From: Jeff King <peff@peff.net>
Date: Wed, 20 Nov 2013 18:54:20 GMT
Subject: Re: [PATCH] t7507-*.sh: Fix test #8 (could not run '"$FAKE_EDITOR"')
Message-ID: <20131120185420.GA11404@sigill.intra.peff.net>
In-Reply-To: <xmqqhab67wjr.fsf@gitster.dls.corp.google.com>

On Wed, Nov 20, 2013 at 10:33:28AM -0800, Junio C Hamano wrote:

> Jeff King <peff@peff.net> writes:
> 
> > Am I misremembering the issues with one-shot variables and functions?
> 
> I think there are two problems involved.

OK, I was misremembering. I recalled the "does not unset afterwards"
part, but not the "does not export" part. I think because:

> test_must_fail () {
> 	(
> 		env | sed -n -e '/EDITOR/s/^/>> /p'
> 	)
> }

...here we _do_ have GIT_EDITOR set properly in the function itself, but
not in the subprocess.

Previous discussion and links to POSIX are here:

  http://article.gmane.org/gmane.comp.version-control.git/137095

Not that they matter compared to the code you demonstrated, but I was
digging them up when you responded. :)

> Another is that EDITOR="$FAKE_EDITOR" that is set up earlier in the
> is having trouble launching (I have a feeling that it never was
> actually used because everybody uses "commit -F <file>").

I think it is used, as there are several "git commit --amend -v"
invocations. Which makes sense, as you should not be able to test "-v"
with "-F", I would think.

I'm not sure why the old $FAKE_EDITOR doesn't work there, though (not
that it would make the test pass anyway, as it does something different
than what the test wants, but I would not expect the shell to complain
of failure).

-Peff