Added integration test to Chrome Proxy to verify direct connection on remote site timeout

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 9c12daa1c4781d54ce840f90646d562c81ff11c0..54821f2d0b3e5530627099b8df41bccc0b06a979 100644
--- a/tools/chrome_proxy/integration_tests/chrome_proxy_metrics.py
+++ b/tools/chrome_proxy/integration_tests/chrome_proxy_metrics.py
@@ -872,6 +872,24 @@ class ChromeProxyMetric(network_metrics.NetworkMetric):
     results.AddValue(scalar.ScalarValue(
         results.current_page, 'succeeded_sum', 'count', succeeded))
 
+  def AddResultsForBypassOnTimeout(self, tab, results):
+    bypass_count = 0
+    tab.WaitForDocumentReadyStateToBeComplete(timeout=30)

[sclittle, 2016/08/09 21:31:31]

Is 30s long enough? Doesn't the origin site (e.g. your test page) take 60s to
respond? Wouldn't this depend on how long it would take for the proxy to timeout
and give up on the origin?

Could you also add a comment here explaining how long the timeout is expected to
be?

[Robert Ogden, 2016/08/09 21:56:22]

Comment added. 30 seconds comes out of a concern for user experience. The test
site takes 60s so that waiting 30s will force a timeout. Either way, for the
sake of the user, the proxy should bypass pretty quickly and let the user use a
direct connection.

[sclittle, 2016/08/09 22:12:03]

I'm worried about the negative case, e.g. what has to break for this test to
start failing, and how will it fail? 90s might be a better timeout here, so that
if the 60s-delayed response goes through, this might be easier to debug.

Also, >1 minute pageloads aren't uncommon on really really slow networks :). I'm
a little worried about flakiness here, e.g. we don't want the test to fail just
because the user's running on a slow network.

[Robert Ogden, 2016/08/09 22:17:55]

Good point. I'll make it 90 second timeout

+    for resp in self.IterResponses(tab):
+      if resp.HasChromeProxyViaHeader() and not resp.response.url.endswith(
+          'favicon.ico'):
+        r = resp.response
+        raise ChromeProxyMetricException, (
+            'Response for %s should not have via header after HTTP timeout.\n'
+            'Reponse: status=(%d)\nHeaders:\n %s' % (
+                r.url, r.status, r.headers))
+      elif not resp.response.url.endswith('favicon.ico'):
+        bypass_count += 1
+    if bypass_count == 0:
+      raise ChromeProxyMetricException('No pages were tested!')
+    results.AddValue(scalar.ScalarValue(
+        results.current_page, 'bypass', 'count', bypass_count))
+
 PROXIED = 'proxied'
 DIRECT = 'direct'