format-patch: take sequence of ranges.

This enhances set of revs you can give format-patch.

Originally, format-patch took either one rev, or two revs:

    format-patch rev1
    format-patch rev1 rev2

The first format was a short-hand for "format-patch rev1 HEAD"
(i.e. rev2==HEAD).  What this meant was to find commits that are
in branch rev2 that has not been merged to branch rev1.

The above notation is still supported, but now it takes sequence
of "from1..to1 from2..to2 ...".  In short, the second format has
become a short-hand for "format-patch rev1..rev2".  Commits in
to1 but not in from1, to2 but not in from2, ... are formatted as
emailable patches.

With this, cherry-picking from other branch can be written as:

    git-format-patch -k --stdout master..branch1 master..branch2 |
    git-am -k -3

which is generally faster than traditional cherry-pick (which
always did 3-way merge) if patches apply cleanly, and still
falls back on 3-way merge if some of them do not.

Signed-off-by: Junio C Hamano <junkio@cox.net>

git-format-patch.sh

@@ -6,7 +6,9 @@
 . git-sh-setup || die "Not a git archive."
 
 usage () {
-    echo >&2 "usage: $0"' [-n] [-o dir | --stdout] [--keep-subject] [--mbox] [--check] [--signoff] [-<diff options>...] upstream [ our-head ]
+    echo >&2 "usage: $0"' [-n] [-o dir | --stdout] [--keep-subject] [--mbox]
+    [--check] [--signoff] [-<diff options>...]
+    ( from..to ... | upstream [ our-head ] )
 
 Prepare each commit with its patch since our-head forked from upstream,
 one file per patch, for e-mail submission.  Each output file is
@@ -75,25 +77,70 @@ tt)
 	die '--keep-subject and --numbered are incompatible.' ;;
 esac
 
-rev1= rev2=
-case "$#" in
-2)
-    rev1="$1" rev2="$2" ;;
-1)
-    case "$1" in
-    *..*)
-	rev1=`expr "$1" : '\(.*\)\.\.'`
-	rev2=`expr "$1" : '.*\.\.\(.*\)'`
+tmp=.tmp-series$$
+trap 'rm -f $tmp-*' 0 1 2 3 15
+
+series=$tmp-series
+commsg=$tmp-commsg
+filelist=$tmp-files
+
+# Backward compatible argument parsing hack.
+#
+# Historically, we supported:
+# 1. "rev1"		is equivalent to "rev1..HEAD"
+# 2. "rev1..rev2"
+# 3. "rev1" "rev2	is equivalent to "rev1..rev2"
+#
+# We want to take a sequence of "rev1..rev2" in general.
+
+case "$#,$1" in
+1,?*..?*)
+	# single "rev1..rev2"
 	;;
-    *)
-        rev1="$1"
-	rev2="HEAD"
+1,*)
+	# single rev1
+	set x "$1..HEAD"
+	shift
+	;;
+2,?*..?*)
+	# not traditional "rev1" "rev2"
+	;;
+2,*)
+	set x "$1..$2"
+	shift
 	;;
-    esac ;;
-*)
-    usage ;;
 esac
 
+# Now we have what we want in $@
+for revpair
+do
+	case "$revpair" in
+	?*..?*)
+		rev1=`expr "$revpair" : '\(.*\)\.\.'`
+		rev2=`expr "$revpair" : '.*\.\.\(.*\)'`
+		;;
+	*)
+		usage
+		;;
+	esac
+	git-rev-parse --verify "$rev1^0" >/dev/null 2>&1 ||
+		die "Not a valid rev $rev1 ($revpair)"
+	git-rev-parse --verify "$rev2^0" >/dev/null 2>&1 ||
+		die "Not a valid rev $rev2 ($revpair)"
+	git-cherry -v "$rev1" "$rev2" |
+	while read sign rev comment
+	do
+		case "$sign" in
+		'-')
+			echo >&2 "Merged already: $comment"
+			;;
+		*)
+			echo $rev
+			;;
+		esac
+	done
+done >$series
+
 me=`git-var GIT_AUTHOR_IDENT | sed -e 's/>.*/>/'`
 
 case "$outdir" in
@@ -102,13 +149,6 @@ case "$outdir" in
 esac
 test -d "$outdir" || mkdir -p "$outdir" || exit
 
-tmp=.tmp-series$$
-trap 'rm -f $tmp-*' 0 1 2 3 15
-
-series=$tmp-series
-commsg=$tmp-commsg
-filelist=$tmp-files
-
 titleScript='
 	/./d
 	/^$/n
@@ -130,19 +170,6 @@ whosepatchScript='
 	q
 }'
 
-git-cherry -v "$rev1" "$rev2" |
-while read sign rev comment
-do
-	case "$sign" in
-	'-')
-		echo >&2 "Merged already: $comment"
-		;;
-	*)
-		echo $rev
-		;;
-	esac
-done >$series
-
 process_one () {
 	mailScript='
 	/./d