Integration test for bypassing the proxy for a single request

tools/chrome_proxy/integration_tests/chrome_proxy_metrics.py

Index: tools/chrome_proxy/integration_tests/chrome_proxy_metrics.py
diff --git a/tools/chrome_proxy/integration_tests/chrome_proxy_metrics.py b/tools/chrome_proxy/integration_tests/chrome_proxy_metrics.py
index 3f765bf68ad8d10e4cb2509427d6e9816c7a64e6..01d801b0fbd3d59eb0080e9a488556f1ab073659 100644
--- a/tools/chrome_proxy/integration_tests/chrome_proxy_metrics.py
+++ b/tools/chrome_proxy/integration_tests/chrome_proxy_metrics.py
@@ -235,6 +235,39 @@ class ChromeProxyMetric(network_metrics.NetworkMetric):
     results.AddValue(scalar.ScalarValue(
         results.current_page, 'bypass', 'count', bypass_count))
 
+  def AddResultsForBlockSingle(self, tab, results):
+    eligible_response_count = 0
+    bypass_count = 0
+    for resp in self.IterResponses(tab):
+      if resp.ShouldHaveChromeProxyViaHeader():
+        eligible_response_count += 1
+        if not resp.HasChromeProxyViaHeader():
+          bypass_count += 1
+
+    if tab:
+      info = GetProxyInfoFromNetworkInternals(tab)
+      if not info['enabled']:
+        raise ChromeProxyMetricException, (
+            'Chrome proxy should be enabled. proxy info: %s' % info)
+      self.VerifyBadProxies(info['badProxies'], [])
+
+    if eligible_response_count <= 1:

[bolian, 2014/09/16 17:04:06]

Question. Why there will be more than one?  Do you see one from DRP to instruct
bypass another from direct loading? I remember in the past, the first response
was not exposed at this level.

[sclittle, 2014/09/16 19:06:53]

There would be more than one because http://check.googlezip.net/blocksingle
loads the google 404 not found page, which has a google logo image on it. The
idea was to test that the response for http://check.googlezip.net/blocksingle
bypassed the DRP, but the image that was requested immediately after that
response was received did not bypass the DRP.

This check that more that one response was received may not be necessary, since
the test already checks that the bad proxy list is empty which implies that only
one response would bypass the DRP. The eligible_response_count check also makes
the test dependent on the 404 not found page having at least one subresource on
it, which might be bad. I can remove this eligible_response_count check if you
think it's unnecessary.

[bolian, 2014/09/16 20:56:00]

No, don't remove that. :) I wasn't aware that the page contains another
resource. Thanks for explaining this.

+      raise ChromeProxyMetricException, (
+          'There should be more than one DRP eligible response '
+          '(eligible_response_count=%d, bypass_count=%d)\n' % (
+              eligible_response_count, bypass_count))
+    elif bypass_count != 1:
+      raise ChromeProxyMetricException, (
+          'Exactly one response should be bypassed. '
+          '(eligible_response_count=%d, bypass_count=%d)\n' % (
+              eligible_response_count, bypass_count))
+    else:
+      results.AddValue(scalar.ScalarValue(
+          results.current_page, 'eligible_responses', 'count',
+          eligible_response_count))
+      results.AddValue(scalar.ScalarValue(
+          results.current_page, 'bypass', 'count', bypass_count))
+
   def AddResultsForSafebrowsing(self, tab, results):
     count = 0
     safebrowsing_count = 0