Add support for rebaselining from trybots.

gm/rebaseline_server/server.py

 #!/usr/bin/python
 
 """
 Copyright 2013 Google Inc.
 
 Use of this source code is governed by a BSD-style license that can be
 found in the LICENSE file.
 
 HTTP server for our HTML rebaseline viewer.
 """
@@ skipping 244 matching lines @@
 
     file_handle.write('\n</ul></body></html>')
 
 
 class Server(object):
   """ HTTP server for our HTML rebaseline viewer. """
 
   def __init__(self,
                actuals_dir=DEFAULT_ACTUALS_DIR,
                json_filename=DEFAULT_JSON_FILENAME,
+               rietveld_issue=None,
                gm_summaries_bucket=DEFAULT_GM_SUMMARIES_BUCKET,
                port=DEFAULT_PORT, export=False, editable=True,
                reload_seconds=0, config_pairs=None, builder_regex_list=None,
                boto_file_path=None,
                imagediffdb_threads=imagediffdb.DEFAULT_NUM_WORKER_THREADS):
     """
     Args:
       actuals_dir: directory under which we will check out the latest actual
           GM results
       json_filename: basename of the JSON summary file to load for each builder
+      rietveld_issue: download json_filename files from latest trybot runs on
+          this codereview.chromium.org issue.
       gm_summaries_bucket: Google Storage bucket to download json_filename
           files from; if None or '', don't fetch new actual-results files
           at all, just compare to whatever files are already in actuals_dir
       port: which TCP port to listen on for HTTP requests
       export: whether to allow HTTP clients on other hosts to access this server
       editable: whether HTTP clients are allowed to submit new GM baselines
           (SKP baseline modifications are performed using an entirely different
           mechanism, not affected by this parameter)
       reload_seconds: polling interval with which to check for new results;
           if 0, don't check for new results at all
       config_pairs: List of (string, string) tuples; for each tuple, compare
           actual results of these two configs.  If None or empty,
           don't compare configs at all.
       builder_regex_list: List of regular expressions specifying which builders
           we will process. If None, process all builders.
       boto_file_path: Path to .boto file giving us credentials to access
           Google Storage buckets; if None, we will only be able to access
           public GS buckets.
       imagediffdb_threads: How many threads to spin up within imagediffdb.
     """
     self._actuals_dir = actuals_dir
     self._json_filename = json_filename
+    self._rietveld_issue = rietveld_issue
     self._gm_summaries_bucket = gm_summaries_bucket
     self._port = port
     self._export = export
     self._editable = editable
     self._reload_seconds = reload_seconds
     self._config_pairs = config_pairs or []
     self._builder_regex_list = builder_regex_list
     self.truncate_results = False
 
     if boto_file_path:
@@ skipping 71 matching lines @@
     the same time.
 
     Args:
       invalidate: if True, invalidate self._results immediately upon entry;
                   otherwise, we will let readers see those results until we
                   replace them
     """
     with self.results_rlock:
       if invalidate:
         self._results = None
+
+      if self._rietveld_issue:
+        logging.info(
+            'Updating GM result summaries in %s from rietveld_issue %s ...'
+            % (self._actuals_dir, self._rietveld_issue))
+
+        # Clean out actuals_dir first, in case some builders have gone away

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

There is a lot of code duplication here with the section below maybe this will
help:

* Extract 398-401 and 435-438 into a function called clean_actuals_dir
* Extract 407-428 and the same section below into a separate function called
from both places.

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

Done.

+        # since we last ran.
+        if os.path.isdir(self._actuals_dir):
+          shutil.rmtree(self._actuals_dir)
+
+        # Get the list of actuals we care about.
+        all_actuals = download_actuals.get_rietveld_actuals(
+            self._rietveld_issue, self._json_filename)
+
+        if self._builder_regex_list:
+          matching_builders = []
+          for builder in all_actuals:
+            for regex in self._builder_regex_list:
+              if re.match(regex, builder):
+                matching_builders.append(builder)
+                break  # go on to the next builder, no need to try more regexes
+        else:
+          matching_builders = all_actuals.keys()
+
+        # Download the JSON file for each builder we care about.
+        #
+        # TODO(epoger): When this is a large number of builders, we would be
+        # better off downloading them in parallel!
+        for builder in matching_builders:
+          self._gs.download_file(
+              source_bucket=all_actuals[builder][0],
+              source_path=all_actuals[builder][1],
+              source_generation=all_actuals[builder][2],
+              dest_path=os.path.join(self._actuals_dir, builder,
+                                     self._json_filename),
+              create_subdirs_if_needed=True)
+
       if self._gm_summaries_bucket:
         logging.info(
             'Updating GM result summaries in %s from gm_summaries_bucket %s ...'
             % (self._actuals_dir, self._gm_summaries_bucket))
 
         # Clean out actuals_dir first, in case some builders have gone away
         # since we last ran.
         if os.path.isdir(self._actuals_dir):
           shutil.rmtree(self._actuals_dir)
 
@@ skipping 494 matching lines @@
   parser.add_argument('--editable', action='store_true',
                       help=('Allow HTTP clients to submit new GM baselines; '
                             'SKP baselines can be edited regardless of this '
                             'setting.'))
   parser.add_argument('--export', action='store_true',
                       help=('Instead of only allowing access from HTTP clients '
                             'on localhost, allow HTTP clients on other hosts '
                             'to access this server.  WARNING: doing so will '
                             'allow users on other hosts to modify your '
                             'GM expectations, if combined with --editable.'))
+  parser.add_argument('--rietveld_issue',
+                      help=('Download json_filename files from latest trybot'
+                            'runs on this codereview.chromium.org issue.'))
   parser.add_argument('--gm-summaries-bucket',
                     help=('Google Cloud Storage bucket to download '
                           'JSON_FILENAME files from. '
                           'Defaults to %(default)s ; if set to '
                           'empty string, just compare to actual-results '
                           'already found in ACTUALS_DIR.'),
                     default=DEFAULT_GM_SUMMARIES_BUCKET)
   parser.add_argument('--json-filename',
                     help=('JSON summary filename to read for each builder; '
                           'defaults to %(default)s.'),
@@ skipping 20 matching lines @@
                             'process, to speed up testing.'))
   args = parser.parse_args()
   if args.compare_configs:
     config_pairs = CONFIG_PAIRS_TO_COMPARE
   else:
     config_pairs = None
 
   global _SERVER
   _SERVER = Server(actuals_dir=args.actuals_dir,
                    json_filename=args.json_filename,
+                   rietveld_issue=args.rietveld_issue,
                    gm_summaries_bucket=args.gm_summaries_bucket,
                    port=args.port, export=args.export, editable=args.editable,
                    reload_seconds=args.reload, config_pairs=config_pairs,
                    builder_regex_list=args.builders, boto_file_path=args.boto,
                    imagediffdb_threads=args.threads)
   if args.truncate:
     _SERVER.truncate_results = True
   _SERVER.run()
 
 
 if __name__ == '__main__':
   main()