Commit d8bee138 authored by POTTIER Francois's avatar POTTIER Francois

Document a couple traps that people fall into with headers and types.

parent 9afcfb1a
......@@ -502,6 +502,24 @@ their order is preserved. However, when two headers originate in distinct
grammar specification files, the order in which they are copied to the \ml
file is unspecified.
It is important to note that the header is copied by \menhir only to the \ml
file, \emph{not} to the \mli file. Therefore, it should not contain
declarations that affect the meaning of the types that appear in the \mli
file. Here are two problems that people commonly run into:
\item Placing an \kw{open} directive that is required for a \dtype declaration
to make sense. For instance, writing \verb+open Foo+ in the header and
declaring \verb+%type<t> bar+, where the type \verb+t+ is defined in the
module \verb+Foo+, will not work. You must write \verb+%type<Foo.t> bar+.
\item Declaring a module alias that affects a (declared or inferred) type. For
instance, writing \verb+module F = Foo+ in the header and declaring
\verb+%type<Foo.t> bar+ will not work (from
2020/05/25 on). The reason is, OCaml infers that the symbol \verb+bar+ has
type \verb+F.t+, and Menhir relies on this information without realizing
that \verb+F+ is a local name, so in the end, the \mli file contains a
reference to \verb+F.t+ that does not make sense.
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment