It seems like no content is parsed after the beginning of the text area in Wikipedia edit pages, such as the above URL. The text area is empty and none of the rest of the DOM is present.
I'll take a look at this when I get home tonight.
Reduction <http://crypto.stanford.edu/~abarth/research/webkit/textarea/wikipedia.html>: <html> <body> <textarea> <!-- </textarea> </body> </html>
Created attachment 21966 [details] Step 1: Don't allow entities to start comments This fixes the wikipedia issue.
Created attachment 21967 [details] Step 2: Handle unterminated comments correctly
Comment on attachment 21966 [details] Step 1: Don't allow entities to start comments It's be better if more of the new tests were text-only "dumpAsText" tests so they could be more portable cross-platform. r=me
Comment on attachment 21967 [details] Step 2: Handle unterminated comments correctly r=me
Comment on attachment 21966 [details] Step 1: Don't allow entities to start comments Actually, it would be better to put the new condition at the *end* of the if statement to minimize the possibility of a performance regression in code that might be pretty hot.
Created attachment 21975 [details] Test case with properly encoded comment in textarea. This test case has an encoded comment tag in it, so it shouldn't be processing that comment tag.
> This test case has an encoded comment tag in it, so it shouldn't be processing > that comment tag. Thanks for the test case Robert. I believe the change I just committed in r34831 fixes this issue (including the test case you just attached). I'm waiting for the buildbots to cycle before marking this issue as fixed.