http://bugs.winehq.org/show_bug.cgi?id=28038
--- Comment #7 from Austin English austinenglish@gmail.com 2012-03-28 13:31:50 CDT --- (In reply to comment #6)
There is a bug in the code which causes the code to go out of bounds on hash pages. This will either cause an overwrite of existing data, or cause a SIGSEGV if the hash page is close to the end, and the overflow goes outside of the memory mapped area.
I have added a patch for the problem if you want to test it out. It will be part of a longer serie that I am preparing for urlcache.
I was still able to get a failure as in comment #1 when running in a loop after 40 seconds.