Merge remote branch 'origin/master' into nvc0-new

31 files changed, 1437 insertions, 834 deletions

diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c
index 8d1dc792bc8..f1d08a3e166 100644
--- a/src/mesa/state_tracker/st_atom_constbuf.c
+++ b/src/mesa/state_tracker/st_atom_constbuf.c
@@ -66,6 +66,11 @@ void st_upload_constants( struct st_context *st,
    if (params && params->NumParameters) {
       const uint paramBytes = params->NumParameters * sizeof(GLfloat) * 4;
 
+      /* Update the constants which come from fixed-function state, such as
+       * transformation matrices, fog factors, etc.  The rest of the values in
+       * the parameters list are explicitly set by the user with glUniform,
+       * glProgramParameter(), etc.
+       */
       _mesa_load_state_parameters(st->ctx, params);
 
       /* We always need to get a new buffer, to keep the drivers simple and
diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c
index 036bc60049a..2843b7b1764 100644
--- a/src/mesa/state_tracker/st_atom_framebuffer.c
+++ b/src/mesa/state_tracker/st_atom_framebuffer.c
@@ -51,7 +51,7 @@ static void
 update_renderbuffer_surface(struct st_context *st,
                             struct st_renderbuffer *strb)
 {
-   struct pipe_screen *screen = st->pipe->screen;
+   struct pipe_context *pipe = st->pipe;
    struct pipe_resource *resource = strb->rtt->pt;
    int rtt_width = strb->Base.Width;
    int rtt_height = strb->Base.Height;
@@ -65,15 +65,19 @@ update_renderbuffer_surface(struct st_context *st,
       for (level = 0; level <= resource->last_level; level++) {
          if (u_minify(resource->width0, level) == rtt_width &&
              u_minify(resource->height0, level) == rtt_height) {
+            struct pipe_surface surf_tmpl;
+            memset(&surf_tmpl, 0, sizeof(surf_tmpl));
+            surf_tmpl.format = resource->format;
+            surf_tmpl.usage = PIPE_BIND_RENDER_TARGET;
+            surf_tmpl.u.tex.level = level;
+            surf_tmpl.u.tex.first_layer = strb->rtt_face + strb->rtt_slice;
+            surf_tmpl.u.tex.last_layer = strb->rtt_face + strb->rtt_slice;
 
             pipe_surface_reference(&strb->surface, NULL);
 
-            strb->surface = screen->get_tex_surface(screen,
-						    resource,
-						    strb->rtt_face,
-						    level,
-						    strb->rtt_slice,
-						    PIPE_BIND_RENDER_TARGET);
+            strb->surface = pipe->create_surface(pipe,
+                                                 resource,
+                                                 &surf_tmpl);
 #if 0
             printf("-- alloc new surface %d x %d into tex %p\n",
                    strb->surface->width, strb->surface->height,
diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c
index 6be03376d01..378b30e57cc 100644
--- a/src/mesa/state_tracker/st_atom_pixeltransfer.c
+++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c
@@ -122,8 +122,8 @@ load_color_map_texture(struct gl_context *ctx, struct pipe_resource *pt)
    uint i, j;
 
    transfer = pipe_get_transfer(st_context(ctx)->pipe,
-					     pt, 0, 0, 0, PIPE_TRANSFER_WRITE,
-					     0, 0, texSize, texSize);
+                                pt, 0, 0, PIPE_TRANSFER_WRITE,
+                                0, 0, texSize, texSize);
    dest = (uint *) pipe_transfer_map(pipe, transfer);
 
    /* Pack four 1D maps into a 2D texture:
diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c
index f147d768084..b67068df373 100644
--- a/src/mesa/state_tracker/st_atom_sampler.c
+++ b/src/mesa/state_tracker/st_atom_sampler.c
@@ -121,6 +121,18 @@ static void
 xlate_border_color(const GLfloat *colorIn, GLenum baseFormat, GLfloat *colorOut)
 {
    switch (baseFormat) {
+   case GL_RED:
+      colorOut[0] = colorIn[0];
+      colorOut[1] = 0.0F;
+      colorOut[2] = 0.0F;
+      colorOut[3] = 1.0F;
+      break;
+   case GL_RG:
+      colorOut[0] = colorIn[0];
+      colorOut[1] = colorIn[1];
+      colorOut[2] = 0.0F;
+      colorOut[3] = 1.0F;
+      break;
    case GL_RGB:
       colorOut[0] = colorIn[0];
       colorOut[1] = colorIn[1];
diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c
index 05442ef91b5..c311d043931 100644
--- a/src/mesa/state_tracker/st_atom_shader.c
+++ b/src/mesa/state_tracker/st_atom_shader.c
@@ -50,99 +50,6 @@
 #include "st_program.h"
 
 
-
-/**
- * Translate fragment program if needed.
- */
-static void
-translate_fp(struct st_context *st,
-             struct st_fragment_program *stfp)
-{
-   if (!stfp->tgsi.tokens) {
-      assert(stfp->Base.Base.NumInstructions > 0);
-
-      st_translate_fragment_program(st, stfp);
-   }
-}
-
-/*
- * Translate geometry program if needed.
- */
-static void
-translate_gp(struct st_context *st,
-             struct st_geometry_program *stgp)
-{
-   if (!stgp->tgsi.tokens) {
-      assert(stgp->Base.Base.NumInstructions > 1);
-
-      st_translate_geometry_program(st, stgp);
-   }
-}
-
-/**
- * Find a translated vertex program that corresponds to stvp and
- * has outputs matched to stfp's inputs.
- * This performs vertex and fragment translation (to TGSI) when needed.
- */
-static struct st_vp_varient *
-find_translated_vp(struct st_context *st,
-                   struct st_vertex_program *stvp )
-{
-   struct st_vp_varient *vpv;
-   struct st_vp_varient_key key;
-
-   /* Nothing in our key yet.  This will change:
-    */
-   memset(&key, 0, sizeof key);
-
-   /* When this is true, we will add an extra input to the vertex
-    * shader translation (for edgeflags), an extra output with
-    * edgeflag semantics, and extend the vertex shader to pass through
-    * the input to the output.  We'll need to use similar logic to set
-    * up the extra vertex_element input for edgeflags.
-    * _NEW_POLYGON, ST_NEW_EDGEFLAGS_DATA
-    */
-   key.passthrough_edgeflags = (st->vertdata_edgeflags && (
-                                st->ctx->Polygon.FrontMode != GL_FILL ||
-                                st->ctx->Polygon.BackMode != GL_FILL));
-
-
-   /* Do we need to throw away old translations after a change in the
-    * GL program string?
-    */
-   if (stvp->serialNo != stvp->lastSerialNo) {
-      /* These may have changed if the program string changed.
-       */
-      st_prepare_vertex_program( st, stvp );
-
-      /* We are now up-to-date:
-       */
-      stvp->lastSerialNo = stvp->serialNo;
-   }
-   
-   /* See if we've got a translated vertex program whose outputs match
-    * the fragment program's inputs.
-    */
-   for (vpv = stvp->varients; vpv; vpv = vpv->next) {
-      if (memcmp(&vpv->key, &key, sizeof key) == 0) {
-         break;
-      }
-   }
-
-   /* No?  Perform new translation here. */
-   if (!vpv) {
-      vpv = st_translate_vertex_program(st, stvp, &key);
-      if (!vpv)
-         return NULL;
-      
-      vpv->next = stvp->varients;
-      stvp->varients = vpv;
-   }
-
-   return vpv;
-}
-
-
 /**
  * Return pointer to a pass-through fragment shader.
  * This shader is used when a texture is missing/incomplete.
@@ -167,12 +74,16 @@ static void
 update_fp( struct st_context *st )
 {
    struct st_fragment_program *stfp;
+   struct st_fp_variant_key key;
 
    assert(st->ctx->FragmentProgram._Current);
    stfp = st_fragment_program(st->ctx->FragmentProgram._Current);
    assert(stfp->Base.Base.Target == GL_FRAGMENT_PROGRAM_ARB);
 
-   translate_fp(st, stfp);
+   memset(&key, 0, sizeof(key));
+   key.st = st;
+
+   st->fp_variant = st_get_fp_variant(st, stfp, &key);
 
    st_reference_fragprog(st, &st->fp, stfp);
 
@@ -182,7 +93,8 @@ update_fp( struct st_context *st )
       cso_set_fragment_shader_handle(st->cso_context, fs);
    }
    else {
-      cso_set_fragment_shader_handle(st->cso_context, stfp->driver_shader);
+      cso_set_fragment_shader_handle(st->cso_context,
+                                     st->fp_variant->driver_shader);
    }
 }
 
@@ -206,6 +118,7 @@ static void
 update_vp( struct st_context *st )
 {
    struct st_vertex_program *stvp;
+   struct st_vp_variant_key key;
 
    /* find active shader and params -- Should be covered by
     * ST_NEW_VERTEX_PROGRAM
@@ -214,12 +127,26 @@ update_vp( struct st_context *st )
    stvp = st_vertex_program(st->ctx->VertexProgram._Current);
    assert(stvp->Base.Base.Target == GL_VERTEX_PROGRAM_ARB);
 
-   st->vp_varient = find_translated_vp(st, stvp);
+   memset(&key, 0, sizeof key);
+   key.st = st;  /* variants are per-context */
+
+   /* When this is true, we will add an extra input to the vertex
+    * shader translation (for edgeflags), an extra output with
+    * edgeflag semantics, and extend the vertex shader to pass through
+    * the input to the output.  We'll need to use similar logic to set
+    * up the extra vertex_element input for edgeflags.
+    * _NEW_POLYGON, ST_NEW_EDGEFLAGS_DATA
+    */
+   key.passthrough_edgeflags = (st->vertdata_edgeflags && (
+                                st->ctx->Polygon.FrontMode != GL_FILL ||
+                                st->ctx->Polygon.BackMode != GL_FILL));
+
+   st->vp_variant = st_get_vp_variant(st, stvp, &key);
 
    st_reference_vertprog(st, &st->vp, stvp);
 
    cso_set_vertex_shader_handle(st->cso_context, 
-                                st->vp_varient->driver_shader);
+                                st->vp_variant->driver_shader);
 
    st->vertex_result_to_slot = stvp->result_to_output;
 }
@@ -231,14 +158,16 @@ const struct st_tracked_state st_update_vp = {
       _NEW_POLYGON,					/* mesa */
       ST_NEW_VERTEX_PROGRAM | ST_NEW_EDGEFLAGS_DATA	/* st */
    },
-   update_vp					/* update */
+   update_vp						/* update */
 };
 
+
+
 static void
 update_gp( struct st_context *st )
 {
-
    struct st_geometry_program *stgp;
+   struct st_gp_variant_key key;
 
    if (!st->ctx->GeometryProgram._Current) {
       cso_set_geometry_shader_handle(st->cso_context, NULL);
@@ -248,18 +177,22 @@ update_gp( struct st_context *st )
    stgp = st_geometry_program(st->ctx->GeometryProgram._Current);
    assert(stgp->Base.Base.Target == MESA_GEOMETRY_PROGRAM);
 
-   translate_gp(st, stgp);
+   memset(&key, 0, sizeof(key));
+   key.st = st;
+
+   st->gp_variant = st_get_gp_variant(st, stgp, &key);
 
    st_reference_geomprog(st, &st->gp, stgp);
 
-   cso_set_geometry_shader_handle(st->cso_context, stgp->driver_shader);
+   cso_set_geometry_shader_handle(st->cso_context,
+                                  st->gp_variant->driver_shader);
 }
 
 const struct st_tracked_state st_update_gp = {
-   "st_update_gp",					/* name */
-   {							/* dirty */
-      0,						/* mesa */
-      ST_NEW_GEOMETRY_PROGRAM                           /* st */
+   "st_update_gp",			/* name */
+   {					/* dirty */
+      0,				/* mesa */
+      ST_NEW_GEOMETRY_PROGRAM           /* st */
    },
-   update_gp  					/* update */
+   update_gp  				/* update */
 };
diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c
index 6c5caf42e35..a76ae92dc3d 100644
--- a/src/mesa/state_tracker/st_cb_accum.c
+++ b/src/mesa/state_tracker/st_cb_accum.c
@@ -138,10 +138,10 @@ accum_accum(struct st_context *st, GLfloat value,
       debug_printf("%s: fallback processing\n", __FUNCTION__);
 
    color_trans = pipe_get_transfer(st->pipe,
-						color_strb->texture,
-						0, 0, 0,
-						PIPE_TRANSFER_READ, xpos, ypos,
-						width, height);
+                                   color_strb->texture,
+                                   0, 0,
+                                   PIPE_TRANSFER_READ, xpos, ypos,
+                                   width, height);
 
    buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
 
@@ -187,9 +187,9 @@ accum_load(struct st_context *st, GLfloat value,
       debug_printf("%s: fallback processing\n", __FUNCTION__);
 
    color_trans = pipe_get_transfer(st->pipe, color_strb->texture,
-						0, 0, 0,
-						PIPE_TRANSFER_READ, xpos, ypos,
-						width, height);
+                                   0, 0,
+                                   PIPE_TRANSFER_READ, xpos, ypos,
+                                   width, height);
 
    buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
 
@@ -241,12 +241,12 @@ accum_return(struct gl_context *ctx, GLfloat value,
       usage = PIPE_TRANSFER_READ_WRITE;
    else
       usage = PIPE_TRANSFER_WRITE;
-   
+
    color_trans = pipe_get_transfer(st_context(ctx)->pipe,
-						color_strb->texture, 0, 0, 0,
-						usage,
-						xpos, ypos,
-						width, height);
+                                   color_strb->texture, 0, 0,
+                                   usage,
+                                   xpos, ypos,
+                                   width, height);
 
    if (usage & PIPE_TRANSFER_READ)
       pipe_get_tile_rgba(pipe, color_trans, 0, 0, width, height, buf);
diff --git a/src/mesa/state_tracker/st_cb_accum.h b/src/mesa/state_tracker/st_cb_accum.h
index b8c9c350031..050a21483e4 100644
--- a/src/mesa/state_tracker/st_cb_accum.h
+++ b/src/mesa/state_tracker/st_cb_accum.h
@@ -30,7 +30,11 @@
 #define ST_CB_ACCUM_H
 
 
-#include "main/mtypes.h"
+#include "main/mfeatures.h"
+
+struct dd_function_table;
+struct gl_context;
+struct gl_renderbuffer;
 
 #if FEATURE_accum
 
diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c
index 3c0ee6c2883..516346c8c7a 100644
--- a/src/mesa/state_tracker/st_cb_bitmap.c
+++ b/src/mesa/state_tracker/st_cb_bitmap.c
@@ -185,48 +185,47 @@ find_free_bit(uint bitfield)
 
 /**
  * Combine basic bitmap fragment program with the user-defined program.
+ * \param st  current context
+ * \param fpIn  the incoming fragment program
+ * \param fpOut  the new fragment program which does fragment culling
+ * \param bitmap_sampler  sampler number for the bitmap texture
  */
-static struct st_fragment_program *
-combined_bitmap_fragment_program(struct gl_context *ctx)
+void
+st_make_bitmap_fragment_program(struct st_context *st,
+                                struct gl_fragment_program *fpIn,
+                                struct gl_fragment_program **fpOut,
+                                GLuint *bitmap_sampler)
 {
-   struct st_context *st = st_context(ctx);
-   struct st_fragment_program *stfp = st->fp;
-
-   if (!stfp->bitmap_program) {
-      /*
-       * Generate new program which is the user-defined program prefixed
-       * with the bitmap sampler/kill instructions.
-       */
-      struct st_fragment_program *bitmap_prog;
-      uint sampler;
-
-      sampler = find_free_bit(st->fp->Base.Base.SamplersUsed);
-      bitmap_prog = make_bitmap_fragment_program(ctx, sampler);
+   struct st_fragment_program *bitmap_prog;
+   struct gl_program *newProg;
+   uint sampler;
 
-      stfp->bitmap_program = (struct st_fragment_program *)
-         _mesa_combine_programs(ctx,
-                                &bitmap_prog->Base.Base, &stfp->Base.Base);
-      stfp->bitmap_program->bitmap_sampler = sampler;
+   /*
+    * Generate new program which is the user-defined program prefixed
+    * with the bitmap sampler/kill instructions.
+    */
+   sampler = find_free_bit(fpIn->Base.SamplersUsed);
+   bitmap_prog = make_bitmap_fragment_program(st->ctx, sampler);
 
-      /* done with this after combining */
-      st_reference_fragprog(st, &bitmap_prog, NULL);
+   newProg = _mesa_combine_programs(st->ctx,
+                                    &bitmap_prog->Base.Base,
+                                    &fpIn->Base);
+   /* done with this after combining */
+   st_reference_fragprog(st, &bitmap_prog, NULL);
 
 #if 0
-      {
-         struct gl_program *p = &stfp->bitmap_program->Base.Base;
-         printf("Combined bitmap program:\n");
-         _mesa_print_program(p);
-         printf("InputsRead: 0x%x\n", p->InputsRead);
-         printf("OutputsWritten: 0x%x\n", p->OutputsWritten);
-         _mesa_print_parameter_list(p->Parameters);
-      }
-#endif
-
-      /* translate to TGSI tokens */
-      st_translate_fragment_program(st, stfp->bitmap_program);
+   {
+      printf("Combined bitmap program:\n");
+      _mesa_print_program(newProg);
+      printf("InputsRead: 0x%x\n", newProg->InputsRead);
+      printf("OutputsWritten: 0x%x\n", newProg->OutputsWritten);
+      _mesa_print_parameter_list(newProg->Parameters);
    }
+#endif
 
-   return stfp->bitmap_program;
+   /* return results */
+   *fpOut = (struct gl_fragment_program *) newProg;
+   *bitmap_sampler = sampler;
 }
 
 
@@ -283,9 +282,9 @@ make_bitmap_texture(struct gl_context *ctx, GLsizei width, GLsizei height,
       return NULL;
    }
 
-   transfer = pipe_get_transfer(st->pipe, pt, 0, 0, 0,
-					   PIPE_TRANSFER_WRITE,
-					   0, 0, width, height);
+   transfer = pipe_get_transfer(st->pipe, pt, 0, 0,
+                                PIPE_TRANSFER_WRITE,
+                                0, 0, width, height);
 
    dest = pipe_transfer_map(pipe, transfer);
 
@@ -349,7 +348,8 @@ setup_bitmap_vertex_data(struct st_context *st, bool normalized,
    if (!st->bitmap.vbuf) {
       st->bitmap.vbuf = pipe_buffer_create(pipe->screen, 
                                            PIPE_BIND_VERTEX_BUFFER,
-                                           max_slots * sizeof(st->bitmap.vertices));
+                                           max_slots *
+                                           sizeof(st->bitmap.vertices));
    }
 
    /* Positions are in clip coords since we need to do clipping in case
@@ -389,10 +389,11 @@ setup_bitmap_vertex_data(struct st_context *st, bool normalized,
 
    /* put vertex data into vbuf */
    pipe_buffer_write_nooverlap(st->pipe,
-                                           st->bitmap.vbuf,
-                                           st->bitmap.vbuf_slot * sizeof st->bitmap.vertices,
-                                           sizeof st->bitmap.vertices,
-                                           st->bitmap.vertices);
+                               st->bitmap.vbuf,
+                               st->bitmap.vbuf_slot
+                               * sizeof(st->bitmap.vertices),
+                               sizeof st->bitmap.vertices,
+                               st->bitmap.vertices);
 
    return st->bitmap.vbuf_slot++ * sizeof st->bitmap.vertices;
 }
@@ -411,11 +412,16 @@ draw_bitmap_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
    struct st_context *st = st_context(ctx);
    struct pipe_context *pipe = st->pipe;
    struct cso_context *cso = st->cso_context;
-   struct st_fragment_program *stfp;
+   struct st_fp_variant *fpv;
+   struct st_fp_variant_key key;
    GLuint maxSize;
    GLuint offset;
 
-   stfp = combined_bitmap_fragment_program(ctx);
+   memset(&key, 0, sizeof(key));
+   key.st = st;
+   key.bitmap = GL_TRUE;
+
+   fpv = st_get_fp_variant(st, st->fp, &key);
 
    /* As an optimization, Mesa's fragment programs will sometimes get the
     * primary color from a statevar/constant rather than a varying variable.
@@ -428,7 +434,7 @@ draw_bitmap_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
       GLfloat colorSave[4];
       COPY_4V(colorSave, ctx->Current.Attrib[VERT_ATTRIB_COLOR0]);
       COPY_4V(ctx->Current.Attrib[VERT_ATTRIB_COLOR0], color);
-      st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT);
+      st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
       COPY_4V(ctx->Current.Attrib[VERT_ATTRIB_COLOR0], colorSave);
    }
 
@@ -437,7 +443,8 @@ draw_bitmap_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
    /* XXX if the bitmap is larger than the max texture size, break
     * it up into chunks.
     */
-   maxSize = 1 << (pipe->screen->get_param(pipe->screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
+   maxSize = 1 << (pipe->screen->get_param(pipe->screen,
+                                    PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
    assert(width <= (GLsizei)maxSize);
    assert(height <= (GLsizei)maxSize);
 
@@ -454,7 +461,7 @@ draw_bitmap_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
    cso_set_rasterizer(cso, &st->bitmap.rasterizer);
 
    /* fragment shader state: TEX lookup program */
-   cso_set_fragment_shader_handle(cso, stfp->driver_shader);
+   cso_set_fragment_shader_handle(cso, fpv->driver_shader);
 
    /* vertex shader state: position + texcoord pass-through */
    cso_set_vertex_shader_handle(cso, st->bitmap.vs);
@@ -462,21 +469,22 @@ draw_bitmap_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
    /* user samplers, plus our bitmap sampler */
    {
       struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS];
-      uint num = MAX2(stfp->bitmap_sampler + 1, st->state.num_samplers);
+      uint num = MAX2(fpv->bitmap_sampler + 1, st->state.num_samplers);
       uint i;
       for (i = 0; i < st->state.num_samplers; i++) {
          samplers[i] = &st->state.samplers[i];
       }
-      samplers[stfp->bitmap_sampler] = &st->bitmap.samplers[sv->texture->target != PIPE_TEXTURE_RECT];
+      samplers[fpv->bitmap_sampler] =
+         &st->bitmap.samplers[sv->texture->target != PIPE_TEXTURE_RECT];
       cso_set_samplers(cso, num, (const struct pipe_sampler_state **) samplers);
    }
 
    /* user textures, plus the bitmap texture */
    {
       struct pipe_sampler_view *sampler_views[PIPE_MAX_SAMPLERS];
-      uint num = MAX2(stfp->bitmap_sampler + 1, st->state.num_textures);
+      uint num = MAX2(fpv->bitmap_sampler + 1, st->state.num_textures);
       memcpy(sampler_views, st->state.sampler_views, sizeof(sampler_views));
-      sampler_views[stfp->bitmap_sampler] = sv;
+      sampler_views[fpv->bitmap_sampler] = sv;
       cso_set_fragment_sampler_views(cso, num, sampler_views);
    }
 
@@ -504,7 +512,9 @@ draw_bitmap_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
    z = z * 2.0 - 1.0;
 
    /* draw textured quad */
-   offset = setup_bitmap_vertex_data(st, sv->texture->target != PIPE_TEXTURE_RECT, x, y, width, height, z, color);
+   offset = setup_bitmap_vertex_data(st,
+                                     sv->texture->target != PIPE_TEXTURE_RECT,
+                                     x, y, width, height, z, color);
 
    util_draw_vertex_buffer(pipe, st->bitmap.vbuf, offset,
                            PIPE_PRIM_TRIANGLE_FAN,
@@ -573,6 +583,9 @@ print_cache(const struct bitmap_cache *cache)
 }
 
 
+/**
+ * Create gallium pipe_transfer object for the bitmap cache.
+ */
 static void
 create_cache_trans(struct st_context *st)
 {
@@ -585,10 +598,10 @@ create_cache_trans(struct st_context *st)
    /* Map the texture transfer.
     * Subsequent glBitmap calls will write into the texture image.
     */
-   cache->trans = pipe_get_transfer(st->pipe, cache->texture, 0, 0, 0,
-					       PIPE_TRANSFER_WRITE, 0, 0,
-					       BITMAP_CACHE_WIDTH,
-					       BITMAP_CACHE_HEIGHT);
+   cache->trans = pipe_get_transfer(st->pipe, cache->texture, 0, 0,
+                                    PIPE_TRANSFER_WRITE, 0, 0,
+                                    BITMAP_CACHE_WIDTH,
+                                    BITMAP_CACHE_HEIGHT);
    cache->buffer = pipe_transfer_map(pipe, cache->trans);
 
    /* init image to all 0xff */
@@ -651,7 +664,9 @@ st_flush_bitmap_cache(struct st_context *st)
    }
 }
 
-/* Flush bitmap cache and release vertex buffer.
+
+/**
+ * Flush bitmap cache and release vertex buffer.
  */
 void
 st_flush_bitmap( struct st_context *st )
@@ -737,7 +752,8 @@ accum_bitmap(struct st_context *st,
  * Called via ctx->Driver.Bitmap()
  */
 static void
-st_Bitmap(struct gl_context *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
+st_Bitmap(struct gl_context *ctx, GLint x, GLint y,
+          GLsizei width, GLsizei height,
           const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap )
 {
    struct st_context *st = st_context(ctx);
@@ -764,7 +780,8 @@ st_Bitmap(struct gl_context *ctx, GLint x, GLint y, GLsizei width, GLsizei heigh
 
    pt = make_bitmap_texture(ctx, width, height, unpack, bitmap);
    if (pt) {
-      struct pipe_sampler_view *sv = st_create_texture_sampler_view(st->pipe, pt);
+      struct pipe_sampler_view *sv =
+         st_create_texture_sampler_view(st->pipe, pt);
 
       assert(pt->target == PIPE_TEXTURE_2D || pt->target == PIPE_TEXTURE_RECT);
 
@@ -814,15 +831,18 @@ st_init_bitmap(struct st_context *st)
    st->bitmap.rasterizer.gl_rasterization_rules = 1;
 
    /* find a usable texture format */
-   if (screen->is_format_supported(screen, PIPE_FORMAT_I8_UNORM, PIPE_TEXTURE_2D, 0,
+   if (screen->is_format_supported(screen, PIPE_FORMAT_I8_UNORM,
+                                   PIPE_TEXTURE_2D, 0,
                                    PIPE_BIND_SAMPLER_VIEW, 0)) {
       st->bitmap.tex_format = PIPE_FORMAT_I8_UNORM;
    }
-   else if (screen->is_format_supported(screen, PIPE_FORMAT_A8_UNORM, PIPE_TEXTURE_2D, 0,
+   else if (screen->is_format_supported(screen, PIPE_FORMAT_A8_UNORM,
+                                        PIPE_TEXTURE_2D, 0,
                                         PIPE_BIND_SAMPLER_VIEW, 0)) {
       st->bitmap.tex_format = PIPE_FORMAT_A8_UNORM;
    }
-   else if (screen->is_format_supported(screen, PIPE_FORMAT_L8_UNORM, PIPE_TEXTURE_2D, 0,
+   else if (screen->is_format_supported(screen, PIPE_FORMAT_L8_UNORM,
+                                        PIPE_TEXTURE_2D, 0,
                                         PIPE_BIND_SAMPLER_VIEW, 0)) {
       st->bitmap.tex_format = PIPE_FORMAT_L8_UNORM;
    }
@@ -845,8 +865,6 @@ st_destroy_bitmap(struct st_context *st)
    struct pipe_context *pipe = st->pipe;
    struct bitmap_cache *cache = st->bitmap.cache;
 
-
-
    if (st->bitmap.vs) {
       cso_delete_vertex_shader(st->cso_context, st->bitmap.vs);
       st->bitmap.vs = NULL;
diff --git a/src/mesa/state_tracker/st_cb_bitmap.h b/src/mesa/state_tracker/st_cb_bitmap.h
index d04b2b67795..2bd63b9b741 100644
--- a/src/mesa/state_tracker/st_cb_bitmap.h
+++ b/src/mesa/state_tracker/st_cb_bitmap.h
@@ -34,6 +34,8 @@
 
 struct dd_function_table;
 struct st_context;
+struct gl_fragment_program;
+struct st_fragment_program;
 
 #if FEATURE_drawpix
 
@@ -47,6 +49,12 @@ extern void
 st_destroy_bitmap(struct st_context *st);
 
 extern void
+st_make_bitmap_fragment_program(struct st_context *st,
+                                struct gl_fragment_program *fpIn,
+                                struct gl_fragment_program **fpOut,
+                                GLuint *bitmap_sampler);
+
+extern void
 st_flush_bitmap_cache(struct st_context *st);
 
 /* Flush bitmap cache and release vertex buffer.  Needed at end of
diff --git a/src/mesa/state_tracker/st_cb_blit.c b/src/mesa/state_tracker/st_cb_blit.c
index af41835326a..06cee520b37 100644
--- a/src/mesa/state_tracker/st_cb_blit.c
+++ b/src/mesa/state_tracker/st_cb_blit.c
@@ -115,17 +115,14 @@ st_BlitFramebuffer(struct gl_context *ctx,
             st_texture_object(srcAtt->Texture);
          struct st_renderbuffer *dstRb =
             st_renderbuffer(drawFB->_ColorDrawBuffers[0]);
-         struct pipe_subresource srcSub;
          struct pipe_surface *dstSurf = dstRb->surface;
 
          if (!srcObj->pt)
             return;
 
-         srcSub.face = srcAtt->CubeMapFace;
-         srcSub.level = srcAtt->TextureLevel;
-
-         util_blit_pixels(st->blit, srcObj->pt, srcSub,
-                          srcX0, srcY0, srcX1, srcY1, srcAtt->Zoffset,
+         util_blit_pixels(st->blit, srcObj->pt, srcAtt->TextureLevel,
+                          srcX0, srcY0, srcX1, srcY1,
+                          srcAtt->Zoffset + srcAtt->CubeMapFace,
                           dstSurf, dstX0, dstY0, dstX1, dstY1,
                           0.0, pFilter);
       }
@@ -136,14 +133,11 @@ st_BlitFramebuffer(struct gl_context *ctx,
             st_renderbuffer(drawFB->_ColorDrawBuffers[0]);
          struct pipe_surface *srcSurf = srcRb->surface;
          struct pipe_surface *dstSurf = dstRb->surface;
-         struct pipe_subresource srcSub;
-
-         srcSub.face = srcSurf->face;
-         srcSub.level = srcSurf->level;
 
          util_blit_pixels(st->blit,
-                          srcRb->texture, srcSub, srcX0, srcY0, srcX1, srcY1,
-                          srcSurf->zslice,
+                          srcRb->texture, srcSurf->u.tex.level,
+                          srcX0, srcY0, srcX1, srcY1,
+                          srcSurf->u.tex.first_layer,
                           dstSurf, dstX0, dstY0, dstX1, dstY1,
                           0.0, pFilter);
       }
@@ -176,11 +170,11 @@ st_BlitFramebuffer(struct gl_context *ctx,
          /* Blitting depth and stencil values between combined
           * depth/stencil buffers.  This is the ideal case for such buffers.
           */
-         util_blit_pixels(st->blit, srcDepthRb->texture,
-                          u_subresource(srcDepthRb->surface->face,
-                                        srcDepthRb->surface->level),
+         util_blit_pixels(st->blit,
+                          srcDepthRb->texture,
+                          srcDepthRb->surface->u.tex.level,
                           srcX0, srcY0, srcX1, srcY1,
-                          srcDepthRb->surface->zslice,
+                          srcDepthRb->surface->u.tex.first_layer,
                           dstDepthSurf, dstX0, dstY0, dstX1, dstY1,
                           0.0, pFilter);
       }
@@ -189,10 +183,9 @@ st_BlitFramebuffer(struct gl_context *ctx,
 
          if (mask & GL_DEPTH_BUFFER_BIT) {
             util_blit_pixels(st->blit, srcDepthRb->texture,
-                             u_subresource(srcDepthRb->surface->face,
-                                           srcDepthRb->surface->level),
+                             srcDepthRb->surface->u.tex.level,
                              srcX0, srcY0, srcX1, srcY1,
-                             srcDepthRb->surface->zslice,
+                             srcDepthRb->surface->u.tex.first_layer,
                              dstDepthSurf, dstX0, dstY0, dstX1, dstY1,
                              0.0, pFilter);
          }
diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c
index 27540c36ce7..8b60f9040d0 100644
--- a/src/mesa/state_tracker/st_cb_bufferobjects.c
+++ b/src/mesa/state_tracker/st_cb_bufferobjects.c
@@ -211,6 +211,13 @@ st_bufferobj_data(struct gl_context *ctx,
 
 
 /**
+ * Dummy data whose's pointer is used for zero size buffers or ranges.
+ */
+static long st_bufferobj_zero_length = 0;
+
+
+
+/**
  * Called via glMapBufferARB().
  */
 static void *
@@ -233,10 +240,16 @@ st_bufferobj_map(struct gl_context *ctx, GLenum target, GLenum access,
       break;      
    }
 
-   obj->Pointer = pipe_buffer_map(st_context(ctx)->pipe,
-                                  st_obj->buffer,
-                                  flags,
-                                  &st_obj->transfer);
+   /* Handle zero-size buffers here rather than in drivers */
+   if (obj->Size == 0) {
+      obj->Pointer = &st_bufferobj_zero_length;
+   }
+   else {
+      obj->Pointer = pipe_buffer_map(st_context(ctx)->pipe,
+                                     st_obj->buffer,
+                                     flags,
+                                     &st_obj->transfer);
+   }
 
    if (obj->Pointer) {
       obj->Offset = 0;
@@ -247,13 +260,6 @@ st_bufferobj_map(struct gl_context *ctx, GLenum target, GLenum access,
 
 
 /**
- * Dummy data whose's pointer is used for zero length ranges.
- */
-static long
-st_bufferobj_zero_length_range = 0;
-
-
-/**
  * Called via glMapBufferRange().
  */
 static void *
@@ -273,6 +279,12 @@ st_bufferobj_map_range(struct gl_context *ctx, GLenum target,
 
    if (access & GL_MAP_FLUSH_EXPLICIT_BIT)
       flags |= PIPE_TRANSFER_FLUSH_EXPLICIT;
+
+   if (access & GL_MAP_INVALIDATE_RANGE_BIT)
+      flags |= PIPE_TRANSFER_DISCARD;
+
+   if (access & GL_MAP_INVALIDATE_BUFFER_BIT)
+      flags |= PIPE_TRANSFER_DISCARD;
    
    if (access & GL_MAP_UNSYNCHRONIZED_BIT)
       flags |= PIPE_TRANSFER_UNSYNCHRONIZED;
@@ -293,7 +305,7 @@ st_bufferobj_map_range(struct gl_context *ctx, GLenum target,
     * length range from the pipe driver.
     */
    if (!length) {
-      obj->Pointer = &st_bufferobj_zero_length_range;
+      obj->Pointer = &st_bufferobj_zero_length;
    }
    else {
       obj->Pointer = pipe_buffer_map_range(pipe, 
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c
index 122d8f07654..1fc47688e4a 100644
--- a/src/mesa/state_tracker/st_cb_drawpixels.c
+++ b/src/mesa/state_tracker/st_cb_drawpixels.c
@@ -94,88 +94,61 @@ is_passthrough_program(const struct gl_fragment_program *prog)
 /**
  * Make fragment shader for glDraw/CopyPixels.  This shader is made
  * by combining the pixel transfer shader with the user-defined shader.
- * \return pointer to Gallium driver fragment shader
+ * \param fpIn  the current/incoming fragment program
+ * \param fpOut  returns the combined fragment program
  */
-static void *
-combined_drawpix_fragment_program(struct gl_context *ctx)
+void
+st_make_drawpix_fragment_program(struct st_context *st,
+                                 struct gl_fragment_program *fpIn,
+                                 struct gl_fragment_program **fpOut)
 {
-   struct st_context *st = st_context(ctx);
-   struct st_fragment_program *stfp;
+   struct gl_program *newProg;
 
-   if (st->pixel_xfer.program->serialNo == st->pixel_xfer.xfer_prog_sn
-       && st->fp->serialNo == st->pixel_xfer.user_prog_sn) {
-      /* the pixel tranfer program has not changed and the user-defined
-       * program has not changed, so re-use the combined program.
-       */
-      stfp = st->pixel_xfer.combined_prog;
+   if (is_passthrough_program(fpIn)) {
+      newProg = (struct gl_program *) _mesa_clone_fragment_program(st->ctx,
+                                             &st->pixel_xfer.program->Base);
    }
    else {
-      /* Concatenate the pixel transfer program with the current user-
-       * defined program.
-       */
-      if (is_passthrough_program(&st->fp->Base)) {
-         stfp = (struct st_fragment_program *)
-            _mesa_clone_fragment_program(ctx, &st->pixel_xfer.program->Base);
-      }
-      else {
 #if 0
-         printf("Base program:\n");
-         _mesa_print_program(&st->fp->Base.Base);
-         printf("DrawPix program:\n");
-         _mesa_print_program(&st->pixel_xfer.program->Base.Base);
+      /* debug */
+      printf("Base program:\n");
+      _mesa_print_program(&fpIn->Base);
+      printf("DrawPix program:\n");
+      _mesa_print_program(&st->pixel_xfer.program->Base.Base);
 #endif
-         stfp = (struct st_fragment_program *)
-            _mesa_combine_programs(ctx,
-                                   &st->pixel_xfer.program->Base.Base,
-                                   &st->fp->Base.Base);
-      }
+      newProg = _mesa_combine_programs(st->ctx,
+                                       &st->pixel_xfer.program->Base.Base,
+                                       &fpIn->Base);
+   }
 
 #if 0
-      {
-         struct gl_program *p = &stfp->Base.Base;
-         printf("Combined DrawPixels program:\n");
-         _mesa_print_program(p);
-         printf("InputsRead: 0x%x\n", p->InputsRead);
-         printf("OutputsWritten: 0x%x\n", p->OutputsWritten);
-         _mesa_print_parameter_list(p->Parameters);
-      }
+   /* debug */
+   printf("Combined DrawPixels program:\n");
+   _mesa_print_program(newProg);
+   printf("InputsRead: 0x%x\n", newProg->InputsRead);
+   printf("OutputsWritten: 0x%x\n", newProg->OutputsWritten);
+   _mesa_print_parameter_list(newProg->Parameters);
 #endif
 
-      /* translate to TGSI tokens */
-      st_translate_fragment_program(st, stfp);
-
-      /* save new program, update serial numbers */
-      st->pixel_xfer.xfer_prog_sn = st->pixel_xfer.program->serialNo;
-      st->pixel_xfer.user_prog_sn = st->fp->serialNo;
-      st->pixel_xfer.combined_prog_sn = stfp->serialNo;
-      /* can't reference new program directly, already have a reference on it */
-      st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL);
-      st->pixel_xfer.combined_prog = stfp;
-   }
-
-   /* Ideally we'd have updated the pipe constants during the normal
-    * st/atom mechanism.  But we can't since this is specific to glDrawPixels.
-    */
-   st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT);
-
-   return stfp->driver_shader;
+   *fpOut = (struct gl_fragment_program *) newProg;
 }
 
 
 /**
- * Create fragment shader that does a TEX() instruction to get a Z and/or
+ * Create fragment program that does a TEX() instruction to get a Z and/or
  * stencil value value, then writes to FRAG_RESULT_DEPTH/FRAG_RESULT_STENCIL.
  * Used for glDrawPixels(GL_DEPTH_COMPONENT / GL_STENCIL_INDEX).
  * Pass fragment color through as-is.
- * \return pointer to the Gallium driver fragment shader
+ * \return pointer to the gl_fragment program
  */
-static void *
-make_fragment_shader_z_stencil(struct st_context *st, GLboolean write_depth,
-                               GLboolean write_stencil)
+struct gl_fragment_program *
+st_make_drawpix_z_stencil_program(struct st_context *st,
+                                  GLboolean write_depth,
+                                  GLboolean write_stencil)
 {
    struct gl_context *ctx = st->ctx;
    struct gl_program *p;
-   struct st_fragment_program *stp;
+   struct gl_fragment_program *fp;
    GLuint ic = 0;
    const GLuint shaderIndex = write_depth * 2 + write_stencil;
 
@@ -183,7 +156,7 @@ make_fragment_shader_z_stencil(struct st_context *st, GLboolean write_depth,
 
    if (st->drawpix.shaders[shaderIndex]) {
       /* already have the proper shader */
-      return st->drawpix.shaders[shaderIndex]->driver_shader;
+      return st->drawpix.shaders[shaderIndex];
    }
 
    /*
@@ -245,18 +218,15 @@ make_fragment_shader_z_stencil(struct st_context *st, GLboolean write_depth,
    if (write_stencil)
       p->SamplersUsed |= 1 << 1;
 
-   stp = st_fragment_program((struct gl_fragment_program *) p);
+   fp = (struct gl_fragment_program *) p;
 
    /* save the new shader */
-   st->drawpix.shaders[shaderIndex] = stp;
+   st->drawpix.shaders[shaderIndex] = fp;
 
-   st_translate_fragment_program(st, stp);
-
-   return stp->driver_shader;
+   return fp;
 }
 
 
-
 /**
  * Create a simple vertex shader that just passes through the
  * vertex position and texcoord (and optionally, color).
@@ -427,9 +397,9 @@ make_texture(struct st_context *st,
       /* we'll do pixel transfer in a fragment shader */
       ctx->_ImageTransferState = 0x0;
 
-      transfer = pipe_get_transfer(st->pipe, pt, 0, 0, 0,
-					      PIPE_TRANSFER_WRITE, 0, 0,
-					      width, height);
+      transfer = pipe_get_transfer(st->pipe, pt, 0, 0,
+                                   PIPE_TRANSFER_WRITE, 0, 0,
+                                   width, height);
 
       /* map texture transfer */
       dest = pipe_transfer_map(pipe, transfer);
@@ -591,7 +561,6 @@ draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
    assert(width <= maxSize);
    assert(height <= maxSize);
 
-   cso_save_depth_stencil_alpha(cso);
    cso_save_rasterizer(cso);
    cso_save_viewport(cso);
    cso_save_samplers(cso);
@@ -599,6 +568,10 @@ draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
    cso_save_fragment_shader(cso);
    cso_save_vertex_shader(cso);
    cso_save_vertex_elements(cso);
+   if (write_stencil) {
+      cso_save_depth_stencil_alpha(cso);
+      cso_save_blend(cso);
+   }
 
    /* rasterizer state: just scissor */
    {
@@ -609,22 +582,30 @@ draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
       cso_set_rasterizer(cso, &rasterizer);
    }
 
-   if (write_depth || write_stencil)
-   {
+   if (write_stencil) {
+      /* Stencil writing bypasses the normal fragment pipeline to
+       * disable color writing and set stencil test to always pass.
+       */
       struct pipe_depth_stencil_alpha_state dsa;
+      struct pipe_blend_state blend;
+
+      /* depth/stencil */
       memset(&dsa, 0, sizeof(dsa));
+      dsa.stencil[0].enabled = 1;
+      dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
+      dsa.stencil[0].writemask = ctx->Stencil.WriteMask[0] & 0xff;
+      dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
       if (write_depth) {
-	 dsa.depth.enabled = 1;
-	 dsa.depth.func = PIPE_FUNC_ALWAYS;
-	 dsa.depth.writemask = 1;
-      }
-      if (write_stencil) {
-	 dsa.stencil[0].enabled = 1;
-	 dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
-	 dsa.stencil[0].writemask = 0xff;
-	 dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
+         /* writing depth+stencil: depth test always passes */
+         dsa.depth.enabled = 1;
+         dsa.depth.writemask = ctx->Depth.Mask;
+         dsa.depth.func = PIPE_FUNC_ALWAYS;
       }
       cso_set_depth_stencil_alpha(cso, &dsa);
+
+      /* blend (colormask) */
+      memset(&blend, 0, sizeof(blend));
+      cso_set_blend(cso, &blend);
    }
 
    /* fragment shader state: TEX lookup program */
@@ -696,7 +677,6 @@ draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
              normalized ? ((GLfloat) height / sv[0]->texture->height0) : (GLfloat)height);
 
    /* restore state */
-   cso_restore_depth_stencil_alpha(cso);
    cso_restore_rasterizer(cso);
    cso_restore_viewport(cso);
    cso_restore_samplers(cso);
@@ -704,6 +684,10 @@ draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
    cso_restore_fragment_shader(cso);
    cso_restore_vertex_shader(cso);
    cso_restore_vertex_elements(cso);
+   if (write_stencil) {
+      cso_restore_depth_stencil_alpha(cso);
+      cso_restore_blend(cso);
+   }
 }
 
 
@@ -749,9 +733,9 @@ draw_stencil_pixels(struct gl_context *ctx, GLint x, GLint y,
    else
       usage = PIPE_TRANSFER_WRITE;
 
-   pt = pipe_get_transfer(st_context(ctx)->pipe, strb->texture, 0, 0, 0,
-				       usage, x, y,
-				       width, height);
+   pt = pipe_get_transfer(st_context(ctx)->pipe, strb->texture, 0, 0,
+                                     usage, x, y,
+                                     width, height);
 
    stmap = pipe_transfer_map(pipe, pt);
 
@@ -859,6 +843,61 @@ draw_stencil_pixels(struct gl_context *ctx, GLint x, GLint y,
 
 
 /**
+ * Get fragment program variant for a glDrawPixels or glCopyPixels
+ * command for RGBA data.
+ */
+static struct st_fp_variant *
+get_color_fp_variant(struct st_context *st)
+{
+   struct gl_context *ctx = st->ctx;
+   struct st_fp_variant_key key;
+   struct st_fp_variant *fpv;
+
+   memset(&key, 0, sizeof(key));
+
+   key.st = st;
+   key.drawpixels = 1;
+   key.scaleAndBias = (ctx->Pixel.RedBias != 0.0 ||
+                       ctx->Pixel.RedScale != 1.0 ||
+                       ctx->Pixel.GreenBias != 0.0 ||
+                       ctx->Pixel.GreenScale != 1.0 ||
+                       ctx->Pixel.BlueBias != 0.0 ||
+                       ctx->Pixel.BlueScale != 1.0 ||
+                       ctx->Pixel.AlphaBias != 0.0 ||
+                       ctx->Pixel.AlphaScale != 1.0);
+   key.pixelMaps = ctx->Pixel.MapColorFlag;
+
+   fpv = st_get_fp_variant(st, st->fp, &key);
+
+   return fpv;
+}
+
+
+/**
+ * Get fragment program variant for a glDrawPixels or glCopyPixels
+ * command for depth/stencil data.
+ */
+static struct st_fp_variant *
+get_depth_stencil_fp_variant(struct st_context *st, GLboolean write_depth,
+                             GLboolean write_stencil)
+{
+   struct st_fp_variant_key key;
+   struct st_fp_variant *fpv;
+
+   memset(&key, 0, sizeof(key));
+
+   key.st = st;
+   key.drawpixels = 1;
+   key.drawpixels_z = write_depth;
+   key.drawpixels_stencil = write_stencil;
+
+   fpv = st_get_fp_variant(st, st->fp, &key);
+
+   return fpv;
+}
+
+
+/**
  * Called via ctx->Driver.DrawPixels()
  */
 static void
@@ -875,6 +914,7 @@ st_DrawPixels(struct gl_context *ctx, GLint x, GLint y,
    struct pipe_sampler_view *sv[2];
    int num_sampler_view = 1;
    enum pipe_format stencil_format = PIPE_FORMAT_NONE;
+   struct st_fp_variant *fpv;
 
    if (format == GL_DEPTH_STENCIL)
       write_stencil = write_depth = GL_TRUE;
@@ -907,14 +947,25 @@ st_DrawPixels(struct gl_context *ctx, GLint x, GLint y,
 
    st_validate_state(st);
 
+   /*
+    * Get vertex/fragment shaders
+    */
    if (write_depth || write_stencil) {
-      driver_fp = make_fragment_shader_z_stencil(st, write_depth, write_stencil);
+      fpv = get_depth_stencil_fp_variant(st, write_depth, write_stencil);
+
+      driver_fp = fpv->driver_shader;
+
       driver_vp = make_passthrough_vertex_shader(st, GL_TRUE);
+
       color = ctx->Current.RasterColor;
    }
    else {
-      driver_fp = combined_drawpix_fragment_program(ctx);
+      fpv = get_color_fp_variant(st);
+
+      driver_fp = fpv->driver_shader;
+
       driver_vp = make_passthrough_vertex_shader(st, GL_FALSE);
+
       color = NULL;
       if (st->pixel_xfer.pixelmap_enabled) {
 	  sv[1] = st->pixel_xfer.pixelmap_sampler_view;
@@ -922,6 +973,9 @@ st_DrawPixels(struct gl_context *ctx, GLint x, GLint y,
       }
    }
 
+   /* update fragment program constants */
+   st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
+
    /* draw with textured quad */
    {
       struct pipe_resource *pt
@@ -994,20 +1048,32 @@ copy_stencil_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
                           GL_STENCIL_INDEX, GL_UNSIGNED_BYTE,
                           &ctx->DefaultPacking, buffer);
 
+   if (0) {
+      /* debug code: dump stencil values */
+      GLint row, col;
+      for (row = 0; row < height; row++) {
+         printf("%3d: ", row);
+         for (col = 0; col < width; col++) {
+            printf("%02x ", buffer[col + row * width]);
+         }
+         printf("\n");
+      }
+   }
+
    if (util_format_get_component_bits(rbDraw->format,
                                      UTIL_FORMAT_COLORSPACE_ZS, 0) != 0)
       usage = PIPE_TRANSFER_READ_WRITE;
    else
       usage = PIPE_TRANSFER_WRITE;
-   
+
    if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
       dsty = rbDraw->Base.Height - dsty - height;
    }
 
    ptDraw = pipe_get_transfer(st_context(ctx)->pipe,
-					   rbDraw->texture, 0, 0, 0,
-					   usage, dstx, dsty,
-					   width, height);
+                              rbDraw->texture, 0, 0,
+                              usage, dstx, dsty,
+                              width, height);
 
    assert(util_format_get_blockwidth(ptDraw->resource->format) == 1);
    assert(util_format_get_blockheight(ptDraw->resource->format) == 1);
@@ -1090,6 +1156,7 @@ st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
    GLint readX, readY, readW, readH;
    GLuint sample_count;
    struct gl_pixelstore_attrib pack = ctx->DefaultPacking;
+   struct st_fp_variant *fpv;
 
    st_validate_state(st);
 
@@ -1099,11 +1166,18 @@ st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
       return;
    }
 
+   /*
+    * Get vertex/fragment shaders
+    */
    if (type == GL_COLOR) {
       rbRead = st_get_color_read_renderbuffer(ctx);
       color = NULL;
-      driver_fp = combined_drawpix_fragment_program(ctx);
+
+      fpv = get_color_fp_variant(st);
+      driver_fp = fpv->driver_shader;
+
       driver_vp = make_passthrough_vertex_shader(st, GL_FALSE);
+
       if (st->pixel_xfer.pixelmap_enabled) {
 	  sv[1] = st->pixel_xfer.pixelmap_sampler_view;
 	  num_sampler_view++;
@@ -1113,10 +1187,17 @@ st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
       assert(type == GL_DEPTH);
       rbRead = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer);
       color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
-      driver_fp = make_fragment_shader_z_stencil(st, GL_TRUE, GL_FALSE);
+
+      fpv = get_depth_stencil_fp_variant(st, GL_TRUE, GL_FALSE);
+      driver_fp = fpv->driver_shader;
+
       driver_vp = make_passthrough_vertex_shader(st, GL_TRUE);
    }
 
+   /* update fragment program constants */
+   st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
+
+
    if (rbRead->Base.Wrapped)
       rbRead = st_renderbuffer(rbRead->Base.Wrapped);
 
@@ -1183,27 +1264,24 @@ st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
    /* Make temporary texture which is a copy of the src region.
     */
    if (srcFormat == texFormat) {
-      struct pipe_subresource srcsub, dstsub;
-      srcsub.face = 0;
-      srcsub.level = 0;
-      dstsub.face = 0;
-      dstsub.level = 0;
-      /* copy source framebuffer surface into mipmap/texture */
+      struct pipe_box src_box;
+      u_box_2d(readX, readY, readW, readH, &src_box);
+    /* copy source framebuffer surface into mipmap/texture */
       pipe->resource_copy_region(pipe,
                                  pt,                                /* dest tex */
-                                 dstsub,
+                                 0,
                                  pack.SkipPixels, pack.SkipRows, 0, /* dest pos */
                                  rbRead->texture,                   /* src tex */
-                                 srcsub,
-                                 readX, readY, 0, readW, readH);    /* src region */
+                                 0,
+                                 &src_box);
 
    }
    else {
       /* CPU-based fallback/conversion */
       struct pipe_transfer *ptRead =
-         pipe_get_transfer(st->pipe, rbRead->texture, 0, 0, 0,
-                                        PIPE_TRANSFER_READ,
-                                        readX, readY, readW, readH);
+         pipe_get_transfer(st->pipe, rbRead->texture, 0, 0,
+                           PIPE_TRANSFER_READ,
+                           readX, readY, readW, readH);
       struct pipe_transfer *ptTex;
       enum pipe_transfer_usage transfer_usage;
 
@@ -1215,8 +1293,8 @@ st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
       else
          transfer_usage = PIPE_TRANSFER_WRITE;
 
-      ptTex = pipe_get_transfer(st->pipe, pt, 0, 0, 0, transfer_usage,
-                                             0, 0, width, height);
+      ptTex = pipe_get_transfer(st->pipe, pt, 0, 0, transfer_usage,
+                                0, 0, width, height);
 
       /* copy image from ptRead surface to ptTex surface */
       if (type == GL_COLOR) {
@@ -1271,7 +1349,7 @@ st_destroy_drawpix(struct st_context *st)
 
    for (i = 0; i < Elements(st->drawpix.shaders); i++) {
       if (st->drawpix.shaders[i])
-         st_reference_fragprog(st, &st->drawpix.shaders[i], NULL);
+         _mesa_reference_fragprog(st->ctx, &st->drawpix.shaders[i], NULL);
    }
 
    st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL);
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.h b/src/mesa/state_tracker/st_cb_drawpixels.h
index 575f169e08e..8f73e626240 100644
--- a/src/mesa/state_tracker/st_cb_drawpixels.h
+++ b/src/mesa/state_tracker/st_cb_drawpixels.h
@@ -42,6 +42,16 @@ extern void st_init_drawpixels_functions(struct dd_function_table *functions);
 extern void
 st_destroy_drawpix(struct st_context *st);
 
+extern void
+st_make_drawpix_fragment_program(struct st_context *st,
+                                 struct gl_fragment_program *fpIn,
+                                 struct gl_fragment_program **fpOut);
+
+extern struct gl_fragment_program *
+st_make_drawpix_z_stencil_program(struct st_context *st,
+                                  GLboolean write_depth,
+                                  GLboolean write_stencil);
+
 #else
 
 static INLINE void
diff --git a/src/mesa/state_tracker/st_cb_eglimage.c b/src/mesa/state_tracker/st_cb_eglimage.c
index b39a624a612..c4b84de790c 100644
--- a/src/mesa/state_tracker/st_cb_eglimage.c
+++ b/src/mesa/state_tracker/st_cb_eglimage.c
@@ -106,6 +106,7 @@ st_bind_surface(struct gl_context *ctx, GLenum target,
    struct st_texture_object *stObj;
    struct st_texture_image *stImage;
    GLenum internalFormat;
+   gl_format texFormat;
 
    /* map pipe format to base format */
    if (util_format_get_component_bits(ps->format, UTIL_FORMAT_COLORSPACE_RGB, 3) > 0)
@@ -122,10 +123,11 @@ st_bind_surface(struct gl_context *ctx, GLenum target,
       stObj->surface_based = GL_TRUE;
    }
 
+   texFormat = st_pipe_format_to_mesa_format(ps->format);
+
    _mesa_init_teximage_fields(ctx, target, texImage,
-                              ps->width, ps->height, 1, 0, internalFormat);
-   texImage->TexFormat = st_pipe_format_to_mesa_format(ps->format);
-   _mesa_set_fetch_functions(texImage, 2);
+                              ps->width, ps->height, 1, 0, internalFormat,
+                              texFormat);
 
    /* FIXME create a non-default sampler view from the pipe_surface? */
    pipe_resource_reference(&stObj->pt, ps->texture);
diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c
index 9425f07aee6..cd718a31a14 100644
--- a/src/mesa/state_tracker/st_cb_fbo.c
+++ b/src/mesa/state_tracker/st_cb_fbo.c
@@ -52,6 +52,7 @@
 
 #include "util/u_format.h"
 #include "util/u_inlines.h"
+#include "util/u_surface.h"
 
 
 /**
@@ -65,9 +66,11 @@ st_renderbuffer_alloc_storage(struct gl_context * ctx, struct gl_renderbuffer *r
                               GLuint width, GLuint height)
 {
    struct st_context *st = st_context(ctx);
+   struct pipe_context *pipe = st->pipe;
    struct pipe_screen *screen = st->pipe->screen;
    struct st_renderbuffer *strb = st_renderbuffer(rb);
    enum pipe_format format;
+   struct pipe_surface surf_tmpl;
 
    if (strb->format != PIPE_FORMAT_NONE)
       format = strb->format;
@@ -113,6 +116,7 @@ st_renderbuffer_alloc_storage(struct gl_context * ctx, struct gl_renderbuffer *r
       template.width0 = width;
       template.height0 = height;
       template.depth0 = 1;
+      template.array_size = 1;
       template.last_level = 0;
       template.nr_samples = rb->NumSamples;
       if (util_format_is_depth_or_stencil(format)) {
@@ -120,7 +124,7 @@ st_renderbuffer_alloc_storage(struct gl_context * ctx, struct gl_renderbuffer *r
       }
       else {
          template.bind = (PIPE_BIND_DISPLAY_TARGET |
-			  PIPE_BIND_RENDER_TARGET);
+                          PIPE_BIND_RENDER_TARGET);
       }
 
       strb->texture = screen->resource_create(screen, &template);
@@ -128,10 +132,11 @@ st_renderbuffer_alloc_storage(struct gl_context * ctx, struct gl_renderbuffer *r
       if (!strb->texture) 
          return FALSE;
 
-      strb->surface = screen->get_tex_surface(screen,
-                                              strb->texture,
-                                              0, 0, 0,
-                                              template.bind);
+      memset(&surf_tmpl, 0, sizeof(surf_tmpl));
+      u_surface_default_template(&surf_tmpl, strb->texture, template.bind);
+      strb->surface = pipe->create_surface(pipe,
+                                           strb->texture,
+                                           &surf_tmpl);
       if (strb->surface) {
          assert(strb->surface->texture);
          assert(strb->surface->format);
@@ -327,12 +332,12 @@ st_render_texture(struct gl_context *ctx,
 {
    struct st_context *st = st_context(ctx);
    struct pipe_context *pipe = st->pipe;
-   struct pipe_screen *screen = pipe->screen;
    struct st_renderbuffer *strb;
    struct gl_renderbuffer *rb;
    struct pipe_resource *pt = st_get_texobj_resource(att->Texture);
    struct st_texture_object *stObj;
    const struct gl_texture_image *texImage;
+   struct pipe_surface surf_tmpl;
 
    /* When would this fail?  Perhaps assert? */
    if (!pt) 
@@ -381,12 +386,15 @@ st_render_texture(struct gl_context *ctx,
    assert(strb->rtt_level <= strb->texture->last_level);
 
    /* new surface for rendering into the texture */
-   strb->surface = screen->get_tex_surface(screen,
-                                           strb->texture,
-                                           strb->rtt_face,
-                                           strb->rtt_level,
-                                           strb->rtt_slice,
-                                           PIPE_BIND_RENDER_TARGET);
+   memset(&surf_tmpl, 0, sizeof(surf_tmpl));
+   surf_tmpl.format = strb->texture->format;
+   surf_tmpl.usage = PIPE_BIND_RENDER_TARGET;
+   surf_tmpl.u.tex.level = strb->rtt_level;
+   surf_tmpl.u.tex.first_layer = strb->rtt_face + strb->rtt_slice;
+   surf_tmpl.u.tex.last_layer = strb->rtt_face + strb->rtt_slice;
+   strb->surface = pipe->create_surface(pipe,
+                                        strb->texture,
+                                        &surf_tmpl);
 
    strb->format = pt->format;
 
diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c
index 4d83fcc6ccb..32694975d17 100644
--- a/src/mesa/state_tracker/st_cb_program.c
+++ b/src/mesa/state_tracker/st_cb_program.c
@@ -46,16 +46,13 @@
 #include "st_cb_program.h"
 
 
-static GLuint SerialNo = 1;
-
 
 /**
  * Called via ctx->Driver.BindProgram() to bind an ARB vertex or
  * fragment program.
  */
-static void st_bind_program( struct gl_context *ctx,
-			     GLenum target, 
-			     struct gl_program *prog )
+static void
+st_bind_program(struct gl_context *ctx, GLenum target, struct gl_program *prog)
 {
    struct st_context *st = st_context(ctx);
 
@@ -77,7 +74,8 @@ static void st_bind_program( struct gl_context *ctx,
  * Called via ctx->Driver.UseProgram() to bind a linked GLSL program
  * (vertex shader + fragment shader).
  */
-static void st_use_program( struct gl_context *ctx, struct gl_shader_program *shProg)
+static void
+st_use_program(struct gl_context *ctx, struct gl_shader_program *shProg)
 {
    struct st_context *st = st_context(ctx);
 
@@ -87,48 +85,28 @@ static void st_use_program( struct gl_context *ctx, struct gl_shader_program *sh
 }
 
 
-
 /**
  * Called via ctx->Driver.NewProgram() to allocate a new vertex or
  * fragment program.
  */
-static struct gl_program *st_new_program( struct gl_context *ctx,
-					  GLenum target,
-					  GLuint id )
+static struct gl_program *
+st_new_program(struct gl_context *ctx, GLenum target, GLuint id)
 {
    switch (target) {
    case GL_VERTEX_PROGRAM_ARB: {
       struct st_vertex_program *prog = ST_CALLOC_STRUCT(st_vertex_program);
-
-      prog->serialNo = SerialNo++;
-
-      return _mesa_init_vertex_program( ctx, 
-					&prog->Base,
-					target, 
-					id );
+      return _mesa_init_vertex_program(ctx, &prog->Base, target, id);
    }
 
    case GL_FRAGMENT_PROGRAM_ARB:
    case GL_FRAGMENT_PROGRAM_NV: {
       struct st_fragment_program *prog = ST_CALLOC_STRUCT(st_fragment_program);
-
-      prog->serialNo = SerialNo++;
-
-      return _mesa_init_fragment_program( ctx, 
-					  &prog->Base,
-					  target, 
-					  id );
+      return _mesa_init_fragment_program(ctx, &prog->Base, target, id);
    }
 
    case MESA_GEOMETRY_PROGRAM: {
       struct st_geometry_program *prog = ST_CALLOC_STRUCT(st_geometry_program);
-
-      prog->serialNo = SerialNo++;
-
-      return _mesa_init_geometry_program( ctx,
-                                          &prog->Base,
-                                          target,
-                                          id );
+      return _mesa_init_geometry_program(ctx, &prog->Base, target, id);
    }
 
    default:
@@ -138,7 +116,10 @@ static struct gl_program *st_new_program( struct gl_context *ctx,
 }
 
 
-void
+/**
+ * Called via ctx->Driver.DeleteProgram()
+ */
+static void
 st_delete_program(struct gl_context *ctx, struct gl_program *prog)
 {
    struct st_context *st = st_context(ctx);
@@ -147,17 +128,15 @@ st_delete_program(struct gl_context *ctx, struct gl_program *prog)
    case GL_VERTEX_PROGRAM_ARB:
       {
          struct st_vertex_program *stvp = (struct st_vertex_program *) prog;
-         st_vp_release_varients( st, stvp );
+         st_release_vp_variants( st, stvp );
       }
       break;
    case MESA_GEOMETRY_PROGRAM:
       {
-         struct st_geometry_program *stgp = (struct st_geometry_program *) prog;
+         struct st_geometry_program *stgp =
+            (struct st_geometry_program *) prog;
 
-         if (stgp->driver_shader) {
-            cso_delete_geometry_shader(st->cso_context, stgp->driver_shader);
-            stgp->driver_shader = NULL;
-         }
+         st_release_gp_variants(st, stgp);
 
          if (stgp->tgsi.tokens) {
             st_free_tokens((void *) stgp->tgsi.tokens);
@@ -167,23 +146,15 @@ st_delete_program(struct gl_context *ctx, struct gl_program *prog)
       break;
    case GL_FRAGMENT_PROGRAM_ARB:
       {
-         struct st_fragment_program *stfp = (struct st_fragment_program *) prog;
+         struct st_fragment_program *stfp =
+            (struct st_fragment_program *) prog;
 
-         if (stfp->driver_shader) {
-            cso_delete_fragment_shader(st->cso_context, stfp->driver_shader);
-            stfp->driver_shader = NULL;
-         }
+         st_release_fp_variants(st, stfp);
          
          if (stfp->tgsi.tokens) {
             st_free_tokens(stfp->tgsi.tokens);
             stfp->tgsi.tokens = NULL;
          }
-
-         if (stfp->bitmap_program) {
-            struct gl_program *prg = &stfp->bitmap_program->Base.Base;
-            _mesa_reference_program(ctx, &prg, NULL);
-            stfp->bitmap_program = NULL;
-         }
       }
       break;
    default:
@@ -195,15 +166,25 @@ st_delete_program(struct gl_context *ctx, struct gl_program *prog)
 }
 
 
-static GLboolean st_is_program_native( struct gl_context *ctx,
-				       GLenum target, 
-				       struct gl_program *prog )
+/**
+ * Called via ctx->Driver.IsProgramNative()
+ */
+static GLboolean
+st_is_program_native(struct gl_context *ctx,
+                     GLenum target, 
+                     struct gl_program *prog)
 {
    return GL_TRUE;
 }
 
 
-static GLboolean st_program_string_notify( struct gl_context *ctx,
+/**
+ * Called via ctx->Driver.ProgramStringNotify()
+ * Called when the program's text/code is changed.  We have to free
+ * all shader variants and corresponding gallium shaders when this happens.
+ */
+static GLboolean
+st_program_string_notify( struct gl_context *ctx,
                                            GLenum target,
                                            struct gl_program *prog )
 {
@@ -212,12 +193,7 @@ static GLboolean st_program_string_notify( struct gl_context *ctx,
    if (target == GL_FRAGMENT_PROGRAM_ARB) {
       struct st_fragment_program *stfp = (struct st_fragment_program *) prog;
 
-      stfp->serialNo++;
-
-      if (stfp->driver_shader) {
-         cso_delete_fragment_shader(st->cso_context, stfp->driver_shader);
-         stfp->driver_shader = NULL;
-      }
+      st_release_fp_variants(st, stfp);
 
       if (stfp->tgsi.tokens) {
          st_free_tokens(stfp->tgsi.tokens);
@@ -230,12 +206,7 @@ static GLboolean st_program_string_notify( struct gl_context *ctx,
    else if (target == MESA_GEOMETRY_PROGRAM) {
       struct st_geometry_program *stgp = (struct st_geometry_program *) prog;
 
-      stgp->serialNo++;
-
-      if (stgp->driver_shader) {
-         cso_delete_geometry_shader(st->cso_context, stgp->driver_shader);
-         stgp->driver_shader = NULL;
-      }
+      st_release_gp_variants(st, stgp);
 
       if (stgp->tgsi.tokens) {
          st_free_tokens((void *) stgp->tgsi.tokens);
@@ -248,9 +219,7 @@ static GLboolean st_program_string_notify( struct gl_context *ctx,
    else if (target == GL_VERTEX_PROGRAM_ARB) {
       struct st_vertex_program *stvp = (struct st_vertex_program *) prog;
 
-      stvp->serialNo++;
-
-      st_vp_release_varients( st, stvp );
+      st_release_vp_variants( st, stvp );
 
       if (st->vp == stvp)
 	 st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
@@ -261,8 +230,11 @@ static GLboolean st_program_string_notify( struct gl_context *ctx,
 }
 
 
-
-void st_init_program_functions(struct dd_function_table *functions)
+/**
+ * Plug in the program and shader-related device driver functions.
+ */
+void
+st_init_program_functions(struct dd_function_table *functions)
 {
    functions->BindProgram = st_bind_program;
    functions->UseProgram = st_use_program;
diff --git a/src/mesa/state_tracker/st_cb_program.h b/src/mesa/state_tracker/st_cb_program.h
index 004afb6d812..091a4439c48 100644
--- a/src/mesa/state_tracker/st_cb_program.h
+++ b/src/mesa/state_tracker/st_cb_program.h
@@ -29,15 +29,10 @@
 #define ST_CB_PROGRAM_H
 
 
-#include "main/mtypes.h"
-
 struct dd_function_table;
 
 extern void
 st_init_program_functions(struct dd_function_table *functions);
 
-extern void
-st_delete_program(struct gl_context *ctx, struct gl_program *prog);
-
 
 #endif
diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c
index bcd46ac9d54..0507be74578 100644
--- a/src/mesa/state_tracker/st_cb_readpixels.c
+++ b/src/mesa/state_tracker/st_cb_readpixels.c
@@ -80,7 +80,7 @@ st_read_stencil_pixels(struct gl_context *ctx, GLint x, GLint y,
    /* Create a read transfer from the renderbuffer's texture */
 
    pt = pipe_get_transfer(pipe, strb->texture,
-                          0, 0, 0,  /* face, level, zslice */
+                          0, 0,
                           PIPE_TRANSFER_READ,
                           x, y, width, height);
 
@@ -236,7 +236,7 @@ st_fast_readpixels(struct gl_context *ctx, struct st_renderbuffer *strb,
       }
 
       trans = pipe_get_transfer(pipe, strb->texture,
-                                0, 0, 0,  /* face, level, zslice */
+                                0, 0,
                                 PIPE_TRANSFER_READ,
                                 x, y, width, height);
       if (!trans) {
@@ -328,7 +328,7 @@ st_readpixels(struct gl_context *ctx, GLint x, GLint y, GLsizei width, GLsizei h
 {
    struct st_context *st = st_context(ctx);
    struct pipe_context *pipe = st->pipe;
-   GLfloat temp[MAX_WIDTH][4];
+   GLfloat (*temp)[4];
    const GLbitfield transferOps = ctx->_ImageTransferState;
    GLsizei i, j;
    GLint yStep, dfStride;
@@ -381,6 +381,13 @@ st_readpixels(struct gl_context *ctx, GLint x, GLint y, GLsizei width, GLsizei h
       return;
    }
 
+   /* allocate temp pixel row buffer */
+   temp = (GLfloat (*)[4]) malloc(4 * width * sizeof(GLfloat));
+   if (!temp) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glReadPixels");
+      return;
+   }
+
    if (format == GL_RGBA && type == GL_FLOAT) {
       /* write tile(row) directly into user's buffer */
       df = (GLfloat *) _mesa_image_address2d(&clippedPacking, dest, width,
@@ -400,7 +407,7 @@ st_readpixels(struct gl_context *ctx, GLint x, GLint y, GLsizei width, GLsizei h
 
    /* Create a read transfer from the renderbuffer's texture */
    trans = pipe_get_transfer(pipe, strb->texture,
-                             0, 0, 0,  /* face, level, zslice */
+                             0, 0,
                              PIPE_TRANSFER_READ,
                              x, y, width, height);
 
@@ -533,6 +540,8 @@ st_readpixels(struct gl_context *ctx, GLint x, GLint y, GLsizei width, GLsizei h
       }
    }
 
+   free(temp);
+
    pipe->transfer_destroy(pipe, trans);
 
    _mesa_unmap_pbo_dest(ctx, &clippedPacking);
diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c
index 4c2f7d78960..866426a7549 100644
--- a/src/mesa/state_tracker/st_cb_texture.c
+++ b/src/mesa/state_tracker/st_cb_texture.c
@@ -63,7 +63,7 @@
 #include "util/u_surface.h"
 #include "util/u_sampler.h"
 #include "util/u_math.h"
-
+#include "util/u_box.h"
 
 #define DBG if (0) printf
 
@@ -431,7 +431,7 @@ compress_with_blit(struct gl_context * ctx,
    struct pipe_resource *src_tex;
    struct pipe_sampler_view view_templ;
    struct pipe_sampler_view *src_view;
-   struct pipe_surface *dst_surface;
+   struct pipe_surface *dst_surface, surf_tmpl;
    struct pipe_transfer *tex_xfer;
    void *map;
 
@@ -441,9 +441,13 @@ compress_with_blit(struct gl_context * ctx,
    }
 
    /* get destination surface (in the compressed texture) */
-   dst_surface = screen->get_tex_surface(screen, stImage->pt,
-                                         stImage->face, stImage->level, 0,
-                                         0 /* flags */);
+   memset(&surf_tmpl, 0, sizeof(surf_tmpl));
+   surf_tmpl.format = stImage->pt->format;
+   surf_tmpl.usage = PIPE_BIND_RENDER_TARGET;
+   surf_tmpl.u.tex.level = stImage->level;
+   surf_tmpl.u.tex.first_layer = stImage->face;
+   surf_tmpl.u.tex.last_layer = stImage->face;
+   dst_surface = pipe->create_surface(pipe, stImage->pt, &surf_tmpl);
    if (!dst_surface) {
       /* can't render into this format (or other problem) */
       return GL_FALSE;
@@ -464,6 +468,7 @@ compress_with_blit(struct gl_context * ctx,
    templ.width0 = width;
    templ.height0 = height;
    templ.depth0 = 1;
+   templ.array_size = 1;
    templ.last_level = 0;
    templ.usage = PIPE_USAGE_DEFAULT;
    templ.bind = PIPE_BIND_SAMPLER_VIEW;
@@ -475,9 +480,9 @@ compress_with_blit(struct gl_context * ctx,
    /* Put user's tex data into the temporary texture
     */
    tex_xfer = pipe_get_transfer(st_context(ctx)->pipe, src_tex,
-					     0, 0, 0, /* face, level are zero */
-					     PIPE_TRANSFER_WRITE,
-					     0, 0, width, height); /* x, y, w, h */
+                                0, 0, /* layer, level are zero */
+                                PIPE_TRANSFER_WRITE,
+                                0, 0, width, height); /* x, y, w, h */
    map = pipe_transfer_map(pipe, tex_xfer);
 
    _mesa_texstore(ctx, 2, GL_RGBA, mesa_format,
@@ -549,14 +554,18 @@ st_TexImage(struct gl_context * ctx,
 
    /* switch to "normal" */
    if (stObj->surface_based) {
+      gl_format texFormat;
+
       _mesa_clear_texture_object(ctx, texObj);
       pipe_resource_reference(&stObj->pt, NULL);
 
       /* oops, need to init this image again */
+      texFormat = _mesa_choose_texture_format(ctx, texObj, target, level,
+                                              internalFormat, format, type);
+
       _mesa_init_teximage_fields(ctx, target, texImage,
-            width, height, depth, border, internalFormat);
-      _mesa_choose_texture_format(ctx, texObj, texImage, target, level,
-            internalFormat, format, type);
+                                 width, height, depth, border,
+                                 internalFormat, texFormat);
 
       stObj->surface_based = GL_FALSE;
    }
@@ -853,7 +862,6 @@ decompress_with_blit(struct gl_context * ctx, GLenum target, GLint level,
 {
    struct st_context *st = st_context(ctx);
    struct pipe_context *pipe = st->pipe;
-   struct pipe_screen *screen = pipe->screen;
    struct st_texture_image *stImage = st_texture_image(texImage);
    struct st_texture_object *stObj = st_texture_object(texObj);
    struct pipe_sampler_view *src_view =
@@ -867,7 +875,7 @@ decompress_with_blit(struct gl_context * ctx, GLenum target, GLint level,
 		    PIPE_BIND_TRANSFER_READ);
 
    /* create temp / dest surface */
-   if (!util_create_rgba_surface(screen, width, height, bind,
+   if (!util_create_rgba_surface(pipe, width, height, bind,
                                  &dst_texture, &dst_surface)) {
       _mesa_problem(ctx, "util_create_rgba_surface() failed "
                     "in decompress_with_blit()");
@@ -887,9 +895,9 @@ decompress_with_blit(struct gl_context * ctx, GLenum target, GLint level,
 
    /* map the dst_surface so we can read from it */
    tex_xfer = pipe_get_transfer(st_context(ctx)->pipe,
-					     dst_texture, 0, 0, 0,
-					     PIPE_TRANSFER_READ,
-					     0, 0, width, height);
+                                dst_texture, 0, 0,
+                                PIPE_TRANSFER_READ,
+                                0, 0, width, height);
 
    pixels = _mesa_map_pbo_dest(ctx, &ctx->Pack, pixels);
 
@@ -1306,7 +1314,7 @@ fallback_copy_texsubimage(struct gl_context *ctx, GLenum target, GLint level,
    struct pipe_transfer *src_trans;
    GLvoid *texDest;
    enum pipe_transfer_usage transfer_usage;
-   
+
    if (ST_DEBUG & DEBUG_FALLBACK)
       debug_printf("%s: fallback processing\n", __FUNCTION__);
 
@@ -1317,11 +1325,11 @@ fallback_copy_texsubimage(struct gl_context *ctx, GLenum target, GLint level,
    }
 
    src_trans = pipe_get_transfer(st_context(ctx)->pipe,
-					       strb->texture,
-					       0, 0, 0,
-					       PIPE_TRANSFER_READ,
-					       srcX, srcY,
-					       width, height);
+                                 strb->texture,
+                                 0, 0,
+                                 PIPE_TRANSFER_READ,
+                                 srcX, srcY,
+                                 width, height);
 
    if ((baseFormat == GL_DEPTH_COMPONENT ||
         baseFormat == GL_DEPTH_STENCIL) &&
@@ -1330,7 +1338,8 @@ fallback_copy_texsubimage(struct gl_context *ctx, GLenum target, GLint level,
    else
       transfer_usage = PIPE_TRANSFER_WRITE;
 
-   texDest = st_texture_image_map(st, stImage, 0, transfer_usage,
+   /* XXX this used to ignore destZ param */
+   texDest = st_texture_image_map(st, stImage, destZ, transfer_usage,
                                   destX, destY, width, height);
 
    if (baseFormat == GL_DEPTH_COMPONENT ||
@@ -1588,27 +1597,23 @@ st_copy_texsubimage(struct gl_context *ctx,
 
       if (matching_base_formats &&
           src_format == dest_format &&
-          !do_flip) 
+          !do_flip)
       {
          /* use surface_copy() / blit */
-         struct pipe_subresource subdst, subsrc;
-         subdst.face = stImage->face;
-         subdst.level = stImage->level;
-         subsrc.face = strb->surface->face;
-         subsrc.level = strb->surface->level;
+         struct pipe_box src_box;
+         u_box_2d_zslice(srcX, srcY, strb->surface->u.tex.first_layer,
+                         width, height, &src_box);
 
          /* for resource_copy_region(), y=0=top, always */
          pipe->resource_copy_region(pipe,
                                     /* dest */
                                     stImage->pt,
-                                    subdst,
-                                    destX, destY, destZ,
+                                    stImage->level,
+                                    destX, destY, destZ + stImage->face,
                                     /* src */
                                     strb->texture,
-                                    subsrc,
-                                    srcX, srcY, strb->surface->zslice,
-                                    /* size */
-                                    width, height);
+                                    strb->surface->u.tex.level,
+                                    &src_box);
          use_fallback = GL_FALSE;
       }
       else if (format_writemask &&
@@ -1624,12 +1629,16 @@ st_copy_texsubimage(struct gl_context *ctx,
                                            0)) {
          /* draw textured quad to do the copy */
          GLint srcY0, srcY1;
-         struct pipe_subresource subsrc;
+         struct pipe_surface surf_tmpl;
+         memset(&surf_tmpl, 0, sizeof(surf_tmpl));
+         surf_tmpl.format = stImage->pt->format;
+         surf_tmpl.usage = PIPE_BIND_RENDER_TARGET;
+         surf_tmpl.u.tex.level = stImage->level;
+         surf_tmpl.u.tex.first_layer = stImage->face + destZ;
+         surf_tmpl.u.tex.last_layer = stImage->face + destZ;
 
-         dest_surface = screen->get_tex_surface(screen, stImage->pt,
-                                                stImage->face, stImage->level,
-                                                destZ,
-                                                PIPE_BIND_RENDER_TARGET);
+         dest_surface = pipe->create_surface(pipe, stImage->pt,
+                                             &surf_tmpl);
 
          if (do_flip) {
             srcY1 = strb->Base.Height - srcY - height;
@@ -1639,15 +1648,13 @@ st_copy_texsubimage(struct gl_context *ctx,
             srcY0 = srcY;
             srcY1 = srcY0 + height;
          }
-         subsrc.face = strb->surface->face;
-         subsrc.level = strb->surface->level;
 
          util_blit_pixels_writemask(st->blit,
                                     strb->texture,
-                                    subsrc,
+                                    strb->surface->u.tex.level,
                                     srcX, srcY0,
                                     srcX + width, srcY1,
-                                    strb->surface->zslice,
+                                    strb->surface->u.tex.first_layer,
                                     dest_surface,
                                     destX, destY,
                                     destX + width, destY + height,
@@ -1848,8 +1855,9 @@ st_finalize_texture(struct gl_context *ctx,
     * will match.
     */
    if (firstImage->pt &&
+       stObj->pt &&
        firstImage->pt != stObj->pt &&
-       firstImage->pt->last_level >= stObj->lastLevel) {
+       firstImage->pt->last_level >= stObj->pt->last_level) {
       pipe_resource_reference(&stObj->pt, firstImage->pt);
       pipe_sampler_view_reference(&stObj->sampler_view, NULL);
    }
@@ -1950,7 +1958,7 @@ st_get_default_texture(struct st_context *st)
 
       _mesa_init_teximage_fields(st->ctx, target, texImg,
                                  16, 16, 1, 0,  /* w, h, d, border */
-                                 GL_RGBA);
+                                 GL_RGBA, MESA_FORMAT_RGBA8888);
 
       st_TexImage(st->ctx, 2, target,
                   0, GL_RGBA,    /* level, intformat */
diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c
index d0dcdd4e29b..21bb91f47a8 100644
--- a/src/mesa/state_tracker/st_context.c
+++ b/src/mesa/state_tracker/st_context.c
@@ -170,6 +170,11 @@ struct st_context *st_create_context(gl_api api, struct pipe_context *pipe,
    struct gl_context *shareCtx = share ? share->ctx : NULL;
    struct dd_function_table funcs;
 
+   /* Sanity checks */
+   assert(MESA_SHADER_VERTEX == PIPE_SHADER_VERTEX);
+   assert(MESA_SHADER_FRAGMENT == PIPE_SHADER_FRAGMENT);
+   assert(MESA_SHADER_GEOMETRY == PIPE_SHADER_GEOMETRY);
+
    memset(&funcs, 0, sizeof(funcs));
    st_init_driver_functions(&funcs);
 
@@ -239,14 +244,16 @@ void st_destroy_context( struct st_context *st )
    pipe->set_index_buffer(pipe, NULL);
 
    for (i = 0; i < PIPE_SHADER_TYPES; i++) {
-      pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, NULL);
-      pipe_resource_reference(&st->state.constants[PIPE_SHADER_VERTEX], NULL);
+      pipe->set_constant_buffer(pipe, i, 0, NULL);
+      pipe_resource_reference(&st->state.constants[i], NULL);
    }
 
    _mesa_delete_program_cache(st->ctx, st->pixel_xfer.cache);
 
    _vbo_DestroyContext(st->ctx);
 
+   st_destroy_program_variants(st);
+
    _mesa_free_context_data(ctx);
 
    st_destroy_context_priv(st);
diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h
index d342c0cff1e..68b52286d94 100644
--- a/src/mesa/state_tracker/st_context.h
+++ b/src/mesa/state_tracker/st_context.h
@@ -130,7 +130,9 @@ struct st_context
    struct st_fragment_program *fp;  /**< Currently bound fragment program */
    struct st_geometry_program *gp;  /**< Currently bound geometry program */
 
-   struct st_vp_varient *vp_varient;
+   struct st_vp_variant *vp_variant;
+   struct st_fp_variant *fp_variant;
+   struct st_gp_variant *gp_variant;
 
    struct gl_texture_object *default_texture;
 
@@ -160,7 +162,7 @@ struct st_context
 
    /** for glDraw/CopyPixels */
    struct {
-      struct st_fragment_program *shaders[4];
+      struct gl_fragment_program *shaders[4];
       void *vert_shaders[2];   /**< ureg shaders */
    } drawpix;
 
diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c
index df32491d044..e1e373d07d4 100644
--- a/src/mesa/state_tracker/st_debug.c
+++ b/src/mesa/state_tracker/st_debug.c
@@ -89,8 +89,8 @@ st_print_current(void)
    }
 #endif
 
-   if (st->vp->varients)
-      tgsi_dump( st->vp->varients[0].tgsi.tokens, 0 );
+   if (st->vp->variants)
+      tgsi_dump( st->vp->variants[0].tgsi.tokens, 0 );
    if (st->vp->Base.Base.Parameters)
       _mesa_print_parameter_list(st->vp->Base.Base.Parameters);
 
diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c
index 61a0e1b0877..f4bf1802390 100644
--- a/src/mesa/state_tracker/st_draw.c
+++ b/src/mesa/state_tracker/st_draw.c
@@ -241,7 +241,7 @@ st_pipe_vertex_format(GLenum type, GLuint size, GLenum format,
  */
 static GLboolean
 is_interleaved_arrays(const struct st_vertex_program *vp,
-                      const struct st_vp_varient *vpv,
+                      const struct st_vp_variant *vpv,
                       const struct gl_client_array **arrays,
                       GLboolean *userSpace)
 {
@@ -297,7 +297,7 @@ is_interleaved_arrays(const struct st_vertex_program *vp,
  */
 static void
 get_arrays_bounds(const struct st_vertex_program *vp,
-                  const struct st_vp_varient *vpv,
+                  const struct st_vp_variant *vpv,
                   const struct gl_client_array **arrays,
                   GLuint max_index,
                   const GLubyte **low, const GLubyte **high)
@@ -343,7 +343,7 @@ get_arrays_bounds(const struct st_vertex_program *vp,
 static void
 setup_interleaved_attribs(struct gl_context *ctx,
                           const struct st_vertex_program *vp,
-                          const struct st_vp_varient *vpv,
+                          const struct st_vp_variant *vpv,
                           const struct gl_client_array **arrays,
                           GLuint max_index,
                           GLboolean userSpace,
@@ -409,7 +409,7 @@ setup_interleaved_attribs(struct gl_context *ctx,
 static void
 setup_non_interleaved_attribs(struct gl_context *ctx,
                               const struct st_vertex_program *vp,
-                              const struct st_vp_varient *vpv,
+                              const struct st_vp_variant *vpv,
                               const struct gl_client_array **arrays,
                               GLuint max_index,
                               GLboolean *userSpace,
@@ -617,7 +617,7 @@ st_draw_vbo(struct gl_context *ctx,
    struct st_context *st = st_context(ctx);
    struct pipe_context *pipe = st->pipe;
    const struct st_vertex_program *vp;
-   const struct st_vp_varient *vpv;
+   const struct st_vp_variant *vpv;
    struct pipe_vertex_buffer vbuffer[PIPE_MAX_SHADER_INPUTS];
    GLuint attr;
    struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS];
@@ -650,7 +650,7 @@ st_draw_vbo(struct gl_context *ctx,
 
    /* must get these after state validation! */
    vp = st->vp;
-   vpv = st->vp_varient;
+   vpv = st->vp_variant;
 
 #if 0
    if (MESA_VERBOSE & VERBOSE_GLSL) {
diff --git a/src/mesa/state_tracker/st_draw_feedback.c b/src/mesa/state_tracker/st_draw_feedback.c
index 7f392fc4916..0a6cebb84cc 100644
--- a/src/mesa/state_tracker/st_draw_feedback.c
+++ b/src/mesa/state_tracker/st_draw_feedback.c
@@ -123,10 +123,10 @@ st_feedback_draw_vbo(struct gl_context *ctx,
 
    /* must get these after state validation! */
    vp = st->vp;
-   vs = &st->vp_varient->tgsi;
+   vs = &st->vp_variant->tgsi;
 
-   if (!st->vp_varient->draw_shader) {
-      st->vp_varient->draw_shader = draw_create_vertex_shader(draw, vs);
+   if (!st->vp_variant->draw_shader) {
+      st->vp_variant->draw_shader = draw_create_vertex_shader(draw, vs);
    }
 
    /*
@@ -139,7 +139,7 @@ st_feedback_draw_vbo(struct gl_context *ctx,
    draw_set_viewport_state(draw, &st->state.viewport);
    draw_set_clip_state(draw, &st->state.clip);
    draw_set_rasterizer_state(draw, &st->state.rasterizer, NULL);
-   draw_bind_vertex_shader(draw, st->vp_varient->draw_shader);
+   draw_bind_vertex_shader(draw, st->vp_variant->draw_shader);
    set_feedback_vertex_format(ctx);
 
    /* loop over TGSI shader inputs to determine vertex buffer
diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c
index 132749130af..62c9ce7273d 100644
--- a/src/mesa/state_tracker/st_extensions.c
+++ b/src/mesa/state_tracker/st_extensions.c
@@ -67,7 +67,7 @@ void st_init_limits(struct st_context *st)
 {
    struct pipe_screen *screen = st->pipe->screen;
    struct gl_constants *c = &st->ctx->Const;
-   unsigned i;
+   gl_shader_type sh;
 
    c->MaxTextureLevels
       = _min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS),
@@ -137,11 +137,12 @@ void st_init_limits(struct st_context *st)
    /* Quads always follow GL provoking rules. */
    c->QuadsFollowProvokingVertexConvention = GL_FALSE;
 
-   for(i = 0; i < MESA_SHADER_TYPES; ++i) {
-      struct gl_shader_compiler_options *options = &st->ctx->ShaderCompilerOptions[i];
+   for (sh = 0; sh < MESA_SHADER_TYPES; ++sh) {
+      struct gl_shader_compiler_options *options =
+         &st->ctx->ShaderCompilerOptions[sh];
       struct gl_program_constants *pc;
-      switch(i)
-      {
+
+      switch (sh) {
       case PIPE_SHADER_FRAGMENT:
          pc = &c->FragmentProgram;
          break;
@@ -156,36 +157,37 @@ void st_init_limits(struct st_context *st)
          continue;
       }
 
-      pc->MaxNativeInstructions    = screen->get_shader_param(screen, i, PIPE_SHADER_CAP_MAX_INSTRUCTIONS);
-      pc->MaxNativeAluInstructions = screen->get_shader_param(screen, i, PIPE_SHADER_CAP_MAX_ALU_INSTRUCTIONS);
-      pc->MaxNativeTexInstructions = screen->get_shader_param(screen, i, PIPE_SHADER_CAP_MAX_TEX_INSTRUCTIONS);
-      pc->MaxNativeTexIndirections = screen->get_shader_param(screen, i, PIPE_SHADER_CAP_MAX_TEX_INDIRECTIONS);
-      pc->MaxNativeAttribs         = screen->get_shader_param(screen, i, PIPE_SHADER_CAP_MAX_INPUTS);
-      pc->MaxNativeTemps           = screen->get_shader_param(screen, i, PIPE_SHADER_CAP_MAX_TEMPS);
-      pc->MaxNativeAddressRegs     = screen->get_shader_param(screen, i, PIPE_SHADER_CAP_MAX_ADDRS);
-      pc->MaxNativeParameters      = screen->get_shader_param(screen, i, PIPE_SHADER_CAP_MAX_CONSTS);
+      pc->MaxNativeInstructions    = screen->get_shader_param(screen, sh, PIPE_SHADER_CAP_MAX_INSTRUCTIONS);
+      pc->MaxNativeAluInstructions = screen->get_shader_param(screen, sh, PIPE_SHADER_CAP_MAX_ALU_INSTRUCTIONS);
+      pc->MaxNativeTexInstructions = screen->get_shader_param(screen, sh, PIPE_SHADER_CAP_MAX_TEX_INSTRUCTIONS);
+      pc->MaxNativeTexIndirections = screen->get_shader_param(screen, sh, PIPE_SHADER_CAP_MAX_TEX_INDIRECTIONS);
+      pc->MaxNativeAttribs         = screen->get_shader_param(screen, sh, PIPE_SHADER_CAP_MAX_INPUTS);
+      pc->MaxNativeTemps           = screen->get_shader_param(screen, sh, PIPE_SHADER_CAP_MAX_TEMPS);
+      pc->MaxNativeAddressRegs     = screen->get_shader_param(screen, sh, PIPE_SHADER_CAP_MAX_ADDRS);
+      pc->MaxNativeParameters      = screen->get_shader_param(screen, sh, PIPE_SHADER_CAP_MAX_CONSTS);
+      pc->MaxUniformComponents     = 4 * MIN2(pc->MaxNativeParameters, MAX_UNIFORMS);
 
       options->EmitNoNoise = TRUE;
 
       /* TODO: make these more fine-grained if anyone needs it */
-      options->EmitNoIfs = !screen->get_shader_param(screen, i, PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH);
-      options->EmitNoFunctions = !screen->get_shader_param(screen, i, PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH);
-      options->EmitNoLoops = !screen->get_shader_param(screen, i, PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH);
-      options->EmitNoMainReturn = !screen->get_shader_param(screen, i, PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH);
+      options->EmitNoIfs = !screen->get_shader_param(screen, sh, PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH);
+      options->EmitNoLoops = !screen->get_shader_param(screen, sh, PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH);
+      options->EmitNoFunctions = !screen->get_shader_param(screen, sh, PIPE_SHADER_CAP_SUBROUTINES);
+      options->EmitNoMainReturn = !screen->get_shader_param(screen, sh, PIPE_SHADER_CAP_SUBROUTINES);
 
-      options->EmitNoCont = !screen->get_shader_param(screen, i, PIPE_SHADER_CAP_TGSI_CONT_SUPPORTED);
+      options->EmitNoCont = !screen->get_shader_param(screen, sh, PIPE_SHADER_CAP_TGSI_CONT_SUPPORTED);
 
-      options->EmitNoIndirectInput = !screen->get_shader_param(screen, i,
+      options->EmitNoIndirectInput = !screen->get_shader_param(screen, sh,
                                         PIPE_SHADER_CAP_INDIRECT_INPUT_ADDR);
-      options->EmitNoIndirectOutput = !screen->get_shader_param(screen, i,
+      options->EmitNoIndirectOutput = !screen->get_shader_param(screen, sh,
                                         PIPE_SHADER_CAP_INDIRECT_OUTPUT_ADDR);
-      options->EmitNoIndirectTemp = !screen->get_shader_param(screen, i,
+      options->EmitNoIndirectTemp = !screen->get_shader_param(screen, sh,
                                         PIPE_SHADER_CAP_INDIRECT_TEMP_ADDR);
-      options->EmitNoIndirectUniform = !screen->get_shader_param(screen, i,
+      options->EmitNoIndirectUniform = !screen->get_shader_param(screen, sh,
                                         PIPE_SHADER_CAP_INDIRECT_CONST_ADDR);
 
       if(options->EmitNoLoops)
-         options->MaxUnrollIterations = MIN2(screen->get_shader_param(screen, i, PIPE_SHADER_CAP_MAX_INSTRUCTIONS), 65536);
+         options->MaxUnrollIterations = MIN2(screen->get_shader_param(screen, sh, PIPE_SHADER_CAP_MAX_INSTRUCTIONS), 65536);
    }
 
    /* PIPE_CAP_MAX_FS_INPUTS specifies the number of COLORn + GENERICn inputs
@@ -298,6 +300,8 @@ void st_init_extensions(struct st_context *st)
       ctx->Extensions.ARB_vertex_shader = GL_TRUE;
       ctx->Extensions.ARB_shader_objects = GL_TRUE;
       ctx->Extensions.ARB_shading_language_100 = GL_TRUE;
+      ctx->Extensions.ARB_explicit_attrib_location = GL_TRUE;
+      ctx->Extensions.EXT_separate_shader_objects = GL_TRUE;
    }
 
    if (screen->get_param(screen, PIPE_CAP_TEXTURE_MIRROR_REPEAT) > 0) {
@@ -339,6 +343,7 @@ void st_init_extensions(struct st_context *st)
 
    if (screen->get_param(screen, PIPE_CAP_OCCLUSION_QUERY)) {
       ctx->Extensions.ARB_occlusion_query = GL_TRUE;
+      ctx->Extensions.ARB_occlusion_query2 = GL_TRUE;
    }
    if (screen->get_param(screen, PIPE_CAP_TIMER_QUERY)) {
      ctx->Extensions.EXT_timer_query = GL_TRUE;
@@ -438,7 +443,9 @@ void st_init_extensions(struct st_context *st)
 #endif
 
    if (screen->get_shader_param(screen, PIPE_SHADER_GEOMETRY, PIPE_SHADER_CAP_MAX_INSTRUCTIONS) > 0) {
+#if 0 /* XXX re-enable when GLSL compiler again supports geometry shaders */
       ctx->Extensions.ARB_geometry_shader4 = GL_TRUE;
+#endif
    }
 
    if (screen->get_param(screen, PIPE_CAP_PRIMITIVE_RESTART)) {
diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c
index fe31418ddd1..c5f6008a222 100644
--- a/src/mesa/state_tracker/st_gen_mipmap.c
+++ b/src/mesa/state_tracker/st_gen_mipmap.c
@@ -29,7 +29,6 @@
 #include "main/imports.h"
 #include "main/mipmap.h"
 #include "main/teximage.h"
-#include "main/texformat.h"
 
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
@@ -80,11 +79,15 @@ st_render_mipmap(struct st_context *st,
    const uint face = _mesa_tex_target_to_face(target);
 
    assert(psv->texture == stObj->pt);
-   assert(target != GL_TEXTURE_3D); /* not done yet */
+#if 0
+   assert(target != GL_TEXTURE_3D); /* implemented but untested */
+#endif
 
    /* check if we can render in the texture's format */
-   if (!screen->is_format_supported(screen, psv->format, psv->texture->target, 0,
-                                    PIPE_BIND_RENDER_TARGET, 0)) {
+   /* XXX should probably kill this and always use util_gen_mipmap
+      since this implements a sw fallback as well */
+   if (!screen->is_format_supported(screen, psv->format, psv->texture->target,
+                                    0, PIPE_BIND_RENDER_TARGET, 0)) {
       return FALSE;
    }
 
@@ -162,12 +165,12 @@ fallback_generate_mipmap(struct gl_context *ctx, GLenum target,
    struct pipe_resource *pt = st_get_texobj_resource(texObj);
    const uint baseLevel = texObj->BaseLevel;
    const uint lastLevel = pt->last_level;
-   const uint face = _mesa_tex_target_to_face(target), zslice = 0;
+   const uint face = _mesa_tex_target_to_face(target);
    uint dstLevel;
    GLenum datatype;
    GLuint comps;
    GLboolean compressed;
-   
+
    if (ST_DEBUG & DEBUG_FALLBACK)
       debug_printf("%s: fallback processing\n", __FUNCTION__);
 
@@ -199,16 +202,15 @@ fallback_generate_mipmap(struct gl_context *ctx, GLenum target,
       ubyte *dstData;
       int srcStride, dstStride;
 
-      srcTrans = pipe_get_transfer(st_context(ctx)->pipe, pt, face,
-						srcLevel, zslice,
-						PIPE_TRANSFER_READ, 0, 0,
-                                                srcWidth, srcHeight);
-						
+      srcTrans = pipe_get_transfer(st_context(ctx)->pipe, pt, srcLevel,
+                                   face,
+                                   PIPE_TRANSFER_READ, 0, 0,
+                                   srcWidth, srcHeight);
 
-      dstTrans = pipe_get_transfer(st_context(ctx)->pipe, pt, face,
-						dstLevel, zslice,
-						PIPE_TRANSFER_WRITE, 0, 0,
-						dstWidth, dstHeight);
+      dstTrans = pipe_get_transfer(st_context(ctx)->pipe, pt, dstLevel,
+                                   face,
+                                   PIPE_TRANSFER_WRITE, 0, 0,
+                                   dstWidth, dstHeight);
 
       srcData = (ubyte *) pipe_transfer_map(pipe, srcTrans);
       dstData = (ubyte *) pipe_transfer_map(pipe, dstTrans);
@@ -216,6 +218,8 @@ fallback_generate_mipmap(struct gl_context *ctx, GLenum target,
       srcStride = srcTrans->stride / util_format_get_blocksize(srcTrans->resource->format);
       dstStride = dstTrans->stride / util_format_get_blocksize(dstTrans->resource->format);
 
+     /* this cannot work correctly for 3d since it does
+        not respect layerStride. */
       if (compressed) {
          const enum pipe_format format = pt->format;
          const uint bw = util_format_get_blockwidth(format);
@@ -366,6 +370,12 @@ st_generate_mipmap(struct gl_context *ctx, GLenum target,
 
       pt = stObj->pt;
    }
+   else {
+      /* Make sure that the base texture image data is present in the
+       * texture buffer.
+       */
+      st_finalize_texture(ctx, st->pipe, texObj);
+   }
 
    assert(pt->last_level >= lastLevel);
 
@@ -373,6 +383,8 @@ st_generate_mipmap(struct gl_context *ctx, GLenum target,
     * use the software fallback.
     */
    if (!st_render_mipmap(st, target, stObj, baseLevel, lastLevel)) {
+      /* since the util code actually also has a fallback, should
+         probably make it never fail and kill this */
       fallback_generate_mipmap(ctx, target, texObj);
    }
 
@@ -400,9 +412,8 @@ st_generate_mipmap(struct gl_context *ctx, GLenum target,
 
       /* initialize new image */
       _mesa_init_teximage_fields(ctx, target, dstImage, dstWidth, dstHeight,
-                                 dstDepth, border, srcImage->InternalFormat);
-
-      dstImage->TexFormat = srcImage->TexFormat;
+                                 dstDepth, border, srcImage->InternalFormat,
+                                 srcImage->TexFormat);
 
       stImage = st_texture_image(dstImage);
       stImage->level = dstLevel;
diff --git a/src/mesa/state_tracker/st_manager.c b/src/mesa/state_tracker/st_manager.c
index 183477a3f31..0307b48978b 100644
--- a/src/mesa/state_tracker/st_manager.c
+++ b/src/mesa/state_tracker/st_manager.c
@@ -34,13 +34,13 @@
 #include "util/u_pointer.h"
 #include "util/u_inlines.h"
 #include "util/u_atomic.h"
+#include "util/u_surface.h"
 
 #include "main/mtypes.h"
 #include "main/context.h"
 #include "main/texobj.h"
 #include "main/teximage.h"
 #include "main/texstate.h"
-#include "main/texfetch.h"
 #include "main/framebuffer.h"
 #include "main/fbobject.h"
 #include "main/renderbuffer.h"
@@ -143,7 +143,7 @@ buffer_index_to_attachment(gl_buffer_index index)
 static void
 st_framebuffer_validate(struct st_framebuffer *stfb, struct st_context *st)
 {
-   struct pipe_screen *screen = st->pipe->screen;
+   struct pipe_context *pipe = st->pipe;
    struct pipe_resource *textures[ST_ATTACHMENT_COUNT];
    uint width, height;
    unsigned i;
@@ -161,7 +161,7 @@ st_framebuffer_validate(struct st_framebuffer *stfb, struct st_context *st)
 
    for (i = 0; i < stfb->num_statts; i++) {
       struct st_renderbuffer *strb;
-      struct pipe_surface *ps;
+      struct pipe_surface *ps, surf_tmpl;
       gl_buffer_index idx;
 
       if (!textures[i])
@@ -180,8 +180,10 @@ st_framebuffer_validate(struct st_framebuffer *stfb, struct st_context *st)
          continue;
       }
 
-      ps = screen->get_tex_surface(screen, textures[i], 0, 0, 0,
-				   PIPE_BIND_RENDER_TARGET);
+      memset(&surf_tmpl, 0, sizeof(surf_tmpl));
+      u_surface_default_template(&surf_tmpl, textures[i],
+                                 PIPE_BIND_RENDER_TARGET);
+      ps = pipe->create_surface(pipe, textures[i], &surf_tmpl);
       if (ps) {
          pipe_surface_reference(&strb->surface, ps);
          pipe_resource_reference(&strb->texture, ps->texture);
@@ -556,6 +558,8 @@ st_context_teximage(struct st_context_iface *stctxi, enum st_texture_type target
    texImage = _mesa_get_tex_image(ctx, texObj, target, level);
    stImage = st_texture_image(texImage);
    if (tex) {
+      gl_format texFormat;
+
       /*
        * XXX When internal_format and tex->format differ, st_finalize_texture
        * needs to allocate a new texture with internal_format and copy the
@@ -573,11 +577,13 @@ st_context_teximage(struct st_context_iface *stctxi, enum st_texture_type target
          internalFormat = GL_RGBA;
       else
          internalFormat = GL_RGB;
+
+      texFormat = st_ChooseTextureFormat(ctx, internalFormat,
+                                         GL_RGBA, GL_UNSIGNED_BYTE);
+
       _mesa_init_teximage_fields(ctx, target, texImage,
-            tex->width0, tex->height0, 1, 0, internalFormat);
-      texImage->TexFormat = st_ChooseTextureFormat(ctx, internalFormat,
-            GL_RGBA, GL_UNSIGNED_BYTE);
-      _mesa_set_fetch_functions(texImage, 2);
+                                 tex->width0, tex->height0, 1, 0,
+                                 internalFormat, texFormat);
 
       width = tex->width0;
       height = tex->height0;
@@ -611,6 +617,26 @@ st_context_teximage(struct st_context_iface *stctxi, enum st_texture_type target
 }
 
 static void
+st_context_copy(struct st_context_iface *stctxi,
+                struct st_context_iface *stsrci, unsigned mask)
+{
+   struct st_context *st = (struct st_context *) stctxi;
+   struct st_context *src = (struct st_context *) stsrci;
+
+   _mesa_copy_context(src->ctx, st->ctx, mask);
+}
+
+static boolean
+st_context_share(struct st_context_iface *stctxi,
+                 struct st_context_iface *stsrci)
+{
+   struct st_context *st = (struct st_context *) stctxi;
+   struct st_context *src = (struct st_context *) stsrci;
+
+   return _mesa_share_state(st->ctx, src->ctx);
+}
+
+static void
 st_context_destroy(struct st_context_iface *stctxi)
 {
    struct st_context *st = (struct st_context *) stctxi;
@@ -677,7 +703,8 @@ st_api_create_context(struct st_api *stapi, struct st_manager *smapi,
       st_context_notify_invalid_framebuffer;
    st->iface.flush = st_context_flush;
    st->iface.teximage = st_context_teximage;
-   st->iface.copy = NULL;
+   st->iface.copy = st_context_copy;
+   st->iface.share = st_context_share;
    st->iface.st_context_private = (void *) smapi;
 
    return &st->iface;
@@ -789,6 +816,7 @@ st_manager_flush_frontbuffer(struct st_context *st)
 
 /**
  * Return the surface of an EGLImage.
+ * FIXME: I think this should operate on resources, not surfaces
  */
 struct pipe_surface *
 st_manager_get_egl_image_surface(struct st_context *st,
@@ -797,7 +825,7 @@ st_manager_get_egl_image_surface(struct st_context *st,
    struct st_manager *smapi =
       (struct st_manager *) st->iface.st_context_private;
    struct st_egl_image stimg;
-   struct pipe_surface *ps;
+   struct pipe_surface *ps, surf_tmpl;
 
    if (!smapi || !smapi->get_egl_image)
       return NULL;
@@ -806,8 +834,13 @@ st_manager_get_egl_image_surface(struct st_context *st,
    if (!smapi->get_egl_image(smapi, eglimg, &stimg))
       return NULL;
 
-   ps = smapi->screen->get_tex_surface(smapi->screen,
-         stimg.texture, stimg.face, stimg.level, stimg.zslice, usage);
+   memset(&surf_tmpl, 0, sizeof(surf_tmpl));
+   surf_tmpl.format = stimg.texture->format;
+   surf_tmpl.usage = usage;
+   surf_tmpl.u.tex.level = stimg.level;
+   surf_tmpl.u.tex.first_layer = stimg.layer;
+   surf_tmpl.u.tex.last_layer = stimg.layer;
+   ps = st->pipe->create_surface(st->pipe, stimg.texture, &surf_tmpl);
    pipe_resource_reference(&stimg.texture, NULL);
 
    return ps;
@@ -865,6 +898,7 @@ st_manager_add_color_renderbuffer(struct st_context *st, struct gl_framebuffer *
 }
 
 static const struct st_api st_gl_api = {
+   "Mesa " MESA_VERSION_STRING,
    ST_API_OPENGL,
 #if FEATURE_GL
    ST_PROFILE_DEFAULT_MASK |
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c
index c5c239b2c95..f848462310e 100644
--- a/src/mesa/state_tracker/st_mesa_to_tgsi.c
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c
@@ -760,10 +760,13 @@ emit_adjusted_wpos( struct st_translate *t,
 
 /**
  * Emit the TGSI instructions for inverting the WPOS y coordinate.
+ * This code is unavoidable because it also depends on whether
+ * a FBO is bound (STATE_FB_WPOS_Y_TRANSFORM).
  */
 static void
-emit_inverted_wpos( struct st_translate *t,
-                    const struct gl_program *program )
+emit_wpos_inversion( struct st_translate *t,
+                     const struct gl_program *program,
+                     boolean invert)
 {
    struct ureg_program *ureg = t->ureg;
 
@@ -771,17 +774,17 @@ emit_inverted_wpos( struct st_translate *t,
     * Need to replace instances of INPUT[WPOS] with temp T
     * where T = INPUT[WPOS] by y is inverted.
     */
-   static const gl_state_index winSizeState[STATE_LENGTH]
-      = { STATE_INTERNAL, STATE_FB_SIZE, 0, 0, 0 };
+   static const gl_state_index wposTransformState[STATE_LENGTH]
+      = { STATE_INTERNAL, STATE_FB_WPOS_Y_TRANSFORM, 0, 0, 0 };
    
    /* XXX: note we are modifying the incoming shader here!  Need to
     * do this before emitting the constant decls below, or this
     * will be missed:
     */
-   unsigned winHeightConst = _mesa_add_state_reference(program->Parameters,
-                                                       winSizeState);
+   unsigned wposTransConst = _mesa_add_state_reference(program->Parameters,
+                                                       wposTransformState);
 
-   struct ureg_src winsize = ureg_DECL_constant( ureg, winHeightConst );
+   struct ureg_src wpostrans = ureg_DECL_constant( ureg, wposTransConst );
    struct ureg_dst wpos_temp;
    struct ureg_src wpos_input = t->inputs[t->inputMapping[FRAG_ATTRIB_WPOS]];
 
@@ -794,12 +797,23 @@ emit_inverted_wpos( struct st_translate *t,
       ureg_MOV( ureg, wpos_temp, wpos_input );
    }
 
-   /* SUB wpos_temp.y, winsize_const, wpos_input
-    */
-   ureg_SUB( ureg,
-             ureg_writemask(wpos_temp, TGSI_WRITEMASK_Y ),
-             winsize,
-             wpos_input);
+   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));
+   } 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));
+   }
 
    /* Use wpos_temp as position input from here on:
     */
@@ -861,8 +875,7 @@ emit_wpos(struct st_context *st,
 
    /* we invert after adjustment so that we avoid the MOV to temporary,
     * and reuse the adjustment ADD instead */
-   if (invert)
-      emit_inverted_wpos(t, program);
+   emit_wpos_inversion(t, program, invert);
 }
 
 
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
index 76799287fe1..cfdc96b9dbe 100644
--- a/src/mesa/state_tracker/st_program.c
+++ b/src/mesa/state_tracker/st_program.c
@@ -32,7 +32,9 @@
 
 
 #include "main/imports.h"
+#include "main/hash.h"
 #include "main/mtypes.h"
+#include "program/prog_parameter.h"
 #include "program/prog_print.h"
 #include "program/programopt.h"
 
@@ -44,6 +46,8 @@
 #include "tgsi/tgsi_ureg.h"
 
 #include "st_debug.h"
+#include "st_cb_bitmap.h"
+#include "st_cb_drawpixels.h"
 #include "st_context.h"
 #include "st_program.h"
 #include "st_mesa_to_tgsi.h"
@@ -52,34 +56,109 @@
 
 
 /**
+ * Delete a vertex program variant.  Note the caller must unlink
+ * the variant from the linked list.
+ */
+static void
+delete_vp_variant(struct st_context *st, struct st_vp_variant *vpv)
+{
+   if (vpv->driver_shader) 
+      cso_delete_vertex_shader(st->cso_context, vpv->driver_shader);
+      
+#if FEATURE_feedback || FEATURE_rastpos
+   if (vpv->draw_shader)
+      draw_delete_vertex_shader( st->draw, vpv->draw_shader );
+#endif
+      
+   if (vpv->tgsi.tokens)
+      st_free_tokens(vpv->tgsi.tokens);
+      
+   FREE( vpv );
+}
+
+
+
+/**
  * Clean out any old compilations:
  */
 void
-st_vp_release_varients( struct st_context *st,
+st_release_vp_variants( struct st_context *st,
                         struct st_vertex_program *stvp )
 {
-   struct st_vp_varient *vpv;
+   struct st_vp_variant *vpv;
 
-   for (vpv = stvp->varients; vpv; ) {
-      struct st_vp_varient *next = vpv->next;
+   for (vpv = stvp->variants; vpv; ) {
+      struct st_vp_variant *next = vpv->next;
+      delete_vp_variant(st, vpv);
+      vpv = next;
+   }
 
-      if (vpv->driver_shader) 
-         cso_delete_vertex_shader(st->cso_context, vpv->driver_shader);
-      
-#if FEATURE_feedback || FEATURE_rastpos
-      if (vpv->draw_shader)
-         draw_delete_vertex_shader( st->draw, vpv->draw_shader );
-#endif
+   stvp->variants = NULL;
+}
+
+
+
+/**
+ * Delete a fragment program variant.  Note the caller must unlink
+ * the variant from the linked list.
+ */
+static void
+delete_fp_variant(struct st_context *st, struct st_fp_variant *fpv)
+{
+   if (fpv->driver_shader) 
+      cso_delete_fragment_shader(st->cso_context, fpv->driver_shader);
       
-      if (vpv->tgsi.tokens)
-         st_free_tokens(vpv->tgsi.tokens);
+   FREE(fpv);
+}
+
+
+/**
+ * Free all variants of a fragment program.
+ */
+void
+st_release_fp_variants(struct st_context *st, struct st_fragment_program *stfp)
+{
+   struct st_fp_variant *fpv;
+
+   for (fpv = stfp->variants; fpv; ) {
+      struct st_fp_variant *next = fpv->next;
+      delete_fp_variant(st, fpv);
+      fpv = next;
+   }
+
+   stfp->variants = NULL;
+}
+
+
+/**
+ * Delete a geometry program variant.  Note the caller must unlink
+ * the variant from the linked list.
+ */
+static void
+delete_gp_variant(struct st_context *st, struct st_gp_variant *gpv)
+{
+   if (gpv->driver_shader) 
+      cso_delete_geometry_shader(st->cso_context, gpv->driver_shader);
       
-      FREE( vpv );
+   FREE(gpv);
+}
 
-      vpv = next;
+
+/**
+ * Free all variants of a geometry program.
+ */
+void
+st_release_gp_variants(struct st_context *st, struct st_geometry_program *stgp)
+{
+   struct st_gp_variant *gpv;
+
+   for (gpv = stgp->variants; gpv; ) {
+      struct st_gp_variant *next = gpv->next;
+      delete_gp_variant(st, gpv);
+      gpv = next;
    }
 
-   stvp->varients = NULL;
+   stgp->variants = NULL;
 }
 
 
@@ -92,7 +171,7 @@ st_vp_release_varients( struct st_context *st,
  * \param tokensOut  destination for TGSI tokens
  * \return  pointer to cached pipe_shader object.
  */
-void
+static void
 st_prepare_vertex_program(struct st_context *st,
                             struct st_vertex_program *stvp)
 {
@@ -196,17 +275,22 @@ st_prepare_vertex_program(struct st_context *st,
 }
 
 
-struct st_vp_varient *
+/**
+ * Translate a vertex program to create a new variant.
+ */
+static struct st_vp_variant *
 st_translate_vertex_program(struct st_context *st,
                             struct st_vertex_program *stvp,
-                            const struct st_vp_varient_key *key)
+                            const struct st_vp_variant_key *key)
 {
-   struct st_vp_varient *vpv = CALLOC_STRUCT(st_vp_varient);
+   struct st_vp_variant *vpv = CALLOC_STRUCT(st_vp_variant);
    struct pipe_context *pipe = st->pipe;
    struct ureg_program *ureg;
    enum pipe_error error;
    unsigned num_outputs;
 
+   st_prepare_vertex_program( st, stvp );
+
    _mesa_remove_output_reads(&stvp->Base.Base, PROGRAM_OUTPUT);
    _mesa_remove_output_reads(&stvp->Base.Base, PROGRAM_VARYING);
 
@@ -216,6 +300,8 @@ st_translate_vertex_program(struct st_context *st,
       return NULL;
    }
 
+   vpv->key = *key;
+
    vpv->num_inputs = stvp->num_inputs;
    num_outputs = stvp->num_outputs;
    if (key->passthrough_edgeflags) {
@@ -229,23 +315,22 @@ st_translate_vertex_program(struct st_context *st,
       debug_printf("\n");
    }
 
-   error = 
-      st_translate_mesa_program(st->ctx,
-                                TGSI_PROCESSOR_VERTEX,
-                                ureg,
-                                &stvp->Base.Base,
-                                /* inputs */
-                                vpv->num_inputs,
-                                stvp->input_to_index,
-                                NULL, /* input semantic name */
-                                NULL, /* input semantic index */
-                                NULL,
-                                /* outputs */
-                                num_outputs,
-                                stvp->result_to_output,
-                                stvp->output_semantic_name,
-                                stvp->output_semantic_index,
-                                key->passthrough_edgeflags );
+   error = st_translate_mesa_program(st->ctx,
+                                     TGSI_PROCESSOR_VERTEX,
+                                     ureg,
+                                     &stvp->Base.Base,
+                                     /* inputs */
+                                     vpv->num_inputs,
+                                     stvp->input_to_index,
+                                     NULL, /* input semantic name */
+                                     NULL, /* input semantic index */
+                                     NULL,
+                                     /* outputs */
+                                     num_outputs,
+                                     stvp->result_to_output,
+                                     stvp->output_semantic_name,
+                                     stvp->output_semantic_index,
+                                     key->passthrough_edgeflags );
 
    if (error)
       goto fail;
@@ -275,201 +360,310 @@ fail:
 }
 
 
+/**
+ * Find/create a vertex program variant.
+ */
+struct st_vp_variant *
+st_get_vp_variant(struct st_context *st,
+                  struct st_vertex_program *stvp,
+                  const struct st_vp_variant_key *key)
+{
+   struct st_vp_variant *vpv;
+
+   /* Search for existing variant */
+   for (vpv = stvp->variants; vpv; vpv = vpv->next) {
+      if (memcmp(&vpv->key, key, sizeof(*key)) == 0) {
+         break;
+      }
+   }
+
+   if (!vpv) {
+      /* create now */
+      vpv = st_translate_vertex_program(st, stvp, key);
+      if (vpv) {
+         /* insert into list */
+         vpv->next = stvp->variants;
+         stvp->variants = vpv;
+      }
+   }
+
+   return vpv;
+}
+
 
 /**
- * Translate a Mesa fragment shader into a TGSI shader.
- * \return  pointer to cached pipe_shader object.
+ * Translate a Mesa fragment shader into a TGSI shader using extra info in
+ * the key.
+ * \return  new fragment program variant
  */
-void
+static struct st_fp_variant *
 st_translate_fragment_program(struct st_context *st,
-                              struct st_fragment_program *stfp )
+                              struct st_fragment_program *stfp,
+                              const struct st_fp_variant_key *key)
 {
    struct pipe_context *pipe = st->pipe;
-   GLuint outputMapping[FRAG_RESULT_MAX];
-   GLuint inputMapping[FRAG_ATTRIB_MAX];
-   GLuint interpMode[PIPE_MAX_SHADER_INPUTS];  /* XXX size? */
-   GLuint attr;
-   enum pipe_error error;
-   const GLbitfield inputsRead = stfp->Base.Base.InputsRead;
-   struct ureg_program *ureg;
+   struct st_fp_variant *variant = CALLOC_STRUCT(st_fp_variant);
 
-   ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS];
-   ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS];
-   uint fs_num_inputs = 0;
+   if (!variant)
+      return NULL;
 
-   ubyte fs_output_semantic_name[PIPE_MAX_SHADER_OUTPUTS];
-   ubyte fs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS];
-   uint fs_num_outputs = 0;
+   assert(!(key->bitmap && key->drawpixels));
 
-   _mesa_remove_output_reads(&stfp->Base.Base, PROGRAM_OUTPUT);
+   if (key->bitmap) {
+      /* glBitmap drawing */
+      struct gl_fragment_program *fp;
 
-   /*
-    * Convert Mesa program inputs to TGSI input register semantics.
-    */
-   for (attr = 0; attr < FRAG_ATTRIB_MAX; attr++) {
-      if (inputsRead & (1 << attr)) {
-         const GLuint slot = fs_num_inputs++;
+      st_make_bitmap_fragment_program(st, &stfp->Base,
+                                      &fp, &variant->bitmap_sampler);
 
-         inputMapping[attr] = slot;
+      variant->parameters = _mesa_clone_parameter_list(fp->Base.Parameters);
+      stfp = st_fragment_program(fp);
+   }
+   else if (key->drawpixels) {
+      /* glDrawPixels drawing */
+      struct gl_fragment_program *fp;
 
-         switch (attr) {
-         case FRAG_ATTRIB_WPOS:
-            input_semantic_name[slot] = TGSI_SEMANTIC_POSITION;
-            input_semantic_index[slot] = 0;
-            interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
-            break;
-         case FRAG_ATTRIB_COL0:
-            input_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
-            input_semantic_index[slot] = 0;
-            interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
-            break;
-         case FRAG_ATTRIB_COL1:
-            input_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
-            input_semantic_index[slot] = 1;
-            interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
-            break;
-         case FRAG_ATTRIB_FOGC:
-            input_semantic_name[slot] = TGSI_SEMANTIC_FOG;
-            input_semantic_index[slot] = 0;
-            interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
-            break;
-         case FRAG_ATTRIB_FACE:
-            input_semantic_name[slot] = TGSI_SEMANTIC_FACE;
-            input_semantic_index[slot] = 0;
-            interpMode[slot] = TGSI_INTERPOLATE_CONSTANT;
-            break;
-            /* In most cases, there is nothing special about these
-             * inputs, so adopt a convention to use the generic
-             * semantic name and the mesa FRAG_ATTRIB_ number as the
-             * index. 
-             * 
-             * All that is required is that the vertex shader labels
-             * its own outputs similarly, and that the vertex shader
-             * generates at least every output required by the
-             * fragment shader plus fixed-function hardware (such as
-             * BFC).
-             * 
-             * There is no requirement that semantic indexes start at
-             * zero or be restricted to a particular range -- nobody
-             * should be building tables based on semantic index.
-             */
-         case FRAG_ATTRIB_PNTC:
-         case FRAG_ATTRIB_TEX0:
-         case FRAG_ATTRIB_TEX1:
-         case FRAG_ATTRIB_TEX2:
-         case FRAG_ATTRIB_TEX3:
-         case FRAG_ATTRIB_TEX4:
-         case FRAG_ATTRIB_TEX5:
-         case FRAG_ATTRIB_TEX6:
-         case FRAG_ATTRIB_TEX7:
-         case FRAG_ATTRIB_VAR0:
-         default:
-            /* Actually, let's try and zero-base this just for
-             * readability of the generated TGSI.
-             */
-            assert(attr >= FRAG_ATTRIB_TEX0);
-            input_semantic_index[slot] = (attr - FRAG_ATTRIB_TEX0);
-            input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
-            if (attr == FRAG_ATTRIB_PNTC)
-               interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
-            else
-               interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
-            break;
-         }
+      if (key->drawpixels_z || key->drawpixels_stencil) {
+         fp = st_make_drawpix_z_stencil_program(st, key->drawpixels_z,
+                                                key->drawpixels_stencil);
       }
       else {
-	 inputMapping[attr] = -1;
+         /* RGBA */
+         st_make_drawpix_fragment_program(st, &stfp->Base, &fp);
+         variant->parameters = _mesa_clone_parameter_list(fp->Base.Parameters);
       }
+      stfp = st_fragment_program(fp);
    }
 
-   /*
-    * Semantics and mapping for outputs
-    */
-   {
-      uint numColors = 0;
-      GLbitfield64 outputsWritten = stfp->Base.Base.OutputsWritten;
-
-      /* if z is written, emit that first */
-      if (outputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) {
-         fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_POSITION;
-         fs_output_semantic_index[fs_num_outputs] = 0;
-         outputMapping[FRAG_RESULT_DEPTH] = fs_num_outputs;
-         fs_num_outputs++;
-         outputsWritten &= ~(1 << FRAG_RESULT_DEPTH);
-      }
+   if (!stfp->tgsi.tokens) {
+      /* need to translate Mesa instructions to TGSI now */
+      GLuint outputMapping[FRAG_RESULT_MAX];
+      GLuint inputMapping[FRAG_ATTRIB_MAX];
+      GLuint interpMode[PIPE_MAX_SHADER_INPUTS];  /* XXX size? */
+      GLuint attr;
+      enum pipe_error error;
+      const GLbitfield inputsRead = stfp->Base.Base.InputsRead;
+      struct ureg_program *ureg;
 
-      if (outputsWritten & BITFIELD64_BIT(FRAG_RESULT_STENCIL)) {
-         fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_STENCIL;
-         fs_output_semantic_index[fs_num_outputs] = 0;
-         outputMapping[FRAG_RESULT_STENCIL] = fs_num_outputs;
-         fs_num_outputs++;
-         outputsWritten &= ~(1 << FRAG_RESULT_STENCIL);
-      }
+      ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS];
+      ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS];
+      uint fs_num_inputs = 0;
+
+      ubyte fs_output_semantic_name[PIPE_MAX_SHADER_OUTPUTS];
+      ubyte fs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS];
+      uint fs_num_outputs = 0;
+
+
+      _mesa_remove_output_reads(&stfp->Base.Base, PROGRAM_OUTPUT);
+
+      /*
+       * Convert Mesa program inputs to TGSI input register semantics.
+       */
+      for (attr = 0; attr < FRAG_ATTRIB_MAX; attr++) {
+         if (inputsRead & (1 << attr)) {
+            const GLuint slot = fs_num_inputs++;
+
+            inputMapping[attr] = slot;
 
-      /* handle remaning outputs (color) */
-      for (attr = 0; attr < FRAG_RESULT_MAX; attr++) {
-         if (outputsWritten & BITFIELD64_BIT(attr)) {
             switch (attr) {
-            case FRAG_RESULT_DEPTH:
-            case FRAG_RESULT_STENCIL:
-               /* handled above */
-               assert(0);
+            case FRAG_ATTRIB_WPOS:
+               input_semantic_name[slot] = TGSI_SEMANTIC_POSITION;
+               input_semantic_index[slot] = 0;
+               interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
+               break;
+            case FRAG_ATTRIB_COL0:
+               input_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
+               input_semantic_index[slot] = 0;
+               interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
+               break;
+            case FRAG_ATTRIB_COL1:
+               input_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
+               input_semantic_index[slot] = 1;
+               interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
+               break;
+            case FRAG_ATTRIB_FOGC:
+               input_semantic_name[slot] = TGSI_SEMANTIC_FOG;
+               input_semantic_index[slot] = 0;
+               interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
+               break;
+            case FRAG_ATTRIB_FACE:
+               input_semantic_name[slot] = TGSI_SEMANTIC_FACE;
+               input_semantic_index[slot] = 0;
+               interpMode[slot] = TGSI_INTERPOLATE_CONSTANT;
                break;
+               /* In most cases, there is nothing special about these
+                * inputs, so adopt a convention to use the generic
+                * semantic name and the mesa FRAG_ATTRIB_ number as the
+                * index. 
+                * 
+                * All that is required is that the vertex shader labels
+                * its own outputs similarly, and that the vertex shader
+                * generates at least every output required by the
+                * fragment shader plus fixed-function hardware (such as
+                * BFC).
+                * 
+                * There is no requirement that semantic indexes start at
+                * zero or be restricted to a particular range -- nobody
+                * should be building tables based on semantic index.
+                */
+            case FRAG_ATTRIB_PNTC:
+            case FRAG_ATTRIB_TEX0:
+            case FRAG_ATTRIB_TEX1:
+            case FRAG_ATTRIB_TEX2:
+            case FRAG_ATTRIB_TEX3:
+            case FRAG_ATTRIB_TEX4:
+            case FRAG_ATTRIB_TEX5:
+            case FRAG_ATTRIB_TEX6:
+            case FRAG_ATTRIB_TEX7:
+            case FRAG_ATTRIB_VAR0:
             default:
-               assert(attr == FRAG_RESULT_COLOR ||
-                      (FRAG_RESULT_DATA0 <= attr && attr < FRAG_RESULT_MAX));
-               fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_COLOR;
-               fs_output_semantic_index[fs_num_outputs] = numColors;
-               outputMapping[attr] = fs_num_outputs;
-               numColors++;
+               /* Actually, let's try and zero-base this just for
+                * readability of the generated TGSI.
+                */
+               assert(attr >= FRAG_ATTRIB_TEX0);
+               input_semantic_index[slot] = (attr - FRAG_ATTRIB_TEX0);
+               input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+               if (attr == FRAG_ATTRIB_PNTC)
+                  interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
+               else
+                  interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
                break;
             }
+         }
+         else {
+            inputMapping[attr] = -1;
+         }
+      }
 
+      /*
+       * Semantics and mapping for outputs
+       */
+      {
+         uint numColors = 0;
+         GLbitfield64 outputsWritten = stfp->Base.Base.OutputsWritten;
+
+         /* if z is written, emit that first */
+         if (outputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) {
+            fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_POSITION;
+            fs_output_semantic_index[fs_num_outputs] = 0;
+            outputMapping[FRAG_RESULT_DEPTH] = fs_num_outputs;
             fs_num_outputs++;
+            outputsWritten &= ~(1 << FRAG_RESULT_DEPTH);
+         }
+
+         if (outputsWritten & BITFIELD64_BIT(FRAG_RESULT_STENCIL)) {
+            fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_STENCIL;
+            fs_output_semantic_index[fs_num_outputs] = 0;
+            outputMapping[FRAG_RESULT_STENCIL] = fs_num_outputs;
+            fs_num_outputs++;
+            outputsWritten &= ~(1 << FRAG_RESULT_STENCIL);
+         }
+
+         /* handle remaning outputs (color) */
+         for (attr = 0; attr < FRAG_RESULT_MAX; attr++) {
+            if (outputsWritten & BITFIELD64_BIT(attr)) {
+               switch (attr) {
+               case FRAG_RESULT_DEPTH:
+               case FRAG_RESULT_STENCIL:
+                  /* handled above */
+                  assert(0);
+                  break;
+               default:
+                  assert(attr == FRAG_RESULT_COLOR ||
+                         (FRAG_RESULT_DATA0 <= attr && attr < FRAG_RESULT_MAX));
+                  fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_COLOR;
+                  fs_output_semantic_index[fs_num_outputs] = numColors;
+                  outputMapping[attr] = fs_num_outputs;
+                  numColors++;
+                  break;
+               }
+
+               fs_num_outputs++;
+            }
          }
       }
-   }
 
-   ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT );
-   if (ureg == NULL)
-      return;
+      ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT );
+      if (ureg == NULL)
+         return NULL;
 
-   if (ST_DEBUG & DEBUG_MESA) {
-      _mesa_print_program(&stfp->Base.Base);
-      _mesa_print_program_parameters(st->ctx, &stfp->Base.Base);
-      debug_printf("\n");
+      if (ST_DEBUG & DEBUG_MESA) {
+         _mesa_print_program(&stfp->Base.Base);
+         _mesa_print_program_parameters(st->ctx, &stfp->Base.Base);
+         debug_printf("\n");
+      }
+
+      error = st_translate_mesa_program(st->ctx,
+                                        TGSI_PROCESSOR_FRAGMENT,
+                                        ureg,
+                                        &stfp->Base.Base,
+                                        /* inputs */
+                                        fs_num_inputs,
+                                        inputMapping,
+                                        input_semantic_name,
+                                        input_semantic_index,
+                                        interpMode,
+                                        /* outputs */
+                                        fs_num_outputs,
+                                        outputMapping,
+                                        fs_output_semantic_name,
+                                        fs_output_semantic_index, FALSE );
+
+      stfp->tgsi.tokens = ureg_get_tokens( ureg, NULL );
+      ureg_destroy( ureg );
    }
 
-   error = 
-      st_translate_mesa_program(st->ctx,
-                                TGSI_PROCESSOR_FRAGMENT,
-                                ureg,
-                                &stfp->Base.Base,
-                                /* inputs */
-                                fs_num_inputs,
-                                inputMapping,
-                                input_semantic_name,
-                                input_semantic_index,
-                                interpMode,
-                                /* outputs */
-                                fs_num_outputs,
-                                outputMapping,
-                                fs_output_semantic_name,
-                                fs_output_semantic_index, FALSE );
-
-   stfp->tgsi.tokens = ureg_get_tokens( ureg, NULL );
-   ureg_destroy( ureg );
-   stfp->driver_shader = pipe->create_fs_state(pipe, &stfp->tgsi);
+   /* fill in variant */
+   variant->driver_shader = pipe->create_fs_state(pipe, &stfp->tgsi);
+   variant->key = *key;
 
    if (ST_DEBUG & DEBUG_TGSI) {
       tgsi_dump( stfp->tgsi.tokens, 0/*TGSI_DUMP_VERBOSE*/ );
       debug_printf("\n");
    }
+
+   return variant;
 }
 
-void
+
+/**
+ * Translate fragment program if needed.
+ */
+struct st_fp_variant *
+st_get_fp_variant(struct st_context *st,
+                  struct st_fragment_program *stfp,
+                  const struct st_fp_variant_key *key)
+{
+   struct st_fp_variant *fpv;
+
+   /* Search for existing variant */
+   for (fpv = stfp->variants; fpv; fpv = fpv->next) {
+      if (memcmp(&fpv->key, key, sizeof(*key)) == 0) {
+         break;
+      }
+   }
+
+   if (!fpv) {
+      /* create new */
+      fpv = st_translate_fragment_program(st, stfp, key);
+      if (fpv) {
+         /* insert into list */
+         fpv->next = stfp->variants;
+         stfp->variants = fpv;
+      }
+   }
+
+   return fpv;
+}
+
+
+/**
+ * Translate a geometry program to create a new variant.
+ */
+static struct st_gp_variant *
 st_translate_geometry_program(struct st_context *st,
-                              struct st_geometry_program *stgp)
+                              struct st_geometry_program *stgp,
+                              const struct st_gp_variant_key *key)
 {
    GLuint inputMapping[GEOM_ATTRIB_MAX];
    GLuint outputMapping[GEOM_RESULT_MAX];
@@ -492,12 +686,19 @@ st_translate_geometry_program(struct st_context *st,
    GLuint maxSlot = 0;
    struct ureg_program *ureg;
 
+   struct st_gp_variant *gpv;
+
+   gpv = CALLOC_STRUCT(st_gp_variant);
+   if (!gpv)
+      return NULL;
+
    _mesa_remove_output_reads(&stgp->Base.Base, PROGRAM_OUTPUT);
    _mesa_remove_output_reads(&stgp->Base.Base, PROGRAM_VARYING);
 
    ureg = ureg_create( TGSI_PROCESSOR_GEOMETRY );
    if (ureg == NULL) {
-      return;
+      FREE(gpv);
+      return NULL;
    }
 
    /* which vertex output goes to the first geometry input */
@@ -527,7 +728,7 @@ st_translate_geometry_program(struct st_context *st,
          } else
             ++gs_builtin_inputs;
 
-#if 1
+#if 0
          debug_printf("input map at %d = %d\n",
                       slot + gs_array_offset, stgp->input_map[slot + gs_array_offset]);
 #endif
@@ -669,37 +870,35 @@ st_translate_geometry_program(struct st_context *st,
       st_free_tokens(stgp->tgsi.tokens);
       stgp->tgsi.tokens = NULL;
    }
-   if (stgp->driver_shader) {
-      cso_delete_geometry_shader(st->cso_context, stgp->driver_shader);
-      stgp->driver_shader = NULL;
-   }
 
    ureg_property_gs_input_prim(ureg, stgp->Base.InputType);
    ureg_property_gs_output_prim(ureg, stgp->Base.OutputType);
    ureg_property_gs_max_vertices(ureg, stgp->Base.VerticesOut);
 
-   error  = st_translate_mesa_program(st->ctx,
-                                      TGSI_PROCESSOR_GEOMETRY,
-                                      ureg,
-                                      &stgp->Base.Base,
-                                      /* inputs */
-                                      gs_num_inputs,
-                                      inputMapping,
-                                      stgp->input_semantic_name,
-                                      stgp->input_semantic_index,
-                                      NULL,
-                                      /* outputs */
-                                      gs_num_outputs,
-                                      outputMapping,
-                                      gs_output_semantic_name,
-                                      gs_output_semantic_index,
-                                      FALSE);
-
+   error = st_translate_mesa_program(st->ctx,
+                                     TGSI_PROCESSOR_GEOMETRY,
+                                     ureg,
+                                     &stgp->Base.Base,
+                                     /* inputs */
+                                     gs_num_inputs,
+                                     inputMapping,
+                                     stgp->input_semantic_name,
+                                     stgp->input_semantic_index,
+                                     NULL,
+                                     /* outputs */
+                                     gs_num_outputs,
+                                     outputMapping,
+                                     gs_output_semantic_name,
+                                     gs_output_semantic_index,
+                                     FALSE);
 
    stgp->num_inputs = gs_num_inputs;
    stgp->tgsi.tokens = ureg_get_tokens( ureg, NULL );
    ureg_destroy( ureg );
-   stgp->driver_shader = pipe->create_gs_state(pipe, &stgp->tgsi);
+
+   /* fill in new variant */
+   gpv->driver_shader = pipe->create_gs_state(pipe, &stgp->tgsi);
+   gpv->key = *key;
 
    if ((ST_DEBUG & DEBUG_TGSI) && (ST_DEBUG & DEBUG_MESA)) {
       _mesa_print_program(&stgp->Base.Base);
@@ -710,8 +909,44 @@ st_translate_geometry_program(struct st_context *st,
       tgsi_dump(stgp->tgsi.tokens, 0);
       debug_printf("\n");
    }
+
+   return gpv;
 }
 
+
+/**
+ * Get/create geometry program variant.
+ */
+struct st_gp_variant *
+st_get_gp_variant(struct st_context *st,
+                  struct st_geometry_program *stgp,
+                  const struct st_gp_variant_key *key)
+{
+   struct st_gp_variant *gpv;
+
+   /* Search for existing variant */
+   for (gpv = stgp->variants; gpv; gpv = gpv->next) {
+      if (memcmp(&gpv->key, key, sizeof(*key)) == 0) {
+         break;
+      }
+   }
+
+   if (!gpv) {
+      /* create new */
+      gpv = st_translate_geometry_program(st, stgp, key);
+      if (gpv) {
+         /* insert into list */
+         gpv->next = stgp->variants;
+         stgp->variants = gpv;
+      }
+   }
+
+   return gpv;
+}
+
+
+
+
 /**
  * Debug- print current shader text
  */
@@ -757,3 +992,155 @@ st_print_shaders(struct gl_context *ctx)
       }
    }
 }
+
+
+/**
+ * Vert/Geom/Frag programs have per-context variants.  Free all the
+ * variants attached to the given program which match the given context.
+ */
+static void
+destroy_program_variants(struct st_context *st, struct gl_program *program)
+{
+   if (!program)
+      return;
+
+   switch (program->Target) {
+   case GL_VERTEX_PROGRAM_ARB:
+      {
+         struct st_vertex_program *stvp = (struct st_vertex_program *) program;
+         struct st_vp_variant *vpv, **prevPtr = &stvp->variants;
+
+         for (vpv = stvp->variants; vpv; ) {
+            struct st_vp_variant *next = vpv->next;
+            if (vpv->key.st == st) {
+               /* unlink from list */
+               *prevPtr = next;
+               /* destroy this variant */
+               delete_vp_variant(st, vpv);
+            }
+            else {
+               prevPtr = &vpv->next;
+            }
+            vpv = next;
+         }
+      }
+      break;
+   case GL_FRAGMENT_PROGRAM_ARB:
+      {
+         struct st_fragment_program *stfp =
+            (struct st_fragment_program *) program;
+         struct st_fp_variant *fpv, **prevPtr = &stfp->variants;
+
+         for (fpv = stfp->variants; fpv; ) {
+            struct st_fp_variant *next = fpv->next;
+            if (fpv->key.st == st) {
+               /* unlink from list */
+               *prevPtr = next;
+               /* destroy this variant */
+               delete_fp_variant(st, fpv);
+            }
+            else {
+               prevPtr = &fpv->next;
+            }
+            fpv = next;
+         }
+      }
+      break;
+   case MESA_GEOMETRY_PROGRAM:
+      {
+         struct st_geometry_program *stgp =
+            (struct st_geometry_program *) program;
+         struct st_gp_variant *gpv, **prevPtr = &stgp->variants;
+
+         for (gpv = stgp->variants; gpv; ) {
+            struct st_gp_variant *next = gpv->next;
+            if (gpv->key.st == st) {
+               /* unlink from list */
+               *prevPtr = next;
+               /* destroy this variant */
+               delete_gp_variant(st, gpv);
+            }
+            else {
+               prevPtr = &gpv->next;
+            }
+            gpv = next;
+         }
+      }
+      break;
+   default:
+      _mesa_problem(NULL, "Unexpected program target in "
+                    "destroy_program_variants_cb()");
+   }
+}
+
+
+/**
+ * Callback for _mesa_HashWalk.  Free all the shader's program variants
+ * which match the given context.
+ */
+static void
+destroy_shader_program_variants_cb(GLuint key, void *data, void *userData)
+{
+   struct st_context *st = (struct st_context *) userData;
+   struct gl_shader *shader = (struct gl_shader *) data;
+
+   switch (shader->Type) {
+   case GL_SHADER_PROGRAM_MESA:
+      {
+         struct gl_shader_program *shProg = (struct gl_shader_program *) data;
+         GLuint i;
+
+         for (i = 0; i < shProg->NumShaders; i++) {
+            destroy_program_variants(st, shProg->Shaders[i]->Program);
+         }
+
+         destroy_program_variants(st, (struct gl_program *)
+                                  shProg->VertexProgram);
+         destroy_program_variants(st, (struct gl_program *)
+                                  shProg->FragmentProgram);
+         destroy_program_variants(st, (struct gl_program *)
+                                  shProg->GeometryProgram);
+      }
+      break;
+   case GL_VERTEX_SHADER:
+   case GL_FRAGMENT_SHADER:
+   case GL_GEOMETRY_SHADER:
+      {
+         destroy_program_variants(st, shader->Program);
+      }
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+/**
+ * Callback for _mesa_HashWalk.  Free all the program variants which match
+ * the given context.
+ */
+static void
+destroy_program_variants_cb(GLuint key, void *data, void *userData)
+{
+   struct st_context *st = (struct st_context *) userData;
+   struct gl_program *program = (struct gl_program *) data;
+   destroy_program_variants(st, program);
+}
+
+
+/**
+ * Walk over all shaders and programs to delete any variants which
+ * belong to the given context.
+ * This is called during context tear-down.
+ */
+void
+st_destroy_program_variants(struct st_context *st)
+{
+   /* ARB vert/frag program */
+   _mesa_HashWalk(st->ctx->Shared->Programs,
+                  destroy_program_variants_cb, st);
+
+   /* GLSL vert/frag/geom shaders */
+   _mesa_HashWalk(st->ctx->Shared->ShaderObjects,
+                  destroy_shader_program_variants_cb, st);
+}
diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h
index 72dbc715fe1..c4244df939e 100644
--- a/src/mesa/state_tracker/st_program.h
+++ b/src/mesa/state_tracker/st_program.h
@@ -40,26 +40,61 @@
 #include "st_context.h"
 
 
+/** Fragment program variant key */
+struct st_fp_variant_key
+{
+   struct st_context *st;         /**< variants are per-context */
+
+   /** for glBitmap */
+   GLuint bitmap:1;               /**< glBitmap variant? */
+
+   /** for glDrawPixels */
+   GLuint drawpixels:1;           /**< glDrawPixels variant */
+   GLuint scaleAndBias:1;         /**< glDrawPixels w/ scale and/or bias? */
+   GLuint pixelMaps:1;            /**< glDrawPixels w/ pixel lookup map? */
+   GLuint drawpixels_z:1;         /**< glDrawPixels(GL_DEPTH) */
+   GLuint drawpixels_stencil:1;   /**< glDrawPixels(GL_STENCIL) */
+};
+
+
+/**
+ * Variant of a fragment program.
+ */
+struct st_fp_variant
+{
+   /** Parameters which generated this version of fragment program */
+   struct st_fp_variant_key key;
+
+   /** Driver's compiled shader */
+   void *driver_shader;
+
+   /** For glBitmap variants */
+   struct gl_program_parameter_list *parameters;
+   uint bitmap_sampler;
+
+   /** next in linked list */
+   struct st_fp_variant *next;
+};
+
+
 /**
  * Derived from Mesa gl_fragment_program:
  */
 struct st_fragment_program
 {
    struct gl_fragment_program Base;
-   GLuint serialNo;
 
    struct pipe_shader_state tgsi;
-   void *driver_shader;
 
-   /** Program prefixed with glBitmap prologue */
-   struct st_fragment_program *bitmap_program;
-   uint bitmap_sampler;
+   struct st_fp_variant *variants;
 };
 
 
 
-struct st_vp_varient_key
+/** Vertex program variant key */
+struct st_vp_variant_key
 {
+   struct st_context *st;          /**< variants are per-context */
    boolean passthrough_edgeflags;
 };
 
@@ -68,12 +103,12 @@ struct st_vp_varient_key
  * This represents a vertex program, especially translated to match
  * the inputs of a particular fragment shader.
  */
-struct st_vp_varient
+struct st_vp_variant
 {
    /* Parameters which generated this translated version of a vertex
     * shader:
     */
-   struct st_vp_varient_key key;
+   struct st_vp_variant_key key;
 
    /**
     * TGSI tokens (to later generate a 'draw' module shader for
@@ -88,9 +123,9 @@ struct st_vp_varient
    struct draw_vertex_shader *draw_shader;
 
    /** Next in linked list */
-   struct st_vp_varient *next;  
+   struct st_vp_variant *next;  
 
-   /** similar to that in st_vertex_program, but with information about edgeflags too */
+   /** similar to that in st_vertex_program, but with edgeflags info too */
    GLuint num_inputs;
 };
 
@@ -101,7 +136,6 @@ struct st_vp_varient
 struct st_vertex_program
 {
    struct gl_vertex_program Base;  /**< The Mesa vertex program */
-   GLuint serialNo, lastSerialNo;
 
    /** maps a Mesa VERT_ATTRIB_x to a packed TGSI input index */
    GLuint input_to_index[VERT_ATTRIB_MAX];
@@ -115,18 +149,41 @@ struct st_vertex_program
    ubyte output_semantic_index[VERT_RESULT_MAX];
    GLuint num_outputs;
 
-   /** List of translated varients of this vertex program.
+   /** List of translated variants of this vertex program.
     */
-   struct st_vp_varient *varients;
+   struct st_vp_variant *variants;
+};
+
+
+
+/** Geometry program variant key */
+struct st_gp_variant_key
+{
+   struct st_context *st;          /**< variants are per-context */
+   /* no other fields yet */
+};
+
+
+/**
+ * Geometry program variant.
+ */
+struct st_gp_variant
+{
+   /* Parameters which generated this translated version of a vertex */
+   struct st_gp_variant_key key;
+
+   void *driver_shader;
+
+   struct st_gp_variant *next;
 };
 
+
 /**
  * Derived from Mesa gl_geometry_program:
  */
 struct st_geometry_program
 {
    struct gl_geometry_program Base;  /**< The Mesa geometry program */
-   GLuint serialNo;
 
    /** map GP input back to VP output */
    GLuint input_map[PIPE_MAX_SHADER_INPUTS];
@@ -145,9 +202,12 @@ struct st_geometry_program
    ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS];
 
    struct pipe_shader_state tgsi;
-   void *driver_shader;
+
+   struct st_gp_variant *variants;
 };
 
+
+
 static INLINE struct st_fragment_program *
 st_fragment_program( struct gl_fragment_program *fp )
 {
@@ -162,9 +222,9 @@ st_vertex_program( struct gl_vertex_program *vp )
 }
 
 static INLINE struct st_geometry_program *
-st_geometry_program( struct gl_geometry_program *vp )
+st_geometry_program( struct gl_geometry_program *gp )
 {
-   return (struct st_geometry_program *)vp;
+   return (struct st_geometry_program *)gp;
 }
 
 static INLINE void
@@ -198,32 +258,43 @@ st_reference_fragprog(struct st_context *st,
 }
 
 
-extern void
-st_translate_fragment_program(struct st_context *st,
-                              struct st_fragment_program *fp);
+extern struct st_vp_variant *
+st_get_vp_variant(struct st_context *st,
+                  struct st_vertex_program *stvp,
+                  const struct st_vp_variant_key *key);
+
+
+extern struct st_fp_variant *
+st_get_fp_variant(struct st_context *st,
+                  struct st_fragment_program *stfp,
+                  const struct st_fp_variant_key *key);
+
+
+extern struct st_gp_variant *
+st_get_gp_variant(struct st_context *st,
+                  struct st_geometry_program *stgp,
+                  const struct st_gp_variant_key *key);
+
+
 
 extern void
-st_translate_geometry_program(struct st_context *st,
-                              struct st_geometry_program *stgp);
+st_release_vp_variants( struct st_context *st,
+                        struct st_vertex_program *stvp );
 
-/* Called after program string change, discard all previous
- * compilation results.
- */
 extern void
-st_prepare_vertex_program(struct st_context *st,
-                          struct st_vertex_program *stvp);
+st_release_fp_variants( struct st_context *st,
+                        struct st_fragment_program *stfp );
 
-extern struct st_vp_varient *
-st_translate_vertex_program(struct st_context *st,
-                            struct st_vertex_program *stvp,
-                            const struct st_vp_varient_key *key);
+extern void
+st_release_gp_variants(struct st_context *st,
+                       struct st_geometry_program *stgp);
 
-void
-st_vp_release_varients( struct st_context *st,
-                        struct st_vertex_program *stvp );
 
 extern void
 st_print_shaders(struct gl_context *ctx);
 
+extern void
+st_destroy_program_variants(struct st_context *st);
+
 
 #endif
diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c
index c6cf2ba061b..155ea39f18c 100644
--- a/src/mesa/state_tracker/st_texture.c
+++ b/src/mesa/state_tracker/st_texture.c
@@ -84,6 +84,7 @@ st_texture_create(struct st_context *st,
    pt.width0 = width0;
    pt.height0 = height0;
    pt.depth0 = depth0;
+   pt.array_size = (target == PIPE_TEXTURE_CUBE ? 6 : 1);
    pt.usage = PIPE_USAGE_DEFAULT;
    pt.bind = bind;
    pt.flags = 0;
@@ -136,7 +137,7 @@ st_texture_match_image(const struct pipe_resource *pt,
  */
 GLubyte *
 st_texture_image_map(struct st_context *st, struct st_texture_image *stImage,
-		     GLuint zoffset, enum pipe_transfer_usage usage,
+                     GLuint zoffset, enum pipe_transfer_usage usage,
                      GLuint x, GLuint y, GLuint w, GLuint h)
 {
    struct pipe_context *pipe = st->pipe;
@@ -144,9 +145,9 @@ st_texture_image_map(struct st_context *st, struct st_texture_image *stImage,
 
    DBG("%s \n", __FUNCTION__);
 
-   stImage->transfer = pipe_get_transfer(st->pipe, pt, stImage->face,
-						    stImage->level, zoffset,
-						    usage, x, y, w, h);
+   stImage->transfer = pipe_get_transfer(st->pipe, pt, stImage->level,
+                                         stImage->face + zoffset,
+                                         usage, x, y, w, h);
 
    if (stImage->transfer)
       return pipe_transfer_map(pipe, stImage->transfer);
@@ -219,10 +220,10 @@ st_texture_image_data(struct st_context *st,
    DBG("%s\n", __FUNCTION__);
 
    for (i = 0; i < depth; i++) {
-      dst_transfer = pipe_get_transfer(st->pipe, dst, face, level, i,
-						  PIPE_TRANSFER_WRITE, 0, 0,
-						  u_minify(dst->width0, level),
-                                                  u_minify(dst->height0, level));
+      dst_transfer = pipe_get_transfer(st->pipe, dst, level, face + i,
+                                       PIPE_TRANSFER_WRITE, 0, 0,
+                                       u_minify(dst->width0, level),
+                                       u_minify(dst->height0, level));
 
       st_surface_data(pipe, dst_transfer,
 		      0, 0,                             /* dstx, dsty */
@@ -230,7 +231,7 @@ st_texture_image_data(struct st_context *st,
 		      src_row_stride,
 		      0, 0,                             /* source x, y */
 		      u_minify(dst->width0, level),
-                      u_minify(dst->height0, level));      /* width, height */
+                      u_minify(dst->height0, level));    /* width, height */
 
       pipe->transfer_destroy(pipe, dst_transfer);
 
@@ -245,14 +246,10 @@ st_texture_image_data(struct st_context *st,
 static void
 print_center_pixel(struct pipe_context *pipe, struct pipe_resource *src)
 {
-   struct pipe_subresource rect;
    struct pipe_transfer *xfer;
    struct pipe_box region;
    ubyte *map;
 
-   rect.face = 0;
-   rect.level = 0;
-
    region.x = src->width0 / 2;
    region.y = src->height0 / 2;
    region.z = 0;
@@ -260,7 +257,7 @@ print_center_pixel(struct pipe_context *pipe, struct pipe_resource *src)
    region.height = 1;
    region.depth = 1;
 
-   xfer = pipe->get_transfer(pipe, src, rect, PIPE_TRANSFER_READ, &region);
+   xfer = pipe->get_transfer(pipe, src, 0, PIPE_TRANSFER_READ, &region);
    map = pipe->transfer_map(pipe, xfer);
 
    printf("center pixel: %d %d %d %d\n", map[0], map[1], map[2], map[3]);
@@ -282,22 +279,26 @@ st_texture_image_copy(struct pipe_context *pipe,
                       struct pipe_resource *src, GLuint srcLevel,
                       GLuint face)
 {
-   GLuint width = u_minify(dst->width0, dstLevel); 
-   GLuint height = u_minify(dst->height0, dstLevel); 
-   GLuint depth = u_minify(dst->depth0, dstLevel); 
-   struct pipe_subresource dstsub, srcsub;
+   GLuint width = u_minify(dst->width0, dstLevel);
+   GLuint height = u_minify(dst->height0, dstLevel);
+   GLuint depth = u_minify(dst->depth0, dstLevel);
+   struct pipe_box src_box;
    GLuint i;
 
    assert(u_minify(src->width0, srcLevel) == width);
    assert(u_minify(src->height0, srcLevel) == height);
    assert(u_minify(src->depth0, srcLevel) == depth);
 
-   dstsub.face = face;
-   dstsub.level = dstLevel;
-   srcsub.face = face;
-   srcsub.level = srcLevel;
+   src_box.x = 0;
+   src_box.y = 0;
+   src_box.width = width;
+   src_box.height = height;
+   src_box.depth = 1;
    /* Loop over 3D image slices */
-   for (i = 0; i < depth; i++) {
+   /* could (and probably should) use "true" 3d box here -
+      but drivers can't quite handle it yet */
+   for (i = face; i < face + depth; i++) {
+      src_box.z = i;
 
       if (0)  {
          print_center_pixel(pipe, src);
@@ -305,12 +306,11 @@ st_texture_image_copy(struct pipe_context *pipe,
 
       pipe->resource_copy_region(pipe,
                                  dst,
-                                 dstsub,
+                                 dstLevel,
                                  0, 0, i,/* destX, Y, Z */
                                  src,
-                                 srcsub,
-                                 0, 0, i,/* srcX, Y, Z */
-                                 width, height);
+                                 srcLevel,
+                                 &src_box);
    }
 }