B. Szilvasy

Thank you very much for the review. > - Implement `caml_plat_futex` and `caml_plat_barrier` > > 1. This commits looks good. I am however wondering why you are adding an explicit...

What I'm going to do is push a commit to address all comments first, and then I will rebase to fix conflicts ~and to combine the changes with previous commits~...

I believe I've marked as resolved all those review comments I've affirmatively addressed or that aren't relevant with new changes. See the latest commit for details, but notably: - I've...

Looking over it again, I've grown sceptical of 51bccd2c66f5715dabcb6129bdff44d2fc05495a. I think I will drop it here and perhaps address it in another PR.

The rebase is just to fix conflicts with d77bc97a0a90cd70babf047d63d0f35775de3874. @gasche Regarding the measurements, I did a run of your script on my machine (with 12 hardware threads). Whether it improved...

The last commit is new, doing away with `caml_wait_interrupt_serviced` and the leader arriving into the barrier the same way as other domains do: [diff](https://github.com/ocaml/ocaml/pull/12579/commits/86994c02d2c058b555ae650ba06a3cabdb34f537#diff-67115925103982a8ebeb085cfab5ef31a182c9a442bc51e053934364d3750dafR1665-R1669)

Rebasing to fix conflicts, mainly with #13063 (all Mutex/CV uses in this PR are within STW sections, or are used as a fallback implementation for the otherwise fully blocking futex,...

It looks like, perhaps unsurprisingly, not all platforms work with the new test (though MSVC just had a quite disappointing network error), so I intend to remove it later, since...

I've spotted and fixed another bug, notably `Runtime_events.create_cursor None` never closing the file descriptor on Unix. On trunk this crashes after opening too many file descriptors, causing `create_cursor None` to...

Another situation where runtime events aren't cleaned up is on an unhandled exception; the fix would be much simpler than for signals, by just performing the needed cleanup in `caml_fatal_uncaught_exception`...