A few days ago I decided to join the CPAN-testers. I wonder how long my enthusiasm will last :-) But in the meantime, it's given me far more of an appreciation for just how useful the testers are. I've helped three authors fix bugs now, plus I don't know how many who didn't need any help other than getting the email from me notifying them of the test failure.
Testing stuff on an old version of perl (5.6.2) seems to be smoking out a fair number of problems.
Testing stuff on non-x86 has so far only shown up two problems, one of which I (and others) already knew about (perl doesn't like Unicos) but the other I gleefully submitted to the perl5-porters - perl 5.8.8 builds but fails a test if you tell Configure to just go with all the defaults on NetBSD on Alpha hardware. No doubt when I actually get a testing environment *working* on either of those platforms, I'll smoke out a load of "all the world's a PeeCee" bugs.
Some of the problems I'm finding are really subtle - see if you can spot what's going on here and here - they are ostensibly the same version of a module passing and failing on the same OS and the same build of perl.
Running diff on the two results shows this:
- PERL5LIB =
+ PERL5LIB =/home/david/cpantesting/perl-5.6.2/.cpan/build/Test-Simple-0.67/blib/arch:/home / david/cpantesting/perl-5.6.2/.cpan/build/Test-Simple-0.67/blib/lib
If you test a module successfully, then I believe that the latest version of CPAN adds that to your PERL5LIB for the rest of that session, even if you don't install it.
