I seem to be the guy who everyone is waiting on for a (finally) Module::Build release, which will allow the rest of the world access to all of the fine features which have already shipped in the Module::Build in perl 5.10.0, as well as the new TAP::Harness support and various nifty bugfixes which would make installing Perl modules a great deal less painful. After a couple of alpha releases, I realized that the only feature I've tried to add in the last 1.5 years needed to be reverted because the existing documentation had never been fully tested (so when I added my code and a test for my code, everything passed!) and the code involved had too many implicit behaviors which won't yield to anything resembling light refactoring.
And that used up what remaining enthusiasm I had.
Meanwhile, to look at the cpantesters results for most of my distributions, I still need to build a bot to filter out all of the testers who are running an old or misconfigured toolchain which either invents a Makefile.PL or prefers the Makefile.PL over Build.PL -- your basic rock and a hard place.
And that has left me with little in reserve.
But I'm sort of getting over all of that by playing with rakudo, which gives me a shred of confidence that progress actually can be made and that software does not need to be forever locked in the death grip of compatibility.
So, I figured I would give maintaining Module::Build another chance, and I shipped another alpha build and looked at some test reports and noticed that Perl 5.6.2 has been failing because of a new test which probably only points out a breakage which was already there in t/ext.t (and some other issue in t/xs.t that I haven't had a will to chase.)
And the thing is: I don't care about 5.6.2.
Perl 5.6.2 was already outdated when I started using Perl in 2002, and while I can certainly build one (and already did long ago) on my newer machine, I have absolutely zero interest in supporting it or debugging for it simply for the sake of compatibility. I want progress.
But due to the nature of the CPAN, I cannot release a new Module::Build without breaking every upgrades and installations for every Module::Build based distribution (over 2000 of them) on 5.6.2.
So, is anybody actively using 5.6.2? If so, how do you propose we solve this?
It's not like this is a big bug or something I can't figure out, I just do not have any incentive to deal with 5.6.2 because I would have to be a masochist to put extra effort toward enabling behavior which I would prefer to discourage. If somebody wants to sort out the problem and send a patch, I'll happily apply it and ship it.
In my opinion, the problem runs deeper than just this release and somebody needs to actively keep on top of this sort of thing or I'll constantly be sitting here holding the release axe and muttering "grumble ... 5.6.2 ... what have you done for me lately".
As far as I can tell, 5.6.2 users are phantoms, yet anything touching the toolchain has to tiptoe around them ever so carefully? We could, of course, "just fix the CPAN", but this too would primarily benefit users who don't upgrade their software. So, if nobody is motivated enough to do the work or pay to have it done, the easy answer is to not do it.
Maintaining Module::Build for multiple platforms is difficult enough (especially with e.g. nobody testing svn on Win32), but I can at least pretend to care about other platforms (mostly because M::B gets patches for them.)
