Bug 76254 – REGRESSION: svg/custom/use-instanceRoot-event-listeners.xhtml & svg/custom/pointer-events-invalid-fill.svg broken on the Bots

• Component: SVG
• Status: RESOLVED FIXED

[Nikolas Zimmermann]

REGRESSION: svg/custom/use-instanceRoot-event-listeners.xhtml & svg/custom/pointer-events-invalid-fill.svg broken on the Bots.
It broke between r103854 and r103869. Will bisect.

[Nikolas Zimmermann]

Error on SL bot first appears here: http://build.webkit.org/old-results/SnowLeopard%20Intel%20Debug%20(Tests)/r103867%20(4059)/results.html

Wasn't present yet in the previous build: http://build.webkit.org/old-results/SnowLeopard%20Intel%20Debug%20(Tests)/r103864%20(4058)/results.html.

Either:
- http://trac.webkit.org/changeset/103865 (document collections change)
- http://trac.webkit.org/changeset/103866 (MiniBrowser change, seems not guilty)
- http://trac.webkit.org/changeset/103867 (cursor changes)

It's not obvious who is guilty, I'll look for another bot with results in that revision range.

[Nikolas Zimmermann]

Aha, both tests use onmouesover handlers, so Dans r103867 is likely the cause.
CC'ing Dan.

[Nikolas Zimmermann]

I revert r103867 locally, and the problems are gone. Reopening the bug associated with r103867.

svg/custom/pointer-events-invalid-fill.svg appears to be relying on a mouseMove event firing late enough after the test loaded (perhaps after all subresources loaded). It used to get that late-firing event as a result of the timer in WebHTMLView.mm (rather than as a result of its own call to eventSender.mouseMoveTo()). It can be made to pass by delaying the call to eventSender.mouseMoveTo() with a timer. If this is about the font resource, then forcing layout before the load event fires should work as well.

The root cause appears to be that the document doesn’t reliably finish loading in the API sense, following the changes in <http://trac.webkit.org/r101858>.

…which is the same issue indicated in bug 73688 comment #7.

(In reply to comment #5)
> The root cause appears to be that the document doesn’t reliably finish loading in the API sense, following the changes in <http://trac.webkit.org/r101858>.

Actually, that issue existed prior to r101858.

Created attachment 122482 [details]
Patch

Fixed in <http://trac.webkit.org/r105005>.

[Nikolas Zimmermann]

(In reply to comment #9)
> Fixed in <http://trac.webkit.org/r105005>.
Thanks for the quick fix Dan!