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

# Copyright (C) 2014-2023 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 os
import logging
import binascii
import configparser
import urllib.request
import urllib.parse
import urllib.error

from zope.cachedescriptors.property import Lazy as LazyProperty

from collections import OrderedDict
from rhodecode.lib.datelib import (
    date_to_timestamp_plus_offset, utcdate_fromtimestamp, makedate)
from rhodecode.lib.str_utils import safe_str
from rhodecode.lib.utils2 import CachedProperty
from rhodecode.lib.vcs import connection, exceptions
from rhodecode.lib.vcs.backends.base import (
    BaseRepository, CollectionGenerator, Config, MergeResponse,
    MergeFailureReason, Reference, BasePathPermissionChecker)
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, SubrepoMergeError, UnresolvedFilesInRepo)

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,
                 do_workspace_checkout=False, with_wire=None, bare=False):
        """
        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 do_workspace_checkout=False: sets update of working copy after
           making a clone
        :param bare: not used, compatible with other VCS
        """

        self.path = safe_str(os.path.abspath(repo_path))
        # mercurial since 4.4.X requires certain configuration to be present
        # because sometimes we init the repos with config we need to meet
        # special requirements
        self.config = config if config else self.get_default_config(
            default=[('extensions', 'largefiles', '')])
        self.with_wire = with_wire or {"cache": False}  # default should not use cache

        self._init_repo(create, src_url, do_workspace_checkout)

        # caches
        self._commit_ids = {}

    @LazyProperty
    def _remote(self):
        repo_id = self.path
        return connection.Hg(self.path, repo_id, self.config, with_wire=self.with_wire)

    @CachedProperty
    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 = {commit_id: index
                                for index, commit_id in enumerate(commit_ids)}

    @CachedProperty
    def branches(self):
        return self._get_branches()

    @CachedProperty
    def branches_closed(self):
        return self._get_branches(active=False, closed=True)

    @CachedProperty
    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 = [(n, h,) for n, h in
                     self._remote.branches(active, closed).items()]

        return OrderedDict(sorted(_branches, key=get_name, reverse=False))

    @CachedProperty
    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 = [(n, 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 = f"Added tag {name} for commit {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._invalidate_prop_cache('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._invalidate_prop_cache('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 = [
            (n, 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=False)

        self._remote.invalidate_vcs_cache()
        # clear cache
        self._invalidate_prop_cache('commit_ids')

        return len(self.commit_ids)

    def verify(self):
        verify = self._remote.verify()

        self._remote.invalidate_vcs_cache()
        return verify

    def hg_update_cache(self):
        update_cache = self._remote.hg_update_cache()

        self._remote.invalidate_vcs_cache()
        return update_cache

    def hg_rebuild_fn_cache(self):
        update_cache = self._remote.hg_rebuild_fn_cache()

        self._remote.invalidate_vcs_cache()
        return update_cache

    def get_common_ancestor(self, commit_id1, commit_id2, repo2):
        log.debug('Calculating common ancestor between %sc1:%s and %sc2:%s',
                  self, commit_id1, repo2, commit_id2)

        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)

        ancestor_id = repo2[ancestors[0]].raw_id if ancestors else None

        log.debug('Found common ancestor with sha: %s', ancestor_id)
        return ancestor_id

    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, do_workspace_checkout=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 `do_workspace_checkout` flag.
        """
        if create and os.path.exists(self.path):
            raise RepositoryError(
                f"Cannot create repository at {self.path}, location already exist")

        if src_url:
            url = str(self._get_url(src_url))
            MercurialRepository.check_url(url, self.config)

            self._remote.clone(url, self.path, do_workspace_checkout)

            # Don't try to create if we've already cloned repo
            create = False

        if create:
            os.makedirs(self.path, mode=0o755)

        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_str(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_str(contact or self.DEFAULT_CONTACT)

    @LazyProperty
    def last_change(self):
        """
        Returns last change made on this repository as
        `datetime.datetime` object.
        """
        try:
            return self.get_commit().date
        except RepositoryError:
            tzoffset = makedate()[1]
            return utcdate_fromtimestamp(self._get_fs_mtime(), tzoffset)

    def _get_fs_mtime(self):
        # 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 _get_url(self, url):
        """
        Returns normalized url. If schema is not given, would fall
        to filesystem
        (``file:///``) schema.
        """
        if url != 'default' and '://' not in url:
            url = "file:" + urllib.request.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,
                   translate_tag=None, maybe_unreachable=False, reference_obj=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:
                # we have cached idx, use it without contacting the remote
                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)
            try:
                _commit_id = self.commit_ids[commit_idx]
                if commit_idx < 0:
                    commit_idx = self.commit_ids.index(_commit_id)

                return MercurialCommit(self, _commit_id, commit_idx, pre_load=pre_load)
            except IndexError:
                commit_id = commit_idx
        else:
            commit_id = "tip"

        # case here is no cached version, do an actual lookup instead
        try:
            raw_id, idx = self._remote.lookup(commit_id, both=True)
        except CommitDoesNotExistError:
            msg = "Commit {} does not exist for `{}`".format(
                *map(safe_str, [commit_id, self.name]))
            raise CommitDoesNotExistError(msg)

        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, show_hidden=False, pre_load=None, translate_tags=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
        :param show_hidden: Show hidden commits such as obsolete or hidden from
            Mercurial evolve
        :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
        if self.is_empty():
            raise EmptyRepositoryError("There are no commits yet")
        self._validate_branch_name(branch_name)

        branch_ancestors = False
        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(f'branch("{branch_name}")')
        elif branch_name and branch_ancestors:
            commit_filter.append(f'ancestors(branch("{branch_name}"))')

        if start_date and not end_date:
            commit_filter.append(f'date(">{start_date}")')
        if end_date and not start_date:
            commit_filter.append(f'date("<{end_date}")')
        if start_date and end_date:
            commit_filter.append(
                f'date(">{start_date}") and date("<{end_date}")')

        if not show_hidden:
            commit_filter.append('not obsolete()')
            commit_filter.append('not hidden()')

        # TODO: johbo: Figure out a simpler way for this solution
        collection_generator = CollectionGenerator
        if commit_filter:
            commit_filter = ' and '.join(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):
        """
        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 fetch(self, url, commit_ids=None):
        """
        Backward compatibility with GIT fetch==pull
        """
        return self.pull(url, commit_ids=commit_ids)

    def push(self, url):
        url = self._get_url(url)
        self._remote.sync_push(url)

    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 copy to the specified revision.
        """
        log.debug('Doing checkout to commit: `%s` for %s', revision, self)
        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)

    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, close_commit_id=None, dry_run=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.
        """

        source_ref_commit_id = source_ref.commit_id
        target_ref_commit_id = target_ref.commit_id

        # update our workdir to target ref, for proper merge
        self._update(target_ref_commit_id, clean=True)

        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 close_commit_id:
            # NOTE(marcink): if we get the close commit, this is our new source
            # which will include the close commit itself.
            source_ref_commit_id = close_commit_id

        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

        unresolved = None
        if use_rebase:
            try:
                bookmark_name = f'rcbook{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, clean=True)
                return self._identify(), True
            except RepositoryError as e:
                # 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.')
                if 'unresolved conflicts' in safe_str(e):
                    unresolved = self._remote.get_unresolved_files()
                    log.debug('unresolved files: %s', unresolved)

                # Cleanup any rebase leftovers
                self._remote.invalidate_vcs_cache()
                self._remote.rebase(abort=True)
                self._remote.invalidate_vcs_cache()
                self._remote.update(clean=True)
                if unresolved:
                    raise UnresolvedFilesInRepo(unresolved)
                else:
                    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(f'{user_name} <{user_email}>'))
                self._remote.invalidate_vcs_cache()
                return self._identify(), True
            except RepositoryError as e:
                # The merge-abort may raise another exception which 'hides'
                # the original one, therefore we log it here.
                log.exception('Error while merging shadow repo during merge.')
                if 'unresolved merge conflicts' in safe_str(e):
                    unresolved = self._remote.get_unresolved_files()
                    log.debug('unresolved files: %s', unresolved)

                # Cleanup any merge leftovers
                self._remote.update(clean=True)
                if unresolved:
                    raise UnresolvedFilesInRepo(unresolved)
                else:
                    raise

    def _local_close(self, target_ref, user_name, user_email,
                     source_ref, close_message=''):
        """
        Close the branch of the given source_revision

        Returns the commit id of the close and a boolean indicating if the
        commit needs to be pushed.
        """
        self._update(source_ref.commit_id)
        message = close_message or f"Closing branch: `{source_ref.name}`"
        try:
            self._remote.commit(
                message=safe_str(message),
                username=safe_str(f'{user_name} <{user_email}>'),
                close_branch=True)
            self._remote.invalidate_vcs_cache()
            return self._identify(), True
        except RepositoryError:
            # Cleanup any commit 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 _maybe_prepare_merge_workspace(
            self, repo_id, workspace_id, unused_target_ref, unused_source_ref):
        shadow_repository_path = self._get_shadow_repository_path(
            self.path, repo_id, 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 _merge_repo(self, repo_id, workspace_id, target_ref,
                    source_repo, source_ref, merge_message,
                    merger_name, merger_email, dry_run=False,
                    use_rebase=False, close_branch=False):

        log.debug('Executing merge_repo with %s strategy, dry_run mode:%s',
                  'rebase' if use_rebase else 'merge', dry_run)

        if target_ref.commit_id not in self._heads():
            return MergeResponse(
                False, False, None, MergeFailureReason.TARGET_IS_NOT_HEAD,
                metadata={'target_ref': target_ref})

        try:
            if target_ref.type == 'branch' and len(self._heads(target_ref.name)) != 1:
                heads_all = self._heads(target_ref.name)
                max_heads = 10
                if len(heads_all) > max_heads:
                    heads = '\n,'.join(
                        heads_all[:max_heads] +
                        [f'and {len(heads_all)-max_heads} more.'])
                else:
                    heads = '\n,'.join(heads_all)
                metadata = {
                    'target_ref': target_ref,
                    'source_ref': source_ref,
                    'heads': heads
                }
                return MergeResponse(
                    False, False, None,
                    MergeFailureReason.HG_TARGET_HAS_MULTIPLE_HEADS,
                    metadata=metadata)
        except CommitDoesNotExistError:
            log.exception('Failure when looking up branch heads on hg target')
            return MergeResponse(
                False, False, None, MergeFailureReason.MISSING_TARGET_REF,
                metadata={'target_ref': target_ref})

        shadow_repository_path = self._maybe_prepare_merge_workspace(
            repo_id, workspace_id, target_ref, source_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_SOURCE_REF,
                metadata={'source_ref': source_ref})

        merge_ref = None
        merge_commit_id = None
        close_commit_id = None
        merge_failure_reason = MergeFailureReason.NONE
        metadata = {}

        # enforce that close branch should be used only in case we source from
        # an actual Branch
        close_branch = close_branch and source_ref.type == 'branch'

        # don't allow to close branch if source and target are the same
        close_branch = close_branch and source_ref.name != target_ref.name

        needs_push_on_close = False
        if close_branch and not use_rebase and not dry_run:
            try:
                close_commit_id, needs_push_on_close = shadow_repo._local_close(
                    target_ref, merger_name, merger_email, source_ref)
                merge_possible = True
            except RepositoryError:
                log.exception('Failure when doing close branch on '
                              'shadow repo: %s', shadow_repo)
                merge_possible = False
                merge_failure_reason = MergeFailureReason.MERGE_FAILED
        else:
            merge_possible = True

        needs_push = False
        if merge_possible:

            try:
                merge_commit_id, needs_push = shadow_repo._local_merge(
                    target_ref, merge_message, merger_name, merger_email,
                    source_ref, use_rebase=use_rebase,
                    close_commit_id=close_commit_id, dry_run=dry_run)
                merge_possible = True

                # read the state of the close action, if it
                # maybe required a push
                needs_push = needs_push or needs_push_on_close

                # 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 SubrepoMergeError:
                log.exception(
                    'Subrepo merge error during local merge on hg shadow repo.')
                merge_possible = False
                merge_failure_reason = MergeFailureReason.SUBREPO_MERGE_FAILED
                needs_push = False
            except RepositoryError as e:
                log.exception('Failure when doing local merge on hg shadow repo')
                metadata['unresolved_files'] = 'no unresolved files found'

                if isinstance(e, UnresolvedFilesInRepo):
                    all_conflicts = list(e.args[0])
                    max_conflicts = 20
                    if len(all_conflicts) > max_conflicts:
                        conflicts = all_conflicts[:max_conflicts] \
                                    + [f'and {len(all_conflicts)-max_conflicts} more.']
                    else:
                        conflicts = all_conflicts
                    metadata['unresolved_files'] = \
                        '\n* conflict: ' + \
                        ('\n * conflict: '.join(conflicts))

                merge_possible = False
                merge_failure_reason = MergeFailureReason.MERGE_FAILED
                needs_push = False

        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)
                    # This is the actual merge action, we push from shadow
                    # into origin.
                    # 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)

                    # maybe we also need to push the close_commit_id
                    if close_commit_id:
                        shadow_repo_with_hooks._local_push(
                            close_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 at %s.', self.path)
                    merge_succeeded = False
                    merge_failure_reason = MergeFailureReason.PUSH_FAILED
                    metadata['target'] = 'hg shadow repo'
                    metadata['merge_commit'] = merge_commit_id
            else:
                merge_succeeded = True
        else:
            merge_succeeded = False

        return MergeResponse(
            merge_possible, merge_succeeded, merge_ref, merge_failure_reason,
            metadata=metadata)

    def get_shadow_instance(self, shadow_repository_path, enable_hooks=False, cache=False):
        config = self.config.copy()
        if not enable_hooks:
            config.clear_section('hooks')
        return MercurialRepository(shadow_repository_path, config, with_wire={"cache": cache})

    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, str):
            bookmark = safe_str(bookmark)
        self._remote.bookmark(bookmark, revision=revision)
        self._remote.invalidate_vcs_cache()

    def get_path_permissions(self, username):
        hgacl_file = os.path.join(self.path, '.hg/hgacl')

        def read_patterns(suffix):
            svalue = None
            for section, option in [
                ('narrowacl', username + suffix),
                ('narrowacl', 'default' + suffix),
                ('narrowhgacl', username + suffix),
                ('narrowhgacl', 'default' + suffix)
            ]:
                try:
                    svalue = hgacl.get(section, option)
                    break  # stop at the first value we find
                except configparser.NoOptionError:
                    pass
            if not svalue:
                return None
            result = ['/']
            for pattern in svalue.split():
                result.append(pattern)
                if '*' not in pattern and '?' not in pattern:
                    result.append(pattern + '/*')
            return result

        if os.path.exists(hgacl_file):
            try:
                hgacl = configparser.RawConfigParser()
                hgacl.read(hgacl_file)

                includes = read_patterns('.includes')
                excludes = read_patterns('.excludes')
                return BasePathPermissionChecker.create_from_patterns(
                    includes, excludes)
            except BaseException as e:
                msg = 'Cannot read ACL settings from {} on {}: {}'.format(
                    hgacl_file, self.name, e)
                raise exceptions.RepositoryRequirementError(msg)
        else:
            return None


class MercurialIndexBasedCollectionGenerator(CollectionGenerator):

    def _commit_factory(self, commit_id):
        if isinstance(commit_id, int):
            return self.repo.get_commit(
                commit_idx=commit_id, pre_load=self.pre_load)
        else:
            return self.repo.get_commit(
                commit_id=commit_id, pre_load=self.pre_load)