fix(gateway): remove watch-mode build/start race (#18782)

docs/help/debugging.md

@@ -34,13 +34,13 @@ Examples:
 For fast iteration, run the gateway under the file watcher:
 
 ```bash
-pnpm gateway:watch --force
+pnpm gateway:watch
 ```
 
 This maps to:
 
 ```bash
-tsx watch src/entry.ts gateway --force
+node --watch-path src --watch-path tsconfig.json --watch-path package.json --watch-preserve-output scripts/run-node.mjs gateway --force
 ```
 
 Add any gateway CLI flags after `gateway:watch` and they will be passed through
@@ -113,13 +113,13 @@ This is the best way to see whether reasoning is arriving as plain text deltas
 Enable it via CLI:
 
 ```bash
-pnpm gateway:watch --force --raw-stream
+pnpm gateway:watch --raw-stream
 ```
 
 Optional path override:
 
 ```bash
-pnpm gateway:watch --force --raw-stream --raw-stream-path ~/.openclaw/logs/raw-stream.jsonl
+pnpm gateway:watch --raw-stream --raw-stream-path ~/.openclaw/logs/raw-stream.jsonl
 ```
 
 Equivalent env vars:

scripts/run-node.mjs

@@ -8,7 +8,7 @@ import { pathToFileURL } from "node:url";
 const compiler = "tsdown";
 const compilerArgs = ["exec", compiler, "--no-clean"];
 
-const gitWatchedPaths = ["src", "tsconfig.json", "package.json"];
+export const runNodeWatchedPaths = ["src", "tsconfig.json", "package.json"];
 
 const statMtime = (filePath, fsImpl = fs) => {
   try {
@@ -91,7 +91,7 @@ const resolveGitHead = (deps) => {
 
 const hasDirtySourceTree = (deps) => {
   const output = runGit(
-    ["status", "--porcelain", "--untracked-files=normal", "--", ...gitWatchedPaths],
+    ["status", "--porcelain", "--untracked-files=normal", "--", ...runNodeWatchedPaths],
     deps,
   );
   if (output === null) {

scripts/watch-node.mjs

@@ -1,65 +1,92 @@
 #!/usr/bin/env node
-import { spawn, spawnSync } from "node:child_process";
+import { spawn } from "node:child_process";
 import process from "node:process";
+import { pathToFileURL } from "node:url";
+import { runNodeWatchedPaths } from "./run-node.mjs";
 
-const args = process.argv.slice(2);
-const env = { ...process.env };
-const cwd = process.cwd();
-const compiler = "tsdown";
-const watchSession = `${Date.now()}-${process.pid}`;
-env.OPENCLAW_WATCH_MODE = "1";
-env.OPENCLAW_WATCH_SESSION = watchSession;
-if (args.length > 0) {
-  env.OPENCLAW_WATCH_COMMAND = args.join(" ");
+const WATCH_NODE_RUNNER = "scripts/run-node.mjs";
+
+const buildWatchArgs = (args) => [
+  ...runNodeWatchedPaths.flatMap((watchPath) => ["--watch-path", watchPath]),
+  "--watch-preserve-output",
+  WATCH_NODE_RUNNER,
+  ...args,
+];
+
+export async function runWatchMain(params = {}) {
+  const deps = {
+    spawn: params.spawn ?? spawn,
+    process: params.process ?? process,
+    cwd: params.cwd ?? process.cwd(),
+    args: params.args ?? process.argv.slice(2),
+    env: params.env ? { ...params.env } : { ...process.env },
+    now: params.now ?? Date.now,
+  };
+
+  const childEnv = { ...deps.env };
+  const watchSession = `${deps.now()}-${deps.process.pid}`;
+  childEnv.OPENCLAW_WATCH_MODE = "1";
+  childEnv.OPENCLAW_WATCH_SESSION = watchSession;
+  if (deps.args.length > 0) {
+    childEnv.OPENCLAW_WATCH_COMMAND = deps.args.join(" ");
+  }
+
+  const watchProcess = deps.spawn(deps.process.execPath, buildWatchArgs(deps.args), {
+    cwd: deps.cwd,
+    env: childEnv,
+    stdio: "inherit",
+  });
+
+  let settled = false;
+  let onSigInt;
+  let onSigTerm;
+
+  const settle = (resolve, code) => {
+    if (settled) {
+      return;
+    }
+    settled = true;
+    if (onSigInt) {
+      deps.process.off("SIGINT", onSigInt);
+    }
+    if (onSigTerm) {
+      deps.process.off("SIGTERM", onSigTerm);
+    }
+    resolve(code);
+  };
+
+  return await new Promise((resolve) => {
+    onSigInt = () => {
+      if (typeof watchProcess.kill === "function") {
+        watchProcess.kill("SIGTERM");
+      }
+      settle(resolve, 130);
+    };
+    onSigTerm = () => {
+      if (typeof watchProcess.kill === "function") {
+        watchProcess.kill("SIGTERM");
+      }
+      settle(resolve, 143);
+    };
+
+    deps.process.on("SIGINT", onSigInt);
+    deps.process.on("SIGTERM", onSigTerm);
+
+    watchProcess.on("exit", (code, signal) => {
+      if (signal) {
+        settle(resolve, 1);
+        return;
+      }
+      settle(resolve, code ?? 1);
+    });
+  });
 }
 
-const initialBuild = spawnSync("pnpm", ["exec", compiler], {
-  cwd,
-  env,
-  stdio: "inherit",
-});
-
-if (initialBuild.status !== 0) {
-  process.exit(initialBuild.status ?? 1);
+if (import.meta.url === pathToFileURL(process.argv[1] ?? "").href) {
+  void runWatchMain()
+    .then((code) => process.exit(code))
+    .catch((err) => {
+      console.error(err);
+      process.exit(1);
+    });
 }
-
-const compilerProcess = spawn("pnpm", ["exec", compiler, "--watch"], {
-  cwd,
-  env,
-  stdio: "inherit",
-});
-
-const nodeProcess = spawn(process.execPath, ["--watch", "openclaw.mjs", ...args], {
-  cwd,
-  env,
-  stdio: "inherit",
-});
-
-let exiting = false;
-
-function cleanup(code = 0) {
-  if (exiting) {
-    return;
-  }
-  exiting = true;
-  nodeProcess.kill("SIGTERM");
-  compilerProcess.kill("SIGTERM");
-  process.exit(code);
-}
-
-process.on("SIGINT", () => cleanup(130));
-process.on("SIGTERM", () => cleanup(143));
-
-compilerProcess.on("exit", (code) => {
-  if (exiting) {
-    return;
-  }
-  cleanup(code ?? 1);
-});
-
-nodeProcess.on("exit", (code, signal) => {
-  if (signal || exiting) {
-    return;
-  }
-  cleanup(code ?? 1);
-});

src/infra/watch-node.test.ts

@@ -0,0 +1,77 @@
+import { EventEmitter } from "node:events";
+import { describe, expect, it, vi } from "vitest";
+import { runNodeWatchedPaths } from "../../scripts/run-node.mjs";
+import { runWatchMain } from "../../scripts/watch-node.mjs";
+
+const createFakeProcess = () =>
+  Object.assign(new EventEmitter(), {
+    pid: 4242,
+    execPath: "/usr/local/bin/node",
+  }) as unknown as NodeJS.Process;
+
+describe("watch-node script", () => {
+  it("wires node watch to run-node with watched source/config paths", async () => {
+    const child = Object.assign(new EventEmitter(), {
+      kill: vi.fn(),
+    });
+    const spawn = vi.fn(() => child);
+    const fakeProcess = createFakeProcess();
+
+    const runPromise = runWatchMain({
+      args: ["gateway", "--force"],
+      cwd: "/tmp/openclaw",
+      env: { PATH: "/usr/bin" },
+      now: () => 1700000000000,
+      process: fakeProcess,
+      spawn,
+    });
+
+    queueMicrotask(() => child.emit("exit", 0, null));
+    const exitCode = await runPromise;
+
+    expect(exitCode).toBe(0);
+    expect(spawn).toHaveBeenCalledTimes(1);
+    expect(spawn).toHaveBeenCalledWith(
+      "/usr/local/bin/node",
+      [
+        ...runNodeWatchedPaths.flatMap((watchPath) => ["--watch-path", watchPath]),
+        "--watch-preserve-output",
+        "scripts/run-node.mjs",
+        "gateway",
+        "--force",
+      ],
+      expect.objectContaining({
+        cwd: "/tmp/openclaw",
+        stdio: "inherit",
+        env: expect.objectContaining({
+          PATH: "/usr/bin",
+          OPENCLAW_WATCH_MODE: "1",
+          OPENCLAW_WATCH_SESSION: "1700000000000-4242",
+          OPENCLAW_WATCH_COMMAND: "gateway --force",
+        }),
+      }),
+    );
+  });
+
+  it("terminates child on SIGINT and returns shell interrupt code", async () => {
+    const child = Object.assign(new EventEmitter(), {
+      kill: vi.fn(),
+    });
+    const spawn = vi.fn(() => child);
+    const fakeProcess = createFakeProcess();
+
+    const runPromise = runWatchMain({
+      args: ["gateway", "--force"],
+      process: fakeProcess,
+      spawn,
+    });
+
+    fakeProcess.emit("SIGINT");
+    const exitCode = await runPromise;
+
+    expect(exitCode).toBe(130);
+    expect(child.kill).toHaveBeenCalledWith("SIGTERM");
+    expect(fakeProcess.listenerCount("SIGINT")).toBe(0);
+    expect(fakeProcess.listenerCount("SIGTERM")).toBe(0);
+  });
+});