error report cites line-number which is not anywhere close to the actual error. (Mantis 7063)

[damiendoligez]

This is http://caml.inria.fr/mantis/view.php?id=7063 :

In some cases camlp4 reports very odd line-numbers when there is an error.

When I compile the attached file:

ocamlfind c -g -c -syntax camlp4o -package mikmatch_str main.ml

I get the following error:

File "main.ml", line 8, characters 7-8:
Parse error: [regexp] expected after "=" (in [str_item])
File "main.ml", line 1:
Error: Error while running external preprocessor
Command line: camlp4 '-I' '/home/me/.opam/4.02.3/lib/ocaml/camlp4' '-I' '/home/me/.opam/4.02.3/lib/ocaml' '-I' '/home/me/.opam/4.02.3/lib/ocaml' '-I' '/home/me/.opam/4.02.3/lib/mikmatch_str' '-parser' 'o' '-parser' 'op' '-printer' 'p' 'str.cma' 'unix.cma' 'pa_mikmatch_str.cma' 'main.ml' > /tmp/ocamlppca2754

There, camlp4 tells me that something is wrong on line 8, column 7
which is very far from the actual error (line 30, column 54).

In this case, it seems that the reported line number with the error
and the line with the actual error can be arbitrarily far.

This can be a source of frustration for programmers.

The attached file is very artificial, but I stumbled upon this with a real program I am working on.

type r1 = {l1a : r4; l1b : r2}
and  r2 = {l2a : string; l2b : r3}
and  r3 = {l3a : string; l3b : string}
and  r4 = {l4a : string; l4b : string}
and  r5 = {l5a : r1; l5bbbb : r1}
and  r6 = {l6a : r5; l6cc : r5}

RE foo = 'bar'

let _ = let foo = 42 in
        {l6a = {l5a = {l1a = {l4a = "s" ^ "1" ^ "a";
                              l4b = "sab" ^ "1bb" ^ "accc"};
                       l1b = {l2a = "sa" ^ "2" ^ "acc";
                              l2b = {l3a = "saaa" ^ "3bbb" ^ "a";
                                     l3b = "s" ^ "3" ^ "b"}}};
                l5bbbb = {l1a = {l4a = "s" ^ "1" ^ "a";
                                 l4b = "sab" ^ "1bb" ^ "accc"};
                          l1b = {l2a = "sa" ^ "2" ^ "acc";
                                 l2b = {l3a = "saaa" ^ "3bbb" ^ "a";
                                        l3b = "s" ^ "3" ^ "b"}}}};
         l6cc = {l5a = {l1a = {l4a = "s" ^ "1" ^ "a";
                               l4b = "sab" ^ "1bb" ^ "accc"};
                        l1b = {l2a = "sa" ^ "2" ^ "acc";
                               l2b = {l3a = "saaa" ^ "3bbb" ^ "a";
                                      l3b = "s" ^ "3" ^ "b"}}};
                 l5bbbb = {l1a = {l4a = "s" ^ "1" ^ "a";
                                  l4b = "sab" ^ "1bb" ^ "accc"};
                           l1b = {l2a = "sa" ^ "2" ^ "acc";
                                  l2b = {l3a = "saaa" ^ "3bbb" ^ "a";
                                         l3b = "s" ^ ^ "3" ^ "b"}}}}}

[ghost]

I think it's because of the TRY in the grammar, I'll try to debug that

[ghost]

The problem comes from the fact that camlp4 accepts { expr with ... } without parentheses for complicated expr. To parse { something, it tries both the {expr with ...} form and { label = expr; ... } form and then report errors at this level. I tried a hack to always choose the latter form when label = follows {. I'll need to check it doesn't break code in opam