bug#41788: 28.0.50; emacsclient -t doesn't work with vipe from moreutils

[Eli Zaretskii]

> Resent-From: Andreas Schwab <schwab@linux-m68k.org>
> Original-Sender: "Debbugs-submit" <debbugs-submit-bounces@debbugs.gnu.org>
> Resent-CC: bug-gnu-emacs@gnu.org
> Resent-Sender: help-debbugs@gnu.org
> Cc: Platon Pronko <platon7pronko@gmail.com>, 41788@debbugs.gnu.org
> From: Andreas Schwab <schwab@linux-m68k.org>
> Date: Mon, 06 Jun 2022 18:25:00 +0200
> 
> On Jun 06 2022, Lars Ingebrigtsen wrote:
> 
> > open(STDIN, "</dev/tty") || die "reopen stdin: $!";
> > open(OUT, ">&STDOUT") || die "save stdout: $!";
> > close STDOUT;
> > open(STDOUT, ">/dev/tty") || die "reopen stdout: $!";
> 
> That associates the process with /dev/tty, not the real terminal.  But
> the running Emacs process usually does not have a controlling terminal,
> thus it cannot open /dev/tty.  It needs to know the real terminal to be
> able to open a tty frame on it.

So you are saying that, instead of using "/dev/tty", vipe should use
the actual name of the terminal device, obtained via the equivalent of
the 'tty' command?  Or is this impossible for what vipe wants to do?