Should "div :is(*)" in a scoped style block produce a different selector?

[kleinfreund]

In a scoped style block, the following:

<style scoped>
div * {
  color: red;
}

div :is(*) {
  color: blue;
}

:is(*) {
  color: green;
}

div *:is(*) {
  color: green;
}
</style>

... produces this CSS:

div *[data-v-1cb5f508] {
  color: red;
}
div[data-v-1cb5f508] :is(*) {
  color: blue;
}
[data-v-1cb5f508]:is(*) {
  color: green;
}
div *[data-v-1cb5f508]:is(*) {
  color: green;
}

This is unexpected to me. The red and green examples make perfect sense, but especially considering how the green examples work, I expected the blue example to have the data-v-* attribute attached to the key part of the selector (i.e. preceding the :is pseudo class) and not the div tag selector.

Should div :is(*) produce div [data-v-1cb5f508]:is(*) instead?

Currently, this allows to break out of scoped styles. Consider the component to have a slot as a child of a div. Currently, the blue example will apply color: blue to all elements in this slot. I don't think that should be the case (i.e. I think this is a bug), but I'm unsure so I'm asking here instead of reporting an issue.

Is there a reason why this should behave like it does today?

Notes:

• This is general behavior for selectors of the style div :pseudo-class (e.g. you can swap out :is for something else in the blue example and get similar results)

[adoin]

:is 并不是实体选择器，是修饰选择器，对之前的补充，所以上面编译的没问题啊。
Normal logic, because :is is not a entity selector, but a adjective selector. Same as :after