On Tue, Sep 30, 2008 at 7:32 AM, Sisyphus <[EMAIL PROTECTED]> wrote: > How is it that the unlocatability of mpc.h is now resulting in an > "UNKNOWN" - when it used to result in a "FAIL" ? > Is this a new *feature* of the testing that's now being conducted ? Is it > (instead) a new bug ? Or is it just something special that Andreas has > configured ? > > If it's a new feature, then I think that's fantastic !! - and congrats to > whoever (David Golden, you're the main suspect ) implemented it.
It's a change to the CPAN Tester's grade definitions following a lengthy thread on the perl-qa mailing list. > The diagnostic that's being reported suggests that the"UNKNOWN" is given > because the 'make' step failed. Is that the basis upon which Math-MPC-0.50 > gets its "UNKNOWN" reports ? I'm not sure that's an ideal basis ... but > it's bloody close to it - and far closer than anything I could have come up > with, despite (or "because of") the fact that it's so simple :-) It turns out that the "original" intent of UNKNOWN was meant to include build failures according to Graham Barr. And since re-grading doesn't break any of the other parts of the ecosystem, we thought that was more pragmatic than trying to introduce "BUILD_FAILS" and upgrade everything. > The super-sensitive module author will still read that as implying that > "there's something wrong with your distro", even though the real problem > could be that the tester's environment is lacking something (eg the mpc > library). I'll consider it, but it's not top priority right now. My reserve of good-will towards super-sensitive module authors is vastly diminished at the moment. Maybe when more authors send "hey, thanks!" emails like yours to outweigh all the carping elsewhere, then I'll reconsider. David -- David
