diff -r 0ecb3a22c361 -r 52f8b84c5bc6 suites/libvirt-cim/lib/XenKvmLib/assoc.py --- a/suites/libvirt-cim/lib/XenKvmLib/assoc.py Fri Sep 26 02:44:52 2008 -0700 +++ b/suites/libvirt-cim/lib/XenKvmLib/assoc.py Tue Sep 30 05:06:02 2008 -0700 @@ -42,13 +42,16 @@ def AssociatorNames(host, basetype, base type = basetype.split('_') if len(type) == 2: basetype = type[1] - - obj = baseobj.split('_') - if len(type) == 2: - baseobj = obj[1] + + if baseobj != 'Linux_ComputerSystem': + obj = baseobj.split('_') + if len(obj) == 2: + baseobj = obj[1] + obj = get_typed_class(virt, baseobj) + else: + obj = baseobj
I don't think the Linux_ComputerSystem classname should be embedded in the association calls. The ConcreteComponent tests will need to support Linux_Processor and Linux_EthernetPort, which would mean this would need to be updated to include those as well. Really, the association calls need to be updated so that they take a classname param instead of base and virt parameters. That way, we build the classname we need in the test case itself and pass that to the association call. That's a lot to change at the moment, but I think it will be good in the long run. It'll allow us to support other SBLIM classnames (besides just Linux_ComputerSystem) and anything else we need to support in the future. -- Kaitlin Rupert IBM Linux Technology Center kaitlin@linux.vnet.ibm.com