By not cloning when building the base image, we speed up the initial builds a little. We could also kick off the base image pipeline in a separate workflow, and it would surely complete in time for a full spack build (the spack job only needs it to upload to the remote registry.
There's a potential race condition with that if the spack pipeline fails quickly... so I don't see a need to do that.
In debugging https://github.com/LLNL/hiop/pull/695, I realised that the clone happens at multiple steps in the pipelines, and doesn't need to.
By not cloning when building the base image, we speed up the initial builds a little. We could also kick off the base image pipeline in a separate workflow, and it would surely complete in time for a full spack build (the spack job only needs it to upload to the remote registry.
There's a potential race condition with that if the spack pipeline fails quickly... so I don't see a need to do that.