Merge branch 'upstream-experimental' into debian-experimental

44 files changed, 544 insertions, 331 deletions

diff --git a/docs/relnotes-7.7.1.html b/docs/relnotes-7.7.1.html
index b20c8a7724f..959efd9e57c 100644
--- a/docs/relnotes-7.7.1.html
+++ b/docs/relnotes-7.7.1.html
@@ -43,6 +43,7 @@ tbd
 <li>Allocate constants more tightly in GL_ARB_vertex/fragment parser.
 <li>Fixed mipmap generation bug caused by invalid viewport state.
 <li>Gallium SSE codegen for XPD didn't always work.
+<li>Fixed Windows build.
 </ul>
 
 
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_arit.c b/src/gallium/drivers/llvmpipe/lp_bld_arit.c
index f5d9db70fb5..1aee9b35f31 100644
--- a/src/gallium/drivers/llvmpipe/lp_bld_arit.c
+++ b/src/gallium/drivers/llvmpipe/lp_bld_arit.c
@@ -628,7 +628,8 @@ lp_build_abs(struct lp_build_context *bld,
    if(type.floating) {
       /* Mask out the sign bit */
       LLVMTypeRef int_vec_type = lp_build_int_vec_type(type);
-      LLVMValueRef mask = lp_build_int_const_scalar(type, ((unsigned long long)1 << type.width) - 1);
+      unsigned long absMask = ~(1 << (type.width - 1));
+      LLVMValueRef mask = lp_build_int_const_scalar(type, ((unsigned long long) absMask));
       a = LLVMBuildBitCast(bld->builder, a, int_vec_type, "");
       a = LLVMBuildAnd(bld->builder, a, mask, "");
       a = LLVMBuildBitCast(bld->builder, a, vec_type, "");
@@ -1082,7 +1083,7 @@ lp_build_log(struct lp_build_context *bld,
              LLVMValueRef x)
 {
    /* log(2) */
-   LLVMValueRef log2 = lp_build_const_scalar(bld->type, 1.4426950408889634);
+   LLVMValueRef log2 = lp_build_const_scalar(bld->type, 0.69314718055994529);
 
    return lp_build_mul(bld, log2, lp_build_exp2(bld, x));
 }
@@ -1094,7 +1095,7 @@ lp_build_log(struct lp_build_context *bld,
 
 /**
  * Generate polynomial.
- * Ex:  x^2 * coeffs[0] + x * coeffs[1] + coeffs[2].
+ * Ex:  coeffs[0] + x * coeffs[1] + x^2 * coeffs[2].
  */
 static LLVMValueRef
 lp_build_polynomial(struct lp_build_context *bld,
@@ -1284,13 +1285,13 @@ lp_build_log2_approx(struct lp_build_context *bld,
       /* mant = (float) mantissa(x) */
       mant = LLVMBuildAnd(bld->builder, i, mantmask, "");
       mant = LLVMBuildOr(bld->builder, mant, one, "");
-      mant = LLVMBuildSIToFP(bld->builder, mant, vec_type, "");
+      mant = LLVMBuildBitCast(bld->builder, mant, vec_type, "");
 
       logmant = lp_build_polynomial(bld, mant, lp_build_log2_polynomial,
                                     Elements(lp_build_log2_polynomial));
 
       /* This effectively increases the polynomial degree by one, but ensures that log2(1) == 0*/
-      logmant = LLVMBuildMul(bld->builder, logmant, LLVMBuildMul(bld->builder, mant, bld->one, ""), "");
+      logmant = LLVMBuildMul(bld->builder, logmant, LLVMBuildSub(bld->builder, mant, bld->one, ""), "");
 
       res = LLVMBuildAdd(bld->builder, logmant, logexp, "");
    }
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c b/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c
index e722bca1a6c..88cdf9e349f 100644
--- a/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c
+++ b/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c
@@ -766,7 +766,7 @@ emit_instruction(
       FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
          src0 = emit_fetch( bld, inst, 0, chan_index );
          tmp0 = lp_build_floor(&bld->base, src0);
-         tmp0 = lp_build_sub(&bld->base, tmp0, src0);
+         tmp0 = lp_build_sub(&bld->base, src0, tmp0);
          dst0[chan_index] = tmp0;
       }
       break;
diff --git a/src/gallium/drivers/llvmpipe/lp_debug.h b/src/gallium/drivers/llvmpipe/lp_debug.h
new file mode 100644
index 00000000000..74b27574942
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_debug.h
@@ -0,0 +1,71 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef LP_DEBUG_H
+#define LP_DEBUG_H
+
+#include "pipe/p_compiler.h"
+#include "util/u_debug.h"
+
+extern void
+st_print_current(void);
+
+
+#define DEBUG_PIPE      0x1
+#define DEBUG_TGSI      0x2
+#define DEBUG_TEX       0x4
+#define DEBUG_ASM       0x8
+#define DEBUG_SETUP     0x10
+#define DEBUG_RAST      0x20
+#define DEBUG_QUERY     0x40
+#define DEBUG_SCREEN    0x80
+#define DEBUG_JIT       0x100
+
+#ifdef DEBUG
+extern int LP_DEBUG;
+#else
+#define LP_DEBUG 0
+#endif
+
+void st_debug_init( void );
+
+static INLINE void
+LP_DBG( unsigned flag, const char *fmt, ... )
+{
+    if (LP_DEBUG & flag)
+    {
+        va_list args;
+
+        va_start( args, fmt );
+        debug_vprintf( fmt, args );
+        va_end( args );
+    }
+}
+
+
+#endif /* LP_DEBUG_H */
diff --git a/src/gallium/drivers/llvmpipe/lp_draw_arrays.c b/src/gallium/drivers/llvmpipe/lp_draw_arrays.c
index a43e438064c..e539e713226 100644
--- a/src/gallium/drivers/llvmpipe/lp_draw_arrays.c
+++ b/src/gallium/drivers/llvmpipe/lp_draw_arrays.c
@@ -101,7 +101,7 @@ llvmpipe_draw_range_elements(struct pipe_context *pipe,
    draw_arrays(draw, mode, start, count);
 
    /*
-    * unmap vertex/index buffers - will cause draw module to flush
+    * unmap vertex/index buffers
     */
    for (i = 0; i < lp->num_vertex_buffers; i++) {
       draw_set_mapped_vertex_buffer(draw, i, NULL);
@@ -110,6 +110,12 @@ llvmpipe_draw_range_elements(struct pipe_context *pipe,
       draw_set_mapped_element_buffer(draw, 0, NULL);
    }
 
+   /*
+    * TODO: Flush only when a user vertex/index buffer is present
+    * (or even better, modify draw module to do this
+    * internally when this condition is seen?)
+    */
+   draw_flush(draw);
 
    /* Note: leave drawing surfaces mapped */
 
diff --git a/src/gallium/drivers/llvmpipe/lp_jit.c b/src/gallium/drivers/llvmpipe/lp_jit.c
index b5aa7d680f1..7e019d4eda2 100644
--- a/src/gallium/drivers/llvmpipe/lp_jit.c
+++ b/src/gallium/drivers/llvmpipe/lp_jit.c
@@ -166,7 +166,7 @@ lp_jit_screen_init(struct llvmpipe_screen *screen)
    if (LLVMCreateJITCompiler(&screen->engine, screen->provider, 1, &error)) {
       _debug_printf("%s\n", error);
       LLVMDisposeMessage(error);
-      abort();
+      assert(0);
    }
 
    screen->target = LLVMGetExecutionEngineTargetData(screen->engine);
diff --git a/src/gallium/drivers/llvmpipe/lp_screen.c b/src/gallium/drivers/llvmpipe/lp_screen.c
index 05189274589..190dad0f26a 100644
--- a/src/gallium/drivers/llvmpipe/lp_screen.c
+++ b/src/gallium/drivers/llvmpipe/lp_screen.c
@@ -27,6 +27,7 @@
 
 
 #include "util/u_memory.h"
+#include "util/u_format.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_screen.h"
 
@@ -35,6 +36,24 @@
 #include "lp_winsys.h"
 #include "lp_jit.h"
 #include "lp_screen.h"
+#include "lp_debug.h"
+
+#ifdef DEBUG
+int LP_DEBUG = 0;
+
+static const struct debug_named_value lp_debug_flags[] = {
+   { "pipe",   DEBUG_PIPE },
+   { "tgsi",   DEBUG_TGSI },
+   { "tex",    DEBUG_TEX },
+   { "asm",    DEBUG_ASM },
+   { "setup",  DEBUG_SETUP },
+   { "rast",   DEBUG_RAST },
+   { "query",  DEBUG_QUERY },
+   { "screen", DEBUG_SCREEN },
+   { "jit",    DEBUG_JIT },
+   {NULL, 0}
+};
+#endif
 
 
 static const char *
@@ -131,17 +150,17 @@ llvmpipe_is_format_supported( struct pipe_screen *_screen,
 {
    struct llvmpipe_screen *screen = llvmpipe_screen(_screen);
    struct llvmpipe_winsys *winsys = screen->winsys;
+   const struct util_format_description *format_desc;
+
+   format_desc = util_format_description(format);
+   if(!format_desc)
+      return FALSE;
 
    assert(target == PIPE_TEXTURE_1D ||
           target == PIPE_TEXTURE_2D ||
           target == PIPE_TEXTURE_3D ||
           target == PIPE_TEXTURE_CUBE);
 
-   if(format == PIPE_FORMAT_Z16_UNORM)
-      return FALSE;
-   if(format == PIPE_FORMAT_S8_UNORM)
-      return FALSE;
-
    switch(format) {
    case PIPE_FORMAT_DXT1_RGB:
    case PIPE_FORMAT_DXT1_RGBA:
@@ -152,8 +171,51 @@ llvmpipe_is_format_supported( struct pipe_screen *_screen,
       break;
    }
 
-   if(tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET)
-      return winsys->is_displaytarget_format_supported(winsys, format);
+   if(tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) {
+      if(format_desc->block.width != 1 ||
+         format_desc->block.height != 1)
+         return FALSE;
+
+      if(format_desc->layout != UTIL_FORMAT_LAYOUT_SCALAR &&
+         format_desc->layout != UTIL_FORMAT_LAYOUT_ARITH &&
+         format_desc->layout != UTIL_FORMAT_LAYOUT_ARRAY)
+         return FALSE;
+
+      if(format_desc->colorspace != UTIL_FORMAT_COLORSPACE_RGB &&
+         format_desc->colorspace != UTIL_FORMAT_COLORSPACE_SRGB)
+         return FALSE;
+   }
+
+   if(tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) {
+      if(!winsys->is_displaytarget_format_supported(winsys, format))
+         return FALSE;
+   }
+
+   if(tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) {
+      if(format_desc->colorspace != UTIL_FORMAT_COLORSPACE_ZS)
+         return FALSE;
+
+      /* FIXME: Temporary restriction. See lp_state_fs.c. */
+      if(format_desc->block.bits != 32)
+         return FALSE;
+   }
+
+   /* FIXME: Temporary restrictions. See lp_bld_sample_soa.c */
+   if(tex_usage & PIPE_TEXTURE_USAGE_SAMPLER) {
+      if(format_desc->block.width != 1 ||
+         format_desc->block.height != 1)
+         return FALSE;
+
+      if(format_desc->layout != UTIL_FORMAT_LAYOUT_SCALAR &&
+         format_desc->layout != UTIL_FORMAT_LAYOUT_ARITH &&
+         format_desc->layout != UTIL_FORMAT_LAYOUT_ARRAY)
+         return FALSE;
+
+      if(format_desc->colorspace != UTIL_FORMAT_COLORSPACE_RGB &&
+         format_desc->colorspace != UTIL_FORMAT_COLORSPACE_SRGB &&
+         format_desc->colorspace != UTIL_FORMAT_COLORSPACE_ZS)
+         return FALSE;
+   }
 
    return TRUE;
 }
@@ -213,6 +275,10 @@ llvmpipe_create_screen(struct llvmpipe_winsys *winsys)
 {
    struct llvmpipe_screen *screen = CALLOC_STRUCT(llvmpipe_screen);
 
+#ifdef DEBUG
+   LP_DEBUG = debug_get_flags_option("LP_DEBUG", lp_debug_flags, 0 );
+#endif
+
    if (!screen)
       return NULL;
 
diff --git a/src/gallium/drivers/llvmpipe/lp_state_blend.c b/src/gallium/drivers/llvmpipe/lp_state_blend.c
index b2e75d3b14e..a94cd05ef20 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_blend.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_blend.c
@@ -34,6 +34,7 @@
 #include "util/u_memory.h"
 #include "util/u_math.h"
 #include "util/u_debug_dump.h"
+#include "draw/draw_context.h"
 #include "lp_screen.h"
 #include "lp_context.h"
 #include "lp_state.h"
@@ -51,6 +52,11 @@ void llvmpipe_bind_blend_state( struct pipe_context *pipe,
 {
    struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
 
+   if (llvmpipe->blend == blend)
+      return;
+
+   draw_flush(llvmpipe->draw);
+
    llvmpipe->blend = blend;
 
    llvmpipe->dirty |= LP_NEW_BLEND;
@@ -69,6 +75,11 @@ void llvmpipe_set_blend_color( struct pipe_context *pipe,
    struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
    unsigned i, j;
 
+   if(memcmp(&llvmpipe->blend_color, blend_color, sizeof *blend_color) == 0)
+      return;
+
+   draw_flush(llvmpipe->draw);
+
    memcpy(&llvmpipe->blend_color, blend_color, sizeof *blend_color);
 
    if(!llvmpipe->jit_context.blend_color)
@@ -99,7 +110,12 @@ llvmpipe_bind_depth_stencil_state(struct pipe_context *pipe,
 {
    struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
 
-   llvmpipe->depth_stencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil;
+   if (llvmpipe->depth_stencil == depth_stencil)
+      return;
+
+   draw_flush(llvmpipe->draw);
+
+   llvmpipe->depth_stencil = depth_stencil;
 
    if(llvmpipe->depth_stencil)
       llvmpipe->jit_context.alpha_ref_value = llvmpipe->depth_stencil->alpha.ref_value;
diff --git a/src/gallium/drivers/llvmpipe/lp_state_fs.c b/src/gallium/drivers/llvmpipe/lp_state_fs.c
index 2297cbb76f8..2b6965deca8 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_fs.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_fs.c
@@ -85,6 +85,7 @@
 #include "lp_buffer.h"
 #include "lp_state.h"
 #include "lp_tex_sample.h"
+#include "lp_debug.h"
 
 
 static const unsigned char quad_offset_x[4] = {0, 1, 0, 1};
@@ -146,6 +147,20 @@ generate_depth(LLVMBuilderRef builder,
    format_desc = util_format_description(key->zsbuf_format);
    assert(format_desc);
 
+   /*
+    * Depths are expected to be between 0 and 1, even if they are stored in
+    * floats. Setting these bits here will ensure that the lp_build_conv() call
+    * below won't try to unnecessarily clamp the incoming values.
+    */
+   if(src_type.floating) {
+      src_type.sign = FALSE;
+      src_type.norm = TRUE;
+   }
+   else {
+      assert(!src_type.sign);
+      assert(src_type.norm);
+   }
+
    /* Pick the depth type. */
    dst_type = lp_depth_type(format_desc, src_type.width*src_type.length);
 
@@ -153,14 +168,11 @@ generate_depth(LLVMBuilderRef builder,
    assert(dst_type.width == src_type.width);
    assert(dst_type.length == src_type.length);
 
-#if 1
-   src = lp_build_clamped_float_to_unsigned_norm(builder,
-                                                 src_type,
-                                                 dst_type.width,
-                                                 src);
-#else
    lp_build_conv(builder, src_type, dst_type, &src, 1, &src, 1);
-#endif
+
+   dst_ptr = LLVMBuildBitCast(builder,
+                              dst_ptr,
+                              LLVMPointerType(lp_build_vec_type(dst_type), 0), "");
 
    lp_build_depth_test(builder,
                        &key->depth,
@@ -395,59 +407,58 @@ generate_fragment(struct llvmpipe_context *lp,
    unsigned i;
    unsigned chan;
 
-#ifdef DEBUG
-   tgsi_dump(shader->base.tokens, 0);
-   if(key->depth.enabled) {
-      debug_printf("depth.format = %s\n", pf_name(key->zsbuf_format));
-      debug_printf("depth.func = %s\n", debug_dump_func(key->depth.func, TRUE));
-      debug_printf("depth.writemask = %u\n", key->depth.writemask);
-   }
-   if(key->alpha.enabled) {
-      debug_printf("alpha.func = %s\n", debug_dump_func(key->alpha.func, TRUE));
-      debug_printf("alpha.ref_value = %f\n", key->alpha.ref_value);
-   }
-   if(key->blend.logicop_enable) {
-      debug_printf("blend.logicop_func = %u\n", key->blend.logicop_func);
-   }
-   else if(key->blend.blend_enable) {
-      debug_printf("blend.rgb_func = %s\n",   debug_dump_blend_func  (key->blend.rgb_func, TRUE));
-      debug_printf("rgb_src_factor = %s\n",   debug_dump_blend_factor(key->blend.rgb_src_factor, TRUE));
-      debug_printf("rgb_dst_factor = %s\n",   debug_dump_blend_factor(key->blend.rgb_dst_factor, TRUE));
-      debug_printf("alpha_func = %s\n",       debug_dump_blend_func  (key->blend.alpha_func, TRUE));
-      debug_printf("alpha_src_factor = %s\n", debug_dump_blend_factor(key->blend.alpha_src_factor, TRUE));
-      debug_printf("alpha_dst_factor = %s\n", debug_dump_blend_factor(key->blend.alpha_dst_factor, TRUE));
-   }
-   debug_printf("blend.colormask = 0x%x\n", key->blend.colormask);
-   for(i = 0; i < PIPE_MAX_SAMPLERS; ++i) {
-      if(key->sampler[i].format) {
-         debug_printf("sampler[%u] = \n", i);
-         debug_printf("  .format = %s\n",
-                      pf_name(key->sampler[i].format));
-         debug_printf("  .target = %s\n",
-                      debug_dump_tex_target(key->sampler[i].target, TRUE));
-         debug_printf("  .pot = %u %u %u\n",
-                      key->sampler[i].pot_width,
-                      key->sampler[i].pot_height,
-                      key->sampler[i].pot_depth);
-         debug_printf("  .wrap = %s %s %s\n",
-                      debug_dump_tex_wrap(key->sampler[i].wrap_s, TRUE),
-                      debug_dump_tex_wrap(key->sampler[i].wrap_t, TRUE),
-                      debug_dump_tex_wrap(key->sampler[i].wrap_r, TRUE));
-         debug_printf("  .min_img_filter = %s\n",
-                      debug_dump_tex_filter(key->sampler[i].min_img_filter, TRUE));
-         debug_printf("  .min_mip_filter = %s\n",
-                      debug_dump_tex_mipfilter(key->sampler[i].min_mip_filter, TRUE));
-         debug_printf("  .mag_img_filter = %s\n",
-                      debug_dump_tex_filter(key->sampler[i].mag_img_filter, TRUE));
-         if(key->sampler[i].compare_mode)
-            debug_printf("  .compare_mode = %s\n", debug_dump_func(key->sampler[i].compare_func, TRUE));
-         debug_printf("  .normalized_coords = %u\n", key->sampler[i].normalized_coords);
-         debug_printf("  .prefilter = %u\n", key->sampler[i].prefilter);
+   if (LP_DEBUG & DEBUG_JIT) {
+      tgsi_dump(shader->base.tokens, 0);
+      if(key->depth.enabled) {
+         debug_printf("depth.format = %s\n", pf_name(key->zsbuf_format));
+         debug_printf("depth.func = %s\n", debug_dump_func(key->depth.func, TRUE));
+         debug_printf("depth.writemask = %u\n", key->depth.writemask);
+      }
+      if(key->alpha.enabled) {
+         debug_printf("alpha.func = %s\n", debug_dump_func(key->alpha.func, TRUE));
+         debug_printf("alpha.ref_value = %f\n", key->alpha.ref_value);
+      }
+      if(key->blend.logicop_enable) {
+         debug_printf("blend.logicop_func = %u\n", key->blend.logicop_func);
+      }
+      else if(key->blend.blend_enable) {
+         debug_printf("blend.rgb_func = %s\n",   debug_dump_blend_func  (key->blend.rgb_func, TRUE));
+         debug_printf("rgb_src_factor = %s\n",   debug_dump_blend_factor(key->blend.rgb_src_factor, TRUE));
+         debug_printf("rgb_dst_factor = %s\n",   debug_dump_blend_factor(key->blend.rgb_dst_factor, TRUE));
+         debug_printf("alpha_func = %s\n",       debug_dump_blend_func  (key->blend.alpha_func, TRUE));
+         debug_printf("alpha_src_factor = %s\n", debug_dump_blend_factor(key->blend.alpha_src_factor, TRUE));
+         debug_printf("alpha_dst_factor = %s\n", debug_dump_blend_factor(key->blend.alpha_dst_factor, TRUE));
+      }
+      debug_printf("blend.colormask = 0x%x\n", key->blend.colormask);
+      for(i = 0; i < PIPE_MAX_SAMPLERS; ++i) {
+         if(key->sampler[i].format) {
+            debug_printf("sampler[%u] = \n", i);
+            debug_printf("  .format = %s\n",
+                         pf_name(key->sampler[i].format));
+            debug_printf("  .target = %s\n",
+                         debug_dump_tex_target(key->sampler[i].target, TRUE));
+            debug_printf("  .pot = %u %u %u\n",
+                         key->sampler[i].pot_width,
+                         key->sampler[i].pot_height,
+                         key->sampler[i].pot_depth);
+            debug_printf("  .wrap = %s %s %s\n",
+                         debug_dump_tex_wrap(key->sampler[i].wrap_s, TRUE),
+                         debug_dump_tex_wrap(key->sampler[i].wrap_t, TRUE),
+                         debug_dump_tex_wrap(key->sampler[i].wrap_r, TRUE));
+            debug_printf("  .min_img_filter = %s\n",
+                         debug_dump_tex_filter(key->sampler[i].min_img_filter, TRUE));
+            debug_printf("  .min_mip_filter = %s\n",
+                         debug_dump_tex_mipfilter(key->sampler[i].min_mip_filter, TRUE));
+            debug_printf("  .mag_img_filter = %s\n",
+                         debug_dump_tex_filter(key->sampler[i].mag_img_filter, TRUE));
+            if(key->sampler[i].compare_mode)
+               debug_printf("  .compare_mode = %s\n", debug_dump_func(key->sampler[i].compare_func, TRUE));
+            debug_printf("  .normalized_coords = %u\n", key->sampler[i].normalized_coords);
+            debug_printf("  .prefilter = %u\n", key->sampler[i].prefilter);
+         }
       }
    }
 
-#endif
-
    variant = CALLOC_STRUCT(lp_fragment_shader_variant);
    if(!variant)
       return NULL;
@@ -586,8 +597,8 @@ generate_fragment(struct llvmpipe_context *lp,
    }
 
    lp_build_conv_mask(builder, fs_type, blend_type,
-                               fs_mask, num_fs,
-                               &blend_mask, 1);
+                      fs_mask, num_fs,
+                      &blend_mask, 1);
 
    /*
     * Blending.
@@ -609,23 +620,24 @@ generate_fragment(struct llvmpipe_context *lp,
     * Translate the LLVM IR into machine code.
     */
 
+#ifdef DEBUG
    if(LLVMVerifyFunction(variant->function, LLVMPrintMessageAction)) {
       LLVMDumpValue(variant->function);
-      abort();
+      assert(0);
    }
+#endif
 
    LLVMRunFunctionPassManager(screen->pass, variant->function);
 
-#ifdef DEBUG
-   LLVMDumpValue(variant->function);
-   debug_printf("\n");
-#endif
+   if (LP_DEBUG & DEBUG_JIT) {
+      LLVMDumpValue(variant->function);
+      debug_printf("\n");
+   }
 
    variant->jit_function = (lp_jit_frag_func)LLVMGetPointerToGlobal(screen->engine, variant->function);
 
-#ifdef DEBUG
-   lp_disassemble(variant->jit_function);
-#endif
+   if (LP_DEBUG & DEBUG_ASM)
+      lp_disassemble(variant->jit_function);
 
    variant->next = shader->variants;
    shader->variants = variant;
@@ -659,7 +671,12 @@ llvmpipe_bind_fs_state(struct pipe_context *pipe, void *fs)
 {
    struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
 
-   llvmpipe->fs = (struct lp_fragment_shader *) fs;
+   if (llvmpipe->fs == fs)
+      return;
+
+   draw_flush(llvmpipe->draw);
+
+   llvmpipe->fs = fs;
 
    llvmpipe->dirty |= LP_NEW_FS;
 }
@@ -710,8 +727,7 @@ llvmpipe_set_constant_buffer(struct pipe_context *pipe,
    assert(shader < PIPE_SHADER_TYPES);
    assert(index == 0);
 
-   if(shader == PIPE_SHADER_VERTEX)
-      draw_flush(llvmpipe->draw);
+   draw_flush(llvmpipe->draw);
 
    /* note: reference counting */
    pipe_buffer_reference(&llvmpipe->constants[shader].buffer, buffer);
diff --git a/src/gallium/drivers/llvmpipe/lp_state_rasterizer.c b/src/gallium/drivers/llvmpipe/lp_state_rasterizer.c
index 4561c6b8456..aa3b5a3f91e 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_rasterizer.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_rasterizer.c
@@ -41,14 +41,17 @@ llvmpipe_create_rasterizer_state(struct pipe_context *pipe,
 }
 
 void llvmpipe_bind_rasterizer_state(struct pipe_context *pipe,
-                                    void *setup)
+                                    void *rasterizer)
 {
    struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
 
+   if (llvmpipe->rasterizer == rasterizer)
+      return;
+
    /* pass-through to draw module */
-   draw_set_rasterizer_state(llvmpipe->draw, setup);
+   draw_set_rasterizer_state(llvmpipe->draw, rasterizer);
 
-   llvmpipe->rasterizer = (struct pipe_rasterizer_state *)setup;
+   llvmpipe->rasterizer = rasterizer;
 
    llvmpipe->dirty |= LP_NEW_RASTERIZER;
 }
diff --git a/src/gallium/drivers/llvmpipe/lp_state_surface.c b/src/gallium/drivers/llvmpipe/lp_state_surface.c
index 9f745856742..4eb45970a4d 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_surface.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_surface.c
@@ -48,6 +48,8 @@ llvmpipe_set_framebuffer_state(struct pipe_context *pipe,
    struct llvmpipe_context *lp = llvmpipe_context(pipe);
    uint i;
 
+   draw_flush(lp->draw);
+
    for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) {
       /* check if changing cbuf */
       if (lp->framebuffer.cbufs[i] != fb->cbufs[i]) {
diff --git a/src/gallium/drivers/llvmpipe/lp_state_vs.c b/src/gallium/drivers/llvmpipe/lp_state_vs.c
index 15c30296144..884e3878e62 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_vs.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_vs.c
@@ -70,14 +70,18 @@ fail:
 
 
 void
-llvmpipe_bind_vs_state(struct pipe_context *pipe, void *vs)
+llvmpipe_bind_vs_state(struct pipe_context *pipe, void *_vs)
 {
    struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+   const struct lp_vertex_shader *vs = (const struct lp_vertex_shader *)_vs;
 
-   llvmpipe->vs = (const struct lp_vertex_shader *)vs;
+   if (llvmpipe->vs == vs)
+      return;
 
-   draw_bind_vertex_shader(llvmpipe->draw,
-                           (llvmpipe->vs ? llvmpipe->vs->draw_data : NULL));
+   draw_bind_vertex_shader(llvmpipe->draw, 
+                           vs ? vs->draw_data : NULL);
+
+   llvmpipe->vs = vs;
 
    llvmpipe->dirty |= LP_NEW_VS;
 }
@@ -92,5 +96,6 @@ llvmpipe_delete_vs_state(struct pipe_context *pipe, void *vs)
       (struct lp_vertex_shader *)vs;
 
    draw_delete_vertex_shader(llvmpipe->draw, state->draw_data);
+   FREE( (void *)state->shader.tokens );
    FREE( state );
 }
diff --git a/src/gallium/drivers/softpipe/sp_state_blend.c b/src/gallium/drivers/softpipe/sp_state_blend.c
index efed082f823..95ab3234337 100644
--- a/src/gallium/drivers/softpipe/sp_state_blend.c
+++ b/src/gallium/drivers/softpipe/sp_state_blend.c
@@ -29,6 +29,7 @@
  */
 
 #include "util/u_memory.h"
+#include "draw/draw_context.h"
 #include "sp_context.h"
 #include "sp_state.h"
 
@@ -45,6 +46,8 @@ void softpipe_bind_blend_state( struct pipe_context *pipe,
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
 
+   draw_flush(softpipe->draw);
+
    softpipe->blend = (struct pipe_blend_state *)blend;
 
    softpipe->dirty |= SP_NEW_BLEND;
@@ -62,6 +65,8 @@ void softpipe_set_blend_color( struct pipe_context *pipe,
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
 
+   draw_flush(softpipe->draw);
+
    softpipe->blend_color = *blend_color;
 
    softpipe->dirty |= SP_NEW_BLEND;
diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c
index b41f7e8ab72..8ea48ed3aae 100644
--- a/src/gallium/drivers/softpipe/sp_state_fs.c
+++ b/src/gallium/drivers/softpipe/sp_state_fs.c
@@ -69,7 +69,14 @@ softpipe_bind_fs_state(struct pipe_context *pipe, void *fs)
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
 
-   softpipe->fs = (struct sp_fragment_shader *) fs;
+   draw_flush(softpipe->draw);
+
+   if (softpipe->fs == fs)
+      return;
+
+   draw_flush(softpipe->draw);
+
+   softpipe->fs = fs;
 
    softpipe->dirty |= SP_NEW_FS;
 }
@@ -159,6 +166,8 @@ softpipe_set_constant_buffer(struct pipe_context *pipe,
    assert(shader < PIPE_SHADER_TYPES);
    assert(index == 0);
 
+   draw_flush(softpipe->draw);
+
    /* note: reference counting */
    pipe_buffer_reference(&softpipe->constants[shader].buffer,
 			 buf ? buf->buffer : NULL);
diff --git a/src/gallium/drivers/softpipe/sp_state_rasterizer.c b/src/gallium/drivers/softpipe/sp_state_rasterizer.c
index 87b72196838..a5b00336d44 100644
--- a/src/gallium/drivers/softpipe/sp_state_rasterizer.c
+++ b/src/gallium/drivers/softpipe/sp_state_rasterizer.c
@@ -41,14 +41,17 @@ softpipe_create_rasterizer_state(struct pipe_context *pipe,
 }
 
 void softpipe_bind_rasterizer_state(struct pipe_context *pipe,
-                                    void *setup)
+                                    void *rasterizer)
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
 
+   if (softpipe->rasterizer == rasterizer)
+      return;
+
    /* pass-through to draw module */
-   draw_set_rasterizer_state(softpipe->draw, setup);
+   draw_set_rasterizer_state(softpipe->draw, rasterizer);
 
-   softpipe->rasterizer = (struct pipe_rasterizer_state *)setup;
+   softpipe->rasterizer = rasterizer;
 
    softpipe->dirty |= SP_NEW_RASTERIZER;
 }
diff --git a/src/gallium/drivers/softpipe/sp_state_surface.c b/src/gallium/drivers/softpipe/sp_state_surface.c
index 794a9225b86..d29b6aeb78b 100644
--- a/src/gallium/drivers/softpipe/sp_state_surface.c
+++ b/src/gallium/drivers/softpipe/sp_state_surface.c
@@ -48,6 +48,8 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe,
    struct softpipe_context *sp = softpipe_context(pipe);
    uint i;
 
+   draw_flush(sp->draw);
+
    for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) {
       /* check if changing cbuf */
       if (sp->framebuffer.cbufs[i] != fb->cbufs[i]) {
diff --git a/src/gallium/drivers/svga/svga_context.c b/src/gallium/drivers/svga/svga_context.c
index af99c9de37c..1e76592ec60 100644
--- a/src/gallium/drivers/svga/svga_context.c
+++ b/src/gallium/drivers/svga/svga_context.c
@@ -215,7 +215,6 @@ struct pipe_context *svga_context_create( struct pipe_screen *screen )
    svga->state.hw_draw.num_views = 0;
 
    svga->dirty = ~0;
-   svga->state.white_fs_id = SVGA3D_INVALID_ID;
 
    LIST_INITHEAD(&svga->dirty_buffers);
 
diff --git a/src/gallium/drivers/svga/svga_context.h b/src/gallium/drivers/svga/svga_context.h
index 32e9304f819..52d47046e74 100644
--- a/src/gallium/drivers/svga/svga_context.h
+++ b/src/gallium/drivers/svga/svga_context.h
@@ -329,10 +329,6 @@ struct svga_context
 
       unsigned texture_timestamp;
 
-      /* Internally generated shaders:
-       */
-      unsigned white_fs_id;
-
       /* 
        */
       struct svga_sw_state          sw;
diff --git a/src/gallium/drivers/svga/svga_pipe_blend.c b/src/gallium/drivers/svga/svga_pipe_blend.c
index 3ad3f97816c..c0cbda3af4b 100644
--- a/src/gallium/drivers/svga/svga_pipe_blend.c
+++ b/src/gallium/drivers/svga/svga_pipe_blend.c
@@ -92,6 +92,7 @@ svga_create_blend_state(struct pipe_context *pipe,
       if (templ->logicop_enable) {
          switch (templ->logicop_func) {
          case PIPE_LOGICOP_XOR:
+         case PIPE_LOGICOP_INVERT:
             blend->need_white_fragments = TRUE;
             blend->rt[i].blend_enable = TRUE;
             blend->rt[i].srcblend       = SVGA3D_BLENDOP_ONE;
@@ -125,12 +126,6 @@ svga_create_blend_state(struct pipe_context *pipe,
             blend->rt[i].dstblend       = SVGA3D_BLENDOP_ONE;
             blend->rt[i].blendeq        = SVGA3D_BLENDEQ_MAXIMUM;
             break;
-         case PIPE_LOGICOP_INVERT:
-            blend->rt[i].blend_enable = TRUE;
-            blend->rt[i].srcblend       = SVGA3D_BLENDOP_INVSRCCOLOR;
-            blend->rt[i].dstblend       = SVGA3D_BLENDOP_ZERO;
-            blend->rt[i].blendeq        = SVGA3D_BLENDEQ_ADD;
-            break;
          case PIPE_LOGICOP_AND:
             /* Approximate with minimum - works for the 0 & anything case: */
             blend->rt[i].blend_enable = TRUE;
diff --git a/src/gallium/drivers/svga/svga_pipe_draw.c b/src/gallium/drivers/svga/svga_pipe_draw.c
index 719b3419f86..8794deeb0a3 100644
--- a/src/gallium/drivers/svga/svga_pipe_draw.c
+++ b/src/gallium/drivers/svga/svga_pipe_draw.c
@@ -218,9 +218,6 @@ svga_draw_range_elements( struct pipe_context *pipe,
    if (SVGA_DEBUG & DEBUG_FLUSH) {
       static unsigned id;
       debug_printf("%s %d\n", __FUNCTION__, id++);
-      if (id > 1300)
-         util_time_sleep( 2000 );
-
       svga_hwtnl_flush_retry( svga );
       svga_context_flush(svga, NULL);
    }
diff --git a/src/gallium/drivers/svga/svga_screen_buffer.c b/src/gallium/drivers/svga/svga_screen_buffer.c
index 719adde27ef..6a58ff238ad 100644
--- a/src/gallium/drivers/svga/svga_screen_buffer.c
+++ b/src/gallium/drivers/svga/svga_screen_buffer.c
@@ -310,10 +310,20 @@ svga_buffer_upload_queue(struct svga_buffer *sbuf,
                          unsigned end)
 {
    unsigned i;
+   unsigned nearest_range;
+   unsigned nearest_dist;
 
    assert(sbuf->hw.buf);
    assert(end > start);
    
+   if (sbuf->hw.num_ranges < SVGA_BUFFER_MAX_RANGES) {
+      nearest_range = sbuf->hw.num_ranges;
+      nearest_dist = ~0;
+   } else {
+      nearest_range = SVGA_BUFFER_MAX_RANGES - 1;
+      nearest_dist = 0;
+   }
+
    /*
     * Try to grow one of the ranges.
     *
@@ -325,11 +335,33 @@ svga_buffer_upload_queue(struct svga_buffer *sbuf,
     */
 
    for(i = 0; i < sbuf->hw.num_ranges; ++i) {
-      if(start <= sbuf->hw.ranges[i].end && sbuf->hw.ranges[i].start <= end) {
+      int left_dist;
+      int right_dist;
+      int dist;
+
+      left_dist = start - sbuf->hw.ranges[i].end;
+      right_dist = sbuf->hw.ranges[i].start - end;
+      dist = MAX2(left_dist, right_dist);
+
+      if (dist <= 0) {
+         /*
+          * Ranges are contiguous or overlapping -- extend this one and return.
+          */
+
          sbuf->hw.ranges[i].start = MIN2(sbuf->hw.ranges[i].start, start);
-         sbuf->hw.ranges[i].end   = MAX2(sbuf->hw.ranges[i].end,    end);
+         sbuf->hw.ranges[i].end   = MAX2(sbuf->hw.ranges[i].end,   end);
          return;
       }
+      else {
+         /*
+          * Discontiguous ranges -- keep track of the nearest range.
+          */
+
+         if (dist < nearest_dist) {
+            nearest_range = i;
+            nearest_dist = dist;
+         }
+      }
    }
 
    /*
@@ -344,13 +376,27 @@ svga_buffer_upload_queue(struct svga_buffer *sbuf,
    assert(!sbuf->hw.svga);
    assert(!sbuf->hw.boxes);
 
-   /*
-    * Add a new range.
-    */
+   if (sbuf->hw.num_ranges < SVGA_BUFFER_MAX_RANGES) {
+      /*
+       * Add a new range.
+       */
 
-   sbuf->hw.ranges[sbuf->hw.num_ranges].start = start;
-   sbuf->hw.ranges[sbuf->hw.num_ranges].end = end;
-   ++sbuf->hw.num_ranges;
+      sbuf->hw.ranges[sbuf->hw.num_ranges].start = start;
+      sbuf->hw.ranges[sbuf->hw.num_ranges].end = end;
+      ++sbuf->hw.num_ranges;
+   } else {
+      /*
+       * Everything else failed, so just extend the nearest range.
+       *
+       * It is OK to do this because we always keep a local copy of the
+       * host buffer data, for SW TNL, and the host never modifies the buffer.
+       */
+
+      assert(nearest_range < SVGA_BUFFER_MAX_RANGES);
+      assert(nearest_range < sbuf->hw.num_ranges);
+      sbuf->hw.ranges[nearest_range].start = MIN2(sbuf->hw.ranges[nearest_range].start, start);
+      sbuf->hw.ranges[nearest_range].end   = MAX2(sbuf->hw.ranges[nearest_range].end,   end);
+   }
 }
 
 
diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c
index 9ad4edafef2..c3819882b41 100644
--- a/src/gallium/drivers/svga/svga_screen_texture.c
+++ b/src/gallium/drivers/svga/svga_screen_texture.c
@@ -203,7 +203,7 @@ svga_transfer_dma(struct svga_transfer *st,
       if(transfer == SVGA3D_READ_HOST_VRAM) {
          svga_screen_flush(screen, &fence);
          sws->fence_finish(sws, fence, 0);
-         //sws->fence_reference(sws, &fence, NULL);
+         sws->fence_reference(sws, &fence, NULL);
       }
    }
    else {
@@ -232,7 +232,7 @@ svga_transfer_dma(struct svga_transfer *st,
             if(y) {
                svga_screen_flush(screen, &fence);
                sws->fence_finish(sws, fence, 0);
-               //sws->fence_reference(sws, &fence, NULL);
+               sws->fence_reference(sws, &fence, NULL);
             }
 
             hw = sws->buffer_map(sws, st->hwbuf, PIPE_BUFFER_USAGE_CPU_WRITE);
diff --git a/src/gallium/drivers/svga/svga_state_fs.c b/src/gallium/drivers/svga/svga_state_fs.c
index 4fe91416778..3b700cd0227 100644
--- a/src/gallium/drivers/svga/svga_state_fs.c
+++ b/src/gallium/drivers/svga/svga_state_fs.c
@@ -103,70 +103,6 @@ fail:
    return ret;
 }
 
-/* The blend workaround for simulating logicop xor behaviour requires
- * that the incoming fragment color be white.  This change achieves
- * that by hooking up a hard-wired fragment shader that just emits
- * color 1,1,1,1
- *   
- * This is a slightly incomplete solution as it assumes that the
- * actual bound shader has no other effects beyond generating a
- * fragment color.  In particular shaders containing TEXKIL and/or
- * depth-write will not have the correct behaviour, nor will those
- * expecting to use alphatest.
- *   
- * These are avoidable issues, but they are not much worse than the
- * unavoidable ones associated with this technique, so it's not clear
- * how much effort should be expended trying to resolve them - the
- * ultimate result will still not be correct in most cases.
- *
- * Shader below was generated with:
- *   SVGA_DEBUG=tgsi ./mesa/progs/fp/fp-tri white.txt
- */
-static int emit_white_fs( struct svga_context *svga )
-{
-   int ret = PIPE_ERROR;
-
-   /* ps_3_0
-    * def c0, 1.000000, 0.000000, 0.000000, 1.000000
-    * mov oC0, c0.x
-    * end
-    */
-   static const unsigned white_tokens[] = {
-      0xffff0300,
-      0x05000051,
-      0xa00f0000,
-      0x3f800000,
-      0x00000000,
-      0x00000000,
-      0x3f800000,
-      0x02000001,
-      0x800f0800,
-      0xa0000000,
-      0x0000ffff,
-   };
-
-   assert(SVGA3D_INVALID_ID == UTIL_BITMASK_INVALID_INDEX);
-   svga->state.white_fs_id = util_bitmask_add(svga->fs_bm);
-   if(svga->state.white_fs_id == SVGA3D_INVALID_ID)
-      goto no_fs_id;
-
-   ret = SVGA3D_DefineShader(svga->swc, 
-                             svga->state.white_fs_id,
-                             SVGA3D_SHADERTYPE_PS,
-                             white_tokens, 
-                             sizeof(white_tokens));
-   if (ret)
-      goto no_definition;
-
-   return 0;
-
-no_definition:
-   util_bitmask_clear(svga->fs_bm, svga->state.white_fs_id);
-   svga->state.white_fs_id = SVGA3D_INVALID_ID;
-no_fs_id:
-   return ret;
-}
-
 
 /* SVGA_NEW_TEXTURE_BINDING
  * SVGA_NEW_RAST
@@ -194,6 +130,23 @@ static int make_fs_key( const struct svga_context *svga,
                        PIPE_WINDING_CW);
    }
 
+   /* The blend workaround for simulating logicop xor behaviour
+    * requires that the incoming fragment color be white.  This change
+    * achieves that by creating a varient of the current fragment
+    * shader that overrides all output colors with 1,1,1,1
+    *   
+    * This will work for most shaders, including those containing
+    * TEXKIL and/or depth-write.  However, it will break on the
+    * combination of xor-logicop plus alphatest.
+    *
+    * Ultimately, we could implement alphatest in the shader using
+    * texkil prior to overriding the outgoing fragment color.
+    *   
+    * SVGA_NEW_BLEND
+    */
+   if (svga->curr.blend->need_white_fragments) {
+      key->white_fragments = 1;
+   }
    
    /* XXX: want to limit this to the textures that the shader actually
     * refers to.
@@ -233,40 +186,29 @@ static int emit_hw_fs( struct svga_context *svga,
    unsigned id = SVGA3D_INVALID_ID;
    int ret = 0;
 
+   struct svga_fragment_shader *fs = svga->curr.fs;
+   struct svga_fs_compile_key key;
+
    /* SVGA_NEW_BLEND
+    * SVGA_NEW_TEXTURE_BINDING
+    * SVGA_NEW_RAST
+    * SVGA_NEW_NEED_SWTNL
+    * SVGA_NEW_SAMPLER
     */
-   if (svga->curr.blend->need_white_fragments) {
-      if (svga->state.white_fs_id == SVGA3D_INVALID_ID) {
-         ret = emit_white_fs( svga );
-         if (ret)
-            return ret;
-      }
-      id = svga->state.white_fs_id;
-   }
-   else {
-      struct svga_fragment_shader *fs = svga->curr.fs;
-      struct svga_fs_compile_key key;
-
-      /* SVGA_NEW_TEXTURE_BINDING
-       * SVGA_NEW_RAST
-       * SVGA_NEW_NEED_SWTNL
-       * SVGA_NEW_SAMPLER
-       */
-      ret = make_fs_key( svga, &key );
+   ret = make_fs_key( svga, &key );
+   if (ret)
+      return ret;
+
+   result = search_fs_key( fs, &key );
+   if (!result) {
+      ret = compile_fs( svga, fs, &key, &result );
       if (ret)
          return ret;
-
-      result = search_fs_key( fs, &key );
-      if (!result) {
-         ret = compile_fs( svga, fs, &key, &result );
-         if (ret)
-            return ret;
-      }
-
-      assert (result);
-      id = result->id;
    }
 
+   assert (result);
+   id = result->id;
+
    assert(id != SVGA3D_INVALID_ID);
 
    if (result != svga->state.hw_draw.fs) {
diff --git a/src/gallium/drivers/svga/svga_tgsi.h b/src/gallium/drivers/svga/svga_tgsi.h
index 896c90a89ae..25f60c78b7c 100644
--- a/src/gallium/drivers/svga/svga_tgsi.h
+++ b/src/gallium/drivers/svga/svga_tgsi.h
@@ -49,6 +49,7 @@ struct svga_fs_compile_key
 {
    boolean light_twoside:1;
    boolean front_cw:1;
+   boolean white_fragments:1;
    ubyte num_textures;
    ubyte num_unnormalized_coords;
    struct {
diff --git a/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c b/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c
index 2291cf116dd..4c078ba9117 100644
--- a/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c
+++ b/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c
@@ -194,8 +194,19 @@ static boolean ps30_output( struct svga_shader_emitter *emit,
 
    switch (semantic.SemanticName) {
    case TGSI_SEMANTIC_COLOR:
-      emit->output_map[idx] = dst_register( SVGA3DREG_COLOROUT, 
-                                            semantic.SemanticIndex );
+      if (emit->unit == PIPE_SHADER_FRAGMENT &&
+          emit->key.fkey.white_fragments) {
+
+         emit->output_map[idx] = dst_register( SVGA3DREG_TEMP,
+                                               emit->nr_hw_temp++ );
+         emit->temp_col[idx] = emit->output_map[idx];
+         emit->true_col[idx] = dst_register( SVGA3DREG_COLOROUT, 
+                                              semantic.SemanticIndex );
+      }
+      else {
+         emit->output_map[idx] = dst_register( SVGA3DREG_COLOROUT, 
+                                               semantic.SemanticIndex );
+      }
       break;
    case TGSI_SEMANTIC_POSITION:
       emit->output_map[idx] = dst_register( SVGA3DREG_TEMP,
diff --git a/src/gallium/drivers/svga/svga_tgsi_insn.c b/src/gallium/drivers/svga/svga_tgsi_insn.c
index ea409b7e165..b5424dacf7d 100644
--- a/src/gallium/drivers/svga/svga_tgsi_insn.c
+++ b/src/gallium/drivers/svga/svga_tgsi_insn.c
@@ -1011,10 +1011,10 @@ static boolean emit_kilp(struct svga_shader_emitter *emit,
 {
    SVGA3dShaderInstToken inst;
    SVGA3dShaderDestToken temp;
-   struct src_register one = get_zero_immediate( emit );
+   struct src_register one = scalar( get_zero_immediate( emit ),
+                                     TGSI_SWIZZLE_W );
 
    inst = inst_token( SVGA3DOP_TEXKILL );
-   one = scalar( one, TGSI_SWIZZLE_W );
 
    /* texkill doesn't allow negation on the operand so lets move
     * negation of {1} to a temp register */
@@ -2254,11 +2254,28 @@ static boolean emit_ps_postamble( struct svga_shader_emitter *emit )
    for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) {
       if (SVGA3dShaderGetRegType(emit->true_col[i].value) != 0) {
 
-         if (!submit_op1( emit,
-                          inst_token(SVGA3DOP_MOV),
-                          emit->true_col[i],
-                          src(emit->temp_col[i]) ))
-            return FALSE;
+         /* Potentially override output colors with white for XOR
+          * logicop workaround.
+          */
+         if (emit->unit == PIPE_SHADER_FRAGMENT &&
+             emit->key.fkey.white_fragments) {
+
+            struct src_register one = scalar( get_zero_immediate( emit ),
+                                              TGSI_SWIZZLE_W );
+
+            if (!submit_op1( emit,
+                             inst_token(SVGA3DOP_MOV),
+                             emit->true_col[i],
+                             one ))
+               return FALSE;
+         }
+         else {
+            if (!submit_op1( emit,
+                             inst_token(SVGA3DOP_MOV),
+                             emit->true_col[i],
+                             src(emit->temp_col[i]) ))
+               return FALSE;
+         }
       }
    }
 
@@ -2467,6 +2484,9 @@ needs_to_create_zero( struct svga_shader_emitter *emit )
       if (emit->key.fkey.light_twoside)
          return TRUE;
 
+      if (emit->key.fkey.white_fragments)
+         return TRUE;
+
       if (emit->emit_frontface)
          return TRUE;
 
diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py
index 348f2e43683..18fa84f2a4c 100755
--- a/src/gallium/state_trackers/python/retrace/interpreter.py
+++ b/src/gallium/state_trackers/python/retrace/interpreter.py
@@ -557,7 +557,7 @@ class Context(Object):
 
         sys.stdout.write('\t{\n')
         for i in range(start, start + count):
-            if i >= start + 16:
+            if i >= start + 16 and not self.interpreter.verbosity(3):
                 sys.stdout.write('\t...\n')
                 break
             offset = i*isize
diff --git a/src/gallium/state_trackers/xorg/xorg_crtc.c b/src/gallium/state_trackers/xorg/xorg_crtc.c
index 0160b1aa59f..daa9f8b8204 100644
--- a/src/gallium/state_trackers/xorg/xorg_crtc.c
+++ b/src/gallium/state_trackers/xorg/xorg_crtc.c
@@ -243,7 +243,11 @@ crtc_load_cursor_argb_kms(xf86CrtcPtr crtc, CARD32 * image)
 	unsigned attr[8];
 
 	attr[0] = KMS_BO_TYPE;
+#ifdef KMS_BO_TYPE_CURSOR_64X64_A8R8G8B8
+	attr[1] = KMS_BO_TYPE_CURSOR_64X64_A8R8G8B8;
+#else
 	attr[1] = KMS_BO_TYPE_CURSOR;
+#endif
 	attr[2] = KMS_WIDTH;
 	attr[3] = 64;
 	attr[4] = KMS_HEIGHT;
diff --git a/src/gallium/state_trackers/xorg/xorg_driver.c b/src/gallium/state_trackers/xorg/xorg_driver.c
index e4ad789e9bc..f53a879a14a 100644
--- a/src/gallium/state_trackers/xorg/xorg_driver.c
+++ b/src/gallium/state_trackers/xorg/xorg_driver.c
@@ -1008,7 +1008,11 @@ drv_create_front_buffer_kms(ScrnInfoPtr pScrn)
     int ret;
 
     attr[0] = KMS_BO_TYPE;
+#ifdef KMS_BO_TYPE_SCANOUT_X8R8G8B8
+    attr[1] = KMS_BO_TYPE_SCANOUT_X8R8G8B8;
+#else
     attr[1] = KMS_BO_TYPE_SCANOUT;
+#endif
     attr[2] = KMS_WIDTH;
     attr[3] = pScrn->virtualX;
     attr[4] = KMS_HEIGHT;
diff --git a/src/gallium/winsys/drm/vmware/core/vmwgfx_drm.h b/src/gallium/winsys/drm/vmware/core/vmwgfx_drm.h
index 2be7e1249b6..681d6b5f996 100644
--- a/src/gallium/winsys/drm/vmware/core/vmwgfx_drm.h
+++ b/src/gallium/winsys/drm/vmware/core/vmwgfx_drm.h
@@ -87,49 +87,6 @@ struct drm_vmw_getparam_arg {
 
 /*************************************************************************/
 /**
- * DRM_VMW_EXTENSION - Query device extensions.
- */
-
-/**
- * struct drm_vmw_extension_rep
- *
- * @exists: The queried extension exists.
- * @driver_ioctl_offset: Ioctl number of the first ioctl in the extension.
- * @driver_sarea_offset: Offset to any space in the DRI SAREA
- * used by the extension.
- * @major: Major version number of the extension.
- * @minor: Minor version number of the extension.
- * @pl: Patch level version number of the extension.
- *
- * Output argument to the DRM_VMW_EXTENSION Ioctl.
- */
-
-struct drm_vmw_extension_rep {
-	int32_t exists;
-	uint32_t driver_ioctl_offset;
-	uint32_t driver_sarea_offset;
-	uint32_t major;
-	uint32_t minor;
-	uint32_t pl;
-	uint32_t pad64;
-};
-
-/**
- * union drm_vmw_extension_arg
- *
- * @extension - Ascii name of the extension to be queried. //In
- * @rep - Reply as defined above. //Out
- *
- * Argument to the DRM_VMW_EXTENSION Ioctl.
- */
-
-union drm_vmw_extension_arg {
-	char extension[DRM_VMW_EXT_NAME_LEN];
-	struct drm_vmw_extension_rep rep;
-};
-
-/*************************************************************************/
-/**
  * DRM_VMW_CREATE_CONTEXT - Create a host context.
  *
  * Allocates a device unique context id, and queues a create context command
diff --git a/src/glx/x11/dri2_glx.c b/src/glx/x11/dri2_glx.c
index 89efe3ab292..9988fa1b132 100644
--- a/src/glx/x11/dri2_glx.c
+++ b/src/glx/x11/dri2_glx.c
@@ -437,8 +437,10 @@ dri2CreateScreen(__GLXscreenConfigs * psc, int screen,
    psc->ext_list_first_time = GL_TRUE;
 
    if (!DRI2Connect(psc->dpy, RootWindow(psc->dpy, screen),
-                    &driverName, &deviceName))
+                    &driverName, &deviceName)) {
+      XFree(psp);
       return NULL;
+   }
 
    psc->driver = driOpenDriver(driverName);
    if (psc->driver == NULL) {
@@ -467,17 +469,17 @@ dri2CreateScreen(__GLXscreenConfigs * psc, int screen,
    psc->fd = open(deviceName, O_RDWR);
    if (psc->fd < 0) {
       ErrorMessageF("failed to open drm device: %s\n", strerror(errno));
-      return NULL;
+      goto handle_error;
    }
 
    if (drmGetMagic(psc->fd, &magic)) {
       ErrorMessageF("failed to get magic\n");
-      return NULL;
+      goto handle_error;
    }
 
    if (!DRI2Authenticate(psc->dpy, RootWindow(psc->dpy, screen), magic)) {
       ErrorMessageF("failed to authenticate magic %d\n", magic);
-      return NULL;
+      goto handle_error;
    }
 
    /* If the server does not support the protocol for
@@ -491,7 +493,7 @@ dri2CreateScreen(__GLXscreenConfigs * psc, int screen,
 
    if (psc->__driScreen == NULL) {
       ErrorMessageF("failed to create dri screen\n");
-      return NULL;
+      goto handle_error;
    }
 
    driBindExtensions(psc, 1);
@@ -521,6 +523,7 @@ dri2CreateScreen(__GLXscreenConfigs * psc, int screen,
  handle_error:
    Xfree(driverName);
    Xfree(deviceName);
+   XFree(psp);
 
    /* FIXME: clean up here */
 
diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c
index dbd233729c0..5cfa490a7e0 100644
--- a/src/mesa/drivers/dri/r600/r600_context.c
+++ b/src/mesa/drivers/dri/r600/r600_context.c
@@ -235,12 +235,11 @@ static void r600_init_vtbl(radeonContextPtr radeon)
 
 static void r600InitConstValues(GLcontext *ctx, radeonScreenPtr screen)
 {
-	context_t *r600 = R700_CONTEXT(ctx);
-
-	ctx->Const.MaxTextureImageUnits =
-	    driQueryOptioni(&r600->radeon.optionCache, "texture_image_units");
-	ctx->Const.MaxTextureCoordUnits =
-	    driQueryOptioni(&r600->radeon.optionCache, "texture_coord_units");
+	ctx->Const.MaxTextureImageUnits = 16;
+	/* 8 per clause on r6xx, 16 on r7xx
+	 * but I think mesa only supports 8 at the moment
+	 */
+	ctx->Const.MaxTextureCoordUnits = 8;
 	ctx->Const.MaxTextureUnits =
 	    MIN2(ctx->Const.MaxTextureImageUnits,
 		 ctx->Const.MaxTextureCoordUnits);
@@ -275,9 +274,8 @@ static void r600InitConstValues(GLcontext *ctx, radeonScreenPtr screen)
 	ctx->Const.FragmentProgram.MaxNativeAttribs = 32;
 	ctx->Const.FragmentProgram.MaxNativeParameters = 256;
 	ctx->Const.FragmentProgram.MaxNativeAluInstructions = 8192;
-	/* 8 per clause on r6xx, 16 on rv670/r7xx */
-	if ((screen->chip_family == CHIP_FAMILY_RV670) ||
-	    (screen->chip_family >= CHIP_FAMILY_RV770))
+	/* 8 per clause on r6xx, 16 on r7xx */
+	if (screen->chip_family >= CHIP_FAMILY_RV770)
 		ctx->Const.FragmentProgram.MaxNativeTexInstructions = 16;
 	else
 		ctx->Const.FragmentProgram.MaxNativeTexInstructions = 8;
diff --git a/src/mesa/drivers/dri/r600/r600_tex.h b/src/mesa/drivers/dri/r600/r600_tex.h
index fb0e1a023e1..0e3d32892a0 100644
--- a/src/mesa/drivers/dri/r600/r600_tex.h
+++ b/src/mesa/drivers/dri/r600/r600_tex.h
@@ -42,7 +42,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 /* Texel pitch is 8 alignment. */
 #define R700_TEXEL_PITCH_ALIGNMENT_MASK 0x7
 
-#define R700_MAX_TEXTURE_UNITS 8 /* TODO : should be 16, lets make it work, review later */
+#define R700_MAX_TEXTURE_UNITS 16
 
 extern void r600SetDepthTexMode(struct gl_texture_object *tObj);
 
diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c
index b0b38927e67..a6b656af951 100644
--- a/src/mesa/drivers/dri/r600/r700_assembler.c
+++ b/src/mesa/drivers/dri/r600/r700_assembler.c
@@ -4164,20 +4164,21 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm)
 
 GLboolean assemble_XPD(r700_AssemblerBase *pAsm) 
 {
-    BITS tmp;
+    BITS tmp1;
+    BITS tmp2 = 0;
 
     if( GL_FALSE == checkop2(pAsm) )
     {
 	    return GL_FALSE;
     }
 
-    tmp = gethelpr(pAsm);
+    tmp1 = gethelpr(pAsm);
 
     pAsm->D.dst.opcode = SQ_OP2_INST_MUL;
 
     setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
     pAsm->D.dst.rtype = DST_REG_TEMPORARY;
-    pAsm->D.dst.reg   = tmp;
+    pAsm->D.dst.reg   = tmp1;
     nomask_PVSDST(&(pAsm->D.dst));
   
     if( GL_FALSE == assemble_src(pAsm, 0, -1) )
@@ -4203,11 +4204,11 @@ GLboolean assemble_XPD(r700_AssemblerBase *pAsm)
 
     if(0xF != pAsm->pILInst[pAsm->uiCurInst].DstReg.WriteMask)
     {
-        tmp = gethelpr(pAsm);
+        tmp2 = gethelpr(pAsm);
 
         setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
         pAsm->D.dst.rtype = DST_REG_TEMPORARY;
-        pAsm->D.dst.reg   = tmp;
+        pAsm->D.dst.reg   = tmp2;
 
         nomask_PVSDST(&(pAsm->D.dst));
     }
@@ -4235,7 +4236,7 @@ GLboolean assemble_XPD(r700_AssemblerBase *pAsm)
     // result1 + (neg) result0
     setaddrmode_PVSSRC(&(pAsm->S[2].src),ADDR_ABSOLUTE);
     pAsm->S[2].src.rtype = SRC_REG_TEMPORARY;
-    pAsm->S[2].src.reg   = tmp;
+    pAsm->S[2].src.reg   = tmp1;
 
     neg_PVSSRC(&(pAsm->S[2].src));
     noswizzle_PVSSRC(&(pAsm->S[2].src));
@@ -4258,7 +4259,7 @@ GLboolean assemble_XPD(r700_AssemblerBase *pAsm)
         // Use tmp as source
         setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
         pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
-        pAsm->S[0].src.reg   = tmp;
+        pAsm->S[0].src.reg   = tmp2;
 
         noneg_PVSSRC(&(pAsm->S[0].src));
         noswizzle_PVSSRC(&(pAsm->S[0].src));
diff --git a/src/mesa/drivers/windows/gdi/mesa.def b/src/mesa/drivers/windows/gdi/mesa.def
index 62f75d9541f..700e3334295 100644
--- a/src/mesa/drivers/windows/gdi/mesa.def
+++ b/src/mesa/drivers/windows/gdi/mesa.def
@@ -870,7 +870,6 @@ EXPORTS
 	_mesa_bzero
 	_mesa_calloc
 	_mesa_choose_tex_format
-	_mesa_compressed_texture_size
 	_mesa_create_framebuffer
 	_mesa_create_visual
 	_mesa_delete_array_object
@@ -932,6 +931,8 @@ EXPORTS
 	_mesa_update_framebuffer_visual
 	_mesa_use_program
 	_mesa_Viewport
+	_mesa_meta_init
+	_mesa_meta_free
 	_mesa_meta_CopyColorSubTable
 	_mesa_meta_CopyColorTable
 	_mesa_meta_CopyConvolutionFilter1D
@@ -941,7 +942,6 @@ EXPORTS
 	_mesa_meta_CopyTexSubImage1D
 	_mesa_meta_CopyTexSubImage2D
 	_mesa_meta_CopyTexSubImage3D
-	_mesa_wait_query
 	_swrast_Accum
 	_swrast_Bitmap
 	_swrast_BlitFramebuffer
diff --git a/src/mesa/drivers/windows/gdi/wmesa.c b/src/mesa/drivers/windows/gdi/wmesa.c
index 8929b22af16..1788313b297 100644
--- a/src/mesa/drivers/windows/gdi/wmesa.c
+++ b/src/mesa/drivers/windows/gdi/wmesa.c
@@ -1286,9 +1286,6 @@ void wmesa_set_renderbuffer_funcs(struct gl_renderbuffer *rb, int pixelformat,
 	    rb->PutMonoValues = write_mono_rgba_pixels_16;
 	    rb->GetRow = read_rgba_span_16;
 	    rb->GetValues = read_rgba_pixels_16;
-            rb->RedBits = 5;
-            rb->GreenBits = 6;
-            rb->BlueBits = 5;
 	    break;
 	case PF_8R8G8B:
 		if (cColorBits == 24)
@@ -1300,9 +1297,6 @@ void wmesa_set_renderbuffer_funcs(struct gl_renderbuffer *rb, int pixelformat,
 		    rb->PutMonoValues = write_mono_rgba_pixels_24;
 		    rb->GetRow = read_rgba_span_24;
 		    rb->GetValues = read_rgba_pixels_24;
-	        rb->RedBits = 8;
-	        rb->GreenBits = 8;
-	        rb->BlueBits = 8;		
 		}
 		else
 		{
@@ -1313,9 +1307,6 @@ void wmesa_set_renderbuffer_funcs(struct gl_renderbuffer *rb, int pixelformat,
 	        rb->PutMonoValues = write_mono_rgba_pixels_32;
 	        rb->GetRow = read_rgba_span_32;
 	        rb->GetValues = read_rgba_pixels_32;
-            rb->RedBits = 8;
-            rb->GreenBits = 8;
-            rb->BlueBits = 8;
 		}
 	    break;
 	default:
@@ -1331,9 +1322,6 @@ void wmesa_set_renderbuffer_funcs(struct gl_renderbuffer *rb, int pixelformat,
 	rb->PutMonoValues = write_mono_rgba_pixels_front;
 	rb->GetRow = read_rgba_span_front;
 	rb->GetValues = read_rgba_pixels_front;
-        rb->RedBits = 8; /* XXX fix these (565?) */
-        rb->GreenBits = 8;
-        rb->BlueBits = 8;
     }
 }
 
diff --git a/src/mesa/main/api_validate.c b/src/mesa/main/api_validate.c
index e71e5a6ce86..013048bbc16 100644
--- a/src/mesa/main/api_validate.c
+++ b/src/mesa/main/api_validate.c
@@ -190,9 +190,6 @@ _mesa_validate_DrawElements(GLcontext *ctx,
       return GL_FALSE;
    }
 
-   if (ctx->NewState)
-      _mesa_update_state(ctx);
-
    if (!check_valid_to_render(ctx, "glDrawElements"))
       return GL_FALSE;
 
@@ -254,9 +251,6 @@ _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
       return GL_FALSE;
    }
 
-   if (ctx->NewState)
-      _mesa_update_state(ctx);
-
    if (!check_valid_to_render(ctx, "glDrawRangeElements"))
       return GL_FALSE;
 
@@ -304,9 +298,6 @@ _mesa_validate_DrawArrays(GLcontext *ctx,
       return GL_FALSE;
    }
 
-   if (ctx->NewState)
-      _mesa_update_state(ctx);
-
    if (!check_valid_to_render(ctx, "glDrawArrays"))
       return GL_FALSE;
 
diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c
index 87eae966392..f5d9a30efbc 100644
--- a/src/mesa/main/context.c
+++ b/src/mesa/main/context.c
@@ -1580,6 +1580,10 @@ _mesa_set_mvp_with_dp4( GLcontext *ctx,
 GLboolean
 _mesa_valid_to_render(GLcontext *ctx, const char *where)
 {
+   /* This depends on having up to date derived state (shaders) */
+   if (ctx->NewState)
+      _mesa_update_state(ctx);
+
    if (ctx->Shader.CurrentProgram) {
       /* using shaders */
       if (!ctx->Shader.CurrentProgram->LinkStatus) {
diff --git a/src/mesa/shader/slang/slang_preprocess.c b/src/mesa/shader/slang/slang_preprocess.c
index e9a24cc009a..ba40bad446e 100644
--- a/src/mesa/shader/slang/slang_preprocess.c
+++ b/src/mesa/shader/slang/slang_preprocess.c
@@ -914,6 +914,35 @@ parse_if (slang_string *output, const byte *prod, GLuint *pi, GLint *result, pp_
 #define PRAGMA_PARAM     1
 
 
+/**
+ * Return the length of the given string, stopping at any C++-style comments.
+ * This step fixes bugs with macro definitions such as:
+ *   #define PI 3.14159 // this is pi
+ * The preprocessor includes the comment in the definition of PI so
+ * when we plug in PI somewhere, we get the comment too.
+ * This function effectively strips of the // comment from the given string.
+ * It might also be possible to fix this in the preprocessor grammar.
+ * This bug is not present in the new Mesa 7.8 preprocessor.
+ */
+static int
+strlen_without_comments(const char *s)
+{
+   char pred = 0;
+   int len = 0;
+   while (*s) {
+      if (*s == '/' && pred == '/') {
+         return len - 1;
+      }
+      pred = *s;
+      s++;
+      len++;
+   }
+   return len;
+}
+
+
+
+
 static GLboolean
 preprocess_source (slang_string *output, const char *source,
                    grammar pid, grammar eid,
@@ -1055,11 +1084,12 @@ preprocess_source (slang_string *output, const char *source,
                if (state.cond.top->effective) {
                   slang_string replacement;
                   expand_state es;
+                  int idlen2 = strlen_without_comments((char*)id);
 
                   pp_annotate (output, ") %s", id);
 
                   slang_string_init(&replacement);
-                  slang_string_pushs(&replacement, id, idlen);
+                  slang_string_pushs(&replacement, id, idlen2);
 
                   /* Expand macro replacement. */
                   es.output = &symbol->replacement;
diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c
index 5c6af1125a3..d3b22db121c 100644
--- a/src/mesa/state_tracker/st_draw.c
+++ b/src/mesa/state_tracker/st_draw.c
@@ -569,6 +569,9 @@ st_draw_vbo(GLcontext *ctx,
    /* sanity check for pointer arithmetic below */
    assert(sizeof(arrays[0]->Ptr[0]) == 1);
 
+   /* Mesa core state should have been validated already */
+   assert(ctx->NewState == 0x0);
+
    st_validate_state(ctx->st);
 
    /* must get these after state validation! */
diff --git a/src/mesa/tnl/t_draw.c b/src/mesa/tnl/t_draw.c
index 9a888ce19f4..b0e31c06a9f 100644
--- a/src/mesa/tnl/t_draw.c
+++ b/src/mesa/tnl/t_draw.c
@@ -394,6 +394,9 @@ void _tnl_draw_prims( GLcontext *ctx,
    GLuint max_basevertex = prim->basevertex;
    GLuint i;
 
+   /* Mesa core state should have been validated already */
+   assert(ctx->NewState == 0x0);
+
    for (i = 1; i < nr_prims; i++)
       max_basevertex = MAX2(max_basevertex, prim[i].basevertex);
 
diff --git a/src/mesa/vbo/vbo_exec_array.c b/src/mesa/vbo/vbo_exec_array.c
index bd2fccdba14..d08976ca298 100644
--- a/src/mesa/vbo/vbo_exec_array.c
+++ b/src/mesa/vbo/vbo_exec_array.c
@@ -443,6 +443,13 @@ recalculate_input_bindings(GLcontext *ctx)
 }
 
 
+/**
+ * Examine the enabled vertex arrays to set the exec->array.inputs[] values.
+ * These will point to the arrays to actually use for drawing.  Some will
+ * be user-provided arrays, other will be zero-stride const-valued arrays.
+ * Note that this might set the _NEW_ARRAY dirty flag so state validation
+ * must be done after this call.
+ */
 static void
 bind_arrays(GLcontext *ctx)
 {
@@ -484,9 +491,6 @@ vbo_exec_DrawArrays(GLenum mode, GLint start, GLsizei count)
 
    FLUSH_CURRENT( ctx, 0 );
 
-   if (ctx->NewState)
-      _mesa_update_state( ctx );
-      
    if (!_mesa_valid_to_render(ctx, "glDrawArrays")) {
       return;
    }
@@ -600,18 +604,16 @@ vbo_validated_drawrangeelements(GLcontext *ctx, GLenum mode,
 
    FLUSH_CURRENT( ctx, 0 );
 
-   if (ctx->NewState)
-      _mesa_update_state( ctx );
-
    if (!_mesa_valid_to_render(ctx, "glDraw[Range]Elements")) {
       return;
    }
 
+   bind_arrays( ctx );
+
+   /* check for dirty state again */
    if (ctx->NewState)
       _mesa_update_state( ctx );
 
-   bind_arrays( ctx );
-
    ib.count = count;
    ib.type = type;
    ib.obj = ctx->Array.ElementArrayBufferObj;
@@ -722,8 +724,7 @@ vbo_exec_DrawRangeElementsBaseVertex(GLenum mode,
 
 #ifdef DEBUG
       /* 'end' was out of bounds, but now let's check the actual array
-       * indexes to see if any of them are out of bounds.  If so, warn
-       * and skip the draw to avoid potential segfault, etc.
+       * indexes to see if any of them are out of bounds.
        */
       {
          GLuint max = _mesa_max_buffer_index(ctx, count, type, indices,
@@ -740,7 +741,6 @@ vbo_exec_DrawRangeElementsBaseVertex(GLenum mode,
                              ctx->Array.ElementArrayBufferObj->Name,
                              ctx->Array.ElementArrayBufferObj->Size);
             }
-            return;
          }
          /* XXX we could also find the min index and compare to 'start'
           * to see if start is correct.  But it's more likely to get the
@@ -850,16 +850,10 @@ vbo_validated_multidrawelements(GLcontext *ctx, GLenum mode,
 
    FLUSH_CURRENT( ctx, 0 );
 
-   if (ctx->NewState)
-      _mesa_update_state( ctx );
-
    if (!_mesa_valid_to_render(ctx, "glMultiDrawElements")) {
       return;
    }
 
-   if (ctx->NewState)
-      _mesa_update_state( ctx );
-
    prim = _mesa_calloc(primcount * sizeof(*prim));
    if (prim == NULL) {
       _mesa_error(ctx, GL_OUT_OF_MEMORY, "glMultiDrawElements");
@@ -871,6 +865,10 @@ vbo_validated_multidrawelements(GLcontext *ctx, GLenum mode,
     */
    bind_arrays( ctx );
 
+   /* check for dirty state again */
+   if (ctx->NewState)
+      _mesa_update_state( ctx );
+
    switch (type) {
    case GL_UNSIGNED_INT:
       index_type_size = 4;
diff --git a/windows/VC8/mesa/gdi/gdi.vcproj b/windows/VC8/mesa/gdi/gdi.vcproj
index a3dd5ef5b62..220923f9051 100644
--- a/windows/VC8/mesa/gdi/gdi.vcproj
+++ b/windows/VC8/mesa/gdi/gdi.vcproj
@@ -397,6 +397,10 @@
 				>

 			</File>

 			<File

+				RelativePath="..\..\..\..\src\mesa\drivers\common\meta.c"

+				>

+			</File>

+			<File

 				RelativePath="..\..\..\..\src\mesa\drivers\windows\gdi\wgl.c"

 				>

 			</File>

@@ -418,6 +422,10 @@
 				>

 			</File>

 			<File

+				RelativePath="..\..\..\..\src\mesa\drivers\common\meta.h"

+				>

+			</File>

+			<File

 				RelativePath="..\..\..\..\include\GL\wmesa.h"

 				>

 			</File>

diff --git a/windows/VC8/mesa/mesa/mesa.vcproj b/windows/VC8/mesa/mesa/mesa.vcproj
index 993c28ddc17..19de4ebd02f 100644
--- a/windows/VC8/mesa/mesa/mesa.vcproj
+++ b/windows/VC8/mesa/mesa/mesa.vcproj
@@ -451,6 +451,10 @@
 				>

 			</File>

 			<File

+				RelativePath="..\..\..\..\src\mesa\main\formats.c"

+				>

+			</File>

+			<File

 				RelativePath="..\..\..\..\src\mesa\main\framebuffer.c"

 				>

 			</File>

@@ -1324,6 +1328,10 @@
 				>

 			</File>

 			<File

+				RelativePath="..\..\..\..\src\mesa\main\formats.h"

+				>

+			</File>

+			<File

 				RelativePath="..\..\..\..\src\mesa\main\framebuffer.h"

 				>

 			</File>

@@ -1896,11 +1904,11 @@
 				>

 			</File>

 			<File

-				RelativePath="..\..\..\..\src\mesa\main\texformat.h"

+				RelativePath="..\..\..\..\src\mesa\main\texfetch_tmp.h"

 				>

 			</File>

 			<File

-				RelativePath="..\..\..\..\src\mesa\main\texfetch_tmp.h"

+				RelativePath="..\..\..\..\src\mesa\main\texformat.h"

 				>

 			</File>

 			<File