David Allsopp

> Is that too big to add to every OCaml binaries? If so, would it sound reasonable to split the `Marshal` module in two (one builtin, and the other with...

I've experimented, too - all with no definite conclusion in either direction. A few thoughts/musings: - Moving the zstd-specific parts of `extern.c` and `intern.c` to a separate library was less...

We're on it. As a temporary workaround, if the static binaries are being built in Docker or another CI system, is it not possible just to ensure that libzstd-dev (or...

Was gating this change on ocaml/ocaml/RFCs#27 discussed? The keyword part has an implementation in #11252 (which I think is aiming to be ready for 5.2 as well, so this shouldn't...

Thanks for this! As it stands, this is definitely a large contribution, which I think will still require a CLA prior to merging (@xavierleroy?). There's also the testing story -...

@kant2002 - `configure` hasn't been committed in the PR; running `tools/autogen` should regenerate it.

In addition to `configure` being regenerated and committed, `loongarch64` also needs adding to the `ARCHES` variable in `Makefile`: https://github.com/ocaml/ocaml/blob/0e4290c162e3f7a6a8489b9526d32770065793a3/Makefile#L34

> Yes, we also found this problem, where selection.ml subsequent we will add features according to the loongarch architecture features, other issues may require higher abstraction to reduce code duplication...

There are various news stories[^1][^2] which suggest it's the case. However, the s390x backend is maintained on a similar basis, so this is a possibility, yes. [^1]: https://uk.pcmag.com/processors/144404/china-bans-exports-of-its-homegrown-loongson-chips [^2]: https://www.scmp.com/tech/article/3203454/china-bans-export-loongson-chips-due-strategic-importance-and-military-sensitivity-russian-newspaper

Looking only at the second failure, and it may now be hard to answer, but are you certain that you'd re-run configure at this stage? The recent changes altered `runtime/caml/m.h.in`...