test: fix cluster-disconnect-handles flakiness

Sometimes the test was timing out because the worker process remained
stuck in the breakpoint and didn't exit. This could happen because the
continue was sent before the breakpoint was set. If that's the case,
with this change, a new continue command is sent so the worker process
can end.

PR-URL: #4009
Reviewed-By: Ben Noordhuis <[email protected]>

test/parallel/test-cluster-disconnect-handles.js

@@ -33,6 +33,7 @@ if (cluster.isMaster) {
  worker.on('message', common.mustCall(message => {
  assert.strictEqual(Array.isArray(message), true);
  assert.strictEqual(message[0], 'listening');
+ let continueRecv = false;
  const address = message[1];
  const host = address.address;
  const debugClient = net.connect({ host, port: common.PORT });
@@ -41,13 +42,25 @@ if (cluster.isMaster) {
  debugClient.on('data', data => protocol.execute(data));
  debugClient.once('connect', common.mustCall(() => {
  protocol.onResponse = common.mustCall(res => {
- protocol.onResponse = () => {};
+ protocol.onResponse = (res) => {
+ // It can happen that the first continue was sent before the break
+ // event was received. If that's the case, send also a continue from
+ // here so the worker exits
+ if (res.body.command === 'continue') {
+ continueRecv = true;
+ } else if (res.body.event === 'break' && continueRecv) {
+ const req = protocol.serialize({ command: 'continue' });
+ debugClient.write(req);
+ }
+ };
  const conn = net.connect({ host, port: address.port });
  conn.once('connect', common.mustCall(() => {
  conn.destroy();
  assert.notDeepStrictEqual(handles, {});
  worker.disconnect();
  assert.deepStrictEqual(handles, {});
+ // Always send the continue, as the break event might have already
+ // been received.
  const req = protocol.serialize({ command: 'continue' });
  debugClient.write(req);
  }));