From 38544cc2f27e20c9ff70b8095b9ab879f4be6771 Mon Sep 17 00:00:00 2001
From: "Gregory P. Smith" <greg@krypto.org>
Date: Thu, 22 Oct 2026 05:25:24 +0000
Subject: [PATCH] gh-161104: Check the http.client tunnel host after IDNA
 encoding

IDNA normalization maps U+00A0 and similar characters to an ASCII space,
so check the encoded host that is put into the CONNECT request line.
---
 Lib/http/client.py                                        | 8 ++++++--
 Lib/test/test_httplib.py                                  | 5 +++++
 .../2026-10-22-05-23-04.gh-issue-161104.rvNPjJ.rst        | 4 ++++
 3 files changed, 15 insertions(+), 2 deletions(-)
 create mode 100644 Misc/NEWS.d/next/Library/2026-10-22-05-23-04.gh-issue-161104.rvNPjJ.rst

diff --git a/Lib/http/client.py b/Lib/http/client.py
index fe67d0f7963f78..0d7300e51e41a4 100644
--- a/Lib/http/client.py
+++ b/Lib/http/client.py
@@ -1010,11 +1010,15 @@ def _wrap_ipv6(self, ip):
         return ip
 
     def _tunnel(self):
-        if _contains_disallowed_url_pchar_re.search(self._tunnel_host):
+        # Check the host as it is sent: the IDNA codec normalizes the name
+        # first (NFKC), which maps some non-ASCII characters to ASCII ones,
+        # e.g. U+00A0 (NO-BREAK SPACE) to a space.
+        tunnel_host = self._tunnel_host.encode("idna")
+        if _contains_disallowed_url_pchar_re.search(tunnel_host.decode("ascii")):
             raise ValueError('Tunnel host can\'t contain control characters %r'
                              % (self._tunnel_host,))
         connect = b"CONNECT %s:%d %s\r\n" % (
-            self._wrap_ipv6(self._tunnel_host.encode("idna")),
+            self._wrap_ipv6(tunnel_host),
             self._tunnel_port,
             self._http_vsn_str.encode("ascii"))
         headers = [connect]
diff --git a/Lib/test/test_httplib.py b/Lib/test/test_httplib.py
index 1e5ec0d4e43c27..6979bc8edc7fdc 100644
--- a/Lib/test/test_httplib.py
+++ b/Lib/test/test_httplib.py
@@ -405,6 +405,11 @@ def test_invalid_tunnel_host(self):
             'invalid.host\r\n',
             'invalid.host\x00',
             'invalid host',
+            # The IDNA codec normalizes these to an ASCII space
+            'invalid\xa0host',
+            'invalid\u3000host',
+            # ... and this one to a space and a combining character
+            'invalid\xa8host',
         )
         for tunnel_host in cases:
             with self.subTest(tunnel_host):
diff --git a/Misc/NEWS.d/next/Library/2026-10-22-05-23-04.gh-issue-161104.rvNPjJ.rst b/Misc/NEWS.d/next/Library/2026-10-22-05-23-04.gh-issue-161104.rvNPjJ.rst
new file mode 100644
index 00000000000000..acae526043a297
--- /dev/null
+++ b/Misc/NEWS.d/next/Library/2026-10-22-05-23-04.gh-issue-161104.rvNPjJ.rst
@@ -0,0 +1,4 @@
+Fix :meth:`http.client.HTTPConnection.set_tunnel` accepting a tunnel host
+containing a non-ASCII character that IDNA encoding maps to a space, such as
+U+00A0 NO-BREAK SPACE. Connecting now raises :exc:`ValueError`, as it already
+does for an ASCII space.