driver/gl/cogl-pipeline-opengl.c

/*
 * Cogl
 *
 * A Low Level GPU Graphics and Utilities API
 *
 * Copyright (C) 2008,2009,2010 Intel Corporation.
 *
 * Permission is hereby granted, free of charge, to any person
 * obtaining a copy of this software and associated documentation
 * files (the "Software"), to deal in the Software without
 * restriction, including without limitation the rights to use, copy,
 * modify, merge, publish, distribute, sublicense, and/or sell copies
 * of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice 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.
 *
 *
 *
 * Authors:
 *   Robert Bragg <robert@linux.intel.com>
 */

#include "cogl-config.h"

#include "cogl-debug.h"
#include "cogl-util-gl-private.h"
#include "cogl-pipeline-opengl-private.h"
#include "cogl-pipeline-private.h"
#include "cogl-context-private.h"
#include "cogl-texture-private.h"
#include "cogl-framebuffer-private.h"
#include "cogl-offscreen.h"
#include "cogl-texture-gl-private.h"

#include "cogl-pipeline-progend-glsl-private.h"

#include <test-fixtures/test-unit.h>

#include <glib.h>
#include <string.h>

/*
 * GL/GLES compatability defines for pipeline thingies:
 */

/* These aren't defined in the GLES headers */
#ifndef GL_POINT_SPRITE
#define GL_POINT_SPRITE 0x8861
#endif
#ifndef GL_COORD_REPLACE
#define GL_COORD_REPLACE 0x8862
#endif
#ifndef GL_CLAMP_TO_BORDER
#define GL_CLAMP_TO_BORDER 0x812d
#endif
#ifndef GL_PROGRAM_POINT_SIZE
#define GL_PROGRAM_POINT_SIZE 0x8642
#endif

static void
texture_unit_init (CoglContext *ctx,
                   CoglTextureUnit *unit,
                   int index_)
{
  unit->index = index_;
  unit->enabled_gl_target = 0;
  unit->gl_texture = 0;
  unit->gl_target = 0;
  unit->is_foreign = FALSE;
  unit->dirty_gl_texture = FALSE;
  unit->matrix_stack = cogl_matrix_stack_new (ctx);

  unit->layer = NULL;
  unit->layer_changes_since_flush = 0;
  unit->texture_storage_changed = FALSE;
}

static void
texture_unit_free (CoglTextureUnit *unit)
{
  if (unit->layer)
    cogl_object_unref (unit->layer);
  cogl_object_unref (unit->matrix_stack);
}

CoglTextureUnit *
_cogl_get_texture_unit (int index_)
{
  _COGL_GET_CONTEXT (ctx, NULL);

  if (ctx->texture_units->len < (index_ + 1))
    {
      int i;
      int prev_len = ctx->texture_units->len;
      ctx->texture_units = g_array_set_size (ctx->texture_units, index_ + 1);
      for (i = prev_len; i <= index_; i++)
        {
          CoglTextureUnit *unit =
            &g_array_index (ctx->texture_units, CoglTextureUnit, i);

          texture_unit_init (ctx, unit, i);
        }
    }

  return &g_array_index (ctx->texture_units, CoglTextureUnit, index_);
}

void
_cogl_destroy_texture_units (void)
{
  int i;

  _COGL_GET_CONTEXT (ctx, NO_RETVAL);

  for (i = 0; i < ctx->texture_units->len; i++)
    {
      CoglTextureUnit *unit =
        &g_array_index (ctx->texture_units, CoglTextureUnit, i);
      texture_unit_free (unit);
    }
  g_array_free (ctx->texture_units, TRUE);
}

void
_cogl_set_active_texture_unit (int unit_index)
{
  _COGL_GET_CONTEXT (ctx, NO_RETVAL);

  if (ctx->active_texture_unit != unit_index)
    {
      GE (ctx, glActiveTexture (GL_TEXTURE0 + unit_index));
      ctx->active_texture_unit = unit_index;
    }
}

/* Note: _cogl_bind_gl_texture_transient conceptually has slightly
 * different semantics to OpenGL's glBindTexture because Cogl never
 * cares about tracking multiple textures bound to different targets
 * on the same texture unit.
 *
 * glBindTexture lets you bind multiple textures to a single texture
 * unit if they are bound to different targets. So it does something
 * like:
 *   unit->current_texture[target] = texture;
 *
 * Cogl only lets you associate one texture with the currently active
 * texture unit, so the target is basically a redundant parameter
 * that's implicitly set on that texture.
 *
 * Technically this is just a thin wrapper around glBindTexture so
 * actually it does have the GL semantics but it seems worth
 * mentioning the conceptual difference in case anyone wonders why we
 * don't associate the gl_texture with a gl_target in the
 * CoglTextureUnit.
 */
void
_cogl_bind_gl_texture_transient (GLenum gl_target,
                                 GLuint gl_texture,
                                 CoglBool is_foreign)
{
  CoglTextureUnit *unit;

  _COGL_GET_CONTEXT (ctx, NO_RETVAL);

  /* We choose to always make texture unit 1 active for transient
   * binds so that in the common case where multitexturing isn't used
   * we can simply ignore the state of this texture unit. Notably we
   * didn't use a large texture unit (.e.g. (GL_MAX_TEXTURE_UNITS - 1)
   * in case the driver doesn't have a sparse data structure for
   * texture units.
   */
  _cogl_set_active_texture_unit (1);
  unit = _cogl_get_texture_unit (1);

  /* NB: If we have previously bound a foreign texture to this texture
   * unit we don't know if that texture has since been deleted and we
   * are seeing the texture name recycled */
  if (unit->gl_texture == gl_texture &&
      !unit->dirty_gl_texture &&
      !unit->is_foreign)
    return;

  GE (ctx, glBindTexture (gl_target, gl_texture));

  unit->dirty_gl_texture = TRUE;
  unit->is_foreign = is_foreign;
}

void
_cogl_delete_gl_texture (GLuint gl_texture)
{
  int i;

  _COGL_GET_CONTEXT (ctx, NO_RETVAL);

  for (i = 0; i < ctx->texture_units->len; i++)
    {
      CoglTextureUnit *unit =
        &g_array_index (ctx->texture_units, CoglTextureUnit, i);

      if (unit->gl_texture == gl_texture)
        {
          unit->gl_texture = 0;
          unit->gl_target = 0;
          unit->dirty_gl_texture = FALSE;
        }
    }

  GE (ctx, glDeleteTextures (1, &gl_texture));
}

/* Whenever the underlying GL texture storage of a CoglTexture is
 * changed (e.g. due to migration out of a texture atlas) then we are
 * notified. This lets us ensure that we reflush that texture's state
 * if it is reused again with the same texture unit.
 */
void
_cogl_pipeline_texture_storage_change_notify (CoglTexture *texture)
{
  int i;

  _COGL_GET_CONTEXT (ctx, NO_RETVAL);

  for (i = 0; i < ctx->texture_units->len; i++)
    {
      CoglTextureUnit *unit =
        &g_array_index (ctx->texture_units, CoglTextureUnit, i);

      if (unit->layer &&
          _cogl_pipeline_layer_get_texture (unit->layer) == texture)
        unit->texture_storage_changed = TRUE;

      /* NB: the texture may be bound to multiple texture units so
       * we continue to check the rest */
    }
}

static void
set_glsl_program (GLuint gl_program)
{
  _COGL_GET_CONTEXT (ctx, NO_RETVAL);

  if (ctx->current_gl_program != gl_program)
    {
      _cogl_gl_util_clear_gl_errors (ctx);
      ctx->glUseProgram (gl_program);
      if (_cogl_gl_util_get_error (ctx) == GL_NO_ERROR)
        ctx->current_gl_program = gl_program;
      else
        {
          GE( ctx, glUseProgram (0) );
          ctx->current_gl_program = 0;
        }
    }
}

void
_cogl_use_fragment_program (GLuint gl_program, CoglPipelineProgramType type)
{
  _COGL_GET_CONTEXT (ctx, NO_RETVAL);

  /* If we're changing program type... */
  if (type != ctx->current_fragment_program_type)
    {
      /* ... disable the old type */
      switch (ctx->current_fragment_program_type)
        {
        case COGL_PIPELINE_PROGRAM_TYPE_GLSL:
          /* If the program contains a vertex shader then we shouldn't
             disable it */
          if (ctx->current_vertex_program_type !=
              COGL_PIPELINE_PROGRAM_TYPE_GLSL)
            set_glsl_program (0);
          break;

        case COGL_PIPELINE_PROGRAM_TYPE_ARBFP:
#ifdef HAVE_COGL_GL
          GE( ctx, glDisable (GL_FRAGMENT_PROGRAM_ARB) );
#endif
          break;

        case COGL_PIPELINE_PROGRAM_TYPE_FIXED:
          /* don't need to to anything */
          break;
        }

      /* ... and enable the new type */
      switch (type)
        {
        case COGL_PIPELINE_PROGRAM_TYPE_ARBFP:
#ifdef HAVE_COGL_GL
          GE( ctx, glEnable (GL_FRAGMENT_PROGRAM_ARB) );
#endif
          break;

        case COGL_PIPELINE_PROGRAM_TYPE_GLSL:
        case COGL_PIPELINE_PROGRAM_TYPE_FIXED:
          /* don't need to to anything */
          break;
        }
    }

  if (type == COGL_PIPELINE_PROGRAM_TYPE_GLSL)
    {
#ifdef COGL_PIPELINE_FRAGEND_GLSL
      set_glsl_program (gl_program);

#else

      g_warning ("Unexpected use of GLSL fragend!");

#endif /* COGL_PIPELINE_FRAGEND_GLSL */
    }
#ifndef COGL_PIPELINE_FRAGEND_ARBFP
  else if (type == COGL_PIPELINE_PROGRAM_TYPE_ARBFP)
    g_warning ("Unexpected use of ARBFP fragend!");
#endif /* COGL_PIPELINE_FRAGEND_ARBFP */

  ctx->current_fragment_program_type = type;
}

void
_cogl_use_vertex_program (GLuint gl_program, CoglPipelineProgramType type)
{
  _COGL_GET_CONTEXT (ctx, NO_RETVAL);

  /* If we're changing program type... */
  if (type != ctx->current_vertex_program_type)
    {
      /* ... disable the old type */
      switch (ctx->current_vertex_program_type)
        {
        case COGL_PIPELINE_PROGRAM_TYPE_GLSL:
          /* If the program contains a fragment shader then we shouldn't
             disable it */
          if (ctx->current_fragment_program_type !=
              COGL_PIPELINE_PROGRAM_TYPE_GLSL)
            set_glsl_program (0);
          break;

        case COGL_PIPELINE_PROGRAM_TYPE_ARBFP:
          /* It doesn't make sense to enable ARBfp for the vertex program */
          g_assert_not_reached ();
          break;

        case COGL_PIPELINE_PROGRAM_TYPE_FIXED:
          /* don't need to to anything */
          break;
        }

      /* ... and enable the new type */
      switch (type)
        {
        case COGL_PIPELINE_PROGRAM_TYPE_ARBFP:
          /* It doesn't make sense to enable ARBfp for the vertex program */
          g_assert_not_reached ();
          break;

        case COGL_PIPELINE_PROGRAM_TYPE_GLSL:
        case COGL_PIPELINE_PROGRAM_TYPE_FIXED:
          /* don't need to to anything */
          break;
        }
    }

  if (type == COGL_PIPELINE_PROGRAM_TYPE_GLSL)
    {
#ifdef COGL_PIPELINE_VERTEND_GLSL
      set_glsl_program (gl_program);

#else

      g_warning ("Unexpected use of GLSL vertend!");

#endif /* COGL_PIPELINE_VERTEND_GLSL */
    }
#ifndef COGL_PIPELINE_VERTEND_ARBFP
  else if (type == COGL_PIPELINE_PROGRAM_TYPE_ARBFP)
    g_warning ("Unexpected use of ARBFP vertend!");
#endif /* COGL_PIPELINE_VERTEND_ARBFP */

  ctx->current_vertex_program_type = type;
}

#if defined(HAVE_COGL_GLES2) || defined(HAVE_COGL_GL)

static CoglBool
blend_factor_uses_constant (GLenum blend_factor)
{
  return (blend_factor == GL_CONSTANT_COLOR ||
          blend_factor == GL_ONE_MINUS_CONSTANT_COLOR ||
          blend_factor == GL_CONSTANT_ALPHA ||
          blend_factor == GL_ONE_MINUS_CONSTANT_ALPHA);
}

#endif

static void
flush_depth_state (CoglContext *ctx,
                   CoglDepthState *depth_state)
{
  CoglBool depth_writing_enabled = depth_state->write_enabled;

  if (ctx->current_draw_buffer)
    depth_writing_enabled &= ctx->current_draw_buffer->depth_writing_enabled;

  if (ctx->depth_test_enabled_cache != depth_state->test_enabled)
    {
      if (depth_state->test_enabled == TRUE)
        GE (ctx, glEnable (GL_DEPTH_TEST));
      else
        GE (ctx, glDisable (GL_DEPTH_TEST));
      ctx->depth_test_enabled_cache = depth_state->test_enabled;
    }

  if (ctx->depth_test_function_cache != depth_state->test_function &&
      depth_state->test_enabled == TRUE)
    {
      GE (ctx, glDepthFunc (depth_state->test_function));
      ctx->depth_test_function_cache = depth_state->test_function;
    }

  if (ctx->depth_writing_enabled_cache != depth_writing_enabled)
    {
      GE (ctx, glDepthMask (depth_writing_enabled ?
                            GL_TRUE : GL_FALSE));
      ctx->depth_writing_enabled_cache = depth_writing_enabled;
    }

  if (ctx->driver != COGL_DRIVER_GLES1 &&
      (ctx->depth_range_near_cache != depth_state->range_near ||
       ctx->depth_range_far_cache != depth_state->range_far))
    {
      if (_cogl_has_private_feature (ctx, COGL_PRIVATE_FEATURE_GL_EMBEDDED))
        GE (ctx, glDepthRangef (depth_state->range_near,
                                depth_state->range_far));
      else
        GE (ctx, glDepthRange (depth_state->range_near,
                               depth_state->range_far));

      ctx->depth_range_near_cache = depth_state->range_near;
      ctx->depth_range_far_cache = depth_state->range_far;
    }
}

UNIT_TEST (check_gl_blend_enable,
           0 /* no requirements */,
           0 /* no failure cases */)
{
  CoglPipeline *pipeline = cogl_pipeline_new (test_ctx);

  /* By default blending should be disabled */
  g_assert_cmpint (test_ctx->gl_blend_enable_cache, ==, 0);

  cogl_framebuffer_draw_rectangle (test_fb, pipeline, 0, 0, 1, 1);
  _cogl_framebuffer_flush_journal (test_fb);

  /* After drawing an opaque rectangle blending should still be
   * disabled */
  g_assert_cmpint (test_ctx->gl_blend_enable_cache, ==, 0);

  cogl_pipeline_set_color4f (pipeline, 0, 0, 0, 0);
  cogl_framebuffer_draw_rectangle (test_fb, pipeline, 0, 0, 1, 1);
  _cogl_framebuffer_flush_journal (test_fb);

  /* After drawing a transparent rectangle blending should be enabled */
  g_assert_cmpint (test_ctx->gl_blend_enable_cache, ==, 1);

  cogl_pipeline_set_blend (pipeline, "RGBA=ADD(SRC_COLOR, 0)", NULL);
  cogl_framebuffer_draw_rectangle (test_fb, pipeline, 0, 0, 1, 1);
  _cogl_framebuffer_flush_journal (test_fb);

  /* After setting a blend string that effectively disables blending
   * then blending should be disabled */
  g_assert_cmpint (test_ctx->gl_blend_enable_cache, ==, 0);
}

static void
_cogl_pipeline_flush_color_blend_alpha_depth_state (
                                            CoglPipeline *pipeline,
                                            unsigned long pipelines_difference,
                                            CoglBool      with_color_attrib)
{
  _COGL_GET_CONTEXT (ctx, NO_RETVAL);

  /* On GLES2 we'll flush the color later */
  if (_cogl_has_private_feature (ctx, COGL_PRIVATE_FEATURE_GL_FIXED) &&
      !with_color_attrib)
    {
      if ((pipelines_difference & COGL_PIPELINE_STATE_COLOR) ||
          /* Assume if we were previously told to skip the color, then
           * the current color needs updating... */
          ctx->current_pipeline_with_color_attrib)
        {
          CoglPipeline *authority =
            _cogl_pipeline_get_authority (pipeline, COGL_PIPELINE_STATE_COLOR);
          GE (ctx, glColor4ub (cogl_color_get_red_byte (&authority->color),
                               cogl_color_get_green_byte (&authority->color),
                               cogl_color_get_blue_byte (&authority->color),
                               cogl_color_get_alpha_byte (&authority->color)));
        }
    }

  if (pipelines_difference & COGL_PIPELINE_STATE_BLEND)
    {
      CoglPipeline *authority =
        _cogl_pipeline_get_authority (pipeline, COGL_PIPELINE_STATE_BLEND);
      CoglPipelineBlendState *blend_state =
        &authority->big_state->blend_state;

      /* GLES 1 only has glBlendFunc */
      if (ctx->driver == COGL_DRIVER_GLES1)
        {
          GE (ctx, glBlendFunc (blend_state->blend_src_factor_rgb,
                                blend_state->blend_dst_factor_rgb));
        }
#if defined(HAVE_COGL_GLES2) || defined(HAVE_COGL_GL)
      else
        {
          if (blend_factor_uses_constant (blend_state->blend_src_factor_rgb) ||
              blend_factor_uses_constant (blend_state
                                          ->blend_src_factor_alpha) ||
              blend_factor_uses_constant (blend_state->blend_dst_factor_rgb) ||
              blend_factor_uses_constant (blend_state->blend_dst_factor_alpha))
            {
              float red =
                cogl_color_get_red_float (&blend_state->blend_constant);
              float green =
                cogl_color_get_green_float (&blend_state->blend_constant);
              float blue =
                cogl_color_get_blue_float (&blend_state->blend_constant);
              float alpha =
                cogl_color_get_alpha_float (&blend_state->blend_constant);


              GE (ctx, glBlendColor (red, green, blue, alpha));
            }

          if (ctx->glBlendEquationSeparate &&
              blend_state->blend_equation_rgb !=
              blend_state->blend_equation_alpha)
            GE (ctx,
                glBlendEquationSeparate (blend_state->blend_equation_rgb,
                                         blend_state->blend_equation_alpha));
          else
            GE (ctx, glBlendEquation (blend_state->blend_equation_rgb));

          if (ctx->glBlendFuncSeparate &&
              (blend_state->blend_src_factor_rgb !=
               blend_state->blend_src_factor_alpha ||
               (blend_state->blend_dst_factor_rgb !=
                blend_state->blend_dst_factor_alpha)))
            GE (ctx, glBlendFuncSeparate (blend_state->blend_src_factor_rgb,
                                          blend_state->blend_dst_factor_rgb,
                                          blend_state->blend_src_factor_alpha,
                                          blend_state->blend_dst_factor_alpha));
          else
            GE (ctx, glBlendFunc (blend_state->blend_src_factor_rgb,
                                  blend_state->blend_dst_factor_rgb));
        }
#endif
    }

#if defined (HAVE_COGL_GL) || defined (HAVE_COGL_GLES)

  if (_cogl_has_private_feature (ctx, COGL_PRIVATE_FEATURE_ALPHA_TEST))
    {
      /* Under GLES2 the alpha function is implemented as part of the
         fragment shader */
      if (pipelines_difference & (COGL_PIPELINE_STATE_ALPHA_FUNC |
                                  COGL_PIPELINE_STATE_ALPHA_FUNC_REFERENCE))
        {
          CoglPipeline *authority =
            _cogl_pipeline_get_authority (pipeline,
                                          COGL_PIPELINE_STATE_ALPHA_FUNC);
          CoglPipelineAlphaFuncState *alpha_state =
            &authority->big_state->alpha_state;

          /* NB: Currently the Cogl defines are compatible with the GL ones: */
          GE (ctx, glAlphaFunc (alpha_state->alpha_func,
                                alpha_state->alpha_func_reference));
        }

      /* Under GLES2 the lighting parameters are implemented as uniforms
         in the progend */
      if (pipelines_difference & COGL_PIPELINE_STATE_LIGHTING)
        {
          CoglPipeline *authority =
            _cogl_pipeline_get_authority (pipeline,
                                          COGL_PIPELINE_STATE_LIGHTING);
          CoglPipelineLightingState *lighting_state =
            &authority->big_state->lighting_state;

          GE (ctx, glMaterialfv (GL_FRONT_AND_BACK, GL_AMBIENT,
                                 lighting_state->ambient));
          GE (ctx, glMaterialfv (GL_FRONT_AND_BACK, GL_DIFFUSE,
                                 lighting_state->diffuse));
          GE (ctx, glMaterialfv (GL_FRONT_AND_BACK, GL_SPECULAR,
                                 lighting_state->specular));
          GE (ctx, glMaterialfv (GL_FRONT_AND_BACK, GL_EMISSION,
                                 lighting_state->emission));
          GE (ctx, glMaterialfv (GL_FRONT_AND_BACK, GL_SHININESS,
                                 &lighting_state->shininess));
        }
    }

#endif

  if (pipelines_difference & COGL_PIPELINE_STATE_DEPTH)
    {
      CoglPipeline *authority =
        _cogl_pipeline_get_authority (pipeline, COGL_PIPELINE_STATE_DEPTH);
      CoglDepthState *depth_state = &authority->big_state->depth_state;

      flush_depth_state (ctx, depth_state);
    }

  if (pipelines_difference & COGL_PIPELINE_STATE_LOGIC_OPS)
    {
      CoglPipeline *authority =
        _cogl_pipeline_get_authority (pipeline, COGL_PIPELINE_STATE_LOGIC_OPS);
      CoglPipelineLogicOpsState *logic_ops_state = &authority->big_state->logic_ops_state;
      CoglColorMask color_mask = logic_ops_state->color_mask;

      if (ctx->current_draw_buffer)
        color_mask &= ctx->current_draw_buffer->color_mask;

      GE (ctx, glColorMask (!!(color_mask & COGL_COLOR_MASK_RED),
                            !!(color_mask & COGL_COLOR_MASK_GREEN),
                            !!(color_mask & COGL_COLOR_MASK_BLUE),
                            !!(color_mask & COGL_COLOR_MASK_ALPHA)));
      ctx->current_gl_color_mask = color_mask;
    }

  if (pipelines_difference & COGL_PIPELINE_STATE_CULL_FACE)
    {
      CoglPipeline *authority =
        _cogl_pipeline_get_authority (pipeline, COGL_PIPELINE_STATE_CULL_FACE);
      CoglPipelineCullFaceState *cull_face_state
        = &authority->big_state->cull_face_state;

      if (cull_face_state->mode == COGL_PIPELINE_CULL_FACE_MODE_NONE)
        GE( ctx, glDisable (GL_CULL_FACE) );
      else
        {
          CoglBool invert_winding;

          GE( ctx, glEnable (GL_CULL_FACE) );

          switch (cull_face_state->mode)
            {
            case COGL_PIPELINE_CULL_FACE_MODE_NONE:
              g_assert_not_reached ();

            case COGL_PIPELINE_CULL_FACE_MODE_FRONT:
              GE( ctx, glCullFace (GL_FRONT) );
              break;

            case COGL_PIPELINE_CULL_FACE_MODE_BACK:
              GE( ctx, glCullFace (GL_BACK) );
              break;

            case COGL_PIPELINE_CULL_FACE_MODE_BOTH:
              GE( ctx, glCullFace (GL_FRONT_AND_BACK) );
              break;
            }

          /* If we are painting to an offscreen framebuffer then we
             need to invert the winding of the front face because
             everything is painted upside down */
          invert_winding = cogl_is_offscreen (ctx->current_draw_buffer);

          switch (cull_face_state->front_winding)
            {
            case COGL_WINDING_CLOCKWISE:
              GE( ctx, glFrontFace (invert_winding ? GL_CCW : GL_CW) );
              break;

            case COGL_WINDING_COUNTER_CLOCKWISE:
              GE( ctx, glFrontFace (invert_winding ? GL_CW : GL_CCW) );
              break;
            }
        }
    }

#ifdef HAVE_COGL_GL
  if (_cogl_has_private_feature
      (ctx, COGL_PRIVATE_FEATURE_ENABLE_PROGRAM_POINT_SIZE) &&
      (pipelines_difference & COGL_PIPELINE_STATE_PER_VERTEX_POINT_SIZE))
    {
      unsigned long state = COGL_PIPELINE_STATE_PER_VERTEX_POINT_SIZE;
      CoglPipeline *authority = _cogl_pipeline_get_authority (pipeline, state);

      if (authority->big_state->per_vertex_point_size)
        GE( ctx, glEnable (GL_PROGRAM_POINT_SIZE) );
      else
        GE( ctx, glDisable (GL_PROGRAM_POINT_SIZE) );
    }
#endif

  if (pipeline->real_blend_enable != ctx->gl_blend_enable_cache)
    {
      if (pipeline->real_blend_enable)
        GE (ctx, glEnable (GL_BLEND));
      else
        GE (ctx, glDisable (GL_BLEND));
      /* XXX: we shouldn't update any other blend state if blending
       * is disabled! */
      ctx->gl_blend_enable_cache = pipeline->real_blend_enable;
    }
}

static int
get_max_activateable_texture_units (void)
{
  _COGL_GET_CONTEXT (ctx, 0);

  if (G_UNLIKELY (ctx->max_activateable_texture_units == -1))
    {
      GLint values[3];
      int n_values = 0;
      int i;

#ifdef HAVE_COGL_GL
      if (!_cogl_has_private_feature (ctx, COGL_PRIVATE_FEATURE_GL_EMBEDDED))
        {
          /* GL_MAX_TEXTURE_COORDS is provided for both GLSL and ARBfp. It
             defines the number of texture coordinates that can be
             uploaded (but doesn't necessarily relate to how many texture
             images can be sampled) */
          if (cogl_has_feature (ctx, COGL_FEATURE_ID_GLSL) ||
              cogl_has_feature (ctx, COGL_FEATURE_ID_ARBFP))
            /* Previously this code subtracted the value by one but there
               was no explanation for why it did this and it doesn't seem
               to make sense so it has been removed */
            GE (ctx, glGetIntegerv (GL_MAX_TEXTURE_COORDS,
                                    values + n_values++));

          /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS is defined for GLSL but
             not ARBfp */
          if (cogl_has_feature (ctx, COGL_FEATURE_ID_GLSL))
            GE (ctx, glGetIntegerv (GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS,
                                    values + n_values++));
        }
#endif /* HAVE_COGL_GL */

#ifdef HAVE_COGL_GLES2
      if (_cogl_has_private_feature (ctx, COGL_PRIVATE_FEATURE_GL_EMBEDDED) &&
          _cogl_has_private_feature (ctx, COGL_PRIVATE_FEATURE_GL_PROGRAMMABLE))
        {
          GE (ctx, glGetIntegerv (GL_MAX_VERTEX_ATTRIBS, values + n_values));
          /* Two of the vertex attribs need to be used for the position
             and color */
          values[n_values++] -= 2;

          GE (ctx, glGetIntegerv (GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS,
                                  values + n_values++));
        }
#endif

#if defined (HAVE_COGL_GL) || defined (HAVE_COGL_GLES)
      if (_cogl_has_private_feature (ctx, COGL_PRIVATE_FEATURE_GL_FIXED))
        {
          /* GL_MAX_TEXTURE_UNITS defines the number of units that are
             usable from the fixed function pipeline, therefore it isn't
             available in GLES2. These are also tied to the number of
             texture coordinates that can be uploaded so it should be less
             than that available from the shader extensions */
          GE (ctx, glGetIntegerv (GL_MAX_TEXTURE_UNITS,
                                  values + n_values++));

        }
#endif

      g_assert (n_values <= G_N_ELEMENTS (values) &&
                n_values > 0);

      /* Use the maximum value */
      ctx->max_activateable_texture_units = values[0];
      for (i = 1; i < n_values; i++)
        ctx->max_activateable_texture_units =
          MAX (values[i], ctx->max_activateable_texture_units);
    }

  return ctx->max_activateable_texture_units;
}

typedef struct
{
  int i;
  unsigned long *layer_differences;
} CoglPipelineFlushLayerState;

static CoglBool
flush_layers_common_gl_state_cb (CoglPipelineLayer *layer, void *user_data)
{
  CoglPipelineFlushLayerState *flush_state = user_data;
  int                          unit_index = flush_state->i;
  CoglTextureUnit             *unit = _cogl_get_texture_unit (unit_index);
  unsigned long                layers_difference =
    flush_state->layer_differences[unit_index];

  _COGL_GET_CONTEXT (ctx, FALSE);

  /* There may not be enough texture units so we can bail out if
   * that's the case...
   */
  if (G_UNLIKELY (unit_index >= get_max_activateable_texture_units ()))
    {
      static CoglBool shown_warning = FALSE;

      if (!shown_warning)
        {
          g_warning ("Your hardware does not have enough texture units"
                     "to handle this many texture layers");
          shown_warning = TRUE;
        }
      return FALSE;
    }

  if (layers_difference & COGL_PIPELINE_LAYER_STATE_TEXTURE_DATA)
    {
      CoglTexture *texture = _cogl_pipeline_layer_get_texture_real (layer);
      GLuint gl_texture;
      GLenum gl_target;

      if (texture == NULL)
        switch (_cogl_pipeline_layer_get_texture_type (layer))
          {
          case COGL_TEXTURE_TYPE_2D:
            texture = COGL_TEXTURE (ctx->default_gl_texture_2d_tex);
            break;
          case COGL_TEXTURE_TYPE_3D:
            texture = COGL_TEXTURE (ctx->default_gl_texture_3d_tex);
            break;
          case COGL_TEXTURE_TYPE_RECTANGLE:
            texture = COGL_TEXTURE (ctx->default_gl_texture_rect_tex);
            break;
          }

      cogl_texture_get_gl_texture (texture,
                                   &gl_texture,
                                   &gl_target);

      _cogl_set_active_texture_unit (unit_index);

      /* NB: There are several Cogl components and some code in
       * Clutter that will temporarily bind arbitrary GL textures to
       * query and modify texture object parameters. If you look at
       * _cogl_bind_gl_texture_transient() you can see we make sure
       * that such code always binds to texture unit 1 which means we
       * can't rely on the unit->gl_texture state if unit->index == 1.
       *
       * Because texture unit 1 is a bit special we actually defer any
       * necessary glBindTexture for it until the end of
       * _cogl_pipeline_flush_gl_state().
       *
       * NB: we get notified whenever glDeleteTextures is used (see
       * _cogl_delete_gl_texture()) where we invalidate
       * unit->gl_texture references to deleted textures so it's safe
       * to compare unit->gl_texture with gl_texture.  (Without the
       * hook it would be possible to delete a GL texture and create a
       * new one with the same name and comparing unit->gl_texture and
       * gl_texture wouldn't detect that.)
       *
       * NB: for foreign textures we don't know how the deletion of
       * the GL texture objects correspond to the deletion of the
       * CoglTextures so if there was previously a foreign texture
       * associated with the texture unit then we can't assume that we
       * aren't seeing a recycled texture name so we have to bind.
       */
      if (unit->gl_texture != gl_texture || unit->is_foreign)
        {
          if (unit_index == 1)
            unit->dirty_gl_texture = TRUE;
          else
            GE (ctx, glBindTexture (gl_target, gl_texture));
          unit->gl_texture = gl_texture;
          unit->gl_target = gl_target;
        }

      unit->is_foreign = _cogl_texture_is_foreign (texture);

      /* The texture_storage_changed boolean indicates if the
       * CoglTexture's underlying GL texture storage has changed since
       * it was flushed to the texture unit. We've just flushed the
       * latest state so we can reset this. */
      unit->texture_storage_changed = FALSE;
    }

  if ((layers_difference & COGL_PIPELINE_LAYER_STATE_SAMPLER) &&
      _cogl_has_private_feature (ctx, COGL_PRIVATE_FEATURE_SAMPLER_OBJECTS))
    {
      const CoglSamplerCacheEntry *sampler_state;

      sampler_state = _cogl_pipeline_layer_get_sampler_state (layer);

      GE( ctx, glBindSampler (unit_index, sampler_state->sampler_object) );
    }

  /* FIXME: If using GLSL the progend we will use gl_PointCoord
   * instead of us needing to replace the texture coordinates but at
   * this point we can't currently tell if we are using the fixed or
   * glsl progend.
   */
#if defined (HAVE_COGL_GLES) || defined (HAVE_COGL_GL)
  if (_cogl_has_private_feature (ctx, COGL_PRIVATE_FEATURE_GL_FIXED) &&
      (layers_difference & COGL_PIPELINE_LAYER_STATE_POINT_SPRITE_COORDS))
    {
      CoglPipelineState change = COGL_PIPELINE_LAYER_STATE_POINT_SPRITE_COORDS;
      CoglPipelineLayer *authority =
        _cogl_pipeline_layer_get_authority (layer, change);
      CoglPipelineLayerBigState *big_state = authority->big_state;

      _cogl_set_active_texture_unit (unit_index);

      GE (ctx, glTexEnvi (GL_POINT_SPRITE, GL_COORD_REPLACE,
                          big_state->point_sprite_coords));
    }
#endif

  cogl_object_ref (layer);
  if (unit->layer != NULL)
    cogl_object_unref (unit->layer);

  unit->layer = layer;
  unit->layer_changes_since_flush = 0;

  flush_state->i++;

  return TRUE;
}

static void
_cogl_pipeline_flush_common_gl_state (CoglPipeline  *pipeline,
                                      unsigned long  pipelines_difference,
                                      unsigned long *layer_differences,
                                      CoglBool       with_color_attrib)
{
  CoglPipelineFlushLayerState state;

  _COGL_GET_CONTEXT (ctx, NO_RETVAL);

  _cogl_pipeline_flush_color_blend_alpha_depth_state (pipeline,
                                                      pipelines_difference,
                                                      with_color_attrib);

  state.i = 0;
  state.layer_differences = layer_differences;
  _cogl_pipeline_foreach_layer_internal (pipeline,
                                         flush_layers_common_gl_state_cb,
                                         &state);
}

/* Re-assert the layer's wrap modes on the given CoglTexture.
 *
 * Note: we don't simply forward the wrap modes to layer->texture
 * since the actual texture being used may have been overridden.
 */
static void
_cogl_pipeline_layer_forward_wrap_modes (CoglPipelineLayer *layer,
                                         CoglTexture *texture)
{
  CoglSamplerCacheWrapMode wrap_mode_s, wrap_mode_t, wrap_mode_p;
  GLenum gl_wrap_mode_s, gl_wrap_mode_t, gl_wrap_mode_p;

  if (texture == NULL)
    return;

  _cogl_pipeline_layer_get_wrap_modes (layer,
                                       &wrap_mode_s,
                                       &wrap_mode_t,
                                       &wrap_mode_p);

  /* Update the wrap mode on the texture object. The texture backend
     should cache the value so that it will be a no-op if the object
     already has the same wrap mode set. The backend is best placed to
     do this because it knows how many of the coordinates will
     actually be used (ie, a 1D texture only cares about the 's'
     coordinate but a 3D texture would use all three). GL uses the
     wrap mode as part of the texture object state but we are
     pretending it's part of the per-layer environment state. This
     will break if the application tries to use different modes in
     different layers using the same texture. */

  if (wrap_mode_s == COGL_SAMPLER_CACHE_WRAP_MODE_AUTOMATIC)
    gl_wrap_mode_s = GL_CLAMP_TO_EDGE;
  else
    gl_wrap_mode_s = wrap_mode_s;

  if (wrap_mode_t == COGL_SAMPLER_CACHE_WRAP_MODE_AUTOMATIC)
    gl_wrap_mode_t = GL_CLAMP_TO_EDGE;
  else
    gl_wrap_mode_t = wrap_mode_t;

  if (wrap_mode_p == COGL_SAMPLER_CACHE_WRAP_MODE_AUTOMATIC)
    gl_wrap_mode_p = GL_CLAMP_TO_EDGE;
  else
    gl_wrap_mode_p = wrap_mode_p;

  _cogl_texture_gl_flush_legacy_texobj_wrap_modes (texture,
                                                   gl_wrap_mode_s,
                                                   gl_wrap_mode_t,
                                                   gl_wrap_mode_p);
}

/* OpenGL associates the min/mag filters and repeat modes with the
 * texture object not the texture unit so we always have to re-assert
 * the filter and repeat modes whenever we use a texture since it may
 * be referenced by multiple pipelines with different modes.
 *
 * This function is bypassed in favour of sampler objects if
 * GL_ARB_sampler_objects is advertised. This fallback won't work if
 * the same texture is bound to multiple layers with different sampler
 * state.
 */
static void
foreach_texture_unit_update_filter_and_wrap_modes (void)
{
  int i;

  _COGL_GET_CONTEXT (ctx, NO_RETVAL);

  for (i = 0; i < ctx->texture_units->len; i++)
    {
      CoglTextureUnit *unit =
        &g_array_index (ctx->texture_units, CoglTextureUnit, i);

      if (unit->layer)
        {
          CoglTexture *texture = _cogl_pipeline_layer_get_texture (unit->layer);

          if (texture != NULL)
            {
              CoglPipelineFilter min;
              CoglPipelineFilter mag;

              _cogl_pipeline_layer_get_filters (unit->layer, &min, &mag);
              _cogl_texture_gl_flush_legacy_texobj_filters (texture, min, mag);

              _cogl_pipeline_layer_forward_wrap_modes (unit->layer, texture);
            }
        }
    }
}

typedef struct
{
  int i;
  unsigned long *layer_differences;
} CoglPipelineCompareLayersState;

static CoglBool
compare_layer_differences_cb (CoglPipelineLayer *layer, void *user_data)
{
  CoglPipelineCompareLayersState *state = user_data;
  CoglTextureUnit *unit = _cogl_get_texture_unit (state->i);

  if (unit->layer == layer)
    state->layer_differences[state->i] = unit->layer_changes_since_flush;
  else if (unit->layer)
    {
      state->layer_differences[state->i] = unit->layer_changes_since_flush;
      state->layer_differences[state->i] |=
        _cogl_pipeline_layer_compare_differences (layer, unit->layer);
    }
  else
    state->layer_differences[state->i] = COGL_PIPELINE_LAYER_STATE_ALL_SPARSE;

  /* XXX: There is always a possibility that a CoglTexture's
   * underlying GL texture storage has been changed since it was last
   * bound to a texture unit which is why we have a callback into
   * _cogl_pipeline_texture_storage_change_notify whenever a textures
   * underlying GL texture storage changes which will set the
   * unit->texture_intern_changed flag. If we see that's been set here
   * then we force an update of the texture state...
   */
  if (unit->texture_storage_changed)
    state->layer_differences[state->i] |=
      COGL_PIPELINE_LAYER_STATE_TEXTURE_DATA;

  state->i++;

  return TRUE;
}

typedef struct
{
  CoglFramebuffer *framebuffer;
  const CoglPipelineVertend *vertend;
  const CoglPipelineFragend *fragend;
  CoglPipeline *pipeline;
  unsigned long *layer_differences;
  CoglBool error_adding_layer;
  CoglBool added_layer;
} CoglPipelineAddLayerState;

static CoglBool
vertend_add_layer_cb (CoglPipelineLayer *layer,
                      void *user_data)
{
  CoglPipelineAddLayerState *state = user_data;
  const CoglPipelineVertend *vertend = state->vertend;
  CoglPipeline *pipeline = state->pipeline;
  int unit_index = _cogl_pipeline_layer_get_unit_index (layer);

  /* Either generate per layer code snippets or setup the
   * fixed function glTexEnv for each layer... */
  if (G_LIKELY (vertend->add_layer (pipeline,
                                    layer,
                                    state->layer_differences[unit_index],
                                    state->framebuffer)))
    state->added_layer = TRUE;
  else
    {
      state->error_adding_layer = TRUE;
      return FALSE;
    }

  return TRUE;
}

static CoglBool
fragend_add_layer_cb (CoglPipelineLayer *layer,
                      void *user_data)
{
  CoglPipelineAddLayerState *state = user_data;
  const CoglPipelineFragend *fragend = state->fragend;
  CoglPipeline *pipeline = state->pipeline;
  int unit_index = _cogl_pipeline_layer_get_unit_index (layer);

  /* Either generate per layer code snippets or setup the
   * fixed function glTexEnv for each layer... */
  if (G_LIKELY (fragend->add_layer (pipeline,
                                    layer,
                                    state->layer_differences[unit_index])))
    state->added_layer = TRUE;
  else
    {
      state->error_adding_layer = TRUE;
      return FALSE;
    }

  return TRUE;
}

/*
 * _cogl_pipeline_flush_gl_state:
 *
 * Details of override options:
 * ->fallback_mask: is a bitmask of the pipeline layers that need to be
 *    replaced with the default, fallback textures. The fallback textures are
 *    fully transparent textures so they hopefully wont contribute to the
 *    texture combining.
 *
 *    The intention of fallbacks is to try and preserve
 *    the number of layers the user is expecting so that texture coordinates
 *    they gave will mostly still correspond to the textures they intended, and
 *    have a fighting chance of looking close to their originally intended
 *    result.
 *
 * ->disable_mask: is a bitmask of the pipeline layers that will simply have
 *    texturing disabled. It's only really intended for disabling all layers
 *    > X; i.e. we'd expect to see a contiguous run of 0 starting from the LSB
 *    and at some point the remaining bits flip to 1. It might work to disable
 *    arbitrary layers; though I'm not sure a.t.m how OpenGL would take to
 *    that.
 *
 *    The intention of the disable_mask is for emitting geometry when the user
 *    hasn't supplied enough texture coordinates for all the layers and it's
 *    not possible to auto generate default texture coordinates for those
 *    layers.
 *
 * ->layer0_override_texture: forcibly tells us to bind this GL texture name for
 *    layer 0 instead of plucking the gl_texture from the CoglTexture of layer
 *    0.
 *
 *    The intention of this is for any primitives that supports sliced textures.
 *    The code will can iterate each of the slices and re-flush the pipeline
 *    forcing the GL texture of each slice in turn.
 *
 * ->wrap_mode_overrides: overrides the wrap modes set on each
 *    layer. This is used to implement the automatic wrap mode.
 *
 * XXX: It might also help if we could specify a texture matrix for code
 *    dealing with slicing that would be multiplied with the users own matrix.
 *
 *    Normaly texture coords in the range [0, 1] refer to the extents of the
 *    texture, but when your GL texture represents a slice of the real texture
 *    (from the users POV) then a texture matrix would be a neat way of
 *    transforming the mapping for each slice.
 *
 *    Currently for textured rectangles we manually calculate the texture
 *    coords for each slice based on the users given coords, but this solution
 *    isn't ideal, and can't be used with CoglVertexBuffers.
 */
void
_cogl_pipeline_flush_gl_state (CoglContext *ctx,
                               CoglPipeline *pipeline,
                               CoglFramebuffer *framebuffer,
                               CoglBool with_color_attrib,
                               CoglBool unknown_color_alpha)
{
  CoglPipeline *current_pipeline = ctx->current_pipeline;
  unsigned long pipelines_difference;
  int n_layers;
  unsigned long *layer_differences;
  int i;
  CoglTextureUnit *unit1;
  const CoglPipelineProgend *progend;

  COGL_STATIC_TIMER (pipeline_flush_timer,
                     "Mainloop", /* parent */
                     "Material Flush",
                     "The time spent flushing material state",
                     0 /* no application private data */);

  COGL_TIMER_START (_cogl_uprof_context, pipeline_flush_timer);

  /* Bail out asap if we've been asked to re-flush the already current
   * pipeline and we can see the pipeline hasn't changed */
  if (current_pipeline == pipeline &&
      ctx->current_pipeline_age == pipeline->age &&
      ctx->current_pipeline_with_color_attrib == with_color_attrib &&
      ctx->current_pipeline_unknown_color_alpha == unknown_color_alpha)
    goto done;
  else
    {
      /* Update derived state (currently just the 'real_blend_enable'
       * state) and determine a mask of state that differs between the
       * current pipeline and the one we are flushing.
       *
       * Note updating the derived state is done before doing any
       * pipeline comparisons so that we can correctly compare the
       * 'real_blend_enable' state itself.
       */

      if (current_pipeline == pipeline)
        {
          pipelines_difference = ctx->current_pipeline_changes_since_flush;

          if (pipelines_difference & COGL_PIPELINE_STATE_AFFECTS_BLENDING ||
              pipeline->unknown_color_alpha != unknown_color_alpha)
            {
              CoglBool save_real_blend_enable = pipeline->real_blend_enable;

              _cogl_pipeline_update_real_blend_enable (pipeline,
                                                       unknown_color_alpha);

              if (save_real_blend_enable != pipeline->real_blend_enable)
                pipelines_difference |= COGL_PIPELINE_STATE_REAL_BLEND_ENABLE;
            }
        }
      else if (current_pipeline)
        {
          pipelines_difference = ctx->current_pipeline_changes_since_flush;

          _cogl_pipeline_update_real_blend_enable (pipeline,
                                                   unknown_color_alpha);

          pipelines_difference |=
            _cogl_pipeline_compare_differences (ctx->current_pipeline,
                                                pipeline);
        }
      else
        {
          _cogl_pipeline_update_real_blend_enable (pipeline,
                                                   unknown_color_alpha);

          pipelines_difference = COGL_PIPELINE_STATE_ALL;
        }
    }

  /* Get a layer_differences mask for each layer to be flushed */
  n_layers = cogl_pipeline_get_n_layers (pipeline);
  if (n_layers)
    {
      CoglPipelineCompareLayersState state;
      layer_differences = g_alloca (sizeof (unsigned long) * n_layers);
      memset (layer_differences, 0, sizeof (unsigned long) * n_layers);
      state.i = 0;
      state.layer_differences = layer_differences;
      _cogl_pipeline_foreach_layer_internal (pipeline,
                                             compare_layer_differences_cb,
                                             &state);
    }
  else
    layer_differences = NULL;

  /* First flush everything that's the same regardless of which
   * pipeline backend is being used...
   *
   * 1) top level state:
   *  glColor (or skip if a vertex attribute is being used for color)
   *  blend state
   *  alpha test state (except for GLES 2.0)
   *
   * 2) then foreach layer:
   *  determine gl_target/gl_texture
   *  bind texture
   *
   *  Note: After _cogl_pipeline_flush_common_gl_state you can expect
   *  all state of the layers corresponding texture unit to be
   *  updated.
   */
  _cogl_pipeline_flush_common_gl_state (pipeline,
                                        pipelines_difference,
                                        layer_differences,
                                        with_color_attrib);

  /* Now flush the fragment, vertex and program state according to the
   * current progend backend.
   *
   * Note: Some backends may not support the current pipeline
   * configuration and in that case it will report and error and we
   * will look for a different backend.
   *
   * NB: if pipeline->progend != COGL_PIPELINE_PROGEND_UNDEFINED then
   * we have previously managed to successfully flush this pipeline
   * with the given progend so we will simply use that to avoid
   * fallback code paths.
   */
  if (pipeline->progend == COGL_PIPELINE_PROGEND_UNDEFINED)
    _cogl_pipeline_set_progend (pipeline, COGL_PIPELINE_PROGEND_DEFAULT);

  for (i = pipeline->progend;
       i < COGL_PIPELINE_N_PROGENDS;
       i++, _cogl_pipeline_set_progend (pipeline, i))
    {
      const CoglPipelineVertend *vertend;
      const CoglPipelineFragend *fragend;
      CoglPipelineAddLayerState state;

      progend = _cogl_pipeline_progends[i];

      if (G_UNLIKELY (!progend->start (pipeline)))
        continue;

      vertend = _cogl_pipeline_vertends[progend->vertend];

      vertend->start (pipeline,
                      n_layers,
                      pipelines_difference);

      state.framebuffer = framebuffer;
      state.vertend = vertend;
      state.pipeline = pipeline;
      state.layer_differences = layer_differences;
      state.error_adding_layer = FALSE;
      state.added_layer = FALSE;

      _cogl_pipeline_foreach_layer_internal (pipeline,
                                             vertend_add_layer_cb,
                                             &state);

      if (G_UNLIKELY (state.error_adding_layer))
        continue;

      if (G_UNLIKELY (!vertend->end (pipeline, pipelines_difference)))
        continue;

      /* Now prepare the fragment processing state (fragend)
       *
       * NB: We can't combine the setup of the vertend and fragend
       * since the backends that do code generation share
       * ctx->codegen_source_buffer as a scratch buffer.
       */

      fragend = _cogl_pipeline_fragends[progend->fragend];
      state.fragend = fragend;

      fragend->start (pipeline,
                      n_layers,
                      pipelines_difference);

      _cogl_pipeline_foreach_layer_internal (pipeline,
                                             fragend_add_layer_cb,
                                             &state);

      if (G_UNLIKELY (state.error_adding_layer))
        continue;

      if (!state.added_layer)
        {
          if (fragend->passthrough &&
              G_UNLIKELY (!fragend->passthrough (pipeline)))
            continue;
        }

      if (G_UNLIKELY (!fragend->end (pipeline, pipelines_difference)))
        continue;

      if (progend->end)
        progend->end (pipeline, pipelines_difference);
      break;
    }

  /* FIXME: This reference is actually resulting in lots of
   * copy-on-write reparenting because one-shot pipelines end up
   * living for longer than necessary and so any later modification of
   * the parent will cause a copy-on-write.
   *
   * XXX: The issue should largely go away when we switch to using
   * weak pipelines for overrides.
   */
  cogl_object_ref (pipeline);
  if (ctx->current_pipeline != NULL)
    cogl_object_unref (ctx->current_pipeline);
  ctx->current_pipeline = pipeline;
  ctx->current_pipeline_changes_since_flush = 0;
  ctx->current_pipeline_with_color_attrib = with_color_attrib;
  ctx->current_pipeline_unknown_color_alpha = unknown_color_alpha;
  ctx->current_pipeline_age = pipeline->age;

done:

  progend = _cogl_pipeline_progends[pipeline->progend];

  /* We can't assume the color will be retained between flushes when
   * using the glsl progend because the generic attribute values are
   * not stored as part of the program object so they could be
   * overridden by any attribute changes in another program */
  if (pipeline->progend == COGL_PIPELINE_PROGEND_GLSL && !with_color_attrib)
    {
      int attribute;
      CoglPipeline *authority =
        _cogl_pipeline_get_authority (pipeline, COGL_PIPELINE_STATE_COLOR);
      int name_index = COGL_ATTRIBUTE_COLOR_NAME_INDEX;

      attribute =
        _cogl_pipeline_progend_glsl_get_attrib_location (pipeline, name_index);
      if (attribute != -1)
        GE (ctx,
            glVertexAttrib4f (attribute,
                              cogl_color_get_red_float (&authority->color),
                              cogl_color_get_green_float (&authority->color),
                              cogl_color_get_blue_float (&authority->color),
                              cogl_color_get_alpha_float (&authority->color)));
    }

  /* Give the progend a chance to update any uniforms that might not
   * depend on the material state. This is used on GLES2 to update the
   * matrices */
  if (progend->pre_paint)
    progend->pre_paint (pipeline, framebuffer);

  /* Handle the fact that OpenGL associates texture filter and wrap
   * modes with the texture objects not the texture units... */
  if (!_cogl_has_private_feature (ctx, COGL_PRIVATE_FEATURE_SAMPLER_OBJECTS))
    foreach_texture_unit_update_filter_and_wrap_modes ();

  /* If this pipeline has more than one layer then we always need
   * to make sure we rebind the texture for unit 1.
   *
   * NB: various components of Cogl may temporarily bind arbitrary
   * textures to texture unit 1 so they can query and modify texture
   * object parameters. cogl-pipeline.c (See
   * _cogl_bind_gl_texture_transient)
   */
  unit1 = _cogl_get_texture_unit (1);
  if (cogl_pipeline_get_n_layers (pipeline) > 1 && unit1->dirty_gl_texture)
    {
      _cogl_set_active_texture_unit (1);
      GE (ctx, glBindTexture (unit1->gl_target, unit1->gl_texture));
      unit1->dirty_gl_texture = FALSE;
    }

  COGL_TIMER_STOP (_cogl_uprof_context, pipeline_flush_timer);
}