Merge pull request #2286 from Keerthana-64/commit-encoding-header

fix(commit): parse the `encoding` header when deserializing

Author: Byron

git/objects/commit.py

@@ -896,8 +896,19 @@ def _deserialize(self, stream: BytesIO) -> "Commit":
         enc = next_line
         buf = enc.strip()
         while buf:
-            if buf[0:10] == b"encoding ":
-                self.encoding = buf[buf.find(b" ") + 1 :].decode(self.encoding, "ignore")
+            if buf[0:9] == b"encoding ":
+                raw_encoding = buf[9:]
+                try:
+                    # Codec names are ASCII, independent of the content encoding.
+                    encoding = raw_encoding.decode("ascii")
+                    # Probe one byte: decoding b"" skips the codec lookup entirely.
+                    b" ".decode(encoding, "replace")
+                except (LookupError, ValueError):
+                    # Reject malformed names and non-text codecs without changing
+                    # the current encoding (the default unless a valid header was read).
+                    _logger.error("Ignoring invalid commit encoding %r", raw_encoding)
+                else:
+                    self.encoding = encoding
             elif buf[0:7] == b"gpgsig ":
                 sig_lines = [buf[buf.find(b" ") + 1 :] + b"\n"]
                 is_next_header = False

test/test_commit.py

@@ -458,6 +458,61 @@ def test_invalid_commit(self):
         self.assertEqual(cmt.author.name, "E.Azer Ko�o�o�oculu", cmt.author.name)
         self.assertEqual(cmt.author.email, "azer@kodfabrik.com", cmt.author.email)
 
+    def test_deserialization_declared_encoding(self):
+        # The "encoding" header names the encoding of the identity and message bytes.
+        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"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.committer.name, "André")
+        self.assertEqual(cmt.message, "Résumé\n")
+
+    def test_deserialization_unknown_declared_encoding(self):
+        # An encoding Python has no text codec for falls back to the default encoding.
+        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"encoding pony-speak\n"
+            b"\n"
+            b"R\xe9sum\xe9\n"
+        )
+        cmt = self.rorepo.commit()
+        cmt._deserialize(BytesIO(data))
+
+        self.assertEqual(cmt.encoding, cmt.default_encoding)
+        self.assertEqual(cmt.author.name, "Andr�")
+        self.assertEqual(cmt.message, "R�sum�\n")
+
+    def test_deserialization_multiple_encoding_headers(self):
+        for headers in (
+            b"encoding utf-16-be\nencoding ISO-8859-1\n",
+            b"encoding ISO-8859-1\nencoding utf-\xff8\n",
+        ):
+            with self.subTest(headers=headers):
+                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" + headers + 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.committer.name, "André")
+                self.assertEqual(cmt.message, "Résumé\n")
+
     @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."""