[BUG] debug locations sometime wrong, affecting sourcemaps

[jordwalke]

While modifying parse_bytecode.ml for rehp, I might have stumbled upon some evidence that parse_bytecode.ml is responsible for source maps being
difficult to use in practice (they jump all over the place in very unintuitive ways).

Problem: For the following Reason code:

let greeting = "hello world";
print_endline(greeting);
let greeting = "hello world with unicode: Ɋ";
print_endline(greeting);       /* THIS IS LINE 4 */

let unicodeLength = String.length(String.make(Random.int(30), 'i'));
print_endline(
 "String.length(\"Ɋ\") should be two:" ++ string_of_int(unicodeLength),
);
print_endline(String.make(1, "Ɋ".[0]) ++ String.make(1, "Ɋ".[1]));

We get the following debug data: (Notice the many things claiming to be on line 4 that aren't!
For example _g_ (the number 30))

  /*<<rehack_tests/strings/strings.re 2 0>>*/  /*<<rehack_tests/strings/strings.re 2 0>>*/ print_endline(
   greeting
 );
  /*<<rehack_tests/strings/strings.re 4 0>>*/  /*<<rehack_tests/strings/strings.re 4 0>>*/ print_endline(
   greeting__0
 );
  /*<<rehack_tests/strings/strings.re 4 0>>*/ var  /*<<rehack_tests/strings/strings.re 4 0>>*/ _f_ =
   105,
   /*<<rehack_tests/strings/strings.re 4 0>>*/ _g_ = 30,
   /*<<rehack_tests/strings/strings.re 6 46>>*/ _h_ =
    /*<<rehack_tests/strings/strings.re 6 46>>*/ int__1(_g_),
   /*<<rehack_tests/strings/strings.re 6 34>>*/ unicodeLength =
    /*<<rehack_tests/strings/strings.re 6 34>>*/ caml_ml_string_length(
      /*<<rehack_tests/strings/strings.re 6 34>>*/ make__0(_h_, _f_)
   ),
   /*<<rehack_tests/strings/strings.re 8 44>>*/ _i_ =
    /*<<rehack_tests/strings/strings.re 8 44>>*/ string_of_int(unicodeLength),
   /*<<rehack_tests/strings/strings.re 8 2>>*/ _k_ =
    /*<<rehack_tests/strings/strings.re 8 2>>*/ _a_(_j_, _i_);
  /*<<rehack_tests/strings/strings.re 7 0>>*/  /*<<rehack_tests/strings/strings.re 7 0>>*/ print_endline(
   _k_
 );
  /*<<rehack_tests/strings/strings.re 4 0>>*/ var  /*<<rehack_tests/strings/strings.re 10 57>>*/ _l_ =
   138,
   /*<<rehack_tests/strings/strings.re 10 57>>*/ _m_ = 1,
   /*<<rehack_tests/strings/strings.re 10 42>>*/ _n_ =
    /*<<rehack_tests/strings/strings.re 10 42>>*/ make__0(_m_, _l_),
   /*<<rehack_tests/strings/strings.re 10 29>>*/ _o_ = 201,
   /*<<rehack_tests/strings/strings.re 10 29>>*/ _p_ = 1,
   /*<<rehack_tests/strings/strings.re 10 14>>*/ _q_ =
    /*<<rehack_tests/strings/strings.re 10 14>>*/ make__0(_p_, _o_),
   /*<<rehack_tests/strings/strings.re 10 14>>*/ _r_ =
    /*<<rehack_tests/strings/strings.re 10 14>>*/ _a_(_q_, _n_);
  /*<<rehack_tests/strings/strings.re 10 0>>*/  /*<<rehack_tests/strings/strings.re 10 0>>*/ print_endline(
   _r_
 );

But if you add the following print statements right after this line in parse_bytecode.ml
You will see that there are only two debug events for line 4, (likely the correct line print_endline(greeting))

let _ = print_string ((if after then "<after>" else "") ^ (if before then "<before>" else "") ^ " DEBUG EVENT FOUND AT: " ^ (pos.Lexing.pos_fname) ^ ":" ^ string_of_int(pos.Lexing.pos_lnum) ^ ":" ^ string_of_int (pos.Lexing.pos_cnum - pos.Lexing.pos_bol)) in
let _ = print_string ("  " ^ (if after then "<after>" else "") ^ (if before then "<before>" else "") ^ " DEBUG END   FOUND AT: " ^ (loc.Location.loc_end.Lexing.pos_fname) ^ ":" ^ string_of_int(loc.Location.loc_end.Lexing.pos_lnum) ^ ":" ^ string_of_int (loc.Location.loc_end.Lexing.pos_cnum - loc.Location.loc_end.Lexing.pos_bol)) in
let _ = print_newline () in

[jordwalke]

This site is really good for debugging issues visually: http://sokra.github.io/source-map-visualization/#custom

[hhugo]

Also, OCaml locations are bytes based where sourcemap seems to expect code-point based.
ocaml/ocaml#11899
terser/terser#960