Workflow jobs: Cache miss should cause job failure


CircleCI 2018-01-22 17-19-00

While diagnosing random job failures caused by the cache key file being changed as part of our build_dependencies workflow job, I was misled for a while because the cache restore didn’t abort the job. That seems like an oversight and an easy fix.



How would you ever expect the job to run if a cache fetch failure would cause it to abort before it got the chance to populate the cache?

Also IMO by definition, a cache is “just” an optimization and not a required condition.


If a job success depends on a cache then there’s something off with your build process / cache setup.

Moreover, if a cache miss causes the job to fail, who would the cache get populated in the first place?