On 2020-01-26 at 22:08:18, Johannes Schindelin wrote:
> > diff --git a/t/t4013-diff-various.sh b/t/t4013-diff-various.sh
> > index 5ac94b390d..6f5f05c3a8 100755
> > --- a/t/t4013-diff-various.sh
> > +++ b/t/t4013-diff-various.sh
> > @@ -120,6 +120,30 @@ test_expect_success setup '
> >  +*++ [initial] Initial
> >  EOF
> >
> > +process_diffs () {
> > +	x04="[0-9a-f][0-9a-f][0-9a-f][0-9a-f]" &&
> > +	x07="$_x05[0-9a-f][0-9a-f]" &&
> 
> Any reason not to stay with the convention, i.e. using `_x04` and `_x07`
> here (with leading underscores)?

None in particular.  I have a slight bias against initial underscores
from C, where that has a specific meaning, but I agree consistency is
good, so I'll make that change.

> > +	sed -e "s/$OID_REGEX/$ZERO_OID/g" \
> > +	    -e "s/From $_x40 /From $ZERO_OID /" \
> > +	    -e "s/from $_x40)/from $ZERO_OID)/" \
> > +	    -e "s/commit $_x40\$/commit $ZERO_OID/" \
> > +	    -e "s/commit $_x40 (/commit $ZERO_OID (/" \
> > +	    -e "s/$_x40 $_x40 $_x40/$ZERO_OID $ZERO_OID $ZERO_OID/" \
> > +	    -e "s/$_x40 $_x40 /$ZERO_OID $ZERO_OID /" \
> > +	    -e "s/^$_x40 $_x40$/$ZERO_OID $ZERO_OID/" \
> > +	    -e "s/^$_x40 /$ZERO_OID /" \
> > +	    -e "s/^$_x40$/$ZERO_OID/" \
> > +	    -e "s/$x07\.\.$x07/fffffff..fffffff/g" \
> > +	    -e "s/$x07,$x07\.\.$x07/fffffff,fffffff..fffffff/g" \
> > +	    -e "s/$x07 $x07 $x07/fffffff fffffff fffffff/g" \
> > +	    -e "s/$x07 $x07 /fffffff fffffff /g" \
> > +	    -e "s/Merge: $x07 $x07/Merge: fffffff fffffff/g" \
> > +	    -e "s/$x07\.\.\./fffffff.../g" \
> > +	    -e "s/ $x04\.\.\./ ffff.../g" \
> > +	    -e "s/ $x04/ ffff/g" \
> > +	    "$1"
> > +}
> > +
> >  V=$(git version | sed -e 's/^git version //' -e 's/\./\\./g')
> >  while read magic cmd
> >  do
> > @@ -158,13 +182,15 @@ do
> >  		} >"$actual" &&
> >  		if test -f "$expect"
> >  		then
> > +			process_diffs "$actual" >actual &&
> > +			process_diffs "$expect" >expect &&
> >  			case $cmd in
> >  			*format-patch* | *-stat*)
> > -				test_i18ncmp "$expect" "$actual";;
> > +				test_i18ncmp expect actual;;
> >  			*)
> > -				test_cmp "$expect" "$actual";;
> > +				test_cmp expect actual;;
> >  			esac &&
> > -			rm -f "$actual"
> > +			rm -f "$actual" actual expect
> >  		else
> >  			# this is to help developing new tests.
> >  			cp "$actual" "$expect"
> > @@ -383,16 +409,22 @@ test_expect_success 'log -S requires an argument' '
> >  test_expect_success 'diff --cached on unborn branch' '
> >  	echo ref: refs/heads/unborn >.git/HEAD &&
> >  	git diff --cached >result &&
> > -	test_cmp "$TEST_DIRECTORY/t4013/diff.diff_--cached" result
> > +	process_diffs result >actual &&
> > +	process_diffs "$TEST_DIRECTORY/t4013/diff.diff_--cached" >expected &&
> 
> I was about to suggest letting `process_diffs` work in-place, but this
> line makes that idea moot.

If I could have done that, I would.

> Another idea I had was to implement a `test_cmp_diff` that processes the
> diffs and then compares them, but I guess that would be _less_ concise
> than this patch.

Yeah, this is a tricky test to work with because it does so many
different things and trying to handle all of them in a tidy way is hard
(as one can intuit from the giant sed statement).  As part of the patch
you quoted above, we sometimes use test_i18ncmp and sometimes use
test_cmp here, so it's not easy hard to pick something that works for
all of these cases without some duplication.
-- 
brian m. carlson: Houston, Texas, US
OpenPGP: https://keybase.io/bk2204