Bug #71548 for Math-BigInt-GMP: libmath-bigint-gmp-perl: new() from UV wrong value

This is a different defect caused by the same code. In this case it is converting 18446744073709551615 to -1 because of the mpz_init_set_si: perl -Iblib/lib -Iblib/arch -E 'use Math::BigInt lib=>"GMP"; my $n = Math::BigInt->new(10); say $n; say $n->bacmp(~0); say $n->copy->bsub(~0);' 10 1 11 perl -E 'use Math::BigInt lib=>"Calc"; my $n = Math::BigInt->new(10); say $n; say $n->bacmp(~0); say $n->copy->bsub(~0);' 10 -1 -18446744073709551605 perl -E 'use Math::BigInt lib=>"Pari"; my $n = Math::BigInt->new(10); say $n; say $n->bacmp(~0); say $n->copy->bsub(~0);' 10 -1 -18446744073709551605 Changing to mpz_init_set_ui makes it work correctly, and still passes the test suite. I think a test like this ought to be added to the test suite.

By the way, a hacky workaround for both is to put ''. in front of the number. So: if ($_[0]->bacmp(''.~0) <= 0) { ... } or in the foo.pl case: my $b = Math::BigInt->new(''.$n); # alternately: my $b = Math::BigInt->new("$n"); I've been using the latter a lot to work around this bug.