fix(caches): actually obey wire cache flag on local method caching

rhodecode/lib/vcs/client_http.py

@@ -1,430 +1,433 b''
 # Copyright (C) 2016-2023 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/
 
 """
 Client for the VCSServer implemented based on HTTP.
 """
 
 import copy
 import logging
 import threading
 import time
 import urllib.request
 import urllib.error
 import urllib.parse
 import urllib.parse
 import uuid
 import traceback
 
 import pycurl
 import msgpack
 import requests
 from requests.packages.urllib3.util.retry import Retry
 
 import rhodecode
 from rhodecode.lib import rc_cache
 from rhodecode.lib.rc_cache.utils import compute_key_from_params
 from rhodecode.lib.system_info import get_cert_path
 from rhodecode.lib.vcs import exceptions, CurlSession
 from rhodecode.lib.utils2 import str2bool
 
 log = logging.getLogger(__name__)
 
 
 # TODO: mikhail: Keep it in sync with vcsserver's
 # HTTPApplication.ALLOWED_EXCEPTIONS
 EXCEPTIONS_MAP = {
     'KeyError': KeyError,
     'URLError': urllib.error.URLError,
 }
 
 
 def _remote_call(url, payload, exceptions_map, session, retries=3):
 
     for attempt in range(retries):
         try:
             response = session.post(url, data=msgpack.packb(payload))
             break
         except pycurl.error as e:
             error_code, error_message = e.args
             if error_code == pycurl.E_RECV_ERROR:
                 log.warning(f'Received a "Connection reset by peer" error. '
                             f'Retrying... ({attempt + 1}/{retries})')
                 continue  # Retry if connection reset error.
             msg = f'{e}. \npycurl traceback: {traceback.format_exc()}'
             raise exceptions.HttpVCSCommunicationError(msg)
         except Exception as e:
             message = getattr(e, 'message', '')
             if 'Failed to connect' in message:
                 # gevent doesn't return proper pycurl errors
                 raise exceptions.HttpVCSCommunicationError(e)
             else:
                 raise
 
     if response.status_code >= 400:
         content_type = response.content_type
         log.error('Call to %s returned non 200 HTTP code: %s [%s]',
                   url, response.status_code, content_type)
         raise exceptions.HttpVCSCommunicationError(repr(response.content))
 
     try:
         response = msgpack.unpackb(response.content)
     except Exception:
         log.exception('Failed to decode response from msgpack')
         raise
 
     error = response.get('error')
     if error:
         type_ = error.get('type', 'Exception')
         exc = exceptions_map.get(type_, Exception)
         exc = exc(error.get('message'))
         try:
             exc._vcs_kind = error['_vcs_kind']
         except KeyError:
             pass
 
         try:
             exc._vcs_server_traceback = error['traceback']
             exc._vcs_server_org_exc_name = error['org_exc']
             exc._vcs_server_org_exc_tb = error['org_exc_tb']
         except KeyError:
             pass
 
         exc.add_note(attach_exc_details(error))
         raise exc  # raising the org exception from vcsserver
     return response.get('result')
 
 
 def attach_exc_details(error):
     note = '-- EXC NOTE -- :\n'
     note += f'vcs_kind: {error.get("_vcs_kind")}\n'
     note += f'org_exc: {error.get("_vcs_kind")}\n'
     note += f'tb: {error.get("traceback")}\n'
     note += '-- END EXC NOTE --'
     return note
 
 
 def _streaming_remote_call(url, payload, exceptions_map, session, chunk_size):
     try:
         headers = {
             'X-RC-Method': payload.get('method'),
             'X-RC-Repo-Name': payload.get('_repo_name')
         }
         response = session.post(url, data=msgpack.packb(payload), headers=headers)
     except pycurl.error as e:
         error_code, error_message = e.args
         msg = f'{e}. \npycurl traceback: {traceback.format_exc()}'
         raise exceptions.HttpVCSCommunicationError(msg)
     except Exception as e:
         message = getattr(e, 'message', '')
         if 'Failed to connect' in message:
             # gevent doesn't return proper pycurl errors
             raise exceptions.HttpVCSCommunicationError(e)
         else:
             raise
 
     if response.status_code >= 400:
         log.error('Call to %s returned non 200 HTTP code: %s',
                   url, response.status_code)
         raise exceptions.HttpVCSCommunicationError(repr(response.content))
 
     return response.iter_content(chunk_size=chunk_size)
 
 
 class ServiceConnection(object):
     def __init__(self, server_and_port, backend_endpoint, session_factory):
         self.url = urllib.parse.urljoin(f'http://{server_and_port}', backend_endpoint)
         self._session_factory = session_factory
 
     def __getattr__(self, name):
         def f(*args, **kwargs):
             return self._call(name, *args, **kwargs)
         return f
 
     @exceptions.map_vcs_exceptions
     def _call(self, name, *args, **kwargs):
         payload = {
             'id': str(uuid.uuid4()),
             'method': name,
             'params': {'args': args, 'kwargs': kwargs}
         }
         return _remote_call(
             self.url, payload, EXCEPTIONS_MAP, self._session_factory())
 
 
 class RemoteVCSMaker(object):
 
     def __init__(self, server_and_port, backend_endpoint, backend_type, session_factory):
         self.url = urllib.parse.urljoin(f'http://{server_and_port}', backend_endpoint)
         self.stream_url = urllib.parse.urljoin(f'http://{server_and_port}', backend_endpoint+'/stream')
 
         self._session_factory = session_factory
         self.backend_type = backend_type
 
     @classmethod
     def init_cache_region(cls, repo_id):
         cache_namespace_uid = f'repo.{repo_id}'
         region = rc_cache.get_or_create_region('cache_repo', cache_namespace_uid)
         return region, cache_namespace_uid
 
     def __call__(self, path, repo_id, config, with_wire=None):
         log.debug('%s RepoMaker call on %s', self.backend_type.upper(), path)
         return RemoteRepo(path, repo_id, config, self, with_wire=with_wire)
 
     def __getattr__(self, name):
         def remote_attr(*args, **kwargs):
             return self._call(name, *args, **kwargs)
         return remote_attr
 
     @exceptions.map_vcs_exceptions
     def _call(self, func_name, *args, **kwargs):
         payload = {
             'id': str(uuid.uuid4()),
             'method': func_name,
             'backend': self.backend_type,
             'params': {'args': args, 'kwargs': kwargs}
         }
         url = self.url
         return _remote_call(url, payload, EXCEPTIONS_MAP, self._session_factory())
 
 
 class RemoteRepo(object):
     CHUNK_SIZE = 16384
 
     def __init__(self, path, repo_id, config, remote_maker, with_wire=None):
         self.url = remote_maker.url
         self.stream_url = remote_maker.stream_url
         self._session = remote_maker._session_factory()
 
         cache_repo_id = self._repo_id_sanitizer(repo_id)
         _repo_name = self._get_repo_name(config, path)
         self._cache_region, self._cache_namespace = \
             remote_maker.init_cache_region(cache_repo_id)
 
         with_wire = with_wire or {}
 
         repo_state_uid = with_wire.get('repo_state_uid') or 'state'
 
         self._wire = {
             "_repo_name": _repo_name,
             "path": path,  # repo path
             "repo_id": repo_id,
             "cache_repo_id": cache_repo_id,
             "config": config,
             "repo_state_uid": repo_state_uid,
             "context": self._create_vcs_cache_context(path, repo_state_uid)
         }
 
         if with_wire:
             self._wire.update(with_wire)
 
         # NOTE(johbo): Trading complexity for performance. Avoiding the call to
         # log.debug brings a few percent gain even if is is not active.
         if log.isEnabledFor(logging.DEBUG):
             self._call_with_logging = True
 
         self.cert_dir = get_cert_path(rhodecode.CONFIG.get('__file__'))
 
     def _get_repo_name(self, config, path):
         repo_store = config.get('paths', '/')
         return path.split(repo_store)[-1].lstrip('/')
 
     def _repo_id_sanitizer(self, repo_id):
         pathless = repo_id.replace('/', '__').replace('-', '_')
         return ''.join(char if ord(char) < 128 else '_{}_'.format(ord(char)) for char in pathless)
 
     def __getattr__(self, name):
 
         if name.startswith('stream:'):
             def repo_remote_attr(*args, **kwargs):
                 return self._call_stream(name, *args, **kwargs)
         else:
             def repo_remote_attr(*args, **kwargs):
                 return self._call(name, *args, **kwargs)
 
         return repo_remote_attr
 
     def _base_call(self, name, *args, **kwargs):
         # TODO: oliver: This is currently necessary pre-call since the
         # config object is being changed for hooking scenarios
         wire = copy.deepcopy(self._wire)
         wire["config"] = wire["config"].serialize()
         wire["config"].append(('vcs', 'ssl_dir', self.cert_dir))
 
         payload = {
             'id': str(uuid.uuid4()),
             'method': name,
             "_repo_name":  wire['_repo_name'],
             'params': {'wire': wire, 'args': args, 'kwargs': kwargs}
         }
 
         context_uid = wire.get('context')
         return context_uid, payload
 
     def get_local_cache(self, name, args):
         cache_on = False
         cache_key = ''
         local_cache_on = rhodecode.ConfigGet().get_bool('vcs.methods.cache')
 
         cache_methods = [
             'branches', 'tags', 'bookmarks',
             'is_large_file', 'is_binary',
             'fctx_size', 'stream:fctx_node_data', 'blob_raw_length',
             'node_history',
             'revision', 'tree_items',
             'ctx_list', 'ctx_branch', 'ctx_description',
             'bulk_request',
             'assert_correct_path',
             'is_path_valid_repository',
         ]
 
-        if local_cache_on and name in cache_methods:
+        wire_cache = self._wire['cache']
+
+        if local_cache_on and wire_cache and name in cache_methods:
             cache_on = True
             repo_state_uid = self._wire['repo_state_uid']
             call_args = [a for a in args]
             cache_key = compute_key_from_params(repo_state_uid, name, *call_args)
 
         return cache_on, cache_key
 
     @exceptions.map_vcs_exceptions
     def _call(self, name, *args, **kwargs):
+
         context_uid, payload = self._base_call(name, *args, **kwargs)
         url = self.url
 
         start = time.time()
         cache_on, cache_key = self.get_local_cache(name, args)
 
         @self._cache_region.conditional_cache_on_arguments(
             namespace=self._cache_namespace, condition=cache_on and cache_key)
         def remote_call(_cache_key):
             if self._call_with_logging:
                 args_repr = f'ARG: {str(args):.512}|KW: {str(kwargs):.512}'
                 log.debug('Calling %s@%s with args:%r. wire_context: %s cache_on: %s',
                           url, name, args_repr, context_uid, cache_on)
             return _remote_call(url, payload, EXCEPTIONS_MAP, self._session)
 
         result = remote_call(cache_key)
         if self._call_with_logging:
             log.debug('Call %s@%s took: %.4fs. wire_context: %s',
                       url, name, time.time()-start, context_uid)
         return result
 
     @exceptions.map_vcs_exceptions
     def _call_stream(self, name, *args, **kwargs):
         context_uid, payload = self._base_call(name, *args, **kwargs)
         payload['chunk_size'] = self.CHUNK_SIZE
         url = self.stream_url
 
         start = time.time()
         cache_on, cache_key = self.get_local_cache(name, args)
 
         # Cache is a problem because this is a stream
         def streaming_remote_call(_cache_key):
             if self._call_with_logging:
                 args_repr = f'ARG: {str(args):.512}|KW: {str(kwargs):.512}'
                 log.debug('Calling %s@%s with args:%r. wire_context: %s cache_on: %s',
                           url, name, args_repr, context_uid, cache_on)
             return _streaming_remote_call(url, payload, EXCEPTIONS_MAP, self._session, self.CHUNK_SIZE)
 
         result = streaming_remote_call(cache_key)
         if self._call_with_logging:
             log.debug('Call %s@%s took: %.4fs. wire_context: %s',
                       url, name, time.time()-start, context_uid)
         return result
 
     def __getitem__(self, key):
         return self.revision(key)
 
     def _create_vcs_cache_context(self, *args):
         """
         Creates a unique string which is passed to the VCSServer on every
         remote call. It is used as cache key in the VCSServer.
         """
         hash_key = '-'.join(map(str, args))
         return str(uuid.uuid5(uuid.NAMESPACE_URL, hash_key))
 
     def invalidate_vcs_cache(self):
         """
         This invalidates the context which is sent to the VCSServer on every
         call to a remote method. It forces the VCSServer to create a fresh
         repository instance on the next call to a remote method.
         """
         self._wire['context'] = str(uuid.uuid4())
 
 
 class VcsHttpProxy(object):
 
     CHUNK_SIZE = 16384
 
     def __init__(self, server_and_port, backend_endpoint):
         retries = Retry(total=5, connect=None, read=None, redirect=None)
 
         adapter = requests.adapters.HTTPAdapter(max_retries=retries)
         self.base_url = urllib.parse.urljoin('http://%s' % server_and_port, backend_endpoint)
         self.session = requests.Session()
         self.session.mount('http://', adapter)
 
     def handle(self, environment, input_data, *args, **kwargs):
         data = {
             'environment': environment,
             'input_data': input_data,
             'args': args,
             'kwargs': kwargs
         }
         result = self.session.post(
             self.base_url, msgpack.packb(data), stream=True)
         return self._get_result(result)
 
     def _deserialize_and_raise(self, error):
         exception = Exception(error['message'])
         try:
             exception._vcs_kind = error['_vcs_kind']
         except KeyError:
             pass
         raise exception
 
     def _iterate(self, result):
         unpacker = msgpack.Unpacker()
         for line in result.iter_content(chunk_size=self.CHUNK_SIZE):
             unpacker.feed(line)
             yield from unpacker
 
     def _get_result(self, result):
         iterator = self._iterate(result)
         error = next(iterator)
         if error:
             self._deserialize_and_raise(error)
 
         status = next(iterator)
         headers = next(iterator)
 
         return iterator, status, headers
 
 
 class ThreadlocalSessionFactory(object):
     """
     Creates one CurlSession per thread on demand.
     """
 
     def __init__(self):
         self._thread_local = threading.local()
 
     def __call__(self):
         if not hasattr(self._thread_local, 'curl_session'):
             self._thread_local.curl_session = CurlSession()
         return self._thread_local.curl_session

rhodecode/model/scm.py

@@ -1,1053 +1,1051 b''
 # Copyright (C) 2010-2023 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/
 
 """
 Scm model for RhodeCode
 """
 
 import os.path
 import traceback
 import logging
 import io
 
 from sqlalchemy import func
 from zope.cachedescriptors.property import Lazy as LazyProperty
 
 import rhodecode
 from rhodecode.lib.str_utils import safe_bytes
 from rhodecode.lib.vcs import get_backend
 from rhodecode.lib.vcs.exceptions import RepositoryError, NodeNotChangedError
 from rhodecode.lib.vcs.nodes import FileNode
 from rhodecode.lib.vcs.backends.base import EmptyCommit
 from rhodecode.lib import helpers as h, rc_cache
 from rhodecode.lib.auth import (
     HasRepoPermissionAny, HasRepoGroupPermissionAny,
     HasUserGroupPermissionAny)
 from rhodecode.lib.exceptions import NonRelativePathError, IMCCommitError
 from rhodecode.lib import hooks_utils
 from rhodecode.lib.utils import (
     get_filesystem_repos, make_db_config)
 from rhodecode.lib.str_utils import safe_str
 from rhodecode.lib.system_info import get_system_info
 from rhodecode.model import BaseModel
 from rhodecode.model.db import (
     or_, false, null,
     Repository, CacheKey, UserFollowing, UserLog, User, RepoGroup,
     PullRequest, FileStore)
 from rhodecode.model.settings import VcsSettingsModel
 from rhodecode.model.validation_schema.validators import url_validator, InvalidCloneUrl
 
 log = logging.getLogger(__name__)
 
 
 class UserTemp(object):
     def __init__(self, user_id):
         self.user_id = user_id
 
     def __repr__(self):
         return "<{}('id:{}')>".format(self.__class__.__name__, self.user_id)
 
 
 class RepoTemp(object):
     def __init__(self, repo_id):
         self.repo_id = repo_id
 
     def __repr__(self):
         return "<{}('id:{}')>".format(self.__class__.__name__, self.repo_id)
 
 
 class SimpleCachedRepoList(object):
     """
     Lighter version of of iteration of repos without the scm initialisation,
     and with cache usage
     """
     def __init__(self, db_repo_list, repos_path, order_by=None, perm_set=None):
         self.db_repo_list = db_repo_list
         self.repos_path = repos_path
         self.order_by = order_by
         self.reversed = (order_by or '').startswith('-')
         if not perm_set:
             perm_set = ['repository.read', 'repository.write',
                         'repository.admin']
         self.perm_set = perm_set
 
     def __len__(self):
         return len(self.db_repo_list)
 
     def __repr__(self):
         return '<{} ({})>'.format(self.__class__.__name__, self.__len__())
 
     def __iter__(self):
         for dbr in self.db_repo_list:
             # check permission at this level
             has_perm = HasRepoPermissionAny(*self.perm_set)(
                 dbr.repo_name, 'SimpleCachedRepoList check')
             if not has_perm:
                 continue
 
             tmp_d = {
                 'name': dbr.repo_name,
                 'dbrepo': dbr.get_dict(),
                 'dbrepo_fork': dbr.fork.get_dict() if dbr.fork else {}
             }
             yield tmp_d
 
 
 class _PermCheckIterator(object):
 
     def __init__(
             self, obj_list, obj_attr, perm_set, perm_checker,
             extra_kwargs=None):
         """
         Creates iterator from given list of objects, additionally
         checking permission for them from perm_set var
 
         :param obj_list: list of db objects
         :param obj_attr: attribute of object to pass into perm_checker
         :param perm_set: list of permissions to check
         :param perm_checker: callable to check permissions against
         """
         self.obj_list = obj_list
         self.obj_attr = obj_attr
         self.perm_set = perm_set
         self.perm_checker = perm_checker(*self.perm_set)
         self.extra_kwargs = extra_kwargs or {}
 
     def __len__(self):
         return len(self.obj_list)
 
     def __repr__(self):
         return '<{} ({})>'.format(self.__class__.__name__, self.__len__())
 
     def __iter__(self):
         for db_obj in self.obj_list:
             # check permission at this level
             # NOTE(marcink): the __dict__.get() is ~4x faster then getattr()
             name = db_obj.__dict__.get(self.obj_attr, None)
             if not self.perm_checker(name, self.__class__.__name__, **self.extra_kwargs):
                 continue
 
             yield db_obj
 
 
 class RepoList(_PermCheckIterator):
 
     def __init__(self, db_repo_list, perm_set=None, extra_kwargs=None):
         if not perm_set:
             perm_set = ['repository.read', 'repository.write', 'repository.admin']
 
         super().__init__(
             obj_list=db_repo_list,
             obj_attr='_repo_name', perm_set=perm_set,
             perm_checker=HasRepoPermissionAny,
             extra_kwargs=extra_kwargs)
 
 
 class RepoGroupList(_PermCheckIterator):
 
     def __init__(self, db_repo_group_list, perm_set=None, extra_kwargs=None):
         if not perm_set:
             perm_set = ['group.read', 'group.write', 'group.admin']
 
         super().__init__(
             obj_list=db_repo_group_list,
             obj_attr='_group_name', perm_set=perm_set,
             perm_checker=HasRepoGroupPermissionAny,
             extra_kwargs=extra_kwargs)
 
 
 class UserGroupList(_PermCheckIterator):
 
     def __init__(self, db_user_group_list, perm_set=None, extra_kwargs=None):
         if not perm_set:
             perm_set = ['usergroup.read', 'usergroup.write', 'usergroup.admin']
 
         super().__init__(
             obj_list=db_user_group_list,
             obj_attr='users_group_name', perm_set=perm_set,
             perm_checker=HasUserGroupPermissionAny,
             extra_kwargs=extra_kwargs)
 
 
 class ScmModel(BaseModel):
     """
     Generic Scm Model
     """
 
     def repo_scan(self, repos_path=None):
         """
         Listing of repositories in given path. This path should not be a
         repository itself. Return a dictionary of repository objects
 
         :param repos_path: path to directory containing repositories
         """
 
         if repos_path is None:
             repos_path = self.repos_path
 
         log.info('scanning for repositories in %s', repos_path)
 
         config = make_db_config()
         config.set('extensions', 'largefiles', '')
         repos = {}
 
         for name, path in get_filesystem_repos(repos_path, recursive=True):
             # name need to be decomposed and put back together using the /
             # since this is internal storage separator for rhodecode
             name = Repository.normalize_repo_name(name)
 
             try:
                 if name in repos:
-                    raise RepositoryError('Duplicate repository name %s '
-                                          'found in %s' % (name, path))
+                    raise RepositoryError(f'Duplicate repository name {name} found in {path}')
                 elif path[0] in rhodecode.BACKENDS:
                     backend = get_backend(path[0])
-                    repos[name] = backend(path[1], config=config,
-                                          with_wire={"cache": False})
+                    repos[name] = backend(path[1], config=config, with_wire={"cache": False})
             except OSError:
                 continue
             except RepositoryError:
                 log.exception('Failed to create a repo')
                 continue
 
         log.debug('found %s paths with repositories', len(repos))
         return repos
 
     def get_repos(self, all_repos=None, sort_key=None):
         """
         Get all repositories from db and for each repo create it's
         backend instance and fill that backed with information from database
 
         :param all_repos: list of repository names as strings
             give specific repositories list, good for filtering
 
         :param sort_key: initial sorting of repositories
         """
         if all_repos is None:
             all_repos = self.sa.query(Repository)\
                 .filter(Repository.group_id == null())\
                 .order_by(func.lower(Repository.repo_name)).all()
         repo_iter = SimpleCachedRepoList(
             all_repos, repos_path=self.repos_path, order_by=sort_key)
         return repo_iter
 
     @staticmethod
     def get_parent_commits(parent_commit, scm_instance):
         if not parent_commit:
             parent_commit = EmptyCommit(alias=scm_instance.alias)
 
         if isinstance(parent_commit, EmptyCommit):
             # EmptyCommit means we're editing empty repository
             parents = None
         else:
             parents = [parent_commit]
         return parent_commit, parents
 
     def initialize_inmemory_vars(self, user, repo, message, author):
         """
         Initialize node specific objects for further usage
         """
         user = self._get_user(user)
         scm_instance = repo.scm_instance(cache=False)
         message = safe_str(message)
         commiter = user.full_contact
         author = safe_str(author) if author else commiter
         imc = scm_instance.in_memory_commit
 
         return user, scm_instance, message, commiter, author, imc
 
     def get_repo_groups(self, all_groups=None):
         if all_groups is None:
             all_groups = RepoGroup.query()\
                 .filter(RepoGroup.group_parent_id == null()).all()
         return [x for x in RepoGroupList(all_groups)]
 
     def mark_for_invalidation(self, repo_name, delete=False):
         """
         Mark caches of this repo invalid in the database. `delete` flag
         removes the cache entries
 
         :param repo_name: the repo_name for which caches should be marked
             invalid, or deleted
         :param delete: delete the entry keys instead of setting bool
             flag on them, and also purge caches used by the dogpile
         """
         repo = Repository.get_by_repo_name(repo_name)
 
         if repo:
             repo_namespace_key = CacheKey.REPO_INVALIDATION_NAMESPACE.format(repo_id=repo.repo_id)
             CacheKey.set_invalidate(repo_namespace_key, delete=delete)
 
             repo_id = repo.repo_id
             config = repo._config
             config.set('extensions', 'largefiles', '')
             repo.update_commit_cache(config=config, cs_cache=None)
             if delete:
                 cache_namespace_uid = f'cache_repo.{repo_id}'
                 rc_cache.clear_cache_namespace('cache_repo', cache_namespace_uid, method=rc_cache.CLEAR_INVALIDATE)
 
     def toggle_following_repo(self, follow_repo_id, user_id):
 
         f = self.sa.query(UserFollowing)\
             .filter(UserFollowing.follows_repo_id == follow_repo_id)\
             .filter(UserFollowing.user_id == user_id).scalar()
 
         if f is not None:
             try:
                 self.sa.delete(f)
                 return
             except Exception:
                 log.error(traceback.format_exc())
                 raise
 
         try:
             f = UserFollowing()
             f.user_id = user_id
             f.follows_repo_id = follow_repo_id
             self.sa.add(f)
         except Exception:
             log.error(traceback.format_exc())
             raise
 
     def toggle_following_user(self, follow_user_id, user_id):
         f = self.sa.query(UserFollowing)\
             .filter(UserFollowing.follows_user_id == follow_user_id)\
             .filter(UserFollowing.user_id == user_id).scalar()
 
         if f is not None:
             try:
                 self.sa.delete(f)
                 return
             except Exception:
                 log.error(traceback.format_exc())
                 raise
 
         try:
             f = UserFollowing()
             f.user_id = user_id
             f.follows_user_id = follow_user_id
             self.sa.add(f)
         except Exception:
             log.error(traceback.format_exc())
             raise
 
     def is_following_repo(self, repo_name, user_id, cache=False):
         r = self.sa.query(Repository)\
             .filter(Repository.repo_name == repo_name).scalar()
 
         f = self.sa.query(UserFollowing)\
             .filter(UserFollowing.follows_repository == r)\
             .filter(UserFollowing.user_id == user_id).scalar()
 
         return f is not None
 
     def is_following_user(self, username, user_id, cache=False):
         u = User.get_by_username(username)
 
         f = self.sa.query(UserFollowing)\
             .filter(UserFollowing.follows_user == u)\
             .filter(UserFollowing.user_id == user_id).scalar()
 
         return f is not None
 
     def get_followers(self, repo):
         repo = self._get_repo(repo)
 
         return self.sa.query(UserFollowing)\
             .filter(UserFollowing.follows_repository == repo).count()
 
     def get_forks(self, repo):
         repo = self._get_repo(repo)
         return self.sa.query(Repository)\
             .filter(Repository.fork == repo).count()
 
     def get_pull_requests(self, repo):
         repo = self._get_repo(repo)
         return self.sa.query(PullRequest)\
             .filter(PullRequest.target_repo == repo)\
             .filter(PullRequest.status != PullRequest.STATUS_CLOSED).count()
 
     def get_artifacts(self, repo):
         repo = self._get_repo(repo)
         return self.sa.query(FileStore)\
             .filter(FileStore.repo == repo)\
             .filter(or_(FileStore.hidden == null(), FileStore.hidden == false())).count()
 
     def mark_as_fork(self, repo, fork, user):
         repo = self._get_repo(repo)
         fork = self._get_repo(fork)
         if fork and repo.repo_id == fork.repo_id:
             raise Exception("Cannot set repository as fork of itself")
 
         if fork and repo.repo_type != fork.repo_type:
             raise RepositoryError(
                 "Cannot set repository as fork of repository with other type")
 
         repo.fork = fork
         self.sa.add(repo)
         return repo
 
     def pull_changes(self, repo, username, remote_uri=None, validate_uri=True, **kwargs):
         dbrepo = self._get_repo(repo)
         remote_uri = remote_uri or dbrepo.clone_uri
         if not remote_uri:
             raise Exception("This repository doesn't have a clone uri")
 
         repo = dbrepo.scm_instance(cache=False)
         repo.config.clear_section('hooks')
 
         try:
             # NOTE(marcink): add extra validation so we skip invalid urls
             # this is due this tasks can be executed via scheduler without
             # proper validation of remote_uri
             if validate_uri:
                 config = make_db_config(clear_session=False)
                 url_validator(remote_uri, dbrepo.repo_type, config)
         except InvalidCloneUrl:
             raise
 
         repo_name = dbrepo.repo_name
         try:
             # TODO: we need to make sure those operations call proper hooks !
             repo.fetch(remote_uri, **kwargs)
 
             self.mark_for_invalidation(repo_name)
         except Exception:
             log.error(traceback.format_exc())
             raise
 
     def push_changes(self, repo, username, remote_uri=None, validate_uri=True, **kwargs):
         dbrepo = self._get_repo(repo)
         remote_uri = remote_uri or dbrepo.push_uri
         if not remote_uri:
             raise Exception("This repository doesn't have a clone uri")
 
         repo = dbrepo.scm_instance(cache=False)
         repo.config.clear_section('hooks')
 
         try:
             # NOTE(marcink): add extra validation so we skip invalid urls
             # this is due this tasks can be executed via scheduler without
             # proper validation of remote_uri
             if validate_uri:
                 config = make_db_config(clear_session=False)
                 url_validator(remote_uri, dbrepo.repo_type, config)
         except InvalidCloneUrl:
             raise
 
         try:
             repo.push(remote_uri, **kwargs)
         except Exception:
             log.error(traceback.format_exc())
             raise
 
     def commit_change(self, repo, repo_name, commit, user, author, message,
                       content: bytes, f_path: bytes, branch: str = None):
         """
         Commits changes
         """
         user = self._get_user(user)
 
         # message and author needs to be unicode
         # proper backend should then translate that into required type
         message = safe_str(message)
         author = safe_str(author)
         imc = repo.in_memory_commit
         imc.change(FileNode(f_path, content, mode=commit.get_file_mode(f_path)))
         try:
             # TODO: handle pre-push action !
             tip = imc.commit(
                 message=message, author=author, parents=[commit],
                 branch=branch or commit.branch)
         except Exception as e:
             log.error(traceback.format_exc())
             raise IMCCommitError(str(e))
         finally:
             # always clear caches, if commit fails we want fresh object also
             self.mark_for_invalidation(repo_name)
 
         # We trigger the post-push action
         hooks_utils.trigger_post_push_hook(
             username=user.username, action='push_local', hook_type='post_push',
             repo_name=repo_name, repo_type=repo.alias, commit_ids=[tip.raw_id])
         return tip
 
     def _sanitize_path(self, f_path: bytes):
         if f_path.startswith(b'/') or f_path.startswith(b'./') or b'../' in f_path:
             raise NonRelativePathError(b'%b is not an relative path' % f_path)
         if f_path:
             f_path = os.path.normpath(f_path)
         return f_path
 
     def get_dirnode_metadata(self, request, commit, dir_node):
         if not dir_node.is_dir():
             return []
 
         data = []
         for node in dir_node:
             if not node.is_file():
                 # we skip file-nodes
                 continue
 
             last_commit = node.last_commit
             last_commit_date = last_commit.date
             data.append({
                 'name': node.name,
                 'size': h.format_byte_size_binary(node.size),
                 'modified_at': h.format_date(last_commit_date),
                 'modified_ts': last_commit_date.isoformat(),
                 'revision': last_commit.revision,
                 'short_id': last_commit.short_id,
                 'message': h.escape(last_commit.message),
                 'author': h.escape(last_commit.author),
                 'user_profile': h.gravatar_with_user(
                     request, last_commit.author),
             })
 
         return data
 
     def get_nodes(self, repo_name, commit_id, root_path='/', flat=True,
                   extended_info=False, content=False, max_file_bytes=None):
         """
         recursive walk in root dir and return a set of all path in that dir
         based on repository walk function
 
         :param repo_name: name of repository
         :param commit_id: commit id for which to list nodes
         :param root_path: root path to list
         :param flat: return as a list, if False returns a dict with description
         :param extended_info: show additional info such as md5, binary, size etc
         :param content: add nodes content to the return data
         :param max_file_bytes: will not return file contents over this limit
 
         """
         _files = list()
         _dirs = list()
 
         try:
             _repo = self._get_repo(repo_name)
             commit = _repo.scm_instance().get_commit(commit_id=commit_id)
             root_path = root_path.lstrip('/')
 
             # get RootNode, inject pre-load options before walking
             top_node = commit.get_node(root_path)
             extended_info_pre_load = []
             if extended_info:
                 extended_info_pre_load += ['md5']
             top_node.default_pre_load = ['is_binary', 'size'] + extended_info_pre_load
 
             for __, dirs, files in commit.walk(top_node):
 
                 for f in files:
                     _content = None
                     _data = f_name = f.str_path
 
                     if not flat:
                         _data = {
                             "name": h.escape(f_name),
                             "type": "file",
                             }
                         if extended_info:
                             _data.update({
                                 "md5": f.md5,
                                 "binary": f.is_binary,
                                 "size": f.size,
                                 "extension": f.extension,
                                 "mimetype": f.mimetype,
                                 "lines": f.lines()[0]
                             })
 
                         if content:
                             over_size_limit = (max_file_bytes is not None
                                                and f.size > max_file_bytes)
                             full_content = None
                             if not f.is_binary and not over_size_limit:
                                 full_content = f.str_content
 
                             _data.update({
                                 "content": full_content,
                             })
                     _files.append(_data)
 
                 for d in dirs:
                     _data = d_name = d.str_path
                     if not flat:
                         _data = {
                             "name": h.escape(d_name),
                             "type": "dir",
                             }
                     if extended_info:
                         _data.update({
                             "md5": "",
                             "binary": False,
                             "size": 0,
                             "extension": "",
                         })
                     if content:
                         _data.update({
                             "content": None
                         })
                     _dirs.append(_data)
         except RepositoryError:
             log.exception("Exception in get_nodes")
             raise
 
         return _dirs, _files
 
     def get_quick_filter_nodes(self, repo_name, commit_id, root_path='/'):
         """
         Generate files for quick filter in files view
         """
 
         _files = list()
         _dirs = list()
         try:
             _repo = self._get_repo(repo_name)
             commit = _repo.scm_instance().get_commit(commit_id=commit_id)
             root_path = root_path.lstrip('/')
 
             top_node = commit.get_node(root_path)
             top_node.default_pre_load = []
 
             for __, dirs, files in commit.walk(top_node):
                 for f in files:
 
                     _data = {
                         "name": h.escape(f.str_path),
                         "type": "file",
                         }
 
                     _files.append(_data)
 
                 for d in dirs:
 
                     _data = {
                         "name": h.escape(d.str_path),
                         "type": "dir",
                         }
 
                     _dirs.append(_data)
         except RepositoryError:
             log.exception("Exception in get_quick_filter_nodes")
             raise
 
         return _dirs, _files
 
     def get_node(self, repo_name, commit_id, file_path,
                  extended_info=False, content=False, max_file_bytes=None, cache=True):
         """
         retrieve single node from commit
         """
 
         try:
 
             _repo = self._get_repo(repo_name)
             commit = _repo.scm_instance().get_commit(commit_id=commit_id)
 
             file_node = commit.get_node(file_path)
             if file_node.is_dir():
                 raise RepositoryError('The given path is a directory')
 
             _content = None
             f_name = file_node.str_path
 
             file_data = {
                 "name": h.escape(f_name),
                 "type": "file",
             }
 
             if extended_info:
                 file_data.update({
                     "extension": file_node.extension,
                     "mimetype": file_node.mimetype,
                 })
 
                 if cache:
                     md5 = file_node.md5
                     is_binary = file_node.is_binary
                     size = file_node.size
                 else:
                     is_binary, md5, size, _content = file_node.metadata_uncached()
 
                 file_data.update({
                     "md5": md5,
                     "binary": is_binary,
                     "size": size,
                 })
 
             if content and cache:
                 # get content + cache
                 size = file_node.size
                 over_size_limit = (max_file_bytes is not None and size > max_file_bytes)
                 full_content = None
                 all_lines = 0
                 if not file_node.is_binary and not over_size_limit:
                     full_content = safe_str(file_node.content)
                     all_lines, empty_lines = file_node.count_lines(full_content)
 
                 file_data.update({
                     "content": full_content,
                     "lines": all_lines
                 })
             elif content:
                 # get content *without* cache
                 if _content is None:
                     is_binary, md5, size, _content = file_node.metadata_uncached()
 
                 over_size_limit = (max_file_bytes is not None and size > max_file_bytes)
                 full_content = None
                 all_lines = 0
                 if not is_binary and not over_size_limit:
                     full_content = safe_str(_content)
                     all_lines, empty_lines = file_node.count_lines(full_content)
 
                 file_data.update({
                     "content": full_content,
                     "lines": all_lines
                 })
 
         except RepositoryError:
             log.exception("Exception in get_node")
             raise
 
         return file_data
 
     def get_fts_data(self, repo_name, commit_id, root_path='/'):
         """
         Fetch node tree for usage in full text search
         """
 
         tree_info = list()
 
         try:
             _repo = self._get_repo(repo_name)
             commit = _repo.scm_instance().get_commit(commit_id=commit_id)
             root_path = root_path.lstrip('/')
             top_node = commit.get_node(root_path)
             top_node.default_pre_load = []
 
             for __, dirs, files in commit.walk(top_node):
 
                 for f in files:
                     is_binary, md5, size, _content = f.metadata_uncached()
                     _data = {
                         "name": f.str_path,
                         "md5": md5,
                         "extension": f.extension,
                         "binary": is_binary,
                         "size": size
                     }
 
                     tree_info.append(_data)
 
         except RepositoryError:
             log.exception("Exception in get_nodes")
             raise
 
         return tree_info
 
     def create_nodes(self, user, repo, message, nodes, parent_commit=None,
                      author=None, trigger_push_hook=True):
         """
         Commits given multiple nodes into repo
 
         :param user: RhodeCode User object or user_id, the commiter
         :param repo: RhodeCode Repository object
         :param message: commit message
         :param nodes: mapping {filename:{'content':content},...}
         :param parent_commit: parent commit, can be empty than it's
            initial commit
         :param author: author of commit, cna be different that commiter
            only for git
         :param trigger_push_hook: trigger push hooks
 
         :returns: new committed commit
         """
         user, scm_instance, message, commiter, author, imc = self.initialize_inmemory_vars(
             user, repo, message, author)
 
         parent_commit, parents = self.get_parent_commits(parent_commit, scm_instance)
 
         upload_file_types = (io.BytesIO, io.BufferedRandom)
         processed_nodes = []
         for filename, content_dict in nodes.items():
             if not isinstance(filename, bytes):
                 raise ValueError(f'filename key in nodes needs to be bytes , or {upload_file_types}')
             content = content_dict['content']
             if not isinstance(content, upload_file_types + (bytes,)):
                 raise ValueError('content key value in nodes needs to be bytes')
 
         for f_path in nodes:
             f_path = self._sanitize_path(f_path)
             content = nodes[f_path]['content']
 
             # decoding here will force that we have proper encoded values
             # in any other case this will throw exceptions and deny commit
 
             if isinstance(content, bytes):
                 pass
             elif isinstance(content, upload_file_types):
                 content = content.read()
             else:
                 raise Exception(f'Content is of unrecognized type {type(content)}, expected {upload_file_types}')
             processed_nodes.append((f_path, content))
 
         # add multiple nodes
         for path, content in processed_nodes:
             imc.add(FileNode(path, content=content))
 
         # TODO: handle pre push scenario
         tip = imc.commit(message=message,
                          author=author,
                          parents=parents,
                          branch=parent_commit.branch)
 
         self.mark_for_invalidation(repo.repo_name)
         if trigger_push_hook:
             hooks_utils.trigger_post_push_hook(
                 username=user.username, action='push_local',
                 repo_name=repo.repo_name, repo_type=scm_instance.alias,
                 hook_type='post_push',
                 commit_ids=[tip.raw_id])
         return tip
 
     def update_nodes(self, user, repo, message, nodes, parent_commit=None,
                      author=None, trigger_push_hook=True):
         user, scm_instance, message, commiter, author, imc = self.initialize_inmemory_vars(
             user, repo, message, author)
 
         parent_commit, parents = self.get_parent_commits(parent_commit, scm_instance)
 
         # add multiple nodes
         for _filename, data in nodes.items():
             # new filename, can be renamed from the old one, also sanitaze
             # the path for any hack around relative paths like ../../ etc.
             filename = self._sanitize_path(data['filename'])
             old_filename = self._sanitize_path(_filename)
             content = data['content']
             file_mode = data.get('mode')
             filenode = FileNode(old_filename, content=content, mode=file_mode)
             op = data['op']
             if op == 'add':
                 imc.add(filenode)
             elif op == 'del':
                 imc.remove(filenode)
             elif op == 'mod':
                 if filename != old_filename:
                     # TODO: handle renames more efficient, needs vcs lib changes
                     imc.remove(filenode)
                     imc.add(FileNode(filename, content=content, mode=file_mode))
                 else:
                     imc.change(filenode)
 
         try:
             # TODO: handle pre push scenario commit changes
             tip = imc.commit(message=message,
                              author=author,
                              parents=parents,
                              branch=parent_commit.branch)
         except NodeNotChangedError:
             raise
         except Exception as e:
             log.exception("Unexpected exception during call to imc.commit")
             raise IMCCommitError(str(e))
         finally:
             # always clear caches, if commit fails we want fresh object also
             self.mark_for_invalidation(repo.repo_name)
 
         if trigger_push_hook:
             hooks_utils.trigger_post_push_hook(
                 username=user.username, action='push_local', hook_type='post_push',
                 repo_name=repo.repo_name, repo_type=scm_instance.alias,
                 commit_ids=[tip.raw_id])
 
         return tip
 
     def update_binary_node(self, user, repo, message, node, parent_commit=None, author=None):
         user, scm_instance, message, commiter, author, imc = self.initialize_inmemory_vars(
             user, repo, message, author)
 
         parent_commit, parents = self.get_parent_commits(parent_commit, scm_instance)
 
         file_path = node.get('file_path')
         if isinstance(raw_content := node.get('content'), (io.BytesIO, io.BufferedRandom)):
             content = raw_content.read()
         else:
             raise Exception("Wrong content was provided")
         file_node = FileNode(file_path, content=content)
         imc.change(file_node)
 
         try:
             tip = imc.commit(message=message,
                              author=author,
                              parents=parents,
                              branch=parent_commit.branch)
         except NodeNotChangedError:
             raise
         except Exception as e:
             log.exception("Unexpected exception during call to imc.commit")
             raise IMCCommitError(str(e))
         finally:
             self.mark_for_invalidation(repo.repo_name)
 
         hooks_utils.trigger_post_push_hook(
             username=user.username, action='push_local', hook_type='post_push',
             repo_name=repo.repo_name, repo_type=scm_instance.alias,
             commit_ids=[tip.raw_id])
         return tip
 
     def delete_nodes(self, user, repo, message, nodes, parent_commit=None,
                      author=None, trigger_push_hook=True):
         """
         Deletes given multiple nodes into `repo`
 
         :param user: RhodeCode User object or user_id, the committer
         :param repo: RhodeCode Repository object
         :param message: commit message
         :param nodes: mapping {filename:{'content':content},...}
         :param parent_commit: parent commit, can be empty than it's initial
            commit
         :param author: author of commit, cna be different that commiter only
            for git
         :param trigger_push_hook: trigger push hooks
 
         :returns: new commit after deletion
         """
 
         user, scm_instance, message, commiter, author, imc = self.initialize_inmemory_vars(
             user, repo, message, author)
 
         processed_nodes = []
         for f_path in nodes:
             f_path = self._sanitize_path(f_path)
             # content can be empty but for compatibility it allows same dicts
             # structure as add_nodes
             content = nodes[f_path].get('content')
             processed_nodes.append((safe_bytes(f_path), content))
 
         parent_commit, parents = self.get_parent_commits(parent_commit, scm_instance)
 
         # add multiple nodes
         for path, content in processed_nodes:
             imc.remove(FileNode(path, content=content))
 
         # TODO: handle pre push scenario
         tip = imc.commit(message=message,
                          author=author,
                          parents=parents,
                          branch=parent_commit.branch)
 
         self.mark_for_invalidation(repo.repo_name)
         if trigger_push_hook:
             hooks_utils.trigger_post_push_hook(
                 username=user.username, action='push_local', hook_type='post_push',
                 repo_name=repo.repo_name, repo_type=scm_instance.alias,
                 commit_ids=[tip.raw_id])
         return tip
 
     def strip(self, repo, commit_id, branch):
         scm_instance = repo.scm_instance(cache=False)
         scm_instance.config.clear_section('hooks')
         scm_instance.strip(commit_id, branch)
         self.mark_for_invalidation(repo.repo_name)
 
     def get_unread_journal(self):
         return self.sa.query(UserLog).count()
 
     @classmethod
     def backend_landing_ref(cls, repo_type):
         """
         Return a default landing ref based on a repository type.
         """
 
         landing_ref = {
             'hg': ('branch:default', 'default'),
             'git': ('branch:master', 'master'),
             'svn': ('rev:tip', 'latest tip'),
             'default': ('rev:tip', 'latest tip'),
         }
 
         return landing_ref.get(repo_type) or landing_ref['default']
 
     def get_repo_landing_revs(self, translator, repo=None):
         """
         Generates select option with tags branches and bookmarks (for hg only)
         grouped by type
 
         :param repo:
         """
         from rhodecode.lib.vcs.backends.git import GitRepository
 
         _ = translator
         repo = self._get_repo(repo)
 
         if repo:
             repo_type = repo.repo_type
         else:
             repo_type = 'default'
 
         default_landing_ref, landing_ref_lbl = self.backend_landing_ref(repo_type)
 
         default_ref_options = [
             [default_landing_ref, landing_ref_lbl]
         ]
         default_choices = [
             default_landing_ref
         ]
 
         if not repo:
             # presented at NEW repo creation
             return default_choices, default_ref_options
 
         repo = repo.scm_instance()
 
         ref_options = [(default_landing_ref, landing_ref_lbl)]
         choices = [default_landing_ref]
 
         # branches
         branch_group = [(f'branch:{safe_str(b)}', safe_str(b)) for b in repo.branches]
         if not branch_group:
             # new repo, or without maybe a branch?
             branch_group = default_ref_options
 
         branches_group = (branch_group, _("Branches"))
         ref_options.append(branches_group)
         choices.extend([x[0] for x in branches_group[0]])
 
         # bookmarks for HG
         if repo.alias == 'hg':
             bookmarks_group = (
                 [(f'book:{safe_str(b)}', safe_str(b))
                     for b in repo.bookmarks],
                 _("Bookmarks"))
             ref_options.append(bookmarks_group)
             choices.extend([x[0] for x in bookmarks_group[0]])
 
         # tags
         tags_group = (
             [(f'tag:{safe_str(t)}', safe_str(t))
                 for t in repo.tags],
             _("Tags"))
         ref_options.append(tags_group)
         choices.extend([x[0] for x in tags_group[0]])
 
         return choices, ref_options
 
     def get_server_info(self, environ=None):
         server_info = get_system_info(environ)
         return server_info