tool for repackaging chrome.perf builds for manual bisect

tools/bisect_repackage/bisect_repackage_linux.py

Index: tools/bisect_repackage/bisect_repackage_linux.py
diff --git a/tools/bisect_repackage/bisect_repackage_linux.py b/tools/bisect_repackage/bisect_repackage_linux.py
new file mode 100644
index 0000000000000000000000000000000000000000..4ec67e91f1bfc3cccae07b91fb8a5bf87ee9b749
--- /dev/null
+++ b/tools/bisect_repackage/bisect_repackage_linux.py
@@ -0,0 +1,393 @@
+# Copyright (c) 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.
+
+""" Bisect repackage tool for Linux
+
+This script repacakges chrome builds for manual bisect script.
+"""
+
+#Declares required files to run manual bisect script on chrome Linux
+#builds in perf. Binary files that should be stripped to reduce zip file
+#size are declared. The file list was gotten from the local chrome
+#executable path in Linux. (This can be retrieved by typing 'chrome://version'
+#in chrome and following the executable path. The list needs to be updated if
+#future chrome versions require additional files.
+FILES_TO_ARCHIVE = [
+  'chrome',
+  'chrome_100_percent.pak',
+  'chrome_200_percent.pak',
+  'default_apps',
+  'icudtl.dat',
+  'libwidevinecdm.so',
+  'locales',
+  'nacl_helper',
+  'nacl_helper_bootstrap',
+  'nacl_irt_x86_64.nexe',
+  'natives_blob.bin',
+  'PepperFlash',
+  'product_logo_48.png',
+  'resources.pak',
+  'snapshot_blob.bin',
+  'xdg-mime',
+  'xdg-settings',
+]
+
+# Declares what files should be stripped of symbols to reduce the archive size.
+STRIP_LIST = [
+  'chrome',
+  'nacl_helper'
+]
+
+# API to convert Githash to Commit position number.
+CHROMIUM_GITHASH_TO_SVN_URL = (
+    'https://cr-rev.appspot.com/_ah/api/crrev/v1/commit/%s')
+
+BUILDER_NAME = 'Linux Builder'
+ARCHIVE_PREFIX = 'full-build-linux'
+REVISION_MAP_FILE = 'revision_map.json'
+
+################################################################################
+import os
+import bisect_repackage_utils
+import zipfile
+import zlib
+import json
+import sys
+import optparse
+import re
+import urllib
+import threading
+from multiprocessing import Pool
+from functools import partial
+
+# Declares where files should be stored temporarily for staging.
+# TODO (miimnk): use tempfile.mkdtemp to make temporary folders
+STAGING_DIR= os.path.join('.', 'a', 'tmp_forward')

[dimu, 2016/08/19 04:14:05]

is there any cleanup for the staging dir?

[miimnk, 2016/08/19 20:29:43]

Erased the two global variables and used 'tempfile.mkdtemp' instead. 

+STAGING_DIR_BACKWARD = os.path.join('.', 'a', 'tmp_backward')
+
+
+class PathContext(object):
+  """A PathContext is used to carry the information used to construct URLs and
+  paths when dealing with the storage server and archives."""
+  def __init__(self, original_gs_url, repackage_gs_url,
+               builder_name=BUILDER_NAME, file_prefix=ARCHIVE_PREFIX,
+               revision_map=REVISION_MAP_FILE):
+    super(PathContext, self).__init__()
+    self.original_gs_url = original_gs_url
+    self.repackage_gs_url = repackage_gs_url
+    self.builder_name = builder_name
+    self.file_prefix = file_prefix
+    self.revision_map = revision_map
+
+
+def get_cp_from_hash(git_hash):
+  """Converts a GitHash to Commit position number"""
+  json_url = CHROMIUM_GITHASH_TO_SVN_URL % git_hash
+  response = urllib.urlopen(json_url)
+  if response.getcode() == 200:
+    try:
+      data = json.loads(response.read())
+    except ValueError:
+      print 'ValueError for JSON URL: %s' % json_url

[dimu, 2016/08/19 04:14:05]

It would be better to use logging instead of print.
(https://docs.python.org/2/howto/logging.html)

[miimnk, 2016/08/19 20:29:43]

Done.

+      return None

[dimu, 2016/08/19 04:14:05]

throw exception for failure, and use try/catch in the caller.

[miimnk, 2016/08/19 20:29:43]

Done.

+  else:
+    print 'ValueError for JSON URL: %s' % json_url
+    return None
+  if 'number' in data:
+    return data['number']
+  print 'Failed to get svn revision number for %s' % git_hash
+  return None
+
+def create_cp_from_hash_map(hash_list):
+  """Creates a dictionary that maps from Commit positio number
+  to corresponding GitHash"""
+  hash_map = {}
+  count = 0
+  for git_hash in hash_list:
+    cp_num = get_cp_from_hash(git_hash)
+    hash_map[cp_num] = git_hash
+    print "Downloaded %s / %s git hash" %(count, len(hash_list))
+    count += 1
+  return hash_map
+
+def isProperHash(regex_match):

[dimu, 2016/08/19 04:14:05]

move it to utils

[dimu, 2016/08/19 04:14:05]

isGitCommitHash

[miimnk, 2016/08/19 20:29:42]

Done.

[miimnk, 2016/08/19 20:29:43]

Done.

+  """Checks if match is correct SHA1 hash"""
+  if len(regex_match) == 40: return True

[dimu, 2016/08/19 04:14:05]

use re.match(r"^[0-9,A-F]{40}$", regex_match.upper())

[miimnk, 2016/08/19 20:29:43]

Done.

+  return False
+
+def isProperRevision(regex_match):

[dimu, 2016/08/19 04:14:05]

move this to utils

[miimnk, 2016/08/19 20:29:43]

Done.

+  """Checks if match is correct revision(Cp number) format"""
+  if len(regex_match) == 6: return True

[dimu, 2016/08/19 04:14:05]

also use regex

[miimnk, 2016/08/19 20:29:43]

Done.

+  return False
+
+def get_list_of_suffix(bucket_address, prefix, filter_function):
+  """Gets the list of suffixes in files in a google storage bucket.
+  For example, a google storage bucket containing one file
+  'full-build-linux_20983' will return ['20983'] if prefix is
+  provided as 'full-build-linux'. Google Storage bucket
+  containing multiple files will return multiple suffixes. """
+  file_list = bisect_repackage_utils.GSutilList(bucket_address)
+  suffix_list = []
+  extract_suffix = '.*?%s_(.*?)\.zip' %(prefix)
+  for file in file_list:
+    match = re.match(extract_suffix, file)
+    if match and filter_function(match.groups()[0]):
+      suffix_list.append(match.groups()[0])
+  return suffix_list
+
+def download_build(cp_num, revision_map, zip_file_name, context):
+  """ Download a single build corresponding to the cp_num and context."""
+  file_url = '%s/%s/%s_%s.zip' %(context.original_gs_url, context.builder_name,
+                                 context.file_prefix, revision_map[cp_num])
+  bisect_repackage_utils.GSUtilDownloadFile(file_url, zip_file_name)
+
+def upload_build(zip_file, context):
+  """Uploads a single build in zip_file to the repackage_gs_url in context."""
+  gs_base_url = '%s/%s' %(context.repackage_gs_url, context.builder_name)
+  upload_url = gs_base_url + '/'
+  bisect_repackage_utils.GSUtilCopy(zip_file, upload_url)
+
+def download_revision_map(context):
+  """Downloads the revision map in original_gs_url in context."""
+  gs_base_url = '%s/%s' %(context.repackage_gs_url, context.builder_name)
+  download_url = gs_base_url + '/' + context.revision_map
+  try:
+    bisect_repackage_utils.GSUtilDownloadFile(download_url,
+                                              context.revision_map)
+  except Exception, e:
+    raise ValueError('Revision map does not exist. Re run the program with'
+                     '-c option to upload a revision map to google storage')
+
+def get_revision_map(context):
+  """Downloads and returns the revision map in repackage_gs_url in context."""
+  bisect_repackage_utils.RemoveFile(context.revision_map)
+  download_revision_map(context)
+  with open(context.revision_map, 'r') as f:

[dimu, 2016/08/19 04:14:05]

nit: better naming: revision_file

[miimnk, 2016/08/19 20:29:43]

Done.

+    revision_map = json.load(f)
+  bisect_repackage_utils.RemoveFile(context.revision_map)
+  return revision_map
+
+def upload_revision_map(revision_map, context):
+  """Upload the given revision_map to the repackage_gs_url in context."""
+  with open(context.revision_map, 'w') as fp:
+    json.dump(revision_map, fp)
+  gs_base_url = '%s/%s' %(context.repackage_gs_url, context.builder_name)
+  upload_url = gs_base_url + '/'
+  bisect_repackage_utils.GSUtilCopy(context.revision_map, upload_url)
+  bisect_repackage_utils.RemoveFile(context.revision_map)
+
+def create_upload_revision_map(context):
+  """ Creates and uploads a dictionary that maps from GitHash to CP number."""
+  gs_base_url = '%s/%s' %(context.original_gs_url, context.builder_name)
+  hash_list = get_list_of_suffix(gs_base_url, context.file_prefix, isProperHash)
+  cp_num_to_hash_map = create_cp_from_hash_map(hash_list)
+  upload_revision_map(cp_num_to_hash_map, context)
+
+def update_upload_revision_map(context):
+  """ Updates and uploads a dictionary that maps from GitHash to CP number."""
+  gs_base_url = '%s/%s' %(context.original_gs_url, context.builder_name)
+  revision_map = get_revision_map(context)
+  hash_list = get_list_of_suffix(gs_base_url, context.file_prefix, isProperHash)
+  hash_list = list(set(hash_list)-set(revision_map.values()))
+  cp_num_to_hash_map = create_cp_from_hash_map(hash_list)
+  upload_revision_map(cp_num_to_hash_map, context)
+
+def make_filtered_archive(file_archive, archive_name):

[dimu, 2016/08/19 04:14:05]

better naming

[miimnk, 2016/08/19 20:29:43]

Done. Changed to 'make_lightweight_archive'. 

+  """Repackages and strips the archive according to FILES_TO_ARCHIVE and
+  STRIP_LIST defined on the top."""
+  return bisect_repackage_utils.MakeZip('.',
+                                archive_name,
+                                FILES_TO_ARCHIVE,
+                                file_archive,
+                                raise_error=False,
+                                strip_files=STRIP_LIST)
+
+def remove_created_files_and_path(files, paths):
+  """ Removes all the files and paths passed in."""
+  for file in files:
+    bisect_repackage_utils.RemoveFile(file)
+  for path in paths:
+    bisect_repackage_utils.RemovePath(path)
+
+def verify_chrome_run(zip_dir):
+  """This function executes chrome executable in zip_dir. Raises error if the
+  execution fails for any reason."""
+  try:
+    command = [os.path.join(zip_dir, 'chrome')]
+    code = bisect_repackage_utils.RunCommand(command)
+    if code != 0:
+      raise ValueError("Executing Chrome Failed: Need to check ")
+  except Exception, e:
+    raise ValueError("Executing Chrome Failed: Need to check ")
+
+def repackage_single_revision(revision_map, isForwardArchive, verify_run,
+                              staging_dir, context, cp_num):
+  """Repackages a single Chrome build for manual bisect."""
+  archive_name = '%s_%s' %(context.file_prefix, cp_num)
+  file_archive = os.path.join(staging_dir, archive_name)
+  zip_file_name = '%s.zip' %file_archive
+
+  bisect_repackage_utils.RemoveFile(zip_file_name)
+  download_build(cp_num, revision_map, zip_file_name, context)
+  extract_dir = os.path.join(staging_dir, archive_name)
+  bisect_repackage_utils.ExtractZip(zip_file_name, extract_dir)
+  extracted_folder = os.path.join(extract_dir, context.file_prefix)
+
+  (zip_dir, zip_file) = make_filtered_archive(extracted_folder, archive_name)
+
+  if verify_run:
+    verify_chrome_run(zip_dir)
+
+  upload_build(zip_file, context)
+  # Removed temporary files created during repackaging process.
+  remove_created_files_and_path([zip_file, zip_file_name],
+                                [zip_dir, extract_dir])
+
+def repackage_revisions(revisions, revision_map, isForwardArchive, verify_run,
+                        staging_dir, context, quit_event=None,
+                        progress_event=None):
+  """ Repackages all Chrome builds listed in revisions. This function calls
+  'repackage_single_revision' with multithreading pool.'"""
+  p = Pool(3)
+  func = partial(repackage_single_revision, revision_map, isForwardArchive,
+                 verify_run, staging_dir, context)
+  p.imap(func, revisions)
+  p.close()
+  p.join()
+
+def get_uploaded_builds(context):
+  """Returns already uploaded commit positions numbers in
+  context.repackage_gs_url"""
+  gs_base_url = '%s/%s' %(context.repackage_gs_url, context.builder_name)
+  return get_list_of_suffix(gs_base_url, context.file_prefix, isProperRevision)
+
+def get_revisions_to_package(revision_map, context):
+  """ Returns revisions that need to be repackaged. The first half of
+  the revisions will be sorted in ascending order and the second half of
+  the revisions will be sorted in desending order.
+
+  The first half will be used for repackaging revisions in forward direction,
+  and the second half will be used for repackaging revisions in backward
+  direction."""
+  already_packaged = get_uploaded_builds(context)
+  not_already_packaged = list(set(revision_map.keys())-set(already_packaged))
+  revisions_to_package = sorted(not_already_packaged)
+  revisions_to_package = filter(lambda a: a != 'null', revisions_to_package)
+
+  forward_rev = revisions_to_package[:len(revisions_to_package)/2]
+  backward_rev = sorted(revisions_to_package[len(revisions_to_package)/2:],
+                        reverse=True)
+  return (forward_rev, backward_rev)
+
+class RepackageJob(object):
+  def __init__(self, name, revisions_to_package, revision_map, isForwardArchive,
+               verify_run, staging_dir, context):
+    super(RepackageJob, self).__init__()
+    self.name = name
+    self.revisions_to_package = revisions_to_package
+    self.revision_map = revision_map
+    self.isForwardArchive = isForwardArchive
+    self.verify_run = verify_run
+    self.staging_dir = staging_dir
+    self.context = context
+    self.quit_event = threading.Event()
+    self.progress_event = threading.Event()
+    self.thread = None
+
+  def Start(self):
+    """Starts the download."""
+    fetchargs = (self.revisions_to_package,
+                 self.revision_map,
+                 self.isForwardArchive,
+                 self.verify_run,
+                 self.staging_dir,
+                 self.context,
+                 self.quit_event,
+                 self.progress_event)
+    self.thread = threading.Thread(target=repackage_revisions,
+                                   name=self.name,
+                                   args=fetchargs)
+    self.thread.start()
+
+  def Stop(self):
+    """Stops the download which must have been started previously."""
+    assert self.thread, 'DownloadJob must be started before Stop is called.'
+    self.quit_event.set()
+    self.thread.join()
+
+  def WaitFor(self):
+    """Prints a message and waits for the download to complete. The download
+    must have been started previously."""
+    assert self.thread, 'DownloadJob must be started before WaitFor is called.'
+    self.progress_event.set()  # Display progress of download.  def Stop(self):
+    assert self.thread, 'DownloadJob must be started before Stop is called.'
+    self.quit_event.set()
+    self.thread.join()
+
+def main(argv):
+  option_parser = optparse.OptionParser()
+
+  # Verifies that the chrome executable runs
+  option_parser.add_option('-v', '--verify',
+                           action='store_true',
+                           help='Verifies that the Chrome executes normally'
+                                'without errors')
+
+  # This option will update the revision map.
+  option_parser.add_option('-u', '--update',
+                           action='store_true',
+                           help='Updates the list of revisions to repackage')
+
+  # This option will creates the revision map.
+  option_parser.add_option('-c', '--create',
+                           action='store_true',
+                           help='Creates the list of revisions to repackage')
+
+  # Original bucket that contains perf builds
+  option_parser.add_option('-o', '--original',
+                    type='str',
+                    help='Google storage url containing original Chrome builds')
+
+  # Bucket that should archive lightweight perf builds
+  option_parser.add_option('-r', '--repackage',
+                    type='str',
+                    help='Google storage url to re-archive Chrome builds')
+
+  verify_run = False
+  (opts, args) = option_parser.parse_args()
+  if not opts.original or not opts.repackage:
+    raise ValueError('Need to specify original gs bucket url and'
+                     'repackage gs bucket url')
+  context = PathContext(opts.original, opts.repackage)
+
+  if opts.create:
+    create_upload_revision_map(context)
+
+  if opts.update:
+    update_upload_revision_map(context)
+
+  if opts.verify:
+    verify_run = True
+
+  revision_map = get_revision_map(context)
+  (forward_rev, backward_rev) = get_revisions_to_package(revision_map, context)
+  bisect_repackage_utils.MaybeMakeDirectory(STAGING_DIR)
+  bisect_repackage_utils.MaybeMakeDirectory(STAGING_DIR_BACKWARD)
+
+  # Spwan two threads: One repackaging from oldest builds. The other repackaging
+  # from the newest builds.
+  forward_fetch = RepackageJob('forward_fetch', forward_rev, revision_map, True,
+                               verify_run, os.path.abspath(STAGING_DIR),
+                               context)
+  backward_fetch = RepackageJob('backward_fetch', backward_rev, revision_map,
+                                False, verify_run,
+                                os.path.abspath(STAGING_DIR_BACKWARD),
+                                context)
+  forward_fetch.Start()
+  backward_fetch.Start()
+  forward_fetch.WaitFor()
+  backward_fetch.WaitFor()
+
+if '__main__' == __name__:
+  sys.exit(main(sys.argv))