bug#24633: highlight-region func using (window-hscroll) in :align-to spec can cause inf loop

[Eli Zaretskii]

> From: npostavs@users.sourceforge.net
> Cc: 24633@debbugs.gnu.org
> Date: Sat, 08 Oct 2016 15:18:26 -0400
> 
> > Automatic hscrolling is based on length of point's line at the end of
> > redisplaying a window.  Here, "line length" means the horizontal space
> > required for displaying all the glyphs of the line, which includes the
> > glyphs that come from the after-string.  Since you extend your
> > after-string each time the window is hscrolled, you in effect enlarge
> > the line each time the display engine hscrolls the window, which
> > causes an endless loop.
> >
> > Can you restrict the after-string's length so it always ends at the
> > last character of the line, or some fixed number of columns afterward?
> 
> Then it won't reach the edge of the screen when the buffer is scrolled.

Yes, but is it so bad?

> I guess what I would expect is that the automatic hscrolling only goes
> as far as the cursor is rendered, which should be the end of text where
> I put the `cursor' property.

The implementation doesn't actually use the cursor position (it can't:
by the time hscroll_windows runs the cursor is not yet in its final
horizontal position, only its vertical position is correct, because
the horizontal scrolling was not yet done).  The implementation uses a
function that simulates redisplay, and that one doesn't pay attention
to the 'cursor' property.  So it stops only when the after-string is
exhausted.

> If you think this would be too complex/slow to implement, we can
> mark this bug as wontfix.

It'd be messy, yes.  We'd need to search for before- and after-strings
that begin/end right at point, then augment the call to the display
emulating function accordingly, deal with complications like strings
that include newlines, etc.

> >> >  In particular, I don't
> >> > understand the align-to expression: e.g., window-hscroll returns its
> >> > value in columns, while align-to needs pixels, AFAIU.
> >> 
> >> According to `(elisp) Pixel Specification',
> >> 
> >>        The form NUM specifies a fraction of the default frame font height
> >>     or width.  The form `(NUM)' specifies an absolute number of pixels.
> >
> > I admire your courage in reading that documentation and then writing
> > stuff like the above, which the documentation doesn't mention even
> > remotely.
> 
> Uh, not sure how to read this, is it irony?

Only a little.  I find this area severely under-documented.