Implement the Chrome Proxy bypass integration test with ChromeDriver

tools/chrome_proxy/webdriver/bypass.py

Index: tools/chrome_proxy/webdriver/bypass.py
diff --git a/tools/chrome_proxy/webdriver/bypass.py b/tools/chrome_proxy/webdriver/bypass.py
new file mode 100644
index 0000000000000000000000000000000000000000..9c7355aa6fa4ac1e98c50f2887e66c7b1339e9e8
--- /dev/null
+++ b/tools/chrome_proxy/webdriver/bypass.py
@@ -0,0 +1,22 @@
+# Copyright 2017 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+import common
+from common import TestDriver
+from common import IntegrationTest
+
+
+class Bypass(IntegrationTest):
+
+  # Ensure Chrome does not use Data Saver for block=0, which uses the default
+  # proxy retry delay.
+  def testBypass(self):
+    with TestDriver() as t:
+      t.AddChromeArg('--enable-spdy-proxy-auth')
+      t.LoadURL('http://check.googlezip.net/block/')
+      for response in t.GetHTTPResponses():

[RyanSturm, 2017/01/31 19:48:23]

tl;dr: Should we check if the CP header is present in request headers? Should we
call LoadURL on a different check.googlezip.net page afterwards to verify that
it does not use DRP?

If I understand this test correctly (which I might not), when LoadURL is called,
subsequent requests will not be attempted via DRP for 1-5 minutes (randomly in
that interval), and this request will be retried without DRP.

If that is the case, it might be worth also checking if checking if the CP
header is also not in the request headers.

Since the main frame request is retried with two different sets of headers, I am
guessing that only the second set of headers is considered when fetching the
headers. Since not having CP header implies not having via in the CP header, you
could replace the assertNotHasChromeProxyViaHeader.

The syntax would be something like:

self.assert(response['request_headers']['chrome-proxy'])

In the same vein, it might be  worth adding a check to make sure we bypass other
navigations that happen within a minute of this one, so maybe
LoadURL("http://check.googlezip.net/test.html") and verify that the CP header is
not present.

[RyanSturm, 2017/01/31 20:39:10]

Can you check if the retried request is treated as a separate request/response?
If so, maybe just leave the Via header checks and add a subsequent navigation
that also just checks the via header in the response.

+        self.assertNotHasChromeProxyViaHeader(response)
+
+if __name__ == '__main__':
+  IntegrationTest.RunAllTests()