--version command should be safe, and bare no modifications

- improved subprocess calls error detection
- fixed I/O read on closed file errors

rhodecode/lib/subprocessio.py

@@ -1,409 +1,415 b''
 '''
 Module provides a class allowing to wrap communication over subprocess.Popen
 input, output, error streams into a meaningfull, non-blocking, concurrent
 stream processor exposing the output data as an iterator fitting to be a
 return value passed by a WSGI applicaiton to a WSGI server per PEP 3333.
 
 Copyright (c) 2011  Daniel Dotsenko <[e-mail]>
 
 This file is part of git_http_backend.py Project.
 
 git_http_backend.py Project is free software: you can redistribute it and/or
 modify it under the terms of the GNU Lesser General Public License as
 published by the Free Software Foundation, either version 2.1 of the License,
 or (at your option) any later version.
 
 git_http_backend.py Project is distributed in the hope that it will be useful,
 but WITHOUT ANY WARRANTY; without even the implied warranty of
 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
 GNU Lesser General Public License for more details.
 
 You should have received a copy of the GNU Lesser General Public License
 along with git_http_backend.py Project.
 If not, see <http://www.gnu.org/licenses/>.
 '''
 import os
 import subprocess
 from rhodecode.lib.compat import deque, Event, Thread, _bytes, _bytearray
 
 
 class StreamFeeder(Thread):
     """
     Normal writing into pipe-like is blocking once the buffer is filled.
     This thread allows a thread to seep data from a file-like into a pipe
     without blocking the main thread.
     We close inpipe once the end of the source stream is reached.
     """
     def __init__(self, source):
         super(StreamFeeder, self).__init__()
         self.daemon = True
         filelike = False
         self.bytes = _bytes()
         if type(source) in (type(''), _bytes, _bytearray):  # string-like
             self.bytes = _bytes(source)
         else:  # can be either file pointer or file-like
             if type(source) in (int, long):  # file pointer it is
                 ## converting file descriptor (int) stdin into file-like
                 try:
                     source = os.fdopen(source, 'rb', 16384)
                 except Exception:
                     pass
             # let's see if source is file-like by now
             try:
                 filelike = source.read
             except Exception:
                 pass
         if not filelike and not self.bytes:
             raise TypeError("StreamFeeder's source object must be a readable "
                             "file-like, a file descriptor, or a string-like.")
         self.source = source
         self.readiface, self.writeiface = os.pipe()
 
     def run(self):
         t = self.writeiface
         if self.bytes:
             os.write(t, self.bytes)
         else:
             s = self.source
             b = s.read(4096)
             while b:
                 os.write(t, b)
                 b = s.read(4096)
         os.close(t)
 
     @property
     def output(self):
         return self.readiface
 
 
 class InputStreamChunker(Thread):
     def __init__(self, source, target, buffer_size, chunk_size):
 
         super(InputStreamChunker, self).__init__()
 
         self.daemon = True  # die die die.
 
         self.source = source
         self.target = target
         self.chunk_count_max = int(buffer_size / chunk_size) + 1
         self.chunk_size = chunk_size
 
         self.data_added = Event()
         self.data_added.clear()
 
         self.keep_reading = Event()
         self.keep_reading.set()
 
         self.EOF = Event()
         self.EOF.clear()
 
         self.go = Event()
         self.go.set()
 
     def stop(self):
         self.go.clear()
         self.EOF.set()
         try:
             # this is not proper, but is done to force the reader thread let
             # go of the input because, if successful, .close() will send EOF
             # down the pipe.
             self.source.close()
         except:
             pass
 
     def run(self):
         s = self.source
         t = self.target
         cs = self.chunk_size
         ccm = self.chunk_count_max
         kr = self.keep_reading
         da = self.data_added
         go = self.go
+
+        try:
         b = s.read(cs)
+        except ValueError:
+            b = ''
 
         while b and go.is_set():
             if len(t) > ccm:
                 kr.clear()
                 kr.wait(2)
 #                # this only works on 2.7.x and up
 #                if not kr.wait(10):
 #                    raise Exception("Timed out while waiting for input to be read.")
                 # instead we'll use this
                 if len(t) > ccm + 3:
                     raise IOError("Timed out while waiting for input from subprocess.")
             t.append(b)
             da.set()
             b = s.read(cs)
         self.EOF.set()
         da.set()  # for cases when done but there was no input.
 
 
 class BufferedGenerator():
     '''
     Class behaves as a non-blocking, buffered pipe reader.
     Reads chunks of data (through a thread)
     from a blocking pipe, and attaches these to an array (Deque) of chunks.
     Reading is halted in the thread when max chunks is internally buffered.
     The .next() may operate in blocking or non-blocking fashion by yielding
     '' if no data is ready
     to be sent or by not returning until there is some data to send
     When we get EOF from underlying source pipe we raise the marker to raise
     StopIteration after the last chunk of data is yielded.
     '''
 
     def __init__(self, source, buffer_size=65536, chunk_size=4096,
                  starting_values=[], bottomless=False):
 
         if bottomless:
             maxlen = int(buffer_size / chunk_size)
         else:
             maxlen = None
 
         self.data = deque(starting_values, maxlen)
 
         self.worker = InputStreamChunker(source, self.data, buffer_size,
                                          chunk_size)
         if starting_values:
             self.worker.data_added.set()
         self.worker.start()
 
     ####################
     # Generator's methods
     ####################
 
     def __iter__(self):
         return self
 
     def next(self):
         while not len(self.data) and not self.worker.EOF.is_set():
             self.worker.data_added.clear()
             self.worker.data_added.wait(0.2)
         if len(self.data):
             self.worker.keep_reading.set()
             return _bytes(self.data.popleft())
         elif self.worker.EOF.is_set():
             raise StopIteration
 
     def throw(self, type, value=None, traceback=None):
         if not self.worker.EOF.is_set():
             raise type(value)
 
     def start(self):
         self.worker.start()
 
     def stop(self):
         self.worker.stop()
 
     def close(self):
         try:
             self.worker.stop()
             self.throw(GeneratorExit)
         except (GeneratorExit, StopIteration):
             pass
 
     def __del__(self):
         self.close()
 
     ####################
     # Threaded reader's infrastructure.
     ####################
     @property
     def input(self):
         return self.worker.w
 
     @property
     def data_added_event(self):
         return self.worker.data_added
 
     @property
     def data_added(self):
         return self.worker.data_added.is_set()
 
     @property
     def reading_paused(self):
         return not self.worker.keep_reading.is_set()
 
     @property
     def done_reading_event(self):
         '''
         Done_reding does not mean that the iterator's buffer is empty.
         Iterator might have done reading from underlying source, but the read
         chunks might still be available for serving through .next() method.
 
         @return An Event class instance.
         '''
         return self.worker.EOF
 
     @property
     def done_reading(self):
         '''
         Done_reding does not mean that the iterator's buffer is empty.
         Iterator might have done reading from underlying source, but the read
         chunks might still be available for serving through .next() method.
 
         @return An Bool value.
         '''
         return self.worker.EOF.is_set()
 
     @property
     def length(self):
         '''
         returns int.
 
         This is the lenght of the que of chunks, not the length of
         the combined contents in those chunks.
 
         __len__() cannot be meaningfully implemented because this
         reader is just flying throuh a bottomless pit content and
         can only know the lenght of what it already saw.
 
         If __len__() on WSGI server per PEP 3333 returns a value,
         the responce's length will be set to that. In order not to
         confuse WSGI PEP3333 servers, we will not implement __len__
         at all.
         '''
         return len(self.data)
 
     def prepend(self, x):
         self.data.appendleft(x)
 
     def append(self, x):
         self.data.append(x)
 
     def extend(self, o):
         self.data.extend(o)
 
     def __getitem__(self, i):
         return self.data[i]
 
 
 class SubprocessIOChunker(object):
     '''
     Processor class wrapping handling of subprocess IO.
 
     In a way, this is a "communicate()" replacement with a twist.
 
     - We are multithreaded. Writing in and reading out, err are all sep threads.
     - We support concurrent (in and out) stream processing.
     - The output is not a stream. It's a queue of read string (bytes, not unicode)
       chunks. The object behaves as an iterable. You can "for chunk in obj:" us.
     - We are non-blocking in more respects than communicate()
       (reading from subprocess out pauses when internal buffer is full, but
        does not block the parent calling code. On the flip side, reading from
        slow-yielding subprocess may block the iteration until data shows up. This
        does not block the parallel inpipe reading occurring parallel thread.)
 
     The purpose of the object is to allow us to wrap subprocess interactions into
     and interable that can be passed to a WSGI server as the application's return
     value. Because of stream-processing-ability, WSGI does not have to read ALL
     of the subprocess's output and buffer it, before handing it to WSGI server for
     HTTP response. Instead, the class initializer reads just a bit of the stream
     to figure out if error ocurred or likely to occur and if not, just hands the
     further iteration over subprocess output to the server for completion of HTTP
     response.
 
     The real or perceived subprocess error is trapped and raised as one of
     EnvironmentError family of exceptions
 
     Example usage:
     #    try:
     #        answer = SubprocessIOChunker(
     #            cmd,
     #            input,
     #            buffer_size = 65536,
     #            chunk_size = 4096
     #            )
     #    except (EnvironmentError) as e:
     #        print str(e)
     #        raise e
     #
     #    return answer
 
 
     '''
     def __init__(self, cmd, inputstream=None, buffer_size=65536,
                  chunk_size=4096, starting_values=[], **kwargs):
         '''
         Initializes SubprocessIOChunker
 
         :param cmd: A Subprocess.Popen style "cmd". Can be string or array of strings
         :param inputstream: (Default: None) A file-like, string, or file pointer.
         :param buffer_size: (Default: 65536) A size of total buffer per stream in bytes.
         :param chunk_size: (Default: 4096) A max size of a chunk. Actual chunk may be smaller.
         :param starting_values: (Default: []) An array of strings to put in front of output que.
         '''
 
         if inputstream:
             input_streamer = StreamFeeder(inputstream)
             input_streamer.start()
             inputstream = input_streamer.output
 
         if isinstance(cmd, (list, tuple)):
             cmd = ' '.join(cmd)
 
         _shell = kwargs.get('shell') or True
         kwargs['shell'] = _shell
         _p = subprocess.Popen(cmd,
             bufsize=-1,
             stdin=inputstream,
             stdout=subprocess.PIPE,
             stderr=subprocess.PIPE,
             **kwargs
             )
 
         bg_out = BufferedGenerator(_p.stdout, buffer_size, chunk_size, starting_values)
         bg_err = BufferedGenerator(_p.stderr, 16000, 1, bottomless=True)
 
         while not bg_out.done_reading and not bg_out.reading_paused and not bg_err.length:
             # doing this until we reach either end of file, or end of buffer.
             bg_out.data_added_event.wait(1)
             bg_out.data_added_event.clear()
 
         # at this point it's still ambiguous if we are done reading or just full buffer.
         # Either way, if error (returned by ended process, or implied based on
         # presence of stuff in stderr output) we error out.
         # Else, we are happy.
         _returncode = _p.poll()
         if _returncode or (_returncode == None and bg_err.length):
             try:
                 _p.terminate()
             except:
                 pass
             bg_out.stop()
             bg_err.stop()
             err = '%s' % ''.join(bg_err)
+            if err:
             raise EnvironmentError("Subprocess exited due to an error:\n" + err)
+            raise EnvironmentError("Subprocess exited with non 0 ret code:%s" % _returncode)
 
         self.process = _p
         self.output = bg_out
         self.error = bg_err
 
     def __iter__(self):
         return self
 
     def next(self):
         if self.process.poll():
             err = '%s' % ''.join(self.error)
             raise EnvironmentError("Subprocess exited due to an error:\n" + err)
         return self.output.next()
 
     def throw(self, type, value=None, traceback=None):
         if self.output.length or not self.output.done_reading:
             raise type(value)
 
     def close(self):
         try:
             self.process.terminate()
         except:
             pass
         try:
             self.output.close()
         except:
             pass
         try:
             self.error.close()
         except:
             pass
 
     def __del__(self):
         self.close()

rhodecode/lib/utils.py

@@ -1,800 +1,801 b''
 # -*- coding: utf-8 -*-
 """
     rhodecode.lib.utils
     ~~~~~~~~~~~~~~~~~~~
 
     Utilities library for RhodeCode
 
     :created_on: Apr 18, 2010
     :author: marcink
     :copyright: (C) 2010-2012 Marcin Kuzminski <[e-mail]>
     :license: GPLv3, see COPYING for more details.
 """
 # This program is free software: you can redistribute it and/or modify
 # it under the terms of the GNU General Public License as published by
 # the Free Software Foundation, either version 3 of the License, or
 # (at your option) any later version.
 #
 # This program is distributed in the hope that it will be useful,
 # but WITHOUT ANY WARRANTY; without even the implied warranty of
 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 # GNU General Public License for more details.
 #
 # You should have received a copy of the GNU General Public License
 # along with this program.  If not, see <http://www.gnu.org/licenses/>.
 
 import os
 import re
 import logging
 import datetime
 import traceback
 import paste
 import beaker
 import tarfile
 import shutil
 import decorator
 import warnings
 from os.path import abspath
 from os.path import dirname as dn, join as jn
 
 from paste.script.command import Command, BadCommand
 
 from mercurial import ui, config
 
 from webhelpers.text import collapse, remove_formatting, strip_tags
 
 from rhodecode.lib.vcs import get_backend
 from rhodecode.lib.vcs.backends.base import BaseChangeset
 from rhodecode.lib.vcs.utils.lazy import LazyProperty
 from rhodecode.lib.vcs.utils.helpers import get_scm
 from rhodecode.lib.vcs.exceptions import VCSError
 
 from rhodecode.lib.caching_query import FromCache
 
 from rhodecode.model import meta
 from rhodecode.model.db import Repository, User, RhodeCodeUi, \
     UserLog, RepoGroup, RhodeCodeSetting, CacheInvalidation
 from rhodecode.model.meta import Session
 from rhodecode.model.repos_group import ReposGroupModel
 from rhodecode.lib.utils2 import safe_str, safe_unicode
 from rhodecode.lib.vcs.utils.fakemod import create_module
 
 log = logging.getLogger(__name__)
 
 REMOVED_REPO_PAT = re.compile(r'rm__\d{8}_\d{6}_\d{6}__.*')
 
 
 def recursive_replace(str_, replace=' '):
     """
     Recursive replace of given sign to just one instance
 
     :param str_: given string
     :param replace: char to find and replace multiple instances
 
     Examples::
     >>> recursive_replace("Mighty---Mighty-Bo--sstones",'-')
     'Mighty-Mighty-Bo-sstones'
     """
 
     if str_.find(replace * 2) == -1:
         return str_
     else:
         str_ = str_.replace(replace * 2, replace)
         return recursive_replace(str_, replace)
 
 
 def repo_name_slug(value):
     """
     Return slug of name of repository
     This function is called on each creation/modification
     of repository to prevent bad names in repo
     """
 
     slug = remove_formatting(value)
     slug = strip_tags(slug)
 
     for c in """`?=[]\;'"<>,/~!@#$%^&*()+{}|: """:
         slug = slug.replace(c, '-')
     slug = recursive_replace(slug, '-')
     slug = collapse(slug, '-')
     return slug
 
 
 def get_repo_slug(request):
     _repo = request.environ['pylons.routes_dict'].get('repo_name')
     if _repo:
         _repo = _repo.rstrip('/')
     return _repo
 
 
 def get_repos_group_slug(request):
     _group = request.environ['pylons.routes_dict'].get('group_name')
     if _group:
         _group = _group.rstrip('/')
     return _group
 
 
 def action_logger(user, action, repo, ipaddr='', sa=None, commit=False):
     """
     Action logger for various actions made by users
 
     :param user: user that made this action, can be a unique username string or
         object containing user_id attribute
     :param action: action to log, should be on of predefined unique actions for
         easy translations
     :param repo: string name of repository or object containing repo_id,
         that action was made on
     :param ipaddr: optional ip address from what the action was made
     :param sa: optional sqlalchemy session
 
     """
 
     if not sa:
         sa = meta.Session()
 
     try:
         if hasattr(user, 'user_id'):
             user_obj = User.get(user.user_id)
         elif isinstance(user, basestring):
             user_obj = User.get_by_username(user)
         else:
             raise Exception('You have to provide a user object or a username')
 
         if hasattr(repo, 'repo_id'):
             repo_obj = Repository.get(repo.repo_id)
             repo_name = repo_obj.repo_name
         elif  isinstance(repo, basestring):
             repo_name = repo.lstrip('/')
             repo_obj = Repository.get_by_repo_name(repo_name)
         else:
             repo_obj = None
             repo_name = ''
 
         user_log = UserLog()
         user_log.user_id = user_obj.user_id
         user_log.username = user_obj.username
         user_log.action = safe_unicode(action)
 
         user_log.repository = repo_obj
         user_log.repository_name = repo_name
 
         user_log.action_date = datetime.datetime.now()
         user_log.user_ip = ipaddr
         sa.add(user_log)
 
         log.info('Logging action %s on %s by %s' %
                  (action, safe_unicode(repo), user_obj))
         if commit:
             sa.commit()
     except:
         log.error(traceback.format_exc())
         raise
 
 
 def get_repos(path, recursive=False, skip_removed_repos=True):
     """
     Scans given path for repos and return (name,(type,path)) tuple
 
     :param path: path to scan for repositories
     :param recursive: recursive search and return names with subdirs in front
     """
 
     # remove ending slash for better results
     path = path.rstrip(os.sep)
     log.debug('now scanning in %s location recursive:%s...' % (path, recursive))
 
     def _get_repos(p):
         if not os.access(p, os.W_OK):
             return
         for dirpath in os.listdir(p):
             if os.path.isfile(os.path.join(p, dirpath)):
                 continue
             cur_path = os.path.join(p, dirpath)
 
             # skip removed repos
             if skip_removed_repos and REMOVED_REPO_PAT.match(dirpath):
                 continue
 
             #skip .<somethin> dirs
             if dirpath.startswith('.'):
                 continue
 
             try:
                 scm_info = get_scm(cur_path)
                 yield scm_info[1].split(path, 1)[-1].lstrip(os.sep), scm_info
             except VCSError:
                 if not recursive:
                     continue
                 #check if this dir containts other repos for recursive scan
                 rec_path = os.path.join(p, dirpath)
                 if os.path.isdir(rec_path):
                     for inner_scm in _get_repos(rec_path):
                         yield inner_scm
 
     return _get_repos(path)
 
 #alias for backward compat
 get_filesystem_repos = get_repos
 
 
 def is_valid_repo(repo_name, base_path, scm=None):
     """
     Returns True if given path is a valid repository False otherwise.
     If scm param is given also compare if given scm is the same as expected
     from scm parameter
 
     :param repo_name:
     :param base_path:
     :param scm:
 
     :return True: if given path is a valid repository
     """
     full_path = os.path.join(safe_str(base_path), safe_str(repo_name))
 
     try:
         scm_ = get_scm(full_path)
         if scm:
             return scm_[0] == scm
         return True
     except VCSError:
         return False
 
 
 def is_valid_repos_group(repos_group_name, base_path):
     """
     Returns True if given path is a repos group False otherwise
 
     :param repo_name:
     :param base_path:
     """
     full_path = os.path.join(safe_str(base_path), safe_str(repos_group_name))
 
     # check if it's not a repo
     if is_valid_repo(repos_group_name, base_path):
         return False
 
     try:
         # we need to check bare git repos at higher level
         # since we might match branches/hooks/info/objects or possible
         # other things inside bare git repo
         get_scm(os.path.dirname(full_path))
         return False
     except VCSError:
         pass
 
     # check if it's a valid path
     if os.path.isdir(full_path):
         return True
 
     return False
 
 
 def ask_ok(prompt, retries=4, complaint='Yes or no please!'):
     while True:
         ok = raw_input(prompt)
         if ok in ('y', 'ye', 'yes'):
             return True
         if ok in ('n', 'no', 'nop', 'nope'):
             return False
         retries = retries - 1
         if retries < 0:
             raise IOError
         print complaint
 
 #propagated from mercurial documentation
 ui_sections = ['alias', 'auth',
                 'decode/encode', 'defaults',
                 'diff', 'email',
                 'extensions', 'format',
                 'merge-patterns', 'merge-tools',
                 'hooks', 'http_proxy',
                 'smtp', 'patch',
                 'paths', 'profiling',
                 'server', 'trusted',
                 'ui', 'web', ]
 
 
 def make_ui(read_from='file', path=None, checkpaths=True, clear_session=True):
     """
     A function that will read python rc files or database
     and make an mercurial ui object from read options
 
     :param path: path to mercurial config file
     :param checkpaths: check the path
     :param read_from: read from 'file' or 'db'
     """
 
     baseui = ui.ui()
 
     # clean the baseui object
     baseui._ocfg = config.config()
     baseui._ucfg = config.config()
     baseui._tcfg = config.config()
 
     if read_from == 'file':
         if not os.path.isfile(path):
             log.debug('hgrc file is not present at %s, skipping...' % path)
             return False
         log.debug('reading hgrc from %s' % path)
         cfg = config.config()
         cfg.read(path)
         for section in ui_sections:
             for k, v in cfg.items(section):
                 log.debug('settings ui from file: [%s] %s=%s' % (section, k, v))
                 baseui.setconfig(safe_str(section), safe_str(k), safe_str(v))
 
     elif read_from == 'db':
         sa = meta.Session()
         ret = sa.query(RhodeCodeUi)\
             .options(FromCache("sql_cache_short", "get_hg_ui_settings"))\
             .all()
 
         hg_ui = ret
         for ui_ in hg_ui:
             if ui_.ui_active:
                 log.debug('settings ui from db: [%s] %s=%s', ui_.ui_section,
                           ui_.ui_key, ui_.ui_value)
                 baseui.setconfig(safe_str(ui_.ui_section), safe_str(ui_.ui_key),
                                  safe_str(ui_.ui_value))
             if ui_.ui_key == 'push_ssl':
                 # force set push_ssl requirement to False, rhodecode
                 # handles that
                 baseui.setconfig(safe_str(ui_.ui_section), safe_str(ui_.ui_key),
                                  False)
         if clear_session:
             meta.Session.remove()
     return baseui
 
 
 def set_rhodecode_config(config):
     """
     Updates pylons config with new settings from database
 
     :param config:
     """
     hgsettings = RhodeCodeSetting.get_app_settings()
 
     for k, v in hgsettings.items():
         config[k] = v
 
 
 def invalidate_cache(cache_key, *args):
     """
     Puts cache invalidation task into db for
     further global cache invalidation
     """
 
     from rhodecode.model.scm import ScmModel
 
     if cache_key.startswith('get_repo_cached_'):
         name = cache_key.split('get_repo_cached_')[-1]
         ScmModel().mark_for_invalidation(name)
 
 
 def map_groups(path):
     """
     Given a full path to a repository, create all nested groups that this
     repo is inside. This function creates parent-child relationships between
     groups and creates default perms for all new groups.
 
     :param paths: full path to repository
     """
     sa = meta.Session()
     groups = path.split(Repository.url_sep())
     parent = None
     group = None
 
     # last element is repo in nested groups structure
     groups = groups[:-1]
     rgm = ReposGroupModel(sa)
     for lvl, group_name in enumerate(groups):
         group_name = '/'.join(groups[:lvl] + [group_name])
         group = RepoGroup.get_by_group_name(group_name)
         desc = '%s group' % group_name
 
         # skip folders that are now removed repos
         if REMOVED_REPO_PAT.match(group_name):
             break
 
         if group is None:
             log.debug('creating group level: %s group_name: %s' % (lvl,
                                                                    group_name))
             group = RepoGroup(group_name, parent)
             group.group_description = desc
             sa.add(group)
             rgm._create_default_perms(group)
             sa.flush()
         parent = group
     return group
 
 
 def repo2db_mapper(initial_repo_list, remove_obsolete=False,
                    install_git_hook=False):
     """
     maps all repos given in initial_repo_list, non existing repositories
     are created, if remove_obsolete is True it also check for db entries
     that are not in initial_repo_list and removes them.
 
     :param initial_repo_list: list of repositories found by scanning methods
     :param remove_obsolete: check for obsolete entries in database
     :param install_git_hook: if this is True, also check and install githook
         for a repo if missing
     """
     from rhodecode.model.repo import RepoModel
     from rhodecode.model.scm import ScmModel
     sa = meta.Session()
     rm = RepoModel()
     user = sa.query(User).filter(User.admin == True).first()
     if user is None:
         raise Exception('Missing administrative account!')
     added = []
 
 #    # clear cache keys
 #    log.debug("Clearing cache keys now...")
 #    CacheInvalidation.clear_cache()
 #    sa.commit()
 
     ##creation defaults
     defs = RhodeCodeSetting.get_default_repo_settings(strip_prefix=True)
     enable_statistics = defs.get('repo_enable_statistics')
     enable_locking = defs.get('repo_enable_locking')
     enable_downloads = defs.get('repo_enable_downloads')
     private = defs.get('repo_private')
 
     for name, repo in initial_repo_list.items():
         group = map_groups(name)
         db_repo = rm.get_by_repo_name(name)
         # found repo that is on filesystem not in RhodeCode database
         if not db_repo:
             log.info('repository %s not found, creating now' % name)
             added.append(name)
             desc = (repo.description
                     if repo.description != 'unknown'
                     else '%s repository' % name)
 
             new_repo = rm.create_repo(
                 repo_name=name,
                 repo_type=repo.alias,
                 description=desc,
                 repos_group=getattr(group, 'group_id', None),
                 owner=user,
                 just_db=True,
                 enable_locking=enable_locking,
                 enable_downloads=enable_downloads,
                 enable_statistics=enable_statistics,
                 private=private
             )
             # we added that repo just now, and make sure it has githook
             # installed
             if new_repo.repo_type == 'git':
                 ScmModel().install_git_hook(new_repo.scm_instance)
             new_repo.update_changeset_cache()
         elif install_git_hook:
             if db_repo.repo_type == 'git':
                 ScmModel().install_git_hook(db_repo.scm_instance)
         # during starting install all cache keys for all repositories in the
         # system, this will register all repos and multiple instances
         key, _prefix, _org_key = CacheInvalidation._get_key(name)
         CacheInvalidation.invalidate(name)
         log.debug("Creating a cache key for %s, instance_id %s"
                   % (name, _prefix or 'unknown'))
 
     sa.commit()
     removed = []
     if remove_obsolete:
         # remove from database those repositories that are not in the filesystem
         for repo in sa.query(Repository).all():
             if repo.repo_name not in initial_repo_list.keys():
                 log.debug("Removing non-existing repository found in db `%s`" %
                           repo.repo_name)
                 try:
                     sa.delete(repo)
                     sa.commit()
                     removed.append(repo.repo_name)
                 except:
                     #don't hold further removals on error
                     log.error(traceback.format_exc())
                     sa.rollback()
 
     return added, removed
 
 
 # set cache regions for beaker so celery can utilise it
 def add_cache(settings):
     cache_settings = {'regions': None}
     for key in settings.keys():
         for prefix in ['beaker.cache.', 'cache.']:
             if key.startswith(prefix):
                 name = key.split(prefix)[1].strip()
                 cache_settings[name] = settings[key].strip()
     if cache_settings['regions']:
         for region in cache_settings['regions'].split(','):
             region = region.strip()
             region_settings = {}
             for key, value in cache_settings.items():
                 if key.startswith(region):
                     region_settings[key.split('.')[1]] = value
             region_settings['expire'] = int(region_settings.get('expire',
                                                                 60))
             region_settings.setdefault('lock_dir',
                                        cache_settings.get('lock_dir'))
             region_settings.setdefault('data_dir',
                                        cache_settings.get('data_dir'))
 
             if 'type' not in region_settings:
                 region_settings['type'] = cache_settings.get('type',
                                                              'memory')
             beaker.cache.cache_regions[region] = region_settings
 
 
 def load_rcextensions(root_path):
     import rhodecode
     from rhodecode.config import conf
 
     path = os.path.join(root_path, 'rcextensions', '__init__.py')
     if os.path.isfile(path):
         rcext = create_module('rc', path)
         EXT = rhodecode.EXTENSIONS = rcext
         log.debug('Found rcextensions now loading %s...' % rcext)
 
         # Additional mappings that are not present in the pygments lexers
         conf.LANGUAGES_EXTENSIONS_MAP.update(getattr(EXT, 'EXTRA_MAPPINGS', {}))
 
         #OVERRIDE OUR EXTENSIONS FROM RC-EXTENSIONS (if present)
 
         if getattr(EXT, 'INDEX_EXTENSIONS', []) != []:
             log.debug('settings custom INDEX_EXTENSIONS')
             conf.INDEX_EXTENSIONS = getattr(EXT, 'INDEX_EXTENSIONS', [])
 
         #ADDITIONAL MAPPINGS
         log.debug('adding extra into INDEX_EXTENSIONS')
         conf.INDEX_EXTENSIONS.extend(getattr(EXT, 'EXTRA_INDEX_EXTENSIONS', []))
 
         # auto check if the module is not missing any data, set to default if is
         # this will help autoupdate new feature of rcext module
         from rhodecode.config import rcextensions
         for k in dir(rcextensions):
             if not k.startswith('_') and not hasattr(EXT, k):
                 setattr(EXT, k, getattr(rcextensions, k))
 
 
 def get_custom_lexer(extension):
     """
     returns a custom lexer if it's defined in rcextensions module, or None
     if there's no custom lexer defined
     """
     import rhodecode
     from pygments import lexers
     #check if we didn't define this extension as other lexer
     if rhodecode.EXTENSIONS and extension in rhodecode.EXTENSIONS.EXTRA_LEXERS:
         _lexer_name = rhodecode.EXTENSIONS.EXTRA_LEXERS[extension]
         return lexers.get_lexer_by_name(_lexer_name)
 
 
 #==============================================================================
 # TEST FUNCTIONS AND CREATORS
 #==============================================================================
 def create_test_index(repo_location, config, full_index):
     """
     Makes default test index
 
     :param config: test config
     :param full_index:
     """
 
     from rhodecode.lib.indexers.daemon import WhooshIndexingDaemon
     from rhodecode.lib.pidlock import DaemonLock, LockHeld
 
     repo_location = repo_location
 
     index_location = os.path.join(config['app_conf']['index_dir'])
     if not os.path.exists(index_location):
         os.makedirs(index_location)
 
     try:
         l = DaemonLock(file_=jn(dn(index_location), 'make_index.lock'))
         WhooshIndexingDaemon(index_location=index_location,
                              repo_location=repo_location)\
             .run(full_index=full_index)
         l.release()
     except LockHeld:
         pass
 
 
 def create_test_env(repos_test_path, config):
     """
     Makes a fresh database and
     install test repository into tmp dir
     """
     from rhodecode.lib.db_manage import DbManage
     from rhodecode.tests import HG_REPO, GIT_REPO, TESTS_TMP_PATH
 
     # PART ONE create db
     dbconf = config['sqlalchemy.db1.url']
     log.debug('making test db %s' % dbconf)
 
     # create test dir if it doesn't exist
     if not os.path.isdir(repos_test_path):
         log.debug('Creating testdir %s' % repos_test_path)
         os.makedirs(repos_test_path)
 
     dbmanage = DbManage(log_sql=True, dbconf=dbconf, root=config['here'],
                         tests=True)
     dbmanage.create_tables(override=True)
     dbmanage.create_settings(dbmanage.config_prompt(repos_test_path))
     dbmanage.create_default_user()
     dbmanage.admin_prompt()
     dbmanage.create_permissions()
     dbmanage.populate_default_permissions()
     Session().commit()
     # PART TWO make test repo
     log.debug('making test vcs repositories')
 
     idx_path = config['app_conf']['index_dir']
     data_path = config['app_conf']['cache_dir']
 
     #clean index and data
     if idx_path and os.path.exists(idx_path):
         log.debug('remove %s' % idx_path)
         shutil.rmtree(idx_path)
 
     if data_path and os.path.exists(data_path):
         log.debug('remove %s' % data_path)
         shutil.rmtree(data_path)
 
     #CREATE DEFAULT TEST REPOS
     cur_dir = dn(dn(abspath(__file__)))
     tar = tarfile.open(jn(cur_dir, 'tests', "vcs_test_hg.tar.gz"))
     tar.extractall(jn(TESTS_TMP_PATH, HG_REPO))
     tar.close()
 
     cur_dir = dn(dn(abspath(__file__)))
     tar = tarfile.open(jn(cur_dir, 'tests', "vcs_test_git.tar.gz"))
     tar.extractall(jn(TESTS_TMP_PATH, GIT_REPO))
     tar.close()
 
     #LOAD VCS test stuff
     from rhodecode.tests.vcs import setup_package
     setup_package()
 
 
 #==============================================================================
 # PASTER COMMANDS
 #==============================================================================
 class BasePasterCommand(Command):
     """
     Abstract Base Class for paster commands.
 
     The celery commands are somewhat aggressive about loading
     celery.conf, and since our module sets the `CELERY_LOADER`
     environment variable to our loader, we have to bootstrap a bit and
     make sure we've had a chance to load the pylons config off of the
     command line, otherwise everything fails.
     """
     min_args = 1
     min_args_error = "Please provide a paster config file as an argument."
     takes_config_file = 1
     requires_config_file = True
 
     def notify_msg(self, msg, log=False):
         """Make a notification to user, additionally if logger is passed
         it logs this action using given logger
 
         :param msg: message that will be printed to user
         :param log: logging instance, to use to additionally log this message
 
         """
         if log and isinstance(log, logging):
             log(msg)
 
     def run(self, args):
         """
         Overrides Command.run
 
         Checks for a config file argument and loads it.
         """
         if len(args) < self.min_args:
             raise BadCommand(
                 self.min_args_error % {'min_args': self.min_args,
                                        'actual_args': len(args)})
 
         # Decrement because we're going to lob off the first argument.
         # @@ This is hacky
         self.min_args -= 1
         self.bootstrap_config(args[0])
         self.update_parser()
         return super(BasePasterCommand, self).run(args[1:])
 
     def update_parser(self):
         """
         Abstract method.  Allows for the class's parser to be updated
         before the superclass's `run` method is called.  Necessary to
         allow options/arguments to be passed through to the underlying
         celery command.
         """
         raise NotImplementedError("Abstract Method.")
 
     def bootstrap_config(self, conf):
         """
         Loads the pylons configuration.
         """
         from pylons import config as pylonsconfig
 
         self.path_to_ini_file = os.path.realpath(conf)
         conf = paste.deploy.appconfig('config:' + self.path_to_ini_file)
         pylonsconfig.init_app(conf.global_conf, conf.local_conf)
 
     def _init_session(self):
         """
         Inits SqlAlchemy Session
         """
         logging.config.fileConfig(self.path_to_ini_file)
         from pylons import config
         from rhodecode.model import init_model
         from rhodecode.lib.utils2 import engine_from_config
 
         #get to remove repos !!
         add_cache(config)
         engine = engine_from_config(config, 'sqlalchemy.db1.')
         init_model(engine)
 
 
 def check_git_version():
     """
     Checks what version of git is installed in system, and issues a warning
     if it's too old for RhodeCode to properly work.
     """
     from rhodecode import BACKENDS
     from rhodecode.lib.vcs.backends.git.repository import GitRepository
     from distutils.version import StrictVersion
 
-    stdout, stderr = GitRepository._run_git_command('--version')
+    stdout, stderr = GitRepository._run_git_command('--version', _bare=True,
+                                                    _safe=True)
 
     ver = (stdout.split(' ')[-1] or '').strip() or '0.0.0'
     if len(ver.split('.')) > 3:
         #StrictVersion needs to be only 3 element type
         ver = '.'.join(ver.split('.')[:3])
     try:
         _ver = StrictVersion(ver)
     except:
         _ver = StrictVersion('0.0.0')
         stderr = traceback.format_exc()
 
     req_ver = '1.7.4'
     to_old_git = False
     if  _ver < StrictVersion(req_ver):
         to_old_git = True
 
     if 'git' in BACKENDS:
         log.debug('GIT version detected: %s' % stdout)
         if stderr:
             log.warning('Unable to detect git version org error was:%r' % stderr)
         elif to_old_git:
             log.warning('RhodeCode detected git version %s, which is too old '
                         'for the system to function properly. Make sure '
                         'its version is at least %s' % (ver, req_ver))
     return _ver
 
 
 @decorator.decorator
 def jsonify(func, *args, **kwargs):
     """Action decorator that formats output for JSON
 
     Given a function that will return content, this decorator will turn
     the result into JSON, with a content-type of 'application/json' and
     output it.
 
     """
     from pylons.decorators.util import get_pylons
     from rhodecode.lib.ext_json import json
     pylons = get_pylons(args)
     pylons.response.headers['Content-Type'] = 'application/json; charset=utf-8'
     data = func(*args, **kwargs)
     if isinstance(data, (list, tuple)):
         msg = "JSON responses with Array envelopes are susceptible to " \
               "cross-site data leak attacks, see " \
               "http://wiki.pylonshq.com/display/pylonsfaq/Warnings"
         warnings.warn(msg, Warning, 2)
         log.warning(msg)
     log.debug("Returning JSON wrapped action output")
     return json.dumps(data, encoding='utf-8')

rhodecode/lib/vcs/backends/git/repository.py

@@ -1,673 +1,687 b''
 # -*- coding: utf-8 -*-
 """
     vcs.backends.git
     ~~~~~~~~~~~~~~~~
 
     Git backend implementation.
 
     :created_on: Apr 8, 2010
     :copyright: (c) 2010-2011 by Marcin Kuzminski, Lukasz Balcerzak.
 """
 
 import os
 import re
 import time
 import posixpath
 import logging
 import traceback
 import urllib
 import urllib2
 from dulwich.repo import Repo, NotGitRepository
 from dulwich.objects import Tag
 from string import Template
 
 import rhodecode
 from rhodecode.lib.vcs.backends.base import BaseRepository
 from rhodecode.lib.vcs.exceptions import BranchDoesNotExistError
 from rhodecode.lib.vcs.exceptions import ChangesetDoesNotExistError
 from rhodecode.lib.vcs.exceptions import EmptyRepositoryError
 from rhodecode.lib.vcs.exceptions import RepositoryError
 from rhodecode.lib.vcs.exceptions import TagAlreadyExistError
 from rhodecode.lib.vcs.exceptions import TagDoesNotExistError
 from rhodecode.lib.vcs.utils import safe_unicode, makedate, date_fromtimestamp
 from rhodecode.lib.vcs.utils.lazy import LazyProperty, ThreadLocalLazyProperty
 from rhodecode.lib.vcs.utils.ordered_dict import OrderedDict
 from rhodecode.lib.vcs.utils.paths import abspath
 from rhodecode.lib.vcs.utils.paths import get_user_home
 from .workdir import GitWorkdir
 from .changeset import GitChangeset
 from .inmemory import GitInMemoryChangeset
 from .config import ConfigFile
 from rhodecode.lib import subprocessio
 
 
 log = logging.getLogger(__name__)
 
 
 class GitRepository(BaseRepository):
     """
     Git repository backend.
     """
     DEFAULT_BRANCH_NAME = 'master'
     scm = 'git'
 
     def __init__(self, repo_path, create=False, src_url=None,
                  update_after_clone=False, bare=False):
 
         self.path = abspath(repo_path)
         repo = self._get_repo(create, src_url, update_after_clone, bare)
         self.bare = repo.bare
 
         self._config_files = [
             bare and abspath(self.path, 'config')
                      or abspath(self.path, '.git', 'config'),
             abspath(get_user_home(), '.gitconfig'),
         ]
 
     @ThreadLocalLazyProperty
     def _repo(self):
         repo = Repo(self.path)
         #temporary set that to now at later we will move it to constructor
         baseui = None
         if baseui is None:
             from mercurial.ui import ui
             baseui = ui()
         # patch the instance of GitRepo with an "FAKE" ui object to add
         # compatibility layer with Mercurial
         setattr(repo, 'ui', baseui)
         return repo
 
     @property
     def head(self):
         try:
             return self._repo.head()
         except KeyError:
             return None
 
     @LazyProperty
     def revisions(self):
         """
         Returns list of revisions' ids, in ascending order.  Being lazy
         attribute allows external tools to inject shas from cache.
         """
         return self._get_all_revisions()
 
     @classmethod
     def _run_git_command(cls, cmd, **opts):
         """
         Runs given ``cmd`` as git command and returns tuple
         (stdout, stderr).
 
         :param cmd: git command to be executed
         :param opts: env options to pass into Subprocess command
         """
 
+        if '_bare' in opts:
+            _copts = []
+            del opts['_bare']
+        else:
         _copts = ['-c', 'core.quotepath=false', ]
+        safe_call = False
+        if '_safe' in opts:
+            #no exc on failure
+            del opts['_safe']
+            safe_call = True
+
         _str_cmd = False
         if isinstance(cmd, basestring):
             cmd = [cmd]
             _str_cmd = True
 
         gitenv = os.environ
         # need to clean fix GIT_DIR !
         if 'GIT_DIR' in gitenv:
             del gitenv['GIT_DIR']
         gitenv['GIT_CONFIG_NOGLOBAL'] = '1'
 
         _git_path = rhodecode.CONFIG.get('git_path', 'git')
         cmd = [_git_path] + _copts + cmd
         if _str_cmd:
             cmd = ' '.join(cmd)
         try:
             _opts = dict(
                 env=gitenv,
                 shell=False,
             )
             _opts.update(opts)
             p = subprocessio.SubprocessIOChunker(cmd, **_opts)
         except (EnvironmentError, OSError), err:
-            log.error(traceback.format_exc())
-            raise RepositoryError("Couldn't run git command (%s).\n"
-                                  "Original error was:%s" % (cmd, err))
+            tb_err = ("Couldn't run git command (%s).\n"
+                      "Original error was:%s\n" % (cmd, err))
+            log.error(tb_err)
+            if safe_call:
+                return '', err
+            else:
+                raise RepositoryError(tb_err)
 
         return ''.join(p.output), ''.join(p.error)
 
     def run_git_command(self, cmd):
         opts = {}
         if os.path.isdir(self.path):
             opts['cwd'] = self.path
         return self._run_git_command(cmd, **opts)
 
     @classmethod
     def _check_url(cls, url):
         """
         Functon will check given url and try to verify if it's a valid
         link. Sometimes it may happened that mercurial will issue basic
         auth request that can cause whole API to hang when used from python
         or other external calls.
 
         On failures it'll raise urllib2.HTTPError
         """
         from mercurial.util import url as Url
 
         # those authnadlers are patched for python 2.6.5 bug an
         # infinit looping when given invalid resources
         from mercurial.url import httpbasicauthhandler, httpdigestauthhandler
 
         # check first if it's not an local url
         if os.path.isdir(url) or url.startswith('file:'):
             return True
 
         if('+' in url[:url.find('://')]):
             url = url[url.find('+') + 1:]
 
         handlers = []
         test_uri, authinfo = Url(url).authinfo()
         if not test_uri.endswith('info/refs'):
             test_uri = test_uri.rstrip('/') + '/info/refs'
         if authinfo:
             #create a password manager
             passmgr = urllib2.HTTPPasswordMgrWithDefaultRealm()
             passmgr.add_password(*authinfo)
 
             handlers.extend((httpbasicauthhandler(passmgr),
                              httpdigestauthhandler(passmgr)))
 
         o = urllib2.build_opener(*handlers)
         o.addheaders = [('User-Agent', 'git/1.7.8.0')]  # fake some git
 
         q = {"service": 'git-upload-pack'}
         qs = '?%s' % urllib.urlencode(q)
         cu = "%s%s" % (test_uri, qs)
         req = urllib2.Request(cu, None, {})
 
         try:
             resp = o.open(req)
             return resp.code == 200
         except Exception, e:
             # means it cannot be cloned
             raise urllib2.URLError("[%s] %s" % (url, e))
 
     def _get_repo(self, create, src_url=None, update_after_clone=False,
                   bare=False):
         if create and os.path.exists(self.path):
             raise RepositoryError("Location already exist")
         if src_url and not create:
             raise RepositoryError("Create should be set to True if src_url is "
                                   "given (clone operation creates repository)")
         try:
             if create and src_url:
                 GitRepository._check_url(src_url)
                 self.clone(src_url, update_after_clone, bare)
                 return Repo(self.path)
             elif create:
                 os.mkdir(self.path)
                 if bare:
                     return Repo.init_bare(self.path)
                 else:
                     return Repo.init(self.path)
             else:
                 return self._repo
         except (NotGitRepository, OSError), err:
             raise RepositoryError(err)
 
     def _get_all_revisions(self):
         # we must check if this repo is not empty, since later command
         # fails if it is. And it's cheaper to ask than throw the subprocess
         # errors
         try:
             self._repo.head()
         except KeyError:
             return []
         cmd = 'rev-list --all --reverse --date-order'
         try:
             so, se = self.run_git_command(cmd)
         except RepositoryError:
             # Can be raised for empty repositories
             return []
         return so.splitlines()
 
     def _get_all_revisions2(self):
         #alternate implementation using dulwich
         includes = [x[1][0] for x in self._parsed_refs.iteritems()
                     if x[1][1] != 'T']
         return [c.commit.id for c in self._repo.get_walker(include=includes)]
 
     def _get_revision(self, revision):
         """
         For git backend we always return integer here. This way we ensure
         that changset's revision attribute would become integer.
         """
         pattern = re.compile(r'^[[0-9a-fA-F]{12}|[0-9a-fA-F]{40}]$')
         is_bstr = lambda o: isinstance(o, (str, unicode))
         is_null = lambda o: len(o) == revision.count('0')
 
         if len(self.revisions) == 0:
             raise EmptyRepositoryError("There are no changesets yet")
 
         if revision in (None, '', 'tip', 'HEAD', 'head', -1):
             revision = self.revisions[-1]
 
         if ((is_bstr(revision) and revision.isdigit() and len(revision) < 12)
             or isinstance(revision, int) or is_null(revision)):
             try:
                 revision = self.revisions[int(revision)]
             except:
                 raise ChangesetDoesNotExistError("Revision %r does not exist "
                     "for this repository %s" % (revision, self))
 
         elif is_bstr(revision):
             # get by branch/tag name
             _ref_revision = self._parsed_refs.get(revision)
             _tags_shas = self.tags.values()
             if _ref_revision:  # and _ref_revision[1] in ['H', 'RH', 'T']:
                 return _ref_revision[0]
 
             # maybe it's a tag ? we don't have them in self.revisions
             elif revision in _tags_shas:
                 return _tags_shas[_tags_shas.index(revision)]
 
             elif not pattern.match(revision) or revision not in self.revisions:
                 raise ChangesetDoesNotExistError("Revision %r does not exist "
                     "for this repository %s" % (revision, self))
 
         # Ensure we return full id
         if not pattern.match(str(revision)):
             raise ChangesetDoesNotExistError("Given revision %r not recognized"
                 % revision)
         return revision
 
     def _get_archives(self, archive_name='tip'):
 
         for i in [('zip', '.zip'), ('gz', '.tar.gz'), ('bz2', '.tar.bz2')]:
                 yield {"type": i[0], "extension": i[1], "node": archive_name}
 
     def _get_url(self, url):
         """
         Returns normalized url. If schema is not given, would fall to
         filesystem (``file:///``) schema.
         """
         url = str(url)
         if url != 'default' and not '://' in url:
             url = ':///'.join(('file', url))
         return url
 
     @LazyProperty
     def name(self):
         return os.path.basename(self.path)
 
     @LazyProperty
     def last_change(self):
         """
         Returns last change made on this repository as datetime object
         """
         return date_fromtimestamp(self._get_mtime(), makedate()[1])
 
     def _get_mtime(self):
         try:
             return time.mktime(self.get_changeset().date.timetuple())
         except RepositoryError:
             idx_loc = '' if self.bare else '.git'
             # fallback to filesystem
             in_path = os.path.join(self.path, idx_loc, "index")
             he_path = os.path.join(self.path, idx_loc, "HEAD")
             if os.path.exists(in_path):
                 return os.stat(in_path).st_mtime
             else:
                 return os.stat(he_path).st_mtime
 
     @LazyProperty
     def description(self):
         idx_loc = '' if self.bare else '.git'
         undefined_description = u'unknown'
         description_path = os.path.join(self.path, idx_loc, 'description')
         if os.path.isfile(description_path):
             return safe_unicode(open(description_path).read())
         else:
             return undefined_description
 
     @LazyProperty
     def contact(self):
         undefined_contact = u'Unknown'
         return undefined_contact
 
     @property
     def branches(self):
         if not self.revisions:
             return {}
         sortkey = lambda ctx: ctx[0]
         _branches = [(x[0], x[1][0])
                      for x in self._parsed_refs.iteritems() if x[1][1] == 'H']
         return OrderedDict(sorted(_branches, key=sortkey, reverse=False))
 
     @LazyProperty
     def tags(self):
         return self._get_tags()
 
     def _get_tags(self):
         if not self.revisions:
             return {}
 
         sortkey = lambda ctx: ctx[0]
         _tags = [(x[0], x[1][0])
                  for x in self._parsed_refs.iteritems() if x[1][1] == 'T']
         return OrderedDict(sorted(_tags, key=sortkey, reverse=True))
 
     def tag(self, name, user, revision=None, message=None, date=None,
             **kwargs):
         """
         Creates and returns a tag for the given ``revision``.
 
         :param name: name for new tag
         :param user: full username, i.e.: "Joe Doe <[e-mail]>"
         :param revision: changeset id for which new tag would be created
         :param message: message of the tag's commit
         :param date: date of tag's commit
 
         :raises TagAlreadyExistError: if tag with same name already exists
         """
         if name in self.tags:
             raise TagAlreadyExistError("Tag %s already exists" % name)
         changeset = self.get_changeset(revision)
         message = message or "Added tag %s for commit %s" % (name,
             changeset.raw_id)
         self._repo.refs["refs/tags/%s" % name] = changeset._commit.id
 
         self._parsed_refs = self._get_parsed_refs()
         self.tags = self._get_tags()
         return changeset
 
     def remove_tag(self, name, user, message=None, date=None):
         """
         Removes tag with the given ``name``.
 
         :param name: name of the tag to be removed
         :param user: full username, i.e.: "Joe Doe <[e-mail]>"
         :param message: message of the tag's removal commit
         :param date: date of tag's removal commit
 
         :raises TagDoesNotExistError: if tag with given name does not exists
         """
         if name not in self.tags:
             raise TagDoesNotExistError("Tag %s does not exist" % name)
         tagpath = posixpath.join(self._repo.refs.path, 'refs', 'tags', name)
         try:
             os.remove(tagpath)
             self._parsed_refs = self._get_parsed_refs()
             self.tags = self._get_tags()
         except OSError, e:
             raise RepositoryError(e.strerror)
 
     @LazyProperty
     def _parsed_refs(self):
         return self._get_parsed_refs()
 
     def _get_parsed_refs(self):
         refs = self._repo.get_refs()
         keys = [('refs/heads/', 'H'),
                 ('refs/remotes/origin/', 'RH'),
                 ('refs/tags/', 'T')]
         _refs = {}
         for ref, sha in refs.iteritems():
             for k, type_ in keys:
                 if ref.startswith(k):
                     _key = ref[len(k):]
                     if type_ == 'T':
                         obj = self._repo.get_object(sha)
                         if isinstance(obj, Tag):
                             sha = self._repo.get_object(sha).object[1]
                     _refs[_key] = [sha, type_]
                     break
         return _refs
 
     def _heads(self, reverse=False):
         refs = self._repo.get_refs()
         heads = {}
 
         for key, val in refs.items():
             for ref_key in ['refs/heads/', 'refs/remotes/origin/']:
                 if key.startswith(ref_key):
                     n = key[len(ref_key):]
                     if n not in ['HEAD']:
                         heads[n] = val
 
         return heads if reverse else dict((y, x) for x, y in heads.iteritems())
 
     def get_changeset(self, revision=None):
         """
         Returns ``GitChangeset`` object representing commit from git repository
         at the given revision or head (most recent commit) if None given.
         """
         if isinstance(revision, GitChangeset):
             return revision
         revision = self._get_revision(revision)
         changeset = GitChangeset(repository=self, revision=revision)
         return changeset
 
     def get_changesets(self, start=None, end=None, start_date=None,
            end_date=None, branch_name=None, reverse=False):
         """
         Returns iterator of ``GitChangeset`` objects from start to end (both
         are inclusive), in ascending date order (unless ``reverse`` is set).
 
         :param start: changeset ID, as str; first returned changeset
         :param end: changeset ID, as str; last returned changeset
         :param start_date: if specified, changesets with commit date less than
           ``start_date`` would be filtered out from returned set
         :param end_date: if specified, changesets with commit date greater than
           ``end_date`` would be filtered out from returned set
         :param branch_name: if specified, changesets not reachable from given
           branch would be filtered out from returned set
         :param reverse: if ``True``, returned generator would be reversed
           (meaning that returned changesets would have descending date order)
 
         :raise BranchDoesNotExistError: If given ``branch_name`` does not
             exist.
         :raise ChangesetDoesNotExistError: If changeset for given ``start`` or
           ``end`` could not be found.
 
         """
         if branch_name and branch_name not in self.branches:
             raise BranchDoesNotExistError("Branch '%s' not found" \
                                           % branch_name)
         # %H at format means (full) commit hash, initial hashes are retrieved
         # in ascending date order
         cmd_template = 'log --date-order --reverse --pretty=format:"%H"'
         cmd_params = {}
         if start_date:
             cmd_template += ' --since "$since"'
             cmd_params['since'] = start_date.strftime('%m/%d/%y %H:%M:%S')
         if end_date:
             cmd_template += ' --until "$until"'
             cmd_params['until'] = end_date.strftime('%m/%d/%y %H:%M:%S')
         if branch_name:
             cmd_template += ' $branch_name'
             cmd_params['branch_name'] = branch_name
         else:
             cmd_template += ' --all'
 
         cmd = Template(cmd_template).safe_substitute(**cmd_params)
         revs = self.run_git_command(cmd)[0].splitlines()
         start_pos = 0
         end_pos = len(revs)
         if start:
             _start = self._get_revision(start)
             try:
                 start_pos = revs.index(_start)
             except ValueError:
                 pass
 
         if end is not None:
             _end = self._get_revision(end)
             try:
                 end_pos = revs.index(_end)
             except ValueError:
                 pass
 
         if None not in [start, end] and start_pos > end_pos:
             raise RepositoryError('start cannot be after end')
 
         if end_pos is not None:
             end_pos += 1
 
         revs = revs[start_pos:end_pos]
         if reverse:
             revs = reversed(revs)
         for rev in revs:
             yield self.get_changeset(rev)
 
     def get_diff(self, rev1, rev2, path=None, ignore_whitespace=False,
                  context=3):
         """
         Returns (git like) *diff*, as plain text. Shows changes introduced by
         ``rev2`` since ``rev1``.
 
         :param rev1: Entry point from which diff is shown. Can be
           ``self.EMPTY_CHANGESET`` - in this case, patch showing all
           the changes since empty state of the repository until ``rev2``
         :param rev2: Until which revision changes should be shown.
         :param ignore_whitespace: If set to ``True``, would not show whitespace
           changes. Defaults to ``False``.
         :param context: How many lines before/after changed lines should be
           shown. Defaults to ``3``.
         """
         flags = ['-U%s' % context, '--full-index', '--binary', '-p', '-M', '--abbrev=40']
         if ignore_whitespace:
             flags.append('-w')
 
         if hasattr(rev1, 'raw_id'):
             rev1 = getattr(rev1, 'raw_id')
 
         if hasattr(rev2, 'raw_id'):
             rev2 = getattr(rev2, 'raw_id')
 
         if rev1 == self.EMPTY_CHANGESET:
             rev2 = self.get_changeset(rev2).raw_id
             cmd = ' '.join(['show'] + flags + [rev2])
         else:
             rev1 = self.get_changeset(rev1).raw_id
             rev2 = self.get_changeset(rev2).raw_id
             cmd = ' '.join(['diff'] + flags + [rev1, rev2])
 
         if path:
             cmd += ' -- "%s"' % path
 
         stdout, stderr = self.run_git_command(cmd)
         # If we used 'show' command, strip first few lines (until actual diff
         # starts)
         if rev1 == self.EMPTY_CHANGESET:
             lines = stdout.splitlines()
             x = 0
             for line in lines:
                 if line.startswith('diff'):
                     break
                 x += 1
             # Append new line just like 'diff' command do
             stdout = '\n'.join(lines[x:]) + '\n'
         return stdout
 
     @LazyProperty
     def in_memory_changeset(self):
         """
         Returns ``GitInMemoryChangeset`` object for this repository.
         """
         return GitInMemoryChangeset(self)
 
     def clone(self, url, update_after_clone=True, bare=False):
         """
         Tries to clone changes from external location.
 
         :param update_after_clone: If set to ``False``, git won't checkout
           working directory
         :param bare: If set to ``True``, repository would be cloned into
           *bare* git repository (no working directory at all).
         """
         url = self._get_url(url)
         cmd = ['clone']
         if bare:
             cmd.append('--bare')
         elif not update_after_clone:
             cmd.append('--no-checkout')
         cmd += ['--', '"%s"' % url, '"%s"' % self.path]
         cmd = ' '.join(cmd)
         # If error occurs run_git_command raises RepositoryError already
         self.run_git_command(cmd)
 
     def pull(self, url):
         """
         Tries to pull changes from external location.
         """
         url = self._get_url(url)
         cmd = ['pull']
         cmd.append("--ff-only")
         cmd.append(url)
         cmd = ' '.join(cmd)
         # If error occurs run_git_command raises RepositoryError already
         self.run_git_command(cmd)
 
     def fetch(self, url):
         """
         Tries to pull changes from external location.
         """
         url = self._get_url(url)
         so, se = self.run_git_command('ls-remote -h %s' % url)
         refs = []
         for line in (x for x in so.splitlines()):
             sha, ref = line.split('\t')
             refs.append(ref)
         refs = ' '.join(('+%s:%s' % (r, r) for r in refs))
         cmd = '''fetch %s -- %s''' % (url, refs)
         self.run_git_command(cmd)
 
     @LazyProperty
     def workdir(self):
         """
         Returns ``Workdir`` instance for this repository.
         """
         return GitWorkdir(self)
 
     def get_config_value(self, section, name, config_file=None):
         """
         Returns configuration value for a given [``section``] and ``name``.
 
         :param section: Section we want to retrieve value from
         :param name: Name of configuration we want to retrieve
         :param config_file: A path to file which should be used to retrieve
           configuration from (might also be a list of file paths)
         """
         if config_file is None:
             config_file = []
         elif isinstance(config_file, basestring):
             config_file = [config_file]
 
         def gen_configs():
             for path in config_file + self._config_files:
                 try:
                     yield ConfigFile.from_path(path)
                 except (IOError, OSError, ValueError):
                     continue
 
         for config in gen_configs():
             try:
                 return config.get(section, name)
             except KeyError:
                 continue
         return None
 
     def get_user_name(self, config_file=None):
         """
         Returns user's name from global configuration file.
 
         :param config_file: A path to file which should be used to retrieve
           configuration from (might also be a list of file paths)
         """
         return self.get_config_value('user', 'name', config_file)
 
     def get_user_email(self, config_file=None):
         """
         Returns user's email from global configuration file.
 
         :param config_file: A path to file which should be used to retrieve
           configuration from (might also be a list of file paths)
         """
         return self.get_config_value('user', 'email', config_file)