Skip generation of stats, since code is now shared across threads

net/base/connection_type_histograms.cc

 // Copyright (c) 2009 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.
 
 #include "net/base/connection_type_histograms.h"
 
 #include "base/metrics/histogram.h"
+#include "content/browser/browser_thread.h"
 
 namespace net {
 
 // We're using a histogram as a group of counters.  We're only interested in
 // the values of the counters.  Ignore the shape, average, and standard
 // deviation of the histograms because they are meaningless.
 //
 // We use two groups of counters.  In the first group (counter1), each counter
 // is a boolean (0 or 1) that indicates whether the user has seen a connection
 // of that type during that session.  In the second group (counter2), each
 // counter is the number of connections of that type the user has seen during
 // that session.
 //
 // Each histogram has an unused bucket at the end to allow seamless future
 // expansion.
 void UpdateConnectionTypeHistograms(ConnectionType type) {
   static bool had_connection_type[NUM_OF_CONNECTION_TYPES];
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));

[willchan no longer on Chromium, 2011/03/04 19:15:34]

Unfortunately, net/ is not aware of browser stuff. This is a layering violation.

Oh, I see the bug thread now. The sync code is running from a different thread
(the sync thread) and it uses TCPClientSocketLibevent. So yes, I guess this code
here is definitely racy.

 
   if (type >= 0 && type < NUM_OF_CONNECTION_TYPES) {
     if (!had_connection_type[type]) {
       had_connection_type[type] = true;
       UMA_HISTOGRAM_ENUMERATION("Net.HadConnectionType3",
           type, NUM_OF_CONNECTION_TYPES);
     }
 
     UMA_HISTOGRAM_ENUMERATION("Net.ConnectionTypeCount3",
         type, NUM_OF_CONNECTION_TYPES);
   } else {
     NOTREACHED();  // Someone's logging an invalid type!
   }
 }
 
 }  // namespace net