8 files changed, 1282 insertions, 1285 deletions
diff --git a/mesalib/src/mesa/vbo/vbo_exec_api.c b/mesalib/src/mesa/vbo/vbo_exec_api.c
index 2b8d38ef2..8474c787a 100644
--- a/mesalib/src/mesa/vbo/vbo_exec_api.c
+++ b/mesalib/src/mesa/vbo/vbo_exec_api.c
@@ -431,6 +431,24 @@ do {									\
 #include "vbo_attrib_tmp.h"
 
 
+/**
+ * Flush (draw) vertices.
+ * \param  unmap - leave VBO unmapped after flushing?
+ */
+static void
+vbo_exec_FlushVertices_internal(struct vbo_exec_context *exec, GLboolean unmap)
+{
+   if (exec->vtx.vert_count || unmap) {
+      vbo_exec_vtx_flush( exec, unmap );
+   }
+
+   if (exec->vtx.vertex_size) {
+      vbo_exec_copy_to_current( exec );
+      reset_attrfv( exec );
+   }
+}
+
+
 #if FEATURE_beginend
 
 
@@ -535,24 +553,6 @@ static void GLAPIENTRY vbo_exec_EvalPoint2( GLint i, GLint j )
 
 
 /**
- * Flush (draw) vertices.
- * \param  unmap - leave VBO unmapped after flushing?
- */
-static void
-vbo_exec_FlushVertices_internal(struct vbo_exec_context *exec, GLboolean unmap)
-{
-   if (exec->vtx.vert_count || unmap) {
-      vbo_exec_vtx_flush( exec, unmap );
-   }
-
-   if (exec->vtx.vertex_size) {
-      vbo_exec_copy_to_current( exec );
-      reset_attrfv( exec );
-   }
-}
-
-
-/**
  * Called via glBegin.
  */
 static void GLAPIENTRY vbo_exec_Begin( GLenum mode )
@@ -947,7 +947,7 @@ void vbo_exec_vtx_destroy( struct vbo_exec_context *exec )
    /* Free the vertex buffer.  Unmap first if needed.
     */
    if (_mesa_bufferobj_mapped(exec->vtx.bufferobj)) {
-      ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, exec->vtx.bufferobj);
+      ctx->Driver.UnmapBuffer(ctx, exec->vtx.bufferobj);
    }
    _mesa_reference_buffer_object(ctx, &exec->vtx.bufferobj, NULL);
 }
diff --git a/mesalib/src/mesa/vbo/vbo_exec_array.c b/mesalib/src/mesa/vbo/vbo_exec_array.c
index 32ce0e4a8..18719d5f5 100644
--- a/mesalib/src/mesa/vbo/vbo_exec_array.c
+++ b/mesalib/src/mesa/vbo/vbo_exec_array.c
@@ -95,10 +95,25 @@ vbo_get_minmax_index(struct gl_context *ctx,
    GLuint i;
 
    if (_mesa_is_bufferobj(ib->obj)) {
-      const GLvoid *map =
-         ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER_ARB,
-                               GL_READ_ONLY, ib->obj);
-      indices = ADD_POINTERS(map, ib->ptr);
+      unsigned map_size;
+
+      switch (ib->type) {
+      case GL_UNSIGNED_INT:
+	 map_size = count * sizeof(GLuint);
+	 break;
+      case GL_UNSIGNED_SHORT:
+	 map_size = count * sizeof(GLushort);
+	 break;
+      case GL_UNSIGNED_BYTE:
+	 map_size = count * sizeof(GLubyte);
+	 break;
+      default:
+	 assert(0);
+	 map_size = 0;
+      }
+
+      indices = ctx->Driver.MapBufferRange(ctx, (GLsizeiptr) ib->ptr, map_size,
+					   GL_MAP_READ_BIT, ib->obj);
    } else {
       indices = ib->ptr;
    }
@@ -176,7 +191,7 @@ vbo_get_minmax_index(struct gl_context *ctx,
    }
 
    if (_mesa_is_bufferobj(ib->obj)) {
-      ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER_ARB, ib->obj);
+      ctx->Driver.UnmapBuffer(ctx, ib->obj);
    }
 }
 
@@ -196,8 +211,8 @@ check_array_data(struct gl_context *ctx, struct gl_client_array *array,
          if (!array->BufferObj->Pointer) {
             /* need to map now */
             array->BufferObj->Pointer =
-               ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER_ARB,
-                                     GL_READ_ONLY, array->BufferObj);
+               ctx->Driver.MapBufferRange(ctx, 0, array->BufferObj->Size,
+					  GL_MAP_READ_BIT, array->BufferObj);
          }
          data = ADD_POINTERS(data, array->BufferObj->Pointer);
       }
@@ -238,7 +253,7 @@ unmap_array_buffer(struct gl_context *ctx, struct gl_client_array *array)
    if (array->Enabled &&
        _mesa_is_bufferobj(array->BufferObj) &&
        _mesa_bufferobj_mapped(array->BufferObj)) {
-      ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER_ARB, array->BufferObj);
+      ctx->Driver.UnmapBuffer(ctx, array->BufferObj);
    }
 }
 
@@ -256,10 +271,10 @@ check_draw_elements_data(struct gl_context *ctx, GLsizei count, GLenum elemType,
    GLint i, k;
 
    if (_mesa_is_bufferobj(ctx->Array.ElementArrayBufferObj)) {
-      elemMap = ctx->Driver.MapBuffer(ctx,
-                                      GL_ELEMENT_ARRAY_BUFFER_ARB,
-                                      GL_READ_ONLY,
-                                      ctx->Array.ElementArrayBufferObj);
+      elemMap = ctx->Driver.MapBufferRange(ctx, 0,
+					   ctx->Array.ElementArrayBufferObj->Size,
+					   GL_MAP_READ_BIT,
+					   ctx->Array.ElementArrayBufferObj);
       elements = ADD_POINTERS(elements, elemMap);
    }
 
@@ -296,8 +311,7 @@ check_draw_elements_data(struct gl_context *ctx, GLsizei count, GLenum elemType,
    }
 
    if (_mesa_is_bufferobj(ctx->Array.ElementArrayBufferObj)) {
-      ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER_ARB,
-			      ctx->Array.ElementArrayBufferObj);
+      ctx->Driver.UnmapBuffer(ctx, ctx->Array.ElementArrayBufferObj);
    }
 
    unmap_array_buffer(ctx, &arrayObj->Vertex);
@@ -351,8 +365,8 @@ print_draw_arrays(struct gl_context *ctx,
 	     bufName);
 
       if (bufName) {
-         GLubyte *p = ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER_ARB,
-                                            GL_READ_ONLY_ARB, bufObj);
+         GLubyte *p = ctx->Driver.MapBufferRange(ctx, 0, bufObj->Size,
+						 GL_MAP_READ_BIT, bufObj);
          int offset = (int) (GLintptr) exec->array.inputs[i]->Ptr;
          float *f = (float *) (p + offset);
          int *k = (int *) f;
@@ -364,7 +378,7 @@ print_draw_arrays(struct gl_context *ctx,
          for (i = 0; i < n; i++) {
             printf("    float[%d] = 0x%08x %f\n", i, k[i], f[i]);
          }
-         ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER_ARB, bufObj);
+         ctx->Driver.UnmapBuffer(ctx, bufObj);
       }
    }
 }
@@ -715,10 +729,11 @@ vbo_exec_DrawArraysInstanced(GLenum mode, GLint start, GLsizei count,
 static void
 dump_element_buffer(struct gl_context *ctx, GLenum type)
 {
-   const GLvoid *map = ctx->Driver.MapBuffer(ctx,
-                                             GL_ELEMENT_ARRAY_BUFFER_ARB,
-                                             GL_READ_ONLY,
-                                             ctx->Array.ElementArrayBufferObj);
+   const GLvoid *map =
+      ctx->Driver.MapBufferRange(ctx, 0,
+				 ctx->Array.ElementArrayBufferObj->Size,
+				 GL_MAP_READ_BIT,
+				 ctx->Array.ElementArrayBufferObj);
    switch (type) {
    case GL_UNSIGNED_BYTE:
       {
@@ -760,8 +775,7 @@ dump_element_buffer(struct gl_context *ctx, GLenum type)
       ;
    }
 
-   ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER_ARB,
-                           ctx->Array.ElementArrayBufferObj);
+   ctx->Driver.UnmapBuffer(ctx, ctx->Array.ElementArrayBufferObj);
 }
 
 
diff --git a/mesalib/src/mesa/vbo/vbo_exec_draw.c b/mesalib/src/mesa/vbo/vbo_exec_draw.c
index 98f585792..8ffaaaa48 100644
--- a/mesalib/src/mesa/vbo/vbo_exec_draw.c
+++ b/mesalib/src/mesa/vbo/vbo_exec_draw.c
@@ -1,427 +1,416 @@
-/*
- * Mesa 3-D graphics library
- * Version:  7.2
- *
- * Copyright (C) 1999-2008  Brian Paul   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
- * BRIAN PAUL 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.
- *
- * Authors:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "main/glheader.h"
-#include "main/bufferobj.h"
-#include "main/compiler.h"
-#include "main/enums.h"
-#include "main/mfeatures.h"
-#include "main/state.h"
-
-#include "vbo_context.h"
-
-
-#if FEATURE_beginend
-
-
-static void
-vbo_exec_debug_verts( struct vbo_exec_context *exec )
-{
-   GLuint count = exec->vtx.vert_count;
-   GLuint i;
-
-   printf("%s: %u vertices %d primitives, %d vertsize\n",
-	  __FUNCTION__,
-	  count,
-	  exec->vtx.prim_count,
-	  exec->vtx.vertex_size);
-
-   for (i = 0 ; i < exec->vtx.prim_count ; i++) {
-      struct _mesa_prim *prim = &exec->vtx.prim[i];
-      printf("   prim %d: %s%s %d..%d %s %s\n",
-	     i, 
-	     _mesa_lookup_prim_by_nr(prim->mode),
-	     prim->weak ? " (weak)" : "",
-	     prim->start, 
-	     prim->start + prim->count,
-	     prim->begin ? "BEGIN" : "(wrap)",
-	     prim->end ? "END" : "(wrap)");
-   }
-}
-
-
-/*
- * NOTE: Need to have calculated primitives by this point -- do it on the fly.
- * NOTE: Old 'parity' issue is gone.
- */
-static GLuint
-vbo_copy_vertices( struct vbo_exec_context *exec )
-{
-   GLuint nr = exec->vtx.prim[exec->vtx.prim_count-1].count;
-   GLuint ovf, i;
-   GLuint sz = exec->vtx.vertex_size;
-   GLfloat *dst = exec->vtx.copied.buffer;
-   const GLfloat *src = (exec->vtx.buffer_map + 
-                         exec->vtx.prim[exec->vtx.prim_count-1].start * 
-                         exec->vtx.vertex_size);
-
-
-   switch (exec->ctx->Driver.CurrentExecPrimitive) {
-   case GL_POINTS:
-      return 0;
-   case GL_LINES:
-      ovf = nr&1;
-      for (i = 0 ; i < ovf ; i++)
-	 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case GL_TRIANGLES:
-      ovf = nr%3;
-      for (i = 0 ; i < ovf ; i++)
-	 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case GL_QUADS:
-      ovf = nr&3;
-      for (i = 0 ; i < ovf ; i++)
-	 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case GL_LINE_STRIP:
-      if (nr == 0) {
-	 return 0;
-      }
-      else {
-	 memcpy( dst, src+(nr-1)*sz, sz * sizeof(GLfloat) );
-	 return 1;
-      }
-   case GL_LINE_LOOP:
-   case GL_TRIANGLE_FAN:
-   case GL_POLYGON:
-      if (nr == 0) {
-	 return 0;
-      }
-      else if (nr == 1) {
-	 memcpy( dst, src+0, sz * sizeof(GLfloat) );
-	 return 1;
-      }
-      else {
-	 memcpy( dst, src+0, sz * sizeof(GLfloat) );
-	 memcpy( dst+sz, src+(nr-1)*sz, sz * sizeof(GLfloat) );
-	 return 2;
-      }
-   case GL_TRIANGLE_STRIP:
-      /* no parity issue, but need to make sure the tri is not drawn twice */
-      if (nr & 1) {
-	 exec->vtx.prim[exec->vtx.prim_count-1].count--;
-      }
-      /* fallthrough */
-   case GL_QUAD_STRIP:
-      switch (nr) {
-      case 0:
-         ovf = 0;
-         break;
-      case 1:
-         ovf = 1;
-         break;
-      default:
-         ovf = 2 + (nr & 1);
-         break;
-      }
-      for (i = 0 ; i < ovf ; i++)
-	 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case PRIM_OUTSIDE_BEGIN_END:
-      return 0;
-   default:
-      assert(0);
-      return 0;
-   }
-}
-
-
-
-/* TODO: populate these as the vertex is defined:
- */
-static void
-vbo_exec_bind_arrays( struct gl_context *ctx )
-{
-   struct vbo_context *vbo = vbo_context(ctx);
-   struct vbo_exec_context *exec = &vbo->exec;
-   struct gl_client_array *arrays = exec->vtx.arrays;
-   const GLuint count = exec->vtx.vert_count;
-   const GLuint *map;
-   GLuint attr;
-   GLbitfield varying_inputs = 0x0;
-
-   /* Install the default (ie Current) attributes first, then overlay
-    * all active ones.
-    */
-   switch (get_program_mode(exec->ctx)) {
-   case VP_NONE:
-      for (attr = 0; attr < 16; attr++) {
-         exec->vtx.inputs[attr] = &vbo->legacy_currval[attr];
-      }
-      for (attr = 0; attr < MAT_ATTRIB_MAX; attr++) {
-         ASSERT(attr + 16 < Elements(exec->vtx.inputs));
-         exec->vtx.inputs[attr + 16] = &vbo->mat_currval[attr];
-      }
-      map = vbo->map_vp_none;
-      break;
-   case VP_NV:
-   case VP_ARB:
-      /* The aliasing of attributes for NV vertex programs has already
-       * occurred.  NV vertex programs cannot access material values,
-       * nor attributes greater than VERT_ATTRIB_TEX7.  
-       */
-      for (attr = 0; attr < 16; attr++) {
-         exec->vtx.inputs[attr] = &vbo->legacy_currval[attr];
-         ASSERT(attr + 16 < Elements(exec->vtx.inputs));
-         exec->vtx.inputs[attr + 16] = &vbo->generic_currval[attr];
-      }
-      map = vbo->map_vp_arb;
-
-      /* check if VERT_ATTRIB_POS is not read but VERT_BIT_GENERIC0 is read.
-       * In that case we effectively need to route the data from
-       * glVertexAttrib(0, val) calls to feed into the GENERIC0 input.
-       */
-      if ((ctx->VertexProgram._Current->Base.InputsRead & VERT_BIT_POS) == 0 &&
-          (ctx->VertexProgram._Current->Base.InputsRead & VERT_BIT_GENERIC0)) {
-         exec->vtx.inputs[16] = exec->vtx.inputs[0];
-         exec->vtx.attrsz[16] = exec->vtx.attrsz[0];
-         exec->vtx.attrptr[16] = exec->vtx.attrptr[0];
-         exec->vtx.attrsz[0] = 0;
-      }
-      break;
-   default:
-      assert(0);
-   }
-
-   /* Make all active attributes (including edgeflag) available as
-    * arrays of floats.
-    */
-   for (attr = 0; attr < VERT_ATTRIB_MAX ; attr++) {
-      const GLuint src = map[attr];
-
-      if (exec->vtx.attrsz[src]) {
-	 GLsizeiptr offset = (GLbyte *)exec->vtx.attrptr[src] -
-	    (GLbyte *)exec->vtx.vertex;
-
-         /* override the default array set above */
-         ASSERT(attr < Elements(exec->vtx.inputs));
-         ASSERT(attr < Elements(exec->vtx.arrays)); /* arrays[] */
-         exec->vtx.inputs[attr] = &arrays[attr];
-
-         if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
-            /* a real buffer obj: Ptr is an offset, not a pointer*/
-            assert(exec->vtx.bufferobj->Pointer);  /* buf should be mapped */
-            assert(offset >= 0);
-            arrays[attr].Ptr = (GLubyte *)exec->vtx.bufferobj->Offset + offset;
-         }
-         else {
-            /* Ptr into ordinary app memory */
-            arrays[attr].Ptr = (GLubyte *)exec->vtx.buffer_map + offset;
-         }
-	 arrays[attr].Size = exec->vtx.attrsz[src];
-	 arrays[attr].StrideB = exec->vtx.vertex_size * sizeof(GLfloat);
-	 arrays[attr].Stride = exec->vtx.vertex_size * sizeof(GLfloat);
-	 arrays[attr].Type = GL_FLOAT;
-         arrays[attr].Format = GL_RGBA;
-	 arrays[attr].Enabled = 1;
-         arrays[attr]._ElementSize = arrays[attr].Size * sizeof(GLfloat);
-         _mesa_reference_buffer_object(ctx,
-                                       &arrays[attr].BufferObj,
-                                       exec->vtx.bufferobj);
-	 arrays[attr]._MaxElement = count; /* ??? */
-
-         varying_inputs |= 1 << attr;
-         ctx->NewState |= _NEW_ARRAY;
-      }
-   }
-
-   _mesa_set_varying_vp_inputs( ctx, varying_inputs );
-}
-
-
-/**
- * Unmap the VBO.  This is called before drawing.
- */
-static void
-vbo_exec_vtx_unmap( struct vbo_exec_context *exec )
-{
-   GLenum target = GL_ARRAY_BUFFER_ARB;
-
-   if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
-      struct gl_context *ctx = exec->ctx;
-      
-      if (ctx->Driver.FlushMappedBufferRange) {
-         GLintptr offset = exec->vtx.buffer_used - exec->vtx.bufferobj->Offset;
-         GLsizeiptr length = (exec->vtx.buffer_ptr - exec->vtx.buffer_map) * sizeof(float);
-
-         if (length)
-            ctx->Driver.FlushMappedBufferRange(ctx, target,
-                                               offset, length,
-                                               exec->vtx.bufferobj);
-      }
-
-      exec->vtx.buffer_used += (exec->vtx.buffer_ptr -
-                                exec->vtx.buffer_map) * sizeof(float);
-
-      assert(exec->vtx.buffer_used <= VBO_VERT_BUFFER_SIZE);
-      assert(exec->vtx.buffer_ptr != NULL);
-      
-      ctx->Driver.UnmapBuffer(ctx, target, exec->vtx.bufferobj);
-      exec->vtx.buffer_map = NULL;
-      exec->vtx.buffer_ptr = NULL;
-      exec->vtx.max_vert = 0;
-   }
-}
-
-
-/**
- * Map the vertex buffer to begin storing glVertex, glColor, etc data.
- */
-void
-vbo_exec_vtx_map( struct vbo_exec_context *exec )
-{
-   struct gl_context *ctx = exec->ctx;
-   const GLenum target = GL_ARRAY_BUFFER_ARB;
-   const GLenum access = GL_READ_WRITE_ARB; /* for MapBuffer */
-   const GLenum accessRange = GL_MAP_WRITE_BIT |  /* for MapBufferRange */
-                              GL_MAP_INVALIDATE_RANGE_BIT |
-                              GL_MAP_UNSYNCHRONIZED_BIT |
-                              GL_MAP_FLUSH_EXPLICIT_BIT |
-                              MESA_MAP_NOWAIT_BIT;
-   const GLenum usage = GL_STREAM_DRAW_ARB;
-   
-   if (!_mesa_is_bufferobj(exec->vtx.bufferobj))
-      return;
-
-   assert(!exec->vtx.buffer_map);
-   assert(!exec->vtx.buffer_ptr);
-
-   if (VBO_VERT_BUFFER_SIZE > exec->vtx.buffer_used + 1024 &&
-       ctx->Driver.MapBufferRange) {
-      /* The VBO exists and there's room for more */
-      exec->vtx.buffer_map = 
-         (GLfloat *)ctx->Driver.MapBufferRange(ctx, 
-                                               target, 
-                                               exec->vtx.buffer_used,
-                                               (VBO_VERT_BUFFER_SIZE - 
-                                                exec->vtx.buffer_used),
-                                               accessRange,
-                                               exec->vtx.bufferobj);
-      exec->vtx.buffer_ptr = exec->vtx.buffer_map;
-   }
-   
-   if (!exec->vtx.buffer_map) {
-      /* Need to allocate a new VBO */
-      exec->vtx.buffer_used = 0;
-
-      ctx->Driver.BufferData(ctx, target, 
-                             VBO_VERT_BUFFER_SIZE, 
-                             NULL, usage, exec->vtx.bufferobj);
-
-
-      if (ctx->Driver.MapBufferRange)
-         exec->vtx.buffer_map = 
-            (GLfloat *)ctx->Driver.MapBufferRange(ctx, target,
-                                                  0, VBO_VERT_BUFFER_SIZE,
-                                                  accessRange,
-                                                  exec->vtx.bufferobj);
-      if (!exec->vtx.buffer_map)
-         exec->vtx.buffer_map =
-            (GLfloat *)ctx->Driver.MapBuffer(ctx, target, access, exec->vtx.bufferobj);
-      assert(exec->vtx.buffer_map);
-      exec->vtx.buffer_ptr = exec->vtx.buffer_map;
-   }
-
-   if (0)
-      printf("map %d..\n", exec->vtx.buffer_used);
-}
-
-
-
-/**
- * Execute the buffer and save copied verts.
- * \param keep_unmapped  if true, leave the VBO unmapped when we're done.
- */
-void
-vbo_exec_vtx_flush(struct vbo_exec_context *exec, GLboolean keepUnmapped)
-{
-   if (0)
-      vbo_exec_debug_verts( exec );
-
-   if (exec->vtx.prim_count && 
-       exec->vtx.vert_count) {
-
-      exec->vtx.copied.nr = vbo_copy_vertices( exec ); 
-
-      if (exec->vtx.copied.nr != exec->vtx.vert_count) {
-	 struct gl_context *ctx = exec->ctx;
-	 
-	 /* Before the update_state() as this may raise _NEW_ARRAY
-          * from _mesa_set_varying_vp_inputs().
-	  */
-	 vbo_exec_bind_arrays( ctx );
-
-         if (ctx->NewState)
-            _mesa_update_state( ctx );
-
-         if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
-            vbo_exec_vtx_unmap( exec );
-         }
-
-         if (0)
-            printf("%s %d %d\n", __FUNCTION__, exec->vtx.prim_count,
-		   exec->vtx.vert_count);
-
-	 vbo_context(ctx)->draw_prims( ctx, 
-				       exec->vtx.inputs, 
-				       exec->vtx.prim, 
-				       exec->vtx.prim_count,
-				       NULL,
-				       GL_TRUE,
-				       0,
-				       exec->vtx.vert_count - 1);
-
-	 /* If using a real VBO, get new storage -- unless asked not to.
-          */
-         if (_mesa_is_bufferobj(exec->vtx.bufferobj) && !keepUnmapped) {
-            vbo_exec_vtx_map( exec );
-         }
-      }
-   }
-
-   /* May have to unmap explicitly if we didn't draw:
-    */
-   if (keepUnmapped &&
-       _mesa_is_bufferobj(exec->vtx.bufferobj) &&
-       exec->vtx.buffer_map) {
-      vbo_exec_vtx_unmap( exec );
-   }
-
-   if (keepUnmapped || exec->vtx.vertex_size == 0)
-      exec->vtx.max_vert = 0;
-   else
-      exec->vtx.max_vert = ((VBO_VERT_BUFFER_SIZE - exec->vtx.buffer_used) / 
-                            (exec->vtx.vertex_size * sizeof(GLfloat)));
-
-   exec->vtx.buffer_ptr = exec->vtx.buffer_map;
-   exec->vtx.prim_count = 0;
-   exec->vtx.vert_count = 0;
-}
-
-
-#endif /* FEATURE_beginend */
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.2
+ *
+ * Copyright (C) 1999-2008  Brian Paul   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
+ * BRIAN PAUL 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.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "main/glheader.h"
+#include "main/bufferobj.h"
+#include "main/compiler.h"
+#include "main/enums.h"
+#include "main/mfeatures.h"
+#include "main/state.h"
+
+#include "vbo_context.h"
+
+
+#if FEATURE_beginend
+
+
+static void
+vbo_exec_debug_verts( struct vbo_exec_context *exec )
+{
+   GLuint count = exec->vtx.vert_count;
+   GLuint i;
+
+   printf("%s: %u vertices %d primitives, %d vertsize\n",
+	  __FUNCTION__,
+	  count,
+	  exec->vtx.prim_count,
+	  exec->vtx.vertex_size);
+
+   for (i = 0 ; i < exec->vtx.prim_count ; i++) {
+      struct _mesa_prim *prim = &exec->vtx.prim[i];
+      printf("   prim %d: %s%s %d..%d %s %s\n",
+	     i, 
+	     _mesa_lookup_prim_by_nr(prim->mode),
+	     prim->weak ? " (weak)" : "",
+	     prim->start, 
+	     prim->start + prim->count,
+	     prim->begin ? "BEGIN" : "(wrap)",
+	     prim->end ? "END" : "(wrap)");
+   }
+}
+
+
+/*
+ * NOTE: Need to have calculated primitives by this point -- do it on the fly.
+ * NOTE: Old 'parity' issue is gone.
+ */
+static GLuint
+vbo_copy_vertices( struct vbo_exec_context *exec )
+{
+   GLuint nr = exec->vtx.prim[exec->vtx.prim_count-1].count;
+   GLuint ovf, i;
+   GLuint sz = exec->vtx.vertex_size;
+   GLfloat *dst = exec->vtx.copied.buffer;
+   const GLfloat *src = (exec->vtx.buffer_map + 
+                         exec->vtx.prim[exec->vtx.prim_count-1].start * 
+                         exec->vtx.vertex_size);
+
+
+   switch (exec->ctx->Driver.CurrentExecPrimitive) {
+   case GL_POINTS:
+      return 0;
+   case GL_LINES:
+      ovf = nr&1;
+      for (i = 0 ; i < ovf ; i++)
+	 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
+      return i;
+   case GL_TRIANGLES:
+      ovf = nr%3;
+      for (i = 0 ; i < ovf ; i++)
+	 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
+      return i;
+   case GL_QUADS:
+      ovf = nr&3;
+      for (i = 0 ; i < ovf ; i++)
+	 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
+      return i;
+   case GL_LINE_STRIP:
+      if (nr == 0) {
+	 return 0;
+      }
+      else {
+	 memcpy( dst, src+(nr-1)*sz, sz * sizeof(GLfloat) );
+	 return 1;
+      }
+   case GL_LINE_LOOP:
+   case GL_TRIANGLE_FAN:
+   case GL_POLYGON:
+      if (nr == 0) {
+	 return 0;
+      }
+      else if (nr == 1) {
+	 memcpy( dst, src+0, sz * sizeof(GLfloat) );
+	 return 1;
+      }
+      else {
+	 memcpy( dst, src+0, sz * sizeof(GLfloat) );
+	 memcpy( dst+sz, src+(nr-1)*sz, sz * sizeof(GLfloat) );
+	 return 2;
+      }
+   case GL_TRIANGLE_STRIP:
+      /* no parity issue, but need to make sure the tri is not drawn twice */
+      if (nr & 1) {
+	 exec->vtx.prim[exec->vtx.prim_count-1].count--;
+      }
+      /* fallthrough */
+   case GL_QUAD_STRIP:
+      switch (nr) {
+      case 0:
+         ovf = 0;
+         break;
+      case 1:
+         ovf = 1;
+         break;
+      default:
+         ovf = 2 + (nr & 1);
+         break;
+      }
+      for (i = 0 ; i < ovf ; i++)
+	 memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
+      return i;
+   case PRIM_OUTSIDE_BEGIN_END:
+      return 0;
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+
+
+/* TODO: populate these as the vertex is defined:
+ */
+static void
+vbo_exec_bind_arrays( struct gl_context *ctx )
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_exec_context *exec = &vbo->exec;
+   struct gl_client_array *arrays = exec->vtx.arrays;
+   const GLuint count = exec->vtx.vert_count;
+   const GLuint *map;
+   GLuint attr;
+   GLbitfield varying_inputs = 0x0;
+
+   /* Install the default (ie Current) attributes first, then overlay
+    * all active ones.
+    */
+   switch (get_program_mode(exec->ctx)) {
+   case VP_NONE:
+      for (attr = 0; attr < 16; attr++) {
+         exec->vtx.inputs[attr] = &vbo->legacy_currval[attr];
+      }
+      for (attr = 0; attr < MAT_ATTRIB_MAX; attr++) {
+         ASSERT(attr + 16 < Elements(exec->vtx.inputs));
+         exec->vtx.inputs[attr + 16] = &vbo->mat_currval[attr];
+      }
+      map = vbo->map_vp_none;
+      break;
+   case VP_NV:
+   case VP_ARB:
+      /* The aliasing of attributes for NV vertex programs has already
+       * occurred.  NV vertex programs cannot access material values,
+       * nor attributes greater than VERT_ATTRIB_TEX7.  
+       */
+      for (attr = 0; attr < 16; attr++) {
+         exec->vtx.inputs[attr] = &vbo->legacy_currval[attr];
+         ASSERT(attr + 16 < Elements(exec->vtx.inputs));
+         exec->vtx.inputs[attr + 16] = &vbo->generic_currval[attr];
+      }
+      map = vbo->map_vp_arb;
+
+      /* check if VERT_ATTRIB_POS is not read but VERT_BIT_GENERIC0 is read.
+       * In that case we effectively need to route the data from
+       * glVertexAttrib(0, val) calls to feed into the GENERIC0 input.
+       */
+      if ((ctx->VertexProgram._Current->Base.InputsRead & VERT_BIT_POS) == 0 &&
+          (ctx->VertexProgram._Current->Base.InputsRead & VERT_BIT_GENERIC0)) {
+         exec->vtx.inputs[16] = exec->vtx.inputs[0];
+         exec->vtx.attrsz[16] = exec->vtx.attrsz[0];
+         exec->vtx.attrptr[16] = exec->vtx.attrptr[0];
+         exec->vtx.attrsz[0] = 0;
+      }
+      break;
+   default:
+      assert(0);
+   }
+
+   /* Make all active attributes (including edgeflag) available as
+    * arrays of floats.
+    */
+   for (attr = 0; attr < VERT_ATTRIB_MAX ; attr++) {
+      const GLuint src = map[attr];
+
+      if (exec->vtx.attrsz[src]) {
+	 GLsizeiptr offset = (GLbyte *)exec->vtx.attrptr[src] -
+	    (GLbyte *)exec->vtx.vertex;
+
+         /* override the default array set above */
+         ASSERT(attr < Elements(exec->vtx.inputs));
+         ASSERT(attr < Elements(exec->vtx.arrays)); /* arrays[] */
+         exec->vtx.inputs[attr] = &arrays[attr];
+
+         if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
+            /* a real buffer obj: Ptr is an offset, not a pointer*/
+            assert(exec->vtx.bufferobj->Pointer);  /* buf should be mapped */
+            assert(offset >= 0);
+            arrays[attr].Ptr = (GLubyte *)exec->vtx.bufferobj->Offset + offset;
+         }
+         else {
+            /* Ptr into ordinary app memory */
+            arrays[attr].Ptr = (GLubyte *)exec->vtx.buffer_map + offset;
+         }
+	 arrays[attr].Size = exec->vtx.attrsz[src];
+	 arrays[attr].StrideB = exec->vtx.vertex_size * sizeof(GLfloat);
+	 arrays[attr].Stride = exec->vtx.vertex_size * sizeof(GLfloat);
+	 arrays[attr].Type = GL_FLOAT;
+         arrays[attr].Format = GL_RGBA;
+	 arrays[attr].Enabled = 1;
+         arrays[attr]._ElementSize = arrays[attr].Size * sizeof(GLfloat);
+         _mesa_reference_buffer_object(ctx,
+                                       &arrays[attr].BufferObj,
+                                       exec->vtx.bufferobj);
+	 arrays[attr]._MaxElement = count; /* ??? */
+
+         varying_inputs |= 1 << attr;
+         ctx->NewState |= _NEW_ARRAY;
+      }
+   }
+
+   _mesa_set_varying_vp_inputs( ctx, varying_inputs );
+}
+
+
+/**
+ * Unmap the VBO.  This is called before drawing.
+ */
+static void
+vbo_exec_vtx_unmap( struct vbo_exec_context *exec )
+{
+   if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
+      struct gl_context *ctx = exec->ctx;
+      
+      if (ctx->Driver.FlushMappedBufferRange) {
+         GLintptr offset = exec->vtx.buffer_used - exec->vtx.bufferobj->Offset;
+         GLsizeiptr length = (exec->vtx.buffer_ptr - exec->vtx.buffer_map) * sizeof(float);
+
+         if (length)
+            ctx->Driver.FlushMappedBufferRange(ctx, offset, length,
+                                               exec->vtx.bufferobj);
+      }
+
+      exec->vtx.buffer_used += (exec->vtx.buffer_ptr -
+                                exec->vtx.buffer_map) * sizeof(float);
+
+      assert(exec->vtx.buffer_used <= VBO_VERT_BUFFER_SIZE);
+      assert(exec->vtx.buffer_ptr != NULL);
+      
+      ctx->Driver.UnmapBuffer(ctx, exec->vtx.bufferobj);
+      exec->vtx.buffer_map = NULL;
+      exec->vtx.buffer_ptr = NULL;
+      exec->vtx.max_vert = 0;
+   }
+}
+
+
+/**
+ * Map the vertex buffer to begin storing glVertex, glColor, etc data.
+ */
+void
+vbo_exec_vtx_map( struct vbo_exec_context *exec )
+{
+   struct gl_context *ctx = exec->ctx;
+   const GLenum accessRange = GL_MAP_WRITE_BIT |  /* for MapBufferRange */
+                              GL_MAP_INVALIDATE_RANGE_BIT |
+                              GL_MAP_UNSYNCHRONIZED_BIT |
+                              GL_MAP_FLUSH_EXPLICIT_BIT |
+                              MESA_MAP_NOWAIT_BIT;
+   const GLenum usage = GL_STREAM_DRAW_ARB;
+   
+   if (!_mesa_is_bufferobj(exec->vtx.bufferobj))
+      return;
+
+   assert(!exec->vtx.buffer_map);
+   assert(!exec->vtx.buffer_ptr);
+
+   if (VBO_VERT_BUFFER_SIZE > exec->vtx.buffer_used + 1024) {
+      /* The VBO exists and there's room for more */
+      exec->vtx.buffer_map = 
+         (GLfloat *)ctx->Driver.MapBufferRange(ctx, 
+                                               exec->vtx.buffer_used,
+                                               (VBO_VERT_BUFFER_SIZE - 
+                                                exec->vtx.buffer_used),
+                                               accessRange,
+                                               exec->vtx.bufferobj);
+      exec->vtx.buffer_ptr = exec->vtx.buffer_map;
+   }
+   
+   if (!exec->vtx.buffer_map) {
+      /* Need to allocate a new VBO */
+      exec->vtx.buffer_used = 0;
+
+      ctx->Driver.BufferData(ctx, GL_ARRAY_BUFFER_ARB,
+                             VBO_VERT_BUFFER_SIZE, 
+                             NULL, usage, exec->vtx.bufferobj);
+
+
+      exec->vtx.buffer_map =
+	 (GLfloat *)ctx->Driver.MapBufferRange(ctx,
+					       0, VBO_VERT_BUFFER_SIZE,
+					       accessRange,
+					       exec->vtx.bufferobj);
+      assert(exec->vtx.buffer_map);
+      exec->vtx.buffer_ptr = exec->vtx.buffer_map;
+   }
+
+   if (0)
+      printf("map %d..\n", exec->vtx.buffer_used);
+}
+
+
+
+/**
+ * Execute the buffer and save copied verts.
+ * \param keep_unmapped  if true, leave the VBO unmapped when we're done.
+ */
+void
+vbo_exec_vtx_flush(struct vbo_exec_context *exec, GLboolean keepUnmapped)
+{
+   if (0)
+      vbo_exec_debug_verts( exec );
+
+   if (exec->vtx.prim_count && 
+       exec->vtx.vert_count) {
+
+      exec->vtx.copied.nr = vbo_copy_vertices( exec ); 
+
+      if (exec->vtx.copied.nr != exec->vtx.vert_count) {
+	 struct gl_context *ctx = exec->ctx;
+	 
+	 /* Before the update_state() as this may raise _NEW_ARRAY
+          * from _mesa_set_varying_vp_inputs().
+	  */
+	 vbo_exec_bind_arrays( ctx );
+
+         if (ctx->NewState)
+            _mesa_update_state( ctx );
+
+         if (_mesa_is_bufferobj(exec->vtx.bufferobj)) {
+            vbo_exec_vtx_unmap( exec );
+         }
+
+         if (0)
+            printf("%s %d %d\n", __FUNCTION__, exec->vtx.prim_count,
+		   exec->vtx.vert_count);
+
+	 vbo_context(ctx)->draw_prims( ctx, 
+				       exec->vtx.inputs, 
+				       exec->vtx.prim, 
+				       exec->vtx.prim_count,
+				       NULL,
+				       GL_TRUE,
+				       0,
+				       exec->vtx.vert_count - 1);
+
+	 /* If using a real VBO, get new storage -- unless asked not to.
+          */
+         if (_mesa_is_bufferobj(exec->vtx.bufferobj) && !keepUnmapped) {
+            vbo_exec_vtx_map( exec );
+         }
+      }
+   }
+
+   /* May have to unmap explicitly if we didn't draw:
+    */
+   if (keepUnmapped &&
+       _mesa_is_bufferobj(exec->vtx.bufferobj) &&
+       exec->vtx.buffer_map) {
+      vbo_exec_vtx_unmap( exec );
+   }
+
+   if (keepUnmapped || exec->vtx.vertex_size == 0)
+      exec->vtx.max_vert = 0;
+   else
+      exec->vtx.max_vert = ((VBO_VERT_BUFFER_SIZE - exec->vtx.buffer_used) / 
+                            (exec->vtx.vertex_size * sizeof(GLfloat)));
+
+   exec->vtx.buffer_ptr = exec->vtx.buffer_map;
+   exec->vtx.prim_count = 0;
+   exec->vtx.vert_count = 0;
+}
+
+
+#endif /* FEATURE_beginend */
diff --git a/mesalib/src/mesa/vbo/vbo_rebase.c b/mesalib/src/mesa/vbo/vbo_rebase.c
index 1de290ff6..a1eab752a 100644
--- a/mesalib/src/mesa/vbo/vbo_rebase.c
+++ b/mesalib/src/mesa/vbo/vbo_rebase.c
@@ -159,10 +159,8 @@ void vbo_rebase_prims( struct gl_context *ctx,
       void *ptr;
 
       if (map_ib) 
-	 ctx->Driver.MapBuffer(ctx, 
-			       GL_ELEMENT_ARRAY_BUFFER,
-			       GL_READ_ONLY_ARB,
-			       ib->obj);
+	 ctx->Driver.MapBufferRange(ctx, 0, ib->obj->Size, GL_MAP_READ_BIT,
+				    ib->obj);
 
 
       ptr = ADD_POINTERS(ib->obj->Pointer, ib->ptr);
@@ -183,9 +181,7 @@ void vbo_rebase_prims( struct gl_context *ctx,
       }      
 
       if (map_ib) 
-	 ctx->Driver.UnmapBuffer(ctx, 
-				 GL_ELEMENT_ARRAY_BUFFER,
-				 ib->obj);
+	 ctx->Driver.UnmapBuffer(ctx, ib->obj);
 
       tmp_ib.obj = ctx->Shared->NullBufferObj;
       tmp_ib.ptr = tmp_indices;
diff --git a/mesalib/src/mesa/vbo/vbo_save_api.c b/mesalib/src/mesa/vbo/vbo_save_api.c
index 9041f791e..ad36e9332 100644
--- a/mesalib/src/mesa/vbo/vbo_save_api.c
+++ b/mesalib/src/mesa/vbo/vbo_save_api.c
@@ -232,11 +232,10 @@ map_vertex_store(struct gl_context *ctx,
    assert(vertex_store->bufferobj);
    assert(!vertex_store->buffer);
    vertex_store->buffer =
-      (GLfloat *) ctx->Driver.MapBuffer(ctx,
-                                        GL_ARRAY_BUFFER_ARB,   /* not used */
-                                        GL_WRITE_ONLY,      /* not used */
-                                        vertex_store->
-                                        bufferobj);
+      (GLfloat *) ctx->Driver.MapBufferRange(ctx, 0,
+					     vertex_store->bufferobj->Size,
+					     GL_MAP_WRITE_BIT,    /* not used */
+					     vertex_store->bufferobj);
 
    assert(vertex_store->buffer);
    return vertex_store->buffer + vertex_store->used;
@@ -247,7 +246,7 @@ static void
 unmap_vertex_store(struct gl_context *ctx,
                    struct vbo_save_vertex_store *vertex_store)
 {
-   ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER_ARB, vertex_store->bufferobj);
+   ctx->Driver.UnmapBuffer(ctx, vertex_store->bufferobj);
    vertex_store->buffer = NULL;
 }
 
diff --git a/mesalib/src/mesa/vbo/vbo_save_draw.c b/mesalib/src/mesa/vbo/vbo_save_draw.c
index a37af73e0..6cda831aa 100644
--- a/mesalib/src/mesa/vbo/vbo_save_draw.c
+++ b/mesalib/src/mesa/vbo/vbo_save_draw.c
@@ -70,7 +70,7 @@ _playback_copy_to_current(struct gl_context *ctx,
       else
          offset = node->buffer_offset;
 
-      ctx->Driver.GetBufferSubData( ctx, 0, offset, 
+      ctx->Driver.GetBufferSubData( ctx, offset,
                                     node->vertex_size * sizeof(GLfloat), 
                                     data, node->vertex_store->bufferobj );
 
@@ -217,10 +217,11 @@ static void
 vbo_save_loopback_vertex_list(struct gl_context *ctx,
                               const struct vbo_save_vertex_list *list)
 {
-   const char *buffer = ctx->Driver.MapBuffer(ctx, 
-					      GL_ARRAY_BUFFER_ARB, 
-					      GL_READ_ONLY, /* ? */
-                                              list->vertex_store->bufferobj);
+   const char *buffer =
+      ctx->Driver.MapBufferRange(ctx, 0,
+				 list->vertex_store->bufferobj->Size,
+				 GL_MAP_READ_BIT, /* ? */
+				 list->vertex_store->bufferobj);
 
    vbo_loopback_vertex_list(ctx,
                             (const GLfloat *)(buffer + list->buffer_offset),
@@ -230,8 +231,7 @@ vbo_save_loopback_vertex_list(struct gl_context *ctx,
                             list->wrap_count,
                             list->vertex_size);
 
-   ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER_ARB, 
-			   list->vertex_store->bufferobj);
+   ctx->Driver.UnmapBuffer(ctx, list->vertex_store->bufferobj);
 }
 
 
diff --git a/mesalib/src/mesa/vbo/vbo_split.c b/mesalib/src/mesa/vbo/vbo_split.c
index e28f9afcf..c933a3f9a 100644
--- a/mesalib/src/mesa/vbo/vbo_split.c
+++ b/mesalib/src/mesa/vbo/vbo_split.c
@@ -1,169 +1,168 @@
-
-/*
- * Mesa 3-D graphics library
- * Version:  6.5
- *
- * Copyright (C) 1999-2006  Brian Paul   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
- * BRIAN PAUL 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.
- *
- * Authors:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-/* Deal with hardware and/or swtnl maximums:
- * - maximum number of vertices in buffer
- * - maximum number of elements (maybe zero)
- *
- * The maximums may vary with opengl state (eg if a larger hardware
- * vertex is required in this state, the maximum number of vertices
- * may be smaller than in another state).
- *
- * We want buffer splitting to be a convenience function for the code
- * actually drawing the primitives rather than a system-wide maximum,
- * otherwise it is hard to avoid pessimism.  
- *
- * For instance, if a driver has no hardware limits on vertex buffer
- * dimensions, it would not ordinarily want to split vbos.  But if
- * there is an unexpected fallback, eg memory manager fails to upload
- * textures, it will want to pass the drawing commands onto swtnl,
- * which does have limitations.  A convenience function allows swtnl
- * to split the drawing and vbos internally without imposing its
- * limitations on drivers which want to use it as a fallback path.
- */
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/mtypes.h"
-#include "main/macros.h"
-
-#include "vbo_split.h"
-#include "vbo.h"
-
-/* True if a primitive can be split without copying of vertices, false
- * otherwise.
- */
-GLboolean split_prim_inplace(GLenum mode, GLuint *first, GLuint *incr)
-{
-   switch (mode) {
-   case GL_POINTS:
-      *first = 1;
-      *incr = 1;
-      return GL_TRUE;
-   case GL_LINES:
-      *first = 2;
-      *incr = 2;
-      return GL_TRUE;
-   case GL_LINE_STRIP:
-      *first = 2;
-      *incr = 1;
-      return GL_TRUE;
-   case GL_TRIANGLES:
-      *first = 3;
-      *incr = 3;
-      return GL_TRUE;
-   case GL_TRIANGLE_STRIP:
-      *first = 3;
-      *incr = 1;
-      return GL_TRUE;
-   case GL_QUADS:
-      *first = 4;
-      *incr = 4;
-      return GL_TRUE;
-   case GL_QUAD_STRIP:
-      *first = 4;
-      *incr = 2;
-      return GL_TRUE;
-   default:
-      *first = 0;
-      *incr = 1;		/* so that count % incr works */
-      return GL_FALSE;
-   }
-}
-
-
-
-void vbo_split_prims( struct gl_context *ctx,
-		      const struct gl_client_array *arrays[],
-		      const struct _mesa_prim *prim,
-		      GLuint nr_prims,
-		      const struct _mesa_index_buffer *ib,
-		      GLuint min_index,
-		      GLuint max_index,
-		      vbo_draw_func draw,
-		      const struct split_limits *limits )
-{
-   GLint max_basevertex = prim->basevertex;
-   GLuint i;
-
-   for (i = 1; i < nr_prims; i++)
-      max_basevertex = MAX2(max_basevertex, prim[i].basevertex);
-
-   /* XXX max_basevertex is computed but not used, why? */
-
-   if (ib) {
-      if (limits->max_indices == 0) {
-	 /* Could traverse the indices, re-emitting vertices in turn.
-	  * But it's hard to see why this case would be needed - for
-	  * software tnl, it is better to convert to non-indexed
-	  * rendering after transformation is complete, as is done in
-	  * the t_dd_rendertmp.h templates.  Are there any devices
-	  * with hardware tnl that cannot do indexed rendering?
-	  *
-	  * For now, this path is disabled.
-	  */
-	 assert(0);
-      }
-      else if (max_index - min_index >= limits->max_verts) {
-	 /* The vertex buffers are too large for hardware (or the
-	  * swtnl module).  Traverse the indices, re-emitting vertices
-	  * in turn.  Use a vertex cache to preserve some of the
-	  * sharing from the original index list.
-	  */
-	 vbo_split_copy(ctx, arrays, prim, nr_prims, ib,
-			draw, limits );
-      }
-      else if (ib->count > limits->max_indices) {
-	 /* The index buffer is too large for hardware.  Try to split
-	  * on whole-primitive boundaries, otherwise try to split the
-	  * individual primitives.
-	  */
-	 vbo_split_inplace(ctx, arrays, prim, nr_prims, ib,
-			   min_index, max_index, draw, limits );
-      }
-      else {
-	 /* Why were we called? */
-	 assert(0);
-      }
-   }
-   else {
-      if (max_index - min_index >= limits->max_verts) {
-	 /* The vertex buffer is too large for hardware (or the swtnl
-	  * module).  Try to split on whole-primitive boundaries,
-	  * otherwise try to split the individual primitives.
-	  */
-	 vbo_split_inplace(ctx, arrays, prim, nr_prims, ib,
-			   min_index, max_index, draw, limits );
-      }
-      else {
-	 /* Why were we called? */
-	 assert(0);
-      }
-   }
-}
-
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  Brian Paul   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
+ * BRIAN PAUL 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.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+/* Deal with hardware and/or swtnl maximums:
+ * - maximum number of vertices in buffer
+ * - maximum number of elements (maybe zero)
+ *
+ * The maximums may vary with opengl state (eg if a larger hardware
+ * vertex is required in this state, the maximum number of vertices
+ * may be smaller than in another state).
+ *
+ * We want buffer splitting to be a convenience function for the code
+ * actually drawing the primitives rather than a system-wide maximum,
+ * otherwise it is hard to avoid pessimism.  
+ *
+ * For instance, if a driver has no hardware limits on vertex buffer
+ * dimensions, it would not ordinarily want to split vbos.  But if
+ * there is an unexpected fallback, eg memory manager fails to upload
+ * textures, it will want to pass the drawing commands onto swtnl,
+ * which does have limitations.  A convenience function allows swtnl
+ * to split the drawing and vbos internally without imposing its
+ * limitations on drivers which want to use it as a fallback path.
+ */
+
+#include "main/glheader.h"
+#include "main/imports.h"
+#include "main/mtypes.h"
+#include "main/macros.h"
+
+#include "vbo_split.h"
+#include "vbo.h"
+
+/* True if a primitive can be split without copying of vertices, false
+ * otherwise.
+ */
+GLboolean split_prim_inplace(GLenum mode, GLuint *first, GLuint *incr)
+{
+   switch (mode) {
+   case GL_POINTS:
+      *first = 1;
+      *incr = 1;
+      return GL_TRUE;
+   case GL_LINES:
+      *first = 2;
+      *incr = 2;
+      return GL_TRUE;
+   case GL_LINE_STRIP:
+      *first = 2;
+      *incr = 1;
+      return GL_TRUE;
+   case GL_TRIANGLES:
+      *first = 3;
+      *incr = 3;
+      return GL_TRUE;
+   case GL_TRIANGLE_STRIP:
+      *first = 3;
+      *incr = 1;
+      return GL_TRUE;
+   case GL_QUADS:
+      *first = 4;
+      *incr = 4;
+      return GL_TRUE;
+   case GL_QUAD_STRIP:
+      *first = 4;
+      *incr = 2;
+      return GL_TRUE;
+   default:
+      *first = 0;
+      *incr = 1;		/* so that count % incr works */
+      return GL_FALSE;
+   }
+}
+
+
+
+void vbo_split_prims( struct gl_context *ctx,
+		      const struct gl_client_array *arrays[],
+		      const struct _mesa_prim *prim,
+		      GLuint nr_prims,
+		      const struct _mesa_index_buffer *ib,
+		      GLuint min_index,
+		      GLuint max_index,
+		      vbo_draw_func draw,
+		      const struct split_limits *limits )
+{
+   GLint max_basevertex = prim->basevertex;
+   GLuint i;
+
+   for (i = 1; i < nr_prims; i++)
+      max_basevertex = MAX2(max_basevertex, prim[i].basevertex);
+
+   /* XXX max_basevertex is computed but not used, why? */
+
+   if (ib) {
+      if (limits->max_indices == 0) {
+	 /* Could traverse the indices, re-emitting vertices in turn.
+	  * But it's hard to see why this case would be needed - for
+	  * software tnl, it is better to convert to non-indexed
+	  * rendering after transformation is complete.  Are there any devices
+	  * with hardware tnl that cannot do indexed rendering?
+	  *
+	  * For now, this path is disabled.
+	  */
+	 assert(0);
+      }
+      else if (max_index - min_index >= limits->max_verts) {
+	 /* The vertex buffers are too large for hardware (or the
+	  * swtnl module).  Traverse the indices, re-emitting vertices
+	  * in turn.  Use a vertex cache to preserve some of the
+	  * sharing from the original index list.
+	  */
+	 vbo_split_copy(ctx, arrays, prim, nr_prims, ib,
+			draw, limits );
+      }
+      else if (ib->count > limits->max_indices) {
+	 /* The index buffer is too large for hardware.  Try to split
+	  * on whole-primitive boundaries, otherwise try to split the
+	  * individual primitives.
+	  */
+	 vbo_split_inplace(ctx, arrays, prim, nr_prims, ib,
+			   min_index, max_index, draw, limits );
+      }
+      else {
+	 /* Why were we called? */
+	 assert(0);
+      }
+   }
+   else {
+      if (max_index - min_index >= limits->max_verts) {
+	 /* The vertex buffer is too large for hardware (or the swtnl
+	  * module).  Try to split on whole-primitive boundaries,
+	  * otherwise try to split the individual primitives.
+	  */
+	 vbo_split_inplace(ctx, arrays, prim, nr_prims, ib,
+			   min_index, max_index, draw, limits );
+      }
+      else {
+	 /* Why were we called? */
+	 assert(0);
+      }
+   }
+}
+
diff --git a/mesalib/src/mesa/vbo/vbo_split_copy.c b/mesalib/src/mesa/vbo/vbo_split_copy.c
index 3a869d712..40906e389 100644
--- a/mesalib/src/mesa/vbo/vbo_split_copy.c
+++ b/mesalib/src/mesa/vbo/vbo_split_copy.c
@@ -1,627 +1,627 @@
-
-/*
- * Mesa 3-D graphics library
- * Version:  6.5
- *
- * Copyright (C) 1999-2006  Brian Paul   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
- * BRIAN PAUL 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.
- *
- * Authors:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-/* Split indexed primitives with per-vertex copying.
- */
-
-#include "main/glheader.h"
-#include "main/bufferobj.h"
-#include "main/imports.h"
-#include "main/image.h"
-#include "main/macros.h"
-#include "main/mtypes.h"
-
-#include "vbo_split.h"
-#include "vbo.h"
-
-
-#define ELT_TABLE_SIZE 16
-
-/**
- * Used for vertex-level splitting of indexed buffers.  Note that
- * non-indexed primitives may be converted to indexed in some cases
- * (eg loops, fans) in order to use this splitting path.
- */
-struct copy_context {
-
-   struct gl_context *ctx;
-   const struct gl_client_array **array;
-   const struct _mesa_prim *prim;
-   GLuint nr_prims;
-   const struct _mesa_index_buffer *ib;
-   vbo_draw_func draw;
-
-   const struct split_limits *limits;
-
-   struct {
-      GLuint attr;
-      GLuint size;
-      const struct gl_client_array *array;
-      const GLubyte *src_ptr;
-
-      struct gl_client_array dstarray;
-
-   } varying[VERT_ATTRIB_MAX];
-   GLuint nr_varying;
-
-   const struct gl_client_array *dstarray_ptr[VERT_ATTRIB_MAX];
-   struct _mesa_index_buffer dstib;
-
-   GLuint *translated_elt_buf;
-   const GLuint *srcelt;
-
-   /** A baby hash table to avoid re-emitting (some) duplicate
-    * vertices when splitting indexed primitives.
-    */
-   struct { 
-      GLuint in;
-      GLuint out;
-   } vert_cache[ELT_TABLE_SIZE];
-
-   GLuint vertex_size;
-   GLubyte *dstbuf;
-   GLubyte *dstptr;     /**< dstptr == dstbuf + dstelt_max * vertsize */
-   GLuint dstbuf_size;  /**< in vertices */
-   GLuint dstbuf_nr;    /**< count of emitted vertices, also the largest value
-                         * in dstelt.  Our MaxIndex.
-                         */
-
-   GLuint *dstelt;
-   GLuint dstelt_nr;
-   GLuint dstelt_size;
-
-#define MAX_PRIM 32
-   struct _mesa_prim dstprim[MAX_PRIM];
-   GLuint dstprim_nr;
-
-};
-
-
-static GLuint attr_size( const struct gl_client_array *array )
-{
-   return array->Size * _mesa_sizeof_type(array->Type);
-}
-
-
-/**
- * Starts returning true slightly before the buffer fills, to ensure
- * that there is sufficient room for any remaining vertices to finish
- * off the prim:
- */
-static GLboolean
-check_flush( struct copy_context *copy )
-{
-   GLenum mode = copy->dstprim[copy->dstprim_nr].mode; 
-
-   if (GL_TRIANGLE_STRIP == mode &&
-       copy->dstelt_nr & 1) { /* see bug9962 */
-       return GL_FALSE;
-   }
-
-   if (copy->dstbuf_nr + 4 > copy->dstbuf_size)
-      return GL_TRUE;
-
-   if (copy->dstelt_nr + 4 > copy->dstelt_size)
-      return GL_TRUE;
-
-   return GL_FALSE;
-}
-
-
-/**
- * Dump the parameters/info for a vbo->draw() call.
- */
-static void
-dump_draw_info(struct gl_context *ctx,
-               const struct gl_client_array **arrays,
-               const struct _mesa_prim *prims,
-               GLuint nr_prims,
-               const struct _mesa_index_buffer *ib,
-               GLuint min_index,
-               GLuint max_index)
-{
-   GLuint i, j;
-
-   printf("VBO Draw:\n");
-   for (i = 0; i < nr_prims; i++) {
-      printf("Prim %u of %u\n", i, nr_prims);
-      printf("  Prim mode 0x%x\n", prims[i].mode);
-      printf("  IB: %p\n", (void*) ib);
-      for (j = 0; j < VERT_ATTRIB_MAX; j++) {
-         printf("    array %d at %p:\n", j, (void*) arrays[j]);
-         printf("      enabled %d, ptr %p, size %d, type 0x%x, stride %d\n",
-		arrays[j]->Enabled, arrays[j]->Ptr,
-		arrays[j]->Size, arrays[j]->Type, arrays[j]->StrideB);
-         if (0) {
-            GLint k = prims[i].start + prims[i].count - 1;
-            GLfloat *last = (GLfloat *) (arrays[j]->Ptr + arrays[j]->Stride * k);
-            printf("        last: %f %f %f\n",
-		   last[0], last[1], last[2]);
-         }
-      }
-   }
-}
-
-
-static void
-flush( struct copy_context *copy )
-{
-   GLuint i;
-
-   /* Set some counters: 
-    */
-   copy->dstib.count = copy->dstelt_nr;
-
-#if 0
-   dump_draw_info(copy->ctx,
-                  copy->dstarray_ptr,
-                  copy->dstprim,
-                  copy->dstprim_nr,
-                  &copy->dstib,
-                  0,
-                  copy->dstbuf_nr);
-#else
-   (void) dump_draw_info;
-#endif
-
-   copy->draw( copy->ctx,
-	       copy->dstarray_ptr,
-	       copy->dstprim,
-	       copy->dstprim_nr,
-	       &copy->dstib,
-	       GL_TRUE,
-	       0,
-	       copy->dstbuf_nr - 1 );
-
-   /* Reset all pointers: 
-    */
-   copy->dstprim_nr = 0;
-   copy->dstelt_nr = 0;
-   copy->dstbuf_nr = 0;
-   copy->dstptr = copy->dstbuf;
-
-   /* Clear the vertex cache:
-    */
-   for (i = 0; i < ELT_TABLE_SIZE; i++)
-      copy->vert_cache[i].in = ~0;
-}
-
-
-/**
- * Called at begin of each primitive during replay.
- */
-static void
-begin( struct copy_context *copy, GLenum mode, GLboolean begin_flag )
-{
-   struct _mesa_prim *prim = &copy->dstprim[copy->dstprim_nr];
-
-   prim->mode = mode;
-   prim->begin = begin_flag;
-   prim->num_instances = 1;
-}
-
-
-/**
- * Use a hashtable to attempt to identify recently-emitted vertices
- * and avoid re-emitting them.
- */
-static GLuint
-elt(struct copy_context *copy, GLuint elt_idx)
-{
-   GLuint elt = copy->srcelt[elt_idx];
-   GLuint slot = elt & (ELT_TABLE_SIZE-1);
-
-/*    printf("elt %d\n", elt); */
-
-   /* Look up the incoming element in the vertex cache.  Re-emit if
-    * necessary.   
-    */
-   if (copy->vert_cache[slot].in != elt) {
-      GLubyte *csr = copy->dstptr;
-      GLuint i;
-
-/*       printf("  --> emit to dstelt %d\n", copy->dstbuf_nr); */
-
-      for (i = 0; i < copy->nr_varying; i++) {
-	 const struct gl_client_array *srcarray = copy->varying[i].array;
-	 const GLubyte *srcptr = copy->varying[i].src_ptr + elt * srcarray->StrideB;
-
-	 memcpy(csr, srcptr, copy->varying[i].size);
-	 csr += copy->varying[i].size;
-
-#ifdef NAN_CHECK
-         if (srcarray->Type == GL_FLOAT) {
-            GLuint k;
-            GLfloat *f = (GLfloat *) srcptr;
-            for (k = 0; k < srcarray->Size; k++) {
-               assert(!IS_INF_OR_NAN(f[k]));
-               assert(f[k] <= 1.0e20 && f[k] >= -1.0e20);
-            }
-         }
-#endif
-
-	 if (0) 
-	 {
-	    const GLuint *f = (const GLuint *)srcptr;
-	    GLuint j;
-	    printf("  varying %d: ", i);
-	    for(j = 0; j < copy->varying[i].size / 4; j++)
-	       printf("%x ", f[j]);
-	    printf("\n");
-	 }
-      }
-
-      copy->vert_cache[slot].in = elt;
-      copy->vert_cache[slot].out = copy->dstbuf_nr++;
-      copy->dstptr += copy->vertex_size;
-
-      assert(csr == copy->dstptr);
-      assert(copy->dstptr == (copy->dstbuf +
-                              copy->dstbuf_nr * copy->vertex_size));
-   }
-/*    else */
-/*       printf("  --> reuse vertex\n"); */
-   
-/*    printf("  --> emit %d\n", copy->vert_cache[slot].out); */
-   copy->dstelt[copy->dstelt_nr++] = copy->vert_cache[slot].out;
-   return check_flush(copy);
-}
-
-
-/**
- * Called at end of each primitive during replay.
- */
-static void
-end( struct copy_context *copy, GLboolean end_flag )
-{
-   struct _mesa_prim *prim = &copy->dstprim[copy->dstprim_nr];
-
-/*    printf("end (%d)\n", end_flag); */
-
-   prim->end = end_flag;
-   prim->count = copy->dstelt_nr - prim->start;
-
-   if (++copy->dstprim_nr == MAX_PRIM ||
-       check_flush(copy)) 
-      flush(copy);
-}
-
-
-static void
-replay_elts( struct copy_context *copy )
-{
-   GLuint i, j, k;
-   GLboolean split;
-
-   for (i = 0; i < copy->nr_prims; i++) {
-      const struct _mesa_prim *prim = &copy->prim[i];
-      const GLuint start = prim->start;
-      GLuint first, incr;
-
-      switch (prim->mode) {
-	 
-      case GL_LINE_LOOP:
-	 /* Convert to linestrip and emit the final vertex explicitly,
-	  * but only in the resultant strip that requires it.
-	  */
-	 j = 0;
-	 while (j != prim->count) {
-	    begin(copy, GL_LINE_STRIP, prim->begin && j == 0);
-
-	    for (split = GL_FALSE; j != prim->count && !split; j++)
-	       split = elt(copy, start + j);
-
-	    if (j == prim->count) {
-	       /* Done, emit final line.  Split doesn't matter as
-		* it is always raised a bit early so we can emit
-		* the last verts if necessary!
-		*/
-	       if (prim->end) 
-		  (void)elt(copy, start + 0);
-
-	       end(copy, prim->end);
-	    }
-	    else {
-	       /* Wrap
-		*/
-	       assert(split);
-	       end(copy, 0);
-	       j--;
-	    }
-	 }
-	 break;
-
-      case GL_TRIANGLE_FAN:
-      case GL_POLYGON:
-	 j = 2;
-	 while (j != prim->count) {
-	    begin(copy, prim->mode, prim->begin && j == 0);
-
-	    split = elt(copy, start+0); 
-	    assert(!split);
-
-	    split = elt(copy, start+j-1); 
-	    assert(!split);
-
-	    for (; j != prim->count && !split; j++)
-	       split = elt(copy, start+j);
-
-	    end(copy, prim->end && j == prim->count);
-
-	    if (j != prim->count) {
-	       /* Wrapped the primitive, need to repeat some vertices:
-		*/
-	       j -= 1;
-	    }
-	 }
-	 break;
-
-      default:
-	 (void)split_prim_inplace(prim->mode, &first, &incr);
-	 
-	 j = 0;
-	 while (j != prim->count) {
-
-	    begin(copy, prim->mode, prim->begin && j == 0);
-
-	    split = 0;
-	    for (k = 0; k < first; k++, j++)
-	       split |= elt(copy, start+j);
-
-	    assert(!split);
-
-	    for (; j != prim->count && !split; )
-	       for (k = 0; k < incr; k++, j++)
-		  split |= elt(copy, start+j);
-
-	    end(copy, prim->end && j == prim->count);
-
-	    if (j != prim->count) {
-	       /* Wrapped the primitive, need to repeat some vertices:
-		*/
-	       assert(j > first - incr);
-	       j -= (first - incr);
-	    }
-	 }
-	 break;
-      }
-   }
-
-   if (copy->dstprim_nr)
-      flush(copy);
-}
-
-
-static void
-replay_init( struct copy_context *copy )
-{
-   struct gl_context *ctx = copy->ctx;
-   GLuint i;
-   GLuint offset;
-   const GLvoid *srcptr;
-
-   /* Make a list of varying attributes and their vbo's.  Also
-    * calculate vertex size.
-    */
-   copy->vertex_size = 0;
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
-      struct gl_buffer_object *vbo = copy->array[i]->BufferObj;
-
-      if (copy->array[i]->StrideB == 0) {
-	 copy->dstarray_ptr[i] = copy->array[i];
-      }
-      else {
-	 GLuint j = copy->nr_varying++;
-	 
-	 copy->varying[j].attr = i;
-	 copy->varying[j].array = copy->array[i];
-	 copy->varying[j].size = attr_size(copy->array[i]);
-	 copy->vertex_size += attr_size(copy->array[i]);
-      
-	 if (_mesa_is_bufferobj(vbo) && !_mesa_bufferobj_mapped(vbo)) 
-	    ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY, vbo);
-
-	 copy->varying[j].src_ptr = ADD_POINTERS(vbo->Pointer,
-						 copy->array[i]->Ptr);
-
-	 copy->dstarray_ptr[i] = &copy->varying[j].dstarray;
-      }
-   }
-
-   /* There must always be an index buffer.  Currently require the
-    * caller convert non-indexed prims to indexed.  Could alternately
-    * do it internally.
-    */
-   if (_mesa_is_bufferobj(copy->ib->obj) &&
-       !_mesa_bufferobj_mapped(copy->ib->obj)) 
-      ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, GL_READ_ONLY,
-			    copy->ib->obj);
-
-   srcptr = (const GLubyte *) ADD_POINTERS(copy->ib->obj->Pointer,
-                                           copy->ib->ptr);
-
-   switch (copy->ib->type) {
-   case GL_UNSIGNED_BYTE:
-      copy->translated_elt_buf = malloc(sizeof(GLuint) * copy->ib->count);
-      copy->srcelt = copy->translated_elt_buf;
-
-      for (i = 0; i < copy->ib->count; i++)
-	 copy->translated_elt_buf[i] = ((const GLubyte *)srcptr)[i];
-      break;
-
-   case GL_UNSIGNED_SHORT:
-      copy->translated_elt_buf = malloc(sizeof(GLuint) * copy->ib->count);
-      copy->srcelt = copy->translated_elt_buf;
-
-      for (i = 0; i < copy->ib->count; i++)
-	 copy->translated_elt_buf[i] = ((const GLushort *)srcptr)[i];
-      break;
-
-   case GL_UNSIGNED_INT:
-      copy->translated_elt_buf = NULL;
-      copy->srcelt = (const GLuint *)srcptr;
-      break;
-   }
-
-   /* Figure out the maximum allowed vertex buffer size:
-    */
-   if (copy->vertex_size * copy->limits->max_verts <= copy->limits->max_vb_size) {
-      copy->dstbuf_size = copy->limits->max_verts;
-   }
-   else {
-      copy->dstbuf_size = copy->limits->max_vb_size / copy->vertex_size;
-   }
-
-   /* Allocate an output vertex buffer:
-    *
-    * XXX:  This should be a VBO!
-    */
-   copy->dstbuf = malloc(copy->dstbuf_size * copy->vertex_size);   
-   copy->dstptr = copy->dstbuf;
-
-   /* Setup new vertex arrays to point into the output buffer: 
-    */
-   for (offset = 0, i = 0; i < copy->nr_varying; i++) {
-      const struct gl_client_array *src = copy->varying[i].array;
-      struct gl_client_array *dst = &copy->varying[i].dstarray;
-
-      dst->Size = src->Size;
-      dst->Type = src->Type;
-      dst->Format = GL_RGBA;
-      dst->Stride = copy->vertex_size;
-      dst->StrideB = copy->vertex_size;
-      dst->Ptr = copy->dstbuf + offset;
-      dst->Enabled = GL_TRUE;
-      dst->Normalized = src->Normalized; 
-      dst->BufferObj = ctx->Shared->NullBufferObj;
-      dst->_ElementSize = src->_ElementSize;
-      dst->_MaxElement = copy->dstbuf_size; /* may be less! */
-
-      offset += copy->varying[i].size;
-   }
-
-   /* Allocate an output element list:
-    */
-   copy->dstelt_size = MIN2(65536,
-			    copy->ib->count * 2 + 3);
-   copy->dstelt_size = MIN2(copy->dstelt_size,
-			    copy->limits->max_indices);
-   copy->dstelt = malloc(sizeof(GLuint) * copy->dstelt_size);
-   copy->dstelt_nr = 0;
-
-   /* Setup the new index buffer to point to the allocated element
-    * list:
-    */
-   copy->dstib.count = 0;	/* duplicates dstelt_nr */
-   copy->dstib.type = GL_UNSIGNED_INT;
-   copy->dstib.obj = ctx->Shared->NullBufferObj;
-   copy->dstib.ptr = copy->dstelt;
-}
-
-
-/**
- * Free up everything allocated during split/replay.
- */
-static void
-replay_finish( struct copy_context *copy )
-{
-   struct gl_context *ctx = copy->ctx;
-   GLuint i;
-
-   /* Free our vertex and index buffers: 
-    */
-   free(copy->translated_elt_buf);
-   free(copy->dstbuf);
-   free(copy->dstelt);
-
-   /* Unmap VBO's 
-    */
-   for (i = 0; i < copy->nr_varying; i++) {
-      struct gl_buffer_object *vbo = copy->varying[i].array->BufferObj;
-      if (_mesa_is_bufferobj(vbo) && _mesa_bufferobj_mapped(vbo)) 
-	 ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, vbo);
-   }
-
-   /* Unmap index buffer:
-    */
-   if (_mesa_is_bufferobj(copy->ib->obj) &&
-       _mesa_bufferobj_mapped(copy->ib->obj)) {
-      ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, copy->ib->obj);
-   }
-}
-
-
-/**
- * Split VBO into smaller pieces, draw the pieces.
- */
-void vbo_split_copy( struct gl_context *ctx,
-		     const struct gl_client_array *arrays[],
-		     const struct _mesa_prim *prim,
-		     GLuint nr_prims,
-		     const struct _mesa_index_buffer *ib,
-		     vbo_draw_func draw,
-		     const struct split_limits *limits )
-{
-   struct copy_context copy;
-   GLuint i, this_nr_prims;
-
-   for (i = 0; i < nr_prims;) {
-      /* Our SW TNL pipeline doesn't handle basevertex yet, so bind_indices
-       * will rebase the elements to the basevertex, and we'll only
-       * emit strings of prims with the same basevertex in one draw call.
-       */
-      for (this_nr_prims = 1; i + this_nr_prims < nr_prims;
-	   this_nr_prims++) {
-	 if (prim[i].basevertex != prim[i + this_nr_prims].basevertex)
-	    break;
-      }
-
-      memset(&copy, 0, sizeof(copy));
-
-      /* Require indexed primitives:
-       */
-      assert(ib);
-
-      copy.ctx = ctx;
-      copy.array = arrays;
-      copy.prim = &prim[i];
-      copy.nr_prims = this_nr_prims;
-      copy.ib = ib;
-      copy.draw = draw;
-      copy.limits = limits;
-
-      /* Clear the vertex cache:
-       */
-      for (i = 0; i < ELT_TABLE_SIZE; i++)
-	 copy.vert_cache[i].in = ~0;
-
-      replay_init(&copy);
-      replay_elts(&copy);
-      replay_finish(&copy);
-   }
-}
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  Brian Paul   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
+ * BRIAN PAUL 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.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+/* Split indexed primitives with per-vertex copying.
+ */
+
+#include "main/glheader.h"
+#include "main/bufferobj.h"
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/macros.h"
+#include "main/mtypes.h"
+
+#include "vbo_split.h"
+#include "vbo.h"
+
+
+#define ELT_TABLE_SIZE 16
+
+/**
+ * Used for vertex-level splitting of indexed buffers.  Note that
+ * non-indexed primitives may be converted to indexed in some cases
+ * (eg loops, fans) in order to use this splitting path.
+ */
+struct copy_context {
+
+   struct gl_context *ctx;
+   const struct gl_client_array **array;
+   const struct _mesa_prim *prim;
+   GLuint nr_prims;
+   const struct _mesa_index_buffer *ib;
+   vbo_draw_func draw;
+
+   const struct split_limits *limits;
+
+   struct {
+      GLuint attr;
+      GLuint size;
+      const struct gl_client_array *array;
+      const GLubyte *src_ptr;
+
+      struct gl_client_array dstarray;
+
+   } varying[VERT_ATTRIB_MAX];
+   GLuint nr_varying;
+
+   const struct gl_client_array *dstarray_ptr[VERT_ATTRIB_MAX];
+   struct _mesa_index_buffer dstib;
+
+   GLuint *translated_elt_buf;
+   const GLuint *srcelt;
+
+   /** A baby hash table to avoid re-emitting (some) duplicate
+    * vertices when splitting indexed primitives.
+    */
+   struct { 
+      GLuint in;
+      GLuint out;
+   } vert_cache[ELT_TABLE_SIZE];
+
+   GLuint vertex_size;
+   GLubyte *dstbuf;
+   GLubyte *dstptr;     /**< dstptr == dstbuf + dstelt_max * vertsize */
+   GLuint dstbuf_size;  /**< in vertices */
+   GLuint dstbuf_nr;    /**< count of emitted vertices, also the largest value
+                         * in dstelt.  Our MaxIndex.
+                         */
+
+   GLuint *dstelt;
+   GLuint dstelt_nr;
+   GLuint dstelt_size;
+
+#define MAX_PRIM 32
+   struct _mesa_prim dstprim[MAX_PRIM];
+   GLuint dstprim_nr;
+
+};
+
+
+static GLuint attr_size( const struct gl_client_array *array )
+{
+   return array->Size * _mesa_sizeof_type(array->Type);
+}
+
+
+/**
+ * Starts returning true slightly before the buffer fills, to ensure
+ * that there is sufficient room for any remaining vertices to finish
+ * off the prim:
+ */
+static GLboolean
+check_flush( struct copy_context *copy )
+{
+   GLenum mode = copy->dstprim[copy->dstprim_nr].mode; 
+
+   if (GL_TRIANGLE_STRIP == mode &&
+       copy->dstelt_nr & 1) { /* see bug9962 */
+       return GL_FALSE;
+   }
+
+   if (copy->dstbuf_nr + 4 > copy->dstbuf_size)
+      return GL_TRUE;
+
+   if (copy->dstelt_nr + 4 > copy->dstelt_size)
+      return GL_TRUE;
+
+   return GL_FALSE;
+}
+
+
+/**
+ * Dump the parameters/info for a vbo->draw() call.
+ */
+static void
+dump_draw_info(struct gl_context *ctx,
+               const struct gl_client_array **arrays,
+               const struct _mesa_prim *prims,
+               GLuint nr_prims,
+               const struct _mesa_index_buffer *ib,
+               GLuint min_index,
+               GLuint max_index)
+{
+   GLuint i, j;
+
+   printf("VBO Draw:\n");
+   for (i = 0; i < nr_prims; i++) {
+      printf("Prim %u of %u\n", i, nr_prims);
+      printf("  Prim mode 0x%x\n", prims[i].mode);
+      printf("  IB: %p\n", (void*) ib);
+      for (j = 0; j < VERT_ATTRIB_MAX; j++) {
+         printf("    array %d at %p:\n", j, (void*) arrays[j]);
+         printf("      enabled %d, ptr %p, size %d, type 0x%x, stride %d\n",
+		arrays[j]->Enabled, arrays[j]->Ptr,
+		arrays[j]->Size, arrays[j]->Type, arrays[j]->StrideB);
+         if (0) {
+            GLint k = prims[i].start + prims[i].count - 1;
+            GLfloat *last = (GLfloat *) (arrays[j]->Ptr + arrays[j]->Stride * k);
+            printf("        last: %f %f %f\n",
+		   last[0], last[1], last[2]);
+         }
+      }
+   }
+}
+
+
+static void
+flush( struct copy_context *copy )
+{
+   GLuint i;
+
+   /* Set some counters: 
+    */
+   copy->dstib.count = copy->dstelt_nr;
+
+#if 0
+   dump_draw_info(copy->ctx,
+                  copy->dstarray_ptr,
+                  copy->dstprim,
+                  copy->dstprim_nr,
+                  &copy->dstib,
+                  0,
+                  copy->dstbuf_nr);
+#else
+   (void) dump_draw_info;
+#endif
+
+   copy->draw( copy->ctx,
+	       copy->dstarray_ptr,
+	       copy->dstprim,
+	       copy->dstprim_nr,
+	       &copy->dstib,
+	       GL_TRUE,
+	       0,
+	       copy->dstbuf_nr - 1 );
+
+   /* Reset all pointers: 
+    */
+   copy->dstprim_nr = 0;
+   copy->dstelt_nr = 0;
+   copy->dstbuf_nr = 0;
+   copy->dstptr = copy->dstbuf;
+
+   /* Clear the vertex cache:
+    */
+   for (i = 0; i < ELT_TABLE_SIZE; i++)
+      copy->vert_cache[i].in = ~0;
+}
+
+
+/**
+ * Called at begin of each primitive during replay.
+ */
+static void
+begin( struct copy_context *copy, GLenum mode, GLboolean begin_flag )
+{
+   struct _mesa_prim *prim = &copy->dstprim[copy->dstprim_nr];
+
+   prim->mode = mode;
+   prim->begin = begin_flag;
+   prim->num_instances = 1;
+}
+
+
+/**
+ * Use a hashtable to attempt to identify recently-emitted vertices
+ * and avoid re-emitting them.
+ */
+static GLuint
+elt(struct copy_context *copy, GLuint elt_idx)
+{
+   GLuint elt = copy->srcelt[elt_idx];
+   GLuint slot = elt & (ELT_TABLE_SIZE-1);
+
+/*    printf("elt %d\n", elt); */
+
+   /* Look up the incoming element in the vertex cache.  Re-emit if
+    * necessary.   
+    */
+   if (copy->vert_cache[slot].in != elt) {
+      GLubyte *csr = copy->dstptr;
+      GLuint i;
+
+/*       printf("  --> emit to dstelt %d\n", copy->dstbuf_nr); */
+
+      for (i = 0; i < copy->nr_varying; i++) {
+	 const struct gl_client_array *srcarray = copy->varying[i].array;
+	 const GLubyte *srcptr = copy->varying[i].src_ptr + elt * srcarray->StrideB;
+
+	 memcpy(csr, srcptr, copy->varying[i].size);
+	 csr += copy->varying[i].size;
+
+#ifdef NAN_CHECK
+         if (srcarray->Type == GL_FLOAT) {
+            GLuint k;
+            GLfloat *f = (GLfloat *) srcptr;
+            for (k = 0; k < srcarray->Size; k++) {
+               assert(!IS_INF_OR_NAN(f[k]));
+               assert(f[k] <= 1.0e20 && f[k] >= -1.0e20);
+            }
+         }
+#endif
+
+	 if (0) 
+	 {
+	    const GLuint *f = (const GLuint *)srcptr;
+	    GLuint j;
+	    printf("  varying %d: ", i);
+	    for(j = 0; j < copy->varying[i].size / 4; j++)
+	       printf("%x ", f[j]);
+	    printf("\n");
+	 }
+      }
+
+      copy->vert_cache[slot].in = elt;
+      copy->vert_cache[slot].out = copy->dstbuf_nr++;
+      copy->dstptr += copy->vertex_size;
+
+      assert(csr == copy->dstptr);
+      assert(copy->dstptr == (copy->dstbuf +
+                              copy->dstbuf_nr * copy->vertex_size));
+   }
+/*    else */
+/*       printf("  --> reuse vertex\n"); */
+   
+/*    printf("  --> emit %d\n", copy->vert_cache[slot].out); */
+   copy->dstelt[copy->dstelt_nr++] = copy->vert_cache[slot].out;
+   return check_flush(copy);
+}
+
+
+/**
+ * Called at end of each primitive during replay.
+ */
+static void
+end( struct copy_context *copy, GLboolean end_flag )
+{
+   struct _mesa_prim *prim = &copy->dstprim[copy->dstprim_nr];
+
+/*    printf("end (%d)\n", end_flag); */
+
+   prim->end = end_flag;
+   prim->count = copy->dstelt_nr - prim->start;
+
+   if (++copy->dstprim_nr == MAX_PRIM ||
+       check_flush(copy)) 
+      flush(copy);
+}
+
+
+static void
+replay_elts( struct copy_context *copy )
+{
+   GLuint i, j, k;
+   GLboolean split;
+
+   for (i = 0; i < copy->nr_prims; i++) {
+      const struct _mesa_prim *prim = &copy->prim[i];
+      const GLuint start = prim->start;
+      GLuint first, incr;
+
+      switch (prim->mode) {
+	 
+      case GL_LINE_LOOP:
+	 /* Convert to linestrip and emit the final vertex explicitly,
+	  * but only in the resultant strip that requires it.
+	  */
+	 j = 0;
+	 while (j != prim->count) {
+	    begin(copy, GL_LINE_STRIP, prim->begin && j == 0);
+
+	    for (split = GL_FALSE; j != prim->count && !split; j++)
+	       split = elt(copy, start + j);
+
+	    if (j == prim->count) {
+	       /* Done, emit final line.  Split doesn't matter as
+		* it is always raised a bit early so we can emit
+		* the last verts if necessary!
+		*/
+	       if (prim->end) 
+		  (void)elt(copy, start + 0);
+
+	       end(copy, prim->end);
+	    }
+	    else {
+	       /* Wrap
+		*/
+	       assert(split);
+	       end(copy, 0);
+	       j--;
+	    }
+	 }
+	 break;
+
+      case GL_TRIANGLE_FAN:
+      case GL_POLYGON:
+	 j = 2;
+	 while (j != prim->count) {
+	    begin(copy, prim->mode, prim->begin && j == 0);
+
+	    split = elt(copy, start+0); 
+	    assert(!split);
+
+	    split = elt(copy, start+j-1); 
+	    assert(!split);
+
+	    for (; j != prim->count && !split; j++)
+	       split = elt(copy, start+j);
+
+	    end(copy, prim->end && j == prim->count);
+
+	    if (j != prim->count) {
+	       /* Wrapped the primitive, need to repeat some vertices:
+		*/
+	       j -= 1;
+	    }
+	 }
+	 break;
+
+      default:
+	 (void)split_prim_inplace(prim->mode, &first, &incr);
+	 
+	 j = 0;
+	 while (j != prim->count) {
+
+	    begin(copy, prim->mode, prim->begin && j == 0);
+
+	    split = 0;
+	    for (k = 0; k < first; k++, j++)
+	       split |= elt(copy, start+j);
+
+	    assert(!split);
+
+	    for (; j != prim->count && !split; )
+	       for (k = 0; k < incr; k++, j++)
+		  split |= elt(copy, start+j);
+
+	    end(copy, prim->end && j == prim->count);
+
+	    if (j != prim->count) {
+	       /* Wrapped the primitive, need to repeat some vertices:
+		*/
+	       assert(j > first - incr);
+	       j -= (first - incr);
+	    }
+	 }
+	 break;
+      }
+   }
+
+   if (copy->dstprim_nr)
+      flush(copy);
+}
+
+
+static void
+replay_init( struct copy_context *copy )
+{
+   struct gl_context *ctx = copy->ctx;
+   GLuint i;
+   GLuint offset;
+   const GLvoid *srcptr;
+
+   /* Make a list of varying attributes and their vbo's.  Also
+    * calculate vertex size.
+    */
+   copy->vertex_size = 0;
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      struct gl_buffer_object *vbo = copy->array[i]->BufferObj;
+
+      if (copy->array[i]->StrideB == 0) {
+	 copy->dstarray_ptr[i] = copy->array[i];
+      }
+      else {
+	 GLuint j = copy->nr_varying++;
+	 
+	 copy->varying[j].attr = i;
+	 copy->varying[j].array = copy->array[i];
+	 copy->varying[j].size = attr_size(copy->array[i]);
+	 copy->vertex_size += attr_size(copy->array[i]);
+      
+	 if (_mesa_is_bufferobj(vbo) && !_mesa_bufferobj_mapped(vbo)) 
+	    ctx->Driver.MapBufferRange(ctx, 0, vbo->Size, GL_MAP_READ_BIT, vbo);
+
+	 copy->varying[j].src_ptr = ADD_POINTERS(vbo->Pointer,
+						 copy->array[i]->Ptr);
+
+	 copy->dstarray_ptr[i] = &copy->varying[j].dstarray;
+      }
+   }
+
+   /* There must always be an index buffer.  Currently require the
+    * caller convert non-indexed prims to indexed.  Could alternately
+    * do it internally.
+    */
+   if (_mesa_is_bufferobj(copy->ib->obj) &&
+       !_mesa_bufferobj_mapped(copy->ib->obj)) 
+      ctx->Driver.MapBufferRange(ctx, 0, copy->ib->obj->Size, GL_MAP_READ_BIT,
+				 copy->ib->obj);
+
+   srcptr = (const GLubyte *) ADD_POINTERS(copy->ib->obj->Pointer,
+                                           copy->ib->ptr);
+
+   switch (copy->ib->type) {
+   case GL_UNSIGNED_BYTE:
+      copy->translated_elt_buf = malloc(sizeof(GLuint) * copy->ib->count);
+      copy->srcelt = copy->translated_elt_buf;
+
+      for (i = 0; i < copy->ib->count; i++)
+	 copy->translated_elt_buf[i] = ((const GLubyte *)srcptr)[i];
+      break;
+
+   case GL_UNSIGNED_SHORT:
+      copy->translated_elt_buf = malloc(sizeof(GLuint) * copy->ib->count);
+      copy->srcelt = copy->translated_elt_buf;
+
+      for (i = 0; i < copy->ib->count; i++)
+	 copy->translated_elt_buf[i] = ((const GLushort *)srcptr)[i];
+      break;
+
+   case GL_UNSIGNED_INT:
+      copy->translated_elt_buf = NULL;
+      copy->srcelt = (const GLuint *)srcptr;
+      break;
+   }
+
+   /* Figure out the maximum allowed vertex buffer size:
+    */
+   if (copy->vertex_size * copy->limits->max_verts <= copy->limits->max_vb_size) {
+      copy->dstbuf_size = copy->limits->max_verts;
+   }
+   else {
+      copy->dstbuf_size = copy->limits->max_vb_size / copy->vertex_size;
+   }
+
+   /* Allocate an output vertex buffer:
+    *
+    * XXX:  This should be a VBO!
+    */
+   copy->dstbuf = malloc(copy->dstbuf_size * copy->vertex_size);   
+   copy->dstptr = copy->dstbuf;
+
+   /* Setup new vertex arrays to point into the output buffer: 
+    */
+   for (offset = 0, i = 0; i < copy->nr_varying; i++) {
+      const struct gl_client_array *src = copy->varying[i].array;
+      struct gl_client_array *dst = &copy->varying[i].dstarray;
+
+      dst->Size = src->Size;
+      dst->Type = src->Type;
+      dst->Format = GL_RGBA;
+      dst->Stride = copy->vertex_size;
+      dst->StrideB = copy->vertex_size;
+      dst->Ptr = copy->dstbuf + offset;
+      dst->Enabled = GL_TRUE;
+      dst->Normalized = src->Normalized; 
+      dst->BufferObj = ctx->Shared->NullBufferObj;
+      dst->_ElementSize = src->_ElementSize;
+      dst->_MaxElement = copy->dstbuf_size; /* may be less! */
+
+      offset += copy->varying[i].size;
+   }
+
+   /* Allocate an output element list:
+    */
+   copy->dstelt_size = MIN2(65536,
+			    copy->ib->count * 2 + 3);
+   copy->dstelt_size = MIN2(copy->dstelt_size,
+			    copy->limits->max_indices);
+   copy->dstelt = malloc(sizeof(GLuint) * copy->dstelt_size);
+   copy->dstelt_nr = 0;
+
+   /* Setup the new index buffer to point to the allocated element
+    * list:
+    */
+   copy->dstib.count = 0;	/* duplicates dstelt_nr */
+   copy->dstib.type = GL_UNSIGNED_INT;
+   copy->dstib.obj = ctx->Shared->NullBufferObj;
+   copy->dstib.ptr = copy->dstelt;
+}
+
+
+/**
+ * Free up everything allocated during split/replay.
+ */
+static void
+replay_finish( struct copy_context *copy )
+{
+   struct gl_context *ctx = copy->ctx;
+   GLuint i;
+
+   /* Free our vertex and index buffers: 
+    */
+   free(copy->translated_elt_buf);
+   free(copy->dstbuf);
+   free(copy->dstelt);
+
+   /* Unmap VBO's 
+    */
+   for (i = 0; i < copy->nr_varying; i++) {
+      struct gl_buffer_object *vbo = copy->varying[i].array->BufferObj;
+      if (_mesa_is_bufferobj(vbo) && _mesa_bufferobj_mapped(vbo)) 
+	 ctx->Driver.UnmapBuffer(ctx, vbo);
+   }
+
+   /* Unmap index buffer:
+    */
+   if (_mesa_is_bufferobj(copy->ib->obj) &&
+       _mesa_bufferobj_mapped(copy->ib->obj)) {
+      ctx->Driver.UnmapBuffer(ctx, copy->ib->obj);
+   }
+}
+
+
+/**
+ * Split VBO into smaller pieces, draw the pieces.
+ */
+void vbo_split_copy( struct gl_context *ctx,
+		     const struct gl_client_array *arrays[],
+		     const struct _mesa_prim *prim,
+		     GLuint nr_prims,
+		     const struct _mesa_index_buffer *ib,
+		     vbo_draw_func draw,
+		     const struct split_limits *limits )
+{
+   struct copy_context copy;
+   GLuint i, this_nr_prims;
+
+   for (i = 0; i < nr_prims;) {
+      /* Our SW TNL pipeline doesn't handle basevertex yet, so bind_indices
+       * will rebase the elements to the basevertex, and we'll only
+       * emit strings of prims with the same basevertex in one draw call.
+       */
+      for (this_nr_prims = 1; i + this_nr_prims < nr_prims;
+	   this_nr_prims++) {
+	 if (prim[i].basevertex != prim[i + this_nr_prims].basevertex)
+	    break;
+      }
+
+      memset(&copy, 0, sizeof(copy));
+
+      /* Require indexed primitives:
+       */
+      assert(ib);
+
+      copy.ctx = ctx;
+      copy.array = arrays;
+      copy.prim = &prim[i];
+      copy.nr_prims = this_nr_prims;
+      copy.ib = ib;
+      copy.draw = draw;
+      copy.limits = limits;
+
+      /* Clear the vertex cache:
+       */
+      for (i = 0; i < ELT_TABLE_SIZE; i++)
+	 copy.vert_cache[i].in = ~0;
+
+      replay_init(&copy);
+      replay_elts(&copy);
+      replay_finish(&copy);
+   }
+}