It was apparently a mistake to try to combine this PR with also eliminating MKEXE_ANSI from the build system. Assuming this finally builds, I ask for this to be merged first, and then a PR will follow with the additional work for removing MKEXE_ANSI completely.

@nojb - ocamlyacc was already using the runtime headers, except that before there was duplication between what was in ocamlyacc and what was in the runtime (with a non-fatal error where mingw was concerned). I don't think there's any particular reason for ocamlyacc to be kept self-contained, it's just that it happens not to be "original"?

I don't have a strong opinion either way on windows_unicode_enabled - it was just that in having to lift that function out of the runtime (since it uses OCaml's memory management functions), I either had to hard-code the value or just duplicate the variable. The only reason for doing it this way is that you then don't end up with an ocamlyacc which does recognise UTF8 alongside a compiler which does not... but it's a largely irrelevant distinction. Does it bother you enough to require changing?

Oups! Very sorry I completely forgot about this one. Yes, I had a few slight questions/remarks about the PR but was convinced by @dra27 answers, so I'm approving as well :)

I'll merge once CI gives the nod.

Contrary to @XVilka's excitement in #8797, this PR adds very little of interest for the Windows platform but, as noted above, it does allow me to dust off the revisions to autoconf for FlexDLL bootstrapping, so thanks for the approval, @nojb!