Re: the GPG_TTY line in bashrc produces error message in chroot environment

On Sun, 2025-10-19 at 23:02 +0800, Xi Ruoyao wrote:
> On Sun, 2025-10-19 at 16:39 +0800, Zhang Wen wrote:
> > Xi Ruoyao <[email protected]> 于2025年10月18日周六 19:00写道：
> > > 
> > > On Sat, 2025-10-18 at 12:17 +0800, Zhang Wen wrote:
> > > > The root cause is while the tty command returns non-zero in the chroot
> > > > environment, the 'tty -s' command returns a zero value.
> > > > I changed 'tty -s' to 'tty > /dev/null 2>&1' and the error message
> > > > disappeared, not sure if there's a better solution.
> > > 
> > > I'd say it's a coreutils bug and it should be reported to the coreutils
> > > developers.  "tty -s" should have the same exit status as "tty".
> > > 
> > > --
> > > Xi Ruoyao <[email protected]>
> > > --
> > > http://lists.linuxfromscratch.org/sympa/info/blfs-dev
> > > Unsubscribe: See the above information page
> > 
> > It's not necessarily the truth, there's a code structure overview link
> > below the the git line on https://www.gnu.org/software/coreutils:
> > https://www.maizure.org/projects/decoded-gnu-coreutils/tty.html
> > It says that the silent mode of tty only return result of the isatty
> > function in the c library, while the normal mode return result of the
> > ttyname one.
> 
> It's not relevant.  The info page says:
> 
> ‘-s’
> ‘--silent’
> ‘--quiet’
>      Print nothing; only return an exit status.
> 
>    Exit status:
> 
>      0 if standard input is a terminal
>      1 if standard input is a non-terminal file
>      2 if given incorrect arguments
>      3 if a write error occurs
>      4 if the terminal's name cannot be determined
> 
> It's not related to how the code is structured.  It's simply the code
> isn't doing what the document says.  If this wasn't a bug, it would mean
> "the terminal's name can be determined with -s, but it cannot without -
> s."  That would be just ridiculous.
> 
> Either the code should be fixed to ensure the exit status consistent
> with and without -s, or the documentation should be updated to document
> the current behavior.  In either case we've encountered a bug here.

I've updated the book to use &>/dev/null instead of -s.  But still I
think coreutils should be fixed.

-- 
Xi Ruoyao <[email protected]>

-- 
http://lists.linuxfromscratch.org/sympa/info/blfs-dev
Unsubscribe: See the above information page