[comp.unix.questions] temporary redirection under a Bourne-shell using "exec"

eh@ign.UUCP (eh@phenix pour news) (09/06/90)

Hi,

Here is a small trick in Bourne Shell, concerning temporary redirections
(and correct restoration) without entering a sub-shell.

In a shell-script (Sys V.3, Bourne shell), I wanted to temporarly
desactivate "stdout" without entering a sub-shell, because I assigned
variables in it.
The order to close stdout for the given shell is "exec >&-", but once it is
closed, it is lost till the end of the shell, unless you save it in a brand new
File Descriptor.
This is an application of the "exec" command.

Here is a short example :

exec 3<&1 >&-	# open File Descriptor 3 (stdout), and close stdout
echo invisible	# shell stdout is closed, local stdout is FD 3
exec 1<&3 3>&-	# reopen stdout (and close FD 3)
echo hello

In this example, you assign "stdout" to FD 3 and you close FD 1 (stdout of the
shell).

When I run this shell-script (toto), "hello" appears.
When I run toto >/dev/null, nothing appears, which proves that "stdout" is
correctly restored.

Of course, you can use files for temporary redirection, or devices...

If you want to desactivate the echo on the console (e.g. reading a password),
you can use "stty -echo" :
echo "Password: \c"
stty -echo
read word
stty echo
# then crypt it and compare it to the crypted entry


Yours,
Eric

---------------------------------------------------------------------------
Eric HURTEBIS (Institut Geographique National - France)
Net: eh@ign.uucp
Tel: +33 1 43 98 80 00 ext.7366
---------------------------------------------------------------------------

martin@mwtech.UUCP (Martin Weitzel) (09/07/90)

In article <977@ign.UUCP> eh@ign.UUCP (eh@phenix pour news) writes:
[about temporarily supressing outout to stdout in shell scripts]
>Here is a short example :
>
>exec 3<&1 >&-	# open File Descriptor 3 (stdout), and close stdout
>echo invisible	# shell stdout is closed, local stdout is FD 3
>exec 1<&3 3>&-	# reopen stdout (and close FD 3)
>echo hello

I would rather change this as follows:

>exec 3>&1 >/dev/null
>echo invisible	# shell stdout is closed, local stdout is FD 3
>exec 1>&3 3>&-	# reopen stdout (and close FD 3)
>echo hello

The original example is fine so far, except that a output-channel should
stay a output-channel (though duping to an input-channel mostly works, as
the example demonstrates) and I would not CLOSE stdout (...>&-...) but
REDIRECT it to /dev/null in the first exec.

Maybe many programs don't care for error returns from their output
operations, but *some* may do (and IMHO these are the ones written
by the more careful programmers). Such a program may assume that a
severe error occured if sending something to stdout returns an error
and terminate in an unexpected way.
-- 
Martin Weitzel, email: martin@mwtech.UUCP, voice: 49-(0)6151-6 56 83