pjw@usna.NAVY.MIL (Prof. Peter Welcher) (05/02/90)
I'm interested in finding out if other people are having comparable problems with MSC 6.0 (and 5.1, for that matter). Reply to pjw@math2.sma.usna.navy.mil. When I compile and run Microsoft's example SIGFP.C, I get a runtime stack overflow. My impression, after a lot of tinkering, is that signal trapping of SIGFPE does not work on my Zenith Z-248. It does not work on various versions of Zenith's MSDOS and various Zenith machines at my office. It does not work on an IBM AT (DOS 3.1). I use signal(SIGFPE,...) in Lattice C. I'm not a novice, but there is that 0.01% chance I'm screwing up. So: has anyone else observed this phenomenon ? Microsoft admitted to this being a bug when I reported it two years ago (in version 5.1). It appears it has not been fixed. What do people using MS C do to keep their programs from dying on floating point exceptions ? (Mine is math-intensivve. Testing for zero before every division slows it down a good bit. Also kind of misses the point of IEEE floating point standard, which is that you should be able to do math, then check and see if an error occurred.)