g_dbus_error_new_for_dbus_error() describes three cases (1. error domain is registered, 2. @dbus_error_name is in the form returned by g_dbus_error_encode_gerror(), 3. anything else). For the second two cases, it explicitly notes "@dbus_error_name is added to the error message such that it can be recovered with g_dbus_error_get_remote_error()." It *does not* say that for the first case, and in fact, g_dbus_error_get_remote_error() is perfectly capable of recovering the D-Bus error name from a registered error domain without the message being prefixed. This all would tend to imply that g_dbus_error_new_for_dbus_error() should not be prefixing the error message in the registered-error-domain case.

But it does. At first I assumed that this was a bug, and wrote a patch, but then "make check" fails because gio/tests/gdbus-error.c actually explicitly verifies the current behavior. So maybe its the docs that are wrong, and that functionality was decided against? (And/or, we're stuck with that behavior now for compatibility reasons?)