New issue
Advanced search Search tips
Note: Color blocks (like █ or █) mean that a user may not be available. Tooltip shows the reason.

Issue 858907 link

Starred by 4 users
Status: Fixed
Owner:
█ kbr@chromium.org
OOO until 2019-01-24
Closed: Aug 24
Components:
EstimatedDays: ----
NextAction: ----
OS: Mac
Pri: 2
Type: Bug

Blocked on:
issue 575305
Blocking:
issue 832886
issue 851213
issue 871749
issue 879423
issue 882103
issue 883194



Sign in to add a comment

ContextLost_WebGLUnblockedAfterUserInitiatedReload flake seen on macOS with Intel GPU

Project Member Reported by kbr@chromium.org, Jun 29 2018 
Seen in this tryjob:
https://ci.chromium.org/p/chromium/builders/luci.chromium.try/mac_chromium_rel_ng/83408

Failing shard:
https://chromium-swarm.appspot.com/task?id=3e605020329ffb10&refresh=10&show_raw=1

Log excerpt:

[7/8] gpu_tests.context_lost_integration_test.ContextLostIntegrationTest.ContextLost_WebGLUnblockedAfterUserInitiatedReload failed unexpectedly 46.9539s:
...
  Traceback (most recent call last):
    _RunGpuTest at content/test/gpu/gpu_tests/gpu_integration_test.py:132
      self.RunActualGpuTest(url, *args)
    RunActualGpuTest at content/test/gpu/gpu_tests/context_lost_integration_test.py:93
      getattr(self, test_name)(test_path)
    _ContextLost_WebGLUnblockedAfterUserInitiatedReload at content/test/gpu/gpu_tests/context_lost_integration_test.py:327
      self.fail('WebGL should have been blocked after a context loss')
    fail at .swarming_module/lib/python2.7/unittest/case.py:410
      raise self.failureException(msg)
  AssertionError: WebGL should have been blocked after a context loss
  
  Locals:
    msg : 'WebGL should have been blocked after a context loss'



Looking at the test, maybe there's a race condition in blocking of WebGL contexts. Seems possible that the context lost event may be delivered when the domain hasn't been forbidden yet from creating new WebGL contexts.

Comment 2 by kbr@chromium.org, Jun 29 2018

Labels: OS-Mac
Project Member

Comment 3 by bugdroid1@chromium.org, Aug 7 

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/3689f3e57b85ca683cc6b61841715fbd0c6af559

commit 3689f3e57b85ca683cc6b61841715fbd0c6af559
Author: Kenneth Russell <kbr@chromium.org>
Date: Tue Aug 07 00:45:40 2018

Add crashGpuProcess() to gpu_benchmarking_extension.

Most of the various Telemetry-based GPU tests which test context loss
do so by creating a new tab, navigating it to chrome://gpucrash, and
switching back to the original tab. This tab manipulation is
introducing too many variables, and unreliability, into these tests.

Instead, allow the already-privileged gpu_benchmarking_extension to
crash the GPU process.

Bug:  858907 
Cq-Include-Trybots: luci.chromium.try:android_optional_gpu_tests_rel;luci.chromium.try:linux_optional_gpu_tests_rel;luci.chromium.try:mac_optional_gpu_tests_rel;luci.chromium.try:win_optional_gpu_tests_rel
Change-Id: I174694c359d253e5b62f50f8e48cda0129731f1e
Reviewed-on: https://chromium-review.googlesource.com/1161558
Reviewed-by: Antoine Labour <piman@chromium.org>
Reviewed-by: Daniel Cheng <dcheng@chromium.org>
Commit-Queue: Kenneth Russell <kbr@chromium.org>
Cr-Commit-Position: refs/heads/master@{#581068}
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/components/viz/service/gl/gpu_service_impl.cc
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/content/public/browser/gpu_utils.cc
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/content/renderer/gpu/gpu_benchmarking_extension.cc
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/content/renderer/gpu/gpu_benchmarking_extension.h
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/content/test/gpu/gpu_tests/context_lost_integration_test.py
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/gpu/config/gpu_preferences.h
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/gpu/config/gpu_preferences_unittest.cc
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/gpu/ipc/client/gpu_channel_host.cc
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/gpu/ipc/client/gpu_channel_host.h
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/gpu/ipc/common/gpu_messages.h
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/gpu/ipc/common/gpu_preferences.mojom
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/gpu/ipc/common/gpu_preferences_struct_traits.h
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/gpu/ipc/service/gpu_channel.cc
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/gpu/ipc/service/gpu_channel.h
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/ui/gl/gl_utils.cc
[modify] https://crrev.com/3689f3e57b85ca683cc6b61841715fbd0c6af559/ui/gl/gl_utils.h

Comment 4 by kbr@chromium.org, Aug 7

Labels: Hotlist-PixelWrangler
Status: Started (was: Assigned)
Need to watch this for a few days to see if context_lost is still flaky.

Comment 5 by kbr@chromium.org, Aug 7

Blocking: 871749

Comment 6 by kbr@chromium.org, Aug 24

Status: Fixed (was: Started)
There are no flakes of this test in the last 200 builds on either of these bots:

https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Mac%20Release%20%28Intel%29?limit=200
https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Mac%20FYI%20Release%20%28Intel%29?limit=200

Closing as fixed.

Comment 7 by kbr@chromium.org, Aug 24

Blocking: 832886

Comment 8 by kbr@chromium.org, Aug 29

Blocking: 851213

Comment 9 by kbr@chromium.org, Aug 31

Blocking: 879423

Comment 10 by kbr@chromium.org, Sep 10

Blocking: 882103

Comment 11 by kbr@chromium.org, Sep 14

Blocking: 883194
►

Sign in to add a comment