On Mon, May 12, 2008 at 05:29:23PM +0200, gregor herrmann wrote: > On Mon, 12 May 2008 15:39:19 +0300, Niko Tyni wrote: > > > > Hi, with a huge number of reverse build-dependencies, I think this needs > > > a better > > > plan. > > > As for building, perl-modules is currently always installed because > > dpkg-dev is build-essential and pulls it in. It's not build-essential > > itself, though. > > > > Conclusion: everything should Just Work (tm), but there's an argument > > for requiring build dependencies on 'perl-modules (>= 5.10.0)' for > > these. Possibly even 'perl-modules (>= 5.10.0) | libversion-perl' to > > ease backporting, but that's mostly cosmetics as at least sbuild won't > > look at the alternative anyway. > > Hm, I guess as long as perl-modules is not build-essential we have to > use 'perl-modules (>= 5.10.0) | libversion-perl' for policy > compliance (and being nice to backporters).
Which policy section says Build-Depends can't contain virtual packages? FWIW, sbuild seems to handle a build-dependency on a virtual package just fine. If there are several alternatives providing the same virtual package, policy 7.4 suggests listing the preferred alternative first, but that's not the case here. Thomas, what's the tool you're using for checking dependencies? Does it take virtual packages into account? Are you looking at sid only or both sid and lenny? (The real libversion-perl is needed in testing until perl 5.10 gets in.) -- Niko Tyni [EMAIL PROTECTED] -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]