Move HTML Attribute case-sensitivity logic out of SelectorChecker to HTMLDocument.
Created attachment 183803 [details] Patch
Comment on attachment 183803 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=183803&action=review This is no longer inline, but I doubt it matters. The more I look at style resolve perf the more I realize its the tree walk which is so expensive (that, and a few super hot functions, like cansharestylewith) > Source/WebCore/html/HTMLDocument.cpp:344 > + HashSet<AtomicStringImpl*>* attrSet = new HashSet<AtomicStringImpl*>; I don't think this needs to be atomicstringimpl anymore. That was an old limitation of our collection classes, but now we have traits for atomicstring. I also don't think it hurts to use impl either. :-)
Comment on attachment 183803 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=183803&action=review >> Source/WebCore/html/HTMLDocument.cpp:344 >> + HashSet<AtomicStringImpl*>* attrSet = new HashSet<AtomicStringImpl*>; > > I don't think this needs to be atomicstringimpl anymore. That was an old limitation of our collection classes, but now we have traits for atomicstring. I also don't think it hurts to use impl either. :-) Well, AtomicString would mean we’d ref and deref each item as it goes into the set, and even ref and deref keys as we check if they are in the map. So I still think this going one level lower to avoid all the ref and deref is worthwhile.
Created attachment 183880 [details] Patch for landing
Comment on attachment 183880 [details] Patch for landing Clearing flags on attachment: 183880 Committed r140382: <http://trac.webkit.org/changeset/140382>
All reviewed patches have been landed. Closing bug.