[PATCH v2 10/14] systemd-logind: Add systemd-logind "core"

[David Herrmann]

Hi

On Tue, Feb 11, 2014 at 9:10 AM, Hans de Goede <hdegoede at redhat.com> wrote:
> Hi,
>
> On 02/11/2014 04:16 AM, Jasper St. Pierre wrote:
>> As said before, with systemd user sessions, this will not happen. Unless
>> I'm misremembering, Lennart has said that the only thing that should be
>> inside the PAM session environment proper should be the session leader like
>> "gnome-session". (Disregarding the session worker process like
>> gdm-session-worker)
>
> That sounds very wrong, many things will break if they are not run in a
> proper login session.
>
> Anyways lets bring Lennart himself into the loop.
>
> Lennart, my current version of the xserver logind integration patches use
> GetSessionByPID(getpid()), Jasper is afraid this will break in the future,
> because the xserver will be running outside of the user-session, see below.
> Is he right?
>
> To be clear, the goal of the GetSessionByPID() call is to determine the
> session on which to issues the TakeControl() call.
>
> Also how does this relate to the fake-sessions spawning multiple heads we
> were talking about before?

It's fine to use GetSessionByPID() for now. Lets get this working with
Xorg and deal with any user-bus transitions later. We use this call
all over the place so no need to make a big deal out of it here.
If we want a transition to a user-bus, the idea is to run a compositor
per user, not per session. Thus, it's no longer running *in* a
session. If Xorg breaks with this, well then we don't do that with
Xorg..

Thanks
David