about summary refs log tree commit diff
path: root/compiler/rustc_parse
diff options
context:
space:
mode:
Diffstat (limited to 'compiler/rustc_parse')
-rw-r--r--compiler/rustc_parse/messages.ftl4
-rw-r--r--compiler/rustc_parse/src/errors.rs18
-rw-r--r--compiler/rustc_parse/src/lexer/mod.rs84
3 files changed, 105 insertions, 1 deletions
diff --git a/compiler/rustc_parse/messages.ftl b/compiler/rustc_parse/messages.ftl
index 948199fd55c..ba5e2ddf4fc 100644
--- a/compiler/rustc_parse/messages.ftl
+++ b/compiler/rustc_parse/messages.ftl
@@ -706,6 +706,10 @@ parse_require_colon_after_labeled_expression = labeled expression must be follow
     .label = the label
     .suggestion = add `:` after the label
 
+parse_reserved_string = invalid string literal
+    .note = unprefixed guarded string literals are reserved for future use since Rust 2024
+    .suggestion_whitespace = consider inserting whitespace here
+
 parse_return_types_use_thin_arrow = return types are denoted using `->`
     .suggestion = use `->` instead
 
diff --git a/compiler/rustc_parse/src/errors.rs b/compiler/rustc_parse/src/errors.rs
index dade3912751..124975f67f1 100644
--- a/compiler/rustc_parse/src/errors.rs
+++ b/compiler/rustc_parse/src/errors.rs
@@ -2111,6 +2111,24 @@ pub(crate) enum UnknownPrefixSugg {
 }
 
 #[derive(Diagnostic)]
+#[diag(parse_reserved_string)]
+#[note]
+pub(crate) struct ReservedString {
+    #[primary_span]
+    pub span: Span,
+    #[subdiagnostic]
+    pub sugg: Option<GuardedStringSugg>,
+}
+#[derive(Subdiagnostic)]
+#[suggestion(
+    parse_suggestion_whitespace,
+    code = " ",
+    applicability = "maybe-incorrect",
+    style = "verbose"
+)]
+pub(crate) struct GuardedStringSugg(#[primary_span] pub Span);
+
+#[derive(Diagnostic)]
 #[diag(parse_too_many_hashes)]
 pub(crate) struct TooManyHashes {
     #[primary_span]
diff --git a/compiler/rustc_parse/src/lexer/mod.rs b/compiler/rustc_parse/src/lexer/mod.rs
index 3e46fc93fa4..d627ef3d2cb 100644
--- a/compiler/rustc_parse/src/lexer/mod.rs
+++ b/compiler/rustc_parse/src/lexer/mod.rs
@@ -10,7 +10,8 @@ use rustc_lexer::unescape::{self, EscapeError, Mode};
 use rustc_lexer::{Base, Cursor, DocStyle, LiteralKind, RawStrError};
 use rustc_session::lint::BuiltinLintDiag;
 use rustc_session::lint::builtin::{
-    RUST_2021_PREFIXES_INCOMPATIBLE_SYNTAX, TEXT_DIRECTION_CODEPOINT_IN_COMMENT,
+    RUST_2021_PREFIXES_INCOMPATIBLE_SYNTAX, RUST_2024_GUARDED_STRING_INCOMPATIBLE_SYNTAX,
+    TEXT_DIRECTION_CODEPOINT_IN_COMMENT,
 };
 use rustc_session::parse::ParseSess;
 use rustc_span::symbol::Symbol;
@@ -251,6 +252,7 @@ impl<'psess, 'src> StringReader<'psess, 'src> {
                     let prefix_span = self.mk_sp(start, lit_start);
                     return (Token::new(self.ident(start), prefix_span), preceded_by_whitespace);
                 }
+                rustc_lexer::TokenKind::GuardedStrPrefix => self.maybe_report_guarded_str(start, str_before),
                 rustc_lexer::TokenKind::Literal { kind, suffix_start } => {
                     let suffix_start = start + BytePos(suffix_start);
                     let (kind, symbol) = self.cook_lexer_literal(start, suffix_start, kind);
@@ -781,6 +783,86 @@ impl<'psess, 'src> StringReader<'psess, 'src> {
         }
     }
 
+    /// Detect guarded string literal syntax
+    ///
+    /// RFC 3598 reserved this syntax for future use. As of Rust 2024,
+    /// using this syntax produces an error. In earlier editions, however, it
+    /// only results in an (allowed by default) lint, and is treated as
+    /// separate tokens.
+    fn maybe_report_guarded_str(&mut self, start: BytePos, str_before: &'src str) -> TokenKind {
+        let span = self.mk_sp(start, self.pos);
+        let edition2024 = span.edition().at_least_rust_2024();
+
+        let space_pos = start + BytePos(1);
+        let space_span = self.mk_sp(space_pos, space_pos);
+
+        let mut cursor = Cursor::new(str_before);
+
+        let (span, unterminated) = match cursor.guarded_double_quoted_string() {
+            Some(rustc_lexer::GuardedStr { n_hashes, terminated, token_len }) => {
+                let end = start + BytePos(token_len);
+                let span = self.mk_sp(start, end);
+                let str_start = start + BytePos(n_hashes);
+
+                if edition2024 {
+                    self.cursor = cursor;
+                    self.pos = end;
+                }
+
+                let unterminated = if terminated { None } else { Some(str_start) };
+
+                (span, unterminated)
+            }
+            _ => {
+                // We should only get here in the `##+` case.
+                debug_assert_eq!(self.str_from_to(start, start + BytePos(2)), "##");
+
+                (span, None)
+            }
+        };
+        if edition2024 {
+            if let Some(str_start) = unterminated {
+                // Only a fatal error if string is unterminated.
+                self.dcx()
+                    .struct_span_fatal(
+                        self.mk_sp(str_start, self.pos),
+                        "unterminated double quote string",
+                    )
+                    .with_code(E0765)
+                    .emit()
+            }
+
+            let sugg = if span.from_expansion() {
+                None
+            } else {
+                Some(errors::GuardedStringSugg(space_span))
+            };
+
+            // In Edition 2024 and later, emit a hard error.
+            let err = self.dcx().emit_err(errors::ReservedString { span, sugg });
+
+            token::Literal(token::Lit {
+                kind: token::Err(err),
+                symbol: self.symbol_from_to(start, self.pos),
+                suffix: None,
+            })
+        } else {
+            // Before Rust 2024, only emit a lint for migration.
+            self.psess.buffer_lint(
+                RUST_2024_GUARDED_STRING_INCOMPATIBLE_SYNTAX,
+                span,
+                ast::CRATE_NODE_ID,
+                BuiltinLintDiag::ReservedString(space_span),
+            );
+
+            // For backwards compatibility, roll back to after just the first `#`
+            // and return the `Pound` token.
+            self.pos = start + BytePos(1);
+            self.cursor = Cursor::new(&str_before[1..]);
+            token::Pound
+        }
+    }
+
     fn report_too_many_hashes(&self, start: BytePos, num: u32) -> ! {
         self.dcx().emit_fatal(errors::TooManyHashes { span: self.mk_sp(start, self.pos), num });
     }