Use fortran compilers for preprocessing instead of cpp

[anton-seaice]

First I run some experiments and then change is finalised, see

#83 (comment)

[anton-seaice]

 ==> [2025-12-19-03:32:44.628171] No patches needed for cice5
==> [2025-12-19-03:32:44.650460] cice5: Executing phase: 'set_deps_targets'
==> [2025-12-19-03:32:44.713073] cice5: Executing phase: 'edit'
==> [2025-12-19-03:32:44.754735] cice5: Executing phase: 'build'
==> [2025-12-19-03:32:45.975931] Error: ProcessError: Command exited with status 2:
    '/tmp/root/spack-stage/spack-stage-cice5-git.0126c7ac338f3c5d05519764cecf925cbc5d1422_stable-obk6sniut3dcoxtoh5u5dkimpmbl5ifj/spack-src/bld/spack-build.sh' 'auscom' '360x300' '24x1' '24'

3 errors found in build log:
     260    Building dependency for sat_vapor_pres_mod.d
     261    Building dependency for shr_orb_mod.d
     262    Building dependency for surface_flux_mod.d
     263    Building dependency for version_mod.d
     264    mpifort -P -traditional -DLINUX -DPAROPT -DNXGLOB=360 -DNYGLOB=300 
                        -DNUMIN=11 -DNUMAX=99             -DTRAGE=1 -DTRFY=1 -D
            TRLVL=1             -DTRPND=1 -DNTRAERO=0 -DTRBRI=1             -DN
            BGCLYR=0 -DTRBGCS=2             -DNICECAT=5 -DNICELYR=4            
             -DNSNWLYR=1             -DBLCKX=15 -DBLCKY=300 -DMXBLCKS=1 -Dgathe
            r_scatter_barrier -Dncdf -DPIO -DAusCOM -Dcoupled -DOASIS3_MCT -I. 
            -I/tmp/root/spack-stage/spack-stage-cice5-git.0126c7ac338f3c5d05519
            764cecf925cbc5d1422_stable-obk6sniut3dcoxtoh5u5dkimpmbl5ifj/spack-s
            rc/drivers/auscom -I/tmp/root/spack-stage/spack-stage-cice5-git.012
            6c7ac338f3c5d05519764cecf925cbc5d1422_stable-obk6sniut3dcoxtoh5u5dk
            impmbl5ifj/spack-src/source -I/tmp/root/spack-stage/spack-stage-cic
            e5-git.0126c7ac338f3c5d05519764cecf925cbc5d1422_stable-obk6sniut3dc
            oxtoh5u5dkimpmbl5ifj/spack-src/mpi -I/tmp/root/spack-stage/spack-st
            age-cice5-git.0126c7ac338f3c5d05519764cecf925cbc5d1422_stable-obk6s
            niut3dcoxtoh5u5dkimpmbl5ifj/spack-src/io_pio -I/tmp/root/spack-stag
            e/spack-stage-cice5-git.0126c7ac338f3c5d05519764cecf925cbc5d1422_st
            able-obk6sniut3dcoxtoh5u5dkimpmbl5ifj/spack-src/csm_share -I. -I/op
            t/release/linux-rocky8-x86_64/gcc-13.2.0/oasis3-mct-stable-azws3fnu
            qlfgq2p5v2d64wdvur5zobdt/include/psmile.MPI1 -I/opt/release/linux-r
            ocky8-x86_64/gcc-13.2.0/parallelio-2.6.2-gkrdy2kdydaql3x634ew4namno
            kq267p/include -I/opt/release/linux-rocky8-x86_64/gcc-13.2.0/oasis3
            -mct-stable-azws3fnuqlfgq2p5v2d64wdvur5zobdt/include -I/opt/release
            /linux-rocky8-x86_64/gcc-13.2.0/libaccessom2-access-om2-jw7h6mxv362
            txj54a6qg6x5347ab2yah/include -I/opt/release/linux-rocky8-x86_64/gc
            c-13.2.0/netcdf-fortran-4.6.1-ff7gembhstandrvuicgw44lbaukzl44y/incl
            ude /tmp/root/spack-stage/spack-stage-cice5-git.0126c7ac338f3c5d055
            19764cecf925cbc5d1422_stable-obk6sniut3dcoxtoh5u5dkimpmbl5ifj/spack
            -src/source/ice_kinds_mod.F90 -o ice_kinds_mod.f90
     265    /lib/../lib64/crt1.o: In function `_start':
  >> 266    (.text+0x24): undefined reference to `main'
  >> 267    collect2: error: ld returned 1 exit status
  >> 268    make: *** [/tmp/root/spack-stage/spack-stage-cice5-git.0126c7ac338f
            3c5d05519764cecf925cbc5d1422_stable-obk6sniut3dcoxtoh5u5dkimpmbl5if
            j/spack-src/bld/Makefile:120: ice_kinds_mod.o] Error 1
     269    exit 2

See build log for details:
  /tmp/root/spack-stage/spack-stage-cice5-git.0126c7ac338f3c5d05519764cecf925cbc5d1422_stable-obk6sniut3dcoxtoh5u5dkimpmbl5ifj/spack-build-out.txt

[anton-seaice]

@manodeep - do you have any ideas why this change fails for gcc and not for intel?

The -P flag is there, this is failing trying to do preprocessing

This doesn't impact when using CMake, maybe I should just not worry about the Makefile build ?

[anton-seaice]

-P

Inhibit generation of linemarkers in the output from the preprocessor. This might be useful when running the preprocessor on something that is not C code, and is sent to a program that might be confused by the linemarkers.

[anton-seaice]

https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html

[manodeep]

Yes - the -P flag should do the trick of pre-processing - not sure why it fails. May be a -c is needed?

Can you manually run that line and see what it does?

[anton-seaice]

hmm

[as2285@gadi-login-02 source]$ gfortran ice_kinds_mod.F90 
/half-root/usr/bin/../lib/gcc/x86_64-redhat-linux/8/../../../../lib64/crt1.o: In function `_start':
(.text+0x24): undefined reference to `main'
collect2: error: ld returned 1 exit status
[as2285@gadi-login-02 source]$ gfortran -P ice_kinds_mod.F90 
/half-root/usr/bin/../lib/gcc/x86_64-redhat-linux/8/../../../../lib64/crt1.o: In function `_start':
(.text+0x24): undefined reference to `main'
collect2: error: ld returned 1 exit status
[as2285@gadi-login-02 source]$ gfortran -cpp ice_kinds_mod.F90 
/half-root/usr/bin/../lib/gcc/x86_64-redhat-linux/8/../../../../lib64/crt1.o: In function `_start':
(.text+0x24): undefined reference to `main'
collect2: error: ld returned 1 exit status
[as2285@gadi-login-02 source]$ gfortran -P -traditional ice_kinds_mod.F90 
/half-root/usr/bin/../lib/gcc/x86_64-redhat-linux/8/../../../../lib64/crt1.o: In function `_start':
(.text+0x24): undefined reference to `main'
collect2: error: ld returned 1 exit status
[as2285@gadi-login-02 source]$ which gcc
/opt/nci/bin/gcc

[anton-seaice]

gfortran -E -P ice_kinds_mod.F90 -o ice_kinds_mod.f90

works

however

ifx -E -P ice_kinds_mod.F90 -o ice_kinds_mod.f90

does not, it writes output to the terminal (See IFX manual)

[anton-seaice]

This changes closes #51 by setting different arguments to conduct preprocessing for intel and gfortran compilers.

intel needs "-fpp -P" whilst gfortran needs "-cpp -E"

Using the intel compiler for preprocessing defines some compiler specific directives (e.g. __INTEL_COMPILER__) which are not available if preprocessing using cpp directely.

[dougiesquire]

LGTM. Does the compiler detection also need to be added to the access-esm1.6 branch?

[anton-seaice]

LGTM. Does the compiler detection also need to be added to the access-esm1.6 branch?

Planning to fix this by merging the branches in #68 :)