Re: Fixing gnome-session

[Michael ROGERS <M Rogers cs ucl ac uk>]

>I've been playing with gnome-session to try an make it handle remote
>clients properly (or at least stop trying to run them locally). I
>tried calling SmsClientHostName() on the connection in new_client, 
>but it always returns the local host. Possibly, this is because the 
>ICE connection is being relayed by the X server, in spite of what 
>the SM protocol doc says.
>
>So, am I missing something? Is this a bug in XFree86? Is the SM
>protocol fatally flawed?
>
>I would think that for a protocol whose goal is to be able to restart
>you desktop exactly as it was, the host the client was running on
>would be important enough to have it's own property!

 I don't know if this is relevant to Gnome, but the client should set 
WM_CLIENT_MACHINE along with WM_COMMAND to allow the session manager 
to restart it on the right machine. However, I guess that it's quite 
rare for session managers to want to restart a client on a remote 
machine, and it could be that the client doesn't bother to set this 
property. You could try reading the raw X property and compare the 
result with what SmsClientHostName gives you. Alternatively, maybe 
the session manager isn't bothering to read WM_CLIENT_MACHINE after 
WM_COMMAND is changed?

 Btw, can anyone point me to documentation on using gnome-session 
without going through GTK? (What are the X events associated with the 
save_yourself and die signals, for example?)


 - Michael Rogers