Always apply UniformMatrix4fvStreamTextureMatrixCHROMIUM matrix argument.

gpu/command_buffer/service/gles2_cmd_decoder.cc

 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "gpu/command_buffer/service/gles2_cmd_decoder.h"
 
 #include <limits.h>
 #include <stddef.h>
 #include <stdint.h>
 #include <stdio.h>
@@ skipping 7517 matching lines @@
                                    &type,
                                    &count)) {
     return;
   }
   glUniformMatrix4fv(real_location, count, transpose, value);
 }
 
 void GLES2DecoderImpl::DoUniformMatrix4fvStreamTextureMatrixCHROMIUM(
     GLint fake_location,
     GLboolean transpose,
-    const GLfloat* default_value) {
+    const GLfloat* post_transform) {
   float gl_matrix[16];
 
-  // If we can't get a matrix from the texture, then use a default.
-  // TODO(liberato): remove |default_value| and replace with an identity matrix.
-  // It is only present as a transitionary step until StreamTexture supplies
-  // the matrix via GLImage.  Once that happens, GLRenderer can quit sending
-  // in a default.
-  memcpy(gl_matrix, default_value, sizeof(gl_matrix));
-
   // This refers to the bound external texture on the active unit.
   TextureUnit& unit = state_.texture_units[state_.active_texture_unit];
   if (TextureRef* texture_ref = unit.bound_texture_external_oes.get()) {
     if (GLStreamTextureImage* image =
             texture_ref->texture()->GetLevelStreamTextureImage(
                 GL_TEXTURE_EXTERNAL_OES, 0)) {
       image->GetTextureMatrix(gl_matrix);
+      // Pre-multiply |gl_matrix| by |post_transform|.

[liberato (no reviews please), 2016/03/21 16:21:56]

are you sure that it should be premultiplied?  if |gl_matrix| is a translation,
and |post_transform|| is a scale, then we don't want to scale the translation,
do we?  that should be applied last.

i think that's |gl_matrix| * |post_transform| * uv_vector_in_fragment_shader,
assuming that the copier shader is also M * v.

now, it should be said that i (somehow) have a track record of < 50% with matrix
order.

+      GLfloat tmp[4];
+      for (int c = 0; c < 16; c += 4) {

[liberato (no reviews please), 2016/03/21 16:21:56]

maybe sievers has a way to avoid writing matrix multiplication in-line here. 
wrap in skmatrix?

+        for (int r = 0; r < 4; ++r) {
+          tmp[r] = 0.0f;
+          for (int i = 0; i < 4; ++i) {
+            tmp[r] += post_transform[r + 4 * i] * gl_matrix[c + i];
+          }
+        }
+        std::copy(tmp, tmp + 4, gl_matrix + c);
+      }
+    } else {
+      // Missing stream texture. Treat matrix as identity.
+      memcpy(gl_matrix, post_transform, sizeof(gl_matrix));
     }
   } else {
     LOCAL_SET_GL_ERROR(GL_INVALID_OPERATION,
                        "DoUniformMatrix4vStreamTextureMatrix",
                        "no texture bound");
     return;
   }
 
   GLenum type = 0;
   GLint real_location = -1;
@@ skipping 8683 matching lines @@
 }
 
 // Include the auto-generated part of this file. We split this because it means
 // we can easily edit the non-auto generated parts right here in this file
 // instead of having to edit some template or the code generator.
 #include "base/macros.h"
 #include "gpu/command_buffer/service/gles2_cmd_decoder_autogen.h"
 
 }  // namespace gles2
 }  // namespace gpu