Request for review: perl-Number-Compare

[Paul Howarth]

Ralf Corsepius wrote:
> This package is the first of a lengthy series of perl modules I'd like
> to see added to FE: perl-Number-Compare.
> 
> %description
> Number::Compare compiles a simple comparison to an anonymous subroutine,
> which you can call with a value to be tested again.
> 
> Files:
> ftp://packman.iu-bremen.de/fedora/SRPMS/perl-Number-Compare.spec
> ftp://packman.iu-bremen.de/fedora/SRPMS/perl-Number-Compare-0.01-1.src.rpm
> 
> Review (and approval) appreciated, TIA

Review:

Good:
- rpmlint silent
- package naming meets naming guidelines
- spec file name matches package name
- package meets packaging guidelines
- package has suitable license (Artistic or GPL)
- spec file written in English and is legible, though I'd personally 
like to see the header fields line up...
- source md5sums match upstream
- builds OK for me on FC4 i386
- no explicit buildrequires needed
- no locale data in package
- no shared libraries in package
- no relocations
- directory ownership looks fine
- no duplicate files
- permissions fine
- package has %clean section
- no compiler optimization macros needed
- package contains code, not content
- documentation present and not excessively large
- nothing in %doc affects runtime
- no subpackages needed or present
- no .la files included

Bad:
- license text not included in package. This is very common with perl 
modules and probably not really a blocker. The pod documentation clearly 
states "This module is free software; you can redistribute it and/or 
modify it under the same terms as Perl itself."
- use of macros not quite consistent, e.g. using %{__perl} but not %{__make}

Suggestions:

- I'd expand %description, to explain more the purpose of the module:

   Number::Compare compiles a simple comparison to an anonymous
   subroutine, which you can call with a value to be tested again. Now
   this would be very pointless, if Number::Compare didn't understand
   magnitudes. The target value may use magnitudes of kilobytes (k, ki),
   megabytes (m, mi), or gigabytes (g, gi). Those suffixed with an i use
   the appropriate 2**n version in accordance with the IEC standard:
   http://physics.nist.gov/cuu/Units/binary.html

- I'd put %check after %install rather than after %clean, so that builds 
on rpm versions not supporting %check could still work.

- I'd base %install on the template in fedora-rpmdevtools rather than 
the "old-fashioned" way using find. I don't see why a manual 
INSTALLARCHLIB setting should be necessary for a noarch package either.

I'd like to hear further comments on the license text issue here.

Paul.