Merge branch 'python/unicode_decode_errors' into 'master'

Python unicode decode errors when decrypting.

See merge request matrix-org/olm!4

python/Makefile

@@ -43,6 +43,9 @@ test: olm-python2 olm-python3
 	PYTHONPATH=install-temp/3 python3 -m pytest --cov --cov-branch --benchmark-disable
 	rm -rf install-temp
 
+isort:
+	isort -y -p olm
+
 clean:
 	rm -rf python_olm.egg-info/ dist/ __pycache__/
 	rm -rf *.so _libolm.o

python/olm/_compat.py

@@ -44,3 +44,24 @@ def to_bytes(string):
         return bytes(string, "utf-8")
 
     raise TypeError("Invalid type {}".format(type(string)))
+
+
+def to_unicode_str(byte_string, errors="replace"):
+    """Turn a byte string into a unicode string.
+
+    Should be used everywhere where the input byte string might not be trusted
+    and may contain invalid unicode values.
+
+    Args:
+        byte_string (bytes): The bytestring that will be converted to a native
+            string.
+        errors (str, optional): The error handling scheme that should be used
+            to handle unicode decode errors. Can be one of "strict" (raise an
+            UnicodeDecodeError exception, "ignore" (remove the offending
+            characters), "replace" (replace the offending character with
+            U+FFFD), "xmlcharrefreplace" as well as any other name registered
+            with codecs.register_error that can handle UnicodeEncodeErrors.
+
+    Returns the decoded native string.
+    """
+    return byte_string.decode(encoding="utf-8", errors=errors)

python/olm/group_session.py

@@ -33,7 +33,7 @@ from future.utils import bytes_to_native_str
 # pylint: disable=no-name-in-module
 from _libolm import ffi, lib  # type: ignore
 
-from ._compat import URANDOM, to_bytearray, to_bytes
+from ._compat import URANDOM, to_bytearray, to_bytes, to_unicode_str
 from ._finalize import track_for_finalization
 
 
@@ -176,8 +176,8 @@ class InboundGroupSession(object):
 
         raise OlmGroupSessionError(last_error)
 
-    def decrypt(self, ciphertext):
-        # type: (AnyStr) -> Tuple[str, int]
+    def decrypt(self, ciphertext, unicode_errors="replace"):
+        # type: (AnyStr, str) -> Tuple[str, int]
         """Decrypt a message
 
         Returns a tuple of the decrypted plain-text and the message index of
@@ -197,6 +197,13 @@ class InboundGroupSession(object):
         Args:
             ciphertext(str): Base64 encoded ciphertext containing the encrypted
                 message
+            unicode_errors(str, optional): The error handling scheme to use for
+                unicode decoding errors. The default is "replace" meaning that
+                the character that was unable to decode will be replaced with
+                the unicode replacement character (U+FFFD). Other possible
+                values are "strict", "ignore" and "xmlcharrefreplace" as well
+                as any other name registered with codecs.register_error that
+                can handle UnicodeEncodeErrors.
         """
         if not ciphertext:
             raise ValueError("Ciphertext can't be empty.")
@@ -223,10 +230,10 @@ class InboundGroupSession(object):
 
         self._check_error(plaintext_length)
 
-        plaintext = bytes_to_native_str(ffi.unpack(
-            plaintext_buffer,
-            plaintext_length
-        ))
+        plaintext = to_unicode_str(
+            ffi.unpack(plaintext_buffer, plaintext_length),
+            errors=unicode_errors
+        )
 
         # clear out copies of the plaintext
         lib.memset(plaintext_buffer, 0, max_plaintext_length)

python/olm/pk.py

@@ -40,7 +40,7 @@ from future.utils import bytes_to_native_str
 
 from _libolm import ffi, lib  # type: ignore
 
-from ._compat import URANDOM, to_bytearray
+from ._compat import URANDOM, to_bytearray, to_unicode_str
 from ._finalize import track_for_finalization
 
 
@@ -313,8 +313,8 @@ class PkDecryption(object):
 
         return obj
 
-    def decrypt(self, message):
-        # type (PkMessage) -> str
+    def decrypt(self, message, unicode_errors="replace"):
+        # type (PkMessage, str) -> str
         """Decrypt a previously encrypted Pk message.
 
         Returns the decrypted plaintext.
@@ -322,6 +322,13 @@ class PkDecryption(object):
 
         Args:
             message(PkMessage): the pk message to decrypt.
+            unicode_errors(str, optional): The error handling scheme to use for
+                unicode decoding errors. The default is "replace" meaning that
+                the character that was unable to decode will be replaced with
+                the unicode replacement character (U+FFFD). Other possible
+                values are "strict", "ignore" and "xmlcharrefreplace" as well
+                as any other name registered with codecs.register_error that
+                can handle UnicodeEncodeErrors.
         """
         ephemeral_key = to_bytearray(message.ephemeral_key)
         ephemeral_key_size = len(ephemeral_key)
@@ -354,7 +361,7 @@ class PkDecryption(object):
         # clear out copies of the plaintext
         lib.memset(plaintext_buffer, 0, max_plaintext_length)
 
-        return bytes_to_native_str(plaintext)
+        return to_unicode_str(plaintext, errors=unicode_errors)
 
 
 def _clear_pk_signing(pk_struct):

python/olm/sas.py

@@ -30,15 +30,15 @@ Examples:
 
 """
 
-from functools import wraps
 from builtins import bytes
+from functools import wraps
 from typing import Optional
 
 from future.utils import bytes_to_native_str
 
 from _libolm import ffi, lib
 
-from ._compat import URANDOM, to_bytes, to_bytearray
+from ._compat import URANDOM, to_bytearray, to_bytes
 from ._finalize import track_for_finalization
 
 

python/olm/session.py

@@ -40,7 +40,7 @@ from future.utils import bytes_to_native_str
 # pylint: disable=no-name-in-module
 from _libolm import ffi, lib  # type: ignore
 
-from ._compat import URANDOM, to_bytearray, to_bytes
+from ._compat import URANDOM, to_bytearray, to_bytes, to_unicode_str
 from ._finalize import track_for_finalization
 
 # This is imported only for type checking purposes
@@ -273,8 +273,8 @@ class Session(object):
         else:  # pragma: no cover
             raise ValueError("Unknown message type")
 
-    def decrypt(self, message):
-        # type: (_OlmMessage) -> str
+    def decrypt(self, message, unicode_errors="replace"):
+        # type: (_OlmMessage, str) -> str
         """Decrypts a message using the session. Returns the plaintext string
         on success. Raises OlmSessionError on failure. If the base64 couldn't
         be decoded then the error message will be "INVALID_BASE64". If the
@@ -285,7 +285,14 @@ class Session(object):
 
         Args:
             message(OlmMessage): The Olm message that will be decrypted. It can
-            be either an OlmPreKeyMessage or an OlmMessage.
+                be either an OlmPreKeyMessage or an OlmMessage.
+            unicode_errors(str, optional): The error handling scheme to use for
+                unicode decoding errors. The default is "replace" meaning that
+                the character that was unable to decode will be replaced with
+                the unicode replacement character (U+FFFD). Other possible
+                values are "strict", "ignore" and "xmlcharrefreplace" as well
+                as any other name registered with codecs.register_error that
+                can handle UnicodeEncodeErrors.
         """
         if not message.ciphertext:
             raise ValueError("Ciphertext can't be empty")
@@ -311,8 +318,10 @@ class Session(object):
             plaintext_buffer, max_plaintext_length
         )
         self._check_error(plaintext_length)
-        plaintext = bytes_to_native_str(
-            ffi.unpack(plaintext_buffer, plaintext_length))
+        plaintext = to_unicode_str(
+            ffi.unpack(plaintext_buffer, plaintext_length),
+            errors=unicode_errors
+        )
 
         # clear out copies of the plaintext
         lib.memset(plaintext_buffer, 0, max_plaintext_length)

python/olm/utility.py

@@ -32,6 +32,7 @@ Examples:
 
 # pylint: disable=redefined-builtin,unused-import
 from typing import AnyStr, Type
+
 from future.utils import bytes_to_native_str
 
 # pylint: disable=no-name-in-module

python/tests/group_session_test.py

+# -*- coding: utf-8 -*-
 import pytest
 
 from olm import InboundGroupSession, OlmGroupSessionError, OutboundGroupSession
@@ -112,3 +113,16 @@ class TestClass(object):
         outbound = OutboundGroupSession()
         inbound = InboundGroupSession(outbound.session_key)
         del inbound
+
+    def test_invalid_unicode_decrypt(self):
+        outbound = OutboundGroupSession()
+        inbound = InboundGroupSession(outbound.session_key)
+
+        text = outbound.encrypt(b"\xed")
+        plaintext, _ = inbound.decrypt(text)
+
+        print(plaintext)
+        assert plaintext == u"�"
+
+        plaintext, _ = inbound.decrypt(text, "ignore")
+        assert plaintext == ""

python/tests/pk_test.py

+# -*- coding: utf-8 -*-
 import pytest
 
 from olm import (PkDecryption, PkDecryptionError, PkEncryption, PkSigning,
@@ -55,3 +56,10 @@ class TestClass(object):
         message = "This statement is true"
         signature = signing.sign(message)
         ed25519_verify(signing.public_key, message, signature)
+
+    def test_invalid_unicode_decrypt(self):
+        decryption = PkDecryption()
+        encryption = PkEncryption(decryption.public_key)
+        message = encryption.encrypt(b"\xed")
+        plaintext = decryption.decrypt(message)
+        assert plaintext == u"�"

python/tests/session_test.py

+# -*- coding: utf-8 -*-
 import pytest
 
 from olm import (Account, InboundSession, OlmMessage, OlmPreKeyMessage,
@@ -141,3 +142,11 @@ class TestClass(object):
 
         new_message = new_session.encrypt(plaintext)
         assert bob_session.matches(new_message) is False
+
+    def test_invalid_unicode_decrypt(self):
+        alice, bob, session = self._create_session()
+        message = session.encrypt(b"\xed")
+
+        bob_session = InboundSession(bob, message)
+        plaintext = bob_session.decrypt(message)
+        assert plaintext == u"�"