On Fri, 11 Jun 2010 07:39:01 -0400 Joseph Jezak <jos...@gentoo.org> wrote:
> Your preferred method is exactly how (as a ppc keyworder) I like to > see these kind of bugs handled. Dropping keywords makes an awful lot > more work for us and hurts our users, especially since we're not > always very prompt at handling bugs. Well, reasoning for the HPPA team, which maintains an architecture that is dying rather more quickly than PPC32 (which still has all kinds of embedded uses and so on),, in favour of IA64, I'd rather see dropped keywords than new profile entries, possibly with the keyworded ebuilds being gradually removed after an OK. That way I can make a choice to keep a package (set) for a bit or to stop supporting it immediately. Since there is no "unveiling" effect in re-adding dropped keywords, as opposed to using profile masks that you can only remove safely by first revdep-checking the entire tree again, I'd rather have people file bug reports than touching the HPPA profile files themselves. Since we (HPPA) basically agreed to drop support for the major desktop environments in due time already (we still need to make that official some time soon and then actually work on the problem for the last time), dropping those keywords is a lot better than masking specific versions of ebuilds or specific uses of USE flags. Funnily enough, I've expressed these wishes to the people who are doing the *DEPEND checks before they commit (hundreds of ebuilds) time and again, and still ended up with sometimes years old entries in package.{,use.}mask files. In fact I think there's a bug open about it and I tried to get some discussion about it going on this very mailing list. :) jer