Fixed #26158 -- Rewrote http.parse_cookie() to better match browsers.

tests/httpwrappers/tests.py

@@ -676,6 +676,8 @@ class CookieTests(unittest.TestCase):
         c2 = SimpleCookie()
         c2.load(c.output()[12:])
         self.assertEqual(c['test'].value, c2['test'].value)
+        c3 = parse_cookie(c.output()[12:])
+        self.assertEqual(c['test'].value, c3['test'])
 
     def test_decode_2(self):
         """
@@ -686,6 +688,8 @@ class CookieTests(unittest.TestCase):
         c2 = SimpleCookie()
         c2.load(c.output()[12:])
         self.assertEqual(c['test'].value, c2['test'].value)
+        c3 = parse_cookie(c.output()[12:])
+        self.assertEqual(c['test'].value, c3['test'])
 
     def test_nonstandard_keys(self):
         """
@@ -699,6 +703,52 @@ class CookieTests(unittest.TestCase):
         """
         self.assertIn('good_cookie', parse_cookie('a:=b; a:=c; good_cookie=yes').keys())
 
+    def test_python_cookies(self):
+        """
+        Test cases copied from Python's Lib/test/test_http_cookies.py
+        """
+        self.assertEqual(parse_cookie('chips=ahoy; vienna=finger'), {'chips': 'ahoy', 'vienna': 'finger'})
+        # Here parse_cookie() differs from Python's cookie parsing in that it
+        # treats all semicolons as delimiters, even within quotes.
+        self.assertEqual(
+            parse_cookie('keebler="E=mc2; L=\\"Loves\\"; fudge=\\012;"'),
+            {'keebler': '"E=mc2', 'L': '\\"Loves\\"', 'fudge': '\\012', '': '"'}
+        )
+        # Illegal cookies that have an '=' char in an unquoted value.
+        self.assertEqual(parse_cookie('keebler=E=mc2'), {'keebler': 'E=mc2'})
+        # Cookies with ':' character in their name.
+        self.assertEqual(parse_cookie('key:term=value:term'), {'key:term': 'value:term'})
+        # Cookies with '[' and ']'.
+        self.assertEqual(parse_cookie('a=b; c=[; d=r; f=h'), {'a': 'b', 'c': '[', 'd': 'r', 'f': 'h'})
+
+    def test_cookie_edgecases(self):
+        # Cookies that RFC6265 allows.
+        self.assertEqual(parse_cookie('a=b; Domain=example.com'), {'a': 'b', 'Domain': 'example.com'})
+        # parse_cookie() has historically kept only the last cookie with the
+        # same name.
+        self.assertEqual(parse_cookie('a=b; h=i; a=c'), {'a': 'c', 'h': 'i'})
+
+    def test_invalid_cookies(self):
+        """
+        Cookie strings that go against RFC6265 but browsers will send if set
+        via document.cookie.
+        """
+        # Chunks without an equals sign appear as unnamed values per
+        # https://bugzilla.mozilla.org/show_bug.cgi?id=169091
+        self.assertIn('django_language', parse_cookie('abc=def; unnamed; django_language=en').keys())
+        # Even a double quote may be an unamed value.
+        self.assertEqual(parse_cookie('a=b; "; c=d'), {'a': 'b', '': '"', 'c': 'd'})
+        # Spaces in names and values, and an equals sign in values.
+        self.assertEqual(parse_cookie('a b c=d e = f; gh=i'), {'a b c': 'd e = f', 'gh': 'i'})
+        # More characters the spec forbids.
+        self.assertEqual(parse_cookie('a   b,c<>@:/[]?{}=d  "  =e,f g'), {'a   b,c<>@:/[]?{}': 'd  "  =e,f g'})
+        # Unicode characters. The spec only allows ASCII.
+        self.assertEqual(parse_cookie('saint=André Bessette'), {'saint': force_str('André Bessette')})
+        # Browsers don't send extra whitespace or semicolons in Cookie headers,
+        # but parse_cookie() should parse whitespace the same way
+        # document.cookie parses whitespace.
+        self.assertEqual(parse_cookie('  =  b  ;  ;  =  ;   c  =  ;  '), {'': 'b', 'c': ''})
+
     def test_httponly_after_load(self):
         """
         Test that we can use httponly attribute on cookies that we load