makedep: Include object file dependencies

[marshallward]

The current implementation of makedep contains something like a race condition, where the creation of the .mod and .o files may be in an order which breaks the current dependency tree. Currently, .mod depends on .o, and changes to a module do not trigger rebuilds of dependent source.

Rather than try to sort out the rule order, which could even depend on compiler internals, this patch just adds both object and module output files as dependencies, and rebuilds if either changes.

We might want to come back to this someday and understand the actual order of rule execution.

Thanks to Alistair Adcroft (@adcroft) for proposing this solution.

[codecov]

Codecov Report

Merging #288 (e713bb9) into dev/gfdl (40c24c5) will increase coverage by 0.04%.
The diff coverage is n/a.

❗ Current head e713bb9 differs from pull request most recent head f6db061. Consider uploading reports for the commit f6db061 to get more accurate results

@@             Coverage Diff              @@
##           dev/gfdl     #288      +/-   ##
============================================
+ Coverage     37.07%   37.11%   +0.04%     
============================================
  Files           263      263              
  Lines         73623    73493     -130     
  Branches      13720    13700      -20     
============================================
- Hits          27296    27278      -18     
+ Misses        41290    41185     -105     
+ Partials       5037     5030       -7     

Impacted Files	Coverage Δ
src/framework/MOM_document.F90	74.10% <0.00%> (-0.23%)	⬇️
...ig_src/drivers/solo_driver/MOM_surface_forcing.F90	24.42% <0.00%> (-0.11%)	⬇️
...eterizations/lateral/MOM_lateral_mixing_coeffs.F90	42.49% <0.00%> (-0.09%)	⬇️
src/tracer/MOM_tracer_hor_diff.F90	76.48% <0.00%> (-0.07%)	⬇️
src/user/Kelvin_initialization.F90	0.00% <0.00%> (ø)
src/user/dumbbell_initialization.F90	0.00% <0.00%> (ø)
src/user/tidal_bay_initialization.F90	0.00% <0.00%> (ø)
...rc/parameterizations/vertical/MOM_tidal_mixing.F90	1.40% <0.00%> (ø)
src/user/MOM_wave_interface.F90	1.32% <0.00%> (+0.01%)	⬆️
src/parameterizations/vertical/MOM_CVMix_KPP.F90	0.80% <0.00%> (+0.03%)	⬆️
... and 6 more

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[adcroft]

I agree with the change to the normal makefile output, but when debugging is turned on (makedep -d ...) I think it better to report both found_mods and found_deps. Could you add that debugging?

[marshallward]

found_mods and found_objs? or do you also want found_deps?

[adcroft]

I thought found_obs would be very long but I see not, so maybe that too? I just want to make sure there's enough information to understand what led to the local rule.

[marshallward]

I just did found_mods and found_objs since found_deps is just the combination of the two

[marshallward]

Not quite on-topic, but I noticed that Cray module files are in ALL CAPS (MOM.mod) so perhaps including the .o files is essential, just to preserve some overall functionality of the Makefile.

(Of course a better solution is to take account of the filename,but that's a separate issue...)

[adcroft]

Looks good.

[Hallberg-NOAA]

This PR has passed pipeline testing at https://gitlab.gfdl.noaa.gov/ogrp/MOM6/-/pipelines/17825.