bug#66128: 28.2; visible-bell breaks setterm --inversescreen on

[Eli Zaretskii <eliz@gnu.org>]

> From: Tomas Hlavaty <tom@logand.com>
> Cc: 66128@debbugs.gnu.org
> Date: Fri, 22 Sep 2023 22:18:14 +0200
> 
> On Thu 21 Sep 2023 at 08:00, Eli Zaretskii <eliz@gnu.org> wrote:
> > Emacs just sends to the terminal the sequence defined by the "vb"
> > termcap capability.  Is that not correct when the screen colors are
> > inverted?
> 
> I do not know.
> 
> > Is this perhaps the problem of the terminal?
> 
> No, I get the same behaviour in xfce4-terminal and kitty, so this does
> not seem to be specific problem with the linux console.
> 
> I think I wrote it in the bug report too, under an X based terminal:
> 
>    $ emacs -nw -Q --eval '(setq visible-bell t)'
> 
> then press PgUp.
> 
> > Btw, we always use the termcap's "vb", even when terminfo is
> > available; should we use the terminfo's "flash" instead?
> 
> I do not know.

Thomas, could you perhaps help us out here?  This is about sending the
"visible bell" sequence to a terminal after "setterm --inversescreen on".
The original report is:

    $ setterm --inversescreen on
    $ emacs -Q --eval '(setq visible-bell t)'

  Then in Emacs do something that causes a bell, like try moving beyond
  the buffer's end.  This causes the Emacs background to become white,
  i.e. the visible-bell somehow countermands the inversescreen state.

Can you think of any reason for this behavior?  Do terminals honor
inversescreen when they perform the visible-bell function?  Emacs just
sends the sequence reported by the "vb" termcap capability of the
terminal when the visible-bell is triggered.

Thanks.