Xavier Leroy

Thanks for the `Global` fix. The warnings are all in the vendored copy of the MMaps library, which is used only for benchmarking purposes and is now maintained upstream (https://github.com/coq-community/coq-mmaps)....

I have the impression this is a Dune issue, and submitted it: https://github.com/ocaml/dune/issues/9152 . This is such a trivial project (2 small OCaml files...) that we could force a sequential...

> I did notice that stream.cmti is not an explicit output but a side-output of the creation of steam.cmi . Is that file really used at runtime? It's a compact,...

> Wild guess: the interface of Genlex (in the standard library) refers to Stream. Could it be that that reference means two different things depending on whether stream.cmi has already...

> physical equality is a subtle operation, used by people who care about the performance implication of their code, Physical equality is essential for the correctness of many algorithms over...

I'm still hesitating between `phys_equal` and an infix operator, probably `===`. I think we should gather more opinions (beyond those that have already been vehemently exposed here), in particular from...

I agree this test is cursed and should probably remain deactivated. Historically, it was written for the "green thread", bytecode-level thread library, as a test for the scheduling of `Thread.delay`...

> the interpreter should do a goto check_stacks right before entering the main loop Well spotted! See the latest commit on this PR. > shouldn't caml_callbackN_exn do a stack limit...

Yes, `Stack_threshold_words`. I pushed an alternative to @stedolan's suggestion, based on @NickBarnes' remark, where `caml_callbackN` checks for stack space and reallocates the stack if needed. I doubt the test suite...

I added a test that triggers the stack resizing, and a Changes entry. This PR is ready for a final review.