At 03:02 PM 8/9/2001 -0700, Brad Hughes wrote:
>"Craig A. Berry" wrote:
> >
> > At 2:16 PM -0700 8/9/01, Brad Hughes wrote:
> > >Jarkko Hietaniemi wrote:
> > >[...]
> > >>
> ____________________________________________________________________________
> > >> [ 11613] By: jhi on
> 2001/08/08 13:22:13
> > >> Log: Make VMS use IEEE math by default;
> > >
> > >This is a major change; is it because of the bigintpm.t test failing?
> >
> > Basically, yes. They want underflows and overflows to return funny special
> > numbers rather than causing exceptions. I've told him IEEE is not
> > available on VAX. There is also the issue of binary data
> > compatibility, but I think we're already inconsistent here because
> > D_FLOAT is the default on VAX and G_FLOAT on Alpha.
>
>I don't agree with defaulting to a rarely used floating poijnt format
>to accommodate a rarely used module. Shouldn't we revert to VAX floats
>by default, code around the tests for VMS, and await your fixes for the
>Math:: modules? ;-)
It's not just the math modules, unfortunately. Any operation on ints/floats
that over or underflows the float max/min will cause perl to die. The only
way around that is to kick in denorm results, and that means IEEE floats.
Unfortunately.
OTOH, there's only a little bit of difference between G_FLOAT and IEEE.
(Literally... :) And it does mean that Spreadsheet::WriteExcel will work
without hacks. Whether this is a good reason or not is a separate issue
entirely, of course.
Dan
--------------------------------------"it's like this"-------------------
Dan Sugalski even samurai
[EMAIL PROTECTED] have teddy bears and even
teddy bears get drunk
