Discrepancy in definition/use of match-data?

[David Kastrup]

We have the following excerpt here:

static Lisp_Object
match_limit (num, beginningp)
     Lisp_Object num;
     int beginningp;
{
  register int n;

  CHECK_NUMBER (num);
  n = XINT (num);
  if (n < 0 || n >= search_regs.num_regs)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    args_out_of_range (num, make_number (search_regs.num_regs));
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  if (search_regs.num_regs <= 0
      || search_regs.start[n] < 0)
    return Qnil;
  return (make_number ((beginningp) ? search_regs.start[n]
                                    : search_regs.end[n]));
}

DEFUN ("match-beginning", Fmatch_beginning, Smatch_beginning, 1, 1, 0,
       doc: /* Return position of start of text matched by last search.
SUBEXP, a number, specifies which parenthesized expression in the last
  regexp.
Value is nil if SUBEXPth pair didn't match, or there were less than
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  SUBEXP pairs.
  ^^^^^^^^^^^^^

Same for match-end and probably match-string.  But if there was no
previous match with at least the same number of SUBEXPs in the
history of the Emacs session, we will get an args_out_of_range error.

I think the fix would be to have match-limit return Qnil instead of
flagging an error for the condition

  n >= search_regs.num_regs

Correct?

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum