[test] Migrate regressions to emotion

[vicasas]

• I have followed (at least) the PR section of the contributing guide.

One of #16947

[mui-pr-bot]

No bundle size changes (experimental)

Generated by 🚫 dangerJS against 0d2d95f

[oliviertassinari]

We need no visual changes in Argos :)

[eps1lon]

Let's not use too much API surface. Especially most of the sx stuff can be removed. It's unrelated to the test and just frustrating to deal with if it breaks.

[oliviertassinari]

@eps1lon I think that we have tried to keep the CSS to its minimum in the visual regression tests so far, but there is likely more room for simplification 👍. I would add a note of caution: we added some CSS to create the right environment to reproduce bugs, these will still need to be here.

[eps1lon]

I think that we have tried to keep the CSS to its minimum in the visual regression tests so far, but there is likely more room for simplification 👍.

Hiding more stuff behind an abstraction (e.g. sx or Box) does not make a test simpler.

[eps1lon]

The style -> sx changes should be reverted. sx is unrelated to the tests.

[vicasas]

The style -> sx changes should be reverted. sx is unrelated to the tests.

I think it is not clear to me. Should we remove sx and use style in all tests including components being tested

-<Input sx={{ margin: 1 }} />
+<Input style={{ margin: 1 }} />

or only for Box components?

-<Box sx={{ margin: 1 }} />
+<div style={{ margin: 1 }} />

[eps1lon]

I think it is not clear to me. Should we remove sx and use style in all tests including components being tested

It's not about removing. It's about keeping it the way it is. We just want to migrate the parts that are using @material-ui/styles. All the rest should stay that way (e.g. style prop usage).

[vicasas]

What is missing from this?

[oliviertassinari]

Using style over sx in the visual regression tests could make sense. It will be less refactoring in the future if we change the API. The style API is more reliable (assuming a 10 years window)

looks good

updated

[eps1lon]

Thanks for sticking with it. Much appreciated.