Link argument parsing functions from libraries, instead of duplicating their source files

[dmikushin]

This PR basically removes

-set(SHARED_SOURCES
-  ${SHARED_SOURCES}
-  ${LIB_SOURCE_DIR}/ADT/hash.c
-  ${LIB_SOURCE_DIR}/ArgParser/arg_parser.c
-  ${LIB_SOURCE_DIR}/ArgParser/debug_action.c
-  ${LIB_SOURCE_DIR}/ArgParser/xflag.c
-  )

and changes a few more things accordingly.

The idea of SHARED_SOURCES is to have the same source files compiled with different macro definitions. For this reason, the source files in SHARED_SOURCES cannot be shipped normally, as a static library.

But the purpose above does not apply to the argument parsing functions - they are universal. Therefore, this PR is taking them out of the SHARED_SOURCES.

[dmikushin]

Need something extra on Windows:

lld-link: error: undefined symbol: flg
lld-link: error: undefined symbol: gbl
lld-link: error: undefined symbol: version

[bryanpkc]

Can we avoid including these headers within gbldefs.h, and instead include them in the .c or .cpp files that actually need the function declarations?

[bryanpkc]

This rename doesn't seem necessary at this point. In a future PR, perhaps we could move listing.c and listing.cpp from under tools/flang1/flang1exe/ and tools/flang2/flang2exe/, and merge them into one file under tools/shared/.

[dmikushin]

I think I did this to come up with a proper name mangling and resolve link errors.

[bryanpkc]

Will you fix the Windows build failure as well?

[dmikushin]

I don't have a Windows builder yet, but I'm looking forward to getting it!

Currently, we are still discussing with you the enablement of Linux builds through Docker. How could I encourage you to have them, or accept mine? This is related, because I'm not able to use the CI and develop in it.

[pawosm-arm]

I wouldn't do that. The original authors wanted to decouple these two as so they wouldn't be stopped with improving flang2 by the anchors deeply rooted in flang1 (and e.g. making flang2 compiled with a C++ compiler (versus C compiler) was one of the means for achieving such improvement). Some duplication was inevitable. This commit goes against that tide and restores the coupling.

[bryanpkc]

Closing the PR as it will not be accepted as-is.