On Sun, Sep 28, 2008 at 3:43 AM, Dmitry Karasik <[EMAIL PROTECTED]> wrote:
> That's an overkill, I think. I also received lots of PASS, which means that
> automated testing works more ofthen than not. I only started receiving fails
> generated by YACSmoke, and by YACSmoke only. I suppose there are other testing
> frameworks, that do their jobs fine, and of course I'm interested in their 
> test
> output.

There's really two issues here.    Testers can use either CPANPLUS or
CPAN+CPAN::Reporter.

Some CPAN::Reporter testers have written a framework to actually
install modules and then throw away the Perl lib directories and untar
a clean copy.  Those are probably the tests that you're getting PASS
reports for.

CPAN::Reporter::Smoker doesn't do that -- it lets CPAN use the same
PERL5LIB trick with blib, but CPAN::Reporter will discard any FAIL or
UNKNOWN report that is missing prerequisites, so you aren't seeing any
of those.

What I don't understand is why CPANPLUS isn't doing the discard with
the Prima dependency not found.

> Yes, me too. However I'm fairly sure that rewriting Prima Makefile.PL in order
> to use blib is a large task. Also, if the sole reason to do that is to be
> compatible with YACSmoke, and no real productivity gain, I don't know, I guess
> it would be much more reasonable to fix YACSmoke to do the proper thing
> instead. After all, are modules for testers, or testers for modules?

The challenge is defining what "do the proper thing" means in a
sufficiently general way.  Testers are for modules, but since every
module is free to do it's own thing, modules need to follow some
standard conventions.  What I like to see is that for the modules that
don't follow conventions, testers at least don't spew useless reports.

Let me think about what assumptions CPAN/CPAN plus could make if a
blib directory isn't found, which is at least a signal that the
convention isn't being followed.

-- David

Reply via email to