On this subject, we've found a bunch of regressions where ansible's playbook semantics (especially when it comes to the behaviour of variables in included roles) don't appear to be nailed down.
Is there a regression framework for test cases? I'm happy to submit a set of playbooks that illustrate the variability we're seeing. At the moment we're kind of in the position of having to choose which set of bugs is the least egregious :-) Cheers, jan On Thu, 4 Jun 2015, j...@ioctl.org wrote: > See the attachment. With ansible 1.7.2 this works as expected: > > % ansible-playbook --version > ansible-playbook 1.7.2 > % ansible-playbook -i hosts/localhost a.yml --tags b > > PLAY [localhost] > ************************************************************** > > GATHERING FACTS > *************************************************************** > ok: [localhost] > > TASK: [debug msg="b"] > ********************************************************* > ok: [localhost] => { > "msg": "b" > } > > PLAY RECAP > ******************************************************************** > localhost : ok=2 changed=0 unreachable=0 > failed=0 > > > With 1.9.1 it's broken: > > % ansible-playbook --version > ansible-playbook 1.9.1 > % ansible-playbook -i hosts/localhost a.yml --tags b > ERROR: tag(s) not found in playbook: b. possible values: > > > > > > -- > j...@ioctl.org http://ioctl.org/jan/ Short, dark, ugly: pick any three > Whenever I see a dog salivate I get an insatiable urge to ring a bell. -- j...@ioctl.org http://ioctl.org/jan/ Short, dark, ugly: pick any three Don't annihilate, assimilate: MacDonalds, not missiles.