Changes for ESM1.5 historical minor release 1.2

[blimlim]

This PR brings the changes from the historical+concentrations development branch into the release branch, to setup the release of version 1.2.

The main changes are:

• Historical: add basin mask for basin specific diagnostics  #102
  Adding the basin_mask.nc input, allowing for correct calculation of basin-specific ocean diagnostics
• Add spinup diagnostic presets to historical+concentrations configuration #106
  Adding spin-up diagnostic presets which produce less output
• Add path to fixed STASHmaster #108
  Using an updated STASHmaster file which fixes a bug in the saved variable 2,290 TOTAL DOWNWARD PAR FLUX AT SURFACE
• Historical - increase atmosphere processors #110
  Increasing the number of processors assigned to the atmosphere, speeding up the model
• Historical: update model version to 2024.12.0 #116
  Updating the model version, allowing for runtime orbital parameters

[blimlim]

!test repro

[github-actions]

✅ The Bitwise Reproducibility Check Succeeded ✅

When comparing:

• dev-historical+concentrations (checksums created using commit ab62130), against
• release-historical+concentrations (checksums in commit 4356563)

Further information

The experiment can be found on Gadi at /scratch/tm70/repro-ci/experiments/access-esm1.5-configs/ab621309daec29bbff8db628630edba5dea1f980, and the test results at https://github.com/ACCESS-NRI/access-esm1.5-configs/runs/36556621929.

The checksums generated by this !test command are found in the testing/checksum directory of https://github.com/ACCESS-NRI/access-esm1.5-configs/actions/runs/13104215337/artifacts/2524297532.

The checksums compared against are found here https://github.com/ACCESS-NRI/access-esm1.5-configs/tree/43565639dd6e698463d023437e015a5d52cbaeac/testing/checksum

[aidanheerdegen]

The check is still listed as failing despite the !test repro passing. Is this a GH actions bug/artefact?

[CodeGat]

Hey @aidanheerdegen, we don't yet update the statuses of the commit used in the !test command - see ACCESS-NRI/model-config-tests#88.
But even when that functionality is implemented, a failing status overrides any potential passing statuses added. The hierarchy is pending > failure > success. There is also no way to remove statuses (for example, initial failing ones) - not even in the GitHub API. Unfortunately, if there is a failing status in a commit, it is there to stay. Luckily, the failing checks are not required, so it is still mergable.

[aidanheerdegen]

Thanks for the clarification @CodeGat.

Is this a failure mode that would persist if another commit were pushed to the PR? i.e. is it something to do with the CI itself?

If not, then in this situation would a no-op commit pushed to trigger a new round of checks would pass and be unambiguously ok?

[blimlim]

no-op commit pushed to trigger a new round of checks would pass and be unambiguously ok?

Should I try this out? Because this PR is merging from the dev to release branch, I'd just need to add a separate PR into the dev branch containing an empty commit

[aidanheerdegen]

I'd just need to add a separate PR into the dev branch containing an empty commit

Ah, well that does seem a bit OTT. Don't bother in this case. I'll approve.

[CodeGat]

@aidanheerdegen,

in this situation would a no-op commit pushed to trigger a new round of checks would pass and be unambiguously ok?

that is correct

[aidanheerdegen]

I was planning on using the Release Build DB to check the exe.yaml manifest is correct, but I can't find access-esm1p5/2024.12.0.

Is this a known issue @CodeGat?

[aidanheerdegen]

LGTM, just a clarifying question.

[blimlim]

Thanks @aidanheerdegen for the review! Merging now