Add script which gathers JSON output from a Swarming task's shards.

content/test/gpu/gather_swarming_json_results.py

Index: content/test/gpu/gather_swarming_json_results.py
diff --git a/content/test/gpu/gather_swarming_json_results.py b/content/test/gpu/gather_swarming_json_results.py
new file mode 100755
index 0000000000000000000000000000000000000000..f46d8b5ad053c2f8812f13423f52e4b30b5e0422
--- /dev/null
+++ b/content/test/gpu/gather_swarming_json_results.py
@@ -0,0 +1,231 @@
+#!/usr/bin/env python
+# 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.
+
+"""Script which gathers and merges the JSON results from multiple
+swarming shards of a step on the waterfall.
+
+This is used to feed in the per-test times of previous runs of tests
+to the browser_test_runner's sharding algorithm, to improve shard
+distribution.
+"""
+
+import argparse
+import json
+import os
+import shutil
+import subprocess
+import sys
+import tempfile
+import urllib
+import urllib2
+
+SWARMING_SERVICE = 'https://chromium-swarm.appspot.com'
+
+THIS_DIR = os.path.dirname(os.path.abspath(__file__))
+SRC_DIR = os.path.dirname(os.path.dirname(os.path.dirname(THIS_DIR)))
+SWARMING_CLIENT_DIR = os.path.join(SRC_DIR, 'tools', 'swarming_client')
+
+class Swarming:
+  @staticmethod
+  def CheckAuth():
+    output = subprocess.check_output([
+      os.path.join(SWARMING_CLIENT_DIR, 'auth.py'),
+      'check',
+      '--service',
+      SWARMING_SERVICE])
+    if not output.startswith('user:'):
+      print 'Must run:'
+      print '  tools/swarming_client/auth.py login --service ' + \
+          SWARMING_SERVICE
+      print 'and authenticate with @google.com credentials.'
+      sys.exit(1)
+
+  @staticmethod
+  def Collect(hashes, output_dir, verbose):
+    cmd = [
+      os.path.join(SWARMING_CLIENT_DIR, 'swarming.py'),
+      'collect',
+      '-S',
+      SWARMING_SERVICE,
+      '--task-output-dir',
+      output_dir] + hashes
+    if verbose:
+      print 'Collecting Swarming results:'
+      print cmd
+    if verbose > 1:
+      # Print stdout from the collect command.
+      stdout = None
+    else:
+      fnull = open(os.devnull, 'w')
+      stdout = fnull
+    subprocess.check_call(cmd, stdout=stdout, stderr=subprocess.STDOUT)
+
+  @staticmethod
+  def ExtractShardHashes(urls):
+    SWARMING_URL = 'https://chromium-swarm.appspot.com/user/task/'
+    hashes = []

[Vadim Sh., 2016/07/02 01:19:51]

technically these are not hashes but task IDs

[Ken Russell (switch to Gerrit), 2016/07/02 03:17:01]

Thanks, renamed throughout.

+    for k,v in urls.iteritems():
+      if not k.startswith('shard'):
+        raise Exception('Illegally formatted \'urls\' key %s' % k)
+      if not v.startswith(SWARMING_URL):
+        raise Exception('Illegally formatted \'urls\' value %s' % v)
+      hashes.append(v[len(SWARMING_URL):])
+    return hashes
+
+class Waterfall:
+  def __init__(self, waterfall):
+    self._waterfall = waterfall
+    self.BASE_URL = 'http://build.chromium.org/p/'
+    self.BASE_BUILD_URL = self.BASE_URL + '%s/builders/%s'

[Zhenyao Mo, 2016/07/01 23:29:38]

unused variable

[Ken Russell (switch to Gerrit), 2016/07/02 03:17:01]

Removed.

+    self.BASE_JSON_BUILDERS_URL = self.BASE_URL + '%s/json/builders'
+    self.BASE_JSON_BUILDS_URL = self.BASE_JSON_BUILDERS_URL + '/%s/builds'
+
+  def GetJsonFromUrl(self, url):
+    conn = urllib2.urlopen(url)
+    result = conn.read()
+    conn.close()
+    return json.loads(result)
+
+  def GetBuildNumbersForBot(self, bot):
+    builds_json = self.GetJsonFromUrl(
+      self.BASE_JSON_BUILDS_URL %
+      (self._waterfall, urllib.quote(bot)))
+    build_numbers = [int(k) for k in builds_json.keys()]
+    build_numbers.sort()
+    return build_numbers
+
+  def GetMostRecentlyCompletedBuildNumberForBot(self, bot):
+    builds = self.GetBuildNumbersForBot(bot)
+    return builds[len(builds) - 1]
+
+  def GetJsonForBuild(self, bot, build):
+    return self.GetJsonFromUrl(
+      (self.BASE_JSON_BUILDS_URL + '/%d') %
+      (self._waterfall, urllib.quote(bot), build))
+
+
+def JsonLoadStrippingUnicode(file, **kwargs):
+  def StripUnicode(obj):
+    if isinstance(obj, unicode):
+      try:
+        return obj.encode('ascii')
+      except UnicodeEncodeError:
+        return obj
+
+    if isinstance(obj, list):
+      return map(StripUnicode, obj)
+
+    if isinstance(obj, dict):
+      new_obj = type(obj)(
+          (StripUnicode(k), StripUnicode(v)) for k, v in obj.iteritems() )
+      return new_obj
+
+    return obj
+
+  return StripUnicode(json.load(file, **kwargs))
+
+
+def Merge(dest, src):
+  if isinstance(dest, list):
+    if not isinstance(src, list):
+      raise Exception('Both must be lists: ' + dest + ' and ' + src)
+    return dest + src
+
+  if isinstance(dest, dict):
+    if not isinstance(src, dict):
+      raise Exception('Both must be dicts: ' + dest + ' and ' + src)
+    for k in src.iterkeys():
+      if k not in dest:
+        dest[k] = src[k]
+      else:
+        dest[k] = Merge(dest[k], src[k])
+    return dest
+
+  return src
+
+
+def main():
+  rest_args = sys.argv[1:]
+  parser = argparse.ArgumentParser(
+    description='Gather JSON results from a run of a Swarming test.',
+    formatter_class=argparse.ArgumentDefaultsHelpFormatter)
+  parser.add_argument('-v', '--verbose', action='count', default=0,
+                      help='Enable verbose output (specify multiple times '
+                      'for more output)')
+  parser.add_argument('--waterfall', type=str, default='chromium.gpu.fyi',
+                      help='Which waterfall to examine')
+  parser.add_argument('--bot', type=str, default='Linux Release (NVIDIA)',
+                      help='Which bot on the waterfall to examine')
+  parser.add_argument('--build', default=-1, type=int,
+                      help='Which build to fetch (-1 means most recent)')
+  parser.add_argument('--step', type=str, default='webgl2_conformance_tests',
+                      help='Which step to fetch (treated as a prefix)')
+  parser.add_argument('--output', type=str, default='output.json',
+                      help='Name of output file')
+  parser.add_argument('--leak-temp-dir', action='store_true', default=False,
+                      help='Deliberately leak temporary directory')
+
+  options = parser.parse_args(rest_args)
+
+  Swarming.CheckAuth()
+
+  waterfall = Waterfall(options.waterfall)
+  build = options.build
+  if build < 0:
+    build = waterfall.GetMostRecentlyCompletedBuildNumberForBot(options.bot)
+
+  build_json = waterfall.GetJsonForBuild(options.bot, build)
+
+  if options.verbose:
+    print 'Fetching information from %s, bot %s, build %s' % (
+      options.waterfall, options.bot, build)
+
+  hashes = []
+  for s in build_json['steps']:
+    if s['name'].startswith(options.step):
+      # Found the step. Now iterate down the URLs.

[Zhenyao Mo, 2016/07/01 23:29:38]

URL name isn't apparent. Can you document why?

[Ken Russell (switch to Gerrit), 2016/07/02 03:17:01]

Added a comment.

+      if 'urls' not in s or not s['urls']:
+        # Note: we could also just download json.output if it exists.
+        print ('%s on waterfall %s, bot %s, build %s doesn\'t '
+               'look like a Swarmed task') % (
+                 s['name'], options.waterfall, options.bot, build)
+        return 1
+      hashes = Swarming.ExtractShardHashes(s['urls'])
+      if options.verbose:
+        print 'Found Swarming hashes for step %s' % s['name']
+
+      break
+  if not hashes:
+    print 'Problem gathering the Swarming hashes for %s' % options.step
+    return 1
+
+  # Collect the results.
+  tmpdir = tempfile.mkdtemp()
+  Swarming.Collect(hashes, tmpdir, options.verbose)
+
+  # Shards' JSON outputs are in sequentially-numbered subdirectories
+  # of the output directory.
+  merged_json = None
+  for i in xrange(len(hashes)):
+    with open(os.path.join(tmpdir, str(i), 'output.json')) as f:
+      cur_json = JsonLoadStrippingUnicode(f)
+      if not merged_json:
+        merged_json = cur_json
+      else:
+        merged_json = Merge(merged_json, cur_json)
+
+  with open(options.output, 'w') as f:
+    json.dump(merged_json, f)
+
+  if options.leak_temp_dir:
+    print 'Temporary directory: %s' % tmpdir
+  else:
+    shutil.rmtree(tmpdir)
+
+  return 0
+
+
+if __name__ == "__main__":
+  sys.exit(main())