1283026 - Converted TimeStamp value will be null when using long effect time and negative playbackRate.

Status: RESOLVED INVALID

(Core :: DOM: Animation, defect, P3)

Description

[Mantaroh Yoshinaga[:mantaroh]]

In bug 1267510, We faced the problem which the result of adding NavigationStartTimeStamp and TimeDuration value was underflow.

This phenomenon will occur when setting too long effect time and negative playbackRate.
(The crash occurred when playing the transition after setting negative playbackRate[1].)

[1] http://searchfox.org/mozilla-central/rev/261fe13dcd88cfd2e99e65755e7ca4b7a2e583df/dom/animation/test/css-transitions/file_animation-finished.html#49

Going into details, when firing the first tick after playing transition, we calculated TimeStamp of transition's effect end. This calculated timestamp value will be a large negative number.

Because timestamp implementations of several platforms use high time step dependent accuracy. For example, Linux uses the nanoseconds. So If we specified 1000s duration and negative playbackRate(like -1.0), the calculated value will be -1000000000000ns.[2]
Then the procedure of converting to TimeStamp will be underflow when specifying those large negative number.[3]

[2] http://searchfox.org/mozilla-central/rev/ef24c234ed53b3ba50a1734f6b946942e4434b5b/dom/animation/Animation.cpp#632
[3] http://searchfox.org/mozilla-central/rev/261fe13dcd88cfd2e99e65755e7ca4b7a2e583df/dom/animation/DocumentTimeline.cpp#207

Comment 1

[Mantaroh Yoshinaga[:mantaroh]]

I tried this case, but this is not occurring on transition.
In the transition, we can't set the playbackRate without Web Animation API.
So we couldn't reproduce phenomenon of comment #0.

But negative small playbackRate will cause other problem.
For detail, see bug 1293945.

Comment 2

[Hiroyuki Ikezoe (:hiro)]

Do you mean null TimeStamp will never happen even if it's in case of web animations?

Comment 3

[Mantaroh Yoshinaga[:mantaroh]]

Hi hiro,

(In reply to Hiroyuki Ikezoe (:hiro) from comment #2)
> Do you mean null TimeStamp will never happen even if it's in case of web
> animations?
I think that perhaps this problem won't occur *now*.
Perhaps we will face bug 1293945 before this phenomenon.

I'm not sure that there is similar limit value problem.
But I think that perhaps we need to investigate to whether we have a similar problem or not.
So I'm going to investigate it.