fix(commit): end commit headers only at an empty line

git/objects/commit.py

@@ -892,10 +892,13 @@ def _deserialize(self, stream: BytesIO) -> "Commit":
         self.encoding = self.default_encoding
         self.gpgsig = ""
 
-        # Read headers.
-        enc = next_line
-        buf = enc.strip()
-        while buf:
+        # Read headers. Like git, end them only at an empty line. A line holding only
+        # whitespace (such as "\r" or a bare continuation space) is still a header
+        # line, and only trailing whitespace is stripped, so a continuation line is
+        # never read as a header of its own.
+        line = next_line
+        while line not in (b"", b"\n"):
+            buf = line.rstrip()
             if buf[0:9] == b"encoding ":
                 raw_encoding = buf[9:]
                 try:
@@ -917,15 +920,15 @@ def _deserialize(self, stream: BytesIO) -> "Commit":
                     if not sigbuf:
                         break
                     if sigbuf[0:1] != b" ":
-                        buf = sigbuf.strip()
+                        line = sigbuf
                         is_next_header = True
                         break
                     sig_lines.append(sigbuf[1:])
                 # END read all signature
                 self.gpgsig = b"".join(sig_lines).rstrip(b"\n").decode(self.encoding, "ignore")
                 if is_next_header:
                     continue
-            buf = readline().strip()
+            line = readline()
 
         # Decode the author's name.
         try:

test/test_commit.py

@@ -513,6 +513,28 @@ def test_deserialization_multiple_encoding_headers(self):
                 self.assertEqual(cmt.committer.name, "André")
                 self.assertEqual(cmt.message, "Résumé\n")
 
+    def test_deserialization_headers_end_at_empty_line(self):
+        # Git ends the headers only at an empty line. A line holding only whitespace
+        # belongs to the headers, so what follows it is not the message.
+        for blank in (b"\r\n", b" \n", b"\t\n"):
+            with self.subTest(blank=blank):
+                data = (
+                    b"tree 3385f7b1e5f91930a3bd3e07105e113fca29981d\n"
+                    b"author Andr\xe9 <andre@example.com> 1112591163 +0200\n"
+                    b"committer Andr\xe9 <andre@example.com> 1112591163 +0200\n"
+                    b"x-note a\n" + blank + b"Co-authored-by: Mallory <mallory@example.com>\n"
+                    b"encoding ISO-8859-1\n"
+                    b"\n"
+                    b"R\xe9sum\xe9\n"
+                )
+                cmt = self.rorepo.commit()
+                cmt._deserialize(BytesIO(data))
+
+                self.assertEqual(cmt.encoding, "ISO-8859-1")
+                self.assertEqual(cmt.author.name, "André")
+                self.assertEqual(cmt.message, "Résumé\n")
+                self.assertEqual(cmt.co_authors, [])
+
     @with_rw_directory
     def test_identity_cannot_alter_headers(self, rw_dir):
         """A name or email must not add header lines or present another identity."""