Hi, Martin.

There was a recent "policy change" on grades -- FAIL is now *only* for
actual test failures.  Build failures are graded 'UNKNOWN'.  Some
authors had complained that it's hard to differentiate between actual
test failures and build failures that might simply be due to a poorly
configured Perl or CPAN toolchain.

This example is a great one -- because of the EU::MM problem, your
module doesn't compile.  By grading that UNKNOWN, you can choose to
pay attention or ignore it and focus your energies on FAIL.

This was hashed out in some massive threads on perl-qa a couple weeks
ago.  As it turns out, Graham Barr weighed in to say that this new
approach was the original intent behind the grades.

You'll still see some build failures as FAIL until everyone upgrades
CPAN::Reporter and CPANPLUS.  The latter is fixed only in a dev
version 0.85_04 and not everyone will upgrade until that goes to a
full release.

-- David

On Thu, Sep 25, 2008 at 5:43 AM, Martin J. Evans
<[EMAIL PROTECTED]> wrote:
> Hi,
>
> Here
> http://www.nntp.perl.org/group/perl.cpan.testers/2008/09/msg2285905.html
> you will find a submitted test result of UNKNOWN which totally failed
> compilation. Is that right? Would you not normally class this as a FAIL?
>
> BTW, the reason it fails compilation is that it is strawberry perl, which
> sets makepl_arg to include INC and ExtUtils::MakeMaker cannot override (or
> add to) and INC added to the command line (so Makefile.PL cannot add the
> path to the DBI header files).
>
> Martin
>

Reply via email to