Make zombie children clean themselves up

The parent wds process can crash sometimes when watching too many files, or due to memory pressure on the system, or really any number of reasons. When this happens, the children become zombies and live forever. If the parent crashes though, they'll never be reloaded, and signals from the terminal won't be sent along to them, so they just accumulate and suck up memory on developer machines! Yuck!

This adds active monitoring to the child wds processes so that they notice if the parent is gone, and kill themselves right away if so. This isn't 100% reliable, but this will at least reap most of em so we stop accumulating so much cruft.

.github/workflows/test.yml

@@ -13,7 +13,7 @@ jobs:
       - uses: actions/checkout@v2
       - uses: ./.github/actions/setup-test-env
       - run: pnpm build
-      - run: test/test.sh
+      - run: integration-test/test.sh
       - run: pnpm jest
 
   lint:

integration-test/parent-crash/package.json

@@ -0,0 +1,10 @@
+{
+  "name": "parent-crash",
+  "version": "1.0.0",
+  "description": "",
+  "main": "index.js",
+  "scripts": {
+    "test": "echo \"Error: no test specified\" && exit 1"
+  },
+  "license": "ISC"
+}

integration-test/parent-crash/run.ts

@@ -0,0 +1,4 @@
+process.stderr.write("child started\n")
+setInterval(() => {
+  process.stderr.write("child still alive\n")
+}, 200)

integration-test/parent-crash/test.js

@@ -0,0 +1,50 @@
+#!/usr/bin/env zx
+const assert = require("assert");
+const { exec } = require("child_process");
+
+async function getChildPids(parentPid, callback) {
+  const result = await $`pgrep -P ${parentPid}`;
+  return result.stdout
+    .split("\n")
+    .filter((pid) => pid)
+    .map((pid) => parseInt(pid, 10));
+}
+
+function processIsRunning(pid) {
+  try {
+    process.kill(pid, 0);
+    return true;
+  } catch (e) {
+    return false;
+  }
+}
+
+const { setTimeout } = require("timers/promises");
+
+const main = async () => {
+  // launch the wds process
+  const parent = $`${__dirname}/../../pkg/wds.bin.js --watch ${__dirname}/run.ts`.nothrow();
+
+  // wait for the wds process to start
+  await setTimeout(500);
+
+  // get the pid of the child process that the parent wds supervisor will have started
+  const pids = await getChildPids(parent.child.pid);
+  assert(pids.length > 0, "no child pids found for supervisor process");
+
+  // SIGKILL the parent process, as if it OOMed or something like that to simulate a zombie child
+  console.log(`killing parent (${parent.child.pid})`);
+  await parent.kill(9);
+  assert.ok(processIsRunning(pids[0]), "test is broken, child process is not running immediately after parent is dead");
+
+  // ensure the children are dead too after their monitoring delay
+  await setTimeout(3000);
+
+  for (const pid of pids) {
+    assert.ok(!processIsRunning(pid), `child process ${pid} is still running after parent has been killed`);
+  }
+
+  await parent;
+};
+
+void main();

integration-test/reload/run-scratch.ts

@@ -0,0 +1,10 @@
+const http = require("http");
+
+const requestListener = function (req, res) {
+  res.writeHead(200);
+  res.end("Hey, Pluto!");
+};
+
+const server = http.createServer(requestListener);
+server.listen(8080);
+console.warn("Listening on 8080");

test/test.sh → integration-test/test.sh

@@ -21,4 +21,9 @@ echo
 echo "::group::Reload test ${args}"
 bash $DIR/reload/test.sh $args
 echo "::endgroup::"
+echo
+
+echo "::group::ParentCrash test ${args}"
+node_modules/.bin/zx $DIR/parent-crash/test.js $args
+echo "::endgroup::"
 echo

jest.config.js

@@ -3,4 +3,5 @@ module.exports = {
     "^.+\\.[jt]sx?$": ["@swc/jest", { sourceMaps: "inline" }],
   },
   testEnvironment: "node",
+  testMatch: ["<rootDir>/spec/**/?(*.)+(spec|test).[tj]s?(x)"],
 };

package.json

@@ -61,7 +61,8 @@
     "gitpkg": "^1.0.0-beta.2",
     "jest": "^27.4.7",
     "prettier": "^2.8.8",
-    "typescript": "^5.1.3"
+    "typescript": "^5.1.3",
+    "zx": "^7.2.3"
   },
   "packageManager": "pnpm@8.12.1+sha256.28ca61ece5a496148b73fabc9afb820f9c3fec4f55f04ce45a2cea0a5219f2e1"
 }