On Thu, Feb 09, 2006 at 04:21:01PM -0800, Tyler MacDonald wrote: > > However, this reveals a more sinister problem with the Makefile.PL > in DBI: > > use Test::Harness; # tests use it so check it's loadable early > use Test::More 0.4; # tests use it so check it's loadable early
I'd hardly call that a 'sinister problem'. > Those should probably just be handled by PREREQ_PM; CPAN will then > attempt to download the latest version if it's not installed; the > Makefile.PL will advise the user that these dependancies haven't been met; > if the user plows ahead and installs anyways, the tests may fail but at > least the code will still work. > > IMHO the only timea package should be "use"d in a Makefile.PL is if > the Makefile.PL itself needs it to make decision, and then it should fail > gracefully. I agree. (Though I believe there were good reasons at the time the original lines were added - but I don't recall them now.) I've removed them. Tim.
