Files
PaperClipAI/server/src/services/native-runtime/plugin-runner-task.test.ts
T
DottaandPaperclip 09ad9adce2 feat(server): adopt running remote agents through durable controller checkpoints (#17439)
## Thinking Path

> - Paperclip must keep admitted remote work alive while its main server
is replaced.
> - Runner owns recovery mechanics and exposes opaque checkpoints
through #17097.
> - The server owns company scope, encryption, Stop, tool authority and
execution ownership.
> - A successor must prove current ownership before it adopts the
original execution.
> - Checkpoint bursts must also leave capacity for normal requests and
lease renewal.
> - This PR persists and adopts that authority through the existing
runtime boundaries.

## Linked Issues or Issue Description

**What existing behavior does this improve?**

Remote agents retain the original execution during a bounded main-server
outage.

**Current behavior**

A server restart can lose the controller's local state and connection.
An unsafe replacement could repeat a turn or an uncertain mutation.
Checkpoint writes in the request pool can also delay ordinary requests
during an aligned burst.

**Proposed behavior**

Restore encrypted control records from Postgres and adopt the original
runner under a fenced lease. Recover known tool receipts through current
tool authority. Give control checkpoints a shared pool capped at two
connections per database owner.

Depends on #17097. Successors are #17455 and #17456. Qualification and
presentation remain separate review slices.

## What Changed

- Add company/run-bound encrypted checkpoint records and migration 0394.
Keep the SQL idempotent after renumbering against current master.
- Recover startup and expired ownership through native runtime and
heartbeat. Preserve the sandbox lease, admitted runner artifact, harness
configuration, native session and active turn.
- Keep Stop, finalization, preparation ownership and stale-writer
fencing transactional. Preserve graceful detach and natural crash
takeover.
- Recover committed tool receipts through the existing authority.
Reauthorize reads and hold uncertain mutations.
- Bound startup retention with the existing coordinator and preserve
zero idle scans.
- Share at most two checkpoint connections per database client. Keep
application and liveness capacity separate. Close the checkpoint pool
during shutdown; late writes cannot reopen it.
- Count SQL across both application and checkpoint clients. Prove
request/liveness progress, command recovery and failure after pool
shutdown with real Postgres.

## Verification

Project creation recovery now uses the existing project activity receipt
owner. It retains authenticated input normalization, fingerprint checks
and current run authority after the receipt lock. An absent receipt
returns no result before mutation. The new lost-HTTP-response and
no-effect regression uses real API routes and Postgres. The seven
transport cases pass locally. This Mac skips all 20 database cases. On
current source `5de803c2cd`, Linux [job
122593415890](https://github.com/paperclipai/paperclip/actions/runs/40620576702/job/122593415890)
executes all 20 project-tool cases successfully (9.480 seconds),
including the committed-project lost-response regression. Its shard
passes 1,627 cases with three unrelated skips. Server typecheck passes.
The transport regression initially has two failures; a subsequent
incorrect expectation about schema defaults was corrected before the
final pass. Ordinary execution adds no checkpoint writes or recurring
work.

Current candidate `5de803c2cd` includes the project-creation receipt
recovery fix and master `2cd747d1b6`. Migration 0394 retains the prior
idempotent checkpoint DDL. Every master SQL file and journal entry is
preserved. Snapshots 0390–0394, migration checks and server typecheck
pass.

The ordinary warm-turn regression fails on its second turn before
`a219ab51ad`. After correction, three completed warm turns select the
correct prior checkpoint and preserve the original session. All 288
provider-wiring cases pass. The control plane, external persistence
callback, identity rotation and filesystem are real; the backend and
database-load seam are scripted. The 13 warm admission/configuration
variants also pass after final master integration. This does not qualify
an underlying harness.

At current source `5de803c2cd`, Linux [job
122593415907](https://github.com/paperclipai/paperclip/actions/runs/40620576702/job/122593415907)
passes all 940 native-session-executor cases (61.431 seconds), including
the warm external-checkpoint behavior; the shard passes 2,358 cases with
one skip. Current [Runner job
122593415475](https://github.com/paperclipai/paperclip/actions/runs/40620576702/job/122593415475)
passes the native recovery lane: 115 passes and five container-only
skips, including the encrypted restore guard, fresh preparation (63.189
seconds), blocked claim lock (1.226 seconds) and recovery beyond 101
blocked owners (3.897 seconds). The first lane passes 2,215 cases with
one skip. The SQL cost test reports zero idle and streaming SQL for both
output volumes; this is component evidence, not a production capacity
claim. At prior candidate `1ff7bb44f4`, Linux job 122549307215 passes
all 938 native-executor cases, including the three-turn
external-checkpoint regression. Job 122549306469 passes the native
recovery lane with 115 passes and five container-only skips, including
the encrypted-restore session guard, fresh preparation (63.664 seconds),
claim-lock contention (3.961 seconds), and 101-owner recovery (3.824
seconds). This Mac skips the real-Postgres suite. The earlier full local
executor attempt has 921 passes and 16 failures during embedded-Postgres
initialization before assertions. The original fixture collisions and
their corrections remain recorded.

At earlier candidate `8fbe62b41d`, server shard job 122535673841 has
3,876 passes, one failure and one skip in an unchanged expiry-worker
retry/drain timing test. The same server/database tree on qualification
source `b64cac6951` passes all 20 expiry cases and all 3,877 shard
cases. The single diagnostic retry, job 122541891915, passes all 20
expiry cases and all 3,877 shard cases (one skip). No assertion was
weakened. The original failure remains recorded and its exact timing
race is not proven. All Actions checks and legacy status passed on
`8fbe62b41d`, with completed Greptile 5/5 and no open feedback. The next
candidate `1ff7bb44f4` also passed all Actions checks and Greptile 5/5.
Its Paperclip Review remained queued before the new master conflict
required integration. The new candidate needs fresh exact-head CI and
reviews; old checks do not qualify it. Migration generation is
checkpoint-only, all master SQL and journal entries remain unchanged,
and migration checks, server typecheck and all 13 warm-owner variants
pass after integration. The first migration check was blocked by TSX
sandbox IPC; its permitted repeat passes. The first warm-test command
used an incorrect path and ran no tests; the corrected command supplies
the 13-case receipt.

All 16 database pool-lifetime/options/teardown cases pass after granting
loopback permission; the first batch had two EPERM failures and 14
passes. Earlier semantic-tool/HTTP recovery, adoption, encryption-batch,
zero-streaming-write and idle-SQL receipts remain source-attributed in
the [qualification
record](https://github.com/paperclipai/paperclip/blob/2ff66f57da0da35bac1c19a2366ea4b84d87a986/doc/plans/2026-10-22-hot-upgrade-qualification.md).
That record retains earlier assertion, setup and cleanup failures and
their corrections. No failed gate is waived.

The historical 300-session component campaign passes four after phases:
steady read p99 is 107–108 ms at 500 reads/s and 100 writes/s;
aligned-burst read p99 is 103–105 ms after the bounded pool, versus
584–689 ms before. This comparison isolates the persistence corrections.
It does not measure the final stack's hosted active-agent capacity. The
hosted zero-agent API pair leaves SQL and CPU essentially unchanged, but
both arms fail absolute latency gates and candidate peak cgroup memory
fails its gate. The active-agent/viewer matrix, different-artifact
upgrade and rollback remain open.

Run `PAPERCLIP_TEST_DATABASE_URL=<owned-disposable-postgres> pnpm
--filter @paperclipai/server exec vitest run
src/services/native-runtime/native-runner-restart-recovery.integration.test.ts`.

Previous head `f2db8a6568` had one failed workspace setup gate: [job
122555672441](https://github.com/paperclipai/paperclip/actions/runs/40608984238/job/122555672441)
passes all 4,169 UI tests and 285 CLI cases (one skip), but the company
import/export suite reaches its 30-second health deadline during doctor
startup. Its output does not reach the server-start step. The CLI,
server, database, shared, adapter and plugin trees are identical on
qualification source `a866619d7b`, whose [job
122555886663](https://github.com/paperclipai/paperclip/actions/runs/40609057937/job/122555886663)
passes that fixture and all 286 CLI cases. The one diagnostic retry,
[job
122561057225](https://github.com/paperclipai/paperclip/actions/runs/40608984238/job/122561057225),
passes the formerly failing fixture and all 286 CLI cases. All Actions
checks and legacy status on `f2db8a6568` pass. Its exact-head Greptile
completed at 5/5 with no open feedback. The current integrated head
requires fresh checks and a completed Paperclip Review before merge. The
original failure remains recorded; its exact timing cause is unproven. A
repeated failure requires investigation and repair rather than another
retry.

All current-head CI, legacy status, exact-head completed Greptile 5/5,
resolved feedback and ordinary GitHub merge eligibility remain required
before merge.

The latest integration preserves master’s published Copilot v70 profiles
and every published migration. Checkpoint migration
`0394_conscious_whizzer.sql` was generated with `pnpm db:generate` and
retains the prior idempotent DDL. The Runner vendor rebuild and replay
parity checks pass. Server typecheck and all 13 warm-owner variants pass
after that rebuild. The imported report-lock patch is byte-identical to
master; its intentional patch-context whitespace is retained, and the
feature delta passes whitespace checks. No hosted experiment was run.

## Risks

The pool adds up to two active connections per instance, shared across
agents. Checkpoint delivery can wait behind that bound; it still
requires its commit. Steady component-test Postgres CPU rises by about
1.7 percentage points of a two-CPU allocation. These corrections do not
remove the stack's control and renewal writes. The migration and key
must survive replacement. Fresh-database tests and local load do not
establish different-artifact hosted upgrade or rollback continuity.

## Model Used

OpenAI GPT-6 through Codex, with reasoning, repository tools and code
execution. The runtime does not expose a more specific model ID or
context-window size.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [ ] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [ ] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

---------

Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-10-27 21:39:23 -05:00

489 lines
31 KiB
TypeScript

import { createHash } from "node:crypto";
import { describe, expect, it, vi } from "vitest";
import type { RunnerProviderLaunchRequirements } from "@paperclipai/paperclip-runner";
import type { RunnerProcessLaunchSpec } from "../../vendor/paperclip-runner/index.js";
import { createPluginRunnerTaskOwner } from "./plugin-runner-task.js";
const ingress = vi.hoisted(() => ({ connect: vi.fn() }));
vi.mock("../../realtime/runner-prp-outbound.js", () => ({ connectRunnerPrpIngress: ingress.connect }));
function launchSpec(): RunnerProcessLaunchSpec {
return { command: "unused", args: [], cwd: "/unused", environment: {}, nativeRunner: {
identity: { runnerInstanceId: "runner", environmentLeaseId: "lease", runId: "run",
normalizedSessionId: "session", turnId: "turn", itemId: "item" }, ticket: "synthetic-ticket",
diagnosticsDirectory: "/controller/private/diagnostics",
stateDirectory: "/state", runnerVersion: "fixture", runnerDigest: `sha256:${"a".repeat(64)}`,
maxOutboxBytes: 262144, p0ReserveBytes: 65536, lifecyclePolicy: { mode: "per_turn", idleTimeoutMs: null },
} };
}
const requirements: RunnerProviderLaunchRequirements = {
schema: "paperclip.runner.launch-requirements.v1", identity: launchSpec().nativeRunner!.identity,
provider: "codex", model: "fixture", instructions: "Core policy", runtimeContext: null,
settings: {}, approvalPolicy: "never", readOnly: false, preparedConversation: false,
includeCollaborationModeInstructions: false,
};
function launchRequirements(harness: string): RunnerProviderLaunchRequirements {
return harness === "claude" ? { ...requirements, provider: "acpx", settings: { acpxAgent: "claude" } } : requirements;
}
async function prepare(owner: ReturnType<typeof createPluginRunnerTaskOwner>, current = requirements) {
await owner.driver.prepareLaunch!(current);
}
function provider() {
let stopped = false;
const task = vi.fn(async (operation: any): Promise<unknown> => {
if (operation.kind === "check_project_auth") return { kind: "project_auth", taskId: "task", allowed: true,
provider: operation.provider, configurationId: operation.configurationId };
if (operation.kind === "prepare") return { kind: "prepared", taskId: "task", launch: {
configurationId: "configuration", requirementsDigest: `sha256:${"a".repeat(64)}`, provider: { kind: "codex" },
} };
if (operation.kind === "read_state") return { kind: "state", taskId: "task", state: {
schema: "paperclip.runner.durable.state.v1", ...operation.identity, lifecycle: "suspended",
} };
if (operation.kind === "stop") stopped = true;
if (operation.kind === "status") return { kind: "status", taskId: "task", phase: stopped ? "cancelled" : "running", executionStopped: stopped, sessionStateAvailable: stopped };
if (operation.kind === "connection") return { kind: "connection", taskId: "task", endpoint: {
kind: "authenticated_websocket", websocketUrl: "wss://fixture.example.test/connect", generation: "fixture", secretHeaders: [],
} };
return { kind: "accepted", taskId: "task" };
});
return { task };
}
describe("plugin native Runner ownership", () => {
it("rechecks revoked grants for unchanged project IDs before adopting retained resources", async () => {
const { task } = provider();
const perform = task.getMockImplementation()!;
let allowed = true;
task.mockImplementation(async op => op.kind === "check_project_auth"
? { ...await perform(op) as object, allowed } : perform(op));
const input = { task, projectIds: ["project-retained"], harness: "codex", projectAuthCheckVersion: 1 as const,
adoptIdentity: requirements.identity };
const owner = createPluginRunnerTaskOwner(input);
await owner.driver.authorizeRunnerAdoption!("retained-configuration", requirements.identity);
allowed = false;
owner.rebind({ ...input });
await expect(owner.driver.authorizeRunnerAdoption!("retained-configuration", requirements.identity)).rejects.toThrow("project authentication conflicts");
expect(task.mock.calls.map(([op]) => op)).toEqual(Array(2).fill({ kind: "check_project_auth",
configurationId: "retained-configuration", provider: "openai", projectIds: ["project-retained"] }));
});
it.each(["identity", "unsupported", "released", "late-release"])("refuses adoption authorization after %s", async mode => {
const { task } = provider();
const perform = task.getMockImplementation()!;
let finish!: () => void;
const gate = new Promise<void>(resolve => { finish = resolve; });
task.mockImplementation(async op => { if (mode === "late-release" && op.kind === "check_project_auth") await gate; return perform(op); });
const owner = createPluginRunnerTaskOwner({ task, projectIds: ["project-current"], harness: "codex",
projectAuthCheckVersion: mode === "unsupported" ? undefined : 1, adoptIdentity: requirements.identity });
const registration = await owner.registration({} as never, "run");
if (mode === "released") await registration.release("detach");
const attempt = owner.driver.authorizeRunnerAdoption!("retained-configuration",
mode === "identity" ? { ...requirements.identity, runId: "unrelated" } : requirements.identity);
const rejection = expect(attempt).rejects.toThrow(mode === "identity" ? "identity mismatch"
: mode === "unsupported" ? "validation unavailable" : "preparation cancelled");
if (mode === "late-release") { await registration.release("detach"); finish(); }
await rejection;
await registration.release("detach");
expect(task.mock.calls.map(([op]) => op.kind)).toEqual(mode === "late-release" ? ["check_project_auth"] : []);
});
it("adopts the saved task and detaches without preparing, submitting, or stopping it", async () => {
const { task } = provider();
const close = vi.fn(async () => {});
ingress.connect.mockReturnValue({ ready: Promise.resolve(), close });
const owner = createPluginRunnerTaskOwner({ task, projectIds: [], harness: "codex",
adoptIdentity: requirements.identity });
const registration = await owner.registration({} as never, "run");
registration.activate(); await registration.ready();
expect(await owner.driver.adoptExistingRunner!.isAlive()).toBe(true);
expect(owner.driver.adoptExistingRunner!.pid).toBeNull();
expect(() => owner.launcher(launchSpec())).toThrow("replacement requires a new task lease");
await registration.release("detach");
expect(close).toHaveBeenCalledOnce();
expect(task.mock.calls.map(([op]) => op.kind)).toEqual(["status", "read_state", "connection", "status"]);
});
it.each(["stopped", "wrong-identity", "revoked", "cancelled"])("refuses adoption when the saved task is %s", async mode => {
const { task } = provider();
const perform = task.getMockImplementation()!;
const abort = new AbortController();
task.mockImplementation(async op => {
if (mode === "revoked" && op.kind === "read_state") throw new Error("execution lease is not active");
const result = await perform(op) as any;
if (mode === "stopped" && op.kind === "status") result.executionStopped = true;
if (mode === "wrong-identity" && op.kind === "read_state") result.state.runId = "other-run";
if (mode === "cancelled" && op.kind === "read_state") abort.abort();
return result;
});
ingress.connect.mockClear();
const owner = createPluginRunnerTaskOwner({ task, projectIds: [], harness: "codex",
adoptIdentity: requirements.identity });
const registration = await owner.registration({} as never, "run", abort.signal);
registration.activate();
await expect(registration.ready()).rejects.toThrow();
await registration.release("detach");
expect(ingress.connect).not.toHaveBeenCalled();
expect(task.mock.calls.some(([op]) => ["prepare", "submit", "stop", "complete"].includes(op.kind))).toBe(false);
});
it("settles an adopted task through confirmed termination and retained-session ownership", async () => {
const { task } = provider();
const retained = vi.fn(async () => {});
ingress.connect.mockReturnValue({ ready: Promise.resolve(), close: async () => {} });
const owner = createPluginRunnerTaskOwner({ task, projectIds: [], harness: "codex",
adoptIdentity: requirements.identity, onSessionRetained: retained });
const registration = await owner.registration({} as never, "run");
registration.activate(); await registration.ready();
await registration.release(); await registration.checkpoint("settled");
expect(retained).toHaveBeenCalledWith(expect.objectContaining(requirements.identity));
expect(task.mock.calls.map(([op]) => op.kind)).toEqual(["status", "read_state", "connection", "stop", "status", "read_state", "complete"]);
});
it("submits only a prepared identity and configuration ID", async () => {
const { task } = provider();
const owner = createPluginRunnerTaskOwner({ projectAuthCheckVersion: 1, task, projectIds: [], harness: "codex", pollMs: 1 });
expect(() => owner.launcher(launchSpec())).toThrow("requires preparation");
await prepare(owner);
expect(task.mock.calls[0][0]).toEqual({ kind: "prepare", requirements, projectIds: [] });
expect(task.mock.calls[1][0]).toMatchObject({ kind: "check_project_auth", configurationId: "configuration", provider: "openai" });
expect(() => owner.launcher({ ...launchSpec(), nativeRunner: { ...launchSpec().nativeRunner!,
identity: { ...requirements.identity, runId: "other" } } })).toThrow("structured per-turn");
ingress.connect.mockReturnValue({ ready: Promise.resolve(), close: async () => {} });
const registration = await owner.registration({} as never, "run");
const handle = owner.launcher(launchSpec());
registration.activate(); await registration.ready();
await registration.release(); await registration.checkpoint("unsettled"); await handle.completion;
expect(task.mock.calls.find(([op]) => op.kind === "submit")?.[0]).toMatchObject({ preparedLaunchId: "configuration" });
});
it("omits duplicate project validation only when the provider enforces it at submission", async () => {
const { task } = provider();
const perform = task.getMockImplementation()!;
task.mockImplementation(async op => op.kind === "prepare"
? { ...await perform(op) as object, projectAuthCheckAtSubmit: true } : perform(op));
const owner = createPluginRunnerTaskOwner({ projectAuthCheckVersion: 1,
task, projectIds: [], harness: "codex", pollMs: 1 });
await prepare(owner);
expect(task.mock.calls.map(([op]) => op.kind)).toEqual(["prepare"]);
ingress.connect.mockReturnValue({ ready: Promise.resolve(), close: async () => {} });
const registration = await owner.registration({} as never, "run");
const handle = owner.launcher(launchSpec());
registration.activate(); await registration.ready();
await registration.release(); await registration.checkpoint("unsettled"); await handle.completion;
expect(task.mock.calls.filter(([op]) => op.kind === "submit")).toHaveLength(1);
const unsupported = createPluginRunnerTaskOwner({ task, projectIds: [], harness: "codex" });
await expect(prepare(unsupported)).rejects.toThrow("validation unavailable");
});
it("does not accept a late preparation after task release", async () => {
let finish!: () => void;
const gate = new Promise<void>(resolve => { finish = resolve; });
const { task } = provider(); const perform = task.getMockImplementation()!;
task.mockImplementation(async op => { if (op.kind === "prepare") await gate; return perform(op); });
const owner = createPluginRunnerTaskOwner({ projectAuthCheckVersion: 1, task, projectIds: [], harness: "codex" });
const prepared = prepare(owner);
const registration = await owner.registration({} as never, "run");
await registration.release(); finish();
await expect(prepared).rejects.toThrow("preparation cancelled");
expect(task.mock.calls.some(([op]) => op.kind === "submit")).toBe(false);
});
it.each(["unsupported", "denied", "wrong-configuration", "wrong-provider"])("withholds credentials when project validation is %s", async mode => {
const { task } = provider(); const perform = task.getMockImplementation()!;
task.mockImplementation(async op => op.kind === "check_project_auth"
? { kind: "project_auth", taskId: "task", allowed: mode !== "denied",
configurationId: mode === "wrong-configuration" ? "other" : op.configurationId,
provider: mode === "wrong-provider" ? "anthropic" : op.provider } : perform(op));
const owner = createPluginRunnerTaskOwner({ task, projectIds: [], harness: "codex",
projectAuthCheckVersion: mode === "unsupported" ? undefined : 1,
credentialIdentity: "revision", accountIdentity: "account", credential: { type: "codex-auth-json", authJson: '{"OPENAI_API_KEY":"never-transfer"}' } });
await expect(prepare(owner)).rejects.toThrow(mode === "unsupported" ? "validation unavailable" : "conflicts");
expect(() => owner.launcher(launchSpec())).toThrow("requires preparation");
expect(JSON.stringify(task.mock.calls)).not.toContain("never-transfer");
expect(task.mock.calls.some(([op]) => op.kind === "submit")).toBe(false);
});
it("rejects an installed account when the plugin lacks project validation", async () => {
const { task } = provider();
const owner = createPluginRunnerTaskOwner({ task, projectIds: [], harness: "codex" });
await expect(prepare(owner)).rejects.toThrow("validation unavailable");
expect(task.mock.calls.some(([op]) => op.kind === "submit")).toBe(false);
});
it("collects refreshed credentials after process termination and before releasing task grants", async () => {
const { task } = provider();
const perform = task.getMockImplementation()!;
const credential = { type: "codex-auth-json" as const, authJson: '{"OPENAI_API_KEY":"fixture-key"}' };
let copied = false;
task.mockImplementation(async operation => {
if (operation.kind === "credentials") {
expect(task.mock.calls.some(([op]) => op.kind === "stop")).toBe(true);
return { kind: "credentials", taskId: "task", credential };
}
if (operation.kind === "complete") expect(copied).toBe(true);
return perform(operation);
});
ingress.connect.mockReturnValue({ ready: Promise.resolve(), close: async () => {} });
const owner = createPluginRunnerTaskOwner({ projectAuthCheckVersion: 1, task, projectIds: [], harness: "codex", credential,
credentialIdentity: "revision", accountIdentity: "stable-account", pollMs: 1,
onCredentials: async value => { expect(value).toEqual(credential); copied = true; } });
await prepare(owner);
const registration = await owner.registration({} as never, "run");
const process = owner.launcher(launchSpec()); registration.activate(); await registration.ready();
expect(task.mock.calls.find(([op]) => op.kind === "submit")?.[0]).toMatchObject({ credential, credentialIdentity: "revision", accountIdentity: "stable-account" });
await registration.release(); await registration.checkpoint("settled"); await process.completion;
expect(copied).toBe(true);
});
it("retains grants when the credential handoff fails", async () => {
const { task } = provider();
const perform = task.getMockImplementation()!;
task.mockImplementation(async op => op.kind === "credentials" ? { kind: "credentials", taskId: "task", credential: { type: "codex-auth-json", authJson: "{}" } } : perform(op));
ingress.connect.mockReturnValue({ ready: Promise.resolve(), close: async () => {} });
const owner = createPluginRunnerTaskOwner({ projectAuthCheckVersion: 1, task, projectIds: [], harness: "codex", accountIdentity: "account", pollMs: 1,
onCredentials: async () => { throw new Error("credential handoff failed"); } });
await prepare(owner);
const registration = await owner.registration({} as never, "run");
const process = owner.launcher(launchSpec()); registration.activate(); await registration.ready(); await registration.release();
await expect(registration.checkpoint("settled")).rejects.toThrow("credential handoff failed");
await process.completion;
expect(task.mock.calls.some(([op]) => op.kind === "complete")).toBe(false);
});
it.each([false, true])("records retained sessions after status failures (retry exhausted: %s)", async exhausted => {
const { task } = provider();
const perform = task.getMockImplementation()!;
let failStatus = false;
let clock = Date.now();
const now = vi.spyOn(Date, "now").mockImplementation(() => clock);
task.mockImplementation(async operation => {
if (operation.kind === "stop") failStatus = false;
if (operation.kind === "status" && failStatus) {
failStatus = exhausted;
if (exhausted) clock += 16_000;
throw new Error("status connection lost");
}
return perform(operation);
});
ingress.connect.mockReturnValue({ ready: Promise.resolve(), close: async () => {} });
const checkpoint = vi.fn(async () => {});
const owner = createPluginRunnerTaskOwner({ projectAuthCheckVersion: 1, task, projectIds: [], harness: "codex", onSessionRetained: checkpoint, pollMs: 1,
});
await prepare(owner);
const registration = await owner.registration({} as never, "run");
const handle = owner.launcher(launchSpec());
const completed = vi.fn(); void handle.completion.then(completed);
try {
registration.activate(); await registration.ready();
expect(handle.restart).toBeNull();
failStatus = true;
if (exhausted) await expect(registration.failure).rejects.toThrow("reconcile the saved task");
else await vi.waitFor(() => expect(failStatus).toBe(false));
expect(completed).not.toHaveBeenCalled();
await registration.release(); await registration.checkpoint("settled");
await expect(handle.completion).resolves.toMatchObject({ code: 1 });
expect(checkpoint).toHaveBeenCalledOnce();
expect(task.mock.calls.filter(([op]) => op.kind === "submit")).toHaveLength(1);
expect(task.mock.calls.filter(([op]) => op.kind === "complete")).toHaveLength(1);
} finally { now.mockRestore(); handle.child.kill(); }
});
it("exposes connection failure before activation and forwards late recovery exhaustion", async () => {
const { task } = provider();
let fail!: (error: Error) => void;
const failure = new Promise<never>((_resolve, reject) => { fail = reject; });
ingress.connect.mockReturnValue({ ready: Promise.resolve(), failure, close: async () => {} });
const owner = createPluginRunnerTaskOwner({ projectAuthCheckVersion: 1, task, projectIds: [], harness: "codex", pollMs: 1 });
await prepare(owner);
const registration = await owner.registration({} as never, "run");
const observed = registration.failure.catch(error => error);
const handle = owner.launcher(launchSpec());
registration.activate(); await registration.ready();
const error = new Error("runner ingress recovery exhausted"); fail(error);
expect(await observed).toBe(error);
expect(task.mock.calls.find(([op]) => op.kind === "stop")).toBeUndefined();
await registration.release(); await registration.checkpoint("settled"); await handle.completion;
});
it("reconciles a lost submit reply without launching another attempt", async () => {
const { task } = provider();
const perform = task.getMockImplementation()!;
let lost = true, statusLost = true;
task.mockImplementation(async op => {
if (op.kind === "submit" && lost) { lost = false; throw new Error("lost submit reply"); }
if (op.kind === "status" && statusLost) { statusLost = false; throw new Error("status temporarily unavailable"); }
return perform(op);
});
ingress.connect.mockReturnValue({ ready: Promise.resolve(), close: async () => {} });
const owner = createPluginRunnerTaskOwner({ projectAuthCheckVersion: 1, task, projectIds: [], harness: "codex",
pollMs: 1 });
await prepare(owner);
const registration = await owner.registration({} as never, "run");
const process = owner.launcher(launchSpec());
try {
registration.activate(); await registration.ready();
expect(task.mock.calls.filter(([operation]) => operation.kind === "submit")).toHaveLength(1);
expect(task.mock.calls.find(([operation]) => operation.kind === "submit")?.[0].launch).not.toHaveProperty("diagnosticsDirectory");
expect(() => owner.launcher(launchSpec())).toThrow("new task lease");
await registration.release(); await registration.checkpoint("settled"); await process.completion;
expect(task.mock.calls.filter(([operation]) => operation.kind === "complete")).toHaveLength(1);
} finally { process.child.kill(); await process.completion.catch(() => {}); }
});
it("retains task grants when retention recording fails after route release", async () => {
const { task } = provider();
ingress.connect.mockReturnValue({ ready: Promise.resolve(), close: async () => {} });
const owner = createPluginRunnerTaskOwner({ projectAuthCheckVersion: 1, task, projectIds: [], harness: "codex", onSessionRetained: async () => { throw new Error("retention write failed"); }, pollMs: 1 });
await prepare(owner);
const registration = await owner.registration({} as never, "run");
const process = owner.launcher(launchSpec());
registration.activate(); await registration.ready(); await registration.release();
expect(task.mock.calls.some(([operation]) => operation.kind === "complete")).toBe(false);
await expect(registration.checkpoint("settled")).rejects.toThrow("retention write failed");
await process.completion;
expect(task.mock.calls.some(([operation]) => operation.kind === "complete")).toBe(false);
});
it("detaches a controller without stopping its durable task", async () => {
const { task } = provider();
const close = vi.fn(async () => {});
ingress.connect.mockReturnValue({ ready: Promise.resolve(), close });
const owner = createPluginRunnerTaskOwner({ projectAuthCheckVersion: 1, task, projectIds: [], harness: "codex",
pollMs: 1 });
await prepare(owner);
const registration = await owner.registration({} as never, "run");
owner.launcher(launchSpec());
registration.activate(); await registration.ready(); await registration.release("detach");
expect(close).toHaveBeenCalledOnce();
expect(task.mock.calls.some(([operation]) => operation.kind === "stop" || operation.kind === "complete")).toBe(false);
});
it("prevents a late submission after cancellation during dispatch recording", async () => {
let finish!: () => void;
const prepared = new Promise<void>(resolve => { finish = resolve; });
const { task } = provider();
const owner = createPluginRunnerTaskOwner({ projectAuthCheckVersion: 1, task, projectIds: [], harness: "codex",
onDispatched: () => prepared, pollMs: 1 });
await prepare(owner);
const registration = await owner.registration({} as never, "run");
const process = owner.launcher(launchSpec());
registration.activate();
process.child.kill(); await registration.release(); finish();
await expect(process.completion).resolves.toMatchObject({ code: 1 });
await expect(registration.ready()).rejects.toThrow("cancelled");
expect(task.mock.calls.some(([op]) => op.kind === "submit")).toBe(false);
});
it("does not connect a released or aborted startup to the PRP authority", async () => {
const { task } = provider();
ingress.connect.mockClear();
const controller = new AbortController(); controller.abort();
const owner = createPluginRunnerTaskOwner({ projectAuthCheckVersion: 1, task, projectIds: [], harness: "codex",
pollMs: 1 });
await prepare(owner);
const registration = await owner.registration({} as never, "run", controller.signal);
const process = owner.launcher(launchSpec());
try {
registration.activate();
await expect(registration.ready()).rejects.toThrow("cancelled");
expect(task.mock.calls.filter(([operation]) => operation.kind === "connection")).toHaveLength(0);
expect(ingress.connect).not.toHaveBeenCalled();
await registration.release(); await registration.checkpoint("settled"); await process.completion;
} finally { process.child.kill(); await process.completion.catch(() => {}); }
});
});
describe("plugin task workspace file transfer", () => {
it("reads only a submitted task, validates bytes, and rejects late responses after release", async () => {
const {task}=provider(); const perform=task.getMockImplementation()!;
const bytes=Buffer.from("VM proof\n");
let file={kind:"file",taskId:"task",base64:bytes.toString("base64"),byteSize:bytes.length,sha256:createHash("sha256").update(bytes).digest("hex")};
let afterRead=async()=>{};
task.mockImplementation(async op=>{if(op.kind==="read_file"){await afterRead();return file;}return perform(op);});
const owner=createPluginRunnerTaskOwner({projectAuthCheckVersion:1,task,projectIds:[],harness:"codex",pollMs:1});
await expect(owner.readWorkspaceFile("proof.txt",1024)).rejects.toThrow("unavailable");
await prepare(owner);
ingress.connect.mockReturnValue({ready:Promise.resolve(),close:async()=>{}});
const registration=await owner.registration({} as never,"run");
const handle=owner.launcher(launchSpec());registration.activate();await registration.ready();
try {
await expect(owner.readWorkspaceFile("proof.txt",1024)).resolves.toEqual(bytes);
await expect(owner.readWorkspaceFile("/home/agent/proof.txt",1024)).resolves.toEqual(bytes);
expect(task.mock.calls.find(([op])=>op.kind==="read_file")?.[0]).toMatchObject({identity:requirements.identity,path:"proof.txt",projectIds:[]});
for(const path of ["../escape","//etc/passwd","file:/secret","a/../b","a\\b"]) await expect(owner.readWorkspaceFile(path,1024)).rejects.toThrow();
await expect(owner.readWorkspaceFile("proof.txt",1)).rejects.toThrow("unavailable");
file={...file,sha256:"f".repeat(64)};
await expect(owner.readWorkspaceFile("proof.txt",1024)).rejects.toThrow("integrity");
afterRead=()=>registration.release();
await expect(owner.readWorkspaceFile("proof.txt",1024)).rejects.toThrow("unavailable");
} finally {await registration.release();await registration.checkpoint("unsettled");await handle.completion;}
});
});
it.each(["codex", "claude"])("reuses a warm %s process across run authorities without idle status polling", async harness => {
const {task} = provider();
ingress.connect.mockReturnValue({ready:Promise.resolve(),close:async()=>{}});
const retained = vi.fn(async()=>{});
const input = {task,projectIds:[],harness,projectAuthCheckVersion:1 as const,warmTaskVersion:1 as const,
accountIdentity:"account",credentialIdentity:"revision",pollMs:1,onSessionRetained:retained};
const owner = createPluginRunnerTaskOwner(input);
await prepare(owner, launchRequirements(harness));
const registration = await owner.registration({} as never,"run");
const spec = launchSpec(); spec.nativeRunner!.lifecyclePolicy = {mode:"warm",idleTimeoutMs:300000};
const handle = owner.launcher(spec); registration.activate(); await registration.ready();
const observed = task.mock.calls.length;
await new Promise(resolve=>setTimeout(resolve,10));
expect(task.mock.calls).toHaveLength(observed);
const originalTask=task.getMockImplementation()!;
const nextTask=vi.fn(async op=>{
const result=await originalTask(op) as Record<string,unknown>;
return op.kind === "prepare" ? {...result,projectAuthCheckAtSubmit:true} : result;
});
owner.rebind({...input,task:nextTask});
const nextRegistration=await owner.registration({} as never,"next-run");
const nextIdentity={...requirements.identity,runId:"next-run",turnId:"next-turn",itemId:"next-item"};
await owner.driver.prepareLaunch!({...launchRequirements(harness),identity:nextIdentity});
expect(nextTask.mock.calls[0][0]).toMatchObject({kind:"prepare",priorIdentity:requirements.identity,accountIdentity:"account"});
expect(nextTask.mock.calls.map(([op])=>op.kind)).toEqual(["prepare","check_project_auth"]);
await registration.release(); nextRegistration.activate(); await nextRegistration.ready();
expect(nextTask.mock.calls.some(([op])=>op.kind==="submit" || op.kind==="stop" || op.kind==="complete")).toBe(false);
await nextRegistration.release(); await nextRegistration.checkpoint("settled"); await handle.completion;
expect(retained).toHaveBeenCalledWith(expect.objectContaining(nextIdentity));
expect(nextTask.mock.calls.filter(([op])=>op.kind==="stop")).toHaveLength(1);
});
it.each(["codex", "claude"])("adopts a verified retained %s task without submitting a replacement", async harness => {
const {task}=provider();
ingress.connect.mockReturnValue({ready:Promise.resolve(),close:async()=>{}});
const owner=createPluginRunnerTaskOwner({task,projectIds:[],harness,projectAuthCheckVersion:1,
warmTaskVersion:1,accountIdentity:"account",adoptIdentity:requirements.identity,pollMs:1});
const recovered = {configurationId:"saved",requirementsDigest:`sha256:${"a".repeat(64)}`,provider:{}};
await expect(owner.driver.prepareLaunch!(launchRequirements(harness),recovered)).resolves.toEqual(recovered);
expect(task.mock.calls.map(([op])=>op.kind)).toEqual(["check_project_auth"]);
expect(task.mock.calls[0][0]).toMatchObject({configurationId:"saved",projectIds:[]});
expect(await owner.driver.adoptExistingRunner!.isAlive()).toBe(true);
const registration=await owner.registration({} as never,"run");
registration.activate(); await registration.ready();
await registration.release(); await registration.checkpoint("settled");
expect(task.mock.calls.some(([op])=>op.kind==="submit")).toBe(false);
expect(task.mock.calls.some(([op])=>op.kind==="complete")).toBe(true);
});
it.each(["codex", "claude"])("denies retained %s process adoption when current project authentication is revoked",async harness=>{
const {task}=provider();
task.mockImplementation(async op=>op.kind === "check_project_auth"
? {kind:"project_auth",taskId:"task",configurationId:op.configurationId,provider:op.provider,allowed:false} as never
: {kind:"accepted",taskId:"task"} as never);
const owner=createPluginRunnerTaskOwner({task,projectIds:["project"],harness,projectAuthCheckVersion:1,
warmTaskVersion:1,accountIdentity:"account",adoptIdentity:requirements.identity});
await expect(owner.driver.prepareLaunch!(launchRequirements(harness),{configurationId:"saved",requirementsDigest:`sha256:${"a".repeat(64)}`,provider:{}}))
.rejects.toThrow("project authentication conflicts");
expect(task.mock.calls.map(([op])=>op.kind)).toEqual(["check_project_auth"]);
});
it.each(["codex", "claude"])("refuses warm %s account, harness, project, or credential generation changes", async harness=>{
const {task}=provider();
const input={task,projectIds:[],harness,projectAuthCheckVersion:1 as const,warmTaskVersion:1 as const,
accountIdentity:"account",credentialIdentity:"revision",pollMs:1};
const owner=createPluginRunnerTaskOwner(input);
await prepare(owner, launchRequirements(harness));
const spec=launchSpec();spec.nativeRunner!.lifecyclePolicy={mode:"warm",idleTimeoutMs:300000};
const process=owner.launcher(spec);
expect(()=>owner.rebind({...input,accountIdentity:"other"})).toThrow("cannot be rebound");
expect(()=>owner.rebind({...input,credentialIdentity:"rotated"})).toThrow("cannot be rebound");
expect(()=>owner.rebind({...input,harness:harness === "codex" ? "claude" : "codex"})).toThrow("cannot be rebound");
expect(()=>owner.rebind({...input,projectIds:["foreign"]})).toThrow("cannot be rebound");
const registration=await owner.registration({} as never,"run");
await registration.checkpoint("unsettled"); await process.completion;
});