fix(run-hardening): let PMs re-claim needs_revision coordination roots

The lifecycle spec (CLAIM_RULES) grants CELL_PM/MAIN_PM claim of NEEDS_REVISION so a rejected coordination root (pr_fail / qa_fail / ceo_reject) can be re-claimed via i_will_plan and re-delegated. The runtime mapping _ROLE_CLAIM_STATUSES omitted it for PMs, so the spec gate allowed i_will_plan on a needs_revision root while the composed claim() rejected it -> returned None -> INVALID_STATE: the PM could neither plan nor idle its own rejected root and respawn-looped (observed live on cell root 0e49e04e, ~143 INVALID_STATE rejections across 11 PM sessions; the tail of the 2026-06-24/25 run).

Add NEEDS_REVISION to the cell_pm/main_pm runtime claim statuses, and a parity test locking _ROLE_CLAIM_STATUSES to lifecycle.CLAIM_RULES so the two can't drift again.
This commit is contained in:
Renn F
2026-06-25 07:21:11 +02:00
parent dfbb8649d0
commit bdb5e1a93a
2 changed files with 65 additions and 2 deletions
+17 -2
View File
@@ -75,8 +75,23 @@ _UUID_HYPHEN_COUNT = 4 # Number of hyphens in a UUID
_ROLE_CLAIM_STATUSES: dict[str, set[TaskStatus]] = {
"qa": {TaskStatus.PENDING, TaskStatus.AWAITING_QA},
"documenter": {TaskStatus.PENDING, TaskStatus.AWAITING_DOCUMENTATION},
"cell_pm": {TaskStatus.PENDING, TaskStatus.AWAITING_PM_REVIEW},
"main_pm": {TaskStatus.PENDING, TaskStatus.AWAITING_PM_REVIEW},
# PMs re-claim NEEDS_REVISION to recover a rejected coordination/assembled
# task (pr_fail / qa_fail / ceo_reject lands it there): the spec
# (lifecycle.CLAIM_RULES) grants it, and the runtime must match or the spec
# gate passes i_will_plan on a needs_revision root while the composed
# claim() returns None -> INVALID_STATE -> the PM respawn-loops on its own
# rejected root, unable to plan or idle it. Parity is locked by
# tests/unit/services/test_pm_claim_needs_revision.py.
"cell_pm": {
TaskStatus.PENDING,
TaskStatus.NEEDS_REVISION,
TaskStatus.AWAITING_PM_REVIEW,
},
"main_pm": {
TaskStatus.PENDING,
TaskStatus.NEEDS_REVISION,
TaskStatus.AWAITING_PM_REVIEW,
},
}
@@ -0,0 +1,48 @@
"""Runtime claim-status parity: PMs may re-claim a NEEDS_REVISION coordination task.
The lifecycle spec (``lifecycle.CLAIM_RULES``) lets ``CELL_PM`` / ``MAIN_PM``
claim ``NEEDS_REVISION`` so a rejected coordination / assembled task (pr_fail,
qa_fail, ceo_reject) can be re-claimed via ``i_will_plan`` and re-delegated.
The runtime claim path (``TaskService.claim`` ->
``_get_valid_claim_statuses`` -> ``_ROLE_CLAIM_STATUSES``) must honour the same
authority. Otherwise the spec gate *allows* ``i_will_plan`` on a
``needs_revision`` root, but the composed ``claim()`` inside the verb returns
``None`` (source status not in the runtime mapping) -> the verb runner raises
``INVALID_STATE`` -> the PM can neither plan nor idle its own rejected root and
respawn-loops on it (observed live 2026-06-25 on the ``0e49e04e`` cell root,
~143 INVALID_STATE rejections across 11 PM sessions).
"""
from __future__ import annotations
from types import SimpleNamespace
import pytest
from roboco.foundation.policy import lifecycle as spec
from roboco.models.base import TaskStatus
from roboco.services.task import _default_claim_statuses, _get_valid_claim_statuses
@pytest.mark.parametrize("role", ["cell_pm", "main_pm"])
def test_pm_runtime_claim_statuses_include_needs_revision(role: str) -> None:
agent = SimpleNamespace(role=role)
assert TaskStatus.NEEDS_REVISION in _get_valid_claim_statuses(
agent, allow_reassign=False
)
assert TaskStatus.NEEDS_REVISION in _default_claim_statuses(role)
@pytest.mark.parametrize("role", [spec.Role.CELL_PM, spec.Role.MAIN_PM])
def test_runtime_pm_claim_mapping_covers_spec_claim_rules(role: spec.Role) -> None:
"""The runtime mapping must cover every status the spec grants the role.
Guards against the spec (CLAIM_RULES) and the runtime mapping
(_ROLE_CLAIM_STATUSES) drifting apart again — the parity invariant.
"""
runtime_values = {s.value for s in _default_claim_statuses(role.value)}
for status in spec.CLAIM_RULES[role]:
assert status.value in runtime_values, (
f"runtime claim mapping for {role.value} is missing spec-allowed "
f"status '{status.value}'"
)