http.client: bound the number of interim 1xx responses skipped

HTTPResponse.begin() skipped 100 Continue responses in an unbounded
loop, so a server streaming them forever would hang getresponse()
regardless of any socket timeout.  At most 100 interim responses are
now skipped before HTTPException is raised.

(cherry picked from commit 508e86d8811e8e17350bf2c6fb23d3bb2d3ff795)

Author: gpshead

Lib/http/client.py

@@ -111,6 +111,13 @@
 _MAXLINE = 65536
 _MAXHEADERS = 100
 
+# maximal number of interim (1xx) responses tolerated before the final
+# response.  Real servers send at most a few; without a bound, a server
+# streaming "100 Continue" responses would hang getresponse() forever.
+# A socket timeout cannot detect that as data keeps arriving within every
+# timeout window.
+_MAXINTERIMRESPONSES = 100
+
 # Data larger than this will be read in chunks, to prevent extreme
 # overallocation.
 _MIN_READ_BUF_SIZE = 1 << 20
@@ -332,7 +339,7 @@ def begin(self):
             return
 
         # read until we get a non-100 response
-        while True:
+        for _ in range(_MAXINTERIMRESPONSES):
             version, status, reason = self._read_status()
             if status != CONTINUE:
                 break
@@ -341,6 +348,9 @@ def begin(self):
             if self.debuglevel > 0:
                 print("headers:", skipped_headers)
             del skipped_headers
+        else:
+            raise HTTPException(
+                f"got more than {_MAXINTERIMRESPONSES} interim responses")
 
         self.code = self.status = status
         self.reason = reason.strip()

Lib/test/test_httplib.py

@@ -1332,6 +1332,35 @@ def test_overflowing_header_limit_after_100(self):
         self.assertIn('got more than ', str(cm.exception))
         self.assertIn('headers', str(cm.exception))
 
+    def test_too_many_interim_responses(self):
+        # A server streaming "100 Continue" responses forever must not
+        # hang getresponse().
+        body = (
+            'HTTP/1.1 100 Continue\r\n\r\n'
+            * (client._MAXINTERIMRESPONSES + 1)
+        )
+        resp = client.HTTPResponse(FakeSocket(body))
+        with self.assertRaises(client.HTTPException) as cm:
+            resp.begin()
+        self.assertIn('got more than ', str(cm.exception))
+        self.assertIn('interim responses', str(cm.exception))
+
+    def test_multiple_interim_responses(self):
+        # A reasonable number of interim responses before the final
+        # response is skipped as before.
+        body = (
+            'HTTP/1.1 100 Continue\r\n\r\n' * 3 +
+            'HTTP/1.1 200 OK\r\n'
+            'Content-Length: 5\r\n'
+            '\r\n'
+            'hello'
+        )
+        resp = client.HTTPResponse(FakeSocket(body), method="GET")
+        resp.begin()
+        self.assertEqual(resp.status, 200)
+        self.assertEqual(resp.read(), b'hello')
+        resp.close()
+
     def test_overflowing_chunked_line(self):
         body = (
             'HTTP/1.1 200 OK\r\n'