[Webkit-unassigned] [Bug 164303] [GTK] BadDamage X Window System error in WebKit::AcceleratedBackingStoreX11::update when called from WebPageProxy::exitAcceleratedCompositingMode

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=164303

--- Comment #13 from Carlos Garcia Campos <cgarcia at igalia.com> ---
(In reply to comment #12)
> Comment on attachment 295406 [details]
> Patch
> 
> I'd like Zan to look at this.
> 
> Questions:
> 
>  (a) Isn't this equivalent to trapping the entire function, now? Both calls
> to X11 APIs are within the scope of one trapper or another, we could just
> put the trapper at the top so we only need one instead of two.

Not exactly the same, first one only happens if we already have a surface and second one if there's a pixmap and drawing area. We could just move it, though, but I prefer to make more expliit what we are protecting.

>  (b) It occurs to me that it's probably a bad idea to trap errors
> unconditionally here. We really want to ignore errors only if shutting down,
> right? Is there any way we can do that, or are we just out of luck and have
> to trap always?

Not really, there are a lot of situation in which this can happen, for example if the web process crashes in the middle of update, the pixmap is gone but we have a pixmap id in the Ui process. Also because of the async way X11 works, sometimes errors happen later, on a sync, so we could be failing due to previous errors. If you take a look at other projects using X11 directly like GTK+, cairo or clutter you will see they are full of xerror trappers.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20161124/9387a2b8/attachment.html>