REGISTER_UNLIMITED_COMPRESSED_AXIS does not exist or is not accessible with FMS2 compiler

[Zeracesharon]

when I need FMS2 to do the data override for your input files, the compiling failed. When FMS2 compiler is used for "-DUSE_FMS2_IO", the icebergs_fms2io.F90 needs register_unlimited_compressed_axis while in the FMS2 source code, the declaration and subroutine for this is missing, make the compiling failed.
The following is my solution (I can of course make a pull request for this only if necessary):
You will need to add src/FMS2/fms2_io/fms2_io.F90 at around line 55:
public :: register_unlimited_compressed_axis

And add src/FMS2/fms2_io/netcdf_io.F90:
public :: register_unlimited_compressed_axis
subroutine register_unlimited_compressed_axis(fileobj, dimension_name, dimension_length)
class(FmsNetcdfFile_t), intent(inout) :: fileobj !< File object.
character(len=*), intent(in) :: dimension_name !< Dimension name.
integer, intent(in) :: dimension_length !< Dimension length for the current rank

integer, dimension(:), allocatable :: npes_start !< The starting index of the dimension for each of the PEs
integer, dimension(:), allocatable :: npes_count !< The size of the dimension for each of the PEs
integer :: i !< For do loops
integer :: err !< Netcdf error
integer :: dimid !< Netcdf id for the dimension

!Gather all local dimension lengths on the I/O root pe.
allocate(npes_start(size(fileobj%pelist)))
allocate(npes_count(size(fileobj%pelist)))

call mpp_gather((/dimension_length/),npes_count,pelist=fileobj%pelist)

npes_start(1) = 1
do i = 1, size(fileobj%pelist)-1
npes_start(i+1) = npes_start(i) + npes_count(i)
enddo
call append_compressed_dimension(fileobj, dimension_name, npes_start, &
npes_count)

if (fileobj%is_root .and. .not. fileobj%is_readonly) then
call set_netcdf_mode(fileobj%ncid, define_mode)
err = nf90_def_dim(fileobj%ncid, trim(dimension_name), unlimited, dimid)
call check_netcdf_code(err, "Netcdf_add_dimension: file:"//trim(fileobj%path)//" dimension name:"// &
& trim(dimension_name))
endif
end subroutine register_unlimited_compressed_axis

After making this changes, recompile FMS, with FMS linked to FMS2 instead of FMS1.
This way, it will allow you to compile for the ocean_SIS2 with FMS2.

[marshallward]

icebergs also requires a preprocessor flag (-DUSE_FMS2_IO). Have you tried that?

Also, the Makefile in ice_ocean_SIS2/ should be preconfigured to use FMS2 but perhaps there could be some issues on your end.

We currently have to retain the FMS1 default for testing reasons. But we have discussed some options to transition most of the tests to FMS2.

[Zeracesharon]

Thanks a lot for replying. My compiling file for FMS is:
#!/bin/csh

set version=REPRO
set platform="ncrc5.intel23"
set rootdir = dirname $0
source ${rootdir}/envs.${platform}

mkdir -p build/fms/

(cd build/fms/; rm -f path_names;
../../src/mkmf/bin/list_paths -l ../../src/FMS;
../../src/mkmf/bin/mkmf -t ../../src/mkmf/templates/ncrc-intel.mk -p libfms.a -c "-DUSE_FMS2_IO -Duse_libMPI -Duse_netCDF -DHAVE_GETTID" path_names)
(cd build/fms/; make NETCDF=3 $version=1 libfms.a -j)
For ocean_ice is:
#!/bin/csh
set platform="ncrc5.intel23"
set rootdir = dirname $0
source ${rootdir}/envs.${platform}
mkdir -p build/ice_ocean_SIS2/
(cd build/ice_ocean_SIS2/; rm -f path_names;
../../src/mkmf/bin/list_paths -l ./ ../../src/MOM6/config_src/{infra/FMS2,memory/dynamic_symmetric,drivers/FMS_cap,external} ../../src/SIS2/config_src/dynamic_symmetric ../../src/MOM6/src/{,/*}/ ../../src/{atmos_null,coupler,land_null,ice_param,icebergs/src,SIS2,FMS/coupler,FMS/include}/)
(cd build/ice_ocean_SIS2/;
../../src/mkmf/bin/mkmf -t ../../src/mkmf/templates/ncrc-intel.mk -o '-I../fms' -p MOM6 -l '-L../fms -lfms' -c '-DUSE_FMS2_IO -Duse_AM3_physics -D_USE_LEGACY_LAND_' path_names )
(cd build/ice_ocean_SIS2/; make NETCDF=3 REPRO=1 MOM6 -j)
I have also use the soft link in src/ link the FMS to FMS2
i believe I have added the compiling flag that you mentioned about. If you go to look at the icebergs source code, you will realized it uses fms2_io mode with register_unlimited_compressed_axis where register_unlimited_compressed_axis is never found in FMS2 source code IO source file. I have tried to use FMS1 to read the my input files but it fails the run, not support the data type.
But If I try the method above by adding the function, declaration to the corresponding files, and then compiling it, it will solve the problem.

I hope this can help anyone that tries to use FMS2 to do the data override with ocean_ice.

[marshallward]

I think you've shown that our FMS2 source tree is more out of date than we had thought. We've known for a while that FMS2 and icebergs are not being properly updated, but this highlights the problem.

In practice, people using icebergs and FMS2 tend to manually set the tag of these repos, if not outright copy them into new projects (e.g OM5), so most of these problems in MOM6-examples go untested and unreported.

The simplest solution may be to update your FMS2 repo to main or a recent tagged release, which includes register_unlimited_compressed_axis, although I can't be sure which version will work. 🤞

I will try to see if this can be fixed on our end. Although not used yet in production, it looks like the Makefile of ice_ocean_SIS2 mentioned above may have been building icebergs with FMS1 without us realizing it.

[Zeracesharon]

Dear Marshall,
thanks a lot for the reply. The FMS2 source code this MOM6-example linked to 0721d59 ,
but this is included in the main repo of https://github.com/NOAA-GFDL/FMS/blob/main/fms2_io/.
Just to report this issue so that someone meets the same problem with me knows the solution.