Gabriel Scherer

Designing language support for this idiom is tricky. A natural idea would be to write something like `foo[@reuse bar]`, for example `{a with foo}[@reuse a]` or `Fmapp(f, args')[@reuse tm]`. But...

A further comment regarding "simplifying the runtime code". Currently the C runtime code for threads contains a stateful initialization function that *modifies* the rest of the runtime state to now...

> Should the signal and wait code (basically, threads_posix.h) stay in systhreads? They're not required by the runtime, and you wouldn't have to carry the definition of the primitives. This...

It's not easy to tell which errors come from a lack of annotation and which are completely unrelated. The best I can think of is to locate whether the error...

> That explains the mismatch: I was using Merlin and assumed that Merlin just used the type-checker directly, so I didn’t think to use ocamlc. Is this a Merlin bug?...

I don't have the bandwith to work on this, and no one has volunteered to work on it. (Maybe @Octachron should have a second look just in case.) I propose...

Using the instrumented runtime enables extra runtime events that are not enabled by default (because they are emitted too often and would hurt performance). The two events are `alloc`, which...

I think we want more protection in the code: - `write_to_ring` should fail clearly when this happens, instead of looping (or maybe it should drop the event and continue, and...

In #13242, @MisterDA simplified some code in the runtime by *assuming* that the `CAMLunreachable()` macro is `[[noreturn]]`. On the other hand, #13344 goes in the opposite direction by making `CAMLassert(0)`...

(My comment on compilers is not quite right: even on systems that currently define `CAMLunreachable` as `__builtin_trap`, we could reasonably prefer to have some debug output with a source location...