Closed Bug 1211084 Opened 5 years ago Closed 3 years ago

Intermittent e10s browser_async_remove_tab.js | closed tab saved

Categories

(Firefox :: Session Restore, defect, P5)

defect

Tracking

()

RESOLVED FIXED
Firefox 55
Tracking Status
e10s + ---
firefox52 --- wontfix
firefox-esr52 --- fixed
firefox53 --- fixed
firefox54 --- fixed
firefox55 --- fixed

People

(Reporter: philor, Assigned: mikedeboer, NeedInfo)

References

(Blocks 1 open bug)

Details

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

Attachments

(1 file)

Blocks: e10s-tests
tracking-e10s: --- → +
Intermittent e10s test failure
Priority: -- → P5
this test increased in frequency, do some investigation
Flags: needinfo?(jmaher)
Flags: needinfo?(jmaher)
Whiteboard: [stockwell needswork]
check out bug 1282990, same test, different failure.
See Also: → 1282990
doing a lot of retriggers didn't help me find a root cause :(  This seems to happen about 1/10 runs consistently.

:ttaubert, can you look at this test case and see if you can fix the errors here and in bug 1282990?
Flags: needinfo?(ttaubert)
(In reply to Joel Maher ( :jmaher) from comment #47)
> :ttaubert, can you look at this test case and see if you can fix the errors
> here and in bug 1282990?

Sorry, I don't work on this stuff or build Firefox these days anymore :( I wonder if Mike has time or can find someone to take a look.
Flags: needinfo?(ttaubert) → needinfo?(mdeboer)
_very_ short analysis: the two tasks where things fail both use the utility function `promiseNewLocationAndHistoryEntryReplaced()`, which is suspicious. I'm pretty sure the race is in there.

I'll add some logging and push that to try tomorrow.
Assignee: nobody → mdeboer
Status: NEW → ASSIGNED
Flags: needinfo?(mdeboer)
Between this bug and bug 1282990, this test case is responsible for 100+ failures/week :(  Can you increase the priority here?  right now this is failing on all platforms primarily debug e10s.

Mike, how is this coming along?
Flags: needinfo?(mdeboer)
:wlach, could you use your retrigger with logging options for this bug?
Flags: needinfo?(wlachance)
(In reply to Joel Maher ( :jmaher) from comment #54)
> :wlach, could you use your retrigger with logging options for this bug?

Tried this on linux32 debug, didn't reproduce the failure but got a pile of warnings about leaks:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=f9f1ae807d384f42912547621443e46db90b00f7&selectedJob=83758884
https://treeherder.mozilla.org/logviewer.html#?job_id=83758884&repo=try&lineNumber=12746

[task 2017-03-14T18:20:04.400538Z] 1375 ERROR TEST-UNEXPECTED-FAIL | browser/components/sessionstore/test/browser_async_remove_tab.js | leaked 16 window(s) until shutdown [url = about:blank]
[task 2017-03-14T18:20:04.402273Z] 1376 ERROR TEST-UNEXPECTED-FAIL | browser/components/sessionstore/test/browser_async_remove_tab.js | leaked 9 window(s) until shutdown [url = https://example.com/]
[task 2017-03-14T18:20:04.405039Z] TEST-INFO | browser/components/sessionstore/test/browser_async_remove_tab.js | windows(s) leaked: [pid = 1332] [serial = 224], [pid = 1332] [serial = 235], [pid = 1332] [serial = 241], [pid = 1332] [serial = 243], [pid = 1332] [serial = 231], [pid = 1332] [serial = 225], [pid = 1332] [serial = 234], [pid = 1332] [serial = 228], [pid = 1332] [serial = 221], [pid = 1332] [serial = 239], [pid = 1332] [serial = 230], [pid = 1332] [serial = 244], [pid = 1332] [serial = 237], [pid = 1332] [serial = 229], [pid = 1332] [serial = 226], [pid = 1332] [serial = 238], [pid = 1332] [serial = 233], [pid = 1332] [serial = 222], [pid = 1332] [serial = 236], [pid = 1332] [serial = 242], [pid = 1332] [serial = 227], [pid = 1332] [serial = 232], [pid = 1332] [serial = 240], [pid = 1332] [serial = 223], [pid = 1332] [serial = 245]
[task 2017-03-14T18:20:04.406702Z] 1377 ERROR TEST-UNEXPECTED-FAIL | browser/components/sessionstore/test/browser_async_remove_tab.js | leaked 5 docShell(s) until shutdown
[task 2017-03-14T18:20:04.408789Z] TEST-INFO | browser/components/sessionstore/test/browser_async_remove_tab.js | docShell(s) leaked: [pid = 1332] [id = {be7bafd6-7e2a-4c93-951c-227716eb5fc3}], [pid = 1332] [id = {fc55d1cf-0d5f-456f-be74-e1a389d7ee44}], [pid = 1332] [id = {22397e39-2266-4fa4-8e8d-77e1d580d0ec}], [pid = 1332] [id = {31dcf68d-3c65-465b-9cfc-8b1ba546e0e5}], [pid = 1332] [id = {cc286fb7-b2c7-488f-a1f1-0bda7d3d25b0}]
Flags: needinfo?(wlachance)
(In reply to William Lachance (:wlach) (use needinfo!) from comment #55)
> (In reply to Joel Maher ( :jmaher) from comment #54)
> > :wlach, could you use your retrigger with logging options for this bug?
> 
> Tried this on linux32 debug, didn't reproduce the failure but got a pile of
> warnings about leaks:
> ...

I'm feeling rather uncertain that this is a problem with the test, as I'm seeing similar errors elsewhere. So take this with a grain of salt.
I'm feeling rather inclined to just disable it on e10s for now. To be sure, I'm going to ask Will if has time to look at this.
Flags: needinfo?(mdeboer) → needinfo?(wiwang)
between this bug and bug 1282990, we have 100+ failures in the last week- I would recommend leaving the ni? on the bug and disabling this until there is time to get a proper fix in!
Attachment #8849119 - Flags: review?(mdeboer)
Hi Mike! 
I am more than happy to look into this bug to improve test quality, from the function you mentioned in comment 50, and I am currently working on some bugs including a P1 bug of Quantum Flow, plus session store tier-1(ss-reliability) bugs; Would it be possible that I keep the needinfo and back to this bug once those bugs are done, or is there any people can work on this at the moment?

Thanks! :)
Attachment #8849119 - Flags: review?(mdeboer) → review+
Pushed by jmaher@mozilla.com:
https://hg.mozilla.org/integration/mozilla-inbound/rev/a1807a4fcc23
Intermittent e10s browser_async_remove_tab.js. disable on debug. r=mikedeboer
Whiteboard: [stockwell needswork] → [stockwell disabled]
https://hg.mozilla.org/mozilla-central/rev/a1807a4fcc23
Status: ASSIGNED → RESOLVED
Closed: 3 years ago
Resolution: --- → FIXED
Target Milestone: --- → Firefox 55
Was it a fluke?
Flags: needinfo?(mdeboer) → needinfo?(aryx.bugmail)
You need to log in before you can comment on or make changes to this bug.