GPU service interprets glFinish as glFlush.

This is so the single threaded GPU service does not get delayed executing glFinish, which could prevent progress in rendering all other GL contexts. Instead just issue a glFlush to ensure that the pending commands will be processed in the future.

From a client side perspective the semantics of glFlush and glFinish are still different. glFlush is asynchronous whereas glFinish is synchronous. The latter waits for the service to acknowledge that it has issued the glFlush. glFinish can therefore be used to synchronize multiple threads that are issuing commands to GL contexts that share a common namespace.

TEST=trybots
BUG=none

Review URL: http://codereview.chromium.org/1719019

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@45724 0039d316-1c4b-4281-b951-d872f2087c98

gpu/command_buffer/build_gles2_cmd_buffer.py

@@ -1032,7 +1032,11 @@
     'decoder_func': 'DoEnableVertexAttribArray',
     'impl_decl': False,
   },
-  'Finish': {'impl_func': False},
+  'Finish': {
+    'impl_func': False,
+    'decoder_func': 'glFlush',
+    'gl_test_func': 'glFlush',
+  },
   'Flush': {'impl_func': False},
   'FramebufferRenderbuffer': {
     'decoder_func': 'DoFramebufferRenderbuffer',

gpu/command_buffer/service/gles2_cmd_decoder_autogen.h

@@ -671,7 +671,7 @@ error::Error GLES2DecoderImpl::HandleEnableVertexAttribArray(
 
 error::Error GLES2DecoderImpl::HandleFinish(
     uint32 immediate_data_size, const gles2::Finish& c) {
-  glFinish();
+  glFlush();
   return error::kNoError;
 }
 

gpu/command_buffer/service/gles2_cmd_decoder_unittest_1_autogen.h

@@ -763,7 +763,7 @@ TEST_F(GLES2DecoderTest1, EnableVertexAttribArrayValidArgs) {
 }
 
 TEST_F(GLES2DecoderTest1, FinishValidArgs) {
-  EXPECT_CALL(*gl_, Finish());
+  EXPECT_CALL(*gl_, Flush());
   SpecializedSetup<Finish, 0>();
   Finish cmd;
   cmd.Init();