diff options
| author | Simon McVittie <simon.mcvittie@collabora.co.uk> | 2009-01-28 10:27:29 +0000 |
|---|---|---|
| committer | Simon McVittie <simon.mcvittie@collabora.co.uk> | 2009-01-28 10:27:29 +0000 |
| commit | 5e3f547b948e4f9f36aa0154b5565494b3da0b22 (patch) | |
| tree | 251e4c259fc8aee93f6119143c5d0a077b2d071d /tests/dbus/text-respawn.c | |
| parent | 1f4242baa58082cc2e96f560682f67718005336a (diff) | |
tests: call tp_get_bus() earlier, to initialize specialized types
Instantiating a DBusGConnection (which is done by tp_get_bus()) is
the only safe way to ensure that dbus-glib is initialized (see
fd.o #13908 for why dbus_g_type_specialized_init() isn't safe).
In principle, we need to do this initialization before creating
TpBaseConnection subclasses.
Previously, this was coincidentally avoided because Connection
didn't actually use any specialized types, but the ConnectionError
signal that will be in the next spec release introduces a
specialized type.
Mitigating factor: connection managers should all be using
tp_run_connection_manager(), which already calls tp_get_bus().
Diffstat (limited to 'tests/dbus/text-respawn.c')
| -rw-r--r-- | tests/dbus/text-respawn.c | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/tests/dbus/text-respawn.c b/tests/dbus/text-respawn.c index f04045789..f2b5ee8cc 100644 --- a/tests/dbus/text-respawn.c +++ b/tests/dbus/text-respawn.c @@ -103,6 +103,7 @@ main (int argc, g_type_init (); /* tp_debug_set_flags ("all"); */ + dbus = tp_dbus_daemon_new (tp_get_bus ()); service_conn = EXAMPLE_ECHO_CONNECTION (g_object_new ( EXAMPLE_TYPE_ECHO_CONNECTION, @@ -117,7 +118,6 @@ main (int argc, &name, &conn_path, &error), ""); MYASSERT_NO_ERROR (error); - dbus = tp_dbus_daemon_new (tp_get_bus ()); conn = tp_connection_new (dbus, name, conn_path, &error); MYASSERT (conn != NULL, ""); MYASSERT_NO_ERROR (error); |