Re: [patch] fix up behavior on panel crash

[Dan Williams <dcbw redhat com>]

On Tue, 2005-08-09 at 16:13 -0400, Robert Love wrote:
> Dan,
> 
> It looks like we try to gracefully handle panel crashes (we catch the
> destroy signal, etc.) but we don't.
> 
> Three problems:
> 
> 	- The GtkWindow is already gone by the time we call
> 	  nmwa_destroy, generating warnings.
> 	- g_thread_join() fails because the dbus thread is not
> 	  joinable.
> 	- We never actually exit.
> 
> Don't know how to fix the first thing, or if we want to.

I'm not sure that we can either, if our window is dead, we pretty much
just want to die ourselves.  I don't think we can create another applet
instance anyway.

> The second is fixed by either not waiting on the thread, or making the
> thread joinable.   Since it makes sense to wait for the thread to die
> (and if we do wait, it will die), we should make the thread joinable and
> then wait on it.  One-line patch.
> 
> The third is fixed by simply calling exit() at the end of
> nmwa_destroy().  This seems right, but are we supposed to die through
> some other implicit method?
> 
> Patch is attached.
> 
> With it, we successfully regenerate (via session management) when the
> panel dies.
> 
> Look right?  May I commit?

Sure, looks file.  Go ahead.

Dan