Bug #74820 for PathTools: The XS part of Cwd ought to be optional

Thu Feb 09 09:35:04 2012 ribasushi [...] leporine.io - Ticket created

CC:	rt-cpan [...] trout.me.uk
Subject:	The XS part of Cwd ought to be optional

There is a number of different "can we build XS" checks available in the wild currently. All of them end up simply relying on ExtUtils::CBuilder->new->has_compiler (with various fallbacks when EU::CB is not available, most of them plain suck). While EU::CB itself properly installs and tests in an environment without a compiler, it depends on a specific version of File::Spec, not available on older perls. In that line of events, File::Spec itself is a pure-perl module, but Cwd has an XS portion, which PathTools insists on building, even though a PP falback is available. As a result if someone with a gcc-less system tries to install a module which configure-requires EU::CB on e.g. perl 5.8.5 using something that does respect configure-requires (e.g. cpanminus) - the installation will ultimately fail, without any clear workarounds. Is it possible to make the lack of XS a soft-failure for PathTools?

Thu Feb 09 09:36:58 2012 ribasushi [...] leporine.io - Dependency by ticket #74706 added

Thu Feb 09 09:54:50 2012 kwilliams [...] cpan.org - Correspondence added

I support the idea in theory. What I wouldn't want is for the twisty turns of CWD code to get even more complicated as a result, though. It's already pretty yucky. Maybe this is an opportunity to make things tidier? -Ken

Thu Feb 09 09:54:51 2012 The RT System itself - Status changed from 'new' to 'open'

Thu Feb 09 09:59:49 2012 ribasushi [...] leporine.io - Correspondence added

CC:	rt-cpan [...] trout.me.uk
Subject:	Re: [rt.cpan.org #74820] The XS part of Cwd ought to be optional
Date:	Thu, 09 Feb 2012 15:59:37 +0100
To:	bug-PathTools [...] rt.cpan.org
From:	Peter Rabbitson <ribasushi [...] cpan.org>

Ken Williams via RT wrote: Show quoted text

> <URL: https://rt.cpan.org/Ticket/Display.html?id=74820 > > > I support the idea in theory. What I wouldn't want is for the twisty > turns of CWD code to get even more complicated as a result, though. > It's already pretty yucky. Maybe this is an opportunity to make things > tidier? >

This concerns only the build-stage doesn't it? I am not proposing to change anything in Cwd - it already has all the necessary PP fallbacks.

Thu Feb 09 13:11:35 2012 dagolden [...] cpan.org - Correspondence added

Makes sense to me. We should try to ensure everything builds clean on a gcc-less system.

Thu Feb 09 14:24:57 2012 ribasushi [...] leporine.io - Correspondence added

RT-Send-CC:

rt-cpan [...] trout.me.uk

So what would be a way forward here? I can furnish a patch to the Makefile.PL of PathTools, which re-implements the latest EU::BC compile check and decides whether to build the .xs. Checking before I dive into this, as there may be more elegant solutions I have not thought of.