HTML5 tree builder should pass some LayoutTests
Created attachment 61101 [details] Patch
Comment on attachment 61101 [details] Patch WebCore/html/HTMLConstructionSite.cpp:103 + RefPtr<Node> child = prpChild; This causes us to end up taking an extra reference. We should steal a raw pointer for our ASSERTs. parserAddChild can never fail. Why do we need to be attaching text nodes? Is something wrong in the lazy attach codepath? Does DRT need to be calling some layout method that it's not?
Comment on attachment 61101 [details] Patch You need to at least add a FIXME about the extra ref problem. But it seems silly/lazy to not just fix it now. :p CC'd James R. about understanding the Text node lazy-attach failure problem.
Created attachment 61105 [details] Patch for landing
Comment on attachment 61105 [details] Patch for landing Clearing flags on attachment: 61105 Committed r63037: <http://trac.webkit.org/changeset/63037>
All reviewed patches have been landed. Closing bug.