Sadiq Jaffer

1. The proposed `Gc.freeze` API has some very sharp edges that I think will make it difficult to use, especially in larger codebases. 2. The 5.x runtime already supports off-heap...

This is curious, as it looks like the issue is happening before we even reach compactor code. The stack trace indicates this is happening in the minor collection required at...

Do we want to make some specific documentation changes that would avoid people hitting this in future?

Thanks for creating an issue for this @copy . So there's some tension in the API and design for `Runtime_events` at the moment in that it was designed both for...

> This looks rather obscure and brittle to me. What's wrong with specifying the file on `Runtime_events.start` with an optional argument ? Or even have a separate fun call to...

> Then why provide `Runtime_events.start` at all? Because we still want to support the usecase of programs monitoring themselves. > I mean it doesn't sound complicated to have a design...

> I've not used "runtime events" at all, only observed from a distance that this feature has been merged in 5.0. I wonder why the API is so different from...

The PR https://github.com/ocaml/ocaml/pull/11110 adds a selection of `CAMLno_tsan` attributes and relaxed atomics. It also adds an ignorelist for functions where the races reported are from `Field` and `Tag_val` macros. There's...

> Would making them relaxed atomics fix the issue? Yes, I think that would work.

> At first I thought that there cannot be a race here: if the domain is alone, then only that domain can spawn new domains, and so non-atomic `if (caml_domain_alone())`...