Re: OpenBSD shell, errexit, and && in if clause

[Alexandre Duret-Lutz]

>>> "RW" == Ralf Wildenhues <address@hidden> writes:

 RW> [ upstream report:
 RW> http://cvs.openbsd.org/cgi-bin/query-pr-wrapper?full=yes&numbers=5093
 RW> ]

 RW> OpenBSD /bin/sh aka /bin/ksh has an interesting bug: the second "false"
 RW> in the script below wrongly causes the shell to exit, but not the first
 RW> one, so "two" isn't printed.

 RW> #! /bin/sh
 RW> set -e
 RW> false && exit 1
 RW> echo one
 RW> if :; then
 RW> false && exit 1
 RW> fi
 RW> echo two

 RW> It doesn't matter if "false" is a builtin here, or replaced by an
 RW> external process that happens to fail.  Same issue with a "for" loop
 RW> (but, surprisingly, not with an "until" loop).

This looks like an issue similar to the one that caused this :

2002-03-17  Alexandre Duret-Lutz  <address@hidden>

        * tests/dirforbid.test: Workaround for NetBSD sh bug.
        Fixes PR automake/305.

http://sources.redhat.com/cgi-bin/cvsweb.cgi/automake/tests/dirforbid.test.diff?cvsroot=automake&r1=1.1&r2=1.2

http://www.netbsd.org/cgi-bin/query-pr-single.pl?number=11542
(I love the suggested "fix" on the initial submission :))

 RW> OK to apply these patches to Autoconf and Automake, respectively?
 RW> Apologies to Alexandre for changing the same lines of depcomp[67].test
 RW> for the third time; at least now I have a good reason to do so.  ;-)

Sure :)
-- 
Alexandre Duret-Lutz

Shared books are happy books.     http://www.bookcrossing.com/friend/gadl