Using branches instead of SourceMods

[gold2718]

There are several beta01 tuning experiments which are using Source Mods instead of branches (#163, #165, #166, #167, #202).

Passing around SourceMods makes it harder to document an experiment and to make it reproducible. It can lead to confusion about what is needed to run a case (e.g., #159) and changes during an experiment are not documented in CaseStatus.

It takes less than a minute to create a branch from a modified source tree and these can be referred to directly in an experiment's issue text.

Is there a science reason to continue using SourceMods instead of documented branches? Please discuss it here.

[oyvindseland]

No scientific reasons for my cases. As far as I remember there were some mismatches between branches so I did not see the solution to bring them together

[gold2718]

No scientific reasons for my cases. As far as I remember there were some mismatches between branches so I did not see the solution to bring them together

The simple solution is to make the changes in your source tree instead of SourceMods and commit the changes to a branch. However, in your case (at least the one I looked at), you have modifications to two CAM externals (PUMAS and Oslo Aero). You could make a branch for each, however, a complete solution for documentation is more steps:

1. Create a branch in PUMAS with the modifications (and push to fork)
2. Create a branch in OSLO_AERO with the modifications (and push to fork)
3. Create a branch in CAM which points to the PUMAS and OSLO_AERO branches (and push to fork)
4. Create a branch in NorESM which points to the CAM branch (and push to fork)

I think it would help if we could skip steps 3 and 4 while running test experiments and still have documentation of the code that was built and run. I have some ideas on that. It might also be nice to have a script that would automatically do 3 and 4 and maybe even 1-4 if we decide on what we want there.

[oyvindseland]

I think the technical reason was that I needed to bring in an older version for compability reasons with previous versions and then an updated parameterisation that was not accepted in the main version.

I should probably have made another branch that branched of from a previous tag.