Validating input namelists, field tables, and diag_tables

[nbren12]

Certain combinations of input configurations are not compatible. For example,

• including 3D tendency outputs in the diag_table requires this namelist settinggfs_physics_nml.ldiag3d = true
• Some physics schemes depend on specific field_tables (see e.g. Segmentation fault when running with satedmf fv3net#928)

Often times these misconfigurations only cause the model to crash with a segmentation fault when it tries to access some non-existing allocatable array or pointer, which makes it hard to debug.

Ideally the model would be smart enough to e.g. turn on ldiag3d if one of the diagnostics that flag controls is requested in the diag_table or select a field_table appropriate for it's namelist options, but some input data validation with helpful error messages would go a long way.

Are there any established routines for such checks?

[spencerkclark]

I'm not aware of any established routines for such checks, but it is fairly straightforward to catch these issues before the segmentation faults occur. See #132 and #133 for PRs addressing the two examples given.

These two PRs go the route of adding more helpful error messages. I briefly looked into what it might take do address either automatically, but I think it would require a decent amount of refactoring, which I don't think is necessarily worth doing for this issue.