Re: [miles@lsi.nec.co.jp: redisplay abort if GLYPH_DEBUG turned on]

[Richard Stallman <rms@gnu.org>]

    > There are other places which set stop_charpos to 0, so
    > there is IT_CHARPOS <= stop_charpos is not an invariant.

    The thing that makes me nervous is that there seem to also be other
    similar xasserts, e.g. [xdisp.c:2727]:

It's good to be nervous when you see something like that.

Looking at the places that set stop_charpos to 0, I see that they
usually also set it->string to a string.  Here's one possible
exception in reseat_to_string,

  else
    {
      it->s = s;
      it->string = Qnil;

but it sets it->s instead.

The only other exception is the newly added code in set_iterator_to_next.
So I think that setting stop_charpos to 0 is not the right way to
do the job there.