Original bug ID: 420
Reporter: administrator
Status: closed (set by @damiendoligez on 2005-12-15T14:49:26Z)
Resolution: won't fix
Priority: normal
Severity: minor
Category: ~DO NOT USE (was: OCaml general)
Related to: #2327

Bug description

Full_Name: Ørjan Johansen
Version: ocaml-3.01-2
OS: linux
Submission from: fimfpc67.math.ntnu.no (129.241.15.112)

Dear Caml developers,

I have had certain problems using/understanding ocamlyacc, which I
now believe have to do with an inconsistency in whether a parsed
expression needs to be followed by EOF or not.

I noticed a similar bug report (id 4) by Prof. Courant, which,
although from December 1999, is still open.

I did some further investigations, which I hope may be of help in
correcting the bug.

The problem seems to occur when the parser is able to reduce an
expression to the goal %entry% without considering the lookahead.
This reduction causes a semantic action raising the YYexit exception,
and thus finishing parsing.

This is a slight difference from what the *.output file would imply,
which is that one should then pass to the final state, containing the
action

$end accept

I suspect that it is the lack of this final test for $end=EOF that is
the cause of the bug.

I have tried to think of a fix, but I have not been able to understand
the C code. My best guess would be to modify the automaton so that
the final state exists and tests for $end.

Otherwise, I would like to add my voice to those who would like
ocamlyacc to allow the passing of an extra argument to semantic
actions.

Yours sincerely,
Ørjan Johansen.