nsz Git - cparser/blob - lexer.c

   1 #include <config.h>
   2
   3 #include "lexer.h"
   4 #include "token_t.h"
   5 #include "symbol_table_t.h"
   6 #include "adt/error.h"
   7 #include "adt/strset.h"
   8 #include "adt/util.h"
   9 #include "types.h"
  10 #include "type_t.h"
  11 #include "target_architecture.h"
  12 #include "parser.h"
  13
  14 #include <assert.h>
  15 #include <errno.h>
  16 #include <string.h>
  17 #include <stdbool.h>
  18 #include <ctype.h>
  19
  20 //#define DEBUG_CHARS
  21 #define MAX_PUTBACK 3
  22
  23 #ifdef _WIN32
  24 /* No strtold on windows and no replacement yet */
  25 #define strtold(s, e) strtod(s, e)
  26 #endif
  27
  28 #if defined HAS_SIGNED_CHAR
  29 typedef signed char char_type;
  30 #elif defined HAS_UNSIGNED_CHAR
  31 typedef unsigned char char_type;
  32 #else
  33 #       error signedness of char not determined
  34 #endif
  35
  36 static int         c;
  37 token_t            lexer_token;
  38 symbol_t          *symbol_L;
  39 static FILE       *input;
  40 static char        buf[1024 + MAX_PUTBACK];
  41 static const char *bufend;
  42 static const char *bufpos;
  43 static strset_t    stringset;
  44
  45 static void error_prefix_at(const char *input_name, unsigned linenr)
  46 {
  47         fprintf(stderr, "%s:%u: Error: ", input_name, linenr);
  48 }
  49
  50 static void error_prefix(void)
  51 {
  52         error_prefix_at(lexer_token.source_position.input_name,
  53                         lexer_token.source_position.linenr);
  54 }
  55
  56 static void parse_error(const char *msg)
  57 {
  58         error_prefix();
  59         fprintf(stderr, "%s\n", msg);
  60 }
  61
  62 static inline void next_real_char(void)
  63 {
  64         bufpos++;
  65         if(bufpos >= bufend) {
  66                 size_t s = fread(buf + MAX_PUTBACK, 1, sizeof(buf) - MAX_PUTBACK,
  67                                  input);
  68                 if(s == 0) {
  69                         c = EOF;
  70                         return;
  71                 }
  72                 bufpos = buf + MAX_PUTBACK;
  73                 bufend = buf + MAX_PUTBACK + s;
  74         }
  75         c = *(bufpos);
  76 }
  77
  78 static inline void put_back(int pc)
  79 {
  80         assert(bufpos >= buf);
  81         //assert(bufpos < buf+MAX_PUTBACK || *bufpos == pc);
  82
  83         char *p = buf + (bufpos - buf);
  84         *p = (char) pc;
  85
  86         /* going backwards in the buffer is legal as long as it's not more often
  87          * than MAX_PUTBACK */
  88         bufpos--;
  89
  90 #ifdef DEBUG_CHARS
  91         printf("putback '%c'\n", pc);
  92 #endif
  93 }
  94
  95 static inline void next_char(void);
  96
  97 #define MATCH_NEWLINE(code)                   \
  98         case '\r':                                \
  99                 next_char();                          \
 100                 if(c == '\n') {                       \
 101                         next_char();                      \
 102                 }                                     \
 103                 lexer_token.source_position.linenr++; \
 104                 code;                                 \
 105         case '\n':                                \
 106                 next_char();                          \
 107                 lexer_token.source_position.linenr++; \
 108                 code;
 109
 110 #define eat(c_type)  do { assert(c == c_type); next_char(); } while(0)
 111
 112 static void maybe_concat_lines(void)
 113 {
 114         eat('\\');
 115
 116         switch(c) {
 117         MATCH_NEWLINE(return;)
 118
 119         default:
 120                 break;
 121         }
 122
 123         put_back(c);
 124         c = '\\';
 125 }
 126
 127 static inline void next_char(void)
 128 {
 129         next_real_char();
 130
 131         /* filter trigraphs */
 132         if(UNLIKELY(c == '\\')) {
 133                 maybe_concat_lines();
 134                 goto end_of_next_char;
 135         }
 136
 137         if(LIKELY(c != '?'))
 138                 goto end_of_next_char;
 139
 140         next_real_char();
 141         if(LIKELY(c != '?')) {
 142                 put_back(c);
 143                 c = '?';
 144                 goto end_of_next_char;
 145         }
 146
 147         next_real_char();
 148         switch(c) {
 149         case '=': c = '#'; break;
 150         case '(': c = '['; break;
 151         case '/': c = '\\'; maybe_concat_lines(); break;
 152         case ')': c = ']'; break;
 153         case '\'': c = '^'; break;
 154         case '<': c = '{'; break;
 155         case '!': c = '|'; break;
 156         case '>': c = '}'; break;
 157         case '-': c = '~'; break;
 158         default:
 159                 put_back('?');
 160                 put_back(c);
 161                 c = '?';
 162                 break;
 163         }
 164
 165 end_of_next_char:;
 166 #ifdef DEBUG_CHARS
 167         printf("nchar '%c'\n", c);
 168 #endif
 169 }
 170
 171 #define SYMBOL_CHARS  \
 172         case 'a':         \
 173         case 'b':         \
 174         case 'c':         \
 175         case 'd':         \
 176         case 'e':         \
 177         case 'f':         \
 178         case 'g':         \
 179         case 'h':         \
 180         case 'i':         \
 181         case 'j':         \
 182         case 'k':         \
 183         case 'l':         \
 184         case 'm':         \
 185         case 'n':         \
 186         case 'o':         \
 187         case 'p':         \
 188         case 'q':         \
 189         case 'r':         \
 190         case 's':         \
 191         case 't':         \
 192         case 'u':         \
 193         case 'v':         \
 194         case 'w':         \
 195         case 'x':         \
 196         case 'y':         \
 197         case 'z':         \
 198         case 'A':         \
 199         case 'B':         \
 200         case 'C':         \
 201         case 'D':         \
 202         case 'E':         \
 203         case 'F':         \
 204         case 'G':         \
 205         case 'H':         \
 206         case 'I':         \
 207         case 'J':         \
 208         case 'K':         \
 209         case 'L':         \
 210         case 'M':         \
 211         case 'N':         \
 212         case 'O':         \
 213         case 'P':         \
 214         case 'Q':         \
 215         case 'R':         \
 216         case 'S':         \
 217         case 'T':         \
 218         case 'U':         \
 219         case 'V':         \
 220         case 'W':         \
 221         case 'X':         \
 222         case 'Y':         \
 223         case 'Z':         \
 224         case '_':
 225
 226 #define DIGITS        \
 227         case '0':         \
 228         case '1':         \
 229         case '2':         \
 230         case '3':         \
 231         case '4':         \
 232         case '5':         \
 233         case '6':         \
 234         case '7':         \
 235         case '8':         \
 236         case '9':
 237
 238 static void parse_symbol(void)
 239 {
 240         symbol_t *symbol;
 241         char     *string;
 242
 243         obstack_1grow(&symbol_obstack, (char) c);
 244         next_char();
 245
 246         while(1) {
 247                 switch(c) {
 248                 DIGITS
 249                 SYMBOL_CHARS
 250                         obstack_1grow(&symbol_obstack, (char) c);
 251                         next_char();
 252                         break;
 253
 254                 default:
 255                         goto end_symbol;
 256                 }
 257         }
 258
 259 end_symbol:
 260         obstack_1grow(&symbol_obstack, '\0');
 261
 262         string = obstack_finish(&symbol_obstack);
 263         symbol = symbol_table_insert(string);
 264
 265         lexer_token.type     = symbol->ID;
 266         lexer_token.v.symbol = symbol;
 267
 268         if(symbol->string != string) {
 269                 obstack_free(&symbol_obstack, string);
 270         }
 271 }
 272
 273 static void parse_integer_suffix(bool is_oct_hex)
 274 {
 275         bool is_unsigned  = false;
 276         bool min_long     = false;
 277         bool min_longlong = false;
 278
 279         if(c == 'U' || c == 'u') {
 280                 is_unsigned = true;
 281                 next_char();
 282                 if(c == 'L' || c == 'l') {
 283                         min_long = true;
 284                         next_char();
 285                         if(c == 'L' || c == 'l') {
 286                                 min_longlong = true;
 287                                 next_char();
 288                         }
 289                 }
 290         } else if(c == 'l' || c == 'L') {
 291                 min_long = true;
 292                 next_char();
 293                 if(c == 'l' || c == 'L') {
 294                         min_longlong = true;
 295                         next_char();
 296                         if(c == 'u' || c == 'U') {
 297                                 is_unsigned = true;
 298                                 next_char();
 299                         }
 300                 } else if(c == 'u' || c == 'U') {
 301                         is_unsigned = true;
 302                         next_char();
 303                         lexer_token.datatype = type_unsigned_long;
 304                 }
 305         }
 306
 307         if(!is_unsigned) {
 308                 long long v = lexer_token.v.intvalue;
 309                 if(!min_long) {
 310                         if(v >= TARGET_INT_MIN && v <= TARGET_INT_MAX) {
 311                                 lexer_token.datatype = type_int;
 312                                 return;
 313                         } else if(is_oct_hex && v >= 0 && v <= TARGET_UINT_MAX) {
 314                                 lexer_token.datatype = type_unsigned_int;
 315                                 return;
 316                         }
 317                 }
 318                 if(!min_longlong) {
 319                         if(v >= TARGET_LONG_MIN && v <= TARGET_LONG_MAX) {
 320                                 lexer_token.datatype = type_long;
 321                                 return;
 322                         } else if(is_oct_hex && v >= 0 && v <= TARGET_ULONG_MAX) {
 323                                 lexer_token.datatype = type_unsigned_long;
 324                                 return;
 325                         }
 326                 }
 327                 unsigned long long uv = (unsigned long long) v;
 328                 if(is_oct_hex && uv > (unsigned long long) TARGET_LONGLONG_MAX) {
 329                         lexer_token.datatype = type_unsigned_long_long;
 330                         return;
 331                 }
 332
 333                 lexer_token.datatype = type_long_long;
 334         } else {
 335                 unsigned long long v = (unsigned long long) lexer_token.v.intvalue;
 336                 if(!min_long && v <= TARGET_UINT_MAX) {
 337                         lexer_token.datatype = type_unsigned_int;
 338                         return;
 339                 }
 340                 if(!min_longlong && v <= TARGET_ULONG_MAX) {
 341                         lexer_token.datatype = type_unsigned_long;
 342                         return;
 343                 }
 344                 lexer_token.datatype = type_unsigned_long_long;
 345         }
 346 }
 347
 348 static void parse_floating_suffix(void)
 349 {
 350         switch(c) {
 351         /* TODO: do something usefull with the suffixes... */
 352         case 'f':
 353         case 'F':
 354                 next_char();
 355                 lexer_token.datatype = type_float;
 356                 break;
 357         case 'l':
 358         case 'L':
 359                 next_char();
 360                 lexer_token.datatype = type_long_double;
 361                 break;
 362         default:
 363                 lexer_token.datatype = type_double;
 364                 break;
 365         }
 366 }
 367
 368 /**
 369  * A replacement for strtoull. Only those parts needed for
 370  * our parser are implemented.
 371  */
 372 static unsigned long long parse_int_string(const char *s, const char **endptr, int base) {
 373         unsigned long long v = 0;
 374
 375         switch (base) {
 376         case 16:
 377                 for (;; ++s) {
 378                         /* check for overrun */
 379                         if (v >= 0x1000000000000000ULL)
 380                                 break;
 381                         switch (tolower(*s)) {
 382                         case '0': v <<= 4; break;
 383                         case '1': v <<= 4; v |= 0x1; break;
 384                         case '2': v <<= 4; v |= 0x2; break;
 385                         case '3': v <<= 4; v |= 0x3; break;
 386                         case '4': v <<= 4; v |= 0x4; break;
 387                         case '5': v <<= 4; v |= 0x5; break;
 388                         case '6': v <<= 4; v |= 0x6; break;
 389                         case '7': v <<= 4; v |= 0x7; break;
 390                         case '8': v <<= 4; v |= 0x8; break;
 391                         case '9': v <<= 4; v |= 0x9; break;
 392                         case 'a': v <<= 4; v |= 0xa; break;
 393                         case 'b': v <<= 4; v |= 0xb; break;
 394                         case 'c': v <<= 4; v |= 0xc; break;
 395                         case 'd': v <<= 4; v |= 0xd; break;
 396                         case 'e': v <<= 4; v |= 0xe; break;
 397                         case 'f': v <<= 4; v |= 0xf; break;
 398                         default:
 399                                 goto end;
 400                         }
 401                 }
 402                 break;
 403         case 8:
 404                 for (;; ++s) {
 405                         /* check for overrun */
 406                         if (v >= 0x2000000000000000ULL)
 407                                 break;
 408                         switch (tolower(*s)) {
 409                         case '0': v <<= 3; break;
 410                         case '1': v <<= 3; v |= 1; break;
 411                         case '2': v <<= 3; v |= 2; break;
 412                         case '3': v <<= 3; v |= 3; break;
 413                         case '4': v <<= 3; v |= 4; break;
 414                         case '5': v <<= 3; v |= 5; break;
 415                         case '6': v <<= 3; v |= 6; break;
 416                         case '7': v <<= 3; v |= 7; break;
 417                         default:
 418                                 goto end;
 419                         }
 420                 }
 421                 break;
 422         case 10:
 423                 for (;; ++s) {
 424                         /* check for overrun */
 425                         if (v > 0x1999999999999999ULL)
 426                                 break;
 427                         switch (tolower(*s)) {
 428                         case '0': v *= 10; break;
 429                         case '1': v *= 10; v += 1; break;
 430                         case '2': v *= 10; v += 2; break;
 431                         case '3': v *= 10; v += 3; break;
 432                         case '4': v *= 10; v += 4; break;
 433                         case '5': v *= 10; v += 5; break;
 434                         case '6': v *= 10; v += 6; break;
 435                         case '7': v *= 10; v += 7; break;
 436                         case '8': v *= 10; v += 8; break;
 437                         case '9': v *= 10; v += 9; break;
 438                         default:
 439                                 goto end;
 440                         }
 441                 }
 442                 break;
 443         default:
 444                 assert(0);
 445                 break;
 446         }
 447 end:
 448         *endptr = s;
 449         return v;
 450 }
 451
 452 static void parse_number_hex(void)
 453 {
 454         assert(c == 'x' || c == 'X');
 455         next_char();
 456
 457         while(isxdigit(c)) {
 458                 obstack_1grow(&symbol_obstack, (char) c);
 459                 next_char();
 460         }
 461         obstack_1grow(&symbol_obstack, '\0');
 462         char *string = obstack_finish(&symbol_obstack);
 463
 464         if(c == '.' || c == 'p' || c == 'P') {
 465                 next_char();
 466                 panic("Hex floating point numbers not implemented yet");
 467         }
 468         if(*string == '\0') {
 469                 parse_error("invalid hex number");
 470                 lexer_token.type = T_ERROR;
 471         }
 472
 473         const char *endptr;
 474         lexer_token.type       = T_INTEGER;
 475         lexer_token.v.intvalue = parse_int_string(string, &endptr, 16);
 476         if(*endptr != '\0') {
 477                 parse_error("hex number literal too long");
 478         }
 479
 480         obstack_free(&symbol_obstack, string);
 481         parse_integer_suffix(true);
 482 }
 483
 484 static inline bool is_octal_digit(int chr)
 485 {
 486         return '0' <= chr && chr <= '7';
 487 }
 488
 489 static void parse_number_oct(void)
 490 {
 491         while(is_octal_digit(c)) {
 492                 obstack_1grow(&symbol_obstack, (char) c);
 493                 next_char();
 494         }
 495         obstack_1grow(&symbol_obstack, '\0');
 496         char *string = obstack_finish(&symbol_obstack);
 497
 498         const char *endptr;
 499         lexer_token.type       = T_INTEGER;
 500         lexer_token.v.intvalue = parse_int_string(string, &endptr, 8);
 501         if(*endptr != '\0') {
 502                 parse_error("octal number literal too long");
 503         }
 504
 505         obstack_free(&symbol_obstack, string);
 506         parse_integer_suffix(true);
 507 }
 508
 509 static void parse_number_dec(void)
 510 {
 511         bool is_float = false;
 512         while(isdigit(c)) {
 513                 obstack_1grow(&symbol_obstack, (char) c);
 514                 next_char();
 515         }
 516
 517         if(c == '.') {
 518                 obstack_1grow(&symbol_obstack, '.');
 519                 next_char();
 520
 521                 while(isdigit(c)) {
 522                         obstack_1grow(&symbol_obstack, (char) c);
 523                         next_char();
 524                 }
 525                 is_float = true;
 526         }
 527         if(c == 'e' || c == 'E') {
 528                 obstack_1grow(&symbol_obstack, 'e');
 529                 next_char();
 530
 531                 if(c == '-' || c == '+') {
 532                         obstack_1grow(&symbol_obstack, (char) c);
 533                         next_char();
 534                 }
 535
 536                 while(isdigit(c)) {
 537                         obstack_1grow(&symbol_obstack, (char) c);
 538                         next_char();
 539                 }
 540                 is_float = true;
 541         }
 542
 543         obstack_1grow(&symbol_obstack, '\0');
 544         char *string = obstack_finish(&symbol_obstack);
 545
 546         if(is_float) {
 547                 char *endptr;
 548                 lexer_token.type         = T_FLOATINGPOINT;
 549                 lexer_token.v.floatvalue = strtold(string, &endptr);
 550
 551                 if(*endptr != '\0') {
 552                         parse_error("invalid number literal");
 553                 }
 554
 555                 parse_floating_suffix();
 556         } else {
 557                 const char *endptr;
 558                 lexer_token.type       = T_INTEGER;
 559                 lexer_token.v.intvalue = parse_int_string(string, &endptr, 10);
 560
 561                 if(*endptr != '\0') {
 562                         parse_error("invalid number literal");
 563                 }
 564
 565                 parse_integer_suffix(false);
 566         }
 567         obstack_free(&symbol_obstack, string);
 568 }
 569
 570 static void parse_number(void)
 571 {
 572         if (c == '0') {
 573                 next_char();
 574                 switch (c) {
 575                         case 'X':
 576                         case 'x':
 577                                 parse_number_hex();
 578                                 break;
 579                         case '0':
 580                         case '1':
 581                         case '2':
 582                         case '3':
 583                         case '4':
 584                         case '5':
 585                         case '6':
 586                         case '7':
 587                                 parse_number_oct();
 588                                 break;
 589                         case '8':
 590                         case '9':
 591                                 next_char();
 592                                 parse_error("invalid octal number");
 593                                 lexer_token.type = T_ERROR;
 594                                 return;
 595                         case '.':
 596                         case 'e':
 597                         case 'E':
 598                         default:
 599                                 obstack_1grow(&symbol_obstack, '0');
 600                                 parse_number_dec();
 601                                 return;
 602                 }
 603         } else {
 604                 parse_number_dec();
 605         }
 606 }
 607
 608 static int parse_octal_sequence(const int first_digit)
 609 {
 610         assert(is_octal_digit(first_digit));
 611         int value = first_digit - '0';
 612         if (!is_octal_digit(c)) return value;
 613         value = 8 * value + c - '0';
 614         next_char();
 615         if (!is_octal_digit(c)) return value;
 616         value = 8 * value + c - '0';
 617         next_char();
 618         return (char_type)value;
 619 }
 620
 621 static int parse_hex_sequence(void)
 622 {
 623         int value = 0;
 624         while(1) {
 625                 if (c >= '0' && c <= '9') {
 626                         value = 16 * value + c - '0';
 627                 } else if ('A' <= c && c <= 'F') {
 628                         value = 16 * value + c - 'A' + 10;
 629                 } else if ('a' <= c && c <= 'f') {
 630                         value = 16 * value + c - 'a' + 10;
 631                 } else {
 632                         break;
 633                 }
 634                 next_char();
 635         }
 636
 637         return (char_type)value;
 638 }
 639
 640 static int parse_escape_sequence(void)
 641 {
 642         eat('\\');
 643
 644         int ec = c;
 645         next_char();
 646
 647         switch(ec) {
 648         case '"':  return '"';
 649         case '\'': return '\'';
 650         case '\\': return '\\';
 651         case '?': return '\?';
 652         case 'a': return '\a';
 653         case 'b': return '\b';
 654         case 'f': return '\f';
 655         case 'n': return '\n';
 656         case 'r': return '\r';
 657         case 't': return '\t';
 658         case 'v': return '\v';
 659         case 'x':
 660                 return parse_hex_sequence();
 661         case '0':
 662         case '1':
 663         case '2':
 664         case '3':
 665         case '4':
 666         case '5':
 667         case '6':
 668         case '7':
 669                 return parse_octal_sequence(ec);
 670         case EOF:
 671                 parse_error("reached end of file while parsing escape sequence");
 672                 return EOF;
 673         default:
 674                 parse_error("unknown escape sequence");
 675                 return EOF;
 676         }
 677 }
 678
 679 const char *concat_strings(const char *s1, const char *s2)
 680 {
 681         size_t  len1   = strlen(s1);
 682         size_t  len2   = strlen(s2);
 683
 684         char   *concat = obstack_alloc(&symbol_obstack, len1 + len2 + 1);
 685         memcpy(concat, s1, len1);
 686         memcpy(concat + len1, s2, len2 + 1);
 687
 688         const char *result = strset_insert(&stringset, concat);
 689         if(result != concat) {
 690                 obstack_free(&symbol_obstack, concat);
 691         }
 692
 693         return result;
 694 }
 695
 696 static void parse_string_literal(void)
 697 {
 698         unsigned    start_linenr = lexer_token.source_position.linenr;
 699         char       *string;
 700         const char *result;
 701
 702         assert(c == '"');
 703         next_char();
 704
 705         int tc;
 706         while(1) {
 707                 switch(c) {
 708                 case '\\':
 709                         tc = parse_escape_sequence();
 710                         obstack_1grow(&symbol_obstack, (char) tc);
 711                         break;
 712
 713                 case EOF:
 714                         error_prefix_at(lexer_token.source_position.input_name,
 715                                         start_linenr);
 716                         fprintf(stderr, "string has no end\n");
 717                         lexer_token.type = T_ERROR;
 718                         return;
 719
 720                 case '"':
 721                         next_char();
 722                         goto end_of_string;
 723
 724                 default:
 725                         obstack_1grow(&symbol_obstack, (char) c);
 726                         next_char();
 727                         break;
 728                 }
 729         }
 730
 731 end_of_string:
 732
 733         /* TODO: concatenate multiple strings separated by whitespace... */
 734
 735         /* add finishing 0 to the string */
 736         obstack_1grow(&symbol_obstack, '\0');
 737         string = obstack_finish(&symbol_obstack);
 738
 739         /* check if there is already a copy of the string */
 740         result = strset_insert(&stringset, string);
 741         if(result != string) {
 742                 obstack_free(&symbol_obstack, string);
 743         }
 744
 745         lexer_token.type     = T_STRING_LITERAL;
 746         lexer_token.v.string = result;
 747 }
 748
 749 static void parse_wide_character_constant(void)
 750 {
 751         eat('\'');
 752
 753         int found_char = 0;
 754         while(1) {
 755                 switch(c) {
 756                 case '\\':
 757                         found_char = parse_escape_sequence();
 758                         break;
 759
 760                 MATCH_NEWLINE(
 761                         parse_error("newline while parsing character constant");
 762                         break;
 763                 )
 764
 765                 case '\'':
 766                         next_char();
 767                         goto end_of_wide_char_constant;
 768
 769                 case EOF:
 770                         parse_error("EOF while parsing character constant");
 771                         lexer_token.type = T_ERROR;
 772                         return;
 773
 774                 default:
 775                         if(found_char != 0) {
 776                                 parse_error("more than 1 characters in character "
 777                                             "constant");
 778                                 goto end_of_wide_char_constant;
 779                         } else {
 780                                 found_char = c;
 781                                 next_char();
 782                         }
 783                         break;
 784                 }
 785         }
 786
 787 end_of_wide_char_constant:
 788         lexer_token.type       = T_INTEGER;
 789         lexer_token.v.intvalue = found_char;
 790         lexer_token.datatype   = type_wchar_t;
 791 }
 792
 793 static void parse_wide_string_literal(void)
 794 {
 795         const unsigned start_linenr = lexer_token.source_position.linenr;
 796
 797         assert(c == '"');
 798         next_char();
 799
 800         while(1) {
 801                 switch(c) {
 802                         case '\\': {
 803                                 wchar_rep_t tc = parse_escape_sequence();
 804                                 obstack_grow(&symbol_obstack, &tc, sizeof(tc));
 805                                 break;
 806                         }
 807
 808                         case EOF:
 809                                 error_prefix_at(lexer_token.source_position.input_name,
 810                                                 start_linenr);
 811                                 fprintf(stderr, "string has no end\n");
 812                                 lexer_token.type = T_ERROR;
 813                                 return;
 814
 815                         case '"':
 816                                 next_char();
 817                                 goto end_of_string;
 818
 819                         default: {
 820                                 wchar_rep_t tc = c;
 821                                 obstack_grow(&symbol_obstack, &tc, sizeof(tc));
 822                                 next_char();
 823                                 break;
 824                         }
 825                 }
 826         }
 827
 828 end_of_string:;
 829
 830         /* TODO: concatenate multiple strings separated by whitespace... */
 831
 832         /* add finishing 0 to the string */
 833         wchar_rep_t nul = L'\0';
 834         obstack_grow(&symbol_obstack, &nul, sizeof(nul));
 835         const size_t             size   = (size_t)obstack_object_size(&symbol_obstack) / sizeof(wchar_rep_t);
 836         const wchar_rep_t *const string = obstack_finish(&symbol_obstack);
 837
 838 #if 0 /* TODO hash */
 839         /* check if there is already a copy of the string */
 840         const wchar_rep_t *const result = strset_insert(&stringset, string);
 841         if(result != string) {
 842                 obstack_free(&symbol_obstack, string);
 843         }
 844 #else
 845         const wchar_rep_t *const result = string;
 846 #endif
 847
 848         lexer_token.type                = T_WIDE_STRING_LITERAL;
 849         lexer_token.v.wide_string.begin = result;
 850         lexer_token.v.wide_string.size  = size;
 851 }
 852
 853 static void parse_character_constant(void)
 854 {
 855         eat('\'');
 856
 857         int found_char = 0;
 858         while(1) {
 859                 switch(c) {
 860                 case '\\':
 861                         found_char = parse_escape_sequence();
 862                         break;
 863
 864                 MATCH_NEWLINE(
 865                         parse_error("newline while parsing character constant");
 866                         break;
 867                 )
 868
 869                 case '\'':
 870                         next_char();
 871                         goto end_of_char_constant;
 872
 873                 case EOF:
 874                         parse_error("EOF while parsing character constant");
 875                         lexer_token.type = T_ERROR;
 876                         return;
 877
 878                 default:
 879                         if(found_char != 0) {
 880                                 parse_error("more than 1 characters in character "
 881                                             "constant");
 882                                 goto end_of_char_constant;
 883                         } else {
 884                                 found_char = c;
 885                                 next_char();
 886                         }
 887                         break;
 888                 }
 889         }
 890
 891 end_of_char_constant:
 892         lexer_token.type       = T_INTEGER;
 893         lexer_token.v.intvalue = found_char;
 894         lexer_token.datatype   = type_int;
 895 }
 896
 897 static void skip_multiline_comment(void)
 898 {
 899         unsigned start_linenr = lexer_token.source_position.linenr;
 900
 901         while(1) {
 902                 switch(c) {
 903                 case '*':
 904                         next_char();
 905                         if(c == '/') {
 906                                 next_char();
 907                                 return;
 908                         }
 909                         break;
 910
 911                 MATCH_NEWLINE(break;)
 912
 913                 case EOF:
 914                         error_prefix_at(lexer_token.source_position.input_name,
 915                                         start_linenr);
 916                         fprintf(stderr, "at end of file while looking for comment end\n");
 917                         return;
 918
 919                 default:
 920                         next_char();
 921                         break;
 922                 }
 923         }
 924 }
 925
 926 static void skip_line_comment(void)
 927 {
 928         while(1) {
 929                 switch(c) {
 930                 case EOF:
 931                         return;
 932
 933                 case '\n':
 934                 case '\r':
 935                         return;
 936
 937                 default:
 938                         next_char();
 939                         break;
 940                 }
 941         }
 942 }
 943
 944 static token_t pp_token;
 945
 946 static inline void next_pp_token(void)
 947 {
 948         lexer_next_preprocessing_token();
 949         pp_token = lexer_token;
 950 }
 951
 952 static void eat_until_newline(void)
 953 {
 954         while(pp_token.type != '\n' && pp_token.type != T_EOF) {
 955                 next_pp_token();
 956         }
 957 }
 958
 959 static void error_directive(void)
 960 {
 961         error_prefix();
 962         fprintf(stderr, "#error directive: \n");
 963
 964         /* parse pp-tokens until new-line */
 965 }
 966
 967 static void define_directive(void)
 968 {
 969         lexer_next_preprocessing_token();
 970         if(lexer_token.type != T_IDENTIFIER) {
 971                 parse_error("expected identifier after #define\n");
 972                 eat_until_newline();
 973         }
 974 }
 975
 976 static void ifdef_directive(int is_ifndef)
 977 {
 978         (void) is_ifndef;
 979         lexer_next_preprocessing_token();
 980         //expect_identifier();
 981         //extect_newline();
 982 }
 983
 984 static void endif_directive(void)
 985 {
 986         //expect_newline();
 987 }
 988
 989 static void parse_line_directive(void)
 990 {
 991         if(pp_token.type != T_INTEGER) {
 992                 parse_error("expected integer");
 993         } else {
 994                 lexer_token.source_position.linenr = (unsigned int)(pp_token.v.intvalue - 1);
 995                 next_pp_token();
 996         }
 997         if(pp_token.type == T_STRING_LITERAL) {
 998                 lexer_token.source_position.input_name = pp_token.v.string;
 999                 next_pp_token();
1000         }
1001
1002         eat_until_newline();
1003 }
1004
1005 static void parse_preprocessor_identifier(void)
1006 {
1007         assert(pp_token.type == T_IDENTIFIER);
1008         symbol_t *symbol = pp_token.v.symbol;
1009
1010         switch(symbol->pp_ID) {
1011         case TP_include:
1012                 printf("include - enable header name parsing!\n");
1013                 break;
1014         case TP_define:
1015                 define_directive();
1016                 break;
1017         case TP_ifdef:
1018                 ifdef_directive(0);
1019                 break;
1020         case TP_ifndef:
1021                 ifdef_directive(1);
1022                 break;
1023         case TP_endif:
1024                 endif_directive();
1025                 break;
1026         case TP_line:
1027                 next_pp_token();
1028                 parse_line_directive();
1029                 break;
1030         case TP_if:
1031         case TP_else:
1032         case TP_elif:
1033         case TP_undef:
1034         case TP_error:
1035                 error_directive();
1036                 break;
1037         case TP_pragma:
1038                 break;
1039         }
1040 }
1041
1042 static void parse_preprocessor_directive(void)
1043 {
1044         next_pp_token();
1045
1046         switch(pp_token.type) {
1047         case T_IDENTIFIER:
1048                 parse_preprocessor_identifier();
1049                 break;
1050         case T_INTEGER:
1051                 parse_line_directive();
1052                 break;
1053         default:
1054                 parse_error("invalid preprocessor directive");
1055                 eat_until_newline();
1056                 break;
1057         }
1058 }
1059
1060 #define MAYBE_PROLOG                                       \
1061                         next_char();                                   \
1062                         while(1) {                                     \
1063                                 switch(c) {
1064
1065 #define MAYBE(ch, set_type)                                \
1066                                 case ch:                                   \
1067                                         next_char();                           \
1068                                         lexer_token.type = set_type;           \
1069                                         return;
1070
1071 #define ELSE_CODE(code)                                    \
1072                                 default:                                   \
1073                                         code;                                  \
1074                                 }                                          \
1075                         } /* end of while(1) */                        \
1076                         break;
1077
1078 #define ELSE(set_type)                                     \
1079                 ELSE_CODE(                                         \
1080                         lexer_token.type = set_type;                   \
1081                         return;                                        \
1082                 )
1083
1084 void lexer_next_preprocessing_token(void)
1085 {
1086         while(1) {
1087                 switch(c) {
1088                 case ' ':
1089                 case '\t':
1090                         next_char();
1091                         break;
1092
1093                 MATCH_NEWLINE(
1094                         lexer_token.type = '\n';
1095                         return;
1096                 )
1097
1098                 SYMBOL_CHARS
1099                         parse_symbol();
1100                         /* might be a wide string ( L"string" ) */
1101                         if(lexer_token.type == T_IDENTIFIER &&
1102                             lexer_token.v.symbol == symbol_L) {
1103                             if(c == '"') {
1104                                         parse_wide_string_literal();
1105                                 } else if(c == '\'') {
1106                                         parse_wide_character_constant();
1107                                 }
1108                         }
1109                         return;
1110
1111                 DIGITS
1112                         parse_number();
1113                         return;
1114
1115                 case '"':
1116                         parse_string_literal();
1117                         return;
1118
1119                 case '\'':
1120                         parse_character_constant();
1121                         return;
1122
1123                 case '.':
1124                         MAYBE_PROLOG
1125                                 case '.':
1126                                         MAYBE_PROLOG
1127                                         MAYBE('.', T_DOTDOTDOT)
1128                                         ELSE_CODE(
1129                                                 put_back(c);
1130                                                 c = '.';
1131                                                 lexer_token.type = '.';
1132                                                 return;
1133                                         )
1134                         ELSE('.')
1135                 case '&':
1136                         MAYBE_PROLOG
1137                         MAYBE('&', T_ANDAND)
1138                         MAYBE('=', T_ANDEQUAL)
1139                         ELSE('&')
1140                 case '*':
1141                         MAYBE_PROLOG
1142                         MAYBE('=', T_ASTERISKEQUAL)
1143                         ELSE('*')
1144                 case '+':
1145                         MAYBE_PROLOG
1146                         MAYBE('+', T_PLUSPLUS)
1147                         MAYBE('=', T_PLUSEQUAL)
1148                         ELSE('+')
1149                 case '-':
1150                         MAYBE_PROLOG
1151                         MAYBE('>', T_MINUSGREATER)
1152                         MAYBE('-', T_MINUSMINUS)
1153                         MAYBE('=', T_MINUSEQUAL)
1154                         ELSE('-')
1155                 case '!':
1156                         MAYBE_PROLOG
1157                         MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1158                         ELSE('!')
1159                 case '/':
1160                         MAYBE_PROLOG
1161                         MAYBE('=', T_SLASHEQUAL)
1162                                 case '*':
1163                                         next_char();
1164                                         skip_multiline_comment();
1165                                         lexer_next_preprocessing_token();
1166                                         return;
1167                                 case '/':
1168                                         next_char();
1169                                         skip_line_comment();
1170                                         lexer_next_preprocessing_token();
1171                                         return;
1172                         ELSE('/')
1173                 case '%':
1174                         MAYBE_PROLOG
1175                         MAYBE('>', T_PERCENTGREATER)
1176                         MAYBE('=', T_PERCENTEQUAL)
1177                                 case ':':
1178                                         MAYBE_PROLOG
1179                                                 case '%':
1180                                                         MAYBE_PROLOG
1181                                                         MAYBE(':', T_PERCENTCOLONPERCENTCOLON)
1182                                                         ELSE_CODE(
1183                                                                 put_back(c);
1184                                                                 c = '%';
1185                                                                 lexer_token.type = T_PERCENTCOLON;
1186                                                                 return;
1187                                                         )
1188                                         ELSE(T_PERCENTCOLON)
1189                         ELSE('%')
1190                 case '<':
1191                         MAYBE_PROLOG
1192                         MAYBE(':', T_LESSCOLON)
1193                         MAYBE('%', T_LESSPERCENT)
1194                         MAYBE('=', T_LESSEQUAL)
1195                                 case '<':
1196                                         MAYBE_PROLOG
1197                                         MAYBE('=', T_LESSLESSEQUAL)
1198                                         ELSE(T_LESSLESS)
1199                         ELSE('<')
1200                 case '>':
1201                         MAYBE_PROLOG
1202                         MAYBE('=', T_GREATEREQUAL)
1203                                 case '>':
1204                                         MAYBE_PROLOG
1205                                         MAYBE('=', T_GREATERGREATEREQUAL)
1206                                         ELSE(T_GREATERGREATER)
1207                         ELSE('>')
1208                 case '^':
1209                         MAYBE_PROLOG
1210                         MAYBE('=', T_CARETEQUAL)
1211                         ELSE('^')
1212                 case '|':
1213                         MAYBE_PROLOG
1214                         MAYBE('=', T_PIPEEQUAL)
1215                         MAYBE('|', T_PIPEPIPE)
1216                         ELSE('|')
1217                 case ':':
1218                         MAYBE_PROLOG
1219                         MAYBE('>', T_COLONGREATER)
1220                         ELSE(':')
1221                 case '=':
1222                         MAYBE_PROLOG
1223                         MAYBE('=', T_EQUALEQUAL)
1224                         ELSE('=')
1225                 case '#':
1226                         MAYBE_PROLOG
1227                         MAYBE('#', T_HASHHASH)
1228                         ELSE('#')
1229
1230                 case '?':
1231                 case '[':
1232                 case ']':
1233                 case '(':
1234                 case ')':
1235                 case '{':
1236                 case '}':
1237                 case '~':
1238                 case ';':
1239                 case ',':
1240                 case '\\':
1241                         lexer_token.type = c;
1242                         next_char();
1243                         return;
1244
1245                 case EOF:
1246                         lexer_token.type = T_EOF;
1247                         return;
1248
1249                 default:
1250                         next_char();
1251                         error_prefix();
1252                         fprintf(stderr, "unknown character '%c' found\n", c);
1253                         lexer_token.type = T_ERROR;
1254                         return;
1255                 }
1256         }
1257 }
1258
1259 void lexer_next_token(void)
1260 {
1261         lexer_next_preprocessing_token();
1262         if(lexer_token.type != '\n')
1263                 return;
1264
1265 newline_found:
1266         do {
1267                 lexer_next_preprocessing_token();
1268         } while(lexer_token.type == '\n');
1269
1270         if(lexer_token.type == '#') {
1271                 parse_preprocessor_directive();
1272                 goto newline_found;
1273         }
1274 }
1275
1276 void init_lexer(void)
1277 {
1278         strset_init(&stringset);
1279 }
1280
1281 void lexer_open_stream(FILE *stream, const char *input_name)
1282 {
1283         input                                  = stream;
1284         lexer_token.source_position.linenr     = 0;
1285         lexer_token.source_position.input_name = input_name;
1286
1287         symbol_L = symbol_table_insert("L");
1288
1289         /* place a virtual \n at the beginning so the lexer knows that we're
1290          * at the beginning of a line */
1291         c = '\n';
1292 }
1293
1294 void exit_lexer(void)
1295 {
1296         strset_destroy(&stringset);
1297 }
1298
1299 static __attribute__((unused))
1300 void dbg_pos(const source_position_t source_position)
1301 {
1302         fprintf(stdout, "%s:%u\n", source_position.input_name,
1303                 source_position.linenr);
1304         fflush(stdout);
1305 }