Reduce the number of compilation warnings

[rpseng]

There are numerous compilation warnings during the NWChem build, mostly related to rtdb_get and rtdb_put, as these methods are expected to handle different argument types (int, double, logical, ...).

An example of warning is:
Warning: Type mismatch between actual argument at (1) and actual argument at (2) (REAL(8)/LOGICAL(8))

I’d like to check with you whether reducing these and other types of compilation warnings is something you’re interested in.

My idea would be to create some wrapper methods rtdb_get_int, rtdb_get_log, etc. If that looks ok, I can try to handle that and make a PR.

[jeffhammond]

The issue is the lack of interface definitions. The compiler infers these upon first use and then complains about incompatible usage. Legacy Fortran compiler didn't care about type issues. Modern Fortran compilers do.

If you want to fix this, you should define a module of interfaces for all the types. Don't just create wrappers.

[rpseng]

Thanks for the feedback! My Fortran is pretty rusty. I'll see if I can manage the interface definitions.

[edoapra]

Thanks for the feedback! My Fortran is pretty rusty. I'll see if I can manage the interface definitions.

That would be great.

[rpseng]

Hello, I gave it a try but realized it would likely require making it compiler-specific and adjusting the Makefiles. For now, I'll have to leave it as is.

[edoapra]

Hello, I gave it a try but realized it would likely require making it compiler-specific and adjusting the Makefiles. For now, I'll have to leave it as is.

This is what make V=-1 is doing.

[edoapra]

You can reduce the amount of warnings by typing

make V=-1