windemut@uxh.cso.uiuc.edu (Windemut) (10/10/90)
I have a program where I do a lot of floating point calculations on the NeXT. Unfortunately there is some bug that leads to illegal results. From earlier, older machines I am used to a program aborting when a floating point exception such as a division by zero or an overflow happens. On the NeXT, all I get is a lot of NaNs and no indication whatsoever where the illegal operation might have occurred. GDB seems to be useless. The closest reference to a possible solution of this problem I found is in math(3), the manual entry for the libm library. There the exception handling is described as being up to the programmer. Alas, the information is very vague with no hints given as to actually accomplish that. How can I gat an oldfashioned, but immensely helpful "Floating point exception (core dumped)" from my program? I am quite desperate about this and would appreciate some help from somebody who knows. Andreas Windemuth Theoretical Biophysics University of Illinois at Urbana-Champaign