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

Issue 796024 link

Starred by 1 user

Issue metadata

Status: Available
Owner: ----
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Android
Pri: 2
Type: Bug



Sign in to add a comment

WebglConformance_conformance_canvas_draw_webgl_to_canvas_test flake on android_optional_gpu_tests_rel

Project Member Reported by kbr@chromium.org, Dec 19 2017

Issue description

In this tryjob:
https://ci.chromium.org/buildbot/tryserver.chromium.android/android_optional_gpu_tests_rel/15495

and this shard:
https://chromium-swarm.appspot.com/task?id=3a85e74a4a405810&refresh=10&show_raw=1

for this unrelated CL (a WebGL conformance roll):
https://chromium-review.googlesource.com/832861/

The test WebglConformance_conformance_canvas_draw_webgl_to_canvas_test flaked in a really weird way. See the results:

  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/webgl_conformance_integration_test.py:203
      getattr(self, test_name)(test_path, *args[1:])
    _RunConformanceTest at content/test/gpu/gpu_tests/webgl_conformance_integration_test.py:217
      self._CheckTestCompletion()
    _CheckTestCompletion at content/test/gpu/gpu_tests/webgl_conformance_integration_test.py:213
      self.fail(self._WebGLTestMessages(self.tab))
    fail at /usr/lib/python2.7/unittest/case.py:410
      raise self.failureException(msg)
  AssertionError: drawImage: Should be [255, 0, 0, 255]
  at (63, 30) expected: 255,0,0,255 was 255,128,0,255
  FAIL drawImage: Should be [255, 0, 0, 255]
  at (63, 30) expected: 255,0,0,255 was 255,128,0,255
  drawImage: Should be [64, 128, 192, 255]
  at (299, 12) expected: 64,128,192,255 was 64,0,191,255
  FAIL drawImage: Should be [64, 128, 192, 255]
  at (299, 12) expected: 64,128,192,255 was 64,0,191,255


The green channel is missing from the only canvas that's supposed to have a non-zero value for it, and one of the other canvases contains that green channel.

This is really weird. It looks almost like the canvases contain garbage, because incorrect synchronization wouldn't explain this -- these colors aren't generated by this test. junov@, xlai@, can you think of any explanation for this failure?

Going to retry the job without marking the test flaky, to see if it happens again. I didn't see further evidence of this on https://ci.chromium.org/buildbot/tryserver.chromium.android/android_optional_gpu_tests_rel/?limit=200 .

 

Comment 1 by junov@chromium.org, Jan 9 2018

Status: Available (was: Untriaged)
Could it be bad synchronization combined with texture recycling?
Cc: -junov@chromium.org

Sign in to add a comment