On Saturday 05 July 2008 10:50:59 you wrote:
>  On Saturday 05 July 2008 10:28:54 Bram wrote:
>  > Citeren Marc Lehmann <perlbug-followup@[...].org>:
>  > > -----------------------------------------------------------------
>  > > [Please enter your report here]
>  > >
>  > > When Math::BigInt uses GMP as backend, it gives completely bogus
>  > > results:
>  > >
>  > >    perl -MMath::BigInt=lib,GMP -e 'my $b = 0xc0010203; my $m =
>  > > Math::BigInt->new (2) ** 24 - 1; warn $m + $b'
>  > >    -105689855 at -e line 1.
>  > >
>  > >    perl -MMath::BigInt=lib,GMP -e 'my $b = 0xc0010203; my $m =
>  > > Math::BigInt->new (2) ** 32 - 1; warn $m + $b'
>  > >    3221291522 at -e line 1.
>  > >
>  > > This only seems to happen with 32-bit perls.
> 
>  I am pretty sure that is because the value is in-between treated as
>  signed vs. unsigned integer and thus arrives wrongly in libgmp.
> 
>  Can you please run:
> 
>  	perl -Mbignum,l,GMP=v

Sorry, 

	perl -Mbignum=l,GMP,v

But I think I found the bug, GMP,xs contains:

    if (SvUOK(x))
      {
      mpz_init_set_si(*RETVAL, (UV)SvUV(x));
      }
    else
      {
      ...

That should probably be:

    if (SvUOK(x))
      {
      mpz_init_set_ui(*RETVAL, (UV)SvUV(x));	// _ui to fix [perl #56608]
      }
    else
      {
      ...

Marc, could you please patch, recompile and test with this and see if it 
resolves it? I only have a 64bit Perl on this machine.

All the best,

Tels

-- 
 Signed on Sat Jul  5 10:54:34 2008 with key 0x93B84C15.
 Get one of my photo posters: http://bloodgate.com/posters
 PGP key on http://bloodgate.com/tels.asc or per email.

 "In my opiniation, "burglarize" is a perfectionally validative
 wordification. How else would reportization of the securitial/policial
 forceship appearize to be importantive enoughly to be respectative by
 the massmediaship and influentate the societyness?"

  -- SharpFang (651121) on 2004-12-13 at /. about "burgle" 
vs. "burglarize"