On Tue, Jul 1, 2014, at 12:53 PM, Michael Catanzaro wrote: > Yes, soname bumps are nonevents with OBS, since everything is > automatically rebuilt. Sounds like Koschei is a big step towards that. I've always found it really strange how so many people talk about "rebuilding" for soname bumps. The *entire point* of a soname bump is to communicate that the API/ABI has *changed*, and thus it may require active human intervention to update callers for the change. In other words, actual source code has to be edited and not just package metadata and a build requeue. If the callers work after a rebuild, it calls into question whether the soname should have been bumped at all (see libffi and udev cases). It's of course good to try rebuilds to see if they work or not; but what'd be better is a system that would allow coordinated updates of groups of packages from a branch, with modified source code. In practice at least for the ABI-unstable libraries in GNOME like libgnome-desktop, all callers are ported during the development cycle, and rebuilds will happen manually via mclazy. The more challenging case is components which are part of multiple ecosystems and not coordinated on a schedule, like libpng. Of course, shielding callers from its ABI changes is major rationale for the existence of gdk-pixbuf. Anyways I'd say "port" for soname bumps instead of "rebuild". -- devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
