aps-agent/scripts/smoke-sidecar-local.mjs

// ============================================================
// Sidecar 本地冒烟（矩阵 99 · 本地可测部分）
// 覆盖：20 次冷启动 / 端口冲突 / 崩溃重启（父进程守护）/ 中文+空格路径
// 只读调用 server/sidecar.py（python -m server.sidecar），不改后端逻辑。
// 任一检查失败 → 非零退出码。
// 用法：node scripts/smoke-sidecar-local.mjs
// 可选：APS_SIDECAR_PYTHON=<python> 指定解释器（默认 .sidecar-venv → .venv → PATH）
// ============================================================
import { spawn, spawnSync } from 'node:child_process';
import crypto from 'node:crypto';
import fs from 'node:fs';
import net from 'node:net';
import os from 'node:os';
import path from 'node:path';
import { fileURLToPath } from 'node:url';

const repoRoot = path.resolve(path.dirname(fileURLToPath(import.meta.url)), '..');
const isWin = process.platform === 'win32';
const LOOPBACK_HOST = '127.0.0.1';
const NONCE_HEADER = 'x-aps-sidecar-nonce';
const HEALTH_URL = (port) => `http://${LOOPBACK_HOST}:${port}/api/health`;

const results = [];
const liveChildren = new Set();

// ---------------------------------------------------------------- helpers

function resolvePython() {
  if (process.env.APS_SIDECAR_PYTHON) return process.env.APS_SIDECAR_PYTHON;
  const candidates = [
    path.join(repoRoot, '.sidecar-venv', 'Scripts', 'python.exe'),
    path.join(repoRoot, '.venv', 'Scripts', 'python.exe'),
    path.join(repoRoot, '.venv', 'bin', 'python3'),
  ];
  for (const candidate of candidates) {
    if (fs.existsSync(candidate)) return candidate;
  }
  return isWin ? 'python' : 'python3';
}

function reservePort() {
  return new Promise((resolve, reject) => {
    const server = net.createServer();
    server.once('error', reject);
    server.listen({ host: LOOPBACK_HOST, port: 0, exclusive: true }, () => {
      const address = server.address();
      const port = typeof address === 'object' && address ? address.port : 0;
      server.close((error) => error ? reject(error) : resolve(port));
    });
  });
}

function delay(milliseconds) {
  return new Promise((resolve) => setTimeout(resolve, milliseconds));
}

function nonce() {
  return crypto.randomBytes(32).toString('hex');
}

function probeHealth(port, expectedNonce, timeoutMs = 2_000) {
  return new Promise((resolve) => {
    const controller = new AbortController();
    const timer = setTimeout(() => controller.abort(), timeoutMs);
    fetch(HEALTH_URL(port), {
      headers: { [NONCE_HEADER]: expectedNonce },
      signal: controller.signal,
    })
      .then(async (response) => {
        const echoed = response.headers.get(NONCE_HEADER);
        clearTimeout(timer);
        resolve(response.status === 200 && echoed === expectedNonce);
      })
      .catch(() => {
        clearTimeout(timer);
        resolve(false);
      });
  });
}

async function waitForHealth(port, expectedNonce, timeoutMs = 60_000, child = null) {
  const deadline = Date.now() + timeoutMs;
  let last = null;
  while (Date.now() < deadline) {
    if (child && child.exitCode !== null) {
      throw new Error(`sidecar exited early code=${child.exitCode}: ${child.stderrTail ?? ''}`);
    }
    last = await probeHealth(port, expectedNonce);
    if (last) return true;
    await delay(250);
  }
  throw new Error(`sidecar did not become healthy within ${timeoutMs}ms (last=${last})`);
}

function stopProcessTree(child) {
  if (!child || !Number.isInteger(child.pid) || child.pid <= 0) return;
  liveChildren.delete(child);
  if (isWin) {
    spawnSync('taskkill.exe', ['/pid', String(child.pid), '/T', '/F'], {
      stdio: 'ignore',
      windowsHide: true,
    });
    return;
  }
  try { child.kill('SIGKILL'); } catch { /* best effort */ }
}

async function waitExit(child, timeoutMs = 15_000) {
  if (child.exitCode !== null) return child.exitCode;
  return new Promise((resolve, reject) => {
    const timer = setTimeout(() => reject(new Error('sidecar did not exit in time')), timeoutMs);
    child.once('exit', (code) => {
      clearTimeout(timer);
      resolve(code ?? -1);
    });
  });
}

function sidecarEnv(overrides) {
  return {
    ...process.env,
    // 关键：显式隔离运行数据，绝不触碰仓库 server/data 或用户 ~/.aps
    APS_MODE: 'desktop',
    PYTHONNOUSERSITE: '1',
    PYTHONUTF8: '1',
    ...overrides,
  };
}

function spawnSidecar({ port, token, home, env = {}, cwd = repoRoot, python, capture = true }) {
  const child = spawn(python, ['-m', 'server.sidecar'], {
    cwd,
    env: sidecarEnv({
      APS_API_HOST: LOOPBACK_HOST,
      APS_API_PORT: String(port),
      APS_SIDECAR_NONCE: token,
      APS_PARENT_PID: String(process.pid),
      APS_HOME: home,
      APS_DB_PATH: path.join(home, 'data', 'master.db'),
      APS_KNOWLEDGE_PATH: path.join(home, 'knowledge'),
      APS_APPROVAL_PATH: path.join(home, 'approval'),
      APS_WORLD_PATH: path.join(home, 'data', 'world.json'),
      APS_PROJECTS_PATH: path.join(home, 'sessions', 'workspace.json'),
      APS_UI_DIR: '',
      LLM_PROVIDER: '',      // 离线：禁止任何 LLM 外呼
      PYTHONPATH: repoRoot,  // 支持 cwd 为任意目录（中文/空格路径用例）
      ...env,
    }),
    stdio: ['ignore', 'pipe', 'pipe'],
    windowsHide: true,
  });
  liveChildren.add(child);
  child.nonce = token;
  let stderrTail = '';
  if (capture) {
    child.stderr.on('data', (chunk) => {
      stderrTail = (stderrTail + String(chunk)).slice(-2000);
      child.stderrTail = stderrTail;
    });
  }
  return child;
}

function check(name, ok, detail) {
  results.push({ name, ok: Boolean(ok), detail: detail ?? '' });
  console.log(`[${ok ? 'PASS' : 'FAIL'}] ${name}${detail ? ` — ${detail}` : ''}`);
}

async function removeDir(dir) {
  try { fs.rmSync(dir, { recursive: true, force: true }); } catch { /* best effort */ }
}

// ---------------------------------------------------------------- sections

async function sectionColdStarts(python, home) {
  let passed = 0;
  for (let index = 1; index <= 20; index += 1) {
    const port = await reservePort();
    const child = spawnSidecar({ port, token: nonce(), home, python });
    try {
      await waitForHealth(port, child.nonce, 60_000, child);
      passed += 1;
      console.log(`  cold-start #${index} ok (port=${port} pid=${child.pid})`);
    } catch (error) {
      stopProcessTree(child);
      check(`冷启动 #${index}/20`, false, String(error.message ?? error));
      return;
    } finally {
      stopProcessTree(child);
      await waitExit(child).catch(() => {});
    }
  }
  check('20 次冷启动（进程起/停）', passed === 20, `${passed}/20 成功`);
}

async function sectionPortConflict(python, home) {
  const port = await reservePort();
  // 预占端口并保持监听，制造冲突
  const blocker = net.createServer();
  await new Promise((resolve, reject) => {
    blocker.once('error', reject);
    blocker.listen({ host: LOOPBACK_HOST, port, exclusive: true }, resolve);
  });
  const child = spawnSidecar({ port, token: nonce(), home, python });
  try {
    const code = await waitExit(child, 25_000);
    const tail = (child.stderrTail ?? '').toLowerCase();
    const explicitFailure = code !== 0
      && (tail.includes('10048') || tail.includes('address already in use') || tail.includes('eaddrinuse')
        || tail.includes('bind') || tail.includes('端口') || tail.includes('使用'));
    check('端口冲突（预占后应换端口或显式失败）', explicitFailure,
      `exit=${code}, stderr 尾部: ${(child.stderrTail ?? '').slice(0, 220)}`);
  } catch (error) {
    stopProcessTree(child);
    check('端口冲突（预占后应换端口或显式失败）', false, `未在 25s 内显式失败：${error.message}`);
  } finally {
    blocker.close();
    stopProcessTree(child);
  }
}

async function sectionCrashRestart(python, home) {
  // 崩溃重启：脚本扮演父进程/守护，kill 掉 sidecar 后以新端口+新 nonce 重启
  const port1 = await reservePort();
  const child1 = spawnSidecar({ port: port1, token: nonce(), home, python });
  try {
    await waitForHealth(port1, child1.nonce, 60_000, child1);
  } catch (error) {
    stopProcessTree(child1);
    check('崩溃重启（父进程守护）', false, `首次启动失败：${error.message}`);
    return;
  }
  const firstPid = child1.pid;
  stopProcessTree(child1);
  await waitExit(child1).catch(() => {});
  const port2 = await reservePort();
  const child2 = spawnSidecar({ port: port2, token: nonce(), home, python });
  let restarted = false;
  try {
    await waitForHealth(port2, child2.nonce, 60_000, child2);
    restarted = true;
  } catch (error) {
    check('崩溃重启（父进程守护）', false, `kill 后重启失败：${error.message}`);
    stopProcessTree(child2);
    return;
  } finally {
    stopProcessTree(child2);
  }
  check('崩溃重启（kill 子进程后父进程拉起新实例）', restarted,
    `pid ${firstPid} → 新 pid ${child2.pid}，端口 ${port1} → ${port2}`);

  // 父进程看门狗：父进程退出后 sidecar 应自行退出（守护反向依赖）
  const dummy = spawn(python, ['-c', 'import time; time.sleep(4)'], {
    stdio: 'ignore',
    windowsHide: true,
  });
  const port3 = await reservePort();
  const child3 = spawnSidecar({
    port: port3,
    token: nonce(),
    home,
    python,
    env: { APS_PARENT_PID: String(dummy.pid) },
  });
  try {
    await waitForHealth(port3, child3.nonce, 60_000, child3);
    const code = await waitExit(child3, 20_000);
    check('父进程看门狗（父退出后 sidecar 自退出）', code === 0, `watchdog exit=${code}`);
  } catch (error) {
    stopProcessTree(child3);
    check('父进程看门狗（父退出后 sidecar 自退出）', false, String(error.message ?? error));
  } finally {
    stopProcessTree(dummy);
    stopProcessTree(child3);
  }
}

async function sectionChinesePath(python, home) {
  const base = fs.mkdtempSync(path.join(os.tmpdir(), 'aps 排产 冒烟 测试-'));
  const uiDir = path.join(base, 'ui 界面');
  const dataDir = path.join(base, 'data 数据');
  fs.mkdirSync(uiDir, { recursive: true });
  fs.mkdirSync(dataDir, { recursive: true });
  fs.writeFileSync(path.join(uiDir, 'index.html'), '<!doctype html><title>APS 冒烟</title><h1>中文路径 OK</h1>', 'utf8');
  const port = await reservePort();
  const token = nonce();
  const child = spawnSidecar({
    port,
    token,
    home,
    python,
    cwd: base, // 在含中文+空格的目录启动
    env: {
      APS_UI_DIR: uiDir,
      APS_DB_PATH: path.join(dataDir, 'smoke.db'),
      APS_WORLD_PATH: path.join(dataDir, 'world.json'),
      APS_HOME: path.join(base, 'home'),
      APS_PROJECTS_PATH: path.join(base, 'home', 'sessions', 'workspace.json'),
    },
  });
  try {
    await waitForHealth(port, token, 60_000, child);
    const response = await fetch(`http://${LOOPBACK_HOST}:${port}/`, {
      headers: { [NONCE_HEADER]: token },
      signal: AbortSignal.timeout(8_000),
    });
    const body = await response.text();
    const uiOk = response.status === 200
      && /text\/html/i.test(response.headers.get('content-type') ?? '')
      && /<html|<!doctype/i.test(body);
    // APS_HOME 中文/空格路径被 ensure_aps_home 实际创建（config.json 落盘）
    const homeOk = fs.existsSync(path.join(base, 'home', 'config.json'))
      && fs.existsSync(path.join(base, 'home', 'data'));
    const noTraceback = !(child.stderrTail ?? '').includes('Traceback');
    check('中文/空格路径（cwd + APS_HOME + APS_UI_DIR 全中文空格目录）',
      uiOk && homeOk && noTraceback,
      `ui=${uiOk} home=${homeOk} traceback=${!noTraceback} cwd=${base}（sqlite 懒创建，健康/UI 探针不触发）`);
  } catch (error) {
    check('中文/空格路径（cwd + APS_HOME + APS_UI_DIR 全中文空格目录）', false, String(error.message ?? error));
  } finally {
    stopProcessTree(child);
    await delay(500);
    await removeDir(base);
  }
}

// ---------------------------------------------------------------- main

async function main() {
  const python = resolvePython();
  const home = fs.mkdtempSync(path.join(os.tmpdir(), 'aps-sidecar-smoke-'));
  const startedAt = Date.now();
  console.log(`sidecar smoke: python=${python} platform=${process.platform} home=${home}`);

  try {
    await sectionColdStarts(python, home);
    await sectionPortConflict(python, home);
    await sectionCrashRestart(python, home);
    await sectionChinesePath(python, home);
  } finally {
    for (const child of [...liveChildren]) stopProcessTree(child);
    await delay(500);
    await removeDir(home);
  }

  const failed = results.filter((r) => !r.ok);
  console.log(`\nsidecar smoke summary: ${results.length - failed.length}/${results.length} passed in ${((Date.now() - startedAt) / 1000).toFixed(1)}s`);
  for (const r of results) console.log(`  [${r.ok ? 'PASS' : 'FAIL'}] ${r.name}`);
  if (failed.length > 0) {
    console.error(`sidecar smoke FAILED: ${failed.map((f) => f.name).join('; ')}`);
    process.exitCode = 1;
  }
}

await main();