From: "Darren Duncan" <[EMAIL PROTECTED]> > I am seeing periodically that some/all CPAN testers, which I am not > sure whether they are automated processes or people, have a large > flaw in the way they are testing new uploads. > > And that is, despite the fact that distributions often explicitely > cite the module names and versions they depend on, in their > machine-readable Makefile.PL or META.yml files, the testers are not > looking at these and ensuring that they first have said dependencies > before testing the new distributions.
There seems to be a problem with CPANPLUS (or an associated module) that occasionally doesn't acknowledge the prerequisite lists. I've noticed this happening with my testing in recent weeks, at which point I try manually testing to see if the module installs, which it generally does. On occasion I've also tried the 'test' command from the cpanp shell and that works too. It appears to be something specific to the way cpansmoke hooks into CPANPLUS. On windows this is fraught with problems anyway, as the return string from IPC::Run, gets dropped 99% of the time in a FAIL or UNKNOWN report, which is why my reports are largely cut'n'pasted by hand. I have tried to look into this myself, but have not come close to figuring out why. Unfortunately upgrading to CPANPLUS-0.49 for me at least has broken the ability for cpansmoke to use CPANPLUS, so I can't test it any further at the moment. > Finally, the problem I am noticing seems to be compounded with the > fact that as soon as a distribution fails under one tester, it is > much less likely to be tested by someone else. You might want to check whether that's because the module requires 5.8.x, as several of the testers are still at 5.6.1. Personally I do try and test everything I possibly can. Sorry can't be of more help at the moment, but hopefully someone will have more understanding of the problem than I and be able to fix it. Barbie. -- barbie (@missbarbell.co.uk) | Birmingham Perl Mongers | http://birmingham.pm.org
