About the Segmentation fault (core dumped) when converting netcdf to grib2

When I conduct

cdo -f grb2 copy NC2020121401.nc MP2020121401.grb2

Some parameter name cannot be read so that there is reported error like:

ECCODES ERROR : concept: no match for cfName=fraction_of_land ECCODES ERROR : concept: input handle edition=2 !!! grib_set_string( grib_handle* h, "cfName", "fraction_of_land") !!! Warning (gribapiDefParam): grib_api: No match for cfName=fraction_of_land ECCODES ERROR : concept: no match for shortName=FRLAND ECCODES ERROR : concept: input handle edition=2 !!! grib_set_string( grib_handle* h, "shortName", "FRLAND") !!! Warning (gribapiDefParam): grib_api: No match for shortName=FRLAND

So I try local configuration for the parameter name and creat the dir “merra_yang” in

/usr/local/eccodes/share/eccodes/definitions/grib2/localConcepts/merra_yang

The configuration file goes like(e.g. cfName.def)

> # land-sea mask 'fraction_of_land' = { discipline = 2; parameterCategory = 0; paramterNumber = 0; typeOfFirstFixedSurface = 1; }
and similar in shortName.def, paramId.def, units.def and name.def

The I set the definition path as

export ECCODES_DEFINITION_PATH=/usr/local/eccodes/share/eccodes/definitions/grib2/localConcepts/merra_yang:/usr/local/eccodes/share/eccodes/definitions

The retry the cdo process but the output became

cdo copy/selall : UNCHANGED_RECORD=0
cdo copy/selall : cdiGribDataScanningMode=0; lcopy=0
Segmentation fault (core dumped)

If unset the ECCODES_DEFINITION_PATH the output would be back to the beginning and the parameter names were still unreadable.

Would anyone offer some help on it?

Sincerely
Tianliang Yang