[Webkit-unassigned] [Bug 73350] [chromium] Allow scrolling non-root layers in the compositor thread

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=73350


James Robinson <jamesr at chromium.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #119898|review+                     |review-
               Flag|                            |




--- Comment #30 from James Robinson <jamesr at chromium.org>  2011-12-19 16:17:12 PST ---
(From update of attachment 119898)
View in context: https://bugs.webkit.org/attachment.cgi?id=119898&action=review

> Source/WebCore/platform/graphics/chromium/cc/CCLayerTreeHostImpl.cpp:477
> +CCInputHandlerClient::ScrollStatus CCLayerTreeHostImpl::beginScrollingInnermostLayerAtPoint(CCLayerImpl* layerImpl, const IntPoint& viewportPoint)

sorry to reverse myself earlier here but now that I think about this more I don't think this is right. just finding the first most-nested layer is not going to find what the user really wants, especially if the tree order doesn't match the z-order very well

instead of iterating through the layer tree in tree order, you really need to walk in z-order here by going through the CCRenderSurface list

> Source/WebCore/platform/graphics/chromium/cc/CCLayerTreeHostImpl.cpp:488
> +    IntPoint contentPoint(layerImpl->screenSpaceTransform().inverse().mapPoint(viewportPoint));

if the layer's screenSpaceTransform is not invertible, inverse() will return an identity transform which will probably _not_ do what you expect here. if the screenSpaceTransform is not invertible we probably do not want to let it scroll, since that means that some dimension got mapped to zero and the layer isn't actually visible

also do you need to consider layer backfaces here?

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.