cc_unittests if flaky on windows-7-sp1 |
|||||||||
Issue descriptionBeen flakily flaking for a while now: https://build.chromium.org/p/chromium.win/builders/Win7%20Tests%20%28dbg%29%281%29?numbuilds=200 A few different errors but recent ones are all around: LayerTreeHostTestGpuRasterDeviceSizeChanged.RunSingleThread_DelegatingRenderer LayerTreeHostTestGpuRasterDeviceSizeChanged.RunSingleThread_DirectRenderer with stack traces like: cc::SurfaceLayer::HasDrawableContent [0x06374DB6+987781] cc::SurfaceLayer::HasDrawableContent [0x06378C59+1003816] testing::internal::FunctionMockerBase<void __cdecl(unsigned int,int,int,unsigned int,unsigned int)>::ClearDefaultActionsLocked [0x00FFE0F2+4370] cc::LayerTreeHostImplForTesting::CommitComplete [0x012FA786+102] cc::SurfaceLayer::HasDrawableContent [0x067FB430+5732607] cc::SurfaceLayer::HasDrawableContent [0x067FE1F7+5744326] cc::SingleThreadProxyForTest::ScheduledActionCommit [0x01301FEA+26] cc::SurfaceLayer::HasDrawableContent [0x0661FBA4+3784819] (...) @jaydasika who most recently touched this test file @loyso who most recently touched surface_layer.cc can you please help figure out what's going on or triage to someone who will? These tests will otherwise have to be disabled.. :-( Thanks!
,
Mar 22 2016
,
Mar 22 2016
Sample build: https://build.chromium.org/p/chromium.win/builders/Win7%20Tests%20%28dbg%29%281%29/builds/47100
,
Mar 22 2016
,
Mar 22 2016
,
Mar 22 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/cc4e6e973649401d608c5ffd4b51f286f46aea3e commit cc4e6e973649401d608c5ffd4b51f286f46aea3e Author: dgozman <dgozman@chromium.org> Date: Tue Mar 22 21:00:13 2016 Disable LayerTreeHostTestGpuRasterDeviceSizeChanged test on win. BUG= 596880 TBR=jaydasika@chromium.org NOTRY=true CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel Review URL: https://codereview.chromium.org/1821633006 Cr-Commit-Position: refs/heads/master@{#382677} [modify] https://crrev.com/cc4e6e973649401d608c5ffd4b51f286f46aea3e/cc/trees/layer_tree_host_unittest.cc
,
Mar 22 2016
My change to that test file has nothing to do with the flakes but I think I know the reason for those flakes. That test is trying to access pending tree and for single thread case, we won't have a pending tree. Its only a flake and doesn't crash always as pending tree is accessed only after 2 commits(so it depends on whether the test is ended before that or not). So, should we just make this test run only for multi-threaded case ? +cc reviewers of the CL that added that test(crrev.com/1024633002).
,
Mar 22 2016
Thanks for the analysis. Yes that sounds right, with a comment on the MULTI_THREAD macro about why it's not testing single thread. It sounds like the test is flakily even testing anything correctly then :S
,
Mar 22 2016
@jaydasika would you mind doing a patch as you suggested?
,
Mar 22 2016
That's a bizarre test end condition. It definitely seems like it's trying to end the test too soon, when most of the conditions are in commit complete. I do think this should probably be threaded only, as it's trying to compare pending and active trees. However, this should probably also be deflaked by moving the EndTest into the commit function, I think?
,
Mar 22 2016
I can do the patch.
,
Mar 22 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/66500aad5711bd511a41d5fdb2521fa865888911 commit 66500aad5711bd511a41d5fdb2521fa865888911 Author: jaydasika <jaydasika@chromium.org> Date: Tue Mar 22 23:53:09 2016 Enable LayerTreeHostTestGpuRasterDeviceSizeChanged test on win And make it run only for multi-threaded case as it relies on pending tree. BUG= 596880 CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel Review URL: https://codereview.chromium.org/1827643002 Cr-Commit-Position: refs/heads/master@{#382736} [modify] https://crrev.com/66500aad5711bd511a41d5fdb2521fa865888911/cc/trees/layer_tree_host_unittest.cc
,
Mar 22 2016
,
Mar 28 2016
Thanks! |
|||||||||
►
Sign in to add a comment |
|||||||||
Comment 1 by gab@chromium.org
, Mar 22 2016