8 files changed, 420 insertions, 22 deletions

diff --git a/contrib/diff-highlight/.gitignore b/contrib/diff-highlight/.gitignore
new file mode 100644
index 0000000000..c07454824e
--- /dev/null
+++ b/contrib/diff-highlight/.gitignore
@@ -0,0 +1,2 @@
+shebang.perl
+diff-highlight
diff --git a/contrib/diff-highlight/diff-highlight b/contrib/diff-highlight/DiffHighlight.pm
index ffefc31a98..663992e530 100755..100644
--- a/contrib/diff-highlight/diff-highlight
+++ b/contrib/diff-highlight/DiffHighlight.pm
@@ -1,4 +1,4 @@
-#!/usr/bin/perl
+package DiffHighlight;
 
 use 5.008;
 use warnings FATAL => 'all';
@@ -21,23 +21,28 @@ my $RESET = "\x1b[m";
 my $COLOR = qr/\x1b\[[0-9;]*m/;
 my $BORING = qr/$COLOR|\s/;
 
+# The patch portion of git log -p --graph should only ever have preceding | and
+# not / or \ as merge history only shows up on the commit line.
+my $GRAPH = qr/$COLOR?\|$COLOR?\s+/;
+
 my @removed;
 my @added;
 my $in_hunk;
 
-# Some scripts may not realize that SIGPIPE is being ignored when launching the
-# pager--for instance scripts written in Python.
-$SIG{PIPE} = 'DEFAULT';
+our $line_cb = sub { print @_ };
+our $flush_cb = sub { local $| = 1 };
+
+sub handle_line {
+	local $_ = shift;
 
-while (<>) {
 	if (!$in_hunk) {
-		print;
-		$in_hunk = /^$COLOR*\@/;
+		$line_cb->($_);
+		$in_hunk = /^$GRAPH*$COLOR*\@\@ /;
 	}
-	elsif (/^$COLOR*-/) {
+	elsif (/^$GRAPH*$COLOR*-/) {
 		push @removed, $_;
 	}
-	elsif (/^$COLOR*\+/) {
+	elsif (/^$GRAPH*$COLOR*\+/) {
 		push @added, $_;
 	}
 	else {
@@ -45,8 +50,8 @@ while (<>) {
 		@removed = ();
 		@added = ();
 
-		print;
-		$in_hunk = /^$COLOR*[\@ ]/;
+		$line_cb->($_);
+		$in_hunk = /^$GRAPH*$COLOR*[\@ ]/;
 	}
 
 	# Most of the time there is enough output to keep things streaming,
@@ -58,15 +63,22 @@ while (<>) {
 	# place to flush. Flushing on a blank line is a heuristic that
 	# happens to match git-log output.
 	if (!length) {
-		local $| = 1;
+		$flush_cb->();
 	}
 }
 
-# Flush any queued hunk (this can happen when there is no trailing context in
-# the final diff of the input).
-show_hunk(\@removed, \@added);
+sub flush {
+	# Flush any queued hunk (this can happen when there is no trailing
+	# context in the final diff of the input).
+	show_hunk(\@removed, \@added);
+}
 
-exit 0;
+sub highlight_stdin {
+	while (<STDIN>) {
+		handle_line($_);
+	}
+	flush();
+}
 
 # Ideally we would feed the default as a human-readable color to
 # git-config as the fallback value. But diff-highlight does
@@ -84,7 +96,7 @@ sub show_hunk {
 
 	# If one side is empty, then there is nothing to compare or highlight.
 	if (!@$a || !@$b) {
-		print @$a, @$b;
+		$line_cb->(@$a, @$b);
 		return;
 	}
 
@@ -93,17 +105,17 @@ sub show_hunk {
 	# stupid, and only handle multi-line hunks that remove and add the same
 	# number of lines.
 	if (@$a != @$b) {
-		print @$a, @$b;
+		$line_cb->(@$a, @$b);
 		return;
 	}
 
 	my @queue;
 	for (my $i = 0; $i < @$a; $i++) {
 		my ($rm, $add) = highlight_pair($a->[$i], $b->[$i]);
-		print $rm;
+		$line_cb->($rm);
 		push @queue, $add;
 	}
-	print @queue;
+	$line_cb->(@queue);
 }
 
 sub highlight_pair {
@@ -163,6 +175,9 @@ sub highlight_pair {
 	}
 }
 
+# we split either by $COLOR or by character. This has the side effect of
+# leaving in graph cruft. It works because the graph cruft does not contain "-"
+# or "+"
 sub split_line {
 	local $_ = shift;
 	return utf8::decode($_) ?
@@ -211,8 +226,8 @@ sub is_pair_interesting {
 	my $suffix_a = join('', @$a[($sa+1)..$#$a]);
 	my $suffix_b = join('', @$b[($sb+1)..$#$b]);
 
-	return $prefix_a !~ /^$COLOR*-$BORING*$/ ||
-	       $prefix_b !~ /^$COLOR*\+$BORING*$/ ||
+	return $prefix_a !~ /^$GRAPH*$COLOR*-$BORING*$/ ||
+	       $prefix_b !~ /^$GRAPH*$COLOR*\+$BORING*$/ ||
 	       $suffix_a !~ /^$BORING*$/ ||
 	       $suffix_b !~ /^$BORING*$/;
 }
diff --git a/contrib/diff-highlight/Makefile b/contrib/diff-highlight/Makefile
new file mode 100644
index 0000000000..f2be7cc924
--- /dev/null
+++ b/contrib/diff-highlight/Makefile
@@ -0,0 +1,23 @@
+all: diff-highlight
+
+PERL_PATH = /usr/bin/perl
+-include ../../config.mak
+
+PERL_PATH_SQ = $(subst ','\'',$(PERL_PATH))
+
+diff-highlight: shebang.perl DiffHighlight.pm diff-highlight.perl
+	cat $^ >$@+
+	chmod +x $@+
+	mv $@+ $@
+
+shebang.perl: FORCE
+	@echo '#!$(PERL_PATH_SQ)' >$@+
+	@cmp $@+ $@ >/dev/null 2>/dev/null || mv $@+ $@
+
+test: all
+	$(MAKE) -C t
+
+clean:
+	$(RM) diff-highlight
+
+.PHONY: FORCE
diff --git a/contrib/diff-highlight/README b/contrib/diff-highlight/README
index 836b97a730..d4c2343175 100644
--- a/contrib/diff-highlight/README
+++ b/contrib/diff-highlight/README
@@ -99,6 +99,36 @@ newHighlight = "black #aaffaa"
 ---------------------------------------------
 
 
+Using diff-highlight as a module
+--------------------------------
+
+If you want to pre- or post- process the highlighted lines as part of
+another perl script, you can use the DiffHighlight module. You can
+either "require" it or just cat the module together with your script (to
+avoid run-time dependencies).
+
+Your script may set up one or more of the following variables:
+
+  - $DiffHighlight::line_cb - this should point to a function which is
+    called whenever DiffHighlight has lines (which may contain
+    highlights) to output. The default function prints each line to
+    stdout. Note that the function may be called with multiple lines.
+
+  - $DiffHighlight::flush_cb - this should point to a function which
+    flushes the output (because DiffHighlight believes it has completed
+    processing a logical chunk of input). The default function flushes
+    stdout.
+
+The script may then feed lines, one at a time, to DiffHighlight::handle_line().
+When lines are done processing, they will be fed to $line_cb. Note that
+DiffHighlight may queue up many input lines (to analyze a whole hunk)
+before calling $line_cb. After providing all lines, call
+DiffHighlight::flush() to flush any unprocessed lines.
+
+If you just want to process stdin, DiffHighlight::highlight_stdin()
+is a convenience helper which will loop and flush for you.
+
+
 Bugs
 ----
 
diff --git a/contrib/diff-highlight/diff-highlight.perl b/contrib/diff-highlight/diff-highlight.perl
new file mode 100644
index 0000000000..9b3e9c1f4d
--- /dev/null
+++ b/contrib/diff-highlight/diff-highlight.perl
@@ -0,0 +1,8 @@
+package main;
+
+# Some scripts may not realize that SIGPIPE is being ignored when launching the
+# pager--for instance scripts written in Python.
+$SIG{PIPE} = 'DEFAULT';
+
+DiffHighlight::highlight_stdin();
+exit 0;
diff --git a/contrib/diff-highlight/t/.gitignore b/contrib/diff-highlight/t/.gitignore
new file mode 100644
index 0000000000..7dcbb232cd
--- /dev/null
+++ b/contrib/diff-highlight/t/.gitignore
@@ -0,0 +1,2 @@
+/trash directory*
+/test-results
diff --git a/contrib/diff-highlight/t/Makefile b/contrib/diff-highlight/t/Makefile
new file mode 100644
index 0000000000..5ff5275496
--- /dev/null
+++ b/contrib/diff-highlight/t/Makefile
@@ -0,0 +1,22 @@
+-include ../../../config.mak.autogen
+-include ../../../config.mak
+
+# copied from ../../t/Makefile
+SHELL_PATH ?= $(SHELL)
+SHELL_PATH_SQ = $(subst ','\'',$(SHELL_PATH))
+T = $(wildcard t[0-9][0-9][0-9][0-9]-*.sh)
+
+all: test
+test: $(T)
+
+.PHONY: help clean all test $(T)
+
+help:
+	@echo 'Run "$(MAKE) test" to launch test scripts'
+	@echo 'Run "$(MAKE) clean" to remove trash folders'
+
+$(T):
+	@echo "*** $@ ***"; '$(SHELL_PATH_SQ)' $@ $(GIT_TEST_OPTS)
+
+clean:
+	$(RM) -r 'trash directory'.*
diff --git a/contrib/diff-highlight/t/t9400-diff-highlight.sh b/contrib/diff-highlight/t/t9400-diff-highlight.sh
new file mode 100755
index 0000000000..3b43dbed74
--- /dev/null
+++ b/contrib/diff-highlight/t/t9400-diff-highlight.sh
@@ -0,0 +1,296 @@
+#!/bin/sh
+
+test_description='Test diff-highlight'
+
+CURR_DIR=$(pwd)
+TEST_OUTPUT_DIRECTORY=$(pwd)
+TEST_DIRECTORY="$CURR_DIR"/../../../t
+DIFF_HIGHLIGHT="$CURR_DIR"/../diff-highlight
+
+CW="$(printf "\033[7m")"	# white
+CR="$(printf "\033[27m")"	# reset
+
+. "$TEST_DIRECTORY"/test-lib.sh
+
+if ! test_have_prereq PERL
+then
+	skip_all='skipping diff-highlight tests; perl not available'
+	test_done
+fi
+
+# dh_test is a test helper function which takes 3 file names as parameters. The
+# first 2 files are used to generate diff and commit output, which is then
+# piped through diff-highlight. The 3rd file should contain the expected output
+# of diff-highlight (minus the diff/commit header, ie. everything after and
+# including the first @@ line).
+dh_test () {
+	a="$1" b="$2" &&
+
+	cat >patch.exp &&
+
+	{
+		cat "$a" >file &&
+		git add file &&
+		git commit -m "Add a file" &&
+
+		cat "$b" >file &&
+		git diff file >diff.raw &&
+		git commit -a -m "Update a file" &&
+		git show >commit.raw
+	} >/dev/null &&
+
+	"$DIFF_HIGHLIGHT" <diff.raw | test_strip_patch_header >diff.act &&
+	"$DIFF_HIGHLIGHT" <commit.raw | test_strip_patch_header >commit.act &&
+	test_cmp patch.exp diff.act &&
+	test_cmp patch.exp commit.act
+}
+
+test_strip_patch_header () {
+	sed -n '/^@@/,$p' $*
+}
+
+# dh_test_setup_history generates a contrived graph such that we have at least
+# 1 nesting (E) and 2 nestings (F).
+#
+#	      A branch
+#	     /
+#	D---E---F master
+#
+#	git log --all --graph
+#	* commit
+#	|    A
+#	| * commit
+#	| |    F
+#	| * commit
+#	|/
+#	|    E
+#	* commit
+#	     D
+#
+dh_test_setup_history () {
+	echo "file1" >file1 &&
+	echo "file2" >file2 &&
+	echo "file3" >file3 &&
+
+	cat file1 >file &&
+	git add file &&
+	git commit -m "D" &&
+
+	git checkout -b branch &&
+	cat file2 >file &&
+	git commit -a -m "A" &&
+
+	git checkout master &&
+	cat file2 >file &&
+	git commit -a -m "E" &&
+
+	cat file3 >file &&
+	git commit -a -m "F"
+}
+
+left_trim () {
+	"$PERL_PATH" -pe 's/^\s+//'
+}
+
+trim_graph () {
+	# graphs start with * or |
+	# followed by a space or / or \
+	"$PERL_PATH" -pe 's@^((\*|\|)( |/|\\))+@@'
+}
+
+test_expect_success 'diff-highlight highlights the beginning of a line' '
+	cat >a <<-\EOF &&
+		aaa
+		bbb
+		ccc
+	EOF
+
+	cat >b <<-\EOF &&
+		aaa
+		0bb
+		ccc
+	EOF
+
+	dh_test a b <<-EOF
+		@@ -1,3 +1,3 @@
+		 aaa
+		-${CW}b${CR}bb
+		+${CW}0${CR}bb
+		 ccc
+	EOF
+'
+
+test_expect_success 'diff-highlight highlights the end of a line' '
+	cat >a <<-\EOF &&
+		aaa
+		bbb
+		ccc
+	EOF
+
+	cat >b <<-\EOF &&
+		aaa
+		bb0
+		ccc
+	EOF
+
+	dh_test a b <<-EOF
+		@@ -1,3 +1,3 @@
+		 aaa
+		-bb${CW}b${CR}
+		+bb${CW}0${CR}
+		 ccc
+	EOF
+'
+
+test_expect_success 'diff-highlight highlights the middle of a line' '
+	cat >a <<-\EOF &&
+		aaa
+		bbb
+		ccc
+	EOF
+
+	cat >b <<-\EOF &&
+		aaa
+		b0b
+		ccc
+	EOF
+
+	dh_test a b <<-EOF
+		@@ -1,3 +1,3 @@
+		 aaa
+		-b${CW}b${CR}b
+		+b${CW}0${CR}b
+		 ccc
+	EOF
+'
+
+test_expect_success 'diff-highlight does not highlight whole line' '
+	cat >a <<-\EOF &&
+		aaa
+		bbb
+		ccc
+	EOF
+
+	cat >b <<-\EOF &&
+		aaa
+		000
+		ccc
+	EOF
+
+	dh_test a b <<-EOF
+		@@ -1,3 +1,3 @@
+		 aaa
+		-bbb
+		+000
+		 ccc
+	EOF
+'
+
+test_expect_failure 'diff-highlight highlights mismatched hunk size' '
+	cat >a <<-\EOF &&
+		aaa
+		bbb
+	EOF
+
+	cat >b <<-\EOF &&
+		aaa
+		b0b
+		ccc
+	EOF
+
+	dh_test a b <<-EOF
+		@@ -1,3 +1,3 @@
+		 aaa
+		-b${CW}b${CR}b
+		+b${CW}0${CR}b
+		+ccc
+	EOF
+'
+
+# These two code points share the same leading byte in UTF-8 representation;
+# a naive byte-wise diff would highlight only the second byte.
+#
+#   - U+00f3 ("o" with acute)
+o_accent=$(printf '\303\263')
+#   - U+00f8 ("o" with stroke)
+o_stroke=$(printf '\303\270')
+
+test_expect_success 'diff-highlight treats multibyte utf-8 as a unit' '
+	echo "unic${o_accent}de" >a &&
+	echo "unic${o_stroke}de" >b &&
+	dh_test a b <<-EOF
+		@@ -1 +1 @@
+		-unic${CW}${o_accent}${CR}de
+		+unic${CW}${o_stroke}${CR}de
+	EOF
+'
+
+# Unlike the UTF-8 above, these are combining code points which are meant
+# to modify the character preceding them:
+#
+#   - U+0301 (combining acute accent)
+combine_accent=$(printf '\314\201')
+#   - U+0302 (combining circumflex)
+combine_circum=$(printf '\314\202')
+
+test_expect_failure 'diff-highlight treats combining code points as a unit' '
+	echo "unico${combine_accent}de" >a &&
+	echo "unico${combine_circum}de" >b &&
+	dh_test a b <<-EOF
+		@@ -1 +1 @@
+		-unic${CW}o${combine_accent}${CR}de
+		+unic${CW}o${combine_circum}${CR}de
+	EOF
+'
+
+test_expect_success 'diff-highlight works with the --graph option' '
+	dh_test_setup_history &&
+
+	# topo-order so that the order of the commits is the same as with --graph
+	# trim graph elements so we can do a diff
+	# trim leading space because our trim_graph is not perfect
+	git log --branches -p --topo-order |
+		"$DIFF_HIGHLIGHT" | left_trim >graph.exp &&
+	git log --branches -p --graph |
+		"$DIFF_HIGHLIGHT" | trim_graph | left_trim >graph.act &&
+	test_cmp graph.exp graph.act
+'
+
+# Most combined diffs won't meet diff-highlight's line-number filter. So we
+# create one here where one side drops a line and the other modifies it. That
+# should result in a diff like:
+#
+#    - modified content
+#    ++resolved content
+#
+# which naively looks like one side added "+resolved".
+test_expect_success 'diff-highlight ignores combined diffs' '
+	echo "content" >file &&
+	git add file &&
+	git commit -m base &&
+
+	>file &&
+	git commit -am master &&
+
+	git checkout -b other HEAD^ &&
+	echo "modified content" >file &&
+	git commit -am other &&
+
+	test_must_fail git merge master &&
+	echo "resolved content" >file &&
+	git commit -am resolved &&
+
+	cat >expect <<-\EOF &&
+	--- a/file
+	+++ b/file
+	@@@ -1,1 -1,0 +1,1 @@@
+	- modified content
+	++resolved content
+	EOF
+
+	git show -c | "$DIFF_HIGHLIGHT" >actual.raw &&
+	sed -n "/^---/,\$p" <actual.raw >actual &&
+	test_cmp expect actual
+'
+
+test_done