz;
+
+   for (i = 0, sz = SIZE; sz >= 1; i++, sz /= 2) {
+      makeImage(i, sz, sz);
+      printf("Level %d size: %d x %d\n", i, sz, sz);
+   }
+}
+
+
+static void
+myInit(void)
+{
+
+   initValues();
+
+   glEnable(GL_DEPTH_TEST);
+   glDepthFunc(GL_LESS);
+   glShadeModel(GL_FLAT);
+
+   glTranslatef(0.0, 0.0, -3.6);
+
+   glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
+   glGenTextures(1, &texImage);
+   glBindTexture(GL_TEXTURE_2D, texImage);
+   makeImages();
+
+   glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_REPEAT);
+   glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_REPEAT);
+   glTexEnvf(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_DECAL);
+   glEnable(GL_TEXTURE_2D);
+}
+
+
+static void
+display(void)
+{
+   GLfloat tcm = 1.0;
+   glBindTexture(GL_TEXTURE_2D, texImage);
+
+   printf("BASE_LEVEL=%d  MAX_LEVEL=%d  MIN_LOD=%.2g  MAX_LOD=%.2g  Bias=%.2g  Filter=%s\n",
+         BaseLevel, MaxLevel, MinLod, MaxLod, LodBias,
+         NearestFilter ? "NEAREST" : "LINEAR");
+   glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_BASE_LEVEL, BaseLevel);
+   glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, MaxLevel);
+
+   glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_LOD, MinLod);
+   glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAX_LOD, MaxLod);
+
+   if (NearestFilter) {
+      glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+      glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER,
+            GL_NEAREST_MIPMAP_NEAREST);
+   }
+   else {
+      glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
+      glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER,
+            GL_LINEAR_MIPMAP_LINEAR);
+   }
+
+   glTexEnvf(GL_TEXTURE_FILTER_CONTROL_EXT, GL_TEXTURE_LOD_BIAS_EXT, LodBias);
+
+   glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
+   glBegin(GL_QUADS);
+   glTexCoord2f(0.0, 0.0); glVertex3f(-2.0, -1.0, 0.0);
+   glTexCoord2f(0.0, tcm); glVertex3f(-2.0, 1.0, 0.0);
+   glTexCoord2f(tcm * 3000.0, tcm); glVertex3f(3000.0, 1.0, -6000.0);
+   glTexCoord2f(tcm * 3000.0, 0.0); glVertex3f(3000.0, -1.0, -6000.0);
+   glEnd();
+   glFlush();
+}
+
+
+static void
+myReshape(int w, int h)
+{
+   glViewport(0, 0, w, h);
+   glMatrixMode(GL_PROJECTION);
+   glLoadIdentity();
+   gluPerspective(60.0, 1.0*(GLfloat)w/(GLfloat)h, 1.0, 30000.0);
+   glMatrixMode(GL_MODELVIEW);
+   glLoadIdentity();
+}
+
+
+static void
+key(unsigned char k, int x, int y)
+{
+  (void) x;
+  (void) y;
+  switch (k) {
+  case 'b':
+     BaseLevel--;
+     if (BaseLevel < 0)
+        BaseLevel = 0;
+     break;
+  case 'B':
+     BaseLevel++;
+     if (BaseLevel > 10)
+        BaseLevel = 10;
+     break;
+  case 'm':
+     MaxLevel--;
+     if (MaxLevel < 0)
+        MaxLevel = 0;
+     break;
+  case 'M':
+     MaxLevel++;
+     if (MaxLevel > 10)
+        MaxLevel = 10;
+     break;
+  case 'l':
+     LodBias -= 0.25;
+     break;
+  case 'L':
+     LodBias += 0.25;
+     break;
+  case 'n':
+     MinLod -= 0.25;
+     break;
+  case 'N':
+     MinLod += 0.25;
+     break;
+  case 'x':
+     MaxLod -= 0.25;
+     break;
+  case 'X':
+     MaxLod += 0.25;
+     break;
+  case 'f':
+     NearestFilter = !NearestFilter;
+     break;
+  case ' ':
+     initValues();
+     break;
+  case 27:  /* Escape */
+    exit(0);
+    break;
+  default:
+    return;
+  }
+  glutPostRedisplay();
+}
+
+
+static void
+usage(void)
+{
+   printf("usage:\n");
+   printf("  b/B    decrease/increase GL_TEXTURE_BASE_LEVEL\n");
+   printf("  m/M    decrease/increase GL_TEXTURE_MAX_LEVEL\n");
+   printf("  n/N    decrease/increase GL_TEXTURE_MIN_LOD\n");
+   printf("  x/X    decrease/increase GL_TEXTURE_MAX_LOD\n");
+   printf("  l/L    decrease/increase GL_TEXTURE_LOD_BIAS\n");
+   printf("  f      toggle nearest/linear filtering\n");
+   printf("  SPACE  reset values\n");
+}
+
+
+int
+main(int argc, char** argv)
+{
+    glutInit(&argc, argv);
+    glutInitDisplayMode (GLUT_SINGLE | GLUT_RGB | GLUT_DEPTH);
+    glutInitWindowSize (600, 600);
+    glutCreateWindow (argv[0]);
+    glewInit();
+    myInit();
+    glutReshapeFunc (myReshape);
+    glutDisplayFunc(display);
+    glutKeyboardFunc(key);
+    usage();
+    glutMainLoop();
+    return 0;             /* ANSI C requires main to return int. */
+}
diff --git a/progs/tests/mipmap_view.c b/progs/tests/mipmap_view.c
index 85fc67ac79f..808d348699c 100644
--- a/progs/tests/mipmap_view.c
+++ b/progs/tests/mipmap_view.c
@@ -18,12 +18,27 @@
 
 #define TEXTURE_FILE "../images/arch.rgb"
 
-static int TexWidth = 256, TexHeight = 256;
+#define LEVELS 8
+#define SIZE (1<<LEVELS)
+static int TexWidth = SIZE, TexHeight = SIZE;
 static int WinWidth = 1044, WinHeight = 900;
 static GLfloat Bias = 0.0;
 static GLboolean ScaleQuads = GL_FALSE;
 static GLboolean Linear = GL_FALSE;
 static GLint Win = 0;
+static GLint RenderTextureLevel = 0;
+static GLuint TexObj;
+
+
+
+static void
+CheckError(int line)
+{
+   GLenum err = glGetError();
+   if (err) {
+      printf("GL Error 0x%x at line %d\n", (int) err, line);
+   }
+}
 
 
 
@@ -37,6 +52,178 @@ PrintString(const char *s)
 }
 
 
+
+
+static void
+MipGenTexture( void )
+{
+   /* test auto mipmap generation */
+   GLint width, height, i;
+   GLenum format;
+   GLubyte *image = LoadRGBImage(TEXTURE_FILE, &width, &height, &format);
+   if (!image) {
+      printf("Error: could not load texture image %s\n", TEXTURE_FILE);
+      exit(1);
+   }
+   /* resize to TexWidth x TexHeight */
+   if (width != TexWidth || height != TexHeight) {
+      GLubyte *newImage = malloc(TexWidth * TexHeight * 4);
+      
+      fprintf(stderr, "rescale %d %d to %d %d\n", width, height,
+              TexWidth, TexHeight);
+      fflush(stderr);
+
+      gluScaleImage(format, width, height, GL_UNSIGNED_BYTE, image,
+                    TexWidth, TexHeight, GL_UNSIGNED_BYTE, newImage);
+      free(image);
+      image = newImage;
+   }
+   printf("Using GL_SGIS_generate_mipmap\n");
+   glTexParameteri(GL_TEXTURE_2D, GL_GENERATE_MIPMAP_SGIS, GL_TRUE);
+   glTexImage2D(GL_TEXTURE_2D, 0, format, TexWidth, TexHeight, 0,
+                format, GL_UNSIGNED_BYTE, image);
+   free(image);
+
+   /* make sure mipmap was really generated correctly */
+   width = TexWidth;
+   height = TexHeight;
+   for (i = 0; i < 9; i++) {
+      GLint w, h;
+      glGetTexLevelParameteriv(GL_TEXTURE_2D, i, GL_TEXTURE_WIDTH, &w);
+      glGetTexLevelParameteriv(GL_TEXTURE_2D, i, GL_TEXTURE_HEIGHT, &h);
+      printf("Level %d size: %d x %d\n", i, w, h);
+      assert(w == width);
+      assert(h == height);
+      width /= 2;
+      height /= 2;
+   }
+
+
+   glTexParameteri(GL_TEXTURE_2D, GL_GENERATE_MIPMAP_SGIS, GL_FALSE);
+}
+
+
+
+static void
+ResetTextureLevel( int i )
+{
+   GLubyte tex2d[SIZE*SIZE][4];
+      
+   {
+      GLint Width = TexWidth / (1 << i);
+      GLint Height = TexHeight / (1 << i);
+      GLint s, t;
+         
+      for (s = 0; s < Width; s++) {
+         for (t = 0; t < Height; t++) {
+            tex2d[t*Width+s][0] = ((s / 16) % 2) ? 0 : 255;
+            tex2d[t*Width+s][1] = ((t / 16) % 2) ? 0 : 255;
+            tex2d[t*Width+s][2] = 128;
+            tex2d[t*Width+s][3] = 255;
+         }
+      }
+         
+      glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
+         
+      glTexImage2D(GL_TEXTURE_2D, i, GL_RGB, Width, Height, 0,
+                   GL_RGBA, GL_UNSIGNED_BYTE, tex2d);
+   }
+}
+
+
+static void
+ResetTexture( void )
+{
+#if 0
+   /* This doesn't work so well as the arch texture is 512x512.
+    */
+   LoadRGBMipmaps(TEXTURE_FILE, GL_RGB);
+#else
+   {
+      int i;
+      
+      for (i = 0; i <= LEVELS; i++)
+      {
+         ResetTextureLevel(i);
+      }
+   }
+#endif
+}
+
+
+
+
+
+
+
+static void
+RenderTexture( void )
+{
+   GLenum status;
+   GLuint MyFB;
+
+   fprintf(stderr, "RenderTextureLevel %d\n", RenderTextureLevel);
+   fflush(stderr);
+
+   /* gen framebuffer id, delete it, do some assertions, just for testing */
+   glGenFramebuffersEXT(1, &MyFB);
+   glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, MyFB);
+   assert(glIsFramebufferEXT(MyFB));
+
+   CheckError(__LINE__);
+
+   /* Render color to texture */
+   glFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, 
+                             GL_COLOR_ATTACHMENT0_EXT,
+                             GL_TEXTURE_2D, TexObj, 
+                             RenderTextureLevel);
+
+
+
+   CheckError(__LINE__);
+
+
+   glMatrixMode(GL_PROJECTION);
+   glLoadIdentity();
+   glOrtho(-1.0, 1.0, -1.0, 1.0, 5.0, 25.0);
+   glMatrixMode(GL_MODELVIEW);
+   glLoadIdentity();
+   glTranslatef(0.0, 0.0, -15.0);
+
+   status = glCheckFramebufferStatusEXT(GL_FRAMEBUFFER_EXT);
+   if (status != GL_FRAMEBUFFER_COMPLETE_EXT) {
+      printf("Framebuffer incomplete!!!\n");
+   }
+
+   glViewport(0, 0,
+              TexWidth / (1 << RenderTextureLevel),
+              TexHeight / (1 << RenderTextureLevel));
+
+   glClearColor(0.5, 0.5, 1.0, 0.0);
+   glClear(GL_COLOR_BUFFER_BIT);
+      
+   CheckError(__LINE__);
+
+   glBegin(GL_POLYGON);
+   glColor3f(1, 0, 0);
+   glVertex2f(-1, -1);
+   glColor3f(0, 1, 0);
+   glVertex2f(1, -1);
+   glColor3f(0, 0, 1);
+   glVertex2f(0, 1);
+   glEnd();
+
+
+   /* Bind normal framebuffer */
+   glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, 0);
+   CheckError(__LINE__);
+
+   glDeleteFramebuffersEXT(1, &MyFB);
+   CheckError(__LINE__);
+
+   glClearColor(0, 0, 0, 0);
+}
+
 static void
 Display(void)
 {
@@ -44,6 +231,8 @@ Display(void)
    char str[100];
    int texWidth = TexWidth, texHeight = TexHeight;
 
+   glViewport(0, 0, WinHeight, WinHeight);
+
    glClear(GL_COLOR_BUFFER_BIT);
 
    glMatrixMode(GL_PROJECTION);
@@ -126,7 +315,6 @@ Reshape(int width, int height)
 {
    WinWidth = width;
    WinHeight = height;
-   glViewport(0, 0, width, height);
 }
 
 
@@ -145,6 +333,21 @@ Key(unsigned char key, int x, int y)
       case 'l':
          Linear = !Linear;
          break;
+      case 'v':
+         RenderTextureLevel++;
+         break;
+      case 'V':
+         RenderTextureLevel--;
+         break;
+      case 'r':
+         RenderTexture();
+         break;
+      case 'X':
+         ResetTexture();
+         break;
+      case 'x':
+         ResetTextureLevel(RenderTextureLevel);
+         break;
       case '0':
       case '1':
       case '2':
@@ -160,6 +363,14 @@ Key(unsigned char key, int x, int y)
       case 's':
          ScaleQuads = !ScaleQuads;
          break;
+      case ' ':
+         MipGenTexture();
+         Bias = 0;
+         Linear = 0;
+         RenderTextureLevel = 0;
+         ScaleQuads = 0;
+         break;
+         
       case 27:
          glutDestroyWindow(Win);
          exit(0);
@@ -186,53 +397,13 @@ Init(void)
 
    glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
 
-   if (1) {
-      /* test auto mipmap generation */
-      GLint width, height, i;
-      GLenum format;
-      GLubyte *image = LoadRGBImage(TEXTURE_FILE, &width, &height, &format);
-      if (!image) {
-         printf("Error: could not load texture image %s\n", TEXTURE_FILE);
-         exit(1);
-      }
-      /* resize to TexWidth x TexHeight */
-      if (width != TexWidth || height != TexHeight) {
-         GLubyte *newImage = malloc(TexWidth * TexHeight * 4);
-         gluScaleImage(format, width, height, GL_UNSIGNED_BYTE, image,
-                       TexWidth, TexHeight, GL_UNSIGNED_BYTE, newImage);
-         free(image);
-         image = newImage;
-      }
-      printf("Using GL_SGIS_generate_mipmap\n");
-      glTexParameteri(GL_TEXTURE_2D, GL_GENERATE_MIPMAP_SGIS, GL_TRUE);
-      glTexImage2D(GL_TEXTURE_2D, 0, format, TexWidth, TexHeight, 0,
-                   format, GL_UNSIGNED_BYTE, image);
-      free(image);
-
-      /* make sure mipmap was really generated correctly */
-      width = TexWidth;
-      height = TexHeight;
-      for (i = 0; i < 9; i++) {
-         GLint w, h;
-         glGetTexLevelParameteriv(GL_TEXTURE_2D, i, GL_TEXTURE_WIDTH, &w);
-         glGetTexLevelParameteriv(GL_TEXTURE_2D, i, GL_TEXTURE_HEIGHT, &h);
-         printf("Level %d size: %d x %d\n", i, w, h);
-         assert(w == width);
-         assert(h == height);
-         width /= 2;
-         height /= 2;
-      }
-   }
-   else {
-      if (LoadRGBMipmaps(TEXTURE_FILE, GL_RGB)) {
-         printf("Using gluBuildMipmaps()\n");
-      }
-      else {
-         printf("Error: could not load texture image %s\n", TEXTURE_FILE);
-         exit(1);
-      }
-   }
+   glGenTextures(1, &TexObj);
+   glBindTexture(GL_TEXTURE_2D, TexObj);
 
+   if (1) 
+      MipGenTexture();
+   else
+      ResetTexture();
 
    /* mipmapping required for this extension */
    glTexEnvi(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_MODULATE);
diff --git a/progs/trivial/Makefile b/progs/trivial/Makefile
index 69c71cbaf61..22de83fa790 100644
--- a/progs/trivial/Makefile
+++ b/progs/trivial/Makefile
@@ -30,6 +30,7 @@ SOURCES = \
 	fs-tri.c \
 	line-clip.c \
 	line-cull.c \
+	line-flat.c \
 	line-smooth.c \
 	line-stipple-wide.c \
 	line-userclip-clip.c \
@@ -138,6 +139,7 @@ SOURCES = \
 	tristrip-flat.c \
 	tristrip.c \
 	vbo-drawarrays.c \
+	vbo-noninterleaved.c \
 	vbo-drawelements.c \
 	vbo-drawrange.c \
 	vp-array.c \
@@ -145,6 +147,7 @@ SOURCES = \
 	vp-clip.c \
 	vp-line-clip.c \
 	vp-tri.c \
+	vp-tri-invariant.c \
 	vp-tri-swap.c \
 	vp-tri-tex.c \
 	vp-tri-imm.c \
diff --git a/progs/trivial/SConscript b/progs/trivial/SConscript
index 480630e210e..9a1f3575bdc 100644
--- a/progs/trivial/SConscript
+++ b/progs/trivial/SConscript
@@ -26,6 +26,7 @@ progs = [
 	'fs-tri',
 	'line-clip',
 	'line-cull',
+	'line-flat',
 	'line-smooth',
 	'line-stipple-wide',
 	'line-userclip-clip',
@@ -134,6 +135,7 @@ progs = [
 	'tristrip-flat',
 	'tristrip',
 	'vbo-drawarrays',
+	'vbo-noninterleaved',
 	'vbo-drawelements',
 	'vbo-drawrange',
 	'vp-array',
@@ -141,6 +143,7 @@ progs = [
 	'vp-clip',
 	'vp-line-clip',
 	'vp-tri',
+	'vp-tri-invariant',
 	'vp-tri-swap',
 	'vp-tri-tex',
 	'vp-tri-imm',
diff --git a/progs/trivial/line-flat.c b/progs/trivial/line-flat.c
new file mode 100644
index 00000000000..14f0ac07825
--- /dev/null
+++ b/progs/trivial/line-flat.c
@@ -0,0 +1,147 @@
+/*
+ * Copyright (c) 1991, 1992, 1993 Silicon Graphics, Inc.
+ *
+ * Permission to use, copy, modify, distribute, and sell this software and
+ * its documentation for any purpose is hereby granted without fee, provided
+ * that (i) the above copyright notices and this permission notice appear in
+ * all copies of the software and related documentation, and (ii) the name of
+ * Silicon Graphics may not be used in any advertising or
+ * publicity relating to the software without the specific, prior written
+ * permission of Silicon Graphics.
+ *
+ * THE SOFTWARE IS PROVIDED "AS-IS" AND WITHOUT WARRANTY OF
+ * ANY KIND,
+ * EXPRESS, IMPLIED OR OTHERWISE, INCLUDING WITHOUT LIMITATION, ANY
+ * WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * IN NO EVENT SHALL SILICON GRAPHICS BE LIABLE FOR
+ * ANY SPECIAL, INCIDENTAL, INDIRECT OR CONSEQUENTIAL DAMAGES OF ANY KIND,
+ * OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
+ * WHETHER OR NOT ADVISED OF THE POSSIBILITY OF DAMAGE, AND ON ANY THEORY OF
+ * LIABILITY, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
+ * OF THIS SOFTWARE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+#include <stdlib.h>
+#include <GL/glut.h>
+
+
+#define CI_OFFSET_1 16
+#define CI_OFFSET_2 32
+
+
+GLenum doubleBuffer;
+
+static void Init(void)
+{
+   fprintf(stderr, "GL_RENDERER   = %s\n", (char *) glGetString(GL_RENDERER));
+   fprintf(stderr, "GL_VERSION    = %s\n", (char *) glGetString(GL_VERSION));
+   fprintf(stderr, "GL_VENDOR     = %s\n", (char *) glGetString(GL_VENDOR));
+   fflush(stderr);
+
+    glClearColor(0.0, 0.0, 1.0, 0.0);
+}
+
+static void Reshape(int width, int height)
+{
+
+    glViewport(0, 0, (GLint)width, (GLint)height);
+
+    glMatrixMode(GL_PROJECTION);
+    glLoadIdentity();
+    glOrtho(-1.0, 1.0, -1.0, 1.0, -0.5, 1000.0);
+    glMatrixMode(GL_MODELVIEW);
+}
+
+static void Key(unsigned char key, int x, int y)
+{
+
+    switch (key) {
+      case 27:
+	exit(1);
+      default:
+	return;
+    }
+
+    glutPostRedisplay();
+}
+
+static void Draw(void)
+{
+   glClear(GL_COLOR_BUFFER_BIT); 
+   glShadeModel(GL_FLAT);
+
+   glBegin(GL_LINES);
+   glColor3f(0,0,.7); 
+   glVertex3f( 0.9, -0.9, -30.0);
+   glColor3f(.8,0,0); 
+   glVertex3f( 0.9,  0.9, -30.0);
+
+   glColor3f(.8,0,0); 
+   glVertex3f( 0.9,  0.9, -30.0);
+   glColor3f(0,.9,0); 
+   glVertex3f(-0.9,  0.0, -30.0);
+
+
+   glColor3f(0,.9,0); 
+   glVertex3f(-0.9,  0.0, -30.0);
+   glColor3f(0,0,.7); 
+   glVertex3f( 0.9, -0.9, -30.0);
+   glEnd();
+
+   glFlush();
+
+   if (doubleBuffer) {
+      glutSwapBuffers();
+   }
+}
+
+static GLenum Args(int argc, char **argv)
+{
+    GLint i;
+
+    doubleBuffer = GL_FALSE;
+
+    for (i = 1; i < argc; i++) {
+        if (strcmp(argv[i], "-sb") == 0) {
+	    doubleBuffer = GL_FALSE;
+	} else if (strcmp(argv[i], "-db") == 0) {
+	    doubleBuffer = GL_TRUE;
+	} else {
+	    fprintf(stderr, "%s (Bad option).\n", argv[i]);
+	    return GL_FALSE;
+	}
+    }
+    return GL_TRUE;
+}
+
+int main(int argc, char **argv)
+{
+    GLenum type;
+
+    glutInit(&argc, argv);
+
+    if (Args(argc, argv) == GL_FALSE) {
+	exit(1);
+    }
+
+    glutInitWindowPosition(0, 0); glutInitWindowSize( 250, 250);
+
+    type = GLUT_RGB;
+    type |= (doubleBuffer) ? GLUT_DOUBLE : GLUT_SINGLE;
+    glutInitDisplayMode(type);
+
+    if (glutCreateWindow(*argv) == GL_FALSE) {
+	exit(1);
+    }
+
+    Init();
+
+    glutReshapeFunc(Reshape);
+    glutKeyboardFunc(Key);
+    glutDisplayFunc(Draw);
+    glutMainLoop();
+	return 0;
+}
diff --git a/progs/trivial/vbo-noninterleaved.c b/progs/trivial/vbo-noninterleaved.c
new file mode 100644
index 00000000000..0672ca50ff1
--- /dev/null
+++ b/progs/trivial/vbo-noninterleaved.c
@@ -0,0 +1,139 @@
+/* Basic VBO */
+
+#include <assert.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <GL/glew.h>
+#include <GL/glut.h>
+
+
+struct {
+   GLfloat pos[4][4];
+   GLfloat col[4][4];
+} verts =  
+{
+   /* Position: a quad
+    */
+   { 
+      {  0.9, -0.9, 0.0, 1.0 },
+      {  0.9,  0.9, 0.0, 1.0 },
+      { -0.9,  0.9, 0.0, 1.0 },
+      { -0.9, -0.9, 0.0, 1.0 },
+   },
+
+   /* Color: all red
+    */
+   { 
+      { 1.0, 0.0, 0.0, 1.0 },
+      { 1.0, 0.0, 0.0, 1.0 },
+      { 1.0, 0.0, 0.0, 1.0 },
+      { 1.0, 0.0, 0.0, 1.0 },
+   },
+
+
+};
+
+GLuint arrayObj, elementObj;
+
+static void Init( void )
+{
+   GLint errno;
+   GLuint prognum;
+   
+   static const char *prog1 =
+      "!!ARBvp1.0\n"
+      "MOV  result.color, vertex.color;\n"
+      "MOV  result.position, vertex.position;\n"
+      "END\n";
+
+   glGenProgramsARB(1, &prognum);
+   glBindProgramARB(GL_VERTEX_PROGRAM_ARB, prognum);
+   glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
+		      strlen(prog1), (const GLubyte *) prog1);
+
+   assert(glIsProgramARB(prognum));
+   errno = glGetError();
+   printf("glGetError = %d\n", errno);
+   if (errno != GL_NO_ERROR)
+   {
+      GLint errorpos;
+
+      glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errorpos);
+      printf("errorpos: %d\n", errorpos);
+      printf("%s\n", (char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB));
+   }
+
+
+   glEnableClientState( GL_VERTEX_ARRAY );
+   glEnableClientState( GL_COLOR_ARRAY );
+
+   glGenBuffersARB(1, &arrayObj);
+   glBindBufferARB(GL_ARRAY_BUFFER_ARB, arrayObj);
+   glBufferDataARB(GL_ARRAY_BUFFER_ARB, sizeof(verts), &verts, GL_STATIC_DRAW_ARB);
+
+   glVertexPointer( 4, GL_FLOAT, sizeof(verts.pos[0]), 0 );
+   glColorPointer( 4, GL_FLOAT, sizeof(verts.col[0]), (void *)(4*4*sizeof(float)) );
+
+}
+
+
+
+static void Display( void )
+{
+   glClearColor(0.3, 0.3, 0.3, 1);
+   glClear( GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT );
+
+   glEnable(GL_VERTEX_PROGRAM_ARB);
+
+//   glDrawArrays( GL_TRIANGLES, 0, 3 );
+//   glDrawArrays( GL_TRIANGLES, 1, 3 );
+   glDrawArrays( GL_QUADS, 0, 4 );
+
+   glFlush(); 
+}
+
+
+static void Reshape( int width, int height )
+{
+   glViewport( 0, 0, width, height );
+   glMatrixMode( GL_PROJECTION );
+   glLoadIdentity();
+   glOrtho(-1.0, 1.0, -1.0, 1.0, -0.5, 1000.0);
+   glMatrixMode( GL_MODELVIEW );
+   glLoadIdentity();
+   /*glTranslatef( 0.0, 0.0, -15.0 );*/
+}
+
+
+static void Key( unsigned char key, int x, int y )
+{
+   (void) x;
+   (void) y;
+   switch (key) {
+      case 27:
+         exit(0);
+         break;
+   }
+   glutPostRedisplay();
+}
+
+
+
+
+int main( int argc, char *argv[] )
+{
+   glutInit( &argc, argv );
+   glutInitWindowPosition( 0, 0 );
+   glutInitWindowSize( 250, 250 );
+   glutInitDisplayMode( GLUT_RGB | GLUT_SINGLE | GLUT_DEPTH );
+   glutCreateWindow(argv[0]);
+   glewInit();
+   glutReshapeFunc( Reshape );
+   glutKeyboardFunc( Key );
+   glutDisplayFunc( Display );
+   Init();
+   glutMainLoop();
+   return 0;
+}
diff --git a/progs/trivial/vp-tri-invariant.c b/progs/trivial/vp-tri-invariant.c
new file mode 100644
index 00000000000..ff241393659
--- /dev/null
+++ b/progs/trivial/vp-tri-invariant.c
@@ -0,0 +1,147 @@
+/* Test glGenProgramsNV(), glIsProgramNV(), glLoadProgramNV() */
+
+#include <stdio.h>
+#include <string.h>
+#include <stdlib.h>
+#include <math.h>
+#include <GL/glew.h>
+#include <GL/glut.h>
+
+
+#define CI_OFFSET_1 16
+#define CI_OFFSET_2 32
+
+
+GLenum doubleBuffer;
+
+static void Init(void)
+{
+   GLint errno;
+   GLuint prognum;
+   
+   static const char *prog1 =
+      "!!ARBvp1.0\n"
+      "OPTION ARB_position_invariant ;"
+      "MOV  result.color, vertex.color;\n"
+      "END\n";
+
+
+   glGenProgramsARB(1, &prognum);
+
+   glBindProgramARB(GL_VERTEX_PROGRAM_ARB, prognum);
+   glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB,
+		      strlen(prog1), (const GLubyte *) prog1);
+
+   errno = glGetError();
+   printf("glGetError = %d\n", errno);
+   if (errno != GL_NO_ERROR)
+   {
+      GLint errorpos;
+
+      glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errorpos);
+      printf("errorpos: %d\n", errorpos);
+      printf("%s\n", (char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB));
+   }
+
+   fprintf(stderr, "GL_RENDERER   = %s\n", (char *) glGetString(GL_RENDERER));
+   fprintf(stderr, "GL_VERSION    = %s\n", (char *) glGetString(GL_VERSION));
+   fprintf(stderr, "GL_VENDOR     = %s\n", (char *) glGetString(GL_VENDOR));
+   fflush(stderr);
+
+    glClearColor(0.0, 0.0, 1.0, 0.0);
+}
+
+static void Reshape(int width, int height)
+{
+
+    glViewport(0, 0, (GLint)width, (GLint)height);
+
+    glMatrixMode(GL_PROJECTION);
+    glLoadIdentity();
+/*     glOrtho(-1.0, 1.0, -1.0, 1.0, -0.5, 1000.0); */
+    glMatrixMode(GL_MODELVIEW);
+}
+
+static void Key(unsigned char key, int x, int y)
+{
+
+    switch (key) {
+      case 27:
+	exit(1);
+      default:
+	return;
+    }
+
+    glutPostRedisplay();
+}
+
+static void Draw(void)
+{
+   glClear(GL_COLOR_BUFFER_BIT); 
+
+   glEnable(GL_VERTEX_PROGRAM_ARB);
+
+   glBegin(GL_TRIANGLES);
+   glColor3f(0,0,.7); 
+   glVertex3f( 0.9, -0.9, -0.0);
+   glColor3f(.8,0,0); 
+   glVertex3f( 0.9,  0.9, -0.0);
+   glColor3f(0,.9,0); 
+   glVertex3f(-0.9,  0.0, -0.0);
+   glEnd();
+
+   glFlush();
+
+   if (doubleBuffer) {
+      glutSwapBuffers();
+   }
+}
+
+static GLenum Args(int argc, char **argv)
+{
+    GLint i;
+
+    doubleBuffer = GL_FALSE;
+
+    for (i = 1; i < argc; i++) {
+        if (strcmp(argv[i], "-sb") == 0) {
+	    doubleBuffer = GL_FALSE;
+	} else if (strcmp(argv[i], "-db") == 0) {
+	    doubleBuffer = GL_TRUE;
+	} else {
+	    fprintf(stderr, "%s (Bad option).\n", argv[i]);
+	    return GL_FALSE;
+	}
+    }
+    return GL_TRUE;
+}
+
+int main(int argc, char **argv)
+{
+    GLenum type;
+
+    glutInit(&argc, argv);
+
+    if (Args(argc, argv) == GL_FALSE) {
+	exit(1);
+    }
+
+    glutInitWindowPosition(0, 0); glutInitWindowSize( 250, 250);
+
+    type = GLUT_RGB | GLUT_ALPHA;
+    type |= (doubleBuffer) ? GLUT_DOUBLE : GLUT_SINGLE;
+    glutInitDisplayMode(type);
+
+    if (glutCreateWindow(*argv) == GL_FALSE) {
+	exit(1);
+    }
+
+    glewInit();
+    Init();
+
+    glutReshapeFunc(Reshape);
+    glutKeyboardFunc(Key);
+    glutDisplayFunc(Draw);
+    glutMainLoop();
+	return 0;
+}
diff --git a/progs/xdemos/glxcontexts.c b/progs/xdemos/glxcontexts.c
index a9ff326ed5e..a97b62a9081 100644
--- a/progs/xdemos/glxcontexts.c
+++ b/progs/xdemos/glxcontexts.c
@@ -385,6 +385,10 @@ draw( Display *dpy, Window win )
    } else
       do_draw();
 
+   glDeleteLists(gear1, 1);
+   glDeleteLists(gear2, 1);
+   glDeleteLists(gear3, 1);
+
    glXSwapBuffers(dpy, win);
    glXDestroyContext(dpy, ctx);
 }
diff --git a/scons/gallium.py b/scons/gallium.py
index 696ddd025ff..c7e74d7e599 100644
--- a/scons/gallium.py
+++ b/scons/gallium.py
@@ -317,6 +317,8 @@ def generate(env):
     if gcc:
         if debug:
             cflags += ['-O0', '-g3']
+        elif env['toolchain'] == 'crossmingw':
+            cflags += ['-O0', '-g3'] # mingw 4.2.1 optimizer is broken
         else:
             cflags += ['-O3', '-g3']
         if env['profile']:
diff --git a/scons/generic.py b/scons/generic.py
index 03563e4c620..29ddf76d6ff 100644
--- a/scons/generic.py
+++ b/scons/generic.py
@@ -398,6 +398,8 @@ def generate(env):
     if gcc:
         if debug:
             ccflags += ['-O0', '-g3']
+        elif env['toolchain'] == 'crossmingw':
+            ccflags += ['-O0', '-g3'] # mingw 4.2.1 optimizer is broken
         else:
             ccflags += ['-O3', '-g0']
         if env['profile']:
diff --git a/src/egl/drivers/demo/Makefile b/src/egl/drivers/demo/Makefile
index d908cdb4b25..26694c92fa6 100644
--- a/src/egl/drivers/demo/Makefile
+++ b/src/egl/drivers/demo/Makefile
@@ -26,8 +26,6 @@ $(TOP)/$(LIB_DIR)/demodriver.so: $(OBJECTS)
 		$(OBJECTS)
 
 install:
-	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
-	$(INSTALL) $(TOP)/$(LIB_DIR)/demodriver.so $(DESTDIR)$(INSTALL_LIB_DIR)
 
 clean:
 	-rm -f *.o
diff --git a/src/gallium/SConscript b/src/gallium/SConscript
index 0c632ac2b8b..b6ceaf3edf9 100644
--- a/src/gallium/SConscript
+++ b/src/gallium/SConscript
@@ -29,6 +29,7 @@ for driver in env['drivers']:
 
 SConscript('state_trackers/python/SConscript')
 SConscript('state_trackers/glx/xlib/SConscript')
+SConscript('state_trackers/dri/SConscript')
 
 if platform == 'windows':
 	SConscript('state_trackers/wgl/SConscript')
diff --git a/src/gallium/auxiliary/draw/draw_pipe.h b/src/gallium/auxiliary/draw/draw_pipe.h
index dbad8f98ac7..479250729ff 100644
--- a/src/gallium/auxiliary/draw/draw_pipe.h
+++ b/src/gallium/auxiliary/draw/draw_pipe.h
@@ -57,6 +57,7 @@ struct draw_stage
    struct draw_context *draw;   /**< parent context */
 
    struct draw_stage *next;     /**< next stage in pipeline */
+   const char *name;            /**< for debugging  */
 
    struct vertex_header **tmp;  /**< temp vert storage, such as for clipping */
    unsigned nr_tmps;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
index ca69f0b95e5..9fedeef2d34 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
@@ -746,6 +746,7 @@ draw_aaline_stage(struct draw_context *draw)
       goto fail;
 
    aaline->stage.draw = draw;
+   aaline->stage.name = "aaline";
    aaline->stage.next = NULL;
    aaline->stage.point = draw_pipe_passthrough_point;
    aaline->stage.line = aaline_first_line;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
index 3133abe5dc7..66839f7873c 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
@@ -757,6 +757,7 @@ draw_aapoint_stage(struct draw_context *draw)
       goto fail;
 
    aapoint->stage.draw = draw;
+   aapoint->stage.name = "aapoint";
    aapoint->stage.next = NULL;
    aapoint->stage.point = aapoint_first_point;
    aapoint->stage.line = draw_pipe_passthrough_line;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c
index 3265dcd154a..0670268a196 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_clip.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c
@@ -496,6 +496,7 @@ struct draw_stage *draw_clip_stage( struct draw_context *draw )
       goto fail;
 
    clipper->stage.draw = draw;
+   clipper->stage.name = "clipper";
    clipper->stage.point = clip_point;
    clipper->stage.line = clip_first_line;
    clipper->stage.tri = clip_first_tri;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_cull.c b/src/gallium/auxiliary/draw/draw_pipe_cull.c
index 053be5f050d..0a70483858c 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_cull.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_cull.c
@@ -129,6 +129,7 @@ struct draw_stage *draw_cull_stage( struct draw_context *draw )
       goto fail;
 
    cull->stage.draw = draw;
+   cull->stage.name = "cull";
    cull->stage.next = NULL;
    cull->stage.point = draw_pipe_passthrough_point;
    cull->stage.line = draw_pipe_passthrough_line;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c
index 43d1fecc4dd..34afb1a0b60 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c
@@ -261,6 +261,7 @@ struct draw_stage *draw_flatshade_stage( struct draw_context *draw )
       goto fail;
 
    flatshade->stage.draw = draw;
+   flatshade->stage.name = "flatshade";
    flatshade->stage.next = NULL;
    flatshade->stage.point = draw_pipe_passthrough_point;
    flatshade->stage.line = flatshade_first_line;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_offset.c b/src/gallium/auxiliary/draw/draw_pipe_offset.c
index 62c31532d01..40798a5d6e7 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_offset.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_offset.c
@@ -166,6 +166,7 @@ struct draw_stage *draw_offset_stage( struct draw_context *draw )
    draw_alloc_temp_verts( &offset->stage, 3 );
 
    offset->stage.draw = draw;
+   offset->stage.name = "offset";
    offset->stage.next = NULL;
    offset->stage.point = draw_pipe_passthrough_point;
    offset->stage.line = draw_pipe_passthrough_line;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
index 04e59152c5c..9287fc130ee 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
@@ -586,6 +586,7 @@ draw_pstip_stage(struct draw_context *draw)
    draw_alloc_temp_verts( &pstip->stage, 8 );
 
    pstip->stage.draw = draw;
+   pstip->stage.name = "pstip";
    pstip->stage.next = NULL;
    pstip->stage.point = draw_pipe_passthrough_point;
    pstip->stage.line = draw_pipe_passthrough_line;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_stipple.c b/src/gallium/auxiliary/draw/draw_pipe_stipple.c
index b65e2aa1021..6e921bac278 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_stipple.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_stipple.c
@@ -238,6 +238,7 @@ struct draw_stage *draw_stipple_stage( struct draw_context *draw )
    draw_alloc_temp_verts( &stipple->stage, 2 );
 
    stipple->stage.draw = draw;
+   stipple->stage.name = "stipple";
    stipple->stage.next = NULL;
    stipple->stage.point = stipple_reset_point;
    stipple->stage.line = stipple_first_line;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_twoside.c b/src/gallium/auxiliary/draw/draw_pipe_twoside.c
index c329d923390..eef0238b157 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_twoside.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_twoside.c
@@ -181,6 +181,7 @@ struct draw_stage *draw_twoside_stage( struct draw_context *draw )
       goto fail;
 
    twoside->stage.draw = draw;
+   twoside->stage.name = "twoside";
    twoside->stage.next = NULL;
    twoside->stage.point = draw_pipe_passthrough_point;
    twoside->stage.line = draw_pipe_passthrough_line;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c
index 68835fd1a59..03bb842e20a 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c
@@ -184,6 +184,7 @@ struct draw_stage *draw_unfilled_stage( struct draw_context *draw )
       goto fail;
 
    unfilled->stage.draw = draw;
+   unfilled->stage.name = "unfilled";
    unfilled->stage.next = NULL;
    unfilled->stage.tmp = NULL;
    unfilled->stage.point = draw_pipe_passthrough_point;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_validate.c b/src/gallium/auxiliary/draw/draw_pipe_validate.c
index 03e842ce082..bea90e50d30 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_validate.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_validate.c
@@ -262,7 +262,15 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage )
 
    
    draw->pipeline.first = next;
-   return next;
+
+   if (0) {
+      debug_printf("draw pipeline:\n");
+      for (next = draw->pipeline.first; next ; next = next->next ) 
+         debug_printf("   %s\n", next->name);
+      debug_printf("\n");
+   }
+   
+   return draw->pipeline.first;
 }
 
 static void validate_tri( struct draw_stage *stage, 
@@ -318,6 +326,7 @@ struct draw_stage *draw_validate_stage( struct draw_context *draw )
       return NULL;
 
    stage->draw = draw;
+   stage->name = "validate";
    stage->next = NULL;
    stage->point = validate_point;
    stage->line = validate_line;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
index 12325d30d61..a5d840b96ea 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
@@ -446,6 +446,7 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw,
       goto fail;
    
    vbuf->stage.draw = draw;
+   vbuf->stage.name = "vbuf";
    vbuf->stage.point = vbuf_first_point;
    vbuf->stage.line = vbuf_first_line;
    vbuf->stage.tri = vbuf_first_tri;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c
index 184e363594d..f32cbef983d 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c
@@ -168,6 +168,7 @@ struct draw_stage *draw_wide_line_stage( struct draw_context *draw )
    draw_alloc_temp_verts( &wide->stage, 4 );
 
    wide->stage.draw = draw;
+   wide->stage.name = "wide-line";
    wide->stage.next = NULL;
    wide->stage.point = draw_pipe_passthrough_point;
    wide->stage.line = wideline_line;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
index e1af9e56a24..49034ae86a2 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
@@ -279,6 +279,7 @@ struct draw_stage *draw_wide_point_stage( struct draw_context *draw )
       goto fail;
 
    wide->stage.draw = draw;
+   wide->stage.name = "wide-point";
    wide->stage.next = NULL;
    wide->stage.point = widepoint_first_point;
    wide->stage.line = draw_pipe_passthrough_line;
diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h
index 6f3e1e0289b..8ef0ea8011f 100644
--- a/src/gallium/auxiliary/draw/draw_pt.h
+++ b/src/gallium/auxiliary/draw/draw_pt.h
@@ -187,6 +187,7 @@ struct pt_emit *draw_pt_emit_create( struct draw_context *draw );
 
 struct pt_fetch;
 void draw_pt_fetch_prepare( struct pt_fetch *fetch,
+                            unsigned vertex_input_count,
 			    unsigned vertex_size );
 
 void draw_pt_fetch_run( struct pt_fetch *fetch,
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c
index 505d32f2c32..65c3a34c347 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c
@@ -26,6 +26,7 @@
  **************************************************************************/
 
 #include "util/u_memory.h"
+#include "util/u_math.h"
 #include "draw/draw_context.h"
 #include "draw/draw_private.h"
 #include "draw/draw_vbuf.h"
@@ -56,9 +57,11 @@ struct pt_fetch {
  *
  */
 void draw_pt_fetch_prepare( struct pt_fetch *fetch,
+                            unsigned vs_input_count,
 			    unsigned vertex_size )
 {
    struct draw_context *draw = fetch->draw;
+   unsigned nr_inputs;
    unsigned i, nr = 0;
    unsigned dst_offset = 0;
    struct translate_key key;
@@ -89,8 +92,11 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch,
       dst_offset += 4 * sizeof(float);
    }
       
+   assert( draw->pt.nr_vertex_elements >= vs_input_count );
 
-   for (i = 0; i < draw->pt.nr_vertex_elements; i++) {
+   nr_inputs = MIN2( vs_input_count, draw->pt.nr_vertex_elements );
+
+   for (i = 0; i < nr_inputs; i++) {
       key.element[nr].input_format = draw->pt.vertex_element[i].src_format;
       key.element[nr].input_buffer = draw->pt.vertex_element[i].vertex_buffer_index;
       key.element[nr].input_offset = draw->pt.vertex_element[i].src_offset;
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
index 11ac90fc563..df6c265b7ec 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
@@ -77,8 +77,8 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle,
    
 
    draw_pt_fetch_prepare( fpme->fetch, 
+                          vs->info.num_inputs,
 			  fpme->vertex_size );
-
    /* XXX: it's not really gl rasterization rules we care about here,
     * but gl vs dx9 clip spaces.
     */
diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c
index dbbc33fffa7..f2368dde5c2 100644
--- a/src/gallium/auxiliary/draw/draw_vs_exec.c
+++ b/src/gallium/auxiliary/draw/draw_vs_exec.c
@@ -114,6 +114,12 @@ vs_exec_run_linear( struct draw_vertex_shader *shader,
 #endif
 
          for (slot = 0; slot < shader->info.num_inputs; slot++) {
+#if 0
+            assert(!util_is_inf_or_nan(input[slot][0]));
+            assert(!util_is_inf_or_nan(input[slot][1]));
+            assert(!util_is_inf_or_nan(input[slot][2]));
+            assert(!util_is_inf_or_nan(input[slot][3]));
+#endif
             machine->Inputs[slot].xyzw[0].f[j] = input[slot][0];
             machine->Inputs[slot].xyzw[1].f[j] = input[slot][1];
             machine->Inputs[slot].xyzw[2].f[j] = input[slot][2];
diff --git a/src/gallium/auxiliary/indices/u_unfilled_gen.c b/src/gallium/auxiliary/indices/u_unfilled_gen.c
index fd082ebbb37..93897c98deb 100644
--- a/src/gallium/auxiliary/indices/u_unfilled_gen.c
+++ b/src/gallium/auxiliary/indices/u_unfilled_gen.c
@@ -71,13 +71,10 @@ static void generate_tris_ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i+=3) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1);
       (out+j)[0] = (ushort)(i);
       (out+j)[1] = (ushort)(i+1);
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)(i+1);
       (out+j+2)[1] = (ushort)(i+2);
-      debug_printf("  line %d %d\n", (int)i+2, (int)i);
       (out+j+4)[0] = (ushort)(i+2);
       (out+j+4)[1] = (ushort)(i);
    }
@@ -90,13 +87,10 @@ static void generate_tristrip_ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1/*+(i&1)*/);
       (out+j)[0] = (ushort)(i);
       (out+j)[1] = (ushort)(i+1/*+(i&1)*/);
-      debug_printf("  line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/);
       (out+j+2)[0] = (ushort)(i+1/*+(i&1)*/);
       (out+j+2)[1] = (ushort)(i+2/*-(i&1)*/);
-      debug_printf("  line %d %d\n", (int)i+2/*-(i&1)*/, (int)i);
       (out+j+4)[0] = (ushort)(i+2/*-(i&1)*/);
       (out+j+4)[1] = (ushort)(i);
    }
@@ -109,13 +103,10 @@ static void generate_trifan_ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (ushort)(0);
       (out+j)[1] = (ushort)(i+1);
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)(i+1);
       (out+j+2)[1] = (ushort)(i+2);
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (ushort)(i+2);
       (out+j+4)[1] = (ushort)(0);
    }
@@ -128,16 +119,12 @@ static void generate_quads_ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=4) { 
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j)[0] = (ushort)(i+0);
       (out+j)[1] = (ushort)(i+1);
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)(i+1);
       (out+j+2)[1] = (ushort)(i+2);
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+3);
       (out+j+4)[0] = (ushort)(i+2);
       (out+j+4)[1] = (ushort)(i+3);
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+0);
       (out+j+6)[0] = (ushort)(i+3);
       (out+j+6)[1] = (ushort)(i+0);
    }
@@ -150,16 +137,12 @@ static void generate_quadstrip_ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=2) { 
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+0);
       (out+j)[0] = (ushort)(i+2);
       (out+j)[1] = (ushort)(i+0);
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j+2)[0] = (ushort)(i+0);
       (out+j+2)[1] = (ushort)(i+1);
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+3);
       (out+j+4)[0] = (ushort)(i+1);
       (out+j+4)[1] = (ushort)(i+3);
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+2);
       (out+j+6)[0] = (ushort)(i+3);
       (out+j+6)[1] = (ushort)(i+2);
    }
@@ -172,13 +155,10 @@ static void generate_polygon_ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (ushort)(0);
       (out+j)[1] = (ushort)(i+1);
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)(i+1);
       (out+j+2)[1] = (ushort)(i+2);
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (ushort)(i+2);
       (out+j+4)[1] = (ushort)(0);
    }
@@ -191,13 +171,10 @@ static void generate_tris_uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i+=3) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1);
       (out+j)[0] = (uint)(i);
       (out+j)[1] = (uint)(i+1);
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)(i+1);
       (out+j+2)[1] = (uint)(i+2);
-      debug_printf("  line %d %d\n", (int)i+2, (int)i);
       (out+j+4)[0] = (uint)(i+2);
       (out+j+4)[1] = (uint)(i);
    }
@@ -210,13 +187,10 @@ static void generate_tristrip_uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1/*+(i&1)*/);
       (out+j)[0] = (uint)(i);
       (out+j)[1] = (uint)(i+1/*+(i&1)*/);
-      debug_printf("  line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/);
       (out+j+2)[0] = (uint)(i+1/*+(i&1)*/);
       (out+j+2)[1] = (uint)(i+2/*-(i&1)*/);
-      debug_printf("  line %d %d\n", (int)i+2/*-(i&1)*/, (int)i);
       (out+j+4)[0] = (uint)(i+2/*-(i&1)*/);
       (out+j+4)[1] = (uint)(i);
    }
@@ -229,13 +203,10 @@ static void generate_trifan_uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (uint)(0);
       (out+j)[1] = (uint)(i+1);
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)(i+1);
       (out+j+2)[1] = (uint)(i+2);
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (uint)(i+2);
       (out+j+4)[1] = (uint)(0);
    }
@@ -248,16 +219,12 @@ static void generate_quads_uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=4) { 
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j)[0] = (uint)(i+0);
       (out+j)[1] = (uint)(i+1);
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)(i+1);
       (out+j+2)[1] = (uint)(i+2);
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+3);
       (out+j+4)[0] = (uint)(i+2);
       (out+j+4)[1] = (uint)(i+3);
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+0);
       (out+j+6)[0] = (uint)(i+3);
       (out+j+6)[1] = (uint)(i+0);
    }
@@ -270,16 +237,12 @@ static void generate_quadstrip_uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=2) { 
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+0);
       (out+j)[0] = (uint)(i+2);
       (out+j)[1] = (uint)(i+0);
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j+2)[0] = (uint)(i+0);
       (out+j+2)[1] = (uint)(i+1);
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+3);
       (out+j+4)[0] = (uint)(i+1);
       (out+j+4)[1] = (uint)(i+3);
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+2);
       (out+j+6)[0] = (uint)(i+3);
       (out+j+6)[1] = (uint)(i+2);
    }
@@ -292,13 +255,10 @@ static void generate_polygon_uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (uint)(0);
       (out+j)[1] = (uint)(i+1);
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)(i+1);
       (out+j+2)[1] = (uint)(i+2);
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (uint)(i+2);
       (out+j+4)[1] = (uint)(0);
    }
@@ -313,13 +273,10 @@ static void translate_tris_ubyte2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i+=3) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1);
       (out+j)[0] = (ushort)in[i];
       (out+j)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)in[i+1];
       (out+j+2)[1] = (ushort)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)i);
       (out+j+4)[0] = (ushort)in[i+2];
       (out+j+4)[1] = (ushort)in[i];
    }
@@ -334,13 +291,10 @@ static void translate_tristrip_ubyte2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1/*+(i&1)*/);
       (out+j)[0] = (ushort)in[i];
       (out+j)[1] = (ushort)in[i+1/*+(i&1)*/];
-      debug_printf("  line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/);
       (out+j+2)[0] = (ushort)in[i+1/*+(i&1)*/];
       (out+j+2)[1] = (ushort)in[i+2/*-(i&1)*/];
-      debug_printf("  line %d %d\n", (int)i+2/*-(i&1)*/, (int)i);
       (out+j+4)[0] = (ushort)in[i+2/*-(i&1)*/];
       (out+j+4)[1] = (ushort)in[i];
    }
@@ -355,13 +309,10 @@ static void translate_trifan_ubyte2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (ushort)in[0];
       (out+j)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)in[i+1];
       (out+j+2)[1] = (ushort)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (ushort)in[i+2];
       (out+j+4)[1] = (ushort)in[0];
    }
@@ -376,16 +327,12 @@ static void translate_quads_ubyte2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=4) { 
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j)[0] = (ushort)in[i+0];
       (out+j)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)in[i+1];
       (out+j+2)[1] = (ushort)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+3);
       (out+j+4)[0] = (ushort)in[i+2];
       (out+j+4)[1] = (ushort)in[i+3];
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+0);
       (out+j+6)[0] = (ushort)in[i+3];
       (out+j+6)[1] = (ushort)in[i+0];
    }
@@ -400,16 +347,12 @@ static void translate_quadstrip_ubyte2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=2) { 
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+0);
       (out+j)[0] = (ushort)in[i+2];
       (out+j)[1] = (ushort)in[i+0];
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j+2)[0] = (ushort)in[i+0];
       (out+j+2)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+3);
       (out+j+4)[0] = (ushort)in[i+1];
       (out+j+4)[1] = (ushort)in[i+3];
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+2);
       (out+j+6)[0] = (ushort)in[i+3];
       (out+j+6)[1] = (ushort)in[i+2];
    }
@@ -424,13 +367,10 @@ static void translate_polygon_ubyte2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (ushort)in[0];
       (out+j)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)in[i+1];
       (out+j+2)[1] = (ushort)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (ushort)in[i+2];
       (out+j+4)[1] = (ushort)in[0];
    }
@@ -445,13 +385,10 @@ static void translate_tris_ubyte2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i+=3) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1);
       (out+j)[0] = (uint)in[i];
       (out+j)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)in[i+1];
       (out+j+2)[1] = (uint)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)i);
       (out+j+4)[0] = (uint)in[i+2];
       (out+j+4)[1] = (uint)in[i];
    }
@@ -466,13 +403,10 @@ static void translate_tristrip_ubyte2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1/*+(i&1)*/);
       (out+j)[0] = (uint)in[i];
       (out+j)[1] = (uint)in[i+1/*+(i&1)*/];
-      debug_printf("  line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/);
       (out+j+2)[0] = (uint)in[i+1/*+(i&1)*/];
       (out+j+2)[1] = (uint)in[i+2/*-(i&1)*/];
-      debug_printf("  line %d %d\n", (int)i+2/*-(i&1)*/, (int)i);
       (out+j+4)[0] = (uint)in[i+2/*-(i&1)*/];
       (out+j+4)[1] = (uint)in[i];
    }
@@ -487,13 +421,10 @@ static void translate_trifan_ubyte2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (uint)in[0];
       (out+j)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)in[i+1];
       (out+j+2)[1] = (uint)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (uint)in[i+2];
       (out+j+4)[1] = (uint)in[0];
    }
@@ -508,16 +439,12 @@ static void translate_quads_ubyte2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=4) { 
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j)[0] = (uint)in[i+0];
       (out+j)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)in[i+1];
       (out+j+2)[1] = (uint)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+3);
       (out+j+4)[0] = (uint)in[i+2];
       (out+j+4)[1] = (uint)in[i+3];
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+0);
       (out+j+6)[0] = (uint)in[i+3];
       (out+j+6)[1] = (uint)in[i+0];
    }
@@ -532,16 +459,12 @@ static void translate_quadstrip_ubyte2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=2) { 
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+0);
       (out+j)[0] = (uint)in[i+2];
       (out+j)[1] = (uint)in[i+0];
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j+2)[0] = (uint)in[i+0];
       (out+j+2)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+3);
       (out+j+4)[0] = (uint)in[i+1];
       (out+j+4)[1] = (uint)in[i+3];
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+2);
       (out+j+6)[0] = (uint)in[i+3];
       (out+j+6)[1] = (uint)in[i+2];
    }
@@ -556,13 +479,10 @@ static void translate_polygon_ubyte2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (uint)in[0];
       (out+j)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)in[i+1];
       (out+j+2)[1] = (uint)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (uint)in[i+2];
       (out+j+4)[1] = (uint)in[0];
    }
@@ -577,13 +497,10 @@ static void translate_tris_ushort2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i+=3) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1);
       (out+j)[0] = (ushort)in[i];
       (out+j)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)in[i+1];
       (out+j+2)[1] = (ushort)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)i);
       (out+j+4)[0] = (ushort)in[i+2];
       (out+j+4)[1] = (ushort)in[i];
    }
@@ -598,13 +515,10 @@ static void translate_tristrip_ushort2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1/*+(i&1)*/);
       (out+j)[0] = (ushort)in[i];
       (out+j)[1] = (ushort)in[i+1/*+(i&1)*/];
-      debug_printf("  line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/);
       (out+j+2)[0] = (ushort)in[i+1/*+(i&1)*/];
       (out+j+2)[1] = (ushort)in[i+2/*-(i&1)*/];
-      debug_printf("  line %d %d\n", (int)i+2/*-(i&1)*/, (int)i);
       (out+j+4)[0] = (ushort)in[i+2/*-(i&1)*/];
       (out+j+4)[1] = (ushort)in[i];
    }
@@ -619,13 +533,10 @@ static void translate_trifan_ushort2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (ushort)in[0];
       (out+j)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)in[i+1];
       (out+j+2)[1] = (ushort)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (ushort)in[i+2];
       (out+j+4)[1] = (ushort)in[0];
    }
@@ -640,16 +551,12 @@ static void translate_quads_ushort2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=4) { 
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j)[0] = (ushort)in[i+0];
       (out+j)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)in[i+1];
       (out+j+2)[1] = (ushort)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+3);
       (out+j+4)[0] = (ushort)in[i+2];
       (out+j+4)[1] = (ushort)in[i+3];
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+0);
       (out+j+6)[0] = (ushort)in[i+3];
       (out+j+6)[1] = (ushort)in[i+0];
    }
@@ -664,16 +571,12 @@ static void translate_quadstrip_ushort2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=2) { 
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+0);
       (out+j)[0] = (ushort)in[i+2];
       (out+j)[1] = (ushort)in[i+0];
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j+2)[0] = (ushort)in[i+0];
       (out+j+2)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+3);
       (out+j+4)[0] = (ushort)in[i+1];
       (out+j+4)[1] = (ushort)in[i+3];
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+2);
       (out+j+6)[0] = (ushort)in[i+3];
       (out+j+6)[1] = (ushort)in[i+2];
    }
@@ -688,13 +591,10 @@ static void translate_polygon_ushort2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (ushort)in[0];
       (out+j)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)in[i+1];
       (out+j+2)[1] = (ushort)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (ushort)in[i+2];
       (out+j+4)[1] = (ushort)in[0];
    }
@@ -709,13 +609,10 @@ static void translate_tris_ushort2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i+=3) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1);
       (out+j)[0] = (uint)in[i];
       (out+j)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)in[i+1];
       (out+j+2)[1] = (uint)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)i);
       (out+j+4)[0] = (uint)in[i+2];
       (out+j+4)[1] = (uint)in[i];
    }
@@ -730,13 +627,10 @@ static void translate_tristrip_ushort2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1/*+(i&1)*/);
       (out+j)[0] = (uint)in[i];
       (out+j)[1] = (uint)in[i+1/*+(i&1)*/];
-      debug_printf("  line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/);
       (out+j+2)[0] = (uint)in[i+1/*+(i&1)*/];
       (out+j+2)[1] = (uint)in[i+2/*-(i&1)*/];
-      debug_printf("  line %d %d\n", (int)i+2/*-(i&1)*/, (int)i);
       (out+j+4)[0] = (uint)in[i+2/*-(i&1)*/];
       (out+j+4)[1] = (uint)in[i];
    }
@@ -751,13 +645,10 @@ static void translate_trifan_ushort2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (uint)in[0];
       (out+j)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)in[i+1];
       (out+j+2)[1] = (uint)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (uint)in[i+2];
       (out+j+4)[1] = (uint)in[0];
    }
@@ -772,16 +663,12 @@ static void translate_quads_ushort2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=4) { 
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j)[0] = (uint)in[i+0];
       (out+j)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)in[i+1];
       (out+j+2)[1] = (uint)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+3);
       (out+j+4)[0] = (uint)in[i+2];
       (out+j+4)[1] = (uint)in[i+3];
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+0);
       (out+j+6)[0] = (uint)in[i+3];
       (out+j+6)[1] = (uint)in[i+0];
    }
@@ -796,16 +683,12 @@ static void translate_quadstrip_ushort2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=2) { 
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+0);
       (out+j)[0] = (uint)in[i+2];
       (out+j)[1] = (uint)in[i+0];
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j+2)[0] = (uint)in[i+0];
       (out+j+2)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+3);
       (out+j+4)[0] = (uint)in[i+1];
       (out+j+4)[1] = (uint)in[i+3];
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+2);
       (out+j+6)[0] = (uint)in[i+3];
       (out+j+6)[1] = (uint)in[i+2];
    }
@@ -820,13 +703,10 @@ static void translate_polygon_ushort2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (uint)in[0];
       (out+j)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)in[i+1];
       (out+j+2)[1] = (uint)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (uint)in[i+2];
       (out+j+4)[1] = (uint)in[0];
    }
@@ -841,13 +721,10 @@ static void translate_tris_uint2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i+=3) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1);
       (out+j)[0] = (ushort)in[i];
       (out+j)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)in[i+1];
       (out+j+2)[1] = (ushort)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)i);
       (out+j+4)[0] = (ushort)in[i+2];
       (out+j+4)[1] = (ushort)in[i];
    }
@@ -862,13 +739,10 @@ static void translate_tristrip_uint2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1/*+(i&1)*/);
       (out+j)[0] = (ushort)in[i];
       (out+j)[1] = (ushort)in[i+1/*+(i&1)*/];
-      debug_printf("  line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/);
       (out+j+2)[0] = (ushort)in[i+1/*+(i&1)*/];
       (out+j+2)[1] = (ushort)in[i+2/*-(i&1)*/];
-      debug_printf("  line %d %d\n", (int)i+2/*-(i&1)*/, (int)i);
       (out+j+4)[0] = (ushort)in[i+2/*-(i&1)*/];
       (out+j+4)[1] = (ushort)in[i];
    }
@@ -883,13 +757,10 @@ static void translate_trifan_uint2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (ushort)in[0];
       (out+j)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)in[i+1];
       (out+j+2)[1] = (ushort)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (ushort)in[i+2];
       (out+j+4)[1] = (ushort)in[0];
    }
@@ -904,16 +775,12 @@ static void translate_quads_uint2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=4) { 
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j)[0] = (ushort)in[i+0];
       (out+j)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)in[i+1];
       (out+j+2)[1] = (ushort)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+3);
       (out+j+4)[0] = (ushort)in[i+2];
       (out+j+4)[1] = (ushort)in[i+3];
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+0);
       (out+j+6)[0] = (ushort)in[i+3];
       (out+j+6)[1] = (ushort)in[i+0];
    }
@@ -928,16 +795,12 @@ static void translate_quadstrip_uint2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=2) { 
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+0);
       (out+j)[0] = (ushort)in[i+2];
       (out+j)[1] = (ushort)in[i+0];
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j+2)[0] = (ushort)in[i+0];
       (out+j+2)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+3);
       (out+j+4)[0] = (ushort)in[i+1];
       (out+j+4)[1] = (ushort)in[i+3];
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+2);
       (out+j+6)[0] = (ushort)in[i+3];
       (out+j+6)[1] = (ushort)in[i+2];
    }
@@ -952,13 +815,10 @@ static void translate_polygon_uint2ushort(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (ushort)in[0];
       (out+j)[1] = (ushort)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (ushort)in[i+1];
       (out+j+2)[1] = (ushort)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (ushort)in[i+2];
       (out+j+4)[1] = (ushort)in[0];
    }
@@ -973,13 +833,10 @@ static void translate_tris_uint2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i+=3) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1);
       (out+j)[0] = (uint)in[i];
       (out+j)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)in[i+1];
       (out+j+2)[1] = (uint)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)i);
       (out+j+4)[0] = (uint)in[i+2];
       (out+j+4)[1] = (uint)in[i];
    }
@@ -994,13 +851,10 @@ static void translate_tristrip_uint2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)i, (int)i+1/*+(i&1)*/);
       (out+j)[0] = (uint)in[i];
       (out+j)[1] = (uint)in[i+1/*+(i&1)*/];
-      debug_printf("  line %d %d\n", (int)i+1/*+(i&1)*/, (int)i+2/*-(i&1)*/);
       (out+j+2)[0] = (uint)in[i+1/*+(i&1)*/];
       (out+j+2)[1] = (uint)in[i+2/*-(i&1)*/];
-      debug_printf("  line %d %d\n", (int)i+2/*-(i&1)*/, (int)i);
       (out+j+4)[0] = (uint)in[i+2/*-(i&1)*/];
       (out+j+4)[1] = (uint)in[i];
    }
@@ -1015,13 +869,10 @@ static void translate_trifan_uint2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (uint)in[0];
       (out+j)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)in[i+1];
       (out+j+2)[1] = (uint)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (uint)in[i+2];
       (out+j+4)[1] = (uint)in[0];
    }
@@ -1036,16 +887,12 @@ static void translate_quads_uint2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=4) { 
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j)[0] = (uint)in[i+0];
       (out+j)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)in[i+1];
       (out+j+2)[1] = (uint)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+3);
       (out+j+4)[0] = (uint)in[i+2];
       (out+j+4)[1] = (uint)in[i+3];
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+0);
       (out+j+6)[0] = (uint)in[i+3];
       (out+j+6)[1] = (uint)in[i+0];
    }
@@ -1060,16 +907,12 @@ static void translate_quadstrip_uint2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=8, i+=2) { 
-      debug_printf("  line %d %d\n", (int)i+2, (int)i+0);
       (out+j)[0] = (uint)in[i+2];
       (out+j)[1] = (uint)in[i+0];
-      debug_printf("  line %d %d\n", (int)i+0, (int)i+1);
       (out+j+2)[0] = (uint)in[i+0];
       (out+j+2)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+3);
       (out+j+4)[0] = (uint)in[i+1];
       (out+j+4)[1] = (uint)in[i+3];
-      debug_printf("  line %d %d\n", (int)i+3, (int)i+2);
       (out+j+6)[0] = (uint)in[i+3];
       (out+j+6)[1] = (uint)in[i+2];
    }
@@ -1084,13 +927,10 @@ static void translate_polygon_uint2uint(
   unsigned i, j;
   (void)j;
   for (j = i = 0; j < nr; j+=6, i++) { 
-      debug_printf("  line %d %d\n", (int)0, (int)i+1);
       (out+j)[0] = (uint)in[0];
       (out+j)[1] = (uint)in[i+1];
-      debug_printf("  line %d %d\n", (int)i+1, (int)i+2);
       (out+j+2)[0] = (uint)in[i+1];
       (out+j+2)[1] = (uint)in[i+2];
-      debug_printf("  line %d %d\n", (int)i+2, (int)0);
       (out+j+4)[0] = (uint)in[i+2];
       (out+j+4)[1] = (uint)in[0];
    }
diff --git a/src/gallium/auxiliary/indices/u_unfilled_gen.py b/src/gallium/auxiliary/indices/u_unfilled_gen.py
index d0344fe3131..36896ce605d 100644
--- a/src/gallium/auxiliary/indices/u_unfilled_gen.py
+++ b/src/gallium/auxiliary/indices/u_unfilled_gen.py
@@ -99,7 +99,6 @@ def vert( intype, outtype, v0 ):
         return '(' + outtype + ')in[' + v0 + ']'
 
 def line( intype, outtype, ptr, v0, v1 ):
-    print '      debug_printf("  line %d %d\\n", (int)' + v0 + ', (int)' + v1 + ');'
     print '      (' + ptr + ')[0] = ' + vert( intype, outtype, v0 ) + ';'
     print '      (' + ptr + ')[1] = ' + vert( intype, outtype, v1 ) + ';'
 
diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
index 2cd0b8a8cdf..4698efa69cf 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
@@ -365,21 +365,22 @@ fenced_buffer_validate(struct pb_buffer *buf,
    if(fenced_buf->vl && fenced_buf->vl != vl)
       return PIPE_ERROR_RETRY;
    
+#if 0
    /* Do not validate if buffer is still mapped */
    if(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) {
       /* TODO: wait for the thread that mapped the buffer to unmap it */
       return PIPE_ERROR_RETRY;
    }
+   /* Final sanity checking */
+   assert(!(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE));
+   assert(!fenced_buf->mapcount);
+#endif
 
    if(fenced_buf->vl == vl &&
       (fenced_buf->validation_flags & flags) == flags) {
       /* Nothing to do -- buffer already validated */
       return PIPE_OK;
    }
-
-   /* Final sanity checking */
-   assert(!(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE));
-   assert(!fenced_buf->mapcount);
    
    ret = pb_validate(fenced_buf->buffer, vl, flags);
    if (ret != PIPE_OK)
@@ -530,16 +531,17 @@ fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list)
 
    pipe_mutex_lock(fenced_list->mutex);
 
-   debug_printf("%10s %7s %10s %s\n",
-                "buffer", "reference.count", "fence", "signalled");
+   debug_printf("%10s %7s %7s %10s %s\n",
+                "buffer", "size", "refcount", "fence", "signalled");
    
    curr = fenced_list->unfenced.next;
    next = curr->next;
    while(curr != &fenced_list->unfenced) {
       fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
       assert(!fenced_buf->fence);
-      debug_printf("%10p %7u\n",
+      debug_printf("%10p %7u %7u\n",
                    fenced_buf,
+                   fenced_buf->base.base.size,
                    fenced_buf->base.base.reference.count);
       curr = next; 
       next = curr->next;
@@ -551,8 +553,9 @@ fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list)
       int signaled;
       fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
       signaled = ops->fence_signalled(ops, fenced_buf->fence, 0);
-      debug_printf("%10p %7u %10p %s\n",
+      debug_printf("%10p %7u %7u %10p %s\n",
                    fenced_buf,
+                   fenced_buf->base.base.size,
                    fenced_buf->base.base.reference.count,
                    fenced_buf->fence,
                    signaled == 0 ? "y" : "n");
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c
index f1a05be46e4..cedf745bdac 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c
@@ -40,6 +40,7 @@
 #include "util/u_memory.h"
 #include "util/u_double_list.h"
 #include "util/u_time.h"
+#include "util/u_debug_stack.h"
 
 #include "pb_buffer.h"
 #include "pb_bufmgr.h"
@@ -48,6 +49,10 @@
 #ifdef DEBUG
 
 
+#define PB_DEBUG_CREATE_BACKTRACE 8
+#define PB_DEBUG_MAP_BACKTRACE 8
+
+
 /**
  * Convenience macro (type safe).
  */
@@ -69,6 +74,14 @@ struct pb_debug_buffer
    
    size_t underflow_size;
    size_t overflow_size;
+
+   struct debug_stack_frame create_backtrace[PB_DEBUG_CREATE_BACKTRACE];
+
+   pipe_mutex mutex;
+   unsigned map_count;
+   struct debug_stack_frame map_backtrace[PB_DEBUG_MAP_BACKTRACE];
+   
+   struct list_head head;
 };
 
 
@@ -80,6 +93,9 @@ struct pb_debug_manager
 
    size_t underflow_size;
    size_t overflow_size;
+   
+   pipe_mutex mutex;
+   struct list_head list;
 };
 
 
@@ -189,6 +205,9 @@ pb_debug_buffer_check(struct pb_debug_buffer *buf)
                       max_ofs == buf->overflow_size - 1 ? "+" : "");
       }
       
+      if(underflow || overflow)
+         debug_backtrace_dump(buf->create_backtrace, PB_DEBUG_CREATE_BACKTRACE);
+
       debug_assert(!underflow && !overflow);
 
       /* re-fill if not aborted */
@@ -207,11 +226,18 @@ static void
 pb_debug_buffer_destroy(struct pb_buffer *_buf)
 {
    struct pb_debug_buffer *buf = pb_debug_buffer(_buf);
+   struct pb_debug_manager *mgr = buf->mgr;
    
    assert(!pipe_is_referenced(&buf->base.base.reference));
    
    pb_debug_buffer_check(buf);
 
+   pipe_mutex_lock(mgr->mutex);
+   LIST_DEL(&buf->head);
+   pipe_mutex_unlock(mgr->mutex);
+
+   pipe_mutex_destroy(buf->mutex);
+   
    pb_reference(&buf->buffer, NULL);
    FREE(buf);
 }
@@ -230,6 +256,13 @@ pb_debug_buffer_map(struct pb_buffer *_buf,
    if(!map)
       return NULL;
    
+   if(map) {
+      pipe_mutex_lock(buf->mutex);
+      ++buf->map_count;
+      debug_backtrace_capture(buf->map_backtrace, 1, PB_DEBUG_MAP_BACKTRACE);
+      pipe_mutex_unlock(buf->mutex);
+   }
+   
    return (uint8_t *)map + buf->underflow_size;
 }
 
@@ -238,6 +271,13 @@ static void
 pb_debug_buffer_unmap(struct pb_buffer *_buf)
 {
    struct pb_debug_buffer *buf = pb_debug_buffer(_buf);   
+   
+   pipe_mutex_lock(buf->mutex);
+   assert(buf->map_count);
+   if(buf->map_count)
+      --buf->map_count;
+   pipe_mutex_unlock(buf->mutex);
+   
    pb_unmap(buf->buffer);
    
    pb_debug_buffer_check(buf);
@@ -262,6 +302,14 @@ pb_debug_buffer_validate(struct pb_buffer *_buf,
 {
    struct pb_debug_buffer *buf = pb_debug_buffer(_buf);
    
+   pipe_mutex_lock(buf->mutex);
+   if(buf->map_count) {
+      debug_printf("%s: attempting to validate a mapped buffer\n", __FUNCTION__);
+      debug_printf("last map backtrace is\n");
+      debug_backtrace_dump(buf->map_backtrace, PB_DEBUG_MAP_BACKTRACE);
+   }
+   pipe_mutex_unlock(buf->mutex);
+
    pb_debug_buffer_check(buf);
 
    return pb_validate(buf->buffer, vl, flags);
@@ -288,6 +336,31 @@ pb_debug_buffer_vtbl = {
 };
 
 
+static void
+pb_debug_manager_dump(struct pb_debug_manager *mgr)
+{
+   struct list_head *curr, *next;
+   struct pb_debug_buffer *buf;
+
+   pipe_mutex_lock(mgr->mutex);
+      
+   curr = mgr->list.next;
+   next = curr->next;
+   while(curr != &mgr->list) {
+      buf = LIST_ENTRY(struct pb_debug_buffer, curr, head);
+
+      debug_printf("buffer = %p\n", buf);
+      debug_printf("    .size = %p\n", buf->base.base.size);
+      debug_backtrace_dump(buf->create_backtrace, PB_DEBUG_CREATE_BACKTRACE);
+      
+      curr = next; 
+      next = curr->next;
+   }
+
+   pipe_mutex_unlock(mgr->mutex);
+}
+
+
 static struct pb_buffer *
 pb_debug_manager_create_buffer(struct pb_manager *_mgr, 
                                size_t size,
@@ -312,6 +385,13 @@ pb_debug_manager_create_buffer(struct pb_manager *_mgr,
                                               &real_desc);
    if(!buf->buffer) {
       FREE(buf);
+#if 0
+      pipe_mutex_lock(mgr->mutex);
+      debug_printf("%s: failed to create buffer\n", __FUNCTION__);
+      if(!LIST_IS_EMPTY(&mgr->list))
+         pb_debug_manager_dump(mgr);
+      pipe_mutex_unlock(mgr->mutex);
+#endif
       return NULL;
    }
    
@@ -331,8 +411,16 @@ pb_debug_manager_create_buffer(struct pb_manager *_mgr,
    buf->underflow_size = mgr->underflow_size;
    buf->overflow_size = buf->buffer->base.size - buf->underflow_size - size;
    
+   debug_backtrace_capture(buf->create_backtrace, 1, PB_DEBUG_CREATE_BACKTRACE);
+
    pb_debug_buffer_fill(buf);
    
+   pipe_mutex_init(buf->mutex);
+   
+   pipe_mutex_lock(mgr->mutex);
+   LIST_ADDTAIL(&buf->head, &mgr->list);
+   pipe_mutex_unlock(mgr->mutex);
+
    return &buf->base;
 }
 
@@ -351,6 +439,15 @@ static void
 pb_debug_manager_destroy(struct pb_manager *_mgr)
 {
    struct pb_debug_manager *mgr = pb_debug_manager(_mgr);
+   
+   pipe_mutex_lock(mgr->mutex);
+   if(!LIST_IS_EMPTY(&mgr->list)) {
+      debug_printf("%s: unfreed buffers\n", __FUNCTION__);
+      pb_debug_manager_dump(mgr);
+   }
+   pipe_mutex_unlock(mgr->mutex);
+   
+   pipe_mutex_destroy(mgr->mutex);
    mgr->provider->destroy(mgr->provider);
    FREE(mgr);
 }
@@ -375,7 +472,10 @@ pb_debug_manager_create(struct pb_manager *provider,
    mgr->provider = provider;
    mgr->underflow_size = underflow_size;
    mgr->overflow_size = overflow_size;
-      
+    
+   pipe_mutex_init(mgr->mutex);
+   LIST_INITHEAD(&mgr->list);
+
    return &mgr->base;
 }
 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c
index e8bd7cda3b1..aba7a3f9374 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c
@@ -53,6 +53,7 @@
 #include "pipe/p_compiler.h"
 #include "pipe/p_state.h"
 #include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_dump.h"
 #include "tgsi/tgsi_parse.h"
 #include "tgsi/tgsi_util.h"
 #include "tgsi_exec.h"
@@ -169,6 +170,56 @@ print_temp(const struct tgsi_exec_machine *mach, uint index)
 #endif
 
 
+/**
+ * Check if there's a potential src/dst register data dependency when
+ * using SOA execution.
+ * Example:
+ *   MOV T, T.yxwz;
+ * This would expand into:
+ *   MOV t0, t1;
+ *   MOV t1, t0;
+ *   MOV t2, t3;
+ *   MOV t3, t2;
+ * The second instruction will have the wrong value for t0 if executed as-is.
+ */
+static boolean
+tgsi_check_soa_dependencies(const struct tgsi_full_instruction *inst)
+{
+   uint i, chan;
+
+   uint writemask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+   if (writemask == TGSI_WRITEMASK_X ||
+       writemask == TGSI_WRITEMASK_Y ||
+       writemask == TGSI_WRITEMASK_Z ||
+       writemask == TGSI_WRITEMASK_W ||
+       writemask == TGSI_WRITEMASK_NONE) {
+      /* no chance of data dependency */
+      return FALSE;
+   }
+
+   /* loop over src regs */
+   for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
+      if ((inst->FullSrcRegisters[i].SrcRegister.File ==
+           inst->FullDstRegisters[0].DstRegister.File) &&
+          (inst->FullSrcRegisters[i].SrcRegister.Index ==
+           inst->FullDstRegisters[0].DstRegister.Index)) {
+         /* loop over dest channels */
+         uint channelsWritten = 0x0;
+         FOR_EACH_ENABLED_CHANNEL(*inst, chan) {
+            /* check if we're reading a channel that's been written */
+            uint swizzle = tgsi_util_get_full_src_register_extswizzle(&inst->FullSrcRegisters[i], chan);
+            if (swizzle <= TGSI_SWIZZLE_W &&
+                (channelsWritten & (1 << swizzle))) {
+               return TRUE;
+            }
+
+            channelsWritten |= (1 << chan);
+         }
+      }
+   }
+   return FALSE;
+}
+
 
 /**
  * Initialize machine state by expanding tokens to full instructions,
@@ -280,6 +331,17 @@ tgsi_exec_machine_bind_shader(
          memcpy(instructions + numInstructions,
                 &parse.FullToken.FullInstruction,
                 sizeof(instructions[0]));
+
+#if 0
+         if (tgsi_check_soa_dependencies(&parse.FullToken.FullInstruction)) {
+            debug_printf("SOA dependency in instruction:\n");
+            tgsi_dump_instruction(&parse.FullToken.FullInstruction,
+                                  numInstructions);
+         }
+#else
+         (void) tgsi_check_soa_dependencies;
+#endif
+
          numInstructions++;
          break;
 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c
index 4b4e34b29eb..ba2bfdef062 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c
@@ -1466,6 +1466,31 @@ emit_cmp(
    }
 }
 
+
+/**
+ * Check if inst src/dest regs use indirect addressing into temporary
+ * register file.
+ */
+static boolean
+indirect_temp_reference(const struct tgsi_full_instruction *inst)
+{
+   uint i;
+   for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
+      const struct tgsi_full_src_register *reg = &inst->FullSrcRegisters[i];
+      if (reg->SrcRegister.File == TGSI_FILE_TEMPORARY &&
+          reg->SrcRegister.Indirect)
+         return TRUE;
+   }
+   for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
+      const struct tgsi_full_dst_register *reg = &inst->FullDstRegisters[i];
+      if (reg->DstRegister.File == TGSI_FILE_TEMPORARY &&
+          reg->DstRegister.Indirect)
+         return TRUE;
+   }
+   return FALSE;
+}
+
+
 static int
 emit_instruction(
    struct x86_function *func,
@@ -1473,6 +1498,10 @@ emit_instruction(
 {
    unsigned chan_index;
 
+   /* we can't handle indirect addressing into temp register file yet */
+   if (indirect_temp_reference(inst))
+      return FALSE;
+
    switch (inst->Instruction.Opcode) {
    case TGSI_OPCODE_ARL:
       FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
diff --git a/src/gallium/auxiliary/tgsi/tgsi_transform.c b/src/gallium/auxiliary/tgsi/tgsi_transform.c
index 062c1be938a..bc9c18fd4a7 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_transform.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_transform.c
@@ -198,3 +198,30 @@ tgsi_transform_shader(const struct tgsi_token *tokens_in,
 
    return ctx->ti;
 }
+
+
+#include "tgsi_text.h"
+
+extern int tgsi_transform_foo( struct tgsi_token *tokens_out,
+                               uint max_tokens_out );
+
+/* This function exists only so that tgsi_text_translate() doesn't get
+ * magic-ed out of the libtgsi.a archive by the build system.  Don't
+ * remove unless you know this has been fixed - check on mingw/scons
+ * builds as well.
+ */
+int
+tgsi_transform_foo( struct tgsi_token *tokens_out,
+                    uint max_tokens_out )
+{
+   const char *text = 
+      "FRAG1.1\n"
+      "DCL IN[0], COLOR, CONSTANT\n"
+      "DCL OUT[0], COLOR\n"
+      "  0: MOV OUT[0], IN[0]\n"
+      "  1: END";
+        
+   return tgsi_text_translate( text,
+                               tokens_out,
+                               max_tokens_out );
+}
diff --git a/src/gallium/auxiliary/util/u_debug.c b/src/gallium/auxiliary/util/u_debug.c
index ae47a274a69..18597ef8395 100644
--- a/src/gallium/auxiliary/util/u_debug.c
+++ b/src/gallium/auxiliary/util/u_debug.c
@@ -721,6 +721,7 @@ void
 debug_dump_surface_bmp(const char *filename,
                        struct pipe_surface *surface)
 {
+#ifndef PIPE_SUBSYSTEM_WINDOWS_MINIPORT
    struct pipe_transfer *transfer;
    struct pipe_texture *texture = surface->texture;
    struct pipe_screen *screen = texture->screen;
@@ -733,6 +734,7 @@ debug_dump_surface_bmp(const char *filename,
    debug_dump_transfer_bmp(filename, transfer);
 
    screen->tex_transfer_destroy(transfer);
+#endif
 }
 
 void
@@ -740,32 +742,56 @@ debug_dump_transfer_bmp(const char *filename,
                         struct pipe_transfer *transfer)
 {
 #ifndef PIPE_SUBSYSTEM_WINDOWS_MINIPORT
-   struct util_stream *stream;
-   struct bmp_file_header bmfh;
-   struct bmp_info_header bmih;
    float *rgba;
-   unsigned x, y;
 
    if (!transfer)
       goto error1;
 
    rgba = MALLOC(transfer->width*transfer->height*4*sizeof(float));
+   if(!rgba)
+      goto error1;
+
+   pipe_get_tile_rgba(transfer, 0, 0,
+                      transfer->width, transfer->height,
+                      rgba);
+
+   debug_dump_float_rgba_bmp(filename,
+                             transfer->width, transfer->height,
+                             rgba, transfer->width);
+
+   FREE(rgba);
+error1:
+   ;
+#endif
+}
+
+void
+debug_dump_float_rgba_bmp(const char *filename,
+                          unsigned width, unsigned height,
+                          float *rgba, unsigned stride)
+{
+#ifndef PIPE_SUBSYSTEM_WINDOWS_MINIPORT
+   struct util_stream *stream;
+   struct bmp_file_header bmfh;
+   struct bmp_info_header bmih;
+   unsigned x, y;
+
    if(!rgba)
       goto error1;
 
    bmfh.bfType = 0x4d42;
-   bmfh.bfSize = 14 + 40 + transfer->height*transfer->width*4;
+   bmfh.bfSize = 14 + 40 + height*width*4;
    bmfh.bfReserved1 = 0;
    bmfh.bfReserved2 = 0;
    bmfh.bfOffBits = 14 + 40;
 
    bmih.biSize = 40;
-   bmih.biWidth = transfer->width;
-   bmih.biHeight = transfer->height;
+   bmih.biWidth = width;
+   bmih.biHeight = height;
    bmih.biPlanes = 1;
    bmih.biBitCount = 32;
    bmih.biCompression = 0;
-   bmih.biSizeImage = transfer->height*transfer->width*4;
+   bmih.biSizeImage = height*width*4;
    bmih.biXPelsPerMeter = 0;
    bmih.biYPelsPerMeter = 0;
    bmih.biClrUsed = 0;
@@ -773,19 +799,15 @@ debug_dump_transfer_bmp(const char *filename,
 
    stream = util_stream_create(filename, bmfh.bfSize);
    if(!stream)
-      goto error2;
+      goto error1;
 
    util_stream_write(stream, &bmfh, 14);
    util_stream_write(stream, &bmih, 40);
 
-   pipe_get_tile_rgba(transfer, 0, 0,
-                      transfer->width, transfer->height,
-                      rgba);
-
-   y = transfer->height;
+   y = height;
    while(y--) {
-      float *ptr = rgba + (transfer->width * y * 4);
-      for(x = 0; x < transfer->width; ++x)
+      float *ptr = rgba + (stride * y * 4);
+      for(x = 0; x < width; ++x)
       {
          struct bmp_rgb_quad pixel;
          pixel.rgbRed   = float_to_ubyte(ptr[x*4 + 0]);
@@ -797,8 +819,6 @@ debug_dump_transfer_bmp(const char *filename,
    }
 
    util_stream_close(stream);
-error2:
-   FREE(rgba);
 error1:
    ;
 #endif
diff --git a/src/gallium/auxiliary/util/u_debug.h b/src/gallium/auxiliary/util/u_debug.h
index 8d703e47fce..d42b65ce281 100644
--- a/src/gallium/auxiliary/util/u_debug.h
+++ b/src/gallium/auxiliary/util/u_debug.h
@@ -102,6 +102,22 @@ debug_printf(const char *format, ...)
 }
 
 
+/*
+ * ... isn't portable so we need to pass arguments in parentheses.
+ *
+ * usage:
+ *    debug_printf_once(("awnser: %i\n", 42));
+ */
+#define debug_printf_once(args) \
+   do { \
+      static boolean once = TRUE; \
+      if (once) { \
+         once = FALSE; \
+         debug_printf args; \
+      } \
+   } while (0)
+
+
 #ifdef DEBUG
 #define debug_vprintf(_format, _ap) _debug_vprintf(_format, _ap)
 #else
@@ -347,10 +363,15 @@ void debug_dump_surface_bmp(const char *filename,
                             struct pipe_surface *surface);
 void debug_dump_transfer_bmp(const char *filename,
                              struct pipe_transfer *transfer);
+void debug_dump_float_rgba_bmp(const char *filename,
+                               unsigned width, unsigned height,
+                               float *rgba, unsigned stride);
 #else
 #define debug_dump_image(prefix, format, cpp, width, height, stride, data) ((void)0)
 #define debug_dump_surface(prefix, surface) ((void)0)
 #define debug_dump_surface_bmp(filename, surface) ((void)0)
+#define debug_dump_transfer_bmp(filename, transfer) ((void)0)
+#define debug_dump_rgba_float_bmp(filename, width, height, rgba, stride) ((void)0)
 #endif
 
 
diff --git a/src/gallium/auxiliary/util/u_debug_stack.c b/src/gallium/auxiliary/util/u_debug_stack.c
index e9891fde8a3..528a1c394be 100644
--- a/src/gallium/auxiliary/util/u_debug_stack.c
+++ b/src/gallium/auxiliary/util/u_debug_stack.c
@@ -62,6 +62,8 @@ debug_backtrace_capture(struct debug_stack_frame *backtrace,
    
 #ifdef PIPE_ARCH_X86
    while(nr_frames) {
+      const void **next_frame_pointer;
+
       if(!frame_pointer)
          break;
       
@@ -72,7 +74,14 @@ debug_backtrace_capture(struct debug_stack_frame *backtrace,
          --nr_frames;
       }
       
-      frame_pointer = (const void **)frame_pointer[0];
+      next_frame_pointer = (const void **)frame_pointer[0];
+      
+      /* Limit the stack walk to avoid referencing undefined memory */
+      if((uintptr_t)next_frame_pointer <= (uintptr_t)frame_pointer ||
+         (uintptr_t)next_frame_pointer > (uintptr_t)frame_pointer + 64*1024)
+         break;
+      
+      frame_pointer = next_frame_pointer;
    }
 #endif
    
diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c
index 690412ae7d3..6fa13a8ce11 100644
--- a/src/gallium/auxiliary/util/u_gen_mipmap.c
+++ b/src/gallium/auxiliary/util/u_gen_mipmap.c
@@ -1316,7 +1316,6 @@ util_create_gen_mipmap(struct pipe_context *pipe,
    for (i = 0; i < 4; i++) {
       ctx->vertices[i][0][2] = 0.0f; /* z */
       ctx->vertices[i][0][3] = 1.0f; /* w */
-      ctx->vertices[i][1][2] = 0.0f; /* r */
       ctx->vertices[i][1][3] = 1.0f; /* q */
    }
 
@@ -1350,29 +1349,104 @@ get_next_slot(struct gen_mipmap_state *ctx)
 
 
 static unsigned
-set_vertex_data(struct gen_mipmap_state *ctx, float width, float height)
+set_vertex_data(struct gen_mipmap_state *ctx,
+                enum pipe_texture_target tex_target,
+                uint face, float width, float height)
 {
    unsigned offset;
 
+   /* vert[0].position */
    ctx->vertices[0][0][0] = 0.0f; /*x*/
    ctx->vertices[0][0][1] = 0.0f; /*y*/
-   ctx->vertices[0][1][0] = 0.0f; /*s*/
-   ctx->vertices[0][1][1] = 0.0f; /*t*/
 
+   /* vert[1].position */
    ctx->vertices[1][0][0] = width;
    ctx->vertices[1][0][1] = 0.0f;
-   ctx->vertices[1][1][0] = 1.0f;
-   ctx->vertices[1][1][1] = 0.0f;
 
+   /* vert[2].position */
    ctx->vertices[2][0][0] = width;
    ctx->vertices[2][0][1] = height;
-   ctx->vertices[2][1][0] = 1.0f;
-   ctx->vertices[2][1][1] = 1.0f;
 
+   /* vert[3].position */
    ctx->vertices[3][0][0] = 0.0f;
    ctx->vertices[3][0][1] = height;
-   ctx->vertices[3][1][0] = 0.0f;
-   ctx->vertices[3][1][1] = 1.0f;
+
+   /* Setup vertex texcoords.  This is a little tricky for cube maps. */
+   if (tex_target == PIPE_TEXTURE_CUBE) {
+      static const float st[4][2] = {
+         {0.0f, 0.0f}, {1.0f, 0.0f}, {1.0f, 1.0f}, {0.0f, 1.0f}
+      };
+      float rx, ry, rz;
+      uint i;
+
+      /* loop over quad verts */
+      for (i = 0; i < 4; i++) {
+         /* Compute sc = +/-scale and tc = +/-scale.
+          * Not +/-1 to avoid cube face selection ambiguity near the edges,
+          * though that can still sometimes happen with this scale factor...
+          */
+         const float scale = 0.9999;
+         const float sc = (2.0f * st[i][0] - 1.0f) * scale;
+         const float tc = (2.0f * st[i][1] - 1.0f) * scale;
+
+         switch (face) {
+         case PIPE_TEX_FACE_POS_X:
+            rx = 1.0f;
+            ry = -tc;
+            rz = -sc;
+            break;
+         case PIPE_TEX_FACE_NEG_X:
+            rx = -1.0f;
+            ry = -tc;
+            rz = sc;
+            break;
+         case PIPE_TEX_FACE_POS_Y:
+            rx = sc;
+            ry = 1.0f;
+            rz = tc;
+            break;
+         case PIPE_TEX_FACE_NEG_Y:
+            rx = sc;
+            ry = -1.0f;
+            rz = -tc;
+            break;
+         case PIPE_TEX_FACE_POS_Z:
+            rx = sc;
+            ry = -tc;
+            rz = 1.0f;
+            break;
+         case PIPE_TEX_FACE_NEG_Z:
+            rx = -sc;
+            ry = -tc;
+            rz = -1.0f;
+            break;
+         default:
+            assert(0);
+         }
+
+         ctx->vertices[i][1][0] = rx; /*s*/
+         ctx->vertices[i][1][1] = ry; /*t*/
+         ctx->vertices[i][1][2] = rz; /*r*/
+      }
+   }
+   else {
+      /* 1D/2D */
+      ctx->vertices[0][1][0] = 0.0f; /*s*/
+      ctx->vertices[0][1][1] = 0.0f; /*t*/
+      ctx->vertices[0][1][2] = 0.0f; /*r*/
+
+      ctx->vertices[1][1][0] = 1.0f;
+      ctx->vertices[1][1][1] = 0.0f;
+      ctx->vertices[1][1][2] = 0.0f;
+
+      ctx->vertices[2][1][0] = 1.0f;
+      ctx->vertices[2][1][1] = 1.0f;
+      ctx->vertices[2][1][2] = 0.0f;
+
+      ctx->vertices[3][1][0] = 0.0f;
+      ctx->vertices[3][1][1] = 1.0f;
+      ctx->vertices[3][1][2] = 0.0f;
+   }
 
    offset = get_next_slot( ctx );
 
@@ -1503,6 +1577,8 @@ util_gen_mipmap(struct gen_mipmap_state *ctx,
 
       /* quad coords in window coords (bypassing vs, clip and viewport) */
       offset = set_vertex_data(ctx,
+                               pt->target,
+                               face,
                                (float) pt->width[dstLevel],
                                (float) pt->height[dstLevel]);
 
diff --git a/src/gallium/auxiliary/util/u_linear.c b/src/gallium/auxiliary/util/u_linear.c
index 6be365e53bb..a1dce3f5cf4 100644
--- a/src/gallium/auxiliary/util/u_linear.c
+++ b/src/gallium/auxiliary/util/u_linear.c
@@ -1,3 +1,34 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Functions for converting tiled data to linear and vice versa.
+ */
+
 
 #include "util/u_debug.h"
 #include "u_linear.h"
diff --git a/src/gallium/auxiliary/util/u_linear.h b/src/gallium/auxiliary/util/u_linear.h
index 1589f029bc4..b74308ffa3d 100644
--- a/src/gallium/auxiliary/util/u_linear.h
+++ b/src/gallium/auxiliary/util/u_linear.h
@@ -1,3 +1,34 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Functions for converting tiled data to linear and vice versa.
+ */
+
 
 #ifndef U_LINEAR_H
 #define U_LINEAR_H
diff --git a/src/gallium/auxiliary/util/u_time.c b/src/gallium/auxiliary/util/u_time.c
index 8afe4fccf7b..5268cbf79ce 100644
--- a/src/gallium/auxiliary/util/u_time.c
+++ b/src/gallium/auxiliary/util/u_time.c
@@ -217,4 +217,9 @@ void util_time_sleep(unsigned usecs)
    } while(start <= curr && curr < end || 
 	   end < start && (curr < end || start <= curr));
 }
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+void util_time_sleep(unsigned usecs)
+{
+   Sleep((usecs + 999)/ 1000);
+}
 #endif
diff --git a/src/gallium/auxiliary/util/u_upload_mgr.c b/src/gallium/auxiliary/util/u_upload_mgr.c
index d9c0d7afa89..2eb98068c86 100644
--- a/src/gallium/auxiliary/util/u_upload_mgr.c
+++ b/src/gallium/auxiliary/util/u_upload_mgr.c
@@ -70,7 +70,7 @@ struct u_upload_mgr *u_upload_create( struct pipe_screen *screen,
 }
 
 
-static INLINE void
+static INLINE enum pipe_error
 my_buffer_write(struct pipe_screen *screen,
                 struct pipe_buffer *buf,
                 unsigned offset, unsigned size, unsigned dirty_size,
@@ -84,12 +84,14 @@ my_buffer_write(struct pipe_screen *screen,
    assert(size);
 
    map = pipe_buffer_map_range(screen, buf, offset, size, PIPE_BUFFER_USAGE_CPU_WRITE);
-   assert(map);
-   if(map) {
-      memcpy(map + offset, data, size);
-      pipe_buffer_flush_mapped_range(screen, buf, offset, dirty_size);
-      pipe_buffer_unmap(screen, buf);
-   }
+   if (map == NULL) 
+      return PIPE_ERROR_OUT_OF_MEMORY;
+
+   memcpy(map + offset, data, size);
+   pipe_buffer_flush_mapped_range(screen, buf, offset, dirty_size);
+   pipe_buffer_unmap(screen, buf);
+
+   return PIPE_OK;
 }
 
 /* Release old buffer.
@@ -162,12 +164,14 @@ enum pipe_error u_upload_data( struct u_upload_mgr *upload,
 
    /* Copy the data, using map_range if available:
     */
-   my_buffer_write( upload->screen, 
-                    upload->buffer,
-                    upload->offset,
-                    size, 
-                    alloc_size,
-                    data );
+   ret = my_buffer_write( upload->screen, 
+                          upload->buffer,
+                          upload->offset,
+                          size, 
+                          alloc_size,
+                          data );
+   if (ret)
+      return ret;
 
    /* Emit the return values:
     */
diff --git a/src/gallium/drivers/i915simple/Makefile b/src/gallium/drivers/i915simple/Makefile
index 12821c5a761..8870b398661 100644
--- a/src/gallium/drivers/i915simple/Makefile
+++ b/src/gallium/drivers/i915simple/Makefile
@@ -8,7 +8,6 @@ C_SOURCES = \
 	i915_clear.c \
 	i915_flush.c \
 	i915_context.c \
-	i915_context.c \
 	i915_debug.c \
 	i915_debug_fp.c \
 	i915_state.c \
diff --git a/src/gallium/drivers/nv04/nv04_miptree.c b/src/gallium/drivers/nv04/nv04_miptree.c
index 4da833c25e8..93f752faec9 100644
--- a/src/gallium/drivers/nv04/nv04_miptree.c
+++ b/src/gallium/drivers/nv04/nv04_miptree.c
@@ -31,7 +31,8 @@ nv04_miptree_layout(struct nv04_miptree *nv04mt)
 
 	for (l = 0; l <= pt->last_level; l++) {
 
-		nv04mt->level[l].image_offset = offset;
+		nv04mt->level[l].image_offset = 
+			CALLOC(nr_faces, sizeof(unsigned));
 		offset += nv04mt->level[l].pitch * pt->height[l];
 	}
 
diff --git a/src/gallium/drivers/nv04/nv04_state.h b/src/gallium/drivers/nv04/nv04_state.h
index 0d51439e3ff..399f750dbe7 100644
--- a/src/gallium/drivers/nv04/nv04_state.h
+++ b/src/gallium/drivers/nv04/nv04_state.h
@@ -37,7 +37,7 @@ struct nv04_miptree {
 
 	struct {
 		uint pitch;
-		uint image_offset;
+		uint *image_offset;
 	} level[PIPE_MAX_TEXTURE_LEVELS];
 };
 
diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index 6bdf544a05c..a4e89c37d1a 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -34,10 +34,6 @@ static boolean r300_draw_range_elements(struct pipe_context* pipe,
     struct r300_context* r300 = r300_context(pipe);
     int i;
 
-    if (r300->dirty_state) {
-        r300_emit_dirty_state(r300);
-    }
-
     for (i = 0; i < r300->vertex_buffer_count; i++) {
         void* buf = pipe_buffer_map(pipe->screen,
                                     r300->vertex_buffers[i].buffer,
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 4c695c1195a..96f1f11246c 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -264,6 +264,11 @@ struct r300_context {
     /* Draw module. Used mostly for SW TCL. */
     struct draw_context* draw;
 
+    /* Vertex buffer for rendering. */
+    struct pipe_buffer* vbo;
+    /* Offset into the VBO. */
+    size_t vbo_offset;
+
     /* Various CSO state objects. */
     /* Blend state. */
     struct r300_blend_state* blend_state;
@@ -289,7 +294,7 @@ struct r300_context {
     /* Texture states. */
     struct r300_texture* textures[8];
     int texture_count;
-    /* Vertex buffers. */
+    /* Vertex buffers for Gallium. */
     struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS];
     int vertex_buffer_count;
     /* Vertex information. */
@@ -314,11 +319,4 @@ struct draw_stage* r300_draw_stage(struct r300_context* r300);
 void r300_init_state_functions(struct r300_context* r300);
 void r300_init_surface_functions(struct r300_context* r300);
 
-/* Fun with includes: r300_winsys also declares this prototype.
- * We'll just step out in that case... */
-#ifndef R300_WINSYS_H
-struct pipe_context* r300_create_context(struct pipe_screen* screen,
-                                         struct r300_winsys* r300_winsys);
-#endif
-
 #endif /* R300_CONTEXT_H */
diff --git a/src/gallium/drivers/r300/r300_cs.h b/src/gallium/drivers/r300/r300_cs.h
index 5d9799dd723..82a3942248e 100644
--- a/src/gallium/drivers/r300/r300_cs.h
+++ b/src/gallium/drivers/r300/r300_cs.h
@@ -49,27 +49,27 @@
 
 #define CS_LOCALS(context) \
     struct r300_winsys* cs_winsys = context->winsys; \
-    struct radeon_cs* cs = cs_winsys->cs; \
     int cs_count = 0;
 
 #define CHECK_CS(size) \
-    cs_winsys->check_cs(cs, (size))
+    cs_winsys->check_cs(cs_winsys, (size))
 
 #define BEGIN_CS(size) do { \
     CHECK_CS(size); \
     debug_printf("r300: BEGIN_CS, count %d, in %s (%s:%d)\n", \
         size, __FUNCTION__, __FILE__, __LINE__); \
-    cs_winsys->begin_cs(cs, (size), __FILE__, __FUNCTION__, __LINE__); \
+    cs_winsys->begin_cs(cs_winsys, (size), \
+            __FILE__, __FUNCTION__, __LINE__); \
     cs_count = size; \
 } while (0)
 
 #define OUT_CS(value) do { \
-    cs_winsys->write_cs_dword(cs, (value)); \
+    cs_winsys->write_cs_dword(cs_winsys, (value)); \
     cs_count--; \
 } while (0)
 
 #define OUT_CS_32F(value) do { \
-    cs_winsys->write_cs_dword(cs, fui(value)); \
+    cs_winsys->write_cs_dword(cs_winsys, fui(value)); \
     cs_count--; \
 } while (0)
 
@@ -97,7 +97,7 @@
         bo, offset); \
     assert(bo); \
     OUT_CS(offset); \
-    cs_winsys->write_cs_reloc(cs, bo, rd, wd, flags); \
+    cs_winsys->write_cs_reloc(cs_winsys, bo, rd, wd, flags); \
     cs_count -= 2; \
 } while (0)
 
@@ -106,13 +106,13 @@
         __LINE__); \
     if (cs_count != 0) \
         debug_printf("r300: Warning: cs_count off by %d\n", cs_count); \
-    cs_winsys->end_cs(cs, __FILE__, __FUNCTION__, __LINE__); \
+    cs_winsys->end_cs(cs_winsys, __FILE__, __FUNCTION__, __LINE__); \
 } while (0)
 
 #define FLUSH_CS do { \
     debug_printf("r300: FLUSH_CS in %s (%s:%d)\n\n", __FUNCTION__, __FILE__, \
         __LINE__); \
-    cs_winsys->flush_cs(cs); \
+    cs_winsys->flush_cs(cs_winsys); \
 } while (0)
 
 #define RADEON_ONE_REG_WR        (1 << 15)
@@ -138,7 +138,7 @@
     assert(bo); \
     OUT_CS(offset); \
     OUT_CS(count); \
-    cs_winsys->write_cs_reloc(cs, bo, rd, wd, flags); \
+    cs_winsys->write_cs_reloc(cs_winsys, bo, rd, wd, flags); \
     cs_count -= 2; \
 } while (0)
 
diff --git a/src/gallium/drivers/r300/r300_debug.c b/src/gallium/drivers/r300/r300_debug.c
index dd63136c9d6..1ff72172ebd 100644
--- a/src/gallium/drivers/r300/r300_debug.c
+++ b/src/gallium/drivers/r300/r300_debug.c
@@ -30,81 +30,6 @@ static void r300_dump_fs(struct r300_fragment_shader* fs)
     }
 }
 
-static char* r500_fs_swiz[] = {
-    " R",
-    " G",
-    " B",
-    " A",
-    " 0",
-    ".5",
-    " 1",
-    " U",
-};
-
-static char* r500_fs_op_rgb[] = {
-    "MAD",
-    "DP3",
-    "DP4",
-    "D2A",
-    "MIN",
-    "MAX",
-    "---",
-    "CND",
-    "CMP",
-    "FRC",
-    "SOP",
-    "MDH",
-    "MDV",
-};
-
-static char* r500_fs_op_alpha[] = {
-    "MAD",
-    " DP",
-    "MIN",
-    "MAX",
-    "---",
-    "CND",
-    "CMP",
-    "FRC",
-    "EX2",
-    "LN2",
-    "RCP",
-    "RSQ",
-    "SIN",
-    "COS",
-    "MDH",
-    "MDV",
-};
-
-static char* r500_fs_mask[] = {
-    "NONE",
-    "R   ",
-    " G  ",
-    "RG  ",
-    "  B ",
-    "R B ",
-    " GB ",
-    "RGB ",
-    "   A",
-    "R  A",
-    " G A",
-    "RG A",
-    "  BA",
-    "R BA",
-    " GBA",
-    "RGBA",
-};
-
-static char* r500_fs_tex[] = {
-    "    NOP",
-    "     LD",
-    "TEXKILL",
-    "   PROJ",
-    "LODBIAS",
-    "    LOD",
-    "   DXDY",
-};
-
 void r500_fs_dump(struct r500_fragment_shader* fs)
 {
     int i;
@@ -225,12 +150,25 @@ void r500_fs_dump(struct r500_fragment_shader* fs)
     }
 }
 
+static void r300_vs_op_dump(uint32_t op)
+{
+    if (op & 0x81) {
+        debug_printf("PVS_MACRO_OP_2CLK_M2X_ADD\n");
+    } else if (op & 0x80) {
+        debug_printf("   PVS_MACRO_OP_2CLK_MADD\n");
+    } else if (op & 0x40) {
+        debug_printf("%s\n", r300_vs_me_ops[op & 0x1f]);
+    } else {
+        debug_printf("%s\n", r300_vs_ve_ops[op & 0x1f]);
+    }
+}
+
 void r300_vs_dump(struct r300_vertex_shader* vs)
 {
     int i;
 
     for (i = 0; i < vs->instruction_count; i++) {
-        debug_printf("inst0: 0x%x\n", vs->instructions[i].inst0);
+        r300_vs_op_dump(vs->instructions[i].inst0);
         debug_printf("inst1: 0x%x\n", vs->instructions[i].inst1);
         debug_printf("inst2: 0x%x\n", vs->instructions[i].inst2);
         debug_printf("inst3: 0x%x\n", vs->instructions[i].inst3);
diff --git a/src/gallium/drivers/r300/r300_debug.h b/src/gallium/drivers/r300/r300_debug.h
index a1f873656dc..63065940997 100644
--- a/src/gallium/drivers/r300/r300_debug.h
+++ b/src/gallium/drivers/r300/r300_debug.h
@@ -27,6 +27,152 @@
 #include "r300_state_shader.h"
 #include "r300_state_tcl.h"
 
+static char* r500_fs_swiz[] = {
+    " R",
+    " G",
+    " B",
+    " A",
+    " 0",
+    ".5",
+    " 1",
+    " U",
+};
+
+static char* r500_fs_op_rgb[] = {
+    "MAD",
+    "DP3",
+    "DP4",
+    "D2A",
+    "MIN",
+    "MAX",
+    "---",
+    "CND",
+    "CMP",
+    "FRC",
+    "SOP",
+    "MDH",
+    "MDV",
+};
+
+static char* r500_fs_op_alpha[] = {
+    "MAD",
+    " DP",
+    "MIN",
+    "MAX",
+    "---",
+    "CND",
+    "CMP",
+    "FRC",
+    "EX2",
+    "LN2",
+    "RCP",
+    "RSQ",
+    "SIN",
+    "COS",
+    "MDH",
+    "MDV",
+};
+
+static char* r500_fs_mask[] = {
+    "NONE",
+    "R   ",
+    " G  ",
+    "RG  ",
+    "  B ",
+    "R B ",
+    " GB ",
+    "RGB ",
+    "   A",
+    "R  A",
+    " G A",
+    "RG A",
+    "  BA",
+    "R BA",
+    " GBA",
+    "RGBA",
+};
+
+static char* r500_fs_tex[] = {
+    "    NOP",
+    "     LD",
+    "TEXKILL",
+    "   PROJ",
+    "LODBIAS",
+    "    LOD",
+    "   DXDY",
+};
+
+static char* r300_vs_ve_ops[] = {
+    /* R300 vector ops */
+    "                 VE_NO_OP",
+    "           VE_DOT_PRODUCT",
+    "              VE_MULTIPLY",
+    "                   VE_ADD",
+    "          VE_MULTIPLY_ADD",
+    "       VE_DISTANCE_FACTOR",
+    "              VE_FRACTION",
+    "               VE_MAXIMUM",
+    "               VE_MINIMUM",
+    "VE_SET_GREATER_THAN_EQUAL",
+    "         VE_SET_LESS_THAN",
+    "        VE_MULTIPLYX2_ADD",
+    "        VE_MULTIPLY_CLAMP",
+    "            VE_FLT2FIX_DX",
+    "        VE_FLT2FIX_DX_RND",
+    /* R500 vector ops */
+    "      VE_PRED_SET_EQ_PUSH",
+    "      VE_PRED_SET_GT_PUSH",
+    "     VE_PRED_SET_GTE_PUSH",
+    "     VE_PRED_SET_NEQ_PUSH",
+    "         VE_COND_WRITE_EQ",
+    "         VE_COND_WRITE_GT",
+    "        VE_COND_WRITE_GTE",
+    "        VE_COND_WRITE_NEQ",
+    "      VE_SET_GREATER_THAN",
+    "             VE_SET_EQUAL",
+    "         VE_SET_NOT_EQUAL",
+    "               (reserved)",
+    "               (reserved)",
+    "               (reserved)",
+};
+
+static char* r300_vs_me_ops[] = {
+    /* R300 math ops */
+    "                 ME_NO_OP",
+    "          ME_EXP_BASE2_DX",
+    "          ME_LOG_BASE2_DX",
+    "          ME_EXP_BASEE_FF",
+    "        ME_LIGHT_COEFF_DX",
+    "         ME_POWER_FUNC_FF",
+    "              ME_RECIP_DX",
+    "              ME_RECIP_FF",
+    "         ME_RECIP_SQRT_DX",
+    "         ME_RECIP_SQRT_FF",
+    "              ME_MULTIPLY",
+    "     ME_EXP_BASE2_FULL_DX",
+    "     ME_LOG_BASE2_FULL_DX",
+    " ME_POWER_FUNC_FF_CLAMP_B",
+    "ME_POWER_FUNC_FF_CLAMP_B1",
+    "ME_POWER_FUNC_FF_CLAMP_01",
+    "                   ME_SIN",
+    "                   ME_COS",
+    /* R500 math ops */
+    "        ME_LOG_BASE2_IEEE",
+    "            ME_RECIP_IEEE",
+    "       ME_RECIP_SQRT_IEEE",
+    "           ME_PRED_SET_EQ",
+    "           ME_PRED_SET_GT",
+    "          ME_PRED_SET_GTE",
+    "          ME_PRED_SET_NEQ",
+    "          ME_PRED_SET_CLR",
+    "          ME_PRED_SET_INV",
+    "          ME_PRED_SET_POP",
+    "      ME_PRED_SET_RESTORE",
+    "               (reserved)",
+    "               (reserved)",
+    "               (reserved)",
+};
+
 void r500_fs_dump(struct r500_fragment_shader* fs);
 
 void r300_vs_dump(struct r300_vertex_shader* vs);
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 417d5f6307d..c73d5a0b443 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -152,34 +152,39 @@ void r500_emit_fragment_shader(struct r300_context* r300,
     END_CS;
 }
 
-/* XXX add pitch, stride, clean up */
 void r300_emit_fb_state(struct r300_context* r300,
                         struct pipe_framebuffer_state* fb)
 {
-    int i;
     struct r300_texture* tex;
+    unsigned pixpitch;
+    int i;
     CS_LOCALS(r300);
 
-    BEGIN_CS((6 * fb->nr_cbufs) + (fb->zsbuf ? 6 : 0) + 4);
+    BEGIN_CS((8 * fb->nr_cbufs) + (fb->zsbuf ? 8 : 0) + 4);
     for (i = 0; i < fb->nr_cbufs; i++) {
         tex = (struct r300_texture*)fb->cbufs[i]->texture;
+        pixpitch = tex->stride / tex->tex.block.size;
+
         OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0 + (4 * i), 1);
         OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
 
+        OUT_CS_REG(R300_RB3D_COLORPITCH0 + (4 * i), pixpitch |
+            r300_translate_colorformat(tex->tex.format));
+
         OUT_CS_REG(R300_US_OUT_FMT_0 + (4 * i),
             r300_translate_out_fmt(fb->cbufs[i]->format));
     }
 
     if (fb->zsbuf) {
         tex = (struct r300_texture*)fb->zsbuf->texture;
+        pixpitch = (tex->stride / tex->tex.block.size);
+
         OUT_CS_REG_SEQ(R300_ZB_DEPTHOFFSET, 1);
         OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
-        if (fb->zsbuf->format == PIPE_FORMAT_Z24S8_UNORM) {
-            OUT_CS_REG(R300_ZB_FORMAT,
-                R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL);
-        } else {
-            OUT_CS_REG(R300_ZB_FORMAT, 0x0);
-        }
+
+        OUT_CS_REG(R300_ZB_FORMAT, r300_translate_zsformat(tex->tex.format));
+
+        OUT_CS_REG(R300_ZB_DEPTHPITCH, pixpitch);
     }
 
     OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT,
@@ -291,6 +296,30 @@ void r300_emit_texture(struct r300_context* r300,
     END_CS;
 }
 
+void r300_emit_vertex_buffer(struct r300_context* r300)
+{
+    CS_LOCALS(r300);
+
+    debug_printf("r300: Preparing vertex buffer %p for render, "
+            "vertex size %d\n", r300->vbo,
+            r300->vertex_info.vinfo.size);
+    /* Set the pointer to our vertex buffer. The emitted values are this:
+     * PACKET3 [3D_LOAD_VBPNTR]
+     * COUNT   [1]
+     * FORMAT  [size | stride << 8]
+     * OFFSET  [offset into BO]
+     * VBPNTR  [relocated BO]
+     */
+    BEGIN_CS(7);
+    OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, 3);
+    OUT_CS(1);
+    OUT_CS(r300->vertex_info.vinfo.size |
+            (r300->vertex_info.vinfo.size << 8));
+    OUT_CS(r300->vbo_offset);
+    OUT_CS_RELOC(r300->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0);
+    END_CS;
+}
+
 void r300_emit_vertex_format_state(struct r300_context* r300)
 {
     int i;
@@ -416,16 +445,45 @@ void r300_flush_textures(struct r300_context* r300)
 void r300_emit_dirty_state(struct r300_context* r300)
 {
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
+    struct r300_texture* tex;
     int i;
     int dirty_tex = 0;
 
-    if (!(r300->dirty_state) && !(r300->dirty_hw)) {
+    if (!(r300->dirty_state)) {
         return;
     }
 
     r300_update_derived_state(r300);
 
     /* XXX check size */
+    /* Color buffers... */
+    for (i = 0; i < r300->framebuffer_state.nr_cbufs; i++) {
+        tex = (struct r300_texture*)r300->framebuffer_state.cbufs[i]->texture;
+        assert(tex && tex->buffer && "cbuf is marked, but NULL!");
+        if (!tex->buffer) return;
+        r300->winsys->add_buffer(r300->winsys, tex->buffer,
+                0, RADEON_GEM_DOMAIN_VRAM);
+    }
+    /* ...depth buffer... */
+    if (r300->framebuffer_state.zsbuf) {
+        tex = (struct r300_texture*)r300->framebuffer_state.zsbuf->texture;
+        assert(tex && tex->buffer && "zsbuf is marked, but NULL!");
+        if (!tex->buffer) return;
+        r300->winsys->add_buffer(r300->winsys, tex->buffer,
+                0, RADEON_GEM_DOMAIN_VRAM);
+    }
+    /* ...and vertex buffer. */
+    if (r300->vbo) {
+        r300->winsys->add_buffer(r300->winsys, r300->vbo,
+                RADEON_GEM_DOMAIN_GTT, 0);
+    } else {
+        debug_printf("No VBO while emitting dirty state!\n");
+    }
+
+    if (r300->winsys->validate(r300->winsys)) {
+        /* XXX */
+        r300->context.flush(&r300->context, 0, NULL);
+    }
 
     if (r300->dirty_state & R300_NEW_BLEND) {
         r300_emit_blend_state(r300, r300->blend_state);
@@ -506,4 +564,9 @@ void r300_emit_dirty_state(struct r300_context* r300)
         r300_emit_vertex_format_state(r300);
         r300->dirty_state &= ~R300_NEW_VERTEX_FORMAT;
     }
+
+    /* Finally, emit the VBO. */
+    r300_emit_vertex_buffer(r300);
+
+    r300->dirty_hw++;
 }
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index 31dbc7ab853..36e14f69f78 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -62,6 +62,8 @@ void r300_emit_scissor_state(struct r300_context* r300,
 void r300_emit_texture(struct r300_context* r300,
                        struct r300_texture* tex, unsigned offset);
 
+void r300_emit_vertex_buffer(struct r300_context* r300);
+
 void r300_emit_vertex_format_state(struct r300_context* r300);
 
 void r300_emit_vertex_shader(struct r300_context* r300,
diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c
index 20ca6905ad2..89a5f2b20cf 100644
--- a/src/gallium/drivers/r300/r300_flush.c
+++ b/src/gallium/drivers/r300/r300_flush.c
@@ -29,6 +29,8 @@ static void r300_flush(struct pipe_context* pipe,
     struct r300_context* r300 = r300_context(pipe);
     CS_LOCALS(r300);
 
+    draw_flush(r300->draw);
+
     if (r300->dirty_hw) {
         FLUSH_CS;
         r300_emit_invariant_state(r300);
diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c
index cbd84d7c569..29b66cee7ec 100644
--- a/src/gallium/drivers/r300/r300_render.c
+++ b/src/gallium/drivers/r300/r300_render.c
@@ -180,27 +180,10 @@ static void prepare_render(struct r300_render* render, unsigned count)
 
     CS_LOCALS(r300);
 
-    /* Make sure that all possible state is emitted. */
-    r300_emit_dirty_state(r300);
+    r300->vbo = render->vbo;
+    r300->vbo_offset = render->vbo_offset;
 
-    debug_printf("r300: Preparing vertex buffer %p for render, "
-            "vertex size %d, vertex count %d\n", render->vbo,
-            r300->vertex_info.vinfo.size, count);
-    /* Set the pointer to our vertex buffer. The emitted values are this:
-     * PACKET3 [3D_LOAD_VBPNTR]
-     * COUNT   [1]
-     * FORMAT  [size | stride << 8]
-     * OFFSET  [0]
-     * VBPNTR  [relocated BO]
-     */
-    BEGIN_CS(7);
-    OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, 3);
-    OUT_CS(1);
-    OUT_CS(r300->vertex_info.vinfo.size |
-            (r300->vertex_info.vinfo.size << 8));
-    OUT_CS(render->vbo_offset);
-    OUT_CS_RELOC(render->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0);
-    END_CS;
+    r300_emit_dirty_state(r300);
 }
 
 static void r300_render_draw_arrays(struct vbuf_render* render,
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index c9507ae1937..0143e228c4e 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -421,6 +421,7 @@ static void r300_bind_rs_state(struct pipe_context* pipe, void* state)
     struct r300_context* r300 = r300_context(pipe);
     struct r300_rs_state* rs = (struct r300_rs_state*)state;
 
+    draw_flush(r300->draw);
     draw_set_rasterizer_state(r300->draw, &rs->rs);
 
     r300->rs_state = rs;
@@ -528,7 +529,6 @@ static void r300_set_scissor_state(struct pipe_context* pipe,
                                    const struct pipe_scissor_state* state)
 {
     struct r300_context* r300 = r300_context(pipe);
-    draw_flush(r300->draw);
 
     if (r300_screen(r300->context.screen)->caps->is_r500) {
         r300->scissor_state->scissor_top_left =
@@ -555,19 +555,24 @@ static void r300_set_viewport_state(struct pipe_context* pipe,
 {
     struct r300_context* r300 = r300_context(pipe);
 
+    draw_flush(r300->draw);
+
     if (r300_screen(r300->context.screen)->caps->has_tcl) {
         /* Do the transform in HW. */
         r300->viewport_state->vte_control = R300_VTX_W0_FMT;
 
         if (state->scale[0] != 1.0f) {
+            assert(state->scale[0] != 0.0f);
             r300->viewport_state->xscale = state->scale[0];
             r300->viewport_state->vte_control |= R300_VPORT_X_SCALE_ENA;
         }
         if (state->scale[1] != 1.0f) {
+            assert(state->scale[1] != 0.0f);
             r300->viewport_state->yscale = state->scale[1];
             r300->viewport_state->vte_control |= R300_VPORT_Y_SCALE_ENA;
         }
         if (state->scale[2] != 1.0f) {
+            assert(state->scale[2] != 0.0f);
             r300->viewport_state->zscale = state->scale[2];
             r300->viewport_state->vte_control |= R300_VPORT_Z_SCALE_ENA;
         }
@@ -642,6 +647,8 @@ static void r300_bind_vs_state(struct pipe_context* pipe, void* shader)
 {
     struct r300_context* r300 = r300_context(pipe);
 
+    draw_flush(r300->draw);
+
     if (r300_screen(pipe->screen)->caps->has_tcl) {
         struct r300_vertex_shader* vs = (struct r300_vertex_shader*)shader;
 
diff --git a/src/gallium/drivers/r300/r300_state_tcl.c b/src/gallium/drivers/r300/r300_state_tcl.c
index bb96e2ad67f..ed9164db496 100644
--- a/src/gallium/drivers/r300/r300_state_tcl.c
+++ b/src/gallium/drivers/r300/r300_state_tcl.c
@@ -34,14 +34,20 @@ static void r300_vs_declare(struct r300_vs_asm* assembler,
                     assembler->tab[decl->DeclarationRange.First] = 0;
                     break;
                 case TGSI_SEMANTIC_COLOR:
-                    assembler->tab[decl->DeclarationRange.First] = 2;
+                    assembler->tab[decl->DeclarationRange.First] =
+                        (assembler->point_size ? 1 : 0) +
+                        assembler->out_colors;
                     break;
+                case TGSI_SEMANTIC_FOG:
                 case TGSI_SEMANTIC_GENERIC:
                     /* XXX multiple? */
-                    assembler->tab[decl->DeclarationRange.First] = 6;
+                    assembler->tab[decl->DeclarationRange.First] =
+                        (assembler->point_size ? 1 : 0) +
+                        assembler->out_colors +
+                        assembler->out_texcoords;
                     break;
                 case TGSI_SEMANTIC_PSIZE:
-                    assembler->tab[decl->DeclarationRange.First] = 15;
+                    assembler->tab[decl->DeclarationRange.First] = 1;
                     break;
                 default:
                     debug_printf("r300: vs: Bad semantic declaration %d\n",
@@ -65,16 +71,13 @@ static INLINE unsigned r300_vs_src_type(struct r300_vs_asm* assembler,
 {
     switch (src->File) {
         case TGSI_FILE_NULL:
+        case TGSI_FILE_INPUT:
             /* Probably a zero or one swizzle */
             return R300_PVS_SRC_REG_INPUT;
-            break;
-        case TGSI_FILE_INPUT:
-            return R300_PVS_SRC_REG_INPUT;
-            break;
         case TGSI_FILE_TEMPORARY:
             return R300_PVS_SRC_REG_TEMPORARY;
-            break;
         case TGSI_FILE_CONSTANT:
+        case TGSI_FILE_IMMEDIATE:
             return R300_PVS_SRC_REG_CONSTANT;
         default:
             debug_printf("r300: vs: Unimplemented src type %d\n", src->File);
@@ -83,16 +86,32 @@ static INLINE unsigned r300_vs_src_type(struct r300_vs_asm* assembler,
     return 0;
 }
 
+static INLINE unsigned r300_vs_src(struct r300_vs_asm* assembler,
+                                   struct tgsi_src_register* src)
+{
+    switch (src->File) {
+        case TGSI_FILE_NULL:
+        case TGSI_FILE_INPUT:
+        case TGSI_FILE_TEMPORARY:
+        case TGSI_FILE_CONSTANT:
+            return src->Index;
+        case TGSI_FILE_IMMEDIATE:
+            return src->Index + assembler->imm_offset;
+        default:
+            debug_printf("r300: vs: Unimplemented src type %d\n", src->File);
+            break;
+    }
+    return 0;
+}
+
 static INLINE unsigned r300_vs_dst_type(struct r300_vs_asm* assembler,
                                         struct tgsi_dst_register* dst)
 {
     switch (dst->File) {
         case TGSI_FILE_TEMPORARY:
             return R300_PVS_DST_REG_TEMPORARY;
-            break;
         case TGSI_FILE_OUTPUT:
             return R300_PVS_DST_REG_OUT;
-            break;
         default:
             debug_printf("r300: vs: Unimplemented dst type %d\n", dst->File);
             break;
@@ -106,10 +125,8 @@ static INLINE unsigned r300_vs_dst(struct r300_vs_asm* assembler,
     switch (dst->File) {
         case TGSI_FILE_TEMPORARY:
             return dst->Index;
-            break;
         case TGSI_FILE_OUTPUT:
             return assembler->tab[dst->Index];
-            break;
         default:
             debug_printf("r300: vs: Unimplemented dst %d\n", dst->File);
             break;
@@ -129,6 +146,12 @@ static uint32_t r300_vs_op(unsigned op)
         case TGSI_OPCODE_MOV:
         case TGSI_OPCODE_SWZ:
             return R300_VE_ADD;
+        case TGSI_OPCODE_MAX:
+            return R300_VE_MAXIMUM;
+        case TGSI_OPCODE_SLT:
+            return R300_VE_SET_LESS_THAN;
+        case TGSI_OPCODE_RSQ:
+            return R300_PVS_DST_MATH_INST | R300_ME_RECIP_DX;
         case TGSI_OPCODE_MAD:
             return R300_PVS_DST_MACRO_INST | R300_PVS_MACRO_OP_2CLK_MADD;
         default:
@@ -152,39 +175,62 @@ static uint32_t r300_vs_swiz(struct tgsi_full_src_register* reg)
     }
 }
 
+/* XXX icky icky icky icky */
+static uint32_t r300_vs_scalar_swiz(struct tgsi_full_src_register* reg)
+{
+    if (reg->SrcRegister.Extended) {
+        return reg->SrcRegisterExtSwz.ExtSwizzleX |
+            (reg->SrcRegisterExtSwz.ExtSwizzleX << 3) |
+            (reg->SrcRegisterExtSwz.ExtSwizzleX << 6) |
+            (reg->SrcRegisterExtSwz.ExtSwizzleX << 9);
+    } else {
+        return reg->SrcRegister.SwizzleX |
+            (reg->SrcRegister.SwizzleX << 3) |
+            (reg->SrcRegister.SwizzleX << 6) |
+            (reg->SrcRegister.SwizzleX << 9);
+    }
+}
+
+/* XXX scalar stupidity */
 static void r300_vs_emit_inst(struct r300_vertex_shader* vs,
                               struct r300_vs_asm* assembler,
                               struct tgsi_full_src_register* src,
                               struct tgsi_full_dst_register* dst,
                               unsigned op,
-                              unsigned count)
+                              unsigned count,
+                              boolean is_scalar)
 {
     int i = vs->instruction_count;
     vs->instructions[i].inst0 = R300_PVS_DST_OPCODE(r300_vs_op(op)) |
         R300_PVS_DST_REG_TYPE(r300_vs_dst_type(assembler, &dst->DstRegister)) |
         R300_PVS_DST_OFFSET(r300_vs_dst(assembler, &dst->DstRegister)) |
-        R300_PVS_DST_WE_XYZW;
+        R300_PVS_DST_WE(dst->DstRegister.WriteMask);
     switch (count) {
         case 3:
             vs->instructions[i].inst3 =
                 R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler,
                             &src[2].SrcRegister)) |
-                R300_PVS_SRC_OFFSET(src[2].SrcRegister.Index) |
+                R300_PVS_SRC_OFFSET(r300_vs_src(assembler,
+                            &src[2].SrcRegister)) |
                 R300_PVS_SRC_SWIZZLE(r300_vs_swiz(&src[2]));
             /* Fall through */
         case 2:
             vs->instructions[i].inst2 =
                 R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler,
                             &src[1].SrcRegister)) |
-                R300_PVS_SRC_OFFSET(src[1].SrcRegister.Index) |
+                R300_PVS_SRC_OFFSET(r300_vs_src(assembler,
+                            &src[1].SrcRegister)) |
                 R300_PVS_SRC_SWIZZLE(r300_vs_swiz(&src[1]));
             /* Fall through */
         case 1:
             vs->instructions[i].inst1 =
                 R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler,
                             &src[0].SrcRegister)) |
-                R300_PVS_SRC_OFFSET(src[0].SrcRegister.Index) |
-                R300_PVS_SRC_SWIZZLE(r300_vs_swiz(&src[0]));
+                R300_PVS_SRC_OFFSET(r300_vs_src(assembler,
+                            &src[0].SrcRegister)) |
+                /* XXX the icky, it burns */
+                R300_PVS_SRC_SWIZZLE(is_scalar ? r300_vs_scalar_swiz(&src[0])
+                        : r300_vs_swiz(&src[0]));
             break;
     }
     vs->instruction_count++;
@@ -195,11 +241,18 @@ static void r300_vs_instruction(struct r300_vertex_shader* vs,
                                 struct tgsi_full_instruction* inst)
 {
     switch (inst->Instruction.Opcode) {
-        case TGSI_OPCODE_ADD:
-        case TGSI_OPCODE_MUL:
+        case TGSI_OPCODE_RSQ:
             r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode,
-                    2);
+                    1, TRUE);
+            break;
+        case TGSI_OPCODE_ADD:
+        case TGSI_OPCODE_MUL:
+        case TGSI_OPCODE_MAX:
+        case TGSI_OPCODE_SLT:
+            r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters,
+                    &inst->FullDstRegisters[0], inst->Instruction.Opcode,
+                    2, FALSE);
             break;
         case TGSI_OPCODE_DP3:
             /* Set alpha swizzle to zero for src0 and src1 */
@@ -229,19 +282,19 @@ static void r300_vs_instruction(struct r300_vertex_shader* vs,
         case TGSI_OPCODE_DP4:
             r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode,
-                    2);
+                    2, FALSE);
             break;
         case TGSI_OPCODE_MOV:
         case TGSI_OPCODE_SWZ:
             inst->FullSrcRegisters[1] = r300_constant_zero;
             r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode,
-                    2);
+                    2, FALSE);
             break;
         case TGSI_OPCODE_MAD:
             r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode,
-                    3);
+                    3, FALSE);
             break;
         case TGSI_OPCODE_END:
             break;
@@ -252,6 +305,28 @@ static void r300_vs_instruction(struct r300_vertex_shader* vs,
     }
 }
 
+static void r300_vs_init(struct r300_vertex_shader* vs,
+                         struct r300_vs_asm* assembler)
+{
+    struct tgsi_shader_info* info = &vs->info;
+    int i;
+
+    for (i = 0; i < info->num_outputs; i++) {
+        switch (info->output_semantic_name[i]) {
+            case TGSI_SEMANTIC_PSIZE:
+                assembler->point_size = TRUE;
+                break;
+            case TGSI_SEMANTIC_COLOR:
+                assembler->out_colors++;
+                break;
+            case TGSI_SEMANTIC_FOG:
+            case TGSI_SEMANTIC_GENERIC:
+                assembler->out_texcoords++;
+                break;
+        }
+    }
+}
+
 void r300_translate_vertex_shader(struct r300_context* r300,
                                   struct r300_vertex_shader* vs)
 {
@@ -264,6 +339,10 @@ void r300_translate_vertex_shader(struct r300_context* r300,
     if (assembler == NULL) {
         return;
     }
+
+    /* Init assembler. */
+    r300_vs_init(vs, assembler);
+
     /* Setup starting offset for immediates. */
     assembler->imm_offset = consts->user_count;
 
diff --git a/src/gallium/drivers/r300/r300_state_tcl.h b/src/gallium/drivers/r300/r300_state_tcl.h
index de944028baa..d5d425e9d6c 100644
--- a/src/gallium/drivers/r300/r300_state_tcl.h
+++ b/src/gallium/drivers/r300/r300_state_tcl.h
@@ -35,6 +35,10 @@
 #   define R300_VE_DOT_PRODUCT            1
 #   define R300_VE_MULTIPLY               2
 #   define R300_VE_ADD                    3
+#   define R300_VE_MAXIMUM                7
+#   define R300_VE_SET_LESS_THAN          10
+#define R300_PVS_DST_MATH_INST     (1 << 6)
+#   define R300_ME_RECIP_DX               6
 #define R300_PVS_DST_MACRO_INST    (1 << 7)
 #   define R300_PVS_MACRO_OP_2CLK_MADD    0
 #define R300_PVS_DST_REG_TYPE(x) ((x) << 8)
@@ -99,7 +103,13 @@ struct r300_vs_asm {
     unsigned imm_offset;
     /* Number of immediate constants. */
     unsigned imm_count;
-    /* Offsets into vertex output memory. */
+    /* Number of colors to write. */
+    unsigned out_colors;
+    /* Number of texcoords to write. */
+    unsigned out_texcoords;
+    /* Whether to emit point size. */
+    boolean point_size;
+    /* Tab of declared outputs to OVM outputs. */
     unsigned tab[16];
 };
 
@@ -115,7 +125,7 @@ static struct r300_vertex_shader r300_passthrough_vertex_shader = {
     .instructions[0].inst3 = 0x0,
     .instructions[1].inst0 = R300_PVS_DST_OPCODE(R300_VE_ADD) |
         R300_PVS_DST_REG_TYPE(R300_PVS_DST_REG_OUT) |
-        R300_PVS_DST_OFFSET(2) | R300_PVS_DST_WE_XYZW,
+        R300_PVS_DST_OFFSET(1) | R300_PVS_DST_WE_XYZW,
     .instructions[1].inst1 = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) |
         R300_PVS_SRC_OFFSET(1) | R300_PVS_SRC_SWIZZLE_XYZW,
     .instructions[1].inst2 = R300_PVS_SRC_SWIZZLE_ZERO,
@@ -134,7 +144,7 @@ static struct r300_vertex_shader r300_texture_vertex_shader = {
     .instructions[0].inst3 = 0x0,
     .instructions[1].inst0 = R300_PVS_DST_OPCODE(R300_VE_ADD) |
         R300_PVS_DST_REG_TYPE(R300_PVS_DST_REG_OUT) |
-        R300_PVS_DST_OFFSET(6) | R300_PVS_DST_WE_XYZW,
+        R300_PVS_DST_OFFSET(1) | R300_PVS_DST_WE_XYZW,
     .instructions[1].inst1 = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) |
         R300_PVS_SRC_OFFSET(1) | R300_PVS_SRC_SWIZZLE_XYZW,
     .instructions[1].inst2 = R300_PVS_SRC_SWIZZLE_ZERO,
diff --git a/src/gallium/drivers/r300/r300_surface.c b/src/gallium/drivers/r300/r300_surface.c
index 79bed032538..4dd5b8af99a 100644
--- a/src/gallium/drivers/r300/r300_surface.c
+++ b/src/gallium/drivers/r300/r300_surface.c
@@ -34,6 +34,13 @@ static void r300_surface_setup(struct pipe_context* pipe,
     unsigned pixpitch = tex->stride / tex->tex.block.size;
     CS_LOCALS(r300);
 
+    /* Make sure our target BO is okay. */
+    r300->winsys->add_buffer(r300->winsys, tex->buffer,
+            0, RADEON_GEM_DOMAIN_VRAM);
+    if (r300->winsys->validate(r300->winsys)) {
+        r300->context.flush(&r300->context, 0, NULL);
+    }
+
     r300_emit_blend_state(r300, &blend_clear_state);
     r300_emit_blend_color_state(r300, &blend_color_clear_state);
     r300_emit_dsa_state(r300, &dsa_clear_state);
diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c
index fe91f4e1844..5ea9f56247b 100644
--- a/src/gallium/drivers/r300/r300_texture.c
+++ b/src/gallium/drivers/r300/r300_texture.c
@@ -86,8 +86,6 @@ static struct pipe_texture*
     r300_texture_create(struct pipe_screen* screen,
                         const struct pipe_texture* template)
 {
-    /* XXX struct r300_screen* r300screen = r300_screen(screen); */
-
     struct r300_texture* tex = CALLOC_STRUCT(r300_texture);
 
     if (!tex) {
diff --git a/src/gallium/drivers/r300/r300_winsys.h b/src/gallium/drivers/r300/r300_winsys.h
index baa95282c33..a833bb0399a 100644
--- a/src/gallium/drivers/r300/r300_winsys.h
+++ b/src/gallium/drivers/r300/r300_winsys.h
@@ -35,8 +35,6 @@ extern "C" {
 #include "pipe/p_state.h"
 #include "pipe/internal/p_winsys_screen.h"
 
-struct radeon_cs;
-
 struct r300_winsys {
     /* Parent class */
     struct pipe_winsys base;
@@ -50,39 +48,50 @@ struct r300_winsys {
     /* GB pipe count */
     uint32_t gb_pipes;
 
-    /* CS object. This is very much like Intel's batchbuffer.
-     * Fill it full of dwords and relocs and then submit.
-     * Repeat as needed. */
-    struct radeon_cs* cs;
+    /* GART size. */
+    uint32_t gart_size;
+
+    /* VRAM size. */
+    uint32_t vram_size;
+
+    /* Add a pipe_buffer to the list of buffer objects to validate. */
+    void (*add_buffer)(struct r300_winsys* winsys,
+                       struct pipe_buffer* pbuffer,
+                       uint32_t rd,
+                       uint32_t wd);
+
+    /* Revalidate all currently setup pipe_buffers.
+     * Returns TRUE if a flush is required. */
+    boolean (*validate)(struct r300_winsys* winsys);
 
     /* Check to see if there's room for commands. */
-    boolean (*check_cs)(struct radeon_cs* cs, int size);
+    boolean (*check_cs)(struct r300_winsys* winsys, int size);
 
     /* Start a command emit. */
-    void (*begin_cs)(struct radeon_cs* cs,
-           int size,
-           const char* file,
-           const char* function,
-           int line);
+    void (*begin_cs)(struct r300_winsys* winsys,
+                     int size,
+                     const char* file,
+                     const char* function,
+                     int line);
 
     /* Write a dword to the command buffer. */
-    void (*write_cs_dword)(struct radeon_cs* cs, uint32_t dword);
+    void (*write_cs_dword)(struct r300_winsys* winsys, uint32_t dword);
 
     /* Write a relocated dword to the command buffer. */
-    void (*write_cs_reloc)(struct radeon_cs* cs,
-           struct pipe_buffer* bo,
-           uint32_t rd,
-           uint32_t wd,
-           uint32_t flags);
+    void (*write_cs_reloc)(struct r300_winsys* winsys,
+                           struct pipe_buffer* bo,
+                           uint32_t rd,
+                           uint32_t wd,
+                           uint32_t flags);
 
     /* Finish a command emit. */
-    void (*end_cs)(struct radeon_cs* cs,
-           const char* file,
-           const char* function,
-           int line);
+    void (*end_cs)(struct r300_winsys* winsys,
+                   const char* file,
+                   const char* function,
+                   int line);
 
     /* Flush the CS. */
-    void (*flush_cs)(struct radeon_cs* cs);
+    void (*flush_cs)(struct r300_winsys* winsys);
 };
 
 struct pipe_context* r300_create_context(struct pipe_screen* screen,
diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c
index 11aff814791..62e8d99cfd0 100644
--- a/src/gallium/drivers/softpipe/sp_context.c
+++ b/src/gallium/drivers/softpipe/sp_context.c
@@ -126,14 +126,14 @@ softpipe_is_texture_referenced( struct pipe_context *pipe,
 				struct pipe_texture *texture,
 				unsigned face, unsigned level)
 {
-   return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE;
+   return PIPE_UNREFERENCED;
 }
 
 static unsigned int
 softpipe_is_buffer_referenced( struct pipe_context *pipe,
 			       struct pipe_buffer *buf)
 {
-   return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE;
+   return PIPE_UNREFERENCED;
 }
 
 struct pipe_context *
diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c
index c0113c47adb..9e19745889e 100644
--- a/src/gallium/drivers/softpipe/sp_texture.c
+++ b/src/gallium/drivers/softpipe/sp_texture.c
@@ -399,3 +399,22 @@ softpipe_init_screen_texture_funcs(struct pipe_screen *screen)
    screen->transfer_map = softpipe_transfer_map;
    screen->transfer_unmap = softpipe_transfer_unmap;
 }
+
+
+boolean
+softpipe_get_texture_buffer( struct pipe_texture *texture,
+                             struct pipe_buffer **buf,
+                             unsigned *stride )
+{
+   struct softpipe_texture *tex = (struct softpipe_texture *)texture;
+
+   if (!tex)
+      return FALSE;
+
+   pipe_buffer_reference(buf, tex->buffer);
+
+   if (stride)
+      *stride = tex->stride[0];
+
+   return TRUE;
+}
diff --git a/src/gallium/drivers/softpipe/sp_winsys.h b/src/gallium/drivers/softpipe/sp_winsys.h
index cf91e7782bc..9e571862b75 100644
--- a/src/gallium/drivers/softpipe/sp_winsys.h
+++ b/src/gallium/drivers/softpipe/sp_winsys.h
@@ -52,6 +52,12 @@ struct pipe_screen *
 softpipe_create_screen(struct pipe_winsys *);
 
 
+boolean
+softpipe_get_texture_buffer( struct pipe_texture *texture,
+                             struct pipe_buffer **buf,
+                             unsigned *stride );
+
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h
index 82e23c413c8..47c24f30862 100644
--- a/src/gallium/include/pipe/p_defines.h
+++ b/src/gallium/include/pipe/p_defines.h
@@ -132,6 +132,7 @@ enum pipe_texture_target {
 #define PIPE_TEX_FACE_NEG_Y 3
 #define PIPE_TEX_FACE_POS_Z 4
 #define PIPE_TEX_FACE_NEG_Z 5
+#define PIPE_TEX_FACE_MAX   6
 
 #define PIPE_TEX_WRAP_REPEAT                   0
 #define PIPE_TEX_WRAP_CLAMP                    1
@@ -158,14 +159,6 @@ enum pipe_texture_target {
 #define PIPE_TEX_COMPARE_NONE          0
 #define PIPE_TEX_COMPARE_R_TO_TEXTURE  1
 
-#define PIPE_TEX_FACE_POS_X   0
-#define PIPE_TEX_FACE_NEG_X   1
-#define PIPE_TEX_FACE_POS_Y   2
-#define PIPE_TEX_FACE_NEG_Y   3
-#define PIPE_TEX_FACE_POS_Z   4
-#define PIPE_TEX_FACE_NEG_Z   5
-#define PIPE_TEX_FACE_MAX     6
-
 #define PIPE_TEXTURE_USAGE_RENDER_TARGET   0x1
 #define PIPE_TEXTURE_USAGE_DISPLAY_TARGET  0x2 /* ie a backbuffer */
 #define PIPE_TEXTURE_USAGE_PRIMARY         0x4 /* ie a frontbuffer */
diff --git a/src/gallium/include/state_tracker/dri1_api.h b/src/gallium/include/state_tracker/dri1_api.h
new file mode 100644
index 00000000000..b173ba3683d
--- /dev/null
+++ b/src/gallium/include/state_tracker/dri1_api.h
@@ -0,0 +1,82 @@
+#ifndef _DRI1_API_H_
+#define _DRI1_API_H_
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_format.h"
+
+#include "state_tracker/drm_api.h"
+
+#include <drm.h>
+
+struct pipe_screen;
+struct pipe_winsys;
+struct pipe_buffer;
+struct pipe_context;
+struct pipe_texture;
+
+struct dri1_api_version
+{
+   int major;
+   int minor;
+   int patch_level;
+};
+
+/**
+ * This callback struct is intended for drivers that need to take
+ * the hardware lock on command submission.
+ */
+
+struct dri1_api_lock_funcs
+{
+   void (*lock) (struct pipe_context * pipe);
+   void (*unlock) (struct pipe_context * locked_pipe);
+      boolean(*is_locked) (struct pipe_context * locked_pipe);
+      boolean(*is_lock_lost) (struct pipe_context * locked_pipe);
+   void (*clear_lost_lock) (struct pipe_context * locked_pipe);
+};
+
+struct dri1_api
+{
+   /**
+    * For flushing to the front buffer. A driver should implement one and only
+    * one of the functions below. The present_locked functions allows a dri1
+    * driver to pageflip.
+    */
+
+   /*@{ */
+
+   struct pipe_surface *(*front_srf_locked) (struct pipe_context *
+					     locked_pipe);
+
+   void (*present_locked) (struct pipe_context * locked_pipe,
+			   struct pipe_surface * surf,
+			   const struct drm_clip_rect * rect,
+			   unsigned int num_clip,
+			   int x_draw, int y_draw,
+			   const struct drm_clip_rect * src_bbox,
+			   struct pipe_fence_handle ** fence);
+   /*@} */
+};
+
+struct dri1_create_screen_arg
+{
+   struct drm_create_screen_arg base;
+
+   struct dri1_api_lock_funcs *lf;
+   void *ddx_info;
+   int ddx_info_size;
+   void *sarea;
+
+   struct dri1_api_version ddx_version;
+   struct dri1_api_version dri_version;
+   struct dri1_api_version drm_version;
+
+   /*
+    * out parameters;
+    */
+
+   struct dri1_api *api;
+};
+
+#endif
diff --git a/src/gallium/include/state_tracker/drm_api.h b/src/gallium/include/state_tracker/drm_api.h
index 435435da29c..5790b2f6c7e 100644
--- a/src/gallium/include/state_tracker/drm_api.h
+++ b/src/gallium/include/state_tracker/drm_api.h
@@ -10,13 +10,30 @@ struct pipe_buffer;
 struct pipe_context;
 struct pipe_texture;
 
+enum drm_create_screen_mode {
+	DRM_CREATE_NORMAL = 0,
+	DRM_CREATE_DRI1,
+	DRM_CREATE_DRIVER = 1024,
+	DRM_CREATE_MAX
+};
+
+/**
+ * Modes other than DRM_CREATE_NORMAL derive from this struct.
+ */
+/*@{*/
+struct drm_create_screen_arg {
+	enum drm_create_screen_mode mode;
+};
+/*@}*/
+
 struct drm_api
 {
 	/**
 	 * Special buffer functions
 	 */
 	/*@{*/
-	struct pipe_screen*  (*create_screen)(int drmFB, int pciID);
+	struct pipe_screen*  (*create_screen)(int drm_fd,
+					      struct drm_create_screen_arg *arg);
 	struct pipe_context* (*create_context)(struct pipe_screen *screen);
 	/*@}*/
 
diff --git a/src/gallium/state_trackers/dri2/Makefile b/src/gallium/state_trackers/dri/Makefile
similarity index 97%
rename from src/gallium/state_trackers/dri2/Makefile
rename to src/gallium/state_trackers/dri/Makefile
index 47750e997e9..ef8f19709ab 100644
--- a/src/gallium/state_trackers/dri2/Makefile
+++ b/src/gallium/state_trackers/dri/Makefile
@@ -1,7 +1,7 @@
 TOP = ../../../..
 include $(TOP)/configs/current
 
-LIBNAME = dri2drm
+LIBNAME = dridrm
 
 LIBRARY_INCLUDES = \
 	-I$(TOP)/include \
diff --git a/src/gallium/state_trackers/dri/SConscript b/src/gallium/state_trackers/dri/SConscript
new file mode 100644
index 00000000000..ce2c2735974
--- /dev/null
+++ b/src/gallium/state_trackers/dri/SConscript
@@ -0,0 +1,23 @@
+#######################################################################
+# SConscript for dri state_tracker
+
+Import('*')
+
+if env['dri']:
+
+    env = env.Clone()
+
+    env.Append(CPPPATH = [
+        '#/src/mesa',
+	'#/src/mesa/drivers/dri/common',
+    ])
+
+    st_dri = env.ConvenienceLibrary(
+	target = 'st_dri',
+	source = [ 'dri_context.c',
+		'dri_drawable.c',
+		'dri_extensions.c',
+		'dri_screen.c',
+		]
+    )
+    Export('st_dri')
diff --git a/src/gallium/state_trackers/dri2/dri_context.c b/src/gallium/state_trackers/dri/dri_context.c
similarity index 61%
rename from src/gallium/state_trackers/dri2/dri_context.c
rename to src/gallium/state_trackers/dri/dri_context.c
index 92c26ac70fe..45eaec4ed39 100644
--- a/src/gallium/state_trackers/dri2/dri_context.c
+++ b/src/gallium/state_trackers/dri/dri_context.c
@@ -32,9 +32,8 @@
 #include "dri_screen.h"
 
 #include "dri_drawable.h"
-
-
 #include "state_tracker/drm_api.h"
+#include "state_tracker/dri1_api.h"
 #include "state_tracker/st_public.h"
 #include "state_tracker/st_context.h"
 #include "pipe/p_context.h"
@@ -43,11 +42,9 @@
 
 #include "util/u_memory.h"
 
-
 GLboolean
-dri_create_context(const __GLcontextModes *visual,
-                   __DRIcontextPrivate *cPriv,
-                   void *sharedContextPrivate)
+dri_create_context(const __GLcontextModes * visual,
+		   __DRIcontextPrivate * cPriv, void *sharedContextPrivate)
 {
    __DRIscreenPrivate *sPriv = cPriv->driScreenPriv;
    struct dri_screen *screen = dri_screen(sPriv);
@@ -55,7 +52,7 @@ dri_create_context(const __GLcontextModes *visual,
    struct st_context *st_share = NULL;
 
    if (sharedContextPrivate) {
-      st_share = ((struct dri_context *) sharedContextPrivate)->st;
+      st_share = ((struct dri_context *)sharedContextPrivate)->st;
    }
 
    ctx = CALLOC_STRUCT(dri_context);
@@ -65,11 +62,12 @@ dri_create_context(const __GLcontextModes *visual,
    cPriv->driverPrivate = ctx;
    ctx->cPriv = cPriv;
    ctx->sPriv = sPriv;
+   ctx->lock = screen->drmLock;
+   ctx->d_stamp = -1;
+   ctx->r_stamp = -1;
 
    driParseConfigFiles(&ctx->optionCache,
-                       &screen->optionCache,
-                       sPriv->myNum,
-                       "dri");
+		       &screen->optionCache, sPriv->myNum, "dri");
 
    ctx->pipe = drm_api_hooks.create_context(screen->pipe_screen);
 
@@ -87,7 +85,7 @@ dri_create_context(const __GLcontextModes *visual,
 
    return GL_TRUE;
 
-fail:
+ fail:
    if (ctx && ctx->st)
       st_destroy_context(ctx->st);
 
@@ -98,9 +96,8 @@ fail:
    return FALSE;
 }
 
-
 void
-dri_destroy_context(__DRIcontextPrivate *cPriv)
+dri_destroy_context(__DRIcontextPrivate * cPriv)
 {
    struct dri_context *ctx = dri_context(cPriv);
    struct dri_screen *screen = dri_screen(cPriv->driScreenPriv);
@@ -122,44 +119,64 @@ dri_destroy_context(__DRIcontextPrivate *cPriv)
    FREE(ctx);
 }
 
-
 GLboolean
-dri_unbind_context(__DRIcontextPrivate *cPriv)
+dri_unbind_context(__DRIcontextPrivate * cPriv)
 {
-   struct dri_context *ctx = dri_context(cPriv);
-   st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
-   /* XXX make_current(NULL)? */
+   if (cPriv) {
+      struct dri_context *ctx = dri_context(cPriv);
+
+      if (--ctx->bind_count == 0) {
+	 if (ctx->st && ctx->st == st_get_current()) {
+	    st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
+	    st_make_current(NULL, NULL, NULL);
+	 }
+      }
+   }
+
    return GL_TRUE;
 }
 
-
 GLboolean
-dri_make_current(__DRIcontextPrivate *cPriv,
-                 __DRIdrawablePrivate *driDrawPriv,
-                 __DRIdrawablePrivate *driReadPriv)
+dri_make_current(__DRIcontextPrivate * cPriv,
+		 __DRIdrawablePrivate * driDrawPriv,
+		 __DRIdrawablePrivate * driReadPriv)
 {
    if (cPriv) {
       struct dri_context *ctx = dri_context(cPriv);
       struct dri_screen *screen = dri_screen(cPriv->driScreenPriv);
       struct dri_drawable *draw = dri_drawable(driDrawPriv);
       struct dri_drawable *read = dri_drawable(driReadPriv);
+      struct st_context *old_st = st_get_current();
+
+      if (old_st && old_st != ctx->st)
+	 st_flush(old_st, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+      ++ctx->bind_count;
 
       /* This is for situations in which we need a rendering context but
        * there may not be any currently bound.
        */
       screen->dummyContext = ctx;
 
-      st_make_current(ctx->st,
-                      draw->stfb,
-                      read->stfb);
+      if (ctx->dPriv != driDrawPriv) {
+	 ctx->dPriv = driDrawPriv;
+	 ctx->d_stamp = driDrawPriv->lastStamp - 1;
+      }
+      if (ctx->rPriv != driReadPriv) {
+	 ctx->rPriv = driReadPriv;
+	 ctx->r_stamp = driReadPriv->lastStamp - 1;
+      }
 
-      /* used in dri_flush_frontbuffer */
-      ctx->dPriv = driDrawPriv;
+      st_make_current(ctx->st, draw->stfb, read->stfb);
 
-      if (driDrawPriv)
-         dri_get_buffers(driDrawPriv);
-      if (driDrawPriv != driReadPriv && driReadPriv)
-         dri_get_buffers(driReadPriv);
+      if (__dri1_api_hooks) {
+	 dri1_update_drawables(ctx, draw, read);
+      } else {
+	 if (driDrawPriv)
+	    dri_get_buffers(driDrawPriv);
+	 if (driDrawPriv != driReadPriv && driReadPriv)
+	    dri_get_buffers(driReadPriv);
+      }
    } else {
       st_make_current(NULL, NULL, NULL);
    }
@@ -167,4 +184,42 @@ dri_make_current(__DRIcontextPrivate *cPriv,
    return GL_TRUE;
 }
 
+static void
+st_dri_lock(struct pipe_context *pipe)
+{
+   dri_lock((struct dri_context *)pipe->priv);
+}
+
+static void
+st_dri_unlock(struct pipe_context *pipe)
+{
+   dri_unlock((struct dri_context *)pipe->priv);
+}
+
+static boolean
+st_dri_is_locked(struct pipe_context *pipe)
+{
+   return ((struct dri_context *)pipe->priv)->isLocked;
+}
+
+static boolean
+st_dri_lost_lock(struct pipe_context *pipe)
+{
+   return ((struct dri_context *)pipe->priv)->wsLostLock;
+}
+
+static void
+st_dri_clear_lost_lock(struct pipe_context *pipe)
+{
+   ((struct dri_context *)pipe->priv)->wsLostLock = FALSE;
+}
+
+struct dri1_api_lock_funcs dri1_lf = {
+   .lock = st_dri_lock,
+   .unlock = st_dri_unlock,
+   .is_locked = st_dri_is_locked,
+   .is_lock_lost = st_dri_lost_lock,
+   .clear_lost_lock = st_dri_clear_lost_lock
+};
+
 /* vim: set sw=3 ts=8 sts=3 expandtab: */
diff --git a/src/gallium/state_trackers/dri2/dri_context.h b/src/gallium/state_trackers/dri/dri_context.h
similarity index 65%
rename from src/gallium/state_trackers/dri2/dri_context.h
rename to src/gallium/state_trackers/dri/dri_context.h
index e910472700c..46501787340 100644
--- a/src/gallium/state_trackers/dri2/dri_context.h
+++ b/src/gallium/state_trackers/dri/dri_context.h
@@ -36,60 +36,87 @@
 #include "drm.h"
 #include "dri_util.h"
 
-
 struct pipe_context;
 struct pipe_fence;
 struct st_context;
 struct dri_drawable;
 
-
 struct dri_context
 {
    /* dri */
    __DRIscreenPrivate *sPriv;
    __DRIcontextPrivate *cPriv;
    __DRIdrawablePrivate *dPriv;
+   __DRIdrawablePrivate *rPriv;
 
    driOptionCache optionCache;
 
+   unsigned int d_stamp;
+   unsigned int r_stamp;
+
+   drmLock *lock;
+   boolean isLocked;
+   boolean stLostLock;
+   boolean wsLostLock;
+
+   unsigned int bind_count;
+
    /* gallium */
    struct st_context *st;
    struct pipe_context *pipe;
 };
 
-
 static INLINE struct dri_context *
-dri_context(__DRIcontextPrivate *driContextPriv)
+dri_context(__DRIcontextPrivate * driContextPriv)
 {
-   return (struct dri_context *) driContextPriv->driverPrivate;
+   return (struct dri_context *)driContextPriv->driverPrivate;
 }
 
+static INLINE void
+dri_lock(struct dri_context *ctx)
+{
+   drm_context_t hw_context = ctx->cPriv->hHWContext;
+   char ret = 0;
+
+   DRM_CAS(ctx->lock, hw_context, DRM_LOCK_HELD | hw_context, ret);
+   if (ret) {
+      drmGetLock(ctx->sPriv->fd, hw_context, 0);
+      ctx->stLostLock = TRUE;
+      ctx->wsLostLock = TRUE;
+   }
+   ctx->isLocked = TRUE;
+}
+
+static INLINE void
+dri_unlock(struct dri_context *ctx)
+{
+   ctx->isLocked = FALSE;
+   DRM_UNLOCK(ctx->sPriv->fd, ctx->lock, ctx->cPriv->hHWContext);
+}
 
 /***********************************************************************
  * dri_context.c
  */
-void
-dri_destroy_context(__DRIcontextPrivate * driContextPriv);
+extern struct dri1_api_lock_funcs dri1_lf;
 
-boolean
-dri_unbind_context(__DRIcontextPrivate * driContextPriv);
+void dri_destroy_context(__DRIcontextPrivate * driContextPriv);
+
+boolean dri_unbind_context(__DRIcontextPrivate * driContextPriv);
 
 boolean
 dri_make_current(__DRIcontextPrivate * driContextPriv,
-                 __DRIdrawablePrivate * driDrawPriv,
-                 __DRIdrawablePrivate * driReadPriv);
+		 __DRIdrawablePrivate * driDrawPriv,
+		 __DRIdrawablePrivate * driReadPriv);
 
 boolean
 dri_create_context(const __GLcontextModes * visual,
-                   __DRIcontextPrivate * driContextPriv,
-                   void *sharedContextPrivate);
-
+		   __DRIcontextPrivate * driContextPriv,
+		   void *sharedContextPrivate);
 
 /***********************************************************************
  * dri_extensions.c
  */
-void
-dri_init_extensions(struct dri_context *ctx);
+void dri_init_extensions(struct dri_context *ctx);
 
 #endif
 
diff --git a/src/gallium/state_trackers/dri/dri_drawable.c b/src/gallium/state_trackers/dri/dri_drawable.c
new file mode 100644
index 00000000000..15a2088df51
--- /dev/null
+++ b/src/gallium/state_trackers/dri/dri_drawable.c
@@ -0,0 +1,595 @@
+/**************************************************************************
+ *
+ * Copyright 2009, VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+/*
+ * Author: Keith Whitwell <keithw@vmware.com>
+ * Author: Jakob Bornecrantz <wallbraker@gmail.com>
+ */
+
+#include "dri_screen.h"
+#include "dri_context.h"
+#include "dri_drawable.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_inlines.h"
+#include "state_tracker/drm_api.h"
+#include "state_tracker/dri1_api.h"
+#include "state_tracker/st_public.h"
+#include "state_tracker/st_context.h"
+#include "state_tracker/st_cb_fbo.h"
+
+#include "util/u_memory.h"
+
+static void
+dri_copy_to_front(__DRIdrawablePrivate * dPriv,
+		  struct pipe_surface *from,
+		  int x, int y, unsigned w, unsigned h)
+{
+   /* TODO send a message to the Xserver to copy to the real front buffer */
+}
+
+static struct pipe_surface *
+dri_surface_from_handle(struct pipe_screen *screen,
+			unsigned handle,
+			enum pipe_format format,
+			unsigned width, unsigned height, unsigned pitch)
+{
+   struct pipe_surface *surface = NULL;
+   struct pipe_texture *texture = NULL;
+   struct pipe_texture templat;
+   struct pipe_buffer *buf = NULL;
+
+   buf = drm_api_hooks.buffer_from_handle(screen, "dri2 buffer", handle);
+   if (!buf)
+      return NULL;
+
+   memset(&templat, 0, sizeof(templat));
+   templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET;
+   templat.target = PIPE_TEXTURE_2D;
+   templat.last_level = 0;
+   templat.depth[0] = 1;
+   templat.format = format;
+   templat.width[0] = width;
+   templat.height[0] = height;
+   pf_get_block(templat.format, &templat.block);
+
+   texture = screen->texture_blanket(screen, &templat, &pitch, buf);
+
+   /* we don't need the buffer from this point on */
+   pipe_buffer_reference(&buf, NULL);
+
+   if (!texture)
+      return NULL;
+
+   surface = screen->get_tex_surface(screen, texture, 0, 0, 0,
+				     PIPE_BUFFER_USAGE_GPU_READ |
+				     PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   /* we don't need the texture from this point on */
+   pipe_texture_reference(&texture, NULL);
+   return surface;
+}
+
+/**
+ * This will be called a drawable is known to have been resized.
+ */
+void
+dri_get_buffers(__DRIdrawablePrivate * dPriv)
+{
+   struct dri_drawable *drawable = dri_drawable(dPriv);
+   struct pipe_surface *surface = NULL;
+   struct pipe_screen *screen = dri_screen(drawable->sPriv)->pipe_screen;
+   __DRIbuffer *buffers = NULL;
+   __DRIscreen *dri_screen = drawable->sPriv;
+   __DRIdrawable *dri_drawable = drawable->dPriv;
+   boolean have_depth = FALSE;
+   int i, count;
+
+   buffers = (*dri_screen->dri2.loader->getBuffers) (dri_drawable,
+						     &dri_drawable->w,
+						     &dri_drawable->h,
+						     drawable->attachments,
+						     drawable->
+						     num_attachments, &count,
+						     dri_drawable->
+						     loaderPrivate);
+
+   if (buffers == NULL) {
+      return;
+   }
+
+   /* set one cliprect to cover the whole dri_drawable */
+   dri_drawable->x = 0;
+   dri_drawable->y = 0;
+   dri_drawable->backX = 0;
+   dri_drawable->backY = 0;
+   dri_drawable->numClipRects = 1;
+   dri_drawable->pClipRects[0].x1 = 0;
+   dri_drawable->pClipRects[0].y1 = 0;
+   dri_drawable->pClipRects[0].x2 = dri_drawable->w;
+   dri_drawable->pClipRects[0].y2 = dri_drawable->h;
+   dri_drawable->numBackClipRects = 1;
+   dri_drawable->pBackClipRects[0].x1 = 0;
+   dri_drawable->pBackClipRects[0].y1 = 0;
+   dri_drawable->pBackClipRects[0].x2 = dri_drawable->w;
+   dri_drawable->pBackClipRects[0].y2 = dri_drawable->h;
+
+   for (i = 0; i < count; i++) {
+      enum pipe_format format = 0;
+      int index = 0;
+
+      switch (buffers[i].attachment) {
+      case __DRI_BUFFER_FRONT_LEFT:
+	 index = ST_SURFACE_FRONT_LEFT;
+	 format = PIPE_FORMAT_A8R8G8B8_UNORM;
+	 break;
+      case __DRI_BUFFER_FAKE_FRONT_LEFT:
+	 index = ST_SURFACE_FRONT_LEFT;
+	 format = PIPE_FORMAT_A8R8G8B8_UNORM;
+	 break;
+      case __DRI_BUFFER_BACK_LEFT:
+	 index = ST_SURFACE_BACK_LEFT;
+	 format = PIPE_FORMAT_A8R8G8B8_UNORM;
+	 break;
+      case __DRI_BUFFER_DEPTH:
+	 index = ST_SURFACE_DEPTH;
+	 format = PIPE_FORMAT_Z24S8_UNORM;
+	 break;
+      case __DRI_BUFFER_STENCIL:
+	 index = ST_SURFACE_DEPTH;
+	 format = PIPE_FORMAT_Z24S8_UNORM;
+	 break;
+      case __DRI_BUFFER_ACCUM:
+      default:
+	 assert(0);
+      }
+      assert(buffers[i].cpp == 4);
+
+      if (index == ST_SURFACE_DEPTH) {
+	 if (have_depth)
+	    continue;
+	 else
+	    have_depth = TRUE;
+      }
+
+      surface = dri_surface_from_handle(screen,
+					buffers[i].name,
+					format,
+					dri_drawable->w,
+					dri_drawable->h, buffers[i].pitch);
+
+      st_set_framebuffer_surface(drawable->stfb, index, surface);
+      pipe_surface_reference(&surface, NULL);
+   }
+   /* this needed, or else the state tracker fails to pick the new buffers */
+   st_resize_framebuffer(drawable->stfb, dri_drawable->w, dri_drawable->h);
+}
+
+void
+dri_flush_frontbuffer(struct pipe_screen *screen,
+		      struct pipe_surface *surf, void *context_private)
+{
+   struct dri_context *ctx = (struct dri_context *)context_private;
+
+   dri_copy_to_front(ctx->dPriv, surf, 0, 0, surf->width, surf->height);
+}
+
+/**
+ * This is called when we need to set up GL rendering to a new X window.
+ */
+boolean
+dri_create_buffer(__DRIscreenPrivate * sPriv,
+		  __DRIdrawablePrivate * dPriv,
+		  const __GLcontextModes * visual, boolean isPixmap)
+{
+   enum pipe_format colorFormat, depthFormat, stencilFormat;
+   struct dri_screen *screen = sPriv->private;
+   struct dri_drawable *drawable = NULL;
+   struct pipe_screen *pscreen = screen->pipe_screen;
+   int i;
+
+   if (isPixmap)
+      goto fail;		       /* not implemented */
+
+   drawable = CALLOC_STRUCT(dri_drawable);
+   if (drawable == NULL)
+      goto fail;
+
+   /* XXX: todo: use the pipe_screen queries to figure out which
+    * render targets are supportable.
+    */
+   assert(visual->redBits == 8);
+   assert(visual->depthBits == 24 || visual->depthBits == 0);
+   assert(visual->stencilBits == 8 || visual->stencilBits == 0);
+
+   colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM;
+
+   if (visual->depthBits) {
+      if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM,
+				       PIPE_TEXTURE_2D,
+				       PIPE_TEXTURE_USAGE_RENDER_TARGET |
+				       PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0))
+	 depthFormat = PIPE_FORMAT_Z24S8_UNORM;
+      else
+	 depthFormat = PIPE_FORMAT_S8Z24_UNORM;
+   } else
+      depthFormat = PIPE_FORMAT_NONE;
+
+   if (visual->stencilBits) {
+      if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM,
+				       PIPE_TEXTURE_2D,
+				       PIPE_TEXTURE_USAGE_RENDER_TARGET |
+				       PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0))
+	 stencilFormat = PIPE_FORMAT_Z24S8_UNORM;
+      else
+	 stencilFormat = PIPE_FORMAT_S8Z24_UNORM;
+   } else
+      stencilFormat = PIPE_FORMAT_NONE;
+
+   drawable->stfb = st_create_framebuffer(visual,
+					  colorFormat,
+					  depthFormat,
+					  stencilFormat,
+					  dPriv->w,
+					  dPriv->h, (void *)drawable);
+   if (drawable->stfb == NULL)
+      goto fail;
+
+   drawable->sPriv = sPriv;
+   drawable->dPriv = dPriv;
+   dPriv->driverPrivate = (void *)drawable;
+
+   /* setup dri2 buffers information */
+   i = 0;
+   drawable->attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
+#if 0
+   /* TODO incase of double buffer visual, delay fake creation */
+   drawable->attachments[i++] = __DRI_BUFFER_FAKE_FRONT_LEFT;
+#endif
+   if (visual->doubleBufferMode)
+      drawable->attachments[i++] = __DRI_BUFFER_BACK_LEFT;
+   if (visual->depthBits)
+      drawable->attachments[i++] = __DRI_BUFFER_DEPTH;
+   if (visual->stencilBits)
+      drawable->attachments[i++] = __DRI_BUFFER_STENCIL;
+   drawable->num_attachments = i;
+
+   drawable->desired_fences = 2;
+
+   return GL_TRUE;
+ fail:
+   FREE(drawable);
+   return GL_FALSE;
+}
+
+static struct pipe_fence_handle *
+dri_swap_fences_pop_front(struct dri_drawable *draw)
+{
+   struct pipe_screen *screen = dri_screen(draw->sPriv)->pipe_screen;
+   struct pipe_fence_handle *fence = NULL;
+
+   if (draw->cur_fences >= draw->desired_fences) {
+      screen->fence_reference(screen, &fence, draw->swap_fences[draw->tail]);
+      screen->fence_reference(screen, &draw->swap_fences[draw->tail++], NULL);
+      --draw->cur_fences;
+      draw->tail &= DRI_SWAP_FENCES_MASK;
+   }
+   return fence;
+}
+
+static void
+dri_swap_fences_push_back(struct dri_drawable *draw,
+			  struct pipe_fence_handle *fence)
+{
+   struct pipe_screen *screen = dri_screen(draw->sPriv)->pipe_screen;
+
+   if (!fence)
+      return;
+
+   if (draw->cur_fences < DRI_SWAP_FENCES_MAX) {
+      draw->cur_fences++;
+      screen->fence_reference(screen, &draw->swap_fences[draw->head++],
+			      fence);
+      draw->head &= DRI_SWAP_FENCES_MASK;
+   }
+}
+
+void
+dri_destroy_buffer(__DRIdrawablePrivate * dPriv)
+{
+   struct dri_drawable *drawable = dri_drawable(dPriv);
+   struct pipe_fence_handle *fence;
+   struct pipe_screen *screen = dri_screen(drawable->sPriv)->pipe_screen;
+
+   st_unreference_framebuffer(drawable->stfb);
+   drawable->desired_fences = 0;
+   while (drawable->cur_fences) {
+      fence = dri_swap_fences_pop_front(drawable);
+      screen->fence_reference(screen, &fence, NULL);
+   }
+
+   FREE(drawable);
+}
+
+static void
+dri1_update_drawables_locked(struct dri_context *ctx,
+			     __DRIdrawablePrivate * driDrawPriv,
+			     __DRIdrawablePrivate * driReadPriv)
+{
+   if (ctx->stLostLock) {
+      ctx->stLostLock = FALSE;
+      if (driDrawPriv == driReadPriv)
+	 DRI_VALIDATE_DRAWABLE_INFO(ctx->sPriv, driDrawPriv);
+      else
+	 DRI_VALIDATE_TWO_DRAWABLES_INFO(ctx->sPriv, driDrawPriv,
+					 driReadPriv);
+   }
+}
+
+/**
+ * This ensures all contexts which bind to a drawable pick up the
+ * drawable change and signal new buffer state.
+ * Calling st_resize_framebuffer for each context may seem like overkill,
+ * but no new buffers will actually be allocated if the dimensions don't
+ * change.
+ */
+
+static void
+dri1_propagate_drawable_change(struct dri_context *ctx)
+{
+   __DRIdrawablePrivate *dPriv = ctx->dPriv;
+   __DRIdrawablePrivate *rPriv = ctx->rPriv;
+   boolean flushed = FALSE;
+
+   if (dPriv && ctx->d_stamp != dPriv->lastStamp) {
+
+      st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
+      flushed = TRUE;
+      ctx->d_stamp = dPriv->lastStamp;
+      st_resize_framebuffer(dri_drawable(dPriv)->stfb, dPriv->w, dPriv->h);
+
+   }
+
+   if (rPriv && dPriv != rPriv && ctx->r_stamp != rPriv->lastStamp) {
+
+      if (!flushed)
+	 st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
+      ctx->r_stamp = rPriv->lastStamp;
+      st_resize_framebuffer(dri_drawable(rPriv)->stfb, rPriv->w, rPriv->h);
+
+   } else if (rPriv && dPriv == rPriv) {
+
+      ctx->r_stamp = ctx->d_stamp;
+
+   }
+}
+
+void
+dri1_update_drawables(struct dri_context *ctx,
+		      struct dri_drawable *draw, struct dri_drawable *read)
+{
+   dri_lock(ctx);
+   dri1_update_drawables_locked(ctx, draw->dPriv, read->dPriv);
+   dri_unlock(ctx);
+
+   dri1_propagate_drawable_change(ctx);
+}
+
+static INLINE boolean
+dri1_intersect_src_bbox(struct drm_clip_rect *dst,
+			int dst_x,
+			int dst_y,
+			const struct drm_clip_rect *src,
+			const struct drm_clip_rect *bbox)
+{
+   int xy1;
+   int xy2;
+
+   xy1 = ((int)src->x1 > (int)bbox->x1 + dst_x) ? src->x1 :
+      (int)bbox->x1 + dst_x;
+   xy2 = ((int)src->x2 < (int)bbox->x2 + dst_x) ? src->x2 :
+      (int)bbox->x2 + dst_x;
+   if (xy1 >= xy2 || xy1 < 0)
+      return FALSE;
+
+   dst->x1 = xy1;
+   dst->x2 = xy2;
+
+   xy1 = ((int)src->y1 > (int)bbox->y1 + dst_y) ? src->y1 :
+      (int)bbox->y1 + dst_y;
+   xy2 = ((int)src->y2 < (int)bbox->y2 + dst_y) ? src->y2 :
+      (int)bbox->y2 + dst_y;
+   if (xy1 >= xy2 || xy1 < 0)
+      return FALSE;
+
+   dst->y1 = xy1;
+   dst->y2 = xy2;
+   return TRUE;
+}
+
+static void
+dri1_swap_copy(struct dri_context *ctx,
+	       struct pipe_surface *dst,
+	       struct pipe_surface *src,
+	       __DRIdrawablePrivate * dPriv, const struct drm_clip_rect *bbox)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct drm_clip_rect clip;
+   struct drm_clip_rect *cur;
+   int i;
+
+   cur = dPriv->pClipRects;
+
+   for (i = 0; i < dPriv->numClipRects; ++i) {
+      if (dri1_intersect_src_bbox(&clip, dPriv->x, dPriv->y, cur++, bbox))
+	 pipe->surface_copy(pipe, dst, clip.x1, clip.y1,
+			    src,
+			    (int)clip.x1 - dPriv->x,
+			    (int)clip.y1 - dPriv->y,
+			    clip.x2 - clip.x1, clip.y2 - clip.y1);
+   }
+}
+
+static void
+dri1_copy_to_front(struct dri_context *ctx,
+		   struct pipe_surface *surf,
+		   __DRIdrawablePrivate * dPriv,
+		   const struct drm_clip_rect *sub_box,
+		   struct pipe_fence_handle **fence)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   boolean save_lost_lock;
+   uint cur_w;
+   uint cur_h;
+   struct drm_clip_rect bbox;
+   boolean visible = TRUE;
+
+   *fence = NULL;
+
+   dri_lock(ctx);
+   save_lost_lock = ctx->stLostLock;
+   dri1_update_drawables_locked(ctx, dPriv, dPriv);
+   st_get_framebuffer_dimensions(dri_drawable(dPriv)->stfb, &cur_w, &cur_h);
+
+   bbox.x1 = 0;
+   bbox.x2 = cur_w;
+   bbox.y1 = 0;
+   bbox.y2 = cur_h;
+
+   if (sub_box)
+      visible = dri1_intersect_src_bbox(&bbox, 0, 0, &bbox, sub_box);
+
+   if (visible && __dri1_api_hooks->present_locked) {
+
+      __dri1_api_hooks->present_locked(pipe,
+				       surf,
+				       dPriv->pClipRects,
+				       dPriv->numClipRects,
+				       dPriv->x, dPriv->y, &bbox, fence);
+
+   } else if (visible && __dri1_api_hooks->front_srf_locked) {
+
+      struct pipe_surface *front = __dri1_api_hooks->front_srf_locked(pipe);
+
+      if (front)
+	 dri1_swap_copy(ctx, front, surf, dPriv, &bbox);
+
+      st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, fence);
+   }
+
+   ctx->stLostLock = save_lost_lock;
+
+   /**
+    * FIXME: Revisit this: Update drawables on copy_sub_buffer ?
+    */
+
+   if (!sub_box)
+      dri1_update_drawables_locked(ctx, ctx->dPriv, ctx->rPriv);
+
+   dri_unlock(ctx);
+   dri1_propagate_drawable_change(ctx);
+}
+
+void
+dri1_flush_frontbuffer(struct pipe_screen *screen,
+		       struct pipe_surface *surf, void *context_private)
+{
+   struct dri_context *ctx = (struct dri_context *)context_private;
+   struct pipe_fence_handle *dummy_fence;
+
+   dri1_copy_to_front(ctx, surf, ctx->dPriv, NULL, &dummy_fence);
+   screen->fence_reference(screen, &dummy_fence, NULL);
+
+   /**
+    * FIXME: Do we need swap throttling here?
+    */
+}
+
+void
+dri_swap_buffers(__DRIdrawablePrivate * dPriv)
+{
+   struct dri_context *ctx;
+   struct pipe_surface *back_surf;
+   struct dri_drawable *draw = dri_drawable(dPriv);
+   struct pipe_screen *screen = dri_screen(draw->sPriv)->pipe_screen;
+   struct pipe_fence_handle *fence;
+   struct st_context *st = st_get_current();
+
+   assert(__dri1_api_hooks != NULL);
+
+   if (!st)
+      return;			       /* For now */
+
+   ctx = (struct dri_context *)st->pipe->priv;
+
+   st_get_framebuffer_surface(draw->stfb, ST_SURFACE_BACK_LEFT, &back_surf);
+   if (back_surf) {
+      st_notify_swapbuffers(draw->stfb);
+      st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
+      fence = dri_swap_fences_pop_front(draw);
+      if (fence) {
+	 (void)screen->fence_finish(screen, fence, 0);
+	 screen->fence_reference(screen, &fence, NULL);
+      }
+      dri1_copy_to_front(ctx, back_surf, dPriv, NULL, &fence);
+      dri_swap_fences_push_back(draw, fence);
+      screen->fence_reference(screen, &fence, NULL);
+   }
+}
+
+void
+dri_copy_sub_buffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h)
+{
+   struct pipe_screen *screen = dri_screen(dPriv->driScreenPriv)->pipe_screen;
+   struct drm_clip_rect sub_bbox;
+   struct dri_context *ctx;
+   struct pipe_surface *back_surf;
+   struct dri_drawable *draw = dri_drawable(dPriv);
+   struct pipe_fence_handle *dummy_fence;
+   struct st_context *st = st_get_current();
+
+   assert(__dri1_api_hooks != NULL);
+
+   if (!st)
+      return;
+
+   ctx = (struct dri_context *)st->pipe->priv;
+
+   sub_bbox.x1 = x;
+   sub_bbox.x2 = x + w;
+   sub_bbox.y1 = y;
+   sub_bbox.y2 = y + h;
+
+   st_get_framebuffer_surface(draw->stfb, ST_SURFACE_BACK_LEFT, &back_surf);
+   if (back_surf) {
+      st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
+      dri1_copy_to_front(ctx, back_surf, dPriv, &sub_bbox, &dummy_fence);
+      screen->fence_reference(screen, &dummy_fence, NULL);
+   }
+}
+
+/* vim: set sw=3 ts=8 sts=3 expandtab: */
diff --git a/src/gallium/state_trackers/dri2/dri_drawable.h b/src/gallium/state_trackers/dri/dri_drawable.h
similarity index 67%
rename from src/gallium/state_trackers/dri2/dri_drawable.h
rename to src/gallium/state_trackers/dri/dri_drawable.h
index 185c657b35a..78a66624aaf 100644
--- a/src/gallium/state_trackers/dri2/dri_drawable.h
+++ b/src/gallium/state_trackers/dri/dri_drawable.h
@@ -31,9 +31,11 @@
 #include "pipe/p_compiler.h"
 
 struct pipe_surface;
-struct pipe_fence;
+struct pipe_fence_handle;
 struct st_framebuffer;
 
+#define DRI_SWAP_FENCES_MAX  8
+#define DRI_SWAP_FENCES_MASK 7
 
 struct dri_drawable
 {
@@ -46,44 +48,47 @@ struct dri_drawable
 
    /* gallium */
    struct st_framebuffer *stfb;
+   struct pipe_fence_handle *swap_fences[DRI_SWAP_FENCES_MAX];
+   unsigned int head;
+   unsigned int tail;
+   unsigned int desired_fences;
+   unsigned int cur_fences;
 };
 
-
 static INLINE struct dri_drawable *
 dri_drawable(__DRIdrawablePrivate * driDrawPriv)
 {
-   return (struct dri_drawable *) driDrawPriv->driverPrivate;
+   return (struct dri_drawable *)driDrawPriv->driverPrivate;
 }
 
-
 /***********************************************************************
  * dri_drawable.c
  */
 boolean
-dri_create_buffer(__DRIscreenPrivate *sPriv,
-                  __DRIdrawablePrivate *dPriv,
-                  const __GLcontextModes *visual,
-                  boolean isPixmap);
+dri_create_buffer(__DRIscreenPrivate * sPriv,
+		  __DRIdrawablePrivate * dPriv,
+		  const __GLcontextModes * visual, boolean isPixmap);
 
 void
 dri_flush_frontbuffer(struct pipe_screen *screen,
-                      struct pipe_surface *surf,
-                      void *context_private);
+		      struct pipe_surface *surf, void *context_private);
+
+void dri_swap_buffers(__DRIdrawablePrivate * dPriv);
 
 void
-dri_swap_buffers(__DRIdrawablePrivate * dPriv);
+dri_copy_sub_buffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h);
+
+void dri_get_buffers(__DRIdrawablePrivate * dPriv);
+
+void dri_destroy_buffer(__DRIdrawablePrivate * dPriv);
 
 void
-dri_copy_sub_buffer(__DRIdrawablePrivate * dPriv,
-                    int x, int y,
-                    int w, int h);
+dri1_update_drawables(struct dri_context *ctx,
+		      struct dri_drawable *draw, struct dri_drawable *read);
 
 void
-dri_get_buffers(__DRIdrawablePrivate * dPriv);
-
-void
-dri_destroy_buffer(__DRIdrawablePrivate *dPriv);
-
+dri1_flush_frontbuffer(struct pipe_screen *screen,
+		       struct pipe_surface *surf, void *context_private);
 #endif
 
 /* vim: set sw=3 ts=8 sts=3 expandtab: */
diff --git a/src/gallium/state_trackers/dri2/dri_extensions.c b/src/gallium/state_trackers/dri/dri_extensions.c
similarity index 97%
rename from src/gallium/state_trackers/dri2/dri_extensions.c
rename to src/gallium/state_trackers/dri/dri_extensions.c
index 732d1e89b0c..0c59d42d5c6 100644
--- a/src/gallium/state_trackers/dri2/dri_extensions.c
+++ b/src/gallium/state_trackers/dri/dri_extensions.c
@@ -52,7 +52,6 @@
 #define need_GL_NV_vertex_program
 #include "extension_helper.h"
 
-
 /**
  * Extension strings exported by the driver.
  */
@@ -74,7 +73,8 @@ const struct dri_extension card_extensions[] = {
    {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions},
    {"GL_ARB_window_pos", GL_ARB_window_pos_functions},
    {"GL_EXT_blend_color", GL_EXT_blend_color_functions},
-   {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions},
+   {"GL_EXT_blend_equation_separate",
+    GL_EXT_blend_equation_separate_functions},
    {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions},
    {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions},
    {"GL_EXT_blend_subtract", NULL},
@@ -98,11 +98,10 @@ const struct dri_extension card_extensions[] = {
    {"GL_NV_blend_square", NULL},
    {"GL_NV_vertex_program", GL_NV_vertex_program_functions},
    {"GL_NV_vertex_program1_1", NULL},
-   {"GL_SGIS_generate_mipmap", NULL },
+   {"GL_SGIS_generate_mipmap", NULL},
    {NULL, NULL}
 };
 
-
 void
 dri_init_extensions(struct dri_context *ctx)
 {
diff --git a/src/gallium/state_trackers/dri2/dri_screen.c b/src/gallium/state_trackers/dri/dri_screen.c
similarity index 53%
rename from src/gallium/state_trackers/dri2/dri_screen.c
rename to src/gallium/state_trackers/dri/dri_screen.c
index ab5878a4bce..d3392ee690b 100644
--- a/src/gallium/state_trackers/dri2/dri_screen.c
+++ b/src/gallium/state_trackers/dri/dri_screen.c
@@ -40,68 +40,38 @@
 #include "pipe/p_context.h"
 #include "pipe/p_screen.h"
 #include "pipe/p_inlines.h"
+#include "pipe/p_format.h"
 #include "state_tracker/drm_api.h"
+#include "state_tracker/dri1_api.h"
 #include "state_tracker/st_public.h"
 #include "state_tracker/st_cb_fbo.h"
 
-
 PUBLIC const char __driConfigOptions[] =
    DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE
-    DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS)
-    DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0)
+   DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS)
+   DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0)
    DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY
-    /*DRI_CONF_FORCE_S3TC_ENABLE(false)*/
-    DRI_CONF_ALLOW_LARGE_TEXTURES(1)
+    /*DRI_CONF_FORCE_S3TC_ENABLE(false) */
+   DRI_CONF_ALLOW_LARGE_TEXTURES(1)
    DRI_CONF_SECTION_END DRI_CONF_END;
 
+   const uint __driNConfigOptions = 3;
 
-const uint __driNConfigOptions = 3;
-
-
-static const __DRIextension *dri_screen_extensions[] = {
-    &driReadDrawableExtension,
-    &driCopySubBufferExtension.base,
-    &driSwapControlExtension.base,
-    &driFrameTrackingExtension.base,
-    &driMediaStreamCounterExtension.base,
-    NULL
-};
-
-
-static void
-dri_get_drm_minor(struct dri_screen *screen)
-{
-   /* TODO get the real minor */
-   screen->minor = 0;
-}
-
-
-static void
-dri_get_device_id(struct dri_screen *screen)
-{
-   char path[512];
-   FILE *file;
-
-   /*
-    * There must be a better way to get the deviceID.
-    * XXX this only works on Linux.
-    */
-   snprintf(path, sizeof(path), "/sys/class/drm/card%d/device/device", screen->minor);
-   file = fopen(path, "r");
-   if (!file) {
-      return;
-   }
-
-   fgets(path, sizeof(path), file);
-   sscanf(path, "%x", &screen->deviceID);
-   fclose(file);
-}
+   static const __DRIextension *dri_screen_extensions[] = {
+      &driReadDrawableExtension,
+      &driCopySubBufferExtension.base,
+      &driSwapControlExtension.base,
+      &driFrameTrackingExtension.base,
+      &driMediaStreamCounterExtension.base,
+      NULL
+   };
 
+struct dri1_api *__dri1_api_hooks = NULL;
 
 static const __DRIconfig **
-dri_fill_in_modes(__DRIscreenPrivate *psp,
-                  unsigned pixel_bits, unsigned depth_bits,
-                  unsigned stencil_bits, GLboolean have_back_buffer)
+dri_fill_in_modes(__DRIscreenPrivate * psp,
+		  unsigned pixel_bits, unsigned depth_bits,
+		  unsigned stencil_bits, GLboolean have_back_buffer)
 {
    __DRIconfig **configs;
    __GLcontextModes *m;
@@ -125,9 +95,9 @@ dri_fill_in_modes(__DRIscreenPrivate *psp,
    depth_bits_array[1] = 24;
    depth_bits_array[2] = 24;
 
-   stencil_bits_array[0] = 0;   /* no depth or stencil */
-   stencil_bits_array[1] = 0;   /* z24x8 */
-   stencil_bits_array[2] = 8;   /* z24s8 */
+   stencil_bits_array[0] = 0;	       /* no depth or stencil */
+   stencil_bits_array[1] = 0;	       /* z24x8 */
+   stencil_bits_array[2] = 8;	       /* z24s8 */
 
    msaa_samples_array[0] = 0;
 
@@ -135,22 +105,22 @@ dri_fill_in_modes(__DRIscreenPrivate *psp,
    back_buffer_factor = 3;
    msaa_samples_factor = 1;
 
-   num_modes = depth_buffer_factor * back_buffer_factor * msaa_samples_factor * 4;
+   num_modes =
+      depth_buffer_factor * back_buffer_factor * msaa_samples_factor * 4;
 
    if (pixel_bits == 16) {
       fb_format = GL_RGB;
       fb_type = GL_UNSIGNED_SHORT_5_6_5;
-   }
-   else {
+   } else {
       fb_format = GL_BGRA;
       fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
    }
 
    configs = driCreateConfigs(fb_format, fb_type,
 			      depth_bits_array,
-                              stencil_bits_array, depth_buffer_factor,
-                              back_buffer_modes, back_buffer_factor,
-                              msaa_samples_array, msaa_samples_factor);
+			      stencil_bits_array, depth_buffer_factor,
+			      back_buffer_modes, back_buffer_factor,
+			      msaa_samples_array, msaa_samples_factor);
    if (configs == NULL) {
       debug_printf("%s: driCreateConfigs failed\n", __FUNCTION__);
       return NULL;
@@ -159,49 +129,99 @@ dri_fill_in_modes(__DRIscreenPrivate *psp,
    for (i = 0; configs[i]; i++) {
       m = &configs[i]->modes;
       if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) {
-         m->visualRating = GLX_SLOW_CONFIG;
+	 m->visualRating = GLX_SLOW_CONFIG;
       }
    }
 
-   return (const const __DRIconfig **) configs;
+   return (const const __DRIconfig **)configs;
 }
 
-
 /**
  * Get information about previous buffer swaps.
  */
-int
-dri_get_swap_info(__DRIdrawablePrivate * dPriv,
-                  __DRIswapInfo * sInfo)
+static int
+dri_get_swap_info(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo)
 {
-   if (dPriv == NULL ||
-       dPriv->driverPrivate == NULL ||
-       sInfo == NULL)
+   if (dPriv == NULL || dPriv->driverPrivate == NULL || sInfo == NULL)
       return -1;
    else
       return 0;
 }
 
-
-/**
- * NULL stub for old dri loaders
- */
-const __DRIconfig **
-dri_init_screen(__DRIscreenPrivate *sPriv)
+static INLINE void
+dri_copy_version(struct dri1_api_version *dst,
+		 const struct __DRIversionRec *src)
 {
-   return NULL;
+   dst->major = src->major;
+   dst->minor = src->minor;
+   dst->patch_level = src->patch;
 }
 
+static const __DRIconfig **
+dri_init_screen(__DRIscreenPrivate * sPriv)
+{
+   struct dri_screen *screen;
+   const __DRIconfig **configs;
+   struct dri1_create_screen_arg arg;
+
+   dri_init_extensions(NULL);
+
+   screen = CALLOC_STRUCT(dri_screen);
+   if (!screen)
+      return NULL;
+
+   screen->sPriv = sPriv;
+   screen->fd = sPriv->fd;
+   screen->drmLock = (drmLock *) & sPriv->pSAREA->lock;
+
+   sPriv->private = (void *)screen;
+   sPriv->extensions = dri_screen_extensions;
+
+   arg.base.mode = DRM_CREATE_DRI1;
+   arg.lf = &dri1_lf;
+   arg.ddx_info = sPriv->pDevPriv;
+   arg.ddx_info_size = sPriv->devPrivSize;
+   arg.sarea = sPriv->pSAREA;
+   dri_copy_version(&arg.ddx_version, &sPriv->ddx_version);
+   dri_copy_version(&arg.dri_version, &sPriv->dri_version);
+   dri_copy_version(&arg.drm_version, &sPriv->drm_version);
+   arg.api = NULL;
+
+   screen->pipe_screen = drm_api_hooks.create_screen(screen->fd, &arg.base);
+
+   if (!screen->pipe_screen || !arg.api) {
+      debug_printf("%s: failed to create dri1 screen\n", __FUNCTION__);
+      goto out_no_screen;
+   }
+
+   __dri1_api_hooks = arg.api;
+
+   screen->pipe_screen->flush_frontbuffer = dri1_flush_frontbuffer;
+   driParseOptionInfo(&screen->optionCache,
+		      __driConfigOptions, __driNConfigOptions);
+
+   configs = dri_fill_in_modes(sPriv, sPriv->fbBPP, 24, 8, 1);
+   if (!configs)
+      goto out_no_configs;
+
+   return configs;
+ out_no_configs:
+   screen->pipe_screen->destroy(screen->pipe_screen);
+ out_no_screen:
+   FREE(screen);
+   return NULL;
+}
 
 /**
  * This is the driver specific part of the createNewScreen entry point.
  *
  * Returns the __GLcontextModes supported by this driver.
  */
-const __DRIconfig **
-dri_init_screen2(__DRIscreenPrivate *sPriv)
+static const __DRIconfig **
+dri_init_screen2(__DRIscreenPrivate * sPriv)
 {
    struct dri_screen *screen;
+   struct drm_create_screen_arg arg;
 
    /* Set up dispatch table to cope with all known extensions */
    dri_init_extensions(NULL);
@@ -212,13 +232,11 @@ dri_init_screen2(__DRIscreenPrivate *sPriv)
 
    screen->sPriv = sPriv;
    screen->fd = sPriv->fd;
-   dri_get_drm_minor(screen);
-   dri_get_device_id(screen);
-   sPriv->private = (void *) screen;
+   sPriv->private = (void *)screen;
    sPriv->extensions = dri_screen_extensions;
+   arg.mode = DRM_CREATE_NORMAL;
 
-
-   screen->pipe_screen = drm_api_hooks.create_screen(screen->fd, screen->deviceID);
+   screen->pipe_screen = drm_api_hooks.create_screen(screen->fd, &arg);
    if (!screen->pipe_screen) {
       debug_printf("%s: failed to create pipe_screen\n", __FUNCTION__);
       goto fail;
@@ -228,20 +246,14 @@ dri_init_screen2(__DRIscreenPrivate *sPriv)
    screen->pipe_screen->flush_frontbuffer = dri_flush_frontbuffer;
 
    driParseOptionInfo(&screen->optionCache,
-                      __driConfigOptions,
-                      __driNConfigOptions);
+		      __driConfigOptions, __driNConfigOptions);
 
-   return dri_fill_in_modes(sPriv,
-                            4 * 8,
-                            24,
-                            8,
-                            1);
-fail:
+   return dri_fill_in_modes(sPriv, 4 * 8, 24, 8, 1);
+ fail:
    return NULL;
 }
 
-
-void
+static void
 dri_destroy_screen(__DRIscreenPrivate * sPriv)
 {
    struct dri_screen *screen = dri_screen(sPriv);
@@ -251,22 +263,22 @@ dri_destroy_screen(__DRIscreenPrivate * sPriv)
    sPriv->private = NULL;
 }
 
-
 PUBLIC const struct __DriverAPIRec driDriverAPI = {
-   .InitScreen          = dri_init_screen, /* not supported but exported */
-   .DestroyScreen       = dri_destroy_screen,
-   .CreateContext       = dri_create_context,
-   .DestroyContext      = dri_destroy_context,
-   .CreateBuffer        = dri_create_buffer,
-   .DestroyBuffer       = dri_destroy_buffer,
-   .SwapBuffers         = dri_swap_buffers, /* not supported but exported */
-   .MakeCurrent         = dri_make_current,
-   .UnbindContext       = dri_unbind_context,
-   .GetSwapInfo         = dri_get_swap_info,
-   .GetDrawableMSC      = driDrawableGetMSC32,
-   .WaitForMSC          = driWaitForMSC32,
-   .CopySubBuffer       = dri_copy_sub_buffer, /* not supported but exported */
-   .InitScreen2         = dri_init_screen2,
+   .InitScreen = dri_init_screen,
+   .DestroyScreen = dri_destroy_screen,
+   .CreateContext = dri_create_context,
+   .DestroyContext = dri_destroy_context,
+   .CreateBuffer = dri_create_buffer,
+   .DestroyBuffer = dri_destroy_buffer,
+   .SwapBuffers = dri_swap_buffers,
+   .MakeCurrent = dri_make_current,
+   .UnbindContext = dri_unbind_context,
+   .GetSwapInfo = dri_get_swap_info,
+   .GetDrawableMSC = driDrawableGetMSC32,
+   .WaitForMSC = driWaitForMSC32,
+   .CopySubBuffer = dri_copy_sub_buffer,
+   .InitScreen = dri_init_screen,
+   .InitScreen2 = dri_init_screen2,
 };
 
 /* vim: set sw=3 ts=8 sts=3 expandtab: */
diff --git a/src/gallium/state_trackers/dri2/dri_screen.h b/src/gallium/state_trackers/dri/dri_screen.h
similarity index 87%
rename from src/gallium/state_trackers/dri2/dri_screen.h
rename to src/gallium/state_trackers/dri/dri_screen.h
index fe2676d0be3..100d9e50e07 100644
--- a/src/gallium/state_trackers/dri2/dri_screen.h
+++ b/src/gallium/state_trackers/dri/dri_screen.h
@@ -37,6 +37,8 @@
 
 #include "pipe/p_compiler.h"
 
+#include "state_tracker/dri1_api.h"
+
 struct dri_screen
 {
    /* dri */
@@ -54,36 +56,26 @@ struct dri_screen
    struct dri_context *dummyContext;
 
    /* drm */
-   int deviceID;
    int fd;
-   int minor;
+   drmLock *drmLock;
 
    /* gallium */
    struct pipe_winsys *pipe_winsys;
    struct pipe_screen *pipe_screen;
 };
 
-
 /** cast wrapper */
 static INLINE struct dri_screen *
-dri_screen(__DRIscreenPrivate *sPriv)
+dri_screen(__DRIscreenPrivate * sPriv)
 {
-   return (struct dri_screen *) sPriv->private;
+   return (struct dri_screen *)sPriv->private;
 }
 
-
 /***********************************************************************
  * dri_screen.c
  */
-const __DRIconfig **
-dri_init_screen2(__DRIscreenPrivate *sPriv);
 
-void
-dri_destroy_screen(__DRIscreenPrivate * sPriv);
-
-int
-dri_get_swap_info(__DRIdrawablePrivate * dPriv,
-                  __DRIswapInfo * sInfo);
+extern struct dri1_api *__dri1_api_hooks;
 
 #endif
 
diff --git a/src/gallium/state_trackers/dri2/dri_drawable.c b/src/gallium/state_trackers/dri2/dri_drawable.c
deleted file mode 100644
index 2e3f4099e2a..00000000000
--- a/src/gallium/state_trackers/dri2/dri_drawable.c
+++ /dev/null
@@ -1,325 +0,0 @@
-/**************************************************************************
- *
- * Copyright 2009, VMware, Inc.
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- **************************************************************************/
-/*
- * Author: Keith Whitwell <keithw@vmware.com>
- * Author: Jakob Bornecrantz <wallbraker@gmail.com>
- */
-
-#include "dri_screen.h"
-#include "dri_context.h"
-#include "dri_drawable.h"
-
-#include "pipe/p_context.h"
-#include "pipe/p_screen.h"
-#include "pipe/p_inlines.h"
-#include "state_tracker/drm_api.h"
-#include "state_tracker/st_public.h"
-#include "state_tracker/st_context.h"
-#include "state_tracker/st_cb_fbo.h"
-
-#include "util/u_memory.h"
-
-
-static void
-dri_copy_to_front(__DRIdrawablePrivate *dPriv,
-                  struct pipe_surface *from,
-                  int x, int y, unsigned w, unsigned h)
-{
-   /* TODO send a message to the Xserver to copy to the real front buffer */
-}
-
-
-static struct pipe_surface *
-dri_surface_from_handle(struct pipe_screen *screen,
-                        unsigned handle,
-                        enum pipe_format format,
-                        unsigned width,
-                        unsigned height,
-                        unsigned pitch)
-{
-   struct pipe_surface *surface = NULL;
-   struct pipe_texture *texture = NULL;
-   struct pipe_texture templat;
-   struct pipe_buffer *buf = NULL;
-
-   buf = drm_api_hooks.buffer_from_handle(screen, "dri2 buffer", handle);
-   if (!buf)
-      return NULL;
-
-   memset(&templat, 0, sizeof(templat));
-   templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET;
-   templat.target = PIPE_TEXTURE_2D;
-   templat.last_level = 0;
-   templat.depth[0] = 1;
-   templat.format = format;
-   templat.width[0] = width;
-   templat.height[0] = height;
-   pf_get_block(templat.format, &templat.block);
-
-   texture = screen->texture_blanket(screen,
-                                     &templat,
-                                     &pitch,
-                                     buf);
-
-   /* we don't need the buffer from this point on */
-   pipe_buffer_reference(&buf, NULL);
-
-   if (!texture)
-      return NULL;
-
-   surface = screen->get_tex_surface(screen, texture, 0, 0, 0,
-                                     PIPE_BUFFER_USAGE_GPU_READ |
-                                     PIPE_BUFFER_USAGE_GPU_WRITE);
-
-   /* we don't need the texture from this point on */
-   pipe_texture_reference(&texture, NULL);
-   return surface;
-}
-
-
-/**
- * This will be called a drawable is known to have been resized.
- */
-void
-dri_get_buffers(__DRIdrawablePrivate *dPriv)
-{
-   struct dri_drawable *drawable = dri_drawable(dPriv);
-   struct pipe_surface *surface = NULL;
-   struct pipe_screen *screen = dri_screen(drawable->sPriv)->pipe_screen;
-   __DRIbuffer *buffers = NULL;
-   __DRIscreen *dri_screen = drawable->sPriv;
-   __DRIdrawable *dri_drawable = drawable->dPriv;
-   boolean have_depth = FALSE;
-   int i, count;
-
-   buffers = (*dri_screen->dri2.loader->getBuffers)(dri_drawable,
-                                                    &dri_drawable->w,
-                                                    &dri_drawable->h,
-                                                    drawable->attachments,
-                                                    drawable->num_attachments,
-                                                    &count,
-                                                    dri_drawable->loaderPrivate);
-
-   if (buffers == NULL) {
-      return;
-   }
-
-   /* set one cliprect to cover the whole dri_drawable */
-   dri_drawable->x = 0;
-   dri_drawable->y = 0;
-   dri_drawable->backX = 0;
-   dri_drawable->backY = 0;
-   dri_drawable->numClipRects = 1;
-   dri_drawable->pClipRects[0].x1 = 0;
-   dri_drawable->pClipRects[0].y1 = 0;
-   dri_drawable->pClipRects[0].x2 = dri_drawable->w;
-   dri_drawable->pClipRects[0].y2 = dri_drawable->h;
-   dri_drawable->numBackClipRects = 1;
-   dri_drawable->pBackClipRects[0].x1 = 0;
-   dri_drawable->pBackClipRects[0].y1 = 0;
-   dri_drawable->pBackClipRects[0].x2 = dri_drawable->w;
-   dri_drawable->pBackClipRects[0].y2 = dri_drawable->h;
-
-   for (i = 0; i < count; i++) {
-      enum pipe_format format = 0;
-      int index = 0;
-
-      switch (buffers[i].attachment) {
-         case __DRI_BUFFER_FRONT_LEFT:
-            index = ST_SURFACE_FRONT_LEFT;
-            format = PIPE_FORMAT_A8R8G8B8_UNORM;
-            break;
-         case __DRI_BUFFER_FAKE_FRONT_LEFT:
-            index = ST_SURFACE_FRONT_LEFT;
-            format = PIPE_FORMAT_A8R8G8B8_UNORM;
-            break;
-         case __DRI_BUFFER_BACK_LEFT:
-            index = ST_SURFACE_BACK_LEFT;
-            format = PIPE_FORMAT_A8R8G8B8_UNORM;
-            break;
-         case __DRI_BUFFER_DEPTH:
-            index = ST_SURFACE_DEPTH;
-            format = PIPE_FORMAT_Z24S8_UNORM;
-            break;
-         case __DRI_BUFFER_STENCIL:
-            index = ST_SURFACE_DEPTH;
-            format = PIPE_FORMAT_Z24S8_UNORM;
-            break;
-         case __DRI_BUFFER_ACCUM:
-         default:
-            assert(0);
-      }
-      assert(buffers[i].cpp == 4);
-
-      if (index == ST_SURFACE_DEPTH) {
-         if (have_depth)
-            continue;
-         else
-            have_depth = TRUE;
-      }
-
-      surface = dri_surface_from_handle(screen,
-                                        buffers[i].name,
-                                        format,
-                                        dri_drawable->w,
-                                        dri_drawable->h,
-                                        buffers[i].pitch);
-
-      st_set_framebuffer_surface(drawable->stfb, index, surface);
-      pipe_surface_reference(&surface, NULL);
-   }
-   /* this needed, or else the state tracker fails to pick the new buffers */
-   st_resize_framebuffer(drawable->stfb, dri_drawable->w, dri_drawable->h);
-}
-
-
-void
-dri_flush_frontbuffer(struct pipe_screen *screen,
-                      struct pipe_surface *surf,
-                      void *context_private)
-{
-   struct dri_context *ctx = (struct dri_context *)context_private;
-   dri_copy_to_front(ctx->dPriv, surf, 0, 0, surf->width, surf->height);
-}
-
-
-void
-dri_swap_buffers(__DRIdrawablePrivate * dPriv)
-{
-   /* not needed for dri2 */
-   assert(0);
-}
-
-
-void
-dri_copy_sub_buffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h)
-{
-   /* not needed for dri2 */
-   assert(0);
-}
-
-
-/**
- * This is called when we need to set up GL rendering to a new X window.
- */
-boolean
-dri_create_buffer(__DRIscreenPrivate *sPriv,
-                  __DRIdrawablePrivate *dPriv,
-                  const __GLcontextModes *visual,
-                  boolean isPixmap)
-{
-   enum pipe_format colorFormat, depthFormat, stencilFormat;
-   struct dri_screen *screen = sPriv->private;
-   struct dri_drawable *drawable = NULL;
-   struct pipe_screen *pscreen = screen->pipe_screen;
-   int i;
-
-   if (isPixmap)
-      goto fail; /* not implemented */
-
-   drawable = CALLOC_STRUCT(dri_drawable);
-   if (drawable == NULL)
-      goto fail;
-
-   /* XXX: todo: use the pipe_screen queries to figure out which
-    * render targets are supportable.
-    */
-   assert(visual->redBits == 8);
-   assert(visual->depthBits == 24 || visual->depthBits == 0);
-   assert(visual->stencilBits == 8 || visual->stencilBits == 0);
-
-   colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM;
-
-   if (visual->depthBits) {
-      if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM,
-                                       PIPE_TEXTURE_2D,
-                                       PIPE_TEXTURE_USAGE_RENDER_TARGET |
-                                       PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0))
-         depthFormat = PIPE_FORMAT_Z24S8_UNORM;
-      else
-         depthFormat = PIPE_FORMAT_S8Z24_UNORM;
-   } else
-      depthFormat = PIPE_FORMAT_NONE;
-
-   if (visual->stencilBits) {
-      if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM,
-                                       PIPE_TEXTURE_2D,
-                                       PIPE_TEXTURE_USAGE_RENDER_TARGET |
-                                       PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0))
-         stencilFormat = PIPE_FORMAT_Z24S8_UNORM;
-      else
-         stencilFormat = PIPE_FORMAT_S8Z24_UNORM;
-   } else
-      stencilFormat = PIPE_FORMAT_NONE;
-
-   drawable->stfb = st_create_framebuffer(visual,
-                                          colorFormat,
-                                          depthFormat,
-                                          stencilFormat,
-                                          dPriv->w,
-                                          dPriv->h,
-                                          (void*) drawable);
-   if (drawable->stfb == NULL)
-      goto fail;
-
-   drawable->sPriv = sPriv;
-   drawable->dPriv = dPriv;
-   dPriv->driverPrivate = (void *) drawable;
-
-   /* setup dri2 buffers information */
-   i = 0;
-   drawable->attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
-#if 0
-   /* TODO incase of double buffer visual, delay fake creation */
-   drawable->attachments[i++] = __DRI_BUFFER_FAKE_FRONT_LEFT;
-#endif
-   if (visual->doubleBufferMode)
-      drawable->attachments[i++] = __DRI_BUFFER_BACK_LEFT;
-   if (visual->depthBits)
-      drawable->attachments[i++] = __DRI_BUFFER_DEPTH;
-   if (visual->stencilBits)
-      drawable->attachments[i++] = __DRI_BUFFER_STENCIL;
-   drawable->num_attachments = i;
-
-   return GL_TRUE;
-fail:
-   FREE(drawable);
-   return GL_FALSE;
-}
-
-
-void
-dri_destroy_buffer(__DRIdrawablePrivate *dPriv)
-{
-   struct dri_drawable *drawable = dri_drawable(dPriv);
-
-   st_unreference_framebuffer(drawable->stfb);
-
-   FREE(drawable);
-}
-
-/* vim: set sw=3 ts=8 sts=3 expandtab: */
diff --git a/src/gallium/state_trackers/egl/egl_tracker.c b/src/gallium/state_trackers/egl/egl_tracker.c
index abdf84544f3..8e620084617 100644
--- a/src/gallium/state_trackers/egl/egl_tracker.c
+++ b/src/gallium/state_trackers/egl/egl_tracker.c
@@ -146,7 +146,7 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor)
 	dev->drmFD = fd;
 	drm_get_device_id(dev);
 
-	dev->screen = drm_api_hooks.create_screen(dev->drmFD, dev->deviceID);
+	dev->screen = drm_api_hooks.create_screen(dev->drmFD, NULL);
 	if (!dev->screen)
 		goto err_screen;
 	dev->winsys = dev->screen->winsys;
diff --git a/src/gallium/state_trackers/glx/xlib/fakeglx.c b/src/gallium/state_trackers/glx/xlib/fakeglx.c
index 65e7048188e..85e7ecfb9e8 100644
--- a/src/gallium/state_trackers/glx/xlib/fakeglx.c
+++ b/src/gallium/state_trackers/glx/xlib/fakeglx.c
@@ -97,6 +97,9 @@ struct fake_glx_context {
 
 
 
+#define DEFAULT_DIRECT GL_TRUE
+
+
 /**********************************************************************/
 /***                       GLX Visual Code                          ***/
 /**********************************************************************/
@@ -1059,7 +1062,7 @@ Fake_glXCreateContext( Display *dpy, XVisualInfo *visinfo,
       return NULL;
    }
 
-   glxCtx->glxContext.isDirect = GL_FALSE;
+   glxCtx->glxContext.isDirect = DEFAULT_DIRECT;
    glxCtx->glxContext.currentDpy = dpy;
    glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
 
@@ -1296,9 +1299,9 @@ Fake_glXDestroyContext( Display *dpy, GLXContext ctx )
 static Bool
 Fake_glXIsDirect( Display *dpy, GLXContext ctx )
 {
-   (void) dpy;
+   struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx;
    (void) ctx;
-   return False;
+   return glxCtx->glxContext.isDirect;
 }
 
 
@@ -2055,7 +2058,7 @@ Fake_glXCreateNewContext( Display *dpy, GLXFBConfig config,
       return NULL;
    }
 
-   glxCtx->glxContext.isDirect = GL_FALSE;
+   glxCtx->glxContext.isDirect = DEFAULT_DIRECT;
    glxCtx->glxContext.currentDpy = dpy;
    glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
 
@@ -2277,7 +2280,7 @@ Fake_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int re
       return NULL;
    }
 
-   glxCtx->glxContext.isDirect = GL_FALSE;
+   glxCtx->glxContext.isDirect = DEFAULT_DIRECT;
    glxCtx->glxContext.currentDpy = dpy;
    glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
 
diff --git a/src/gallium/state_trackers/glx/xlib/xm_api.c b/src/gallium/state_trackers/glx/xlib/xm_api.c
index a3d16516531..79c2230588f 100644
--- a/src/gallium/state_trackers/glx/xlib/xm_api.c
+++ b/src/gallium/state_trackers/glx/xlib/xm_api.c
@@ -1100,26 +1100,19 @@ XMesaContext XMesaGetCurrentContext( void )
 
 
 
-
-
-
-/*
- * Copy the back buffer to the front buffer.  If there's no back buffer
- * this is a no-op.
+/**
+ * Swap front and back color buffers and have winsys display front buffer.
+ * If there's no front color buffer no swap actually occurs.
  */
 PUBLIC
 void XMesaSwapBuffers( XMesaBuffer b )
 {
-   struct pipe_surface *surf;
+   struct pipe_surface *frontLeftSurf;
 
-   /* If we're swapping the buffer associated with the current context
-    * we have to flush any pending rendering commands first.
-    */
-   st_notify_swapbuffers(b->stfb);
+   st_swapbuffers(b->stfb, &frontLeftSurf, NULL);
 
-   st_get_framebuffer_surface(b->stfb, ST_SURFACE_BACK_LEFT, &surf);
-   if (surf) {
-      driver.display_surface(b, surf);
+   if (frontLeftSurf) {
+      driver.display_surface(b, frontLeftSurf);
    }
 
    xmesa_check_and_update_buffer_size(NULL, b);
diff --git a/src/gallium/state_trackers/wgl/SConscript b/src/gallium/state_trackers/wgl/SConscript
index 61fd8bfc0c8..5bbcc7175f7 100644
--- a/src/gallium/state_trackers/wgl/SConscript
+++ b/src/gallium/state_trackers/wgl/SConscript
@@ -26,7 +26,7 @@ if env['platform'] in ['windows']:
         'shared/stw_device.c',
         'shared/stw_framebuffer.c',
         'shared/stw_pixelformat.c',
-        'shared/stw_arbextensionsstring.c',
+        'shared/stw_extensionsstring.c',
         'shared/stw_getprocaddress.c',
         'shared/stw_arbpixelformat.c',
         'shared/stw_tls.c',
diff --git a/src/gallium/state_trackers/wgl/icd/stw_icd.c b/src/gallium/state_trackers/wgl/icd/stw_icd.c
index faf7f2f4106..62af7652116 100644
--- a/src/gallium/state_trackers/wgl/icd/stw_icd.c
+++ b/src/gallium/state_trackers/wgl/icd/stw_icd.c
@@ -38,9 +38,6 @@
 
 #define DBG 0
 
-static GLCLTPROCTABLE cpt;
-static boolean cpt_initialized = FALSE;
-
 
 BOOL APIENTRY
 DrvCopyContext(
@@ -165,351 +162,352 @@ DrvSetCallbackProcs(
 }
 
 
-static void init_proc_table( GLCLTPROCTABLE *cpt )
-{ 
-   GLDISPATCHTABLE *disp = &cpt->glDispatchTable;
+/**
+ * Although WGL allows different dispatch entrypoints per context 
+ */
+static const GLCLTPROCTABLE cpt =
+{
+   OPENGL_VERSION_110_ENTRIES,
+   {
+      &glNewList,
+      &glEndList,
+      &glCallList,
+      &glCallLists,
+      &glDeleteLists,
+      &glGenLists,
+      &glListBase,
+      &glBegin,
+      &glBitmap,
+      &glColor3b,
+      &glColor3bv,
+      &glColor3d,
+      &glColor3dv,
+      &glColor3f,
+      &glColor3fv,
+      &glColor3i,
+      &glColor3iv,
+      &glColor3s,
+      &glColor3sv,
+      &glColor3ub,
+      &glColor3ubv,
+      &glColor3ui,
+      &glColor3uiv,
+      &glColor3us,
+      &glColor3usv,
+      &glColor4b,
+      &glColor4bv,
+      &glColor4d,
+      &glColor4dv,
+      &glColor4f,
+      &glColor4fv,
+      &glColor4i,
+      &glColor4iv,
+      &glColor4s,
+      &glColor4sv,
+      &glColor4ub,
+      &glColor4ubv,
+      &glColor4ui,
+      &glColor4uiv,
+      &glColor4us,
+      &glColor4usv,
+      &glEdgeFlag,
+      &glEdgeFlagv,
+      &glEnd,
+      &glIndexd,
+      &glIndexdv,
+      &glIndexf,
+      &glIndexfv,
+      &glIndexi,
+      &glIndexiv,
+      &glIndexs,
+      &glIndexsv,
+      &glNormal3b,
+      &glNormal3bv,
+      &glNormal3d,
+      &glNormal3dv,
+      &glNormal3f,
+      &glNormal3fv,
+      &glNormal3i,
+      &glNormal3iv,
+      &glNormal3s,
+      &glNormal3sv,
+      &glRasterPos2d,
+      &glRasterPos2dv,
+      &glRasterPos2f,
+      &glRasterPos2fv,
+      &glRasterPos2i,
+      &glRasterPos2iv,
+      &glRasterPos2s,
+      &glRasterPos2sv,
+      &glRasterPos3d,
+      &glRasterPos3dv,
+      &glRasterPos3f,
+      &glRasterPos3fv,
+      &glRasterPos3i,
+      &glRasterPos3iv,
+      &glRasterPos3s,
+      &glRasterPos3sv,
+      &glRasterPos4d,
+      &glRasterPos4dv,
+      &glRasterPos4f,
+      &glRasterPos4fv,
+      &glRasterPos4i,
+      &glRasterPos4iv,
+      &glRasterPos4s,
+      &glRasterPos4sv,
+      &glRectd,
+      &glRectdv,
+      &glRectf,
+      &glRectfv,
+      &glRecti,
+      &glRectiv,
+      &glRects,
+      &glRectsv,
+      &glTexCoord1d,
+      &glTexCoord1dv,
+      &glTexCoord1f,
+      &glTexCoord1fv,
+      &glTexCoord1i,
+      &glTexCoord1iv,
+      &glTexCoord1s,
+      &glTexCoord1sv,
+      &glTexCoord2d,
+      &glTexCoord2dv,
+      &glTexCoord2f,
+      &glTexCoord2fv,
+      &glTexCoord2i,
+      &glTexCoord2iv,
+      &glTexCoord2s,
+      &glTexCoord2sv,
+      &glTexCoord3d,
+      &glTexCoord3dv,
+      &glTexCoord3f,
+      &glTexCoord3fv,
+      &glTexCoord3i,
+      &glTexCoord3iv,
+      &glTexCoord3s,
+      &glTexCoord3sv,
+      &glTexCoord4d,
+      &glTexCoord4dv,
+      &glTexCoord4f,
+      &glTexCoord4fv,
+      &glTexCoord4i,
+      &glTexCoord4iv,
+      &glTexCoord4s,
+      &glTexCoord4sv,
+      &glVertex2d,
+      &glVertex2dv,
+      &glVertex2f,
+      &glVertex2fv,
+      &glVertex2i,
+      &glVertex2iv,
+      &glVertex2s,
+      &glVertex2sv,
+      &glVertex3d,
+      &glVertex3dv,
+      &glVertex3f,
+      &glVertex3fv,
+      &glVertex3i,
+      &glVertex3iv,
+      &glVertex3s,
+      &glVertex3sv,
+      &glVertex4d,
+      &glVertex4dv,
+      &glVertex4f,
+      &glVertex4fv,
+      &glVertex4i,
+      &glVertex4iv,
+      &glVertex4s,
+      &glVertex4sv,
+      &glClipPlane,
+      &glColorMaterial,
+      &glCullFace,
+      &glFogf,
+      &glFogfv,
+      &glFogi,
+      &glFogiv,
+      &glFrontFace,
+      &glHint,
+      &glLightf,
+      &glLightfv,
+      &glLighti,
+      &glLightiv,
+      &glLightModelf,
+      &glLightModelfv,
+      &glLightModeli,
+      &glLightModeliv,
+      &glLineStipple,
+      &glLineWidth,
+      &glMaterialf,
+      &glMaterialfv,
+      &glMateriali,
+      &glMaterialiv,
+      &glPointSize,
+      &glPolygonMode,
+      &glPolygonStipple,
+      &glScissor,
+      &glShadeModel,
+      &glTexParameterf,
+      &glTexParameterfv,
+      &glTexParameteri,
+      &glTexParameteriv,
+      &glTexImage1D,
+      &glTexImage2D,
+      &glTexEnvf,
+      &glTexEnvfv,
+      &glTexEnvi,
+      &glTexEnviv,
+      &glTexGend,
+      &glTexGendv,
+      &glTexGenf,
+      &glTexGenfv,
+      &glTexGeni,
+      &glTexGeniv,
+      &glFeedbackBuffer,
+      &glSelectBuffer,
+      &glRenderMode,
+      &glInitNames,
+      &glLoadName,
+      &glPassThrough,
+      &glPopName,
+      &glPushName,
+      &glDrawBuffer,
+      &glClear,
+      &glClearAccum,
+      &glClearIndex,
+      &glClearColor,
+      &glClearStencil,
+      &glClearDepth,
+      &glStencilMask,
+      &glColorMask,
+      &glDepthMask,
+      &glIndexMask,
+      &glAccum,
+      &glDisable,
+      &glEnable,
+      &glFinish,
+      &glFlush,
+      &glPopAttrib,
+      &glPushAttrib,
+      &glMap1d,
+      &glMap1f,
+      &glMap2d,
+      &glMap2f,
+      &glMapGrid1d,
+      &glMapGrid1f,
+      &glMapGrid2d,
+      &glMapGrid2f,
+      &glEvalCoord1d,
+      &glEvalCoord1dv,
+      &glEvalCoord1f,
+      &glEvalCoord1fv,
+      &glEvalCoord2d,
+      &glEvalCoord2dv,
+      &glEvalCoord2f,
+      &glEvalCoord2fv,
+      &glEvalMesh1,
+      &glEvalPoint1,
+      &glEvalMesh2,
+      &glEvalPoint2,
+      &glAlphaFunc,
+      &glBlendFunc,
+      &glLogicOp,
+      &glStencilFunc,
+      &glStencilOp,
+      &glDepthFunc,
+      &glPixelZoom,
+      &glPixelTransferf,
+      &glPixelTransferi,
+      &glPixelStoref,
+      &glPixelStorei,
+      &glPixelMapfv,
+      &glPixelMapuiv,
+      &glPixelMapusv,
+      &glReadBuffer,
+      &glCopyPixels,
+      &glReadPixels,
+      &glDrawPixels,
+      &glGetBooleanv,
+      &glGetClipPlane,
+      &glGetDoublev,
+      &glGetError,
+      &glGetFloatv,
+      &glGetIntegerv,
+      &glGetLightfv,
+      &glGetLightiv,
+      &glGetMapdv,
+      &glGetMapfv,
+      &glGetMapiv,
+      &glGetMaterialfv,
+      &glGetMaterialiv,
+      &glGetPixelMapfv,
+      &glGetPixelMapuiv,
+      &glGetPixelMapusv,
+      &glGetPolygonStipple,
+      &glGetString,
+      &glGetTexEnvfv,
+      &glGetTexEnviv,
+      &glGetTexGendv,
+      &glGetTexGenfv,
+      &glGetTexGeniv,
+      &glGetTexImage,
+      &glGetTexParameterfv,
+      &glGetTexParameteriv,
+      &glGetTexLevelParameterfv,
+      &glGetTexLevelParameteriv,
+      &glIsEnabled,
+      &glIsList,
+      &glDepthRange,
+      &glFrustum,
+      &glLoadIdentity,
+      &glLoadMatrixf,
+      &glLoadMatrixd,
+      &glMatrixMode,
+      &glMultMatrixf,
+      &glMultMatrixd,
+      &glOrtho,
+      &glPopMatrix,
+      &glPushMatrix,
+      &glRotated,
+      &glRotatef,
+      &glScaled,
+      &glScalef,
+      &glTranslated,
+      &glTranslatef,
+      &glViewport,
+      &glArrayElement,
+      &glBindTexture,
+      &glColorPointer,
+      &glDisableClientState,
+      &glDrawArrays,
+      &glDrawElements,
+      &glEdgeFlagPointer,
+      &glEnableClientState,
+      &glIndexPointer,
+      &glIndexub,
+      &glIndexubv,
+      &glInterleavedArrays,
+      &glNormalPointer,
+      &glPolygonOffset,
+      &glTexCoordPointer,
+      &glVertexPointer,
+      &glAreTexturesResident,
+      &glCopyTexImage1D,
+      &glCopyTexImage2D,
+      &glCopyTexSubImage1D,
+      &glCopyTexSubImage2D,
+      &glDeleteTextures,
+      &glGenTextures,
+      &glGetPointerv,
+      &glIsTexture,
+      &glPrioritizeTextures,
+      &glTexSubImage1D,
+      &glTexSubImage2D,
+      &glPopClientAttrib,
+      &glPushClientAttrib
+   }
+};
 
-   memset( cpt, 0, sizeof *cpt );
-   cpt->cEntries = OPENGL_VERSION_110_ENTRIES;
-
-#define GPA_GL( NAME ) disp->NAME = gl##NAME
-   GPA_GL( NewList );
-   GPA_GL( EndList );
-   GPA_GL( CallList );
-   GPA_GL( CallLists );
-   GPA_GL( DeleteLists );
-   GPA_GL( GenLists );
-   GPA_GL( ListBase );
-   GPA_GL( Begin );
-   GPA_GL( Bitmap );
-   GPA_GL( Color3b );
-   GPA_GL( Color3bv );
-   GPA_GL( Color3d );
-   GPA_GL( Color3dv );
-   GPA_GL( Color3f );
-   GPA_GL( Color3fv );
-   GPA_GL( Color3i );
-   GPA_GL( Color3iv );
-   GPA_GL( Color3s );
-   GPA_GL( Color3sv );
-   GPA_GL( Color3ub );
-   GPA_GL( Color3ubv );
-   GPA_GL( Color3ui );
-   GPA_GL( Color3uiv );
-   GPA_GL( Color3us );
-   GPA_GL( Color3usv );
-   GPA_GL( Color4b );
-   GPA_GL( Color4bv );
-   GPA_GL( Color4d );
-   GPA_GL( Color4dv );
-   GPA_GL( Color4f );
-   GPA_GL( Color4fv );
-   GPA_GL( Color4i );
-   GPA_GL( Color4iv );
-   GPA_GL( Color4s );
-   GPA_GL( Color4sv );
-   GPA_GL( Color4ub );
-   GPA_GL( Color4ubv );
-   GPA_GL( Color4ui );
-   GPA_GL( Color4uiv );
-   GPA_GL( Color4us );
-   GPA_GL( Color4usv );
-   GPA_GL( EdgeFlag );
-   GPA_GL( EdgeFlagv );
-   GPA_GL( End );
-   GPA_GL( Indexd );
-   GPA_GL( Indexdv );
-   GPA_GL( Indexf );
-   GPA_GL( Indexfv );
-   GPA_GL( Indexi );
-   GPA_GL( Indexiv );
-   GPA_GL( Indexs );
-   GPA_GL( Indexsv );
-   GPA_GL( Normal3b );
-   GPA_GL( Normal3bv );
-   GPA_GL( Normal3d );
-   GPA_GL( Normal3dv );
-   GPA_GL( Normal3f );
-   GPA_GL( Normal3fv );
-   GPA_GL( Normal3i );
-   GPA_GL( Normal3iv );
-   GPA_GL( Normal3s );
-   GPA_GL( Normal3sv );
-   GPA_GL( RasterPos2d );
-   GPA_GL( RasterPos2dv );
-   GPA_GL( RasterPos2f );
-   GPA_GL( RasterPos2fv );
-   GPA_GL( RasterPos2i );
-   GPA_GL( RasterPos2iv );
-   GPA_GL( RasterPos2s );
-   GPA_GL( RasterPos2sv );
-   GPA_GL( RasterPos3d );
-   GPA_GL( RasterPos3dv );
-   GPA_GL( RasterPos3f );
-   GPA_GL( RasterPos3fv );
-   GPA_GL( RasterPos3i );
-   GPA_GL( RasterPos3iv );
-   GPA_GL( RasterPos3s );
-   GPA_GL( RasterPos3sv );
-   GPA_GL( RasterPos4d );
-   GPA_GL( RasterPos4dv );
-   GPA_GL( RasterPos4f );
-   GPA_GL( RasterPos4fv );
-   GPA_GL( RasterPos4i );
-   GPA_GL( RasterPos4iv );
-   GPA_GL( RasterPos4s );
-   GPA_GL( RasterPos4sv );
-   GPA_GL( Rectd );
-   GPA_GL( Rectdv );
-   GPA_GL( Rectf );
-   GPA_GL( Rectfv );
-   GPA_GL( Recti );
-   GPA_GL( Rectiv );
-   GPA_GL( Rects );
-   GPA_GL( Rectsv );
-   GPA_GL( TexCoord1d );
-   GPA_GL( TexCoord1dv );
-   GPA_GL( TexCoord1f );
-   GPA_GL( TexCoord1fv );
-   GPA_GL( TexCoord1i );
-   GPA_GL( TexCoord1iv );
-   GPA_GL( TexCoord1s );
-   GPA_GL( TexCoord1sv );
-   GPA_GL( TexCoord2d );
-   GPA_GL( TexCoord2dv );
-   GPA_GL( TexCoord2f );
-   GPA_GL( TexCoord2fv );
-   GPA_GL( TexCoord2i );
-   GPA_GL( TexCoord2iv );
-   GPA_GL( TexCoord2s );
-   GPA_GL( TexCoord2sv );
-   GPA_GL( TexCoord3d );
-   GPA_GL( TexCoord3dv );
-   GPA_GL( TexCoord3f );
-   GPA_GL( TexCoord3fv );
-   GPA_GL( TexCoord3i );
-   GPA_GL( TexCoord3iv );
-   GPA_GL( TexCoord3s );
-   GPA_GL( TexCoord3sv );
-   GPA_GL( TexCoord4d );
-   GPA_GL( TexCoord4dv );
-   GPA_GL( TexCoord4f );
-   GPA_GL( TexCoord4fv );
-   GPA_GL( TexCoord4i );
-   GPA_GL( TexCoord4iv );
-   GPA_GL( TexCoord4s );
-   GPA_GL( TexCoord4sv );
-   GPA_GL( Vertex2d );
-   GPA_GL( Vertex2dv );
-   GPA_GL( Vertex2f );
-   GPA_GL( Vertex2fv );
-   GPA_GL( Vertex2i );
-   GPA_GL( Vertex2iv );
-   GPA_GL( Vertex2s );
-   GPA_GL( Vertex2sv );
-   GPA_GL( Vertex3d );
-   GPA_GL( Vertex3dv );
-   GPA_GL( Vertex3f );
-   GPA_GL( Vertex3fv );
-   GPA_GL( Vertex3i );
-   GPA_GL( Vertex3iv );
-   GPA_GL( Vertex3s );
-   GPA_GL( Vertex3sv );
-   GPA_GL( Vertex4d );
-   GPA_GL( Vertex4dv );
-   GPA_GL( Vertex4f );
-   GPA_GL( Vertex4fv );
-   GPA_GL( Vertex4i );
-   GPA_GL( Vertex4iv );
-   GPA_GL( Vertex4s );
-   GPA_GL( Vertex4sv );
-   GPA_GL( ClipPlane );
-   GPA_GL( ColorMaterial );
-   GPA_GL( CullFace );
-   GPA_GL( Fogf );
-   GPA_GL( Fogfv );
-   GPA_GL( Fogi );
-   GPA_GL( Fogiv );
-   GPA_GL( FrontFace );
-   GPA_GL( Hint );
-   GPA_GL( Lightf );
-   GPA_GL( Lightfv );
-   GPA_GL( Lighti );
-   GPA_GL( Lightiv );
-   GPA_GL( LightModelf );
-   GPA_GL( LightModelfv );
-   GPA_GL( LightModeli );
-   GPA_GL( LightModeliv );
-   GPA_GL( LineStipple );
-   GPA_GL( LineWidth );
-   GPA_GL( Materialf );
-   GPA_GL( Materialfv );
-   GPA_GL( Materiali );
-   GPA_GL( Materialiv );
-   GPA_GL( PointSize );
-   GPA_GL( PolygonMode );
-   GPA_GL( PolygonStipple );
-   GPA_GL( Scissor );
-   GPA_GL( ShadeModel );
-   GPA_GL( TexParameterf );
-   GPA_GL( TexParameterfv );
-   GPA_GL( TexParameteri );
-   GPA_GL( TexParameteriv );
-   GPA_GL( TexImage1D );
-   GPA_GL( TexImage2D );
-   GPA_GL( TexEnvf );
-   GPA_GL( TexEnvfv );
-   GPA_GL( TexEnvi );
-   GPA_GL( TexEnviv );
-   GPA_GL( TexGend );
-   GPA_GL( TexGendv );
-   GPA_GL( TexGenf );
-   GPA_GL( TexGenfv );
-   GPA_GL( TexGeni );
-   GPA_GL( TexGeniv );
-   GPA_GL( FeedbackBuffer );
-   GPA_GL( SelectBuffer );
-   GPA_GL( RenderMode );
-   GPA_GL( InitNames );
-   GPA_GL( LoadName );
-   GPA_GL( PassThrough );
-   GPA_GL( PopName );
-   GPA_GL( PushName );
-   GPA_GL( DrawBuffer );
-   GPA_GL( Clear );
-   GPA_GL( ClearAccum );
-   GPA_GL( ClearIndex );
-   GPA_GL( ClearColor );
-   GPA_GL( ClearStencil );
-   GPA_GL( ClearDepth );
-   GPA_GL( StencilMask );
-   GPA_GL( ColorMask );
-   GPA_GL( DepthMask );
-   GPA_GL( IndexMask );
-   GPA_GL( Accum );
-   GPA_GL( Disable );
-   GPA_GL( Enable );
-   GPA_GL( Finish );
-   GPA_GL( Flush );
-   GPA_GL( PopAttrib );
-   GPA_GL( PushAttrib );
-   GPA_GL( Map1d );
-   GPA_GL( Map1f );
-   GPA_GL( Map2d );
-   GPA_GL( Map2f );
-   GPA_GL( MapGrid1d );
-   GPA_GL( MapGrid1f );
-   GPA_GL( MapGrid2d );
-   GPA_GL( MapGrid2f );
-   GPA_GL( EvalCoord1d );
-   GPA_GL( EvalCoord1dv );
-   GPA_GL( EvalCoord1f );
-   GPA_GL( EvalCoord1fv );
-   GPA_GL( EvalCoord2d );
-   GPA_GL( EvalCoord2dv );
-   GPA_GL( EvalCoord2f );
-   GPA_GL( EvalCoord2fv );
-   GPA_GL( EvalMesh1 );
-   GPA_GL( EvalPoint1 );
-   GPA_GL( EvalMesh2 );
-   GPA_GL( EvalPoint2 );
-   GPA_GL( AlphaFunc );
-   GPA_GL( BlendFunc );
-   GPA_GL( LogicOp );
-   GPA_GL( StencilFunc );
-   GPA_GL( StencilOp );
-   GPA_GL( DepthFunc );
-   GPA_GL( PixelZoom );
-   GPA_GL( PixelTransferf );
-   GPA_GL( PixelTransferi );
-   GPA_GL( PixelStoref );
-   GPA_GL( PixelStorei );
-   GPA_GL( PixelMapfv );
-   GPA_GL( PixelMapuiv );
-   GPA_GL( PixelMapusv );
-   GPA_GL( ReadBuffer );
-   GPA_GL( CopyPixels );
-   GPA_GL( ReadPixels );
-   GPA_GL( DrawPixels );
-   GPA_GL( GetBooleanv );
-   GPA_GL( GetClipPlane );
-   GPA_GL( GetDoublev );
-   GPA_GL( GetError );
-   GPA_GL( GetFloatv );
-   GPA_GL( GetIntegerv );
-   GPA_GL( GetLightfv );
-   GPA_GL( GetLightiv );
-   GPA_GL( GetMapdv );
-   GPA_GL( GetMapfv );
-   GPA_GL( GetMapiv );
-   GPA_GL( GetMaterialfv );
-   GPA_GL( GetMaterialiv );
-   GPA_GL( GetPixelMapfv );
-   GPA_GL( GetPixelMapuiv );
-   GPA_GL( GetPixelMapusv );
-   GPA_GL( GetPolygonStipple );
-   GPA_GL( GetString );
-   GPA_GL( GetTexEnvfv );
-   GPA_GL( GetTexEnviv );
-   GPA_GL( GetTexGendv );
-   GPA_GL( GetTexGenfv );
-   GPA_GL( GetTexGeniv );
-   GPA_GL( GetTexImage );
-   GPA_GL( GetTexParameterfv );
-   GPA_GL( GetTexParameteriv );
-   GPA_GL( GetTexLevelParameterfv );
-   GPA_GL( GetTexLevelParameteriv );
-   GPA_GL( IsEnabled );
-   GPA_GL( IsList );
-   GPA_GL( DepthRange );
-   GPA_GL( Frustum );
-   GPA_GL( LoadIdentity );
-   GPA_GL( LoadMatrixf );
-   GPA_GL( LoadMatrixd );
-   GPA_GL( MatrixMode );
-   GPA_GL( MultMatrixf );
-   GPA_GL( MultMatrixd );
-   GPA_GL( Ortho );
-   GPA_GL( PopMatrix );
-   GPA_GL( PushMatrix );
-   GPA_GL( Rotated );
-   GPA_GL( Rotatef );
-   GPA_GL( Scaled );
-   GPA_GL( Scalef );
-   GPA_GL( Translated );
-   GPA_GL( Translatef );
-   GPA_GL( Viewport );
-   GPA_GL( ArrayElement );
-   GPA_GL( BindTexture );
-   GPA_GL( ColorPointer );
-   GPA_GL( DisableClientState );
-   GPA_GL( DrawArrays );
-   GPA_GL( DrawElements );
-   GPA_GL( EdgeFlagPointer );
-   GPA_GL( EnableClientState );
-   GPA_GL( IndexPointer );
-   GPA_GL( Indexub );
-   GPA_GL( Indexubv );
-   GPA_GL( InterleavedArrays );
-   GPA_GL( NormalPointer );
-   GPA_GL( PolygonOffset );
-   GPA_GL( TexCoordPointer );
-   GPA_GL( VertexPointer );
-   GPA_GL( AreTexturesResident );
-   GPA_GL( CopyTexImage1D );
-   GPA_GL( CopyTexImage2D );
-   GPA_GL( CopyTexSubImage1D );
-   GPA_GL( CopyTexSubImage2D );
-   GPA_GL( DeleteTextures );
-   GPA_GL( GenTextures );
-   GPA_GL( GetPointerv );
-   GPA_GL( IsTexture );
-   GPA_GL( PrioritizeTextures );
-   GPA_GL( TexSubImage1D );
-   GPA_GL( TexSubImage2D );
-   GPA_GL( PopClientAttrib );
-   GPA_GL( PushClientAttrib );
-}
 
 PGLCLTPROCTABLE APIENTRY
 DrvSetContext(
@@ -521,17 +519,10 @@ DrvSetContext(
       debug_printf( "%s( 0x%p, %u, 0x%p )\n", 
                     __FUNCTION__, hdc, dhglrc, pfnSetProcTable );
 
-   /* Although WGL allows different dispatch entrypoints per 
-    */
-   if (!cpt_initialized) {
-      init_proc_table( &cpt );
-      cpt_initialized = TRUE;
-   }
-
    if (!stw_make_current( hdc, dhglrc ))
       return NULL;
       
-   return &cpt;
+   return (GLCLTPROCTABLE *)&cpt;
 }
 
 int APIENTRY
@@ -571,7 +562,7 @@ DrvShareLists(
    if (DBG)
       debug_printf( "%s\n", __FUNCTION__ );
 
-   return FALSE;
+   return stw_share_lists(dhglrc1, dhglrc2);
 }
 
 BOOL APIENTRY
@@ -592,7 +583,7 @@ DrvSwapLayerBuffers(
    if (DBG)
       debug_printf( "%s\n", __FUNCTION__ );
 
-   return FALSE;
+   return stw_swap_layer_buffers( hdc, fuPlanes );
 }
 
 BOOL APIENTRY
diff --git a/src/gallium/state_trackers/wgl/shared/stw_context.c b/src/gallium/state_trackers/wgl/shared/stw_context.c
index f3c7af93f50..dd97e48b144 100644
--- a/src/gallium/state_trackers/wgl/shared/stw_context.c
+++ b/src/gallium/state_trackers/wgl/shared/stw_context.c
@@ -64,6 +64,7 @@ stw_copy_context(
 
    if (src && dst) { 
       /* FIXME */
+      assert(0);
       (void) src;
       (void) dst;
       (void) mask;
@@ -74,6 +75,30 @@ stw_copy_context(
    return ret;
 }
 
+BOOL
+stw_share_lists(
+   UINT_PTR hglrc1, 
+   UINT_PTR hglrc2 )
+{
+   struct stw_context *ctx1;
+   struct stw_context *ctx2;
+   BOOL ret = FALSE;
+
+   pipe_mutex_lock( stw_dev->mutex );
+   
+   ctx1 = stw_lookup_context_locked( hglrc1 );
+   ctx2 = stw_lookup_context_locked( hglrc2 );
+
+   if (ctx1 && ctx2 &&
+       ctx1->pfi == ctx2->pfi) { 
+      ret = _mesa_share_state(ctx2->st->ctx, ctx1->st->ctx);
+   }
+
+   pipe_mutex_unlock( stw_dev->mutex );
+   
+   return ret;
+}
+
 UINT_PTR
 stw_create_layer_context(
    HDC hdc,
@@ -85,7 +110,6 @@ stw_create_layer_context(
    GLvisual *visual = NULL;
    struct pipe_screen *screen = NULL;
    struct pipe_context *pipe = NULL;
-   UINT_PTR hglrc = 0;
 
    if(!stw_dev)
       return 0;
@@ -101,7 +125,7 @@ stw_create_layer_context(
 
    ctx = CALLOC_STRUCT( stw_context );
    if (ctx == NULL)
-      return 0;
+      goto no_ctx;
 
    ctx->hdc = hdc;
    ctx->color_bits = GetDeviceCaps( ctx->hdc, BITSPIXEL );
@@ -125,7 +149,7 @@ stw_create_layer_context(
       pf->pfd.cAccumAlphaBits,
       pf->numSamples );
    if (visual == NULL) 
-      goto fail;
+      goto no_visual;
 
    screen = stw_dev->screen;
 
@@ -137,7 +161,7 @@ stw_create_layer_context(
 
    pipe = stw_dev->stw_winsys->create_context( screen );
    if (pipe == NULL) 
-      goto fail;
+      goto no_pipe;
 
 #ifdef DEBUG
    /* Wrap context */
@@ -150,28 +174,29 @@ stw_create_layer_context(
 
    ctx->st = st_create_context( pipe, visual, NULL );
    if (ctx->st == NULL) 
-      goto fail;
+      goto no_st_ctx;
 
    ctx->st->ctx->DriverCtx = ctx;
    ctx->pfi = pf;
 
    pipe_mutex_lock( stw_dev->mutex );
-   hglrc = handle_table_add(stw_dev->ctx_table, ctx);
+   ctx->hglrc = handle_table_add(stw_dev->ctx_table, ctx);
    pipe_mutex_unlock( stw_dev->mutex );
+   if (!ctx->hglrc)
+      goto no_hglrc;
 
-   /* Success?
-    */
-   if (hglrc != 0)
-      return hglrc;
+   return ctx->hglrc;
 
-fail:
-   if (visual)
-      _mesa_destroy_visual( visual );
-   
-   if (pipe)
-      pipe->destroy( pipe );
-      
+no_hglrc:
+   st_destroy_context(ctx->st);
+   goto no_pipe; /* st_context_destroy already destroys pipe */
+no_st_ctx:
+   pipe->destroy( pipe );
+no_pipe:
+   _mesa_destroy_visual( visual );
+no_visual:
    FREE( ctx );
+no_ctx:
    return 0;
 }
 
@@ -253,31 +278,64 @@ stw_release_context(
 /* Find the width and height of the window named by hdc.
  */
 static void
-stw_get_window_size( HDC hdc, GLuint *width, GLuint *height )
+stw_get_window_size( HDC hdc, GLuint *pwidth, GLuint *pheight )
 {
-   if (WindowFromDC( hdc )) {
-      RECT rect;
+   GLuint width, height;
+   HWND hwnd;
 
-      GetClientRect( WindowFromDC( hdc ), &rect );
-      *width = rect.right - rect.left;
-      *height = rect.bottom - rect.top;
+   hwnd = WindowFromDC( hdc );
+   if (hwnd) {
+      RECT rect;
+      GetClientRect( hwnd, &rect );
+      width = rect.right - rect.left;
+      height = rect.bottom - rect.top;
    }
    else {
-      *width = GetDeviceCaps( hdc, HORZRES );
-      *height = GetDeviceCaps( hdc, VERTRES );
+      width = GetDeviceCaps( hdc, HORZRES );
+      height = GetDeviceCaps( hdc, VERTRES );
    }
+
+   if(width < 1)
+      width = 1;
+   if(height < 1)
+      height = 1;
+
+   *pwidth = width; 
+   *pheight = height; 
 }
 
 UINT_PTR
 stw_get_current_context( void )
 {
-   return stw_tls_get_data()->currentGLRC;
+   GET_CURRENT_CONTEXT( glcurctx );
+   struct stw_context *ctx;
+
+   if(!glcurctx)
+      return 0;
+   
+   ctx = (struct stw_context *)glcurctx->DriverCtx;
+   assert(ctx);
+   if(!ctx)
+      return 0;
+   
+   return ctx->hglrc;
 }
 
 HDC
 stw_get_current_dc( void )
 {
-    return stw_tls_get_data()->currentDC;
+   GET_CURRENT_CONTEXT( glcurctx );
+   struct stw_context *ctx;
+
+   if(!glcurctx)
+      return NULL;
+   
+   ctx = (struct stw_context *)glcurctx->DriverCtx;
+   assert(ctx);
+   if(!ctx)
+      return NULL;
+   
+   return ctx->hdc;
 }
 
 BOOL
@@ -290,7 +348,7 @@ stw_make_current(
    struct stw_framebuffer *fb;
    GLuint width = 0;
    GLuint height = 0;
-   struct stw_context *curctx;
+   struct stw_context *curctx = NULL;
 
    if (!stw_dev)
       return FALSE;
@@ -299,9 +357,6 @@ stw_make_current(
    ctx = stw_lookup_context_locked( hglrc );
    pipe_mutex_unlock( stw_dev->mutex );
 
-   stw_tls_get_data()->currentDC = hdc;
-   stw_tls_get_data()->currentGLRC = hglrc;
-
    if (glcurctx != NULL) {
       curctx = (struct stw_context *) glcurctx->DriverCtx;
 
diff --git a/src/gallium/state_trackers/wgl/shared/stw_context.h b/src/gallium/state_trackers/wgl/shared/stw_context.h
index bc3b1dc880d..e276737e85a 100644
--- a/src/gallium/state_trackers/wgl/shared/stw_context.h
+++ b/src/gallium/state_trackers/wgl/shared/stw_context.h
@@ -36,6 +36,7 @@ struct stw_pixelformat_info;
 struct stw_context
 {
    struct st_context *st;
+   UINT_PTR hglrc;
    HDC hdc;
    DWORD color_bits;
    const struct stw_pixelformat_info *pfi;
diff --git a/src/gallium/state_trackers/wgl/shared/stw_arbextensionsstring.c b/src/gallium/state_trackers/wgl/shared/stw_extensionsstring.c
similarity index 82%
rename from src/gallium/state_trackers/wgl/shared/stw_arbextensionsstring.c
rename to src/gallium/state_trackers/wgl/shared/stw_extensionsstring.c
index cd9fe93eeec..2660c591f9f 100644
--- a/src/gallium/state_trackers/wgl/shared/stw_arbextensionsstring.c
+++ b/src/gallium/state_trackers/wgl/shared/stw_extensionsstring.c
@@ -1,5 +1,6 @@
 /**************************************************************************
  * 
+ * Copyright 2009 VMware, Inc.
  * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
  * All Rights Reserved.
  * 
@@ -33,14 +34,25 @@
 #include <GL/wglext.h>
 
 
+static const char *stw_extension_string = 
+   "WGL_ARB_extensions_string "
+   "WGL_ARB_multisample "
+   "WGL_ARB_pixel_format "
+   "WGL_EXT_extensions_string";
+
+
 WINGDIAPI const char * APIENTRY
 wglGetExtensionsStringARB(
    HDC hdc )
 {
    (void) hdc;
 
-   return
-      "WGL_ARB_extensions_string "
-      "WGL_ARB_multisample "
-      "WGL_ARB_pixel_format";
+   return stw_extension_string;
+}
+
+
+WINGDIAPI const char * APIENTRY
+wglGetExtensionsStringEXT( void )
+{
+   return stw_extension_string;
 }
diff --git a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c
index 4348b8f3262..f66f4295424 100644
--- a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c
+++ b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c
@@ -237,6 +237,18 @@ stw_swap_buffers(
 }
 
 
+BOOL
+stw_swap_layer_buffers(
+   HDC hdc,
+   UINT fuPlanes )
+{
+   if(fuPlanes & WGL_SWAP_MAIN_PLANE)
+      return stw_swap_buffers(hdc);
+
+   return FALSE;
+}
+
+
 boolean
 stw_framebuffer_init_thread(void)
 {
diff --git a/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c b/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c
index aa43120955e..4070cbd5c00 100644
--- a/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c
+++ b/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c
@@ -53,6 +53,9 @@ static const struct stw_extension_entry stw_extension_entries[] = {
    STW_EXTENSION_ENTRY( wglGetPixelFormatAttribfvARB ),
    STW_EXTENSION_ENTRY( wglGetPixelFormatAttribivARB ),
 
+   /* WGL_EXT_extensions_string */
+   STW_EXTENSION_ENTRY( wglGetExtensionsStringEXT ),
+
    { NULL, NULL }
 };
 
diff --git a/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c b/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c
index 9e642cbdd48..b81d2b59a48 100644
--- a/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c
+++ b/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c
@@ -119,6 +119,7 @@ stw_pixelformat_add(
    struct stw_device *stw_dev,
    const struct stw_pf_color_info *color,
    const struct stw_pf_depth_info *depth,
+   unsigned accum,
    boolean doublebuffer,
    unsigned samples )
 {
@@ -158,7 +159,7 @@ stw_pixelformat_add(
    
    pfi->pfd.iPixelType = PFD_TYPE_RGBA;
 
-   pfi->pfd.cColorBits = color->bits.red + color->bits.green + color->bits.blue;
+   pfi->pfd.cColorBits = color->bits.red + color->bits.green + color->bits.blue + color->bits.alpha;
    pfi->pfd.cRedBits = color->bits.red;
    pfi->pfd.cRedShift = color->shift.red;
    pfi->pfd.cGreenBits = color->bits.green;
@@ -167,11 +168,11 @@ stw_pixelformat_add(
    pfi->pfd.cBlueShift = color->shift.blue;
    pfi->pfd.cAlphaBits = color->bits.alpha;
    pfi->pfd.cAlphaShift = color->shift.alpha;
-   pfi->pfd.cAccumBits = 0;
-   pfi->pfd.cAccumRedBits = 0;
-   pfi->pfd.cAccumGreenBits = 0;
-   pfi->pfd.cAccumBlueBits = 0;
-   pfi->pfd.cAccumAlphaBits = 0;
+   pfi->pfd.cAccumBits = 4*accum;
+   pfi->pfd.cAccumRedBits = accum;
+   pfi->pfd.cAccumGreenBits = accum;
+   pfi->pfd.cAccumBlueBits = accum;
+   pfi->pfd.cAccumAlphaBits = accum;
    pfi->pfd.cDepthBits = depth->bits.depth;
    pfi->pfd.cStencilBits = depth->bits.stencil;
    pfi->pfd.cAuxBuffers = 0;
@@ -228,7 +229,8 @@ stw_pixelformat_init( void )
                                                PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0))
                   continue;
 
-               stw_pixelformat_add( stw_dev, color, depth, doublebuffer, samples );
+               stw_pixelformat_add( stw_dev, color, depth,  0, doublebuffer, samples );
+               stw_pixelformat_add( stw_dev, color, depth, 16, doublebuffer, samples );
             }
          }
       }
diff --git a/src/gallium/state_trackers/wgl/shared/stw_public.h b/src/gallium/state_trackers/wgl/shared/stw_public.h
index 39d377c16b5..7fe9cfb3561 100644
--- a/src/gallium/state_trackers/wgl/shared/stw_public.h
+++ b/src/gallium/state_trackers/wgl/shared/stw_public.h
@@ -37,6 +37,8 @@ BOOL stw_copy_context( UINT_PTR hglrcSrc,
 UINT_PTR stw_create_layer_context( HDC hdc, 
                                    int iLayerPlane );
 
+BOOL stw_share_lists( UINT_PTR hglrc1, UINT_PTR hglrc2 );
+
 BOOL stw_delete_context( UINT_PTR hglrc );
 
 BOOL
@@ -50,6 +52,9 @@ BOOL stw_make_current( HDC hdc, UINT_PTR hglrc );
 
 BOOL stw_swap_buffers( HDC hdc );
 
+BOOL
+stw_swap_layer_buffers( HDC hdc, UINT fuPlanes );
+
 PROC stw_get_proc_address( LPCSTR lpszProc );
 
 int stw_pixelformat_describe( HDC hdc,
diff --git a/src/gallium/state_trackers/wgl/shared/stw_tls.c b/src/gallium/state_trackers/wgl/shared/stw_tls.c
index e72bafb8804..95863ca9cfb 100644
--- a/src/gallium/state_trackers/wgl/shared/stw_tls.c
+++ b/src/gallium/state_trackers/wgl/shared/stw_tls.c
@@ -44,6 +44,20 @@ stw_tls_init(void)
    return TRUE;
 }
 
+static INLINE struct stw_tls_data *
+stw_tls_data_create()
+{
+   struct stw_tls_data *data;
+
+   data = CALLOC_STRUCT(stw_tls_data);
+   if (!data)
+      return NULL;
+
+   data->currentPixelFormat = 0;
+   
+   return data;
+}
+
 boolean
 stw_tls_init_thread(void)
 {
@@ -53,14 +67,9 @@ stw_tls_init_thread(void)
       return FALSE;
    }
 
-   data = MALLOC(sizeof(*data));
-   if (!data) {
+   data = stw_tls_data_create();
+   if(!data)
       return FALSE;
-   }
-
-   data->currentPixelFormat = 0;
-   data->currentDC = NULL;
-   data->currentGLRC = 0;
 
    TlsSetValue(tlsIndex, data);
 
@@ -93,9 +102,23 @@ stw_tls_cleanup(void)
 struct stw_tls_data *
 stw_tls_get_data(void)
 {
+   struct stw_tls_data *data;
+   
    if (tlsIndex == TLS_OUT_OF_INDEXES) {
       return NULL;
    }
+   
+   data = (struct stw_tls_data *) TlsGetValue(tlsIndex);
+   if(!data) {
+      /* DllMain is called with DLL_THREAD_ATTACH only by threads created after 
+       * the DLL is loaded by the process */
+      
+      data = stw_tls_data_create();
+      if(!data)
+         return NULL;
 
-   return (struct stw_tls_data *) TlsGetValue(tlsIndex);
+      TlsSetValue(tlsIndex, data);
+   }
+
+   return data;
 }
diff --git a/src/gallium/state_trackers/wgl/shared/stw_tls.h b/src/gallium/state_trackers/wgl/shared/stw_tls.h
index f5a6bdf4b1a..6cfb0899f2d 100644
--- a/src/gallium/state_trackers/wgl/shared/stw_tls.h
+++ b/src/gallium/state_trackers/wgl/shared/stw_tls.h
@@ -33,8 +33,6 @@
 struct stw_tls_data
 {
    uint currentPixelFormat;
-   HDC currentDC;
-   UINT_PTR currentGLRC;
    HHOOK hCallWndProcHook;
 };
 
diff --git a/src/gallium/state_trackers/wgl/wgl/stw_wgl.c b/src/gallium/state_trackers/wgl/wgl/stw_wgl.c
index e06d2640b47..a131292f7ae 100644
--- a/src/gallium/state_trackers/wgl/wgl/stw_wgl.c
+++ b/src/gallium/state_trackers/wgl/wgl/stw_wgl.c
@@ -100,10 +100,7 @@ wglSwapLayerBuffers(
    HDC hdc,
    UINT fuPlanes )
 {
-   (void) hdc;
-   (void) fuPlanes;
-
-   return FALSE;
+   return stw_swap_layer_buffers( hdc, fuPlanes );
 }
 
 WINGDIAPI PROC APIENTRY
@@ -189,12 +186,7 @@ wglShareLists(
    HGLRC hglrc1,
    HGLRC hglrc2 )
 {
-   (void) hglrc1;
-   (void) hglrc2;
-
-   assert( 0 );
-
-   return FALSE;
+   return stw_share_lists( (UINT_PTR)hglrc1, (UINT_PTR)hglrc2);;
 }
 
 WINGDIAPI BOOL APIENTRY
diff --git a/src/gallium/state_trackers/xorg/xorg_dri2.c b/src/gallium/state_trackers/xorg/xorg_dri2.c
index d04204e1bfd..401bd39dac0 100644
--- a/src/gallium/state_trackers/xorg/xorg_dri2.c
+++ b/src/gallium/state_trackers/xorg/xorg_dri2.c
@@ -28,6 +28,7 @@
  *
  */
 
+#include "xorg-server.h"
 #include "xf86.h"
 #include "xf86_OSproc.h"
 
@@ -85,7 +86,6 @@ driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count)
 
 	    memset(&template, 0, sizeof(template));
 	    template.target = PIPE_TEXTURE_2D;
-	    template.compressed = 0;
 	    template.format = PIPE_FORMAT_S8Z24_UNORM;
 	    pf_get_block(template.format, &template.block);
 	    template.width[0] = pDraw->width;
@@ -98,7 +98,6 @@ driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count)
 	    struct pipe_texture template;
 	    memset(&template, 0, sizeof(template));
 	    template.target = PIPE_TEXTURE_2D;
-	    template.compressed = 0;
 	    template.format = PIPE_FORMAT_A8R8G8B8_UNORM;
 	    pf_get_block(template.format, &template.block);
 	    template.width[0] = pDraw->width;
diff --git a/src/gallium/state_trackers/xorg/xorg_driver.c b/src/gallium/state_trackers/xorg/xorg_driver.c
index 8a2711e70cd..45e831f0c28 100644
--- a/src/gallium/state_trackers/xorg/xorg_driver.c
+++ b/src/gallium/state_trackers/xorg/xorg_driver.c
@@ -476,7 +476,7 @@ ScreenInit(int scrnIndex, ScreenPtr pScreen, int argc, char **argv)
     }
 
     if (!ms->screen) {
-	ms->screen = drm_api_hooks.create_screen(ms->fd, ms->PciInfo->device_id);
+	ms->screen = drm_api_hooks.create_screen(ms->fd, NULL);
 
 	if (!ms->screen) {
 	    FatalError("Could not init pipe_screen\n");
diff --git a/src/gallium/state_trackers/xorg/xorg_exa.c b/src/gallium/state_trackers/xorg/xorg_exa.c
index 56c8fdccb28..79131743542 100644
--- a/src/gallium/state_trackers/xorg/xorg_exa.c
+++ b/src/gallium/state_trackers/xorg/xorg_exa.c
@@ -28,6 +28,7 @@
  *
  */
 
+#include "xorg-server.h"
 #include "xf86.h"
 #include "xorg_tracker.h"
 
@@ -36,6 +37,8 @@
 #include "pipe/p_state.h"
 #include "pipe/p_inlines.h"
 
+#include "util/u_rect.h"
+
 struct exa_entity
 {
     ExaDriverPtr pExa;
@@ -425,7 +428,6 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height,
 
 	memset(&template, 0, sizeof(template));
 	template.target = PIPE_TEXTURE_2D;
-	template.compressed = 0;
 	exa_get_pipe_format(depth, &template.format, &bitsPerPixel);
 	pf_get_block(template.format, &template.block);
 	template.width[0] = width;
@@ -436,6 +438,18 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height,
 	priv->tex = exa->scrn->texture_create(exa->scrn, &template);
     }
 
+    if (pPixData) {
+	struct pipe_transfer *transfer =
+	    exa->scrn->get_tex_transfer(exa->scrn, priv->tex, 0, 0, 0,
+					PIPE_TRANSFER_WRITE,
+					0, 0, width, height);
+        pipe_copy_rect(exa->scrn->transfer_map(exa->scrn, transfer),
+                       &priv->tex->block, transfer->stride, 0, 0,
+                       width, height, pPixData, pPixmap->devKind, 0, 0);
+        exa->scrn->transfer_unmap(exa->scrn, transfer);
+        exa->scrn->tex_transfer_destroy(transfer);
+    }
+
     return TRUE;
 }
 
diff --git a/src/gallium/winsys/drm/intel/dri2/Makefile b/src/gallium/winsys/drm/intel/dri/Makefile
similarity index 60%
rename from src/gallium/winsys/drm/intel/dri2/Makefile
rename to src/gallium/winsys/drm/intel/dri/Makefile
index 1a021092743..a4704bc0509 100644
--- a/src/gallium/winsys/drm/intel/dri2/Makefile
+++ b/src/gallium/winsys/drm/intel/dri/Makefile
@@ -4,8 +4,9 @@ include $(TOP)/configs/current
 LIBNAME = i915_dri.so
 
 PIPE_DRIVERS = \
-	$(TOP)/src/gallium/state_trackers/dri2/libdri2drm.a \
+	$(TOP)/src/gallium/state_trackers/dri/libdridrm.a \
 	$(TOP)/src/gallium/winsys/drm/intel/gem/libinteldrm.a \
+	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
 	$(TOP)/src/gallium/drivers/i915simple/libi915simple.a
 
 
@@ -19,4 +20,6 @@ include ../../Makefile.template
 
 DRI_LIB_DEPS += -ldrm_intel
 
-symlinks:
+symlinks: $(TOP)/$(LIB_DIR)/gallium
+	@rm -f $(TOP)/lib/gallium/i965_dri.so
+	ln -s i915_dri.so $(TOP)/lib/gallium/i965_dri.so
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_api.h b/src/gallium/winsys/drm/intel/gem/intel_be_api.h
index 73e458d4ba9..1c622f3b978 100644
--- a/src/gallium/winsys/drm/intel/gem/intel_be_api.h
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_api.h
@@ -8,7 +8,8 @@
 
 #include "intel_be_device.h"
 
-struct pipe_screen *intel_be_create_screen(int drmFD, int pciID);
+struct pipe_screen *intel_be_create_screen(int drmFD,
+					   struct drm_create_screen_arg *arg);
 struct pipe_context *intel_be_create_context(struct pipe_screen *screen);
 
 #endif
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_context.c b/src/gallium/winsys/drm/intel/gem/intel_be_context.c
index bb6f1b916c2..fe0b138fbea 100644
--- a/src/gallium/winsys/drm/intel/gem/intel_be_context.c
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_context.c
@@ -1,6 +1,8 @@
 
 #include "pipe/p_screen.h"
 
+#include "softpipe/sp_winsys.h"
+
 #include "intel_be_device.h"
 #include "intel_be_context.h"
 #include "intel_be_batchbuffer.h"
@@ -106,13 +108,13 @@ intel_be_create_context(struct pipe_screen *screen)
 
 	intel_be_init_context(intel, device);
 
-#if 0
-	pipe = intel_create_softpipe(intel, screen->winsys);
-#else
-	pipe = i915_create_context(screen, &device->base, &intel->base);
-#endif
+	if (device->softpipe)
+		pipe = softpipe_create(screen);
+	else
+		pipe = i915_create_context(screen, &device->base, &intel->base);
 
-	pipe->priv = intel;
+	if (pipe)
+		pipe->priv = intel;
 
 	return pipe;
 }
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.c b/src/gallium/winsys/drm/intel/gem/intel_be_device.c
index 0f6300323b0..907ac86637f 100644
--- a/src/gallium/winsys/drm/intel/gem/intel_be_device.c
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_device.c
@@ -6,12 +6,15 @@
 #include "pipe/p_state.h"
 #include "pipe/p_inlines.h"
 #include "util/u_memory.h"
+#include "util/u_debug.h"
 
 #include "intel_be_fence.h"
 
 #include "i915simple/i915_winsys.h"
+#include "softpipe/sp_winsys.h"
 
 #include "intel_be_api.h"
+#include <stdio.h>
 
 /*
  * Buffer
@@ -175,8 +178,6 @@ intel_be_handle_from_buffer(struct pipe_screen *screen,
                             struct pipe_buffer *buffer,
                             unsigned *handle)
 {
-	drm_intel_bo *bo;
-
 	if (!buffer)
 		return FALSE;
 
@@ -285,14 +286,47 @@ intel_be_init_device(struct intel_be_device *dev, int fd, unsigned id)
 
 	dev->pools.gem = drm_intel_bufmgr_gem_init(dev->fd, dev->max_batch_size);
 
+	dev->softpipe = debug_get_bool_option("INTEL_SOFTPIPE", FALSE);
+
 	return true;
 }
 
+static void
+intel_be_get_device_id(unsigned int *device_id)
+{
+   char path[512];
+   FILE *file;
+
+   /*
+    * FIXME: Fix this up to use a drm ioctl or whatever.
+    */
+
+   snprintf(path, sizeof(path), "/sys/class/drm/card0/device/device");
+   file = fopen(path, "r");
+   if (!file) {
+      return;
+   }
+
+   fgets(path, sizeof(path), file);
+   sscanf(path, "%x", device_id);
+   fclose(file);
+}
+
 struct pipe_screen *
-intel_be_create_screen(int drmFD, int deviceID)
+intel_be_create_screen(int drmFD, struct drm_create_screen_arg *arg)
 {
 	struct intel_be_device *dev;
 	struct pipe_screen *screen;
+	unsigned int deviceID;
+
+	if (arg != NULL) {
+		switch(arg->mode) {
+		case DRM_CREATE_NORMAL:
+			break;
+		default:
+			return NULL;
+		}
+	}
 
 	/* Allocate the private area */
 	dev = malloc(sizeof(*dev));
@@ -300,9 +334,14 @@ intel_be_create_screen(int drmFD, int deviceID)
 		return NULL;
 	memset(dev, 0, sizeof(*dev));
 
+	intel_be_get_device_id(&deviceID);
 	intel_be_init_device(dev, drmFD, deviceID);
 
-	screen = i915_create_screen(&dev->base, deviceID);
+	if (dev->softpipe) {
+		screen = softpipe_create_screen(&dev->base);
+		drm_api_hooks.buffer_from_texture = softpipe_get_texture_buffer;
+	} else
+		screen = i915_create_screen(&dev->base, deviceID);
 
 	return screen;
 }
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.h b/src/gallium/winsys/drm/intel/gem/intel_be_device.h
index 47d2176cb42..b32637ece29 100644
--- a/src/gallium/winsys/drm/intel/gem/intel_be_device.h
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_device.h
@@ -16,6 +16,8 @@ struct intel_be_device
 {
 	struct pipe_winsys base;
 
+	boolean softpipe;
+
 	int fd; /**< Drm file discriptor */
 
 	unsigned id;
diff --git a/src/gallium/winsys/drm/intel/xorg/Makefile b/src/gallium/winsys/drm/intel/xorg/Makefile
index a45ca570db1..b1b6b9362b6 100644
--- a/src/gallium/winsys/drm/intel/xorg/Makefile
+++ b/src/gallium/winsys/drm/intel/xorg/Makefile
@@ -21,6 +21,7 @@ LIBS = \
 	$(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a \
 	$(GALLIUMDIR)/winsys/drm/intel/gem/libinteldrm.a \
 	$(TOP)/src/gallium/drivers/i915simple/libi915simple.a \
+	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
 	$(GALLIUM_AUXILIARIES)
 
 #############################################
@@ -29,7 +30,7 @@ LIBS = \
 
 all default: $(TARGET)
 
-$(TARGET): $(OBJECTS) Makefile $(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a
+$(TARGET): $(OBJECTS) Makefile $(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a $(LIBS)
 	$(TOP)/bin/mklib -noprefix -o $@ \
 	$(OBJECTS) $(LIBS) $(shell pkg-config --libs libdrm) -ldrm_intel
 
diff --git a/src/gallium/winsys/drm/intel/xorg/intel_xorg.c b/src/gallium/winsys/drm/intel/xorg/intel_xorg.c
index aea39247e51..46a7971f609 100644
--- a/src/gallium/winsys/drm/intel/xorg/intel_xorg.c
+++ b/src/gallium/winsys/drm/intel/xorg/intel_xorg.c
@@ -37,20 +37,17 @@ static Bool intel_xorg_pci_probe(DriverPtr driver,
 				 intptr_t match_data);
 
 static const struct pci_id_match intel_xorg_device_match[] = {
-    {0x8086, 0x2592, 0xffff, 0xffff, 0, 0, 0},
-    {0x8086, 0x27A2, 0xffff, 0xffff, 0, 0, 0},
+    {0x8086, PCI_MATCH_ANY, PCI_MATCH_ANY, PCI_MATCH_ANY, 0, 0, 0},
     {0, 0, 0},
 };
 
 static SymTabRec intel_xorg_chipsets[] = {
-    {0x2592, "Intel Graphics Device"},
-    {0x27A2, "Intel Graphics Device"},
+    {PCI_MATCH_ANY, "Intel Graphics Device"},
     {-1, NULL}
 };
 
 static PciChipsets intel_xorg_pci_devices[] = {
-    {0x2592, 0x2592, RES_SHARED_VGA},
-    {0x27A2, 0x27A2, RES_SHARED_VGA},
+    {PCI_MATCH_ANY, PCI_MATCH_ANY, RES_SHARED_VGA},
     {-1, -1, RES_UNDEFINED}
 };
 
diff --git a/src/gallium/winsys/drm/nouveau/Makefile b/src/gallium/winsys/drm/nouveau/Makefile
index f8c81358544..6c9cbef26df 100644
--- a/src/gallium/winsys/drm/nouveau/Makefile
+++ b/src/gallium/winsys/drm/nouveau/Makefile
@@ -2,7 +2,7 @@
 TOP = ../../../../..
 include $(TOP)/configs/current
 
-SUBDIRS = drm dri dri2
+SUBDIRS = drm $(GALLIUM_STATE_TRACKERS_DIRS)
 
 default install clean:
 	@for dir in $(SUBDIRS) ; do \
diff --git a/src/gallium/winsys/drm/nouveau/dri/Makefile b/src/gallium/winsys/drm/nouveau/dri/Makefile
index f7db6201fea..024ab150cbd 100644
--- a/src/gallium/winsys/drm/nouveau/dri/Makefile
+++ b/src/gallium/winsys/drm/nouveau/dri/Makefile
@@ -3,9 +3,8 @@ include $(TOP)/configs/current
 
 LIBNAME = nouveau_dri.so
 
-MINIGLX_SOURCES =
-
 PIPE_DRIVERS = \
+	$(TOP)/src/gallium/state_trackers/dri/libdridrm.a \
 	$(TOP)/src/gallium/winsys/drm/nouveau/drm/libnouveaudrm.a \
 	$(TOP)/src/gallium/drivers/nv04/libnv04.a \
 	$(TOP)/src/gallium/drivers/nv10/libnv10.a \
@@ -13,22 +12,15 @@ PIPE_DRIVERS = \
 	$(TOP)/src/gallium/drivers/nv30/libnv30.a \
 	$(TOP)/src/gallium/drivers/nv40/libnv40.a \
 	$(TOP)/src/gallium/drivers/nv50/libnv50.a
-	
-DRIVER_SOURCES = \
-	nouveau_context.c \
-	nouveau_screen.c \
-	nouveau_swapbuffers.c \
-	nouveau_lock.c
+
+DRIVER_SOURCES =
 
 C_SOURCES = \
 	$(COMMON_GALLIUM_SOURCES) \
 	$(DRIVER_SOURCES)
 
-ASM_SOURCES = 
-
-DRIVER_DEFINES = $(shell pkg-config libdrm_nouveau --cflags)
-DRI_LIB_DEPS += $(shell pkg-config libdrm_nouveau --libs)
-
 include ../../Makefile.template
 
+DRI_LIB_DEPS += $(shell pkg-config libdrm_nouveau --libs)
+
 symlinks:
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_context.c
deleted file mode 100644
index deb6ffcff1c..00000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.c
+++ /dev/null
@@ -1,118 +0,0 @@
-#include <main/glheader.h>
-#include <glapi/glthread.h>
-#include <GL/internal/glcore.h>
-#include <utils.h>
-
-#include <state_tracker/st_public.h>
-#include <state_tracker/st_context.h>
-#include <state_tracker/drm_api.h>
-#include <pipe/p_defines.h>
-#include <pipe/p_context.h>
-#include <pipe/p_screen.h>
-
-#include "nouveau_context.h"
-#include "nouveau_screen.h"
-
-#include "nouveau_drmif.h"
-
-GLboolean
-nouveau_context_create(const __GLcontextModes *glVis,
-		       __DRIcontextPrivate *driContextPriv,
-		       void *sharedContextPrivate)
-{
-	__DRIscreenPrivate *driScrnPriv = driContextPriv->driScreenPriv;
-	struct nouveau_screen  *nv_screen = driScrnPriv->private;
-	struct nouveau_context *nv;
-	struct pipe_context *pipe;
-	struct st_context *st_share = NULL;
-
-	if (sharedContextPrivate)
-		st_share = ((struct nouveau_context *)sharedContextPrivate)->st;
-
-	nv = CALLOC_STRUCT(nouveau_context);
-	if (!nv)
-		return GL_FALSE;
-
-	{
-		struct nouveau_device_priv *nvdev =
-			nouveau_device(nv_screen->device);
-
-		nvdev->ctx  = driContextPriv->hHWContext;
-		nvdev->lock = (drmLock *)&driScrnPriv->pSAREA->lock;
-	}
-
-	pipe = drm_api_hooks.create_context(nv_screen->pscreen);
-	if (!pipe) {
-		FREE(nv);
-		return GL_FALSE;
-	}
-	pipe->priv = nv;
-
-	driContextPriv->driverPrivate = nv;
-	nv->dri_screen = driScrnPriv;
-
-	driParseConfigFiles(&nv->dri_option_cache, &nv_screen->option_cache,
-			    nv->dri_screen->myNum, "nouveau");
-
-	nv->st = st_create_context(pipe, glVis, st_share);
-	return GL_TRUE;
-}
-
-void
-nouveau_context_destroy(__DRIcontextPrivate *driContextPriv)
-{
-	struct nouveau_context *nv = driContextPriv->driverPrivate;
-
-	assert(nv);
-
-	st_finish(nv->st);
-	st_destroy_context(nv->st);
-
-	FREE(nv);
-}
-
-GLboolean
-nouveau_context_bind(__DRIcontextPrivate *driContextPriv,
-		     __DRIdrawablePrivate *driDrawPriv,
-		     __DRIdrawablePrivate *driReadPriv)
-{
-	struct nouveau_context *nv;
-	struct nouveau_framebuffer *draw, *read;
-
-	if (!driContextPriv) {
-		st_make_current(NULL, NULL, NULL);
-		return GL_TRUE;
-	}
-
-	nv = driContextPriv->driverPrivate;
-	draw = driDrawPriv->driverPrivate;
-	read = driReadPriv->driverPrivate;
-
-	st_make_current(nv->st, draw->stfb, read->stfb);
-
-	if ((nv->dri_drawable != driDrawPriv) ||
-	    (nv->last_stamp != driDrawPriv->lastStamp)) {
-		nv->dri_drawable = driDrawPriv;
-		st_resize_framebuffer(draw->stfb, driDrawPriv->w,
-				      driDrawPriv->h);
-		nv->last_stamp = driDrawPriv->lastStamp;
-	}
-
-	if (driDrawPriv != driReadPriv) {
-		st_resize_framebuffer(read->stfb, driReadPriv->w,
-				      driReadPriv->h);
-	}
-
-	return GL_TRUE;
-}
-
-GLboolean
-nouveau_context_unbind(__DRIcontextPrivate *driContextPriv)
-{
-	struct nouveau_context *nv = driContextPriv->driverPrivate;
-	(void)nv;
-
-	st_flush(nv->st, 0, NULL);
-	return GL_TRUE;
-}
-
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_context.h
deleted file mode 100644
index 2779b092e64..00000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.h
+++ /dev/null
@@ -1,53 +0,0 @@
-#ifndef __NOUVEAU_CONTEXT_DRI_H__
-#define __NOUVEAU_CONTEXT_DRI_H__
-
-#include <dri_util.h>
-#include <xmlconfig.h>
-
-#include "nouveau/nouveau_winsys.h"
-
-#define NOUVEAU_ERR(fmt, args...) debug_printf("%s: "fmt, __func__, ##args)
-
-struct nouveau_framebuffer {
-	struct st_framebuffer *stfb;
-};
-
-struct nouveau_context {
-	struct st_context *st;
-
-	/* DRI stuff */
-	__DRIscreenPrivate    *dri_screen;
-	__DRIdrawablePrivate  *dri_drawable;
-	unsigned int           last_stamp;
-	driOptionCache         dri_option_cache;
-	drm_context_t          drm_context;
-	drmLock                drm_lock;
-	int                    locked;
-};
-
-extern GLboolean nouveau_context_create(const __GLcontextModes *,
-					__DRIcontextPrivate *, void *);
-extern void nouveau_context_destroy(__DRIcontextPrivate *);
-extern GLboolean nouveau_context_bind(__DRIcontextPrivate *,
-				      __DRIdrawablePrivate *draw,
-				      __DRIdrawablePrivate *read);
-extern GLboolean nouveau_context_unbind(__DRIcontextPrivate *);
-
-extern void nouveau_contended_lock(struct nouveau_context *nv);
-extern void LOCK_HARDWARE(struct nouveau_context *nv);
-extern void UNLOCK_HARDWARE(struct nouveau_context *nv);
-
-#ifdef DEBUG
-extern int __nouveau_debug;
-
-#define DEBUG_BO (1 << 0)
-
-#define DBG(flag, ...) do {                   \
-	if (__nouveau_debug & (DEBUG_##flag)) \
-		NOUVEAU_ERR(__VA_ARGS__);     \
-} while(0)
-#else
-#define DBG(flag, ...)
-#endif
-
-#endif
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_lock.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_lock.c
deleted file mode 100644
index 92f5bd09c9d..00000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_lock.c
+++ /dev/null
@@ -1,73 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
-#include <pipe/p_thread.h>
-#include "nouveau_context.h"
-#include "nouveau_screen.h"
-#include "nouveau_drmif.h"
-
-pipe_static_mutex(lockMutex);
-
-/* Lock the hardware and validate our state.
- */
-void
-LOCK_HARDWARE(struct nouveau_context *nv)
-{
-	struct nouveau_screen *nv_screen = nv->dri_screen->private;
-	struct nouveau_device *dev = nv_screen->device;
-	struct nouveau_device_priv *nvdev = nouveau_device(dev);
-	char __ret=0;
-
-	assert(!nv->locked);
-	pipe_mutex_lock(lockMutex);
-
-	DRM_CAS(nvdev->lock, nvdev->ctx,
-		(DRM_LOCK_HELD | nvdev->ctx), __ret);
-
-	if (__ret) {
-		drmGetLock(nvdev->fd, nvdev->ctx, 0);
-		nouveau_contended_lock(nv);
-	}
-	nv->locked = 1;
-}
-
-/* Unlock the hardware using the global current context 
- */
-void
-UNLOCK_HARDWARE(struct nouveau_context *nv)
-{
-	struct nouveau_screen *nv_screen = nv->dri_screen->private;
-	struct nouveau_device *dev = nv_screen->device;
-	struct nouveau_device_priv *nvdev = nouveau_device(dev);
-
-	assert(nv->locked);
-	nv->locked = 0;
-
-	DRM_UNLOCK(nvdev->fd, nvdev->lock, nvdev->ctx);
-
-	pipe_mutex_unlock(lockMutex);
-} 
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.c
deleted file mode 100644
index 0b45b1ff1f6..00000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.c
+++ /dev/null
@@ -1,330 +0,0 @@
-#include <utils.h>
-#include <vblank.h>
-#include <xmlpool.h>
-
-#include <pipe/p_context.h>
-#include <state_tracker/st_public.h>
-#include <state_tracker/st_cb_fbo.h>
-#include <state_tracker/drm_api.h>
-
-#include "nouveau_context.h"
-#include "nouveau_screen.h"
-#include "nouveau_swapbuffers.h"
-#include "nouveau_dri.h"
-
-#include "nouveau_drm.h"
-#include "nouveau_drmif.h"
-
-#if NOUVEAU_DRM_HEADER_PATCHLEVEL != 12
-#error nouveau_drm.h version does not match expected version
-#endif
-
-/* Extension stuff, enabling of extensions handled by Gallium's GL state
- * tracker.  But, we still need to define the entry points we want.
- */
-#define need_GL_ARB_fragment_program
-#define need_GL_ARB_multisample
-#define need_GL_ARB_occlusion_query
-#define need_GL_ARB_point_parameters
-#define need_GL_ARB_shader_objects
-#define need_GL_ARB_texture_compression
-#define need_GL_ARB_vertex_program
-#define need_GL_ARB_vertex_shader
-#define need_GL_ARB_vertex_buffer_object
-#define need_GL_EXT_compiled_vertex_array
-#define need_GL_EXT_fog_coord
-#define need_GL_EXT_secondary_color
-#define need_GL_EXT_framebuffer_object
-#define need_GL_VERSION_2_0
-#define need_GL_VERSION_2_1
-#include "extension_helper.h"
-
-const struct dri_extension card_extensions[] =
-{
-	{ "GL_ARB_multisample", GL_ARB_multisample_functions },
-	{ "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions },
-	{ "GL_ARB_point_parameters", GL_ARB_point_parameters_functions },
-	{ "GL_ARB_shader_objects", GL_ARB_shader_objects_functions },
-	{ "GL_ARB_shading_language_100", GL_VERSION_2_0_functions },
-	{ "GL_ARB_shading_language_120", GL_VERSION_2_1_functions },
-	{ "GL_ARB_texture_compression", GL_ARB_texture_compression_functions },
-	{ "GL_ARB_vertex_program", GL_ARB_vertex_program_functions },
-	{ "GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions },
-	{ "GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions },
-	{ "GL_EXT_compiled_vertex_array", GL_EXT_compiled_vertex_array_functions },
-	{ "GL_EXT_fog_coord", GL_EXT_fog_coord_functions },
-	{ "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
-	{ "GL_EXT_secondary_color", GL_EXT_secondary_color_functions },
-	{ NULL, 0 }
-};
-
-PUBLIC const char __driConfigOptions[] =
-DRI_CONF_BEGIN
-DRI_CONF_END;
-static const GLuint __driNConfigOptions = 0;
-
-extern const struct dri_extension common_extensions[];
-extern const struct dri_extension nv40_extensions[];
-
-static GLboolean
-nouveau_create_buffer(__DRIscreenPrivate * driScrnPriv,
-		      __DRIdrawablePrivate * driDrawPriv,
-		      const __GLcontextModes *glVis, GLboolean pixmapBuffer)
-{
-	struct nouveau_framebuffer *nvfb;
-	enum pipe_format colour, depth, stencil;
-
-	if (pixmapBuffer)
-		return GL_FALSE;
-
-	nvfb = CALLOC_STRUCT(nouveau_framebuffer);
-	if (!nvfb)
-		return GL_FALSE;
-
-	if (glVis->redBits == 5)
-		colour = PIPE_FORMAT_R5G6B5_UNORM;
-	else
-		colour = PIPE_FORMAT_A8R8G8B8_UNORM;
-
-	if (glVis->depthBits == 16)
-		depth = PIPE_FORMAT_Z16_UNORM;
-	else if (glVis->depthBits == 24)
-		depth = PIPE_FORMAT_Z24S8_UNORM;
-	else
-		depth = PIPE_FORMAT_NONE;
-
-	if (glVis->stencilBits == 8)
-		stencil = PIPE_FORMAT_Z24S8_UNORM;
-	else
-		stencil = PIPE_FORMAT_NONE;
-
-	nvfb->stfb = st_create_framebuffer(glVis, colour, depth, stencil,
-					   driDrawPriv->w, driDrawPriv->h,
-					   (void*)nvfb);
-	if (!nvfb->stfb) {
-		free(nvfb);
-		return  GL_FALSE;
-	}
-
-	driDrawPriv->driverPrivate = (void *)nvfb;
-	return GL_TRUE;
-}
-
-static void
-nouveau_destroy_buffer(__DRIdrawablePrivate * driDrawPriv)
-{
-	struct nouveau_framebuffer *nvfb;
-	
-	nvfb = (struct nouveau_framebuffer *)driDrawPriv->driverPrivate;
-	st_unreference_framebuffer(nvfb->stfb);
-	free(nvfb);
-}
-
-static __DRIconfig **
-nouveau_fill_in_modes(__DRIscreenPrivate *psp,
-		      unsigned pixel_bits, unsigned depth_bits,
-		      unsigned stencil_bits, GLboolean have_back_buffer)
-{
-	__DRIconfig **configs;
-	unsigned depth_buffer_factor;
-	unsigned back_buffer_factor;
-	GLenum fb_format;
-	GLenum fb_type;
-
-	static const GLenum back_buffer_modes[] = {
-		GLX_NONE, GLX_SWAP_UNDEFINED_OML,
-	};
-
-	uint8_t depth_bits_array[3];
-	uint8_t stencil_bits_array[3];
-	uint8_t msaa_samples_array[1];
-
-	depth_bits_array[0] = 0;
-	depth_bits_array[1] = depth_bits;
-	depth_bits_array[2] = depth_bits;
-
-	/* Just like with the accumulation buffer, always provide some modes
-	 * with a stencil buffer.  It will be a sw fallback, but some apps won't
-	 * care about that.
-	 */
-	stencil_bits_array[0] = 0;
-	stencil_bits_array[1] = 0;
-	if (depth_bits == 24)
-		stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
-	stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits;
-
-	msaa_samples_array[0] = 0;
-
-	depth_buffer_factor =
-		((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1;
-	back_buffer_factor = (have_back_buffer) ? 3 : 1;
-
-	if (pixel_bits == 16) {
-		fb_format = GL_RGB;
-		fb_type = GL_UNSIGNED_SHORT_5_6_5;
-	}
-	else {
-		fb_format = GL_BGRA;
-		fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
-	}
-
-	configs = driCreateConfigs(fb_format, fb_type,
-				   depth_bits_array, stencil_bits_array,
-				   depth_buffer_factor, back_buffer_modes,
-				   back_buffer_factor, msaa_samples_array, 1);
-	if (configs == NULL) {
-	 fprintf(stderr, "[%s:%u] Error creating FBConfig!\n",
-			 __func__, __LINE__);
-		return NULL;
-	}
-
-	return configs;
-}
-
-static struct pipe_surface *
-dri_surface_from_handle(struct pipe_screen *screen,
-                        unsigned handle,
-                        enum pipe_format format,
-                        unsigned width,
-                        unsigned height,
-                        unsigned pitch)
-{
-   struct pipe_surface *surface = NULL;
-   struct pipe_texture *texture = NULL;
-   struct pipe_texture templat;
-   struct pipe_buffer *buf = NULL;
-
-   buf = drm_api_hooks.buffer_from_handle(screen, "front buffer", handle);
-   if (!buf)
-      return NULL;
-
-   memset(&templat, 0, sizeof(templat));
-   templat.tex_usage = PIPE_TEXTURE_USAGE_PRIMARY |
-                       NOUVEAU_TEXTURE_USAGE_LINEAR;
-   templat.target = PIPE_TEXTURE_2D;
-   templat.last_level = 0;
-   templat.depth[0] = 1;
-   templat.format = format;
-   templat.width[0] = width;
-   templat.height[0] = height;
-   pf_get_block(templat.format, &templat.block);
-
-   texture = screen->texture_blanket(screen,
-                                     &templat,
-                                     &pitch,
-                                     buf);
-
-   /* we don't need the buffer from this point on */
-   pipe_buffer_reference(&buf, NULL);
-
-   if (!texture)
-      return NULL;
-
-   surface = screen->get_tex_surface(screen, texture, 0, 0, 0,
-                                     PIPE_BUFFER_USAGE_GPU_READ |
-                                     PIPE_BUFFER_USAGE_GPU_WRITE);
-
-   /* we don't need the texture from this point on */
-   pipe_texture_reference(&texture, NULL);
-   return surface;
-}
-
-static const __DRIconfig **
-nouveau_screen_create(__DRIscreenPrivate *psp)
-{
-	struct nouveau_dri *nv_dri = psp->pDevPriv;
-	struct nouveau_screen *nv_screen;
-	static const __DRIversion ddx_expected =
-		{ 0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL };
-	static const __DRIversion dri_expected = { 4, 0, 0 };
-	static const __DRIversion drm_expected =
-		{ 0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL };
-
-	if (!driCheckDriDdxDrmVersions2("nouveau",
-					&psp->dri_version, &dri_expected,
-					&psp->ddx_version, &ddx_expected,
-					&psp->drm_version, &drm_expected)) {
-		return NULL;
-	}
-
-	if (drm_expected.patch != psp->drm_version.patch) {
-		fprintf(stderr, "Incompatible DRM patch level.\n"
-				"Expected: %d\n" "Current : %d\n",
-			drm_expected.patch, psp->drm_version.patch);
-		return NULL;
-	}
-
-	driInitExtensions(NULL, card_extensions, GL_FALSE);
-
-	if (psp->devPrivSize != sizeof(struct nouveau_dri)) {
-		NOUVEAU_ERR("DRI struct mismatch between DDX/DRI\n");
-		return NULL;
-	}
-
-	nv_screen = CALLOC_STRUCT(nouveau_screen);
-	if (!nv_screen)
-		return NULL;
-
-	nouveau_device_open_existing(&nv_screen->device, 0, psp->fd, 0);
-
-	nv_screen->pscreen = drm_api_hooks.create_screen(psp->fd, 0);
-	if (!nv_screen->pscreen) {
-		FREE(nv_screen);
-		return NULL;
-	}
-	nv_screen->pscreen->flush_frontbuffer = nouveau_flush_frontbuffer;
-
-	{
-		enum pipe_format format;
-
-		if (nv_dri->bpp == 16)
-			format = PIPE_FORMAT_R5G6B5_UNORM;
-		else
-			format = PIPE_FORMAT_A8R8G8B8_UNORM;
-
-		nv_screen->fb = dri_surface_from_handle(nv_screen->pscreen,
-							nv_dri->front_offset,
-							format,
-							nv_dri->width,
-							nv_dri->height,
-							nv_dri->front_pitch *
-							nv_dri->bpp / 8);
-	}
-						
-	driParseOptionInfo(&nv_screen->option_cache,
-			   __driConfigOptions, __driNConfigOptions);
-
-	nv_screen->driScrnPriv = psp;
-	psp->private = (void *)nv_screen;
-
-	return (const __DRIconfig **)
-		nouveau_fill_in_modes(psp, nv_dri->bpp,
-				      (nv_dri->bpp == 16) ? 16 : 24,
-				      (nv_dri->bpp == 16) ? 0 : 8, 1);
-}
-
-static void
-nouveau_screen_destroy(__DRIscreenPrivate *driScrnPriv)
-{
-	struct nouveau_screen *nv_screen = driScrnPriv->private;
-
-	driScrnPriv->private = NULL;
-	FREE(nv_screen);
-}
-
-const struct __DriverAPIRec
-driDriverAPI = {
-	.InitScreen	= nouveau_screen_create,
-	.DestroyScreen	= nouveau_screen_destroy,
-	.CreateContext	= nouveau_context_create,
-	.DestroyContext	= nouveau_context_destroy,
-	.CreateBuffer	= nouveau_create_buffer,
-	.DestroyBuffer	= nouveau_destroy_buffer,
-	.SwapBuffers	= nouveau_swap_buffers,
-	.MakeCurrent	= nouveau_context_bind,
-	.UnbindContext	= nouveau_context_unbind,
-	.CopySubBuffer	= nouveau_copy_sub_buffer,
-
-	.InitScreen2	= NULL, /* one day, I promise! */
-};
-
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.h
deleted file mode 100644
index ac078f3c638..00000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.h
+++ /dev/null
@@ -1,16 +0,0 @@
-#ifndef __NOUVEAU_SCREEN_DRI_H__
-#define __NOUVEAU_SCREEN_DRI_H__
-
-#include "xmlconfig.h"
-
-struct nouveau_screen {
-	__DRIscreenPrivate *driScrnPriv;
-	driOptionCache      option_cache;
-
-	struct nouveau_device *device;
-
-	struct pipe_screen *pscreen;
-	struct pipe_surface *fb;
-};
-
-#endif
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c
deleted file mode 100644
index 9c841a0b2d0..00000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c
+++ /dev/null
@@ -1,115 +0,0 @@
-#include <main/glheader.h>
-#include <glapi/glthread.h>
-#include <GL/internal/glcore.h>
-
-#include <pipe/p_context.h>
-#include <state_tracker/st_public.h>
-#include <state_tracker/st_context.h>
-#include <state_tracker/st_cb_fbo.h>
-
-#include "nouveau_context.h"
-#include "nouveau_screen.h"
-#include "nouveau_swapbuffers.h"
-
-#include "nouveau_pushbuf.h"
-
-void
-nouveau_copy_buffer(__DRIdrawablePrivate *dPriv, struct pipe_surface *surf,
-		    const drm_clip_rect_t *rect)
-{
-	struct nouveau_context *nv = dPriv->driContextPriv->driverPrivate;
-	struct nouveau_screen *nv_screen = nv->dri_screen->private;
-	struct pipe_context *pipe = nv->st->pipe;
-	drm_clip_rect_t *pbox;
-	int nbox, i;
-
-	LOCK_HARDWARE(nv);
-	if (!dPriv->numClipRects) {
-		UNLOCK_HARDWARE(nv);
-		return;
-	}
-	pbox = dPriv->pClipRects;
-	nbox = dPriv->numClipRects;
-
-	for (i = 0; i < nbox; i++, pbox++) {
-		int sx, sy, dx, dy, w, h;
-
-		sx = pbox->x1 - dPriv->x;
-		sy = pbox->y1 - dPriv->y;
-		dx = pbox->x1;
-		dy = pbox->y1;
-		w  = pbox->x2 - pbox->x1;
-		h  = pbox->y2 - pbox->y1;
-
-		pipe->surface_copy(pipe, nv_screen->fb, dx, dy, surf,
-				   sx, sy, w, h);
-	}
-
-	pipe->flush(pipe, 0, NULL);
-	UNLOCK_HARDWARE(nv);
-
-	if (nv->last_stamp != dPriv->lastStamp) {
-		struct nouveau_framebuffer *nvfb = dPriv->driverPrivate;
-		st_resize_framebuffer(nvfb->stfb, dPriv->w, dPriv->h);
-		nv->last_stamp = dPriv->lastStamp;
-	}
-}
-
-void
-nouveau_copy_sub_buffer(__DRIdrawablePrivate *dPriv, int x, int y, int w, int h)
-{
-	struct nouveau_framebuffer *nvfb = dPriv->driverPrivate;
-	struct pipe_surface *surf;
-
-	st_get_framebuffer_surface(nvfb->stfb, ST_SURFACE_BACK_LEFT, &surf);
-	if (surf) {
-		drm_clip_rect_t rect;
-		rect.x1 = x;
-		rect.y1 = y;
-		rect.x2 = x + w;
-		rect.y2 = y + h;
-
-		st_notify_swapbuffers(nvfb->stfb);
-		nouveau_copy_buffer(dPriv, surf, &rect);
-	}
-}
-
-void
-nouveau_swap_buffers(__DRIdrawablePrivate *dPriv)
-{
-	struct nouveau_framebuffer *nvfb = dPriv->driverPrivate;
-	struct pipe_surface *surf;
-
-	st_get_framebuffer_surface(nvfb->stfb, ST_SURFACE_BACK_LEFT, &surf);
-	if (surf) {
-		st_notify_swapbuffers(nvfb->stfb);
-		nouveau_copy_buffer(dPriv, surf, NULL);
-	}
-}
-
-void
-nouveau_flush_frontbuffer(struct pipe_screen *pscreen, struct pipe_surface *ps,
-			  void *context_private)
-{
-	struct nouveau_context *nv = context_private;
-	__DRIdrawablePrivate *dPriv = nv->dri_drawable;
-
-	nouveau_copy_buffer(dPriv, ps, NULL);
-}
-
-void
-nouveau_contended_lock(struct nouveau_context *nv)
-{
-	struct nouveau_context *nv_sub = (struct nouveau_context*)nv;
-	__DRIdrawablePrivate *dPriv = nv_sub->dri_drawable;
-	__DRIscreenPrivate *sPriv = nv_sub->dri_screen;
-
-	/* If the window moved, may need to set a new cliprect now.
-	 *
-	 * NOTE: This releases and regains the hw lock, so all state
-	 * checking must be done *after* this call:
-	 */
-	if (dPriv)
-		DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv);
-}
-
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h
deleted file mode 100644
index 4ca9cc22831..00000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h
+++ /dev/null
@@ -1,11 +0,0 @@
-#ifndef __NOUVEAU_SWAPBUFFERS_H__
-#define __NOUVEAU_SWAPBUFFERS_H__
-
-void nouveau_copy_buffer(__DRIdrawablePrivate *, struct pipe_surface *,
-			 const drm_clip_rect_t *);
-void nouveau_copy_sub_buffer(__DRIdrawablePrivate *, int x, int y, int w, int h);
-void nouveau_swap_buffers(__DRIdrawablePrivate *);
-void nouveau_flush_frontbuffer(struct pipe_screen *, struct pipe_surface *,
-			       void *context_private);
-
-#endif
diff --git a/src/gallium/winsys/drm/nouveau/dri2/Makefile b/src/gallium/winsys/drm/nouveau/dri2/Makefile
deleted file mode 100644
index 728870d2e1d..00000000000
--- a/src/gallium/winsys/drm/nouveau/dri2/Makefile
+++ /dev/null
@@ -1,26 +0,0 @@
-TOP = ../../../../../..
-include $(TOP)/configs/current
-
-LIBNAME = nouveau_dri2.so
-
-PIPE_DRIVERS = \
-	$(TOP)/src/gallium/state_trackers/dri2/libdri2drm.a \
-	$(TOP)/src/gallium/winsys/drm/nouveau/drm/libnouveaudrm.a \
-	$(TOP)/src/gallium/drivers/nv04/libnv04.a \
-	$(TOP)/src/gallium/drivers/nv10/libnv10.a \
-	$(TOP)/src/gallium/drivers/nv20/libnv20.a \
-	$(TOP)/src/gallium/drivers/nv30/libnv30.a \
-	$(TOP)/src/gallium/drivers/nv40/libnv40.a \
-	$(TOP)/src/gallium/drivers/nv50/libnv50.a
-
-DRIVER_SOURCES =
-
-C_SOURCES = \
-	$(COMMON_GALLIUM_SOURCES) \
-	$(DRIVER_SOURCES)
-
-include ../../Makefile.template
-
-DRI_LIB_DEPS += $(shell pkg-config libdrm_nouveau --libs)
-
-symlinks:
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_dri.h b/src/gallium/winsys/drm/nouveau/drm/nouveau_dri.h
similarity index 100%
rename from src/gallium/winsys/drm/nouveau/dri/nouveau_dri.h
rename to src/gallium/winsys/drm/nouveau/drm/nouveau_dri.h
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
index c0127e803f1..b355a1391d3 100644
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
+++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
@@ -7,9 +7,68 @@
 #include "nouveau_channel.h"
 #include "nouveau_bo.h"
 
-static struct pipe_screen *
-nouveau_drm_create_screen(int fd, int pciid)
+static struct pipe_surface *
+dri_surface_from_handle(struct pipe_screen *screen,
+                        unsigned handle,
+                        enum pipe_format format,
+                        unsigned width,
+                        unsigned height,
+                        unsigned pitch)
 {
+   struct pipe_surface *surface = NULL;
+   struct pipe_texture *texture = NULL;
+   struct pipe_texture templat;
+   struct pipe_buffer *buf = NULL;
+
+   buf = drm_api_hooks.buffer_from_handle(screen, "front buffer", handle);
+   if (!buf)
+      return NULL;
+
+   memset(&templat, 0, sizeof(templat));
+   templat.tex_usage = PIPE_TEXTURE_USAGE_PRIMARY |
+                       NOUVEAU_TEXTURE_USAGE_LINEAR;
+   templat.target = PIPE_TEXTURE_2D;
+   templat.last_level = 0;
+   templat.depth[0] = 1;
+   templat.format = format;
+   templat.width[0] = width;
+   templat.height[0] = height;
+   pf_get_block(templat.format, &templat.block);
+
+   texture = screen->texture_blanket(screen,
+                                     &templat,
+                                     &pitch,
+                                     buf);
+
+   /* we don't need the buffer from this point on */
+   pipe_buffer_reference(&buf, NULL);
+
+   if (!texture)
+      return NULL;
+
+   surface = screen->get_tex_surface(screen, texture, 0, 0, 0,
+                                     PIPE_BUFFER_USAGE_GPU_READ |
+                                     PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   /* we don't need the texture from this point on */
+   pipe_texture_reference(&texture, NULL);
+   return surface;
+}
+
+static struct pipe_surface *
+nouveau_dri1_front_surface(struct pipe_context *pipe)
+{
+	return nouveau_screen(pipe->screen)->front;
+}
+
+static struct dri1_api nouveau_dri1_api = {
+	nouveau_dri1_front_surface,
+};
+
+static struct pipe_screen *
+nouveau_drm_create_screen(int fd, struct drm_create_screen_arg *arg)
+{
+	struct dri1_create_screen_arg *dri1 = (void *)arg;
 	struct pipe_winsys *ws;
 	struct nouveau_winsys *nvws;
 	struct nouveau_device *dev = NULL;
@@ -67,6 +126,33 @@ nouveau_drm_create_screen(int fd, int pciid)
 		return NULL;
 	}
 
+	if (arg->mode == DRM_CREATE_DRI1) {
+		struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws);
+		struct nouveau_dri *nvdri = dri1->ddx_info;
+		enum pipe_format format;
+
+		if (nvdri->bpp == 16)
+			format = PIPE_FORMAT_R5G6B5_UNORM;
+		else
+			format = PIPE_FORMAT_A8R8G8B8_UNORM;
+
+		nvpws->front = dri_surface_from_handle(nvpws->pscreen,
+						       nvdri->front_offset,
+						       format,
+						       nvdri->width,
+						       nvdri->height,
+						       nvdri->front_pitch *
+						       (nvdri->bpp / 8));
+		if (!nvpws->front) {
+			debug_printf("%s: error referencing front buffer\n",
+				     __func__);
+			ws->destroy(ws);
+			return NULL;
+		}
+
+		dri1->api = &nouveau_dri1_api;
+	}
+
 	return nouveau_pipe_winsys(ws)->pscreen;
 }
 
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h
index 2782c83c0e7..cc237bfc13a 100644
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h
+++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h
@@ -1,5 +1,7 @@
 #ifndef __NOUVEAU_DRM_API_H__
 #define __NOUVEAU_DRM_API_H__
 #include "state_tracker/drm_api.h"
+#include "state_tracker/dri1_api.h"
+#include "nouveau_dri.h"
 
 #endif
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h
index 10e1e269e8c..ec10f1e00c8 100644
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h
+++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h
@@ -29,6 +29,8 @@ struct nouveau_pipe_winsys {
 
 	unsigned nr_pctx;
 	struct pipe_context **pctx;
+
+	struct pipe_surface *front;
 };
 
 static INLINE struct nouveau_pipe_winsys *
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
index 611ee68da63..a15487352b2 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
@@ -68,8 +68,8 @@ static struct pipe_buffer *radeon_buffer_create(struct pipe_winsys *ws,
         domain |= RADEON_GEM_DOMAIN_GTT;
     }
 
-    radeon_buffer->bo = radeon_bo_open(radeon_ws->bom, 0, size, alignment,
-                                       domain, 0);
+    radeon_buffer->bo = radeon_bo_open(radeon_ws->priv->bom, 0, size,
+            alignment, domain, 0);
     if (radeon_buffer->bo == NULL) {
         FREE(radeon_buffer);
     }
@@ -169,8 +169,13 @@ struct radeon_winsys* radeon_pipe_winsys(int fd)
         return NULL;
     }
 
-    bom = radeon_bo_manager_gem_ctor(fd);
-    radeon_ws->bom = bom;
+    radeon_ws->priv = CALLOC_STRUCT(radeon_winsys_priv);
+    if (radeon_ws->priv == NULL) {
+        FREE(radeon_ws);
+        return NULL;
+    }
+
+    radeon_ws->priv->bom = radeon_bo_manager_gem_ctor(fd);
 
     radeon_ws->base.flush_frontbuffer = radeon_flush_frontbuffer;
 
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h
index 163422f2969..ca8bbb3c117 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h
+++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h
@@ -41,6 +41,7 @@
 #include "util/u_memory.h"
 
 #include "radeon_bo.h"
+#include "radeon_cs.h"
 
 #include "radeon_drm.h"
 
@@ -49,13 +50,30 @@ struct radeon_pipe_buffer {
     struct radeon_bo    *bo;
 };
 
+#define RADEON_MAX_BOS 24
+
+struct radeon_winsys_priv {
+    /* Radeon BO manager. */
+    struct radeon_bo_manager* bom;
+
+    /* Radeon BO space checker. */
+    struct radeon_cs_space_check sc[RADEON_MAX_BOS];
+    /* Current BO count. */
+    unsigned bo_count;
+
+    /* Radeon CS manager. */
+    struct radeon_cs_manager* csm;
+
+    /* Current CS. */
+    struct radeon_cs* cs;
+};
+
 struct radeon_winsys {
     /* Parent class. */
     struct pipe_winsys base;
 
-    /* Radeon BO manager.
-     * This corresponds to void* radeon_winsys in r300_winsys. */
-    struct radeon_bo_manager* bom;
+    /* This corresponds to void* radeon_winsys in r300_winsys. */
+    struct radeon_winsys_priv* priv;
 };
 
 struct radeon_winsys* radeon_pipe_winsys(int fb);
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.c b/src/gallium/winsys/drm/radeon/core/radeon_drm.c
index 3446654e287..428d3f65a16 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_drm.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.c
@@ -31,7 +31,8 @@
 #include "radeon_drm.h"
 
 /* Create a pipe_screen. */
-struct pipe_screen* radeon_create_screen(int drmFB, int pciID)
+struct pipe_screen* radeon_create_screen(int drmFB,
+					 struct drm_create_screen_arg *arg)
 {
     struct radeon_winsys* winsys = radeon_pipe_winsys(drmFB);
 
@@ -68,7 +69,7 @@ struct pipe_buffer* radeon_buffer_from_handle(struct pipe_screen* screen,
                                               unsigned handle)
 {
     struct radeon_bo_manager* bom =
-        ((struct radeon_winsys*)screen->winsys)->bom;
+        ((struct radeon_winsys*)screen->winsys)->priv->bom;
     struct radeon_pipe_buffer* radeon_buffer;
     struct radeon_bo* bo = NULL;
 
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.h b/src/gallium/winsys/drm/radeon/core/radeon_drm.h
index ca2d98ed1a1..049f9984dba 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_drm.h
+++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.h
@@ -40,7 +40,8 @@
 #include "radeon_r300.h"
 #include "radeon_winsys_softpipe.h"
 
-struct pipe_screen* radeon_create_screen(int drmFB, int pciID);
+struct pipe_screen* radeon_create_screen(int drmFB,
+					 struct drm_create_screen_arg *arg);
 
 struct pipe_context* radeon_create_context(struct pipe_screen* screen);
 
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.c b/src/gallium/winsys/drm/radeon/core/radeon_r300.c
index 293b6c2d389..be70ead68d0 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_r300.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.c
@@ -22,86 +22,206 @@
 
 #include "radeon_r300.h"
 
-static boolean radeon_r300_check_cs(struct radeon_cs* cs, int size)
+static void radeon_r300_add_buffer(struct r300_winsys* winsys,
+                                   struct pipe_buffer* pbuffer,
+                                   uint32_t rd,
+                                   uint32_t wd)
+{
+    int i;
+    struct radeon_winsys_priv* priv =
+        (struct radeon_winsys_priv*)winsys->radeon_winsys;
+    struct radeon_cs_space_check* sc = priv->sc;
+    struct radeon_bo* bo = ((struct radeon_pipe_buffer*)pbuffer)->bo;
+
+    /* Check to see if this BO is already in line for validation;
+     * find a slot for it otherwise. */
+    assert(priv->bo_count <= RADEON_MAX_BOS);
+    for (i = 0; i < priv->bo_count; i++) {
+        if (sc[i].bo == bo) {
+            sc[i].read_domains |= rd;
+            sc[i].write_domain |= wd;
+            return;
+        }
+    }
+
+    sc[priv->bo_count].bo = bo;
+    sc[priv->bo_count].read_domains = rd;
+    sc[priv->bo_count].write_domain = wd;
+    priv->bo_count++;
+}
+
+static boolean radeon_r300_validate(struct r300_winsys* winsys)
+{
+    int retval, i;
+    struct radeon_winsys_priv* priv =
+        (struct radeon_winsys_priv*)winsys->radeon_winsys;
+    struct radeon_cs_space_check* sc = priv->sc;
+
+    debug_printf("Validation count: %d\n", priv->bo_count);
+    for (i = 0; i < priv->bo_count; i++) {
+        debug_printf("BO %d: %p rd: %d wd: %d\n", i, sc[i].bo, sc[i].read_domains, sc[i].write_domain);
+    }
+    retval = radeon_cs_space_check(priv->cs, sc, priv->bo_count);
+
+    if (retval == RADEON_CS_SPACE_OP_TO_BIG) {
+        /* We might as well HCF, since this is not going to fit in the card,
+         * period. */
+        /* XXX just drop it on the floor instead */
+	exit(1);
+    } else if (retval == RADEON_CS_SPACE_FLUSH) {
+        /* We must flush before more rendering can commence. */
+        return TRUE;
+    }
+
+    /* Things are fine, we can proceed as normal. */
+    return FALSE;
+}
+
+static boolean radeon_r300_check_cs(struct r300_winsys* winsys, int size)
 {
     /* XXX check size here, lazy ass! */
+    /* XXX also validate buffers */
     return TRUE;
 }
 
-static void radeon_r300_write_cs_reloc(struct radeon_cs* cs,
-                                    struct pipe_buffer* pbuffer,
-                                    uint32_t rd,
-                                    uint32_t wd,
-                                    uint32_t flags)
+static void radeon_r300_begin_cs(struct r300_winsys* winsys,
+                                 int size,
+                                 const char* file,
+                                 const char* function,
+                                 int line)
 {
-    radeon_cs_write_reloc(cs, ((struct radeon_pipe_buffer*)pbuffer)->bo, rd, wd, flags);
+    struct radeon_winsys_priv* priv =
+        (struct radeon_winsys_priv*)winsys->radeon_winsys;
+
+    radeon_cs_begin(priv->cs, size, file, function, line);
 }
 
-static void radeon_r300_flush_cs(struct radeon_cs* cs)
+static void radeon_r300_write_cs_dword(struct r300_winsys* winsys,
+                                       uint32_t dword)
 {
-    int retval = 0;
+    struct radeon_winsys_priv* priv =
+        (struct radeon_winsys_priv*)winsys->radeon_winsys;
 
-    retval = radeon_cs_emit(cs);
+    radeon_cs_write_dword(priv->cs, dword);
+}
+
+static void radeon_r300_write_cs_reloc(struct r300_winsys* winsys,
+                                       struct pipe_buffer* pbuffer,
+                                       uint32_t rd,
+                                       uint32_t wd,
+                                       uint32_t flags)
+{
+    struct radeon_winsys_priv* priv =
+        (struct radeon_winsys_priv*)winsys->radeon_winsys;
+
+    radeon_cs_write_reloc(priv->cs,
+            ((struct radeon_pipe_buffer*)pbuffer)->bo, rd, wd, flags);
+}
+
+static void radeon_r300_end_cs(struct r300_winsys* winsys,
+                               const char* file,
+                               const char* function,
+                               int line)
+{
+    struct radeon_winsys_priv* priv =
+        (struct radeon_winsys_priv*)winsys->radeon_winsys;
+
+    radeon_cs_end(priv->cs, file, function, line);
+}
+
+static void radeon_r300_flush_cs(struct r300_winsys* winsys)
+{
+    struct radeon_winsys_priv* priv =
+        (struct radeon_winsys_priv*)winsys->radeon_winsys;
+    struct radeon_cs_space_check* sc = priv->sc;
+    int retval = 1;
+
+    /* Emit the CS. */
+    retval = radeon_cs_emit(priv->cs);
     if (retval) {
         debug_printf("radeon: Bad CS, dumping...\n");
-        radeon_cs_print(cs, stderr);
+        radeon_cs_print(priv->cs, stderr);
     }
-    radeon_cs_erase(cs);
+    radeon_cs_erase(priv->cs);
+
+    /* Clean out BOs. */
+    memset(sc, 0, sizeof(struct radeon_cs_space_check) * RADEON_MAX_BOS);
+    priv->bo_count = 0;
 }
 
 /* Helper function to do the ioctls needed for setup and init. */
 static void do_ioctls(struct r300_winsys* winsys, int fd)
 {
-    drm_radeon_getparam_t gp;
-    uint32_t target;
+    struct drm_radeon_gem_info info = {0};
+    drm_radeon_getparam_t gp = {0};
+    int target = 0;
     int retval;
 
-    /* XXX is this cast safe? */
-    gp.value = (int*)&target;
+    gp.value = &target;
 
-    /* First, get PCI ID */
-    gp.param = RADEON_PARAM_DEVICE_ID;
-    retval = drmCommandWriteRead(fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp));
-    if (retval) {
-        fprintf(stderr, "%s: Failed to get PCI ID, error number %d",
-                __FUNCTION__, retval);
-        exit(1);
-    }
-    winsys->pci_id = target;
-
-    /* Then, get the number of pixel pipes */
+    /* First, get the number of pixel pipes */
     gp.param = RADEON_PARAM_NUM_GB_PIPES;
     retval = drmCommandWriteRead(fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp));
     if (retval) {
-        fprintf(stderr, "%s: Failed to get GB pipe count, error number %d",
+        fprintf(stderr, "%s: Failed to get GB pipe count, error number %d\n",
                 __FUNCTION__, retval);
         exit(1);
     }
     winsys->gb_pipes = target;
 
+    /* Then, get PCI ID */
+    gp.param = RADEON_PARAM_DEVICE_ID;
+    retval = drmCommandWriteRead(fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp));
+    if (retval) {
+        fprintf(stderr, "%s: Failed to get PCI ID, error number %d\n",
+                __FUNCTION__, retval);
+        exit(1);
+    }
+    winsys->pci_id = target;
+
+    /* Finally, retrieve MM info */
+    retval = drmCommandWriteRead(fd, DRM_RADEON_GEM_INFO,
+            &info, sizeof(info));
+    if (retval) {
+        fprintf(stderr, "%s: Failed to get MM info, error number %d\n",
+                __FUNCTION__, retval);
+        exit(1);
+    }
+    winsys->gart_size = info.gart_size;
+    /* XXX */
+    winsys->vram_size = info.vram_visible;
 }
 
 struct r300_winsys*
 radeon_create_r300_winsys(int fd, struct radeon_winsys* old_winsys)
 {
     struct r300_winsys* winsys = CALLOC_STRUCT(r300_winsys);
-    struct radeon_cs_manager* csm;
+    struct radeon_winsys_priv* priv;
 
     if (winsys == NULL) {
         return NULL;
     }
 
+    priv = old_winsys->priv;
+
     do_ioctls(winsys, fd);
 
-    csm = radeon_cs_manager_gem_ctor(fd);
+    priv->csm = radeon_cs_manager_gem_ctor(fd);
 
-    winsys->cs = radeon_cs_create(csm, 1024 * 64 / 4);
+    priv->cs = radeon_cs_create(priv->csm, 1024 * 64 / 4);
+    radeon_cs_set_limit(priv->cs,
+            RADEON_GEM_DOMAIN_GTT, winsys->gart_size);
+    radeon_cs_set_limit(priv->cs,
+            RADEON_GEM_DOMAIN_VRAM, winsys->vram_size);
+
+    winsys->add_buffer = radeon_r300_add_buffer;
+    winsys->validate = radeon_r300_validate;
 
     winsys->check_cs = radeon_r300_check_cs;
-    winsys->begin_cs = radeon_cs_begin;
-    winsys->write_cs_dword = radeon_cs_write_dword;
+    winsys->begin_cs = radeon_r300_begin_cs;
+    winsys->write_cs_dword = radeon_r300_write_cs_dword;
     winsys->write_cs_reloc = radeon_r300_write_cs_reloc;
-    winsys->end_cs = radeon_cs_end;
+    winsys->end_cs = radeon_r300_end_cs;
     winsys->flush_cs = radeon_r300_flush_cs;
 
     memcpy(winsys, old_winsys, sizeof(struct radeon_winsys));
diff --git a/src/gallium/winsys/drm/radeon/dri2/Makefile b/src/gallium/winsys/drm/radeon/dri/Makefile
similarity index 88%
rename from src/gallium/winsys/drm/radeon/dri2/Makefile
rename to src/gallium/winsys/drm/radeon/dri/Makefile
index f471c44349e..c218ee9d010 100644
--- a/src/gallium/winsys/drm/radeon/dri2/Makefile
+++ b/src/gallium/winsys/drm/radeon/dri/Makefile
@@ -7,7 +7,7 @@ LIBNAME = radeon_dri.so
 MINIGLX_SOURCES =
 
 PIPE_DRIVERS = \
-	$(TOP)/src/gallium/state_trackers/dri2/libdri2drm.a \
+	$(TOP)/src/gallium/state_trackers/dri/libdridrm.a \
 	$(TOP)/src/gallium/winsys/drm/radeon/core/libradeonwinsys.a \
 	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
 	$(TOP)/src/gallium/drivers/r300/libr300.a
diff --git a/src/gallium/winsys/drm/radeon/dri2/SConscript b/src/gallium/winsys/drm/radeon/dri/SConscript
similarity index 100%
rename from src/gallium/winsys/drm/radeon/dri2/SConscript
rename to src/gallium/winsys/drm/radeon/dri/SConscript
diff --git a/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c b/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c
index 6f77fbe5dee..36824251f0c 100644
--- a/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c
+++ b/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c
@@ -38,20 +38,17 @@ static Bool radeon_xorg_pci_probe(DriverPtr driver,
 				 intptr_t match_data);
 
 static const struct pci_id_match radeon_xorg_device_match[] = {
-    {0x1002, 0x5E4D, 0xffff, 0xffff, 0, 0, 0},
-    {0x1002, 0x7249, 0xffff, 0xffff, 0, 0, 0},
+    {0x1002, PCI_MATCH_ANY, PCI_MATCH_ANY, PCI_MATCH_ANY, 0, 0, 0},
     {0, 0, 0},
 };
 
 static SymTabRec radeon_xorg_chipsets[] = {
-    {0x5E4D, "Radeon RV410 PCIE (X700)"},
-    {0x7249, "Radeon R580 PCIE (X1900 XT)"},
+    {PCI_MATCH_ANY, "ATI/AMD Radeon Graphics Chipset"},
     {-1, NULL}
 };
 
 static PciChipsets radeon_xorg_pci_devices[] = {
-    {0x5E4D, 0x5E4D, RES_SHARED_VGA},
-    {0x7249, 0x7249, RES_SHARED_VGA},
+    {PCI_MATCH_ANY, PCI_MATCH_ANY, RES_SHARED_VGA},
     {-1, -1, RES_UNDEFINED}
 };
 
diff --git a/src/gallium/winsys/egl_xlib/Makefile b/src/gallium/winsys/egl_xlib/Makefile
index 02ac47caa4e..8646ee3b527 100644
--- a/src/gallium/winsys/egl_xlib/Makefile
+++ b/src/gallium/winsys/egl_xlib/Makefile
@@ -60,7 +60,7 @@ $(TOP)/$(LIB_DIR)/$(DRIVER_NAME): $(WINSYS_OBJECTS) $(LIBS)
 		-noprefix \
 		-install $(TOP)/$(LIB_DIR) \
 		$(MKLIB_OPTIONS) $(WINSYS_OBJECTS) \
-		--whole-archive $(LIBS) --no-whole-archive
+		-Wl,--whole-archive $(LIBS) -Wl,--no-whole-archive
 
 
 depend: $(ALL_SOURCES)
diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c
index b52f427e4a7..c10e3c00ffc 100644
--- a/src/gallium/winsys/egl_xlib/egl_xlib.c
+++ b/src/gallium/winsys/egl_xlib/egl_xlib.c
@@ -588,7 +588,9 @@ find_supported_apis(void)
    EGLint mask = 0;
    void *handle;
 
-   handle = dlopen(NULL, 0);
+   handle = dlopen(NULL, RTLD_LAZY | RTLD_LOCAL);
+   if(!handle)
+      return mask;
 
    if (dlsym(handle, "st_api_OpenGL_ES1"))
       mask |= EGL_OPENGL_ES_BIT;
diff --git a/src/gallium/winsys/xlib/Makefile b/src/gallium/winsys/xlib/Makefile
index bb187cc14a5..04309e67eea 100644
--- a/src/gallium/winsys/xlib/Makefile
+++ b/src/gallium/winsys/xlib/Makefile
@@ -74,7 +74,7 @@ $(TOP)/$(LIB_DIR)/gallium/$(GL_LIB_NAME): $(XLIB_WINSYS_OBJECTS) $(LIBS) Makefil
 		-major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \
 		-install $(TOP)/$(LIB_DIR)/gallium \
 		$(MKLIB_OPTIONS) $(XLIB_WINSYS_OBJECTS) \
-		--start-group $(LIBS) --end-group $(GL_LIB_DEPS)
+		-Wl,--start-group $(LIBS) -Wl,--end-group $(GL_LIB_DEPS)
 
 
 depend: $(XLIB_WINSYS_SOURCES)
diff --git a/src/glu/Makefile b/src/glu/Makefile
index e519dfeec49..5c26ead1bb9 100644
--- a/src/glu/Makefile
+++ b/src/glu/Makefile
@@ -22,7 +22,8 @@ pcedit = sed \
 	-e 's,@GLU_PC_REQ@,$(GLU_PC_REQ),' \
 	-e 's,@GLU_PC_REQ_PRIV@,$(GLU_PC_REQ_PRIV),' \
 	-e 's,@GLU_PC_LIB_PRIV@,$(GLU_PC_LIB_PRIV),' \
-	-e 's,@GLU_PC_CFLAGS@,$(GLU_PC_CFLAGS),'
+	-e 's,@GLU_PC_CFLAGS@,$(GLU_PC_CFLAGS),' \
+	-e 's,@GLU_LIB@,$(GLU_LIB),'
 glu.pc: glu.pc.in
 	$(pcedit) $< > $@
 
diff --git a/src/glu/glu.pc.in b/src/glu/glu.pc.in
index bc2517e90ed..f7d91098232 100644
--- a/src/glu/glu.pc.in
+++ b/src/glu/glu.pc.in
@@ -8,6 +8,6 @@ Description: Mesa OpenGL Utility library
 Requires: @GLU_PC_REQ@
 Requires.private: @GLU_PC_REQ_PRIV@
 Version: @VERSION@
-Libs: -L${libdir} -lGLU
+Libs: -L${libdir} -l@GLU_LIB@
 Libs.private: @GLU_PC_LIB_PRIV@
 Cflags: -I${includedir} @GLU_PC_CFLAGS@
diff --git a/src/glu/sgi/glu.exports.darwin b/src/glu/sgi/glu.exports.darwin
index 62d20ae1fbc..4f56e36f3b8 100644
--- a/src/glu/sgi/glu.exports.darwin
+++ b/src/glu/sgi/glu.exports.darwin
@@ -1,59 +1,59 @@
-_gluBeginCurve
-_gluBeginPolygon
-_gluBeginSurface
-_gluBeginTrim
-_gluBuild1DMipmapLevels
-_gluBuild1DMipmaps
-_gluBuild2DMipmapLevels
-_gluBuild2DMipmaps
-_gluBuild3DMipmapLevels
-_gluBuild3DMipmaps
-_gluCheckExtension
-_gluCylinder
-_gluDeleteNurbsRenderer
-_gluDeleteQuadric
-_gluDeleteTess
-_gluDisk
-_gluEndCurve
-_gluEndPolygon
-_gluEndSurface
-_gluEndTrim
-_gluErrorString
-_gluGetNurbsProperty
-_gluGetString
-_gluGetTessProperty
-_gluLoadSamplingMatrices
-_gluLookAt
-_gluNewNurbsRenderer
-_gluNewQuadric
-_gluNewTess
-_gluNextContour
-_gluNurbsCallback
-_gluNurbsCallbackData
-_gluNurbsCallbackDataEXT
-_gluNurbsCurve
-_gluNurbsProperty
-_gluNurbsSurface
-_gluOrtho2D
-_gluPartialDisk
-_gluPerspective
-_gluPickMatrix
-_gluProject
-_gluPwlCurve
-_gluQuadricCallback
-_gluQuadricDrawStyle
-_gluQuadricNormals
-_gluQuadricOrientation
-_gluQuadricTexture
-_gluScaleImage
-_gluSphere
-_gluTessBeginContour
-_gluTessBeginPolygon
-_gluTessCallback
-_gluTessEndContour
-_gluTessEndPolygon
-_gluTessNormal
-_gluTessProperty
-_gluTessVertex
-_gluUnProject
-_gluUnProject4
+_*gluBeginCurve
+_*gluBeginPolygon
+_*gluBeginSurface
+_*gluBeginTrim
+_*gluBuild1DMipmapLevels
+_*gluBuild1DMipmaps
+_*gluBuild2DMipmapLevels
+_*gluBuild2DMipmaps
+_*gluBuild3DMipmapLevels
+_*gluBuild3DMipmaps
+_*gluCheckExtension
+_*gluCylinder
+_*gluDeleteNurbsRenderer
+_*gluDeleteQuadric
+_*gluDeleteTess
+_*gluDisk
+_*gluEndCurve
+_*gluEndPolygon
+_*gluEndSurface
+_*gluEndTrim
+_*gluErrorString
+_*gluGetNurbsProperty
+_*gluGetString
+_*gluGetTessProperty
+_*gluLoadSamplingMatrices
+_*gluLookAt
+_*gluNewNurbsRenderer
+_*gluNewQuadric
+_*gluNewTess
+_*gluNextContour
+_*gluNurbsCallback
+_*gluNurbsCallbackData
+_*gluNurbsCallbackDataEXT
+_*gluNurbsCurve
+_*gluNurbsProperty
+_*gluNurbsSurface
+_*gluOrtho2D
+_*gluPartialDisk
+_*gluPerspective
+_*gluPickMatrix
+_*gluProject
+_*gluPwlCurve
+_*gluQuadricCallback
+_*gluQuadricDrawStyle
+_*gluQuadricNormals
+_*gluQuadricOrientation
+_*gluQuadricTexture
+_*gluScaleImage
+_*gluSphere
+_*gluTessBeginContour
+_*gluTessBeginPolygon
+_*gluTessCallback
+_*gluTessEndContour
+_*gluTessEndPolygon
+_*gluTessNormal
+_*gluTessProperty
+_*gluTessVertex
+_*gluUnProject
+_*gluUnProject4
diff --git a/src/glut/glx/Makefile b/src/glut/glx/Makefile
index 9a975bbc60d..1b072906c74 100644
--- a/src/glut/glx/Makefile
+++ b/src/glut/glx/Makefile
@@ -107,7 +107,8 @@ pcedit = sed \
 	-e 's,@VERSION@,$(GLUT_MAJOR).$(GLUT_MINOR).$(GLUT_TINY),' \
 	-e 's,@GLUT_PC_REQ_PRIV@,$(GLUT_PC_REQ_PRIV),' \
 	-e 's,@GLUT_PC_LIB_PRIV@,$(GLUT_PC_LIB_PRIV),' \
-	-e 's,@GLUT_PC_CFLAGS@,$(GLUT_PC_CFLAGS),'
+	-e 's,@GLUT_PC_CFLAGS@,$(GLUT_PC_CFLAGS),' \
+	-e 's,@GLUT_LIB@,$(GLUT_LIB),'
 glut.pc: glut.pc.in
 	$(pcedit) $< > $@
 
diff --git a/src/glut/glx/glut.pc.in b/src/glut/glx/glut.pc.in
index ae0689d7e81..151dd0b802b 100644
--- a/src/glut/glx/glut.pc.in
+++ b/src/glut/glx/glut.pc.in
@@ -8,6 +8,6 @@ Description: Mesa OpenGL Utility Toolkit library
 Requires: gl glu
 Requires.private: @GLUT_PC_REQ_PRIV@
 Version: @VERSION@
-Libs: -L${libdir} -lglut
+Libs: -L${libdir} -l@GLUT_LIB@
 Libs.private: @GLUT_PC_LIB_PRIV@
 Cflags: -I${includedir} @GLUT_PC_CFLAGS@
diff --git a/src/glut/mini/Makefile b/src/glut/mini/Makefile
index 841a473f891..0e42436133d 100644
--- a/src/glut/mini/Makefile
+++ b/src/glut/mini/Makefile
@@ -81,7 +81,8 @@ pcedit = sed \
 	-e 's,@VERSION@,$(GLUT_MAJOR).$(GLUT_MINOR).$(GLUT_TINY),' \
 	-e 's,@GLUT_PC_REQ_PRIV@,$(GLUT_PC_REQ_PRIV),' \
 	-e 's,@GLUT_PC_LIB_PRIV@,$(GLUT_PC_LIB_PRIV),' \
-	-e 's,@GLUT_PC_CFLAGS@,$(GLUT_PC_CFLAGS),'
+	-e 's,@GLUT_PC_CFLAGS@,$(GLUT_PC_CFLAGS),' \
+	-e 's,@GLUT_LIB@,$(GLUT_LIB),'
 glut.pc: glut.pc.in
 	$(pcedit) $< > $@
 
diff --git a/src/glut/mini/glut.pc.in b/src/glut/mini/glut.pc.in
index ae0689d7e81..151dd0b802b 100644
--- a/src/glut/mini/glut.pc.in
+++ b/src/glut/mini/glut.pc.in
@@ -8,6 +8,6 @@ Description: Mesa OpenGL Utility Toolkit library
 Requires: gl glu
 Requires.private: @GLUT_PC_REQ_PRIV@
 Version: @VERSION@
-Libs: -L${libdir} -lglut
+Libs: -L${libdir} -l@GLUT_LIB@
 Libs.private: @GLUT_PC_LIB_PRIV@
 Cflags: -I${includedir} @GLUT_PC_CFLAGS@
diff --git a/src/glw/Makefile b/src/glw/Makefile
index b153a6df751..d88d7733131 100644
--- a/src/glw/Makefile
+++ b/src/glw/Makefile
@@ -33,7 +33,8 @@ pcedit = sed \
 	-e 's,@VERSION@,$(MAJOR).$(MINOR).$(TINY),' \
 	-e 's,@GLW_PC_REQ_PRIV@,$(GLW_PC_REQ_PRIV),' \
 	-e 's,@GLW_PC_LIB_PRIV@,$(GLW_PC_LIB_PRIV),' \
-	-e 's,@GLW_PC_CFLAGS@,$(GLW_PC_CFLAGS),'
+	-e 's,@GLW_PC_CFLAGS@,$(GLW_PC_CFLAGS),' \
+	-e 's,@GLW_LIB@,$(GLW_LIB),'
 glw.pc: glw.pc.in
 	$(pcedit) $< > $@
 
diff --git a/src/glw/glw.pc.in b/src/glw/glw.pc.in
index 5493093be19..19a7c307c01 100644
--- a/src/glw/glw.pc.in
+++ b/src/glw/glw.pc.in
@@ -8,6 +8,6 @@ Description: Mesa OpenGL widget library
 Requires: gl
 Requires.private: @GLW_PC_REQ_PRIV@
 Version: @VERSION@
-Libs: -L${libdir} -lGLw
+Libs: -L${libdir} -l@GLW_LIB@
 Libs.private: @GLW_PC_LIB_PRIV@
 Cflags: -I${includedir} @GLW_PC_CFLAGS@
diff --git a/src/glx/x11/dri2.c b/src/glx/x11/dri2.c
index f967432b994..ebb2985924c 100644
--- a/src/glx/x11/dri2.c
+++ b/src/glx/x11/dri2.c
@@ -39,6 +39,16 @@
 #include "xf86drm.h"
 #include "dri2.h"
 
+/* Allow the build to work with an older versions of dri2proto.h and
+ * dri2tokens.h.
+ */
+#if DRI2_MINOR < 1
+#undef DRI2_MINOR
+#define DRI2_MINOR 1
+#define X_DRI2GetBuffersWithFormat 7
+#endif
+
+
 static char dri2ExtensionName[] = DRI2_NAME;
 static XExtensionInfo *dri2Info;
 static XEXT_GENERATE_CLOSE_DISPLAY (DRI2CloseDisplay, dri2Info)
@@ -276,6 +286,66 @@ DRI2Buffer *DRI2GetBuffers(Display *dpy, XID drawable,
     return buffers;
 }
 
+
+DRI2Buffer *DRI2GetBuffersWithFormat(Display *dpy, XID drawable,
+				     int *width, int *height,
+				     unsigned int *attachments, int count,
+				     int *outCount)
+{
+    XExtDisplayInfo *info = DRI2FindDisplay(dpy);
+    xDRI2GetBuffersReply rep;
+    xDRI2GetBuffersReq *req;
+    DRI2Buffer *buffers;
+    xDRI2Buffer repBuffer;
+    CARD32 *p;
+    int i;
+
+    XextCheckExtension (dpy, info, dri2ExtensionName, False);
+
+    LockDisplay(dpy);
+    GetReqExtra(DRI2GetBuffers, count * (4 * 2), req);
+    req->reqType = info->codes->major_opcode;
+    req->dri2ReqType = X_DRI2GetBuffersWithFormat;
+    req->drawable = drawable;
+    req->count = count;
+    p = (CARD32 *) &req[1];
+    for (i = 0; i < (count * 2); i++)
+	p[i] = attachments[i];
+
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+	return NULL;
+    }
+
+    *width = rep.width;
+    *height = rep.height;
+    *outCount = rep.count;
+
+    buffers = Xmalloc(rep.count * sizeof buffers[0]);
+    if (buffers == NULL) {
+	_XEatData(dpy, rep.count * sizeof repBuffer);
+	UnlockDisplay(dpy);
+	SyncHandle();
+	return NULL;
+    }
+
+    for (i = 0; i < rep.count; i++) {
+	_XReadPad(dpy, (char *) &repBuffer, sizeof repBuffer);
+	buffers[i].attachment = repBuffer.attachment;
+	buffers[i].name = repBuffer.name;
+	buffers[i].pitch = repBuffer.pitch;
+	buffers[i].cpp = repBuffer.cpp;
+	buffers[i].flags = repBuffer.flags;
+    }
+
+    UnlockDisplay(dpy);
+    SyncHandle();
+
+    return buffers;
+}
+
+
 void DRI2CopyRegion(Display *dpy, XID drawable, XserverRegion region,
 		    CARD32 dest, CARD32 src)
 {
diff --git a/src/glx/x11/dri2.h b/src/glx/x11/dri2.h
index 356c6bcb55d..b0e61f80d70 100644
--- a/src/glx/x11/dri2.h
+++ b/src/glx/x11/dri2.h
@@ -63,6 +63,16 @@ DRI2GetBuffers(Display *dpy, XID drawable,
 	       unsigned int *attachments, int count,
 	       int *outCount);
 
+/**
+ * \note
+ * This function is only supported with DRI2 version 1.1 or later.
+ */
+extern DRI2Buffer *
+DRI2GetBuffersWithFormat(Display *dpy, XID drawable,
+			 int *width, int *height,
+			 unsigned int *attachments, int count,
+			 int *outCount);
+
 extern void
 DRI2CopyRegion(Display *dpy, XID drawable, XserverRegion region,
 	       CARD32 dest, CARD32 src);
diff --git a/src/glx/x11/dri2_glx.c b/src/glx/x11/dri2_glx.c
index b6eeb913b66..fb31898db2d 100644
--- a/src/glx/x11/dri2_glx.c
+++ b/src/glx/x11/dri2_glx.c
@@ -47,6 +47,9 @@
 #include "dri2.h"
 #include "dri_common.h"
 
+#undef DRI2_MINOR
+#define DRI2_MINOR 1
+
 typedef struct __GLXDRIdisplayPrivateRec __GLXDRIdisplayPrivate;
 typedef struct __GLXDRIcontextPrivateRec __GLXDRIcontextPrivate;
 typedef struct __GLXDRIdrawablePrivateRec __GLXDRIdrawablePrivate;
@@ -289,30 +292,25 @@ static void dri2DestroyScreen(__GLXscreenConfigs *psc)
     psc->__driScreen = NULL;
 }
 
-static __DRIbuffer *
-dri2GetBuffers(__DRIdrawable *driDrawable,
-	       int *width, int *height,
-	       unsigned int *attachments, int count,
-	       int *out_count, void *loaderPrivate)
+/**
+ * Process list of buffer received from the server
+ *
+ * Processes the list of buffers received in a reply from the server to either
+ * \c DRI2GetBuffers or \c DRI2GetBuffersWithFormat.
+ */
+static void
+process_buffers(__GLXDRIdrawablePrivate *pdraw, DRI2Buffer *buffers,
+		unsigned count)
 {
-    __GLXDRIdrawablePrivate *pdraw = loaderPrivate;
-    DRI2Buffer *buffers;
     int i;
 
-    buffers = DRI2GetBuffers(pdraw->base.psc->dpy, pdraw->base.xDrawable,
-			     width, height, attachments, count, out_count);
-    if (buffers == NULL)
-       return NULL;
-
-    pdraw->width = *width;
-    pdraw->height = *height;
-    pdraw->bufferCount = *out_count;
+    pdraw->bufferCount = count;
     pdraw->have_fake_front = 0;
     pdraw->have_back = 0;
 
     /* This assumes the DRI2 buffer attachment tokens matches the
      * __DRIbuffer tokens. */
-    for (i = 0; i < *out_count; i++) {
+    for (i = 0; i < count; i++) {
 	pdraw->buffers[i].attachment = buffers[i].attachment;
 	pdraw->buffers[i].name = buffers[i].name;
 	pdraw->buffers[i].pitch = buffers[i].pitch;
@@ -324,6 +322,51 @@ dri2GetBuffers(__DRIdrawable *driDrawable,
 	    pdraw->have_back = 1;
     }
 
+}
+
+static __DRIbuffer *
+dri2GetBuffers(__DRIdrawable *driDrawable,
+	       int *width, int *height,
+	       unsigned int *attachments, int count,
+	       int *out_count, void *loaderPrivate)
+{
+    __GLXDRIdrawablePrivate *pdraw = loaderPrivate;
+    DRI2Buffer *buffers;
+
+    buffers = DRI2GetBuffers(pdraw->base.psc->dpy, pdraw->base.xDrawable,
+			     width, height, attachments, count, out_count);
+    if (buffers == NULL)
+       return NULL;
+
+    pdraw->width = *width;
+    pdraw->height = *height;
+    process_buffers(pdraw, buffers, *out_count);
+
+    Xfree(buffers);
+
+    return pdraw->buffers;
+}
+
+static __DRIbuffer *
+dri2GetBuffersWithFormat(__DRIdrawable *driDrawable,
+			 int *width, int *height,
+			 unsigned int *attachments, int count,
+			 int *out_count, void *loaderPrivate)
+{
+    __GLXDRIdrawablePrivate *pdraw = loaderPrivate;
+    DRI2Buffer *buffers;
+
+    buffers = DRI2GetBuffersWithFormat(pdraw->base.psc->dpy,
+				       pdraw->base.xDrawable,
+				       width, height, attachments,
+				       count, out_count);
+    if (buffers == NULL)
+       return NULL;
+
+    pdraw->width = *width;
+    pdraw->height = *height;
+    process_buffers(pdraw, buffers, *out_count);
+
     Xfree(buffers);
 
     return pdraw->buffers;
@@ -332,7 +375,15 @@ dri2GetBuffers(__DRIdrawable *driDrawable,
 static const __DRIdri2LoaderExtension dri2LoaderExtension = {
     { __DRI_DRI2_LOADER, __DRI_DRI2_LOADER_VERSION },
     dri2GetBuffers,
-    dri2FlushFrontBuffer
+    dri2FlushFrontBuffer,
+    dri2GetBuffersWithFormat,
+};
+
+static const __DRIdri2LoaderExtension dri2LoaderExtension_old = {
+    { __DRI_DRI2_LOADER, __DRI_DRI2_LOADER_VERSION },
+    dri2GetBuffers,
+    dri2FlushFrontBuffer,
+    NULL,
 };
 
 static const __DRIextension *loader_extensions[] = {
@@ -341,11 +392,19 @@ static const __DRIextension *loader_extensions[] = {
     NULL
 };
 
+static const __DRIextension *loader_extensions_old[] = {
+    &dri2LoaderExtension_old.base,
+    &systemTimeExtension.base,
+    NULL
+};
+
 static __GLXDRIscreen *dri2CreateScreen(__GLXscreenConfigs *psc, int screen,
 					__GLXdisplayPrivate *priv)
 {
     const __DRIconfig **driver_configs;
     const __DRIextension **extensions;
+    const __GLXDRIdisplayPrivate *const pdp = (__GLXDRIdisplayPrivate *)
+      priv->dri2Display;
     __GLXDRIscreen *psp;
     char *driverName, *deviceName;
     drm_magic_t magic;
@@ -402,9 +461,16 @@ static __GLXDRIscreen *dri2CreateScreen(__GLXscreenConfigs *psc, int screen,
 	return NULL;
     }
 
+    /* If the server does not support the protocol for
+     * DRI2GetBuffersWithFormat, don't supply that interface to the driver.
+     */
     psc->__driScreen = 
-	psc->dri2->createNewScreen(screen, psc->fd,
-				   loader_extensions, &driver_configs, psc);
+      psc->dri2->createNewScreen(screen, psc->fd,
+				 ((pdp->driMinor < 1)
+				  ? loader_extensions_old
+				  : loader_extensions),
+				 &driver_configs, psc);
+
     if (psc->__driScreen == NULL) {
 	ErrorMessageF("failed to create dri screen\n");
 	return NULL;
diff --git a/src/glx/x11/dri_glx.c b/src/glx/x11/dri_glx.c
index 87d62ad8468..3ce410d9be3 100644
--- a/src/glx/x11/dri_glx.c
+++ b/src/glx/x11/dri_glx.c
@@ -614,7 +614,7 @@ static __GLXDRIscreen *driCreateScreen(__GLXscreenConfigs *psc, int screen,
     char *driverName;
     int i;
 
-    psp = Xmalloc(sizeof *psp);
+    psp = Xcalloc(1, sizeof *psp);
     if (psp == NULL)
 	return NULL;
 
diff --git a/src/glx/x11/drisw_glx.c b/src/glx/x11/drisw_glx.c
index 35bbd9151ca..5e3d763cff5 100644
--- a/src/glx/x11/drisw_glx.c
+++ b/src/glx/x11/drisw_glx.c
@@ -359,7 +359,7 @@ driCreateScreen(__GLXscreenConfigs * psc, int screen,
    const char *driverName = "swrast";
    int i;
 
-   psp = Xmalloc(sizeof *psp);
+   psp = Xcalloc(1, sizeof *psp);
    if (psp == NULL)
       return NULL;
 
diff --git a/src/mesa/Makefile b/src/mesa/Makefile
index 4ff28dae9bb..bb18bee8eaf 100644
--- a/src/mesa/Makefile
+++ b/src/mesa/Makefile
@@ -103,7 +103,8 @@ gl_pcedit = sed \
 	-e 's,@VERSION@,$(MESA_MAJOR).$(MESA_MINOR).$(MESA_TINY),' \
 	-e 's,@GL_PC_REQ_PRIV@,$(GL_PC_REQ_PRIV),' \
 	-e 's,@GL_PC_LIB_PRIV@,$(GL_PC_LIB_PRIV),' \
-	-e 's,@GL_PC_CFLAGS@,$(GL_PC_CFLAGS),'
+	-e 's,@GL_PC_CFLAGS@,$(GL_PC_CFLAGS),' \
+	-e 's,@GL_LIB@,$(GL_LIB),'
 
 gl.pc: gl.pc.in
 	$(gl_pcedit) $< > $@
diff --git a/src/mesa/drivers/dri/i915/i830_context.c b/src/mesa/drivers/dri/i915/i830_context.c
index 10b9bf371c3..840946f908d 100644
--- a/src/mesa/drivers/dri/i915/i830_context.c
+++ b/src/mesa/drivers/dri/i915/i830_context.c
@@ -98,6 +98,8 @@ i830CreateContext(const __GLcontextModes * mesaVis,
    ctx->Const.MaxTextureRectSize = (1 << 11);
    ctx->Const.MaxTextureUnits = I830_TEX_UNITS;
 
+   ctx->Const.MaxTextureMaxAnisotropy = 2.0;
+
    ctx->Const.MaxDrawBuffers = 1;
 
    _tnl_init_vertices(ctx, ctx->Const.MaxArrayLockSize + 12,
diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c
index fdd2cf61096..367d2a3b648 100644
--- a/src/mesa/drivers/dri/i915/i915_context.c
+++ b/src/mesa/drivers/dri/i915/i915_context.c
@@ -73,7 +73,7 @@ i915InvalidateState(GLcontext * ctx, GLuint new_state)
          p->params_uptodate = 0;
    }
 
-   if (new_state & (_NEW_FOG | _NEW_HINT | _NEW_PROGRAM))
+   if (new_state & (_NEW_FOG | _NEW_HINT | _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS))
       i915_update_fog(ctx);
 }
 
@@ -145,6 +145,8 @@ i915CreateContext(const __GLcontextModes * mesaVis,
    ctx->Const.MaxTextureRectSize = (1 << 11);
    ctx->Const.MaxTextureUnits = I915_TEX_UNITS;
 
+   ctx->Const.MaxTextureMaxAnisotropy = 4.0;
+
    /* GL_ARB_fragment_program limits - don't think Mesa actually
     * validates programs against these, and in any case one ARB
     * instruction can translate to more than one HW instruction, so
diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c
index 43f65392b56..a37dd7f4fb5 100644
--- a/src/mesa/drivers/dri/i915/i915_texstate.c
+++ b/src/mesa/drivers/dri/i915/i915_texstate.c
@@ -132,7 +132,7 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
    struct intel_texture_object *intelObj = intel_texture_object(tObj);
    struct gl_texture_image *firstImage;
    GLuint *state = i915->state.Tex[unit], format, pitch;
-   GLint lodbias;
+   GLint lodbias, aniso = 0;
    GLubyte border[4];
 
    memset(state, 0, sizeof(state));
@@ -230,6 +230,10 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
       if (tObj->MaxAnisotropy > 1.0) {
          minFilt = FILTER_ANISOTROPIC;
          magFilt = FILTER_ANISOTROPIC;
+         if (tObj->MaxAnisotropy > 2.0)
+            aniso = SS2_MAX_ANISO_4;
+         else
+            aniso = SS2_MAX_ANISO_2;
       }
       else {
          switch (tObj->MagFilter) {
@@ -275,7 +279,8 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
 
       state[I915_TEXREG_SS2] |= ((minFilt << SS2_MIN_FILTER_SHIFT) |
                                  (mipFilt << SS2_MIP_FILTER_SHIFT) |
-                                 (magFilt << SS2_MAG_FILTER_SHIFT));
+                                 (magFilt << SS2_MAG_FILTER_SHIFT) |
+                                 aniso);
    }
 
    {
diff --git a/src/mesa/drivers/dri/i915/intel_generatemipmap.c b/src/mesa/drivers/dri/i915/intel_generatemipmap.c
new file mode 120000
index 00000000000..4c6b37ada01
--- /dev/null
+++ b/src/mesa/drivers/dri/i915/intel_generatemipmap.c
@@ -0,0 +1 @@
+../intel/intel_generatemipmap.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile
index 2934414d99a..9712c387254 100644
--- a/src/mesa/drivers/dri/i965/Makefile
+++ b/src/mesa/drivers/dri/i965/Makefile
@@ -14,6 +14,7 @@ DRIVER_SOURCES = \
 	intel_decode.c \
 	intel_extensions.c \
 	intel_fbo.c \
+	intel_generatemipmap.c \
 	intel_mipmap_tree.c \
 	intel_regions.c \
 	intel_screen.c \
@@ -69,6 +70,7 @@ DRIVER_SOURCES = \
 	brw_vs_constval.c \
 	brw_vs_emit.c \
 	brw_vs_state.c \
+	brw_vs_surface_state.c \
 	brw_vtbl.c \
 	brw_wm.c \
 	brw_wm_debug.c \
diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c
index d96ff293102..4dbe551d832 100644
--- a/src/mesa/drivers/dri/i965/brw_context.c
+++ b/src/mesa/drivers/dri/i965/brw_context.c
@@ -118,6 +118,8 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis,
    ctx->Const.MaxCubeTextureLevels = 12;
    ctx->Const.MaxTextureRectSize = (1<<12);
    
+   ctx->Const.MaxTextureMaxAnisotropy = 16.0;
+
    /* if conformance mode is set, swrast can handle any size AA point */
    ctx->Const.MaxPointSizeAA = 255.0;
 
diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
index a0b3b06309f..873fc8ffff6 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -131,6 +131,7 @@ struct brw_context;
 #define BRW_NEW_WM_INPUT_DIMENSIONS     0x100
 #define BRW_NEW_INPUT_VARYING           0x200
 #define BRW_NEW_PSP                     0x800
+#define BRW_NEW_WM_SURFACES		0x1000
 #define BRW_NEW_FENCE                   0x2000
 #define BRW_NEW_INDICES			0x4000
 #define BRW_NEW_VERTICES		0x8000
@@ -161,6 +162,7 @@ struct brw_vertex_program {
    struct gl_vertex_program program;
    GLuint id;
    dri_bo *const_buffer;    /** Program constant buffer/surface */
+   GLboolean use_const_buffer;
 };
 
 
@@ -171,6 +173,7 @@ struct brw_fragment_program {
    GLboolean isGLSL;  /**< really, any IF/LOOP/CONT/BREAK instructions */
 
    dri_bo *const_buffer;    /** Program constant buffer/surface */
+   GLboolean use_const_buffer;
 };
 
 
@@ -243,6 +246,9 @@ struct brw_vs_ouput_sizes {
 };
 
 
+/** Number of general purpose registers (VS, WM, etc) */
+#define BRW_MAX_GRF 128
+
 /** Number of texture sampler units */
 #define BRW_MAX_TEX_UNIT 16
 
@@ -448,8 +454,6 @@ struct brw_context
 
    struct {
       struct brw_state_flags dirty;
-      struct brw_tracked_state **atoms;
-      GLuint nr_atoms;
 
       GLuint nr_color_regions;
       struct intel_region *color_regions[MAX_DRAW_BUFFERS];
@@ -469,7 +473,8 @@ struct brw_context
       int validated_bo_count;
    } state;
 
-   struct brw_cache cache;
+   struct brw_cache cache;  /** non-surface items */
+   struct brw_cache surface_cache;  /* surface items */
    struct brw_cached_batch_item *cached_batch_items;
 
    struct {
@@ -553,11 +558,6 @@ struct brw_context
       GLuint vs_size;
       GLuint total_size;
 
-      /* Dynamic tracker which changes to reflect the state referenced
-       * by active fp and vp program parameters:
-       */
-      struct brw_tracked_state tracked_state;
-
       dri_bo *curbe_bo;
       /** Offset within curbe_bo of space for current curbe entry */
       GLuint curbe_offset;
diff --git a/src/mesa/drivers/dri/i965/brw_curbe.c b/src/mesa/drivers/dri/i965/brw_curbe.c
index 18b187ed1dc..a1a6c53d0e0 100644
--- a/src/mesa/drivers/dri/i965/brw_curbe.c
+++ b/src/mesa/drivers/dri/i965/brw_curbe.c
@@ -36,6 +36,7 @@
 #include "main/macros.h"
 #include "main/enums.h"
 #include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
 #include "shader/prog_statevars.h"
 #include "intel_batchbuffer.h"
 #include "intel_regions.h"
@@ -188,13 +189,6 @@ static void prepare_constant_buffer(struct brw_context *brw)
    GLfloat *buf;
    GLuint i;
 
-   /* Update our own dependency flags.  This works because this
-    * function will also be called whenever fp or vp changes.
-    */
-   brw->curbe.tracked_state.dirty.mesa = (_NEW_TRANSFORM|_NEW_PROJECTION);
-   brw->curbe.tracked_state.dirty.mesa |= vp->program.Base.Parameters->StateFlags;
-   brw->curbe.tracked_state.dirty.mesa |= fp->program.Base.Parameters->StateFlags;
-
    if (sz == 0) {
       if (brw->curbe.last_buf) {
 	 free(brw->curbe.last_buf);
@@ -335,72 +329,11 @@ static void prepare_constant_buffer(struct brw_context *brw)
     */
 }
 
-
-/**
- * Copy Mesa program parameters into given constant buffer.
- */
-static void
-update_constant_buffer(struct brw_context *brw,
-                       const struct gl_program_parameter_list *params,
-                       dri_bo *const_buffer)
-{
-   const int size = params->NumParameters * 4 * sizeof(GLfloat);
-
-   /* copy Mesa program constants into the buffer */
-   if (const_buffer && size > 0) {
-      GLubyte *map;
-
-      assert(const_buffer);
-      assert(const_buffer->size >= size);
-
-      dri_bo_map(const_buffer, GL_TRUE);
-      map = const_buffer->virtual;
-      memcpy(map, params->ParameterValues, size);
-      dri_bo_unmap(const_buffer);
-
-      if (0) {
-         int i;
-         for (i = 0; i < params->NumParameters; i++) {
-            float *p = params->ParameterValues[i];
-            printf("%d: %f %f %f %f\n", i, p[0], p[1], p[2], p[3]);
-         }
-      }
-   }
-}
-
-
-/** Copy current vertex program's parameters into the constant buffer */
-static void
-update_vertex_constant_buffer(struct brw_context *brw)
-{
-   struct brw_vertex_program *vp =
-      (struct brw_vertex_program *) brw->vertex_program;
-   if (0) {
-      printf("update VS constants in buffer %p\n", vp->const_buffer);
-      printf("program %u\n", vp->program.Base.Id);
-   }
-   update_constant_buffer(brw, vp->program.Base.Parameters, vp->const_buffer);
-}
-
-
-/** Copy current fragment program's parameters into the constant buffer */
-static void
-update_fragment_constant_buffer(struct brw_context *brw)
-{
-   struct brw_fragment_program *fp =
-      (struct brw_fragment_program *) brw->fragment_program;
-   update_constant_buffer(brw, fp->program.Base.Parameters, fp->const_buffer);
-}
-
-
 static void emit_constant_buffer(struct brw_context *brw)
 {
    struct intel_context *intel = &brw->intel;
    GLuint sz = brw->curbe.total_size;
 
-   update_vertex_constant_buffer(brw);
-   update_fragment_constant_buffer(brw);
-
    BEGIN_BATCH(2, IGNORE_CLIPRECTS);
    if (sz == 0) {
       OUT_BATCH((CMD_CONST_BUFFER << 16) | (2 - 2));
@@ -422,7 +355,7 @@ static void emit_constant_buffer(struct brw_context *brw)
  */
 const struct brw_tracked_state brw_constant_buffer = {
    .dirty = {
-      .mesa = (_NEW_TRANSFORM|_NEW_PROJECTION),      /* plus fp and vp flags */
+      .mesa = _NEW_PROGRAM_CONSTANTS,
       .brw  = (BRW_NEW_FRAGMENT_PROGRAM |
 	       BRW_NEW_VERTEX_PROGRAM |
 	       BRW_NEW_URB_FENCE | /* Implicit - hardware requires this, not used above */
diff --git a/src/mesa/drivers/dri/i965/brw_misc_state.c b/src/mesa/drivers/dri/i965/brw_misc_state.c
index 9bc5c35139c..4784254bc7d 100644
--- a/src/mesa/drivers/dri/i965/brw_misc_state.c
+++ b/src/mesa/drivers/dri/i965/brw_misc_state.c
@@ -118,7 +118,10 @@ static void upload_binding_table_pointers(struct brw_context *brw)
 
    BEGIN_BATCH(6, IGNORE_CLIPRECTS);
    OUT_BATCH(CMD_BINDING_TABLE_PTRS << 16 | (6 - 2));
-   OUT_RELOC(brw->vs.bind_bo, I915_GEM_DOMAIN_SAMPLER, 0, 0); /* vs */
+   if (brw->vs.bind_bo != NULL)
+      OUT_RELOC(brw->vs.bind_bo, I915_GEM_DOMAIN_SAMPLER, 0, 0); /* vs */
+   else
+      OUT_BATCH(0);
    OUT_BATCH(0); /* gs */
    OUT_BATCH(0); /* clip */
    OUT_BATCH(0); /* sf */
diff --git a/src/mesa/drivers/dri/i965/brw_sf_state.c b/src/mesa/drivers/dri/i965/brw_sf_state.c
index fc4eddda0a5..68fa9820b6f 100644
--- a/src/mesa/drivers/dri/i965/brw_sf_state.c
+++ b/src/mesa/drivers/dri/i965/brw_sf_state.c
@@ -147,7 +147,7 @@ sf_unit_populate_key(struct brw_context *brw, struct brw_sf_unit_key *key)
    key->line_smooth = ctx->Line.SmoothFlag;
 
    key->point_sprite = ctx->Point.PointSprite;
-   key->point_size = ctx->Point.Size;
+   key->point_size = CLAMP(ctx->Point.Size, ctx->Point.MinSize, ctx->Point.MaxSize);
    key->point_attenuated = ctx->Point._Attenuated;
 
    key->render_to_fbo = brw->intel.ctx.DrawBuffer->Name != 0;
diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h
index 81b0a45998f..bf9f6cae55e 100644
--- a/src/mesa/drivers/dri/i965/brw_state.h
+++ b/src/mesa/drivers/dri/i965/brw_state.h
@@ -72,11 +72,13 @@ const struct brw_tracked_state brw_sf_vp;
 const struct brw_tracked_state brw_state_base_address;
 const struct brw_tracked_state brw_urb_fence;
 const struct brw_tracked_state brw_vertex_state;
+const struct brw_tracked_state brw_vs_surfaces;
 const struct brw_tracked_state brw_vs_prog;
 const struct brw_tracked_state brw_vs_unit;
 const struct brw_tracked_state brw_wm_input_sizes;
 const struct brw_tracked_state brw_wm_prog;
 const struct brw_tracked_state brw_wm_samplers;
+const struct brw_tracked_state brw_wm_constant_surface;
 const struct brw_tracked_state brw_wm_surfaces;
 const struct brw_tracked_state brw_wm_unit;
 
@@ -91,6 +93,20 @@ const struct brw_tracked_state brw_drawing_rect;
 const struct brw_tracked_state brw_indices;
 const struct brw_tracked_state brw_vertices;
 
+/**
+ * Use same key for WM and VS surfaces.
+ */
+struct brw_surface_key {
+   GLenum target, depthmode;
+   dri_bo *bo;
+   GLint format, internal_format;
+   GLint first_level, last_level;
+   GLint width, height, depth;
+   GLint pitch, cpp;
+   uint32_t tiling;
+   GLuint offset;
+};
+
 /***********************************************************************
  * brw_state.c
  */
@@ -135,8 +151,8 @@ dri_bo *brw_search_cache( struct brw_cache *cache,
 			  void *aux_return);
 void brw_state_cache_check_size( struct brw_context *brw );
 
-void brw_init_cache( struct brw_context *brw );
-void brw_destroy_cache( struct brw_context *brw );
+void brw_init_caches( struct brw_context *brw );
+void brw_destroy_caches( struct brw_context *brw );
 
 /***********************************************************************
  * brw_state_batch.c
@@ -150,4 +166,9 @@ GLboolean brw_cached_batch_struct( struct brw_context *brw,
 void brw_destroy_batch_cache( struct brw_context *brw );
 void brw_clear_batch_cache_flush( struct brw_context *brw );
 
+/* brw_wm_surface_state.c */
+dri_bo *
+brw_create_constant_surface( struct brw_context *brw,
+                             struct brw_surface_key *key );
+
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_state_cache.c b/src/mesa/drivers/dri/i965/brw_state_cache.c
index d5b51664066..e40d7a04164 100644
--- a/src/mesa/drivers/dri/i965/brw_state_cache.c
+++ b/src/mesa/drivers/dri/i965/brw_state_cache.c
@@ -56,9 +56,9 @@
  * incorrect program is run for the other instance.
  */
 
+#include "main/imports.h"
 #include "brw_state.h"
 #include "intel_batchbuffer.h"
-#include "main/imports.h"
 
 /* XXX: Fixme - have to include these to get the sizes of the prog_key
  * structs:
@@ -69,8 +69,10 @@
 #include "brw_sf.h"
 #include "brw_gs.h"
 
-static GLuint hash_key( const void *key, GLuint key_size,
-			dri_bo **reloc_bufs, GLuint nr_reloc_bufs)
+
+static GLuint
+hash_key(const void *key, GLuint key_size,
+         dri_bo **reloc_bufs, GLuint nr_reloc_bufs)
 {
    GLuint *ikey = (GLuint *)key;
    GLuint hash = 0, i;
@@ -95,6 +97,7 @@ static GLuint hash_key( const void *key, GLuint key_size,
    return hash;
 }
 
+
 /**
  * Marks a new buffer as being chosen for the given cache id.
  */
@@ -111,6 +114,7 @@ update_cache_last(struct brw_cache *cache, enum brw_cache_id cache_id,
    cache->brw->state.dirty.cache |= 1 << cache_id;
 }
 
+
 static struct brw_cache_item *
 search_cache(struct brw_cache *cache, enum brw_cache_id cache_id,
 	     GLuint hash, const void *key, GLuint key_size,
@@ -143,7 +147,8 @@ search_cache(struct brw_cache *cache, enum brw_cache_id cache_id,
 }
 
 
-static void rehash( struct brw_cache *cache )
+static void
+rehash(struct brw_cache *cache)
 {
    struct brw_cache_item **items;
    struct brw_cache_item *c, *next;
@@ -164,15 +169,17 @@ static void rehash( struct brw_cache *cache )
    cache->size = size;
 }
 
+
 /**
  * Returns the buffer object matching cache_id and key, or NULL.
  */
-dri_bo *brw_search_cache( struct brw_cache *cache,
-			  enum brw_cache_id cache_id,
-			  const void *key,
-			  GLuint key_size,
-			  dri_bo **reloc_bufs, GLuint nr_reloc_bufs,
-			  void *aux_return )
+dri_bo *
+brw_search_cache(struct brw_cache *cache,
+                 enum brw_cache_id cache_id,
+                 const void *key,
+                 GLuint key_size,
+                 dri_bo **reloc_bufs, GLuint nr_reloc_bufs,
+                 void *aux_return)
 {
    struct brw_cache_item *item;
    GLuint hash = hash_key(key, key_size, reloc_bufs, nr_reloc_bufs);
@@ -192,6 +199,7 @@ dri_bo *brw_search_cache( struct brw_cache *cache,
    return item->bo;
 }
 
+
 dri_bo *
 brw_upload_cache( struct brw_cache *cache,
 		  enum brw_cache_id cache_id,
@@ -265,7 +273,9 @@ brw_upload_cache( struct brw_cache *cache,
    return bo;
 }
 
-/* This doesn't really work with aux data.  Use search/upload instead
+
+/**
+ * This doesn't really work with aux data.  Use search/upload instead
  */
 dri_bo *
 brw_cache_data_sz(struct brw_cache *cache,
@@ -296,6 +306,7 @@ brw_cache_data_sz(struct brw_cache *cache,
    return bo;
 }
 
+
 /**
  * Wrapper around brw_cache_data_sz using the cache_id's canonical key size.
  *
@@ -319,21 +330,22 @@ enum pool_type {
    DW_GENERAL_STATE
 };
 
-static void
-brw_init_cache_id( struct brw_context *brw,
-		const char *name,
-		enum brw_cache_id id,
-		GLuint key_size,
-		GLuint aux_size)
-{
-   struct brw_cache *cache = &brw->cache;
 
+static void
+brw_init_cache_id(struct brw_cache *cache,
+                  const char *name,
+                  enum brw_cache_id id,
+                  GLuint key_size,
+                  GLuint aux_size)
+{
    cache->name[id] = strdup(name);
    cache->key_size[id] = key_size;
    cache->aux_size[id] = aux_size;
 }
 
-void brw_init_cache( struct brw_context *brw )
+
+static void
+brw_init_non_surface_cache(struct brw_context *brw)
 {
    struct brw_cache *cache = &brw->cache;
 
@@ -342,114 +354,136 @@ void brw_init_cache( struct brw_context *brw )
    cache->size = 7;
    cache->n_items = 0;
    cache->items = (struct brw_cache_item **)
-      _mesa_calloc(cache->size * 
-		   sizeof(struct brw_cache_item));
+      _mesa_calloc(cache->size * sizeof(struct brw_cache_item));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "CC_VP",
 		     BRW_CC_VP,
 		     sizeof(struct brw_cc_viewport),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "CC_UNIT",
 		     BRW_CC_UNIT,
 		     sizeof(struct brw_cc_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "WM_PROG",
 		     BRW_WM_PROG,
 		     sizeof(struct brw_wm_prog_key),
 		     sizeof(struct brw_wm_prog_data));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SAMPLER_DEFAULT_COLOR",
 		     BRW_SAMPLER_DEFAULT_COLOR,
 		     sizeof(struct brw_sampler_default_color),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SAMPLER",
 		     BRW_SAMPLER,
 		     0,		/* variable key/data size */
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "WM_UNIT",
 		     BRW_WM_UNIT,
 		     sizeof(struct brw_wm_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SF_PROG",
 		     BRW_SF_PROG,
 		     sizeof(struct brw_sf_prog_key),
 		     sizeof(struct brw_sf_prog_data));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SF_VP",
 		     BRW_SF_VP,
 		     sizeof(struct brw_sf_viewport),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SF_UNIT",
 		     BRW_SF_UNIT,
 		     sizeof(struct brw_sf_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "VS_UNIT",
 		     BRW_VS_UNIT,
 		     sizeof(struct brw_vs_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "VS_PROG",
 		     BRW_VS_PROG,
 		     sizeof(struct brw_vs_prog_key),
 		     sizeof(struct brw_vs_prog_data));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "CLIP_UNIT",
 		     BRW_CLIP_UNIT,
 		     sizeof(struct brw_clip_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "CLIP_PROG",
 		     BRW_CLIP_PROG,
 		     sizeof(struct brw_clip_prog_key),
 		     sizeof(struct brw_clip_prog_data));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "GS_UNIT",
 		     BRW_GS_UNIT,
 		     sizeof(struct brw_gs_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "GS_PROG",
 		     BRW_GS_PROG,
 		     sizeof(struct brw_gs_prog_key),
 		     sizeof(struct brw_gs_prog_data));
+}
 
-   brw_init_cache_id(brw,
+
+static void
+brw_init_surface_cache(struct brw_context *brw)
+{
+   struct brw_cache *cache = &brw->surface_cache;
+
+   cache->brw = brw;
+
+   cache->size = 7;
+   cache->n_items = 0;
+   cache->items = (struct brw_cache_item **)
+      _mesa_calloc(cache->size * sizeof(struct brw_cache_item));
+
+   brw_init_cache_id(cache,
 		     "SS_SURFACE",
 		     BRW_SS_SURFACE,
 		     sizeof(struct brw_surface_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SS_SURF_BIND",
 		     BRW_SS_SURF_BIND,
 		     0,
 		     0);
 }
 
+
+void
+brw_init_caches(struct brw_context *brw)
+{
+   brw_init_non_surface_cache(brw);
+   brw_init_surface_cache(brw);
+}
+
+
 static void
-brw_clear_cache( struct brw_context *brw )
+brw_clear_cache(struct brw_context *brw, struct brw_cache *cache)
 {
    struct brw_cache_item *c, *next;
    GLuint i;
@@ -457,8 +491,8 @@ brw_clear_cache( struct brw_context *brw )
    if (INTEL_DEBUG & DEBUG_STATE)
       _mesa_printf("%s\n", __FUNCTION__);
 
-   for (i = 0; i < brw->cache.size; i++) {
-      for (c = brw->cache.items[i]; c; c = next) {
+   for (i = 0; i < cache->size; i++) {
+      for (c = cache->items[i]; c; c = next) {
 	 int j;
 
 	 next = c->next;
@@ -468,10 +502,10 @@ brw_clear_cache( struct brw_context *brw )
 	 free((void *)c->key);
 	 free(c);
       }
-      brw->cache.items[i] = NULL;
+      cache->items[i] = NULL;
    }
 
-   brw->cache.n_items = 0;
+   cache->n_items = 0;
 
    if (brw->curbe.last_buf) {
       _mesa_free(brw->curbe.last_buf);
@@ -483,25 +517,46 @@ brw_clear_cache( struct brw_context *brw )
    brw->state.dirty.cache |= ~0;
 }
 
-void brw_state_cache_check_size( struct brw_context *brw )
+
+void
+brw_state_cache_check_size(struct brw_context *brw)
 {
+   if (INTEL_DEBUG & DEBUG_STATE)
+      _mesa_printf("%s (n_items=%d)\n", __FUNCTION__, brw->cache.n_items);
+
    /* un-tuned guess.  We've got around 20 state objects for a total of around
     * 32k, so 1000 of them is around 1.5MB.
     */
    if (brw->cache.n_items > 1000)
-      brw_clear_cache(brw);
+      brw_clear_cache(brw, &brw->cache);
+
+   if (brw->surface_cache.n_items > 1000)
+      brw_clear_cache(brw, &brw->surface_cache);
 }
 
-void brw_destroy_cache( struct brw_context *brw )
+
+static void
+brw_destroy_cache(struct brw_context *brw, struct brw_cache *cache)
 {
    GLuint i;
 
-   brw_clear_cache(brw);
+   if (INTEL_DEBUG & DEBUG_STATE)
+      _mesa_printf("%s\n", __FUNCTION__);
+
+   brw_clear_cache(brw, cache);
    for (i = 0; i < BRW_MAX_CACHE; i++) {
-      dri_bo_unreference(brw->cache.last_bo[i]);
-      free(brw->cache.name[i]);
+      dri_bo_unreference(cache->last_bo[i]);
+      free(cache->name[i]);
    }
-   free(brw->cache.items);
-   brw->cache.items = NULL;
-   brw->cache.size = 0;
+   free(cache->items);
+   cache->items = NULL;
+   cache->size = 0;
+}
+
+
+void
+brw_destroy_caches(struct brw_context *brw)
+{
+   brw_destroy_cache(brw, &brw->cache);
+   brw_destroy_cache(brw, &brw->surface_cache);
 }
diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c
index 5de1450e612..c6dfea4743c 100644
--- a/src/mesa/drivers/dri/i965/brw_state_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_state_upload.c
@@ -59,11 +59,12 @@ const struct brw_tracked_state *atoms[] =
    &brw_curbe_offsets,
    &brw_recalculate_urb_fence,
 
-
    &brw_cc_vp,
    &brw_cc_unit,
 
-   &brw_wm_surfaces,		/* must do before samplers */
+   &brw_vs_surfaces,		/* must do before unit */
+   &brw_wm_constant_surface,	/* must do before wm surfaces/bind bo */
+   &brw_wm_surfaces,		/* must do before samplers and unit */
    &brw_wm_samplers,
 
    &brw_wm_unit,
@@ -88,54 +89,26 @@ const struct brw_tracked_state *atoms[] =
 
    &brw_line_stipple,
    &brw_aa_line_parameters,
-   /* Ordering of the commands below is documented as fixed.  
-    */
-#if 0
-   &brw_pipelined_state_pointers,
-   &brw_urb_fence,
-   &brw_constant_buffer_state,
-#else
+
    &brw_psp_urb_cbs,
-#endif
 
    &brw_drawing_rect,
    &brw_indices,
    &brw_vertices,
 
-   NULL,			/* brw_constant_buffer */
+   &brw_constant_buffer
 };
 
 
 void brw_init_state( struct brw_context *brw )
 {
-   GLuint i;
-
-   brw_init_cache(brw);
-
-   brw->state.atoms = _mesa_malloc(sizeof(atoms));
-   brw->state.nr_atoms = sizeof(atoms)/sizeof(*atoms);
-   _mesa_memcpy(brw->state.atoms, atoms, sizeof(atoms));
-
-   /* Patch in a pointer to the dynamic state atom:
-    */
-   for (i = 0; i < brw->state.nr_atoms; i++)
-      if (brw->state.atoms[i] == NULL)
-	 brw->state.atoms[i] = &brw->curbe.tracked_state;
-
-   _mesa_memcpy(&brw->curbe.tracked_state, 
-		&brw_constant_buffer,
-		sizeof(brw_constant_buffer));
+   brw_init_caches(brw);
 }
 
 
 void brw_destroy_state( struct brw_context *brw )
 {
-   if (brw->state.atoms) {
-      _mesa_free(brw->state.atoms);
-      brw->state.atoms = NULL;
-   }
-
-   brw_destroy_cache(brw);
+   brw_destroy_caches(brw);
    brw_destroy_batch_cache(brw);
 }
 
@@ -218,6 +191,7 @@ static struct dirty_bit_map mesa_bits[] = {
    DEFINE_BIT(_NEW_MULTISAMPLE),
    DEFINE_BIT(_NEW_TRACK_MATRIX),
    DEFINE_BIT(_NEW_PROGRAM),
+   DEFINE_BIT(_NEW_PROGRAM_CONSTANTS),
    {0, 0, 0}
 };
 
@@ -336,7 +310,7 @@ void brw_validate_state( struct brw_context *brw )
 
    /* do prepare stage for all atoms */
    for (i = 0; i < Elements(atoms); i++) {
-      const struct brw_tracked_state *atom = brw->state.atoms[i];
+      const struct brw_tracked_state *atom = atoms[i];
 
       if (brw->intel.Fallback)
          break;
@@ -367,8 +341,8 @@ void brw_upload_state(struct brw_context *brw)
       _mesa_memset(&examined, 0, sizeof(examined));
       prev = *state;
 
-      for (i = 0; i < brw->state.nr_atoms; i++) {	 
-	 const struct brw_tracked_state *atom = brw->state.atoms[i];
+      for (i = 0; i < Elements(atoms); i++) {	 
+	 const struct brw_tracked_state *atom = atoms[i];
 	 struct brw_state_flags generated;
 
 	 assert(atom->dirty.mesa ||
@@ -397,7 +371,7 @@ void brw_upload_state(struct brw_context *brw)
    }
    else {
       for (i = 0; i < Elements(atoms); i++) {	 
-	 const struct brw_tracked_state *atom = brw->state.atoms[i];
+	 const struct brw_tracked_state *atom = atoms[i];
 
 	 if (brw->intel.Fallback)
 	    break;
diff --git a/src/mesa/drivers/dri/i965/brw_vs.h b/src/mesa/drivers/dri/i965/brw_vs.h
index d20cf78b8af..1e4f66091e3 100644
--- a/src/mesa/drivers/dri/i965/brw_vs.h
+++ b/src/mesa/drivers/dri/i965/brw_vs.h
@@ -75,8 +75,6 @@ struct brw_vs_compile {
 
    struct brw_reg userplane[6];
 
-   /** using a real constant buffer? */
-   GLboolean use_const_buffer;
    /** we may need up to 3 constants per instruction (if use_const_buffer) */
    struct {
       GLint index;
diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c
index 524f1211cee..d7f75e3685e 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c
@@ -69,13 +69,18 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
 {
    GLuint i, reg = 0, mrf;
 
-#if 0
-   if (c->vp->program.Base.Parameters->NumParameters >= 6)
-      c->use_const_buffer = 1;
+   /* Determine whether to use a real constant buffer or use a block
+    * of GRF registers for constants.  The later is faster but only
+    * works if everything fits in the GRF.
+    * XXX this heuristic/check may need some fine tuning...
+    */
+   if (c->vp->program.Base.Parameters->NumParameters +
+       c->vp->program.Base.NumTemporaries + 20 > BRW_MAX_GRF)
+      c->vp->use_const_buffer = GL_TRUE;
    else
-#endif
-      c->use_const_buffer = GL_FALSE;
-   /*printf("use_const_buffer = %d\n", c->use_const_buffer);*/
+      c->vp->use_const_buffer = GL_FALSE;
+
+   /*printf("use_const_buffer = %d\n", c->vp->use_const_buffer);*/
 
    /* r0 -- reserved as usual
     */
@@ -96,7 +101,7 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
 
    /* Vertex program parameters from curbe:
     */
-   if (c->use_const_buffer) {
+   if (c->vp->use_const_buffer) {
       /* get constants from a real constant buffer */
       c->prog_data.curb_read_length = 0;
       c->prog_data.nr_params = 4; /* XXX 0 causes a bug elsewhere... */
@@ -172,7 +177,7 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
       reg++;
    }
 
-   if (c->use_const_buffer) {
+   if (c->vp->use_const_buffer) {
       for (i = 0; i < 3; i++) {
          c->current_const[i].index = -1;
          c->current_const[i].reg = brw_vec8_grf(reg, 0);
@@ -709,10 +714,11 @@ get_constant(struct brw_vs_compile *c,
    struct brw_compile *p = &c->func;
    struct brw_reg const_reg;
    struct brw_reg const2_reg;
+   const GLboolean relAddr = src->RelAddr;
 
    assert(argIndex < 3);
 
-   if (c->current_const[argIndex].index != src->Index || src->RelAddr) {
+   if (c->current_const[argIndex].index != src->Index || relAddr) {
       struct brw_reg addrReg = c->regs[PROGRAM_ADDRESS][0];
 
       c->current_const[argIndex].index = src->Index;
@@ -725,13 +731,13 @@ get_constant(struct brw_vs_compile *c,
       brw_dp_READ_4_vs(p,
                        c->current_const[argIndex].reg,/* writeback dest */
                        0,                             /* oword */
-                       src->RelAddr,                  /* relative indexing? */
+                       relAddr,                       /* relative indexing? */
                        addrReg,                       /* address register */
                        16 * src->Index,               /* byte offset */
                        SURF_INDEX_VERT_CONST_BUFFER   /* binding table index */
                        );
 
-      if (src->RelAddr) {
+      if (relAddr) {
          /* second read */
          const2_reg = get_tmp(c);
 
@@ -742,7 +748,7 @@ get_constant(struct brw_vs_compile *c,
          brw_dp_READ_4_vs(p,
                           const2_reg,              /* writeback dest */
                           1,                       /* oword */
-                          src->RelAddr,            /* relative indexing? */
+                          relAddr,                 /* relative indexing? */
                           addrReg,                 /* address register */
                           16 * src->Index,         /* byte offset */
                           SURF_INDEX_VERT_CONST_BUFFER
@@ -752,7 +758,7 @@ get_constant(struct brw_vs_compile *c,
 
    const_reg = c->current_const[argIndex].reg;
 
-   if (src->RelAddr) {
+   if (relAddr) {
       /* merge the two Owords into the constant register */
       /* const_reg[7..4] = const2_reg[7..4] */
       brw_MOV(p,
@@ -869,7 +875,7 @@ get_src_reg( struct brw_vs_compile *c,
    case PROGRAM_STATE_VAR:
    case PROGRAM_CONSTANT:
    case PROGRAM_UNIFORM:
-      if (c->use_const_buffer) {
+      if (c->vp->use_const_buffer) {
          return get_constant(c, inst, argIndex);
       }
       else if (relAddr) {
@@ -1219,7 +1225,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
 
    for (insn = 0; insn < nr_insns; insn++) {
 
-      struct prog_instruction *inst = &c->vp->program.Base.Instructions[insn];
+      const struct prog_instruction *inst = &c->vp->program.Base.Instructions[insn];
       struct brw_reg args[3], dst;
       GLuint i;
       
@@ -1232,7 +1238,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
        */
       if (inst->Opcode != OPCODE_SWZ)
 	  for (i = 0; i < 3; i++) {
-	      struct prog_src_register *src = &inst->SrcReg[i];
+	      const struct prog_src_register *src = &inst->SrcReg[i];
 	      index = src->Index;
 	      file = src->File;	
 	      if (file == PROGRAM_OUTPUT && c->output_regs[index].used_in_src)
diff --git a/src/mesa/drivers/dri/i965/brw_vs_surface_state.c b/src/mesa/drivers/dri/i965/brw_vs_surface_state.c
new file mode 100644
index 00000000000..89f47522a1c
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/brw_vs_surface_state.c
@@ -0,0 +1,226 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "main/mtypes.h"
+#include "main/texformat.h"
+#include "main/texstore.h"
+#include "shader/prog_parameter.h"
+
+#include "brw_context.h"
+#include "brw_state.h"
+#include "brw_defines.h"
+
+/* Creates a new VS constant buffer reflecting the current VS program's
+ * constants, if needed by the VS program.
+ *
+ * Otherwise, constants go through the CURBEs using the brw_constant_buffer
+ * state atom.
+ */
+static drm_intel_bo *
+brw_vs_update_constant_buffer(struct brw_context *brw)
+{
+   struct intel_context *intel = &brw->intel;
+   struct brw_vertex_program *vp =
+      (struct brw_vertex_program *) brw->vertex_program;
+   const struct gl_program_parameter_list *params = vp->program.Base.Parameters;
+   const int size = params->NumParameters * 4 * sizeof(GLfloat);
+   drm_intel_bo *const_buffer;
+
+   /* BRW_NEW_VERTEX_PROGRAM */
+   if (!vp->use_const_buffer)
+      return NULL;
+
+   const_buffer = drm_intel_bo_alloc(intel->bufmgr, "vp_const_buffer",
+				     size, 64);
+
+   /* _NEW_PROGRAM_CONSTANTS */
+   dri_bo_subdata(const_buffer, 0, size, params->ParameterValues);
+
+   return const_buffer;
+}
+
+/**
+ * Update the surface state for a VS constant buffer.
+ *
+ * Sets brw->vs.surf_bo[surf] and brw->vp->const_buffer.
+ */
+static void
+brw_update_vs_constant_surface( GLcontext *ctx,
+                                GLuint surf)
+{
+   struct brw_context *brw = brw_context(ctx);
+   struct brw_surface_key key;
+   struct brw_vertex_program *vp =
+      (struct brw_vertex_program *) brw->vertex_program;
+   const struct gl_program_parameter_list *params = vp->program.Base.Parameters;
+
+   assert(surf == 0);
+
+   /* If we're in this state update atom, we need to update VS constants, so
+    * free the old buffer and create a new one for the new contents.
+    */
+   dri_bo_unreference(vp->const_buffer);
+   vp->const_buffer = brw_vs_update_constant_buffer(brw);
+
+   /* If there's no constant buffer, then no surface BO is needed to point at
+    * it.
+    */
+   if (vp->const_buffer == 0) {
+      drm_intel_bo_unreference(brw->vs.surf_bo[surf]);
+      brw->vs.surf_bo[surf] = NULL;
+      return;
+   }
+
+   memset(&key, 0, sizeof(key));
+
+   key.format = MESA_FORMAT_RGBA_FLOAT32;
+   key.internal_format = GL_RGBA;
+   key.bo = vp->const_buffer;
+   key.depthmode = GL_NONE;
+   key.pitch = params->NumParameters;
+   key.width = params->NumParameters;
+   key.height = 1;
+   key.depth = 1;
+   key.cpp = 16;
+
+   /*
+   printf("%s:\n", __FUNCTION__);
+   printf("  width %d  height %d  depth %d  cpp %d  pitch %d\n",
+          key.width, key.height, key.depth, key.cpp, key.pitch);
+   */
+
+   drm_intel_bo_unreference(brw->vs.surf_bo[surf]);
+   brw->vs.surf_bo[surf] = brw_search_cache(&brw->surface_cache,
+                                            BRW_SS_SURFACE,
+                                            &key, sizeof(key),
+                                            &key.bo, key.bo ? 1 : 0,
+                                            NULL);
+   if (brw->vs.surf_bo[surf] == NULL) {
+      brw->vs.surf_bo[surf] = brw_create_constant_surface(brw, &key);
+   }
+}
+
+
+/**
+ * Constructs the binding table for the VS surface state.
+ */
+static dri_bo *
+brw_vs_get_binding_table(struct brw_context *brw)
+{
+   dri_bo *bind_bo;
+
+   bind_bo = brw_search_cache(&brw->surface_cache, BRW_SS_SURF_BIND,
+			      NULL, 0,
+			      brw->vs.surf_bo, BRW_VS_MAX_SURF,
+			      NULL);
+
+   if (bind_bo == NULL) {
+      GLuint data_size = BRW_VS_MAX_SURF * sizeof(GLuint);
+      uint32_t *data = malloc(data_size);
+      int i;
+
+      for (i = 0; i < BRW_VS_MAX_SURF; i++)
+         if (brw->vs.surf_bo[i])
+            data[i] = brw->vs.surf_bo[i]->offset;
+         else
+            data[i] = 0;
+
+      bind_bo = brw_upload_cache( &brw->surface_cache, BRW_SS_SURF_BIND,
+				  NULL, 0,
+				  brw->vs.surf_bo, BRW_VS_MAX_SURF,
+				  data, data_size,
+				  NULL, NULL);
+
+      /* Emit binding table relocations to surface state */
+      for (i = 0; i < BRW_VS_MAX_SURF; i++) {
+	 if (brw->vs.surf_bo[i] != NULL) {
+	    /* The presumed offsets were set in the data values for
+	     * brw_upload_cache.
+	     */
+	    drm_intel_bo_emit_reloc(bind_bo, i * 4,
+				    brw->vs.surf_bo[i], 0,
+				    I915_GEM_DOMAIN_INSTRUCTION, 0);
+	 }
+      }
+
+      free(data);
+   }
+
+   return bind_bo;
+}
+
+/**
+ * Vertex shader surfaces (constant buffer).
+ *
+ * This consumes the state updates for the constant buffer needing
+ * to be updated, and produces BRW_NEW_NR_VS_SURFACES for the VS unit and
+ * CACHE_NEW_SURF_BIND for the binding table upload.
+ */
+static void prepare_vs_surfaces(struct brw_context *brw )
+{
+   GLcontext *ctx = &brw->intel.ctx;
+   int i;
+   int nr_surfaces = 0;
+
+   brw_update_vs_constant_surface(ctx, SURF_INDEX_VERT_CONST_BUFFER);
+
+   for (i = 0; i < BRW_VS_MAX_SURF; i++) {
+      if (brw->vs.surf_bo[i] != NULL) {
+	 nr_surfaces = i + 1;
+      }
+   }
+
+   if (brw->vs.nr_surfaces != nr_surfaces) {
+      brw->state.dirty.brw |= BRW_NEW_NR_VS_SURFACES;
+      brw->vs.nr_surfaces = nr_surfaces;
+   }
+
+   /* Note that we don't end up updating the bind_bo if we don't have a
+    * surface to be pointing at.  This should be relatively harmless, as it
+    * just slightly increases our working set size.
+    */
+   if (brw->vs.nr_surfaces != 0) {
+      dri_bo_unreference(brw->vs.bind_bo);
+      brw->vs.bind_bo = brw_vs_get_binding_table(brw);
+   }
+}
+
+const struct brw_tracked_state brw_vs_surfaces = {
+   .dirty = {
+      .mesa = (_NEW_PROGRAM_CONSTANTS),
+      .brw = (BRW_NEW_VERTEX_PROGRAM),
+      .cache = 0
+   },
+   .prepare = prepare_vs_surfaces,
+};
+
+
+
diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c
index 90d74c2885c..cd65f57bbc9 100644
--- a/src/mesa/drivers/dri/i965/brw_wm.c
+++ b/src/mesa/drivers/dri/i965/brw_wm.c
@@ -146,6 +146,13 @@ static void do_wm_prog( struct brw_context *brw,
    if (c == NULL) {
       brw->wm.compile_data = calloc(1, sizeof(*brw->wm.compile_data));
       c = brw->wm.compile_data;
+      if (c == NULL) {
+         /* Ouch - big out of memory problem.  Can't continue
+          * without triggering a segfault, no way to signal,
+          * so just return.
+          */
+         return;
+      }
    } else {
       memset(c, 0, sizeof(*brw->wm.compile_data));
    }
diff --git a/src/mesa/drivers/dri/i965/brw_wm.h b/src/mesa/drivers/dri/i965/brw_wm.h
index d0ab3bdc65f..59ead757b51 100644
--- a/src/mesa/drivers/dri/i965/brw_wm.h
+++ b/src/mesa/drivers/dri/i965/brw_wm.h
@@ -240,22 +240,25 @@ struct brw_wm_compile {
    GLuint max_wm_grf;
    GLuint last_scratch;
 
+   GLuint cur_inst;  /**< index of current instruction */
+
+   GLboolean out_of_regs;  /**< ran out of GRF registers? */
+
    /** Mapping from Mesa registers to hardware registers */
    struct {
       GLboolean inited;
       struct brw_reg reg;
    } wm_regs[PROGRAM_PAYLOAD+1][256][4];
 
+   GLboolean used_grf[BRW_WM_MAX_GRF];
+   GLuint first_free_grf;
    struct brw_reg stack;
    struct brw_reg emit_mask_reg;
-   GLuint reg_index;  /**< Index of next free GRF register */
    GLuint tmp_regs[BRW_WM_MAX_GRF];
    GLuint tmp_index;
    GLuint tmp_max;
    GLuint subroutines[BRW_WM_MAX_SUBROUTINE];
 
-   /** using a real constant buffer? */
-   GLboolean use_const_buffer;
    /** we may need up to 3 constants per instruction (if use_const_buffer) */
    struct {
       GLint index;
diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
index 22e17622c6d..efe8b5126c8 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
@@ -1,5 +1,7 @@
 #include "main/macros.h"
 #include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+#include "shader/prog_optimize.h"
 #include "brw_context.h"
 #include "brw_eu.h"
 #include "brw_wm.h"
@@ -42,6 +44,83 @@ GLboolean brw_wm_is_glsl(const struct gl_fragment_program *fp)
 }
 
 
+
+static void
+reclaim_temps(struct brw_wm_compile *c);
+
+
+/** Mark GRF register as used. */
+static void
+prealloc_grf(struct brw_wm_compile *c, int r)
+{
+   c->used_grf[r] = GL_TRUE;
+}
+
+
+/** Mark given GRF register as not in use. */
+static void
+release_grf(struct brw_wm_compile *c, int r)
+{
+   /*assert(c->used_grf[r]);*/
+   c->used_grf[r] = GL_FALSE;
+   c->first_free_grf = MIN2(c->first_free_grf, r);
+}
+
+
+/** Return index of a free GRF, mark it as used. */
+static int
+alloc_grf(struct brw_wm_compile *c)
+{
+   GLuint r;
+   for (r = c->first_free_grf; r < BRW_WM_MAX_GRF; r++) {
+      if (!c->used_grf[r]) {
+         c->used_grf[r] = GL_TRUE;
+         c->first_free_grf = r + 1;  /* a guess */
+         return r;
+      }
+   }
+
+   /* no free temps, try to reclaim some */
+   reclaim_temps(c);
+   c->first_free_grf = 0;
+
+   /* try alloc again */
+   for (r = c->first_free_grf; r < BRW_WM_MAX_GRF; r++) {
+      if (!c->used_grf[r]) {
+         c->used_grf[r] = GL_TRUE;
+         c->first_free_grf = r + 1;  /* a guess */
+         return r;
+      }
+   }
+
+   for (r = 0; r < BRW_WM_MAX_GRF; r++) {
+      assert(c->used_grf[r]);
+   }
+
+   /* really, no free GRF regs found */
+   if (!c->out_of_regs) {
+      /* print warning once per compilation */
+      _mesa_warning(NULL, "i965: ran out of registers for fragment program");
+      c->out_of_regs = GL_TRUE;
+   }
+
+   return -1;
+}
+
+
+/** Return number of GRF registers used */
+static int
+num_grf_used(const struct brw_wm_compile *c)
+{
+   int r;
+   for (r = BRW_WM_MAX_GRF - 1; r >= 0; r--)
+      if (c->used_grf[r])
+         return r + 1;
+   return 0;
+}
+
+
+
 /**
  * Record the mapping of a Mesa register to a hardware register.
  */
@@ -68,11 +147,23 @@ static int get_scalar_dst_index(const struct prog_instruction *inst)
 static struct brw_reg alloc_tmp(struct brw_wm_compile *c)
 {
     struct brw_reg reg;
-    if(c->tmp_index == c->tmp_max)
-	c->tmp_regs[ c->tmp_max++ ] = c->reg_index++;
-    
+
+    /* if we need to allocate another temp, grow the tmp_regs[] array */
+    if (c->tmp_index == c->tmp_max) {
+       int r = alloc_grf(c);
+       if (r < 0) {
+          /*printf("Out of temps in %s\n", __FUNCTION__);*/
+          r = 50; /* XXX random register! */
+       }
+       c->tmp_regs[ c->tmp_max++ ] = r;
+    }
+
+    /* form the GRF register */
     reg = brw_vec8_grf(c->tmp_regs[ c->tmp_index++ ], 0);
+    /*printf("alloc_temp %d\n", reg.nr);*/
+    assert(reg.nr < BRW_WM_MAX_GRF);
     return reg;
+
 }
 
 /**
@@ -130,35 +221,29 @@ get_reg(struct brw_wm_compile *c, int file, int index, int component,
 	    return brw_null_reg();
     }
 
+    assert(index < 256);
+    assert(component < 4);
+
     /* see if we've already allocated a HW register for this Mesa register */
     if (c->wm_regs[file][index][component].inited) {
-	/* yes, re-use */
-	reg = c->wm_regs[file][index][component].reg;
+       /* yes, re-use */
+       reg = c->wm_regs[file][index][component].reg;
     }
     else {
 	/* no, allocate new register */
-	reg = brw_vec8_grf(c->reg_index, 0);
+       int grf = alloc_grf(c);
+       /*printf("alloc grf %d for reg %d:%d.%d\n", grf, file, index, component);*/
+       if (grf < 0) {
+          /* totally out of temps */
+          grf = 51; /* XXX random register! */
+       }
+
+       reg = brw_vec8_grf(grf, 0);
+       /*printf("Alloc new grf %d for %d.%d\n", reg.nr, index, component);*/
+
+       set_reg(c, file, index, component, reg);
     }
 
-    /* if this is a new register allocation, record it in the table */
-    if (!c->wm_regs[file][index][component].inited) {
-	set_reg(c, file, index, component, reg);
-	c->reg_index++;
-    }
-
-    if (c->reg_index >= BRW_WM_MAX_GRF - 12) {
-	/* ran out of temporary registers! */
-#if 1
-        /* This is a big hack for now.
-         * Return bad register index, just don't hang the GPU.
-         */
-        _mesa_fprintf(stderr, "out of regs %d\n", c->reg_index);
-        c->reg_index = BRW_WM_MAX_GRF - 13;
-#else
-	return brw_null_reg();
-#endif
-    }
- 
     if (neg & (1 << component)) {
 	reg = negate(reg);
     }
@@ -168,6 +253,46 @@ get_reg(struct brw_wm_compile *c, int file, int index, int component,
 }
 
 
+
+/**
+ * This is called if we run out of GRF registers.  Examine the live intervals
+ * of temp regs in the program and free those which won't be used again.
+ */
+static void
+reclaim_temps(struct brw_wm_compile *c)
+{
+   GLint intBegin[MAX_PROGRAM_TEMPS];
+   GLint intEnd[MAX_PROGRAM_TEMPS];
+   int index;
+
+   /*printf("Reclaim temps:\n");*/
+
+   _mesa_find_temp_intervals(c->prog_instructions, c->nr_fp_insns,
+                             intBegin, intEnd);
+
+   for (index = 0; index < MAX_PROGRAM_TEMPS; index++) {
+      if (intEnd[index] != -1 && intEnd[index] < c->cur_inst) {
+         /* program temp[i] can be freed */
+         int component;
+         /*printf("  temp[%d] is dead\n", index);*/
+         for (component = 0; component < 4; component++) {
+            if (c->wm_regs[PROGRAM_TEMPORARY][index][component].inited) {
+               int r = c->wm_regs[PROGRAM_TEMPORARY][index][component].reg.nr;
+               release_grf(c, r);
+               /*
+               printf("  Reclaim temp %d, reg %d at inst %d\n",
+                      index, r, c->cur_inst);
+               */
+               c->wm_regs[PROGRAM_TEMPORARY][index][component].inited = GL_FALSE;
+            }
+         }
+      }
+   }
+}
+
+
+
+
 /**
  * Preallocate registers.  This sets up the Mesa to hardware register
  * mapping for certain registers, such as constants (uniforms/state vars)
@@ -179,6 +304,10 @@ static void prealloc_reg(struct brw_wm_compile *c)
     struct brw_reg reg;
     int nr_interp_regs = 0;
     GLuint inputs = FRAG_BIT_WPOS | c->fp_interp_emitted | c->fp_deriv_emitted;
+    GLuint reg_index = 0;
+
+    memset(c->used_grf, GL_FALSE, sizeof(c->used_grf));
+    c->first_free_grf = 0;
 
     for (i = 0; i < 4; i++) {
         if (i < c->key.nr_depth_regs) 
@@ -187,16 +316,22 @@ static void prealloc_reg(struct brw_wm_compile *c)
             reg = brw_vec8_grf(0, 0);
 	set_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, i, reg);
     }
-    c->reg_index += 2 * c->key.nr_depth_regs;
+    reg_index += 2 * c->key.nr_depth_regs;
 
     /* constants */
     {
-        const int nr_params = c->fp->program.Base.Parameters->NumParameters;
+        const GLuint nr_params = c->fp->program.Base.Parameters->NumParameters;
+        const GLuint nr_temps = c->fp->program.Base.NumTemporaries;
 
         /* use a real constant buffer, or just use a section of the GRF? */
-        c->use_const_buffer = GL_FALSE; /* (nr_params > 8);*/
+        /* XXX this heuristic may need adjustment... */
+        if ((nr_params + nr_temps) * 4 + reg_index > 80)
+           c->fp->use_const_buffer = GL_TRUE;
+        else
+           c->fp->use_const_buffer = GL_FALSE;
+        /*printf("WM use_const_buffer = %d\n", c->fp->use_const_buffer);*/
 
-        if (c->use_const_buffer) {
+        if (c->fp->use_const_buffer) {
            /* We'll use a real constant buffer and fetch constants from
             * it with a dataport read message.
             */
@@ -216,7 +351,7 @@ static void prealloc_reg(struct brw_wm_compile *c)
            for (i = 0; i < nr_params; i++) {
               /* loop over XYZW channels */
               for (j = 0; j < 4; j++, index++) {
-                 reg = brw_vec1_grf(c->reg_index + index / 8, index % 8);
+                 reg = brw_vec1_grf(reg_index + index / 8, index % 8);
                  /* Save pointer to parameter/constant value.
                   * Constants will be copied in prepare_constant_buffer()
                   */
@@ -226,7 +361,7 @@ static void prealloc_reg(struct brw_wm_compile *c)
            }
            /* number of constant regs used (each reg is float[8]) */
            c->nr_creg = 2 * ((4 * nr_params + 15) / 16);
-           c->reg_index += c->nr_creg;
+           reg_index += c->nr_creg;
         }
     }
 
@@ -234,34 +369,42 @@ static void prealloc_reg(struct brw_wm_compile *c)
     for (i = 0; i < FRAG_ATTRIB_MAX; i++) {
 	if (inputs & (1<<i)) {
 	    nr_interp_regs++;
-	    reg = brw_vec8_grf(c->reg_index, 0);
+	    reg = brw_vec8_grf(reg_index, 0);
 	    for (j = 0; j < 4; j++)
 		set_reg(c, PROGRAM_PAYLOAD, i, j, reg);
-	    c->reg_index += 2;
+	    reg_index += 2;
 	}
     }
 
     c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2;
     c->prog_data.urb_read_length = nr_interp_regs * 2;
     c->prog_data.curb_read_length = c->nr_creg;
-    c->emit_mask_reg = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0);
-    c->reg_index++;
-    c->stack =  brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0);
-    c->reg_index += 2;
+    c->emit_mask_reg = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, reg_index, 0);
+    reg_index++;
+    c->stack =  brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, reg_index, 0);
+    reg_index += 2;
+
+    /* mark GRF regs [0..reg_index-1] as in-use */
+    for (i = 0; i < reg_index; i++)
+       prealloc_grf(c, i);
+
+    /* Don't use GRF 126, 127.  Using them seems to lead to GPU lock-ups */
+    prealloc_grf(c, 126);
+    prealloc_grf(c, 127);
 
     /* An instruction may reference up to three constants.
      * They'll be found in these registers.
      * XXX alloc these on demand!
      */
-    if (c->use_const_buffer) {
+    if (c->fp->use_const_buffer) {
        for (i = 0; i < 3; i++) {
           c->current_const[i].index = -1;
-          c->current_const[i].reg = alloc_tmp(c);
+          c->current_const[i].reg = brw_vec8_grf(alloc_grf(c), 0);
        }
     }
 #if 0
-    printf("USE CONST BUFFER? %d\n", c->use_const_buffer);
-    printf("AFTER PRE_ALLOC, reg_index = %d\n", c->reg_index);
+    printf("USE CONST BUFFER? %d\n", c->fp->use_const_buffer);
+    printf("AFTER PRE_ALLOC, reg_index = %d\n", reg_index);
 #endif
 }
 
@@ -283,23 +426,21 @@ static void fetch_constants(struct brw_wm_compile *c,
       if (src->File == PROGRAM_STATE_VAR ||
           src->File == PROGRAM_CONSTANT ||
           src->File == PROGRAM_UNIFORM) {
-         if (c->current_const[i].index != src->Index) {
-            c->current_const[i].index = src->Index;
+	 c->current_const[i].index = src->Index;
 
 #if 0
-            printf("  fetch const[%d] for arg %d into reg %d\n",
-                   src->Index, i, c->current_const[i].reg.nr);
+	 printf("  fetch const[%d] for arg %d into reg %d\n",
+		src->Index, i, c->current_const[i].reg.nr);
 #endif
 
-            /* need to fetch the constant now */
-            brw_dp_READ_4(p,
-                          c->current_const[i].reg,  /* writeback dest */
-                          1,                        /* msg_reg */
-                          src->RelAddr,             /* relative indexing? */
-                          16 * src->Index,          /* byte offset */
-                          SURF_INDEX_FRAG_CONST_BUFFER/* binding table index */
-                          );
-         }
+	 /* need to fetch the constant now */
+	 brw_dp_READ_4(p,
+		       c->current_const[i].reg,  /* writeback dest */
+		       1,                        /* msg_reg */
+		       src->RelAddr,             /* relative indexing? */
+		       16 * src->Index,          /* byte offset */
+		       SURF_INDEX_FRAG_CONST_BUFFER/* binding table index */
+		       );
       }
    }
 }
@@ -368,7 +509,7 @@ static struct brw_reg get_src_reg(struct brw_wm_compile *c,
     const GLuint nr = 1;
     const GLuint component = GET_SWZ(src->Swizzle, channel);
 
-    if (c->use_const_buffer &&
+    if (c->fp->use_const_buffer &&
         (src->File == PROGRAM_STATE_VAR ||
          src->File == PROGRAM_CONSTANT ||
          src->File == PROGRAM_UNIFORM)) {
@@ -2595,7 +2736,8 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
     struct brw_compile *p = &c->func;
     struct brw_indirect stack_index = brw_indirect(0, 0);
 
-    c->reg_index = 0;
+    c->out_of_regs = GL_FALSE;
+
     prealloc_reg(c);
     brw_set_compression_control(p, BRW_COMPRESSION_NONE);
     brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack));
@@ -2603,13 +2745,15 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
     for (i = 0; i < c->nr_fp_insns; i++) {
         const struct prog_instruction *inst = &c->prog_instructions[i];
 
+        c->cur_inst = i;
+
 #if 0
         _mesa_printf("Inst %d: ", i);
         _mesa_print_instruction(inst);
 #endif
 
         /* fetch any constants that this instruction needs */
-        if (c->use_const_buffer)
+        if (c->fp->use_const_buffer)
            fetch_constants(c, inst);
 
 	if (inst->CondUpdate)
@@ -2833,17 +2977,13 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
 		_mesa_printf("unsupported IR in fragment shader %d\n",
 			inst->Opcode);
 	}
+
 	if (inst->CondUpdate)
 	    brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
 	else
 	    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
     }
     post_wm_emit(c);
-
-    if (c->reg_index >= BRW_WM_MAX_GRF) {
-        _mesa_problem(NULL, "Ran out of registers in brw_wm_emit_glsl()");
-        /* XXX we need to do some proper error recovery here */
-    }
 }
 
 
@@ -2867,6 +3007,6 @@ void brw_wm_glsl_emit(struct brw_context *brw, struct brw_wm_compile *c)
         brw_wm_print_program(c, "brw_wm_glsl_emit done");
     }
 
-    c->prog_data.total_grf = c->reg_index;
+    c->prog_data.total_grf = num_grf_used(c);
     c->prog_data.total_scratch = 0;
 }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
index 1fc9f013727..c604ef0162a 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
@@ -152,7 +152,7 @@ static void brw_update_sampler_state(struct wm_sampler_entry *key,
       sampler->ss0.mag_filter = BRW_MAPFILTER_ANISOTROPIC;
 
       if (key->max_aniso > 2.0) {
-	 sampler->ss3.max_aniso = MAX2((key->max_aniso - 2) / 2,
+	 sampler->ss3.max_aniso = MIN2((key->max_aniso - 2) / 2,
 				       BRW_ANISORATIO_16);
       }
    }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
index 71840d1e4e8..c49a5f6b4ec 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
@@ -176,22 +176,6 @@ static GLuint translate_tex_format( GLuint mesa_format, GLenum internal_format,
    }
 }
 
-
-/**
- * Use same key for WM and VS surfaces.
- */
-struct brw_surface_key {
-   GLenum target, depthmode;
-   dri_bo *bo;
-   GLint format, internal_format;
-   GLint first_level, last_level;
-   GLint width, height, depth;
-   GLint pitch, cpp;
-   uint32_t tiling;
-   GLuint offset;
-};
-
-
 static void
 brw_set_surface_tiling(struct brw_surface_state *surf, uint32_t tiling)
 {
@@ -268,7 +252,7 @@ brw_create_texture_surface( struct brw_context *brw,
       surf.ss0.cube_neg_z = 1;
    }
 
-   bo = brw_upload_cache(&brw->cache, BRW_SS_SURFACE,
+   bo = brw_upload_cache(&brw->surface_cache, BRW_SS_SURFACE,
 			 key, sizeof(*key),
 			 &key->bo, key->bo ? 1 : 0,
 			 &surf, sizeof(surf),
@@ -321,10 +305,11 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit )
    key.tiling = intelObj->mt->region->tiling;
 
    dri_bo_unreference(brw->wm.surf_bo[surf]);
-   brw->wm.surf_bo[surf] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
-                                         &key, sizeof(key),
-                                         &key.bo, key.bo ? 1 : 0,
-                                         NULL);
+   brw->wm.surf_bo[surf] = brw_search_cache(&brw->surface_cache,
+                                            BRW_SS_SURFACE,
+                                            &key, sizeof(key),
+                                            &key.bo, key.bo ? 1 : 0,
+                                            NULL);
    if (brw->wm.surf_bo[surf] == NULL) {
       brw->wm.surf_bo[surf] = brw_create_texture_surface(brw, &key);
    }
@@ -336,7 +321,7 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit )
  * Create the constant buffer surface.  Vertex/fragment shader constants will be
  * read from this buffer with Data Port Read instructions/messages.
  */
-static dri_bo *
+dri_bo *
 brw_create_constant_surface( struct brw_context *brw,
                              struct brw_surface_key *key )
 {
@@ -362,7 +347,7 @@ brw_create_constant_surface( struct brw_context *brw,
    surf.ss3.pitch = (key->pitch * key->cpp) - 1; /* ignored?? */
    brw_set_surface_tiling(&surf, key->tiling); /* tiling now allowed */
  
-   bo = brw_upload_cache(&brw->cache, BRW_SS_SURFACE,
+   bo = brw_upload_cache(&brw->surface_cache, BRW_SS_SURFACE,
 			 key, sizeof(*key),
 			 &key->bo, key->bo ? 1 : 0,
 			 &surf, sizeof(surf),
@@ -380,39 +365,70 @@ brw_create_constant_surface( struct brw_context *brw,
    return bo;
 }
 
+/* Creates a new WM constant buffer reflecting the current fragment program's
+ * constants, if needed by the fragment program.
+ *
+ * Otherwise, constants go through the CURBEs using the brw_constant_buffer
+ * state atom.
+ */
+static drm_intel_bo *
+brw_wm_update_constant_buffer(struct brw_context *brw)
+{
+   struct intel_context *intel = &brw->intel;
+   struct brw_fragment_program *fp =
+      (struct brw_fragment_program *) brw->fragment_program;
+   const struct gl_program_parameter_list *params = fp->program.Base.Parameters;
+   const int size = params->NumParameters * 4 * sizeof(GLfloat);
+   drm_intel_bo *const_buffer;
+
+   /* BRW_NEW_FRAGMENT_PROGRAM */
+   if (!fp->use_const_buffer)
+      return NULL;
+
+   const_buffer = drm_intel_bo_alloc(intel->bufmgr, "fp_const_buffer",
+				     size, 64);
+
+   /* _NEW_PROGRAM_CONSTANTS */
+   dri_bo_subdata(const_buffer, 0, size, params->ParameterValues);
+
+   return const_buffer;
+}
 
 /**
  * Update the surface state for a WM constant buffer.
  * The constant buffer will be (re)allocated here if needed.
  */
-static dri_bo *
+static void
 brw_update_wm_constant_surface( GLcontext *ctx,
-                                GLuint surf,
-                                dri_bo *const_buffer,
-                                const struct gl_program_parameter_list *params)
+                                GLuint surf)
 {
    struct brw_context *brw = brw_context(ctx);
    struct brw_surface_key key;
-   struct intel_context *intel = &brw->intel;
-   const int size = params->NumParameters * 4 * sizeof(GLfloat);
+   struct brw_fragment_program *fp =
+      (struct brw_fragment_program *) brw->fragment_program;
+   const struct gl_program_parameter_list *params =
+      fp->program.Base.Parameters;
 
-   /* free old const buffer if too small */
-   if (const_buffer && const_buffer->size < size) {
-      dri_bo_unreference(const_buffer);
-      const_buffer = NULL;
-   }
+   /* If we're in this state update atom, we need to update WM constants, so
+    * free the old buffer and create a new one for the new contents.
+    */
+   dri_bo_unreference(fp->const_buffer);
+   fp->const_buffer = brw_wm_update_constant_buffer(brw);
 
-   /* alloc new buffer if needed */
-   if (!const_buffer) {
-      const_buffer =
-         drm_intel_bo_alloc(intel->bufmgr, "fp_const_buffer", size, 64);
+   /* If there's no constant buffer, then no surface BO is needed to point at
+    * it.
+    */
+   if (fp->const_buffer == 0) {
+      drm_intel_bo_unreference(brw->wm.surf_bo[surf]);
+      brw->wm.surf_bo[surf] = NULL;
+      return;
    }
 
    memset(&key, 0, sizeof(key));
 
    key.format = MESA_FORMAT_RGBA_FLOAT32;
    key.internal_format = GL_RGBA;
-   key.bo = const_buffer;
+   key.bo = fp->const_buffer;
    key.depthmode = GL_NONE;
    key.pitch = params->NumParameters;
    key.width = params->NumParameters;
@@ -427,77 +443,59 @@ brw_update_wm_constant_surface( GLcontext *ctx,
    */
 
    dri_bo_unreference(brw->wm.surf_bo[surf]);
-   brw->wm.surf_bo[surf] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
+   brw->wm.surf_bo[surf] = brw_search_cache(&brw->surface_cache,
+                                            BRW_SS_SURFACE,
                                             &key, sizeof(key),
                                             &key.bo, key.bo ? 1 : 0,
                                             NULL);
    if (brw->wm.surf_bo[surf] == NULL) {
       brw->wm.surf_bo[surf] = brw_create_constant_surface(brw, &key);
    }
-
-   return const_buffer;
+   brw->state.dirty.brw |= BRW_NEW_WM_SURFACES;
 }
 
-
 /**
- * Update the surface state for a VS constant buffer.
- * The constant buffer will be (re)allocated here if needed.
+ * Updates surface / buffer for fragment shader constant buffer, if
+ * one is required.
+ *
+ * This consumes the state updates for the constant buffer, and produces
+ * BRW_NEW_WM_SURFACES to get picked up by brw_prepare_wm_surfaces for
+ * inclusion in the binding table.
  */
-static dri_bo *
-brw_update_vs_constant_surface( GLcontext *ctx,
-                                GLuint surf,
-                                dri_bo *const_buffer,
-                                const struct gl_program_parameter_list *params)
+static void prepare_wm_constant_surface(struct brw_context *brw )
 {
-   struct brw_context *brw = brw_context(ctx);
-   struct brw_surface_key key;
-   struct intel_context *intel = &brw->intel;
-   const int size = params->NumParameters * 4 * sizeof(GLfloat);
+   GLcontext *ctx = &brw->intel.ctx;
+   struct brw_fragment_program *fp =
+      (struct brw_fragment_program *) brw->fragment_program;
+   GLuint surf = SURF_INDEX_FRAG_CONST_BUFFER;
 
-   assert(surf == 0);
+   drm_intel_bo_unreference(fp->const_buffer);
+   fp->const_buffer = brw_wm_update_constant_buffer(brw);
 
-   /* free old const buffer if too small */
-   if (const_buffer && const_buffer->size < size) {
-      dri_bo_unreference(const_buffer);
-      const_buffer = NULL;
+   /* If there's no constant buffer, then no surface BO is needed to point at
+    * it.
+    */
+   if (fp->const_buffer == 0) {
+      if (brw->wm.surf_bo[surf] != NULL) {
+	 drm_intel_bo_unreference(brw->wm.surf_bo[surf]);
+	 brw->wm.surf_bo[surf] = NULL;
+	 brw->state.dirty.brw |= BRW_NEW_WM_SURFACES;
+      }
+      return;
    }
 
-   /* alloc new buffer if needed */
-   if (!const_buffer) {
-      const_buffer =
-         drm_intel_bo_alloc(intel->bufmgr, "vp_const_buffer", size, 64);
-   }
-
-   memset(&key, 0, sizeof(key));
-
-   key.format = MESA_FORMAT_RGBA_FLOAT32;
-   key.internal_format = GL_RGBA;
-   key.bo = const_buffer;
-   key.depthmode = GL_NONE;
-   key.pitch = params->NumParameters;
-   key.width = params->NumParameters;
-   key.height = 1;
-   key.depth = 1;
-   key.cpp = 16;
-
-   /*
-   printf("%s:\n", __FUNCTION__);
-   printf("  width %d  height %d  depth %d  cpp %d  pitch %d\n",
-          key.width, key.height, key.depth, key.cpp, key.pitch);
-   */
-
-   dri_bo_unreference(brw->vs.surf_bo[surf]);
-   brw->vs.surf_bo[surf] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
-                                            &key, sizeof(key),
-                                            &key.bo, key.bo ? 1 : 0,
-                                            NULL);
-   if (brw->vs.surf_bo[surf] == NULL) {
-      brw->vs.surf_bo[surf] = brw_create_constant_surface(brw, &key);
-   }
-
-   return const_buffer;
+   brw_update_wm_constant_surface(ctx, surf);
 }
 
+const struct brw_tracked_state brw_wm_constant_surface = {
+   .dirty = {
+      .mesa = (_NEW_PROGRAM_CONSTANTS),
+      .brw = (BRW_NEW_FRAGMENT_PROGRAM),
+      .cache = 0
+   },
+   .prepare = prepare_wm_constant_surface,
+};
+
 
 /**
  * Sets up a surface state structure to point at the given region.
@@ -507,7 +505,7 @@ brw_update_vs_constant_surface( GLcontext *ctx,
 static void
 brw_update_renderbuffer_surface(struct brw_context *brw,
 				struct gl_renderbuffer *rb,
-				unsigned int unit, GLboolean cached)
+				unsigned int unit)
 {
    GLcontext *ctx = &brw->intel.ctx;
    dri_bo *region_bo = NULL;
@@ -520,6 +518,7 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
       GLubyte color_mask[4];
       GLboolean color_blend;
       uint32_t tiling;
+      uint32_t draw_offset;
    } key;
 
    memset(&key, 0, sizeof(key));
@@ -550,6 +549,7 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
       key.height = region->height;
       key.pitch = region->pitch;
       key.cpp = region->cpp;
+      key.draw_offset = region->draw_offset; /* cur 3d or cube face offset */
    } else {
       key.surface_type = BRW_SURFACE_NULL;
       key.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM;
@@ -557,6 +557,7 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
       key.width = 1;
       key.height = 1;
       key.cpp = 4;
+      key.draw_offset = 0;
    }
    memcpy(key.color_mask, ctx->Color.ColorMask,
 	  sizeof(key.color_mask));
@@ -564,12 +565,11 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
 		      ctx->Color.BlendEnabled);
 
    dri_bo_unreference(brw->wm.surf_bo[unit]);
-   brw->wm.surf_bo[unit] = NULL;
-   if (cached) 
-       brw->wm.surf_bo[unit] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
-	       &key, sizeof(key),
-	       &region_bo, 1,
-	       NULL);
+   brw->wm.surf_bo[unit] = brw_search_cache(&brw->surface_cache,
+					    BRW_SS_SURFACE,
+					    &key, sizeof(key),
+					    &region_bo, 1,
+					    NULL);
 
    if (brw->wm.surf_bo[unit] == NULL) {
       struct brw_surface_state surf;
@@ -578,8 +578,9 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
 
       surf.ss0.surface_format = key.surface_format;
       surf.ss0.surface_type = key.surface_type;
+      surf.ss1.base_addr =  key.draw_offset;
       if (region_bo != NULL)
-	 surf.ss1.base_addr = region_bo->offset; /* reloc */
+	 surf.ss1.base_addr += region_bo->offset; /* reloc */
 
       surf.ss2.width = key.width - 1;
       surf.ss2.height = key.height - 1;
@@ -594,7 +595,8 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
       surf.ss0.writedisable_alpha = !key.color_mask[3];
 
       /* Key size will never match key size for textures, so we're safe. */
-      brw->wm.surf_bo[unit] = brw_upload_cache(&brw->cache, BRW_SS_SURFACE,
+      brw->wm.surf_bo[unit] = brw_upload_cache(&brw->surface_cache,
+                                               BRW_SS_SURFACE,
                                                &key, sizeof(key),
 					       &region_bo, 1,
 					       &surf, sizeof(surf),
@@ -604,12 +606,12 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
 	  * them both.  We might be able to figure out from other state
 	  * a more restrictive relocation to emit.
 	  */
-	 dri_bo_emit_reloc(brw->wm.surf_bo[unit],
-			   I915_GEM_DOMAIN_RENDER,
-			   I915_GEM_DOMAIN_RENDER,
-			   0,
-			   offsetof(struct brw_surface_state, ss1),
-			   region_bo);
+	 drm_intel_bo_emit_reloc(brw->wm.surf_bo[unit],
+				 offsetof(struct brw_surface_state, ss1),
+				 region_bo,
+				 key.draw_offset,
+				 I915_GEM_DOMAIN_RENDER,
+				 I915_GEM_DOMAIN_RENDER);
       }
    }
 }
@@ -626,7 +628,7 @@ brw_wm_get_binding_table(struct brw_context *brw)
 
    assert(brw->wm.nr_surfaces <= BRW_WM_MAX_SURF);
 
-   bind_bo = brw_search_cache(&brw->cache, BRW_SS_SURF_BIND,
+   bind_bo = brw_search_cache(&brw->surface_cache, BRW_SS_SURF_BIND,
 			      NULL, 0,
 			      brw->wm.surf_bo, brw->wm.nr_surfaces,
 			      NULL);
@@ -642,7 +644,7 @@ brw_wm_get_binding_table(struct brw_context *brw)
          else
             data[i] = 0;
 
-      bind_bo = brw_upload_cache( &brw->cache, BRW_SS_SURF_BIND,
+      bind_bo = brw_upload_cache( &brw->surface_cache, BRW_SS_SURF_BIND,
 				  NULL, 0,
 				  brw->wm.surf_bo, brw->wm.nr_surfaces,
 				  data, data_size,
@@ -678,27 +680,17 @@ static void prepare_wm_surfaces(struct brw_context *brw )
       for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) {
          brw_update_renderbuffer_surface(brw,
 					 ctx->DrawBuffer->_ColorDrawBuffers[i],
-					 i,
-					 GL_FALSE);
+					 i);
       }
    } else {
-      brw_update_renderbuffer_surface(brw, NULL, 0, GL_TRUE);
+      brw_update_renderbuffer_surface(brw, NULL, 0);
    }
 
    old_nr_surfaces = brw->wm.nr_surfaces;
    brw->wm.nr_surfaces = MAX_DRAW_BUFFERS;
 
-   /* Update surface / buffer for fragment shader constant buffer */
-   {
-      const GLuint surf = SURF_INDEX_FRAG_CONST_BUFFER;
-      struct brw_fragment_program *fp =
-         (struct brw_fragment_program *) brw->fragment_program;
-      fp->const_buffer =
-         brw_update_wm_constant_surface(ctx, surf, fp->const_buffer,
-                                     fp->program.Base.Parameters);
-
-      brw->wm.nr_surfaces = surf + 1;
-   }
+   if (brw->wm.surf_bo[SURF_INDEX_FRAG_CONST_BUFFER] != NULL)
+       brw->wm.nr_surfaces = SURF_INDEX_FRAG_CONST_BUFFER + 1;
 
    /* Update surfaces for textures */
    for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
@@ -731,100 +723,16 @@ static void prepare_wm_surfaces(struct brw_context *brw )
       brw->state.dirty.brw |= BRW_NEW_NR_WM_SURFACES;
 }
 
-
-/**
- * Constructs the binding table for the VS surface state.
- */
-static dri_bo *
-brw_vs_get_binding_table(struct brw_context *brw)
-{
-   dri_bo *bind_bo;
-
-   assert(brw->vs.nr_surfaces <= BRW_VS_MAX_SURF);
-
-   bind_bo = brw_search_cache(&brw->cache, BRW_SS_SURF_BIND,
-			      NULL, 0,
-			      brw->vs.surf_bo, brw->vs.nr_surfaces,
-			      NULL);
-
-   if (bind_bo == NULL) {
-      GLuint data_size = brw->vs.nr_surfaces * sizeof(GLuint);
-      uint32_t *data = malloc(data_size);
-      int i;
-
-      for (i = 0; i < brw->vs.nr_surfaces; i++)
-         if (brw->vs.surf_bo[i])
-            data[i] = brw->vs.surf_bo[i]->offset;
-         else
-            data[i] = 0;
-
-      bind_bo = brw_upload_cache( &brw->cache, BRW_SS_SURF_BIND,
-				  NULL, 0,
-				  brw->vs.surf_bo, brw->vs.nr_surfaces,
-				  data, data_size,
-				  NULL, NULL);
-
-      /* Emit binding table relocations to surface state */
-      for (i = 0; i < BRW_VS_MAX_SURF; i++) {
-	 if (brw->vs.surf_bo[i] != NULL) {
-	    dri_bo_emit_reloc(bind_bo,
-			      I915_GEM_DOMAIN_INSTRUCTION, 0,
-			      0,
-			      i * sizeof(GLuint),
-			      brw->vs.surf_bo[i]);
-	 }
-      }
-
-      free(data);
-   }
-
-   return bind_bo;
-}
-
-
-/**
- * Vertex shader surfaces.  Just constant buffer for now.  Could add vertex 
- * shader textures in the future.
- */
-static void prepare_vs_surfaces(struct brw_context *brw )
-{
-   GLcontext *ctx = &brw->intel.ctx;
-
-   /* Update surface / buffer for vertex shader constant buffer */
-   {
-      const GLuint surf = SURF_INDEX_VERT_CONST_BUFFER;
-      struct brw_vertex_program *vp =
-         (struct brw_vertex_program *) brw->vertex_program;
-      vp->const_buffer =
-         brw_update_vs_constant_surface(ctx, surf, vp->const_buffer,
-                                        vp->program.Base.Parameters);
-
-      brw->vs.nr_surfaces = 1;
-   }
-
-   dri_bo_unreference(brw->vs.bind_bo);
-   brw->vs.bind_bo = brw_vs_get_binding_table(brw);
-
-   if (1)
-      brw->state.dirty.brw |= BRW_NEW_NR_VS_SURFACES;
-}
-
-
-static void
-prepare_surfaces(struct brw_context *brw)
-{
-   prepare_wm_surfaces(brw);
-   prepare_vs_surfaces(brw);
-}
-
-
 const struct brw_tracked_state brw_wm_surfaces = {
    .dirty = {
-      .mesa = _NEW_COLOR | _NEW_TEXTURE | _NEW_BUFFERS | _NEW_PROGRAM,
-      .brw = BRW_NEW_CONTEXT,
+      .mesa = (_NEW_COLOR |
+               _NEW_TEXTURE |
+               _NEW_BUFFERS),
+      .brw = (BRW_NEW_CONTEXT |
+	      BRW_NEW_WM_SURFACES),
       .cache = 0
    },
-   .prepare = prepare_surfaces,
+   .prepare = prepare_wm_surfaces,
 };
 
 
diff --git a/src/mesa/drivers/dri/i965/intel_generatemipmap.c b/src/mesa/drivers/dri/i965/intel_generatemipmap.c
new file mode 120000
index 00000000000..4c6b37ada01
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/intel_generatemipmap.c
@@ -0,0 +1 @@
+../intel/intel_generatemipmap.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/intel/intel_buffer_objects.c b/src/mesa/drivers/dri/intel/intel_buffer_objects.c
index b7c7eeb368f..f6b0d769c6b 100644
--- a/src/mesa/drivers/dri/intel/intel_buffer_objects.c
+++ b/src/mesa/drivers/dri/intel/intel_buffer_objects.c
@@ -35,6 +35,9 @@
 #include "intel_batchbuffer.h"
 #include "intel_regions.h"
 
+static GLboolean
+intel_bufferobj_unmap(GLcontext * ctx,
+                      GLenum target, struct gl_buffer_object *obj);
 
 /** Allocates a new dri_bo to store the data for the buffer object. */
 static void
@@ -100,7 +103,13 @@ intel_bufferobj_free(GLcontext * ctx, struct gl_buffer_object *obj)
    struct intel_buffer_object *intel_obj = intel_buffer_object(obj);
 
    assert(intel_obj);
-   assert(!obj->Pointer); /* Mesa should have unmapped it */
+
+   /* Buffer objects are automatically unmapped when deleting according
+    * to the spec, but Mesa doesn't do UnmapBuffer for us at context destroy
+    * (though it does if you call glDeleteBuffers)
+    */
+   if (obj->Pointer)
+      intel_bufferobj_unmap(ctx, 0, obj);
 
    if (intel_obj->region) {
       intel_bufferobj_release_region(intel, intel_obj);
@@ -204,9 +213,8 @@ intel_bufferobj_map(GLcontext * ctx,
 {
    struct intel_context *intel = intel_context(ctx);
    struct intel_buffer_object *intel_obj = intel_buffer_object(obj);
+   GLboolean read_only = (access == GL_READ_ONLY_ARB);
 
-   /* XXX: Translate access to flags arg below:
-    */
    assert(intel_obj);
 
    if (intel_obj->region)
@@ -217,7 +225,7 @@ intel_bufferobj_map(GLcontext * ctx,
       return NULL;
    }
 
-   dri_bo_map(intel_obj->buffer, GL_TRUE);
+   dri_bo_map(intel_obj->buffer, !read_only);
    obj->Pointer = intel_obj->buffer->virtual;
    return obj->Pointer;
 }
diff --git a/src/mesa/drivers/dri/intel/intel_buffers.c b/src/mesa/drivers/dri/intel/intel_buffers.c
index 90964df3553..4f4ea45b74f 100644
--- a/src/mesa/drivers/dri/intel/intel_buffers.c
+++ b/src/mesa/drivers/dri/intel/intel_buffers.c
@@ -157,7 +157,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
    /* Do this here, not core Mesa, since this function is called from
     * many places within the driver.
     */
-   if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
+   if (ctx->NewState & _NEW_BUFFERS) {
       /* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */
       _mesa_update_framebuffer(ctx);
       /* this updates the DrawBuffer's Width/Height if it's a FBO */
@@ -323,8 +323,19 @@ intelDrawBuffer(GLcontext * ctx, GLenum mode)
 {
    if ((ctx->DrawBuffer != NULL) && (ctx->DrawBuffer->Name == 0)) {
       struct intel_context *const intel = intel_context(ctx);
+      const GLboolean was_front_buffer_rendering =
+	intel->is_front_buffer_rendering;
 
-      intel->is_front_buffer_rendering = (mode == GL_FRONT_LEFT);
+      intel->is_front_buffer_rendering = (mode == GL_FRONT_LEFT)
+	|| (mode == GL_FRONT);
+
+      /* If we weren't front-buffer rendering before but we are now, make sure
+       * that the front-buffer has actually been allocated.
+       */
+      if (!was_front_buffer_rendering && intel->is_front_buffer_rendering) {
+	 intel_update_renderbuffers(intel->driContext,
+				    intel->driContext->driDrawablePriv);
+      }
    }
 
    intel_draw_buffer(ctx, ctx->DrawBuffer);
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index 3436b8ecd30..8b3e50f9b62 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -173,6 +173,24 @@ intelGetString(GLcontext * ctx, GLenum name)
    }
 }
 
+static unsigned
+intel_bits_per_pixel(const struct intel_renderbuffer *rb)
+{
+   switch (rb->Base._ActualFormat) {
+   case GL_RGB5:
+   case GL_DEPTH_COMPONENT16:
+      return 16;
+   case GL_RGB8:
+   case GL_RGBA8:
+   case GL_DEPTH_COMPONENT24:
+   case GL_DEPTH24_STENCIL8_EXT:
+   case GL_STENCIL_INDEX8_EXT:
+      return 32;
+   default:
+      return 0;
+   }
+}
+
 void
 intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 {
@@ -180,7 +198,7 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
    struct intel_renderbuffer *rb;
    struct intel_region *region, *depth_region;
    struct intel_context *intel = context->driverPrivate;
-   __DRIbuffer *buffers;
+   __DRIbuffer *buffers = NULL;
    __DRIscreen *screen;
    int i, count;
    unsigned int attachments[10];
@@ -192,22 +210,63 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 
    screen = intel->intelScreen->driScrnPriv;
 
-   i = 0;
-   if (intel_fb->color_rb[0])
-      attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
-   if (intel_fb->color_rb[1])
-      attachments[i++] = __DRI_BUFFER_BACK_LEFT;
-   if (intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH))
-      attachments[i++] = __DRI_BUFFER_DEPTH;
-   if (intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL))
-      attachments[i++] = __DRI_BUFFER_STENCIL;
+   if (screen->dri2.loader
+       && (screen->dri2.loader->base.version > 2)
+       && (screen->dri2.loader->getBuffersWithFormat != NULL)) {
+      struct intel_renderbuffer *depth_rb;
+      struct intel_renderbuffer *stencil_rb;
 
-   buffers = (*screen->dri2.loader->getBuffers)(drawable,
-						&drawable->w,
-						&drawable->h,
-						attachments, i,
-						&count,
-						drawable->loaderPrivate);
+      i = 0;
+      if ((intel->is_front_buffer_rendering || !intel_fb->color_rb[1])
+	   && intel_fb->color_rb[0]) {
+	 attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
+	 attachments[i++] = intel_bits_per_pixel(intel_fb->color_rb[0]);
+      }
+
+      if (intel_fb->color_rb[1]) {
+	 attachments[i++] = __DRI_BUFFER_BACK_LEFT;
+	 attachments[i++] = intel_bits_per_pixel(intel_fb->color_rb[1]);
+      }
+
+      depth_rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
+      stencil_rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL);
+
+      if ((depth_rb != NULL) && (stencil_rb != NULL)) {
+	 attachments[i++] = __DRI_BUFFER_DEPTH_STENCIL;
+	 attachments[i++] = intel_bits_per_pixel(depth_rb);
+      } else if (depth_rb != NULL) {
+	 attachments[i++] = __DRI_BUFFER_DEPTH;
+	 attachments[i++] = intel_bits_per_pixel(depth_rb);
+      } else if (stencil_rb != NULL) {
+	 attachments[i++] = __DRI_BUFFER_STENCIL;
+	 attachments[i++] = intel_bits_per_pixel(stencil_rb);
+      }
+
+      buffers =
+	 (*screen->dri2.loader->getBuffersWithFormat)(drawable,
+						      &drawable->w,
+						      &drawable->h,
+						      attachments, i / 2,
+						      &count,
+						      drawable->loaderPrivate);
+   } else if (screen->dri2.loader) {
+      i = 0;
+      if (intel_fb->color_rb[0])
+	 attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
+      if (intel_fb->color_rb[1])
+	 attachments[i++] = __DRI_BUFFER_BACK_LEFT;
+      if (intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH))
+	 attachments[i++] = __DRI_BUFFER_DEPTH;
+      if (intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL))
+	 attachments[i++] = __DRI_BUFFER_STENCIL;
+
+      buffers = (*screen->dri2.loader->getBuffers)(drawable,
+						   &drawable->w,
+						   &drawable->h,
+						   attachments, i,
+						   &count,
+						   drawable->loaderPrivate);
+   }
 
    if (buffers == NULL)
       return;
@@ -250,6 +309,11 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 	   region_name = "dri2 depth buffer";
 	   break;
 
+       case __DRI_BUFFER_DEPTH_STENCIL:
+	   rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
+	   region_name = "dri2 depth / stencil buffer";
+	   break;
+
        case __DRI_BUFFER_STENCIL:
 	   rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL);
 	   region_name = "dri2 stencil buffer";
@@ -296,6 +360,23 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 
        intel_renderbuffer_set_region(rb, region);
        intel_region_release(&region);
+
+       if (buffers[i].attachment == __DRI_BUFFER_DEPTH_STENCIL) {
+	  rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL);
+	  if (rb != NULL) {
+	     struct intel_region *stencil_region = NULL;
+
+	     if (rb->region) {
+		dri_bo_flink(rb->region->buffer, &name);
+		if (name == buffers[i].name)
+		   continue;
+	     }
+
+	     intel_region_reference(&stencil_region, region);
+	     intel_renderbuffer_set_region(rb, stencil_region);
+	     intel_region_release(&stencil_region);
+	  }
+       }
    }
 
    driUpdateFramebufferSize(&intel->ctx, drawable);
@@ -528,8 +609,6 @@ intelInitContext(struct intel_context *intel,
       }
    }
 
-   ctx->Const.MaxTextureMaxAnisotropy = 2.0;
-
    /* This doesn't yet catch all non-conformant rendering, but it's a
     * start.
     */
@@ -695,13 +774,64 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
       intel->prim.vb_bo = NULL;
 
       if (release_texture_heaps) {
-         /* This share group is about to go away, free our private
-          * texture object data.
+         /* Nothing is currently done here to free texture heaps;
+          * but we're not using the texture heap utilities, so I
+          * rather think we shouldn't.  I've taken a look, and can't
+          * find any private texture data hanging around anywhere, but
+          * I'm not yet certain there isn't any at all...
           */
-         if (INTEL_DEBUG & DEBUG_TEXTURE)
+         /* if (INTEL_DEBUG & DEBUG_TEXTURE)
             fprintf(stderr, "do something to free texture heaps\n");
+          */
       }
 
+      /* XXX In intelMakeCurrent() below, the context's static regions are 
+       * referenced inside the frame buffer; it's listed as a hack,
+       * with a comment of "XXX FBO temporary fix-ups!", but
+       * as long as it's there, we should release the regions here.
+       * The do/while loop around the block is used to allow the
+       * "continue" statements inside the block to exit the block,
+       * to avoid many layers of "if" constructs.
+       */
+      do {
+         __DRIdrawablePrivate * driDrawPriv = intel->driDrawable;
+         struct intel_framebuffer *intel_fb;
+         struct intel_renderbuffer *irbDepth, *irbStencil;
+         if (!driDrawPriv) {
+            /* We're already detached from the drawable; exit this block. */
+            continue;
+         }
+         intel_fb = (struct intel_framebuffer *) driDrawPriv->driverPrivate;
+         if (!intel_fb) {
+            /* The frame buffer is already gone; exit this block. */
+            continue;
+         }
+         irbDepth = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
+         irbStencil = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL);
+
+         /* If the regions of the frame buffer still match the regions
+          * of the context, release them.  If they've changed somehow,
+          * leave them alone.
+          */
+         if (intel_fb->color_rb[0] && intel_fb->color_rb[0]->region == intel->front_region) {
+	    intel_renderbuffer_set_region(intel_fb->color_rb[0], NULL);
+         }
+         if (intel_fb->color_rb[1] && intel_fb->color_rb[1]->region == intel->back_region) {
+	    intel_renderbuffer_set_region(intel_fb->color_rb[1], NULL);
+         }
+
+         if (irbDepth && irbDepth->region == intel->depth_region) {
+	    intel_renderbuffer_set_region(irbDepth, NULL);
+         }
+         /* Usually, the stencil buffer is the same as the depth buffer;
+          * but they're handled separately in MakeCurrent, so we'll
+          * handle them separately here.
+          */
+         if (irbStencil && irbStencil->region == intel->depth_region) {
+	    intel_renderbuffer_set_region(irbStencil, NULL);
+         }
+      } while (0);
+
       intel_region_release(&intel->front_region);
       intel_region_release(&intel->back_region);
       intel_region_release(&intel->depth_region);
@@ -710,6 +840,8 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
 
       /* free the Mesa context */
       _mesa_free_context_data(&intel->ctx);
+
+      
    }
 }
 
@@ -738,7 +870,10 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
           if (driDrawPriv != driReadPriv)
               intel_update_renderbuffers(driContextPriv, driReadPriv);
       } else {
-          /* XXX FBO temporary fix-ups! */
+          /* XXX FBO temporary fix-ups!  These are released in 
+           * intelDextroyContext(), above.  Changes here should be
+           * reflected there.
+           */
           /* if the renderbuffers don't have regions, init them from the context */
          struct intel_renderbuffer *irbDepth
             = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index d798225ddd9..c16732d7b52 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -161,12 +161,22 @@ struct intel_context
    struct {
       struct gl_fragment_program *bitmap_fp;
       struct gl_vertex_program *passthrough_vp;
+      struct gl_buffer_object *texcoord_vbo;
 
       struct gl_fragment_program *saved_fp;
       GLboolean saved_fp_enable;
       struct gl_vertex_program *saved_vp;
       GLboolean saved_vp_enable;
 
+      struct gl_fragment_program *tex2d_fp;
+
+      GLboolean saved_texcoord_enable;
+      struct gl_buffer_object *saved_array_vbo, *saved_texcoord_vbo;
+      GLenum saved_texcoord_type;
+      GLsizei saved_texcoord_size, saved_texcoord_stride;
+      const void *saved_texcoord_ptr;
+      int saved_active_texture;
+
       GLint saved_vp_x, saved_vp_y;
       GLsizei saved_vp_width, saved_vp_height;
       GLenum saved_matrix_mode;
diff --git a/src/mesa/drivers/dri/intel/intel_generatemipmap.c b/src/mesa/drivers/dri/intel/intel_generatemipmap.c
new file mode 100644
index 00000000000..02804b51fa8
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_generatemipmap.c
@@ -0,0 +1,283 @@
+/*
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Eric Anholt <eric@anholt.net>
+ *
+ */
+
+#include "main/glheader.h"
+#include "main/enums.h"
+#include "main/image.h"
+#include "main/mtypes.h"
+#include "main/macros.h"
+#include "main/bufferobj.h"
+#include "main/teximage.h"
+#include "main/texenv.h"
+#include "main/texobj.h"
+#include "main/texstate.h"
+#include "main/texparam.h"
+#include "main/varray.h"
+#include "main/attrib.h"
+#include "main/enable.h"
+#include "main/buffers.h"
+#include "main/fbobject.h"
+#include "main/framebuffer.h"
+#include "main/renderbuffer.h"
+#include "main/depth.h"
+#include "main/hash.h"
+#include "main/mipmap.h"
+#include "main/blend.h"
+#include "glapi/dispatch.h"
+#include "swrast/swrast.h"
+
+#include "intel_screen.h"
+#include "intel_context.h"
+#include "intel_batchbuffer.h"
+#include "intel_pixel.h"
+#include "intel_tex.h"
+#include "intel_mipmap_tree.h"
+
+static const char *intel_fp_tex2d =
+      "!!ARBfp1.0\n"
+      "TEX result.color, fragment.texcoord[0], texture[0], 2D;\n"
+      "END\n";
+
+static GLboolean
+intel_generate_mipmap_level(GLcontext *ctx, GLuint tex_name,
+			    int level, int width, int height)
+{
+   struct intel_context *intel = intel_context(ctx);
+   GLfloat vertices[4][2];
+   GLint status;
+
+   /* Set to source from the previous level */
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_BASE_LEVEL, level - 1);
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, level - 1);
+
+   /* Set to draw into the current level */
+   _mesa_FramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT,
+				 GL_COLOR_ATTACHMENT0_EXT,
+				 GL_TEXTURE_2D,
+				 tex_name,
+				 level);
+   /* Choose to render to the color attachment. */
+   _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0_EXT);
+
+   status = glCheckFramebufferStatusEXT (GL_FRAMEBUFFER_EXT);
+   if (status != GL_FRAMEBUFFER_COMPLETE_EXT)
+      return GL_FALSE;
+
+   intel_meta_set_passthrough_transform(intel);
+
+   /* XXX: Doing it right would involve setting up the transformation to do
+    * 0-1 mapping or something, and not changing the vertex data.
+    */
+   vertices[0][0] = 0;
+   vertices[0][1] = 0;
+   vertices[1][0] = width;
+   vertices[1][1] = 0;
+   vertices[2][0] = width;
+   vertices[2][1] = height;
+   vertices[3][0] = 0;
+   vertices[3][1] = height;
+
+   _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices);
+   _mesa_Enable(GL_VERTEX_ARRAY);
+   intel_meta_set_default_texrect(intel);
+
+   CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4));
+
+   intel_meta_restore_texcoords(intel);
+   intel_meta_restore_transform(intel);
+
+   return GL_TRUE;
+}
+
+static GLboolean
+intel_generate_mipmap_2d(GLcontext *ctx,
+			 GLenum target,
+			 struct gl_texture_object *texObj)
+{
+   struct intel_context *intel = intel_context(ctx);
+   GLint old_active_texture;
+   int level, max_levels, start_level, end_level;
+   GLuint fb_name;
+   GLboolean success = GL_FALSE;
+   struct gl_framebuffer *saved_fbo = NULL;
+
+   _mesa_PushAttrib(GL_ENABLE_BIT | GL_TEXTURE_BIT |
+		    GL_CURRENT_BIT | GL_COLOR_BUFFER_BIT |
+		    GL_DEPTH_BUFFER_BIT);
+   _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
+   old_active_texture = ctx->Texture.CurrentUnit;
+   _mesa_reference_framebuffer(&saved_fbo, ctx->DrawBuffer);
+
+   _mesa_Disable(GL_POLYGON_STIPPLE);
+   _mesa_Disable(GL_DEPTH_TEST);
+   _mesa_Disable(GL_STENCIL_TEST);
+   _mesa_ColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
+   _mesa_DepthMask(GL_FALSE);
+
+   /* Bind the given texture to GL_TEXTURE_2D with linear filtering for our
+    * minification.
+    */
+   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB);
+   _mesa_Enable(GL_TEXTURE_2D);
+   _mesa_BindTexture(GL_TEXTURE_2D, texObj->Name);
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER,
+		       GL_LINEAR_MIPMAP_NEAREST);
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
+
+   /* Bind the new renderbuffer to the color attachment point. */
+   _mesa_GenFramebuffersEXT(1, &fb_name);
+   _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, fb_name);
+
+   intel_meta_set_fragment_program(intel, &intel->meta.tex2d_fp,
+				   intel_fp_tex2d);
+   intel_meta_set_passthrough_vertex_program(intel);
+
+   max_levels = _mesa_max_texture_levels(ctx, texObj->Target);
+   start_level = texObj->BaseLevel;
+   end_level = texObj->MaxLevel;
+
+   /* Loop generating level+1 from level. */
+   for (level = start_level; level < end_level && level < max_levels - 1; level++) {
+      const struct gl_texture_image *srcImage;
+      int width, height;
+
+      srcImage = _mesa_select_tex_image(ctx, texObj, target, level);
+      if (srcImage->Border != 0)
+	 goto fail;
+
+      width = srcImage->Width / 2;
+      if (width < 1)
+	 width = 1;
+      height = srcImage->Height / 2;
+      if (height < 1)
+	 height = 1;
+
+      if (width == srcImage->Width &&
+	  height == srcImage->Height) {
+	 /* Neither _mesa_max_texture_levels nor texObj->MaxLevel are the
+	  * maximum texture level for the object, so break out when we've gone
+	  * over the edge.
+	  */
+	 break;
+      }
+
+      /* Make sure that there's space allocated for the target level.
+       * We could skip this if there's already space allocated and save some
+       * time.
+       */
+      _mesa_TexImage2D(GL_TEXTURE_2D, level + 1, srcImage->InternalFormat,
+		       width, height, 0,
+		       GL_RGBA, GL_UNSIGNED_INT, NULL);
+
+      if (!intel_generate_mipmap_level(ctx, texObj->Name, level + 1,
+				       width, height))
+	 goto fail;
+   }
+
+   success = GL_TRUE;
+
+fail:
+   intel_meta_restore_fragment_program(intel);
+   intel_meta_restore_vertex_program(intel);
+
+   _mesa_DeleteFramebuffersEXT(1, &fb_name);
+   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
+   if (saved_fbo)
+      _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, saved_fbo->Name);
+   _mesa_reference_framebuffer(&saved_fbo, NULL);
+   _mesa_PopClientAttrib();
+   _mesa_PopAttrib();
+
+   return success;
+}
+
+
+/**
+ * Generate new mipmap data from BASE+1 to BASE+p (the minimally-sized mipmap
+ * level).
+ *
+ * The texture object's miptree must be mapped.
+ *
+ * It would be really nice if this was just called by Mesa whenever mipmaps
+ * needed to be regenerated, rather than us having to remember to do so in
+ * each texture image modification path.
+ *
+ * This function should also include an accelerated path.
+ */
+void
+intel_generate_mipmap(GLcontext *ctx, GLenum target,
+                      struct gl_texture_object *texObj)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_texture_object *intelObj = intel_texture_object(texObj);
+   GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
+   int face, i;
+
+   /* HW path */
+   if (target == GL_TEXTURE_2D &&
+       ctx->Extensions.EXT_framebuffer_object &&
+       ctx->Extensions.ARB_fragment_program &&
+       ctx->Extensions.ARB_vertex_program) {
+      GLboolean success;
+
+      /* We'll be accessing this texture using GL entrypoints, which should
+       * be resilient against other access to this texture.
+       */
+      _mesa_unlock_texture(ctx, texObj);
+      success = intel_generate_mipmap_2d(ctx, target, texObj);
+      _mesa_lock_texture(ctx, texObj);
+
+      if (success)
+	 return;
+   }
+
+   /* SW path */
+   intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel);
+   _mesa_generate_mipmap(ctx, target, texObj);
+   intel_tex_unmap_level_images(intel, intelObj, texObj->BaseLevel);
+
+   /* Update the level information in our private data in the new images, since
+    * it didn't get set as part of a normal TexImage path.
+    */
+   for (face = 0; face < nr_faces; face++) {
+      for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) {
+         struct intel_texture_image *intelImage;
+
+	 intelImage = intel_texture_image(texObj->Image[face][i]);
+	 if (intelImage == NULL)
+	    break;
+
+	 intelImage->level = i;
+	 intelImage->face = face;
+	 /* Unreference the miptree to signal that the new Data is a bare
+	  * pointer from mesa.
+	  */
+	 intel_miptree_release(intel, &intelImage->mt);
+      }
+   }
+}
diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
index 6e1e034e53d..f3652720ece 100644
--- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
+++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
@@ -62,9 +62,10 @@ intel_miptree_create_internal(struct intel_context *intel,
    GLboolean ok;
    struct intel_mipmap_tree *mt = calloc(sizeof(*mt), 1);
 
-   DBG("%s target %s format %s level %d..%d\n", __FUNCTION__,
+   DBG("%s target %s format %s level %d..%d <-- %p\n", __FUNCTION__,
        _mesa_lookup_enum_by_nr(target),
-       _mesa_lookup_enum_by_nr(internal_format), first_level, last_level);
+       _mesa_lookup_enum_by_nr(internal_format), 
+       first_level, last_level, mt);
 
    mt->target = target_to_target(target);
    mt->internal_format = internal_format;
@@ -89,6 +90,7 @@ intel_miptree_create_internal(struct intel_context *intel,
 
    if (!ok) {
       free(mt);
+      DBG("%s not okay - returning NULL\n", __FUNCTION__);
       return NULL;
    }
 
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.c b/src/mesa/drivers/dri/intel/intel_pixel.c
index fc0ac0b79c0..36a684b3b85 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel.c
@@ -27,9 +27,12 @@
 
 #include "main/enums.h"
 #include "main/state.h"
+#include "main/bufferobj.h"
 #include "main/context.h"
 #include "main/enable.h"
 #include "main/matrix.h"
+#include "main/texstate.h"
+#include "main/varray.h"
 #include "main/viewport.h"
 #include "swrast/swrast.h"
 #include "shader/arbprogram.h"
@@ -334,6 +337,85 @@ intel_meta_restore_fragment_program(struct intel_context *intel)
       _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
 }
 
+static const float default_texcoords[4][2] = { { 0.0, 0.0 },
+					       { 1.0, 0.0 },
+					       { 1.0, 1.0 },
+					       { 0.0, 1.0 } };
+
+void
+intel_meta_set_default_texrect(struct intel_context *intel)
+{
+   GLcontext *ctx = &intel->ctx;
+   struct gl_client_array *old_texcoord_array;
+
+   intel->meta.saved_active_texture = ctx->Texture.CurrentUnit;
+   if (intel->meta.saved_array_vbo == NULL) {
+      _mesa_reference_buffer_object(ctx, &intel->meta.saved_array_vbo,
+				    ctx->Array.ArrayBufferObj);
+   }
+
+   old_texcoord_array = &ctx->Array.ArrayObj->TexCoord[0];
+   intel->meta.saved_texcoord_type = old_texcoord_array->Type;
+   intel->meta.saved_texcoord_size = old_texcoord_array->Size;
+   intel->meta.saved_texcoord_stride = old_texcoord_array->Stride;
+   intel->meta.saved_texcoord_enable = old_texcoord_array->Enabled;
+   intel->meta.saved_texcoord_ptr = old_texcoord_array->Ptr;
+   _mesa_reference_buffer_object(ctx, &intel->meta.saved_texcoord_vbo,
+				 old_texcoord_array->BufferObj);
+
+   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
+
+   if (intel->meta.texcoord_vbo == NULL) {
+      GLuint vbo_name;
+
+      _mesa_GenBuffersARB(1, &vbo_name);
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, vbo_name);
+      _mesa_BufferDataARB(GL_ARRAY_BUFFER_ARB, sizeof(default_texcoords),
+			  default_texcoords, GL_STATIC_DRAW_ARB);
+      _mesa_reference_buffer_object(ctx, &intel->meta.texcoord_vbo,
+				    ctx->Array.ArrayBufferObj);
+   } else {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
+			  intel->meta.texcoord_vbo->Name);
+   }
+   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), NULL);
+
+   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
+}
+
+void
+intel_meta_restore_texcoords(struct intel_context *intel)
+{
+   GLcontext *ctx = &intel->ctx;
+
+   /* Restore the old TexCoordPointer */
+   if (intel->meta.saved_texcoord_vbo) {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
+			  intel->meta.saved_texcoord_vbo->Name);
+      _mesa_reference_buffer_object(ctx, &intel->meta.saved_texcoord_vbo, NULL);
+   } else {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
+   }
+
+   _mesa_TexCoordPointer(intel->meta.saved_texcoord_size,
+			 intel->meta.saved_texcoord_type,
+			 intel->meta.saved_texcoord_stride,
+			 intel->meta.saved_texcoord_ptr);
+   if (!intel->meta.saved_texcoord_enable)
+      _mesa_Disable(GL_TEXTURE_COORD_ARRAY);
+
+   _mesa_ClientActiveTextureARB(GL_TEXTURE0 +
+				intel->meta.saved_active_texture);
+
+   if (intel->meta.saved_array_vbo) {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
+			  intel->meta.saved_array_vbo->Name);
+      _mesa_reference_buffer_object(ctx, &intel->meta.saved_array_vbo, NULL);
+   } else {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
+   }
+}
+
 void
 intelInitPixelFuncs(struct dd_function_table *functions)
 {
@@ -355,5 +437,7 @@ intel_free_pixel_state(struct intel_context *intel)
 
    _mesa_reference_vertprog(ctx, &intel->meta.passthrough_vp, NULL);
    _mesa_reference_fragprog(ctx, &intel->meta.bitmap_fp, NULL);
+   _mesa_reference_fragprog(ctx, &intel->meta.tex2d_fp, NULL);
+   _mesa_reference_buffer_object(ctx, &intel->meta.texcoord_vbo, NULL);
 }
 
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.h b/src/mesa/drivers/dri/intel/intel_pixel.h
index cb41fa182cb..6acf0813c8c 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.h
+++ b/src/mesa/drivers/dri/intel/intel_pixel.h
@@ -40,6 +40,9 @@ void intel_meta_set_fragment_program(struct intel_context *intel,
 				     const char *prog_string);
 void intel_meta_restore_fragment_program(struct intel_context *intel);
 void intel_free_pixel_state(struct intel_context *intel);
+void intel_meta_set_default_texrect(struct intel_context *intel);
+void intel_meta_set_default_texrect(struct intel_context *intel);
+void intel_meta_restore_texcoords(struct intel_context *intel);
 
 GLboolean intel_check_blit_fragment_ops(GLcontext * ctx,
 					GLboolean src_alpha_is_one);
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
index 1db7f5594e9..b20840b9a06 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
@@ -360,7 +360,6 @@ intel_texture_bitmap(GLcontext * ctx,
       "END\n";
    GLuint texname;
    GLfloat vertices[4][4];
-   GLfloat texcoords[4][2];
    GLint old_active_texture;
    GLubyte *unpacked_bitmap;
    GLubyte *a8_bitmap;
@@ -485,22 +484,12 @@ intel_texture_bitmap(GLcontext * ctx,
    vertices[3][2] = dst_z;
    vertices[3][3] = 1.0;
 
-   texcoords[0][0] = 0.0;
-   texcoords[0][1] = 0.0;
-   texcoords[1][0] = 1.0;
-   texcoords[1][1] = 0.0;
-   texcoords[2][0] = 1.0;
-   texcoords[2][1] = 1.0;
-   texcoords[3][0] = 0.0;
-   texcoords[3][1] = 1.0;
-
    _mesa_VertexPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &vertices);
-   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
-   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
    _mesa_Enable(GL_VERTEX_ARRAY);
-   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
+   intel_meta_set_default_texrect(intel);
    CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4));
 
+   intel_meta_restore_texcoords(intel);
    intel_meta_restore_transform(intel);
    intel_meta_restore_fragment_program(intel);
    intel_meta_restore_vertex_program(intel);
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_draw.c b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
index e8d5ac8569d..abcdcd5724c 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_draw.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
@@ -70,7 +70,6 @@ intel_texture_drawpixels(GLcontext * ctx,
    struct intel_context *intel = intel_context(ctx);
    GLuint texname;
    GLfloat vertices[4][4];
-   GLfloat texcoords[4][2];
    GLfloat z;
    GLint old_active_texture;
    GLenum internalFormat;
@@ -169,22 +168,13 @@ intel_texture_drawpixels(GLcontext * ctx,
    vertices[3][2] = z;
    vertices[3][3] = 1.0;
 
-   texcoords[0][0] = 0.0;
-   texcoords[0][1] = 0.0;
-   texcoords[1][0] = 1.0;
-   texcoords[1][1] = 0.0;
-   texcoords[2][0] = 1.0;
-   texcoords[2][1] = 1.0;
-   texcoords[3][0] = 0.0;
-   texcoords[3][1] = 1.0;
-
    _mesa_VertexPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &vertices);
-   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
-   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
    _mesa_Enable(GL_VERTEX_ARRAY);
-   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
+   intel_meta_set_default_texrect(intel);
+
    CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4));
 
+   intel_meta_restore_texcoords(intel);
    intel_meta_restore_transform(intel);
 
    _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
@@ -208,7 +198,6 @@ intel_stencil_drawpixels(GLcontext * ctx,
    struct intel_context *intel = intel_context(ctx);
    GLuint texname, rb_name, fb_name, old_fb_name;
    GLfloat vertices[4][2];
-   GLfloat texcoords[4][2];
    struct intel_renderbuffer *irb;
    struct intel_renderbuffer *depth_irb;
    struct gl_renderbuffer *rb;
@@ -343,7 +332,6 @@ intel_stencil_drawpixels(GLcontext * ctx,
    _mesa_free(stencil_pixels);
 
    intel_meta_set_passthrough_transform(intel);
-
    vertices[0][0] = x;
    vertices[0][1] = y;
    vertices[1][0] = x + width * ctx->Pixel.ZoomX;
@@ -353,22 +341,13 @@ intel_stencil_drawpixels(GLcontext * ctx,
    vertices[3][0] = x;
    vertices[3][1] = y + height * ctx->Pixel.ZoomY;
 
-   texcoords[0][0] = 0.0;
-   texcoords[0][1] = 0.0;
-   texcoords[1][0] = 1.0;
-   texcoords[1][1] = 0.0;
-   texcoords[2][0] = 1.0;
-   texcoords[2][1] = 1.0;
-   texcoords[3][0] = 0.0;
-   texcoords[3][1] = 1.0;
-
    _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices);
-   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
-   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
    _mesa_Enable(GL_VERTEX_ARRAY);
-   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
+   intel_meta_set_default_texrect(intel);
+
    CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4));
 
+   intel_meta_restore_texcoords(intel);
    intel_meta_restore_transform(intel);
 
    _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c
index 0aa5b8c02c9..534e75efe1f 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.c
+++ b/src/mesa/drivers/dri/intel/intel_regions.c
@@ -52,12 +52,66 @@
 
 #define FILE_DEBUG_FLAG DEBUG_REGION
 
+/* This should be set to the maximum backtrace size desired.
+ * Set it to 0 to disable backtrace debugging.
+ */
+#define DEBUG_BACKTRACE_SIZE 0
+
+#if DEBUG_BACKTRACE_SIZE == 0
+/* Use the standard debug output */
+#define _DBG(...) DBG(__VA_ARGS__)
+#else
+/* Use backtracing debug output */
+#define _DBG(...) {debug_backtrace(); DBG(__VA_ARGS__);}
+
+/* Backtracing debug support */
+#include <execinfo.h>
+
+static void
+debug_backtrace(void)
+{
+   void *trace[DEBUG_BACKTRACE_SIZE];
+   char **strings = NULL;
+   int traceSize;
+   register int i;
+
+   traceSize = backtrace(trace, DEBUG_BACKTRACE_SIZE);
+   strings = backtrace_symbols(trace, traceSize);
+   if (strings == NULL) {
+      DBG("no backtrace:");
+      return;
+   }
+
+   /* Spit out all the strings with a colon separator.  Ignore
+    * the first, since we don't really care about the call
+    * to debug_backtrace() itself.  Skip until the final "/" in
+    * the trace to avoid really long lines.
+    */
+   for (i = 1; i < traceSize; i++) {
+      char *p = strings[i], *slash = strings[i];
+      while (*p) {
+         if (*p++ == '/') {
+            slash = p;
+         }
+      }
+
+      DBG("%s:", slash);
+   }
+
+   /* Free up the memory, and we're done */
+   free(strings);
+}
+
+#endif
+
+
+
 /* XXX: Thread safety?
  */
 GLubyte *
 intel_region_map(struct intel_context *intel, struct intel_region *region)
 {
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s %p\n", __FUNCTION__, region);
    if (!region->map_refcount++) {
       if (region->pbo)
          intel_region_cow(intel, region);
@@ -72,7 +126,7 @@ intel_region_map(struct intel_context *intel, struct intel_region *region)
 void
 intel_region_unmap(struct intel_context *intel, struct intel_region *region)
 {
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s %p\n", __FUNCTION__, region);
    if (!--region->map_refcount) {
       dri_bo_unmap(region->buffer);
       region->map = NULL;
@@ -87,10 +141,10 @@ intel_region_alloc_internal(struct intel_context *intel,
 {
    struct intel_region *region;
 
-   DBG("%s\n", __FUNCTION__);
-
-   if (buffer == NULL)
+   if (buffer == NULL) {
+      _DBG("%s <-- NULL\n", __FUNCTION__);
       return NULL;
+   }
 
    region = calloc(sizeof(*region), 1);
    region->cpp = cpp;
@@ -104,6 +158,7 @@ intel_region_alloc_internal(struct intel_context *intel,
    region->tiling = I915_TILING_NONE;
    region->bit_6_swizzle = I915_BIT_6_SWIZZLE_NONE;
 
+   _DBG("%s <-- %p\n", __FUNCTION__, region);
    return region;
 }
 
@@ -158,7 +213,7 @@ void
 intel_region_reference(struct intel_region **dst, struct intel_region *src)
 {
    if (src)
-      DBG("%s %p %d\n", __FUNCTION__, src, src->refcount);
+      _DBG("%s %p %d\n", __FUNCTION__, src, src->refcount);
 
    assert(*dst == NULL);
    if (src) {
@@ -172,10 +227,12 @@ intel_region_release(struct intel_region **region_handle)
 {
    struct intel_region *region = *region_handle;
 
-   if (region == NULL)
+   if (region == NULL) {
+      _DBG("%s NULL\n", __FUNCTION__);
       return;
+   }
 
-   DBG("%s %p %d\n", __FUNCTION__, region, region->refcount - 1);
+   _DBG("%s %p %d\n", __FUNCTION__, region, region->refcount - 1);
 
    ASSERT(region->refcount > 0);
    region->refcount--;
@@ -251,7 +308,7 @@ intel_region_data(struct intel_context *intel,
 {
    GLboolean locked = GL_FALSE;
 
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s\n", __FUNCTION__);
 
    if (intel == NULL)
       return;
@@ -293,7 +350,7 @@ intel_region_copy(struct intel_context *intel,
                   GLuint src_offset,
                   GLuint srcx, GLuint srcy, GLuint width, GLuint height)
 {
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s\n", __FUNCTION__);
 
    if (intel == NULL)
       return;
@@ -326,7 +383,7 @@ intel_region_fill(struct intel_context *intel,
                   GLuint dstx, GLuint dsty,
                   GLuint width, GLuint height, GLuint color)
 {
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s\n", __FUNCTION__);
 
    if (intel == NULL)
       return;   
@@ -356,6 +413,8 @@ intel_region_attach_pbo(struct intel_context *intel,
    if (region->pbo == pbo)
       return;
 
+   _DBG("%s %p %p\n", __FUNCTION__, region, pbo);
+
    /* If there is already a pbo attached, break the cow tie now.
     * Don't call intel_region_release_pbo() as that would
     * unnecessarily allocate a new buffer we would have to immediately
@@ -385,6 +444,7 @@ void
 intel_region_release_pbo(struct intel_context *intel,
                          struct intel_region *region)
 {
+   _DBG("%s %p\n", __FUNCTION__, region);
    assert(region->buffer == region->pbo->buffer);
    region->pbo->region = NULL;
    region->pbo = NULL;
@@ -412,7 +472,7 @@ intel_region_cow(struct intel_context *intel, struct intel_region *region)
 
    assert(region->cpp * region->pitch * region->height == pbo->Base.Size);
 
-   DBG("%s (%d bytes)\n", __FUNCTION__, pbo->Base.Size);
+   _DBG("%s %p (%d bytes)\n", __FUNCTION__, region, pbo->Base.Size);
 
    /* Now blit from the texture buffer to the new buffer: 
     */
@@ -459,6 +519,10 @@ intel_recreate_static(struct intel_context *intel,
    if (region == NULL) {
       region = calloc(sizeof(*region), 1);
       region->refcount = 1;
+      _DBG("%s creating new region %p\n", __FUNCTION__, region);
+   }
+   else {
+      _DBG("%s %p\n", __FUNCTION__, region);
    }
 
    if (intel->ctx.Visual.rgbBits == 24)
diff --git a/src/mesa/drivers/dri/intel/intel_tex.c b/src/mesa/drivers/dri/intel/intel_tex.c
index ae0994b183a..fbd6e1d0c36 100644
--- a/src/mesa/drivers/dri/intel/intel_tex.c
+++ b/src/mesa/drivers/dri/intel/intel_tex.c
@@ -158,60 +158,6 @@ timed_memcpy(void *dest, const void *src, size_t n)
 }
 #endif /* DO_DEBUG */
 
-/**
- * Generate new mipmap data from BASE+1 to BASE+p (the minimally-sized mipmap
- * level).
- *
- * The texture object's miptree must be mapped.
- *
- * It would be really nice if this was just called by Mesa whenever mipmaps
- * needed to be regenerated, rather than us having to remember to do so in
- * each texture image modification path.
- *
- * This function should also include an accelerated path.
- */
-void
-intel_generate_mipmap(GLcontext *ctx, GLenum target,
-                      struct gl_texture_object *texObj)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct intel_texture_object *intelObj = intel_texture_object(texObj);
-   GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
-   int face, i;
-
-   _mesa_generate_mipmap(ctx, target, texObj);
-
-   /* Update the level information in our private data in the new images, since
-    * it didn't get set as part of a normal TexImage path.
-    */
-   for (face = 0; face < nr_faces; face++) {
-      for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) {
-         struct intel_texture_image *intelImage;
-
-	 intelImage = intel_texture_image(texObj->Image[face][i]);
-	 if (intelImage == NULL)
-	    break;
-
-	 intelImage->level = i;
-	 intelImage->face = face;
-	 /* Unreference the miptree to signal that the new Data is a bare
-	  * pointer from mesa.
-	  */
-	 intel_miptree_release(intel, &intelImage->mt);
-      }
-   }
-}
-
-static void intelGenerateMipmap(GLcontext *ctx, GLenum target, struct gl_texture_object *texObj)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct intel_texture_object *intelObj = intel_texture_object(texObj);
-
-   intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel);
-   intel_generate_mipmap(ctx, target, texObj);
-   intel_tex_unmap_level_images(intel, intelObj, texObj->BaseLevel);
-}
-
 void
 intelInitTextureFuncs(struct dd_function_table *functions)
 {
@@ -227,7 +173,7 @@ intelInitTextureFuncs(struct dd_function_table *functions)
    functions->CopyTexSubImage1D = intelCopyTexSubImage1D;
    functions->CopyTexSubImage2D = intelCopyTexSubImage2D;
    functions->GetTexImage = intelGetTexImage;
-   functions->GenerateMipmap = intelGenerateMipmap;
+   functions->GenerateMipmap = intel_generate_mipmap;
 
    /* compressed texture functions */
    functions->CompressedTexImage2D = intelCompressedTexImage2D;
diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c
index 08437aa0e2b..7c2b26ef1d4 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c
@@ -158,7 +158,7 @@ do_copy_texsubimage(struct intel_context *intel,
 
    /* GL_SGIS_generate_mipmap */
    if (intelImage->level == texObj->BaseLevel && texObj->GenerateMipmap) {
-      ctx->Driver.GenerateMipmap(ctx, target, texObj);
+      intel_generate_mipmap(ctx, target, texObj);
    }
 
    return GL_TRUE;
diff --git a/src/mesa/drivers/dri/intel/intel_tex_image.c b/src/mesa/drivers/dri/intel/intel_tex_image.c
index 1f192dafbe1..b71fe2a7ae0 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_image.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_image.c
@@ -316,7 +316,6 @@ intelTexImage(GLcontext * ctx,
    GLint postConvHeight = height;
    GLint texelBytes, sizeInBytes;
    GLuint dstRowStride = 0, srcRowStride = texImage->RowStride;
-   GLboolean needs_map;
 
    DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__,
        _mesa_lookup_enum_by_nr(target), level, width, height, depth, border);
@@ -482,15 +481,8 @@ intelTexImage(GLcontext * ctx,
 
    LOCK_HARDWARE(intel);
 
-   /* Two cases where we need a mapping of the miptree: when the user supplied
-    * data is mapped as well (non-PBO, memcpy upload) or when we're going to do
-    * (software) mipmap generation.
-    */
-   needs_map = (pixels != NULL) || (level == texObj->BaseLevel &&
-				  texObj->GenerateMipmap);
-
    if (intelImage->mt) {
-      if (needs_map)
+      if (pixels != NULL)
          texImage->Data = intel_miptree_image_map(intel,
                                                   intelImage->mt,
                                                   intelImage->face,
@@ -547,22 +539,22 @@ intelTexImage(GLcontext * ctx,
 						   format, type, pixels, unpack)) {
 	   _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
        }
-
-       /* GL_SGIS_generate_mipmap */
-       if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
-	  intel_generate_mipmap(ctx, target, texObj);
-       }
    }
 
    _mesa_unmap_teximage_pbo(ctx, unpack);
 
    if (intelImage->mt) {
-      if (needs_map)
+      if (pixels != NULL)
          intel_miptree_image_unmap(intel, intelImage->mt);
       texImage->Data = NULL;
    }
 
    UNLOCK_HARDWARE(intel);
+
+   /* GL_SGIS_generate_mipmap */
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      intel_generate_mipmap(ctx, target, texObj);
+   }
 }
 
 void
diff --git a/src/mesa/drivers/dri/intel/intel_tex_subimage.c b/src/mesa/drivers/dri/intel/intel_tex_subimage.c
index f86de568976..48104de2a9b 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_subimage.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_subimage.c
@@ -101,11 +101,6 @@ intelTexSubimage(GLcontext * ctx,
       _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage");
    }
 
-   /* GL_SGIS_generate_mipmap */
-   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
-      intel_generate_mipmap(ctx, target, texObj);
-   }
-
    _mesa_unmap_teximage_pbo(ctx, packing);
 
    if (intelImage->mt) {
@@ -114,6 +109,11 @@ intelTexSubimage(GLcontext * ctx,
    }
 
    UNLOCK_HARDWARE(intel);
+
+   /* GL_SGIS_generate_mipmap */
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      intel_generate_mipmap(ctx, target, texObj);
+   }
 }
 
 
diff --git a/src/mesa/drivers/dri/r200/r200_state.c b/src/mesa/drivers/dri/r200/r200_state.c
index ebf389efe26..a06ea17e799 100644
--- a/src/mesa/drivers/dri/r200/r200_state.c
+++ b/src/mesa/drivers/dri/r200/r200_state.c
@@ -2327,7 +2327,7 @@ GLboolean r200ValidateState( GLcontext *ctx )
       R200_STATECHANGE(rmesa, ctx);
    }
 
-   if (new_state & (_NEW_TEXTURE | _NEW_PROGRAM)) {
+   if (new_state & (_NEW_TEXTURE | _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS)) {
       r200UpdateTextureState( ctx );
       new_state |= rmesa->radeon.NewGLState; /* may add TEXTURE_MATRIX */
       r200UpdateLocalViewer( ctx );
@@ -2370,6 +2370,7 @@ GLboolean r200ValidateState( GLcontext *ctx )
    }
 
    if (new_state & (_NEW_PROGRAM|
+                    _NEW_PROGRAM_CONSTANTS |
    /* need to test for pretty much anything due to possible parameter bindings */
 	_NEW_MODELVIEW|_NEW_PROJECTION|_NEW_TRANSFORM|
 	_NEW_LIGHT|_NEW_TEXTURE|_NEW_TEXTURE_MATRIX|
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index 330f1dab493..217a16818e3 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -464,7 +464,9 @@ static void r300SetEarlyZState(GLcontext * ctx)
 
 	if (ctx->Color.AlphaEnabled && ctx->Color.AlphaFunc != GL_ALWAYS)
 		topZ = R300_ZTOP_DISABLE;
-	if (current_fragment_program_writes_depth(ctx))
+	else if (current_fragment_program_writes_depth(ctx))
+		topZ = R300_ZTOP_DISABLE;
+	else if (ctx->FragmentProgram._Current && ctx->FragmentProgram._Current->UsesKill)
 		topZ = R300_ZTOP_DISABLE;
 
 	if (topZ != r300->hw.zstencil_format.cmd[2]) {
@@ -1060,7 +1062,7 @@ void r300UpdateStateParameters(GLcontext * ctx, GLuint new_state)
 	struct gl_program_parameter_list *paramList;
 	GLuint i;
 
-	if (!(new_state & (_NEW_BUFFERS | _NEW_PROGRAM)))
+	if (!(new_state & (_NEW_BUFFERS | _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS)))
 		return;
 
 	fp = (struct r300_fragment_program *)ctx->FragmentProgram._Current;
@@ -2218,11 +2220,12 @@ void r300UpdateShaders(r300ContextPtr rmesa)
 			hw_tcl_on = future_hw_tcl_on = 0;
 			r300ResetHwState(rmesa);
 
-			r300UpdateStateParameters(ctx, _NEW_PROGRAM);
+			r300UpdateStateParameters(ctx, _NEW_PROGRAM |
+                                                  _NEW_PROGRAM_CONSTANTS);
 			return;
 		}
 	}
-	r300UpdateStateParameters(ctx, _NEW_PROGRAM);
+	r300UpdateStateParameters(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
 }
 
 static const GLfloat *get_fragmentprogram_constant(GLcontext *ctx,
diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c
index 305df548fa2..9a01465bdf9 100644
--- a/src/mesa/drivers/x11/xm_dd.c
+++ b/src/mesa/drivers/x11/xm_dd.c
@@ -912,8 +912,9 @@ xmesa_update_state( GLcontext *ctx, GLbitfield new_state )
    /*
     * GL_DITHER, GL_READ/DRAW_BUFFER, buffer binding state, etc. effect
     * renderbuffer span/clear funcs.
+    * Check _NEW_COLOR to detect dither enable/disable.
     */
-   if (new_state & (_NEW_COLOR | _NEW_PIXEL | _NEW_BUFFERS)) {
+   if (new_state & (_NEW_COLOR | _NEW_BUFFERS)) {
       XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer);
       struct xmesa_renderbuffer *front_xrb, *back_xrb;
 
diff --git a/src/mesa/gl.pc.in b/src/mesa/gl.pc.in
index 0462b9fca2b..97b86596fcf 100644
--- a/src/mesa/gl.pc.in
+++ b/src/mesa/gl.pc.in
@@ -7,6 +7,6 @@ Name: gl
 Description: Mesa OpenGL library
 Requires.private: @GL_PC_REQ_PRIV@
 Version: @VERSION@
-Libs: -L${libdir} -lGL
+Libs: -L${libdir} -l@GL_LIB@
 Libs.private: @GL_PC_LIB_PRIV@
 Cflags: -I${includedir} @GL_PC_CFLAGS@
diff --git a/src/mesa/main/api_validate.c b/src/mesa/main/api_validate.c
index 42d1e579e08..d5c604c56a2 100644
--- a/src/mesa/main/api_validate.c
+++ b/src/mesa/main/api_validate.c
@@ -30,6 +30,26 @@
 #include "state.h"
 
 
+
+/**
+ * \return  number of bytes in array [count] of type.
+ */
+static GLsizei
+index_bytes(GLenum type, GLsizei count)
+{
+   if (type == GL_UNSIGNED_INT) {
+      return count * sizeof(GLuint);
+   }
+   else if (type == GL_UNSIGNED_BYTE) {
+      return count * sizeof(GLubyte);
+   }
+   else {
+      ASSERT(type == GL_UNSIGNED_SHORT);
+      return count * sizeof(GLushort);
+   }
+}
+
+
 /**
  * Find the max index in the given element/index buffer
  */
@@ -44,10 +64,8 @@ max_buffer_index(GLcontext *ctx, GLuint count, GLenum type,
 
    if (elementBuf->Name) {
       /* elements are in a user-defined buffer object.  need to map it */
-      map = ctx->Driver.MapBuffer(ctx,
-                                  GL_ELEMENT_ARRAY_BUFFER_ARB,
-                                  GL_READ_ONLY,
-                                  elementBuf);
+      map = ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER,
+                                  GL_READ_ONLY, elementBuf);
       /* Actual address is the sum of pointers */
       indices = (const GLvoid *) ADD_POINTERS(map, (const GLubyte *) indices);
    }
@@ -70,14 +88,16 @@ max_buffer_index(GLcontext *ctx, GLuint count, GLenum type,
    }
 
    if (map) {
-      ctx->Driver.UnmapBuffer(ctx,
-                              GL_ELEMENT_ARRAY_BUFFER_ARB,
-                              ctx->Array.ElementArrayBufferObj);
+      ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER_ARB, elementBuf);
    }
 
    return max;
 }
 
+
+/**
+ * Check if OK to render by examining framebuffer status and vertex arrays.
+ */
 static GLboolean
 check_valid_to_render(GLcontext *ctx, char *function)
 {
@@ -105,6 +125,12 @@ check_valid_to_render(GLcontext *ctx, char *function)
    return GL_TRUE;
 }
 
+
+/**
+ * Error checking for glDrawElements().  Includes parameter checking
+ * and VBO bounds checking.
+ * \return GL_TRUE if OK to render, GL_FALSE if error found
+ */
 GLboolean
 _mesa_validate_DrawElements(GLcontext *ctx,
 			    GLenum mode, GLsizei count, GLenum type,
@@ -140,27 +166,8 @@ _mesa_validate_DrawElements(GLcontext *ctx,
    /* Vertex buffer object tests */
    if (ctx->Array.ElementArrayBufferObj->Name) {
       /* use indices in the buffer object */
-      GLuint indexBytes;
-
-      if (!ctx->Array.ElementArrayBufferObj->Size) {
-         _mesa_warning(ctx,
-                       "glDrawElements called with empty array elements buffer");
-         return GL_FALSE;
-      }
-
-      if (type == GL_UNSIGNED_INT) {
-         indexBytes = count * sizeof(GLuint);
-      }
-      else if (type == GL_UNSIGNED_BYTE) {
-         indexBytes = count * sizeof(GLubyte);
-      }
-      else {
-         ASSERT(type == GL_UNSIGNED_SHORT);
-         indexBytes = count * sizeof(GLushort);
-      }
-
       /* make sure count doesn't go outside buffer bounds */
-      if (indexBytes > (GLuint) ctx->Array.ElementArrayBufferObj->Size) {
+      if (index_bytes(type, count) > ctx->Array.ElementArrayBufferObj->Size) {
          _mesa_warning(ctx, "glDrawElements index out of buffer bounds");
          return GL_FALSE;
       }
@@ -177,6 +184,8 @@ _mesa_validate_DrawElements(GLcontext *ctx,
                                     ctx->Array.ElementArrayBufferObj);
       if (max >= ctx->Array._MaxElement) {
          /* the max element is out of bounds of one or more enabled arrays */
+         _mesa_warning(ctx, "glDrawElements() index=%u is "
+                       "out of bounds (max=%u)", max, ctx->Array._MaxElement);
          return GL_FALSE;
       }
    }
@@ -184,6 +193,12 @@ _mesa_validate_DrawElements(GLcontext *ctx,
    return GL_TRUE;
 }
 
+
+/**
+ * Error checking for glDrawRangeElements().  Includes parameter checking
+ * and VBO bounds checking.
+ * \return GL_TRUE if OK to render, GL_FALSE if error found
+ */
 GLboolean
 _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
 				 GLuint start, GLuint end,
@@ -224,21 +239,8 @@ _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
    /* Vertex buffer object tests */
    if (ctx->Array.ElementArrayBufferObj->Name) {
       /* use indices in the buffer object */
-      GLuint indexBytes;
-
-      if (type == GL_UNSIGNED_INT) {
-         indexBytes = count * sizeof(GLuint);
-      }
-      else if (type == GL_UNSIGNED_BYTE) {
-         indexBytes = count * sizeof(GLubyte);
-      }
-      else {
-         ASSERT(type == GL_UNSIGNED_SHORT);
-         indexBytes = count * sizeof(GLushort);
-      }
-
       /* make sure count doesn't go outside buffer bounds */
-      if (indexBytes > ctx->Array.ElementArrayBufferObj->Size) {
+      if (index_bytes(type, count) > ctx->Array.ElementArrayBufferObj->Size) {
          _mesa_warning(ctx, "glDrawRangeElements index out of buffer bounds");
          return GL_FALSE;
       }
@@ -265,6 +267,7 @@ _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
 /**
  * Called from the tnl module to error check the function parameters and
  * verify that we really can draw something.
+ * \return GL_TRUE if OK to render, GL_FALSE if error found
  */
 GLboolean
 _mesa_validate_DrawArrays(GLcontext *ctx,
diff --git a/src/mesa/main/arrayobj.c b/src/mesa/main/arrayobj.c
index b04095fd16b..8ec73b9526c 100644
--- a/src/mesa/main/arrayobj.c
+++ b/src/mesa/main/arrayobj.c
@@ -68,6 +68,32 @@ lookup_arrayobj(GLcontext *ctx, GLuint id)
 }
 
 
+/**
+ * For all the vertex arrays in the array object, unbind any pointers
+ * to any buffer objects (VBOs).
+ * This is done just prior to array object destruction.
+ */
+static void
+unbind_array_object_vbos(GLcontext *ctx, struct gl_array_object *obj)
+{
+   GLuint i;
+
+   _mesa_reference_buffer_object(ctx, &obj->Vertex.BufferObj, NULL);
+   _mesa_reference_buffer_object(ctx, &obj->Normal.BufferObj, NULL);
+   _mesa_reference_buffer_object(ctx, &obj->Color.BufferObj, NULL);
+   _mesa_reference_buffer_object(ctx, &obj->SecondaryColor.BufferObj, NULL);
+   _mesa_reference_buffer_object(ctx, &obj->FogCoord.BufferObj, NULL);
+   _mesa_reference_buffer_object(ctx, &obj->Index.BufferObj, NULL);
+   _mesa_reference_buffer_object(ctx, &obj->EdgeFlag.BufferObj, NULL);
+
+   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++)
+      _mesa_reference_buffer_object(ctx, &obj->TexCoord[i].BufferObj, NULL);
+
+   for (i = 0; i < VERT_ATTRIB_MAX; i++)
+      _mesa_reference_buffer_object(ctx, &obj->VertexAttrib[i].BufferObj,NULL);
+}
+
+
 /**
  * Allocate and initialize a new vertex array object.
  * 
@@ -94,10 +120,70 @@ void
 _mesa_delete_array_object( GLcontext *ctx, struct gl_array_object *obj )
 {
    (void) ctx;
+   unbind_array_object_vbos(ctx, obj);
+   _glthread_DESTROY_MUTEX(obj->Mutex);
    _mesa_free(obj);
 }
 
 
+/**
+ * Set ptr to arrayObj w/ reference counting.
+ */
+void
+_mesa_reference_array_object(GLcontext *ctx,
+                             struct gl_array_object **ptr,
+                             struct gl_array_object *arrayObj)
+{
+   if (*ptr == arrayObj)
+      return;
+
+   if (*ptr) {
+      /* Unreference the old array object */
+      GLboolean deleteFlag = GL_FALSE;
+      struct gl_array_object *oldObj = *ptr;
+
+      _glthread_LOCK_MUTEX(oldObj->Mutex);
+      ASSERT(oldObj->RefCount > 0);
+      oldObj->RefCount--;
+#if 0
+      printf("ArrayObj %p %d DECR to %d\n",
+             (void *) oldObj, oldObj->Name, oldObj->RefCount);
+#endif
+      deleteFlag = (oldObj->RefCount == 0);
+      _glthread_UNLOCK_MUTEX(oldObj->Mutex);
+
+      if (deleteFlag) {
+	 ASSERT(ctx->Driver.DeleteArrayObject);
+         ctx->Driver.DeleteArrayObject(ctx, oldObj);
+      }
+
+      *ptr = NULL;
+   }
+   ASSERT(!*ptr);
+
+   if (arrayObj) {
+      /* reference new array object */
+      _glthread_LOCK_MUTEX(arrayObj->Mutex);
+      if (arrayObj->RefCount == 0) {
+         /* this array's being deleted (look just above) */
+         /* Not sure this can every really happen.  Warn if it does. */
+         _mesa_problem(NULL, "referencing deleted array object");
+         *ptr = NULL;
+      }
+      else {
+         arrayObj->RefCount++;
+#if 0
+         printf("ArrayObj %p %d INCR to %d\n",
+                (void *) arrayObj, arrayObj->Name, arrayObj->RefCount);
+#endif
+         *ptr = arrayObj;
+      }
+      _glthread_UNLOCK_MUTEX(arrayObj->Mutex);
+   }
+}
+
+
+
 static void
 init_array(GLcontext *ctx,
            struct gl_client_array *array, GLint size, GLint type)
@@ -112,7 +198,8 @@ init_array(GLcontext *ctx,
    array->Normalized = GL_FALSE;
 #if FEATURE_ARB_vertex_buffer_object
    /* Vertex array buffers */
-   array->BufferObj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx, &array->BufferObj,
+                                 ctx->Shared->NullBufferObj);
 #endif
 }
 
@@ -129,6 +216,9 @@ _mesa_initialize_array_object( GLcontext *ctx,
 
    obj->Name = name;
 
+   _glthread_INIT_MUTEX(obj->Mutex);
+   obj->RefCount = 1;
+
    /* Init the individual arrays */
    init_array(ctx, &obj->Vertex, 4, GL_FLOAT);
    init_array(ctx, &obj->Normal, 3, GL_FLOAT);
@@ -153,8 +243,8 @@ _mesa_initialize_array_object( GLcontext *ctx,
 /**
  * Add the given array object to the array object pool.
  */
-void
-_mesa_save_array_object( GLcontext *ctx, struct gl_array_object *obj )
+static void
+save_array_object( GLcontext *ctx, struct gl_array_object *obj )
 {
    if (obj->Name > 0) {
       /* insert into hash table */
@@ -167,8 +257,8 @@ _mesa_save_array_object( GLcontext *ctx, struct gl_array_object *obj )
  * Remove the given array object from the array object pool.
  * Do not deallocate the array object though.
  */
-void
-_mesa_remove_array_object( GLcontext *ctx, struct gl_array_object *obj )
+static void
+remove_array_object( GLcontext *ctx, struct gl_array_object *obj )
 {
    if (obj->Name > 0) {
       /* remove from hash table */
@@ -177,15 +267,6 @@ _mesa_remove_array_object( GLcontext *ctx, struct gl_array_object *obj )
 }
 
 
-static void
-unbind_buffer_object( GLcontext *ctx, struct gl_buffer_object *bufObj )
-{
-   if (bufObj != ctx->Array.NullBufferObj) {
-      _mesa_reference_buffer_object(ctx, &bufObj, NULL);
-   }
-}
-
-
 /**********************************************************************/
 /* API Functions                                                      */
 /**********************************************************************/
@@ -212,7 +293,7 @@ _mesa_BindVertexArrayAPPLE( GLuint id )
       return;   /* rebinding the same array object- no change */
 
    /*
-    * Get pointer to new array object (newBufObj)
+    * Get pointer to new array object (newObj)
     */
    if (id == 0) {
       /* The spec says there is no array object named 0, but we use
@@ -226,21 +307,18 @@ _mesa_BindVertexArrayAPPLE( GLuint id )
       if (!newObj) {
          /* If this is a new array object id, allocate an array object now.
 	  */
-
 	 newObj = (*ctx->Driver.NewArrayObject)(ctx, id);
          if (!newObj) {
             _mesa_error(ctx, GL_OUT_OF_MEMORY, "glBindVertexArrayAPPLE");
             return;
          }
-         _mesa_save_array_object(ctx, newObj);
+         save_array_object(ctx, newObj);
       }
    }
 
-
    ctx->NewState |= _NEW_ARRAY;
    ctx->Array.NewState |= _NEW_ARRAY_ALL;
-   ctx->Array.ArrayObj = newObj;
-
+   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, newObj);
 
    /* Pass BindVertexArray call to device driver */
    if (ctx->Driver.BindArrayObject && newObj)
@@ -274,7 +352,6 @@ _mesa_DeleteVertexArraysAPPLE(GLsizei n, const GLuint *ids)
       if ( obj != NULL ) {
 	 ASSERT( obj->Name == ids[i] );
 
-
 	 /* If the array object is currently bound, the spec says "the binding
 	  * for that object reverts to zero and the default vertex array
 	  * becomes current."
@@ -283,28 +360,13 @@ _mesa_DeleteVertexArraysAPPLE(GLsizei n, const GLuint *ids)
 	    CALL_BindVertexArrayAPPLE( ctx->Exec, (0) );
 	 }
 
-#if FEATURE_ARB_vertex_buffer_object
-	 /* Unbind any buffer objects that might be bound to arrays in
-	  * this array object.
-	  */
-	 unbind_buffer_object( ctx, obj->Vertex.BufferObj );
-	 unbind_buffer_object( ctx, obj->Normal.BufferObj );
-	 unbind_buffer_object( ctx, obj->Color.BufferObj );
-	 unbind_buffer_object( ctx, obj->SecondaryColor.BufferObj );
-	 unbind_buffer_object( ctx, obj->FogCoord.BufferObj );
-	 unbind_buffer_object( ctx, obj->Index.BufferObj );
-	 for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
-	    unbind_buffer_object( ctx, obj->TexCoord[i].BufferObj );
-	 }
-	 unbind_buffer_object( ctx, obj->EdgeFlag.BufferObj );
-	 for (i = 0; i < VERT_ATTRIB_MAX; i++) {
-	    unbind_buffer_object( ctx, obj->VertexAttrib[i].BufferObj );
-	 }
-#endif
-
 	 /* The ID is immediately freed for re-use */
-	 _mesa_remove_array_object(ctx, obj);
-	 ctx->Driver.DeleteArrayObject(ctx, obj);
+	 remove_array_object(ctx, obj);
+
+         /* Unreference the array object. 
+          * If refcount hits zero, the object will be deleted.
+          */
+         _mesa_reference_array_object(ctx, &obj, NULL);
       }
    }
 
@@ -353,7 +415,7 @@ _mesa_GenVertexArraysAPPLE(GLsizei n, GLuint *arrays)
          _mesa_error(ctx, GL_OUT_OF_MEMORY, "glGenVertexArraysAPPLE");
          return;
       }
-      _mesa_save_array_object(ctx, obj);
+      save_array_object(ctx, obj);
       arrays[i] = first + i;
    }
 
diff --git a/src/mesa/main/arrayobj.h b/src/mesa/main/arrayobj.h
index c7d66ec1669..9c2f340cfae 100644
--- a/src/mesa/main/arrayobj.h
+++ b/src/mesa/main/arrayobj.h
@@ -41,18 +41,20 @@
  * Internal functions
  */
 
-struct gl_array_object * _mesa_new_array_object( GLcontext *ctx,
-    GLuint name );
+extern struct gl_array_object *
+_mesa_new_array_object( GLcontext *ctx, GLuint name );
 
-void _mesa_delete_array_object( GLcontext *ctx, struct gl_array_object *obj );
+extern void
+_mesa_delete_array_object( GLcontext *ctx, struct gl_array_object *obj );
 
-void _mesa_initialize_array_object( GLcontext *ctx,
-    struct gl_array_object *obj, GLuint name );
-
-void _mesa_save_array_object( GLcontext *ctx, struct gl_array_object *obj );
-
-void _mesa_remove_array_object( GLcontext *ctx, struct gl_array_object *obj );
+extern void
+_mesa_reference_array_object(GLcontext *ctx,
+                             struct gl_array_object **ptr,
+                             struct gl_array_object *arrayObj);
 
+extern void
+_mesa_initialize_array_object( GLcontext *ctx,
+                               struct gl_array_object *obj, GLuint name );
 
 
 /*
diff --git a/src/mesa/main/bufferobj.c b/src/mesa/main/bufferobj.c
index c8d160baa9a..1f2070ef473 100644
--- a/src/mesa/main/bufferobj.c
+++ b/src/mesa/main/bufferobj.c
@@ -194,7 +194,7 @@ _mesa_reference_buffer_object(GLcontext *ctx,
       return;
 
    if (*ptr) {
-      /* Unreference the old texture */
+      /* Unreference the old buffer */
       GLboolean deleteFlag = GL_FALSE;
       struct gl_buffer_object *oldObj = *ptr;
 
@@ -227,7 +227,7 @@ _mesa_reference_buffer_object(GLcontext *ctx,
    ASSERT(!*ptr);
 
    if (bufObj) {
-      /* reference new texture */
+      /* reference new buffer */
       /*_glthread_LOCK_MUTEX(tex->Mutex);*/
       if (bufObj->RefCount == 0) {
          /* this buffer's being deleted (look just above) */
@@ -389,7 +389,6 @@ _mesa_buffer_map( GLcontext *ctx, GLenum target, GLenum access,
    (void) ctx;
    (void) target;
    (void) access;
-   ASSERT(!bufObj->OnCard);
    /* Just return a direct pointer to the data */
    if (bufObj->Pointer) {
       /* already mapped! */
@@ -413,7 +412,6 @@ _mesa_buffer_unmap( GLcontext *ctx, GLenum target,
 {
    (void) ctx;
    (void) target;
-   ASSERT(!bufObj->OnCard);
    /* XXX we might assert here that bufObj->Pointer is non-null */
    bufObj->Pointer = NULL;
    return GL_TRUE;
@@ -426,16 +424,8 @@ _mesa_buffer_unmap( GLcontext *ctx, GLenum target,
 void
 _mesa_init_buffer_objects( GLcontext *ctx )
 {
-   /* Allocate the default buffer object and set refcount so high that
-    * it never gets deleted.
-    * XXX with recent/improved refcounting this may not longer be needed.
-    */
-   ctx->Array.NullBufferObj = _mesa_new_buffer_object(ctx, 0, 0);
-   if (ctx->Array.NullBufferObj)
-      ctx->Array.NullBufferObj->RefCount = 1000;
-
-   ctx->Array.ArrayBufferObj = ctx->Array.NullBufferObj;
-   ctx->Array.ElementArrayBufferObj = ctx->Array.NullBufferObj;
+   ctx->Array.ArrayBufferObj = ctx->Shared->NullBufferObj;
+   ctx->Array.ElementArrayBufferObj = ctx->Shared->NullBufferObj;
 }
 
 
@@ -479,7 +469,7 @@ bind_buffer_object(GLcontext *ctx, GLenum target, GLuint buffer)
       /* The spec says there's not a buffer object named 0, but we use
        * one internally because it simplifies things.
        */
-      newBufObj = ctx->Array.NullBufferObj;
+      newBufObj = ctx->Shared->NullBufferObj;
    }
    else {
       /* non-default buffer object */
@@ -746,7 +736,7 @@ unbind(GLcontext *ctx,
        struct gl_buffer_object *obj)
 {
    if (*ptr == obj) {
-      _mesa_reference_buffer_object(ctx, ptr, ctx->Array.NullBufferObj);
+      _mesa_reference_buffer_object(ctx, ptr, ctx->Shared->NullBufferObj);
    }
 }
 
@@ -958,8 +948,12 @@ _mesa_BufferDataARB(GLenum target, GLsizeiptrARB size,
       bufObj->Pointer = NULL;
    }  
 
+   FLUSH_VERTICES(ctx, _NEW_BUFFER_OBJECT);
+
    ASSERT(ctx->Driver.BufferData);
 
+   bufObj->Written = GL_TRUE;
+
    /* Give the buffer object to the driver!  <data> may be null! */
    ctx->Driver.BufferData( ctx, target, size, data, usage, bufObj );
 }
@@ -980,6 +974,8 @@ _mesa_BufferSubDataARB(GLenum target, GLintptrARB offset,
       return;
    }
 
+   bufObj->Written = GL_TRUE;
+
    ASSERT(ctx->Driver.BufferSubData);
    ctx->Driver.BufferSubData( ctx, target, offset, size, data, bufObj );
 }
@@ -1044,6 +1040,8 @@ _mesa_MapBufferARB(GLenum target, GLenum access)
    }
 
    bufObj->Access = access;
+   if (access == GL_WRITE_ONLY_ARB || access == GL_READ_WRITE_ARB)
+      bufObj->Written = GL_TRUE;
 
    return bufObj->Pointer;
 }
diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c
index c5f13345f04..d8b5f3b1f4a 100644
--- a/src/mesa/main/buffers.c
+++ b/src/mesa/main/buffers.c
@@ -443,7 +443,7 @@ _mesa_readbuffer(GLcontext *ctx, GLenum buffer, GLint bufferIndex)
    fb->ColorReadBuffer = buffer;
    fb->_ColorReadBufferIndex = bufferIndex;
 
-   ctx->NewState |= _NEW_PIXEL;
+   ctx->NewState |= _NEW_BUFFERS;
 }
 
 
diff --git a/src/mesa/main/config.h b/src/mesa/main/config.h
index 2a9fdf9ca05..888e08ff934 100644
--- a/src/mesa/main/config.h
+++ b/src/mesa/main/config.h
@@ -187,7 +187,7 @@
 #define MAX_PROGRAM_TEMPS 256
 #define MAX_PROGRAM_ADDRESS_REGS 2
 #define MAX_UNIFORMS 1024  /**< number of vec4 uniforms */
-#define MAX_VARYING 8      /**< number of float[4] vectors */
+#define MAX_VARYING 16     /**< number of float[4] vectors */
 #define MAX_SAMPLERS MAX_TEXTURE_IMAGE_UNITS
 #define MAX_PROGRAM_INPUTS 32
 #define MAX_PROGRAM_OUTPUTS 32
diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c
index 5726dbd983e..5e0f2d7b1bd 100644
--- a/src/mesa/main/context.c
+++ b/src/mesa/main/context.c
@@ -434,7 +434,7 @@ one_time_init( GLcontext *ctx )
       }
 
 #if defined(DEBUG) && defined(__DATE__) && defined(__TIME__)
-      _mesa_debug(ctx, "Mesa %s DEBUG build %s %s",
+      _mesa_debug(ctx, "Mesa %s DEBUG build %s %s\n",
                   MESA_VERSION_STRING, __DATE__, __TIME__);
 #endif
 
@@ -602,6 +602,10 @@ _mesa_init_constants(GLcontext *ctx)
    ASSERT(MAX_NV_VERTEX_PROGRAM_TEMPS <= MAX_PROGRAM_TEMPS);
    ASSERT(MAX_NV_VERTEX_PROGRAM_INPUTS <= VERT_ATTRIB_MAX);
    ASSERT(MAX_NV_VERTEX_PROGRAM_OUTPUTS <= VERT_RESULT_MAX);
+
+   /* check that we don't exceed various 32-bit bitfields */
+   ASSERT(VERT_RESULT_MAX <= 32);
+   ASSERT(FRAG_ATTRIB_MAX <= 32);
 }
 
 
@@ -1005,9 +1009,6 @@ _mesa_free_context_data( GLcontext *ctx )
    _mesa_free_query_data(ctx);
 #endif
 
-#if FEATURE_ARB_vertex_buffer_object
-   _mesa_delete_buffer_object(ctx, ctx->Array.NullBufferObj);
-#endif
    _mesa_delete_array_object(ctx, ctx->Array.DefaultArrayObj);
 
    /* free dispatch tables */
@@ -1397,14 +1398,21 @@ _mesa_share_state(GLcontext *ctx, GLcontext *ctxToShare)
 {
    if (ctx && ctxToShare && ctx->Shared && ctxToShare->Shared) {
       struct gl_shared_state *oldSharedState = ctx->Shared;
+      GLint RefCount;
 
       ctx->Shared = ctxToShare->Shared;
+      
+      _glthread_LOCK_MUTEX(ctx->Shared->Mutex);
       ctx->Shared->RefCount++;
+      _glthread_UNLOCK_MUTEX(ctx->Shared->Mutex);
 
       update_default_objects(ctx);
 
-      oldSharedState->RefCount--;
-      if (oldSharedState->RefCount == 0) {
+      _glthread_LOCK_MUTEX(oldSharedState->Mutex);
+      RefCount = --oldSharedState->RefCount;
+      _glthread_UNLOCK_MUTEX(oldSharedState->Mutex);
+
+      if (RefCount == 0) {
          _mesa_free_shared_state(ctx, oldSharedState);
       }
 
@@ -1497,6 +1505,7 @@ _mesa_Finish(void)
 {
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
+   FLUSH_CURRENT( ctx, 0 );
    if (ctx->Driver.Finish) {
       ctx->Driver.Finish(ctx);
    }
@@ -1521,4 +1530,17 @@ _mesa_Flush(void)
 }
 
 
+/**
+ * Set mvp_with_dp4 flag.  If a driver has a preference for DP4 over
+ * MUL/MAD, or vice versa, call this function to register that.
+ * Otherwise we default to MUL/MAD.
+ */
+void
+_mesa_set_mvp_with_dp4( GLcontext *ctx,
+                        GLboolean flag )
+{
+   ctx->mvp_with_dp4 = flag;
+}
+
+
 /*@}*/
diff --git a/src/mesa/main/context.h b/src/mesa/main/context.h
index ecc1cec7799..5b57d88029c 100644
--- a/src/mesa/main/context.h
+++ b/src/mesa/main/context.h
@@ -151,6 +151,10 @@ extern struct _glapi_table *
 _mesa_get_dispatch(GLcontext *ctx);
 
 
+void
+_mesa_set_mvp_with_dp4( GLcontext *ctx,
+                        GLboolean flag );
+
 
 /** \name Miscellaneous */
 /*@{*/
diff --git a/src/mesa/main/debug.c b/src/mesa/main/debug.c
index fdd10dd3074..2eabcdaf493 100644
--- a/src/mesa/main/debug.c
+++ b/src/mesa/main/debug.c
@@ -3,6 +3,7 @@
  * Version:  6.5
  *
  * Copyright (C) 1999-2005  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -262,7 +263,7 @@ static void
 write_texture_image(struct gl_texture_object *texObj)
 {
    const struct gl_texture_image *img = texObj->Image[0][0];
-   if (img) {
+   if (img && img->Data) {
       char s[100];
 
       /* make filename */
@@ -338,5 +339,5 @@ _mesa_dump_textures(GLboolean dumpImages)
 {
    GET_CURRENT_CONTEXT(ctx);
    DumpImages = dumpImages;
-   _mesa_HashDeleteAll(ctx->Shared->TexObjects, dump_texture_cb, ctx);
+   _mesa_HashWalk(ctx->Shared->TexObjects, dump_texture_cb, ctx);
 }
diff --git a/src/mesa/main/enable.c b/src/mesa/main/enable.c
index f432be183cb..2e7baa48ff5 100644
--- a/src/mesa/main/enable.c
+++ b/src/mesa/main/enable.c
@@ -222,14 +222,16 @@ get_texcoord_unit(GLcontext *ctx)
 
 /**
  * Helper function to enable or disable a texture target.
+ * \param bit  one of the TEXTURE_x_BIT values
+ * \return GL_TRUE if state is changing or GL_FALSE if no change
  */
 static GLboolean
-enable_texture(GLcontext *ctx, GLboolean state, GLbitfield bit)
+enable_texture(GLcontext *ctx, GLboolean state, GLbitfield texBit)
 {
    const GLuint curr = ctx->Texture.CurrentUnit;
    struct gl_texture_unit *texUnit = &ctx->Texture.Unit[curr];
-   const GLuint newenabled = (!state)
-       ? (texUnit->Enabled & ~bit) :  (texUnit->Enabled | bit);
+   const GLbitfield newenabled = state
+      ? (texUnit->Enabled | texBit) : (texUnit->Enabled & ~texBit);
 
    if (!ctx->DrawBuffer->Visual.rgbMode || texUnit->Enabled == newenabled)
        return GL_FALSE;
diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c
index 1ce5685af45..43325b13529 100644
--- a/src/mesa/main/ffvertex_prog.c
+++ b/src/mesa/main/ffvertex_prog.c
@@ -315,12 +315,6 @@ static void make_state_key( GLcontext *ctx, struct state_key *key )
  */
 #define DISASSEM 0
 
-/* Should be tunable by the driver - do we want to do matrix
- * multiplications with DP4's or with MUL/MAD's?  SSE works better
- * with the latter, drivers may differ.
- */
-#define PREFER_DP4 0
-
 
 /* Use uregs to represent registers internally, translate to Mesa's
  * expected formats on emit.  
@@ -348,6 +342,7 @@ struct tnl_program {
    const struct state_key *state;
    struct gl_vertex_program *program;
    GLint max_inst;  /** number of instructions allocated for program */
+   GLboolean mvp_with_dp4;
    
    GLuint temp_in_use;
    GLuint temp_reserved;
@@ -775,7 +770,7 @@ static struct ureg get_eye_position( struct tnl_program *p )
 
       p->eye_position = reserve_temp(p);
 
-      if (PREFER_DP4) {
+      if (p->mvp_with_dp4) {
 	 register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3,
                                  0, modelview );
 
@@ -881,7 +876,7 @@ static void build_hpos( struct tnl_program *p )
    struct ureg hpos = register_output( p, VERT_RESULT_HPOS );
    struct ureg mvp[4];
 
-   if (PREFER_DP4) {
+   if (p->mvp_with_dp4) {
       register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3, 
 			      0, mvp );
       emit_matrix_transform_vec4( p, hpos, mvp, pos );
@@ -1574,7 +1569,7 @@ static void build_texture_transform( struct tnl_program *p )
 	    struct ureg in = (!is_undef(out_texgen) ? 
 			      out_texgen : 
 			      register_input(p, VERT_ATTRIB_TEX0+i));
-	    if (PREFER_DP4) {
+	    if (p->mvp_with_dp4) {
 	       register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3,
 				       0, texmat );
 	       emit_matrix_transform_vec4( p, out, texmat, in );
@@ -1708,6 +1703,7 @@ static void build_tnl_program( struct tnl_program *p )
 static void
 create_new_program( const struct state_key *key,
                     struct gl_vertex_program *program,
+                    GLboolean mvp_with_dp4,
                     GLuint max_temps)
 {
    struct tnl_program p;
@@ -1721,6 +1717,7 @@ create_new_program( const struct state_key *key,
    p.transformed_normal = undef;
    p.identity = undef;
    p.temp_in_use = 0;
+   p.mvp_with_dp4 = mvp_with_dp4;
    
    if (max_temps >= sizeof(int) * 8)
       p.temp_reserved = 0;
@@ -1776,6 +1773,7 @@ _mesa_get_fixed_func_vertex_program(GLcontext *ctx)
          return NULL;
 
       create_new_program( &key, prog,
+                          ctx->mvp_with_dp4,
                           ctx->Const.VertexProgram.MaxTemps );
 
 #if 0
diff --git a/src/mesa/main/histogram.c b/src/mesa/main/histogram.c
index 905c1ad8301..5fee4fd0e34 100644
--- a/src/mesa/main/histogram.c
+++ b/src/mesa/main/histogram.c
@@ -975,6 +975,8 @@ _mesa_Histogram(GLenum target, GLsizei width, GLenum internalFormat, GLboolean s
       }
    }
 
+   FLUSH_VERTICES(ctx, _NEW_PIXEL);
+
    /* reset histograms */
    for (i = 0; i < HISTOGRAM_TABLE_SIZE; i++) {
       ctx->Histogram.Count[i][0] = 0;
@@ -1002,8 +1004,6 @@ _mesa_Histogram(GLenum target, GLsizei width, GLenum internalFormat, GLboolean s
       ctx->Histogram.AlphaSize     = 8 * sizeof(GLuint);
       ctx->Histogram.LuminanceSize = 8 * sizeof(GLuint);
    }
-
-   ctx->NewState |= _NEW_PIXEL;
 }
 
 
@@ -1058,8 +1058,6 @@ _mesa_ResetHistogram(GLenum target)
       ctx->Histogram.Count[i][2] = 0;
       ctx->Histogram.Count[i][3] = 0;
    }
-
-   ctx->NewState |= _NEW_PIXEL;
 }
 
 
@@ -1083,7 +1081,6 @@ _mesa_ResetMinmax(GLenum target)
    ctx->MinMax.Min[GCOMP] = 1000;    ctx->MinMax.Max[GCOMP] = -1000;
    ctx->MinMax.Min[BCOMP] = 1000;    ctx->MinMax.Max[BCOMP] = -1000;
    ctx->MinMax.Min[ACOMP] = 1000;    ctx->MinMax.Max[ACOMP] = -1000;
-   ctx->NewState |= _NEW_PIXEL;
 }
 
 
diff --git a/src/mesa/main/image.c b/src/mesa/main/image.c
index ddae456fa12..ea76ed04e4f 100644
--- a/src/mesa/main/image.c
+++ b/src/mesa/main/image.c
@@ -181,6 +181,8 @@ _mesa_sizeof_type( GLenum type )
 	 return sizeof(GLint);
       case GL_FLOAT:
 	 return sizeof(GLfloat);
+      case GL_DOUBLE:
+	 return sizeof(GLdouble);
       case GL_HALF_FLOAT_ARB:
 	 return sizeof(GLhalfARB);
       default:
diff --git a/src/mesa/main/imports.c b/src/mesa/main/imports.c
index 2ac93a52371..615f7c9a6da 100644
--- a/src/mesa/main/imports.c
+++ b/src/mesa/main/imports.c
@@ -980,7 +980,8 @@ _mesa_vsprintf( char *str, const char *fmt, va_list args )
 /*@{*/
 
 static void
-output_if_debug(const char *prefixString, const char *outputString)
+output_if_debug(const char *prefixString, const char *outputString,
+                GLboolean newline)
 {
    static int debug = -1;
 
@@ -1004,16 +1005,19 @@ output_if_debug(const char *prefixString, const char *outputString)
 
    /* Now only print the string if we're required to do so. */
    if (debug) {
-      fprintf(stderr, "%s: %s\n", prefixString, outputString);
+      fprintf(stderr, "%s: %s", prefixString, outputString);
+      if (newline)
+         fprintf(stderr, "\n");
    }
 }
 
+
 /**
  * Report a warning (a recoverable error condition) to stderr if
  * either DEBUG is defined or the MESA_DEBUG env var is set.
  *
  * \param ctx GL context.
- * \param fmtString printf() alike format string.
+ * \param fmtString printf()-like format string.
  */
 void
 _mesa_warning( GLcontext *ctx, const char *fmtString, ... )
@@ -1025,11 +1029,12 @@ _mesa_warning( GLcontext *ctx, const char *fmtString, ... )
    (void) vsnprintf( str, MAXSTRING, fmtString, args );
    va_end( args );
 
-   output_if_debug("Mesa warning", str);
+   output_if_debug("Mesa warning", str, GL_TRUE);
 }
 
+
 /**
- * Report an internla implementation problem.
+ * Report an internal implementation problem.
  * Prints the message to stderr via fprintf().
  *
  * \param ctx GL context.
@@ -1050,8 +1055,9 @@ _mesa_problem( const GLcontext *ctx, const char *fmtString, ... )
    fprintf(stderr, "Please report at bugzilla.freedesktop.org\n");
 }
 
+
 /**
- * Record an OpenGL state error.  These usually occur when the users
+ * Record an OpenGL state error.  These usually occur when the user
  * passes invalid parameters to a GL function.
  *
  * If debugging is enabled (either at compile-time via the DEBUG macro, or
@@ -1123,11 +1129,22 @@ _mesa_error( GLcontext *ctx, GLenum error, const char *fmtString, ... )
 	    errstr = "unknown";
 	    break;
       }
-      _mesa_debug(ctx, "User error: %s in %s\n", errstr, where);
+
+      {
+         char s[MAXSTRING], s2[MAXSTRING];
+         va_list args;
+         va_start(args, fmtString);
+         vsnprintf(s, MAXSTRING, fmtString, args);
+         va_end(args);
+
+         _mesa_snprintf(s2, MAXSTRING, "%s in %s", errstr, s);
+         output_if_debug("Mesa: User error", s2, GL_TRUE);
+      }
    }
 
    _mesa_record_error(ctx, error);
-}  
+}
+
 
 /**
  * Report debug information.  Print error message to stderr via fprintf().
@@ -1145,7 +1162,7 @@ _mesa_debug( const GLcontext *ctx, const char *fmtString, ... )
    va_start(args, fmtString);
    vsnprintf(s, MAXSTRING, fmtString, args);
    va_end(args);
-   output_if_debug("Mesa", s);
+   output_if_debug("Mesa", s, GL_FALSE);
 #endif /* DEBUG */
    (void) ctx;
    (void) fmtString;
diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
index 30c7cca3b5a..d11df535f24 100644
--- a/src/mesa/main/mtypes.h
+++ b/src/mesa/main/mtypes.h
@@ -1028,7 +1028,7 @@ struct gl_stencil_attrib
 
 /**
  * An index for each type of texture object.  These correspond to the GL
- * target target enums, such as GL_TEXTURE_2D, GL_TEXTURE_CUBE_MAP, etc.
+ * texture target enums, such as GL_TEXTURE_2D, GL_TEXTURE_CUBE_MAP, etc.
  * Note: the order is from highest priority to lowest priority.
  */
 typedef enum
@@ -1505,7 +1505,7 @@ struct gl_buffer_object
    GLsizeiptr Length;        /**< mapped length */
    GLsizeiptrARB Size;       /**< Size of storage in bytes */
    GLubyte *Data;            /**< Location of storage either in RAM or VRAM. */
-   GLboolean OnCard;         /**< Is buffer in VRAM? (hardware drivers) */
+   GLboolean Written;        /**< Ever written to? (for debugging) */
 };
 
 
@@ -1541,10 +1541,10 @@ struct gl_client_array
    const GLubyte *Ptr;          /**< Points to array data */
    GLboolean Enabled;		/**< Enabled flag is a boolean */
    GLboolean Normalized;        /**< GL_ARB_vertex_program */
+   GLuint _ElementSize;         /**< size of each element in bytes */
 
-   /**< GL_ARB_vertex_buffer_object */
-   struct gl_buffer_object *BufferObj;
-   GLuint _MaxElement;
+   struct gl_buffer_object *BufferObj;/**< GL_ARB_vertex_buffer_object */
+   GLuint _MaxElement;          /**< max element index into array buffer */
 };
 
 
@@ -1557,6 +1557,9 @@ struct gl_array_object
    /** Name of the array object as received from glGenVertexArrayAPPLE. */
    GLuint Name;
 
+   GLint RefCount;
+   _glthread_Mutex Mutex;
+
    /** Conventional vertex arrays */
    /*@{*/
    struct gl_client_array Vertex;
@@ -1583,7 +1586,10 @@ struct gl_array_object
  */
 struct gl_array_attrib
 {
+   /** Currently bound array object. See _mesa_BindVertexArrayAPPLE() */
    struct gl_array_object *ArrayObj;
+
+   /** The default vertex array object */
    struct gl_array_object *DefaultArrayObj;
 
    GLint ActiveTexture;		/**< Client Active Texture */
@@ -1593,7 +1599,6 @@ struct gl_array_attrib
    GLbitfield NewState;		/**< mask of _NEW_ARRAY_* values */
 
 #if FEATURE_ARB_vertex_buffer_object
-   struct gl_buffer_object *NullBufferObj;
    struct gl_buffer_object *ArrayBufferObj;
    struct gl_buffer_object *ElementArrayBufferObj;
 #endif
@@ -2050,6 +2055,9 @@ struct gl_shared_state
    /** Default texture objects (shared by all texture units) */
    struct gl_texture_object *DefaultTex[NUM_TEXTURE_TARGETS];
 
+   /** Fallback texture used when a bound texture is incomplete */
+   struct gl_texture_object *FallbackTex;
+
    /**
     * \name Thread safety and statechange notification for texture
     * objects. 
@@ -2061,6 +2069,8 @@ struct gl_shared_state
    GLuint TextureStateStamp;	        /**< state notification for shared tex */
    /*@}*/
 
+   /** Default buffer object for vertex arrays that aren't in VBOs */
+   struct gl_buffer_object *NullBufferObj;
 
    /**
     * \name Vertex/fragment programs
@@ -2616,6 +2626,7 @@ struct gl_matrix_stack
 #define _NEW_PROGRAM            0x8000000  /**< __GLcontextRec::VertexProgram */
 #define _NEW_CURRENT_ATTRIB     0x10000000  /**< __GLcontextRec::Current */
 #define _NEW_PROGRAM_CONSTANTS  0x20000000
+#define _NEW_BUFFER_OBJECT      0x40000000
 #define _NEW_ALL ~0
 /*@}*/
 
@@ -2977,6 +2988,12 @@ struct __GLcontextRec
    /** software compression/decompression supported or not */
    GLboolean Mesa_DXTn;
 
+   /** 
+    * Use dp4 (rather than mul/mad) instructions for position
+    * transformation?
+    */
+   GLboolean mvp_with_dp4;
+
    /** Core tnl module support */
    struct gl_tnl_module TnlModule;
 
diff --git a/src/mesa/main/pixel.c b/src/mesa/main/pixel.c
index 52781e048ec..d9f3e476e81 100644
--- a/src/mesa/main/pixel.c
+++ b/src/mesa/main/pixel.c
@@ -170,7 +170,7 @@ _mesa_PixelMapfv( GLenum map, GLsizei mapsize, const GLfloat *values )
          return;
       }
       /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+      ctx->DefaultPacking.BufferObj = ctx->Shared->NullBufferObj;
       buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
                                               GL_READ_ONLY_ARB,
                                               ctx->Unpack.BufferObj);
@@ -229,7 +229,7 @@ _mesa_PixelMapuiv(GLenum map, GLsizei mapsize, const GLuint *values )
          return;
       }
       /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+      ctx->DefaultPacking.BufferObj = ctx->Shared->NullBufferObj;
       buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
                                               GL_READ_ONLY_ARB,
                                               ctx->Unpack.BufferObj);
@@ -303,7 +303,7 @@ _mesa_PixelMapusv(GLenum map, GLsizei mapsize, const GLushort *values )
          return;
       }
       /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+      ctx->DefaultPacking.BufferObj = ctx->Shared->NullBufferObj;
       buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
                                               GL_READ_ONLY_ARB,
                                               ctx->Unpack.BufferObj);
@@ -371,7 +371,7 @@ _mesa_GetPixelMapfv( GLenum map, GLfloat *values )
          return;
       }
       /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+      ctx->DefaultPacking.BufferObj = ctx->Shared->NullBufferObj;
       buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
                                               GL_WRITE_ONLY_ARB,
                                               ctx->Pack.BufferObj);
@@ -432,7 +432,7 @@ _mesa_GetPixelMapuiv( GLenum map, GLuint *values )
          return;
       }
       /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+      ctx->DefaultPacking.BufferObj = ctx->Shared->NullBufferObj;
       buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
                                               GL_WRITE_ONLY_ARB,
                                               ctx->Pack.BufferObj);
@@ -494,7 +494,7 @@ _mesa_GetPixelMapusv( GLenum map, GLushort *values )
          return;
       }
       /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+      ctx->DefaultPacking.BufferObj = ctx->Shared->NullBufferObj;
       buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
                                               GL_WRITE_ONLY_ARB,
                                               ctx->Pack.BufferObj);
@@ -819,7 +819,7 @@ update_image_transfer_state(GLcontext *ctx)
 
 
 /**
- * Update meas pixel transfer derived state.
+ * Update mesa pixel transfer derived state.
  */
 void _mesa_update_pixel( GLcontext *ctx, GLuint new_state )
 {
diff --git a/src/mesa/main/pixelstore.c b/src/mesa/main/pixelstore.c
index ff1a6344cc9..6a641f83f27 100644
--- a/src/mesa/main/pixelstore.c
+++ b/src/mesa/main/pixelstore.c
@@ -245,7 +245,8 @@ _mesa_init_pixelstore( GLcontext *ctx )
    ctx->Pack.ClientStorage = GL_FALSE;
    ctx->Pack.Invert = GL_FALSE;
 #if FEATURE_EXT_pixel_buffer_object
-   ctx->Pack.BufferObj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx, &ctx->Pack.BufferObj,
+                                 ctx->Shared->NullBufferObj);
 #endif
    ctx->Unpack.Alignment = 4;
    ctx->Unpack.RowLength = 0;
@@ -258,7 +259,8 @@ _mesa_init_pixelstore( GLcontext *ctx )
    ctx->Unpack.ClientStorage = GL_FALSE;
    ctx->Unpack.Invert = GL_FALSE;
 #if FEATURE_EXT_pixel_buffer_object
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx, &ctx->Unpack.BufferObj,
+                                 ctx->Shared->NullBufferObj);
 #endif
 
    /*
@@ -278,6 +280,7 @@ _mesa_init_pixelstore( GLcontext *ctx )
    ctx->DefaultPacking.ClientStorage = GL_FALSE;
    ctx->DefaultPacking.Invert = GL_FALSE;
 #if FEATURE_EXT_pixel_buffer_object
-   ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx, &ctx->DefaultPacking.BufferObj,
+                                 ctx->Shared->NullBufferObj);
 #endif
 }
diff --git a/src/mesa/main/shared.c b/src/mesa/main/shared.c
index 193ac8970cf..759883743da 100644
--- a/src/mesa/main/shared.c
+++ b/src/mesa/main/shared.c
@@ -33,6 +33,7 @@
 #include "mtypes.h"
 #include "hash.h"
 #include "arrayobj.h"
+#include "bufferobj.h"
 #include "shared.h"
 #include "shader/program.h"
 #include "shader/shader_api.h"
@@ -92,6 +93,13 @@ _mesa_alloc_shared_state(GLcontext *ctx)
    shared->BufferObjects = _mesa_NewHashTable();
 #endif
 
+   /* Allocate the default buffer object and set refcount so high that
+    * it never gets deleted.
+    * XXX with recent/improved refcounting this may not longer be needed.
+    */
+   shared->NullBufferObj = _mesa_new_buffer_object(ctx, 0, 0);
+   shared->NullBufferObj->RefCount = 1000;
+
    shared->ArrayObjects = _mesa_NewHashTable();
 
    /* Create default texture objects */
@@ -190,6 +198,10 @@ delete_bufferobj_cb(GLuint id, void *data, void *userData)
 {
    struct gl_buffer_object *bufObj = (struct gl_buffer_object *) data;
    GLcontext *ctx = (GLcontext *) userData;
+   if (bufObj->Pointer) {
+      ctx->Driver.UnmapBuffer(ctx, 0, bufObj);
+      bufObj->Pointer = NULL;
+   }
    ctx->Driver.DeleteBuffer(ctx, bufObj);
 }
 
@@ -337,6 +349,10 @@ _mesa_free_shared_state(GLcontext *ctx, struct gl_shared_state *shared)
    _mesa_DeleteHashTable(shared->RenderBuffers);
 #endif
 
+#if FEATURE_ARB_vertex_buffer_object
+   _mesa_delete_buffer_object(ctx, shared->NullBufferObj);
+#endif
+
    /*
     * Free texture objects (after FBOs since some textures might have
     * been bound to FBOs).
diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c
index f18fc8f6837..a6411f7b621 100644
--- a/src/mesa/main/state.c
+++ b/src/mesa/main/state.c
@@ -64,110 +64,145 @@ update_separate_specular(GLcontext *ctx)
 
 
 /**
- * Update state dependent on vertex arrays.
+ * Compute the index of the last array element that can be safely accessed
+ * in a vertex array.  We can really only do this when the array lives in
+ * a VBO.
+ * The array->_MaxElement field will be updated.
+ * Later in glDrawArrays/Elements/etc we can do some bounds checking.
+ */
+static void
+compute_max_element(struct gl_client_array *array)
+{
+   assert(array->Enabled);
+   if (array->BufferObj->Name) {
+      /* Compute the max element we can access in the VBO without going
+       * out of bounds.
+       */
+      array->_MaxElement = ((GLsizeiptrARB) array->BufferObj->Size
+                            - (GLsizeiptrARB) array->Ptr + array->StrideB
+                            - array->_ElementSize) / array->StrideB;
+   }
+   else {
+      /* user-space array, no idea how big it is */
+      array->_MaxElement = 2 * 1000 * 1000 * 1000; /* just a big number */
+   }
+}
+
+
+/**
+ * Helper for update_arrays().
+ * \return  min(current min, array->_MaxElement).
+ */
+static GLuint
+update_min(GLuint min, struct gl_client_array *array)
+{
+   compute_max_element(array);
+   return MIN2(min, array->_MaxElement);
+}
+
+
+/**
+ * Update ctx->Array._MaxElement (the max legal index into all enabled arrays).
+ * Need to do this upon new array state or new buffer object state.
  */
 static void
 update_arrays( GLcontext *ctx )
 {
-   GLuint i, min;
+   struct gl_array_object *arrayObj = ctx->Array.ArrayObj;
+   GLuint i, min = ~0;
 
    /* find min of _MaxElement values for all enabled arrays */
 
    /* 0 */
    if (ctx->VertexProgram._Current
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_POS].Enabled) {
-      min = ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_POS]._MaxElement;
+       && arrayObj->VertexAttrib[VERT_ATTRIB_POS].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_POS]);
    }
-   else if (ctx->Array.ArrayObj->Vertex.Enabled) {
-      min = ctx->Array.ArrayObj->Vertex._MaxElement;
-   }
-   else {
-      /* can't draw anything without vertex positions! */
-      min = 0;
+   else if (arrayObj->Vertex.Enabled) {
+      min = update_min(min, &arrayObj->Vertex);
    }
 
    /* 1 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_WEIGHT].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_WEIGHT]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_WEIGHT].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_WEIGHT]);
    }
    /* no conventional vertex weight array */
 
    /* 2 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_NORMAL].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_NORMAL]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_NORMAL].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_NORMAL]);
    }
-   else if (ctx->Array.ArrayObj->Normal.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->Normal._MaxElement);
+   else if (arrayObj->Normal.Enabled) {
+      min = update_min(min, &arrayObj->Normal);
    }
 
    /* 3 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR0].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR0]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_COLOR0].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_COLOR0]);
    }
-   else if (ctx->Array.ArrayObj->Color.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->Color._MaxElement);
+   else if (arrayObj->Color.Enabled) {
+      min = update_min(min, &arrayObj->Color);
    }
 
    /* 4 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR1].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR1]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_COLOR1].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_COLOR1]);
    }
-   else if (ctx->Array.ArrayObj->SecondaryColor.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->SecondaryColor._MaxElement);
+   else if (arrayObj->SecondaryColor.Enabled) {
+      min = update_min(min, &arrayObj->SecondaryColor);
    }
 
    /* 5 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_FOG].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_FOG]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_FOG].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_FOG]);
    }
-   else if (ctx->Array.ArrayObj->FogCoord.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->FogCoord._MaxElement);
+   else if (arrayObj->FogCoord.Enabled) {
+      min = update_min(min, &arrayObj->FogCoord);
    }
 
    /* 6 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR_INDEX].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR_INDEX]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_COLOR_INDEX].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_COLOR_INDEX]);
    }
-   else if (ctx->Array.ArrayObj->Index.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->Index._MaxElement);
+   else if (arrayObj->Index.Enabled) {
+      min = update_min(min, &arrayObj->Index);
    }
 
-
    /* 7 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG]);
    }
 
    /* 8..15 */
    for (i = VERT_ATTRIB_TEX0; i <= VERT_ATTRIB_TEX7; i++) {
       if (ctx->VertexProgram._Enabled
-          && ctx->Array.ArrayObj->VertexAttrib[i].Enabled) {
-         min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[i]._MaxElement);
+          && arrayObj->VertexAttrib[i].Enabled) {
+         min = update_min(min, &arrayObj->VertexAttrib[i]);
       }
       else if (i - VERT_ATTRIB_TEX0 < ctx->Const.MaxTextureCoordUnits
-               && ctx->Array.ArrayObj->TexCoord[i - VERT_ATTRIB_TEX0].Enabled) {
-         min = MIN2(min, ctx->Array.ArrayObj->TexCoord[i - VERT_ATTRIB_TEX0]._MaxElement);
+               && arrayObj->TexCoord[i - VERT_ATTRIB_TEX0].Enabled) {
+         min = update_min(min, &arrayObj->TexCoord[i - VERT_ATTRIB_TEX0]);
       }
    }
 
    /* 16..31 */
    if (ctx->VertexProgram._Current) {
       for (i = VERT_ATTRIB_GENERIC0; i < VERT_ATTRIB_MAX; i++) {
-         if (ctx->Array.ArrayObj->VertexAttrib[i].Enabled) {
-            min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[i]._MaxElement);
+         if (arrayObj->VertexAttrib[i].Enabled) {
+            min = update_min(min, &arrayObj->VertexAttrib[i]);
          }
       }
    }
 
-   if (ctx->Array.ArrayObj->EdgeFlag.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->EdgeFlag._MaxElement);
+   if (arrayObj->EdgeFlag.Enabled) {
+      min = update_min(min, &arrayObj->EdgeFlag);
    }
 
    /* _MaxElement is one past the last legal array element */
@@ -547,7 +582,7 @@ _mesa_update_state_locked( GLcontext *ctx )
    if (new_state & _DD_NEW_SEPARATE_SPECULAR)
       update_separate_specular( ctx );
 
-   if (new_state & (_NEW_ARRAY | _NEW_PROGRAM))
+   if (new_state & (_NEW_ARRAY | _NEW_PROGRAM | _NEW_BUFFER_OBJECT))
       update_arrays( ctx );
 
    if (new_state & (_NEW_BUFFERS | _NEW_VIEWPORT))
diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c
index 8c03c36c753..76b46d700b5 100644
--- a/src/mesa/main/teximage.c
+++ b/src/mesa/main/teximage.c
@@ -51,6 +51,17 @@
 #include "mtypes.h"
 
 
+/**
+ * State changes which we care about for glCopyTex[Sub]Image() calls.
+ * In particular, we care about pixel transfer state and buffer state
+ * (such as glReadBuffer to make sure we read from the right renderbuffer).
+ */
+#define NEW_COPY_TEX_STATE (_MESA_NEW_TRANSFER_STATE | \
+                            _NEW_BUFFERS | \
+                            _NEW_PIXEL)
+
+
+
 /**
  * We allocate texture memory on 512-byte boundaries so we can use MMX/SSE
  * elsewhere.
@@ -3008,7 +3019,7 @@ _mesa_CopyTexImage1D( GLenum target, GLint level,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
-   if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
+   if (ctx->NewState & NEW_COPY_TEX_STATE)
       _mesa_update_state(ctx);
 
 #if FEATURE_convolve
@@ -3073,7 +3084,7 @@ _mesa_CopyTexImage2D( GLenum target, GLint level, GLenum internalFormat,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
-   if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
+   if (ctx->NewState & NEW_COPY_TEX_STATE)
       _mesa_update_state(ctx);
 
 #if FEATURE_convolve
@@ -3141,7 +3152,7 @@ _mesa_CopyTexSubImage1D( GLenum target, GLint level,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
-   if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
+   if (ctx->NewState & NEW_COPY_TEX_STATE)
       _mesa_update_state(ctx);
 
    if (copytexsubimage_error_check1(ctx, 1, target, level))
@@ -3196,7 +3207,7 @@ _mesa_CopyTexSubImage2D( GLenum target, GLint level,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
-   if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
+   if (ctx->NewState & NEW_COPY_TEX_STATE)
       _mesa_update_state(ctx);
 
    if (copytexsubimage_error_check1(ctx, 2, target, level))
@@ -3251,7 +3262,7 @@ _mesa_CopyTexSubImage3D( GLenum target, GLint level,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
-   if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
+   if (ctx->NewState & NEW_COPY_TEX_STATE)
       _mesa_update_state(ctx);
 
    if (copytexsubimage_error_check1(ctx, 3, target, level))
diff --git a/src/mesa/main/texobj.c b/src/mesa/main/texobj.c
index b63f747fe8d..0024efc0e64 100644
--- a/src/mesa/main/texobj.c
+++ b/src/mesa/main/texobj.c
@@ -662,6 +662,59 @@ _mesa_test_texobj_completeness( const GLcontext *ctx,
    }
 }
 
+
+/**
+ * Return pointer to a default/fallback texture.
+ * The texture is a 2D 8x8 RGBA texture with all texels = (0,0,0,1).
+ * That's the value a sampler should get when sampling from an
+ * incomplete texture.
+ */
+struct gl_texture_object *
+_mesa_get_fallback_texture(GLcontext *ctx)
+{
+   if (!ctx->Shared->FallbackTex) {
+      /* create fallback texture now */
+      static GLubyte texels[8 * 8][4];
+      struct gl_texture_object *texObj;
+      struct gl_texture_image *texImage;
+      GLuint i;
+
+      for (i = 0; i < 8 * 8; i++) {
+         texels[i][0] =
+         texels[i][1] =
+         texels[i][2] = 0x0;
+         texels[i][3] = 0xff;
+      }
+
+      /* create texture object */
+      texObj = ctx->Driver.NewTextureObject(ctx, 0, GL_TEXTURE_2D);
+      assert(texObj->RefCount == 1);
+      texObj->MinFilter = GL_NEAREST;
+      texObj->MagFilter = GL_NEAREST;
+
+      /* create level[0] texture image */
+      texImage = _mesa_get_tex_image(ctx, texObj, GL_TEXTURE_2D, 0);
+
+      /* init the image fields */
+      _mesa_init_teximage_fields(ctx, GL_TEXTURE_2D, texImage,
+                                    8, 8, 1, 0, GL_RGBA); 
+
+      /* set image data */
+      ctx->Driver.TexImage2D(ctx, GL_TEXTURE_2D, 0, GL_RGBA,
+                             8, 8, 0,
+                             GL_RGBA, GL_UNSIGNED_BYTE, texels,
+                             &ctx->DefaultPacking, texObj, texImage);
+
+      _mesa_test_texobj_completeness(ctx, texObj);
+      assert(texObj->_Complete);
+
+      ctx->Shared->FallbackTex = texObj;
+   }
+   return ctx->Shared->FallbackTex;
+}
+
+
+
 /*@}*/
 
 
diff --git a/src/mesa/main/texobj.h b/src/mesa/main/texobj.h
index d5374c5d6c4..2599c0816a9 100644
--- a/src/mesa/main/texobj.h
+++ b/src/mesa/main/texobj.h
@@ -65,6 +65,9 @@ extern void
 _mesa_test_texobj_completeness( const GLcontext *ctx,
                                 struct gl_texture_object *obj );
 
+extern struct gl_texture_object *
+_mesa_get_fallback_texture(GLcontext *ctx);
+
 extern void
 _mesa_unlock_context_textures( GLcontext *ctx );
 
diff --git a/src/mesa/main/texparam.c b/src/mesa/main/texparam.c
index 515a35cdfcf..2195a334d3e 100644
--- a/src/mesa/main/texparam.c
+++ b/src/mesa/main/texparam.c
@@ -69,7 +69,7 @@ validate_texture_wrap_mode(GLcontext * ctx, GLenum target, GLenum wrap)
       return GL_TRUE;
    }
 
-   _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param)" );
+   _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param=0x%x)", wrap );
    return GL_FALSE;
 }
 
@@ -209,7 +209,8 @@ set_tex_parameteri(GLcontext *ctx,
          }
          /* fall-through */
       default:
-         _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param)" );
+         _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param=0x%x)",
+                      params[0] );
       }
       return GL_FALSE;
 
@@ -223,7 +224,8 @@ set_tex_parameteri(GLcontext *ctx,
          texObj->MagFilter = params[0];
          return GL_TRUE;
       default:
-         _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param)" );
+         _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param=0x%x)",
+                      params[0]);
       }
       return GL_FALSE;
 
@@ -262,7 +264,8 @@ set_tex_parameteri(GLcontext *ctx,
          return GL_FALSE;
       if (params[0] < 0 ||
           (texObj->Target == GL_TEXTURE_RECTANGLE_ARB && params[0] != 0)) {
-         _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)");
+         _mesa_error(ctx, GL_INVALID_VALUE,
+                     "glTexParameter(param=%d)", params[0]);
          return GL_FALSE;
       }
       flush(ctx, texObj);
@@ -273,7 +276,8 @@ set_tex_parameteri(GLcontext *ctx,
       if (texObj->MaxLevel == params[0])
          return GL_FALSE;
       if (params[0] < 0 || texObj->Target == GL_TEXTURE_RECTANGLE_ARB) {
-         _mesa_error(ctx, GL_INVALID_OPERATION, "glTexParameter(param)");
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                     "glTexParameter(param=%d)", params[0]);
          return GL_FALSE;
       }
       flush(ctx, texObj);
@@ -340,7 +344,7 @@ set_tex_parameteri(GLcontext *ctx,
          }
       }
       else {
-         _mesa_error(ctx, GL_INVALID_ENUM, "glTexParameter(param)");
+         _mesa_error(ctx, GL_INVALID_ENUM, "glTexParameter(pname=0x%x)", pname);
       }
       return GL_FALSE;
 
@@ -465,8 +469,10 @@ set_tex_parameterf(GLcontext *ctx,
          return GL_TRUE;
       }
       else {
-         _mesa_error(ctx, GL_INVALID_ENUM,
-                     "glTexParameter(pname=GL_TEXTURE_MAX_ANISOTROPY_EXT)");
+         static GLuint count = 0;
+         if (count++ < 10)
+            _mesa_error(ctx, GL_INVALID_ENUM,
+                        "glTexParameter(pname=GL_TEXTURE_MAX_ANISOTROPY_EXT)");
       }
       return GL_FALSE;
 
diff --git a/src/mesa/main/texstate.c b/src/mesa/main/texstate.c
index 73f8a5339ec..6e0c0c688a1 100644
--- a/src/mesa/main/texstate.c
+++ b/src/mesa/main/texstate.c
@@ -561,8 +561,19 @@ update_texture_state( GLcontext *ctx )
       }
 
       if (!texUnit->_ReallyEnabled) {
-         _mesa_reference_texobj(&texUnit->_Current, NULL);
-         continue;
+         if (fprog) {
+            /* If we get here it means the shader is expecting a texture
+             * object, but there isn't one (or it's incomplete).  Use the
+             * fallback texture.
+             */
+            struct gl_texture_object *texObj = _mesa_get_fallback_texture(ctx);
+            texUnit->_ReallyEnabled = 1 << TEXTURE_2D_INDEX;
+            _mesa_reference_texobj(&texUnit->_Current, texObj);
+         }
+         else {
+            /* fixed-function: texture unit is really disabled */
+            continue;
+         }
       }
 
       /* if we get here, we know this texture unit is enabled */
@@ -780,6 +791,9 @@ _mesa_free_texture_data(GLcontext *ctx)
 
    /* unreference current textures */
    for (u = 0; u < MAX_TEXTURE_IMAGE_UNITS; u++) {
+      /* The _Current texture could account for another reference */
+      _mesa_reference_texobj(&ctx->Texture.Unit[u]._Current, NULL);
+
       for (tgt = 0; tgt < NUM_TEXTURE_TARGETS; tgt++) {
          _mesa_reference_texobj(&ctx->Texture.Unit[u].CurrentTex[tgt], NULL);
       }
diff --git a/src/mesa/main/varray.c b/src/mesa/main/varray.c
index 106252e4609..a9c9162be1b 100644
--- a/src/mesa/main/varray.c
+++ b/src/mesa/main/varray.c
@@ -63,22 +63,11 @@ update_array(GLcontext *ctx, struct gl_client_array *array,
    array->StrideB = stride ? stride : elementSize;
    array->Normalized = normalized;
    array->Ptr = (const GLubyte *) ptr;
-#if FEATURE_ARB_vertex_buffer_object
+   array->_ElementSize = elementSize;
+
    _mesa_reference_buffer_object(ctx, &array->BufferObj,
                                  ctx->Array.ArrayBufferObj);
 
-   /* Compute the index of the last array element that's inside the buffer.
-    * Later in glDrawArrays we'll check if start + count > _MaxElement to
-    * be sure we won't go out of bounds.
-    */
-   if (ctx->Array.ArrayBufferObj->Name)
-      array->_MaxElement = ((GLsizeiptrARB) ctx->Array.ArrayBufferObj->Size
-                            - (GLsizeiptrARB) array->Ptr + array->StrideB
-                            - elementSize) / array->StrideB;
-   else
-#endif
-      array->_MaxElement = 2 * 1000 * 1000 * 1000; /* just a big number */
-
    ctx->NewState |= _NEW_ARRAY;
    ctx->Array.NewState |= dirtyBit;
 }
@@ -1050,7 +1039,7 @@ void
 _mesa_init_varray(GLcontext *ctx)
 {
    ctx->Array.DefaultArrayObj = _mesa_new_array_object(ctx, 0);
-   ctx->Array.ArrayObj = ctx->Array.DefaultArrayObj;
-
+   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj,
+                                ctx->Array.DefaultArrayObj);
    ctx->Array.ActiveTexture = 0;   /* GL_ARB_multitexture */
 }
diff --git a/src/mesa/main/version.h b/src/mesa/main/version.h
index f5bf6e2c855..d4d3dd1a94e 100644
--- a/src/mesa/main/version.h
+++ b/src/mesa/main/version.h
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.5
+ * Version:  7.6
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -29,9 +30,9 @@
 
 /* Mesa version */
 #define MESA_MAJOR 7
-#define MESA_MINOR 5
+#define MESA_MINOR 6
 #define MESA_PATCH 0
-#define MESA_VERSION_STRING "7.5-devel"
+#define MESA_VERSION_STRING "7.6-devel"
 
 /* To make version comparison easy */
 #define MESA_VERSION(a,b,c) (((a) << 16) + ((b) << 8) + (c))
diff --git a/src/mesa/math/m_vector.c b/src/mesa/math/m_vector.c
index c5e2fd1de12..4cbab11a358 100644
--- a/src/mesa/math/m_vector.c
+++ b/src/mesa/math/m_vector.c
@@ -1,4 +1,3 @@
-
 /*
  * Mesa 3-D graphics library
  * Version:  3.5
@@ -37,11 +36,12 @@
 
 
 
-/*
+/**
  * Given a vector [count][4] of floats, set all the [][elt] values
  * to 0 (if elt = 0, 1, 2) or 1.0 (if elt = 3).
  */
-void _mesa_vector4f_clean_elem( GLvector4f *vec, GLuint count, GLuint elt )
+void
+_mesa_vector4f_clean_elem( GLvector4f *vec, GLuint count, GLuint elt )
 {
    static const GLubyte elem_bits[4] = {
       VEC_DIRTY_0,
@@ -54,12 +54,13 @@ void _mesa_vector4f_clean_elem( GLvector4f *vec, GLuint count, GLuint elt )
    GLfloat (*data)[4] = (GLfloat (*)[4])vec->start;
    GLuint i;
 
-   for (i = 0 ; i < count ; i++)
+   for (i = 0; i < count; i++)
       data[i][elt] = v;
 
    vec->flags &= ~elem_bits[elt];
 }
 
+
 static const GLubyte size_bits[5] = {
    0,
    VEC_SIZE_1,
@@ -69,61 +70,53 @@ static const GLubyte size_bits[5] = {
 };
 
 
-
-/*
+/**
  * Initialize GLvector objects.
- * Input: v - the vector object to initialize.
- *        flags - bitwise-OR of VEC_* flags
- *        storage - pointer to storage for the vector's data
+ * \param v  the vector object to initialize.
+ * \param flags  bitwise-OR of VEC_* flags
+ * \param storage  pointer to storage for the vector's data
  */
-
-
-void _mesa_vector4f_init( GLvector4f *v, GLuint flags, GLfloat (*storage)[4] )
+void
+_mesa_vector4f_init( GLvector4f *v, GLbitfield flags, GLfloat (*storage)[4] )
 {
    v->stride = 4 * sizeof(GLfloat);
    v->size = 2;   /* may change: 2-4 for vertices and 1-4 for texcoords */
    v->data = storage;
    v->start = (GLfloat *) storage;
    v->count = 0;
-   v->flags = size_bits[4] | flags ;
+   v->flags = size_bits[4] | flags;
 }
 
 
-
-
-/*
+/**
  * Initialize GLvector objects and allocate storage.
- * Input: v - the vector object
- *        sz - unused????
- *        flags - bitwise-OR of VEC_* flags
- *        count - number of elements to allocate in vector
- *        alignment - desired memory alignment for the data (in bytes)
+ * \param v  the vector object
+ * \param flags  bitwise-OR of VEC_* flags
+ * \param count  number of elements to allocate in vector
+ * \param alignment  desired memory alignment for the data (in bytes)
  */
-
-
-void _mesa_vector4f_alloc( GLvector4f *v, GLuint flags, GLuint count,
-			GLuint alignment )
+void
+_mesa_vector4f_alloc( GLvector4f *v, GLbitfield flags, GLuint count,
+                      GLuint alignment )
 {
    v->stride = 4 * sizeof(GLfloat);
    v->size = 2;
    v->storage = ALIGN_MALLOC( count * 4 * sizeof(GLfloat), alignment );
+   v->storage_count = count;
    v->start = (GLfloat *) v->storage;
    v->data = (GLfloat (*)[4]) v->storage;
    v->count = 0;
-   v->flags = size_bits[4] | flags | VEC_MALLOC ;
+   v->flags = size_bits[4] | flags | VEC_MALLOC;
 }
 
 
-
-
-/*
+/**
  * Vector deallocation.  Free whatever memory is pointed to by the
  * vector's storage field if the VEC_MALLOC flag is set.
  * DO NOT free the GLvector object itself, though.
  */
-
-
-void _mesa_vector4f_free( GLvector4f *v )
+void
+_mesa_vector4f_free( GLvector4f *v )
 {
    if (v->flags & VEC_MALLOC) {
       ALIGN_FREE( v->storage );
@@ -135,13 +128,15 @@ void _mesa_vector4f_free( GLvector4f *v )
 }
 
 
-/*
+/**
  * For debugging
  */
-void _mesa_vector4f_print( GLvector4f *v, GLubyte *cullmask, GLboolean culling )
+void
+_mesa_vector4f_print( const GLvector4f *v, const GLubyte *cullmask,
+                      GLboolean culling )
 {
-   GLfloat c[4] = { 0, 0, 0, 1 };
-   const char *templates[5] = {
+   static const GLfloat c[4] = { 0, 0, 0, 1 };
+   static const char *templates[5] = {
       "%d:\t0, 0, 0, 1\n",
       "%d:\t%f, 0, 0, 1\n",
       "%d:\t%f, %f, 0, 1\n",
@@ -154,30 +149,32 @@ void _mesa_vector4f_print( GLvector4f *v, GLubyte *cullmask, GLboolean culling )
    GLuint j, i = 0, count;
 
    _mesa_printf("data-start\n");
-   for ( ; d != v->start ; STRIDE_F(d, v->stride), i++)
+   for (; d != v->start; STRIDE_F(d, v->stride), i++)
       _mesa_printf(t, i, d[0], d[1], d[2], d[3]);
 
    _mesa_printf("start-count(%u)\n", v->count);
    count = i + v->count;
 
    if (culling) {
-      for ( ; i < count ; STRIDE_F(d, v->stride), i++)
+      for (; i < count; STRIDE_F(d, v->stride), i++)
 	 if (cullmask[i])
 	    _mesa_printf(t, i, d[0], d[1], d[2], d[3]);
    }
    else {
-      for ( ; i < count ; STRIDE_F(d, v->stride), i++)
+      for (; i < count; STRIDE_F(d, v->stride), i++)
 	 _mesa_printf(t, i, d[0], d[1], d[2], d[3]);
    }
 
-   for (j = v->size ; j < 4; j++) {
+   for (j = v->size; j < 4; j++) {
       if ((v->flags & (1<<j)) == 0) {
 
 	 _mesa_printf("checking col %u is clean as advertised ", j);
 
-	 for (i = 0, d = (GLfloat *) v->data ;
-	      i < count && d[j] == c[j] ;
-	      i++, STRIDE_F(d, v->stride)) {};
+	 for (i = 0, d = (GLfloat *) v->data;
+	      i < count && d[j] == c[j];
+	      i++, STRIDE_F(d, v->stride)) {
+            /* no-op */
+         }
 
 	 if (i == count)
 	    _mesa_printf(" --> ok\n");
@@ -186,5 +183,3 @@ void _mesa_vector4f_print( GLvector4f *v, GLubyte *cullmask, GLboolean culling )
       }
    }
 }
-
-
diff --git a/src/mesa/math/m_vector.h b/src/mesa/math/m_vector.h
index 647388ac7d0..71281d57589 100644
--- a/src/mesa/math/m_vector.h
+++ b/src/mesa/math/m_vector.h
@@ -31,7 +31,6 @@
 #define _M_VECTOR_H_
 
 #include "main/glheader.h"
-#include "main/mtypes.h"		/* hack for GLchan */
 
 
 #define VEC_DIRTY_0        0x1
@@ -50,7 +49,8 @@
 
 
 
-/* Wrap all the information about vectors up in a struct.  Has
+/**
+ * Wrap all the information about vectors up in a struct.  Has
  * additional fields compared to the other vectors to help us track of
  * different vertex sizes, and whether we need to clean columns out
  * because they contain non-(0,0,0,1) values.
@@ -60,29 +60,27 @@
  * the transformation routines.
  */
 typedef struct {
-   GLfloat (*data)[4];	/* may be malloc'd or point to client data */
-   GLfloat *start;	/* points somewhere inside of <data> */
-   GLuint count;	/* size of the vector (in elements) */
-   GLuint stride;	/* stride from one element to the next (in bytes) */
-   GLuint size;		/* 2-4 for vertices and 1-4 for texcoords */
-   GLuint flags;	/* which columns are dirty */
-   void *storage;	/* self-allocated storage */
+   GLfloat (*data)[4];	/**< may be malloc'd or point to client data */
+   GLfloat *start;	/**< points somewhere inside of <data> */
+   GLuint count;	/**< size of the vector (in elements) */
+   GLuint stride;	/**< stride from one element to the next (in bytes) */
+   GLuint size;		/**< 2-4 for vertices and 1-4 for texcoords */
+   GLbitfield flags;	/**< bitmask of VEC_x flags */
+   void *storage;	/**< self-allocated storage */
+   GLuint storage_count; /**< storage size in elements */
 } GLvector4f;
 
 
-extern void _mesa_vector4f_init( GLvector4f *v, GLuint flags,
+extern void _mesa_vector4f_init( GLvector4f *v, GLbitfield flags,
 			      GLfloat (*storage)[4] );
-extern void _mesa_vector4f_alloc( GLvector4f *v, GLuint flags,
+extern void _mesa_vector4f_alloc( GLvector4f *v, GLbitfield flags,
 			       GLuint count, GLuint alignment );
 extern void _mesa_vector4f_free( GLvector4f *v );
-extern void _mesa_vector4f_print( GLvector4f *v, GLubyte *, GLboolean );
+extern void _mesa_vector4f_print( const GLvector4f *v, const GLubyte *, GLboolean );
 extern void _mesa_vector4f_clean_elem( GLvector4f *vec, GLuint nr, GLuint elt );
 
 
-
-
-
-/*
+/**
  * Given vector <v>, return a pointer (cast to <type *> to the <i>-th element.
  *
  * End up doing a lot of slow imuls if not careful.
diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c
index b47bf360cf5..36267c336f6 100644
--- a/src/mesa/shader/arbprogparse.c
+++ b/src/mesa/shader/arbprogparse.c
@@ -563,6 +563,7 @@ struct var_cache
                                  * we take up with our state tokens or constants. Note that
                                  * this is _not_ the same as the number of param registers
                                  * we eventually use */
+   GLuint swizzle;              /**< swizzle to access this variable */
    struct var_cache *next;
 };
 
@@ -581,6 +582,7 @@ var_cache_create (struct var_cache **va)
       (**va).param_binding_begin = ~0;
       (**va).param_binding_length = ~0;
       (**va).alias_binding = NULL;
+      (**va).swizzle = SWIZZLE_XYZW;
       (**va).next = NULL;
    }
 }
@@ -872,15 +874,16 @@ parse_signed_float (const GLubyte ** inst, struct arb_program *Program)
  * This picks out a constant value from the parsed array. The constant vector is r
  * returned in the *values array, which should be of length 4.
  *
- * \param values - The 4 component vector with the constant value in it
+ * \param values - return the vector constant values.
+ * \param size - returns the number elements in valuesOut [1..4]
  */
 static GLvoid
-parse_constant (const GLubyte ** inst, GLfloat *values, struct arb_program *Program,
-                GLboolean use)
+parse_constant(const GLubyte ** inst, GLfloat *values, GLint *size,
+               struct arb_program *Program,
+               GLboolean use)
 {
    GLuint components, i;
 
-
    switch (*(*inst)++) {
       case CONSTANT_SCALAR:
          if (use == GL_TRUE) {
@@ -893,7 +896,7 @@ parse_constant (const GLubyte ** inst, GLfloat *values, struct arb_program *Prog
                values[1] =
                values[2] = values[3] = parse_signed_float (inst, Program);
          }
-
+         *size = 1;
          break;
       case CONSTANT_VECTOR:
          values[0] = values[1] = values[2] = 0;
@@ -902,7 +905,12 @@ parse_constant (const GLubyte ** inst, GLfloat *values, struct arb_program *Prog
          for (i = 0; i < components; i++) {
             values[i] = parse_signed_float (inst, Program);
          }
+         *size = 4;
          break;
+      default:
+         _mesa_problem(NULL, "unexpected case in parse_constant()");
+         values[0] = 0.0F;
+         *size = 0;
    }
 }
 
@@ -1513,10 +1521,16 @@ generic_attrib_check(struct var_cache *vc_head)
    curr = vc_head;
    while (curr) {
       if (curr->type == vt_attrib) {
-         if (curr->attrib_is_generic)
-            genericAttrib[ curr->attrib_binding ] = GL_TRUE;
-         else
+         if (curr->attrib_is_generic) {
+            GLuint attr = (curr->attrib_binding == 0)
+               ? 0 : (curr->attrib_binding - VERT_ATTRIB_GENERIC0);
+            assert(attr < MAX_VERTEX_PROGRAM_ATTRIBS);
+            genericAttrib[attr] = GL_TRUE;
+         }
+         else {
+            assert(curr->attrib_binding < MAX_VERTEX_PROGRAM_ATTRIBS);
             explicitAttrib[ curr->attrib_binding ] = GL_TRUE;
+         }
       }
 
       curr = curr->next;
@@ -1810,7 +1824,6 @@ parse_param_elements (GLcontext * ctx, const GLubyte ** inst,
    GLint idx;
    GLuint err = 0;
    gl_state_index state_tokens[STATE_LENGTH] = {0, 0, 0, 0, 0};
-   GLfloat const_values[4];
 
    GLubyte token = *(*inst)++;
 
@@ -1902,18 +1915,31 @@ parse_param_elements (GLcontext * ctx, const GLubyte ** inst,
 
       case PARAM_CONSTANT:
          /* parsing something like {1.0, 2.0, 3.0, 4.0} */
-         parse_constant (inst, const_values, Program, use);
-         idx = _mesa_add_named_constant(Program->Base.Parameters,
-                                        (char *) param_var->name,
-                                        const_values, 4);
-         if (param_var->param_binding_begin == ~0U)
-            param_var->param_binding_begin = idx;
-         param_var->param_binding_type = PROGRAM_STATE_VAR;
-         /* Note: when we reference this parameter in an instruction later,
-          * we'll check if it's really a constant/immediate and set the
-          * instruction register type appropriately.
-          */
-         param_var->param_binding_length++;
+         {
+            GLfloat const_values[4];
+            GLint size;
+            parse_constant(inst, const_values, &size, Program, use);
+            if (param_var->name[0] == ' ') {
+               /* this is an unnamed constant */
+               idx = _mesa_add_unnamed_constant(Program->Base.Parameters,
+                                                const_values, size,
+                                                &param_var->swizzle);
+            }
+            else {
+               /* named parameter/constant */
+               idx = _mesa_add_named_constant(Program->Base.Parameters,
+                                              (char *) param_var->name,
+                                              const_values, size);
+            }
+            if (param_var->param_binding_begin == ~0U)
+               param_var->param_binding_begin = idx;
+            param_var->param_binding_type = PROGRAM_STATE_VAR;
+            /* Note: when we reference this parameter in an instruction later,
+             * we'll check if it's really a constant/immediate and set the
+             * instruction register type appropriately.
+             */
+            param_var->param_binding_length++;
+         }
          break;
 
       default:
@@ -2422,6 +2448,9 @@ parse_swizzle_mask(const GLubyte ** inst, GLubyte *swizzle, GLint len)
             return;
       }
    }
+
+   if (len == 1)
+      swizzle[1] = swizzle[2] = swizzle[3] = swizzle[0];
 }
 
 
@@ -2476,7 +2505,7 @@ static GLuint
 parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
                struct var_cache **vc_head,
                struct arb_program *Program,
-               gl_register_file * File, GLint * Index,
+               gl_register_file * File, GLint * Index, GLuint *swizzle,
                GLboolean *IsRelOffset )
 {
    struct var_cache *src;
@@ -2485,6 +2514,8 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
 
    *IsRelOffset = 0;
 
+   *swizzle = SWIZZLE_XYZW; /* default */
+
    /* And the binding for the src */
    switch (*(*inst)++) {
       case REGISTER_ATTRIB:
@@ -2540,6 +2571,7 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
                      }
 
                      *Index = src->param_binding_begin + offset;
+                     *swizzle = src->swizzle;
                      break;
 
                   case ARRAY_INDEX_RELATIVE:
@@ -2562,6 +2594,7 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
                         /* And store it properly */
                         *Index = src->param_binding_begin + rel_off;
                         *IsRelOffset = 1;
+                        *swizzle = src->swizzle;
                      }
                      break;
                }
@@ -2573,6 +2606,7 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
 
                *File = (gl_register_file) src->param_binding_type;
                *Index = src->param_binding_begin;
+               *swizzle = src->swizzle;
                break;
          }
          break;
@@ -2641,6 +2675,21 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
 }
 
 
+static GLuint
+swizzle_swizzle(GLuint baseSwizzle, const GLubyte swizzle[4])
+{
+   GLuint i, swz, s[4];
+   for (i = 0; i < 4; i++) {
+      GLuint c = swizzle[i];
+      if (c <= SWIZZLE_W)
+         s[i] = GET_SWZ(baseSwizzle, c);
+      else
+         s[i] = c;
+   }
+   swz = MAKE_SWIZZLE4(s[0], s[1], s[2], s[3]);
+   return swz;
+}
+
 /**
  * Parse vertex/fragment program vector source register.
  */
@@ -2655,12 +2704,14 @@ parse_vector_src_reg(GLcontext *ctx, const GLubyte **inst,
    GLubyte negateMask;
    GLubyte swizzle[4];
    GLboolean isRelOffset;
+   GLuint baseSwizzle;
 
    /* Grab the sign */
    negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE;
 
    /* And the src reg */
-   if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &isRelOffset))
+   if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &baseSwizzle,
+                     &isRelOffset))
       return 1;
 
    /* finally, the swizzle */
@@ -2668,7 +2719,7 @@ parse_vector_src_reg(GLcontext *ctx, const GLubyte **inst,
 
    reg->File = file;
    reg->Index = index;
-   reg->Swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1], swizzle[2], swizzle[3]);
+   reg->Swizzle = swizzle_swizzle(baseSwizzle, swizzle);
    reg->Negate = negateMask;
    reg->RelAddr = isRelOffset;
    return 0;
@@ -2689,12 +2740,14 @@ parse_scalar_src_reg(GLcontext *ctx, const GLubyte **inst,
    GLubyte negateMask;
    GLubyte swizzle[4];
    GLboolean isRelOffset;
+   GLuint baseSwizzle;
 
    /* Grab the sign */
    negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE;
 
    /* And the src reg */
-   if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &isRelOffset))
+   if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &baseSwizzle,
+                     &isRelOffset))
       return 1;
 
    /* finally, the swizzle */
@@ -2702,7 +2755,7 @@ parse_scalar_src_reg(GLcontext *ctx, const GLubyte **inst,
 
    reg->File = file;
    reg->Index = index;
-   reg->Swizzle = (swizzle[0] << 0);
+   reg->Swizzle = swizzle_swizzle(baseSwizzle, swizzle);
    reg->Negate = negateMask;
    reg->RelAddr = isRelOffset;
    return 0;
@@ -3013,8 +3066,10 @@ parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst,
 	    GLubyte negateMask;
             gl_register_file file;
 	    GLint index;
+            GLuint baseSwizzle;
 
-	    if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index, &rel))
+	    if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index,
+                              &baseSwizzle, &rel))
 	       return 1;
 	    parse_extended_swizzle_mask(inst, swizzle, &negateMask);
 	    fp->SrcReg[0].File = file;
@@ -3354,11 +3409,13 @@ parse_vp_instruction (GLcontext * ctx, const GLubyte ** inst,
 	    GLboolean relAddr;
             gl_register_file file;
 	    GLint index;
+            GLuint baseSwizzle;
 
 	    if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg))
 	       return 1;
 
-	    if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index, &relAddr))
+	    if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index,
+                              &baseSwizzle, &relAddr))
 	       return 1;
 	    parse_extended_swizzle_mask (inst, swizzle, &negateMask);
 	    vp->SrcReg[0].File = file;
diff --git a/src/mesa/shader/arbprogram.c b/src/mesa/shader/arbprogram.c
index 981565ab8f1..317d623a228 100644
--- a/src/mesa/shader/arbprogram.c
+++ b/src/mesa/shader/arbprogram.c
@@ -74,8 +74,6 @@ _mesa_BindProgram(GLenum target, GLuint id)
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-
    /* Error-check target and get curProg */
    if ((target == GL_VERTEX_PROGRAM_ARB) && /* == GL_VERTEX_PROGRAM_NV */
         (ctx->Extensions.NV_vertex_program ||
@@ -132,6 +130,9 @@ _mesa_BindProgram(GLenum target, GLuint id)
       return;
    }
 
+   /* signal new program (and its new constants) */
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+
    /* bind newProg */
    if (target == GL_VERTEX_PROGRAM_ARB) { /* == GL_VERTEX_PROGRAM_NV */
       _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
@@ -489,7 +490,7 @@ _mesa_ProgramEnvParameter4fARB(GLenum target, GLuint index,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    if (target == GL_FRAGMENT_PROGRAM_ARB
        && ctx->Extensions.ARB_fragment_program) {
@@ -537,7 +538,7 @@ _mesa_ProgramEnvParameters4fvEXT(GLenum target, GLuint index, GLsizei count,
    GLfloat * dest;
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    if (count <= 0) {
       _mesa_error(ctx, GL_INVALID_VALUE, "glProgramEnvParameters4fv(count)");
@@ -631,7 +632,7 @@ _mesa_ProgramLocalParameter4fARB(GLenum target, GLuint index,
    struct gl_program *prog;
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    if ((target == GL_FRAGMENT_PROGRAM_NV
         && ctx->Extensions.NV_fragment_program) ||
@@ -685,7 +686,7 @@ _mesa_ProgramLocalParameters4fvEXT(GLenum target, GLuint index, GLsizei count,
    GLint i;
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    if (count <= 0) {
       _mesa_error(ctx, GL_INVALID_VALUE, "glProgramLocalParameters4fv(count)");
diff --git a/src/mesa/shader/nvprogram.c b/src/mesa/shader/nvprogram.c
index 5142c2a4a59..8ba521182b8 100644
--- a/src/mesa/shader/nvprogram.c
+++ b/src/mesa/shader/nvprogram.c
@@ -706,7 +706,7 @@ _mesa_ProgramNamedParameter4fNV(GLuint id, GLsizei len, const GLubyte *name,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    prog = _mesa_lookup_program(ctx, id);
    if (!prog || prog->Target != GL_FRAGMENT_PROGRAM_NV) {
diff --git a/src/mesa/shader/prog_instruction.c b/src/mesa/shader/prog_instruction.c
index ca7565c0911..ae3a003feed 100644
--- a/src/mesa/shader/prog_instruction.c
+++ b/src/mesa/shader/prog_instruction.c
@@ -285,6 +285,56 @@ _mesa_is_tex_instruction(gl_inst_opcode opcode)
 }
 
 
+/**
+ * Check if there's a potential src/dst register data dependency when
+ * using SOA execution.
+ * Example:
+ *   MOV T, T.yxwz;
+ * This would expand into:
+ *   MOV t0, t1;
+ *   MOV t1, t0;
+ *   MOV t2, t3;
+ *   MOV t3, t2;
+ * The second instruction will have the wrong value for t0 if executed as-is.
+ */
+GLboolean
+_mesa_check_soa_dependencies(const struct prog_instruction *inst)
+{
+   GLuint i, chan;
+
+   if (inst->DstReg.WriteMask == WRITEMASK_X ||
+       inst->DstReg.WriteMask == WRITEMASK_Y ||
+       inst->DstReg.WriteMask == WRITEMASK_Z ||
+       inst->DstReg.WriteMask == WRITEMASK_W ||
+       inst->DstReg.WriteMask == 0x0) {
+      /* no chance of data dependency */
+      return GL_FALSE;
+   }
+
+   /* loop over src regs */
+   for (i = 0; i < 3; i++) {
+      if (inst->SrcReg[i].File == inst->DstReg.File &&
+          inst->SrcReg[i].Index == inst->DstReg.Index) {
+         /* loop over dest channels */
+         GLuint channelsWritten = 0x0;
+         for (chan = 0; chan < 4; chan++) {
+            if (inst->DstReg.WriteMask & (1 << chan)) {
+               /* check if we're reading a channel that's been written */
+               GLuint swizzle = GET_SWZ(inst->SrcReg[i].Swizzle, chan);
+               if (swizzle <= SWIZZLE_W &&
+                   (channelsWritten & (1 << swizzle))) {
+                  return GL_TRUE;
+               }
+
+               channelsWritten |= (1 << chan);
+            }
+         }
+      }
+   }
+   return GL_FALSE;
+}
+
+
 /**
  * Return string name for given program opcode.
  */
diff --git a/src/mesa/shader/prog_instruction.h b/src/mesa/shader/prog_instruction.h
index 3109f6cbae5..40ad998f79d 100644
--- a/src/mesa/shader/prog_instruction.h
+++ b/src/mesa/shader/prog_instruction.h
@@ -428,6 +428,9 @@ _mesa_num_inst_dst_regs(gl_inst_opcode opcode);
 extern GLboolean
 _mesa_is_tex_instruction(gl_inst_opcode opcode);
 
+extern GLboolean
+_mesa_check_soa_dependencies(const struct prog_instruction *inst);
+
 extern const char *
 _mesa_opcode_string(gl_inst_opcode opcode);
 
diff --git a/src/mesa/shader/prog_optimize.c b/src/mesa/shader/prog_optimize.c
index 6ba2e76ff96..be903106a08 100644
--- a/src/mesa/shader/prog_optimize.c
+++ b/src/mesa/shader/prog_optimize.c
@@ -547,15 +547,13 @@ update_interval(GLint intBegin[], GLint intEnd[], GLuint index, GLuint ic)
 
 
 /**
- * Find the live intervals for each temporary register in the program.
- * For register R, the interval [A,B] indicates that R is referenced
- * from instruction A through instruction B.
- * Special consideration is needed for loops and subroutines.
- * \return GL_TRUE if success, GL_FALSE if we cannot proceed for some reason
+ * Find first/last instruction that references each temporary register.
  */
-static GLboolean
-find_live_intervals(struct gl_program *prog,
-                    struct interval_list *liveIntervals)
+GLboolean
+_mesa_find_temp_intervals(const struct prog_instruction *instructions,
+                          GLuint numInstructions,
+                          GLint intBegin[MAX_PROGRAM_TEMPS],
+                          GLint intEnd[MAX_PROGRAM_TEMPS])
 {
    struct loop_info
    {
@@ -563,26 +561,15 @@ find_live_intervals(struct gl_program *prog,
    };
    struct loop_info loopStack[MAX_LOOP_NESTING];
    GLuint loopStackDepth = 0;
-   GLint intBegin[MAX_PROGRAM_TEMPS], intEnd[MAX_PROGRAM_TEMPS];
    GLuint i;
 
-   /*
-    * Note: we'll return GL_FALSE below if we find relative indexing
-    * into the TEMP register file.  We can't handle that yet.
-    * We also give up on subroutines for now.
-    */
-
-   if (dbg) {
-      _mesa_printf("Optimize: Begin find intervals\n");
-   }
-
    for (i = 0; i < MAX_PROGRAM_TEMPS; i++){
       intBegin[i] = intEnd[i] = -1;
    }
 
    /* Scan instructions looking for temporary registers */
-   for (i = 0; i < prog->NumInstructions; i++) {
-      const struct prog_instruction *inst = prog->Instructions + i;
+   for (i = 0; i < numInstructions; i++) {
+      const struct prog_instruction *inst = instructions + i;
       if (inst->Opcode == OPCODE_BGNLOOP) {
          loopStack[loopStackDepth].Start = i;
          loopStack[loopStackDepth].End = inst->BranchTarget;
@@ -595,7 +582,7 @@ find_live_intervals(struct gl_program *prog,
          return GL_FALSE;
       }
       else {
-         const GLuint numSrc = _mesa_num_inst_src_regs(inst->Opcode);
+         const GLuint numSrc = 3;/*_mesa_num_inst_src_regs(inst->Opcode);*/
          GLuint j;
          for (j = 0; j < numSrc; j++) {
             if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) {
@@ -624,6 +611,39 @@ find_live_intervals(struct gl_program *prog,
       }
    }
 
+   return GL_TRUE;
+}
+
+
+/**
+ * Find the live intervals for each temporary register in the program.
+ * For register R, the interval [A,B] indicates that R is referenced
+ * from instruction A through instruction B.
+ * Special consideration is needed for loops and subroutines.
+ * \return GL_TRUE if success, GL_FALSE if we cannot proceed for some reason
+ */
+static GLboolean
+find_live_intervals(struct gl_program *prog,
+                    struct interval_list *liveIntervals)
+{
+   GLint intBegin[MAX_PROGRAM_TEMPS], intEnd[MAX_PROGRAM_TEMPS];
+   GLuint i;
+
+   /*
+    * Note: we'll return GL_FALSE below if we find relative indexing
+    * into the TEMP register file.  We can't handle that yet.
+    * We also give up on subroutines for now.
+    */
+
+   if (dbg) {
+      _mesa_printf("Optimize: Begin find intervals\n");
+   }
+
+   /* build intermediate arrays */
+   if (!_mesa_find_temp_intervals(prog->Instructions, prog->NumInstructions,
+                                  intBegin, intEnd))
+      return GL_FALSE;
+
    /* Build live intervals list from intermediate arrays */
    liveIntervals->Num = 0;
    for (i = 0; i < MAX_PROGRAM_TEMPS; i++) {
@@ -792,8 +812,6 @@ _mesa_reallocate_registers(struct gl_program *prog)
 }
 
 
-
-
 /**
  * Apply optimizations to the given program to eliminate unnecessary
  * instructions, temp regs, etc.
diff --git a/src/mesa/shader/prog_optimize.h b/src/mesa/shader/prog_optimize.h
index d102cfd9fc1..43894a27237 100644
--- a/src/mesa/shader/prog_optimize.h
+++ b/src/mesa/shader/prog_optimize.h
@@ -25,7 +25,19 @@
 #ifndef PROG_OPT_H
 #define PROG_OPT_H
 
+
+#include "main/config.h"
+
+
 struct gl_program;
+struct prog_instruction;
+
+
+extern GLboolean
+_mesa_find_temp_intervals(const struct prog_instruction *instructions,
+                          GLuint numInstructions,
+                          GLint intBegin[MAX_PROGRAM_TEMPS],
+                          GLint intEnd[MAX_PROGRAM_TEMPS]);
 
 extern void
 _mesa_optimize_program(GLcontext *ctx, struct gl_program *program);
diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c
index 9967f2978de..e6f9a910690 100644
--- a/src/mesa/shader/prog_print.c
+++ b/src/mesa/shader/prog_print.c
@@ -941,6 +941,10 @@ _mesa_write_shader_to_file(const struct gl_shader *shader)
       fprintf(f, "/*\n");
       _mesa_fprint_program_opt(f, shader->Program, PROG_PRINT_DEBUG, GL_TRUE);
       fprintf(f, "*/\n");
+      fprintf(f, "/* Parameters / constants */\n");
+      fprintf(f, "/*\n");
+      _mesa_fprint_parameter_list(f, shader->Program->Parameters);
+      fprintf(f, "*/\n");
    }
 
    fclose(f);
diff --git a/src/mesa/shader/programopt.c b/src/mesa/shader/programopt.c
index ecd98dc85c5..f70c75cec8e 100644
--- a/src/mesa/shader/programopt.c
+++ b/src/mesa/shader/programopt.c
@@ -45,8 +45,8 @@
  * into a vertex program.
  * May be used to implement the position_invariant option.
  */
-void
-_mesa_insert_mvp_code(GLcontext *ctx, struct gl_vertex_program *vprog)
+static void
+_mesa_insert_mvp_dp4_code(GLcontext *ctx, struct gl_vertex_program *vprog)
 {
    struct prog_instruction *newInst;
    const GLuint origLen = vprog->Base.NumInstructions;
@@ -113,6 +113,121 @@ _mesa_insert_mvp_code(GLcontext *ctx, struct gl_vertex_program *vprog)
 }
 
 
+static void
+_mesa_insert_mvp_mad_code(GLcontext *ctx, struct gl_vertex_program *vprog)
+{
+   struct prog_instruction *newInst;
+   const GLuint origLen = vprog->Base.NumInstructions;
+   const GLuint newLen = origLen + 4;
+   GLuint hposTemp;
+   GLuint i;
+
+   /*
+    * Setup state references for the modelview/projection matrix.
+    * XXX we should check if these state vars are already declared.
+    */
+   static const gl_state_index mvpState[4][STATE_LENGTH] = {
+      { STATE_MVP_MATRIX, 0, 0, 0, STATE_MATRIX_TRANSPOSE },
+      { STATE_MVP_MATRIX, 0, 1, 1, STATE_MATRIX_TRANSPOSE },
+      { STATE_MVP_MATRIX, 0, 2, 2, STATE_MATRIX_TRANSPOSE },
+      { STATE_MVP_MATRIX, 0, 3, 3, STATE_MATRIX_TRANSPOSE },
+   };
+   GLint mvpRef[4];
+
+   for (i = 0; i < 4; i++) {
+      mvpRef[i] = _mesa_add_state_reference(vprog->Base.Parameters,
+                                            mvpState[i]);
+   }
+
+   /* Alloc storage for new instructions */
+   newInst = _mesa_alloc_instructions(newLen);
+   if (!newInst) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY,
+                  "glProgramString(inserting position_invariant code)");
+      return;
+   }
+
+   /* TEMP hposTemp; */
+   hposTemp = vprog->Base.NumTemporaries++;
+
+   /*
+    * Generated instructions:
+    *    emit_op2(p, OPCODE_MUL, tmp, 0, swizzle1(src,X), mat[0]);
+    *    emit_op3(p, OPCODE_MAD, tmp, 0, swizzle1(src,Y), mat[1], tmp);
+    *    emit_op3(p, OPCODE_MAD, tmp, 0, swizzle1(src,Z), mat[2], tmp);
+    *    emit_op3(p, OPCODE_MAD, dest, 0, swizzle1(src,W), mat[3], tmp);
+    */
+   _mesa_init_instructions(newInst, 4);
+
+   newInst[0].Opcode = OPCODE_MUL;
+   newInst[0].DstReg.File = PROGRAM_TEMPORARY;
+   newInst[0].DstReg.Index = hposTemp;
+   newInst[0].DstReg.WriteMask = WRITEMASK_XYZW;
+   newInst[0].SrcReg[0].File = PROGRAM_INPUT;
+   newInst[0].SrcReg[0].Index = VERT_ATTRIB_POS;
+   newInst[0].SrcReg[0].Swizzle = SWIZZLE_XXXX;
+   newInst[0].SrcReg[1].File = PROGRAM_STATE_VAR;
+   newInst[0].SrcReg[1].Index = mvpRef[0];
+   newInst[0].SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+   for (i = 1; i <= 2; i++) {
+      newInst[i].Opcode = OPCODE_MAD;
+      newInst[i].DstReg.File = PROGRAM_TEMPORARY;
+      newInst[i].DstReg.Index = hposTemp;
+      newInst[i].DstReg.WriteMask = WRITEMASK_XYZW;
+      newInst[i].SrcReg[0].File = PROGRAM_INPUT;
+      newInst[i].SrcReg[0].Index = VERT_ATTRIB_POS;
+      newInst[i].SrcReg[0].Swizzle = MAKE_SWIZZLE4(i,i,i,i);
+      newInst[i].SrcReg[1].File = PROGRAM_STATE_VAR;
+      newInst[i].SrcReg[1].Index = mvpRef[i];
+      newInst[i].SrcReg[1].Swizzle = SWIZZLE_NOOP;
+      newInst[i].SrcReg[2].File = PROGRAM_TEMPORARY;
+      newInst[i].SrcReg[2].Index = hposTemp;
+      newInst[1].SrcReg[2].Swizzle = SWIZZLE_NOOP;
+   }
+
+   newInst[3].Opcode = OPCODE_MAD;
+   newInst[3].DstReg.File = PROGRAM_OUTPUT;
+   newInst[3].DstReg.Index = VERT_RESULT_HPOS;
+   newInst[3].DstReg.WriteMask = WRITEMASK_XYZW;
+   newInst[3].SrcReg[0].File = PROGRAM_INPUT;
+   newInst[3].SrcReg[0].Index = VERT_ATTRIB_POS;
+   newInst[3].SrcReg[0].Swizzle = SWIZZLE_WWWW;
+   newInst[3].SrcReg[1].File = PROGRAM_STATE_VAR;
+   newInst[3].SrcReg[1].Index = mvpRef[3];
+   newInst[3].SrcReg[1].Swizzle = SWIZZLE_NOOP;
+   newInst[3].SrcReg[2].File = PROGRAM_TEMPORARY;
+   newInst[3].SrcReg[2].Index = hposTemp;
+   newInst[3].SrcReg[2].Swizzle = SWIZZLE_NOOP;
+
+
+   /* Append original instructions after new instructions */
+   _mesa_copy_instructions (newInst + 4, vprog->Base.Instructions, origLen);
+
+   /* free old instructions */
+   _mesa_free_instructions(vprog->Base.Instructions, origLen);
+
+   /* install new instructions */
+   vprog->Base.Instructions = newInst;
+   vprog->Base.NumInstructions = newLen;
+   vprog->Base.InputsRead |= VERT_BIT_POS;
+   vprog->Base.OutputsWritten |= (1 << VERT_RESULT_HPOS);
+}
+
+
+void
+_mesa_insert_mvp_code(GLcontext *ctx, struct gl_vertex_program *vprog)
+{
+   if (ctx->mvp_with_dp4) 
+      _mesa_insert_mvp_dp4_code( ctx, vprog );
+   else
+      _mesa_insert_mvp_mad_code( ctx, vprog );
+}
+      
+
+
+
+
 
 /**
  * Append extra instructions onto the given fragment program to implement
diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c
index 644cd39185c..a8390d30942 100644
--- a/src/mesa/shader/shader_api.c
+++ b/src/mesa/shader/shader_api.c
@@ -1487,7 +1487,7 @@ _mesa_use_program(GLcontext *ctx, GLuint program)
       return;
    }
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
 
    if (program) {
       shProg = _mesa_lookup_shader_program_err(ctx, program, "glUseProgram");
@@ -1509,6 +1509,10 @@ _mesa_use_program(GLcontext *ctx, GLuint program)
                          shProg->Shaders[i]->Name,
                          shProg->Shaders[i]->Type);
          }
+         if (shProg->VertexProgram)
+            printf(" vert prog %u\n", shProg->VertexProgram->Base.Id);
+         if (shProg->FragmentProgram)
+            printf(" frag prog %u\n", shProg->FragmentProgram->Base.Id);
       }
    }
    else {
@@ -1789,7 +1793,7 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count,
       return;
    }
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    uniform = &shProg->Uniforms->Uniforms[location];
 
@@ -1929,7 +1933,7 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows,
       return;
    }
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    uniform = &shProg->Uniforms->Uniforms[location];
 
diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c
index ba2fc4f85c9..d7ad879e97a 100644
--- a/src/mesa/shader/slang/slang_compile.c
+++ b/src/mesa/shader/slang/slang_compile.c
@@ -2161,6 +2161,12 @@ parse_function(slang_parse_ctx * C, slang_output_ctx * O, int definition,
                                            (O->funs->num_functions + 1)
                                            * sizeof(slang_function));
       if (O->funs->functions == NULL) {
+         /* Make sure that there are no functions marked, as the
+          * allocation is currently NULL, in order to avoid
+          * a potental segfault as we clean up later.
+          */
+         O->funs->num_functions = 0;
+
          slang_info_log_memory(C->L);
          slang_function_destruct(&parsed_func);
          return GL_FALSE;
diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c
index 1fdf4db054c..2bc8809661d 100644
--- a/src/mesa/shader/slang/slang_link.c
+++ b/src/mesa/shader/slang/slang_link.c
@@ -97,7 +97,8 @@ bits_agree(GLbitfield flags1, GLbitfield flags2, GLbitfield bit)
  * which inputs are centroid-sampled, invariant, etc.
  */
 static GLboolean
-link_varying_vars(struct gl_shader_program *shProg, struct gl_program *prog)
+link_varying_vars(GLcontext *ctx,
+                  struct gl_shader_program *shProg, struct gl_program *prog)
 {
    GLuint *map, i, firstVarying, newFile;
    GLbitfield *inOutFlags;
@@ -156,8 +157,12 @@ link_varying_vars(struct gl_shader_program *shProg, struct gl_program *prog)
                                var->Flags);
       }
 
+      if (shProg->Varying->NumParameters > ctx->Const.MaxVarying) {
+         link_error(shProg, "Too many varying variables");
+         return GL_FALSE;
+      }
+
       /* Map varying[i] to varying[j].
-       * Plus, set prog->Input/OutputFlags[] as described above.
        * Note: the loop here takes care of arrays or large (sz>4) vars.
        */
       {
@@ -712,6 +717,8 @@ _slang_link(GLcontext *ctx,
       struct gl_vertex_program *linked_vprog =
          vertex_program(_mesa_clone_program(ctx, &vertProg->Base));
       shProg->VertexProgram = linked_vprog; /* refcount OK */
+      /* vertex program ID not significant; just set Id for debugging purposes */
+      shProg->VertexProgram->Base.Id = shProg->Name;
       ASSERT(shProg->VertexProgram->Base.RefCount == 1);
    }
 
@@ -720,16 +727,18 @@ _slang_link(GLcontext *ctx,
       struct gl_fragment_program *linked_fprog = 
          fragment_program(_mesa_clone_program(ctx, &fragProg->Base));
       shProg->FragmentProgram = linked_fprog; /* refcount OK */
+      /* vertex program ID not significant; just set Id for debugging purposes */
+      shProg->FragmentProgram->Base.Id = shProg->Name;
       ASSERT(shProg->FragmentProgram->Base.RefCount == 1);
    }
 
    /* link varying vars */
    if (shProg->VertexProgram) {
-      if (!link_varying_vars(shProg, &shProg->VertexProgram->Base))
+      if (!link_varying_vars(ctx, shProg, &shProg->VertexProgram->Base))
          return;
    }
    if (shProg->FragmentProgram) {
-      if (!link_varying_vars(shProg, &shProg->FragmentProgram->Base))
+      if (!link_varying_vars(ctx, shProg, &shProg->FragmentProgram->Base))
          return;
    }
 
diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c
index 77ecd0719e6..3ba7b269285 100644
--- a/src/mesa/state_tracker/st_atom_constbuf.c
+++ b/src/mesa/state_tracker/st_atom_constbuf.c
@@ -105,7 +105,7 @@ static void update_vs_constants(struct st_context *st )
 const struct st_tracked_state st_update_vs_constants = {
    "st_update_vs_constants",				/* name */
    {							/* dirty */
-      _NEW_PROGRAM_CONSTANTS,
+      (_NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS),          /* mesa */
       ST_NEW_VERTEX_PROGRAM,				/* st */
    },
    update_vs_constants					/* update */
diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c
index df0f0931eab..f23186c73de 100644
--- a/src/mesa/state_tracker/st_atom_framebuffer.c
+++ b/src/mesa/state_tracker/st_atom_framebuffer.c
@@ -123,6 +123,7 @@ update_framebuffer_state( struct st_context *st )
             framebuffer->cbufs[framebuffer->nr_cbufs] = strb->surface;
             framebuffer->nr_cbufs++;
          }
+         strb->defined = GL_TRUE; /* we'll be drawing something */
       }
    }
 
diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c
index 61687fbc3e2..4e70510c0c0 100644
--- a/src/mesa/state_tracker/st_atom_rasterizer.c
+++ b/src/mesa/state_tracker/st_atom_rasterizer.c
@@ -180,22 +180,7 @@ static void update_raster_state( struct st_context *st )
 
    if (ctx->Polygon.StippleFlag)
       raster->poly_stipple_enable = 1;
-
-
-   /* _NEW_BUFFERS, _NEW_POLYGON
-    */
-   if (raster->fill_cw != PIPE_POLYGON_MODE_FILL ||
-       raster->fill_ccw != PIPE_POLYGON_MODE_FILL)
-   {
-      GLfloat mrd = (ctx->DrawBuffer ? 
-		     ctx->DrawBuffer->_MRD : 
-		     1.0f);
-
-      raster->offset_units = ctx->Polygon.OffsetFactor * mrd;
-      raster->offset_scale = (ctx->Polygon.OffsetUnits * mrd *
-			    st->polygon_offset_scale);
-   }
-      
+     
    /* _NEW_POINT
     */
    raster->point_size = ctx->Point.Size;
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c
index 08dc7c930e2..89725cfe8dc 100644
--- a/src/mesa/state_tracker/st_cb_drawpixels.c
+++ b/src/mesa/state_tracker/st_cb_drawpixels.c
@@ -910,6 +910,34 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
 
    st_validate_state(st);
 
+   if (srcx < 0) {
+      width -= -srcx;
+      dstx += -srcx;
+      srcx = 0;
+   }
+
+   if (srcy < 0) {
+      height -= -srcy;
+      dsty += -srcy;
+      srcy = 0;
+   }
+
+   if (dstx < 0) {
+      width -= -dstx;
+      srcx += -dstx;
+      dstx = 0;
+   }
+
+   if (dsty < 0) {
+      height -= -dsty;
+      srcy += -dsty;
+      dsty = 0;
+   }
+
+   if (width < 0 || height < 0)
+      return;
+
+
    if (type == GL_STENCIL) {
       /* can't use texturing to do stencil */
       copy_stencil_pixels(ctx, srcx, srcy, width, height, dstx, dsty);
@@ -951,15 +979,24 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
       }
    }
 
+   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+      srcy = ctx->DrawBuffer->Height - srcy - height;
+
+      if (srcy < 0) {
+         height -= -srcy;
+         srcy = 0;
+      }
+      
+      if (height < 0)
+         return;
+   }
+
    pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, texFormat, 0,
                           width, height, 1,
                           PIPE_TEXTURE_USAGE_SAMPLER);
    if (!pt)
       return;
 
-   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
-      srcy = ctx->DrawBuffer->Height - srcy - height;
-   }
 
    if (srcFormat == texFormat) {
       /* copy source framebuffer surface into mipmap/texture */
diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c
index 1590f275e2a..c249f3b3578 100644
--- a/src/mesa/state_tracker/st_cb_fbo.c
+++ b/src/mesa/state_tracker/st_cb_fbo.c
@@ -125,6 +125,8 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
    strb->Base.Height = height;
    init_renderbuffer_bits(strb, template.format);
 
+   strb->defined = GL_FALSE;  /* undefined contents now */
+
    /* Probably need dedicated flags for surface usage too: 
     */
    surface_usage = (PIPE_BUFFER_USAGE_GPU_READ |
@@ -343,6 +345,7 @@ st_render_texture(GLcontext *ctx,
                   struct gl_framebuffer *fb,
                   struct gl_renderbuffer_attachment *att)
 {
+   struct pipe_screen *screen = ctx->st->pipe->screen;
    struct st_renderbuffer *strb;
    struct gl_renderbuffer *rb;
    struct pipe_texture *pt = st_get_texobj_texture(att->Texture);
@@ -365,6 +368,8 @@ st_render_texture(GLcontext *ctx,
    rb->AllocStorage = NULL; /* should not get called */
    strb = st_renderbuffer(rb);
 
+   assert(strb->Base.RefCount > 0);
+
    /* get the texture for the texture object */
    stObj = st_texture_object(att->Texture);
 
@@ -384,7 +389,14 @@ st_render_texture(GLcontext *ctx,
 
    pipe_surface_reference(&strb->surface, NULL);
 
-   /* the new surface will be created during framebuffer validation */
+   /* new surface for rendering into the texture */
+   strb->surface = screen->get_tex_surface(screen,
+                                           strb->texture,
+                                           strb->rtt_face,
+                                           strb->rtt_level,
+                                           strb->rtt_slice,
+                                           PIPE_BUFFER_USAGE_GPU_READ |
+                                           PIPE_BUFFER_USAGE_GPU_WRITE);
 
    init_renderbuffer_bits(strb, pt->format);
 
@@ -452,6 +464,134 @@ st_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb)
 }
 
 
+/**
+ * Copy back color buffer to front color buffer.
+ */
+static void
+copy_back_to_front(struct st_context *st,
+                   struct gl_framebuffer *fb,
+                   gl_buffer_index frontIndex,
+                   gl_buffer_index backIndex)
+
+{
+   struct st_framebuffer *stfb = (struct st_framebuffer *) fb;
+   struct pipe_surface *surf_front, *surf_back;
+
+   (void) st_get_framebuffer_surface(stfb, frontIndex, &surf_front);
+   (void) st_get_framebuffer_surface(stfb, backIndex, &surf_back);
+
+   if (surf_front && surf_back) {
+      st->pipe->surface_copy(st->pipe,
+                             surf_front, 0, 0,  /* dest */
+                             surf_back, 0, 0,   /* src */
+                             fb->Width, fb->Height);
+   }
+}
+
+
+/**
+ * Check if we're drawing into, or read from, a front color buffer.  If the
+ * front buffer is missing, create it now.
+ *
+ * The back color buffer must exist since we'll use its format/samples info
+ * for creating the front buffer.
+ *
+ * \param frontIndex  either BUFFER_FRONT_LEFT or BUFFER_FRONT_RIGHT
+ * \param backIndex  either BUFFER_BACK_LEFT or BUFFER_BACK_RIGHT
+ */
+static void
+check_create_front_buffer(GLcontext *ctx, struct gl_framebuffer *fb,
+                          gl_buffer_index frontIndex,
+                          gl_buffer_index backIndex)
+{
+   if (fb->Attachment[frontIndex].Renderbuffer == NULL) {
+      GLboolean create = GL_FALSE;
+
+      /* check if drawing to or reading from front buffer */
+      if (fb->_ColorReadBufferIndex == frontIndex) {
+         create = GL_TRUE;
+      }
+      else {
+         GLuint b;
+         for (b = 0; b < fb->_NumColorDrawBuffers; b++) {
+            if (fb->_ColorDrawBufferIndexes[b] == frontIndex) {
+               create = GL_TRUE;
+               break;
+            }
+         }
+      }
+
+      if (create) {
+         struct st_renderbuffer *back;
+         struct gl_renderbuffer *front;
+         enum pipe_format colorFormat;
+         uint samples;
+
+         if (0)
+            _mesa_debug(ctx, "Allocate new front buffer\n");
+
+         /* get back renderbuffer info */
+         back = st_renderbuffer(fb->Attachment[backIndex].Renderbuffer);
+         colorFormat = back->format;
+         samples = back->Base.NumSamples;
+
+         /* create front renderbuffer */
+         front = st_new_renderbuffer_fb(colorFormat, samples);
+         _mesa_add_renderbuffer(fb, frontIndex, front);
+
+         /* alloc texture/surface for new front buffer */
+         front->AllocStorage(ctx, front, front->InternalFormat,
+                             fb->Width, fb->Height);
+
+         /* initialize the front color buffer contents by copying
+          * the back buffer.
+          */
+         copy_back_to_front(ctx->st, fb, frontIndex, backIndex);
+      }
+   }
+}
+
+
+/**
+ * If front left/right color buffers are missing, create them now.
+ */
+static void
+check_create_front_buffers(GLcontext *ctx, struct gl_framebuffer *fb)
+{
+   /* check if we need to create the front left buffer now */
+   check_create_front_buffer(ctx, fb, BUFFER_FRONT_LEFT, BUFFER_BACK_LEFT);
+
+   if (fb->Visual.stereoMode) {
+      check_create_front_buffer(ctx, fb, BUFFER_FRONT_RIGHT, BUFFER_BACK_RIGHT);
+   }
+
+   st_invalidate_state(ctx, _NEW_BUFFERS);
+}
+
+
+/**
+ * Called via glDrawBuffer.
+ */
+static void
+st_DrawBuffers(GLcontext *ctx, GLsizei count, const GLenum *buffers)
+{
+   (void) count;
+   (void) buffers;
+   check_create_front_buffers(ctx, ctx->DrawBuffer);
+}
+
+
+/**
+ * Called via glReadBuffer.
+ */
+static void
+st_ReadBuffer(GLcontext *ctx, GLenum buffer)
+{
+   (void) buffer;
+   check_create_front_buffers(ctx, ctx->ReadBuffer);
+}
+
+
 void st_init_fbo_functions(struct dd_function_table *functions)
 {
    functions->NewFramebuffer = st_new_framebuffer;
@@ -464,4 +604,7 @@ void st_init_fbo_functions(struct dd_function_table *functions)
    /* no longer needed by core Mesa, drivers handle resizes...
    functions->ResizeBuffers = st_resize_buffers;
    */
+
+   functions->DrawBuffers = st_DrawBuffers;
+   functions->ReadBuffer = st_ReadBuffer;
 }
diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h
index 44fa9fe9a4f..fd77d0a95b0 100644
--- a/src/mesa/state_tracker/st_cb_fbo.h
+++ b/src/mesa/state_tracker/st_cb_fbo.h
@@ -44,6 +44,7 @@ struct st_renderbuffer
    struct pipe_texture *texture;
    struct pipe_surface *surface; /* temporary view into texture */
    enum pipe_format format;  /** preferred format, or PIPE_FORMAT_NONE */
+   GLboolean defined;        /**< defined contents? */
 
    struct st_texture_object *rtt;  /**< GL render to texture's texture */
    int rtt_level, rtt_face, rtt_slice;
diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c
index 7d7d3823c99..8ceeeabcd37 100644
--- a/src/mesa/state_tracker/st_cb_flush.c
+++ b/src/mesa/state_tracker/st_cb_flush.c
@@ -47,10 +47,19 @@
 #include "util/u_blit.h"
 
 
+/** Check if we have a front color buffer and if it's been drawn to. */
 static INLINE GLboolean
 is_front_buffer_dirty(struct st_context *st)
 {
-   return st->frontbuffer_status == FRONT_STATUS_DIRTY;
+   if (st->frontbuffer_status == FRONT_STATUS_DIRTY) {
+      return GL_TRUE;
+   }
+   else {
+      GLframebuffer *fb = st->ctx->DrawBuffer;
+      struct st_renderbuffer *strb
+         = st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+      return strb && strb->defined;
+   }
 }
 
 
@@ -82,7 +91,7 @@ display_front_buffer(struct st_context *st)
 void st_flush( struct st_context *st, uint pipeFlushFlags,
                struct pipe_fence_handle **fence )
 {
-   FLUSH_VERTICES(st->ctx, 0);
+   FLUSH_CURRENT(st->ctx, 0);
 
    /* Release any vertex buffers that might potentially be accessed in
     * successive frames:
diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c
index c3e990e0775..b182106fd56 100644
--- a/src/mesa/state_tracker/st_cb_texture.c
+++ b/src/mesa/state_tracker/st_cb_texture.c
@@ -51,6 +51,7 @@
 #include "state_tracker/st_texture.h"
 #include "state_tracker/st_gen_mipmap.h"
 #include "state_tracker/st_inlines.h"
+#include "state_tracker/st_atom.h"
 
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
@@ -418,7 +419,6 @@ compress_with_blit(GLcontext * ctx,
    const GLuint dstImageOffsets[1] = {0};
    struct st_texture_image *stImage = st_texture_image(texImage);
    struct pipe_screen *screen = ctx->st->pipe->screen;
-   const GLuint face = _mesa_tex_target_to_face(target);
    const struct gl_texture_format *mesa_format;
    struct pipe_texture templ;
    struct pipe_texture *src_tex;
@@ -467,7 +467,7 @@ compress_with_blit(GLcontext * ctx,
    /* Put user's tex data into the temporary texture
     */
    tex_xfer = st_cond_flush_get_tex_transfer(st_context(ctx), src_tex,
-					     face, level, 0,
+					     0, 0, 0, /* face, level are zero */
 					     PIPE_TRANSFER_WRITE,
 					     0, 0, width, height); /* x, y, w, h */
    map = screen->transfer_map(screen, tex_xfer);
@@ -1315,8 +1315,9 @@ st_copy_texsubimage(GLcontext *ctx,
    GLboolean use_fallback = GL_TRUE;
    GLboolean matching_base_formats;
 
-   /* any rendering in progress must complete before we grab the fb image */
-   st_finish(ctx->st);
+   /* make sure finalize_textures has been called? 
+    */
+   if (0) st_validate_state(ctx->st);
 
    /* determine if copying depth or color data */
    if (texBaseFormat == GL_DEPTH_COMPONENT ||
@@ -1331,6 +1332,39 @@ st_copy_texsubimage(GLcontext *ctx,
       strb = st_renderbuffer(fb->_ColorReadBuffer);
    }
 
+   if (!strb || !strb->surface || !stImage->pt) {
+      debug_printf("%s: null strb or stImage\n", __FUNCTION__);
+      return;
+   }
+
+   if (srcX < 0) {
+      width -= -srcX;
+      destX += -srcX;
+      srcX = 0;
+   }
+
+   if (srcY < 0) {
+      height -= -srcY;
+      destY += -srcY;
+      srcY = 0;
+   }
+
+   if (destX < 0) {
+      width -= -destX;
+      srcX += -destX;
+      destX = 0;
+   }
+
+   if (destY < 0) {
+      height -= -destY;
+      srcY += -destY;
+      destY = 0;
+   }
+
+   if (width < 0 || height < 0)
+      return;
+
+
    assert(strb);
    assert(strb->surface);
    assert(stImage->pt);
diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c
index b27274725fc..2a1f21c51ca 100644
--- a/src/mesa/state_tracker/st_context.c
+++ b/src/mesa/state_tracker/st_context.c
@@ -177,6 +177,12 @@ struct st_context *st_create_context(struct pipe_context *pipe,
 
    ctx = _mesa_create_context(visual, shareCtx, &funcs, NULL);
 
+   /* XXX: need a capability bit in gallium to query if the pipe
+    * driver prefers DP4 or MUL/MAD for vertex transformation.
+    */
+   if (debug_get_bool_option("MESA_MVP_DP4", FALSE))
+      _mesa_set_mvp_with_dp4( ctx, GL_TRUE );
+
    return st_create_context_priv(ctx, pipe);
 }
 
@@ -279,6 +285,12 @@ void st_make_current(struct st_context *st,
    }
 }
 
+struct st_context *st_get_current(void)
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   return (ctx == NULL) ? NULL : ctx->st;
+}
 
 void st_copy_context_state(struct st_context *dst,
                            struct st_context *src,
diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h
index f840579a404..18adb35e872 100644
--- a/src/mesa/state_tracker/st_context.h
+++ b/src/mesa/state_tracker/st_context.h
@@ -45,6 +45,7 @@ struct blit_state;
 struct bitmap_cache;
 
 
+/** XXX we'd like to get rid of these */
 #define FRONT_STATUS_UNDEFINED    0
 #define FRONT_STATUS_DIRTY        1
 #define FRONT_STATUS_COPY_OF_BACK 2
@@ -111,7 +112,7 @@ struct st_context
       struct gl_fragment_program *fragment_program;
    } cb;
 
-   GLuint frontbuffer_status;  /**< one of FRONT_STATUS_ */
+   GLuint frontbuffer_status;  /**< one of FRONT_STATUS_ (XXX to be removed) */
 
    char vendor[100];
    char renderer[100];
@@ -120,8 +121,6 @@ struct st_context
 
    GLboolean missing_textures;
 
-   GLfloat polygon_offset_scale; /* ?? */
-
    /** Mapping from VERT_RESULT_x to post-transformed vertex slot */
    const GLuint *vertex_result_to_slot;
 
diff --git a/src/mesa/state_tracker/st_draw_feedback.c b/src/mesa/state_tracker/st_draw_feedback.c
index e533afd051e..32502a9cda4 100644
--- a/src/mesa/state_tracker/st_draw_feedback.c
+++ b/src/mesa/state_tracker/st_draw_feedback.c
@@ -196,13 +196,10 @@ st_feedback_draw_vbo(GLcontext *ctx,
    draw_set_vertex_elements(draw, vp->num_inputs, velements);
 
    if (ib) {
-      unsigned indexSize;
       struct gl_buffer_object *bufobj = ib->obj;
-      struct st_buffer_object *stobj = st_buffer_object(bufobj);
+      unsigned indexSize;
       void *map;
 
-      index_buffer_handle = stobj->buffer;
-
       switch (ib->type) {
       case GL_UNSIGNED_INT:
          indexSize = 4;
@@ -215,9 +212,19 @@ st_feedback_draw_vbo(GLcontext *ctx,
 	 return;
       }
 
-      map = pipe_buffer_map(pipe->screen, index_buffer_handle,
-                            PIPE_BUFFER_USAGE_CPU_READ);
-      draw_set_mapped_element_buffer(draw, indexSize, map);
+      if (bufobj && bufobj->Name) {
+         struct st_buffer_object *stobj = st_buffer_object(bufobj);
+
+         index_buffer_handle = stobj->buffer;
+
+         map = pipe_buffer_map(pipe->screen, index_buffer_handle,
+                               PIPE_BUFFER_USAGE_CPU_READ);
+
+         draw_set_mapped_element_buffer(draw, indexSize, map);
+      }
+      else {
+         draw_set_mapped_element_buffer(draw, indexSize, (void *) ib->ptr);
+      }
    }
    else {
       /* no index/element buffer */
@@ -252,7 +259,7 @@ st_feedback_draw_vbo(GLcontext *ctx,
          draw_set_mapped_vertex_buffer(draw, i, NULL);
       }
    }
-   if (ib) {
+   if (index_buffer_handle) {
       pipe_buffer_unmap(pipe->screen, index_buffer_handle);
       draw_set_mapped_element_buffer(draw, 0, NULL);
    }
diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c
index daaad65ccaf..ef800291ccd 100644
--- a/src/mesa/state_tracker/st_framebuffer.c
+++ b/src/mesa/state_tracker/st_framebuffer.c
@@ -58,20 +58,20 @@ st_create_framebuffer( const __GLcontextModes *visual,
 
       _mesa_initialize_framebuffer(&stfb->Base, visual);
 
-      {
-         /* fake frontbuffer */
-         /* XXX allocation should only happen in the unusual case
-            it's actually needed */
-         struct gl_renderbuffer *rb
-            = st_new_renderbuffer_fb(colorFormat, samples);
-         _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb);
-      }
-
       if (visual->doubleBufferMode) {
          struct gl_renderbuffer *rb
             = st_new_renderbuffer_fb(colorFormat, samples);
          _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb);
       }
+      else {
+         /* Only allocate front buffer right now if we're single buffered.
+          * If double-buffered, allocate front buffer on demand later.
+          * See check_create_front_buffers().
+          */
+         struct gl_renderbuffer *rb
+            = st_new_renderbuffer_fb(colorFormat, samples);
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb);
+      }
 
       if (depthFormat == stencilFormat && depthFormat != PIPE_FORMAT_NONE) {
          /* combined depth/stencil buffer */
@@ -293,6 +293,115 @@ st_notify_swapbuffers(struct st_framebuffer *stfb)
 }
 
 
+/**
+ * Swap the front/back color buffers.  Exchange the front/back pointers
+ * and update some derived state.
+ * No need to call st_notify_swapbuffers() first.
+ *
+ * For a single-buffered framebuffer, no swap occurs, but we still return
+ * the pointer(s) to the front color buffer(s).
+ *
+ * \param front_left  returns pointer to front-left renderbuffer after swap
+ * \param front_right  returns pointer to front-right renderbuffer after swap
+ */
+void
+st_swapbuffers(struct st_framebuffer *stfb,
+               struct pipe_surface **front_left,
+               struct pipe_surface **front_right)
+{
+   struct gl_framebuffer *fb = &stfb->Base;
+
+   GET_CURRENT_CONTEXT(ctx);
+
+   if (ctx && ctx->DrawBuffer == &stfb->Base) {
+      st_flush( ctx->st, 
+		PIPE_FLUSH_RENDER_CACHE | 
+		PIPE_FLUSH_SWAPBUFFERS |
+		PIPE_FLUSH_FRAME,
+                NULL );
+   }
+
+   if (!fb->Visual.doubleBufferMode) {
+      /* single buffer mode - return pointers to front surfaces */
+      if (front_left) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+         *front_left = strb->surface;
+      }
+      if (front_right) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer);
+         *front_right = strb ? strb->surface : NULL;
+      }
+      return;
+   }
+
+   /* swap left buffers */
+   if (fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer &&
+       fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer) {
+      struct gl_renderbuffer *rbTemp;
+      rbTemp = fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer;
+      fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer =
+         fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+      fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer = rbTemp;
+      if (front_left) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+         *front_left = strb->surface;
+      }
+      /* mark back buffer contents as undefined */
+      {
+         struct st_renderbuffer *back =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+         back->defined = GL_FALSE;
+      }
+   }
+   else {
+      /* no front buffer, display the back buffer */
+      if (front_left) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+         *front_left = strb->surface;
+      }
+   }
+
+   /* swap right buffers (for stereo) */
+   if (fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer &&
+       fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer) {
+      struct gl_renderbuffer *rbTemp;
+      rbTemp = fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer;
+      fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer =
+         fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer;
+      fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer = rbTemp;
+      if (front_right) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer);
+         *front_right = strb->surface;
+      }
+      /* mark back buffer contents as undefined */
+      {
+         struct st_renderbuffer *back =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer);
+         back->defined = GL_FALSE;
+      }
+   }
+   else {
+      /* no front right buffer, display back right buffer (if exists) */
+      if (front_right) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer);
+         *front_right = strb ? strb->surface : NULL;
+      }
+   }
+
+   /* Update the _ColorDrawBuffers[] array and _ColorReadBuffer pointer */
+   _mesa_update_framebuffer(ctx);
+
+   /* Make sure we draw into the new back surface */
+   st_invalidate_state(ctx, _NEW_BUFFERS);
+}
+
+
 void *st_framebuffer_private( struct st_framebuffer *stfb )
 {
    return stfb->Private;
diff --git a/src/mesa/state_tracker/st_inlines.h b/src/mesa/state_tracker/st_inlines.h
index 0322d5dfa6e..a41cfeb96f7 100644
--- a/src/mesa/state_tracker/st_inlines.h
+++ b/src/mesa/state_tracker/st_inlines.h
@@ -1,3 +1,35 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Functions for checking if buffers/textures are referenced when we need
+ * to read/write from/to them.  Flush when needed.
+ */
+
 #ifndef ST_INLINES_H
 #define ST_INLINES_H
 
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
index 2795570cf10..6ec633c0b46 100644
--- a/src/mesa/state_tracker/st_program.c
+++ b/src/mesa/state_tracker/st_program.c
@@ -169,6 +169,14 @@ st_translate_vertex_program(struct st_context *st,
          case VERT_ATTRIB_GENERIC5:
          case VERT_ATTRIB_GENERIC6:
          case VERT_ATTRIB_GENERIC7:
+         case VERT_ATTRIB_GENERIC8:
+         case VERT_ATTRIB_GENERIC9:
+         case VERT_ATTRIB_GENERIC10:
+         case VERT_ATTRIB_GENERIC11:
+         case VERT_ATTRIB_GENERIC12:
+         case VERT_ATTRIB_GENERIC13:
+         case VERT_ATTRIB_GENERIC14:
+         case VERT_ATTRIB_GENERIC15:
             assert(attr < VERT_ATTRIB_MAX);
             vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
             vs_input_semantic_index[slot] = num_generic++;
diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h
index 030314372f9..174fbc63941 100644
--- a/src/mesa/state_tracker/st_public.h
+++ b/src/mesa/state_tracker/st_public.h
@@ -95,12 +95,18 @@ void st_make_current(struct st_context *st,
                      struct st_framebuffer *draw,
                      struct st_framebuffer *read);
 
+struct st_context *st_get_current(void);
+
 void st_flush( struct st_context *st, uint pipeFlushFlags,
                struct pipe_fence_handle **fence );
 void st_finish( struct st_context *st );
 
 void st_notify_swapbuffers(struct st_framebuffer *stfb);
 
+void st_swapbuffers(struct st_framebuffer *stfb,
+                    struct pipe_surface **front_left,
+                    struct pipe_surface **front_right);
+
 int st_set_teximage(struct pipe_texture *pt, int target);
 
 /** Redirect rendering into stfb's surface to a texture image */
diff --git a/src/mesa/swrast/s_imaging.c b/src/mesa/swrast/s_imaging.c
index d6be3aa022e..3578b713f61 100644
--- a/src/mesa/swrast/s_imaging.c
+++ b/src/mesa/swrast/s_imaging.c
@@ -60,7 +60,7 @@ _swrast_CopyColorTable( GLcontext *ctx,
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
 
    _mesa_ColorTable(target, internalformat, width, GL_RGBA, CHAN_TYPE, data);
 
@@ -94,7 +94,7 @@ _swrast_CopyColorSubTable( GLcontext *ctx,GLenum target, GLsizei start,
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
 
    _mesa_ColorSubTable(target, start, width, GL_RGBA, CHAN_TYPE, data);
 
@@ -126,7 +126,7 @@ _swrast_CopyConvolutionFilter1D(GLcontext *ctx, GLenum target,
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
 
    /* store as convolution filter */
    _mesa_ConvolutionFilter1D(target, internalFormat, width,
@@ -178,12 +178,12 @@ _swrast_CopyConvolutionFilter2D(GLcontext *ctx, GLenum target,
    ctx->Unpack.SkipImages = 0;
    ctx->Unpack.SwapBytes = GL_FALSE;
    ctx->Unpack.LsbFirst = GL_FALSE;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
    ctx->NewState |= _NEW_PACKUNPACK;
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
 
    _mesa_ConvolutionFilter2D(target, internalFormat, width, height,
                              GL_RGBA, CHAN_TYPE, rgba);
diff --git a/src/mesa/swrast/s_span.c b/src/mesa/swrast/s_span.c
index cfff82b0513..fa8ca1d0e26 100644
--- a/src/mesa/swrast/s_span.c
+++ b/src/mesa/swrast/s_span.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.5
  *
- * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -188,10 +189,10 @@ interpolate_active_attribs(GLcontext *ctx, SWspan *span, GLbitfield attrMask)
          const GLfloat dv1dx = span->attrStepX[attr][1];
          const GLfloat dv2dx = span->attrStepX[attr][2];
          const GLfloat dv3dx = span->attrStepX[attr][3];
-         GLfloat v0 = span->attrStart[attr][0];
-         GLfloat v1 = span->attrStart[attr][1];
-         GLfloat v2 = span->attrStart[attr][2];
-         GLfloat v3 = span->attrStart[attr][3];
+         GLfloat v0 = span->attrStart[attr][0] + span->leftClip * dv0dx;
+         GLfloat v1 = span->attrStart[attr][1] + span->leftClip * dv1dx;
+         GLfloat v2 = span->attrStart[attr][2] + span->leftClip * dv2dx;
+         GLfloat v3 = span->attrStart[attr][3] + span->leftClip * dv3dx;
          GLuint k;
          for (k = 0; k < span->end; k++) {
             const GLfloat invW = 1.0f / w;
@@ -521,10 +522,10 @@ interpolate_texcoords(GLcontext *ctx, SWspan *span)
          const GLfloat drdx = span->attrStepX[attr][2];
          const GLfloat dqdx = span->attrStepX[attr][3];
          const GLfloat dqdy = span->attrStepY[attr][3];
-         GLfloat s = span->attrStart[attr][0];
-         GLfloat t = span->attrStart[attr][1];
-         GLfloat r = span->attrStart[attr][2];
-         GLfloat q = span->attrStart[attr][3];
+         GLfloat s = span->attrStart[attr][0] + span->leftClip * dsdx;
+         GLfloat t = span->attrStart[attr][1] + span->leftClip * dtdx;
+         GLfloat r = span->attrStart[attr][2] + span->leftClip * drdx;
+         GLfloat q = span->attrStart[attr][3] + span->leftClip * dqdx;
 
          if (obj) {
             const struct gl_texture_image *img = obj->Image[0][obj->BaseLevel];
@@ -546,7 +547,7 @@ interpolate_texcoords(GLcontext *ctx, SWspan *span)
                 || ctx->ATIFragmentShader._Enabled) {
                /* do perspective correction but don't divide s, t, r by q */
                const GLfloat dwdx = span->attrStepX[FRAG_ATTRIB_WPOS][3];
-               GLfloat w = span->attrStart[FRAG_ATTRIB_WPOS][3];
+               GLfloat w = span->attrStart[FRAG_ATTRIB_WPOS][3] + span->leftClip * dwdx;
                for (i = 0; i < span->end; i++) {
                   const GLfloat invW = 1.0F / w;
                   texcoord[i][0] = s * invW;
@@ -587,7 +588,7 @@ interpolate_texcoords(GLcontext *ctx, SWspan *span)
                 ctx->ATIFragmentShader._Enabled) {
                /* do perspective correction but don't divide s, t, r by q */
                const GLfloat dwdx = span->attrStepX[FRAG_ATTRIB_WPOS][3];
-               GLfloat w = span->attrStart[FRAG_ATTRIB_WPOS][3];
+               GLfloat w = span->attrStart[FRAG_ATTRIB_WPOS][3] + span->leftClip * dwdx;
                for (i = 0; i < span->end; i++) {
                   const GLfloat invW = 1.0F / w;
                   texcoord[i][0] = s * invW;
@@ -660,8 +661,8 @@ interpolate_wpos(GLcontext *ctx, SWspan *span)
       }
    }
 
-   w = span->attrStart[FRAG_ATTRIB_WPOS][3];
    dw = span->attrStepX[FRAG_ATTRIB_WPOS][3];
+   w = span->attrStart[FRAG_ATTRIB_WPOS][3] + span->leftClip * dw;
    for (i = 0; i < span->end; i++) {
       wpos[i][2] = (GLfloat) span->array->z[i] * zScale;
       wpos[i][3] = w;
@@ -726,6 +727,8 @@ clip_span( GLcontext *ctx, SWspan *span )
    const GLint ymin = ctx->DrawBuffer->_Ymin;
    const GLint ymax = ctx->DrawBuffer->_Ymax;
 
+   span->leftClip = 0;
+
    if (span->arrayMask & SPAN_XY) {
       /* arrays of x/y pixel coords */
       const GLint *x = span->array->x;
@@ -753,7 +756,7 @@ clip_span( GLcontext *ctx, SWspan *span )
       /* horizontal span of pixels */
       const GLint x = span->x;
       const GLint y = span->y;
-      const GLint n = span->end;
+      GLint n = span->end;
 
       /* Trivial rejection tests */
       if (y < ymin || y >= ymax || x + n <= xmin || x >= xmax) {
@@ -761,19 +764,45 @@ clip_span( GLcontext *ctx, SWspan *span )
          return GL_FALSE;  /* all pixels clipped */
       }
 
-      /* Clip to the left */
-      if (x < xmin) {
-         ASSERT(x + n > xmin);
-         span->writeAll = GL_FALSE;
-         _mesa_bzero(span->array->mask, (xmin - x) * sizeof(GLubyte));
-      }
-
       /* Clip to right */
       if (x + n > xmax) {
          ASSERT(x < xmax);
-         span->end = xmax - x;
+         n = span->end = xmax - x;
       }
 
+      /* Clip to the left */
+      if (x < xmin) {
+         const GLint leftClip = xmin - x;
+         GLuint i;
+
+         ASSERT(leftClip > 0);
+         ASSERT(x + n > xmin);
+
+         /* Clip 'leftClip' pixels from the left side.
+          * The span->leftClip field will be applied when we interpolate
+          * fragment attributes.
+          * For arrays of values, shift them left.
+          */
+         for (i = 0; i < FRAG_ATTRIB_MAX; i++) {
+            if (span->arrayAttribs & (1 << i)) {
+               /* shift array elements left by 'leftClip' */
+               _mesa_memcpy(span->array->attribs[i],
+                            span->array->attribs[i] + leftClip,
+                            (n - leftClip) * 4 * sizeof(GLfloat));
+            }
+         }
+
+         span->leftClip = leftClip;
+         span->x = xmin;
+         span->end -= leftClip;
+         span->writeAll = GL_FALSE;
+      }
+
+      ASSERT(span->x >= xmin);
+      ASSERT(span->x + span->end <= xmax);
+      ASSERT(span->y >= ymin);
+      ASSERT(span->y < ymax);
+
       return GL_TRUE;  /* some pixels visible */
    }
 }
@@ -818,6 +847,12 @@ _swrast_write_index_span( GLcontext *ctx, SWspan *span)
       }
    }
 
+   if (!(span->arrayMask & SPAN_MASK)) {
+      /* post-clip sanity check */
+      assert(span->x >= 0);
+      assert(span->y >= 0);
+   }
+
    /* Depth bounds test */
    if (ctx->Depth.BoundsTest && fb->Visual.depthBits > 0) {
       if (!_swrast_depth_bounds_test(ctx, span)) {
@@ -1284,6 +1319,7 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
 #ifdef DEBUG
    /* Make sure all fragments are within window bounds */
    if (span->arrayMask & SPAN_XY) {
+      /* array of pixel locations */
       GLuint i;
       for (i = 0; i < span->end; i++) {
          if (span->array->mask[i]) {
@@ -1321,6 +1357,14 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
       if (!(span->arrayMask & SPAN_Z))
          _swrast_span_interpolate_z(ctx, span);
 
+      if ((span->arrayMask & SPAN_XY) == 0) {
+         if (span->x < fb->_Xmin || span->x + span->end > fb->_Xmax ||
+             span->y < fb->_Ymin || span->y >= fb->_Ymax) {
+            printf("Bad span clipping at %d, %d\n", span->x, span->y);
+            return;
+         }
+      }
+
       if (ctx->Stencil._Enabled) {
          /* Combined Z/stencil tests */
          if (!_swrast_stencil_and_ztest_span(ctx, span)) {
diff --git a/src/mesa/swrast/s_span.h b/src/mesa/swrast/s_span.h
index c4b47df58ff..0eabae20e03 100644
--- a/src/mesa/swrast/s_span.h
+++ b/src/mesa/swrast/s_span.h
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5
+ * Version:  7.5
  *
- * Copyright (C) 1999-2005  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -106,6 +107,9 @@ typedef struct sw_span
    /** Number of fragments in the span */
    GLuint end;
 
+   /** for clipping left edge of spans */
+   GLuint leftClip;
+
    /** This flag indicates that mask[] array is effectively filled with ones */
    GLboolean writeAll;
 
@@ -165,6 +169,7 @@ do {						\
    (S).arrayMask = 0x0;				\
    (S).arrayAttribs = 0x0;			\
    (S).end = 0;					\
+   (S).leftClip = 0;				\
    (S).facing = 0;				\
    (S).array = SWRAST_CONTEXT(ctx)->SpanArrays;	\
 } while (0)
diff --git a/src/mesa/swrast/s_texfilter.c b/src/mesa/swrast/s_texfilter.c
index a483023a503..31bfb5c9520 100644
--- a/src/mesa/swrast/s_texfilter.c
+++ b/src/mesa/swrast/s_texfilter.c
@@ -462,6 +462,7 @@ clamp_rect_coord_linear(GLenum wrapMode, GLfloat coord, GLint max,
       fcol -= 0.5F;
       i0 = IFLOOR(fcol);
       i1 = i0 + 1;
+      break;
    default:
       _mesa_problem(NULL, "bad wrapMode in clamp_rect_coord_linear");
       i0 = i1 = 0;
diff --git a/src/mesa/tnl/t_vb_cliptmp.h b/src/mesa/tnl/t_vb_cliptmp.h
index 788fe329ed8..61b0a89554c 100644
--- a/src/mesa/tnl/t_vb_cliptmp.h
+++ b/src/mesa/tnl/t_vb_cliptmp.h
@@ -127,7 +127,7 @@ TAG(clip_line)( GLcontext *ctx, GLuint v0, GLuint v1, GLubyte mask )
    GLuint p;
    const GLuint v0_orig = v0;
 
-   if (mask & 0x3f) {
+   if (mask & CLIP_FRUSTUM_BITS) {
       LINE_CLIP( CLIP_RIGHT_BIT,  -1,  0,  0, 1 );
       LINE_CLIP( CLIP_LEFT_BIT,    1,  0,  0, 1 );
       LINE_CLIP( CLIP_TOP_BIT,     0, -1,  0, 1 );
@@ -199,7 +199,7 @@ TAG(clip_tri)( GLcontext *ctx, GLuint v0, GLuint v1, GLuint v2, GLubyte mask )
 
    ASSIGN_3V(inlist, v2, v0, v1 ); /* pv rotated to slot zero */
 
-   if (mask & 0x3f) {
+   if (mask & CLIP_FRUSTUM_BITS) {
       POLY_CLIP( CLIP_RIGHT_BIT,  -1,  0,  0, 1 );
       POLY_CLIP( CLIP_LEFT_BIT,    1,  0,  0, 1 );
       POLY_CLIP( CLIP_TOP_BIT,     0, -1,  0, 1 );
@@ -227,6 +227,25 @@ TAG(clip_tri)( GLcontext *ctx, GLuint v0, GLuint v1, GLuint v2, GLubyte mask )
       }
    }
 
+   if (0) {
+      /* print pre/post-clip vertex coords */
+      GLuint i, j;
+      _mesa_printf("pre clip\n");
+      for (i = 0; i < 3; i++) {
+         j = outlist[i];
+         _mesa_printf("  %u: %u: %f, %f, %f, %f\n",
+                      i, j,
+                      coord[j][0], coord[j][1], coord[j][2], coord[j][3]);
+      }
+      _mesa_printf("post clip\n");
+      for (i = 0; i < n; i++) {
+         j = inlist[i];
+         _mesa_printf("  %u: %u: %f, %f, %f, %f\n",
+                      i, j,
+                      coord[j][0], coord[j][1], coord[j][2], coord[j][3]);
+      }
+   }
+
    tnl->Driver.Render.ClippedPolygon( ctx, inlist, n );
 }
 
@@ -250,7 +269,7 @@ TAG(clip_quad)( GLcontext *ctx, GLuint v0, GLuint v1, GLuint v2, GLuint v3,
 
    ASSIGN_4V(inlist, v3, v0, v1, v2 ); /* pv rotated to slot zero */
 
-   if (mask & 0x3f) {
+   if (mask & CLIP_FRUSTUM_BITS) {
       POLY_CLIP( CLIP_RIGHT_BIT,  -1,  0,  0, 1 );
       POLY_CLIP( CLIP_LEFT_BIT,    1,  0,  0, 1 );
       POLY_CLIP( CLIP_TOP_BIT,     0, -1,  0, 1 );
diff --git a/src/mesa/vbo/vbo_context.c b/src/mesa/vbo/vbo_context.c
index ca8190fd059..f193a4bf1e0 100644
--- a/src/mesa/vbo/vbo_context.c
+++ b/src/mesa/vbo/vbo_context.c
@@ -28,6 +28,7 @@
 #include "main/imports.h"
 #include "main/mtypes.h"
 #include "main/api_arrayelt.h"
+#include "main/bufferobj.h"
 #include "math/m_eval.h"
 #include "vbo.h"
 #include "vbo_context.h"
@@ -81,7 +82,8 @@ static void init_legacy_currval(GLcontext *ctx)
       cl->Type = GL_FLOAT;
       cl->Format = GL_RGBA;
       cl->Ptr = (const void *)ctx->Current.Attrib[i];
-      cl->BufferObj = ctx->Array.NullBufferObj;
+      _mesa_reference_buffer_object(ctx, &cl->BufferObj,
+                                    ctx->Shared->NullBufferObj);
    }
 }
 
@@ -106,7 +108,8 @@ static void init_generic_currval(GLcontext *ctx)
       cl->Stride = 0;
       cl->StrideB = 0;
       cl->Enabled = 1;
-      cl->BufferObj = ctx->Array.NullBufferObj;
+      _mesa_reference_buffer_object(ctx, &cl->BufferObj,
+                                    ctx->Shared->NullBufferObj);
    }
 }
 
@@ -150,7 +153,7 @@ static void init_mat_currval(GLcontext *ctx)
       cl->Stride = 0;
       cl->StrideB = 0;
       cl->Enabled = 1;
-      cl->BufferObj = ctx->Array.NullBufferObj;
+      cl->BufferObj = ctx->Shared->NullBufferObj;
    }
 }
 
diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c
index 5d35ec9c111..6871ee5cab1 100644
--- a/src/mesa/vbo/vbo_exec_api.c
+++ b/src/mesa/vbo/vbo_exec_api.c
@@ -671,7 +671,7 @@ void vbo_use_buffer_objects(GLcontext *ctx)
    GLsizei size = VBO_VERT_BUFFER_SIZE;
 
    /* Make sure this func is only used once */
-   assert(exec->vtx.bufferobj == ctx->Array.NullBufferObj);
+   assert(exec->vtx.bufferobj == ctx->Shared->NullBufferObj);
    if (exec->vtx.buffer_map) {
       _mesa_align_free(exec->vtx.buffer_map);
       exec->vtx.buffer_map = NULL;
@@ -697,7 +697,7 @@ void vbo_exec_vtx_init( struct vbo_exec_context *exec )
     */
    _mesa_reference_buffer_object(ctx,
                                  &exec->vtx.bufferobj,
-                                 ctx->Array.NullBufferObj);
+                                 ctx->Shared->NullBufferObj);
 
    ASSERT(!exec->vtx.buffer_map);
    exec->vtx.buffer_map = (GLfloat *)ALIGN_MALLOC(VBO_VERT_BUFFER_SIZE, 64);
diff --git a/src/mesa/vbo/vbo_exec_array.c b/src/mesa/vbo/vbo_exec_array.c
index 0d4cbe9a1e5..f4ad394f516 100644
--- a/src/mesa/vbo/vbo_exec_array.c
+++ b/src/mesa/vbo/vbo_exec_array.c
@@ -352,6 +352,13 @@ vbo_exec_DrawRangeElements(GLenum mode,
    if (!_mesa_validate_DrawRangeElements( ctx, mode, start, end, count, type, indices ))
       return;
 
+   if (end >= ctx->Array._MaxElement) {
+      /* the max element is out of bounds of one or more enabled arrays */
+      _mesa_warning(ctx, "glDraw[Range]Elements() index=%u is "
+                    "out of bounds (max=%u)", end, ctx->Array._MaxElement);
+      return;
+   }
+
    FLUSH_CURRENT( ctx, 0 );
 
    if (ctx->NewState)
diff --git a/src/mesa/vbo/vbo_rebase.c b/src/mesa/vbo/vbo_rebase.c
index dae778e741e..ea87dede646 100644
--- a/src/mesa/vbo/vbo_rebase.c
+++ b/src/mesa/vbo/vbo_rebase.c
@@ -161,7 +161,7 @@ void vbo_rebase_prims( GLcontext *ctx,
 				 GL_ELEMENT_ARRAY_BUFFER,
 				 ib->obj);
 
-      tmp_ib.obj = ctx->Array.NullBufferObj;
+      tmp_ib.obj = ctx->Shared->NullBufferObj;
       tmp_ib.ptr = tmp_indices;
       tmp_ib.count = ib->count;
       tmp_ib.type = ib->type;
diff --git a/src/mesa/vbo/vbo_split_copy.c b/src/mesa/vbo/vbo_split_copy.c
index 5fb66d3318f..2f6a1998eaa 100644
--- a/src/mesa/vbo/vbo_split_copy.c
+++ b/src/mesa/vbo/vbo_split_copy.c
@@ -31,6 +31,7 @@
 
 #include "main/glheader.h"
 #include "main/imports.h"
+#include "main/image.h"
 #include "main/macros.h"
 #include "main/enums.h"
 #include "main/mtypes.h"
@@ -41,7 +42,8 @@
 
 #define ELT_TABLE_SIZE 16
 
-/* Used for vertex-level splitting of indexed buffers.  Note that
+/**
+ * Used for vertex-level splitting of indexed buffers.  Note that
  * non-indexed primitives may be converted to indexed in some cases
  * (eg loops, fans) in order to use this splitting path.
  */
@@ -73,23 +75,21 @@ struct copy_context {
    GLuint *translated_elt_buf;
    const GLuint *srcelt;
 
-   /* A baby hash table to avoid re-emitting (some) duplicate
+   /** A baby hash table to avoid re-emitting (some) duplicate
     * vertices when splitting indexed primitives.
     */
    struct { 
       GLuint in;
       GLuint out;
    } vert_cache[ELT_TABLE_SIZE];
-      
 
    GLuint vertex_size;
    GLubyte *dstbuf;
-   GLubyte *dstptr;		/* dstptr == dstbuf + dstelt_max * vertsize */
-   GLuint dstbuf_size;	/* in vertices */
-   GLuint dstbuf_nr;		/* count of emitted vertices, also the
-				 * largest value in dstelt.  Our
-				 * MaxIndex.
-				 */
+   GLubyte *dstptr;     /**< dstptr == dstbuf + dstelt_max * vertsize */
+   GLuint dstbuf_size;  /**< in vertices */
+   GLuint dstbuf_nr;    /**< count of emitted vertices, also the largest value
+                         * in dstelt.  Our MaxIndex.
+                         */
 
    GLuint *dstelt;
    GLuint dstelt_nr;
@@ -102,32 +102,19 @@ struct copy_context {
 };
 
 
-static GLuint type_size( GLenum type )
-{
-   switch(type) {
-   case GL_BYTE: return sizeof(GLbyte);
-   case GL_UNSIGNED_BYTE: return sizeof(GLubyte);
-   case GL_SHORT: return sizeof(GLshort);
-   case GL_UNSIGNED_SHORT: return sizeof(GLushort);
-   case GL_INT: return sizeof(GLint);
-   case GL_UNSIGNED_INT: return sizeof(GLuint);
-   case GL_FLOAT: return sizeof(GLfloat);
-   case GL_DOUBLE: return sizeof(GLdouble);
-   default: return 0;
-   }
-}
-
 static GLuint attr_size( const struct gl_client_array *array )
 {
-   return array->Size * type_size(array->Type);
+   return array->Size * _mesa_sizeof_type(array->Type);
 }
 
 
-/* Starts returning true slightly before the buffer fills, to ensure
+/**
+ * Starts returning true slightly before the buffer fills, to ensure
  * that there is sufficient room for any remaining vertices to finish
  * off the prim:
  */
-static GLboolean check_flush( struct copy_context *copy )
+static GLboolean
+check_flush( struct copy_context *copy )
 {
    GLenum mode = copy->dstprim[copy->dstprim_nr].mode; 
 
@@ -145,7 +132,9 @@ static GLboolean check_flush( struct copy_context *copy )
    return GL_FALSE;
 }
 
-static void flush( struct copy_context *copy )
+
+static void
+flush( struct copy_context *copy )
 {
    GLuint i;
 
@@ -175,8 +164,11 @@ static void flush( struct copy_context *copy )
 }
 
 
-
-static void begin( struct copy_context *copy, GLenum mode, GLboolean begin_flag )
+/**
+ * Called at begin of each primitive during replay.
+ */
+static void
+begin( struct copy_context *copy, GLenum mode, GLboolean begin_flag )
 {
    struct _mesa_prim *prim = &copy->dstprim[copy->dstprim_nr];
 
@@ -187,10 +179,12 @@ static void begin( struct copy_context *copy, GLenum mode, GLboolean begin_flag
 }
 
 
-/* Use a hashtable to attempt to identify recently-emitted vertices
+/**
+ * Use a hashtable to attempt to identify recently-emitted vertices
  * and avoid re-emitting them.
  */
-static GLuint elt(struct copy_context *copy, GLuint elt_idx)
+static GLuint
+elt(struct copy_context *copy, GLuint elt_idx)
 {
    GLuint elt = copy->srcelt[elt_idx];
    GLuint slot = elt & (ELT_TABLE_SIZE-1);
@@ -222,7 +216,6 @@ static GLuint elt(struct copy_context *copy, GLuint elt_idx)
 	       _mesa_printf("%x ", f[j]);
 	    _mesa_printf("\n");
 	 }
-	       
       }
 
       copy->vert_cache[slot].in = elt;
@@ -230,9 +223,8 @@ static GLuint elt(struct copy_context *copy, GLuint elt_idx)
       copy->dstptr += copy->vertex_size;
 
       assert(csr == copy->dstptr);
-      assert(copy->dstptr == (copy->dstbuf + 
-				    copy->dstbuf_nr * 
-				    copy->vertex_size));
+      assert(copy->dstptr == (copy->dstbuf +
+                              copy->dstbuf_nr * copy->vertex_size));
    }
 /*    else */
 /*       _mesa_printf("  --> reuse vertex\n"); */
@@ -242,7 +234,12 @@ static GLuint elt(struct copy_context *copy, GLuint elt_idx)
    return check_flush(copy);
 }
 
-static void end( struct copy_context *copy, GLboolean end_flag )
+
+/**
+ * Called at end of each primitive during replay.
+ */
+static void
+end( struct copy_context *copy, GLboolean end_flag )
 {
    struct _mesa_prim *prim = &copy->dstprim[copy->dstprim_nr];
 
@@ -257,8 +254,8 @@ static void end( struct copy_context *copy, GLboolean end_flag )
 }
 
 
-
-static void replay_elts( struct copy_context *copy )
+static void
+replay_elts( struct copy_context *copy )
 {
    GLuint i, j, k;
    GLboolean split;
@@ -362,7 +359,8 @@ static void replay_elts( struct copy_context *copy )
 }
 
 
-static void replay_init( struct copy_context *copy )
+static void
+replay_init( struct copy_context *copy )
 {
    GLcontext *ctx = copy->ctx;
    GLuint i;
@@ -388,10 +386,7 @@ static void replay_init( struct copy_context *copy )
 	 copy->vertex_size += attr_size(copy->array[i]);
       
 	 if (vbo->Name && !vbo->Pointer) 
-	    ctx->Driver.MapBuffer(ctx,
-				  GL_ARRAY_BUFFER_ARB, 
-				  GL_WRITE_ONLY, /* XXX */
-				  vbo);
+	    ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY, vbo);
 
 	 copy->varying[j].src_ptr = ADD_POINTERS(vbo->Pointer,
 						 copy->array[i]->Ptr);
@@ -405,12 +400,11 @@ static void replay_init( struct copy_context *copy )
     * do it internally.
     */
    if (copy->ib->obj->Name && !copy->ib->obj->Pointer) 
-      ctx->Driver.MapBuffer(ctx, 
-			    GL_ARRAY_BUFFER_ARB, /* XXX */
-			    GL_WRITE_ONLY, /* XXX */
+      ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, GL_READ_ONLY,
 			    copy->ib->obj);
 
-   srcptr = (const GLubyte *)ADD_POINTERS(copy->ib->obj->Pointer, copy->ib->ptr);
+   srcptr = (const GLubyte *) ADD_POINTERS(copy->ib->obj->Pointer,
+                                           copy->ib->ptr);
 
    switch (copy->ib->type) {
    case GL_UNSIGNED_BYTE:
@@ -434,7 +428,6 @@ static void replay_init( struct copy_context *copy )
       copy->srcelt = (const GLuint *)srcptr;
       break;
    }
-   
 
    /* Figure out the maximum allowed vertex buffer size:
     */
@@ -449,8 +442,7 @@ static void replay_init( struct copy_context *copy )
     *
     * XXX:  This should be a VBO!
     */
-   copy->dstbuf = _mesa_malloc(copy->dstbuf_size * 
-			       copy->vertex_size);   
+   copy->dstbuf = _mesa_malloc(copy->dstbuf_size * copy->vertex_size);   
    copy->dstptr = copy->dstbuf;
 
    /* Setup new vertex arrays to point into the output buffer: 
@@ -467,7 +459,7 @@ static void replay_init( struct copy_context *copy )
       dst->Ptr = copy->dstbuf + offset;
       dst->Enabled = GL_TRUE;
       dst->Normalized = src->Normalized; 
-      dst->BufferObj = ctx->Array.NullBufferObj;
+      dst->BufferObj = ctx->Shared->NullBufferObj;
       dst->_MaxElement = copy->dstbuf_size; /* may be less! */
 
       offset += copy->varying[i].size;
@@ -487,12 +479,16 @@ static void replay_init( struct copy_context *copy )
     */
    copy->dstib.count = 0;	/* duplicates dstelt_nr */
    copy->dstib.type = GL_UNSIGNED_INT;
-   copy->dstib.obj = ctx->Array.NullBufferObj;
+   copy->dstib.obj = ctx->Shared->NullBufferObj;
    copy->dstib.ptr = copy->dstelt;
 }
 
 
-static void replay_finish( struct copy_context *copy )
+/**
+ * Free up everything allocated during split/replay.
+ */
+static void
+replay_finish( struct copy_context *copy )
 {
    GLcontext *ctx = copy->ctx;
    GLuint i;
@@ -502,25 +498,26 @@ static void replay_finish( struct copy_context *copy )
    _mesa_free(copy->translated_elt_buf);
    _mesa_free(copy->dstbuf);
    _mesa_free(copy->dstelt);
-   
+
    /* Unmap VBO's 
     */
    for (i = 0; i < copy->nr_varying; i++) {
       struct gl_buffer_object *vbo = copy->varying[i].array->BufferObj;
-
       if (vbo->Name && vbo->Pointer) 
-	 ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER_ARB, vbo);
+	 ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, vbo);
    }
 
    /* Unmap index buffer:
     */
    if (copy->ib->obj->Name && copy->ib->obj->Pointer) {
-      ctx->Driver.UnmapBuffer(ctx, 
-			      GL_ARRAY_BUFFER_ARB, /* XXX */
-			      copy->ib->obj);
+      ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, copy->ib->obj);
    }
 }
 
+
+/**
+ * Split VBO into smaller pieces, draw the pieces.
+ */
 void vbo_split_copy( GLcontext *ctx,
 		     const struct gl_client_array *arrays[],
 		     const struct _mesa_prim *prim,
@@ -546,13 +543,11 @@ void vbo_split_copy( GLcontext *ctx,
    copy.draw = draw;
    copy.limits = limits;
 
-
    /* Clear the vertex cache:
     */
    for (i = 0; i < ELT_TABLE_SIZE; i++)
       copy.vert_cache[i].in = ~0;
 
-
    replay_init(&copy);
    replay_elts(&copy);
    replay_finish(&copy);
diff --git a/src/mesa/vbo/vbo_split_inplace.c b/src/mesa/vbo/vbo_split_inplace.c
index fbc856e93b0..3ed6b34fbf0 100644
--- a/src/mesa/vbo/vbo_split_inplace.c
+++ b/src/mesa/vbo/vbo_split_inplace.c
@@ -221,7 +221,7 @@ static void split_prims( struct split_context *split)
 
 	 ib.count = count;
 	 ib.type = GL_UNSIGNED_INT;
-	 ib.obj = split->ctx->Array.NullBufferObj;
+	 ib.obj = split->ctx->Shared->NullBufferObj;
 	 ib.ptr = elts;
 	    
 	 tmpprim = *prim;