fix: conditional promote + real SessionDB tests

Address review feedback on #63068:

1. Replace unconditional reopen_session() + end_session() with a
   conditional promote_to_session_reset() method in SessionDB.
   The new method only promotes live rows or rows ended with
   agent_close — explicit boundaries (compression, session_reset,
   new_command) are preserved via first-writer-wins semantics.

2. Rewrite tests to use real SessionDB instead of MagicMock:
   - 7 unit tests for promote_to_session_reset edge cases
   - 3 integration tests verifying the actual recovery contract
     in find_latest_gateway_session_for_peer after promotion
This commit is contained in:
saitsuki 2026-07-16 07:53:59 +08:00 committed by Teknium
parent e701cdc86e
commit 3305dcedbb
3 changed files with 199 additions and 49 deletions

View file

@ -1587,12 +1587,15 @@ class SessionStore:
# a durable ``session_reset`` end_reason, later agent cleanup can
# close the row as ``agent_close``; stale-route recovery treats
# that as resumable and resurrects the expired full history.
# Reopen first because SessionDB.end_session is first-writer-wins.
self._db.reopen_session(entry.session_id)
self._db.end_session(entry.session_id, "session_reset")
#
# promote_to_session_reset is conditional: it only promotes
# live rows or rows ended with ``agent_close``. Explicit
# boundaries (compression, session_reset, new_command, etc.)
# are preserved — the first writer wins.
self._db.promote_to_session_reset(entry.session_id)
except Exception as exc:
logger.debug(
"Session DB end_session(session_reset) failed for %s: %s",
"Session DB promote_to_session_reset failed for %s: %s",
entry.session_id, exc,
)

View file

@ -2286,6 +2286,36 @@ class SessionDB:
)
self._execute_write(_do)
def promote_to_session_reset(self, session_id: str) -> bool:
"""Mark a session as ended by session_reset — but only when safe.
Promotes *only* live rows (``ended_at IS NULL``) or rows ended with
``agent_close``. Explicit conversation boundaries such as
``compression``, ``session_reset``, ``new_command``, etc. are
preserved the first writer wins for those, and a later expiry
finalization must not silently overwrite them.
Returns ``True`` when the row was promoted, ``False`` when skipped
(already has a different explicit end_reason, or row not found).
"""
if not session_id:
return False
now = time.time()
def _do(conn):
cursor = conn.execute(
"UPDATE sessions SET ended_at = ?, end_reason = 'session_reset' "
"WHERE id = ? AND (ended_at IS NULL OR end_reason = 'agent_close')",
(now, session_id),
)
return cursor.rowcount
try:
rows = self._execute_write(_do)
return bool(rows)
except Exception:
return False
def update_session_cwd(
self, session_id: str, cwd: str, git_branch: str = None, git_repo_root: str = None
) -> None:

View file

@ -5,63 +5,180 @@ then agent cleanup can close it as ``agent_close``. Stale routing recovery treat
``agent_close`` as recoverable, so expired sessions were reopened with full
history unless expiry finalization also persisted the real conversation boundary
as ``end_reason='session_reset'``.
These tests use a real ``SessionDB`` (in-memory) to verify the actual recovery
contract in ``find_latest_gateway_session_for_peer`` not just call counts on
a MagicMock.
"""
from datetime import datetime, timedelta
from unittest.mock import MagicMock, patch
from __future__ import annotations
from gateway.config import GatewayConfig, Platform, SessionResetPolicy
from gateway.session import SessionEntry, SessionStore
import time
from pathlib import Path
import pytest
from hermes_state import SessionDB
def _make_store_with_db(tmp_path, db_mock) -> SessionStore:
config = GatewayConfig(default_reset_policy=SessionResetPolicy(mode="daily"))
with patch("gateway.session.SessionStore._ensure_loaded"):
store = SessionStore(sessions_dir=tmp_path, config=config)
store._db = db_mock
store._loaded = True
return store
@pytest.fixture
def db(tmp_path: Path) -> SessionDB:
return SessionDB(tmp_path / "state.db")
def _entry(session_id: str = "sid-expired") -> SessionEntry:
now = datetime.now()
return SessionEntry(
session_key="agent:main:telegram:dm:8494508720",
session_id=session_id,
created_at=now - timedelta(days=1),
updated_at=now - timedelta(days=1),
platform=Platform.TELEGRAM,
chat_type="dm",
model_override={"provider": "openrouter", "model": "test/model"},
)
_SESSION_KEY = "agent:main:telegram:dm:8494508720"
_SOURCE = "telegram"
_USER_ID = "8494508720"
def test_set_expiry_finalized_persists_session_reset_boundary(tmp_path):
db = MagicMock()
db.set_expiry_finalized.return_value = None
db.reopen_session.return_value = None
db.end_session.return_value = None
store = _make_store_with_db(tmp_path, db)
entry = _entry()
# ------------------------------------------------------------------
# promote_to_session_reset — unit tests on real DB
# ------------------------------------------------------------------
store.set_expiry_finalized(entry)
class TestPromoteToSessionReset:
"""promote_to_session_reset promotes only safe rows."""
assert entry.expiry_finalized is True
assert entry.model_override is None
db.set_expiry_finalized.assert_called_once_with("sid-expired", True)
db.reopen_session.assert_called_once_with("sid-expired")
db.end_session.assert_called_once_with("sid-expired", "session_reset")
def test_promotes_live_row(self, db: SessionDB) -> None:
"""A live row (ended_at IS NULL) is promoted to session_reset."""
db.create_session(
"sid-live", _SOURCE,
user_id=_USER_ID, session_key=_SESSION_KEY,
chat_id="8494508720", chat_type="dm",
)
# Seed a message so recovery would match
db.append_message("sid-live", "user", "hello")
assert db.promote_to_session_reset("sid-live") is True
row = db.get_session("sid-live")
assert row["end_reason"] == "session_reset"
assert row["ended_at"] is not None
def test_promotes_agent_close_row(self, db: SessionDB) -> None:
"""A row ended with agent_close is promoted to session_reset."""
db.create_session(
"sid-ac", _SOURCE,
user_id=_USER_ID, session_key=_SESSION_KEY,
chat_id="8494508720", chat_type="dm",
)
db.append_message("sid-ac", "user", "hello")
db.end_session("sid-ac", "agent_close")
assert db.promote_to_session_reset("sid-ac") is True
row = db.get_session("sid-ac")
assert row["end_reason"] == "session_reset"
def test_does_not_overwrite_compression(self, db: SessionDB) -> None:
"""An existing compression boundary must not be overwritten."""
db.create_session(
"sid-comp", _SOURCE,
user_id=_USER_ID, session_key=_SESSION_KEY,
chat_id="8494508720", chat_type="dm",
)
db.end_session("sid-comp", "compression")
assert db.promote_to_session_reset("sid-comp") is False
row = db.get_session("sid-comp")
assert row["end_reason"] == "compression"
def test_does_not_overwrite_existing_session_reset(self, db: SessionDB) -> None:
"""Already-promoted rows are idempotently skipped."""
db.create_session(
"sid-reset", _SOURCE,
user_id=_USER_ID, session_key=_SESSION_KEY,
chat_id="8494508720", chat_type="dm",
)
db.end_session("sid-reset", "session_reset")
# Should be a no-op (rowcount = 0)
assert db.promote_to_session_reset("sid-reset") is False
row = db.get_session("sid-reset")
assert row["end_reason"] == "session_reset"
def test_does_not_overwrite_new_command(self, db: SessionDB) -> None:
"""A /new-command boundary is preserved."""
db.create_session(
"sid-new", _SOURCE,
user_id=_USER_ID, session_key=_SESSION_KEY,
chat_id="8494508720", chat_type="dm",
)
db.end_session("sid-new", "new_command")
assert db.promote_to_session_reset("sid-new") is False
row = db.get_session("sid-new")
assert row["end_reason"] == "new_command"
def test_noop_on_missing_session(self, db: SessionDB) -> None:
"""Non-existent session_id returns False without error."""
assert db.promote_to_session_reset("nonexistent") is False
def test_noop_on_empty_session_id(self, db: SessionDB) -> None:
assert db.promote_to_session_reset("") is False
def test_set_expiry_finalized_still_sets_flag_if_end_session_fails(tmp_path):
db = MagicMock()
db.set_expiry_finalized.return_value = None
db.reopen_session.side_effect = RuntimeError("database locked")
store = _make_store_with_db(tmp_path, db)
entry = _entry()
# ------------------------------------------------------------------
# Integration: promotion blocks stale-route recovery
# ------------------------------------------------------------------
store.set_expiry_finalized(entry)
class TestPromotionBlocksRecovery:
"""After promotion, find_latest_gateway_session_for_peer must NOT
recover the session it is now ended with session_reset, which
the recovery query excludes (only live/agent_close are recoverable).
"""
assert entry.expiry_finalized is True
db.set_expiry_finalized.assert_called_once_with("sid-expired", True)
db.end_session.assert_not_called()
def test_live_session_recoverable_before_promotion(self, db: SessionDB) -> None:
db.create_session(
"sid-pre", _SOURCE,
user_id=_USER_ID, session_key=_SESSION_KEY,
chat_id="8494508720", chat_type="dm",
)
db.append_message("sid-pre", "user", "hello")
recovered = db.find_latest_gateway_session_for_peer(
source=_SOURCE, session_key=_SESSION_KEY,
user_id=_USER_ID, chat_id="8494508720", chat_type="dm",
)
assert recovered is not None
assert recovered["id"] == "sid-pre"
def test_promoted_session_not_recoverable(self, db: SessionDB) -> None:
db.create_session(
"sid-post", _SOURCE,
user_id=_USER_ID, session_key=_SESSION_KEY,
chat_id="8494508720", chat_type="dm",
)
db.append_message("sid-post", "user", "hello")
db.promote_to_session_reset("sid-post")
recovered = db.find_latest_gateway_session_for_peer(
source=_SOURCE, session_key=_SESSION_KEY,
user_id=_USER_ID, chat_id="8494508720", chat_type="dm",
)
# session_reset rows are not in the recovery set
assert recovered is None
def test_agent_close_session_recoverable_but_not_after_promotion(
self, db: SessionDB
) -> None:
db.create_session(
"sid-ac-rec", _SOURCE,
user_id=_USER_ID, session_key=_SESSION_KEY,
chat_id="8494508720", chat_type="dm",
)
db.append_message("sid-ac-rec", "user", "hello")
db.end_session("sid-ac-rec", "agent_close")
# agent_close is recoverable
recovered = db.find_latest_gateway_session_for_peer(
source=_SOURCE, session_key=_SESSION_KEY,
user_id=_USER_ID, chat_id="8494508720", chat_type="dm",
)
assert recovered is not None
assert recovered["id"] == "sid-ac-rec"
# After promotion, it is no longer recoverable
db.promote_to_session_reset("sid-ac-rec")
recovered2 = db.find_latest_gateway_session_for_peer(
source=_SOURCE, session_key=_SESSION_KEY,
user_id=_USER_ID, chat_id="8494508720", chat_type="dm",
)
assert recovered2 is None