truncated error message with Emacs+Merlin on OCaml 5.04 and v5.7.1-504

[gasche]

Repro case:

module type A = sig type a end
module type B = sig type b end

module A : A = struct type a end
module B : B = struct type b end

module F (_ : A) (_ : B) = struct end

module Test = F(B)(A)

Merlin shows the following error message:

2. Modules do not match: A : sig type a = A.a end is not included in B

(The exact same thing is in the *Messages* and *merlin-errors* buffer.)

This is only the last line of the full error message shown by the compiler:

Error: This application of the functor F is ill-typed.
       These arguments:
         B A
       do not match these parameters:
         (Arg : A) (Arg : B) -> ...
       1. Modules do not match:
            B : sig type b = B.b end
          is not included in
            A
          The type a is required but not provided
          File "test.ml", line 1, characters 20-26: Expected declaration
       2. Modules do not match: A : sig type a = A.a end is not included in B

[gasche]

Note: does Merlin have a testsuite checking that error messages are included correctly in the output? Would it help if we compiler developers (cc @Octachron who makes our error messages greater, and also, well, longer) could help by letting you know about new or improved error messages so that you can include them in the testsuite, to detect this kind of issues earlier.

(Maybe the ongoing work on upstreaming type recovery will improve this? Maybe the marathon work of Octachron to get proper structured (sexp, json) outputs from the compiler (ocaml/ocaml#9979, ocaml/RFCs#45, ocaml/ocaml#13766) will make this category of errors go away completely?)

[xvw]

I think, but I am maybe wrong, this is related to the client (merlin.el or ocaml-eglot) because that part of the error is included as a .. "footnote" and in OCaml LSP, they are included in the diagnostic. Unfortunately, Eglot (not ocaml-eglot) discard that part of the diagnostic.
Since Emacs is not particularly colorful in the error buffer, I think that include application ill can be hard to read btw

[gasche]

(Here I am using merlin.el, not ocaml-eglot.)

I'm not sure if you suggest that the responsibility to fix this lies in another piece of software maintained elsewhere, and I should report the issue there. My guess would be that ocaml/merlin maintains both merlin-the-server and merlin.el, so the fix would be here I suppose.

[xvw]

I was not clear, sorry, my goal was to give more context, not to says that we need to move the error in an other repo: merlin client (eglot, merlin.el and probably OCaml.nvim/vim) does not handle additional hints in error diagnostics.

[voodoos]

An option for LSP-based clients we could have a client capability to tell the server not to use the "footnote" field and have the server not use that field if the client doesn't support it.

[gasche]

But here the client is merlin.el, could it work harder to support actually showing the error message in full?

(Or: maybe merlin could actually never use the footnote field and just include all of the error message in the main part of the error message payload, whatever that is?)

I am still a bit confused by the explanation proposed. The error message that is shown is just the very last line of the error message, which starts with 2. .... If the problem is that footnotes are not shown, then that suggests that this is somehow the "main" part of the error message, and all the rest above is a footnote? Would Merlin really decide that for example "Error: This application of the functor F is ill-typed." should be a footnote, while "2. Modules do not match: A : sig type a = A.a end is not included in B" is not? If that was going on, it should be fixed, shouldn't it?

[xvw]

Yes, I think that errors are truncated (with an arbitrary rule) on merlin.el.
We can try to fix it on merlin.el but the mode is not well maintain, since the error seems also present in ocaml-eglot (tarides/ocaml-eglot#78) and I will try to fix it, is there other things that block you from moving from merlin.el to ocaml-eglot?