nsz Git - cparser/blob - lexer.c

   1 #include <config.h>
   2
   3 #include "diagnostic.h"
   4 #include "lexer.h"
   5 #include "token_t.h"
   6 #include "symbol_table_t.h"
   7 #include "adt/error.h"
   8 #include "adt/strset.h"
   9 #include "adt/util.h"
  10 #include "types.h"
  11 #include "type_t.h"
  12 #include "target_architecture.h"
  13 #include "parser.h"
  14 #include "warning.h"
  15
  16 #include <assert.h>
  17 #include <errno.h>
  18 #include <string.h>
  19 #include <stdbool.h>
  20 #include <ctype.h>
  21
  22 //#define DEBUG_CHARS
  23 #define MAX_PUTBACK 3
  24
  25 #ifdef _WIN32
  26 /* No strtold on windows and no replacement yet */
  27 #define strtold(s, e) strtod(s, e)
  28 #endif
  29
  30 #if defined HAS_SIGNED_CHAR
  31 typedef signed char char_type;
  32 #elif defined HAS_UNSIGNED_CHAR
  33 typedef unsigned char char_type;
  34 #else
  35 #       error signedness of char not determined
  36 #endif
  37
  38 static int         c;
  39 token_t            lexer_token;
  40 symbol_t          *symbol_L;
  41 static FILE       *input;
  42 static char        buf[1024 + MAX_PUTBACK];
  43 static const char *bufend;
  44 static const char *bufpos;
  45 static strset_t    stringset;
  46
  47 static void error_prefix_at(const char *input_name, unsigned linenr)
  48 {
  49         fprintf(stderr, "%s:%u: Error: ", input_name, linenr);
  50 }
  51
  52 static void error_prefix(void)
  53 {
  54         error_prefix_at(lexer_token.source_position.input_name,
  55                         lexer_token.source_position.linenr);
  56 }
  57
  58 static void parse_error(const char *msg)
  59 {
  60         error_prefix();
  61         fprintf(stderr, "%s\n", msg);
  62 }
  63
  64 static inline void next_real_char(void)
  65 {
  66         assert(bufpos <= bufend);
  67         if (bufpos >= bufend) {
  68                 size_t s = fread(buf + MAX_PUTBACK, 1, sizeof(buf) - MAX_PUTBACK,
  69                                  input);
  70                 if(s == 0) {
  71                         c = EOF;
  72                         return;
  73                 }
  74                 bufpos = buf + MAX_PUTBACK;
  75                 bufend = buf + MAX_PUTBACK + s;
  76         }
  77         c = *bufpos++;
  78 }
  79
  80 static inline void put_back(int pc)
  81 {
  82         assert(bufpos > buf);
  83         *(--bufpos - buf + buf) = (char) pc;
  84
  85 #ifdef DEBUG_CHARS
  86         printf("putback '%c'\n", pc);
  87 #endif
  88 }
  89
  90 static inline void next_char(void);
  91
  92 #define MATCH_NEWLINE(code)                   \
  93         case '\r':                                \
  94                 next_char();                          \
  95                 if(c == '\n') {                       \
  96                         next_char();                      \
  97                 }                                     \
  98                 lexer_token.source_position.linenr++; \
  99                 code                                  \
 100         case '\n':                                \
 101                 next_char();                          \
 102                 lexer_token.source_position.linenr++; \
 103                 code
 104
 105 #define eat(c_type)  do { assert(c == c_type); next_char(); } while(0)
 106
 107 static void maybe_concat_lines(void)
 108 {
 109         eat('\\');
 110
 111         switch(c) {
 112         MATCH_NEWLINE(return;)
 113
 114         default:
 115                 break;
 116         }
 117
 118         put_back(c);
 119         c = '\\';
 120 }
 121
 122 static inline void next_char(void)
 123 {
 124         next_real_char();
 125
 126         /* filter trigraphs */
 127         if(UNLIKELY(c == '\\')) {
 128                 maybe_concat_lines();
 129                 goto end_of_next_char;
 130         }
 131
 132         if(LIKELY(c != '?'))
 133                 goto end_of_next_char;
 134
 135         next_real_char();
 136         if(LIKELY(c != '?')) {
 137                 put_back(c);
 138                 c = '?';
 139                 goto end_of_next_char;
 140         }
 141
 142         next_real_char();
 143         switch(c) {
 144         case '=': c = '#'; break;
 145         case '(': c = '['; break;
 146         case '/': c = '\\'; maybe_concat_lines(); break;
 147         case ')': c = ']'; break;
 148         case '\'': c = '^'; break;
 149         case '<': c = '{'; break;
 150         case '!': c = '|'; break;
 151         case '>': c = '}'; break;
 152         case '-': c = '~'; break;
 153         default:
 154                 put_back(c);
 155                 put_back('?');
 156                 c = '?';
 157                 break;
 158         }
 159
 160 end_of_next_char:;
 161 #ifdef DEBUG_CHARS
 162         printf("nchar '%c'\n", c);
 163 #endif
 164 }
 165
 166 #define SYMBOL_CHARS  \
 167         case 'a':         \
 168         case 'b':         \
 169         case 'c':         \
 170         case 'd':         \
 171         case 'e':         \
 172         case 'f':         \
 173         case 'g':         \
 174         case 'h':         \
 175         case 'i':         \
 176         case 'j':         \
 177         case 'k':         \
 178         case 'l':         \
 179         case 'm':         \
 180         case 'n':         \
 181         case 'o':         \
 182         case 'p':         \
 183         case 'q':         \
 184         case 'r':         \
 185         case 's':         \
 186         case 't':         \
 187         case 'u':         \
 188         case 'v':         \
 189         case 'w':         \
 190         case 'x':         \
 191         case 'y':         \
 192         case 'z':         \
 193         case 'A':         \
 194         case 'B':         \
 195         case 'C':         \
 196         case 'D':         \
 197         case 'E':         \
 198         case 'F':         \
 199         case 'G':         \
 200         case 'H':         \
 201         case 'I':         \
 202         case 'J':         \
 203         case 'K':         \
 204         case 'L':         \
 205         case 'M':         \
 206         case 'N':         \
 207         case 'O':         \
 208         case 'P':         \
 209         case 'Q':         \
 210         case 'R':         \
 211         case 'S':         \
 212         case 'T':         \
 213         case 'U':         \
 214         case 'V':         \
 215         case 'W':         \
 216         case 'X':         \
 217         case 'Y':         \
 218         case 'Z':         \
 219         case '_':
 220
 221 #define DIGITS        \
 222         case '0':         \
 223         case '1':         \
 224         case '2':         \
 225         case '3':         \
 226         case '4':         \
 227         case '5':         \
 228         case '6':         \
 229         case '7':         \
 230         case '8':         \
 231         case '9':
 232
 233 static void parse_symbol(void)
 234 {
 235         symbol_t *symbol;
 236         char     *string;
 237
 238         obstack_1grow(&symbol_obstack, (char) c);
 239         next_char();
 240
 241         while(1) {
 242                 switch(c) {
 243                 DIGITS
 244                 SYMBOL_CHARS
 245                         obstack_1grow(&symbol_obstack, (char) c);
 246                         next_char();
 247                         break;
 248
 249                 default:
 250                         goto end_symbol;
 251                 }
 252         }
 253
 254 end_symbol:
 255         obstack_1grow(&symbol_obstack, '\0');
 256
 257         string = obstack_finish(&symbol_obstack);
 258         symbol = symbol_table_insert(string);
 259
 260         lexer_token.type     = symbol->ID;
 261         lexer_token.v.symbol = symbol;
 262
 263         if(symbol->string != string) {
 264                 obstack_free(&symbol_obstack, string);
 265         }
 266 }
 267
 268 static void parse_integer_suffix(bool is_oct_hex)
 269 {
 270         bool is_unsigned  = false;
 271         bool min_long     = false;
 272         bool min_longlong = false;
 273
 274         if(c == 'U' || c == 'u') {
 275                 is_unsigned = true;
 276                 next_char();
 277                 if(c == 'L' || c == 'l') {
 278                         min_long = true;
 279                         next_char();
 280                         if(c == 'L' || c == 'l') {
 281                                 min_longlong = true;
 282                                 next_char();
 283                         }
 284                 }
 285         } else if(c == 'l' || c == 'L') {
 286                 min_long = true;
 287                 next_char();
 288                 if(c == 'l' || c == 'L') {
 289                         min_longlong = true;
 290                         next_char();
 291                         if(c == 'u' || c == 'U') {
 292                                 is_unsigned = true;
 293                                 next_char();
 294                         }
 295                 } else if(c == 'u' || c == 'U') {
 296                         is_unsigned = true;
 297                         next_char();
 298                         lexer_token.datatype = type_unsigned_long;
 299                 }
 300         }
 301
 302         if(!is_unsigned) {
 303                 long long v = lexer_token.v.intvalue;
 304                 if(!min_long) {
 305                         if(v >= TARGET_INT_MIN && v <= TARGET_INT_MAX) {
 306                                 lexer_token.datatype = type_int;
 307                                 return;
 308                         } else if(is_oct_hex && v >= 0 && v <= TARGET_UINT_MAX) {
 309                                 lexer_token.datatype = type_unsigned_int;
 310                                 return;
 311                         }
 312                 }
 313                 if(!min_longlong) {
 314                         if(v >= TARGET_LONG_MIN && v <= TARGET_LONG_MAX) {
 315                                 lexer_token.datatype = type_long;
 316                                 return;
 317                         } else if(is_oct_hex && v >= 0 && v <= TARGET_ULONG_MAX) {
 318                                 lexer_token.datatype = type_unsigned_long;
 319                                 return;
 320                         }
 321                 }
 322                 unsigned long long uv = (unsigned long long) v;
 323                 if(is_oct_hex && uv > (unsigned long long) TARGET_LONGLONG_MAX) {
 324                         lexer_token.datatype = type_unsigned_long_long;
 325                         return;
 326                 }
 327
 328                 lexer_token.datatype = type_long_long;
 329         } else {
 330                 unsigned long long v = (unsigned long long) lexer_token.v.intvalue;
 331                 if(!min_long && v <= TARGET_UINT_MAX) {
 332                         lexer_token.datatype = type_unsigned_int;
 333                         return;
 334                 }
 335                 if(!min_longlong && v <= TARGET_ULONG_MAX) {
 336                         lexer_token.datatype = type_unsigned_long;
 337                         return;
 338                 }
 339                 lexer_token.datatype = type_unsigned_long_long;
 340         }
 341 }
 342
 343 static void parse_floating_suffix(void)
 344 {
 345         switch(c) {
 346         /* TODO: do something usefull with the suffixes... */
 347         case 'f':
 348         case 'F':
 349                 next_char();
 350                 lexer_token.datatype = type_float;
 351                 break;
 352         case 'l':
 353         case 'L':
 354                 next_char();
 355                 lexer_token.datatype = type_long_double;
 356                 break;
 357         default:
 358                 lexer_token.datatype = type_double;
 359                 break;
 360         }
 361 }
 362
 363 /**
 364  * A replacement for strtoull. Only those parts needed for
 365  * our parser are implemented.
 366  */
 367 static unsigned long long parse_int_string(const char *s, const char **endptr, int base) {
 368         unsigned long long v = 0;
 369
 370         switch (base) {
 371         case 16:
 372                 for (;; ++s) {
 373                         /* check for overrun */
 374                         if (v >= 0x1000000000000000ULL)
 375                                 break;
 376                         switch (tolower(*s)) {
 377                         case '0': v <<= 4; break;
 378                         case '1': v <<= 4; v |= 0x1; break;
 379                         case '2': v <<= 4; v |= 0x2; break;
 380                         case '3': v <<= 4; v |= 0x3; break;
 381                         case '4': v <<= 4; v |= 0x4; break;
 382                         case '5': v <<= 4; v |= 0x5; break;
 383                         case '6': v <<= 4; v |= 0x6; break;
 384                         case '7': v <<= 4; v |= 0x7; break;
 385                         case '8': v <<= 4; v |= 0x8; break;
 386                         case '9': v <<= 4; v |= 0x9; break;
 387                         case 'a': v <<= 4; v |= 0xa; break;
 388                         case 'b': v <<= 4; v |= 0xb; break;
 389                         case 'c': v <<= 4; v |= 0xc; break;
 390                         case 'd': v <<= 4; v |= 0xd; break;
 391                         case 'e': v <<= 4; v |= 0xe; break;
 392                         case 'f': v <<= 4; v |= 0xf; break;
 393                         default:
 394                                 goto end;
 395                         }
 396                 }
 397                 break;
 398         case 8:
 399                 for (;; ++s) {
 400                         /* check for overrun */
 401                         if (v >= 0x2000000000000000ULL)
 402                                 break;
 403                         switch (tolower(*s)) {
 404                         case '0': v <<= 3; break;
 405                         case '1': v <<= 3; v |= 1; break;
 406                         case '2': v <<= 3; v |= 2; break;
 407                         case '3': v <<= 3; v |= 3; break;
 408                         case '4': v <<= 3; v |= 4; break;
 409                         case '5': v <<= 3; v |= 5; break;
 410                         case '6': v <<= 3; v |= 6; break;
 411                         case '7': v <<= 3; v |= 7; break;
 412                         default:
 413                                 goto end;
 414                         }
 415                 }
 416                 break;
 417         case 10:
 418                 for (;; ++s) {
 419                         /* check for overrun */
 420                         if (v > 0x1999999999999999ULL)
 421                                 break;
 422                         switch (tolower(*s)) {
 423                         case '0': v *= 10; break;
 424                         case '1': v *= 10; v += 1; break;
 425                         case '2': v *= 10; v += 2; break;
 426                         case '3': v *= 10; v += 3; break;
 427                         case '4': v *= 10; v += 4; break;
 428                         case '5': v *= 10; v += 5; break;
 429                         case '6': v *= 10; v += 6; break;
 430                         case '7': v *= 10; v += 7; break;
 431                         case '8': v *= 10; v += 8; break;
 432                         case '9': v *= 10; v += 9; break;
 433                         default:
 434                                 goto end;
 435                         }
 436                 }
 437                 break;
 438         default:
 439                 assert(0);
 440                 break;
 441         }
 442 end:
 443         *endptr = s;
 444         return v;
 445 }
 446
 447 static void parse_number_hex(void)
 448 {
 449         assert(c == 'x' || c == 'X');
 450         next_char();
 451
 452         while(isxdigit(c)) {
 453                 obstack_1grow(&symbol_obstack, (char) c);
 454                 next_char();
 455         }
 456         obstack_1grow(&symbol_obstack, '\0');
 457         char *string = obstack_finish(&symbol_obstack);
 458
 459         if(c == '.' || c == 'p' || c == 'P') {
 460                 next_char();
 461                 panic("Hex floating point numbers not implemented yet");
 462         }
 463         if(*string == '\0') {
 464                 parse_error("invalid hex number");
 465                 lexer_token.type = T_ERROR;
 466         }
 467
 468         const char *endptr;
 469         lexer_token.type       = T_INTEGER;
 470         lexer_token.v.intvalue = parse_int_string(string, &endptr, 16);
 471         if(*endptr != '\0') {
 472                 parse_error("hex number literal too long");
 473         }
 474
 475         obstack_free(&symbol_obstack, string);
 476         parse_integer_suffix(true);
 477 }
 478
 479 static inline bool is_octal_digit(int chr)
 480 {
 481         return '0' <= chr && chr <= '7';
 482 }
 483
 484 static void parse_number_oct(void)
 485 {
 486         while(is_octal_digit(c)) {
 487                 obstack_1grow(&symbol_obstack, (char) c);
 488                 next_char();
 489         }
 490         obstack_1grow(&symbol_obstack, '\0');
 491         char *string = obstack_finish(&symbol_obstack);
 492
 493         const char *endptr;
 494         lexer_token.type       = T_INTEGER;
 495         lexer_token.v.intvalue = parse_int_string(string, &endptr, 8);
 496         if(*endptr != '\0') {
 497                 parse_error("octal number literal too long");
 498         }
 499
 500         obstack_free(&symbol_obstack, string);
 501         parse_integer_suffix(true);
 502 }
 503
 504 static void parse_number_dec(void)
 505 {
 506         bool is_float = false;
 507         while(isdigit(c)) {
 508                 obstack_1grow(&symbol_obstack, (char) c);
 509                 next_char();
 510         }
 511
 512         if(c == '.') {
 513                 obstack_1grow(&symbol_obstack, '.');
 514                 next_char();
 515
 516                 while(isdigit(c)) {
 517                         obstack_1grow(&symbol_obstack, (char) c);
 518                         next_char();
 519                 }
 520                 is_float = true;
 521         }
 522         if(c == 'e' || c == 'E') {
 523                 obstack_1grow(&symbol_obstack, 'e');
 524                 next_char();
 525
 526                 if(c == '-' || c == '+') {
 527                         obstack_1grow(&symbol_obstack, (char) c);
 528                         next_char();
 529                 }
 530
 531                 while(isdigit(c)) {
 532                         obstack_1grow(&symbol_obstack, (char) c);
 533                         next_char();
 534                 }
 535                 is_float = true;
 536         }
 537
 538         obstack_1grow(&symbol_obstack, '\0');
 539         char *string = obstack_finish(&symbol_obstack);
 540
 541         if(is_float) {
 542                 char *endptr;
 543                 lexer_token.type         = T_FLOATINGPOINT;
 544                 lexer_token.v.floatvalue = strtold(string, &endptr);
 545
 546                 if(*endptr != '\0') {
 547                         parse_error("invalid number literal");
 548                 }
 549
 550                 parse_floating_suffix();
 551         } else {
 552                 const char *endptr;
 553                 lexer_token.type       = T_INTEGER;
 554                 lexer_token.v.intvalue = parse_int_string(string, &endptr, 10);
 555
 556                 if(*endptr != '\0') {
 557                         parse_error("invalid number literal");
 558                 }
 559
 560                 parse_integer_suffix(false);
 561         }
 562         obstack_free(&symbol_obstack, string);
 563 }
 564
 565 static void parse_number(void)
 566 {
 567         if (c == '0') {
 568                 next_char();
 569                 switch (c) {
 570                         case 'X':
 571                         case 'x':
 572                                 parse_number_hex();
 573                                 break;
 574                         case '0':
 575                         case '1':
 576                         case '2':
 577                         case '3':
 578                         case '4':
 579                         case '5':
 580                         case '6':
 581                         case '7':
 582                                 parse_number_oct();
 583                                 break;
 584                         case '8':
 585                         case '9':
 586                                 next_char();
 587                                 parse_error("invalid octal number");
 588                                 lexer_token.type = T_ERROR;
 589                                 return;
 590                         case '.':
 591                         case 'e':
 592                         case 'E':
 593                         default:
 594                                 obstack_1grow(&symbol_obstack, '0');
 595                                 parse_number_dec();
 596                                 return;
 597                 }
 598         } else {
 599                 parse_number_dec();
 600         }
 601 }
 602
 603 static int parse_octal_sequence(const int first_digit)
 604 {
 605         assert(is_octal_digit(first_digit));
 606         int value = first_digit - '0';
 607         if (!is_octal_digit(c)) return value;
 608         value = 8 * value + c - '0';
 609         next_char();
 610         if (!is_octal_digit(c)) return value;
 611         value = 8 * value + c - '0';
 612         next_char();
 613         return (char_type)value;
 614 }
 615
 616 static int parse_hex_sequence(void)
 617 {
 618         int value = 0;
 619         while(1) {
 620                 if (c >= '0' && c <= '9') {
 621                         value = 16 * value + c - '0';
 622                 } else if ('A' <= c && c <= 'F') {
 623                         value = 16 * value + c - 'A' + 10;
 624                 } else if ('a' <= c && c <= 'f') {
 625                         value = 16 * value + c - 'a' + 10;
 626                 } else {
 627                         break;
 628                 }
 629                 next_char();
 630         }
 631
 632         return (char_type)value;
 633 }
 634
 635 static int parse_escape_sequence(void)
 636 {
 637         eat('\\');
 638
 639         int ec = c;
 640         next_char();
 641
 642         switch(ec) {
 643         case '"':  return '"';
 644         case '\'': return '\'';
 645         case '\\': return '\\';
 646         case '?': return '\?';
 647         case 'a': return '\a';
 648         case 'b': return '\b';
 649         case 'f': return '\f';
 650         case 'n': return '\n';
 651         case 'r': return '\r';
 652         case 't': return '\t';
 653         case 'v': return '\v';
 654         case 'x':
 655                 return parse_hex_sequence();
 656         case '0':
 657         case '1':
 658         case '2':
 659         case '3':
 660         case '4':
 661         case '5':
 662         case '6':
 663         case '7':
 664                 return parse_octal_sequence(ec);
 665         case EOF:
 666                 parse_error("reached end of file while parsing escape sequence");
 667                 return EOF;
 668         default:
 669                 parse_error("unknown escape sequence");
 670                 return EOF;
 671         }
 672 }
 673
 674 string_t concat_strings(const string_t *const s1, const string_t *const s2)
 675 {
 676         const size_t len1 = s1->size - 1;
 677         const size_t len2 = s2->size - 1;
 678
 679         char *const concat = obstack_alloc(&symbol_obstack, len1 + len2 + 1);
 680         memcpy(concat, s1->begin, len1);
 681         memcpy(concat + len1, s2->begin, len2 + 1);
 682
 683 #if 0 /* TODO hash */
 684         const char *result = strset_insert(&stringset, concat);
 685         if(result != concat) {
 686                 obstack_free(&symbol_obstack, concat);
 687         }
 688
 689         return result;
 690 #else
 691         return (string_t){ concat, len1 + len2 + 1 };
 692 #endif
 693 }
 694
 695 static void parse_string_literal(void)
 696 {
 697         const unsigned start_linenr = lexer_token.source_position.linenr;
 698
 699         assert(c == '"');
 700         next_char();
 701
 702         int tc;
 703         while(1) {
 704                 switch(c) {
 705                 case '\\':
 706                         tc = parse_escape_sequence();
 707                         obstack_1grow(&symbol_obstack, (char) tc);
 708                         break;
 709
 710                 case EOF:
 711                         error_prefix_at(lexer_token.source_position.input_name,
 712                                         start_linenr);
 713                         fprintf(stderr, "string has no end\n");
 714                         lexer_token.type = T_ERROR;
 715                         return;
 716
 717                 case '"':
 718                         next_char();
 719                         goto end_of_string;
 720
 721                 default:
 722                         obstack_1grow(&symbol_obstack, (char) c);
 723                         next_char();
 724                         break;
 725                 }
 726         }
 727
 728 end_of_string:
 729
 730         /* TODO: concatenate multiple strings separated by whitespace... */
 731
 732         /* add finishing 0 to the string */
 733         obstack_1grow(&symbol_obstack, '\0');
 734         const size_t      size   = (size_t)obstack_object_size(&symbol_obstack);
 735         const char *const string = obstack_finish(&symbol_obstack);
 736
 737 #if 0 /* TODO hash */
 738         /* check if there is already a copy of the string */
 739         result = strset_insert(&stringset, string);
 740         if(result != string) {
 741                 obstack_free(&symbol_obstack, string);
 742         }
 743 #else
 744         const char *const result = string;
 745 #endif
 746
 747         lexer_token.type           = T_STRING_LITERAL;
 748         lexer_token.v.string.begin = result;
 749         lexer_token.v.string.size  = size;
 750 }
 751
 752 static void parse_wide_character_constant(void)
 753 {
 754         eat('\'');
 755
 756         int found_char = 0;
 757         while(1) {
 758                 switch(c) {
 759                 case '\\':
 760                         found_char = parse_escape_sequence();
 761                         break;
 762
 763                 MATCH_NEWLINE(
 764                         parse_error("newline while parsing character constant");
 765                         break;
 766                 )
 767
 768                 case '\'':
 769                         next_char();
 770                         goto end_of_wide_char_constant;
 771
 772                 case EOF:
 773                         parse_error("EOF while parsing character constant");
 774                         lexer_token.type = T_ERROR;
 775                         return;
 776
 777                 default:
 778                         if(found_char != 0) {
 779                                 parse_error("more than 1 characters in character "
 780                                             "constant");
 781                                 goto end_of_wide_char_constant;
 782                         } else {
 783                                 found_char = c;
 784                                 next_char();
 785                         }
 786                         break;
 787                 }
 788         }
 789
 790 end_of_wide_char_constant:
 791         lexer_token.type       = T_INTEGER;
 792         lexer_token.v.intvalue = found_char;
 793         lexer_token.datatype   = type_wchar_t;
 794 }
 795
 796 static void parse_wide_string_literal(void)
 797 {
 798         const unsigned start_linenr = lexer_token.source_position.linenr;
 799
 800         assert(c == '"');
 801         next_char();
 802
 803         while(1) {
 804                 switch(c) {
 805                         case '\\': {
 806                                 wchar_rep_t tc = parse_escape_sequence();
 807                                 obstack_grow(&symbol_obstack, &tc, sizeof(tc));
 808                                 break;
 809                         }
 810
 811                         case EOF:
 812                                 error_prefix_at(lexer_token.source_position.input_name,
 813                                                 start_linenr);
 814                                 fprintf(stderr, "string has no end\n");
 815                                 lexer_token.type = T_ERROR;
 816                                 return;
 817
 818                         case '"':
 819                                 next_char();
 820                                 goto end_of_string;
 821
 822                         default: {
 823                                 wchar_rep_t tc = c;
 824                                 obstack_grow(&symbol_obstack, &tc, sizeof(tc));
 825                                 next_char();
 826                                 break;
 827                         }
 828                 }
 829         }
 830
 831 end_of_string:;
 832
 833         /* TODO: concatenate multiple strings separated by whitespace... */
 834
 835         /* add finishing 0 to the string */
 836         wchar_rep_t nul = L'\0';
 837         obstack_grow(&symbol_obstack, &nul, sizeof(nul));
 838         const size_t             size   = (size_t)obstack_object_size(&symbol_obstack) / sizeof(wchar_rep_t);
 839         const wchar_rep_t *const string = obstack_finish(&symbol_obstack);
 840
 841 #if 0 /* TODO hash */
 842         /* check if there is already a copy of the string */
 843         const wchar_rep_t *const result = strset_insert(&stringset, string);
 844         if(result != string) {
 845                 obstack_free(&symbol_obstack, string);
 846         }
 847 #else
 848         const wchar_rep_t *const result = string;
 849 #endif
 850
 851         lexer_token.type                = T_WIDE_STRING_LITERAL;
 852         lexer_token.v.wide_string.begin = result;
 853         lexer_token.v.wide_string.size  = size;
 854 }
 855
 856 static void parse_character_constant(void)
 857 {
 858         eat('\'');
 859
 860         int found_char = 0;
 861         while(1) {
 862                 switch(c) {
 863                 case '\\':
 864                         found_char = parse_escape_sequence();
 865                         break;
 866
 867                 MATCH_NEWLINE(
 868                         parse_error("newline while parsing character constant");
 869                         break;
 870                 )
 871
 872                 case '\'':
 873                         next_char();
 874                         goto end_of_char_constant;
 875
 876                 case EOF:
 877                         parse_error("EOF while parsing character constant");
 878                         lexer_token.type = T_ERROR;
 879                         return;
 880
 881                 default:
 882                         if(found_char != 0) {
 883                                 parse_error("more than 1 characters in character "
 884                                             "constant");
 885                                 goto end_of_char_constant;
 886                         } else {
 887                                 found_char = c;
 888                                 next_char();
 889                         }
 890                         break;
 891                 }
 892         }
 893
 894 end_of_char_constant:
 895         lexer_token.type       = T_INTEGER;
 896         lexer_token.v.intvalue = found_char;
 897         lexer_token.datatype   = type_int;
 898 }
 899
 900 static void skip_multiline_comment(void)
 901 {
 902         unsigned start_linenr = lexer_token.source_position.linenr;
 903
 904         while(1) {
 905                 switch(c) {
 906                 case '*':
 907                         next_char();
 908                         if(c == '/') {
 909                                 next_char();
 910                                 return;
 911                         }
 912                         break;
 913
 914                 MATCH_NEWLINE(break;)
 915
 916                 case EOF:
 917                         error_prefix_at(lexer_token.source_position.input_name,
 918                                         start_linenr);
 919                         fprintf(stderr, "at end of file while looking for comment end\n");
 920                         return;
 921
 922                 default:
 923                         next_char();
 924                         break;
 925                 }
 926         }
 927 }
 928
 929 static void skip_line_comment(void)
 930 {
 931         while(1) {
 932                 switch(c) {
 933                 case EOF:
 934                         return;
 935
 936                 case '\n':
 937                 case '\r':
 938                         return;
 939
 940                 default:
 941                         next_char();
 942                         break;
 943                 }
 944         }
 945 }
 946
 947 static token_t pp_token;
 948
 949 static inline void next_pp_token(void)
 950 {
 951         lexer_next_preprocessing_token();
 952         pp_token = lexer_token;
 953 }
 954
 955 static void eat_until_newline(void)
 956 {
 957         while(pp_token.type != '\n' && pp_token.type != T_EOF) {
 958                 next_pp_token();
 959         }
 960 }
 961
 962 static void error_directive(void)
 963 {
 964         error_prefix();
 965         fprintf(stderr, "#error directive: \n");
 966
 967         /* parse pp-tokens until new-line */
 968 }
 969
 970 static void define_directive(void)
 971 {
 972         lexer_next_preprocessing_token();
 973         if(lexer_token.type != T_IDENTIFIER) {
 974                 parse_error("expected identifier after #define\n");
 975                 eat_until_newline();
 976         }
 977 }
 978
 979 static void ifdef_directive(int is_ifndef)
 980 {
 981         (void) is_ifndef;
 982         lexer_next_preprocessing_token();
 983         //expect_identifier();
 984         //extect_newline();
 985 }
 986
 987 static void endif_directive(void)
 988 {
 989         //expect_newline();
 990 }
 991
 992 static void parse_line_directive(void)
 993 {
 994         if(pp_token.type != T_INTEGER) {
 995                 parse_error("expected integer");
 996         } else {
 997                 lexer_token.source_position.linenr = (unsigned int)(pp_token.v.intvalue - 1);
 998                 next_pp_token();
 999         }
1000         if(pp_token.type == T_STRING_LITERAL) {
1001                 lexer_token.source_position.input_name = pp_token.v.string.begin;
1002                 next_pp_token();
1003         }
1004
1005         eat_until_newline();
1006 }
1007
1008 static void parse_preprocessor_identifier(void)
1009 {
1010         assert(pp_token.type == T_IDENTIFIER);
1011         symbol_t *symbol = pp_token.v.symbol;
1012
1013         switch(symbol->pp_ID) {
1014         case TP_include:
1015                 printf("include - enable header name parsing!\n");
1016                 break;
1017         case TP_define:
1018                 define_directive();
1019                 break;
1020         case TP_ifdef:
1021                 ifdef_directive(0);
1022                 break;
1023         case TP_ifndef:
1024                 ifdef_directive(1);
1025                 break;
1026         case TP_endif:
1027                 endif_directive();
1028                 break;
1029         case TP_line:
1030                 next_pp_token();
1031                 parse_line_directive();
1032                 break;
1033         case TP_if:
1034         case TP_else:
1035         case TP_elif:
1036         case TP_undef:
1037         case TP_error:
1038                 error_directive();
1039                 break;
1040         case TP_pragma:
1041                 if (warning.unknown_pragmas) {
1042                         warningf(lexer_token.source_position, "encountered unknown #pragma");
1043                 }
1044                 eat_until_newline();
1045                 break;
1046         }
1047 }
1048
1049 static void parse_preprocessor_directive(void)
1050 {
1051         next_pp_token();
1052
1053         switch(pp_token.type) {
1054         case T_IDENTIFIER:
1055                 parse_preprocessor_identifier();
1056                 break;
1057         case T_INTEGER:
1058                 parse_line_directive();
1059                 break;
1060         default:
1061                 parse_error("invalid preprocessor directive");
1062                 eat_until_newline();
1063                 break;
1064         }
1065 }
1066
1067 #define MAYBE_PROLOG                                       \
1068                         next_char();                                   \
1069                         while(1) {                                     \
1070                                 switch(c) {
1071
1072 #define MAYBE(ch, set_type)                                \
1073                                 case ch:                                   \
1074                                         next_char();                           \
1075                                         lexer_token.type = set_type;           \
1076                                         return;
1077
1078 #define ELSE_CODE(code)                                    \
1079                                 default:                                   \
1080                                         code;                                  \
1081                                 }                                          \
1082                         } /* end of while(1) */                        \
1083                         break;
1084
1085 #define ELSE(set_type)                                     \
1086                 ELSE_CODE(                                         \
1087                         lexer_token.type = set_type;                   \
1088                         return;                                        \
1089                 )
1090
1091 void lexer_next_preprocessing_token(void)
1092 {
1093         while(1) {
1094                 switch(c) {
1095                 case ' ':
1096                 case '\t':
1097                         next_char();
1098                         break;
1099
1100                 MATCH_NEWLINE(
1101                         lexer_token.type = '\n';
1102                         return;
1103                 )
1104
1105                 SYMBOL_CHARS
1106                         parse_symbol();
1107                         /* might be a wide string ( L"string" ) */
1108                         if(lexer_token.type == T_IDENTIFIER &&
1109                             lexer_token.v.symbol == symbol_L) {
1110                             if(c == '"') {
1111                                         parse_wide_string_literal();
1112                                 } else if(c == '\'') {
1113                                         parse_wide_character_constant();
1114                                 }
1115                         }
1116                         return;
1117
1118                 DIGITS
1119                         parse_number();
1120                         return;
1121
1122                 case '"':
1123                         parse_string_literal();
1124                         return;
1125
1126                 case '\'':
1127                         parse_character_constant();
1128                         return;
1129
1130                 case '.':
1131                         MAYBE_PROLOG
1132                                 case '0':
1133                                 case '1':
1134                                 case '2':
1135                                 case '3':
1136                                 case '4':
1137                                 case '5':
1138                                 case '6':
1139                                 case '7':
1140                                 case '8':
1141                                 case '9':
1142                                         put_back(c);
1143                                         c = '.';
1144                                         parse_number_dec();
1145                                         return;
1146
1147                                 case '.':
1148                                         MAYBE_PROLOG
1149                                         MAYBE('.', T_DOTDOTDOT)
1150                                         ELSE_CODE(
1151                                                 put_back(c);
1152                                                 c = '.';
1153                                                 lexer_token.type = '.';
1154                                                 return;
1155                                         )
1156                         ELSE('.')
1157                 case '&':
1158                         MAYBE_PROLOG
1159                         MAYBE('&', T_ANDAND)
1160                         MAYBE('=', T_ANDEQUAL)
1161                         ELSE('&')
1162                 case '*':
1163                         MAYBE_PROLOG
1164                         MAYBE('=', T_ASTERISKEQUAL)
1165                         ELSE('*')
1166                 case '+':
1167                         MAYBE_PROLOG
1168                         MAYBE('+', T_PLUSPLUS)
1169                         MAYBE('=', T_PLUSEQUAL)
1170                         ELSE('+')
1171                 case '-':
1172                         MAYBE_PROLOG
1173                         MAYBE('>', T_MINUSGREATER)
1174                         MAYBE('-', T_MINUSMINUS)
1175                         MAYBE('=', T_MINUSEQUAL)
1176                         ELSE('-')
1177                 case '!':
1178                         MAYBE_PROLOG
1179                         MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1180                         ELSE('!')
1181                 case '/':
1182                         MAYBE_PROLOG
1183                         MAYBE('=', T_SLASHEQUAL)
1184                                 case '*':
1185                                         next_char();
1186                                         skip_multiline_comment();
1187                                         lexer_next_preprocessing_token();
1188                                         return;
1189                                 case '/':
1190                                         next_char();
1191                                         skip_line_comment();
1192                                         lexer_next_preprocessing_token();
1193                                         return;
1194                         ELSE('/')
1195                 case '%':
1196                         MAYBE_PROLOG
1197                         MAYBE('>', T_PERCENTGREATER)
1198                         MAYBE('=', T_PERCENTEQUAL)
1199                                 case ':':
1200                                         MAYBE_PROLOG
1201                                                 case '%':
1202                                                         MAYBE_PROLOG
1203                                                         MAYBE(':', T_PERCENTCOLONPERCENTCOLON)
1204                                                         ELSE_CODE(
1205                                                                 put_back(c);
1206                                                                 c = '%';
1207                                                                 lexer_token.type = T_PERCENTCOLON;
1208                                                                 return;
1209                                                         )
1210                                         ELSE(T_PERCENTCOLON)
1211                         ELSE('%')
1212                 case '<':
1213                         MAYBE_PROLOG
1214                         MAYBE(':', T_LESSCOLON)
1215                         MAYBE('%', T_LESSPERCENT)
1216                         MAYBE('=', T_LESSEQUAL)
1217                                 case '<':
1218                                         MAYBE_PROLOG
1219                                         MAYBE('=', T_LESSLESSEQUAL)
1220                                         ELSE(T_LESSLESS)
1221                         ELSE('<')
1222                 case '>':
1223                         MAYBE_PROLOG
1224                         MAYBE('=', T_GREATEREQUAL)
1225                                 case '>':
1226                                         MAYBE_PROLOG
1227                                         MAYBE('=', T_GREATERGREATEREQUAL)
1228                                         ELSE(T_GREATERGREATER)
1229                         ELSE('>')
1230                 case '^':
1231                         MAYBE_PROLOG
1232                         MAYBE('=', T_CARETEQUAL)
1233                         ELSE('^')
1234                 case '|':
1235                         MAYBE_PROLOG
1236                         MAYBE('=', T_PIPEEQUAL)
1237                         MAYBE('|', T_PIPEPIPE)
1238                         ELSE('|')
1239                 case ':':
1240                         MAYBE_PROLOG
1241                         MAYBE('>', T_COLONGREATER)
1242                         ELSE(':')
1243                 case '=':
1244                         MAYBE_PROLOG
1245                         MAYBE('=', T_EQUALEQUAL)
1246                         ELSE('=')
1247                 case '#':
1248                         MAYBE_PROLOG
1249                         MAYBE('#', T_HASHHASH)
1250                         ELSE('#')
1251
1252                 case '?':
1253                 case '[':
1254                 case ']':
1255                 case '(':
1256                 case ')':
1257                 case '{':
1258                 case '}':
1259                 case '~':
1260                 case ';':
1261                 case ',':
1262                 case '\\':
1263                         lexer_token.type = c;
1264                         next_char();
1265                         return;
1266
1267                 case EOF:
1268                         lexer_token.type = T_EOF;
1269                         return;
1270
1271                 default:
1272                         next_char();
1273                         error_prefix();
1274                         fprintf(stderr, "unknown character '%c' found\n", c);
1275                         lexer_token.type = T_ERROR;
1276                         return;
1277                 }
1278         }
1279 }
1280
1281 void lexer_next_token(void)
1282 {
1283         lexer_next_preprocessing_token();
1284         if(lexer_token.type != '\n')
1285                 return;
1286
1287 newline_found:
1288         do {
1289                 lexer_next_preprocessing_token();
1290         } while(lexer_token.type == '\n');
1291
1292         if(lexer_token.type == '#') {
1293                 parse_preprocessor_directive();
1294                 goto newline_found;
1295         }
1296 }
1297
1298 void init_lexer(void)
1299 {
1300         strset_init(&stringset);
1301 }
1302
1303 void lexer_open_stream(FILE *stream, const char *input_name)
1304 {
1305         input                                  = stream;
1306         lexer_token.source_position.linenr     = 0;
1307         lexer_token.source_position.input_name = input_name;
1308
1309         symbol_L = symbol_table_insert("L");
1310         bufpos = NULL;
1311         bufend = NULL;
1312
1313         /* place a virtual \n at the beginning so the lexer knows that we're
1314          * at the beginning of a line */
1315         c = '\n';
1316 }
1317
1318 void exit_lexer(void)
1319 {
1320         strset_destroy(&stringset);
1321 }
1322
1323 static __attribute__((unused))
1324 void dbg_pos(const source_position_t source_position)
1325 {
1326         fprintf(stdout, "%s:%u\n", source_position.input_name,
1327                 source_position.linenr);
1328         fflush(stdout);
1329 }