Bug 210213 – Release WebGLLayer earlier in ~GraphicsContextGLOpenGL

• Component: WebGL
• Status: RESOLVED WONTFIX

[Kenneth Russell]

In GraphicsContextGLOpenGL::~GraphicsContextGLOpenGL in Source/WebCore/platform/graphics/cocoa/GraphicsContextGLOpenGLCocoa.mm, the native context is deleted and [WebGLLayer setContext:] called immediately afterward to clear out its NakedPtr<GraphicsContextGLOpenGL>. In order to allow WebGLLayer's cleanup code to potentially do work with the native context, it seems appropriate to release the WebGLLayer earlier.

May be related to other recently filed crashes.

[Kenneth Russell]

Created attachment 395871 [details]
Patch

[EWS]

Committed r259764: <https://trac.webkit.org/changeset/259764>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 395871 [details].

[Radar WebKit Bug Importer]

<rdar://problem/61482594>

[Ryan Haddad]

Reverted r259764 for reason:

Causes layout test crashes under GuardMalloc

Committed r259915: <https://trac.webkit.org/changeset/259915>

[Ryan Haddad]

(In reply to Ryan Haddad from comment #4)
> Reverted r259764 for reason:
> 
> Causes layout test crashes under GuardMalloc
> 
> Committed r259915: <https://trac.webkit.org/changeset/259915>
We were seeing crashes on iOS Simulator when running tests under GuardMalloc

These are some of the tests that hit crashes:
fast/canvas/webgl/no-info-log-for-simple-shaders.html
fast/canvas/webgl/tex-image-and-sub-image-2d-with-image.html
webgl/1.0.3/conformance/extensions/ext-blend-minmax.html
fast/canvas/webgl/no-info-log-for-simple-shaders.html

[Kenneth Russell]

Thanks Ryan and sorry about the crashes. Was worried about that.

How can I run my local iOS Simulator build with GuardMalloc?

[Kenneth Russell]

Note - per fix in Bug 210222, if that one lands and sticks, I'll likely close this as WontFix rather than trying to pursue it further.

[Kenneth Russell]

The fix for Bug 210222 seems that it's stable and won't be reverted, so closing this out as WontFix.