On Tue, Jan 08, 2008 at 04:06:18PM -0500, David Golden wrote: > [Breaking this out into a separate subject] > > Perhaps this is how the logic has been done in the past, but I don't > think it's a good idea. Example: > > * Foo-1.23 depends on Bar > * Bar-3.14 tests NA on Win32, Perl 5.8.8 > * Foo-1.23 gets marked NA on Win32, Perl 5.8.8 > * Bar-3.15 fixes Win32 problems and tests PASS on Perl 5.8.8
I can see where you're going with this, but in my experience this situation has not happened. If anything more authors have made the effort to ensure their dists are set to NA and left OSs unsupported. In pretty much every instance where it was implemented the NA was based on the Perl requirement, not the OS. I can only remember one instance where a dist depended upon a Linux only dist, and thus wouldn't work on Win32. > If prerequisites can't be satisfied -- whether due to failure > of tests (FAIL) or due to lack of platform support (NA) -- no report > should be sent. The same hold true for prerequisites that can't be > installed because of missing libraries. All these three are very different, and IMO result in very different outcomes. Missing libraries is not a FAILed prereq, nor is it NA. It would be the same if a module prereq hadn't been uploaded to CPAN in time. With a FAIL or NA the knowledge of the prereq is understood. With a FAIL we have to wait for it to be fixed. With an NA the author is saying that they don't intend to fix, or it may not be possible to fix. Are you aware of any dists that have subsequently enabled support for an OS? I'm willing to be converted, but I see the NA follow through as giving more value to the reporting for users. Cheers, Barbie. -- Birmingham Perl Mongers - http://birmingham.pm.org Memoirs Of A Roadie - http://barbie.missbarbell.co.uk
