![Boden R](/assets/img/avatar_default.png)
While the intention of the incubating hacking check factory was to provide a means to roll-out new checks without breaking consumers [1], the complications involved with running 2 factories in a gate job [2] perhaps negate the benefits of this approach. Therefore, as per discussion with Armando (see the neutron IRC logs from Jan 9, 2017 at 18:58) this patch proposed we only have a single hacking check factory for all consumers. The implications are as follows: - All new adopter hacking checks go into the single factory. - Consumers failing the new pep8 hacking check will see the failure via neutron-lib periodic job. - Consumers fix failures ASAP. IFF a special case arises and we need to release neutron-lib before consumers adopt check(s), we can comment them out of the factory to mitigate as a whole. Or individual consumer project's can 'ignore' failing checks as outlined in the usage.rst of this patch. This patch updates the dev-ref as well as the code to implement the above scheme. [1] https://review.openstack.org/#/c/350723/ [2] https://review.openstack.org/#/c/412628/ Change-Id: I0a8b39052c0f81f2ecbab3923353be1537b89a65
7 lines
224 B
YAML
7 lines
224 B
YAML
---
|
|
other:
|
|
- The hacking check factory function
|
|
``neutron_lib.hacking.checks.incubating_factory`` has been removed.
|
|
All consumers should use ``neutron_lib.hacking.checks.factory`` as per
|
|
the ``usage`` dev-ref.
|