Florian Angeletti

The split in the two commits was exactly what I had in mind. Merged, thanks for your contribution !

The printer is installed but will fail to match any type until all inference variables are unified to a ground type. I agree that switching to an error would be...

I have pushed an example of code where this PR introduces a regression by using the same printed name for different identifiers in the same error message. My point of...

After some hesitation, I am now planning to close this approximative solution and implement a correct one in 5.4, after we replace the implementation of short paths.

@kit-ty-kate , I have mostly two use cases in mind: - an easy and small fix at the build system or package manager for code base using the `effect`, without...

@stedolan , note that this PR also allow to set the `keywords` flag through environment variable via `OCAMLPARAM`. My aim for the `-keywords` flag is mostly to be able to...

Even without optional argument, this is also a type system change that can break existing code: ```ocaml 0 |> Fun.id (fun ~x y -> x + y) ``` is currently...

Since for packages we are only applying the constraint on abstract types, the meaning should be really the same. Maybe one small disadvantage with supporting nested `with type ...` is...

One issue is that I see with that approach is that the situation is more symmetrical than the spellchecking hint for missing identifiers. Typically, for the example above, a better...

For module error messages, we already compute all missing items and print all of them, thus we are already outside the "stop on the first error" case. If we wanted...