548022 - test_animLengthReadonly.xhtml sporadically aborts [ABORT: Negative sample time for active animation: 'mSampleTime >= 0 || !mIsActive']

Status: RESOLVED FIXED

(Core :: SVG, defect)

Description

[Shawn Wilsher :sdwilsh]

47798 INFO Running /tests/content/svg/content/test/test_animLengthReadonly.xhtml...
++DOMWINDOW == 14 (0E961AE8) [serial = 1263] [outer = 07343170]
WARNING: wrong frame type: file e:/builds/moz2_slave/mozilla-central-win32-debug/build/content/svg/content/src/nsSVGSVGElement.cpp, line 1103
###!!! ABORT: Negative sample time for active animation: 'mSampleTime >= 0 || !mIsActive', file e:/builds/moz2_slave/mozilla-central-win32-debug/build/content/smil/nsSMILAnimationFunction.cpp, line 257
###!!! ABORT: Tear-off objects remain in hashtable at shutdown.: 'mTable.Count() == 0', file e:\builds\moz2_slave\mozilla-central-win32-debug\build\content\svg\content\src\nsSVGAttrTearoffTable.h, line 54

WINNT 5.2 mozilla-central debug test mochitests-1/5 on 2010/02/23 09:22:14  
http://tinderbox.mozilla.org/showlog.cgi?tree=Firefox&errorparser=unittest&logfile=1266945734.1266947599.10782.gz&buildtime=1266945734&buildname=WINNT%205.2%20mozilla-central%20debug%20test%20mochitests-1%2f5&fulltext=1

Comment 1

[Daniel Holbert [:dholbert] (blocking needinfo/review, due to AFK on parental leave 12/9 - 1/6)]

Note that this test already calls PauseAnimations before doing anything else.  

So, unlike some past SMIL randomorange bugs, this one doesn't appear to be a simple timing issue (i.e. "the-tinderbox-took-longer-than-we-expected-to-evaluate-these-lines-of-js").

Comment 2

[Daniel Holbert [:dholbert] (blocking needinfo/review, due to AFK on parental leave 12/9 - 1/6)]

It looks like this happens when the onload event fires very late -- after we're already begun our animation -- and then we pause & seek back to a time before our animation, and we fail this NS_ABORT_IF_FALSE.

Two things:
 (1) We can prevent the late handler firing by listening for SVGLoad instead of for load, as mentioned in bug 535850 comment 6. (This removes the need to seek back to 0, also)
 (2) I need to investigate this NS_ABORT_IF_FALSE. (perhaps it should be a NS_WARNING instead?) It's probably related to the fact that we're not great with seeking back before the current interval yet (bug 492458).

Comment 3

[Daniel Holbert [:dholbert] (blocking needinfo/review, due to AFK on parental leave 12/9 - 1/6)]

(In reply to comment #2)
>  (1) We can prevent the late handler firing by listening for SVGLoad instead of

I filed bug 550975 on this.  (Marking dependency, since I'll fix that first, which will make this work.)

>  (2) I need to investigate this NS_ABORT_IF_FALSE.

I filed Bug 550978 on this, with a reduced testcase.

Comment 4

[Phil Ringnalda (:philor)]

http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1268100217.1268102484.9513.gz
Linux mozilla-central debug test mochitests-1/5 on 2010/03/08 18:03:37
s: moz2-linux-slave24

49472 INFO Running /tests/content/svg/content/test/test_animLengthUnits.xhtml...
++DOMWINDOW == 17 (0xa86f6f0) [serial = 1373] [outer = 0xa4fcbc8]
###!!! ABORT: Negative sample time for active animation: 'mSampleTime >= 0 || !mIsActive', file /builds/slave/mozilla-central-linux-debug/build/content/smil/nsSMILAnimationFunction.cpp, line 257

(not quite the same test, but I figured it was close enough)

Comment 5

[Daniel Holbert [:dholbert] (blocking needinfo/review, due to AFK on parental leave 12/9 - 1/6)]

I just landed bug 550975, which should fix this randomorange per comment 3.