Re: pkg/48375: etckeeper fails to run with /bin/sh

[Robert Elz <kre%munnari.OZ.AU@localhost>]

The following reply was made to PR pkg/48375; it has been noted by GNATS.

From: Robert Elz <kre%munnari.OZ.AU@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: pkg/48375: etckeeper fails to run with /bin/sh
Date: Sun, 10 Nov 2013 10:45:40 +0700

     Date:        Sun, 10 Nov 2013 01:45:01 +0000 (UTC)
     From:        Aleksej Saushev <asau%inbox.ru@localhost>
     Message-ID:  <20131110014501.74ECBA61B1%mollari.NetBSD.org@localhost>
 
   |  PR bin/48312.
   |  The fix should be pulled into supported versions rather.
 
 That is not a solution to the PR in the Subject.   By all means, fix
 (already done) and pull up (not sure on that) the fix, but pkgsrc neither
 can nor should rely upon that, it needs to work even on older broken
 NetBSDs.
 
 If there was a sh bug on Solaris (well, millions...) or on Darwin, or
 OpenBSD, ... we wouldn't even comtemplate just saying "upgrade to a fixed
 shell", we'd find some kind of a workaround, and incorporate that, one way
 or another (that is, in the bootstrap for that system, or in the package
 Makefile or patches or ...) into pkgsrc, so it works on NetBSD 6.0 as
 promised, not just 6.0++
 
 On the other hand, requiring bash (just using ksh would probably work, and
 not require a dependency, but even that is overboard) for this trivial
 problem is overkill.
 
 If there's a variable that's being unset, without ever being set, just
 install a patch to set the thing before the unset happens, and no more
 problems, that is, if
 
        unset x
 
 is the issue, just make it be
 
        x= ; unset x
 
 and we're done.  (Don't omit the ';' or we get the weird special effects
 of assignments on special builtins in the sh definition).
 
 The fact that this problem has (apparently) been in the shell so long,
 and caused so few problems suggests that this isn't going to require
 zillions of patches all over the place, just the off one, here and there.
 
 kre