[OpenStack-Infra] Too much optimizations in project-config jobs ; (

Jeremy Stanley fungi at yuggoth.org
Tue Dec 1 16:33:56 UTC 2015


On 2015-12-01 09:22:38 +0100 (+0100), Andreas Jaeger wrote:
> We recently merged changes to limit the running of jobs on
> project-config to run only when a job is needed.
> 
> But now we have a change [1] to data/* which is not covered by any
> of the existing jobs and thus no job is run.
> 
> What shall we do? Always run one of the existing jobs? Extend a
> job to cover all files? Revert the last change for this?
[...]

I'm not really a fan of skipping jobs based on which files are
changed since it can lead to gaps in effective test coverage (or not
running any jobs at all in cases like this one). It's too easy to
get wrong, or to fall out of step with reorganizations/refactors
within the repository being tested. I'm in favor of running all
project-config jobs on all project-config changes, or at worst only
applying this workaround to jobs which take a really long time to
complete (e.g., nodepool integration testing).
-- 
Jeremy Stanley



More information about the OpenStack-Infra mailing list