Hi Aleksander,
Yep, that works too. How about this?
diff --git a/src/mm-iface-modem.c b/src/mm-iface-modem.c
index 49a18a5..7b44d6a 100644
--- a/src/mm-iface-modem.c
+++ b/src/mm-iface-modem.c
@@ -3676,6 +3676,14 @@ mm_iface_modem_shutdown (MMIfaceModem *self)
{
g_return_if_fail (MM_IS_IFACE_MODEM (self));
+ /* Remove SignalQualityUpdateContext object to make sure any pending
+ * invocation of expire_signal_quality is canceled before the DBus
skeleton
+ * is removed. */
+ if (G_LIKELY (signal_quality_update_context_quark))
+ g_object_set_qdata (G_OBJECT (self),
+ signal_quality_update_context_quark,
+ NULL);
+
/* Remove SIM object */
g_object_set (self,
MM_IFACE_MODEM_SIM, NULL,
Thanks,
Ben
On Wed, Jul 11, 2012 at 9:49 PM, Aleksander Morgado
<[email protected]>wrote:
> Hey Ben,
>
> > expire_signal_quality() may be called after the DBus object skeleton is
> > removed in mm_iface_modem_shutdown() and before
> > signal_quality_update_context_free() is called due to the destruction of
> > the modem object. This patch prevents expire_signal_quality() from
> > accessing an already removed skeleton.
>
> Ugly race it is :-/ So the data set with g_object_set_qdata_full() gets
> actually destroyed *after* dispose() has run, right? Just wondering if
> it isn't better to ensure in each interface's _shutdown() method that we
> clear any data association we may have set previously, by calling again
> g_object_set_qdata() with a NULL pointer. That should call the
> GDestroyNotify of each data associated previously, if any, and it will
> get the timeouts removed from the main loop as well. That would also
> help for the cases where, due to some bug in some other place, we let a
> reference to the modem object around, effectively leaving the timeouts
> running even after we think everything has been disposed.
>
> What do you think?
>
> Cheers!
>
> --
> Aleksander
>
>
>
_______________________________________________
networkmanager-list mailing list
[email protected]
https://mail.gnome.org/mailman/listinfo/networkmanager-list
