LGTM, thanks for the update!
Should be SCM_I_FIXNUM_BIT < 32, and the last two commits should logically go together.
I think SCM_I_FIXNUM_BIT < 32 is the correct condition here: if it was equal to 32, we could just create an inum and be done. To make the link, you should probably squash the last two commits to...
LGTM, thanks for detailed explanations!
I missed at first that you tested "with prebuilt binaries"... For me, without the build time improves from 14m24s to 13m12s on my 8 core laptop processor. With prebuilt bytecode, it's 6m57s to 6m.
Yes, I'm not convinced that all of this is necessary. Can we first take a step back and assess what is the current situation: Does Guile build or not? I'm only talking about make, not yet about...
This PR is, once again, a wild mix of "obviously correct" fixes (such as using scm_from_uintptr_t) and intricate changes in integers.c that I'm not sure are needed. Can these be split into...
There is an easier approach that we use in LilyPond (not yet upstreamed): https://gitlab.com/lilypond/lilypond/-/blob/master/release/binaries/patches/0009-mpz-integer-do-not-convert-to-bignum-if-in...
This is incorrect because it's possible to build Guile with "normal" GMP on 64-bit Windows.
@spk121 if you can share some more details about what doesn't compile, I may be able to help / make educated guesses of what's missing.
AFAICT the configure-flags pass --enable-mini-gmp unconditionally, so you probably want to modify that as well?