[Release] Distinguish between merges and follow-up CLs

tools/release/mergeinfo.py

 #!/usr/bin/env python
 # Copyright 2015 the V8 project 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 argparse
 import os
 import sys
 
-from subprocess import call
+from search_related_commits import git_execute
 
-def print_analysis(gitWorkingDir, hashToSearch):
-  print '1.) Info'
-  git_execute(gitWorkingDir, ['status'])
-  print '2.) Searching for "' + hashToSearch + '"'
-  print '=====================ORIGINAL COMMIT START====================='
-  git_execute(gitWorkingDir, ['show', hashToSearch])
+GIT_OPTION_HASH_ONLY = '--pretty=format:%H'
+GIT_OPTION_NO_DIFF = '--quiet'
+GIT_OPTION_ONELINE = '--oneline'
+
+def describe_commit(git_working_dir, hash_to_search, one_line=False):
+  if one_line:
+    return git_execute(git_working_dir, ['show',
+                                         GIT_OPTION_NO_DIFF,
+                                         GIT_OPTION_ONELINE,
+                                         hash_to_search]).strip()
+  return git_execute(git_working_dir, ['show',
+                                       GIT_OPTION_NO_DIFF,
+                                       hash_to_search]).strip()
+
+
+def get_followup_commits(git_working_dir, hash_to_search):
+  return git_execute(git_working_dir, ['log',
+                                       '--grep='+hash_to_search,

[Michael Achenbach, 2015/10/01 08:24:18]

nit: space around +

[Michael Achenbach, 2015/10/01 08:24:18]

How often does it happen that developers use the full hash in a commit message
to refer to another commit. And how often does it happen that they refer to
another commit by other means, e.g. shortened hash, cr-commit-position, text of
original cl, actual link of cl, etc?

[Michael Hablich, 2015/10/01 10:33:25]

Done.

[Michael Hablich, 2015/10/01 10:33:25]

As far as I have seen the most prominent form is referring to full commit
hashes. The tools we are using for doing a revert and merge are also using the
full commit hash. Our port suppliers do the same.

It might make sense to invest in commit position and a reduced hash in future
iterations.

+                                       GIT_OPTION_HASH_ONLY,
+                                       'master']).strip().splitlines()
+
+def get_merge_commits(git_working_dir, hash_to_search):
+  merges = get_related_commits_not_on_master(git_working_dir, hash_to_search)
+  false_merges = get_related_commits_not_on_master(
+    git_working_dir, 'Cr-Branched-From: ' + hash_to_search)
+  return list(set(merges) - set(false_merges))

[Michael Achenbach, 2015/10/01 08:24:18]

you might loose the original order while going through the sets

[Michael Hablich, 2015/10/01 10:33:25]

Done.

+
+def get_related_commits_not_on_master(git_working_dir, grep_command):
+  commits = git_execute(git_working_dir, ['log',
+                                       '--all',

[Michael Achenbach, 2015/10/01 08:24:18]

nit: indentation

[Michael Hablich, 2015/10/01 10:33:25]

Done.

+                                       '--grep=' + grep_command,
+                                       GIT_OPTION_ONELINE,
+                                       '--decorate',
+                                       '--not',
+                                       'master',
+                                       GIT_OPTION_HASH_ONLY])
+  return commits.splitlines()
+
+def get_branches_for_commit(git_working_dir, hash_to_search):
+  branches = git_execute(git_working_dir, ['branch',
+                                           '--contains',
+                                           hash_to_search,
+                                           '-a']).strip()
+  branches = branches.splitlines()
+  branches = [ current.strip() for current in branches]

[Michael Achenbach, 2015/10/01 08:24:18]

3 in 1:
return map(str.strip, branches.splitlines())

[Michael Hablich, 2015/10/01 10:33:25]

Did a mixture because I like it a little bit more verbose

[Michael Achenbach, 2015/10/01 10:40:29]

ok - 2 in 1 :)

+  return branches
+
+def is_rolled(git_working_dir, hash_to_search):
+  branches = get_branches_for_commit(git_working_dir, hash_to_search)
+  return 'remotes/origin/roll' in branches
+
+def is_lkgr(git_working_dir, hash_to_search):
+  branches = get_branches_for_commit(git_working_dir, hash_to_search)
+  return 'remotes/origin/lkgr' in branches
+
+def get_first_canary(git_working_dir, hash_to_search):
+  branches = get_branches_for_commit(git_working_dir, hash_to_search)
+  canaries = ([currentBranch for currentBranch in branches if
+    currentBranch.startswith('remotes/origin/chromium/')])
+  canaries.sort()

[Michael Achenbach, 2015/10/01 08:24:18]

or put sorted() around the call above...

[Michael Hablich, 2015/10/01 10:33:25]

I don't want to nest the list operation above even more.

+  if len(canaries) == 0:
+    return 'No Canary coverage'
+  return canaries[0].split('/')[-1]
+
+def print_analysis(git_working_dir, hash_to_search):
+  print '1.) Searching for "' + hash_to_search + '"'
+  print '=====================ORIGINAL COMMIT START==================='
+  print describe_commit(git_working_dir, hash_to_search)
   print '=====================ORIGINAL COMMIT END====================='
-  print '#####################FOUND MERGES & REVERTS START#####################'
-  git_execute(gitWorkingDir, ["log",'--all', '--grep='+hashToSearch])
-  print '#####################FOUND MERGES & REVERTS END#####################'
+
+  print '2.) General information:'
+  print 'Is rolled: ' + str(is_rolled(git_working_dir, hash_to_search))
+  print 'Is LKGR: ' + str(is_lkgr(git_working_dir, hash_to_search))
+  print 'Is on Canary: ' + (
+    str(get_first_canary(git_working_dir, hash_to_search)))
+  print '3.) Found follow-up commits, reverts and ports:'
+  followups = get_followup_commits(git_working_dir, hash_to_search)
+  for followup in followups:
+    print describe_commit(git_working_dir, followup, True)
+
+  print '4.) Found merges:'
+  merges = get_merge_commits(git_working_dir, hash_to_search)
+  for currentMerge in merges:
+    print describe_commit(git_working_dir, currentMerge, True)
+    print '---Merged to:'
+    mergeOutput = git_execute(git_working_dir, ['branch',
+                                                '--contains',
+                                                currentMerge,
+                                                '-r']).strip()
+    print mergeOutput
   print 'Finished successfully'
 
-def git_execute(workingDir, commands):
-  return call(["git", '-C', workingDir] + commands)
+if __name__ == '__main__':  # pragma: no cover
+  parser = argparse.ArgumentParser('Tool to check where a git commit was'
+ ' merged and reverted.')
 
-if __name__ == "__main__":  # pragma: no cover
-  parser = argparse.ArgumentParser('Tool to check where a git commit was merged and reverted.')
-  parser.add_argument("-g", "--git-dir", required=False, default='.',
-                        help="The path to your git working directory.")
+  parser.add_argument('-g', '--git-dir', required=False, default='.',
+                        help='The path to your git working directory.')
 
-  parser.add_argument('hash', nargs=1, help="Hash of the commit to be searched.")
+  parser.add_argument('hash',
+                      nargs=1,
+                      help='Hash of the commit to be searched.')
 
   args = sys.argv[1:]
   options = parser.parse_args(args)
 
   sys.exit(print_analysis(options.git_dir, options.hash[0]))