fix: ensure gateway restarts on install failure via ERR trap

apps/memos-local-openclaw/install.sh

@@ -19,11 +19,11 @@ success() {
 }
 
 warn() {
-  echo -e "${YELLOW}$1${NC}"
+  echo -e "${YELLOW}$1${NC}" >&2
 }
 
 error() {
-  echo -e "${RED}$1${NC}"
+  echo -e "${RED}$1${NC}" >&2
 }
 
 node_major_version() {
@@ -225,6 +225,25 @@ success "Using global OpenClaw CLI, 使用全局 OpenClaw CLI: ${OPENCLAW_BIN}"
 PACKAGE_SPEC="${PLUGIN_PACKAGE}@${PLUGIN_VERSION}"
 EXTENSION_DIR="${OPENCLAW_HOME}/extensions/${PLUGIN_ID}"
 OPENCLAW_CONFIG_PATH="${OPENCLAW_HOME}/openclaw.json"
+TMP_PACK_DIR=""
+GATEWAY_RECOVERY_STATE="inactive"
+
+cleanup_on_exit() {
+  if [[ "${GATEWAY_RECOVERY_STATE:-inactive}" == "needs_recovery" ]]; then
+    local recovery_out=""
+    if ! recovery_out="$("${OPENCLAW_BIN}" gateway start 2>&1)"; then
+      warn "Gateway recovery start failed; OpenClaw Gateway may still be stopped. Gateway 恢复启动失败，服务可能仍处于停止状态。"
+      if [[ -n "${recovery_out}" ]]; then
+        printf '%s\n' "${recovery_out}" | sed 's/^/       /' >&2
+      fi
+    fi
+  fi
+  if [[ -n "${TMP_PACK_DIR:-}" ]]; then
+    rm -rf "${TMP_PACK_DIR}"
+  fi
+}
+
+trap cleanup_on_exit EXIT
 
 update_openclaw_config() {
   info "Update OpenClaw config, 更新 OpenClaw 配置..."
@@ -317,6 +336,7 @@ NODE
 
 info "Stop OpenClaw Gateway, 停止 OpenClaw Gateway..."
 "${OPENCLAW_BIN}" gateway stop >/dev/null 2>&1 || true
+GATEWAY_RECOVERY_STATE="needs_recovery"
 
 if command -v lsof >/dev/null 2>&1; then
   PIDS="$(lsof -i :"${PORT}" -t 2>/dev/null || true)"
@@ -334,7 +354,6 @@ fi
 
 info "Install plugin ${PACKAGE_SPEC}, 安装插件 ${PACKAGE_SPEC}..."
 TMP_PACK_DIR="$(mktemp -d)"
-trap 'rm -rf "${TMP_PACK_DIR}"' EXIT
 
 if [[ -f "${PLUGIN_VERSION}" ]]; then
   info "Using local tarball, 使用本地包: ${PLUGIN_VERSION}"
@@ -404,8 +423,15 @@ update_openclaw_config
 info "Install OpenClaw Gateway service, 安装 OpenClaw Gateway 服务..."
 "${OPENCLAW_BIN}" gateway install --port "${PORT}" --force 2>&1 || true
 
-success "Start OpenClaw Gateway service, 启动 OpenClaw Gateway 服务..."
-"${OPENCLAW_BIN}" gateway start 2>&1
+info "Starting OpenClaw Gateway service, 正在启动 OpenClaw Gateway 服务..."
+if ! "${OPENCLAW_BIN}" gateway start 2>&1; then
+  # Do not repeat the same failed final start from the EXIT cleanup.
+  GATEWAY_RECOVERY_STATE="final_failed"
+  error "Failed to start OpenClaw Gateway, OpenClaw Gateway 启动失败"
+  exit 1
+fi
+GATEWAY_RECOVERY_STATE="inactive"
+success "OpenClaw Gateway started, OpenClaw Gateway 已启动"
 
 info "Starting Memory Viewer, 正在启动记忆面板..."
 VIEWER_URL="http://127.0.0.1:18799"

apps/memos-local-plugin/install.sh

@@ -264,10 +264,28 @@ BUILT_TARBALL=""
 STAGE_DIR=""
 SOURCE_KIND=""   # "path" for a local file, "npm" otherwise
 SOURCE_SPEC=""
+GATEWAY_RECOVERY_BIN=""
+GATEWAY_RECOVERY_STATE="inactive"
+
+cleanup_on_exit() {
+  if [[ "${GATEWAY_RECOVERY_STATE:-inactive}" == "needs_recovery" && -n "${GATEWAY_RECOVERY_BIN:-}" ]]; then
+    local recovery_out=""
+    if ! recovery_out="$("${GATEWAY_RECOVERY_BIN}" gateway start 2>&1)"; then
+      warn "OpenClaw gateway recovery failed; the gateway may still be stopped."
+      if [[ -n "${recovery_out}" ]]; then
+        printf '%s\n' "${recovery_out}" | sed 's/^/       /' >&2
+      fi
+    fi
+  fi
+  if [[ -n "${STAGE_DIR:-}" ]]; then
+    rm -rf "${STAGE_DIR}"
+  fi
+}
+
+trap cleanup_on_exit EXIT
 
 resolve_tarball() {
   STAGE_DIR="$(mktemp -d)"
-  trap 'rm -rf "${STAGE_DIR}"' EXIT
 
   if [[ -n "${VERSION_ARG}" && -f "${VERSION_ARG}" ]]; then
     BUILT_TARBALL="$(cd "$(dirname "${VERSION_ARG}")" && pwd)/$(basename "${VERSION_ARG}")"
@@ -429,11 +447,17 @@ install_openclaw() {
   mkdir -p "${HOME}/.openclaw"
 
   local oc_bin=""
+  # These remain global because the EXIT trap reads them after this function returns.
+  # The top-level agent dispatch invokes this installer at most once.
+  GATEWAY_RECOVERY_BIN=""
+  GATEWAY_RECOVERY_STATE="inactive"
   if oc_bin="$(find_openclaw_cli)"; then
     step "Stopping OpenClaw gateway"
     "${oc_bin}" gateway stop >/dev/null 2>&1 || true
     sleep 1
     success "Gateway stopped"
+    GATEWAY_RECOVERY_BIN="${oc_bin}"
+    GATEWAY_RECOVERY_STATE="needs_recovery"
   fi
 
   deploy_tarball_to_prefix "${prefix}"
@@ -608,6 +632,8 @@ NODE
        || (command -v lsof >/dev/null 2>&1 && lsof -i ":18789" -t >/dev/null 2>&1); then
       success "OpenClaw gateway already running"
     else
+      # Do not repeat the same failed final start from the EXIT cleanup.
+      GATEWAY_RECOVERY_STATE="final_failed"
       error "openclaw gateway start failed:"
       echo "${start_out}" | sed 's/^/       /' >&2
       warn "Inspect ~/.openclaw/logs/gateway.err.log for the full reason."
@@ -616,6 +642,10 @@ NODE
   else
     success "OpenClaw gateway started"
   fi
+  # The service started (or was already running), so the viewer fallback must not
+  # trigger another service start from the EXIT trap.
+  GATEWAY_RECOVERY_STATE="inactive"
+  GATEWAY_RECOVERY_BIN=""
 
   step "Waiting for Memory Viewer"
   if wait_for_viewer "${OPENCLAW_PORT}"; then

apps/memos-local-plugin/tests/unit/install-gateway-recovery.test.ts

@@ -0,0 +1,211 @@
+import { spawnSync } from "node:child_process";
+import {
+  chmodSync,
+  existsSync,
+  mkdirSync,
+  mkdtempSync,
+  readFileSync,
+  readdirSync,
+  rmSync,
+  writeFileSync,
+} from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { fileURLToPath } from "node:url";
+
+import { describe, expect, it } from "vitest";
+
+const pluginInstaller = fileURLToPath(new URL("../../install.sh", import.meta.url));
+const legacyInstaller = fileURLToPath(
+  new URL("../../../memos-local-openclaw/install.sh", import.meta.url),
+);
+
+interface InstallerHarness {
+  root: string;
+  home: string;
+  bin: string;
+  temp: string;
+  gatewayLog: string;
+}
+
+function writeExecutable(path: string, body: string): void {
+  writeFileSync(path, `#!/usr/bin/env bash\nset -u\n${body}\n`, "utf8");
+  chmodSync(path, 0o755);
+}
+
+function createHarness(): InstallerHarness {
+  const root = mkdtempSync(join(tmpdir(), "memos-installer-recovery-"));
+  const home = join(root, "home");
+  const bin = join(root, "bin");
+  const temp = join(root, "tmp");
+  const gatewayLog = join(root, "gateway.log");
+  mkdirSync(join(home, ".openclaw"), { recursive: true });
+  mkdirSync(bin);
+  mkdirSync(temp);
+
+  writeExecutable(
+    join(bin, "node"),
+    'if [[ "${1:-}" == "-v" ]]; then printf "v22.0.0\\n"; fi\nexit 0',
+  );
+  writeExecutable(join(bin, "npx"), "exit 0");
+  writeExecutable(
+    join(bin, "npm"),
+    `case "\${1:-}" in
+  pack) exit "\${FAKE_NPM_PACK_EXIT:-0}" ;;
+  install) mkdir -p node_modules/better-sqlite3; exit 0 ;;
+  rebuild) exit 0 ;;
+  *) exit 0 ;;
+esac`,
+  );
+  writeExecutable(
+    join(bin, "openclaw"),
+    `printf '%s\\n' "$*" >> "\${FAKE_GATEWAY_LOG:?}"
+if [[ "$*" == "gateway start" ]]; then
+  if [[ "\${FAKE_GATEWAY_START_EXIT:-0}" != "0" ]]; then
+    printf 'fake gateway start failure\\n' >&2
+  fi
+  exit "\${FAKE_GATEWAY_START_EXIT:-0}"
+fi
+exit 0`,
+  );
+  writeExecutable(join(bin, "sleep"), "exit 0");
+  writeExecutable(join(bin, "lsof"), "exit 1");
+  writeExecutable(join(bin, "curl"), 'exit "${FAKE_CURL_EXIT:-0}"');
+
+  return { root, home, bin, temp, gatewayLog };
+}
+
+function runInstaller(
+  harness: InstallerHarness,
+  script: string,
+  args: string[],
+  extraEnv: NodeJS.ProcessEnv = {},
+) {
+  return spawnSync("bash", [script, ...args], {
+    cwd: harness.root,
+    encoding: "utf8",
+    timeout: 30_000,
+    env: {
+      ...process.env,
+      ...extraEnv,
+      HOME: harness.home,
+      TMPDIR: harness.temp,
+      PATH: `${harness.bin}:${process.env.PATH ?? ""}`,
+      FAKE_GATEWAY_LOG: harness.gatewayLog,
+    },
+  });
+}
+
+function gatewayCalls(harness: InstallerHarness): string[] {
+  if (!existsSync(harness.gatewayLog)) return [];
+  return readFileSync(harness.gatewayLog, "utf8").trim().split("\n").filter(Boolean);
+}
+
+function expectTempCleaned(harness: InstallerHarness): void {
+  expect(readdirSync(harness.temp)).toEqual([]);
+}
+
+describe.skipIf(process.platform === "win32")("installer gateway recovery", () => {
+  it("restarts the gateway when the unified installer cannot extract the package", () => {
+    const harness = createHarness();
+    try {
+      const brokenTarball = join(harness.root, "broken.tgz");
+      writeFileSync(brokenTarball, "not a tarball", "utf8");
+
+      const result = runInstaller(harness, pluginInstaller, [
+        "--agent",
+        "openclaw",
+        "--version",
+        brokenTarball,
+      ]);
+
+      expect(result.status).not.toBe(0);
+      expect(gatewayCalls(harness)).toEqual(["gateway stop", "gateway start"]);
+      expectTempCleaned(harness);
+    } finally {
+      rmSync(harness.root, { recursive: true, force: true });
+    }
+  });
+
+  it("warns when legacy-installer gateway recovery also fails", () => {
+    const harness = createHarness();
+    try {
+      const result = runInstaller(
+        harness,
+        legacyInstaller,
+        ["--version", "missing-test-version", "--openclaw-home", join(harness.home, ".openclaw")],
+        { FAKE_GATEWAY_START_EXIT: "17", FAKE_NPM_PACK_EXIT: "23" },
+      );
+
+      expect(result.status).not.toBe(0);
+      expect(gatewayCalls(harness)).toEqual(["gateway stop", "gateway start"]);
+      expect(result.stderr).toContain("Gateway recovery start failed");
+      expect(result.stderr).toContain("fake gateway start failure");
+      expectTempCleaned(harness);
+    } finally {
+      rmSync(harness.root, { recursive: true, force: true });
+    }
+  });
+
+  it("does not restart the gateway after the unified installer enters foreground fallback", () => {
+    const harness = createHarness();
+    try {
+      const packageRoot = join(harness.root, "package");
+      const runtimeDir = join(packageRoot, "dist", "adapters", "openclaw");
+      const tarball = join(harness.root, "plugin.tgz");
+      mkdirSync(runtimeDir, { recursive: true });
+      writeFileSync(join(packageRoot, "package.json"), '{"name":"test-plugin"}\n', "utf8");
+      writeFileSync(join(runtimeDir, "index.js"), "export {};\n", "utf8");
+      const tar = spawnSync("tar", ["-czf", tarball, "-C", harness.root, "package"], {
+        encoding: "utf8",
+      });
+      expect(tar.status, tar.stderr).toBe(0);
+
+      const result = runInstaller(
+        harness,
+        pluginInstaller,
+        ["--agent", "openclaw", "--version", tarball],
+        { FAKE_CURL_EXIT: "1" },
+      );
+
+      expect(result.status).not.toBe(0);
+      expect(gatewayCalls(harness)).toEqual(["gateway stop", "gateway start", "gateway"]);
+      expectTempCleaned(harness);
+    } finally {
+      rmSync(harness.root, { recursive: true, force: true });
+    }
+  });
+
+  it("does not retry a failed final gateway start from the exit trap", () => {
+    const harness = createHarness();
+    try {
+      const packageRoot = join(harness.root, "package");
+      const tarball = join(harness.root, "plugin.tgz");
+      mkdirSync(packageRoot);
+      writeFileSync(join(packageRoot, "package.json"), '{"name":"test-plugin"}\n', "utf8");
+      const tar = spawnSync("tar", ["-czf", tarball, "-C", harness.root, "package"], {
+        encoding: "utf8",
+      });
+      expect(tar.status, tar.stderr).toBe(0);
+
+      const result = runInstaller(
+        harness,
+        legacyInstaller,
+        ["--version", tarball, "--openclaw-home", join(harness.home, ".openclaw")],
+        { FAKE_GATEWAY_START_EXIT: "17" },
+      );
+
+      expect(result.status).not.toBe(0);
+      const calls = gatewayCalls(harness);
+      expect(calls[0]).toBe("gateway stop");
+      expect(calls.filter((call) => call === "gateway start")).toHaveLength(1);
+      expect(result.stdout).toContain("Starting OpenClaw Gateway service");
+      expect(result.stdout).not.toContain("OpenClaw Gateway started");
+      expect(result.stdout).not.toContain("Start OpenClaw Gateway service");
+      expect(result.stderr).toContain("Failed to start OpenClaw Gateway");
+      expectTempCleaned(harness);
+    } finally {
+      rmSync(harness.root, { recursive: true, force: true });
+    }
+  });
+});