Example:

  ocaml-odbc-devel.i386: W: file-not-utf8
/usr/share/doc/ocaml-odbc-devel-2.15/Exemples/monitor.ml

The situation here is complicated.  A standard OCaml source file
can usually be considered ISO-8859-1 encoded.  More precisely:

Identifiers may contain ISO-8859-1 characters.  In particular if
identifiers are converted to UTF-8 the program will no longer
compile:

  $ echo -e 'let m\xe9 = 1' > test.ml
  $ hexdump -C test.ml
  00000000  6c 65 74 20 6d e9 20 3d  20 31 0a     |let m. = 1.|
  0000000b
  $ ocamlc test.ml
  $ iconv -f iso-8859-1 -t utf-8 < test.ml > testu.ml
  $ ocamlc testu.ml
  File "testu.ml", line 1, characters 6-7:
  Illegal character (\169)

Comments in the source can contain ISO-8859-1 characters (and
given that the primary developers are French, this not just a
theoretical consideration).

Literal strings in OCaml programs are really byte arrays and
as such could contain just about anything.

Literal strings in, say, OCaml GTK2 programs might contain UTF-8
because GTK itself would be expecting UTF-8 for labels, messages, etc.

All of the above are (in my opinion) very bad practice -- one shouldn't
be using ISO-8859-1 for identifiers for example, and strings which
could contain foreign characters are better stored either as \escapes
or better still as external resources.  And using ISO-8859-1 in
identifiers is just insane.  Nevertheless, all of the above
are possible.

So I think the best thing is to disable this warning for *.ml, *.mli,
*.mly and *.mll files, unless you can think of a better way of handling
this.

I'm open to discussing changing Fedora OCaml policy to forbid this
sort of thing.  Hopefully it's fairly rare outside comments.