Hello guys!

I'm using Revision: 3321 (nemo_v3_4) and it seems that:

• In contrast to the documentation I can link NEMO to !NetCdf4 libs and produce netcdf classic files without using key_netcdf4
• When I link to !NetCdf4 (4.1.3 or 4.2) and define key_netcdf4 when running opa I get only the restart file as hdf(nc4) type but the U,V,T and W files are still nc3 (classic).

Another problem (wich is probably for separated ticket):

• while running NEMO it doesn't "flush" the written record so unless it finishes and save the file properly the first several bytes of the nc files give number of record to 0 as described here: ​http://www.unidata.ucar.edu/software/netcdf/docs/faq.html#recover

Here is my test configuration:

makenemo -r GYRE -n TEST2_nc4.2_hdf5 -m ifort_nc4.2_debug -j 6 add_key "key_netcdf4"

and my ifort_nc4.2_debug file:

%SZIP /usr/opt/szip/intel/12.1.1.256/2.1       

%ZLIB /usr/opt/zlib/intel/12.1.1.256/1.2.6
%HDF5 /usr/opt/hdf5/intel/12.1.1.256/1.8.8_sz2.1_z1.2.6       %NETCDF_C /usr/opt/netcdf/intel/12.1.1.256/4.2_hdf5-1.8.8_sz2.1_z1.2.6 
%NETCDF_F /usr/opt/netcdf-fortran/intel/12.1.1.256/4.2_netcdf-4.2_hdf5-1.8.8_sz2.1_z1.2.6                                             
%NCDF_INC            -I%SZIP/include -I%ZLIB/include -I%HDF5/include -I%NETCDF_C/include -I%NETCDF_F/include
        
%NCDF_LIB            -L%NETCDF_C/lib -L%NETCDF_F/lib -L%HDF5/lib -L%ZLIB/lib -L%SZIP/lib  -Wl,--start-group -lnetcdff -lnetcdf -lhdf5 -lhdf5_fortran -lhdf5_hl -lz -lsz -Wl,--end-group                                       
%FC                  ifort                                       
%FCFLAGS             -r8 -g -traceback                           
%FFLAGS              %FCFLAGS                                    
%LD                  ifort                                       
%FPPFLAGS            -P -C -traditional                          
%LDFLAGS             -static                                       
%AR                  ar                                          
%ARFLAGS             -r                                          
%MK                  make                                        
%USER_INC            %NCDF_INC
%USER_LIB            %NCDF_LIB 

The s-coord code has a bug relating to domain decomposition. This limits the way the domain can be decomposed. The issue occurs when Processor interface pass through areas where the envelope bathy is employed near the boundary.

It looks like the code may be incorrectly undoing lbc links in certain cases.

When running an ORCA2_LIM3 reference configuration on an SGI-ALITIX-ICE system the job can fail with a claim to have exceeded the maximum permitted number of MPI_GROUPS. Raising the MPI_GROUP_MAX environmental parameter only delays this happening. The problem can be traced to the mpp_ini_ice routine which creates groups every call but never frees them. Possibly not a problem on all systems but the groups should be free'd after they've been used to create the new communicator (which is itself eventually free'd in limthd). The required changes will be made in lib_mpp.F90 on the trunk.