emrg: isolate GUI test port file + alert on scheduler connect failures (rant 08:03)

Agent.md

@@ -93,7 +93,7 @@ Community needs voiced in HN agent-UI discussions map directly to EMRG's design:
 pkill -f "emrg.server"; rm -f ~/.emrg/emrgd.port; python -m emrg
 ```
 
-Python: `uv run pytest tests/ -v` (634) — import check: `uv run python -c "from emrg.client.app import run_client"`
+Python: `uv run pytest tests/ -v` (636) — import check: `uv run python -c "from emrg.client.app import run_client"`
 GUI: `cd emrg/gui && npm test` (96: 22 daemon_client + 22 app-commands + 27 renderer smoke + 15 i18n + 7 integration + 3 commands) — syntax: `node --check main.js preload.js daemon_client.js renderer/js/*.js`
 CI: `uv run pytest` + GUI tests + **actionlint workflow lint** (`rhysd/actionlint@v1.7.12` gate, #444 — workflow 解析错误在 PR CI 即失败，如 `if:` secrets 上下文)
 Re-trigger: `scripts/re-trigger-ci.sh [branch]` (workflow_dispatch, #527 — 替代空 commit 重触发：Actions outage 会整段丢弃 push 事件，dispatch 走 API 路径不受影响)

README.cn.md

@@ -274,15 +274,15 @@ EMRG 不只是追赶——它自己追上来。
 git clone https://github.com/argszero/emrg.git
 cd emrg
 uv sync # 安装依赖
-uv run pytest tests/ -v # 跑测试（当前 634 项）
+uv run pytest tests/ -v # 跑测试（当前 636 项）
 uv run python -m emrg # 启动 TUI
 # CI 含 actionlint workflow 门禁（#444）：workflow 解析错误在 PR 即失败
 
 # 可选：Electron GUI（非开发者主入口，Phase 3）
 cd emrg/gui
 npm ci # 安装依赖（生产模式可 --omit=dev）
 npm start # 启动 GUI（自动拉起 daemon）
-npm test # 运行 Node 测试（91 项：22 daemon_client + 22 app-commands + 22 renderer smoke + 15 i18n + 7 integration + 3 commands；集成测试在 CI 跑，本地可 npm run test:integration）
+npm test # 运行 Node 测试（96 项：22 daemon_client + 22 app-commands + 27 renderer smoke + 15 i18n + 7 integration + 3 commands；集成测试在 CI 跑，本地可 npm run test:integration）
 ```
 
 CI 通过 GitHub Actions 自动运行测试并检查冲突标记（`.github/workflows/test.yml`）。

README.md

@@ -273,7 +273,7 @@ EMRG doesn't just keep up — it catches up on its own.
 git clone https://github.com/argszero/emrg.git
 cd emrg
 uv sync # install deps
-uv run pytest tests/ -v # run tests (currently 634 items)
+uv run pytest tests/ -v # run tests (currently 636 items)
 uv run python -m emrg # launch TUI
 # CI includes actionlint workflow gate (#444): workflow parse errors fail PR CI
 

emrg/gui/daemon_client.js

@@ -18,7 +18,12 @@ const crypto = require("crypto");
 const { spawn } = require("child_process");
 const WebSocket = require("ws");
 
-const PORT_FILE = () => path.join(os.homedir(), ".emrg", "emrgd.port");
+// G129 (rant 2026-08-09T08:03:46): PORT_FILE 必须接受 projectDir——硬编码
+// os.homedir() 时，Windows 测试的 setupTempHome() 只设 HOME 不设 USERPROFILE，
+// os.homedir() 仍读 USERPROFILE（真实用户目录）→ 测试把假 port/token 写进
+// 真实的 ~/.emrg/emrgd.port → 演化周期 10 小时连不上 daemon（WinError 1225）。
+// 所有调用点必须传 this.projectDir（默认 os.homedir() 保持生产行为不变）。
+const PORT_FILE = (projectDir = os.homedir()) => path.join(projectDir, ".emrg", "emrgd.port");
 const MAX_PAYLOAD = 16 * 1024 * 1024; // G62/G105：16MB 双向一致（工具输出上限 200KB）
 const AUTH_TIMEOUT_MS = 10_000;
 const SPAWN_WAIT_MS = 5_000;
@@ -74,7 +79,7 @@ class DaemonClient {
  isRunning(timeoutMs = 1500) {
  // G43/G90：TCP 探测（不可简化为 port 文件存在）
  try {
- const port = Number(fs.readFileSync(PORT_FILE(), "utf8").split("\n")[0]);
+ const port = Number(fs.readFileSync(PORT_FILE(this.projectDir), "utf8").split("\n")[0]);
  return new Promise((resolve) => {
  const sock = net.connect({ host: "127.0.0.1", port, timeout: timeoutMs });
  sock.once("connect", () => { sock.destroy(); resolve(true); });
@@ -166,12 +171,12 @@ class DaemonClient {
  // 1. 读 port 文件 → 无则拉 daemon
  let port, token;
  try {
- const text = fs.readFileSync(PORT_FILE(), "utf8");
+ const text = fs.readFileSync(PORT_FILE(this.projectDir), "utf8");
  [port, token] = text.split(/\s+/);
  if (!port || !token) throw new Error("malformed port file");
  } catch {
  await this.startDaemon();
- const text = fs.readFileSync(PORT_FILE(), "utf8");
+ const text = fs.readFileSync(PORT_FILE(this.projectDir), "utf8");
  [port, token] = text.split(/\s+/);
  }
 
@@ -188,9 +193,9 @@ class DaemonClient {
  // G43：port 文件存在但连不上（daemon 已死/端口被占）→ 删文件重拉一次
  this.logger.warn(`[gui] ws connect failed: ${e.message} — stale port, respawning daemon`);
  try { this.ws.close(); } catch { /* ignore */ }
- try { fs.unlinkSync(PORT_FILE()); } catch { /* ignore */ }
+ try { fs.unlinkSync(PORT_FILE(this.projectDir)); } catch { /* ignore */ }
  await this.startDaemon();
- const text = fs.readFileSync(PORT_FILE(), "utf8");
+ const text = fs.readFileSync(PORT_FILE(this.projectDir), "utf8");
  [port, token] = text.split(/\s+/);
  this.ws = new WebSocket(`ws://127.0.0.1:${port}`, { maxPayload: MAX_PAYLOAD });
  await this._awaitOpen();

emrg/gui/test/daemon_client.test.js

@@ -63,18 +63,39 @@ Module._load = function (request, parent, isMain) {
 const { DaemonClient, generateSessionId, PORT_FILE } = require("../daemon_client.js");
 let tmpHome = null;
 let origHome = null;
+let origUserProfile = null;
+
+// G129 (rant 2026-08-09T08:03:46): 测试隔离守卫——写 port 文件前断言目标路径
+// 位于临时目录内。Windows 上 Node os.homedir() 优先读 USERPROFILE（HOME 无效），
+// 若无此守卫，PORT_FILE() 会解析到真实 ~/.emrg/emrgd.port，测试假值
+// ("41234\nseekrit-token") 会覆盖真实 daemon 端口文件 → 演化周期 10h 连不上。
+function assertPortFileInTmp(portFile) {
+ const resolved = path.resolve(portFile);
+ const tmpResolved = path.resolve(tmpHome);
+ assert.ok(
+ resolved.startsWith(tmpResolved + path.sep),
+ `port file ${resolved} escapes tmpHome ${tmpResolved} — refusing to write`,
+ );
+}
 
 function setupTempHome() {
  tmpHome = fs.mkdtempSync(path.join(os.tmpdir(), "emrg-gui-test-"));
  fs.mkdirSync(path.join(tmpHome, ".emrg"), { recursive: true });
  origHome = process.env.HOME;
+ origUserProfile = process.env.USERPROFILE;
  process.env.HOME = tmpHome;
- // 预写 port 文件（模拟已运行 daemon）
- fs.writeFileSync(PORT_FILE(), "41234\nseekrit-token");
+ // G129: Windows os.homedir() 读 USERPROFILE —— 必须一并重定向，否则
+ // os.homedir() 仍返回真实用户目录（这是 10h daemon gap 的直接根因）。
+ process.env.USERPROFILE = tmpHome;
+ // 预写 port 文件（模拟已运行 daemon）—— 路径必须落在 tmpHome 内
+ const portFile = PORT_FILE(tmpHome);
+ assertPortFileInTmp(portFile);
+ fs.writeFileSync(portFile, "41234\nseekrit-token");
 }
 
 function teardownTempHome() {
  if (origHome !== undefined) process.env.HOME = origHome; else delete process.env.HOME;
+ if (origUserProfile !== undefined) process.env.USERPROFILE = origUserProfile; else delete process.env.USERPROFILE;
  if (tmpHome) fs.rmSync(tmpHome, { recursive: true, force: true });
  tmpHome = null;
 }
@@ -130,13 +151,13 @@ test("ensureConnected: port 文件读取 + auth 首帧 + auth_ok", async () => {
 });
 
 test("ensureConnected: port 文件缺失 → 拉起 daemon（spawn 参数正确 G28/G68/G125）", async () => {
- fs.rmSync(PORT_FILE(), { force: true });
+ fs.rmSync(PORT_FILE(tmpHome), { force: true });
  const client = new DaemonClient({ projectDir: tmpHome });
  // stub startDaemon：模拟拉起后写 port 文件
  let spawnCalls = null;
  client.startDaemon = async function () {
  spawnCalls = { python: this._findPython(), projectDir: this.projectDir };
- fs.writeFileSync(PORT_FILE(), "41235\nseekrit-token");
+ fs.writeFileSync(PORT_FILE(tmpHome), "41235\nseekrit-token");
  };
  await connectClient(client);
  assert.ok(spawnCalls, "startDaemon should be called");
@@ -205,7 +226,7 @@ test("G43 stale port: 连接失败（port 文件存在但拒绝）→ 删文件
  let respawned = false;
  client.startDaemon = async function () {
  respawned = true;
- fs.writeFileSync(PORT_FILE(), "41236\nseekrit-token");
+ fs.writeFileSync(PORT_FILE(tmpHome), "41236\nseekrit-token");
  };
  const p = client.ensureConnected();
  await waitForWs();
@@ -214,7 +235,7 @@ test("G43 stale port: 连接失败（port 文件存在但拒绝）→ 删文件
  // 重拉后创建第二个 ws → open → auth → auth_ok
  await waitForWs(() => currentMockWs !== firstWs);
  assert.ok(respawned, "startDaemon should respawn after stale port");
- assert.strictEqual(fs.existsSync(PORT_FILE()), true);
+ assert.strictEqual(fs.existsSync(PORT_FILE(tmpHome)), true);
  assert.strictEqual(currentMockWs.url, "ws://127.0.0.1:41236");
  currentMockWs.emit("open");
  await waitForAuthSent(currentMockWs);
@@ -512,7 +533,7 @@ test("isRunning：TCP 探测（G43/G90）", async () => {
  net.connect = origConnect;
  }
  // port 文件缺失 → false
- fs.rmSync(PORT_FILE(), { force: true });
+ fs.rmSync(PORT_FILE(tmpHome), { force: true });
  assert.strictEqual(await client.isRunning(), false);
 });
 

emrg/gui/test/integration.test.js

@@ -41,7 +41,11 @@ process.env.HOME = tmp;
 process.env.USERPROFILE = tmp;
 
 function findPython() {
- const root = path.resolve(__dirname, "..", ".."); // emrg/gui → 项目根
+ // G129: 本文件位于 emrg/gui/test/ —— 必须上溯 3 级到仓库根（emrg/gui/ 下的
+ // daemon_client.js 才是 2 级）。此前只上溯 2 级解析到 emrg/ 包目录，找不到
+ // .venv → 回退 PATH python3/python；本机 PATH python 是损坏的
+ // ~/.emrg/install/bin/python.exe（Failed to import encodings）→ daemon 起不来。
+ const root = path.resolve(__dirname, "..", "..", ".."); // test → gui → emrg → 仓库根
  const candidates = [
  path.join(root, ".venv", "bin", "python"),
  path.join(root, ".venv", "Scripts", "python.exe"),
@@ -65,7 +69,7 @@ function waitForPortFile(timeoutMs = 15000) {
  return new Promise((resolve, reject) => {
  const check = () => {
  try {
- const text = fs.readFileSync(PORT_FILE(), "utf8");
+ const text = fs.readFileSync(PORT_FILE(tmp), "utf8");
  if (text && text.trim()) return resolve(text);
  } catch { /* not yet */ }
  if (Date.now() > deadline) return reject(new Error("daemon port file timeout"));
@@ -180,7 +184,7 @@ test("daemon 被杀 → ensureConnected 重连（G43 stale port 流程）", { sk
  // ensureConnected 应自动重拉（G43 stale port）
  await client.ensureConnected();
  assert.strictEqual(client.connected, true);
- assert.ok(fs.existsSync(PORT_FILE()), "new port file written");
+ assert.ok(fs.existsSync(PORT_FILE(tmp)), "new port file written");
  // 重连后可继续通信
  const frame = await client.sendCommandAndWait("list_sessions", { cwd: tmp }, 5000);
  assert.strictEqual(frame.type, "sessions_list");

emrg/server/scheduler.py

@@ -142,6 +142,10 @@ def __init__(
  #
  # Counter is persisted to disk to survive daemon restarts.
  self._IDLE_HALT_THRESHOLD = 30
+ # G129: 连续连接失败告警阈值——达到后升级为 ERROR（防静默吞掉，
+ # rant 2026-08-09T08:03:46：GUI 测试覆盖真实 emrgd.port 致 10h 连不上）。
+ self._CONNECT_FAIL_ALERT = 3
+ self._connect_failures = 0
  self._saturation_dir = config_dir() / "saturation"
  self._saturation_dir.mkdir(parents=True, exist_ok=True)
  self._saturation_file = self._saturation_dir / f"{self.name}.json"
@@ -684,10 +688,27 @@ async def _run_evolution_cycle(self) -> None:
  try:
  ws = await connect_to_server()
  logger.info("EvolutionHandler[%s]: connected", self.name)
- except (ConnectionRefusedError, FileNotFoundError) as e:
- logger.warning(
- "EvolutionHandler[%s]: cannot connect: %s", self.name, e
- )
+ self._connect_failures = 0
+ except (ConnectionRefusedError, FileNotFoundError, OSError) as e:
+ # G129 (rant 2026-08-09T08:03:46): 连接失败不得静默吞掉——GUI 测试曾把
+ # 假 port 值写进真实 ~/.emrg/emrgd.port，导致演化周期 10 小时连不上
+ # daemon（WinError 1225）只留下 WARNING。累计失败达到阈值后升级为
+ # ERROR 告警，提示 port 文件可能被外部覆盖（检查 ~/.emrg/emrgd.port）。
+ self._connect_failures += 1
+ port_path = config_dir() / "emrgd.port"
+ if self._connect_failures >= self._CONNECT_FAIL_ALERT:
+ logger.error(
+ "EvolutionHandler[%s]: cannot connect for %d consecutive cycles "
+ "(%s) — daemon unreachable. Check %s (may have been overwritten "
+ "by GUI tests or stale after daemon restart); run 'emrg server' "
+ "or restart the daemon to recover.",
+ self.name, self._connect_failures, e, port_path,
+ )
+ else:
+ logger.warning(
+ "EvolutionHandler[%s]: cannot connect (%d/%d): %s",
+ self.name, self._connect_failures, self._CONNECT_FAIL_ALERT, e,
+ )
  return
 
  task_msg = json.dumps(

tests/test_doc_counts.py

@@ -4,6 +4,12 @@
 removed, the documented counts drift and require a follow-up doc PR. This
 module asserts the documented Python count matches the real collection, and
 that the documented GUI breakdown sums to its headline number.
+
+#584: README.cn.md was the only test-count doc NOT guarded — it drifted to
+91 (22 renderer smoke) while README.md/Agent.md said 96 (27 renderer smoke)
+after #580 added 3 GUI tests. Both checks now cover all three docs
+(README.md, README.cn.md, Agent.md); CJK full-width parens and the
+"项：" separator are normalized before matching.
 """
 
 import re
@@ -29,11 +35,14 @@ def _collected_pytest_count() -> int:
 def _gui_breakdowns() -> list[tuple[str, int, list[int]]]:
  """Extract (label, headline, parts) for every documented GUI count."""
  found = []
- for doc in ("README.md", "Agent.md"):
+ for doc in ("README.md", "README.cn.md", "Agent.md"):
  text = (REPO_ROOT / doc).read_text(encoding="utf-8")
  for line in text.splitlines():
  if "npm test" not in line:
  continue
+ # CJK docs use full-width parens and "（N 项：..." instead of "(N: ..."
+ line = line.replace("（", "(").replace("）", ")")
+ line = re.sub(r"(\d+) 项：", r"\1: ", line)
  m = re.search(r"\((\d+): ([^)]+)\)", line)
  if not m:
  continue
@@ -51,10 +60,13 @@ def _gui_breakdowns() -> list[tuple[str, int, list[int]]]:
 
 def test_python_count_matches_docs() -> None:
  collected = _collected_pytest_count()
- for doc in ("README.md", "Agent.md"):
+ for doc in ("README.md", "README.cn.md", "Agent.md"):
  text = (REPO_ROOT / doc).read_text(encoding="utf-8")
- # README: "run tests (currently N items)" | Agent.md: "pytest tests/ -v` (N)"
+ # README: "run tests (currently N items)" | README.cn: "（当前 N 项）"
+ # Agent.md: "pytest tests/ -v` (N)"
  m = re.search(r"currently (\d+) items", text) or re.search(
+ r"当前 (\d+) 项", text
+ ) or re.search(
  r"uv run pytest tests/ -v` \((\d+)\)", text
  )
  assert m, f"no documented Python count found in {doc}"

tests/test_scheduler.py

@@ -1137,6 +1137,87 @@ def test_evolution_cycle_aborted_resets_empty_streak(tmp_path):
  assert handler._empty_cycles == 0, "abort resets the streak (not a real empty cycle)"
 
 
+# ── Connect-failure alerting (G129, rant 2026-08-09T08:03:46) ─────
+# GUI tests once overwrote the real ~/.emrg/emrgd.port with fake values,
+# so the evolution cycle failed to reach the daemon for 10 hours with only
+# a WARNING log. Consecutive failures must escalate to ERROR + carry an
+# actionable hint (check the port file), never silently swallow.
+
+def test_evolution_cycle_connect_failure_escalates_to_error(tmp_path, caplog):
+ """Repeated connect failures must escalate from warning to error alert."""
+ import logging
+
+ from emrg.server import scheduler as mod
+
+ handler, captured = _make_cycle_handler(tmp_path, frames=[])
+ async def _refuse():
+ raise ConnectionRefusedError("no daemon")
+ mod.connect_to_server = _refuse
+ try:
+ for i in range(handler._CONNECT_FAIL_ALERT):
+ with caplog.at_level(logging.ERROR, logger="emrg.server.scheduler"):
+ asyncio.run(handler._run_evolution_cycle())
+ assert "log" not in captured, "connect failure must not write an evolution log"
+ assert handler.evolutions == []
+ assert handler._empty_cycles == 0, "connect failure ≠ empty cycle"
+ assert handler._connect_failures == handler._CONNECT_FAIL_ALERT
+ # 第 3 次（达到阈值）必须出现 ERROR 告警，且提示检查 port 文件
+ error_msgs = [r.message for r in caplog.records if r.levelno >= logging.ERROR]
+ assert error_msgs, "expected an ERROR alert after threshold"
+ assert any("emrgd.port" in m for m in error_msgs), error_msgs
+ assert any("consecutive" in m for m in error_msgs), error_msgs
+ finally:
+ mod.connect_to_server = _original_connect_to_server()
+
+
+def test_evolution_cycle_connect_failure_resets_on_success(tmp_path, caplog):
+ """A successful connection resets the consecutive-failure counter."""
+ from emrg.server import scheduler as mod
+
+ handler, captured = _make_cycle_handler(tmp_path, frames=[])
+ async def _refuse():
+ raise ConnectionRefusedError("no daemon")
+ mod.connect_to_server = _refuse
+ try:
+ asyncio.run(handler._run_evolution_cycle())
+ assert handler._connect_failures == 1
+ # 成功连接 → 计数归零
+ async def _fake_connect():
+ return _FakeWsForCycle([{"request_id": "r1", "content": "Done", "done": True,
+ "delta": False, "session_id": "s"}])
+ mod.connect_to_server = _fake_connect
+ asyncio.run(handler._run_evolution_cycle())
+ assert handler._connect_failures == 0, "success must reset the failure counter"
+ finally:
+ mod.connect_to_server = _original_connect_to_server()
+
+
+class _FakeWsForCycle:
+ """Minimal ws stand-in for the reset-on-success test."""
+ def __init__(self, frames):
+ import json as _json
+ from websockets.exceptions import ConnectionClosed as _Closed
+ self._frames = list(frames)
+ self._json = _json
+ self._Closed = _Closed
+ self.sent = []
+ async def send(self, msg):
+ self.sent.append(msg)
+ async def recv(self):
+ if self._frames:
+ return self._json.dumps(self._frames.pop(0), ensure_ascii=False)
+ raise self._Closed()
+ async def close(self):
+ pass
+
+
+def _original_connect_to_server():
+ """Restore the real connect_to_server after a test replaced it."""
+ import importlib
+ from emrg.server import scheduler as mod
+ return importlib.import_module("emrg.connect").connect_to_server
+
+
 # ── Saturation halt auto-resume on upstream advance ───────────────
 # The halt skips scheduled runs entirely, so a halted handler can never
 # detect a HEAD change itself (only /trigger could resume it). If every