bug#54450: 29.0.50; redisplay--pre-redisplay-functions issue

[Ergus]

On Fri, Mar 18, 2022 at 07:01:06PM +0200, Eli Zaretskii wrote:
> > Date: Fri, 18 Mar 2022 16:38:45 +0100
> > From:  Ergus via "Bug reports for GNU Emacs,
> >  the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>
> >
> > emacs -Q
> > (setq highlight-nonselected-windows t)
> >
> > C-x 2
> > C-x o
> > C-x C-f /tmp/test
> > (insert some text)
> > C-a C-SPC C-f C-f C-f
> > C-x o
> >
> > [at this point we should see a region highlighted over the 3 first
> > letters of the inserted text in the other winfow buffer (test)]
> >
> > then eval this several times:
> >
> > (with-selected-window (get-buffer-window "test" 0)
> >     (right-char))
> >
> > This will move the point in the test window, but the region won't be
> > updated.
>
> Thanks.  Does the patch below solve the problem?
>
> diff --git a/src/intervals.c b/src/intervals.c
> index 687b237..cbcc100 100644
> --- a/src/intervals.c
> +++ b/src/intervals.c
> @@ -44,6 +44,7 @@
> #include "lisp.h"
> #include "intervals.h"
> #include "buffer.h"
> +#include "window.h"
> #include "puresize.h"
> #include "keymap.h"
>
> @@ -1885,6 +1886,11 @@ set_point_both (ptrdiff_t charpos, ptrdiff_t bytepos)
>       return;
>     }
>
> +  if (highlight_nonselected_windows
> +      && !NILP (Vtransient_mark_mode)
> +      && !NILP (BVAR (current_buffer, mark_active)))
> +    bset_redisplay (current_buffer);
> +
>   /* Set TO to the interval containing the char after CHARPOS,
>      and TOPREV to the interval containing the char before CHARPOS.
>      Either one may be null.  They may be equal.  */

Hi Eli:

Yes, this could solve the issue for the region, but not in general for
redisplay--update-cursor-face-highlight which was the initial problem
with Juri's code. Same for any other function that goes into the
pre-redisplay-functions hook.

The region example reported in this issue was only the simplest way I
found to give you a simple reproducible snippet, but the issue is a bit
more general.

With the cursor-face we may have a similar issue cause such hook works
in the same way, so this condition will need to be extended and extended
and extended on every case for every function in
pre-redisplay-functions hook.

else

For Juri's code maybe there is a way that he can explicitly do
bset_redisplay at the end of the with-selected-window?

Does it makes sense?

Best,
Ergus