[DBIx-Class-Devel] Bug in Makefile.PL?

[Peter Rabbitson]

On Wed, Jan 23, 2013 at 09:29:33AM -0600, fREW Schmidt wrote:
> On Thu, Jan 24, 2013 at 01:04:36AM +1100, Peter Rabbitson wrote:
> > On Wed, Jan 23, 2013 at 07:58:43AM -0600, fREW Schmidt wrote:
> > > frew at frew.chickenkiller.com:2222, dbic is at ~/code/DBIx-Class.
> >
> > Will look later today, but possibly not - $work is not working and
> > I got to go to a pm meeting in couple hours.
> >
> > >
> > > BTW I still think Makefile.PL should exit non-zero if there was an
> > > error.
> >
> > This is not an error, this is a "do you want me to install X+Y on
> > next `make`". It is akin to ./configure. If we change it to non-0
> > then you will not be able to install this module anywhere, since the
> > normal workflow (even for cpanm) is:
> >
> > perl Makefile.PL && make && make test && make install
> >
> > In fact `exit 1` from Makefile.PL is how you get a "N/A" smoke
> > result.
> >
> 
> This sure looks like an error to me:
> 
>   !!! FAILED: Can't locate Pod/Tree.pm in @INC (@INC contains: /home/frew/code/DBIx-Class/lib inc /home/frew/code/dbic-w-datetime/lib /home/frew/perl5/perlbrew/perls/perl-5.16.1/lib/site_perl/5.16.1/i686-linux /home/frew/perl5/perlbrew/perls/perl-5.16.1/lib/site_perl/5.16.1 /home/frew/perl5/perlbrew/perls/perl-5.16.1/lib/5.16.1/i686-linux /home/frew/perl5/perlbrew/perls/perl-5.16.1/lib/5.16.1 .) at /home/frew/perl5/perlbrew/perls/perl-5.16.1/lib/site_perl/5.16.1/Pod/Inherit.pm line 383, <> line 9.
> 

https://github.com/dbsrgits/dbix-class/blob/master/maint/Makefile.PL.inc/53_autogen_pod.pl#L71

If we do not do this, then in author mode (i.e. in a git checkout) one 
would not be able to run Makefile.PL to completion at all to see what 
deps they need in the 1st place. We could certainly silence this 
entirely - it is impossible to build an actual dist due to 
https://github.com/dbsrgits/dbix-class/blob/master/maint/Makefile.PL.inc/53_autogen_pod.pl#L76 
No error trapping and the entire thing falls through.

My idea was that showing the non-fatal error was a good thing. If it is 
confusing instead - I am open to silencing it.