On 8/17/2014 3:36 PM, Alan Pevec wrote:
2014-08-17 22:25 GMT+02:00 Matt Riedemann <mriedem@linux.vnet.ibm.com>:
The other thing I thought was we could cap the version of python-keystoneclient in stable/havana, would that be bad? stable/havana is going to be end of life pretty soon anyway.
No, we had cap on some clients and it was creating situations with conflicting requirements, last example was swiftclient<2. Another alternative was to start stable/* from clients but that was rejected in the past. Theory is that *clients are backward compatible but I'm not sure if addition of new dependencies was considered when decision to go with master-only clients was made.
I think it's fine to add new test-requirements on stable, we should just somehow get an early warning that client change is going to break stable branch and update test-req preemptively.
Cheers, Alan
OK, so here is where we appear to be: 1. We need the oslo.utils changes in python-keystoneclient reverted on master to get the stable/havana backports for global-requirements to pass Jenkins. The revert is here: https://review.openstack.org/#/q/status:open+project:openstack/python-keysto... 2. The backports for oslo.i18n and oslo.utils to stable/havana are here: https://review.openstack.org/#/q/status:open+project:openstack/requirements+... 3. Once 1 and 2 are done, we can restore the changes to python-keystoneclient on master. -- Thanks, Matt Riedemann