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             {
 877               // Check for an invalid character here, as we get better
 878               // error behaviour if we swallow them as part of the
 879               // identifier we are building.
 880               if ((cc >= ' ' && cc < 0x7f)
 881                   || cc == '\t'
 882                   || cc == '\r'
 883                   || cc == '\n')
 884                 break;
 885
 886               this->lineoff_ = p - this->linebuf_;
 887               error_at(this->location(),
 888                        "invalid character 0x%x in identifier",
 889                        cc);
 890               if (!has_non_ascii_char)
 891                 {
 892                   buf.assign(pstart, p - pstart);
 893                   has_non_ascii_char = true;
 894                 }
 895               if (!Lex::is_invalid_identifier(buf))
 896                 buf.append("$INVALID$");
 897             }
 898           ++p;
 899           if (is_first)
 900             {
 901               is_exported = cc >= 'A' && cc <= 'Z';
 902               is_first = false;
 903             }
 904           if (has_non_ascii_char)
 905             buf.push_back(cc);
 906         }
 907       else
 908         {
 909           unsigned int ci;
 910           bool issued_error;
 911           this->lineoff_ = p - this->linebuf_;
 912           const char* pnext = this->advance_one_utf8_char(p, &ci,
 913                                                           &issued_error);
 914           bool is_invalid = false;
 915           if (!Lex::is_unicode_letter(ci) && !Lex::is_unicode_digit(ci))
 916             {
 917               // There is no valid place for a non-ASCII character
 918               // other than an identifier, so we get better error
 919               // handling behaviour if we swallow this character after
 920               // giving an error.
 921               if (!issued_error)
 922                 error_at(this->location(),
 923                          "invalid character 0x%x in identifier",
 924                          ci);
 925               is_invalid = true;
 926             }
 927           if (is_first)
 928             {
 929               is_exported = Lex::is_unicode_uppercase(ci);
 930               is_first = false;
 931             }
 932           if (!has_non_ascii_char)
 933             {
 934               buf.assign(pstart, p - pstart);
 935               has_non_ascii_char = true;
 936             }
 937           if (is_invalid && !Lex::is_invalid_identifier(buf))
 938             buf.append("$INVALID$");
 939           buf.append(p, pnext - p);
 940           p = pnext;
 941         }
 942     }
 943   Location location = this->location();
 944   this->add_semi_at_eol_ = true;
 945   this->lineoff_ = p - this->linebuf_;
 946   if (has_non_ascii_char)
 947     return Token::make_identifier_token(buf, is_exported, location);
 948   else
 949     {
 950       Keyword code = keywords.keyword_to_code(pstart, p - pstart);
 951       if (code == KEYWORD_INVALID)
 952         return Token::make_identifier_token(std::string(pstart, p - pstart),
 953                                             is_exported, location);
 954       else
 955         {
 956           switch (code)
 957             {
 958             case KEYWORD_BREAK:
 959             case KEYWORD_CONTINUE:
 960             case KEYWORD_FALLTHROUGH:
 961             case KEYWORD_RETURN:
 962               break;
 963             default:
 964               this->add_semi_at_eol_ = false;
 965               break;
 966             }
 967           return Token::make_keyword_token(code, location);
 968         }
 969     }
 970 }
 971
 972 // Return whether C is a hex digit.
 973
 974 bool
 975 Lex::is_hex_digit(char c)
 976 {
 977   return ((c >= '0' && c <= '9')
 978           || (c >= 'A' && c <= 'F')
 979           || (c >= 'a' && c <= 'f'));
 980 }
 981
 982 // Return whether an exponent could start at P.
 983
 984 bool
 985 Lex::could_be_exponent(const char* p, const char* pend)
 986 {
 987   if (*p != 'e' && *p != 'E')
 988     return false;
 989   ++p;
 990   if (p >= pend)
 991     return false;
 992   if (*p == '+' || *p == '-')
 993     {
 994       ++p;
 995       if (p >= pend)
 996         return false;
 997     }
 998   return *p >= '0' && *p <= '9';
 999 }
1000
1001 // Pick up a number.
1002
1003 Token
1004 Lex::gather_number()
1005 {
1006   const char* pstart = this->linebuf_ + this->lineoff_;
1007   const char* p = pstart;
1008   const char* pend = this->linebuf_ + this->linesize_;
1009
1010   Location location = this->location();
1011
1012   bool neg = false;
1013   if (*p == '+')
1014     ++p;
1015   else if (*p == '-')
1016     {
1017       ++p;
1018       neg = true;
1019     }
1020
1021   const char* pnum = p;
1022   if (*p == '0')
1023     {
1024       int base;
1025       if ((p[1] == 'x' || p[1] == 'X')
1026           && Lex::is_hex_digit(p[2]))
1027         {
1028           base = 16;
1029           p += 2;
1030           pnum = p;
1031           while (p < pend)
1032             {
1033               if (!Lex::is_hex_digit(*p))
1034                 break;
1035               ++p;
1036             }
1037         }
1038       else
1039         {
1040           base = 8;
1041           pnum = p;
1042           while (p < pend)
1043             {
1044               if (*p < '0' || *p > '7')
1045                 break;
1046               ++p;
1047             }
1048         }
1049
1050       // A partial token that looks like an octal literal might actually be the
1051       // beginning of a floating-point or imaginary literal.
1052       if (base == 16 || (*p != '.' && *p != 'i' && !Lex::could_be_exponent(p, pend)))
1053         {
1054           std::string s(pnum, p - pnum);
1055           mpz_t val;
1056           int r = mpz_init_set_str(val, s.c_str(), base);
1057           go_assert(r == 0);
1058
1059           if (neg)
1060             mpz_neg(val, val);
1061
1062           this->lineoff_ = p - this->linebuf_;
1063           Token ret = Token::make_integer_token(val, location);
1064           mpz_clear(val);
1065           return ret;
1066         }
1067     }
1068
1069   while (p < pend)
1070     {
1071       if (*p < '0' || *p > '9')
1072         break;
1073       ++p;
1074     }
1075
1076   if (*p != '.' && *p != 'i' && !Lex::could_be_exponent(p, pend))
1077     {
1078       std::string s(pnum, p - pnum);
1079       mpz_t val;
1080       int r = mpz_init_set_str(val, s.c_str(), 10);
1081       go_assert(r == 0);
1082
1083       if (neg)
1084         mpz_neg(val, val);
1085
1086       this->lineoff_ = p - this->linebuf_;
1087       Token ret = Token::make_integer_token(val, location);
1088       mpz_clear(val);
1089       return ret;
1090     }
1091
1092   if (*p != 'i')
1093     {
1094       bool dot = *p == '.';
1095
1096       ++p;
1097
1098       if (!dot)
1099         {
1100           if (*p == '+' || *p == '-')
1101             ++p;
1102         }
1103
1104       while (p < pend)
1105         {
1106           if (*p < '0' || *p > '9')
1107             break;
1108           ++p;
1109         }
1110
1111       if (dot && Lex::could_be_exponent(p, pend))
1112         {
1113           ++p;
1114           if (*p == '+' || *p == '-')
1115             ++p;
1116           while (p < pend)
1117             {
1118               if (*p < '0' || *p > '9')
1119                 break;
1120               ++p;
1121             }
1122         }
1123     }
1124
1125   std::string s(pnum, p - pnum);
1126   mpfr_t val;
1127   int r = mpfr_init_set_str(val, s.c_str(), 10, GMP_RNDN);
1128   go_assert(r == 0);
1129
1130   if (neg)
1131     mpfr_neg(val, val, GMP_RNDN);
1132
1133   bool is_imaginary = *p == 'i';
1134   if (is_imaginary)
1135     ++p;
1136
1137   this->lineoff_ = p - this->linebuf_;
1138   if (is_imaginary)
1139     {
1140       Token ret = Token::make_imaginary_token(val, location);
1141       mpfr_clear(val);
1142       return ret;
1143     }
1144   else
1145     {
1146       Token ret = Token::make_float_token(val, location);
1147       mpfr_clear(val);
1148       return ret;
1149     }
1150 }
1151
1152 // Advance one character, possibly escaped.  Return the pointer beyond
1153 // the character.  Set *VALUE to the character.  Set *IS_CHARACTER if
1154 // this is a character (e.g., 'a' or '\u1234') rather than a byte
1155 // value (e.g., '\001').
1156
1157 const char*
1158 Lex::advance_one_char(const char* p, bool is_single_quote, unsigned int* value,
1159                       bool* is_character)
1160 {
1161   *value = 0;
1162   *is_character = true;
1163   if (*p != '\\')
1164     {
1165       bool issued_error;
1166       const char* ret = this->advance_one_utf8_char(p, value, &issued_error);
1167       if (is_single_quote
1168           && (*value == '\'' || *value == '\n')
1169           && !issued_error)
1170         error_at(this->location(), "invalid character literal");
1171       return ret;
1172     }
1173   else
1174     {
1175       ++p;
1176       switch (*p)
1177         {
1178         case '0': case '1': case '2': case '3':
1179         case '4': case '5': case '6': case '7':
1180           *is_character = false;
1181           if (p[1] >= '0' && p[1] <= '7'
1182               && p[2] >= '0' && p[2] <= '7')
1183             {
1184               *value = ((Lex::octal_value(p[0]) << 6)
1185                         + (Lex::octal_value(p[1]) << 3)
1186                         + Lex::octal_value(p[2]));
1187               if (*value > 255)
1188                 {
1189                   error_at(this->location(), "invalid octal constant");
1190                   *value = 255;
1191                 }
1192               return p + 3;
1193             }
1194               error_at(this->location(), "invalid octal character");
1195           return (p[1] >= '0' && p[1] <= '7'
1196                   ? p + 2
1197                   : p + 1);
1198
1199         case 'x':
1200         case 'X':
1201           *is_character = false;
1202           if (Lex::is_hex_digit(p[1]) && Lex::is_hex_digit(p[2]))
1203             {
1204               *value = (hex_value(p[1]) << 4) + hex_value(p[2]);
1205               return p + 3;
1206             }
1207           error_at(this->location(), "invalid hex character");
1208           return (Lex::is_hex_digit(p[1])
1209                   ? p + 2
1210                   : p + 1);
1211
1212         case 'a':
1213           *value = '\a';
1214           return p + 1;
1215         case 'b':
1216           *value = '\b';
1217           return p + 1;
1218         case 'f':
1219           *value = '\f';
1220           return p + 1;
1221         case 'n':
1222           *value = '\n';
1223           return p + 1;
1224         case 'r':
1225           *value = '\r';
1226           return p + 1;
1227         case 't':
1228           *value = '\t';
1229           return p + 1;
1230         case 'v':
1231           *value = '\v';
1232           return p + 1;
1233         case '\\':
1234           *value = '\\';
1235           return p + 1;
1236         case '\'':
1237           if (!is_single_quote)
1238             error_at(this->location(), "invalid quoted character");
1239           *value = '\'';
1240           return p + 1;
1241         case '"':
1242           if (is_single_quote)
1243             error_at(this->location(), "invalid quoted character");
1244           *value = '"';
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             {
1251               *value = ((hex_value(p[1]) << 12)
1252                         + (hex_value(p[2]) << 8)
1253                         + (hex_value(p[3]) << 4)
1254                         + hex_value(p[4]));
1255               if (*value >= 0xd800 && *value < 0xe000)
1256                 {
1257                   error_at(this->location(),
1258                            "invalid unicode code point 0x%x",
1259                            *value);
1260                   // Use the replacement character.
1261                   *value = 0xfffd;
1262                 }
1263               return p + 5;
1264             }
1265           error_at(this->location(), "invalid little unicode code point");
1266           return p + 1;
1267
1268         case 'U':
1269           if (Lex::is_hex_digit(p[1]) && Lex::is_hex_digit(p[2])
1270               && Lex::is_hex_digit(p[3]) && Lex::is_hex_digit(p[4])
1271               && Lex::is_hex_digit(p[5]) && Lex::is_hex_digit(p[6])
1272               && Lex::is_hex_digit(p[7]) && Lex::is_hex_digit(p[8]))
1273             {
1274               *value = ((hex_value(p[1]) << 28)
1275                         + (hex_value(p[2]) << 24)
1276                         + (hex_value(p[3]) << 20)
1277                         + (hex_value(p[4]) << 16)
1278                         + (hex_value(p[5]) << 12)
1279                         + (hex_value(p[6]) << 8)
1280                         + (hex_value(p[7]) << 4)
1281                         + hex_value(p[8]));
1282               if (*value > 0x10ffff
1283                   || (*value >= 0xd800 && *value < 0xe000))
1284                 {
1285                   error_at(this->location(), "invalid unicode code point 0x%x",
1286                            *value);
1287                   // Use the replacement character.
1288                   *value = 0xfffd;
1289                 }
1290               return p + 9;
1291             }
1292           error_at(this->location(), "invalid big unicode code point");
1293           return p + 1;
1294
1295         default:
1296           error_at(this->location(), "invalid character after %<\\%>");
1297           *value = *p;
1298           return p + 1;
1299         }
1300     }
1301 }
1302
1303 // Append V to STR.  IS_CHARACTER is true for a character which should
1304 // be stored in UTF-8, false for a general byte value which should be
1305 // stored directly.
1306
1307 void
1308 Lex::append_char(unsigned int v, bool is_character, std::string* str,
1309                  Location location)
1310 {
1311   char buf[4];
1312   size_t len;
1313   if (v <= 0x7f || !is_character)
1314     {
1315       buf[0] = v;
1316       len = 1;
1317     }
1318   else if (v <= 0x7ff)
1319     {
1320       buf[0] = 0xc0 + (v >> 6);
1321       buf[1] = 0x80 + (v & 0x3f);
1322       len = 2;
1323     }
1324   else
1325     {
1326       if (v > 0x10ffff)
1327         {
1328           warning_at(location, 0,
1329                      "unicode code point 0x%x out of range in string", v);
1330           // Turn it into the "replacement character".
1331           v = 0xfffd;
1332         }
1333       if (v >= 0xd800 && v < 0xe000)
1334         {
1335           warning_at(location, 0,
1336                      "unicode code point 0x%x is invalid surrogate pair", v);
1337           v = 0xfffd;
1338         }
1339       if (v <= 0xffff)
1340         {
1341           buf[0] = 0xe0 + (v >> 12);
1342           buf[1] = 0x80 + ((v >> 6) & 0x3f);
1343           buf[2] = 0x80 + (v & 0x3f);
1344           len = 3;
1345         }
1346       else
1347         {
1348           buf[0] = 0xf0 + (v >> 18);
1349           buf[1] = 0x80 + ((v >> 12) & 0x3f);
1350           buf[2] = 0x80 + ((v >> 6) & 0x3f);
1351           buf[3] = 0x80 + (v & 0x3f);
1352           len = 4;
1353         }
1354     }
1355   str->append(buf, len);
1356 }
1357
1358 // Pick up a character literal.
1359
1360 Token
1361 Lex::gather_character()
1362 {
1363   ++this->lineoff_;
1364   const char* pstart = this->linebuf_ + this->lineoff_;
1365   const char* p = pstart;
1366
1367   unsigned int value;
1368   bool is_character;
1369   p = this->advance_one_char(p, true, &value, &is_character);
1370
1371   if (*p != '\'')
1372     {
1373       error_at(this->location(), "unterminated character constant");
1374       this->lineoff_ = p - this->linebuf_;
1375       return this->make_invalid_token();
1376     }
1377
1378   mpz_t val;
1379   mpz_init_set_ui(val, value);
1380
1381   Location location = this->location();
1382   this->lineoff_ = p + 1 - this->linebuf_;
1383   Token ret = Token::make_character_token(val, location);
1384   mpz_clear(val);
1385   return ret;
1386 }
1387
1388 // Pick up a quoted string.
1389
1390 Token
1391 Lex::gather_string()
1392 {
1393   const char* pstart = this->linebuf_ + this->lineoff_ + 1;
1394   const char* p = pstart;
1395   const char* pend = this->linebuf_ + this->linesize_;
1396
1397   std::string value;
1398   while (*p != '"')
1399     {
1400       Location loc = this->location();
1401       unsigned int c;
1402       bool is_character;
1403       this->lineoff_ = p - this->linebuf_;
1404       p = this->advance_one_char(p, false, &c, &is_character);
1405       if (p >= pend)
1406         {
1407           error_at(this->location(), "unterminated string");
1408           --p;
1409           break;
1410         }
1411       Lex::append_char(c, is_character, &value, loc);
1412     }
1413
1414   Location location = this->location();
1415   this->lineoff_ = p + 1 - this->linebuf_;
1416   return Token::make_string_token(value, location);
1417 }
1418
1419 // Pick up a raw string.
1420
1421 Token
1422 Lex::gather_raw_string()
1423 {
1424   const char* p = this->linebuf_ + this->lineoff_ + 1;
1425   const char* pend = this->linebuf_ + this->linesize_;
1426   Location location = this->location();
1427
1428   std::string value;
1429   while (true)
1430     {
1431       while (p < pend)
1432         {
1433           if (*p == '`')
1434             {
1435               this->lineoff_ = p + 1 - this->linebuf_;
1436               return Token::make_string_token(value, location);
1437             }
1438           Location loc = this->location();
1439           unsigned int c;
1440           bool issued_error;
1441           this->lineoff_ = p - this->linebuf_;
1442           p = this->advance_one_utf8_char(p, &c, &issued_error);
1443           Lex::append_char(c, true, &value, loc);
1444         }
1445       this->lineoff_ = p - this->linebuf_;
1446       if (!this->require_line())
1447         {
1448           error_at(location, "unterminated raw string");
1449           return Token::make_string_token(value, location);
1450         }
1451       p = this->linebuf_ + this->lineoff_;
1452       pend = this->linebuf_ + this->linesize_;
1453     }
1454 }
1455
1456 // If C1 C2 C3 are a three character operator, return the code.
1457
1458 Operator
1459 Lex::three_character_operator(char c1, char c2, char c3)
1460 {
1461   if (c3 == '=')
1462     {
1463       if (c1 == '<' && c2 == '<')
1464         return OPERATOR_LSHIFTEQ;
1465       else if (c1 == '>' && c2 == '>')
1466         return OPERATOR_RSHIFTEQ;
1467       else if (c1 == '&' && c2 == '^')
1468         return OPERATOR_BITCLEAREQ;
1469     }
1470   return OPERATOR_INVALID;
1471 }
1472
1473 // If C1 C2 are a two character operator, return the code.
1474
1475 Operator
1476 Lex::two_character_operator(char c1, char c2)
1477 {
1478   switch (c1)
1479     {
1480     case '|':
1481       if (c2 == '|')
1482         return OPERATOR_OROR;
1483       else if (c2 == '=')
1484         return OPERATOR_OREQ;
1485       break;
1486     case '&':
1487       if (c2 == '&')
1488         return OPERATOR_ANDAND;
1489       else if (c2 == '^')
1490         return OPERATOR_BITCLEAR;
1491       else if (c2 == '=')
1492         return OPERATOR_ANDEQ;
1493       break;
1494     case '^':
1495       if (c2 == '=')
1496         return OPERATOR_XOREQ;
1497       break;
1498     case '=':
1499       if (c2 == '=')
1500         return OPERATOR_EQEQ;
1501       break;
1502     case '!':
1503       if (c2 == '=')
1504         return OPERATOR_NOTEQ;
1505       break;
1506     case '<':
1507       if (c2 == '=')
1508         return OPERATOR_LE;
1509       else if (c2 == '<')
1510         return OPERATOR_LSHIFT;
1511       else if (c2 == '-')
1512         return OPERATOR_CHANOP;
1513       break;
1514     case '>':
1515       if (c2 == '=')
1516         return OPERATOR_GE;
1517       else if (c2 == '>')
1518         return OPERATOR_RSHIFT;
1519       break;
1520     case '*':
1521       if (c2 == '=')
1522         return OPERATOR_MULTEQ;
1523       break;
1524     case '/':
1525       if (c2 == '=')
1526         return OPERATOR_DIVEQ;
1527       break;
1528     case '%':
1529       if (c2 == '=')
1530         return OPERATOR_MODEQ;
1531       break;
1532     case '+':
1533       if (c2 == '+')
1534         {
1535           this->add_semi_at_eol_ = true;
1536           return OPERATOR_PLUSPLUS;
1537         }
1538       else if (c2 == '=')
1539         return OPERATOR_PLUSEQ;
1540       break;
1541     case '-':
1542       if (c2 == '-')
1543         {
1544           this->add_semi_at_eol_ = true;
1545           return OPERATOR_MINUSMINUS;
1546         }
1547       else if (c2 == '=')
1548         return OPERATOR_MINUSEQ;
1549       break;
1550     case ':':
1551       if (c2 == '=')
1552         return OPERATOR_COLONEQ;
1553       break;
1554     default:
1555       break;
1556     }
1557   return OPERATOR_INVALID;
1558 }
1559
1560 // If character C is an operator, return the code.
1561
1562 Operator
1563 Lex::one_character_operator(char c)
1564 {
1565   switch (c)
1566     {
1567     case '<':
1568       return OPERATOR_LT;
1569     case '>':
1570       return OPERATOR_GT;
1571     case '+':
1572       return OPERATOR_PLUS;
1573     case '-':
1574       return OPERATOR_MINUS;
1575     case '|':
1576       return OPERATOR_OR;
1577     case '^':
1578       return OPERATOR_XOR;
1579     case '*':
1580       return OPERATOR_MULT;
1581     case '/':
1582       return OPERATOR_DIV;
1583     case '%':
1584       return OPERATOR_MOD;
1585     case '&':
1586       return OPERATOR_AND;
1587     case '!':
1588       return OPERATOR_NOT;
1589     case '=':
1590       return OPERATOR_EQ;
1591     case ':':
1592       return OPERATOR_COLON;
1593     case ';':
1594       return OPERATOR_SEMICOLON;
1595     case '.':
1596       return OPERATOR_DOT;
1597     case ',':
1598       return OPERATOR_COMMA;
1599     case '(':
1600       return OPERATOR_LPAREN;
1601     case ')':
1602       this->add_semi_at_eol_ = true;
1603       return OPERATOR_RPAREN;
1604     case '{':
1605       return OPERATOR_LCURLY;
1606     case '}':
1607       this->add_semi_at_eol_ = true;
1608       return OPERATOR_RCURLY;
1609     case '[':
1610       return OPERATOR_LSQUARE;
1611     case ']':
1612       this->add_semi_at_eol_ = true;
1613       return OPERATOR_RSQUARE;
1614     default:
1615       return OPERATOR_INVALID;
1616     }
1617 }
1618
1619 // Skip a C-style comment.
1620
1621 bool
1622 Lex::skip_c_comment()
1623 {
1624   while (true)
1625     {
1626       if (!this->require_line())
1627         {
1628           error_at(this->location(), "unterminated comment");
1629           return false;
1630         }
1631
1632       const char* p = this->linebuf_ + this->lineoff_;
1633       const char* pend = this->linebuf_ + this->linesize_;
1634
1635       while (p < pend)
1636         {
1637           if (p[0] == '*' && p + 1 < pend && p[1] == '/')
1638             {
1639               this->lineoff_ = p + 2 - this->linebuf_;
1640               return true;
1641             }
1642
1643           this->lineoff_ = p - this->linebuf_;
1644           unsigned int c;
1645           bool issued_error;
1646           p = this->advance_one_utf8_char(p, &c, &issued_error);
1647         }
1648
1649       this->lineoff_ = p - this->linebuf_;
1650     }
1651 }
1652
1653 // Skip a C++-style comment.
1654
1655 void
1656 Lex::skip_cpp_comment()
1657 {
1658   // Ensure that if EXTERN_ is set, it means that we just saw a
1659   // //extern comment.
1660   this->extern_.clear();
1661
1662   const char* p = this->linebuf_ + this->lineoff_;
1663   const char* pend = this->linebuf_ + this->linesize_;
1664
1665   // By convention, a C++ comment at the start of the line of the form
1666   //   //line FILE:LINENO
1667   // is interpreted as setting the file name and line number of the
1668   // next source line.
1669
1670   if (this->lineoff_ == 2
1671       && pend - p > 5
1672       && memcmp(p, "line ", 5) == 0)
1673     {
1674       p += 5;
1675       while (p < pend && *p == ' ')
1676         ++p;
1677       const char* pcolon = static_cast<const char*>(memchr(p, ':', pend - p));
1678       if (pcolon != NULL
1679           && pcolon[1] >= '0'
1680           && pcolon[1] <= '9')
1681         {
1682           char* plend;
1683           long lineno = strtol(pcolon + 1, &plend, 10);
1684           if (plend > pcolon + 1
1685               && (plend == pend
1686                   || *plend < '0'
1687                   || *plend > '9')
1688               && lineno > 0
1689               && lineno < 0x7fffffff)
1690             {
1691               unsigned int filelen = pcolon - p;
1692               char* file = new char[filelen + 1];
1693               memcpy(file, p, filelen);
1694               file[filelen] = '\0';
1695
1696               this->linemap_->start_file(file, lineno);
1697               this->lineno_ = lineno - 1;
1698
1699               p = plend;
1700             }
1701         }
1702     }
1703
1704   // As a special gccgo extension, a C++ comment at the start of the
1705   // line of the form
1706   //   //extern NAME
1707   // which immediately precedes a function declaration means that the
1708   // external name of the function declaration is NAME.  This is
1709   // normally used to permit Go code to call a C function.
1710   if (this->lineoff_ == 2
1711       && pend - p > 7
1712       && memcmp(p, "extern ", 7) == 0)
1713     {
1714       p += 7;
1715       while (p < pend && (*p == ' ' || *p == '\t'))
1716         ++p;
1717       const char* plend = pend;
1718       while (plend > p
1719              && (plend[-1] == ' ' || plend[-1] == '\t' || plend[-1] == '\n'))
1720         --plend;
1721       if (plend > p)
1722         this->extern_ = std::string(p, plend - p);
1723     }
1724
1725   // For field tracking analysis: a //go:nointerface comment means
1726   // that the next interface method should not be stored in the type
1727   // descriptor.  This permits it to be discarded if it is not needed.
1728   if (this->lineoff_ == 2 && memcmp(p, "go:nointerface", 14) == 0)
1729     this->saw_nointerface_ = true;
1730
1731   while (p < pend)
1732     {
1733       this->lineoff_ = p - this->linebuf_;
1734       unsigned int c;
1735       bool issued_error;
1736       p = this->advance_one_utf8_char(p, &c, &issued_error);
1737       if (issued_error)
1738         this->extern_.clear();
1739     }
1740 }
1741
1742 // The Unicode tables use this struct.
1743
1744 struct Unicode_range
1745 {
1746   // The low end of the range.
1747   unsigned int low;
1748   // The high end of the range.
1749   unsigned int high;
1750   // The stride.  This entries represents low, low + stride, low + 2 *
1751   // stride, etc., up to high.
1752   unsigned int stride;
1753 };
1754
1755 // A table of whitespace characters--Unicode code points classified as
1756 // "Space", "C" locale whitespace characters, the "next line" control
1757 // character (0085), the line separator (2028), the paragraph
1758 // separator (2029), and the "zero-width non-break space" (feff).
1759
1760 static const Unicode_range unicode_space[] =
1761 {
1762   { 0x0009, 0x000d, 1 },
1763   { 0x0020, 0x0020, 1 },
1764   { 0x0085, 0x0085, 1 },
1765   { 0x00a0, 0x00a0, 1 },
1766   { 0x1680, 0x1680, 1 },
1767   { 0x180e, 0x180e, 1 },
1768   { 0x2000, 0x200a, 1 },
1769   { 0x2028, 0x2029, 1 },
1770   { 0x202f, 0x202f, 1 },
1771   { 0x205f, 0x205f, 1 },
1772   { 0x3000, 0x3000, 1 },
1773   { 0xfeff, 0xfeff, 1 },
1774 };
1775
1776 // A table of Unicode digits--Unicode code points classified as
1777 // "Digit".
1778
1779 static const Unicode_range unicode_digits[] =
1780 {
1781   { 0x0030, 0x0039, 1},
1782   { 0x0660, 0x0669, 1},
1783   { 0x06f0, 0x06f9, 1},
1784   { 0x07c0, 0x07c9, 1},
1785   { 0x0966, 0x096f, 1},
1786   { 0x09e6, 0x09ef, 1},
1787   { 0x0a66, 0x0a6f, 1},
1788   { 0x0ae6, 0x0aef, 1},
1789   { 0x0b66, 0x0b6f, 1},
1790   { 0x0be6, 0x0bef, 1},
1791   { 0x0c66, 0x0c6f, 1},
1792   { 0x0ce6, 0x0cef, 1},
1793   { 0x0d66, 0x0d6f, 1},
1794   { 0x0e50, 0x0e59, 1},
1795   { 0x0ed0, 0x0ed9, 1},
1796   { 0x0f20, 0x0f29, 1},
1797   { 0x1040, 0x1049, 1},
1798   { 0x17e0, 0x17e9, 1},
1799   { 0x1810, 0x1819, 1},
1800   { 0x1946, 0x194f, 1},
1801   { 0x19d0, 0x19d9, 1},
1802   { 0x1b50, 0x1b59, 1},
1803   { 0xff10, 0xff19, 1},
1804   { 0x104a0, 0x104a9, 1},
1805   { 0x1d7ce, 0x1d7ff, 1},
1806 };
1807
1808 // A table of Unicode letters--Unicode code points classified as
1809 // "Letter".
1810
1811 static const Unicode_range unicode_letters[] =
1812 {
1813   { 0x0041, 0x005a, 1},
1814   { 0x0061, 0x007a, 1},
1815   { 0x00aa, 0x00b5, 11},
1816   { 0x00ba, 0x00ba, 1},
1817   { 0x00c0, 0x00d6, 1},
1818   { 0x00d8, 0x00f6, 1},
1819   { 0x00f8, 0x02c1, 1},
1820   { 0x02c6, 0x02d1, 1},
1821   { 0x02e0, 0x02e4, 1},
1822   { 0x02ec, 0x02ee, 2},
1823   { 0x0370, 0x0374, 1},
1824   { 0x0376, 0x0377, 1},
1825   { 0x037a, 0x037d, 1},
1826   { 0x0386, 0x0386, 1},
1827   { 0x0388, 0x038a, 1},
1828   { 0x038c, 0x038c, 1},
1829   { 0x038e, 0x03a1, 1},
1830   { 0x03a3, 0x03f5, 1},
1831   { 0x03f7, 0x0481, 1},
1832   { 0x048a, 0x0523, 1},
1833   { 0x0531, 0x0556, 1},
1834   { 0x0559, 0x0559, 1},
1835   { 0x0561, 0x0587, 1},
1836   { 0x05d0, 0x05ea, 1},
1837   { 0x05f0, 0x05f2, 1},
1838   { 0x0621, 0x064a, 1},
1839   { 0x066e, 0x066f, 1},
1840   { 0x0671, 0x06d3, 1},
1841   { 0x06d5, 0x06d5, 1},
1842   { 0x06e5, 0x06e6, 1},
1843   { 0x06ee, 0x06ef, 1},
1844   { 0x06fa, 0x06fc, 1},
1845   { 0x06ff, 0x0710, 17},
1846   { 0x0712, 0x072f, 1},
1847   { 0x074d, 0x07a5, 1},
1848   { 0x07b1, 0x07b1, 1},
1849   { 0x07ca, 0x07ea, 1},
1850   { 0x07f4, 0x07f5, 1},
1851   { 0x07fa, 0x07fa, 1},
1852   { 0x0904, 0x0939, 1},
1853   { 0x093d, 0x0950, 19},
1854   { 0x0958, 0x0961, 1},
1855   { 0x0971, 0x0972, 1},
1856   { 0x097b, 0x097f, 1},
1857   { 0x0985, 0x098c, 1},
1858   { 0x098f, 0x0990, 1},
1859   { 0x0993, 0x09a8, 1},
1860   { 0x09aa, 0x09b0, 1},
1861   { 0x09b2, 0x09b2, 1},
1862   { 0x09b6, 0x09b9, 1},
1863   { 0x09bd, 0x09ce, 17},
1864   { 0x09dc, 0x09dd, 1},
1865   { 0x09df, 0x09e1, 1},
1866   { 0x09f0, 0x09f1, 1},
1867   { 0x0a05, 0x0a0a, 1},
1868   { 0x0a0f, 0x0a10, 1},
1869   { 0x0a13, 0x0a28, 1},
1870   { 0x0a2a, 0x0a30, 1},
1871   { 0x0a32, 0x0a33, 1},
1872   { 0x0a35, 0x0a36, 1},
1873   { 0x0a38, 0x0a39, 1},
1874   { 0x0a59, 0x0a5c, 1},
1875   { 0x0a5e, 0x0a5e, 1},
1876   { 0x0a72, 0x0a74, 1},
1877   { 0x0a85, 0x0a8d, 1},
1878   { 0x0a8f, 0x0a91, 1},
1879   { 0x0a93, 0x0aa8, 1},
1880   { 0x0aaa, 0x0ab0, 1},
1881   { 0x0ab2, 0x0ab3, 1},
1882   { 0x0ab5, 0x0ab9, 1},
1883   { 0x0abd, 0x0ad0, 19},
1884   { 0x0ae0, 0x0ae1, 1},
1885   { 0x0b05, 0x0b0c, 1},
1886   { 0x0b0f, 0x0b10, 1},
1887   { 0x0b13, 0x0b28, 1},
1888   { 0x0b2a, 0x0b30, 1},
1889   { 0x0b32, 0x0b33, 1},
1890   { 0x0b35, 0x0b39, 1},
1891   { 0x0b3d, 0x0b3d, 1},
1892   { 0x0b5c, 0x0b5d, 1},
1893   { 0x0b5f, 0x0b61, 1},
1894   { 0x0b71, 0x0b83, 18},
1895   { 0x0b85, 0x0b8a, 1},
1896   { 0x0b8e, 0x0b90, 1},
1897   { 0x0b92, 0x0b95, 1},
1898   { 0x0b99, 0x0b9a, 1},
1899   { 0x0b9c, 0x0b9c, 1},
1900   { 0x0b9e, 0x0b9f, 1},
1901   { 0x0ba3, 0x0ba4, 1},
1902   { 0x0ba8, 0x0baa, 1},
1903   { 0x0bae, 0x0bb9, 1},
1904   { 0x0bd0, 0x0bd0, 1},
1905   { 0x0c05, 0x0c0c, 1},
1906   { 0x0c0e, 0x0c10, 1},
1907   { 0x0c12, 0x0c28, 1},
1908   { 0x0c2a, 0x0c33, 1},
1909   { 0x0c35, 0x0c39, 1},
1910   { 0x0c3d, 0x0c3d, 1},
1911   { 0x0c58, 0x0c59, 1},
1912   { 0x0c60, 0x0c61, 1},
1913   { 0x0c85, 0x0c8c, 1},
1914   { 0x0c8e, 0x0c90, 1},
1915   { 0x0c92, 0x0ca8, 1},
1916   { 0x0caa, 0x0cb3, 1},
1917   { 0x0cb5, 0x0cb9, 1},
1918   { 0x0cbd, 0x0cde, 33},
1919   { 0x0ce0, 0x0ce1, 1},
1920   { 0x0d05, 0x0d0c, 1},
1921   { 0x0d0e, 0x0d10, 1},
1922   { 0x0d12, 0x0d28, 1},
1923   { 0x0d2a, 0x0d39, 1},
1924   { 0x0d3d, 0x0d3d, 1},
1925   { 0x0d60, 0x0d61, 1},
1926   { 0x0d7a, 0x0d7f, 1},
1927   { 0x0d85, 0x0d96, 1},
1928   { 0x0d9a, 0x0db1, 1},
1929   { 0x0db3, 0x0dbb, 1},
1930   { 0x0dbd, 0x0dbd, 1},
1931   { 0x0dc0, 0x0dc6, 1},
1932   { 0x0e01, 0x0e30, 1},
1933   { 0x0e32, 0x0e33, 1},
1934   { 0x0e40, 0x0e46, 1},
1935   { 0x0e81, 0x0e82, 1},
1936   { 0x0e84, 0x0e84, 1},
1937   { 0x0e87, 0x0e88, 1},
1938   { 0x0e8a, 0x0e8d, 3},
1939   { 0x0e94, 0x0e97, 1},
1940   { 0x0e99, 0x0e9f, 1},
1941   { 0x0ea1, 0x0ea3, 1},
1942   { 0x0ea5, 0x0ea7, 2},
1943   { 0x0eaa, 0x0eab, 1},
1944   { 0x0ead, 0x0eb0, 1},
1945   { 0x0eb2, 0x0eb3, 1},
1946   { 0x0ebd, 0x0ebd, 1},
1947   { 0x0ec0, 0x0ec4, 1},
1948   { 0x0ec6, 0x0ec6, 1},
1949   { 0x0edc, 0x0edd, 1},
1950   { 0x0f00, 0x0f00, 1},
1951   { 0x0f40, 0x0f47, 1},
1952   { 0x0f49, 0x0f6c, 1},
1953   { 0x0f88, 0x0f8b, 1},
1954   { 0x1000, 0x102a, 1},
1955   { 0x103f, 0x103f, 1},
1956   { 0x1050, 0x1055, 1},
1957   { 0x105a, 0x105d, 1},
1958   { 0x1061, 0x1061, 1},
1959   { 0x1065, 0x1066, 1},
1960   { 0x106e, 0x1070, 1},
1961   { 0x1075, 0x1081, 1},
1962   { 0x108e, 0x108e, 1},
1963   { 0x10a0, 0x10c5, 1},
1964   { 0x10d0, 0x10fa, 1},
1965   { 0x10fc, 0x10fc, 1},
1966   { 0x1100, 0x1159, 1},
1967   { 0x115f, 0x11a2, 1},
1968   { 0x11a8, 0x11f9, 1},
1969   { 0x1200, 0x1248, 1},
1970   { 0x124a, 0x124d, 1},
1971   { 0x1250, 0x1256, 1},
1972   { 0x1258, 0x1258, 1},
1973   { 0x125a, 0x125d, 1},
1974   { 0x1260, 0x1288, 1},
1975   { 0x128a, 0x128d, 1},
1976   { 0x1290, 0x12b0, 1},
1977   { 0x12b2, 0x12b5, 1},
1978   { 0x12b8, 0x12be, 1},
1979   { 0x12c0, 0x12c0, 1},
1980   { 0x12c2, 0x12c5, 1},
1981   { 0x12c8, 0x12d6, 1},
1982   { 0x12d8, 0x1310, 1},
1983   { 0x1312, 0x1315, 1},
1984   { 0x1318, 0x135a, 1},
1985   { 0x1380, 0x138f, 1},
1986   { 0x13a0, 0x13f4, 1},
1987   { 0x1401, 0x166c, 1},
1988   { 0x166f, 0x1676, 1},
1989   { 0x1681, 0x169a, 1},
1990   { 0x16a0, 0x16ea, 1},
1991   { 0x1700, 0x170c, 1},
1992   { 0x170e, 0x1711, 1},
1993   { 0x1720, 0x1731, 1},
1994   { 0x1740, 0x1751, 1},
1995   { 0x1760, 0x176c, 1},
1996   { 0x176e, 0x1770, 1},
1997   { 0x1780, 0x17b3, 1},
1998   { 0x17d7, 0x17dc, 5},
1999   { 0x1820, 0x1877, 1},
2000   { 0x1880, 0x18a8, 1},
2001   { 0x18aa, 0x18aa, 1},
2002   { 0x1900, 0x191c, 1},
2003   { 0x1950, 0x196d, 1},
2004   { 0x1970, 0x1974, 1},
2005   { 0x1980, 0x19a9, 1},
2006   { 0x19c1, 0x19c7, 1},
2007   { 0x1a00, 0x1a16, 1},
2008   { 0x1b05, 0x1b33, 1},
2009   { 0x1b45, 0x1b4b, 1},
2010   { 0x1b83, 0x1ba0, 1},
2011   { 0x1bae, 0x1baf, 1},
2012   { 0x1c00, 0x1c23, 1},
2013   { 0x1c4d, 0x1c4f, 1},
2014   { 0x1c5a, 0x1c7d, 1},
2015   { 0x1d00, 0x1dbf, 1},
2016   { 0x1e00, 0x1f15, 1},
2017   { 0x1f18, 0x1f1d, 1},
2018   { 0x1f20, 0x1f45, 1},
2019   { 0x1f48, 0x1f4d, 1},
2020   { 0x1f50, 0x1f57, 1},
2021   { 0x1f59, 0x1f5d, 2},
2022   { 0x1f5f, 0x1f7d, 1},
2023   { 0x1f80, 0x1fb4, 1},
2024   { 0x1fb6, 0x1fbc, 1},
2025   { 0x1fbe, 0x1fbe, 1},
2026   { 0x1fc2, 0x1fc4, 1},
2027   { 0x1fc6, 0x1fcc, 1},
2028   { 0x1fd0, 0x1fd3, 1},
2029   { 0x1fd6, 0x1fdb, 1},
2030   { 0x1fe0, 0x1fec, 1},
2031   { 0x1ff2, 0x1ff4, 1},
2032   { 0x1ff6, 0x1ffc, 1},
2033   { 0x2071, 0x207f, 14},
2034   { 0x2090, 0x2094, 1},
2035   { 0x2102, 0x2107, 5},
2036   { 0x210a, 0x2113, 1},
2037   { 0x2115, 0x2115, 1},
2038   { 0x2119, 0x211d, 1},
2039   { 0x2124, 0x2128, 2},
2040   { 0x212a, 0x212d, 1},
2041   { 0x212f, 0x2139, 1},
2042   { 0x213c, 0x213f, 1},
2043   { 0x2145, 0x2149, 1},
2044   { 0x214e, 0x214e, 1},
2045   { 0x2183, 0x2184, 1},
2046   { 0x2c00, 0x2c2e, 1},
2047   { 0x2c30, 0x2c5e, 1},
2048   { 0x2c60, 0x2c6f, 1},
2049   { 0x2c71, 0x2c7d, 1},
2050   { 0x2c80, 0x2ce4, 1},
2051   { 0x2d00, 0x2d25, 1},
2052   { 0x2d30, 0x2d65, 1},
2053   { 0x2d6f, 0x2d6f, 1},
2054   { 0x2d80, 0x2d96, 1},
2055   { 0x2da0, 0x2da6, 1},
2056   { 0x2da8, 0x2dae, 1},
2057   { 0x2db0, 0x2db6, 1},
2058   { 0x2db8, 0x2dbe, 1},
2059   { 0x2dc0, 0x2dc6, 1},
2060   { 0x2dc8, 0x2dce, 1},
2061   { 0x2dd0, 0x2dd6, 1},
2062   { 0x2dd8, 0x2dde, 1},
2063   { 0x2e2f, 0x2e2f, 1},
2064   { 0x3005, 0x3006, 1},
2065   { 0x3031, 0x3035, 1},
2066   { 0x303b, 0x303c, 1},
2067   { 0x3041, 0x3096, 1},
2068   { 0x309d, 0x309f, 1},
2069   { 0x30a1, 0x30fa, 1},
2070   { 0x30fc, 0x30ff, 1},
2071   { 0x3105, 0x312d, 1},
2072   { 0x3131, 0x318e, 1},
2073   { 0x31a0, 0x31b7, 1},
2074   { 0x31f0, 0x31ff, 1},
2075   { 0x3400, 0x4db5, 1},
2076   { 0x4e00, 0x9fc3, 1},
2077   { 0xa000, 0xa48c, 1},
2078   { 0xa500, 0xa60c, 1},
2079   { 0xa610, 0xa61f, 1},
2080   { 0xa62a, 0xa62b, 1},
2081   { 0xa640, 0xa65f, 1},
2082   { 0xa662, 0xa66e, 1},
2083   { 0xa67f, 0xa697, 1},
2084   { 0xa717, 0xa71f, 1},
2085   { 0xa722, 0xa788, 1},
2086   { 0xa78b, 0xa78c, 1},
2087   { 0xa7fb, 0xa801, 1},
2088   { 0xa803, 0xa805, 1},
2089   { 0xa807, 0xa80a, 1},
2090   { 0xa80c, 0xa822, 1},
2091   { 0xa840, 0xa873, 1},
2092   { 0xa882, 0xa8b3, 1},
2093   { 0xa90a, 0xa925, 1},
2094   { 0xa930, 0xa946, 1},
2095   { 0xaa00, 0xaa28, 1},
2096   { 0xaa40, 0xaa42, 1},
2097   { 0xaa44, 0xaa4b, 1},
2098   { 0xac00, 0xd7a3, 1},
2099   { 0xf900, 0xfa2d, 1},
2100   { 0xfa30, 0xfa6a, 1},
2101   { 0xfa70, 0xfad9, 1},
2102   { 0xfb00, 0xfb06, 1},
2103   { 0xfb13, 0xfb17, 1},
2104   { 0xfb1d, 0xfb1d, 1},
2105   { 0xfb1f, 0xfb28, 1},
2106   { 0xfb2a, 0xfb36, 1},
2107   { 0xfb38, 0xfb3c, 1},
2108   { 0xfb3e, 0xfb3e, 1},
2109   { 0xfb40, 0xfb41, 1},
2110   { 0xfb43, 0xfb44, 1},
2111   { 0xfb46, 0xfbb1, 1},
2112   { 0xfbd3, 0xfd3d, 1},
2113   { 0xfd50, 0xfd8f, 1},
2114   { 0xfd92, 0xfdc7, 1},
2115   { 0xfdf0, 0xfdfb, 1},
2116   { 0xfe70, 0xfe74, 1},
2117   { 0xfe76, 0xfefc, 1},
2118   { 0xff21, 0xff3a, 1},
2119   { 0xff41, 0xff5a, 1},
2120   { 0xff66, 0xffbe, 1},
2121   { 0xffc2, 0xffc7, 1},
2122   { 0xffca, 0xffcf, 1},
2123   { 0xffd2, 0xffd7, 1},
2124   { 0xffda, 0xffdc, 1},
2125   { 0x10000, 0x1000b, 1},
2126   { 0x1000d, 0x10026, 1},
2127   { 0x10028, 0x1003a, 1},
2128   { 0x1003c, 0x1003d, 1},
2129   { 0x1003f, 0x1004d, 1},
2130   { 0x10050, 0x1005d, 1},
2131   { 0x10080, 0x100fa, 1},
2132   { 0x10280, 0x1029c, 1},
2133   { 0x102a0, 0x102d0, 1},
2134   { 0x10300, 0x1031e, 1},
2135   { 0x10330, 0x10340, 1},
2136   { 0x10342, 0x10349, 1},
2137   { 0x10380, 0x1039d, 1},
2138   { 0x103a0, 0x103c3, 1},
2139   { 0x103c8, 0x103cf, 1},
2140   { 0x10400, 0x1049d, 1},
2141   { 0x10800, 0x10805, 1},
2142   { 0x10808, 0x10808, 1},
2143   { 0x1080a, 0x10835, 1},
2144   { 0x10837, 0x10838, 1},
2145   { 0x1083c, 0x1083f, 3},
2146   { 0x10900, 0x10915, 1},
2147   { 0x10920, 0x10939, 1},
2148   { 0x10a00, 0x10a00, 1},
2149   { 0x10a10, 0x10a13, 1},
2150   { 0x10a15, 0x10a17, 1},
2151   { 0x10a19, 0x10a33, 1},
2152   { 0x12000, 0x1236e, 1},
2153   { 0x1d400, 0x1d454, 1},
2154   { 0x1d456, 0x1d49c, 1},
2155   { 0x1d49e, 0x1d49f, 1},
2156   { 0x1d4a2, 0x1d4a2, 1},
2157   { 0x1d4a5, 0x1d4a6, 1},
2158   { 0x1d4a9, 0x1d4ac, 1},
2159   { 0x1d4ae, 0x1d4b9, 1},
2160   { 0x1d4bb, 0x1d4bb, 1},
2161   { 0x1d4bd, 0x1d4c3, 1},
2162   { 0x1d4c5, 0x1d505, 1},
2163   { 0x1d507, 0x1d50a, 1},
2164   { 0x1d50d, 0x1d514, 1},
2165   { 0x1d516, 0x1d51c, 1},
2166   { 0x1d51e, 0x1d539, 1},
2167   { 0x1d53b, 0x1d53e, 1},
2168   { 0x1d540, 0x1d544, 1},
2169   { 0x1d546, 0x1d546, 1},
2170   { 0x1d54a, 0x1d550, 1},
2171   { 0x1d552, 0x1d6a5, 1},
2172   { 0x1d6a8, 0x1d6c0, 1},
2173   { 0x1d6c2, 0x1d6da, 1},
2174   { 0x1d6dc, 0x1d6fa, 1},
2175   { 0x1d6fc, 0x1d714, 1},
2176   { 0x1d716, 0x1d734, 1},
2177   { 0x1d736, 0x1d74e, 1},
2178   { 0x1d750, 0x1d76e, 1},
2179   { 0x1d770, 0x1d788, 1},
2180   { 0x1d78a, 0x1d7a8, 1},
2181   { 0x1d7aa, 0x1d7c2, 1},
2182   { 0x1d7c4, 0x1d7cb, 1},
2183   { 0x20000, 0x2a6d6, 1},
2184   { 0x2f800, 0x2fa1d, 1},
2185 };
2186
2187 // A table of Unicode uppercase letters--Unicode code points
2188 // classified as "Letter, uppercase".
2189
2190 static const Unicode_range unicode_uppercase_letters[] =
2191 {
2192   { 0x0041, 0x005a, 1},
2193   { 0x00c0, 0x00d6, 1},
2194   { 0x00d8, 0x00de, 1},
2195   { 0x0100, 0x0136, 2},
2196   { 0x0139, 0x0147, 2},
2197   { 0x014a, 0x0176, 2},
2198   { 0x0178, 0x0179, 1},
2199   { 0x017b, 0x017d, 2},
2200   { 0x0181, 0x0182, 1},
2201   { 0x0184, 0x0184, 1},
2202   { 0x0186, 0x0187, 1},
2203   { 0x0189, 0x018b, 1},
2204   { 0x018e, 0x0191, 1},
2205   { 0x0193, 0x0194, 1},
2206   { 0x0196, 0x0198, 1},
2207   { 0x019c, 0x019d, 1},
2208   { 0x019f, 0x01a0, 1},
2209   { 0x01a2, 0x01a4, 2},
2210   { 0x01a6, 0x01a7, 1},
2211   { 0x01a9, 0x01ac, 3},
2212   { 0x01ae, 0x01af, 1},
2213   { 0x01b1, 0x01b3, 1},
2214   { 0x01b5, 0x01b5, 1},
2215   { 0x01b7, 0x01b8, 1},
2216   { 0x01bc, 0x01c4, 8},
2217   { 0x01c7, 0x01cd, 3},
2218   { 0x01cf, 0x01db, 2},
2219   { 0x01de, 0x01ee, 2},
2220   { 0x01f1, 0x01f4, 3},
2221   { 0x01f6, 0x01f8, 1},
2222   { 0x01fa, 0x0232, 2},
2223   { 0x023a, 0x023b, 1},
2224   { 0x023d, 0x023e, 1},
2225   { 0x0241, 0x0241, 1},
2226   { 0x0243, 0x0246, 1},
2227   { 0x0248, 0x024e, 2},
2228   { 0x0370, 0x0372, 2},
2229   { 0x0376, 0x0386, 16},
2230   { 0x0388, 0x038a, 1},
2231   { 0x038c, 0x038c, 1},
2232   { 0x038e, 0x038f, 1},
2233   { 0x0391, 0x03a1, 1},
2234   { 0x03a3, 0x03ab, 1},
2235   { 0x03cf, 0x03cf, 1},
2236   { 0x03d2, 0x03d4, 1},
2237   { 0x03d8, 0x03ee, 2},
2238   { 0x03f4, 0x03f7, 3},
2239   { 0x03f9, 0x03fa, 1},
2240   { 0x03fd, 0x042f, 1},
2241   { 0x0460, 0x0480, 2},
2242   { 0x048a, 0x04be, 2},
2243   { 0x04c0, 0x04c1, 1},
2244   { 0x04c3, 0x04cd, 2},
2245   { 0x04d0, 0x0522, 2},
2246   { 0x0531, 0x0556, 1},
2247   { 0x10a0, 0x10c5, 1},
2248   { 0x1e00, 0x1e94, 2},
2249   { 0x1e9e, 0x1efe, 2},
2250   { 0x1f08, 0x1f0f, 1},
2251   { 0x1f18, 0x1f1d, 1},
2252   { 0x1f28, 0x1f2f, 1},
2253   { 0x1f38, 0x1f3f, 1},
2254   { 0x1f48, 0x1f4d, 1},
2255   { 0x1f59, 0x1f5f, 2},
2256   { 0x1f68, 0x1f6f, 1},
2257   { 0x1fb8, 0x1fbb, 1},
2258   { 0x1fc8, 0x1fcb, 1},
2259   { 0x1fd8, 0x1fdb, 1},
2260   { 0x1fe8, 0x1fec, 1},
2261   { 0x1ff8, 0x1ffb, 1},
2262   { 0x2102, 0x2107, 5},
2263   { 0x210b, 0x210d, 1},
2264   { 0x2110, 0x2112, 1},
2265   { 0x2115, 0x2115, 1},
2266   { 0x2119, 0x211d, 1},
2267   { 0x2124, 0x2128, 2},
2268   { 0x212a, 0x212d, 1},
2269   { 0x2130, 0x2133, 1},
2270   { 0x213e, 0x213f, 1},
2271   { 0x2145, 0x2183, 62},
2272   { 0x2c00, 0x2c2e, 1},
2273   { 0x2c60, 0x2c60, 1},
2274   { 0x2c62, 0x2c64, 1},
2275   { 0x2c67, 0x2c6b, 2},
2276   { 0x2c6d, 0x2c6f, 1},
2277   { 0x2c72, 0x2c75, 3},
2278   { 0x2c80, 0x2ce2, 2},
2279   { 0xa640, 0xa65e, 2},
2280   { 0xa662, 0xa66c, 2},
2281   { 0xa680, 0xa696, 2},
2282   { 0xa722, 0xa72e, 2},
2283   { 0xa732, 0xa76e, 2},
2284   { 0xa779, 0xa77b, 2},
2285   { 0xa77d, 0xa77e, 1},
2286   { 0xa780, 0xa786, 2},
2287   { 0xa78b, 0xa78b, 1},
2288   { 0xff21, 0xff3a, 1},
2289   { 0x10400, 0x10427, 1},
2290   { 0x1d400, 0x1d419, 1},
2291   { 0x1d434, 0x1d44d, 1},
2292   { 0x1d468, 0x1d481, 1},
2293   { 0x1d49c, 0x1d49c, 1},
2294   { 0x1d49e, 0x1d49f, 1},
2295   { 0x1d4a2, 0x1d4a2, 1},
2296   { 0x1d4a5, 0x1d4a6, 1},
2297   { 0x1d4a9, 0x1d4ac, 1},
2298   { 0x1d4ae, 0x1d4b5, 1},
2299   { 0x1d4d0, 0x1d4e9, 1},
2300   { 0x1d504, 0x1d505, 1},
2301   { 0x1d507, 0x1d50a, 1},
2302   { 0x1d50d, 0x1d514, 1},
2303   { 0x1d516, 0x1d51c, 1},
2304   { 0x1d538, 0x1d539, 1},
2305   { 0x1d53b, 0x1d53e, 1},
2306   { 0x1d540, 0x1d544, 1},
2307   { 0x1d546, 0x1d546, 1},
2308   { 0x1d54a, 0x1d550, 1},
2309   { 0x1d56c, 0x1d585, 1},
2310   { 0x1d5a0, 0x1d5b9, 1},
2311   { 0x1d5d4, 0x1d5ed, 1},
2312   { 0x1d608, 0x1d621, 1},
2313   { 0x1d63c, 0x1d655, 1},
2314   { 0x1d670, 0x1d689, 1},
2315   { 0x1d6a8, 0x1d6c0, 1},
2316   { 0x1d6e2, 0x1d6fa, 1},
2317   { 0x1d71c, 0x1d734, 1},
2318   { 0x1d756, 0x1d76e, 1},
2319   { 0x1d790, 0x1d7a8, 1},
2320   { 0x1d7ca, 0x1d7ca, 1},
2321 };
2322
2323 // Return true if C is in RANGES.
2324
2325 bool
2326 Lex::is_in_unicode_range(unsigned int c, const Unicode_range* ranges,
2327                          size_t range_size)
2328 {
2329   if (c < 0x100)
2330     {
2331       // The common case is a small value, and we know that it will be
2332       // in the first few entries of the table.  Do a linear scan
2333       // rather than a binary search.
2334       for (size_t i = 0; i < range_size; ++i)
2335         {
2336           const Unicode_range* p = &ranges[i];
2337           if (c <= p->high)
2338             {
2339               if (c < p->low)
2340                 return false;
2341               return (c - p->low) % p->stride == 0;
2342             }
2343         }
2344       return false;
2345     }
2346   else
2347     {
2348       size_t lo = 0;
2349       size_t hi = range_size;
2350       while (lo < hi)
2351         {
2352           size_t mid = lo + (hi - lo) / 2;
2353           const Unicode_range* p = &ranges[mid];
2354           if (c < p->low)
2355             hi = mid;
2356           else if (c > p->high)
2357             lo = mid + 1;
2358           else
2359             return (c - p->low) % p->stride == 0;
2360         }
2361       return false;
2362     }
2363 }
2364
2365 // Return whether C is a space character.
2366
2367 bool
2368 Lex::is_unicode_space(unsigned int c)
2369 {
2370   return Lex::is_in_unicode_range(c, unicode_space,
2371                                   ARRAY_SIZE(unicode_space));
2372 }
2373
2374 // Return whether C is a Unicode digit--a Unicode code point
2375 // classified as "Digit".
2376
2377 bool
2378 Lex::is_unicode_digit(unsigned int c)
2379 {
2380   return Lex::is_in_unicode_range(c, unicode_digits,
2381                                   ARRAY_SIZE(unicode_digits));
2382 }
2383
2384 // Return whether C is a Unicode letter--a Unicode code point
2385 // classified as "Letter".
2386
2387 bool
2388 Lex::is_unicode_letter(unsigned int c)
2389 {
2390   return Lex::is_in_unicode_range(c, unicode_letters,
2391                                   ARRAY_SIZE(unicode_letters));
2392 }
2393
2394 // Return whether C is a Unicode uppercase letter.  a Unicode code
2395 // point classified as "Letter, uppercase".
2396
2397 bool
2398 Lex::is_unicode_uppercase(unsigned int c)
2399 {
2400   return Lex::is_in_unicode_range(c, unicode_uppercase_letters,
2401                                   ARRAY_SIZE(unicode_uppercase_letters));
2402 }
2403
2404 // Return whether the identifier NAME should be exported.  NAME is a
2405 // mangled name which includes only ASCII characters.
2406
2407 bool
2408 Lex::is_exported_name(const std::string& name)
2409 {
2410   unsigned char c = name[0];
2411   if (c != '$')
2412     return c >= 'A' && c <= 'Z';
2413   else
2414     {
2415       const char* p = name.data();
2416       size_t len = name.length();
2417       if (len < 2 || p[1] != 'U')
2418         return false;
2419       unsigned int ci = 0;
2420       for (size_t i = 2; i < len && p[i] != '$'; ++i)
2421         {
2422           c = p[i];
2423           if (!hex_p(c))
2424             return false;
2425           ci <<= 4;
2426           ci |= hex_value(c);
2427         }
2428       return Lex::is_unicode_uppercase(ci);
2429     }
2430 }
2431
2432 // Return whether the identifier NAME contains an invalid character.
2433 // This is based on how we handle invalid characters in
2434 // gather_identifier.
2435
2436 bool
2437 Lex::is_invalid_identifier(const std::string& name)
2438 {
2439   return name.find("$INVALID$") != std::string::npos;
2440 }