gmtb/develop (--> dtc/develop): static build and out-of-source-build capabilities

[climbfuji]

This PR targets gmtb/develop for code review purposes. Once merged, dtc/develop should be created from gtmtb/develop, made a protected branch and gmtb/develop should be frozen.

This PR:

• updates gmtb-scm to work with the latest changes in ccpp-framework and ccpp-physics from master (out-of-source build capabilities, although these are not used for SCM)
• allows to build SCM with both dynamic and static CCPP
• includes an update for ccpp-physics to prevent mismatches between the ozone physics namelist variables and suite definition file entries (dtc/develop: prevent mismatch of namelist and suite definition file settings for ozone physics ccpp-physics#350); this is not yet in dtc/develop and still needs to be tested with ufs-weather-model
• the ccpp-physics PR also includes the CIRES UGWP namelist fix from @grantfirl (CIRES UGWP namelist fix ccpp-physics#343), which also needs to be tested with the ufs-weather-model
• includes an update for the static build in ccpp-framework, which is already in dtc/develop and has been tested with ufs-weather-model when the dtc/develop branch was created (ccpp_prebuild: validate suite name in static build, fix legacy Fortran code, improve error handling ccpp-framework#240)

This PR does not update the documentation wrt building SCM with the static CCPP option.

No changes are required for buidling SCM with dynamic CCPP, i.e. the following still works:

$ ./ccpp/framework/scripts/ccpp_prebuild.py \
    --config=ccpp/config/ccpp_prebuild_config.py \
    2>&1 | tee ccpp_prebuild.log
$ mkdir -p scm/bin
$ cd scm/bin
$ cmake -DCMAKE_BUILD_TYPE=Debug -DOPENMP=ON ../src 2>&1 | tee log.cmake
$ make VERBOSE=1 2>&1 | tee log.make

To build SCM in static mode and use the multi_run_gmtb_scm.py script, the following commands are required (compiling all available suites into the executable):

$ ./ccpp/framework/scripts/ccpp_prebuild.py \
    --config=ccpp/config/ccpp_prebuild_config.py \
    --static \
    --suites=SCM_GFS_2017_myj,SCM_GFS_v15,SCM_GFS_v15_noahmp,SCM_GFS_v15_prescribed_surface,SCM_GFS_v15plus,SCM_GFS_v15plus_prescribed_surface,SCM_GSD_v0,SCM_GSD_v0_prescribed_surface,SCM_csawmg,SCM_csawmg_prescribed_surface \
    2>&1 | tee ccpp_prebuild.log
$ mkdir -p scm/bin
$ cd scm/bin
$ cmake -DCMAKE_BUILD_TYPE=Debug -DOPENMP=ON -DSTATIC=ON ../src 2>&1 | tee log.cmake
$ make VERBOSE=1 2>&1 | tee log.make

With both -DCMAKE_BUILD_TYPE=Debug and -DCMAKE_BUILD_TYPE=Release, the results are identical between the dynamic and static CCPP builds:

$ pwd
/Users/dom.heinzeller/scratch/gmtb-scm/gmtb-scm_develop_static-build/static/scm/bin
$ for dir in output_*; do echo $dir; cmp /Users/dom.heinzeller/scratch/gmtb-scm/gmtb-scm_develop_static-build/dynamic/scm/bin/$dir/output.nc $dir/output.nc; done
output_LASSO_2016051812_SCM_GFS_v15
output_LASSO_2016051812_SCM_GFS_v15plus
output_LASSO_2016051812_SCM_GSD_v0
output_LASSO_2016051812_SCM_csawmg
output_arm_sgp_summer_1997_A_SCM_GFS_v15
output_arm_sgp_summer_1997_A_SCM_GFS_v15plus
output_arm_sgp_summer_1997_A_SCM_GSD_v0
output_arm_sgp_summer_1997_A_SCM_csawmg
output_astex_SCM_GFS_v15
output_astex_SCM_GFS_v15plus
output_astex_SCM_GSD_v0
output_astex_SCM_csawmg
output_bomex_SCM_GFS_v15
output_bomex_SCM_GFS_v15plus
output_bomex_SCM_GSD_v0
output_bomex_SCM_csawmg
output_twpice_SCM_GFS_v15
output_twpice_SCM_GFS_v15plus
output_twpice_SCM_GSD_v0
output_twpice_SCM_csawmg

[climbfuji]

Note: because the two PRs for ccpp-physics, NCAR/ccpp-physics#350 and NCAR/ccpp-physics#343, still need to be tested with ufs-weather-model, this PR cannot be merged until the testing has been completed.

[grantfirl]

Thanks @climbfuji for making modifications to work with the static build. I started to look at this last week and was a little lost, so thanks for doing this. I tested this on my machine and it works as advertised once a small fix for ccpp_prebuild_config.py was included.

As per the instructions, I'm waiting to merge this until NCAR/ccpp-physics#350 is tested in ufs-weather-model, which is currently ongoing on Hera. If tests pass, I'll merge this PR and create the dtc/develop branch from it.

[grantfirl]

@climbfuji When I started looking at the static build last week, I played around with calling ccpp_prebuild.py from cmake. With the lack of a more complicated build system as in the UFS, would it make sense for us to cut out the manual step of calling ccpp_prebuild by making cmake do it?

[grantfirl]

ufs-weather-model RT passes on hera/intel with rt.conf for NCAR/ccpp-physics#350 so, I'm going to merge this and create dtc/develop from it.

[climbfuji]

@climbfuji When I started looking at the static build last week, I played around with calling ccpp_prebuild.py from cmake. With the lack of a more complicated build system as in the UFS, would it make sense for us to cut out the manual step of calling ccpp_prebuild by making cmake do it?

Hi @grantfirl, the problem I had was that cmake wouldn't allow me to execute a script before anything else happens in cmake - it would allow me to run it pre_build (but after cmake). I believe Dusan found a solution for this for UFS - can you touch base with him? Thanks!