[webkit-reviews] review granted: [Bug 109639] Padding and border changes doesn't trigger relayout of children : [Attachment 188941] Patch

[bugzilla-daemon at webkit.org]

Ryosuke Niwa <rniwa at webkit.org> has granted Dave Hyatt <hyatt at apple.com>'s
request for review:
Bug 109639: Padding and border changes doesn't trigger relayout of children
https://bugs.webkit.org/show_bug.cgi?id=109639

Attachment 188941: Patch
https://bugs.webkit.org/attachment.cgi?id=188941&action=review

------- Additional Comments from Ryosuke Niwa <rniwa at webkit.org>
View in context: https://bugs.webkit.org/attachment.cgi?id=188941&action=review


> Source/WebCore/rendering/RenderBlock.cpp:420
> +	   // First walk our normal flow objects and see if there is any
change.
> +	   for (RenderBox* childBox = firstChildBox(); childBox; childBox =
childBox->nextSiblingBox()) {
> +	       if (childBox->isFloatingOrOutOfFlowPositioned())
> +		   continue;
> +	       childBox->setChildNeedsLayout(true, MarkOnlyThis);
> +	   }

I would be good to mention why we don't have to set floating objects dirty per
IRC discussion.

> Source/WebCore/rendering/RenderBlock.cpp:423
> +	   // Now walk the positioned objects for which we are the containing
block.
> +	   TrackedRendererListHashSet* positionedDescendants =
positionedObjects();

It would be great if you could come up with a test case for floating or
position objects.

> Source/WebCore/rendering/RenderTableRow.cpp:89
> +		   if (!childBox->isTableCell())
> +		       continue;

On a completely unrelated note, it's weird that we can ever have a
non-RenderTableCell RenderObject in RenderTableRow.