can't parse group names that contain a period about f90nml

[marshallward]

Surprised that even works, AFAIK that is not a valid namelist name. From the standard,

• namelist-group-object is a variable-name (5.6, line 9)
• variable-name is a name (6.2, line 20)
• name is letter [alphannumeric-character] ... (3.2.2, line 14)

But it wouldn't be the first time that non-standard stuff has worked.

Supporting this should probably be done with a strictness flag.

[marshallward]

I can't get gfortran or ifort to compile something with the dot, am I interpreting this right?

[aekiss]

Steve sent me the namelist containing that group - it's from a GFDL ESM2M config but that's about all I know. I was only interested in the MOM groups so the offending &diag_inESM2_Control_216.xmltegral_nml is not actually of interest to me. I don't know if they process it somehow before sending to their model - perhaps the diag_inESM2_Control_216. is stripped off by a preprocessor?
So anyway if this is not a valid name and gfortran & ifort don't handle it then it probably isn't worth fixing...? Though if GFDL use f90nml they might encounter this.

[marshallward]

There seems to be a "diag_integral_nml" entries in MOM.. but anyway this is not looking like a f90nml issue.

It might be worth supporting this, I think it's valuable to support preprocessed namelists, though as mentioned before it should be flag-controlled.

[marshallward]

Actually this is starting to look like a typo.. like the word ESM2_Control_216.xml got pasted in the diag_integral_nml name. Suggest we close this?

[aekiss]

D'oh! Yes that's what it is. Sorry to have wasted your time - I'll let Steve know. Happy to close this.