systemctl reload-or-try-restart does not fail silently as per documentation

[coling]

The man page states:

   reload-or-try-restart PATTERN...
       Reload one or more units if they support it. If not, restart them instead.
       This does nothing if the units are not running. Note that, for compatibility
       with SysV init scripts, force-reload is equivalent to this command.

However:

[root@jimmy ~]# systemctl is-active httpd
inactive
[root@jimmy ~]# systemctl reload-or-try-restart httpd
Job for httpd.service invalid.
[root@jimmy ~]# echo $?
1

cf.

[root@jimmy ~]# systemctl try-restart httpd
[root@jimmy ~]# echo $?
0

So it's not behaving as described. This is with systemd 222

[poettering]

humm, that must be recent breakage. What's your systemd version?

[coling]

Sorry, was editing comments to include that! v222 here. But it's been like that for ages (at least 217). If it's not happening for you then it might be a downstream issue (in which case sorry!) but I don't think any of our patches would touch that area of the code.

[coling]

(for reference downstream Mageia bug

[dvdhrm]

This is very easy to reproduce. Just write a random service file with ExecReload=/bin/true. If the unit is NOT running, then systemctl reload-or-try-restart foobar.service fails with the given error code.

The underlying issue is bus_unit_queue_job() drops the try- information if the unit supports reloading. That is, if the unit supports reloading, the reload-or-try-restart (which is represented as type == JOB_TRY_RESTART && reload_if_possible) is folded into JOB_RELOAD. This clearly lacks the information that the call should not fail if the unit is not running.

I'm not sure how to fix this properly, and I cannot see how this ever worked. No patch from me, as I have really no clue how this is properly fixed. @poettering any ideas?