Title: [243374] trunk/Source/WebCore
- Revision
- 243374
- Author
- [email protected]
- Date
- 2019-03-22 06:46:56 -0700 (Fri, 22 Mar 2019)
Log Message
[Web Animations] com.apple.WebKit.WebContent.Development at com.apple.WebCore: WebCore::WebAnimation::timeToNextTick const + 757
https://bugs.webkit.org/show_bug.cgi?id=196125
<rdar://problem/46520059>
Patch by Antoine Quint <[email protected]> on 2019-03-22
Reviewed by Dean Jackson.
Because of floating point math in playState() that wouldn't account for timeEpsilon, we would get into a state where the animation phase
was "after" but the play state was "running" when the current time was about a microsecond away from the active time boundary. This meant
that the early return statement in WebAnimation::timeToNextTick() would not be hit while we would not handle the possibility of being in
the "after" phase in the rest of the function, therefore eventually hitting the ASSERT_NOT_REACHED() at the end of the function.
We now account for timeEpsilon in playState() and correctly report we're in the "finished" play state when in the "after" phase also.
* animation/WebAnimation.cpp:
(WebCore::WebAnimation::playState const):
Modified Paths
Diff
Modified: trunk/Source/WebCore/ChangeLog (243373 => 243374)
--- trunk/Source/WebCore/ChangeLog 2019-03-22 12:56:32 UTC (rev 243373)
+++ trunk/Source/WebCore/ChangeLog 2019-03-22 13:46:56 UTC (rev 243374)
@@ -1,3 +1,21 @@
+2019-03-22 Antoine Quint <[email protected]>
+
+ [Web Animations] com.apple.WebKit.WebContent.Development at com.apple.WebCore: WebCore::WebAnimation::timeToNextTick const + 757
+ https://bugs.webkit.org/show_bug.cgi?id=196125
+ <rdar://problem/46520059>
+
+ Reviewed by Dean Jackson.
+
+ Because of floating point math in playState() that wouldn't account for timeEpsilon, we would get into a state where the animation phase
+ was "after" but the play state was "running" when the current time was about a microsecond away from the active time boundary. This meant
+ that the early return statement in WebAnimation::timeToNextTick() would not be hit while we would not handle the possibility of being in
+ the "after" phase in the rest of the function, therefore eventually hitting the ASSERT_NOT_REACHED() at the end of the function.
+
+ We now account for timeEpsilon in playState() and correctly report we're in the "finished" play state when in the "after" phase also.
+
+ * animation/WebAnimation.cpp:
+ (WebCore::WebAnimation::playState const):
+
2019-03-22 Alicia Boya García <[email protected]>
[MSE][GStreamer] Don't construct segments on PlaybackPipeline::flush
Modified: trunk/Source/WebCore/animation/WebAnimation.cpp (243373 => 243374)
--- trunk/Source/WebCore/animation/WebAnimation.cpp 2019-03-22 12:56:32 UTC (rev 243373)
+++ trunk/Source/WebCore/animation/WebAnimation.cpp 2019-03-22 13:46:56 UTC (rev 243374)
@@ -528,7 +528,7 @@
// animation's effective playback rate > 0 and current time ≥ target effect end; or
// animation's effective playback rate < 0 and current time ≤ 0,
// → finished
- if (animationCurrentTime && ((effectivePlaybackRate() > 0 && animationCurrentTime.value() >= effectEndTime()) || (effectivePlaybackRate() < 0 && animationCurrentTime.value() <= 0_s)))
+ if (animationCurrentTime && ((effectivePlaybackRate() > 0 && (*animationCurrentTime + timeEpsilon) >= effectEndTime()) || (effectivePlaybackRate() < 0 && (*animationCurrentTime - timeEpsilon) <= 0_s)))
return PlayState::Finished;
// Otherwise → running
_______________________________________________
webkit-changes mailing list
[email protected]
https://lists.webkit.org/mailman/listinfo/webkit-changes
