Confict of size of INTEGER with efc on TX

[Dave Offiler]

Despite the configure for netCDF using efc -i4 (32-bit intergers - the default) compiler switch, ropp_io code using it (notably the ncdf interface, like is_netcdf() ) complains of incorrect arguments. This can be cured by compiling (say) is_netcdf.f90 with the -i8 (64-bit integers) option, which suggests netCDF is compiled effectively with -i8 regardless.

BUT flib.f90 won't compile usder -i8 as (some) routines don't support this data type.

Inpection of flib.f90 shows 4-5 subroutines dealing with low-level i/o and having overloading for I4, R4, R8 types. One even has an I8. It would be a simple matter to add to the routines with missing I8 support. The problem occurs in the licencing as to whther this code can be modifed and if it can, can it be distributed?

Purely as a test out of curiosity, I've edited flib.f90 as sugested, and efc -i8 will compile it. Further, the whole of ropp_tools and ropp_io can now be built (but not tested if the tools actually run - I don't know if or where flib might be used.

This version of flib has not been committed, and will not be included in the distro unless the licence clearly allows it. At worst we should refer the mods back to the authors.

[Dave Offiler]

Gone ahead and committed the change anyway. Re-open if it becomes an issue code- or licence-wise.