[b84330808]

The kind of change this PR does introduce

• a bug fix
• a new feature
• an update to the documentation
• a code change that improves performance
• other

Current behavior

With retries but no ackTimeout, the client reads the server acknowledgement as an error:

// server
socket.on("echo", (arg, cb) => cb(arg));

// client
const socket = io({ retries: 3 });
socket.emit("echo", 42, (err, val) => {
  // err === 42, val === undefined, and the server handler ran 4 times
});
await socket.emitWithAck("echo", 42); // rejects with 42

_addToQueue() pushes a handler that expects the error as its first argument. Without a timeout, _registerAckCallback() stores it as-is. So onack() does not prepend null, and the packet is re-sent until the retries run out.

New behavior

The queue handler is flagged with withError, like the emitWithAck() handler and the wrapper used when a timeout is set. The callback now gets (null, 42) and the event is sent once. On disconnection, _clearAcks() now also notifies the queue handler, as it already does with ackTimeout. So a disconnection counts as one try, and the packet is still re-sent on reconnection.

Other information (e.g. related issues)

No existing issue found. I added two tests in test/retry.ts. Before the fix they fail with expected 1 to equal null, and emitWithAck() rejects with 2.

I used Claude Code to write this change.