Bug #89435 for Dist-Zilla-Plugin-MakeMaker-Fallback: Warning on top of Makefile.PL

Sun Oct 13 10:15:07 2013 SREZIC [...] cpan.org - Ticket #89434: Ticket created

Subject:

Warning on top of Makefile.PL

The warning on top of the Makefile.PL should be removed. If both Makefile.PL and Build.PL are shipped with a distribution, then both should work. If any of both does not work, then it should be removed. At least CPAN.pm does the right thing here if just Build.PL exists, and users installing manually should read the README or INSTALL document anyway (which should point to Build.PL). If Makefile.PL works, then the warning is meaningless anyway. Please note that CPAN.pm has the option prefer_installer which can be set to EUMM (prefer Makefile.PL over Build.PL) and RAND (choose randomly either Makefile.PL or Build.PL). Please also note that it seems that Module::Build will be removed from perl core, so I don't see why Build.PL should be preferred over Makefile.PL nowadays. Regards, Slaven

Sun Oct 13 10:26:20 2013 SREZIC [...] cpan.org - Ticket created

Subject:

Warning on top of Makefile.PL

I just realized that the warning described in https://rt.cpan.org/Ticket/Display.html?id=89434 is generated by this Dist-Zilla recipe. Probably it's better to discuss the issue here. Regards, Slaven

Sun Oct 13 10:26:50 2013 SREZIC [...] cpan.org - Ticket #89434: Merged into ticket #89435

Sun Oct 13 10:26:50 2013 SREZIC [...] cpan.org - Merged into ticket #89435

Sun Oct 13 12:54:15 2013 ether [...] cpan.org - Taken

Mon Oct 14 21:30:50 2013 ether [...] cpan.org - Correspondence added

On 2013-10-13 07:15:07, SREZIC wrote: Show quoted text

> The warning on top of the Makefile.PL should be removed. > > If both Makefile.PL and Build.PL are shipped with a distribution, then > both should work.

Both do work, so that's not thee issue here. Show quoted text

> If Makefile.PL works, then the warning is meaningless anyway.

I think you're missing a bit of context here -- the fallback Makefile.PL is present explicitly for clients that are too old to understand 'configure_requires', which points to Module::Build::Tiny. Show quoted text

> Please also note that it > seems that Module::Build will be removed from perl core, so I don't > see why Build.PL should be preferred over Makefile.PL nowadays.

Being removed from core is a long way from being deprecated, and it's the Module::Build *implementation itself* that has fallen into disfavour, not the general idea of an extensible module builder. I'm using Module::Build::Tiny, which bears no resemblance to Module::Build other than both being driven by a Build.PL file that follows the same interface. I've updated the plugin that generates the Makefile.PL to only display the warning if the configure_requires prerequisites genuinely are missing, which I think should answer the root of your concern.

Mon Oct 14 21:30:52 2013 The RT System itself - Status changed from 'new' to 'open'

Mon Oct 14 21:30:53 2013 ether [...] cpan.org - Status changed from 'open' to 'resolved'

Tue Oct 15 03:58:20 2013 slaven [...] rezic.de - Correspondence added

CC:	SREZIC [...] cpan.org
Subject:	Re: [rt.cpan.org #89435] Warning on top of Makefile.PL
Date:	Tue, 15 Oct 2013 09:57:42 +0200
To:	bug-Dist-Zilla-Plugin-MakeMaker-Fallback [...] rt.cpan.org
From:	Slaven Rezic <slaven [...] rezic.de>

"Karen Etheridge via RT" <bug-Dist-Zilla-Plugin-MakeMaker-Fallback@rt.cpan.org> writes: Show quoted text

> <URL: https://rt.cpan.org/Ticket/Display.html?id=89435 > > > On 2013-10-13 07:15:07, SREZIC wrote:

>> The warning on top of the Makefile.PL should be removed. >> >> If both Makefile.PL and Build.PL are shipped with a distribution, then >> both should work.

> > Both do work, so that's not thee issue here. >

>> If Makefile.PL works, then the warning is meaningless anyway.

> > I think you're missing a bit of context here -- the fallback Makefile.PL is present explicitly for clients that are too old to understand 'configure_requires', which points to Module::Build::Tiny. >

Makefile.PL is run if CPAN.pm is configured with prefer_installer=EUMM or RAND. This has nothing to do with "old clients". In fact, I saw the warning with a perl 5.18 installation, which is probably much newer than the average installed perl. Show quoted text

>> Please also note that it >> seems that Module::Build will be removed from perl core, so I don't >> see why Build.PL should be preferred over Makefile.PL nowadays.

> > Being removed from core is a long way from being deprecated, and > it's the Module::Build *implementation itself* that has fallen into > disfavour, not the general idea of an extensible module builder. > I'm using Module::Build::Tiny, which bears no resemblance to > Module::Build other than both being driven by a Build.PL file > that follows the same interface. > > I've updated the plugin that generates the Makefile.PL to only > display the warning if the configure_requires prerequisites > genuinely are missing, which I think should answer the root > of your concern.

Still I think the warning text is misleading. It would still be shown if a configure_requires prerequisite is missing due to reasons which are not in the realm of the client (e.g. a test failure happened, so the prerequisite was not installed). A more defensive warning text would probably be better here. Regards, Slaven -- Slaven Rezic - slaven <at> rezic <dot> de Berlin Perl Mongers - http://berlin.pm.org

Tue Oct 15 05:55:06 2013 rt-cpan [...] trout.me.uk - Correspondence added

On Tue Oct 15 03:58:20 2013, slaven@rezic.de wrote: Show quoted text

> Makefile.PL is run if CPAN.pm is configured with prefer_installer=EUMM > or RAND. This has nothing to do with "old clients". In fact, I saw the > warning with a perl 5.18 installation, which is probably much newer > than > the average installed perl.

Unless you're explicitly and intentionally testing the Makefile.PL that's almost certainly the wrong configuration to use. If you want to propose an alternate wording that expands to 'old or probably misconfigured', please go ahead - but really nobody should be touching prefer_installer without being willing to inform themselves about the results thereof, and calling a setting of anything except MB valid except for experts who already know what they're doing would be foolish.

Tue Oct 15 10:51:35 2013 slaven [...] rezic.de - Correspondence added

CC:	SREZIC [...] cpan.org
Subject:	Re: [rt.cpan.org #89435] Warning on top of Makefile.PL
Date:	Tue, 15 Oct 2013 16:51:04 +0200
To:	bug-Dist-Zilla-Plugin-MakeMaker-Fallback [...] rt.cpan.org
From:	Slaven Rezic <slaven [...] rezic.de>

"MSTROUT via RT" <bug-Dist-Zilla-Plugin-MakeMaker-Fallback@rt.cpan.org> writes: Show quoted text

> <URL: https://rt.cpan.org/Ticket/Display.html?id=89435 > > > On Tue Oct 15 03:58:20 2013, slaven@rezic.de wrote:

>> Makefile.PL is run if CPAN.pm is configured with prefer_installer=EUMM >> or RAND. This has nothing to do with "old clients". In fact, I saw the >> warning with a perl 5.18 installation, which is probably much newer >> than >> the average installed perl.

> > Unless you're explicitly and intentionally testing the Makefile.PL that's almost certainly the wrong configuration to use. > > If you want to propose an alternate wording that expands to 'old or probably misconfigured', please go ahead - but really nobody should be touching prefer_installer without being willing to inform themselves about the results thereof, and calling a setting of anything except MB valid except for experts who already know what they're doing would be foolish. >

Actually, with the 0.03 release the warning is now issued only if some configure_requires modules are missing. It's not anymore a Makefile.PL vs. Build.PL issue. In fact, the Build.PL could also be improved by prepending a similar BEGIN block (for users running Build.PL manually). I created https://github.com/karenetheridge/Dist-Zilla-Plugin-MakeMaker-Fallback/pull/1 Regards, Slaven -- Slaven Rezic - slaven <at> rezic <dot> de BBBike - route planner for cyclists in Berlin WWW version: http://www.bbbike.de Perl/Tk version for Unix and Windows: http://bbbike.sourceforge.net

Tue Oct 15 14:46:01 2013 rt-cpan [...] trout.me.uk - Correspondence added

On Tue Oct 15 10:51:35 2013, slaven@rezic.de wrote: Show quoted text

> > If you want to propose an alternate wording that expands to 'old or > > probably misconfigured', please go ahead

> > I created > https://github.com/karenetheridge/Dist-Zilla-Plugin-MakeMaker- > Fallback/pull/1

Which is a complete waste of time. I am asking you to add text to note that there is a possibility you have a recent toolchain with a non-standard configuration. Deleting the most important part of the warning is not acceptable. Also, please *propose* alternate wording, as I requested - blind pull requests will not be accepted.

Tue Oct 15 18:32:29 2013 slaven [...] rezic.de - Correspondence added

CC:	SREZIC [...] cpan.org
Subject:	Re: [rt.cpan.org #89435] Warning on top of Makefile.PL
Date:	Wed, 16 Oct 2013 00:29:24 +0200
To:	bug-Dist-Zilla-Plugin-MakeMaker-Fallback [...] rt.cpan.org
From:	Slaven Rezic <slaven [...] rezic.de>

"MSTROUT via RT" <bug-Dist-Zilla-Plugin-MakeMaker-Fallback@rt.cpan.org> writes: Show quoted text

> <URL: https://rt.cpan.org/Ticket/Display.html?id=89435 > > > On Tue Oct 15 10:51:35 2013, slaven@rezic.de wrote:

>> > If you want to propose an alternate wording that expands to 'old or >> > probably misconfigured', please go ahead

>> >> I created >> https://github.com/karenetheridge/Dist-Zilla-Plugin-MakeMaker- >> Fallback/pull/1

> > Which is a complete waste of time. > > I am asking you to add text to note that there is a possibility you have a recent toolchain with a non-standard configuration. > > Deleting the most important part of the warning is not acceptable. > > Also, please *propose* alternate wording, as I requested - blind pull requests will not be accepted. >

Matt, Can you show me the place in CPAN.pm's documentation proving that setting prefer_installer=EUMM is considered a "misconfiguration"? Or in the CPANPLUS documentation which has a similar option? I did not find such a place. So I was not really inclined to propose a wording based on this claim. Actually, I am not deeply interested in this issue, and it seems that there are huge cultural differences in getting a solution here. This RT ticket is resolved, the github pull request rejected & closed --- probably we should leave it at this. Regards, Slaven -- Slaven Rezic - slaven <at> rezic <dot> de Berlin Perl Mongers - http://berlin.pm.org

Thu Oct 23 18:23:24 2014 ether [...] cpan.org - Fixed in 0.003 added