Fix MPAS standalone compile issues

[mark-petersen]

These two small changes allow MPAS components to compile in stand-alone mode. It should not affect E3SM compile or simulations. The fixes are:

1. A missing endif was added to the Makefile
2. A redundant corelist declaration was removed from mpas.F because it was added to mpas_subdriver.F in PR Introduce MOAB coupler interfaces and update moab driver #5979, and the subdriver is used by mpas.F.

These fixes should work for ocean, sea ice, and MALI cores.

Fixes #6133
Fixes #6138
[BFB]

[mark-petersen]

With these changes, MPAS-Ocean standalone compiles with gnu debug on chrysalis and chicoma and passes the nightly test suite on both. I can successfully compile mpas-seaice with gnu debug. I was not able to compile MALI, but I'm not familiar with it, so probably did it wrong.

[mark-petersen]

This PR also compiles MPAS-Ocean with intel debug on chrysalis and passes the nightly test suite. MALI compiles successfully with intel on chrysalis but not gnu.

[xylar]

I'm also not able to build MALI on Chrysalis with gnu. I'm using a compass load script with albany support. When I run:

make ALBANY=true gfortran

I see:

gcc -D_MPI -DUSE_PIO2 -DHAVE_ADIOS -DCORE_LANDICE -DUSE_EXTERNAL_FIRSTORDER -DMPAS_LI_BUILD_INTERFACE -DMPAS_NAMELIST_SUFFIX=landice -DMPAS_EXE_NAME=landice_model -DMPAS_OPENMP -DMPAS_NATIVE_TIMERS -DMPAS_GIT_VERSION=v2.1.0-11503-g1496cabcea -O3 -m64 -fopenmp -I../../external/ezxml -o namelist_gen namelist_gen.o test_functions.o ../../external/ezxml/ezxml.o
namelist_gen.o: In function `main':
namelist_gen.c:(.text+0x27): undefined reference to `__intel_new_feature_proc_init'
namelist_gen.c:(.text+0x44): undefined reference to `__kmpc_begin'
namelist_gen.c:(.text+0x104): undefined reference to `__intel_sse2_strdup'
namelist_gen.c:(.text+0x356): undefined reference to `__intel_sse2_strdup'
namelist_gen.c:(.text+0x42c): undefined reference to `__kmpc_end'
namelist_gen.c:(.text+0x4bd): undefined reference to `__kmpc_end'
namelist_gen.c:(.text+0x4ec): undefined reference to `__intel_sse2_strdup'
namelist_gen.c:(.text+0x537): undefined reference to `__kmpc_end'
namelist_gen.c:(.text+0x579): undefined reference to `__kmpc_end'
namelist_gen.o: In function `generate_namelist':
namelist_gen.c:(.text+0x757): undefined reference to `__intel_sse2_strdup'
namelist_gen.c:(.text+0x868): undefined reference to `__intel_sse2_strdup'
namelist_gen.o: In function `get_option_value':
namelist_gen.c:(.text+0x91c): undefined reference to `__intel_sse2_strdup'
namelist_gen.c:(.text+0x94d): undefined reference to `__intel_sse2_strdup'
test_functions.o: In function `is_structure_writable':
test_functions.c:(.text+0x64): undefined reference to `__intel_sse2_strdup'
collect2: error: ld returned 1 exit status

[xylar]

I tested the compass pr suite on Chrysalis with Intel and OpenMPI using this branch, and all tests ran successfully.

[mark-petersen]

I confirmed that E3SM builds are unaffected. This passes

./create_test SMS_Ld62.T62_oQU240.GMPAS-IAF.chrysalis_gnu -q acme-small --walltime 1:00:00

[darincomeau]

Passes the polaris mpas-seaice single_column test suite.

[hyungyukang]

FYI, I was able to compile the standalone MPAS-O with gnu on Perlmutter and my Mac. Some standalone test cases were successfully terminated.

[mark-petersen]

I checked LISIO, since that one tests MALI within E3SM. It passes:

./create_test SMS_D.T62_oQU120_ais20.MPAS_LISIO_TEST.chrysalis_gnu -q acme-small --walltime 30:00

[matthewhoffman]

I successfully compiling MALI on Anvil with gnu and mvapich and without Albany. I'm getting a compass error trying to set up a fresh compass env on Perlmutter, but using an existing 1.2.0-alpha.8_pm-cpu_gnu_mpich_albany env, I am able to successfully compile MALI with Albany on Perlmutter using this branch.

@mark-petersen , @xylar , should we be worried about the MALI compile problems you are both seeing?

[mark-petersen]

should we be worried about the MALI compile problems you are both seeing?

@matthewhoffman I tested commit 09491ec, which was the last commit working before stand-alone compile issues were introduced. The MALI problem existed at that point, i.e. this command on chrysalis

make ALBANY=true gfortran

fails in exactly the same way as it does with this PR. So the issue seems to be unrelated. Since you were able to compile on Anvil, and we compiled MALI on chrysalis with intel, I think we should proceed with the fixes in this PR.

[matthewhoffman]

Based on @mark-petersen's information, I'm approving this based on the MALI compilation tests that pass and that the ones that don't are a problem pre-existing this PR.

[jonbob]

Passes:

• SMS_Ld1.ne30pg2_EC30to60E2r2.WCYCL1850.chrysalis_intel.allactive-wcprod_1850
• ERS_Ld5.T62_oQU120.CMPASO-NYF.chrysalis_intel
• SMS.T62_oQU120_ais20.MPAS_LISIO_TEST.chrysalis_intel

merged to next

[jonbob]

merged to master