On Thu, Nov 14, 2013 at 08:17:06AM -0500, Marc-André Lureau wrote: > > > ----- Original Message ----- > > On Wed, Nov 13, 2013 at 10:13:10PM +0100, Marc-André Lureau wrote: > > > When migrating a channel, all channels are disconnected during reset() > > > after MAIN_MIGRATED_END. The channel swap on destination will happen > > > only after MAIN_INIT is completed. During that time, the channel > > > coroutine can be waiting for socket events, and will have its socket > > > closed and cleared. > > > > > > Add a missing channel state check to fix a warning. (fwiw, the coroutine > > > is then blocked in wait_migration() until migration is fully completed) > > > > > > (remote-viewer:7061): GLib-GIO-CRITICAL **: g_socket_condition_check: > > > assertion `G_IS_SOCKET (socket)' failed > > > --- > > > gtk/spice-channel.c | 3 +++ > > > 1 file changed, 3 insertions(+) > > > > > > diff --git a/gtk/spice-channel.c b/gtk/spice-channel.c > > > index 41d5eab..9eb7ce5 100644 > > > --- a/gtk/spice-channel.c > > > +++ b/gtk/spice-channel.c > > > @@ -2106,6 +2106,9 @@ static gboolean spice_channel_iterate(SpiceChannel > > > *channel) > > > SPICE_CHANNEL_GET_CLASS(channel)->iterate_write(channel); > > > SPICE_CHANNEL_GET_CLASS(channel)->iterate_read(channel); > > > > > > + if (c->state == SPICE_CHANNEL_STATE_MIGRATING) > > > + return TRUE; > > > + > > > > spice_channel_iterate_read() which may be called right before the check you > > add is calling > > g_coroutine_socket_wait(&c->coroutine, c->sock, G_IO_IN); > > Is it ok to do in the situation you describe? > > It's ok, since it should never happen. Ok then, ACK. Christophe
Attachment:
pgplFJExuY0gM.pgp
Description: PGP signature
_______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/spice-devel