On Tue, Mar 05, 2019 at 05:58:30PM +0100, Michael Biebl wrote: > Am 05.03.19 um 15:34 schrieb Mark Hindley: > > > > reassign -1 elogind > > retitle -1 Make libelogind0 ABI compatible with libsystemd0 > > tags -1 + pending > > thanks > > > > On Wed, Feb 27, 2019 at 05:12:50PM +0000, Mark Hindley wrote: > >> We are currently liasing with elogind upstream who are making libelogind > >> ABI > >> compatible with libsystemd. See > >> https://github.com/elogind/elogind/issues/97 > > > > ABI compatibility between libelogind.so and libsystemd.so is complete > > upstream. > > Initial testing with the necessary symlinks now produces full function with > > stock policykit-1. > > Can you eloborate on that a bit? > Why exactly is libelogind0 needed at all in Debian, where we have > libsystemd0?
Because the elogind implementation of sd-login is necessarily different. For example, elogind doesn't support the concept of slices or scopes. > I would very much prefer if we could simply drop libelogind0 and make > sure elogind + libsystemd0 works. Has this route been tried? Yes. It can't work. See https://github.com/elogind/elogind/issues/95#issuecomment-438335747 > It sounds like you want to make libelogind0 Conflicts/Replaces/Provides: > libsystemd0. Yes, versioned provides. > This has the potential for some interesting breakage, like the real > systemd being installed and used together with libelogind0. > We would need to make sure something like that can never happen (ideas > how to solve that?) Obviously that wouldn't work. At the moment elogind conflicts with systemd and depends on libelogind0. So you either end up with systemd + libsystemd0 or elogind + libelogind0. But to be sure we could add libelogind0 conflicts systemd? Mark
