[net.unix-wizards] Floating-point overflow

johns (11/24/82)

Does anyone know how to recover from a floating-point overflow?
Using signal, sigset, or sigsys does not appear to help
as the error signal can be caught (or ignored) but the overflowed
value then causes an illegal instruction error on the next 
instruction executed.  How can any numerical routines be run
on these vax's if overflows always cause the core to be dumped.