tools/hook-scripts/mailer/mailer.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-
#
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements.  See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership.  The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License.  You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied.  See the License for the
# specific language governing permissions and limitations
# under the License.
#
#
# mailer.py: send email describing a commit
#
# $HeadURL$
# $LastChangedDate$
# $LastChangedBy$
# $LastChangedRevision$
#
# USAGE: mailer.py commit      REPOS REVISION [CONFIG-FILE]
#        mailer.py propchange  REPOS REVISION AUTHOR REVPROPNAME [CONFIG-FILE]
#        mailer.py propchange2 REPOS REVISION AUTHOR REVPROPNAME ACTION \
#                              [CONFIG-FILE]
#        mailer.py lock        REPOS AUTHOR [CONFIG-FILE]
#        mailer.py unlock      REPOS AUTHOR [CONFIG-FILE]
#
#   Using CONFIG-FILE, deliver an email describing the changes between
#   REV and REV-1 for the repository REPOS.
#
#   ACTION was added as a fifth argument to the post-revprop-change hook
#   in Subversion 1.2.0.  Its value is one of 'A', 'M' or 'D' to indicate
#   if the property was added, modified or deleted, respectively.
#
#   See _MIN_SVN_VERSION below for which version of Subversion's Python
#   bindings are required by this version of mailer.py.

import os
import sys
if sys.hexversion >= 0x3000000:
  PY3 = True
  import configparser
  from urllib.parse import quote as _url_quote
else:
  PY3 = False
  import ConfigParser as configparser
  from urllib import quote as  _url_quote
import time
import subprocess
from io import BytesIO
import smtplib
import re
import tempfile
import codecs

# Minimal version of Subversion's bindings required
_MIN_SVN_VERSION = [1, 5, 0]

# Import the Subversion Python bindings, making sure they meet our
# minimum version requirements.
import svn.fs
import svn.delta
import svn.repos
import svn.core
if _MIN_SVN_VERSION > [svn.core.SVN_VER_MAJOR,
                       svn.core.SVN_VER_MINOR,
                       svn.core.SVN_VER_PATCH]:
  sys.stderr.write(
    "You need version %s or better of the Subversion Python bindings.\n" \
    % ".".join([str(x) for x in _MIN_SVN_VERSION]))
  sys.exit(1)

# Absorb difference between Python 2 and Python >= 3
if PY3:
  def to_bytes(x):
    return x.encode('utf-8')

  def to_str(x):
    return x.decode('utf-8')

  # We never use sys.stdin nor sys.stdout TextIOwrapper.
  _stdin = sys.stdin.buffer
  _stdout = sys.stdout.buffer
else:
  # Python 2
  def to_bytes(x):
    return x

  def to_str(x):
    return x

  _stdin = sys.stdin
  _stdout = sys.stdout


SEPARATOR = '=' * 78

def main(pool, cmd, config_fname, repos_dir, cmd_args):
  ### TODO:  Sanity check the incoming args

  if cmd == 'commit':
    revision = int(cmd_args[0])
    repos = Repository(repos_dir, revision, pool)
    cfg = Config(config_fname, repos_dir,
                 {'author': repos.author,
                  'repos_basename': os.path.basename(repos.repos_dir)
                 })
    messenger = Commit(pool, cfg, repos)
  elif cmd == 'propchange' or cmd == 'propchange2':
    revision = int(cmd_args[0])
    author = cmd_args[1]
    propname = cmd_args[2]
    if cmd == 'propchange2' and cmd_args[3]:
      action = cmd_args[3]
    else:
      action = 'A'
    repos = Repository(repos_dir, revision, pool)
    # Override the repos revision author with the author of the propchange
    repos.author = author
    cfg = Config(config_fname, repos_dir,
                 {'author': author,
                  'repos_basename': os.path.basename(repos.repos_dir)
                 })
    messenger = PropChange(cfg, repos, author, propname, action)
  elif cmd == 'lock' or cmd == 'unlock':
    author = cmd_args[0]
    repos = Repository(repos_dir, 0, pool) ### any old revision will do
    # Override the repos revision author with the author of the lock/unlock
    repos.author = author
    cfg = Config(config_fname, repos_dir,
                 {'author': author,
                  'repos_basename': os.path.basename(repos.repos_dir)
                 })
    messenger = Lock(pool, cfg, repos, author, cmd == 'lock')
  else:
    raise UnknownSubcommand(cmd)

  output = create_output(cfg, repos)
  return messenger.generate(output, pool)


def create_output(cfg, repos):
    if cfg.is_set('general.mail_command'):
      cls = PipeOutput
    elif cfg.is_set('general.smtp_hostname'):
      cls = SMTPOutput
    else:
      cls = StandardOutput

    return cls(cfg, repos)


def remove_leading_slashes(path):
  while path and path[0:1] == b'/':
    path = path[1:]
  return path


class Writer:
  "Simple class for writing strings/binary, with optional encoding."

  def __init__(self, maxbytes, encoding):
    self.maxbytes = maxbytes
    self.buffer = BytesIO()

    # Attach a couple functions to SELF, rather than methods.
    self.write_binary = self.buffer.write

    if codecs.lookup(encoding) != codecs.lookup('utf-8'):
      def _write(s):
        "Write text string S using the given encoding."
        return self.buffer.write(s.encode(encoding, 'backslashreplace'))
    else:
      def _write(s):
        "Write text string S using the *default* encoding (utf-8)."
        return self.buffer.write(to_bytes(s))

    def write_limited(s):
        # If it looks like this write() will surpass the maximum length,
        # then bail out.
        if len(self.buffer.getbuffer()) + len(s) > self.maxbytes:
            raise MessageTooLarge
        return _write(s)

    self.write = write_limited


class OutputBase:
  "Abstract base class to formalize the interface of output methods"

  def __init__(self, cfg, repos):
    self.cfg = cfg
    self.repos = repos
    self._CHUNKSIZE = 128 * 1024

  def send(self, subject_line, group, params, long_func, short_func):
      ### get the MAXBYTEs from the configuration
      writer = Writer(90000, self.get_encoding())

      try:
          try:
              long_func(writer)
          except MessageTooLarge:
              writer.buffer.truncate(0)
              writer.buffer.seek(0)
              try:
                  short_func(writer)
              except MessageTooLarge:
                  # NOTE: don't use the Writer() API, or it will check the
                  # length again. Reach inside.
                  writer.buffer.write(b'\n\n\n... message too long. Truncated.\n')
                  # FALLTHRU
      except MessageSendFailure:
        return True  # failed

      self.deliver(subject_line, group, params, writer.buffer.getvalue())

      return False  # succeeded

  def get_encoding(self):
    """Get the encoding for text-to-bytes in the output.

    This will default to UTF-8. If the output mechanism needs a different
    encoding, then override this method to provide the custom encoding.
    """
    return 'utf-8'

  def deliver(self, subject_line, group, params, body):
    """Override this method.

    ### FIX THIS DOCSTRING

    Begin writing an output representation. SUBJECT_LINE is a subject line
    describing the action (commit, properties, lock), which may be tweaked
    given other conditions. GROUP is the name of the configuration file
    group which is causing this output to be produced. PARAMS is a
    dictionary of any named subexpressions of regular expressions defined
    in the configuration file, plus the key 'author' contains the author
    of the action being reported.

    Return bytes() for the prefix of the content to deliver.
    """
    raise NotImplementedError


class MailedOutput(OutputBase):

  def get_prefix(self, subject_line, group, params):
    # whitespace (or another character) separated list of addresses
    # which must be split into a clean list
    to_addr_in = self.cfg.get('to_addr', group, params)
    # if list of addresses starts with '[.]'
    # use the character between the square brackets as split char
    # else use whitespaces
    if len(to_addr_in) >= 3 and to_addr_in[0] == '[' \
                            and to_addr_in[2] == ']':
      self.to_addrs = \
        [_f for _f in to_addr_in[3:].split(to_addr_in[1]) if _f]
    else:
      self.to_addrs = [_f for _f in to_addr_in.split() if _f]
    self.from_addr = self.cfg.get('from_addr', group, params) \
                     or self.repos.author or 'no_author'
    # if the from_addr (also) starts with '[.]' (may happen if one
    # map is used for both to_addr and from_addr) remove '[.]'
    if len(self.from_addr) >= 3 and self.from_addr[0] == '[' \
                                and self.from_addr[2] == ']':
      self.from_addr = self.from_addr[3:]
    self.reply_to = self.cfg.get('reply_to', group, params)
    # if the reply_to (also) starts with '[.]' (may happen if one
    # map is used for both to_addr and reply_to) remove '[.]'
    if len(self.reply_to) >= 3 and self.reply_to[0] == '[' \
                               and self.reply_to[2] == ']':
      self.reply_to = self.reply_to[3:]

    # Return the prefix for the mail message.
    return self.mail_headers(subject_line, group)

  def _rfc2047_encode(self, hdr):
    # Return the result of splitting HDR into tokens (on space
    # characters), encoding (per RFC2047) each token as necessary, and
    # slapping 'em back to together again.
    from email.header import Header

    def _maybe_encode_header(hdr_token):
      try:
        hdr_token.encode('ascii')
        return hdr_token
      except UnicodeError:
        return Header(hdr_token, 'utf-8').encode()

    return ' '.join(map(_maybe_encode_header, hdr.split()))

  def mail_headers(self, subject_line, group):
    from email import utils

    subject  = self._rfc2047_encode(subject_line)
    from_hdr = self._rfc2047_encode(self.from_addr)
    to_hdr   = self._rfc2047_encode(', '.join(self.to_addrs))

    hdrs = 'From: %s\n' \
           'To: %s\n' \
           'Subject: %s\n' \
           'Date: %s\n' \
           'Message-ID: %s\n' \
           'MIME-Version: 1.0\n' \
           'Content-Type: text/plain; charset=UTF-8\n' \
           'Content-Transfer-Encoding: 8bit\n' \
           'X-Svn-Commit-Project: %s\n' \
           'X-Svn-Commit-Author: %s\n' \
           'X-Svn-Commit-Revision: %d\n' \
           'X-Svn-Commit-Repository: %s\n' \
           % (from_hdr, to_hdr, subject,
              utils.formatdate(), utils.make_msgid(), group,
              self.repos.author or 'no_author', self.repos.rev,
              os.path.basename(self.repos.repos_dir))
    if self.reply_to:
      hdrs = '%sReply-To: %s\n' % (hdrs, self.reply_to)
    return (hdrs + '\n').encode()


class SMTPOutput(MailedOutput):
  "Deliver a mail message to an MTA using SMTP."

  def deliver(self, subject_line, group, params, body):
    """
    Send email via SMTP or SMTP_SSL, logging in if username is
    specified.

    Errors such as invalid recipient, which affect a particular email,
    are reported to stderr and raise MessageSendFailure. If the caller
    has other emails to send, it may continue doing so.

    Errors caused by bad configuration, such as login failures, for
    which too many occurrences could lead to SMTP server lockout, are
    reported to stderr and re-raised. These should be considered fatal
    (to minimize the chances of said lockout).
    """

    prefix = self.get_prefix(subject_line, group, params)

    if self.cfg.is_set('general.smtp_port'):
       smtp_port = self.cfg.general.smtp_port
    else:
       smtp_port = 0
    try:
      if self.cfg.is_set('general.smtp_ssl') and self.cfg.general.smtp_ssl == 'yes':
        server = smtplib.SMTP_SSL(self.cfg.general.smtp_hostname, smtp_port)
      else:
        server = smtplib.SMTP(self.cfg.general.smtp_hostname, smtp_port)
    except Exception as detail:
      sys.stderr.write("mailer.py: Failed to instantiate SMTP object: %s\n" % (detail,))
      # Any error to instantiate is fatal
      raise

    try:
      if self.cfg.is_set('general.smtp_username'):
        try:
          server.login(self.cfg.general.smtp_username,
                       self.cfg.general.smtp_password)
        except smtplib.SMTPException as detail:
          sys.stderr.write("mailer.py: SMTP login failed with username %s and/or password: %s\n"
                           % (self.cfg.general.smtp_username, detail,))
          # Any error at login is fatal
          raise

      server.sendmail(self.from_addr, self.to_addrs, prefix + body)

    ### TODO: 'raise .. from' is Python 3+. When we convert this
    ###       script to Python 3, uncomment 'from detail' below
    ###       (2 instances):

    except smtplib.SMTPRecipientsRefused as detail:
      sys.stderr.write("mailer.py: SMTP recipient(s) refused: %s: %s\n"
                           % (self.to_addrs, detail,))
      raise MessageSendFailure ### from detail

    except smtplib.SMTPSenderRefused as detail:
      sys.stderr.write("mailer.py: SMTP sender refused: %s: %s\n"
                           % (self.from_addr, detail,))
      raise MessageSendFailure ### from detail

    except smtplib.SMTPException as detail:
      # All other errors are fatal; this includes:
      # SMTPHeloError, SMTPDataError, SMTPNotSupportedError
      sys.stderr.write("mailer.py: SMTP error occurred: %s\n" % (detail,))
      raise

    finally:
      try:
        server.quit()
      except smtplib.SMTPException as detail:
        sys.stderr.write("mailer.py: Error occurred during SMTP session cleanup: %s\n"
                             % (detail,))


class StandardOutput(OutputBase):
  "Print the commit message to stdout."

  def get_encoding(self):
    return sys.stdout.encoding if PY3 else 'utf-8'

  def deliver(self, subject_line, group, params, body):
      _stdout.write((
                        ("Group: " + (group or "defaults") + "\n")
                      + ("Subject: %s\n\n" % (subject_line,))
                    ).encode()  ### whoops. use the encoding
                    + body)


class PipeOutput(MailedOutput):
  "Deliver a mail message to an MTA via a pipe."

  def __init__(self, cfg, repos):
    MailedOutput.__init__(self, cfg, repos)

    # figure out the command for delivery
    self.cmd = cfg.general.mail_command.split()

  def deliver(self, subject_line, group, params, body):
    prefix = self.get_prefix(subject_line, group, params)

    ### gotta fix this. this is pretty specific to sendmail and qmail's
    ### mailwrapper program. should be able to use option param substitution
    cmd = self.cmd + [ '-f', self.from_addr ] + self.to_addrs

    # construct the pipe for talking to the mailer
    pipe = subprocess.Popen(cmd, stdin=subprocess.PIPE,
                            close_fds=sys.platform != "win32")

    # Send the content to the mailer, and wait for completion.
    pipe.communicate(prefix + body)


class Messenger:
  def __init__(self, cfg, repos, prefix_param):
    self.cfg = cfg
    self.repos = repos
    self.prefix_param = prefix_param

    # Subclasses should set this instance variable to describe the action
    # being performed. See OutputBase.start() docstring.
    self.basic_subject = ''

  def make_subject(self, basic_subject, group, params):
    prefix = self.cfg.get(self.prefix_param, group, params)
    if prefix:
      subject = prefix + ' ' + basic_subject
    else:
      subject = basic_subject

    try:
      truncate_subject = int(
          self.cfg.get('truncate_subject', group, params))
    except ValueError:
      truncate_subject = 0

    # truncate subject as UTF-8 string.
    # Note: there still exists an issue on combining characters.
    if truncate_subject:
      bsubject = to_bytes(subject)
      if len(bsubject) > truncate_subject:
        idx = truncate_subject - 2
        while b'\x80' <= bsubject[idx-1:idx] <= b'\xbf':
          idx -= 1
        subject = to_str(bsubject[:idx-1]) + "..."

    return subject


class Commit(Messenger):
  def __init__(self, pool, cfg, repos):
    Messenger.__init__(self, cfg, repos, 'commit_subject_prefix')

    # get all the changes and sort by path
    editor = svn.repos.ChangeCollector(repos.fs_ptr, repos.root_this, pool)
    e_ptr, e_baton = svn.delta.make_editor(editor, pool)
    svn.repos.replay2(repos.root_this, "", svn.core.SVN_INVALID_REVNUM, 1, e_ptr, e_baton, None, pool)

    self.changelist = sorted(editor.get_changes().items())

    log = to_str(repos.get_rev_prop(svn.core.SVN_PROP_REVISION_LOG, pool) or b'')

    # collect the set of groups and the unique sets of params for the options
    self.groups = { }
    for path, change in self.changelist:
      for (group, params) in self.cfg.which_groups(to_str(path), log):
        # turn the params into a hashable object and stash it away
        param_list = sorted(params.items())
        # collect the set of paths belonging to this group
        if (group, tuple(param_list)) in self.groups:
          old_param, paths = self.groups[group, tuple(param_list)]
        else:
          paths = { }
        paths[path] = None
        self.groups[group, tuple(param_list)] = (params, paths)

    # figure out the changed directories
    dirs = { }
    for path, change in self.changelist:
      path = to_str(path)
      if change.item_kind == svn.core.svn_node_dir:
        dirs[path] = None
      else:
        idx = path.rfind('/')
        if idx == -1:
          dirs[''] = None
        else:
          dirs[path[:idx]] = None

    dirlist = list(dirs.keys())

    commondir, dirlist = get_commondir(dirlist)

    # compose the basic subject line. later, we can prefix it.
    dirlist_s = ' '.join(sorted(dirlist))
    if commondir:
      self.basic_subject = 'r%d - in %s: %s' % (repos.rev, commondir, dirlist_s)
    else:
      self.basic_subject = 'r%d - %s' % (repos.rev, dirlist_s)

  def generate(self, output, scratch_pool):
    "Generate email for the various groups and option-params."

    ### the groups need to be further compressed. if the headers and
    ### body are the same across groups, then we can have multiple To:
    ### addresses. SMTPOutput holds the entire message body in memory,
    ### so if the body doesn't change, then it can be sent N times
    ### rather than rebuilding it each time.

    iterpool = svn.core.svn_pool_create(scratch_pool)
    failed = False

    for (group, param_tuple), (params, paths) in sorted(self.groups.items()):
      subject_line = self.make_subject(self.basic_subject, group, params)

      def long_commit(writer):
        # generate commit message (with diffs) for this group and params
        generate_commit(writer, self.cfg, self.repos, self.changelist, False,
                        group, params, paths, iterpool)

      def short_commit(writer):
        # generate a shorter message, using URLs instead of diffs
        generate_commit(writer, self.cfg, self.repos, self.changelist, True,
                        group, params, paths, iterpool)

      failed |= output.send(subject_line, group, params,
                            long_commit, short_commit)
      svn.core.svn_pool_clear(iterpool)

    svn.core.svn_pool_destroy(iterpool)
    return failed


class PropChange(Messenger):
  def __init__(self, cfg, repos, author, propname, action):
    Messenger.__init__(self, cfg, repos, 'propchange_subject_prefix')
    self.author = author
    self.propname = propname
    self.action = action

    # collect the set of groups and the unique sets of params for the options
    self.groups = { }
    for (group, params) in self.cfg.which_groups('', None):
      # turn the params into a hashable object and stash it away
      param_list = sorted(params.items())
      self.groups[group, tuple(param_list)] = params

    self.basic_subject = 'r%d - %s' % (repos.rev, propname)

  def generate(self, output, scratch_pool):
    actions = { 'A': 'added', 'M': 'modified', 'D': 'deleted' }
    failed = False
    ### maybe create an iterpool?

    for (group, param_tuple), params in self.groups.items():
      subject_line = self.make_subject(self.basic_subject, group, params)

      def long_propchange(writer):
        writer.write('Author: %s\n'
                          'Revision: %s\n'
                          'Property Name: %s\n'
                          'Action: %s\n'
                          '\n'
                          % (self.author, self.repos.rev, self.propname,
                             actions.get(self.action, 'Unknown (\'%s\')' \
                                         % self.action)))
        if self.action == 'A' or self.action not in actions:
          writer.write('Property value:\n')
          propvalue = self.repos.get_rev_prop(self.propname, scratch_pool)
          writer.write(propvalue)
        elif self.action == 'M':
          writer.write('Property diff:\n')
          tempfile1 = tempfile.NamedTemporaryFile()
          tempfile1.write(_stdin.read())
          tempfile1.flush()
          tempfile2 = tempfile.NamedTemporaryFile()
          tempfile2.write(self.repos.get_rev_prop(self.propname, scratch_pool))
          tempfile2.flush()
          for diffs in generate_diff(self.cfg.get_diff_cmd(group, {
              'label_from' : 'old property value',
              'label_to' : 'new property value',
              'from' : tempfile1.name,
              'to' : tempfile2.name,
              })):
              writer.write(to_str(diffs.raw))
      failed |= output.send(subject_line, group, params, long_propchange, None)

    return failed


def get_commondir(dirlist):
  """Figure out the common portion/parent (commondir) of all the paths
  in DIRLIST and return a tuple consisting of commondir, dirlist.  If
  a commondir is found, the dirlist returned is rooted in that
  commondir.  If no commondir is found, dirlist is returned unchanged,
  and commondir is the empty string."""
  if len(dirlist) < 2 or '/' in dirlist:
    commondir = ''
    newdirs = dirlist
  else:
    common = dirlist[0].split('/')
    for j in range(1, len(dirlist)):
      d = dirlist[j]
      parts = d.split('/')
      for i in range(len(common)):
        if i == len(parts) or common[i] != parts[i]:
          del common[i:]
          break
    commondir = '/'.join(common)
    if commondir:
      # strip the common portion from each directory
      l = len(commondir) + 1
      newdirs = [ ]
      for d in dirlist:
        if d == commondir:
          newdirs.append('.')
        else:
          newdirs.append(d[l:])
    else:
      # nothing in common, so reset the list of directories
      newdirs = dirlist

  return commondir, newdirs


class Lock(Messenger):
  def __init__(self, pool, cfg, repos, author, do_lock):
    self.author = author
    self.do_lock = do_lock

    Messenger.__init__(self, cfg, repos,
                       (do_lock and 'lock_subject_prefix'
                        or 'unlock_subject_prefix'))

    # read all the locked paths from STDIN and strip off the trailing newlines
    self.dirlist = [to_str(x).rstrip() for x in _stdin.readlines()]

    # collect the set of groups and the unique sets of params for the options
    self.groups = { }
    for path in self.dirlist:
      for (group, params) in self.cfg.which_groups(path, None):
        # turn the params into a hashable object and stash it away
        param_list = sorted(params.items())
        # collect the set of paths belonging to this group
        if (group, tuple(param_list)) in self.groups:
          old_param, paths = self.groups[group, tuple(param_list)]
        else:
          paths = { }
        paths[path] = None
        self.groups[group, tuple(param_list)] = (params, paths)

    commondir, dirlist = get_commondir(self.dirlist)

    # compose the basic subject line. later, we can prefix it.
    dirlist_s = ' '.join(sorted(dirlist))
    if commondir:
      self.basic_subject = '%s: %s' % (commondir, dirlist_s)
    else:
      self.basic_subject = dirlist_s

    # The lock comment is the same for all paths, so we can just pull
    # the comment for the first path in the dirlist and cache it.
    self.lock = svn.fs.svn_fs_get_lock(self.repos.fs_ptr,
                                       to_bytes(self.dirlist[0]),
                                       pool)

  def generate(self, output, scratch_pool):
    failed = False

    for (group, param_tuple), (params, paths) in sorted(self.groups.items()):
      subject_line = self.make_subject(self.basic_subject, group, params)

      def long_lock(writer):
        writer.write('Author: %s\n'
                          '%s paths:\n' %
                          (self.author, self.do_lock and 'Locked' or 'Unlocked'))

        self.dirlist.sort()
        for dir in self.dirlist:
          writer.write('   %s\n\n' % dir)

        if self.do_lock:
          writer.write('Comment:\n%s\n' % (self.lock.comment or ''))

      failed |= output.send(subject_line, group, params, long_lock, None)

    return failed


class DiffSelections:
  def __init__(self, cfg, group, params):
    self.add = False
    self.copy = False
    self.delete = False
    self.modify = False

    gen_diffs = cfg.get('generate_diffs', group, params)

    ### Do a little dance for deprecated options.  Note that even if you
    ### don't have an option anywhere in your configuration file, it
    ### still gets returned as non-None.
    if len(gen_diffs):
      list = gen_diffs.split(" ")
      for item in list:
        if item == 'add':
          self.add = True
        if item == 'copy':
          self.copy = True
        if item == 'delete':
          self.delete = True
        if item == 'modify':
          self.modify = True
    else:
      self.add = True
      self.copy = True
      self.delete = True
      self.modify = True
      ### These options are deprecated
      suppress = cfg.get('suppress_deletes', group, params)
      if suppress == 'yes':
        self.delete = False
      suppress = cfg.get('suppress_adds', group, params)
      if suppress == 'yes':
        self.add = False


class DiffURLSelections:
  def __init__(self, cfg, group, params):
    self.cfg = cfg
    self.group = group
    self.params = params

  def _get_url(self, action, repos_rev, change):
    # The parameters for the URLs generation need to be placed in the
    # parameters for the configuration module, otherwise we may get
    # KeyError exceptions.
    params = self.params.copy()
    params['path'] = _url_quote(change.path) if change.path else None
    params['base_path'] = (_url_quote(change.base_path)
                           if change.base_path else None)
    params['rev'] = repos_rev
    params['base_rev'] = change.base_rev

    return self.cfg.get("diff_%s_url" % action, self.group, params)

  def get_add_url(self, repos_rev, change):
    return self._get_url('add', repos_rev, change)

  def get_copy_url(self, repos_rev, change):
    return self._get_url('copy', repos_rev, change)

  def get_delete_url(self, repos_rev, change):
    return self._get_url('delete', repos_rev, change)

  def get_modify_url(self, repos_rev, change):
    return self._get_url('modify', repos_rev, change)


def generate_commit(writer, cfg, repos, changelist, no_diff_content,
                    group, params, paths, pool):

  svndate = repos.get_rev_prop(svn.core.SVN_PROP_REVISION_DATE, pool)
  ### pick a different date format?
  date = time.ctime(svn.core.secs_from_timestr(svndate, pool))

  show_nonmatching_paths = cfg.get('show_nonmatching_paths', group, params) \
      or 'yes'

  params_with_rev = params.copy()
  params_with_rev['rev'] = repos.rev
  commit_url = cfg.get('commit_url', group, params_with_rev)

  # figure out the lists of changes outside the selected path-space
  if len(paths) != len(changelist) and show_nonmatching_paths != 'no':
    other_summary = generate_summary(changelist, paths, False)
  else:
    other_summary = None

  if len(paths) != len(changelist) and show_nonmatching_paths == 'yes':
    other_diffs = generate_changelist_diffs(cfg, repos, changelist,
                                            no_diff_content, False,
                                            group, params, paths,
                                            date, pool)
  else:
    other_diffs = None

  summary = generate_summary(changelist, paths, True)

  data = _data(
    author=repos.author,
    date=date,
    rev=repos.rev,
    log=to_str(repos.get_rev_prop(svn.core.SVN_PROP_REVISION_LOG, pool) or b''),
    commit_url=commit_url,
    summary=summary,
    no_diff_content=no_diff_content,
    show_nonmatching_paths=show_nonmatching_paths,
    other_summary=other_summary,
    diffs=generate_changelist_diffs(cfg, repos, changelist,
                                    no_diff_content, True,
                                    group, params, paths, date, pool),
    other_diffs=other_diffs,
    )
  ### clean this up in future rev. Just use wb
  w = writer.write
  wb = writer.write_binary
  render_commit(w, wb, data)


def generate_summary(changelist, paths, in_paths):
  def gather_info(action):
    return _gather_paths(action, changelist, paths, in_paths)
  return _data(
    added=gather_info(svn.repos.CHANGE_ACTION_ADD),
    replaced=gather_info(svn.repos.CHANGE_ACTION_REPLACE),
    deleted=gather_info(svn.repos.CHANGE_ACTION_DELETE),
    modified=gather_info(svn.repos.CHANGE_ACTION_MODIFY),
    )


def _gather_paths(action, changelist, paths, in_paths):
  items = [ ]
  for path, change in changelist:
    if change.action == action and (path in paths) == in_paths:
      item = _data(
        path=path,
        is_dir=change.item_kind == svn.core.svn_node_dir,
        props_changed=change.prop_changes,
        text_changed=change.text_changed,
        copied=(change.action == svn.repos.CHANGE_ACTION_ADD \
                or change.action == svn.repos.CHANGE_ACTION_REPLACE) \
               and change.base_path,
        base_path=remove_leading_slashes(change.base_path),
        base_rev=change.base_rev,
        )
      items.append(item)

  return items


def generate_changelist_diffs(cfg, repos, changelist,
                              no_diff_content, in_paths,
                              group, params, paths, date, pool):
    "This is a generator returning diffs for each change."

    diffsels = DiffSelections(cfg, group, params)
    diffurls = DiffURLSelections(cfg, group, params)

    for path, change in changelist:

      diff = diff_url = None
      kind = None
      label1 = None
      label2 = None
      src_fname = None
      dst_fname = None
      binary = None
      singular = None
      content = None

      # just skip directories. they have no diffs.
      if change.item_kind == svn.core.svn_node_dir:
        continue

      # is this change in (or out of) the set of matched paths?
      if (path in paths) != in_paths:
        continue

      if change.base_rev != -1:
        svndate = repos.get_rev_prop(svn.core.SVN_PROP_REVISION_DATE,
                                     pool, change.base_rev)
        ### pick a different date format?
        base_date = time.ctime(svn.core.secs_from_timestr(svndate, pool))
      else:
        base_date = ''

      # figure out if/how to generate a diff

      base_path_bytes = remove_leading_slashes(change.base_path)
      base_path = (to_str(base_path_bytes)
                   if base_path_bytes is not None else None)
      if change.action == svn.repos.CHANGE_ACTION_DELETE:
        # it was delete.
        kind = 'D'

        # get the diff url, if any is specified
        diff_url = diffurls.get_delete_url(repos.rev, change)

        # show the diff?
        if diffsels.delete:
          diff = svn.fs.FileDiff(repos.get_root(change.base_rev),
                                 base_path_bytes, None, None, pool)

          label1 = '%s\t%s\t(r%s)' % (base_path, date, change.base_rev)
          label2 = '/dev/null\t00:00:00 1970\t(deleted)'
          singular = True

      elif change.action == svn.repos.CHANGE_ACTION_ADD \
           or change.action == svn.repos.CHANGE_ACTION_REPLACE:
        if base_path and (change.base_rev != -1):

          # any diff of interest?
          if change.text_changed:
            # this file was copied and modified.
            kind = 'W'

            # get the diff url, if any is specified
            diff_url = diffurls.get_copy_url(repos.rev, change)

            # show the diff?
            if diffsels.modify:
              diff = svn.fs.FileDiff(repos.get_root(change.base_rev),
                                     base_path_bytes,
                                     repos.root_this, change.path,
                                     pool)
              label1 = ('%s\t%s\t(r%s, copy source)'
                        % (base_path, base_date, change.base_rev))
              label2 = ('%s\t%s\t(r%s)'
                        % (to_str(change.path), date, repos.rev))
              singular = False
          else:
            # this file was copied.
            kind = 'C'
            if diffsels.copy:
              diff = svn.fs.FileDiff(None, None, repos.root_this,
                                     change.path, pool)
              label1 = ('/dev/null\t00:00:00 1970\t'
                        '(empty, because file is newly added)')
              label2 = ('%s\t%s\t(r%s, copy of r%s, %s)'
                        % (to_str(change.path),
                           date, repos.rev, change.base_rev,
                           base_path))
              singular = False
        else:
          # the file was added.
          kind = 'A'

          # get the diff url, if any is specified
          diff_url = diffurls.get_add_url(repos.rev, change)

          # show the diff?
          if diffsels.add:
            diff = svn.fs.FileDiff(None, None, repos.root_this,
                                   change.path, pool)
            label1 = '/dev/null\t00:00:00 1970\t' \
                     '(empty, because file is newly added)'
            label2 = '%s\t%s\t(r%s)' \
                     % (to_str(change.path), date, repos.rev)
            singular = True

      elif not change.text_changed:
        # the text didn't change, so nothing to show.
        continue
      else:
        # a simple modification.
        kind = 'M'

        # get the diff url, if any is specified
        diff_url = diffurls.get_modify_url(repos.rev, change)

        # show the diff?
        if diffsels.modify:
          diff = svn.fs.FileDiff(repos.get_root(change.base_rev),
                                 base_path,
                                 repos.root_this, change.path,
                                 pool)
          label1 = '%s\t%s\t(r%s)' \
                   % (base_path, base_date, change.base_rev)
          label2 = '%s\t%s\t(r%s)' \
                   % (to_str(change.path), date, repos.rev)
          singular = False

      if diff:
        binary = diff.either_binary()
        if binary or no_diff_content:
          content = src_fname = dst_fname = None
        else:
            src_fname, dst_fname = diff.get_files()
            content = generate_diff(cfg.get_diff_cmd(group, {
              'label_from' : label1,
              'label_to' : label2,
              'from' : src_fname,
              'to' : dst_fname,
              }))

      # return a data item for this diff
      yield _data(
        path=change.path,
        base_path=base_path_bytes,
        base_rev=change.base_rev,
        diff=diff,
        diff_url=diff_url,
        kind=kind,
        label_from=label1,
        label_to=label2,
        from_fname=src_fname,
        to_fname=dst_fname,
        binary=binary,
        singular=singular,
        content=content,
        )


def _classify_diff_line(line, seen_change):
  # classify the type of line.
  first = line[:1]
  ltype = ''
  if first == '@':
    seen_change = True
    ltype = 'H'
  elif first == '-':
    if seen_change:
      ltype = 'D'
    else:
      ltype = 'F'
  elif first == '+':
    if seen_change:
      ltype = 'A'
    else:
      ltype = 'T'
  elif first == ' ':
    ltype = 'C'
  else:
    ltype = 'U'

  if line[-2] == '\r':
    line=line[0:-2] + '\n' # remove carriage return

  return line, ltype, seen_change


def generate_diff(cmd):
    seen_change = False

    # By default we choose to incorporate child stderr into the output
    pipe = subprocess.Popen(cmd,
                            stdout=subprocess.PIPE,
                            stderr=subprocess.STDOUT,
                            close_fds=sys.platform != "win32")

    while True:
        line = pipe.stdout.readline()
        if not line:
            # wait on the child so we don't end up with a billion zombies
            pipe.wait()
            return  # will raise StopIteration

        line, ltype, seen_change = _classify_diff_line(line, seen_change)
        yield _data(
            raw=line,
            text=line[1:-1],  # remove indicator and newline
            type=ltype,
        )


def render_commit(w, wb, data):
    "Call W and/or WB to render the commit defined by DATA."

    w('Author: %s\nDate: %s\nNew Revision: %s\n' % (data.author,
                                                      data.date,
                                                      data.rev))

    if data.commit_url:
      w('URL: %s\n\n' % data.commit_url)
    else:
      w('\n')

    w('Log:\n%s\n\n' % data.log.strip())

    # print summary sections
    _render_summary(w, data.summary)

    if data.other_summary:
      if data.show_nonmatching_paths:
        w('\nChanges in other areas also in this revision:\n')
        _render_summary(w, data.other_summary)
      else:
        w('and changes in other areas\n')

    if data.no_diff_content:
        w('\nNOTE: this message was too long when including "diff" contents.'
          '\n      The contents have been replaced with URLs to display the'
          '\n      diff contents on a web page.'
          '\n\n')

    _render_diffs(w, wb, data.diffs, '')
    if data.other_diffs:
        _render_diffs(w, wb, data.other_diffs,
                         '\nDiffs of changes in other areas also'
                         ' in this revision:\n')


def _render_summary(w, summary):
    _render_list(w, 'Added', summary.added)
    _render_list(w, 'Replaced', summary.replaced)
    _render_list(w, 'Deleted', summary.deleted)
    _render_list(w, 'Modified', summary.modified)


def _render_list(w, header, data_list):
    if not data_list:
      return

    w(header + ':\n')
    for d in data_list:
      if d.is_dir:
        is_dir = '/'
      else:
        is_dir = ''
      if d.props_changed:
        if d.text_changed:
          props = '   (contents, props changed)'
        else:
          props = '   (props changed)'
      else:
        props = ''
      w('   %s%s%s\n' % (to_str(d.path), is_dir, props))
      if d.copied:
        if is_dir:
          text = ''
        elif d.text_changed:
          text = ', changed'
        else:
          text = ' unchanged'
        w('      - copied%s from r%d, %s%s\n'
          % (text, d.base_rev, to_str(d.base_path), is_dir))


def _render_diffs(w, wb, diffs, section_header):
    """Render diffs. Write the SECTION_HEADER if there are actually
    any diffs to render."""
    if not diffs:
      return

    section_header_printed = False

    for diff in diffs:
      if not diff.diff and not diff.diff_url:
        continue
      if not section_header_printed:
        w(section_header)
        section_header_printed = True
      if diff.kind == 'D':
        w('\nDeleted: %s\n' % to_str(diff.base_path))
      elif diff.kind == 'A':
        w('\nAdded: %s\n' % to_str(diff.path))
      elif diff.kind == 'C':
        w('\nCopied: %s (from r%d, %s)\n'
          % (to_str(diff.path), diff.base_rev,
             to_str(diff.base_path)))
      elif diff.kind == 'W':
        w('\nCopied and modified: %s (from r%d, %s)\n'
          % (to_str(diff.path), diff.base_rev,
             to_str(diff.base_path)))
      else:
        # kind == 'M'
        w('\nModified: %s\n' % to_str(diff.path))

      if diff.diff_url:
        w('URL: %s\n' % diff.diff_url)

      if not diff.diff:
        continue

      if diff.binary:
        w(SEPARATOR + '\n')
        if diff.singular:
          w('Binary file. No diff available.\n')
        else:
          w('Binary file (source and/or target). No diff available.\n')
        continue

      if diff.content is None:
          continue

      w(SEPARATOR + '\n')
      for line in diff.content:
        wb(line.raw)


class Repository:
  "Hold roots and other information about the repository."

  def __init__(self, repos_dir, rev, pool):
    self.repos_dir = repos_dir
    self.rev = rev

    # Any data that we HOLD will be allocated in this pool.
    self.hold_pool = pool

    self.repos_ptr = svn.repos.open(repos_dir, pool)
    self.fs_ptr = svn.repos.fs(self.repos_ptr)

    self.roots = { }

    self.root_this = self.get_root(rev)

    self.author = self.get_rev_prop(svn.core.SVN_PROP_REVISION_AUTHOR, pool)
    if self.author is not None:
      self.author = to_str(self.author)

  def get_rev_prop(self, propname, scratch_pool, rev=None):
    if not rev:
      rev = self.rev
    return svn.fs.revision_prop(self.fs_ptr, rev, propname, scratch_pool)

  def get_root(self, rev):
    try:
      return self.roots[rev]
    except KeyError:
      pass
    root = self.roots[rev] = svn.fs.revision_root(self.fs_ptr, rev, self.hold_pool)
    return root


class Config:

  # The predefined configuration sections. These are omitted from the
  # set of groups.
  PREDEFINED = ('general', 'defaults', 'maps')

  def __init__(self, fname, repos_dir, default_params):
    cp = configparser.ConfigParser()
    cp.read(fname)

    # record the (non-default) groups that we find
    ordered_groups = [ ]

    def build_section_ob(section_name):
        ob = _sub_section()
        # Get the raw values. We use the same format for *our* interpolation.
        vars(ob).update(cp.items(section_name, raw=True))
        return ob

    # The config file MUST contain [general] and [defaults]; maybe [maps]
    self.general = build_section_ob('general')
    self.defaults = build_section_ob('defaults')
    if cp.has_section('maps'):
        self.maps = build_section_ob('maps')

    # Process all other sections as groups.
    for section in cp.sections():
        if section in self.PREDEFINED:
            continue
        assert not hasattr(self, section)

        ordered_groups.append(section)

        setattr(self, section, build_section_ob(section))

    # be compatible with old format config files
    if hasattr(self.general, 'diff') and not hasattr(self.defaults, 'diff'):
      self.defaults.diff = self.general.diff
    if not hasattr(self, 'maps'):
      self.maps = _sub_section()

    # prepare maps. this may remove sections from consideration as a group.
    mapsections = self._prep_maps()
    for sectname in mapsections:
        ordered_groups.remove(sectname)

    # process all the group sections.
    # NOTE: path-group testing and precedence is in file-order. Use a list
    #   instead of a set to maintain the match ordering.
    self._prep_groups(ordered_groups, repos_dir, default_params)

    ### for t3, provide this instance var
    self._groups = ordered_groups

  def is_set(self, option):
    """Return None if the option is not set; otherwise, its value is returned.

    The option is specified as a dotted symbol, such as 'general.mail_command'
    """
    ob = self
    for part in option.split('.'):
      if not hasattr(ob, part):
        return None
      ob = getattr(ob, part)
    return ob

  def get(self, option, group, params):
    "Get a config value with appropriate substitutions and value mapping."

    # find the right value
    value = None
    if group:
      sub = getattr(self, group)
      value = getattr(sub, option, None)
    if value is None:
      value = getattr(self.defaults, option, '')

    # parameterize it
    if params is not None:
      value = value % params

    # apply any mapper
    mapper = getattr(self.maps, option, None)
    if mapper is not None:
      value = mapper(value)

      # Apply any parameters that may now be available for
      # substitution that were not before the mapping.
      if value is not None and params is not None:
        value = value % params

    return value

  def get_diff_cmd(self, group, args):
    "Get a diff command as a list of argv elements."
    ### do some better splitting to enable quoting of spaces
    diff_cmd = self.get('diff', group, None).split()

    cmd = [ ]
    for part in diff_cmd:
      cmd.append(part % args)
    return cmd

  def _prep_maps(self):
    """Rewrite the [maps] options into callables that look up values.

    Returns a set of section names that are used for mappings, which
    should not be considered as path-match groups.
    """

    mapsections = set()

    for optname, mapvalue in vars(self.maps).items():
      if mapvalue[:1] == '[':
        # a section is acting as a mapping
        sectname = mapvalue[1:-1]
        if not hasattr(self, sectname):
          raise UnknownMappingSection(sectname)
        # construct a lambda to look up the given value as an option name,
        # and return the option's value. if the option is not present,
        # then just return the value unchanged.
        setattr(self.maps, optname,
                lambda value,
                       sect=getattr(self, sectname): getattr(sect,
                                                             value.lower(),
                                                             value))
        # mark for removal when all optnames are done
        mapsections.add(sectname)

      # elif test for other mapper types. possible examples:
      #   dbm:filename.db
      #   file:two-column-file.txt
      #   ldap:some-query-spec
      # just craft a mapper function and insert it appropriately

      else:
        raise UnknownMappingSpec(mapvalue)

    # remove each mapping section from consideration as a group
    return mapsections

  def _prep_groups(self, groups, repos_dir, default_params):
    self._group_re = [ ]

    ### does it arrive as an abspath?
    repos_dir = os.path.abspath(repos_dir)

    def repos_params(section_name, defaults):
        "Build key/value params for this section, based on current repos."

        section = getattr(self, section_name)  # should exist
        if hasattr(section, 'for_repos'):
            match = re.match(section.for_repos, repos_dir)
            if not match:
                # The FOR_REPOS selector does not apply to this repository.
                return None  # no params at all

            # Extract key/value pairs from the regex match of this
            # repository, and merge them into the default params.
            # Make sure to copy() to avoid mutation of the argument.
            return defaults.copy().update(match.groupdict())

        # There are no repository-specific key/value params, to add.
        return defaults

    # compute the default repository-based parameters. start with some
    # basic parameters, then bring in the regex-based params.
    # Note: use the defaults, even if selected-against by FOR_REPOS
    self._default_params = (repos_params('defaults', default_params)
                            or default_params)

    # select the groups that apply to this repository
    for group in groups:
      params = repos_params(group, self._default_params)
      if params is None:
          # There was a FOR_REPOS, but this repos does not match.
          # Thus, ignore this param group.
          continue

      sub = getattr(self, group)

      # if a matching rule hasn't been given, then use the empty string
      # as it will match all paths
      for_paths = getattr(sub, 'for_paths', '')

      # Build an optional regex to exclude some change paths.
      exclude_paths = getattr(sub, 'exclude_paths', None)
      if exclude_paths:
        exclude_paths_re = re.compile(exclude_paths)
      else:
        exclude_paths_re = None

      # Build an optional regex to extract key/value pairs from the
      # log message to augment the params.
      search_logmsg = getattr(sub, 'search_logmsg', None)
      if search_logmsg is not None:
        search_logmsg_re = re.compile(search_logmsg)
      else:
        search_logmsg_re = None

      self._group_re.append((group,
                             re.compile(for_paths),
                             exclude_paths_re,
                             params,
                             search_logmsg_re))

    # after all the groups are done, add in the default group
    try:
      self._group_re.append((None,
                             re.compile(self.defaults.for_paths),
                             None,
                             self._default_params,
                             None))
    except AttributeError:
      # there is no self.defaults.for_paths
      pass

  def which_groups(self, path, logmsg):
    "Return the path's associated groups."
    groups = []
    for group, pattern, exclude_pattern, repos_params, search_logmsg_re in self._group_re:
      match = pattern.match(path)
      if match:
        if exclude_pattern and exclude_pattern.match(path):
          continue
        params = repos_params.copy()
        params.update(match.groupdict())

        if search_logmsg_re is None:
          groups.append((group, params))
        else:
          if logmsg is None:
            logmsg = ''

          for match in search_logmsg_re.finditer(logmsg):
            # Add captured variables to (a copy of) params
            msg_params = params.copy()
            msg_params.update(match.groupdict())
            groups.append((group, msg_params))

    if not groups:
      groups.append((None, self._default_params))

    return groups


class _sub_section:
  pass

class _data:
  "Helper class to define an attribute-based hunk o' data."
  def __init__(self, **kw):
    vars(self).update(kw)

class MissingConfig(Exception):
  pass
class UnknownMappingSection(Exception):
  pass
class UnknownMappingSpec(Exception):
  pass
class UnknownSubcommand(Exception):
  pass
class MessageSendFailure(Exception):
  pass
class MessageTooLarge(Exception):
  pass


if __name__ == '__main__':
  def usage():
    scriptname = os.path.basename(sys.argv[0])
    sys.stderr.write(
"""USAGE: %s commit      REPOS REVISION [CONFIG-FILE]
       %s propchange  REPOS REVISION AUTHOR REVPROPNAME [CONFIG-FILE]
       %s propchange2 REPOS REVISION AUTHOR REVPROPNAME ACTION [CONFIG-FILE]
       %s lock        REPOS AUTHOR [CONFIG-FILE]
       %s unlock      REPOS AUTHOR [CONFIG-FILE]

If no CONFIG-FILE is provided, the script will first search for a mailer.conf
file in REPOS/conf/.  Failing that, it will search the directory in which
the script itself resides.

ACTION was added as a fifth argument to the post-revprop-change hook
in Subversion 1.2.0.  Its value is one of 'A', 'M' or 'D' to indicate
if the property was added, modified or deleted, respectively.

""" % (scriptname, scriptname, scriptname, scriptname, scriptname))
    sys.exit(1)

  # Command list:  subcommand -> number of arguments expected (not including
  #                              the repository directory and config-file)
  cmd_list = {'commit'     : 1,
              'propchange' : 3,
              'propchange2': 4,
              'lock'       : 1,
              'unlock'     : 1,
              }

  config_fname = None
  argc = len(sys.argv)
  if argc < 3:
    usage()

  cmd = sys.argv[1]
  repos_dir = to_str(svn.core.svn_path_canonicalize(to_bytes(sys.argv[2])))
  try:
    expected_args = cmd_list[cmd]
  except KeyError:
    usage()

  if argc < (expected_args + 3):
    usage()
  elif argc > expected_args + 4:
    usage()
  elif argc == (expected_args + 4):
    config_fname = sys.argv[expected_args + 3]

  # Settle on a config file location, and open it.
  if config_fname is None:
    # Default to REPOS-DIR/conf/mailer.conf.
    config_fname = os.path.join(repos_dir, 'conf', 'mailer.conf')
    if not os.path.exists(config_fname):
      # Okay.  Look for 'mailer.conf' as a sibling of this script.
      config_fname = os.path.join(os.path.dirname(sys.argv[0]), 'mailer.conf')
  if not os.path.exists(config_fname):
    raise MissingConfig(config_fname)

  failed = svn.core.run_app(main, cmd, config_fname, repos_dir,
                            sys.argv[3:3+expected_args])
  sys.exit(1 if failed else 0)


# ------------------------------------------------------------------------
# TODO
#
# * add configuration options
#   - each group defines delivery info:
#     o whether to set Reply-To and/or Mail-Followup-To
#       (btw: it is legal do set Reply-To since this is the originator of the
#        mail; i.e. different from MLMs that munge it)
#   - each group defines content construction:
#     o max size of diff before trimming
#     o max size of entire commit message before truncation
#   - per-repository configuration
#     o extra config living in repos
#     o optional, non-mail log file
#     o look up authors (username -> email; for the From: header) in a
#       file(s) or DBM
# * get rid of global functions that should properly be class methods

#
# For Emacs, we want to move towards the standard 4-space indent. It
# inspects the current formatting of this file, and sets 2-space.
# Override that with a 4-space indent.
#
# Local Variables:
# python-indent-offset: 4
# End:
#