Format specifier macros for cgsize_t

[jedbrown]

HDF5 has the following definitions, which allow reading and formatting of type hsize_t.

typedef uint64_t hsize_t;
typedef int64_t hssize_t;
#define PRIdHSIZE          PRId64
#define PRIiHSIZE          PRIi64
#define PRIoHSIZE          PRIo64
#define PRIuHSIZE          PRIu64
#define PRIxHSIZE          PRIx64
#define PRIXHSIZE          PRIX64
#define H5_SIZEOF_HSIZE_T  8
#define H5_SIZEOF_HSSIZE_T 8

CGNS has

# if CG_BUILD_64BIT
#  define CG_SIZEOF_SIZE    64
#  define CG_SIZE_DATATYPE "I8"
   typedef CG_LONG_T cgsize_t;
# else
#  define CG_SIZEOF_SIZE    32
#  define CG_SIZE_DATATYPE "I4"
   typedef int cgsize_t;
# endif

Unfortunately, CG_LONG_T can take on different values depend on whether it's on Windows. It's also incompatible with hssize_t because int64_t usually maps to long (e.g., on x86-64) instead of long long. (I'm not saying they should be compatible, but HDF5 changed how they handle these types in HDFGroup/hdf5#709 and I think that rationale also applies to CGNS.)

This is a minor headache for us in PETSc. Would you be amenable to a change similar to the HDF5 change above and adding PRIdCGSIZE (and others if someone wants them)?

Also, I'm assuming for now that 32-bit cgsize_t is important for some users. That makes it harder to package/distribute and requires libraries like ours to test both versions.

[brtnfld]

I don't see any issues with making similar changes; we already require C99.

[brtnfld]

This has been fixed and should be in the next release. Thanks for the report.