Regression on 'componentDidUpdate' not being called when calling 'setProps' on shallow wrapped component

[BertrandFritsch]

Calling setProps on a shallow wrapper may result in not calling componentDidUpdate of the wrapped component. It's the case when the state depends on a prop and when that specific setProps call does change the state. If that call doesn't change the state, componentDidUpdate is called. I suspect this is a regression of #2007. It worked fine in version 3.8.0

To reproduce

import { shallow } from 'enzyme';
import * as React from 'react';

class DummyComp extends React.PureComponent {
  constructor(props) {
    super(props);
    this.state = { stateCounter: 0 };
  }

  static getDerivedStateFromProps(nextProps) {
    // set stateCounter if asked for
    if (nextProps.propCounter > -1) {
      return {
        stateCounter: nextProps.propCounter
      };
    }

    return null;
  }

  componentDidUpdate() {
    console.log('Counter:', this.state.stateCounter);
  }

  render() {
    return <p>Counter: { this.state.stateCounter }</p>;
  }
}

describe('Counter', () => {
  it('should call cDU on setProps', () => {

    const wrapper = shallow(<DummyComp propCounter={ 0 } />);

    const cDUSpy = jest.spyOn(wrapper.instance(), 'componentDidUpdate');

    // does not change the state, so cDU will be called
    wrapper.setProps({ propCounter: -1 });

    expect(cDUSpy).toBeCalledTimes(1);

    // does change the state, cDU should be called anyway
    wrapper.setProps({ propCounter: 1 });

    expect(cDUSpy).toBeCalledTimes(2); // this test will fail
  });
});

Your environment

API

• shallow
• mount
• render

Version

library	version
enzyme	3.9.0
react	16.8.2
react-dom	16.8.2
react-test-renderer	16.8.2
adapter (below)

Adapter

• enzyme-adapter-react-16
• enzyme-adapter-react-16.3
• enzyme-adapter-react-16.2
• enzyme-adapter-react-16.1
• enzyme-adapter-react-15
• enzyme-adapter-react-15.4
• enzyme-adapter-react-14
• enzyme-adapter-react-13
• enzyme-adapter-react-helper
• others ( )

[ljharb]

cc @peanutenthusiast; see also #2007 (comment)

[ghost]

@ljharb thanks for the cc will look into this.

[ghost]

Replicated the above test case, which results in the aforementioned regression. With regards to the fix and the previous discussions from 1862, is it possible to apply a fix without implementing batched updates? @ljharb any pointers would be appreciated

[ljharb]

@peanutenthusiast i'm honestly not sure; the first thing ideally would be a PR that contains the failing test case, and we can iterate from there?