Thomas Refis

Also, could you add tests? (That `-I` works when given a single file, that positional arguments are interpreted as your last commit suggests, ...)

If I followed the discussion correctly, the idea of passing cmi/cmx files as positional arguments has been abandoned. So the remaining propositions are: 1. reuse `-I` 2. use explicit flags...

If you have a better sed line to offer, I'll take it. Notice that it already special cased one particular custom root, but I don't think the approach is scalable....

Unless I missed something, your first commit doesn't expose the buggy behavior? (I think this might be due to the fact that using `server` in tests is unfortunately somewhat flaky...

If people want to try the "let's just reload everything when the config changes" approach first (which makes sense, it's much simpler), I'd make the request that we hold off...

(By hold off I mean: before merging in master, you're off course free to implement it as soon as you can :p)

> I'd like to suggest to render identifiers mostly as they are found in the .mli, that is strip the open modules at most until the first prefix (so that...

Ping. I've just hit this, as I was thinking about how a merlin manual (currently being written) might render on the new v3.ocaml.org It seems *pretty important* to have this.

(And indeed https://v3.ocaml.org/p/vg/0.9.4/doc/Vg/index.html#paths vs https://erratique.ch/software/vg/doc/Vg/index.html#paths )

Ah, getting the ping one year late thanks to the stale bot, nice! Indeed I had already looked at this, apparently even pushed a few commits [here](https://github.com/trefis/ocaml/tree/coerce-patterns), although they don't...