This caused a regression after r165037. When we have to clear :active style of shadow DOM, we should clear host's style.
Created attachment 227625 [details] Patch
Comment on attachment 227625 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=227625&action=review > Source/WebCore/dom/Document.cpp:5844 > + for (Element* curr = oldActiveElement; curr; curr = curr->parentOrShadowHostElement()) { To set these flags we use the render tree. To clear them we are using the DOM tree. I suspect this inconsistency could cause more problems than just this one.
Darin, Thanks for your review. I'll check it and upload another patch.
Comment on attachment 227625 [details] Patch As Darin pointed out, the inconsistent tree traversing between set and clear phase needs to be investigated, but it may be followed in other bug. So I set cq+ for this patch.
Comment on attachment 227625 [details] Patch Clearing flags on attachment: 227625 Committed r166277: <http://trac.webkit.org/changeset/166277>
All reviewed patches have been landed. Closing bug.