Skip to content

slow-agent-device: the simlock-kill test claims the lease ends, but ADR 0071 §4 says a standalone worker restart keeps it #1578

Open

Description

@V3RON

e2e/slow-agent-device.test.ts:517 ("after the simlock daemon is killed and started again, … the agent's next command fails with lease ended by simlock, and its old simlock lease is released") still claims ADR 0023 §3. ADR 0071 §4 and #1468 say a standalone worker's restart rebuilds the lease (adoptable, src/agent-device/front/host-link.ts:99) and the next heartbeat succeeds on the same device; agent-device.mdx:209 says the same. A SIGKILLed holder keeps its lease (src/leasing/AGENTS.md, Known gaps), so the adopt should apply.

Not run: it needs the slow lane. #1577's report says only the agent-device-kill test failed in three runs. If this one passes, the code contradicts the ADR; if it fails, its claim is stale like #1577's.

Found while triaging #1577.

Written by an agent.

Activity

  1. added
    bug:newReported. An agent may triage it (rule 8); the fast track may set it ready (rule 17).
    on Nov 15, 2026
  2. self-assigned this
    on Nov 15, 2026
  3. V3RON commented on Nov 15, 2026

    @V3RON
    ContributorAuthor

    Reproduction

    It fails: the test is stale, and the code follows ADR 0071 §4. The test is e2e/slow-agent-device.test.ts:517, unchanged on main. bug/1578-repro points at origin/main. I did not run the slow lane. I read a slow-lane run that was already going on this host (#1569's worktree at 2c6d716, with #1573's helper, Android). The old-pid wait passes and :535 fails:

    # /var/folders/tb/j39z273s5kn7lb9tz1dhk8l00000gn/T/simlock-slow-e2e-20261115-180422.log
    SLOW END failed 48s … (Android) > after the simlock daemon is killed and started again, …
    SLOW ERROR expected +0 not to be +0 // Object.is equality      # expect(next.code).not.toBe(0)
    # passed 2026-11-03, before #1466: simlock-slow-e2e-20261103-223944.log
    

    The right behaviour already passes at unit level:

    pnpm exec vitest run --project unit src/agent-device/edge-restart.test.ts -t "after a standalone worker restarts, the agent's next heartbeat succeeds"
    Tests  2 passed | 20 skipped (22)
    

    Root cause

    Defect, in the test. e2e/slow-agent-device.test.ts:534-541 still asserts ADR 0023 §3, which ADR 0071 replaced. The emulator outlives the SIGKILL, and the reconciler keeps a lease whose device is running (ADR 0019). The restarted front rebuilds the entry (adoptable, src/agent-device/front/host-link.ts:99), and the next heartbeat re-allocates the lease (agent-device.mdx:209). #1466 did not update this case, the twin of #1577.

    Simplest fix

    Use #1577's shape. Before the kill: install Appduct <app> and record listLeases. After the old pid is gone: ok(open <bundleId>) and ok(snapshot). Assert one lease, with the same id and deviceId. New title: "…pid is gone within 60 s; the agent opens the installed Appduct fixture app again and takes a snapshot, and its simlock lease stays, on the same device". The open also covers iOS (SESSION_NOT_FOUND comes before a new open).

    Alternatives rejected

    • Flip not.toBe(0) to ok(next) only: fails on iOS.
    • Delete the case: it is the only real-device check of ADR 0071 §4.

    Risk

    Fast track

    • Defect: yes
    • Fix tried: no — the test runs only on the slow lane
    • Nothing public added or broken: yes
    • No ADR or rule changed: yes
    • One fix: yes

    Written by an agent.

  4. removed their assignment
    on Nov 15, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    bug:newReported. An agent may triage it (rule 8); the fast track may set it ready (rule 17).

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions