Re: [PATCH 3/3] gnu: Add perl-log-any-adapter-log4perl.

[Alex Sassmannshausen <alex@pompo.co>]

On 17 Jul 2016 12:17, Alex Sassmannshausen <alex@pompo.co> wrote:
>
> Hi Ricardo,
>
> On 17 Jul 2016 10:44, Ricardo Wurmus <rekado@elephly.net> wrote:
> >
> >
> > Alex Sassmannshausen <alex.sassmannshausen@gmail.com> writes: 
> >
> > > * gnu/packages/perl.scm (perl-log-any-adapter-log4perl): New variable. 
> > > --- 
> > >  gnu/packages/perl.scm | 27 +++++++++++++++++++++++++++ 
> > >  1 file changed, 27 insertions(+) 
> > > 
> > > diff --git a/gnu/packages/perl.scm b/gnu/packages/perl.scm 
> > > index fc63df3..1f18214 100644 
> > > --- a/gnu/packages/perl.scm 
> > > +++ b/gnu/packages/perl.scm 
> > > @@ -2970,6 +2970,33 @@ whether the application has chosen (or will ever choose) a logging 
> > >  mechanism.") 
> > >      (license (package-license perl)))) 
> > >  
> > > +(define-public perl-log-any-adapter-log4perl 
> > > +  (package 
> > > +    (name "perl-log-any-adapter-log4perl") 
> > > +    (version "0.08") 
> > > +    (source 
> > > +     (origin 
> > > +       (method url-fetch) 
> > > +       (uri (string-append 
> > > +             "mirror://cpan/authors/id/D/DA/DAGOLDEN/Log-Any-Adapter-Log4perl-" 
> > > +             version 
> > > +             ".tar.gz")) 
> > > +       (sha256 
> > > +        (base32 
> > > +         "0xf4fnrsznvg0hf36q481124ja1hp3lybki1xjgnk82z9990jmxn")))) 
> > > +    (build-system perl-build-system) 
> > > +    (native-inputs `(("perl-log-any" ,perl-log-any))) 
> > > +    (inputs 
> > > +     `(("perl-log-any" ,perl-log-any) 
> > > +       ("perl-log-log4perl" ,perl-log-log4perl))) 
> >
> > Shouldn’t these inputs be propagated inputs instead?  Perl doesn’t have 
> > anything comparable to the RUNPATH feature, so inputs must be 
> > propagated. 
> >
> > Have you tested using this module, e.g. in a container?
>
> You are quite right, I have now tested loading the module from within a container in a test script, and I do not to propagate the inputs.
>
> I will update the patch - and also look at the perl importer, as it seems to add inputs rather than propagated inputs by default.
>
> Thanks for your review!
>
> Alex

Looking at this, there seems to be a larger problem in perl.scm with this. Many packages simply specify inputs, and I believe the same reasoning that applied to my package her will apply to those too.

Binaries are not affected by this problem, as we generally use wrapping scripts to explicitly include all dependencies in PERL5LIB.

Am I right in thinking that we should be systematically changing inputs to propagated-inputs in the whole of perl.scm?

Confused but eager to help,

Alex