Description of problem: bonobo-activation-server fails to exit at the end of the user's session. Version-Release number of selected component (if applicable): libbonobo-2.8.0-2 How reproducible: Always Steps to Reproduce: 1. Log in to GNOME. 2. Log out. 3. Switch to a command-line with Ctrl-Alt-F[1-6] and log in as the same user you logged in to GNOME as, and run 'ps x' Actual results: bonobo-activation-server is still alive and well Expected results: bonobo-activation-server should be dead and gone
I've done some testing with a rawhide nightly install 20041014: if I login, do nothing and logout, b-a-s goes away correctly (as does e-d-s) If I click on the panel clock (which should use evolution-data-server to get your tasks/appointments), then logout, b-a-s and e-d-s also go away correctly. However, if I login, start Evolution, and close Evolution, then logout, THEN b-a-s and e-d-s survive the session. Do you see this behaviour? Do you have to run Evolution to get b-a-s and e-d-s to achieve immortality?
I've confirmed this on another machine; the b-a-s problem seems to be related to e-d-s, when e-d-s is launched from evolution.
The e-d-s bug is bug #134851
The fix for the e-d-s bug fixes this
This appears to also happen with RHEL3. I can reproduce it easily just by starting evolution then logging out. Would you prefer I open a new bug since it's for RHEL3?
Dave Lehman: yes, please do it for RHEL3 as a separate bug, since so much of the code is different (and it can then appear in different trackers, etc)
OK, resolving this one as CURRENTRELEASE. The not-yet-fixed RHEL3 bug (bug #134144) looks similar, but I believe the exact underlying cause is different.
Sorry to ask but which version of libbonobo was this fixed in? I am still seeing immortal /usr/libexec/bonobo-activation-server with libbonobo-2.4.0-1 (FC3 final)...
From your comments on bug #139834 it sounds like your've discovered another way to keep evolution-data-server from exiting, and it's that that keepds bonobo-activation-server alive. I believe the real problem you're seeing is with e-d-s, not with b-a-s (although it's arguable that managing object lifetimes based on distributed reference counting is an inherently broken design).