Bug 106415

Summary: REGRESSION(r139145): Multiple fast/workers tests fail
Product: WebKit Reporter: Stephanie Lewis <slewis>
Component: JavaScriptCoreAssignee: Oliver Hunt <oliver>
Status: RESOLVED WORKSFORME    
Severity: Critical CC: abarth, fpizlo, mark.lam, oliver, ossy, rniwa, slewis, timothy
Priority: P1 Keywords: InRadar
Version: 528+ (Nightly build)   
Hardware: Mac   
OS: Unspecified   
Bug Depends on:    
Bug Blocks: 79668, 98898    
Attachments:
Description Flags
Patch none

Stephanie Lewis
Reported 2013-01-08 21:23:35 PST
Due to http://trac.webkit.org/projects/webkit/changeset/139145 Failing on Mac WK1 and WK2, Release and Debug, Mountain Lion and Lion. The tests are timing out waiting for notifyDone and flaky. Every once in awhile they'll pass. [1/3] fast/workers/worker-close-more.html failed unexpectedly (text diff) [2/3] fast/workers/worker-document-leak.html failed unexpectedly (text diff) [3/3] fast/workers/worker-lifecycle.html failed unexpectedly (text diff)
Attachments
Patch (1.48 KB, patch)
2013-01-28 11:20 PST, Oliver Hunt 		no flags
DetailsFormatted DiffDiff
View All Add attachment proposed patch, testcase, etc.
Stephanie Lewis
Comment 1 2013-01-08 21:32:01 PST
Updated Expectations in http://trac.webkit.org/projects/webkit/changeset/139160
Csaba Osztrogonác
Comment 3 2013-01-08 22:32:14 PST
Similar flakiness on EFL, GTK and Qt. :-/
Csaba Osztrogonác
Comment 4 2013-01-08 22:42:22 PST
I skipped them on Qt - r139164 to paint the bots green. Please unskip them with the proper fix. (It is P1/critical bug, because it is a regression and tests fail with timeout)
Alexey Proskuryakov
Comment 5 2013-01-09 10:00:02 PST
<rdar://problem/12981718>
Csaba Osztrogonác
Comment 6 2013-01-13 01:44:24 PST
Are you working on fixing this serious regression?
Csaba Osztrogonác
Comment 7 2013-01-15 02:37:52 PST
ping?
Oliver Hunt
Comment 8 2013-01-15 09:21:05 PST
(In reply to comment #7) > ping? Eventually. I'm side tracked with other more important bugs for a couple of days. Could you do some analysis yourself? Maybe you'll find some obvious cause of the problem. Otherwise causing two layout tests to hang, but not having real world site breakage means that this can wait a few more days.
Csaba Osztrogonác
Comment 9 2013-01-23 16:07:10 PST
I checked this bug, it seems it isn't valid on ToT with Qt 64 bit release. I have no idea which revision fixed it and don't have time to bisect 1000+ revision. Unfortunately it's hard to reproduce, because these tests passed if you ran only fast/workers tests, but failed if you ran all fast tests. (on r139145)
Oliver Hunt
Comment 10 2013-01-23 16:11:58 PST
(In reply to comment #9) > I checked this bug, it seems it isn't valid on ToT with Qt 64 bit release. > I have no idea which revision fixed it and don't have time to bisect 1000+ > revision. Unfortunately it's hard to reproduce, because these tests passed > if you ran only fast/workers tests, but failed if you ran all fast tests. (on r139145) Thanks for checking. I'll look around and see if there's anything obviously stupid happening. It's possible that this was just a temporary thing before i fixed the massive gc bug i had introduced :-/
Oliver Hunt
Comment 11 2013-01-28 11:20:53 PST
Created attachment 185020 [details] Patch
Oliver Hunt
Comment 12 2013-01-28 11:21:17 PST
Committed r140981: <http://trac.webkit.org/changeset/140981>
Alexey Proskuryakov
Comment 13 2013-01-28 19:31:45 PST
Comment on attachment 185020 [details] Patch Clearing review flag from a landed patch.
Oliver Hunt
Comment 15 2013-01-28 20:56:49 PST
(In reply to comment #14) > These tests are still failing :( See http://build.webkit.org/results/Apple%20MountainLion%20Release%20WK2%20(Tests)/r141034%20(5309)/results.html goddammit
Csaba Osztrogonác
Comment 16 2013-01-28 21:41:01 PST
(In reply to comment #14) > These tests are still failing :( See http://build.webkit.org/results/Apple%20MountainLion%20Release%20WK2%20(Tests)/r141034%20(5309)/results.html Reopen, because it seems it is still valid bug.
Ryosuke Niwa
Comment 17 2013-01-28 22:04:25 PST
*** Bug 108144 has been marked as a duplicate of this bug. ***
Ryosuke Niwa
Comment 18 2013-01-28 22:05:38 PST
I think we need to add these failing expectations back.
Oliver Hunt
Comment 19 2013-01-29 08:51:39 PST
(In reply to comment #18) > I think we need to add these failing expectations back. yup. this makes no sense.
Ryosuke Niwa
Comment 20 2013-01-29 10:21:05 PST
This bug is causing Mac EWS bots to erroneously reject patches.
Oliver Hunt
Comment 21 2013-01-29 10:22:04 PST
(In reply to comment #20) > This bug is causing Mac EWS bots to erroneously reject patches. Sorry, i thought you'd said you were going to revert the patch, i'll be able to do so in half an hour
Oliver Hunt
Comment 22 2013-01-29 10:22:28 PST
(patch being the expectation patch)
Ryosuke Niwa
Comment 23 2013-01-29 10:23:45 PST
(In reply to comment #21) > (In reply to comment #20) > > This bug is causing Mac EWS bots to erroneously reject patches. > > Sorry, i thought you'd said you were going to revert the patch, i'll be able to do so in half an hour Done that in r141133.
Ryosuke Niwa
Comment 24 2013-01-29 10:26:42 PST
In the future, I recommend using http://test-results.appspot.com/dashboards/flakiness_dashboard.html#group=%40ToT%20-%20webkit.org&tests=fast%2Fworkers%2Fworker-close-more.html%2Cfast%2Fworkers%2Fworker-document-leak.html%2Cfast%2Fworkers%2Fworker-lifecycle.html It shows historical actual failure / pass status of these tests reported by bots. As you can see, these tests occasionally pass.
Ryosuke Niwa
Comment 25 2013-01-29 13:10:34 PST
*** Bug 108217 has been marked as a duplicate of this bug. ***
Timothy Hatcher
Comment 26 2013-03-29 11:56:24 PDT
These are also failing most of the time on Windows. I marked them as flaky in https://trac.webkit.org/r147232.
Alexey Proskuryakov
Comment 27 2013-09-26 17:02:25 PDT
Still timing out.
Alexey Proskuryakov
Comment 28 2015-08-29 00:21:21 PDT
Looks like r189009 took care of fast/workers/worker-close-more.html fast/workers/worker-document-leak.html has been passing for a long time now and fast/workers/worker-lifecycle.html too.
Note You need to log in before you can comment on or make changes to this bug.