RESOLVED WONTFIX 72134
[chromium] WorkerTest flakiness 
https://bugs.webkit.org/show_bug.cgi?id=72134
Summary [chromium] WorkerTest flakiness
Sadrul Habib Chowdhury
Reported 2011-11-11 08:37:36 PST
The worker tests have been flaking in different platforms for a while. This is a speculative fix.
Attachments
patch (4.71 KB, patch)
2011-11-11 08:38 PST, Sadrul Habib Chowdhury 		levin: review-
DetailsFormatted DiffDiff
patch (6.22 KB, patch)
2011-11-11 09:09 PST, Sadrul Habib Chowdhury 		no flags
DetailsFormatted DiffDiff
Show Obsolete (1) View All Add attachment proposed patch, testcase, etc.
Sadrul Habib Chowdhury
Comment 1 2011-11-11 08:38:52 PST
Created attachment 114705 [details] patch
Sadrul Habib Chowdhury
Comment 2 2011-11-11 08:43:14 PST
If this does fix the flakiness, then presumably similar fixes can be applied to a number of other worker tests.
David Levin
Comment 3 2011-11-11 08:52:53 PST
Comment on attachment 114705 [details] patch The patch doesn't seem to apply (purple bots) and this needs a changelog (which would hopefully explain why this is thought to possibly fix the problem).
Sadrul Habib Chowdhury
Comment 4 2011-11-11 09:09:27 PST
Created attachment 114713 [details] patch I created the diff from the chrome tree, so it didn't apply correctly. I have prepared this new patch from the webkit tree, and included a description in the ChangeLog. (a sample log from a failure: http://build.chromium.org/p/chromium/builders/Linux%20Tests%20%28dbg%29%28shared%29/builds/3742/steps/ui_tests/logs/WorkerCall)
Dmitry Lomov
Comment 5 2011-11-11 09:44:16 PST
Cool - really sounds plausible. I wonder why we haven't seen these before.
Tony Chang
Comment 6 2011-11-11 10:22:37 PST
Comment on attachment 114713 [details] patch Seem worth trying. If this does fix the flakiness, it sounds like it's a bug in our ui_test test harness. We should try to fix the test harness or people will add new tests that are flaky in the same way.
Tony Chang
Comment 7 2011-11-11 10:32:08 PST
CQ appears to be down. I will try to land manually.
Tony Chang
Comment 8 2011-11-11 10:35:28 PST
Committed r99990: <http://trac.webkit.org/changeset/99990>
Tony Chang
Comment 9 2011-11-11 10:36:06 PST
Comment on attachment 114713 [details] patch I fixed the formatting of the ChangeLog and the indenting before landing.
Andrew Wilson
Comment 10 2011-11-11 10:46:42 PST
I don't think this change is bad, but if we ever have a case where: <p>Some text</p> <script>...some javascript...</script> executes the javascript before the HTML is populated in the DOM, then that indicates a fundamental issue with our HTML rendering that should be fixed, not worked around. So, yeah, if this "speculative fix" actually fixes the flakiness, then we really need to log a separate bug to address that behavior.
Dmitry Lomov
Comment 11 2011-11-11 11:33:29 PST
(In reply to comment #10) > I don't think this change is bad, but if we ever have a case where: > > <p>Some text</p> > <script>...some javascript...</script> > > executes the javascript before the HTML is populated in the DOM, then that indicates a fundamental issue with our HTML rendering that should be fixed, not worked around. > > So, yeah, if this "speculative fix" actually fixes the flakiness, then we really need to log a separate bug to address that behavior. Right, it must be a ui_test harness bug, as Tony suggests.
Ojan Vafai
Comment 12 2011-11-11 14:20:51 PST
Reverted r99990 for reason: Caused worker tests to fail on the chromium bots. Committed r100023: <http://trac.webkit.org/changeset/100023>
Stephen Chenney
Comment 14 2013-04-09 16:28:25 PDT
Marking test failures as WontFix. Bug is still accessible and recording in TestExpectations.
Note You need to log in before you can comment on or make changes to this bug.