Confusing error when using non-`XID_Start` characters in lifetime names

[jieyouxu]

Code

fn bad_lifetime_name<'🐛>(_: &'🐛 ()) {}

Current output

error: character literal may only contain one codepoint
 --> src/lib.rs:1:22
  |
1 | fn bad_lifetime_name<'🐛>(_: &'🐛 ()) {}
  |                      ^^^^^^^^^^
  |
help: if you meant to write a string literal, use double quotes
  |
1 - fn bad_lifetime_name<'🐛>(_: &'🐛 ()) {}
1 + fn bad_lifetime_name<"🐛>(_: &"🐛 ()) {}
  |

error: unexpected closing delimiter: `)`
 --> src/lib.rs:1:35
  |
1 | fn bad_lifetime_name<'🐛>(_: &'🐛 ()) {}
  |                                     ^ unexpected closing delimiter

Desired output

error: a lifetime name cannot begin with a character that is neither `_` nor a `XID_Start` character

 --> src/lib.rs:1:22
  |
1 | fn bad_lifetime_name<'🐛>(_: &'🐛 ()) {}
  |                       ^--
  |                        help: `\u{1F41B}` is neither `_` nor in `XID_Start`

Rationale and extra context

See also https://doc.rust-lang.org/reference/identifiers.html.
Initially reported at #general > Compiler isn't helpful with invalid lifetimes.

Looks like this recovery path(?) is hit when starting a lifetime name (not including the ') with a non-XID_Start character fails both is_id_start check and ASCII digit check hitting the !can_be_a_lifetime codepath.

rust/compiler/rustc_lexer/src/lib.rs

Lines 824 to 842 in c79bbfa

	let can_be_a_lifetime = if self.second() == '\'' {
	// It's surely not a lifetime.
	false
	} else {
	// If the first symbol is valid for identifier, it can be a lifetime.
	// Also check if it's a number for a better error reporting (so '0 will
	// be reported as invalid lifetime and not as unterminated char literal).
	is_id_start(self.first()) || self.first().is_ascii_digit()
	};
	if !can_be_a_lifetime {
	let terminated = self.single_quoted_string();
	let suffix_start = self.pos_within_token();
	if terminated {
	self.eat_literal_suffix();
	}
	let kind = Char { terminated };
	return Literal { kind, suffix_start };
	}

Other cases

Marginally related is loop labels, this complains about "unterminated character literal".

fn main() {
    '🐛: {
        todo!();  
    };
}

Rust Version

$ rustc --version --verbose
rustc 1.87.0 (17067e9ac 2025-05-09)
binary: rustc
commit-hash: 17067e9ac6d7ecb70e50f92c1944e545188d2359
commit-date: 2025-05-09
host: aarch64-unknown-linux-gnu
release: 1.87.0
LLVM version: 20.1.1

Anything else?

No response

[mu001999]

@rustbot claim

[mu001999]

@rustbot release-assignment

[estebank]

#153893 introduces a barebones error for these cases. I'd like to keep the ticket open for tracking adding more information to it.