Re: signal(SIGFPE,SIG_IGN), possible solution?

Matthew J Brown (mjb@sophos.com)
Sun, 28 Apr 1996 21:14:49 +0100

H. Peter Anvin writes:
> No kidding. However, it seems to me that either SIGILL, SIGIOT or
> SIGBUS (in decreasing order of personal preference) would be a better
> signal than SIGFPE; I think it is reasonable that a program should be
> able to expect that receiving a SIGFPE means an FP exception has
> occurred.

I think this is another of those Historical Reasons things ... this is
what various other flavours of UNIX do, and changing it would break
compatibility. It seems to me that SIGFPE should really be named
SIGMATH or something like that, to indicate that it's a general math
error signal (integer and float) rather than float specifically, but I
guess we're stuck with it now.

-Matt