Revert 227977 "Enable GPU control lists in tests."

Revert 227977 "Enable GPU control lists in tests."

Broke GpuFeatureTest.Canvas2DAllowed on Vista.

GpuFeatureTest.Canvas2DAllowed: 
[1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list match for rule #64 in gpu_blacklist.
[1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list match for rule #65 in gpu_blacklist.
[1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list match for rule #3 in gpu_driver_bug_list.
[1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list match for rule #4 in gpu_driver_bug_list.
[1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list match for rule #17 in gpu_driver_bug_list.
[1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list match for rule #26 in gpu_driver_bug_list.
[1636:5464:1010/153734:2795640:ERROR:chrome_views_delegate.cc(158)] NOT IMPLEMENTED
[1636:5464:1010/153734:2795640:ERROR:desktop_root_window_host_win.cc(690)] NOT IMPLEMENTED
test\gpu\gpu_feature_browsertest.cc(141): error: Expected: (event_count) > (0U), actual: 0 vs 0

> Enable GPU control lists in tests.
> 
> Reland of https://codereview.chromium.org/23534006/ (see that and the bug for more details).
> 
> Also adding INFO-level logging to spell out each control list rule being applied on the bots (makes debugging and understanding failures much easier).
> 
> R=piman@chromium.org
> TBR=jcivelli, piman, zmo
> BUG=295799, 304833
> 
> Previously Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=224810
> Reverted: https://src.chromium.org/viewvc/chrome?view=rev&revision=224845
> 
> Review URL: https://codereview.chromium.org/23703017

TBR=gab@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=228014

[Albert Bodenhamer, 2013-10-10 21:55:19]

Committed patchset #1 manually as r228014.

[gab, 2013-10-10 22:29:55]

Potential other failures' logs for reference:

VISTA
(http://build.chromium.org/p/chromium.win/builders/Vista%20Tests%20%281%29/bui...):

CompositingRenderWidgetHostViewBrowserTestTabCapture.CopyFromCompositingSurface_Cropped_Scaled:

[6620:8680:1010/144905:2436828:INFO:gpu_control_list.cc(1038)] Control list
match for rule #64 in gpu_blacklist.
[6620:8680:1010/144905:2436828:INFO:gpu_control_list.cc(1038)] Control list
match for rule #65 in gpu_blacklist.
[6620:8680:1010/144905:2436828:INFO:gpu_control_list.cc(1038)] Control list
match for rule #3 in gpu_driver_bug_list.
[6620:8680:1010/144905:2436828:INFO:gpu_control_list.cc(1038)] Control list
match for rule #4 in gpu_driver_bug_list.
[6620:8680:1010/144905:2436828:INFO:gpu_control_list.cc(1038)] Control list
match for rule #17 in gpu_driver_bug_list.
[6620:8680:1010/144905:2436828:INFO:gpu_control_list.cc(1038)] Control list
match for rule #26 in gpu_driver_bug_list.
[6620:8680:1010/144905:2436828:ERROR:desktop_root_window_host_win.cc(690)] NOT
IMPLEMENTED
browser\renderer_host\render_widget_host_view_browsertest.cc(556): error: The
difference between expected_red and red is 1, which exceeds allowable_error_,
where
expected_red evaluates to 0,
red evaluates to 1, and
allowable_error_ evaluates to 0.
expected_color: ff00ffff color: ff01fffd Failed at 7, 0 Failure 1
browser\renderer_host\render_widget_host_view_browsertest.cc(566): error: The
difference between expected_blue and blue is 2, which exceeds allowable_error_,
where
expected_blue evaluates to 255,
blue evaluates to 253, and
allowable_error_ evaluates to 0.
expected_color: ff00ffff color: ff01fffd Failed at 7, 0 Failure 2
browser\renderer_host\render_widget_host_view_browsertest.cc(556): error: The
difference between expected_red and red is 1, which exceeds allowable_error_,
where
expected_red evaluates to 0,
red evaluates to 1, and
allowable_error_ evaluates to 0.
expected_color: ff00ffff color: ff01fffd Failed at 7, 1 Failure 3
browser\renderer_host\render_widget_host_view_browsertest.cc(566): error: The
difference between expected_blue and blue is 2, which exceeds allowable_error_,
where
expected_blue evaluates to 255,
blue evaluates to 253, and
allowable_error_ evaluates to 0.
expected_color: ff00ffff color: ff01fffd Failed at 7, 1 Failure 4
browser\renderer_host\render_widget_host_view_browsertest.cc(556): error: The
difference between expected_red and red is 1, which exceeds allowable_error_,
where
expected_red evaluates to 0,
red evaluates to 1, and
allowable_error_ evaluates to 0.
expected_color: ff00ffff color: ff01fffd Failed at 7, 2 Failure 5
browser\renderer_host\render_widget_host_view_browsertest.cc(566): error: The
difference between expected_blue and blue is 2, which exceeds allowable_error_,
where
expected_blue evaluates to 255,
blue evaluates to 253, and
allowable_error_ evaluates to 0.
expected_color: ff00ffff color: ff01fffd Failed at 7, 2 Failure 6
browser\renderer_host\render_widget_host_view_browsertest.cc(556): error: The
difference between expected_red and red is 1, which exceeds allowable_error_,
where
expected_red evaluates to 0,
red evaluates to 1, and
allowable_error_ evaluates to 0.
expected_color: ff00ffff color: ff01fffd Failed at 7, 3 Failure 7
browser\renderer_host\render_widget_host_view_browsertest.cc(566): error: The
difference between expected_blue and blue is 2, which exceeds allowable_error_,
where
expected_blue evaluates to 255,
blue evaluates to 253, and
allowable_error_ evaluates to 0.
expected_color: ff00ffff color: ff01fffd Failed at 7, 3 Failure 8
browser\renderer_host\render_widget_host_view_browsertest.cc(556): error: The
difference between expected_red and red is 1, which exceeds allowable_error_,
where
expected_red evaluates to 0,
red evaluates to 1, and
allowable_error_ evaluates to 0.
expected_color: ff00ffff color: ff01fffd Failed at 7, 4 Failure 9
browser\renderer_host\render_widget_host_view_browsertest.cc(566): error: The
difference between expected_blue and blue is 2, which exceeds allowable_error_,
where
expected_blue evaluates to 255,
blue evaluates to 253, and
allowable_error_ evaluates to 0.
expected_color: ff00ffff color: ff01fffd Failed at 7, 4 Failure 10
browser\renderer_host\render_widget_host_view_browsertest.cc(573): error:
Expected: (fails) < (10), actual: 10 vs 10

And a similar failure for
CompositingRenderWidgetHostViewBrowserTestTabCapture.CopyFromCompositingSurface_Origin_Scaled


ImageTransportFactoryBrowserTest.TestLostContext: 
Did not complete.

[Ken Russell (switch to Gerrit), 2013-10-10 22:30:17]

This also broke gpu_throughput_tests on the Linux Debug (NVIDIA) GPU bot:

http://build.chromium.org/p/chromium.gpu/builders/Linux%20Release%20%28NVIDIA...

Please send re-lands through the win_gpu, linux_gpu, mac_gpu, and mac_gpu_retina
try servers. (Unfortunately these are debug bots and don't run the throughput
tests though...)

[Ken Russell (switch to Gerrit), 2013-10-10 22:31:12]

On 2013/10/10 22:30:17, Ken Russell wrote:
> This also broke gpu_throughput_tests on the Linux Debug (NVIDIA) GPU bot:
> 
>
http://build.chromium.org/p/chromium.gpu/builders/Linux%2520Release%2520%2528...
> 
> Please send re-lands through the win_gpu, linux_gpu, mac_gpu, and
mac_gpu_retina
> try servers. (Unfortunately these are debug bots and don't run the throughput
> tests though...)

URL escaping failed. Correct URL:

http://goo.gl/ZIfOTV

[gab, 2013-10-10 22:37:07]

+vangelis, cpu, erg

The linux gpu bot failures are probably a bad sign as this means these tests
don't pass under the default blacklist config we ship users today :(.

I'm going to have to make my change Windows only for now I guess while we sort
everything out for each platform (I had already taken Mac out of this last
attempt, but it looks like Linux is part of the issue as well now)...

The problem is not this CL though, it's the fact that it highlights existing
problems in the configuration we currently ship..!

[Ken Russell (switch to Gerrit), 2013-10-10 22:41:26]

lgtm

[Ken Russell (switch to Gerrit), 2013-10-10 22:43:27]

Crap, wrong button.

+alokp, ernstm

I think the gpu_throughput_tests are being migrated to Telemetry which should
hopefully make them more reliable and correct. There have been problems with
these tests on some platforms which already resulted in their being disabled on
Windows.

[gab, 2013-10-10 22:48:25]

Other failures (now failed twice in a row on tries pre-revert and hadn't failed
much before):

On VISTA
http://build.chromium.org/p/chromium.win/builders/Vista%20Tests%20%282%29/bui...:

(view as text)
OutOfProcessPPAPITest.Graphics3D: 
[61540:61560:1010/153904:3046828:INFO:gpu_control_list.cc(1038)] Control list
match for rule #64 in gpu_blacklist.
[61540:61560:1010/153904:3046828:INFO:gpu_control_list.cc(1038)] Control list
match for rule #65 in gpu_blacklist.
[61540:61560:1010/153905:3046843:INFO:gpu_control_list.cc(1038)] Control list
match for rule #3 in gpu_driver_bug_list.
[61540:61560:1010/153905:3046843:INFO:gpu_control_list.cc(1038)] Control list
match for rule #4 in gpu_driver_bug_list.
[61540:61560:1010/153905:3046843:INFO:gpu_control_list.cc(1038)] Control list
match for rule #17 in gpu_driver_bug_list.
[61540:61560:1010/153905:3046843:INFO:gpu_control_list.cc(1038)] Control list
match for rule #26 in gpu_driver_bug_list.
[61540:61560:1010/153905:3046890:ERROR:chrome_views_delegate.cc(158)] NOT
IMPLEMENTED
[61540:61560:1010/153905:3046890:ERROR:desktop_root_window_host_win.cc(690)] NOT
IMPLEMENTED
[61540:61560:1010/153905:3047140:INFO:CONSOLE(0)] "ppapi_tests.dll: 0x7B is not
a valid resource ID.", source:
file:///E:/b/build/slave/Vista_Tests__2_/build/src/ppapi/tests/test_case.html?testcase=Graphics3D
(0)
[61540:61560:1010/153905:3047140:INFO:CONSOLE(0)] "ppapi_tests.dll: 0x7B is not
a valid resource ID.", source:
file:///E:/b/build/slave/Vista_Tests__2_/build/src/ppapi/tests/test_case.html?testcase=Graphics3D
(0)
(and a ton more identical INFO_CONSOLE logs which make this go past the 10000
chars limit...)
test\ppapi\ppapi_test.cc(268): error: Value of: handler.message().c_str()
Actual: "FramePPAPIForceAsync FAIL: Failure in tests\\\\test_graphics_3d.cc(49):
!(context.is_null()), FramePPAPI FAIL: Failure in
tests\\\\test_graphics_3d.cc(49): !(context.is_null()), FramePPAPIBlocking FAIL:
Failure in tests\\\\test_graphics_3d.cc(49): !(context.is_null()),
FramePPAPIBackground FAIL: Failure in tests\\\\test_graphics_3d.cc(49):
!(context.is_null()), FrameGLForceAsync FAIL: Failure in
tests\\\\test_graphics_3d.cc(77): !(context.is_null()), FrameGL FAIL: Failure in
tests\\\\test_graphics_3d.cc(77): !(context.is_null()), FrameGLBlocking FAIL:
Failure in tests\\\\test_graphics_3d.cc(77): !(context.is_null()),
FrameGLBackground FAIL: Failure in tests\\\\test_graphics_3d.cc(77):
!(context.is_null()), ExtensionsGLForceAsync FAIL: Failure in
tests\\\\test_graphics_3d.cc(105): !(context.is_null()), ExtensionsGL FAIL:
Failure in tests\\\\test_graphics_3d.cc(105): !(context.is_null()),
ExtensionsGLBlocking FAIL: Failure in tests\\\\test_graphics_3d.cc(105):
!(context.is_null()), ExtensionsGLBackground FAIL: Failure in
tests\\\\test_graphics_3d.cc(105): !(context.is_null())"
Expected: "PASS"

[gab, 2013-10-10 23:02:12]

@vangelis: Regarding the original failure for Canvas2DAllowed, is it possible
that rules #64/65 which blacklist: accelerated_video_decode, flash_3d,
flash_stage3d, and force_compositing_mode would have forced the
kAcceleratedCanvasCreationEvent to not occur in that test?

i.e., accelerated_2d_canvas itself isn't disabled, yet the event isn't
triggered...

[gab, 2013-10-10 23:30:54]

On 2013/10/10 22:48:25, gab wrote:
> Other failures (now failed twice in a row on tries pre-revert and hadn't
failed
> much before):
> 
> On VISTA
>
http://build.chromium.org/p/chromium.win/builders/Vista%2520Tests%2520%25282%...:
> 
> (view as text)
> OutOfProcessPPAPITest.Graphics3D: 



This test also fails on XP (repro'able locally).



> [61540:61560:1010/153904:3046828:INFO:gpu_control_list.cc(1038)] Control list
> match for rule #64 in gpu_blacklist.
> [61540:61560:1010/153904:3046828:INFO:gpu_control_list.cc(1038)] Control list
> match for rule #65 in gpu_blacklist.
> [61540:61560:1010/153905:3046843:INFO:gpu_control_list.cc(1038)] Control list
> match for rule #3 in gpu_driver_bug_list.
> [61540:61560:1010/153905:3046843:INFO:gpu_control_list.cc(1038)] Control list
> match for rule #4 in gpu_driver_bug_list.
> [61540:61560:1010/153905:3046843:INFO:gpu_control_list.cc(1038)] Control list
> match for rule #17 in gpu_driver_bug_list.
> [61540:61560:1010/153905:3046843:INFO:gpu_control_list.cc(1038)] Control list
> match for rule #26 in gpu_driver_bug_list.
> [61540:61560:1010/153905:3046890:ERROR:chrome_views_delegate.cc(158)] NOT
> IMPLEMENTED
> [61540:61560:1010/153905:3046890:ERROR:desktop_root_window_host_win.cc(690)]
NOT
> IMPLEMENTED
> [61540:61560:1010/153905:3047140:INFO:CONSOLE(0)] "ppapi_tests.dll: 0x7B is
not
> a valid resource ID.", source:
>
file:///E:/b/build/slave/Vista_Tests__2_/build/src/ppapi/tests/test_case.html?testcase=Graphics3D
> (0)
> [61540:61560:1010/153905:3047140:INFO:CONSOLE(0)] "ppapi_tests.dll: 0x7B is
not
> a valid resource ID.", source:
>
file:///E:/b/build/slave/Vista_Tests__2_/build/src/ppapi/tests/test_case.html?testcase=Graphics3D
> (0)
> (and a ton more identical INFO_CONSOLE logs which make this go past the 10000
> chars limit...)
> test\ppapi\ppapi_test.cc(268): error: Value of: handler.message().c_str()
> Actual: "FramePPAPIForceAsync FAIL: Failure in
tests\\\\test_graphics_3d.cc(49):
> !(context.is_null()), FramePPAPI FAIL: Failure in
> tests\\\\test_graphics_3d.cc(49): !(context.is_null()), FramePPAPIBlocking
FAIL:
> Failure in tests\\\\test_graphics_3d.cc(49): !(context.is_null()),
> FramePPAPIBackground FAIL: Failure in tests\\\\test_graphics_3d.cc(49):
> !(context.is_null()), FrameGLForceAsync FAIL: Failure in
> tests\\\\test_graphics_3d.cc(77): !(context.is_null()), FrameGL FAIL: Failure
in
> tests\\\\test_graphics_3d.cc(77): !(context.is_null()), FrameGLBlocking FAIL:
> Failure in tests\\\\test_graphics_3d.cc(77): !(context.is_null()),
> FrameGLBackground FAIL: Failure in tests\\\\test_graphics_3d.cc(77):
> !(context.is_null()), ExtensionsGLForceAsync FAIL: Failure in
> tests\\\\test_graphics_3d.cc(105): !(context.is_null()), ExtensionsGL FAIL:
> Failure in tests\\\\test_graphics_3d.cc(105): !(context.is_null()),
> ExtensionsGLBlocking FAIL: Failure in tests\\\\test_graphics_3d.cc(105):
> !(context.is_null()), ExtensionsGLBackground FAIL: Failure in
> tests\\\\test_graphics_3d.cc(105): !(context.is_null())"
> Expected: "PASS"

[gab, 2013-10-10 23:40:03]

Note control list matches are as follows:

XP:
gpu_blacklist #41 (flash_3d), 64(accelerated_video_decode),
67(accelerated_2d_canvas), 79(force_compositing_mode) 
gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
4(clear_alpha_in_readpixels), 17(exit_on_context_lost), 26(disable_d3d11)

VISTA:
gpu_blaclist #64(accelerated_video_decode), #65(flash_3d, flash_stage3d,
force_compositing_mode)
gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
4(clear_alpha_in_readpixels), 17(exit_on_context_lost), 26(disable_d3d11)


Playing with some of these locally will most likely reveal what's causing which
failures.

Or maybe some of you more experienced graphics folks know what could cause what
just by looking at it?

Cheers!
Gab

[vangelis, 2013-10-11 04:51:12]

On 2013/10/10 23:40:03, gab wrote:
> Note control list matches are as follows:
> 
> XP:
> gpu_blacklist #41 (flash_3d), 64(accelerated_video_decode),
> 67(accelerated_2d_canvas), 79(force_compositing_mode) 
> gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
> 4(clear_alpha_in_readpixels), 17(exit_on_context_lost), 26(disable_d3d11)
> 
> VISTA:
> gpu_blaclist #64(accelerated_video_decode), #65(flash_3d, flash_stage3d,
> force_compositing_mode)
> gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
> 4(clear_alpha_in_readpixels), 17(exit_on_context_lost), 26(disable_d3d11)
> 
> 
> Playing with some of these locally will most likely reveal what's causing
which
> failures.
> 
> Or maybe some of you more experienced graphics folks know what could cause
what
> just by looking at it?
> 
> Cheers!
> Gab

Gab, I would have to look at the individual tests to see what's going on. Is
there a complete list of failures somewhere as I'm not seeing the bot results.
Thanks!

[vangelis, 2013-10-11 04:51:14]

On 2013/10/10 23:40:03, gab wrote:
> Note control list matches are as follows:
> 
> XP:
> gpu_blacklist #41 (flash_3d), 64(accelerated_video_decode),
> 67(accelerated_2d_canvas), 79(force_compositing_mode) 
> gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
> 4(clear_alpha_in_readpixels), 17(exit_on_context_lost), 26(disable_d3d11)
> 
> VISTA:
> gpu_blaclist #64(accelerated_video_decode), #65(flash_3d, flash_stage3d,
> force_compositing_mode)
> gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
> 4(clear_alpha_in_readpixels), 17(exit_on_context_lost), 26(disable_d3d11)
> 
> 
> Playing with some of these locally will most likely reveal what's causing
which
> failures.
> 
> Or maybe some of you more experienced graphics folks know what could cause
what
> just by looking at it?
> 
> Cheers!
> Gab

Gab, I would have to look at the individual tests to see what's going on. Is
there a complete list of failures somewhere as I'm not seeing the bot results.
Thanks!

[gab, 2013-10-11 13:38:53]

On 2013/10/11 04:51:14, vangelis wrote:
> On 2013/10/10 23:40:03, gab wrote:
> > Note control list matches are as follows:
> > 
> > XP:
> > gpu_blacklist #41 (flash_3d), 64(accelerated_video_decode),
> > 67(accelerated_2d_canvas), 79(force_compositing_mode) 
> > gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
> > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost), 26(disable_d3d11)
> > 
> > VISTA:
> > gpu_blaclist #64(accelerated_video_decode), #65(flash_3d, flash_stage3d,
> > force_compositing_mode)
> > gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
> > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost), 26(disable_d3d11)
> > 
> > 
> > Playing with some of these locally will most likely reveal what's causing
> which
> > failures.
> > 
> > Or maybe some of you more experienced graphics folks know what could cause
> what
> > just by looking at it?
> > 
> > Cheers!
> > Gab
> 
> Gab, I would have to look at the individual tests to see what's going on. Is
> there a complete list of failures somewhere as I'm not seeing the bot results.
> Thanks!

I listed the failures with detailed logs in the comments of this reverted CL
(and added the log for the initial failure causing this revert in the CL
description).

I'm guessing it doesn't make sense to run OutOfProcessPPAPITest.Graphics3D on
machines where flash_3d is disabled, right?

[vangelis, 2013-10-11 16:21:55]

On 2013/10/11 13:38:53, gab wrote:
> On 2013/10/11 04:51:14, vangelis wrote:
> > On 2013/10/10 23:40:03, gab wrote:
> > > Note control list matches are as follows:
> > > 
> > > XP:
> > > gpu_blacklist #41 (flash_3d), 64(accelerated_video_decode),
> > > 67(accelerated_2d_canvas), 79(force_compositing_mode) 
> > > gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
> > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost), 26(disable_d3d11)
> > > 
> > > VISTA:
> > > gpu_blaclist #64(accelerated_video_decode), #65(flash_3d, flash_stage3d,
> > > force_compositing_mode)
> > > gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
> > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost), 26(disable_d3d11)
> > > 
> > > 
> > > Playing with some of these locally will most likely reveal what's causing
> > which
> > > failures.
> > > 
> > > Or maybe some of you more experienced graphics folks know what could cause
> > what
> > > just by looking at it?
> > > 
> > > Cheers!
> > > Gab
> > 
> > Gab, I would have to look at the individual tests to see what's going on. Is
> > there a complete list of failures somewhere as I'm not seeing the bot
results.
> > Thanks!
> 
> I listed the failures with detailed logs in the comments of this reverted CL
> (and added the log for the initial failure causing this revert in the CL
> description).

Thanks, Gab! I don't think any of the blacklist entries would affect canvas but
it makes me wonder whether this particular test would ever succeed on a bot that
doesn't have a GPU. This Vista bot presumably is a VM that runs the compositor
in Software mode? Mo, any ideas? 
 

> 
> I'm guessing it doesn't make sense to run OutOfProcessPPAPITest.Graphics3D on
> machines where flash_3d is disabled, right?

Yes, I don't think we could run these tests if flash_3d is disabled. It looks
like they have mostly been disabled for Windows altogether anyway? 

https://code.google.com/p/chromium/codesearch#chromium/src/chrome/test/ppapi/...

[Zhenyao Mo, 2013-10-11 19:38:27]

On 2013/10/11 16:21:55, vangelis wrote:
> On 2013/10/11 13:38:53, gab wrote:
> > On 2013/10/11 04:51:14, vangelis wrote:
> > > On 2013/10/10 23:40:03, gab wrote:
> > > > Note control list matches are as follows:
> > > > 
> > > > XP:
> > > > gpu_blacklist #41 (flash_3d), 64(accelerated_video_decode),
> > > > 67(accelerated_2d_canvas), 79(force_compositing_mode) 
> > > > gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
> > > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost),
26(disable_d3d11)
> > > > 
> > > > VISTA:
> > > > gpu_blaclist #64(accelerated_video_decode), #65(flash_3d, flash_stage3d,
> > > > force_compositing_mode)
> > > > gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
> > > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost),
26(disable_d3d11)
> > > > 
> > > > 
> > > > Playing with some of these locally will most likely reveal what's
causing
> > > which
> > > > failures.
> > > > 
> > > > Or maybe some of you more experienced graphics folks know what could
cause
> > > what
> > > > just by looking at it?
> > > > 
> > > > Cheers!
> > > > Gab
> > > 
> > > Gab, I would have to look at the individual tests to see what's going on.
Is
> > > there a complete list of failures somewhere as I'm not seeing the bot
> results.
> > > Thanks!
> > 
> > I listed the failures with detailed logs in the comments of this reverted CL
> > (and added the log for the initial failure causing this revert in the CL
> > description).
> 
> Thanks, Gab! I don't think any of the blacklist entries would affect canvas
but
> it makes me wonder whether this particular test would ever succeed on a bot
that
> doesn't have a GPU. This Vista bot presumably is a VM that runs the compositor
> in Software mode? Mo, any ideas? 

Gab, do you have a link to the failing vista bot?  I'd like to have a look.

>  
> 
> > 
> > I'm guessing it doesn't make sense to run OutOfProcessPPAPITest.Graphics3D
on
> > machines where flash_3d is disabled, right?
> 
> Yes, I don't think we could run these tests if flash_3d is disabled. It looks
> like they have mostly been disabled for Windows altogether anyway? 
> 
>
https://code.google.com/p/chromium/codesearch#chromium/src/chrome/test/ppapi/...

[gab, 2013-10-11 20:05:32]

On 2013/10/11 19:38:27, Zhenyao Mo wrote:
> On 2013/10/11 16:21:55, vangelis wrote:
> > On 2013/10/11 13:38:53, gab wrote:
> > > On 2013/10/11 04:51:14, vangelis wrote:
> > > > On 2013/10/10 23:40:03, gab wrote:
> > > > > Note control list matches are as follows:
> > > > > 
> > > > > XP:
> > > > > gpu_blacklist #41 (flash_3d), 64(accelerated_video_decode),
> > > > > 67(accelerated_2d_canvas), 79(force_compositing_mode) 
> > > > > gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
> > > > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost),
> 26(disable_d3d11)
> > > > > 
> > > > > VISTA:
> > > > > gpu_blaclist #64(accelerated_video_decode), #65(flash_3d,
flash_stage3d,
> > > > > force_compositing_mode)
> > > > > gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
> > > > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost),
> 26(disable_d3d11)
> > > > > 
> > > > > 
> > > > > Playing with some of these locally will most likely reveal what's
> causing
> > > > which
> > > > > failures.
> > > > > 
> > > > > Or maybe some of you more experienced graphics folks know what could
> cause
> > > > what
> > > > > just by looking at it?
> > > > > 
> > > > > Cheers!
> > > > > Gab
> > > > 
> > > > Gab, I would have to look at the individual tests to see what's going
on.
> Is
> > > > there a complete list of failures somewhere as I'm not seeing the bot
> > results.
> > > > Thanks!
> > > 
> > > I listed the failures with detailed logs in the comments of this reverted
CL
> > > (and added the log for the initial failure causing this revert in the CL
> > > description).
> > 
> > Thanks, Gab! I don't think any of the blacklist entries would affect canvas
> but
> > it makes me wonder whether this particular test would ever succeed on a bot
> that
> > doesn't have a GPU. This Vista bot presumably is a VM that runs the
compositor
> > in Software mode? Mo, any ideas? 
> 
> Gab, do you have a link to the failing vista bot?  I'd like to have a look.

Failing Vista bot:
http://build.chromium.org/p/chromium.win/builders/Vista%20Tests%20%281%29/bui...

> 
> >  
> > 
> > > 
> > > I'm guessing it doesn't make sense to run OutOfProcessPPAPITest.Graphics3D
> on
> > > machines where flash_3d is disabled, right?
> > 
> > Yes, I don't think we could run these tests if flash_3d is disabled. It
looks
> > like they have mostly been disabled for Windows altogether anyway? 
> > 
> >
>
https://code.google.com/p/chromium/codesearch#chromium/src/chrome/test/ppapi/...

[Zhenyao Mo, 2013-10-11 21:38:27]

On 2013/10/11 20:05:32, gab wrote:
> On 2013/10/11 19:38:27, Zhenyao Mo wrote:
> > On 2013/10/11 16:21:55, vangelis wrote:
> > > On 2013/10/11 13:38:53, gab wrote:
> > > > On 2013/10/11 04:51:14, vangelis wrote:
> > > > > On 2013/10/10 23:40:03, gab wrote:
> > > > > > Note control list matches are as follows:
> > > > > > 
> > > > > > XP:
> > > > > > gpu_blacklist #41 (flash_3d), 64(accelerated_video_decode),
> > > > > > 67(accelerated_2d_canvas), 79(force_compositing_mode) 
> > > > > > gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
> > > > > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost),
> > 26(disable_d3d11)
> > > > > > 
> > > > > > VISTA:
> > > > > > gpu_blaclist #64(accelerated_video_decode), #65(flash_3d,
> flash_stage3d,
> > > > > > force_compositing_mode)
> > > > > > gpu_driver_bug_list #3(set_texture_filter_before_generating_mipmap),
> > > > > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost),
> > 26(disable_d3d11)
> > > > > > 
> > > > > > 
> > > > > > Playing with some of these locally will most likely reveal what's
> > causing
> > > > > which
> > > > > > failures.
> > > > > > 
> > > > > > Or maybe some of you more experienced graphics folks know what could
> > cause
> > > > > what
> > > > > > just by looking at it?
> > > > > > 
> > > > > > Cheers!
> > > > > > Gab
> > > > > 
> > > > > Gab, I would have to look at the individual tests to see what's going
> on.
> > Is
> > > > > there a complete list of failures somewhere as I'm not seeing the bot
> > > results.
> > > > > Thanks!
> > > > 
> > > > I listed the failures with detailed logs in the comments of this
reverted
> CL
> > > > (and added the log for the initial failure causing this revert in the CL
> > > > description).
> > > 
> > > Thanks, Gab! I don't think any of the blacklist entries would affect
canvas
> > but
> > > it makes me wonder whether this particular test would ever succeed on a
bot
> > that
> > > doesn't have a GPU. This Vista bot presumably is a VM that runs the
> compositor
> > > in Software mode? Mo, any ideas? 
> > 
> > Gab, do you have a link to the failing vista bot?  I'd like to have a look.
> 
> Failing Vista bot:
>
http://build.chromium.org/p/chromium.win/builders/Vista%2520Tests%2520%25281%...
> 
> > 
> > >  
> > > 
> > > > 
> > > > I'm guessing it doesn't make sense to run
OutOfProcessPPAPITest.Graphics3D
> > on
> > > > machines where flash_3d is disabled, right?
> > > 
> > > Yes, I don't think we could run these tests if flash_3d is disabled. It
> looks
> > > like they have mostly been disabled for Windows altogether anyway? 
> > > 
> > >
> >
>
https://code.google.com/p/chromium/codesearch#chromium/src/chrome/test/ppapi/...

The canvas isn't accelerated because accelerated compositor is turned off due to
it's a software renderer (blacklist entry 3).

[gab, 2013-10-11 22:20:35]

On 2013/10/11 21:38:27, Zhenyao Mo wrote:
> On 2013/10/11 20:05:32, gab wrote:
> > On 2013/10/11 19:38:27, Zhenyao Mo wrote:
> > > On 2013/10/11 16:21:55, vangelis wrote:
> > > > On 2013/10/11 13:38:53, gab wrote:
> > > > > On 2013/10/11 04:51:14, vangelis wrote:
> > > > > > On 2013/10/10 23:40:03, gab wrote:
> > > > > > > Note control list matches are as follows:
> > > > > > > 
> > > > > > > XP:
> > > > > > > gpu_blacklist #41 (flash_3d), 64(accelerated_video_decode),
> > > > > > > 67(accelerated_2d_canvas), 79(force_compositing_mode) 
> > > > > > > gpu_driver_bug_list
#3(set_texture_filter_before_generating_mipmap),
> > > > > > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost),
> > > 26(disable_d3d11)
> > > > > > > 
> > > > > > > VISTA:
> > > > > > > gpu_blaclist #64(accelerated_video_decode), #65(flash_3d,
> > flash_stage3d,
> > > > > > > force_compositing_mode)
> > > > > > > gpu_driver_bug_list
#3(set_texture_filter_before_generating_mipmap),
> > > > > > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost),
> > > 26(disable_d3d11)
> > > > > > > 
> > > > > > > 
> > > > > > > Playing with some of these locally will most likely reveal what's
> > > causing
> > > > > > which
> > > > > > > failures.
> > > > > > > 
> > > > > > > Or maybe some of you more experienced graphics folks know what
could
> > > cause
> > > > > > what
> > > > > > > just by looking at it?
> > > > > > > 
> > > > > > > Cheers!
> > > > > > > Gab
> > > > > > 
> > > > > > Gab, I would have to look at the individual tests to see what's
going
> > on.
> > > Is
> > > > > > there a complete list of failures somewhere as I'm not seeing the
bot
> > > > results.
> > > > > > Thanks!
> > > > > 
> > > > > I listed the failures with detailed logs in the comments of this
> reverted
> > CL
> > > > > (and added the log for the initial failure causing this revert in the
CL
> > > > > description).
> > > > 
> > > > Thanks, Gab! I don't think any of the blacklist entries would affect
> canvas
> > > but
> > > > it makes me wonder whether this particular test would ever succeed on a
> bot
> > > that
> > > > doesn't have a GPU. This Vista bot presumably is a VM that runs the
> > compositor
> > > > in Software mode? Mo, any ideas? 
> > > 
> > > Gab, do you have a link to the failing vista bot?  I'd like to have a
look.
> > 
> > Failing Vista bot:
> >
>
http://build.chromium.org/p/chromium.win/builders/Vista%252520Tests%252520%25...
> > 
> > > 
> > > >  
> > > > 
> > > > > 
> > > > > I'm guessing it doesn't make sense to run
> OutOfProcessPPAPITest.Graphics3D
> > > on
> > > > > machines where flash_3d is disabled, right?
> > > > 
> > > > Yes, I don't think we could run these tests if flash_3d is disabled. It
> > looks
> > > > like they have mostly been disabled for Windows altogether anyway? 
> > > > 
> > > >
> > >
> >
>
https://code.google.com/p/chromium/codesearch#chromium/src/chrome/test/ppapi/...
> 
> The canvas isn't accelerated because accelerated compositor is turned off due
to
> it's a software renderer (blacklist entry 3).

Hmmm,
http://build.chromium.org/p/chromium.win/builders/Vista%20Tests%20%281%29/bui...
doesn't suggest that rule #3 was applied (in fact rule #3 is a linux-only rule).

[gab, 2013-10-11 22:24:36]

On 2013/10/11 22:20:35, gab wrote:
> On 2013/10/11 21:38:27, Zhenyao Mo wrote:
> > On 2013/10/11 20:05:32, gab wrote:
> > > On 2013/10/11 19:38:27, Zhenyao Mo wrote:
> > > > On 2013/10/11 16:21:55, vangelis wrote:
> > > > > On 2013/10/11 13:38:53, gab wrote:
> > > > > > On 2013/10/11 04:51:14, vangelis wrote:
> > > > > > > On 2013/10/10 23:40:03, gab wrote:
> > > > > > > > Note control list matches are as follows:
> > > > > > > > 
> > > > > > > > XP:
> > > > > > > > gpu_blacklist #41 (flash_3d), 64(accelerated_video_decode),
> > > > > > > > 67(accelerated_2d_canvas), 79(force_compositing_mode) 
> > > > > > > > gpu_driver_bug_list
> #3(set_texture_filter_before_generating_mipmap),
> > > > > > > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost),
> > > > 26(disable_d3d11)
> > > > > > > > 
> > > > > > > > VISTA:
> > > > > > > > gpu_blaclist #64(accelerated_video_decode), #65(flash_3d,
> > > flash_stage3d,
> > > > > > > > force_compositing_mode)
> > > > > > > > gpu_driver_bug_list
> #3(set_texture_filter_before_generating_mipmap),
> > > > > > > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost),
> > > > 26(disable_d3d11)
> > > > > > > > 
> > > > > > > > 
> > > > > > > > Playing with some of these locally will most likely reveal
what's
> > > > causing
> > > > > > > which
> > > > > > > > failures.
> > > > > > > > 
> > > > > > > > Or maybe some of you more experienced graphics folks know what
> could
> > > > cause
> > > > > > > what
> > > > > > > > just by looking at it?
> > > > > > > > 
> > > > > > > > Cheers!
> > > > > > > > Gab
> > > > > > > 
> > > > > > > Gab, I would have to look at the individual tests to see what's
> going
> > > on.
> > > > Is
> > > > > > > there a complete list of failures somewhere as I'm not seeing the
> bot
> > > > > results.
> > > > > > > Thanks!
> > > > > > 
> > > > > > I listed the failures with detailed logs in the comments of this
> > reverted
> > > CL
> > > > > > (and added the log for the initial failure causing this revert in
the
> CL
> > > > > > description).
> > > > > 
> > > > > Thanks, Gab! I don't think any of the blacklist entries would affect
> > canvas
> > > > but
> > > > > it makes me wonder whether this particular test would ever succeed on
a
> > bot
> > > > that
> > > > > doesn't have a GPU. This Vista bot presumably is a VM that runs the
> > > compositor
> > > > > in Software mode? Mo, any ideas? 
> > > > 
> > > > Gab, do you have a link to the failing vista bot?  I'd like to have a
> look.
> > > 
> > > Failing Vista bot:
> > >
> >
>
http://build.chromium.org/p/chromium.win/builders/Vista%25252520Tests%2525252...
> > > 
> > > > 
> > > > >  
> > > > > 
> > > > > > 
> > > > > > I'm guessing it doesn't make sense to run
> > OutOfProcessPPAPITest.Graphics3D
> > > > on
> > > > > > machines where flash_3d is disabled, right?
> > > > > 
> > > > > Yes, I don't think we could run these tests if flash_3d is disabled.
It
> > > looks
> > > > > like they have mostly been disabled for Windows altogether anyway? 
> > > > > 
> > > > >
> > > >
> > >
> >
>
https://code.google.com/p/chromium/codesearch#chromium/src/chrome/test/ppapi/...
> > 
> > The canvas isn't accelerated because accelerated compositor is turned off
due
> to
> > it's a software renderer (blacklist entry 3).
> 
> Hmmm,
>
http://build.chromium.org/p/chromium.win/builders/Vista%2520Tests%2520%25281%...
> doesn't suggest that rule #3 was applied (in fact rule #3 is a linux-only
rule).

You might have mistakenly read the log entry that mentions that the
gpu_driver_bug_list #3 was applied; this is not the gpu_blacklist, it refers to:


gpu_driver_bug_list_json.cc:52-57:
    {
      "id": 3,
      "features": [
        "set_texture_filter_before_generating_mipmap"
      ]
    },

[gab, 2013-10-11 22:25:31]

On 2013/10/11 22:24:36, gab wrote:
> On 2013/10/11 22:20:35, gab wrote:
> > On 2013/10/11 21:38:27, Zhenyao Mo wrote:
> > > On 2013/10/11 20:05:32, gab wrote:
> > > > On 2013/10/11 19:38:27, Zhenyao Mo wrote:
> > > > > On 2013/10/11 16:21:55, vangelis wrote:
> > > > > > On 2013/10/11 13:38:53, gab wrote:
> > > > > > > On 2013/10/11 04:51:14, vangelis wrote:
> > > > > > > > On 2013/10/10 23:40:03, gab wrote:
> > > > > > > > > Note control list matches are as follows:
> > > > > > > > > 
> > > > > > > > > XP:
> > > > > > > > > gpu_blacklist #41 (flash_3d), 64(accelerated_video_decode),
> > > > > > > > > 67(accelerated_2d_canvas), 79(force_compositing_mode) 
> > > > > > > > > gpu_driver_bug_list
> > #3(set_texture_filter_before_generating_mipmap),
> > > > > > > > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost),
> > > > > 26(disable_d3d11)
> > > > > > > > > 
> > > > > > > > > VISTA:
> > > > > > > > > gpu_blaclist #64(accelerated_video_decode), #65(flash_3d,
> > > > flash_stage3d,
> > > > > > > > > force_compositing_mode)
> > > > > > > > > gpu_driver_bug_list
> > #3(set_texture_filter_before_generating_mipmap),
> > > > > > > > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost),
> > > > > 26(disable_d3d11)
> > > > > > > > > 
> > > > > > > > > 
> > > > > > > > > Playing with some of these locally will most likely reveal
> what's
> > > > > causing
> > > > > > > > which
> > > > > > > > > failures.
> > > > > > > > > 
> > > > > > > > > Or maybe some of you more experienced graphics folks know what
> > could
> > > > > cause
> > > > > > > > what
> > > > > > > > > just by looking at it?
> > > > > > > > > 
> > > > > > > > > Cheers!
> > > > > > > > > Gab
> > > > > > > > 
> > > > > > > > Gab, I would have to look at the individual tests to see what's
> > going
> > > > on.
> > > > > Is
> > > > > > > > there a complete list of failures somewhere as I'm not seeing
the
> > bot
> > > > > > results.
> > > > > > > > Thanks!
> > > > > > > 
> > > > > > > I listed the failures with detailed logs in the comments of this
> > > reverted
> > > > CL
> > > > > > > (and added the log for the initial failure causing this revert in
> the
> > CL
> > > > > > > description).
> > > > > > 
> > > > > > Thanks, Gab! I don't think any of the blacklist entries would affect
> > > canvas
> > > > > but
> > > > > > it makes me wonder whether this particular test would ever succeed
on
> a
> > > bot
> > > > > that
> > > > > > doesn't have a GPU. This Vista bot presumably is a VM that runs the
> > > > compositor
> > > > > > in Software mode? Mo, any ideas? 
> > > > > 
> > > > > Gab, do you have a link to the failing vista bot?  I'd like to have a
> > look.
> > > > 
> > > > Failing Vista bot:
> > > >
> > >
> >
>
http://build.chromium.org/p/chromium.win/builders/Vista%2525252520Tests%25252...
> > > > 
> > > > > 
> > > > > >  
> > > > > > 
> > > > > > > 
> > > > > > > I'm guessing it doesn't make sense to run
> > > OutOfProcessPPAPITest.Graphics3D
> > > > > on
> > > > > > > machines where flash_3d is disabled, right?
> > > > > > 
> > > > > > Yes, I don't think we could run these tests if flash_3d is disabled.
> It
> > > > looks
> > > > > > like they have mostly been disabled for Windows altogether anyway? 
> > > > > > 
> > > > > >
> > > > >
> > > >
> > >
> >
>
https://code.google.com/p/chromium/codesearch#chromium/src/chrome/test/ppapi/...
> > > 
> > > The canvas isn't accelerated because accelerated compositor is turned off
> due
> > to
> > > it's a software renderer (blacklist entry 3).
> > 
> > Hmmm,
> >
>
http://build.chromium.org/p/chromium.win/builders/Vista%252520Tests%252520%25...
> > doesn't suggest that rule #3 was applied (in fact rule #3 is a linux-only
> rule).
> 
> You might have mistakenly read the log entry that mentions that the
> gpu_driver_bug_list #3 was applied; this is not the gpu_blacklist, it refers
to:
> 
> 
> gpu_driver_bug_list_json.cc:52-57:
>     {
>       "id": 3,
>       "features": [
>         "set_texture_filter_before_generating_mipmap"
>       ]
>     },

FTR, here is the log since the code review is too stupid to parse the link
properly:

GpuFeatureTest.Canvas2DAllowed: 
[1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list
match for rule #64 in gpu_blacklist.
[1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list
match for rule #65 in gpu_blacklist.
[1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list
match for rule #3 in gpu_driver_bug_list.
[1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list
match for rule #4 in gpu_driver_bug_list.
[1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list
match for rule #17 in gpu_driver_bug_list.
[1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list
match for rule #26 in gpu_driver_bug_list.
[1636:5464:1010/153734:2795640:ERROR:chrome_views_delegate.cc(158)] NOT
IMPLEMENTED
[1636:5464:1010/153734:2795640:ERROR:desktop_root_window_host_win.cc(690)] NOT
IMPLEMENTED
test\gpu\gpu_feature_browsertest.cc(141): error: Expected: (event_count) > (0U),
actual: 0 vs 0

[Zhenyao Mo, 2013-10-11 22:33:10]

On 2013/10/11 22:25:31, gab wrote:
> On 2013/10/11 22:24:36, gab wrote:
> > On 2013/10/11 22:20:35, gab wrote:
> > > On 2013/10/11 21:38:27, Zhenyao Mo wrote:
> > > > On 2013/10/11 20:05:32, gab wrote:
> > > > > On 2013/10/11 19:38:27, Zhenyao Mo wrote:
> > > > > > On 2013/10/11 16:21:55, vangelis wrote:
> > > > > > > On 2013/10/11 13:38:53, gab wrote:
> > > > > > > > On 2013/10/11 04:51:14, vangelis wrote:
> > > > > > > > > On 2013/10/10 23:40:03, gab wrote:
> > > > > > > > > > Note control list matches are as follows:
> > > > > > > > > > 
> > > > > > > > > > XP:
> > > > > > > > > > gpu_blacklist #41 (flash_3d), 64(accelerated_video_decode),
> > > > > > > > > > 67(accelerated_2d_canvas), 79(force_compositing_mode) 
> > > > > > > > > > gpu_driver_bug_list
> > > #3(set_texture_filter_before_generating_mipmap),
> > > > > > > > > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost),
> > > > > > 26(disable_d3d11)
> > > > > > > > > > 
> > > > > > > > > > VISTA:
> > > > > > > > > > gpu_blaclist #64(accelerated_video_decode), #65(flash_3d,
> > > > > flash_stage3d,
> > > > > > > > > > force_compositing_mode)
> > > > > > > > > > gpu_driver_bug_list
> > > #3(set_texture_filter_before_generating_mipmap),
> > > > > > > > > > 4(clear_alpha_in_readpixels), 17(exit_on_context_lost),
> > > > > > 26(disable_d3d11)
> > > > > > > > > > 
> > > > > > > > > > 
> > > > > > > > > > Playing with some of these locally will most likely reveal
> > what's
> > > > > > causing
> > > > > > > > > which
> > > > > > > > > > failures.
> > > > > > > > > > 
> > > > > > > > > > Or maybe some of you more experienced graphics folks know
what
> > > could
> > > > > > cause
> > > > > > > > > what
> > > > > > > > > > just by looking at it?
> > > > > > > > > > 
> > > > > > > > > > Cheers!
> > > > > > > > > > Gab
> > > > > > > > > 
> > > > > > > > > Gab, I would have to look at the individual tests to see
what's
> > > going
> > > > > on.
> > > > > > Is
> > > > > > > > > there a complete list of failures somewhere as I'm not seeing
> the
> > > bot
> > > > > > > results.
> > > > > > > > > Thanks!
> > > > > > > > 
> > > > > > > > I listed the failures with detailed logs in the comments of this
> > > > reverted
> > > > > CL
> > > > > > > > (and added the log for the initial failure causing this revert
in
> > the
> > > CL
> > > > > > > > description).
> > > > > > > 
> > > > > > > Thanks, Gab! I don't think any of the blacklist entries would
affect
> > > > canvas
> > > > > > but
> > > > > > > it makes me wonder whether this particular test would ever succeed
> on
> > a
> > > > bot
> > > > > > that
> > > > > > > doesn't have a GPU. This Vista bot presumably is a VM that runs
the
> > > > > compositor
> > > > > > > in Software mode? Mo, any ideas? 
> > > > > > 
> > > > > > Gab, do you have a link to the failing vista bot?  I'd like to have
a
> > > look.
> > > > > 
> > > > > Failing Vista bot:
> > > > >
> > > >
> > >
> >
>
http://build.chromium.org/p/chromium.win/builders/Vista%252525252520Tests%252...
> > > > > 
> > > > > > 
> > > > > > >  
> > > > > > > 
> > > > > > > > 
> > > > > > > > I'm guessing it doesn't make sense to run
> > > > OutOfProcessPPAPITest.Graphics3D
> > > > > > on
> > > > > > > > machines where flash_3d is disabled, right?
> > > > > > > 
> > > > > > > Yes, I don't think we could run these tests if flash_3d is
disabled.
> > It
> > > > > looks
> > > > > > > like they have mostly been disabled for Windows altogether anyway?

> > > > > > > 
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>
https://code.google.com/p/chromium/codesearch#chromium/src/chrome/test/ppapi/...
> > > > 
> > > > The canvas isn't accelerated because accelerated compositor is turned
off
> > due
> > > to
> > > > it's a software renderer (blacklist entry 3).
> > > 
> > > Hmmm,
> > >
> >
>
http://build.chromium.org/p/chromium.win/builders/Vista%25252520Tests%2525252...
> > > doesn't suggest that rule #3 was applied (in fact rule #3 is a linux-only
> > rule).
> > 
> > You might have mistakenly read the log entry that mentions that the
> > gpu_driver_bug_list #3 was applied; this is not the gpu_blacklist, it refers
> to:
> > 
> > 
> > gpu_driver_bug_list_json.cc:52-57:
> >     {
> >       "id": 3,
> >       "features": [
> >         "set_texture_filter_before_generating_mipmap"
> >       ]
> >     },
> 
> FTR, here is the log since the code review is too stupid to parse the link
> properly:
> 
> GpuFeatureTest.Canvas2DAllowed: 
> [1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list
> match for rule #64 in gpu_blacklist.
> [1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list
> match for rule #65 in gpu_blacklist.
> [1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list
> match for rule #3 in gpu_driver_bug_list.
> [1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list
> match for rule #4 in gpu_driver_bug_list.
> [1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list
> match for rule #17 in gpu_driver_bug_list.
> [1636:5464:1010/153734:2795609:INFO:gpu_control_list.cc(1038)] Control list
> match for rule #26 in gpu_driver_bug_list.
> [1636:5464:1010/153734:2795640:ERROR:chrome_views_delegate.cc(158)] NOT
> IMPLEMENTED
> [1636:5464:1010/153734:2795640:ERROR:desktop_root_window_host_win.cc(690)] NOT
> IMPLEMENTED
> test\gpu\gpu_feature_browsertest.cc(141): error: Expected: (event_count) >
(0U),
> actual: 0 vs 0

You are right. Sorry about the noise.

Then I think the reason might be GPU process just failed to launch on Vista bot
with software renderer.

If that's the only bot that prevents you from landing this, you should just turn
the test off on vista.