Bug 37494 – REGRESSION (r57340): fast/events/mouse-drag-from-frame-to-other-frame.html fails on Windows

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Adam Roben (:aroben)]

Since r57340 <http://trac.webkit.org/changeset/57340>, fast/events/mouse-drag-from-frame-to-other-frame.html has been failing on Windows.

[Adam Roben (:aroben)]

<rdar://problem/7857060>

[Alexey Proskuryakov]

Hmm. What about bug 29963?

I don't see how r57340 could affect this.

[Adam Roben (:aroben)]

(In reply to comment #2)
> Hmm. What about bug 29963?

That bug was fixed in r54400, and the expected failure result was removed in r54459. But now we're failing the test again.

> I don't see how r57340 could affect this.

Given that r48884 likely caused bug 29963, and r54400 fixed it, it seems this test is sensitive to some details of our implementation of event listeners. Since r57340 changes some of those details, I'm not that surprised that it could have caused this.

[Adam Roben (:aroben)]

Created attachment 53312 [details]
Add Windows failing results for mouse-drag-from-frame-to-other-frame.html

[Eric Seidel (no email)]

Comment on attachment 53312 [details]
Add Windows failing results for mouse-drag-from-frame-to-other-frame.html

I assume you're going to leave this bug open.

[Adam Roben (:aroben)]

(In reply to comment #5)
> I assume you're going to leave this bug open.

Yes, exactly.

Thanks for the review!

[Adam Roben (:aroben)]

Comment on attachment 53312 [details]
Add Windows failing results for mouse-drag-from-frame-to-other-frame.html

Committed r57563.

[Alexey Proskuryakov]

The problem here is that after r57340, ScriptController for one of the frames is created later than it used to be. It's created during replaySavedEvents in DRT, which results in EventSender state being lost:

JSObjectRef makeEventSender(JSContextRef context) 
{
    down = false; 
    dragMode = true; 
    replayingSavedEvents = false; 
...

We don't have such code on the Mac, because the state machine for replaySavedEvents is different there, and resetting wasn't needed. See <http://trac.webkit.org/changeset/29724> for the original fix that introduced resetting.

[Alexey Proskuryakov]

Created attachment 53488 [details]
proposed fix

[Geoffrey Garen]

Comment on attachment 53488 [details]
proposed fix

r=me

[Alexey Proskuryakov]

Committed <http://trac.webkit.org/changeset/57685>.