Re: [PATCH 12/10] support pager.* for external commands

From: Jeff King <peff@peff.net>
Date: Feb 14, 2012, 19:13 UTC
Message-ID: <20120214191340.GC12072@sigill.intra.peff.net>
In-Reply-To: <CACBZZX596wnk2KE9QzUPMc=A6Mt8HbUs7F4rnAZbw1_RrcKHnw@mail.gmail.com>

On Sun, Feb 12, 2012 at 01:46:34AM +0100, Ævar Arnfjörð Bjarmason wrote:

Show 24 quoted lines

> On Fri, Aug 19, 2011 at 00:01, Jeff King <peff@peff.net> wrote:
> 
> > +test_expect_success TTY 'command-specific pager works for external commands' '
> > +       sane_unset PAGER GIT_PAGER &&
> > +       echo "foo:initial" >expect &&
> > +       >actual &&
> > +       test_config pager.external "sed s/^/foo:/ >actual" &&
> > +       test_terminal git --exec-path="`pwd`" external log --format=%s -1 &&
> > +       test_cmp expect actual
> 
> For reasons that I haven't looked into using sed like that breaks
> under /usr/bin/ksh on Solaris. Just using:
> 
>     sed -e \"s/^/foo:/\"
> 
> Instead fixes it, it's not broken with /usr/xpg4/bin/sh, so it's some
> ksh peculiarity.
> 
> The error it gives is:
> 
>     sed s/^/foo:/ >actual: Not found
> 
> Indicating that for some reason it's considering that whole "sed
> s/^/foo:/ >actual" string to be a single command.

Hrm. Is the problem on the git-executing side, or is it on the setting up the config side?

Sadly (or perhaps not) I no longer have any Solaris machines to test on. Can you confirm that "git config pager.external" looks OK inside that test? Can you confirm via GIT_TRACE=1 what is being sent to the shell?

Also, it looks like we actually run commands internally from git using "sh -c". So if it is the executing side that is wrong, I don't see how /usr/bin/ksh would be involved at all (it would either be /bin/sh, or /usr/xpg4/bin/sh if you have your PATH set).

-Peff