On Thu, Aug 8, 2013 at 2:26 PM, Tom Wijsman <tom...@gentoo.org> wrote: >> On Thu, Aug 8, 2013 at 8:41 PM, Rich Freeman <ri...@gentoo.org> wrote: >> > Stability is about the quality of the ebuilds and the user >> > experience in general. It is not a statement that all Gentoo >> > developers think that the package is useful. Many would say that >> > nobody should be using MySQL/MariaDB for production work, but that >> > has nothing to do with its stability as a package either. >> >> This is not entirely correct. >> >> If from now on, a bug with systemd of new version of a package blocks >> that package stabilization, it means that all developers must support >> systemd. > > Not necessarily, one can opt to mask this combination and stabilize > this combination later by removing the mask; it's an implementation > detail, but certainly there's no need to imply that they must.
Package updates that break other packages is not an issue unique to the stable tree - we just have less tolerance for it there. If libfoo-5 breaks stable systemd, then there needs to be coordination, just as is the case if libfoo-5 breaks stable xeyes or openoffice. Usually in these situations things get straightened out and we're usually the better for it because such incompatibilities tend to be the result of brain-dead behavior in one upstream or the other. If it can't be straightened out then sometimes we accept blocking deps/etc. I'm sure in the history of every glibc upgrade in Gentoo there has been some stable package or another that couldn't keep up, and I even recall one that basically required rebuilding everything in ages past. And this is cooperation - everybody benefits from it. > > Another example is that when you add a package to the tree, you are not > required to initially commit both an OpenRC unit and systemd service > file; you are suggested to provide them for the convenience of the > user, > Indeed, when you add a package to the tree you're not required to initially commit any kind of service files for it at all (openrc or systemd). Supporting both is to be encouraged all the same, and accepting patches to add service files should not be discretionary (if another maintainer/proxy is willing to do the work). Rich
