Re: standards on SIGFPE for divide by zero?

[Jeff Rizzo <riz%tastylime.net@localhost>]

On 2/28/11 12:38 PM, Greg Troxel wrote:
>    My question is:  what do the relevant standards say about this?  Is it
>    required to SIGFPE?  If it's required, I'll file a PR against
>    port-powerpc and mark the tests expected-fail on ppc.  If not, then
>    I'll skip them on ppc.
>
> I dimly recall that it's required to be able to set this mode, but that
> it doesn't have to be the default.  I think this is controlled by
> IEEE754, but note that vax is different.
>
> See fpgetmask(3).  Maybe the test should set FP_X_DZ.

Someone else pointed me there, but perhaps I'm using it wrong... after doing
fpsetmask(FP_X_DZ);
The only thing different is that "1.0/0.0" prints "0" instead of "inf".  :/


> Regardless of standards, it's probably good for NetBSD to be consistent
> across ports on default fp behavior.

I agree, but am willing to yield to more experienced heads on this.

+j