On Wed, 10 Dec 2003, Robert Elz wrote:

> "Set -e" (which I half remember actually predates
> && and stuff in shells) is really intended to allow scripts comprised of
> mostly a long list of commands to abort when one of the commands fails.
> It isn't really the way that any reasonable shell program should operate
> (and that includes shell programs generated by makefiles).

Having applied my brain, I think you're right -- it's impossible for "-e"
to do the right thing at all times.  The problem is that a non-zero exit
status can mean two things: "false" and "error".  A perfect "-e" would
kill the shell on the second, but not the first, but it's obviously
impossible to reliably tell the difference between them.

>   | or make "(false && true)" behave differently from "false && true"
>   | (historical behaviour, exemplified by ksh).
>
> AT&T ksh perhaps - not the pdksh that NetBSD ships as ksh, that one
> agrees with NetBSD's sh and with bash (and posix).

Yes, I meant AT&T ksh (as shipped with Solaris 9).

-- 
Ben Harris                                                   <bjh21@NetBSD.org>
Portmaster, NetBSD/acorn26           <URL:http://www.NetBSD.org/Ports/acorn26/>