Originally Posted by uau
Start the program first, then attach gdb to the running process. Looks like the the crash happens when the process forks, probably in a pthread_atfork() handler registered by the nvidia driver.
Yeah that's what I did ... maybe it is some kind of race that just didn't trigger due to gdb slowing it down.
Anyway a crash at fork looks like it is the same bug as reported in http://www.nvnews.net/vbulletin/showthread.php?t=174244