Merge remote branch 'origin/master' into pipe-video

26 files changed, 992 insertions, 1427 deletions

diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c
index 7fd1287368f..ae7f633b8d8 100644
--- a/src/mesa/main/attrib.c
+++ b/src/mesa/main/attrib.c
@@ -125,9 +125,6 @@ struct gl_enable_attrib
    GLbitfield Texture[MAX_TEXTURE_UNITS];
    GLbitfield TexGen[MAX_TEXTURE_UNITS];
 
-   /* SGI_texture_color_table */
-   GLboolean TextureColorTable[MAX_TEXTURE_UNITS];
-
    /* GL_ARB_vertex_program / GL_NV_vertex_program */
    GLboolean VertexProgram;
    GLboolean VertexProgramPointSize;
@@ -311,7 +308,6 @@ _mesa_PushAttrib(GLbitfield mask)
       for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
          attr->Texture[i] = ctx->Texture.Unit[i].Enabled;
          attr->TexGen[i] = ctx->Texture.Unit[i].TexGenEnabled;
-         attr->TextureColorTable[i] = ctx->Texture.Unit[i].ColorTableEnabled;
       }
       /* GL_NV_vertex_program */
       attr->VertexProgram = ctx->VertexProgram.Enabled;
@@ -657,9 +653,6 @@ pop_enable_group(struct gl_context *ctx, const struct gl_enable_attrib *enable)
          _mesa_set_enable(ctx, GL_TEXTURE_GEN_Q,
                           (genEnabled & Q_BIT) ? GL_TRUE : GL_FALSE);
       }
-
-      /* GL_SGI_texture_color_table */
-      ctx->Texture.Unit[i].ColorTableEnabled = enable->TextureColorTable[i];
    }
 
    _mesa_ActiveTextureARB(GL_TEXTURE0 + curTexUnitSave);
@@ -702,10 +695,6 @@ pop_texture_group(struct gl_context *ctx, struct texture_state *texstate)
                  (unit->Enabled & TEXTURE_2D_ARRAY_BIT) ? GL_TRUE : GL_FALSE);
       }
 
-      if (ctx->Extensions.SGI_texture_color_table) {
-         _mesa_set_enable(ctx, GL_TEXTURE_COLOR_TABLE_SGI,
-                          unit->ColorTableEnabled);
-      }
       _mesa_TexEnvi(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, unit->EnvMode);
       _mesa_TexEnvfv(GL_TEXTURE_ENV, GL_TEXTURE_ENV_COLOR, unit->EnvColor);
       _mesa_TexGeni(GL_S, GL_TEXTURE_GEN_MODE, unit->GenS.Mode);
diff --git a/src/mesa/main/bufferobj.c b/src/mesa/main/bufferobj.c
index 75afae0add1..35d92616f45 100644
--- a/src/mesa/main/bufferobj.c
+++ b/src/mesa/main/bufferobj.c
@@ -658,248 +658,6 @@ _mesa_update_default_objects_buffer_objects(struct gl_context *ctx)
 }
 
 
-/**
- * When we're about to read pixel data out of a PBO (via glDrawPixels,
- * glTexImage, etc) or write data into a PBO (via glReadPixels,
- * glGetTexImage, etc) we call this function to check that we're not
- * going to read out of bounds.
- *
- * XXX This would also be a convenient time to check that the PBO isn't
- * currently mapped.  Whoever calls this function should check for that.
- * Remember, we can't use a PBO when it's mapped!
- *
- * If we're not using a PBO, this is a no-op.
- *
- * \param width  width of image to read/write
- * \param height  height of image to read/write
- * \param depth  depth of image to read/write
- * \param format  format of image to read/write
- * \param type  datatype of image to read/write
- * \param ptr  the user-provided pointer/offset
- * \return GL_TRUE if the PBO access is OK, GL_FALSE if the access would
- *         go out of bounds.
- */
-GLboolean
-_mesa_validate_pbo_access(GLuint dimensions,
-                          const struct gl_pixelstore_attrib *pack,
-                          GLsizei width, GLsizei height, GLsizei depth,
-                          GLenum format, GLenum type, const GLvoid *ptr)
-{
-   GLvoid *start, *end;
-   const GLubyte *sizeAddr; /* buffer size, cast to a pointer */
-
-   if (!_mesa_is_bufferobj(pack->BufferObj))
-      return GL_TRUE;  /* no PBO, OK */
-
-   if (pack->BufferObj->Size == 0)
-      /* no buffer! */
-      return GL_FALSE;
-
-   /* get address of first pixel we'll read */
-   start = _mesa_image_address(dimensions, pack, ptr, width, height,
-                               format, type, 0, 0, 0);
-
-   /* get address just past the last pixel we'll read */
-   end =  _mesa_image_address(dimensions, pack, ptr, width, height,
-                              format, type, depth-1, height-1, width);
-
-
-   sizeAddr = ((const GLubyte *) 0) + pack->BufferObj->Size;
-
-   if ((const GLubyte *) start > sizeAddr) {
-      /* This will catch negative values / wrap-around */
-      return GL_FALSE;
-   }
-   if ((const GLubyte *) end > sizeAddr) {
-      /* Image read goes beyond end of buffer */
-      return GL_FALSE;
-   }
-
-   /* OK! */
-   return GL_TRUE;
-}
-
-
-/**
- * For commands that read from a PBO (glDrawPixels, glTexImage,
- * glPolygonStipple, etc), if we're reading from a PBO, map it read-only
- * and return the pointer into the PBO.  If we're not reading from a
- * PBO, return \p src as-is.
- * If non-null return, must call _mesa_unmap_pbo_source() when done.
- *
- * \return NULL if error, else pointer to start of data
- */
-const GLvoid *
-_mesa_map_pbo_source(struct gl_context *ctx,
-                     const struct gl_pixelstore_attrib *unpack,
-                     const GLvoid *src)
-{
-   const GLubyte *buf;
-
-   if (_mesa_is_bufferobj(unpack->BufferObj)) {
-      /* unpack from PBO */
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                                              GL_READ_ONLY_ARB,
-                                              unpack->BufferObj);
-      if (!buf)
-         return NULL;
-
-      buf = ADD_POINTERS(buf, src);
-   }
-   else {
-      /* unpack from normal memory */
-      buf = src;
-   }
-
-   return buf;
-}
-
-
-/**
- * Combine PBO-read validation and mapping.
- * If any GL errors are detected, they'll be recorded and NULL returned.
- * \sa _mesa_validate_pbo_access
- * \sa _mesa_map_pbo_source
- * A call to this function should have a matching call to
- * _mesa_unmap_pbo_source().
- */
-const GLvoid *
-_mesa_map_validate_pbo_source(struct gl_context *ctx,
-                              GLuint dimensions,
-                              const struct gl_pixelstore_attrib *unpack,
-                              GLsizei width, GLsizei height, GLsizei depth,
-                              GLenum format, GLenum type, const GLvoid *ptr,
-                              const char *where)
-{
-   ASSERT(dimensions == 1 || dimensions == 2 || dimensions == 3);
-
-   if (!_mesa_is_bufferobj(unpack->BufferObj)) {
-      /* non-PBO access: no validation to be done */
-      return ptr;
-   }
-
-   if (!_mesa_validate_pbo_access(dimensions, unpack,
-                                  width, height, depth, format, type, ptr)) {
-      _mesa_error(ctx, GL_INVALID_OPERATION,
-                  "%s(out of bounds PBO access)", where);
-      return NULL;
-   }
-
-   if (_mesa_bufferobj_mapped(unpack->BufferObj)) {
-      /* buffer is already mapped - that's an error */
-      _mesa_error(ctx, GL_INVALID_OPERATION, "%s(PBO is mapped)", where);
-      return NULL;
-   }
-
-   ptr = _mesa_map_pbo_source(ctx, unpack, ptr);
-   return ptr;
-}
-
-
-/**
- * Counterpart to _mesa_map_pbo_source()
- */
-void
-_mesa_unmap_pbo_source(struct gl_context *ctx,
-                       const struct gl_pixelstore_attrib *unpack)
-{
-   ASSERT(unpack != &ctx->Pack); /* catch pack/unpack mismatch */
-   if (_mesa_is_bufferobj(unpack->BufferObj)) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                              unpack->BufferObj);
-   }
-}
-
-
-/**
- * For commands that write to a PBO (glReadPixels, glGetColorTable, etc),
- * if we're writing to a PBO, map it write-only and return the pointer
- * into the PBO.  If we're not writing to a PBO, return \p dst as-is.
- * If non-null return, must call _mesa_unmap_pbo_dest() when done.
- *
- * \return NULL if error, else pointer to start of data
- */
-void *
-_mesa_map_pbo_dest(struct gl_context *ctx,
-                   const struct gl_pixelstore_attrib *pack,
-                   GLvoid *dest)
-{
-   void *buf;
-
-   if (_mesa_is_bufferobj(pack->BufferObj)) {
-      /* pack into PBO */
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                                              GL_WRITE_ONLY_ARB,
-                                              pack->BufferObj);
-      if (!buf)
-         return NULL;
-
-      buf = ADD_POINTERS(buf, dest);
-   }
-   else {
-      /* pack to normal memory */
-      buf = dest;
-   }
-
-   return buf;
-}
-
-
-/**
- * Combine PBO-write validation and mapping.
- * If any GL errors are detected, they'll be recorded and NULL returned.
- * \sa _mesa_validate_pbo_access
- * \sa _mesa_map_pbo_dest
- * A call to this function should have a matching call to
- * _mesa_unmap_pbo_dest().
- */
-GLvoid *
-_mesa_map_validate_pbo_dest(struct gl_context *ctx,
-                            GLuint dimensions,
-                            const struct gl_pixelstore_attrib *unpack,
-                            GLsizei width, GLsizei height, GLsizei depth,
-                            GLenum format, GLenum type, GLvoid *ptr,
-                            const char *where)
-{
-   ASSERT(dimensions == 1 || dimensions == 2 || dimensions == 3);
-
-   if (!_mesa_is_bufferobj(unpack->BufferObj)) {
-      /* non-PBO access: no validation to be done */
-      return ptr;
-   }
-
-   if (!_mesa_validate_pbo_access(dimensions, unpack,
-                                  width, height, depth, format, type, ptr)) {
-      _mesa_error(ctx, GL_INVALID_OPERATION,
-                  "%s(out of bounds PBO access)", where);
-      return NULL;
-   }
-
-   if (_mesa_bufferobj_mapped(unpack->BufferObj)) {
-      /* buffer is already mapped - that's an error */
-      _mesa_error(ctx, GL_INVALID_OPERATION, "%s(PBO is mapped)", where);
-      return NULL;
-   }
-
-   ptr = _mesa_map_pbo_dest(ctx, unpack, ptr);
-   return ptr;
-}
-
-
-/**
- * Counterpart to _mesa_map_pbo_dest()
- */
-void
-_mesa_unmap_pbo_dest(struct gl_context *ctx,
-                     const struct gl_pixelstore_attrib *pack)
-{
-   ASSERT(pack != &ctx->Unpack); /* catch pack/unpack mismatch */
-   if (_mesa_is_bufferobj(pack->BufferObj)) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT, pack->BufferObj);
-   }
-}
-
-
 
 /**
  * Return the gl_buffer_object for the given ID.
diff --git a/src/mesa/main/bufferobj.h b/src/mesa/main/bufferobj.h
index 09ccab31742..91fa073b649 100644
--- a/src/mesa/main/bufferobj.h
+++ b/src/mesa/main/bufferobj.h
@@ -79,47 +79,6 @@ _mesa_reference_buffer_object(struct gl_context *ctx,
                               struct gl_buffer_object **ptr,
                               struct gl_buffer_object *bufObj);
 
-extern GLboolean
-_mesa_validate_pbo_access(GLuint dimensions,
-                          const struct gl_pixelstore_attrib *pack,
-                          GLsizei width, GLsizei height, GLsizei depth,
-                          GLenum format, GLenum type, const GLvoid *ptr);
-
-extern const GLvoid *
-_mesa_map_pbo_source(struct gl_context *ctx,
-                     const struct gl_pixelstore_attrib *unpack,
-                     const GLvoid *src);
-
-extern const GLvoid *
-_mesa_map_validate_pbo_source(struct gl_context *ctx,
-                              GLuint dimensions,
-                              const struct gl_pixelstore_attrib *unpack,
-                              GLsizei width, GLsizei height, GLsizei depth,
-                              GLenum format, GLenum type, const GLvoid *ptr,
-                              const char *where);
-
-extern void
-_mesa_unmap_pbo_source(struct gl_context *ctx,
-                       const struct gl_pixelstore_attrib *unpack);
-
-extern void *
-_mesa_map_pbo_dest(struct gl_context *ctx,
-                   const struct gl_pixelstore_attrib *pack,
-                   GLvoid *dest);
-
-extern GLvoid *
-_mesa_map_validate_pbo_dest(struct gl_context *ctx,
-                            GLuint dimensions,
-                            const struct gl_pixelstore_attrib *unpack,
-                            GLsizei width, GLsizei height, GLsizei depth,
-                            GLenum format, GLenum type, GLvoid *ptr,
-                            const char *where);
-
-extern void
-_mesa_unmap_pbo_dest(struct gl_context *ctx,
-                     const struct gl_pixelstore_attrib *pack);
-
-
 extern void
 _mesa_init_buffer_object_functions(struct dd_function_table *driver);
 
diff --git a/src/mesa/main/colortab.c b/src/mesa/main/colortab.c
index b0ba31c732f..d0c865735ac 100644
--- a/src/mesa/main/colortab.c
+++ b/src/mesa/main/colortab.c
@@ -32,6 +32,7 @@
 #include "mfeatures.h"
 #include "mtypes.h"
 #include "pack.h"
+#include "pbo.h"
 #include "state.h"
 #include "teximage.h"
 #include "texstate.h"
@@ -300,23 +301,6 @@ _mesa_ColorTable( GLenum target, GLenum internalFormat,
       case GL_SHARED_TEXTURE_PALETTE_EXT:
          table = &ctx->Texture.Palette;
          break;
-      case GL_TEXTURE_COLOR_TABLE_SGI:
-         if (!ctx->Extensions.SGI_texture_color_table) {
-            _mesa_error(ctx, GL_INVALID_ENUM, "glColorTable(target)");
-            return;
-         }
-         table = &(texUnit->ColorTable);
-         scale = ctx->Pixel.TextureColorTableScale;
-         bias = ctx->Pixel.TextureColorTableBias;
-         break;
-      case GL_PROXY_TEXTURE_COLOR_TABLE_SGI:
-         if (!ctx->Extensions.SGI_texture_color_table) {
-            _mesa_error(ctx, GL_INVALID_ENUM, "glColorTable(target)");
-            return;
-         }
-         table = &(texUnit->ProxyColorTable);
-         proxy = GL_TRUE;
-         break;
       default:
          /* try texture targets */
          {
@@ -435,15 +419,6 @@ _mesa_ColorSubTable( GLenum target, GLsizei start,
       case GL_SHARED_TEXTURE_PALETTE_EXT:
          table = &ctx->Texture.Palette;
          break;
-      case GL_TEXTURE_COLOR_TABLE_SGI:
-         if (!ctx->Extensions.SGI_texture_color_table) {
-            _mesa_error(ctx, GL_INVALID_ENUM, "glColorSubTable(target)");
-            return;
-         }
-         table = &(texUnit->ColorTable);
-         scale = ctx->Pixel.TextureColorTableScale;
-         bias = ctx->Pixel.TextureColorTableBias;
-         break;
       default:
          /* try texture targets */
          texObj = _mesa_select_tex_object(ctx, texUnit, target);
@@ -551,13 +526,6 @@ _mesa_GetColorTable( GLenum target, GLenum format,
       case GL_SHARED_TEXTURE_PALETTE_EXT:
          table = &ctx->Texture.Palette;
          break;
-      case GL_TEXTURE_COLOR_TABLE_SGI:
-         if (!ctx->Extensions.SGI_texture_color_table) {
-            _mesa_error(ctx, GL_INVALID_ENUM, "glGetColorTable(target)");
-            return;
-         }
-         table = &(texUnit->ColorTable);
-         break;
       default:
          /* try texture targets */
          {
@@ -661,32 +629,10 @@ _mesa_GetColorTable( GLenum target, GLenum format,
 static void GLAPIENTRY
 _mesa_ColorTableParameterfv(GLenum target, GLenum pname, const GLfloat *params)
 {
-   GLfloat *scale, *bias;
+   /* no extensions use this function */
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
-
-   switch (target) {
-   case GL_TEXTURE_COLOR_TABLE_SGI:
-      scale = ctx->Pixel.TextureColorTableScale;
-      bias  = ctx->Pixel.TextureColorTableBias;
-      break;
-   default:
-      _mesa_error(ctx, GL_INVALID_ENUM, "glColorTableParameter(target)");
-      return;
-   }
-
-   if (pname == GL_COLOR_TABLE_SCALE_SGI) {
-      COPY_4V(scale, params);
-   }
-   else if (pname == GL_COLOR_TABLE_BIAS_SGI) {
-      COPY_4V(bias, params);
-   }
-   else {
-      _mesa_error(ctx, GL_INVALID_ENUM, "glColorTableParameterfv(pname)");
-      return;
-   }
-
-   ctx->NewState |= _NEW_PIXEL;
+   _mesa_error(ctx, GL_INVALID_ENUM, "glColorTableParameterfv(target)");
 }
 
 
@@ -694,19 +640,10 @@ _mesa_ColorTableParameterfv(GLenum target, GLenum pname, const GLfloat *params)
 static void GLAPIENTRY
 _mesa_ColorTableParameteriv(GLenum target, GLenum pname, const GLint *params)
 {
-   GLfloat fparams[4];
-   if (pname == GL_TEXTURE_COLOR_TABLE_SGI) {
-      /* four values */
-      fparams[0] = (GLfloat) params[0];
-      fparams[1] = (GLfloat) params[1];
-      fparams[2] = (GLfloat) params[2];
-      fparams[3] = (GLfloat) params[3];
-   }
-   else {
-      /* one values */
-      fparams[0] = (GLfloat) params[0];
-   }
-   _mesa_ColorTableParameterfv(target, pname, fparams);
+   /* no extensions use this function */
+   GET_CURRENT_CONTEXT(ctx);
+   ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
+   _mesa_error(ctx, GL_INVALID_ENUM, "glColorTableParameteriv(target)");
 }
 
 
@@ -723,28 +660,6 @@ _mesa_GetColorTableParameterfv( GLenum target, GLenum pname, GLfloat *params )
       case GL_SHARED_TEXTURE_PALETTE_EXT:
          table = &ctx->Texture.Palette;
          break;
-      case GL_TEXTURE_COLOR_TABLE_SGI:
-         if (!ctx->Extensions.SGI_texture_color_table) {
-            _mesa_error(ctx, GL_INVALID_ENUM, "glGetColorTableParameter(target)");
-            return;
-         }
-         table = &(texUnit->ColorTable);
-         if (pname == GL_COLOR_TABLE_SCALE_SGI) {
-            COPY_4V(params, ctx->Pixel.TextureColorTableScale);
-            return;
-         }
-         else if (pname == GL_COLOR_TABLE_BIAS_SGI) {
-            COPY_4V(params, ctx->Pixel.TextureColorTableBias);
-            return;
-         }
-         break;
-      case GL_PROXY_TEXTURE_COLOR_TABLE_SGI:
-         if (!ctx->Extensions.SGI_texture_color_table) {
-            _mesa_error(ctx, GL_INVALID_ENUM, "glGetColorTableParameter(target)");
-            return;
-         }
-         table = &(texUnit->ProxyColorTable);
-         break;
       default:
          /* try texture targets */
          {
@@ -808,34 +723,6 @@ _mesa_GetColorTableParameteriv( GLenum target, GLenum pname, GLint *params )
       case GL_SHARED_TEXTURE_PALETTE_EXT:
          table = &ctx->Texture.Palette;
          break;
-      case GL_TEXTURE_COLOR_TABLE_SGI:
-         if (!ctx->Extensions.SGI_texture_color_table) {
-            _mesa_error(ctx, GL_INVALID_ENUM, "glGetColorTableParameter(target)");
-            return;
-         }
-         table = &(texUnit->ColorTable);
-         if (pname == GL_COLOR_TABLE_SCALE_SGI) {
-            params[0] = (GLint) ctx->Pixel.TextureColorTableScale[0];
-            params[1] = (GLint) ctx->Pixel.TextureColorTableScale[1];
-            params[2] = (GLint) ctx->Pixel.TextureColorTableScale[2];
-            params[3] = (GLint) ctx->Pixel.TextureColorTableScale[3];
-            return;
-         }
-         else if (pname == GL_COLOR_TABLE_BIAS_SGI) {
-            params[0] = (GLint) ctx->Pixel.TextureColorTableBias[0];
-            params[1] = (GLint) ctx->Pixel.TextureColorTableBias[1];
-            params[2] = (GLint) ctx->Pixel.TextureColorTableBias[2];
-            params[3] = (GLint) ctx->Pixel.TextureColorTableBias[3];
-            return;
-         }
-         break;
-      case GL_PROXY_TEXTURE_COLOR_TABLE_SGI:
-         if (!ctx->Extensions.SGI_texture_color_table) {
-            _mesa_error(ctx, GL_INVALID_ENUM, "glGetColorTableParameter(target)");
-            return;
-         }
-         table = &(texUnit->ProxyColorTable);
-         break;
       default:
          /* Try texture targets */
          {
diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c
index a942314552e..5d581c84002 100644
--- a/src/mesa/main/context.c
+++ b/src/mesa/main/context.c
@@ -487,6 +487,7 @@ init_program_limits(GLenum type, struct gl_program_constants *prog)
    prog->MaxEnvParams = MAX_PROGRAM_ENV_PARAMS;
    prog->MaxLocalParams = MAX_PROGRAM_LOCAL_PARAMS;
    prog->MaxUniformComponents = 4 * MAX_UNIFORMS;
+   prog->MaxAddressOffset = MAX_PROGRAM_LOCAL_PARAMS;
 
    switch (type) {
    case GL_VERTEX_PROGRAM_ARB:
diff --git a/src/mesa/main/dlist.c b/src/mesa/main/dlist.c
index 8e904c7787a..4e463dd06a1 100644
--- a/src/mesa/main/dlist.c
+++ b/src/mesa/main/dlist.c
@@ -54,6 +54,7 @@
 #include "light.h"
 #include "macros.h"
 #include "pack.h"
+#include "pbo.h"
 #include "queryobj.h"
 #include "teximage.h"
 #include "mtypes.h"
diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c
index df4712de894..98e82ef852b 100644
--- a/src/mesa/main/drawpix.c
+++ b/src/mesa/main/drawpix.c
@@ -31,6 +31,7 @@
 #include "feedback.h"
 #include "framebuffer.h"
 #include "mfeatures.h"
+#include "pbo.h"
 #include "readpix.h"
 #include "state.h"
 #include "dispatch.h"
diff --git a/src/mesa/main/enable.c b/src/mesa/main/enable.c
index ea3b8214c3b..f247092410b 100644
--- a/src/mesa/main/enable.c
+++ b/src/mesa/main/enable.c
@@ -703,15 +703,6 @@ _mesa_set_enable(struct gl_context *ctx, GLenum cap, GLboolean state)
          client_state( ctx, cap, state );
          return;
 
-      /* GL_SGI_texture_color_table */
-      case GL_TEXTURE_COLOR_TABLE_SGI:
-         CHECK_EXTENSION(SGI_texture_color_table, cap);
-         if (ctx->Texture.Unit[ctx->Texture.CurrentUnit].ColorTableEnabled == state)
-            return;
-         FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-         ctx->Texture.Unit[ctx->Texture.CurrentUnit].ColorTableEnabled = state;
-         break;
-
       /* GL_ARB_texture_cube_map */
       case GL_TEXTURE_CUBE_MAP_ARB:
          CHECK_EXTENSION(ARB_texture_cube_map, cap);
@@ -1308,11 +1299,6 @@ _mesa_IsEnabled( GLenum cap )
          return (ctx->Array.ArrayObj->PointSize.Enabled != 0);
 #endif
 
-      /* GL_SGI_texture_color_table */
-      case GL_TEXTURE_COLOR_TABLE_SGI:
-         CHECK_EXTENSION(SGI_texture_color_table);
-         return ctx->Texture.Unit[ctx->Texture.CurrentUnit].ColorTableEnabled;
-
       /* GL_ARB_texture_cube_map */
       case GL_TEXTURE_CUBE_MAP_ARB:
          CHECK_EXTENSION(ARB_texture_cube_map);
diff --git a/src/mesa/main/extensions.c b/src/mesa/main/extensions.c
index b8bb2555acd..3840cdc5d36 100644
--- a/src/mesa/main/extensions.c
+++ b/src/mesa/main/extensions.c
@@ -290,7 +290,6 @@ static const struct extension extension_table[] = {
    { "GL_SGIS_texture_border_clamp",               o(ARB_texture_border_clamp),                GL             },
    { "GL_SGIS_texture_edge_clamp",                 o(SGIS_texture_edge_clamp),                 GL             },
    { "GL_SGIS_texture_lod",                        o(SGIS_texture_lod),                        GL             },
-   { "GL_SGI_texture_color_table",                 o(SGI_texture_color_table),                 GL             },
    { "GL_SUN_multi_draw_arrays",                   o(EXT_multi_draw_arrays),                   GL             },
 
    { 0, 0, 0 },
@@ -520,7 +519,6 @@ _mesa_enable_sw_extensions(struct gl_context *ctx)
 #if FEATURE_NV_fragment_program && FEATURE_ARB_fragment_program
    ctx->Extensions.NV_fragment_program_option = GL_TRUE;
 #endif
-   ctx->Extensions.SGI_texture_color_table = GL_TRUE;
    /*ctx->Extensions.SGIS_generate_mipmap = GL_TRUE;*/
    ctx->Extensions.SGIS_texture_edge_clamp = GL_TRUE;
 #if FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program
diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c
index ab6b2a9b17b..c242e568d08 100644
--- a/src/mesa/main/fbobject.c
+++ b/src/mesa/main/fbobject.c
@@ -160,6 +160,29 @@ invalidate_framebuffer(struct gl_framebuffer *fb)
 
 
 /**
+ * Return the gl_framebuffer object which corresponds to the given
+ * framebuffer target, such as GL_DRAW_FRAMEBUFFER.
+ * Check support for GL_EXT_framebuffer_blit to determine if certain
+ * targets are legal.
+ * \return gl_framebuffer pointer or NULL if target is illegal
+ */
+static struct gl_framebuffer *
+get_framebuffer_target(struct gl_context *ctx, GLenum target)
+{
+   switch (target) {
+   case GL_DRAW_FRAMEBUFFER:
+      return ctx->Extensions.EXT_framebuffer_blit ? ctx->DrawBuffer : NULL;
+   case GL_READ_FRAMEBUFFER:
+      return ctx->Extensions.EXT_framebuffer_blit ? ctx->ReadBuffer : NULL;
+   case GL_FRAMEBUFFER_EXT:
+      return ctx->DrawBuffer;
+   default:
+      return NULL;
+   }
+}
+
+
+/**
  * Given a GL_*_ATTACHMENTn token, return a pointer to the corresponding
  * gl_renderbuffer_attachment object.
  * This function is only used for user-created FB objects, not the
@@ -652,7 +675,7 @@ _mesa_test_framebuffer_completeness(struct gl_context *ctx,
    for (i = -2; i < (GLint) ctx->Const.MaxColorAttachments; i++) {
       struct gl_renderbuffer_attachment *att;
       GLenum f;
-      gl_format mesaFormat;
+      gl_format attFormat;
 
       /*
        * XXX for ARB_fbo, only check color buffers that are named by
@@ -699,7 +722,7 @@ _mesa_test_framebuffer_completeness(struct gl_context *ctx,
          minHeight = MIN2(minHeight, texImg->Height);
          maxHeight = MAX2(maxHeight, texImg->Height);
          f = texImg->_BaseFormat;
-         mesaFormat = texImg->TexFormat;
+         attFormat = texImg->TexFormat;
          numImages++;
          if (!_mesa_is_legal_color_format(ctx, f) &&
              !is_legal_depth_format(ctx, f)) {
@@ -714,7 +737,7 @@ _mesa_test_framebuffer_completeness(struct gl_context *ctx,
          minHeight = MIN2(minHeight, att->Renderbuffer->Height);
          maxHeight = MAX2(minHeight, att->Renderbuffer->Height);
          f = att->Renderbuffer->InternalFormat;
-         mesaFormat = att->Renderbuffer->Format;
+         attFormat = att->Renderbuffer->Format;
          numImages++;
       }
       else {
@@ -728,7 +751,7 @@ _mesa_test_framebuffer_completeness(struct gl_context *ctx,
       }
 
       /* check if integer color */
-      fb->_IntegerColor = _mesa_is_format_integer_color(mesaFormat);
+      fb->_IntegerColor = _mesa_is_format_integer_color(attFormat);
 
       /* Error-check width, height, format, samples
        */
@@ -1677,29 +1700,10 @@ _mesa_CheckFramebufferStatusEXT(GLenum target)
 
    ASSERT_OUTSIDE_BEGIN_END_WITH_RETVAL(ctx, 0);
 
-   switch (target) {
-#if FEATURE_EXT_framebuffer_blit
-   case GL_DRAW_FRAMEBUFFER_EXT:
-      if (!ctx->Extensions.EXT_framebuffer_blit) {
-         _mesa_error(ctx, GL_INVALID_ENUM, "glCheckFramebufferStatus(target)");
-         return 0;
-      }
-      buffer = ctx->DrawBuffer;
-      break;
-   case GL_READ_FRAMEBUFFER_EXT:
-      if (!ctx->Extensions.EXT_framebuffer_blit) {
-         _mesa_error(ctx, GL_INVALID_ENUM, "glCheckFramebufferStatus(target)");
-         return 0;
-      }
-      buffer = ctx->ReadBuffer;
-      break;
-#endif
-   case GL_FRAMEBUFFER_EXT:
-      buffer = ctx->DrawBuffer;
-      break;
-   default:
+   buffer = get_framebuffer_target(ctx, target);
+   if (!buffer) {
       _mesa_error(ctx, GL_INVALID_ENUM, "glCheckFramebufferStatus(target)");
-      return 0; /* formerly GL_FRAMEBUFFER_STATUS_ERROR_EXT */
+      return 0;
    }
 
    if (buffer->Name == 0) {
@@ -1729,33 +1733,16 @@ framebuffer_texture(struct gl_context *ctx, const char *caller, GLenum target,
    struct gl_renderbuffer_attachment *att;
    struct gl_texture_object *texObj = NULL;
    struct gl_framebuffer *fb;
-   GLboolean error = GL_FALSE;
 
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   switch (target) {
-   case GL_READ_FRAMEBUFFER_EXT:
-      error = !ctx->Extensions.EXT_framebuffer_blit;
-      fb = ctx->ReadBuffer;
-      break;
-   case GL_DRAW_FRAMEBUFFER_EXT:
-      error = !ctx->Extensions.EXT_framebuffer_blit;
-      /* fall-through */
-   case GL_FRAMEBUFFER_EXT:
-      fb = ctx->DrawBuffer;
-      break;
-   default:
-      error = GL_TRUE;
-   }
-
-   if (error) {
+   fb = get_framebuffer_target(ctx, target);
+   if (!fb) {
       _mesa_error(ctx, GL_INVALID_ENUM,
                   "glFramebufferTexture%sEXT(target=0x%x)", caller, target);
       return;
    }
 
-   ASSERT(fb);
-
    /* check framebuffer binding */
    if (fb->Name == 0) {
       _mesa_error(ctx, GL_INVALID_OPERATION,
@@ -1936,31 +1923,9 @@ _mesa_FramebufferRenderbufferEXT(GLenum target, GLenum attachment,
 
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   switch (target) {
-#if FEATURE_EXT_framebuffer_blit
-   case GL_DRAW_FRAMEBUFFER_EXT:
-      if (!ctx->Extensions.EXT_framebuffer_blit) {
-         _mesa_error(ctx, GL_INVALID_ENUM,
-                     "glFramebufferRenderbufferEXT(target)");
-         return;
-      }
-      fb = ctx->DrawBuffer;
-      break;
-   case GL_READ_FRAMEBUFFER_EXT:
-      if (!ctx->Extensions.EXT_framebuffer_blit) {
-         _mesa_error(ctx, GL_INVALID_ENUM,
-                     "glFramebufferRenderbufferEXT(target)");
-         return;
-      }
-      fb = ctx->ReadBuffer;
-      break;
-#endif
-   case GL_FRAMEBUFFER_EXT:
-      fb = ctx->DrawBuffer;
-      break;
-   default:
-      _mesa_error(ctx, GL_INVALID_ENUM,
-                  "glFramebufferRenderbufferEXT(target)");
+   fb = get_framebuffer_target(ctx, target);
+   if (!fb) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glFramebufferRenderbufferEXT(target)");
       return;
    }
 
@@ -2040,29 +2005,8 @@ _mesa_GetFramebufferAttachmentParameterivEXT(GLenum target, GLenum attachment,
 
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   switch (target) {
-#if FEATURE_EXT_framebuffer_blit
-   case GL_DRAW_FRAMEBUFFER_EXT:
-      if (!ctx->Extensions.EXT_framebuffer_blit) {
-         _mesa_error(ctx, GL_INVALID_ENUM,
-                     "glGetFramebufferAttachmentParameterivEXT(target)");
-         return;
-      }
-      buffer = ctx->DrawBuffer;
-      break;
-   case GL_READ_FRAMEBUFFER_EXT:
-      if (!ctx->Extensions.EXT_framebuffer_blit) {
-         _mesa_error(ctx, GL_INVALID_ENUM,
-                     "glGetFramebufferAttachmentParameterivEXT(target)");
-         return;
-      }
-      buffer = ctx->ReadBuffer;
-      break;
-#endif
-   case GL_FRAMEBUFFER_EXT:
-      buffer = ctx->DrawBuffer;
-      break;
-   default:
+   buffer = get_framebuffer_target(ctx, target);
+   if (!buffer) {
       _mesa_error(ctx, GL_INVALID_ENUM,
                   "glGetFramebufferAttachmentParameterivEXT(target)");
       return;
diff --git a/src/mesa/main/framebuffer.c b/src/mesa/main/framebuffer.c
index 948b3b7b5a2..8916441f455 100644
--- a/src/mesa/main/framebuffer.c
+++ b/src/mesa/main/framebuffer.c
@@ -794,9 +794,6 @@ update_framebuffer(struct gl_context *ctx, struct gl_framebuffer *fb)
          _mesa_drawbuffers(ctx, ctx->Const.MaxDrawBuffers,
                            ctx->Color.DrawBuffer, NULL);
       }
-      if (fb->ColorReadBuffer != ctx->Pixel.ReadBuffer) {
-         
-      }
    }
    else {
       /* This is a user-created framebuffer.
diff --git a/src/mesa/main/get.c b/src/mesa/main/get.c
index fa7aa1121aa..336304c591d 100644
--- a/src/mesa/main/get.c
+++ b/src/mesa/main/get.c
@@ -290,7 +290,6 @@ EXTRA_EXT(NV_fragment_program);
 EXTRA_EXT(NV_texture_rectangle);
 EXTRA_EXT(EXT_stencil_two_side);
 EXTRA_EXT(NV_light_max_exponent);
-EXTRA_EXT(SGI_texture_color_table);
 EXTRA_EXT(EXT_depth_bounds_test);
 EXTRA_EXT(ARB_depth_clamp);
 EXTRA_EXT(ATI_fragment_shader);
@@ -911,11 +910,6 @@ static const struct value_desc values[] = {
      CONTEXT_MATRIX_T(ProjectionMatrixStack.Top), NO_EXTRA },
    { GL_TRANSPOSE_TEXTURE_MATRIX_ARB, CONTEXT_MATRIX_T(TextureMatrixStack), NO_EXTRA },
 
-   /* GL_SGI_texture_color_table */
-   { GL_TEXTURE_COLOR_TABLE_SGI, LOC_TEXUNIT, TYPE_BOOLEAN,
-     offsetof(struct gl_texture_unit, ColorTableEnabled),
-     extra_SGI_texture_color_table },
-
    /* GL_EXT_secondary_color */
    { GL_COLOR_SUM_EXT, CONTEXT_BOOL(Fog.ColorSumEnabled),
      extra_EXT_secondary_color_ARB_vertex_program },
diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c
index d8a56103800..0727e1818f1 100644
--- a/src/mesa/main/mipmap.c
+++ b/src/mesa/main/mipmap.c
@@ -1755,8 +1755,13 @@ _mesa_generate_mipmap(struct gl_context *ctx, GLenum target,
       if (srcImage->_BaseFormat == GL_RGB) {
          convertFormat = MESA_FORMAT_RGB888;
          components = 3;
-      }
-      else if (srcImage->_BaseFormat == GL_RGBA) {
+      } else if (srcImage->_BaseFormat == GL_RED) {
+         convertFormat = MESA_FORMAT_R8;
+         components = 1;
+      } else if (srcImage->_BaseFormat == GL_RG) {
+         convertFormat = MESA_FORMAT_RG88;
+         components = 2;
+      } else if (srcImage->_BaseFormat == GL_RGBA) {
          convertFormat = MESA_FORMAT_RGBA8888;
          components = 4;
       }
diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
index b237063a6d8..db3eba20c68 100644
--- a/src/mesa/main/mtypes.h
+++ b/src/mesa/main/mtypes.h
@@ -1031,10 +1031,6 @@ struct gl_pixel_attrib
 
    /** glPixelZoom */
    GLfloat ZoomX, ZoomY;
-
-   /** GL_SGI_texture_color_table */
-   GLfloat TextureColorTableScale[4]; /**< RGBA */
-   GLfloat TextureColorTableBias[4];  /**< RGBA */
 };
 
 
@@ -1435,13 +1431,6 @@ struct gl_texture_unit
 
    /** Points to highest priority, complete and enabled texture object */
    struct gl_texture_object *_Current;
-
-   /** GL_SGI_texture_color_table */
-   /*@{*/
-   struct gl_color_table ColorTable;
-   struct gl_color_table ProxyColorTable;
-   GLboolean ColorTableEnabled;
-   /*@}*/
 };
 
 
@@ -2584,6 +2573,7 @@ struct gl_program_constants
    GLuint MaxAttribs;
    GLuint MaxTemps;
    GLuint MaxAddressRegs;
+   GLuint MaxAddressOffset;  /**< [-MaxAddressOffset, MaxAddressOffset-1] */
    GLuint MaxParameters;
    GLuint MaxLocalParams;
    GLuint MaxEnvParams;
@@ -2859,7 +2849,6 @@ struct gl_extensions
    GLboolean NV_vertex_program;
    GLboolean NV_vertex_program1_1;
    GLboolean OES_read_format;
-   GLboolean SGI_texture_color_table;
    GLboolean SGIS_generate_mipmap;
    GLboolean SGIS_texture_edge_clamp;
    GLboolean SGIS_texture_lod;
diff --git a/src/mesa/main/pbo.c b/src/mesa/main/pbo.c
new file mode 100644
index 00000000000..dc00d423ba9
--- /dev/null
+++ b/src/mesa/main/pbo.c
@@ -0,0 +1,372 @@
+/*
+ * Mesa 3-D graphics library
+ *
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009-2011  VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/**
+ * \file pbo.c
+ * \brief Functions related to Pixel Buffer Objects.
+ */
+
+
+
+#include "glheader.h"
+#include "bufferobj.h"
+#include "image.h"
+#include "imports.h"
+#include "mtypes.h"
+#include "pbo.h"
+
+
+
+/**
+ * When we're about to read pixel data out of a PBO (via glDrawPixels,
+ * glTexImage, etc) or write data into a PBO (via glReadPixels,
+ * glGetTexImage, etc) we call this function to check that we're not
+ * going to read out of bounds.
+ *
+ * XXX This would also be a convenient time to check that the PBO isn't
+ * currently mapped.  Whoever calls this function should check for that.
+ * Remember, we can't use a PBO when it's mapped!
+ *
+ * If we're not using a PBO, this is a no-op.
+ *
+ * \param width  width of image to read/write
+ * \param height  height of image to read/write
+ * \param depth  depth of image to read/write
+ * \param format  format of image to read/write
+ * \param type  datatype of image to read/write
+ * \param ptr  the user-provided pointer/offset
+ * \return GL_TRUE if the PBO access is OK, GL_FALSE if the access would
+ *         go out of bounds.
+ */
+GLboolean
+_mesa_validate_pbo_access(GLuint dimensions,
+                          const struct gl_pixelstore_attrib *pack,
+                          GLsizei width, GLsizei height, GLsizei depth,
+                          GLenum format, GLenum type, const GLvoid *ptr)
+{
+   GLvoid *start, *end;
+   const GLubyte *sizeAddr; /* buffer size, cast to a pointer */
+
+   if (!_mesa_is_bufferobj(pack->BufferObj))
+      return GL_TRUE;  /* no PBO, OK */
+
+   if (pack->BufferObj->Size == 0)
+      /* no buffer! */
+      return GL_FALSE;
+
+   /* get address of first pixel we'll read */
+   start = _mesa_image_address(dimensions, pack, ptr, width, height,
+                               format, type, 0, 0, 0);
+
+   /* get address just past the last pixel we'll read */
+   end =  _mesa_image_address(dimensions, pack, ptr, width, height,
+                              format, type, depth-1, height-1, width);
+
+
+   sizeAddr = ((const GLubyte *) 0) + pack->BufferObj->Size;
+
+   if ((const GLubyte *) start > sizeAddr) {
+      /* This will catch negative values / wrap-around */
+      return GL_FALSE;
+   }
+   if ((const GLubyte *) end > sizeAddr) {
+      /* Image read goes beyond end of buffer */
+      return GL_FALSE;
+   }
+
+   /* OK! */
+   return GL_TRUE;
+}
+
+
+/**
+ * For commands that read from a PBO (glDrawPixels, glTexImage,
+ * glPolygonStipple, etc), if we're reading from a PBO, map it read-only
+ * and return the pointer into the PBO.  If we're not reading from a
+ * PBO, return \p src as-is.
+ * If non-null return, must call _mesa_unmap_pbo_source() when done.
+ *
+ * \return NULL if error, else pointer to start of data
+ */
+const GLvoid *
+_mesa_map_pbo_source(struct gl_context *ctx,
+                     const struct gl_pixelstore_attrib *unpack,
+                     const GLvoid *src)
+{
+   const GLubyte *buf;
+
+   if (_mesa_is_bufferobj(unpack->BufferObj)) {
+      /* unpack from PBO */
+      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
+                                              GL_READ_ONLY_ARB,
+                                              unpack->BufferObj);
+      if (!buf)
+         return NULL;
+
+      buf = ADD_POINTERS(buf, src);
+   }
+   else {
+      /* unpack from normal memory */
+      buf = src;
+   }
+
+   return buf;
+}
+
+
+/**
+ * Combine PBO-read validation and mapping.
+ * If any GL errors are detected, they'll be recorded and NULL returned.
+ * \sa _mesa_validate_pbo_access
+ * \sa _mesa_map_pbo_source
+ * A call to this function should have a matching call to
+ * _mesa_unmap_pbo_source().
+ */
+const GLvoid *
+_mesa_map_validate_pbo_source(struct gl_context *ctx,
+                              GLuint dimensions,
+                              const struct gl_pixelstore_attrib *unpack,
+                              GLsizei width, GLsizei height, GLsizei depth,
+                              GLenum format, GLenum type, const GLvoid *ptr,
+                              const char *where)
+{
+   ASSERT(dimensions == 1 || dimensions == 2 || dimensions == 3);
+
+   if (!_mesa_is_bufferobj(unpack->BufferObj)) {
+      /* non-PBO access: no validation to be done */
+      return ptr;
+   }
+
+   if (!_mesa_validate_pbo_access(dimensions, unpack,
+                                  width, height, depth, format, type, ptr)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "%s(out of bounds PBO access)", where);
+      return NULL;
+   }
+
+   if (_mesa_bufferobj_mapped(unpack->BufferObj)) {
+      /* buffer is already mapped - that's an error */
+      _mesa_error(ctx, GL_INVALID_OPERATION, "%s(PBO is mapped)", where);
+      return NULL;
+   }
+
+   ptr = _mesa_map_pbo_source(ctx, unpack, ptr);
+   return ptr;
+}
+
+
+/**
+ * Counterpart to _mesa_map_pbo_source()
+ */
+void
+_mesa_unmap_pbo_source(struct gl_context *ctx,
+                       const struct gl_pixelstore_attrib *unpack)
+{
+   ASSERT(unpack != &ctx->Pack); /* catch pack/unpack mismatch */
+   if (_mesa_is_bufferobj(unpack->BufferObj)) {
+      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
+                              unpack->BufferObj);
+   }
+}
+
+
+/**
+ * For commands that write to a PBO (glReadPixels, glGetColorTable, etc),
+ * if we're writing to a PBO, map it write-only and return the pointer
+ * into the PBO.  If we're not writing to a PBO, return \p dst as-is.
+ * If non-null return, must call _mesa_unmap_pbo_dest() when done.
+ *
+ * \return NULL if error, else pointer to start of data
+ */
+void *
+_mesa_map_pbo_dest(struct gl_context *ctx,
+                   const struct gl_pixelstore_attrib *pack,
+                   GLvoid *dest)
+{
+   void *buf;
+
+   if (_mesa_is_bufferobj(pack->BufferObj)) {
+      /* pack into PBO */
+      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
+                                              GL_WRITE_ONLY_ARB,
+                                              pack->BufferObj);
+      if (!buf)
+         return NULL;
+
+      buf = ADD_POINTERS(buf, dest);
+   }
+   else {
+      /* pack to normal memory */
+      buf = dest;
+   }
+
+   return buf;
+}
+
+
+/**
+ * Combine PBO-write validation and mapping.
+ * If any GL errors are detected, they'll be recorded and NULL returned.
+ * \sa _mesa_validate_pbo_access
+ * \sa _mesa_map_pbo_dest
+ * A call to this function should have a matching call to
+ * _mesa_unmap_pbo_dest().
+ */
+GLvoid *
+_mesa_map_validate_pbo_dest(struct gl_context *ctx,
+                            GLuint dimensions,
+                            const struct gl_pixelstore_attrib *unpack,
+                            GLsizei width, GLsizei height, GLsizei depth,
+                            GLenum format, GLenum type, GLvoid *ptr,
+                            const char *where)
+{
+   ASSERT(dimensions == 1 || dimensions == 2 || dimensions == 3);
+
+   if (!_mesa_is_bufferobj(unpack->BufferObj)) {
+      /* non-PBO access: no validation to be done */
+      return ptr;
+   }
+
+   if (!_mesa_validate_pbo_access(dimensions, unpack,
+                                  width, height, depth, format, type, ptr)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "%s(out of bounds PBO access)", where);
+      return NULL;
+   }
+
+   if (_mesa_bufferobj_mapped(unpack->BufferObj)) {
+      /* buffer is already mapped - that's an error */
+      _mesa_error(ctx, GL_INVALID_OPERATION, "%s(PBO is mapped)", where);
+      return NULL;
+   }
+
+   ptr = _mesa_map_pbo_dest(ctx, unpack, ptr);
+   return ptr;
+}
+
+
+/**
+ * Counterpart to _mesa_map_pbo_dest()
+ */
+void
+_mesa_unmap_pbo_dest(struct gl_context *ctx,
+                     const struct gl_pixelstore_attrib *pack)
+{
+   ASSERT(pack != &ctx->Unpack); /* catch pack/unpack mismatch */
+   if (_mesa_is_bufferobj(pack->BufferObj)) {
+      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT, pack->BufferObj);
+   }
+}
+
+
+
+/**
+ * Check if an unpack PBO is active prior to fetching a texture image.
+ * If so, do bounds checking and map the buffer into main memory.
+ * Any errors detected will be recorded.
+ * The caller _must_ call _mesa_unmap_teximage_pbo() too!
+ */
+const GLvoid *
+_mesa_validate_pbo_teximage(struct gl_context *ctx, GLuint dimensions,
+			    GLsizei width, GLsizei height, GLsizei depth,
+			    GLenum format, GLenum type, const GLvoid *pixels,
+			    const struct gl_pixelstore_attrib *unpack,
+			    const char *funcName)
+{
+   GLubyte *buf;
+
+   if (!_mesa_is_bufferobj(unpack->BufferObj)) {
+      /* no PBO */
+      return pixels;
+   }
+   if (!_mesa_validate_pbo_access(dimensions, unpack, width, height, depth,
+                                  format, type, pixels)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, funcName, "(invalid PBO access)");
+      return NULL;
+   }
+
+   buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
+                                          GL_READ_ONLY_ARB, unpack->BufferObj);
+   if (!buf) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, funcName, "(PBO is mapped)");
+      return NULL;
+   }
+
+   return ADD_POINTERS(buf, pixels);
+}
+
+
+/**
+ * Check if an unpack PBO is active prior to fetching a compressed texture
+ * image.
+ * If so, do bounds checking and map the buffer into main memory.
+ * Any errors detected will be recorded.
+ * The caller _must_ call _mesa_unmap_teximage_pbo() too!
+ */
+const GLvoid *
+_mesa_validate_pbo_compressed_teximage(struct gl_context *ctx,
+                                 GLsizei imageSize, const GLvoid *pixels,
+                                 const struct gl_pixelstore_attrib *packing,
+                                 const char *funcName)
+{
+   GLubyte *buf;
+
+   if (!_mesa_is_bufferobj(packing->BufferObj)) {
+      /* not using a PBO - return pointer unchanged */
+      return pixels;
+   }
+   if ((const GLubyte *) pixels + imageSize >
+       ((const GLubyte *) 0) + packing->BufferObj->Size) {
+      /* out of bounds read! */
+      _mesa_error(ctx, GL_INVALID_OPERATION, funcName, "(invalid PBO access)");
+      return NULL;
+   }
+
+   buf = (GLubyte*) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
+                                         GL_READ_ONLY_ARB, packing->BufferObj);
+   if (!buf) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, funcName, "(PBO is mapped");
+      return NULL;
+   }
+
+   return ADD_POINTERS(buf, pixels);
+}
+
+
+/**
+ * This function must be called after either of the validate_pbo_*_teximage()
+ * functions.  It unmaps the PBO buffer if it was mapped earlier.
+ */
+void
+_mesa_unmap_teximage_pbo(struct gl_context *ctx,
+                         const struct gl_pixelstore_attrib *unpack)
+{
+   if (_mesa_is_bufferobj(unpack->BufferObj)) {
+      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
+                              unpack->BufferObj);
+   }
+}
+
+
diff --git a/src/mesa/main/pbo.h b/src/mesa/main/pbo.h
new file mode 100644
index 00000000000..0cddd72ba7f
--- /dev/null
+++ b/src/mesa/main/pbo.h
@@ -0,0 +1,92 @@
+/*
+ * Mesa 3-D graphics library
+ *
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009-2011  VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+#ifndef PBO_H
+#define PBO_H
+
+
+#include "mtypes.h"
+
+
+extern GLboolean
+_mesa_validate_pbo_access(GLuint dimensions,
+                          const struct gl_pixelstore_attrib *pack,
+                          GLsizei width, GLsizei height, GLsizei depth,
+                          GLenum format, GLenum type, const GLvoid *ptr);
+
+extern const GLvoid *
+_mesa_map_pbo_source(struct gl_context *ctx,
+                     const struct gl_pixelstore_attrib *unpack,
+                     const GLvoid *src);
+
+extern const GLvoid *
+_mesa_map_validate_pbo_source(struct gl_context *ctx,
+                              GLuint dimensions,
+                              const struct gl_pixelstore_attrib *unpack,
+                              GLsizei width, GLsizei height, GLsizei depth,
+                              GLenum format, GLenum type, const GLvoid *ptr,
+                              const char *where);
+
+extern void
+_mesa_unmap_pbo_source(struct gl_context *ctx,
+                       const struct gl_pixelstore_attrib *unpack);
+
+extern void *
+_mesa_map_pbo_dest(struct gl_context *ctx,
+                   const struct gl_pixelstore_attrib *pack,
+                   GLvoid *dest);
+
+extern GLvoid *
+_mesa_map_validate_pbo_dest(struct gl_context *ctx,
+                            GLuint dimensions,
+                            const struct gl_pixelstore_attrib *unpack,
+                            GLsizei width, GLsizei height, GLsizei depth,
+                            GLenum format, GLenum type, GLvoid *ptr,
+                            const char *where);
+
+extern void
+_mesa_unmap_pbo_dest(struct gl_context *ctx,
+                     const struct gl_pixelstore_attrib *pack);
+
+
+extern const GLvoid *
+_mesa_validate_pbo_teximage(struct gl_context *ctx, GLuint dimensions,
+			    GLsizei width, GLsizei height, GLsizei depth,
+			    GLenum format, GLenum type, const GLvoid *pixels,
+			    const struct gl_pixelstore_attrib *unpack,
+			    const char *funcName);
+
+extern const GLvoid *
+_mesa_validate_pbo_compressed_teximage(struct gl_context *ctx,
+                                    GLsizei imageSize, const GLvoid *pixels,
+                                    const struct gl_pixelstore_attrib *packing,
+                                    const char *funcName);
+
+extern void
+_mesa_unmap_teximage_pbo(struct gl_context *ctx,
+                         const struct gl_pixelstore_attrib *unpack);
+
+
+#endif
diff --git a/src/mesa/main/pixel.c b/src/mesa/main/pixel.c
index 0254980b89d..195fa234be2 100644
--- a/src/mesa/main/pixel.c
+++ b/src/mesa/main/pixel.c
@@ -35,6 +35,7 @@
 #include "macros.h"
 #include "mfeatures.h"
 #include "pixel.h"
+#include "pbo.h"
 #include "mtypes.h"
 #include "main/dispatch.h"
 
@@ -685,9 +686,6 @@ _mesa_init_pixel( struct gl_context *ctx )
    init_pixelmap(&ctx->PixelMaps.GtoG);
    init_pixelmap(&ctx->PixelMaps.BtoB);
    init_pixelmap(&ctx->PixelMaps.AtoA);
-   /* GL_SGI_texture_color_table */
-   ASSIGN_4V(ctx->Pixel.TextureColorTableScale, 1.0, 1.0, 1.0, 1.0);
-   ASSIGN_4V(ctx->Pixel.TextureColorTableBias, 0.0, 0.0, 0.0, 0.0);
 
    if (ctx->Visual.doubleBufferMode) {
       ctx->Pixel.ReadBuffer = GL_BACK;
diff --git a/src/mesa/main/polygon.c b/src/mesa/main/polygon.c
index 9c213b9b4c0..ff4232ecc39 100644
--- a/src/mesa/main/polygon.c
+++ b/src/mesa/main/polygon.c
@@ -30,11 +30,11 @@
 
 #include "glheader.h"
 #include "imports.h"
-#include "bufferobj.h"
 #include "context.h"
 #include "image.h"
 #include "enums.h"
 #include "pack.h"
+#include "pbo.h"
 #include "polygon.h"
 #include "mtypes.h"
 
diff --git a/src/mesa/main/readpix.c b/src/mesa/main/readpix.c
index 9a4f15f7279..6e09a52c88a 100644
--- a/src/mesa/main/readpix.c
+++ b/src/mesa/main/readpix.c
@@ -32,6 +32,7 @@
 #include "formats.h"
 #include "image.h"
 #include "mtypes.h"
+#include "pbo.h"
 #include "state.h"
 
 
diff --git a/src/mesa/main/texcompress_rgtc.c b/src/mesa/main/texcompress_rgtc.c
index 1a01755f14d..26dca2d760b 100644
--- a/src/mesa/main/texcompress_rgtc.c
+++ b/src/mesa/main/texcompress_rgtc.c
@@ -46,11 +46,17 @@
 
 #define RGTC_DEBUG 0
 
-static void encode_rgtc_chan_u(GLubyte *blkaddr, GLubyte srccolors[4][4],
-			     GLint numxpixels, GLint numypixels);
-static void encode_rgtc_chan_s(GLbyte *blkaddr, GLbyte srccolors[4][4],
+static void unsigned_encode_rgtc_chan(GLubyte *blkaddr, GLubyte srccolors[4][4],
+					GLint numxpixels, GLint numypixels);
+static void signed_encode_rgtc_chan(GLbyte *blkaddr, GLbyte srccolors[4][4],
 			     GLint numxpixels, GLint numypixels);
 
+static void unsigned_fetch_texel_rgtc(unsigned srcRowStride, const GLubyte *pixdata,
+				      unsigned i, unsigned j, GLubyte *value, unsigned comps);
+
+static void signed_fetch_texel_rgtc(unsigned srcRowStride, const GLbyte *pixdata,
+				      unsigned i, unsigned j, GLbyte *value, unsigned comps);
+
 static void extractsrc_u( GLubyte srcpixels[4][4], const GLchan *srcaddr,
 			  GLint srcRowStride, GLint numxpixels, GLint numypixels, GLint comps)
 {
@@ -123,7 +129,7 @@ _mesa_texstore_red_rgtc1(TEXSTORE_PARAMS)
 	 if (srcWidth > i + 3) numxpixels = 4;
 	 else numxpixels = srcWidth - i;
 	 extractsrc_u(srcpixels, srcaddr, srcWidth, numxpixels, numypixels, 1);
-	 encode_rgtc_chan_u(blkaddr, srcpixels, numxpixels, numypixels);
+	 unsigned_encode_rgtc_chan(blkaddr, srcpixels, numxpixels, numypixels);
 	 srcaddr += numxpixels;
 	 blkaddr += 8;
       }
@@ -177,7 +183,7 @@ _mesa_texstore_signed_red_rgtc1(TEXSTORE_PARAMS)
 	 if (srcWidth > i + 3) numxpixels = 4;
 	 else numxpixels = srcWidth - i;
 	 extractsrc_s(srcpixels, srcaddr, srcWidth, numxpixels, numypixels, 1);
-	 encode_rgtc_chan_s(blkaddr, srcpixels, numxpixels, numypixels);
+	 signed_encode_rgtc_chan(blkaddr, srcpixels, numxpixels, numypixels);
 	 srcaddr += numxpixels;
 	 blkaddr += 8;
       }
@@ -232,11 +238,11 @@ _mesa_texstore_rg_rgtc2(TEXSTORE_PARAMS)
 	 if (srcWidth > i + 3) numxpixels = 4;
 	 else numxpixels = srcWidth - i;
 	 extractsrc_u(srcpixels, srcaddr, srcWidth, numxpixels, numypixels, 2);
-	 encode_rgtc_chan_u(blkaddr, srcpixels, numxpixels, numypixels);
+	 unsigned_encode_rgtc_chan(blkaddr, srcpixels, numxpixels, numypixels);
 
 	 blkaddr += 8;
 	 extractsrc_u(srcpixels, (GLchan *)srcaddr + 1, srcWidth, numxpixels, numypixels, 2);
-	 encode_rgtc_chan_u(blkaddr, srcpixels, numxpixels, numypixels);
+	 unsigned_encode_rgtc_chan(blkaddr, srcpixels, numxpixels, numypixels);
 
 	 blkaddr += 8;
 
@@ -294,11 +300,11 @@ _mesa_texstore_signed_rg_rgtc2(TEXSTORE_PARAMS)
 	 else numxpixels = srcWidth - i;
 
 	 extractsrc_s(srcpixels, srcaddr, srcWidth, numxpixels, numypixels, 2);
-	 encode_rgtc_chan_s(blkaddr, srcpixels, numxpixels, numypixels);
+	 signed_encode_rgtc_chan(blkaddr, srcpixels, numxpixels, numypixels);
 	 blkaddr += 8;
 
 	 extractsrc_s(srcpixels, srcaddr + 1, srcWidth, numxpixels, numypixels, 2);
-	 encode_rgtc_chan_s(blkaddr, srcpixels, numxpixels, numypixels);
+	 signed_encode_rgtc_chan(blkaddr, srcpixels, numxpixels, numypixels);
 	 blkaddr += 8;
 
 	 srcaddr += numxpixels * 2;
@@ -312,73 +318,14 @@ _mesa_texstore_signed_rg_rgtc2(TEXSTORE_PARAMS)
    return GL_TRUE;
 }
 
-static void _fetch_texel_rgtc_u(GLint srcRowStride, const GLubyte *pixdata,
-				GLint i, GLint j, GLchan *value, int comps)
-{
-   GLchan decode;
-   const GLubyte *blksrc = (pixdata + ((srcRowStride + 3) / 4 * (j / 4) + (i / 4)) * 8 * comps);
-   const GLubyte alpha0 = blksrc[0];
-   const GLubyte alpha1 = blksrc[1];
-   const GLubyte bit_pos = ((j&3) * 4 + (i&3)) * 3;
-   const GLubyte acodelow = blksrc[2 + bit_pos / 8];
-   const GLubyte acodehigh = blksrc[3 + bit_pos / 8];
-   const GLubyte code = (acodelow >> (bit_pos & 0x7) |
-      (acodehigh  << (8 - (bit_pos & 0x7)))) & 0x7;
-
-   if (code == 0)
-      decode = UBYTE_TO_CHAN( alpha0 );
-   else if (code == 1)
-      decode = UBYTE_TO_CHAN( alpha1 );
-   else if (alpha0 > alpha1)
-      decode = UBYTE_TO_CHAN( ((alpha0 * (8 - code) + (alpha1 * (code - 1))) / 7) );
-   else if (code < 6)
-      decode = UBYTE_TO_CHAN( ((alpha0 * (6 - code) + (alpha1 * (code - 1))) / 5) );
-   else if (code == 6)
-      decode = 0;
-   else
-      decode = CHAN_MAX;
-
-   *value = decode;
-}
-
-
-static void _fetch_texel_rgtc_s(GLint srcRowStride, const GLbyte *pixdata,
-				GLint i, GLint j, GLbyte *value, int comps)
-{
-   GLbyte decode;
-   const GLbyte *blksrc = (pixdata + ((srcRowStride + 3) / 4 * (j / 4) + (i / 4)) * 8 * comps);
-   const GLbyte alpha0 = blksrc[0];
-   const GLbyte alpha1 = blksrc[1];
-   const GLbyte bit_pos = ((j&3) * 4 + (i&3)) * 3;
-   const GLbyte acodelow = blksrc[2 + bit_pos / 8];
-   const GLbyte acodehigh = blksrc[3 + bit_pos / 8];
-   const GLbyte code = (acodelow >> (bit_pos & 0x7) |
-      (acodehigh  << (8 - (bit_pos & 0x7)))) & 0x7;
-
-   if (code == 0)
-      decode = alpha0;
-   else if (code == 1)
-      decode = alpha1;
-   else if (alpha0 > alpha1)
-      decode = ((alpha0 * (8 - code) + (alpha1 * (code - 1))) / 7);
-   else if (code < 6)
-      decode = ((alpha0 * (6 - code) + (alpha1 * (code - 1))) / 5);
-   else if (code == 6)
-      decode = -128;
-   else
-      decode = 127;
-
-   *value = decode;
-}
-
 void
 _mesa_fetch_texel_2d_f_red_rgtc1(const struct gl_texture_image *texImage,
 				 GLint i, GLint j, GLint k, GLfloat *texel)
 {
-   GLchan red;
-   _fetch_texel_rgtc_u(texImage->RowStride, (GLubyte *)(texImage->Data),
+   GLubyte red;
+   unsigned_fetch_texel_rgtc(texImage->RowStride, (GLubyte *)(texImage->Data),
 		       i, j, &red, 1);
-   texel[RCOMP] = CHAN_TO_FLOAT(red);
+   texel[RCOMP] = UBYTE_TO_FLOAT(red);
    texel[GCOMP] = 0.0;
    texel[BCOMP] = 0.0;
    texel[ACOMP] = 1.0;
@@ -389,7 +336,7 @@ _mesa_fetch_texel_2d_f_signed_red_rgtc1(const struct gl_texture_image *texImage,
 					GLint i, GLint j, GLint k, GLfloat *texel)
 {
    GLbyte red;
-   _fetch_texel_rgtc_s(texImage->RowStride, (GLbyte *)(texImage->Data),
+   signed_fetch_texel_rgtc(texImage->RowStride, (GLbyte *)(texImage->Data),
 		       i, j, &red, 1);
    texel[RCOMP] = BYTE_TO_FLOAT_TEX(red);
    texel[GCOMP] = 0.0;
@@ -401,13 +348,13 @@ void
 _mesa_fetch_texel_2d_f_rg_rgtc2(const struct gl_texture_image *texImage,
 				 GLint i, GLint j, GLint k, GLfloat *texel)
 {
-   GLchan red, green;
-   _fetch_texel_rgtc_u(texImage->RowStride, (GLubyte *)(texImage->Data),
+   GLubyte red, green;
+   unsigned_fetch_texel_rgtc(texImage->RowStride, (GLubyte *)(texImage->Data),
 		     i, j, &red, 2);
-   _fetch_texel_rgtc_u(texImage->RowStride, (GLubyte *)(texImage->Data) + 8,
+   unsigned_fetch_texel_rgtc(texImage->RowStride, (GLubyte *)(texImage->Data) + 8,
 		     i, j, &green, 2);
-   texel[RCOMP] = CHAN_TO_FLOAT(red);
-   texel[GCOMP] = CHAN_TO_FLOAT(green);
+   texel[RCOMP] = UBYTE_TO_FLOAT(red);
+   texel[GCOMP] = UBYTE_TO_FLOAT(green);
    texel[BCOMP] = 0.0;
    texel[ACOMP] = 1.0;
 }
@@ -417,9 +364,9 @@ _mesa_fetch_texel_2d_f_signed_rg_rgtc2(const struct gl_texture_image *texImage,
 				       GLint i, GLint j, GLint k, GLfloat *texel)
 {
    GLbyte red, green;
-   _fetch_texel_rgtc_s(texImage->RowStride, (GLbyte *)(texImage->Data),
+   signed_fetch_texel_rgtc(texImage->RowStride, (GLbyte *)(texImage->Data),
 		     i, j, &red, 2);
-   _fetch_texel_rgtc_s(texImage->RowStride, (GLbyte *)(texImage->Data) + 8,
+   signed_fetch_texel_rgtc(texImage->RowStride, (GLbyte *)(texImage->Data) + 8,
 		     i, j, &green, 2);
    texel[RCOMP] = BYTE_TO_FLOAT_TEX(red);
    texel[GCOMP] = BYTE_TO_FLOAT_TEX(green);
@@ -427,696 +374,27 @@ _mesa_fetch_texel_2d_f_signed_rg_rgtc2(const struct gl_texture_image *texImage,
    texel[ACOMP] = 1.0;
 }
 
-static void write_rgtc_encoded_channel(GLubyte *blkaddr,
-				       GLubyte alphabase1,
-				       GLubyte alphabase2,
-				       GLubyte alphaenc[16])
-{
-   *blkaddr++ = alphabase1;
-   *blkaddr++ = alphabase2;
-   *blkaddr++ = alphaenc[0] | (alphaenc[1] << 3) | ((alphaenc[2] & 3) << 6);
-   *blkaddr++ = (alphaenc[2] >> 2) | (alphaenc[3] << 1) | (alphaenc[4] << 4) | ((alphaenc[5] & 1) << 7);
-   *blkaddr++ = (alphaenc[5] >> 1) | (alphaenc[6] << 2) | (alphaenc[7] << 5);
-   *blkaddr++ = alphaenc[8] | (alphaenc[9] << 3) | ((alphaenc[10] & 3) << 6);
-   *blkaddr++ = (alphaenc[10] >> 2) | (alphaenc[11] << 1) | (alphaenc[12] << 4) | ((alphaenc[13] & 1) << 7);
-   *blkaddr++ = (alphaenc[13] >> 1) | (alphaenc[14] << 2) | (alphaenc[15] << 5);
-}
+#define TAG(x) unsigned_##x
 
-static void encode_rgtc_chan_u(GLubyte *blkaddr, GLubyte srccolors[4][4],
-			     GLint numxpixels, GLint numypixels)
-{
-   GLubyte alphabase[2], alphause[2];
-   GLshort alphatest[2] = { 0 };
-   GLuint alphablockerror1, alphablockerror2, alphablockerror3;
-   GLubyte i, j, aindex, acutValues[7];
-   GLubyte alphaenc1[16], alphaenc2[16], alphaenc3[16];
-   GLboolean alphaabsmin = GL_FALSE;
-   GLboolean alphaabsmax = GL_FALSE;
-   GLshort alphadist;
+#define TYPE GLubyte
+#define T_MIN 0
+#define T_MAX 0xff
 
-   /* find lowest and highest alpha value in block, alphabase[0] lowest, alphabase[1] highest */
-   alphabase[0] = 0xff; alphabase[1] = 0x0;
-   for (j = 0; j < numypixels; j++) {
-      for (i = 0; i < numxpixels; i++) {
-         if (srccolors[j][i] == 0)
-            alphaabsmin = GL_TRUE;
-         else if (srccolors[j][i] == 255)
-            alphaabsmax = GL_TRUE;
-         else {
-            if (srccolors[j][i] > alphabase[1])
-               alphabase[1] = srccolors[j][i];
-            if (srccolors[j][i] < alphabase[0])
-               alphabase[0] = srccolors[j][i];
-         }
-      }
-   }
+#include "texcompress_rgtc_tmp.h"
 
+#undef TAG
+#undef TYPE
+#undef T_MIN
+#undef T_MAX
 
-   if ((alphabase[0] > alphabase[1]) && !(alphaabsmin && alphaabsmax)) { /* one color, either max or min */
-      /* shortcut here since it is a very common case (and also avoids later problems) */
-      /* || (alphabase[0] == alphabase[1] && !alphaabsmin && !alphaabsmax) */
-      /* could also thest for alpha0 == alpha1 (and not min/max), but probably not common, so don't bother */
+#define TAG(x) signed_##x
+#define TYPE GLbyte
+#define T_MIN (GLbyte)-128
+#define T_MAX (GLbyte)127
 
-      *blkaddr++ = srccolors[0][0];
-      blkaddr++;
-      *blkaddr++ = 0;
-      *blkaddr++ = 0;
-      *blkaddr++ = 0;
-      *blkaddr++ = 0;
-      *blkaddr++ = 0;
-      *blkaddr++ = 0;
-#if RGTC_DEBUG
-      fprintf(stderr, "enc0 used\n");
-#endif
-      return;
-   }
+#include "texcompress_rgtc_tmp.h"
 
-   /* find best encoding for alpha0 > alpha1 */
-   /* it's possible this encoding is better even if both alphaabsmin and alphaabsmax are true */
-   alphablockerror1 = 0x0;
-   alphablockerror2 = 0xffffffff;
-   alphablockerror3 = 0xffffffff;
-   if (alphaabsmin) alphause[0] = 0;
-   else alphause[0] = alphabase[0];
-   if (alphaabsmax) alphause[1] = 255;
-   else alphause[1] = alphabase[1];
-   /* calculate the 7 cut values, just the middle between 2 of the computed alpha values */
-   for (aindex = 0; aindex < 7; aindex++) {
-      /* don't forget here is always rounded down */
-      acutValues[aindex] = (alphause[0] * (2*aindex + 1) + alphause[1] * (14 - (2*aindex + 1))) / 14;
-   }
-
-   for (j = 0; j < numypixels; j++) {
-      for (i = 0; i < numxpixels; i++) {
-         /* maybe it's overkill to have the most complicated calculation just for the error
-            calculation which we only need to figure out if encoding1 or encoding2 is better... */
-         if (srccolors[j][i] > acutValues[0]) {
-            alphaenc1[4*j + i] = 0;
-            alphadist = srccolors[j][i] - alphause[1];
-         }
-         else if (srccolors[j][i] > acutValues[1]) {
-            alphaenc1[4*j + i] = 2;
-            alphadist = srccolors[j][i] - (alphause[1] * 6 + alphause[0] * 1) / 7;
-         }
-         else if (srccolors[j][i] > acutValues[2]) {
-            alphaenc1[4*j + i] = 3;
-            alphadist = srccolors[j][i] - (alphause[1] * 5 + alphause[0] * 2) / 7;
-         }
-         else if (srccolors[j][i] > acutValues[3]) {
-            alphaenc1[4*j + i] = 4;
-            alphadist = srccolors[j][i] - (alphause[1] * 4 + alphause[0] * 3) / 7;
-         }
-         else if (srccolors[j][i] > acutValues[4]) {
-            alphaenc1[4*j + i] = 5;
-            alphadist = srccolors[j][i] - (alphause[1] * 3 + alphause[0] * 4) / 7;
-         }
-         else if (srccolors[j][i] > acutValues[5]) {
-            alphaenc1[4*j + i] = 6;
-            alphadist = srccolors[j][i] - (alphause[1] * 2 + alphause[0] * 5) / 7;
-         }
-         else if (srccolors[j][i] > acutValues[6]) {
-            alphaenc1[4*j + i] = 7;
-            alphadist = srccolors[j][i] - (alphause[1] * 1 + alphause[0] * 6) / 7;
-         }
-         else {
-            alphaenc1[4*j + i] = 1;
-            alphadist = srccolors[j][i] - alphause[0];
-         }
-         alphablockerror1 += alphadist * alphadist;
-      }
-   }
-
-#if RGTC_DEBUG
-   for (i = 0; i < 16; i++) {
-      fprintf(stderr, "%d ", alphaenc1[i]);
-   }
-   fprintf(stderr, "cutVals ");
-   for (i = 0; i < 8; i++) {
-      fprintf(stderr, "%d ", acutValues[i]);
-   }
-   fprintf(stderr, "srcVals ");
-   for (j = 0; j < numypixels; j++) {
-      for (i = 0; i < numxpixels; i++) {
-	 fprintf(stderr, "%d ", srccolors[j][i]);
-      }
-   }
-   fprintf(stderr, "\n");
-#endif
-
-   /* it's not very likely this encoding is better if both alphaabsmin and alphaabsmax
-      are false but try it anyway */
-   if (alphablockerror1 >= 32) {
-
-      /* don't bother if encoding is already very good, this condition should also imply
-      we have valid alphabase colors which we absolutely need (alphabase[0] <= alphabase[1]) */
-      alphablockerror2 = 0;
-      for (aindex = 0; aindex < 5; aindex++) {
-         /* don't forget here is always rounded down */
-         acutValues[aindex] = (alphabase[0] * (10 - (2*aindex + 1)) + alphabase[1] * (2*aindex + 1)) / 10;
-      }
-      for (j = 0; j < numypixels; j++) {
-         for (i = 0; i < numxpixels; i++) {
-             /* maybe it's overkill to have the most complicated calculation just for the error
-               calculation which we only need to figure out if encoding1 or encoding2 is better... */
-            if (srccolors[j][i] == 0) {
-               alphaenc2[4*j + i] = 6;
-               alphadist = 0;
-            }
-            else if (srccolors[j][i] == 255) {
-               alphaenc2[4*j + i] = 7;
-               alphadist = 0;
-            }
-            else if (srccolors[j][i] <= acutValues[0]) {
-               alphaenc2[4*j + i] = 0;
-               alphadist = srccolors[j][i] - alphabase[0];
-            }
-            else if (srccolors[j][i] <= acutValues[1]) {
-               alphaenc2[4*j + i] = 2;
-               alphadist = srccolors[j][i] - (alphabase[0] * 4 + alphabase[1] * 1) / 5;
-            }
-            else if (srccolors[j][i] <= acutValues[2]) {
-               alphaenc2[4*j + i] = 3;
-               alphadist = srccolors[j][i] - (alphabase[0] * 3 + alphabase[1] * 2) / 5;
-            }
-            else if (srccolors[j][i] <= acutValues[3]) {
-               alphaenc2[4*j + i] = 4;
-               alphadist = srccolors[j][i] - (alphabase[0] * 2 + alphabase[1] * 3) / 5;
-            }
-            else if (srccolors[j][i] <= acutValues[4]) {
-               alphaenc2[4*j + i] = 5;
-               alphadist = srccolors[j][i] - (alphabase[0] * 1 + alphabase[1] * 4) / 5;
-            }
-            else {
-               alphaenc2[4*j + i] = 1;
-               alphadist = srccolors[j][i] - alphabase[1];
-            }
-            alphablockerror2 += alphadist * alphadist;
-         }
-      }
-
-
-      /* skip this if the error is already very small
-         this encoding is MUCH better on average than #2 though, but expensive! */
-      if ((alphablockerror2 > 96) && (alphablockerror1 > 96)) {
-         GLshort blockerrlin1 = 0;
-         GLshort blockerrlin2 = 0;
-         GLubyte nralphainrangelow = 0;
-         GLubyte nralphainrangehigh = 0;
-         alphatest[0] = 0xff;
-         alphatest[1] = 0x0;
-         /* if we have large range it's likely there are values close to 0/255, try to map them to 0/255 */
-         for (j = 0; j < numypixels; j++) {
-            for (i = 0; i < numxpixels; i++) {
-               if ((srccolors[j][i] > alphatest[1]) && (srccolors[j][i] < (255 -(alphabase[1] - alphabase[0]) / 28)))
-                  alphatest[1] = srccolors[j][i];
-               if ((srccolors[j][i] < alphatest[0]) && (srccolors[j][i] > (alphabase[1] - alphabase[0]) / 28))
-                  alphatest[0] = srccolors[j][i];
-            }
-         }
-          /* shouldn't happen too often, don't really care about those degenerated cases */
-          if (alphatest[1] <= alphatest[0]) {
-             alphatest[0] = 1;
-             alphatest[1] = 254;
-         }
-         for (aindex = 0; aindex < 5; aindex++) {
-         /* don't forget here is always rounded down */
-            acutValues[aindex] = (alphatest[0] * (10 - (2*aindex + 1)) + alphatest[1] * (2*aindex + 1)) / 10;
-         }
-
-         /* find the "average" difference between the alpha values and the next encoded value.
-            This is then used to calculate new base values.
-            Should there be some weighting, i.e. those values closer to alphatest[x] have more weight,
-            since they will see more improvement, and also because the values in the middle are somewhat
-            likely to get no improvement at all (because the base values might move in different directions)?
-            OTOH it would mean the values in the middle are even less likely to get an improvement
-         */
-         for (j = 0; j < numypixels; j++) {
-            for (i = 0; i < numxpixels; i++) {
-               if (srccolors[j][i] <= alphatest[0] / 2) {
-               }
-               else if (srccolors[j][i] > ((255 + alphatest[1]) / 2)) {
-               }
-               else if (srccolors[j][i] <= acutValues[0]) {
-                  blockerrlin1 += (srccolors[j][i] - alphatest[0]);
-                  nralphainrangelow += 1;
-               }
-               else if (srccolors[j][i] <= acutValues[1]) {
-                  blockerrlin1 += (srccolors[j][i] - (alphatest[0] * 4 + alphatest[1] * 1) / 5);
-                  blockerrlin2 += (srccolors[j][i] - (alphatest[0] * 4 + alphatest[1] * 1) / 5);
-                  nralphainrangelow += 1;
-                  nralphainrangehigh += 1;
-               }
-               else if (srccolors[j][i] <= acutValues[2]) {
-                  blockerrlin1 += (srccolors[j][i] - (alphatest[0] * 3 + alphatest[1] * 2) / 5);
-                  blockerrlin2 += (srccolors[j][i] - (alphatest[0] * 3 + alphatest[1] * 2) / 5);
-                  nralphainrangelow += 1;
-                  nralphainrangehigh += 1;
-               }
-               else if (srccolors[j][i] <= acutValues[3]) {
-                  blockerrlin1 += (srccolors[j][i] - (alphatest[0] * 2 + alphatest[1] * 3) / 5);
-                  blockerrlin2 += (srccolors[j][i] - (alphatest[0] * 2 + alphatest[1] * 3) / 5);
-                  nralphainrangelow += 1;
-                  nralphainrangehigh += 1;
-               }
-               else if (srccolors[j][i] <= acutValues[4]) {
-                  blockerrlin1 += (srccolors[j][i] - (alphatest[0] * 1 + alphatest[1] * 4) / 5);
-                  blockerrlin2 += (srccolors[j][i] - (alphatest[0] * 1 + alphatest[1] * 4) / 5);
-                  nralphainrangelow += 1;
-                  nralphainrangehigh += 1;
-                  }
-               else {
-                  blockerrlin2 += (srccolors[j][i] - alphatest[1]);
-                  nralphainrangehigh += 1;
-               }
-            }
-         }
-         /* shouldn't happen often, needed to avoid div by zero */
-         if (nralphainrangelow == 0) nralphainrangelow = 1;
-         if (nralphainrangehigh == 0) nralphainrangehigh = 1;
-         alphatest[0] = alphatest[0] + (blockerrlin1 / nralphainrangelow);
-#if RGTC_DEBUG
-         fprintf(stderr, "block err lin low %d, nr %d\n", blockerrlin1, nralphainrangelow);
-         fprintf(stderr, "block err lin high %d, nr %d\n", blockerrlin2, nralphainrangehigh);
-#endif
-         /* again shouldn't really happen often... */
-         if (alphatest[0] < 0) {
-            alphatest[0] = 0;
-         }
-         alphatest[1] = alphatest[1] + (blockerrlin2 / nralphainrangehigh);
-         if (alphatest[1] > 255) {
-            alphatest[1] = 255;
-         }
-
-         alphablockerror3 = 0;
-         for (aindex = 0; aindex < 5; aindex++) {
-         /* don't forget here is always rounded down */
-            acutValues[aindex] = (alphatest[0] * (10 - (2*aindex + 1)) + alphatest[1] * (2*aindex + 1)) / 10;
-         }
-         for (j = 0; j < numypixels; j++) {
-            for (i = 0; i < numxpixels; i++) {
-                /* maybe it's overkill to have the most complicated calculation just for the error
-                  calculation which we only need to figure out if encoding1 or encoding2 is better... */
-               if (srccolors[j][i] <= alphatest[0] / 2) {
-                  alphaenc3[4*j + i] = 6;
-                  alphadist = srccolors[j][i];
-               }
-               else if (srccolors[j][i] > ((255 + alphatest[1]) / 2)) {
-                  alphaenc3[4*j + i] = 7;
-                  alphadist = 255 - srccolors[j][i];
-               }
-               else if (srccolors[j][i] <= acutValues[0]) {
-                  alphaenc3[4*j + i] = 0;
-                  alphadist = srccolors[j][i] - alphatest[0];
-               }
-               else if (srccolors[j][i] <= acutValues[1]) {
-                 alphaenc3[4*j + i] = 2;
-                 alphadist = srccolors[j][i] - (alphatest[0] * 4 + alphatest[1] * 1) / 5;
-               }
-               else if (srccolors[j][i] <= acutValues[2]) {
-                  alphaenc3[4*j + i] = 3;
-                  alphadist = srccolors[j][i] - (alphatest[0] * 3 + alphatest[1] * 2) / 5;
-               }
-               else if (srccolors[j][i] <= acutValues[3]) {
-                  alphaenc3[4*j + i] = 4;
-                  alphadist = srccolors[j][i] - (alphatest[0] * 2 + alphatest[1] * 3) / 5;
-               }
-               else if (srccolors[j][i] <= acutValues[4]) {
-                  alphaenc3[4*j + i] = 5;
-                  alphadist = srccolors[j][i] - (alphatest[0] * 1 + alphatest[1] * 4) / 5;
-               }
-               else {
-                  alphaenc3[4*j + i] = 1;
-                  alphadist = srccolors[j][i] - alphatest[1];
-               }
-               alphablockerror3 += alphadist * alphadist;
-            }
-         }
-      }
-   }
-  /* write the alpha values and encoding back. */
-   if ((alphablockerror1 <= alphablockerror2) && (alphablockerror1 <= alphablockerror3)) {
-#if RGTC_DEBUG
-      if (alphablockerror1 > 96) fprintf(stderr, "enc1 used, error %d\n", alphablockerror1);
-#endif
-      write_rgtc_encoded_channel( blkaddr, alphause[1], alphause[0], alphaenc1 );
-   }
-   else if (alphablockerror2 <= alphablockerror3) {
-#if RGTC_DEBUG
-      if (alphablockerror2 > 96) fprintf(stderr, "enc2 used, error %d\n", alphablockerror2);
-#endif
-      write_rgtc_encoded_channel( blkaddr, alphabase[0], alphabase[1], alphaenc2 );
-   }
-   else {
-#if RGTC_DEBUG
-      fprintf(stderr, "enc3 used, error %d\n", alphablockerror3);
-#endif
-      write_rgtc_encoded_channel( blkaddr, (GLubyte)alphatest[0], (GLubyte)alphatest[1], alphaenc3 );
-   }
-}
-
-
-static void write_rgtc_encoded_channel_s(GLbyte *blkaddr,
-					 GLbyte alphabase1,
-					 GLbyte alphabase2,
-					 GLbyte alphaenc[16])
-{
-   *blkaddr++ = alphabase1;
-   *blkaddr++ = alphabase2;
-   *blkaddr++ = alphaenc[0] | (alphaenc[1] << 3) | ((alphaenc[2] & 3) << 6);
-   *blkaddr++ = (alphaenc[2] >> 2) | (alphaenc[3] << 1) | (alphaenc[4] << 4) | ((alphaenc[5] & 1) << 7);
-   *blkaddr++ = (alphaenc[5] >> 1) | (alphaenc[6] << 2) | (alphaenc[7] << 5);
-   *blkaddr++ = alphaenc[8] | (alphaenc[9] << 3) | ((alphaenc[10] & 3) << 6);
-   *blkaddr++ = (alphaenc[10] >> 2) | (alphaenc[11] << 1) | (alphaenc[12] << 4) | ((alphaenc[13] & 1) << 7);
-   *blkaddr++ = (alphaenc[13] >> 1) | (alphaenc[14] << 2) | (alphaenc[15] << 5);
-}
-
-static void encode_rgtc_chan_s(GLbyte *blkaddr, GLbyte srccolors[4][4],
-			       GLint numxpixels, GLint numypixels)
-{
-   GLbyte alphabase[2], alphause[2];
-   GLshort alphatest[2] = { 0 };
-   GLuint alphablockerror1, alphablockerror2, alphablockerror3;
-   GLbyte i, j, aindex, acutValues[7];
-   GLbyte alphaenc1[16], alphaenc2[16], alphaenc3[16];
-   GLboolean alphaabsmin = GL_FALSE;
-   GLboolean alphaabsmax = GL_FALSE;
-   GLshort alphadist;
-
-   /* find lowest and highest alpha value in block, alphabase[0] lowest, alphabase[1] highest */
-   alphabase[0] = 0xff; alphabase[1] = 0x0;
-   for (j = 0; j < numypixels; j++) {
-      for (i = 0; i < numxpixels; i++) {
-         if (srccolors[j][i] == 0)
-            alphaabsmin = GL_TRUE;
-         else if (srccolors[j][i] == 255)
-            alphaabsmax = GL_TRUE;
-         else {
-            if (srccolors[j][i] > alphabase[1])
-               alphabase[1] = srccolors[j][i];
-            if (srccolors[j][i] < alphabase[0])
-               alphabase[0] = srccolors[j][i];
-         }
-      }
-   }
-
-
-   if ((alphabase[0] > alphabase[1]) && !(alphaabsmin && alphaabsmax)) { /* one color, either max or min */
-      /* shortcut here since it is a very common case (and also avoids later problems) */
-      /* || (alphabase[0] == alphabase[1] && !alphaabsmin && !alphaabsmax) */
-      /* could also thest for alpha0 == alpha1 (and not min/max), but probably not common, so don't bother */
-
-      *blkaddr++ = srccolors[0][0];
-      blkaddr++;
-      *blkaddr++ = 0;
-      *blkaddr++ = 0;
-      *blkaddr++ = 0;
-      *blkaddr++ = 0;
-      *blkaddr++ = 0;
-      *blkaddr++ = 0;
-#if RGTC_DEBUG
-      fprintf(stderr, "enc0 used\n");
-#endif
-      return;
-   }
-
-   /* find best encoding for alpha0 > alpha1 */
-   /* it's possible this encoding is better even if both alphaabsmin and alphaabsmax are true */
-   alphablockerror1 = 0x0;
-   alphablockerror2 = 0xffffffff;
-   alphablockerror3 = 0xffffffff;
-   if (alphaabsmin) alphause[0] = 0;
-   else alphause[0] = alphabase[0];
-   if (alphaabsmax) alphause[1] = 255;
-   else alphause[1] = alphabase[1];
-   /* calculate the 7 cut values, just the middle between 2 of the computed alpha values */
-   for (aindex = 0; aindex < 7; aindex++) {
-      /* don't forget here is always rounded down */
-      acutValues[aindex] = (alphause[0] * (2*aindex + 1) + alphause[1] * (14 - (2*aindex + 1))) / 14;
-   }
-
-   for (j = 0; j < numypixels; j++) {
-      for (i = 0; i < numxpixels; i++) {
-         /* maybe it's overkill to have the most complicated calculation just for the error
-            calculation which we only need to figure out if encoding1 or encoding2 is better... */
-         if (srccolors[j][i] > acutValues[0]) {
-            alphaenc1[4*j + i] = 0;
-            alphadist = srccolors[j][i] - alphause[1];
-         }
-         else if (srccolors[j][i] > acutValues[1]) {
-            alphaenc1[4*j + i] = 2;
-            alphadist = srccolors[j][i] - (alphause[1] * 6 + alphause[0] * 1) / 7;
-         }
-         else if (srccolors[j][i] > acutValues[2]) {
-            alphaenc1[4*j + i] = 3;
-            alphadist = srccolors[j][i] - (alphause[1] * 5 + alphause[0] * 2) / 7;
-         }
-         else if (srccolors[j][i] > acutValues[3]) {
-            alphaenc1[4*j + i] = 4;
-            alphadist = srccolors[j][i] - (alphause[1] * 4 + alphause[0] * 3) / 7;
-         }
-         else if (srccolors[j][i] > acutValues[4]) {
-            alphaenc1[4*j + i] = 5;
-            alphadist = srccolors[j][i] - (alphause[1] * 3 + alphause[0] * 4) / 7;
-         }
-         else if (srccolors[j][i] > acutValues[5]) {
-            alphaenc1[4*j + i] = 6;
-            alphadist = srccolors[j][i] - (alphause[1] * 2 + alphause[0] * 5) / 7;
-         }
-         else if (srccolors[j][i] > acutValues[6]) {
-            alphaenc1[4*j + i] = 7;
-            alphadist = srccolors[j][i] - (alphause[1] * 1 + alphause[0] * 6) / 7;
-         }
-         else {
-            alphaenc1[4*j + i] = 1;
-            alphadist = srccolors[j][i] - alphause[0];
-         }
-         alphablockerror1 += alphadist * alphadist;
-      }
-   }
-#if RGTC_DEBUG
-   for (i = 0; i < 16; i++) {
-      fprintf(stderr, "%d ", alphaenc1[i]);
-   }
-   fprintf(stderr, "cutVals ");
-   for (i = 0; i < 8; i++) {
-      fprintf(stderr, "%d ", acutValues[i]);
-   }
-   fprintf(stderr, "srcVals ");
-   for (j = 0; j < numypixels; j++)
-      for (i = 0; i < numxpixels; i++) {
-	 fprintf(stderr, "%d ", srccolors[j][i]);
-      }
-   
-   fprintf(stderr, "\n");
-#endif
-
-   /* it's not very likely this encoding is better if both alphaabsmin and alphaabsmax
-      are false but try it anyway */
-   if (alphablockerror1 >= 32) {
-
-      /* don't bother if encoding is already very good, this condition should also imply
-      we have valid alphabase colors which we absolutely need (alphabase[0] <= alphabase[1]) */
-      alphablockerror2 = 0;
-      for (aindex = 0; aindex < 5; aindex++) {
-         /* don't forget here is always rounded down */
-         acutValues[aindex] = (alphabase[0] * (10 - (2*aindex + 1)) + alphabase[1] * (2*aindex + 1)) / 10;
-      }
-      for (j = 0; j < numypixels; j++) {
-         for (i = 0; i < numxpixels; i++) {
-             /* maybe it's overkill to have the most complicated calculation just for the error
-               calculation which we only need to figure out if encoding1 or encoding2 is better... */
-            if (srccolors[j][i] == 0) {
-               alphaenc2[4*j + i] = 6;
-               alphadist = 0;
-            }
-            else if (srccolors[j][i] == 255) {
-               alphaenc2[4*j + i] = 7;
-               alphadist = 0;
-            }
-            else if (srccolors[j][i] <= acutValues[0]) {
-               alphaenc2[4*j + i] = 0;
-               alphadist = srccolors[j][i] - alphabase[0];
-            }
-            else if (srccolors[j][i] <= acutValues[1]) {
-               alphaenc2[4*j + i] = 2;
-               alphadist = srccolors[j][i] - (alphabase[0] * 4 + alphabase[1] * 1) / 5;
-            }
-            else if (srccolors[j][i] <= acutValues[2]) {
-               alphaenc2[4*j + i] = 3;
-               alphadist = srccolors[j][i] - (alphabase[0] * 3 + alphabase[1] * 2) / 5;
-            }
-            else if (srccolors[j][i] <= acutValues[3]) {
-               alphaenc2[4*j + i] = 4;
-               alphadist = srccolors[j][i] - (alphabase[0] * 2 + alphabase[1] * 3) / 5;
-            }
-            else if (srccolors[j][i] <= acutValues[4]) {
-               alphaenc2[4*j + i] = 5;
-               alphadist = srccolors[j][i] - (alphabase[0] * 1 + alphabase[1] * 4) / 5;
-            }
-            else {
-               alphaenc2[4*j + i] = 1;
-               alphadist = srccolors[j][i] - alphabase[1];
-            }
-            alphablockerror2 += alphadist * alphadist;
-         }
-      }
-
-
-      /* skip this if the error is already very small
-         this encoding is MUCH better on average than #2 though, but expensive! */
-      if ((alphablockerror2 > 96) && (alphablockerror1 > 96)) {
-         GLshort blockerrlin1 = 0;
-         GLshort blockerrlin2 = 0;
-         GLubyte nralphainrangelow = 0;
-         GLubyte nralphainrangehigh = 0;
-         alphatest[0] = 0xff;
-         alphatest[1] = 0x0;
-         /* if we have large range it's likely there are values close to 0/255, try to map them to 0/255 */
-         for (j = 0; j < numypixels; j++) {
-            for (i = 0; i < numxpixels; i++) {
-               if ((srccolors[j][i] > alphatest[1]) && (srccolors[j][i] < (255 -(alphabase[1] - alphabase[0]) / 28)))
-                  alphatest[1] = srccolors[j][i];
-               if ((srccolors[j][i] < alphatest[0]) && (srccolors[j][i] > (alphabase[1] - alphabase[0]) / 28))
-                  alphatest[0] = srccolors[j][i];
-            }
-         }
-          /* shouldn't happen too often, don't really care about those degenerated cases */
-          if (alphatest[1] <= alphatest[0]) {
-             alphatest[0] = 1;
-             alphatest[1] = 254;
-         }
-         for (aindex = 0; aindex < 5; aindex++) {
-         /* don't forget here is always rounded down */
-            acutValues[aindex] = (alphatest[0] * (10 - (2*aindex + 1)) + alphatest[1] * (2*aindex + 1)) / 10;
-         }
-
-         /* find the "average" difference between the alpha values and the next encoded value.
-            This is then used to calculate new base values.
-            Should there be some weighting, i.e. those values closer to alphatest[x] have more weight,
-            since they will see more improvement, and also because the values in the middle are somewhat
-            likely to get no improvement at all (because the base values might move in different directions)?
-            OTOH it would mean the values in the middle are even less likely to get an improvement
-         */
-         for (j = 0; j < numypixels; j++) {
-            for (i = 0; i < numxpixels; i++) {
-               if (srccolors[j][i] <= alphatest[0] / 2) {
-               }
-               else if (srccolors[j][i] > ((255 + alphatest[1]) / 2)) {
-               }
-               else if (srccolors[j][i] <= acutValues[0]) {
-                  blockerrlin1 += (srccolors[j][i] - alphatest[0]);
-                  nralphainrangelow += 1;
-               }
-               else if (srccolors[j][i] <= acutValues[1]) {
-                  blockerrlin1 += (srccolors[j][i] - (alphatest[0] * 4 + alphatest[1] * 1) / 5);
-                  blockerrlin2 += (srccolors[j][i] - (alphatest[0] * 4 + alphatest[1] * 1) / 5);
-                  nralphainrangelow += 1;
-                  nralphainrangehigh += 1;
-               }
-               else if (srccolors[j][i] <= acutValues[2]) {
-                  blockerrlin1 += (srccolors[j][i] - (alphatest[0] * 3 + alphatest[1] * 2) / 5);
-                  blockerrlin2 += (srccolors[j][i] - (alphatest[0] * 3 + alphatest[1] * 2) / 5);
-                  nralphainrangelow += 1;
-                  nralphainrangehigh += 1;
-               }
-               else if (srccolors[j][i] <= acutValues[3]) {
-                  blockerrlin1 += (srccolors[j][i] - (alphatest[0] * 2 + alphatest[1] * 3) / 5);
-                  blockerrlin2 += (srccolors[j][i] - (alphatest[0] * 2 + alphatest[1] * 3) / 5);
-                  nralphainrangelow += 1;
-                  nralphainrangehigh += 1;
-               }
-               else if (srccolors[j][i] <= acutValues[4]) {
-                  blockerrlin1 += (srccolors[j][i] - (alphatest[0] * 1 + alphatest[1] * 4) / 5);
-                  blockerrlin2 += (srccolors[j][i] - (alphatest[0] * 1 + alphatest[1] * 4) / 5);
-                  nralphainrangelow += 1;
-                  nralphainrangehigh += 1;
-                  }
-               else {
-                  blockerrlin2 += (srccolors[j][i] - alphatest[1]);
-                  nralphainrangehigh += 1;
-               }
-            }
-         }
-         /* shouldn't happen often, needed to avoid div by zero */
-         if (nralphainrangelow == 0) nralphainrangelow = 1;
-         if (nralphainrangehigh == 0) nralphainrangehigh = 1;
-         alphatest[0] = alphatest[0] + (blockerrlin1 / nralphainrangelow);
-#if RGTC_DEBUG
-	 fprintf(stderr, "block err lin low %d, nr %d\n", blockerrlin1, nralphainrangelow);
-         fprintf(stderr, "block err lin high %d, nr %d\n", blockerrlin2, nralphainrangehigh);
-#endif
-         /* again shouldn't really happen often... */
-         if (alphatest[0] < 0) {
-            alphatest[0] = 0;
-         }
-         alphatest[1] = alphatest[1] + (blockerrlin2 / nralphainrangehigh);
-         if (alphatest[1] > 255) {
-            alphatest[1] = 255;
-         }
-
-         alphablockerror3 = 0;
-         for (aindex = 0; aindex < 5; aindex++) {
-         /* don't forget here is always rounded down */
-            acutValues[aindex] = (alphatest[0] * (10 - (2*aindex + 1)) + alphatest[1] * (2*aindex + 1)) / 10;
-         }
-         for (j = 0; j < numypixels; j++) {
-            for (i = 0; i < numxpixels; i++) {
-                /* maybe it's overkill to have the most complicated calculation just for the error
-                  calculation which we only need to figure out if encoding1 or encoding2 is better... */
-               if (srccolors[j][i] <= alphatest[0] / 2) {
-                  alphaenc3[4*j + i] = 6;
-                  alphadist = srccolors[j][i];
-               }
-               else if (srccolors[j][i] > ((255 + alphatest[1]) / 2)) {
-                  alphaenc3[4*j + i] = 7;
-                  alphadist = 255 - srccolors[j][i];
-               }
-               else if (srccolors[j][i] <= acutValues[0]) {
-                  alphaenc3[4*j + i] = 0;
-                  alphadist = srccolors[j][i] - alphatest[0];
-               }
-               else if (srccolors[j][i] <= acutValues[1]) {
-                 alphaenc3[4*j + i] = 2;
-                 alphadist = srccolors[j][i] - (alphatest[0] * 4 + alphatest[1] * 1) / 5;
-               }
-               else if (srccolors[j][i] <= acutValues[2]) {
-                  alphaenc3[4*j + i] = 3;
-                  alphadist = srccolors[j][i] - (alphatest[0] * 3 + alphatest[1] * 2) / 5;
-               }
-               else if (srccolors[j][i] <= acutValues[3]) {
-                  alphaenc3[4*j + i] = 4;
-                  alphadist = srccolors[j][i] - (alphatest[0] * 2 + alphatest[1] * 3) / 5;
-               }
-               else if (srccolors[j][i] <= acutValues[4]) {
-                  alphaenc3[4*j + i] = 5;
-                  alphadist = srccolors[j][i] - (alphatest[0] * 1 + alphatest[1] * 4) / 5;
-               }
-               else {
-                  alphaenc3[4*j + i] = 1;
-                  alphadist = srccolors[j][i] - alphatest[1];
-               }
-               alphablockerror3 += alphadist * alphadist;
-            }
-         }
-      }
-   }
-  /* write the alpha values and encoding back. */
-   if ((alphablockerror1 <= alphablockerror2) && (alphablockerror1 <= alphablockerror3)) {
-#if RGTC_DEBUG
-      if (alphablockerror1 > 96) fprintf(stderr, "enc1 used, error %d\n", alphablockerror1);
-#endif
-      write_rgtc_encoded_channel_s( blkaddr, alphause[1], alphause[0], alphaenc1 );
-   }
-   else if (alphablockerror2 <= alphablockerror3) {
-#if RGTC_DEBUG
-      if (alphablockerror2 > 96) fprintf(stderr, "enc2 used, error %d\n", alphablockerror2);
-#endif
-      write_rgtc_encoded_channel_s( blkaddr, alphabase[0], alphabase[1], alphaenc2 );
-   }
-   else {
-#if RGTC_DEBUG
-      fprintf(stderr, "enc3 used, error %d\n", alphablockerror3);
-#endif
-      write_rgtc_encoded_channel_s( blkaddr, (GLubyte)alphatest[0], (GLubyte)alphatest[1], alphaenc3 );
-   }
-}
+#undef TAG
+#undef TYPE
+#undef T_MIN
+#undef T_MAX
diff --git a/src/mesa/main/texcompress_rgtc_tmp.h b/src/mesa/main/texcompress_rgtc_tmp.h
new file mode 100644
index 00000000000..c8bf082a158
--- /dev/null
+++ b/src/mesa/main/texcompress_rgtc_tmp.h
@@ -0,0 +1,418 @@
+/*
+ * Copyright (C) 2011 Red Hat Inc.
+ * 
+ * block compression parts are:
+ * Copyright (C) 2004  Roland Scheidegger   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Author:
+ *    Dave Airlie
+ */
+
+/* included by texcompress_rgtc to define byte/ubyte compressors */
+
+static void TAG(fetch_texel_rgtc)(unsigned srcRowStride, const TYPE *pixdata,
+				  unsigned i, unsigned j, TYPE *value, unsigned comps)
+{
+   TYPE decode;
+   const TYPE *blksrc = (pixdata + ((srcRowStride + 3) / 4 * (j / 4) + (i / 4)) * 8 * comps);
+   const TYPE alpha0 = blksrc[0];
+   const TYPE alpha1 = blksrc[1];
+   const char bit_pos = ((j&3) * 4 + (i&3)) * 3;
+   const TYPE acodelow = blksrc[2 + bit_pos / 8];
+   const TYPE acodehigh = (3 + bit_pos / 8) < 8 ? blksrc[3 + bit_pos / 8] : 0;
+   const TYPE code = (acodelow >> (bit_pos & 0x7) |
+      (acodehigh  << (8 - (bit_pos & 0x7)))) & 0x7;
+
+   if (code == 0)
+      decode = alpha0;
+   else if (code == 1)
+      decode = alpha1;
+   else if (alpha0 > alpha1)
+      decode = ((alpha0 * (8 - code) + (alpha1 * (code - 1))) / 7);
+   else if (code < 6)
+      decode = ((alpha0 * (6 - code) + (alpha1 * (code - 1))) / 5);
+   else if (code == 6)
+      decode = T_MIN;
+   else
+      decode = T_MAX;
+
+   *value = decode;
+}
+
+static void TAG(write_rgtc_encoded_channel)(TYPE *blkaddr,
+					    TYPE alphabase1,
+					    TYPE alphabase2,
+					    TYPE alphaenc[16])
+{
+   *blkaddr++ = alphabase1;
+   *blkaddr++ = alphabase2;
+   *blkaddr++ = alphaenc[0] | (alphaenc[1] << 3) | ((alphaenc[2] & 3) << 6);
+   *blkaddr++ = (alphaenc[2] >> 2) | (alphaenc[3] << 1) | (alphaenc[4] << 4) | ((alphaenc[5] & 1) << 7);
+   *blkaddr++ = (alphaenc[5] >> 1) | (alphaenc[6] << 2) | (alphaenc[7] << 5);
+   *blkaddr++ = alphaenc[8] | (alphaenc[9] << 3) | ((alphaenc[10] & 3) << 6);
+   *blkaddr++ = (alphaenc[10] >> 2) | (alphaenc[11] << 1) | (alphaenc[12] << 4) | ((alphaenc[13] & 1) << 7);
+   *blkaddr++ = (alphaenc[13] >> 1) | (alphaenc[14] << 2) | (alphaenc[15] << 5);
+}
+
+static void TAG(encode_rgtc_chan)(TYPE *blkaddr, TYPE srccolors[4][4],
+			     int numxpixels, int numypixels)
+{
+   TYPE alphabase[2], alphause[2];
+   short alphatest[2] = { 0 };
+   unsigned int alphablockerror1, alphablockerror2, alphablockerror3;
+   TYPE i, j, aindex, acutValues[7];
+   TYPE alphaenc1[16], alphaenc2[16], alphaenc3[16];
+   int alphaabsmin = 0, alphaabsmax = 0;
+   short alphadist;
+
+   /* find lowest and highest alpha value in block, alphabase[0] lowest, alphabase[1] highest */
+   alphabase[0] = T_MAX; alphabase[1] = T_MIN;
+   for (j = 0; j < numypixels; j++) {
+      for (i = 0; i < numxpixels; i++) {
+	 if (srccolors[j][i] == T_MIN)
+            alphaabsmin = 1;
+         else if (srccolors[j][i] == T_MAX)
+            alphaabsmax = 1;
+         else {
+            if (srccolors[j][i] > alphabase[1])
+               alphabase[1] = srccolors[j][i];
+            if (srccolors[j][i] < alphabase[0])
+               alphabase[0] = srccolors[j][i];
+         }
+      }
+   }
+
+
+   if (((alphabase[0] > alphabase[1]) && !(alphaabsmin && alphaabsmax))
+       || (alphabase[0] == alphabase[1] && !alphaabsmin && !alphaabsmax)) { /* one color, either max or min */
+      /* shortcut here since it is a very common case (and also avoids later problems) */
+      /* could also thest for alpha0 == alpha1 (and not min/max), but probably not common, so don't bother */
+
+      *blkaddr++ = srccolors[0][0];
+      blkaddr++;
+      *blkaddr++ = 0;
+      *blkaddr++ = 0;
+      *blkaddr++ = 0;
+      *blkaddr++ = 0;
+      *blkaddr++ = 0;
+      *blkaddr++ = 0;
+#if RGTC_DEBUG
+      fprintf(stderr, "enc0 used\n");
+#endif
+      return;
+   }
+
+   /* find best encoding for alpha0 > alpha1 */
+   /* it's possible this encoding is better even if both alphaabsmin and alphaabsmax are true */
+   alphablockerror1 = 0x0;
+   alphablockerror2 = 0xffffffff;
+   alphablockerror3 = 0xffffffff;
+   if (alphaabsmin) alphause[0] = T_MIN;
+   else alphause[0] = alphabase[0];
+   if (alphaabsmax) alphause[1] = T_MAX;
+   else alphause[1] = alphabase[1];
+   /* calculate the 7 cut values, just the middle between 2 of the computed alpha values */
+   for (aindex = 0; aindex < 7; aindex++) {
+      /* don't forget here is always rounded down */
+      acutValues[aindex] = (alphause[0] * (2*aindex + 1) + alphause[1] * (14 - (2*aindex + 1))) / 14;
+   }
+
+   for (j = 0; j < numypixels; j++) {
+      for (i = 0; i < numxpixels; i++) {
+         /* maybe it's overkill to have the most complicated calculation just for the error
+            calculation which we only need to figure out if encoding1 or encoding2 is better... */
+         if (srccolors[j][i] > acutValues[0]) {
+            alphaenc1[4*j + i] = 0;
+            alphadist = srccolors[j][i] - alphause[1];
+         }
+         else if (srccolors[j][i] > acutValues[1]) {
+            alphaenc1[4*j + i] = 2;
+            alphadist = srccolors[j][i] - (alphause[1] * 6 + alphause[0] * 1) / 7;
+         }
+         else if (srccolors[j][i] > acutValues[2]) {
+            alphaenc1[4*j + i] = 3;
+            alphadist = srccolors[j][i] - (alphause[1] * 5 + alphause[0] * 2) / 7;
+         }
+         else if (srccolors[j][i] > acutValues[3]) {
+            alphaenc1[4*j + i] = 4;
+            alphadist = srccolors[j][i] - (alphause[1] * 4 + alphause[0] * 3) / 7;
+         }
+         else if (srccolors[j][i] > acutValues[4]) {
+            alphaenc1[4*j + i] = 5;
+            alphadist = srccolors[j][i] - (alphause[1] * 3 + alphause[0] * 4) / 7;
+         }
+         else if (srccolors[j][i] > acutValues[5]) {
+            alphaenc1[4*j + i] = 6;
+            alphadist = srccolors[j][i] - (alphause[1] * 2 + alphause[0] * 5) / 7;
+         }
+         else if (srccolors[j][i] > acutValues[6]) {
+            alphaenc1[4*j + i] = 7;
+            alphadist = srccolors[j][i] - (alphause[1] * 1 + alphause[0] * 6) / 7;
+         }
+         else {
+            alphaenc1[4*j + i] = 1;
+            alphadist = srccolors[j][i] - alphause[0];
+         }
+         alphablockerror1 += alphadist * alphadist;
+      }
+   }
+
+#if RGTC_DEBUG
+   for (i = 0; i < 16; i++) {
+      fprintf(stderr, "%d ", alphaenc1[i]);
+   }
+   fprintf(stderr, "cutVals ");
+   for (i = 0; i < 8; i++) {
+      fprintf(stderr, "%d ", acutValues[i]);
+   }
+   fprintf(stderr, "srcVals ");
+   for (j = 0; j < numypixels; j++) {
+      for (i = 0; i < numxpixels; i++) {
+	 fprintf(stderr, "%d ", srccolors[j][i]);
+      }
+   }
+   fprintf(stderr, "\n");
+#endif
+
+   /* it's not very likely this encoding is better if both alphaabsmin and alphaabsmax
+      are false but try it anyway */
+   if (alphablockerror1 >= 32) {
+
+      /* don't bother if encoding is already very good, this condition should also imply
+      we have valid alphabase colors which we absolutely need (alphabase[0] <= alphabase[1]) */
+      alphablockerror2 = 0;
+      for (aindex = 0; aindex < 5; aindex++) {
+         /* don't forget here is always rounded down */
+         acutValues[aindex] = (alphabase[0] * (10 - (2*aindex + 1)) + alphabase[1] * (2*aindex + 1)) / 10;
+      }
+      for (j = 0; j < numypixels; j++) {
+         for (i = 0; i < numxpixels; i++) {
+             /* maybe it's overkill to have the most complicated calculation just for the error
+               calculation which we only need to figure out if encoding1 or encoding2 is better... */
+            if (srccolors[j][i] == T_MIN) {
+               alphaenc2[4*j + i] = 6;
+               alphadist = 0;
+            }
+            else if (srccolors[j][i] == T_MAX) {
+               alphaenc2[4*j + i] = 7;
+               alphadist = 0;
+            }
+            else if (srccolors[j][i] <= acutValues[0]) {
+               alphaenc2[4*j + i] = 0;
+               alphadist = srccolors[j][i] - alphabase[0];
+            }
+            else if (srccolors[j][i] <= acutValues[1]) {
+               alphaenc2[4*j + i] = 2;
+               alphadist = srccolors[j][i] - (alphabase[0] * 4 + alphabase[1] * 1) / 5;
+            }
+            else if (srccolors[j][i] <= acutValues[2]) {
+               alphaenc2[4*j + i] = 3;
+               alphadist = srccolors[j][i] - (alphabase[0] * 3 + alphabase[1] * 2) / 5;
+            }
+            else if (srccolors[j][i] <= acutValues[3]) {
+               alphaenc2[4*j + i] = 4;
+               alphadist = srccolors[j][i] - (alphabase[0] * 2 + alphabase[1] * 3) / 5;
+            }
+            else if (srccolors[j][i] <= acutValues[4]) {
+               alphaenc2[4*j + i] = 5;
+               alphadist = srccolors[j][i] - (alphabase[0] * 1 + alphabase[1] * 4) / 5;
+            }
+            else {
+               alphaenc2[4*j + i] = 1;
+               alphadist = srccolors[j][i] - alphabase[1];
+            }
+            alphablockerror2 += alphadist * alphadist;
+         }
+      }
+
+
+      /* skip this if the error is already very small
+         this encoding is MUCH better on average than #2 though, but expensive! */
+      if ((alphablockerror2 > 96) && (alphablockerror1 > 96)) {
+         short blockerrlin1 = 0;
+         short blockerrlin2 = 0;
+         TYPE nralphainrangelow = 0;
+         TYPE nralphainrangehigh = 0;
+         alphatest[0] = T_MAX;
+         alphatest[1] = T_MIN;
+         /* if we have large range it's likely there are values close to 0/255, try to map them to 0/255 */
+         for (j = 0; j < numypixels; j++) {
+            for (i = 0; i < numxpixels; i++) {
+               if ((srccolors[j][i] > alphatest[1]) && (srccolors[j][i] < (T_MAX -(alphabase[1] - alphabase[0]) / 28)))
+                  alphatest[1] = srccolors[j][i];
+               if ((srccolors[j][i] < alphatest[0]) && (srccolors[j][i] > (alphabase[1] - alphabase[0]) / 28))
+                  alphatest[0] = srccolors[j][i];
+            }
+         }
+          /* shouldn't happen too often, don't really care about those degenerated cases */
+          if (alphatest[1] <= alphatest[0]) {
+             alphatest[0] = T_MIN+1;
+             alphatest[1] = T_MAX-1;
+         }
+         for (aindex = 0; aindex < 5; aindex++) {
+         /* don't forget here is always rounded down */
+            acutValues[aindex] = (alphatest[0] * (10 - (2*aindex + 1)) + alphatest[1] * (2*aindex + 1)) / 10;
+         }
+
+         /* find the "average" difference between the alpha values and the next encoded value.
+            This is then used to calculate new base values.
+            Should there be some weighting, i.e. those values closer to alphatest[x] have more weight,
+            since they will see more improvement, and also because the values in the middle are somewhat
+            likely to get no improvement at all (because the base values might move in different directions)?
+            OTOH it would mean the values in the middle are even less likely to get an improvement
+         */
+         for (j = 0; j < numypixels; j++) {
+            for (i = 0; i < numxpixels; i++) {
+               if (srccolors[j][i] <= alphatest[0] / 2) {
+               }
+               else if (srccolors[j][i] > ((T_MAX + alphatest[1]) / 2)) {
+               }
+               else if (srccolors[j][i] <= acutValues[0]) {
+                  blockerrlin1 += (srccolors[j][i] - alphatest[0]);
+                  nralphainrangelow += 1;
+               }
+               else if (srccolors[j][i] <= acutValues[1]) {
+                  blockerrlin1 += (srccolors[j][i] - (alphatest[0] * 4 + alphatest[1] * 1) / 5);
+                  blockerrlin2 += (srccolors[j][i] - (alphatest[0] * 4 + alphatest[1] * 1) / 5);
+                  nralphainrangelow += 1;
+                  nralphainrangehigh += 1;
+               }
+               else if (srccolors[j][i] <= acutValues[2]) {
+                  blockerrlin1 += (srccolors[j][i] - (alphatest[0] * 3 + alphatest[1] * 2) / 5);
+                  blockerrlin2 += (srccolors[j][i] - (alphatest[0] * 3 + alphatest[1] * 2) / 5);
+                  nralphainrangelow += 1;
+                  nralphainrangehigh += 1;
+               }
+               else if (srccolors[j][i] <= acutValues[3]) {
+                  blockerrlin1 += (srccolors[j][i] - (alphatest[0] * 2 + alphatest[1] * 3) / 5);
+                  blockerrlin2 += (srccolors[j][i] - (alphatest[0] * 2 + alphatest[1] * 3) / 5);
+                  nralphainrangelow += 1;
+                  nralphainrangehigh += 1;
+               }
+               else if (srccolors[j][i] <= acutValues[4]) {
+                  blockerrlin1 += (srccolors[j][i] - (alphatest[0] * 1 + alphatest[1] * 4) / 5);
+                  blockerrlin2 += (srccolors[j][i] - (alphatest[0] * 1 + alphatest[1] * 4) / 5);
+                  nralphainrangelow += 1;
+                  nralphainrangehigh += 1;
+                  }
+               else {
+                  blockerrlin2 += (srccolors[j][i] - alphatest[1]);
+                  nralphainrangehigh += 1;
+               }
+            }
+         }
+         /* shouldn't happen often, needed to avoid div by zero */
+         if (nralphainrangelow == 0) nralphainrangelow = 1;
+         if (nralphainrangehigh == 0) nralphainrangehigh = 1;
+         alphatest[0] = alphatest[0] + (blockerrlin1 / nralphainrangelow);
+#if RGTC_DEBUG
+         fprintf(stderr, "block err lin low %d, nr %d\n", blockerrlin1, nralphainrangelow);
+         fprintf(stderr, "block err lin high %d, nr %d\n", blockerrlin2, nralphainrangehigh);
+#endif
+         /* again shouldn't really happen often... */
+         if (alphatest[0] < T_MIN) {
+            alphatest[0] = T_MIN;
+         }
+         alphatest[1] = alphatest[1] + (blockerrlin2 / nralphainrangehigh);
+         if (alphatest[1] > T_MAX) {
+            alphatest[1] = T_MAX;
+         }
+
+         alphablockerror3 = 0;
+         for (aindex = 0; aindex < 5; aindex++) {
+         /* don't forget here is always rounded down */
+            acutValues[aindex] = (alphatest[0] * (10 - (2*aindex + 1)) + alphatest[1] * (2*aindex + 1)) / 10;
+         }
+         for (j = 0; j < numypixels; j++) {
+            for (i = 0; i < numxpixels; i++) {
+                /* maybe it's overkill to have the most complicated calculation just for the error
+                  calculation which we only need to figure out if encoding1 or encoding2 is better... */
+               if (srccolors[j][i] <= alphatest[0] / 2) {
+                  alphaenc3[4*j + i] = 6;
+                  alphadist = srccolors[j][i];
+               }
+               else if (srccolors[j][i] > ((T_MAX + alphatest[1]) / 2)) {
+                  alphaenc3[4*j + i] = 7;
+                  alphadist = T_MAX - srccolors[j][i];
+               }
+               else if (srccolors[j][i] <= acutValues[0]) {
+                  alphaenc3[4*j + i] = 0;
+                  alphadist = srccolors[j][i] - alphatest[0];
+               }
+               else if (srccolors[j][i] <= acutValues[1]) {
+                 alphaenc3[4*j + i] = 2;
+                 alphadist = srccolors[j][i] - (alphatest[0] * 4 + alphatest[1] * 1) / 5;
+               }
+               else if (srccolors[j][i] <= acutValues[2]) {
+                  alphaenc3[4*j + i] = 3;
+                  alphadist = srccolors[j][i] - (alphatest[0] * 3 + alphatest[1] * 2) / 5;
+               }
+               else if (srccolors[j][i] <= acutValues[3]) {
+                  alphaenc3[4*j + i] = 4;
+                  alphadist = srccolors[j][i] - (alphatest[0] * 2 + alphatest[1] * 3) / 5;
+               }
+               else if (srccolors[j][i] <= acutValues[4]) {
+                  alphaenc3[4*j + i] = 5;
+                  alphadist = srccolors[j][i] - (alphatest[0] * 1 + alphatest[1] * 4) / 5;
+               }
+               else {
+                  alphaenc3[4*j + i] = 1;
+                  alphadist = srccolors[j][i] - alphatest[1];
+               }
+               alphablockerror3 += alphadist * alphadist;
+            }
+         }
+      }
+   }
+
+  /* write the alpha values and encoding back. */
+   if ((alphablockerror1 <= alphablockerror2) && (alphablockerror1 <= alphablockerror3)) {
+#if RGTC_DEBUG
+      if (alphablockerror1 > 96) fprintf(stderr, "enc1 used, error %d\n", alphablockerror1);
+      fprintf(stderr,"w1: min %d max %d au0 %d au1 %d\n",
+	      T_MIN, T_MAX,
+	      alphause[1], alphause[0]);
+#endif
+
+      TAG(write_rgtc_encoded_channel)( blkaddr, alphause[1], alphause[0], alphaenc1 );
+   }
+   else if (alphablockerror2 <= alphablockerror3) {
+#if RGTC_DEBUG
+      if (alphablockerror2 > 96) fprintf(stderr, "enc2 used, error %d\n", alphablockerror2);
+      fprintf(stderr,"w2: min %d max %d au0 %d au1 %d\n",
+	      T_MIN, T_MAX,
+	      alphabase[0], alphabase[1]);
+#endif
+
+      TAG(write_rgtc_encoded_channel)( blkaddr, alphabase[0], alphabase[1], alphaenc2 );
+   }
+   else {
+#if RGTC_DEBUG
+      fprintf(stderr, "enc3 used, error %d\n", alphablockerror3);
+      fprintf(stderr,"w3: min %d max %d au0 %d au1 %d\n",
+	      T_MIN, T_MAX,
+	      alphatest[0], alphatest[1]);
+#endif
+
+      TAG(write_rgtc_encoded_channel)( blkaddr, (TYPE)alphatest[0], (TYPE)alphatest[1], alphaenc3 );
+   }
+}
diff --git a/src/mesa/main/texgetimage.c b/src/mesa/main/texgetimage.c
index 28829694153..21d9140c550 100644
--- a/src/mesa/main/texgetimage.c
+++ b/src/mesa/main/texgetimage.c
@@ -38,6 +38,7 @@
 #include "mfeatures.h"
 #include "mtypes.h"
 #include "pack.h"
+#include "pbo.h"
 #include "texgetimage.h"
 #include "teximage.h"
 
diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c
index 1f2ad7f266b..d4ae6dd69d8 100644
--- a/src/mesa/main/teximage.c
+++ b/src/mesa/main/teximage.c
@@ -1898,7 +1898,9 @@ copytexture_error_check( struct gl_context *ctx, GLuint dimensions,
 
    /* Check that the source buffer is complete */
    if (ctx->ReadBuffer->Name) {
-      _mesa_test_framebuffer_completeness(ctx, ctx->ReadBuffer);
+      if (ctx->ReadBuffer->_Status == 0) {
+         _mesa_test_framebuffer_completeness(ctx, ctx->ReadBuffer);
+      }
       if (ctx->ReadBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
          _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT,
                      "glCopyTexImage%dD(invalid readbuffer)", dimensions);
@@ -2000,7 +2002,9 @@ copytexsubimage_error_check1( struct gl_context *ctx, GLuint dimensions,
 {
    /* Check that the source buffer is complete */
    if (ctx->ReadBuffer->Name) {
-      _mesa_test_framebuffer_completeness(ctx, ctx->ReadBuffer);
+      if (ctx->ReadBuffer->_Status == 0) {
+         _mesa_test_framebuffer_completeness(ctx, ctx->ReadBuffer);
+      }
       if (ctx->ReadBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
          _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT,
                      "glCopyTexImage%dD(invalid readbuffer)", dimensions);
diff --git a/src/mesa/main/texstate.c b/src/mesa/main/texstate.c
index f4d77189f29..41d531f5976 100644
--- a/src/mesa/main/texstate.c
+++ b/src/mesa/main/texstate.c
@@ -818,9 +818,6 @@ _mesa_free_texture_data(struct gl_context *ctx)
    /* Free proxy texture objects */
    for (tgt = 0; tgt < NUM_TEXTURE_TARGETS; tgt++)
       ctx->Driver.DeleteTexture(ctx, ctx->Texture.ProxyTex[tgt]);
-
-   for (u = 0; u < Elements(ctx->Texture.Unit); u++)
-      _mesa_free_colortable_data(&ctx->Texture.Unit[u].ColorTable);
 }
 
 
diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c
index 8a3e5f77979..cd30fa02149 100644
--- a/src/mesa/main/texstore.c
+++ b/src/mesa/main/texstore.c
@@ -61,6 +61,7 @@
 #include "mfeatures.h"
 #include "mtypes.h"
 #include "pack.h"
+#include "pbo.h"
 #include "imports.h"
 #include "pack.h"
 #include "texcompress.h"
@@ -4200,94 +4201,6 @@ _mesa_texstore(TEXSTORE_PARAMS)
 }
 
 
-/**
- * Check if an unpack PBO is active prior to fetching a texture image.
- * If so, do bounds checking and map the buffer into main memory.
- * Any errors detected will be recorded.
- * The caller _must_ call _mesa_unmap_teximage_pbo() too!
- */
-const GLvoid *
-_mesa_validate_pbo_teximage(struct gl_context *ctx, GLuint dimensions,
-			    GLsizei width, GLsizei height, GLsizei depth,
-			    GLenum format, GLenum type, const GLvoid *pixels,
-			    const struct gl_pixelstore_attrib *unpack,
-			    const char *funcName)
-{
-   GLubyte *buf;
-
-   if (!_mesa_is_bufferobj(unpack->BufferObj)) {
-      /* no PBO */
-      return pixels;
-   }
-   if (!_mesa_validate_pbo_access(dimensions, unpack, width, height, depth,
-                                  format, type, pixels)) {
-      _mesa_error(ctx, GL_INVALID_OPERATION, funcName, "(invalid PBO access)");
-      return NULL;
-   }
-
-   buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                                          GL_READ_ONLY_ARB, unpack->BufferObj);
-   if (!buf) {
-      _mesa_error(ctx, GL_INVALID_OPERATION, funcName, "(PBO is mapped)");
-      return NULL;
-   }
-
-   return ADD_POINTERS(buf, pixels);
-}
-
-
-/**
- * Check if an unpack PBO is active prior to fetching a compressed texture
- * image.
- * If so, do bounds checking and map the buffer into main memory.
- * Any errors detected will be recorded.
- * The caller _must_ call _mesa_unmap_teximage_pbo() too!
- */
-const GLvoid *
-_mesa_validate_pbo_compressed_teximage(struct gl_context *ctx,
-                                 GLsizei imageSize, const GLvoid *pixels,
-                                 const struct gl_pixelstore_attrib *packing,
-                                 const char *funcName)
-{
-   GLubyte *buf;
-
-   if (!_mesa_is_bufferobj(packing->BufferObj)) {
-      /* not using a PBO - return pointer unchanged */
-      return pixels;
-   }
-   if ((const GLubyte *) pixels + imageSize >
-       ((const GLubyte *) 0) + packing->BufferObj->Size) {
-      /* out of bounds read! */
-      _mesa_error(ctx, GL_INVALID_OPERATION, funcName, "(invalid PBO access)");
-      return NULL;
-   }
-
-   buf = (GLubyte*) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                                         GL_READ_ONLY_ARB, packing->BufferObj);
-   if (!buf) {
-      _mesa_error(ctx, GL_INVALID_OPERATION, funcName, "(PBO is mapped");
-      return NULL;
-   }
-
-   return ADD_POINTERS(buf, pixels);
-}
-
-
-/**
- * This function must be called after either of the validate_pbo_*_teximage()
- * functions.  It unmaps the PBO buffer if it was mapped earlier.
- */
-void
-_mesa_unmap_teximage_pbo(struct gl_context *ctx,
-                         const struct gl_pixelstore_attrib *unpack)
-{
-   if (_mesa_is_bufferobj(unpack->BufferObj)) {
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                              unpack->BufferObj);
-   }
-}
-
-
 /** Return texture size in bytes */
 static GLuint
 texture_size(const struct gl_texture_image *texImage)
diff --git a/src/mesa/main/texstore.h b/src/mesa/main/texstore.h
index 2f3c4e821fc..d563187098c 100644
--- a/src/mesa/main/texstore.h
+++ b/src/mesa/main/texstore.h
@@ -206,22 +206,4 @@ _mesa_store_compressed_texsubimage3d(struct gl_context *ctx, GLenum target,
                                 struct gl_texture_image *texImage);
 
 
-extern const GLvoid *
-_mesa_validate_pbo_teximage(struct gl_context *ctx, GLuint dimensions,
-			    GLsizei width, GLsizei height, GLsizei depth,
-			    GLenum format, GLenum type, const GLvoid *pixels,
-			    const struct gl_pixelstore_attrib *unpack,
-			    const char *funcName);
-
-extern const GLvoid *
-_mesa_validate_pbo_compressed_teximage(struct gl_context *ctx,
-                                    GLsizei imageSize, const GLvoid *pixels,
-                                    const struct gl_pixelstore_attrib *packing,
-                                    const char *funcName);
-
-extern void
-_mesa_unmap_teximage_pbo(struct gl_context *ctx,
-                         const struct gl_pixelstore_attrib *unpack);
-
-
 #endif