New issue
Advanced search Search tips

Issue 769344 link

Starred by 3 users

Issue metadata

Status: Fixed
Owner:
Closed: Apr 2018
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug

Blocking:
issue 703263



Sign in to add a comment

LayerTreeHostPictureTestTwinLayer is flaky

Project Member Reported by danakj@chromium.org, Sep 27 2017

Issue description

[ RUN      ] LayerTreeHostPictureTestTwinLayer.RunMultiThread_DelegatingRenderer
../../cc/trees/layer_tree_host_unittest_picture.cc:136: Failure
      Expected: 5
To be equal to: activates_
      Which is: 4
[  FAILED  ] LayerTreeHostPictureTestTwinLayer.RunMultiThread_DelegatingRenderer (20 ms)

Somehow the test is exiting with fewer activates that anticipated?

https://cs.chromium.org/chromium/src/cc/trees/layer_tree_host_unittest_picture.cc?rcl=df35ca0c4c57ce22fb03310bcfb2125d983e6ee9&l=136
 

Comment 1 by danakj@chromium.org, Nov 15 2017

Blocking: 703263

Comment 2 by xing...@intel.com, Nov 27 2017

I tried this case 60 times (Ubuntu 16.04), finally got one fail.
How about add one --test-launcher-retry-till-fail flag, case will keep retrying till fail?  This would be helpful for flaky case.
danakj@ shared this script to retry a command till it fails:

#!/bin/bash
COUNTER=0
until [ $? -ne 0 ]; do
    let COUNTER+=1
    echo "fail: attempt #$COUNTER"
    "$@"
done

Usage: fail out/Release/cc_unittests --gtest_filter="LayerTreeHostPictureTestTwinLayer.RunMultiThread_DelegatingRenderer"
Components: Internals>Compositing
Status: Available (was: Untriaged)

Comment 8 by danakj@chromium.org, Apr 25 2018

Owner: danakj@chromium.org
Status: Started (was: Available)
Project Member

Comment 9 by bugdroid1@chromium.org, Apr 26 2018

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/155ed73f14c7b9fa58d366197fe335d06b9d5775

commit 155ed73f14c7b9fa58d366197fe335d06b9d5775
Author: danakj <danakj@chromium.org>
Date: Thu Apr 26 18:33:43 2018

Deflake LayerTreeHostPictureTestTwinLayer test.

The test was doing EndTest() on the main thread after the 5th
DidCommit(), but then verifying the number of /activations/ that
happened, which was incremented on the compositor thread in
DidActivateTreeOnThread(). Since the DidCommit() races with the
DidActivateTreeOnThread(), then number of reported activations
after the test ended could be wrong.

Fix this by moving the EndTest() call to the compositor thread, after
the 5th activation.

R=enne@chromium.org

Bug:  769344 
Cq-Include-Trybots: luci.chromium.try:android_optional_gpu_tests_rel;master.tryserver.blink:linux_trusty_blink_rel
Change-Id: Id78f72b9a3d7a7a5ae140dcca74e396e72114a2e
Reviewed-on: https://chromium-review.googlesource.com/1028519
Commit-Queue: enne <enne@chromium.org>
Reviewed-by: enne <enne@chromium.org>
Cr-Commit-Position: refs/heads/master@{#554093}
[modify] https://crrev.com/155ed73f14c7b9fa58d366197fe335d06b9d5775/cc/trees/layer_tree_host_unittest_picture.cc

Status: Fixed (was: Started)

Sign in to add a comment