REGRESSION(r200171?, r200164?) animations/animation-events-not-cancelable.html is flaky on Mac/iOS

• Component: Animations
• Status: NEW ---

[Ryan Haddad]

LayoutTest animations/animation-events-not-cancelable.html has become flaky on Mac and iOS as of r200171. Since it is a flaky failure, I am not sure if that change is directly related.

Example failure:
<https://build.webkit.org/results/Apple%20iOS%209%20Simulator%20Debug%20WK2%20(Tests)/r200191%20(1438)/results.html>

Flakiness dashboard:
<http://webkit-test-results.webkit.org/dashboards/flakiness_dashboard.html#showAllRuns=true&tests=animations%2Fanimation-events-not-cancelable.html>

--- /Volumes/Data/slave/ios-simulator-9-debug-tests-wk2/build/layout-test-results/animations/animation-events-not-cancelable-expected.txt
+++ /Volumes/Data/slave/ios-simulator-9-debug-tests-wk2/build/layout-test-results/animations/animation-events-not-cancelable-actual.txt
@@ -5,8 +5,6 @@
 
 animationstart:
 PASS event.cancelable is false
-animationiteration:
-PASS event.cancelable is false
 animationend:
 PASS event.cancelable is false
 PASS successfullyParsed is true

[Simon Fraser (smfr)]

Attempted improvement in https://trac.webkit.org/r200213

[Alexey Proskuryakov]

This made it much, much better. But I still found one failure on iOS (WK1 production, same diff).

[Simon Fraser (smfr)]

Yeah. On a loaded bot, the animation won't get serviced frequently enough to fire the iteration event. I don't know how to fix this.

[Alexey Proskuryakov]

This particular test seems like it would be nearly as good if it didn't log anything on success. There is the risk it will somehow never fire the animationiteration at some point in the future and thus stop testing it, but that seems unlikely.

Also, perhaps it's a bug that we allow skipping an iteration?