[3.6] bpo-38216, bpo-36274: Allow subclasses to separately override validation and encoding behavior (GH-16448) (GH-16462)

(cherry picked from commit 7774d78)

Co-authored-by: Jason R. Coombs <[email protected]>

Author: jaraco

Lib/http/client.py

@@ -1117,19 +1117,15 @@ def putrequest(self, method, url, skip_host=False,
         else:
             raise CannotSendRequest(self.__state)
 
-        # Save the method we use, we need it later in the response phase
+        # Save the method for use later in the response phase
         self._method = method
-        if not url:
-            url = '/'
-        # Prevent CVE-2019-9740.
-        match = _contains_disallowed_url_pchar_re.search(url)
-        if match:
-            raise InvalidURL(f"URL can't contain control characters. {url!r} "
-                             f"(found at least {match.group()!r})")
+
+        url = url or '/'
+        self._validate_path(url)
+
         request = '%s %s %s' % (method, url, self._http_vsn_str)
 
-        # Non-ASCII characters should have been eliminated earlier
-        self._output(request.encode('ascii'))
+        self._output(self._encode_request(request))
 
         if self._http_vsn == 11:
             # Issue some standard headers for better HTTP/1.1 compliance
@@ -1207,6 +1203,18 @@ def putrequest(self, method, url, skip_host=False,
             # For HTTP/1.0, the server will assume "not chunked"
             pass
 
+    def _encode_request(self, request):
+        # ASCII also helps prevent CVE-2019-9740.
+        return request.encode('ascii')
+
+    def _validate_path(self, url):
+        """Validate a url for putrequest."""
+        # Prevent CVE-2019-9740.
+        match = _contains_disallowed_url_pchar_re.search(url)
+        if match:
+            raise InvalidURL(f"URL can't contain control characters. {url!r} "
+                             f"(found at least {match.group()!r})")
+
     def putheader(self, header, *values):
         """Send a request header line to the server.
 

Lib/test/test_httplib.py

@@ -1132,6 +1132,34 @@ def run_server():
         thread.join()
         self.assertEqual(result, b"proxied data\n")
 
+    def test_putrequest_override_validation(self):
+        """
+        It should be possible to override the default validation
+        behavior in putrequest (bpo-38216).
+        """
+        class UnsafeHTTPConnection(client.HTTPConnection):
+            def _validate_path(self, url):
+                pass
+
+        conn = UnsafeHTTPConnection('example.com')
+        conn.sock = FakeSocket('')
+        conn.putrequest('GET', '/\x00')
+
+    def test_putrequest_override_encoding(self):
+        """
+        It should be possible to override the default encoding
+        to transmit bytes in another encoding even if invalid
+        (bpo-36274).
+        """
+        class UnsafeHTTPConnection(client.HTTPConnection):
+            def _encode_request(self, str_url):
+                return str_url.encode('utf-8')
+
+        conn = UnsafeHTTPConnection('example.com')
+        conn.sock = FakeSocket('')
+        conn.putrequest('GET', '/☃')
+
+
 class ExtendedReadTest(TestCase):
     """
     Test peek(), read1(), readline()
@@ -1256,6 +1284,7 @@ def test_peek_0(self):
         p = self.resp.peek(0)
         self.assertLessEqual(0, len(p))
 
+
 class ExtendedReadTestChunked(ExtendedReadTest):
     """
     Test peek(), read1(), readline() in chunked mode

Misc/NEWS.d/next/Library/2019-09-27-15-24-45.bpo-38216.-7yvZR.rst

@@ -0,0 +1,4 @@
+Allow the rare code that wants to send invalid http requests from the
+`http.client` library a way to do so.  The fixes for bpo-30458 led to
+breakage for some projects that were relying on this ability to test their
+own behavior in the face of bad requests.