Merge lp:~muktupavels/lightdm/improve-multiseat-support into lp:lightdm

[Alberts Muktupāvels]

Improve multiseat support

With current lightdm version I have two following problems in multiseat setup:
1) First seat user logs in, all is ok. When second seat user logs in, first seat vt gets switched. User needs use Ctrl+Alt+F7 to get back to desktop.
2) Second seat user can not access audio devices. In sound settings only 'dummy output' is listed.

This commit will fix both problems.

[Robert Ancell (robert-ancell)]

This fails when you run 'make check'. Specifically the xdg-seat test (cd tests ; ./test-xdg-seat). The fix is quite trivial but you should make it to show that you know the tests and think if this is the correct behaviour.

The removal of the VT check in seat_xlocal_get_active_session () doesn't really make sense. In a multi-seat case you should probably check seat_get_string_property (seat, "xdg-seat") instead and do something different than checking if any session is on this VT.

You should replace:
(xdg_seat && strcmp (xdg_seat, "seat0") == 0)
with
g_strcmp0 (xdg_seat, "seat0") == 0
as this is more readable.

review: Needs Fixing

[Alberts Muktupāvels (muktupavels)]

> This fails when you run 'make check'. Specifically the xdg-seat test (cd tests
> ; ./test-xdg-seat). The fix is quite trivial but you should make it to show
> that you know the tests and think if this is the correct behaviour.

Is it good now? Now all tests completes successfully. From xdg-seat.conf removed VT check. Only main seat - seat0 should know about VT (from docomentation - http://www.freedesktop.org/wiki/Software/systemd/writing-display-managers/). So if seat is not seat0 than vt should not be set.

> The removal of the VT check in seat_xlocal_get_active_session () doesn't
> really make sense. In a multi-seat case you should probably check
> seat_get_string_property (seat, "xdg-seat") instead and do something different
> than checking if any session is on this VT.

Added back VT check.

> You should replace:
> (xdg_seat && strcmp (xdg_seat, "seat0") == 0)
> with
> g_strcmp0 (xdg_seat, "seat0") == 0
> as this is more readable.

Replaced.

[Robert Ancell (robert-ancell)]

Looks good, thanks!

review: Approve

[Alberts Muktupāvels (muktupavels)]

> Looks good, thanks!

Is there any reason why you did not merge this branch?

[Robert Ancell (robert-ancell)]

I was waiting for the CI system to confirm it, but it didn't happen as I thought.

[PS Jenkins bot (ps-jenkins)]

review: Approve (continuous-integration)

=== modified file 'src/seat-xlocal.c'
--- src/seat-xlocal.c	2013-12-11 03:28:54 +0000
+++ src/seat-xlocal.c	2014-02-05 15:30:28 +0000
@@ -61,6 +61,7 @@
 get_vt (Seat *seat, DisplayServer *display_server)
 {
     gint vt = -1;
+    const gchar *xdg_seat = seat_get_string_property (seat, "xdg-seat");
 
     /* If Plymouth is running, stop it */
     if (plymouth_get_is_active () && plymouth_has_active_vt ())
@@ -78,7 +79,9 @@
     }
     if (plymouth_get_is_active ())
         plymouth_quit (FALSE);
-    if (vt < 0)
+    if (!xdg_seat)
+        xdg_seat = "seat0";
+    if (vt < 0 && g_strcmp0 (xdg_seat, "seat0") == 0)
         vt = vt_get_unused ();
 
     return vt;

=== modified file 'tests/scripts/xdg-seat.conf'
--- tests/scripts/xdg-seat.conf	2013-10-03 21:28:05 +0000
+++ tests/scripts/xdg-seat.conf	2014-02-05 15:30:28 +0000
@@ -10,7 +10,7 @@
 #?RUNNER DAEMON-START
 
 # X server starts
-#?XSERVER-0 START VT=7 SEAT=seat1
+#?XSERVER-0 START SEAT=seat1
 
 # Daemon connects when X server is ready
 #?*XSERVER-0 INDICATE-READY
@@ -18,7 +18,7 @@
 #?XSERVER-0 ACCEPT-CONNECT
 
 # Session starts
-#?SESSION-X-0 START XDG_SEAT=seat1 XDG_VTNR=7 DESKTOP_SESSION=default USER=have-password1
+#?SESSION-X-0 START XDG_SEAT=seat1 DESKTOP_SESSION=default USER=have-password1
 #?XSERVER-0 ACCEPT-CONNECT
 #?SESSION-X-0 CONNECT-XSERVER