Jeffrey Thalhammer <[EMAIL PROTECTED]> wrote:
> Randy Kobes distributes Win32 PPMs for some of the
> modules that ActiveState doesn't provide. It is not
> entirely automated, so the latest code isn't always
> available. But Randy is very helpful if there's
> anything you want to see.
>
> http://theoryx5.uwinnipeg.ca/ppms/
What is actually happening on ppm.activestate.com, is that only
modules that pass all their unit tests are packaged for the general public.
IMHO this is a great idea, since then people who are "ppm installing" stuff
from activestate repoes can be reasonably(*) confident that the package will
work on their system.
Part of the problem is that a lot of modules out there are fully
functional even when a few of their tests fail due to assumptions about the
environment they are being tested in. Another part is that the ActiveState
perl package build process ("cpanrun") doesn't behave exactly the same way
as CPAN::YACSmoke. So, a lot of packages that build successfully for CPAN
testers don't for ppm.activestate.com (and sometimes, the opposite is true).
Gozer (the new ActiveState CPAN<->PPM guru) is working hard to get
as many of these failing modules working properly ASAP. In the past few
weeks he's implemented the PERL_MM_USE_DEFAULT and AUTOMATED_TESTING
environment variables in cpanrun. He's also working on getting as many
non-perl dependancies as possible (eg; libgd, libpg, etc) installed on his
build boxes so that XS-based packages that depend on these things can be
built. A lot of packages still dont get released through activeperl, but the
situation is getting better.
There's still always going to be packages that fail unit tests, that
people want anyways; I think keeping those packages out of a "quality
assured" repo is a neccessary sacrifice to maintain integrity. Maybe it
would be a good idea for there to be an official "unstable" PPM repo where
packages that built, but failed unit tests, get placed -- then somebody who
wants to be on the bleeding edge can add that repo to their list to get at
the packages, and maybe even lend a hand in figuring out why the tests are
failing.
Cheers,
Tyler
