[comp.sys.dec] Problems with floating point routines in Ultrix 2.0

jon@applix.UUCP (Jon Dale) (09/23/87)

we have found some problems with the math libraries in Ultrix 2.0
running on a VAX 8600.  Specifically, using the "power" function "pow",
an overflow condition results in a SIGILL (illegal instruction), and
and underflow does not result in any notification at all, you just
get a zero back.  The first of these causes the most significant problems.
Ultrix 1.2 correctly reported the overflow as a SIGFPE (floating point
exception), and did not report the underflow.
   The test routine is written in "C" and linked with the standard -lm
switch.  Anyone know a fix or workaround so I can avoid crashing on a
floating point overflow??  What's the problem here?

Jon Dale   APPLiX
...ames!harvard!m2c!applix!jon