Bug#1043234: libmath-bigint-perl: subclassing breaks numeric comparison

[Niko Tyni]

Package: libmath-bigint-perl
Version: 1.999838-1
Severity: important
Tags: ftbfs upstream
User: debian-perl at lists.debian.org
Usertags: perl-5.38-transition
Control: affects -1 ledgersmb

ledgersmb fails to build with new versions of Math-BigInt,
including the one separately packaged (both in stable and unstable)
as libmath-bigint-perl, and the one bundled with Perl 5.38 (currently
in experimental.)

   #   Failed test 'form: 9.999,99 parsed as 1.000,00 - 9999.99'
   #   at t/02-number-handling.t line 224.
   #          got: 9999.99
   #     expected: 9999.99
   
   [...]
   
   Test Summary Report
   -------------------
   t/02-number-handling.t        (Wstat: 5120 (exited 20) Tests: 374 Failed: 20)
     Failed tests:  336-340, 345-349, 354-358, 365-369
     Non-zero exit status: 20
   Files=20, Tests=1802, 17 wallclock secs ( 0.39 usr  0.06 sys + 14.92 cusr  1.89 csys = 17.26 CPU)
   Result: FAIL

I was able to narrow it down to this test, which fails with newer versions:

---------------
package MyFloat;
use base qw(Math::BigFloat);
1;

print MyFloat->new(9999.99) == Math::BigFloat->new(9999.99) ? "ok\n": "not ok\n";
---------------

It bisects down to upstream commit 

  https://github.com/pjacklam/p5-Math-BigInt/commit/46d1252c98f565ae787c840173e5f98acf8953f1

so it regressed with upstream version 1.999836.
-- 
Niko Tyni   ntyni at debian.org