<halfline> kay: hey around?
<halfline> kay: any idea why there is sd_uid_get_state and not
sd_session_get_state for each individual session ?
<kay> halfline: guess we just don't really care about multiple sessions
belonging to the same uid. you would need that for?
<halfline> kay: well, some sessions "linger" after logout
<halfline> kay: and i don't want gdm to ever jump to those
<halfline> right now gdm will jump to existing running sessions if they exist
instead of creating a new one
<halfline> after the user types their username and password
<kay> halfline: oh, i see
<kay> halfline: vc logins would probably confuse that too
<halfline> well we explicitly check the type
<halfline> and ignore vc logins
<kay> ah. so we need a 'linger' type?
<halfline> well there are two user states that need to map to session states it
hink
<halfline> lingering and closing
<halfline> lingering is when someone wants to keep screen running but otherwise
log out
<halfline> and closing is when pulseaudio stays alive for 10 seconds after
logout
<halfline> both states are currently "user states"
<halfline> but that's kind of wrong, since they're about specific sessions
<kay> yeah

in systemd git there's now a new state property on all seats which is one of "online", "active", "closing".

online = logged in
active = logged in and in the fg
closing = nominally logged out but still some left-over around from this session

This is exposed both on the Session bus object as well as in sd_session_get_state().

This should provide everything gdm needs, right?

yup sounds perfect.

Created attachment 218250 [details] [review]
patch

The patch contains an unrelated bugfix, btw: Don't use g_free on a string returned by sd_session_get_type

Review of attachment 218250 [details] [review]:

Looks fine, thanks !

::: daemon/gdm-slave.c
@@ +1436,1 @@
 

would be good if this was a separate commit, but good either way.

Pushed as two commits