Bug #105742 for IO-Socket-SSL: IO::Socket::IP support does not work without network

Wed Jul 08 03:48:51 2015 https://www.opensuse.org/openid/user/coolo - Ticket created

Subject:

IO::Socket::IP support does not work without network

As we updated perl to 5.22.0 in the openSUSE distribution, we noticed that IO-Socket-SSL fails to run its test suite. After some debugging, it turned out that with new perl, IO-Socket-SSL switches from ::INET to ::IP and that module uses getaddrinfo, which fails to resolve the port numbers given by the ::SSL test suite if there is no proper network in the build environment. We disable network for reproduciblity and security. The good news: you can disable that feature of ::IP, but you have to add it in every constructor call in the test suite. Attached is the patch that makes it build in the Open Build Service again. Hope you can integrate that in one way or another in a future upstream release. (I used spaces where you used tabs, I noticed too late - so don't get fooled by applying as is)

Subject:

perl-IO-Socket-SSL_fix_offline.patch

Message body is not shown because it is too large.

Wed Jul 08 04:37:34 2015 Steffen_Ullrich [...] genua.de - Correspondence added

Am Mi 08. Jul 2015, 03:48:51, https://www.opensuse.org/openid/user/coolo schrieb: Show quoted text

> As we updated perl to 5.22.0 in the openSUSE distribution, we noticed > that IO-Socket-SSL fails to run its test suite. > > After some debugging, it turned out that with new perl, IO-Socket-SSL > switches from ::INET to ::IP and that module uses getaddrinfo, which > fails to resolve the port numbers given by the ::SSL test suite if > there is no proper network in the build environment. We disable > network for reproduciblity and security. > > The good news: you can disable that feature of ::IP, but you have to > add it in every constructor call in the test suite.

I would assume that this would not only affect IO::Socket::SSL but any module and test using IO::Socket::IP. I also don't think that a default of using AI_ADDRCONFIG is a bad idea, since it requires anybody to explicitly disable this default when using software on an unconnected system with only localhost, which is not uncommon setup. It also behaves differently to existing modules like IO::Socket::INET. Thus I would recommend that this should be fixed in IO::Socket::IP and not in every module using IO::Socket::IP. Please consider to file this bug against IO::Socket::IP instead.

Wed Jul 08 04:37:35 2015 The RT System itself - Status changed from 'new' to 'open'

Wed Jul 08 04:45:22 2015 Steffen_Ullrich [...] genua.de - Correspondence added

Show quoted text

> Thus I would recommend that this should be fixed in IO::Socket::IP > and not in every module using IO::Socket::IP. Please consider to > file this bug against IO::Socket::IP instead.

Actually it looks like there is already a similar bug report against IO::Socket::IP, which is unrelated to IO::Socket::SSL. https://rt.cpan.org/Public/Bug/Display.html?id=104793

Wed Jul 08 05:08:26 2015 https://www.opensuse.org/openid/user/coolo - Correspondence added

From:

coolo [...] suse.de

Am Mi 08. Jul 2015, 04:45:22, SULLR schrieb: Show quoted text

>

> > Thus I would recommend that this should be fixed in IO::Socket::IP > > and not in every module using IO::Socket::IP. Please consider to > > file this bug against IO::Socket::IP instead.

> > Actually it looks like there is already a similar bug report against > IO::Socket::IP, which is unrelated to IO::Socket::SSL. > https://rt.cpan.org/Public/Bug/Display.html?id=104793

Actually Socket::IP's test suite uses the same flag all over the place. So the author is aware of the problem - but the bug shows little activity ;( Greetings, Stephan

Sun Aug 23 15:10:00 2015 Steffen_Ullrich [...] genua.de - Correspondence added

On Wed Jul 08 05:08:26 2015, https://www.opensuse.org/openid/user/coolo wrote: Show quoted text

> Am Mi 08. Jul 2015, 04:45:22, SULLR schrieb:

> >

> > > Thus I would recommend that this should be fixed in IO::Socket::IP > > > and not in every module using IO::Socket::IP. Please consider to > > > file this bug against IO::Socket::IP instead.

> > > > Actually it looks like there is already a similar bug report against > > IO::Socket::IP, which is unrelated to IO::Socket::SSL. > > https://rt.cpan.org/Public/Bug/Display.html?id=104793

> > Actually Socket::IP's test suite uses the same flag all over the > place. So the author is aware of the problem - but the bug shows > little activity ;(

I currently don't think that I should work around bugs in other modules if the author is already aware of the problem. It would probably be different if the module author actively declared that this is not a bug but a feature and one need to use the module the correct way. But thanks for pointing out the issue and if it gets too bad I either try to convince the module author to fix the problem or need to work around it myself.

Sun Aug 23 15:10:08 2015 Steffen_Ullrich [...] genua.de - Status changed from 'open' to 'rejected'