fix(engine): split stuck requeue fallback handling

This commit is contained in:
Phil Larson
2026-05-31 10:57:54 -07:00
parent 5a163ce07a
commit a7bd44ca96
3 changed files with 54 additions and 8 deletions

View File

@@ -4,6 +4,7 @@ import "../executor-test-helpers.js";
import type { Task, TaskStore } from "@fusion/core";
import { TaskExecutor } from "../../executor.js";
import { SelfHealingManager } from "../../self-healing.js";
import { isRunnableQueuedOverlapCandidate } from "../../scheduler.js";
import { StuckTaskDetector } from "../../stuck-task-detector.js";
type MockTaskStore = TaskStore & EventEmitter & {
@@ -219,6 +220,7 @@ describe("reliability interactions: non-progress churn", () => {
expect(task.steps).toEqual([{ name: "Implement", status: "in-progress" }]);
expect(task.log?.some((entry) => entry.action.includes("incomplete task exhausted stuck kill budget"))).toBe(true);
expect(store.handoffToReview).not.toHaveBeenCalled();
expect(isRunnableQueuedOverlapCandidate(task, [task])).toBe(true);
manager.stop();
});

View File

@@ -494,6 +494,40 @@ describe("SelfHealingManager", () => {
);
});
it("logs post-move requeue patch failures without executor fallback", async () => {
(store.getTask as ReturnType<typeof vi.fn>).mockResolvedValue({
id: "FN-001",
column: "in-progress",
stuckKillCount: 6,
steps: [
{ name: "Preflight", status: "done" },
{ name: "Delivery", status: "in-progress" },
],
} as unknown as Task);
(store.updateTask as ReturnType<typeof vi.fn>)
.mockResolvedValueOnce({} as Task)
.mockRejectedValueOnce(new Error("write conflict"));
manager.start();
const result = await manager.checkStuckBudget("FN-001", "loop");
expect(result).toBe(false);
expect(store.moveTask).toHaveBeenCalledWith("FN-001", "todo", {
preserveProgress: true,
preserveStatus: true,
});
expect(store.logEntry).toHaveBeenCalledWith(
"FN-001",
"STUCK_LOOP_EXHAUSTED: incomplete task moved to todo with progress preserved, but post-move requeue patch failed (write conflict); scheduler retry may wait for the next state repair pass.",
);
expect(store.logEntry).toHaveBeenCalledWith(
"FN-001",
"STUCK_LOOP_EXHAUSTED: incomplete task exhausted stuck kill budget (7/6), last reason=loop. Re-queued in todo with progress preserved; scheduler may retry without manual unpause.",
);
expect(store.handoffToReview).not.toHaveBeenCalled();
});
it("terminalizes no-progress churn without incrementing stuck kill budget", async () => {
(store.getTask as ReturnType<typeof vi.fn>).mockResolvedValue({
id: "FN-001",

View File

@@ -1079,14 +1079,6 @@ export class SelfHealingManager {
preserveProgress: true,
preserveStatus: true,
});
const requeueUpdate = {
stuckKillCount: newCount,
paused: false,
userPaused: false,
pausedReason: null,
status: "queued",
} satisfies Parameters<typeof this.store.updateTask>[1] & { userPaused: boolean };
await this.store.updateTask(taskId, requeueUpdate);
} catch (moveErr: unknown) {
const moveErrMessage = moveErr instanceof Error ? moveErr.message : String(moveErr);
log.warn(`${taskId} moveTask(todo) failed (${moveErrMessage}) after incomplete STUCK_LOOP_EXHAUSTED terminalization — falling back to executor stuck-kill requeue`);
@@ -1096,6 +1088,24 @@ export class SelfHealingManager {
);
return true;
}
const requeueUpdate = {
stuckKillCount: newCount,
paused: false,
userPaused: false,
pausedReason: null,
status: "queued",
} satisfies Parameters<typeof this.store.updateTask>[1] & { userPaused: boolean };
try {
await this.store.updateTask(taskId, requeueUpdate);
} catch (patchErr: unknown) {
const patchErrMessage = patchErr instanceof Error ? patchErr.message : String(patchErr);
log.warn(`${taskId} post-move requeue patch failed after incomplete STUCK_LOOP_EXHAUSTED terminalization: ${patchErrMessage}`);
await this.store.logEntry(
taskId,
`STUCK_LOOP_EXHAUSTED: incomplete task moved to todo with progress preserved, but post-move requeue patch failed (${patchErrMessage}); scheduler retry may wait for the next state repair pass.`,
);
}
await this.store.logEntry(
taskId,
`STUCK_LOOP_EXHAUSTED: incomplete task exhausted stuck kill budget (${newCount}/${maxKills}), last reason=${reason}. Re-queued in todo with progress preserved; scheduler may retry without manual unpause.`,