It does it's job for a while and then I've just noticed that NMI will disappear
so I ran NM in gdb and caught this.  It's pretty reproduceable, just takes some
time to happen again.

Here's some backtrace stuff:

NetworkManager: <information>   Restarting mDNSResponder.
NetworkManager: <information>   Activation (eth0/wireless): Success!  Connected
to access point 'Eric The Red' and got an IP address.
NetworkManager: <information>   Activation (eth0) IP configuration/DHCP successful!
NetworkManager: <information>   Activation (eth0) ended.
NetworkManager: <information>   Activation (eth0) successful, device activated.

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread -1208047040 (LWP 15232)]
0x00000000 in ?? ()
(gdb) bt
#0  0x00000000 in ?? ()
#1  0x00d9e9f1 in libhal_psi_get_strlist () from /usr/lib/libhal.so.1
#2  0x05e5315a in dbus_connection_dispatch () from /usr/lib/libdbus-1.so.1
#3  0x0087b21e in dbus_g_pending_call_cancel ()
   from /usr/lib/libdbus-glib-1.so.1
#4  0x00b25d4c in g_io_unix_dispatch (source=0x97bee80, callback=0,
    user_data=0x97c1c00) at giounix.c:162
#5  0x00b0046e in IA__g_main_context_dispatch (context=0x97bc590)
    at gmain.c:1947
#6  0x00b03476 in g_main_context_iterate (context=0x97bc590, block=1,
    dispatch=1, self=0x97ba608) at gmain.c:2578
#7  0x00b03763 in IA__g_main_loop_run (loop=0x97bbf78) at gmain.c:2782
#8  0x0805ab87 in main (argc=2, argv=0xbfee9a64) at NetworkManager.c:910
(gdb) list
910             g_main_loop_run (nm_data->main_loop);
911
912             /* Cleanup */
913             libhal_ctx_shutdown (nm_data->hal_ctx, &dbus_error);
914
915             if (dbus_error_is_set (&dbus_error)) {
916                     nm_warning ("libhal shutdown failed - %s",
917                                 dbus_error.message);
918                     dbus_error_free (&dbus_error);
919             }