Fix CI. Handle the special case of file://C:\\ parsed by url2pathname

function where ":" should not be treated as port separator.

Author: orsenthil

Lib/test/test_urllib.py

@@ -7,6 +7,7 @@
 import email.message
 import io
 import unittest
+import unittest.mock
 from test import support
 from test.support import os_helper
 from test.support import socket_helper
@@ -1678,6 +1679,17 @@ def test_url2pathname_resolve_host(self):
         self.assertEqual(fn(f'//{socket.gethostname()}/foo/bar'), f'{sep}foo{sep}bar')
         self.assertEqual(fn(f'//{socket.gethostname()}/foo/bar', resolve_host=True), f'{sep}foo{sep}bar')
 
+    def test_url2pathname_win_netloc_not_validated_as_port(self):
+        # The Windows branch of url2pathname() stuffs things like drive
+        # letters into the netloc (e.g. '//C:\\' -> authority 'C:\\'),
+        # which is not a real host[:port] pair. url2pathname() must not
+        # reject it as an invalid port; run the Windows branch directly
+        # so this is exercised on every platform, not just win32.
+        with unittest.mock.patch('os.name', 'nt'):
+            fn = urllib.request.url2pathname
+            self.assertEqual(fn('//C:\\'), 'C:\\')
+            self.assertEqual(fn('//C:80/'), 'C:80\\')
+
     @unittest.skipUnless(sys.platform == 'win32',
                          'test specific to Windows pathnames.')
     def test_url2pathname_win(self):

Lib/urllib/request.py

@@ -101,7 +101,8 @@
 from urllib.parse import (
     urlparse, urlsplit, urljoin, unwrap, quote, unquote,
     _splittype, _splithost, _splitport, _splituser, _splitpasswd,
-    _splitattr, _splitvalue, _splittag,
+    _splitattr, _splitvalue, _splittag, _splitnetloc,
+    _WHATWG_C0_CONTROL_OR_SPACE, _UNSAFE_URL_BYTES_TO_REMOVE,
     unquote_to_bytes, urlunparse)
 from urllib.response import addinfourl, addclosehook
 
@@ -1681,7 +1682,22 @@ def url2pathname(url, *, require_scheme=False, resolve_host=False):
     """
     if not require_scheme:
         url = 'file:' + url
-    scheme, authority, url = urlsplit(url)[:3]  # Discard query and fragment.
+    # Match urlsplit()'s own control-character handling up front, so the
+    # except branch below can reuse this same cleaned-up string.
+    url = url.lstrip(_WHATWG_C0_CONTROL_OR_SPACE)
+    for b in _UNSAFE_URL_BYTES_TO_REMOVE:
+        url = url.replace(b, '')
+    try:
+        scheme, authority, url = urlsplit(url)[:3]  # Discard query and fragment.
+    except ValueError:
+        # The authority here is not a real host[:port] pair -- it may
+        # hold a Windows drive letter (e.g. 'C:\\') -- so urlsplit()'s
+        # eager port check doesn't apply; split without it instead.
+        scheme, _, url = url.partition(':')
+        scheme = scheme.lower()
+        authority = ''
+        if url[:2] == '//':
+            authority, url = _splitnetloc(url, 2)
     if scheme != 'file':
         raise URLError("URL is missing a 'file:' scheme")
     if os.name == 'nt':

Misc/NEWS.d/next/Library/2026-10-11-13-08-07.gh-issue-88037.CzKcxh.rst

@@ -1,3 +1,3 @@
 :func:`urllib.parse.urlsplit` and :func:`urllib.parse.urlparse` now raise
 :exc:`ValueError` for an invalid port when the URL is parsed, instead of
-when the :attr:`!port` attribute is read. Patch by Miguel Brito.
+when the :attr:`!port` attribute is read. Patch by Miguel Brito and Senthil Kumaran.