Two tickets on CPAN::Reporter have prompted the question of the role
of CPAN Testers in capturing build failures versus test failures --
i.e. failures in Makefile.PL/Build.PL or in make/Build.

* http://rt.cpan.org/Public/Bug/Display.html?id=21691
* http://rt.cpan.org/Public/Bug/Display.html?id=21566

Right now, CPAN::Reporter only wraps the "test" action in CPAN.pm and
CPAN.pm doesn't even run tests if the make process wasn't successful.
It wouldn't be terribly hard to add support to capture the earlier
stages, but I'm not sure if CPAN Testers is the right place to capture
it and, if so, how it should be coded.

If failures happen before tests are run, should that be counted as an
"NA" -- meaning that tests could not be run on that platform?
Generally, I've seen that to mean that prerequisites could not be
satisfied.  Or should there be a new failure category "BUILDFAIL" (or
"WONTBUILD") that distinguishes these cases?  Or something else
entirely

Before I hack something into CPAN::Reporter and CPAN.pm, I'd like to
get some community consensus, particularly from those involved in CPAN
Testers.

Regards,
David Golden

P.S. As a side note, Chorny (from win32.perl.org and #win32) mentioned
that he's looking into HTTP transport for Test::Reporter -- that will
be a nice upgrade for getting reports from behind ISP firewalls.

Reply via email to