46.6 Watch Out for Bourne Shell -e Bug

The Bourne shell -e option should stop execution when a command returns a non-zero status. Does your -e option seem to cause some if commands to abort scripts? If so, you have a copy of the Buggy Bourne Shell, as distributed with 4.2BSD, 4.3BSD, and probably several other systems. It can be identified by running:

$ 

set -e




$ 

if false; then echo yipe; else echo ok; fi

and noting that the shell exits instead of printing ok , and by:

||
 

$ 

set -e


$ 

false || echo ok

which also exits and should not, and by:

$ 

set -e


$ 

while false; do :; done

To fix it, first get the source, and then change it in the obvious three places in xec.c . You will have to learn Bournegol [the ALGOL-like dialect of C that Steve Bourne used to write the original Bourne shell -JP  ]. Another alternative is to replace /bin/sh with one of the free sh look-alikes ( 1.8 ) , provided you can find one that is enough alike.

As a workaround, you can set +e around all the tests that might fail. Unfortunately, some versions of the Buggy Bourne Shell do not even support set +e ; here the only workaround is to run a subshell ( 38.4 ) without the -e flag.