After bug 88888, we have a new descendant dependent flag (has-self-painting-layer-descendant) on RenderLayer that behaves slightly differently than the existing flags (visible descendant mostly and visible content to some extend). One of the review comment of bug 88888 was to bring their implementation and naming closer to avoid un-intended difference in behavior.
Created attachment 147892 [details] Proposed change v1.
Comment on attachment 147892 [details] Proposed change v1. View in context: https://bugs.webkit.org/attachment.cgi?id=147892&action=review > Source/WebCore/ChangeLog:16 > + Changed this method to not take a boolean as every callers where passing "as every caller was passing" > Source/WebCore/rendering/RenderLayer.cpp:648 > + sc->dirtyZOrderLists(); > + if (sc->hasVisibleContent()) I'd like to see a comment explaining why z-order lists are dirtied. I think we also don't omit hidden layers from the z-order lists when compositing, so you may be able to avoid this dirty in some cases.
Comment on attachment 147892 [details] Proposed change v1. View in context: https://bugs.webkit.org/attachment.cgi?id=147892&action=review >> Source/WebCore/rendering/RenderLayer.cpp:648 >> + if (sc->hasVisibleContent()) > > I'd like to see a comment explaining why z-order lists are dirtied. I think we also don't omit hidden layers from the z-order lists when compositing, so you may be able to avoid this dirty in some cases. Here is what I have added to the patch for landing: // We don't collect invisible layers in z-order lists if we are not in compositing mode. // As we became visible, we need to dirty our stacking contexts ancestors to be properly // collected. FIXME: When compositing, we could skip this dirtying phase.
Created attachment 148424 [details] Patch for landing
Comment on attachment 148424 [details] Patch for landing Clearing flags on attachment: 148424 Committed r120770: <http://trac.webkit.org/changeset/120770>
All reviewed patches have been landed. Closing bug.