Regresses the time for the #speech5 test on webkit.org/perf/slickspeed from 1-3ms to 10-12+ ms.
Created attachment 22488 [details] Patch; Passes run-webkit-tests and doesn't regress http://ejohn.org/apps/selectortest/#target
Comment on attachment 22488 [details] Patch; Passes run-webkit-tests and doesn't regress http://ejohn.org/apps/selectortest/#target Looks good.
Comment on attachment 22488 [details] Patch; Passes run-webkit-tests and doesn't regress http://ejohn.org/apps/selectortest/#target The check shouldn't be necessary, and can be made an ASSERT as far as I can tell. It should also read querySelector->m_attr== idAttr, not querySelector->m_attr.localName() == idAttr. I believe this will also break some uses of querySelector("#camalCase"); in quirks mode. We probably need to disable the optimization in quirks mode methinks.
Created attachment 22511 [details] patch
Fixed in r35406.