13 files changed, 313 insertions, 314 deletions

diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c
index fc1ff5be04..ee649be885 100644
--- a/src/mesa/state_tracker/st_atom_shader.c
+++ b/src/mesa/state_tracker/st_atom_shader.c
@@ -175,11 +175,12 @@ find_translated_vp(struct st_context *st,
 
    /* See if we need to translate vertex program to TGSI form */
    if (xvp->serialNo != stvp->serialNo) {
-      GLuint outAttr, dummySlot;
+      GLuint outAttr;
       const GLbitfield outputsWritten = stvp->Base.Base.OutputsWritten;
       GLuint numVpOuts = 0;
       GLboolean emitPntSize = GL_FALSE, emitBFC0 = GL_FALSE, emitBFC1 = GL_FALSE;
-      GLint maxGeneric;
+      GLbitfield usedGenerics = 0x0;
+      GLbitfield usedOutputSlots = 0x0;
 
       /* Compute mapping of vertex program outputs to slots, which depends
        * on the fragment program's input->slot mapping.
@@ -192,10 +193,12 @@ find_translated_vp(struct st_context *st,
 
          if (outAttr == VERT_RESULT_HPOS) {
             /* always put xformed position into slot zero */
-            xvp->output_to_slot[VERT_RESULT_HPOS] = 0;
+            GLuint slot = 0;
+            xvp->output_to_slot[VERT_RESULT_HPOS] = slot;
             xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_POSITION;
             xvp->output_to_semantic_index[outAttr] = 0;
             numVpOuts++;
+            usedOutputSlots |= (1 << slot);
          }
          else if (outputsWritten & (1 << outAttr)) {
             /* see if the frag prog wants this vert output */
@@ -209,6 +212,12 @@ find_translated_vp(struct st_context *st,
                   xvp->output_to_semantic_name[outAttr] = stfp->input_semantic_name[fpInSlot];
                   xvp->output_to_semantic_index[outAttr] = stfp->input_semantic_index[fpInSlot];
                   numVpOuts++;
+                  usedOutputSlots |= (1 << vpOutSlot);
+               }
+               else {
+#if 0 /*debug*/
+                  printf("VP output %d not used by FP\n", outAttr);
+#endif
                }
             }
             else if (outAttr == VERT_RESULT_PSIZ)
@@ -226,45 +235,49 @@ find_translated_vp(struct st_context *st,
 
       /* must do these last */
       if (emitPntSize) {
-         xvp->output_to_slot[VERT_RESULT_PSIZ] = numVpOuts++;
+         GLuint slot = numVpOuts++;
+         xvp->output_to_slot[VERT_RESULT_PSIZ] = slot;
          xvp->output_to_semantic_name[VERT_RESULT_PSIZ] = TGSI_SEMANTIC_PSIZE;
          xvp->output_to_semantic_index[VERT_RESULT_PSIZ] = 0;
+         usedOutputSlots |= (1 << slot);
       }
       if (emitBFC0) {
-         xvp->output_to_slot[VERT_RESULT_BFC0] = numVpOuts++;
+         GLuint slot = numVpOuts++;
+         xvp->output_to_slot[VERT_RESULT_BFC0] = slot;
          xvp->output_to_semantic_name[VERT_RESULT_BFC0] = TGSI_SEMANTIC_COLOR;
          xvp->output_to_semantic_index[VERT_RESULT_BFC0] = 0;
+         usedOutputSlots |= (1 << slot);
       }
       if (emitBFC1) {
-         xvp->output_to_slot[VERT_RESULT_BFC1] = numVpOuts++;
-         xvp->output_to_semantic_name[VERT_RESULT_BFC0] = TGSI_SEMANTIC_COLOR;
-         xvp->output_to_semantic_index[VERT_RESULT_BFC0] = 1;
+         GLuint slot = numVpOuts++;
+         xvp->output_to_slot[VERT_RESULT_BFC1] = slot;
+         xvp->output_to_semantic_name[VERT_RESULT_BFC1] = TGSI_SEMANTIC_COLOR;
+         xvp->output_to_semantic_index[VERT_RESULT_BFC1] = 1;
+         usedOutputSlots |= (1 << slot);
       }
 
-      /* Unneeded vertex program outputs will go to this slot.
-       * We could use this info to do dead code elimination in the
-       * vertex program.
-       */
-      dummySlot = numVpOuts;
-
-      /* find max GENERIC slot index */
-      maxGeneric = -1;
+      /* build usedGenerics mask */
+      usedGenerics = 0x0;
       for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) {
          if (xvp->output_to_semantic_name[outAttr] == TGSI_SEMANTIC_GENERIC) {
-            maxGeneric = MAX2(maxGeneric,
-                              xvp->output_to_semantic_index[outAttr]);
+            usedGenerics |= (1 << xvp->output_to_semantic_index[outAttr]);
          }
       }
 
-      /* Map vert program outputs that aren't used to the dummy slot
-       * (and an unused generic attribute slot).
+      /* For each vertex program output that doesn't match up to a fragment
+       * program input, map the vertex program output to a free slot and
+       * free generic attribute.
        */
       for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) {
          if (outputsWritten & (1 << outAttr)) {
             if (xvp->output_to_slot[outAttr] == UNUSED) {
-               xvp->output_to_slot[outAttr] = dummySlot;
+               GLint freeGeneric = _mesa_ffs(~usedGenerics) - 1;
+               GLint freeSlot = _mesa_ffs(~usedOutputSlots) - 1;
+               usedGenerics |= (1 << freeGeneric);
+               usedOutputSlots |= (1 << freeSlot);
+               xvp->output_to_slot[outAttr] = freeSlot;
                xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_GENERIC;
-               xvp->output_to_semantic_index[outAttr] = maxGeneric + 1;
+               xvp->output_to_semantic_index[outAttr] = freeGeneric;
             }
          }
 
diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c
index 7f793cf08d..95181578f6 100644
--- a/src/mesa/state_tracker/st_cb_accum.c
+++ b/src/mesa/state_tracker/st_cb_accum.c
@@ -48,10 +48,6 @@
 #include "util/u_tile.h"
 
 
-#define UNCLAMPED_FLOAT_TO_SHORT(us, f)  \
-   us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) )
-
-
 /**
  * For hardware that supports deep color buffers, we could accelerate
  * most/all the accum operations with blending/texturing.
@@ -59,74 +55,20 @@
  */
 
 
-/**
- * Wrapper for pipe_get_tile_rgba().  Do format/cpp override to make the
- * tile util function think the surface is 16bit/channel, even if it's not.
- * See also: st_renderbuffer_alloc_storage()
- */
-static void
-acc_get_tile_rgba(struct pipe_context *pipe, struct pipe_transfer *acc_pt,
-                  uint x, uint y, uint w, uint h, float *p)
-{
-   const enum pipe_format f = acc_pt->format;
-   const struct pipe_format_block b = acc_pt->block;
-
-   acc_pt->format = DEFAULT_ACCUM_PIPE_FORMAT;
-   acc_pt->block.size = 8;
-   acc_pt->block.width = 1;
-   acc_pt->block.height = 1;
-
-   pipe_get_tile_rgba(acc_pt, x, y, w, h, p);
-
-   acc_pt->format = f;
-   acc_pt->block = b;
-}
-
-
-/**
- * Wrapper for pipe_put_tile_rgba().  Do format/cpp override to make the
- * tile util function think the surface is 16bit/channel, even if it's not.
- * See also: st_renderbuffer_alloc_storage()
- */
-static void
-acc_put_tile_rgba(struct pipe_context *pipe, struct pipe_transfer *acc_pt,
-                  uint x, uint y, uint w, uint h, const float *p)
-{
-   enum pipe_format f = acc_pt->format;
-   const struct pipe_format_block b = acc_pt->block;
-
-   acc_pt->format = DEFAULT_ACCUM_PIPE_FORMAT;
-   acc_pt->block.size = 8;
-   acc_pt->block.width = 1;
-   acc_pt->block.height = 1;
-
-   pipe_put_tile_rgba(acc_pt, x, y, w, h, p);
-
-   acc_pt->format = f;
-   acc_pt->block = b;
-}
-
-
-
 void
 st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
 {
    struct st_renderbuffer *acc_strb = st_renderbuffer(rb);
-   struct pipe_transfer *acc_pt;
-   struct pipe_screen *screen = ctx->st->pipe->screen;
    const GLint xpos = ctx->DrawBuffer->_Xmin;
    const GLint ypos = ctx->DrawBuffer->_Ymin;
    const GLint width = ctx->DrawBuffer->_Xmax - xpos;
    const GLint height = ctx->DrawBuffer->_Ymax - ypos;
-   GLubyte *map;
-
-   acc_pt = st_cond_flush_get_tex_transfer(st_context(ctx), acc_strb->texture,
-					   0, 0, 0,
-					   PIPE_TRANSFER_WRITE, xpos, ypos,
-					   width, height);
-   map = screen->transfer_map(screen, acc_pt);
+   size_t stride = acc_strb->stride;
+   GLubyte *data = acc_strb->data;
 
-   /* note acc_strb->format might not equal acc_pt->format */
+   if(!data)
+      return;
+   
    switch (acc_strb->format) {
    case PIPE_FORMAT_R16G16B16A16_SNORM:
       {
@@ -136,7 +78,7 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
          GLshort a = FLOAT_TO_SHORT(ctx->Accum.ClearColor[3]);
          int i, j;
          for (i = 0; i < height; i++) {
-            GLshort *dst = (GLshort *) (map + i * acc_pt->stride + xpos * 8);
+            GLshort *dst = (GLshort *) (data + (ypos + i) * stride + xpos * 8);
             for (j = 0; j < width; j++) {
                dst[0] = r;
                dst[1] = g;
@@ -150,9 +92,6 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
    default:
       _mesa_problem(ctx, "unexpected format in st_clear_accum_buffer()");
    }
-
-   screen->transfer_unmap(screen, acc_pt);
-   screen->tex_transfer_destroy(acc_pt);
 }
 
 
@@ -162,27 +101,18 @@ accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias,
           GLint xpos, GLint ypos, GLint width, GLint height,
           struct st_renderbuffer *acc_strb)
 {
-   struct pipe_screen *screen = ctx->st->pipe->screen;
-   struct pipe_transfer *acc_pt;
-   GLubyte *map;
-
-   acc_pt = st_cond_flush_get_tex_transfer(st_context(ctx), acc_strb->texture,
-					   0, 0, 0,
-					   PIPE_TRANSFER_READ_WRITE,
-					   xpos, ypos,
-					   width, height);
-   map = screen->transfer_map(screen, acc_pt);
+   size_t stride = acc_strb->stride;
+   GLubyte *data = acc_strb->data;
 
-   /* note acc_strb->format might not equal acc_pt->format */
    switch (acc_strb->format) {
    case PIPE_FORMAT_R16G16B16A16_SNORM:
       {
          int i, j;
          for (i = 0; i < height; i++) {
-            GLshort *acc = (GLshort *) (map + (ypos + i) * acc_pt->stride + xpos * 8);
+            GLshort *acc = (GLshort *) (data + (ypos + i) * stride + xpos * 8);
             for (j = 0; j < width * 4; j++) {
-               float val = SHORT_TO_FLOAT(acc[j]) * scale + bias;
-               acc[j] = FLOAT_TO_SHORT(val);
+               float val = SHORT_TO_FLOAT(*acc) * scale + bias;
+               *acc++ = FLOAT_TO_SHORT(val);
             }
          }
       }
@@ -190,9 +120,6 @@ accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias,
    default:
       _mesa_problem(NULL, "unexpected format in st_clear_accum_buffer()");
    }
-
-   screen->transfer_unmap(screen, acc_pt);
-   screen->tex_transfer_destroy(acc_pt);
 }
 
 
@@ -204,39 +131,39 @@ accum_accum(struct st_context *st, GLfloat value,
 {
    struct pipe_context *pipe = st->pipe;
    struct pipe_screen *screen = pipe->screen;
-   struct pipe_transfer *acc_trans, *color_trans;
-   GLfloat *colorBuf, *accBuf;
-   GLint i;
-
-   acc_trans = st_cond_flush_get_tex_transfer(st, acc_strb->texture, 0, 0, 0,
-					      PIPE_TRANSFER_READ, xpos, ypos,
-					      width, height);
+   struct pipe_transfer *color_trans;
+   size_t stride = acc_strb->stride;
+   GLubyte *data = acc_strb->data;
+   GLfloat *buf;
 
    color_trans = st_cond_flush_get_tex_transfer(st, color_strb->texture,
 						0, 0, 0,
 						PIPE_TRANSFER_READ, xpos, ypos,
 						width, height);
 
-   colorBuf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
-   accBuf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
+   buf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
 
-   pipe_get_tile_rgba(color_trans, 0, 0, width, height, colorBuf);
-   acc_get_tile_rgba(pipe, acc_trans, 0, 0, width, height, accBuf);
+   pipe_get_tile_rgba(color_trans, 0, 0, width, height, buf);
 
-   for (i = 0; i < 4 * width * height; i++) {
-      accBuf[i] = accBuf[i] + colorBuf[i] * value;
+   switch (acc_strb->format) {
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      {
+         const GLfloat *color = buf;
+         int i, j;
+         for (i = 0; i < height; i++) {
+            GLshort *acc = (GLshort *) (data + (ypos + i) * stride + xpos * 8);
+            for (j = 0; j < width * 4; j++) {
+               float val = *color++ * value;
+               *acc++ += FLOAT_TO_SHORT(val);
+            }
+         }
+      }
+      break;
+   default:
+      _mesa_problem(NULL, "unexpected format in st_clear_accum_buffer()");
    }
 
-   screen->tex_transfer_destroy(acc_trans);
-   acc_trans = st_no_flush_get_tex_transfer(st, acc_strb->texture, 0, 0, 0,
-					    PIPE_TRANSFER_WRITE, xpos, ypos,
-					    width, height);
-
-   acc_put_tile_rgba(pipe, acc_trans, 0, 0, width, height, accBuf);
-
-   _mesa_free(colorBuf);
-   _mesa_free(accBuf);
-   screen->tex_transfer_destroy(acc_trans);
+   _mesa_free(buf);
    screen->tex_transfer_destroy(color_trans);
 }
 
@@ -249,13 +176,10 @@ accum_load(struct st_context *st, GLfloat value,
 {
    struct pipe_context *pipe = st->pipe;
    struct pipe_screen *screen = pipe->screen;
-   struct pipe_transfer *acc_trans, *color_trans;
+   struct pipe_transfer *color_trans;
+   size_t stride = acc_strb->stride;
+   GLubyte *data = acc_strb->data;
    GLfloat *buf;
-   GLint i;
-
-   acc_trans = st_cond_flush_get_tex_transfer(st, acc_strb->texture, 0, 0, 0,
-					      PIPE_TRANSFER_WRITE, xpos, ypos,
-					      width, height);
 
    color_trans = st_cond_flush_get_tex_transfer(st, color_strb->texture,
 						0, 0, 0,
@@ -266,14 +190,25 @@ accum_load(struct st_context *st, GLfloat value,
 
    pipe_get_tile_rgba(color_trans, 0, 0, width, height, buf);
 
-   for (i = 0; i < 4 * width * height; i++) {
-      buf[i] = buf[i] * value;
+   switch (acc_strb->format) {
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      {
+         const GLfloat *color = buf;
+         int i, j;
+         for (i = 0; i < height; i++) {
+            GLshort *acc = (GLshort *) (data + (ypos + i) * stride + xpos * 8);
+            for (j = 0; j < width * 4; j++) {
+               float val = *color++ * value;
+               *acc++ = FLOAT_TO_SHORT(val);
+            }
+         }
+      }
+      break;
+   default:
+      _mesa_problem(NULL, "unexpected format in st_clear_accum_buffer()");
    }
 
-   acc_put_tile_rgba(pipe, acc_trans, 0, 0, width, height, buf);
-
    _mesa_free(buf);
-   screen->tex_transfer_destroy(acc_trans);
    screen->tex_transfer_destroy(color_trans);
 }
 
@@ -287,48 +222,58 @@ accum_return(GLcontext *ctx, GLfloat value,
    struct pipe_context *pipe = ctx->st->pipe;
    struct pipe_screen *screen = pipe->screen;
    const GLubyte *colormask = ctx->Color.ColorMask;
-   struct pipe_transfer *acc_trans, *color_trans;
-   GLfloat *abuf, *cbuf = NULL;
-   GLint i, ch;
-
-   abuf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
+   enum pipe_transfer_usage usage;
+   struct pipe_transfer *color_trans;
+   size_t stride = acc_strb->stride;
+   const GLubyte *data = acc_strb->data;
+   GLfloat *buf;
 
-   acc_trans = st_cond_flush_get_tex_transfer(st_context(ctx),
-					      acc_strb->texture, 0, 0, 0,
-					      PIPE_TRANSFER_READ, xpos, ypos,
-					      width, height);
+   buf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
 
+   if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3])
+      usage = PIPE_TRANSFER_READ_WRITE;
+   else
+      usage = PIPE_TRANSFER_WRITE;
+   
    color_trans = st_cond_flush_get_tex_transfer(st_context(ctx),
 						color_strb->texture, 0, 0, 0,
-						PIPE_TRANSFER_READ_WRITE,
+						usage,
 						xpos, ypos,
 						width, height);
 
-   acc_get_tile_rgba(pipe, acc_trans, 0, 0, width, height, abuf);
-
-   if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3]) {
-      cbuf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
-      pipe_get_tile_rgba(color_trans, 0, 0, width, height, cbuf);
-   }
+   if (usage != PIPE_TRANSFER_WRITE)
+      pipe_get_tile_rgba(color_trans, 0, 0, width, height, buf);
 
-   for (i = 0; i < width * height; i++) {
-      for (ch = 0; ch < 4; ch++) {
-         if (colormask[ch]) {
-            GLfloat val = abuf[i * 4 + ch] * value;
-            abuf[i * 4 + ch] = CLAMP(val, 0.0f, 1.0f);
-         }
-         else {
-            abuf[i * 4 + ch] = cbuf[i * 4 + ch];
+   switch (acc_strb->format) {
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      {
+         GLfloat *color = buf;
+         int i, j, ch;
+         for (i = 0; i < height; i++) {
+            const GLshort *acc = (const GLshort *) (data + (ypos + i) * stride + xpos * 8);
+            for (j = 0; j < width; j++) {
+               for (ch = 0; ch < 4; ch++) {
+                  if (colormask[ch]) {
+                     GLfloat val = SHORT_TO_FLOAT(*acc * value);
+                     *color = CLAMP(val, 0.0f, 1.0f);
+                  }
+                  else {
+                     /* No change */
+                  }
+                  ++acc;
+                  ++color;
+               }
+            }
          }
       }
+      break;
+   default:
+      _mesa_problem(NULL, "unexpected format in st_clear_accum_buffer()");
    }
 
-   pipe_put_tile_rgba(color_trans, 0, 0, width, height, abuf);
+   pipe_put_tile_rgba(color_trans, 0, 0, width, height, buf);
 
-   _mesa_free(abuf);
-   if (cbuf)
-      _mesa_free(cbuf);
-   screen->tex_transfer_destroy(acc_trans);
+   _mesa_free(buf);
    screen->tex_transfer_destroy(color_trans);
 }
 
@@ -347,6 +292,9 @@ st_Accum(GLcontext *ctx, GLenum op, GLfloat value)
    const GLint width = ctx->DrawBuffer->_Xmax - xpos;
    const GLint height = ctx->DrawBuffer->_Ymax - ypos;
 
+   if(!acc_strb->data)
+      return;
+   
    /* make sure color bufs aren't cached */
    st_flush( st, PIPE_FLUSH_RENDER_CACHE, NULL );
 
diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c
index f140641d7e..7021d73208 100644
--- a/src/mesa/state_tracker/st_cb_bufferobjects.c
+++ b/src/mesa/state_tracker/st_cb_bufferobjects.c
@@ -245,11 +245,10 @@ st_bufferobj_map_range(GLcontext *ctx, GLenum target,
    assert(offset < obj->Size);
    assert(offset + length <= obj->Size);
 
-   map = obj->Pointer = pipe_buffer_map_range(pipe->screen, st_obj->buffer,
-                                              offset, length, flags);
-   if (obj->Pointer) {
-      obj->Offset = 0;
-      obj->Length = obj->Size;
+   map = obj->Pointer = pipe_buffer_map_range(pipe->screen, st_obj->buffer, offset, length, flags);
+   if(obj->Pointer) {
+      obj->Offset = offset;
+      obj->Length = length;
       map += offset;
    }
    
@@ -268,7 +267,6 @@ st_bufferobj_flush_mapped_range(GLcontext *ctx, GLenum target,
    /* Subrange is relative to mapped range */
    assert(offset >= 0);
    assert(length >= 0);
-   assert(offset < obj->Length);
    assert(offset + length <= obj->Length);
    
    pipe_buffer_flush_mapped_range(pipe->screen, st_obj->buffer, 
diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c
index 880e83108c..668acbccb8 100644
--- a/src/mesa/state_tracker/st_cb_clear.c
+++ b/src/mesa/state_tracker/st_cb_clear.c
@@ -103,20 +103,6 @@ st_destroy_clear(struct st_context *st)
 }
 
 
-static GLboolean
-is_depth_stencil_format(enum pipe_format pipeFormat)
-{
-   switch (pipeFormat) {
-   case PIPE_FORMAT_S8Z24_UNORM:
-   case PIPE_FORMAT_Z24S8_UNORM:
-      return GL_TRUE;
-   default:
-      return GL_FALSE;
-   }
-}
-
-
-
 /**
  * Draw a screen-aligned quadrilateral.
  * Coords are window coords with y=0=bottom.  These will be passed
@@ -331,7 +317,7 @@ static INLINE GLboolean
 check_clear_depth_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
 {
    const struct st_renderbuffer *strb = st_renderbuffer(rb);
-   const GLboolean isDS = is_depth_stencil_format(strb->surface->format);
+   const GLboolean isDS = pf_is_depth_and_stencil(strb->surface->format);
 
    if (ctx->Scissor.Enabled)
       return TRUE;
@@ -351,7 +337,7 @@ static INLINE GLboolean
 check_clear_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
 {
    const struct st_renderbuffer *strb = st_renderbuffer(rb);
-   const GLboolean isDS = is_depth_stencil_format(strb->surface->format);
+   const GLboolean isDS = pf_is_depth_and_stencil(strb->surface->format);
    const GLuint stencilMax = (1 << rb->StencilBits) - 1;
    const GLboolean maskStencil
       = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax;
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c
index 89725cfe8d..2027b713ce 100644
--- a/src/mesa/state_tracker/st_cb_drawpixels.c
+++ b/src/mesa/state_tracker/st_cb_drawpixels.c
@@ -630,6 +630,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
    struct pipe_context *pipe = st->pipe;
    struct pipe_screen *screen = pipe->screen;
    struct st_renderbuffer *strb;
+   enum pipe_transfer_usage usage;
    struct pipe_transfer *pt;
    const GLboolean zoom = ctx->Pixel.ZoomX != 1.0 || ctx->Pixel.ZoomY != 1.0;
    GLint skipPixels;
@@ -642,8 +643,14 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
       y = ctx->DrawBuffer->Height - y - height;
    }
 
+   if(format != GL_DEPTH_STENCIL && 
+      pf_get_component_bits( strb->format, PIPE_FORMAT_COMP_Z ) != 0)
+      usage = PIPE_TRANSFER_READ_WRITE;
+   else
+      usage = PIPE_TRANSFER_WRITE;
+   
    pt = st_cond_flush_get_tex_transfer(st_context(ctx), strb->texture, 0, 0, 0,
-				       PIPE_TRANSFER_WRITE, x, y,
+				       usage, x, y,
 				       width, height);
 
    stmap = screen->transfer_map(screen, pt);
@@ -694,6 +701,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
             case PIPE_FORMAT_S8_UNORM:
                {
                   ubyte *dest = stmap + spanY * pt->stride + spanX;
+                  assert(usage == PIPE_TRANSFER_WRITE);
                   memcpy(dest, sValues, spanWidth);
                }
                break;
@@ -701,6 +709,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
                if (format == GL_DEPTH_STENCIL) {
                   uint *dest = (uint *) (stmap + spanY * pt->stride + spanX*4);
                   GLint k;
+                  assert(usage == PIPE_TRANSFER_WRITE);
                   for (k = 0; k < spanWidth; k++) {
                      dest[k] = zValues[k] | (sValues[k] << 24);
                   }
@@ -708,6 +717,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
                else {
                   uint *dest = (uint *) (stmap + spanY * pt->stride + spanX*4);
                   GLint k;
+                  assert(usage == PIPE_TRANSFER_READ_WRITE);
                   for (k = 0; k < spanWidth; k++) {
                      dest[k] = (dest[k] & 0xffffff) | (sValues[k] << 24);
                   }
@@ -717,13 +727,15 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
                if (format == GL_DEPTH_STENCIL) {
                   uint *dest = (uint *) (stmap + spanY * pt->stride + spanX*4);
                   GLint k;
+                  assert(usage == PIPE_TRANSFER_WRITE);
                   for (k = 0; k < spanWidth; k++) {
-                     dest[k] = zValues[k] | (sValues[k] << 24);
+                     dest[k] = (zValues[k] << 8) | (sValues[k] & 0xff);
                   }
                }
                else {
                   uint *dest = (uint *) (stmap + spanY * pt->stride + spanX*4);
                   GLint k;
+                  assert(usage == PIPE_TRANSFER_READ_WRITE);
                   for (k = 0; k < spanWidth; k++) {
                      dest[k] = (dest[k] & 0xffffff00) | (sValues[k] & 0xff);
                   }
@@ -757,7 +769,6 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
    struct st_vertex_program *stvp;
    struct st_context *st = ctx->st;
    struct pipe_surface *ps;
-   GLuint bufferFormat;
    const GLfloat *color;
 
    if (format == GL_STENCIL_INDEX ||
@@ -785,8 +796,6 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
       color = NULL;
    }
 
-   bufferFormat = ps->format;
-
    /* draw with textured quad */
    {
       struct pipe_texture *pt
@@ -811,6 +820,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
 {
    struct st_renderbuffer *rbDraw = st_renderbuffer(ctx->DrawBuffer->_StencilBuffer);
    struct pipe_screen *screen = ctx->st->pipe->screen;
+   enum pipe_transfer_usage usage;
    struct pipe_transfer *ptDraw;
    ubyte *drawMap;
    ubyte *buffer;
@@ -827,9 +837,14 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
                           GL_STENCIL_INDEX, GL_UNSIGNED_BYTE,
                           &ctx->DefaultPacking, buffer);
 
+   if(pf_get_component_bits( rbDraw->format, PIPE_FORMAT_COMP_Z ) != 0)
+      usage = PIPE_TRANSFER_READ_WRITE;
+   else
+      usage = PIPE_TRANSFER_WRITE;
+   
    ptDraw = st_cond_flush_get_tex_transfer(st_context(ctx),
 					   rbDraw->texture, 0, 0, 0,
-					   PIPE_TRANSFER_WRITE, dstx, dsty,
+					   usage, dstx, dsty,
 					   width, height);
 
    assert(ptDraw->block.width == 1);
@@ -859,6 +874,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
          {
             uint *dst4 = (uint *) dst;
             int j;
+            assert(usage == PIPE_TRANSFER_READ_WRITE);
             for (j = 0; j < width; j++) {
                *dst4 = (*dst4 & 0xffffff) | (src[j] << 24);
                dst4++;
@@ -869,6 +885,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
          {
             uint *dst4 = (uint *) dst;
             int j;
+            assert(usage == PIPE_TRANSFER_READ_WRITE);
             for (j = 0; j < width; j++) {
                *dst4 = (*dst4 & 0xffffff00) | (src[j] & 0xff);
                dst4++;
@@ -876,6 +893,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
          }
          break;
       case PIPE_FORMAT_S8_UNORM:
+         assert(usage == PIPE_TRANSFER_WRITE);
          memcpy(dst, src, width);
          break;
       default:
diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c
index c249f3b357..ecdb988033 100644
--- a/src/mesa/state_tracker/st_cb_fbo.c
+++ b/src/mesa/state_tracker/st_cb_fbo.c
@@ -88,93 +88,92 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
 {
    struct pipe_context *pipe = ctx->st->pipe;
    struct st_renderbuffer *strb = st_renderbuffer(rb);
-   struct pipe_texture template;
-   unsigned surface_usage;
-
-   /* Free the old surface and texture
-    */
-   pipe_surface_reference( &strb->surface, NULL );
-   pipe_texture_reference( &strb->texture, NULL );
-
-   /* Setup new texture template.
-    */
-   memset(&template, 0, sizeof(template));
-   template.target = PIPE_TEXTURE_2D;
-   if (strb->format != PIPE_FORMAT_NONE) {
-      template.format = strb->format;
-   }
-   else {
-      template.format = st_choose_renderbuffer_format(pipe, internalFormat);
-   }
-   pf_get_block(template.format, &template.block);
-   template.width[0] = width;
-   template.height[0] = height;
-   template.depth[0] = 1;
-   template.last_level = 0;
-   template.nr_samples = rb->NumSamples;
-   if (pf_is_depth_stencil(template.format)) {
-      template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
-   }
-   else {
-      template.tex_usage = (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
-                            PIPE_TEXTURE_USAGE_RENDER_TARGET);
-   }
+   enum pipe_format format;
 
+   if (strb->format != PIPE_FORMAT_NONE)
+      format = strb->format;
+   else
+      format = st_choose_renderbuffer_format(pipe, internalFormat);
+      
    /* init renderbuffer fields */
    strb->Base.Width  = width;
    strb->Base.Height = height;
-   init_renderbuffer_bits(strb, template.format);
+   init_renderbuffer_bits(strb, format);
 
    strb->defined = GL_FALSE;  /* undefined contents now */
 
-   /* Probably need dedicated flags for surface usage too: 
-    */
-   surface_usage = (PIPE_BUFFER_USAGE_GPU_READ |
-                    PIPE_BUFFER_USAGE_GPU_WRITE);
-#if 0
-                    PIPE_BUFFER_USAGE_CPU_READ |
-                    PIPE_BUFFER_USAGE_CPU_WRITE);
-#endif
+   if(strb->software) {
+      struct pipe_format_block block;
+      size_t size;
+      
+      _mesa_free(strb->data);
 
-   strb->texture = pipe->screen->texture_create( pipe->screen,
-                                                 &template );
+      assert(strb->format != PIPE_FORMAT_NONE);
+      pf_get_block(strb->format, &block);
+      
+      strb->stride = pf_get_stride(&block, width);
+      size = pf_get_2d_size(&block, strb->stride, height);
+      
+      strb->data = _mesa_malloc(size);
+      
+      return strb->data != NULL;
+   }
+   else {
+      struct pipe_texture template;
+      unsigned surface_usage;
+    
+      /* Free the old surface and texture
+       */
+      pipe_surface_reference( &strb->surface, NULL );
+      pipe_texture_reference( &strb->texture, NULL );
 
-   /* Special path for accum buffers.  
-    *
-    * Try a different surface format.  Since accum buffers are s/w
-    * only for now, the surface pixel format doesn't really matter,
-    * only that the buffer is large enough.
-    */
-   if (!strb->texture && template.format == DEFAULT_ACCUM_PIPE_FORMAT) 
-   {
-      /* Actually, just setting this usage value should be sufficient
-       * to tell the driver to go ahead and allocate the buffer, even
-       * if HW doesn't support the format.
+      /* Setup new texture template.
+       */
+      memset(&template, 0, sizeof(template));
+      template.target = PIPE_TEXTURE_2D;
+      template.format = format;
+      pf_get_block(format, &template.block);
+      template.width[0] = width;
+      template.height[0] = height;
+      template.depth[0] = 1;
+      template.last_level = 0;
+      template.nr_samples = rb->NumSamples;
+      if (pf_is_depth_stencil(format)) {
+         template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
+      }
+      else {
+         template.tex_usage = (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                               PIPE_TEXTURE_USAGE_RENDER_TARGET);
+      }
+
+      /* Probably need dedicated flags for surface usage too: 
        */
-      template.tex_usage = 0;
-      surface_usage = (PIPE_BUFFER_USAGE_CPU_READ |
+      surface_usage = (PIPE_BUFFER_USAGE_GPU_READ |
+                       PIPE_BUFFER_USAGE_GPU_WRITE);
+#if 0
+                       PIPE_BUFFER_USAGE_CPU_READ |
                        PIPE_BUFFER_USAGE_CPU_WRITE);
+#endif
 
       strb->texture = pipe->screen->texture_create( pipe->screen,
                                                     &template );
 
-   }
+      if (!strb->texture) 
+         return FALSE;
 
-   if (!strb->texture) 
-      return FALSE;
+      strb->surface = pipe->screen->get_tex_surface( pipe->screen,
+                                                     strb->texture,
+                                                     0, 0, 0,
+                                                     surface_usage );
 
-   strb->surface = pipe->screen->get_tex_surface( pipe->screen,
-                                                  strb->texture,
-                                                  0, 0, 0,
-                                                  surface_usage );
+      assert(strb->surface->texture);
+      assert(strb->surface->format);
+      assert(strb->surface->width == width);
+      assert(strb->surface->height == height);
 
-   assert(strb->surface->texture);
-   assert(strb->surface->format);
-   assert(strb->surface->width == width);
-   assert(strb->surface->height == height);
 
-
-   return strb->surface != NULL;
+      return strb->surface != NULL;
+   }
 }
 
 
@@ -188,6 +187,7 @@ st_renderbuffer_delete(struct gl_renderbuffer *rb)
    ASSERT(strb);
    pipe_surface_reference(&strb->surface, NULL);
    pipe_texture_reference(&strb->texture, NULL);
+   _mesa_free(strb->data);
    _mesa_free(strb);
 }
 
@@ -244,7 +244,7 @@ st_new_renderbuffer(GLcontext *ctx, GLuint name)
  * renderbuffer).  The window system code determines the format.
  */
 struct gl_renderbuffer *
-st_new_renderbuffer_fb(enum pipe_format format, int samples)
+st_new_renderbuffer_fb(enum pipe_format format, int samples, boolean sw)
 {
    struct st_renderbuffer *strb;
 
@@ -258,7 +258,8 @@ st_new_renderbuffer_fb(enum pipe_format format, int samples)
    strb->Base.ClassID = 0x4242; /* just a unique value */
    strb->Base.NumSamples = samples;
    strb->format = format;
-
+   strb->software = sw;
+   
    switch (format) {
    case PIPE_FORMAT_A8R8G8B8_UNORM:
    case PIPE_FORMAT_B8G8R8A8_UNORM:
@@ -289,7 +290,7 @@ st_new_renderbuffer_fb(enum pipe_format format, int samples)
       strb->Base.InternalFormat = GL_STENCIL_INDEX8_EXT;
       strb->Base._BaseFormat = GL_STENCIL_INDEX;
       break;
-   case DEFAULT_ACCUM_PIPE_FORMAT: /*PIPE_FORMAT_R16G16B16A16_SNORM*/
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
       strb->Base.InternalFormat = GL_RGBA16;
       strb->Base._BaseFormat = GL_RGBA;
       break;
@@ -536,7 +537,7 @@ check_create_front_buffer(GLcontext *ctx, struct gl_framebuffer *fb,
          samples = back->Base.NumSamples;
 
          /* create front renderbuffer */
-         front = st_new_renderbuffer_fb(colorFormat, samples);
+         front = st_new_renderbuffer_fb(colorFormat, samples, FALSE);
          _mesa_add_renderbuffer(fb, frontIndex, front);
 
          /* alloc texture/surface for new front buffer */
diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h
index fd77d0a95b..bea6eb89c3 100644
--- a/src/mesa/state_tracker/st_cb_fbo.h
+++ b/src/mesa/state_tracker/st_cb_fbo.h
@@ -30,10 +30,6 @@
 #define ST_CB_FBO_H
 
 
-#define DEFAULT_ACCUM_PIPE_FORMAT PIPE_FORMAT_R16G16B16A16_SNORM
-
-
-
 /**
  * Derived renderbuffer class.  Just need to add a pointer to the
  * pipe surface.
@@ -46,6 +42,13 @@ struct st_renderbuffer
    enum pipe_format format;  /** preferred format, or PIPE_FORMAT_NONE */
    GLboolean defined;        /**< defined contents? */
 
+   /**
+    * Used only when hardware accumulation buffers are not supported.
+    */
+   boolean software;
+   size_t stride;
+   void *data;
+   
    struct st_texture_object *rtt;  /**< GL render to texture's texture */
    int rtt_level, rtt_face, rtt_slice;
 
@@ -63,7 +66,7 @@ st_renderbuffer(struct gl_renderbuffer *rb)
 
 
 extern struct gl_renderbuffer *
-st_new_renderbuffer_fb(enum pipe_format format, int samples);
+st_new_renderbuffer_fb(enum pipe_format format, int samples, boolean sw);
 
 extern void
 st_init_fbo_functions(struct dd_function_table *functions);
diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c
index 7a4bbf5ce3..ccf1a0b563 100644
--- a/src/mesa/state_tracker/st_cb_readpixels.c
+++ b/src/mesa/state_tracker/st_cb_readpixels.c
@@ -112,7 +112,7 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
       case PIPE_FORMAT_S8Z24_UNORM:
          if (format == GL_DEPTH_STENCIL) {
             const uint *src = (uint *) (stmap + srcY * pt->stride);
-            const GLfloat scale = 1.0 / (0xffffff);
+            const GLfloat scale = 1.0f / (0xffffff);
             GLint k;
             for (k = 0; k < width; k++) {
                sValues[k] = src[k] >> 24;
@@ -130,7 +130,7 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
       case PIPE_FORMAT_Z24S8_UNORM:
          if (format == GL_DEPTH_STENCIL) {
             const uint *src = (uint *) (stmap + srcY * pt->stride);
-            const GLfloat scale = 1.0 / (0xffffff);
+            const GLfloat scale = 1.0f / (0xffffff);
             GLint k;
             for (k = 0; k < width; k++) {
                sValues[k] = src[k] & 0xff;
@@ -445,11 +445,16 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
             }
          }
          else {
-            /* untested, but simple: */
+            /* XXX: unreachable code -- should be before st_read_stencil_pixels */
             assert(format == GL_DEPTH_STENCIL_EXT);
             for (i = 0; i < height; i++) {
+               GLuint *zshort = (GLuint *)dst;
                pipe_get_tile_raw(trans, 0, y, width, 1, dst, 0);
                y += yStep;
+               /* Reverse into 24/8 */
+               for (j = 0; j < width; j++) {
+                  zshort[j] = (zshort[j] << 8) | (zshort[j] >> 24);
+               }
                dst += dstStride;
             }
          }
@@ -472,7 +477,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
             }
          }
          else {
-            /* untested, but simple: */
+            /* XXX: unreachable code -- should be before st_read_stencil_pixels */
             assert(format == GL_DEPTH_STENCIL_EXT);
             for (i = 0; i < height; i++) {
                pipe_get_tile_raw(trans, 0, y, width, 1, dst, 0);
diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c
index dc65b8b5f5..8f63819e63 100644
--- a/src/mesa/state_tracker/st_extensions.c
+++ b/src/mesa/state_tracker/st_extensions.c
@@ -175,6 +175,7 @@ void st_init_extensions(struct st_context *st)
 
    ctx->Extensions.NV_blend_square = GL_TRUE;
    ctx->Extensions.NV_texgen_reflection = GL_TRUE;
+   ctx->Extensions.NV_texture_env_combine4 = GL_TRUE;
 
    ctx->Extensions.SGI_color_matrix = GL_TRUE;
    ctx->Extensions.SGIS_generate_mipmap = GL_TRUE;
diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c
index d507e3e58d..b243c249e3 100644
--- a/src/mesa/state_tracker/st_format.c
+++ b/src/mesa/state_tracker/st_format.c
@@ -392,7 +392,7 @@ default_depth_format(struct pipe_screen *screen,
  *                   or PIPE_TEXTURE_USAGE_SAMPLER
  */
 enum pipe_format
-st_choose_format(struct pipe_context *pipe, GLint internalFormat,
+st_choose_format(struct pipe_context *pipe, GLenum internalFormat,
                  enum pipe_texture_target target, unsigned tex_usage)
 {
    struct pipe_screen *screen = pipe->screen;
@@ -594,9 +594,13 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat,
 
 
 static GLboolean
-is_stencil_format(GLenum format)
+is_depth_or_stencil_format(GLenum internalFormat)
 {
-   switch (format) {
+   switch (internalFormat) {
+   case GL_DEPTH_COMPONENT:
+   case GL_DEPTH_COMPONENT16:
+   case GL_DEPTH_COMPONENT24:
+   case GL_DEPTH_COMPONENT32:
    case GL_STENCIL_INDEX:
    case GL_STENCIL_INDEX1_EXT:
    case GL_STENCIL_INDEX4_EXT:
@@ -614,10 +618,10 @@ is_stencil_format(GLenum format)
  * Called by FBO code to choose a PIPE_FORMAT_ for drawing surfaces.
  */
 enum pipe_format
-st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat)
+st_choose_renderbuffer_format(struct pipe_context *pipe, GLenum internalFormat)
 {
    uint usage;
-   if (is_stencil_format(internalFormat))
+   if (is_depth_or_stencil_format(internalFormat))
       usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
    else
       usage = PIPE_TEXTURE_USAGE_RENDER_TARGET;
diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h
index 7bbbe2d570..9d9e02fe9b 100644
--- a/src/mesa/state_tracker/st_format.h
+++ b/src/mesa/state_tracker/st_format.h
@@ -64,11 +64,11 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat);
 
 
 extern enum pipe_format
-st_choose_format(struct pipe_context *pipe, GLint internalFormat,
+st_choose_format(struct pipe_context *pipe, GLenum internalFormat,
                  enum pipe_texture_target target, unsigned tex_usage);
 
 extern enum pipe_format
-st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat);
+st_choose_renderbuffer_format(struct pipe_context *pipe, GLenum internalFormat);
 
 
 extern const struct gl_texture_format *
diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c
index ef800291cc..7072cbe62c 100644
--- a/src/mesa/state_tracker/st_framebuffer.c
+++ b/src/mesa/state_tracker/st_framebuffer.c
@@ -60,7 +60,7 @@ st_create_framebuffer( const __GLcontextModes *visual,
 
       if (visual->doubleBufferMode) {
          struct gl_renderbuffer *rb
-            = st_new_renderbuffer_fb(colorFormat, samples);
+            = st_new_renderbuffer_fb(colorFormat, samples, FALSE);
          _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb);
       }
       else {
@@ -69,14 +69,14 @@ st_create_framebuffer( const __GLcontextModes *visual,
           * See check_create_front_buffers().
           */
          struct gl_renderbuffer *rb
-            = st_new_renderbuffer_fb(colorFormat, samples);
+            = st_new_renderbuffer_fb(colorFormat, samples, FALSE);
          _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb);
       }
 
       if (depthFormat == stencilFormat && depthFormat != PIPE_FORMAT_NONE) {
          /* combined depth/stencil buffer */
          struct gl_renderbuffer *depthStencilRb
-            = st_new_renderbuffer_fb(depthFormat, samples);
+            = st_new_renderbuffer_fb(depthFormat, samples, FALSE);
          /* note: bind RB to two attachment points */
          _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthStencilRb);
          _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, depthStencilRb);
@@ -87,34 +87,35 @@ st_create_framebuffer( const __GLcontextModes *visual,
          if (visual->depthBits == 32) {
             /* 32-bit depth buffer */
             struct gl_renderbuffer *depthRb
-               = st_new_renderbuffer_fb(depthFormat, samples);
+               = st_new_renderbuffer_fb(depthFormat, samples, FALSE);
             _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb);
          }
          else if (visual->depthBits == 24) {
             /* 24-bit depth buffer, ignore stencil bits */
             struct gl_renderbuffer *depthRb
-               = st_new_renderbuffer_fb(depthFormat, samples);
+               = st_new_renderbuffer_fb(depthFormat, samples, FALSE);
             _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb);
          }
          else if (visual->depthBits > 0) {
             /* 16-bit depth buffer */
             struct gl_renderbuffer *depthRb
-               = st_new_renderbuffer_fb(depthFormat, samples);
+               = st_new_renderbuffer_fb(depthFormat, samples, FALSE);
             _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb);
          }
 
          if (visual->stencilBits > 0) {
             /* 8-bit stencil */
             struct gl_renderbuffer *stencilRb
-               = st_new_renderbuffer_fb(stencilFormat, samples);
+               = st_new_renderbuffer_fb(stencilFormat, samples, FALSE);
             _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, stencilRb);
          }
       }
 
       if (visual->accumRedBits > 0) {
          /* 16-bit/channel accum */
+         /* TODO: query the pipe screen for accumulation buffer format support */
          struct gl_renderbuffer *accumRb
-            = st_new_renderbuffer_fb(DEFAULT_ACCUM_PIPE_FORMAT, 0); /* XXX accum isn't multisampled right? */
+            = st_new_renderbuffer_fb(PIPE_FORMAT_R16G16B16A16_SNORM, 0, TRUE);
          _mesa_add_renderbuffer(&stfb->Base, BUFFER_ACCUM, accumRb);
       }
 
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
index 34926101ed..263dcc03b5 100644
--- a/src/mesa/state_tracker/st_program.c
+++ b/src/mesa/state_tracker/st_program.c
@@ -55,20 +55,6 @@
 #define TGSI_DEBUG 0
 
 
-/** XXX we should use the version of this from u_memory.h but including
- * that header causes symbol collisions.
- */
-static INLINE void *
-mem_dup(const void *src, uint size)
-{
-   void *dup = _mesa_malloc(size);
-   if (dup)
-      _mesa_memcpy(dup, src, size);
-   return dup;
-}
-
-
-
 /**
  * Translate a Mesa vertex shader into a TGSI shader.
  * \param outputMapping  to map vertex program output registers (VERT_RESULT_x)
@@ -84,7 +70,7 @@ st_translate_vertex_program(struct st_context *st,
                             const ubyte *outputSemanticIndex)
 {
    struct pipe_context *pipe = st->pipe;
-   struct tgsi_token tokens[ST_MAX_SHADER_TOKENS];
+   struct tgsi_token *tokens;
    GLuint defaultOutputMapping[VERT_RESULT_MAX];
    struct pipe_shader_state vs;
    GLuint attr, i;
@@ -102,6 +88,13 @@ st_translate_vertex_program(struct st_context *st,
    GLbitfield input_flags[MAX_PROGRAM_INPUTS];
    GLbitfield output_flags[MAX_PROGRAM_OUTPUTS];
 
+   tokens =  (struct tgsi_token *)MALLOC(ST_MAX_SHADER_TOKENS * sizeof *tokens);
+   if(!tokens) {
+      /* FIXME: propagate error to the caller */
+      assert(0);
+      return;
+   }
+
    memset(&vs, 0, sizeof(vs));
    memset(input_flags, 0, sizeof(input_flags));
    memset(output_flags, 0, sizeof(output_flags));
@@ -297,9 +290,6 @@ st_translate_vertex_program(struct st_context *st,
       }
    }
 
-   assert(vs_output_semantic_name[0] == TGSI_SEMANTIC_POSITION);
-
-
    if (outputMapping) {
       /* find max output slot referenced to compute vs_num_outputs */
       GLuint maxSlot = 0;
@@ -313,6 +303,26 @@ st_translate_vertex_program(struct st_context *st,
       outputMapping = defaultOutputMapping;
    }
 
+#if 0 /* debug */
+   {
+      GLuint i;
+      printf("outputMapping? %d\n", outputMapping ? 1 : 0);
+      if (outputMapping) {
+         printf("attr -> slot\n");
+         for (i = 0; i < 16;  i++) {
+            printf(" %2d       %3d\n", i, outputMapping[i]);
+         }
+      }
+      printf("slot    sem_name  sem_index\n");
+      for (i = 0; i < vs_num_outputs; i++) {
+         printf(" %2d         %d         %d\n",
+                i,
+                vs_output_semantic_name[i],
+                vs_output_semantic_index[i]);
+      }
+   }
+#endif
+
    /* free old shader state, if any */
    if (stvp->state.tokens) {
       _mesa_free((void *) stvp->state.tokens);
@@ -347,7 +357,9 @@ st_translate_vertex_program(struct st_context *st,
    assert(num_tokens < ST_MAX_SHADER_TOKENS);
 
    vs.tokens = (struct tgsi_token *)
-      mem_dup(tokens, num_tokens * sizeof(tokens[0]));
+      _mesa_realloc(tokens,
+                    ST_MAX_SHADER_TOKENS * sizeof *tokens,
+                    num_tokens * sizeof *tokens);
 
    stvp->num_inputs = vs_num_inputs;
    stvp->state = vs; /* struct copy */
@@ -375,7 +387,7 @@ st_translate_fragment_program(struct st_context *st,
                               const GLuint inputMapping[])
 {
    struct pipe_context *pipe = st->pipe;
-   struct tgsi_token tokens[ST_MAX_SHADER_TOKENS];
+   struct tgsi_token *tokens;
    GLuint outputMapping[FRAG_RESULT_MAX];
    GLuint defaultInputMapping[FRAG_ATTRIB_MAX];
    struct pipe_shader_state fs;
@@ -395,6 +407,13 @@ st_translate_fragment_program(struct st_context *st,
    GLbitfield input_flags[MAX_PROGRAM_INPUTS];
    GLbitfield output_flags[MAX_PROGRAM_OUTPUTS];
 
+   tokens =  (struct tgsi_token *)MALLOC(ST_MAX_SHADER_TOKENS * sizeof *tokens);
+   if(!tokens) {
+      /* FIXME: propagate error to the caller */
+      assert(0);
+      return;
+   }
+
    memset(&fs, 0, sizeof(fs));
    memset(input_flags, 0, sizeof(input_flags));
    memset(output_flags, 0, sizeof(output_flags));
@@ -536,7 +555,9 @@ st_translate_fragment_program(struct st_context *st,
    assert(num_tokens < ST_MAX_SHADER_TOKENS);
 
    fs.tokens = (struct tgsi_token *)
-      mem_dup(tokens, num_tokens * sizeof(tokens[0]));
+      _mesa_realloc(tokens,
+                    ST_MAX_SHADER_TOKENS * sizeof *tokens,
+                    num_tokens * sizeof *tokens);
 
    stfp->state = fs; /* struct copy */
    stfp->driver_shader = pipe->create_fs_state(pipe, &fs);