Emacs logs AT-SPI warnings to console on startup

[Tuure Laurinolli]

When starting up certain (GTK?) applications from console in a custom X session, such as emacs or ubuntu-bug, there is a lot of log output like:

"""
** (emacs:4437): WARNING **: 15:05:10.026: AT-SPI: Could not obtain desktop path or name
"""

or

"""
** (apport-gtk:175957): WARNING **: 15:16:16.747: AT-SPI: Could not obtain desktop path or name
"""

The issue seems to be related to at-spi2-atk, or at least that's where the code that prints the error is:

In register_reply in bridge.c:
  if (reply)
    {
      gchar *app_name, *obj_path;

      if (strcmp (dbus_message_get_signature (reply), "(so)") != 0)
        {
          g_warning ("AT-SPI: Could not obtain desktop path or name\n");
        }

Which seems to be triggered by an attempt to register the application to ATK:
static gboolean
register_application (SpiBridge * app)
{
  DBusMessage *message;
  DBusMessageIter iter;
  DBusPendingCall *pending;

  g_free (app->desktop_name);
  g_free (app->desktop_path);

  /* These will be overridden when we get a reply, but in practice these
     defaults should always be correct */
  app->desktop_name = g_strdup (ATSPI_DBUS_NAME_REGISTRY);
  app->desktop_path = g_strdup (ATSPI_DBUS_PATH_ROOT);

  message = dbus_message_new_method_call (SPI_DBUS_NAME_REGISTRY,
                                          ATSPI_DBUS_PATH_ROOT,
                                          ATSPI_DBUS_INTERFACE_SOCKET,
                                          "Embed");

  dbus_message_iter_init_append (message, &iter);
  spi_object_append_reference (&iter, app->root);

    if (!dbus_connection_send_with_reply (app->bus, message, &pending, -1)
        || !pending)
    {
        if (pending)
          dbus_pending_call_unref (pending);

        dbus_message_unref (message);
        return FALSE;
    }

    dbus_pending_call_set_notify (pending, register_reply, app, NULL);

It seems to be caused by nobody being available at the ATSPI_DBUS_PATH_ROOT to handle the message. I have not yet been able to figure out what *should* be handling the message.

[Tuure Laurinolli (tuure.laurinolli+vaisala.com)]

Apparently this is solved by running /usr/lib/at-spi2-core/at-spi2-registryd. Normally it seems to be launched by gdm, but my session is not started through gdm.

[Nicholas P (dgtlcmo)]

Launching Emacs outside of gdm seems like a typical use case for a headless client, for instance using Virtualbox and ssh. I also noticed placing this in the environment:

export NO_AT_BRIDGE=1

resolves the error by preventing this AT extension from loading. My Emacs load time is much faster.

Why was the change integrated and should it be backed out?