On Oct 15, 2011, at 12:25 AM, Andreas J. Koenig wrote: >>>>>> On Fri, 14 Oct 2011 11:01:23 -0500, brian d foy <[email protected]> >>>>>> said: > > bdf> [[ This message was both posted and mailed: see > bdf> the "To," "Cc," and "Newsgroups" headers for details. ]] > > bdf> In article <[email protected]>, > bdf> Christopher Fields <[email protected]> wrote: > >>> I have no idea why this is not being indexed; is it due to the module >>> versions in the META.yml file? >>> >>> Copy of META.yml and error message here: >>> >>> https://gist.github.com/1284795 >>> >>> I had this problem before with BioPerl-DB, it seems specific to the >>> 'provides' section, which is being autogenerated via Module::Build and >>> './Build dist'. Maybe the 'version' = 0? > > bdf> Why are the module versions 0? That seems really odd. Is that the > bdf> version in the module? > > I find at least two bugs involved. One is that META.yml is lying: it > reports versions to be 0 even though the modules in that package all > assign no version at all. The second bug was in the indexer that did not > treat a version of zero as a proper version number. > > I've fixed the bug in the indexer and now > CJFIELDS/BioPerl-Network-1.006901.tar.gz is indexed. I cannot guarantee > that this will work well for the enduser because endusers will have > modules without version numbers installed and they (or their software) > will possibly conclude that CPAN has indexed a different version than > what they have installed. > > If I were a user I would probably report the missing versions as a bug > in the BioPerl-Network package. > > All that said, thanks to Christopher for finding and reporting the bug > in the indexer. Would you fancy to bring the issue also to the attention > of the Module::Build people? > > bdf> What happens if they are not zero? > > Now this would be boring, everything would simply work:) > > -- > andreas
Oh, I'm sure lack of version is a bug in the main bioperl code; there is a long-standing and very hacky way of assigning VERSION to all bioperl modules that has caused headaches, and the current build process seems to bypass that. We're switching to Dist::Zilla for assigning specific versions, as we're breaking up BioPerl into smaller easier-to-maintain distributions. But the "version => '0'" is definitely off, so I'll bring it up with the Module::Build folks. chris
