Session daemon quitting

[Havoc Pennington]

On Sat, 2004-04-10 at 15:59, Thomas Leonard wrote:
> BTW, I noticed that the default policy has:
> 
>     <!-- Allow everything to be received -->
> 	<allow eavesdrop="true"/>
> 
> This seems to block errors. I had to change it to:
> 
>     <!-- Allow everything to be received -->
> 	<allow eavesdrop="true" requested_reply="false"/>
> 
> Otherwise, you don't even get to see the error telling you that the
> security policy stopped you from seeing the original error, which isn't
> too friendly... (tested with dbus-send).

I think this may be just a bug - an error reply should be considered a
reply and allowed. You'd have to investigate in more detail.

> Finally, calling _exit in a client automatically when the dbus daemon dies
> seems a bit unfriendly. Although we never want the daemon to exit, users
> should really be able to save their work. I realise this can be disabled,
> but how many authors will think to change it?

The reason everything exits when your X session dies is this same policy
(Xlib calls exit()) - only with Xlib it can't even be disabled.

One of the main purposes of dbus in fact is to be the "lifecycle
grounding point" for a user session. My expectation if dbus is
successful/useful per-session is that the session will be irrevocably
hosed if the daemon crashes, just as with the X server, so it's best to
avoid orphan processes.

Havoc