sequencer: fix "rebase -i --root" corrupting author header timezone

When "git rebase -i --root" creates a new root commit, it corrupts the
"author" header's timezone by repeating the last digit:

    author A U Thor <[email protected]> @1112912773 -07000

This is due to two bugs.

First, write_author_script() neglects to add the closing quote to the
value of GIT_AUTHOR_DATE when generating "rebase-merge/author-script".

Second, although sq_dequote() correctly diagnoses the missing closing
quote, read_author_ident() ignores sq_dequote()'s return value and
blindly uses the result of the aborted dequote.

sq_dequote() performs dequoting in-place by removing quoting and
shifting content downward. When it detects misquoting (lack of closing
quote, in this case), it gives up and returns an error without inserting
a NUL-terminator at the end of the shifted content, which explains the
duplicated last digit in the timezone.

(Note that the "@" preceding the timestamp is a separate bug which
will be fixed subsequently.)

Signed-off-by: Eric Sunshine <[email protected]>
Signed-off-by: Junio C Hamano <[email protected]>

Author: sunshineco

sequencer.c

@@ -651,6 +651,7 @@ static int write_author_script(const char *message)
 			strbuf_addch(&buf, *(message++));
 		else
 			strbuf_addf(&buf, "'\\\\%c'", *(message++));
+	strbuf_addch(&buf, '\'');
 	res = write_message(buf.buf, buf.len, rebase_path_author_script(), 1);
 	strbuf_release(&buf);
 	return res;
@@ -721,7 +722,11 @@ static const char *read_author_ident(struct strbuf *buf)
 
 		eol = strchrnul(in, '\n');
 		*eol = '\0';
-		sq_dequote(in);
+		if (!sq_dequote(in)) {
+			warning(_("bad quoting on %s value in '%s'"),
+				keys[i], rebase_path_author_script());
+			return NULL;
+		}
 		len = strlen(in);
 
 		if (i > 0) /* separate values by spaces */

t/t3404-rebase-interactive.sh

@@ -1384,7 +1384,7 @@ test_expect_success 'valid author header after --root swap' '
 	set_fake_editor &&
 	FAKE_LINES="2 1" git rebase -i --root &&
 	git cat-file commit HEAD^ >out &&
-	grep "^author ..*> @[0-9][0-9]* [-+][0-9][0-9]*$" out
+	grep "^author ..*> @[0-9][0-9]* [-+][0-9][0-9][0-9][0-9]$" out
 '
 
 test_done