gcc/go/gofrontend/lex.cc

   1 // lex.cc -- Go frontend lexer.
   2
   3 // Copyright 2009 The Go Authors. All rights reserved.
   4 // Use of this source code is governed by a BSD-style
   5 // license that can be found in the LICENSE file.
   6
   7 #include "go-system.h"
   8
   9 #include "lex.h"
  10
  11 // Manage mapping from keywords to the Keyword codes.
  12
  13 class Keywords
  14 {
  15  public:
  16   // The structure which maps keywords to codes.
  17   struct Mapping
  18   {
  19     // Keyword string.
  20     const char* keystring;
  21     // Keyword code.
  22     Keyword keycode;
  23   };
  24
  25   // Return the parsecode corresponding to KEYSTRING, or
  26   // KEYWORD_INVALID if it is not a keyword.
  27   Keyword
  28   keyword_to_code(const char* keyword, size_t len) const;
  29
  30   // Return the string for a keyword.
  31   const char*
  32   keyword_to_string(Keyword) const;
  33
  34  private:
  35   static const Mapping mapping_[];
  36   static const int count_;
  37 };
  38
  39 // Mapping from keyword string to keyword code.  This array must be
  40 // kept in sorted order, and the order must match the Keyword enum.
  41 // Strings are looked up using bsearch.
  42
  43 const Keywords::Mapping
  44 Keywords::mapping_[] =
  45 {
  46   { NULL,          KEYWORD_INVALID },
  47   { "__asm__",     KEYWORD_ASM },
  48   { "break",       KEYWORD_BREAK },
  49   { "case",        KEYWORD_CASE },
  50   { "chan",        KEYWORD_CHAN },
  51   { "const",       KEYWORD_CONST },
  52   { "continue",    KEYWORD_CONTINUE },
  53   { "default",     KEYWORD_DEFAULT },
  54   { "defer",       KEYWORD_DEFER },
  55   { "else",        KEYWORD_ELSE },
  56   { "fallthrough", KEYWORD_FALLTHROUGH },
  57   { "for",         KEYWORD_FOR },
  58   { "func",        KEYWORD_FUNC },
  59   { "go",          KEYWORD_GO },
  60   { "goto",        KEYWORD_GOTO },
  61   { "if",          KEYWORD_IF },
  62   { "import",      KEYWORD_IMPORT },
  63   { "interface",   KEYWORD_INTERFACE },
  64   { "map",         KEYWORD_MAP },
  65   { "package",     KEYWORD_PACKAGE },
  66   { "range",       KEYWORD_RANGE },
  67   { "return",      KEYWORD_RETURN },
  68   { "select",      KEYWORD_SELECT },
  69   { "struct",      KEYWORD_STRUCT },
  70   { "switch",      KEYWORD_SWITCH },
  71   { "type",        KEYWORD_TYPE },
  72   { "var",         KEYWORD_VAR }
  73 };
  74
  75 // Number of entries in the map.
  76
  77 const int Keywords::count_ =
  78   sizeof(Keywords::mapping_) / sizeof(Keywords::mapping_[0]);
  79
  80 // Comparison function passed to bsearch.
  81
  82 extern "C"
  83 {
  84
  85 struct Keywords_search_key
  86 {
  87   const char* str;
  88   size_t len;
  89 };
  90
  91 static int
  92 keyword_compare(const void* keyv, const void* mapv)
  93 {
  94   const Keywords_search_key* key =
  95     static_cast<const Keywords_search_key*>(keyv);
  96   const Keywords::Mapping* map =
  97     static_cast<const Keywords::Mapping*>(mapv);
  98   if (map->keystring == NULL)
  99     return 1;
 100   int i = strncmp(key->str, map->keystring, key->len);
 101   if (i != 0)
 102     return i;
 103   if (map->keystring[key->len] != '\0')
 104     return -1;
 105   return 0;
 106 }
 107
 108 } // End extern "C".
 109
 110 // Convert a string to a keyword code.  Return KEYWORD_INVALID if the
 111 // string is not a keyword.
 112
 113 Keyword
 114 Keywords::keyword_to_code(const char* keyword, size_t len) const
 115 {
 116   Keywords_search_key key;
 117   key.str = keyword;
 118   key.len = len;
 119   void* mapv = bsearch(&key,
 120                        this->mapping_,
 121                        this->count_,
 122                        sizeof(this->mapping_[0]),
 123                        keyword_compare);
 124   if (mapv == NULL)
 125     return KEYWORD_INVALID;
 126   Mapping* map = static_cast<Mapping*>(mapv);
 127   return map->keycode;
 128 }
 129
 130 // Convert a keyword code to a string.
 131
 132 const char*
 133 Keywords::keyword_to_string(Keyword code) const
 134 {
 135   go_assert(code > KEYWORD_INVALID && code < this->count_);
 136   const Mapping* map = &this->mapping_[code];
 137   go_assert(map->keycode == code);
 138   return map->keystring;
 139 }
 140
 141 // There is one instance of the Keywords class.
 142
 143 static Keywords keywords;
 144
 145 // Class Token.
 146
 147 // Make a general token.
 148
 149 Token::Token(Classification classification, Location location)
 150   : classification_(classification), location_(location)
 151 {
 152 }
 153
 154 // Destroy a token.
 155
 156 Token::~Token()
 157 {
 158   this->clear();
 159 }
 160
 161 // Clear a token--release memory.
 162
 163 void
 164 Token::clear()
 165 {
 166   if (this->classification_ == TOKEN_INTEGER
 167       || this->classification_ == TOKEN_CHARACTER)
 168     mpz_clear(this->u_.integer_value);
 169   else if (this->classification_ == TOKEN_FLOAT
 170            || this->classification_ == TOKEN_IMAGINARY)
 171     mpfr_clear(this->u_.float_value);
 172 }
 173
 174 // Construct a token.
 175
 176 Token::Token(const Token& tok)
 177   : classification_(tok.classification_), location_(tok.location_)
 178 {
 179   switch (this->classification_)
 180     {
 181     case TOKEN_INVALID:
 182     case TOKEN_EOF:
 183       break;
 184     case TOKEN_KEYWORD:
 185       this->u_.keyword = tok.u_.keyword;
 186       break;
 187     case TOKEN_IDENTIFIER:
 188     case TOKEN_STRING:
 189       this->u_.string_value = tok.u_.string_value;
 190       break;
 191     case TOKEN_OPERATOR:
 192       this->u_.op = tok.u_.op;
 193       break;
 194     case TOKEN_CHARACTER:
 195     case TOKEN_INTEGER:
 196       mpz_init_set(this->u_.integer_value, tok.u_.integer_value);
 197       break;
 198     case TOKEN_FLOAT:
 199     case TOKEN_IMAGINARY:
 200       mpfr_init_set(this->u_.float_value, tok.u_.float_value, GMP_RNDN);
 201       break;
 202     default:
 203       go_unreachable();
 204     }
 205 }
 206
 207 // Assign to a token.
 208
 209 Token&
 210 Token::operator=(const Token& tok)
 211 {
 212   this->clear();
 213   this->classification_ = tok.classification_;
 214   this->location_ = tok.location_;
 215   switch (tok.classification_)
 216     {
 217     case TOKEN_INVALID:
 218     case TOKEN_EOF:
 219       break;
 220     case TOKEN_KEYWORD:
 221       this->u_.keyword = tok.u_.keyword;
 222       break;
 223     case TOKEN_IDENTIFIER:
 224       this->u_.identifier_value.name = tok.u_.identifier_value.name;
 225       this->u_.identifier_value.is_exported =
 226         tok.u_.identifier_value.is_exported;
 227       break;
 228     case TOKEN_STRING:
 229       this->u_.string_value = tok.u_.string_value;
 230       break;
 231     case TOKEN_OPERATOR:
 232       this->u_.op = tok.u_.op;
 233       break;
 234     case TOKEN_CHARACTER:
 235     case TOKEN_INTEGER:
 236       mpz_init_set(this->u_.integer_value, tok.u_.integer_value);
 237       break;
 238     case TOKEN_FLOAT:
 239     case TOKEN_IMAGINARY:
 240       mpfr_init_set(this->u_.float_value, tok.u_.float_value, GMP_RNDN);
 241       break;
 242     default:
 243       go_unreachable();
 244     }
 245   return *this;
 246 }
 247
 248 // Print the token for debugging.
 249
 250 void
 251 Token::print(FILE* file) const
 252 {
 253   switch (this->classification_)
 254     {
 255     case TOKEN_INVALID:
 256       fprintf(file, "invalid");
 257       break;
 258     case TOKEN_EOF:
 259       fprintf(file, "EOF");
 260       break;
 261     case TOKEN_KEYWORD:
 262       fprintf(file, "keyword %s", keywords.keyword_to_string(this->u_.keyword));
 263       break;
 264     case TOKEN_IDENTIFIER:
 265       fprintf(file, "identifier \"%s\"", this->u_.string_value->c_str());
 266       break;
 267     case TOKEN_STRING:
 268       fprintf(file, "quoted string \"%s\"", this->u_.string_value->c_str());
 269       break;
 270     case TOKEN_CHARACTER:
 271       fprintf(file, "character ");
 272       mpz_out_str(file, 10, this->u_.integer_value);
 273       break;
 274     case TOKEN_INTEGER:
 275       fprintf(file, "integer ");
 276       mpz_out_str(file, 10, this->u_.integer_value);
 277       break;
 278     case TOKEN_FLOAT:
 279       fprintf(file, "float ");
 280       mpfr_out_str(file, 10, 0, this->u_.float_value, GMP_RNDN);
 281       break;
 282     case TOKEN_IMAGINARY:
 283       fprintf(file, "imaginary ");
 284       mpfr_out_str(file, 10, 0, this->u_.float_value, GMP_RNDN);
 285       break;
 286     case TOKEN_OPERATOR:
 287       fprintf(file, "operator ");
 288       switch (this->u_.op)
 289         {
 290         case OPERATOR_INVALID:
 291           fprintf(file, "invalid");
 292           break;
 293         case OPERATOR_OROR:
 294           fprintf(file, "||");
 295           break;
 296         case OPERATOR_ANDAND:
 297           fprintf(file, "&&");
 298           break;
 299         case OPERATOR_EQEQ:
 300           fprintf(file, "==");
 301           break;
 302         case OPERATOR_NOTEQ:
 303           fprintf(file, "!=");
 304           break;
 305         case OPERATOR_LT:
 306           fprintf(file, "<");
 307           break;
 308         case OPERATOR_LE:
 309           fprintf(file, "<=");
 310           break;
 311         case OPERATOR_GT:
 312           fprintf(file, ">");
 313           break;
 314         case OPERATOR_GE:
 315           fprintf(file, ">=");
 316           break;
 317         case OPERATOR_PLUS:
 318           fprintf(file, "+");
 319           break;
 320         case OPERATOR_MINUS:
 321           fprintf(file, "-");
 322           break;
 323         case OPERATOR_OR:
 324           fprintf(file, "|");
 325           break;
 326         case OPERATOR_XOR:
 327           fprintf(file, "^");
 328           break;
 329         case OPERATOR_MULT:
 330           fprintf(file, "*");
 331           break;
 332         case OPERATOR_DIV:
 333           fprintf(file, "/");
 334           break;
 335         case OPERATOR_MOD:
 336           fprintf(file, "%%");
 337           break;
 338         case OPERATOR_LSHIFT:
 339           fprintf(file, "<<");
 340           break;
 341         case OPERATOR_RSHIFT:
 342           fprintf(file, ">>");
 343           break;
 344         case OPERATOR_AND:
 345           fprintf(file, "&");
 346           break;
 347         case OPERATOR_BITCLEAR:
 348           fprintf(file, "&^");
 349           break;
 350         case OPERATOR_NOT:
 351           fprintf(file, "!");
 352           break;
 353         case OPERATOR_CHANOP:
 354           fprintf(file, "<-");
 355           break;
 356         case OPERATOR_EQ:
 357           fprintf(file, "=");
 358           break;
 359         case OPERATOR_PLUSEQ:
 360           fprintf(file, "+=");
 361           break;
 362         case OPERATOR_MINUSEQ:
 363           fprintf(file, "-=");
 364           break;
 365         case OPERATOR_OREQ:
 366           fprintf(file, "|=");
 367           break;
 368         case OPERATOR_XOREQ:
 369           fprintf(file, "^=");
 370           break;
 371         case OPERATOR_MULTEQ:
 372           fprintf(file, "*=");
 373           break;
 374         case OPERATOR_DIVEQ:
 375           fprintf(file, "/=");
 376           break;
 377         case OPERATOR_MODEQ:
 378           fprintf(file, "%%=");
 379           break;
 380         case OPERATOR_LSHIFTEQ:
 381           fprintf(file, "<<=");
 382           break;
 383         case OPERATOR_RSHIFTEQ:
 384           fprintf(file, ">>=");
 385           break;
 386         case OPERATOR_ANDEQ:
 387           fprintf(file, "&=");
 388           break;
 389         case OPERATOR_BITCLEAREQ:
 390           fprintf(file, "&^=");
 391           break;
 392         case OPERATOR_PLUSPLUS:
 393           fprintf(file, "++");
 394           break;
 395         case OPERATOR_MINUSMINUS:
 396           fprintf(file, "--");
 397           break;
 398         case OPERATOR_COLON:
 399           fprintf(file, ":");
 400           break;
 401         case OPERATOR_COLONEQ:
 402           fprintf(file, ":=");
 403           break;
 404         case OPERATOR_SEMICOLON:
 405           fprintf(file, ";");
 406           break;
 407         case OPERATOR_DOT:
 408           fprintf(file, ".");
 409           break;
 410         case OPERATOR_COMMA:
 411           fprintf(file, ",");
 412           break;
 413         case OPERATOR_LPAREN:
 414           fprintf(file, "(");
 415           break;
 416         case OPERATOR_RPAREN:
 417           fprintf(file, ")");
 418           break;
 419         case OPERATOR_LCURLY:
 420           fprintf(file, "{");
 421           break;
 422         case OPERATOR_RCURLY:
 423           fprintf(file, "}");
 424           break;
 425         case OPERATOR_LSQUARE:
 426           fprintf(file, "[");
 427           break;
 428         case OPERATOR_RSQUARE:
 429           fprintf(file, "]");
 430           break;
 431         default:
 432           go_unreachable();
 433         }
 434       break;
 435     default:
 436       go_unreachable();
 437     }
 438 }
 439
 440 // Class Lex.
 441
 442 Lex::Lex(const char* input_file_name, FILE* input_file, Linemap* linemap)
 443   : input_file_name_(input_file_name), input_file_(input_file),
 444     linemap_(linemap), linebuf_(NULL), linebufsize_(120), linesize_(0),
 445     lineoff_(0), lineno_(0), add_semi_at_eol_(false), saw_nointerface_(false),
 446     extern_()
 447 {
 448   this->linebuf_ = new char[this->linebufsize_];
 449   this->linemap_->start_file(input_file_name, 0);
 450 }
 451
 452 Lex::~Lex()
 453 {
 454   delete[] this->linebuf_;
 455 }
 456
 457 // Read a new line from the file.
 458
 459 ssize_t
 460 Lex::get_line()
 461 {
 462   char* buf = this->linebuf_;
 463   size_t size = this->linebufsize_;
 464
 465   FILE* file = this->input_file_;
 466   size_t cur = 0;
 467   while (true)
 468     {
 469       int c = getc(file);
 470       if (c == EOF)
 471         {
 472           if (cur == 0)
 473             return -1;
 474           break;
 475         }
 476       if (cur + 1 >= size)
 477         {
 478           size_t ns = 2 * size + 1;
 479           if (ns < size || static_cast<ssize_t>(ns) < 0)
 480             error_at(this->location(), "out of memory");
 481           char* nb = new char[ns];
 482           memcpy(nb, buf, cur);
 483           delete[] buf;
 484           buf = nb;
 485           size = ns;
 486         }
 487       buf[cur] = c;
 488       ++cur;
 489
 490       if (c == '\n')
 491         break;
 492     }
 493
 494   buf[cur] = '\0';
 495
 496   this->linebuf_ = buf;
 497   this->linebufsize_ = size;
 498
 499   return cur;
 500 }
 501
 502 // See if we need to read a new line.  Return true if there is a new
 503 // line, false if we are at EOF.
 504
 505 bool
 506 Lex::require_line()
 507 {
 508   if (this->lineoff_ < this->linesize_)
 509     return true;
 510
 511   ssize_t got = this->get_line();
 512   if (got < 0)
 513     return false;
 514   ++this->lineno_;
 515   this->linesize_= got;
 516   this->lineoff_ = 0;
 517
 518   this->linemap_->start_line(this->lineno_, this->linesize_);
 519
 520   return true;
 521 }
 522
 523 // Get the current location.
 524
 525 Location
 526 Lex::location() const
 527 {
 528   return this->linemap_->get_location(this->lineoff_ + 1);
 529 }
 530
 531 // Get a location slightly before the current one.  This is used for
 532 // slightly more efficient handling of operator tokens.
 533
 534 Location
 535 Lex::earlier_location(int chars) const
 536 {
 537   return this->linemap_->get_location(this->lineoff_ + 1 - chars);
 538 }
 539
 540 // Get the next token.
 541
 542 Token
 543 Lex::next_token()
 544 {
 545   bool saw_cpp_comment = false;
 546   while (true)
 547     {
 548       if (!this->require_line())
 549         {
 550           bool add_semi_at_eol = this->add_semi_at_eol_;
 551           this->add_semi_at_eol_ = false;
 552           if (add_semi_at_eol)
 553             return this->make_operator(OPERATOR_SEMICOLON, 1);
 554           return this->make_eof_token();
 555         }
 556
 557       if (!saw_cpp_comment)
 558         this->extern_.clear();
 559       saw_cpp_comment = false;
 560
 561       const char* p = this->linebuf_ + this->lineoff_;
 562       const char* pend = this->linebuf_ + this->linesize_;
 563
 564       while (p < pend)
 565         {
 566           unsigned char cc = *p;
 567           switch (cc)
 568             {
 569             case ' ': case '\t': case '\r':
 570               ++p;
 571               // Skip whitespace quickly.
 572               while (*p == ' ' || *p == '\t' || *p == '\r')
 573                 ++p;
 574               break;
 575
 576             case '\n':
 577               {
 578                 ++p;
 579                 bool add_semi_at_eol = this->add_semi_at_eol_;
 580                 this->add_semi_at_eol_ = false;
 581                 if (add_semi_at_eol)
 582                   {
 583                     this->lineoff_ = p - this->linebuf_;
 584                     return this->make_operator(OPERATOR_SEMICOLON, 1);
 585                   }
 586               }
 587               break;
 588
 589             case '/':
 590               if (p[1] == '/')
 591                 {
 592                   this->lineoff_ = p + 2 - this->linebuf_;
 593                   this->skip_cpp_comment();
 594                   p = pend;
 595                   if (p[-1] == '\n' && this->add_semi_at_eol_)
 596                     --p;
 597                   saw_cpp_comment = true;
 598                 }
 599               else if (p[1] == '*')
 600                 {
 601                   this->lineoff_ = p - this->linebuf_;
 602                   Location location = this->location();
 603                   if (!this->skip_c_comment())
 604                     return Token::make_invalid_token(location);
 605                   p = this->linebuf_ + this->lineoff_;
 606                   pend = this->linebuf_ + this->linesize_;
 607                 }
 608               else if (p[1] == '=')
 609                 {
 610                   this->add_semi_at_eol_ = false;
 611                   this->lineoff_ = p + 2 - this->linebuf_;
 612                   return this->make_operator(OPERATOR_DIVEQ, 2);
 613                 }
 614               else
 615                 {
 616                   this->add_semi_at_eol_ = false;
 617                   this->lineoff_ = p + 1 - this->linebuf_;
 618                   return this->make_operator(OPERATOR_DIV, 1);
 619                 }
 620               break;
 621
 622             case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
 623             case 'G': case 'H': case 'I': case 'J': case 'K': case 'L':
 624             case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R':
 625             case 'S': case 'T': case 'U': case 'V': case 'W': case 'X':
 626             case 'Y': case 'Z':
 627             case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
 628             case 'g': case 'h': case 'i': case 'j': case 'k': case 'l':
 629             case 'm': case 'n': case 'o': case 'p': case 'q': case 'r':
 630             case 's': case 't': case 'u': case 'v': case 'w': case 'x':
 631             case 'y': case 'z':
 632             case '_':
 633               this->lineoff_ = p - this->linebuf_;
 634               return this->gather_identifier();
 635
 636             case '0': case '1': case '2': case '3': case '4':
 637             case '5': case '6': case '7': case '8': case '9':
 638               this->add_semi_at_eol_ = true;
 639               this->lineoff_ = p - this->linebuf_;
 640               return this->gather_number();
 641
 642             case '\'':
 643               this->add_semi_at_eol_ = true;
 644               this->lineoff_ = p - this->linebuf_;
 645               return this->gather_character();
 646
 647             case '"':
 648               this->add_semi_at_eol_ = true;
 649               this->lineoff_ = p - this->linebuf_;
 650               return this->gather_string();
 651
 652             case '`':
 653               this->add_semi_at_eol_ = true;
 654               this->lineoff_ = p - this->linebuf_;
 655               return this->gather_raw_string();
 656
 657             case '<':
 658             case '>':
 659             case '&':
 660               if (p + 2 < pend)
 661                 {
 662                   this->add_semi_at_eol_ = false;
 663                   Operator op = this->three_character_operator(cc, p[1], p[2]);
 664                   if (op != OPERATOR_INVALID)
 665                     {
 666                       this->lineoff_ = p + 3 - this->linebuf_;
 667                       return this->make_operator(op, 3);
 668                     }
 669                 }
 670               // Fall through.
 671             case '|':
 672             case '=':
 673             case '!':
 674             case '+':
 675             case '-':
 676             case '^':
 677             case '*':
 678               // '/' handled above.
 679             case '%':
 680             case ':':
 681             case ';':
 682             case ',':
 683             case '(': case ')':
 684             case '{': case '}':
 685             case '[': case ']':
 686               {
 687                 this->add_semi_at_eol_ = false;
 688                 Operator op = this->two_character_operator(cc, p[1]);
 689                 int chars;
 690                 if (op != OPERATOR_INVALID)
 691                   {
 692                     ++p;
 693                     chars = 2;
 694                   }
 695                 else
 696                   {
 697                     op = this->one_character_operator(cc);
 698                     chars = 1;
 699                   }
 700                 this->lineoff_ = p + 1 - this->linebuf_;
 701                 return this->make_operator(op, chars);
 702               }
 703
 704             case '.':
 705               if (p[1] >= '0' && p[1] <= '9')
 706                 {
 707                   this->add_semi_at_eol_ = true;
 708                   this->lineoff_ = p - this->linebuf_;
 709                   return this->gather_number();
 710                 }
 711               if (p[1] == '.' && p[2] == '.')
 712                 {
 713                   this->add_semi_at_eol_ = false;
 714                   this->lineoff_ = p + 3 - this->linebuf_;
 715                   return this->make_operator(OPERATOR_ELLIPSIS, 3);
 716                 }
 717               this->add_semi_at_eol_ = false;
 718               this->lineoff_ = p + 1 - this->linebuf_;
 719               return this->make_operator(OPERATOR_DOT, 1);
 720
 721             default:
 722               {
 723                 unsigned int ci;
 724                 bool issued_error;
 725                 this->lineoff_ = p - this->linebuf_;
 726                 const char *pnext = this->advance_one_utf8_char(p, &ci,
 727                                                                 &issued_error);
 728
 729                 // Ignore byte order mark at start of file.
 730                 if (ci == 0xfeff)
 731                   {
 732                     p = pnext;
 733                     break;
 734                   }
 735
 736                 if (Lex::is_unicode_letter(ci))
 737                   return this->gather_identifier();
 738
 739                 if (!issued_error)
 740                   error_at(this->location(),
 741                            "invalid character 0x%x in input file",
 742                            ci);
 743
 744                 p = pend;
 745
 746                 break;
 747               }
 748             }
 749         }
 750
 751       this->lineoff_ = p - this->linebuf_;
 752     }
 753 }
 754
 755 // Fetch one UTF-8 character from a string.  Set *VALUE to the value.
 756 // Return the number of bytes read from the string.  Returns 0 if the
 757 // string does not point to a valid UTF-8 character.
 758
 759 int
 760 Lex::fetch_char(const char* p, unsigned int* value)
 761 {
 762   unsigned char c = *p;
 763   if (c <= 0x7f)
 764     {
 765       *value = c;
 766       return 1;
 767     }
 768   else if ((c & 0xe0) == 0xc0
 769            && (p[1] & 0xc0) == 0x80)
 770     {
 771       *value = (((c & 0x1f) << 6)
 772                 + (p[1] & 0x3f));
 773       if (*value <= 0x7f)
 774         {
 775           *value = 0xfffd;
 776           return 0;
 777         }
 778       return 2;
 779     }
 780   else if ((c & 0xf0) == 0xe0
 781            && (p[1] & 0xc0) == 0x80
 782            && (p[2] & 0xc0) == 0x80)
 783     {
 784       *value = (((c & 0xf) << 12)
 785                 + ((p[1] & 0x3f) << 6)
 786                 + (p[2] & 0x3f));
 787       if (*value <= 0x7ff)
 788         {
 789           *value = 0xfffd;
 790           return 0;
 791         }
 792       return 3;
 793     }
 794   else if ((c & 0xf8) == 0xf0
 795            && (p[1] & 0xc0) == 0x80
 796            && (p[2] & 0xc0) == 0x80
 797            && (p[3] & 0xc0) == 0x80)
 798     {
 799       *value = (((c & 0x7) << 18)
 800                 + ((p[1] & 0x3f) << 12)
 801                 + ((p[2] & 0x3f) << 6)
 802                 + (p[3] & 0x3f));
 803       if (*value <= 0xffff)
 804         {
 805           *value = 0xfffd;
 806           return 0;
 807         }
 808       return 4;
 809     }
 810   else
 811     {
 812       /* Invalid encoding. Return the Unicode replacement
 813          character.  */
 814       *value = 0xfffd;
 815       return 0;
 816     }
 817 }
 818
 819 // Advance one UTF-8 character.  Return the pointer beyond the
 820 // character.  Set *VALUE to the value.  Set *ISSUED_ERROR if an error
 821 // was issued.
 822
 823 const char*
 824 Lex::advance_one_utf8_char(const char* p, unsigned int* value,
 825                            bool* issued_error)
 826 {
 827   *issued_error = false;
 828
 829   if (*p == '\0')
 830     {
 831       error_at(this->location(), "invalid NUL byte");
 832       *issued_error = true;
 833       *value = 0;
 834       return p + 1;
 835     }
 836
 837   int adv = Lex::fetch_char(p, value);
 838   if (adv == 0)
 839     {
 840       error_at(this->location(), "invalid UTF-8 encoding");
 841       *issued_error = true;
 842       return p + 1;
 843     }
 844
 845   // Warn about byte order mark, except at start of file.
 846   if (*value == 0xfeff && (this->lineno_ != 1 || this->lineoff_ != 0))
 847     {
 848       error_at(this->location(), "Unicode (UTF-8) BOM in middle of file");
 849       *issued_error = true;
 850     }
 851
 852   return p + adv;
 853 }
 854
 855 // Pick up an identifier.
 856
 857 Token
 858 Lex::gather_identifier()
 859 {
 860   const char* pstart = this->linebuf_ + this->lineoff_;
 861   const char* p = pstart;
 862   const char* pend = this->linebuf_ + this->linesize_;
 863   bool is_first = true;
 864   bool is_exported = false;
 865   bool has_non_ascii_char = false;
 866   std::string buf;
 867   while (p < pend)
 868     {
 869       unsigned char cc = *p;
 870       if (cc <= 0x7f)
 871         {
 872           if ((cc < 'A' || cc > 'Z')
 873               && (cc < 'a' || cc > 'z')
 874               && cc != '_'
 875               && (cc < '0' || cc > '9'))
 876             break;
 877           ++p;
 878           if (is_first)
 879             {
 880               is_exported = cc >= 'A' && cc <= 'Z';
 881               is_first = false;
 882             }
 883           if (has_non_ascii_char)
 884             buf.push_back(cc);
 885         }
 886       else
 887         {
 888           unsigned int ci;
 889           bool issued_error;
 890           this->lineoff_ = p - this->linebuf_;
 891           const char* pnext = this->advance_one_utf8_char(p, &ci,
 892                                                           &issued_error);
 893           bool is_invalid = false;
 894           if (!Lex::is_unicode_letter(ci) && !Lex::is_unicode_digit(ci))
 895             {
 896               // There is no valid place for a non-ASCII character
 897               // other than an identifier, so we get better error
 898               // handling behaviour if we swallow this character after
 899               // giving an error.
 900               if (!issued_error)
 901                 error_at(this->location(),
 902                          "invalid character 0x%x in identifier",
 903                          ci);
 904               is_invalid = true;
 905             }
 906           if (is_first)
 907             {
 908               is_exported = Lex::is_unicode_uppercase(ci);
 909               is_first = false;
 910             }
 911           if (!has_non_ascii_char)
 912             {
 913               buf.assign(pstart, p - pstart);
 914               has_non_ascii_char = true;
 915             }
 916           if (is_invalid && !Lex::is_invalid_identifier(buf))
 917             buf.append("$INVALID$");
 918           buf.append(p, pnext - p);
 919           p = pnext;
 920         }
 921     }
 922   Location location = this->location();
 923   this->add_semi_at_eol_ = true;
 924   this->lineoff_ = p - this->linebuf_;
 925   if (has_non_ascii_char)
 926     return Token::make_identifier_token(buf, is_exported, location);
 927   else
 928     {
 929       Keyword code = keywords.keyword_to_code(pstart, p - pstart);
 930       if (code == KEYWORD_INVALID)
 931         return Token::make_identifier_token(std::string(pstart, p - pstart),
 932                                             is_exported, location);
 933       else
 934         {
 935           switch (code)
 936             {
 937             case KEYWORD_BREAK:
 938             case KEYWORD_CONTINUE:
 939             case KEYWORD_FALLTHROUGH:
 940             case KEYWORD_RETURN:
 941               break;
 942             default:
 943               this->add_semi_at_eol_ = false;
 944               break;
 945             }
 946           return Token::make_keyword_token(code, location);
 947         }
 948     }
 949 }
 950
 951 // Return whether C is a hex digit.
 952
 953 bool
 954 Lex::is_hex_digit(char c)
 955 {
 956   return ((c >= '0' && c <= '9')
 957           || (c >= 'A' && c <= 'F')
 958           || (c >= 'a' && c <= 'f'));
 959 }
 960
 961 // Return whether an exponent could start at P.
 962
 963 bool
 964 Lex::could_be_exponent(const char* p, const char* pend)
 965 {
 966   if (*p != 'e' && *p != 'E')
 967     return false;
 968   ++p;
 969   if (p >= pend)
 970     return false;
 971   if (*p == '+' || *p == '-')
 972     {
 973       ++p;
 974       if (p >= pend)
 975         return false;
 976     }
 977   return *p >= '0' && *p <= '9';
 978 }
 979
 980 // Pick up a number.
 981
 982 Token
 983 Lex::gather_number()
 984 {
 985   const char* pstart = this->linebuf_ + this->lineoff_;
 986   const char* p = pstart;
 987   const char* pend = this->linebuf_ + this->linesize_;
 988
 989   Location location = this->location();
 990
 991   bool neg = false;
 992   if (*p == '+')
 993     ++p;
 994   else if (*p == '-')
 995     {
 996       ++p;
 997       neg = true;
 998     }
 999
1000   const char* pnum = p;
1001   if (*p == '0')
1002     {
1003       int base;
1004       if ((p[1] == 'x' || p[1] == 'X')
1005           && Lex::is_hex_digit(p[2]))
1006         {
1007           base = 16;
1008           p += 2;
1009           pnum = p;
1010           while (p < pend)
1011             {
1012               if (!Lex::is_hex_digit(*p))
1013                 break;
1014               ++p;
1015             }
1016         }
1017       else
1018         {
1019           base = 8;
1020           pnum = p;
1021           while (p < pend)
1022             {
1023               if (*p < '0' || *p > '7')
1024                 break;
1025               ++p;
1026             }
1027         }
1028
1029       // A partial token that looks like an octal literal might actually be the
1030       // beginning of a floating-point or imaginary literal.
1031       if (base == 16 || (*p != '.' && *p != 'i' && !Lex::could_be_exponent(p, pend)))
1032         {
1033           std::string s(pnum, p - pnum);
1034           mpz_t val;
1035           int r = mpz_init_set_str(val, s.c_str(), base);
1036           go_assert(r == 0);
1037
1038           if (neg)
1039             mpz_neg(val, val);
1040
1041           this->lineoff_ = p - this->linebuf_;
1042           Token ret = Token::make_integer_token(val, location);
1043           mpz_clear(val);
1044           return ret;
1045         }
1046     }
1047
1048   while (p < pend)
1049     {
1050       if (*p < '0' || *p > '9')
1051         break;
1052       ++p;
1053     }
1054
1055   if (*p != '.' && *p != 'i' && !Lex::could_be_exponent(p, pend))
1056     {
1057       std::string s(pnum, p - pnum);
1058       mpz_t val;
1059       int r = mpz_init_set_str(val, s.c_str(), 10);
1060       go_assert(r == 0);
1061
1062       if (neg)
1063         mpz_neg(val, val);
1064
1065       this->lineoff_ = p - this->linebuf_;
1066       Token ret = Token::make_integer_token(val, location);
1067       mpz_clear(val);
1068       return ret;
1069     }
1070
1071   if (*p != 'i')
1072     {
1073       bool dot = *p == '.';
1074
1075       ++p;
1076
1077       if (!dot)
1078         {
1079           if (*p == '+' || *p == '-')
1080             ++p;
1081         }
1082
1083       while (p < pend)
1084         {
1085           if (*p < '0' || *p > '9')
1086             break;
1087           ++p;
1088         }
1089
1090       if (dot && Lex::could_be_exponent(p, pend))
1091         {
1092           ++p;
1093           if (*p == '+' || *p == '-')
1094             ++p;
1095           while (p < pend)
1096             {
1097               if (*p < '0' || *p > '9')
1098                 break;
1099               ++p;
1100             }
1101         }
1102     }
1103
1104   std::string s(pnum, p - pnum);
1105   mpfr_t val;
1106   int r = mpfr_init_set_str(val, s.c_str(), 10, GMP_RNDN);
1107   go_assert(r == 0);
1108
1109   if (neg)
1110     mpfr_neg(val, val, GMP_RNDN);
1111
1112   bool is_imaginary = *p == 'i';
1113   if (is_imaginary)
1114     ++p;
1115
1116   this->lineoff_ = p - this->linebuf_;
1117   if (is_imaginary)
1118     {
1119       Token ret = Token::make_imaginary_token(val, location);
1120       mpfr_clear(val);
1121       return ret;
1122     }
1123   else
1124     {
1125       Token ret = Token::make_float_token(val, location);
1126       mpfr_clear(val);
1127       return ret;
1128     }
1129 }
1130
1131 // Advance one character, possibly escaped.  Return the pointer beyond
1132 // the character.  Set *VALUE to the character.  Set *IS_CHARACTER if
1133 // this is a character (e.g., 'a' or '\u1234') rather than a byte
1134 // value (e.g., '\001').
1135
1136 const char*
1137 Lex::advance_one_char(const char* p, bool is_single_quote, unsigned int* value,
1138                       bool* is_character)
1139 {
1140   *value = 0;
1141   *is_character = true;
1142   if (*p != '\\')
1143     {
1144       bool issued_error;
1145       const char* ret = this->advance_one_utf8_char(p, value, &issued_error);
1146       if (is_single_quote
1147           && (*value == '\'' || *value == '\n')
1148           && !issued_error)
1149         error_at(this->location(), "invalid character literal");
1150       return ret;
1151     }
1152   else
1153     {
1154       ++p;
1155       switch (*p)
1156         {
1157         case '0': case '1': case '2': case '3':
1158         case '4': case '5': case '6': case '7':
1159           *is_character = false;
1160           if (p[1] >= '0' && p[1] <= '7'
1161               && p[2] >= '0' && p[2] <= '7')
1162             {
1163               *value = ((Lex::octal_value(p[0]) << 6)
1164                         + (Lex::octal_value(p[1]) << 3)
1165                         + Lex::octal_value(p[2]));
1166               if (*value > 255)
1167                 {
1168                   error_at(this->location(), "invalid octal constant");
1169                   *value = 255;
1170                 }
1171               return p + 3;
1172             }
1173               error_at(this->location(), "invalid octal character");
1174           return (p[1] >= '0' && p[1] <= '7'
1175                   ? p + 2
1176                   : p + 1);
1177
1178         case 'x':
1179         case 'X':
1180           *is_character = false;
1181           if (Lex::is_hex_digit(p[1]) && Lex::is_hex_digit(p[2]))
1182             {
1183               *value = (hex_value(p[1]) << 4) + hex_value(p[2]);
1184               return p + 3;
1185             }
1186           error_at(this->location(), "invalid hex character");
1187           return (Lex::is_hex_digit(p[1])
1188                   ? p + 2
1189                   : p + 1);
1190
1191         case 'a':
1192           *value = '\a';
1193           return p + 1;
1194         case 'b':
1195           *value = '\b';
1196           return p + 1;
1197         case 'f':
1198           *value = '\f';
1199           return p + 1;
1200         case 'n':
1201           *value = '\n';
1202           return p + 1;
1203         case 'r':
1204           *value = '\r';
1205           return p + 1;
1206         case 't':
1207           *value = '\t';
1208           return p + 1;
1209         case 'v':
1210           *value = '\v';
1211           return p + 1;
1212         case '\\':
1213           *value = '\\';
1214           return p + 1;
1215         case '\'':
1216           if (!is_single_quote)
1217             error_at(this->location(), "invalid quoted character");
1218           *value = '\'';
1219           return p + 1;
1220         case '"':
1221           if (is_single_quote)
1222             error_at(this->location(), "invalid quoted character");
1223           *value = '"';
1224           return p + 1;
1225
1226         case 'u':
1227           if (Lex::is_hex_digit(p[1]) && Lex::is_hex_digit(p[2])
1228               && Lex::is_hex_digit(p[3]) && Lex::is_hex_digit(p[4]))
1229             {
1230               *value = ((hex_value(p[1]) << 12)
1231                         + (hex_value(p[2]) << 8)
1232                         + (hex_value(p[3]) << 4)
1233                         + hex_value(p[4]));
1234               if (*value >= 0xd800 && *value < 0xe000)
1235                 {
1236                   error_at(this->location(),
1237                            "invalid unicode code point 0x%x",
1238                            *value);
1239                   // Use the replacement character.
1240                   *value = 0xfffd;
1241                 }
1242               return p + 5;
1243             }
1244           error_at(this->location(), "invalid little unicode code point");
1245           return p + 1;
1246
1247         case 'U':
1248           if (Lex::is_hex_digit(p[1]) && Lex::is_hex_digit(p[2])
1249               && Lex::is_hex_digit(p[3]) && Lex::is_hex_digit(p[4])
1250               && Lex::is_hex_digit(p[5]) && Lex::is_hex_digit(p[6])
1251               && Lex::is_hex_digit(p[7]) && Lex::is_hex_digit(p[8]))
1252             {
1253               *value = ((hex_value(p[1]) << 28)
1254                         + (hex_value(p[2]) << 24)
1255                         + (hex_value(p[3]) << 20)
1256                         + (hex_value(p[4]) << 16)
1257                         + (hex_value(p[5]) << 12)
1258                         + (hex_value(p[6]) << 8)
1259                         + (hex_value(p[7]) << 4)
1260                         + hex_value(p[8]));
1261               if (*value > 0x10ffff
1262                   || (*value >= 0xd800 && *value < 0xe000))
1263                 {
1264                   error_at(this->location(), "invalid unicode code point 0x%x",
1265                            *value);
1266                   // Use the replacement character.
1267                   *value = 0xfffd;
1268                 }
1269               return p + 9;
1270             }
1271           error_at(this->location(), "invalid big unicode code point");
1272           return p + 1;
1273
1274         default:
1275           error_at(this->location(), "invalid character after %<\\%>");
1276           *value = *p;
1277           return p + 1;
1278         }
1279     }
1280 }
1281
1282 // Append V to STR.  IS_CHARACTER is true for a character which should
1283 // be stored in UTF-8, false for a general byte value which should be
1284 // stored directly.
1285
1286 void
1287 Lex::append_char(unsigned int v, bool is_character, std::string* str,
1288                  Location location)
1289 {
1290   char buf[4];
1291   size_t len;
1292   if (v <= 0x7f || !is_character)
1293     {
1294       buf[0] = v;
1295       len = 1;
1296     }
1297   else if (v <= 0x7ff)
1298     {
1299       buf[0] = 0xc0 + (v >> 6);
1300       buf[1] = 0x80 + (v & 0x3f);
1301       len = 2;
1302     }
1303   else
1304     {
1305       if (v > 0x10ffff)
1306         {
1307           warning_at(location, 0,
1308                      "unicode code point 0x%x out of range in string", v);
1309           // Turn it into the "replacement character".
1310           v = 0xfffd;
1311         }
1312       if (v >= 0xd800 && v < 0xe000)
1313         {
1314           warning_at(location, 0,
1315                      "unicode code point 0x%x is invalid surrogate pair", v);
1316           v = 0xfffd;
1317         }
1318       if (v <= 0xffff)
1319         {
1320           buf[0] = 0xe0 + (v >> 12);
1321           buf[1] = 0x80 + ((v >> 6) & 0x3f);
1322           buf[2] = 0x80 + (v & 0x3f);
1323           len = 3;
1324         }
1325       else
1326         {
1327           buf[0] = 0xf0 + (v >> 18);
1328           buf[1] = 0x80 + ((v >> 12) & 0x3f);
1329           buf[2] = 0x80 + ((v >> 6) & 0x3f);
1330           buf[3] = 0x80 + (v & 0x3f);
1331           len = 4;
1332         }
1333     }
1334   str->append(buf, len);
1335 }
1336
1337 // Pick up a character literal.
1338
1339 Token
1340 Lex::gather_character()
1341 {
1342   ++this->lineoff_;
1343   const char* pstart = this->linebuf_ + this->lineoff_;
1344   const char* p = pstart;
1345
1346   unsigned int value;
1347   bool is_character;
1348   p = this->advance_one_char(p, true, &value, &is_character);
1349
1350   if (*p != '\'')
1351     {
1352       error_at(this->location(), "unterminated character constant");
1353       this->lineoff_ = p - this->linebuf_;
1354       return this->make_invalid_token();
1355     }
1356
1357   mpz_t val;
1358   mpz_init_set_ui(val, value);
1359
1360   Location location = this->location();
1361   this->lineoff_ = p + 1 - this->linebuf_;
1362   Token ret = Token::make_character_token(val, location);
1363   mpz_clear(val);
1364   return ret;
1365 }
1366
1367 // Pick up a quoted string.
1368
1369 Token
1370 Lex::gather_string()
1371 {
1372   const char* pstart = this->linebuf_ + this->lineoff_ + 1;
1373   const char* p = pstart;
1374   const char* pend = this->linebuf_ + this->linesize_;
1375
1376   std::string value;
1377   while (*p != '"')
1378     {
1379       Location loc = this->location();
1380       unsigned int c;
1381       bool is_character;
1382       this->lineoff_ = p - this->linebuf_;
1383       p = this->advance_one_char(p, false, &c, &is_character);
1384       if (p >= pend)
1385         {
1386           error_at(this->location(), "unterminated string");
1387           --p;
1388           break;
1389         }
1390       Lex::append_char(c, is_character, &value, loc);
1391     }
1392
1393   Location location = this->location();
1394   this->lineoff_ = p + 1 - this->linebuf_;
1395   return Token::make_string_token(value, location);
1396 }
1397
1398 // Pick up a raw string.
1399
1400 Token
1401 Lex::gather_raw_string()
1402 {
1403   const char* p = this->linebuf_ + this->lineoff_ + 1;
1404   const char* pend = this->linebuf_ + this->linesize_;
1405   Location location = this->location();
1406
1407   std::string value;
1408   while (true)
1409     {
1410       while (p < pend)
1411         {
1412           if (*p == '`')
1413             {
1414               this->lineoff_ = p + 1 - this->linebuf_;
1415               return Token::make_string_token(value, location);
1416             }
1417           Location loc = this->location();
1418           unsigned int c;
1419           bool issued_error;
1420           this->lineoff_ = p - this->linebuf_;
1421           p = this->advance_one_utf8_char(p, &c, &issued_error);
1422           Lex::append_char(c, true, &value, loc);
1423         }
1424       this->lineoff_ = p - this->linebuf_;
1425       if (!this->require_line())
1426         {
1427           error_at(location, "unterminated raw string");
1428           return Token::make_string_token(value, location);
1429         }
1430       p = this->linebuf_ + this->lineoff_;
1431       pend = this->linebuf_ + this->linesize_;
1432     }
1433 }
1434
1435 // If C1 C2 C3 are a three character operator, return the code.
1436
1437 Operator
1438 Lex::three_character_operator(char c1, char c2, char c3)
1439 {
1440   if (c3 == '=')
1441     {
1442       if (c1 == '<' && c2 == '<')
1443         return OPERATOR_LSHIFTEQ;
1444       else if (c1 == '>' && c2 == '>')
1445         return OPERATOR_RSHIFTEQ;
1446       else if (c1 == '&' && c2 == '^')
1447         return OPERATOR_BITCLEAREQ;
1448     }
1449   return OPERATOR_INVALID;
1450 }
1451
1452 // If C1 C2 are a two character operator, return the code.
1453
1454 Operator
1455 Lex::two_character_operator(char c1, char c2)
1456 {
1457   switch (c1)
1458     {
1459     case '|':
1460       if (c2 == '|')
1461         return OPERATOR_OROR;
1462       else if (c2 == '=')
1463         return OPERATOR_OREQ;
1464       break;
1465     case '&':
1466       if (c2 == '&')
1467         return OPERATOR_ANDAND;
1468       else if (c2 == '^')
1469         return OPERATOR_BITCLEAR;
1470       else if (c2 == '=')
1471         return OPERATOR_ANDEQ;
1472       break;
1473     case '^':
1474       if (c2 == '=')
1475         return OPERATOR_XOREQ;
1476       break;
1477     case '=':
1478       if (c2 == '=')
1479         return OPERATOR_EQEQ;
1480       break;
1481     case '!':
1482       if (c2 == '=')
1483         return OPERATOR_NOTEQ;
1484       break;
1485     case '<':
1486       if (c2 == '=')
1487         return OPERATOR_LE;
1488       else if (c2 == '<')
1489         return OPERATOR_LSHIFT;
1490       else if (c2 == '-')
1491         return OPERATOR_CHANOP;
1492       break;
1493     case '>':
1494       if (c2 == '=')
1495         return OPERATOR_GE;
1496       else if (c2 == '>')
1497         return OPERATOR_RSHIFT;
1498       break;
1499     case '*':
1500       if (c2 == '=')
1501         return OPERATOR_MULTEQ;
1502       break;
1503     case '/':
1504       if (c2 == '=')
1505         return OPERATOR_DIVEQ;
1506       break;
1507     case '%':
1508       if (c2 == '=')
1509         return OPERATOR_MODEQ;
1510       break;
1511     case '+':
1512       if (c2 == '+')
1513         {
1514           this->add_semi_at_eol_ = true;
1515           return OPERATOR_PLUSPLUS;
1516         }
1517       else if (c2 == '=')
1518         return OPERATOR_PLUSEQ;
1519       break;
1520     case '-':
1521       if (c2 == '-')
1522         {
1523           this->add_semi_at_eol_ = true;
1524           return OPERATOR_MINUSMINUS;
1525         }
1526       else if (c2 == '=')
1527         return OPERATOR_MINUSEQ;
1528       break;
1529     case ':':
1530       if (c2 == '=')
1531         return OPERATOR_COLONEQ;
1532       break;
1533     default:
1534       break;
1535     }
1536   return OPERATOR_INVALID;
1537 }
1538
1539 // If character C is an operator, return the code.
1540
1541 Operator
1542 Lex::one_character_operator(char c)
1543 {
1544   switch (c)
1545     {
1546     case '<':
1547       return OPERATOR_LT;
1548     case '>':
1549       return OPERATOR_GT;
1550     case '+':
1551       return OPERATOR_PLUS;
1552     case '-':
1553       return OPERATOR_MINUS;
1554     case '|':
1555       return OPERATOR_OR;
1556     case '^':
1557       return OPERATOR_XOR;
1558     case '*':
1559       return OPERATOR_MULT;
1560     case '/':
1561       return OPERATOR_DIV;
1562     case '%':
1563       return OPERATOR_MOD;
1564     case '&':
1565       return OPERATOR_AND;
1566     case '!':
1567       return OPERATOR_NOT;
1568     case '=':
1569       return OPERATOR_EQ;
1570     case ':':
1571       return OPERATOR_COLON;
1572     case ';':
1573       return OPERATOR_SEMICOLON;
1574     case '.':
1575       return OPERATOR_DOT;
1576     case ',':
1577       return OPERATOR_COMMA;
1578     case '(':
1579       return OPERATOR_LPAREN;
1580     case ')':
1581       this->add_semi_at_eol_ = true;
1582       return OPERATOR_RPAREN;
1583     case '{':
1584       return OPERATOR_LCURLY;
1585     case '}':
1586       this->add_semi_at_eol_ = true;
1587       return OPERATOR_RCURLY;
1588     case '[':
1589       return OPERATOR_LSQUARE;
1590     case ']':
1591       this->add_semi_at_eol_ = true;
1592       return OPERATOR_RSQUARE;
1593     default:
1594       return OPERATOR_INVALID;
1595     }
1596 }
1597
1598 // Skip a C-style comment.
1599
1600 bool
1601 Lex::skip_c_comment()
1602 {
1603   while (true)
1604     {
1605       if (!this->require_line())
1606         {
1607           error_at(this->location(), "unterminated comment");
1608           return false;
1609         }
1610
1611       const char* p = this->linebuf_ + this->lineoff_;
1612       const char* pend = this->linebuf_ + this->linesize_;
1613
1614       while (p < pend)
1615         {
1616           if (p[0] == '*' && p + 1 < pend && p[1] == '/')
1617             {
1618               this->lineoff_ = p + 2 - this->linebuf_;
1619               return true;
1620             }
1621
1622           this->lineoff_ = p - this->linebuf_;
1623           unsigned int c;
1624           bool issued_error;
1625           p = this->advance_one_utf8_char(p, &c, &issued_error);
1626         }
1627
1628       this->lineoff_ = p - this->linebuf_;
1629     }
1630 }
1631
1632 // Skip a C++-style comment.
1633
1634 void
1635 Lex::skip_cpp_comment()
1636 {
1637   // Ensure that if EXTERN_ is set, it means that we just saw a
1638   // //extern comment.
1639   this->extern_.clear();
1640
1641   const char* p = this->linebuf_ + this->lineoff_;
1642   const char* pend = this->linebuf_ + this->linesize_;
1643
1644   // By convention, a C++ comment at the start of the line of the form
1645   //   //line FILE:LINENO
1646   // is interpreted as setting the file name and line number of the
1647   // next source line.
1648
1649   if (this->lineoff_ == 2
1650       && pend - p > 5
1651       && memcmp(p, "line ", 5) == 0)
1652     {
1653       p += 5;
1654       while (p < pend && *p == ' ')
1655         ++p;
1656       const char* pcolon = static_cast<const char*>(memchr(p, ':', pend - p));
1657       if (pcolon != NULL
1658           && pcolon[1] >= '0'
1659           && pcolon[1] <= '9')
1660         {
1661           char* plend;
1662           long lineno = strtol(pcolon + 1, &plend, 10);
1663           if (plend > pcolon + 1
1664               && (plend == pend
1665                   || *plend < '0'
1666                   || *plend > '9')
1667               && lineno > 0
1668               && lineno < 0x7fffffff)
1669             {
1670               unsigned int filelen = pcolon - p;
1671               char* file = new char[filelen + 1];
1672               memcpy(file, p, filelen);
1673               file[filelen] = '\0';
1674
1675               this->linemap_->start_file(file, lineno);
1676               this->lineno_ = lineno - 1;
1677
1678               p = plend;
1679             }
1680         }
1681     }
1682
1683   // As a special gccgo extension, a C++ comment at the start of the
1684   // line of the form
1685   //   //extern NAME
1686   // which immediately precedes a function declaration means that the
1687   // external name of the function declaration is NAME.  This is
1688   // normally used to permit Go code to call a C function.
1689   if (this->lineoff_ == 2
1690       && pend - p > 7
1691       && memcmp(p, "extern ", 7) == 0)
1692     {
1693       p += 7;
1694       while (p < pend && (*p == ' ' || *p == '\t'))
1695         ++p;
1696       const char* plend = pend;
1697       while (plend > p
1698              && (plend[-1] == ' ' || plend[-1] == '\t' || plend[-1] == '\n'))
1699         --plend;
1700       if (plend > p)
1701         this->extern_ = std::string(p, plend - p);
1702     }
1703
1704   // For field tracking analysis: a //go:nointerface comment means
1705   // that the next interface method should not be stored in the type
1706   // descriptor.  This permits it to be discarded if it is not needed.
1707   if (this->lineoff_ == 2 && memcmp(p, "go:nointerface", 14) == 0)
1708     this->saw_nointerface_ = true;
1709
1710   while (p < pend)
1711     {
1712       this->lineoff_ = p - this->linebuf_;
1713       unsigned int c;
1714       bool issued_error;
1715       p = this->advance_one_utf8_char(p, &c, &issued_error);
1716       if (issued_error)
1717         this->extern_.clear();
1718     }
1719 }
1720
1721 // The Unicode tables use this struct.
1722
1723 struct Unicode_range
1724 {
1725   // The low end of the range.
1726   unsigned int low;
1727   // The high end of the range.
1728   unsigned int high;
1729   // The stride.  This entries represents low, low + stride, low + 2 *
1730   // stride, etc., up to high.
1731   unsigned int stride;
1732 };
1733
1734 // A table of whitespace characters--Unicode code points classified as
1735 // "Space", "C" locale whitespace characters, the "next line" control
1736 // character (0085), the line separator (2028), the paragraph
1737 // separator (2029), and the "zero-width non-break space" (feff).
1738
1739 static const Unicode_range unicode_space[] =
1740 {
1741   { 0x0009, 0x000d, 1 },
1742   { 0x0020, 0x0020, 1 },
1743   { 0x0085, 0x0085, 1 },
1744   { 0x00a0, 0x00a0, 1 },
1745   { 0x1680, 0x1680, 1 },
1746   { 0x180e, 0x180e, 1 },
1747   { 0x2000, 0x200a, 1 },
1748   { 0x2028, 0x2029, 1 },
1749   { 0x202f, 0x202f, 1 },
1750   { 0x205f, 0x205f, 1 },
1751   { 0x3000, 0x3000, 1 },
1752   { 0xfeff, 0xfeff, 1 },
1753 };
1754
1755 // A table of Unicode digits--Unicode code points classified as
1756 // "Digit".
1757
1758 static const Unicode_range unicode_digits[] =
1759 {
1760   { 0x0030, 0x0039, 1},
1761   { 0x0660, 0x0669, 1},
1762   { 0x06f0, 0x06f9, 1},
1763   { 0x07c0, 0x07c9, 1},
1764   { 0x0966, 0x096f, 1},
1765   { 0x09e6, 0x09ef, 1},
1766   { 0x0a66, 0x0a6f, 1},
1767   { 0x0ae6, 0x0aef, 1},
1768   { 0x0b66, 0x0b6f, 1},
1769   { 0x0be6, 0x0bef, 1},
1770   { 0x0c66, 0x0c6f, 1},
1771   { 0x0ce6, 0x0cef, 1},
1772   { 0x0d66, 0x0d6f, 1},
1773   { 0x0e50, 0x0e59, 1},
1774   { 0x0ed0, 0x0ed9, 1},
1775   { 0x0f20, 0x0f29, 1},
1776   { 0x1040, 0x1049, 1},
1777   { 0x17e0, 0x17e9, 1},
1778   { 0x1810, 0x1819, 1},
1779   { 0x1946, 0x194f, 1},
1780   { 0x19d0, 0x19d9, 1},
1781   { 0x1b50, 0x1b59, 1},
1782   { 0xff10, 0xff19, 1},
1783   { 0x104a0, 0x104a9, 1},
1784   { 0x1d7ce, 0x1d7ff, 1},
1785 };
1786
1787 // A table of Unicode letters--Unicode code points classified as
1788 // "Letter".
1789
1790 static const Unicode_range unicode_letters[] =
1791 {
1792   { 0x0041, 0x005a, 1},
1793   { 0x0061, 0x007a, 1},
1794   { 0x00aa, 0x00b5, 11},
1795   { 0x00ba, 0x00ba, 1},
1796   { 0x00c0, 0x00d6, 1},
1797   { 0x00d8, 0x00f6, 1},
1798   { 0x00f8, 0x02c1, 1},
1799   { 0x02c6, 0x02d1, 1},
1800   { 0x02e0, 0x02e4, 1},
1801   { 0x02ec, 0x02ee, 2},
1802   { 0x0370, 0x0374, 1},
1803   { 0x0376, 0x0377, 1},
1804   { 0x037a, 0x037d, 1},
1805   { 0x0386, 0x0386, 1},
1806   { 0x0388, 0x038a, 1},
1807   { 0x038c, 0x038c, 1},
1808   { 0x038e, 0x03a1, 1},
1809   { 0x03a3, 0x03f5, 1},
1810   { 0x03f7, 0x0481, 1},
1811   { 0x048a, 0x0523, 1},
1812   { 0x0531, 0x0556, 1},
1813   { 0x0559, 0x0559, 1},
1814   { 0x0561, 0x0587, 1},
1815   { 0x05d0, 0x05ea, 1},
1816   { 0x05f0, 0x05f2, 1},
1817   { 0x0621, 0x064a, 1},
1818   { 0x066e, 0x066f, 1},
1819   { 0x0671, 0x06d3, 1},
1820   { 0x06d5, 0x06d5, 1},
1821   { 0x06e5, 0x06e6, 1},
1822   { 0x06ee, 0x06ef, 1},
1823   { 0x06fa, 0x06fc, 1},
1824   { 0x06ff, 0x0710, 17},
1825   { 0x0712, 0x072f, 1},
1826   { 0x074d, 0x07a5, 1},
1827   { 0x07b1, 0x07b1, 1},
1828   { 0x07ca, 0x07ea, 1},
1829   { 0x07f4, 0x07f5, 1},
1830   { 0x07fa, 0x07fa, 1},
1831   { 0x0904, 0x0939, 1},
1832   { 0x093d, 0x0950, 19},
1833   { 0x0958, 0x0961, 1},
1834   { 0x0971, 0x0972, 1},
1835   { 0x097b, 0x097f, 1},
1836   { 0x0985, 0x098c, 1},
1837   { 0x098f, 0x0990, 1},
1838   { 0x0993, 0x09a8, 1},
1839   { 0x09aa, 0x09b0, 1},
1840   { 0x09b2, 0x09b2, 1},
1841   { 0x09b6, 0x09b9, 1},
1842   { 0x09bd, 0x09ce, 17},
1843   { 0x09dc, 0x09dd, 1},
1844   { 0x09df, 0x09e1, 1},
1845   { 0x09f0, 0x09f1, 1},
1846   { 0x0a05, 0x0a0a, 1},
1847   { 0x0a0f, 0x0a10, 1},
1848   { 0x0a13, 0x0a28, 1},
1849   { 0x0a2a, 0x0a30, 1},
1850   { 0x0a32, 0x0a33, 1},
1851   { 0x0a35, 0x0a36, 1},
1852   { 0x0a38, 0x0a39, 1},
1853   { 0x0a59, 0x0a5c, 1},
1854   { 0x0a5e, 0x0a5e, 1},
1855   { 0x0a72, 0x0a74, 1},
1856   { 0x0a85, 0x0a8d, 1},
1857   { 0x0a8f, 0x0a91, 1},
1858   { 0x0a93, 0x0aa8, 1},
1859   { 0x0aaa, 0x0ab0, 1},
1860   { 0x0ab2, 0x0ab3, 1},
1861   { 0x0ab5, 0x0ab9, 1},
1862   { 0x0abd, 0x0ad0, 19},
1863   { 0x0ae0, 0x0ae1, 1},
1864   { 0x0b05, 0x0b0c, 1},
1865   { 0x0b0f, 0x0b10, 1},
1866   { 0x0b13, 0x0b28, 1},
1867   { 0x0b2a, 0x0b30, 1},
1868   { 0x0b32, 0x0b33, 1},
1869   { 0x0b35, 0x0b39, 1},
1870   { 0x0b3d, 0x0b3d, 1},
1871   { 0x0b5c, 0x0b5d, 1},
1872   { 0x0b5f, 0x0b61, 1},
1873   { 0x0b71, 0x0b83, 18},
1874   { 0x0b85, 0x0b8a, 1},
1875   { 0x0b8e, 0x0b90, 1},
1876   { 0x0b92, 0x0b95, 1},
1877   { 0x0b99, 0x0b9a, 1},
1878   { 0x0b9c, 0x0b9c, 1},
1879   { 0x0b9e, 0x0b9f, 1},
1880   { 0x0ba3, 0x0ba4, 1},
1881   { 0x0ba8, 0x0baa, 1},
1882   { 0x0bae, 0x0bb9, 1},
1883   { 0x0bd0, 0x0bd0, 1},
1884   { 0x0c05, 0x0c0c, 1},
1885   { 0x0c0e, 0x0c10, 1},
1886   { 0x0c12, 0x0c28, 1},
1887   { 0x0c2a, 0x0c33, 1},
1888   { 0x0c35, 0x0c39, 1},
1889   { 0x0c3d, 0x0c3d, 1},
1890   { 0x0c58, 0x0c59, 1},
1891   { 0x0c60, 0x0c61, 1},
1892   { 0x0c85, 0x0c8c, 1},
1893   { 0x0c8e, 0x0c90, 1},
1894   { 0x0c92, 0x0ca8, 1},
1895   { 0x0caa, 0x0cb3, 1},
1896   { 0x0cb5, 0x0cb9, 1},
1897   { 0x0cbd, 0x0cde, 33},
1898   { 0x0ce0, 0x0ce1, 1},
1899   { 0x0d05, 0x0d0c, 1},
1900   { 0x0d0e, 0x0d10, 1},
1901   { 0x0d12, 0x0d28, 1},
1902   { 0x0d2a, 0x0d39, 1},
1903   { 0x0d3d, 0x0d3d, 1},
1904   { 0x0d60, 0x0d61, 1},
1905   { 0x0d7a, 0x0d7f, 1},
1906   { 0x0d85, 0x0d96, 1},
1907   { 0x0d9a, 0x0db1, 1},
1908   { 0x0db3, 0x0dbb, 1},
1909   { 0x0dbd, 0x0dbd, 1},
1910   { 0x0dc0, 0x0dc6, 1},
1911   { 0x0e01, 0x0e30, 1},
1912   { 0x0e32, 0x0e33, 1},
1913   { 0x0e40, 0x0e46, 1},
1914   { 0x0e81, 0x0e82, 1},
1915   { 0x0e84, 0x0e84, 1},
1916   { 0x0e87, 0x0e88, 1},
1917   { 0x0e8a, 0x0e8d, 3},
1918   { 0x0e94, 0x0e97, 1},
1919   { 0x0e99, 0x0e9f, 1},
1920   { 0x0ea1, 0x0ea3, 1},
1921   { 0x0ea5, 0x0ea7, 2},
1922   { 0x0eaa, 0x0eab, 1},
1923   { 0x0ead, 0x0eb0, 1},
1924   { 0x0eb2, 0x0eb3, 1},
1925   { 0x0ebd, 0x0ebd, 1},
1926   { 0x0ec0, 0x0ec4, 1},
1927   { 0x0ec6, 0x0ec6, 1},
1928   { 0x0edc, 0x0edd, 1},
1929   { 0x0f00, 0x0f00, 1},
1930   { 0x0f40, 0x0f47, 1},
1931   { 0x0f49, 0x0f6c, 1},
1932   { 0x0f88, 0x0f8b, 1},
1933   { 0x1000, 0x102a, 1},
1934   { 0x103f, 0x103f, 1},
1935   { 0x1050, 0x1055, 1},
1936   { 0x105a, 0x105d, 1},
1937   { 0x1061, 0x1061, 1},
1938   { 0x1065, 0x1066, 1},
1939   { 0x106e, 0x1070, 1},
1940   { 0x1075, 0x1081, 1},
1941   { 0x108e, 0x108e, 1},
1942   { 0x10a0, 0x10c5, 1},
1943   { 0x10d0, 0x10fa, 1},
1944   { 0x10fc, 0x10fc, 1},
1945   { 0x1100, 0x1159, 1},
1946   { 0x115f, 0x11a2, 1},
1947   { 0x11a8, 0x11f9, 1},
1948   { 0x1200, 0x1248, 1},
1949   { 0x124a, 0x124d, 1},
1950   { 0x1250, 0x1256, 1},
1951   { 0x1258, 0x1258, 1},
1952   { 0x125a, 0x125d, 1},
1953   { 0x1260, 0x1288, 1},
1954   { 0x128a, 0x128d, 1},
1955   { 0x1290, 0x12b0, 1},
1956   { 0x12b2, 0x12b5, 1},
1957   { 0x12b8, 0x12be, 1},
1958   { 0x12c0, 0x12c0, 1},
1959   { 0x12c2, 0x12c5, 1},
1960   { 0x12c8, 0x12d6, 1},
1961   { 0x12d8, 0x1310, 1},
1962   { 0x1312, 0x1315, 1},
1963   { 0x1318, 0x135a, 1},
1964   { 0x1380, 0x138f, 1},
1965   { 0x13a0, 0x13f4, 1},
1966   { 0x1401, 0x166c, 1},
1967   { 0x166f, 0x1676, 1},
1968   { 0x1681, 0x169a, 1},
1969   { 0x16a0, 0x16ea, 1},
1970   { 0x1700, 0x170c, 1},
1971   { 0x170e, 0x1711, 1},
1972   { 0x1720, 0x1731, 1},
1973   { 0x1740, 0x1751, 1},
1974   { 0x1760, 0x176c, 1},
1975   { 0x176e, 0x1770, 1},
1976   { 0x1780, 0x17b3, 1},
1977   { 0x17d7, 0x17dc, 5},
1978   { 0x1820, 0x1877, 1},
1979   { 0x1880, 0x18a8, 1},
1980   { 0x18aa, 0x18aa, 1},
1981   { 0x1900, 0x191c, 1},
1982   { 0x1950, 0x196d, 1},
1983   { 0x1970, 0x1974, 1},
1984   { 0x1980, 0x19a9, 1},
1985   { 0x19c1, 0x19c7, 1},
1986   { 0x1a00, 0x1a16, 1},
1987   { 0x1b05, 0x1b33, 1},
1988   { 0x1b45, 0x1b4b, 1},
1989   { 0x1b83, 0x1ba0, 1},
1990   { 0x1bae, 0x1baf, 1},
1991   { 0x1c00, 0x1c23, 1},
1992   { 0x1c4d, 0x1c4f, 1},
1993   { 0x1c5a, 0x1c7d, 1},
1994   { 0x1d00, 0x1dbf, 1},
1995   { 0x1e00, 0x1f15, 1},
1996   { 0x1f18, 0x1f1d, 1},
1997   { 0x1f20, 0x1f45, 1},
1998   { 0x1f48, 0x1f4d, 1},
1999   { 0x1f50, 0x1f57, 1},
2000   { 0x1f59, 0x1f5d, 2},
2001   { 0x1f5f, 0x1f7d, 1},
2002   { 0x1f80, 0x1fb4, 1},
2003   { 0x1fb6, 0x1fbc, 1},
2004   { 0x1fbe, 0x1fbe, 1},
2005   { 0x1fc2, 0x1fc4, 1},
2006   { 0x1fc6, 0x1fcc, 1},
2007   { 0x1fd0, 0x1fd3, 1},
2008   { 0x1fd6, 0x1fdb, 1},
2009   { 0x1fe0, 0x1fec, 1},
2010   { 0x1ff2, 0x1ff4, 1},
2011   { 0x1ff6, 0x1ffc, 1},
2012   { 0x2071, 0x207f, 14},
2013   { 0x2090, 0x2094, 1},
2014   { 0x2102, 0x2107, 5},
2015   { 0x210a, 0x2113, 1},
2016   { 0x2115, 0x2115, 1},
2017   { 0x2119, 0x211d, 1},
2018   { 0x2124, 0x2128, 2},
2019   { 0x212a, 0x212d, 1},
2020   { 0x212f, 0x2139, 1},
2021   { 0x213c, 0x213f, 1},
2022   { 0x2145, 0x2149, 1},
2023   { 0x214e, 0x214e, 1},
2024   { 0x2183, 0x2184, 1},
2025   { 0x2c00, 0x2c2e, 1},
2026   { 0x2c30, 0x2c5e, 1},
2027   { 0x2c60, 0x2c6f, 1},
2028   { 0x2c71, 0x2c7d, 1},
2029   { 0x2c80, 0x2ce4, 1},
2030   { 0x2d00, 0x2d25, 1},
2031   { 0x2d30, 0x2d65, 1},
2032   { 0x2d6f, 0x2d6f, 1},
2033   { 0x2d80, 0x2d96, 1},
2034   { 0x2da0, 0x2da6, 1},
2035   { 0x2da8, 0x2dae, 1},
2036   { 0x2db0, 0x2db6, 1},
2037   { 0x2db8, 0x2dbe, 1},
2038   { 0x2dc0, 0x2dc6, 1},
2039   { 0x2dc8, 0x2dce, 1},
2040   { 0x2dd0, 0x2dd6, 1},
2041   { 0x2dd8, 0x2dde, 1},
2042   { 0x2e2f, 0x2e2f, 1},
2043   { 0x3005, 0x3006, 1},
2044   { 0x3031, 0x3035, 1},
2045   { 0x303b, 0x303c, 1},
2046   { 0x3041, 0x3096, 1},
2047   { 0x309d, 0x309f, 1},
2048   { 0x30a1, 0x30fa, 1},
2049   { 0x30fc, 0x30ff, 1},
2050   { 0x3105, 0x312d, 1},
2051   { 0x3131, 0x318e, 1},
2052   { 0x31a0, 0x31b7, 1},
2053   { 0x31f0, 0x31ff, 1},
2054   { 0x3400, 0x4db5, 1},
2055   { 0x4e00, 0x9fc3, 1},
2056   { 0xa000, 0xa48c, 1},
2057   { 0xa500, 0xa60c, 1},
2058   { 0xa610, 0xa61f, 1},
2059   { 0xa62a, 0xa62b, 1},
2060   { 0xa640, 0xa65f, 1},
2061   { 0xa662, 0xa66e, 1},
2062   { 0xa67f, 0xa697, 1},
2063   { 0xa717, 0xa71f, 1},
2064   { 0xa722, 0xa788, 1},
2065   { 0xa78b, 0xa78c, 1},
2066   { 0xa7fb, 0xa801, 1},
2067   { 0xa803, 0xa805, 1},
2068   { 0xa807, 0xa80a, 1},
2069   { 0xa80c, 0xa822, 1},
2070   { 0xa840, 0xa873, 1},
2071   { 0xa882, 0xa8b3, 1},
2072   { 0xa90a, 0xa925, 1},
2073   { 0xa930, 0xa946, 1},
2074   { 0xaa00, 0xaa28, 1},
2075   { 0xaa40, 0xaa42, 1},
2076   { 0xaa44, 0xaa4b, 1},
2077   { 0xac00, 0xd7a3, 1},
2078   { 0xf900, 0xfa2d, 1},
2079   { 0xfa30, 0xfa6a, 1},
2080   { 0xfa70, 0xfad9, 1},
2081   { 0xfb00, 0xfb06, 1},
2082   { 0xfb13, 0xfb17, 1},
2083   { 0xfb1d, 0xfb1d, 1},
2084   { 0xfb1f, 0xfb28, 1},
2085   { 0xfb2a, 0xfb36, 1},
2086   { 0xfb38, 0xfb3c, 1},
2087   { 0xfb3e, 0xfb3e, 1},
2088   { 0xfb40, 0xfb41, 1},
2089   { 0xfb43, 0xfb44, 1},
2090   { 0xfb46, 0xfbb1, 1},
2091   { 0xfbd3, 0xfd3d, 1},
2092   { 0xfd50, 0xfd8f, 1},
2093   { 0xfd92, 0xfdc7, 1},
2094   { 0xfdf0, 0xfdfb, 1},
2095   { 0xfe70, 0xfe74, 1},
2096   { 0xfe76, 0xfefc, 1},
2097   { 0xff21, 0xff3a, 1},
2098   { 0xff41, 0xff5a, 1},
2099   { 0xff66, 0xffbe, 1},
2100   { 0xffc2, 0xffc7, 1},
2101   { 0xffca, 0xffcf, 1},
2102   { 0xffd2, 0xffd7, 1},
2103   { 0xffda, 0xffdc, 1},
2104   { 0x10000, 0x1000b, 1},
2105   { 0x1000d, 0x10026, 1},
2106   { 0x10028, 0x1003a, 1},
2107   { 0x1003c, 0x1003d, 1},
2108   { 0x1003f, 0x1004d, 1},
2109   { 0x10050, 0x1005d, 1},
2110   { 0x10080, 0x100fa, 1},
2111   { 0x10280, 0x1029c, 1},
2112   { 0x102a0, 0x102d0, 1},
2113   { 0x10300, 0x1031e, 1},
2114   { 0x10330, 0x10340, 1},
2115   { 0x10342, 0x10349, 1},
2116   { 0x10380, 0x1039d, 1},
2117   { 0x103a0, 0x103c3, 1},
2118   { 0x103c8, 0x103cf, 1},
2119   { 0x10400, 0x1049d, 1},
2120   { 0x10800, 0x10805, 1},
2121   { 0x10808, 0x10808, 1},
2122   { 0x1080a, 0x10835, 1},
2123   { 0x10837, 0x10838, 1},
2124   { 0x1083c, 0x1083f, 3},
2125   { 0x10900, 0x10915, 1},
2126   { 0x10920, 0x10939, 1},
2127   { 0x10a00, 0x10a00, 1},
2128   { 0x10a10, 0x10a13, 1},
2129   { 0x10a15, 0x10a17, 1},
2130   { 0x10a19, 0x10a33, 1},
2131   { 0x12000, 0x1236e, 1},
2132   { 0x1d400, 0x1d454, 1},
2133   { 0x1d456, 0x1d49c, 1},
2134   { 0x1d49e, 0x1d49f, 1},
2135   { 0x1d4a2, 0x1d4a2, 1},
2136   { 0x1d4a5, 0x1d4a6, 1},
2137   { 0x1d4a9, 0x1d4ac, 1},
2138   { 0x1d4ae, 0x1d4b9, 1},
2139   { 0x1d4bb, 0x1d4bb, 1},
2140   { 0x1d4bd, 0x1d4c3, 1},
2141   { 0x1d4c5, 0x1d505, 1},
2142   { 0x1d507, 0x1d50a, 1},
2143   { 0x1d50d, 0x1d514, 1},
2144   { 0x1d516, 0x1d51c, 1},
2145   { 0x1d51e, 0x1d539, 1},
2146   { 0x1d53b, 0x1d53e, 1},
2147   { 0x1d540, 0x1d544, 1},
2148   { 0x1d546, 0x1d546, 1},
2149   { 0x1d54a, 0x1d550, 1},
2150   { 0x1d552, 0x1d6a5, 1},
2151   { 0x1d6a8, 0x1d6c0, 1},
2152   { 0x1d6c2, 0x1d6da, 1},
2153   { 0x1d6dc, 0x1d6fa, 1},
2154   { 0x1d6fc, 0x1d714, 1},
2155   { 0x1d716, 0x1d734, 1},
2156   { 0x1d736, 0x1d74e, 1},
2157   { 0x1d750, 0x1d76e, 1},
2158   { 0x1d770, 0x1d788, 1},
2159   { 0x1d78a, 0x1d7a8, 1},
2160   { 0x1d7aa, 0x1d7c2, 1},
2161   { 0x1d7c4, 0x1d7cb, 1},
2162   { 0x20000, 0x2a6d6, 1},
2163   { 0x2f800, 0x2fa1d, 1},
2164 };
2165
2166 // A table of Unicode uppercase letters--Unicode code points
2167 // classified as "Letter, uppercase".
2168
2169 static const Unicode_range unicode_uppercase_letters[] =
2170 {
2171   { 0x0041, 0x005a, 1},
2172   { 0x00c0, 0x00d6, 1},
2173   { 0x00d8, 0x00de, 1},
2174   { 0x0100, 0x0136, 2},
2175   { 0x0139, 0x0147, 2},
2176   { 0x014a, 0x0176, 2},
2177   { 0x0178, 0x0179, 1},
2178   { 0x017b, 0x017d, 2},
2179   { 0x0181, 0x0182, 1},
2180   { 0x0184, 0x0184, 1},
2181   { 0x0186, 0x0187, 1},
2182   { 0x0189, 0x018b, 1},
2183   { 0x018e, 0x0191, 1},
2184   { 0x0193, 0x0194, 1},
2185   { 0x0196, 0x0198, 1},
2186   { 0x019c, 0x019d, 1},
2187   { 0x019f, 0x01a0, 1},
2188   { 0x01a2, 0x01a4, 2},
2189   { 0x01a6, 0x01a7, 1},
2190   { 0x01a9, 0x01ac, 3},
2191   { 0x01ae, 0x01af, 1},
2192   { 0x01b1, 0x01b3, 1},
2193   { 0x01b5, 0x01b5, 1},
2194   { 0x01b7, 0x01b8, 1},
2195   { 0x01bc, 0x01c4, 8},
2196   { 0x01c7, 0x01cd, 3},
2197   { 0x01cf, 0x01db, 2},
2198   { 0x01de, 0x01ee, 2},
2199   { 0x01f1, 0x01f4, 3},
2200   { 0x01f6, 0x01f8, 1},
2201   { 0x01fa, 0x0232, 2},
2202   { 0x023a, 0x023b, 1},
2203   { 0x023d, 0x023e, 1},
2204   { 0x0241, 0x0241, 1},
2205   { 0x0243, 0x0246, 1},
2206   { 0x0248, 0x024e, 2},
2207   { 0x0370, 0x0372, 2},
2208   { 0x0376, 0x0386, 16},
2209   { 0x0388, 0x038a, 1},
2210   { 0x038c, 0x038c, 1},
2211   { 0x038e, 0x038f, 1},
2212   { 0x0391, 0x03a1, 1},
2213   { 0x03a3, 0x03ab, 1},
2214   { 0x03cf, 0x03cf, 1},
2215   { 0x03d2, 0x03d4, 1},
2216   { 0x03d8, 0x03ee, 2},
2217   { 0x03f4, 0x03f7, 3},
2218   { 0x03f9, 0x03fa, 1},
2219   { 0x03fd, 0x042f, 1},
2220   { 0x0460, 0x0480, 2},
2221   { 0x048a, 0x04be, 2},
2222   { 0x04c0, 0x04c1, 1},
2223   { 0x04c3, 0x04cd, 2},
2224   { 0x04d0, 0x0522, 2},
2225   { 0x0531, 0x0556, 1},
2226   { 0x10a0, 0x10c5, 1},
2227   { 0x1e00, 0x1e94, 2},
2228   { 0x1e9e, 0x1efe, 2},
2229   { 0x1f08, 0x1f0f, 1},
2230   { 0x1f18, 0x1f1d, 1},
2231   { 0x1f28, 0x1f2f, 1},
2232   { 0x1f38, 0x1f3f, 1},
2233   { 0x1f48, 0x1f4d, 1},
2234   { 0x1f59, 0x1f5f, 2},
2235   { 0x1f68, 0x1f6f, 1},
2236   { 0x1fb8, 0x1fbb, 1},
2237   { 0x1fc8, 0x1fcb, 1},
2238   { 0x1fd8, 0x1fdb, 1},
2239   { 0x1fe8, 0x1fec, 1},
2240   { 0x1ff8, 0x1ffb, 1},
2241   { 0x2102, 0x2107, 5},
2242   { 0x210b, 0x210d, 1},
2243   { 0x2110, 0x2112, 1},
2244   { 0x2115, 0x2115, 1},
2245   { 0x2119, 0x211d, 1},
2246   { 0x2124, 0x2128, 2},
2247   { 0x212a, 0x212d, 1},
2248   { 0x2130, 0x2133, 1},
2249   { 0x213e, 0x213f, 1},
2250   { 0x2145, 0x2183, 62},
2251   { 0x2c00, 0x2c2e, 1},
2252   { 0x2c60, 0x2c60, 1},
2253   { 0x2c62, 0x2c64, 1},
2254   { 0x2c67, 0x2c6b, 2},
2255   { 0x2c6d, 0x2c6f, 1},
2256   { 0x2c72, 0x2c75, 3},
2257   { 0x2c80, 0x2ce2, 2},
2258   { 0xa640, 0xa65e, 2},
2259   { 0xa662, 0xa66c, 2},
2260   { 0xa680, 0xa696, 2},
2261   { 0xa722, 0xa72e, 2},
2262   { 0xa732, 0xa76e, 2},
2263   { 0xa779, 0xa77b, 2},
2264   { 0xa77d, 0xa77e, 1},
2265   { 0xa780, 0xa786, 2},
2266   { 0xa78b, 0xa78b, 1},
2267   { 0xff21, 0xff3a, 1},
2268   { 0x10400, 0x10427, 1},
2269   { 0x1d400, 0x1d419, 1},
2270   { 0x1d434, 0x1d44d, 1},
2271   { 0x1d468, 0x1d481, 1},
2272   { 0x1d49c, 0x1d49c, 1},
2273   { 0x1d49e, 0x1d49f, 1},
2274   { 0x1d4a2, 0x1d4a2, 1},
2275   { 0x1d4a5, 0x1d4a6, 1},
2276   { 0x1d4a9, 0x1d4ac, 1},
2277   { 0x1d4ae, 0x1d4b5, 1},
2278   { 0x1d4d0, 0x1d4e9, 1},
2279   { 0x1d504, 0x1d505, 1},
2280   { 0x1d507, 0x1d50a, 1},
2281   { 0x1d50d, 0x1d514, 1},
2282   { 0x1d516, 0x1d51c, 1},
2283   { 0x1d538, 0x1d539, 1},
2284   { 0x1d53b, 0x1d53e, 1},
2285   { 0x1d540, 0x1d544, 1},
2286   { 0x1d546, 0x1d546, 1},
2287   { 0x1d54a, 0x1d550, 1},
2288   { 0x1d56c, 0x1d585, 1},
2289   { 0x1d5a0, 0x1d5b9, 1},
2290   { 0x1d5d4, 0x1d5ed, 1},
2291   { 0x1d608, 0x1d621, 1},
2292   { 0x1d63c, 0x1d655, 1},
2293   { 0x1d670, 0x1d689, 1},
2294   { 0x1d6a8, 0x1d6c0, 1},
2295   { 0x1d6e2, 0x1d6fa, 1},
2296   { 0x1d71c, 0x1d734, 1},
2297   { 0x1d756, 0x1d76e, 1},
2298   { 0x1d790, 0x1d7a8, 1},
2299   { 0x1d7ca, 0x1d7ca, 1},
2300 };
2301
2302 // Return true if C is in RANGES.
2303
2304 bool
2305 Lex::is_in_unicode_range(unsigned int c, const Unicode_range* ranges,
2306                          size_t range_size)
2307 {
2308   if (c < 0x100)
2309     {
2310       // The common case is a small value, and we know that it will be
2311       // in the first few entries of the table.  Do a linear scan
2312       // rather than a binary search.
2313       for (size_t i = 0; i < range_size; ++i)
2314         {
2315           const Unicode_range* p = &ranges[i];
2316           if (c <= p->high)
2317             {
2318               if (c < p->low)
2319                 return false;
2320               return (c - p->low) % p->stride == 0;
2321             }
2322         }
2323       return false;
2324     }
2325   else
2326     {
2327       size_t lo = 0;
2328       size_t hi = range_size;
2329       while (lo < hi)
2330         {
2331           size_t mid = lo + (hi - lo) / 2;
2332           const Unicode_range* p = &ranges[mid];
2333           if (c < p->low)
2334             hi = mid;
2335           else if (c > p->high)
2336             lo = mid + 1;
2337           else
2338             return (c - p->low) % p->stride == 0;
2339         }
2340       return false;
2341     }
2342 }
2343
2344 // Return whether C is a space character.
2345
2346 bool
2347 Lex::is_unicode_space(unsigned int c)
2348 {
2349   return Lex::is_in_unicode_range(c, unicode_space,
2350                                   ARRAY_SIZE(unicode_space));
2351 }
2352
2353 // Return whether C is a Unicode digit--a Unicode code point
2354 // classified as "Digit".
2355
2356 bool
2357 Lex::is_unicode_digit(unsigned int c)
2358 {
2359   return Lex::is_in_unicode_range(c, unicode_digits,
2360                                   ARRAY_SIZE(unicode_digits));
2361 }
2362
2363 // Return whether C is a Unicode letter--a Unicode code point
2364 // classified as "Letter".
2365
2366 bool
2367 Lex::is_unicode_letter(unsigned int c)
2368 {
2369   return Lex::is_in_unicode_range(c, unicode_letters,
2370                                   ARRAY_SIZE(unicode_letters));
2371 }
2372
2373 // Return whether C is a Unicode uppercase letter.  a Unicode code
2374 // point classified as "Letter, uppercase".
2375
2376 bool
2377 Lex::is_unicode_uppercase(unsigned int c)
2378 {
2379   return Lex::is_in_unicode_range(c, unicode_uppercase_letters,
2380                                   ARRAY_SIZE(unicode_uppercase_letters));
2381 }
2382
2383 // Return whether the identifier NAME should be exported.  NAME is a
2384 // mangled name which includes only ASCII characters.
2385
2386 bool
2387 Lex::is_exported_name(const std::string& name)
2388 {
2389   unsigned char c = name[0];
2390   if (c != '$')
2391     return c >= 'A' && c <= 'Z';
2392   else
2393     {
2394       const char* p = name.data();
2395       size_t len = name.length();
2396       if (len < 2 || p[1] != 'U')
2397         return false;
2398       unsigned int ci = 0;
2399       for (size_t i = 2; i < len && p[i] != '$'; ++i)
2400         {
2401           c = p[i];
2402           if (!hex_p(c))
2403             return false;
2404           ci <<= 4;
2405           ci |= hex_value(c);
2406         }
2407       return Lex::is_unicode_uppercase(ci);
2408     }
2409 }
2410
2411 // Return whether the identifier NAME contains an invalid character.
2412 // This is based on how we handle invalid characters in
2413 // gather_identifier.
2414
2415 bool
2416 Lex::is_invalid_identifier(const std::string& name)
2417 {
2418   return name.find("$INVALID$") != std::string::npos;
2419 }