Rust 2015 and 2018 allow emoji in identifiers in "Unknown prefix" position

[mattheww]

In the 2015 and 2018 editions, the following compiles (with warnings):

macro_rules! lexes {($($_:tt)*) => {}}

lexes!(🐛#);
lexes!(🐛"foo");
lexes!(🐛'q');
lexes!(🐛'q);

playground

The 🐛 is taken as an identifier, although emoji aren't generally permitted in identifiers in any edition.

I tested with rustc 1.77.1.

I think the underlying problem is that ident_or_unknown_prefix() and fake_ident_or_unknown_prefix() in rustc_lexer distinguish "identifiers" containing emoji (as InvalidIdent rather than Ident), but don't have a way to make that distinction for UnknownPrefix.

[fmease]

Regression from 1.57 to 1.58 (stable to stable). Very likely in #88781, cc @estebank.
Regression from error to pass. Previous error: unknown start of token: \u{1f41b}.

[apiraino]

WG-prioritization assigning priority (Zulip discussion).

@rustbot label -I-prioritize +P-low