Re: interactive shell detection in shrc

At Wed, 09 Oct 2024 09:57:31 +0700, Robert Elz <[email protected]> wrote:
Subject: Re: interactive shell detection in shrc
>
> But if the idea is to use just that to determine if a shell
> is interactive, as in:
>
>   | When I decided this is what interactive means
>   | all of the edge cases went away. Is there a
>   | case I am not considering?
>
> what that might be intending to mean, then no, that won't
> work at all.
>
> Run any script (from any interactive shell, as a command line) as
>
> 	sh script
> 		(or just ./script if it has 'x' perms & perhaps a #!)
>
> and in the script if you do "test -t 0" you'll get "true" - in that
> script, standard input is a terminal, the same one as the shell from
> which the script was run from.   Yet the shell running the script is
> certainly not interactive.   To be interactive, then as well as stderr
> also needing to be a terminal, the shell needs to be reading its commands
> from that standard input (being a terminal).   There's no shell operator
> or function to tell you which file descriptor the shell is reading its
> command input from (none of the script's business.)

Indeed.

I've given up on testing for TTYs and the "test $0 = $SHELL" bit as I've
realized that for shells I care about that'll be true for shell
fragments run from system(3) too.  Only the ksh's (pdksh and
derivatives, as well as AT&T Ksh) and Bash put 'c' in $- when passed
'-c' on the command line, which would be the only other sure way to
test, and even with the 'c' they still leave $#=0.  None of the true
Bourne shells or derivatives, nor Ash and its derivatives add the 'c'.

Anyway it's pretty academic and meaningless in the real world as $ENV is
pretty much only supported by shells which do always add 'i' to "$-"
whenever they are interactive (Ash has done so since its initial 1989
release on comp.sources.unix), and furthermore only a very few shells
source $ENV when they are not interactive (and thus they don't need any
interactivity test at all).

Even AT&T Ksh gave up on always expanding and sourcing $ENV with the
original ksh93 release (i.e. as of 1993/12/28), so it is always POSIX
compliant now.  (I hadn't realized that until I read the manual more
carefully and tested it more explicitly.)

--
					Greg A. Woods <[email protected]>

Kelowna, BC     +1 250 762-7675           RoboHack <[email protected]>
Planix, Inc. <[email protected]>     Avoncote Farms <[email protected]>

-----BEGIN PGP SIGNATURE-----

iF0EABECAB0WIQRuK6dmwVAucmRxuh9mfXG3eL/0fwUCZwdasQAKCRBmfXG3eL/0
f5ajAKCtq6FDREdZsZPS86DWcU6aZLlggQCg84P3osi9mM6xcvItFXnne+GlGOI=
=xNVt
-----END PGP SIGNATURE-----