1268319 - The responsive.html mochitest-dt tests are permaleaking the world on OSX debug e10s

Status: RESOLVED FIXED

(DevTools :: Responsive Design Mode, defect, P1)

Description

[Ryan VanderMeulen [:RyanVM]]

Try bisection confirms that bug 1250084 is the culprit. From IRC:
<mccr8> RyanVM: I'm going to guess this line is the culprit: exports.startup = startup;

Currently, OSX debug e10s m-dt is only running on the Ash branch. If you want to experiment on Try, the patch below will force e10s on for the regular test suites. Otherwise, local tests default to e10s-enabled these days, so hopefully it'll reproduce locally without any fiddling around if you have a Mac handy.
https://hg.mozilla.org/try/rev/2a1e8f59a80e3a376bf4cdd35fe65bd2275ab0a5

https://treeherder.mozilla.org/logviewer.html#?job_id=171081&repo=ash
TEST-UNEXPECTED-FAIL | leakcheck | default process: 4934659 bytes leaked (AnimationTimeline, AsyncLatencyLogger, AsyncStatement, BackstagePass, BoxObject, ...)

I did a Try push with Cycle Collector shutdown logging enabled as well. Here's a link to those logs:

Comment 1

[Ryan VanderMeulen [:RyanVM]]

(In reply to Ryan VanderMeulen [:RyanVM] from comment #0)
> I did a Try push with Cycle Collector shutdown logging enabled as well.
> Here's a link to those logs:

Now with 100% more CC logs:
https://treeherder.mozilla.org/logviewer.html#?job_id=20074427&repo=try
cc-edges.1744-5.log: http://mozilla-releng-blobs.s3.amazonaws.com/blobs/try/sha512/e6abe1488db71746d0d0f1a1b305a88e4bc53d5afa778d124e5d764efa4f445f5309afeba43ccb42806056daf6f012cf9fe97dc0bd8462d97b3bd0a9fe7b43df
cc-edges.1744-6.log: http://mozilla-releng-blobs.s3.amazonaws.com/blobs/try/sha512/e6ba2e8980f404a1fcbac6e2a260e17e09ef8cc10e8f106fafba0b05884f6f02884f68abd4a3a44d4a7b1c7dc2a53fa2e32cda9019f433e2fc6c0c7b4f699a7f

The other cc-edges.1744 logs look useless. Andrew, does either one of the above logs help confirm your suspicion?

Comment 2

[Matteo Ferretti [:zer0] [:matteo]]

I'm working on bug 1267278, it's likely that fixing that would fixes also this bug.

Comment 3

[Andrew McCreight [:mccr8]]

The leaking windows are being held alive by a XUL element in chrome://browser/content/browser.xul. The description in the CC log is "menuitem id='helpSafeMode'". Which does sound related to bug 1267278.

Comment 4

[Intermittent Failures Robot]

7 automation job failures were associated with this bug in the last 7 days.

Repository breakdown:
* ash: 7

Platform breakdown:
* osx-10-10: 7

For more details, see:
https://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=1268319&startday=2016-04-25&endday=2016-05-01&tree=all

Comment 5

[Ryan VanderMeulen [:RyanVM]]

I'm enabling OSX debug e10s tests in production very soon, so I'm disabling this directory for now.
https://hg.mozilla.org/integration/fx-team/rev/cecc6f0f0560

Comment 6

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/cecc6f0f0560

Comment 7

[Intermittent Failures Robot]

5 automation job failures were associated with this bug in the last 7 days.

Repository breakdown:
* ash: 5

Platform breakdown:
* osx-10-10: 5

For more details, see:
https://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=1268319&startday=2016-05-10&endday=2016-05-16&tree=all

Comment 8

[J. Ryan Stinnett [:jryans] (Use needinfo, replies may be slow)]

We believe this issue is fixed by bug 1267278.  The skip-if was removed there.

Please reopen if it is still present.