[RenderTreeBuilder] Move styleDidChange mutation logic to RenderTreeUpdater

[zalan]

ssia.

[Radar WebKit Bug Importer]

<rdar://problem/38054892>

[zalan]

Created attachment 334877 [details]
Patch

[EWS Watchlist]

Comment on attachment 334877 [details]
Patch

Attachment 334877 [details] did not pass mac-debug-ews (mac):
Output: http://webkit-queues.webkit.org/results/6729312

New failing tests:
fullscreen/full-screen-fixed-pos-parent.html

[EWS Watchlist]

Created attachment 334879 [details]
Archive of layout-test-results from ews114 for mac-sierra

The attached test failures were seen while running run-webkit-tests on the mac-debug-ews.
Bot: ews114  Port: mac-sierra  Platform: Mac OS X 10.12.6

[zalan]

Created attachment 334906 [details]
Patch

[zalan]

Created attachment 334907 [details]
Patch

[Antti Koivisto]

Comment on attachment 334907 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=334907&action=review

> Source/WebCore/rendering/updating/RenderTreeUpdater.cpp:341
> +    auto& parent = *renderer.parent();
> +    bool isFloating = renderer.style().isFloating();
> +    bool isOutOfFlowPositioned = renderer.style().hasOutOfFlowPosition();
> +    bool startsAffectingParent = false;
> +    bool noLongerAffectsParent = false;
> +
> +    if (is<RenderBlock>(parent))
> +        noLongerAffectsParent = (!wasFloating && isFloating) || (!wasOufOfFlowPositioned && isOutOfFlowPositioned);
> +
> +    if (is<RenderBlockFlow>(parent) || is<RenderInline>(parent)) {
> +        startsAffectingParent = (wasFloating || wasOufOfFlowPositioned) && !isFloating && !isOutOfFlowPositioned;
> +        ASSERT(!startsAffectingParent || !noLongerAffectsParent);
> +    }
> +
> +    if (startsAffectingParent) {
> +        // We have gone from not affecting the inline status of the parent flow to suddenly
> +        // having an impact. See if there is a mismatch between the parent flow's
> +        // childrenInline() state and our state.
> +        renderer.setInline(renderer.style().isDisplayInlineType());
> +        if (renderer.isInline() != renderer.parent()->childrenInline())
> +            m_builder.childFlowStateChangesAndAffectsParentBlock(renderer);
> +        return;
> +    }
> +
> +    if (noLongerAffectsParent) {
> +        m_builder.childFlowStateChangesAndNoLongerAffectsParentBlock(renderer);
> +
> +        if (is<RenderBlockFlow>(renderer)) {
> +            // Fresh floats need to be reparented if they actually belong to the previous anonymous block.
> +            // It copies the logic of RenderBlock::addChildIgnoringContinuation
> +            if (isFloating && renderer.previousSibling() && renderer.previousSibling()->isAnonymousBlock())
> +                m_builder.move(downcast<RenderBoxModelObject>(parent), downcast<RenderBoxModelObject>(*renderer.previousSibling()), renderer, RenderTreeBuilder::NormalizeAfterInsertion::No);
> +        }
> +    }
> +}

Maybe this stuff could go to the Builder as mutateTreeAfterStyleChange() or something?

[zalan]

Created attachment 334957 [details]
Patch

[WebKit Commit Bot]

Comment on attachment 334957 [details]
Patch

Clearing flags on attachment: 334957

Committed r229200: <https://trac.webkit.org/changeset/229200>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.