Strange code in find-composition-internal

Strange code in find-composition-internal

[Eli Zaretskii]

This is a fragment from find-composition-internal that attempts to
find a composition, either static (i.e. defined by a `composition'
text property) or automatic, near the position POS:

  if (!find_composition (from, to, &start, &end, &prop, string))
    {
      if (!NILP (current_buffer->enable_multibyte_characters)
	  && ! NILP (Vauto_composition_mode)
	  && find_automatic_composition (from, to, &start, &end, &gstring,
					 string))
	return list3 (make_number (start), make_number (end), gstring);
      return Qnil;
    }
  if ((end <= XINT (pos) || start > XINT (pos)))
    {
      EMACS_INT s, e;

      if (find_automatic_composition (from, to, &s, &e, &gstring, string)
	  && (e <= XINT (pos) ? e > end : s < start))
	return list3 (make_number (start), make_number (end), gstring);
    }

The doc string of find-composition says, among other things:

    When Automatic Compostion mode is on, this function also finds a
    chunk of text that is automatically composed.  If such a chunk is
    found closer to POS than the position that has `composition'
    property, the value is a list of FROM, TO, and a glyph gstring
    the specify how the chunk is composed.

But the code fragment above seems to return `start' and `end' from the
static composition even if find_automatic_composition finds an
automatic composition that is closer to POS, which is contrary to the
doc.

Shouldn't the last line be

	return list3 (make_number (s), make_number (e), gstring);

?  Or am I missing something?

Re: Strange code in find-composition-internal

[Kenichi Handa]

In article <838w8c25hz.fsf@gnu.org>, Eli Zaretskii <eliz@gnu.org> writes:

> This is a fragment from find-composition-internal that attempts to
> find a composition, either static (i.e. defined by a `composition'
> text property) or automatic, near the position POS:

>   if (!find_composition (from, to, &start, &end, &prop, string))
>     {
>       if (!NILP (current_buffer->enable_multibyte_characters)
> 	  && ! NILP (Vauto_composition_mode)
> 	  && find_automatic_composition (from, to, &start, &end, &gstring,
> 					 string))
> 	return list3 (make_number (start), make_number (end), gstring);
>       return Qnil;
>     }
>   if ((end <= XINT (pos) || start > XINT (pos)))
>     {
>       EMACS_INT s, e;

>       if (find_automatic_composition (from, to, &s, &e, &gstring, string)
> 	  && (e <= XINT (pos) ? e > end : s < start))
> 	return list3 (make_number (start), make_number (end), gstring);
>     }
[...]
> Shouldn't the last line be

> 	return list3 (make_number (s), make_number (e), gstring);

You are right.  I've just committed a fix for the trunk.

---
Kenichi Handa
handa@m17n.org