Bug 168660 – REGRESSION(~r212322): LayoutTest media/track/track-cue-container-rendering-position.html is a flaky timeout

• Status: RESOLVED
• Resolution: FIXED
• Component: Media

[Ryan Haddad]

LayoutTest media/track/track-cue-container-rendering-position.html is a flaky timeout

https://build.webkit.org/results/Apple%20Sierra%20Debug%20WK1%20(Tests)/r212702%20(3335)/results.html

https://webkit-test-results.webkit.org/dashboards/flakiness_dashboard.html#showAllRuns=true&tests=media%2Ftrack%2Ftrack-cue-container-rendering-position.html

[Alexey Proskuryakov]

First occurrence in r212322. Not sure if it can be the culprit, unless it messed up some feature initialization flags.

But even if this change is not to blame, this was definitely broken not long before - the test was absolutely stable earlier, and fails very often now. The test fails especially often on slower bots (ASan, leaks, GuardMalloc).

@@ -1,7 +1,7 @@
+CONSOLE MESSAGE: line 83: No text track cue with display id '-webkit-media-text-track-display' is currently visible
+FAIL: Timed out waiting for notifyDone to be called
 
 The top of the text track container should be in the bottom 30% of the video element.
 EVENT(canplaythrough)
 
-EXPECTED (cueDisplayElement.offsetTop > (video.videoHeight * .70) == 'true') OK
-END OF TEST

[Radar WebKit Bug Importer]

<rdar://problem/30726032>

[Per Arne Vollan]

I am unable to reproduce this timeout with a debug ASAN build.

[Per Arne Vollan]

Created attachment 305728 [details]
Patch

[Per Arne Vollan]

Created attachment 305729 [details]
Patch

[Eric Carlson]

Comment on attachment 305729 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=305729&action=review

> LayoutTests/media/track/track-cue-container-rendering-position.html:25
> +            } catch (exception) {

I think it would be better to try this a finite number of times and fail with an error message instead of having the test timeout in the event of an error.

[Per Arne Vollan]

Created attachment 305834 [details]
Patch

[Per Arne Vollan]

(In reply to Eric Carlson from comment #6)
> Comment on attachment 305729 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=305729&action=review
> 
> > LayoutTests/media/track/track-cue-container-rendering-position.html:25
> > +            } catch (exception) {
> 
> I think it would be better to try this a finite number of times and fail
> with an error message instead of having the test timeout in the event of an
> error.

Thanks for reviewing! I will update the patch.

[Per Arne Vollan]

Committed <https://trac.webkit.org/changeset/214600/webkit>.

[Eric Carlson]

Comment on attachment 305834 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=305834&action=review

> LayoutTests/media/track/track-cue-container-rendering-position.html:14
> +        var maxRetries = 256;

This means the test won't fail for 25 seconds. What made you choose such a long time, it seems *extremely* unlikely that it could take that long for the cue display element to be created.