gh-150474: Fix RFC 2231 charset handling in the email header parser

Lib/email/_header_value_parser.py

@@ -2580,12 +2580,15 @@ def get_parameter(value):
             return param, value
     else:
         if token is not None:
-            for t in token:
+            # The charset is parsed as a Value child, but belongs to the
+            # parameter itself. Retype its extended-attrtext as plain attrtext.
+            charset = token[0]
+            for t in charset:
                 if t.token_type == 'extended-attrtext':
+                    t.token_type = 'attrtext'
                     break
-            t.token_type == 'attrtext'
-            appendto.append(t)
-            param.charset = t.value
+            appendto.append(charset)
+            param.charset = charset.stripped_value
         if value[0] != "'":
             raise errors.HeaderParseError("Expected RFC2231 char/lang encoding "
                                           "delimiter, but found {!r}".format(value))

Lib/test/test_email/test__header_value_parser.py

@@ -3030,6 +3030,37 @@ def test_parse_message_ids_broken_ang(self):
         self.assertEqual(message_ids.token_type, 'message-id-list')
 
 
+class Test_get_parameter(TestParserMixin, TestEmailBase):
+
+    def test_rfc2231_charset_is_retyped_as_attrtext(self):
+        # gh-150474: the charset of an initial extended parameter is moved
+        # out of the value and into the parameter itself, and while doing so
+        # its text is retyped as an attrtext token.
+        param = self._test_get_x(parser.get_parameter,
+            "title*=us-ascii'en'This%20is%20a%20test",
+            "title*=us-ascii'en'This%20is%20a%20test",
+            "title*=us-ascii'en'This%20is%20a%20test",
+            [],
+            '')
+        self.assertEqual(param.token_type, 'parameter')
+        self.assertEqual(param[3].token_type, 'attribute')
+        self.assertEqual(param[3][0].token_type, 'attrtext')
+        self.assertEqual(param.charset, 'us-ascii')
+
+    def test_rfc2231_charset_strips_cfws(self):
+        # gh-150474: CFWS around the charset is part of the parse tree, but
+        # must not end up in the charset itself.
+        param = self._test_get_x(parser.get_parameter,
+            "title*=us-ascii 'en'This%20is%20a%20test",
+            "title*=us-ascii 'en'This%20is%20a%20test",
+            "title*=us-ascii 'en'This%20is%20a%20test",
+            [],
+            '')
+        self.assertEqual(param[3].token_type, 'attribute')
+        self.assertEqual(param[3][0].token_type, 'attrtext')
+        self.assertEqual(param[3][1].token_type, 'cfws')
+        self.assertEqual(param.charset, 'us-ascii')
+
 
 @parameterize
 class Test_parse_mime_parameters(TestParserMixin, TestEmailBase):

Misc/NEWS.d/next/Library/2026-09-18-23-41-48.gh-issue-150474.uJ3cL-.rst

@@ -0,0 +1,4 @@
+Fix the parsing of the character set of :rfc:`2231` encoded parameters in the
+:mod:`email` header parser.  The character set is now correctly located in
+the parsed value, so whitespace or a comment following it is no longer
+included in the parameter's character set name.