On 11/26/2014 04:09 AM, Martin Kletzander wrote:
>>>>
>>> The problem is with expected results that are coded for these tests.
>>> On distros that support these flags, the issue would go away if the
>>> expected results take into account the locking flags. However,
>>> adding a permanent change to the expected args string would break
>>> older distros.
>> Actually, no, I wanted to unconditionally add the parameters there
>> only for tests.
>>
>> Looking at it more closely, this can fail only if you are building as
>> root, is that correct?
>
> Yes, that is correct.
>
I'm testing a patch with different approach, I'll Cc you on that when
it is done. Let me know whether that works for you then.
It looks like your work is duplicated with Stefan's patches:
https://www.redhat.com/archives/libvir-list/2014-November/msg01022.html
Hmm, Stefan's approach was to unconditionally DISABLE the locking
parameter in the tests, but I'm now wondering if unconditionally
ENABLING the parameter, and adjusting the expected test output, is the
better course of action. At any rate, I agree that we want the tests to
be independent of what the actual installation provides.
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library
http://libvirt.org