vcs: Use more specific merge error reason (either target or source ref missing)

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

@@ -1,928 +1,928 b''
 # -*- coding: utf-8 -*-
 
 # Copyright (C) 2014-2016  RhodeCode GmbH
 #
 # This program is free software: you can redistribute it and/or modify
 # it under the terms of the GNU Affero General Public License, version 3
 # (only), as published by the Free Software Foundation.
 #
 # 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 Affero General Public License
 # along with this program.  If not, see <http://www.gnu.org/licenses/>.
 #
 # This program is dual-licensed. If you wish to learn more about the
 # RhodeCode Enterprise Edition, including its added features, Support services,
 # and proprietary license terms, please see https://rhodecode.com/licenses/
 
 """
 GIT repository module
 """
 
 import logging
 import os
 import re
 import shutil
 import time
 
 from zope.cachedescriptors.property import Lazy as LazyProperty
 
 from rhodecode.lib.compat import OrderedDict
 from rhodecode.lib.datelib import makedate, utcdate_fromtimestamp
 from rhodecode.lib.utils import safe_unicode, safe_str
 from rhodecode.lib.vcs import connection, path as vcspath
 from rhodecode.lib.vcs.backends.base import (
     BaseRepository, CollectionGenerator, Config, MergeResponse,
     MergeFailureReason, Reference)
 from rhodecode.lib.vcs.backends.git.commit import GitCommit
 from rhodecode.lib.vcs.backends.git.diff import GitDiff
 from rhodecode.lib.vcs.backends.git.inmemory import GitInMemoryCommit
 from rhodecode.lib.vcs.exceptions import (
     CommitDoesNotExistError, EmptyRepositoryError,
     RepositoryError, TagAlreadyExistError, TagDoesNotExistError, VCSError)
 
 
 SHA_PATTERN = re.compile(r'^[[0-9a-fA-F]{12}|[0-9a-fA-F]{40}]$')
 
 log = logging.getLogger(__name__)
 
 
 class GitRepository(BaseRepository):
     """
     Git repository backend.
     """
     DEFAULT_BRANCH_NAME = 'master'
 
     contact = BaseRepository.DEFAULT_CONTACT
 
     def __init__(self, repo_path, config=None, create=False, src_url=None,
                  update_after_clone=False, with_wire=None, bare=False):
 
         self.path = safe_str(os.path.abspath(repo_path))
         self.config = config if config else Config()
         self._remote = connection.Git(
             self.path, self.config, with_wire=with_wire)
 
         self._init_repo(create, src_url, update_after_clone, bare)
 
         # caches
         self._commit_ids = {}
 
         self.bookmarks = {}
 
     @LazyProperty
     def bare(self):
         return self._remote.bare()
 
     @LazyProperty
     def head(self):
         return self._remote.head()
 
     @LazyProperty
     def commit_ids(self):
         """
         Returns list of commit ids, in ascending order.  Being lazy
         attribute allows external tools to inject commit ids from cache.
         """
         commit_ids = self._get_all_commit_ids()
         self._rebuild_cache(commit_ids)
         return commit_ids
 
     def _rebuild_cache(self, commit_ids):
         self._commit_ids = dict((commit_id, index)
                                 for index, commit_id in enumerate(commit_ids))
 
     def run_git_command(self, 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 not isinstance(cmd, list):
             raise ValueError('cmd must be a list, got %s instead' % type(cmd))
 
         out, err = self._remote.run_git_command(cmd, **opts)
         log.debug('Stderr output of git command "%s":\n%s', cmd, err)
         return out, err
 
     @staticmethod
     def check_url(url, config):
         """
         Function will check given url and try to verify if it's a valid
         link. Sometimes it may happened that git 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, exception is also thrown
         when the return code is non 200
         """
         # check first if it's not an url
         if os.path.isdir(url) or url.startswith('file:'):
             return True
 
         if '+' in url.split('://', 1)[0]:
             url = url.split('+', 1)[1]
 
         # Request the _remote to verify the url
         return connection.Git.check_url(url, config.serialize())
 
     @staticmethod
     def is_valid_repository(path):
         if os.path.isdir(os.path.join(path, '.git')):
             return True
         # check case of bare repository
         try:
             GitRepository(path)
             return True
         except VCSError:
             pass
         return False
 
     def _init_repo(self, create, src_url=None, update_after_clone=False,
                    bare=False):
         if create and os.path.exists(self.path):
             raise RepositoryError(
                 "Cannot create repository at %s, location already exist"
                 % self.path)
 
         try:
             if create and src_url:
                 GitRepository.check_url(src_url, self.config)
                 self.clone(src_url, update_after_clone, bare)
             elif create:
                 os.makedirs(self.path, mode=0755)
 
                 if bare:
                     self._remote.init_bare()
                 else:
                     self._remote.init()
             else:
                 self._remote.assert_correct_path()
         # TODO: johbo: check if we have to translate the OSError here
         except OSError as err:
             raise RepositoryError(err)
 
     def _get_all_commit_ids(self, filters=None):
         # 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._remote.head()
         except KeyError:
             return []
 
         rev_filter = ['--branches', '--tags']
         extra_filter = []
 
         if filters:
             if filters.get('since'):
                 extra_filter.append('--since=%s' % (filters['since']))
             if filters.get('until'):
                 extra_filter.append('--until=%s' % (filters['until']))
             if filters.get('branch_name'):
                 rev_filter = ['--tags']
                 extra_filter.append(filters['branch_name'])
         rev_filter.extend(extra_filter)
 
             # if filters.get('start') or filters.get('end'):
             #     # skip is offset, max-count is limit
             #     if filters.get('start'):
             #         extra_filter += ' --skip=%s' % filters['start']
             #     if filters.get('end'):
             #         extra_filter += ' --max-count=%s' % (filters['end'] - (filters['start'] or 0))
 
         cmd = ['rev-list', '--reverse', '--date-order'] + rev_filter
         try:
             output, __ = self.run_git_command(cmd)
         except RepositoryError:
             # Can be raised for empty repositories
             return []
         return output.splitlines()
 
     def _get_commit_id(self, commit_id_or_idx):
         def is_null(value):
             return len(value) == commit_id_or_idx.count('0')
 
         if self.is_empty():
             raise EmptyRepositoryError("There are no commits yet")
 
         if commit_id_or_idx in (None, '', 'tip', 'HEAD', 'head', -1):
             return self.commit_ids[-1]
 
         is_bstr = isinstance(commit_id_or_idx, (str, unicode))
         if ((is_bstr and commit_id_or_idx.isdigit() and len(commit_id_or_idx) < 12)
             or isinstance(commit_id_or_idx, int) or is_null(commit_id_or_idx)):
             try:
                 commit_id_or_idx = self.commit_ids[int(commit_id_or_idx)]
             except Exception:
                 msg = "Commit %s does not exist for %s" % (
                     commit_id_or_idx, self)
                 raise CommitDoesNotExistError(msg)
 
         elif is_bstr:
             # check full path ref, eg. refs/heads/master
             ref_id = self._refs.get(commit_id_or_idx)
             if ref_id:
                 return ref_id
 
             # check branch name
             branch_ids = self.branches.values()
             ref_id = self._refs.get('refs/heads/%s' % commit_id_or_idx)
             if ref_id:
                 return ref_id
 
             # check tag name
             ref_id = self._refs.get('refs/tags/%s' % commit_id_or_idx)
             if ref_id:
                 return ref_id
 
             if (not SHA_PATTERN.match(commit_id_or_idx) or
                     commit_id_or_idx not in self.commit_ids):
                 msg = "Commit %s does not exist for %s" % (
                     commit_id_or_idx, self)
                 raise CommitDoesNotExistError(msg)
 
         # Ensure we return full id
         if not SHA_PATTERN.match(str(commit_id_or_idx)):
             raise CommitDoesNotExistError(
                 "Given commit id %s not recognized" % commit_id_or_idx)
         return commit_id_or_idx
 
     def get_hook_location(self):
         """
         returns absolute path to location where hooks are stored
         """
         loc = os.path.join(self.path, 'hooks')
         if not self.bare:
             loc = os.path.join(self.path, '.git', 'hooks')
         return loc
 
     @LazyProperty
     def last_change(self):
         """
         Returns last change made on this repository as
         `datetime.datetime` object.
         """
         return utcdate_fromtimestamp(self._get_mtime(), makedate()[1])
 
     def _get_mtime(self):
         try:
             return time.mktime(self.get_commit().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):
         description = self._remote.get_description()
         return safe_unicode(description or self.DEFAULT_DESCRIPTION)
 
     def _get_refs_entries(self, prefix='', reverse=False, strip_prefix=True):
         if self.is_empty():
             return OrderedDict()
 
         result = []
         for ref, sha in self._refs.iteritems():
             if ref.startswith(prefix):
                 ref_name = ref
                 if strip_prefix:
                     ref_name = ref[len(prefix):]
                 result.append((safe_unicode(ref_name), sha))
 
         def get_name(entry):
             return entry[0]
 
         return OrderedDict(sorted(result, key=get_name, reverse=reverse))
 
     def _get_branches(self):
         return self._get_refs_entries(prefix='refs/heads/', strip_prefix=True)
 
     @LazyProperty
     def branches(self):
         return self._get_branches()
 
     @LazyProperty
     def branches_closed(self):
         return {}
 
     @LazyProperty
     def branches_all(self):
         all_branches = {}
         all_branches.update(self.branches)
         all_branches.update(self.branches_closed)
         return all_branches
 
     @LazyProperty
     def tags(self):
         return self._get_tags()
 
     def _get_tags(self):
         return self._get_refs_entries(
             prefix='refs/tags/', strip_prefix=True, reverse=True)
 
     def tag(self, name, user, commit_id=None, message=None, date=None,
             **kwargs):
         # TODO: fix this method to apply annotated tags correct with message
         """
         Creates and returns a tag for the given ``commit_id``.
 
         :param name: name for new tag
         :param user: full username, i.e.: "Joe Doe <[e-mail]>"
         :param commit_id: commit 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)
         commit = self.get_commit(commit_id=commit_id)
         message = message or "Added tag %s for commit %s" % (
             name, commit.raw_id)
         self._remote.set_refs('refs/tags/%s' % name, commit._commit['id'])
 
         self._refs = self._get_refs()
         self.tags = self._get_tags()
         return commit
 
     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 = vcspath.join(
             self._remote.get_refs_path(), 'refs', 'tags', name)
         try:
             os.remove(tagpath)
             self._refs = self._get_refs()
             self.tags = self._get_tags()
         except OSError as e:
             raise RepositoryError(e.strerror)
 
     def _get_refs(self):
         return self._remote.get_refs()
 
     @LazyProperty
     def _refs(self):
         return self._get_refs()
 
     @property
     def _ref_tree(self):
         node = tree = {}
         for ref, sha in self._refs.iteritems():
             path = ref.split('/')
             for bit in path[:-1]:
                 node = node.setdefault(bit, {})
             node[path[-1]] = sha
             node = tree
         return tree
 
     def get_commit(self, commit_id=None, commit_idx=None, pre_load=None):
         """
         Returns `GitCommit` object representing commit from git repository
         at the given `commit_id` or head (most recent commit) if None given.
         """
         if commit_id is not None:
             self._validate_commit_id(commit_id)
         elif commit_idx is not None:
             self._validate_commit_idx(commit_idx)
             commit_id = commit_idx
         commit_id = self._get_commit_id(commit_id)
         try:
             # Need to call remote to translate id for tagging scenario
             commit_id = self._remote.get_object(commit_id)["commit_id"]
             idx = self._commit_ids[commit_id]
         except KeyError:
             raise RepositoryError("Cannot get object with id %s" % commit_id)
 
         return GitCommit(self, commit_id, idx, pre_load=pre_load)
 
     def get_commits(
             self, start_id=None, end_id=None, start_date=None, end_date=None,
             branch_name=None, pre_load=None):
         """
         Returns generator of `GitCommit` objects from start to end (both
         are inclusive), in ascending date order.
 
         :param start_id: None, str(commit_id)
         :param end_id: None, str(commit_id)
         :param start_date: if specified, commits with commit date less than
           ``start_date`` would be filtered out from returned set
         :param end_date: if specified, commits with commit date greater than
           ``end_date`` would be filtered out from returned set
         :param branch_name: if specified, commits not reachable from given
           branch would be filtered out from returned set
 
         :raise BranchDoesNotExistError: If given `branch_name` does not
             exist.
         :raise CommitDoesNotExistError: If commits for given `start` or
           `end` could not be found.
 
         """
         if self.is_empty():
             raise EmptyRepositoryError("There are no commits yet")
         self._validate_branch_name(branch_name)
 
         if start_id is not None:
             self._validate_commit_id(start_id)
         if end_id is not None:
             self._validate_commit_id(end_id)
 
         start_raw_id = self._get_commit_id(start_id)
         start_pos = self._commit_ids[start_raw_id] if start_id else None
         end_raw_id = self._get_commit_id(end_id)
         end_pos = max(0, self._commit_ids[end_raw_id]) if end_id else None
 
         if None not in [start_id, end_id] and start_pos > end_pos:
             raise RepositoryError(
                 "Start commit '%s' cannot be after end commit '%s'" %
                 (start_id, end_id))
 
         if end_pos is not None:
             end_pos += 1
 
         filter_ = []
         if branch_name:
             filter_.append({'branch_name': branch_name})
         if start_date and not end_date:
             filter_.append({'since': start_date})
         if end_date and not start_date:
             filter_.append({'until': end_date})
         if start_date and end_date:
             filter_.append({'since': start_date})
             filter_.append({'until': end_date})
 
         # if start_pos or end_pos:
         #     filter_.append({'start': start_pos})
         #     filter_.append({'end': end_pos})
 
         if filter_:
             revfilters = {
                 'branch_name': branch_name,
                 'since': start_date.strftime('%m/%d/%y %H:%M:%S') if start_date else None,
                 'until': end_date.strftime('%m/%d/%y %H:%M:%S') if end_date else None,
                 'start': start_pos,
                 'end': end_pos,
             }
             commit_ids = self._get_all_commit_ids(filters=revfilters)
 
             # pure python stuff, it's slow due to walker walking whole repo
             # def get_revs(walker):
             #     for walker_entry in walker:
             #         yield walker_entry.commit.id
             # revfilters = {}
             # commit_ids = list(reversed(list(get_revs(self._repo.get_walker(**revfilters)))))
         else:
             commit_ids = self.commit_ids
 
         if start_pos or end_pos:
             commit_ids = commit_ids[start_pos: end_pos]
 
         return CollectionGenerator(self, commit_ids, pre_load=pre_load)
 
     def get_diff(
             self, commit1, commit2, path='', ignore_whitespace=False,
             context=3, path1=None):
         """
         Returns (git like) *diff*, as plain text. Shows changes introduced by
         ``commit2`` since ``commit1``.
 
         :param commit1: Entry point from which diff is shown. Can be
           ``self.EMPTY_COMMIT`` - in this case, patch showing all
           the changes since empty state of the repository until ``commit2``
         :param commit2: Until which commits 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``.
         """
         self._validate_diff_commits(commit1, commit2)
         if path1 is not None and path1 != path:
             raise ValueError("Diff of two different paths not supported.")
 
         flags = [
             '-U%s' % context, '--full-index', '--binary', '-p',
             '-M', '--abbrev=40']
         if ignore_whitespace:
             flags.append('-w')
 
         if commit1 == self.EMPTY_COMMIT:
             cmd = ['show'] + flags + [commit2.raw_id]
         else:
             cmd = ['diff'] + flags + [commit1.raw_id, commit2.raw_id]
 
         if path:
             cmd.extend(['--', path])
 
         stdout, __ = self.run_git_command(cmd)
         # If we used 'show' command, strip first few lines (until actual diff
         # starts)
         if commit1 == self.EMPTY_COMMIT:
             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 GitDiff(stdout)
 
     def strip(self, commit_id, branch_name):
         commit = self.get_commit(commit_id=commit_id)
         if commit.merge:
             raise Exception('Cannot reset to merge commit')
 
         # parent is going to be the new head now
         commit = commit.parents[0]
         self._remote.set_refs('refs/heads/%s' % branch_name, commit.raw_id)
 
         self.commit_ids = self._get_all_commit_ids()
         self._rebuild_cache(self.commit_ids)
 
     def get_common_ancestor(self, commit_id1, commit_id2, repo2):
         if commit_id1 == commit_id2:
             return commit_id1
 
         if self != repo2:
             commits = self._remote.get_missing_revs(
                 commit_id1, commit_id2, repo2.path)
             if commits:
                 commit = repo2.get_commit(commits[-1])
                 if commit.parents:
                     ancestor_id = commit.parents[0].raw_id
                 else:
                     ancestor_id = None
             else:
                 # no commits from other repo, ancestor_id is the commit_id2
                 ancestor_id = commit_id2
         else:
             output, __ = self.run_git_command(
                 ['merge-base', commit_id1, commit_id2])
             ancestor_id = re.findall(r'[0-9a-fA-F]{40}', output)[0]
 
         return ancestor_id
 
     def compare(self, commit_id1, commit_id2, repo2, merge, pre_load=None):
         repo1 = self
         ancestor_id = None
 
         if commit_id1 == commit_id2:
             commits = []
         elif repo1 != repo2:
             missing_ids = self._remote.get_missing_revs(commit_id1, commit_id2,
                                                         repo2.path)
             commits = [
                 repo2.get_commit(commit_id=commit_id, pre_load=pre_load)
                 for commit_id in reversed(missing_ids)]
         else:
             output, __ = repo1.run_git_command(
                 ['log', '--reverse', '--pretty=format: %H', '-s',
                  '%s..%s' % (commit_id1, commit_id2)])
             commits = [
                 repo1.get_commit(commit_id=commit_id, pre_load=pre_load)
                 for commit_id in re.findall(r'[0-9a-fA-F]{40}', output)]
 
         return commits
 
     @LazyProperty
     def in_memory_commit(self):
         """
         Returns ``GitInMemoryCommit`` object for this repository.
         """
         return GitInMemoryCommit(self)
 
     def clone(self, url, update_after_clone=True, bare=False):
         """
         Tries to clone commits 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).
         """
         # init_bare and init expect empty dir created to proceed
         if not os.path.exists(self.path):
             os.mkdir(self.path)
 
         if bare:
             self._remote.init_bare()
         else:
             self._remote.init()
 
         deferred = '^{}'
         valid_refs = ('refs/heads', 'refs/tags', 'HEAD')
 
         return self._remote.clone(
             url, deferred, valid_refs, update_after_clone)
 
     def pull(self, url, commit_ids=None):
         """
         Tries to pull changes from external location. We use fetch here since
         pull in get does merges and we want to be compatible with hg backend so
         pull == fetch in this case
         """
         self.fetch(url, commit_ids=commit_ids)
 
     def fetch(self, url, commit_ids=None):
         """
         Tries to fetch changes from external location.
         """
         refs = None
 
         if commit_ids is not None:
             remote_refs = self._remote.get_remote_refs(url)
             refs = [
                 ref for ref in remote_refs if remote_refs[ref] in commit_ids]
         self._remote.fetch(url, refs=refs)
 
     def set_refs(self, ref_name, commit_id):
         self._remote.set_refs(ref_name, commit_id)
 
     def remove_ref(self, ref_name):
         self._remote.remove_ref(ref_name)
 
     def _update_server_info(self):
         """
         runs gits update-server-info command in this repo instance
         """
         self._remote.update_server_info()
 
     def _current_branch(self):
         """
         Return the name of the current branch.
 
         It only works for non bare repositories (i.e. repositories with a
         working copy)
         """
         if self.bare:
             raise RepositoryError('Bare git repos do not have active branches')
 
         if self.is_empty():
             return None
 
         stdout, _ = self.run_git_command(['rev-parse', '--abbrev-ref', 'HEAD'])
         return stdout.strip()
 
     def _checkout(self, branch_name, create=False):
         """
         Checkout a branch in the working directory.
 
         It tries to create the branch if create is True, failing if the branch
         already exists.
 
         It only works for non bare repositories (i.e. repositories with a
         working copy)
         """
         if self.bare:
             raise RepositoryError('Cannot checkout branches in a bare git repo')
 
         cmd = ['checkout']
         if create:
             cmd.append('-b')
         cmd.append(branch_name)
         self.run_git_command(cmd, fail_on_stderr=False)
 
     def _local_clone(self, clone_path, branch_name):
         """
         Create a local clone of the current repo.
         """
         # N.B.(skreft): the --branch option is required as otherwise the shallow
         # clone will only fetch the active branch.
         cmd = ['clone', '--branch', branch_name, '--single-branch',
                self.path, os.path.abspath(clone_path)]
         self.run_git_command(cmd, fail_on_stderr=False)
 
     def _local_fetch(self, repository_path, branch_name):
         """
         Fetch a branch from a local repository.
         """
         repository_path = os.path.abspath(repository_path)
         if repository_path == self.path:
             raise ValueError('Cannot fetch from the same repository')
 
         cmd = ['fetch', '--no-tags', repository_path, branch_name]
         self.run_git_command(cmd, fail_on_stderr=False)
 
     def _last_fetch_heads(self):
         """
         Return the last fetched heads that need merging.
 
         The algorithm is defined at
         https://github.com/git/git/blob/v2.1.3/git-pull.sh#L283
         """
         if not self.bare:
             fetch_heads_path = os.path.join(self.path, '.git', 'FETCH_HEAD')
         else:
             fetch_heads_path = os.path.join(self.path, 'FETCH_HEAD')
 
         heads = []
         with open(fetch_heads_path) as f:
             for line in f:
                 if '    not-for-merge   ' in line:
                     continue
                 line = re.sub('\t.*', '', line, flags=re.DOTALL)
                 heads.append(line)
 
         return heads
 
     def _local_pull(self, repository_path, branch_name):
         """
         Pull a branch from a local repository.
         """
         if self.bare:
             raise RepositoryError('Cannot pull into a bare git repository')
         # N.B.(skreft): The --ff-only option is to make sure this is a
         # fast-forward (i.e., we are only pulling new changes and there are no
         # conflicts with our current branch)
         # Additionally, that option needs to go before --no-tags, otherwise git
         # pull complains about it being an unknown flag.
         cmd = ['pull', '--ff-only', '--no-tags', repository_path, branch_name]
         self.run_git_command(cmd, fail_on_stderr=False)
 
     def _local_merge(self, merge_message, user_name, user_email, heads):
         """
         Merge the given head into the checked out branch.
 
         It will force a merge commit.
 
         Currently it raises an error if the repo is empty, as it is not possible
         to create a merge commit in an empty repo.
 
         :param merge_message: The message to use for the merge commit.
         :param heads: the heads to merge.
         """
         if self.bare:
             raise RepositoryError('Cannot merge into a bare git repository')
 
         if not heads:
             return
 
         if self.is_empty():
             # TODO(skreft): do somehting more robust in this case.
             raise RepositoryError(
                 'Do not know how to merge into empty repositories yet')
 
         # N.B.(skreft): the --no-ff option is used to enforce the creation of a
         # commit message. We also specify the user who is doing the merge.
         cmd = ['-c', 'user.name=%s' % safe_str(user_name),
                '-c', 'user.email=%s' % safe_str(user_email),
                'merge', '--no-ff', '-m', safe_str(merge_message)]
         cmd.extend(heads)
         try:
             self.run_git_command(cmd, fail_on_stderr=False)
         except RepositoryError:
             # Cleanup any merge leftovers
             self.run_git_command(['merge', '--abort'], fail_on_stderr=False)
             raise
 
     def _local_push(
             self, source_branch, repository_path, target_branch,
             enable_hooks=False, rc_scm_data=None):
         """
         Push the source_branch to the given repository and target_branch.
 
         Currently it if the target_branch is not master and the target repo is
         empty, the push will work, but then GitRepository won't be able to find
         the pushed branch or the commits. As the HEAD will be corrupted (i.e.,
         pointing to master, which does not exist).
 
         It does not run the hooks in the target repo.
         """
         # TODO(skreft): deal with the case in which the target repo is empty,
         # and the target_branch is not master.
         target_repo = GitRepository(repository_path)
         if (not target_repo.bare and
                 target_repo._current_branch() == target_branch):
             # Git prevents pushing to the checked out branch, so simulate it by
             # pulling into the target repository.
             target_repo._local_pull(self.path, source_branch)
         else:
             cmd = ['push', os.path.abspath(repository_path),
                    '%s:%s' % (source_branch, target_branch)]
             gitenv = {}
             if rc_scm_data:
                 gitenv.update({'RC_SCM_DATA': rc_scm_data})
 
             if not enable_hooks:
                 gitenv['RC_SKIP_HOOKS'] = '1'
             self.run_git_command(cmd, fail_on_stderr=False, extra_env=gitenv)
 
     def _get_new_pr_branch(self, source_branch, target_branch):
         prefix = 'pr_%s-%s_' % (source_branch, target_branch)
         pr_branches = []
         for branch in self.branches:
             if branch.startswith(prefix):
                 pr_branches.append(int(branch[len(prefix):]))
 
         if not pr_branches:
             branch_id = 0
         else:
             branch_id = max(pr_branches) + 1
 
         return '%s%d' % (prefix, branch_id)
 
     def _merge_repo(self, shadow_repository_path, target_ref,
                     source_repo, source_ref, merge_message,
                     merger_name, merger_email, dry_run=False,
                     use_rebase=False):
         if target_ref.commit_id != self.branches[target_ref.name]:
             return MergeResponse(
                 False, False, None, MergeFailureReason.TARGET_IS_NOT_HEAD)
 
         shadow_repo = GitRepository(shadow_repository_path)
         shadow_repo._checkout(target_ref.name)
         shadow_repo._local_pull(self.path, target_ref.name)
         # Need to reload repo to invalidate the cache, or otherwise we cannot
         # retrieve the last target commit.
         shadow_repo = GitRepository(shadow_repository_path)
         if target_ref.commit_id != shadow_repo.branches[target_ref.name]:
             return MergeResponse(
                 False, False, None, MergeFailureReason.TARGET_IS_NOT_HEAD)
 
         pr_branch = shadow_repo._get_new_pr_branch(
             source_ref.name, target_ref.name)
         shadow_repo._checkout(pr_branch, create=True)
         try:
             shadow_repo._local_fetch(source_repo.path, source_ref.name)
         except RepositoryError:
             log.exception('Failure when doing local fetch on git shadow repo')
             return MergeResponse(
-                False, False, None, MergeFailureReason.MISSING_COMMIT)
+                False, False, None, MergeFailureReason.MISSING_SOURCE_REF)
 
         merge_ref = None
         merge_failure_reason = MergeFailureReason.NONE
         try:
             shadow_repo._local_merge(merge_message, merger_name, merger_email,
                                      [source_ref.commit_id])
             merge_possible = True
 
             # Need to reload repo to invalidate the cache, or otherwise we
             # cannot retrieve the merge commit.
             shadow_repo = GitRepository(shadow_repository_path)
             merge_commit_id = shadow_repo.branches[pr_branch]
 
             # Set a reference pointing to the merge commit. This reference may
             # be used to easily identify the last successful merge commit in
             # the shadow repository.
             shadow_repo.set_refs('refs/heads/pr-merge', merge_commit_id)
             merge_ref = Reference('branch', 'pr-merge', merge_commit_id)
         except RepositoryError:
             log.exception('Failure when doing local merge on git shadow repo')
             merge_possible = False
             merge_failure_reason = MergeFailureReason.MERGE_FAILED
 
         if merge_possible and not dry_run:
             try:
                 shadow_repo._local_push(
                     pr_branch, self.path, target_ref.name, enable_hooks=True,
                     rc_scm_data=self.config.get('rhodecode', 'RC_SCM_DATA'))
                 merge_succeeded = True
             except RepositoryError:
                 log.exception(
                     'Failure when doing local push on git shadow repo')
                 merge_succeeded = False
                 merge_failure_reason = MergeFailureReason.PUSH_FAILED
         else:
             merge_succeeded = False
 
         return MergeResponse(
             merge_possible, merge_succeeded, merge_ref,
             merge_failure_reason)
 
     def _get_shadow_repository_path(self, workspace_id):
         # The name of the shadow repository must start with '.', so it is
         # skipped by 'rhodecode.lib.utils.get_filesystem_repos'.
         return os.path.join(
             os.path.dirname(self.path),
             '.__shadow_%s_%s' % (os.path.basename(self.path), workspace_id))
 
     def _maybe_prepare_merge_workspace(self, workspace_id, target_ref):
         shadow_repository_path = self._get_shadow_repository_path(workspace_id)
         if not os.path.exists(shadow_repository_path):
             self._local_clone(shadow_repository_path, target_ref.name)
 
         return shadow_repository_path
 
     def cleanup_merge_workspace(self, workspace_id):
         shadow_repository_path = self._get_shadow_repository_path(workspace_id)
         shutil.rmtree(shadow_repository_path, ignore_errors=True)

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

@@ -1,798 +1,803 b''
 # -*- coding: utf-8 -*-
 
 # Copyright (C) 2014-2016  RhodeCode GmbH
 #
 # This program is free software: you can redistribute it and/or modify
 # it under the terms of the GNU Affero General Public License, version 3
 # (only), as published by the Free Software Foundation.
 #
 # 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 Affero General Public License
 # along with this program.  If not, see <http://www.gnu.org/licenses/>.
 #
 # This program is dual-licensed. If you wish to learn more about the
 # RhodeCode Enterprise Edition, including its added features, Support services,
 # and proprietary license terms, please see https://rhodecode.com/licenses/
 
 """
 HG repository module
 """
 
 import logging
 import binascii
 import os
 import shutil
 import urllib
 
 from zope.cachedescriptors.property import Lazy as LazyProperty
 
 from rhodecode.lib.compat import OrderedDict
 from rhodecode.lib.datelib import (
     date_to_timestamp_plus_offset, utcdate_fromtimestamp, makedate,
     date_astimestamp)
 from rhodecode.lib.utils import safe_unicode, safe_str
 from rhodecode.lib.vcs import connection
 from rhodecode.lib.vcs.backends.base import (
     BaseRepository, CollectionGenerator, Config, MergeResponse,
     MergeFailureReason, Reference)
 from rhodecode.lib.vcs.backends.hg.commit import MercurialCommit
 from rhodecode.lib.vcs.backends.hg.diff import MercurialDiff
 from rhodecode.lib.vcs.backends.hg.inmemory import MercurialInMemoryCommit
 from rhodecode.lib.vcs.exceptions import (
     EmptyRepositoryError, RepositoryError, TagAlreadyExistError,
     TagDoesNotExistError, CommitDoesNotExistError)
 
 hexlify = binascii.hexlify
 nullid = "\0" * 20
 
 log = logging.getLogger(__name__)
 
 
 class MercurialRepository(BaseRepository):
     """
     Mercurial repository backend
     """
     DEFAULT_BRANCH_NAME = 'default'
 
     def __init__(self, repo_path, config=None, create=False, src_url=None,
                  update_after_clone=False, with_wire=None):
         """
         Raises RepositoryError if repository could not be find at the given
         ``repo_path``.
 
         :param repo_path: local path of the repository
         :param config: config object containing the repo configuration
         :param create=False: if set to True, would try to create repository if
            it does not exist rather than raising exception
         :param src_url=None: would try to clone repository from given location
         :param update_after_clone=False: sets update of working copy after
            making a clone
         """
         self.path = safe_str(os.path.abspath(repo_path))
         self.config = config if config else Config()
         self._remote = connection.Hg(
             self.path, self.config, with_wire=with_wire)
 
         self._init_repo(create, src_url, update_after_clone)
 
         # caches
         self._commit_ids = {}
 
     @LazyProperty
     def commit_ids(self):
         """
         Returns list of commit ids, in ascending order.  Being lazy
         attribute allows external tools to inject shas from cache.
         """
         commit_ids = self._get_all_commit_ids()
         self._rebuild_cache(commit_ids)
         return commit_ids
 
     def _rebuild_cache(self, commit_ids):
         self._commit_ids = dict((commit_id, index)
                                 for index, commit_id in enumerate(commit_ids))
 
     @LazyProperty
     def branches(self):
         return self._get_branches()
 
     @LazyProperty
     def branches_closed(self):
         return self._get_branches(active=False, closed=True)
 
     @LazyProperty
     def branches_all(self):
         all_branches = {}
         all_branches.update(self.branches)
         all_branches.update(self.branches_closed)
         return all_branches
 
     def _get_branches(self, active=True, closed=False):
         """
         Gets branches for this repository
         Returns only not closed active branches by default
 
         :param active: return also active branches
         :param closed: return also closed branches
 
         """
         if self.is_empty():
             return {}
 
         def get_name(ctx):
             return ctx[0]
 
         _branches = [(safe_unicode(n), hexlify(h),) for n, h in
                      self._remote.branches(active, closed).items()]
 
         return OrderedDict(sorted(_branches, key=get_name, reverse=False))
 
     @LazyProperty
     def tags(self):
         """
         Gets tags for this repository
         """
         return self._get_tags()
 
     def _get_tags(self):
         if self.is_empty():
             return {}
 
         def get_name(ctx):
             return ctx[0]
 
         _tags = [(safe_unicode(n), hexlify(h),) for n, h in
                  self._remote.tags().items()]
 
         return OrderedDict(sorted(_tags, key=get_name, reverse=True))
 
     def tag(self, name, user, commit_id=None, message=None, date=None,
             **kwargs):
         """
         Creates and returns a tag for the given ``commit_id``.
 
         :param name: name for new tag
         :param user: full username, i.e.: "Joe Doe <[e-mail]>"
         :param commit_id: commit 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)
         commit = self.get_commit(commit_id=commit_id)
         local = kwargs.setdefault('local', False)
 
         if message is None:
             message = "Added tag %s for commit %s" % (name, commit.short_id)
 
         date, tz = date_to_timestamp_plus_offset(date)
 
         self._remote.tag(
             name, commit.raw_id, message, local, user, date, tz)
         self._remote.invalidate_vcs_cache()
 
         # Reinitialize tags
         self.tags = self._get_tags()
         tag_id = self.tags[name]
 
         return self.get_commit(commit_id=tag_id)
 
     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)
         if message is None:
             message = "Removed tag %s" % name
         local = False
 
         date, tz = date_to_timestamp_plus_offset(date)
 
         self._remote.tag(name, nullid, message, local, user, date, tz)
         self._remote.invalidate_vcs_cache()
         self.tags = self._get_tags()
 
     @LazyProperty
     def bookmarks(self):
         """
         Gets bookmarks for this repository
         """
         return self._get_bookmarks()
 
     def _get_bookmarks(self):
         if self.is_empty():
             return {}
 
         def get_name(ctx):
             return ctx[0]
 
         _bookmarks = [
             (safe_unicode(n), hexlify(h)) for n, h in
             self._remote.bookmarks().items()]
 
         return OrderedDict(sorted(_bookmarks, key=get_name))
 
     def _get_all_commit_ids(self):
         return self._remote.get_all_commit_ids('visible')
 
     def get_diff(
             self, commit1, commit2, path='', ignore_whitespace=False,
             context=3, path1=None):
         """
         Returns (git like) *diff*, as plain text. Shows changes introduced by
         `commit2` since `commit1`.
 
         :param commit1: Entry point from which diff is shown. Can be
           ``self.EMPTY_COMMIT`` - in this case, patch showing all
           the changes since empty state of the repository until `commit2`
         :param commit2: Until which commit 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``.
         """
         self._validate_diff_commits(commit1, commit2)
         if path1 is not None and path1 != path:
             raise ValueError("Diff of two different paths not supported.")
 
         if path:
             file_filter = [self.path, path]
         else:
             file_filter = None
 
         diff = self._remote.diff(
             commit1.raw_id, commit2.raw_id, file_filter=file_filter,
             opt_git=True, opt_ignorews=ignore_whitespace,
             context=context)
         return MercurialDiff(diff)
 
     def strip(self, commit_id, branch=None):
         self._remote.strip(commit_id, update=False, backup="none")
 
         self._remote.invalidate_vcs_cache()
         self.commit_ids = self._get_all_commit_ids()
         self._rebuild_cache(self.commit_ids)
 
     def get_common_ancestor(self, commit_id1, commit_id2, repo2):
         if commit_id1 == commit_id2:
             return commit_id1
 
         ancestors = self._remote.revs_from_revspec(
             "ancestor(id(%s), id(%s))", commit_id1, commit_id2,
             other_path=repo2.path)
         return repo2[ancestors[0]].raw_id if ancestors else None
 
     def compare(self, commit_id1, commit_id2, repo2, merge, pre_load=None):
         if commit_id1 == commit_id2:
             commits = []
         else:
             if merge:
                 indexes = self._remote.revs_from_revspec(
                     "ancestors(id(%s)) - ancestors(id(%s)) - id(%s)",
                     commit_id2, commit_id1, commit_id1, other_path=repo2.path)
             else:
                 indexes = self._remote.revs_from_revspec(
                     "id(%s)..id(%s) - id(%s)", commit_id1, commit_id2,
                     commit_id1, other_path=repo2.path)
 
             commits = [repo2.get_commit(commit_idx=idx, pre_load=pre_load)
                        for idx in indexes]
 
         return commits
 
     @staticmethod
     def check_url(url, config):
         """
         Function 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, exception is also thrown
         when the return code is non 200
         """
         # check first if it's not an local url
         if os.path.isdir(url) or url.startswith('file:'):
             return True
 
         # Request the _remote to verify the url
         return connection.Hg.check_url(url, config.serialize())
 
     @staticmethod
     def is_valid_repository(path):
         return os.path.isdir(os.path.join(path, '.hg'))
 
     def _init_repo(self, create, src_url=None, update_after_clone=False):
         """
         Function will check for mercurial repository in given path. If there
         is no repository in that path it will raise an exception unless
         `create` parameter is set to True - in that case repository would
         be created.
 
         If `src_url` is given, would try to clone repository from the
         location at given clone_point. Additionally it'll make update to
         working copy accordingly to `update_after_clone` flag.
         """
         if create and os.path.exists(self.path):
             raise RepositoryError(
                 "Cannot create repository at %s, location already exist"
                 % self.path)
 
         if src_url:
             url = str(self._get_url(src_url))
             MercurialRepository.check_url(url, self.config)
 
             self._remote.clone(url, self.path, update_after_clone)
 
             # Don't try to create if we've already cloned repo
             create = False
 
         if create:
             os.makedirs(self.path, mode=0755)
 
         self._remote.localrepository(create)
 
     @LazyProperty
     def in_memory_commit(self):
         return MercurialInMemoryCommit(self)
 
     @LazyProperty
     def description(self):
         description = self._remote.get_config_value(
             'web', 'description', untrusted=True)
         return safe_unicode(description or self.DEFAULT_DESCRIPTION)
 
     @LazyProperty
     def contact(self):
         contact = (
             self._remote.get_config_value("web", "contact") or
             self._remote.get_config_value("ui", "username"))
         return safe_unicode(contact or self.DEFAULT_CONTACT)
 
     @LazyProperty
     def last_change(self):
         """
         Returns last change made on this repository as
         `datetime.datetime` object
         """
         return utcdate_fromtimestamp(self._get_mtime(), makedate()[1])
 
     def _get_mtime(self):
         try:
             return date_astimestamp(self.get_commit().date)
         except RepositoryError:
             # fallback to filesystem
             cl_path = os.path.join(self.path, '.hg', "00changelog.i")
             st_path = os.path.join(self.path, '.hg', "store")
             if os.path.exists(cl_path):
                 return os.stat(cl_path).st_mtime
             else:
                 return os.stat(st_path).st_mtime
 
     def _sanitize_commit_idx(self, idx):
         # Note: Mercurial has ``int(-1)`` reserved as not existing id_or_idx
         # number. A `long` is treated in the correct way though. So we convert
         # `int` to `long` here to make sure it is handled correctly.
         if isinstance(idx, int):
             return long(idx)
         return idx
 
     def _get_url(self, url):
         """
         Returns normalized url. If schema is not given, would fall
         to filesystem
         (``file:///``) schema.
         """
         url = url.encode('utf8')
         if url != 'default' and '://' not in url:
             url = "file:" + urllib.pathname2url(url)
         return url
 
     def get_hook_location(self):
         """
         returns absolute path to location where hooks are stored
         """
         return os.path.join(self.path, '.hg', '.hgrc')
 
     def get_commit(self, commit_id=None, commit_idx=None, pre_load=None):
         """
         Returns ``MercurialCommit`` object representing repository's
         commit at the given `commit_id` or `commit_idx`.
         """
         if self.is_empty():
             raise EmptyRepositoryError("There are no commits yet")
 
         if commit_id is not None:
             self._validate_commit_id(commit_id)
             try:
                 idx = self._commit_ids[commit_id]
                 return MercurialCommit(self, commit_id, idx, pre_load=pre_load)
             except KeyError:
                 pass
         elif commit_idx is not None:
             self._validate_commit_idx(commit_idx)
             commit_idx = self._sanitize_commit_idx(commit_idx)
             try:
                 id_ = self.commit_ids[commit_idx]
                 if commit_idx < 0:
                     commit_idx += len(self.commit_ids)
                 return MercurialCommit(
                     self, id_, commit_idx, pre_load=pre_load)
             except IndexError:
                 commit_id = commit_idx
         else:
             commit_id = "tip"
 
         # TODO Paris: Ugly hack to "serialize" long for msgpack
         if isinstance(commit_id, long):
             commit_id = float(commit_id)
 
         if isinstance(commit_id, unicode):
             commit_id = safe_str(commit_id)
 
         raw_id, idx = self._remote.lookup(commit_id, both=True)
 
         return MercurialCommit(self, raw_id, idx, pre_load=pre_load)
 
     def get_commits(
             self, start_id=None, end_id=None, start_date=None, end_date=None,
             branch_name=None, pre_load=None):
         """
         Returns generator of ``MercurialCommit`` objects from start to end
         (both are inclusive)
 
         :param start_id: None, str(commit_id)
         :param end_id: None, str(commit_id)
         :param start_date: if specified, commits with commit date less than
           ``start_date`` would be filtered out from returned set
         :param end_date: if specified, commits with commit date greater than
           ``end_date`` would be filtered out from returned set
         :param branch_name: if specified, commits not reachable from given
           branch would be filtered out from returned set
 
         :raise BranchDoesNotExistError: If given ``branch_name`` does not
             exist.
         :raise CommitDoesNotExistError: If commit for given ``start`` or
           ``end`` could not be found.
         """
         # actually we should check now if it's not an empty repo
         branch_ancestors = False
         if self.is_empty():
             raise EmptyRepositoryError("There are no commits yet")
         self._validate_branch_name(branch_name)
 
         if start_id is not None:
             self._validate_commit_id(start_id)
             c_start = self.get_commit(commit_id=start_id)
             start_pos = self._commit_ids[c_start.raw_id]
         else:
             start_pos = None
 
         if end_id is not None:
             self._validate_commit_id(end_id)
             c_end = self.get_commit(commit_id=end_id)
             end_pos = max(0, self._commit_ids[c_end.raw_id])
         else:
             end_pos = None
 
         if None not in [start_id, end_id] and start_pos > end_pos:
             raise RepositoryError(
                 "Start commit '%s' cannot be after end commit '%s'" %
                 (start_id, end_id))
 
         if end_pos is not None:
             end_pos += 1
 
         commit_filter = []
         if branch_name and not branch_ancestors:
             commit_filter.append('branch("%s")' % branch_name)
         elif branch_name and branch_ancestors:
             commit_filter.append('ancestors(branch("%s"))' % branch_name)
         if start_date and not end_date:
             commit_filter.append('date(">%s")' % start_date)
         if end_date and not start_date:
             commit_filter.append('date("<%s")' % end_date)
         if start_date and end_date:
             commit_filter.append(
                 'date(">%s") and date("<%s")' % (start_date, end_date))
 
         # TODO: johbo: Figure out a simpler way for this solution
         collection_generator = CollectionGenerator
         if commit_filter:
             commit_filter = map(safe_str, commit_filter)
             revisions = self._remote.rev_range(commit_filter)
             collection_generator = MercurialIndexBasedCollectionGenerator
         else:
             revisions = self.commit_ids
 
         if start_pos or end_pos:
             revisions = revisions[start_pos:end_pos]
 
         return collection_generator(self, revisions, pre_load=pre_load)
 
     def pull(self, url, commit_ids=None):
         """
         Tries to pull changes from external location.
 
         :param commit_ids: Optional. Can be set to a list of commit ids
            which shall be pulled from the other repository.
         """
         url = self._get_url(url)
         self._remote.pull(url, commit_ids=commit_ids)
         self._remote.invalidate_vcs_cache()
 
     def _local_clone(self, clone_path):
         """
         Create a local clone of the current repo.
         """
         self._remote.clone(self.path, clone_path, update_after_clone=True,
                            hooks=False)
 
     def _update(self, revision, clean=False):
         """
         Update the working copty to the specified revision.
         """
         self._remote.update(revision, clean=clean)
 
     def _identify(self):
         """
         Return the current state of the working directory.
         """
         return self._remote.identify().strip().rstrip('+')
 
     def _heads(self, branch=None):
         """
         Return the commit ids of the repository heads.
         """
         return self._remote.heads(branch=branch).strip().split(' ')
 
     def _ancestor(self, revision1, revision2):
         """
         Return the common ancestor of the two revisions.
         """
         return self._remote.ancestor(
             revision1, revision2).strip().split(':')[-1]
 
     def _local_push(
             self, revision, repository_path, push_branches=False,
             enable_hooks=False):
         """
         Push the given revision to the specified repository.
 
         :param push_branches: allow to create branches in the target repo.
         """
         self._remote.push(
             [revision], repository_path, hooks=enable_hooks,
             push_branches=push_branches)
 
     def _local_merge(self, target_ref, merge_message, user_name, user_email,
                      source_ref, use_rebase=False):
         """
         Merge the given source_revision into the checked out revision.
 
         Returns the commit id of the merge and a boolean indicating if the
         commit needs to be pushed.
         """
         self._update(target_ref.commit_id)
 
         ancestor = self._ancestor(target_ref.commit_id, source_ref.commit_id)
         is_the_same_branch = self._is_the_same_branch(target_ref, source_ref)
 
         if ancestor == source_ref.commit_id:
             # Nothing to do, the changes were already integrated
             return target_ref.commit_id, False
 
         elif ancestor == target_ref.commit_id and is_the_same_branch:
             # In this case we should force a commit message
             return source_ref.commit_id, True
 
         if use_rebase:
             try:
                 bookmark_name = 'rcbook%s%s' % (source_ref.commit_id,
                                                 target_ref.commit_id)
                 self.bookmark(bookmark_name, revision=source_ref.commit_id)
                 self._remote.rebase(
                     source=source_ref.commit_id, dest=target_ref.commit_id)
                 self._remote.invalidate_vcs_cache()
                 self._update(bookmark_name)
                 return self._identify(), True
             except RepositoryError:
                 # The rebase-abort may raise another exception which 'hides'
                 # the original one, therefore we log it here.
                 log.exception('Error while rebasing shadow repo during merge.')
 
                 # Cleanup any rebase leftovers
                 self._remote.invalidate_vcs_cache()
                 self._remote.rebase(abort=True)
                 self._remote.invalidate_vcs_cache()
                 self._remote.update(clean=True)
                 raise
         else:
             try:
                 self._remote.merge(source_ref.commit_id)
                 self._remote.invalidate_vcs_cache()
                 self._remote.commit(
                     message=safe_str(merge_message),
                     username=safe_str('%s <%s>' % (user_name, user_email)))
                 self._remote.invalidate_vcs_cache()
                 return self._identify(), True
             except RepositoryError:
                 # Cleanup any merge leftovers
                 self._remote.update(clean=True)
                 raise
 
     def _is_the_same_branch(self, target_ref, source_ref):
         return (
             self._get_branch_name(target_ref) ==
             self._get_branch_name(source_ref))
 
     def _get_branch_name(self, ref):
         if ref.type == 'branch':
             return ref.name
         return self._remote.ctx_branch(ref.commit_id)
 
     def _get_shadow_repository_path(self, workspace_id):
         # The name of the shadow repository must start with '.', so it is
         # skipped by 'rhodecode.lib.utils.get_filesystem_repos'.
         return os.path.join(
             os.path.dirname(self.path),
             '.__shadow_%s_%s' % (os.path.basename(self.path), workspace_id))
 
     def _maybe_prepare_merge_workspace(self, workspace_id, unused_target_ref):
         shadow_repository_path = self._get_shadow_repository_path(workspace_id)
         if not os.path.exists(shadow_repository_path):
             self._local_clone(shadow_repository_path)
             log.debug(
                 'Prepared shadow repository in %s', shadow_repository_path)
 
         return shadow_repository_path
 
     def cleanup_merge_workspace(self, workspace_id):
         shadow_repository_path = self._get_shadow_repository_path(workspace_id)
         shutil.rmtree(shadow_repository_path, ignore_errors=True)
 
     def _merge_repo(self, shadow_repository_path, target_ref,
                     source_repo, source_ref, merge_message,
                     merger_name, merger_email, dry_run=False,
                     use_rebase=False):
         if target_ref.commit_id not in self._heads():
             return MergeResponse(
                 False, False, None, MergeFailureReason.TARGET_IS_NOT_HEAD)
 
+        try:
         if (target_ref.type == 'branch' and
                 len(self._heads(target_ref.name)) != 1):
             return MergeResponse(
                 False, False, None,
                 MergeFailureReason.HG_TARGET_HAS_MULTIPLE_HEADS)
+        except CommitDoesNotExistError as e:
+            log.exception('Failure when looking up branch heads on hg target')
+            return MergeResponse(
+                False, False, None, MergeFailureReason.MISSING_TARGET_REF)
 
         shadow_repo = self._get_shadow_instance(shadow_repository_path)
 
         log.debug('Pulling in target reference %s', target_ref)
         self._validate_pull_reference(target_ref)
         shadow_repo._local_pull(self.path, target_ref)
         try:
             log.debug('Pulling in source reference %s', source_ref)
             source_repo._validate_pull_reference(source_ref)
             shadow_repo._local_pull(source_repo.path, source_ref)
         except CommitDoesNotExistError:
             log.exception('Failure when doing local pull on hg shadow repo')
             return MergeResponse(
-                False, False, None, MergeFailureReason.MISSING_COMMIT)
+                False, False, None, MergeFailureReason.MISSING_SOURCE_REF)
 
         merge_ref = None
         merge_failure_reason = MergeFailureReason.NONE
 
         try:
             merge_commit_id, needs_push = shadow_repo._local_merge(
                 target_ref, merge_message, merger_name, merger_email,
                 source_ref, use_rebase=use_rebase)
             merge_possible = True
 
             # Set a bookmark pointing to the merge commit. This bookmark may be
             # used to easily identify the last successful merge commit in the
             # shadow repository.
             shadow_repo.bookmark('pr-merge', revision=merge_commit_id)
             merge_ref = Reference('book', 'pr-merge', merge_commit_id)
         except RepositoryError:
             log.exception('Failure when doing local merge on hg shadow repo')
             merge_possible = False
             merge_failure_reason = MergeFailureReason.MERGE_FAILED
 
         if merge_possible and not dry_run:
             if needs_push:
                 # In case the target is a bookmark, update it, so after pushing
                 # the bookmarks is also updated in the target.
                 if target_ref.type == 'book':
                     shadow_repo.bookmark(
                         target_ref.name, revision=merge_commit_id)
 
                 try:
                     shadow_repo_with_hooks = self._get_shadow_instance(
                         shadow_repository_path,
                         enable_hooks=True)
                     # Note: the push_branches option will push any new branch
                     # defined in the source repository to the target. This may
                     # be dangerous as branches are permanent in Mercurial.
                     # This feature was requested in issue #441.
                     shadow_repo_with_hooks._local_push(
                         merge_commit_id, self.path, push_branches=True,
                         enable_hooks=True)
                     merge_succeeded = True
                 except RepositoryError:
                     log.exception(
                         'Failure when doing local push from the shadow '
                         'repository to the target repository.')
                     merge_succeeded = False
                     merge_failure_reason = MergeFailureReason.PUSH_FAILED
             else:
                 merge_succeeded = True
         else:
             merge_succeeded = False
 
         return MergeResponse(
             merge_possible, merge_succeeded, merge_ref, merge_failure_reason)
 
     def _get_shadow_instance(
             self, shadow_repository_path, enable_hooks=False):
         config = self.config.copy()
         if not enable_hooks:
             config.clear_section('hooks')
         return MercurialRepository(shadow_repository_path, config)
 
     def _validate_pull_reference(self, reference):
         if not (reference.name in self.bookmarks or
                 reference.name in self.branches or
                 self.get_commit(reference.commit_id)):
             raise CommitDoesNotExistError(
                 'Unknown branch, bookmark or commit id')
 
     def _local_pull(self, repository_path, reference):
         """
         Fetch a branch, bookmark or commit from a local repository.
         """
         repository_path = os.path.abspath(repository_path)
         if repository_path == self.path:
             raise ValueError('Cannot pull from the same repository')
 
         reference_type_to_option_name = {
             'book': 'bookmark',
             'branch': 'branch',
         }
         option_name = reference_type_to_option_name.get(
             reference.type, 'revision')
 
         if option_name == 'revision':
             ref = reference.commit_id
         else:
             ref = reference.name
 
         options = {option_name: [ref]}
         self._remote.pull_cmd(repository_path, hooks=False, **options)
         self._remote.invalidate_vcs_cache()
 
     def bookmark(self, bookmark, revision=None):
         if isinstance(bookmark, unicode):
             bookmark = safe_str(bookmark)
         self._remote.bookmark(bookmark, revision=revision)
         self._remote.invalidate_vcs_cache()
 
 
 class MercurialIndexBasedCollectionGenerator(CollectionGenerator):
 
     def _commit_factory(self, commit_id):
         return self.repo.get_commit(
             commit_idx=commit_id, pre_load=self.pre_load)

rhodecode/tests/vcs/test_repository.py

@@ -1,532 +1,532 b''
 # -*- coding: utf-8 -*-
 
 # Copyright (C) 2010-2016  RhodeCode GmbH
 #
 # This program is free software: you can redistribute it and/or modify
 # it under the terms of the GNU Affero General Public License, version 3
 # (only), as published by the Free Software Foundation.
 #
 # 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 Affero General Public License
 # along with this program.  If not, see <http://www.gnu.org/licenses/>.
 #
 # This program is dual-licensed. If you wish to learn more about the
 # RhodeCode Enterprise Edition, including its added features, Support services,
 # and proprietary license terms, please see https://rhodecode.com/licenses/
 
 import datetime
 from urllib2 import URLError
 
 import mock
 import pytest
 
 from rhodecode.lib.vcs import backends
 from rhodecode.lib.vcs.backends.base import (
     Config, BaseInMemoryCommit, Reference, MergeResponse, MergeFailureReason)
 from rhodecode.lib.vcs.exceptions import VCSError, RepositoryError
 from rhodecode.lib.vcs.nodes import FileNode
 from rhodecode.tests.vcs.base import BackendTestMixin
 
 
 class TestRepositoryBase(BackendTestMixin):
     recreate_repo_per_test = False
 
     def test_init_accepts_unicode_path(self, tmpdir):
         path = unicode(tmpdir.join(u'unicode ä'))
         self.Backend(path, create=True)
 
     def test_init_accepts_str_path(self, tmpdir):
         path = str(tmpdir.join('str ä'))
         self.Backend(path, create=True)
 
     def test_init_fails_if_path_does_not_exist(self, tmpdir):
         path = unicode(tmpdir.join('i-do-not-exist'))
         with pytest.raises(VCSError):
             self.Backend(path)
 
     def test_init_fails_if_path_is_not_a_valid_repository(self, tmpdir):
         path = unicode(tmpdir.mkdir(u'unicode ä'))
         with pytest.raises(VCSError):
             self.Backend(path)
 
     def test_has_commits_attribute(self):
         self.repo.commit_ids
 
     def test_name(self):
         assert self.repo.name.startswith('vcs-test')
 
     @pytest.mark.backends("hg", "git")
     def test_has_default_branch_name(self):
         assert self.repo.DEFAULT_BRANCH_NAME is not None
 
     @pytest.mark.backends("svn")
     def test_has_no_default_branch_name(self):
         assert self.repo.DEFAULT_BRANCH_NAME is None
 
     def test_has_empty_commit(self):
         assert self.repo.EMPTY_COMMIT_ID is not None
         assert self.repo.EMPTY_COMMIT is not None
 
     def test_empty_changeset_is_deprecated(self):
         def get_empty_changeset(repo):
             return repo.EMPTY_CHANGESET
         pytest.deprecated_call(get_empty_changeset, self.repo)
 
     def test_bookmarks(self):
         assert len(self.repo.bookmarks) == 0
 
     # TODO: Cover two cases: Local repo path, remote URL
     def test_check_url(self):
         config = Config()
         assert self.Backend.check_url(self.repo.path, config)
 
     def test_check_url_invalid(self):
         config = Config()
         with pytest.raises(URLError):
             self.Backend.check_url(self.repo.path + "invalid", config)
 
     def test_get_contact(self):
         self.repo.contact
 
     def test_get_description(self):
         self.repo.description
 
     def test_get_hook_location(self):
         assert len(self.repo.get_hook_location()) != 0
 
     def test_last_change(self):
         assert self.repo.last_change >= datetime.datetime(2010, 1, 1, 21, 0)
 
     def test_last_change_in_empty_repository(self, vcsbackend):
         delta = datetime.timedelta(seconds=1)
         start = datetime.datetime.now()
         empty_repo = vcsbackend.create_repo()
         now = datetime.datetime.now()
         assert empty_repo.last_change >= start - delta
         assert empty_repo.last_change <= now + delta
 
     def test_repo_equality(self):
         assert self.repo == self.repo
 
     def test_repo_equality_broken_object(self):
         import copy
         _repo = copy.copy(self.repo)
         delattr(_repo, 'path')
         assert self.repo != _repo
 
     def test_repo_equality_other_object(self):
         class dummy(object):
             path = self.repo.path
         assert self.repo != dummy()
 
     def test_get_commit_is_implemented(self):
         self.repo.get_commit()
 
     def test_get_commits_is_implemented(self):
         commit_iter = iter(self.repo.get_commits())
         commit = next(commit_iter)
         assert commit.idx == 0
 
     def test_supports_iteration(self):
         repo_iter = iter(self.repo)
         commit = next(repo_iter)
         assert commit.idx == 0
 
     def test_in_memory_commit(self):
         imc = self.repo.in_memory_commit
         assert isinstance(imc, BaseInMemoryCommit)
 
     @pytest.mark.backends("hg")
     def test__get_url_unicode(self):
         url = u'/home/repos/malmö'
         assert self.repo._get_url(url)
 
 
 class TestDeprecatedRepositoryAPI(BackendTestMixin):
     recreate_repo_per_test = False
 
     def test_revisions_is_deprecated(self):
         def get_revisions(repo):
             return repo.revisions
         pytest.deprecated_call(get_revisions, self.repo)
 
     def test_get_changeset_is_deprecated(self):
         pytest.deprecated_call(self.repo.get_changeset)
 
     def test_get_changesets_is_deprecated(self):
         pytest.deprecated_call(self.repo.get_changesets)
 
     def test_in_memory_changeset_is_deprecated(self):
         def get_imc(repo):
             return repo.in_memory_changeset
         pytest.deprecated_call(get_imc, self.repo)
 
 
 # TODO: these tests are incomplete, must check the resulting compare result for
 # correcteness
 class TestRepositoryCompare:
 
     @pytest.mark.parametrize('merge', [True, False])
     def test_compare_commits_of_same_repository(self, vcsbackend, merge):
         target_repo = vcsbackend.create_repo(number_of_commits=5)
         target_repo.compare(
             target_repo[1].raw_id, target_repo[3].raw_id, target_repo,
             merge=merge)
 
     @pytest.mark.xfail_backends('svn')
     @pytest.mark.parametrize('merge', [True, False])
     def test_compare_cloned_repositories(self, vcsbackend, merge):
         target_repo = vcsbackend.create_repo(number_of_commits=5)
         source_repo = vcsbackend.clone_repo(target_repo)
         assert target_repo != source_repo
 
         vcsbackend.add_file(source_repo, 'newfile', 'somecontent')
         source_commit = source_repo.get_commit()
 
         target_repo.compare(
             target_repo[1].raw_id, source_repo[3].raw_id, source_repo,
             merge=merge)
 
     @pytest.mark.xfail_backends('svn')
     @pytest.mark.parametrize('merge', [True, False])
     def test_compare_unrelated_repositories(self, vcsbackend, merge):
         orig = vcsbackend.create_repo(number_of_commits=5)
         unrelated = vcsbackend.create_repo(number_of_commits=5)
         assert orig != unrelated
 
         orig.compare(
             orig[1].raw_id, unrelated[3].raw_id, unrelated, merge=merge)
 
 
 class TestRepositoryGetCommonAncestor:
 
     def test_get_common_ancestor_from_same_repo_existing(self, vcsbackend):
         target_repo = vcsbackend.create_repo(number_of_commits=5)
 
         expected_ancestor = target_repo[2].raw_id
 
         assert target_repo.get_common_ancestor(
             commit_id1=target_repo[2].raw_id,
             commit_id2=target_repo[4].raw_id,
             repo2=target_repo
         ) == expected_ancestor
 
         assert target_repo.get_common_ancestor(
             commit_id1=target_repo[4].raw_id,
             commit_id2=target_repo[2].raw_id,
             repo2=target_repo
         ) == expected_ancestor
 
     @pytest.mark.xfail_backends("svn")
     def test_get_common_ancestor_from_cloned_repo_existing(self, vcsbackend):
         target_repo = vcsbackend.create_repo(number_of_commits=5)
         source_repo = vcsbackend.clone_repo(target_repo)
         assert target_repo != source_repo
 
         vcsbackend.add_file(source_repo, 'newfile', 'somecontent')
         source_commit = source_repo.get_commit()
 
         expected_ancestor = target_repo[4].raw_id
 
         assert target_repo.get_common_ancestor(
             commit_id1=target_repo[4].raw_id,
             commit_id2=source_commit.raw_id,
             repo2=source_repo
         ) == expected_ancestor
 
         assert target_repo.get_common_ancestor(
             commit_id1=source_commit.raw_id,
             commit_id2=target_repo[4].raw_id,
             repo2=target_repo
         ) == expected_ancestor
 
     @pytest.mark.xfail_backends("svn")
     def test_get_common_ancestor_from_unrelated_repo_missing(self, vcsbackend):
         original = vcsbackend.create_repo(number_of_commits=5)
         unrelated = vcsbackend.create_repo(number_of_commits=5)
         assert original != unrelated
 
         assert original.get_common_ancestor(
             commit_id1=original[0].raw_id,
             commit_id2=unrelated[0].raw_id,
             repo2=unrelated
         ) == None
 
         assert original.get_common_ancestor(
             commit_id1=original[-1].raw_id,
             commit_id2=unrelated[-1].raw_id,
             repo2=unrelated
         ) == None
 
 
 @pytest.mark.backends("git", "hg")
 class TestRepositoryMerge:
     def prepare_for_success(self, vcsbackend):
         self.target_repo = vcsbackend.create_repo(number_of_commits=1)
         self.source_repo = vcsbackend.clone_repo(self.target_repo)
         vcsbackend.add_file(self.target_repo, 'README_MERGE1', 'Version 1')
         vcsbackend.add_file(self.source_repo, 'README_MERGE2', 'Version 2')
         imc = self.source_repo.in_memory_commit
         imc.add(FileNode('file_x', content=self.source_repo.name))
         imc.commit(
             message=u'Automatic commit from repo merge test',
             author=u'Automatic')
         self.target_commit = self.target_repo.get_commit()
         self.source_commit = self.source_repo.get_commit()
         # This only works for Git and Mercurial
         default_branch = self.target_repo.DEFAULT_BRANCH_NAME
         self.target_ref = Reference(
             'branch', default_branch, self.target_commit.raw_id)
         self.source_ref = Reference(
             'branch', default_branch, self.source_commit.raw_id)
         self.workspace = 'test-merge'
 
     def prepare_for_conflict(self, vcsbackend):
         self.target_repo = vcsbackend.create_repo(number_of_commits=1)
         self.source_repo = vcsbackend.clone_repo(self.target_repo)
         vcsbackend.add_file(self.target_repo, 'README_MERGE', 'Version 1')
         vcsbackend.add_file(self.source_repo, 'README_MERGE', 'Version 2')
         self.target_commit = self.target_repo.get_commit()
         self.source_commit = self.source_repo.get_commit()
         # This only works for Git and Mercurial
         default_branch = self.target_repo.DEFAULT_BRANCH_NAME
         self.target_ref = Reference(
             'branch', default_branch, self.target_commit.raw_id)
         self.source_ref = Reference(
             'branch', default_branch, self.source_commit.raw_id)
         self.workspace = 'test-merge'
 
     def test_merge_success(self, vcsbackend):
         self.prepare_for_success(vcsbackend)
 
         merge_response = self.target_repo.merge(
             self.target_ref, self.source_repo, self.source_ref, self.workspace,
             'test user', '[e-mail]', 'merge message 1',
             dry_run=False)
         expected_merge_response = MergeResponse(
             True, True, merge_response.merge_ref,
             MergeFailureReason.NONE)
         assert merge_response == expected_merge_response
 
         target_repo = backends.get_backend(vcsbackend.alias)(
             self.target_repo.path)
         target_commits = list(target_repo.get_commits())
         commit_ids = [c.raw_id for c in target_commits[:-1]]
         assert self.source_ref.commit_id in commit_ids
         assert self.target_ref.commit_id in commit_ids
 
         merge_commit = target_commits[-1]
         assert merge_commit.raw_id == merge_response.merge_ref.commit_id
         assert merge_commit.message.strip() == 'merge message 1'
         assert merge_commit.author == 'test user <[e-mail]>'
 
         # We call it twice so to make sure we can handle updates
         target_ref = Reference(
             self.target_ref.type, self.target_ref.name,
             merge_response.merge_ref.commit_id)
 
         merge_response = target_repo.merge(
             target_ref, self.source_repo, self.source_ref, self.workspace,
             'test user', '[e-mail]', 'merge message 2',
             dry_run=False)
         expected_merge_response = MergeResponse(
             True, True, merge_response.merge_ref,
             MergeFailureReason.NONE)
         assert merge_response == expected_merge_response
 
         target_repo = backends.get_backend(
             vcsbackend.alias)(self.target_repo.path)
         merge_commit = target_repo.get_commit(
             merge_response.merge_ref.commit_id)
         assert merge_commit.message.strip() == 'merge message 1'
         assert merge_commit.author == 'test user <[e-mail]>'
 
     def test_merge_success_dry_run(self, vcsbackend):
         self.prepare_for_success(vcsbackend)
 
         merge_response = self.target_repo.merge(
             self.target_ref, self.source_repo, self.source_ref, self.workspace,
             dry_run=True)
 
         # We call it twice so to make sure we can handle updates
         merge_response_update = self.target_repo.merge(
             self.target_ref, self.source_repo, self.source_ref, self.workspace,
             dry_run=True)
 
         # Multiple merges may differ in their commit id. Therefore we set the
         # commit id to `None` before comparing the merge responses.
         merge_response = merge_response._replace(
             merge_ref=merge_response.merge_ref._replace(commit_id=None))
         merge_response_update = merge_response_update._replace(
             merge_ref=merge_response_update.merge_ref._replace(commit_id=None))
 
         assert merge_response == merge_response_update
         assert merge_response.possible is True
         assert merge_response.executed is False
         assert merge_response.merge_ref
         assert merge_response.failure_reason is MergeFailureReason.NONE
 
     @pytest.mark.parametrize('dry_run', [True, False])
     def test_merge_conflict(self, vcsbackend, dry_run):
         self.prepare_for_conflict(vcsbackend)
         expected_merge_response = MergeResponse(
             False, False, None, MergeFailureReason.MERGE_FAILED)
 
         merge_response = self.target_repo.merge(
             self.target_ref, self.source_repo, self.source_ref, self.workspace,
             'test_user', '[e-mail]', 'test message', dry_run=dry_run)
         assert merge_response == expected_merge_response
 
         # We call it twice so to make sure we can handle updates
         merge_response = self.target_repo.merge(
             self.target_ref, self.source_repo, self.source_ref, self.workspace,
             'test_user', '[e-mail]', 'test message', dry_run=dry_run)
         assert merge_response == expected_merge_response
 
     def test_merge_target_is_not_head(self, vcsbackend):
         self.prepare_for_success(vcsbackend)
         expected_merge_response = MergeResponse(
             False, False, None, MergeFailureReason.TARGET_IS_NOT_HEAD)
 
         target_ref = Reference(
             self.target_ref.type, self.target_ref.name, '0' * 40)
 
         merge_response = self.target_repo.merge(
             target_ref, self.source_repo, self.source_ref, self.workspace,
             dry_run=True)
 
         assert merge_response == expected_merge_response
 
-    def test_merge_missing_commit(self, vcsbackend):
+    def test_merge_missing_source_reference(self, vcsbackend):
         self.prepare_for_success(vcsbackend)
         expected_merge_response = MergeResponse(
-            False, False, None, MergeFailureReason.MISSING_COMMIT)
+            False, False, None, MergeFailureReason.MISSING_SOURCE_REF)
 
         source_ref = Reference(
             self.source_ref.type, 'not_existing', self.source_ref.commit_id)
 
         merge_response = self.target_repo.merge(
             self.target_ref, self.source_repo, source_ref, self.workspace,
             dry_run=True)
 
         assert merge_response == expected_merge_response
 
     def test_merge_raises_exception(self, vcsbackend):
         self.prepare_for_success(vcsbackend)
         expected_merge_response = MergeResponse(
             False, False, None, MergeFailureReason.UNKNOWN)
 
         with mock.patch.object(self.target_repo, '_merge_repo',
                                side_effect=RepositoryError()):
             merge_response = self.target_repo.merge(
                 self.target_ref, self.source_repo, self.source_ref,
                 self.workspace, dry_run=True)
 
         assert merge_response == expected_merge_response
 
     def test_merge_invalid_user_name(self, vcsbackend):
         repo = vcsbackend.create_repo(number_of_commits=1)
         ref = Reference('branch', 'master', 'not_used')
         with pytest.raises(ValueError):
             repo.merge(ref, self, ref, 'workspace_id')
 
     def test_merge_invalid_user_email(self, vcsbackend):
         repo = vcsbackend.create_repo(number_of_commits=1)
         ref = Reference('branch', 'master', 'not_used')
         with pytest.raises(ValueError):
             repo.merge(ref, self, ref, 'workspace_id', 'user name')
 
     def test_merge_invalid_message(self, vcsbackend):
         repo = vcsbackend.create_repo(number_of_commits=1)
         ref = Reference('branch', 'master', 'not_used')
         with pytest.raises(ValueError):
             repo.merge(
                 ref, self, ref, 'workspace_id', 'user name', '[e-mail]')
 
 
 class TestRepositoryStrip(BackendTestMixin):
     recreate_repo_per_test = True
 
     @classmethod
     def _get_commits(cls):
         commits = [
             {
                 'message': 'Initial commit',
                 'author': 'Joe Doe <[e-mail]>',
                 'date': datetime.datetime(2010, 1, 1, 20),
                 'branch': 'master',
                 'added': [
                     FileNode('foobar', content='foobar'),
                     FileNode('foobar2', content='foobar2'),
                 ],
             },
         ]
         for x in xrange(10):
             commit_data = {
                 'message': 'Changed foobar - commit%s' % x,
                 'author': 'Jane Doe <[e-mail]>',
                 'date': datetime.datetime(2010, 1, 1, 21, x),
                 'branch': 'master',
                 'changed': [
                     FileNode('foobar', 'FOOBAR - %s' % x),
                 ],
             }
             commits.append(commit_data)
         return commits
 
     @pytest.mark.backends("git", "hg")
     def test_strip_commit(self):
         tip = self.repo.get_commit()
         assert tip.idx == 10
         self.repo.strip(tip.raw_id, self.repo.DEFAULT_BRANCH_NAME)
 
         tip = self.repo.get_commit()
         assert tip.idx == 9
 
     @pytest.mark.backends("git", "hg")
     def test_strip_multiple_commits(self):
         tip = self.repo.get_commit()
         assert tip.idx == 10
 
         old = self.repo.get_commit(commit_idx=5)
         self.repo.strip(old.raw_id, self.repo.DEFAULT_BRANCH_NAME)
 
         tip = self.repo.get_commit()
         assert tip.idx == 4
 
 
 @pytest.mark.backends('hg', 'git')
 class TestRepositoryPull:
 
     def test_pull(self, vcsbackend):
         source_repo = vcsbackend.repo
         target_repo = vcsbackend.create_repo()
         assert len(source_repo.commit_ids) > len(target_repo.commit_ids)
 
         target_repo.pull(source_repo.path)
         # Note: Get a fresh instance, avoids caching trouble
         target_repo = vcsbackend.backend(target_repo.path)
         assert len(source_repo.commit_ids) == len(target_repo.commit_ids)
 
     def test_pull_wrong_path(self, vcsbackend):
         target_repo = vcsbackend.create_repo()
         with pytest.raises(RepositoryError):
             target_repo.pull(target_repo.path + "wrong")
 
     def test_pull_specific_commits(self, vcsbackend):
         source_repo = vcsbackend.repo
         target_repo = vcsbackend.create_repo()
 
         second_commit = source_repo[1].raw_id
         if vcsbackend.alias == 'git':
             second_commit_ref = 'refs/test-refs/a'
             source_repo.set_refs(second_commit_ref, second_commit)
 
         target_repo.pull(source_repo.path, commit_ids=[second_commit])
         target_repo = vcsbackend.backend(target_repo.path)
         assert 2 == len(target_repo.commit_ids)
         assert second_commit == target_repo.get_commit().raw_id