Wednesday, April 11, 2012

I hope I'm wrong about this

I've been spending a lot of time lately writing C (yay!) to be called from MATLAB (boo!). Something deep in my code was seg faulting, but I couldn't get close to it because MATLAB's got signal handlers set. Okay, fine, I'll set the handler for SIGSEGV to SIG_DFL and then I can look at the core file.

Getting rid of MATLAB's SIGSEGV handler leads to crash and dump within seconds, even if my code is just the signal() call. Which makes me think that there's something in MATLAB that's seg faulting more or less continuously, and they're putting the train back on the tracks every few seconds.

The grass is so very much greener over in NumPy/SciPy/NiPy land.

No comments:

Post a Comment