Bug #97050 for IO-Socket-IP: version 0.30 breaks cpanm-reporter and cpan-upload

Tue Jul 08 06:19:17 2014 KENTNL [...] cpan.org - Ticket created

Subject:

version 0.30 breaks cpanm-reporter and cpan-upload

There's something seriously awry here.

0.30 installed cause both cpan-upload and cpanm-reporter to consistently return 500 "Connection reset by peer" and other problems:

http://paste.perldancer.org/2wQw5qhUf1XYo

Its not glaringly obvious why, but upgrading broke many things, and a downgrade to 0.29 unbroke all the things.

Unfortunately, both the stacks for those things are massive, and I only found the POF by luck because I had a perl that wasn't broken, and a perl that was, and it was a cause of liberal application of PERL5OPT=-d:TraceUse and process of elimination, which quickly found a this module as being the common failure.

Marking critical because this fundementally breaks two of my most common toolchain modules, and it caused a minor twitter explosion. https://twitter.com/creaktive/status/486059242416504832

Tue Jul 08 06:28:53 2014 leonerd-cpan [...] leonerd.org.uk - Correspondence added

On Tue Jul 08 06:19:17 2014, KENTNL wrote: Show quoted text

> Marking critical because this fundementally breaks two of my most > common > toolchain modules, and it caused a minor twitter explosion. > https://twitter.com/creaktive/status/486059242416504832

The differences are: https://metacpan.org/diff/file?target=PEVANS/IO-Socket-IP-0.30/&source=PEVANS/IO-Socket-IP-0.29/ -- Paul Evans

Tue Jul 08 06:28:54 2014 The RT System itself - Status changed from 'new' to 'open'

Tue Jul 08 06:30:54 2014 leonerd-cpan [...] leonerd.org.uk - Correspondence added

On Tue Jul 08 06:28:53 2014, PEVANS wrote: Show quoted text

> On Tue Jul 08 06:19:17 2014, KENTNL wrote:

> > Marking critical because this fundementally breaks two of my most > > common > > toolchain modules, and it caused a minor twitter explosion. > > https://twitter.com/creaktive/status/486059242416504832

> > The differences are: > > https://metacpan.org/diff/file?target=PEVANS/IO-Socket-IP- > 0.30/&source=PEVANS/IO-Socket-IP-0.29/

Seems to be caused by the bugfix for RT95988 (https://rt.cpan.org/Ticket/Display.html?id=95983) -- Paul Evans

Tue Jul 08 06:33:49 2014 KENTNL [...] cpan.org - Correspondence added

It would appear the change in 0.30 is breaking IO::Socket::SSL, because IO::Socket::SSL overrides 'connect' and calls connect later via ->SUPER::connect() in the middle of a complex method.

https://metacpan.org/source/SULLR/IO-Socket-SSL-1.994/lib/IO/Socket/SSL.pm#L495

... which I think means SSL handshake stops happening entirely.

And both those tools are connecting via SSL.

So of course, the end point is going "Thats not SSL, Goodbye".

Tue Jul 08 06:51:11 2014 leonerd-cpan [...] leonerd.org.uk - Correspondence added

$ perlmodversion IO::Socket::IP 0.29 leo@shy:~/src/perl/IO-Socket-SSL-1.994 $ make test ... Result: PASS $ perlmodversion IO::Socket::IP 0.30 leo@shy:~/src/perl/IO-Socket-SSL-1.994 $ make test PERL_DL_NONLAZY=1 /usr/bin/perl "-MExtUtils::Command::MM" "-e" "test_harness(0, 'blib/lib', 'blib/arch')" t/*.t t/external/*.t t/01loadmodule.t .................. 1/3 # openssl version=0x10001080 # Net::SSLeay::VERSION=1.63 t/01loadmodule.t .................. ok t/acceptSSL-timeout.t ............. Dubious, test returned 1 (wstat 256, 0x100) Failed 9/15 subtests t/auto_verify_hostname.t .......... 1/30 ^C (just sat there and I SIGINTed it) -- Paul Evans

Tue Jul 08 06:51:12 2014 leonerd-cpan [...] leonerd.org.uk - Fixed in 0.30 deleted

Tue Jul 08 07:04:31 2014 leonerd-cpan [...] leonerd.org.uk - Cc SULLR added

Tue Jul 08 07:05:09 2014 leonerd-cpan [...] leonerd.org.uk - Correspondence added

Subject:

Version 0.30 breaks IO-Socket-SSL

+Cc SULLR as it relates closely to IO-Socket-SSL -- Paul Evans

Tue Jul 08 12:59:09 2014 CARNIL [...] cpan.org - Cc CARNIL added

Thu Jul 10 01:49:29 2014 Steffen_Ullrich [...] genua.de - Correspondence added

Am Di 08. Jul 2014, 07:05:09, PEVANS schrieb: Show quoted text

> +Cc SULLR as it relates closely to IO-Socket-SSL

The problem lies in the interaction between IO::Socket calling hierarchy (new -> configure -> connect), sub-classing IO::Socket, the intention do to multi-homing inside IO::Socket::IP and the difference between transport layer problems (fixable by multi-homing) and application layer problems (not fixable by multi-homing). A more detailed description and a proposed patch can be found at https://rt.cpan.org/Ticket/Display.html?id=95983#txn-1384383.

Thu Jul 10 13:02:02 2014 ether [...] cpan.org - Correspondence added

Please delete this problematic release from the CPAN until a new release is available with a fix - that will help contain the damage - thanks!

Fri Jul 11 13:07:19 2014 leonerd-cpan [...] leonerd.org.uk - Correspondence added

Fixed by 0.31 -- Paul Evans

Fri Jul 11 13:07:21 2014 leonerd-cpan [...] leonerd.org.uk - Status changed from 'open' to 'resolved'

Fri Jul 11 13:07:22 2014 leonerd-cpan [...] leonerd.org.uk - Fixed in 0.31 added

Fri Jul 11 17:27:48 2014 Steffen_Ullrich [...] genua.de - Correspondence added

Subject:	Re: [rt.cpan.org #97050] version 0.30 breaks cpanm-reporter and cpan-upload
Date:	Fri, 11 Jul 2014 23:27:26 +0200
To:	Paul Evans via RT <bug-IO-Socket-IP [...] rt.cpan.org>
From:	Steffen Ullrich <Steffen_Ullrich [...] genua.de>

On Fri, Jul 11, 2014 at 01:07:20PM -0400, Paul Evans via RT <bug-IO-Socket-IP@rt.cpan.org> wrote: Show quoted text

> <URL: https://rt.cpan.org/Ticket/Display.html?id=97050 > > > Fixed by 0.31

Thanks a lot. Steffen -- genua Gesellschaft fuer Netzwerk- und Unix-Administration mbH Domagkstrasse 7, 85551 Kirchheim bei Muenchen tel +49 89 991950-0, fax -999, www.genua.de Geschaeftsfuehrer: Dr. Magnus Harlander, Dr. Michaela Harlander, Bernhard Schneck. Amtsgericht Muenchen HRB 98238