react-testing-library: React 18. set state in finally throws "act" warning, though test is passing

  • @testing-library/react version: 13.1.1
  • Testing Framework and version: jest 27.5.1
  • DOM Environment: jest-dom 27.5.1

Relevant code or config:

  const handleSavedCardPayment = () => {
    const paymentInfo = {
      amount: formDataRef.current.amount,
      userId,
      accountId: formDataRef.current.selectedCardAccountId,
    };

    dispatch(makePayment(paymentInfo))
      .then(handlePaymentSuccess)
      .catch(noop) // already caught by thunk
      .finally(() => {
        formDataRef.current = null;
        setSubmitting(false);
      });
  };

What you did:

NOTE: This error only shows up after updating to react 18 and testing library to 13.1.1, this was not an issue in earlier version.

setSubmitting toggles the button disability. When the promise is complete, it turns the re-enables the button. so the test is doing a waitFor button to be re-enabled.

the promise looks more like this

dispatch(makePayment(paymentInfo)) returns a promise, with a catch inside.

so the whole promise chain looks like this promisedFunction().then(() => do something).catch(error => show error). then(() => componentLevel).catch(do nothing here).finally(reset state)

What happened:

image

the “act” error is thrown, even though i’ve added a waitFor and that is passing to prove that that statement has already been rendered.

await waitFor(() => expect(screen.getByRole('button', { name: /submit/i })).not.toBeDisabled());

Problem description:

set state in finally seems to cause testing library to think act is incomplete, even though it is.

currently solving the error by putting the set state inside of catch image

However, there’s the WET code, i had to put the set state into handlePaymentSuccess too, to achieve the same results.

About this issue

  • Original URL
  • State: closed
  • Created 2 years ago
  • Reactions: 41
  • Comments: 32 (6 by maintainers)

Commits related to this issue

Most upvoted comments

I appreciate all the additional info here, both from maintainers and other users. I have never really wrapped my head around act(). I understand when it should be used, but I don’t fully understand what it’s actually doing. And I think because of that, I’m a bit unclear on what the right thing to do here is - to remove all of these act() warnings which started showing up after upgrading to React 18.

Is it something we as consumers should be doing differently in our tests? Something that needs to change on the React side? Something that needs to change in react-testing-library? Some combination of things? Something else entirely? Not clear yet?

I’m essentially delaying my upgrade to React 18, as a result of this issue, but it’s not quite clear to me what the eventual fix will be. So if someone is able to add some clarity there, I would really appreciate that. 🙏

I have a test roughly like this:

    setup();
    await userEvent.click(someButton);
    await waitFor(() => 
      expect(getCell()).toHaveTextContent('...');
    );

Clicking the button triggers an function with an API call roughly like this:

    async function onClick() {
      try {
        setLoading(true);

        const data = await api('POST', ...);

        // update the swr cache with the new data
        // this is what will update the DOM
        await mutate(...);
      } catch (error) {
        setError(error);
      } finally {
        setLoading(false);
      }
    }

The test works fine, but React still gives me an act warning about the setLoading(false) in the finally block. Calling unmount() doesn’t help either.

What I had to do was waitFor the loading state to be set to false by testing it

    setup();
    await userEvent.click(someButton);
    await waitForElementToBeRemoved(someSpinner);
    expect(getCell()).toHaveTextContent('...');

Now the act warning is gone, but this isn’t ideal IMO as

  1. I don’t care to test for the loading state itself
  2. I don’t get a warning for this with React 17

🤷‍♂️

To be sure I’m not missing anything, I created two small codesandboxes with the same test.

In @testing-lib/react@^12 with react@^17, the following simple test does not produce an act warning. (presumably because waitFor is wrapped in an act)

https://codesandbox.io/s/act-warning-waitfor-react-17-btzykv

import { render, screen, waitFor } from "@testing-library/react";
import App from "./App";
import userEvent from "@testing-library/user-event";

describe("app", () => {
  it("turns on", async () => {
    render(<App />);

    await userEvent.click(screen.getByText("Turn on"));

   // I'm aware of find*, but the point is about `waitFor`
    await waitFor(async () => {
      screen.getByText("It is on");
    });

   // that's just here so the test doesn't finish prematurely 
    await new Promise((resolve) => {
      setTimeout(resolve, 300);
    });
  });
});

The component under test, shows some text on click after a few ms

import { useState } from "react";
import "./styles.css";

export default function App() {
  const [isOn, setIsOn] = useState(false);

  return (
    <div className="App">
      {isOn && <div>It is on</div>}
      <button
        onClick={() => {
          setTimeout(() => {
            setIsOn(true);
          }, 1000);
        }}
      >
        Turn on
      </button>
    </div>
  );
}

The same setup with @testing-library/react@^13 and react@18, now produces a warning.

https://codesandbox.io/s/act-warning-wait-for-react-18-w4e521

image

@eps1lon is this really the intended behaviour from now on or am I missing/misunderstanding something?

It also seems like waitFor sets global.IS_REACT_ACT_ENVIRONMENT to be false while it executes, which can cause the The current testing environment is not configured to support act(...) warning when one tries to manually wrap certain calls with act (I still need to wrap my head around this one 😅)

Example sandbox for this behaviour: https://codesandbox.io/s/act-warning-wait-for-react-18-forked-nc2mm8?file=/src/App.test.js

Based on my tests, it looks like cleanup isn’t being called at the end of every test as it used to. For me, if I manually unmount the component or call the cleanup function, the act errors go away.


Edit:

Note that calling cleanup in an afterEach will not work. You will have to actually call it at the bottom of the test. Something like this

test('should do something', () => {
  render(<Component />);
  // test stuff
  cleanup();
});

I am having a similar problem after updating to react 18.2.0 and @testing-library/react 13.3.0. Some of my calls to dispatch store actions need to be wrapped in await act(async () => {...} ). I thought that @testing-library/react took care of the act wrapping for us.

Hey guys, i made some tests because I am having the same problem: This is what I got.

@testing-library/react”: “^13.3.0”, “react”: “^18.2.0”, “react-dom”: “^18.2.0”, “@testing-library/user-event”: “^14.2.1”,

// it pass without act() warning
fireEvent.change(loginField, { target: { value: '' } })
expect(loginField).toHaveValue('')

// it gives the act() warning
fireEvent.change(loginField, { target: { value: '' } })
await waitFor(async () => {
  expect(loginField).toHaveValue('')
})

// don't wait the render - error
fireEvent.click(loginButton)
expect(screen.queryAllByText('Campo obrigatório')).toHaveLength(2)

// it pass all fine :)
fireEvent.click(loginButton)
await waitFor(async () => {
  expect(screen.queryAllByText('Campo obrigatório')).toHaveLength(2)
})

This way it works fine:

// it pass without act() warning
fireEvent.change(loginField, { target: { value: '' } })
expect(loginField).toHaveValue('') 

But if I change to userEvent, I get the act() warning all over again.

userEvent.clear(loginField)
expect(loginField).toHaveValue('')

We are able to spot-fix tests by wrapping in act, but tests then fail previously passing cases with strange outputs that do not match manual testing.

Not to mention in our large-ish codebase, its nigh impossible to manually update the 1000+ tests we have…

I’m using react-hook-form and experience the same act warning. Using either waitFor or waitForElementToBeRemoved doesn’t remove the warning.

The only thing that worked is to wrap act with sleep.

  await act(async () => {
    await new Promise((resolve) => {
      setTimeout(resolve, 50);
    });
  });

credit to https://bufferings.hatenablog.com/entry/2021/11/18/015809

@eps1lon Is my example in this comment not sufficient to demonstrate the issue?

Having struggled with a lot of test was not wrapped in act(...) warnings in our tests as well, I noticed we had added @testing-library/dom in our project dependencies. On re-reading the @testing-library/user-event install guide it advises not to do this:

If you use one of the framework wrappers, it is important that @testing-library/dom is resolved to the same installation required by the framework wrapper of your choice. Usually this means that if you use one of the framework wrappers, you should not add @testing-library/dom to your project dependencies.

On removing the dependency from our package.json all the warnings were gone. This may not be a fix for all the described issues, though adding comment here incase this helps any others having similar struggles.

For info, currently using:

@testing-library/jest-dom”: “5.16.4”, “@testing-library/react”: “13.3.0”, “@testing-library/user-event”: “14.3.0”, “react”: “18.2.0”, “react-dom”: “18.2.0”

I’m running into the same problem upgrading my App to React 18, and RTL to 13.3.0.

In my case, I’ve even got a test which triggers an act warning for every letter pressed from await user.type() triggering an onChange handler, which just calls a setter from a useState hook.

Even if I add await waitFor(() => expect(myInput).toHaveValue(finalValue) it still throws the warnings (amongst hundreds of other act warnings in my other tests).

I have observed the same behavior as @JoeyAtSS. After update to the latest React 18 and the latest RTL, we had to change following across whole application, otherwise error from above is shown:

image

We also observed https://github.com/testing-library/react-testing-library/issues/1057. Unfortunately, I haven’t observed pattern in which it happens.

@eps1lon I replicated our issue here: https://codesandbox.io/s/rtl-react-18-act-issue-forked-l0bcj2?file=/src/__tests__/App.test.js.

You can see this error when you run tests and check the output in the console. In our case it’s triggered by a component that uses react-query with dynamic import.

We’re getting these warnings/errors in other places as well (e.g. we had 2 getByText(...).click() calls right after each other and after updating to React 18, I had to wrap each in a separate act call), but I wasn’t able to create a minimal example for these.

I encountered something similar while using Formik, I suspect this is related. Here is a repo which reproduces this issue:

This only happens on the new version of RTL and React 18.

https://github.com/testing-library/react-testing-library/blob/9171163fccf0a7ea43763475ca2980898b4079a5/src/pure.js#L19-L24 It looks like this is the culprit, it disables the “act environment” while running waitFor. This doesn’t make sense to me, any particular reason for doing this @eps1lon?

It was already mentioned by others, but making sure that we only have one version of @testing-library/dom library helped us significantly reduce number of these warnings.

If you use yarn, you can run yarn why @testing-library/dom to see how many versions of this library you have and why. If you have more than one, it’s probably best to uninstall all libraries that depend on it and reinstall them.

Is this documented somewhere?

I guess the documentation for act() could be more explicit about this. It’s somewhat implied by “state updates have been processed at the end” but I can see how this is fairly ambigious

Do we need to flush state updates? Don’t they run anyway when they can?

If inside an act scope no state updates, effects etc are flushed. Only at the end. Which is why it wouldn’t make sense to wrap a waiting period in act

Like others have mentioned, we’re also seeing this pervasively after upgrading to react 18 and latest RTL.

"@testing-library/jest-dom": "5.16.4",
"@testing-library/react": "13.3.0",
"@testing-library/user-event": "14.4.0",
"react": "18.2.0",
"react-dom": "18.2.0",

Hoping for a solution that doesn’t involve wrapping all our tests in additional act or adding cleanup() to every test or what have you.

Edit: In my case, it seems it’s caused by https://github.com/remix-run/react-router/issues/7634 as we’re also upgrading to react router v6 as part of this.

The useNavigate() hook does not return a stable reference, and using it it in the dependency array of a useEffect causes any useEffect to run again if the location changes. The end result of this is that if a test navigates as part of the user interaction, the useEffect gets triggered again and if it updates state, it may be doing so after the test has passed and completed.

Just putting this here in case it helps any one else dropping by.

Just to add to this, we’ve noticed our tests riddled with act errors since moving to react 18. I’ve been able to solve most by swapping getBy to findBys and using waitFors however it has been a pain updating pretty much every test. In addition, I also had to enable globalThis.IS_REACT_ACT_ENVIRONMENT = true; so that I could use act errors to fix the state updates that I couldn’t control via a findBy.

@eps1lon here is what I found out…

29.4.3 have act warning… "jest": "^29.4.3" "jest-environment-jsdom": "^29.4.3"

but 29.3.1 is working perfectly "jest": "^29.3.1" "jest-environment-jsdom": "^29.3.1" @smellai give it a try, hope can help to solve it

I’m getting more act warnings now with @testing-library@14.0.0. Tested with await waitFor, and still throwing act warnings with userEvent that triggers a state change.

Tested version: “jest”: “29.4.3”, “@testing-library/jest-dom”: “5.16.5”, “@testing-library/react”: “14.0.0”, “@testing-library/user-event”: “14.4.3”,