On Wed, Sep 22, 2010 at 17:59, Pat Notz <patnotz@xxxxxxxxx> wrote: > On Tue, Sep 21, 2010 at 2:36 PM, Ãvar ArnfjÃrà Bjarmason > <avarab@xxxxxxxxx> wrote: >> >> On Tue, Sep 21, 2010 at 20:25, Pat Notz <patnotz@xxxxxxxxx> wrote: >> >> > +cat >editor <<\EOF >> > +#!/bin/sh >> > +sed -e "s/intermediate/edited/g" <"$1" >"$1-" >> > +mv "$1-" "$1" >> > +EOF >> > +chmod 755 editor >> > + >> > +test_expect_success 'commit --squash works with -c' ' >> > +    commit_for_rebase_autosquash_setup && >> > +    EDITOR=./editor git commit --squash HEAD~1 -c HEAD && >> > +    commit_msg_is "squash! target message subject lineedited commit" >> > +' >> >> Why not put the editor in t/t7500/ and use test_set_editor() like the >> other tests? > > The real reason is that I'm new enough that I wasn't aware of this > pattern. ÂI saw what was done in t7501-commit.sh and followed along. > I missed the use of test_set_editor() right there in t7500-commit.sh. > Doh! > > I can certainly do that if it's preferred. ÂI must say, though, that I > find it odd to put test inputs in a separate file in a separate > directory from where the test transforms those into expected outputs. > To see what the test is doing you have to load both files and trace > through it. > > Still, I'd be happy to change do this if that's the preferred way. It's a bit odd, but it's best to following existing style within a test. Then maybe submit fixup patches to fix the whole thing later. -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html