Linux TSAN as well: https://build.chromium.org/p/chromium.memory/builders/Linux%20TSan%20Tests/builds/14127 for example.
The reason for suspecting this CL is that the flakiness on Mac10.10 started on a build that only had this CL on it, and on no bot do we see this flake *before* this CL landed.
Unfortunately the logs from the failing builds inexplicably have no information. I'm going to revert the suspected CL to see if the flake disappears; if the flake is still there I'll reland it (or feel free to reland yourself if it's after EU working hours and you see that the flake is still there).
Lucas, can you look at this? One of your pointer lock tests became flaky when I landed OOPIF frame throttling.
Do you think the test just needs a call to WaitForChildFrameSurfaceReady()? I haven't been able to reproduce the flake locally so it's hard to verify.
That seems plausible. If we get to here https://cs.chromium.org/chromium/src/content/browser/pointer_lock_browsertest.cc?l=257 before the child frame has submitted a frame the test would fail. And there's nothing before that that would guarantee that the child has submitted a frame.
The only thing that I'm a bit worried is that the test times out in ExecuteScriptAndExtractInt(). This happens immediately after MainThreadFrameObserver, which implies the rendere was/is alive at the time. I'm not sure why ExecuteScript would fail.
Comment 1 by tasak@google.com
, Nov 13 2017