Use API unit tests requests on all backends

Do we? It seems like if we have tests verifying that all of the backends behave the same way, then we don't need to run the API tests against all of the backends.

It would really be a good idea. There's no check nor conversion between what's returned by the storage backend and what's return though the API, so we can have some driver returning valid Python values but not valid for WSME. This already happened twice with bug #1132628 and an older one.

I agree there's a problem with the database API, I'm just not certain fixing it requires making the test suite run take so much longer. For example, I've opened a blueprint to define database models for the storage API (https://blueprints.launchpad.net/ceilometer/+spec/storage-api-models).

I wouldn't disagree with such a solution Doug. I just prefer to "fix" or possibly detect failure of our current existing code first, and that just requires us to enable more tests. :)

Reviewed: https://review.openstack.org/23302
Committed: http://github.com/openstack/ceilometer/commit/c41249e488c553c3cbd69fa54873fb4ebcaf86f1
Submitter: Jenkins
Branch: master

commit c41249e488c553c3cbd69fa54873fb4ebcaf86f1
Author: Julien Danjou <email address hidden>
Date: Fri Mar 1 18:09:11 2013 +0100

    Run some APIv1 tests on different backends

    This is a first change towards implementing bug #1131638. It clears a lot of
    stuff in the various DB mocking stuff, and I plan to do more after.

    In the same process, this implements a unit test and a fix for bug #1138137

    Change-Id: I36a6741393455c764913e0388d73d834db06da8c
    Signed-off-by: Julien Danjou <email address hidden>