compiler/rustc_parse/src/lexer/mod.rs

   1 use crate::lexer::unicode_chars::UNICODE_ARRAY;
   2 use rustc_ast::ast::{self, AttrStyle};
   3 use rustc_ast::token::{self, CommentKind, Delimiter, Token, TokenKind};
   4 use rustc_ast::tokenstream::{Spacing, TokenStream};
   5 use rustc_ast::util::unicode::contains_text_flow_control_chars;
   6 use rustc_errors::{error_code, Applicability, DiagnosticBuilder, ErrorGuaranteed, PResult};
   7 use rustc_lexer::unescape::{self, Mode};
   8 use rustc_lexer::{Base, DocStyle, RawStrError};
   9 use rustc_session::lint::builtin::{
  10     RUST_2021_PREFIXES_INCOMPATIBLE_SYNTAX, TEXT_DIRECTION_CODEPOINT_IN_COMMENT,
  11 };
  12 use rustc_session::lint::BuiltinLintDiagnostics;
  13 use rustc_session::parse::ParseSess;
  14 use rustc_span::symbol::{sym, Symbol};
  15 use rustc_span::{edition::Edition, BytePos, Pos, Span};
  16
  17 mod tokentrees;
  18 mod unescape_error_reporting;
  19 mod unicode_chars;
  20
  21 use unescape_error_reporting::{emit_unescape_error, escaped_char};
  22
  23 // This type is used a lot. Make sure it doesn't unintentionally get bigger.
  24 //
  25 // This assertion is in this crate, rather than in `rustc_lexer`, because that
  26 // crate cannot depend on `rustc_data_structures`.
  27 #[cfg(all(target_arch = "x86_64", target_pointer_width = "64"))]
  28 rustc_data_structures::static_assert_size!(rustc_lexer::Token, 12);
  29
  30 #[derive(Clone, Debug)]
  31 pub struct UnmatchedBrace {
  32     pub expected_delim: Delimiter,
  33     pub found_delim: Option<Delimiter>,
  34     pub found_span: Span,
  35     pub unclosed_span: Option<Span>,
  36     pub candidate_span: Option<Span>,
  37 }
  38
  39 pub(crate) fn parse_token_trees<'a>(
  40     sess: &'a ParseSess,
  41     src: &'a str,
  42     start_pos: BytePos,
  43     override_span: Option<Span>,
  44 ) -> (PResult<'a, TokenStream>, Vec<UnmatchedBrace>) {
  45     StringReader { sess, start_pos, pos: start_pos, src, override_span }.into_token_trees()
  46 }
  47
  48 struct StringReader<'a> {
  49     sess: &'a ParseSess,
  50     /// Initial position, read-only.
  51     start_pos: BytePos,
  52     /// The absolute offset within the source_map of the current character.
  53     pos: BytePos,
  54     /// Source text to tokenize.
  55     src: &'a str,
  56     override_span: Option<Span>,
  57 }
  58
  59 impl<'a> StringReader<'a> {
  60     fn mk_sp(&self, lo: BytePos, hi: BytePos) -> Span {
  61         self.override_span.unwrap_or_else(|| Span::with_root_ctxt(lo, hi))
  62     }
  63
  64     /// Returns the next token, and info about preceding whitespace, if any.
  65     fn next_token(&mut self) -> (Spacing, Token) {
  66         let mut spacing = Spacing::Joint;
  67
  68         // Skip `#!` at the start of the file
  69         if self.pos == self.start_pos
  70             && let Some(shebang_len) = rustc_lexer::strip_shebang(self.src)
  71         {
  72             self.pos = self.pos + BytePos::from_usize(shebang_len);
  73             spacing = Spacing::Alone;
  74         }
  75
  76         // Skip trivial (whitespace & comments) tokens
  77         loop {
  78             let start_src_index = self.src_index(self.pos);
  79             let text: &str = &self.src[start_src_index..];
  80
  81             if text.is_empty() {
  82                 let span = self.mk_sp(self.pos, self.pos);
  83                 return (spacing, Token::new(token::Eof, span));
  84             }
  85
  86             let token = rustc_lexer::first_token(text);
  87
  88             let start = self.pos;
  89             self.pos = self.pos + BytePos(token.len);
  90
  91             debug!("next_token: {:?}({:?})", token.kind, self.str_from(start));
  92
  93             match self.cook_lexer_token(token.kind, start) {
  94                 Some(kind) => {
  95                     let span = self.mk_sp(start, self.pos);
  96                     return (spacing, Token::new(kind, span));
  97                 }
  98                 None => spacing = Spacing::Alone,
  99             }
 100         }
 101     }
 102
 103     /// Report a fatal lexical error with a given span.
 104     fn fatal_span(&self, sp: Span, m: &str) -> ! {
 105         self.sess.span_diagnostic.span_fatal(sp, m)
 106     }
 107
 108     /// Report a lexical error with a given span.
 109     fn err_span(&self, sp: Span, m: &str) {
 110         self.sess.span_diagnostic.struct_span_err(sp, m).emit();
 111     }
 112
 113     /// Report a fatal error spanning [`from_pos`, `to_pos`).
 114     fn fatal_span_(&self, from_pos: BytePos, to_pos: BytePos, m: &str) -> ! {
 115         self.fatal_span(self.mk_sp(from_pos, to_pos), m)
 116     }
 117
 118     /// Report a lexical error spanning [`from_pos`, `to_pos`).
 119     fn err_span_(&self, from_pos: BytePos, to_pos: BytePos, m: &str) {
 120         self.err_span(self.mk_sp(from_pos, to_pos), m)
 121     }
 122
 123     fn struct_fatal_span_char(
 124         &self,
 125         from_pos: BytePos,
 126         to_pos: BytePos,
 127         m: &str,
 128         c: char,
 129     ) -> DiagnosticBuilder<'a, !> {
 130         self.sess
 131             .span_diagnostic
 132             .struct_span_fatal(self.mk_sp(from_pos, to_pos), &format!("{}: {}", m, escaped_char(c)))
 133     }
 134
 135     fn struct_err_span_char(
 136         &self,
 137         from_pos: BytePos,
 138         to_pos: BytePos,
 139         m: &str,
 140         c: char,
 141     ) -> DiagnosticBuilder<'a, ErrorGuaranteed> {
 142         self.sess
 143             .span_diagnostic
 144             .struct_span_err(self.mk_sp(from_pos, to_pos), &format!("{}: {}", m, escaped_char(c)))
 145     }
 146
 147     /// Detect usages of Unicode codepoints changing the direction of the text on screen and loudly
 148     /// complain about it.
 149     fn lint_unicode_text_flow(&self, start: BytePos) {
 150         // Opening delimiter of the length 2 is not included into the comment text.
 151         let content_start = start + BytePos(2);
 152         let content = self.str_from(content_start);
 153         if contains_text_flow_control_chars(content) {
 154             let span = self.mk_sp(start, self.pos);
 155             self.sess.buffer_lint_with_diagnostic(
 156                 &TEXT_DIRECTION_CODEPOINT_IN_COMMENT,
 157                 span,
 158                 ast::CRATE_NODE_ID,
 159                 "unicode codepoint changing visible direction of text present in comment",
 160                 BuiltinLintDiagnostics::UnicodeTextFlow(span, content.to_string()),
 161             );
 162         }
 163     }
 164
 165     /// Turns simple `rustc_lexer::TokenKind` enum into a rich
 166     /// `rustc_ast::TokenKind`. This turns strings into interned
 167     /// symbols and runs additional validation.
 168     fn cook_lexer_token(&self, token: rustc_lexer::TokenKind, start: BytePos) -> Option<TokenKind> {
 169         Some(match token {
 170             rustc_lexer::TokenKind::LineComment { doc_style } => {
 171                 // Skip non-doc comments
 172                 let Some(doc_style) = doc_style else {
 173                     self.lint_unicode_text_flow(start);
 174                     return None;
 175                 };
 176
 177                 // Opening delimiter of the length 3 is not included into the symbol.
 178                 let content_start = start + BytePos(3);
 179                 let content = self.str_from(content_start);
 180                 self.cook_doc_comment(content_start, content, CommentKind::Line, doc_style)
 181             }
 182             rustc_lexer::TokenKind::BlockComment { doc_style, terminated } => {
 183                 if !terminated {
 184                     self.report_unterminated_block_comment(start, doc_style);
 185                 }
 186
 187                 // Skip non-doc comments
 188                 let Some(doc_style) = doc_style else {
 189                     self.lint_unicode_text_flow(start);
 190                     return None;
 191                 };
 192
 193                 // Opening delimiter of the length 3 and closing delimiter of the length 2
 194                 // are not included into the symbol.
 195                 let content_start = start + BytePos(3);
 196                 let content_end = self.pos - BytePos(if terminated { 2 } else { 0 });
 197                 let content = self.str_from_to(content_start, content_end);
 198                 self.cook_doc_comment(content_start, content, CommentKind::Block, doc_style)
 199             }
 200             rustc_lexer::TokenKind::Whitespace => return None,
 201             rustc_lexer::TokenKind::Ident
 202             | rustc_lexer::TokenKind::RawIdent
 203             | rustc_lexer::TokenKind::UnknownPrefix => {
 204                 let is_raw_ident = token == rustc_lexer::TokenKind::RawIdent;
 205                 let is_unknown_prefix = token == rustc_lexer::TokenKind::UnknownPrefix;
 206                 let mut ident_start = start;
 207                 if is_raw_ident {
 208                     ident_start = ident_start + BytePos(2);
 209                 }
 210                 if is_unknown_prefix {
 211                     self.report_unknown_prefix(start);
 212                 }
 213                 let sym = nfc_normalize(self.str_from(ident_start));
 214                 let span = self.mk_sp(start, self.pos);
 215                 self.sess.symbol_gallery.insert(sym, span);
 216                 if is_raw_ident {
 217                     if !sym.can_be_raw() {
 218                         self.err_span(span, &format!("`{}` cannot be a raw identifier", sym));
 219                     }
 220                     self.sess.raw_identifier_spans.borrow_mut().push(span);
 221                 }
 222                 token::Ident(sym, is_raw_ident)
 223             }
 224             rustc_lexer::TokenKind::InvalidIdent
 225                 // Do not recover an identifier with emoji if the codepoint is a confusable
 226                 // with a recoverable substitution token, like `➖`.
 227                 if !UNICODE_ARRAY
 228                     .iter()
 229                     .any(|&(c, _, _)| {
 230                         let sym = self.str_from(start);
 231                         sym.chars().count() == 1 && c == sym.chars().next().unwrap()
 232                     })
 233                      =>
 234             {
 235                 let sym = nfc_normalize(self.str_from(start));
 236                 let span = self.mk_sp(start, self.pos);
 237                 self.sess.bad_unicode_identifiers.borrow_mut().entry(sym).or_default().push(span);
 238                 token::Ident(sym, false)
 239             }
 240             rustc_lexer::TokenKind::Literal { kind, suffix_start } => {
 241                 let suffix_start = start + BytePos(suffix_start);
 242                 let (kind, symbol) = self.cook_lexer_literal(start, suffix_start, kind);
 243                 let suffix = if suffix_start < self.pos {
 244                     let string = self.str_from(suffix_start);
 245                     if string == "_" {
 246                         self.sess
 247                             .span_diagnostic
 248                             .struct_span_warn(
 249                                 self.mk_sp(suffix_start, self.pos),
 250                                 "underscore literal suffix is not allowed",
 251                             )
 252                             .warn(
 253                                 "this was previously accepted by the compiler but is \
 254                                    being phased out; it will become a hard error in \
 255                                    a future release!",
 256                             )
 257                             .note(
 258                                 "see issue #42326 \
 259                                  <https://github.com/rust-lang/rust/issues/42326> \
 260                                  for more information",
 261                             )
 262                             .emit();
 263                         None
 264                     } else {
 265                         Some(Symbol::intern(string))
 266                     }
 267                 } else {
 268                     None
 269                 };
 270                 token::Literal(token::Lit { kind, symbol, suffix })
 271             }
 272             rustc_lexer::TokenKind::Lifetime { starts_with_number } => {
 273                 // Include the leading `'` in the real identifier, for macro
 274                 // expansion purposes. See #12512 for the gory details of why
 275                 // this is necessary.
 276                 let lifetime_name = self.str_from(start);
 277                 if starts_with_number {
 278                     self.err_span_(start, self.pos, "lifetimes cannot start with a number");
 279                 }
 280                 let ident = Symbol::intern(lifetime_name);
 281                 token::Lifetime(ident)
 282             }
 283             rustc_lexer::TokenKind::Semi => token::Semi,
 284             rustc_lexer::TokenKind::Comma => token::Comma,
 285             rustc_lexer::TokenKind::Dot => token::Dot,
 286             rustc_lexer::TokenKind::OpenParen => token::OpenDelim(Delimiter::Parenthesis),
 287             rustc_lexer::TokenKind::CloseParen => token::CloseDelim(Delimiter::Parenthesis),
 288             rustc_lexer::TokenKind::OpenBrace => token::OpenDelim(Delimiter::Brace),
 289             rustc_lexer::TokenKind::CloseBrace => token::CloseDelim(Delimiter::Brace),
 290             rustc_lexer::TokenKind::OpenBracket => token::OpenDelim(Delimiter::Bracket),
 291             rustc_lexer::TokenKind::CloseBracket => token::CloseDelim(Delimiter::Bracket),
 292             rustc_lexer::TokenKind::At => token::At,
 293             rustc_lexer::TokenKind::Pound => token::Pound,
 294             rustc_lexer::TokenKind::Tilde => token::Tilde,
 295             rustc_lexer::TokenKind::Question => token::Question,
 296             rustc_lexer::TokenKind::Colon => token::Colon,
 297             rustc_lexer::TokenKind::Dollar => token::Dollar,
 298             rustc_lexer::TokenKind::Eq => token::Eq,
 299             rustc_lexer::TokenKind::Bang => token::Not,
 300             rustc_lexer::TokenKind::Lt => token::Lt,
 301             rustc_lexer::TokenKind::Gt => token::Gt,
 302             rustc_lexer::TokenKind::Minus => token::BinOp(token::Minus),
 303             rustc_lexer::TokenKind::And => token::BinOp(token::And),
 304             rustc_lexer::TokenKind::Or => token::BinOp(token::Or),
 305             rustc_lexer::TokenKind::Plus => token::BinOp(token::Plus),
 306             rustc_lexer::TokenKind::Star => token::BinOp(token::Star),
 307             rustc_lexer::TokenKind::Slash => token::BinOp(token::Slash),
 308             rustc_lexer::TokenKind::Caret => token::BinOp(token::Caret),
 309             rustc_lexer::TokenKind::Percent => token::BinOp(token::Percent),
 310
 311             rustc_lexer::TokenKind::Unknown | rustc_lexer::TokenKind::InvalidIdent => {
 312                 let c = self.str_from(start).chars().next().unwrap();
 313                 let mut err =
 314                     self.struct_err_span_char(start, self.pos, "unknown start of token", c);
 315                 // FIXME: the lexer could be used to turn the ASCII version of unicode homoglyphs,
 316                 // instead of keeping a table in `check_for_substitution`into the token. Ideally,
 317                 // this should be inside `rustc_lexer`. However, we should first remove compound
 318                 // tokens like `<<` from `rustc_lexer`, and then add fancier error recovery to it,
 319                 // as there will be less overall work to do this way.
 320                 let token = unicode_chars::check_for_substitution(self, start, c, &mut err);
 321                 if c == '\x00' {
 322                     err.help("source files must contain UTF-8 encoded text, unexpected null bytes might occur when a different encoding is used");
 323                 }
 324                 err.emit();
 325                 token?
 326             }
 327         })
 328     }
 329
 330     fn cook_doc_comment(
 331         &self,
 332         content_start: BytePos,
 333         content: &str,
 334         comment_kind: CommentKind,
 335         doc_style: DocStyle,
 336     ) -> TokenKind {
 337         if content.contains('\r') {
 338             for (idx, _) in content.char_indices().filter(|&(_, c)| c == '\r') {
 339                 self.err_span_(
 340                     content_start + BytePos(idx as u32),
 341                     content_start + BytePos(idx as u32 + 1),
 342                     match comment_kind {
 343                         CommentKind::Line => "bare CR not allowed in doc-comment",
 344                         CommentKind::Block => "bare CR not allowed in block doc-comment",
 345                     },
 346                 );
 347             }
 348         }
 349
 350         let attr_style = match doc_style {
 351             DocStyle::Outer => AttrStyle::Outer,
 352             DocStyle::Inner => AttrStyle::Inner,
 353         };
 354
 355         token::DocComment(comment_kind, attr_style, Symbol::intern(content))
 356     }
 357
 358     fn cook_lexer_literal(
 359         &self,
 360         start: BytePos,
 361         suffix_start: BytePos,
 362         kind: rustc_lexer::LiteralKind,
 363     ) -> (token::LitKind, Symbol) {
 364         // prefix means `"` or `br"` or `r###"`, ...
 365         let (lit_kind, mode, prefix_len, postfix_len) = match kind {
 366             rustc_lexer::LiteralKind::Char { terminated } => {
 367                 if !terminated {
 368                     self.sess.span_diagnostic.span_fatal_with_code(
 369                         self.mk_sp(start, suffix_start),
 370                         "unterminated character literal",
 371                         error_code!(E0762),
 372                     )
 373                 }
 374                 (token::Char, Mode::Char, 1, 1) // ' '
 375             }
 376             rustc_lexer::LiteralKind::Byte { terminated } => {
 377                 if !terminated {
 378                     self.sess.span_diagnostic.span_fatal_with_code(
 379                         self.mk_sp(start + BytePos(1), suffix_start),
 380                         "unterminated byte constant",
 381                         error_code!(E0763),
 382                     )
 383                 }
 384                 (token::Byte, Mode::Byte, 2, 1) // b' '
 385             }
 386             rustc_lexer::LiteralKind::Str { terminated } => {
 387                 if !terminated {
 388                     self.sess.span_diagnostic.span_fatal_with_code(
 389                         self.mk_sp(start, suffix_start),
 390                         "unterminated double quote string",
 391                         error_code!(E0765),
 392                     )
 393                 }
 394                 (token::Str, Mode::Str, 1, 1) // " "
 395             }
 396             rustc_lexer::LiteralKind::ByteStr { terminated } => {
 397                 if !terminated {
 398                     self.sess.span_diagnostic.span_fatal_with_code(
 399                         self.mk_sp(start + BytePos(1), suffix_start),
 400                         "unterminated double quote byte string",
 401                         error_code!(E0766),
 402                     )
 403                 }
 404                 (token::ByteStr, Mode::ByteStr, 2, 1) // b" "
 405             }
 406             rustc_lexer::LiteralKind::RawStr { n_hashes } => {
 407                 if let Some(n_hashes) = n_hashes {
 408                     let n = u32::from(n_hashes);
 409                     (token::StrRaw(n_hashes), Mode::RawStr, 2 + n, 1 + n) // r##" "##
 410                 } else {
 411                     self.report_raw_str_error(start, 1);
 412                 }
 413             }
 414             rustc_lexer::LiteralKind::RawByteStr { n_hashes } => {
 415                 if let Some(n_hashes) = n_hashes {
 416                     let n = u32::from(n_hashes);
 417                     (token::ByteStrRaw(n_hashes), Mode::RawByteStr, 3 + n, 1 + n) // br##" "##
 418                 } else {
 419                     self.report_raw_str_error(start, 2);
 420                 }
 421             }
 422             rustc_lexer::LiteralKind::Int { base, empty_int } => {
 423                 return if empty_int {
 424                     self.sess
 425                         .span_diagnostic
 426                         .struct_span_err_with_code(
 427                             self.mk_sp(start, suffix_start),
 428                             "no valid digits found for number",
 429                             error_code!(E0768),
 430                         )
 431                         .emit();
 432                     (token::Integer, sym::integer(0))
 433                 } else {
 434                     self.validate_int_literal(base, start, suffix_start);
 435                     (token::Integer, self.symbol_from_to(start, suffix_start))
 436                 };
 437             }
 438             rustc_lexer::LiteralKind::Float { base, empty_exponent } => {
 439                 if empty_exponent {
 440                     self.err_span_(start, self.pos, "expected at least one digit in exponent");
 441                 }
 442
 443                 match base {
 444                     Base::Hexadecimal => self.err_span_(
 445                         start,
 446                         suffix_start,
 447                         "hexadecimal float literal is not supported",
 448                     ),
 449                     Base::Octal => {
 450                         self.err_span_(start, suffix_start, "octal float literal is not supported")
 451                     }
 452                     Base::Binary => {
 453                         self.err_span_(start, suffix_start, "binary float literal is not supported")
 454                     }
 455                     _ => (),
 456                 }
 457
 458                 let id = self.symbol_from_to(start, suffix_start);
 459                 return (token::Float, id);
 460             }
 461         };
 462         let content_start = start + BytePos(prefix_len);
 463         let content_end = suffix_start - BytePos(postfix_len);
 464         let id = self.symbol_from_to(content_start, content_end);
 465         self.validate_literal_escape(mode, content_start, content_end, prefix_len, postfix_len);
 466         (lit_kind, id)
 467     }
 468
 469     #[inline]
 470     fn src_index(&self, pos: BytePos) -> usize {
 471         (pos - self.start_pos).to_usize()
 472     }
 473
 474     /// Slice of the source text from `start` up to but excluding `self.pos`,
 475     /// meaning the slice does not include the character `self.ch`.
 476     fn str_from(&self, start: BytePos) -> &str {
 477         self.str_from_to(start, self.pos)
 478     }
 479
 480     /// As symbol_from, with an explicit endpoint.
 481     fn symbol_from_to(&self, start: BytePos, end: BytePos) -> Symbol {
 482         debug!("taking an ident from {:?} to {:?}", start, end);
 483         Symbol::intern(self.str_from_to(start, end))
 484     }
 485
 486     /// Slice of the source text spanning from `start` up to but excluding `end`.
 487     fn str_from_to(&self, start: BytePos, end: BytePos) -> &str {
 488         &self.src[self.src_index(start)..self.src_index(end)]
 489     }
 490
 491     fn report_raw_str_error(&self, start: BytePos, prefix_len: u32) -> ! {
 492         match rustc_lexer::validate_raw_str(self.str_from(start), prefix_len) {
 493             Err(RawStrError::InvalidStarter { bad_char }) => {
 494                 self.report_non_started_raw_string(start, bad_char)
 495             }
 496             Err(RawStrError::NoTerminator { expected, found, possible_terminator_offset }) => self
 497                 .report_unterminated_raw_string(start, expected, possible_terminator_offset, found),
 498             Err(RawStrError::TooManyDelimiters { found }) => {
 499                 self.report_too_many_hashes(start, found)
 500             }
 501             Ok(()) => panic!("no error found for supposedly invalid raw string literal"),
 502         }
 503     }
 504
 505     fn report_non_started_raw_string(&self, start: BytePos, bad_char: char) -> ! {
 506         self.struct_fatal_span_char(
 507             start,
 508             self.pos,
 509             "found invalid character; only `#` is allowed in raw string delimitation",
 510             bad_char,
 511         )
 512         .emit()
 513     }
 514
 515     fn report_unterminated_raw_string(
 516         &self,
 517         start: BytePos,
 518         n_hashes: u32,
 519         possible_offset: Option<u32>,
 520         found_terminators: u32,
 521     ) -> ! {
 522         let mut err = self.sess.span_diagnostic.struct_span_fatal_with_code(
 523             self.mk_sp(start, start),
 524             "unterminated raw string",
 525             error_code!(E0748),
 526         );
 527
 528         err.span_label(self.mk_sp(start, start), "unterminated raw string");
 529
 530         if n_hashes > 0 {
 531             err.note(&format!(
 532                 "this raw string should be terminated with `\"{}`",
 533                 "#".repeat(n_hashes as usize)
 534             ));
 535         }
 536
 537         if let Some(possible_offset) = possible_offset {
 538             let lo = start + BytePos(possible_offset as u32);
 539             let hi = lo + BytePos(found_terminators as u32);
 540             let span = self.mk_sp(lo, hi);
 541             err.span_suggestion(
 542                 span,
 543                 "consider terminating the string here",
 544                 "#".repeat(n_hashes as usize),
 545                 Applicability::MaybeIncorrect,
 546             );
 547         }
 548
 549         err.emit()
 550     }
 551
 552     fn report_unterminated_block_comment(&self, start: BytePos, doc_style: Option<DocStyle>) {
 553         let msg = match doc_style {
 554             Some(_) => "unterminated block doc-comment",
 555             None => "unterminated block comment",
 556         };
 557         let last_bpos = self.pos;
 558         let mut err = self.sess.span_diagnostic.struct_span_fatal_with_code(
 559             self.mk_sp(start, last_bpos),
 560             msg,
 561             error_code!(E0758),
 562         );
 563         let mut nested_block_comment_open_idxs = vec![];
 564         let mut last_nested_block_comment_idxs = None;
 565         let mut content_chars = self.str_from(start).char_indices().peekable();
 566
 567         while let Some((idx, current_char)) = content_chars.next() {
 568             match content_chars.peek() {
 569                 Some((_, '*')) if current_char == '/' => {
 570                     nested_block_comment_open_idxs.push(idx);
 571                 }
 572                 Some((_, '/')) if current_char == '*' => {
 573                     last_nested_block_comment_idxs =
 574                         nested_block_comment_open_idxs.pop().map(|open_idx| (open_idx, idx));
 575                 }
 576                 _ => {}
 577             };
 578         }
 579
 580         if let Some((nested_open_idx, nested_close_idx)) = last_nested_block_comment_idxs {
 581             err.span_label(self.mk_sp(start, start + BytePos(2)), msg)
 582                 .span_label(
 583                     self.mk_sp(
 584                         start + BytePos(nested_open_idx as u32),
 585                         start + BytePos(nested_open_idx as u32 + 2),
 586                     ),
 587                     "...as last nested comment starts here, maybe you want to close this instead?",
 588                 )
 589                 .span_label(
 590                     self.mk_sp(
 591                         start + BytePos(nested_close_idx as u32),
 592                         start + BytePos(nested_close_idx as u32 + 2),
 593                     ),
 594                     "...and last nested comment terminates here.",
 595                 );
 596         }
 597
 598         err.emit();
 599     }
 600
 601     // RFC 3101 introduced the idea of (reserved) prefixes. As of Rust 2021,
 602     // using a (unknown) prefix is an error. In earlier editions, however, they
 603     // only result in a (allowed by default) lint, and are treated as regular
 604     // identifier tokens.
 605     fn report_unknown_prefix(&self, start: BytePos) {
 606         let prefix_span = self.mk_sp(start, self.pos);
 607         let prefix_str = self.str_from_to(start, self.pos);
 608         let msg = format!("prefix `{}` is unknown", prefix_str);
 609
 610         let expn_data = prefix_span.ctxt().outer_expn_data();
 611
 612         if expn_data.edition >= Edition::Edition2021 {
 613             // In Rust 2021, this is a hard error.
 614             let mut err = self.sess.span_diagnostic.struct_span_err(prefix_span, &msg);
 615             err.span_label(prefix_span, "unknown prefix");
 616             if prefix_str == "rb" {
 617                 err.span_suggestion_verbose(
 618                     prefix_span,
 619                     "use `br` for a raw byte string",
 620                     "br",
 621                     Applicability::MaybeIncorrect,
 622                 );
 623             } else if expn_data.is_root() {
 624                 err.span_suggestion_verbose(
 625                     prefix_span.shrink_to_hi(),
 626                     "consider inserting whitespace here",
 627                     " ",
 628                     Applicability::MaybeIncorrect,
 629                 );
 630             }
 631             err.note("prefixed identifiers and literals are reserved since Rust 2021");
 632             err.emit();
 633         } else {
 634             // Before Rust 2021, only emit a lint for migration.
 635             self.sess.buffer_lint_with_diagnostic(
 636                 &RUST_2021_PREFIXES_INCOMPATIBLE_SYNTAX,
 637                 prefix_span,
 638                 ast::CRATE_NODE_ID,
 639                 &msg,
 640                 BuiltinLintDiagnostics::ReservedPrefix(prefix_span),
 641             );
 642         }
 643     }
 644
 645     fn report_too_many_hashes(&self, start: BytePos, found: u32) -> ! {
 646         self.fatal_span_(
 647             start,
 648             self.pos,
 649             &format!(
 650                 "too many `#` symbols: raw strings may be delimited \
 651                 by up to 255 `#` symbols, but found {}",
 652                 found
 653             ),
 654         )
 655     }
 656
 657     fn validate_literal_escape(
 658         &self,
 659         mode: Mode,
 660         content_start: BytePos,
 661         content_end: BytePos,
 662         prefix_len: u32,
 663         postfix_len: u32,
 664     ) {
 665         let lit_content = self.str_from_to(content_start, content_end);
 666         unescape::unescape_literal(lit_content, mode, &mut |range, result| {
 667             // Here we only check for errors. The actual unescaping is done later.
 668             if let Err(err) = result {
 669                 let span_with_quotes = self
 670                     .mk_sp(content_start - BytePos(prefix_len), content_end + BytePos(postfix_len));
 671                 let (start, end) = (range.start as u32, range.end as u32);
 672                 let lo = content_start + BytePos(start);
 673                 let hi = lo + BytePos(end - start);
 674                 let span = self.mk_sp(lo, hi);
 675                 emit_unescape_error(
 676                     &self.sess.span_diagnostic,
 677                     lit_content,
 678                     span_with_quotes,
 679                     span,
 680                     mode,
 681                     range,
 682                     err,
 683                 );
 684             }
 685         });
 686     }
 687
 688     fn validate_int_literal(&self, base: Base, content_start: BytePos, content_end: BytePos) {
 689         let base = match base {
 690             Base::Binary => 2,
 691             Base::Octal => 8,
 692             _ => return,
 693         };
 694         let s = self.str_from_to(content_start + BytePos(2), content_end);
 695         for (idx, c) in s.char_indices() {
 696             let idx = idx as u32;
 697             if c != '_' && c.to_digit(base).is_none() {
 698                 let lo = content_start + BytePos(2 + idx);
 699                 let hi = content_start + BytePos(2 + idx + c.len_utf8() as u32);
 700                 self.err_span_(lo, hi, &format!("invalid digit for a base {} literal", base));
 701             }
 702         }
 703     }
 704 }
 705
 706 pub fn nfc_normalize(string: &str) -> Symbol {
 707     use unicode_normalization::{is_nfc_quick, IsNormalized, UnicodeNormalization};
 708     match is_nfc_quick(string.chars()) {
 709         IsNormalized::Yes => Symbol::intern(string),
 710         _ => {
 711             let normalized_str: String = string.chars().nfc().collect();
 712             Symbol::intern(&normalized_str)
 713         }
 714     }
 715 }