[webkit-reviews] review requested: [Bug 109302] Simplify logic for automatically opting into composited scrolling : [Attachment 192274] Patch

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Fri Mar 8 12:37:32 PST 2013


Glenn Hartmann <hartmanng at chromium.org> has asked  for review:
Bug 109302: Simplify logic for automatically opting into composited scrolling
https://bugs.webkit.org/show_bug.cgi?id=109302

Attachment 192274: Patch
https://bugs.webkit.org/attachment.cgi?id=192274&action=review

------- Additional Comments from Glenn Hartmann <hartmanng at chromium.org>
> > > Source/WebCore/rendering/RenderLayer.cpp:609
> > > +    updateNormalFlowList();
> > 
> > Was the original ASSERT wrong? (AFAICT the rest of the calling code is the
same so this shouldn't be needed, unless the ASSERT was indeed wrong).
> > 
> > > Source/WebCore/rendering/RenderLayer.cpp:611
> > > +    ASSERT(!m_normalFlowListDirty);
> > 
> > This ASSERT is trivially passing as you call updateNormalFlowList() above,
did you remove the wrong one?
> 
> Fixed, the original ASSERT was correct, I put it back in and removed the call
to updateNormalFlowList().

No, I was wrong. The diff there is misleading, as the old function is being
called from stacking contexts, whereas the new one is being called from all the
stacking contexts' children. The children do not necessarily have clean normal
flow lists. However, this call to updateNormalFlowLists() seems to be outdated
from a previous version of the patch. I've gone through the function and we
don't actually use the normal flow lists for anything, we only need z-order
lists. So I took out all the normalFlow stuff from the beginning of that
function.

I also found and fixed an indexing bug in getStackingOrderElementAt() that
tripped assertions in debug mode.

I updated updateCanBeStackingContainerRecursively, as well, to mirror the
behavior of collectLayers().

Finally, I updated the code that builds our pre/post promotion paint order
lists to never make use of the 'stacking container' concept (which depends on
the promotion decision for other layers).


More information about the webkit-reviews mailing list