New issue
Advanced search Search tips

Issue 713257 link

Starred by 1 user

Issue metadata

Status: Fixed
Owner:
Closed: May 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug

Blocking:
issue 703263



Sign in to add a comment

LayerTreeHostTestStartPageScaleAnimation.RunMultiThread_DelegatingRenderer is flaky

Project Member Reported by brettw@chromium.org, Apr 19 2017

Issue description

https://test-results.appspot.com/dashboards/flakiness_dashboard.html#testType=cc_unittests&tests=LayerTreeHostTestStartPageScaleAnimation.RunMultiThread_DelegatingRenderer

This test crashes and times out. Crash stack reported by Chromium Win7:

	std::unique_ptr<cc::LayerTreeHost,std::default_delete<cc::LayerTreeHost> >::get [0x006D749F+15]

	std::unique_ptr<cc::LayerTreeHost,std::default_delete<cc::LayerTreeHost> >::operator bool [0x0121B641+17]

	cc::LayerTreeTest::DestroyLayerTreeHost [0x0121D187+23]

	cc::LayerTreeTest::RunTest [0x0122025B+1195]

	base::internal::Invoker<base::internal::BindState<void (__thiscall cc::`anonymous namespace'::SurfaceLayerSwapPromise::*)(void),base::internal::UnretainedWrapper<cc::`anonymous namespace'::SurfaceLayerSwapPromise> >,void __cdecl(void)>::RunOnce [0x00782B86+518]

	testing::internal::HandleExceptionsInMethodIfSupported<testing::Test,void> [0x013053D4+52]

	testing::Test::Run [0x01320907+135]

	testing::TestInfo::Run [0x01320B8D+173]

	testing::TestCase::Run [0x01320A2F+191]

	testing::internal::UnitTestImpl::RunAllTests [0x01321015+661]

	testing::internal::HandleExceptionsInMethodIfSupported<testing::internal::UnitTestImpl,bool> [0x013054C4+52]

	testing::UnitTest::Run [0x01320D0F+207]

	RUN_ALL_TESTS [0x012A2AFF+15]

	base::TestSuite::Run [0x012A2C33+115]

	??$Invoke@PAVCCTestSuite@cc@@$$V@?$FunctorTraits@P8TestSuite@base@@AEHXZX@internal@base@@SAHP8TestSuite@2@AEHXZ$$QAPAVCCTestSuite@cc@@@Z [0x0105ED02+18]

	base::internal::InvokeHelper<0,int>::MakeItSo<int (__thiscall base::TestSuite::*const &)(void),cc::CCTestSuite *> [0x0105ED64+36]

	base::internal::Invoker<base::internal::BindState<int (__thiscall base::TestSuite::*)(void),base::internal::UnretainedWrapper<cc::CCTestSuite> >,int __cdecl(void)>::RunImpl<int (__thiscall base::TestSuite::*const &)(void),std::tuple<base::internal::Unreta [0x0105EDBB+59]

	base::internal::Invoker<base::internal::BindState<int (__thiscall base::TestSuite::*)(void),base::internal::UnretainedWrapper<cc::CCTestSuite> >,int __cdecl(void)>::Run [0x0105EFC4+36]

	base::Callback<int __cdecl(void),1,1>::Run [0x012ABEE1+33]

	base::LaunchUnitTests [0x012AB1F8+408]

	base::LaunchUnitTests [0x012AB031+81]

	main [0x0105F06C+92]
 

Comment 1 by brettw@chromium.org, Apr 19 2017

Labels: -Pri-3 Pri-2
This is also reported as being flaky under ASAN in issue 703263.

Comment 2 by brettw@chromium.org, Apr 19 2017

Blocking: 703263
Cc: danakj@chromium.org vmi...@chromium.org enne@chromium.org
This test is no longer crashing, suspecting because the fix for the ASAN flake. However it does time out.
Status: Available (was: Untriaged)

Comment 5 by danakj@chromium.org, May 10 2017

Owner: danakj@chromium.org
Status: Assigned (was: Available)
let's not leave flaky tests unowned

Comment 6 by danakj@chromium.org, May 10 2017

Status: Started (was: Assigned)
This test is flaky when sometimes a main frame occurs during/before the page scale animation where LayerTreeHost::ApplyViewportDeltas early outs instead of calling to the unit test. In this case, the source frame number increments and throws off the test.

Comment 7 by danakj@chromium.org, May 10 2017

In the failure case, active_tree_->page_scale_factor()->PullDeltaForMainThread() returns 1 instead of 1.25 for a BeginMainFrame (then 1.25 for another BeginMainFrame shortly after).

I am investigating why the BeginMainFrame after the PageScaleAnimation is flaky wrt running the PageScaleAnimation first.

Comment 8 by danakj@chromium.org, May 10 2017

Cc: sunn...@chromium.org
It looks like this is due to high latency mode, we're sending beginframes outside of impl frames. Some test runs take longer to fall into high latency mode than others, which I think must contribute to the timing of main frames relative to impl frames being semi-random.
Project Member

Comment 9 by bugdroid1@chromium.org, May 10 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/44afd4d12b169f43c3c0dade63486a4541bcc147

commit 44afd4d12b169f43c3c0dade63486a4541bcc147
Author: danakj <danakj@chromium.org>
Date: Wed May 10 22:27:04 2017

Fix flaky LayerTreeHostTestStartPageScaleAnimation test.

The test can get a main frame before the page scale animation starts,
which would be aborted. In this case the source frame number increments
twice before we see the page scale animation run.

R=vmpstr@chromium.org
BUG= 713257 
CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_trusty_blink_rel

Review-Url: https://codereview.chromium.org/2871323002
Cr-Commit-Position: refs/heads/master@{#470731}

[modify] https://crrev.com/44afd4d12b169f43c3c0dade63486a4541bcc147/cc/trees/layer_tree_host_unittest.cc

Status: Fixed (was: Started)

Sign in to add a comment