Compute API version >= 2.37 requires a 'networks' value in
the server create request. The novaclient CLI defaults this
to 'auto' if not specified, but the novaclient ServerManager.create
python API binding code does not, as it wants clients to be explicit.

For the purposes of the OSC CLI, we should follow suit and if the
user is requesting OS_COMPUTE_API_VERSION>=2.37 without specific
nics, we should just default to 'auto'.

Change-Id: Ib760c55e31209223338a4086ff1f4fee88dc6959
Closes-Bug: #1750395
(cherry picked from commit 1008544882fbdae16b045abca05cf3e2e8a14787)
(cherry picked from commit 4944d4eaa9aaeffdee358680550ac7d7f0c6e8d8)

Change-Id: I27afca9e826378dbcb7feb7528e0c65c528b04b0
Closes-Bug: #1703278

So yeah, this is not kosher for functional tests, but we're testing the
client interaction, not the raciness of Nova or Neutron.  Failure to delete
is not our problem.

Change-Id: I21043f1de0fbacee1aec63110fb12a7cff42e0a0

To ensure project imports are placed after third party import,
we need to specify application-import-names.
Previously flake8-import-check checked only standard imports or not.

Change-Id: Iad7afa456cec7cf5b44955f1ea03c593a4c0e426

Use fixtures to restore the API version changes of os.environ
in each functional tests, aims to avoid the following test cases
failing in unexpected context.

And make sure setUpClass/tearDownClass call super class's
corresponding methods first.

Change-Id: Ie248fe9d3a9e25f1b076c9f2c363200f29a83817
Closes-Bug: #1696080

Deleting snapshot may take time. The current volume API does not allow
to delete volumes with snapshots, so if deleting snapshot may take time,
a delete request for a parent volume will fail.
This sometimes causes functional test failures in slow environments.

wait_for_status() checks whether volume status is in error statuses
but previously the expected error status was wrong. Cinder API uses
lower case as volume status, so it did not work expectedly.

Change-Id: I095894ba39f23bf81d71351818d24dbb5ca459fb

Both resource ID and name are supported to identify an object
in openstackclient to make user easy to input, for security group,
nova only support security group name in API when launch a new server,
this patch convert ID to name, then pass name to nova API, and check
the security group exist before creating server.

Change-Id: I1ed4a967fb9de3f91c8945a1ef63f6c7b6b2dfb2
Closes-Bug: #1687814

The patch fix the following issues:

1. ValueError is raised if input don't contain "=". Sometimes the whole
"server create" command is very complex, it's difficult to find out root
reason directly.
2. Don't support to add block device from snapshot, like:
--block-device-mapping
vdb=0c8ae9d8-cadc-4a23-8337-4254614d277e:snapshot:1, it's supported by
novaclient, but not in osc.
3. If input "vdb=", not add any mapping information, the server will be
launched successfully, not raise error message to let use add
volume/snapshot id, just ignore "--block-device-mapping" option.
4. The help message of "block-device-mapping" option is so simple, need
to add some details about how to add <type>, <delete_on_terminate>
contains.

Change-Id: Ib7f7a654c3dc2a8272545f168b4c4ced230ce39e
Depends-On: Ib37913891bbf7a31b570404c4668c490d5ac859b
Closes-Bug: #1667266

As of Ocata release Nova forces nova-network to run in a cells v1
configuration.  Floating IP and network functions currently do not
work in the gate jobs so we have to skip this.  It is known to work
tested against a Mitaka nova-net DevStack without cells.

Change-Id: I74f67ac8eb12c7a649ddcbd7979cf745fb35cc0c