Update buildbots to parse new telemetry JSON format.

scripts/slave/telemetry_utils.py

Index: scripts/slave/telemetry_utils.py
diff --git a/scripts/slave/telemetry_utils.py b/scripts/slave/telemetry_utils.py
new file mode 100644
index 0000000000000000000000000000000000000000..d683fef33404329843c8ecabeb0dfd27782ac4fc
--- /dev/null
+++ b/scripts/slave/telemetry_utils.py
@@ -0,0 +1,43 @@
+#! /usr/bin/env python
+# Copyright (c) 2014 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.
+
+"""Log parsing for telemetry tests."""
+
+import json
+import logging
+
+
+class TelemetryLogParser(object):
+
+  def __init__(self):
+    self._chart = None
+    self._ref_chart = None
+
+  def IsChartJson(self):
+    """This is the new telemetry --chartjson output format."""
+    return True
+
+  def Chart(self):
+    return self._chart
+
+  def RefChart(self):
+    return self._ref_chart
+
+  def ProcessLine(self, line):

[sullivan, 2014/09/05 04:30:00]

Here is sample output from running the test:
https://paste.googleplex.com/6080686586855424

It looks to me like the first test results are the build under test, and the
second (if any) are the ref build. Is that a reasonable assumption? Is there a
better way to get the two results?

[ghost stip (do not use), 2014/09/05 21:39:06]

I was under the impression that telemetry would drop its json to a file, instead
of spitting out to stdout. We're *strongly* discouraging log parsing, and
emitting the result json to a file is the easiest way to do this.

[sullivan, 2014/09/05 21:47:03]

Oops, I think I gave eakuefner some out-of-date advice here. Ethan, would you be
okay re-implementing the telemetry side to spit out files?

I assume we'll need to pass some argument for the filename to telemetry, right?
And a different one for the ref vs non-ref builds. What's the best way to do
that?

+    try:
+      results = json.loads(line.strip())
+    except ValueError:
+      return
+    if results.has_key('charts'):
+      if not self._chart:
+        self._chart = results
+      elif not self._ref_chart:
+        self._ref_chart = results
+      else:
+        logging.error('Too many result lines, dropping.')
+
+  def FailedTests(self):
+    # TODO(sullivan): Does this need to be implemented? Not in process_log_utils
+    return []