Re: [PATCH OSSTEST v4 3/3] Create a flight to test OpenStack with xen-unstable and libvirt

[Anthony PERARD <anthony.perard@citrix.com>]

On Tue, Dec 08, 2015 at 10:23:49AM +0000, Ian Campbell wrote:
> On Fri, 2015-11-20 at 14:55 +0000, Anthony PERARD wrote:
> 
> I think this will also add a devstack job to most other flights? That's a
> good thing, I think.

That was not my intention. It's true that the more test, the better, but on
the other hands it takes about 1h to deploy+test openstack, after
installing the host.

> I wonder if the flight ought to be called openstack-nova, to leave open the
> possibility of other openstack-$foo flights in the future?

Yeah, that probably good to do.

> > Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
> > 
> > ---
> > Change in V4:
> > - also skip build-*-oldkern in make flight
> 
> Those should already be caught by cr-daily-branch
> setting REVISION_LINUX_OLD=disable for all but the xen-unstable flight
> (arguably that default ought to be inverted, such that make-flight by
> default doesn't make such flights unless asked to, but that's not your Yakk
> I think)
> 
> > - fix select_xenbranch
> > - set revision_*=$REVISION_OPENSTACK_* in make-flight
> >   (was revision_*=master before)
> >   only REVISION_OPENSTACK_NOVA is set, the others are unset.
> >   empty revision_* runvar would clone the default branch, which should
> >   be master for every openstack repos
> 
> The sort of info in this last bullet would be useful in the commit message,
> I think.

Yes, I should add that to the commit, to at least let know the intention of
the author.

> > diff --git a/make-flight b/make-flight
> > index 8523995..5a4fc0c 100755
> > --- a/make-flight
> > +++ b/make-flight
> > @@ -54,6 +54,12 @@ job_create_build_filter_callback () {
> >          *)			return 1 ;;
> >        esac
> >      ;;
> > +    openstack)
> > +      case "$job" in
> > +        *-xsm) return 1;;
> 
> I wonder, would a test-$xenarch$kern-$dom0arch-devstack-xsm be a useful
> think to have though?

Probably, that would test xsm with a different scenario.

> > +        *-oldkern) return 1;;
> 
> See above.

Ok, I guest I can remove this.

> > diff --git a/mfi-common b/mfi-common
> > index 5fbe195..f11302b 100644
> > --- a/mfi-common
> > +++ b/mfi-common
> > @@ -59,6 +59,7 @@ xenbranch_xsm_variants () {
> >      xen-4.3-testing) echo "false";;
> >      xen-4.4-testing) echo "false";;
> >      xen-4.5-testing) echo "false";;
> > +    openstack)       echo "false";;
> >      *) echo "false true";
> >      esac
> >  }
> > @@ -102,6 +103,7 @@ create_build_jobs () {
> >        rumpuserxen) continue;;
> >        seabios) continue;;
> >        ovmf) continue;;
> > +      openstack) continue;;
> >        esac
> >        case "$xenbranch" in
> >        xen-3.*-testing) continue;;
> > @@ -127,6 +129,9 @@ create_build_jobs () {
> >        "
> >        ;;
> >      esac
> > +    if [ "$arch" = i386 ] && [ "$branch" = openstack ]; then
> > +      continue
> 
> This accepts ARM, but I think you filter the test cases for that? The
> filtering of build vs. test jobs in make-flight/mfi-common is a bit of a
> mess, recently e21625f79d33 "make-flight: move in-lined branch vs arch
> filtering into callbacks" make it a bit less bad, which might be useful to
> you here?

Yes, ARM should be filter elsewere. I'll give a look at this commit.

Thanks,

-- 
Anthony PERARD