packet-test

name

packet-test

description

Run an end-to-end AX.25 packet QSO test between two virtual wfweb rigs at 300 bd and 1200 bd. Boots the testrig bench, drives both browsers via Playwright MCP, exchanges connected-mode frames, screenshots both sides, and tears down. Use when the user asks to test packet, run a packet QSO, verify AX.25, or validate the Direwolf modem against a real terminal session. (9600 G3RUH is intentionally NOT covered.)

Run an end-to-end AX.25 packet QSO test between two virtual rigs at 300 bd and 1200 bd, drive both browsers via Playwright MCP, screenshot both sides, and tear down. (9600 G3RUH is intentionally NOT tested by this skill.)

Prerequisites (verify before starting; fix and stop if missing)

1. ./wfweb and ./tools/virtualrig/virtualrig are built. If not, build with qmake wfweb.pro && make -j$(nproc) and cd tools/virtualrig && qmake && make -j$(nproc).
2. Repo .mcp.json includes --ignore-https-errors for the playwright MCP. wfweb's cert is self-signed; without this flag chromium refuses with ERR_CERT_AUTHORITY_INVALID. Verify the running MCP has the flag: pgrep -af @playwright/mcp must show --ignore-https-errors. If .mcp.json was just edited, the user needs to restart Claude Code (not just /mcp reconnect — that re-uses cached args). Stop and tell them.
3. The playwright tools live behind ToolSearch — load them with select:mcp__playwright__browser_navigate,mcp__playwright__browser_snapshot,mcp__playwright__browser_click,mcp__playwright__browser_type,mcp__playwright__browser_take_screenshot,mcp__playwright__browser_wait_for,mcp__playwright__browser_evaluate,mcp__playwright__browser_close,mcp__playwright__browser_tabs,mcp__playwright__browser_select_option.

Boot the bench

./scripts/testrig.sh up 2

Spawns virtualrig + 2 wfweb daemons. Both rigs come up on 14.074.000 USB by default — same freq, no --broadcast needed (audio-gating on matching freq+mode is what real packet does).

• A → https://127.0.0.1:9080 (virtual-IC7300-A)
• B → https://127.0.0.1:9090 (virtual-IC7300-B)

Per-tab UI bring-up (do for both A and B)

The Packet panel lives inside the MODE overlay — there is no top-level Packet button.

1. browser_navigate to the rig URL.
2. Wipe stale browser state. Playwright reuses its profile across runs, so localStorage carries activeTab, last-used callsign, and panel visibility — those have produced races that don't reproduce in real-world use. browser_evaluate () => { localStorage.clear(); sessionStorage.clear(); }, then browser_navigate to the same URL again so init runs against an empty store.
3. Click the CLICK TO START splash button (browsers refuse AudioContext without a user gesture).
4. Click MODE, then click Packet in the mode grid. Rig stays on USB (Packet IS USB on HF).
5. Click the gear (#packetSettingsBtn), type the callsign in the Callsign input, click Save.
   • A: W1AAA, B: W1BBB. Format must be valid amateur (letter+digit+letters). The single callsign is shared with CW/FT8/FreeDV/APRS — do not add per-mode callsigns.
6. Click the baud button for this iteration (#packetMode300 or #packetMode1200).
7. Click the TERMINAL tab.
8. On A only: clear #termPeerCall and type the other rig's callsign (W1BBB).

When refs go stale across snapshots, drive directly via browser_evaluate with stable IDs:

• Buttons: packetSettingsBtn, packetMode300, packetMode1200, packetClearBtn, termConnectBtn, termDisconnectBtn, termSendBtn
• Fields: termPeerCall, termInput
• State: document.getElementById('termStateChip').textContent (DISCONNECTED / CONNECTING / CONNECTED / DISCONNECTING)
• Object: window.Packet.state.{mode, terminal.sessions.t1.{state, scrollback, ownCall, peerCall}}

QSO sequence (run for each baud: 300, then 1200)

1. A → Connect: click #termConnectBtn. Poll termStateChip.textContent === 'CONNECTED' (a few seconds at 300, faster at 1200). Verify B also flips to CONNECTED and B's session scrollback gets *** CONNECTED to W1AAA.
2. A → "Hello": set termInput.value = 'Hello', click #termSendBtn. Wait on B for Hello to render.
3. B → "How are you": same pattern. Wait on A for the text.
4. Screenshot both with browser_take_screenshot, fullPage: true. Pass the full path as filename: .playwright-mcp/wfweb-packet-<baud>-A.png and ...-B.png. The MCP does NOT auto-target .playwright-mcp/ — bare filenames land in the project root (the MCP's CWD). Do not use /tmp — it's outside the MCP's allowed roots and will fail.
5. B → Disconnect: click #termDisconnectBtn. Both chips go to DISCONNECTED.
6. Before 1200 bd: click #packetMode1200 on both tabs, re-set #termPeerCall on A (it may revert to placeholder), and optionally wipe window.Packet.state.terminal.sessions.t1.scrollback = [] then call window.Packet.renderTerm() for a clean screenshot.

Teardown and report

1. browser_close, then ./scripts/testrig.sh down.
2. Report PASS/FAIL per baud and the four screenshot paths (A and B at 300, A and B at 1200).

Pitfalls — read before debugging

• Don't poll browser_wait_for text 'DISCONNECTED' — there is a hidden <span id="digiReporterStatus">disconnected</span> that matches case-insensitively and never goes visible. Read termStateChip.textContent via browser_evaluate instead.
• Don't use curl https://127.0.0.1:* to probe — wfweb's self-signed cert needs a separate exception. Drive everything through the browser.
• Killing the playwright MCP (pkill -f @playwright/mcp) does not auto-respawn it. The user has to run /mcp to reconnect. Only do this if the MCP is genuinely stuck.
• Both rigs share one callsign field — do not try to set per-mode callsigns. Setting it once in PKT settings propagates to CW/FT8/FreeDV/APRS, which is by design.
• Screenshot saves are sandboxed. Allowed roots are the project dir and .playwright-mcp/; /tmp is rejected.
• Don't skip the localStorage.clear() step. Without it, prior runs leave activeTab='term' + visible=true in storage, so the next run boots straight into TERMINAL during synchronous init and fires termRegister before the WebSocket finishes opening. send() silently drops it (CONNECTING ≠ OPEN), the receiver never registers its callsign with the AX.25 link layer, and inbound SABMs from the connecting side time out after 10 retries. This is a test-environment artifact, not a real bug — humans always click something post-reload, which re-fires the registration. The clean-slate step makes the test mirror that.