Red Hat Bugzilla – Bug 229113
compiz session management conflicts with desktop-effects
Last modified: 2007-11-30 17:11:57 EST
Description of problem:
When I switch between compiz <> metacity more than 2 times its not possible to
start metacity without a X restart because compiz keeps getting started from
gnome-session (replaces metacity).
The patch from gandalfn fixes this for me.
Is it possible to include this in the fedora rpms?
(works fine for me)
Created attachment 148266 [details]
patch from gandalfn
Does this work correctly ? We do want compiz to be restarted immediately if it
So this probably broke from fixing bug 200280.
The fix doesn't look right though. The desktop-effects capplet should invoke
metacity and compiz with the --replace arguments. Compiz should notice that
it's getting replaced when the capplet does "metacity --replace" and then
unregister itself with the session-manager before exiting.
yeah, thats the right fix.
how is this done in metacity? (where in the code / which file)
The window manager that's currently running for a screen takes ownership over a
well known named selection (WM_Sn where n is a screen number).
When it gets a SelectionClear event for that selection it's responsible for exiting.
If you look in display.c in the event_callback function it handles the
SelectionClear event by calling process_selection_clear which calls
meta_display_unmanage_screen. After all the screens are no longer managed by
metacity (which would happen when compiz --replace is called) then metacity
calls meta_quit () (from the meta_display_close function). meta_quit makes the
event loop that's run in main() return, so then meta_session_shutdown () gets
called. meta_session_shutdown tells the sesion manager not to restart metacity
after it exits.
I added this patch to the RPM:
I believe it's the right fix, but please give it a spin.
(In reply to comment #7)
> I added this patch to the RPM:
> I believe it's the right fix, but please give it a spin.
Is it possible to update the compiz to 0.5.0 ?
(there is no link on the hp yet but its released) , should include the session
fix + others.
We are already in feature freeze, and 0.5.0 does not look like a bugfix-only
release. So we are probably going to wait until after F7 and put it out as an
update with sufficient time in updates-testing.
(In reply to comment #9)
> We are already in feature freeze, and 0.5.0 does not look like a bugfix-only
> release. So we are probably going to wait until after F7 and put it out as an
> update with sufficient time in updates-testing.
the session fix in head does not work (same bug ) ....
Hey, this one is fixed, right?
yes confirmed it works fine with the current version in rawhide.