Bug #57224 for CGI: CGI::Carp works incorrectly if we have an exception in a package of one of parents

Wed May 05 08:24:20 2010 UNERA [...] cpan.org - Ticket created

Subject:

CGI::Carp works incorrectly if we have an exception in a package of one of parents

Debian bugreport: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=580356 Today I've found an unpleasant bug in CGI::Carp. In attache You can find test script showing the problem. If we have more than one packages in parent list, and one of them throws an exception from BEGIN section, CGI::Carp can't detect $@ error-text properly. If You start test.pl You will see the following error-text: ... [Wed May 5 15:33:31 2010] test.pl: Died at /usr/share/perl/5.10/base.pm line 93. [Wed May 5 15:33:31 2010] test.pl: BEGIN failed--compilation aborted at Package1.pm line 5. Compilation failed in require at test.pl line 11. If You comment text 'use CGI::Carp' You can see the following text: 123 at Package3.pm line 12. Compilation failed in require at (eval 3) line 3. ...propagated at /usr/share/perl/5.10/base.pm line 93. BEGIN failed--compilation aborted at Package1.pm line 5. Compilation failed in require at test.pl line 11. This bug appears only if we have more than one packages in parent-list.

Subject:

eval.tgz

Download eval.tgz
application/x-gzip 417b

Message body not shown because it is not plain text.

Wed May 05 10:35:44 2010 yanick+cpan [...] babyl.dyndns.org - Correspondence added

I only had time to quickly throw a look at it, but it seems that the problem has to do with the fact that 'base' also muck with the 'die' sighandler. In base.pm, the crux seems to be in the 'import' function. I'll have to dig deeper to say something more enlightning, but it could be a case where the fix for this corner case might be a rather big headache.

Wed May 05 10:35:46 2010 The RT System itself - Status changed from 'new' to 'open'

Thu May 06 04:10:42 2010 UNERA [...] cpan.org - Correspondence added

Срд Май 05 10:35:44 2010, YANICK писал: Show quoted text

> I only had time to quickly throw a look at it, > but it seems that the problem has to do with the fact that > 'base' also muck with the 'die' sighandler. > > In base.pm, the crux seems to be in the 'import' function. > > I'll have to dig deeper to say something more enlightning, > but it could be a case where the fix for this corner case > might be a rather big headache.

No :) I think this is the CGI::Carp's bug, not in 'base'. In attache You can find another script showes this problem. There is no dependencises exclude CGI::Carp. If we use CGI::Carp, we can see the following error-text: <h1>Software error:</h1> <pre> at test.pl line 18.</pre> If we comment 'use CGI::Carp', we can see properly error-text: 123 at test.pl line 13. ...propagated at test.pl line 18.

Subject:

test.pl

#!/usr/bin/perl use warnings; use strict; use utf8; use open qw(:std :utf8); use CGI::Carp qw(fatalsToBrowser); sub foo() { die 123; } eval { foo(); }; die if $@;

Thu May 06 19:50:57 2010 yanick [...] babyl.dyndns.org - Correspondence added

Subject:	Re: [rt.cpan.org #57224] CGI::Carp works incorrectly if we have an exception in a package of one of parents
Date:	Thu, 6 May 2010 19:50:41 -0400
To:	bug-CGI.pm [...] rt.cpan.org
From:	Yanick Champoux <yanick [...] babyl.dyndns.org>

On May 6, 2010 04:10:45 am you wrote: Show quoted text

> No :) I think this is the CGI::Carp's bug, not in 'base'.

I didn't say it was a bug in base, I said it was probably caused by the meddling of 'die' on both side, and I was somewhat right. In both your examples, CGI::Carp::die is called without argument. The real die propagate the already-existing value of $@, if there is any. C::C::die wasn't doing the same thing. Tada. Mystery solved. :-) A patch to have C::C::die do the right thing is resting at http://github.com/yanick/CGI.pm/tree/rt-57224 Joy, `/anick

Fri Dec 24 16:52:21 2010 MARKSTOS [...] cpan.org - Taken

Fri Dec 24 16:52:38 2010 MARKSTOS [...] cpan.org - Status changed from 'open' to 'patched'

Sun Jan 23 20:45:34 2011 MARKSTOS [...] cpan.org - Correspondence added

Subject:

patch released for CGI.pm

Thanks for the bug report. A patch for it appeared in 3.51, if not sooner. Resolving. Mark

Sun Jan 23 20:45:35 2011 The RT System itself - Status changed from 'patched' to 'open'

Sun Jan 23 20:45:35 2011 MARKSTOS [...] cpan.org - Status changed from 'open' to 'resolved'

Fri May 23 14:29:35 2014 The RT System itself - Queue changed from CGI.pm to CGI

Bug #57224 for CGI: CGI::Carp works incorrectly if we have an exception in a package of one of parents

Preferred bug tracker