Closed Bug 940381 Opened 6 years ago Closed 5 years ago
Intermittent TEST-UNEXPECTED-FAIL | /tests/content/svg/content/test/test
_viewport .html | Test timed out .
Rev4 MacOSX Lion 10.7 mozilla-aurora opt test mochitest-1 on 2013-11-18 16:10:05 PST for push 1c1fcdc50acc slave: talos-r4-lion-069 https://tbpl.mozilla.org/php/getParsedLog.php?id=30730024&tree=Mozilla-Aurora 198574 ERROR TEST-UNEXPECTED-FAIL | /tests/content/svg/content/test/test_viewport.html | Test timed out. 198577 ERROR TEST-UNEXPECTED-FAIL | /tests/content/svg/content/test/test_viewport.html | uncaught exception - TypeError: inner is null at http://mochi.test:8888/tests/content/svg/content/test/test_viewport.html:37 198578 ERROR TEST-UNEXPECTED-FAIL | /tests/content/svg/content/test/test_viewport.html | [SimpleTest.finish()] this test already called finish! 198579 ERROR TEST-UNEXPECTED-FAIL | /tests/content/svg/content/test/test_viewport.html | called finish() multiple times
inner is null at ... test_viewport.html:37 http://mxr.mozilla.org/mozilla-central/source/content/svg/content/test/test_viewport.html?force=1#37 Maybe the iframe isn't loaded yet when we execute line 37? Should the "load" listener on line 74 be on the iframe instead?
Priority: -- → P5
Yes, I imagine it should be on the iframe.
Assignee: nobody → longsonr
Attachment #8340279 - Flags: review?(dholbert)
I think the iframe's load event blocks the outer document's load event, right? At least, that seems to be the case for: data:text/html,<body onload="alert('body')"><iframe src="https://people.mozilla.com" onload="alert('iframe')"> Still, I guess this could be worth a shot, and it may get us closer to the problem if it's not already addressing the problem.
Attachment #8340279 - Flags: review?(dholbert) → review+
(In reply to Daniel Holbert [:dholbert] from comment #4) > I think the iframe's load event blocks the outer document's load event, > right? I think you're right, my mistake. So maybe the iframe load simply hangs for some reason? Looking at the log again, I see that the "inner is null" error occurs after the 5 minute timeout. So when the test framework is aborting this test (I don't know what that entails) it's causing the load event to occur? there's also this: [SimpleTest.finish()] this test already called finish! that must be our call in runTest(), right? but the framework already called finish() as part of aborting the test?
> it's causing the load event to occur? Aborting a load will in fact fire onload on the aborted document. > but the framework already called finish() as part of aborting the test? Entirely possible. The new test after the comment 6 changes seems _less_ robust to me, because the iframe can finish loading before the script ever runs, so the load listener it adds would never get triggered. As in, the new code is _definitely_ buggy, while the old code at least has a chance at being correct.
(In reply to Boris Zbarsky [:bz] from comment #8) > > but the framework already called finish() as part of aborting the test? > > Entirely possible. Yeah, I recall seeing this in other mochitest test timeouts (in cases where there's clearly only one code-path in the test that could reach Finish()). So I wouldn't pay too much attention to that > The new test after the comment 6 changes seems _less_ robust to me, because > the iframe can finish loading before the script [that sets up the load listener] ever runs Oh, I didn't think of that -- good point. With that realization, I agree -- this is now less-good, and we should probably back out. :-/ On the bright side, the test-timeout has only ever happened once, AFAICT, so whatever's causing it is pretty anomalous.
Backed out part 1: https://hg.mozilla.org/integration/mozilla-inbound/rev/27c14a1b1d4c (part 2 is unrelated test-simplification and should stay in)
My apologies for causing unnecessary work here. :-(
Status: NEW → RESOLVED
Closed: 6 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla28
Canceling FIXED setting, as this was backed out after it landed (on inbound).
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Target Milestone: mozilla28 → ---
(and the merged-to-central cset for part 2 in comment 12 is tangential and didn't fix this)
The missing iframe load event doesn't seem likely to be an SVG bug. It seems more likely the problem is on the server side, in the mochitest framework I presume.
Component: SVG → Mochitest
Product: Core → Testing
I doubt this is a Mochitest framework bug. It's likely either a platform bug or a bug in the test.
Closing bugs where TBPLbot has previously commented, but have now not been modified for >3 months & do not contain the whiteboard strings for disabled/annotated tests or use the keyword leave-open. Filter on: mass-intermittent-bug-closure-2014-07
Status: REOPENED → RESOLVED
Closed: 6 years ago → 5 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.