convert_mpas warning: IEEE_INVALID_FLAG

[cgrudz]

Dear MPAS forum,

I have built the convert_mpas utility in a minimal Ubuntu containerized environment for portability across machines in post-processing data. This container builds Zlib / HDF5 / NetCDF-C / NetCDF-F libs from scratch against gcc / gfortran. The convert_mpas utility appears to run successfully from the singularity container based on preliminary testing, but I have encountered a new warning that I did not see earlier when building the convert_mpas utility against system modules / Intel compilers on the SDSC Expanse super computer, with an example output copied below:
```
Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG
Reading MPAS mesh information from file '/in_dir/x20.835586.WestCoast.history.2022-12-23_00_00_00.nc'

Reading target domain specification from file 'target_domain'
Setting nlat = 750
Setting nlon = 1375
Setting startlat = 4.96000004
Setting startlon = 162.000000
Setting endlat = 65.0000000
Setting endlon = 272.000000

Computing remapping weights
Time to compute remap weights: 6.365722 s
Created a new output file

Reading list of fields to include in remapped output
Found 2 lines in 'include_fields' file
List of fields to be excluded from output ('exclude_fields') not found
Remapping MPAS fields from file '/in_dir/x20.835586.WestCoast.history.2022-12-23_00_00_00.nc'
Input file has 1 records

Defining fields in output file
Remapping field rainnc, frame 1
read: 0.006000 s
remap: 0.006357 s
write: 0.000599 s
Remapping field rainc, frame 1
read: 0.005438 s
remap: 0.006001 s
write: 0.000757 s

Total runtime: 6.711883
```
I wanted to see if the `IEEE_INVALID_FLAG` should be cause for concern -- preliminary results using the GNU / containerized convert_mpas look identical to the Intel / system module version, but I've only made a cursory glance. Thanks for your consideration.

Cheers,
Colin

 

[cgrudz]

Following up on this, it appears that I get the `IEEE_INVALID_FLAG` warning when I compile this from scratch outside of a container too. That is to say, I've loosely followed the definition file recipe to install zlib / HDF5 / NetCDF-C / NetCDF-F libraries from scratch on the SDSC Expanse supercomputer using just the basic gcc module load for this system. The checks pass for these dependences and I am able to successfully build the convert_mpas executable on top of these, and I continue to get warning message below
```
Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG
```
The results for regridded precip fields are all point-wise identical between the system libraries Intel build / containerized build with gcc / build from scratch with system gcc, but it is only the gcc builds that throw this warning. Is this any cause for concern?

Thanks,
Colin