• This migration could be part of the jasmine-vitest migration. Should I moved it there and add an opt-out?

Makes sense to me.

👌 It's also much less effort than having to factorize 50% of what jasmine-vitest migration is doing in terms of test file collection etc...

• flushMicrotasks can only be partially reproduced (i.e. explicit queueMicrotask calls), transforming it have high chances of producing a broken test. Should we just skip transforming tests that use it?

Hmm, wouldn't await vi.advanceTimersByTimeAsync(0) be close enough?

It's the closest thing we can get.
From the usage examples I found on github (such as on angular/components repo), I didn't notice any test where executing the 0ms timers in addition to the microtask queue would cause a side effect that would break the test.

• Should we generate code that measures time "manually" when the return value of flush is used?

Like with Date? I suppose we could and that'd get us there. Do people really use the return value....? Maybe just add a TODO if people ask for it.

👌

• Should we generate an "ad-hoc" function in the test file that reproduces flush's maxTurns option?

like with fakeTimers.loopLimit? Maybe this can also be a TODO and we just see if people really need it.

I was thinking of generating a function such as this one in every test file that is using the maxTurns option.

async function flush(maxTurns: number) {
  let count = 0;
  while (vi.getTimerCount() > 0) {
    ++count;
    if (count > maxTurns) {
     throw ...
    }
    await vi.advanceTimersToNextTimerAsync();
  }
}

I'll go with the TODO 😊

• What should we do about processNewMacroTasksSynchronously option?

Kill it 🔥

☠️

• This only migrates flush and tick if used inside fakeAsync but it's probably better to just replace in the whole file as users might create some wrappers. What do you think?

Yea, I had a comment about this and am seeing this note now. Probably should migrate it. They need to be in a fakeAsync somewhere up the chain.

👌

Perfect! Thank you very much for the review.
You've answered all my questions 😊.
I'll take care of this as soon as I can.

One last thing.
What do you think of replacing flushMicrotasks() with vi.advanceTimersToNextTimerAsync() instead of vi.advanceTimersByTimeAsync(0)? It's a tiny bit closer to it.

One last thing.
What do you think of replacing flushMicrotasks() with vi.advanceTimersToNextTimerAsync() instead of vi.advanceTimersByTimeAsync(0)? It's a tiny bit closer to it.

maybe I’m misremembering what that does but isn’t it not closer? advanceTimersToNextTimerAsync Would go all the way to the next timer, potentially to 5000 if that’s where the next one is rather than only flushing tasks at 0.

maybe I’m misremembering what that does but isn’t it not closer? advanceTimersToNextTimerAsync Would go all the way to the next timer, potentially to 5000 if that’s where the next one is rather than only flushing tasks at 0.

My bad! I should have gone to bed instead of writing my previous comment 😅.
I made two wrong assumptions:

• I was only focused on the scenario of a world where all timers have a 0ms delay 🫠.
• I assumed that advanceTimersToNextTimerAsync only triggers one timer but it actually ticks all timers with same delay (https://github.com/vitest-dev/vitest/blob/fe9ed7ac857c2e3d44217f87ad78bc2f20c6b2be/packages/vitest/src/integrations/mock/timers.ts#L101)

From a fresh morning perspective, I think that the closest thing to flushMicrotasks would be:

import { getSafeTimers } from '@vitest/utils/timers';

async function flushMicrotasks() {
  return new Promise((resolve) => {
    const { setTimeout: safeSetTimeout } = getSafeTimers();
    safeSetTimeout(resolve);
  }); 
}

The drawbacks I see are:

• having to add @vitest/utils as a dependency
• the boilerplate of generating the flushMicrotasks utility in each test that needs it

Unless! Unless it makes sense for vitest to add vi.flushMicrotasks() 🤔

But back to the usages of flushMicrotasks that I've observed, vi.advanceTimersByTimeAsync(0) works.
So pragmatically speaking, it could be the way to go and we keep the — bit more realistic — flushMicrotasks as a TODO for later.