bug#5609: 23.1.92; segfault in composition_compute_stop_pos

[YAMAMOTO Mitsuharu <mituharu@math.s.chiba-u.ac.jp>]

>>>>> On Thu, 25 Feb 2010 11:33:22 +0900, Kenichi Handa <handa@m17n.org> said:

> In article <871vgfomwq.fsf@turtle.gmx.de>, Sven Joachim
> <svenjoac@gmx.de> writes:
>> During the last weeks I experienced several segfaults, mostly in
>> Dired, but unfortunately I did not run Emacs under GDB.  This time
>> I did.  I visited a file named "Makefile" and started an I-search
>> for "distcheck" when Emacs segfaulted.  This is not reproducible.

> Although I can't reproduce that bug, I found a suspicious code and
> just fixed it.  Please try the latest version.

Actually I suspected this out-of-boundary `endpos' value in
http://lists.gnu.org/archive/html/bug-gnu-emacs/2010-01/msg00522.html

  I suspect the problematic case is that `charpos + field_width' in
  reseat_to_string exceeds the length of the given string.  The value of
  `field_width' is 12 by default when displaying buffer names.

Also, I guess the upper limit of `endpos' is `SCHARS (it->string)'
instead of `charpos + SCHARS (it->string)' at line 5614 below.  Could
you confirm if it is correct?

  5611	  it->stop_charpos = charpos;
  5612	  if (s == NULL && it->multibyte_p)
  5613	    {
  5614	      EMACS_INT endpos = charpos + SCHARS (it->string);
  5615	      if (endpos > it->end_charpos)
  5616		endpos = it->end_charpos;
  5617	      composition_compute_stop_pos (&it->cmp_it, charpos, -1, endpos,
  5618					    it->string);
  5619	    }

				     YAMAMOTO Mitsuharu
				mituharu@math.s.chiba-u.ac.jp