On Thu, 07 Jul 2016 at 15:56:15 +0200, Yves-Alexis Perez wrote: > When running the second time, I get: > > corsac 28763 0.0 0.0 43696 388 ? S 15:52 0:00 > /usr/bin/dbus- launch --exit-with-session --sh-syntax > > > so no dbus-daemon indeed
Huh. If dbus-launch had killed the dbus-daemon for whatever reason, like for instance losing control of its X11 window, it should have also exited. So maybe it's not that... > + export DBUS_SESSION_BUS_ADDRESS > + DBUS_SESSION_BUS_PID=28764 > + DBUS_SESSION_BUS_WINDOWID=4194305 > > so dbus-launch was started correctly and fork()ed dbus-daemon as 28764 but > then something happened Here is how X11 autolaunch works: https://bugs.freedesktop.org/show_bug.cgi?id=35887#c6 So if you use "xprop -id 4194305" (in this case) to inspect that window, you should hopefully find a consistent address and pid attached to it. > I'll try to investigate why dbus-daemon actually dies the second time (and why > only with lightdm 1.18.2) but any clue welcome. This might help: cat > /etc/dbus-1/session-local.conf <<EOF <busconfig> <syslog/> </busconfig> EOF (Actually, I'm quite tempted to enable that for all Debian users via a drop-in in /etc/dbus-1/session.d - dbus-user-session already logs to syslog, via the systemd Journal, so it would be consistent.) Or you could try running dbus-launch with DBUS_DEBUG_OUTPUT=1 in the environment, which makes it not close its dbus-daemon's stderr. If all else fails you could try installing dbus-1-dbg and running /usr/lib/x86_64-linux-gnu/dbus-1.0/debug-build/bin/dbus-launch (or whatever is right for your architecture) with DBUS_VERBOSE=1 in addition to DBUS_DEBUG_OUTPUT=1, which will result in spammier debug output than you could possibly want. I hope one of those helps to shed some light on this... S
