docs: Add 8.0.4 release notes

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>

Makefile

@@ -184,7 +184,7 @@ ultrix-gcc:
 
 # Rules for making release tarballs
 
-PACKAGE_VERSION=8.0.1
+PACKAGE_VERSION=8.0.4
 PACKAGE_DIR = Mesa-$(PACKAGE_VERSION)
 PACKAGE_NAME = MesaLib-$(PACKAGE_VERSION)
 
@@ -199,6 +199,12 @@ EXTRA_FILES = \
 	src/glsl/glcpp/glcpp-lex.c			\
 	src/glsl/glcpp/glcpp-parse.c			\
 	src/glsl/glcpp/glcpp-parse.h			\
+	src/mesa/main/api_exec_es1.c			\
+	src/mesa/main/api_exec_es1_dispatch.h		\
+	src/mesa/main/api_exec_es1_remap_helper.h	\
+	src/mesa/main/api_exec_es2.c			\
+	src/mesa/main/api_exec_es2_dispatch.h		\
+	src/mesa/main/api_exec_es2_remap_helper.h	\
 	src/mesa/program/lex.yy.c			\
 	src/mesa/program/program_parse.tab.c		\
 	src/mesa/program/program_parse.tab.h

bin/mklib

@@ -334,7 +334,9 @@ case $ARCH in
 	    # environment.  If so, pass -m32 flag to linker.
 	    set ${OBJECTS}
 	    ABI32=`file $1 | grep 32-bit`
-	    if [ "${ABI32}" -a `uname -m` = "x86_64" ] ; then
+	    ARM=`file $1 | grep ARM`
+	    # Do not add "-m32" option for arm.
+            if [ -z "$ARM" -a "${ABI32}" -a `uname -m` = "x86_64" ] ; then
 		OPTS="-m32 ${OPTS}"
 	    fi
 
@@ -391,7 +393,9 @@ case $ARCH in
 	    # environment.  If so, pass -m32 flag to linker.
 	    set ${OBJECTS}
 	    ABI32=`file $1 | grep 32-bit`
-	    if [ "${ABI32}" -a `uname -m` = "x86_64" ] ; then
+            ARM=`file $1 | grep ARM`
+	    # Do not add "-m32" option for arm.
+            if [ -z "$ARM" -a "${ABI32}" -a `uname -m` = "x86_64" ] ; then
 		OPTS="-m32 ${OPTS}"
 	    fi
             if [ "${ALTOPTS}" ] ; then

configs/darwin

@@ -9,8 +9,8 @@ INSTALL_DIR = /usr/X11
 X11_DIR = $(INSTALL_DIR)
 
 # Compiler and flags
-CC = gcc
-CXX = g++
+CC = $(shell xcrun -find cc)
+CXX = $(shell xcrun -find c++)
 PIC_FLAGS = -fPIC
 DEFINES =  -D_DARWIN_C_SOURCE -DPTHREADS -D_GNU_SOURCE \
 	   -DGLX_ALIAS_UNSUPPORTED \
@@ -24,11 +24,14 @@ DEFINES =  -D_DARWIN_C_SOURCE -DPTHREADS -D_GNU_SOURCE \
 # -DIN_DRI_DRIVER
 
 ARCH_FLAGS += $(RC_CFLAGS)
+INCLUDE_FLAGS = -I$(INSTALL_DIR)/include -I$(X11_DIR)/include
+OPT_FLAGS = -g3 -gdwarf-2 -Os -ffast-math -fno-strict-aliasing
+WARN_FLAGS = -Wall -Wmissing-prototypes
 
-CFLAGS =  -ggdb3 -Os -Wall -Wmissing-prototypes -std=c99 -ffast-math -fno-strict-aliasing -fvisibility=hidden \
-	-I$(INSTALL_DIR)/include -I$(X11_DIR)/include $(OPT_FLAGS) $(PIC_FLAGS) $(ARCH_FLAGS) $(ASM_FLAGS) $(DEFINES)
-CXXFLAGS =  -ggdb3 -Os -Wall -fno-strict-aliasing -fvisibility=hidden \
-	-I$(INSTALL_DIR)/include -I$(X11_DIR)/include $(OPT_FLAGS) $(PIC_FLAGS) $(ARCH_FLAGS) $(ASM_FLAGS) $(DEFINES)
+CFLAGS = -std=c99 -fvisibility=hidden \
+	$(OPT_FLAGS) $(WARN_FLAGS) $(INCLUDE_FLAGS) $(PIC_FLAGS) $(ARCH_FLAGS) $(ASM_FLAGS) $(DEFINES) $(EXTRA_CFLAGS)
+CXXFLAGS = -fvisibility=hidden \
+	$(OPT_FLAGS) $(WARN_FLAGS) $(INCLUDE_FLAGS) $(PIC_FLAGS) $(ARCH_FLAGS) $(ASM_FLAGS) $(DEFINES) $(EXTRA_CFLAGS)
 
 # Library names (actual file names)
 GL_LIB_NAME = lib$(GL_LIB).dylib
@@ -44,10 +47,10 @@ GLW_LIB_GLOB = lib$(GLW_LIB).*dylib
 OSMESA_LIB_GLOB = lib$(OSMESA_LIB).*dylib
 VG_LIB_GLOB = lib$(VG_LIB).*dylib
 
-GL_LIB_DEPS = -L$(INSTALL_DIR)/$(LIB_DIR) -L$(X11_DIR)/$(LIB_DIR) -lX11 -lXext -lm -lpthread
-OSMESA_LIB_DEPS =
-GLU_LIB_DEPS = -L$(TOP)/$(LIB_DIR) -l$(GL_LIB)
-GLW_LIB_DEPS = -L$(TOP)/$(LIB_DIR) -l$(GL_LIB) -L$(INSTALL_DIR)/$(LIB_DIR) -L$(X11_DIR)/$(LIB_DIR) -lX11 -lXt
+GL_LIB_DEPS = -L$(INSTALL_DIR)/$(LIB_DIR) -L$(X11_DIR)/$(LIB_DIR) -lX11-xcb -lxcb -lX11 -lXext $(EXTRA_LDFLAGS)
+OSMESA_LIB_DEPS = $(EXTRA_LDFLAGS)
+GLU_LIB_DEPS = -L$(TOP)/$(LIB_DIR) -l$(GL_LIB) $(EXTRA_LDFLAGS)
+GLW_LIB_DEPS = -L$(TOP)/$(LIB_DIR) -l$(GL_LIB) -L$(INSTALL_DIR)/$(LIB_DIR) -L$(X11_DIR)/$(LIB_DIR) -lX11 -lXt $(EXTRA_LDFLAGS)
 
 SRC_DIRS = glsl mapi/glapi mapi/vgapi glx/apple mesa gallium glu
 GLU_DIRS = sgi

configs/darwin-fat-intel

@@ -0,0 +1,7 @@
+# Configuration for Darwin / MacOS X, making 32bit and 64bit fat dynamic libs for intel
+
+RC_CFLAGS=-arch i386 -arch x86_64
+
+include $(TOP)/configs/darwin
+
+CONFIG_NAME = darwin-fat-intel

configs/default

@@ -10,7 +10,7 @@ CONFIG_NAME = default
 # Version info
 MESA_MAJOR=8
 MESA_MINOR=0
-MESA_TINY=0
+MESA_TINY=4
 MESA_VERSION = $(MESA_MAJOR).$(MESA_MINOR).$(MESA_TINY)
 
 # external projects.  This should be useless now that we use libdrm.

configure.ac

@@ -168,9 +168,6 @@ esac
 dnl Add flags for gcc and g++
 if test "x$GCC" = xyes; then
     CFLAGS="$CFLAGS -Wall -Wmissing-prototypes -std=c99"
-    if test "x$CLANG" = "xno"; then
-       CFLAGS="$CFLAGS -ffast-math"
-    fi
 
     # Enable -fvisibility=hidden if using a gcc that supports it
     save_CFLAGS="$CFLAGS"

docs/news.html

@@ -11,6 +11,22 @@
 <H1>News</H1>
 
 
+<h2>March 21, 2012</h2>
+
+<p>
+<a href="relnotes-8.0.2.html">Mesa 8.0.2</a> is released.
+This is a bug fix release.
+</p>
+
+
+<h2>February 16, 2012</h2>
+
+<p>
+<a href="relnotes-8.0.1.html">Mesa 8.0.1</a> is released.
+This is a bug fix release.
+</p>
+
+
 <h2>February 9, 2012</h2>
 
 <p>

docs/relnotes-8.0.1.html

@@ -28,7 +28,9 @@ for DRI hardware acceleration.
 
 <h2>MD5 checksums</h2>
 <pre>
-tdb
+4855c2d93bd2ebd43f384bdcc92c9a27  MesaLib-8.0.1.tar.gz
+24eeebf66971809d8f40775a379b36c9  MesaLib-8.0.1.tar.bz2
+54e745d14dac5717f7f65b4e2d5c1df2  MesaLib-8.0.1.zip
 </pre>
 
 <h2>New features</h2>

docs/relnotes-8.0.2.html

@@ -0,0 +1,160 @@
+<HTML>
+
+<head>
+<TITLE>Mesa Release Notes</TITLE>
+<link rel="stylesheet" type="text/css" href="mesa.css">
+<meta http-equiv="content-type" content="text/html; charset=utf-8" />
+</head>
+
+<BODY>
+
+<body bgcolor="#eeeeee">
+
+<H1>Mesa 8.0.2 Release Notes / March 21, 2012</H1>
+
+<p>
+Mesa 8.0.2 is a bug fix release which fixes bugs found since the 8.0.1 release.
+</p>
+<p>
+Mesa 8.0.2 implements the OpenGL 3.0 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 3.0.
+</p>
+<p>
+See the <a href="install.html">Compiling/Installing page</a> for prerequisites
+for DRI hardware acceleration.
+</p>
+
+
+<h2>MD5 checksums</h2>
+<pre>
+70eb3dc74fbfcd72f6776268ee1db52e  MesaLib-8.0.2.tar.gz
+a368104e5700707048dc3e8691a9a7a1  MesaLib-8.0.2.tar.bz2
+d5e5cdb85d2afdbcd1c0623d3ed1c54d  MesaLib-8.0.2.zip
+</pre>
+
+<h2>New features</h2>
+<p>None.</p>
+
+<h2>Bug fixes</h2>
+
+<p>This list is likely incomplete.</p>
+
+<ul>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=38720">Bug 38720</a> - [SNB] Trine triggers a GPU hang</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=40059">Bug 40059</a> - [SNB] hang in "Amnesia: The Dark Descent" demo</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=45216">Bug 45216</a> - [SNB] GPU hang in OilRush</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=46631">Bug 46631</a> - It's really hard to hit the fast path for the fallback glReadPixels code</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=46679">Bug 46679</a> - glReadPixels on a luminance texture returns the wrong values</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=46311">Bug 46311</a> - Missing support of point size in Mesa core</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=46665">Bug 46665</a> - [PNV] webgl conformance case max texture fails</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=45975">Bug 45975</a> - [Gen4 + ILK] render with pointcoord will fail to render</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=46666">Bug 46666</a> - [PNV] webgl conformance case NPOT case fails with TEXTURE_MIN_FILTER set to LINEAR</li>
+
+<!-- <li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=">Bug </a> - </li> -->
+
+</ul>
+
+
+<h2>Changes</h2>
+<p>The full set of changes can be viewed by using the following GIT command:</p>
+
+<pre>
+  git log mesa-8.0.1..mesa-8.0.2
+</pre>
+
+<p>Brian Paul (7):
+<ul>
+      <li>svga: add null vs pointer check in update_need_pipeline()</li>
+      <li>util: add mutex lock in u_debug_memory.c code</li>
+      <li>mesa: add _mesa_rebase_rgba_float/uint() functions</li>
+      <li>mesa: use _mesa_rebase_rgba_float/uint() in glReadPixels code</li>
+      <li>mesa: use _mesa_rebase_rgba_float/uint() in glGetTexImage code</li>
+      <li>mesa: fix GL_LUMINANCE handling in glGetTexImage</li>
+      <li>docs: add links to 8.0.1 release notes</li>
+</ul></p>
+
+<p>Daniel Vetter (1):
+<ul>
+      <li>i965: fixup W-tile offset computation to take swizzling into account</li>
+<ul></p>
+
+<p>Dylan Noblesmith (1):
+<ul>
+      <li>mesa: add back glGetnUniform*v() overflow error reporting</li>
+</ul></p>
+
+<p>Ian Romanick (1):
+<ul>
+      <li>docs: Add 8.0.1 release md5sums</li>
+</ul></p>
+
+<p>Jakob Bornecrantz (3):
+<ul>
+      <li>mesa: Include mesa ES mapi generated files</li>
+      <li>mesa: Bump version number to 8.0.2</li>
+      <li>docs: Add 8.0.2 release notes</li>
+</ul></p>
+
+<p>Jeremy Huddleston (3):
+<ul>
+      <li>darwin: config file cleanups</li>
+      <li>darwin: Build create_context.c</li>
+      <li>darwin: Link against libxcb</li>
+</ul></p>
+
+<p>José Fonseca (1):
+<ul>
+      <li>svga: Clamp advertised PIPE_SHADER_CAP_MAX_TEMPS to SVGA3D_TEMPREG_MAX.</li>
+</ul></p>
+
+<p>Kenneth Graunke (2):
+<ul>
+      <li>i965: Only set Last Render Target Select on the last FB write.</li>
+      <li>i965: Fix Gen6+ dynamic state upper bound on older kernels.</li>
+</ul></p>
+
+<p>Marek Olšák (1):
+<ul>
+      <li>gallium/rtasm: properly detect SSE and SSE2</li>
+</ul></p>
+
+<p>Neil Roberts (1):
+<ul>
+      <li>mesa: Don't disable fast path for normalized types</li>
+</ul></p>
+
+<p>Tom Stellard (1):
+<ul>
+      <li>r300/compiler: Fix bug when lowering KILP on r300 cards</li>
+</ul></p>
+
+<p>Yuanhan Liu (6):
+<ul>
+      <li>mesa: let GL3 buf obj queries not depend on opengl major version</li>
+      <li>tnl: let _TNL_ATTRIB_POINTSIZE do not depend on ctx-&gt;VertexProgram._Enabled</li>
+      <li>i915: fix wrong rendering of gl_PointSize on Pineview</li>
+      <li>i915: move the FALLBACK_DRAW_OFFSET check outside the drawing rect check</li>
+      <li>i965: handle gl_PointCoord for Gen4 and Gen5 platforms</li>
+      <li>i915: fallback for NPOT cubemap texture</li>
+</ul></p>
+
+<p>Zack Rusin (3):
+<ul>
+      <li>svga: fix a crash happening before setting fragment shaders.</li>
+      <li>svga: Fix stencil op mapping</li>
+      <li>svga: fix the rasterizer state resets</li>
+</ul></p>
+
+
+</body>
+</html>

docs/relnotes-8.0.3.html

@@ -0,0 +1,319 @@
+<HTML>
+
+<head>
+<TITLE>Mesa Release Notes</TITLE>
+<link rel="stylesheet" type="text/css" href="mesa.css">
+<meta http-equiv="content-type" content="text/html; charset=utf-8" />
+</head>
+
+<BODY>
+
+<body bgcolor="#eeeeee">
+
+<H1>Mesa 8.0.3 Release Notes / May 18, 2012</H1>
+
+<p>
+Mesa 8.0.3 is a bug fix release which fixes bugs found since the 8.0.2 release.
+</p>
+<p>
+Mesa 8.0.3 implements the OpenGL 3.0 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 3.0.
+</p>
+<p>
+See the <a href="install.html">Compiling/Installing page</a> for prerequisites
+for DRI hardware acceleration.
+</p>
+
+
+<h2>MD5 checksums</h2>
+<pre>
+b7210a6d6e4584bd852ab29294ce717e  MesaLib-8.0.3.tar.gz
+cc5ee15e306b8c15da6a478923797171  MesaLib-8.0.3.tar.bz2
+32773634a0f7e70a680505a81426eccf  MesaLib-8.0.3.zip
+</pre>
+
+<h2>New features</h2>
+<p>None.</p>
+
+<h2>Bug fixes</h2>
+
+<p>This list is likely incomplete.</p>
+
+<ul>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=28138">Bug 28138</a> - [G45] Regnum Online, sparkling in in-game rendering</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=30102">Bug 30102</a> - glean depthStencil test fails BadLength with indirect non-swrast rendering</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=40361">Bug 40361</a> - Glitches on X3100 after upgrade to 7.11</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=41152">Bug 41152</a> - [glsl] Shader backend in Regnum Online does not work</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=41216">Bug 41216</a> - [bisected pineview]oglc filtercubemin(basic.sizedRGBA) fails</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=41372">Bug 41372</a> - i830_state.c PBO crash</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=41495">Bug 41495</a> - i830: intel_get_vb_max / intel_batchbuffer_space mismatch.</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=44701">Bug 44701</a> - Regnum online textures flickering</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=44961">Bug 44961</a> - [bisected i965] oglc sRGB(Mipmap.1D_textures) regressed</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=44970">Bug 44970</a> - [i965]oglc max_values(negative.textureSize.textureCube) segfaults</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=45214">Bug 45214</a> - Textures disappearing or missing in RegnumOnline OpenGL game</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=45558">Bug 45558</a> - cannot render on a drawable of size equal the max framebuffer size</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=45921">Bug 45921</a> - [r300g, bisected] Multiple piglit regressions after glsl_to_tgsi changes</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=46303">Bug 46303</a> - [SNB] segfault in intel_miptree_release()</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=46739">Bug 46739</a> - [snb-m-gt2+] compiz crashed with SIGSEGV in intel_miptree_release()</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=46834">Bug 46834</a> - small performance when playing flightgear (swrast fallback through GTT mapping)</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=47126">Bug 47126</a> - tests/fbo/fbo-array.c:109: create_array_fbo: Assertion `glGetError() == 0' failed.</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=48218">Bug 48218</a> - brw_fs_schedule_instructions.cpp segfault due to accessing not allocated last_mrf_write[16]</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=48545">Bug 48545</a> - LLVMpipe glReadPixels Firefox hits the slow path (WebGL rendering)</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=49124">Bug 49124</a> - swrast/s_texfetch.c:1156: set_fetch_functions: Assertion `texImage-&gt;FetchTexel' failed.</li>
+
+<!-- <li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=">Bug </a> - </li> -->
+
+</ul>
+
+
+<h2>Changes</h2>
+<p>The full set of changes can be viewed by using the following GIT command:</p>
+
+<pre>
+  git log mesa-8.0.2..mesa-8.0.3
+</pre>
+
+<p>Alban Browaeys (1):
+<ul>
+  <li>dri/i915: Fix off-by-one in i830 clip region size.</li>
+</ul></p>
+
+<p>Alex Deucher (2):
+<ul>
+  <li>r200: fix fog coordinate emit</li>
+  <li>radeon: fix fog coordinate emit</li>
+</ul></p>
+
+<p>Alexander von Gluck (4):
+<ul>
+  <li>llvmpipe: fix symbol conflict on Haiku</li>
+  <li>svga: fix typedef conflicts on Haiku</li>
+  <li>mesa: Don't use newlocale on Haiku</li>
+  <li>glsl: Don't use newlocale on Haiku</li>
+</ul></p>
+
+<p>Anuj Phogat (4):
+<ul>
+  <li>mesa: fix issues with texture border and array textures</li>
+  <li>mesa: Fix valid texture target test in _mesa_GetTexLevelParameteriv()</li>
+  <li>mesa: Fix the cause of piglit test fbo-array failure</li>
+  <li>intel: Fix a case when mapping large texture fails</li>
+</ul></p>
+
+<p>Brian Paul (17):
+<ul>
+  <li>mesa: add a couple fast-paths to fast_read_rgba_pixels_memcpy()</li>
+  <li>mesa/gdi: remove wmesa_set_renderbuffer_funcs() function</li>
+  <li>mesa/gdi: remove clear_color() function</li>
+  <li>mesa: bump version to 8.0.2 in configs/default</li>
+  <li>swrast: include s_fragprog.h to silence warnings</li>
+  <li>mesa: remove LSB-first pixel packing check in glReadPixels</li>
+  <li>mesa: fix error in _mesa_format_matches_format_and_type() for RGB888</li>
+  <li>mesa: add BGR888 code in _mesa_format_matches_format_and_type()</li>
+  <li>vbo: fix node_attrsz[] usage in vbo_bind_vertex_list()</li>
+  <li>mesa: add missing texture integer test in glTexSubImage()</li>
+  <li>mesa: add missing return after _mesa_error() in update_array()</li>
+  <li>glsl: propagate MaxUnrollIterations to the optimizer's loop unroller</li>
+  <li>st/mesa: set MaxUnrollIterations = 255</li>
+  <li>st/mesa: no-op glCopyPixels if source region is out of bounds</li>
+  <li>mesa: do more teximage error checking for generic compressed formats</li>
+  <li>mesa: fix/add error check in _mesa_ColorMaterial()</li>
+  <li>mesa: fix glMaterial / dlist bug</li>
+</ul></p>
+
+<p>Chad Versace (3):
+<ul>
+  <li>glsl: Fix Android build</li>
+  <li>main: Fix memory leak in _mesa_make_extension_string()</li>
+  <li>intel: Disable ARB_framebuffer_object in ES contexts</li>
+</ul></p>
+
+<p>Chris Wilson (1):
+<ul>
+  <li>i830: Compute initial number of vertices from remaining batch space</li>
+</ul></p>
+
+<p>Dave Airlie (4):
+<ul>
+  <li>mesa/format_unpack: add LUMINANCE 8/16 UINT/INT</li>
+  <li>glx/drisw: avoid segfaults when we fail to get visual</li>
+  <li>drisw: fix image stride calculation for 16-bit.</li>
+  <li>intel: fix TFP at 16-bpp</li>
+</ul></p>
+
+<p>Dylan Noblesmith (7):
+<ul>
+  <li>intel: fix null dereference processing HiZ buffer</li>
+  <li>util: fix undefined behavior</li>
+  <li>util: fix uninitialized table</li>
+  <li>egl: fix uninitialized values</li>
+  <li>st/vega: fix uninitialized values</li>
+  <li>egl-static: fix printf warning</li>
+  <li>i965: fix typo</li>
+</ul></p>
+
+<p>Eric Anholt (19):
+<ul>
+  <li>i965/fs: Jump from discard statements to the end of the program when done.</li>
+  <li>intel: Fix rendering from textures after RenderTexture().</li>
+  <li>mesa: Fix handling of glCopyBufferSubData() for src == dst.</li>
+  <li>i965/fs: Move GL_CLAMP handling to coordinate setup.</li>
+  <li>i965/fs: Implement GL_CLAMP behavior on texture rectangles on gen6+.</li>
+  <li>mesa: Fix push/pop of multisample coverage invert.</li>
+  <li>mesa: Include the multisample enables under GL_MULTISAMPLE_BIT attrib as well.</li>
+  <li>mesa: Fix display list handling for GL_ARB_draw_instanced.</li>
+  <li>mesa: Fix display lists for draw_elements_base_vertex with draw_instanced.</li>
+  <li>mesa: Add missing error check for first &lt 0 in glDrawArraysInstanced().</li>
+  <li>i915: Fix piglit fbo-nodepth-test on i830.</li>
+  <li>intel: Return success when asked to allocate a 0-width/height renderbuffer.</li>
+  <li>mesa: Throw error on glGetActiveUniform inside Begin/End.</li>
+  <li>i965/vs: Fix up swizzle for dereference_array of matrices.</li>
+  <li>glsl: Fix indentation of switch code.</li>
+  <li>glsl: Let the constructor figure out the types of switch-related expressions.</li>
+  <li>glsl: Reject non-scalar switch expressions.</li>
+  <li>glsl: Fix assertion failure on handling switch on uint expressions.</li>
+  <li>mesa: Check for framebuffer completeness before looking at the rb.</li>
+</ul></p>
+
+<p>Eugeni Dodonov (1):
+<ul>
+  <li>intel: add PCI IDs for Ivy Bridge GT2 server variant</li>
+</ul></p>
+
+<p>Han Shen(沈涵) (1):
+<ul>
+  <li>bin/mklib: remove '-m32' for arm linux</li>
+</ul></p>
+
+<p>Ian Romanick (1):
+<ul>
+  <li>mesa: Bump version number to 8.0.3</li>
+</ul></p>
+
+<p>Jakob Bornecrantz (1):
+<ul>
+  <li>docs: Add 8.0.2 md5sums</li>
+</ul></p>
+
+<p>Jeremy Huddleston (7):
+<ul>
+  <li>darwin: Eliminate a pthread mutex leak</li>
+  <li>darwin: Fix an error message</li>
+  <li>darwin: Make reported errors more user-friendly</li>
+  <li>darwin: Use ASL for logging</li>
+  <li>darwin: Unlock our mutex before destroying it</li>
+  <li>darwin: Eliminate a possible race condition while destroying a surface</li>
+  <li>darwin: Address a build failure on Leopard and earlier OS versions</li>
+</ul></p>
+
+<p>Jon TURNEY (1):
+<ul>
+  <li>Have __glImageSize handle format GL_DEPTH_STENCIL_NV the same way as the server does</li>
+</ul></p>
+
+<p>Jonas Maebe (2):
+<ul>
+  <li>glapi: Correct size of allocated _glapi_table struct</li>
+  <li>apple: Fix a use after free</li>
+</ul></p>
+
+<p>Jordan Justen (1):
+<ul>
+  <li>mesa: Add primitive restart support to glArrayElement</li>
+</ul></p>
+
+<p>Kenneth Graunke (12):
+<ul>
+  <li>i965: Actually upload sampler state pointers for the VS unit on Gen6.</li>
+  <li>i965/fs: Fix FB writes that tried to use the non-existent m16 register.</li>
+  <li>vbo: Remove pedantic warning about 'end' beind out of bounds.</li>
+  <li>vbo: Ignore invalid element ranges which are outside VBO bounds.</li>
+  <li>vbo: Rework checking of 'end' against _MaxElement.</li>
+  <li>vbo: Eliminate short-circuiting in invalid-start case.</li>
+  <li>i965: Fix GPU hangs in the dummy fragment shader.</li>
+  <li>i965: Make the dummy fragment shader work in SIMD16 mode.</li>
+  <li>drirc: Add force_glsl_extensions_warn workaround for Unigine Heaven.</li>
+  <li>i965: Avoid explicit accumulator operands in SIMD16 mode on Gen7.</li>
+  <li>intel: Remove pointless software fallback for glBitmap on Gen6.</li>
+  <li>glsl: Fix broken constant expression handling for &lt, &lt=, &gt;, and &gt;=.</li>
+</ul></p>
+
+<p>Kurt Roeckx (2):
+<ul>
+  <li>i915: Compute maximum number of verts using the actual batchbuffer size.</li>
+  <li>i915: Fix i830 polygon stipple from PBOs.</li>
+</ul></p>
+
+<p>Marek Olšák (5):
+<ul>
+  <li>r300g/swtcl: don't print an error when getting ClipVertex</li>
+  <li>r300g/swtcl: don't enter u_vbuf_mgr</li>
+  <li>r300g/swtcl: don't expose shader subroutine support</li>
+  <li>r300g/swtcl: fix polygon offset</li>
+  <li>r300g/swtcl: fix crash when back color is present in vertex shader</li>
+</ul></p>
+
+<p>Mathias Fröhlich (1):
+<ul>
+  <li>glsl: Avoid excessive loop unrolling.</li>
+</ul></p>
+
+<p>Matt Turner (1):
+<ul>
+  <li>Remove -ffast-math from default CFLAGS</li>
+</ul></p>
+
+<p>Paul Berry (1):
+<ul>
+  <li>i915: Initialize swrast_texture_image structure fields.</li>
+</ul></p>
+
+<p>Roland Scheidegger (1):
+<ul>
+  <li>mesa: check_index_bounds off-by-one fix</li>
+</ul></p>
+
+<p>Tom Stellard (2):
+<ul>
+  <li>r300/compiler: Clear loop registers in vertex shaders w/o loops</li>
+  <li>r300/compiler: Copy all instruction attributes during local transfoms</li>
+</ul></p>
+
+<p>Vinson Lee (5):
+<ul>
+  <li>mesa: Fix memory leak in _mesa_get_uniform_location.</li>
+  <li>linker: Fix memory leak in count_uniform_size::visit_field.</li>
+  <li>swrast: Fix memory leaks in blit_linear.</li>
+  <li>ir_to_mesa: Fix uninitialized member in add_uniform_to_shader.</li>
+  <li>mesa: Fix memory leak in generate_mipmap_compressed.</li>
+</ul></p>
+
+<p>Yuanhan Liu (2):
+<ul>
+  <li>i915: set SPRITE_POINT_ENABLE bit correctly</li>
+  <li>i965: fix wrong cube/3D texture layout</li>
+</ul></p>
+
+</body>
+</html>

docs/relnotes-8.0.4.html

@@ -0,0 +1,202 @@
+<HTML>
+
+<head>
+<TITLE>Mesa Release Notes</TITLE>
+<link rel="stylesheet" type="text/css" href="mesa.css">
+<meta http-equiv="content-type" content="text/html; charset=utf-8" />
+</head>
+
+<BODY>
+
+<body bgcolor="#eeeeee">
+
+<H1>Mesa 8.0.4 Release Notes / July 10, 2012</H1>
+
+<p>
+Mesa 8.0.4 is a bug fix release which fixes bugs found since the 8.0.2 release.
+</p>
+<p>
+Mesa 8.0.4 implements the OpenGL 3.0 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 3.0.
+</p>
+<p>
+See the <a href="install.html">Compiling/Installing page</a> for prerequisites
+for DRI hardware acceleration.
+</p>
+
+
+<h2>MD5 checksums</h2>
+<pre>
+TBD
+</pre>
+
+<h2>New features</h2>
+<p>None.</p>
+
+<h2>Bug fixes</h2>
+
+<p>This list is likely incomplete.</p>
+
+<ul>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=45967">Bug 45967</a> - piglit getteximage-invalid-format-for-packed-type regression</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=47742">Bug 47742</a> - [softpipe] piglit fbo-generatemipmap-array regression</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=48141">Bug 48141</a> - [vmwgfx] src/gallium/auxiliary/util/u_inlines.h:256:pipe_buffer_map_range: Assertion `offset + length &lt;= buffer-&gt;width0' failed.</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=48472">Bug 48472</a> - GPU Lockup while running demo (rzr - the scene is dead) in wine</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=50033">Bug 50033</a> - src/mesa/state_tracker/st_cb_fbo.c:379:st_render_texture: Assertion `strb-&gt;rtt_level &lt;= strb-&gt;texture-&gt;last_level' failed.</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=50621">Bug 50621</a> - Mesa fails its test suite with a buffer overflow.</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=50298">Bug 50298</a> - [ILK IVB bisected]Ogles2conform GL/sin/sin_float_vert_xvary.test regressed</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=51574">Bug 51574</a> - ir_loop_jump constructor assigns member variable to itself</li>
+
+<!-- <li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=">Bug </a> - </li> -->
+
+</ul>
+
+
+<h2>Changes</h2>
+<p>The full set of changes can be viewed by using the following GIT command:</p>
+
+<pre>
+  git log mesa-8.0.3..mesa-8.0.4
+</pre>
+
+<p>Andreas Betz (1):
+<ul>
+  <li>vega: fix 565 color unpacking bug</li>
+</ul></p>
+
+<p>Antoine Labour (2):
+<ul>
+  <li>meta: Cleanup the resources we allocate.</li>
+  <li>mesa: Free uniforms correclty.</li>
+</ul></p>
+
+<p>Brian Paul (22):
+<ul>
+  <li>docs: add link to 8.0.3 release notes</li>
+  <li>mesa: fix Z32_FLOAT -&gt; uint conversion functions</li>
+  <li>draw: fix primitive restart bug by using the index buffer offset</li>
+  <li>st/mesa: fix glDrawPixels(GL_DEPTH_COMPONENT) color output</li>
+  <li>svga: fix synchronization bug between sampler views and surfaces</li>
+  <li>mesa: new _mesa_error_check_format_and_type() function</li>
+  <li>mesa: add missing GL_UNSIGNED_INT_10F_11F_11F_REV case</li>
+  <li>mesa: fix missing return value in getteximage_error_check()</li>
+  <li>st/mesa: pass GL_MAP_INVALIDATE_RANGE_BIT to gallium drivers</li>
+  <li>svga: add 0.5 in float-&gt;int conversion of sample min/max lod</li>
+  <li>svga: fix min/max lod clamping</li>
+  <li>svga: change PIPE_CAPF_MAX_TEXTURE_LOD_BIAS from 16.0 to 15.0</li>
+  <li>st/mesa: add fallback pipe formats for (compressed) R, RG formats</li>
+  <li>st/mesa: copy num_immediates field when copying the immediates array</li>
+  <li>svga: move svga_texture() casts/calls in svga_surface_copy()</li>
+  <li>svga: reset vertex buffer offset in svga_release_user_upl_buffers()</li>
+  <li>st/mesa: don't set PIPE_BIND_DISPLAY_TARGET for user-created renderbuffers</li>
+  <li>st/mesa: use private pipe_sampler_view in decompress_with_blit()</li>
+  <li>st/mesa: add null pointer check in st_texture_image_map()</li>
+  <li>st/mesa: fix mipmap image size computation w.r.t. texture arrays</li>
+  <li>draw: fix missing immediates bug in polygon stipple code</li>
+  <li>st/mesa: fix max_offset computation for base vertex</li>
+</ul></p>
+
+<p>Christoph Bumiller (1):
+<ul>
+  <li>nv50: handle NEG,ABS modifiers for short RCP encoding</li>
+</ul></p>
+
+<p>Dylan Noblesmith (1):
+<ul>
+  <li>mesa: require GL_MAX_SAMPLES &gt;= 4 for GL 3.0</li>
+</ul></p>
+
+<p>Eric Anholt (1):
+<ul>
+  <li>i965/vs: Fix texelFetchOffset()</li>
+</ul></p>
+
+<p>Ian Romanick (5):
+<ul>
+  <li>docs: Add 8.0.3 release md5sums</li>
+  <li>glx/tests: Fix off-by-one error in allocating extension string buffer</li>
+  <li>glsl: Remove spurious printf messages</li>
+  <li>glsl: Fix pi/2 constant in acos built-in function</li>
+  <li>mesa: Bump version number to 8.0.4</li>
+</ul></p>
+
+<p>José Fonseca (2):
+<ul>
+  <li>mesa: Avoid void acinclude.m4 Android.common.mk Android.mk autogen.sh bin common.py configs configure.ac docs doxygen include Makefile scons SConstruct src tests arithmetic.</li>
+  <li>draw: Ensure that prepare is always run after LLVM garbagge collection.</li>
+</ul></p>
+
+<p>Kenneth Graunke (15):
+<ul>
+  <li>mesa: Check for a negative "size" parameter in glCopyBufferSubData().</li>
+  <li>i965: Fix brw_swap_cmod() for LE/GE comparisons.</li>
+  <li>glsl: Remove unused ir_loop_jump::loop pointer.</li>
+  <li>ralloc: Fix ralloc_parent() of memory allocated out of the NULL context.</li>
+  <li>mesa: Restore depth texture state on glPopAttrib(GL_TEXTURE_BIT).</li>
+  <li>glsl/builtins: Fix textureGrad() for Array samplers.</li>
+  <li>mesa: Unbind ARB_copy_buffer and transform feedback buffers on delete.</li>
+  <li>mesa: Support BindBuffer{Base,Offset,Range} with a buffer of 0.</li>
+  <li>mesa: Unbind ARB_transform_feedback2 binding points on Delete too.</li>
+  <li>meta: Fix GL_RENDERBUFFER binding in decompress_texture_image().</li>
+  <li>i965/fs: Fix texelFetchOffset() on pre-Gen7.</li>
+  <li>i965/vs: Fix texelFetchOffset() on pre-Gen7.</li>
+  <li>i965/fs: Fix user-defined FS outputs with less than four components.</li>
+  <li>glsl: Hook up loop_variable_state destructor to plug a memory leak.</li>
+  <li>glsl: Don't trust loop analysis in the presence of function calls.</li>
+</ul></p>
+
+<p>Kurt Roeckx (1):
+<ul>
+  <li>i830: Fix crash for GL_STENCIL_TEST in i830Enable()</li>
+</ul></p>
+
+<p>Lukas Rössler (1):
+<ul>
+  <li>glu: fix two Clang warnings</li>
+</ul></p>
+
+<p>Marek Olšák (2):
+<ul>
+  <li>mesa: allow exposing GL3 without EXT_texture_integer</li>
+  <li>st/mesa: don't do srgb-&gt;linear conversion in decompress_with_blit</li>
+</ul></p>
+
+<p>Paul Seidler (1):
+<ul>
+  <li>tests: include mesa headers</li>
+</ul></p>
+
+<p>Stéphane Marchesin (3):
+<ul>
+  <li>glx: Handle a null reply in QueryVersion.</li>
+  <li>i915g: Don't invert signalled/unsignalled fences</li>
+  <li>i915g: Don't avoid flushing when we have a pending fence.</li>
+</ul></p>
+
+<p>Thomas Gstädtner (1):
+<ul>
+  <li>gallium/targets: pass ldflags parameter to MKLIB</li>
+</ul></p>
+
+<p>Vadim Girlin (2):
+<ul>
+  <li>st/mesa: set stObj-&gt;lastLevel in guess_and_alloc_texture</li>
+  <li>r600g: check gpr count limit</li>
+</ul></p>
+
+<p>Vinson Lee (1):
+<ul>
+  <li>st/mesa: Fix uninitialized members in glsl_to_tgsi_visitor constructor.</li>
+</ul></p>
+
+</body>
+</html>

docs/relnotes.html

@@ -13,6 +13,10 @@ The release notes summarize what's new or changed in each Mesa release.
 </p>
 
 <UL>
+<LI><A HREF="relnotes-8.0.4.html">8.0.4 release notes</A>
+<LI><A HREF="relnotes-8.0.3.html">8.0.3 release notes</A>
+<LI><A HREF="relnotes-8.0.2.html">8.0.2 release notes</A>
+<LI><A HREF="relnotes-8.0.1.html">8.0.1 release notes</A>
 <LI><A HREF="relnotes-8.0.html">8.0 release notes</A>
 <LI><A HREF="relnotes-7.11.html">7.11 release notes</A>
 <LI><A HREF="relnotes-7.10.3.html">7.10.3 release notes</A>

include/pci_ids/i965_pci_ids.h

@@ -25,3 +25,4 @@ CHIPSET(0x0162, IVYBRIDGE_GT2, ivb_gt2)
 CHIPSET(0x0156, IVYBRIDGE_M_GT1, ivb_gt1)
 CHIPSET(0x0166, IVYBRIDGE_M_GT2, ivb_gt2)
 CHIPSET(0x015a, IVYBRIDGE_S_GT1, ivb_gt1)
+CHIPSET(0x016a, IVYBRIDGE_S_GT2, ivb_gt2)

src/egl/main/eglimage.c

@@ -45,7 +45,7 @@ _eglParseImageAttribList(_EGLImageAttribs *attrs, _EGLDisplay *dpy,
 
    (void) dpy;
 
-   memset(attrs, 0, sizeof(attrs));
+   memset(attrs, 0, sizeof(*attrs));
    attrs->ImagePreserved = EGL_FALSE;
    attrs->GLTextureLevel = 0;
    attrs->GLTextureZOffset = 0;

src/gallium/auxiliary/draw/draw_context.h

@@ -80,6 +80,21 @@ void draw_set_viewport_state( struct draw_context *draw,
 void draw_set_clip_state( struct draw_context *pipe,
                           const struct pipe_clip_state *clip );
 
+/**
+ * Sets the rasterization state used by the draw module.
+ * The rast_handle is used to pass the driver specific representation
+ * of the rasterization state. It's going to be used when the
+ * draw module sets the state back on the driver itself using the
+ * pipe::bind_rasterizer_state method.
+ *
+ * NOTE: if you're calling this function from within the pipe's
+ * bind_rasterizer_state you should always call it before binding
+ * the actual state - that's because the draw module can try to
+ * bind its own rasterizer state which would reset your newly
+ * set state. i.e. always do
+ * draw_set_rasterizer_state(driver->draw, state->pipe_state, state);
+ * driver->state.raster = state;
+ */
 void draw_set_rasterizer_state( struct draw_context *draw,
                                 const struct pipe_rasterizer_state *raster,
                                 void *rast_handle );

src/gallium/auxiliary/draw/draw_llvm.c

@@ -65,8 +65,13 @@ static void
 draw_llvm_garbage_collect_callback(void *cb_data)
 {
    struct draw_llvm *llvm = (struct draw_llvm *) cb_data;
+   struct draw_context *draw = llvm->draw;
    struct draw_llvm_variant_list_item *li;
 
+   /* Ensure prepare will be run and shaders recompiled */
+   assert(!draw->suspend_flushing);
+   draw_do_flush(draw, DRAW_FLUSH_STATE_CHANGE);
+
    /* free all shader variants */
    li = first_elem(&llvm->vs_variants_list);
    while (!at_end(&llvm->vs_variants_list, li)) {

src/gallium/auxiliary/draw/draw_pipe_pstipple.c

@@ -165,11 +165,16 @@ pstip_transform_decl(struct tgsi_transform_context *ctx,
 }
 
 
+/**
+ * TGSI immediate declaration transform callback.
+ * We're just counting the number of immediates here.
+ */
 static void
 pstip_transform_immed(struct tgsi_transform_context *ctx,
                       struct tgsi_full_immediate *immed)
 {
    struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx;
+   ctx->emit_immediate(ctx, immed); /* emit to output shader */
    pctx->numImmed++;
 }
 

src/gallium/auxiliary/draw/draw_pt.c

@@ -329,25 +329,28 @@ draw_pt_arrays_restart(struct draw_context *draw,
 
    if (draw->pt.user.elts) {
       /* indexed prims (draw_elements) */
+      const char *elts =
+         (const char *) draw->pt.user.elts + draw->pt.index_buffer.offset;
+
       cur_start = start;
       cur_count = 0;
 
       switch (draw->pt.user.eltSize) {
       case 1:
          {
-            const ubyte *elt_ub = (const ubyte *) draw->pt.user.elts;
+            const ubyte *elt_ub = (const ubyte *) elts;
             PRIM_RESTART_LOOP(elt_ub);
          }
          break;
       case 2:
          {
-            const ushort *elt_us = (const ushort *) draw->pt.user.elts;
+            const ushort *elt_us = (const ushort *) elts;
             PRIM_RESTART_LOOP(elt_us);
          }
          break;
       case 4:
          {
-            const uint *elt_ui = (const uint *) draw->pt.user.elts;
+            const uint *elt_ui = (const uint *) elts;
             PRIM_RESTART_LOOP(elt_ui);
          }
          break;

src/gallium/auxiliary/rtasm/rtasm_cpu.c

@@ -25,43 +25,43 @@
  *
  **************************************************************************/
 
-
-#include "util/u_debug.h"
+#include "pipe/p_config.h"
 #include "rtasm_cpu.h"
 
-
 #if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
-static boolean rtasm_sse_enabled(void)
+
+#include "util/u_debug.h"
+#include "util/u_cpu_detect.h"
+
+DEBUG_GET_ONCE_BOOL_OPTION(nosse, "GALLIUM_NOSSE", FALSE);
+
+static struct util_cpu_caps *get_cpu_caps(void)
 {
-   static boolean firsttime = 1;
-   static boolean enabled;
-   
-   /* This gets called quite often at the moment:
-    */
-   if (firsttime) {
-      enabled =  !debug_get_bool_option("GALLIUM_NOSSE", FALSE);
-      firsttime = FALSE;
-   }
-   return enabled;
+   util_cpu_detect();
+   return &util_cpu_caps;
 }
-#endif
 
 int rtasm_cpu_has_sse(void)
 {
-   /* FIXME: actually detect this at run-time */
-#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
-   return rtasm_sse_enabled();
-#else
-   return 0;
-#endif
+   return !debug_get_option_nosse() && get_cpu_caps()->has_sse;
 }
 
 int rtasm_cpu_has_sse2(void) 
 {
-   /* FIXME: actually detect this at run-time */
-#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
-   return rtasm_sse_enabled();
-#else
-   return 0;
-#endif
+   return !debug_get_option_nosse() && get_cpu_caps()->has_sse2;
 }
+
+
+#else
+
+int rtasm_cpu_has_sse(void)
+{
+   return 0;
+}
+
+int rtasm_cpu_has_sse2(void)
+{
+   return 0;
+}
+
+#endif

src/gallium/auxiliary/util/u_debug_memory.c

@@ -38,6 +38,7 @@
 
 #include "os/os_memory.h"
 #include "os/os_memory_debug.h"
+#include "os/os_thread.h"
 
 #include "util/u_debug.h" 
 #include "util/u_debug_stack.h" 
@@ -72,6 +73,8 @@ struct debug_memory_footer
 
 static struct list_head list = { &list, &list };
 
+pipe_static_mutex(list_mutex);
+
 static unsigned long last_no = 0;
 
 
@@ -132,7 +135,9 @@ debug_malloc(const char *file, unsigned line, const char *function,
    ftr = footer_from_header(hdr);
    ftr->magic = DEBUG_MEMORY_MAGIC;
    
+   pipe_mutex_lock(list_mutex);
    LIST_ADDTAIL(&hdr->head, &list);
+   pipe_mutex_unlock(list_mutex);
    
    return data_from_header(hdr);
 }
@@ -164,7 +169,9 @@ debug_free(const char *file, unsigned line, const char *function,
       debug_assert(0);
    }
 
+   pipe_mutex_lock(list_mutex);
    LIST_DEL(&hdr->head);
+   pipe_mutex_unlock(list_mutex);
    hdr->magic = 0;
    ftr->magic = 0;
    
@@ -232,7 +239,9 @@ debug_realloc(const char *file, unsigned line, const char *function,
    new_ftr = footer_from_header(new_hdr);
    new_ftr->magic = DEBUG_MEMORY_MAGIC;
    
+   pipe_mutex_lock(list_mutex);
    LIST_REPLACE(&old_hdr->head, &new_hdr->head);
+   pipe_mutex_unlock(list_mutex);
 
    /* copy data */
    new_ptr = data_from_header(new_hdr);

src/gallium/auxiliary/util/u_double_list.h

@@ -105,6 +105,11 @@ static INLINE void list_delinit(struct list_head *item)
 #define LIST_IS_EMPTY(__list)                   \
     ((__list)->next == (__list))
 
+/**
+ * Cast from a pointer to a member of a struct back to the containing struct.
+ *
+ * 'sample' MUST be initialized, or else the result is undefined!
+ */
 #ifndef container_of
 #define container_of(ptr, sample, member)				\
     (void *)((char *)(ptr)						\
@@ -112,29 +117,29 @@ static INLINE void list_delinit(struct list_head *item)
 #endif
 
 #define LIST_FOR_EACH_ENTRY(pos, head, member)				\
-   for (pos = container_of((head)->next, pos, member);			\
+   for (pos = NULL, pos = container_of((head)->next, pos, member);	\
 	&pos->member != (head);						\
 	pos = container_of(pos->member.next, pos, member))
 
 #define LIST_FOR_EACH_ENTRY_SAFE(pos, storage, head, member)	\
-   for (pos = container_of((head)->next, pos, member),			\
+   for (pos = NULL, pos = container_of((head)->next, pos, member),	\
 	storage = container_of(pos->member.next, pos, member);	\
 	&pos->member != (head);						\
 	pos = storage, storage = container_of(storage->member.next, storage, member))
 
 #define LIST_FOR_EACH_ENTRY_SAFE_REV(pos, storage, head, member)	\
-   for (pos = container_of((head)->prev, pos, member),			\
+   for (pos = NULL, pos = container_of((head)->prev, pos, member),	\
 	storage = container_of(pos->member.prev, pos, member);		\
 	&pos->member != (head);						\
 	pos = storage, storage = container_of(storage->member.prev, storage, member))
 
 #define LIST_FOR_EACH_ENTRY_FROM(pos, start, head, member)		\
-   for (pos = container_of((start), pos, member);			\
+   for (pos = NULL, pos = container_of((start), pos, member);		\
 	&pos->member != (head);						\
 	pos = container_of(pos->member.next, pos, member))
 
 #define LIST_FOR_EACH_ENTRY_FROM_REV(pos, start, head, member)		\
-   for (pos = container_of((start), pos, member);			\
+   for (pos = NULL, pos = container_of((start), pos, member);		\
 	&pos->member != (head);						\
 	pos = container_of(pos->member.prev, pos, member))
 

src/gallium/auxiliary/util/u_linkage.h

@@ -49,15 +49,16 @@ unsigned util_semantic_set_from_program_file(struct util_semantic_set *set, cons
  *
  * num_slots is the size of the layout array and hardware limit instead.
  *
- * efficient_slots == 0 or efficient_solts == num_slots are typical settings.
+ * efficient_slots == 0 or efficient_slots == num_slots are typical settings.
  */
 void util_semantic_layout_from_set(unsigned char *layout, const struct util_semantic_set *set, unsigned efficient_slots, unsigned num_slots);
 
 static INLINE void
-util_semantic_table_from_layout(unsigned char *table, unsigned char *layout, unsigned char first_slot_value, unsigned char num_slots)
+util_semantic_table_from_layout(unsigned char *table, size_t table_size, unsigned char *layout,
+                                unsigned char first_slot_value, unsigned char num_slots)
 {
-   int i;
-   memset(table, 0xff, sizeof(table));
+   unsigned char i;
+   memset(table, 0xff, table_size);
 
    for(i = 0; i < num_slots; ++i)
       table[layout[i]] = first_slot_value + i;

src/gallium/drivers/i915/i915_flush.c

@@ -45,7 +45,10 @@ static void i915_flush_pipe( struct pipe_context *pipe,
 
    draw_flush(i915->draw);
 
-   if (i915->batch->map == i915->batch->ptr) {
+   /* Only shortcut this if we have no fence, otherwise we must flush the
+    * empty batchbuffer to get our fence back.
+    */
+   if (!fence && i915->batch && (i915->batch->map == i915->batch->ptr)) {
       return;
    }
 

src/gallium/drivers/i915/i915_screen.c

@@ -368,7 +368,7 @@ i915_fence_signalled(struct pipe_screen *screen,
 {
    struct i915_screen *is = i915_screen(screen);
 
-   return is->iws->fence_signalled(is->iws, fence) == 0;
+   return is->iws->fence_signalled(is->iws, fence) == 1;
 }
 
 static boolean
@@ -378,7 +378,7 @@ i915_fence_finish(struct pipe_screen *screen,
 {
    struct i915_screen *is = i915_screen(screen);
 
-   return is->iws->fence_finish(is->iws, fence) == 0;
+   return is->iws->fence_finish(is->iws, fence) == 1;
 }
 
 

src/gallium/drivers/llvmpipe/lp_rast.c

@@ -792,7 +792,7 @@ lp_rast_finish( struct lp_rasterizer *rast )
  *   2. do work
  *   3. signal that we're done
  */
-static PIPE_THREAD_ROUTINE( thread_func, init_data )
+static PIPE_THREAD_ROUTINE( thread_function, init_data )
 {
    struct lp_rasterizer_task *task = (struct lp_rasterizer_task *) init_data;
    struct lp_rasterizer *rast = task->rast;
@@ -860,7 +860,7 @@ create_rast_threads(struct lp_rasterizer *rast)
    for (i = 0; i < rast->num_threads; i++) {
       pipe_semaphore_init(&rast->tasks[i].work_ready, 0);
       pipe_semaphore_init(&rast->tasks[i].work_done, 0);
-      rast->threads[i] = pipe_thread_create(thread_func,
+      rast->threads[i] = pipe_thread_create(thread_function,
                                             (void *) &rast->tasks[i]);
    }
 }

src/gallium/drivers/nv50/nv50_pc_emit.c

@@ -865,8 +865,10 @@ emit_flop(struct nv_pc *pc, struct nv_instruction *i)
    assert(SFILE(i, 0) == NV_FILE_GPR);
 
    if (!i->is_long) {
+      assert(i->opcode == NV_OP_RCP);
       emit_form_MUL(pc, i);
-      assert(i->opcode == NV_OP_RCP && !src0->mod);
+      if (src0->mod & NV_MOD_NEG) pc->emit[0] |= 0x00400000;
+      if (src0->mod & NV_MOD_ABS) pc->emit[0] |= 0x00008000;
       return;
    }
 

src/gallium/drivers/nvfx/nvfx_fragprog.c

@@ -977,7 +977,8 @@ nvfx_fragprog_prepare(struct nvfx_context* nvfx, struct nvfx_fpc *fpc)
 	if(fpc->fp->num_slots > num_texcoords)
 		return FALSE;
 	util_semantic_layout_from_set(fpc->fp->slot_to_generic, &set, 0, num_texcoords);
-	util_semantic_table_from_layout(fpc->generic_to_slot, fpc->fp->slot_to_generic, 0, num_texcoords);
+	util_semantic_table_from_layout(fpc->generic_to_slot, sizeof fpc->generic_to_slot,
+                                        fpc->fp->slot_to_generic, 0, num_texcoords);
 
 	memset(fpc->fp->slot_to_fp_input, 0xff, sizeof(fpc->fp->slot_to_fp_input));
 

src/gallium/drivers/r300/compiler/radeon_program_alu.c

@@ -41,13 +41,16 @@
 
 static struct rc_instruction *emit1(
 	struct radeon_compiler * c, struct rc_instruction * after,
-	rc_opcode Opcode, rc_saturate_mode Saturate, struct rc_dst_register DstReg,
-	struct rc_src_register SrcReg)
+	rc_opcode Opcode, struct rc_sub_instruction * base,
+	struct rc_dst_register DstReg, struct rc_src_register SrcReg)
 {
 	struct rc_instruction *fpi = rc_insert_new_instruction(c, after);
 
+	if (base) {
+		memcpy(&fpi->U.I, base, sizeof(struct rc_sub_instruction));
+	}
+
 	fpi->U.I.Opcode = Opcode;
-	fpi->U.I.SaturateMode = Saturate;
 	fpi->U.I.DstReg = DstReg;
 	fpi->U.I.SrcReg[0] = SrcReg;
 	return fpi;
@@ -55,13 +58,17 @@ static struct rc_instruction *emit1(
 
 static struct rc_instruction *emit2(
 	struct radeon_compiler * c, struct rc_instruction * after,
-	rc_opcode Opcode, rc_saturate_mode Saturate, struct rc_dst_register DstReg,
+	rc_opcode Opcode, struct rc_sub_instruction * base,
+	struct rc_dst_register DstReg,
 	struct rc_src_register SrcReg0, struct rc_src_register SrcReg1)
 {
 	struct rc_instruction *fpi = rc_insert_new_instruction(c, after);
 
+	if (base) {
+		memcpy(&fpi->U.I, base, sizeof(struct rc_sub_instruction));
+	}
+
 	fpi->U.I.Opcode = Opcode;
-	fpi->U.I.SaturateMode = Saturate;
 	fpi->U.I.DstReg = DstReg;
 	fpi->U.I.SrcReg[0] = SrcReg0;
 	fpi->U.I.SrcReg[1] = SrcReg1;
@@ -70,14 +77,18 @@ static struct rc_instruction *emit2(
 
 static struct rc_instruction *emit3(
 	struct radeon_compiler * c, struct rc_instruction * after,
-	rc_opcode Opcode, rc_saturate_mode Saturate, struct rc_dst_register DstReg,
+	rc_opcode Opcode, struct rc_sub_instruction * base,
+	struct rc_dst_register DstReg,
 	struct rc_src_register SrcReg0, struct rc_src_register SrcReg1,
 	struct rc_src_register SrcReg2)
 {
 	struct rc_instruction *fpi = rc_insert_new_instruction(c, after);
 
+	if (base) {
+		memcpy(&fpi->U.I, base, sizeof(struct rc_sub_instruction));
+	}
+
 	fpi->U.I.Opcode = Opcode;
-	fpi->U.I.SaturateMode = Saturate;
 	fpi->U.I.DstReg = DstReg;
 	fpi->U.I.SrcReg[0] = SrcReg0;
 	fpi->U.I.SrcReg[1] = SrcReg1;
@@ -221,7 +232,7 @@ static void transform_ABS(struct radeon_compiler* c,
 	struct rc_src_register src = inst->U.I.SrcReg[0];
 	src.Abs = 1;
 	src.Negate = RC_MASK_NONE;
-	emit1(c, inst->Prev, RC_OPCODE_MOV, inst->U.I.SaturateMode, inst->U.I.DstReg, src);
+	emit1(c, inst->Prev, RC_OPCODE_MOV, &inst->U.I, inst->U.I.DstReg, src);
 	rc_remove_instruction(inst);
 }
 
@@ -240,7 +251,7 @@ static void transform_CEIL(struct radeon_compiler* c,
 
 	struct rc_dst_register dst = try_to_reuse_dst(c, inst);
 	emit1(c, inst->Prev, RC_OPCODE_FRC, 0, dst, negate(inst->U.I.SrcReg[0]));
-	emit2(c, inst->Prev, RC_OPCODE_ADD, inst->U.I.SaturateMode, inst->U.I.DstReg,
+	emit2(c, inst->Prev, RC_OPCODE_ADD, &inst->U.I, inst->U.I.DstReg,
 		inst->U.I.SrcReg[0], srcreg(RC_FILE_TEMPORARY, dst.Index));
 	rc_remove_instruction(inst);
 }
@@ -256,7 +267,7 @@ static void transform_CLAMP(struct radeon_compiler *c,
 	struct rc_dst_register dst = try_to_reuse_dst(c, inst);
 	emit2(c, inst->Prev, RC_OPCODE_MIN, 0, dst,
 		inst->U.I.SrcReg[0], inst->U.I.SrcReg[2]);
-	emit2(c, inst->Prev, RC_OPCODE_MAX, inst->U.I.SaturateMode, inst->U.I.DstReg,
+	emit2(c, inst->Prev, RC_OPCODE_MAX, &inst->U.I, inst->U.I.DstReg,
 		srcreg(RC_FILE_TEMPORARY, dst.Index), inst->U.I.SrcReg[1]);
 	rc_remove_instruction(inst);
 }
@@ -272,7 +283,7 @@ static void transform_DP2(struct radeon_compiler* c,
 	src1.Negate &= ~(RC_MASK_Z | RC_MASK_W);
 	src1.Swizzle &= ~(63 << (3 * 2));
 	src1.Swizzle |= (RC_SWIZZLE_ZERO << (3 * 2)) | (RC_SWIZZLE_ZERO << (3 * 3));
-	emit2(c, inst->Prev, RC_OPCODE_DP3, inst->U.I.SaturateMode, inst->U.I.DstReg, src0, src1);
+	emit2(c, inst->Prev, RC_OPCODE_DP3, &inst->U.I, inst->U.I.DstReg, src0, src1);
 	rc_remove_instruction(inst);
 }
 
@@ -283,7 +294,7 @@ static void transform_DPH(struct radeon_compiler* c,
 	src0.Negate &= ~RC_MASK_W;
 	src0.Swizzle &= ~(7 << (3 * 3));
 	src0.Swizzle |= RC_SWIZZLE_ONE << (3 * 3);
-	emit2(c, inst->Prev, RC_OPCODE_DP4, inst->U.I.SaturateMode, inst->U.I.DstReg, src0, inst->U.I.SrcReg[1]);
+	emit2(c, inst->Prev, RC_OPCODE_DP4, &inst->U.I, inst->U.I.DstReg, src0, inst->U.I.SrcReg[1]);
 	rc_remove_instruction(inst);
 }
 
@@ -294,7 +305,7 @@ static void transform_DPH(struct radeon_compiler* c,
 static void transform_DST(struct radeon_compiler* c,
 	struct rc_instruction* inst)
 {
-	emit2(c, inst->Prev, RC_OPCODE_MUL, inst->U.I.SaturateMode, inst->U.I.DstReg,
+	emit2(c, inst->Prev, RC_OPCODE_MUL, &inst->U.I, inst->U.I.DstReg,
 		swizzle(inst->U.I.SrcReg[0], RC_SWIZZLE_ONE, RC_SWIZZLE_Y, RC_SWIZZLE_Z, RC_SWIZZLE_ONE),
 		swizzle(inst->U.I.SrcReg[1], RC_SWIZZLE_ONE, RC_SWIZZLE_Y, RC_SWIZZLE_ONE, RC_SWIZZLE_W));
 	rc_remove_instruction(inst);
@@ -305,7 +316,7 @@ static void transform_FLR(struct radeon_compiler* c,
 {
 	struct rc_dst_register dst = try_to_reuse_dst(c, inst);
 	emit1(c, inst->Prev, RC_OPCODE_FRC, 0, dst, inst->U.I.SrcReg[0]);
-	emit2(c, inst->Prev, RC_OPCODE_ADD, inst->U.I.SaturateMode, inst->U.I.DstReg,
+	emit2(c, inst->Prev, RC_OPCODE_ADD, &inst->U.I, inst->U.I.DstReg,
 		inst->U.I.SrcReg[0], negate(srcreg(RC_FILE_TEMPORARY, dst.Index)));
 	rc_remove_instruction(inst);
 }
@@ -379,14 +390,14 @@ static void transform_LIT(struct radeon_compiler* c,
 		swizzle_wwww(srctemp));
 
 	/* tmp.z = (tmp.x > 0) ? tmp.w : 0.0 */
-	emit3(c, inst->Prev, RC_OPCODE_CMP, inst->U.I.SaturateMode,
+	emit3(c, inst->Prev, RC_OPCODE_CMP, &inst->U.I,
 		dstregtmpmask(temp, RC_MASK_Z),
 		negate(swizzle_xxxx(srctemp)),
 		swizzle_wwww(srctemp),
 		builtin_zero);
 
 	/* tmp.x, tmp.y, tmp.w = 1.0, tmp.x, 1.0 */
-	emit1(c, inst->Prev, RC_OPCODE_MOV, inst->U.I.SaturateMode,
+	emit1(c, inst->Prev, RC_OPCODE_MOV, &inst->U.I,
 		dstregtmpmask(temp, RC_MASK_XYW),
 		swizzle(srctemp, RC_SWIZZLE_ONE, RC_SWIZZLE_X, RC_SWIZZLE_ONE, RC_SWIZZLE_ONE));
 
@@ -401,7 +412,7 @@ static void transform_LRP(struct radeon_compiler* c,
 	emit2(c, inst->Prev, RC_OPCODE_ADD, 0,
 		dst,
 		inst->U.I.SrcReg[1], negate(inst->U.I.SrcReg[2]));
-	emit3(c, inst->Prev, RC_OPCODE_MAD, inst->U.I.SaturateMode,
+	emit3(c, inst->Prev, RC_OPCODE_MAD, &inst->U.I,
 		inst->U.I.DstReg,
 		inst->U.I.SrcReg[0], srcreg(RC_FILE_TEMPORARY, dst.Index), inst->U.I.SrcReg[2]);
 
@@ -418,7 +429,7 @@ static void transform_POW(struct radeon_compiler* c,
 
 	emit1(c, inst->Prev, RC_OPCODE_LG2, 0, tempdst, swizzle_xxxx(inst->U.I.SrcReg[0]));
 	emit2(c, inst->Prev, RC_OPCODE_MUL, 0, tempdst, tempsrc, swizzle_xxxx(inst->U.I.SrcReg[1]));
-	emit1(c, inst->Prev, RC_OPCODE_EX2, inst->U.I.SaturateMode, inst->U.I.DstReg, tempsrc);
+	emit1(c, inst->Prev, RC_OPCODE_EX2, &inst->U.I, inst->U.I.DstReg, tempsrc);
 
 	rc_remove_instruction(inst);
 }
@@ -472,7 +483,7 @@ static void transform_SEQ(struct radeon_compiler* c,
 	struct rc_dst_register dst = try_to_reuse_dst(c, inst);
 
 	emit2(c, inst->Prev, RC_OPCODE_ADD, 0, dst, inst->U.I.SrcReg[0], negate(inst->U.I.SrcReg[1]));
-	emit3(c, inst->Prev, RC_OPCODE_CMP, inst->U.I.SaturateMode, inst->U.I.DstReg,
+	emit3(c, inst->Prev, RC_OPCODE_CMP, &inst->U.I, inst->U.I.DstReg,
 		negate(absolute(srcreg(RC_FILE_TEMPORARY, dst.Index))), builtin_zero, builtin_one);
 
 	rc_remove_instruction(inst);
@@ -481,7 +492,7 @@ static void transform_SEQ(struct radeon_compiler* c,
 static void transform_SFL(struct radeon_compiler* c,
 	struct rc_instruction* inst)
 {
-	emit1(c, inst->Prev, RC_OPCODE_MOV, inst->U.I.SaturateMode, inst->U.I.DstReg, builtin_zero);
+	emit1(c, inst->Prev, RC_OPCODE_MOV, &inst->U.I, inst->U.I.DstReg, builtin_zero);
 	rc_remove_instruction(inst);
 }
 
@@ -491,7 +502,7 @@ static void transform_SGE(struct radeon_compiler* c,
 	struct rc_dst_register dst = try_to_reuse_dst(c, inst);
 
 	emit2(c, inst->Prev, RC_OPCODE_ADD, 0, dst, inst->U.I.SrcReg[0], negate(inst->U.I.SrcReg[1]));
-	emit3(c, inst->Prev, RC_OPCODE_CMP, inst->U.I.SaturateMode, inst->U.I.DstReg,
+	emit3(c, inst->Prev, RC_OPCODE_CMP, &inst->U.I, inst->U.I.DstReg,
 		srcreg(RC_FILE_TEMPORARY, dst.Index), builtin_zero, builtin_one);
 
 	rc_remove_instruction(inst);
@@ -503,7 +514,7 @@ static void transform_SGT(struct radeon_compiler* c,
 	struct rc_dst_register dst = try_to_reuse_dst(c, inst);
 
 	emit2(c, inst->Prev, RC_OPCODE_ADD, 0, dst, negate(inst->U.I.SrcReg[0]), inst->U.I.SrcReg[1]);
-	emit3(c, inst->Prev, RC_OPCODE_CMP, inst->U.I.SaturateMode, inst->U.I.DstReg,
+	emit3(c, inst->Prev, RC_OPCODE_CMP, &inst->U.I, inst->U.I.DstReg,
 		srcreg(RC_FILE_TEMPORARY, dst.Index), builtin_one, builtin_zero);
 
 	rc_remove_instruction(inst);
@@ -515,7 +526,7 @@ static void transform_SLE(struct radeon_compiler* c,
 	struct rc_dst_register dst = try_to_reuse_dst(c, inst);
 
 	emit2(c, inst->Prev, RC_OPCODE_ADD, 0, dst, negate(inst->U.I.SrcReg[0]), inst->U.I.SrcReg[1]);
-	emit3(c, inst->Prev, RC_OPCODE_CMP, inst->U.I.SaturateMode, inst->U.I.DstReg,
+	emit3(c, inst->Prev, RC_OPCODE_CMP, &inst->U.I, inst->U.I.DstReg,
 		srcreg(RC_FILE_TEMPORARY, dst.Index), builtin_zero, builtin_one);
 
 	rc_remove_instruction(inst);
@@ -527,7 +538,7 @@ static void transform_SLT(struct radeon_compiler* c,
 	struct rc_dst_register dst = try_to_reuse_dst(c, inst);
 
 	emit2(c, inst->Prev, RC_OPCODE_ADD, 0, dst, inst->U.I.SrcReg[0], negate(inst->U.I.SrcReg[1]));
-	emit3(c, inst->Prev, RC_OPCODE_CMP, inst->U.I.SaturateMode, inst->U.I.DstReg,
+	emit3(c, inst->Prev, RC_OPCODE_CMP, &inst->U.I, inst->U.I.DstReg,
 		srcreg(RC_FILE_TEMPORARY, dst.Index), builtin_one, builtin_zero);
 
 	rc_remove_instruction(inst);
@@ -539,7 +550,7 @@ static void transform_SNE(struct radeon_compiler* c,
 	struct rc_dst_register dst = try_to_reuse_dst(c, inst);
 
 	emit2(c, inst->Prev, RC_OPCODE_ADD, 0, dst, inst->U.I.SrcReg[0], negate(inst->U.I.SrcReg[1]));
-	emit3(c, inst->Prev, RC_OPCODE_CMP, inst->U.I.SaturateMode, inst->U.I.DstReg,
+	emit3(c, inst->Prev, RC_OPCODE_CMP, &inst->U.I, inst->U.I.DstReg,
 		negate(absolute(srcreg(RC_FILE_TEMPORARY, dst.Index))), builtin_one, builtin_zero);
 
 	rc_remove_instruction(inst);
@@ -604,7 +615,7 @@ static void transform_XPD(struct radeon_compiler* c,
 	emit2(c, inst->Prev, RC_OPCODE_MUL, 0, dst,
 		swizzle(inst->U.I.SrcReg[0], RC_SWIZZLE_Z, RC_SWIZZLE_X, RC_SWIZZLE_Y, RC_SWIZZLE_W),
 		swizzle(inst->U.I.SrcReg[1], RC_SWIZZLE_Y, RC_SWIZZLE_Z, RC_SWIZZLE_X, RC_SWIZZLE_W));
-	emit3(c, inst->Prev, RC_OPCODE_MAD, inst->U.I.SaturateMode, inst->U.I.DstReg,
+	emit3(c, inst->Prev, RC_OPCODE_MAD, &inst->U.I, inst->U.I.DstReg,
 		swizzle(inst->U.I.SrcReg[0], RC_SWIZZLE_Y, RC_SWIZZLE_Z, RC_SWIZZLE_X, RC_SWIZZLE_W),
 		swizzle(inst->U.I.SrcReg[1], RC_SWIZZLE_Z, RC_SWIZZLE_X, RC_SWIZZLE_Y, RC_SWIZZLE_W),
 		negate(srcreg(RC_FILE_TEMPORARY, dst.Index)));
@@ -719,7 +730,7 @@ static void transform_r300_vertex_DP3(struct radeon_compiler* c,
 	src1.Negate &= ~RC_MASK_W;
 	src1.Swizzle &= ~(7 << (3 * 3));
 	src1.Swizzle |= RC_SWIZZLE_ZERO << (3 * 3);
-	emit2(c, inst->Prev, RC_OPCODE_DP4, inst->U.I.SaturateMode, inst->U.I.DstReg, src0, src1);
+	emit2(c, inst->Prev, RC_OPCODE_DP4, &inst->U.I, inst->U.I.DstReg, src0, src1);
 	rc_remove_instruction(inst);
 }
 
@@ -1043,22 +1054,22 @@ static void r300_transform_SIN_COS_SCS(struct radeon_compiler *c,
 	unsigned srctmp)
 {
 	if (inst->U.I.Opcode == RC_OPCODE_COS) {
-		emit1(c, inst->Prev, RC_OPCODE_COS, inst->U.I.SaturateMode, inst->U.I.DstReg,
+		emit1(c, inst->Prev, RC_OPCODE_COS, &inst->U.I, inst->U.I.DstReg,
 			srcregswz(RC_FILE_TEMPORARY, srctmp, RC_SWIZZLE_WWWW));
 	} else if (inst->U.I.Opcode == RC_OPCODE_SIN) {
-		emit1(c, inst->Prev, RC_OPCODE_SIN, inst->U.I.SaturateMode,
+		emit1(c, inst->Prev, RC_OPCODE_SIN, &inst->U.I,
 			inst->U.I.DstReg, srcregswz(RC_FILE_TEMPORARY, srctmp, RC_SWIZZLE_WWWW));
 	} else if (inst->U.I.Opcode == RC_OPCODE_SCS) {
 		struct rc_dst_register moddst = inst->U.I.DstReg;
 
 		if (inst->U.I.DstReg.WriteMask & RC_MASK_X) {
 			moddst.WriteMask = RC_MASK_X;
-			emit1(c, inst->Prev, RC_OPCODE_COS, inst->U.I.SaturateMode, moddst,
+			emit1(c, inst->Prev, RC_OPCODE_COS, &inst->U.I, moddst,
 				srcregswz(RC_FILE_TEMPORARY, srctmp, RC_SWIZZLE_WWWW));
 		}
 		if (inst->U.I.DstReg.WriteMask & RC_MASK_Y) {
 			moddst.WriteMask = RC_MASK_Y;
-			emit1(c, inst->Prev, RC_OPCODE_SIN, inst->U.I.SaturateMode, moddst,
+			emit1(c, inst->Prev, RC_OPCODE_SIN, &inst->U.I, moddst,
 				srcregswz(RC_FILE_TEMPORARY, srctmp, RC_SWIZZLE_WWWW));
 		}
 	}
@@ -1165,35 +1176,79 @@ int radeonTransformDeriv(struct radeon_compiler* c,
 }
 
 /**
+ * IF Temp[0].x -> IF Temp[0].x
+ * ...          -> ...
+ * KILP         -> KIL -abs(Temp[0].x)
+ * ...          -> ...
+ * ENDIF        -> ENDIF
+ *
+ * === OR ===
+ *
  * IF Temp[0].x -\
  * KILP         - > KIL -abs(Temp[0].x)
  * ENDIF        -/
  *
- * This needs to be done in its own pass, because it modifies the instructions
- * before and after KILP.
+ * === OR ===
+ *
+ * IF Temp[0].x -> IF Temp[0].x
+ * ...          -> ...
+ * ELSE         -> ELSE
+ * ...	        -> ...
+ * KILP	        -> KIL -abs(Temp[0].x)
+ * ...          -> ...
+ * ENDIF        -> ENDIF
+ *
+ * === OR ===
+ *
+ * KILP         -> KIL -none.1111
+ *
+ * This needs to be done in its own pass, because it might modify the
+ * instructions before and after KILP.
  */
 void rc_transform_KILP(struct radeon_compiler * c, void *user)
 {
 	struct rc_instruction * inst;
 	for (inst = c->Program.Instructions.Next;
 			inst != &c->Program.Instructions; inst = inst->Next) {
+		struct rc_instruction * if_inst;
+		unsigned in_if = 0;
 
 		if (inst->U.I.Opcode != RC_OPCODE_KILP)
 			continue;
 
+		for (if_inst = inst->Prev; if_inst != &c->Program.Instructions;
+						if_inst = if_inst->Prev) {
+
+			if (if_inst->U.I.Opcode == RC_OPCODE_IF) {
+				in_if = 1;
+				break;
+			}
+		}
+
 		inst->U.I.Opcode = RC_OPCODE_KIL;
 
-		if (inst->Prev->U.I.Opcode != RC_OPCODE_IF
-				|| inst->Next->U.I.Opcode != RC_OPCODE_ENDIF) {
+		if (!in_if) {
 			inst->U.I.SrcReg[0] = negate(builtin_one);
 		} else {
-
+			/* This should work even if the KILP is inside the ELSE
+			 * block, because -0.0 is considered negative. */
 			inst->U.I.SrcReg[0] =
-				negate(absolute(inst->Prev->U.I.SrcReg[0]));
-			/* Remove IF */
-			rc_remove_instruction(inst->Prev);
-			/* Remove ENDIF */
-			rc_remove_instruction(inst->Next);
+				negate(absolute(if_inst->U.I.SrcReg[0]));
+
+			if (inst->Prev->U.I.Opcode != RC_OPCODE_IF
+				&& inst->Next->U.I.Opcode != RC_OPCODE_ENDIF) {
+
+				/* Optimize the special case:
+				 * IF Temp[0].x
+				 * KILP
+				 * ENDIF
+				 */
+
+				/* Remove IF */
+				rc_remove_instruction(inst->Prev);
+				/* Remove ENDIF */
+				rc_remove_instruction(inst->Next);
+			}
 		}
 	}
 }

src/gallium/drivers/r300/r300_blit.c

@@ -63,8 +63,13 @@ static void r300_blitter_begin(struct r300_context* r300, enum r300_blitter_op o
     util_blitter_save_vertex_shader(r300->blitter, r300->vs_state.state);
     util_blitter_save_viewport(r300->blitter, &r300->viewport);
     util_blitter_save_vertex_elements(r300->blitter, r300->velems);
-    util_blitter_save_vertex_buffers(r300->blitter, r300->vbuf_mgr->nr_vertex_buffers,
-                                     r300->vbuf_mgr->vertex_buffer);
+    if (r300->vbuf_mgr) {
+        util_blitter_save_vertex_buffers(r300->blitter, r300->vbuf_mgr->nr_vertex_buffers,
+                                         r300->vbuf_mgr->vertex_buffer);
+    } else {
+        util_blitter_save_vertex_buffers(r300->blitter, r300->swtcl_nr_vertex_buffers,
+                                         r300->swtcl_vertex_buffer);
+    }
 
     if (op & R300_SAVE_FRAMEBUFFER) {
         util_blitter_save_framebuffer(r300->blitter, r300->fb_state.state);

src/gallium/drivers/r300/r300_context.c

@@ -419,17 +419,19 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
     r300_init_query_functions(r300);
     r300_init_state_functions(r300);
     r300_init_resource_functions(r300);
-    
+
     r300->context.create_video_decoder = vl_create_decoder;
     r300->context.create_video_buffer = vl_video_buffer_create;
 
-    r300->vbuf_mgr = u_vbuf_create(&r300->context, 1024 * 1024, 16,
+    if (r300->screen->caps.has_tcl) {
+        r300->vbuf_mgr = u_vbuf_create(&r300->context, 1024 * 1024, 16,
                                        PIPE_BIND_VERTEX_BUFFER |
                                        PIPE_BIND_INDEX_BUFFER,
                                        U_VERTEX_FETCH_DWORD_ALIGNED);
-    if (!r300->vbuf_mgr)
-        goto fail;
-    r300->vbuf_mgr->caps.format_fixed32 = 0;
+        if (!r300->vbuf_mgr)
+            goto fail;
+        r300->vbuf_mgr->caps.format_fixed32 = 0;
+    }
 
     r300->blitter = util_blitter_create(&r300->context);
     if (r300->blitter == NULL)

src/gallium/drivers/r300/r300_context.h

@@ -581,6 +581,9 @@ struct r300_context {
     void *dsa_decompress_zmask;
 
     struct u_vbuf *vbuf_mgr;
+    struct pipe_index_buffer swtcl_index_buffer;
+    struct pipe_vertex_buffer swtcl_vertex_buffer[PIPE_MAX_ATTRIBS];
+    unsigned swtcl_nr_vertex_buffers;
 
     struct util_slab_mempool pool_transfers;
 

src/gallium/drivers/r300/r300_emit.c

@@ -1030,20 +1030,18 @@ void r300_emit_vs_state(struct r300_context* r300, unsigned size, void* state)
             R300_PVS_VF_MAX_VTX_NUM(12) |
             (r300screen->caps.is_r500 ? R500_TCL_STATE_OPTIMIZATION : 0));
 
-    /* Emit flow control instructions. */
-    if (code->num_fc_ops) {
-
-        OUT_CS_REG(R300_VAP_PVS_FLOW_CNTL_OPC, code->fc_ops);
-        if (r300screen->caps.is_r500) {
-            OUT_CS_REG_SEQ(R500_VAP_PVS_FLOW_CNTL_ADDRS_LW_0, code->num_fc_ops * 2);
-            OUT_CS_TABLE(code->fc_op_addrs.r500, code->num_fc_ops * 2);
-        } else {
-            OUT_CS_REG_SEQ(R300_VAP_PVS_FLOW_CNTL_ADDRS_0, code->num_fc_ops);
-            OUT_CS_TABLE(code->fc_op_addrs.r300, code->num_fc_ops);
-        }
-        OUT_CS_REG_SEQ(R300_VAP_PVS_FLOW_CNTL_LOOP_INDEX_0, code->num_fc_ops);
-        OUT_CS_TABLE(code->fc_loop_index, code->num_fc_ops);
+    /* Emit flow control instructions.  Even if there are no fc instructions,
+     * we still need to write the registers to make sure they are cleared. */
+    OUT_CS_REG(R300_VAP_PVS_FLOW_CNTL_OPC, code->fc_ops);
+    if (r300screen->caps.is_r500) {
+        OUT_CS_REG_SEQ(R500_VAP_PVS_FLOW_CNTL_ADDRS_LW_0, R300_VS_MAX_FC_OPS * 2);
+        OUT_CS_TABLE(code->fc_op_addrs.r500, R300_VS_MAX_FC_OPS * 2);
+    } else {
+        OUT_CS_REG_SEQ(R300_VAP_PVS_FLOW_CNTL_ADDRS_0, R300_VS_MAX_FC_OPS);
+        OUT_CS_TABLE(code->fc_op_addrs.r300, R300_VS_MAX_FC_OPS);
     }
+    OUT_CS_REG_SEQ(R300_VAP_PVS_FLOW_CNTL_LOOP_INDEX_0, R300_VS_MAX_FC_OPS);
+    OUT_CS_TABLE(code->fc_loop_index, R300_VS_MAX_FC_OPS);
 
     END_CS;
 }

src/gallium/drivers/r300/r300_render.c

@@ -818,7 +818,7 @@ static void r300_swtcl_draw_vbo(struct pipe_context* pipe,
     struct pipe_transfer *ib_transfer = NULL;
     int i;
     void *indices = NULL;
-    boolean indexed = info->indexed && r300->vbuf_mgr->index_buffer.buffer;
+    boolean indexed = info->indexed && r300->swtcl_index_buffer.buffer;
 
     if (r300->skip_rendering) {
         return;
@@ -831,10 +831,10 @@ static void r300_swtcl_draw_vbo(struct pipe_context* pipe,
             (indexed ? PREP_INDEXED : 0),
             indexed ? 256 : 6);
 
-    for (i = 0; i < r300->vbuf_mgr->nr_vertex_buffers; i++) {
-        if (r300->vbuf_mgr->vertex_buffer[i].buffer) {
+    for (i = 0; i < r300->swtcl_nr_vertex_buffers; i++) {
+        if (r300->swtcl_vertex_buffer[i].buffer) {
             void *buf = pipe_buffer_map(pipe,
-                                  r300->vbuf_mgr->vertex_buffer[i].buffer,
+                                  r300->swtcl_vertex_buffer[i].buffer,
                                   PIPE_TRANSFER_READ |
                                   PIPE_TRANSFER_UNSYNCHRONIZED,
                                   &vb_transfer[i]);
@@ -843,7 +843,7 @@ static void r300_swtcl_draw_vbo(struct pipe_context* pipe,
     }
 
     if (indexed) {
-        indices = pipe_buffer_map(pipe, r300->vbuf_mgr->index_buffer.buffer,
+        indices = pipe_buffer_map(pipe, r300->swtcl_index_buffer.buffer,
                                   PIPE_TRANSFER_READ |
                                   PIPE_TRANSFER_UNSYNCHRONIZED, &ib_transfer);
     }
@@ -856,8 +856,8 @@ static void r300_swtcl_draw_vbo(struct pipe_context* pipe,
     draw_flush(r300->draw);
     r300->draw_vbo_locked = FALSE;
 
-    for (i = 0; i < r300->vbuf_mgr->nr_vertex_buffers; i++) {
-        if (r300->vbuf_mgr->vertex_buffer[i].buffer) {
+    for (i = 0; i < r300->swtcl_nr_vertex_buffers; i++) {
+        if (r300->swtcl_vertex_buffer[i].buffer) {
             pipe_buffer_unmap(pipe, vb_transfer[i]);
             draw_set_mapped_vertex_buffer(r300->draw, i, NULL);
         }

src/gallium/drivers/r300/r300_screen.c

@@ -212,6 +212,7 @@ static int r300_get_shader_param(struct pipe_screen *pscreen, unsigned shader, e
         switch (param)
         {
         case PIPE_SHADER_CAP_MAX_TEXTURE_SAMPLERS:
+        case PIPE_SHADER_CAP_SUBROUTINES:
             return 0;
         default:;
         }

src/gallium/drivers/r300/r300_state.c

@@ -1048,6 +1048,10 @@ static void* r300_create_rs_state(struct pipe_context* pipe,
 
     /* Override some states for Draw. */
     rs->rs_draw.sprite_coord_enable = 0; /* We can do this in HW. */
+    rs->rs_draw.offset_point = 0;
+    rs->rs_draw.offset_line = 0;
+    rs->rs_draw.offset_tri = 0;
+    rs->rs_draw.offset_clamp = 0;
 
 #ifdef PIPE_ARCH_LITTLE_ENDIAN
     vap_control_status = R300_VC_NO_SWAP;
@@ -1595,7 +1599,6 @@ static void r300_set_vertex_buffers(struct pipe_context* pipe,
                                     const struct pipe_vertex_buffer* buffers)
 {
     struct r300_context* r300 = r300_context(pipe);
-    unsigned i;
     struct pipe_vertex_buffer dummy_vb = {0};
 
     /* There must be at least one vertex buffer set, otherwise it locks up. */
@@ -1605,18 +1608,13 @@ static void r300_set_vertex_buffers(struct pipe_context* pipe,
         count = 1;
     }
 
-    u_vbuf_set_vertex_buffers(r300->vbuf_mgr, count, buffers);
-
     if (r300->screen->caps.has_tcl) {
-        /* HW TCL. */
-        for (i = 0; i < count; i++) {
-            if (buffers[i].buffer &&
-		!r300_resource(buffers[i].buffer)->b.user_ptr) {
-            }
-        }
+        u_vbuf_set_vertex_buffers(r300->vbuf_mgr, count, buffers);
         r300->vertex_arrays_dirty = TRUE;
     } else {
-        /* SW TCL. */
+        util_copy_vertex_buffers(r300->swtcl_vertex_buffer,
+                                 &r300->swtcl_nr_vertex_buffers,
+                                 buffers, count);
         draw_set_vertex_buffers(r300->draw, count, buffers);
     }
 }
@@ -1626,9 +1624,15 @@ static void r300_set_index_buffer(struct pipe_context* pipe,
 {
     struct r300_context* r300 = r300_context(pipe);
 
-    u_vbuf_set_index_buffer(r300->vbuf_mgr, ib);
-
-    if (!r300->screen->caps.has_tcl) {
+    if (r300->screen->caps.has_tcl) {
+        u_vbuf_set_index_buffer(r300->vbuf_mgr, ib);
+    } else {
+        if (ib) {
+            pipe_resource_reference(&r300->swtcl_index_buffer.buffer, ib->buffer);
+            memcpy(&r300->swtcl_index_buffer, ib, sizeof(*ib));
+        } else {
+            pipe_resource_reference(&r300->swtcl_index_buffer.buffer, NULL);
+        }
         draw_set_index_buffer(r300->draw, ib);
     }
 }
@@ -1702,11 +1706,11 @@ static void* r300_create_vertex_elements_state(struct pipe_context* pipe,
         return NULL;
 
     velems->count = count;
-    velems->vmgr_elements =
-        u_vbuf_create_vertex_elements(r300->vbuf_mgr, count, attribs,
-                                          velems->velem);
 
     if (r300_screen(pipe->screen)->caps.has_tcl) {
+        velems->vmgr_elements =
+            u_vbuf_create_vertex_elements(r300->vbuf_mgr, count, attribs,
+                                          velems->velem);
         /* Setup PSC.
          * The unused components will be replaced by (..., 0, 1). */
         r300_vertex_psc(velems);
@@ -1716,6 +1720,8 @@ static void* r300_create_vertex_elements_state(struct pipe_context* pipe,
                 align(util_format_get_blocksize(velems->velem[i].src_format), 4);
             velems->vertex_size_dwords += velems->format_size[i] / 4;
         }
+    } else {
+        memcpy(velems->velem, attribs, count * sizeof(struct pipe_vertex_element));
     }
 
     return velems;
@@ -1733,9 +1739,9 @@ static void r300_bind_vertex_elements_state(struct pipe_context *pipe,
 
     r300->velems = velems;
 
-    u_vbuf_bind_vertex_elements(r300->vbuf_mgr, state, velems->vmgr_elements);
-
-    if (r300->draw) {
+    if (r300->screen->caps.has_tcl) {
+        u_vbuf_bind_vertex_elements(r300->vbuf_mgr, state, velems->vmgr_elements);
+    } else {
         draw_set_vertex_elements(r300->draw, velems->count, velems->velem);
         return;
     }
@@ -1750,7 +1756,9 @@ static void r300_delete_vertex_elements_state(struct pipe_context *pipe, void *s
     struct r300_context *r300 = r300_context(pipe);
     struct r300_vertex_element_state *velems = state;
 
-    u_vbuf_destroy_vertex_elements(r300->vbuf_mgr, velems->vmgr_elements);
+    if (r300->screen->caps.has_tcl) {
+        u_vbuf_destroy_vertex_elements(r300->vbuf_mgr, velems->vmgr_elements);
+    }
     FREE(state);
 }
 
@@ -1765,10 +1773,10 @@ static void* r300_create_vs_state(struct pipe_context* pipe,
     vs->state.tokens = tgsi_dup_tokens(shader->tokens);
 
     if (r300->screen->caps.has_tcl) {
-        r300_init_vs_outputs(vs);
+        r300_init_vs_outputs(r300, vs);
         r300_translate_vertex_shader(r300, vs);
     } else {
-        r300_draw_init_vertex_shader(r300->draw, vs);
+        r300_draw_init_vertex_shader(r300, vs);
     }
 
     return vs;
@@ -1794,9 +1802,8 @@ static void r300_bind_vs_state(struct pipe_context* pipe, void* shader)
     if (r300->screen->caps.has_tcl) {
         unsigned fc_op_dwords = r300->screen->caps.is_r500 ? 3 : 2;
         r300_mark_atom_dirty(r300, &r300->vs_state);
-        r300->vs_state.size =
-                vs->code.length + 9 +
-        (vs->code.num_fc_ops ? vs->code.num_fc_ops * fc_op_dwords + 4 : 0);
+        r300->vs_state.size = vs->code.length + 9 +
+			(R300_VS_MAX_FC_OPS * fc_op_dwords + 4);
 
         r300_mark_atom_dirty(r300, &r300->vs_constants);
         r300->vs_constants.size =

src/gallium/drivers/r300/r300_vs.c

@@ -36,6 +36,7 @@
 
 /* Convert info about VS output semantics into r300_shader_semantics. */
 static void r300_shader_read_vs_outputs(
+    struct r300_context *r300,
     struct tgsi_shader_info* info,
     struct r300_shader_semantics* vs_outputs)
 {
@@ -83,6 +84,14 @@ static void r300_shader_read_vs_outputs(
                 fprintf(stderr, "r300 VP: cannot handle edgeflag output.\n");
                 break;
 
+            case TGSI_SEMANTIC_CLIPVERTEX:
+                assert(index == 0);
+                /* Draw does clip vertex for us. */
+                if (r300->screen->caps.has_tcl) {
+                    fprintf(stderr, "r300 VP: cannot handle clip vertex output.\n");
+                }
+                break;
+
             default:
                 fprintf(stderr, "r300 VP: unknown vertex output semantic: %i.\n",
                         info->output_semantic_name[i]);
@@ -160,10 +169,11 @@ static void set_vertex_inputs_outputs(struct r300_vertex_program_compiler * c)
     c->code->outputs[outputs->wpos] = reg++;
 }
 
-void r300_init_vs_outputs(struct r300_vertex_shader *vs)
+void r300_init_vs_outputs(struct r300_context *r300,
+                          struct r300_vertex_shader *vs)
 {
     tgsi_scan_shader(vs->state.tokens, &vs->info);
-    r300_shader_read_vs_outputs(&vs->info, &vs->outputs);
+    r300_shader_read_vs_outputs(r300, &vs->info, &vs->outputs);
 }
 
 static void r300_dummy_vertex_shader(
@@ -187,7 +197,7 @@ static void r300_dummy_vertex_shader(
     ureg_destroy(ureg);
 
     shader->dummy = TRUE;
-    r300_init_vs_outputs(shader);
+    r300_init_vs_outputs(r300, shader);
     r300_translate_vertex_shader(r300, shader);
 }
 

src/gallium/drivers/r300/r300_vs.h

@@ -56,12 +56,13 @@ struct r300_vertex_shader {
     void *draw_vs;
 };
 
-void r300_init_vs_outputs(struct r300_vertex_shader *vs);
+void r300_init_vs_outputs(struct r300_context *r300,
+                          struct r300_vertex_shader *vs);
 
 void r300_translate_vertex_shader(struct r300_context *r300,
                                   struct r300_vertex_shader *vs);
 
-void r300_draw_init_vertex_shader(struct draw_context *draw,
+void r300_draw_init_vertex_shader(struct r300_context *r300,
                                   struct r300_vertex_shader *vs);
 
 #endif /* R300_VS_H */

src/gallium/drivers/r300/r300_vs_draw.c

@@ -29,7 +29,7 @@
  *
  * Transformations:
  * 1) If the secondary color output is present, the primary color must be
- *    inserted before it.
+ *    present too.
  * 2) If any back-face color output is present, there must be all 4 color
  *    outputs and missing ones must be inserted.
  * 3) Insert a trailing texcoord output containing a copy of POS, for WPOS.
@@ -52,7 +52,6 @@ struct vs_transform_context {
 
     boolean color_used[2];
     boolean bcolor_used[2];
-    boolean temp_used[128];
 
     /* Index of the pos output, typically 0. */
     unsigned pos_output;
@@ -72,6 +71,8 @@ struct vs_transform_context {
     boolean first_instruction;
     /* End instruction processed? */
     boolean end_instruction;
+
+    boolean temp_used[1024];
 };
 
 static void emit_temp(struct tgsi_transform_context *ctx, unsigned reg)
@@ -102,9 +103,9 @@ static void emit_output(struct tgsi_transform_context *ctx,
     ++vsctx->num_outputs;
 }
 
-static void insert_output(struct tgsi_transform_context *ctx,
-                          struct tgsi_full_declaration *before,
-                          unsigned name, unsigned index, unsigned interp)
+static void insert_output_before(struct tgsi_transform_context *ctx,
+                                 struct tgsi_full_declaration *before,
+                                 unsigned name, unsigned index, unsigned interp)
 {
     struct vs_transform_context *vsctx = (struct vs_transform_context *)ctx;
     unsigned i;
@@ -115,28 +116,29 @@ static void insert_output(struct tgsi_transform_context *ctx,
     }
 
     /* Insert the new output. */
-    emit_output(ctx, name, index, interp, before->Range.First);
+    emit_output(ctx, name, index, interp,
+                before->Range.First + vsctx->decl_shift);
 
     ++vsctx->decl_shift;
 }
 
-static void insert_trailing_bcolor(struct tgsi_transform_context *ctx,
-                                   struct tgsi_full_declaration *before)
+static void insert_output_after(struct tgsi_transform_context *ctx,
+                                struct tgsi_full_declaration *after,
+                                unsigned name, unsigned index, unsigned interp)
 {
     struct vs_transform_context *vsctx = (struct vs_transform_context *)ctx;
+    unsigned i;
 
-    /* If BCOLOR0 is used, make sure BCOLOR1 is present too. Otherwise
-     * the rasterizer doesn't do the color selection correctly. */
-    if (vsctx->bcolor_used[0] && !vsctx->bcolor_used[1]) {
-        if (before) {
-            insert_output(ctx, before, TGSI_SEMANTIC_BCOLOR, 1,
-                          TGSI_INTERPOLATE_LINEAR);
-        } else {
-            emit_output(ctx, TGSI_SEMANTIC_BCOLOR, 1,
-                        TGSI_INTERPOLATE_LINEAR, vsctx->num_outputs);
-        }
-        vsctx->bcolor_used[1] = TRUE;
+    /* Make a place for the new output. */
+    for (i = after->Range.First+1; i < Elements(vsctx->out_remap); i++) {
+        ++vsctx->out_remap[i];
     }
+
+    /* Insert the new output. */
+    emit_output(ctx, name, index, interp,
+                after->Range.First + 1);
+
+    ++vsctx->decl_shift;
 }
 
 static void transform_decl(struct tgsi_transform_context *ctx,
@@ -153,41 +155,38 @@ static void transform_decl(struct tgsi_transform_context *ctx,
 
             case TGSI_SEMANTIC_COLOR:
                 assert(decl->Semantic.Index < 2);
-                vsctx->color_used[decl->Semantic.Index] = TRUE;
 
                 /* We must rasterize the first color if the second one is
                  * used, otherwise the rasterizer doesn't do the color
                  * selection correctly. Declare it, but don't write to it. */
                 if (decl->Semantic.Index == 1 && !vsctx->color_used[0]) {
-                    insert_output(ctx, decl, TGSI_SEMANTIC_COLOR, 0,
-                                  TGSI_INTERPOLATE_LINEAR);
+                    insert_output_before(ctx, decl, TGSI_SEMANTIC_COLOR, 0,
+                                         TGSI_INTERPOLATE_LINEAR);
                     vsctx->color_used[0] = TRUE;
                 }
                 break;
 
             case TGSI_SEMANTIC_BCOLOR:
                 assert(decl->Semantic.Index < 2);
-                vsctx->bcolor_used[decl->Semantic.Index] = TRUE;
 
                 /* We must rasterize all 4 colors if back-face colors are
                  * used, otherwise the rasterizer doesn't do the color
                  * selection correctly. Declare it, but don't write to it. */
                 if (!vsctx->color_used[0]) {
-                    insert_output(ctx, decl, TGSI_SEMANTIC_COLOR, 0,
-                                  TGSI_INTERPOLATE_LINEAR);
+                    insert_output_before(ctx, decl, TGSI_SEMANTIC_COLOR, 0,
+                                         TGSI_INTERPOLATE_LINEAR);
                     vsctx->color_used[0] = TRUE;
                 }
                 if (!vsctx->color_used[1]) {
-                    insert_output(ctx, decl, TGSI_SEMANTIC_COLOR, 1,
-                                  TGSI_INTERPOLATE_LINEAR);
+                    insert_output_before(ctx, decl, TGSI_SEMANTIC_COLOR, 1,
+                                         TGSI_INTERPOLATE_LINEAR);
                     vsctx->color_used[1] = TRUE;
                 }
                 if (decl->Semantic.Index == 1 && !vsctx->bcolor_used[0]) {
-                    insert_output(ctx, decl, TGSI_SEMANTIC_BCOLOR, 0,
-                                  TGSI_INTERPOLATE_LINEAR);
+                    insert_output_before(ctx, decl, TGSI_SEMANTIC_BCOLOR, 0,
+                                         TGSI_INTERPOLATE_LINEAR);
                     vsctx->bcolor_used[0] = TRUE;
                 }
-                /* One more case is handled in insert_trailing_bcolor. */
                 break;
 
             case TGSI_SEMANTIC_GENERIC:
@@ -195,11 +194,6 @@ static void transform_decl(struct tgsi_transform_context *ctx,
                 break;
         }
 
-        if (decl->Semantic.Name != TGSI_SEMANTIC_BCOLOR) {
-            /* Insert it as soon as possible. */
-            insert_trailing_bcolor(ctx, decl);
-        }
-
         /* Since we're inserting new outputs in between, the following outputs
          * should be moved to the right so that they don't overlap with
          * the newly added ones. */
@@ -214,6 +208,14 @@ static void transform_decl(struct tgsi_transform_context *ctx,
     }
 
     ctx->emit_declaration(ctx, decl);
+
+    /* Insert BCOLOR1 if needed. */
+    if (decl->Declaration.File == TGSI_FILE_OUTPUT &&
+        decl->Semantic.Name == TGSI_SEMANTIC_BCOLOR &&
+        !vsctx->bcolor_used[1]) {
+        insert_output_after(ctx, decl, TGSI_SEMANTIC_BCOLOR, 1,
+                            TGSI_INTERPOLATE_LINEAR);
+    }
 }
 
 static void transform_inst(struct tgsi_transform_context *ctx,
@@ -226,10 +228,6 @@ static void transform_inst(struct tgsi_transform_context *ctx,
     if (!vsctx->first_instruction) {
         vsctx->first_instruction = TRUE;
 
-        /* The trailing BCOLOR should be inserted before the code
-         * if it hasn't already been done so. */
-        insert_trailing_bcolor(ctx, NULL);
-
         /* Insert the generic output for WPOS. */
         emit_output(ctx, TGSI_SEMANTIC_GENERIC, vsctx->last_generic + 1,
                     TGSI_INTERPOLATE_PERSPECTIVE, vsctx->num_outputs);
@@ -309,14 +307,18 @@ static void transform_inst(struct tgsi_transform_context *ctx,
     ctx->emit_instruction(ctx, inst);
 }
 
-void r300_draw_init_vertex_shader(struct draw_context *draw,
+void r300_draw_init_vertex_shader(struct r300_context *r300,
                                   struct r300_vertex_shader *vs)
 {
+    struct draw_context *draw = r300->draw;
     struct pipe_shader_state new_vs;
+    struct tgsi_shader_info info;
     struct vs_transform_context transform;
     const uint newLen = tgsi_num_tokens(vs->state.tokens) + 100 /* XXX */;
     unsigned i;
 
+    tgsi_scan_shader(vs->state.tokens, &info);
+
     new_vs.tokens = tgsi_alloc_tokens(newLen);
     if (new_vs.tokens == NULL)
         return;
@@ -329,6 +331,22 @@ void r300_draw_init_vertex_shader(struct draw_context *draw,
     transform.base.transform_instruction = transform_inst;
     transform.base.transform_declaration = transform_decl;
 
+    for (i = 0; i < info.num_outputs; i++) {
+        unsigned index = info.output_semantic_index[i];
+
+        switch (info.output_semantic_name[i]) {
+            case TGSI_SEMANTIC_COLOR:
+                assert(index < 2);
+                transform.color_used[index] = TRUE;
+                break;
+
+            case TGSI_SEMANTIC_BCOLOR:
+                assert(index < 2);
+                transform.bcolor_used[index] = TRUE;
+                break;
+        }
+    }
+
     tgsi_transform_shader(vs->state.tokens,
                           (struct tgsi_token*)new_vs.tokens,
                           newLen, &transform.base);
@@ -350,7 +368,7 @@ void r300_draw_init_vertex_shader(struct draw_context *draw,
     vs->state.tokens = new_vs.tokens;
 
     /* Init the VS output table for the rasterizer. */
-    r300_init_vs_outputs(vs);
+    r300_init_vs_outputs(r300, vs);
 
     /* Make the last generic be WPOS. */
     vs->outputs.wpos = vs->outputs.generic[transform.last_generic + 1];

src/gallium/drivers/r600/r600_shader.c

@@ -1277,6 +1277,14 @@ static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pi
 	if (ctx.bc->chip_class == CAYMAN)
 		cm_bytecode_add_cf_end(ctx.bc);
 
+	/* check GPR limit - we have 124 = 128 - 4
+	 * (4 are reserved as alu clause temporary registers) */
+	if (ctx.bc->ngpr > 124) {
+		R600_ERR("GPR limit exceeded - shader requires %d registers\n", ctx.bc->ngpr);
+		r = -ENOMEM;
+		goto out_err;
+	}
+
 	free(ctx.literals);
 	tgsi_parse_free(&ctx.parse);
 	return 0;

src/gallium/drivers/svga/include/svga_types.h

@@ -28,6 +28,7 @@
 
 #include "pipe/p_compiler.h"
 
+#ifndef __HAIKU__
 typedef int64_t int64;
 typedef uint64_t uint64;
 
@@ -39,6 +40,9 @@ typedef uint16_t uint16;
 
 typedef int8_t int8;
 typedef uint8_t uint8;
+#else
+#include <OS.h>
+#endif /* HAIKU */
 
 typedef uint8_t Bool;
 

src/gallium/drivers/svga/svga_pipe_blit.c

@@ -44,8 +44,7 @@ static void svga_surface_copy(struct pipe_context *pipe,
                               const struct pipe_box *src_box)
  {
    struct svga_context *svga = svga_context(pipe);
-   struct svga_texture *stex = svga_texture(src_tex);
-   struct svga_texture *dtex = svga_texture(dst_tex);
+   struct svga_texture *stex, *dtex;
 /*   struct pipe_screen *screen = pipe->screen;
    SVGA3dCopyBox *box;
    enum pipe_error ret;
@@ -63,6 +62,9 @@ static void svga_surface_copy(struct pipe_context *pipe,
       return;
    }
 
+   stex = svga_texture(src_tex);
+   dtex = svga_texture(dst_tex);
+
 #if 0
    srcsurf = screen->get_tex_surface(screen, src_tex,
                                      src_level, src_box->z, src_box->z,

src/gallium/drivers/svga/svga_pipe_depthstencil.c

@@ -57,10 +57,10 @@ svga_translate_stencil_op(unsigned op)
    case PIPE_STENCIL_OP_KEEP:      return SVGA3D_STENCILOP_KEEP;
    case PIPE_STENCIL_OP_ZERO:      return SVGA3D_STENCILOP_ZERO;
    case PIPE_STENCIL_OP_REPLACE:   return SVGA3D_STENCILOP_REPLACE;
-   case PIPE_STENCIL_OP_INCR:      return SVGA3D_STENCILOP_INCR;
-   case PIPE_STENCIL_OP_DECR:      return SVGA3D_STENCILOP_DECR;
-   case PIPE_STENCIL_OP_INCR_WRAP: return SVGA3D_STENCILOP_INCRSAT; /* incorrect? */
-   case PIPE_STENCIL_OP_DECR_WRAP: return SVGA3D_STENCILOP_DECRSAT; /* incorrect? */
+   case PIPE_STENCIL_OP_INCR:      return SVGA3D_STENCILOP_INCRSAT;
+   case PIPE_STENCIL_OP_DECR:      return SVGA3D_STENCILOP_DECRSAT;
+   case PIPE_STENCIL_OP_INCR_WRAP: return SVGA3D_STENCILOP_INCR;
+   case PIPE_STENCIL_OP_DECR_WRAP: return SVGA3D_STENCILOP_DECR;
    case PIPE_STENCIL_OP_INVERT:    return SVGA3D_STENCILOP_INVERT;
    default:
       assert(0);

src/gallium/drivers/svga/svga_pipe_draw.c

@@ -201,6 +201,17 @@ svga_release_user_upl_buffers(struct svga_context *svga)
       if (vb->buffer && svga_buffer_is_user_buffer(vb->buffer)) {
          struct svga_buffer *buffer = svga_buffer(vb->buffer);
 
+         /* The buffer_offset is relative to the uploaded buffer.
+          * Since we're discarding that buffer we need to reset this offset
+          * so it's not inadvertantly applied to a subsequent draw.
+          *
+          * XXX a root problem here is that the svga->curr.vb[] information
+          * is getting set both by gallium API calls and by code in
+          * svga_upload_user_buffers().  We should instead have two copies
+          * of the vertex buffer information and choose between as needed.
+          */
+         vb->buffer_offset = 0;
+
          buffer->uploaded.start = ~0;
          buffer->uploaded.end = 0;
          if (buffer->uploaded.buffer)

src/gallium/drivers/svga/svga_pipe_rasterizer.c

@@ -237,11 +237,11 @@ static void svga_bind_rasterizer_state( struct pipe_context *pipe,
    struct svga_context *svga = svga_context(pipe);
    struct svga_rasterizer_state *raster = (struct svga_rasterizer_state *)state;
 
-   svga->curr.rast = raster;
 
    draw_set_rasterizer_state(svga->swtnl.draw, raster ? &raster->templ : NULL,
                              state);
-   
+   svga->curr.rast = raster;
+
    svga->dirty |= SVGA_NEW_RAST;
 }
 

src/gallium/drivers/svga/svga_pipe_sampler.c

@@ -125,8 +125,8 @@ svga_create_sampler_state(struct pipe_context *pipe,
     *    - min/max LOD clamping
     */
    cso->min_lod = 0;
-   cso->view_min_lod = MAX2(sampler->min_lod, 0);
-   cso->view_max_lod = MAX2(sampler->max_lod, 0);
+   cso->view_min_lod = MAX2((int) (sampler->min_lod + 0.5), 0);
+   cso->view_max_lod = MAX2((int) (sampler->max_lod + 0.5), 0);
 
    /* Use min_mipmap */
    if (svga->debug.use_min_mipmap) {

src/gallium/drivers/svga/svga_screen.c

@@ -120,7 +120,7 @@ svga_get_paramf(struct pipe_screen *screen, enum pipe_capf param)
       return result.u;
 
    case PIPE_CAPF_MAX_TEXTURE_LOD_BIAS:
-      return 16.0;
+      return 15.0;
 
    default:
       return 0;
@@ -235,7 +235,7 @@ static int svga_get_shader_param(struct pipe_screen *screen, unsigned shader, en
       case PIPE_SHADER_CAP_MAX_TEMPS:
          if (!sws->get_cap(sws, SVGA3D_DEVCAP_MAX_FRAGMENT_SHADER_TEMPS, &result))
             return 32;
-         return result.u;
+         return MIN2(result.u, SVGA3D_TEMPREG_MAX);
       case PIPE_SHADER_CAP_MAX_ADDRS:
       case PIPE_SHADER_CAP_INDIRECT_INPUT_ADDR:
 	 /* 
@@ -286,7 +286,7 @@ static int svga_get_shader_param(struct pipe_screen *screen, unsigned shader, en
       case PIPE_SHADER_CAP_MAX_TEMPS:
          if (!sws->get_cap(sws, SVGA3D_DEVCAP_MAX_VERTEX_SHADER_TEMPS, &result))
             return 32;
-         return result.u;
+         return MIN2(result.u, SVGA3D_TEMPREG_MAX);
       case PIPE_SHADER_CAP_MAX_ADDRS:
          return 1;
       case PIPE_SHADER_CAP_MAX_PREDS:

src/gallium/drivers/svga/svga_state_need_swtnl.c

@@ -136,7 +136,7 @@ update_need_pipeline( struct svga_context *svga,
 
    /* EDGEFLAGS
     */
-    if (vs->base.info.writes_edgeflag) {
+    if (vs && vs->base.info.writes_edgeflag) {
       SVGA_DBG(DEBUG_SWTNL, "%s: edgeflags\n", __FUNCTION__);
       need_pipeline = TRUE;
    }
@@ -145,7 +145,8 @@ update_need_pipeline( struct svga_context *svga,
     */
    if (svga->curr.reduced_prim == PIPE_PRIM_POINTS) {
       unsigned sprite_coord_gen = svga->curr.rast->templ.sprite_coord_enable;
-      unsigned generic_inputs = svga->curr.fs->generic_inputs;
+      unsigned generic_inputs =
+         svga->curr.fs ? svga->curr.fs->generic_inputs : 0;
 
       if (sprite_coord_gen &&
           (generic_inputs & ~sprite_coord_gen)) {

src/gallium/drivers/svga/svga_state_tss.c

@@ -82,12 +82,14 @@ update_tss_binding(struct svga_context *svga,
       const struct svga_sampler_state *s = svga->curr.sampler[i];
       struct svga_hw_view_state *view = &svga->state.hw_draw.views[i];
       struct pipe_resource *texture = NULL;
+      struct pipe_sampler_view *sv = svga->curr.sampler_views[i];
 
       /* get min max lod */
-      if (svga->curr.sampler_views[i]) {
-         min_lod = MAX2(s->view_min_lod, 0);
-         max_lod = MIN2(s->view_max_lod, svga->curr.sampler_views[i]->texture->last_level);
-         texture = svga->curr.sampler_views[i]->texture;
+      if (sv) {
+         min_lod = MAX2(0, (s->view_min_lod + sv->u.tex.first_level));
+         max_lod = MIN2(s->view_max_lod, sv->texture->last_level);
+         max_lod += sv->u.tex.first_level;
+         texture = sv->texture;
       } else {
          min_lod = 0;
          max_lod = 0;

src/gallium/drivers/svga/svga_surface.c

@@ -320,6 +320,11 @@ svga_mark_surface_dirty(struct pipe_surface *surf)
       else {
          /* this will happen later in svga_propagate_surface */
       }
+
+      /* Increment the view_age and texture age for this surface's slice
+       * so that any sampler views into the texture are re-validated too.
+       */
+      tex->view_age[surf->u.tex.first_layer] = ++(tex->age);
    }
 }
 

src/gallium/state_trackers/dri/sw/drisw.c

@@ -253,6 +253,7 @@ drisw_update_tex_buffer(struct dri_drawable *drawable,
    char *map;
    int x, y, w, h;
    int ximage_stride, line;
+   int cpp = util_format_get_blocksize(res->format);
 
    get_drawable_info(dPriv, &x, &y, &w, &h);
 
@@ -265,9 +266,8 @@ drisw_update_tex_buffer(struct dri_drawable *drawable,
    /* Copy the Drawable content to the mapped texture buffer */
    get_image(dPriv, x, y, w, h, map);
 
-   /* The pipe transfer has a pitch rounded up to the nearest 64 pixels.
-      We assume 32 bit pixels. */
-   ximage_stride = w * 4;
+   /* The pipe transfer has a pitch rounded up to the nearest 64 pixels. */
+   ximage_stride = w * cpp;
    for (line = h-1; line; --line) {
       memmove(&map[line * transfer->stride],
               &map[line * ximage_stride],

src/gallium/state_trackers/vega/text.c

@@ -73,8 +73,8 @@ static void add_glyph(struct vg_font *font,
    glyph = CALLOC_STRUCT(vg_glyph);
    glyph->object = obj;
    glyph->is_hinted = isHinted;
-   memcpy(glyph->glyph_origin, glyphOrigin, sizeof(glyphOrigin));
-   memcpy(glyph->escapement, escapement, sizeof(escapement));
+   memcpy(glyph->glyph_origin, glyphOrigin, sizeof(glyph->glyph_origin));
+   memcpy(glyph->escapement, escapement, sizeof(glyph->glyph_origin));
 
    cso_hash_insert(font->glyphs, (unsigned) glyphIndex, glyph);
 }

src/gallium/state_trackers/vega/vg_translate.c

@@ -526,8 +526,8 @@ void _vega_unpack_float_span_rgba(struct vg_context *ctx,
       src += offset;
       for (i = 0; i < n; ++i) {
          VGfloat clr[4];
-         clr[0] = ((*src >> 10) & 31)/31.;
-         clr[1] = ((*src >>  5) & 95)/95.;
+         clr[0] = ((*src >> 11) & 31)/31.;
+         clr[1] = ((*src >>  5) & 63)/63.;
          clr[2] = ((*src >>  0) & 31)/31.;
          clr[3] = 1.f;
 

src/gallium/targets/Makefile.xorg

@@ -46,7 +46,7 @@ endif
 default: depend $(TOP)/$(LIB_DIR)/gallium $(LIBNAME) $(LIBNAME_STAGING)
 
 $(LIBNAME): $(OBJECTS) Makefile ../Makefile.xorg $(LIBS) $(DRIVER_PIPES) $(GALLIUM_AUXILIARIES)
-	$(MKLIB) -linker '$(LD)' -noprefix -o $@ $(LDFLAGS) $(OBJECTS) $(DRIVER_PIPES) $(GALLIUM_AUXILIARIES) $(DRIVER_LINKS)
+	$(MKLIB) -linker '$(LD)' -noprefix -o $@ -ldflags '$(LDFLAGS)' $(OBJECTS) $(DRIVER_PIPES) $(GALLIUM_AUXILIARIES) $(DRIVER_LINKS)
 
 depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) $(GENERATED_SOURCES)
 	rm -f depend

src/gallium/targets/egl-static/egl_st.c

@@ -54,8 +54,9 @@ dlopen_gl_lib_cb(const char *dir, size_t len, void *callback_data)
    int ret;
 
    if (len) {
+      assert(len <= INT_MAX && "path is insanely long!");
       ret = util_snprintf(path, sizeof(path), "%.*s/%s" UTIL_DL_EXT,
-            len, dir, name);
+            (int)len, dir, name);
    }
    else {
       ret = util_snprintf(path, sizeof(path), "%s" UTIL_DL_EXT, name);

src/glsl/Android.mk

@@ -39,6 +39,7 @@ LOCAL_SRC_FILES := \
 	$(LIBGLSL_CXX_SOURCES)
 
 LOCAL_C_INCLUDES := \
+	external/astl/include \
 	$(MESA_TOP)/src/mapi \
 	$(MESA_TOP)/src/mesa
 

src/glsl/ast_to_hir.cpp

@@ -3516,11 +3516,9 @@ ast_switch_statement::hir(exec_list *instructions,
     *
     *    "The type of init-expression in a switch statement must be a 
     *     scalar integer." 
-    *
-    * The checks are separated so that higher quality diagnostics can be
-    * generated for cases where the rule is violated.
     */
-   if (!test_expression->type->is_integer()) {
+   if (!test_expression->type->is_scalar() ||
+       !test_expression->type->is_integer()) {
       YYLTYPE loc = this->test_expression->get_location();
 
       _mesa_glsl_error(& loc,
@@ -3571,7 +3569,7 @@ ast_switch_statement::hir(exec_list *instructions,
    /* Cache test expression.
     */
    test_to_hir(instructions, state);
-   
+
    /* Emit code for body of switch stmt.
     */
    body->hir(instructions, state);
@@ -3580,290 +3578,265 @@ ast_switch_statement::hir(exec_list *instructions,
 
    state->switch_state = saved;
 
-     /* Switch statements do not have r-values.
-      */
-     return NULL;
-  }
+   /* Switch statements do not have r-values. */
+   return NULL;
+}
 
 
-  void
-  ast_switch_statement::test_to_hir(exec_list *instructions,
-				    struct _mesa_glsl_parse_state *state)
-  {
-     void *ctx = state;
+void
+ast_switch_statement::test_to_hir(exec_list *instructions,
+				  struct _mesa_glsl_parse_state *state)
+{
+   void *ctx = state;
 
-     /* Cache value of test expression.
-      */
-     ir_rvalue *const test_val =
-	test_expression->hir(instructions,
-			     state);
+   /* Cache value of test expression. */
+   ir_rvalue *const test_val =
+      test_expression->hir(instructions,
+			   state);
 
-     state->switch_state.test_var = new(ctx) ir_variable(glsl_type::int_type,
-							 "switch_test_tmp",
-							 ir_var_temporary);
-     ir_dereference_variable *deref_test_var =
-	new(ctx) ir_dereference_variable(state->switch_state.test_var);
+   state->switch_state.test_var = new(ctx) ir_variable(test_val->type,
+						       "switch_test_tmp",
+						       ir_var_temporary);
+   ir_dereference_variable *deref_test_var =
+      new(ctx) ir_dereference_variable(state->switch_state.test_var);
 
-     instructions->push_tail(state->switch_state.test_var);
-     instructions->push_tail(new(ctx) ir_assignment(deref_test_var,
-						    test_val,
-						    NULL));
-  }
+   instructions->push_tail(state->switch_state.test_var);
+   instructions->push_tail(new(ctx) ir_assignment(deref_test_var, test_val,
+						  NULL));
+}
 
 
-  ir_rvalue *
-  ast_switch_body::hir(exec_list *instructions,
-		       struct _mesa_glsl_parse_state *state)
-  {
-     if (stmts != NULL)
-	stmts->hir(instructions, state);
+ir_rvalue *
+ast_switch_body::hir(exec_list *instructions,
+		     struct _mesa_glsl_parse_state *state)
+{
+   if (stmts != NULL)
+      stmts->hir(instructions, state);
 
-     /* Switch bodies do not have r-values.
-      */
-     return NULL;
-  }
+   /* Switch bodies do not have r-values. */
+   return NULL;
+}
+
+ir_rvalue *
+ast_case_statement_list::hir(exec_list *instructions,
+			     struct _mesa_glsl_parse_state *state)
+{
+   foreach_list_typed (ast_case_statement, case_stmt, link, & this->cases)
+      case_stmt->hir(instructions, state);
+
+   /* Case statements do not have r-values. */
+   return NULL;
+}
+
+ir_rvalue *
+ast_case_statement::hir(exec_list *instructions,
+			struct _mesa_glsl_parse_state *state)
+{
+   labels->hir(instructions, state);
+
+   /* Conditionally set fallthru state based on break state. */
+   ir_constant *const false_val = new(state) ir_constant(false);
+   ir_dereference_variable *const deref_is_fallthru_var =
+      new(state) ir_dereference_variable(state->switch_state.is_fallthru_var);
+   ir_dereference_variable *const deref_is_break_var =
+      new(state) ir_dereference_variable(state->switch_state.is_break_var);
+   ir_assignment *const reset_fallthru_on_break =
+      new(state) ir_assignment(deref_is_fallthru_var,
+			       false_val,
+			       deref_is_break_var);
+   instructions->push_tail(reset_fallthru_on_break);
+
+   /* Guard case statements depending on fallthru state. */
+   ir_dereference_variable *const deref_fallthru_guard =
+      new(state) ir_dereference_variable(state->switch_state.is_fallthru_var);
+   ir_if *const test_fallthru = new(state) ir_if(deref_fallthru_guard);
+
+   foreach_list_typed (ast_node, stmt, link, & this->stmts)
+      stmt->hir(& test_fallthru->then_instructions, state);
+
+   instructions->push_tail(test_fallthru);
+
+   /* Case statements do not have r-values. */
+   return NULL;
+}
 
 
-  ir_rvalue *
-  ast_case_statement_list::hir(exec_list *instructions,
-			       struct _mesa_glsl_parse_state *state)
-  {
-     foreach_list_typed (ast_case_statement, case_stmt, link, & this->cases)
-	case_stmt->hir(instructions, state);
+ir_rvalue *
+ast_case_label_list::hir(exec_list *instructions,
+			 struct _mesa_glsl_parse_state *state)
+{
+   foreach_list_typed (ast_case_label, label, link, & this->labels)
+      label->hir(instructions, state);
 
-     /* Case statements do not have r-values.
-      */
-     return NULL;
-  }
+   /* Case labels do not have r-values. */
+   return NULL;
+}
 
+ir_rvalue *
+ast_case_label::hir(exec_list *instructions,
+		    struct _mesa_glsl_parse_state *state)
+{
+   void *ctx = state;
 
-  ir_rvalue *
-  ast_case_statement::hir(exec_list *instructions,
-			  struct _mesa_glsl_parse_state *state)
-  {
-     labels->hir(instructions, state);
+   ir_dereference_variable *deref_fallthru_var =
+      new(ctx) ir_dereference_variable(state->switch_state.is_fallthru_var);
 
-     /* Conditionally set fallthru state based on break state.
-      */
-     ir_constant *const false_val = new(state) ir_constant(false);
-     ir_dereference_variable *const deref_is_fallthru_var =
-	new(state) ir_dereference_variable(state->switch_state.is_fallthru_var);
-     ir_dereference_variable *const deref_is_break_var =
-	new(state) ir_dereference_variable(state->switch_state.is_break_var);
-     ir_assignment *const reset_fallthru_on_break =
-	new(state) ir_assignment(deref_is_fallthru_var,
-				 false_val,
-				 deref_is_break_var);
-     instructions->push_tail(reset_fallthru_on_break);
+   ir_rvalue *const true_val = new(ctx) ir_constant(true);
 
-     /* Guard case statements depending on fallthru state.
-      */
-     ir_dereference_variable *const deref_fallthru_guard =
-	new(state) ir_dereference_variable(state->switch_state.is_fallthru_var);
-     ir_if *const test_fallthru = new(state) ir_if(deref_fallthru_guard);
+   /* If not default case, ... */
+   if (this->test_value != NULL) {
+      /* Conditionally set fallthru state based on
+       * comparison of cached test expression value to case label.
+       */
+      ir_rvalue *const label_rval = this->test_value->hir(instructions, state);
+      ir_constant *label_const = label_rval->constant_expression_value();
 
-     foreach_list_typed (ast_node, stmt, link, & this->stmts)
-	stmt->hir(& test_fallthru->then_instructions, state);
+      if (!label_const) {
+	 YYLTYPE loc = this->test_value->get_location();
 
-     instructions->push_tail(test_fallthru);
+	 _mesa_glsl_error(& loc, state,
+			  "switch statement case label must be a "
+			  "constant expression");
 
-     /* Case statements do not have r-values.
-      */
-     return NULL;
-  }
+	 /* Stuff a dummy value in to allow processing to continue. */
+	 label_const = new(ctx) ir_constant(0);
+      } else {
+	 ast_expression *previous_label = (ast_expression *)
+	    hash_table_find(state->switch_state.labels_ht,
+			    (void *)(uintptr_t)label_const->value.u[0]);
 
+	 if (previous_label) {
+	    YYLTYPE loc = this->test_value->get_location();
+	    _mesa_glsl_error(& loc, state,
+			     "duplicate case value");
 
-  ir_rvalue *
-  ast_case_label_list::hir(exec_list *instructions,
-			   struct _mesa_glsl_parse_state *state)
-  {
-     foreach_list_typed (ast_case_label, label, link, & this->labels)
-	label->hir(instructions, state);
-
-     /* Case labels do not have r-values.
-      */
-     return NULL;
-  }
-
-
-  ir_rvalue *
-  ast_case_label::hir(exec_list *instructions,
-		      struct _mesa_glsl_parse_state *state)
-  {
-     void *ctx = state;
-
-     ir_dereference_variable *deref_fallthru_var =
-	new(ctx) ir_dereference_variable(state->switch_state.is_fallthru_var);
-
-     ir_rvalue *const true_val = new(ctx) ir_constant(true);
-
-     /* If not default case, ...
-      */
-     if (this->test_value != NULL) {
-	/* Conditionally set fallthru state based on
-	 * comparison of cached test expression value to case label.
-	 */
-	ir_rvalue *const label_rval = this->test_value->hir(instructions, state);
-	ir_constant *label_const = label_rval->constant_expression_value();
-
-	if (!label_const) {
-	   YYLTYPE loc = this->test_value->get_location();
-
-	   _mesa_glsl_error(& loc, state,
-			    "switch statement case label must be a "
-			    "constant expression");
-
-	   /* Stuff a dummy value in to allow processing to continue. */
-	   label_const = new(ctx) ir_constant(0);
-	} else {
-	   ast_expression *previous_label = (ast_expression *)
-	      hash_table_find(state->switch_state.labels_ht,
+	    loc = previous_label->get_location();
+	    _mesa_glsl_error(& loc, state,
+			     "this is the previous case label");
+	 } else {
+	    hash_table_insert(state->switch_state.labels_ht,
+			      this->test_value,
 			      (void *)(uintptr_t)label_const->value.u[0]);
+	 }
+      }
 
-	   if (previous_label) {
-	      YYLTYPE loc = this->test_value->get_location();
-	      _mesa_glsl_error(& loc, state,
-			       "duplicate case value");
+      ir_dereference_variable *deref_test_var =
+	 new(ctx) ir_dereference_variable(state->switch_state.test_var);
 
-	      loc = previous_label->get_location();
-	      _mesa_glsl_error(& loc, state,
-			       "this is the previous case label");
-	   } else {
-	      hash_table_insert(state->switch_state.labels_ht,
-				this->test_value,
-				(void *)(uintptr_t)label_const->value.u[0]);
-	   }
-	}
+      ir_rvalue *const test_cond = new(ctx) ir_expression(ir_binop_all_equal,
+							  label_const,
+							  deref_test_var);
 
-	ir_dereference_variable *deref_test_var =
-	   new(ctx) ir_dereference_variable(state->switch_state.test_var);
+      ir_assignment *set_fallthru_on_test =
+	 new(ctx) ir_assignment(deref_fallthru_var,
+				true_val,
+				test_cond);
 
-	ir_rvalue *const test_cond = new(ctx) ir_expression(ir_binop_all_equal,
-							    glsl_type::bool_type,
-							    label_const,
-							    deref_test_var);
+      instructions->push_tail(set_fallthru_on_test);
+   } else { /* default case */
+      if (state->switch_state.previous_default) {
+	 YYLTYPE loc = this->get_location();
+	 _mesa_glsl_error(& loc, state,
+			  "multiple default labels in one switch");
 
-	ir_assignment *set_fallthru_on_test =
-	   new(ctx) ir_assignment(deref_fallthru_var,
-				  true_val,
-				  test_cond);
+	 loc = state->switch_state.previous_default->get_location();
+	 _mesa_glsl_error(& loc, state,
+			  "this is the first default label");
+      }
+      state->switch_state.previous_default = this;
 
-	instructions->push_tail(set_fallthru_on_test);
-     } else { /* default case */
-	if (state->switch_state.previous_default) {
-	   printf("a\n");
-	   YYLTYPE loc = this->get_location();
-	   _mesa_glsl_error(& loc, state,
-			       "multiple default labels in one switch");
+      /* Set falltrhu state. */
+      ir_assignment *set_fallthru =
+	 new(ctx) ir_assignment(deref_fallthru_var, true_val, NULL);
 
-	   printf("b\n");
+      instructions->push_tail(set_fallthru);
+   }
 
-	   loc = state->switch_state.previous_default->get_location();
-	   _mesa_glsl_error(& loc, state,
-			    "this is the first default label");
-	}
-	state->switch_state.previous_default = this;
+   /* Case statements do not have r-values. */
+   return NULL;
+}
 
-	/* Set falltrhu state.
-	 */
-	ir_assignment *set_fallthru =
-	   new(ctx) ir_assignment(deref_fallthru_var,
-				  true_val,
-				  NULL);
+void
+ast_iteration_statement::condition_to_hir(ir_loop *stmt,
+					  struct _mesa_glsl_parse_state *state)
+{
+   void *ctx = state;
 
-	instructions->push_tail(set_fallthru);
-     }
+   if (condition != NULL) {
+      ir_rvalue *const cond =
+	 condition->hir(& stmt->body_instructions, state);
 
-     /* Case statements do not have r-values.
-      */
-     return NULL;
-  }
+      if ((cond == NULL)
+	  || !cond->type->is_boolean() || !cond->type->is_scalar()) {
+	 YYLTYPE loc = condition->get_location();
+
+	 _mesa_glsl_error(& loc, state,
+			  "loop condition must be scalar boolean");
+      } else {
+	 /* As the first code in the loop body, generate a block that looks
+	  * like 'if (!condition) break;' as the loop termination condition.
+	  */
+	 ir_rvalue *const not_cond =
+	    new(ctx) ir_expression(ir_unop_logic_not, cond);
+
+	 ir_if *const if_stmt = new(ctx) ir_if(not_cond);
+
+	 ir_jump *const break_stmt =
+	    new(ctx) ir_loop_jump(ir_loop_jump::jump_break);
+
+	 if_stmt->then_instructions.push_tail(break_stmt);
+	 stmt->body_instructions.push_tail(if_stmt);
+      }
+   }
+}
 
 
-  void
-  ast_iteration_statement::condition_to_hir(ir_loop *stmt,
-					    struct _mesa_glsl_parse_state *state)
-  {
-     void *ctx = state;
+ir_rvalue *
+ast_iteration_statement::hir(exec_list *instructions,
+			     struct _mesa_glsl_parse_state *state)
+{
+   void *ctx = state;
 
-     if (condition != NULL) {
-	ir_rvalue *const cond =
-	   condition->hir(& stmt->body_instructions, state);
+   /* For-loops and while-loops start a new scope, but do-while loops do not.
+    */
+   if (mode != ast_do_while)
+      state->symbols->push_scope();
 
-	if ((cond == NULL)
-	    || !cond->type->is_boolean() || !cond->type->is_scalar()) {
-	   YYLTYPE loc = condition->get_location();
+   if (init_statement != NULL)
+      init_statement->hir(instructions, state);
 
-	   _mesa_glsl_error(& loc, state,
-			    "loop condition must be scalar boolean");
-	} else {
-	   /* As the first code in the loop body, generate a block that looks
-	    * like 'if (!condition) break;' as the loop termination condition.
-	    */
-	   ir_rvalue *const not_cond =
-	      new(ctx) ir_expression(ir_unop_logic_not, glsl_type::bool_type, cond,
-				     NULL);
+   ir_loop *const stmt = new(ctx) ir_loop();
+   instructions->push_tail(stmt);
 
-	   ir_if *const if_stmt = new(ctx) ir_if(not_cond);
+   /* Track the current loop nesting. */
+   ast_iteration_statement *nesting_ast = state->loop_nesting_ast;
 
-	   ir_jump *const break_stmt =
-	      new(ctx) ir_loop_jump(ir_loop_jump::jump_break);
+   state->loop_nesting_ast = this;
 
-	   if_stmt->then_instructions.push_tail(break_stmt);
-	   stmt->body_instructions.push_tail(if_stmt);
-	}
-     }
-  }
+   /* Likewise, indicate that following code is closest to a loop,
+    * NOT closest to a switch.
+    */
+   bool saved_is_switch_innermost = state->switch_state.is_switch_innermost;
+   state->switch_state.is_switch_innermost = false;
 
+   if (mode != ast_do_while)
+      condition_to_hir(stmt, state);
 
-  ir_rvalue *
-  ast_iteration_statement::hir(exec_list *instructions,
-			       struct _mesa_glsl_parse_state *state)
-  {
-     void *ctx = state;
+   if (body != NULL)
+      body->hir(& stmt->body_instructions, state);
 
-     /* For-loops and while-loops start a new scope, but do-while loops do not.
-      */
-     if (mode != ast_do_while)
-	state->symbols->push_scope();
+   if (rest_expression != NULL)
+      rest_expression->hir(& stmt->body_instructions, state);
 
-     if (init_statement != NULL)
-	init_statement->hir(instructions, state);
+   if (mode == ast_do_while)
+      condition_to_hir(stmt, state);
 
-     ir_loop *const stmt = new(ctx) ir_loop();
-     instructions->push_tail(stmt);
+   if (mode != ast_do_while)
+      state->symbols->pop_scope();
 
-     /* Track the current loop nesting.
-      */
-     ast_iteration_statement *nesting_ast = state->loop_nesting_ast;
-
-     state->loop_nesting_ast = this;
-
-     /* Likewise, indicate that following code is closest to a loop,
-      * NOT closest to a switch.
-      */
-     bool saved_is_switch_innermost = state->switch_state.is_switch_innermost;
-     state->switch_state.is_switch_innermost = false;
-
-     if (mode != ast_do_while)
-	condition_to_hir(stmt, state);
-
-     if (body != NULL)
-	body->hir(& stmt->body_instructions, state);
-
-     if (rest_expression != NULL)
-	rest_expression->hir(& stmt->body_instructions, state);
-
-     if (mode == ast_do_while)
-	condition_to_hir(stmt, state);
-
-     if (mode != ast_do_while)
-	state->symbols->pop_scope();
-
-     /* Restore previous nesting before returning.
-      */
-     state->loop_nesting_ast = nesting_ast;
-     state->switch_state.is_switch_innermost = saved_is_switch_innermost;
+   /* Restore previous nesting before returning. */
+   state->loop_nesting_ast = nesting_ast;
+   state->switch_state.is_switch_innermost = saved_is_switch_innermost;
 
    /* Loops do not have r-values.
     */

src/glsl/builtins/ir/acos.ir

@@ -2,21 +2,21 @@
    (signature float
      (parameters
        (declare (in) float x))
-     ((return (expression float - (constant float (1.5707963))
+     ((return (expression float - (constant float (1.5707964))
                                   (call asin ((var_ref x)))))))
    (signature vec2
      (parameters
        (declare (in) vec2 x))
-     ((return (expression vec2 - (constant float (1.5707963))
+     ((return (expression vec2 - (constant float (1.5707964))
                                  (call asin ((var_ref x)))))))
    (signature vec3
      (parameters
        (declare (in) vec3 x))
-     ((return (expression vec3 - (constant float (1.5707963))
+     ((return (expression vec3 - (constant float (1.5707964))
                                  (call asin ((var_ref x)))))))
    (signature vec4
      (parameters
        (declare (in) vec4 x))
-     ((return (expression vec4 - (constant float (1.5707963))
+     ((return (expression vec4 - (constant float (1.5707964))
                                  (call asin ((var_ref x)))))))
 ))

src/glsl/builtins/tools/texture_builtins.py

@@ -54,7 +54,7 @@ def get_txs_dim(sampler_type):
 def generate_sigs(g, tex_inst, sampler_type, variant = 0, unused_fields = 0):
     coord_dim = get_coord_dim(sampler_type)
     extra_dim = get_extra_dim(sampler_type, variant & Proj, unused_fields)
-    offset_dim = get_sampler_dim(sampler_type)
+    sampler_dim = get_sampler_dim(sampler_type)
 
     if variant & Single:
         return_type = "float"
@@ -74,12 +74,12 @@ def generate_sigs(g, tex_inst, sampler_type, variant = 0, unused_fields = 0):
     elif tex_inst == "txf" or tex_inst == "txs":
         print "\n       (declare (in) int lod)",
     elif tex_inst == "txd":
-        grad_type = vec_type("", coord_dim)
+        grad_type = vec_type("", sampler_dim)
         print "\n       (declare (in) " + grad_type + " dPdx)",
         print "\n       (declare (in) " + grad_type + " dPdy)",
 
     if variant & Offset:
-        print "\n       (declare (const_in) " + vec_type("i", offset_dim) + " offset)",
+        print "\n       (declare (const_in) " + vec_type("i", sampler_dim) + " offset)",
     if tex_inst == "txb":
         print "\n       (declare (in) float bias)",
 

src/glsl/ir.h

@@ -1158,7 +1158,6 @@ public:
    {
       this->ir_type = ir_type_loop_jump;
       this->mode = mode;
-      this->loop = loop;
    }
 
    virtual ir_loop_jump *clone(void *mem_ctx, struct hash_table *) const;
@@ -1182,9 +1181,6 @@ public:
 
    /** Mode selector for the jump instruction. */
    enum jump_mode mode;
-private:
-   /** Loop containing this break instruction. */
-   ir_loop *loop;
 };
 
 /**

src/glsl/ir_constant_expression.cpp

@@ -633,13 +633,13 @@ ir_expression::constant_expression_value()
       for (unsigned c = 0; c < op[0]->type->components(); c++) {
 	 switch (op[0]->type->base_type) {
 	 case GLSL_TYPE_UINT:
-	    data.b[0] = op[0]->value.u[0] < op[1]->value.u[0];
+	    data.b[c] = op[0]->value.u[c] < op[1]->value.u[c];
 	    break;
 	 case GLSL_TYPE_INT:
-	    data.b[0] = op[0]->value.i[0] < op[1]->value.i[0];
+	    data.b[c] = op[0]->value.i[c] < op[1]->value.i[c];
 	    break;
 	 case GLSL_TYPE_FLOAT:
-	    data.b[0] = op[0]->value.f[0] < op[1]->value.f[0];
+	    data.b[c] = op[0]->value.f[c] < op[1]->value.f[c];
 	    break;
 	 default:
 	    assert(0);
@@ -669,13 +669,13 @@ ir_expression::constant_expression_value()
       for (unsigned c = 0; c < op[0]->type->components(); c++) {
 	 switch (op[0]->type->base_type) {
 	 case GLSL_TYPE_UINT:
-	    data.b[0] = op[0]->value.u[0] <= op[1]->value.u[0];
+	    data.b[c] = op[0]->value.u[c] <= op[1]->value.u[c];
 	    break;
 	 case GLSL_TYPE_INT:
-	    data.b[0] = op[0]->value.i[0] <= op[1]->value.i[0];
+	    data.b[c] = op[0]->value.i[c] <= op[1]->value.i[c];
 	    break;
 	 case GLSL_TYPE_FLOAT:
-	    data.b[0] = op[0]->value.f[0] <= op[1]->value.f[0];
+	    data.b[c] = op[0]->value.f[c] <= op[1]->value.f[c];
 	    break;
 	 default:
 	    assert(0);
@@ -687,13 +687,13 @@ ir_expression::constant_expression_value()
       for (unsigned c = 0; c < op[0]->type->components(); c++) {
 	 switch (op[0]->type->base_type) {
 	 case GLSL_TYPE_UINT:
-	    data.b[0] = op[0]->value.u[0] >= op[1]->value.u[0];
+	    data.b[c] = op[0]->value.u[c] >= op[1]->value.u[c];
 	    break;
 	 case GLSL_TYPE_INT:
-	    data.b[0] = op[0]->value.i[0] >= op[1]->value.i[0];
+	    data.b[c] = op[0]->value.i[c] >= op[1]->value.i[c];
 	    break;
 	 case GLSL_TYPE_FLOAT:
-	    data.b[0] = op[0]->value.f[0] >= op[1]->value.f[0];
+	    data.b[c] = op[0]->value.f[c] >= op[1]->value.f[c];
 	    break;
 	 default:
 	    assert(0);

src/glsl/link_uniforms.cpp

@@ -174,8 +174,7 @@ private:
       if (this->map->get(id, name))
 	 return;
 
-      char *key = strdup(name);
-      this->map->put(this->num_active_uniforms, key);
+      this->map->put(this->num_active_uniforms, name);
 
       /* Each leaf uniform occupies one entry in the list of active
        * uniforms.

src/glsl/linker.cpp

@@ -2263,7 +2263,9 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
       if (ctx->ShaderCompilerOptions[i].LowerClipDistance)
          lower_clip_distance(prog->_LinkedShaders[i]->ir);
 
-      while (do_common_optimization(prog->_LinkedShaders[i]->ir, true, false, 32))
+      unsigned max_unroll = ctx->ShaderCompilerOptions[i].MaxUnrollIterations;
+
+      while (do_common_optimization(prog->_LinkedShaders[i]->ir, true, false, max_unroll))
 	 ;
    }
 

src/glsl/loop_analysis.cpp

@@ -110,6 +110,8 @@ public:
    virtual ir_visitor_status visit(ir_loop_jump *);
    virtual ir_visitor_status visit(ir_dereference_variable *);
 
+   virtual ir_visitor_status visit_enter(ir_call *);
+
    virtual ir_visitor_status visit_enter(ir_loop *);
    virtual ir_visitor_status visit_leave(ir_loop *);
    virtual ir_visitor_status visit_enter(ir_assignment *);
@@ -152,6 +154,21 @@ loop_analysis::visit(ir_loop_jump *ir)
 }
 
 
+ir_visitor_status
+loop_analysis::visit_enter(ir_call *ir)
+{
+   /* If we're not somewhere inside a loop, there's nothing to do. */
+   if (this->state.is_empty())
+      return visit_continue;
+
+   loop_variable_state *const ls =
+      (loop_variable_state *) this->state.get_head();
+
+   ls->contains_calls = true;
+   return visit_continue_with_parent;
+}
+
+
 ir_visitor_status
 loop_analysis::visit(ir_dereference_variable *ir)
 {
@@ -209,6 +226,17 @@ loop_analysis::visit_leave(ir_loop *ir)
    loop_variable_state *const ls =
       (loop_variable_state *) this->state.pop_head();
 
+   /* Function calls may contain side effects.  These could alter any of our
+    * variables in ways that cannot be known, and may even terminate shader
+    * execution (say, calling discard in the fragment shader).  So we can't
+    * rely on any of our analysis about assignments to variables.
+    *
+    * We could perform some conservative analysis (prove there's no statically
+    * possible assignment, etc.) but it isn't worth it for now; function
+    * inlining will allow us to unroll loops anyway.
+    */
+   if (ls->contains_calls)
+      return visit_continue;
 
    foreach_list(node, &ir->body_instructions) {
       /* Skip over declarations at the start of a loop.

src/glsl/loop_analysis.h

@@ -122,10 +122,16 @@ public:
     */
    unsigned num_loop_jumps;
 
+   /**
+    * Whether this loop contains any function calls.
+    */
+   bool contains_calls;
+
    loop_variable_state()
    {
       this->max_iterations = -1;
       this->num_loop_jumps = 0;
+      this->contains_calls = false;
       this->var_hash = hash_table_ctor(0, hash_table_pointer_hash,
 				       hash_table_pointer_compare);
    }
@@ -134,6 +140,23 @@ public:
    {
       hash_table_dtor(this->var_hash);
    }
+
+   static void* operator new(size_t size, void *ctx)
+   {
+      void *lvs = ralloc_size(ctx, size);
+      assert(lvs != NULL);
+
+      ralloc_set_destructor(lvs, (void (*)(void*)) destructor);
+
+      return lvs;
+   }
+
+private:
+   static void
+   destructor(loop_variable_state *lvs)
+   {
+      lvs->~loop_variable_state();
+   }
 };
 
 

src/glsl/loop_unroll.cpp

@@ -56,6 +56,7 @@ loop_unroll_visitor::visit_leave(ir_loop *ir)
 {
    loop_variable_state *const ls = this->state->get(ir);
    int iterations;
+   unsigned ir_count;
 
    /* If we've entered a loop that hasn't been analyzed, something really,
     * really bad has happened.
@@ -78,6 +79,20 @@ loop_unroll_visitor::visit_leave(ir_loop *ir)
    if (iterations > (int) max_iterations)
       return visit_continue;
 
+   /* Don't try to unroll nested loops and loops with a huge body.
+    */
+   ir_count = 0;
+   foreach_list(node, &ir->body_instructions) {
+      ++ir_count;
+
+      /* If the loop body gets to huge, do not unroll. */
+      if (5*max_iterations < ir_count*iterations)
+          return visit_continue;
+      /* Do not unroll loops with child loop nodes. */
+      if (((ir_instruction *) node)->as_loop())
+          return visit_continue;
+   }
+
    if (ls->num_loop_jumps > 1)
       return visit_continue;
    else if (ls->num_loop_jumps) {

src/glsl/ralloc.c

@@ -278,7 +278,7 @@ ralloc_parent(const void *ptr)
       return NULL;
 
    info = get_header(ptr);
-   return PTR_FROM_HEADER(info->parent);
+   return info->parent ? PTR_FROM_HEADER(info->parent) : NULL;
 }
 
 static void *autofree_context = NULL;

src/glsl/strtod.c

@@ -44,7 +44,8 @@
 double
 glsl_strtod(const char *s, char **end)
 {
-#if defined(_GNU_SOURCE) && !defined(__CYGWIN__) && !defined(__FreeBSD__)
+#if defined(_GNU_SOURCE) && !defined(__CYGWIN__) && !defined(__FreeBSD__) && \
+   !defined(__HAIKU__)
    static locale_t loc = NULL;
    if (!loc) {
       loc = newlocale(LC_CTYPE_MASK, "C", NULL);

src/glu/sgi/libnurbs/internals/mapdesc.cc

@@ -90,7 +90,7 @@ Mapdesc::setBboxsize( INREAL *mat )
 void
 Mapdesc::identify( REAL dest[MAXCOORDS][MAXCOORDS] )
 {
-    memset( dest, 0, sizeof( dest ) );
+    memset( dest, 0, sizeof( REAL ) * MAXCOORDS * MAXCOORDS );
     for( int i=0; i != hcoords; i++ )
 	dest[i][i] = 1.0;
 }

src/glu/sgi/libnurbs/internals/nurbstess.cc

@@ -505,7 +505,7 @@ NurbsTessellator::do_pwlcurve( O_pwlcurve *o_pwlcurve )
 	o_pwlcurve->owner = currentCurve;
     }
 
-    if( (inCurve == 2) ) 
+    if( inCurve == 2 ) 
 	endcurve();
 }
 

src/glx/apple/Makefile

@@ -26,6 +26,7 @@ SOURCES = \
 	apple_glx.c \
 	apple_glx_context.c \
 	apple_glx_drawable.c \
+	apple_glx_log.c \
 	apple_glx_pbuffer.c \
 	apple_glx_pixmap.c \
 	apple_glx_surface.c \
@@ -35,6 +36,7 @@ SOURCES = \
 	apple_xgl_api_stereo.c \
 	apple_xgl_api_viewport.c \
 	appledri.c \
+	../create_context.c \
 	../clientattrib.c \
 	../compsize.c \
 	../glxconfig.c \

src/glx/apple/apple_glx.c

@@ -33,6 +33,8 @@
 #include <assert.h>
 #include <stdarg.h>
 #include <dlfcn.h>
+#include <pthread.h>
+#include <inttypes.h>
 #include "appledri.h"
 #include "apple_glx.h"
 #include "apple_glx_context.h"
@@ -43,22 +45,6 @@ static int dri_event_base = 0;
 
 const GLuint __glXDefaultPixelStore[9] = { 0, 0, 0, 0, 0, 0, 0, 0, 1 };
 
-static bool diagnostic = false;
-
-void
-apple_glx_diagnostic(const char *fmt, ...)
-{
-   va_list vl;
-
-   if (diagnostic) {
-      fprintf(stderr, "DIAG: ");
-
-      va_start(vl, fmt);
-      vfprintf(stderr, fmt, vl);
-      va_end(vl);
-   }
-}
-
 int
 apple_get_dri_event_base(void)
 {
@@ -125,10 +111,9 @@ apple_init_glx(Display * dpy)
    if (initialized)
       return false;
 
-   if (getenv("LIBGL_DIAGNOSTIC")) {
-      printf("initializing libGL in %s\n", __func__);
-      diagnostic = true;
-   }
+   apple_glx_log_init();
+
+   apple_glx_log(ASL_LEVEL_INFO, "Initializing libGL.");
 
    apple_cgl_init();
    (void) apple_glx_get_client_id();

src/glx/apple/apple_glx.h

@@ -38,7 +38,8 @@
 #define XP_NO_X_HEADERS
 #include <Xplugin.h>
 
-void apple_glx_diagnostic(const char *fmt, ...);
+#include "apple_glx_log.h"
+
 xp_client_id apple_glx_get_client_id(void);
 bool apple_init_glx(Display * dpy);
 void apple_glx_swap_buffers(void *ptr);

src/glx/apple/apple_glx_context.c

@@ -421,7 +421,7 @@ apple_glx_make_current_context(Display * dpy, void *oldptr, void *ptr,
     */
 
    if (same_drawable && ac->is_current) {
-      apple_glx_diagnostic("%s: same_drawable and ac->is_current\n");
+      apple_glx_diagnostic("same_drawable and ac->is_current\n");
       return false;
    }
 

src/glx/apple/apple_glx_drawable.c

@@ -32,6 +32,7 @@
 #include <stdlib.h>
 #include <assert.h>
 #include <pthread.h>
+#include <string.h>
 #include "apple_glx.h"
 #include "apple_glx_context.h"
 #include "apple_glx_drawable.h"
@@ -48,8 +49,8 @@ lock_drawables_list(void)
    err = pthread_mutex_lock(&drawables_lock);
 
    if (err) {
-      fprintf(stderr, "pthread_mutex_lock failure in %s: %d\n",
-              __func__, err);
+      fprintf(stderr, "pthread_mutex_lock failure in %s: %s\n",
+              __func__, strerror(err));
       abort();
    }
 }
@@ -62,8 +63,8 @@ unlock_drawables_list(void)
    err = pthread_mutex_unlock(&drawables_lock);
 
    if (err) {
-      fprintf(stderr, "pthread_mutex_unlock failure in %s: %d\n",
-              __func__, err);
+      fprintf(stderr, "pthread_mutex_unlock failure in %s: %s\n",
+              __func__, strerror(err));
       abort();
    }
 }
@@ -95,7 +96,7 @@ drawable_lock(struct apple_glx_drawable *agd)
    err = pthread_mutex_lock(&agd->mutex);
 
    if (err) {
-      fprintf(stderr, "pthread_mutex_lock error: %d\n", err);
+      fprintf(stderr, "pthread_mutex_lock error: %s\n", strerror(err));
       abort();
    }
 }
@@ -108,7 +109,7 @@ drawable_unlock(struct apple_glx_drawable *d)
    err = pthread_mutex_unlock(&d->mutex);
 
    if (err) {
-      fprintf(stderr, "pthread_mutex_unlock error: %d\n", err);
+      fprintf(stderr, "pthread_mutex_unlock error: %s\n", strerror(err));
       abort();
    }
 }
@@ -135,6 +136,7 @@ release_drawable(struct apple_glx_drawable *d)
 static bool
 destroy_drawable(struct apple_glx_drawable *d)
 {
+   int err;
 
    d->lock(d);
 
@@ -172,6 +174,15 @@ destroy_drawable(struct apple_glx_drawable *d)
 
    apple_glx_diagnostic("%s: freeing %p\n", __func__, (void *) d);
 
+   /* Stupid recursive locks */
+   while (pthread_mutex_unlock(&d->mutex) == 0);
+
+   err = pthread_mutex_destroy(&d->mutex);
+   if (err) {
+      fprintf(stderr, "pthread_mutex_destroy error: %s\n", strerror(err));
+      abort();
+   }
+   
    free(d);
 
    /* So that the locks are balanced and the caller correctly unlocks. */
@@ -238,7 +249,7 @@ common_init(Display * dpy, GLXDrawable drawable, struct apple_glx_drawable *d)
    err = pthread_mutexattr_init(&attr);
 
    if (err) {
-      fprintf(stderr, "pthread_mutexattr_init error: %d\n", err);
+      fprintf(stderr, "pthread_mutexattr_init error: %s\n", strerror(err));
       abort();
    }
 
@@ -250,14 +261,14 @@ common_init(Display * dpy, GLXDrawable drawable, struct apple_glx_drawable *d)
    err = pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_RECURSIVE);
 
    if (err) {
-      fprintf(stderr, "error: setting pthread mutex type: %d\n", err);
+      fprintf(stderr, "error: setting pthread mutex type: %s\n", strerror(err));
       abort();
    }
 
    err = pthread_mutex_init(&d->mutex, &attr);
 
    if (err) {
-      fprintf(stderr, "pthread_mutex_init error: %d\n", err);
+      fprintf(stderr, "pthread_mutex_init error: %s\n", strerror(err));
       abort();
    }
 

src/glx/apple/apple_glx_log.c

@@ -0,0 +1,128 @@
+/*
+ * Copyright (c) 2012 Apple Inc.
+ *
+ * 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 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 ABOVE LISTED COPYRIGHT
+ * HOLDER(S) 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.
+ *
+ * Except as contained in this notice, the name(s) of the above
+ * copyright holders shall not be used in advertising or otherwise to
+ * promote the sale, use or other dealings in this Software without
+ * prior written authorization.
+ */
+
+#include <sys/cdefs.h>
+#include <asl.h>
+#include <stdio.h>
+#include <stdbool.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <inttypes.h>
+#include <pthread.h>
+#include "apple_glx_log.h"
+
+static bool diagnostic = false;
+static aslclient aslc;
+
+void apple_glx_log_init(void) {
+    if (getenv("LIBGL_DIAGNOSTIC")) {
+        diagnostic = true;
+    }
+
+    aslc = asl_open(NULL, NULL, 0);
+}
+
+void _apple_glx_log(int level, const char *file, const char *function,
+                    int line, const char *fmt, ...) {
+    va_list v;
+    va_start(v, fmt);
+    _apple_glx_vlog(level, file, function, line, fmt, v);
+    va_end(v);
+}
+
+static const char *
+_asl_level_string(int level)
+{
+        if (level == ASL_LEVEL_EMERG) return ASL_STRING_EMERG;
+        if (level == ASL_LEVEL_ALERT) return ASL_STRING_ALERT;
+        if (level == ASL_LEVEL_CRIT) return ASL_STRING_CRIT;
+        if (level == ASL_LEVEL_ERR) return ASL_STRING_ERR;
+        if (level == ASL_LEVEL_WARNING) return ASL_STRING_WARNING;
+        if (level == ASL_LEVEL_NOTICE) return ASL_STRING_NOTICE;
+        if (level == ASL_LEVEL_INFO) return ASL_STRING_INFO;
+        if (level == ASL_LEVEL_DEBUG) return ASL_STRING_DEBUG;
+        return "unknown";
+}
+
+void _apple_glx_vlog(int level, const char *file, const char *function,
+                     int line, const char *fmt, va_list args) {
+    aslmsg msg;
+    uint64_t thread = 0;
+
+    if (pthread_is_threaded_np()) {
+#if MAC_OS_X_VERSION_MAX_ALLOWED < 1060
+        thread = (uint64_t)(uintptr_t)pthread_self();
+#elif MAC_OS_X_VERSION_MIN_REQUIRED < 1060
+        if (&pthread_threadid_np) {
+            pthread_threadid_np(NULL, &thread);
+        } else {
+            thread = (uint64_t)(uintptr_t)pthread_self();
+        }
+#else
+        pthread_threadid_np(NULL, &thread);
+#endif
+    }
+
+    if (diagnostic) {
+        va_list args2;
+        va_copy(args2, args);
+
+        fprintf(stderr, "%-9s %24s:%-4d %s(%"PRIu64"): ",
+                _asl_level_string(level), file, line, function, thread);
+        vfprintf(stderr, fmt, args2);
+    }
+
+    msg = asl_new(ASL_TYPE_MSG);
+    if (msg) {
+        if (file)
+            asl_set(msg, "File", file);
+        if (function)
+            asl_set(msg, "Function", function);
+        if (line) {
+            char *_line;
+            asprintf(&_line, "%d", line);
+            if (_line) {
+                asl_set(msg, "Line", _line);
+                free(_line);
+            }
+        }
+        if (pthread_is_threaded_np()) {
+            char *_thread;
+            asprintf(&_thread, "%"PRIu64, thread);
+            if (_thread) {
+                asl_set(msg, "Thread", _thread);
+                free(_thread);
+            }
+        }
+    }
+
+    asl_vlog(aslc, msg, level, fmt, args);
+    if (msg)
+        asl_free(msg);
+}

src/glx/apple/apple_glx_log.h

@@ -0,0 +1,57 @@
+/*
+ * Copyright (c) 2012 Apple Inc.
+ *
+ * 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 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 ABOVE LISTED COPYRIGHT
+ * HOLDER(S) 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.
+ *
+ * Except as contained in this notice, the name(s) of the above
+ * copyright holders shall not be used in advertising or otherwise to
+ * promote the sale, use or other dealings in this Software without
+ * prior written authorization.
+ */
+
+#ifndef APPLE_GLX_LOG_H
+#define APPLE_GLX_LOG_H
+
+#include <sys/cdefs.h>
+#include <asl.h>
+
+void apple_glx_log_init(void);
+
+__printflike(5, 6)
+void _apple_glx_log(int level, const char *file, const char *function,
+                    int line, const char *fmt, ...);
+#define apple_glx_log(l, f, args ...) \
+    _apple_glx_log(l, __FILE__, __FUNCTION__, __LINE__, f, ## args)
+
+
+__printflike(5, 0)
+void _apple_glx_vlog(int level, const char *file, const char *function,
+                     int line, const char *fmt, va_list v);
+#define apple_glx_vlog(l, f, v) \
+    _apple_glx_vlog(l, __FILE__, __FUNCTION__, __LINE__, f, v)
+
+/* This is just here to help the transition.
+ * TODO: Replace calls to apple_glx_diagnostic
+ */
+#define apple_glx_diagnostic(f, args ...) \
+    apple_glx_log(ASL_LEVEL_DEBUG, f, ## args)
+
+#endif

src/glx/apple/apple_glx_surface.c

@@ -206,6 +206,7 @@ apple_glx_surface_destroy(unsigned int uid)
    if (d) {
       d->types.surface.pending_destroy = true;
       d->release(d);
+
       /* 
        * We release 2 references to the surface.  One was acquired by
        * the find, and the other was leftover from a context, or 
@@ -216,8 +217,9 @@ apple_glx_surface_destroy(unsigned int uid)
        * to actually destroy it when the pending_destroy is processed
        * by a glViewport callback (see apple_glx_context_update()).
        */
-      d->destroy(d);
-
-      d->unlock(d);
+      if (!d->destroy(d)) {
+          /* apple_glx_drawable_find_by_uid returns a locked drawable */
+          d->unlock(d);
+      }
    }
 }

src/glx/compsize.c

@@ -71,6 +71,7 @@ __glElementsPerGroup(GLenum format, GLenum type)
    case GL_422_REV_EXT:
    case GL_422_AVERAGE_EXT:
    case GL_422_REV_AVERAGE_EXT:
+   case GL_DEPTH_STENCIL_NV:
    case GL_YCBCR_422_APPLE:
    case GL_LUMINANCE_ALPHA:
       return 2;

src/glx/create_context.c

@@ -80,8 +80,13 @@ glXCreateContextAttribsARB(Display *dpy, GLXFBConfig config,
 					       &dummy_err);
    }
 
-   if (gc == NULL)
+   if (gc == NULL) {
+#ifdef GLX_USE_APPLEGL
+      gc = applegl_create_context(psc, cfg, share, 0);
+#else
       gc = indirect_create_context(psc, cfg, share, 0);
+#endif
+   }
 
    gc->xid = xcb_generate_id(c);
    gc->share_xid = (share != NULL) ? share->xid : 0;

src/glx/drisw_glx.c

@@ -91,6 +91,9 @@ XCreateDrawable(struct drisw_drawable * pdp,
    visMask = VisualIDMask;
    pdp->visinfo = XGetVisualInfo(dpy, visMask, &visTemp, &num_visuals);
 
+   if (!pdp->visinfo || num_visuals == 0)
+      return False;
+
    /* create XImage */
    pdp->ximage = XCreateImage(dpy,
                               pdp->visinfo->visual,
@@ -513,7 +516,7 @@ driswCreateDrawable(struct glx_screen *base, XID xDrawable,
    struct drisw_drawable *pdp;
    __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) modes;
    struct drisw_screen *psc = (struct drisw_screen *) base;
-
+   Bool ret;
    const __DRIswrastExtension *swrast = psc->swrast;
 
    pdp = Xmalloc(sizeof(*pdp));
@@ -525,7 +528,11 @@ driswCreateDrawable(struct glx_screen *base, XID xDrawable,
    pdp->base.drawable = drawable;
    pdp->base.psc = &psc->base;
 
-   XCreateDrawable(pdp, psc->base.dpy, xDrawable, modes->visualID);
+   ret = XCreateDrawable(pdp, psc->base.dpy, xDrawable, modes->visualID);
+   if (!ret) {
+      Xfree(pdp);
+      return NULL;
+   }
 
    /* Create a new drawable */
    pdp->driDrawable =

src/glx/glxext.c

@@ -293,6 +293,9 @@ QueryVersion(Display * dpy, int opcode, int *major, int *minor)
                                                                        GLX_MINOR_VERSION),
                                                                       NULL);
 
+   if (!reply)
+     return GL_FALSE;
+
    if (reply->major_version != GLX_MAJOR_VERSION) {
       free(reply);
       return GL_FALSE;

src/mapi/glapi/glapi_gentable.c

@@ -105,7 +105,7 @@ __glapi_gentable_set_remaining_noop(struct _glapi_table *disp) {
 
 struct _glapi_table *
 _glapi_create_table_from_handle(void *handle, const char *symbol_prefix) {
-    struct _glapi_table *disp = calloc(1, sizeof(struct _glapi_table));
+    struct _glapi_table *disp = calloc(_glapi_get_dispatch_table_size(), sizeof(void *));
     char symboln[512];
 
     if(!disp)

src/mesa/drivers/common/meta.c

@@ -315,6 +315,10 @@ struct gl_meta_state
    struct drawtex_state DrawTex;  /**< For _mesa_meta_DrawTex() */
 };
 
+static void meta_glsl_blit_cleanup(struct gl_context *ctx, struct blit_state *blit);
+static void cleanup_temp_texture(struct gl_context *ctx, struct temp_texture *tex);
+static void meta_glsl_clear_cleanup(struct gl_context *ctx, struct clear_state *clear);
+
 static GLuint
 compile_shader_with_debug(struct gl_context *ctx, GLenum target, const GLcharARB *source)
 {
@@ -331,12 +335,16 @@ compile_shader_with_debug(struct gl_context *ctx, GLenum target, const GLcharARB
       return shader;
 
    _mesa_GetShaderiv(shader, GL_INFO_LOG_LENGTH, &size);
-   if (size == 0)
+   if (size == 0) {
+      _mesa_DeleteObjectARB(shader);
       return 0;
+   }
 
    info = malloc(size);
-   if (!info)
+   if (!info) {
+      _mesa_DeleteObjectARB(shader);
       return 0;
+   }
 
    _mesa_GetProgramInfoLog(shader, size, NULL, info);
    _mesa_problem(ctx,
@@ -345,6 +353,7 @@ compile_shader_with_debug(struct gl_context *ctx, GLenum target, const GLcharARB
 		 info, source);
 
    free(info);
+   _mesa_DeleteObjectARB(shader);
 
    return 0;
 }
@@ -397,10 +406,15 @@ _mesa_meta_init(struct gl_context *ctx)
 void
 _mesa_meta_free(struct gl_context *ctx)
 {
-   /* Note: Any textures, VBOs, etc, that we allocate should get
-    * freed by the normal context destruction code.  But this would be
-    * the place to free other meta data someday.
-    */
+   GET_CURRENT_CONTEXT(old_context);
+   _mesa_make_current(ctx, NULL, NULL);
+   meta_glsl_blit_cleanup(ctx, &ctx->Meta->Blit);
+   meta_glsl_clear_cleanup(ctx, &ctx->Meta->Clear);
+   cleanup_temp_texture(ctx, &ctx->Meta->TempTex);
+   if (old_context)
+      _mesa_make_current(old_context, old_context->WinSysDrawBuffer, old_context->WinSysReadBuffer);
+   else
+      _mesa_make_current(NULL, NULL, NULL);
    free(ctx->Meta);
    ctx->Meta = NULL;
 }
@@ -1056,6 +1070,15 @@ init_temp_texture(struct gl_context *ctx, struct temp_texture *tex)
    _mesa_GenTextures(1, &tex->TexObj);
 }
 
+static void
+cleanup_temp_texture(struct gl_context *ctx, struct temp_texture *tex)
+{
+   if (!tex->TexObj)
+     return;
+   _mesa_DeleteTextures(1, &tex->TexObj);
+   tex->TexObj = 0;
+}
+
 
 /**
  * Return pointer to temp_texture info for non-bitmap ops.
@@ -1592,6 +1615,21 @@ _mesa_meta_BlitFramebuffer(struct gl_context *ctx,
    }
 }
 
+static void
+meta_glsl_blit_cleanup(struct gl_context *ctx, struct blit_state *blit)
+{
+   if (blit->ArrayObj) {
+      _mesa_DeleteVertexArraysAPPLE(1, &blit->ArrayObj);
+      blit->ArrayObj = 0;
+      _mesa_DeleteBuffersARB(1, &blit->VBO);
+      blit->VBO = 0;
+   }
+   if (blit->DepthFP) {
+      _mesa_DeletePrograms(1, &blit->DepthFP);
+      blit->DepthFP = 0;
+   }
+}
+
 
 /**
  * Meta implementation of ctx->Driver.Clear() in terms of polygon rendering.
@@ -1774,7 +1812,9 @@ meta_glsl_clear_init(struct gl_context *ctx, struct clear_state *clear)
 
    clear->ShaderProg = _mesa_CreateProgramObjectARB();
    _mesa_AttachShader(clear->ShaderProg, fs);
+   _mesa_DeleteObjectARB(fs);
    _mesa_AttachShader(clear->ShaderProg, vs);
+   _mesa_DeleteObjectARB(vs);
    _mesa_BindAttribLocationARB(clear->ShaderProg, 0, "position");
    _mesa_LinkProgramARB(clear->ShaderProg);
 
@@ -1787,7 +1827,9 @@ meta_glsl_clear_init(struct gl_context *ctx, struct clear_state *clear)
 
       clear->IntegerShaderProg = _mesa_CreateProgramObjectARB();
       _mesa_AttachShader(clear->IntegerShaderProg, fs);
+      _mesa_DeleteObjectARB(fs);
       _mesa_AttachShader(clear->IntegerShaderProg, vs);
+      _mesa_DeleteObjectARB(vs);
       _mesa_BindAttribLocationARB(clear->IntegerShaderProg, 0, "position");
 
       /* Note that user-defined out attributes get automatically assigned
@@ -1802,6 +1844,24 @@ meta_glsl_clear_init(struct gl_context *ctx, struct clear_state *clear)
    }
 }
 
+static void
+meta_glsl_clear_cleanup(struct gl_context *ctx, struct clear_state *clear)
+{
+   if (clear->ArrayObj == 0)
+      return;
+   _mesa_DeleteVertexArraysAPPLE(1, &clear->ArrayObj);
+   clear->ArrayObj = 0;
+   _mesa_DeleteBuffersARB(1, &clear->VBO);
+   clear->VBO = 0;
+   _mesa_DeleteObjectARB(clear->ShaderProg);
+   clear->ShaderProg = 0;
+
+   if (clear->IntegerShaderProg) {
+      _mesa_DeleteObjectARB(clear->IntegerShaderProg);
+      clear->IntegerShaderProg = 0;
+   }
+}
+
 /**
  * Meta implementation of ctx->Driver.Clear() in terms of polygon rendering.
  */
@@ -3256,6 +3316,7 @@ decompress_texture_image(struct gl_context *ctx,
    };
    struct vertex verts[4];
    GLuint fboDrawSave, fboReadSave;
+   GLuint rbSave;
 
    if (slice > 0) {
       assert(target == GL_TEXTURE_3D ||
@@ -3272,6 +3333,7 @@ decompress_texture_image(struct gl_context *ctx,
    /* save fbo bindings (not saved by _mesa_meta_begin()) */
    fboDrawSave = ctx->DrawBuffer->Name;
    fboReadSave = ctx->ReadBuffer->Name;
+   rbSave = ctx->CurrentRenderbuffer ? ctx->CurrentRenderbuffer->Name : 0;
 
    _mesa_meta_begin(ctx, MESA_META_ALL & ~MESA_META_PIXEL_STORE);
 
@@ -3292,6 +3354,7 @@ decompress_texture_image(struct gl_context *ctx,
 
    /* alloc dest surface */
    if (width > decompress->Width || height > decompress->Height) {
+      _mesa_BindRenderbufferEXT(GL_RENDERBUFFER_EXT, decompress->RBO);
       _mesa_RenderbufferStorageEXT(GL_RENDERBUFFER_EXT, GL_RGBA,
                                    width, height);
       decompress->Width = width;
@@ -3424,6 +3487,7 @@ decompress_texture_image(struct gl_context *ctx,
       _mesa_BindFramebufferEXT(GL_DRAW_FRAMEBUFFER_EXT, fboDrawSave);
       _mesa_BindFramebufferEXT(GL_READ_FRAMEBUFFER_EXT, fboReadSave);
    }
+   _mesa_BindRenderbufferEXT(GL_RENDERBUFFER_EXT, rbSave);
 }
 
 

src/mesa/drivers/dri/common/drirc

@@ -6,5 +6,11 @@
         <application executable="Tropics">
             <option name="force_glsl_extensions_warn" value="true" />
 	</application>
+        <application executable="heaven_x86">
+            <option name="force_glsl_extensions_warn" value="true" />
+	</application>
+        <application executable="heaven_x64">
+            <option name="force_glsl_extensions_warn" value="true" />
+	</application>
     </device>
 </driconf>

src/mesa/drivers/dri/i915/i830_state.c

@@ -436,7 +436,10 @@ i830DepthMask(struct gl_context * ctx, GLboolean flag)
    struct i830_context *i830 = i830_context(ctx);
 
    DBG("%s flag (%d)\n", __FUNCTION__, flag);
-   
+
+   if (!ctx->DrawBuffer || !ctx->DrawBuffer->Visual.depthBits)
+      flag = false;
+
    I830_STATECHANGE(i830, I830_UPLOAD_CTX);
 
    i830->state.Ctx[I830_CTXREG_ENABLES_2] &= ~ENABLE_DIS_DEPTH_WRITE_MASK;
@@ -473,7 +476,7 @@ static void
 i830PolygonStipple(struct gl_context * ctx, const GLubyte * mask)
 {
    struct i830_context *i830 = i830_context(ctx);
-   const GLubyte *m = mask;
+   const GLubyte *m;
    GLubyte p[4];
    int i, j, k;
    int active = (ctx->Polygon.StippleFlag &&
@@ -485,6 +488,12 @@ i830PolygonStipple(struct gl_context * ctx, const GLubyte * mask)
       i830->state.Stipple[I830_STPREG_ST1] &= ~ST1_ENABLE;
    }
 
+   /* Use the already unpacked stipple data from the context rather than the
+    * uninterpreted mask passed in.
+    */
+   mask = (const GLubyte *)ctx->PolygonStipple;
+   m = mask;
+
    p[0] = mask[12] & 0xf;
    p[0] |= p[0] << 4;
    p[1] = mask[8] & 0xf;
@@ -795,6 +804,9 @@ i830Enable(struct gl_context * ctx, GLenum cap, GLboolean state)
       I830_STATECHANGE(i830, I830_UPLOAD_CTX);
       i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_DIS_DEPTH_TEST_MASK;
 
+      if (!ctx->DrawBuffer || !ctx->DrawBuffer->Visual.depthBits)
+	 state = false;
+
       if (state)
          i830->state.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_DEPTH_TEST;
       else
@@ -849,7 +861,7 @@ i830Enable(struct gl_context * ctx, GLenum cap, GLboolean state)
          if (ctx->DrawBuffer) {
             struct intel_renderbuffer *irbStencil
                = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL);
-            hw_stencil = (irbStencil && irbStencil->mt->region);
+            hw_stencil = (irbStencil && irbStencil->mt);
          }
          if (hw_stencil) {
             I830_STATECHANGE(i830, I830_UPLOAD_CTX);

src/mesa/drivers/dri/i915/i830_vtbl.c

@@ -692,8 +692,8 @@ i830_set_draw_region(struct intel_context *intel,
    state->Buffer[I830_DESTREG_DRAWRECT1] = 0;
    state->Buffer[I830_DESTREG_DRAWRECT2] = (draw_y << 16) | draw_x;
    state->Buffer[I830_DESTREG_DRAWRECT3] =
-      ((ctx->DrawBuffer->Width + draw_x) & 0xffff) |
-      ((ctx->DrawBuffer->Height + draw_y) << 16);
+      ((ctx->DrawBuffer->Width + draw_x - 1) & 0xffff) |
+      ((ctx->DrawBuffer->Height + draw_y - 1) << 16);
    state->Buffer[I830_DESTREG_DRAWRECT4] = (draw_y << 16) | draw_x;
    state->Buffer[I830_DESTREG_DRAWRECT5] = MI_NOOP;
 
@@ -826,8 +826,7 @@ i830_update_draw_buffer(struct intel_context *intel)
    /*
     * Update depth and stencil test state
     */
-   ctx->Driver.Enable(ctx, GL_DEPTH_TEST,
-		      (ctx->Depth.Test && fb->Visual.depthBits > 0));
+   ctx->Driver.Enable(ctx, GL_DEPTH_TEST, ctx->Depth.Test);
    ctx->Driver.Enable(ctx, GL_STENCIL_TEST,
 		      (ctx->Stencil.Enabled && fb->Visual.stencilBits > 0));
 

src/mesa/drivers/dri/i915/i915_context.c

@@ -76,6 +76,8 @@ i915InvalidateState(struct gl_context * ctx, GLuint new_state)
        i915_update_provoking_vertex(ctx);
    if (new_state & (_NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS))
        i915_update_program(ctx);
+   if (new_state & (_NEW_PROGRAM | _NEW_POINT))
+       i915_update_sprite_point_enable(ctx);
 }
 
 

src/mesa/drivers/dri/i915/i915_context.h

@@ -40,6 +40,7 @@
 #define I915_FALLBACK_POINT_SMOOTH	 0x80000
 #define I915_FALLBACK_POINT_SPRITE_COORD_ORIGIN	 0x100000
 #define I915_FALLBACK_DRAW_OFFSET	 0x200000
+#define I915_FALLBACK_COORD_REPLACE	 0x400000
 
 #define I915_UPLOAD_CTX              0x1
 #define I915_UPLOAD_BUFFERS          0x2
@@ -338,6 +339,7 @@ extern void i915InitStateFunctions(struct dd_function_table *functions);
 extern void i915InitState(struct i915_context *i915);
 extern void i915_update_stencil(struct gl_context * ctx);
 extern void i915_update_provoking_vertex(struct gl_context *ctx);
+extern void i915_update_sprite_point_enable(struct gl_context *ctx);
 
 
 /*======================================================================

src/mesa/drivers/dri/i915/i915_fragprog.c

@@ -1361,6 +1361,10 @@ i915ValidateFragmentProgram(struct i915_context *i915)
       EMIT_ATTR(_TNL_ATTRIB_POS, EMIT_3F_VIEWPORT, S4_VFMT_XYZ, 12);
    }
 
+   /* Handle gl_PointSize builtin var here */
+   if (ctx->Point._Attenuated || ctx->VertexProgram.PointSizeEnabled)
+      EMIT_ATTR(_TNL_ATTRIB_POINTSIZE, EMIT_1F, S4_VFMT_POINT_WIDTH, 4);
+
    if (inputsRead & FRAG_BIT_COL0) {
       intel->coloroffset = offset / 4;
       EMIT_ATTR(_TNL_ATTRIB_COLOR0, EMIT_4UB_4F_BGRA, S4_VFMT_COLOR, 4);

src/mesa/drivers/dri/i915/i915_state.c

@@ -652,6 +652,48 @@ i915PointParameterfv(struct gl_context * ctx, GLenum pname, const GLfloat *param
    }
 }
 
+void
+i915_update_sprite_point_enable(struct gl_context *ctx)
+{
+   struct intel_context *intel = intel_context(ctx);
+   /* _NEW_PROGRAM */
+   struct i915_fragment_program *p =
+      (struct i915_fragment_program *) ctx->FragmentProgram._Current;
+   const GLbitfield64 inputsRead = p->FragProg.Base.InputsRead;
+   struct i915_context *i915 = i915_context(ctx);
+   GLuint s4 = i915->state.Ctx[I915_CTXREG_LIS4] & ~S4_VFMT_MASK;
+   int i;
+   GLuint coord_replace_bits = 0x0;
+   GLuint tex_coord_unit_bits = 0x0;
+
+   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
+      /* _NEW_POINT */
+      if (ctx->Point.CoordReplace[i] && ctx->Point.PointSprite)
+         coord_replace_bits |= (1 << i);
+      if (inputsRead & FRAG_BIT_TEX(i))
+         tex_coord_unit_bits |= (1 << i);
+   }
+
+   /*
+    * Here we can't enable the SPRITE_POINT_ENABLE bit when the mis-match
+    * of tex_coord_unit_bits and coord_replace_bits, or this will make all
+    * the other non-point-sprite coords(like varying inputs, as we now use
+    * tex coord to implement varying inputs) be replaced to value (0, 0)-(1, 1).
+    *
+    * Thus, do fallback when needed.
+    */
+   FALLBACK(intel, I915_FALLBACK_COORD_REPLACE,
+            coord_replace_bits && coord_replace_bits != tex_coord_unit_bits);
+
+   s4 &= ~S4_SPRITE_POINT_ENABLE;
+   s4 |= (coord_replace_bits && coord_replace_bits == tex_coord_unit_bits) ?
+         S4_SPRITE_POINT_ENABLE : 0;
+   if (s4 != i915->state.Ctx[I915_CTXREG_LIS4]) {
+      i915->state.Ctx[I915_CTXREG_LIS4] = s4;
+      I915_STATECHANGE(i915, I915_UPLOAD_CTX);
+   }
+}
+
 
 /* =============================================================
  * Color masks
@@ -869,18 +911,7 @@ i915Enable(struct gl_context * ctx, GLenum cap, GLboolean state)
       break;
 
    case GL_POINT_SPRITE:
-      /* This state change is handled in i915_reduced_primitive_state because
-       * the hardware bit should only be set when rendering points.
-       */
-	 dw = i915->state.Ctx[I915_CTXREG_LIS4];
-      if (state)
-	 dw |= S4_SPRITE_POINT_ENABLE;
-      else
-	 dw &= ~S4_SPRITE_POINT_ENABLE;
-      if (dw != i915->state.Ctx[I915_CTXREG_LIS4]) {
-	 i915->state.Ctx[I915_CTXREG_LIS4] = dw;
-	 I915_STATECHANGE(i915, I915_UPLOAD_CTX);
-      }
+      /* Handle it at i915_update_sprite_point_enable () */
       break;
 
    case GL_POINT_SMOOTH:

src/mesa/drivers/dri/i915/i915_texstate.c

@@ -319,6 +319,28 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
            ((wt != GL_CLAMP) && (wt != GL_CLAMP_TO_EDGE))))
           return false;
 
+      /*
+       * According to 3DSTATE_MAP_STATE at page of 104 in Bspec
+       * Vol3d 3D Instructions:
+       *   [DevGDG and DevAlv]: Must be a power of 2 for cube maps.
+       *   [DevLPT, DevCST and DevBLB]: If not a power of 2, cube maps
+       *      must have all faces enabled.
+       *
+       * But, as I tested on pineview(DevBLB derived), the rendering is
+       * bad(you will find the color isn't samplered right in some
+       * fragments). After checking, it seems that the texture layout is
+       * wrong: making the width and height align of 4(although this
+       * doesn't make much sense) will fix this issue and also broke some
+       * others. Well, Bspec mentioned nothing about the layout alignment
+       * and layout for NPOT cube map.  I guess the Bspec just assume it's
+       * a POT cube map.
+       *
+       * Thus, I guess we need do this for other platforms as well.
+       */
+      if (tObj->Target == GL_TEXTURE_CUBE_MAP_ARB &&
+          !is_power_of_two(firstImage->Height))
+         return false;
+
       state[I915_TEXREG_SS3] = ss3;     /* SS3_NORMALIZED_COORDS */
 
       state[I915_TEXREG_SS3] |=

src/mesa/drivers/dri/i915/i915_vtbl.c

@@ -665,12 +665,11 @@ i915_set_draw_region(struct intel_context *intel,
 
    draw_offset = (draw_y << 16) | draw_x;
 
+   FALLBACK(intel, I915_FALLBACK_DRAW_OFFSET,
+            (ctx->DrawBuffer->Width + draw_x > 2048) ||
+            (ctx->DrawBuffer->Height + draw_y > 2048));
    /* When changing drawing rectangle offset, an MI_FLUSH is first required. */
    if (draw_offset != i915->last_draw_offset) {
-      FALLBACK(intel, I915_FALLBACK_DRAW_OFFSET,
-               (ctx->DrawBuffer->Width + draw_x > 2048) ||
-               (ctx->DrawBuffer->Height + draw_y > 2048));
-
       state->Buffer[I915_DESTREG_DRAWRECT0] = MI_FLUSH | INHIBIT_FLUSH_RENDER_CACHE;
       i915->last_draw_offset = draw_offset;
    } else

src/mesa/drivers/dri/i915/intel_render.c

@@ -119,13 +119,15 @@ intelDmaPrimitive(struct intel_context *intel, GLenum prim)
    intel_set_prim(intel, hw_prim[prim]);
 }
 
+#define INTEL_NO_VBO_STATE_RESERVED 1500
+
 static INLINE GLuint intel_get_vb_max(struct intel_context *intel)
 {
    GLuint ret;
 
-   if (intel->intelScreen->no_vbo)
-      ret = sizeof(intel->batch.map) - 1500;
-   else
+   if (intel->intelScreen->no_vbo) {
+      ret = intel->batch.bo->size - INTEL_NO_VBO_STATE_RESERVED;
+   } else
       ret = INTEL_VB_SIZE;
    ret /= (intel->vertex_size * 4);
    return ret;
@@ -133,11 +135,15 @@ static INLINE GLuint intel_get_vb_max(struct intel_context *intel)
 
 static INLINE GLuint intel_get_current_max(struct intel_context *intel)
 {
+   GLuint ret;
 
-   if (intel->intelScreen->no_vbo)
-      return intel_get_vb_max(intel);
-   else
-      return (INTEL_VB_SIZE - intel->prim.current_offset) / (intel->vertex_size * 4);
+   if (intel->intelScreen->no_vbo) {
+      ret = intel_batchbuffer_space(intel);
+      ret = ret <= INTEL_NO_VBO_STATE_RESERVED ? 0 : ret - INTEL_NO_VBO_STATE_RESERVED;
+   } else
+      ret = (INTEL_VB_SIZE - intel->prim.current_offset);
+
+   return ret / (intel->vertex_size * 4);
 }
 
 #define LOCAL_VARS struct intel_context *intel = intel_context(ctx)

src/mesa/drivers/dri/i915/intel_tris.c

@@ -1198,6 +1198,7 @@ static char *fallbackStrings[] = {
    [19] = "Smooth point",
    [20] = "point sprite coord origin",
    [21] = "depth/color drawing offset",
+   [22] = "coord replace(SPRITE POINT ENABLE)",
 };
 
 

src/mesa/drivers/dri/i965/brw_context.h

@@ -290,6 +290,12 @@ typedef enum
    BRW_VERT_RESULT_NDC = VERT_RESULT_MAX,
    BRW_VERT_RESULT_HPOS_DUPLICATE,
    BRW_VERT_RESULT_PAD,
+   /*
+    * It's actually not a vert_result but just a _mark_ to let sf aware that
+    * he need do something special to handle gl_PointCoord builtin variable
+    * correctly. see compile_sf_prog() for more info.
+    */
+   BRW_VERT_RESULT_PNTC,
    BRW_VERT_RESULT_MAX
 } brw_vert_result;
 

src/mesa/drivers/dri/i965/brw_eu.c

@@ -47,13 +47,13 @@ brw_swap_cmod(uint32_t cmod)
    case BRW_CONDITIONAL_NZ:
       return cmod;
    case BRW_CONDITIONAL_G:
-      return BRW_CONDITIONAL_LE;
-   case BRW_CONDITIONAL_GE:
       return BRW_CONDITIONAL_L;
+   case BRW_CONDITIONAL_GE:
+      return BRW_CONDITIONAL_LE;
    case BRW_CONDITIONAL_L:
-      return BRW_CONDITIONAL_GE;
-   case BRW_CONDITIONAL_LE:
       return BRW_CONDITIONAL_G;
+   case BRW_CONDITIONAL_LE:
+      return BRW_CONDITIONAL_GE;
    default:
       return ~0;
    }

src/mesa/drivers/dri/i965/brw_eu.h

@@ -1048,6 +1048,7 @@ struct brw_instruction *brw_WHILE(struct brw_compile *p);
 struct brw_instruction *brw_BREAK(struct brw_compile *p);
 struct brw_instruction *brw_CONT(struct brw_compile *p);
 struct brw_instruction *gen6_CONT(struct brw_compile *p);
+struct brw_instruction *gen6_HALT(struct brw_compile *p);
 /* Forward jumps:
  */
 void brw_land_fwd_jump(struct brw_compile *p, int jmp_insn_idx);

src/mesa/drivers/dri/i965/brw_eu_emit.c

@@ -1339,6 +1339,20 @@ struct brw_instruction *brw_CONT(struct brw_compile *p)
    return insn;
 }
 
+struct brw_instruction *gen6_HALT(struct brw_compile *p)
+{
+   struct brw_instruction *insn;
+
+   insn = next_insn(p, BRW_OPCODE_HALT);
+   brw_set_dest(p, insn, retype(brw_null_reg(), BRW_REGISTER_TYPE_D));
+   brw_set_src0(p, insn, retype(brw_null_reg(), BRW_REGISTER_TYPE_D));
+   brw_set_src1(p, insn, brw_imm_d(0x0)); /* UIP and JIP, updated later. */
+
+   insn->header.compression_control = BRW_COMPRESSION_NONE;
+   insn->header.execution_size = BRW_EXECUTE_8;
+   return insn;
+}
+
 /* DO/WHILE loop:
  *
  * The DO/WHILE is just an unterminated loop -- break or continue are
@@ -2188,7 +2202,7 @@ void brw_fb_WRITE(struct brw_compile *p,
 			    msg_type,
 			    msg_length,
 			    header_present,
-			    1, /* last render target write */
+			    eot, /* last render target write */
 			    response_length,
 			    eot,
 			    0 /* send_commit_msg */);
@@ -2395,8 +2409,8 @@ brw_find_next_block_end(struct brw_compile *p, int start)
 	 return ip;
       }
    }
-   assert(!"not reached");
-   return start + 1;
+
+   return 0;
 }
 
 /* There is no DO instruction on gen6, so to find the end of the loop
@@ -2425,7 +2439,7 @@ brw_find_loop_end(struct brw_compile *p, int start)
 }
 
 /* After program generation, go back and update the UIP and JIP of
- * BREAK and CONT instructions to their correct locations.
+ * BREAK, CONT, and HALT instructions to their correct locations.
  */
 void
 brw_set_uip_jip(struct brw_compile *p)
@@ -2439,18 +2453,47 @@ brw_set_uip_jip(struct brw_compile *p)
 
    for (ip = 0; ip < p->nr_insn; ip++) {
       struct brw_instruction *insn = &p->store[ip];
+      int block_end_ip = 0;
+
+      if (insn->header.opcode == BRW_OPCODE_BREAK ||
+	  insn->header.opcode == BRW_OPCODE_CONTINUE ||
+	  insn->header.opcode == BRW_OPCODE_HALT) {
+	 block_end_ip = brw_find_next_block_end(p, ip);
+      }
 
       switch (insn->header.opcode) {
       case BRW_OPCODE_BREAK:
-	 insn->bits3.break_cont.jip = br * (brw_find_next_block_end(p, ip) - ip);
+	 assert(block_end_ip != 0);
+	 insn->bits3.break_cont.jip = br * (block_end_ip - ip);
 	 /* Gen7 UIP points to WHILE; Gen6 points just after it */
 	 insn->bits3.break_cont.uip =
 	    br * (brw_find_loop_end(p, ip) - ip + (intel->gen == 6 ? 1 : 0));
 	 break;
       case BRW_OPCODE_CONTINUE:
-	 insn->bits3.break_cont.jip = br * (brw_find_next_block_end(p, ip) - ip);
+	 assert(block_end_ip != 0);
+	 insn->bits3.break_cont.jip = br * (block_end_ip - ip);
 	 insn->bits3.break_cont.uip = br * (brw_find_loop_end(p, ip) - ip);
 
+	 assert(insn->bits3.break_cont.uip != 0);
+	 assert(insn->bits3.break_cont.jip != 0);
+	 break;
+      case BRW_OPCODE_HALT:
+	 /* From the Sandy Bridge PRM (volume 4, part 2, section 8.3.19):
+	  *
+	  *    "In case of the halt instruction not inside any conditional code
+	  *     block, the value of <JIP> and <UIP> should be the same. In case
+	  *     of the halt instruction inside conditional code block, the <UIP>
+	  *     should be the end of the program, and the <JIP> should be end of
+	  *     the most inner conditional code block."
+	  *
+	  * The uip will have already been set by whoever set up the
+	  * instruction.
+	  */
+	 if (block_end_ip == 0) {
+	    insn->bits3.break_cont.jip = insn->bits3.break_cont.uip;
+	 } else {
+	    insn->bits3.break_cont.jip = br * (block_end_ip - ip);
+	 }
 	 assert(insn->bits3.break_cont.uip != 0);
 	 assert(insn->bits3.break_cont.jip != 0);
 	 break;

src/mesa/drivers/dri/i965/brw_fs.cpp

@@ -710,6 +710,15 @@ fs_visitor::calculate_urb_setup()
 	       urb_setup[fp_index] = urb_next++;
 	 }
       }
+
+      /*
+       * It's a FS only attribute, and we did interpolation for this attribute
+       * in SF thread. So, count it here, too.
+       *
+       * See compile_sf_prog() for more info.
+       */
+      if (brw->fragment_program->Base.InputsRead & BITFIELD64_BIT(FRAG_ATTRIB_PNTC))
+         urb_setup[FRAG_ATTRIB_PNTC] = urb_next++;
    }
 
    /* Each attribute is 4 setup channels, each of which is half a reg. */

src/mesa/drivers/dri/i965/brw_fs.h

@@ -171,6 +171,26 @@ static const fs_reg reg_undef;
 static const fs_reg reg_null_f(ARF, BRW_ARF_NULL, BRW_REGISTER_TYPE_F);
 static const fs_reg reg_null_d(ARF, BRW_ARF_NULL, BRW_REGISTER_TYPE_D);
 
+class ip_record : public exec_node {
+public:
+   static void* operator new(size_t size, void *ctx)
+   {
+      void *node;
+
+      node = rzalloc_size(ctx, size);
+      assert(node != NULL);
+
+      return node;
+   }
+
+   ip_record(int ip)
+   {
+      this->ip = ip;
+   }
+
+   int ip;
+};
+
 class fs_inst : public exec_node {
 public:
    /* Callers of this ralloc-based new need not call delete. It's
@@ -489,6 +509,7 @@ public:
    bool remove_duplicate_mrf_writes();
    bool virtual_grf_interferes(int a, int b);
    void schedule_instructions();
+   void patch_discard_jumps_to_fb_writes();
    void fail(const char *msg, ...);
 
    void push_force_uncompressed();
@@ -571,6 +592,7 @@ public:
    struct gl_shader_program *prog;
    void *mem_ctx;
    exec_list instructions;
+   exec_list discard_halt_patches;
 
    /* Delayed setup of c->prog_data.params[] due to realloc of
     * ParamValues[] during compile.
@@ -595,6 +617,7 @@ public:
    struct hash_table *variable_ht;
    ir_variable *frag_depth;
    fs_reg outputs[BRW_MAX_DRAW_BUFFERS];
+   unsigned output_components[BRW_MAX_DRAW_BUFFERS];
    int first_non_payload_grf;
    int max_grf;
    int urb_setup[FRAG_ATTRIB_MAX];