Fix crash when @importing a Sass file that @uses another file

[Goodwine]

The crash happened when @importing a Sass file that @uses a user-defined module, only when the Sass file @includes a mixin at the root level if the mixin emits a declaration that would end up emitted at the root. This is allowed because the @import within a style rule would end up causing that top-level declaration to be wrapped by the style rule, and therefore won't be a top-level declaration. However, this crashed the compiler instead.

Fixes: #2588

Specs: sass/sass-spec#2060

[nex3]

Please fix the CI

[Goodwine]

I believe sass-parser is broken at main rather than being broken by my PR

[nex3]

Please make the commit message shorter. The Linux git style guide recommends 50 characters for the first line of a commit.

[nex3]

It seems like the issue here is that _parent is reset to _root. What happens if we don't do that, rather than adding an extra variable to track (and then need to un-track) a different type of parent?

[Goodwine]

It seems like this is used to track ordering when doing @import containing @use.

e.g.

// This file is @imported into main.scss
@use 'uptream'
.foo-midstream{}

Results in

.foo-midstream{}
.foo-upstream{}

[nex3]

Why does the existence of _parent cause that?

[Goodwine]

I updated the tests but I'm still not clear whether @include meta.load-css() should pass and fail exactly like @import whether it's nested or not respectively. Currently it fails in both cases without crashing

[Goodwine]

I updated the tests but I'm still not clear whether @include meta.load-css() should pass and fail exactly like @import whether it's nested or not respectively. Currently it fails in both cases without crashing

discussed offline, it should fail regardless. It is only @import that gets treated differently. So it already works fine. I will just update the test.

I'll also try to figure out why was _parent introduced in the first place

[Goodwine]

I think this is now ready, the crash actually went away recently. The Sass compiler still fails but the crash was less crashy :)