Bug 143379 – fast/fixed-layout/fixed-layout.html is flaky

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Alexey Proskuryakov]

fast/fixed-layout/fixed-layout.html is flaky on EWS and Mac bots, seems to always fail elsewhere, and even to break subsequent tests.

http://webkit-test-results.appspot.com/dashboards/flakiness_dashboard.html#tests=fast%2Ffixed-layout%2Ffixed-layout.html

[Alexey Proskuryakov]

See also http://trac.webkit.org/changeset/182311

Not sure if we should roll out <http://trac.webkit.org/changeset/182307>. I'm going to mark the test as flaky in root test expectations, keep it skipped where it's skipped, and wait to Tim to weigh in.

[Alexey Proskuryakov]

Added an expectation in r182319.

[Tim Horton]

run-webkit-tests --repeat-each=100 --fully-parallel fast/fixed-layout

reproduces the flakiness quite reliably for me

[Tim Horton]

Created attachment 250098 [details]
Patch

[Tim Horton]

http://trac.webkit.org/changeset/182339

[Alexey Proskuryakov]

On Windows, the test started to time out after this patch, and it seems to have started to crash on Gtk:

https://webkit-test-results.appspot.com/dashboards/flakiness_dashboard.html#showAllRuns=true&tests=fast%2Ffixed-layout%2Ffixed-layout.html

[Tim Horton]

(In reply to comment #6)
> On Windows, the test started to time out after this patch, and it seems to
> have started to crash on Gtk:
> 
> https://webkit-test-results.appspot.com/dashboards/flakiness_dashboard.
> html#showAllRuns=true&tests=fast%2Ffixed-layout%2Ffixed-layout.html

GTK makes "fixed layout" mean a lot more than it should, and in really bad ways (randomly setting prefs in random places, etc.), so I've mostly been disregarding them.  Should probably skip the test there. Windows is more interesting.