fix(session): serialize canonical file access (#5383)

This commit is contained in:
chengyongru
2026-08-14 10:32:17 +08:00
committed by GitHub
parent e3d1819a2b
commit e226242dfc
10 changed files with 262 additions and 77 deletions
+24 -13
View File
@@ -29,8 +29,11 @@ class TestPruneDreamSessions:
import os
import time
sessions_dir = tmp_path / "sessions"
sessions_dir.mkdir()
manager = SessionManager(
tmp_path / "workspace",
sessions_root=tmp_path / "runtime",
)
sessions_dir = manager.sessions_dir
base_time = time.time() - 100
dream_paths = []
@@ -50,7 +53,7 @@ class TestPruneDreamSessions:
normal_path = sessions_dir / "telegram_123.jsonl"
normal_path.write_text('{"_type": "metadata"}\n', encoding="utf-8")
MemoryStore.prune_dream_sessions(sessions_dir, keep=10)
MemoryStore.prune_dream_sessions(manager, keep=10)
assert [path.exists() for path in dream_paths] == [False] * 5 + [True] * 10
assert normal_path.exists()
@@ -59,8 +62,11 @@ class TestPruneDreamSessions:
import os
import time
sessions_dir = tmp_path / "sessions"
sessions_dir.mkdir()
manager = SessionManager(
tmp_path / "workspace",
sessions_root=tmp_path / "runtime",
)
sessions_dir = manager.sessions_dir
base_time = time.time() - 100
current_paths = []
@@ -81,24 +87,29 @@ class TestPruneDreamSessions:
)
os.utime(legacy_path, (base_time - 1, base_time - 1))
MemoryStore.prune_dream_sessions(sessions_dir, keep=1)
MemoryStore.prune_dream_sessions(manager, keep=1)
assert [path.exists() for path in current_paths] == [False, True]
assert legacy_path.exists()
def test_noop_when_under_limit(self, tmp_path):
sessions_dir = tmp_path / "sessions"
sessions_dir.mkdir()
manager = SessionManager(
tmp_path / "workspace",
sessions_root=tmp_path / "runtime",
)
sessions_dir = manager.sessions_dir
for i in range(3):
key = f"dream:20260528-{100000 + i:06d}"
path = sessions_dir / f"{SessionManager._storage_key(key)}.jsonl"
path.write_text("{}", encoding="utf-8")
MemoryStore.prune_dream_sessions(sessions_dir, keep=10)
MemoryStore.prune_dream_sessions(manager, keep=10)
assert len(list(sessions_dir.glob("*.jsonl"))) == 3
def test_empty_dir_noop(self, tmp_path):
sessions_dir = tmp_path / "sessions"
sessions_dir.mkdir()
MemoryStore.prune_dream_sessions(sessions_dir, keep=10)
assert list(sessions_dir.iterdir()) == []
manager = SessionManager(
tmp_path / "workspace",
sessions_root=tmp_path / "runtime",
)
MemoryStore.prune_dream_sessions(manager, keep=10)
assert list(manager.sessions_dir.glob("*.jsonl")) == []
+40 -9
View File
@@ -4,6 +4,9 @@ import json
from datetime import datetime
from pathlib import Path
import pytest
from filelock import Timeout
from nanobot.providers.base import ProviderConversationState
from nanobot.session.manager import Session, SessionManager
@@ -37,14 +40,23 @@ class TestAtomicSave:
tmp_files = list(mgr.sessions_dir.glob("*.tmp"))
assert tmp_files == []
def test_tmp_file_cleaned_up_on_write_failure(self, tmp_path: Path):
def test_unique_tmp_file_cleaned_up_on_write_failure(
self,
tmp_path: Path,
monkeypatch: pytest.MonkeyPatch,
):
mgr = SessionManager(tmp_path)
session = Session(key="test:fail")
path = mgr._get_session_path("test:fail")
tmp_path_file = path.with_suffix(".jsonl.tmp")
stale_shared_tmp = path.with_suffix(".jsonl.tmp")
unique_tmp = path.with_name(f".{path.name}.save-failure.tmp")
path.parent.mkdir(parents=True, exist_ok=True)
tmp_path_file.write_text("stale")
stale_shared_tmp.write_text("stale", encoding="utf-8")
monkeypatch.setattr(
"nanobot.session.manager.secrets.token_hex",
lambda _length: "save-failure",
)
class BadMessage:
def __init__(self, data):
@@ -64,13 +76,17 @@ class TestAtomicSave:
]
import unittest.mock
with unittest.mock.patch("nanobot.session.manager.json.dumps", side_effect=failing_dumps):
try:
mgr.save(session)
except OSError:
pass
with (
unittest.mock.patch(
"nanobot.session.manager.json.dumps",
side_effect=failing_dumps,
),
pytest.raises(OSError, match="simulated disk full"),
):
mgr.save(session)
assert not tmp_path_file.exists()
assert not unique_tmp.exists()
assert stale_shared_tmp.read_text(encoding="utf-8") == "stale"
def test_overwrite_preserves_latest_data(self, tmp_path: Path):
mgr = SessionManager(tmp_path)
@@ -102,6 +118,21 @@ class TestAtomicSave:
for i in range(5):
assert loaded.messages[i]["content"] == f"msg{i}"
def test_managers_for_same_directory_coordinate_saves(self, tmp_path: Path):
workspace = tmp_path / "workspace"
sessions_root = tmp_path / "runtime"
owner = SessionManager(workspace, sessions_root=sessions_root)
peer = SessionManager(workspace, sessions_root=sessions_root)
assert owner.sessions_dir == peer.sessions_dir
session = Session(key="test:peer-manager")
peer._jsonl_store._session_files_lock.timeout = 0
with owner.locked_session_files(), pytest.raises(Timeout):
peer.save(session)
peer.save(session)
assert peer._get_session_path(session.key).is_file()
def test_provider_state_round_trips_in_private_record_only(self, tmp_path: Path):
mgr = SessionManager(tmp_path)
secret = "encrypted-reasoning-blob"
+12 -12
View File
@@ -16,6 +16,7 @@ from nanobot.command.builtin import (
cmd_dream_restore,
)
from nanobot.command.router import CommandContext
from nanobot.session.manager import SessionManager
from nanobot.utils.gitstore import CommitInfo
@@ -107,6 +108,13 @@ class _FakeBus:
self.outbound.append(message)
def _make_sessions(tmp_path) -> SessionManager:
return SessionManager(
tmp_path / "workspace",
sessions_root=tmp_path / "runtime",
)
def _make_ctx(raw: str, git: _FakeGit, *, args: str = "", last_dream_cursor: int = 1) -> CommandContext:
msg = InboundMessage(channel="cli", sender_id="u1", chat_id="direct", content=raw)
store = _FakeStore(git, last_dream_cursor=last_dream_cursor)
@@ -118,12 +126,10 @@ def _make_dream_ctx(tmp_path) -> tuple[CommandContext, _FakeBus]:
msg = InboundMessage(channel="cli", sender_id="u1", chat_id="direct", content="/dream")
store = _FakeStore(_FakeGit(initialized=False), dream_prompt_result=None)
bus = _FakeBus()
sessions_dir = tmp_path / "sessions"
sessions_dir.mkdir()
loop = SimpleNamespace(
bus=bus,
context=SimpleNamespace(memory=store, timezone="UTC"),
sessions=SimpleNamespace(sessions_dir=sessions_dir),
sessions=_make_sessions(tmp_path),
)
ctx = CommandContext(msg=msg, session=None, key=msg.session_key, raw="/dream", args="", loop=loop)
return ctx, bus
@@ -169,13 +175,11 @@ async def test_dream_internal_run_silences_progress(tmp_path) -> None:
metadata={"_stop_reason": "completed"},
)
sessions_dir = tmp_path / "sessions"
sessions_dir.mkdir()
dream_runtime = object()
loop = SimpleNamespace(
bus=bus,
context=SimpleNamespace(memory=store, timezone="UTC"),
sessions=SimpleNamespace(sessions_dir=sessions_dir),
sessions=_make_sessions(tmp_path),
process_direct=process_direct,
dream_runtime=lambda: dream_runtime,
)
@@ -224,12 +228,10 @@ def _build_runnable_dream(
)
bus = _FakeBus()
sessions_dir = tmp_path / "sessions"
sessions_dir.mkdir()
loop = SimpleNamespace(
bus=bus,
context=SimpleNamespace(memory=store, timezone="UTC"),
sessions=SimpleNamespace(sessions_dir=sessions_dir),
sessions=_make_sessions(tmp_path),
process_direct=process_direct,
dream_runtime=lambda: None,
)
@@ -311,12 +313,10 @@ async def test_dream_noop_batch_unlocks_following_history(tmp_path) -> None:
msg = InboundMessage(channel="cli", sender_id="u1", chat_id="direct", content="/dream")
bus = _FakeBus()
sessions_dir = tmp_path / "sessions"
sessions_dir.mkdir()
loop = SimpleNamespace(
bus=bus,
context=SimpleNamespace(memory=store, timezone="UTC"),
sessions=SimpleNamespace(sessions_dir=sessions_dir),
sessions=_make_sessions(tmp_path),
process_direct=process_direct,
dream_runtime=lambda: None,
)
+5 -3
View File
@@ -6,7 +6,7 @@ import errno
import os
import sys
from pathlib import Path
from unittest.mock import patch
from unittest.mock import call, patch
import pytest
@@ -65,6 +65,7 @@ class TestSaveFsync:
session.add_message("user", "hello")
directory_fd = 987654
with (
manager.locked_session_files(),
patch("nanobot.session.manager.os.open", return_value=directory_fd) as open_dir,
patch(
"nanobot.session.manager.os.fsync",
@@ -76,7 +77,7 @@ class TestSaveFsync:
assert manager._get_session_path(session.key).exists()
open_dir.assert_called_once_with(str(manager.sessions_dir), os.O_RDONLY)
close_dir.assert_called_once_with(directory_fd)
assert close_dir.call_args_list.count(call(directory_fd)) == 1
def test_save_propagates_other_directory_fsync_errors(
self, manager: SessionManager
@@ -85,6 +86,7 @@ class TestSaveFsync:
session = manager.get_or_create("test:directory-fsync-io-error")
directory_fd = 987654
with (
manager.locked_session_files(),
patch("nanobot.session.manager.os.open", return_value=directory_fd),
patch(
"nanobot.session.manager.os.fsync",
@@ -95,7 +97,7 @@ class TestSaveFsync:
):
manager.save(session, fsync=True)
close_dir.assert_called_once_with(directory_fd)
assert close_dir.call_args_list.count(call(directory_fd)) == 1
class TestFlushAll:
+98
View File
@@ -3,6 +3,8 @@ from __future__ import annotations
import io
import json
import os
import threading
from concurrent.futures import ThreadPoolExecutor
from datetime import datetime
from pathlib import Path
@@ -50,6 +52,20 @@ def test_webui_session_list_reuses_valid_index_without_scanning_files(
assert rows[0]["model_preset"] == "fast"
def test_webui_session_index_uses_unique_temp_file(tmp_path: Path) -> None:
manager = SessionManager(tmp_path)
session = manager.get_or_create("websocket:unique-index-temp")
session.add_message("user", "hello")
manager.save(session)
stale_shared_tmp = manager.sessions_dir / ".webui_session_index.json.tmp"
stale_shared_tmp.write_text("stale", encoding="utf-8")
assert list_webui_sessions(manager)[0]["preview"] == "hello"
assert stale_shared_tmp.read_text(encoding="utf-8") == "stale"
assert not list(manager.sessions_dir.glob(".webui_session_index.json.*.tmp"))
def test_webui_session_list_indexes_workspace_scope_and_preserves_null(
tmp_path: Path,
) -> None:
@@ -133,6 +149,88 @@ def test_webui_session_list_does_not_cache_old_snapshot_with_new_signature(
assert session_list_index.indexed_workspace_scope(second)[1]["access_mode"] == "restricted"
def test_webui_session_scan_does_not_overlap_session_save(
tmp_path: Path,
monkeypatch: pytest.MonkeyPatch,
) -> None:
manager = SessionManager(
tmp_path / "workspace",
sessions_root=tmp_path / "runtime",
)
session = manager.get_or_create("websocket:windows-reader")
session.add_message("user", "before")
manager.save(session)
session_path = manager._get_session_path(session.key)
session.messages[0]["content"] = "after"
reader_open = threading.Event()
release_reader = threading.Event()
save_started = threading.Event()
save_lock_attempted = threading.Event()
write_entered = threading.Event()
original_open = open
store = manager._jsonl_store
original_acquire = store._session_files_lock.acquire
original_save_unlocked = store._save_unlocked
class BlockingReader:
def __init__(self, file):
self.file = file
def __enter__(self):
entered = self.file.__enter__()
reader_open.set()
if not release_reader.wait(5):
raise AssertionError("timed out waiting to release the session reader")
return entered
def __exit__(self, *args):
try:
return self.file.__exit__(*args)
finally:
reader_open.clear()
def blocking_open(path, *args, **kwargs):
file = original_open(path, *args, **kwargs)
if Path(path) == session_path:
return BlockingReader(file)
return file
def observed_acquire(*args, **kwargs):
if save_started.is_set():
save_lock_attempted.set()
return original_acquire(*args, **kwargs)
def observed_save_unlocked(session, *, fsync=False):
write_entered.set()
assert not reader_open.is_set(), "save entered while the canonical file was open"
return original_save_unlocked(session, fsync=fsync)
monkeypatch.setattr(session_list_index, "open", blocking_open, raising=False)
monkeypatch.setattr(store._session_files_lock, "acquire", observed_acquire)
monkeypatch.setattr(store, "_save_unlocked", observed_save_unlocked)
def save_session() -> None:
save_started.set()
manager.save(session)
with ThreadPoolExecutor(max_workers=2) as executor:
list_future = executor.submit(list_webui_sessions, manager)
try:
assert reader_open.wait(5)
save_future = executor.submit(save_session)
assert save_lock_attempted.wait(5)
assert not write_entered.is_set()
finally:
release_reader.set()
assert list_future.result(timeout=5)[0]["preview"] == "before"
save_future.result(timeout=5)
assert write_entered.is_set()
assert list_webui_sessions(manager)[0]["preview"] == "after"
def test_webui_session_list_rejects_invalid_internal_model_preset_metadata(
tmp_path: Path,
) -> None: