misc/181151: builtins/locale1.0 fails

[Jilles Tjoelker]

The following reply was made to PR misc/181151; it has been noted by GNATS.

From: Jilles Tjoelker <jilles at stack.nl>
To: Garrett Cooper <yaneurabeya at gmail.com>
Cc: bug-followup at FreeBSD.org, Julio Merino <julio at meroh.net>
Subject: Re: misc/181151: builtins/locale1.0 fails
Date: Tue, 21 Jan 2014 23:23:20 +0100

 On Mon, Jan 20, 2014 at 10:37:40AM -0800, Garrett Cooper wrote:
 > On Jan 19, 2014, at 5:27 AM, Jilles Tjoelker <jilles at stack.nl> wrote:
 
 > > On Fri, Jan 17, 2014 at 11:05:28PM -0800, Garrett Cooper wrote:
 > >> On Aug 9, 2013, at 6:06 AM, Jilles Tjoelker <jilles at stack.nl> wrote:
 
 > >>> In PR misc/181151, you wrote:
 > >>>> [sh test builtins/locale1.0 fails]
 
 > >>> I think you have disabled some parts of locale support, since it works
 > >>> fine here. The test uses nl_NL.ISO8859-1 as its non-English locale.
 
 > >> What exactly do I need in order to make this test pass? It’s the one
 > >> test that’s not passing on my CURRENT VM now..
 
 > > You can debug by running the test manually. From bin/sh/tests/, run
 > >  SH=sh sh builtins/locale1.0
 > > or
 > >  SH=sh sh -x builtins/locale1.0
 > > and look at what it does not like.
 
 > Ok, some issues.
 
 > 1. I set SH to /usr/local/bin/bash in .shrc-local, so that’s not going
 > to work with the tests out of the box. The tests need to unset/use a
 > different variable name other than ${SH} IMHO as it’s a standard shell
 > variable.
 
 I don't think it is standard. However, ${SH} should be unconditionally
 overwritten if the testing framework is used. This used to be the case
 in the prove based framework via an extra level of indirection.
 
 On the other hand, passing a value for SH is useful for testing
 (possibly broken) sh changes without installing them. For example,
   SH=${.OBJDIR}/sh sh tests/legacy_test.sh
 This could be useful as a 'regress' target if there is no other way to
 run tests without installing everything. (Note that I would like this to
 work on a stable/9 system.)
 
 > 2. Checking for /var/empty/foo might not be such a great idea. Why not
 > create a temporary file, then delete it?
 
 Because creating a temporary file makes the test more complicated and
 slower. I just need a nonexistent file, and I think I can rely on the
 nonexistence or emptiness of /var/empty. (OpenBSD has a /var/empty/dev,
 but...)
 
 > 3. Other than that, the locale tests don’t seem to be working as
 > expected (it’s acting like LANG=C — output’s available below).
 
 I think you are compiling world WITHOUT_NLS, since there seems to be no
 way to get a message in Dutch language from your sh.
 
 -- 
 Jilles Tjoelker