Comment 23 for bug 578265

Revision history for this message
In , Simon McVittie (smcv) wrote :

Relevant CM parameters:

(telepathy-gabble:10152): tp-glib/params-DEBUG: tp_cm_param_setter_offset: account = "<email address hidden>"
(telepathy-gabble:10152): tp-glib/params-DEBUG: tp_cm_param_setter_offset: password = <hidden>
(telepathy-gabble:10152): tp-glib/params-DEBUG: parse_parameters: server not given, using default behaviour
(telepathy-gabble:10152): tp-glib/params-DEBUG: tp_cm_param_setter_offset: port = 5222 = 0x1466
(telepathy-gabble:10152): tp-glib/params-DEBUG: tp_cm_param_setter_offset: old-ssl = FALSE

So we'll connect to the result of a SRV lookup for _xmpp-client._tcp.chat.facebook.com, or failing that, to chat.facebook.com. However, the SRV lookup fails, and fails with an error that doesn't cause us to fall back.

For anyone who gets this bug, I'd be interested to see the results of these two commands on the same machine:

host -t SRV _xmpp._tcp.chat.facebook.com

host -v -t SRV _xmpp._tcp.chat.facebook.com

From the GIO source code, it seems that the error you got, "Unknown error on connect", is raised with code G_IO_ERROR_FAILED if g_socket_address_enumerator_next or g_socket_address_enumerator_next_finish returns NULL without setting its @error parameter. That looks like a GIO bug, because g_socket_address_enumerator_next returns NULL without setting @error if there are no more addresses... GIO should probably set a more specific error code for that.

Wocky, in turn, interprets the domain G_IO_ERROR to be fatal:

      /* An IO error implies there IS a SRV record but we could not
       * connect: we do not fall through in this case: */