Dan Smith wrote:
KR> The change to "from_str()" here is so that one would be able to KR> pass the classname in. The logic for determining the device / KR> pool type is the same for the InstanceID and the classname.
Right, but we could potentially need to look at more of the ID in some cases to determine the type, right? If we had disk pools that represent LVMs separately from file images or something like that. I suppose it's reasonable to think that the classname would always indicate the subtype, but it just seems cleaner to me to always do it From an ID.
Ah, yes - this is a really good point. Hmm.. cu_compare_ref() seems like it could be used , except that the handler would need to construct an instance just for the purpose of comparing it against the ref, which seems a little unnecessary to me. It's probably best to leave device_type_from_poolid() as is, but devise some other way to further make sure the ref is a valid one. Ideally, GetInstance() from Virt_DevicePool should be able to check whether the ref is valid. However GetInstance() also supports refs with mis-matching InstanceIDs and class types. =) So, if this issue is fixed in Virt_DevicePool, the same logic / function could be reused in RAFP and EAFP. I'll see if I can cook up something a little more elegant in Virt_DevicePool. Thanks for the feedback! =) -- Kaitlin Rupert IBM Linux Technology Center karupert@us.ibm.com