Guido Günther [2018-01-15 12:14 +0100]: > > > This seems to be caused by the fact that libnss-systemd is not a hard > > > dependency of systemd. I'm not sure what the best solution is? Having a > > > service that is enabled by fails to start looks weird though. Maybe > > > providing a static user isn't that bad? > > > > > > > It requires libnss-systemd, yes. Do you not have it installed? > > It's a recommends, so should be installed by default > > See above: "without installing recommends". My whole point is that the > systemd package installs a service that won't even start without the > recommends which looks somewhat wrong to me.
Note that *in general*, DynamicUser=yes does not *require* libnss-systemd. Services start without it, the only effect is that showing the process with tools like "ps" will not be able to resolve a dynamic user ID to a name - it will just be shown as an ID. This might be a bit confusing, but acceptable for some environments, hence I just made it a Recommends:, not a Depends:. If timesyncd in particular somehow wants to resolve the systemd-timesyncd system user in its own code, then that either should be fixed, or systemd needs to raise libnss-systemd to a Depends: for that particular bug/reason. Martin