Bug 139826

Summary: REGRESSION (177368): Some tests started to immediately time out
Product: WebKit Reporter: Alexey Proskuryakov <ap>
Component: Tools / TestsAssignee: Alexey Proskuryakov <ap>
Status: RESOLVED FIXED    
Severity: Normal CC: cdumez, dbates, koivisto
Priority: P2    
Version: 528+ (Nightly build)   
Hardware: Unspecified   
OS: Unspecified   
See Also: https://bugs.webkit.org/show_bug.cgi?id=96741
Attachments:
Description Flags
proposed fix dbates: review+

Description Alexey Proskuryakov 2014-12-19 12:03:03 PST 
fast/repaint/no-animation-outside-viewport-subframe.html has started to frequently fail on December 16.

It's not a true failure, but a problem anyway. For some reason, waitUntilDone watchdog timer fires almost immediately - logs say that the test timed out, but also that it only took 0.1 second!

https://build.webkit.org/builders/Apple%20Mavericks%20Release%20WK2%20(Tests)/builds/10221/steps/layout-test/logs/stdio

11:28:23.653 74860 Tests that timed out or crashed:
...
11:28:23.654 74860   fast/repaint/no-animation-outside-viewport-subframe.html took 0.1 seconds
Comment 1 Chris Dumez 2014-12-19 12:13:11 PST 
Antti wrote this test, maybe he has an idea.
Comment 2 Alexey Proskuryakov 2014-12-19 12:14:52 PST 
This looks like a problem with deeper WebCore or DumpRenderTree machinery, like DOM timers being broken.
Comment 3 Alexey Proskuryakov 2014-12-19 12:46:23 PST 
This happens on other tests too, just happened for the first time on fast/text/international/lang-sensitive-fonts-xml.xhtml.

So far, looks like WebKit2 only.
Comment 4 Alexey Proskuryakov 2014-12-19 13:45:59 PST 
My theory is that this is some deficiency in WebKitTestRunner watchdog timer behavior, we might leak it across tests. Looking into this.
Comment 5 Alexey Proskuryakov 2014-12-19 14:40:54 PST 
The reason why this regressed is that I unskipped a test that breaks a consequent test 30 seconds later!
Comment 6 Alexey Proskuryakov 2014-12-19 14:50:37 PST 
Created attachment 243579 [details]
proposed fix
Comment 7 Daniel Bates 2014-12-19 14:54:07 PST 
Comment on attachment 243579 [details]
proposed fix

View in context: https://bugs.webkit.org/attachment.cgi?id=243579&action=review

> Tools/WebKitTestRunner/InjectedBundle/TestRunner.cpp:165
> +    // than to let webkitpy do that, becasue WebKitTestRunner will dump partial results.

Nit: ", because" => because
Comment 8 Daniel Bates 2014-12-19 14:57:13 PST 
Do we need something similar for DumpRenderTree?
Comment 9 Alexey Proskuryakov 2014-12-19 15:25:06 PST 
> Do we need something similar for DumpRenderTree?

I'm checking that. It looks like the bug exists there too, but it's much less of a problem due to the timer being a global.
Comment 10 Alexey Proskuryakov 2014-12-19 15:26:06 PST 
Committed <http://trac.webkit.org/r177598>.
Comment 11 Alexey Proskuryakov 2014-12-19 15:28:15 PST 
Actually, DumpRenderTree seems fine, we always go through dumpTestResults() there.