Open Bug 1157576 Opened 9 years ago Updated 2 years ago

Intermittent browser_bug706743.js | Test timed out

Categories

(Core :: Layout, defect, P3)

x86
Linux
defect

Tracking

()

Tracking Status
firefox40 --- disabled
firefox41 --- disabled

People

(Reporter: philor, Unassigned)

References

Details

(Keywords: intermittent-failure, leave-open, Whiteboard: [stockwell disabled])

+++ This bug was initially created as a clone of Bug #1141549 +++
Probably should disable this one on e10s again for now.
OS: Windows XP → Linux
Something in the last day or two has caused this test to start failing on non-e10s at a high rate. Before June 15 it only failed a few times on June 3 inbound (suggesting a patch that was backed out) and then a few times starting June 10 on fx-team.

It feels like there is a regressing bug here, so let's see if we can find it. Joel, can you help here?

Note that the test has always failed frequently on e10s. It should be disabled there.
Flags: needinfo?(jmaher)
my first attempt at bisecting this:
https://treeherder.mozilla.org/#/jobs?repo=mozilla-central&fromchange=a7913ddbdd70&tochange=bd930c91d042&filter-searchStr=Ubuntu%20VM%2012.04%20mozilla-central%20opt%20test%20mochitest-browser-chrome-3

we will see if there are patterns there, and can then narrow down to the integration branch.  Likewise we could do this on debug or ASAN, I chose opt as opt runs faster.
Flags: needinfo?(jmaher)
and the winner is:
https://hg.mozilla.org/integration/fx-team/rev/fd59f91fb8c9

I don't see an easy correlation between that revision and the test file:
https://dxr.mozilla.org/mozilla-central/source/layout/xul/test/browser_bug706743.js?from=browser_bug706743.js&case=true#1

since this is in run-by-dir mode, there should be no bleedthrough in the other tests.

:Gijs, could you take a look and see if anything pops out here?
Flags: needinfo?(gijskruitbosch+bugs)
The lack of layout flush on page loads will be influencing this, but that change is a good thing and points to something in the test being broken since before this landed (ie backing it out wouldn't help anyone much). The original cause might have hidden by bug 1160775 landing.

FWIW, it's almost impossible to see what's actually going wrong here because the test logs no information messages at all, so it's not clear where we're timing out (early, I guess?). I would suggest disabling on e10s, requesting the complete log within the test, and adding some info() statements to it to figure out what's breaking.
Flags: needinfo?(gijskruitbosch+bugs)
Looks like this has been happening since bug 1155355 ( https://hg.mozilla.org/mozilla-central/rev/c36ce7aed6d9 ) enabled this for e10s testing.
Blocks: 1155355
Flags: needinfo?(enndeakin)
(Oops, missed some of the other comments, but maybe the non-e10s failure should be a separate bug.)
Disabled on e10s per comment #293.
Keywords: leave-open
It's too awful on Linux non-e10s to leave enabled. Disabling there as well.
https://hg.mozilla.org/integration/fx-team/rev/ff40745e2aee
Whiteboard: [test disabled on linux and e10s][leave open]
I suspect this was actually a permafail on linux opt just before it got disabled:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=df22362dc20f

It's possible my patch made it permafail, but more likely it was something in:
https://hg.mozilla.org/mozilla-central/pushloghtml?changeset=099d6cd6725e

I'll retrigger it on m-c a couple times to confirm.
Flags: needinfo?(enndeakin)
Bulk assigning P3 to all open intermittent bugs without a priority set in Firefox components per bug 1298978.
Priority: -- → P3
This test was skipped on (linux || e10s) - comment 421.

Bug 1255235 enabled this on e10s over a year ago, so now skipped on (linux).

In recent months, there have been ongoing failures on Windows 8, e10s. These were generally infrequent, Windows 8 Debug, e10s failures, until May 24

https://treeherder.mozilla.org/#/jobs?repo=autoland&filter-searchStr=windows%20x64%20browser-chrome&tochange=c39826d3493f20dca39f41d2e164149e3dc17ebe&fromchange=3676568d6311b37bae297eee8685cd4b463acc89 suggests this became more frequent with the change for bug 1359733...but perhaps that is not significant.


:mrbkap - Since you enabled this on e10s (bug 1255235), pinging you to see if you want to investigate. If not, would it be okay to go back to skipping on (linux || e10s)? Maybe just (linux || (windows && e10s))??
Flags: needinfo?(mrbkap)
See Also: → 1255235
I'm fine going back to skipping this test. I don't have a setup that I could debug this on (and am not personally invested in this test).
Flags: needinfo?(mrbkap)
Pushed by gbrown@mozilla.com:
https://hg.mozilla.org/integration/mozilla-inbound/rev/9421074e0074
Skip test browser_bug706743.js on e10s, again, for ongoing failures; r=me,test-only
Whiteboard: [test disabled on linux and e10s][leave open] → [stockwell disabled] [leave open]
Whiteboard: [stockwell disabled] [leave open] → [stockwell disabled]
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.