[MacPorts] #51591: perl5 modules: problems with livecheck versions

[MacPorts]

#51591: perl5 modules: problems with livecheck versions
----------------------+--------------------------------
  Reporter:  mojca@…  |      Owner:  macports-tickets@…
      Type:  defect   |     Status:  closed
  Priority:  Normal   |  Milestone:
 Component:  ports    |    Version:  2.3.4
Resolution:  fixed    |   Keywords:
      Port:           |
----------------------+--------------------------------
Changes (by devans@…):

 * status:  new => closed
 * resolution:   => fixed


Comment:

 I've made some changes to the perl5 port group to help with this.  The
 thing to realize is that search.cpan.org and api.metacpan.org are two
 different search engines but they both operate on the same CPAN
 repository.  metacpan has some advantages over the traditional
 search.cpan.org but is not perfect and it is acknowledged that there are
 cases where search.cpan.org works better but its hard to tell why because,
 while the source for metacpan is open, the source for search.cpan.org is
 not.

 After getting all the perl modules up to date after my vacation, I found 3
 categories of problems to address of which the subject of this ticket is
 one.


 {{{
 1) found by search.cpan but not by metacpan

 Error: cannot check if was updated (The requested URL returned error: 404
 Not Found)
 Error: cannot check if p5-pgtop was updated (The requested URL returned
 error: 404 Not Found)
 Error: cannot check if p5-podtohtml was updated (The requested URL
 returned error: 404 Not Found)
 Error: cannot check if p5-sgmlspm was updated (The requested URL returned
 error: 404 Not Found)

 2) search.cpan version > metacpan version  (this ticket)

 p5-business-issn seems to have been updated (port version: 1.001, new
 version: 0.91)
 p5-class-dbi-pg seems to have been updated (port version: 0.09, new
 version: 0.08)
 p5-config-simple seems to have been updated (port version: 4.59, new
 version: 4.58)
 p5-list-uniq seems to have been updated (port version: v0.21.0, new
 version: 0.20)
 p5-net-senderbase seems to have been updated (port version: 1.02, new
 version: 1.01)
 p5-svg-graph seems to have been updated (port version: 0.04, new version:
 0.02)
 p5-sys-meminfo seems to have been updated (port version: 0.98, new
 version: 0.91)

 3) metacpan version > search.cpanversion

 p5-attribute-handlers seems to have been updated (port version: 0.96, new
 version: 0.99)
 p5-class-makemethods seems to have been updated (port version: 1.009, new
 version: 1.01)
 p5-mail-sender seems to have been updated (port version: 0.900003, new
 version: 0.901)
 p5-net-ftpserver seems to have been updated (port version: 1.122, new
 version: 1.125)

 }}}

 Category 1 & 2 address ports that are better served by search.cpan.org
 than metacpan.org.  Category 3 is where metacpan.org works better.

 To address these issues, I've made two changes to the perl5 port group:

 In r150080, I changed the default homepage to use metacpan.org to match
 livecheck.  This fixed the problem associated with Category 3 where
 previously the metacpan livecheck (which was the correct latest version)
 disagreed with the search.cpan.org homepage as to module version.

 In r150321, I added an additional boolean option
 (perl5.use_search_cpan_org) to perl5.setup that allows one to specify that
 search.cpan.org be
 used instead of metacpan.org.  The default here is false, use
 metacpan.org, which works best for the vast majority on ports.  But for
 the few ports in
 Category 1 & 2, setting this option to yes in the Portfile fixes the
 problem by using search.cpan.org for livecheck and also for the homepage
 so
 that again homepage and livecheck agree on latest version.

 This is an empirical approach, not an analytical one but seems to fix the
 problems at hand.

 From what I can tell, many of these ports are older ones that may not
 provide all the metadata that metacpan is looking for and I suspect have
 been special cased in search.cpan.org but I haven't done the work to be
 able to see exactly what is missing or what specifically metacpan is
 looking for.
 As you say, getting the upstream author to fix any deficiencies so that
 the offending ports are correctly indexed by metacpan would be the best
 solution.

 After applying these changes to the portgroup and adjusting the various
 problem ports, the only livecheck failure is p5-mecab which is a different
 issue altogether (#47338).

 For now, calling this fixed.

-- 
Ticket URL: <https://trac.macports.org/ticket/51591#comment:6>
MacPorts <https://www.macports.org/>
Ports system for OS X