This error seems to be wrong, since Bar is Foo and thus it has the #m static field.

I can't tell if it's completely correct or not.
In this PR, I convert SyntaxError into TypeError.
Maybe it should be counted as "external" access.

According to the latest nightly for TS, import(..).X.#priv is technically syntactically valid, but it will never be semantically valid because it counts as "external" access to the private member, which is disallowed.
...

Originally posted by @bradzacher in babel/babel#14454 (comment)

I will try to add more test cases to determine which case is more reasonable and self-consistent.

I think a comparable way to test this would be to stick private onto the non-# one and see what happens.

I think a comparable way to test this would be to stick private onto the non-# one and see what happens.

I didn't manage to do it.
At first I tried to alias type Bar = typeof import("./lib").Bar, then I found I cannot use Bar.#m.
typeof import("./lib").Bar.#m does not imply (typeof import("./lib").Bar).#m.
it means typeof (import("./lib").Bar.#m).

What do you think about it?

For private, I just meant to write something like:

// @strict: true
// @target: esnext

// @filename: main.ts
export class Foo {
  private static m = 1;
  static #f(x: typeof import("./lib").Bar.m) {}
}

// @filename: lib.ts
export { Foo as Bar } from "./main";

Where the static thing is "private" via TS, not via ECMAScript's private identifiers. I was replying on my phone so couldn't write it myself, sorry about that.

Regarding the precedence, that's #19707 (mentioned in #48640). All of this is getting really hairy and inconsistent, though.

The AST should be changed. This PR is blocked by #48640.