[WinCairo] Crash in ATI display driver in accelerated compositing mode.

• Component: WebKit Misc.
• Status: RESOLVED FIXED

I got a crash in the ATI display driver in accelerated compositing mode, which resulted in a Windows blue screen.
This patch is a speculative fix for the crash; a call to use antialiasing when drawing a texture is removed.
I haven't seen the crash after making this change, and I don't think antialiasing is really needed.
It is taken from the GTK implementation, and I can't see any visual differences after removing the call.
Accelerated compositing should also be faster when not doing antialiasing.

Created attachment 240121 [details]
Patch

[Darin Adler]

Comment on attachment 240121 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=240121&action=review

> Source/WebKit/win/WebCoreSupport/AcceleratedCompositingContext.cpp:-106
> -    static_cast<TextureMapperGL*>(m_textureMapper.get())->setEnableEdgeDistanceAntialiasing(true);

Is this code used only in the WinCairo configuration or is it also used in Apple’s Windows port?

[Brent Fulgham]

Comment on attachment 240121 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=240121&action=review

>> Source/WebKit/win/WebCoreSupport/AcceleratedCompositingContext.cpp:-106
>> -    static_cast<TextureMapperGL*>(m_textureMapper.get())->setEnableEdgeDistanceAntialiasing(true);
> 
> Is this code used only in the WinCairo configuration or is it also used in Apple’s Windows port?

This file is only used in the WinCairo port; we don't build it on our port because we do not enable the TEXTURE_MAPPER_GL feature (we use our own compositing code).

[Alex Christensen]

This seems like it is just covering up a bigger problem, but I'd rather have no antialiasing than frequent crashes. Is there an equivalent call in the gtk code?

[WebKit Commit Bot]

Comment on attachment 240121 [details]
Patch

Clearing flags on attachment: 240121

Committed r174919: <http://trac.webkit.org/changeset/174919>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

(In reply to comment #4)
> This seems like it is just covering up a bigger problem, but I'd rather have
> no antialiasing than frequent crashes. Is there an equivalent call in the
> gtk code?

Yes, the gtk code has the same call.