[CI] Add symbols.yml and modules.yml

[amontoison]

@jfowkes @nimgould
I add two new CI builds with this pull request.

The goal of the first new build (symbols.yml) is to check if the following shared libraries export the same symbols:

• libgalahad_single.so with 32-bit integer
• libgalahad_double.so with 32-bit integer
• libgalahad_single.so with 64-bit integer
• libgalahad_double.so with 64-bit integer

Two libraries can export the same symbols ONLY IF it's exactly the same routine.
For example libgalahad_single.so and libgalahad_double.so can both export iamax_ ONLY IF both libraries are
compiled with 32-bit integer.

Note that I don't compile GALAHAD with SSIDS for now but it seems that many C++ symbols are shared by both single / double libraries.
We probably need to preprocess some symbols in C++ files...

The goal of the second new build (modules.yml) is to check that each library generates Fortran modules with different names.

[amontoison]

An example of what we can conclude from the output of the two new builds:

• modules.yml : https://github.com/ralna/GALAHAD/actions/runs/7897536131/job/21553279373?pr=261#step:9:20
  We need to add macros for HSL modules with 64-bit integer.

• symbols.yml : https://github.com/ralna/GALAHAD/actions/runs/7897536129/job/21553279371?pr=261#step:9:33
  bsc_information and fit_information don't have the suffix _s in single precision.

[amontoison]

@nimgould Can you confirm that the following Fortran modules are the same in single and double precision?

• "galahad_blas_interface"
• "galahad_lapack_interface"
• "spral_ssids_blas_iface"
• "spral_ssids_lapack_iface"
• "galahad_clock"
• "galahad_string"
• "galahad_symbols"
• "galahad_tools"
• "galahad_copyright"
• "galahad_hash"
• "galahad_common_ciface"
• "galahad_hash_ciface"
• "galahad_hsl_kb22_long_integer"
• "galahad_hsl_mc68_integer"
• "galahad_hsl_mc78_integer"
• "galahad_hsl_of01_integer"
• "galahad_hsl_zb01_integer"
• "galahad_hsl_mc68_integer_ciface"
• "galahad_kinds"
• "galahad_kinds_single"
• "galahad_kinds_double"
• "spral_kinds"
• "spral_kinds_single"
• "spral_kinds_double"
• "mkl_pardiso"
• "mkl_pardiso_private"
• "lancelot_hsl_routines"
• "pastixf_enums"
• "pastixf_interfaces" (x)
• "spmf_enums"
• "spmf_interfaces" (x)
• "spral_core_analyse"
• "spral_hw_topology"
• "spral_metis_wrapper"
• "spral_pgm"
• "spral_ssids_profile"

I'm not sur for the last ones.

[amontoison]

@nimgould I checked the modules of my previous message tonight and the modules pastixf_interfaces and spmf_interfaces are different in single and double precision.

They should have a different name to not collapse if a library / binary is linked with both -lgalahad_single and -lgalahad_double.

The other modules are only using integers so it's fine if they have the same name in single / double precision.

[amontoison]

Nick, we renamed the Fortran modules for 64-bit integer but we did nothing for the C routines.
Should we also add a suffix _64?
https://github.com/ralna/GALAHAD/actions/runs/7906337216/job/21580936425?pr=261#step:9:1007

[nimgould]

Yes to both of these, I will sort them out as soon as I have a working draft of the HSL subset library.

[nimgould]

I am not sure I know what to do with the pastix/spm interface modules as these are provided by the pastix people with the same name for single and double. I can change the galahad dummy names, but what happens when somebody links against a binary library compiled by someone else? Any ideas?

[nimgould]

I could presume that the single precision module is pastixf_interfaces_single, while the double one is just pastixf_interfaces (without a _double)?