load problems with cvs HEAD of gconf.

[J Shane Culpepper <pepper eazel com>]

Hi,

I've been working on nautilus and I noticed that when we went to cvs
HEAD of gconf there seem to be huge performance problems related to
gconf.  The size of gconfd grows very large and slows the machine down
too much.  /var/log/messages is spewing tons of problems.  Here is a
small piece of the log that looks a bit strange:

Oct  5 15:37:25 juggernaut gconfd (pepper-20730): starting (version
0.9), pid 20730 user 'pepper'
Oct  5 15:37:25 juggernaut gconfd (pepper-20730): No configuration files
found, trying to use the default config source
`xml:readwrite:/home/pepper/.gconf'
Oct  5 15:37:25 juggernaut gconfd (pepper-20730): Initializing XML
backend module
Oct  5 15:37:25 juggernaut gconfd (pepper-20730): Removing stale lock
`/home/pepper/.gconf/%gconf-xml-backend.lock' because of error pinging
server: IDL:CORBA/COMM_FAILURE:1.0
Oct  5 15:37:25 juggernaut gconfd (pepper-20730): Directory/file
permissions for XML source at root /home/pepper/.gconf are: 700/600
Oct  5 15:37:27 juggernaut gconfd (pepper-20730): Saved state file had a
removal of a listener that wasn't added; ignoring the removal.
Oct  5 15:37:33 juggernaut last message repeated 60277 times
Oct  5 15:37:35 juggernaut gconfd (pepper-20730): Failed to notify
listener 134222862, removing: IDL:CORBA/COMM_FAILURE:1.0


Then there are tons of messages like the last line.  I would file a bug
report but I have not collected enough useful information yet to figure
out what is happening.  Its quite possible that the performace drain is
caused by all the writes to the message log.

-Shane