a bug about format-patch of multibyte characters comment

[xiaozhu <xiaozhu@gmail.com>]

Hi,

I found a bug when I use format-patch to export a patch which contains comment with
some multibyte characters. I also found the relation source, but I can't understand
the source clearly, so I think I need a help to know how can I fix it.

At first, the symptom.

I commit a fix to my repository with comment like following:
-----------------------------------------------------
XXXXXXXXXXXX
YYYYYY
-----------------------------------------------------

two lines of multibyte language comment.

then I use format-patch to export this fix, I get a patch file like following:

------------------------------------------------------------------------------
 From d3532c3263a02a2367a3aa5c9cc3f0bd738b79b1 Mon Sep 17 00:00:00 2001
From: xz <xz>
Date: Fri, 11 Feb 2011 21:30:35 +0900
Subject: [PATCH] =?UTF-8?q?=E6=97=A5=E6=9C=AC=E8=AA=9E=E3=81=8C=E5=A4=A7=E4=B8=88=E5=A4=AB
=20=E6=94=B9=E8=A1=8C=E3=81=99=E3=82=8B?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
  testfile.txt |    4 +++-
  1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/testfile.txt b/testfile.txt
index 1e5d832..da982fd 100644
--- a/testfile.txt
+++ b/testfile.txt
@@ -1 +1,3 @@
-sadfasdf
\ No newline at end of file
+sadfasdf

..........

-------------------------------------------------------------------------------

If I use am to apply this patch, am can't analyze the comment correctly, then the
committed comment will become
"=?UTF-8?q?=E6=97=A5=E6=9C=AC=E8=AA=9E=E3=81=8C=E5=A4=A7=E4=B8=88=E5=A4=AB".

Above is the symptom.

Then I did some try, I modify the comment to 3 lines:
-----------------------------------------------------
XXXXXXXXXXXX

YYYYYY
-----------------------------------------------------

add a empty line, then I get a patch like following:
------------------------------------------------------------------------------
 From d3532c3263a02a2367a3aa5c9cc3f0bd738b79b1 Mon Sep 17 00:00:00 2001
From: xz <xz>
Date: Fri, 11 Feb 2011 21:30:35 +0900
Subject: [PATCH] =?UTF-8?q?=E6=97=A5=E6=9C=AC=E8=AA=9E=E3=81=8C=E5=A4=A7=E4=B8=88=E5=A4=AB?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

YYYYYY
---
  testfile.txt |    4 +++-
  1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/testfile.txt b/testfile.txt
index 1e5d832..da982fd 100644
--- a/testfile.txt
+++ b/testfile.txt
@@ -1 +1,3 @@
-sadfasdf
\ No newline at end of file
+sadfasdf

..........

-------------------------------------------------------------------------------

this patch will be applied successfully. So I know the problem is about the subject creating.
I search the source, then I found the following function at "pretty.c:655":

const char *format_subject(struct strbuf *sb, const char *msg,
			   const char *line_separator)
{
	int first = 1;

	for (;;) {
		const char *line = msg;
		int linelen = get_one_line(line);

		msg += linelen;
		
		if (!linelen || is_empty_line(line, &linelen))
			break;

		if (!sb)cat
			continue;
		strbuf_grow(sb, linelen + 2);
		if (!first)
			strbuf_addstr(sb, line_separator);
		strbuf_add(sb, line, linelen);
		first = 0;
	}
	return msg;
}

At first I want to know: Does this function means that always add the first line
of comment to the argument sb, then return the rest? Is there any other thing that I
didn't considered?

I found 4 place where to call this function, I think there is no problem about 3
of them, but I don't know is there any other problem to the rest one which is
at "pretty.c:931".

At last, if what I think is correct, I plan to fix it as following:

const char *format_subject(struct strbuf *sb, const char *msg,
			   const char *line_separator)
{
	int first = 1;

	//for (;;) {
		const char *line = msg;
		int linelen = get_one_line(line);

		msg += linelen;
		
		if (!linelen || is_empty_line(line, &linelen)) return msg;
			//break;

		if (!sb) return msg;
			//continue;
		strbuf_grow(sb, linelen + 2);
		if (!first)
			strbuf_addstr(sb, line_separator);
		strbuf_add(sb, line, linelen);
		first = 0;
	//}
	return msg;
}

I dont't think it is necessary to have a loop here, so I want to remove
the loop. Is there anybody can confirm my fix for me?