fix(ssr): normalize class in child attrs

[cardoso]

Details

There might be a better way to do this, but it could be enough for now.

I updated the tests added in #4931 to cover this case. It still doesn't fix any expected failures but reduces the diff further:

attribute-class/unstyled/dynamic/index.js (15 to 10)
attribute-class/with-scoped-styles-only-in-child/dynamic/index.js (25 to 20)
attribute-class/with-scoped-styles-only-in-parent/dynamic/index.js (15 to 10)
attribute-class/with-scoped-styles/dynamic/index.js (25 to 20)

Does this pull request introduce a breaking change?

• 😮‍💨 No, it does not introduce a breaking change.

Does this pull request introduce an observable change?

• 🔬 Yes, it does include an observable change.

Template class binding should mostly work for child components now.

GUS work item

[nolanlawson]

Looking great so far, just a small issue with ctx.import() and a request for a new test. Thanks a bunch! 🥇

[nolanlawson]

Another case we should cover here for safety: dynamic components (lwc:component). These technically go down a different code path (LwcComponent.ts), although you should still hit it since it will call into getChildAttrsOrProps. As mentioned elsewhere though, I think this should be a separate standalone fixture.

[cardoso]

I hit an interesting behavior when first doing a single fixture:

If I imported 'x/child' and passed its constructor into lwc:component while also using it directly as <x-child></x-child>.

__SYMBOL__GENERATE_MARKUP is not defined

Is this issue known? Doesn't seem convered by the existing fixtures.

[nolanlawson]

Seems like an unrelated error!

[cardoso]

Opened #4958

[nolanlawson]

LGTM! Thanks 🌞

[nolanlawson]

/nucleus test