Implemented telemetry benchmark that loads page and outputs screenshot.

tools/perf/measurements/pixel_diff_unittest.py

+# Copyright 2017 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.
+import shutil

[wkorman, 2017/06/07 20:15:12]

+1 blank above

+import tempfile
+
+from telemetry.testing import options_for_unittests
+from telemetry.testing import page_test_test_case
+
+from measurements import pixel_diff
+
+class PixelDiffUnitTest(page_test_test_case.PageTestTestCase):
+
+  def setUp(self):
+    self._options = options_for_unittests.GetCopy()
+    self._png_outdir = tempfile.mkdtemp('_png_test')
+
+  def tearDown(self):
+    shutil.rmtree(self._png_outdir)
+
+  def testPixelDiff(self):
+    ps = self.CreateStorySetFromFileInUnittestDataDir('blank.html')

[wkorman, 2017/06/07 20:15:12]

Rename ps -> page_set for better understandability (even though I see it's 'ps'
in the base test class too)

+    measurement = pixel_diff.PixelDiff(self._png_outdir)
+    results = self.RunMeasurement(measurement, ps, options=self._options)
+
+    # Screenshots are not supported on all platforms

[wkorman, 2017/06/07 20:15:12]

We should make sure this test fails somehow if for example we somehow get into a
state, due to subsequent change by someone else, where screenshots are not
supported on any platform.

I could be convinced that it's ok given our primary focus of having this run on
linux to semi hard code this test to check whatever is appropriate to ensure
that, on linux, screenshots work.

Also I wanted to understand how this actually screenshots for unit test purposes
since I don't think it's actually bringing up a content shell or browser. Where
is the fake that for at least one platform pretends it really took a screenshot?

Update from in person discussion: nednguyen@ it looks like PageTestTestCase is
actually more of a base class for integration tests because it does actually
start up a browser. Just wanted to confirm this is reasonable and expected.

+    if results.failures:
+      assert 'not supported' in results.failures[0].exc_info[1].message
+      return
+
+    saved_picture_count = results.FindAllPageSpecificValuesNamed(
+        'saved_picture_count')
+    self.assertEquals(len(saved_picture_count), 1)
+    self.assertGreater(saved_picture_count[0].GetRepresentativeNumber(), 0)