[virt-tools-list] cpu model list population

[Cole Robinson]

On 03/21/2013 07:59 AM, Pete Beardmore wrote:
> 
> hi
> 
> [Thu, 21 Mar 2013 10:53:29 virt-manager 16554] ERROR (details:872) Error
> populating CPU model list
> Traceback (most recent call last):
>   File "/usr/local/share/virt-manager/virtManager/details.py", line 867, in
> init_details
>     cpu_values = caps.get_cpu_values(self.vm.get_arch())
>   File
> "/usr/local/lib64/python2.7/site-packages/virtinst/CapabilitiesParser.py",
> line 672, in get_cpu_values
>     self._cpu_values = CPUValues()
>   File
> "/usr/local/lib64/python2.7/site-packages/virtinst/CapabilitiesParser.py",
> line 115, in __init__
>     xml = file(cpu_filename).read()
> IOError: [Errno 2] No such file or directory: '/usr/share/libvirt/cpu_map.xml'
> 
> '/usr/share/libvirt/cpu_map.xml' is a hard-coded default in
> 'CapabilitiesParser.py' for use when no additional arg specifying the correct
> location is given to instantiate the CPUValues class. trouble is, nowhere is
> this constructor called with that additional arg. hence is this 'wip', waiting
> for the addition of a config file and parse mechanism? or maybe i'm missing
> something obvious?
> 
> as this issue wasn't invasive (error dialog), all my machines have been set up
> with defaults for this parameter. i'm hoping being able to play with cpu model
> may improve my poor success (to date) running guest xyx without crashes
> (fingers crossed anyway) and as such i've just modified the path and am ready
> to move on. i'm raising this on the off chance somebody here cares

Yes, hardcoding this is not very nice, however there isn't really a better
option at the moment. There's an RFE for libvirt to expose this data over an
API, which would solve this in a nice way:

https://bugzilla.redhat.com/show_bug.cgi?id=916786

- Cole