Merge remote-tracking branch 'origin/released'

64 files changed, 1434 insertions, 665 deletions

diff --git a/mesalib/src/gallium/auxiliary/util/u_blitter.c b/mesalib/src/gallium/auxiliary/util/u_blitter.c
index b0d538de6..f5cc5cb3b 100644
--- a/mesalib/src/gallium/auxiliary/util/u_blitter.c
+++ b/mesalib/src/gallium/auxiliary/util/u_blitter.c
@@ -59,7 +59,6 @@ struct blitter_context_priv
    float vertices[4][2][4];   /**< {pos, color} or {pos, texcoord} */
 
    /* Templates for various state objects. */
-   struct pipe_sampler_state template_sampler_state;
 
    /* Constant state objects. */
    /* Vertex shaders. */
@@ -92,8 +91,8 @@ struct blitter_context_priv
    void *velem_uint_state;
    void *velem_sint_state;
 
-   /* Sampler state for clamping to a miplevel. */
-   void *sampler_state[PIPE_MAX_TEXTURE_LEVELS * 2];
+   /* Sampler state. */
+   void *sampler_state;
 
    /* Rasterizer state. */
    void *rs_state;
@@ -126,7 +125,7 @@ struct blitter_context *util_blitter_create(struct pipe_context *pipe)
    struct pipe_blend_state blend;
    struct pipe_depth_stencil_alpha_state dsa;
    struct pipe_rasterizer_state rs_state;
-   struct pipe_sampler_state *sampler_state;
+   struct pipe_sampler_state sampler_state;
    struct pipe_vertex_element velem[2];
    unsigned i;
 
@@ -192,13 +191,12 @@ struct blitter_context *util_blitter_create(struct pipe_context *pipe)
       pipe->create_depth_stencil_alpha_state(pipe, &dsa);
 
    /* sampler state */
-   sampler_state = &ctx->template_sampler_state;
-   sampler_state->wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
-   sampler_state->wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
-   sampler_state->wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
-   sampler_state->normalized_coords = TRUE;
-   /* The sampler state objects which sample from a specified mipmap level
-    * are created on-demand. */
+   memset(&sampler_state, 0, sizeof(sampler_state));
+   sampler_state.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   sampler_state.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   sampler_state.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   sampler_state.normalized_coords = 1;
+   ctx->sampler_state = pipe->create_sampler_state(pipe, &sampler_state);
 
    /* rasterizer state */
    memset(&rs_state, 0, sizeof(rs_state));
@@ -292,10 +290,7 @@ void util_blitter_destroy(struct blitter_context *blitter)
          pipe->delete_fs_state(pipe, ctx->fs_col_int[i]);
    }
 
-   for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS * 2; i++)
-      if (ctx->sampler_state[i])
-         pipe->delete_sampler_state(pipe, ctx->sampler_state[i]);
-
+   pipe->delete_sampler_state(pipe, ctx->sampler_state);
    pipe_resource_reference(&ctx->vbuf, NULL);
    FREE(ctx);
 }
@@ -497,21 +492,22 @@ static void blitter_set_clear_color(struct blitter_context_priv *ctx,
    }
 }
 
-static void get_texcoords(struct pipe_resource *src,
-                          unsigned level,
+static void get_texcoords(struct pipe_sampler_view *src,
+                          unsigned src_width0, unsigned src_height0,
                           unsigned x1, unsigned y1,
                           unsigned x2, unsigned y2,
-                          boolean normalized, float out[4])
+                          float out[4])
 {
-   if(normalized)
-   {
-      out[0] = x1 / (float)u_minify(src->width0,  level);
-      out[1] = y1 / (float)u_minify(src->height0, level);
-      out[2] = x2 / (float)u_minify(src->width0,  level);
-      out[3] = y2 / (float)u_minify(src->height0, level);
-   }
-   else
-   {
+   struct pipe_resource *tex = src->texture;
+   unsigned level = src->u.tex.first_level;
+   boolean normalized = tex->target != PIPE_TEXTURE_RECT;
+
+   if (normalized) {
+      out[0] = x1 / (float)u_minify(src_width0,  level);
+      out[1] = y1 / (float)u_minify(src_height0, level);
+      out[2] = x2 / (float)u_minify(src_width0,  level);
+      out[3] = y2 / (float)u_minify(src_height0, level);
+   } else {
       out[0] = x1;
       out[1] = y1;
       out[2] = x2;
@@ -535,76 +531,52 @@ static void set_texcoords_in_vertices(const float coord[4],
    out[1] = coord[3]; /*t3.t*/
 }
 
-static void blitter_set_texcoords_2d(struct blitter_context_priv *ctx,
-                                     struct pipe_resource *src,
-                                     unsigned level,
-                                     unsigned x1, unsigned y1,
-                                     unsigned x2, unsigned y2)
+static void blitter_set_texcoords(struct blitter_context_priv *ctx,
+                                  struct pipe_sampler_view *src,
+                                  unsigned src_width0, unsigned src_height0,
+                                  unsigned layer,
+                                  unsigned x1, unsigned y1,
+                                  unsigned x2, unsigned y2)
 {
    unsigned i;
    float coord[4];
+   float face_coord[4][2];
 
-   get_texcoords(src, level, x1, y1, x2, y2, TRUE, coord);
-   set_texcoords_in_vertices(coord, &ctx->vertices[0][1][0], 8);
+   get_texcoords(src, src_width0, src_height0, x1, y1, x2, y2, coord);
 
-   for (i = 0; i < 4; i++) {
-      ctx->vertices[i][1][2] = 0; /*r*/
-      ctx->vertices[i][1][3] = 1; /*q*/
+   if (src->texture->target == PIPE_TEXTURE_CUBE) {
+      set_texcoords_in_vertices(coord, &face_coord[0][0], 2);
+      util_map_texcoords2d_onto_cubemap(layer,
+                                        /* pointer, stride in floats */
+                                        &face_coord[0][0], 2,
+                                        &ctx->vertices[0][1][0], 8);
+   } else {
+      set_texcoords_in_vertices(coord, &ctx->vertices[0][1][0], 8);
    }
-}
-
-static void blitter_set_texcoords_3d(struct blitter_context_priv *ctx,
-                                     struct pipe_resource *src,
-                                     unsigned level,
-                                     unsigned zslice,
-                                     unsigned x1, unsigned y1,
-                                     unsigned x2, unsigned y2,
-				     boolean normalized)
-{
-   int i;
-   float r = normalized ? zslice / (float)u_minify(src->depth0, level) : zslice;
-
-   blitter_set_texcoords_2d(ctx, src, level, x1, y1, x2, y2);
-
-   for (i = 0; i < 4; i++)
-      ctx->vertices[i][1][2] = r; /*r*/
-}
 
-static void blitter_set_texcoords_1d_array(struct blitter_context_priv *ctx,
-                                           struct pipe_resource *src,
-                                           unsigned level,
-                                           unsigned zslice,
-                                           unsigned x1, unsigned x2)
-{
-   int i;
-   float r = zslice;
-
-   blitter_set_texcoords_2d(ctx, src, level, x1, 0, x2, 0);
-
-   for (i = 0; i < 4; i++)
-      ctx->vertices[i][1][1] = r; /*r*/
-}
-
-static void blitter_set_texcoords_cube(struct blitter_context_priv *ctx,
-                                       struct pipe_resource *src,
-                                       unsigned level, unsigned face,
-                                       unsigned x1, unsigned y1,
-                                       unsigned x2, unsigned y2)
-{
-   int i;
-   float coord[4];
-   float st[4][2];
+   /* Set the layer. */
+   switch (src->texture->target) {
+   case PIPE_TEXTURE_3D:
+      {
+         float r = layer / (float)u_minify(src->texture->depth0,
+                                           src->u.tex.first_level);
+         for (i = 0; i < 4; i++)
+            ctx->vertices[i][1][2] = r; /*r*/
+      }
+      break;
 
-   get_texcoords(src, level, x1, y1, x2, y2, TRUE, coord);
-   set_texcoords_in_vertices(coord, &st[0][0], 2);
+   case PIPE_TEXTURE_1D_ARRAY:
+      for (i = 0; i < 4; i++)
+         ctx->vertices[i][1][1] = layer; /*t*/
+      break;
 
-   util_map_texcoords2d_onto_cubemap(face,
-                                     /* pointer, stride in floats */
-                                     &st[0][0], 2,
-                                     &ctx->vertices[0][1][0], 8);
+   case PIPE_TEXTURE_2D_ARRAY:
+      for (i = 0; i < 4; i++)
+         ctx->vertices[i][1][2] = layer; /*r*/
+      break;
 
-   for (i = 0; i < 4; i++)
-      ctx->vertices[i][1][3] = 1; /*q*/
+   default:;
+   }
 }
 
 static void blitter_set_dst_dimensions(struct blitter_context_priv *ctx,
@@ -615,31 +587,6 @@ static void blitter_set_dst_dimensions(struct blitter_context_priv *ctx,
 }
 
 static INLINE
-void **blitter_get_sampler_state(struct blitter_context_priv *ctx,
-                                 int miplevel, boolean normalized)
-{
-   struct pipe_context *pipe = ctx->base.pipe;
-   struct pipe_sampler_state *sampler_state = &ctx->template_sampler_state;
-
-   assert(miplevel < PIPE_MAX_TEXTURE_LEVELS);
-
-   /* Create the sampler state on-demand. */
-   if (!ctx->sampler_state[miplevel * 2 + normalized]) {
-      sampler_state->lod_bias = miplevel;
-      sampler_state->min_lod = miplevel;
-      sampler_state->max_lod = miplevel;
-      sampler_state->normalized_coords = normalized;
-
-      ctx->sampler_state[miplevel * 2 + normalized] = pipe->create_sampler_state(pipe,
-                                                                sampler_state);
-   }
-
-   /* Return void** so that it can be passed to bind_fragment_sampler_states
-    * directly. */
-   return &ctx->sampler_state[miplevel * 2 + normalized];
-}
-
-static INLINE
 void *blitter_get_fs_col(struct blitter_context_priv *ctx, unsigned num_cbufs,
                          boolean int_format)
 {
@@ -864,27 +811,16 @@ void util_blitter_copy_texture(struct blitter_context *blitter,
    struct pipe_context *pipe = ctx->base.pipe;
    struct pipe_screen *screen = pipe->screen;
    struct pipe_surface *dstsurf, surf_templ;
-   struct pipe_framebuffer_state fb_state;
    struct pipe_sampler_view viewTempl, *view;
    unsigned bind;
-   unsigned width = srcbox->width;
-   unsigned height = srcbox->height;
    boolean is_stencil, is_depth;
-   boolean normalized;
 
    /* Give up if textures are not set. */
    assert(dst && src);
    if (!dst || !src)
       return;
 
-   /* Sanity checks. */
-   if (dst == src) {
-      assert(!is_overlap(srcbox->x, srcbox->x + width, srcbox->y, srcbox->y + height,
-                         dstx, dstx + width, dsty, dsty + height));
-   }
    assert(src->target < PIPE_MAX_TEXTURE_TYPES);
-   /* XXX should handle 3d regions */
-   assert(srcbox->depth == 1);
 
    /* Is this a ZS format? */
    is_depth = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0;
@@ -902,22 +838,60 @@ void util_blitter_copy_texture(struct blitter_context *blitter,
                                     dst->nr_samples, bind) ||
        !screen->is_format_supported(screen, src->format, src->target,
                                     src->nr_samples, PIPE_BIND_SAMPLER_VIEW)) {
-      ctx->base.running = TRUE;
+      blitter_set_running_flag(ctx);
       util_resource_copy_region(pipe, dst, dstlevel, dstx, dsty, dstz,
                                 src, srclevel, srcbox);
-      ctx->base.running = FALSE;
+      blitter_unset_running_flag(ctx);
       return;
    }
 
-   /* Get surface. */
+   /* Initialize the surface. */
    memset(&surf_templ, 0, sizeof(surf_templ));
    u_surface_default_template(&surf_templ, dst, bind);
    surf_templ.format = util_format_linear(dst->format);
    surf_templ.u.tex.level = dstlevel;
    surf_templ.u.tex.first_layer = dstz;
-   surf_templ.u.tex.last_layer = dstz;
+   surf_templ.u.tex.last_layer = dstz + srcbox->depth - 1;
    dstsurf = pipe->create_surface(pipe, dst, &surf_templ);
 
+   /* Initialize the sampler view. */
+   u_sampler_view_default_template(&viewTempl, src,
+                                   util_format_linear(src->format));
+   viewTempl.u.tex.first_level = srclevel;
+   viewTempl.u.tex.last_level = srclevel;
+   view = pipe->create_sampler_view(pipe, src, &viewTempl);
+
+   /* Copy. */
+   util_blitter_copy_texture_view(blitter, dstsurf, dstx, dsty, view, srcbox,
+                                  src->width0, src->height0);
+
+   pipe_surface_reference(&dstsurf, NULL);
+   pipe_sampler_view_reference(&view, NULL);
+}
+
+void util_blitter_copy_texture_view(struct blitter_context *blitter,
+                                    struct pipe_surface *dst,
+                                    unsigned dstx, unsigned dsty,
+                                    struct pipe_sampler_view *src,
+                                    const struct pipe_box *srcbox,
+                                    unsigned src_width0, unsigned src_height0)
+{
+   struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter;
+   struct pipe_context *pipe = ctx->base.pipe;
+   struct pipe_framebuffer_state fb_state;
+   enum pipe_texture_target src_target = src->texture->target;
+   unsigned width = srcbox->width;
+   unsigned height = srcbox->height;
+
+   /* Sanity checks. */
+   if (dst->texture == src->texture &&
+       dst->u.tex.level == src->u.tex.first_level) {
+      assert(!is_overlap(srcbox->x, srcbox->x + width, srcbox->y, srcbox->y + height,
+                         dstx, dstx + width, dsty, dsty + height));
+   }
+   /* XXX should handle 3d regions */
+   assert(srcbox->depth == 1);
+
    /* Check whether the states are properly saved. */
    blitter_set_running_flag(ctx);
    blitter_check_saved_vertex_states(ctx);
@@ -926,60 +900,55 @@ void util_blitter_copy_texture(struct blitter_context *blitter,
    blitter_check_saved_fb_state(ctx);
 
    /* Initialize framebuffer state. */
-   fb_state.width = dstsurf->width;
-   fb_state.height = dstsurf->height;
+   fb_state.width = dst->width;
+   fb_state.height = dst->height;
 
-   if (is_depth) {
+   if (util_format_is_depth_or_stencil(dst->format)) {
       pipe->bind_blend_state(pipe, ctx->blend_keep_color);
       pipe->bind_depth_stencil_alpha_state(pipe,
                                            ctx->dsa_write_depth_keep_stencil);
       pipe->bind_fs_state(pipe,
-                          blitter_get_fs_texfetch_depth(ctx, src->target));
+            blitter_get_fs_texfetch_depth(ctx, src_target));
 
       fb_state.nr_cbufs = 0;
-      fb_state.zsbuf = dstsurf;
+      fb_state.zsbuf = dst;
    } else {
       pipe->bind_blend_state(pipe, ctx->blend_write_color);
       pipe->bind_depth_stencil_alpha_state(pipe, ctx->dsa_keep_depth_stencil);
       pipe->bind_fs_state(pipe,
-                          blitter_get_fs_texfetch_col(ctx, src->target));
+            blitter_get_fs_texfetch_col(ctx, src_target));
 
       fb_state.nr_cbufs = 1;
-      fb_state.cbufs[0] = dstsurf;
+      fb_state.cbufs[0] = dst;
       fb_state.zsbuf = 0;
    }
 
-   normalized = src->target != PIPE_TEXTURE_RECT;
-
-   /* Initialize sampler view. */
-   u_sampler_view_default_template(&viewTempl, src, util_format_linear(src->format));
-   view = pipe->create_sampler_view(pipe, src, &viewTempl);
-
    /* Set rasterizer state, shaders, and textures. */
    pipe->bind_rasterizer_state(pipe, ctx->rs_state);
    pipe->bind_vs_state(pipe, ctx->vs);
    if (ctx->has_geometry_shader)
       pipe->bind_gs_state(pipe, NULL);
-   pipe->bind_fragment_sampler_states(pipe, 1,
-                                      blitter_get_sampler_state(ctx, srclevel, normalized));
+   pipe->bind_fragment_sampler_states(pipe, 1, &ctx->sampler_state);
    pipe->bind_vertex_elements_state(pipe, ctx->velem_state);
-   pipe->set_fragment_sampler_views(pipe, 1, &view);
+   pipe->set_fragment_sampler_views(pipe, 1, &src);
    pipe->set_framebuffer_state(pipe, &fb_state);
 
-   blitter_set_dst_dimensions(ctx, dstsurf->width, dstsurf->height);
+   blitter_set_dst_dimensions(ctx, dst->width, dst->height);
 
-   switch (src->target) {
+   switch (src_target) {
       /* Draw the quad with the draw_rectangle callback. */
       case PIPE_TEXTURE_1D:
       case PIPE_TEXTURE_2D:
       case PIPE_TEXTURE_RECT:
          {
             /* Set texture coordinates. - use a pipe color union
-             * for interface purposes
+             * for interface purposes.
+             * XXX pipe_color_union is a wrong name since we use that to set
+             * texture coordinates too.
              */
             union pipe_color_union coord;
-            get_texcoords(src, srclevel, srcbox->x, srcbox->y,
-                          srcbox->x+width, srcbox->y+height, normalized, coord.f);
+            get_texcoords(src, src_width0, src_height0, srcbox->x, srcbox->y,
+                          srcbox->x+width, srcbox->y+height, coord.f);
 
             /* Draw. */
             blitter->draw_rectangle(blitter, dstx, dsty, dstx+width, dsty+height, 0,
@@ -990,24 +959,14 @@ void util_blitter_copy_texture(struct blitter_context *blitter,
       /* Draw the quad with the generic codepath. */
       default:
          /* Set texture coordinates. */
-         switch (src->target) {
+         switch (src_target) {
          case PIPE_TEXTURE_1D_ARRAY:
-            blitter_set_texcoords_1d_array(ctx, src, srclevel, srcbox->y,
-                                           srcbox->x, srcbox->x + width);
-            break;
-
          case PIPE_TEXTURE_2D_ARRAY:
          case PIPE_TEXTURE_3D:
-            blitter_set_texcoords_3d(ctx, src, srclevel, srcbox->z,
-                                     srcbox->x, srcbox->y,
-                                     srcbox->x + width, srcbox->y + height,
-                                     src->target == PIPE_TEXTURE_3D);
-            break;
-
          case PIPE_TEXTURE_CUBE:
-            blitter_set_texcoords_cube(ctx, src, srclevel, srcbox->z,
-                                       srcbox->x, srcbox->y,
-                                       srcbox->x + width, srcbox->y + height);
+            blitter_set_texcoords(ctx, src, src_width0, src_height0, srcbox->z,
+                                  srcbox->y, srcbox->x,
+                                  srcbox->x + width, srcbox->y + height);
             break;
 
          default:
@@ -1028,9 +987,6 @@ void util_blitter_copy_texture(struct blitter_context *blitter,
    blitter_restore_textures(ctx);
    blitter_restore_fb_state(ctx);
    blitter_unset_running_flag(ctx);
-
-   pipe_surface_reference(&dstsurf, NULL);
-   pipe_sampler_view_reference(&view, NULL);
 }
 
 /* Clear a region of a color surface to a constant value. */
diff --git a/mesalib/src/gallium/auxiliary/util/u_blitter.h b/mesalib/src/gallium/auxiliary/util/u_blitter.h
index 798096e74..3e1457ae0 100644
--- a/mesalib/src/gallium/auxiliary/util/u_blitter.h
+++ b/mesalib/src/gallium/auxiliary/util/u_blitter.h
@@ -190,6 +190,30 @@ void util_blitter_copy_texture(struct blitter_context *blitter,
                                boolean ignore_stencil);
 
 /**
+ * Same as util_blitter_copy_texture, but dst and src are pipe_surface and
+ * pipe_sampler_view, respectively. The mipmap level and dstz are part of
+ * the views.
+ *
+ * Drivers can use this to change resource properties (like format, width,
+ * height) by changing how the views interpret them, instead of changing
+ * pipe_resource directly. This is usually needed to accelerate copying of
+ * compressed formats.
+ *
+ * src_width0 and src_height0 are sampler_view-private properties that
+ * override pipe_resource. The blitter uses them for computation of texture
+ * coordinates. The dst dimensions are supplied through pipe_surface::width
+ * and height.
+ *
+ * NOTE: There are no checks whether the blit is actually supported.
+ */
+void util_blitter_copy_texture_view(struct blitter_context *blitter,
+                                    struct pipe_surface *dst,
+                                    unsigned dstx, unsigned dsty,
+                                    struct pipe_sampler_view *src,
+                                    const struct pipe_box *srcbox,
+                                    unsigned src_width0, unsigned src_height0);
+
+/**
  * Clear a region of a (color) surface to a constant value.
  *
  * These states must be saved in the blitter in addition to the state objects
diff --git a/mesalib/src/gallium/auxiliary/util/u_upload_mgr.c b/mesalib/src/gallium/auxiliary/util/u_upload_mgr.c
index 71fe53e3a..a2319d0c4 100644
--- a/mesalib/src/gallium/auxiliary/util/u_upload_mgr.c
+++ b/mesalib/src/gallium/auxiliary/util/u_upload_mgr.c
@@ -131,8 +131,9 @@ u_upload_alloc_buffer( struct u_upload_mgr *upload,
                                         upload->bind,
                                         PIPE_USAGE_STREAM,
                                         size );
-   if (upload->buffer == NULL) 
-      goto fail;
+   if (upload->buffer == NULL) {
+      return PIPE_ERROR_OUT_OF_MEMORY;
+   }
 
    /* Map the new buffer. */
    upload->map = pipe_buffer_map_range(upload->pipe, upload->buffer,
@@ -140,17 +141,16 @@ u_upload_alloc_buffer( struct u_upload_mgr *upload,
                                        PIPE_TRANSFER_WRITE |
                                        PIPE_TRANSFER_FLUSH_EXPLICIT,
                                        &upload->transfer);
-   
+   if (upload->map == NULL) {
+      upload->size = 0;
+      pipe_resource_reference(&upload->buffer, NULL);
+      return PIPE_ERROR_OUT_OF_MEMORY;
+   }
+
    upload->size = size;
 
    upload->offset = 0;
-   return 0;
-
-fail:
-   if (upload->buffer)
-      pipe_resource_reference( &upload->buffer, NULL );
-
-   return PIPE_ERROR_OUT_OF_MEMORY;
+   return PIPE_OK;
 }
 
 enum pipe_error u_upload_alloc( struct u_upload_mgr *upload,
@@ -170,7 +170,7 @@ enum pipe_error u_upload_alloc( struct u_upload_mgr *upload,
    if (MAX2(upload->offset, alloc_offset) + alloc_size > upload->size) {
       enum pipe_error ret = u_upload_alloc_buffer(upload,
                                                   alloc_offset + alloc_size);
-      if (ret)
+      if (ret != PIPE_OK)
          return ret;
 
       *flushed = TRUE;
@@ -187,6 +187,11 @@ enum pipe_error u_upload_alloc( struct u_upload_mgr *upload,
 					  PIPE_TRANSFER_FLUSH_EXPLICIT |
 					  PIPE_TRANSFER_UNSYNCHRONIZED,
 					  &upload->transfer);
+      if (!upload->map) {
+         pipe_resource_reference(outbuf, NULL);
+         *ptr = NULL;
+         return PIPE_ERROR_OUT_OF_MEMORY;
+      }
    }
 
    assert(offset < upload->buffer->width0);
@@ -214,7 +219,7 @@ enum pipe_error u_upload_data( struct u_upload_mgr *upload,
    enum pipe_error ret = u_upload_alloc(upload, min_out_offset, size,
                                         out_offset, outbuf, flushed,
                                         (void**)&ptr);
-   if (ret)
+   if (ret != PIPE_OK)
       return ret;
 
    memcpy(ptr, data, size);
@@ -247,8 +252,7 @@ enum pipe_error u_upload_buffer( struct u_upload_mgr *upload,
                                              &transfer);
 
    if (map == NULL) {
-      ret = PIPE_ERROR_OUT_OF_MEMORY;
-      goto done;
+      return PIPE_ERROR_OUT_OF_MEMORY;
    }
 
    if (0)
@@ -261,9 +265,7 @@ enum pipe_error u_upload_buffer( struct u_upload_mgr *upload,
                         out_offset,
                         outbuf, flushed );
 
-done:
-   if (map)
-      pipe_buffer_unmap( upload->pipe, transfer );
+   pipe_buffer_unmap( upload->pipe, transfer );
 
    return ret;
 }
diff --git a/mesalib/src/glsl/Makefile b/mesalib/src/glsl/Makefile
index b2efb2abc..504f1fb43 100644
--- a/mesalib/src/glsl/Makefile
+++ b/mesalib/src/glsl/Makefile
@@ -53,6 +53,7 @@ CXX_SOURCES = \
 	ir_variable_refcount.cpp \
 	linker.cpp \
 	link_functions.cpp \
+	link_uniforms.cpp \
 	loop_analysis.cpp \
 	loop_controls.cpp \
 	loop_unroll.cpp \
diff --git a/mesalib/src/glsl/SConscript b/mesalib/src/glsl/SConscript
index b4786c5e5..09c7edbc4 100644
--- a/mesalib/src/glsl/SConscript
+++ b/mesalib/src/glsl/SConscript
@@ -64,6 +64,7 @@ glsl_sources = [
     'ir_variable_refcount.cpp',
     'linker.cpp',
     'link_functions.cpp',
+    'link_uniforms.cpp',
     'loop_analysis.cpp',
     'loop_controls.cpp',
     'loop_unroll.cpp',
diff --git a/mesalib/src/glsl/builtin_variables.h b/mesalib/src/glsl/builtin_variables.h
index d25bbf42a..237b361d8 100644
--- a/mesalib/src/glsl/builtin_variables.h
+++ b/mesalib/src/glsl/builtin_variables.h
@@ -103,8 +103,8 @@ static const builtin_variable builtin_110_deprecated_uniforms[] = {
    { ir_var_uniform, -1, "gl_LightModelParameters", "gl_LightModel"},
 
    /* Mesa-internal ATI_envmap_bumpmap state. */
-   { ir_var_uniform, -1, "vec2", "gl_MESABumpRotMatrix0"},
-   { ir_var_uniform, -1, "vec2", "gl_MESABumpRotMatrix1"},
-   { ir_var_uniform, -1, "vec4", "gl_MESAFogParamsOptimized"},
+   { ir_var_uniform, -1, "vec2", "gl_BumpRotMatrix0MESA"},
+   { ir_var_uniform, -1, "vec2", "gl_BumpRotMatrix1MESA"},
+   { ir_var_uniform, -1, "vec4", "gl_FogParamsOptimizedMESA"},
 };
 
diff --git a/mesalib/src/glsl/glsl_parser_extras.cpp b/mesalib/src/glsl/glsl_parser_extras.cpp
index a9075b2b1..e2112fe6d 100644
--- a/mesalib/src/glsl/glsl_parser_extras.cpp
+++ b/mesalib/src/glsl/glsl_parser_extras.cpp
@@ -883,8 +883,27 @@ ast_struct_specifier::ast_struct_specifier(char *identifier,
    this->declarations.push_degenerate_list_at_head(&declarator_list->link);
 }
 
+/**
+ * Do the set of common optimizations passes
+ *
+ * \param ir                          List of instructions to be optimized
+ * \param linked                      Is the shader linked?  This enables
+ *                                    optimizations passes that remove code at
+ *                                    global scope and could cause linking to
+ *                                    fail.
+ * \param uniform_locations_assigned  Have locations already been assigned for
+ *                                    uniforms?  This prevents the declarations
+ *                                    of unused uniforms from being removed.
+ *                                    The setting of this flag only matters if
+ *                                    \c linked is \c true.
+ * \param max_unroll_iterations       Maximum number of loop iterations to be
+ *                                    unrolled.  Setting to 0 forces all loops
+ *                                    to be unrolled.
+ */
 bool
-do_common_optimization(exec_list *ir, bool linked, unsigned max_unroll_iterations)
+do_common_optimization(exec_list *ir, bool linked,
+		       bool uniform_locations_assigned,
+		       unsigned max_unroll_iterations)
 {
    GLboolean progress = GL_FALSE;
 
@@ -900,7 +919,7 @@ do_common_optimization(exec_list *ir, bool linked, unsigned max_unroll_iteration
    progress = do_copy_propagation(ir) || progress;
    progress = do_copy_propagation_elements(ir) || progress;
    if (linked)
-      progress = do_dead_code(ir) || progress;
+      progress = do_dead_code(ir, uniform_locations_assigned) || progress;
    else
       progress = do_dead_code_unlinked(ir) || progress;
    progress = do_dead_code_local(ir) || progress;
diff --git a/mesalib/src/glsl/ir_constant_expression.cpp b/mesalib/src/glsl/ir_constant_expression.cpp
index b3fe6cf96..83f084d88 100644
--- a/mesalib/src/glsl/ir_constant_expression.cpp
+++ b/mesalib/src/glsl/ir_constant_expression.cpp
@@ -39,6 +39,25 @@
 #include "ir_visitor.h"
 #include "glsl_types.h"
 
+/* Using C99 rounding functions for roundToEven() implementation is
+ * difficult, because round(), rint, and nearbyint() are affected by
+ * fesetenv(), which the application may have done for its own
+ * purposes.  Mesa's IROUND macro is close to what we want, but it
+ * rounds away from 0 on n + 0.5.
+ */
+static int
+round_to_even(float val)
+{
+   int rounded = IROUND(val);
+
+   if (val - floor(val) == 0.5) {
+      if (rounded % 2 != 0)
+	 rounded += val > 0 ? -1 : 1;
+   }
+
+   return rounded;
+}
+
 static float
 dot(ir_constant *op0, ir_constant *op1)
 {
@@ -196,6 +215,13 @@ ir_expression::constant_expression_value()
       }
       break;
 
+   case ir_unop_round_even:
+      assert(op[0]->type->base_type == GLSL_TYPE_FLOAT);
+      for (unsigned c = 0; c < op[0]->type->components(); c++) {
+	 data.f[c] = round_to_even(op[0]->value.f[c]);
+      }
+      break;
+
    case ir_unop_ceil:
       assert(op[0]->type->base_type == GLSL_TYPE_FLOAT);
       for (unsigned c = 0; c < op[0]->type->components(); c++) {
@@ -1324,6 +1350,9 @@ ir_call::constant_expression_value()
 			    * op[1]->value.f[c];
 	 }
       }
+   } else if (strcmp(callee, "round") == 0 ||
+	      strcmp(callee, "roundEven") == 0) {
+      expr = new(mem_ctx) ir_expression(ir_unop_round_even, op[0]);
    } else if (strcmp(callee, "sign") == 0) {
       expr = new(mem_ctx) ir_expression(ir_unop_sign, type, op[0], NULL);
    } else if (strcmp(callee, "sin") == 0) {
diff --git a/mesalib/src/glsl/ir_optimization.h b/mesalib/src/glsl/ir_optimization.h
index af80e26b9..7b32e84f0 100644
--- a/mesalib/src/glsl/ir_optimization.h
+++ b/mesalib/src/glsl/ir_optimization.h
@@ -37,7 +37,9 @@
 #define MOD_TO_FRACT       0x20
 #define INT_DIV_TO_MUL_RCP 0x40
 
-bool do_common_optimization(exec_list *ir, bool linked, unsigned max_unroll_iterations);
+bool do_common_optimization(exec_list *ir, bool linked,
+			    bool uniform_locations_assigned,
+			    unsigned max_unroll_iterations);
 
 bool do_algebraic(exec_list *instructions);
 bool do_constant_folding(exec_list *instructions);
@@ -46,7 +48,7 @@ bool do_constant_variable_unlinked(exec_list *instructions);
 bool do_copy_propagation(exec_list *instructions);
 bool do_copy_propagation_elements(exec_list *instructions);
 bool do_constant_propagation(exec_list *instructions);
-bool do_dead_code(exec_list *instructions);
+bool do_dead_code(exec_list *instructions, bool uniform_locations_assigned);
 bool do_dead_code_local(exec_list *instructions);
 bool do_dead_code_unlinked(exec_list *instructions);
 bool do_dead_functions(exec_list *instructions);
diff --git a/mesalib/src/glsl/ir_variable.cpp b/mesalib/src/glsl/ir_variable.cpp
index 6ae3b1f9e..1ee84d219 100644
--- a/mesalib/src/glsl/ir_variable.cpp
+++ b/mesalib/src/glsl/ir_variable.cpp
@@ -166,15 +166,15 @@ static struct gl_builtin_uniform_element gl_NormalScale_elements[] = {
    {NULL, {STATE_NORMAL_SCALE}, SWIZZLE_XXXX},
 };
 
-static struct gl_builtin_uniform_element gl_MESABumpRotMatrix0_elements[] = {
+static struct gl_builtin_uniform_element gl_BumpRotMatrix0MESA_elements[] = {
    {NULL, {STATE_INTERNAL, STATE_ROT_MATRIX_0}, SWIZZLE_XYZW},
 };
 
-static struct gl_builtin_uniform_element gl_MESABumpRotMatrix1_elements[] = {
+static struct gl_builtin_uniform_element gl_BumpRotMatrix1MESA_elements[] = {
    {NULL, {STATE_INTERNAL, STATE_ROT_MATRIX_1}, SWIZZLE_XYZW},
 };
 
-static struct gl_builtin_uniform_element gl_MESAFogParamsOptimized_elements[] = {
+static struct gl_builtin_uniform_element gl_FogParamsOptimizedMESA_elements[] = {
    {NULL, {STATE_INTERNAL, STATE_FOG_PARAMS_OPTIMIZED}, SWIZZLE_XYZW},
 };
 
@@ -232,11 +232,11 @@ MATRIX(gl_TextureMatrixInverseTranspose,
 
 static struct gl_builtin_uniform_element gl_NormalMatrix_elements[] = {
    { NULL, { STATE_MODELVIEW_MATRIX, 0, 0, 0, STATE_MATRIX_INVERSE},
-     SWIZZLE_XYZW },
+     MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_Z) },
    { NULL, { STATE_MODELVIEW_MATRIX, 0, 1, 1, STATE_MATRIX_INVERSE},
-     SWIZZLE_XYZW },
+     MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_Z) },
    { NULL, { STATE_MODELVIEW_MATRIX, 0, 2, 2, STATE_MATRIX_INVERSE},
-     SWIZZLE_XYZW },
+     MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_Z) },
 };
 
 #undef MATRIX
@@ -289,9 +289,9 @@ const struct gl_builtin_uniform_desc _mesa_builtin_uniform_desc[] = {
    STATEVAR(gl_NormalMatrix),
    STATEVAR(gl_NormalScale),
 
-   STATEVAR(gl_MESABumpRotMatrix0),
-   STATEVAR(gl_MESABumpRotMatrix1),
-   STATEVAR(gl_MESAFogParamsOptimized),
+   STATEVAR(gl_BumpRotMatrix0MESA),
+   STATEVAR(gl_BumpRotMatrix1MESA),
+   STATEVAR(gl_FogParamsOptimizedMESA),
    STATEVAR(gl_CurrentAttribVertMESA),
    STATEVAR(gl_CurrentAttribFragMESA),
 
@@ -615,6 +615,8 @@ generate_130_uniforms(exec_list *instructions,
 
    add_builtin_constant(instructions, symtab, "gl_MaxClipDistances",
                         state->Const.MaxClipPlanes);
+   add_builtin_constant(instructions, symtab, "gl_MaxVaryingComponents",
+			state->Const.MaxVaryingFloats);
 }
 
 
diff --git a/mesalib/src/glsl/link_uniforms.cpp b/mesalib/src/glsl/link_uniforms.cpp
new file mode 100644
index 000000000..6dd1f5354
--- /dev/null
+++ b/mesalib/src/glsl/link_uniforms.cpp
@@ -0,0 +1,85 @@
+/*
+ * Copyright © 2011 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "main/core.h"
+#include "ir.h"
+#include "linker.h"
+#include "glsl_symbol_table.h"
+#include "program/hash_table.h"
+
+/**
+ * \file link_uniforms.cpp
+ * Assign locations for GLSL uniforms.
+ *
+ * \author Ian Romanick <ian.d.romanick@intel.com>
+ */
+
+void
+uniform_field_visitor::process(ir_variable *var)
+{
+   const glsl_type *t = var->type;
+
+   /* Only strdup the name if we actually will need to modify it. */
+   if (t->is_record() || (t->is_array() && t->fields.array->is_record())) {
+      char *name = ralloc_strdup(NULL, var->name);
+      recursion(var->type, &name, strlen(name));
+      ralloc_free(name);
+   } else {
+      this->visit_field(t, var->name);
+   }
+}
+
+void
+uniform_field_visitor::recursion(const glsl_type *t, char **name,
+				 unsigned name_length)
+{
+   /* Records need to have each field processed individually.
+    *
+    * Arrays of records need to have each array element processed
+    * individually, then each field of the resulting array elements processed
+    * individually.
+    */
+   if (t->is_record()) {
+      for (unsigned i = 0; i < t->length; i++) {
+	 const char *field = t->fields.structure[i].name;
+
+	 /* Append '.field' to the current uniform name. */
+	 ralloc_asprintf_rewrite_tail(name, name_length, ".%s", field);
+
+	 recursion(t->fields.structure[i].type, name,
+		   name_length + 1 + strlen(field));
+      }
+   } else if (t->is_array() && t->fields.array->is_record()) {
+      for (unsigned i = 0; i < t->length; i++) {
+	 char subscript[13];
+
+	 /* Append the subscript to the current uniform name */
+	 const unsigned subscript_length = snprintf(subscript, 13, "[%u]", i);
+	 ralloc_asprintf_rewrite_tail(name, name_length, "%s", subscript);
+
+	 recursion(t->fields.array, name, name_length + subscript_length);
+      }
+   } else {
+      this->visit_field(t, *name);
+   }
+}
diff --git a/mesalib/src/glsl/linker.cpp b/mesalib/src/glsl/linker.cpp
index a7c38a342..beadec6f6 100644
--- a/mesalib/src/glsl/linker.cpp
+++ b/mesalib/src/glsl/linker.cpp
@@ -1742,14 +1742,10 @@ 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, 32))
+      while (do_common_optimization(prog->_LinkedShaders[i]->ir, true, false, 32))
 	 ;
    }
 
-   update_array_sizes(prog);
-
-   assign_uniform_locations(prog);
-
    /* FINISHME: The value of the max_attribute_index parameter is
     * FINISHME: implementation dependent based on the value of
     * FINISHME: GL_MAX_VERTEX_ATTRIBS.  GL_MAX_VERTEX_ATTRIBS must be
@@ -1785,6 +1781,12 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
    if (prog->_LinkedShaders[MESA_SHADER_VERTEX] != NULL) {
       demote_shader_inputs_and_outputs(prog->_LinkedShaders[MESA_SHADER_VERTEX],
 				       ir_var_out);
+
+      /* Eliminate code that is now dead due to unused vertex outputs being
+       * demoted.
+       */
+      while (do_dead_code(prog->_LinkedShaders[MESA_SHADER_VERTEX]->ir, false))
+	 ;
    }
 
    if (prog->_LinkedShaders[MESA_SHADER_GEOMETRY] != NULL) {
@@ -1793,14 +1795,30 @@ link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
       demote_shader_inputs_and_outputs(sh, ir_var_in);
       demote_shader_inputs_and_outputs(sh, ir_var_inout);
       demote_shader_inputs_and_outputs(sh, ir_var_out);
+
+      /* Eliminate code that is now dead due to unused geometry outputs being
+       * demoted.
+       */
+      while (do_dead_code(prog->_LinkedShaders[MESA_SHADER_GEOMETRY]->ir, false))
+	 ;
    }
 
    if (prog->_LinkedShaders[MESA_SHADER_FRAGMENT] != NULL) {
       gl_shader *const sh = prog->_LinkedShaders[MESA_SHADER_FRAGMENT];
 
       demote_shader_inputs_and_outputs(sh, ir_var_in);
+
+      /* Eliminate code that is now dead due to unused fragment inputs being
+       * demoted.  This shouldn't actually do anything other than remove
+       * declarations of the (now unused) global variables.
+       */
+      while (do_dead_code(prog->_LinkedShaders[MESA_SHADER_FRAGMENT]->ir, false))
+	 ;
    }
 
+   update_array_sizes(prog);
+   assign_uniform_locations(prog);
+
    /* OpenGL ES requires that a vertex shader and a fragment shader both be
     * present in a linked program.  By checking for use of shading language
     * version 1.00, we also catch the GL_ARB_ES2_compatibility case.
diff --git a/mesalib/src/glsl/linker.h b/mesalib/src/glsl/linker.h
index 769cf68b6..78c632961 100644
--- a/mesalib/src/glsl/linker.h
+++ b/mesalib/src/glsl/linker.h
@@ -1,3 +1,4 @@
+/* -*- c++ -*- */
 /*
  * Copyright © 2010 Intel Corporation
  *
@@ -29,4 +30,46 @@ extern bool
 link_function_calls(gl_shader_program *prog, gl_shader *main,
 		    gl_shader **shader_list, unsigned num_shaders);
 
+/**
+ * Class for processing all of the leaf fields of an uniform
+ *
+ * Leaves are, roughly speaking, the parts of the uniform that the application
+ * could query with \c glGetUniformLocation (or that could be returned by
+ * \c glGetActiveUniforms).
+ *
+ * Classes my derive from this class to implement specific functionality.
+ * This class only provides the mechanism to iterate over the leaves.  Derived
+ * classes must implement \c ::visit_field and may override \c ::process.
+ */
+class uniform_field_visitor {
+public:
+   /**
+    * Begin processing a uniform
+    *
+    * Classes that overload this function should call \c ::process from the
+    * base class to start the recursive processing of the uniform.
+    *
+    * \param var  The uniform variable that is to be processed
+    *
+    * Calls \c ::visit_field for each leaf of the uniform.
+    */
+   void process(ir_variable *var);
+
+protected:
+   /**
+    * Method invoked for each leaf of the uniform
+    *
+    * \param type  Type of the field.
+    * \param name  Fully qualified name of the field.
+    */
+   virtual void visit_field(const glsl_type *type, const char *name) = 0;
+
+private:
+   /**
+    * \param name_length  Length of the current name \b not including the
+    *                     terminating \c NUL character.
+    */
+   void recursion(const glsl_type *t, char **name, unsigned name_length);
+};
+
 #endif /* GLSL_LINKER_H */
diff --git a/mesalib/src/glsl/lower_if_to_cond_assign.cpp b/mesalib/src/glsl/lower_if_to_cond_assign.cpp
index 7b89a1539..2c5d5612d 100644
--- a/mesalib/src/glsl/lower_if_to_cond_assign.cpp
+++ b/mesalib/src/glsl/lower_if_to_cond_assign.cpp
@@ -79,6 +79,9 @@ public:
 bool
 lower_if_to_cond_assign(exec_list *instructions, unsigned max_depth)
 {
+   if (max_depth == UINT_MAX)
+      return false;
+
    ir_if_to_cond_assign_visitor v(max_depth);
 
    visit_list_elements(&v, instructions);
diff --git a/mesalib/src/glsl/main.cpp b/mesalib/src/glsl/main.cpp
index f0d951514..519abb5b5 100644
--- a/mesalib/src/glsl/main.cpp
+++ b/mesalib/src/glsl/main.cpp
@@ -177,7 +177,7 @@ compile_shader(struct gl_context *ctx, struct gl_shader *shader)
    if (!state->error && !shader->ir->is_empty()) {
       bool progress;
       do {
-	 progress = do_common_optimization(shader->ir, false, 32);
+	 progress = do_common_optimization(shader->ir, false, false, 32);
       } while (progress);
 
       validate_ir_tree(shader->ir);
diff --git a/mesalib/src/glsl/opt_dead_code.cpp b/mesalib/src/glsl/opt_dead_code.cpp
index cb500d2d1..5b9546ad4 100644
--- a/mesalib/src/glsl/opt_dead_code.cpp
+++ b/mesalib/src/glsl/opt_dead_code.cpp
@@ -42,7 +42,7 @@ static bool debug = false;
  * for usage on an unlinked instruction stream.
  */
 bool
-do_dead_code(exec_list *instructions)
+do_dead_code(exec_list *instructions, bool uniform_locations_assigned)
 {
    ir_variable_refcount_visitor v;
    bool progress = false;
@@ -94,10 +94,11 @@ do_dead_code(exec_list *instructions)
 	  */
 
 	 /* uniform initializers are precious, and could get used by another
-	  * stage.
+	  * stage.  Also, once uniform locations have been assigned, the
+	  * declaration cannot be deleted.
 	  */
 	 if (entry->var->mode == ir_var_uniform &&
-	     entry->var->constant_value)
+	     (uniform_locations_assigned || entry->var->constant_value))
 	    continue;
 
 	 entry->var->remove();
@@ -132,7 +133,12 @@ do_dead_code_unlinked(exec_list *instructions)
 	 foreach_iter(exec_list_iterator, sigiter, *f) {
 	    ir_function_signature *sig =
 	       (ir_function_signature *) sigiter.get();
-	    if (do_dead_code(&sig->body))
+	    /* The setting of the uniform_locations_assigned flag here is
+	     * irrelevent.  If there is a uniform declaration encountered
+	     * inside the body of the function, something has already gone
+	     * terribly, terribly wrong.
+	     */
+	    if (do_dead_code(&sig->body, false))
 	       progress = true;
 	 }
       }
diff --git a/mesalib/src/glsl/ralloc.c b/mesalib/src/glsl/ralloc.c
index fb48a91c5..f5f3934ac 100644
--- a/mesalib/src/glsl/ralloc.c
+++ b/mesalib/src/glsl/ralloc.c
@@ -439,7 +439,28 @@ ralloc_asprintf_append(char **str, const char *fmt, ...)
 bool
 ralloc_vasprintf_append(char **str, const char *fmt, va_list args)
 {
-   size_t existing_length, new_length;
+   size_t existing_length;
+   assert(str != NULL);
+   existing_length = *str ? strlen(*str) : 0;
+   return ralloc_vasprintf_rewrite_tail(str, existing_length, fmt, args);
+}
+
+bool
+ralloc_asprintf_rewrite_tail(char **str, size_t start, const char *fmt, ...)
+{
+   bool success;
+   va_list args;
+   va_start(args, fmt);
+   success = ralloc_vasprintf_rewrite_tail(str, start, fmt, args);
+   va_end(args);
+   return success;
+}
+
+bool
+ralloc_vasprintf_rewrite_tail(char **str, size_t start, const char *fmt,
+			      va_list args)
+{
+   size_t new_length;
    char *ptr;
 
    assert(str != NULL);
@@ -450,14 +471,13 @@ ralloc_vasprintf_append(char **str, const char *fmt, va_list args)
       return true;
    }
 
-   existing_length = strlen(*str);
    new_length = printf_length(fmt, args);
 
-   ptr = resize(*str, existing_length + new_length + 1);
+   ptr = resize(*str, start + new_length + 1);
    if (unlikely(ptr == NULL))
       return false;
 
-   vsnprintf(ptr + existing_length, new_length + 1, fmt, args);
+   vsnprintf(ptr + start, new_length + 1, fmt, args);
    *str = ptr;
    return true;
 }
diff --git a/mesalib/src/glsl/ralloc.h b/mesalib/src/glsl/ralloc.h
index d5338152f..1324f3466 100644
--- a/mesalib/src/glsl/ralloc.h
+++ b/mesalib/src/glsl/ralloc.h
@@ -314,9 +314,60 @@ char *ralloc_asprintf (const void *ctx, const char *fmt, ...);
 char *ralloc_vasprintf(const void *ctx, const char *fmt, va_list args);
 
 /**
+ * Rewrite the tail of an existing string, starting at a given index.
+ *
+ * Overwrites the contents of *str starting at \p start with newly formatted
+ * text, including a new null-terminator.  Allocates more memory as necessary.
+ *
+ * This can be used to append formatted text when the length of the existing
+ * string is already known, saving a strlen() call.
+ *
+ * \sa ralloc_asprintf_append
+ *
+ * \param str   The string to be updated.
+ * \param start The index to start appending new data at.
+ * \param fmt   A printf-style formatting string
+ *
+ * \p str will be updated to the new pointer unless allocation fails.
+ *
+ * \return True unless allocation failed.
+ */
+bool ralloc_asprintf_rewrite_tail(char **str, size_t start,
+				  const char *fmt, ...);
+
+/**
+ * Rewrite the tail of an existing string, starting at a given index.
+ *
+ * Overwrites the contents of *str starting at \p start with newly formatted
+ * text, including a new null-terminator.  Allocates more memory as necessary.
+ *
+ * This can be used to append formatted text when the length of the existing
+ * string is already known, saving a strlen() call.
+ *
+ * \sa ralloc_vasprintf_append
+ *
+ * \param str   The string to be updated.
+ * \param start The index to start appending new data at.
+ * \param fmt   A printf-style formatting string
+ * \param args  A va_list containing the data to be formatted
+ *
+ * \p str will be updated to the new pointer unless allocation fails.
+ *
+ * \return True unless allocation failed.
+ */
+bool ralloc_vasprintf_rewrite_tail(char **str, size_t start, const char *fmt,
+				   va_list args);
+
+/**
  * Append formatted text to the supplied string.
  *
+ * This is equivalent to
+ * \code
+ * ralloc_asprintf_rewrite_tail(str, strlen(*str), fmt, ...)
+ * \endcode
+ *
  * \sa ralloc_asprintf
+ * \sa ralloc_asprintf_rewrite_tail
  * \sa ralloc_strcat
  *
  * \p str will be updated to the new pointer unless allocation fails.
@@ -328,7 +379,13 @@ bool ralloc_asprintf_append (char **str, const char *fmt, ...);
 /**
  * Append formatted text to the supplied string, given a va_list.
  *
+ * This is equivalent to
+ * \code
+ * ralloc_vasprintf_rewrite_tail(str, strlen(*str), fmt, args)
+ * \endcode
+ *
  * \sa ralloc_vasprintf
+ * \sa ralloc_vasprintf_rewrite_tail
  * \sa ralloc_strcat
  *
  * \p str will be updated to the new pointer unless allocation fails.
diff --git a/mesalib/src/glsl/test_optpass.cpp b/mesalib/src/glsl/test_optpass.cpp
index 89b7f8338..6abafb5d3 100644
--- a/mesalib/src/glsl/test_optpass.cpp
+++ b/mesalib/src/glsl/test_optpass.cpp
@@ -64,7 +64,7 @@ do_optimization(struct exec_list *ir, const char *optimization)
 
    if (sscanf(optimization, "do_common_optimization ( %d , %d ) ",
               &int_0, &int_1) == 2) {
-      return do_common_optimization(ir, int_0 != 0, int_1);
+      return do_common_optimization(ir, int_0 != 0, false, int_1);
    } else if (strcmp(optimization, "do_algebraic") == 0) {
       return do_algebraic(ir);
    } else if (strcmp(optimization, "do_constant_folding") == 0) {
@@ -80,7 +80,7 @@ do_optimization(struct exec_list *ir, const char *optimization)
    } else if (strcmp(optimization, "do_constant_propagation") == 0) {
       return do_constant_propagation(ir);
    } else if (strcmp(optimization, "do_dead_code") == 0) {
-      return do_dead_code(ir);
+      return do_dead_code(ir, false);
    } else if (strcmp(optimization, "do_dead_code_local") == 0) {
       return do_dead_code_local(ir);
    } else if (strcmp(optimization, "do_dead_code_unlinked") == 0) {
diff --git a/mesalib/src/mesa/drivers/dri/swrast/swrast.c b/mesalib/src/mesa/drivers/dri/swrast/swrast.c
index 1bf39435b..b1abc502f 100644
--- a/mesalib/src/mesa/drivers/dri/swrast/swrast.c
+++ b/mesalib/src/mesa/drivers/dri/swrast/swrast.c
@@ -57,6 +57,7 @@
 #include "main/texstate.h"
 
 #include "swrast_priv.h"
+#include "swrast/s_context.h"
 
 
 /**
@@ -72,6 +73,7 @@ static void swrastSetTexBuffer2(__DRIcontext *pDRICtx, GLint target,
     struct gl_texture_unit *texUnit;
     struct gl_texture_object *texObj;
     struct gl_texture_image *texImage;
+    struct swrast_texture_image *swImage;
     uint32_t internalFormat;
     gl_format texFormat;
 
@@ -82,6 +84,7 @@ static void swrastSetTexBuffer2(__DRIcontext *pDRICtx, GLint target,
     texUnit = _mesa_get_current_tex_unit(&dri_ctx->Base);
     texObj = _mesa_select_tex_object(&dri_ctx->Base, texUnit, target);
     texImage = _mesa_get_tex_image(&dri_ctx->Base, texObj, target, 0);
+    swImage = swrast_texture_image(texImage);
 
     _mesa_lock_texture(&dri_ctx->Base, texObj);
 
@@ -95,7 +98,7 @@ static void swrastSetTexBuffer2(__DRIcontext *pDRICtx, GLint target,
     _mesa_init_teximage_fields(&dri_ctx->Base, target, texImage,
 			       w, h, 1, 0, internalFormat, texFormat);
 
-    sPriv->swrast_loader->getImage(dPriv, x, y, w, h, (char *)texImage->Data,
+    sPriv->swrast_loader->getImage(dPriv, x, y, w, h, (char *)swImage->Data,
 				   dPriv->loaderPrivate);
 
     _mesa_unlock_texture(&dri_ctx->Base, texObj);
diff --git a/mesalib/src/mesa/main/APIspec.xml b/mesalib/src/mesa/main/APIspec.xml
index 574480e28..a92bb437c 100644
--- a/mesalib/src/mesa/main/APIspec.xml
+++ b/mesalib/src/mesa/main/APIspec.xml
@@ -631,12 +631,6 @@
 	<desc name="pname">
 		<value name="GL_RGB_SCALE"/>
 		<value name="GL_ALPHA_SCALE"/>
-
-		<desc name="param" convert="true" error="GL_INVALID_VALUE">
-			<value name="1.0"/>
-			<value name="2.0"/>
-			<value name="4.0"/>
-		</desc>
 	</desc>
 
 	<desc name="pname">
diff --git a/mesalib/src/mesa/main/attrib.c b/mesalib/src/mesa/main/attrib.c
index e67957d4d..1dc1c1b97 100644
--- a/mesalib/src/mesa/main/attrib.c
+++ b/mesalib/src/mesa/main/attrib.c
@@ -1279,30 +1279,6 @@ _mesa_PopAttrib(void)
 
 
 /**
- * Helper for incrementing/decrementing vertex buffer object reference
- * counts when pushing/popping the GL_CLIENT_VERTEX_ARRAY_BIT attribute group.
- */
-static void
-adjust_buffer_object_ref_counts(struct gl_array_object *arrayObj, GLint step)
-{
-   GLuint i;
-
-   arrayObj->Vertex.BufferObj->RefCount += step;
-   arrayObj->Weight.BufferObj->RefCount += step;
-   arrayObj->Normal.BufferObj->RefCount += step;
-   arrayObj->Color.BufferObj->RefCount += step;
-   arrayObj->SecondaryColor.BufferObj->RefCount += step;
-   arrayObj->FogCoord.BufferObj->RefCount += step;
-   arrayObj->Index.BufferObj->RefCount += step;
-   arrayObj->EdgeFlag.BufferObj->RefCount += step;
-   for (i = 0; i < Elements(arrayObj->TexCoord); i++)
-      arrayObj->TexCoord[i].BufferObj->RefCount += step;
-   for (i = 0; i < Elements(arrayObj->VertexAttrib); i++)
-      arrayObj->VertexAttrib[i].BufferObj->RefCount += step;
-}
-
-
-/**
  * Copy gl_pixelstore_attrib from src to dst, updating buffer
  * object refcounts.
  */
@@ -1327,6 +1303,151 @@ copy_pixelstore(struct gl_context *ctx,
 #define GL_CLIENT_PACK_BIT (1<<20)
 #define GL_CLIENT_UNPACK_BIT (1<<21)
 
+/**
+ * Copy gl_array_object from src to dest.
+ * 'dest' must be in an initialized state.
+ */
+static void
+copy_array_object(struct gl_context *ctx,
+                  struct gl_array_object *dest,
+                  struct gl_array_object *src)
+{
+   GLuint i;
+
+   /* skip Name */
+   /* skip RefCount */
+
+   /* In theory must be the same anyway, but on recreate make sure it matches */
+   dest->VBOonly = src->VBOonly;
+
+   _mesa_copy_client_array(ctx, &dest->Vertex, &src->Vertex);
+   _mesa_copy_client_array(ctx, &dest->Weight, &src->Weight);
+   _mesa_copy_client_array(ctx, &dest->Normal, &src->Normal);
+   _mesa_copy_client_array(ctx, &dest->Color, &src->Color);
+   _mesa_copy_client_array(ctx, &dest->SecondaryColor, &src->SecondaryColor);
+   _mesa_copy_client_array(ctx, &dest->FogCoord, &src->FogCoord);
+   _mesa_copy_client_array(ctx, &dest->Index, &src->Index);
+   _mesa_copy_client_array(ctx, &dest->EdgeFlag, &src->EdgeFlag);
+#if FEATURE_point_size_array
+   _mesa_copy_client_array(ctx, &dest->PointSize, &src->PointSize);
+#endif
+   for (i = 0; i < Elements(src->TexCoord); i++)
+      _mesa_copy_client_array(ctx, &dest->TexCoord[i], &src->TexCoord[i]);
+   for (i = 0; i < Elements(src->VertexAttrib); i++)
+      _mesa_copy_client_array(ctx, &dest->VertexAttrib[i], &src->VertexAttrib[i]);
+
+   /* _Enabled must be the same than on push */
+   dest->_Enabled = src->_Enabled;
+   dest->_MaxElement = src->_MaxElement;
+}
+
+/**
+ * Copy gl_array_attrib from src to dest.
+ * 'dest' must be in an initialized state.
+ */
+static void
+copy_array_attrib(struct gl_context *ctx,
+                  struct gl_array_attrib *dest,
+                  struct gl_array_attrib *src)
+{
+   /* skip ArrayObj */
+   /* skip DefaultArrayObj, Objects */
+   dest->ActiveTexture = src->ActiveTexture;
+   dest->LockFirst = src->LockFirst;
+   dest->LockCount = src->LockCount;
+   dest->PrimitiveRestart = src->PrimitiveRestart;
+   dest->RestartIndex = src->RestartIndex;
+   /* skip NewState */
+   /* skip RebindArrays */
+
+   copy_array_object(ctx, dest->ArrayObj, src->ArrayObj);
+
+   /* skip ArrayBufferObj */
+   /* skip ElementArrayBufferObj */
+}
+
+/**
+ * Save the content of src to dest.
+ */
+static void
+save_array_attrib(struct gl_context *ctx,
+                  struct gl_array_attrib *dest,
+                  struct gl_array_attrib *src)
+{
+   /* Set the Name, needed for restore, but do never overwrite.
+    * Needs to match value in the object hash. */
+   dest->ArrayObj->Name = src->ArrayObj->Name;
+   /* And copy all of the rest. */
+   copy_array_attrib(ctx, dest, src);
+
+   /* Just reference them here */
+   _mesa_reference_buffer_object(ctx, &dest->ArrayBufferObj,
+                                 src->ArrayBufferObj);
+   _mesa_reference_buffer_object(ctx, &dest->ElementArrayBufferObj,
+                                 src->ElementArrayBufferObj);
+}
+
+/**
+ * Restore the content of src to dest.
+ */
+static void
+restore_array_attrib(struct gl_context *ctx,
+                     struct gl_array_attrib *dest,
+                     struct gl_array_attrib *src)
+{
+   /* Restore or recreate the array object by its name ... */
+   _mesa_BindVertexArrayAPPLE(src->ArrayObj->Name);
+
+   /* ... and restore its content */
+   copy_array_attrib(ctx, dest, src);
+
+   /* Restore or recreate the buffer objects by the names ... */
+   _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
+                       src->ArrayBufferObj->Name);
+   _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB,
+                       src->ElementArrayBufferObj->Name);
+
+   /* Better safe than sorry?! */
+   dest->RebindArrays = GL_TRUE;
+
+   /* FIXME: Should some bits in ctx->Array->NewState also be set
+    * FIXME: here?  It seems like it should be set to inclusive-or
+    * FIXME: of the old ArrayObj->_Enabled and the new _Enabled.
+    * ... just do it.
+    */
+   dest->NewState |= src->ArrayObj->_Enabled | dest->ArrayObj->_Enabled;
+}
+
+/**
+ * init/alloc the fields of 'attrib'.
+ * Needs to the init part matching free_array_attrib_data below.
+ */
+static void
+init_array_attrib_data(struct gl_context *ctx,
+                       struct gl_array_attrib *attrib)
+{
+   /* Get a non driver gl_array_object. */
+   attrib->ArrayObj = CALLOC_STRUCT( gl_array_object );
+   _mesa_initialize_array_object(ctx, attrib->ArrayObj, 0);
+}
+
+/**
+ * Free/unreference the fields of 'attrib' but don't delete it (that's
+ * done later in the calling code).
+ * Needs to the cleanup part matching init_array_attrib_data above.
+ */
+static void
+free_array_attrib_data(struct gl_context *ctx,
+                       struct gl_array_attrib *attrib)
+{
+   /* We use a non driver array object, so don't just unref since we would
+    * end up using the drivers DeleteArrayObject function for deletion. */
+   _mesa_delete_array_object(ctx, attrib->ArrayObj);
+   attrib->ArrayObj = 0;
+   _mesa_reference_buffer_object(ctx, &attrib->ArrayBufferObj, NULL);
+   _mesa_reference_buffer_object(ctx, &attrib->ElementArrayBufferObj, NULL);
+}
+
 
 void GLAPIENTRY
 _mesa_PushClientAttrib(GLbitfield mask)
@@ -1360,26 +1481,10 @@ _mesa_PushClientAttrib(GLbitfield mask)
 
    if (mask & GL_CLIENT_VERTEX_ARRAY_BIT) {
       struct gl_array_attrib *attr;
-      struct gl_array_object *obj;
-
-      attr = MALLOC_STRUCT( gl_array_attrib );
-      obj = MALLOC_STRUCT( gl_array_object );
-
-#if FEATURE_ARB_vertex_buffer_object
-      /* increment ref counts since we're copying pointers to these objects */
-      ctx->Array.ArrayBufferObj->RefCount++;
-      ctx->Array.ElementArrayBufferObj->RefCount++;
-#endif
-
-      memcpy( attr, &ctx->Array, sizeof(struct gl_array_attrib) );
-      memcpy( obj, ctx->Array.ArrayObj, sizeof(struct gl_array_object) );
-
-      attr->ArrayObj = obj;
-
+      attr = CALLOC_STRUCT( gl_array_attrib );
+      init_array_attrib_data(ctx, attr);
+      save_array_attrib(ctx, attr, &ctx->Array);
       save_attrib_data(&head, GL_CLIENT_VERTEX_ARRAY_BIT, attr);
-
-      /* bump reference counts on buffer objects */
-      adjust_buffer_object_ref_counts(ctx->Array.ArrayObj, 1);
    }
 
    ctx->ClientAttribStack[ctx->ClientAttribStackDepth] = head;
@@ -1426,36 +1531,10 @@ _mesa_PopClientAttrib(void)
 	    ctx->NewState |= _NEW_PACKUNPACK;
             break;
          case GL_CLIENT_VERTEX_ARRAY_BIT: {
-	    struct gl_array_attrib * data =
+	    struct gl_array_attrib * attr =
 	      (struct gl_array_attrib *) node->data;
-
-            adjust_buffer_object_ref_counts(ctx->Array.ArrayObj, -1);
-	 
-            ctx->Array.ActiveTexture = data->ActiveTexture;
-	    if (data->LockCount != 0)
-	       _mesa_LockArraysEXT(data->LockFirst, data->LockCount);
-	    else if (ctx->Array.LockCount)
-	       _mesa_UnlockArraysEXT();
-
-	    _mesa_BindVertexArrayAPPLE( data->ArrayObj->Name );
-	    
-#if FEATURE_ARB_vertex_buffer_object
-            _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
-                                data->ArrayBufferObj->Name);
-            _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB,
-                                data->ElementArrayBufferObj->Name);
-#endif
-
-	    memcpy( ctx->Array.ArrayObj, data->ArrayObj,
-		    sizeof( struct gl_array_object ) );
-
-	    FREE( data->ArrayObj );
-	    
-	    /* FIXME: Should some bits in ctx->Array->NewState also be set
-	     * FIXME: here?  It seems like it should be set to inclusive-or
-	     * FIXME: of the old ArrayObj->_Enabled and the new _Enabled.
-	     */
-
+            restore_array_attrib(ctx, &ctx->Array, attr);
+            free_array_attrib_data(ctx, attr);
 	    ctx->NewState |= _NEW_ARRAY;
             break;
 	 }
diff --git a/mesalib/src/mesa/main/bufferobj.c b/mesalib/src/mesa/main/bufferobj.c
index 67241eb1e..f2e8f568c 100644
--- a/mesalib/src/mesa/main/bufferobj.c
+++ b/mesalib/src/mesa/main/bufferobj.c
@@ -571,7 +571,7 @@ bind_buffer_object(struct gl_context *ctx, GLenum target, GLuint buffer)
 
    /* Get pointer to old buffer object (to be unbound) */
    oldBufObj = *bindTarget;
-   if (oldBufObj && oldBufObj->Name == buffer)
+   if (oldBufObj && oldBufObj->Name == buffer && !oldBufObj->DeletePending)
       return;   /* rebinding the same buffer object- no change */
 
    /*
@@ -773,6 +773,17 @@ _mesa_DeleteBuffersARB(GLsizei n, const GLuint *ids)
 
          /* The ID is immediately freed for re-use */
          _mesa_HashRemove(ctx->Shared->BufferObjects, ids[i]);
+         /* Make sure we do not run into the classic ABA problem on bind.
+          * We don't want to allow re-binding a buffer object that's been
+          * "deleted" by glDeleteBuffers().
+          *
+          * The explicit rebinding to the default object in the current context
+          * prevents the above in the current context, but another context
+          * sharing the same objects might suffer from this problem.
+          * The alternative would be to do the hash lookup in any case on bind
+          * which would introduce more runtime overhead than this.
+          */
+         bufObj->DeletePending = GL_TRUE;
          _mesa_reference_buffer_object(ctx, &bufObj, NULL);
       }
    }
diff --git a/mesalib/src/mesa/main/compiler.h b/mesalib/src/mesa/main/compiler.h
index 89d6cda91..921e30222 100644
--- a/mesalib/src/mesa/main/compiler.h
+++ b/mesalib/src/mesa/main/compiler.h
@@ -150,7 +150,7 @@ extern "C" {
  * inline a static function that we later use in an alias. - ajax
  */
 #ifndef PUBLIC
-#  if defined(__GNUC__) || (defined(__SUNPRO_C) && (__SUNPRO_C >= 0x590))
+#  if (defined(__GNUC__) && __GNUC__ >= 4) || (defined(__SUNPRO_C) && (__SUNPRO_C >= 0x590))
 #    define PUBLIC __attribute__((visibility("default")))
 #    define USED __attribute__((used))
 #  else
diff --git a/mesalib/src/mesa/main/context.h b/mesalib/src/mesa/main/context.h
index a4c7ba2c5..a9df0a866 100644
--- a/mesalib/src/mesa/main/context.h
+++ b/mesalib/src/mesa/main/context.h
@@ -233,7 +233,7 @@ do {								\
  * glBegin()/glEnd() pair, with return value.
  * 
  * \param ctx GL context.
- * \param retval value to return value in case the assertion fails.
+ * \param retval value to return in case the assertion fails.
  */
 #define ASSERT_OUTSIDE_BEGIN_END_WITH_RETVAL(ctx, retval)		\
 do {									\
@@ -274,7 +274,7 @@ do {									\
  * glBegin()/glEnd() pair and flush the vertices, with return value.
  * 
  * \param ctx GL context.
- * \param retval value to return value in case the assertion fails.
+ * \param retval value to return in case the assertion fails.
  */
 #define ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH_WITH_RETVAL(ctx, retval)	\
 do {									\
diff --git a/mesalib/src/mesa/main/dlist.c b/mesalib/src/mesa/main/dlist.c
index 625649e9d..d901bddf8 100644
--- a/mesalib/src/mesa/main/dlist.c
+++ b/mesalib/src/mesa/main/dlist.c
@@ -939,7 +939,9 @@ unpack_image(struct gl_context *ctx, GLuint dimensions,
       }
       return image;
    }
+
    /* bad access! */
+   _mesa_error(ctx, GL_INVALID_OPERATION, "invalid PBO access");
    return NULL;
 }
 
diff --git a/mesalib/src/mesa/main/fbobject.c b/mesalib/src/mesa/main/fbobject.c
index f9da54e73..c56062ac6 100644
--- a/mesalib/src/mesa/main/fbobject.c
+++ b/mesalib/src/mesa/main/fbobject.c
@@ -2486,6 +2486,7 @@ _mesa_GetFramebufferAttachmentParameterivEXT(GLenum target, GLenum attachment,
 void GLAPIENTRY
 _mesa_GenerateMipmapEXT(GLenum target)
 {
+   struct gl_texture_image *srcImage;
    struct gl_texture_object *texObj;
    GLboolean error;
 
@@ -2532,6 +2533,13 @@ _mesa_GenerateMipmapEXT(GLenum target)
    }
 
    _mesa_lock_texture(ctx, texObj);
+
+   srcImage = _mesa_select_tex_image(ctx, texObj, target, texObj->BaseLevel);
+   if (!srcImage) {
+      _mesa_unlock_texture(ctx, texObj);
+      return;
+   }
+
    if (target == GL_TEXTURE_CUBE_MAP) {
       GLuint face;
       for (face = 0; face < 6; face++)
diff --git a/mesalib/src/mesa/main/ff_fragment_shader.cpp b/mesalib/src/mesa/main/ff_fragment_shader.cpp
index 160a97c0c..b5500714b 100644
--- a/mesalib/src/mesa/main/ff_fragment_shader.cpp
+++ b/mesalib/src/mesa/main/ff_fragment_shader.cpp
@@ -1137,8 +1137,8 @@ load_texunit_bumpmap( struct texenv_fragment_program *p, GLuint unit )
    ir_variable *rot_mat_0_var, *rot_mat_1_var;
    ir_dereference_variable *rot_mat_0, *rot_mat_1;
 
-   rot_mat_0_var = p->shader->symbols->get_variable("gl_MESABumpRotMatrix0");
-   rot_mat_1_var = p->shader->symbols->get_variable("gl_MESABumpRotMatrix1");
+   rot_mat_0_var = p->shader->symbols->get_variable("gl_BumpRotMatrix0MESA");
+   rot_mat_1_var = p->shader->symbols->get_variable("gl_BumpRotMatrix1MESA");
    rot_mat_0 = new(p->mem_ctx) ir_dereference_variable(rot_mat_0_var);
    rot_mat_1 = new(p->mem_ctx) ir_dereference_variable(rot_mat_1_var);
 
@@ -1229,7 +1229,7 @@ emit_fog_instructions(struct texenv_fragment_program *p,
    temp = new(p->mem_ctx) ir_dereference_variable(fog_result);
    fragcolor = new(p->mem_ctx) ir_swizzle(temp, 0, 1, 2, 3, 3);
 
-   oparams = p->shader->symbols->get_variable("gl_MESAFogParamsOptimized");
+   oparams = p->shader->symbols->get_variable("gl_FogParamsOptimizedMESA");
    fogcoord = p->shader->symbols->get_variable("gl_FogFragCoord");
    params = p->shader->symbols->get_variable("gl_Fog");
    f = new(p->mem_ctx) ir_dereference_variable(fogcoord);
@@ -1464,7 +1464,7 @@ create_new_program(struct gl_context *ctx, struct state_key *key)
 
    validate_ir_tree(p.shader->ir);
 
-   while (do_common_optimization(p.shader->ir, false, 32))
+   while (do_common_optimization(p.shader->ir, false, false, 32))
       ;
    reparent_ir(p.shader->ir, p.shader->ir);
 
diff --git a/mesalib/src/mesa/main/imports.c b/mesalib/src/mesa/main/imports.c
index 345a1c53e..2469e4265 100644
--- a/mesalib/src/mesa/main/imports.c
+++ b/mesalib/src/mesa/main/imports.c
@@ -514,7 +514,7 @@ _mesa_ffsll(int64_t val)
 #endif
 
 #if !defined(__GNUC__) ||\
-   ((_GNUC__ == 3 && __GNUC_MINOR__ < 4) && __GNUC__ < 4)
+   ((__GNUC__ * 100 + __GNUC_MINOR__) < 304) /* Not gcc 3.4 or later */
 /**
  * Return number of bits set in given GLuint.
  */
diff --git a/mesalib/src/mesa/main/imports.h b/mesalib/src/mesa/main/imports.h
index 20fa148fe..797f35742 100644
--- a/mesalib/src/mesa/main/imports.h
+++ b/mesalib/src/mesa/main/imports.h
@@ -453,7 +453,7 @@ static inline int32_t
 _mesa_next_pow_two_32(uint32_t x)
 {
 #if defined(__GNUC__) && \
-	((__GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
+	((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
 	uint32_t y = (x != 1);
 	return (1 + y) << ((__builtin_clz(x - y) ^ 31) );
 #else
@@ -472,7 +472,7 @@ static inline int64_t
 _mesa_next_pow_two_64(uint64_t x)
 {
 #if defined(__GNUC__) && \
-	((__GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
+	((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
 	uint64_t y = (x != 1);
 	if (sizeof(x) == sizeof(long))
 		return (1 + y) << ((__builtin_clzl(x - y) ^ 63));
@@ -499,7 +499,7 @@ static inline GLuint
 _mesa_logbase2(GLuint n)
 {
 #if defined(__GNUC__) && \
-   ((__GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
+   ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
    return (31 - __builtin_clz(n | 1));
 #else
    GLuint pos = 0;
@@ -576,7 +576,7 @@ _mesa_init_sqrt_table(void);
 #define _mesa_ffs(i)  ffs(i)
 #define _mesa_ffsll(i)  ffsll(i)
 
-#if ((_GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
+#if ((__GNUC__ * 100 + __GNUC_MINOR__) >= 304) /* gcc 3.4 or later */
 #define _mesa_bitcount(i) __builtin_popcount(i)
 #define _mesa_bitcount_64(i) __builtin_popcountll(i)
 #else
diff --git a/mesalib/src/mesa/main/mtypes.h b/mesalib/src/mesa/main/mtypes.h
index 17c645a7e..411768641 100644
--- a/mesalib/src/mesa/main/mtypes.h
+++ b/mesalib/src/mesa/main/mtypes.h
@@ -1261,11 +1261,6 @@ struct gl_texture_image
    GLuint Level;                /**< Which mipmap level am I? */
    /** Cube map face: index into gl_texture_object::Image[] array */
    GLuint Face;
-
-   GLuint RowStride;		/**< Padded width in units of texels */
-   GLuint *ImageOffsets;        /**< if 3D texture: array [Depth] of offsets to
-                                     each 2D slice in 'Data', in texels */
-   GLvoid *Data;		/**< Image data, accessed via FetchTexel() */
 };
 
 
@@ -1535,6 +1530,7 @@ struct gl_buffer_object
    GLintptr Offset;     /**< Mapped offset */
    GLsizeiptr Length;   /**< Mapped length */
    /*@}*/
+   GLboolean DeletePending;   /**< true if buffer object is removed from the hash */
    GLboolean Written;   /**< Ever written to? (for debugging) */
    GLboolean Purgeable; /**< Is the buffer purgeable under memory pressure? */
 };
@@ -2736,6 +2732,20 @@ struct gl_constants
 
    /* GL_ARB_robustness */
    GLenum ResetStrategy;
+
+   /**
+    * Whether the implementation strips out and ignores texture borders.
+    *
+    * Many GPU hardware implementations don't support rendering with texture
+    * borders and mipmapped textures.  (Note: not static border color, but the
+    * old 1-pixel border around each edge).  Implementations then have to do
+    * slow fallbacks to be correct, or just ignore the border and be fast but
+    * wrong.  Setting the flag stripts the border off of TexImage calls,
+    * providing "fast but wrong" at significantly reduced driver complexity.
+    *
+    * Texture borders are deprecated in GL 3.0.
+    **/
+   GLboolean StripTextureBorder;
 };
 
 
diff --git a/mesalib/src/mesa/main/pack.c b/mesalib/src/mesa/main/pack.c
index 092e541c5..ecdeaf5dc 100644
--- a/mesalib/src/mesa/main/pack.c
+++ b/mesalib/src/mesa/main/pack.c
@@ -3717,6 +3717,7 @@ _mesa_unpack_color_span_ubyte(struct gl_context *ctx,
 
          if (!indexes) {
             _mesa_error(ctx, GL_OUT_OF_MEMORY, "pixel unpacking");
+            free(rgba);
             return;
          }
 
diff --git a/mesalib/src/mesa/main/pixelstore.c b/mesalib/src/mesa/main/pixelstore.c
index d957950ed..81474491d 100644
--- a/mesalib/src/mesa/main/pixelstore.c
+++ b/mesalib/src/mesa/main/pixelstore.c
@@ -214,7 +214,7 @@ _mesa_PixelStorei( GLenum pname, GLint param )
 void GLAPIENTRY
 _mesa_PixelStoref( GLenum pname, GLfloat param )
 {
-   _mesa_PixelStorei( pname, (GLint) param );
+   _mesa_PixelStorei( pname, IROUND(param) );
 }
 
 
diff --git a/mesalib/src/mesa/main/state.c b/mesalib/src/mesa/main/state.c
index 98ca733c0..80fd03b91 100644
--- a/mesalib/src/mesa/main/state.c
+++ b/mesalib/src/mesa/main/state.c
@@ -461,7 +461,7 @@ static void
 update_twoside(struct gl_context *ctx)
 {
    if (ctx->Shader.CurrentVertexProgram ||
-       ctx->VertexProgram.Current) {
+       ctx->VertexProgram._Enabled) {
       ctx->VertexProgram._TwoSideEnabled = ctx->VertexProgram.TwoSideEnabled;
    } else {
       ctx->VertexProgram._TwoSideEnabled = (ctx->Light.Enabled &&
diff --git a/mesalib/src/mesa/main/texcompress.c b/mesalib/src/mesa/main/texcompress.c
index 03e05d5ef..0458b9b68 100644
--- a/mesalib/src/mesa/main/texcompress.c
+++ b/mesalib/src/mesa/main/texcompress.c
@@ -461,8 +461,8 @@ _mesa_decompress_image(gl_format format, GLuint width, GLuint height,
 
    /* setup dummy texture image info */
    memset(&texImage, 0, sizeof(texImage));
-   texImage.Base.Data = (void *) src;
-   texImage.Base.RowStride = srcRowStride;
+   texImage.Data = (void *) src;
+   texImage.RowStride = srcRowStride;
 
    switch (format) {
    /* DXT formats */
diff --git a/mesalib/src/mesa/main/texcompress_fxt1.c b/mesalib/src/mesa/main/texcompress_fxt1.c
index 41630a47c..d5c73e3b4 100644
--- a/mesalib/src/mesa/main/texcompress_fxt1.c
+++ b/mesalib/src/mesa/main/texcompress_fxt1.c
@@ -177,7 +177,7 @@ _mesa_fetch_texel_2d_f_rgba_fxt1( const struct swrast_texture_image *texImage,
    /* just sample as GLubyte and convert to float here */
    GLubyte rgba[4];
    (void) k;
-   fxt1_decode_1(texImage->Base.Data, texImage->Base.RowStride, i, j, rgba);
+   fxt1_decode_1(texImage->Data, texImage->RowStride, i, j, rgba);
    texel[RCOMP] = UBYTE_TO_FLOAT(rgba[RCOMP]);
    texel[GCOMP] = UBYTE_TO_FLOAT(rgba[GCOMP]);
    texel[BCOMP] = UBYTE_TO_FLOAT(rgba[BCOMP]);
@@ -192,7 +192,7 @@ _mesa_fetch_texel_2d_f_rgb_fxt1( const struct swrast_texture_image *texImage,
    /* just sample as GLubyte and convert to float here */
    GLubyte rgba[4];
    (void) k;
-   fxt1_decode_1(texImage->Base.Data, texImage->Base.RowStride, i, j, rgba);
+   fxt1_decode_1(texImage->Data, texImage->RowStride, i, j, rgba);
    texel[RCOMP] = UBYTE_TO_FLOAT(rgba[RCOMP]);
    texel[GCOMP] = UBYTE_TO_FLOAT(rgba[GCOMP]);
    texel[BCOMP] = UBYTE_TO_FLOAT(rgba[BCOMP]);
diff --git a/mesalib/src/mesa/main/texcompress_rgtc.c b/mesalib/src/mesa/main/texcompress_rgtc.c
index b03cd28b8..3586fc39d 100644
--- a/mesalib/src/mesa/main/texcompress_rgtc.c
+++ b/mesalib/src/mesa/main/texcompress_rgtc.c
@@ -325,7 +325,7 @@ _mesa_fetch_texel_2d_f_red_rgtc1(const struct swrast_texture_image *texImage,
 				 GLint i, GLint j, GLint k, GLfloat *texel)
 {
    GLubyte red;
-   unsigned_fetch_texel_rgtc(texImage->Base.RowStride, (GLubyte *)(texImage->Base.Data),
+   unsigned_fetch_texel_rgtc(texImage->RowStride, texImage->Data,
 		       i, j, &red, 1);
    texel[RCOMP] = UBYTE_TO_FLOAT(red);
    texel[GCOMP] = 0.0;
@@ -338,7 +338,7 @@ _mesa_fetch_texel_2d_f_signed_red_rgtc1(const struct swrast_texture_image *texIm
 					GLint i, GLint j, GLint k, GLfloat *texel)
 {
    GLbyte red;
-   signed_fetch_texel_rgtc(texImage->Base.RowStride, (GLbyte *)(texImage->Base.Data),
+   signed_fetch_texel_rgtc(texImage->RowStride, (GLbyte *)(texImage->Data),
 		       i, j, &red, 1);
    texel[RCOMP] = BYTE_TO_FLOAT_TEX(red);
    texel[GCOMP] = 0.0;
@@ -351,9 +351,9 @@ _mesa_fetch_texel_2d_f_rg_rgtc2(const struct swrast_texture_image *texImage,
 				 GLint i, GLint j, GLint k, GLfloat *texel)
 {
    GLubyte red, green;
-   unsigned_fetch_texel_rgtc(texImage->Base.RowStride, (GLubyte *)(texImage->Base.Data),
+   unsigned_fetch_texel_rgtc(texImage->RowStride, texImage->Data,
 		     i, j, &red, 2);
-   unsigned_fetch_texel_rgtc(texImage->Base.RowStride, (GLubyte *)(texImage->Base.Data) + 8,
+   unsigned_fetch_texel_rgtc(texImage->RowStride, texImage->Data + 8,
 		     i, j, &green, 2);
    texel[RCOMP] = UBYTE_TO_FLOAT(red);
    texel[GCOMP] = UBYTE_TO_FLOAT(green);
@@ -366,9 +366,9 @@ _mesa_fetch_texel_2d_f_signed_rg_rgtc2(const struct swrast_texture_image *texIma
 				       GLint i, GLint j, GLint k, GLfloat *texel)
 {
    GLbyte red, green;
-   signed_fetch_texel_rgtc(texImage->Base.RowStride, (GLbyte *)(texImage->Base.Data),
+   signed_fetch_texel_rgtc(texImage->RowStride, (GLbyte *)(texImage->Data),
 		     i, j, &red, 2);
-   signed_fetch_texel_rgtc(texImage->Base.RowStride, (GLbyte *)(texImage->Base.Data) + 8,
+   signed_fetch_texel_rgtc(texImage->RowStride, (GLbyte *)(texImage->Data) + 8,
 		     i, j, &green, 2);
    texel[RCOMP] = BYTE_TO_FLOAT_TEX(red);
    texel[GCOMP] = BYTE_TO_FLOAT_TEX(green);
@@ -381,7 +381,7 @@ _mesa_fetch_texel_2d_f_l_latc1(const struct swrast_texture_image *texImage,
                                  GLint i, GLint j, GLint k, GLfloat *texel)
 {
    GLubyte red;
-   unsigned_fetch_texel_rgtc(texImage->Base.RowStride, (GLubyte *)(texImage->Base.Data),
+   unsigned_fetch_texel_rgtc(texImage->RowStride, texImage->Data,
                        i, j, &red, 1);
    texel[RCOMP] =
    texel[GCOMP] =
@@ -394,7 +394,7 @@ _mesa_fetch_texel_2d_f_signed_l_latc1(const struct swrast_texture_image *texImag
                                         GLint i, GLint j, GLint k, GLfloat *texel)
 {
    GLbyte red;
-   signed_fetch_texel_rgtc(texImage->Base.RowStride, (GLbyte *)(texImage->Base.Data),
+   signed_fetch_texel_rgtc(texImage->RowStride, (GLbyte *)(texImage->Data),
                        i, j, &red, 1);
    texel[RCOMP] =
    texel[GCOMP] =
@@ -407,9 +407,9 @@ _mesa_fetch_texel_2d_f_la_latc2(const struct swrast_texture_image *texImage,
                                  GLint i, GLint j, GLint k, GLfloat *texel)
 {
    GLubyte red, green;
-   unsigned_fetch_texel_rgtc(texImage->Base.RowStride, (GLubyte *)(texImage->Base.Data),
+   unsigned_fetch_texel_rgtc(texImage->RowStride, texImage->Data,
                      i, j, &red, 2);
-   unsigned_fetch_texel_rgtc(texImage->Base.RowStride, (GLubyte *)(texImage->Base.Data) + 8,
+   unsigned_fetch_texel_rgtc(texImage->RowStride, texImage->Data + 8,
                      i, j, &green, 2);
    texel[RCOMP] =
    texel[GCOMP] =
@@ -422,9 +422,9 @@ _mesa_fetch_texel_2d_f_signed_la_latc2(const struct swrast_texture_image *texIma
                                        GLint i, GLint j, GLint k, GLfloat *texel)
 {
    GLbyte red, green;
-   signed_fetch_texel_rgtc(texImage->Base.RowStride, (GLbyte *)(texImage->Base.Data),
+   signed_fetch_texel_rgtc(texImage->RowStride, (GLbyte *)(texImage->Data),
                      i, j, &red, 2);
-   signed_fetch_texel_rgtc(texImage->Base.RowStride, (GLbyte *)(texImage->Base.Data) + 8,
+   signed_fetch_texel_rgtc(texImage->RowStride, (GLbyte *)(texImage->Data) + 8,
                      i, j, &green, 2);
    texel[RCOMP] =
    texel[GCOMP] =
diff --git a/mesalib/src/mesa/main/texcompress_s3tc.c b/mesalib/src/mesa/main/texcompress_s3tc.c
index 29386a37d..11c7db4ef 100644
--- a/mesalib/src/mesa/main/texcompress_s3tc.c
+++ b/mesalib/src/mesa/main/texcompress_s3tc.c
@@ -402,8 +402,8 @@ fetch_texel_2d_rgb_dxt1( const struct swrast_texture_image *texImage,
 {
    (void) k;
    if (fetch_ext_rgb_dxt1) {
-      fetch_ext_rgb_dxt1(texImage->Base.RowStride,
-                         (GLubyte *)(texImage)->Base.Data, i, j, texel);
+      fetch_ext_rgb_dxt1(texImage->RowStride,
+                         texImage->Data, i, j, texel);
    }
    else
       _mesa_debug(NULL, "attempted to decode s3tc texture without library available: fetch_texel_2d_rgb_dxt1");
@@ -430,8 +430,8 @@ fetch_texel_2d_rgba_dxt1( const struct swrast_texture_image *texImage,
 {
    (void) k;
    if (fetch_ext_rgba_dxt1) {
-      fetch_ext_rgba_dxt1(texImage->Base.RowStride,
-                          (GLubyte *)(texImage)->Base.Data, i, j, texel);
+      fetch_ext_rgba_dxt1(texImage->RowStride,
+                          texImage->Data, i, j, texel);
    }
    else
       _mesa_debug(NULL, "attempted to decode s3tc texture without library available: fetch_texel_2d_rgba_dxt1\n");
@@ -458,9 +458,8 @@ fetch_texel_2d_rgba_dxt3( const struct swrast_texture_image *texImage,
 {
    (void) k;
    if (fetch_ext_rgba_dxt3) {
-      fetch_ext_rgba_dxt3(texImage->Base.RowStride,
-                          (GLubyte *)(texImage)->Base.Data,
-                          i, j, texel);
+      fetch_ext_rgba_dxt3(texImage->RowStride,
+                          texImage->Data, i, j, texel);
    }
    else
       _mesa_debug(NULL, "attempted to decode s3tc texture without library available: fetch_texel_2d_rgba_dxt3\n");
@@ -487,9 +486,8 @@ fetch_texel_2d_rgba_dxt5( const struct swrast_texture_image *texImage,
 {
    (void) k;
    if (fetch_ext_rgba_dxt5) {
-      fetch_ext_rgba_dxt5(texImage->Base.RowStride,
-                          (GLubyte *)(texImage)->Base.Data,
-                          i, j, texel);
+      fetch_ext_rgba_dxt5(texImage->RowStride,
+                          texImage->Data, i, j, texel);
    }
    else
       _mesa_debug(NULL, "attempted to decode s3tc texture without library available: fetch_texel_2d_rgba_dxt5\n");
diff --git a/mesalib/src/mesa/main/texformat.c b/mesalib/src/mesa/main/texformat.c
index 7f262d6d8..aebe38ee0 100644
--- a/mesalib/src/mesa/main/texformat.c
+++ b/mesalib/src/mesa/main/texformat.c
@@ -34,6 +34,7 @@
 
 
 #include "context.h"
+#include "enums.h"
 #include "mfeatures.h"
 #include "mtypes.h"
 #include "texcompress.h"
@@ -897,7 +898,8 @@ _mesa_choose_tex_format( struct gl_context *ctx, GLint internalFormat,
       }
    }
 
-   _mesa_problem(ctx, "unexpected format in _mesa_choose_tex_format()");
+   _mesa_problem(ctx, "unexpected format %s in _mesa_choose_tex_format()",
+                 _mesa_lookup_enum_by_nr(internalFormat));
    return MESA_FORMAT_NONE;
 }
 
diff --git a/mesalib/src/mesa/main/teximage.c b/mesalib/src/mesa/main/teximage.c
index 2d06f84bf..acf7187fd 100644
--- a/mesalib/src/mesa/main/teximage.c
+++ b/mesalib/src/mesa/main/teximage.c
@@ -58,27 +58,6 @@
 
 
 /**
- * We allocate texture memory on 512-byte boundaries so we can use MMX/SSE
- * elsewhere.
- */
-void *
-_mesa_alloc_texmemory(GLsizei bytes)
-{
-   return _mesa_align_malloc(bytes, 512);
-}
-
-
-/**
- * Free texture memory allocated with _mesa_alloc_texmemory()
- */
-void
-_mesa_free_texmemory(void *m)
-{
-   _mesa_align_free(m);
-}
-
-
-/**
  * Return the simple base format for a given internal texture format.
  * For example, given GL_LUMINANCE12_ALPHA4, return GL_LUMINANCE_ALPHA.
  *
@@ -599,29 +578,6 @@ _mesa_new_texture_image( struct gl_context *ctx )
 
 
 /**
- * Free texture image data.
- * This function is a fallback called via ctx->Driver.FreeTextureImageBuffer().
- *
- * \param texImage texture image.
- *
- * Free the texture image data if it's not marked as client data.
- */
-void
-_mesa_free_texture_image_data(struct gl_context *ctx,
-                              struct gl_texture_image *texImage)
-{
-   (void) ctx;
-
-   if (texImage->Data) {
-      /* free the old texture data */
-      _mesa_free_texmemory(texImage->Data);
-   }
-
-   texImage->Data = NULL;
-}
-
-
-/**
  * Free a gl_texture_image and associated data.
  * This function is a fallback called via ctx->Driver.DeleteTextureImage().
  *
@@ -638,11 +594,6 @@ _mesa_delete_texture_image(struct gl_context *ctx,
     */
    ASSERT(ctx->Driver.FreeTextureImageBuffer);
    ctx->Driver.FreeTextureImageBuffer( ctx, texImage );
-
-   ASSERT(texImage->Data == NULL);
-   if (texImage->ImageOffsets)
-      free(texImage->ImageOffsets);
-   free(texImage);
 }
 
 
@@ -1084,18 +1035,12 @@ clear_teximage_fields(struct gl_texture_image *img)
    img->Width = 0;
    img->Height = 0;
    img->Depth = 0;
-   img->RowStride = 0;
-   if (img->ImageOffsets) {
-      free(img->ImageOffsets);
-      img->ImageOffsets = NULL;
-   }
    img->Width2 = 0;
    img->Height2 = 0;
    img->Depth2 = 0;
    img->WidthLog2 = 0;
    img->HeightLog2 = 0;
    img->DepthLog2 = 0;
-   img->Data = NULL;
    img->TexFormat = MESA_FORMAT_NONE;
 }
 
@@ -1123,8 +1068,6 @@ _mesa_init_teximage_fields(struct gl_context *ctx, GLenum target,
                            GLint border, GLenum internalFormat,
                            gl_format format)
 {
-   GLint i;
-
    ASSERT(img);
    ASSERT(width >= 0);
    ASSERT(height >= 0);
@@ -1161,19 +1104,6 @@ _mesa_init_teximage_fields(struct gl_context *ctx, GLenum target,
 
    img->MaxLog2 = MAX2(img->WidthLog2, img->HeightLog2);
 
-   /* RowStride and ImageOffsets[] describe how to address texels in 'Data' */
-   img->RowStride = width;
-   /* Allocate the ImageOffsets array and initialize to typical values.
-    * We allocate the array for 1D/2D textures too in order to avoid special-
-    * case code in the texstore routines.
-    */
-   if (img->ImageOffsets)
-      free(img->ImageOffsets);
-   img->ImageOffsets = (GLuint *) malloc(depth * sizeof(GLuint));
-   for (i = 0; i < depth; i++) {
-      img->ImageOffsets[i] = i * width * height;
-   }
-
    img->TexFormat = format;
 }
 
@@ -2316,6 +2246,45 @@ _mesa_choose_texture_format(struct gl_context *ctx,
    return f;
 }
 
+/**
+ * Adjust pixel unpack params and image dimensions to strip off the
+ * texture border.
+ *
+ * Gallium and intel don't support texture borders.  They've seldem been used
+ * and seldom been implemented correctly anyway.
+ *
+ * \param unpackNew returns the new pixel unpack parameters
+ */
+static void
+strip_texture_border(GLint *border,
+                     GLint *width, GLint *height, GLint *depth,
+                     const struct gl_pixelstore_attrib *unpack,
+                     struct gl_pixelstore_attrib *unpackNew)
+{
+   assert(*border > 0);  /* sanity check */
+
+   *unpackNew = *unpack;
+
+   if (unpackNew->RowLength == 0)
+      unpackNew->RowLength = *width;
+
+   if (depth && unpackNew->ImageHeight == 0)
+      unpackNew->ImageHeight = *height;
+
+   unpackNew->SkipPixels += *border;
+   if (height)
+      unpackNew->SkipRows += *border;
+   if (depth)
+      unpackNew->SkipImages += *border;
+
+   assert(*width >= 3);
+   *width = *width - 2 * *border;
+   if (height && *height >= 3)
+      *height = *height - 2 * *border;
+   if (depth && *depth >= 3)
+      *depth = *depth - 2 * *border;
+   *border = 0;
+}
 
 /**
  * Common code to implement all the glTexImage1D/2D/3D functions.
@@ -2328,6 +2297,8 @@ teximage(struct gl_context *ctx, GLuint dims,
          const GLvoid *pixels)
 {
    GLboolean error;
+   struct gl_pixelstore_attrib unpack_no_border;
+   const struct gl_pixelstore_attrib *unpack = &ctx->Unpack;
 
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
@@ -2392,6 +2363,16 @@ teximage(struct gl_context *ctx, GLuint dims,
          return;   /* error was recorded */
       }
 
+      /* Allow a hardware driver to just strip out the border, to provide
+       * reliable but slightly incorrect hardware rendering instead of
+       * rarely-tested software fallback rendering.
+       */
+      if (border && ctx->Const.StripTextureBorder) {
+	 strip_texture_border(&border, &width, &height, &depth, unpack,
+			      &unpack_no_border);
+	 unpack = &unpack_no_border;
+      }
+
       if (ctx->NewState & _NEW_PIXEL)
 	 _mesa_update_state(ctx);
 
@@ -2409,7 +2390,6 @@ teximage(struct gl_context *ctx, GLuint dims,
 
             ctx->Driver.FreeTextureImageBuffer(ctx, texImage);
 
-            ASSERT(texImage->Data == NULL);
             texFormat = _mesa_choose_texture_format(ctx, texObj, target, level,
                                                     internalFormat, format,
                                                     type);
@@ -2425,19 +2405,19 @@ teximage(struct gl_context *ctx, GLuint dims,
                case 1:
                   ctx->Driver.TexImage1D(ctx, target, level, internalFormat,
                                          width, border, format,
-                                         type, pixels, &ctx->Unpack, texObj,
+                                         type, pixels, unpack, texObj,
                                          texImage);
                   break;
                case 2:
                   ctx->Driver.TexImage2D(ctx, target, level, internalFormat,
                                          width, height, border, format,
-                                         type, pixels, &ctx->Unpack, texObj,
+                                         type, pixels, unpack, texObj,
                                          texImage);
                   break;
                case 3:
                   ctx->Driver.TexImage3D(ctx, target, level, internalFormat,
                                          width, height, depth, border, format,
-                                         type, pixels, &ctx->Unpack, texObj,
+                                         type, pixels, unpack, texObj,
                                          texImage);
                   break;
                default:
@@ -2548,7 +2528,6 @@ _mesa_EGLImageTargetTexture2DOES (GLenum target, GLeglImageOES image)
    } else {
       ctx->Driver.FreeTextureImageBuffer(ctx, texImage);
 
-      ASSERT(texImage->Data == NULL);
       ctx->Driver.EGLImageTargetTexture2D(ctx, target,
 					  texObj, texImage, image);
 
@@ -2734,6 +2713,16 @@ copyteximage(struct gl_context *ctx, GLuint dims,
 
    texObj = _mesa_get_current_tex_object(ctx, target);
 
+   if (border && ctx->Const.StripTextureBorder) {
+      x += border;
+      width -= border * 2;
+      if (dims == 2) {
+	 y += border;
+	 height -= border * 2;
+      }
+      border = 0;
+   }
+
    _mesa_lock_texture(ctx, texObj);
    {
       texImage = _mesa_get_tex_image(ctx, texObj, target, level);
@@ -3370,7 +3359,6 @@ compressedteximage(struct gl_context *ctx, GLuint dims,
             gl_format texFormat;
 
             ctx->Driver.FreeTextureImageBuffer(ctx, texImage);
-            ASSERT(texImage->Data == NULL);
 
             texFormat = _mesa_choose_texture_format(ctx, texObj, target, level,
                                                     internalFormat, GL_NONE,
diff --git a/mesalib/src/mesa/main/teximage.h b/mesalib/src/mesa/main/teximage.h
index 6ce0fe92c..fd315bea3 100644
--- a/mesalib/src/mesa/main/teximage.h
+++ b/mesalib/src/mesa/main/teximage.h
@@ -36,13 +36,6 @@
 #include "formats.h"
 
 
-extern void *
-_mesa_alloc_texmemory(GLsizei bytes);
-
-extern void
-_mesa_free_texmemory(void *m);
-
-
 /** \name Internal functions */
 /*@{*/
 
@@ -62,10 +55,6 @@ extern void
 _mesa_delete_texture_image( struct gl_context *ctx,
                             struct gl_texture_image *teximage );
 
-extern void
-_mesa_free_texture_image_data( struct gl_context *ctx, 
-			       struct gl_texture_image *texImage );
-
 
 extern void
 _mesa_init_teximage_fields(struct gl_context *ctx, GLenum target,
diff --git a/mesalib/src/mesa/main/texobj.c b/mesalib/src/mesa/main/texobj.c
index 1b90cca9b..4d9942ba8 100644
--- a/mesalib/src/mesa/main/texobj.c
+++ b/mesalib/src/mesa/main/texobj.c
@@ -543,12 +543,13 @@ _mesa_test_texobj_completeness( const struct gl_context *ctx,
                width /= 2;
             }
             if (i >= minLevel && i <= maxLevel) {
-               if (!t->Image[0][i]) {
-                  incomplete(t, "1D Image[0][i] == NULL");
+               const struct gl_texture_image *img = t->Image[0][i];
+               if (!img) {
+                  incomplete(t, "1D Image[%d] is missing", i);
                   return;
                }
-               if (t->Image[0][i]->Width2 != width ) {
-                  incomplete(t, "1D Image[0][i] bad width");
+               if (img->Width2 != width ) {
+                  incomplete(t, "1D Image[%d] bad width %u", i, img->Width2);
                   return;
                }
             }
@@ -570,16 +571,17 @@ _mesa_test_texobj_completeness( const struct gl_context *ctx,
                height /= 2;
             }
             if (i >= minLevel && i <= maxLevel) {
-               if (!t->Image[0][i]) {
-                  incomplete(t, "2D Image[0][i] == NULL");
+               const struct gl_texture_image *img = t->Image[0][i];
+               if (!img) {
+                  incomplete(t, "2D Image[%d of %d] is missing", i, maxLevel);
                   return;
                }
-               if (t->Image[0][i]->Width2 != width) {
-                  incomplete(t, "2D Image[0][i] bad width");
+               if (img->Width2 != width) {
+                  incomplete(t, "2D Image[%d] bad width %u", i, img->Width2);
                   return;
                }
-               if (t->Image[0][i]->Height2 != height) {
-                  incomplete(t, "2D Image[0][i] bad height");
+               if (img->Height2 != height) {
+                  incomplete(t, "2D Image[i] bad height %u", i, img->Height2);
                   return;
                }
                if (width==1 && height==1) {
@@ -604,24 +606,25 @@ _mesa_test_texobj_completeness( const struct gl_context *ctx,
                depth /= 2;
             }
             if (i >= minLevel && i <= maxLevel) {
-               if (!t->Image[0][i]) {
-                  incomplete(t, "3D Image[0][i] == NULL");
+               const struct gl_texture_image *img = t->Image[0][i];
+               if (!img) {
+                  incomplete(t, "3D Image[%d] is missing", i);
                   return;
                }
-               if (t->Image[0][i]->_BaseFormat == GL_DEPTH_COMPONENT) {
+               if (img->_BaseFormat == GL_DEPTH_COMPONENT) {
                   incomplete(t, "GL_DEPTH_COMPONENT only works with 1/2D tex");
                   return;
                }
-               if (t->Image[0][i]->Width2 != width) {
-                  incomplete(t, "3D Image[0][i] bad width");
+               if (img->Width2 != width) {
+                  incomplete(t, "3D Image[%d] bad width %u", i, img->Width2);
                   return;
                }
-               if (t->Image[0][i]->Height2 != height) {
-                  incomplete(t, "3D Image[0][i] bad height");
+               if (img->Height2 != height) {
+                  incomplete(t, "3D Image[%d] bad height %u", i, img->Height2);
                   return;
                }
-               if (t->Image[0][i]->Depth2 != depth) {
-                  incomplete(t, "3D Image[0][i] bad depth");
+               if (img->Depth2 != depth) {
+                  incomplete(t, "3D Image[%d] bad depth %u", i, img->Depth2);
                   return;
                }
             }
diff --git a/mesalib/src/mesa/main/texstore.c b/mesalib/src/mesa/main/texstore.c
index cc9fbc020..05c1964d6 100644
--- a/mesalib/src/mesa/main/texstore.c
+++ b/mesalib/src/mesa/main/texstore.c
@@ -4499,6 +4499,9 @@ _mesa_store_teximage1d(struct gl_context *ctx, GLenum target, GLint level,
 
    (void) border;
 
+   if (width == 0)
+      return;
+
    /* allocate storage for texture data */
    if (!ctx->Driver.AllocTextureImageBuffer(ctx, texImage, texImage->TexFormat,
                                             width, 1, 1)) {
@@ -4560,6 +4563,9 @@ _mesa_store_teximage2d(struct gl_context *ctx, GLenum target, GLint level,
 
    (void) border;
 
+   if (width == 0 || height == 0)
+      return;
+
    /* allocate storage for texture data */
    if (!ctx->Driver.AllocTextureImageBuffer(ctx, texImage, texImage->TexFormat,
                                             width, height, 1)) {
@@ -4651,6 +4657,9 @@ _mesa_store_teximage3d(struct gl_context *ctx, GLenum target, GLint level,
 
    (void) border;
 
+   if (width == 0 || height == 0 || depth == 0)
+      return;
+
    /* allocate storage for texture data */
    if (!ctx->Driver.AllocTextureImageBuffer(ctx, texImage, texImage->TexFormat,
                                             width, height, depth)) {
@@ -4906,7 +4915,6 @@ _mesa_store_compressed_teximage2d(struct gl_context *ctx,
    ASSERT(texImage->Width > 0);
    ASSERT(texImage->Height > 0);
    ASSERT(texImage->Depth == 1);
-   ASSERT(texImage->Data == NULL); /* was freed in glCompressedTexImage2DARB */
 
    /* allocate storage for texture data */
    if (!ctx->Driver.AllocTextureImageBuffer(ctx, texImage, texImage->TexFormat,
diff --git a/mesalib/src/mesa/program/ir_to_mesa.cpp b/mesalib/src/mesa/program/ir_to_mesa.cpp
index fecab50f7..bdbb6b938 100644
--- a/mesalib/src/mesa/program/ir_to_mesa.cpp
+++ b/mesalib/src/mesa/program/ir_to_mesa.cpp
@@ -40,6 +40,7 @@
 #include "../glsl/program.h"
 #include "ir_optimization.h"
 #include "ast.h"
+#include "linker.h"
 
 #include "main/mtypes.h"
 #include "main/shaderobj.h"
@@ -2587,13 +2588,35 @@ check_resources(const struct gl_context *ctx,
    }
 }
 
+class add_uniform_to_shader : public uniform_field_visitor {
+public:
+   add_uniform_to_shader(struct gl_shader_program *shader_program,
+			 struct gl_program_parameter_list *params)
+      : shader_program(shader_program), params(params), next_sampler(0)
+   {
+      /* empty */
+   }
 
-static int
-add_uniform_to_shader(ir_variable *var,
-		      struct gl_program_parameter_list *params,
-		      unsigned int &next_sampler)
+   int process(ir_variable *var)
+   {
+      this->idx = -1;
+      this->uniform_field_visitor::process(var);
+
+      return this->idx;
+   }
+
+private:
+   virtual void visit_field(const glsl_type *type, const char *name);
+
+   struct gl_shader_program *shader_program;
+   struct gl_program_parameter_list *params;
+   int next_sampler;
+   int idx;
+};
+
+void
+add_uniform_to_shader::visit_field(const glsl_type *type, const char *name)
 {
-   const glsl_type *type = var->type;
    unsigned int size;
 
    if (type->is_vector() || type->is_scalar()) {
@@ -2610,10 +2633,9 @@ add_uniform_to_shader(ir_variable *var,
       file = PROGRAM_UNIFORM;
    }
 
-   int index = _mesa_lookup_parameter_index(params, -1, var->name);
+   int index = _mesa_lookup_parameter_index(params, -1, name);
    if (index < 0) {
-      index = _mesa_add_parameter(params, file,
-				  var->name, size, type->gl_type,
+      index = _mesa_add_parameter(params, file, name, size, type->gl_type,
 				  NULL, NULL, 0x0);
 
       /* Sampler uniform values are stored in prog->SamplerUnits,
@@ -2622,11 +2644,15 @@ add_uniform_to_shader(ir_variable *var,
        */
       if (file == PROGRAM_SAMPLER) {
 	 for (unsigned int j = 0; j < size / 4; j++)
-	    params->ParameterValues[index + j][0].f = next_sampler++;
+	    params->ParameterValues[index + j][0].f = this->next_sampler++;
       }
    }
 
-   return index;
+   /* The first part of the uniform that's processed determines the base
+    * location of the whole uniform (for structures).
+    */
+   if (this->idx < 0)
+      this->idx = index;
 }
 
 /**
@@ -2644,7 +2670,7 @@ _mesa_generate_parameters_list_for_uniforms(struct gl_shader_program
 					    struct gl_program_parameter_list
 					    *params)
 {
-   unsigned int next_sampler = 0;
+   add_uniform_to_shader add(shader_program, params);
 
    foreach_list(node, sh->ir) {
       ir_variable *var = ((ir_instruction *) node)->as_variable();
@@ -2653,7 +2679,7 @@ _mesa_generate_parameters_list_for_uniforms(struct gl_shader_program
 	  || (strncmp(var->name, "gl_", 3) == 0))
 	 continue;
 
-      int loc = add_uniform_to_shader(var, params, next_sampler);
+      int loc = add.process(var);
 
       /* The location chosen in the Parameters list here (returned from
        * _mesa_add_parameter) has to match what the linker chose.
@@ -3212,7 +3238,9 @@ _mesa_ir_link_shader(struct gl_context *ctx, struct gl_shader_program *prog)
 
 	 progress = do_lower_jumps(ir, true, true, options->EmitNoMainReturn, options->EmitNoCont, options->EmitNoLoops) || progress;
 
-	 progress = do_common_optimization(ir, true, options->MaxUnrollIterations) || progress;
+	 progress = do_common_optimization(ir, true, true,
+					   options->MaxUnrollIterations)
+	   || progress;
 
 	 progress = lower_quadop_vector(ir, true) || progress;
 
@@ -3321,7 +3349,7 @@ _mesa_glsl_compile_shader(struct gl_context *ctx, struct gl_shader *shader)
       /* Do some optimization at compile time to reduce shader IR size
        * and reduce later work if the same shader is linked multiple times
        */
-      while (do_common_optimization(shader->ir, false, 32))
+      while (do_common_optimization(shader->ir, false, false, 32))
 	 ;
 
       validate_ir_tree(shader->ir);
diff --git a/mesalib/src/mesa/state_tracker/st_cb_texture.c b/mesalib/src/mesa/state_tracker/st_cb_texture.c
index 169e235ac..f82346bc6 100644
--- a/mesalib/src/mesa/state_tracker/st_cb_texture.c
+++ b/mesalib/src/mesa/state_tracker/st_cb_texture.c
@@ -543,45 +543,6 @@ st_AllocTextureImageBuffer(struct gl_context *ctx,
    }
 }
 
-
-/**
- * Adjust pixel unpack params and image dimensions to strip off the
- * texture border.
- * Gallium doesn't support texture borders.  They've seldem been used
- * and seldom been implemented correctly anyway.
- * \param unpackNew  returns the new pixel unpack parameters
- */
-static void
-strip_texture_border(GLint border,
-                     GLint *width, GLint *height, GLint *depth,
-                     const struct gl_pixelstore_attrib *unpack,
-                     struct gl_pixelstore_attrib *unpackNew)
-{
-   assert(border > 0);  /* sanity check */
-
-   *unpackNew = *unpack;
-
-   if (unpackNew->RowLength == 0)
-      unpackNew->RowLength = *width;
-
-   if (depth && unpackNew->ImageHeight == 0)
-      unpackNew->ImageHeight = *height;
-
-   unpackNew->SkipPixels += border;
-   if (height)
-      unpackNew->SkipRows += border;
-   if (depth)
-      unpackNew->SkipImages += border;
-
-   assert(*width >= 3);
-   *width = *width - 2 * border;
-   if (height && *height >= 3)
-      *height = *height - 2 * border;
-   if (depth && *depth >= 3)
-      *depth = *depth - 2 * border;
-}
-
-
 /**
  * Do glTexImage1/2/3D().
  */
@@ -602,7 +563,6 @@ st_TexImage(struct gl_context * ctx,
    struct st_texture_object *stObj = st_texture_object(texObj);
    struct st_texture_image *stImage = st_texture_image(texImage);
    GLuint dstRowStride = 0;
-   struct gl_pixelstore_attrib unpackNB;
    enum pipe_transfer_usage transfer_usage = 0;
    GLubyte *dstMap;
 
@@ -627,21 +587,9 @@ st_TexImage(struct gl_context * ctx,
       stObj->surface_based = GL_FALSE;
    }
 
-   /* gallium does not support texture borders, strip it off */
-   if (border) {
-      strip_texture_border(border, &width, &height, &depth, unpack, &unpackNB);
-      unpack = &unpackNB;
-      texImage->Width = width;
-      texImage->Height = height;
-      texImage->Depth = depth;
-      texImage->Border = 0;
-      border = 0;
-   }
-   else {
-      assert(texImage->Width == width);
-      assert(texImage->Height == height);
-      assert(texImage->Depth == depth);
-   }
+   assert(texImage->Width == width);
+   assert(texImage->Height == height);
+   assert(texImage->Depth == depth);
 
    stImage->base.Face = _mesa_tex_target_to_face(target);
    stImage->base.Level = level;
diff --git a/mesalib/src/mesa/state_tracker/st_context.c b/mesalib/src/mesa/state_tracker/st_context.c
index 6d4bc544d..a1817720d 100644
--- a/mesalib/src/mesa/state_tracker/st_context.c
+++ b/mesalib/src/mesa/state_tracker/st_context.c
@@ -144,6 +144,10 @@ st_create_context_priv( struct gl_context *ctx, struct pipe_context *pipe )
    /* we want all vertex data to be placed in buffer objects */
    vbo_use_buffer_objects(ctx);
 
+
+   /* make sure that no VBOs are left mapped when we're drawing. */
+   vbo_always_unmap_buffers(ctx);
+
    /* Need these flags:
     */
    st->ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
diff --git a/mesalib/src/mesa/state_tracker/st_context.h b/mesalib/src/mesa/state_tracker/st_context.h
index 0a3220221..c60780989 100644
--- a/mesalib/src/mesa/state_tracker/st_context.h
+++ b/mesalib/src/mesa/state_tracker/st_context.h
@@ -75,6 +75,7 @@ struct st_context
    struct draw_stage *feedback_stage;  /**< For GL_FEEDBACK rendermode */
    struct draw_stage *selection_stage;  /**< For GL_SELECT rendermode */
    struct draw_stage *rastpos_stage;  /**< For glRasterPos */
+   GLboolean sw_primitive_restart;
 
 
    /* On old libGL's for linux we need to invalidate the drawables
diff --git a/mesalib/src/mesa/state_tracker/st_draw.c b/mesalib/src/mesa/state_tracker/st_draw.c
index 574802084..ff3008a5f 100644
--- a/mesalib/src/mesa/state_tracker/st_draw.c
+++ b/mesalib/src/mesa/state_tracker/st_draw.c
@@ -353,9 +353,26 @@ setup_interleaved_attribs(struct gl_context *ctx,
    struct pipe_context *pipe = st->pipe;
    GLuint attr;
    const GLubyte *low_addr = NULL;
-
-   /* Find the lowest address of the arrays we're drawing */
+   GLboolean usingVBO;      /* all arrays in a VBO? */
+   struct gl_buffer_object *bufobj;
+   GLuint user_buffer_size = 0;
+   GLuint vertex_size = 0;  /* bytes per vertex, in bytes */
+   GLsizei stride;
+
+   /* Find the lowest address of the arrays we're drawing,
+    * Init bufobj and stride.
+    */
    if (vpv->num_inputs) {
+      const GLuint mesaAttr0 = vp->index_to_input[0];
+      const struct gl_client_array *array = arrays[mesaAttr0];
+
+      /* Since we're doing interleaved arrays, we know there'll be at most
+       * one buffer object and the stride will be the same for all arrays.
+       * Grab them now.
+       */
+      bufobj = array->BufferObj;
+      stride = array->StrideB;
+
       low_addr = arrays[vp->index_to_input[0]]->Ptr;
 
       for (attr = 1; attr < vpv->num_inputs; attr++) {
@@ -363,44 +380,24 @@ setup_interleaved_attribs(struct gl_context *ctx,
          low_addr = MIN2(low_addr, start);
       }
    }
+   else {
+      /* not sure we'll ever have zero inputs, but play it safe */
+      bufobj = NULL;
+      stride = 0;
+      low_addr = 0;
+   }
+
+   /* are the arrays in user space? */
+   usingVBO = bufobj && _mesa_is_bufferobj(bufobj);
 
    for (attr = 0; attr < vpv->num_inputs; attr++) {
       const GLuint mesaAttr = vp->index_to_input[attr];
       const struct gl_client_array *array = arrays[mesaAttr];
-      struct gl_buffer_object *bufobj = array->BufferObj;
-      struct st_buffer_object *stobj = st_buffer_object(bufobj);
       unsigned src_offset = (unsigned) (array->Ptr - low_addr);
       GLuint element_size = array->_ElementSize;
-      GLsizei stride = array->StrideB;
 
       assert(element_size == array->Size * _mesa_sizeof_type(array->Type));
 
-      if (attr == 0) {
-         if (bufobj && _mesa_is_bufferobj(bufobj)) {
-            vbuffer->buffer = NULL;
-            pipe_resource_reference(&vbuffer->buffer, stobj->buffer);
-            vbuffer->buffer_offset = pointer_to_offset(low_addr);
-         }
-         else {
-            uint divisor = array->InstanceDivisor;
-            uint last_index = divisor ? num_instances / divisor : max_index;
-            uint bytes = src_offset + stride * last_index + element_size;
-
-            vbuffer->buffer = pipe_user_buffer_create(pipe->screen,
-                                                      (void*) low_addr,
-                                                      bytes,
-                                                      PIPE_BIND_VERTEX_BUFFER);
-            vbuffer->buffer_offset = 0;
-
-            /* Track user vertex buffers. */
-            pipe_resource_reference(&st->user_attrib[0].buffer, vbuffer->buffer);
-            st->user_attrib[0].element_size = element_size;
-            st->user_attrib[0].stride = stride;
-            st->num_user_attribs = 1;
-         }
-         vbuffer->stride = stride; /* in bytes */
-      }
-
       velements[attr].src_offset = src_offset;
       velements[attr].instance_divisor = array->InstanceDivisor;
       velements[attr].vertex_buffer_index = 0;
@@ -409,6 +406,54 @@ setup_interleaved_attribs(struct gl_context *ctx,
                                                          array->Format,
                                                          array->Normalized);
       assert(velements[attr].src_format);
+
+      if (!usingVBO) {
+         /* how many bytes referenced by this attribute array? */
+         uint divisor = array->InstanceDivisor;
+         uint last_index = divisor ? num_instances / divisor : max_index;
+         uint bytes = src_offset + stride * last_index + element_size;
+
+         user_buffer_size = MAX2(user_buffer_size, bytes);
+
+         /* update vertex size */
+         vertex_size = MAX2(vertex_size, src_offset + element_size);
+      }
+   }
+
+   /*
+    * Return the vbuffer info and setup user-space attrib info, if needed.
+    */
+   if (vpv->num_inputs == 0) {
+      /* just defensive coding here */
+      vbuffer->buffer = NULL;
+      vbuffer->buffer_offset = 0;
+      vbuffer->stride = 0;
+      st->num_user_attribs = 0;
+   }
+   else if (usingVBO) {
+      /* all interleaved arrays in a VBO */
+      struct st_buffer_object *stobj = st_buffer_object(bufobj);
+
+      vbuffer->buffer = NULL;
+      pipe_resource_reference(&vbuffer->buffer, stobj->buffer);
+      vbuffer->buffer_offset = pointer_to_offset(low_addr);
+      vbuffer->stride = stride;
+      st->num_user_attribs = 0;
+   }
+   else {
+      /* all interleaved arrays in user memory */
+      vbuffer->buffer = pipe_user_buffer_create(pipe->screen,
+                                                (void*) low_addr,
+                                                user_buffer_size,
+                                                PIPE_BIND_VERTEX_BUFFER);
+      vbuffer->buffer_offset = 0;
+      vbuffer->stride = stride;
+
+      /* Track user vertex buffers. */
+      pipe_resource_reference(&st->user_attrib[0].buffer, vbuffer->buffer);
+      st->user_attrib[0].element_size = vertex_size;
+      st->user_attrib[0].stride = stride;
+      st->num_user_attribs = 1;
    }
 }
 
@@ -582,6 +627,127 @@ check_uniforms(struct gl_context *ctx)
    }
 }
 
+/** Helper code for primitive restart fallback */
+#define DO_DRAW(pipe, cur_start, cur_count) \
+   do { \
+      info.start = cur_start; \
+      info.count = cur_count; \
+      if (u_trim_pipe_prim(info.mode, &info.count)) { \
+         if (transfer) \
+            pipe_buffer_unmap(pipe, transfer); \
+         pipe->draw_vbo(pipe, &info); \
+         if (transfer) { \
+            ptr = pipe_buffer_map(pipe, ibuffer->buffer, PIPE_TRANSFER_READ, &transfer); \
+            assert(ptr != NULL); \
+            ptr = ADD_POINTERS(ptr, ibuffer->offset); \
+         } \
+      } \
+   } while(0)
+      
+/** More helper code for primitive restart fallback */
+#define PRIM_RESTART_LOOP(elements) \
+   do { \
+      for (i = start; i < end; i++) { \
+         if (elements[i] == info.restart_index) { \
+            if (cur_count > 0) { \
+               /* draw elts up to prev pos */ \
+               DO_DRAW(pipe, cur_start, cur_count); \
+            } \
+            /* begin new prim at next elt */ \
+            cur_start = i + 1; \
+            cur_count = 0; \
+         } \
+         else { \
+            cur_count++; \
+         } \
+      } \
+      if (cur_count > 0) { \
+         DO_DRAW(pipe, cur_start, cur_count); \
+      } \
+   } while (0)
+
+static void
+handle_fallback_primitive_restart(struct pipe_context *pipe,
+                                  const struct _mesa_index_buffer *ib,
+                                  struct pipe_index_buffer *ibuffer,
+                                  struct pipe_draw_info *orig_info)
+{
+   const unsigned start = orig_info->start;
+   const unsigned count = orig_info->count;
+   const unsigned end = start + count;
+   struct pipe_draw_info info = *orig_info;
+   struct pipe_transfer *transfer = NULL;
+   unsigned instance, i, cur_start, cur_count;
+   const void *ptr;
+
+   info.primitive_restart = FALSE;
+
+   if (!info.indexed) {
+      /* Splitting the draw arrays call is handled by the VBO module */
+      if (u_trim_pipe_prim(info.mode, &info.count))
+         pipe->draw_vbo(pipe, &info);
+
+      return;
+   }
+
+   /* info.indexed == TRUE */
+   assert(ibuffer);
+   assert(ibuffer->buffer);
+
+   if (ib) {
+      struct gl_buffer_object *bufobj = ib->obj;
+      if (bufobj && bufobj->Name) {
+         ptr = NULL;
+      }
+      else {
+         ptr = ib->ptr;
+      }
+   } else {
+      ptr = NULL;
+   }
+
+   if (!ptr)
+      ptr = pipe_buffer_map(pipe, ibuffer->buffer, PIPE_TRANSFER_READ, &transfer);
+
+   if (!ptr)
+     return;
+   ptr = ADD_POINTERS(ptr, ibuffer->offset);
+
+   /* Need to loop over instances as well to preserve draw order */
+   for (instance = 0; instance < orig_info->instance_count; instance++) {
+      info.start_instance = instance + orig_info->start_instance;
+      info.instance_count = 1;
+      cur_start = start;
+      cur_count = 0;
+
+      switch (ibuffer->index_size) {
+      case 1:
+         {
+            const ubyte *elt_ub = (const ubyte *)ptr; 
+            PRIM_RESTART_LOOP(elt_ub);
+         }
+         break;
+      case 2:
+         {
+            const ushort *elt_us = (const ushort *)ptr;
+            PRIM_RESTART_LOOP(elt_us);
+         }
+         break;
+      case 4:
+         {
+            const uint *elt_ui = (const uint *)ptr;
+            PRIM_RESTART_LOOP(elt_ui);
+         }
+         break;
+      default:
+         assert(0 && "bad index_size in handle_fallback_primitive_restart()");
+      }
+   }
+
+   if (transfer)
+      pipe_buffer_unmap(pipe, transfer);
+}
+
 
 /**
  * Translate OpenGL primtive type (GL_POINTS, GL_TRIANGLE_STRIP, etc) to
@@ -794,7 +960,22 @@ st_draw_vbo(struct gl_context *ctx,
          info.max_index = info.start + info.count - 1;
       }
 
-      if (u_trim_pipe_prim(info.mode, &info.count))
+      if (info.primitive_restart) {
+         /*
+          * Handle primitive restart for drivers that doesn't support it.
+          *
+          * The VBO module handles restart inside of draw_arrays for us,
+          * but we should still remove the primitive_restart flag on the
+          * info struct, the fallback function does this for us. Just
+          * remove the flag for all drivers in this case as well.
+          */
+         if (st->sw_primitive_restart || !info.indexed)
+            handle_fallback_primitive_restart(pipe, ib, &ibuffer, &info);
+         else
+            /* don't trim, restarts might be inside index list */
+            pipe->draw_vbo(pipe, &info);
+      }
+      else if (u_trim_pipe_prim(info.mode, &info.count))
          pipe->draw_vbo(pipe, &info);
    }
 
diff --git a/mesalib/src/mesa/state_tracker/st_extensions.c b/mesalib/src/mesa/state_tracker/st_extensions.c
index a1f029089..6b9ff6b72 100644
--- a/mesalib/src/mesa/state_tracker/st_extensions.c
+++ b/mesalib/src/mesa/state_tracker/st_extensions.c
@@ -222,6 +222,8 @@ void st_init_limits(struct st_context *st)
       _mesa_override_glsl_version(st->ctx);
       c->UniformBooleanTrue = ~0;
    }
+
+   c->StripTextureBorder = GL_TRUE;
 }
 
 
@@ -555,8 +557,9 @@ void st_init_extensions(struct st_context *st)
 #endif
    }
 
-   if (screen->get_param(screen, PIPE_CAP_PRIMITIVE_RESTART)) {
-      ctx->Extensions.NV_primitive_restart = GL_TRUE;
+   ctx->Extensions.NV_primitive_restart = GL_TRUE;
+   if (!screen->get_param(screen, PIPE_CAP_PRIMITIVE_RESTART)) {
+      st->sw_primitive_restart = GL_TRUE;
    }
 
    if (screen->get_param(screen, PIPE_CAP_DEPTH_CLAMP)) {
diff --git a/mesalib/src/mesa/state_tracker/st_glsl_to_tgsi.cpp b/mesalib/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
index 18e8a1db4..145bd7dcd 100644
--- a/mesalib/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
+++ b/mesalib/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
@@ -4322,37 +4322,15 @@ compile_tgsi_instruction(struct st_translate *t,
 }
 
 /**
- * Emit the TGSI instructions to adjust the WPOS pixel center convention
- * Basically, add (adjX, adjY) to the fragment position.
- */
-static void
-emit_adjusted_wpos(struct st_translate *t,
-                   const struct gl_program *program,
-                   float adjX, float adjY)
-{
-   struct ureg_program *ureg = t->ureg;
-   struct ureg_dst wpos_temp = ureg_DECL_temporary(ureg);
-   struct ureg_src wpos_input = t->inputs[t->inputMapping[FRAG_ATTRIB_WPOS]];
-
-   /* Note that we bias X and Y and pass Z and W through unchanged.
-    * The shader might also use gl_FragCoord.w and .z.
-    */
-   ureg_ADD(ureg, wpos_temp, wpos_input,
-            ureg_imm4f(ureg, adjX, adjY, 0.0f, 0.0f));
-
-   t->inputs[t->inputMapping[FRAG_ATTRIB_WPOS]] = ureg_src(wpos_temp);
-}
-
-
-/**
- * Emit the TGSI instructions for inverting the WPOS y coordinate.
+ * Emit the TGSI instructions for inverting and adjusting WPOS.
  * This code is unavoidable because it also depends on whether
  * a FBO is bound (STATE_FB_WPOS_Y_TRANSFORM).
  */
 static void
-emit_wpos_inversion(struct st_translate *t,
-                    const struct gl_program *program,
-                    bool invert)
+emit_wpos_adjustment( struct st_translate *t,
+                      const struct gl_program *program,
+                      boolean invert,
+                      GLfloat adjX, GLfloat adjY[2])
 {
    struct ureg_program *ureg = t->ureg;
 
@@ -4371,35 +4349,55 @@ emit_wpos_inversion(struct st_translate *t,
    unsigned wposTransConst = _mesa_add_state_reference(program->Parameters,
                                                        wposTransformState);
 
-   struct ureg_src wpostrans = ureg_DECL_constant(ureg, wposTransConst);
-   struct ureg_dst wpos_temp;
+   struct ureg_src wpostrans = ureg_DECL_constant( ureg, wposTransConst );
+   struct ureg_dst wpos_temp = ureg_DECL_temporary( ureg );
    struct ureg_src wpos_input = t->inputs[t->inputMapping[FRAG_ATTRIB_WPOS]];
 
-   /* MOV wpos_temp, input[wpos]
-    */
-   if (wpos_input.File == TGSI_FILE_TEMPORARY)
-      wpos_temp = ureg_dst(wpos_input);
-   else {
-      wpos_temp = ureg_DECL_temporary(ureg);
-      ureg_MOV(ureg, wpos_temp, wpos_input);
+   /* First, apply the coordinate shift: */
+   if (adjX || adjY[0] || adjY[1]) {
+      if (adjY[0] != adjY[1]) {
+         /* Adjust the y coordinate by adjY[1] or adjY[0] respectively
+          * depending on whether inversion is actually going to be applied
+          * or not, which is determined by testing against the inversion
+          * state variable used below, which will be either +1 or -1.
+          */
+         struct ureg_dst adj_temp = ureg_DECL_temporary(ureg);
+
+         ureg_CMP(ureg, adj_temp,
+                  ureg_scalar(wpostrans, invert ? 2 : 0),
+                  ureg_imm4f(ureg, adjX, adjY[0], 0.0f, 0.0f),
+                  ureg_imm4f(ureg, adjX, adjY[1], 0.0f, 0.0f));
+         ureg_ADD(ureg, wpos_temp, wpos_input, ureg_src(adj_temp));
+      } else {
+         ureg_ADD(ureg, wpos_temp, wpos_input,
+                  ureg_imm4f(ureg, adjX, adjY[0], 0.0f, 0.0f));
+      }
+      wpos_input = ureg_src(wpos_temp);
+   } else {
+      /* MOV wpos_temp, input[wpos]
+       */
+      ureg_MOV( ureg, wpos_temp, wpos_input );
    }
 
+   /* Now the conditional y flip: STATE_FB_WPOS_Y_TRANSFORM.xy/zw will be
+    * inversion/identity, or the other way around if we're drawing to an FBO.
+    */
    if (invert) {
       /* MAD wpos_temp.y, wpos_input, wpostrans.xxxx, wpostrans.yyyy
        */
-      ureg_MAD(ureg,
-               ureg_writemask(wpos_temp, TGSI_WRITEMASK_Y),
-               wpos_input,
-               ureg_scalar(wpostrans, 0),
-               ureg_scalar(wpostrans, 1));
+      ureg_MAD( ureg,
+                ureg_writemask(wpos_temp, TGSI_WRITEMASK_Y ),
+                wpos_input,
+                ureg_scalar(wpostrans, 0),
+                ureg_scalar(wpostrans, 1));
    } else {
       /* MAD wpos_temp.y, wpos_input, wpostrans.zzzz, wpostrans.wwww
        */
-      ureg_MAD(ureg,
-               ureg_writemask(wpos_temp, TGSI_WRITEMASK_Y),
-               wpos_input,
-               ureg_scalar(wpostrans, 2),
-               ureg_scalar(wpostrans, 3));
+      ureg_MAD( ureg,
+                ureg_writemask(wpos_temp, TGSI_WRITEMASK_Y ),
+                wpos_input,
+                ureg_scalar(wpostrans, 2),
+                ureg_scalar(wpostrans, 3));
    }
 
    /* Use wpos_temp as position input from here on:
@@ -4420,8 +4418,37 @@ emit_wpos(struct st_context *st,
    const struct gl_fragment_program *fp =
       (const struct gl_fragment_program *) program;
    struct pipe_screen *pscreen = st->pipe->screen;
+   GLfloat adjX = 0.0f;
+   GLfloat adjY[2] = { 0.0f, 0.0f };
    boolean invert = FALSE;
 
+   /* Query the pixel center conventions supported by the pipe driver and set
+    * adjX, adjY to help out if it cannot handle the requested one internally.
+    *
+    * The bias of the y-coordinate depends on whether y-inversion takes place
+    * (adjY[1]) or not (adjY[0]), which is in turn dependent on whether we are
+    * drawing to an FBO (causes additional inversion), and whether the the pipe
+    * driver origin and the requested origin differ (the latter condition is
+    * stored in the 'invert' variable).
+    *
+    * For height = 100 (i = integer, h = half-integer, l = lower, u = upper):
+    *
+    * center shift only:
+    * i -> h: +0.5
+    * h -> i: -0.5
+    *
+    * inversion only:
+    * l,i -> u,i: ( 0.0 + 1.0) * -1 + 100 = 99
+    * l,h -> u,h: ( 0.5 + 0.0) * -1 + 100 = 99.5
+    * u,i -> l,i: (99.0 + 1.0) * -1 + 100 = 0
+    * u,h -> l,h: (99.5 + 0.0) * -1 + 100 = 0.5
+    *
+    * inversion and center shift:
+    * l,i -> u,h: ( 0.0 + 0.5) * -1 + 100 = 99.5
+    * l,h -> u,i: ( 0.5 + 0.5) * -1 + 100 = 99
+    * u,i -> l,h: (99.0 + 0.5) * -1 + 100 = 0.5
+    * u,h -> l,i: (99.5 + 0.5) * -1 + 100 = 0
+    */
    if (fp->OriginUpperLeft) {
       /* Fragment shader wants origin in upper-left */
       if (pscreen->get_param(pscreen, PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT)) {
@@ -4449,12 +4476,17 @@ emit_wpos(struct st_context *st,
    
    if (fp->PixelCenterInteger) {
       /* Fragment shader wants pixel center integer */
-      if (pscreen->get_param(pscreen, PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER))
+      if (pscreen->get_param(pscreen, PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER)) {
          /* the driver supports pixel center integer */
+         adjY[1] = 1.0f;
          ureg_property_fs_coord_pixel_center(ureg, TGSI_FS_COORD_PIXEL_CENTER_INTEGER);
-      else if (pscreen->get_param(pscreen, PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER))
+      }
+      else if (pscreen->get_param(pscreen, PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER)) {
          /* the driver supports pixel center half integer, need to bias X,Y */
-         emit_adjusted_wpos(t, program, 0.5f, invert ? 0.5f : -0.5f);
+         adjX = -0.5f;
+         adjY[0] = -0.5f;
+         adjY[1] = 0.5f;
+      }
       else
          assert(0);
    }
@@ -4465,8 +4497,8 @@ emit_wpos(struct st_context *st,
       }
       else if (pscreen->get_param(pscreen, PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER)) {
          /* the driver supports pixel center integer, need to bias X,Y */
+         adjX = adjY[0] = adjY[1] = 0.5f;
          ureg_property_fs_coord_pixel_center(ureg, TGSI_FS_COORD_PIXEL_CENTER_INTEGER);
-         emit_adjusted_wpos(t, program, 0.5f, invert ? -0.5f : 0.5f);
       }
       else
          assert(0);
@@ -4474,7 +4506,7 @@ emit_wpos(struct st_context *st,
 
    /* we invert after adjustment so that we avoid the MOV to temporary,
     * and reuse the adjustment ADD instead */
-   emit_wpos_inversion(t, program, invert);
+   emit_wpos_adjustment(t, program, invert, adjX, adjY);
 }
 
 /**
@@ -5026,7 +5058,9 @@ st_link_shader(struct gl_context *ctx, struct gl_shader_program *prog)
 
          progress = do_lower_jumps(ir, true, true, options->EmitNoMainReturn, options->EmitNoCont, options->EmitNoLoops) || progress;
 
-         progress = do_common_optimization(ir, true, options->MaxUnrollIterations) || progress;
+         progress = do_common_optimization(ir, true, true,
+					   options->MaxUnrollIterations)
+	   || progress;
 
          progress = lower_quadop_vector(ir, false) || progress;
 
diff --git a/mesalib/src/mesa/swrast/s_context.c b/mesalib/src/mesa/swrast/s_context.c
index 5287671d7..9112cf30d 100644
--- a/mesalib/src/mesa/swrast/s_context.c
+++ b/mesalib/src/mesa/swrast/s_context.c
@@ -747,6 +747,12 @@ _swrast_CreateContext( struct gl_context *ctx )
    swrast->AllowVertexFog = GL_TRUE;
    swrast->AllowPixelFog = GL_TRUE;
 
+   swrast->Driver.SpanRenderStart = _swrast_span_render_start;
+   swrast->Driver.SpanRenderFinish = _swrast_span_render_finish;
+
+   ctx->Driver.MapTexture = _swrast_map_texture;
+   ctx->Driver.UnmapTexture = _swrast_unmap_texture;
+
    /* Optimized Accum buffer */
    swrast->_IntegerAccumMode = GL_FALSE;
    swrast->_IntegerAccumScaler = 0.0;
@@ -837,6 +843,24 @@ _swrast_render_primitive( struct gl_context *ctx, GLenum prim )
 }
 
 
+/** called via swrast->Driver.SpanRenderStart() */
+void
+_swrast_span_render_start(struct gl_context *ctx)
+{
+   _swrast_map_textures(ctx);
+   _swrast_map_renderbuffers(ctx);
+}
+
+
+/** called via swrast->Driver.SpanRenderFinish() */
+void
+_swrast_span_render_finish(struct gl_context *ctx)
+{
+   _swrast_unmap_textures(ctx);
+   _swrast_unmap_renderbuffers(ctx);
+}
+
+
 void
 _swrast_render_start( struct gl_context *ctx )
 {
diff --git a/mesalib/src/mesa/swrast/s_context.h b/mesalib/src/mesa/swrast/s_context.h
index ec8451eb8..d3ba37819 100644
--- a/mesalib/src/mesa/swrast/s_context.h
+++ b/mesalib/src/mesa/swrast/s_context.h
@@ -138,20 +138,17 @@ struct swrast_texture_image
    /** used for mipmap LOD computation */
    GLfloat WidthScale, HeightScale, DepthScale;
 
-#if 0
-   GLubyte *Data;    /**< The actual texture data in malloc'd memory */
+   /** These fields only valid when texture memory is mapped */
+   GLint RowStride;		/**< Padded width in units of texels */
+   GLuint *ImageOffsets;        /**< if 3D texture: array [Depth] of offsets to
+                                     each 2D slice in 'Data', in texels */
+   GLubyte *Data;		/**< Image data, accessed via FetchTexel() */
 
-   GLint TexelSize;  /**< bytes per texel block */
-#endif
+   /** Malloc'd texture memory */
+   GLubyte *Buffer;
 
    FetchTexelFunc FetchTexel;
    StoreTexelFunc Store;
-
-#if 0
-   /** These fields only valid when texture memory is mapped */
-   GLubyte **SliceMaps;  /**< points to OneMap or a malloc'd array */
-   GLint RowStride;  /**< bytes per row of blocks */
-#endif
 };
 
 
@@ -339,6 +336,31 @@ swrast_render_finish(struct gl_context *ctx)
 }
 
 
+extern void
+_swrast_span_render_start(struct gl_context *ctx);
+
+extern void
+_swrast_span_render_finish(struct gl_context *ctx);
+
+extern void
+_swrast_map_textures(struct gl_context *ctx);
+
+extern void
+_swrast_unmap_textures(struct gl_context *ctx);
+
+extern void
+_swrast_map_texture(struct gl_context *ctx, struct gl_texture_object *texObj);
+
+extern void
+_swrast_unmap_texture(struct gl_context *ctx, struct gl_texture_object *texObj);
+
+
+extern void
+_swrast_map_renderbuffers(struct gl_context *ctx);
+
+extern void
+_swrast_unmap_renderbuffers(struct gl_context *ctx);
+
 
 /**
  * Size of an RGBA pixel, in bytes, for given datatype.
diff --git a/mesalib/src/mesa/swrast/s_fragprog.c b/mesalib/src/mesa/swrast/s_fragprog.c
index 7f205a200..1caa0ebc2 100644
--- a/mesalib/src/mesa/swrast/s_fragprog.c
+++ b/mesalib/src/mesa/swrast/s_fragprog.c
@@ -237,7 +237,8 @@ run_program(struct gl_context *ctx, SWspan *span, GLuint start, GLuint end)
                else if (depth >= 1.0)
                   span->array->z[i] = ctx->DrawBuffer->_DepthMax;
                else
-                  span->array->z[i] = IROUND(depth * ctx->DrawBuffer->_DepthMaxF);
+                  span->array->z[i] =
+                     (GLuint) (depth * ctx->DrawBuffer->_DepthMaxF + 0.5F);
             }
          }
          else {
diff --git a/mesalib/src/mesa/swrast/s_span.c b/mesalib/src/mesa/swrast/s_span.c
index e517c9ae1..4124e444e 100644
--- a/mesalib/src/mesa/swrast/s_span.c
+++ b/mesalib/src/mesa/swrast/s_span.c
@@ -1251,7 +1251,10 @@ _swrast_write_rgba_span( struct gl_context *ctx, SWspan *span)
                       4 * span->end * sizeof(GLchan));
             }
 
-            ASSERT(rb->_BaseFormat == GL_RGBA || rb->_BaseFormat == GL_RGB ||
+            ASSERT(rb->_BaseFormat == GL_RGBA ||
+                   rb->_BaseFormat == GL_RGB ||
+                   rb->_BaseFormat == GL_RED ||
+                   rb->_BaseFormat == GL_RG ||
 		   rb->_BaseFormat == GL_ALPHA);
 
             if (ctx->Color.ColorLogicOpEnabled) {
diff --git a/mesalib/src/mesa/swrast/s_texcombine.c b/mesalib/src/mesa/swrast/s_texcombine.c
index c67c356c1..a7cbb4424 100644
--- a/mesalib/src/mesa/swrast/s_texcombine.c
+++ b/mesalib/src/mesa/swrast/s_texcombine.c
@@ -108,6 +108,7 @@ texture_combine( struct gl_context *ctx, GLuint unit, GLuint n,
             i--;
          }
          _mesa_error(ctx, GL_OUT_OF_MEMORY, "texture_combine");
+         free(rgba);
          return;
       }
    }
diff --git a/mesalib/src/mesa/swrast/s_texfetch_tmp.h b/mesalib/src/mesa/swrast/s_texfetch_tmp.h
index c63b2043c..8b7e930f9 100644
--- a/mesalib/src/mesa/swrast/s_texfetch_tmp.h
+++ b/mesalib/src/mesa/swrast/s_texfetch_tmp.h
@@ -43,7 +43,7 @@
 #if DIM == 1
 
 #define TEXEL_ADDR( type, image, i, j, k, size ) \
-	((void) (j), (void) (k), ((type *)(image)->Base.Data + (i) * (size)))
+	((void) (j), (void) (k), ((type *)(image)->Data + (i) * (size)))
 
 #define FETCH(x) fetch_texel_1d_##x
 
@@ -51,15 +51,15 @@
 
 #define TEXEL_ADDR( type, image, i, j, k, size )			\
 	((void) (k),							\
-	 ((type *)(image)->Base.Data + ((image)->Base.RowStride * (j) + (i)) * (size)))
+	 ((type *)(image)->Data + ((image)->RowStride * (j) + (i)) * (size)))
 
 #define FETCH(x) fetch_texel_2d_##x
 
 #elif DIM == 3
 
 #define TEXEL_ADDR( type, image, i, j, k, size )			\
-	((type *)(image)->Base.Data + ((image)->Base.ImageOffsets[k]		\
-             + (image)->Base.RowStride * (j) + (i)) * (size))
+	((type *)(image)->Data + ((image)->ImageOffsets[k]		\
+             + (image)->RowStride * (j) + (i)) * (size))
 
 #define FETCH(x) fetch_texel_3d_##x
 
diff --git a/mesalib/src/mesa/swrast/s_texfilter.c b/mesalib/src/mesa/swrast/s_texfilter.c
index f8b0fa1aa..9de5c0276 100644
--- a/mesalib/src/mesa/swrast/s_texfilter.c
+++ b/mesalib/src/mesa/swrast/s_texfilter.c
@@ -1375,7 +1375,7 @@ opt_sample_rgb_2d(struct gl_context *ctx,
       GLint i = IFLOOR(texcoords[k][0] * width) & colMask;
       GLint j = IFLOOR(texcoords[k][1] * height) & rowMask;
       GLint pos = (j << shift) | i;
-      GLubyte *texel = ((GLubyte *) img->Data) + 3*pos;
+      GLubyte *texel = swImg->Data + 3 * pos;
       rgba[k][RCOMP] = UBYTE_TO_FLOAT(texel[2]);
       rgba[k][GCOMP] = UBYTE_TO_FLOAT(texel[1]);
       rgba[k][BCOMP] = UBYTE_TO_FLOAT(texel[0]);
@@ -1419,7 +1419,7 @@ opt_sample_rgba_2d(struct gl_context *ctx,
       const GLint col = IFLOOR(texcoords[i][0] * width) & colMask;
       const GLint row = IFLOOR(texcoords[i][1] * height) & rowMask;
       const GLint pos = (row << shift) | col;
-      const GLuint texel = *((GLuint *) img->Data + pos);
+      const GLuint texel = *((GLuint *) swImg->Data + pos);
       rgba[i][RCOMP] = UBYTE_TO_FLOAT( (texel >> 24)        );
       rgba[i][GCOMP] = UBYTE_TO_FLOAT( (texel >> 16) & 0xff );
       rgba[i][BCOMP] = UBYTE_TO_FLOAT( (texel >>  8) & 0xff );
@@ -1442,7 +1442,7 @@ sample_lambda_2d(struct gl_context *ctx,
 
    const GLboolean repeatNoBorderPOT = (tObj->Sampler.WrapS == GL_REPEAT)
       && (tObj->Sampler.WrapT == GL_REPEAT)
-      && (tImg->Border == 0 && (tImg->Width == tImg->RowStride))
+      && (tImg->Border == 0 && (tImg->Width == swImg->RowStride))
       && swImg->_IsPowerOfTwo;
 
    ASSERT(lambda != NULL);
diff --git a/mesalib/src/mesa/swrast/s_texrender.c b/mesalib/src/mesa/swrast/s_texrender.c
index 47e458e1c..e2b921512 100644
--- a/mesalib/src/mesa/swrast/s_texrender.c
+++ b/mesalib/src/mesa/swrast/s_texrender.c
@@ -31,6 +31,15 @@ struct texture_renderbuffer
 };
 
 
+/** cast wrapper */
+static inline struct texture_renderbuffer *
+texture_renderbuffer(struct gl_renderbuffer *rb)
+{
+   return (struct texture_renderbuffer *) rb;
+}
+
+
+
 /**
  * Get row of values from the renderbuffer that wraps a texture image.
  */
@@ -38,8 +47,7 @@ static void
 texture_get_row(struct gl_context *ctx, struct gl_renderbuffer *rb, GLuint count,
                 GLint x, GLint y, void *values)
 {
-   const struct texture_renderbuffer *trb
-      = (const struct texture_renderbuffer *) rb;
+   struct texture_renderbuffer *trb = texture_renderbuffer(rb);
    const GLint z = trb->Zoffset;
    GLuint i;
 
@@ -107,8 +115,7 @@ static void
 texture_get_values(struct gl_context *ctx, struct gl_renderbuffer *rb, GLuint count,
                    const GLint x[], const GLint y[], void *values)
 {
-   const struct texture_renderbuffer *trb
-      = (const struct texture_renderbuffer *) rb;
+   struct texture_renderbuffer *trb = texture_renderbuffer(rb);
    const GLint z = trb->Zoffset;
    GLuint i;
 
@@ -174,8 +181,7 @@ static void
 texture_put_row(struct gl_context *ctx, struct gl_renderbuffer *rb, GLuint count,
                 GLint x, GLint y, const void *values, const GLubyte *mask)
 {
-   const struct texture_renderbuffer *trb
-      = (const struct texture_renderbuffer *) rb;
+   struct texture_renderbuffer *trb = texture_renderbuffer(rb);
    const GLint z = trb->Zoffset;
    GLuint i;
 
@@ -236,8 +242,7 @@ static void
 texture_put_row_rgb(struct gl_context *ctx, struct gl_renderbuffer *rb, GLuint count,
                 GLint x, GLint y, const void *values, const GLubyte *mask)
 {
-   const struct texture_renderbuffer *trb
-      = (const struct texture_renderbuffer *) rb;
+   struct texture_renderbuffer *trb = texture_renderbuffer(rb);
    const GLint z = trb->Zoffset;
    GLuint i;
 
@@ -296,8 +301,7 @@ static void
 texture_put_mono_row(struct gl_context *ctx, struct gl_renderbuffer *rb, GLuint count,
                      GLint x, GLint y, const void *value, const GLubyte *mask)
 {
-   const struct texture_renderbuffer *trb
-      = (const struct texture_renderbuffer *) rb;
+   struct texture_renderbuffer *trb = texture_renderbuffer(rb);
    const GLint z = trb->Zoffset;
    GLuint i;
 
@@ -356,8 +360,7 @@ texture_put_values(struct gl_context *ctx, struct gl_renderbuffer *rb, GLuint co
                    const GLint x[], const GLint y[], const void *values,
                    const GLubyte *mask)
 {
-   const struct texture_renderbuffer *trb
-      = (const struct texture_renderbuffer *) rb;
+   struct texture_renderbuffer *trb = texture_renderbuffer(rb);
    const GLint z = trb->Zoffset;
    GLuint i;
 
@@ -415,8 +418,7 @@ texture_put_mono_values(struct gl_context *ctx, struct gl_renderbuffer *rb,
                         GLuint count, const GLint x[], const GLint y[],
                         const void *value, const GLubyte *mask)
 {
-   const struct texture_renderbuffer *trb
-      = (const struct texture_renderbuffer *) rb;
+   struct texture_renderbuffer *trb = texture_renderbuffer(rb);
    const GLint z = trb->Zoffset;
    GLuint i;
 
@@ -610,7 +612,6 @@ update_wrapper(struct gl_context *ctx, struct gl_renderbuffer_attachment *att)
       trb->Base.DataType = CHAN_TYPE;
       trb->Base._BaseFormat = GL_RGBA;
    }
-   trb->Base.Data = trb->TexImage->Base.Data;
 }
 
 
diff --git a/mesalib/src/mesa/swrast/s_texture.c b/mesalib/src/mesa/swrast/s_texture.c
index 36b429cfa..fb1edb318 100644
--- a/mesalib/src/mesa/swrast/s_texture.c
+++ b/mesalib/src/mesa/swrast/s_texture.c
@@ -69,14 +69,32 @@ _swrast_alloc_texture_image_buffer(struct gl_context *ctx,
 {
    struct swrast_texture_image *swImg = swrast_texture_image(texImage);
    GLuint bytes = _mesa_format_image_size(format, width, height, depth);
+   GLuint i;
 
    /* This _should_ be true (revisit if these ever fail) */
    assert(texImage->Width == width);
    assert(texImage->Height == height);
    assert(texImage->Depth == depth);
 
-   assert(!texImage->Data);
-   texImage->Data = _mesa_align_malloc(bytes, 512);
+   assert(!swImg->Buffer);
+   swImg->Buffer = _mesa_align_malloc(bytes, 512);
+   if (!swImg->Buffer)
+      return GL_FALSE;
+
+   /* RowStride and ImageOffsets[] describe how to address texels in 'Data' */
+   swImg->RowStride = width;
+
+   /* Allocate the ImageOffsets array and initialize to typical values.
+    * We allocate the array for 1D/2D textures too in order to avoid special-
+    * case code in the texstore routines.
+    */
+   swImg->ImageOffsets = (GLuint *) malloc(depth * sizeof(GLuint));
+   if (!swImg->ImageOffsets)
+      return GL_FALSE;
+
+   for (i = 0; i < depth; i++) {
+      swImg->ImageOffsets[i] = i * width * height;
+   }
 
    if ((width == 1 || _mesa_is_pow_two(texImage->Width2)) &&
        (height == 1 || _mesa_is_pow_two(texImage->Height2)) &&
@@ -98,7 +116,7 @@ _swrast_alloc_texture_image_buffer(struct gl_context *ctx,
       swImg->DepthScale = (GLfloat) texImage->Depth;
    }
 
-   return texImage->Data != NULL;
+   return GL_TRUE;
 }
 
 
@@ -109,11 +127,16 @@ void
 _swrast_free_texture_image_buffer(struct gl_context *ctx,
                                   struct gl_texture_image *texImage)
 {
-   if (texImage->Data) {
-      _mesa_align_free(texImage->Data);
+   struct swrast_texture_image *swImage = swrast_texture_image(texImage);
+   if (swImage->Buffer) {
+      _mesa_align_free(swImage->Buffer);
+      swImage->Buffer = NULL;
    }
 
-   texImage->Data = NULL;
+   if (swImage->ImageOffsets) {
+      free(swImage->ImageOffsets);
+      swImage->ImageOffsets = NULL;
+   }
 }
 
 
@@ -155,6 +178,7 @@ _swrast_map_teximage(struct gl_context *ctx,
                      GLubyte **mapOut,
                      GLint *rowStrideOut)
 {
+   struct swrast_texture_image *swImage = swrast_texture_image(texImage);
    GLubyte *map;
    GLint stride, texelSize;
    GLuint bw, bh;
@@ -165,9 +189,9 @@ _swrast_map_teximage(struct gl_context *ctx,
    stride = _mesa_format_row_stride(texImage->TexFormat, texImage->Width);
    _mesa_get_format_block_size(texImage->TexFormat, &bw, &bh);
 
-   assert(texImage->Data);
+   assert(swImage->Buffer);
 
-   map = texImage->Data;
+   map = swImage->Buffer;
 
    if (texImage->TexObject->Target == GL_TEXTURE_3D ||
        texImage->TexObject->Target == GL_TEXTURE_2D_ARRAY) {
@@ -200,3 +224,138 @@ _swrast_unmap_teximage(struct gl_context *ctx,
 {
    /* nop */
 }
+
+
+void
+_swrast_map_texture(struct gl_context *ctx, struct gl_texture_object *texObj)
+{
+   const GLuint faces = texObj->Target == GL_TEXTURE_CUBE_MAP ? 6 : 1;
+   GLuint face, level;
+
+   for (face = 0; face < faces; face++) {
+      for (level = texObj->BaseLevel; level < MAX_TEXTURE_LEVELS; level++) {
+         struct gl_texture_image *texImage = texObj->Image[face][level];
+         if (texImage) {
+            struct swrast_texture_image *swImage =
+               swrast_texture_image(texImage);
+
+            /* XXX we'll eventually call _swrast_map_teximage() here */
+            swImage->Data = swImage->Buffer;
+            assert(swImage->Buffer);
+         }
+      }
+   }
+}
+
+
+void
+_swrast_unmap_texture(struct gl_context *ctx, struct gl_texture_object *texObj)
+{
+   const GLuint faces = texObj->Target == GL_TEXTURE_CUBE_MAP ? 6 : 1;
+   GLuint face, level;
+
+   for (face = 0; face < faces; face++) {
+      for (level = texObj->BaseLevel; level < MAX_TEXTURE_LEVELS; level++) {
+         struct gl_texture_image *texImage = texObj->Image[face][level];
+         if (texImage) {
+            struct swrast_texture_image *swImage
+               = swrast_texture_image(texImage);
+
+            /* XXX we'll eventually call _swrast_unmap_teximage() here */
+            swImage->Data = NULL;
+         }
+      }
+   }
+}
+
+
+/**
+ * Map all textures for reading prior to software rendering.
+ */
+void
+_swrast_map_textures(struct gl_context *ctx)
+{
+   GLbitfield enabledUnits = ctx->Texture._EnabledUnits;
+
+   /* loop over enabled texture units */
+   while (enabledUnits) {
+      GLuint unit = _mesa_ffs(enabledUnits) - 1;
+      struct gl_texture_object *texObj = ctx->Texture.Unit[unit]._Current;
+      
+      _swrast_map_texture(ctx, texObj);
+
+      enabledUnits &= ~(1 << unit);
+   }
+}
+
+
+/**
+ * Unmap all textures for reading prior to software rendering.
+ */
+void
+_swrast_unmap_textures(struct gl_context *ctx)
+{
+   GLbitfield enabledUnits = ctx->Texture._EnabledUnits;
+
+   /* loop over enabled texture units */
+   while (enabledUnits) {
+      GLuint unit = _mesa_ffs(enabledUnits) - 1;
+      struct gl_texture_object *texObj = ctx->Texture.Unit[unit]._Current;
+      
+      _swrast_unmap_texture(ctx, texObj);
+
+      enabledUnits &= ~(1 << unit);
+   }
+}
+
+
+/**
+ * Map or unmap any textures that we may be rendering to as renderbuffers.
+ */
+static void
+map_unmap_renderbuffers(struct gl_context *ctx,
+                        struct gl_framebuffer *fb,
+                        GLboolean map)
+{
+   GLuint i;
+
+   for (i = 0; i < Elements(fb->Attachment); i++) {
+      struct gl_texture_object *texObj = fb->Attachment[i].Texture;
+      if (texObj) {
+         const GLuint level = fb->Attachment[i].TextureLevel;
+         const GLuint face = fb->Attachment[i].CubeMapFace;
+         struct gl_texture_image *texImage = texObj->Image[face][level];
+         if (texImage) {
+            struct swrast_texture_image *swImage
+               = swrast_texture_image(texImage);
+
+            if (map) {
+               /* XXX we'll eventually call _swrast_map_teximage() here */
+               swImage->Data = swImage->Buffer;
+            }
+            else {
+               /* XXX we'll eventually call _swrast_unmap_teximage() here */
+               swImage->Data = NULL;
+            }
+         }
+      }
+   }
+}
+
+
+void
+_swrast_map_renderbuffers(struct gl_context *ctx)
+{
+   map_unmap_renderbuffers(ctx, ctx->DrawBuffer, GL_TRUE);
+   if (ctx->ReadBuffer != ctx->DrawBuffer)
+      map_unmap_renderbuffers(ctx, ctx->ReadBuffer, GL_TRUE);
+}
+
+
+void
+_swrast_unmap_renderbuffers(struct gl_context *ctx)
+{
+   map_unmap_renderbuffers(ctx, ctx->DrawBuffer, GL_FALSE);
+   if (ctx->ReadBuffer != ctx->DrawBuffer)
+      map_unmap_renderbuffers(ctx, ctx->ReadBuffer, GL_FALSE);
+}
diff --git a/mesalib/src/mesa/swrast/s_triangle.c b/mesalib/src/mesa/swrast/s_triangle.c
index 839c4fd08..b4f8e7479 100644
--- a/mesalib/src/mesa/swrast/s_triangle.c
+++ b/mesalib/src/mesa/swrast/s_triangle.c
@@ -127,10 +127,12 @@ _swrast_culltriangle( struct gl_context *ctx,
       ctx->Texture.Unit[0].CurrentTex[TEXTURE_2D_INDEX];		\
    const struct gl_texture_image *texImg =				\
       obj->Image[0][obj->BaseLevel];					\
+   const struct swrast_texture_image *swImg =				\
+      swrast_texture_image_const(texImg);				\
    const GLfloat twidth = (GLfloat) texImg->Width;			\
    const GLfloat theight = (GLfloat) texImg->Height;			\
    const GLint twidth_log2 = texImg->WidthLog2;				\
-   const GLubyte *texture = (const GLubyte *) texImg->Data;		\
+   const GLubyte *texture = (const GLubyte *) swImg->Data;		\
    const GLint smask = texImg->Width - 1;				\
    const GLint tmask = texImg->Height - 1;				\
    ASSERT(texImg->TexFormat == MESA_FORMAT_RGB888);			\
@@ -181,10 +183,12 @@ _swrast_culltriangle( struct gl_context *ctx,
       ctx->Texture.Unit[0].CurrentTex[TEXTURE_2D_INDEX];		\
    const struct gl_texture_image *texImg = 				\
        obj->Image[0][obj->BaseLevel]; 					\
+   const struct swrast_texture_image *swImg =				\
+      swrast_texture_image_const(texImg);				\
    const GLfloat twidth = (GLfloat) texImg->Width;			\
    const GLfloat theight = (GLfloat) texImg->Height;			\
    const GLint twidth_log2 = texImg->WidthLog2;				\
-   const GLubyte *texture = (const GLubyte *) texImg->Data;		\
+   const GLubyte *texture = (const GLubyte *) swImg->Data;		\
    const GLint smask = texImg->Width - 1;				\
    const GLint tmask = texImg->Height - 1;				\
    ASSERT(texImg->TexFormat == MESA_FORMAT_RGB888);			\
@@ -533,9 +537,11 @@ affine_span(struct gl_context *ctx, SWspan *span,
       ctx->Texture.Unit[0].CurrentTex[TEXTURE_2D_INDEX];		\
    const struct gl_texture_image *texImg = 				\
       obj->Image[0][obj->BaseLevel]; 					\
+   const struct swrast_texture_image *swImg =				\
+      swrast_texture_image_const(texImg);				\
    const GLfloat twidth = (GLfloat) texImg->Width;			\
    const GLfloat theight = (GLfloat) texImg->Height;			\
-   info.texture = (const GLchan *) texImg->Data;			\
+   info.texture = (const GLchan *) swImg->Data;				\
    info.twidth_log2 = texImg->WidthLog2;				\
    info.smask = texImg->Width - 1;					\
    info.tmask = texImg->Height - 1;					\
@@ -800,7 +806,9 @@ fast_persp_span(struct gl_context *ctx, SWspan *span,
       ctx->Texture.Unit[0].CurrentTex[TEXTURE_2D_INDEX];		\
    const struct gl_texture_image *texImg = 				\
       obj->Image[0][obj->BaseLevel];			 		\
-   info.texture = (const GLchan *) texImg->Data;			\
+   const struct swrast_texture_image *swImg =				\
+      swrast_texture_image_const(texImg);				\
+   info.texture = (const GLchan *) swImg->Data;				\
    info.twidth_log2 = texImg->WidthLog2;				\
    info.smask = texImg->Width - 1;					\
    info.tmask = texImg->Height - 1;					\
@@ -1062,7 +1070,7 @@ _swrast_choose_triangle( struct gl_context *ctx )
              && texObj2D->_Swizzle == SWIZZLE_NOOP
              && swImg->_IsPowerOfTwo
              && texImg->Border == 0
-             && texImg->Width == texImg->RowStride
+             && texImg->Width == swImg->RowStride
              && (format == MESA_FORMAT_RGB888 || format == MESA_FORMAT_RGBA8888)
              && minFilter == magFilter
              && ctx->Light.Model.ColorControl == GL_SINGLE_COLOR
diff --git a/mesalib/src/mesa/tnl/t_vb_program.c b/mesalib/src/mesa/tnl/t_vb_program.c
index 367dfd5bb..836e8e8fa 100644
--- a/mesalib/src/mesa/tnl/t_vb_program.c
+++ b/mesalib/src/mesa/tnl/t_vb_program.c
@@ -68,6 +68,8 @@ struct vp_stage_data {
    GLubyte *clipmask;                 /**< clip flags */
    GLubyte ormask, andmask;           /**< for clipping */
 
+   GLboolean vertex_textures;
+
    struct gl_program_machine machine;
 };
 
diff --git a/mesalib/src/mesa/vbo/vbo_exec.h b/mesalib/src/mesa/vbo/vbo_exec.h
index 9a1b5a127..0b72579a8 100644
--- a/mesalib/src/mesa/vbo/vbo_exec.h
+++ b/mesalib/src/mesa/vbo/vbo_exec.h
@@ -78,9 +78,6 @@ struct vbo_exec_copied_vtx {
 };
 
 
-typedef void (*vbo_attrfv_func)( const GLfloat * );
-
-
 struct vbo_exec_context
 {
    struct gl_context *ctx;   
@@ -113,8 +110,6 @@ struct vbo_exec_context
        * values are squashed down to the 32 attributes passed to the
        * vertex program below:
        */
-      enum vp_mode program_mode;
-      GLuint enabled_flags;
       const struct gl_client_array *inputs[VERT_ATTRIB_MAX];
    } vtx;
 
@@ -126,10 +121,6 @@ struct vbo_exec_context
    } eval;
 
    struct {
-      enum vp_mode program_mode;
-      GLuint enabled_flags;
-      GLuint array_obj;
-
       /* These just mirror the current arrayobj (todo: make arrayobj
        * look like this and remove the mirror):
        */
diff --git a/mesalib/src/mesa/vbo/vbo_exec_array.c b/mesalib/src/mesa/vbo/vbo_exec_array.c
index 4e4f2c947..7023380a1 100644
--- a/mesalib/src/mesa/vbo/vbo_exec_array.c
+++ b/mesalib/src/mesa/vbo/vbo_exec_array.c
@@ -422,8 +422,6 @@ bind_array_obj(struct gl_context *ctx)
       assert(i < Elements(exec->array.generic_array));
       exec->array.generic_array[i] = &arrayObj->VertexAttrib[i];
    }
-   
-   exec->array.array_obj = arrayObj->Name;
 }
 
 
@@ -444,10 +442,7 @@ recalculate_input_bindings(struct gl_context *ctx)
    GLbitfield const_inputs = 0x0;
    GLuint i;
 
-   exec->array.program_mode = get_program_mode(ctx);
-   exec->array.enabled_flags = ctx->Array.ArrayObj->_Enabled;
-
-   switch (exec->array.program_mode) {
+   switch (get_program_mode(ctx)) {
    case VP_NONE:
       /* When no vertex program is active (or the vertex program is generated
        * from fixed-function state).  We put the material values into the
diff --git a/mesalib/src/mesa/vbo/vbo_save.h b/mesalib/src/mesa/vbo/vbo_save.h
index a064090cf..a85a7cbf6 100644
--- a/mesalib/src/mesa/vbo/vbo_save.h
+++ b/mesalib/src/mesa/vbo/vbo_save.h
@@ -146,7 +146,6 @@ struct vbo_save_context {
    GLuint vert_count;
    GLuint max_vert;
    GLboolean dangling_attr_ref;
-   GLboolean have_materials;
 
    GLuint opcode_vertex_list;