http://bugs.winehq.org/show_bug.cgi?id=24241
--- Comment #3 from Bart Van Assche bart.vanassche@gmail.com 2010-09-03 13:15:38 CDT --- (In reply to comment #2)
Can you run again with --track-origins=yes on?
Do you think that would provide additional information ? I might have missed something, but as far as I know --track-origins=yes only provides additional information for uninitialized memory accesses and not for invalid free()'s.
And can you provide a download link for an app that reproduces the problem?
I will try to minimize the source code of the application that allows to trigger the above Valgrind complaint. Is MSVC 2005 SP1 source code acceptable instead of an .exe file ?
Finally, can you add a note to http://wiki.winehq.org/Wine_and_Valgrind about your patch? I was under the impression that Wine+valgrind already knew about heap tail zones. What trick did we miss?
You are right, the Wine code in heap.c already informs Valgrind about heap tail zones. Julian and I are looking at preserving heap bits when RtlReAllocateHeap() reallocates memory in-place. The patch that realizes this will be made available as soon as it is ready and underwent sufficient testing. Note: that patch doesn't affect reports about invalid frees.