Serge Gautherie (:sgautherie) Reporter
	Description • 16 years ago

Bug 467672 comment 29 checkins fail on MacOSX SeaMonkey.

{
[...]
*** 42170 INFO TEST-PASS | /tests/layout/base/tests/test_bug441782.html | Rendering of reftest bug467672-5 is different with bidi.numeral == 3
*** 42171 ERROR TEST-UNEXPECTED-FAIL | /tests/layout/base/tests/test_bug441782.html | Test timed out.
}

On
http://tinderbox.mozilla.org/showbuilds.cgi?tree=SeaMonkey&maxdate=1234493009
Regression timeframe:
http://hg.mozilla.org/releases/mozilla-1.9.1/pushloghtml?startdate=2009-02-12+10%3A53%3A50&enddate=2009-02-12+14%3A28%3A43

	(no longer active) Assignee
	Comment 1 • 16 years ago

Serge: do you have access to a Mac?  Can you reproduce this locally?  If yes, I'd like to attach a patch which catches any exceptions to see what's failing here.

	Serge Gautherie (:sgautherie) Reporter
	Comment 2 • 16 years ago

No, I have my local Windows 2000 only.

	(no longer active) Assignee
	Comment 3 • 16 years ago

Roc: do you have access to a Mac (and time to do the testing) or know someone who does?

	Robert Kaiser
	Comment 4 • 16 years ago

As a note, this unit test box is running Tiger, while I think all Firefox boxes are Leopard.

	Robert O'Callahan (:roc) (email my personal email if necessary)
	Comment 5 • 16 years ago

Jonathan might be able to help

	Jonathan Kew [:jfkthame]
	Comment 6 • 16 years ago

I'll take a look; currently pulling the source so I can start a SeaMonkey build. I don't have a Mac running Tiger, though, if that turns out to be a significant factor.

	Jonathan Kew [:jfkthame]
	Comment 7 • 16 years ago

This test runs ok for me on 10.5 (Leopard) using a 1.9.1 trunk build. I don't have access to Tiger for comparison here.

Is it simply timing out because it takes too long to complete? test_bug441782.html has a lot of sub-tests and runs pretty slowly; I just timed it at slightly over 1 minute on my laptop. What's the timeout used when running the unit tests? (IIRC I've seen 30 seconds mentioned somewhere, in which case a timeout would be quite likely on anything except a pretty fast Mac.) Maybe Ehsan could simply split it into a collection of separate files.

	(no longer active) Assignee
	Comment 8 • 16 years ago

I think Jonathan's suggestion to split the test is worthwhile, here is the patch.  It doesn't include any changes to the test code, it just splits it to multiple files.

	Robert O'Callahan (:roc) (email my personal email if necessary)
	Comment 9 • 16 years ago

Even if it doesn't fix the problem, it will help narrow it down.

	(no longer active) Assignee
	Comment 10 • 16 years ago

http://hg.mozilla.org/mozilla-central/rev/b6debd8f3a0d
http://hg.mozilla.org/releases/mozilla-1.9.1/rev/ab8f309ba4ab

I won't mark it as fixed now in order to wait and see the tests passing.

	(no longer active) Assignee
	Comment 11 • 16 years ago

The SeaMonkey test finally passed: <http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey/1234775152.1234782319.28786.gz>, so it seems like it was indeed a test taking too long to complete.  -> FIXED.  Thanks for the suggestion, Jonathan.

	Robert Kaiser
	Comment 12 • 16 years ago

And thanks for doing that fix, Ehsan.

	(no longer active) Assignee
	Comment 13 • 16 years ago

(In reply to comment #12)
> And thanks for doing that fix, Ehsan.

Sure thing.  Sorry for turning your tree orange in the first place.  :-)

	Serge Gautherie (:sgautherie) Reporter
	Comment 14 • 16 years ago

verified1.9.1, as box turned green.

	Boris Zbarsky [:bzbarsky]
	Comment 15 • 16 years ago

I suspect, for what it's worth, that the slowness was a combination of using slow canvas compare and reflowing the big mochitest document a whole bunch of times during these tests...