[Sorry for sending this again, I think it didn’t go through to the list the first time because I used my old address.] On Jul 25, 2014, at 4:54 AM, Ihar Hrachyshka <ihrachys@redhat.com> wrote:

Signed PGP part On 25/07/14 08:05, jenkins@openstack.org wrote:

...

- periodic-heat-python26-havana http://logs.openstack.org/periodic-stable/periodic-heat-python26-havana/3db9... : FAILURE in 46s

libffi-devel package is probably missing on the node, that results in cffi compilation failure. Should we ask infra to check that it's installed on all nodes?

I thought that those packages are to be installed via [1]. Is it correct? Because the script seems to be Ubuntu specific, and the machine that we got those tests running on is CentOS, where there is no apt-get or libffi-dev. Should we have another integration script for CentOS? Doug?

[1]: https://github.com/openstack-dev/pbr/blob/master/tools/integration.sh

/Ihar

That integration script in the pbr repository is just for testing pbr and requirements changes (and we took out the requirements tests in juno). I’m not sure the heat periodic tests run that script, do they? Packages are installed on the test servers in a couple of different ways. For devstack-gate, the list of packages in devstack is used. The integration script you mentioned above is doing it for some of the packages it uses. But I think the heat tests that are failing are just running on a regular unit test node. I believe you want to update the puppet manifest that describes those test nodes to include the new packages. I see libffi_dev_package in openstack-infra/config/modules/openstack_project/manifests/thick_slave.pp but I don’t know if that’s the right package or manifest. Doug