Comment author: @Octachron

This might warrant a clarification, but the current rule is that toplevel phrase are terminated by ";;"; and whatever comes after does not matter. In other words, if you are in a particularly interrogative mood, you can end your toplevel phrase with

() ;; ???

Comment author: vanto

Yes, this needs to be clarified. If this example above is calculated in Caml Light, the answer is:
#();;???

• : unit = ()
  there is no error. On the other hand if I write this other example in Caml Light then there is an error which is:
  #"er";;???
  Entrée interactive:

....???
.........
L'identificateur ??? n'est pas défini.

It seems that the sequence of characters, after the double-semicolon, is taken into account by OCaml and this generates a loss of time and unnecessary calculation. I think it's better to stop writing after a double semicolon at the end of a line in the Toplevel. Are you sure it does not matter?

Comment author: @Octachron

Caml Light is obsolete and its quirks are not the object of this issue tracker.

The behavior of OCaml's toplevel is consistent (and utop is available as a more user-friendly version of the toplevel).

Comment author: vanto

Caml light is quoted here for comparison and I think we had another approach to programming over thirty years ago. I think you are wrong and this little problem deserves our full attention. Here utop has nothing to do.

Comment author: @Octachron

A Github PR clarifying the behavior of the interactive mode would be welcome if you think that this behavior deserves your full attention.

Comment author: @xavierleroy

There are two orthogonal points here:

1- In OCaml, ";;" without anything before is parsed as a "do nothing" toplevel phrase. This works both in separate compilation (ocamlc, ocamlopt) and in interactive use (ocaml).

2- In interactive use only (ocaml), the rest of the line after the terminating ";;" is ignored.

Both 1- and 2-, and especially 2-, are tolerances: they cause no harm, and might possibly be useful (?), but could also go away at any time, so we're not going to document them. Just don't use those tolerances and you'll be safe.

Comment author: @gasche

The ability to enter an unattached attribute in the toplevel

[@@@foo];;

is useful (it can affect the state of a ppx transformation pass) and can be understood as an empty phrase (annotated).