Vincent Laviron

I'm proposing to disable this closure computation in cases with too many functions. @chambart told me that it's very unlikely we will discover anything useful here, particularly on menhir-generated parsers.

To clarify a bit: the main goal of this piece of code is to discover invariant parameters. If we have invariant parameters in a set of recursive functions, then it...

> Is someone planning to submit a PR that does this? Done in #13379.

@gasche I'm considering reviewing this. I have already noted that I'm not moved by the performance argument, which I think is better solved by my own PR, but I'm fine...

> * the underlying problem is that `caml_shutdown` will segfault if other domains are still running, and that should be tested and fixed, but it should be done in another...

> If the variable is lifted before the class definition, it's cumbersome to hide it for the rest of the module (I don't see how to do it without writing...

> If the variable is lifted before the class definition, it's cumbersome to hide it for the rest of the module (I don't see how to do it without writing...

The opam repository does not contain many tests, so it's likely going to miss a lot of the potentially impacted users. But I guess we might start with that. If...

There are indeed a number of pieces of code where the change of semantics may be observable. A number of these results only bind functions, so the only observable change...

Extensible types can be extended an arbitrary number of times: ```ocaml type t = .. let r : t list ref = ref [] let add_new_value () = let open...