Bug 121889 – [Windows] Attempting to scroll from a non-scrollable layer results the main document getting scrolled

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKit Misc.

[Roger Fong]

This applies to anything that uses RenderLayer::scrollbyRecursive to scroll and where the initial caller of the method is a layer that is not scrollable and has no overflow clip.

The win port uses this for scrolling in response to a panning gesture.

If a layer that for which hasOverflowCiip returns false is first touched when intiating the pan we end up just scrolling the main document.

This is because the method doesn't appropriately recursively search up the RenderLayer tree.  It should walk up the RenderLayer tree in its entirety and not decide to just bail as soon as we reach a layer that has no overflow clip.

[Roger Fong]

<rdar://problem/15072441>

[Roger Fong]

I'm not sure why we haven't seen any issues on mac...hrmmm...will have to look through usages of scrollbyrecursive in the mac port.

[Roger Fong]

(In reply to comment #2)
> I'm not sure why we haven't seen any issues on mac...hrmmm...will have to look through usages of scrollbyrecursive in the mac port.

O! The win port is calling this method from a layer returned by "RenderLayer::enclosingLayer()" instead of "RenderLayer::enclosingScrollableLayer()", that might be the real issue.

[Roger Fong]

I could make scrollbyrecursively calling itself on its enclosingScrollableLayer if the layer itself is not scrollable.

However, it seems the pattern here in general is that whatever calls this method is already a scrollable layer, so I will change the win port to match that for now.

I think we should consider fixing this method in the future though.
Will add a fixme.

[Roger Fong]

Created attachment 212630 [details]
patch

[Roger Fong]

Created attachment 212632 [details]
patch

removed the fixme, filed a separate bug for the issue: https://bugs.webkit.org/show_bug.cgi?id=121929

[Tim Horton]

The changelog could be better.

[Roger Fong]

Committed: http://trac.webkit.org/changeset/156433 
with a better changelog

[Tim Horton]

(In reply to comment #8)
> Committed: http://trac.webkit.org/changeset/156433 
> with a better changelog

agreed