test suite failure - more detail ...

[Havoc Pennington]

Hi,

When GetServiceOwner returns SERVICE_DOES_NOT_EXIST, we're overloading
that error; it would normally mean "the service you invoked the method
on does not exist," which is different from "the service you're asking
for the owner of does not exist."
It might be cleaner to have DBUS_ERROR_SERVICE_HAS_NO_OWNER or
something? Other opinions on this?
Imagine some generic code that tried to detect sending a message to a
nonexistent service and use that info somehow.

For bus_connection_get_base_service(), it may currently be guaranteed
that the first service in the list is always the base service since you
have to have a base service to do anything, and you can never disown
your base service. If so that'd be a nice optimization and code savings,
and I think safe as long as the test suite covers the code.

Otherwise the patch looks nice, thanks.

Havoc