Florian Angeletti

Indeed, those kind of known possible C runtime API changes are one of the reason we have this 0th alpha release with less stability guarantees than an ordinary alpha.

Since the change is quite invasive, I am very tempted to move this PR to the 5.1 milestone to give more time to the reviewers and alternative designs.

If the issue is that you have a complex inferred type that you want to lift to the module level, first-class modules are enough: ```ocaml module type T = sig...

@dra27 how bad was the slowdown for @gasche nospill version on your machine?

The segfaults should be fixed with #11742, even if there are still rooms for (micro-)optimizations at a later point for interested people.

The feature wish seems reasonable to me, and seems to have only be stuck due to naming paralysis.

Another option would be to have a node for the generative application `F()` itself.

The symmetric construct already exists in fact: https://github.com/ocaml/ocaml/blob/trunk/parsing/parsetree.mli#L819 .

- [x] Looking at the stdlib diff with the anchor point, there is a bit of rebasing noise in `stdlib/filename.ml`: in the definition of `check_suffix_name` at line 103, `String.ends_with` has...

It seems probably better to also add `str.ml` to this table even if it not strictly part of the standard library?