ouyangbo/hermes-agent
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

fix(kanban): request_review shares complete_task's live-worker fence

Browse Source 
The PR docstring said complete_task applies "the same fence request_review
applies", but the two disagreed: complete_task keyed on a live worker
process while request_review still refused any running task with a
claim_lock, so a claim whose worker is gone (or a CLI/library claim that
never spawned one) could be completed but not sent to review without
force. Factor the liveness test into _claim_is_live and use it in both.

TTL expiry is deliberately not part of "live": reclaim_stale_tasks extends
(not reclaims) the claim of a live worker, so the process stays the
liveness authority.

Test: claim -> request_review without a worker PID now succeeds (red
before); a live worker's claim is still refused without expected_run_id.
This commit is contained in:
teknium1
2026-09-15 14:53:16 -07:00
committed by Teknium
parent 72916de360
commit 85d4415bed
3 changed files with 39 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files
hermes_cli/kanban_db.py
+22 -21
View File
@@ -2614,6 +2614,20 @@ class LiveClaimError(ValueError):
)
def _claim_is_live(trow) -> bool:
"""True when a ``running`` task's claim still protects a run: the worker process
it spawned exists (PID + start-time fingerprint). A claim whose worker is gone,
or a library/CLI claim that never spawned one, has no run to protect. TTL expiry
is deliberately not consulted: ``reclaim_stale_tasks`` extends, not reclaims, the
claim of a live worker, so the process is the liveness authority here too."""
return bool(
trow["status"] == "running"
and trow["claim_lock"] is not None
and trow["worker_pid"]
and _worker_alive(trow["worker_pid"], trow["worker_started_at"])
)
def complete_task(
conn: sqlite3.Connection, task_id: str, *, result: Optional[str] = None,
summary: Optional[str] = None, metadata: Optional[dict] = None,
@@ -2659,18 +2673,9 @@ def complete_task(
).fetchone()
prior_status = trow["status"] if trow else None
# Refuse to close a LIVE worker's run without proof of ownership
# (expected_run_id) or an explicit human override (force=True). "Live"
# means the spawned worker process still exists: a claim whose worker
# is gone (or a library claim that never spawned one) has no run to
# protect, so manual completion keeps working there.
if (
expected_run_id is None
and not force
and prior_status == "running"
and trow["claim_lock"] is not None
and trow["worker_pid"]
and _worker_alive(trow["worker_pid"], trow["worker_started_at"])
):
# (expected_run_id) or an explicit human override (force=True); see
# _claim_is_live for what "live" means.
if expected_run_id is None and not force and trow and _claim_is_live(trow):
raise LiveClaimError(task_id)
sql = """
UPDATE tasks
@@ -3164,19 +3169,15 @@ def request_review(
if not _parents_satisfied(conn, task_id):
return _ret(False, "parent dependencies are not satisfied")
trow = conn.execute(
"SELECT assignee, status, claim_lock, current_run_id "
"FROM tasks WHERE id = ?", (task_id,),
"SELECT assignee, status, claim_lock, current_run_id, worker_pid, "
"worker_started_at FROM tasks WHERE id = ?", (task_id,),
).fetchone()
if trow is None:
return _ret(False, "task not found")
# Refuse to clear a live worker's claim without proof of ownership
# (expected_run_id) or an explicit human override (force=True).
if (
expected_run_id is None
and not force
and trow["status"] == "running"
and trow["claim_lock"] is not None
):
# (expected_run_id) or an explicit human override (force=True);
# the same fence as complete_task (_claim_is_live).
if expected_run_id is None and not force and _claim_is_live(trow):
return _ret(
False, "task is running under a live claim; pass expected_run_id "
"(worker ownership) or force=True (explicit operator "