Reland: Added a debug info UI for Blimp

blimp/net/blimp_connection_statistics.h

Index: blimp/net/blimp_connection_statistics.h
diff --git a/blimp/net/blimp_connection_statistics.h b/blimp/net/blimp_connection_statistics.h
new file mode 100644
index 0000000000000000000000000000000000000000..fefd96ddcd8439d0172d16394c47580f18ce6592
--- /dev/null
+++ b/blimp/net/blimp_connection_statistics.h
@@ -0,0 +1,49 @@
+// Copyright 2016 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.
+
+#ifndef BLIMP_NET_BLIMP_CONNECTION_STATISTICS_H_
+#define BLIMP_NET_BLIMP_CONNECTION_STATISTICS_H_
+
+#include <map>
+#include "base/macros.h"
+#include "base/metrics/sparse_histogram.h"
+#include "blimp/net/blimp_net_export.h"
+
+namespace blimp {
+
+// Collects network traffic statistics. Maintains a counter for number of
+// completed commits, bytes sent and received over network and notifies its
+// observers on the main thread.
+// This class is supposed to live an entire session, created and destroyed along
+// it. This class is thread-safe.

[Wez, 2016/06/07 00:30:13]

I can't find anything in SparseHistogram that actually says it's thread-safe. 
It's also not clear that you need thread-safety - we only update the network
statistics on the network I/O thread, so provided we can cope with fetching
statistics asynchronously, we can have the histograms updated on that thread and
use PostTaskAndReply() to request a snapshot from which to get the values we
want.

AFAICT this class just creates SparseHistogram with specific flags, and defines
three event types - can't we do just do that directly?

+class BLIMP_NET_EXPORT BlimpConnectionStatistics {
+ public:
+  enum EventType {
+    BYTES_SENT = 0,
+    BYTES_RECEIVED = 1,
+    COMMIT = 2,
+  };
+
+  BlimpConnectionStatistics();
+
+  ~BlimpConnectionStatistics();
+
+  // Accumulates values for a metric.
+  void Add(EventType type, int data);
+
+  // Returns value for a metric.
+  int Get(EventType type);
+
+ private:
+  // A histogram to hold the metrics. It will have one sample for each type of
+  // metric which will be accumulated on each subsequent call. Histograms are
+  // intentionally leaked at the process termination.

[Kevin M, 2016/05/26 23:15:30]

???

Why do we leak them? Is there another task that cleans them up? If so, what is
it?

[Wez, 2016/06/07 00:30:12]

Did this ever get answered?

+  base::HistogramBase* histogram_;
+
+  DISALLOW_COPY_AND_ASSIGN(BlimpConnectionStatistics);
+};
+
+}  // namespace blimp
+
+#endif  // BLIMP_NET_BLIMP_CONNECTION_STATISTICS_H_