pythongh-80222: Fix email address header folding with long quoted-string

Email generators using email.policy.default could incorrectly omit the
quote ('"') characters from a quoted-string during header refolding,
leading to invalid address headers and enabling header spoofing. This
change restores the quote characters on a bare-quoted-string as the
header is refolded, and escapes backslash and quote chars in the string.

Lib/email/_header_value_parser.py

@@ -95,8 +95,16 @@
 NLSET = {'\n', '\r'}
 SPECIALSNL = SPECIALS | NLSET
 
+
+def escape_for_quotes(value):
+    """Escape dquote and backslash for use within a quoted-string."""
+    return str(value).replace('\\', '\\\\').replace('"', '\\"')
+
+
 def quote_string(value):
-    return '"'+str(value).replace('\\', '\\\\').replace('"', r'\"')+'"'
+    escaped = escape_for_quotes(value)
+    return f'"{escaped}"'
+
 
 # Match a RFC 2047 word, looks like =?utf-8?q?someword?=
 rfc2047_matcher = re.compile(r'''
@@ -2905,6 +2913,14 @@ def _refold_parse_tree(parse_tree, *, policy):
         if not hasattr(part, 'encode'):
             # It's not a terminal, try folding the subparts.
             newparts = list(part)
+            if part.token_type == 'bare-quoted-string':
+                # Restore the quotes and escape contents.
+                dquote = ValueTerminal('"', 'ptext')
+                newparts = (
+                    [dquote] +
+                    [ValueTerminal(escape_for_quotes(p), 'ptext')
+                     for p in newparts] +
+                    [dquote])
             if not part.as_ew_allowed:
                 wrap_as_ew_blocked += 1
                 newparts.append(end_ew_not_allowed)

Lib/test/test_email/test__header_value_parser.py

@@ -3082,13 +3082,40 @@ def test_address_list_with_list_separator_after_fold(self):
         self._test(parser.get_address_list(to)[0],
             f'{a},\n =?utf-8?q?H=C3=BCbsch?= Kaktus <[email protected]>\n')
 
-        a = '.' * 79
+        a = '.' * 79  # ('.' is a special, so must be in quoted-string.)
         to = f'"{a}" <[email protected]>, "Hübsch Kaktus" <[email protected]>'
         self._test(parser.get_address_list(to)[0],
-            f'{a}\n'
+            f'"{a}"\n'
             ' <[email protected]>, =?utf-8?q?H=C3=BCbsch?= Kaktus '
             '<[email protected]>\n')
 
+    def test_address_list_with_specials_in_long_quoted_string(self):
+        # Regression for gh-80222.
+        policy = self.policy.clone(max_line_length=40)
+        cases = [
+            # (to, folded)
+            ('"Exfiltrator <[email protected]> (unclosed comment?" <[email protected]>',
+             '"Exfiltrator <[email protected]> (unclosed\n'
+             ' comment?" <[email protected]>\n'),
+            ('"Escaped \\" chars \\\\ in quoted-string stay escaped" <[email protected]>',
+             '"Escaped \\" chars \\\\ in quoted-string\n'
+             ' stay escaped" <[email protected]>\n'),
+            ('This long display name does not need quotes <[email protected]>',
+             'This long display name does not need\n'
+             ' quotes <[email protected]>\n'),
+            ('"Quotes are not required but are retained here" <[email protected]>',
+             '"Quotes are not required but are\n'
+             ' retained here" <[email protected]>\n'),
+            ('"A quoted-string, it can be a valid local-part"@example.com',
+             '"A quoted-string, it can be a valid\n'
+             ' local-part"@example.com\n'),
+            ('"[email protected]"@example.com',
+             '"[email protected]"@example.com\n'),
+        ]
+        for (to, folded) in cases:
+            with self.subTest(to=to):
+                self._test(parser.get_address_list(to)[0], folded, policy=policy)
+
     # XXX Need tests with comments on various sides of a unicode token,
     # and with unicode tokens in the comments.  Spaces inside the quotes
     # currently don't do the right thing.

Misc/NEWS.d/next/Security/2024-08-06-11-43-08.gh-issue-80222.wfR4BU.rst

@@ -0,0 +1,3 @@
+Fix a problem where email.policy.default header refolding could incorrectly
+omit quotes from structured email headers, enabling sender or recipient
+spoofing via a carefully crafted display-name.