Add support for rebaselining from trybots.

gm/rebaseline_server/download_actuals.py

 #!/usr/bin/python
 
 """
 Copyright 2014 Google Inc.
 
 Use of this source code is governed by a BSD-style license that can be
 found in the LICENSE file.
 
 Download actual GM results for a particular builder.
 """
 
 # System-level imports
+import httplib
+import logging
 import optparse
 import os
 import posixpath
 import re
+import traceback
 import urllib2
 
 # Must fix up PYTHONPATH before importing from within Skia
 import rs_fixpypath  # pylint: disable=W0611
 
 # Imports from within Skia
 from py.utils import gs_utils
 from py.utils import url_utils
 import buildbot_globals
 import gm_json
@@ skipping 62 matching lines @@
 def get_builders_list(summaries_bucket=GM_SUMMARIES_BUCKET):
   """ Returns the list of builders we have actual results for.
 
   Args:
     summaries_bucket: Google Cloud Storage bucket containing the summary
         JSON files
   """
   dirs, _ = gs_utils.GSUtils().list_bucket_contents(bucket=GM_SUMMARIES_BUCKET)
   return dirs
 
+def get_rietveld_actuals(issue, json_filename):
+  """ Returns the actuals for the given rietveld issue's tryjobs.
+  {non_try_builder:string -> [ bucket:string, path:string, generation:string ]}
+
+  e.g.
+  {'Test-Android-Xoom-Tegra2-Arm7-Release': (
+      'chromium-skia-gm-summaries',
+      'Test-Android-Xoom-Tegra2-Arm7-Release-Trybot/actual-results.json',
+      '1415041165535000')}
+  """
+  result = dict()
+  json_filename_re = re.compile('^Created: gs://([^/]+)/((?:[^/]+/)+%s)#(\d+)$'
+                                % re.escape(json_filename), re.MULTILINE)
+  codereview_api_url = 'https://codereview.chromium.org/api'
+  upload_gm_results_step_url = '/steps/Upload GM Results/logs/stdio'
+
+  logging.info('Fetching issue %s ...' % (issue,))
+  json_issue_url = '%s/%s' % (codereview_api_url, issue)
+  json_issue_data = urllib2.urlopen(json_issue_url).read()
+  issue_dict = gm_json.LoadFromString(json_issue_data)
+
+  patchsets = issue_dict.get("patchsets", [])
+  patchset = patchsets[-1]
+  if not patchset:
+    logging.warning('No patchsets for rietveld issue %s.' % (issue,))
+    return result
+
+  logging.info('Fetching issue %s patch %s...' % (issue, patchset))
+  json_patchset_url = '%s/%s/%s' % (codereview_api_url, issue, patchset)
+  json_patchset_data = urllib2.urlopen(json_patchset_url).read()
+  patchset_dict = gm_json.LoadFromString(json_patchset_data)
+
+  # try_job_results is ordered reverse chronologically
+  try_job_results = patchset_dict.get('try_job_results', [])
+  for try_job_result in try_job_results:
+    builder = try_job_result.get('builder', '<bad builder>')
+    non_try_builder = builder

[rmistry, 2014/11/05 13:39:16]

Seems confusing to switch non_try_builder's value between a try_builder and then
to a non_try_builder. How about directly doing:

if not builder.endswith('-Trybot'):
  continue

non_try_builder = builder[:-len('-Trybot')]
...

[bungeman-skia, 2014/11/07 22:56:52]

Done.

+    if non_try_builder.endswith('-Trybot'):
+      non_try_builder = non_try_builder[:-len('-Trybot')]
+    if non_try_builder in result:
+      continue
+
+    logging.info('Fetching issue %s patch %s try %s...' %
+                (issue, patchset, builder))
+    build_url = try_job_result.get('url', '<bad url>')
+    gm_upload_output_url = build_url + urllib2.quote(upload_gm_results_step_url)
+    logging.info('Fetching %s ...' % (gm_upload_output_url,))
+
+    # Tryjobs might not produce the step, but don't let that fail everything.
+    gm_upload_output = None
+    try:
+      gm_upload_output = urllib2.urlopen(gm_upload_output_url).read()
+    except urllib2.HTTPError, e:
+      logging.warning('HTTPError: ' + str(e.code))
+    except urllib2.URLError, e:
+      logging.warning('URLError: ' + str(e.reason))
+    except httplib.HTTPException, e:
+      logging.warning('HTTPException')
+    except Exception:
+      logging.warning('generic exception: ' + traceback.format_exc())

[rmistry, 2014/11/05 13:39:16]

traceback.format_exc() returns None so this will probably print:
generic exception: None

It will be useful to output the complete stacktrace here when something goes
wrong. Maybe collapse all the except sections into:

except Exception:
  logging.exception('Error opening %s', gm_upload_output_url)

logging.exception grabs the exception details including the traceback from the
current frame.

[bungeman-skia, 2014/11/07 22:56:53]

Done.

+    if not gm_upload_output:
+      logging.warning('Could not fetch %s .' % (gm_upload_output_url,))
+      continue
+
+    json_filename_match = json_filename_re.search(gm_upload_output)
+    if json_filename_match:
+      logging.info('Found  issue %s patch %s try %s result gs://%s/%s#%s .' %
+                  (issue, patchset, non_try_builder,
+                   json_filename_match.group(1),
+                   json_filename_match.group(2),
+                   json_filename_match.group(3)))
+      result[non_try_builder] = [json_filename_match.group(1),
+                                 json_filename_match.group(2),
+                                 json_filename_match.group(3)]
+    else:
+      logging.warning('Did not find %s for issue %s patch %s try %s.' %
+                     (json_filename, issue, patchset, builder))
+
+  return result
 
 def main():
   parser = optparse.OptionParser()
   required_params = []
   parser.add_option('--actuals-base-url',
                     action='store', type='string',
                     default=DEFAULT_ACTUALS_BASE_URL,
                     help=('Base URL from which to read files containing JSON '
                           'summaries of actual GM results; defaults to '
                           '"%default".'))
@@ skipping 36 matching lines @@
                     remaining_args)
 
   downloader = Download(actuals_base_url=params.actuals_base_url)
   downloader.fetch(builder_name=params.builder,
                    dest_dir=params.dest_dir)
 
 
 
 if __name__ == '__main__':
   main()