netcdf-fortran: NETCDF.mod install location varies with universal variant

[ryandesign (Ryan Carsten Schmidt)]

netcdf-fortran installs NETCDF.mod to a different location depending on whether or not the universal variant is used:

$ port -v -q installed netcdf-fortran
  netcdf-fortran @4.2_6+gcc47 (active) platform='darwin 12' archs='x86_64'
  netcdf-fortran @4.2_6+gcc47+universal platform='darwin 12' archs='i386 x86_64'
$ port contents netcdf-fortran | grep mod
  /opt/local/include/NETCDF.mod
  /opt/local/include/TYPESIZES.mod
$ sudo port activate netcdf-fortran @4.2_6+gcc47+universal
--->  Computing dependencies for netcdf-fortran
--->  Deactivating netcdf-fortran @4.2_6+gcc47
--->  Activating netcdf-fortran @4.2_6+gcc47+universal
--->  Cleaning netcdf-fortran
$ port contents netcdf-fortran | grep mod
  /opt/local/include/TYPESIZES.mod
  /opt/local/include/mods32/NETCDF.mod
  /opt/local/include/mods64/NETCDF.mod

How are ports that want to use NETCDF.mod meant to deal with this situation? Currently it is leading to build failures like comment:ticket:38995:10.

[cooljeanius (Eric Gallager)]

Cc Me!

[cooljeanius (Eric Gallager)]

The muniversal portgroup is the cause for this: browser:trunk/dports/_resources/port1.0/group/muniversal-1.0.tcl#L546​

This would be easier for portfile authors to deal with if #32428 were fixed...

[ryandesign (Ryan Carsten Schmidt)]

Oh. Looks like it's been that way for awhile, since r49727. But I'm still curious how ports are intended to use this. How are ports supposed to find a file provided by another port if that file's location varies?

[ryandesign (Ryan Carsten Schmidt)]

I no longer see this difference between netcdf-fortran @4.2_9+gcc48 and netcdf-fortran @4.2_9+gcc48+universal... so is this fixed?