On Mon, Aug 10, 2015 at 05:12:26PM +0300, Pavel Fedin wrote:
Hello!
> And that's the exact problem. I'm not saying "capabilities are not
> made available when defining", but rather "there are no capabilities
> when defining".
Stop stop stop... If we don't mess up with terminology...
OK, sorry for that, I'll be clear from now on.
The test case contains some hardcoded capabilities, needed by the
test, right? Ok, this means that
Let's say we're speaking about qemuxml2xmltest, just to make it as
easy as possible. That test ultimately checks parsing and formatting
and parsing is what we're talking about here.
That test has *no* hardcoded capabilities. And the reason why it has
no hardcoded capabilities is what I'm trying to express here.
actually we have this information, and we need to make up capabilities
cache from these caps, and
there will be no problem.
With this change the test will exactly mimic real life behavior, where we have the cache,
and this
cache was obtained from the qemu binary.
I don't know whether to explain what's wrong here just so we don't get
confusing again. But I must try.
In libvirt, we are not using capabilities when parsing/formatting XML
because it [parsing/formatting] must a) work without any qemu
binary and b) produce exactly same results no matter what is installed
in the system. Is that any clearer now by any chance?
Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia