On 17-Aug-2012 17:39, mathog wrote:
Hopefully whatever gremlin this was has been permanently cast out of the system!
No such luck. It comes and it goes. It doesn't crash reliably enough to be sure whatever change was just made was a fix or not. I'm not entirely sure the problem is even in my code exactly, it blows up in the Windows variant but not the linux variant, but when the Windows specific pieces are commented out it still blows up. I have seen things like this before where changes in code in a program far far away causes other bits of memory to move around, so that a preexisting memory access violation is moved into a destructive location.
Ran it in valgrind on linux and found a pair of writes outside of allocated memory (not in my patch region.) Will try fixing those on linux, maybe it will heal the windows version.
David Mathog mathog@...1176... Manager, Sequence Analysis Facility, Biology Division, Caltech