gcc 2.95: wrong floating point underflow handling on Alpha

[Pawel Salek]

Hi,

* The GCC version: gcc 2.95 as well as egcs 1.12
* system type:    Linux (RedHat 6.1), Alpha ev6
* options passed: gcc underflow.c
* source code:

int main () {
   double c = 3.958851865662793e-313;
   if(c>1e0) printf("Bigger than 1\n");
}

*Behavior:
crashes in ln.3 with Floating point exception (core dumped) message

*Expected behavior: 
total silence.

*Description:
underflows are sometimes handled inproperly on Alpha (tested on i686 and
works fine). The default behavior is supposed to ignore underflows but
it doesn't.

The same problem appears in g77 but I guess it's because of common
backend.

I would appreciate if you informed me when a patch is available.

/Pawel
-- 
-- 
Pawel Salek (pawsa@theochem.kth.se) http://www.theochem.kth.se/~pawsa/
Theoretical Chemistry Division, KTH voice: +46 8 790-8202