605345 - "ABORT: Got a regular sample during startup state, expected an end sample instead"

Status: RESOLVED FIXED

(Core :: SVG, defect)

Description

[Jesse Ruderman]

Attachment: testcase with sketchy timer

###!!! ABORT: Got a regular sample during startup state, expected an end sample instead: 'mElementState != STATE_STARTUP || aEndOnly', file /builds/slave/mozilla-central-macosx64-debug/build/content/smil/nsSMILTimedElement.cpp, line 506

Comment 1

[Brian Birtles (:birtles)]

Assigning to me for now.

Comment 2

[Brian Birtles (:birtles)]

Nominating as a blocker until I have time to look at it properly.

Comment 3

[Brian Birtles (:birtles)]

Attachment: Patch v1a

The problem is that we end up getting negative container times because we're not being careful about how we handle pause state whilst a time container has no parent.

This patch fixes that behaviour.

I've also added two minor tweaks:

a) Clear the mStartAnimationOnBindToTree flag in nsSVGSVGElement after we've used it since it only makes sense to call Begin on a time container once.

b) Moved the point where we reset our max milestone in nsSMILTimedElement::BindToTree. This isn't strictly necessary, but it's a little more intuitive. Stepping through this with a debugger I noticed that we end up clearing this value and registering the milestone twice: firstly in Rewind and then again at the end of the method. That's no problem since nsSMILTimeContainer allows for that, but if we reset this value at the start of the method, then we'll skip registering the milestone a second time.

Comment 4

[Daniel Holbert [:dholbert]]

Comment on attachment 488396 [details] [diff] [review]
Patch v1a

Looks great! r=dholbert

Comment 5

[Brian Birtles (:birtles)]

Pushed: http://hg.mozilla.org/mozilla-central/rev/32d8cf0c4596