Ed Hartnett

OK, thanks for checking. I have undone the attempted fix. I will have to slowly back away from this now...

See also #734, which may be related.

Or we could return error if user does a write of enum that requires a fill value, and no fill value is provided.

OK, I have been pondering and I don't think using the first or last enum value as a fill value is a good idea. It would be seriously misleading. Agree...

Another thing to consider is a HDF5 file that is not a netCDF-4 file which will be using 0 for fill value, which is the HDF5 default (and a bummer...

If we use zero, we will maintain compatibility with all existing cases, and HDF5. I am not crazy about 0 as a fill value, but I think it's less confusing...

Here's what I think we should do: 1 - For netCDF API enforce that fill value for enums is a valid value for that enum. 2 - ncdump should be...

Well the API takes the easy way out. It does not check for validity of enum values at all. When you do an nc_get_vara() for an enum, and provide an...

Ha! Good one. I didn't think of that. OK, we could pass back "_FillValue" or "_UNDEFINED". Which do you prefer?

It's a feature not a bug! ;-) I will make the modifcations. I will use "_FillValue" as the name, since in all cases, that should be the only reason that...