Currently we invalidate the whole element subtree if there are any attribute selectors for the changed attribute. We should only invalidate what is really needed.
Created attachment 271344 [details] patch
<rdar://problem/24655799>
Created attachment 271345 [details] patch
Comment on attachment 271345 [details] patch View in context: https://bugs.webkit.org/attachment.cgi?id=271345&action=review r=me This is a really solid performance hack! > Source/WebCore/style/AttributeChangeInvalidation.h:47 > + using ClassChangeVector = Vector<AtomicStringImpl*, 4>; why
Created attachment 271360 [details] for landing
Created attachment 271361 [details] for landing
https://trac.webkit.org/r196629