Lift the restriction on installing 5.3.0+BER on macos+arm64

[yallop]

@kayceesrk reports that MetaOCaml 5.3.0+BER can be installed successfully on macos+arm64, so the current constraint is apparently unnecessary.

[yallop]

FWIW the same can be done for 4.14.1+BER:

Thanks. I've made that change in 20f0163

[dra27]

I've pushed two additional commits which will hopefully improve the CI. As it's written at the moment, the 4.14.1 and 5.3.0 packages don't support bytecode-only builds (I'm not sure if MetaOCaml actually supports that itself?), so I added the conflict to the 4.14.1 version as well (that happened to be being caught by CI).

The second commit brings a few of the changes which are in ocaml-compiler.5.3.0 into the BER version - hopefully the 32-bit buillds will be correctly rejected this time.

[dra27]

The lint error is incorrect (as for compiler packages in general). The failures on Arch I think can be ignored, although there's trouble on the horizon! I think that's GCC 15. For the 4.14.1 version, it's missing ocaml/ocaml#12577 (from 4.14.2). For the 5.3.0 version, I'm wondering if the BER patches have accidentally backed out ocaml/ocaml#12509 and ocaml/ocaml#11764 (which are the "trunk" parts of the 4.14.2 PR), but I haven't looked in detail.

[shonfeder]

The outstanding CI errors:

archlinux-ocaml-4.14 ocaml-variants.4.14.1+BER (failed: too many arguments to function ‘(value ()(void))(caml_prim_table.contents + (sizetype)((long unsigned int)*pc * 8))’; expected 0, have 1) archlinux-ocaml-5.3 ocaml-variants.4.14.1+BER (failed: too many arguments to function ‘(value ()(void))(caml_prim_table.contents + (sizetype)((long unsigned int)*pc * 8))’; expected 0, have 1)

IIUC, @dra27 you think we should hold off until these are fixed? (The other option being we mark the package unavailable or in conflict?).

[yallop]

For the 5.3.0 version, I'm wondering if the BER patches have accidentally backed out ocaml/ocaml#12509 and ocaml/ocaml#11764 (which are the "trunk" parts of the 4.14.2 PR), but I haven't looked in detail.

At a glance, it doesn't look that way: the commit https://github.com/metaocaml/ber-metaocaml/commit/d6b868c (which seems to be the reworking of those PRs in 5.3.0) is on the ber-n153 branch, which is the source for the 5.3.0+BER variant.

[dra27]

No, I’d suggesting ignoring those errors - the existing 4.14.1 is already not working on Arch, so nothing here I think makes it worse. Not sure about constraining it away completely

[mseri]

Is my understanding correct that this should be merged then?

[Zeta611]

I have checked that 5.3.0+BER works on macOS as well!

[mseri]

Thanks