Rollup merge of #123752 - estebank:emoji-prefix, r=wesleywiser

Properly handle emojis as literal prefix in macros

Do not accept the following

```rust
macro_rules! lexes {($($_:tt)*) => {}}
lexes!(🐛"foo");
```

Before, invalid emoji identifiers were gated during parsing instead of lexing in all cases, but this didn't account for macro pre-expansion of literal prefixes.

Fix #123696.

1 files changed, 4 insertions, 1 deletions

diff --git a/compiler/rustc_parse/src/lexer/mod.rs b/compiler/rustc_parse/src/lexer/mod.rs
index 058539e1717..1abb1d29562 100644
--- a/compiler/rustc_parse/src/lexer/mod.rs
+++ b/compiler/rustc_parse/src/lexer/mod.rs
@@ -204,6 +204,7 @@ impl<'psess, 'src> StringReader<'psess, 'src> {
                     self.ident(start)
                 }
                 rustc_lexer::TokenKind::InvalidIdent
+                | rustc_lexer::TokenKind::InvalidPrefix
                     // Do not recover an identifier with emoji if the codepoint is a confusable
                     // with a recoverable substitution token, like `➖`.
                     if !UNICODE_ARRAY
@@ -301,7 +302,9 @@ impl<'psess, 'src> StringReader<'psess, 'src> {
                 rustc_lexer::TokenKind::Caret => token::BinOp(token::Caret),
                 rustc_lexer::TokenKind::Percent => token::BinOp(token::Percent),
 
-                rustc_lexer::TokenKind::Unknown | rustc_lexer::TokenKind::InvalidIdent => {
+                rustc_lexer::TokenKind::Unknown
+                | rustc_lexer::TokenKind::InvalidIdent
+                | rustc_lexer::TokenKind::InvalidPrefix => {
                     // Don't emit diagnostics for sequences of the same invalid token
                     if swallow_next_invalid > 0 {
                         swallow_next_invalid -= 1;