* Dave Anderson [2008-05-13 15:36]: > > Bernhard Walle wrote: > > When crash is compiled with gcc 4.3 and -O2, the __builtin_return_address() > > causes crash to crash. See also [1] for a discussion about that. > > It appears that [1] only describes a problem if you ask for more frames > than exist. Given the usage in crash, it's impossible to have a frame count > of less than 4. So it must be some other bug, no? Yes. I only referenced to that bug to state that the behaviour is undefined, and it's not a gcc bug per definition. However, I tried to track down the problem a bit. In a test program, I encountered that when __builtin_frame_address(i) is zero, then the program crashes when __builtin_frame_address(i+1) is executed. However, changing that in crash doesn't help. It really crashes in __builtin_return_address(1). Tracking that down is out of my knowledge. I used the same compiler flags in crash and the test program. Bernhard -- Crash-utility mailing list Crash-utility@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/crash-utility
