Many performance tests are timing out on IsJavaScriptExpressionTrue |
|||
Issue descriptionA recent bisect of mine failed because of various tests flaking with this same timeout: https://build.chromium.org/p/tryserver.chromium.perf/builders/winx64_10_perf_bisect/builds/1388 A casual search reveals at least a half-dozen bugs filed containing a flaky timeout in IsJavaScriptExpressionTrue: * Bug 692663 and bug 750147 (Windows) * Bug 748528 and bug 728081 (Android) * Bug 718427 (Linux) * Bug 728576 (Mac) Due to the variety of platforms and tests it's hard for me to tell if these all have the same underlying issue, or if it's just a common symptom of other problems. If it's the latter, it would be great to expose more information so those other problems can be diagnosed.
,
Aug 8 2017
Telemetry should shows JS stack trace upon IsJavaScriptExpressionTrue failure. Generally this is a sign of test written in a way that's sensitive to timing. The best way to debug these is still trying to produce the test locally.
,
Aug 8 2017
I examined the logs for a few of the linked failures and I don't see any JS stack traces. For example in this build https://build.chromium.org/p/chromium.perf/builders/Android%20Nexus7v2%20Perf/builds/334 blink_perf.paint failed on appending-text.html, and the only other suspicious thing is another python trace. Same thing for this build: https://luci-milo.appspot.com/buildbot/chromium.perf/Mac%20Retina%20Perf/717 system_health.common_desktop failing on test browse:news:cnn@{'case': 'browse', 'group': 'news'}. Is this because the JS stack traces aren't being printed on the bots? Or is it something simple like the target webpage isn't loading in time?
,
Jan 16
,
Jan 16
|
|||
►
Sign in to add a comment |
|||
Comment 1 by sullivan@chromium.org
, Aug 8 2017