[leonardoxr]

Closes #1115

Status

Implement: done Review: fix-check passed (round 1, round 2, fix-check), 5 fixed Hardware: pass at ac517aa (4/4 across a full run and one re-run after an emulator boot hang, #1183) Gate: running

Done when

• The Tests lines pass: src/daemon/tunnel-cli.e2e.test.ts, src/cli/tunnel.test.ts, src/tunnel/worker.test.ts, src/drivers/android/index.test.ts. — run in the fix runs; CI on the head
• e2e/slow-android-tunnel.test.ts passes in the slow lane with a gateway and a worker on one machine: simlock tunnel reverse <p> forward <q>; the device fetches the caller's page on <p>; a device-side server started with (printf 'device-ok'; sleep 30) | nc -l -p <q> after the tunnel prints its lines answers device-ok to a client on the caller's localhost:<q> that retries its connect until it succeeds, for at most 5 s, and receives the bytes that client sends. — slow lane at ac517aa, 2026-11-01: full file 3/4 (this test passed; the lone-worker test hit a boot timeout, Android: a stale emulator crash dump makes every boot wait on a crash-report dialog until the lease times out #1183), then the lone-worker test alone passed; 4/4 at the head
• The CLI reference describes forward. — cli.mdx, checked by the claims review

Assumptions

• Assumption: how the slow-lane test observes "receives the bytes that client sends" — the device server redirects nc's output to a file the test reads back; the spec's command is otherwise unchanged.
• Assumption: wording of the new usage and BAD_REQUEST messages and the RESET reasons (timeout, adb's own closed) — the spec fixes the codes, not the text; chose short messages matching the existing ones.

Review

Spec review: 2 blocking, 2 fixed, 0 notes. Code review: 2 blocking, 2 fixed, 1 notes. Claims review: 1 blocking, 1 fixed, 0 notes.
Rounds: 2. Tokens: 43.4M.
Runs: 25. Cost USD: 16.0376. Agent seconds: 10324.

Rejected:

• spec: src/daemon/tunnel-lease.e2e.test.ts:400, :532 and e2e/slow-android-tunnel.test.ts:446 tests outside simlock tunnel forward: reach a leased device's port from the caller's machine #1115's Tests lines — the maintainer asked for them under A remote caller's leased device reaches ports on the caller's machine, and back #1110 [intentional]
• claims: src/drivers/android/index.test.ts:3612 the title names a phone after its check, only the closed check is asserted — wording; behaviour proven [nonblocking]

Assessment: complete

Written by an agent.

[V3RON]

Review notes

Not blocking, not verified. Each is one reviewer's claim.

• code: src/cli/commands/tunnel.ts:534 on macOS a caller port another process holds on 0.0.0.0 or ::1 still binds on 127.0.0.1, so the CLI prints its forward line instead of TUNNEL_CALLER_PORT_IN_USE; the cause is the shared listener factory (src/ports/tcp-listener.ts:33-37), outside this diff

Written by an agent.