[Merge] lp:~serge-hallyn/ubuntu/utopic/systemd-shim/shim-abandoncg into lp:ubuntu/systemd-shim

[Serge Hallyn]

Quoting James Hunt (james.hunt at canonical.com):
> This fix is also required for bug 1357252.
> 
> The code changes look fine to me (although I'm not very familiar with the codebase). Just a few queries:
> 
> - src/systemd-shim.c : There's a chunk of code that's #if-0'd out. Can this be removed?

It can, but I'd rather not.  It's a reminder that we have got to either
find a way to make it short-lived again, or address any issues which
were ignored previously due to systemd-shim being short-lived.  This
includes closing or at least re-opening on peer-closure of the cgmanager
socket and running systemd-shim under valgrind to check for memleaks.

> - I was imagining when I heard that systemd-shim was now long-running that it would exit when the session exits but I now see that it starts on-demand and then stays running for the duration of the boot to allow it to service requests from multiple sessions. Now that it is maintaining state, what's the implication of it stopping on error and being restarted?

StopUnit will continue to work, Abandon will be ignored.  So on a default
Ubuntu desktop, cgroups for any active sessions will not be deleted.

> - Does the shim need an apparmor policy now it's long-running?

D'oh, that wouldn't hurt.  shim really shouldn't need much access at all.

-- 
https://code.launchpad.net/~serge-hallyn/ubuntu/utopic/systemd-shim/shim-abandoncg/+merge/233280
Your team Ubuntu branches is requested to review the proposed merge of lp:~serge-hallyn/ubuntu/utopic/systemd-shim/shim-abandoncg into lp:ubuntu/systemd-shim.