Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!munnari!uqcspe!tony
From: tony@uqcspe.OZ (Tony O'Hagan)
Newsgroups: comp.unix.wizards
Subject: Re: sh(1) traps & exit status
Message-ID: <1652@uqcspe.OZ>
Date: Wed, 4-Nov-87 22:14:00 EST
Article-I.D.: uqcspe.1652
Posted: Wed Nov  4 22:14:00 1987
Date-Received: Sun, 8-Nov-87 01:48:30 EST
References: <1639@uqcspe.OZ> <20@grand.sm.unisys.com>
Reply-To: tony@uqcspe.oz (Tony O'Hagan)
Organization: Computer Science, Queensland Uni, Australia
Lines: 37

In article <20@grand.sm.unisys.com> you write:
>In article <1639@uqcspe.OZ> tony@uqcspe.oz (Tony O'Hagan) writes:
>|I hate having to assign the exit status to a variable before every exit.
>|This doesn't work:
>|	trap 'ts=$?; rm -fr $tdir; exit $ts' 0 1 2 3 15
>
>You are springing the trap twice.  Use this instead:
>
>	trap exit 1 2 3 15
>	trap 'ts=$?; rm -fr $tdir; exit $ts' 0

Thanks for you thoughts ... 
I realised that interupts cause the trap to be done twice,
I guess I'm just lazy when it's only doing "rm -f ..." .
Although when it does matter, your solution
is cleaner than turning off traps within the trap !!

What I'm really concerned about is that "exit 2" does not set $? for the trap.

I often have something like :-
	trap 'rm -fr $tdir; exit $ts' 0 1 2 3 15
	[ -s "$1" ] || { echo "$1 is empty" >&2; ts=2; exit 2; }

One solution mailed to me by Robert Elz [kre@monet.berkeley.edu]  was
	trap 'ts=$?; rm -fr $tdir; exit $ts' 0 1 2 3 15
	[ -s "$1" ] || { echo "$1 is empty" >&2; (exit 2); exit; }
which could also use the exit status of "mkdir" in something like :-
	mkdir $2 || exit

I'd appeciate any further ideas,
	Tony O'Hagan.

Computer Science Department,	ACSnet:		tony@uqcspe.oz
University of Queensland,	UUCP:		..!uunet!munnari!uqcspe.oz!tony
St Lucia, Brisbane, 4067	JANET:		uqcspe.oz!tony@ukc
Australia.			Internet:	tony%uqcspe.oz@uunet.uu.net
Phone: +61 7 377 4125		Smart:		tony@uqcspe.oz.au