[css-mixins-1] Lexical vs. dynamic env() scoping

[andruud]

If I understand the current spec correctly, you can not access outer mixin parameters from an inner @apply "call", since the target of that @apply call won't have the necessary lexical scope in its parent rule chain:

@mixin --one() {
  & { color: env(--color); } /* --color not defined here */
}

@mixin --two() {
  @env --color: green;
  @apply --one();
}

div { @apply --two(); }

I'm not a huge fan of dynamic scoping personally, and I do get that that authors can just define their envs at the stylesheet-level if they want "globals", but this feels inconsistent with what we decided in #10954, so I'd like an explicit decision on this.

[tabatkins]

Correct, per the current spec, lexical scoping would prevent --one()'s body from seeing the @env --color defined in --two()'s body.

I'm not opposed to the scoping being post-substitution, though. That would enable some use-cases, like what you give, or passing a @contents block that will use the mixin to define some env()s in it.