[PATCH] .gitignore git-describe

[PATCH] .gitignore git-describe

[Andreas Ericsson]

Signed-off-by: Andreas Ericsson <ae@op5.se>

---

 .gitignore |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

265df952d4c61f60eae42261aeea90e9b6d8e414
diff --git a/.gitignore b/.gitignore
index 1a9090b..4335061 100644
--- a/.gitignore
+++ b/.gitignore
@@ -23,6 +23,7 @@ git-count-objects
 git-cvsexportcommit
 git-cvsimport
 git-daemon
+git-describe
 git-diff
 git-diff-files
 git-diff-index
-- 
1.0.7-gb38d1

git-format-patch Date header

[Timo Hirvonen]

git-format-patch uses "date '+%s %z'" format for the Date header.
Sylpheed does not understand this format.  How common this convention
is?  Should we change git-format-patch to use more standard date format?

OTOH, I like the %s format because it is so simple and easy to parse :)

-- 
http://onion.dynserv.net/~timo/

Re: git-format-patch Date header

[Andreas Ericsson]

Timo Hirvonen wrote:
> git-format-patch uses "date '+%s %z'" format for the Date header.
> Sylpheed does not understand this format.  How common this convention
> is?  Should we change git-format-patch to use more standard date format?
> 
> OTOH, I like the %s format because it is so simple and easy to parse :)
> 

Actually, that's what's printed in the commit message, so any change 
would have to be put there and that would break backwards compatibility 
for new tools that might want to use it.

On an unrelated note, please don't start a new thread by replying to an 
old message. It's very confusing to follow and people may not see your 
message. This time it was doubly so because the next mail in my inbox is 
a patch for git-format-patch, to which I thought you replied.

-- 
Andreas Ericsson                   andreas.ericsson@op5.se
OP5 AB                             www.op5.se
Tel: +46 8-230225                  Fax: +46 8-230231

Re: git-format-patch Date header

[Timo Hirvonen]

On Sun, 08 Jan 2006 16:21:13 +0100
Andreas Ericsson <ae@op5.se> wrote:

> Timo Hirvonen wrote:
> > git-format-patch uses "date '+%s %z'" format for the Date header.
> > Sylpheed does not understand this format.  How common this convention
> > is?  Should we change git-format-patch to use more standard date format?
> > 
> > OTOH, I like the %s format because it is so simple and easy to parse :)
> > 
> 
> Actually, that's what's printed in the commit message, so any change 
> would have to be put there and that would break backwards compatibility 
> for new tools that might want to use it.

So fixing Sylpheed (and maybe other email clients too) is the right
solution...

> On an unrelated note, please don't start a new thread by replying to an 
> old message. It's very confusing to follow and people may not see your 
> message. This time it was doubly so because the next mail in my inbox is 
> a patch for git-format-patch, to which I thought you replied.

Sorry.  I replied to your message because it was the message having Date
header in the "%s %z" format.

-- 
http://onion.dynserv.net/~timo/

Re: git-format-patch Date header

[Andreas Ericsson]

Timo Hirvonen wrote:
> On Sun, 08 Jan 2006 16:21:13 +0100
> Andreas Ericsson <ae@op5.se> wrote:
> 
> 
>>Timo Hirvonen wrote:
>>
>>>git-format-patch uses "date '+%s %z'" format for the Date header.
>>>Sylpheed does not understand this format.  How common this convention
>>>is?  Should we change git-format-patch to use more standard date format?
>>>
>>>OTOH, I like the %s format because it is so simple and easy to parse :)
>>>
>>
>>Actually, that's what's printed in the commit message, so any change 
>>would have to be put there and that would break backwards compatibility 
>>for new tools that might want to use it.
> 
> 
> So fixing Sylpheed (and maybe other email clients too) is the right
> solution...
> 

Nopes. In this case fixing my home-hacked git-send-patch script is the 
right thing to do. git-format-patch has always printed that format, 
though, so the error lies with me and the patch I sent. I shall have to 
re-visit it after reading rfc 822 a bit more closely.

> 
>>On an unrelated note, please don't start a new thread by replying to an 
>>old message. It's very confusing to follow and people may not see your 
>>message. This time it was doubly so because the next mail in my inbox is 
>>a patch for git-format-patch, to which I thought you replied.
> 
> 
> Sorry.  I replied to your message because it was the message having Date
> header in the "%s %z" format.
> 

I just noticed. Thanks, and sorry.

-- 
Andreas Ericsson                   andreas.ericsson@op5.se
OP5 AB                             www.op5.se
Tel: +46 8-230225                  Fax: +46 8-230231

Re: git-format-patch Date header

[H. Peter Anvin]

Andreas Ericsson wrote:
> 
> Nopes. In this case fixing my home-hacked git-send-patch script is the 
> right thing to do. git-format-patch has always printed that format, 
> though, so the error lies with me and the patch I sent. I shall have to 
> re-visit it after reading rfc 822 a bit more closely.
> 

RFC 2822, please.

	-hpa

Re: git-format-patch Date header

[Andreas Ericsson]

H. Peter Anvin wrote:
> Andreas Ericsson wrote:
> 
>>
>> Nopes. In this case fixing my home-hacked git-send-patch script is the 
>> right thing to do. git-format-patch has always printed that format, 
>> though, so the error lies with me and the patch I sent. I shall have 
>> to re-visit it after reading rfc 822 a bit more closely.
>>
> 
> RFC 2822, please.
> 

Right you are.

-- 
Andreas Ericsson                   andreas.ericsson@op5.se
OP5 AB                             www.op5.se
Tel: +46 8-230225                  Fax: +46 8-230231

Re: git-format-patch Date header

[Linus Torvalds]

On Sun, 8 Jan 2006, Andreas Ericsson wrote:
> 
> Actually, that's what's printed in the commit message, so any change 
> would have to be put there and that would break backwards compatibility 
> for new tools that might want to use it.

No, git should always take the date in any of a million different formats, 
and always turn it into "seconds + timezone" internally.

The fact that git-format-patch also prints it out in the internal format 
is unambiguous (nice) but human-unreadable (bad). 

There are other unambiguous formats it could use, notably standard rfc2822 
format ("date -R").

The git "show_date()" function hopefully does exactly that rfc2822 format, 
but you'd have to make a helper program to do the conversion in 
git-format-patch.sh.

		Linus

[PATCH] format-patch: Use --mbox by default.

[Andreas Ericsson]

This patch enables a slightly modified and more correct --mbox
output by default. The --date, --author and --mbox options
are obsoleted and removed.

Previously, author-info and date was by default only printed if the
commit was made by someone else, irrespective of the obsoleted options.
Since it does no harm having ones own info there too it shouldn't really
matter much, and should lessen the confusion regarding how to use
git-format-patch with git-am.

The patches generated also had bad mail headers for sendmail (of
postfix 2.2.2 at least) to use directly (From: line must be on top,
Subject: last of header), which isn't really an issue but annoying
just the same. I haven't actually tried this new format, but at least
'file' now recognize it as "news or mail text", rather than "ASCII
English text".

Note that this reverts commit aa66c7ec77d474b737da607d6cb2d07f56628def,
"format-patch/commit: Quote single quote in the author name properly"
which is no longer needed since the whosepatchScript now prints
the headers directly rather than passing them to eval.

These changes allow us to remove two conditional statements
and the invokation of a sub-shell in the process_one function,
so it should be a bit faster too.

Signed-off-by: Andreas Ericsson <ae@op5.se>

---

 Documentation/git-format-patch.txt |   18 ++-----------
 git-format-patch.sh                |   49 +++++++-----------------------------
 2 files changed, 12 insertions(+), 55 deletions(-)

b38d1ffeb27ba32de849f33a9c416fda4cd8eecb
diff --git a/Documentation/git-format-patch.txt b/Documentation/git-format-patch.txt
index 47705de..73e1358 100644
--- a/Documentation/git-format-patch.txt
+++ b/Documentation/git-format-patch.txt
@@ -9,7 +9,7 @@ git-format-patch - Prepare patches for e
 SYNOPSIS
 --------
 [verse]
-'git-format-patch' [-n | -k] [-o <dir> | --stdout] [-s] [-c] [--mbox]
+'git-format-patch' [-n | -k] [-o <dir> | --stdout] [-s] [-c]
 		 [--diff-options] <his> [<mine>]
 
 DESCRIPTION
@@ -27,9 +27,8 @@ When -n is specified, instead of "[PATCH
 line is formatted as "[PATCH N/M] Subject", unless you have only
 one patch.
 
-When --mbox is specified, the output is formatted to resemble
-UNIX mailbox format, and can be concatenated together for
-processing with applymbox.
+The output is formatted to resemble UNIX mailbox format, and can
+be concatenated together for processing with gitlink::git-am[1].
 
 
 OPTIONS
@@ -45,11 +44,6 @@ OPTIONS
 	Do not strip/add '[PATCH]' from the first line of the
 	commit log message.
 
--a|--author, -d|--date::
-	Output From: and Date: headers for commits made by
-	yourself as well.  Usually these are output only for
-	commits made by people other than yourself.
-
 -s|--signoff::
 	Add `Signed-off-by:` line to the commit message, using
 	the committer identity of yourself.
@@ -61,12 +55,6 @@ OPTIONS
         has a SP character immediately followed by a TAB
         character.
 
--m|--mbox::
-	Format the output files for closer to mbox format by
-	adding a phony Unix "From " line, so they can be
-	concatenated together and fed to `git-applymbox`.
-	Implies --author and --date.
-
 --stdout::
 	This flag generates the mbox formatted output to the
 	standard output, instead of saving them into a file per
diff --git a/git-format-patch.sh b/git-format-patch.sh
index 818059f..047246e 100755
--- a/git-format-patch.sh
+++ b/git-format-patch.sh
@@ -3,7 +3,7 @@
 # Copyright (c) 2005 Junio C Hamano
 #
 
-USAGE='[-n | -k] [-o <dir> | --stdout] [--signoff] [--check] [--mbox] [--diff-options] <upstream> [<our-head>]'
+USAGE='[-n | -k] [-o <dir> | --stdout] [--signoff] [--check] [--diff-options] <upstream> [<our-head>]'
 LONG_USAGE='Prepare each commit with its patch since our-head forked from upstream,
 one file per patch, for e-mail submission.  Each output file is
 numbered sequentially from 1, and uses the first line of the commit
@@ -15,9 +15,8 @@ the current working directory.
 When -n is specified, instead of "[PATCH] Subject", the first line is formatted
 as "[PATCH N/M] Subject", unless you have only one patch.
 
-When --mbox is specified, the output is formatted to resemble
-UNIX mailbox format, and can be concatenated together for processing
-with applymbox.'
+The output is formatted to resemble UNIX mailbox format, and can be
+concatenated together for processing with "git am".'
 . git-sh-setup
 
 # Force diff to run in C locale.
@@ -32,14 +31,8 @@ outdir=./
 while case "$#" in 0) break;; esac
 do
     case "$1" in
-    -a|--a|--au|--aut|--auth|--autho|--author)
-    author=t ;;
     -c|--c|--ch|--che|--chec|--check)
     check=t ;;
-    -d|--d|--da|--dat|--date)
-    date=t ;;
-    -m|--m|--mb|--mbo|--mbox)
-    date=t author=t mbox=t ;;
     -k|--k|--ke|--kee|--keep|--keep-|--keep-s|--keep-su|--keep-sub|\
     --keep-subj|--keep-subje|--keep-subjec|--keep-subject)
     keep_subject=t ;;
@@ -48,7 +41,7 @@ do
     -s|--s|--si|--sig|--sign|--signo|--signof|--signoff)
     signoff=t ;;
     --st|--std|--stdo|--stdou|--stdout)
-    stdout=t mbox=t date=t author=t ;;
+    stdout=t ;;
     -o=*|--o=*|--ou=*|--out=*|--outp=*|--outpu=*|--output=*|--output-=*|\
     --output-d=*|--output-di=*|--output-dir=*|--output-dire=*|\
     --output-direc=*|--output-direct=*|--output-directo=*|\
@@ -173,12 +166,12 @@ titleScript='
 
 whosepatchScript='
 /^author /{
-	s/'\''/'\''\\'\'\''/g
-	s/author \(.*>\) \(.*\)$/au='\''\1'\'' ad='\''\2'\''/p
+	s/^author \([^>]*>\)[\t ]*\(.*\)$/From: \1\nDate: \2/p
 	q
 }'
 
 process_one () {
+	sed -ne "$whosepatchScript" $commsg
 	mailScript='
 	/./d
 	/^$/n'
@@ -191,26 +184,7 @@ process_one () {
 	    ;;
 	esac
 	mailScript="$mailScript"'
-	s|^|Subject: |'
-	case "$mbox" in
-	t)
-	    echo 'From nobody Mon Sep 17 00:00:00 2001' ;# UNIX "From" line
-	    ;;
-	esac
-
-	eval "$(sed -ne "$whosepatchScript" $commsg)"
-	test "$author,$au" = ",$me" || {
-		mailScript="$mailScript"'
-	a\
-From: '"$au"
-	}
-	test "$date,$au" = ",$me" || {
-		mailScript="$mailScript"'
-	a\
-Date: '"$ad"
-	}
-
-	mailScript="$mailScript"'
+	s|^|Subject: |
 	a\
 
 	: body
@@ -218,7 +192,7 @@ Date: '"$ad"
 	n
 	b body'
 
-	(cat $commsg ; echo; echo) |
+	(cat $commsg; echo; echo) |
 	sed -ne "$mailScript" |
 	git-stripspace
 
@@ -239,12 +213,7 @@ Date: '"$ad"
 	git-diff-tree -p $diff_opts "$commit"
 	echo "-- "
 	echo "@@GIT_VERSION@@"
-
-	case "$mbox" in
-	t)
-		echo
-		;;
-	esac
+	echo
 }
 
 total=`wc -l <$series | tr -dc "[0-9]"`
-- 
1.0.7-gb38d1

Re: [PATCH] format-patch: Use --mbox by default.

[Junio C Hamano]

Andreas Ericsson <ae@op5.se> writes:

> This patch enables a slightly modified and more correct --mbox
> output by default. The --date, --author and --mbox options
> are obsoleted and removed.

I always use format-patch with -k -m (or -k --stdout), so I have
no objections to making --mbox the default.  I do not think
anybody uses --date or --author in their scripts, but removing
the --mbox option _might_ break things for people's scripts.
Maybe it is worthwhile to keep it as a no-op option for a while.

> Previously, author-info and date was by default only printed if the
> commit was made by someone else, irrespective of the obsoleted options.
> Since it does no harm having ones own info there too it shouldn't really
> matter much, and should lessen the confusion regarding how to use
> git-format-patch with git-am.

True.  Originally it was really "format my changes to
submission" tool (--signoff option is a sure sign to tell you
that) and propagating authorship info was added as afterthought.

> The patches generated also had bad mail headers for sendmail (of
> postfix 2.2.2 at least) to use directly (From: line must be on top,
> Subject: last of header),...

I do not think RFC2822 specifies that.  See "3.6 Field
definitions", third paragraph, first two sentences.

In that sense, technically they are not "bad mail headers", but
at the same time, since it is valid in any order, if _one_
popular MTA is unconfortable with a particular ordering, we
could order things differently to make it happier (mind you,
this would not work for more than one such broken MTAs).

Another thing I've been considering is to handle UTF-8 (or
i18n.commitencoding in general) a bit better.  We currently do
not have Content-Type: header, so if you pipe the output to
"am", it must be invoked without "-u" to make sure that it takes
things as-is ("am -u" assumes things are in latin1 without
Content-Type:).  If we start to tackle that issue, we might also
want to do RFC2047 in headers.

Re: [PATCH] format-patch: Use --mbox by default.

[Andreas Ericsson]

Junio C Hamano wrote:
> Andreas Ericsson <ae@op5.se> writes:
> 
> 
>>This patch enables a slightly modified and more correct --mbox
>>output by default. The --date, --author and --mbox options
>>are obsoleted and removed.
> 
> 
> I always use format-patch with -k -m (or -k --stdout), so I have
> no objections to making --mbox the default.  I do not think
> anybody uses --date or --author in their scripts, but removing
> the --mbox option _might_ break things for people's scripts.
> Maybe it is worthwhile to keep it as a no-op option for a while.
> 

The reworked patch (which works for sending patches to myself, so far), 
prints a warning to stderr that a deprecated option is used and then 
silently ignores it.

> 
>>The patches generated also had bad mail headers for sendmail (of
>>postfix 2.2.2 at least) to use directly (From: line must be on top,
>>Subject: last of header),...
> 
> 
> I do not think RFC2822 specifies that.  See "3.6 Field
> definitions", third paragraph, first two sentences.
> 
> In that sense, technically they are not "bad mail headers", but
> at the same time, since it is valid in any order, if _one_
> popular MTA is unconfortable with a particular ordering, we
> could order things differently to make it happier (mind you,
> this would not work for more than one such broken MTAs).
> 


The only thing I've seen break is if Subject: is placed on top. It's 
easier, script-wise, to put it last and that seems to work with qmail 
and postfix both.


> Another thing I've been considering is to handle UTF-8 (or
> i18n.commitencoding in general) a bit better.  We currently do
> not have Content-Type: header, so if you pipe the output to
> "am", it must be invoked without "-u" to make sure that it takes
> things as-is ("am -u" assumes things are in latin1 without
> Content-Type:).  If we start to tackle that issue, we might also
> want to do RFC2047 in headers.
> 
> 

You'll have to ask someone else for that. I have no idea how it works 
and I wouldn't know how to begin doing it in shell.

-- 
Andreas Ericsson                   andreas.ericsson@op5.se
OP5 AB                             www.op5.se
Tel: +46 8-230225                  Fax: +46 8-230231

Re: [PATCH] format-patch: Use --mbox by default.

[H. Peter Anvin]

Andreas Ericsson wrote:
> 
> The patches generated also had bad mail headers for sendmail (of
> postfix 2.2.2 at least) to use directly (From: line must be on top,
> Subject: last of header)

Sounds like a bug report against postfix is in order.

	-hpa

Re: [PATCH] format-patch: Use --mbox by default.

[Andreas Ericsson]

H. Peter Anvin wrote:
> Andreas Ericsson wrote:
> 
>>
>> The patches generated also had bad mail headers for sendmail (of
>> postfix 2.2.2 at least) to use directly (From: line must be on top,
>> Subject: last of header)
> 
> 
> Sounds like a bug report against postfix is in order.
> 

No need. It works with 2.2.8. The 2.2.2 it failed with is the default 
for FC4. I have no interest investigating all the possibilities, and I 
doubt Vietse will look for bugs in old versions unless they're security 
related.

-- 
Andreas Ericsson                   andreas.ericsson@op5.se
OP5 AB                             www.op5.se
Tel: +46 8-230225                  Fax: +46 8-230231

[PATCH] format-patch: always --mbox and show sane Date:

[Junio C Hamano]

Make --mbox, --author, and --date options a no-op, and always
use --mbox output, and rewrite the commit log formatting in
Perl.  This makes it easier to output Date: header in RFC 2822
format, so do that as well.

Inspiration for this patch came from Andreas Ericsson's earlier
patch.

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

---

 * I admit that trying to avoid Perl and sticking to multiple
   sed invocations in this script was a lost battle.  This patch
   is still minimalistic and it does not go all the way to
   rewrite the main loop of the program, which probably should
   happen someday.

   I would not mind if someday somebody rewrote most of the
   barebone Porcelainish in Python or Perl, but the rewrite
   should not be for the sake of using that language.  It should
   have other justifications: "there is too much text processing
   there it would not make sense to do that in shell".  I think
   this patch is borderline to qualify to be in that camp.

   As the version string at the end shows, this was prepared
   with the script source itself before building, as a test
   case.

 git-format-patch.sh |  143 +++++++++++++++++++++++++++------------------------
 1 files changed, 75 insertions(+), 68 deletions(-)

8b4271ef24818e6d52926bc615e5ce78ba0ed01b
diff --git a/git-format-patch.sh b/git-format-patch.sh
index d3979d7..7e67c4e 100755
--- a/git-format-patch.sh
+++ b/git-format-patch.sh
@@ -34,14 +34,12 @@ outdir=./
 while case "$#" in 0) break;; esac
 do
     case "$1" in
-    -a|--a|--au|--aut|--auth|--autho|--author)
-    author=t ;;
     -c|--c|--ch|--che|--chec|--check)
     check=t ;;
-    -d|--d|--da|--dat|--date)
-    date=t ;;
-    -m|--m|--mb|--mbo|--mbox)
-    date=t author=t mbox=t ;;
+    -a|--a|--au|--aut|--auth|--autho|--author|\
+    -d|--d|--da|--dat|--date|\
+    -m|--m|--mb|--mbo|--mbox) # now noop
+    ;;
     -k|--k|--ke|--kee|--keep|--keep-|--keep-s|--keep-su|--keep-sub|\
     --keep-subj|--keep-subje|--keep-subjec|--keep-subject)
     keep_subject=t ;;
@@ -173,80 +171,89 @@ titleScript='
 	q
 '
 
-whosepatchScript='
-/^author /{
-	s/'\''/'\''\\'\'\''/g
-	s/author \(.*>\) \(.*\)$/au='\''\1'\'' ad='\''\2'\''/p
-	q
-}'
-
 process_one () {
-	mailScript='
-	/./d
-	/^$/n'
-	case "$keep_subject" in
-	t)  ;;
-	*)
-	    mailScript="$mailScript"'
-	    s|^\[PATCH[^]]*\] *||
-	    s|^|[PATCH'"$num"'] |'
-	    ;;
-	esac
-	mailScript="$mailScript"'
-	s|^|Subject: |'
-	case "$mbox" in
-	t)
-	    echo 'From nobody Mon Sep 17 00:00:00 2001' ;# UNIX "From" line
-	    ;;
-	esac
+	perl -w -e '
+my ($keep_subject, $num, $signoff, $commsg) = @ARGV;
+my ($signoff_pattern, $done_header, $done_subject, $signoff_seen,
+    $last_was_signoff);
+
+if ($signoff) {
+	$signoff = `git-var GIT_COMMITTER_IDENT`;
+	$signoff =~ s/>.*/>/;
+	$signoff_pattern = quotemeta($signoff);
+}
 
-	eval "$(sed -ne "$whosepatchScript" $commsg)"
-	test "$author,$au" = ",$me" || {
-		mailScript="$mailScript"'
-	a\
-From: '"$au"
-	}
-	test "$date,$au" = ",$me" || {
-		mailScript="$mailScript"'
-	a\
-Date: '"$ad"
-	}
+my @weekday_names = qw(Sun Mon Tue Wed Thu Fri Sat);
+my @month_names = qw(Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec);
 
-	mailScript="$mailScript"'
-	a\
+sub show_date {
+    my ($time, $tz) = @_;
+    my $minutes = abs($tz);
+    $minutes = ($minutes / 100) * 60 + ($minutes % 100);
+    if ($tz < 0) {
+        $minutes = -$minutes;
+    }
+    my $t = $time + $minutes * 60;
+    my ($sec,$min,$hour,$mday,$mon,$year,$wday,$yday) = gmtime($t);
+    return sprintf("%s %s %d %02d:%02d:%02d %d %+05d",
+		   $weekday_names[$wday],
+		   $month_names[$mon],
+		   $mday, $hour, $min, $sec,
+		   $year+1900, $tz);
+}
 
-	: body
-	p
-	n
-	b body'
+print "From nobody Mon Sep 17 00:00:00 2001\n";
+open FH, "git stripspace <$commsg |" or die "open $commsg pipe";
+while (<FH>) {
+    unless ($done_header) {
+	if (/^$/) {
+	    $done_header = 1;
+	}
+	elsif (/^author (.*>) (.*)$/) {
+	    my ($author_ident, $author_date) = ($1, $2);
+	    my ($utc, $off) = ($author_date =~ /^(\d+) ([-+]?\d+)$/);
+	    $author_date = show_date($utc, $off);
 
-	(cat $commsg ; echo; echo) |
-	sed -ne "$mailScript" |
-	git-stripspace
-
-	test "$signoff" = "t" && {
-		offsigner=`git-var GIT_COMMITTER_IDENT | sed -e 's/>.*/>/'`
-		line="Signed-off-by: $offsigner"
-		grep -q "^$line\$" $commsg || {
-			echo
-			echo "$line"
-			echo
-		}
+	    print "From: $author_ident\n";
+	    print "Date: $author_date\n";
 	}
-	echo
-	echo '---'
-	echo
+	next;
+    }
+    unless ($done_subject) {
+	unless ($keep_subject) {
+	    s/^\[PATCH[^]]*\]\s*//;
+	    s/^/[PATCH$num] /;
+	}
+        print "Subject: $_";
+	$done_subject = 1;
+	next;
+    }
+
+    $last_was_signoff = 0;
+    if (/Signed-off-by:/i) {
+        if ($signoff ne "" && /Signed-off-by:\s*$signoff_pattern$/i) {
+	    $signoff_seen = 1;
+	}
+    }
+    print $_;
+}
+if (!$signoff_seen && $signoff ne "") {
+    if (!$last_was_signoff) {
+        print "\n";
+    }
+    print "$signoff\n";
+}
+print "\n---\n\n";
+close FH or die "close $commsg pipe";
+' "$keep_subject" "$num" "$signoff" $commsg
+
 	git-diff-tree -p $diff_opts "$commit" | git-apply --stat --summary
 	echo
 	git-diff-tree -p $diff_opts "$commit"
 	echo "-- "
 	echo "@@GIT_VERSION@@"
 
-	case "$mbox" in
-	t)
-		echo
-		;;
-	esac
+	echo
 }
 
 total=`wc -l <$series | tr -dc "[0-9]"`
-- 
@@GIT_VERSION@@