Bug #70743 for Math-BigInt: round_mode test fail on 5.6.2

Mon Sep 05 14:12:27 2011 zefram [...] fysh.org - Ticket created

Subject:	round_mode test fail on 5.6.2
Date:	Mon, 5 Sep 2011 19:12:15 +0100
To:	bug-Math-BigInt [...] rt.cpan.org
From:	Zefram <zefram [...] fysh.org>

The round_mode tests in Math-BigInt are, in 1.997, newly failing on perl 5.6.2 (but not 5.8.1 or later): t/mbimbf.t .......... 1/710 # Failed test at t/mbimbf.t line 38. # 'Not a HASH reference at /home/zefram/usr/perl/perl_install/perl-5.6.2-i32-f52/cpan/build/Math-BigInt-1.997-szzIPv/blib/lib/Math/BigInt.pm line 827, <DATA> line 77. # ' # doesn't match '(?-xism:^Unknown round mode 'huhmbi' at)' # Failed test at t/mbimbf.t line 40. # 'Not a HASH reference at /home/zefram/usr/perl/perl_install/perl-5.6.2-i32-f52/cpan/build/Math-BigInt-1.997-szzIPv/blib/lib/Math/BigFloat.pm line 365, <DATA> line 77. # ' # doesn't match '(?-xism:^Unknown round mode 'huhmbf' at)' # Looks like you failed 2 tests of 710. t/mbimbf.t .......... Dubious, test returned 2 (wstat 512, 0x200) Failed 2/710 subtests -zefram

Thu Sep 08 02:09:46 2011 peter.john.acklam [...] gmail.com - Correspondence added

Thanks for the report. I have not yet been able to build Perl 5.6.2 on Cygwin, so I have not yet been able to reproduce the error. I am working on it. The error puzzles me somewhat, as there have been no changes to the code related to rounding. I'll see what I can do.

Thu Sep 08 02:09:47 2011 The RT System itself - Status changed from 'new' to 'open'

Mon Jul 09 14:46:11 2012 BBYRD [...] cpan.org - Correspondence added

Jesus was still walking around when 5.6.2 was released. Please close this ticket.

Thu Oct 31 14:48:54 2013 DANAJ [...] cpan.org - Correspondence added

I've reproduced in in 5.6.2, but at this point I rather agree with BBYRD. Inside round_mode, when we call Carp::croak ("Unknown round mode '$m'"); on 5.6.2 with Carp 1.26 it tries to stringify $self, which then dies because $round_mode is a sort-kinda-not-really-a-BigInt. If we made $round_mode a hash that had a sign as well as a round_mode then that should work. Alternately, upgrade Carp to 1.32 and it no longer tries to stringify $self, and works fine. The mbimf.t test completes successfully. 64-bit 5.6.2 fails on other tests for me for the usual "SvUV turns things into doubles" reason that make me seriously hate 5.6.2, but that is a different issue.