[iOS Scrolling] Propagate scrolls to non-nested UIScrollViews

• Component: Scrolling
• Status: RESOLVED FIXED

[Antti Koivisto]

With help of some UIKit SPIs.

[Antti Koivisto]

Created attachment 372913 [details]
WIP

[Antti Koivisto]

Created attachment 372914 [details]
test case

[Antti Koivisto]

Created attachment 373030 [details]
test with more nested scrollers

[Antti Koivisto]

Created attachment 373104 [details]
patch

[Simon Fraser (smfr)]

Comment on attachment 373104 [details]
patch

View in context: https://bugs.webkit.org/attachment.cgi?id=373104&action=review

> Source/WebKit/UIProcess/RemoteLayerTree/ios/RemoteLayerTreeViews.mm:141
> +                    if (auto* nonAncestorScrollingNode = host.nodeForID(node->relatedScrollContainerIDs()[0]))

I'm having a hard time convincing myself that using relatedScrollContainerIDs()[0] is correct. What does it mean to have more than one? What is the order?

[Antti Koivisto]

> I'm having a hard time convincing myself that using
> relatedScrollContainerIDs()[0] is correct. What does it mean to have more
> than one? What is the order?

Would knowing that you wrote the code that sets it be help? :)

It is simply the layer tree counterpart of ScrollingTreeOverflowScrollProxyNode::m_overflowScrollingNodeID. Having Moves  and Stationary cases in a single vector matches the old design where PositionedNode handled both cases. It should really be refactored into two separate variables, matching PositionedNode split.

[WebKit Commit Bot]

Comment on attachment 373104 [details]
patch

Clearing flags on attachment: 373104

Committed r246926: <https://trac.webkit.org/changeset/246926>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Radar WebKit Bug Importer]

<rdar://problem/52338107>

[Antti Koivisto]

<rdar://problem/49720738>