Open Stack

----- Original Message -----
> On 03/19/2016 01:43 AM, Tony Breeds wrote:
> > On Sun, Mar 13, 2016 at 12:15:23PM +0000, Amrith Kumar wrote:
> >> Hi,
> >>
> >> Sorry for the blast mail, this is targeted at the stable-maint team. The
> >> following five changes are now ready to merge. Two of them are changes
> >> proposed by the bot and three of them are clearly identified as
> >> cherry-picks
> >> from changes that went into master.
> > <snip>
> >
> >> Add instance create int-tests<https://review.openstack.org/262287>
> >> Add MySQL int-test helper client<https://review.openstack.org/262289>
> > It looks like it's down to just those 2 which are testing only backports.
> >
> > From a stable POV I'm a little on the fence about applicability.
> >
> > Ignoring that right now both reviews seem to be focused around "scenarios"
> > testing but trove doesn't run gate-trove-scenario-functional-dsvm* on
> > stable
> > and on master they seem to be non-voting experimental jobs.
> >
> > Did I misunderstand what these reviews are doing?
> 
> Tony, your understanding is correct. These requests led to a lot of
> discussion and the fact that these are non-voting experimental jobs was
> discussed at length. It was felt that these would be useful for people
> who have to support stable/liberty. But, as one of the reviewers points
> out, this request opens a can of worms. Are we implying that over time
> there are going to be more fixes to get these tests working? And I don't
> think any of us is comfortable with the answer to that being "yes".

No one said that there will be other backports, at this time. It would have made sense if the reviews would have been merged when they have been proposed. It would be nice thought to backports one or two other patches for the integration tests, provided that the two disscused here are merged, but let's focus on them first.

Without the code from the two reviews, the code for scenario tests _shipped_ in Liberty is unusable.
Scenario tests are the new tests which are going to replace the current integration tests. Scenario tests are not used on the CI right now (which also means that the backport is not going to break anything), the plan was to switch to them in Mitaka, but most likely now in Newton.

Now, the integration tests currently used on the CI makes it very difficult, if not impossible, to test any datastore which is not MySQL - including for example MariaDB. Merging the two reviews would make the life a bit easier for people which use Liberty (which is for a while still the stable release, and it will be supported for other few months).

My summary is:
Does they break some code? No.
Does they fix broken code? Yes.
Does this open a can of worm? IMHO no, there is no impact on the functionalities even in the "worst" case of other reviews being merged.

Hope this helps.

Ciao
-- 
Luigi