Comment author: bobot

One possible way would be to add the ability to create a .cmi from a .cmi and a .mli which is included in the signature of the given .cmi.

Given the files internal/a.ml internal/a.mli external/a.mli with external/a.mli included in internal/a.mli

ocamlc -o internal/a.cmi internal/a.mli
ocamlc -o internal/a.cmo internal/a.ml

ocamlsig -o external/a.cmi internal/a.cmi external/a.mli

install external/a.cmi internal/a.cmo

Internally internal/a.cmi is used but the user of the library use external/a.cmi.

PS: We can play with attributes in order to "create" external/a.mli from internal/a.mli

Comment author: @garrigue

What you should do is just create an export version of your module:

mylib_a.ml:
include Mylib_intern_a

mylib_a.mli:
type t = Mylib_intern_a.t

And do not export mylib_intern_a.cmi

Of course the dangling type names are not very nice, but this is not a new.

Comment author: bobot

Thank you, that simply works! It will need some plumbing (module-alias, -open, a ppx that filter the signature) for making that transparent to the library user and the developer, but indeed it can work, it is not too complicated and it requires no modification of ocaml!

Except that:

• It would be handy to have an option for ocamlc/opt that it should use the current .cmi even if there is no .mli. From one .mli we would create two .cmi` using a ppx for filtering the declaration (I think there is also such request for parallelism). If you think the idea is acceptable, I can do the merge request.

• with -no-short-path you have in the library user type errors:

Mylib_a.t = Mylib_intern_a.t

but who doesn't use short-path?

This issue has been open one year with no activity. Consequently, it is being marked with the "stale" label. What this means is that the issue will be automatically closed in 30 days unless more comments are added or the "stale" label is removed. Comments that provide new information on the issue are especially welcome: is it still reproducible? did it appear in other contexts? how critical is it? etc.