[webkit-reviews] review denied: [Bug 87873] [chromium] Single thread proxy should not tick animations unless the layer renderer has been initialized : [Attachment 145103] Patch

[bugzilla-daemon at webkit.org]

James Robinson <jamesr at chromium.org> has denied vollick at chromium.org's request
for review:
Bug 87873: [chromium] Single thread proxy should not tick animations unless the
layer renderer has been initialized
https://bugs.webkit.org/show_bug.cgi?id=87873

Attachment 145103: Patch
https://bugs.webkit.org/attachment.cgi?id=145103&action=review

------- Additional Comments from James Robinson <jamesr at chromium.org>
View in context: https://bugs.webkit.org/attachment.cgi?id=145103&action=review


> Source/WebCore/platform/graphics/chromium/cc/CCLayerTreeHostImpl.h:97
> +    virtual void didStopAnimationTimer() { }

This is an odd part of the API, especially since the notion of animation timer
is CCSingleThreadProxy specific.  Is this just a test hook?  Why not simply
have a test that calls finishAllRendering and then ASSERT()s that the timer
doesn't fire?

> Source/WebCore/platform/graphics/chromium/cc/CCLayerTreeHostImpl.h:126
> +    virtual bool initializeLayerRenderer(PassRefPtr<GraphicsContext3D>,
TextureUploaderOption);

Also a weird test hook - we can make initialization fail by passing a
GraphicsContext3D in that's not useful (say, make the first
makeContextCurrent() call return false) - that's a much more realistic test and
we do it in other places

> Source/WebCore/platform/graphics/chromium/cc/CCSingleThreadProxy.cpp:133
> +	   // We're done rendering and should not process any future animation
events.

I'd much prefer that we cancel the timer when initialization fails.  The fact
that WebViewImpl calls finishAllRendering() is a strange historical artifact of
how shutdown used to work and isn't really a concern of the compositor - in
fact I've been planning to delete that call.