@deeeed/metamask-harness 0.3.8 → 0.3.9

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md CHANGED
@@ -1,19 +1,43 @@
1
1
  # Changelog
2
2
 
3
+ ## 0.3.9 - 2026-07-04
4
+
5
+ ### Fixed
6
+ - **Readiness is presence-authoritative** — a no-baseline mtime-only stale verdict no longer re-runs `yarn setup` inside every launch (warning instead; fingerprint-verified drift still installs), the deps baseline is recorded after any successful setup, and the `ready` decision probes that the dev client is actually running on the target device before doing nothing (`launch` launches).
7
+ - **Slot-context-first resolution** — device/ports resolve from the checkout's `agentic-runtime.json` before pool/formula fallbacks, for mobile (simulator, Metro port) and extension (CDP + watcher ports, overwriting stale inherited env); naming formulas no longer hardcode pool prefixes.
8
+ - **UTF-8 locale for pod-triggering spawns** — CocoaPods no longer dies on multibyte podspecs under locale-less orchestrator environments.
9
+ - **Hermetic contract tests** — tmux/xcrun/adb are stubbed; the Metro log-window opens only in a run-owned session, never an unrelated attached one.
10
+
11
+ ### Added
12
+ - **`mm-harness stop`** — stop the port-scoped Metro this checkout owns and close its log-tail window; idempotent; teaching error on extension/core.
13
+ - **Extension launch streams live** with an upfront intent line (ports named); mobile and extension long operations are never silent.
14
+ - **Doctor reports live runtime state per platform** (deps currency, Metro on the slot's resolved port, dev-client presence) with adapter-scoped, colored runtime-context rows.
15
+ - **Colored grouped help with DEV OVERRIDE banner and SLOT line**; `mmdev-harness` completions; outcome-stating launch summaries with a `[dev]` marker.
16
+ - **docs/UX-PRINCIPLES.md** — the CLI interaction contract (human surface rules; `--json` envelope purity is exempt and inviolable).
17
+
18
+ ## Unreleased
19
+
20
+ ### Fixed
21
+ - **Mobile preflight no longer re-runs `yarn setup` on every launch** — the deps readiness decision escalated an mtime hint to `stale` and ran the FULL `yarn setup` (submodules + inpage bridge + jetify + pods) inside the launch phase on every invocation in orchestrated (farmslot) contexts. `depsCheck` reports `stale` for a no-baseline checkout whenever a manifest is newer than the install markers; the orchestrator's git phase refreshes tracked-file mtimes on every sync while its deps phase leaves `node_modules` untouched, so "manifest newer than markers" is normal and not proof of drift — yet it duplicated the orchestrator's deps work and turned a fast preflight into 15+ minutes (the behavior that also looked like a hang before 0.3.8's streaming). **Presence is now authoritative:** without a recorded baseline an mtime-only `stale` is trusted as `current` (with a one-line stderr warning), so a churned-but-present checkout launches with zero `yarn` in both tiers. Genuine drift is still caught by the recorded-baseline fingerprint. `mm-harness launch` (quick tier / no `--build`) also dropped `preflightMode` before deciding, so the fast contract never took effect; it is threaded through now.
22
+ - **Fast tier never installs; genuine gaps teach the orchestrator** — in the fast tier (quick launch / orchestrated), deps that genuinely need work (absent markers, an absent required package, a fingerprint drift, or Metro unable to resolve a module) return a teaching block naming the orchestrator deps/prepare phase (plus the standalone `yarn setup:expo` / `launch --build` remedies) instead of an implicit setup — deps are the orchestrator's contract. The standalone/full tier still installs.
23
+ - **Deps baseline recorded the instant a setup leaf succeeds** — `prepareMobile` records the deps fingerprint baseline immediately after any successful `yarn-setup`, unconditionally (previously only a later `record: true` re-decide did, gated on the original decision being `install` and on `RECIPE_UP_INSTALL_ATTEMPTED`). Genuine-drift detection is fingerprint-based and self-sustaining after one good install.
24
+ - **CocoaPods no longer crashes on non-UTF-8 podspecs under a locale-less env** — pod-triggering spawns (`yarn setup` / native `yarn start:*`) now pin a UTF-8 `LANG`/`LC_ALL` when the caller has none. A gateway/tmux parent env without a locale made Ruby parse podspecs as US-ASCII and die on the first multibyte byte (`Invalid react-native-mmkv.podspec: "\xE2" on US-ASCII`).
25
+ - **`launch` no longer reports a dead app as ready** — the mobile readiness decision treated a healthy Metro + successful bundle as `ready` without checking the target device, so a dev client that had crashed or was never opened produced `✓ ready` while nothing was running. The decision now probes the device (`xcrun simctl spawn … launchctl list` / `adb shell ps -A`, fail-closed) and returns `launch`/`app-not-running` (relaunch the client) unless the app is actually alive.
26
+ - **Metro log-tail window no longer leaks into an unrelated tmux session** — running `start-metro` outside a tmux client fell back to `tmux display-message` (the last-attached session, e.g. the user's own) and created its `metro-<port>` window — with a `tail -F` — there, outliving the run. The viewer now opens a window only in a session that belongs to this run (a named slot session that exists, or the current session when actually inside tmux) and otherwise skips it; Metro still runs detached to the log. Extracted to `adapters/mobile/lib/tmux-viewer.sh` so the rule is unit-tested; contract tests stub tmux hermetically so no suite run can create a real window.
27
+
28
+ ### Added
29
+ - **`mm-harness stop`** — stop the port-scoped Metro dev server this mobile checkout owns and close its log-tail window (idempotent; extension/core are slot-managed and get a teaching error). Slot-context-aware port resolution.
30
+ - **Slot-context device/port detection** — `resolve-farmslot-ports.sh` reads `agentic-runtime.json` (the mapping the orchestrator's prepare wrote into the checkout) first, so device/port resolution survives pool renames that break any naming formula.
31
+ - **CLI help surface** — grouped/colored `mm-harness` help, a `DEV OVERRIDE ACTIVE` banner when `MM_HARNESS_BIN` is set, and a `SLOT` line when the checkout is a prepared slot; `mmdev-harness` completion alias.
32
+
33
+ ### Changed
34
+ - **Deps install uses a scoped `setup:expo` variant when native isn't wanted** — `adapters/mobile/yarn-setup.sh` gains `--mode full|expo`. `expo` runs `yarn setup:expo --no-build-ios --no-build-android` (JS deps + prebuild, no native — native builds come from `open-device.sh`, never doubled); the full `yarn setup` runs only for a native build (`launch --build`). This is also the standalone Metro-only remedy taught by the fast-tier deps block.
35
+ - **`mm-harness runtime-decision --adapter mobile` accepts `--preflight-mode`** — surfaces the fast (presence, blocks on a genuine gap) vs full (installs) decision for inspection and callers.
36
+
3
37
  ## 0.3.8 - 2026-07-04
4
38
 
5
39
  ### Fixed
6
- - **Mobile launch streams progress in mode** — mobile prepare leaves now run through , which tees the child's stdout+stderr to the parent's stderr live while stdout stays reserved for the JSON envelope. A multi-minute {
7
- "schemaVersion": 1,
8
- "command": "launch",
9
- "status": "fail",
10
- "exitCode": 2,
11
- "error": {
12
- "code": "USAGE",
13
- "message": "could not detect the MetaMask repo type for /private/tmp/claude-501/-Users-deeeed-dev-metamask-principal-pitch-2025/4645d206-3e7f-482d-a5c1-2274f26a7411/scratchpad/emh-review",
14
- "userAction": "cd into a MetaMask checkout or pass --target <path>, or force it with --adapter <mobile|extension|core>"
15
- }
16
- } is no longer silent (previously buffered leaf output and suppressed it entirely under ). Contract test asserts leaf progress reaches stderr, the envelope stays clean, and the streaming conversion preserves the exit-code contract.
40
+ - **Mobile launch streams progress in `--json` mode** — mobile prepare leaves now run through `spawnScriptStreaming`, which tees the child's stdout+stderr to the parent's stderr live while stdout stays reserved for the JSON envelope. A multi-minute `mm-harness launch --json` is no longer silent (previously `spawnSync` buffered leaf output and suppressed it entirely under `--json`). Contract test asserts leaf progress reaches stderr, the envelope stays clean, and the streaming conversion preserves the exit-code contract.
17
41
 
18
42
  ## 0.3.7 - 2026-07-04
19
43
 
@@ -243,7 +243,14 @@ printf '%s\n' "$started_pid" > "$watch_pid_file"
243
243
  tmux_session="${RECIPE_TMUX_SESSION:-}"
244
244
  if [ -z "$tmux_session" ]; then
245
245
  slot="$(slot_number || true)"
246
- [ -z "$slot" ] || tmux_session="mme-$slot"
246
+ if [ -n "$slot" ] && tmux has-session -t "=mme-$slot" 2>/dev/null; then
247
+ tmux_session="mme-$slot"
248
+ fi
249
+ fi
250
+ # Fall back to the session this run is inside: renamed pools and manual runs in
251
+ # any tmux still get their watcher window; outside tmux there is nowhere to put one.
252
+ if [ -z "$tmux_session" ]; then
253
+ tmux_session="$(tmux display-message -p '#S' 2>/dev/null || true)"
247
254
  fi
248
255
  if [ -n "$tmux_session" ] && command -v tmux >/dev/null 2>&1 && tmux has-session -t "$tmux_session" 2>/dev/null; then
249
256
  tmux_window="webpack-${WATCHER_PORT:-default}"
@@ -18,6 +18,14 @@
18
18
  "inputs": "--target --port --clear; env WATCHER_PORT, METRO_PORT, MOBILE_METRO_REQUIRED_ENV, METRO_MAX_WORKERS",
19
19
  "outputs": "Metro PID file + log; exit 0 ready / 1 timeout / 2 bad args"
20
20
  },
21
+ {
22
+ "id": "mobile/stop-metro",
23
+ "entry": "adapters/mobile/stop-metro.sh",
24
+ "kind": "bash",
25
+ "purpose": "Stop the port-scoped Metro bundler this checkout owns and close its log-tail window (idempotent).",
26
+ "inputs": "--target --port; env WATCHER_PORT",
27
+ "outputs": "progress on stderr; exit 0 stopped-or-nothing / 1 stop failed / 2 bad args"
28
+ },
21
29
  {
22
30
  "id": "mobile/prewarm-bundle",
23
31
  "entry": "adapters/mobile/prewarm-bundle.sh",
@@ -0,0 +1,38 @@
1
+ #!/usr/bin/env bash
2
+ # tmux-viewer — open a read-only tmux window tailing the Metro log, but ONLY in a
3
+ # session that belongs to this run. Sourced by start-metro.sh; kept separate so
4
+ # the session-resolution rule (the part that must never leak into a user session)
5
+ # is unit-testable in isolation.
6
+ #
7
+ # Contract: never create a window in "whatever session happens to be attached".
8
+ # Outside a tmux client `tmux display-message -p '#S'` returns the last-attached
9
+ # session (e.g. the user's own), which would land the window — and its `tail -F` —
10
+ # in an unrelated session that outlives the run. When no run-owned session can be
11
+ # resolved, skip the viewer entirely (Metro still runs detached to the log).
12
+ #
13
+ # Inputs (env/args): $1 = metro log path; PORT, LOG_DIR, and optional `slot` from
14
+ # the caller; TMUX indicates we are inside a tmux client.
15
+
16
+ # shellcheck disable=SC2329 # sourced by start-metro.sh; also sourced by tests.
17
+ start_viewer_window() {
18
+ local metro_log="$1"
19
+ command -v tmux >/dev/null 2>&1 || return 0
20
+ local window="metro-${PORT}"
21
+ local metro_tmux="$LOG_DIR/metro.tmux"
22
+ local session=""
23
+ # (1) the slot session, when one is named and actually exists. The mm-<slot>
24
+ # formula predates the mmdev rename; a session named otherwise simply falls
25
+ # through — has-session gates it, so a wrong name never creates a window.
26
+ if [ -n "${slot:-}" ] && tmux has-session -t "=mm-${slot}" 2>/dev/null; then
27
+ session="mm-${slot}"
28
+ # (2) the current session, but ONLY when we are actually inside a tmux client.
29
+ elif [ -n "${TMUX:-}" ]; then
30
+ session="$(tmux display-message -p '#S' 2>/dev/null || true)"
31
+ fi
32
+ # No run-owned session → skip the viewer; no orphaned tail-window is left behind.
33
+ { [ -n "$session" ] && tmux has-session -t "=$session" 2>/dev/null; } || return 0
34
+ tmux kill-window -t "${session}:${window}" >/dev/null 2>&1 || true
35
+ tmux new-window -d -t "$session" -n "$window" "exec tail -n +1 -F $(printf '%q' "$metro_log")"
36
+ printf '%s:%s\n' "$session" "$window" > "$metro_tmux"
37
+ printf 'Metro log viewer → tmux %s:%s\n' "$session" "$window" >&2
38
+ }
@@ -72,24 +72,10 @@ default_metro_workers() {
72
72
  fi
73
73
  }
74
74
 
75
- start_viewer_window() {
76
- local metro_log="$1"
77
- command -v tmux >/dev/null 2>&1 || return 0
78
- local window="metro-${PORT}"
79
- local metro_tmux="$LOG_DIR/metro.tmux"
80
- local session=""
81
- if [ -n "${slot:-}" ] && tmux has-session -t "=mm-${slot}" 2>/dev/null; then
82
- session="mm-${slot}"
83
- fi
84
- if [ -z "$session" ]; then
85
- session="$(tmux display-message -p '#S' 2>/dev/null || true)"
86
- fi
87
- [ -n "$session" ] || return 0
88
- tmux kill-window -t "${session}:${window}" >/dev/null 2>&1 || true
89
- tmux new-window -d -t "$session" -n "$window" "exec tail -n +1 -F $(printf '%q' "$metro_log")"
90
- printf '%s:%s\n' "$session" "$window" > "$metro_tmux"
91
- printf 'Metro log viewer → tmux %s:%s\n' "$session" "$window" >&2
92
- }
75
+ # Log-tail viewer window (session-safe): sourced so its run-owned-session rule is
76
+ # unit-testable and can never leak a window into an unrelated user session.
77
+ # shellcheck disable=SC1091
78
+ . "$SCRIPT_DIR/lib/tmux-viewer.sh"
93
79
 
94
80
  # --- main ---------------------------------------------------------------------
95
81
 
@@ -0,0 +1,66 @@
1
+ #!/usr/bin/env bash
2
+ # stop-metro.sh — stop the Metro dev server this checkout owns and close its
3
+ # log-tail window. Port-scoped like start-metro: only the listener on the
4
+ # resolved port is signalled, so concurrent slots on other ports are untouched.
5
+ # Idempotent: nothing running is a success, not an error.
6
+ set -euo pipefail
7
+
8
+ TARGET="."
9
+ PORT="${WATCHER_PORT:-}"
10
+ while [ $# -gt 0 ]; do
11
+ case "$1" in
12
+ --target) TARGET="$2"; shift 2 ;;
13
+ --port) PORT="$2"; shift 2 ;;
14
+ -h|--help)
15
+ printf 'Usage: stop-metro.sh [--target <dir>] [--port <port>]\n'
16
+ printf ' --target MetaMask Mobile checkout directory\n'
17
+ printf ' --port Metro port (default: WATCHER_PORT env, else the slot context, else 8081)\n'
18
+ exit 0
19
+ ;;
20
+ *) printf 'stop-metro: unknown arg: %s\n' "$1" >&2; exit 2 ;;
21
+ esac
22
+ done
23
+
24
+ TARGET="$(cd "$TARGET" && pwd)"
25
+
26
+ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
27
+ # shellcheck disable=SC1091
28
+ . "$SCRIPT_DIR/../shared/harness-path.sh"
29
+ if ! command -v recipe_runtime_dir >/dev/null 2>&1; then
30
+ echo "stop-metro: shared lib adapters/shared/harness-path.sh not found; reinstall the runner." >&2
31
+ exit 1
32
+ fi
33
+
34
+ # No explicit port: resolve the checkout's own (slot context first, then the
35
+ # pool/formula fallbacks) so we stop THIS slot's Metro, not the 8081 default.
36
+ if [ -z "$PORT" ]; then
37
+ # shellcheck disable=SC1091
38
+ . "$SCRIPT_DIR/../shared/resolve-farmslot-ports.sh"
39
+ resolved_port="$(resolve_mobile_runtime_ports "$TARGET" 2>/dev/null | sed -n 's/^WATCHER_PORT=//p' | head -1 || true)"
40
+ PORT="${resolved_port:-8081}"
41
+ fi
42
+
43
+ LOG_DIR="$TARGET/$(recipe_runtime_dir)"
44
+ PID_FILE="$LOG_DIR/metro.pid"
45
+ TMUX_FILE="$LOG_DIR/metro.tmux"
46
+
47
+ # shellcheck disable=SC1091
48
+ . "$SCRIPT_DIR/lib/metro-listener.sh"
49
+
50
+ pids="$(metro_listener_pids 2>/dev/null || true)"
51
+ if [ -n "$pids" ]; then
52
+ stop_metro_listener || exit 1
53
+ printf 'Stopped Metro on port %s (pid %s)\n' "$PORT" "$pids" >&2
54
+ else
55
+ printf 'Metro not running on port %s — nothing to stop\n' "$PORT" >&2
56
+ fi
57
+ rm -f "$PID_FILE"
58
+
59
+ # Close the read-only log-tail window start-metro opened, if it is still there.
60
+ if [ -f "$TMUX_FILE" ] && command -v tmux >/dev/null 2>&1; then
61
+ win="$(cat "$TMUX_FILE" 2>/dev/null || true)"
62
+ if [ -n "$win" ] && tmux kill-window -t "$win" 2>/dev/null; then
63
+ printf 'Closed Metro log window %s\n' "$win" >&2
64
+ fi
65
+ fi
66
+ rm -f "$TMUX_FILE"
@@ -2,19 +2,31 @@
2
2
  # yarn-setup — install node_modules for MetaMask Mobile.
3
3
  #
4
4
  # Inputs: --target <metamask-mobile dir> (default $PWD)
5
+ # --mode full|expo (default full)
5
6
  # Outputs: yarn install output on stderr; exit 0 pass, 1 install failed, 2 bad args.
6
7
  #
7
8
  # Single device op: install node_modules. No other sequencing.
9
+ # full — `yarn setup` (JS deps + expo prebuild + native pods/jetify). Use only
10
+ # when a native build is wanted (launch --build / rebuild-native).
11
+ # expo — `yarn setup:expo --no-build-ios --no-build-android` (JS deps + prebuild,
12
+ # no native). The Metro-only remedy: native builds come from open-device.
8
13
  set -euo pipefail
9
14
 
10
15
  TARGET="$PWD"
16
+ MODE="full"
11
17
  while [ "$#" -gt 0 ]; do
12
18
  case "$1" in
13
19
  --target) TARGET="$2"; shift 2 ;;
14
- -h|--help) printf 'Usage: yarn-setup.sh [--target <dir>]\n'; exit 0 ;;
20
+ --mode) MODE="$2"; shift 2 ;;
21
+ -h|--help) printf 'Usage: yarn-setup.sh [--target <dir>] [--mode full|expo]\n'; exit 0 ;;
15
22
  *) printf 'yarn-setup: unknown arg: %s\n' "$1" >&2; exit 2 ;;
16
23
  esac
17
24
  done
18
25
 
19
26
  TARGET="$(cd "$TARGET" && pwd)"
20
- cd "$TARGET" && exec yarn setup
27
+ cd "$TARGET"
28
+ case "$MODE" in
29
+ full) exec yarn setup ;;
30
+ expo) exec yarn setup:expo --no-build-ios --no-build-android ;;
31
+ *) printf 'yarn-setup: unknown --mode: %s (want full|expo)\n' "$MODE" >&2; exit 2 ;;
32
+ esac
@@ -155,8 +155,29 @@ resolve_mobile_slot_defaults() {
155
155
  printf 'SLOT_ID=local-mobile-%s\n' "$n"
156
156
  }
157
157
 
158
+ # Slot context the orchestrator's prepare wrote INTO the checkout — the
159
+ # authoritative device/port mapping for this repo (survives pool renames that
160
+ # break any naming formula). Highest-priority detection source.
161
+ resolve_mobile_runtime_context() {
162
+ local repo="$1" ctx
163
+ ctx="$repo/${RECIPE_RUNTIME_DIR:-temp/recipe/runtime}/agentic-runtime.json"
164
+ [ -f "$ctx" ] || return 1
165
+ node -e '
166
+ try {
167
+ const c = JSON.parse(require("fs").readFileSync(process.argv[1], "utf8"));
168
+ if (!c.simulator && !c.metroPort) process.exit(1);
169
+ if (c.metroPort) console.log(`WATCHER_PORT=${c.metroPort}`);
170
+ if (c.simulator) console.log(`IOS_SIMULATOR=${c.simulator}`);
171
+ if (c.slotId) console.log(`SLOT_ID=${c.slotId}`);
172
+ } catch { process.exit(1); }
173
+ ' "$ctx" 2>/dev/null
174
+ }
175
+
158
176
  resolve_mobile_runtime_ports() {
159
177
  local repo="${1:-.}"
178
+ if resolve_mobile_runtime_context "$repo"; then
179
+ return 0
180
+ fi
160
181
  if resolve_farmslot_ports_by_repo "$repo"; then
161
182
  return 0
162
183
  fi
@@ -0,0 +1,61 @@
1
+ # UX principles — the mm-harness interaction contract
2
+
3
+ Every command surface follows these rules. They are review criteria, not aspirations:
4
+ a PR that adds a command or output path is checked against each one.
5
+
6
+ **Scope: the HUMAN surface.** The agent contract is exempt and inviolable: under
7
+ `--json`, stdout carries exactly one machine-readable envelope — no color, no
8
+ banners, no decorative lines, schema per CLI-SPEC.md. Progress/streaming exists
9
+ for agents too, but only ever on stderr. When a rule below conflicts with envelope
10
+ purity, the envelope wins.
11
+
12
+ ## 1. Context-aware by default
13
+ The CLI answers relative to the checkout it is standing in.
14
+ - Adapter is auto-detected everywhere (`--adapter` overrides; a bare command never
15
+ fails on a value detection can supply).
16
+ - The slot context the orchestrator wrote into the checkout
17
+ (`<runtime_dir>/agentic-runtime.json`) is the authoritative source for device,
18
+ ports, and slot identity — naming formulas are last-resort fallbacks and never
19
+ hardcode pool prefixes (they go stale on rename).
20
+ - Output is scoped to the platform: an extension-only field is noise on a mobile
21
+ slot; help, flag lists, and completion candidates shrink to what applies here.
22
+
23
+ ## 2. Never silent, never opaque
24
+ Long operations show intent immediately and progress continuously.
25
+ - First line within milliseconds: what is being done, where (device/port), and
26
+ where the full log lives.
27
+ - Leaf output streams live to stderr — `--json` reserves stdout for the envelope,
28
+ it never buys silence. Quiet stretches get heartbeats with elapsed time.
29
+ - Each dev server a command starts gets its tmux log-tail window at launch time
30
+ (window creation is launch-coupled; users re-create separately when they close one).
31
+
32
+ ## 3. Summaries state outcomes
33
+ The final line says what actually happened: target device, what the tier meant,
34
+ resulting state — never a bare status word. Dev-override runs are marked
35
+ (`[dev: MM_HARNESS_BIN]`) so it is always evident which bin served.
36
+
37
+ ## 4. Dev and prod are explicit
38
+ The global install is prod (`mm-harness`); a dev checkout runs via the documented
39
+ `MM_HARNESS_BIN` switch (conventionally aliased `mmdev-harness`). The dev override
40
+ announces itself in help and summaries. Completions cover both names.
41
+
42
+ ## 5. Errors teach the exact escape
43
+ Every failure states the next command for the caller's actual situation. A wrong
44
+ hint (an escape for a different failure mode) is a bug of the same severity as
45
+ the failure itself.
46
+
47
+ ## 6. Lifecycle verbs are symmetric and idempotent
48
+ What a command starts, a sibling verb stops (`launch`/`stop`), scoped to what this
49
+ checkout owns (port-scoped, session-scoped) — never "whatever is running".
50
+ Nothing-to-do is success. Tests never leak into the operator's real tmux.
51
+
52
+ ## 7. Color is consistent and optional
53
+ All human output styles through `cli-color.ts` (headers, ok/warn/err states,
54
+ provenance, "Next:" hints); `NO_COLOR`/non-TTY degrade to plain text. No surface
55
+ invents its own ANSI.
56
+
57
+ ## 8. Doctor is the status surface
58
+ `doctor` reports, per platform: live runtime state (deps currency, dev-server
59
+ up/stale on the slot's resolved port, app/browser presence), the scoped runtime
60
+ context with provenance (`[file]`/`[env]`/`[default]`), and the exact customize
61
+ hint per row. Read-only unless `--fix`.
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@deeeed/metamask-harness",
3
- "version": "0.3.8",
3
+ "version": "0.3.9",
4
4
  "type": "module",
5
5
  "bin": {
6
6
  "mm-harness": "bin/mm-harness"
@@ -17,8 +17,13 @@
17
17
  "@farmslot/protocol": "^0.7.2",
18
18
  "@farmslot/recipe-harness": "^0.3.3",
19
19
  "commander": "^12.0.0",
20
- "tsx": "^4.19.0",
21
- "viem": "^2.52.2"
20
+ "tsx": "^4.23.0",
21
+ "viem": "^2.54.3"
22
+ },
23
+ "resolutions": {
24
+ "esbuild": "0.28.1",
25
+ "undici": "6.27.0",
26
+ "ws": "8.21.0"
22
27
  },
23
28
  "devDependencies": {
24
29
  "@types/node": "^22.0.0",
@@ -29,12 +29,13 @@ _mmh_bin() {
29
29
  return 1
30
30
  }
31
31
 
32
- _mmh_commands="launch logs debug fixtures actions call flows run doctor install verify cleanup completions"
32
+ _mmh_commands="launch stop logs debug fixtures actions call flows run doctor install verify cleanup completions"
33
33
 
34
34
  # Per-command flags (static, from the mm-harness surface).
35
35
  _mmh_flags_for() {
36
36
  case "$1" in
37
37
  launch) printf '%s' "--build --watch --verify --sidepanel --fullscreen --runway --device --cdp-port --watcher-port --heal --adapter --target --json" ;;
38
+ stop) printf '%s' "--port --target --json" ;;
38
39
  logs) printf '%s' "--full --events --source --adapter --target --json" ;;
39
40
  debug) printf '%s' "--worker --dev-menu --adapter --target --json" ;;
40
41
  fixtures) printf '%s' "--fixture --adapter --target --json" ;;
@@ -83,7 +84,9 @@ if [ -n "${BASH_VERSION:-}" ]; then
83
84
  *) COMPREPLY=($(compgen -W "$(_mmh_flags_for "$cmd")" -- "$cur")) ;;
84
85
  esac
85
86
  }
87
+ # mmdev-harness is the documented dev-checkout alias; complete it identically.
86
88
  complete -F _mmh_complete_bash mm-harness 2>/dev/null || true
89
+ complete -F _mmh_complete_bash mmdev-harness 2>/dev/null || true
87
90
  fi
88
91
 
89
92
  # ── Zsh ─────────────────────────────────────────────────────────────────
@@ -117,9 +120,9 @@ if [ -n "${ZSH_VERSION:-}" ]; then
117
120
  }
118
121
 
119
122
  if (( $+functions[compdef] )); then
120
- compdef _mmh_complete_zsh mm-harness
123
+ compdef _mmh_complete_zsh mm-harness mmdev-harness
121
124
  else
122
125
  autoload -Uz compinit 2>/dev/null && compinit -u 2>/dev/null
123
- (( $+functions[compdef] )) && compdef _mmh_complete_zsh mm-harness
126
+ (( $+functions[compdef] )) && compdef _mmh_complete_zsh mm-harness mmdev-harness
124
127
  fi
125
128
  fi
@@ -7,6 +7,8 @@
7
7
 
8
8
  import path from 'node:path';
9
9
 
10
+ import { recordDepsBaseline } from '@farmslot/recipe-harness/runtime/deps-readiness';
11
+
10
12
  import { EXIT, spawnScriptStreaming } from '../../commands/shared.ts';
11
13
  import type { ScriptResult } from '../../commands/shared.ts';
12
14
  import { runnerDir } from '../../paths.ts';
@@ -24,7 +26,16 @@ export { type MobileRuntimeDecisionReport };
24
26
  // An inherited FORCE_COLOR makes node emit a colorized `undefined`, so VisionCamera
25
27
  // misdetects the worklets pod, enables FrameProcessors, and fails on the missing
26
28
  // pod. Force plain output for pod-triggering spawns so the probe reads `undefined`.
27
- const POD_PROBE_ENV: Record<string, string> = { FORCE_COLOR: '0', NO_COLOR: '1' };
29
+ // CocoaPods reads UTF-8 podspecs; a parent env without a locale (gateway/tmux
30
+ // spawns) makes Ruby parse them as US-ASCII and die on the first multibyte byte
31
+ // ('"\xE2" on US-ASCII' in react-native-mmkv.podspec). Pin a UTF-8 locale for
32
+ // pod-triggering spawns when the caller has none.
33
+ const POD_PROBE_ENV: Record<string, string> = {
34
+ FORCE_COLOR: '0',
35
+ NO_COLOR: '1',
36
+ LANG: process.env.LANG?.includes('UTF-8') ? process.env.LANG : 'en_US.UTF-8',
37
+ LC_ALL: process.env.LC_ALL?.includes('UTF-8') ? process.env.LC_ALL : 'en_US.UTF-8',
38
+ };
28
39
 
29
40
  export interface PrepareMobileOptions extends MobileRuntimeDecisionOptions {
30
41
  /**
@@ -59,6 +70,7 @@ export async function mobileRuntimeStatus(
59
70
  metroLog: opts.metroLog,
60
71
  platform: opts.platform,
61
72
  record: opts.record,
73
+ preflightMode: opts.preflightMode,
62
74
  });
63
75
  }
64
76
 
@@ -81,11 +93,13 @@ export async function prepareMobile(
81
93
  metroLog: opts.metroLog,
82
94
  platform,
83
95
  record: opts.record,
96
+ preflightMode,
84
97
  });
85
98
 
86
99
  if (report.decision === 'blocked') {
87
100
  const reasons = report.reasons.join(' ');
88
- const msg = `mobile prepare blocked: ${reasons}\n Next: fix the bundle error in app code before retrying recipe up.`;
101
+ const next = report.userAction ?? 'fix the bundle error in app code before retrying recipe up.';
102
+ const msg = `mobile prepare blocked: ${reasons}\n Next: ${next}`;
89
103
  if (!json) process.stderr.write(`${msg}\n`);
90
104
  return { status: EXIT.runtime, output: msg };
91
105
  }
@@ -112,6 +126,11 @@ export async function prepareMobile(
112
126
  for (const action of actions) {
113
127
  const result = await dispatchAction(action, target, platform, json, preflightMode);
114
128
  if (result.status !== 0) return result;
129
+ // Record the deps baseline the instant a setup/install leaf succeeds, so the
130
+ // mtime freshness fallback stops firing forever after one good install. This
131
+ // is unconditional (not gated on a later re-decide or an env flag), so the
132
+ // baseline lifecycle is self-sustaining regardless of the caller's path.
133
+ if (action.id === 'yarn-setup') recordDepsBaseline(path.resolve(target));
115
134
  }
116
135
 
117
136
  // After install: re-decide and run appropriate actions.
@@ -122,7 +141,7 @@ export async function prepareMobile(
122
141
  watcherPort: opts.watcherPort,
123
142
  metroLog: opts.metroLog,
124
143
  platform,
125
- record: true,
144
+ preflightMode,
126
145
  });
127
146
  switch (postInstall.decision) {
128
147
  case 'install': {
@@ -1,3 +1,4 @@
1
+ import { execFileSync } from 'node:child_process';
1
2
  import fs from 'node:fs';
2
3
  import path from 'node:path';
3
4
 
@@ -43,6 +44,9 @@ export interface MobileRuntimeDecisionReport {
43
44
  decision: MobileReadinessDecision;
44
45
  reasonCode: string;
45
46
  reasons: string[];
47
+ // A single teaching escape for the caller's actual situation, set on decisions
48
+ // the runner deliberately refuses to auto-resolve (e.g. fast-mode deps gaps).
49
+ userAction?: string;
46
50
  checks: {
47
51
  deps: DepsCheck;
48
52
  metroLog: MetroLogCheck;
@@ -56,6 +60,9 @@ export interface MobileRuntimeDecisionOptions {
56
60
  metroLog?: string;
57
61
  platform?: string;
58
62
  record?: boolean;
63
+ // 'fast' (quick launch / no --build) makes deps readiness a presence check only:
64
+ // freshness is the orchestrator's deps-phase contract, never re-decided here.
65
+ preflightMode?: string;
59
66
  }
60
67
 
61
68
  const BUNDLE_ERR = /Bundling failed|Unable to resolve /u;
@@ -128,38 +135,38 @@ function missingRequiredDeps(target: string): string[] {
128
135
  }
129
136
  }
130
137
 
131
- // Returns true when any manifest input (package.json or yarn.lock) is newer than
132
- // the install markers written by yarn on a successful install. A reliable signal
133
- // that deps changed since the last install, even when depsCheck found no recorded
134
- // baseline drift (e.g. git checkout preserved an old author-date mtime on the
135
- // manifest, then a new dep was committed, making the manifest newer than markers).
136
- function depsStaleByMtime(target: string): boolean {
137
- // Use the same install markers as depsCheck: the files yarn writes on install.
138
- const INSTALL_MARKERS = ['node_modules/.yarn-state.yml', '.yarn/install-state.gz'];
139
- let markerMtime = 0;
140
- for (const rel of INSTALL_MARKERS) {
141
- try {
142
- const mt = fs.statSync(path.join(target, rel)).mtimeMs;
143
- if (mt > markerMtime) markerMtime = mt;
144
- } catch {
145
- // marker absent; keep searching
146
- }
147
- }
148
- if (markerMtime === 0) return false; // no markers → depsCheck already handles via 'missing'
149
- for (const file of ['package.json', 'yarn.lock']) {
150
- try {
151
- if (fs.statSync(path.join(target, file)).mtimeMs > markerMtime) return true;
152
- } catch {
153
- // manifest absent; skip
154
- }
155
- }
156
- return false;
157
- }
158
-
159
138
  const installActions = (target: string): MobileRuntimeDecisionAction[] => [
160
139
  { id: 'yarn-setup', argv: ['yarn', 'setup'], cwd: target },
161
140
  ];
162
141
 
142
+ // True when a MetaMask dev client process is alive on the target device
143
+ // (IOS_SIMULATOR / ADB_SERIAL, falling back to the booted simulator). Probe
144
+ // failures count as not-running: the worst case is an idempotent relaunch,
145
+ // while failing open would declare a dead app "ready".
146
+ function appRunningOnDevice(platform?: string): boolean {
147
+ try {
148
+ if (platform === 'android') {
149
+ const serial = process.env.ADB_SERIAL || process.env.ANDROID_SERIAL;
150
+ const args = serial ? ['-s', serial] : [];
151
+ const out = execFileSync('adb', [...args, 'shell', 'ps', '-A'], {
152
+ encoding: 'utf8',
153
+ stdio: ['ignore', 'pipe', 'ignore'],
154
+ timeout: 5_000,
155
+ });
156
+ return out.includes('io.metamask');
157
+ }
158
+ const device = process.env.IOS_SIMULATOR || 'booted';
159
+ const out = execFileSync('xcrun', ['simctl', 'spawn', device, 'launchctl', 'list'], {
160
+ encoding: 'utf8',
161
+ stdio: ['ignore', 'pipe', 'ignore'],
162
+ timeout: 5_000,
163
+ });
164
+ return out.toLowerCase().includes('io.metamask');
165
+ } catch {
166
+ return false;
167
+ }
168
+ }
169
+
163
170
  const launchActions = (target: string, clearMetro = false): MobileRuntimeDecisionAction[] => {
164
171
  const actions: MobileRuntimeDecisionAction[] = [];
165
172
  // start-metro ensures Metro is running; --clear resets the bundle cache.
@@ -184,22 +191,60 @@ export async function decideMobileReadiness(
184
191
  const resolved = path.resolve(target);
185
192
  if (options.record) recordDepsBaseline(resolved);
186
193
 
194
+ const fast = options.preflightMode === 'fast';
195
+ const report = await computeMobileReadiness(resolved, options, fast);
196
+
197
+ // mtime churn is already neutralized above, so a fast-mode 'install' verdict
198
+ // means deps genuinely need work (absent markers, an absent required package, a
199
+ // fingerprint-baseline drift, or Metro unable to resolve a module). In an
200
+ // orchestrated run the deps phase owns installation; having the runner install
201
+ // mid-launch is inverted authority. Surface a teaching block, not an implicit
202
+ // setup, so a runway preflight never re-does the orchestrator's deps work.
203
+ if (fast && report.decision === 'install') {
204
+ return {
205
+ ...report,
206
+ decision: 'blocked',
207
+ reasonCode: 'deps-not-ready',
208
+ reasons: [
209
+ 'Fast preflight found dependencies not ready and does not install them (the orchestrator deps phase owns installation).',
210
+ ...report.reasons,
211
+ ],
212
+ userAction:
213
+ 'run the slot deps/prepare phase; standalone: `yarn setup:expo --no-build-ios --no-build-android` in the checkout, or `mm-harness launch <platform> --build` to install and build',
214
+ actions: [],
215
+ };
216
+ }
217
+ return report;
218
+ }
219
+
220
+ async function computeMobileReadiness(
221
+ resolved: string,
222
+ options: MobileRuntimeDecisionOptions,
223
+ fast: boolean,
224
+ ): Promise<MobileRuntimeDecisionReport> {
187
225
  const rawDeps = depsCheck(resolved, {
188
226
  productMarkers: mobileProductMarkers(options.platform),
189
227
  });
190
- // Mtime-based fallback: when depsCheck found no recorded baseline, supplement
191
- // with a direct mtime comparison if package.json or yarn.lock is newer than
192
- // node_modules, a dep change landed after the last install. Fail fast at
193
- // pre-flight rather than launching a doomed Metro bundle. Skip when a baseline
194
- // was found (hasBaseline=true) so fingerprint-verified 'current' is trusted.
195
- let deps: DepsCheck =
196
- rawDeps.status === 'current' && !rawDeps.hasBaseline && depsStaleByMtime(resolved)
197
- ? { installed: rawDeps.installed, status: 'stale', hasBaseline: rawDeps.hasBaseline }
198
- : rawDeps;
199
- // Required-dep absence: even when fingerprint + mtime reports 'current', a
200
- // top-level dependency could be absent from node_modules (e.g. added in a
201
- // branch merge but install not re-run). Catch at pre-flight before Metro
202
- // discovers it during a 249s bundle run.
228
+ // Presence is authoritative; mtime is only a hint. Without a recorded baseline
229
+ // the only way depsCheck reports 'stale' is mtime drift (a manifest newer than
230
+ // the install markers). In a managed checkout the orchestrator's git phase
231
+ // refreshes tracked-file mtimes on every sync while its deps phase leaves
232
+ // node_modules untouched, so "manifest newer than markers" is normal and NOT
233
+ // proof of drift — treating it as stale re-ran the full yarn setup inside every
234
+ // launch. Trust the installed node_modules and warn; genuine drift is caught by
235
+ // the recorded-baseline fingerprint when one exists (hasBaseline=true stays 'stale').
236
+ let deps: DepsCheck = rawDeps;
237
+ if (rawDeps.status === 'stale' && !rawDeps.hasBaseline) {
238
+ deps = { installed: rawDeps.installed, status: 'current', hasBaseline: false };
239
+ process.stderr.write(
240
+ '[runtime-decision] manifest is newer than install markers (no recorded baseline); ' +
241
+ 'trusting installed node_modules — run yarn setup manually if deps truly changed.\n',
242
+ );
243
+ }
244
+ // Required-dep absence is a presence check (both modes): a top-level dependency
245
+ // can be absent from node_modules even when markers + fingerprint report
246
+ // 'current' (e.g. added in a branch merge but install not re-run). Catch at
247
+ // pre-flight before Metro discovers it during a long bundle run.
203
248
  if (deps.status === 'current') {
204
249
  const absent = missingRequiredDeps(resolved);
205
250
  if (absent.length > 0) {
@@ -373,13 +418,29 @@ export async function decideMobileReadiness(
373
418
  }
374
419
 
375
420
  if (metro.status === 'up' && metroLog.status === 'ok') {
421
+ // Metro health alone is not runtime readiness: the dev client may be
422
+ // installed but not running on the target device (launch must launch).
423
+ // Probe the device before declaring ready; a healthy Metro with a dead
424
+ // app relaunches the client without touching Metro.
425
+ if (!appRunningOnDevice(options.platform)) {
426
+ return {
427
+ schemaVersion: 1,
428
+ adapter: 'mobile',
429
+ target: resolved,
430
+ decision: 'launch',
431
+ reasonCode: 'app-not-running',
432
+ reasons: ['Metro is healthy but the dev client is not running on the target device; launching it.'],
433
+ checks,
434
+ actions: launchActions(resolved),
435
+ };
436
+ }
376
437
  return {
377
438
  schemaVersion: 1,
378
439
  adapter: 'mobile',
379
440
  target: resolved,
380
441
  decision: 'ready',
381
442
  reasonCode: 'healthy',
382
- reasons: ['Dependencies are current and Metro reports a successful bundle. Verify bridge before replay.'],
443
+ reasons: ['Dependencies are current, Metro reports a successful bundle, and the dev client is running. Verify bridge before replay.'],
383
444
  checks,
384
445
  actions: [],
385
446
  };
package/src/cli.ts CHANGED
@@ -4,6 +4,7 @@ import { mkdtemp } from 'node:fs/promises';
4
4
  import os from 'node:os';
5
5
  import path from 'node:path';
6
6
 
7
+ import { color } from './cli-color.ts';
7
8
  import { createDoctorReport, renderRuntimeContext } from './doctor.ts';
8
9
  import {
9
10
  invalidateCompletionCache,
@@ -420,7 +421,7 @@ async function runSelfTest(options: CliOptions) {
420
421
  }
421
422
 
422
423
  async function handleManifest({ options }: ParsedArgs): Promise<number> {
423
- const adapter = adapterOption(options);
424
+ const { adapter } = resolveAdapter(options);
424
425
  const actionManifestPath = actionManifestPathOption(options, adapter);
425
426
  const manifest = loadActionManifest(adapter, optionString(options, 'actionManifest'));
426
427
  await validateManifest(manifest);
@@ -430,7 +431,7 @@ async function handleManifest({ options }: ParsedArgs): Promise<number> {
430
431
  }
431
432
 
432
433
  async function handleActions({ options }: ParsedArgs): Promise<number> {
433
- const adapter = adapterOption(options);
434
+ const { adapter } = resolveAdapter(options);
434
435
  const manifest = loadActionManifest(adapter, optionString(options, 'actionManifest'));
435
436
  await validateManifest(manifest);
436
437
  const action = optionString(options, 'action');
@@ -475,9 +476,38 @@ async function handleDoctor({ options }: ParsedArgs): Promise<number> {
475
476
  }
476
477
 
477
478
  const result = createDoctorReport(adapter, target, manifestValidation, actionManifestPath);
478
- if (json) console.log(JSON.stringify(result, null, 2));
479
+ // Live runtime state, per platform: mobile reports the same read-only readiness
480
+ // the launch decision uses (deps currency, Metro on the slot's port, decision +
481
+ // reasons). Extension/core gain their sections when their readiness probes exist.
482
+ let runtime: { decision: string; reasonCode?: string; reasons: string[]; deps?: string; metro?: string } | undefined;
483
+ if (adapter === 'mobile') {
484
+ try {
485
+ const { mobileRuntimeStatus } = await import('./adapters/mobile/prepare.ts');
486
+ const { resolveMobileSlotPorts } = await import('./commands/launch.ts');
487
+ resolveMobileSlotPorts(target);
488
+ const watcherPort = process.env.WATCHER_PORT ? parseInt(process.env.WATCHER_PORT, 10) : undefined;
489
+ const report = await mobileRuntimeStatus(target, { watcherPort });
490
+ runtime = {
491
+ decision: report.decision,
492
+ reasonCode: report.reasonCode,
493
+ reasons: report.reasons,
494
+ deps: report.checks?.deps?.status,
495
+ metro: report.checks?.metro?.status,
496
+ };
497
+ } catch { /* readiness probe unavailable — doctor stays useful without it */ }
498
+ }
499
+ if (json) console.log(JSON.stringify({ ...result, runtime }, null, 2));
479
500
  else {
480
- console.log(`${result.status} ${adapter} ${result.compatibilityMode} manifest=${actionManifestPath}`);
501
+ const out = (style: string, text: string) => color(style, text, { stream: process.stdout });
502
+ const stateStyle = (value: string | undefined, good: string) => (value === good ? 'ok' : 'warn');
503
+ console.log(`${out(result.status === 'pass' ? 'ok' : 'err', result.status)} ${out('bold', adapter)} ${result.compatibilityMode} ${out('dim', `manifest=${actionManifestPath}`)}`);
504
+ if (runtime) {
505
+ const decisionStyle = runtime.decision === 'ready' ? 'ok' : runtime.decision === 'blocked' ? 'err' : 'warn';
506
+ console.log(
507
+ `${out('label', 'runtime:')} decision=${out(decisionStyle, runtime.decision)}${runtime.reasonCode ? ` ${out('dim', `(${runtime.reasonCode})`)}` : ''} deps=${out(stateStyle(runtime.deps, 'current'), runtime.deps ?? 'unknown')} metro=${out(stateStyle(runtime.metro, 'up'), runtime.metro ?? 'unprobed')}`,
508
+ );
509
+ for (const reason of runtime.reasons) console.log(` ${out('dim', reason)}`);
510
+ }
481
511
  console.log(renderRuntimeContext(result.runtimeContext));
482
512
  }
483
513
  return result.status === 'pass' ? 0 : 1;
@@ -758,6 +788,7 @@ async function handleRuntimeDecision({ options }: ParsedArgs): Promise<number> {
758
788
  metroLog: optionString(options, 'metroLog'),
759
789
  platform: optionString(options, 'platform') ?? process.env.PLATFORM ?? process.env.RECIPE_HARNESS_PLATFORM,
760
790
  record: optionFlag(options, 'record'),
791
+ preflightMode: optionString(options, 'preflightMode'),
761
792
  });
762
793
  if (optionFlag(options, 'json')) console.log(JSON.stringify(report, null, 2));
763
794
  else console.log(`${report.decision} ${report.reasonCode} — ${report.reasons[0] ?? ''}`);
@@ -1082,6 +1113,47 @@ function emitPlanUsageError(
1082
1113
  // hand it to runRecipe). One execution path, two doors: `call` = one node, `run` =
1083
1114
  // a graph. Inherits run semantics: always-validates (adapter-aware, exit 5), same
1084
1115
  // trace/evidence artifacts, same --json contract.
1116
+ async function handleStop(argv: string[]): Promise<number> {
1117
+ const { options } = parseArgs(argv, 'stop');
1118
+ const json = optionFlag(options, 'json');
1119
+ const { adapter, target } = resolveAdapter(options);
1120
+ if (adapter !== 'mobile') {
1121
+ return usageOut(
1122
+ json,
1123
+ 'stop',
1124
+ `stop owns the mobile Metro dev server; the ${adapter} dev server is slot-managed`,
1125
+ 'stop the extension watcher via its runtime pids: kill $(cat <runtime_dir>/webpack.pid); core runs no dev server',
1126
+ );
1127
+ }
1128
+ const leaf = path.join(runnerDir, 'adapters', 'mobile', 'stop-metro.sh');
1129
+ const args = ['--target', target];
1130
+ const port = optionString(options, 'port') ?? optionString(options, 'watcherPort');
1131
+ if (port) args.push('--port', port);
1132
+ const result = spawnSync('bash', [leaf, ...args], {
1133
+ stdio: json ? 'pipe' : 'inherit',
1134
+ encoding: 'utf8',
1135
+ });
1136
+ const exitCode = result.status ?? 1;
1137
+ if (json) {
1138
+ console.log(
1139
+ JSON.stringify(
1140
+ {
1141
+ schemaVersion: 1,
1142
+ command: 'stop',
1143
+ adapter,
1144
+ target,
1145
+ status: exitCode === 0 ? 'pass' : 'fail',
1146
+ exitCode,
1147
+ output: `${result.stdout ?? ''}${result.stderr ?? ''}`.trim(),
1148
+ },
1149
+ null,
1150
+ 2,
1151
+ ),
1152
+ );
1153
+ }
1154
+ return exitCode;
1155
+ }
1156
+
1085
1157
  async function handleCall(argv: string[]): Promise<number> {
1086
1158
  // Grammar: mm-harness call <action> [--arg k=v ...] [flags].
1087
1159
  // If the first token is a flag, the action positional is missing — parseCallArgs
@@ -1803,6 +1875,7 @@ export async function main(argv: string[]): Promise<number> {
1803
1875
  // launch/logs/debug/fixtures compose adapter scripts directly and own policy,
1804
1876
  // healing, teaching, and the --json contract.
1805
1877
  if (command === 'launch') return handleLaunch(argv.slice(1));
1878
+ if (command === 'stop') return handleStop(argv.slice(1));
1806
1879
  if (command === 'logs') return handleLogs(argv.slice(1));
1807
1880
  if (command === 'debug') return handleDebug(argv.slice(1));
1808
1881
  if (command === 'fixtures') return handleFixtures(argv.slice(1), { runOneNode });
@@ -8,7 +8,8 @@ import { execFileSync } from 'node:child_process';
8
8
  import fs from 'node:fs';
9
9
  import path from 'node:path';
10
10
 
11
- import { handleHarness } from '../harness.ts';
11
+ import { color } from '../cli-color.ts';
12
+ import { handleHarness, readRuntimeContextField, resolveRuntimeContextPath } from '../harness.ts';
12
13
  import { recipeHarnessPath, recipeRuntimeDir, recipeRuntimePath, runnerDir } from '../paths.ts';
13
14
  import type { MetaMaskRecipeAdapter } from '../types.ts';
14
15
  import { prepareMobile } from '../adapters/mobile/prepare.ts';
@@ -19,6 +20,7 @@ import {
19
20
  parseFlags,
20
21
  resolveAdapter,
21
22
  spawnScript,
23
+ spawnScriptStreaming,
22
24
  str,
23
25
  targetOf,
24
26
  usageOut,
@@ -227,15 +229,33 @@ function applyKVLines(output: string, overwrite: boolean): void {
227
229
  case 'SLOT_ID':
228
230
  if (overwrite || !process.env['RECIPE_SLOT_ID']) process.env['RECIPE_SLOT_ID'] = val;
229
231
  break;
232
+ case 'CDP_PORT':
233
+ if (overwrite || !process.env['CDP_PORT']) {
234
+ process.env['CDP_PORT'] = val;
235
+ process.env['RECIPE_CDP_PORT'] = val;
236
+ }
237
+ break;
230
238
  }
231
239
  }
232
240
  }
233
241
 
234
- // Resolve mobile slot port/simulator from the farmslot pool (pool-wins-over-env)
235
- // or from the slot-suffix formula (only fills unset vars). Called before explicit
242
+ // Resolve mobile slot port/simulator: the slot context the orchestrator wrote
243
+ // into the checkout wins first, then the farmslot pool (both overwrite env),
244
+ // then the slot-suffix formula (only fills unset vars). Called before explicit
236
245
  // CLI flag overrides so flags always win at the top.
237
- function resolveMobileSlotPorts(target: string): void {
246
+ export function resolveMobileSlotPorts(target: string): void {
238
247
  const resolveScript = path.join(runnerDir, 'adapters/shared/resolve-farmslot-ports.sh');
248
+ try {
249
+ // The checkout's own runtime context is authoritative — it names the exact
250
+ // simulator/port this slot was prepared with, surviving pool renames.
251
+ const ctxOut = execFileSync('bash', [
252
+ '-c', `source "${resolveScript}" && resolve_mobile_runtime_context "${target}"`,
253
+ ], { encoding: 'utf8', timeout: 5000, stdio: ['ignore', 'pipe', 'ignore'] });
254
+ if (ctxOut.trim()) {
255
+ applyKVLines(ctxOut, true);
256
+ return;
257
+ }
258
+ } catch { /* no runtime context — fall through to pool */ }
239
259
  try {
240
260
  // Pool match always wins — overwrite whatever is in the environment.
241
261
  const poolOut = execFileSync('bash', [
@@ -255,6 +275,43 @@ function resolveMobileSlotPorts(target: string): void {
255
275
  } catch { /* no slot suffix in dir name — stays at env defaults */ }
256
276
  }
257
277
 
278
+ // Resolve extension slot ports the same way as mobile: the checkout's runtime
279
+ // context first (cdpPort/devServerPort written by the orchestrator's prepare),
280
+ // then the farmslot pool, then the directory-suffix formula (fills unset only).
281
+ function resolveExtensionSlotPorts(target: string): void {
282
+ // The prepared checkout's context OVERWRITES inherited env (same authority as
283
+ // mobile's context/pool resolution): a stale CDP_PORT from the shell must not
284
+ // hijack the slot's browser. Explicit CLI flags are applied after and win.
285
+ const contextPath = resolveRuntimeContextPath(target);
286
+ const cdp = readRuntimeContextField(contextPath, 'cdpPort');
287
+ if (cdp) {
288
+ process.env['CDP_PORT'] = cdp;
289
+ process.env['RECIPE_CDP_PORT'] = cdp;
290
+ }
291
+ const dev = readRuntimeContextField(contextPath, 'devServerPort');
292
+ if (dev) {
293
+ process.env['WATCHER_PORT'] = dev;
294
+ process.env['RECIPE_WATCHER_PORT'] = dev;
295
+ }
296
+ if (process.env['CDP_PORT']) return;
297
+ const resolveScript = path.join(runnerDir, 'adapters/shared/resolve-farmslot-ports.sh');
298
+ try {
299
+ const poolOut = execFileSync('bash', [
300
+ '-c', `source "${resolveScript}" && resolve_farmslot_ports_by_repo "${target}"`,
301
+ ], { encoding: 'utf8', timeout: 5000, stdio: ['ignore', 'pipe', 'ignore'] });
302
+ if (poolOut.trim()) {
303
+ applyKVLines(poolOut, true);
304
+ return;
305
+ }
306
+ } catch { /* no pool match — fall through to formula */ }
307
+ try {
308
+ const defOut = execFileSync('bash', [
309
+ '-c', `source "${resolveScript}" && resolve_default_extension_ports "${target}"`,
310
+ ], { encoding: 'utf8', timeout: 5000, stdio: ['ignore', 'pipe', 'ignore'] });
311
+ if (defOut.trim()) applyKVLines(defOut, false);
312
+ } catch { /* no slot suffix in dir name — stays at env defaults */ }
313
+ }
314
+
258
315
  // Map the env-gap flags onto the env vars the leaf scripts read. For mobile, slot
259
316
  // port/simulator defaults are resolved first (pool-wins, then formula-fills-empty)
260
317
  // so that the harness respects slot isolation; explicit CLI flags applied below
@@ -265,9 +322,12 @@ function applyLaunchEnvOverrides(
265
322
  mobileTarget: string | undefined,
266
323
  target: string,
267
324
  ): void {
268
- // Mobile slot isolation: resolve WATCHER_PORT / IOS_SIMULATOR / SLOT_ID from
269
- // the farmslot pool or directory-suffix formula BEFORE applying explicit flags.
325
+ // Slot isolation: resolve ports/device from the checkout's own slot context,
326
+ // the farmslot pool, or the directory-suffix formula BEFORE applying explicit
327
+ // flags — extension needs this as much as mobile (CDP_PORT), so neither
328
+ // adapter hard-fails on a value its slot already knows.
270
329
  if (adapter === 'mobile') resolveMobileSlotPorts(target);
330
+ if (adapter === 'extension') resolveExtensionSlotPorts(target);
271
331
 
272
332
  const device = str(options, 'device');
273
333
  if (device && adapter === 'mobile') {
@@ -346,17 +406,19 @@ async function executeComposition(
346
406
  return extensionRebuild(target, json);
347
407
  }
348
408
 
349
- // Extension build or watch: full webpack compile via start-watch.sh.
409
+ // Extension build or watch: full webpack compile via start-watch.sh. Streams
410
+ // live — a webpack compile is minutes long and must never look hung.
350
411
  const startWatchSh = path.join(runnerDir, 'adapters/extension/start-watch.sh');
351
412
  const watchArgs = ['--target', target];
352
413
  if (process.env.WATCHER_PORT) watchArgs.push('--watcher-port', process.env.WATCHER_PORT);
353
- return spawnScript(startWatchSh, watchArgs, target, json);
414
+ console.error(`→ extension ${wantWatch ? 'watch' : 'build'} — webpack :${process.env.WATCHER_PORT ?? 'default'} (output streams below)`);
415
+ return spawnScriptStreaming(startWatchSh, watchArgs, target);
354
416
  }
355
417
 
356
418
  // Extension rebuild: kill the harness-owned webpack watcher (pid file + ps-scan
357
419
  // for orphans), clear the rebuild log, then drive the installed overlay's
358
420
  // live.sh --start-watch and tee its output into the rebuild log.
359
- function extensionRebuild(target: string, json: boolean): ScriptResult {
421
+ async function extensionRebuild(target: string, json: boolean): Promise<ScriptResult> {
360
422
  const runtimeDirRel = recipeRuntimeDir();
361
423
  const runtimeAbs = path.join(target, runtimeDirRel);
362
424
  const webpackPidFile = path.join(runtimeAbs, 'recipe-harness-webpack.pid');
@@ -418,7 +480,10 @@ function extensionRebuild(target: string, json: boolean): ScriptResult {
418
480
  const liveArgs = ['--target', target, '--start-watch'];
419
481
  if (process.env.CDP_PORT) liveArgs.push('--cdp-port', process.env.CDP_PORT);
420
482
  if (process.env.WATCHER_PORT) liveArgs.push('--watcher-port', process.env.WATCHER_PORT);
421
- const result = spawnScript(liveScript, liveArgs, target, json);
483
+ // Streams live: the watcher restart + Chromium boot runs for minutes and the
484
+ // command must show progress immediately, not a silent prompt.
485
+ console.error(`→ extension quick relaunch — webpack :${process.env.WATCHER_PORT ?? 'default'} · CDP :${process.env.CDP_PORT ?? 'default'} (output streams below)`);
486
+ const result = await spawnScriptStreaming(liveScript, liveArgs, target);
422
487
  // Tee live.sh output into rebuild.log for post-mortem diagnosis.
423
488
  if (result.output) {
424
489
  try { fs.appendFileSync(rebuildLog, result.output); } catch { /* best-effort */ }
@@ -489,7 +554,18 @@ function launchPass(
489
554
  ),
490
555
  );
491
556
  } else {
492
- console.error(`✓ launch ${adapter}${mobileTarget ? ` ${mobileTarget}` : ''} (${tier}) ready`);
557
+ // The summary must say what actually happened where: the target device,
558
+ // what the tier meant, and that the app+bridge are up — "(quick) ready"
559
+ // alone reads as a no-op even when a launch occurred.
560
+ const device =
561
+ adapter === 'mobile'
562
+ ? process.env.IOS_SIMULATOR || process.env.ADB_SERIAL || 'booted device'
563
+ : displayMode;
564
+ const tierNote = tier === 'quick' ? 'quick relaunch, no native build' : tier;
565
+ const devNote = process.env.MM_HARNESS_BIN ? ` ${color('dim', '[dev: MM_HARNESS_BIN]')}` : '';
566
+ console.error(
567
+ `${color('ok', '✓')} launch ${adapter}${mobileTarget ? ` ${mobileTarget}` : ''} — ${color('ok', String(device))} · ${tierNote} · app + bridge ready${devNote}`,
568
+ );
493
569
  }
494
570
  return EXIT.ok;
495
571
  }
package/src/doctor.ts CHANGED
@@ -1,6 +1,7 @@
1
1
  import fs from 'node:fs';
2
2
  import path from 'node:path';
3
3
 
4
+ import { color } from './cli-color.ts';
4
5
  import { readRuntimeContextField, resolveRuntimeContextPath } from './harness.ts';
5
6
  import { manifestPath, readJson, recipeHarnessRoot, recipeRuntimeDir, runnerDir } from './paths.ts';
6
7
  import type {
@@ -100,26 +101,32 @@ const RUNTIME_CONTEXT_FIELDS: ReadonlyArray<{
100
101
  envVars: readonly string[];
101
102
  envVar: string | null;
102
103
  customize: string;
104
+ adapters: readonly MetaMaskRecipeAdapter[];
103
105
  }> = [
104
- { key: 'slotId', envVars: ['RECIPE_SLOT_ID'], envVar: 'RECIPE_SLOT_ID', customize: 'farmslot dispatch writes this' },
105
- { key: 'extensionId', envVars: ['RECIPE_HARNESS_EXTENSION_ID'], envVar: 'RECIPE_HARNESS_EXTENSION_ID', customize: 'auto-resolved; edit file to pin' },
106
- { key: 'cdpPort', envVars: ['RECIPE_CDP_PORT', 'CDP_PORT'], envVar: 'CDP_PORT', customize: 'edit file or pass --cdp-port' },
107
- { key: 'runtimeStart.approved', envVars: ['RECIPE_RUNTIME_START_APPROVED'], envVar: 'RECIPE_RUNTIME_START_APPROVED', customize: 'edit file (true/false)' },
108
- { key: 'runtimeStart.command', envVars: [], envVar: null, customize: 'edit file' },
109
- { key: 'runtimeStart.readyUrl', envVars: ['RECIPE_RUNTIME_READY_URL'], envVar: 'RECIPE_RUNTIME_READY_URL', customize: 'edit file' },
106
+ { key: 'slotId', envVars: ['RECIPE_SLOT_ID'], envVar: 'RECIPE_SLOT_ID', customize: 'farmslot dispatch writes this', adapters: ['mobile', 'extension', 'core'] },
107
+ { key: 'extensionId', envVars: ['RECIPE_HARNESS_EXTENSION_ID'], envVar: 'RECIPE_HARNESS_EXTENSION_ID', customize: 'auto-resolved; edit file to pin', adapters: ['extension'] },
108
+ { key: 'cdpPort', envVars: ['RECIPE_CDP_PORT', 'CDP_PORT'], envVar: 'CDP_PORT', customize: 'edit file or pass --cdp-port', adapters: ['extension'] },
109
+ { key: 'runtimeStart.approved', envVars: ['RECIPE_RUNTIME_START_APPROVED'], envVar: 'RECIPE_RUNTIME_START_APPROVED', customize: 'edit file (true/false)', adapters: ['mobile', 'extension'] },
110
+ { key: 'runtimeStart.command', envVars: [], envVar: null, customize: 'edit file', adapters: ['mobile', 'extension'] },
111
+ { key: 'runtimeStart.readyUrl', envVars: ['RECIPE_RUNTIME_READY_URL'], envVar: 'RECIPE_RUNTIME_READY_URL', customize: 'edit file', adapters: ['mobile', 'extension'] },
110
112
  ];
111
113
 
112
114
  // Report every runtime-context field with its current value and where it came from
113
115
  // (live env override > file > unset default), reusing the harness's own file reader
114
116
  // so doctor and dispatch agree on resolution. Absent file → fileExists:false and
115
117
  // every field falls back to env or default; the path shows where it WOULD live.
116
- export function runtimeContextSummary(target: string): MetaMaskRuntimeContextReport {
118
+ // An adapter scopes the report to that platform's fields — an extension-only row
119
+ // (cdpPort, extensionId) is noise on a mobile or core slot.
120
+ export function runtimeContextSummary(target: string, adapter?: MetaMaskRecipeAdapter): MetaMaskRuntimeContextReport {
117
121
  const contextPath = resolveRuntimeContextPath(target);
118
122
  const envOverride = process.env.RECIPE_RUNTIME_CONTEXT ?? null;
119
123
  const fileExists = fs.existsSync(contextPath);
120
124
  const file = envOverride ?? path.relative(target, contextPath);
121
125
  const fields: Record<string, MetaMaskRuntimeContextField> = {};
122
- for (const spec of RUNTIME_CONTEXT_FIELDS) {
126
+ const specs = adapter
127
+ ? RUNTIME_CONTEXT_FIELDS.filter((spec) => spec.adapters.includes(adapter))
128
+ : RUNTIME_CONTEXT_FIELDS;
129
+ for (const spec of specs) {
123
130
  const envValue = spec.envVars
124
131
  .map((name) => process.env[name])
125
132
  .find((value) => value !== undefined && value !== '');
@@ -137,16 +144,21 @@ export function runtimeContextSummary(target: string): MetaMaskRuntimeContextRep
137
144
 
138
145
  // Human-readable runtime-context section for `doctor` without --json.
139
146
  export function renderRuntimeContext(runtimeContext: MetaMaskRuntimeContextReport): string {
147
+ // Human render: set values read bright with their provenance highlighted;
148
+ // unset defaults read dim so the eye lands on what is actually configured.
149
+ const out = (style: string, text: string) => color(style, text, { stream: process.stdout });
140
150
  const lines: string[] = [];
141
151
  lines.push(
142
152
  runtimeContext.fileExists
143
- ? `runtime-context: ${runtimeContext.file} (present)`
144
- : `runtime-context: ${runtimeContext.file} (absent — written by farmslot prepare/dispatch)`,
153
+ ? `${out('label', 'runtime-context:')} ${runtimeContext.file} ${out('ok', '(present)')}`
154
+ : `${out('label', 'runtime-context:')} ${runtimeContext.file} ${out('dim', '(absent — written by farmslot prepare/dispatch)')}`,
145
155
  );
146
156
  for (const [key, field] of Object.entries(runtimeContext.fields)) {
147
- const value = field.value ?? '(unset)';
157
+ const isSet = field.value !== undefined && field.value !== null && field.value !== '';
158
+ const value = isSet ? out('ok', String(field.value)) : out('dim', '(unset)');
148
159
  const origin = field.source === 'env' && field.envVar ? `env ${field.envVar}` : field.source;
149
- lines.push(` ${key.padEnd(22)} ${value} [${origin}] ${field.customize}`);
160
+ const originTag = isSet ? out('accent', `[${origin}]`) : out('dim', `[${origin}]`);
161
+ lines.push(` ${key.padEnd(22)} ${value} ${originTag} ${out('dim', `— ${field.customize}`)}`);
150
162
  }
151
163
  return lines.join('\n');
152
164
  }
@@ -189,7 +201,7 @@ export function createDoctorReport(
189
201
  compatibilityMode: mode,
190
202
  shape: repoShape(target),
191
203
  fixture: fixtureSummary(target),
192
- runtimeContext: runtimeContextSummary(target),
204
+ runtimeContext: runtimeContextSummary(target, adapter),
193
205
  manifestValidation: manifestValidation.summary,
194
206
  };
195
207
  }
@@ -7,11 +7,13 @@
7
7
  // teach; genuinely-unknown commands fall to commander's default.
8
8
 
9
9
  import { spawnSync } from 'node:child_process';
10
+ import fs from 'node:fs';
10
11
  import path from 'node:path';
11
12
  import { fileURLToPath } from 'node:url';
12
13
 
13
14
  import { Command } from 'commander';
14
15
 
16
+ import { color } from './cli-color.ts';
15
17
  import { handleUpdate, maybeNudge } from './commands/update.ts';
16
18
 
17
19
  // …/src → its parent is the package root that holds scripts/.
@@ -57,6 +59,25 @@ Example:
57
59
  mm-harness actions --adapter mobile
58
60
  mm-harness actions --adapter extension --raw`,
59
61
  },
62
+ {
63
+ name: 'stop',
64
+ summary: 'Stop the dev server this checkout owns (mobile: port-scoped Metro) and close its log window.',
65
+ example: 'mm-harness stop',
66
+ helpText: `mm-harness stop [flags]
67
+
68
+ Stop the Metro dev server this checkout owns and close its tmux log-tail window.
69
+ Port-scoped: only the listener on the checkout's resolved port is signalled, so
70
+ concurrent slots are untouched. Nothing running = success (idempotent).
71
+ Extension/core: teaching error (their dev servers are slot-managed).
72
+
73
+ --port <port> Metro port (default: the checkout's slot context)
74
+ --target <path> Checkout path (default: cwd)
75
+ --json Machine-readable output
76
+
77
+ Example:
78
+ mm-harness stop
79
+ mm-harness stop --port 8061`,
80
+ },
60
81
  {
61
82
  name: 'call',
62
83
  summary: 'Run one action in isolation as a one-node recipe through the real engine path (fuzzy short names; --arg k=v; same trace/evidence as run).',
@@ -345,7 +366,7 @@ const HELP_GROUPS: HelpGroup[] = [
345
366
  {
346
367
  title: 'DAILY LOOP',
347
368
  blurb: 'what a teammate runs many times a day (auto-ensures the overlay; --heal owns recovery)',
348
- commands: ['launch', 'logs', 'debug', 'fixtures'],
369
+ commands: ['launch', 'stop', 'logs', 'debug', 'fixtures'],
349
370
  },
350
371
  {
351
372
  title: 'DISCOVER',
@@ -375,19 +396,51 @@ function commandMeta(name: string): { summary: string; example: string; planned:
375
396
  return { summary: '', example: '', planned: false };
376
397
  }
377
398
 
399
+ // Slot context the orchestrator's prepare wrote into the checkout, when the
400
+ // help is run from inside one. Presence-gated: no context, no line.
401
+ function detectedSlotLine(out: (style: string, text: string) => string): string | null {
402
+ const ctxPath = path.join(
403
+ process.cwd(),
404
+ process.env.RECIPE_RUNTIME_DIR || 'temp/recipe/runtime',
405
+ 'agentic-runtime.json',
406
+ );
407
+ try {
408
+ const ctx = JSON.parse(fs.readFileSync(ctxPath, 'utf8')) as Record<string, unknown>;
409
+ const parts: string[] = [];
410
+ if (ctx.slotId) parts.push(`slot ${out('ok', String(ctx.slotId))}`);
411
+ if (ctx.simulator) parts.push(`device ${out('ok', String(ctx.simulator))}`);
412
+ if (ctx.metroPort) parts.push(`metro :${out('ok', String(ctx.metroPort))}`);
413
+ if (ctx.gitBranch) parts.push(`branch ${out('info', String(ctx.gitBranch))}`);
414
+ if (parts.length === 0) return null;
415
+ return `${out('label', 'SLOT')} — this checkout is a prepared slot: ${parts.join(' · ')}`;
416
+ } catch {
417
+ return null;
418
+ }
419
+ }
420
+
378
421
  function groupedHelp(): string {
422
+ const out = (style: string, text: string) => color(style, text, { stream: process.stdout });
379
423
  const lines: string[] = [];
380
- lines.push('mm-harness — one front door for the MetaMask recipe loop: launch the app, prove behavior, manage the runtime overlay.');
424
+ lines.push(`${out('bold', 'mm-harness')} — one front door for the MetaMask recipe loop: launch the app, prove behavior, manage the runtime overlay.`);
381
425
  lines.push('Run it from inside a MetaMask checkout; the platform (mobile | extension | core) is auto-detected.');
382
- lines.push('Grammar: mm-harness <command> [target] [flags] (target is a positional: ios | android; flags add agent depth; --json is the agent contract)');
426
+ lines.push(`Grammar: ${out('cmd', 'mm-harness <command> [target] [flags]')} (target is a positional: ios | android; flags add agent depth; --json is the agent contract)`);
427
+ if (process.env.MM_HARNESS_BIN) {
428
+ lines.push('');
429
+ lines.push(`${out('warn', 'DEV OVERRIDE ACTIVE')} — this run is served by MM_HARNESS_BIN=${out('path', process.env.MM_HARNESS_BIN)} (unset it to return to the installed/global bin).`);
430
+ }
431
+ const slotLine = detectedSlotLine(out);
432
+ if (slotLine) {
433
+ lines.push('');
434
+ lines.push(slotLine);
435
+ }
383
436
  for (const group of HELP_GROUPS) {
384
437
  lines.push('');
385
- lines.push(`${group.title} — ${group.blurb}:`);
438
+ lines.push(`${out('label', group.title)} — ${group.blurb}:`);
386
439
  for (const name of group.commands) {
387
440
  const meta = commandMeta(name);
388
441
  const planned = meta.planned ? ' (planned)' : '';
389
- lines.push(` ${name.padEnd(10)} ${meta.summary}${planned}`);
390
- lines.push(` ${meta.example}`);
442
+ lines.push(` ${out('cmd', name.padEnd(10))} ${meta.summary}${planned}`);
443
+ lines.push(` ${out('comment', meta.example)}`);
391
444
  }
392
445
  }
393
446
  lines.push('');