Closed Bug 1182072 Opened 9 years ago Closed 8 years ago

Intermittent e10s browser-test.js | Found an unexpected unknown window

Categories

(Firefox :: General, defect)

defect
Not set
normal

Tracking

()

RESOLVED FIXED
Firefox 46
Tracking Status
e10s + ---
firefox43 --- wontfix
firefox44 --- wontfix
firefox45 --- fixed
firefox46 --- fixed
firefox-esr38 --- wontfix

People

(Reporter: cbook, Assigned: Gijs)

References

(Blocks 1 open bug, )

Details

(Keywords: intermittent-failure)

Attachments

(2 files)

https://treeherder.mozilla.org/logviewer.html#?job_id=11520810&repo=mozilla-inbound

06:42:18 INFO - 2 INFO TEST-UNEXPECTED-FAIL | browser-test.js | Found an unexpected unknown window 

06:42:48 INFO - 67 INFO TEST-UNEXPECTED-FAIL | browser/base/content/test/plugins/browser_bug797677.js | Uncaught exception - at :0 - Error: operation not possible on dead CPOW
This is the #7 intermittent orange over the past 3 days.  It would be helpful to get this assigned.
Flags: needinfo?(dolske)
I'll bite. First thing will be figuring out what window this is. The patch I added will add URIs to the output for this test.
Flags: needinfo?(dolske)
(In reply to :Gijs Kruitbosch from comment #568)
> for this test

... harness check.
Assignee: nobody → gijskruitbosch+bugs
Comment on attachment 8702591 [details]
MozReview Request: Bug 1182072 - Add more diagnostic output to test failures caused by extra windows, r?jmaher

https://reviewboard.mozilla.org/r/29139/#review25953

lgtm
Attachment #8702591 - Flags: review+
Comment on attachment 8702591 [details]
MozReview Request: Bug 1182072 - Add more diagnostic output to test failures caused by extra windows, r?jmaher

https://reviewboard.mozilla.org/r/29139/#review25955

this is great, please consider adding extra debugging as well if you can think of it.  I have found getting the documentURI in some other debugging for talos hangs was not as useful as I thought.
Attachment #8702591 - Flags: review?(jmaher) → review+
Keywords: leave-open
	13:28:12 INFO - 2 INFO TEST-UNEXPECTED-FAIL | browser-test.js | Found an unexpected unknown window with document URI: chrome://gfxsanity/content/sanityparent.html

Seems like the simplest thing would be to either wait for that window to go away before starting the test, or to ignore it on exit.
If it helps any, this isn't actually intermittent - every single Windows e10s browser-chrome and devtools-chrome log shows this, it's just that we only see it and star it (sometimes also starring the real failure, sometimes not) when there is also another failure in the same run.
Summary: Intermittent browser-test.js | Found an unexpected unknown window | Uncaught exception - at :0 - Error: operation not possible on dead CPOW → Intermittent e10s browser-test.js | Found an unexpected unknown window
(In reply to Phil Ringnalda (:philor) from comment #575)
> If it helps any, this isn't actually intermittent - every single Windows
> e10s browser-chrome and devtools-chrome log shows this, it's just that we
> only see it and star it (sometimes also starring the real failure, sometimes
> not) when there is also another failure in the same run.

TBH, all the more reason to fix this - which I'm doing right now - though I wonder if there should be a separate bug about these failures not showing up for all the other test runs (ie, why isn't this perma-orange?)

Phil, any idea where the code lives that makes that determination?
Flags: needinfo?(philringnalda)
Matter of opinion where that fault lies: a mochitest run passed if it printed "Failed: 0", http://mxr.mozilla.org/mozilla-central/source/testing/mozharness/mozharness/mozilla/testing/unittest.py#104, but http://mxr.mozilla.org/mozilla-central/source/testing/mochitest/browser-test.js#322 wants to declare failures even when there is no currentTest and thus nothing to addResult to.
Flags: needinfo?(philringnalda)
try push:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=aa13c703b3de

I'll be curious to see how much orange it would cause to make these errors fatal... but there we are. :-)
Comment on attachment 8702907 [details]
MozReview Request: Bug 1182072 - deal with the graphics window that shows up on startup, r?jmaher

https://reviewboard.mozilla.org/r/29215/#review26013

just a small nit, lets see what try has to say.

::: testing/mochitest/browser-test.js:281
(Diff revision 1)
> +  waitForGraphicsTestWindowToBeGone(aCallback) {

the only feedback I have here is that we are hardcoding our initialization tests to be a specific sanityparent.html window- there could be other things.  I would rather not over engineer this now, but in the future we might have a need to do that.

::: testing/mochitest/browser-test.js:395
(Diff revision 1)
> -        this.dumper.structuredLogger.info("Failed:  " + failCount);
> +        this.dumper.structuredLogger.info("Failed:  " + (failCount + this.failuresFromInitialWindowState));

this doesn't read well.  It might be better to have a separate line with a comment, which is more like:
// Include failures prior to running the first test
failCount += this.failuresFromInitialWindowState
Attachment #8702907 - Flags: review?(jmaher) → review+
Blocks: e10s-tests
tracking-e10s: --- → +
https://hg.mozilla.org/mozilla-central/rev/cabf012e0c3b
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → FIXED
Target Milestone: --- → Firefox 46
Should we land these changes on aurora as well? Per comment 583 it seems that that's affected, too.
Flags: needinfo?(ryanvm)
Totally. Wes, can you please add the revs from comment 573 and 582 to your next round of Aurora uplifts, a=test-only?
Flags: needinfo?(ryanvm)
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: