Subject: | Math::Pari problem on Fedora 1 (gcc-3.3.2) |
We are building Math-Pari-2.010501 on a number of platforms, including Solaris 8, RedHat 9, and Fedora Core 1.
On Fedora Core 1, the result of my $I = PARI('34579687721723281952451') is suspicious. After creating $I, I have examined it with print Dumper($I).
SOL8: $VAR1 = bless( do{\(my $o = 5972912)}, 'Math::Pari' );
RH9 : $VAR1 = bless( do{\(my $o = 1082636520)}, 'Math::Pari' );
FC1 : $VAR1 = bless( do{\(my $o = '-1086555928')}, 'Math::Pari' );
$I on FC1 (Fedora) looks suspicious. I believe this is what is returned from sv2pari(). I believe this funny object is causing a segfault when creating a new object that contains this object (Crypt::Random).
The only significant difference that I can see between RH9 and FC1 is that FC1 is using gcc-3.3.2 and RH9 is using gcc-3.2.2. Both platforms use glibc-2.3.2.
An example break in Crypt::Random is:
my $I = PARI('34579687721723281952451');
my $l = makerandom_itv(Lower => $I+1, Upper => 2*$I);
Distribution: Math-Pari-2.010501
Using: pari-2.1.5
Perl: 5.8.3
Linux tkdli001 2.4.22-1.2115.nptlsmp #1 SMP Wed Oct 29 15:30:09 EST 2003 i686 i686 i386 GNU/Linux
#!/opt/perl-5.8.0/bin/perl -sw
#
#
use Math::Pari qw(PARI);
use Data::Dumper;
#my $I = PARI('34579687721723281952451');
#print Dumper($I);