Bug 93077 – [WK2] LayerTreeCoordinator can disallow tile updates in forceRepaint

• Component: Platform
• Status: REOPENED

[Balazs Kelemen]

This has been revealed in connection with bug 90394 and it makes repaint tests failing.

[Balazs Kelemen]

Created attachment 156283 [details]
proposed fix

This was suggested by Joselyn :)
Actually I'm not 100% sure it's correct. It fixes the tests and don't introduce new fails. But is it ok to send tile updates when waiting for the ui process? I guess it works because when running WTR, the UI expects this behavior. As long as the only user of force repaint is wtr this makes no danger but still I wonder whether should we involve the UI side in fixing this.

[Jocelyn Turcotte]

Comment on attachment 156283 [details]
proposed fix

View in context: https://bugs.webkit.org/attachment.cgi?id=156283&action=review

Yep that's what I had in mind. I think it's ok to override m_waitingForUIProcess here since forceRepaint does the same for the rest of the layer sync operations by calling flushPendingLayerChanges directly instead of performScheduledLayerFlush.

> Source/WebKit2/WebProcess/WebPage/LayerTreeCoordinator/LayerTreeCoordinator.cpp:179
> +    WTF::TemporaryChange<bool> inForceRepaint(m_inForceRepaint, true);

Wow, didn't know we had such a class.

[Balazs Kelemen]

Comment on attachment 156283 [details]
proposed fix

Landed in http://trac.webkit.org/changeset/124603

[Csaba Osztrogonác]

(In reply to comment #3)
> (From update of attachment 156283 [details])
> Landed in http://trac.webkit.org/changeset/124603

It made 18 tests crash, here is the brand new bug report for you: https://bugs.webkit.org/show_bug.cgi?id=93121

[Balazs Kelemen]

Reopening because of the crashes.

[Balazs Kelemen]

Reverted r124603 for reason:

Made a few tests crash under LayerTreeCoordinator::flushPendingLayerChanges

Committed r124620: <http://trac.webkit.org/changeset/124620>