Make note of build procedure regarding moab_source.F90

[makeclean]

Need patch file or instructions as how to use moab_source.F90 as my installation fails to find iMesh_f.h. I can't seem to pass the appropriate variable using the MCNP build script.

I can make it work by hacking it, i.e.

1 make build CONFIG='......'
2 when it fails, cd into src, gfortran -c moab_source.F90 -I/path/to/moab/include -fcray-pointers -o moab_source.o
3 manual linking, cd into src, gfortran *.o -o mcnp5

I.e. a bit dirty.

[erelson]

Not sure if this is useful, but referencing info from a few past issues:

From #44:

To summarize the needed changes; in DAG-MCNP/config/Linux.gcf:

add -fcray-pointer to FFLAGS
add -liMesh to DAGMC_LIBS
add $(MOAB_INCLUDES) to FPPFLAGS

This issue is very similar to one I made for DAGMC: svalinn/DAGMC#48

[gonuke]

Do we know why we need the cray-pointer flag? Is this a MOAB thing?

[erelson]

Yup. I never investigated if there was a workaround to this... I think Fortran 90 has more modern, built-in pointers? Efficient bet would probably be to ask Tim.

[makeclean]

I think its the way you declare your pointer, its not std F90 form by the looks of it. Eg

 real, pointer :: bob

 bob => something

I could be wrong but it doesn't look F90ish

[erelson]

Yeah, I'm using cray pointers as that's how all the example Fortran ITAPS usage is done. It might be a F77 compatibility thing.

[makeclean]

You are almost certainly correct. :)