Do we really need an access-esm1.5 branch?

[dougiesquire]

The situation at the moment, as I understand it, is a little messy:

• ACCESS-OM2 uses the master branch which essentially adds some CI on top of mom-ocean/MOM5 master
• ACCESS-ESM1.5 uses the access-esm1.5 branch

Modifications to MOM5 needed only in ACCESS-OM2 require:

• PR to mom-ocean/MOM5 master, review, merge
• PR to update ACCESS-NRI/MOM5 master, review, merge
• Additional PRs to update ACCESS-OM2

Modifications to MOM5 needed only in ACCESS-ESM1.5 require:

• PR to ACCESS-NRI/MOM5 access-esm1.5, review, merge
• Additional PRs to update ACCESS-ESM1.5

Modifications to MOM5 needed in bothACCESS-OM2 and ACCESS-ESM1.5 require:

• PR to mom-ocean/MOM5 master, review, merge
• PR to update ACCESS-NRI/MOM5 master, review, merge
• equivalent PR to ACCESS-NRI/MOM5 access-esm1.5, review, merge (in some cases this could be cherry-picked from master)
• Additional PRs to update ACCESS-OM2 and ACCESS-ESM1.5

The need to open the equivalent PR to ACCESS-NRI/MOM5 access-esm1.5 is time-consuming and error-prone.

So... do we really need an access-esm1.5 branch? MOM5 src is already riddled with compiler preprocessor directives for ACCESS_OM and ACCESS_CM. Could all three models be reconciled into a single codebase or is this too difficult/not worth it at this point?

[dougiesquire]

Also, would ACCESS-ESM1.6 use the access-esm1.5 branch?

[dougiesquire]

An update: MOM5 in ACCESS-ESM1.6 builds from the master branch using the ACCESS_ESM type (this requires these changes)

[CodeGat]

Just putting my hat into the ring to say that I would much prefer not needing an access-esm1.5 branch at all, in any of the model components. It complicates the spack.yaml versioning a bit: https://github.com/ACCESS-NRI/ACCESS-ESM1.5/blob/6ba5764fd9a4937370b24b33f487f0cbb77d8f3b/spack.yaml#L13-L15