On Mon, Feb 28, 2011 at 04:07:13PM +0000, Colin Guthrie wrote: > 'Twas brillig, and Rafa? Mu?y?o at 28/02/11 15:16 did gyre and gimble: > > I start i.e. audacious on first display, switch to the second, > > start another audacious there. Both are playing. I stop audacious on the > > second and logout there. That stops audacious on first too, as while > > pulseaudio gets restarted, audacious still holds connection to the > > server that died. What's more (but that's (argueably) probably a bug in > > audacious), I need to kill audacious, as I can't just close it then. > > Well that's the thing, PA should not die when the session manager dies. > It's meant to behave nicely. > > I susepct it's the session manager that is killing the PA process. > > If you unload the various X11 modules (only the ones for the session > you're logging out of tho' - the same modules will be loaded twice, once > for each session) gracefully via pacmd/pactl, then log out, does PA > still die? > > If not, then the problem is in the session manager. As I said in the ticket, the problem arises only with x11-cork-request and x11-publish loaded - actually, only one of those is necessary to trigger the bug, though I don't quite remember which one. start-pulseaudio-x11 loads both anyway. However, firefox is no longer taken down when on pulseaudio termination, though that may simply be a result of the change in plugin handling in firefox. Same as with rox-session, x11-xsmp doesn't get loaded.
