test_runner: t.after is never called

[KhafraDev]

Version

v21.7.0

Platform

n/a

Subsystem

test_runner

What steps will reproduce the bug?

import { once } from 'node:events'
import { createServer } from 'node:http'
import { test } from 'node:test'

test('after is called', async (t) => {
  const server = createServer((req, res) => res.end())
    .listen(0)

  t.after(() => {
    console.log('Called!')
    server.close()
  })
  await once(server, 'listening')

  await fetch(`http://localhost:${server.address().port}`)
})

node --test test.mjs

How often does it reproduce? Is there a required condition?

always

What is the expected behavior? Why is that the expected behavior?

No response

What do you see instead?

the test hangs, "Called!" is never printed to the console

Additional information

It could be caused by #51389?

[cjihrig]

It could be caused by #51389?

Probably. It looks like #51389 was fixing #51371, which does not look like a bug according to the documentation.

[Uzlopak]

cross referencing to nodejs/undici#2931

[Uzlopak]

@cjihrig

In #51371 the example is

import { test } from 'node:test';
import { open } from 'node:fs/promises';

test('test', async (t) => {
  let filehandle;

  t.before(async () => {
    console.log('before');
    filehandle = await open('./index.mjs', 'r');
  });

  t.after(async () => {
    console.log('after');
    await filehandle.close();
  });

//   await t.test(() => {});
});

This can be simplified to

import { test } from 'node:test';

test('test', async (t) => {

  t.before(async () => {
    console.log('before');
  });

  t.after(async () => {
    console.log('after');
  });

  // await t.test(() => {});
});

So running this on node 21.6.2 results in printing only after but not before. In 21.7.0 it does not print anything.
commenting in the test call then in 21.6.2 it prints before and after, but in 21.7.0 only before.

[cjihrig]

I have opened #51998 with the revert + a regression test.

[Uzlopak]

I am currently building node from main branch and want to understand what we should expect. But I wonder what the correct behavior is. Should it really not trigger before and after if there is not test, or should it anyway run before and after if there is no test?!

[cjihrig]

The documented behavior for context.before() is:

This function is used to create a hook running before subtest of the current test.

The documented behavior for context.after() is:

This function is used to create a hook that runs after the current test finishes.

before(), beforeEach(), and afterEach() only run when there are subtests. after() should run all the time for cleanup purposes (this issue). The reason before() doesn't run all the time is because there are no subtests to run before and it can't run before the test itself that is already running.

[Uzlopak]

Actually I have the feeling that before should also run always like after. But I thinks that your revert PR should land asap.

[cjihrig]

Feel free to change whatever you want 😄

[Uzlopak]

I would rather prefer continue working on undici ;).