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

Issue 601821 link

Starred by 1 user

Issue metadata

Status: Fixed
Owner:
Closed: Apr 2016
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Windows
Pri: 1
Type: Bug



Sign in to add a comment

ANGLE CQ broken because of failing webgl test(s)

Project Member Reported by jmad...@chromium.org, Apr 8 2016

Issue description

Probably from the recent WebGL roll -- zmo, since you did the roll, can you look at this? This is fairly high priority as it is blocking all ANGLE CLs.

Looking at the try waterfall: https://build.chromium.org/p/tryserver.chromium.angle/waterfall

Specifically the windows debug:
https://build.chromium.org/p/tryserver.chromium.angle/builders/win_angle_dbg_ng

first suspicious failure:
https://build.chromium.org/p/tryserver.chromium.angle/builders/win_angle_dbg_ng/builds/1026

this fails webgl2_conformance_tests on NVIDIA GPU on Windows (with patch) on Windows-2008ServerR2-SP1, possibly due to a new ASSERT.

 

Comment 1 by kbr@chromium.org, Apr 9 2016

Cc: nedngu...@google.com
Components: Tests>Telemetry
Owner: jmad...@chromium.org
All of the Windows Debug bots on the chromium.gpu.fyi waterfall seem to be failing the WebGL 2.0 conformance tests:

https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Debug%20(NVIDIA)
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win8%20Debug%20(NVIDIA)
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Debug%20(ATI)
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Debug%20(New%20Intel)

The failure mode is a timeout:

Expected exception while running WebglConformance.deqp_data_gles3_shaders_linkage

Traceback (most recent call last):
  RunStoryWithRetries at content\test\gpu\gpu_tests\gpu_test_base.py:72
    super(cls, shared_page_state).RunStory(results)
  RunStory at third_party\catapult\telemetry\telemetry\page\shared_page_state.py:336
    self._current_page.Run(self)
  Run at third_party\catapult\telemetry\telemetry\page\__init__.py:84
    shared_state.page_test.RunNavigateSteps(self, current_tab)
  RunNavigateSteps at third_party\catapult\telemetry\telemetry\page\page_test.py:182
    page.RunNavigateSteps(action_runner)
  RunNavigateSteps at content\test\gpu\gpu_tests\webgl_conformance.py:143
    'webglTestHarness._finished', timeout_in_seconds=300)
  WaitForJavaScriptCondition at third_party\catapult\telemetry\telemetry\page\action_runner.py:186
    self._tab.WaitForJavaScriptExpression(condition, timeout_in_seconds)
  WaitForJavaScriptExpression at third_party\catapult\telemetry\telemetry\internal\browser\web_contents.py:111
    e.message + '\n' + debug_message)
TimeoutException: 
Exception thrown when trying to capture console output: TimeoutException('',)

Locals:
  IsJavaScriptExpressionTrue : <function IsJavaScriptExpressionTrue at 0x031FBDF0>
  debug_message              : "Exception thrown when trying to capture console output: TimeoutException('',)"
  e                          : TimeoutException('',)
  expr                       : 'webglTestHarness._finished'
  timeout                    : 300

[1868:3960:0408/215843:WARNING:histogram_base.cc(131)] 2 histograms were created before reporting was enabled.
+--------------------------------------------------------------------------------+
| End of shard 0  Pending: 541.4s  Duration: 1992.8s  Bot: build124-m1  Exit: -1 |
+--------------------------------------------------------------------------------+


I'm not 100% sure what is going on but it looks like this particular test may be timing out, and Telemetry may not be handling that timeout correctly, causing the entire test suite to abort.

Jamie, can you please try reproducing this behavior locally in a debug build on Windows? If it's necessary to skip a test in Debug, let's add it to webgl2_conformance_expectations.py.

Project Member

Comment 2 by bugdroid1@chromium.org, Apr 11 2016

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/86e7030b9c0bd7c13661fa3058e21235e1c4bf74

commit 86e7030b9c0bd7c13661fa3058e21235e1c4bf74
Author: jmadill <jmadill@chromium.org>
Date: Mon Apr 11 18:41:49 2016

gpu: Skip the dEQP linkage GLES3 test on Windows Debug.

Seems to be causing an assert failure, unknown as to why.

BUG= 601821 
TBR=kbr@chromium.org,zmo@chromium.org
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

Review URL: https://codereview.chromium.org/1880543002

Cr-Commit-Position: refs/heads/master@{#386426}

[modify] https://crrev.com/86e7030b9c0bd7c13661fa3058e21235e1c4bf74/content/test/gpu/gpu_tests/webgl2_conformance_expectations.py

Status: Fixed (was: Assigned)
ANGLE CQ is fixed, regression remains for unknown reasons - unlikely it's ANGLE.

Comment 4 by zmo@chromium.org, Apr 11 2016

It's not a regression, but I enabled some tests for WebGL 2.

This should be a bug in ANGLE all the time.  Enabling this test triggers this bug.
Ken: I think it's weird that telemetry fails to run subsequent test in this case. I don't have cycle to look into this right now, but this is something to watch out for.

Sign in to add a comment