Bug #53836 for Test-Harness: bootstrapping close to impossible

Wed Jan 20 02:48:49 2010 ANDK [...] cpan.org - Ticket created

CC:	dagolden [...] cpan.org
Subject:	bootstrapping close to impossible

(Test-Harness-3.18) In the META.yml I find configure_requires: Module::Build: 0.35 I don't know who writes it there but it's close to impossible to fulfill unless you have it already. Because Module::Build has build_requires: Test::Harness: 3.16 One of you two should be less demanding otherwise I will get the complaints. Hrm...

Wed Jan 20 06:35:26 2010 andy [...] hexten.net - Correspondence added

On Wed Jan 20 02:48:49 2010, ANDK wrote: Show quoted text

> (Test-Harness-3.18) > > In the META.yml I find > > configure_requires: > Module::Build: 0.35 > > > I don't know who writes it there but it's close to impossible to fulfill > unless you have it already. Because Module::Build has > > build_requires: > Test::Harness: 3.16 > > One of you two should be less demanding otherwise I will get the > complaints. Hrm...

Ah - it looks like that something MB is doing. It seems to have crept in when Schwern regenerated META.yml. I guess I need to quickly get a release out that removes this dep and then figure out how to stop it permanently.

Wed Jan 20 06:35:30 2010 The RT System itself - Status changed from 'new' to 'open'

Wed Jan 20 17:39:18 2010 mschwern [...] cpan.org - Correspondence added

I don't think that just removing Module::Build from configure_requires is going to do it without the CPAN shell being a whole lot smarter. Consider the following... User has a version of Perl with no Module::Build. User fires up the CPAN shell and goes to install Test::Harness. CPAN shell is configured to prefer Build.PL. CPAN shell adds implied Module::Build configure_requires. CPAN shell goes to install Module::Build. Module::Build requires Test::Harness. Circular. The CPAN shell would have to recognize this and fall back to the Makefile.PL.

Wed Jan 20 18:03:23 2010 xdaveg [...] gmail.com - Correspondence added

CC:	ANDK [...] cpan.org
Subject:	Re: [rt.cpan.org #53836] bootstrapping close to impossible
Date:	Wed, 20 Jan 2010 18:02:45 -0500
To:	bug-Test-Harness [...] rt.cpan.org
From:	David Golden <xdaveg [...] gmail.com>

On Wed, Jan 20, 2010 at 5:39 PM, Michael G Schwern via RT <bug-Test-Harness@rt.cpan.org> wrote: Show quoted text

> <URL: https://rt.cpan.org/Ticket/Display.html?id=53836 > > > I don't think that just removing Module::Build from configure_requires > is going to do it without the CPAN shell being a whole lot smarter. > Consider the following... > > User has a version of Perl with no Module::Build. > User fires up the CPAN shell and goes to install Test::Harness. > CPAN shell is configured to prefer Build.PL. > CPAN shell adds implied Module::Build configure_requires. > CPAN shell goes to install Module::Build. > Module::Build requires Test::Harness. > > Circular.

Can Test::Harness just not ship Build.PL? (Add it to MANIFEST.SKIP) It's that or M::B has to bundle all it's dependencies like EU::MM does, I think. David

Thu Jan 21 02:51:19 2010 andreas.koenig.7os6VVqR [...] franz.ak.mind.de - Correspondence added

CC:	ANDK [...] cpan.org, dagolden [...] cpan.org
Subject:	Re: [rt.cpan.org #53836] bootstrapping close to impossible
Date:	Thu, 21 Jan 2010 08:50:54 +0100
To:	bug-Test-Harness [...] rt.cpan.org
From:	andreas.koenig.7os6VVqR [...] franz.ak.mind.de (Andreas J. Koenig)

Show quoted text

>>>>> On Wed, 20 Jan 2010 18:03:26 -0500, "David Golden via RT" <bug-Test-Harness@rt.cpan.org> said:

Show quoted text

> <URL: https://rt.cpan.org/Ticket/Display.html?id=53836 > > On Wed, Jan 20, 2010 at 5:39 PM, Michael G Schwern via RT > <bug-Test-Harness@rt.cpan.org> wrote:

>> <URL: https://rt.cpan.org/Ticket/Display.html?id=53836 > >> >> User has a version of Perl with no Module::Build. >> User fires up the CPAN shell and goes to install Test::Harness. >> CPAN shell is configured to prefer Build.PL. >> CPAN shell adds implied Module::Build configure_requires. >> CPAN shell goes to install Module::Build. >> Module::Build requires Test::Harness. >> >> Circular.

I'm willing to help out from within the cpan shell. Test::Harness has already gotten special support for very old versions. And I find it reasonable to demand from the cpan shell to ignore "prefer_installer=MB" before Module::Build itself is installed. I think this should be doable (haven't yet tried it). Show quoted text

> Can Test::Harness just not ship Build.PL? (Add it to MANIFEST.SKIP)

Show quoted text

> It's that or M::B has to bundle all it's dependencies like EU::MM does, I think.

Both variants seem a bit drastic to me. I'd only recommend them if we cannot find better approaches. Thanks, Andy for 3.19, this makes the problem much smaller. I'd wish Module::Build would not require Test::Harness 3.16. I just ran the Bodule::Build test suite with 2.64 without a problem. Am I missing something? -- andreas

Thu Jan 21 06:36:28 2010 xdaveg [...] gmail.com - Correspondence added

CC:	bug-Test-Harness [...] rt.cpan.org, ANDK [...] cpan.org
Subject:	Re: [rt.cpan.org #53836] bootstrapping close to impossible
Date:	Thu, 21 Jan 2010 06:35:35 -0500
To:	"Andreas J. Koenig" <andreas.koenig.7os6VVqR [...] franz.ak.mind.de>
From:	David Golden <xdaveg [...] gmail.com>

On Thu, Jan 21, 2010 at 2:50 AM, Andreas J. Koenig <andreas.koenig.7os6VVqR@franz.ak.mind.de> wrote: Show quoted text

> I'm willing to help out from within the cpan shell. Test::Harness has > already gotten special support for very old versions. And I find it > reasonable to demand from the cpan shell to ignore "prefer_installer=MB" > before Module::Build itself is installed. I think this should be doable > (haven't yet tried it).

Doesn't help people with older CPAN's, though. Show quoted text

> I'd wish Module::Build would not require Test::Harness 3.16. I just ran > the Bodule::Build test suite with 2.64 without a problem. Am I missing > something?

From M::B Changes: - Bumped Test::Harness prereq to 3.16 for latest PERL5LIB fixes (solves test failures when installing Module::Build using CPANPLUS::Dist::Build) [David Golden] I don't think it's unreasonable to expect a certain set of foundational toolchain modules to be Makefile.PL only -- sufficient to bootstrap M::B. I've tried to keep M::B's prereqs minimal and have bumped things where it's been necessary to ensure consistent, cross-platform behaviors. David

Sat Apr 16 07:44:52 2011 andy [...] hexten.net - Correspondence added

I believe this was fixed in 7b68accc195e159fab1412ca07a8715583e68c75: commit 7b68accc195e159fab1412ca07a8715583e68c75 Author: Andy Armstrong <andy@hexten.net> Date: Wed Apr 14 00:45:58 2010 +0100 Rename Build.PL to NotBuild.PL to avoid pulling in Module::Build as a dependency for people who've told CPAN.pm to prefer MB over EUMM. This is so a recent version of T::H can become a dependency of Test::Simple without Module::Build becoming (effectively) a transitive dependency of Test::Simple.

Sat Apr 16 07:44:53 2011 andy [...] hexten.net - Status changed from 'open' to 'resolved'