fix(e2e): clean telegram credential timeouts

scripts/e2e/telegram-user-credential-io.ts

@@ -21,6 +21,33 @@ type RunCommandOptions = {
 const DEFAULT_OUTPUT_LIMIT = 128 * 1024;
 const DEFAULT_FETCH_BODY_LIMIT = 1024 * 1024;
 const KILL_GRACE_MS = 5_000;
+const SIGNAL_EXIT_CODES = {
+  SIGHUP: 129,
+  SIGINT: 130,
+  SIGTERM: 143,
+};
+const ACTIVE_CHILD_TREE_KILLERS = new Set<(signal: NodeJS.Signals) => void>();
+let forwardedSignalExitCode: number | undefined;
+let forwardedSignalForceKillTimer: NodeJS.Timeout | undefined;
+
+for (const signal of Object.keys(SIGNAL_EXIT_CODES) as Array<keyof typeof SIGNAL_EXIT_CODES>) {
+  process.on(signal, () => {
+    forwardedSignalExitCode ??= SIGNAL_EXIT_CODES[signal];
+    if (ACTIVE_CHILD_TREE_KILLERS.size === 0) {
+      process.exit(forwardedSignalExitCode);
+    }
+    const activeKillers = Array.from(ACTIVE_CHILD_TREE_KILLERS);
+    for (const killChildTree of activeKillers) {
+      killChildTree(signal);
+    }
+    forwardedSignalForceKillTimer ??= setTimeout(() => {
+      for (const killChildTree of activeKillers) {
+        killChildTree("SIGKILL");
+      }
+      process.exit(forwardedSignalExitCode);
+    }, KILL_GRACE_MS);
+  });
+}
 
 function timeoutError(message: string) {
   return Object.assign(new Error(message), { code: "ETIMEDOUT" });
@@ -70,15 +97,19 @@ export function runCommand(
   options: RunCommandOptions,
 ) {
   return new Promise<void>((resolve, reject) => {
+    const useProcessGroup = process.platform !== "win32";
     const child = spawn(command, args, {
       cwd,
       stdio: ["ignore", "pipe", "pipe"],
+      detached: useProcessGroup,
     });
+    const activeChildTree = registerActiveChildProcessTree(child);
     const outputLimit = options.outputLimit ?? DEFAULT_OUTPUT_LIMIT;
     let stdout = "";
     let stderr = "";
     let settled = false;
     let killTimer: NodeJS.Timeout | undefined;
+    let pendingTimeoutReject: (() => void) | undefined;
     let timedOutError: Error | undefined;
     const timeoutMs = Math.max(1, options.timeoutMs);
     const timeoutKillGraceMs = Math.max(0, options.timeoutKillGraceMs ?? KILL_GRACE_MS);
@@ -94,6 +125,7 @@ export function runCommand(
       }
       settled = true;
       clearTimers();
+      activeChildTree.unregister();
       reject(error);
     };
     const timeout: NodeJS.Timeout = setTimeout(() => {
@@ -103,11 +135,12 @@ export function runCommand(
       timedOutError = timeoutError(
         `${command} ${args.join(" ")} timed out after ${timeoutMs}ms\n${stdout}${stderr}`,
       );
-      child.kill("SIGTERM");
+      activeChildTree.killChildTree("SIGTERM");
       killTimer = setTimeout(() => {
-        child.kill("SIGKILL");
+        killTimer = undefined;
+        activeChildTree.killChildTree("SIGKILL");
+        pendingTimeoutReject?.();
       }, timeoutKillGraceMs);
-      killTimer.unref?.();
     }, timeoutMs);
     timeout.unref?.();
 
@@ -122,8 +155,26 @@ export function runCommand(
       if (settled) {
         return;
       }
+      if (forwardedSignalExitCode !== undefined) {
+        activeChildTree.unregister();
+        return;
+      }
+      if (timedOutError && killTimer && childProcessTreeMayStillExist(child)) {
+        const error = timedOutError;
+        pendingTimeoutReject = () => {
+          if (settled) {
+            return;
+          }
+          settled = true;
+          clearTimers();
+          activeChildTree.unregister();
+          reject(error);
+        };
+        return;
+      }
       settled = true;
       clearTimers();
+      activeChildTree.unregister();
       if (timedOutError) {
         reject(timedOutError);
         return;
@@ -138,6 +189,41 @@ export function runCommand(
   });
 }
 
+function signalChildProcessTree(child: ReturnType<typeof spawn>, signal: NodeJS.Signals) {
+  if (process.platform !== "win32" && child.pid) {
+    try {
+      process.kill(-child.pid, signal);
+      return;
+    } catch {
+      // The process group can disappear between timeout and cleanup.
+    }
+  }
+  child.kill(signal);
+}
+
+function childProcessTreeMayStillExist(child: ReturnType<typeof spawn>) {
+  if (process.platform === "win32" || !child.pid) {
+    return false;
+  }
+  try {
+    process.kill(-child.pid, 0);
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+function registerActiveChildProcessTree(child: ReturnType<typeof spawn>) {
+  const killChildTree = (signal: NodeJS.Signals) => signalChildProcessTree(child, signal);
+  ACTIVE_CHILD_TREE_KILLERS.add(killChildTree);
+  return {
+    killChildTree,
+    unregister: () => {
+      ACTIVE_CHILD_TREE_KILLERS.delete(killChildTree);
+    },
+  };
+}
+
 export async function fetchJsonWithTimeout(params: FetchJsonParams) {
   const timeoutMs = Math.max(1, params.timeoutMs);
   const maxBodyBytes = resolveFetchBodyLimit(params.maxBodyBytes);

test/scripts/telegram-user-credential.test.ts

@@ -1,4 +1,4 @@
-import { existsSync, mkdtempSync, rmSync, writeFileSync } from "node:fs";
+import { existsSync, mkdtempSync, readFileSync, rmSync, writeFileSync } from "node:fs";
 import { readFile } from "node:fs/promises";
 import { tmpdir } from "node:os";
 import path, { win32 } from "node:path";
@@ -18,6 +18,41 @@ function makeTempDir(prefix: string) {
   return dir;
 }
 
+function isProcessAlive(pid: number): boolean {
+  try {
+    process.kill(pid, 0);
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+async function waitForFile(filePath: string, timeoutMs: number): Promise<void> {
+  const deadline = Date.now() + timeoutMs;
+  while (Date.now() < deadline) {
+    if (existsSync(filePath)) {
+      return;
+    }
+    await new Promise((resolve) => {
+      setTimeout(resolve, 25);
+    });
+  }
+  throw new Error(`timeout waiting for ${filePath}`);
+}
+
+async function waitForDead(pid: number, timeoutMs: number): Promise<void> {
+  const deadline = Date.now() + timeoutMs;
+  while (Date.now() < deadline) {
+    if (!isProcessAlive(pid)) {
+      return;
+    }
+    await new Promise((resolve) => {
+      setTimeout(resolve, 25);
+    });
+  }
+  throw new Error(`process still alive: ${pid}`);
+}
+
 afterEach(() => {
   for (const dir of tempDirs.splice(0)) {
     rmSync(dir, { force: true, recursive: true });
@@ -112,6 +147,43 @@ setInterval(() => {}, 1000);
     },
   );
 
+  it.runIf(process.platform !== "win32")(
+    "kills timed-out child process groups",
+    async () => {
+      const dir = makeTempDir("openclaw-telegram-credential-tree-timeout-");
+      const childPidPath = path.join(dir, "child.pid");
+      let childPid: number | undefined;
+
+      try {
+        const childScript = "process.on('SIGTERM', () => {}); setInterval(() => {}, 1000);";
+        const parentScript = [
+          "const { spawn } = require('node:child_process');",
+          "const fs = require('node:fs');",
+          `const child = spawn(process.execPath, ['-e', ${JSON.stringify(childScript)}], { stdio: 'ignore' });`,
+          `fs.writeFileSync(${JSON.stringify(childPidPath)}, String(child.pid));`,
+          "setInterval(() => {}, 1000);",
+        ].join("");
+
+        const runPromise = runCommand(process.execPath, ["-e", parentScript], dir, {
+          timeoutKillGraceMs: 25,
+          timeoutMs: 100,
+        });
+        await waitForFile(childPidPath, 2_000);
+        childPid = Number.parseInt(readFileSync(childPidPath, "utf8"), 10);
+
+        await expect(runPromise).rejects.toMatchObject({
+          code: "ETIMEDOUT",
+          message: expect.stringContaining("timed out after 100ms"),
+        });
+        await waitForDead(childPid, 2_000);
+      } finally {
+        if (childPid !== undefined && isProcessAlive(childPid)) {
+          process.kill(childPid, "SIGKILL");
+        }
+      }
+    },
+  );
+
   it("aborts broker fetches that never return", async () => {
     let signal: AbortSignal | undefined;
     await expect(