[virt-tools-list] libosinfo 1.0 not reporting device class?

[Cole Robinson]

On 10/26/2016 11:40 AM, Daniel P. Berrange wrote:
> On Wed, Oct 26, 2016 at 12:00:06AM +0200, Daniel P. Berrange wrote:
>> On Tue, Oct 25, 2016 at 11:58:17PM +0200, Daniel P. Berrange wrote:
>>> On Tue, Oct 25, 2016 at 04:05:34PM -0400, Cole Robinson wrote:
>>>> libosinfo 1.0 seems to be causing regressions with virt-manager on Fedora 25.
>>>> Just run the virt-manager test suite and check the failures. Not sure if this
>>>> is an upstream bug or a packaging issue though. Simple reproducer is:
>>>
>>> Hmm, that's odd - I explicitly got the virt-manager test suite running in
>>> the libvirt centos CI, specifically to test that I didn't cause regressions.
>>>
>>> So builds from GIT at working at least, which suggests packaging bug as
>>> the issue.
>>>
>>>> Notice that get_class() returns None. virt-manager does all the device support
>>>> matching by filtering on device class, and it seems class=None is breaking
>>>> that. get_name() returns empty results as well which will also break
>>>> virt-manager. I haven't dug into the 'why' though
>>>
>>> Most device info comes from the pci.ids file, but class info is added via
>>> an XML override, so seeing the devices without class, means the override
>>> is missing.
>>
>> Oh, I see its a bug in the currently released osinfo-db package, already
>> fixed in git by:
>>
>> commit 47bdf50c47c59eb74d146d08c1062988440f9aa9
>> Author: Daniel P. Berrange <berrange at redhat.com>
>> Date:   Wed Aug 10 13:00:57 2016 +0100
>>
>>     Include data files from 4th level dirs too
>>     
>>     Most XML files are stored at the 3rd level, but for
>>     devices we have some overrides which exist at the
>>     4th level.
>>
>> I need to do a new release with this commit added
> 
> I've done a release and pushed updates to Fedora 25 testing

Thanks Dan, it fixes the breakage here.

- Cole