GH-100884: email/_header_value_parser: don't encode list separators

ListSeparator should not be encoded. This could happen when a long line
pushes its separator to the next line, which would have been encoded.

Fixes #100884

Author: t-8ch

Lib/email/_header_value_parser.py

@@ -949,6 +949,7 @@ class _InvalidEwError(errors.HeaderParseError):
 # up other parse trees.  Maybe should have  tests for that, too.
 DOT = ValueTerminal('.', 'dot')
 ListSeparator = ValueTerminal(',', 'list-separator')
+ListSeparator.as_ew_allowed = False
 RouteComponentMarker = ValueTerminal('@', 'route-component-marker')
 
 #

Lib/test/test_email/test__header_value_parser.py

@@ -2946,6 +2946,11 @@ def test_address_list_with_unicode_names_in_quotes(self):
             '=?utf-8?q?H=C3=BCbsch?= Kaktus <[email protected]>,\n'
                 ' =?utf-8?q?bei=C3=9Ft_bei=C3=9Ft?= <[email protected]>\n')
 
+    def test_address_list_with_list_separator_after_fold(self):
+        to = '0123456789' * 8 + '@foo, ä <foo@bar>'
+        self._test(parser.get_address_list(to)[0],
+                   '0123456789' * 8 + '@foo,\n =?utf-8?q?=C3=A4?= <foo@bar>\n')
+
     # 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/Library/2023-01-09-14-08-02.gh-issue-100884.DcmdLl.rst

@@ -0,0 +1,2 @@
+email: fix misfolding of comma in address-lists over multiple lines in
+combination with unicode encoding.