Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS
JavaScript ai-agentsdashboarddeveloper-toolsdigital-twinquantified-selfsecond-brainself-hostedtailscale
fix(cos): stop failure-artifact reaper from overwriting revived tasks or expiring running investigations (#12059)
[._.]/ Adam Eivy608dc751e297
diff --git a/server/services/cosTaskStore.js b/server/services/cosTaskStore.jsindex 8632cf38c9..346de67bb7 100644--- a/server/services/cosTaskStore.js+++ b/server/services/cosTaskStore.js@@ -1181,6 +1181,9 @@ export function isReapableBlockedFailure(task, { now = Date.now(), maxAgeMs = DEexport function isReapableInvestigation(task, tasksById) {if (!isInvestigationTask(task)) return false;if (task.status === 'completed') return false; // already terminal+ // An in-progress investigation has a live agent diagnosing it; reaping it+ // would let a later finalize regress the status or spawn a second run (#12059).+ if (task.status === 'in_progress') return false;const affected = Array.isArray(task.metadata?.affectedTasks) ? task.metadata.affectedTasks : [];if (affected.length === 0) return false;return affected.every((id) => {@@ -1230,11 +1233,13 @@ export async function sweepResolvedFailureTasks({let staleBlocks = 0;let investigations = 0;for (const { task, type, reason } of targets) {+ // Guard the flip under the write lock: a Resume/Run or dequeue that landed+ // after the snapshot must not be overwritten by a stale auto-expire (#12059).const updated = await updateTask(task.id, {status: 'completed',metadata: { resolution: 'auto-expired', autoExpiredReason: reason, autoExpiredAt: new Date(now).toISOString() }- }, type, { now });- if (updated?.error) continue;+ }, type, { now, expectedStatus: task.status });+ if (updated?.error || updated?.statusChanged) continue;if (reason === 'stale-failure-block') staleBlocks++; else investigations++;}diff --git a/server/services/cosTaskStore.test.js b/server/services/cosTaskStore.test.jsindex 2e73e064e3..29078ac208 100644--- a/server/services/cosTaskStore.test.js+++ b/server/services/cosTaskStore.test.js@@ -30,6 +30,7 @@ const mock = vi.hoisted(() => ({parseCalls: 0,// When set, atomicWrite rejects with it (a write that dies before the rename).atomicFailure: null,+ afterWrite: null,state: null,events: [],// Controls the mocked codeReview.js for resolveTaskChallengeWithRecheck (#2471).@@ -61,6 +62,8 @@ vi.mock('../lib/fileUtils.js', async (importOriginal) => {if (mock.atomicFailure) throw mock.atomicFailure;mock.files.set(p, content);mock.mtimes.set(p, (mock.mtimes.get(p) || 0) + 1000);+ // Lets a test mutate the file between two reaper flips (snapshot race).+ if (mock.afterWrite && mock.afterWrite(p)) mock.afterWrite = null;})};});@@ -162,6 +165,7 @@ const baseState = () => ({beforeEach(() => {mock.atomicFailure = null;+ mock.afterWrite = null;mock.files = new Map();mock.mtimes = new Map();mock.parseCalls = 0;@@ -2323,6 +2327,11 @@ describe('cosTaskStore — stale failure-artifact reaper (#2619)', () => {expect(isReapableInvestigation(investigation(), new Map())).toBe(true);});+ it('never reaps an in-progress investigation, even when its origins are all completed', () => {+ const byId = new Map([['task-a', { id: 'task-a', status: 'completed' }]]);+ expect(isReapableInvestigation(investigation({ status: 'in_progress' }), byId)).toBe(false);+ });+it('does not reap while any originating task is still live', () => {const byId = new Map([['task-a', { id: 'task-a', status: 'blocked' }]]);expect(isReapableInvestigation(investigation(), byId)).toBe(false);@@ -2421,6 +2430,44 @@ describe('cosTaskStore — stale failure-artifact reaper (#2619)', () => {expect(after.tasks.find(t => t.id === 'sys-inv2').status).toBe('pending');});+ it('leaves an in-progress investigation running even after its origin completes', async () => {+ await addTask({ description: 'origin fixed by hand', id: 'task-origin3', priority: 'HIGH' }, 'user', { now: NOW });+ await addTask({+ description: '[Auto] Investigate agent failure [fp]: running',+ id: 'sys-running',+ isInvestigation: true,+ affectedTasks: ['task-origin3']+ }, 'internal', { now: NOW });+ await updateTask('sys-running', { status: 'in_progress' }, 'internal', { now: NOW });+ await updateTask('task-origin3', { status: 'completed' }, 'user', { now: NOW });+ const res = await sweepResolvedFailureTasks({ now: NOW });+ expect(res.reaped).toBe(0);+ const after = await getCosTasks();+ expect(after.tasks.find(t => t.id === 'sys-running').status).toBe('in_progress');+ });++ it('does not auto-expire a target that was revived between the snapshot and its flip', async () => {+ await addTask({ description: 'stale one', id: 'task-race1', priority: 'HIGH' }, 'user', { now: NOW });+ await addTask({ description: 'stale two', id: 'task-race2', priority: 'HIGH' }, 'user', { now: NOW });+ for (const id of ['task-race1', 'task-race2']) {+ await updateTask(id, { status: 'blocked', metadata: { blockedCategory: 'max-retries', blockedAt: daysAgo(20) } }, 'user', { now: NOW });+ }+ // After the first flip lands, the user revives the second target (Resume+ // → in_progress). Its flip was planned from the pre-revive snapshot.+ mock.afterWrite = (p) => {+ if (p !== USER_FILE) return false;+ // Revive whichever target the sweep has not flipped yet.+ mock.files.set(p, mock.files.get(p).replace(/^(- )\[!\](.*#task-race)/m, '$1[~]$2'));+ mock.mtimes.set(p, (mock.mtimes.get(p) || 0) + 1000);+ return true;+ };+ const res = await sweepResolvedFailureTasks({ now: NOW });+ expect(res).toMatchObject({ reaped: 1, staleBlocks: 1 });+ const after = await getUserTasks();+ const statuses = after.tasks.filter(t => t.id.startsWith('task-race')).map(t => t.status).sort();+ expect(statuses).toEqual(['completed', 'in_progress']);+ });+it('bounds the number of flips per sweep to the limit', async () => {for (let i = 0; i < 4; i++) {await addTask({ description: `old ${i}`, id: `task-b${i}`, priority: 'LOW' }, 'user', { now: NOW });