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 expansion of literal prefixes.

Fix #123696.

Author: estebank

compiler/rustc_lexer/src/lib.rs

@@ -88,6 +88,10 @@ pub enum TokenKind {
     /// tokens.
     UnknownPrefix,
 
+    /// Similar to the above, but *always* an error on every edition. This is used
+    /// for emoji identifier recovery, as those are not meant to be ever accepted.
+    InvalidPrefix,
+
     /// Examples: `12u8`, `1.0e-40`, `b"123"`. Note that `_` is an invalid
     /// suffix, but may be present here on string and float literals. Users of
     /// this type will need to check for and reject that case.
@@ -528,7 +532,7 @@ impl Cursor<'_> {
         // Known prefixes must have been handled earlier. So if
         // we see a prefix here, it is definitely an unknown prefix.
         match self.first() {
-            '#' | '"' | '\'' => UnknownPrefix,
+            '#' | '"' | '\'' => InvalidPrefix,
             _ => InvalidIdent,
         }
     }

compiler/rustc_parse/src/lexer/mod.rs

@@ -205,6 +205,7 @@ 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
@@ -302,7 +303,9 @@ 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;

src/librustdoc/html/highlight.rs

@@ -876,9 +876,10 @@ impl<'src> Classifier<'src> {
                 },
                 Some(c) => c,
             },
-            TokenKind::RawIdent | TokenKind::UnknownPrefix | TokenKind::InvalidIdent => {
-                Class::Ident(self.new_span(before, text))
-            }
+            TokenKind::RawIdent
+            | TokenKind::UnknownPrefix
+            | TokenKind::InvalidPrefix
+            | TokenKind::InvalidIdent => Class::Ident(self.new_span(before, text)),
             TokenKind::Lifetime { .. } => Class::Lifetime,
             TokenKind::Eof => panic!("Eof in advance"),
         };

src/tools/rust-analyzer/crates/parser/src/lexed_str.rs

@@ -178,7 +178,7 @@ impl<'a> Converter<'a> {
                 rustc_lexer::TokenKind::Ident => {
                     SyntaxKind::from_keyword(token_text).unwrap_or(IDENT)
                 }
-                rustc_lexer::TokenKind::InvalidIdent => {
+                rustc_lexer::TokenKind::InvalidPrefix | rustc_lexer::TokenKind::InvalidIdent => {
                     err = "Ident contains invalid characters";
                     IDENT
                 }

tests/ui/lexer/emoji-literal-prefix.rs

@@ -0,0 +1,8 @@
+macro_rules! lexes {($($_:tt)*) => {}}
+
+lexes!(🐛#); //~ ERROR identifiers cannot contain emoji
+lexes!(🐛"foo");
+lexes!(🐛'q');
+lexes!(🐛'q);
+
+fn main() {}

tests/ui/lexer/emoji-literal-prefix.stderr

@@ -0,0 +1,14 @@
+error: identifiers cannot contain emoji: `🐛`
+  --> $DIR/emoji-literal-prefix.rs:3:8
+   |
+LL | lexes!(🐛#);
+   |        ^^
+LL | lexes!(🐛"foo");
+   |        ^^
+LL | lexes!(🐛'q');
+   |        ^^
+LL | lexes!(🐛'q);
+   |        ^^
+
+error: aborting due to 1 previous error
+