Bug 175329 – REGRESSION (r220052): [Sierra] API test VideoControlsManager.VideoControlsManagerFullSizeVideoInWideMainFrame is a flaky failure

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

[Ryan Haddad]

API test VideoControlsManager.VideoControlsManagerFullSizeVideoInWideMainFrame is a flaky failure

FAIL VideoControlsManager.VideoControlsManagerFullSizeVideoInWideMainFrame

/Volumes/Data/slave/sierra-debug/build/Tools/TestWebKitAPI/Tests/WebKit2Cocoa/VideoControlsManager.mm:57
Value of: hasVideoForControlsManager
  Actual: false
Expected: true

https://build.webkit.org/builders/Apple%20Sierra%20Debug%20WK1%20(Tests)/builds/2870

[Alexey Proskuryakov]

HAs something changed, or have we simply been not noticing this for so long?

[Ryan Haddad]

(In reply to Alexey Proskuryakov from comment #1)
> HAs something changed, or have we simply been not noticing this for so long?

This seems to be a recent regression, perhaps sharing the same root cause as https://bugs.webkit.org/show_bug.cgi?id=175143

I am trying to reproduce both failures locally and identify a regression point.

[Ryan Haddad]

This regressed with https://trac.webkit.org/changeset/220052/webkit

[Ryan Haddad]

Skipped this flaky test in https://trac.webkit.org/changeset/220711/webkit

[Darin Adler]

Jer, can you perhaps help me figure out why this is failing? I would like to fix it and re-enable the test.

[Darin Adler]

Created attachment 319811 [details]
Patch

[Darin Adler]

Passing all EWS now. Ready for someone to review.

[Wenson Hsieh]

Comment on attachment 319811 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=319811&action=review

> Tools/ChangeLog:9
> +        video controls to appear. With changes to how loading occurrs, it can now

Nit - occurrs => occurs

> Tools/ChangeLog:13
> +        changing the tests to wait one more run loop should get rid of the flakiness.

In the longer term, I think it makes more sense for some of these tests to expect media controls to show up eventually, instead of expecting them after a "playing" message sent after a 0-delay timer in the web process. As you noted, there isn't a guarantee of how many run loop cycles it would take for the web process to be in a state where it wants to show media controls, nor should there be, since touch bar presence for media elements isn't web/API exposed.

[Darin Adler]

Committed r221589: <http://trac.webkit.org/changeset/221589>

[Radar WebKit Bug Importer]

<rdar://problem/34694057>