Nathan Rennie-Waldock [2015-02-23 16:07 -0000]:
> This is actually in systemd, not systemd-shim.
No, it's still in -shim.
> I patched systemd-shim to ignore unknown unit warning, so then I get
> a warning from systemd-logind: systemd-logind[1414]: Failed to start
> user service: Message did not receive a reply (timeout by message
> bus)
Yeah, ignoring the warning is simple, but it's opening a can of worms
indeed as then logind runs into more unimplemented API from -shim. I
wrote an initial (incomplete) patch months ago:
> I'm guessing user@.service is generated at runtime by systemd (which
> would require systemd is init, not upstart).
By logind, yes.
> Suggestions for fixing this:
> 1. Move /etc/pam.d/systemd-user to the systemd-sysv package (assuming none of systemd-logind's features are required)
That would be wrong (as you can boot with systemd without having the
systemd-sysv package), as well as unnecessary (if user@.service isn't
started, this pam file should be a no-op).
> 2. Patch systemd-logind to not start the user service if the system
> wasn't booted with systemd
This is ugly indeed, but would be an acceptable SRU for utopic. For
vivid (or possibly in vivid+1 now) we want to switch to systemd for
booting, so it will become unnecessary.
Hey Nathan,
Nathan Rennie-Waldock [2015-02-23 16:07 -0000]:
> This is actually in systemd, not systemd-shim.
No, it's still in -shim.
> I patched systemd-shim to ignore unknown unit warning, so then I get logind[ 1414]: Failed to start
> a warning from systemd-logind: systemd-
> user service: Message did not receive a reply (timeout by message
> bus)
Yeah, ignoring the warning is simple, but it's opening a can of worms
indeed as then logind runs into more unimplemented API from -shim. I
wrote an initial (incomplete) patch months ago:
http:// people. canonical. com/~pitti/ tmp/0001- Quiesce- warning- about-missing- user-NNNN. service. patch
but it's not sufficient.
> I'm guessing user@.service is generated at runtime by systemd (which
> would require systemd is init, not upstart).
By logind, yes.
> Suggestions for fixing this: d/systemd- user to the systemd-sysv package (assuming none of systemd-logind's features are required)
> 1. Move /etc/pam.
That would be wrong (as you can boot with systemd without having the
systemd-sysv package), as well as unnecessary (if user@.service isn't
started, this pam file should be a no-op).
> 2. Patch systemd-logind to not start the user service if the system
> wasn't booted with systemd
This is ugly indeed, but would be an acceptable SRU for utopic. For
vivid (or possibly in vivid+1 now) we want to switch to systemd for
booting, so it will become unnecessary.
Thanks!
Martin