Enable POSITION_INDEPENDENT_CODE for FMS CMake

[mathomp4]

Is your feature request related to a problem? Please describe.

I've been trying to build GEOS with the latest FMS and after not being able to do it with Autotools (see #805), I went back to CMake build and I got close. So close except I got a message like:

relocation R_X86_64_32 against `.rodata' can not be used when making a shared object; recompile with -fPIC

It wasn't exactly this, but my terminal lost the real error as I stupidly had only like 100 lines of playback. ☹️

The main issue is that we build MOM6 as a shared object library, but as far as I can tell, FMS builds as a static library (libfms_r4.a). Or it least, as I have it set up, that's how it built for me.

We've encountered this before with some of our Baselibs and the usual solution is to "add -fPIC" or in CMake terms...

Describe the solution you'd like

Add:

set(CMAKE_POSITION_INDEPENDENT_CODE ON)

to CMakeLists.txt

I did this in my 2021.03 checkout at line 51 and...it worked.

Describe alternatives you've considered

I guess I could try building FMS as a shared object library, but I'm not sure how. My scan of its CMake and instructions didn't seem to show how.

Now, the question is: does setting this hurt you? I don't think it should as it just adds -fPIC everywhere and that's nigh-standard nowadays.

[mathomp4]

Note: There is also the "per target" version:

set_target_properties(mylib PROPERTIES POSITION_INDEPENDENT_CODE ON)

but I'm usually lazy and do it globally. 😄

[aerorahul]

@mathomp4
We should be able to choose at build time to build a shared or static FMS library (or both).
It will need minimal modifications to CMakeLists.txt and I know what needs to be done to enable this.

If you want, I can provide a patch/branch for you to test with GEOS in the next few weeks.