On Wed, 29 Oct 2003, Melvin Smith wrote:
> At 06:40 PM 10/29/2003 +0100, Leopold Toetsch wrote:
> >Adam Thomason <[EMAIL PROTECTED]> wrote:
> > > The other problematic test is ye =
> > > olde -0 error,
> Its an issue with our (IBM's) compiler. It comes from
> 0.0-0.0*0.0 as a result of the mod test if I recall.
> It is arguably correct, depending on who you ask, except
> every other mainstream compiler disagrees, so go figure.
A quick grep on the perl5 sources for 'negative zero' gets a fair number
of hits mentioning AIX, Digital Unix, and OpenBSD. sv.c contains some
legacy 'FIXNEGATIVEZERO' code as well. In short, it's probably best to
assume that some compilers will give us a negative zero sometimes.
Whether the parrot virtual machine should or should not hide that detail
is a good question -- probably equivalent to whether parrot should spell
infinity "Inf" or "inf" or not care (and similarly for "nan" and "NaN").
--
Andy Dougherty [EMAIL PROTECTED]
