Comment 57 for bug 857153

(In reply to Trevor Saunders (:tbsaunde) from comment #49)
> well, the firstthing that comes to mind is arranging for build machines and
> test machines to have GNOME_ACCESSIBILITY set, do we have an idea how hard
> this is?

It would be good to ensure there is a session bus, because we'd like to test the dbus code. Perhaps DBUS_SESSION_BUS_ADDRESS is not set because of the way our testing connects to the existing gnome session. Or perhaps the CentOS 5.0 machines are just too old to have a session bus running. On these systems the GConf code can provide information on whether accessibility is enabled.
If the test slaves were fine on comment 39's landing and it was only build machines that had failure, then that might be because the tests slaves are running a newer OS, one likely to have a DBus session, and DBUS_SESSION_BUS_ADDRESS may be set there.

You might need to ask someone from releng how our build/test slave starts up and whether that inherits environment variables from the gnome session.

If you only need accessibility enabled for accessibility tests then it probably makes sense to set GNOME_ACCESSIBILITY in the Makefile command that runs the accessiblity tests.

> I'd expect though that in the vast majority of cases that there is a
> session and this is not relavent. Karl would you object if I merge that
> change set?

No I would not object. Please add a comment to explain why the DBUS_SESSION_BUS_ADDRESS environment variable is there though.

The situation with dbus when it can't connect to a session bus is unfortunate, and perhaps something worth avoiding anyway, if we can.