All the Perl that's Practical to Extract and Report

Duplicate test results

drhyde on 2007-04-23T12:08:19

One thing that's a bit annoying about my CPAN testing environment is that it doesn't keep track of previous test results. This means that it sends duplicate PASS results for common pre-requisites like this and this. At least the default setting is to ask me before sending FAIL results, so I can manually check what I've sent before and avoid pissing authors off.

But no more! I've patched my local installation of CPAN::Reporter to not send duplicates and sent the diff to the author. If you want it before Dave uploads a new release to the CPAN, let me know.

YACSmoke

barbie on 2007-04-23T19:00:03

Now if you were using CPAN::YACSmoke you wouldn't have this problem, as it automatically records the distributions you're testing. Hence why I'd like to add CPAN support and not have it tied to CPANPLUS as it is at the moment.

Re:YACSmoke

drhyde on 2007-04-24T10:57:21

My understanding is that CPANPLUS uses Build.PL by preference even when a Makefile.PL is present. CPAN.pm, of course, prefers to use Makefile.PL. If everyone were to use CPANPLUS, then we wouldn't catch errors like this where the author lists all her dependencies in the Build.PL but her Makefile.PL is horribly out of date.

Re:YACSmoke

barbie on 2007-04-24T13:19:10

Not true. You can choose you use Makefile.PL as a preference in CPANPLUS. I'm glad it does as I am not a fan of Module::Build, mainly due to the fact it can't install itself, thus causing numerous crashes and fails for distributions when I was CPAN Testing on Windows.

One idea I've had more recently is to incorporate some standard testing in YACSmoke, so that rather than the distribution including tests for testing POD, POD coverage, Meta.yml or the whole distribution and possibly even the complete test suite coverage or the Perl critic, YACSmoke can run those tests as a separate part of the CPAN Testing. A part of that could also be to ensure the Makefile.PL/Build.PL files are accurate and sane (for some measure of sane) if they both exist. This would then mean that the author doesn't have to worry about including tests that often don't mean a lot to the actual user, but that they get tested on a multitude of platforms that the author may otherwise never have access to. It could then better interface to the CPANTS metrics for those areas, as currently CPANTS only makes a cursory judgement on some of them. That's a long way off, and perhaps ties in with Adam's PITA idea, but it's on my TODO list as a possible new feature.

Fixed in 0.42

dagolden on 2007-04-24T04:42:25

0.42 is uploading to CPAN now with the new "send_duplicates" option. Duplicate reports will be skipped by default, but this behavior may be customized similar to most other CPAN::Reporter configuration options.

should check only for full dupes

chorny on 2007-04-24T07:09:44

IMHO, such code should also check for prerequisites versions - it may fail with some combinations. And in some distant future - versions of C libs.