http://bugs.winehq.org/show_bug.cgi?id=30505
--- Comment #5 from Nikolay Sivov bunglehead@gmail.com 2012-04-24 03:16:11 CDT --- (In reply to comment #4)
Do you think the crash within CDATA event handling deserves a new bug?
Yes, please. It looks unrelated, I suspect two things so far - we mess with input stream a bit before calling handlers, could be buggy. Second thought is used chunk length for characters() callback, and there's no tests for CDATA section for a reader apparently.