Bug 91686 – [BlackBerry] Suspend when there's no target buffer until an external compositor is set

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKit BlackBerry

[Jakob Petsovits]

If we don't have a client window (i.e. rendering to GL directly), and a WebPageCompositor is only set after a rendering operation (e.g. a loadData() call that doesn't need to wait for networking), then we'll try to render to BackingStorePrivate::buffer() which doesn't exist at this point. That's bad, and gets us various assertions and possibly worse. Fix it by starting in a screen-suspended state and only resuming screen and backingstore once a compositor is actually set.

[Jakob Petsovits]

Created attachment 153123 [details]
Patch

[Jakob Petsovits]

Depends on the robustified suspend/resume behavior in bug 91644, therefore I didn't set a commit-queue request yet.

[Jakob Petsovits]

Comment on attachment 153123 [details]
Patch

Bug 91644 has been pushed, so this one's good to go if I get a positive review. Setting cq? flag in addition to the r?. Preferred reviewer is George Staikos as this patch is a reimplementation (slightly different, imho slightly better) of an original suggestion of his, but if others feel it's straightforward enough then I might take somebody else's review as well.

[Antonio Gomes]

Comment on attachment 153123 [details]
Patch

I am not in love with this, but discussed with Jakob offline.

Jakob, please update the commit message with the details you said on IRC (right timing of things, etc)

[Jakob Petsovits]

Created attachment 153344 [details]
Patch as r+ed above, with more verbose commit message

Antonio, please cq+ if the commit message is more what you were thinking of. Thanks :)

[WebKit Review Bot]

Comment on attachment 153344 [details]
Patch as r+ed above, with more verbose commit message

Clearing flags on attachment: 153344

Committed r123172: <http://trac.webkit.org/changeset/123172>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.