mirror of https://github.com/python/cpython
321 lines
11 KiB
Python
321 lines
11 KiB
Python
import textwrap
|
|
import unittest
|
|
import contextlib
|
|
from email import policy
|
|
from email import errors
|
|
from test.test_email import TestEmailBase
|
|
|
|
|
|
class TestDefectsBase:
|
|
|
|
policy = policy.default
|
|
raise_expected = False
|
|
|
|
@contextlib.contextmanager
|
|
def _raise_point(self, defect):
|
|
yield
|
|
|
|
def test_same_boundary_inner_outer(self):
|
|
source = textwrap.dedent("""\
|
|
Subject: XX
|
|
From: xx@xx.dk
|
|
To: XX
|
|
Mime-version: 1.0
|
|
Content-type: multipart/mixed;
|
|
boundary="MS_Mac_OE_3071477847_720252_MIME_Part"
|
|
|
|
--MS_Mac_OE_3071477847_720252_MIME_Part
|
|
Content-type: multipart/alternative;
|
|
boundary="MS_Mac_OE_3071477847_720252_MIME_Part"
|
|
|
|
--MS_Mac_OE_3071477847_720252_MIME_Part
|
|
Content-type: text/plain; charset="ISO-8859-1"
|
|
Content-transfer-encoding: quoted-printable
|
|
|
|
text
|
|
|
|
--MS_Mac_OE_3071477847_720252_MIME_Part
|
|
Content-type: text/html; charset="ISO-8859-1"
|
|
Content-transfer-encoding: quoted-printable
|
|
|
|
<HTML></HTML>
|
|
|
|
--MS_Mac_OE_3071477847_720252_MIME_Part--
|
|
|
|
--MS_Mac_OE_3071477847_720252_MIME_Part
|
|
Content-type: image/gif; name="xx.gif";
|
|
Content-disposition: attachment
|
|
Content-transfer-encoding: base64
|
|
|
|
Some removed base64 encoded chars.
|
|
|
|
--MS_Mac_OE_3071477847_720252_MIME_Part--
|
|
|
|
""")
|
|
# XXX better would be to actually detect the duplicate.
|
|
with self._raise_point(errors.StartBoundaryNotFoundDefect):
|
|
msg = self._str_msg(source)
|
|
if self.raise_expected: return
|
|
inner = msg.get_payload(0)
|
|
self.assertTrue(hasattr(inner, 'defects'))
|
|
self.assertEqual(len(self.get_defects(inner)), 1)
|
|
self.assertIsInstance(self.get_defects(inner)[0],
|
|
errors.StartBoundaryNotFoundDefect)
|
|
|
|
def test_multipart_no_boundary(self):
|
|
source = textwrap.dedent("""\
|
|
Date: Fri, 6 Apr 2001 09:23:06 -0800 (GMT-0800)
|
|
From: foobar
|
|
Subject: broken mail
|
|
MIME-Version: 1.0
|
|
Content-Type: multipart/report; report-type=delivery-status;
|
|
|
|
--JAB03225.986577786/zinfandel.lacita.com
|
|
|
|
One part
|
|
|
|
--JAB03225.986577786/zinfandel.lacita.com
|
|
Content-Type: message/delivery-status
|
|
|
|
Header: Another part
|
|
|
|
--JAB03225.986577786/zinfandel.lacita.com--
|
|
""")
|
|
with self._raise_point(errors.NoBoundaryInMultipartDefect):
|
|
msg = self._str_msg(source)
|
|
if self.raise_expected: return
|
|
self.assertIsInstance(msg.get_payload(), str)
|
|
self.assertEqual(len(self.get_defects(msg)), 2)
|
|
self.assertIsInstance(self.get_defects(msg)[0],
|
|
errors.NoBoundaryInMultipartDefect)
|
|
self.assertIsInstance(self.get_defects(msg)[1],
|
|
errors.MultipartInvariantViolationDefect)
|
|
|
|
multipart_msg = textwrap.dedent("""\
|
|
Date: Wed, 14 Nov 2007 12:56:23 GMT
|
|
From: foo@bar.invalid
|
|
To: foo@bar.invalid
|
|
Subject: Content-Transfer-Encoding: base64 and multipart
|
|
MIME-Version: 1.0
|
|
Content-Type: multipart/mixed;
|
|
boundary="===============3344438784458119861=="{}
|
|
|
|
--===============3344438784458119861==
|
|
Content-Type: text/plain
|
|
|
|
Test message
|
|
|
|
--===============3344438784458119861==
|
|
Content-Type: application/octet-stream
|
|
Content-Transfer-Encoding: base64
|
|
|
|
YWJj
|
|
|
|
--===============3344438784458119861==--
|
|
""")
|
|
|
|
def test_multipart_invalid_cte(self):
|
|
with self._raise_point(
|
|
errors.InvalidMultipartContentTransferEncodingDefect):
|
|
msg = self._str_msg(
|
|
self.multipart_msg.format(
|
|
"\nContent-Transfer-Encoding: base64"))
|
|
if self.raise_expected: return
|
|
self.assertEqual(len(self.get_defects(msg)), 1)
|
|
self.assertIsInstance(self.get_defects(msg)[0],
|
|
errors.InvalidMultipartContentTransferEncodingDefect)
|
|
|
|
def test_multipart_no_cte_no_defect(self):
|
|
if self.raise_expected: return
|
|
msg = self._str_msg(self.multipart_msg.format(''))
|
|
self.assertEqual(len(self.get_defects(msg)), 0)
|
|
|
|
def test_multipart_valid_cte_no_defect(self):
|
|
if self.raise_expected: return
|
|
for cte in ('7bit', '8bit', 'BINary'):
|
|
msg = self._str_msg(
|
|
self.multipart_msg.format("\nContent-Transfer-Encoding: "+cte))
|
|
self.assertEqual(len(self.get_defects(msg)), 0, "cte="+cte)
|
|
|
|
def test_lying_multipart(self):
|
|
source = textwrap.dedent("""\
|
|
From: "Allison Dunlap" <xxx@example.com>
|
|
To: yyy@example.com
|
|
Subject: 64423
|
|
Date: Sun, 11 Jul 2004 16:09:27 -0300
|
|
MIME-Version: 1.0
|
|
Content-Type: multipart/alternative;
|
|
|
|
Blah blah blah
|
|
""")
|
|
with self._raise_point(errors.NoBoundaryInMultipartDefect):
|
|
msg = self._str_msg(source)
|
|
if self.raise_expected: return
|
|
self.assertTrue(hasattr(msg, 'defects'))
|
|
self.assertEqual(len(self.get_defects(msg)), 2)
|
|
self.assertIsInstance(self.get_defects(msg)[0],
|
|
errors.NoBoundaryInMultipartDefect)
|
|
self.assertIsInstance(self.get_defects(msg)[1],
|
|
errors.MultipartInvariantViolationDefect)
|
|
|
|
def test_missing_start_boundary(self):
|
|
source = textwrap.dedent("""\
|
|
Content-Type: multipart/mixed; boundary="AAA"
|
|
From: Mail Delivery Subsystem <xxx@example.com>
|
|
To: yyy@example.com
|
|
|
|
--AAA
|
|
|
|
Stuff
|
|
|
|
--AAA
|
|
Content-Type: message/rfc822
|
|
|
|
From: webmaster@python.org
|
|
To: zzz@example.com
|
|
Content-Type: multipart/mixed; boundary="BBB"
|
|
|
|
--BBB--
|
|
|
|
--AAA--
|
|
|
|
""")
|
|
# The message structure is:
|
|
#
|
|
# multipart/mixed
|
|
# text/plain
|
|
# message/rfc822
|
|
# multipart/mixed [*]
|
|
#
|
|
# [*] This message is missing its start boundary
|
|
with self._raise_point(errors.StartBoundaryNotFoundDefect):
|
|
outer = self._str_msg(source)
|
|
if self.raise_expected: return
|
|
bad = outer.get_payload(1).get_payload(0)
|
|
self.assertEqual(len(self.get_defects(bad)), 1)
|
|
self.assertIsInstance(self.get_defects(bad)[0],
|
|
errors.StartBoundaryNotFoundDefect)
|
|
|
|
def test_first_line_is_continuation_header(self):
|
|
with self._raise_point(errors.FirstHeaderLineIsContinuationDefect):
|
|
msg = self._str_msg(' Line 1\nSubject: test\n\nbody')
|
|
if self.raise_expected: return
|
|
self.assertEqual(msg.keys(), ['Subject'])
|
|
self.assertEqual(msg.get_payload(), 'body')
|
|
self.assertEqual(len(self.get_defects(msg)), 1)
|
|
self.assertDefectsEqual(self.get_defects(msg),
|
|
[errors.FirstHeaderLineIsContinuationDefect])
|
|
self.assertEqual(self.get_defects(msg)[0].line, ' Line 1\n')
|
|
|
|
def test_missing_header_body_separator(self):
|
|
# Our heuristic if we see a line that doesn't look like a header (no
|
|
# leading whitespace but no ':') is to assume that the blank line that
|
|
# separates the header from the body is missing, and to stop parsing
|
|
# headers and start parsing the body.
|
|
with self._raise_point(errors.MissingHeaderBodySeparatorDefect):
|
|
msg = self._str_msg('Subject: test\nnot a header\nTo: abc\n\nb\n')
|
|
if self.raise_expected: return
|
|
self.assertEqual(msg.keys(), ['Subject'])
|
|
self.assertEqual(msg.get_payload(), 'not a header\nTo: abc\n\nb\n')
|
|
self.assertDefectsEqual(self.get_defects(msg),
|
|
[errors.MissingHeaderBodySeparatorDefect])
|
|
|
|
def test_bad_padding_in_base64_payload(self):
|
|
source = textwrap.dedent("""\
|
|
Subject: test
|
|
MIME-Version: 1.0
|
|
Content-Type: text/plain; charset="utf-8"
|
|
Content-Transfer-Encoding: base64
|
|
|
|
dmk
|
|
""")
|
|
msg = self._str_msg(source)
|
|
with self._raise_point(errors.InvalidBase64PaddingDefect):
|
|
payload = msg.get_payload(decode=True)
|
|
if self.raise_expected: return
|
|
self.assertEqual(payload, b'vi')
|
|
self.assertDefectsEqual(self.get_defects(msg),
|
|
[errors.InvalidBase64PaddingDefect])
|
|
|
|
def test_invalid_chars_in_base64_payload(self):
|
|
source = textwrap.dedent("""\
|
|
Subject: test
|
|
MIME-Version: 1.0
|
|
Content-Type: text/plain; charset="utf-8"
|
|
Content-Transfer-Encoding: base64
|
|
|
|
dm\x01k===
|
|
""")
|
|
msg = self._str_msg(source)
|
|
with self._raise_point(errors.InvalidBase64CharactersDefect):
|
|
payload = msg.get_payload(decode=True)
|
|
if self.raise_expected: return
|
|
self.assertEqual(payload, b'vi')
|
|
self.assertDefectsEqual(self.get_defects(msg),
|
|
[errors.InvalidBase64CharactersDefect])
|
|
|
|
def test_missing_ending_boundary(self):
|
|
source = textwrap.dedent("""\
|
|
To: 1@harrydomain4.com
|
|
Subject: Fwd: 1
|
|
MIME-Version: 1.0
|
|
Content-Type: multipart/alternative;
|
|
boundary="------------000101020201080900040301"
|
|
|
|
--------------000101020201080900040301
|
|
Content-Type: text/plain; charset=ISO-8859-1
|
|
Content-Transfer-Encoding: 7bit
|
|
|
|
Alternative 1
|
|
|
|
--------------000101020201080900040301
|
|
Content-Type: text/html; charset=ISO-8859-1
|
|
Content-Transfer-Encoding: 7bit
|
|
|
|
Alternative 2
|
|
|
|
""")
|
|
with self._raise_point(errors.CloseBoundaryNotFoundDefect):
|
|
msg = self._str_msg(source)
|
|
if self.raise_expected: return
|
|
self.assertEqual(len(msg.get_payload()), 2)
|
|
self.assertEqual(msg.get_payload(1).get_payload(), 'Alternative 2\n')
|
|
self.assertDefectsEqual(self.get_defects(msg),
|
|
[errors.CloseBoundaryNotFoundDefect])
|
|
|
|
|
|
class TestDefectDetection(TestDefectsBase, TestEmailBase):
|
|
|
|
def get_defects(self, obj):
|
|
return obj.defects
|
|
|
|
|
|
class TestDefectCapture(TestDefectsBase, TestEmailBase):
|
|
|
|
class CapturePolicy(policy.EmailPolicy):
|
|
captured = None
|
|
def register_defect(self, obj, defect):
|
|
self.captured.append(defect)
|
|
|
|
def setUp(self):
|
|
self.policy = self.CapturePolicy(captured=list())
|
|
|
|
def get_defects(self, obj):
|
|
return self.policy.captured
|
|
|
|
|
|
class TestDefectRaising(TestDefectsBase, TestEmailBase):
|
|
|
|
policy = TestDefectsBase.policy
|
|
policy = policy.clone(raise_on_defect=True)
|
|
raise_expected = True
|
|
|
|
@contextlib.contextmanager
|
|
def _raise_point(self, defect):
|
|
with self.assertRaises(defect):
|
|
yield
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|