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