[Israel.pm] CPAN dependencies are evil

Shmuel Fomberg shmuelfomberg at gmail.com
Sat Nov 3 06:21:07 PDT 2012


Hi Gabor.

> Is this what we get from the famous CPAN and the testers? modules that can
> > not be installed, silenced tests (I wrote about it a month ago) and a
> lot of
> > headache.
>
> The most basic way to help is to configure your CPAN client so
> it will report every success and failure to the CPAN Testers
> and to the authors of the modules.
>

It is not that these failure are not reported - you can see them in CPAN
test results. They are just ignored.
Now I know that is quite difficult to debug using CPAN. you release a
version, see the failures, (which may take a week to arrive) try to
understand what differentiate the passing systems from the failing systems,
(is that the OS? some compilation option?) fix the bug, release another
version, wait another week... been there done that.

But as modules use more and more dependecies, these small 2% failures add
up.

You want to fix it? tell me, if the auto CPAN tester is trying to test a
release, and it fails to install a dependency, what does it do?
maybe it should mark a failure in installing a dependency as failure in
that module too. or maybe in all the dependencies. So people will see the
real picture on the installability of their module, and will avoid using
modules that are buggy.

Oh, and here is a project for someone with free time: automatically
classify the test failures. this is a classic categorization problem.

Shmuel.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.perl.org.il/pipermail/perl/attachments/20121103/67e999e6/attachment.htm 


More information about the Perl mailing list