diff --git a/swh/storage/in_memory.py b/swh/storage/in_memory.py
index d983fee6..a38121a7 100644
--- a/swh/storage/in_memory.py
+++ b/swh/storage/in_memory.py
@@ -1,1212 +1,1212 @@
 # Copyright (C) 2015-2020  The Software Heritage developers
 # See the AUTHORS file at the top-level directory of this distribution
 # License: GNU General Public License version 3, or any later version
 # See top-level LICENSE file for more information
 
 import re
 import bisect
 import collections
 import copy
 import datetime
 import itertools
 import random
 
 from collections import defaultdict
 from datetime import timedelta
 from typing import (
     Any,
     Callable,
     Dict,
     Generic,
     Hashable,
     Iterable,
     Iterator,
     List,
     Optional,
     Tuple,
     TypeVar,
     Union,
 )
 
 import attr
 
 from swh.core.api.serializers import msgpack_loads, msgpack_dumps
 from swh.model.identifiers import SWHID
 from swh.model.model import (
     BaseContent,
     Content,
     SkippedContent,
     Directory,
     Revision,
     Release,
     Snapshot,
     OriginVisit,
     OriginVisitStatus,
     Origin,
     SHA1_SIZE,
     MetadataAuthority,
     MetadataAuthorityType,
     MetadataFetcher,
     MetadataTargetType,
     RawExtrinsicMetadata,
 )
 from swh.model.hashutil import DEFAULT_ALGORITHMS, hash_to_bytes, hash_to_hex
 from swh.storage.objstorage import ObjStorage
 from swh.storage.utils import now
 
 from .converters import origin_url_to_sha1
 from .exc import StorageArgumentException, HashCollision
 from .utils import get_partition_bounds_bytes
 from .writer import JournalWriter
 
 # Max block size of contents to return
 BULK_BLOCK_CONTENT_LEN_MAX = 10000
 
 
 SortedListItem = TypeVar("SortedListItem")
 SortedListKey = TypeVar("SortedListKey")
 
 FetcherKey = Tuple[str, str]
 
 
 class SortedList(collections.UserList, Generic[SortedListKey, SortedListItem]):
     data: List[Tuple[SortedListKey, SortedListItem]]
 
     # https://github.com/python/mypy/issues/708
     # key: Callable[[SortedListItem], SortedListKey]
 
     def __init__(
         self,
         data: List[SortedListItem] = None,
         key: Optional[Callable[[SortedListItem], SortedListKey]] = None,
     ):
         if key is None:
 
             def key(item):
                 return item
 
         assert key is not None  # for mypy
         super().__init__(sorted((key(x), x) for x in data or []))
 
         self.key: Callable[[SortedListItem], SortedListKey] = key
 
     def add(self, item: SortedListItem):
         k = self.key(item)
         bisect.insort(self.data, (k, item))
 
     def __iter__(self) -> Iterator[SortedListItem]:
         for (k, item) in self.data:
             yield item
 
     def iter_from(self, start_key: Any) -> Iterator[SortedListItem]:
         """Returns an iterator over all the elements whose key is greater
         or equal to `start_key`.
         (This is an efficient equivalent to:
         `(x for x in L if key(x) >= start_key)`)
         """
         from_index = bisect.bisect_left(self.data, (start_key,))
         for (k, item) in itertools.islice(self.data, from_index, None):
             yield item
 
     def iter_after(self, start_key: Any) -> Iterator[SortedListItem]:
         """Same as iter_from, but using a strict inequality."""
         it = self.iter_from(start_key)
         for item in it:
             if self.key(item) > start_key:  # type: ignore
                 yield item
                 break
 
         yield from it
 
 
 class InMemoryStorage:
     def __init__(self, journal_writer=None):
 
         self.reset()
         self.journal_writer = JournalWriter(journal_writer)
 
     def reset(self):
         self._contents = {}
         self._content_indexes = defaultdict(lambda: defaultdict(set))
         self._skipped_contents = {}
         self._skipped_content_indexes = defaultdict(lambda: defaultdict(set))
         self._directories = {}
         self._revisions = {}
         self._releases = {}
         self._snapshots = {}
         self._origins = {}
         self._origins_by_id = []
         self._origins_by_sha1 = {}
         self._origin_visits = {}
         self._origin_visit_statuses: Dict[Tuple[str, int], List[OriginVisitStatus]] = {}
         self._persons = {}
 
         # {object_type: {id: {authority: [metadata]}}}
         self._raw_extrinsic_metadata: Dict[
             MetadataTargetType,
             Dict[
                 Union[str, SWHID],
                 Dict[
                     Hashable,
                     SortedList[
                         Tuple[datetime.datetime, FetcherKey], RawExtrinsicMetadata
                     ],
                 ],
             ],
         ] = defaultdict(
             lambda: defaultdict(
                 lambda: defaultdict(
                     lambda: SortedList(
                         key=lambda x: (
                             x.discovery_date,
                             self._metadata_fetcher_key(x.fetcher),
                         )
                     )
                 )
             )
         )  # noqa
 
         self._metadata_fetchers: Dict[FetcherKey, MetadataFetcher] = {}
         self._metadata_authorities: Dict[Hashable, MetadataAuthority] = {}
         self._objects = defaultdict(list)
         self._sorted_sha1s = SortedList[bytes, bytes]()
 
         self.objstorage = ObjStorage({"cls": "memory", "args": {}})
 
     def check_config(self, *, check_write):
         return True
 
     def _content_add(self, contents: Iterable[Content], with_data: bool) -> Dict:
         self.journal_writer.content_add(contents)
 
         content_add = 0
         if with_data:
             summary = self.objstorage.content_add(
                 c for c in contents if c.status != "absent"
             )
             content_add_bytes = summary["content:add:bytes"]
 
         for content in contents:
             key = self._content_key(content)
             if key in self._contents:
                 continue
             for algorithm in DEFAULT_ALGORITHMS:
                 hash_ = content.get_hash(algorithm)
                 if hash_ in self._content_indexes[algorithm] and (
                     algorithm not in {"blake2s256", "sha256"}
                 ):
                     colliding_content_hashes = []
                     # Add the already stored contents
                     for content_hashes_set in self._content_indexes[algorithm][hash_]:
                         hashes = dict(content_hashes_set)
                         colliding_content_hashes.append(hashes)
                     # Add the new colliding content
                     colliding_content_hashes.append(content.hashes())
                     raise HashCollision(algorithm, hash_, colliding_content_hashes)
             for algorithm in DEFAULT_ALGORITHMS:
                 hash_ = content.get_hash(algorithm)
                 self._content_indexes[algorithm][hash_].add(key)
             self._objects[content.sha1_git].append(("content", content.sha1))
             self._contents[key] = content
             self._sorted_sha1s.add(content.sha1)
             self._contents[key] = attr.evolve(self._contents[key], data=None)
             content_add += 1
 
         summary = {
             "content:add": content_add,
         }
         if with_data:
             summary["content:add:bytes"] = content_add_bytes
 
         return summary
 
     def content_add(self, content: Iterable[Content]) -> Dict:
         content = [attr.evolve(c, ctime=now()) for c in content]
         return self._content_add(content, with_data=True)
 
     def content_update(self, content, keys=[]):
         self.journal_writer.content_update(content)
 
         for cont_update in content:
             cont_update = cont_update.copy()
             sha1 = cont_update.pop("sha1")
             for old_key in self._content_indexes["sha1"][sha1]:
                 old_cont = self._contents.pop(old_key)
 
                 for algorithm in DEFAULT_ALGORITHMS:
                     hash_ = old_cont.get_hash(algorithm)
                     self._content_indexes[algorithm][hash_].remove(old_key)
 
                 new_cont = attr.evolve(old_cont, **cont_update)
                 new_key = self._content_key(new_cont)
 
                 self._contents[new_key] = new_cont
 
                 for algorithm in DEFAULT_ALGORITHMS:
                     hash_ = new_cont.get_hash(algorithm)
                     self._content_indexes[algorithm][hash_].add(new_key)
 
     def content_add_metadata(self, content: Iterable[Content]) -> Dict:
         return self._content_add(content, with_data=False)
 
     def content_get(self, content):
         # FIXME: Make this method support slicing the `data`.
         if len(content) > BULK_BLOCK_CONTENT_LEN_MAX:
             raise StorageArgumentException(
                 "Sending at most %s contents." % BULK_BLOCK_CONTENT_LEN_MAX
             )
         yield from self.objstorage.content_get(content)
 
     def content_get_range(self, start, end, limit=1000):
         if limit is None:
             raise StorageArgumentException("limit should not be None")
         sha1s = (
             (sha1, content_key)
             for sha1 in self._sorted_sha1s.iter_from(start)
             for content_key in self._content_indexes["sha1"][sha1]
         )
         matched = []
         next_content = None
         for sha1, key in sha1s:
             if sha1 > end:
                 break
             if len(matched) >= limit:
                 next_content = sha1
                 break
             matched.append(self._contents[key].to_dict())
         return {
             "contents": matched,
             "next": next_content,
         }
 
     def content_get_partition(
         self,
         partition_id: int,
         nb_partitions: int,
         limit: int = 1000,
         page_token: str = None,
     ):
         if limit is None:
             raise StorageArgumentException("limit should not be None")
         (start, end) = get_partition_bounds_bytes(
             partition_id, nb_partitions, SHA1_SIZE
         )
         if page_token:
             start = hash_to_bytes(page_token)
         if end is None:
             end = b"\xff" * SHA1_SIZE
         result = self.content_get_range(start, end, limit)
         result2 = {
             "contents": result["contents"],
             "next_page_token": None,
         }
         if result["next"]:
             result2["next_page_token"] = hash_to_hex(result["next"])
         return result2
 
     def content_get_metadata(self, contents: List[bytes]) -> Dict[bytes, List[Dict]]:
         result: Dict = {sha1: [] for sha1 in contents}
         for sha1 in contents:
             if sha1 in self._content_indexes["sha1"]:
                 objs = self._content_indexes["sha1"][sha1]
                 # only 1 element as content_add_metadata would have raised a
                 # hash collision otherwise
                 for key in objs:
                     d = self._contents[key].to_dict()
                     del d["ctime"]
                     if "data" in d:
                         del d["data"]
                     result[sha1].append(d)
         return result
 
     def content_find(self, content):
         if not set(content).intersection(DEFAULT_ALGORITHMS):
             raise StorageArgumentException(
                 "content keys must contain at least one of: %s"
                 % ", ".join(sorted(DEFAULT_ALGORITHMS))
             )
         found = []
         for algo in DEFAULT_ALGORITHMS:
             hash = content.get(algo)
             if hash and hash in self._content_indexes[algo]:
                 found.append(self._content_indexes[algo][hash])
 
         if not found:
             return []
 
         keys = list(set.intersection(*found))
         return [self._contents[key].to_dict() for key in keys]
 
     def content_missing(self, content, key_hash="sha1"):
         for cont in content:
             for (algo, hash_) in cont.items():
                 if algo not in DEFAULT_ALGORITHMS:
                     continue
                 if hash_ not in self._content_indexes.get(algo, []):
                     yield cont[key_hash]
                     break
             else:
                 for result in self.content_find(cont):
                     if result["status"] == "missing":
                         yield cont[key_hash]
 
     def content_missing_per_sha1(self, contents):
         for content in contents:
             if content not in self._content_indexes["sha1"]:
                 yield content
 
     def content_missing_per_sha1_git(self, contents):
         for content in contents:
             if content not in self._content_indexes["sha1_git"]:
                 yield content
 
     def content_get_random(self):
         return random.choice(list(self._content_indexes["sha1_git"]))
 
     def _skipped_content_add(self, contents: List[SkippedContent]) -> Dict:
         self.journal_writer.skipped_content_add(contents)
 
         summary = {"skipped_content:add": 0}
 
         missing_contents = self.skipped_content_missing([c.hashes() for c in contents])
         missing = {self._content_key(c) for c in missing_contents}
         contents = [c for c in contents if self._content_key(c) in missing]
         for content in contents:
             key = self._content_key(content)
             for algo in DEFAULT_ALGORITHMS:
                 if content.get_hash(algo):
                     self._skipped_content_indexes[algo][content.get_hash(algo)].add(key)
             self._skipped_contents[key] = content
             summary["skipped_content:add"] += 1
 
         return summary
 
     def skipped_content_add(self, content: Iterable[SkippedContent]) -> Dict:
         content = [attr.evolve(c, ctime=now()) for c in content]
         return self._skipped_content_add(content)
 
     def skipped_content_missing(self, contents):
         for content in contents:
             matches = list(self._skipped_contents.values())
             for (algorithm, key) in self._content_key(content):
                 if algorithm == "blake2s256":
                     continue
                 # Filter out skipped contents with the same hash
                 matches = [
                     match for match in matches if match.get_hash(algorithm) == key
                 ]
             # if none of the contents match
             if not matches:
                 yield {algo: content[algo] for algo in DEFAULT_ALGORITHMS}
 
     def directory_add(self, directories: Iterable[Directory]) -> Dict:
         directories = [dir_ for dir_ in directories if dir_.id not in self._directories]
         self.journal_writer.directory_add(directories)
 
         count = 0
         for directory in directories:
             count += 1
             self._directories[directory.id] = directory
             self._objects[directory.id].append(("directory", directory.id))
 
         return {"directory:add": count}
 
     def directory_missing(self, directories):
         for id in directories:
             if id not in self._directories:
                 yield id
 
     def _join_dentry_to_content(self, dentry):
         keys = (
             "status",
             "sha1",
             "sha1_git",
             "sha256",
             "length",
         )
         ret = dict.fromkeys(keys)
         ret.update(dentry)
         if ret["type"] == "file":
             # TODO: Make it able to handle more than one content
             content = self.content_find({"sha1_git": ret["target"]})
             if content:
                 content = content[0]
                 for key in keys:
                     ret[key] = content[key]
         return ret
 
     def _directory_ls(self, directory_id, recursive, prefix=b""):
         if directory_id in self._directories:
             for entry in self._directories[directory_id].entries:
                 ret = self._join_dentry_to_content(entry.to_dict())
                 ret["name"] = prefix + ret["name"]
                 ret["dir_id"] = directory_id
                 yield ret
                 if recursive and ret["type"] == "dir":
                     yield from self._directory_ls(
                         ret["target"], True, prefix + ret["name"] + b"/"
                     )
 
     def directory_ls(self, directory, recursive=False):
         yield from self._directory_ls(directory, recursive)
 
     def directory_entry_get_by_path(self, directory, paths):
         return self._directory_entry_get_by_path(directory, paths, b"")
 
     def directory_get_random(self):
         if not self._directories:
             return None
         return random.choice(list(self._directories))
 
     def _directory_entry_get_by_path(self, directory, paths, prefix):
         if not paths:
             return
 
         contents = list(self.directory_ls(directory))
 
         if not contents:
             return
 
         def _get_entry(entries, name):
             for entry in entries:
                 if entry["name"] == name:
                     entry = entry.copy()
                     entry["name"] = prefix + entry["name"]
                     return entry
 
         first_item = _get_entry(contents, paths[0])
 
         if len(paths) == 1:
             return first_item
 
         if not first_item or first_item["type"] != "dir":
             return
 
         return self._directory_entry_get_by_path(
             first_item["target"], paths[1:], prefix + paths[0] + b"/"
         )
 
     def revision_add(self, revisions: Iterable[Revision]) -> Dict:
         revisions = [rev for rev in revisions if rev.id not in self._revisions]
         self.journal_writer.revision_add(revisions)
 
         count = 0
         for revision in revisions:
             revision = attr.evolve(
                 revision,
                 committer=self._person_add(revision.committer),
                 author=self._person_add(revision.author),
             )
             self._revisions[revision.id] = revision
             self._objects[revision.id].append(("revision", revision.id))
             count += 1
 
         return {"revision:add": count}
 
     def revision_missing(self, revisions):
         for id in revisions:
             if id not in self._revisions:
                 yield id
 
     def revision_get(self, revisions):
         for id in revisions:
             if id in self._revisions:
                 yield self._revisions.get(id).to_dict()
             else:
                 yield None
 
     def _get_parent_revs(self, rev_id, seen, limit):
         if limit and len(seen) >= limit:
             return
         if rev_id in seen or rev_id not in self._revisions:
             return
         seen.add(rev_id)
         yield self._revisions[rev_id].to_dict()
         for parent in self._revisions[rev_id].parents:
             yield from self._get_parent_revs(parent, seen, limit)
 
     def revision_log(self, revisions, limit=None):
         seen = set()
         for rev_id in revisions:
             yield from self._get_parent_revs(rev_id, seen, limit)
 
     def revision_shortlog(self, revisions, limit=None):
         yield from (
             (rev["id"], rev["parents"]) for rev in self.revision_log(revisions, limit)
         )
 
     def revision_get_random(self):
         return random.choice(list(self._revisions))
 
     def release_add(self, releases: Iterable[Release]) -> Dict:
         to_add = []
         for rel in releases:
             if rel.id not in self._releases and rel not in to_add:
                 to_add.append(rel)
         self.journal_writer.release_add(to_add)
 
         for rel in to_add:
             if rel.author:
                 self._person_add(rel.author)
             self._objects[rel.id].append(("release", rel.id))
             self._releases[rel.id] = rel
 
         return {"release:add": len(to_add)}
 
     def release_missing(self, releases):
         yield from (rel for rel in releases if rel not in self._releases)
 
     def release_get(self, releases):
         for rel_id in releases:
             if rel_id in self._releases:
                 yield self._releases[rel_id].to_dict()
             else:
                 yield None
 
     def release_get_random(self):
         return random.choice(list(self._releases))
 
     def snapshot_add(self, snapshots: Iterable[Snapshot]) -> Dict:
         count = 0
         snapshots = (snap for snap in snapshots if snap.id not in self._snapshots)
         for snapshot in snapshots:
             self.journal_writer.snapshot_add([snapshot])
             self._snapshots[snapshot.id] = snapshot
             self._objects[snapshot.id].append(("snapshot", snapshot.id))
             count += 1
 
         return {"snapshot:add": count}
 
     def snapshot_missing(self, snapshots):
         for id in snapshots:
             if id not in self._snapshots:
                 yield id
 
     def snapshot_get(self, snapshot_id):
         return self.snapshot_get_branches(snapshot_id)
 
     def snapshot_get_by_origin_visit(self, origin, visit):
         origin_url = self._get_origin_url(origin)
         if not origin_url:
             return
 
         if origin_url not in self._origins or visit > len(
             self._origin_visits[origin_url]
         ):
             return None
 
         visit = self._origin_visit_get_updated(origin_url, visit)
         snapshot_id = visit["snapshot"]
         if snapshot_id:
             return self.snapshot_get(snapshot_id)
         else:
             return None
 
     def snapshot_count_branches(self, snapshot_id):
         snapshot = self._snapshots[snapshot_id]
         return collections.Counter(
             branch.target_type.value if branch else None
             for branch in snapshot.branches.values()
         )
 
     def snapshot_get_branches(
         self, snapshot_id, branches_from=b"", branches_count=1000, target_types=None
     ):
         snapshot = self._snapshots.get(snapshot_id)
         if snapshot is None:
             return None
         sorted_branches = sorted(snapshot.branches.items())
         sorted_branch_names = [k for (k, v) in sorted_branches]
         from_index = bisect.bisect_left(sorted_branch_names, branches_from)
         if target_types:
             next_branch = None
             branches = {}
             for (branch_name, branch) in sorted_branches:
                 if branch_name in sorted_branch_names[from_index:]:
                     if branch and branch.target_type.value in target_types:
                         if len(branches) < branches_count:
                             branches[branch_name] = branch
                         else:
                             next_branch = branch_name
                             break
         else:
             # As there is no 'target_types', we can do that much faster
             to_index = from_index + branches_count
             returned_branch_names = frozenset(sorted_branch_names[from_index:to_index])
             branches = dict(
                 (branch_name, branch)
                 for (branch_name, branch) in snapshot.branches.items()
                 if branch_name in returned_branch_names
             )
             if to_index >= len(sorted_branch_names):
                 next_branch = None
             else:
                 next_branch = sorted_branch_names[to_index]
 
         branches = {
             name: branch.to_dict() if branch else None
             for (name, branch) in branches.items()
         }
 
         return {
             "id": snapshot_id,
             "branches": branches,
             "next_branch": next_branch,
         }
 
     def snapshot_get_random(self):
         return random.choice(list(self._snapshots))
 
     def object_find_by_sha1_git(self, ids):
         ret = {}
         for id_ in ids:
             objs = self._objects.get(id_, [])
             ret[id_] = [{"sha1_git": id_, "type": obj[0],} for obj in objs]
         return ret
 
     def _convert_origin(self, t):
         if t is None:
             return None
 
         return t.to_dict()
 
     def origin_get_one(self, origin_url: str) -> Optional[Origin]:
         return self._origins.get(origin_url)
 
     def origin_get(self, origins: Iterable[str]) -> Iterable[Optional[Origin]]:
         return [self.origin_get_one(origin_url) for origin_url in origins]
 
     def origin_get_by_sha1(self, sha1s):
         return [self._convert_origin(self._origins_by_sha1.get(sha1)) for sha1 in sha1s]
 
     def origin_get_range(self, origin_from=1, origin_count=100):
         origin_from = max(origin_from, 1)
         if origin_from <= len(self._origins_by_id):
             max_idx = origin_from + origin_count - 1
             if max_idx > len(self._origins_by_id):
                 max_idx = len(self._origins_by_id)
             for idx in range(origin_from - 1, max_idx):
                 origin = self._convert_origin(self._origins[self._origins_by_id[idx]])
                 yield {"id": idx + 1, **origin}
 
     def origin_list(self, page_token: Optional[str] = None, limit: int = 100) -> dict:
         origin_urls = sorted(self._origins)
         if page_token:
             from_ = bisect.bisect_left(origin_urls, page_token)
         else:
             from_ = 0
 
         result = {
             "origins": [
                 {"url": origin_url} for origin_url in origin_urls[from_ : from_ + limit]
             ]
         }
 
         if from_ + limit < len(origin_urls):
             result["next_page_token"] = origin_urls[from_ + limit]
 
         return result
 
     def origin_search(
         self, url_pattern, offset=0, limit=50, regexp=False, with_visit=False
     ):
         origins = map(self._convert_origin, self._origins.values())
         if regexp:
             pat = re.compile(url_pattern)
             origins = [orig for orig in origins if pat.search(orig["url"])]
         else:
             origins = [orig for orig in origins if url_pattern in orig["url"]]
         if with_visit:
             filtered_origins = []
             for orig in origins:
                 visits = (
                     self._origin_visit_get_updated(ov.origin, ov.visit)
                     for ov in self._origin_visits[orig["url"]]
                 )
                 for ov in visits:
                     snapshot = ov["snapshot"]
                     if snapshot and snapshot in self._snapshots:
                         filtered_origins.append(orig)
                         break
         else:
             filtered_origins = origins
 
         return filtered_origins[offset : offset + limit]
 
     def origin_count(self, url_pattern, regexp=False, with_visit=False):
         return len(
             self.origin_search(
                 url_pattern,
                 regexp=regexp,
                 with_visit=with_visit,
                 limit=len(self._origins),
             )
         )
 
     def origin_add(self, origins: Iterable[Origin]) -> Dict[str, int]:
         origins = list(origins)
         added = 0
         for origin in origins:
             if origin.url not in self._origins:
                 self.origin_add_one(origin)
                 added += 1
 
         return {"origin:add": added}
 
     def origin_add_one(self, origin: Origin) -> str:
         if origin.url not in self._origins:
             self.journal_writer.origin_add([origin])
             # generate an origin_id because it is needed by origin_get_range.
             # TODO: remove this when we remove origin_get_range
             origin_id = len(self._origins) + 1
             self._origins_by_id.append(origin.url)
             assert len(self._origins_by_id) == origin_id
 
             self._origins[origin.url] = origin
             self._origins_by_sha1[origin_url_to_sha1(origin.url)] = origin
             self._origin_visits[origin.url] = []
             self._objects[origin.url].append(("origin", origin.url))
 
         return origin.url
 
     def origin_visit_add(self, visits: Iterable[OriginVisit]) -> Iterable[OriginVisit]:
         for visit in visits:
             origin = self.origin_get_one(visit.origin)
             if not origin:  # Cannot add a visit without an origin
                 raise StorageArgumentException("Unknown origin %s", visit.origin)
 
         all_visits = []
         for visit in visits:
             origin_url = visit.origin
             if origin_url in self._origins:
                 origin = self._origins[origin_url]
                 if visit.visit:
                     self.journal_writer.origin_visit_add([visit])
                     while len(self._origin_visits[origin_url]) < visit.visit:
                         self._origin_visits[origin_url].append(None)
                     self._origin_visits[origin_url][visit.visit - 1] = visit
                 else:
                     # visit ids are in the range [1, +inf[
                     visit_id = len(self._origin_visits[origin_url]) + 1
                     visit = attr.evolve(visit, visit=visit_id)
                     self.journal_writer.origin_visit_add([visit])
                     self._origin_visits[origin_url].append(visit)
                     visit_key = (origin_url, visit.visit)
                     self._objects[visit_key].append(("origin_visit", None))
                 assert visit.visit is not None
                 self._origin_visit_status_add_one(
                     OriginVisitStatus(
                         origin=visit.origin,
                         visit=visit.visit,
                         date=visit.date,
                         status="created",
                         snapshot=None,
                     )
                 )
                 all_visits.append(visit)
 
         return all_visits
 
     def _origin_visit_status_add_one(self, visit_status: OriginVisitStatus) -> None:
         """Add an origin visit status without checks. If already present, do nothing.
 
         """
         self.journal_writer.origin_visit_status_add([visit_status])
         visit_key = (visit_status.origin, visit_status.visit)
         self._origin_visit_statuses.setdefault(visit_key, [])
         visit_statuses = self._origin_visit_statuses[visit_key]
         if visit_status not in visit_statuses:
             visit_statuses.append(visit_status)
 
     def origin_visit_status_add(
         self, visit_statuses: Iterable[OriginVisitStatus],
     ) -> None:
         # First round to check existence (fail early if any is ko)
         for visit_status in visit_statuses:
             origin_url = self.origin_get_one(visit_status.origin)
             if not origin_url:
                 raise StorageArgumentException(f"Unknown origin {visit_status.origin}")
 
         for visit_status in visit_statuses:
             self._origin_visit_status_add_one(visit_status)
 
     def _origin_visit_status_get_latest(
         self, origin: str, visit_id: int
     ) -> Tuple[OriginVisit, OriginVisitStatus]:
         """Return a tuple of OriginVisit, latest associated OriginVisitStatus.
 
         """
         assert visit_id >= 1
         visit = self._origin_visits[origin][visit_id - 1]
         assert visit is not None
         visit_key = (origin, visit_id)
 
         visit_update = max(self._origin_visit_statuses[visit_key], key=lambda v: v.date)
         return visit, visit_update
 
     def _origin_visit_get_updated(self, origin: str, visit_id: int) -> Dict[str, Any]:
         """Merge origin visit and latest origin visit status
 
         """
         visit, visit_update = self._origin_visit_status_get_latest(origin, visit_id)
         assert visit is not None and visit_update is not None
         return {
             # default to the values in visit
             **visit.to_dict(),
             # override with the last update
             **visit_update.to_dict(),
             # but keep the date of the creation of the origin visit
             "date": visit.date,
         }
 
     def origin_visit_get(
         self,
         origin: str,
         last_visit: Optional[int] = None,
         limit: Optional[int] = None,
         order: str = "asc",
     ) -> Iterable[Dict[str, Any]]:
         order = order.lower()
         assert order in ["asc", "desc"]
         origin_url = self._get_origin_url(origin)
         if origin_url in self._origin_visits:
             visits = self._origin_visits[origin_url]
             visits = sorted(visits, key=lambda v: v.visit, reverse=(order == "desc"))
             if last_visit is not None:
                 if order == "asc":
                     visits = [v for v in visits if v.visit > last_visit]
                 else:
                     visits = [v for v in visits if v.visit < last_visit]
             if limit is not None:
                 visits = visits[:limit]
             for visit in visits:
                 if not visit:
                     continue
                 visit_id = visit.visit
 
                 visit_update = self._origin_visit_get_updated(origin_url, visit_id)
                 assert visit_update is not None
                 yield visit_update
 
     def origin_visit_find_by_date(
         self, origin: str, visit_date: datetime.datetime
     ) -> Optional[OriginVisit]:
         origin_url = self._get_origin_url(origin)
         if origin_url in self._origin_visits:
             visits = self._origin_visits[origin_url]
             visit = min(visits, key=lambda v: (abs(v.date - visit_date), -v.visit))
             return visit
         return None
 
     def origin_visit_get_by(self, origin: str, visit: int) -> Optional[OriginVisit]:
         origin_url = self._get_origin_url(origin)
         if origin_url in self._origin_visits and visit <= len(
             self._origin_visits[origin_url]
         ):
             found_visit, _ = self._origin_visit_status_get_latest(origin, visit)
             return found_visit
         return None
 
     def origin_visit_get_latest(
         self,
         origin: str,
         type: Optional[str] = None,
         allowed_statuses: Optional[List[str]] = None,
         require_snapshot: bool = False,
     ) -> Optional[OriginVisit]:
         ori = self._origins.get(origin)
         if not ori:
             return None
 
         visits = sorted(
-            self._origin_visits[ori.url], key=lambda v: v.date, reverse=True,
+            self._origin_visits[ori.url], key=lambda v: (v.date, v.visit), reverse=True,
         )
         for visit in visits:
             if type is not None and visit.type != type:
                 continue
             visit_statuses = self._origin_visit_statuses[origin, visit.visit]
 
             if allowed_statuses is not None:
                 visit_statuses = [
                     vs for vs in visit_statuses if vs.status in allowed_statuses
                 ]
             if require_snapshot:
                 visit_statuses = [vs for vs in visit_statuses if vs.snapshot]
 
             if visit_statuses:  # we found visit statuses matching criteria
                 visit_status = max(visit_statuses, key=lambda vs: (vs.date, vs.visit))
                 assert visit.origin == visit_status.origin
                 assert visit.visit == visit_status.visit
                 return visit
 
         return None
 
     def origin_visit_status_get_latest(
         self,
         origin_url: str,
         visit: int,
         allowed_statuses: Optional[List[str]] = None,
         require_snapshot: bool = False,
     ) -> Optional[OriginVisitStatus]:
         ori = self._origins.get(origin_url)
         if not ori:
             return None
 
         visit_key = (origin_url, visit)
         visits = self._origin_visit_statuses.get(visit_key)
         if not visits:
             return None
 
         if allowed_statuses is not None:
             visits = [visit for visit in visits if visit.status in allowed_statuses]
         if require_snapshot:
             visits = [visit for visit in visits if visit.snapshot]
 
         visit_status = max(visits, key=lambda v: (v.date, v.visit), default=None)
         return visit_status
 
     def _select_random_origin_visit_by_type(self, type: str) -> str:
         while True:
             url = random.choice(list(self._origin_visits.keys()))
             random_origin_visits = self._origin_visits[url]
             if random_origin_visits[0].type == type:
                 return url
 
     def origin_visit_status_get_random(
         self, type: str
     ) -> Optional[Tuple[OriginVisit, OriginVisitStatus]]:
 
         url = self._select_random_origin_visit_by_type(type)
         random_origin_visits = copy.deepcopy(self._origin_visits[url])
         random_origin_visits.reverse()
         back_in_the_day = now() - timedelta(weeks=12)  # 3 months back
         # This should be enough for tests
         for visit in random_origin_visits:
             origin_visit, latest_visit_status = self._origin_visit_status_get_latest(
                 url, visit.visit
             )
             assert latest_visit_status is not None
             if (
                 origin_visit.date > back_in_the_day
                 and latest_visit_status.status == "full"
             ):
                 return origin_visit, latest_visit_status
         else:
             return None
 
     def stat_counters(self):
         keys = (
             "content",
             "directory",
             "origin",
             "origin_visit",
             "person",
             "release",
             "revision",
             "skipped_content",
             "snapshot",
         )
         stats = {key: 0 for key in keys}
         stats.update(
             collections.Counter(
                 obj_type
                 for (obj_type, obj_id) in itertools.chain(*self._objects.values())
             )
         )
         return stats
 
     def refresh_stat_counters(self):
         pass
 
     def raw_extrinsic_metadata_add(
         self, metadata: Iterable[RawExtrinsicMetadata],
     ) -> None:
         for metadata_entry in metadata:
             authority_key = self._metadata_authority_key(metadata_entry.authority)
             if authority_key not in self._metadata_authorities:
                 raise StorageArgumentException(
                     f"Unknown authority {metadata_entry.authority}"
                 )
             fetcher_key = self._metadata_fetcher_key(metadata_entry.fetcher)
             if fetcher_key not in self._metadata_fetchers:
                 raise StorageArgumentException(
                     f"Unknown fetcher {metadata_entry.fetcher}"
                 )
 
             raw_extrinsic_metadata_list = self._raw_extrinsic_metadata[
                 metadata_entry.type
             ][metadata_entry.id][authority_key]
 
             for existing_raw_extrinsic_metadata in raw_extrinsic_metadata_list:
                 if (
                     self._metadata_fetcher_key(existing_raw_extrinsic_metadata.fetcher)
                     == fetcher_key
                     and existing_raw_extrinsic_metadata.discovery_date
                     == metadata_entry.discovery_date
                 ):
                     # Duplicate of an existing one; ignore it.
                     break
             else:
                 raw_extrinsic_metadata_list.add(metadata_entry)
 
     def raw_extrinsic_metadata_get(
         self,
         object_type: MetadataTargetType,
         id: Union[str, SWHID],
         authority: MetadataAuthority,
         after: Optional[datetime.datetime] = None,
         page_token: Optional[bytes] = None,
         limit: int = 1000,
     ) -> Dict[str, Union[Optional[bytes], List[RawExtrinsicMetadata]]]:
         authority_key = self._metadata_authority_key(authority)
 
         if object_type == MetadataTargetType.ORIGIN:
             if isinstance(id, SWHID):
                 raise StorageArgumentException(
                     f"raw_extrinsic_metadata_get called with object_type='origin', "
                     f"but provided id is an SWHID: {id!r}"
                 )
         else:
             if not isinstance(id, SWHID):
                 raise StorageArgumentException(
                     f"raw_extrinsic_metadata_get called with object_type!='origin', "
                     f"but provided id is not an SWHID: {id!r}"
                 )
 
         if page_token is not None:
             (after_time, after_fetcher) = msgpack_loads(page_token)
             after_fetcher = tuple(after_fetcher)
             if after is not None and after > after_time:
                 raise StorageArgumentException(
                     "page_token is inconsistent with the value of 'after'."
                 )
             entries = self._raw_extrinsic_metadata[object_type][id][
                 authority_key
             ].iter_after((after_time, after_fetcher))
         elif after is not None:
             entries = self._raw_extrinsic_metadata[object_type][id][
                 authority_key
             ].iter_from((after,))
             entries = (entry for entry in entries if entry.discovery_date > after)
         else:
             entries = iter(self._raw_extrinsic_metadata[object_type][id][authority_key])
 
         if limit:
             entries = itertools.islice(entries, 0, limit + 1)
 
         results = []
         for entry in entries:
             entry_authority = self._metadata_authorities[
                 self._metadata_authority_key(entry.authority)
             ]
             entry_fetcher = self._metadata_fetchers[
                 self._metadata_fetcher_key(entry.fetcher)
             ]
             if after:
                 assert entry.discovery_date > after
             results.append(
                 attr.evolve(
                     entry,
                     authority=attr.evolve(entry_authority, metadata=None),
                     fetcher=attr.evolve(entry_fetcher, metadata=None),
                 )
             )
 
         if len(results) > limit:
             results.pop()
             assert len(results) == limit
             last_result = results[-1]
             next_page_token: Optional[bytes] = msgpack_dumps(
                 (
                     last_result.discovery_date,
                     self._metadata_fetcher_key(last_result.fetcher),
                 )
             )
         else:
             next_page_token = None
 
         return {
             "next_page_token": next_page_token,
             "results": results,
         }
 
     def metadata_fetcher_add(self, fetchers: Iterable[MetadataFetcher]) -> None:
         for fetcher in fetchers:
             if fetcher.metadata is None:
                 raise StorageArgumentException(
                     "MetadataFetcher.metadata may not be None in metadata_fetcher_add."
                 )
             key = self._metadata_fetcher_key(fetcher)
             if key not in self._metadata_fetchers:
                 self._metadata_fetchers[key] = fetcher
 
     def metadata_fetcher_get(
         self, name: str, version: str
     ) -> Optional[MetadataFetcher]:
         return self._metadata_fetchers.get(
             self._metadata_fetcher_key(MetadataFetcher(name=name, version=version))
         )
 
     def metadata_authority_add(self, authorities: Iterable[MetadataAuthority]) -> None:
         for authority in authorities:
             if authority.metadata is None:
                 raise StorageArgumentException(
                     "MetadataAuthority.metadata may not be None in "
                     "metadata_authority_add."
                 )
             key = self._metadata_authority_key(authority)
             self._metadata_authorities[key] = authority
 
     def metadata_authority_get(
         self, type: MetadataAuthorityType, url: str
     ) -> Optional[MetadataAuthority]:
         return self._metadata_authorities.get(
             self._metadata_authority_key(MetadataAuthority(type=type, url=url))
         )
 
     def _get_origin_url(self, origin):
         if isinstance(origin, str):
             return origin
         else:
             raise TypeError("origin must be a string.")
 
     def _person_add(self, person):
         key = ("person", person.fullname)
         if key not in self._objects:
             self._persons[person.fullname] = person
             self._objects[key].append(key)
 
         return self._persons[person.fullname]
 
     @staticmethod
     def _content_key(content):
         """ A stable key and the algorithm for a content"""
         if isinstance(content, BaseContent):
             content = content.to_dict()
         return tuple((key, content.get(key)) for key in sorted(DEFAULT_ALGORITHMS))
 
     @staticmethod
     def _metadata_fetcher_key(fetcher: MetadataFetcher) -> FetcherKey:
         return (fetcher.name, fetcher.version)
 
     @staticmethod
     def _metadata_authority_key(authority: MetadataAuthority) -> Hashable:
         return (authority.type, authority.url)
 
     def diff_directories(self, from_dir, to_dir, track_renaming=False):
         raise NotImplementedError("InMemoryStorage.diff_directories")
 
     def diff_revisions(self, from_rev, to_rev, track_renaming=False):
         raise NotImplementedError("InMemoryStorage.diff_revisions")
 
     def diff_revision(self, revision, track_renaming=False):
         raise NotImplementedError("InMemoryStorage.diff_revision")
 
     def clear_buffers(self, object_types: Optional[Iterable[str]] = None) -> None:
         """Do nothing
 
         """
         return None
 
     def flush(self, object_types: Optional[Iterable[str]] = None) -> Dict:
         return {}
diff --git a/swh/storage/tests/test_storage.py b/swh/storage/tests/test_storage.py
index 79fecfb6..6989e36b 100644
--- a/swh/storage/tests/test_storage.py
+++ b/swh/storage/tests/test_storage.py
@@ -1,4160 +1,4178 @@
 # Copyright (C) 2015-2020  The Software Heritage developers
 # See the AUTHORS file at the top-level directory of this distribution
 # License: GNU General Public License version 3, or any later version
 # See top-level LICENSE file for more information
 
 import datetime
 import inspect
 import itertools
 import math
 import queue
 import random
 import threading
 
 from collections import defaultdict
 from contextlib import contextmanager
 from datetime import timedelta
 from unittest.mock import Mock
 
 import attr
 import pytest
 
 from hypothesis import given, strategies, settings, HealthCheck
 
 from typing import ClassVar, Optional
 
 from swh.model import from_disk
 from swh.model.hashutil import hash_to_bytes
 from swh.model.identifiers import SWHID
 from swh.model.model import (
     Content,
     MetadataTargetType,
     Origin,
     OriginVisit,
     OriginVisitStatus,
     Person,
     Release,
     Revision,
     Snapshot,
 )
 from swh.model.hypothesis_strategies import objects
 from swh.storage import get_storage
 from swh.storage.converters import origin_url_to_sha1 as sha1
 from swh.storage.exc import HashCollision, StorageArgumentException
 from swh.storage.interface import StorageInterface
 from swh.storage.utils import content_hex_hashes, now
 
 
 @contextmanager
 def db_transaction(storage):
     with storage.db() as db:
         with db.transaction() as cur:
             yield db, cur
 
 
 def transform_entries(dir_, *, prefix=b""):
     for ent in dir_.entries:
         yield {
             "dir_id": dir_.id,
             "type": ent.type,
             "target": ent.target,
             "name": prefix + ent.name,
             "perms": ent.perms,
             "status": None,
             "sha1": None,
             "sha1_git": None,
             "sha256": None,
             "length": None,
         }
 
 
 def cmpdir(directory):
     return (directory["type"], directory["dir_id"])
 
 
 def assert_contents_ok(
     expected_contents, actual_contents, keys_to_check={"sha1", "data"}
 ):
     """Assert that a given list of contents matches on a given set of keys.
 
     """
     for k in keys_to_check:
         expected_list = set([c.get(k) for c in expected_contents])
         actual_list = set([c.get(k) for c in actual_contents])
         assert actual_list == expected_list, k
 
 
 def round_to_milliseconds(date):
     """Round datetime to milliseconds before insertion, so equality doesn't fail after a
     round-trip through a DB (eg. Cassandra)
 
     """
     return date.replace(microsecond=(date.microsecond // 1000) * 1000)
 
 
 def test_round_to_milliseconds():
     date = now()
 
     for (ms, expected_ms) in [(0, 0), (1000, 1000), (555555, 555000), (999500, 999000)]:
         date = date.replace(microsecond=ms)
         actual_date = round_to_milliseconds(date)
         assert actual_date.microsecond == expected_ms
 
 
 class LazyContent(Content):
     def with_data(self):
         return Content.from_dict({**self.to_dict(), "data": b"42\n"})
 
 
 class TestStorage:
     """Main class for Storage testing.
 
     This class is used as-is to test local storage (see TestLocalStorage
     below) and remote storage (see TestRemoteStorage in
     test_remote_storage.py.
 
     We need to have the two classes inherit from this base class
     separately to avoid nosetests running the tests from the base
     class twice.
     """
 
     maxDiff = None  # type: ClassVar[Optional[int]]
 
     def test_types(self, swh_storage_backend_config):
         """Checks all methods of StorageInterface are implemented by this
         backend, and that they have the same signature."""
         # Create an instance of the protocol (which cannot be instantiated
         # directly, so this creates a subclass, then instantiates it)
         interface = type("_", (StorageInterface,), {})()
         storage = get_storage(**swh_storage_backend_config)
 
         assert "content_add" in dir(interface)
 
         missing_methods = []
 
         for meth_name in dir(interface):
             if meth_name.startswith("_"):
                 continue
             interface_meth = getattr(interface, meth_name)
             try:
                 concrete_meth = getattr(storage, meth_name)
             except AttributeError:
                 if not getattr(interface_meth, "deprecated_endpoint", False):
                     # The backend is missing a (non-deprecated) endpoint
                     missing_methods.append(meth_name)
                 continue
 
             expected_signature = inspect.signature(interface_meth)
             actual_signature = inspect.signature(concrete_meth)
 
             assert expected_signature == actual_signature, meth_name
 
         assert missing_methods == []
 
     def test_check_config(self, swh_storage):
         assert swh_storage.check_config(check_write=True)
         assert swh_storage.check_config(check_write=False)
 
     def test_content_add(self, swh_storage, sample_data):
         cont = sample_data.content
 
         insertion_start_time = now()
         actual_result = swh_storage.content_add([cont])
         insertion_end_time = now()
 
         assert actual_result == {
             "content:add": 1,
             "content:add:bytes": cont.length,
         }
 
         assert list(swh_storage.content_get([cont.sha1])) == [
             {"sha1": cont.sha1, "data": cont.data}
         ]
 
         expected_cont = attr.evolve(cont, data=None)
 
         contents = [
             obj
             for (obj_type, obj) in swh_storage.journal_writer.journal.objects
             if obj_type == "content"
         ]
         assert len(contents) == 1
         for obj in contents:
             assert insertion_start_time <= obj.ctime
             assert obj.ctime <= insertion_end_time
             assert obj == expected_cont
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["content"] == 1
 
     def test_content_add_from_generator(self, swh_storage, sample_data):
         cont = sample_data.content
 
         def _cnt_gen():
             yield cont
 
         actual_result = swh_storage.content_add(_cnt_gen())
 
         assert actual_result == {
             "content:add": 1,
             "content:add:bytes": cont.length,
         }
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["content"] == 1
 
     def test_content_add_from_lazy_content(self, swh_storage, sample_data):
         cont = sample_data.content
         lazy_content = LazyContent.from_dict(cont.to_dict())
 
         insertion_start_time = now()
 
         actual_result = swh_storage.content_add([lazy_content])
 
         insertion_end_time = now()
 
         assert actual_result == {
             "content:add": 1,
             "content:add:bytes": cont.length,
         }
 
         # the fact that we retrieve the content object from the storage with
         # the correct 'data' field ensures it has been 'called'
         assert list(swh_storage.content_get([cont.sha1])) == [
             {"sha1": cont.sha1, "data": cont.data}
         ]
 
         expected_cont = attr.evolve(lazy_content, data=None, ctime=None)
         contents = [
             obj
             for (obj_type, obj) in swh_storage.journal_writer.journal.objects
             if obj_type == "content"
         ]
         assert len(contents) == 1
         for obj in contents:
             assert insertion_start_time <= obj.ctime
             assert obj.ctime <= insertion_end_time
             assert attr.evolve(obj, ctime=None).to_dict() == expected_cont.to_dict()
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["content"] == 1
 
     def test_content_get_missing(self, swh_storage, sample_data):
         cont, cont2 = sample_data.contents[:2]
 
         swh_storage.content_add([cont])
 
         # Query a single missing content
         results = list(swh_storage.content_get([cont2.sha1]))
         assert results == [None]
 
         # Check content_get does not abort after finding a missing content
         results = list(swh_storage.content_get([cont.sha1, cont2.sha1]))
         assert results == [{"sha1": cont.sha1, "data": cont.data}, None]
 
         # Check content_get does not discard found countent when it finds
         # a missing content.
         results = list(swh_storage.content_get([cont2.sha1, cont.sha1]))
         assert results == [None, {"sha1": cont.sha1, "data": cont.data}]
 
     def test_content_add_different_input(self, swh_storage, sample_data):
         cont, cont2 = sample_data.contents[:2]
 
         actual_result = swh_storage.content_add([cont, cont2])
         assert actual_result == {
             "content:add": 2,
             "content:add:bytes": cont.length + cont2.length,
         }
 
     def test_content_add_twice(self, swh_storage, sample_data):
         cont, cont2 = sample_data.contents[:2]
 
         actual_result = swh_storage.content_add([cont])
         assert actual_result == {
             "content:add": 1,
             "content:add:bytes": cont.length,
         }
         assert len(swh_storage.journal_writer.journal.objects) == 1
 
         actual_result = swh_storage.content_add([cont, cont2])
         assert actual_result == {
             "content:add": 1,
             "content:add:bytes": cont2.length,
         }
         assert 2 <= len(swh_storage.journal_writer.journal.objects) <= 3
 
         assert len(swh_storage.content_find(cont.to_dict())) == 1
         assert len(swh_storage.content_find(cont2.to_dict())) == 1
 
     def test_content_add_collision(self, swh_storage, sample_data):
         cont1 = sample_data.content
 
         # create (corrupted) content with same sha1{,_git} but != sha256
         sha256_array = bytearray(cont1.sha256)
         sha256_array[0] += 1
         cont1b = attr.evolve(cont1, sha256=bytes(sha256_array))
 
         with pytest.raises(HashCollision) as cm:
             swh_storage.content_add([cont1, cont1b])
 
         exc = cm.value
         actual_algo = exc.algo
         assert actual_algo in ["sha1", "sha1_git", "blake2s256"]
         actual_id = exc.hash_id
         assert actual_id == getattr(cont1, actual_algo).hex()
         collisions = exc.args[2]
         assert len(collisions) == 2
         assert collisions == [
             content_hex_hashes(cont1.hashes()),
             content_hex_hashes(cont1b.hashes()),
         ]
         assert exc.colliding_content_hashes() == [
             cont1.hashes(),
             cont1b.hashes(),
         ]
 
     def test_content_add_duplicate(self, swh_storage, sample_data):
         cont = sample_data.content
         swh_storage.content_add([cont, cont])
 
         assert list(swh_storage.content_get([cont.sha1])) == [
             {"sha1": cont.sha1, "data": cont.data}
         ]
 
     def test_content_update(self, swh_storage, sample_data):
         cont1 = sample_data.content
 
         if hasattr(swh_storage, "journal_writer"):
             swh_storage.journal_writer.journal = None  # TODO, not supported
 
         swh_storage.content_add([cont1])
 
         # alter the sha1_git for example
         cont1b = attr.evolve(
             cont1, sha1_git=hash_to_bytes("3a60a5275d0333bf13468e8b3dcab90f4046e654")
         )
 
         swh_storage.content_update([cont1b.to_dict()], keys=["sha1_git"])
 
         results = swh_storage.content_get_metadata([cont1.sha1])
 
         expected_content = attr.evolve(cont1b, data=None).to_dict()
         del expected_content["ctime"]
         assert tuple(results[cont1.sha1]) == (expected_content,)
 
     def test_content_add_metadata(self, swh_storage, sample_data):
         cont = attr.evolve(sample_data.content, data=None, ctime=now())
 
         actual_result = swh_storage.content_add_metadata([cont])
         assert actual_result == {
             "content:add": 1,
         }
 
         expected_cont = cont.to_dict()
         del expected_cont["ctime"]
 
         assert tuple(swh_storage.content_get_metadata([cont.sha1])[cont.sha1]) == (
             expected_cont,
         )
         contents = [
             obj
             for (obj_type, obj) in swh_storage.journal_writer.journal.objects
             if obj_type == "content"
         ]
         assert len(contents) == 1
         for obj in contents:
             obj = attr.evolve(obj, ctime=None)
             assert obj == cont
 
     def test_content_add_metadata_different_input(self, swh_storage, sample_data):
         contents = sample_data.contents[:2]
         cont = attr.evolve(contents[0], data=None, ctime=now())
         cont2 = attr.evolve(contents[1], data=None, ctime=now())
 
         actual_result = swh_storage.content_add_metadata([cont, cont2])
         assert actual_result == {
             "content:add": 2,
         }
 
     def test_content_add_metadata_collision(self, swh_storage, sample_data):
         cont1 = attr.evolve(sample_data.content, data=None, ctime=now())
 
         # create (corrupted) content with same sha1{,_git} but != sha256
         sha1_git_array = bytearray(cont1.sha256)
         sha1_git_array[0] += 1
         cont1b = attr.evolve(cont1, sha256=bytes(sha1_git_array))
 
         with pytest.raises(HashCollision) as cm:
             swh_storage.content_add_metadata([cont1, cont1b])
 
         exc = cm.value
         actual_algo = exc.algo
         assert actual_algo in ["sha1", "sha1_git", "blake2s256"]
         actual_id = exc.hash_id
         assert actual_id == getattr(cont1, actual_algo).hex()
         collisions = exc.args[2]
         assert len(collisions) == 2
         assert collisions == [
             content_hex_hashes(cont1.hashes()),
             content_hex_hashes(cont1b.hashes()),
         ]
         assert exc.colliding_content_hashes() == [
             cont1.hashes(),
             cont1b.hashes(),
         ]
 
     def test_skipped_content_add(self, swh_storage, sample_data):
         contents = sample_data.skipped_contents[:2]
         cont = contents[0]
         cont2 = attr.evolve(contents[1], blake2s256=None)
 
         contents_dict = [c.to_dict() for c in [cont, cont2]]
 
         missing = list(swh_storage.skipped_content_missing(contents_dict))
 
         assert missing == [cont.hashes(), cont2.hashes()]
 
         actual_result = swh_storage.skipped_content_add([cont, cont, cont2])
 
         assert 2 <= actual_result.pop("skipped_content:add") <= 3
         assert actual_result == {}
 
         missing = list(swh_storage.skipped_content_missing(contents_dict))
         assert missing == []
 
     def test_skipped_content_add_missing_hashes(self, swh_storage, sample_data):
         cont, cont2 = [
             attr.evolve(c, sha1_git=None) for c in sample_data.skipped_contents[:2]
         ]
         contents_dict = [c.to_dict() for c in [cont, cont2]]
 
         missing = list(swh_storage.skipped_content_missing(contents_dict))
         assert len(missing) == 2
 
         actual_result = swh_storage.skipped_content_add([cont, cont, cont2])
 
         assert 2 <= actual_result.pop("skipped_content:add") <= 3
         assert actual_result == {}
 
         missing = list(swh_storage.skipped_content_missing(contents_dict))
         assert missing == []
 
     def test_skipped_content_missing_partial_hash(self, swh_storage, sample_data):
         cont = sample_data.skipped_content
         cont2 = attr.evolve(cont, sha1_git=None)
         contents_dict = [c.to_dict() for c in [cont, cont2]]
 
         missing = list(swh_storage.skipped_content_missing(contents_dict))
         assert len(missing) == 2
 
         actual_result = swh_storage.skipped_content_add([cont])
 
         assert actual_result.pop("skipped_content:add") == 1
         assert actual_result == {}
 
         missing = list(swh_storage.skipped_content_missing(contents_dict))
         assert missing == [cont2.hashes()]
 
     @pytest.mark.property_based
     @settings(deadline=None)  # this test is very slow
     @given(
         strategies.sets(
             elements=strategies.sampled_from(["sha256", "sha1_git", "blake2s256"]),
             min_size=0,
         )
     )
     def test_content_missing(self, swh_storage, sample_data, algos):
         algos |= {"sha1"}
         content, missing_content = [sample_data.content2, sample_data.skipped_content]
         swh_storage.content_add([content])
 
         test_contents = [content.to_dict()]
         missing_per_hash = defaultdict(list)
         for i in range(256):
             test_content = missing_content.to_dict()
             for hash in algos:
                 test_content[hash] = bytes([i]) + test_content[hash][1:]
                 missing_per_hash[hash].append(test_content[hash])
             test_contents.append(test_content)
 
         assert set(swh_storage.content_missing(test_contents)) == set(
             missing_per_hash["sha1"]
         )
 
         for hash in algos:
             assert set(
                 swh_storage.content_missing(test_contents, key_hash=hash)
             ) == set(missing_per_hash[hash])
 
     @pytest.mark.property_based
     @given(
         strategies.sets(
             elements=strategies.sampled_from(["sha256", "sha1_git", "blake2s256"]),
             min_size=0,
         )
     )
     def test_content_missing_unknown_algo(self, swh_storage, sample_data, algos):
         algos |= {"sha1"}
         content, missing_content = [sample_data.content2, sample_data.skipped_content]
         swh_storage.content_add([content])
 
         test_contents = [content.to_dict()]
         missing_per_hash = defaultdict(list)
         for i in range(16):
             test_content = missing_content.to_dict()
             for hash in algos:
                 test_content[hash] = bytes([i]) + test_content[hash][1:]
                 missing_per_hash[hash].append(test_content[hash])
             test_content["nonexisting_algo"] = b"\x00"
             test_contents.append(test_content)
 
         assert set(swh_storage.content_missing(test_contents)) == set(
             missing_per_hash["sha1"]
         )
 
         for hash in algos:
             assert set(
                 swh_storage.content_missing(test_contents, key_hash=hash)
             ) == set(missing_per_hash[hash])
 
     def test_content_missing_per_sha1(self, swh_storage, sample_data):
         # given
         cont = sample_data.content
         missing_cont = sample_data.skipped_content
         swh_storage.content_add([cont])
 
         # when
         gen = swh_storage.content_missing_per_sha1([cont.sha1, missing_cont.sha1])
         # then
         assert list(gen) == [missing_cont.sha1]
 
     def test_content_missing_per_sha1_git(self, swh_storage, sample_data):
         cont, cont2 = sample_data.contents[:2]
         missing_cont = sample_data.skipped_content
 
         swh_storage.content_add([cont, cont2])
 
         contents = [cont.sha1_git, cont2.sha1_git, missing_cont.sha1_git]
 
         missing_contents = swh_storage.content_missing_per_sha1_git(contents)
         assert list(missing_contents) == [missing_cont.sha1_git]
 
     def test_content_get_partition(self, swh_storage, swh_contents):
         """content_get_partition paginates results if limit exceeded"""
         expected_contents = [c.to_dict() for c in swh_contents if c.status != "absent"]
 
         actual_contents = []
         for i in range(16):
             actual_result = swh_storage.content_get_partition(i, 16)
             assert actual_result["next_page_token"] is None
             actual_contents.extend(actual_result["contents"])
 
         assert_contents_ok(expected_contents, actual_contents, ["sha1"])
 
     def test_content_get_partition_full(self, swh_storage, swh_contents):
         """content_get_partition for a single partition returns all available
         contents"""
         expected_contents = [c.to_dict() for c in swh_contents if c.status != "absent"]
 
         actual_result = swh_storage.content_get_partition(0, 1)
         assert actual_result["next_page_token"] is None
 
         actual_contents = actual_result["contents"]
         assert_contents_ok(expected_contents, actual_contents, ["sha1"])
 
     def test_content_get_partition_empty(self, swh_storage, swh_contents):
         """content_get_partition when at least one of the partitions is
         empty"""
         expected_contents = {
             cont.sha1 for cont in swh_contents if cont.status != "absent"
         }
         # nb_partitions = smallest power of 2 such that at least one of
         # the partitions is empty
         nb_partitions = 1 << math.floor(math.log2(len(swh_contents)) + 1)
 
         seen_sha1s = []
 
         for i in range(nb_partitions):
             actual_result = swh_storage.content_get_partition(
                 i, nb_partitions, limit=len(swh_contents) + 1
             )
 
             for cont in actual_result["contents"]:
                 seen_sha1s.append(cont["sha1"])
 
             # Limit is higher than the max number of results
             assert actual_result["next_page_token"] is None
 
         assert set(seen_sha1s) == expected_contents
 
     def test_content_get_partition_limit_none(self, swh_storage):
         """content_get_partition call with wrong limit input should fail"""
         with pytest.raises(StorageArgumentException) as e:
             swh_storage.content_get_partition(1, 16, limit=None)
 
         assert e.value.args == ("limit should not be None",)
 
     def test_generate_content_get_partition_pagination(self, swh_storage, swh_contents):
         """content_get_partition returns contents within range provided"""
         expected_contents = [c.to_dict() for c in swh_contents if c.status != "absent"]
 
         # retrieve contents
         actual_contents = []
         for i in range(4):
             page_token = None
             while True:
                 actual_result = swh_storage.content_get_partition(
                     i, 4, limit=3, page_token=page_token
                 )
                 actual_contents.extend(actual_result["contents"])
                 page_token = actual_result["next_page_token"]
 
                 if page_token is None:
                     break
 
         assert_contents_ok(expected_contents, actual_contents, ["sha1"])
 
     def test_content_get_metadata(self, swh_storage, sample_data):
         cont1, cont2 = sample_data.contents[:2]
 
         swh_storage.content_add([cont1, cont2])
 
         actual_md = swh_storage.content_get_metadata([cont1.sha1, cont2.sha1])
 
         # we only retrieve the metadata so no data nor ctime within
         expected_cont1, expected_cont2 = [
             attr.evolve(c, data=None).to_dict() for c in [cont1, cont2]
         ]
         expected_cont1.pop("ctime")
         expected_cont2.pop("ctime")
 
         assert tuple(actual_md[cont1.sha1]) == (expected_cont1,)
         assert tuple(actual_md[cont2.sha1]) == (expected_cont2,)
         assert len(actual_md.keys()) == 2
 
     def test_content_get_metadata_missing_sha1(self, swh_storage, sample_data):
         cont1, cont2 = sample_data.contents[:2]
         missing_cont = sample_data.skipped_content
 
         swh_storage.content_add([cont1, cont2])
 
         actual_contents = swh_storage.content_get_metadata([missing_cont.sha1])
 
         assert len(actual_contents) == 1
         assert tuple(actual_contents[missing_cont.sha1]) == ()
 
     def test_content_get_random(self, swh_storage, sample_data):
         cont, cont2, cont3 = sample_data.contents[:3]
         swh_storage.content_add([cont, cont2, cont3])
 
         assert swh_storage.content_get_random() in {
             cont.sha1_git,
             cont2.sha1_git,
             cont3.sha1_git,
         }
 
     def test_directory_add(self, swh_storage, sample_data):
         directory = sample_data.directories[1]
 
         init_missing = list(swh_storage.directory_missing([directory.id]))
         assert [directory.id] == init_missing
 
         actual_result = swh_storage.directory_add([directory])
         assert actual_result == {"directory:add": 1}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("directory", directory)
         ]
 
         actual_data = list(swh_storage.directory_ls(directory.id))
         expected_data = list(transform_entries(directory))
 
         assert sorted(expected_data, key=cmpdir) == sorted(actual_data, key=cmpdir)
 
         after_missing = list(swh_storage.directory_missing([directory.id]))
         assert after_missing == []
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["directory"] == 1
 
     def test_directory_add_from_generator(self, swh_storage, sample_data):
         directory = sample_data.directories[1]
 
         def _dir_gen():
             yield directory
 
         actual_result = swh_storage.directory_add(directories=_dir_gen())
         assert actual_result == {"directory:add": 1}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("directory", directory)
         ]
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["directory"] == 1
 
     def test_directory_add_twice(self, swh_storage, sample_data):
         directory = sample_data.directories[1]
 
         actual_result = swh_storage.directory_add([directory])
         assert actual_result == {"directory:add": 1}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("directory", directory)
         ]
 
         actual_result = swh_storage.directory_add([directory])
         assert actual_result == {"directory:add": 0}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("directory", directory)
         ]
 
     def test_directory_get_recursive(self, swh_storage, sample_data):
         dir1, dir2, dir3 = sample_data.directories[:3]
 
         init_missing = list(swh_storage.directory_missing([dir1.id]))
         assert init_missing == [dir1.id]
 
         actual_result = swh_storage.directory_add([dir1, dir2, dir3])
         assert actual_result == {"directory:add": 3}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("directory", dir1),
             ("directory", dir2),
             ("directory", dir3),
         ]
 
         # List directory containing a file and an unknown subdirectory
         actual_data = list(swh_storage.directory_ls(dir1.id, recursive=True))
         expected_data = list(transform_entries(dir1))
         assert sorted(expected_data, key=cmpdir) == sorted(actual_data, key=cmpdir)
 
         # List directory containing a file and an unknown subdirectory
         actual_data = list(swh_storage.directory_ls(dir2.id, recursive=True))
         expected_data = list(transform_entries(dir2))
         assert sorted(expected_data, key=cmpdir) == sorted(actual_data, key=cmpdir)
 
         # List directory containing a known subdirectory, entries should
         # be both those of the directory and of the subdir
         actual_data = list(swh_storage.directory_ls(dir3.id, recursive=True))
         expected_data = list(
             itertools.chain(
                 transform_entries(dir3), transform_entries(dir2, prefix=b"subdir/"),
             )
         )
         assert sorted(expected_data, key=cmpdir) == sorted(actual_data, key=cmpdir)
 
     def test_directory_get_non_recursive(self, swh_storage, sample_data):
         dir1, dir2, dir3 = sample_data.directories[:3]
 
         init_missing = list(swh_storage.directory_missing([dir1.id]))
         assert init_missing == [dir1.id]
 
         actual_result = swh_storage.directory_add([dir1, dir2, dir3])
         assert actual_result == {"directory:add": 3}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("directory", dir1),
             ("directory", dir2),
             ("directory", dir3),
         ]
 
         # List directory containing a file and an unknown subdirectory
         actual_data = list(swh_storage.directory_ls(dir1.id))
         expected_data = list(transform_entries(dir1))
         assert sorted(expected_data, key=cmpdir) == sorted(actual_data, key=cmpdir)
 
         # List directory contaiining a single file
         actual_data = list(swh_storage.directory_ls(dir2.id))
         expected_data = list(transform_entries(dir2))
         assert sorted(expected_data, key=cmpdir) == sorted(actual_data, key=cmpdir)
 
         # List directory containing a known subdirectory, entries should
         # only be those of the parent directory, not of the subdir
         actual_data = list(swh_storage.directory_ls(dir3.id))
         expected_data = list(transform_entries(dir3))
         assert sorted(expected_data, key=cmpdir) == sorted(actual_data, key=cmpdir)
 
     def test_directory_entry_get_by_path(self, swh_storage, sample_data):
         cont = sample_data.content
         dir1, dir2, dir3, dir4, dir5 = sample_data.directories[:5]
 
         # given
         init_missing = list(swh_storage.directory_missing([dir3.id]))
         assert init_missing == [dir3.id]
 
         actual_result = swh_storage.directory_add([dir3, dir4])
         assert actual_result == {"directory:add": 2}
 
         expected_entries = [
             {
                 "dir_id": dir3.id,
                 "name": b"foo",
                 "type": "file",
                 "target": cont.sha1_git,
                 "sha1": None,
                 "sha1_git": None,
                 "sha256": None,
                 "status": None,
                 "perms": from_disk.DentryPerms.content,
                 "length": None,
             },
             {
                 "dir_id": dir3.id,
                 "name": b"subdir",
                 "type": "dir",
                 "target": dir2.id,
                 "sha1": None,
                 "sha1_git": None,
                 "sha256": None,
                 "status": None,
                 "perms": from_disk.DentryPerms.directory,
                 "length": None,
             },
             {
                 "dir_id": dir3.id,
                 "name": b"hello",
                 "type": "file",
                 "target": dir5.id,
                 "sha1": None,
                 "sha1_git": None,
                 "sha256": None,
                 "status": None,
                 "perms": from_disk.DentryPerms.content,
                 "length": None,
             },
         ]
 
         # when (all must be found here)
         for entry, expected_entry in zip(dir3.entries, expected_entries):
             actual_entry = swh_storage.directory_entry_get_by_path(
                 dir3.id, [entry.name]
             )
             assert actual_entry == expected_entry
 
         # same, but deeper
         for entry, expected_entry in zip(dir3.entries, expected_entries):
             actual_entry = swh_storage.directory_entry_get_by_path(
                 dir4.id, [b"subdir1", entry.name]
             )
             expected_entry = expected_entry.copy()
             expected_entry["name"] = b"subdir1/" + expected_entry["name"]
             assert actual_entry == expected_entry
 
         # when (nothing should be found here since `dir` is not persisted.)
         for entry in dir2.entries:
             actual_entry = swh_storage.directory_entry_get_by_path(
                 dir2.id, [entry.name]
             )
             assert actual_entry is None
 
     def test_directory_get_random(self, swh_storage, sample_data):
         dir1, dir2, dir3 = sample_data.directories[:3]
         swh_storage.directory_add([dir1, dir2, dir3])
 
         assert swh_storage.directory_get_random() in {
             dir1.id,
             dir2.id,
             dir3.id,
         }
 
     def test_revision_add(self, swh_storage, sample_data):
         revision = sample_data.revision
         init_missing = swh_storage.revision_missing([revision.id])
         assert list(init_missing) == [revision.id]
 
         actual_result = swh_storage.revision_add([revision])
         assert actual_result == {"revision:add": 1}
 
         end_missing = swh_storage.revision_missing([revision.id])
         assert list(end_missing) == []
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("revision", revision)
         ]
 
         # already there so nothing added
         actual_result = swh_storage.revision_add([revision])
         assert actual_result == {"revision:add": 0}
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["revision"] == 1
 
     def test_revision_add_from_generator(self, swh_storage, sample_data):
         revision = sample_data.revision
 
         def _rev_gen():
             yield revision
 
         actual_result = swh_storage.revision_add(_rev_gen())
         assert actual_result == {"revision:add": 1}
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["revision"] == 1
 
     def test_revision_add_twice(self, swh_storage, sample_data):
         revision, revision2 = sample_data.revisions[:2]
 
         actual_result = swh_storage.revision_add([revision])
         assert actual_result == {"revision:add": 1}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("revision", revision)
         ]
 
         actual_result = swh_storage.revision_add([revision, revision2])
         assert actual_result == {"revision:add": 1}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("revision", revision),
             ("revision", revision2),
         ]
 
     def test_revision_add_name_clash(self, swh_storage, sample_data):
         revision, revision2 = sample_data.revisions[:2]
 
         revision1 = attr.evolve(
             revision,
             author=Person(
                 fullname=b"John Doe <john.doe@example.com>",
                 name=b"John Doe",
                 email=b"john.doe@example.com",
             ),
         )
         revision2 = attr.evolve(
             revision2,
             author=Person(
                 fullname=b"John Doe <john.doe@example.com>",
                 name=b"John Doe ",
                 email=b"john.doe@example.com ",
             ),
         )
         actual_result = swh_storage.revision_add([revision1, revision2])
         assert actual_result == {"revision:add": 2}
 
     def test_revision_get_order(self, swh_storage, sample_data):
         revision, revision2 = sample_data.revisions[:2]
 
         add_result = swh_storage.revision_add([revision, revision2])
         assert add_result == {"revision:add": 2}
 
         # order 1
         res1 = swh_storage.revision_get([revision.id, revision2.id])
 
         assert [Revision.from_dict(r) for r in res1] == [revision, revision2]
 
         # order 2
         res2 = swh_storage.revision_get([revision2.id, revision.id])
         assert [Revision.from_dict(r) for r in res2] == [revision2, revision]
 
     def test_revision_log(self, swh_storage, sample_data):
         revision1, revision2, revision3, revision4 = sample_data.revisions[:4]
 
         # rev4 -is-child-of-> rev3 -> rev1, (rev2 -> rev1)
         swh_storage.revision_add([revision1, revision2, revision3, revision4])
 
         # when
         results = list(swh_storage.revision_log([revision4.id]))
 
         # for comparison purposes
         actual_results = [Revision.from_dict(r) for r in results]
         assert len(actual_results) == 4  # rev4 -child-> rev3 -> rev1, (rev2 -> rev1)
         assert actual_results == [revision4, revision3, revision1, revision2]
 
     def test_revision_log_with_limit(self, swh_storage, sample_data):
         revision1, revision2, revision3, revision4 = sample_data.revisions[:4]
 
         # revision4 -is-child-of-> revision3
         swh_storage.revision_add([revision3, revision4])
         results = list(swh_storage.revision_log([revision4.id], 1))
 
         actual_results = [Revision.from_dict(r) for r in results]
         assert len(actual_results) == 1
         assert actual_results[0] == revision4
 
     def test_revision_log_unknown_revision(self, swh_storage, sample_data):
         revision = sample_data.revision
         rev_log = list(swh_storage.revision_log([revision.id]))
         assert rev_log == []
 
     def test_revision_shortlog(self, swh_storage, sample_data):
         revision1, revision2, revision3, revision4 = sample_data.revisions[:4]
 
         # rev4 -is-child-of-> rev3 -> (rev1, rev2); rev2 -> rev1
         swh_storage.revision_add([revision1, revision2, revision3, revision4])
 
         results = list(swh_storage.revision_shortlog([revision4.id]))
         actual_results = [[id, tuple(parents)] for (id, parents) in results]
 
         assert len(actual_results) == 4
         assert actual_results == [
             [revision4.id, revision4.parents],
             [revision3.id, revision3.parents],
             [revision1.id, revision1.parents],
             [revision2.id, revision2.parents],
         ]
 
     def test_revision_shortlog_with_limit(self, swh_storage, sample_data):
         revision1, revision2, revision3, revision4 = sample_data.revisions[:4]
 
         # revision4 -is-child-of-> revision3
         swh_storage.revision_add([revision1, revision2, revision3, revision4])
         results = list(swh_storage.revision_shortlog([revision4.id], 1))
         actual_results = [[id, tuple(parents)] for (id, parents) in results]
 
         assert len(actual_results) == 1
         assert list(actual_results[0]) == [revision4.id, revision4.parents]
 
     def test_revision_get(self, swh_storage, sample_data):
         revision, revision2 = sample_data.revisions[:2]
 
         swh_storage.revision_add([revision])
 
         actual_revisions = list(swh_storage.revision_get([revision.id, revision2.id]))
 
         assert len(actual_revisions) == 2
         assert Revision.from_dict(actual_revisions[0]) == revision
         assert actual_revisions[1] is None
 
     def test_revision_get_no_parents(self, swh_storage, sample_data):
         revision = sample_data.revision
         swh_storage.revision_add([revision])
 
         get = list(swh_storage.revision_get([revision.id]))
 
         assert len(get) == 1
         assert revision.parents == ()
         assert tuple(get[0]["parents"]) == ()  # no parents on this one
 
     def test_revision_get_random(self, swh_storage, sample_data):
         revision1, revision2, revision3 = sample_data.revisions[:3]
 
         swh_storage.revision_add([revision1, revision2, revision3])
 
         assert swh_storage.revision_get_random() in {
             revision1.id,
             revision2.id,
             revision3.id,
         }
 
     def test_release_add(self, swh_storage, sample_data):
         release, release2 = sample_data.releases[:2]
 
         init_missing = swh_storage.release_missing([release.id, release2.id])
         assert list(init_missing) == [release.id, release2.id]
 
         actual_result = swh_storage.release_add([release, release2])
         assert actual_result == {"release:add": 2}
 
         end_missing = swh_storage.release_missing([release.id, release2.id])
         assert list(end_missing) == []
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("release", release),
             ("release", release2),
         ]
 
         # already present so nothing added
         actual_result = swh_storage.release_add([release, release2])
         assert actual_result == {"release:add": 0}
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["release"] == 2
 
     def test_release_add_from_generator(self, swh_storage, sample_data):
         release, release2 = sample_data.releases[:2]
 
         def _rel_gen():
             yield release
             yield release2
 
         actual_result = swh_storage.release_add(_rel_gen())
         assert actual_result == {"release:add": 2}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("release", release),
             ("release", release2),
         ]
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["release"] == 2
 
     def test_release_add_no_author_date(self, swh_storage, sample_data):
         full_release = sample_data.release
 
         release = attr.evolve(full_release, author=None, date=None)
         actual_result = swh_storage.release_add([release])
         assert actual_result == {"release:add": 1}
 
         end_missing = swh_storage.release_missing([release.id])
         assert list(end_missing) == []
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("release", release)
         ]
 
     def test_release_add_twice(self, swh_storage, sample_data):
         release, release2 = sample_data.releases[:2]
 
         actual_result = swh_storage.release_add([release])
         assert actual_result == {"release:add": 1}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("release", release)
         ]
 
         actual_result = swh_storage.release_add([release, release2, release, release2])
         assert actual_result == {"release:add": 1}
 
         assert set(swh_storage.journal_writer.journal.objects) == set(
             [("release", release), ("release", release2),]
         )
 
     def test_release_add_name_clash(self, swh_storage, sample_data):
         release, release2 = [
             attr.evolve(
                 c,
                 author=Person(
                     fullname=b"John Doe <john.doe@example.com>",
                     name=b"John Doe",
                     email=b"john.doe@example.com",
                 ),
             )
             for c in sample_data.releases[:2]
         ]
 
         actual_result = swh_storage.release_add([release, release2])
         assert actual_result == {"release:add": 2}
 
     def test_release_get(self, swh_storage, sample_data):
         release, release2, release3 = sample_data.releases[:3]
 
         # given
         swh_storage.release_add([release, release2])
 
         # when
         releases = list(swh_storage.release_get([release.id, release2.id]))
         actual_releases = [Release.from_dict(r) for r in releases]
 
         # then
         assert actual_releases == [release, release2]
 
         unknown_releases = list(swh_storage.release_get([release3.id]))
         assert unknown_releases[0] is None
 
     def test_release_get_order(self, swh_storage, sample_data):
         release, release2 = sample_data.releases[:2]
 
         add_result = swh_storage.release_add([release, release2])
         assert add_result == {"release:add": 2}
 
         # order 1
         res1 = swh_storage.release_get([release.id, release2.id])
         assert list(res1) == [release.to_dict(), release2.to_dict()]
 
         # order 2
         res2 = swh_storage.release_get([release2.id, release.id])
         assert list(res2) == [release2.to_dict(), release.to_dict()]
 
     def test_release_get_random(self, swh_storage, sample_data):
         release, release2, release3 = sample_data.releases[:3]
 
         swh_storage.release_add([release, release2, release3])
 
         assert swh_storage.release_get_random() in {
             release.id,
             release2.id,
             release3.id,
         }
 
     def test_origin_add(self, swh_storage, sample_data):
         origins = list(sample_data.origins[:2])
         origin_urls = [o.url for o in origins]
 
         assert swh_storage.origin_get(origin_urls) == [None, None]
 
         stats = swh_storage.origin_add(origins)
         assert stats == {"origin:add": 2}
 
         actual_origins = swh_storage.origin_get(origin_urls)
         assert actual_origins == origins
 
         assert set(swh_storage.journal_writer.journal.objects) == set(
             [("origin", origins[0]), ("origin", origins[1]),]
         )
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["origin"] == 2
 
     def test_origin_add_from_generator(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
 
         def _ori_gen():
             yield origin
             yield origin2
 
         stats = swh_storage.origin_add(_ori_gen())
         assert stats == {"origin:add": 2}
 
         actual_origins = swh_storage.origin_get([origin.url, origin2.url])
         assert actual_origins == [origin, origin2]
 
         assert set(swh_storage.journal_writer.journal.objects) == set(
             [("origin", origin), ("origin", origin2),]
         )
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["origin"] == 2
 
     def test_origin_add_twice(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
 
         add1 = swh_storage.origin_add([origin, origin2])
         assert set(swh_storage.journal_writer.journal.objects) == set(
             [("origin", origin), ("origin", origin2),]
         )
         assert add1 == {"origin:add": 2}
 
         add2 = swh_storage.origin_add([origin, origin2])
         assert set(swh_storage.journal_writer.journal.objects) == set(
             [("origin", origin), ("origin", origin2),]
         )
         assert add2 == {"origin:add": 0}
 
     def test_origin_get(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
 
         assert swh_storage.origin_get([origin.url]) == [None]
         swh_storage.origin_add([origin])
 
         actual_origins = swh_storage.origin_get([origin.url])
         assert actual_origins == [origin]
 
         actual_origins = swh_storage.origin_get([origin.url, "not://exists"])
         assert actual_origins == [origin, None]
 
     def _generate_random_visits(self, nb_visits=100, start=0, end=7):
         """Generate random visits within the last 2 months (to avoid
         computations)
 
         """
         visits = []
         today = now()
         for weeks in range(nb_visits, 0, -1):
             hours = random.randint(0, 24)
             minutes = random.randint(0, 60)
             seconds = random.randint(0, 60)
             days = random.randint(0, 28)
             weeks = random.randint(start, end)
             date_visit = today - timedelta(
                 weeks=weeks, hours=hours, minutes=minutes, seconds=seconds, days=days
             )
             visits.append(date_visit)
         return visits
 
     def test_origin_visit_get_all(self, swh_storage, sample_data):
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         visits = swh_storage.origin_visit_add(
             [
                 OriginVisit(
                     origin=origin.url,
                     date=sample_data.date_visit1,
                     type=sample_data.type_visit1,
                 ),
                 OriginVisit(
                     origin=origin.url,
                     date=sample_data.date_visit2,
                     type=sample_data.type_visit2,
                 ),
                 OriginVisit(
                     origin=origin.url,
                     date=sample_data.date_visit2,
                     type=sample_data.type_visit2,
                 ),
             ]
         )
         ov1, ov2, ov3 = [
             {**v.to_dict(), "status": "created", "snapshot": None, "metadata": None,}
             for v in visits
         ]
 
         # order asc, no pagination, no limit
         all_visits = list(swh_storage.origin_visit_get(origin.url))
         assert all_visits == [ov1, ov2, ov3]
 
         # order asc, no pagination, limit
         all_visits2 = list(swh_storage.origin_visit_get(origin.url, limit=2))
         assert all_visits2 == [ov1, ov2]
 
         # order asc, pagination, no limit
         all_visits3 = list(
             swh_storage.origin_visit_get(origin.url, last_visit=ov1["visit"])
         )
         assert all_visits3 == [ov2, ov3]
 
         # order asc, pagination, limit
         all_visits4 = list(
             swh_storage.origin_visit_get(origin.url, last_visit=ov2["visit"], limit=1)
         )
         assert all_visits4 == [ov3]
 
         # order desc, no pagination, no limit
         all_visits5 = list(swh_storage.origin_visit_get(origin.url, order="desc"))
         assert all_visits5 == [ov3, ov2, ov1]
 
         # order desc, no pagination, limit
         all_visits6 = list(
             swh_storage.origin_visit_get(origin.url, limit=2, order="desc")
         )
         assert all_visits6 == [ov3, ov2]
 
         # order desc, pagination, no limit
         all_visits7 = list(
             swh_storage.origin_visit_get(
                 origin.url, last_visit=ov3["visit"], order="desc"
             )
         )
         assert all_visits7 == [ov2, ov1]
 
         # order desc, pagination, limit
         all_visits8 = list(
             swh_storage.origin_visit_get(
                 origin.url, last_visit=ov3["visit"], order="desc", limit=1
             )
         )
         assert all_visits8 == [ov2]
 
     def test_origin_visit_get__unknown_origin(self, swh_storage):
         assert [] == list(swh_storage.origin_visit_get("foo"))
 
     def test_origin_visit_status_get_random(self, swh_storage, sample_data):
         origins = sample_data.origins[:2]
         swh_storage.origin_add(origins)
 
         # Add some random visits within the selection range
         visits = self._generate_random_visits()
         visit_type = "git"
 
         # Add visits to those origins
         for origin in origins:
             for date_visit in visits:
                 visit = swh_storage.origin_visit_add(
                     [OriginVisit(origin=origin.url, date=date_visit, type=visit_type,)]
                 )[0]
                 swh_storage.origin_visit_status_add(
                     [
                         OriginVisitStatus(
                             origin=origin.url,
                             visit=visit.visit,
                             date=now(),
                             status="full",
                             snapshot=None,
                         )
                     ]
                 )
 
         swh_storage.refresh_stat_counters()
 
         stats = swh_storage.stat_counters()
         assert stats["origin"] == len(origins)
         assert stats["origin_visit"] == len(origins) * len(visits)
 
         random_ov, random_ovs = swh_storage.origin_visit_status_get_random(visit_type)
         assert random_ov and random_ovs
         assert random_ov.origin is not None
         assert random_ov.origin == random_ovs.origin
         assert random_ov.origin in [o.url for o in origins]
 
     def test_origin_visit_status_get_random_nothing_found(
         self, swh_storage, sample_data
     ):
         origins = sample_data.origins
         swh_storage.origin_add(origins)
         visit_type = "hg"
         # Add some visits outside of the random generation selection so nothing
         # will be found by the random selection
         visits = self._generate_random_visits(nb_visits=3, start=13, end=24)
         for origin in origins:
             for date_visit in visits:
                 visit = swh_storage.origin_visit_add(
                     [OriginVisit(origin=origin.url, date=date_visit, type=visit_type,)]
                 )[0]
                 swh_storage.origin_visit_status_add(
                     [
                         OriginVisitStatus(
                             origin=origin.url,
                             visit=visit.visit,
                             date=now(),
                             status="full",
                             snapshot=None,
                         )
                     ]
                 )
 
         random_origin_visit = swh_storage.origin_visit_status_get_random(visit_type)
         assert random_origin_visit is None
 
     def test_origin_get_by_sha1(self, swh_storage, sample_data):
         origin = sample_data.origin
         assert swh_storage.origin_get([origin.url])[0] is None
         swh_storage.origin_add([origin])
 
         origins = list(swh_storage.origin_get_by_sha1([sha1(origin.url)]))
         assert len(origins) == 1
         assert origins[0]["url"] == origin.url
 
     def test_origin_get_by_sha1_not_found(self, swh_storage, sample_data):
         origin = sample_data.origin
         assert swh_storage.origin_get([origin.url])[0] is None
         origins = list(swh_storage.origin_get_by_sha1([sha1(origin.url)]))
         assert len(origins) == 1
         assert origins[0] is None
 
     def test_origin_search_single_result(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
 
         found_origins = list(swh_storage.origin_search(origin.url))
         assert len(found_origins) == 0
 
         found_origins = list(swh_storage.origin_search(origin.url, regexp=True))
         assert len(found_origins) == 0
 
         swh_storage.origin_add([origin])
         origin_data = origin.to_dict()
         found_origins = list(swh_storage.origin_search(origin.url))
 
         assert len(found_origins) == 1
         assert found_origins[0] == origin_data
 
         found_origins = list(
             swh_storage.origin_search(f".{origin.url[1:-1]}.", regexp=True)
         )
         assert len(found_origins) == 1
         assert found_origins[0] == origin_data
 
         swh_storage.origin_add([origin2])
         origin2_data = origin2.to_dict()
         found_origins = list(swh_storage.origin_search(origin2.url))
         assert len(found_origins) == 1
         assert found_origins[0] == origin2_data
 
         found_origins = list(
             swh_storage.origin_search(f".{origin2.url[1:-1]}.", regexp=True)
         )
         assert len(found_origins) == 1
         assert found_origins[0] == origin2_data
 
     def test_origin_search_no_regexp(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
         origin_dicts = [o.to_dict() for o in [origin, origin2]]
 
         swh_storage.origin_add([origin, origin2])
 
         # no pagination
         found_origins = list(swh_storage.origin_search("/"))
         assert len(found_origins) == 2
 
         # offset=0
         found_origins0 = list(swh_storage.origin_search("/", offset=0, limit=1))
         assert len(found_origins0) == 1
         assert found_origins0[0] in origin_dicts
 
         # offset=1
         found_origins1 = list(swh_storage.origin_search("/", offset=1, limit=1))
         assert len(found_origins1) == 1
         assert found_origins1[0] in origin_dicts
 
         # check both origins were returned
         assert found_origins0 != found_origins1
 
     def test_origin_search_regexp_substring(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
         origin_dicts = [o.to_dict() for o in [origin, origin2]]
 
         swh_storage.origin_add([origin, origin2])
 
         # no pagination
         found_origins = list(swh_storage.origin_search("/", regexp=True))
         assert len(found_origins) == 2
 
         # offset=0
         found_origins0 = list(
             swh_storage.origin_search("/", offset=0, limit=1, regexp=True)
         )
         assert len(found_origins0) == 1
         assert found_origins0[0] in origin_dicts
 
         # offset=1
         found_origins1 = list(
             swh_storage.origin_search("/", offset=1, limit=1, regexp=True)
         )
         assert len(found_origins1) == 1
         assert found_origins1[0] in origin_dicts
 
         # check both origins were returned
         assert found_origins0 != found_origins1
 
     def test_origin_search_regexp_fullstring(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
         origin_dicts = [o.to_dict() for o in [origin, origin2]]
 
         swh_storage.origin_add([origin, origin2])
 
         # no pagination
         found_origins = list(swh_storage.origin_search(".*/.*", regexp=True))
         assert len(found_origins) == 2
 
         # offset=0
         found_origins0 = list(
             swh_storage.origin_search(".*/.*", offset=0, limit=1, regexp=True)
         )
         assert len(found_origins0) == 1
         assert found_origins0[0] in origin_dicts
 
         # offset=1
         found_origins1 = list(
             swh_storage.origin_search(".*/.*", offset=1, limit=1, regexp=True)
         )
         assert len(found_origins1) == 1
         assert found_origins1[0] in origin_dicts
 
         # check both origins were returned
         assert found_origins0 != found_origins1
 
     def test_origin_visit_add(self, swh_storage, sample_data):
         origin1 = sample_data.origins[1]
         swh_storage.origin_add([origin1])
 
         date_visit = now()
         date_visit2 = date_visit + datetime.timedelta(minutes=1)
 
         date_visit = round_to_milliseconds(date_visit)
         date_visit2 = round_to_milliseconds(date_visit2)
 
         visit1 = OriginVisit(
             origin=origin1.url, date=date_visit, type=sample_data.type_visit1,
         )
         visit2 = OriginVisit(
             origin=origin1.url, date=date_visit2, type=sample_data.type_visit2,
         )
 
         # add once
         ov1, ov2 = swh_storage.origin_visit_add([visit1, visit2])
         # then again (will be ignored as they already exist)
         origin_visit1, origin_visit2 = swh_storage.origin_visit_add([ov1, ov2])
         assert ov1 == origin_visit1
         assert ov2 == origin_visit2
 
         ovs1 = OriginVisitStatus(
             origin=origin1.url,
             visit=ov1.visit,
             date=date_visit,
             status="created",
             snapshot=None,
         )
         ovs2 = OriginVisitStatus(
             origin=origin1.url,
             visit=ov2.visit,
             date=date_visit2,
             status="created",
             snapshot=None,
         )
 
         actual_origin_visits = list(swh_storage.origin_visit_get(origin1.url))
         expected_visits = [
             {**ovs1.to_dict(), "type": ov1.type},
             {**ovs2.to_dict(), "type": ov2.type},
         ]
 
         assert len(expected_visits) == len(actual_origin_visits)
 
         for visit in expected_visits:
             assert visit in actual_origin_visits
 
         actual_objects = list(swh_storage.journal_writer.journal.objects)
         expected_objects = list(
             [("origin", origin1)]
             + [("origin_visit", visit) for visit in [ov1, ov2]] * 2
             + [("origin_visit_status", ovs) for ovs in [ovs1, ovs2]]
         )
 
         for obj in expected_objects:
             assert obj in actual_objects
 
     def test_origin_visit_add_validation(self, swh_storage, sample_data):
         """Unknown origin when adding visits should raise"""
         visit = attr.evolve(sample_data.origin_visit, origin="something-unknonw")
         with pytest.raises(StorageArgumentException, match="Unknown origin"):
             swh_storage.origin_visit_add([visit])
 
         objects = list(swh_storage.journal_writer.journal.objects)
         assert not objects
 
     def test_origin_visit_status_add_validation(self, swh_storage):
         """Wrong origin_visit_status input should raise storage argument error"""
         date_visit = now()
         visit_status1 = OriginVisitStatus(
             origin="unknown-origin-url",
             visit=10,
             date=date_visit,
             status="full",
             snapshot=None,
         )
         with pytest.raises(StorageArgumentException, match="Unknown origin"):
             swh_storage.origin_visit_status_add([visit_status1])
 
         objects = list(swh_storage.journal_writer.journal.objects)
         assert not objects
 
     def test_origin_visit_status_add(self, swh_storage, sample_data):
         """Correct origin visit statuses should add a new visit status
 
         """
         snapshot = sample_data.snapshot
         origin1 = sample_data.origins[1]
         origin2 = Origin(url="new-origin")
         swh_storage.origin_add([origin1, origin2])
 
         ov1, ov2 = swh_storage.origin_visit_add(
             [
                 OriginVisit(
                     origin=origin1.url,
                     date=sample_data.date_visit1,
                     type=sample_data.type_visit1,
                 ),
                 OriginVisit(
                     origin=origin2.url,
                     date=sample_data.date_visit2,
                     type=sample_data.type_visit2,
                 ),
             ]
         )
 
         ovs1 = OriginVisitStatus(
             origin=origin1.url,
             visit=ov1.visit,
             date=sample_data.date_visit1,
             status="created",
             snapshot=None,
         )
         ovs2 = OriginVisitStatus(
             origin=origin2.url,
             visit=ov2.visit,
             date=sample_data.date_visit2,
             status="created",
             snapshot=None,
         )
 
         date_visit_now = round_to_milliseconds(now())
         visit_status1 = OriginVisitStatus(
             origin=ov1.origin,
             visit=ov1.visit,
             date=date_visit_now,
             status="full",
             snapshot=snapshot.id,
         )
 
         date_visit_now = round_to_milliseconds(now())
         visit_status2 = OriginVisitStatus(
             origin=ov2.origin,
             visit=ov2.visit,
             date=date_visit_now,
             status="ongoing",
             snapshot=None,
             metadata={"intrinsic": "something"},
         )
         swh_storage.origin_visit_status_add([visit_status1, visit_status2])
 
         visit = swh_storage.origin_visit_get_latest(origin1.url, require_snapshot=True)
         visit_status = swh_storage.origin_visit_status_get_latest(
             origin1.url, visit.visit, require_snapshot=True
         )
         assert visit_status == visit_status1
 
         visit = swh_storage.origin_visit_get_latest(origin2.url, require_snapshot=False)
         visit_status = swh_storage.origin_visit_status_get_latest(
             origin2.url, visit.visit, require_snapshot=False
         )
         assert origin2.url != origin1.url
         assert visit_status == visit_status2
 
         actual_objects = list(swh_storage.journal_writer.journal.objects)
 
         expected_origins = [origin1, origin2]
         expected_visits = [ov1, ov2]
         expected_visit_statuses = [ovs1, ovs2, visit_status1, visit_status2]
 
         expected_objects = (
             [("origin", o) for o in expected_origins]
             + [("origin_visit", v) for v in expected_visits]
             + [("origin_visit_status", ovs) for ovs in expected_visit_statuses]
         )
 
         for obj in expected_objects:
             assert obj in actual_objects
 
     def test_origin_visit_status_add_twice(self, swh_storage, sample_data):
         """Correct origin visit statuses should add a new visit status
 
         """
         snapshot = sample_data.snapshot
         origin1 = sample_data.origins[1]
         swh_storage.origin_add([origin1])
         ov1 = swh_storage.origin_visit_add(
             [
                 OriginVisit(
                     origin=origin1.url,
                     date=sample_data.date_visit1,
                     type=sample_data.type_visit1,
                 ),
             ]
         )[0]
 
         ovs1 = OriginVisitStatus(
             origin=origin1.url,
             visit=ov1.visit,
             date=sample_data.date_visit1,
             status="created",
             snapshot=None,
         )
         date_visit_now = now()
         visit_status1 = OriginVisitStatus(
             origin=ov1.origin,
             visit=ov1.visit,
             date=date_visit_now,
             status="full",
             snapshot=snapshot.id,
         )
 
         swh_storage.origin_visit_status_add([visit_status1])
         # second call will ignore existing entries (will send to storage though)
         swh_storage.origin_visit_status_add([visit_status1])
 
         origin_visits = list(swh_storage.origin_visit_get(ov1.origin))
 
         assert len(origin_visits) == 1
         origin_visit1 = origin_visits[0]
         assert origin_visit1
         assert origin_visit1["status"] == "full"
         assert origin_visit1["snapshot"] == snapshot.id
 
         actual_objects = list(swh_storage.journal_writer.journal.objects)
 
         expected_origins = [origin1]
         expected_visits = [ov1]
         expected_visit_statuses = [ovs1, visit_status1, visit_status1]
 
         # write twice in the journal
         expected_objects = (
             [("origin", o) for o in expected_origins]
             + [("origin_visit", v) for v in expected_visits]
             + [("origin_visit_status", ovs) for ovs in expected_visit_statuses]
         )
 
         for obj in expected_objects:
             assert obj in actual_objects
 
     def test_origin_visit_find_by_date(self, swh_storage, sample_data):
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         visit1 = OriginVisit(
             origin=origin.url,
             date=sample_data.date_visit2,
             type=sample_data.type_visit1,
         )
         visit2 = OriginVisit(
             origin=origin.url,
             date=sample_data.date_visit3,
             type=sample_data.type_visit2,
         )
         visit3 = OriginVisit(
             origin=origin.url,
             date=sample_data.date_visit2,
             type=sample_data.type_visit3,
         )
         ov1, ov2, ov3 = swh_storage.origin_visit_add([visit1, visit2, visit3])
 
         ovs1 = OriginVisitStatus(
             origin=origin.url,
             visit=ov1.visit,
             date=sample_data.date_visit2,
             status="ongoing",
             snapshot=None,
         )
         ovs2 = OriginVisitStatus(
             origin=origin.url,
             visit=ov2.visit,
             date=sample_data.date_visit3,
             status="ongoing",
             snapshot=None,
         )
         ovs3 = OriginVisitStatus(
             origin=origin.url,
             visit=ov3.visit,
             date=sample_data.date_visit2,
             status="ongoing",
             snapshot=None,
         )
         swh_storage.origin_visit_status_add([ovs1, ovs2, ovs3])
 
         # Simple case
         actual_visit = swh_storage.origin_visit_find_by_date(
             origin.url, sample_data.date_visit3
         )
         assert actual_visit == ov2
 
         # There are two visits at the same date, the latest must be returned
         actual_visit = swh_storage.origin_visit_find_by_date(
             origin.url, sample_data.date_visit2
         )
         assert actual_visit == ov3
 
     def test_origin_visit_find_by_date__unknown_origin(self, swh_storage, sample_data):
         actual_visit = swh_storage.origin_visit_find_by_date(
             "foo", sample_data.date_visit2
         )
         assert actual_visit is None
 
     def test_origin_visit_get_by(self, swh_storage, sample_data):
         snapshot = sample_data.snapshot
         origins = sample_data.origins[:2]
         swh_storage.origin_add(origins)
         origin_url, origin_url2 = [o.url for o in origins]
 
         visit = OriginVisit(
             origin=origin_url,
             date=sample_data.date_visit2,
             type=sample_data.type_visit2,
         )
         origin_visit1 = swh_storage.origin_visit_add([visit])[0]
 
         swh_storage.snapshot_add([snapshot])
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin_url,
                     visit=origin_visit1.visit,
                     date=now(),
                     status="ongoing",
                     snapshot=snapshot.id,
                 )
             ]
         )
 
         # Add some other {origin, visit} entries
         visit2 = OriginVisit(
             origin=origin_url,
             date=sample_data.date_visit3,
             type=sample_data.type_visit3,
         )
         visit3 = OriginVisit(
             origin=origin_url2,
             date=sample_data.date_visit3,
             type=sample_data.type_visit3,
         )
         swh_storage.origin_visit_add([visit2, visit3])
 
         # when
         visit1_metadata = {
             "contents": 42,
             "directories": 22,
         }
 
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin_url,
                     visit=origin_visit1.visit,
                     date=now(),
                     status="full",
                     snapshot=snapshot.id,
                     metadata=visit1_metadata,
                 )
             ]
         )
 
         actual_visit = swh_storage.origin_visit_get_by(origin_url, origin_visit1.visit)
         assert actual_visit == origin_visit1
 
     def test_origin_visit_get_by__no_result(self, swh_storage, sample_data):
         actual_visit = swh_storage.origin_visit_get_by("unknown", 10)  # unknown origin
         assert actual_visit is None
 
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         actual_visit = swh_storage.origin_visit_get_by(origin.url, 999)  # unknown visit
         assert actual_visit is None
 
     def test_origin_visit_get_latest_none(self, swh_storage, sample_data):
         """Origin visit get latest on unknown objects should return nothing
 
         """
         # unknown origin so no result
         assert swh_storage.origin_visit_get_latest("unknown-origin") is None
 
         # unknown type
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         assert swh_storage.origin_visit_get_latest(origin.url, type="unknown") is None
 
     def test_origin_visit_get_latest_filter_type(self, swh_storage, sample_data):
         """Filtering origin visit get latest with filter type should be ok
 
         """
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         visit1 = OriginVisit(
             origin=origin.url, date=sample_data.date_visit1, type="git",
         )
         visit2 = OriginVisit(
             origin=origin.url, date=sample_data.date_visit2, type="hg",
         )
         date_now = round_to_milliseconds(now())
         visit3 = OriginVisit(origin=origin.url, date=date_now, type="hg",)
         assert sample_data.date_visit1 < sample_data.date_visit2
         assert sample_data.date_visit2 < date_now
 
         ov1, ov2, ov3 = swh_storage.origin_visit_add([visit1, visit2, visit3])
 
         # Check type filter is ok
         actual_visit = swh_storage.origin_visit_get_latest(origin.url, type="git")
         assert actual_visit == ov1
         actual_visit = swh_storage.origin_visit_get_latest(origin.url, type="hg")
         assert actual_visit == ov3
         actual_visit_unknown_type = swh_storage.origin_visit_get_latest(
             origin.url, type="npm",  # no visit matching that type
         )
         assert actual_visit_unknown_type is None
 
     def test_origin_visit_get_latest(self, swh_storage, sample_data):
         empty_snapshot, complete_snapshot = sample_data.snapshots[1:3]
         origin = sample_data.origin
 
         swh_storage.origin_add([origin])
         visit1 = OriginVisit(
             origin=origin.url, date=sample_data.date_visit1, type="git",
         )
         visit2 = OriginVisit(
             origin=origin.url, date=sample_data.date_visit2, type="hg",
         )
         date_now = round_to_milliseconds(now())
         visit3 = OriginVisit(origin=origin.url, date=date_now, type="hg",)
         assert visit1.date < visit2.date
         assert visit2.date < visit3.date
 
         ov1, ov2, ov3 = swh_storage.origin_visit_add([visit1, visit2, visit3])
 
         # no filters, latest visit is the last one (whose date is most recent)
         actual_visit = swh_storage.origin_visit_get_latest(origin.url)
         assert actual_visit == ov3
 
         # 3 visits, none has snapshot so nothing is returned
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, require_snapshot=True
         )
         assert actual_visit is None
 
         # visit are created with "created" status, so nothing will get returned
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["partial"]
         )
         assert actual_visit is None
 
         # visit are created with "created" status, so most recent again
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["created"]
         )
         assert actual_visit == ov3
 
         # Add snapshot to visit1; require_snapshot=True makes it return first visit
         swh_storage.snapshot_add([complete_snapshot])
         visit_status_with_snapshot = OriginVisitStatus(
             origin=origin.url,
             visit=ov1.visit,
             date=round_to_milliseconds(now()),
             status="ongoing",
             snapshot=complete_snapshot.id,
         )
         swh_storage.origin_visit_status_add([visit_status_with_snapshot])
         # only the first visit has a snapshot now
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, require_snapshot=True
         )
         assert actual_visit == ov1
 
         # only the first visit has a status ongoing now
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["ongoing"]
         )
         assert actual_visit == ov1
 
         actual_visit_status = swh_storage.origin_visit_status_get_latest(
             origin.url, ov1.visit, require_snapshot=True
         )
         assert actual_visit_status == visit_status_with_snapshot
 
         # ... and require_snapshot=False (defaults) still returns latest visit (3rd)
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, require_snapshot=False
         )
         assert actual_visit == ov3
         # no specific filter, this returns as before the latest visit
         actual_visit = swh_storage.origin_visit_get_latest(origin.url)
         assert actual_visit == ov3
 
         # Status filter: all three visits are status=ongoing, so no visit
         # returned
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["full"]
         )
         assert actual_visit is None
 
         visit_status1_full = OriginVisitStatus(
             origin=origin.url,
             visit=ov1.visit,
             date=round_to_milliseconds(now()),
             status="full",
             snapshot=complete_snapshot.id,
         )
         # Mark the first visit as completed and check status filter again
         swh_storage.origin_visit_status_add([visit_status1_full])
 
         # only the first visit has the full status
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["full"]
         )
         assert actual_visit == ov1
 
         actual_visit_status = swh_storage.origin_visit_status_get_latest(
             origin.url, ov1.visit, allowed_statuses=["full"]
         )
         assert actual_visit_status == visit_status1_full
 
         # no specific filter, this returns as before the latest visit
         actual_visit = swh_storage.origin_visit_get_latest(origin.url)
         assert actual_visit == ov3
 
         # Add snapshot to visit2 and check that the new snapshot is returned
         swh_storage.snapshot_add([empty_snapshot])
 
         visit_status2_full = OriginVisitStatus(
             origin=origin.url,
             visit=ov2.visit,
             date=round_to_milliseconds(now()),
             status="ongoing",
             snapshot=empty_snapshot.id,
         )
         swh_storage.origin_visit_status_add([visit_status2_full])
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, require_snapshot=True
         )
         # 2nd visit is most recent with a snapshot
         assert actual_visit == ov2
         actual_visit_status = swh_storage.origin_visit_status_get_latest(
             origin.url, ov2.visit, require_snapshot=True
         )
         assert actual_visit_status == visit_status2_full
 
         # no specific filter, this returns as before the latest visit, 3rd one
         actual_origin = swh_storage.origin_visit_get_latest(origin.url)
         assert actual_origin == ov3
 
         # full status is still the first visit
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["full"]
         )
         assert actual_visit == ov1
 
         # Add snapshot to visit3 (same date as visit2)
         visit_status3_with_snapshot = OriginVisitStatus(
             origin=origin.url,
             visit=ov3.visit,
             date=round_to_milliseconds(now()),
             status="ongoing",
             snapshot=complete_snapshot.id,
         )
         swh_storage.origin_visit_status_add([visit_status3_with_snapshot])
 
         # full status is still the first visit
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, allowed_statuses=["full"], require_snapshot=True,
         )
         assert actual_visit == ov1
 
         actual_visit_status = swh_storage.origin_visit_status_get_latest(
             origin.url,
             visit=actual_visit.visit,
             allowed_statuses=["full"],
             require_snapshot=True,
         )
         assert actual_visit_status == visit_status1_full
 
         # most recent is still the 3rd visit
         actual_visit = swh_storage.origin_visit_get_latest(origin.url)
         assert actual_visit == ov3
 
         # 3rd visit has a snapshot now, so it's elected
         actual_visit = swh_storage.origin_visit_get_latest(
             origin.url, require_snapshot=True
         )
         assert actual_visit == ov3
 
         actual_visit_status = swh_storage.origin_visit_status_get_latest(
             origin.url, ov3.visit, require_snapshot=True
         )
         assert actual_visit_status == visit_status3_with_snapshot
 
+    def test_origin_visit_get_latest__same_date(self, swh_storage, sample_data):
+        empty_snapshot, complete_snapshot = sample_data.snapshots[1:3]
+        origin = sample_data.origin
+
+        swh_storage.origin_add([origin])
+        visit1 = OriginVisit(
+            origin=origin.url, date=sample_data.date_visit1, type="git",
+        )
+        visit2 = OriginVisit(
+            origin=origin.url, date=sample_data.date_visit1, type="hg",
+        )
+
+        ov1, ov2 = swh_storage.origin_visit_add([visit1, visit2])
+
+        # ties should be broken by using the visit id
+        actual_visit = swh_storage.origin_visit_get_latest(origin.url)
+        assert actual_visit == ov2
+
     def test_origin_visit_status_get_latest(self, swh_storage, sample_data):
         snapshot = sample_data.snapshots[2]
         origin1 = sample_data.origin
         swh_storage.origin_add([origin1])
 
         # to have some reference visits
 
         ov1, ov2 = swh_storage.origin_visit_add(
             [
                 OriginVisit(
                     origin=origin1.url,
                     date=sample_data.date_visit1,
                     type=sample_data.type_visit1,
                 ),
                 OriginVisit(
                     origin=origin1.url,
                     date=sample_data.date_visit2,
                     type=sample_data.type_visit2,
                 ),
             ]
         )
         swh_storage.snapshot_add([snapshot])
 
         date_now = round_to_milliseconds(now())
         assert sample_data.date_visit1 < sample_data.date_visit2
         assert sample_data.date_visit2 < date_now
 
         ovs1 = OriginVisitStatus(
             origin=origin1.url,
             visit=ov1.visit,
             date=sample_data.date_visit1,
             status="partial",
             snapshot=None,
         )
         ovs2 = OriginVisitStatus(
             origin=origin1.url,
             visit=ov1.visit,
             date=sample_data.date_visit2,
             status="ongoing",
             snapshot=None,
         )
         ovs3 = OriginVisitStatus(
             origin=origin1.url,
             visit=ov2.visit,
             date=sample_data.date_visit2
             + datetime.timedelta(minutes=1),  # to not be ignored
             status="ongoing",
             snapshot=None,
         )
         ovs4 = OriginVisitStatus(
             origin=origin1.url,
             visit=ov2.visit,
             date=date_now,
             status="full",
             snapshot=snapshot.id,
             metadata={"something": "wicked"},
         )
 
         swh_storage.origin_visit_status_add([ovs1, ovs2, ovs3, ovs4])
 
         # unknown origin so no result
         actual_origin_visit = swh_storage.origin_visit_status_get_latest(
             "unknown-origin", ov1.visit
         )
         assert actual_origin_visit is None
 
         # unknown visit so no result
         actual_origin_visit = swh_storage.origin_visit_status_get_latest(
             ov1.origin, ov1.visit + 10
         )
         assert actual_origin_visit is None
 
         # Two visits, both with no snapshot, take the most recent
         actual_origin_visit2 = swh_storage.origin_visit_status_get_latest(
             origin1.url, ov1.visit
         )
         assert isinstance(actual_origin_visit2, OriginVisitStatus)
         assert actual_origin_visit2 == ovs2
         assert ovs2.origin == origin1.url
         assert ovs2.visit == ov1.visit
 
         actual_origin_visit = swh_storage.origin_visit_status_get_latest(
             origin1.url, ov1.visit, require_snapshot=True
         )
         # there is no visit with snapshot yet for that visit
         assert actual_origin_visit is None
 
         actual_origin_visit2 = swh_storage.origin_visit_status_get_latest(
             origin1.url, ov1.visit, allowed_statuses=["partial", "ongoing"]
         )
         # visit status with partial status visit elected
         assert actual_origin_visit2 == ovs2
         assert actual_origin_visit2.status == "ongoing"
 
         actual_origin_visit4 = swh_storage.origin_visit_status_get_latest(
             origin1.url, ov2.visit, require_snapshot=True
         )
         assert actual_origin_visit4 == ovs4
         assert actual_origin_visit4.snapshot == snapshot.id
 
         actual_origin_visit = swh_storage.origin_visit_status_get_latest(
             origin1.url, ov2.visit, require_snapshot=True, allowed_statuses=["ongoing"]
         )
         # nothing matches so nothing
         assert actual_origin_visit is None  # there is no visit with status full
 
         actual_origin_visit3 = swh_storage.origin_visit_status_get_latest(
             origin1.url, ov2.visit, allowed_statuses=["ongoing"]
         )
         assert actual_origin_visit3 == ovs3
 
     def test_person_fullname_unicity(self, swh_storage, sample_data):
         revision, rev2 = sample_data.revisions[0:2]
         # create a revision with same committer fullname but wo name and email
         revision2 = attr.evolve(
             rev2,
             committer=Person(
                 fullname=revision.committer.fullname, name=None, email=None
             ),
         )
 
         swh_storage.revision_add([revision, revision2])
 
         # when getting added revisions
         revisions = list(swh_storage.revision_get([revision.id, revision2.id]))
 
         # then check committers are the same
         assert revisions[0]["committer"] == revisions[1]["committer"]
 
     def test_snapshot_add_get_empty(self, swh_storage, sample_data):
         empty_snapshot = sample_data.snapshots[1]
         empty_snapshot_dict = empty_snapshot.to_dict()
 
         origin = sample_data.origin
         swh_storage.origin_add([origin])
         ov1 = swh_storage.origin_visit_add(
             [
                 OriginVisit(
                     origin=origin.url,
                     date=sample_data.date_visit1,
                     type=sample_data.type_visit1,
                 )
             ]
         )[0]
 
         actual_result = swh_storage.snapshot_add([empty_snapshot])
         assert actual_result == {"snapshot:add": 1}
 
         date_now = now()
 
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin.url,
                     visit=ov1.visit,
                     date=date_now,
                     status="full",
                     snapshot=empty_snapshot.id,
                 )
             ]
         )
 
         by_id = swh_storage.snapshot_get(empty_snapshot.id)
         assert by_id == {**empty_snapshot_dict, "next_branch": None}
 
         by_ov = swh_storage.snapshot_get_by_origin_visit(origin.url, ov1.visit)
         assert by_ov == {**empty_snapshot_dict, "next_branch": None}
 
         ovs1 = OriginVisitStatus.from_dict(
             {
                 "origin": origin.url,
                 "date": sample_data.date_visit1,
                 "visit": ov1.visit,
                 "status": "created",
                 "snapshot": None,
                 "metadata": None,
             }
         )
         ovs2 = OriginVisitStatus.from_dict(
             {
                 "origin": origin.url,
                 "date": date_now,
                 "visit": ov1.visit,
                 "status": "full",
                 "metadata": None,
                 "snapshot": empty_snapshot.id,
             }
         )
         actual_objects = list(swh_storage.journal_writer.journal.objects)
 
         expected_objects = [
             ("origin", origin),
             ("origin_visit", ov1),
             ("origin_visit_status", ovs1,),
             ("snapshot", empty_snapshot),
             ("origin_visit_status", ovs2,),
         ]
         for obj in expected_objects:
             assert obj in actual_objects
 
     def test_snapshot_add_get_complete(self, swh_storage, sample_data):
         complete_snapshot = sample_data.snapshots[2]
         complete_snapshot_dict = complete_snapshot.to_dict()
         origin = sample_data.origin
 
         swh_storage.origin_add([origin])
         visit = OriginVisit(
             origin=origin.url,
             date=sample_data.date_visit1,
             type=sample_data.type_visit1,
         )
         origin_visit1 = swh_storage.origin_visit_add([visit])[0]
         visit_id = origin_visit1.visit
 
         actual_result = swh_storage.snapshot_add([complete_snapshot])
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin.url,
                     visit=origin_visit1.visit,
                     date=now(),
                     status="ongoing",
                     snapshot=complete_snapshot.id,
                 )
             ]
         )
         assert actual_result == {"snapshot:add": 1}
 
         by_id = swh_storage.snapshot_get(complete_snapshot.id)
         assert by_id == {**complete_snapshot_dict, "next_branch": None}
 
         by_ov = swh_storage.snapshot_get_by_origin_visit(origin.url, visit_id)
         assert by_ov == {**complete_snapshot_dict, "next_branch": None}
 
     def test_snapshot_add_many(self, swh_storage, sample_data):
         snapshot, _, complete_snapshot = sample_data.snapshots[:3]
 
         actual_result = swh_storage.snapshot_add([snapshot, complete_snapshot])
         assert actual_result == {"snapshot:add": 2}
 
         assert swh_storage.snapshot_get(complete_snapshot.id) == {
             **complete_snapshot.to_dict(),
             "next_branch": None,
         }
 
         assert swh_storage.snapshot_get(snapshot.id) == {
             **snapshot.to_dict(),
             "next_branch": None,
         }
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["snapshot"] == 2
 
     def test_snapshot_add_many_from_generator(self, swh_storage, sample_data):
         snapshot, _, complete_snapshot = sample_data.snapshots[:3]
 
         def _snp_gen():
             yield from [snapshot, complete_snapshot]
 
         actual_result = swh_storage.snapshot_add(_snp_gen())
         assert actual_result == {"snapshot:add": 2}
 
         swh_storage.refresh_stat_counters()
         assert swh_storage.stat_counters()["snapshot"] == 2
 
     def test_snapshot_add_many_incremental(self, swh_storage, sample_data):
         snapshot, _, complete_snapshot = sample_data.snapshots[:3]
 
         actual_result = swh_storage.snapshot_add([complete_snapshot])
         assert actual_result == {"snapshot:add": 1}
 
         actual_result2 = swh_storage.snapshot_add([snapshot, complete_snapshot])
         assert actual_result2 == {"snapshot:add": 1}
 
         assert swh_storage.snapshot_get(complete_snapshot.id) == {
             **complete_snapshot.to_dict(),
             "next_branch": None,
         }
 
         assert swh_storage.snapshot_get(snapshot.id) == {
             **snapshot.to_dict(),
             "next_branch": None,
         }
 
     def test_snapshot_add_twice(self, swh_storage, sample_data):
         snapshot, empty_snapshot = sample_data.snapshots[:2]
 
         actual_result = swh_storage.snapshot_add([empty_snapshot])
         assert actual_result == {"snapshot:add": 1}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("snapshot", empty_snapshot)
         ]
 
         actual_result = swh_storage.snapshot_add([snapshot])
         assert actual_result == {"snapshot:add": 1}
 
         assert list(swh_storage.journal_writer.journal.objects) == [
             ("snapshot", empty_snapshot),
             ("snapshot", snapshot),
         ]
 
     def test_snapshot_add_count_branches(self, swh_storage, sample_data):
         complete_snapshot = sample_data.snapshots[2]
 
         actual_result = swh_storage.snapshot_add([complete_snapshot])
         assert actual_result == {"snapshot:add": 1}
 
         snp_size = swh_storage.snapshot_count_branches(complete_snapshot.id)
 
         expected_snp_size = {
             "alias": 1,
             "content": 1,
             "directory": 2,
             "release": 1,
             "revision": 1,
             "snapshot": 1,
             None: 1,
         }
         assert snp_size == expected_snp_size
 
     def test_snapshot_add_get_paginated(self, swh_storage, sample_data):
         complete_snapshot = sample_data.snapshots[2]
 
         swh_storage.snapshot_add([complete_snapshot])
 
         snp_id = complete_snapshot.id
         branches = complete_snapshot.to_dict()["branches"]
         branch_names = list(sorted(branches))
 
         # Test branch_from
         snapshot = swh_storage.snapshot_get_branches(snp_id, branches_from=b"release")
 
         rel_idx = branch_names.index(b"release")
         expected_snapshot = {
             "id": snp_id,
             "branches": {name: branches[name] for name in branch_names[rel_idx:]},
             "next_branch": None,
         }
 
         assert snapshot == expected_snapshot
 
         # Test branches_count
         snapshot = swh_storage.snapshot_get_branches(snp_id, branches_count=1)
 
         expected_snapshot = {
             "id": snp_id,
             "branches": {branch_names[0]: branches[branch_names[0]],},
             "next_branch": b"content",
         }
         assert snapshot == expected_snapshot
 
         # test branch_from + branches_count
 
         snapshot = swh_storage.snapshot_get_branches(
             snp_id, branches_from=b"directory", branches_count=3
         )
 
         dir_idx = branch_names.index(b"directory")
         expected_snapshot = {
             "id": snp_id,
             "branches": {
                 name: branches[name] for name in branch_names[dir_idx : dir_idx + 3]
             },
             "next_branch": branch_names[dir_idx + 3],
         }
 
         assert snapshot == expected_snapshot
 
     def test_snapshot_add_get_filtered(self, swh_storage, sample_data):
         origin = sample_data.origin
         complete_snapshot = sample_data.snapshots[2]
 
         swh_storage.origin_add([origin])
         visit = OriginVisit(
             origin=origin.url,
             date=sample_data.date_visit1,
             type=sample_data.type_visit1,
         )
         origin_visit1 = swh_storage.origin_visit_add([visit])[0]
 
         swh_storage.snapshot_add([complete_snapshot])
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin.url,
                     visit=origin_visit1.visit,
                     date=now(),
                     status="ongoing",
                     snapshot=complete_snapshot.id,
                 )
             ]
         )
 
         snp_id = complete_snapshot.id
         branches = complete_snapshot.to_dict()["branches"]
 
         snapshot = swh_storage.snapshot_get_branches(
             snp_id, target_types=["release", "revision"]
         )
 
         expected_snapshot = {
             "id": snp_id,
             "branches": {
                 name: tgt
                 for name, tgt in branches.items()
                 if tgt and tgt["target_type"] in ["release", "revision"]
             },
             "next_branch": None,
         }
 
         assert snapshot == expected_snapshot
 
         snapshot = swh_storage.snapshot_get_branches(snp_id, target_types=["alias"])
 
         expected_snapshot = {
             "id": snp_id,
             "branches": {
                 name: tgt
                 for name, tgt in branches.items()
                 if tgt and tgt["target_type"] == "alias"
             },
             "next_branch": None,
         }
 
         assert snapshot == expected_snapshot
 
     def test_snapshot_add_get_filtered_and_paginated(self, swh_storage, sample_data):
         complete_snapshot = sample_data.snapshots[2]
 
         swh_storage.snapshot_add([complete_snapshot])
 
         snp_id = complete_snapshot.id
         branches = complete_snapshot.to_dict()["branches"]
         branch_names = list(sorted(branches))
 
         # Test branch_from
 
         snapshot = swh_storage.snapshot_get_branches(
             snp_id, target_types=["directory", "release"], branches_from=b"directory2"
         )
 
         expected_snapshot = {
             "id": snp_id,
             "branches": {name: branches[name] for name in (b"directory2", b"release")},
             "next_branch": None,
         }
 
         assert snapshot == expected_snapshot
 
         # Test branches_count
 
         snapshot = swh_storage.snapshot_get_branches(
             snp_id, target_types=["directory", "release"], branches_count=1
         )
 
         expected_snapshot = {
             "id": snp_id,
             "branches": {b"directory": branches[b"directory"]},
             "next_branch": b"directory2",
         }
         assert snapshot == expected_snapshot
 
         # Test branches_count
 
         snapshot = swh_storage.snapshot_get_branches(
             snp_id, target_types=["directory", "release"], branches_count=2
         )
 
         expected_snapshot = {
             "id": snp_id,
             "branches": {
                 name: branches[name] for name in (b"directory", b"directory2")
             },
             "next_branch": b"release",
         }
         assert snapshot == expected_snapshot
 
         # test branch_from + branches_count
 
         snapshot = swh_storage.snapshot_get_branches(
             snp_id,
             target_types=["directory", "release"],
             branches_from=b"directory2",
             branches_count=1,
         )
 
         dir_idx = branch_names.index(b"directory2")
         expected_snapshot = {
             "id": snp_id,
             "branches": {branch_names[dir_idx]: branches[branch_names[dir_idx]],},
             "next_branch": b"release",
         }
 
         assert snapshot == expected_snapshot
 
     def test_snapshot_add_get_branch_by_type(self, swh_storage, sample_data):
         complete_snapshot = sample_data.snapshots[2]
         snapshot = complete_snapshot.to_dict()
 
         alias1 = b"alias1"
         alias2 = b"alias2"
         target1 = random.choice(list(snapshot["branches"].keys()))
         target2 = random.choice(list(snapshot["branches"].keys()))
 
         snapshot["branches"][alias2] = {
             "target": target2,
             "target_type": "alias",
         }
 
         snapshot["branches"][alias1] = {
             "target": target1,
             "target_type": "alias",
         }
 
         new_snapshot = Snapshot.from_dict(snapshot)
         swh_storage.snapshot_add([new_snapshot])
 
         branches = swh_storage.snapshot_get_branches(
             new_snapshot.id,
             target_types=["alias"],
             branches_from=alias1,
             branches_count=1,
         )["branches"]
 
         assert len(branches) == 1
         assert alias1 in branches
 
     def test_snapshot_add_get(self, swh_storage, sample_data):
         snapshot = sample_data.snapshot
         origin = sample_data.origin
 
         swh_storage.origin_add([origin])
         visit = OriginVisit(
             origin=origin.url,
             date=sample_data.date_visit1,
             type=sample_data.type_visit1,
         )
         ov1 = swh_storage.origin_visit_add([visit])[0]
 
         swh_storage.snapshot_add([snapshot])
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin.url,
                     visit=ov1.visit,
                     date=now(),
                     status="ongoing",
                     snapshot=snapshot.id,
                 )
             ]
         )
 
         expected_snapshot = {**snapshot.to_dict(), "next_branch": None}
 
         by_id = swh_storage.snapshot_get(snapshot.id)
         assert by_id == expected_snapshot
 
         by_ov = swh_storage.snapshot_get_by_origin_visit(origin.url, ov1.visit)
         assert by_ov == expected_snapshot
 
         actual_visit = swh_storage.origin_visit_get_by(origin.url, ov1.visit)
         assert actual_visit == ov1
 
         visit_status = swh_storage.origin_visit_status_get_latest(
             origin.url, ov1.visit, require_snapshot=True
         )
         assert visit_status.snapshot == snapshot.id
 
     def test_snapshot_add_twice__by_origin_visit(self, swh_storage, sample_data):
         snapshot = sample_data.snapshot
         origin = sample_data.origin
 
         swh_storage.origin_add([origin])
         ov1 = swh_storage.origin_visit_add(
             [
                 OriginVisit(
                     origin=origin.url,
                     date=sample_data.date_visit1,
                     type=sample_data.type_visit1,
                 )
             ]
         )[0]
         swh_storage.snapshot_add([snapshot])
         date_now2 = now()
 
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin.url,
                     visit=ov1.visit,
                     date=date_now2,
                     status="ongoing",
                     snapshot=snapshot.id,
                 )
             ]
         )
 
         expected_snapshot = {**snapshot.to_dict(), "next_branch": None}
 
         by_ov1 = swh_storage.snapshot_get_by_origin_visit(origin.url, ov1.visit)
         assert by_ov1 == expected_snapshot
 
         ov2 = swh_storage.origin_visit_add(
             [
                 OriginVisit(
                     origin=origin.url,
                     date=sample_data.date_visit2,
                     type=sample_data.type_visit2,
                 )
             ]
         )[0]
 
         date_now4 = now()
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin.url,
                     visit=ov2.visit,
                     date=date_now4,
                     status="ongoing",
                     snapshot=snapshot.id,
                 )
             ]
         )
 
         by_ov2 = swh_storage.snapshot_get_by_origin_visit(origin.url, ov2.visit)
         assert by_ov2 == expected_snapshot
 
         ovs1 = OriginVisitStatus.from_dict(
             {
                 "origin": origin.url,
                 "date": sample_data.date_visit1,
                 "visit": ov1.visit,
                 "status": "created",
                 "metadata": None,
                 "snapshot": None,
             }
         )
         ovs2 = OriginVisitStatus.from_dict(
             {
                 "origin": origin.url,
                 "date": date_now2,
                 "visit": ov1.visit,
                 "status": "ongoing",
                 "metadata": None,
                 "snapshot": snapshot.id,
             }
         )
         ovs3 = OriginVisitStatus.from_dict(
             {
                 "origin": origin.url,
                 "date": sample_data.date_visit2,
                 "visit": ov2.visit,
                 "status": "created",
                 "metadata": None,
                 "snapshot": None,
             }
         )
         ovs4 = OriginVisitStatus.from_dict(
             {
                 "origin": origin.url,
                 "date": date_now4,
                 "visit": ov2.visit,
                 "status": "ongoing",
                 "metadata": None,
                 "snapshot": snapshot.id,
             }
         )
         actual_objects = list(swh_storage.journal_writer.journal.objects)
         expected_objects = [
             ("origin", origin),
             ("origin_visit", ov1),
             ("origin_visit_status", ovs1),
             ("snapshot", snapshot),
             ("origin_visit_status", ovs2),
             ("origin_visit", ov2),
             ("origin_visit_status", ovs3),
             ("origin_visit_status", ovs4),
         ]
 
         for obj in expected_objects:
             assert obj in actual_objects
 
     def test_snapshot_get_random(self, swh_storage, sample_data):
         snapshot, empty_snapshot, complete_snapshot = sample_data.snapshots[:3]
         swh_storage.snapshot_add([snapshot, empty_snapshot, complete_snapshot])
 
         assert swh_storage.snapshot_get_random() in {
             snapshot.id,
             empty_snapshot.id,
             complete_snapshot.id,
         }
 
     def test_snapshot_missing(self, swh_storage, sample_data):
         snapshot, missing_snapshot = sample_data.snapshots[:2]
         snapshots = [snapshot.id, missing_snapshot.id]
         swh_storage.snapshot_add([snapshot])
 
         missing_snapshots = swh_storage.snapshot_missing(snapshots)
 
         assert list(missing_snapshots) == [missing_snapshot.id]
 
     def test_stat_counters(self, swh_storage, sample_data):
         origin = sample_data.origin
         snapshot = sample_data.snapshot
         revision = sample_data.revision
         release = sample_data.release
         directory = sample_data.directory
         content = sample_data.content
 
         expected_keys = ["content", "directory", "origin", "revision"]
 
         # Initially, all counters are 0
 
         swh_storage.refresh_stat_counters()
         counters = swh_storage.stat_counters()
         assert set(expected_keys) <= set(counters)
         for key in expected_keys:
             assert counters[key] == 0
 
         # Add a content. Only the content counter should increase.
 
         swh_storage.content_add([content])
 
         swh_storage.refresh_stat_counters()
         counters = swh_storage.stat_counters()
 
         assert set(expected_keys) <= set(counters)
         for key in expected_keys:
             if key != "content":
                 assert counters[key] == 0
         assert counters["content"] == 1
 
         # Add other objects. Check their counter increased as well.
 
         swh_storage.origin_add([origin])
         visit = OriginVisit(
             origin=origin.url,
             date=sample_data.date_visit2,
             type=sample_data.type_visit2,
         )
         origin_visit1 = swh_storage.origin_visit_add([visit])[0]
 
         swh_storage.snapshot_add([snapshot])
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin.url,
                     visit=origin_visit1.visit,
                     date=now(),
                     status="ongoing",
                     snapshot=snapshot.id,
                 )
             ]
         )
         swh_storage.directory_add([directory])
         swh_storage.revision_add([revision])
         swh_storage.release_add([release])
 
         swh_storage.refresh_stat_counters()
         counters = swh_storage.stat_counters()
         assert counters["content"] == 1
         assert counters["directory"] == 1
         assert counters["snapshot"] == 1
         assert counters["origin"] == 1
         assert counters["origin_visit"] == 1
         assert counters["revision"] == 1
         assert counters["release"] == 1
         assert counters["snapshot"] == 1
         if "person" in counters:
             assert counters["person"] == 3
 
     def test_content_find_ctime(self, swh_storage, sample_data):
         origin_content = sample_data.content
         ctime = round_to_milliseconds(now())
         content = attr.evolve(origin_content, data=None, ctime=ctime)
         swh_storage.content_add_metadata([content])
 
         actually_present = swh_storage.content_find({"sha1": content.sha1})
         assert actually_present[0] == content.to_dict()
 
     def test_content_find_with_present_content(self, swh_storage, sample_data):
         content = sample_data.content
         expected_content = content.to_dict()
         del expected_content["data"]
         del expected_content["ctime"]
 
         # 1. with something to find
         swh_storage.content_add([content])
 
         actually_present = swh_storage.content_find({"sha1": content.sha1})
         assert 1 == len(actually_present)
         actually_present[0].pop("ctime")
         assert actually_present[0] == expected_content
 
         # 2. with something to find
         actually_present = swh_storage.content_find({"sha1_git": content.sha1_git})
         assert 1 == len(actually_present)
         actually_present[0].pop("ctime")
         assert actually_present[0] == expected_content
 
         # 3. with something to find
         actually_present = swh_storage.content_find({"sha256": content.sha256})
         assert 1 == len(actually_present)
         actually_present[0].pop("ctime")
         assert actually_present[0] == expected_content
 
         # 4. with something to find
         actually_present = swh_storage.content_find(content.hashes())
         assert 1 == len(actually_present)
         actually_present[0].pop("ctime")
         assert actually_present[0] == expected_content
 
     def test_content_find_with_non_present_content(self, swh_storage, sample_data):
         missing_content = sample_data.skipped_content
         # 1. with something that does not exist
         actually_present = swh_storage.content_find({"sha1": missing_content.sha1})
 
         assert actually_present == []
 
         # 2. with something that does not exist
         actually_present = swh_storage.content_find(
             {"sha1_git": missing_content.sha1_git}
         )
         assert actually_present == []
 
         # 3. with something that does not exist
         actually_present = swh_storage.content_find({"sha256": missing_content.sha256})
         assert actually_present == []
 
     def test_content_find_with_duplicate_input(self, swh_storage, sample_data):
         content = sample_data.content
 
         # Create fake data with colliding sha256 and blake2s256
         sha1_array = bytearray(content.sha1)
         sha1_array[0] += 1
         sha1git_array = bytearray(content.sha1_git)
         sha1git_array[0] += 1
         duplicated_content = attr.evolve(
             content, sha1=bytes(sha1_array), sha1_git=bytes(sha1git_array)
         )
 
         # Inject the data
         swh_storage.content_add([content, duplicated_content])
 
         actual_result = list(
             swh_storage.content_find(
                 {
                     "blake2s256": duplicated_content.blake2s256,
                     "sha256": duplicated_content.sha256,
                 }
             )
         )
 
         expected_content = content.to_dict()
         expected_duplicated_content = duplicated_content.to_dict()
 
         for key in ["data", "ctime"]:  # so we can compare
             for dict_ in [
                 expected_content,
                 expected_duplicated_content,
                 actual_result[0],
                 actual_result[1],
             ]:
                 dict_.pop(key, None)
 
         expected_result = [expected_content, expected_duplicated_content]
         for result in expected_result:
             assert result in actual_result
 
     def test_content_find_with_duplicate_sha256(self, swh_storage, sample_data):
         content = sample_data.content
 
         hashes = {}
         # Create fake data with colliding sha256
         for hashalgo in ("sha1", "sha1_git", "blake2s256"):
             value = bytearray(getattr(content, hashalgo))
             value[0] += 1
             hashes[hashalgo] = bytes(value)
 
         duplicated_content = attr.evolve(
             content,
             sha1=hashes["sha1"],
             sha1_git=hashes["sha1_git"],
             blake2s256=hashes["blake2s256"],
         )
         swh_storage.content_add([content, duplicated_content])
 
         actual_result = list(
             swh_storage.content_find({"sha256": duplicated_content.sha256})
         )
 
         assert len(actual_result) == 2
 
         expected_content = content.to_dict()
         expected_duplicated_content = duplicated_content.to_dict()
 
         for key in ["data", "ctime"]:  # so we can compare
             for dict_ in [
                 expected_content,
                 expected_duplicated_content,
                 actual_result[0],
                 actual_result[1],
             ]:
                 dict_.pop(key, None)
 
         assert sorted(actual_result, key=lambda x: x["sha1"]) == [
             expected_content,
             expected_duplicated_content,
         ]
 
         # Find with both sha256 and blake2s256
         actual_result = list(
             swh_storage.content_find(
                 {
                     "sha256": duplicated_content.sha256,
                     "blake2s256": duplicated_content.blake2s256,
                 }
             )
         )
 
         assert len(actual_result) == 1
         actual_result[0].pop("ctime")
 
         assert actual_result == [expected_duplicated_content]
 
     def test_content_find_with_duplicate_blake2s256(self, swh_storage, sample_data):
         content = sample_data.content
 
         # Create fake data with colliding sha256 and blake2s256
         sha1_array = bytearray(content.sha1)
         sha1_array[0] += 1
         sha1git_array = bytearray(content.sha1_git)
         sha1git_array[0] += 1
         sha256_array = bytearray(content.sha256)
         sha256_array[0] += 1
 
         duplicated_content = attr.evolve(
             content,
             sha1=bytes(sha1_array),
             sha1_git=bytes(sha1git_array),
             sha256=bytes(sha256_array),
         )
 
         swh_storage.content_add([content, duplicated_content])
 
         actual_result = list(
             swh_storage.content_find({"blake2s256": duplicated_content.blake2s256})
         )
 
         expected_content = content.to_dict()
         expected_duplicated_content = duplicated_content.to_dict()
 
         for key in ["data", "ctime"]:  # so we can compare
             for dict_ in [
                 expected_content,
                 expected_duplicated_content,
                 actual_result[0],
                 actual_result[1],
             ]:
                 dict_.pop(key, None)
 
         expected_result = [expected_content, expected_duplicated_content]
         for result in expected_result:
             assert result in actual_result
 
         # Find with both sha256 and blake2s256
         actual_result = list(
             swh_storage.content_find(
                 {
                     "sha256": duplicated_content.sha256,
                     "blake2s256": duplicated_content.blake2s256,
                 }
             )
         )
 
         actual_result[0].pop("ctime")
         assert actual_result == [expected_duplicated_content]
 
     def test_content_find_bad_input(self, swh_storage):
         # 1. with bad input
         with pytest.raises(StorageArgumentException):
             swh_storage.content_find({})  # empty is bad
 
         # 2. with bad input
         with pytest.raises(StorageArgumentException):
             swh_storage.content_find({"unknown-sha1": "something"})  # not the right key
 
     def test_object_find_by_sha1_git(self, swh_storage, sample_data):
         content = sample_data.content
         directory = sample_data.directory
         revision = sample_data.revision
         release = sample_data.release
 
         sha1_gits = [b"00000000000000000000"]
         expected = {
             b"00000000000000000000": [],
         }
 
         swh_storage.content_add([content])
         sha1_gits.append(content.sha1_git)
 
         expected[content.sha1_git] = [
             {"sha1_git": content.sha1_git, "type": "content",}
         ]
 
         swh_storage.directory_add([directory])
         sha1_gits.append(directory.id)
         expected[directory.id] = [{"sha1_git": directory.id, "type": "directory",}]
 
         swh_storage.revision_add([revision])
         sha1_gits.append(revision.id)
         expected[revision.id] = [{"sha1_git": revision.id, "type": "revision",}]
 
         swh_storage.release_add([release])
         sha1_gits.append(release.id)
         expected[release.id] = [{"sha1_git": release.id, "type": "release",}]
 
         ret = swh_storage.object_find_by_sha1_git(sha1_gits)
 
         assert expected == ret
 
     def test_metadata_fetcher_add_get(self, swh_storage, sample_data):
         fetcher = sample_data.metadata_fetcher
         actual_fetcher = swh_storage.metadata_fetcher_get(fetcher.name, fetcher.version)
         assert actual_fetcher is None  # does not exist
 
         swh_storage.metadata_fetcher_add([fetcher])
 
         res = swh_storage.metadata_fetcher_get(fetcher.name, fetcher.version)
         assert res == fetcher
 
     def test_metadata_fetcher_add_zero(self, swh_storage, sample_data):
         fetcher = sample_data.metadata_fetcher
         actual_fetcher = swh_storage.metadata_fetcher_get(fetcher.name, fetcher.version)
         assert actual_fetcher is None  # does not exist
 
         swh_storage.metadata_fetcher_add([])
 
     def test_metadata_authority_add_get(self, swh_storage, sample_data):
         authority = sample_data.metadata_authority
 
         actual_authority = swh_storage.metadata_authority_get(
             authority.type, authority.url
         )
         assert actual_authority is None  # does not exist
 
         swh_storage.metadata_authority_add([authority])
 
         res = swh_storage.metadata_authority_get(authority.type, authority.url)
         assert res == authority
 
     def test_metadata_authority_add_zero(self, swh_storage, sample_data):
         authority = sample_data.metadata_authority
 
         actual_authority = swh_storage.metadata_authority_get(
             authority.type, authority.url
         )
         assert actual_authority is None  # does not exist
 
         swh_storage.metadata_authority_add([])
 
     def test_content_metadata_add(self, swh_storage, sample_data):
         content = sample_data.content
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         content_metadata = sample_data.content_metadata[:2]
 
         content_swhid = SWHID(
             object_type="content", object_id=hash_to_bytes(content.sha1_git)
         )
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
 
         swh_storage.raw_extrinsic_metadata_add(content_metadata)
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.CONTENT, content_swhid, authority
         )
         assert result["next_page_token"] is None
         assert list(sorted(result["results"], key=lambda x: x.discovery_date,)) == list(
             content_metadata
         )
 
     def test_content_metadata_add_duplicate(self, swh_storage, sample_data):
         """Duplicates should be silently updated."""
         content = sample_data.content
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         content_metadata, content_metadata2 = sample_data.content_metadata[:2]
         content_swhid = SWHID(
             object_type="content", object_id=hash_to_bytes(content.sha1_git)
         )
 
         new_content_metadata2 = attr.evolve(
             content_metadata2, format="new-format", metadata=b"new-metadata",
         )
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
 
         swh_storage.raw_extrinsic_metadata_add([content_metadata, content_metadata2])
         swh_storage.raw_extrinsic_metadata_add([new_content_metadata2])
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.CONTENT, content_swhid, authority
         )
         assert result["next_page_token"] is None
 
         expected_results1 = (content_metadata, new_content_metadata2)
         expected_results2 = (content_metadata, content_metadata2)
 
         assert tuple(sorted(result["results"], key=lambda x: x.discovery_date,)) in (
             expected_results1,  # cassandra
             expected_results2,  # postgresql
         )
 
     def test_content_metadata_get(self, swh_storage, sample_data):
         content, content2 = sample_data.contents[:2]
         fetcher, fetcher2 = sample_data.fetchers[:2]
         authority, authority2 = sample_data.authorities[:2]
         (
             content1_metadata1,
             content1_metadata2,
             content1_metadata3,
         ) = sample_data.content_metadata[:3]
 
         content1_swhid = SWHID(object_type="content", object_id=content.sha1_git)
         content2_swhid = SWHID(object_type="content", object_id=content2.sha1_git)
         content2_metadata = attr.evolve(content1_metadata2, id=content2_swhid)
 
         swh_storage.metadata_authority_add([authority, authority2])
         swh_storage.metadata_fetcher_add([fetcher, fetcher2])
 
         swh_storage.raw_extrinsic_metadata_add(
             [
                 content1_metadata1,
                 content1_metadata2,
                 content1_metadata3,
                 content2_metadata,
             ]
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.CONTENT, content1_swhid, authority
         )
         assert result["next_page_token"] is None
         assert [content1_metadata1, content1_metadata2] == list(
             sorted(result["results"], key=lambda x: x.discovery_date,)
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.CONTENT, content1_swhid, authority2
         )
         assert result["next_page_token"] is None
         assert [content1_metadata3] == list(
             sorted(result["results"], key=lambda x: x.discovery_date,)
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.CONTENT, content2_swhid, authority
         )
         assert result["next_page_token"] is None
         assert [content2_metadata] == list(result["results"],)
 
     def test_content_metadata_get_after(self, swh_storage, sample_data):
         content = sample_data.content
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         content_metadata, content_metadata2 = sample_data.content_metadata[:2]
 
         content_swhid = SWHID(object_type="content", object_id=content.sha1_git)
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
 
         swh_storage.raw_extrinsic_metadata_add([content_metadata, content_metadata2])
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.CONTENT,
             content_swhid,
             authority,
             after=content_metadata.discovery_date - timedelta(seconds=1),
         )
         assert result["next_page_token"] is None
         assert [content_metadata, content_metadata2] == list(
             sorted(result["results"], key=lambda x: x.discovery_date,)
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.CONTENT,
             content_swhid,
             authority,
             after=content_metadata.discovery_date,
         )
         assert result["next_page_token"] is None
         assert result["results"] == [content_metadata2]
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.CONTENT,
             content_swhid,
             authority,
             after=content_metadata2.discovery_date,
         )
         assert result["next_page_token"] is None
         assert result["results"] == []
 
     def test_content_metadata_get_paginate(self, swh_storage, sample_data):
         content = sample_data.content
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         content_metadata, content_metadata2 = sample_data.content_metadata[:2]
 
         content_swhid = SWHID(object_type="content", object_id=content.sha1_git)
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
         swh_storage.raw_extrinsic_metadata_add([content_metadata, content_metadata2])
         swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.CONTENT, content_swhid, authority
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.CONTENT, content_swhid, authority, limit=1
         )
         assert result["next_page_token"] is not None
         assert result["results"] == [content_metadata]
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.CONTENT,
             content_swhid,
             authority,
             limit=1,
             page_token=result["next_page_token"],
         )
         assert result["next_page_token"] is None
         assert result["results"] == [content_metadata2]
 
     def test_content_metadata_get_paginate_same_date(self, swh_storage, sample_data):
         content = sample_data.content
         fetcher1, fetcher2 = sample_data.fetchers[:2]
         authority = sample_data.metadata_authority
         content_metadata, content_metadata2 = sample_data.content_metadata[:2]
 
         content_swhid = SWHID(object_type="content", object_id=content.sha1_git)
 
         swh_storage.metadata_fetcher_add([fetcher1, fetcher2])
         swh_storage.metadata_authority_add([authority])
 
         new_content_metadata2 = attr.evolve(
             content_metadata2,
             discovery_date=content_metadata2.discovery_date,
             fetcher=attr.evolve(fetcher2, metadata=None),
         )
 
         swh_storage.raw_extrinsic_metadata_add(
             [content_metadata, new_content_metadata2]
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.CONTENT, content_swhid, authority, limit=1
         )
         assert result["next_page_token"] is not None
         assert result["results"] == [content_metadata]
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.CONTENT,
             content_swhid,
             authority,
             limit=1,
             page_token=result["next_page_token"],
         )
         assert result["next_page_token"] is None
         assert result["results"] == [new_content_metadata2]
 
     def test_content_metadata_get__invalid_id(self, swh_storage, sample_data):
         origin = sample_data.origin
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         content_metadata, content_metadata2 = sample_data.content_metadata[:2]
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
         swh_storage.raw_extrinsic_metadata_add([content_metadata, content_metadata2])
 
         with pytest.raises(StorageArgumentException, match="SWHID"):
             swh_storage.raw_extrinsic_metadata_get(
                 MetadataTargetType.CONTENT, origin.url, authority
             )
 
     def test_origin_metadata_add(self, swh_storage, sample_data):
         origin = sample_data.origin
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
 
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
 
         swh_storage.raw_extrinsic_metadata_add([origin_metadata, origin_metadata2])
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.ORIGIN, origin.url, authority
         )
         assert result["next_page_token"] is None
         assert list(sorted(result["results"], key=lambda x: x.discovery_date)) == [
             origin_metadata,
             origin_metadata2,
         ]
 
     def test_origin_metadata_add_duplicate(self, swh_storage, sample_data):
         """Duplicates should be silently updated."""
         origin = sample_data.origin
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         new_origin_metadata2 = attr.evolve(
             origin_metadata2, format="new-format", metadata=b"new-metadata",
         )
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
 
         swh_storage.raw_extrinsic_metadata_add([origin_metadata, origin_metadata2])
         swh_storage.raw_extrinsic_metadata_add([new_origin_metadata2])
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.ORIGIN, origin.url, authority
         )
         assert result["next_page_token"] is None
 
         # which of the two behavior happens is backend-specific.
         expected_results1 = (origin_metadata, new_origin_metadata2)
         expected_results2 = (origin_metadata, origin_metadata2)
 
         assert tuple(sorted(result["results"], key=lambda x: x.discovery_date,)) in (
             expected_results1,  # cassandra
             expected_results2,  # postgresql
         )
 
     def test_origin_metadata_get(self, swh_storage, sample_data):
         origin, origin2 = sample_data.origins[:2]
         fetcher, fetcher2 = sample_data.fetchers[:2]
         authority, authority2 = sample_data.authorities[:2]
         (
             origin1_metadata1,
             origin1_metadata2,
             origin1_metadata3,
         ) = sample_data.origin_metadata[:3]
 
         assert swh_storage.origin_add([origin, origin2]) == {"origin:add": 2}
 
         origin2_metadata = attr.evolve(origin1_metadata2, id=origin2.url)
 
         swh_storage.metadata_authority_add([authority, authority2])
         swh_storage.metadata_fetcher_add([fetcher, fetcher2])
 
         swh_storage.raw_extrinsic_metadata_add(
             [origin1_metadata1, origin1_metadata2, origin1_metadata3, origin2_metadata]
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.ORIGIN, origin.url, authority
         )
         assert result["next_page_token"] is None
         assert [origin1_metadata1, origin1_metadata2] == list(
             sorted(result["results"], key=lambda x: x.discovery_date,)
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.ORIGIN, origin.url, authority2
         )
         assert result["next_page_token"] is None
         assert [origin1_metadata3] == list(
             sorted(result["results"], key=lambda x: x.discovery_date,)
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.ORIGIN, origin2.url, authority
         )
         assert result["next_page_token"] is None
         assert [origin2_metadata] == list(result["results"],)
 
     def test_origin_metadata_get_after(self, swh_storage, sample_data):
         origin = sample_data.origin
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
 
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
         swh_storage.raw_extrinsic_metadata_add([origin_metadata, origin_metadata2])
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.ORIGIN,
             origin.url,
             authority,
             after=origin_metadata.discovery_date - timedelta(seconds=1),
         )
         assert result["next_page_token"] is None
         assert list(sorted(result["results"], key=lambda x: x.discovery_date,)) == [
             origin_metadata,
             origin_metadata2,
         ]
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.ORIGIN,
             origin.url,
             authority,
             after=origin_metadata.discovery_date,
         )
         assert result["next_page_token"] is None
         assert result["results"] == [origin_metadata2]
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.ORIGIN,
             origin.url,
             authority,
             after=origin_metadata2.discovery_date,
         )
         assert result["next_page_token"] is None
         assert result["results"] == []
 
     def test_origin_metadata_get_paginate(self, swh_storage, sample_data):
         origin = sample_data.origin
         fetcher = sample_data.metadata_fetcher
         authority = sample_data.metadata_authority
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
 
         swh_storage.raw_extrinsic_metadata_add([origin_metadata, origin_metadata2])
 
         swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.ORIGIN, origin.url, authority
         )
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.ORIGIN, origin.url, authority, limit=1
         )
         assert result["next_page_token"] is not None
         assert result["results"] == [origin_metadata]
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.ORIGIN,
             origin.url,
             authority,
             limit=1,
             page_token=result["next_page_token"],
         )
         assert result["next_page_token"] is None
         assert result["results"] == [origin_metadata2]
 
     def test_origin_metadata_get_paginate_same_date(self, swh_storage, sample_data):
         origin = sample_data.origin
         fetcher1, fetcher2 = sample_data.fetchers[:2]
         authority = sample_data.metadata_authority
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         swh_storage.metadata_fetcher_add([fetcher1, fetcher2])
         swh_storage.metadata_authority_add([authority])
 
         new_origin_metadata2 = attr.evolve(
             origin_metadata2,
             discovery_date=origin_metadata2.discovery_date,
             fetcher=attr.evolve(fetcher2, metadata=None),
         )
 
         swh_storage.raw_extrinsic_metadata_add([origin_metadata, new_origin_metadata2])
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.ORIGIN, origin.url, authority, limit=1
         )
         assert result["next_page_token"] is not None
         assert result["results"] == [origin_metadata]
 
         result = swh_storage.raw_extrinsic_metadata_get(
             MetadataTargetType.ORIGIN,
             origin.url,
             authority,
             limit=1,
             page_token=result["next_page_token"],
         )
         assert result["next_page_token"] is None
         assert result["results"] == [new_origin_metadata2]
 
     def test_origin_metadata_add_missing_authority(self, swh_storage, sample_data):
         origin = sample_data.origin
         fetcher = sample_data.metadata_fetcher
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         swh_storage.metadata_fetcher_add([fetcher])
 
         with pytest.raises(StorageArgumentException, match="authority"):
             swh_storage.raw_extrinsic_metadata_add([origin_metadata, origin_metadata2])
 
     def test_origin_metadata_add_missing_fetcher(self, swh_storage, sample_data):
         origin = sample_data.origin
         authority = sample_data.metadata_authority
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         swh_storage.metadata_authority_add([authority])
 
         with pytest.raises(StorageArgumentException, match="fetcher"):
             swh_storage.raw_extrinsic_metadata_add([origin_metadata, origin_metadata2])
 
     def test_origin_metadata_get__invalid_id_type(self, swh_storage, sample_data):
         origin = sample_data.origin
         authority = sample_data.metadata_authority
         fetcher = sample_data.metadata_fetcher
         origin_metadata, origin_metadata2 = sample_data.origin_metadata[:2]
         content_metadata = sample_data.content_metadata[0]
         assert swh_storage.origin_add([origin]) == {"origin:add": 1}
 
         swh_storage.metadata_fetcher_add([fetcher])
         swh_storage.metadata_authority_add([authority])
 
         swh_storage.raw_extrinsic_metadata_add([origin_metadata, origin_metadata2])
 
         with pytest.raises(StorageArgumentException, match="SWHID"):
             swh_storage.raw_extrinsic_metadata_get(
                 MetadataTargetType.ORIGIN, content_metadata.id, authority,
             )
 
 
 class TestStorageGeneratedData:
     def test_generate_content_get(self, swh_storage, swh_contents):
         contents_with_data = [c.to_dict() for c in swh_contents if c.status != "absent"]
         # input the list of sha1s we want from storage
         get_sha1s = [c["sha1"] for c in contents_with_data]
 
         # retrieve contents
         actual_contents = list(swh_storage.content_get(get_sha1s))
         assert None not in actual_contents
         assert_contents_ok(contents_with_data, actual_contents)
 
     def test_generate_content_get_metadata(self, swh_storage, swh_contents):
         # input the list of sha1s we want from storage
         expected_contents = [c.to_dict() for c in swh_contents if c.status != "absent"]
         get_sha1s = [c["sha1"] for c in expected_contents]
 
         # retrieve contents
         meta_contents = swh_storage.content_get_metadata(get_sha1s)
 
         assert len(list(meta_contents)) == len(get_sha1s)
 
         actual_contents = []
         for contents in meta_contents.values():
             actual_contents.extend(contents)
 
         keys_to_check = {"length", "status", "sha1", "sha1_git", "sha256", "blake2s256"}
 
         assert_contents_ok(
             expected_contents, actual_contents, keys_to_check=keys_to_check
         )
 
     def test_generate_content_get_range(self, swh_storage, swh_contents):
         """content_get_range returns complete range"""
         present_contents = [c.to_dict() for c in swh_contents if c.status != "absent"]
 
         get_sha1s = sorted([c.sha1 for c in swh_contents if c.status != "absent"])
         start = get_sha1s[2]
         end = get_sha1s[-2]
         actual_result = swh_storage.content_get_range(start, end)
 
         assert actual_result["next"] is None
 
         actual_contents = actual_result["contents"]
         expected_contents = [c for c in present_contents if start <= c["sha1"] <= end]
         if expected_contents:
             assert_contents_ok(expected_contents, actual_contents, ["sha1"])
         else:
             assert actual_contents == []
 
     def test_generate_content_get_range_full(self, swh_storage, swh_contents):
         """content_get_range for a full range returns all available contents"""
         present_contents = [c.to_dict() for c in swh_contents if c.status != "absent"]
 
         start = b"0" * 40
         end = b"f" * 40
         actual_result = swh_storage.content_get_range(start, end)
         assert actual_result["next"] is None
 
         actual_contents = actual_result["contents"]
         expected_contents = [c for c in present_contents if start <= c["sha1"] <= end]
         if expected_contents:
             assert_contents_ok(expected_contents, actual_contents, ["sha1"])
         else:
             assert actual_contents == []
 
     def test_generate_content_get_range_empty(self, swh_storage, swh_contents):
         """content_get_range for an empty range returns nothing"""
         start = b"0" * 40
         end = b"f" * 40
         actual_result = swh_storage.content_get_range(end, start)
         assert actual_result["next"] is None
         assert len(actual_result["contents"]) == 0
 
     def test_generate_content_get_range_limit_none(self, swh_storage):
         """content_get_range call with wrong limit input should fail"""
         with pytest.raises(StorageArgumentException) as e:
             swh_storage.content_get_range(start=None, end=None, limit=None)
 
         assert e.value.args == ("limit should not be None",)
 
     def test_generate_content_get_range_no_limit(self, swh_storage, swh_contents):
         """content_get_range returns contents within range provided"""
         # input the list of sha1s we want from storage
         get_sha1s = sorted([c.sha1 for c in swh_contents if c.status != "absent"])
         start = get_sha1s[0]
         end = get_sha1s[-1]
 
         # retrieve contents
         actual_result = swh_storage.content_get_range(start, end)
 
         actual_contents = actual_result["contents"]
         assert actual_result["next"] is None
         assert len(actual_contents) == len(get_sha1s)
 
         expected_contents = [c.to_dict() for c in swh_contents if c.status != "absent"]
         assert_contents_ok(expected_contents, actual_contents, ["sha1"])
 
     def test_generate_content_get_range_limit(self, swh_storage, swh_contents):
         """content_get_range paginates results if limit exceeded"""
         contents_map = {c.sha1: c.to_dict() for c in swh_contents}
 
         # input the list of sha1s we want from storage
         get_sha1s = sorted([c.sha1 for c in swh_contents if c.status != "absent"])
         start = get_sha1s[0]
         end = get_sha1s[-1]
 
         # retrieve contents limited to n-1 results
         limited_results = len(get_sha1s) - 1
         actual_result = swh_storage.content_get_range(start, end, limit=limited_results)
 
         actual_contents = actual_result["contents"]
         assert actual_result["next"] == get_sha1s[-1]
         assert len(actual_contents) == limited_results
 
         expected_contents = [contents_map[sha1] for sha1 in get_sha1s[:-1]]
         assert_contents_ok(expected_contents, actual_contents, ["sha1"])
 
         # retrieve next part
         actual_results2 = swh_storage.content_get_range(start=end, end=end)
         assert actual_results2["next"] is None
         actual_contents2 = actual_results2["contents"]
         assert len(actual_contents2) == 1
 
         assert_contents_ok([contents_map[get_sha1s[-1]]], actual_contents2, ["sha1"])
 
     def test_origin_get_range_from_zero(self, swh_storage, swh_origins):
         actual_origins = list(
             swh_storage.origin_get_range(origin_from=0, origin_count=0)
         )
         assert len(actual_origins) == 0
 
         actual_origins = list(
             swh_storage.origin_get_range(origin_from=0, origin_count=1)
         )
         assert len(actual_origins) == 1
         assert actual_origins[0]["id"] == 1
         assert actual_origins[0]["url"] == swh_origins[0].url
 
     @pytest.mark.parametrize(
         "origin_from,origin_count",
         [(1, 1), (1, 10), (1, 20), (1, 101), (11, 0), (11, 10), (91, 11)],
     )
     def test_origin_get_range(
         self, swh_storage, swh_origins, origin_from, origin_count
     ):
         actual_origins = list(
             swh_storage.origin_get_range(
                 origin_from=origin_from, origin_count=origin_count
             )
         )
 
         origins_with_id = list(enumerate(swh_origins, start=1))
         expected_origins = [
             {"url": origin.url, "id": origin_id,}
             for (origin_id, origin) in origins_with_id[
                 origin_from - 1 : origin_from + origin_count - 1
             ]
         ]
 
         assert actual_origins == expected_origins
 
     @pytest.mark.parametrize("limit", [1, 7, 10, 100, 1000])
     def test_origin_list(self, swh_storage, swh_origins, limit):
         returned_origins = []
 
         page_token = None
         i = 0
         while True:
             result = swh_storage.origin_list(page_token=page_token, limit=limit)
             assert len(result["origins"]) <= limit
 
             returned_origins.extend(origin["url"] for origin in result["origins"])
 
             i += 1
             page_token = result.get("next_page_token")
 
             if page_token is None:
                 assert i * limit >= len(swh_origins)
                 break
             else:
                 assert len(result["origins"]) == limit
 
         expected_origins = [origin.url for origin in swh_origins]
         assert sorted(returned_origins) == sorted(expected_origins)
 
     def test_origin_count(self, swh_storage, sample_data):
         swh_storage.origin_add(sample_data.origins)
 
         assert swh_storage.origin_count("github") == 3
         assert swh_storage.origin_count("gitlab") == 2
         assert swh_storage.origin_count(".*user.*", regexp=True) == 5
         assert swh_storage.origin_count(".*user.*", regexp=False) == 0
         assert swh_storage.origin_count(".*user1.*", regexp=True) == 2
         assert swh_storage.origin_count(".*user1.*", regexp=False) == 0
 
     def test_origin_count_with_visit_no_visits(self, swh_storage, sample_data):
         swh_storage.origin_add(sample_data.origins)
 
         # none of them have visits, so with_visit=True => 0
         assert swh_storage.origin_count("github", with_visit=True) == 0
         assert swh_storage.origin_count("gitlab", with_visit=True) == 0
         assert swh_storage.origin_count(".*user.*", regexp=True, with_visit=True) == 0
         assert swh_storage.origin_count(".*user.*", regexp=False, with_visit=True) == 0
         assert swh_storage.origin_count(".*user1.*", regexp=True, with_visit=True) == 0
         assert swh_storage.origin_count(".*user1.*", regexp=False, with_visit=True) == 0
 
     def test_origin_count_with_visit_with_visits_no_snapshot(
         self, swh_storage, sample_data
     ):
         swh_storage.origin_add(sample_data.origins)
 
         origin_url = "https://github.com/user1/repo1"
         visit = OriginVisit(origin=origin_url, date=now(), type="git",)
         swh_storage.origin_visit_add([visit])
 
         assert swh_storage.origin_count("github", with_visit=False) == 3
         # it has a visit, but no snapshot, so with_visit=True => 0
         assert swh_storage.origin_count("github", with_visit=True) == 0
 
         assert swh_storage.origin_count("gitlab", with_visit=False) == 2
         # these gitlab origins have no visit
         assert swh_storage.origin_count("gitlab", with_visit=True) == 0
 
         assert (
             swh_storage.origin_count("github.*user1", regexp=True, with_visit=False)
             == 1
         )
         assert (
             swh_storage.origin_count("github.*user1", regexp=True, with_visit=True) == 0
         )
         assert swh_storage.origin_count("github", regexp=True, with_visit=True) == 0
 
     def test_origin_count_with_visit_with_visits_and_snapshot(
         self, swh_storage, sample_data
     ):
         snapshot = sample_data.snapshot
         swh_storage.origin_add(sample_data.origins)
 
         swh_storage.snapshot_add([snapshot])
         origin_url = "https://github.com/user1/repo1"
         visit = OriginVisit(origin=origin_url, date=now(), type="git",)
         visit = swh_storage.origin_visit_add([visit])[0]
         swh_storage.origin_visit_status_add(
             [
                 OriginVisitStatus(
                     origin=origin_url,
                     visit=visit.visit,
                     date=now(),
                     status="ongoing",
                     snapshot=snapshot.id,
                 )
             ]
         )
 
         assert swh_storage.origin_count("github", with_visit=False) == 3
         # github/user1 has a visit and a snapshot, so with_visit=True => 1
         assert swh_storage.origin_count("github", with_visit=True) == 1
 
         assert (
             swh_storage.origin_count("github.*user1", regexp=True, with_visit=False)
             == 1
         )
         assert (
             swh_storage.origin_count("github.*user1", regexp=True, with_visit=True) == 1
         )
         assert swh_storage.origin_count("github", regexp=True, with_visit=True) == 1
 
     @settings(suppress_health_check=[HealthCheck.too_slow])
     @given(strategies.lists(objects(split_content=True), max_size=2))
     def test_add_arbitrary(self, swh_storage, objects):
         for (obj_type, obj) in objects:
             if obj.object_type == "origin_visit":
                 swh_storage.origin_add([Origin(url=obj.origin)])
                 visit = OriginVisit(origin=obj.origin, date=obj.date, type=obj.type,)
                 swh_storage.origin_visit_add([visit])
             else:
                 method = getattr(swh_storage, obj_type + "_add")
                 try:
                     method([obj])
                 except HashCollision:
                     pass
 
 
 @pytest.mark.db
 class TestLocalStorage:
     """Test the local storage"""
 
     # This test is only relevant on the local storage, with an actual
     # objstorage raising an exception
     def test_content_add_objstorage_exception(self, swh_storage, sample_data):
         content = sample_data.content
 
         swh_storage.objstorage.content_add = Mock(
             side_effect=Exception("mocked broken objstorage")
         )
 
         with pytest.raises(Exception, match="mocked broken"):
             swh_storage.content_add([content])
 
         missing = list(swh_storage.content_missing([content.hashes()]))
         assert missing == [content.sha1]
 
 
 @pytest.mark.db
 class TestStorageRaceConditions:
     @pytest.mark.xfail
     def test_content_add_race(self, swh_storage, sample_data):
         content = sample_data.content
 
         results = queue.Queue()
 
         def thread():
             try:
                 with db_transaction(swh_storage) as (db, cur):
                     ret = swh_storage.content_add([content], db=db, cur=cur)
                 results.put((threading.get_ident(), "data", ret))
             except Exception as e:
                 results.put((threading.get_ident(), "exc", e))
 
         t1 = threading.Thread(target=thread)
         t2 = threading.Thread(target=thread)
         t1.start()
         # this avoids the race condition
         # import time
         # time.sleep(1)
         t2.start()
         t1.join()
         t2.join()
 
         r1 = results.get(block=False)
         r2 = results.get(block=False)
 
         with pytest.raises(queue.Empty):
             results.get(block=False)
         assert r1[0] != r2[0]
         assert r1[1] == "data", "Got exception %r in Thread%s" % (r1[2], r1[0])
         assert r2[1] == "data", "Got exception %r in Thread%s" % (r2[2], r2[0])
 
 
 @pytest.mark.db
 class TestPgStorage:
     """This class is dedicated for the rare case where the schema needs to
        be altered dynamically.
 
        Otherwise, the tests could be blocking when ran altogether.
 
     """
 
     def test_content_update_with_new_cols(self, swh_storage, sample_data):
         content, content2 = sample_data.contents[:2]
 
         swh_storage.journal_writer.journal = None  # TODO, not supported
 
         with db_transaction(swh_storage) as (_, cur):
             cur.execute(
                 """alter table content
                            add column test text default null,
                            add column test2 text default null"""
             )
 
         swh_storage.content_add([content])
 
         cont = content.to_dict()
         cont["test"] = "value-1"
         cont["test2"] = "value-2"
 
         swh_storage.content_update([cont], keys=["test", "test2"])
         with db_transaction(swh_storage) as (_, cur):
             cur.execute(
                 """SELECT sha1, sha1_git, sha256, length, status,
                    test, test2
                    FROM content WHERE sha1 = %s""",
                 (cont["sha1"],),
             )
 
             datum = cur.fetchone()
 
         assert datum == (
             cont["sha1"],
             cont["sha1_git"],
             cont["sha256"],
             cont["length"],
             "visible",
             cont["test"],
             cont["test2"],
         )
 
         with db_transaction(swh_storage) as (_, cur):
             cur.execute(
                 """alter table content drop column test,
                                                drop column test2"""
             )
 
     def test_content_add_db(self, swh_storage, sample_data):
         content = sample_data.content
 
         actual_result = swh_storage.content_add([content])
 
         assert actual_result == {
             "content:add": 1,
             "content:add:bytes": content.length,
         }
 
         if hasattr(swh_storage, "objstorage"):
             assert content.sha1 in swh_storage.objstorage.objstorage
 
         with db_transaction(swh_storage) as (_, cur):
             cur.execute(
                 "SELECT sha1, sha1_git, sha256, length, status"
                 " FROM content WHERE sha1 = %s",
                 (content.sha1,),
             )
             datum = cur.fetchone()
 
         assert datum == (
             content.sha1,
             content.sha1_git,
             content.sha256,
             content.length,
             "visible",
         )
 
         contents = [
             obj
             for (obj_type, obj) in swh_storage.journal_writer.journal.objects
             if obj_type == "content"
         ]
         assert len(contents) == 1
         assert contents[0] == attr.evolve(content, data=None)
 
     def test_content_add_metadata_db(self, swh_storage, sample_data):
         content = attr.evolve(sample_data.content, data=None, ctime=now())
 
         actual_result = swh_storage.content_add_metadata([content])
 
         assert actual_result == {
             "content:add": 1,
         }
 
         if hasattr(swh_storage, "objstorage"):
             assert content.sha1 not in swh_storage.objstorage.objstorage
         with db_transaction(swh_storage) as (_, cur):
             cur.execute(
                 "SELECT sha1, sha1_git, sha256, length, status"
                 " FROM content WHERE sha1 = %s",
                 (content.sha1,),
             )
             datum = cur.fetchone()
         assert datum == (
             content.sha1,
             content.sha1_git,
             content.sha256,
             content.length,
             "visible",
         )
 
         contents = [
             obj
             for (obj_type, obj) in swh_storage.journal_writer.journal.objects
             if obj_type == "content"
         ]
         assert len(contents) == 1
         assert contents[0] == content
 
     def test_skipped_content_add_db(self, swh_storage, sample_data):
         content, cont2 = sample_data.skipped_contents[:2]
         content2 = attr.evolve(cont2, blake2s256=None)
 
         actual_result = swh_storage.skipped_content_add([content, content, content2])
 
         assert 2 <= actual_result.pop("skipped_content:add") <= 3
         assert actual_result == {}
 
         with db_transaction(swh_storage) as (_, cur):
             cur.execute(
                 "SELECT sha1, sha1_git, sha256, blake2s256, "
                 "length, status, reason "
                 "FROM skipped_content ORDER BY sha1_git"
             )
 
             dbdata = cur.fetchall()
 
         assert len(dbdata) == 2
         assert dbdata[0] == (
             content.sha1,
             content.sha1_git,
             content.sha256,
             content.blake2s256,
             content.length,
             "absent",
             "Content too long",
         )
 
         assert dbdata[1] == (
             content2.sha1,
             content2.sha1_git,
             content2.sha256,
             content2.blake2s256,
             content2.length,
             "absent",
             "Content too long",
         )
 
     def test_clear_buffers(self, swh_storage):
         """Calling clear buffers on real storage does nothing
 
         """
         assert swh_storage.clear_buffers() is None
 
     def test_flush(self, swh_storage):
         """Calling clear buffers on real storage does nothing
 
         """
         assert swh_storage.flush() == {}