Always define WINDOWS_UNICODE internally

[dra27]

Rather than just defining WINDOWS_UNICODE for the Windows builds, define it for all builds (it's just always 0 on Unix). Split off from #9284.

[xavierleroy]

I can see many a Linux or macOS users surprised by seeing WINDOWS_UNICODE on the command line when doing ocamlopt -verbose or watching build logs. What is the motivation for this change?

[dra27]

They won't see that - this only goes to OC_CPPFLAGS (calling the C compiler during the build), not OCAMLC_CFLAGS/OCAMLOPT_CFLAGS (ocamlc/ocamlopt calling the C compiler)

#9284 tests if WINDOWS_UNICODE is 1 in the pre-processor - the behaviour will be correct if it's not defined, but it seemed better (to me) to have it always defined.

[xavierleroy]

They won't see that - this only goes to OC_CPPFLAGS (calling the C compiler during the build), not OCAMLC_CFLAGS/OCAMLOPT_CFLAGS (ocamlc/ocamlopt calling the C compiler)

OK, but it will still show up in the build logs for OCaml itself.

As you mention, #if WINDOWS_UNICODE will evaluate to false if WINDOWS_UNICODE is undefined, and nothing bad will happen. If you find this too obscure, #if defined(WINDOWS_UNICODE) && WINDOWS_UNICODE will do the job. None of this warrants a change to configure and even longer command lines when building OCaml.

[dra27]

I originally didn't think I'd get away with relying on the defined behaviour for undefined variables 🙂