Re: [PATCH] Allow shell scripts to run with non-Bash /bin/sh

From: Eygene Ryabinkin <rea-git@codelabs.ru>
Date: Sep 23, 2007, 08:31 UTC
Message-ID: <20070923083118.GB99140@void.codelabs.ru>
In-Reply-To: <7vtzpnf6c9.fsf@gitster.siamese.dyndns.org>

Junio, *, good day.

Sat, Sep 22, 2007 at 01:32:38AM -0700, Junio C Hamano wrote:

Show 13 quoted lines

> > OK, you're right.  Especially if /bin/sh from Solaris and OpenBSD
> > are working and they are not Bash.  But I would not tell that
> > the shell is broken now -- I had not seen the POSIX specification.
> > Does it specifies how the shell should work in this case?
> 
> I have always been assuming it to be the case (this construct is
> not my invention but is an old school idiom I just inherited
> from my mentor) and never looked at the spec recently, but I
> re-read it just to make sure.  The answer is yes.
> 
> Visit http://www.opengroup.org/onlinepubs/000095399/ and follow
> "Shell and Utilities volume (XCU)" and then "Case conditional
> construct".

Yes, thanks for the pointer.

Show 10 quoted lines

> So, as David suggests, if
> 
>         false
>         case Ultra in
>         Super) false ;;
>         Hyper) true ;;
>         esac && echo case returned ok
> 
> does not say "case returned ok", then the shell has a bit of
> problem.

Correct: the current /bin/sh for FreeBSD does not set zero exit
code if no case patterns were matched.  So, I apologize for my quick
decision on the non-brokenness of the /bin/sh -- it is broken.

I had fixed the shell and filed the problem report. May be the change will be incorporated into the future release of FreeBSD. Meanwhile, I had added workarounds to the other places Junio mentioned in his follow-up and will try to push this patch to the FreeBSD port of Git. The explanation had been changed too ;))

Thanks to all people who helped me to realize what is wrong and where!

-- 
Eygene