Fake compute driver cannot deploy image with hypervisor_type attribute

[Cedric Brandily]

The fake compute driver does not provide the attribute supported_instances to ImagePropertiesFilter scheduler filter.
So ImagePropertiesFilter refuses to deploy images with hypervisor_type=fake property on fake computes.

Consequently, fake computes can not be used in multi hypervisor_types deployments because in this case hypervisor_type property on image is mandatory to avoid mixing one hypervisor_type image with another hypervisor_type compute.

[Cedric Brandily (cbrandily)]

Associated code review is here: https://review.openstack.org/62024

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/62024
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=bdecc8d2339e3e0dd87c7258244ac8568b5b965e
Submitter: Jenkins
Branch: master

commit bdecc8d2339e3e0dd87c7258244ac8568b5b965e
Author: Cedric Brandily <email address hidden>
Date: Fri Dec 13 17:03:19 2013 +0100

    Define "supported_instances" for fake compute

    Allow fake computes to deploy images with hypervisor_type
    defined and set to "fake" and by extension to be used in
    multi hypervisor_type deployments.
    Mixing fake and "real" computes on the same OpenStack
    deployment allows to create many fake instances and some
    working instances for testing applications on top of
    OpenStack (ie: fake instances for stress tests, working
    ones for functional tests).

    Closes-bug: #1260771
    Change-Id: Idc337c4a7ed024f236ca2b60d91e2c30f7d54536