[test] Do not enforce the presence of ownerState.className in describeConformance

[flaviendelangle]

In @mui/x-date-pickers and @mui/x-date-pickers-pro, we are updating all our ownerState (both the one used in slotProps and the one used in styled() or styleOverrides to be a small object shared across all the slots of the library.

The idea behind this change is that, from a user point of view, it is unclear which part of this UI resolves those slots since the component are very large. So the props used for the ownerState is hardly predictable and does not contain the same information from one slot to another, often missing key information of the current state of the picker.

I encountered a small problem because describeConformance uses ownerState.className which is no longer present in our components for some tests. Do you see a problem hardcoding a value instead?

[mui-bot]

Netlify deploy preview

https://deploy-preview-44479--material-ui.netlify.app/

Bundle size report

No bundle size changes (Toolpad)
No bundle size changes

Generated by 🚫 dangerJS against 1d88653

[michaldudak]

This changes the meaning of this test entirely. The purpose of it is to test if the ownerState in callback has the props passed to the component. Since your components work differently, IMO skipping this test would make more sense.

[flaviendelangle]

We should update the name of the test then, IMHO it does not describe what you are saying.
It would be great if we could have 2 tests then, one that tests the props and what that only test the firing of the callback.
That way we can skip the one that tests the props without loosing what is interesting to us (the firing of the callback).

[michaldudak]

Yup, 2 different tests are a good option as well.

[siriwatknp]

This looks good to me.

[flaviendelangle]

I added the 2nd test 👍
And the CI is passing

[LukasTy]

LGTM. 👍

A good compromise to cater for both the @mui/material as well as the transitional approach on mui-x@v8 👌