Add generated and hand-written passthrough command handlers with stub Doers.

Add generated and hand-written passthrough command handlers with stub Doers.

Use the build_gles2_cmd_buffer.py script to generate as many handlers as
possible and hand-write the rest.

BUG=602688
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel

Committed: https://crrev.com/d00f0b244d1bbb3d63daa2aa3ae72a83c1e17561
Cr-Commit-Position: refs/heads/master@{#397725}

[Geoff Lang, 2016-05-25 14:19:51]

Description was changed from

==========
Add generated and hand-written passthrough command handlers with stub Doers.

Use the build_gles2_cmd_buffer.py script to generate as many handlers as
possible and hand-write the rest.

BUG=602688
==========

to

==========
Add generated and hand-written passthrough command handlers with stub Doers.

Use the build_gles2_cmd_buffer.py script to generate as many handlers as
possible and hand-write the rest.

BUG=602688
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
==========

[Geoff Lang, 2016-05-25 17:04:16]

Description was changed from

==========
Add generated and hand-written passthrough command handlers with stub Doers.

Use the build_gles2_cmd_buffer.py script to generate as many handlers as
possible and hand-write the rest.

BUG=602688
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
==========

to

==========
Add generated and hand-written passthrough command handlers with stub Doers.

Use the build_gles2_cmd_buffer.py script to generate as many handlers as
possible and hand-write the rest.

BUG=602688
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
==========

[Geoff Lang, 2016-05-25 17:04:16]

geofflang@chromium.org changed reviewers:
+ piman@chromium.org, zmo@chromium.org

[Geoff Lang, 2016-05-25 17:04:42]

PTAL

[Zhenyao Mo, 2016-05-25 20:05:49]

I haven't looked into handlers yet because there are a lot of functions and it
will take me some time to go through them, but here are some preliminary
comments for your consideration.

https://codereview.chromium.org/2012533004/diff/60001/gpu/command_buffer/comm...
File gpu/command_buffer/common/buffer.cc (right):

https://codereview.chromium.org/2012533004/diff/60001/gpu/command_buffer/comm...
gpu/command_buffer/common/buffer.cc:50: return NULL;
I don't really understand why this is necessary (assume you copy the example
from Buffer::GetDataAddress above). There is no numeric operation, so
data_offset should always be valid.

https://codereview.chromium.org/2012533004/diff/60001/gpu/command_buffer/serv...
File gpu/command_buffer/service/gles2_cmd_decoder_passthrough_doer_prototypes.h
(right):

https://codereview.chromium.org/2012533004/diff/60001/gpu/command_buffer/serv...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_doer_prototypes.h:5:
error::Error DoActiveTexture(GLenum texture);
Are you able to auto-generate this file?

[Geoff Lang, 2016-05-25 20:44:30]

Thanks for looking, no rush on the review.  Most of the hand-written handlers
were done by taking a look at how the current command buffer unpacks the
command.

https://codereview.chromium.org/2012533004/diff/60001/gpu/command_buffer/comm...
File gpu/command_buffer/common/buffer.cc (right):

https://codereview.chromium.org/2012533004/diff/60001/gpu/command_buffer/comm...
gpu/command_buffer/common/buffer.cc:50: return NULL;
On 2016/05/25 20:05:49, Zhenyao Mo wrote:
> I don't really understand why this is necessary (assume you copy the example
> from Buffer::GetDataAddress above). There is no numeric operation, so
> data_offset should always be valid.

Removed the checks, probably a result of an initial implementation that was
copy-pasted.

https://codereview.chromium.org/2012533004/diff/60001/gpu/command_buffer/serv...
File gpu/command_buffer/service/gles2_cmd_decoder_passthrough_doer_prototypes.h
(right):

https://codereview.chromium.org/2012533004/diff/60001/gpu/command_buffer/serv...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_doer_prototypes.h:5:
error::Error DoActiveTexture(GLenum texture);
On 2016/05/25 20:05:49, Zhenyao Mo wrote:
> Are you able to auto-generate this file?

This fine was initially created by modifying the script.  After that I made some
changes for commands that require extra parameters or were not fully covered by
the script.

https://codereview.chromium.org/2012533004/diff/60001/gpu/command_buffer/serv...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_doer_prototypes.h:247:
GLsizei* length,
These are some examples of extra parameters.

https://codereview.chromium.org/2012533004/diff/60001/gpu/command_buffer/serv...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_doer_prototypes.h:304:
error::Error DoGetShaderSource(GLuint shader, std::string* source);
More custom stuff, using std::string.

[piman, 2016-05-25 23:34:41]

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/buil...
File gpu/command_buffer/build_gles2_cmd_buffer.py (right):

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/buil...
gpu/command_buffer/build_gles2_cmd_buffer.py:5007:
self.WritePassthroughServiceFunctionHeader(func, f)
I see that you don't have an extension check here. Does that mean that if ANGLE
doesn't support a given extension (e.g. because the underlying GL/DX driver
doesn't support it), it will still accept the function calls given by that
extension, but just maybe fail them (e.g. with a GL error, or a lost context)?
It means we change from a command buffer error (error::kUnknownCommand) to a GL
error, but I think that's acceptable. I just want to make sure this is the
intent, and not an oversight.

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/buil...
gpu/command_buffer/build_gles2_cmd_buffer.py:6941: %(last_arg_type)s
%(last_arg_name)s = result ? result->GetData() : NULL;
Is it the responsibility of the Do* function to check NULL? Is there a case
where NULL is valid? Otherwise, it would be useful I think to just report the
error in the generated code so that we don't miss it.

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/buil...
gpu/command_buffer/build_gles2_cmd_buffer.py:6954: code = """  if
(written_values > bufsize) {
Is the intent that the function will only write up to bufsize entries, but
report how much it would have written in *length?

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/comm...
File gpu/command_buffer/common/buffer.cc (right):

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/comm...
gpu/command_buffer/common/buffer.cc:56: if (!offset.IsValid() ||
offset.ValueOrDie() > static_cast<uint32_t>(size_))
Not sure of the goal of the offset.IsValid() check here. data_offset is a
uint32_t, so moving into a CheckedNumeric<uint32_t> will always be valid.

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/comm...
File gpu/command_buffer/common/gles2_cmd_utils.h (right):

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/comm...
gpu/command_buffer/common/gles2_cmd_utils.h:62: // Returns the address of the
frst byte after the struct or NULL if size >
nit: frst->first

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/serv...
File gpu/command_buffer/service/gles2_cmd_decoder_passthrough_doers.h (right):

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/serv...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_doers.h:10: //
Implementations of commands
Maybe make this a .cc

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/serv...
File gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers.h
(right):

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/serv...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers.h:11: //
Custom Handlers
This too, make it a .cc

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/serv...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers.h:295:
(void)c;
nit: no need for this

[Geoff Lang, 2016-05-27 14:31:28]

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/buil...
File gpu/command_buffer/build_gles2_cmd_buffer.py (right):

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/buil...
gpu/command_buffer/build_gles2_cmd_buffer.py:5007:
self.WritePassthroughServiceFunctionHeader(func, f)
On 2016/05/25 23:34:42, piman wrote:
> I see that you don't have an extension check here. Does that mean that if
ANGLE
> doesn't support a given extension (e.g. because the underlying GL/DX driver
> doesn't support it), it will still accept the function calls given by that
> extension, but just maybe fail them (e.g. with a GL error, or a lost context)?
> It means we change from a command buffer error (error::kUnknownCommand) to a
GL
> error, but I think that's acceptable. I just want to make sure this is the
> intent, and not an oversight.

I think we can continue to report command errors in this case, but I'll probably
do it in the Doer functions.  Reporting GL errors opens up a whole new group of
security issues because it's undefined behavour to call an entry point for an
extension that isn't present (if the GetProcAdress call returns anything to
begin with).

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/buil...
gpu/command_buffer/build_gles2_cmd_buffer.py:6941: %(last_arg_type)s
%(last_arg_name)s = result ? result->GetData() : NULL;
On 2016/05/25 23:34:42, piman wrote:
> Is it the responsibility of the Do* function to check NULL? Is there a case
> where NULL is valid? Otherwise, it would be useful I think to just report the
> error in the generated code so that we don't miss it.

Good point, added the null check here.

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/buil...
gpu/command_buffer/build_gles2_cmd_buffer.py:6954: code = """  if
(written_values > bufsize) {
On 2016/05/25 23:34:42, piman wrote:
> Is the intent that the function will only write up to bufsize entries, but
> report how much it would have written in *length?

Yes, exactly.

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/comm...
File gpu/command_buffer/common/buffer.cc (right):

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/comm...
gpu/command_buffer/common/buffer.cc:56: if (!offset.IsValid() ||
offset.ValueOrDie() > static_cast<uint32_t>(size_))
On 2016/05/25 23:34:42, piman OOO back 2016-5-31 wrote:
> Not sure of the goal of the offset.IsValid() check here. data_offset is a
> uint32_t, so moving into a CheckedNumeric<uint32_t> will always be valid.

Whoops, yes, forgot to remove the CheckedNumerics from both functions.

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/serv...
File gpu/command_buffer/service/gles2_cmd_decoder_passthrough_doers.h (right):

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/serv...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_doers.h:10: //
Implementations of commands
On 2016/05/25 23:34:42, piman OOO back 2016-5-31 wrote:
> Maybe make this a .cc

Doing so makes it harder to hide this class entirely in the
gles2_cmd_decoder_passthrough.cc object.  I could put the
GLES2DecoderPassthroughImpl class in a header and change these to .cc. 
Thoughts?

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/serv...
File gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers.h
(right):

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/serv...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers.h:295:
(void)c;
On 2016/05/25 23:34:42, piman OOO back 2016-5-31 wrote:
> nit: no need for this

Done.

[Zhenyao Mo, 2016-05-31 01:37:22]

Sorry for the delay of reviewing this CL.

Mostly looks good with some minor issues.

https://codereview.chromium.org/2012533004/diff/100001/gpu/command_buffer/ser...
File gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers_autogen.h
(right):

https://codereview.chromium.org/2012533004/diff/100001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers_autogen.h:20:
(void)c;
nit: get rid of this. Here and below.

https://codereview.chromium.org/2012533004/diff/100001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers_autogen.h:26:
return error::kNoError;
This is rather unfortunate that we can't just "return DoActiveTexture(texture)".
But for simplicity in code generator, I guess that's OK.

https://codereview.chromium.org/2012533004/diff/100001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers_autogen.h:1313:
GLsizei bufsize = Result::ComputeMaxResults(buffer_size);
Isn't this the wrong value? The function only works for 32 bit results.

https://codereview.chromium.org/2012533004/diff/100001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers_autogen.h:1373:
GLsizei bufsize = Result::ComputeMaxResults(buffer_size);
Same here.  It seems to be ComputeMaxResults() does not work for 64 bit.

https://codereview.chromium.org/2012533004/diff/100001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers_autogen.h:2305:
(void)length;
nit: this is unnecessary.

[piman, 2016-05-31 15:57:34]

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/serv...
File gpu/command_buffer/service/gles2_cmd_decoder_passthrough_doers.h (right):

https://codereview.chromium.org/2012533004/diff/80001/gpu/command_buffer/serv...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_doers.h:10: //
Implementations of commands
On 2016/05/27 14:31:28, Geoff Lang wrote:
> On 2016/05/25 23:34:42, piman OOO back 2016-5-31 wrote:
> > Maybe make this a .cc
> 
> Doing so makes it harder to hide this class entirely in the
> gles2_cmd_decoder_passthrough.cc object.  I could put the
> GLES2DecoderPassthroughImpl class in a header and change these to .cc. 
> Thoughts? 

Yeah, I'd suggest moving GLES2DecoderPassthroughImpl into a header.
gles2_cmd_decoder.cc is humongous and is very slow to compile. So I think
starting with a split passthrough version is healthier long term.

[Geoff Lang, 2016-06-01 19:40:59]

Remove some.gyp change.

[Geoff Lang, 2016-06-01 20:11:13]

Fix presubmit

[Geoff Lang, 2016-06-01 20:11:51]

https://codereview.chromium.org/2012533004/diff/100001/gpu/command_buffer/ser...
File gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers_autogen.h
(right):

https://codereview.chromium.org/2012533004/diff/100001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers_autogen.h:20:
(void)c;
On 2016/05/31 01:37:22, Zhenyao Mo wrote:
> nit: get rid of this. Here and below.

I can remove it from the generator, but this generation code also generates
gles2_cmd_decoder_autogen.h

https://codereview.chromium.org/2012533004/diff/100001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers_autogen.h:26:
return error::kNoError;
On 2016/05/31 01:37:22, Zhenyao Mo wrote:
> This is rather unfortunate that we can't just "return
DoActiveTexture(texture)".
> But for simplicity in code generator, I guess that's OK.

Agreed, not so simple to detect which commands do extra logic after the Do call.

https://codereview.chromium.org/2012533004/diff/100001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers_autogen.h:1313:
GLsizei bufsize = Result::ComputeMaxResults(buffer_size);
On 2016/05/31 01:37:22, Zhenyao Mo wrote:
> Isn't this the wrong value? The function only works for 32 bit results.

Could you elaborate?  bufsize here is the number of GLint64's that can fit in
the params pointer and it looks like the ComputeMaxResults here does the right
thing (divides by sizeof(GLint64)).

https://codereview.chromium.org/2012533004/diff/100001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers_autogen.h:2305:
(void)length;
On 2016/05/31 01:37:22, Zhenyao Mo wrote:
> nit: this is unnecessary.

Hmm, this is also shared autogen code, I can try removing it.

[Zhenyao Mo, 2016-06-01 20:17:45]

lgtm

https://codereview.chromium.org/2012533004/diff/100001/gpu/command_buffer/ser...
File gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers_autogen.h
(right):

https://codereview.chromium.org/2012533004/diff/100001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers_autogen.h:20:
(void)c;
On 2016/06/01 20:11:51, Geoff Lang wrote:
> On 2016/05/31 01:37:22, Zhenyao Mo wrote:
> > nit: get rid of this. Here and below.
> 
> I can remove it from the generator, but this generation code also generates
> gles2_cmd_decoder_autogen.h

I think it's most likely not needed in gles2_cmd_decoder_autogen.h, but let's
not change it in this CL, to avoid complicating things.

https://codereview.chromium.org/2012533004/diff/100001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers_autogen.h:1313:
GLsizei bufsize = Result::ComputeMaxResults(buffer_size);
On 2016/06/01 20:11:51, Geoff Lang wrote:
> On 2016/05/31 01:37:22, Zhenyao Mo wrote:
> > Isn't this the wrong value? The function only works for 32 bit results.
> 
> Could you elaborate?  bufsize here is the number of GLint64's that can fit in
> the params pointer and it looks like the ComputeMaxResults here does the right
> thing (divides by sizeof(GLint64)).

Never mind.  I thought I read / sizeof(uint32_t), now I recheck the code, it's
not.  You are correct here.

[piman, 2016-06-03 04:23:37]

lgtm

https://codereview.chromium.org/2012533004/diff/180001/gpu/command_buffer/ser...
File gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers.cc
(right):

https://codereview.chromium.org/2012533004/diff/180001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers.cc:2491: }
nit: blank line between functions

https://codereview.chromium.org/2012533004/diff/180001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers.cc:2516: }
nit: blank line between functions

https://codereview.chromium.org/2012533004/diff/180001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers.cc:2538: }
nit: blank line between functions

[Geoff Lang, 2016-06-03 13:33:13]

https://codereview.chromium.org/2012533004/diff/180001/gpu/command_buffer/ser...
File gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers.cc
(right):

https://codereview.chromium.org/2012533004/diff/180001/gpu/command_buffer/ser...
gpu/command_buffer/service/gles2_cmd_decoder_passthrough_handlers.cc:2491: }
On 2016/06/03 04:23:37, piman OOO back 2016-6-2 wrote:
> nit: blank line between functions

Done, and the rest in this file.

[Geoff Lang, 2016-06-03 13:33:56]

The CQ bit was checked by geofflang@chromium.org

[Geoff Lang, 2016-06-03 13:33:57]

The patchset sent to the CQ was uploaded after l-g-t-m from zmo@chromium.org,
piman@chromium.org
Link to the patchset: https://codereview.chromium.org/2012533004/#ps200001
(title: " ")

[commit-bot: I haz the power, 2016-06-03 13:34:16]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2012533004/200001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/2012533004/200001

[commit-bot: I haz the power, 2016-06-03 13:43:55]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-06-03 13:43:56]

Try jobs failed on following builders:
  mac_chromium_gn_rel on tryserver.chromium.mac (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_gn_r...)

[Geoff Lang, 2016-06-03 15:04:52]

The CQ bit was checked by geofflang@chromium.org

[Geoff Lang, 2016-06-03 15:04:53]

The patchset sent to the CQ was uploaded after l-g-t-m from zmo@chromium.org,
piman@chromium.org
Link to the patchset: https://codereview.chromium.org/2012533004/#ps220001
(title: "rebase")

[commit-bot: I haz the power, 2016-06-03 15:05:11]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2012533004/220001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/2012533004/220001

[commit-bot: I haz the power, 2016-06-03 16:35:05]

Description was changed from

==========
Add generated and hand-written passthrough command handlers with stub Doers.

Use the build_gles2_cmd_buffer.py script to generate as many handlers as
possible and hand-write the rest.

BUG=602688
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
==========

to

==========
Add generated and hand-written passthrough command handlers with stub Doers.

Use the build_gles2_cmd_buffer.py script to generate as many handlers as
possible and hand-write the rest.

BUG=602688
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
==========

[commit-bot: I haz the power, 2016-06-03 16:35:07]

Committed patchset #12 (id:220001)

[commit-bot: I haz the power, 2016-06-03 16:36:54]

Description was changed from

==========
Add generated and hand-written passthrough command handlers with stub Doers.

Use the build_gles2_cmd_buffer.py script to generate as many handlers as
possible and hand-write the rest.

BUG=602688
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
==========

to

==========
Add generated and hand-written passthrough command handlers with stub Doers.

Use the build_gles2_cmd_buffer.py script to generate as many handlers as
possible and hand-write the rest.

BUG=602688
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel

Committed: https://crrev.com/d00f0b244d1bbb3d63daa2aa3ae72a83c1e17561
Cr-Commit-Position: refs/heads/master@{#397725}
==========

[commit-bot: I haz the power, 2016-06-03 16:36:55]

Patchset 12 (id:??) landed as
https://crrev.com/d00f0b244d1bbb3d63daa2aa3ae72a83c1e17561
Cr-Commit-Position: refs/heads/master@{#397725}