nsz Git - cparser/blob - lexer.c

   1 /*
   2  * This file is part of cparser.
   3  * Copyright (C) 2007-2008 Matthias Braun <matze@braunis.de>
   4  *
   5  * This program is free software; you can redistribute it and/or
   6  * modify it under the terms of the GNU General Public License
   7  * as published by the Free Software Foundation; either version 2
   8  * of the License, or (at your option) any later version.
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License
  16  * along with this program; if not, write to the Free Software
  17  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
  18  * 02111-1307, USA.
  19  */
  20 #include <config.h>
  21
  22 #include "diagnostic.h"
  23 #include "lexer.h"
  24 #include "symbol_t.h"
  25 #include "token_t.h"
  26 #include "symbol_table_t.h"
  27 #include "adt/error.h"
  28 #include "adt/strset.h"
  29 #include "adt/util.h"
  30 #include "types.h"
  31 #include "type_t.h"
  32 #include "target_architecture.h"
  33 #include "parser.h"
  34 #include "warning.h"
  35 #include "lang_features.h"
  36
  37 #include <assert.h>
  38 #include <errno.h>
  39 #include <string.h>
  40 #include <stdbool.h>
  41 #include <ctype.h>
  42
  43 //#define DEBUG_CHARS
  44 #define MAX_PUTBACK 3
  45
  46 #ifdef _WIN32
  47 /* No strtold on windows and no replacement yet */
  48 #define strtold(s, e) strtod(s, e)
  49 #endif
  50
  51 static int         c;
  52 token_t            lexer_token;
  53 symbol_t          *symbol_L;
  54 static FILE       *input;
  55 static char        buf[1024 + MAX_PUTBACK];
  56 static const char *bufend;
  57 static const char *bufpos;
  58 static strset_t    stringset;
  59 bool               allow_dollar_in_symbol = true;
  60
  61 /**
  62  * Prints a parse error message at the current token.
  63  *
  64  * @param msg   the error message
  65  */
  66 static void parse_error(const char *msg)
  67 {
  68         errorf(&lexer_token.source_position, "%s", msg);
  69 }
  70
  71 /**
  72  * Prints an internal error message at the current token.
  73  *
  74  * @param msg   the error message
  75  */
  76 static NORETURN internal_error(const char *msg)
  77 {
  78         internal_errorf(&lexer_token.source_position, "%s", msg);
  79 }
  80
  81 static inline void next_real_char(void)
  82 {
  83         assert(bufpos <= bufend);
  84         if (bufpos >= bufend) {
  85                 if (input == NULL) {
  86                         c = EOF;
  87                         return;
  88                 }
  89
  90                 size_t s = fread(buf + MAX_PUTBACK, 1, sizeof(buf) - MAX_PUTBACK,
  91                                  input);
  92                 if(s == 0) {
  93                         c = EOF;
  94                         return;
  95                 }
  96                 bufpos = buf + MAX_PUTBACK;
  97                 bufend = buf + MAX_PUTBACK + s;
  98         }
  99         c = *bufpos++;
 100 }
 101
 102 /**
 103  * Put a character back into the buffer.
 104  *
 105  * @param pc  the character to put back
 106  */
 107 static inline void put_back(int pc)
 108 {
 109         assert(bufpos > buf);
 110         *(--bufpos - buf + buf) = (char) pc;
 111
 112 #ifdef DEBUG_CHARS
 113         printf("putback '%c'\n", pc);
 114 #endif
 115 }
 116
 117 static inline void next_char(void);
 118
 119 #define MATCH_NEWLINE(code)                   \
 120         case '\r':                                \
 121                 next_char();                          \
 122                 if(c == '\n') {                       \
 123                         next_char();                      \
 124                 }                                     \
 125                 lexer_token.source_position.linenr++; \
 126                 code                                  \
 127         case '\n':                                \
 128                 next_char();                          \
 129                 lexer_token.source_position.linenr++; \
 130                 code
 131
 132 #define eat(c_type)  do { assert(c == c_type); next_char(); } while(0)
 133
 134 static void maybe_concat_lines(void)
 135 {
 136         eat('\\');
 137
 138         switch(c) {
 139         MATCH_NEWLINE(return;)
 140
 141         default:
 142                 break;
 143         }
 144
 145         put_back(c);
 146         c = '\\';
 147 }
 148
 149 /**
 150  * Set c to the next input character, ie.
 151  * after expanding trigraphs.
 152  */
 153 static inline void next_char(void)
 154 {
 155         next_real_char();
 156
 157         /* filter trigraphs */
 158         if(UNLIKELY(c == '\\')) {
 159                 maybe_concat_lines();
 160                 goto end_of_next_char;
 161         }
 162
 163         if(LIKELY(c != '?'))
 164                 goto end_of_next_char;
 165
 166         next_real_char();
 167         if(LIKELY(c != '?')) {
 168                 put_back(c);
 169                 c = '?';
 170                 goto end_of_next_char;
 171         }
 172
 173         next_real_char();
 174         switch(c) {
 175         case '=': c = '#'; break;
 176         case '(': c = '['; break;
 177         case '/': c = '\\'; maybe_concat_lines(); break;
 178         case ')': c = ']'; break;
 179         case '\'': c = '^'; break;
 180         case '<': c = '{'; break;
 181         case '!': c = '|'; break;
 182         case '>': c = '}'; break;
 183         case '-': c = '~'; break;
 184         default:
 185                 put_back(c);
 186                 put_back('?');
 187                 c = '?';
 188                 break;
 189         }
 190
 191 end_of_next_char:;
 192 #ifdef DEBUG_CHARS
 193         printf("nchar '%c'\n", c);
 194 #endif
 195 }
 196
 197 #define SYMBOL_CHARS  \
 198         case '$': if (!allow_dollar_in_symbol) goto dollar_sign; \
 199         case 'a':         \
 200         case 'b':         \
 201         case 'c':         \
 202         case 'd':         \
 203         case 'e':         \
 204         case 'f':         \
 205         case 'g':         \
 206         case 'h':         \
 207         case 'i':         \
 208         case 'j':         \
 209         case 'k':         \
 210         case 'l':         \
 211         case 'm':         \
 212         case 'n':         \
 213         case 'o':         \
 214         case 'p':         \
 215         case 'q':         \
 216         case 'r':         \
 217         case 's':         \
 218         case 't':         \
 219         case 'u':         \
 220         case 'v':         \
 221         case 'w':         \
 222         case 'x':         \
 223         case 'y':         \
 224         case 'z':         \
 225         case 'A':         \
 226         case 'B':         \
 227         case 'C':         \
 228         case 'D':         \
 229         case 'E':         \
 230         case 'F':         \
 231         case 'G':         \
 232         case 'H':         \
 233         case 'I':         \
 234         case 'J':         \
 235         case 'K':         \
 236         case 'L':         \
 237         case 'M':         \
 238         case 'N':         \
 239         case 'O':         \
 240         case 'P':         \
 241         case 'Q':         \
 242         case 'R':         \
 243         case 'S':         \
 244         case 'T':         \
 245         case 'U':         \
 246         case 'V':         \
 247         case 'W':         \
 248         case 'X':         \
 249         case 'Y':         \
 250         case 'Z':         \
 251         case '_':
 252
 253 #define DIGITS        \
 254         case '0':         \
 255         case '1':         \
 256         case '2':         \
 257         case '3':         \
 258         case '4':         \
 259         case '5':         \
 260         case '6':         \
 261         case '7':         \
 262         case '8':         \
 263         case '9':
 264
 265 /**
 266  * Read a symbol from the input and build
 267  * the lexer_token.
 268  */
 269 static void parse_symbol(void)
 270 {
 271         symbol_t *symbol;
 272         char     *string;
 273
 274         obstack_1grow(&symbol_obstack, (char) c);
 275         next_char();
 276
 277         while(1) {
 278                 switch(c) {
 279                 DIGITS
 280                 SYMBOL_CHARS
 281                         obstack_1grow(&symbol_obstack, (char) c);
 282                         next_char();
 283                         break;
 284
 285                 default:
 286 dollar_sign:
 287                         goto end_symbol;
 288                 }
 289         }
 290
 291 end_symbol:
 292         obstack_1grow(&symbol_obstack, '\0');
 293
 294         string = obstack_finish(&symbol_obstack);
 295         symbol = symbol_table_insert(string);
 296
 297         lexer_token.type     = symbol->ID;
 298         lexer_token.v.symbol = symbol;
 299
 300         if(symbol->string != string) {
 301                 obstack_free(&symbol_obstack, string);
 302         }
 303 }
 304
 305 static void parse_integer_suffix(bool is_oct_hex)
 306 {
 307         bool is_unsigned     = false;
 308         bool min_long        = false;
 309         bool min_longlong    = false;
 310         bool not_traditional = false;
 311         int  pos             = 0;
 312         char suffix[4];
 313
 314         if (c == 'U' || c == 'u') {
 315                 not_traditional = true;
 316                 suffix[pos++]   = toupper(c);
 317                 is_unsigned     = true;
 318                 next_char();
 319                 if (c == 'L' || c == 'l') {
 320                         suffix[pos++] = toupper(c);
 321                         min_long = true;
 322                         next_char();
 323                         if (c == 'L' || c == 'l') {
 324                                 suffix[pos++] = toupper(c);
 325                                 min_longlong = true;
 326                                 next_char();
 327                         }
 328                 }
 329         } else if (c == 'l' || c == 'L') {
 330                 suffix[pos++] = toupper(c);
 331                 min_long = true;
 332                 next_char();
 333                 if (c == 'l' || c == 'L') {
 334                         not_traditional = true;
 335                         suffix[pos++]   = toupper(c);
 336                         min_longlong    = true;
 337                         next_char();
 338                         if (c == 'u' || c == 'U') {
 339                                 suffix[pos++] = toupper(c);
 340                                 is_unsigned   = true;
 341                                 next_char();
 342                         }
 343                 } else if (c == 'u' || c == 'U') {
 344                         not_traditional = true;
 345                         suffix[pos++]   = toupper(c);
 346                         is_unsigned     = true;
 347                         next_char();
 348                         lexer_token.datatype = type_unsigned_long;
 349                 }
 350         }
 351
 352         if (warning.traditional && not_traditional) {
 353                 suffix[pos] = '\0';
 354                 warningf(&lexer_token.source_position,
 355                         "traditional C rejects the '%s' suffix", suffix);
 356         }
 357         if (!is_unsigned) {
 358                 long long v = lexer_token.v.intvalue;
 359                 if (!min_long) {
 360                         if (v >= TARGET_INT_MIN && v <= TARGET_INT_MAX) {
 361                                 lexer_token.datatype = type_int;
 362                                 return;
 363                         } else if (is_oct_hex && v >= 0 && v <= TARGET_UINT_MAX) {
 364                                 lexer_token.datatype = type_unsigned_int;
 365                                 return;
 366                         }
 367                 }
 368                 if (!min_longlong) {
 369                         if (v >= TARGET_LONG_MIN && v <= TARGET_LONG_MAX) {
 370                                 lexer_token.datatype = type_long;
 371                                 return;
 372                         } else if (is_oct_hex && v >= 0 && (unsigned long long)v <= (unsigned long long)TARGET_ULONG_MAX) {
 373                                 lexer_token.datatype = type_unsigned_long;
 374                                 return;
 375                         }
 376                 }
 377                 unsigned long long uv = (unsigned long long) v;
 378                 if (is_oct_hex && uv > (unsigned long long) TARGET_LONGLONG_MAX) {
 379                         lexer_token.datatype = type_unsigned_long_long;
 380                         return;
 381                 }
 382
 383                 lexer_token.datatype = type_long_long;
 384         } else {
 385                 unsigned long long v = (unsigned long long) lexer_token.v.intvalue;
 386                 if (!min_long && v <= TARGET_UINT_MAX) {
 387                         lexer_token.datatype = type_unsigned_int;
 388                         return;
 389                 }
 390                 if (!min_longlong && v <= TARGET_ULONG_MAX) {
 391                         lexer_token.datatype = type_unsigned_long;
 392                         return;
 393                 }
 394                 lexer_token.datatype = type_unsigned_long_long;
 395         }
 396 }
 397
 398 static void parse_floating_suffix(void)
 399 {
 400         switch(c) {
 401         /* TODO: do something useful with the suffixes... */
 402         case 'f':
 403         case 'F':
 404                 if (warning.traditional) {
 405                         warningf(&lexer_token.source_position,
 406                                 "traditional C rejects the 'F' suffix");
 407                 }
 408                 next_char();
 409                 lexer_token.datatype = type_float;
 410                 break;
 411         case 'l':
 412         case 'L':
 413                 if (warning.traditional) {
 414                         warningf(&lexer_token.source_position,
 415                                 "traditional C rejects the 'F' suffix");
 416                 }
 417                 next_char();
 418                 lexer_token.datatype = type_long_double;
 419                 break;
 420         default:
 421                 lexer_token.datatype = type_double;
 422                 break;
 423         }
 424 }
 425
 426 /**
 427  * A replacement for strtoull. Only those parts needed for
 428  * our parser are implemented.
 429  */
 430 static unsigned long long parse_int_string(const char *s, const char **endptr, int base) {
 431         unsigned long long v = 0;
 432
 433         switch (base) {
 434         case 16:
 435                 for (;; ++s) {
 436                         /* check for overrun */
 437                         if (v >= 0x1000000000000000ULL)
 438                                 break;
 439                         switch (tolower(*s)) {
 440                         case '0': v <<= 4; break;
 441                         case '1': v <<= 4; v |= 0x1; break;
 442                         case '2': v <<= 4; v |= 0x2; break;
 443                         case '3': v <<= 4; v |= 0x3; break;
 444                         case '4': v <<= 4; v |= 0x4; break;
 445                         case '5': v <<= 4; v |= 0x5; break;
 446                         case '6': v <<= 4; v |= 0x6; break;
 447                         case '7': v <<= 4; v |= 0x7; break;
 448                         case '8': v <<= 4; v |= 0x8; break;
 449                         case '9': v <<= 4; v |= 0x9; break;
 450                         case 'a': v <<= 4; v |= 0xa; break;
 451                         case 'b': v <<= 4; v |= 0xb; break;
 452                         case 'c': v <<= 4; v |= 0xc; break;
 453                         case 'd': v <<= 4; v |= 0xd; break;
 454                         case 'e': v <<= 4; v |= 0xe; break;
 455                         case 'f': v <<= 4; v |= 0xf; break;
 456                         default:
 457                                 goto end;
 458                         }
 459                 }
 460                 break;
 461         case 8:
 462                 for (;; ++s) {
 463                         /* check for overrun */
 464                         if (v >= 0x2000000000000000ULL)
 465                                 break;
 466                         switch (tolower(*s)) {
 467                         case '0': v <<= 3; break;
 468                         case '1': v <<= 3; v |= 1; break;
 469                         case '2': v <<= 3; v |= 2; break;
 470                         case '3': v <<= 3; v |= 3; break;
 471                         case '4': v <<= 3; v |= 4; break;
 472                         case '5': v <<= 3; v |= 5; break;
 473                         case '6': v <<= 3; v |= 6; break;
 474                         case '7': v <<= 3; v |= 7; break;
 475                         default:
 476                                 goto end;
 477                         }
 478                 }
 479                 break;
 480         case 10:
 481                 for (;; ++s) {
 482                         /* check for overrun */
 483                         if (v > 0x1999999999999999ULL)
 484                                 break;
 485                         switch (tolower(*s)) {
 486                         case '0': v *= 10; break;
 487                         case '1': v *= 10; v += 1; break;
 488                         case '2': v *= 10; v += 2; break;
 489                         case '3': v *= 10; v += 3; break;
 490                         case '4': v *= 10; v += 4; break;
 491                         case '5': v *= 10; v += 5; break;
 492                         case '6': v *= 10; v += 6; break;
 493                         case '7': v *= 10; v += 7; break;
 494                         case '8': v *= 10; v += 8; break;
 495                         case '9': v *= 10; v += 9; break;
 496                         default:
 497                                 goto end;
 498                         }
 499                 }
 500                 break;
 501         default:
 502                 assert(0);
 503                 break;
 504         }
 505 end:
 506         *endptr = s;
 507         return v;
 508 }
 509
 510 /**
 511  * Parses a hex number including hex floats and set the
 512  * lexer_token.
 513  */
 514 static void parse_number_hex(void)
 515 {
 516         bool is_float = false;
 517         assert(c == 'x' || c == 'X');
 518         next_char();
 519
 520         obstack_1grow(&symbol_obstack, '0');
 521         obstack_1grow(&symbol_obstack, 'x');
 522
 523         while(isxdigit(c)) {
 524                 obstack_1grow(&symbol_obstack, (char) c);
 525                 next_char();
 526         }
 527
 528         if (c == '.') {
 529                 obstack_1grow(&symbol_obstack, (char) c);
 530                 next_char();
 531
 532                 while (isxdigit(c)) {
 533                         obstack_1grow(&symbol_obstack, (char) c);
 534                         next_char();
 535                 }
 536                 is_float = true;
 537         }
 538         if (c == 'p' || c == 'P') {
 539                 obstack_1grow(&symbol_obstack, (char) c);
 540                 next_char();
 541
 542                 if (c == '-' || c == '+') {
 543                         obstack_1grow(&symbol_obstack, (char) c);
 544                         next_char();
 545                 }
 546
 547                 while (isxdigit(c)) {
 548                         obstack_1grow(&symbol_obstack, (char) c);
 549                         next_char();
 550                 }
 551                 is_float = true;
 552         }
 553
 554         obstack_1grow(&symbol_obstack, '\0');
 555         char *string = obstack_finish(&symbol_obstack);
 556         if(*string == '\0') {
 557                 parse_error("invalid hex number");
 558                 lexer_token.type = T_ERROR;
 559                 obstack_free(&symbol_obstack, string);
 560                 return;
 561         }
 562
 563         if (is_float) {
 564                 char *endptr;
 565                 lexer_token.type         = T_FLOATINGPOINT;
 566                 lexer_token.v.floatvalue = strtold(string, &endptr);
 567
 568                 if(*endptr != '\0') {
 569                         parse_error("invalid hex float literal");
 570                 }
 571
 572                 parse_floating_suffix();
 573         } else {
 574                 const char *endptr;
 575                 lexer_token.type       = T_INTEGER;
 576                 lexer_token.v.intvalue = parse_int_string(string + 2, &endptr, 16);
 577                 if(*endptr != '\0') {
 578                         parse_error("hex number literal too long");
 579                 }
 580                 parse_integer_suffix(true);
 581         }
 582
 583         obstack_free(&symbol_obstack, string);
 584 }
 585
 586 /**
 587  * Returns true if the given char is a octal digit.
 588  *
 589  * @param char  the character to check
 590  */
 591 static inline bool is_octal_digit(int chr)
 592 {
 593         switch(chr) {
 594         case '0':
 595         case '1':
 596         case '2':
 597         case '3':
 598         case '4':
 599         case '5':
 600         case '6':
 601         case '7':
 602                 return true;
 603         default:
 604                 return false;
 605         }
 606 }
 607
 608 /**
 609  * Parses a octal number and set the lexer_token.
 610  */
 611 static void parse_number_oct(void)
 612 {
 613         while(is_octal_digit(c)) {
 614                 obstack_1grow(&symbol_obstack, (char) c);
 615                 next_char();
 616         }
 617         obstack_1grow(&symbol_obstack, '\0');
 618         char *string = obstack_finish(&symbol_obstack);
 619
 620         const char *endptr;
 621         lexer_token.type       = T_INTEGER;
 622         lexer_token.v.intvalue = parse_int_string(string, &endptr, 8);
 623         if(*endptr != '\0') {
 624                 parse_error("octal number literal too long");
 625         }
 626
 627         obstack_free(&symbol_obstack, string);
 628         parse_integer_suffix(true);
 629 }
 630
 631 /**
 632  * Parses a decimal including float number and set the
 633  * lexer_token.
 634  */
 635 static void parse_number_dec(void)
 636 {
 637         bool is_float = false;
 638         while (isdigit(c)) {
 639                 obstack_1grow(&symbol_obstack, (char) c);
 640                 next_char();
 641         }
 642
 643         if (c == '.') {
 644                 obstack_1grow(&symbol_obstack, '.');
 645                 next_char();
 646
 647                 while (isdigit(c)) {
 648                         obstack_1grow(&symbol_obstack, (char) c);
 649                         next_char();
 650                 }
 651                 is_float = true;
 652         }
 653         if(c == 'e' || c == 'E') {
 654                 obstack_1grow(&symbol_obstack, (char) c);
 655                 next_char();
 656
 657                 if(c == '-' || c == '+') {
 658                         obstack_1grow(&symbol_obstack, (char) c);
 659                         next_char();
 660                 }
 661
 662                 while(isdigit(c)) {
 663                         obstack_1grow(&symbol_obstack, (char) c);
 664                         next_char();
 665                 }
 666                 is_float = true;
 667         }
 668
 669         obstack_1grow(&symbol_obstack, '\0');
 670         char *string = obstack_finish(&symbol_obstack);
 671
 672         if(is_float) {
 673                 char *endptr;
 674                 lexer_token.type         = T_FLOATINGPOINT;
 675                 lexer_token.v.floatvalue = strtold(string, &endptr);
 676
 677                 if(*endptr != '\0') {
 678                         parse_error("invalid number literal");
 679                 }
 680
 681                 parse_floating_suffix();
 682         } else {
 683                 const char *endptr;
 684                 lexer_token.type       = T_INTEGER;
 685                 lexer_token.v.intvalue = parse_int_string(string, &endptr, 10);
 686
 687                 if(*endptr != '\0') {
 688                         parse_error("invalid number literal");
 689                 }
 690
 691                 parse_integer_suffix(false);
 692         }
 693         obstack_free(&symbol_obstack, string);
 694 }
 695
 696 /**
 697  * Parses a number and sets the lexer_token.
 698  */
 699 static void parse_number(void)
 700 {
 701         if (c == '0') {
 702                 next_char();
 703                 switch (c) {
 704                         case 'X':
 705                         case 'x':
 706                                 parse_number_hex();
 707                                 break;
 708                         case '0':
 709                         case '1':
 710                         case '2':
 711                         case '3':
 712                         case '4':
 713                         case '5':
 714                         case '6':
 715                         case '7':
 716                                 parse_number_oct();
 717                                 break;
 718                         case '8':
 719                         case '9':
 720                                 next_char();
 721                                 parse_error("invalid octal number");
 722                                 lexer_token.type = T_ERROR;
 723                                 return;
 724                         case '.':
 725                         case 'e':
 726                         case 'E':
 727                         default:
 728                                 obstack_1grow(&symbol_obstack, '0');
 729                                 parse_number_dec();
 730                                 return;
 731                 }
 732         } else {
 733                 parse_number_dec();
 734         }
 735 }
 736
 737 /**
 738  * Returns the value of a digit.
 739  * The only portable way to do it ...
 740  */
 741 static int digit_value(int digit) {
 742         switch (digit) {
 743         case '0': return 0;
 744         case '1': return 1;
 745         case '2': return 2;
 746         case '3': return 3;
 747         case '4': return 4;
 748         case '5': return 5;
 749         case '6': return 6;
 750         case '7': return 7;
 751         case '8': return 8;
 752         case '9': return 9;
 753         case 'a':
 754         case 'A': return 10;
 755         case 'b':
 756         case 'B': return 11;
 757         case 'c':
 758         case 'C': return 12;
 759         case 'd':
 760         case 'D': return 13;
 761         case 'e':
 762         case 'E': return 14;
 763         case 'f':
 764         case 'F': return 15;
 765         default:
 766                 internal_error("wrong character given");
 767         }
 768 }
 769
 770 /**
 771  * Parses an octal character sequence.
 772  *
 773  * @param first_digit  the already read first digit
 774  */
 775 static int parse_octal_sequence(const int first_digit)
 776 {
 777         assert(is_octal_digit(first_digit));
 778         int value = digit_value(first_digit);
 779         if (!is_octal_digit(c)) return value;
 780         value = 8 * value + digit_value(c);
 781         next_char();
 782         if (!is_octal_digit(c)) return value;
 783         value = 8 * value + digit_value(c);
 784         next_char();
 785
 786         if(char_is_signed) {
 787                 return (signed char) value;
 788         } else {
 789                 return (unsigned char) value;
 790         }
 791 }
 792
 793 /**
 794  * Parses a hex character sequence.
 795  */
 796 static int parse_hex_sequence(void)
 797 {
 798         int value = 0;
 799         while(isxdigit(c)) {
 800                 value = 16 * value + digit_value(c);
 801                 next_char();
 802         }
 803
 804         if(char_is_signed) {
 805                 return (signed char) value;
 806         } else {
 807                 return (unsigned char) value;
 808         }
 809 }
 810
 811 /**
 812  * Parse an escape sequence.
 813  */
 814 static int parse_escape_sequence(void)
 815 {
 816         eat('\\');
 817
 818         int ec = c;
 819         next_char();
 820
 821         switch (ec) {
 822         case '"':  return '"';
 823         case '\'': return '\'';
 824         case '\\': return '\\';
 825         case '?': return '\?';
 826         case 'a': return '\a';
 827         case 'b': return '\b';
 828         case 'f': return '\f';
 829         case 'n': return '\n';
 830         case 'r': return '\r';
 831         case 't': return '\t';
 832         case 'v': return '\v';
 833         case 'x':
 834                 return parse_hex_sequence();
 835         case '0':
 836         case '1':
 837         case '2':
 838         case '3':
 839         case '4':
 840         case '5':
 841         case '6':
 842         case '7':
 843                 return parse_octal_sequence(ec);
 844         case EOF:
 845                 parse_error("reached end of file while parsing escape sequence");
 846                 return EOF;
 847         case 'e':
 848                 if (c_mode & _GNUC)
 849                         return 27;   /* hopefully 27 is ALWAYS the code for ESACAPE */
 850                 /*fallthrough*/
 851         default:
 852                 parse_error("unknown escape sequence");
 853                 return EOF;
 854         }
 855 }
 856
 857 /**
 858  * Concatenate two strings.
 859  */
 860 string_t concat_strings(const string_t *const s1, const string_t *const s2)
 861 {
 862         const size_t len1 = s1->size - 1;
 863         const size_t len2 = s2->size - 1;
 864
 865         char *const concat = obstack_alloc(&symbol_obstack, len1 + len2 + 1);
 866         memcpy(concat, s1->begin, len1);
 867         memcpy(concat + len1, s2->begin, len2 + 1);
 868
 869         if (warning.traditional) {
 870                 warningf(&lexer_token.source_position,
 871                         "traditional C rejects string constant concatenation");
 872         }
 873 #if 0 /* TODO hash */
 874         const char *result = strset_insert(&stringset, concat);
 875         if(result != concat) {
 876                 obstack_free(&symbol_obstack, concat);
 877         }
 878
 879         return result;
 880 #else
 881         return (string_t){ concat, len1 + len2 + 1 };
 882 #endif
 883 }
 884
 885 /**
 886  * Concatenate a string and a wide string.
 887  */
 888 wide_string_t concat_string_wide_string(const string_t *const s1, const wide_string_t *const s2)
 889 {
 890         const size_t len1 = s1->size - 1;
 891         const size_t len2 = s2->size - 1;
 892
 893         wchar_rep_t *const concat = obstack_alloc(&symbol_obstack, (len1 + len2 + 1) * sizeof(*concat));
 894         const char *const src = s1->begin;
 895         for (size_t i = 0; i != len1; ++i) {
 896                 concat[i] = src[i];
 897         }
 898         memcpy(concat + len1, s2->begin, (len2 + 1) * sizeof(*concat));
 899         if (warning.traditional) {
 900                 warningf(&lexer_token.source_position,
 901                         "traditional C rejects string constant concatenation");
 902         }
 903
 904         return (wide_string_t){ concat, len1 + len2 + 1 };
 905 }
 906
 907 /**
 908  * Concatenate two wide strings.
 909  */
 910 wide_string_t concat_wide_strings(const wide_string_t *const s1, const wide_string_t *const s2)
 911 {
 912         const size_t len1 = s1->size - 1;
 913         const size_t len2 = s2->size - 1;
 914
 915         wchar_rep_t *const concat = obstack_alloc(&symbol_obstack, (len1 + len2 + 1) * sizeof(*concat));
 916         memcpy(concat,        s1->begin, len1       * sizeof(*concat));
 917         memcpy(concat + len1, s2->begin, (len2 + 1) * sizeof(*concat));
 918         if (warning.traditional) {
 919                 warningf(&lexer_token.source_position,
 920                         "traditional C rejects string constant concatenation");
 921         }
 922
 923         return (wide_string_t){ concat, len1 + len2 + 1 };
 924 }
 925
 926 /**
 927  * Concatenate a wide string and a string.
 928  */
 929 wide_string_t concat_wide_string_string(const wide_string_t *const s1, const string_t *const s2)
 930 {
 931         const size_t len1 = s1->size - 1;
 932         const size_t len2 = s2->size - 1;
 933
 934         wchar_rep_t *const concat = obstack_alloc(&symbol_obstack, (len1 + len2 + 1) * sizeof(*concat));
 935         memcpy(concat, s1->begin, len1 * sizeof(*concat));
 936         const char *const src = s2->begin;
 937         for (size_t i = 0; i != len2 + 1; ++i) {
 938                 concat[i] = src[i];
 939         }
 940         if (warning.traditional) {
 941                 warningf(&lexer_token.source_position,
 942                         "traditional C rejects string constant concatenation");
 943         }
 944
 945         return (wide_string_t){ concat, len1 + len2 + 1 };
 946 }
 947
 948 /**
 949  * Parse a string literal and set lexer_token.
 950  */
 951 static void parse_string_literal(void)
 952 {
 953         const unsigned start_linenr = lexer_token.source_position.linenr;
 954
 955         eat('"');
 956
 957         int tc;
 958         while(1) {
 959                 switch(c) {
 960                 case '\\':
 961                         tc = parse_escape_sequence();
 962                         obstack_1grow(&symbol_obstack, (char) tc);
 963                         break;
 964
 965                 case EOF: {
 966                         source_position_t source_position;
 967                         source_position.input_name = lexer_token.source_position.input_name;
 968                         source_position.linenr     = start_linenr;
 969                         errorf(&source_position, "string has no end");
 970                         lexer_token.type = T_ERROR;
 971                         return;
 972                 }
 973
 974                 case '"':
 975                         next_char();
 976                         goto end_of_string;
 977
 978                 default:
 979                         obstack_1grow(&symbol_obstack, (char) c);
 980                         next_char();
 981                         break;
 982                 }
 983         }
 984
 985 end_of_string:
 986
 987         /* TODO: concatenate multiple strings separated by whitespace... */
 988
 989         /* add finishing 0 to the string */
 990         obstack_1grow(&symbol_obstack, '\0');
 991         const size_t      size   = (size_t)obstack_object_size(&symbol_obstack);
 992         const char *const string = obstack_finish(&symbol_obstack);
 993
 994 #if 0 /* TODO hash */
 995         /* check if there is already a copy of the string */
 996         result = strset_insert(&stringset, string);
 997         if(result != string) {
 998                 obstack_free(&symbol_obstack, string);
 999         }
1000 #else
1001         const char *const result = string;
1002 #endif
1003
1004         lexer_token.type           = T_STRING_LITERAL;
1005         lexer_token.v.string.begin = result;
1006         lexer_token.v.string.size  = size;
1007 }
1008
1009 /**
1010  * Parse a wide character constant and set lexer_token.
1011  */
1012 static void parse_wide_character_constant(void)
1013 {
1014         const unsigned start_linenr = lexer_token.source_position.linenr;
1015
1016         eat('\'');
1017
1018         while(1) {
1019                 switch(c) {
1020                 case '\\': {
1021                         wchar_rep_t tc = parse_escape_sequence();
1022                         obstack_grow(&symbol_obstack, &tc, sizeof(tc));
1023                         break;
1024                 }
1025
1026                 MATCH_NEWLINE(
1027                         parse_error("newline while parsing character constant");
1028                         break;
1029                 )
1030
1031                 case '\'':
1032                         next_char();
1033                         goto end_of_wide_char_constant;
1034
1035                 case EOF: {
1036                         source_position_t source_position = lexer_token.source_position;
1037                         source_position.linenr = start_linenr;
1038                         errorf(&source_position, "EOF while parsing character constant");
1039                         lexer_token.type = T_ERROR;
1040                         return;
1041                 }
1042
1043                 default: {
1044                         wchar_rep_t tc = (wchar_rep_t) c;
1045                         obstack_grow(&symbol_obstack, &tc, sizeof(tc));
1046                         next_char();
1047                         break;
1048                 }
1049                 }
1050         }
1051
1052 end_of_wide_char_constant:;
1053         size_t             size   = (size_t) obstack_object_size(&symbol_obstack);
1054         assert(size % sizeof(wchar_rep_t) == 0);
1055         size /= sizeof(wchar_rep_t);
1056
1057         const wchar_rep_t *string = obstack_finish(&symbol_obstack);
1058
1059         lexer_token.type                = T_WIDE_CHARACTER_CONSTANT;
1060         lexer_token.v.wide_string.begin = string;
1061         lexer_token.v.wide_string.size  = size;
1062         lexer_token.datatype            = type_wchar_t;
1063 }
1064
1065 /**
1066  * Parse a wide string literal and set lexer_token.
1067  */
1068 static void parse_wide_string_literal(void)
1069 {
1070         const unsigned start_linenr = lexer_token.source_position.linenr;
1071
1072         assert(c == '"');
1073         next_char();
1074
1075         while(1) {
1076                 switch(c) {
1077                 case '\\': {
1078                         wchar_rep_t tc = parse_escape_sequence();
1079                         obstack_grow(&symbol_obstack, &tc, sizeof(tc));
1080                         break;
1081                 }
1082
1083                 case EOF: {
1084                         source_position_t source_position;
1085                         source_position.input_name = lexer_token.source_position.input_name;
1086                         source_position.linenr     = start_linenr;
1087                         errorf(&source_position, "string has no end");
1088                         lexer_token.type = T_ERROR;
1089                         return;
1090                 }
1091
1092                 case '"':
1093                         next_char();
1094                         goto end_of_string;
1095
1096                 default: {
1097                         wchar_rep_t tc = c;
1098                         obstack_grow(&symbol_obstack, &tc, sizeof(tc));
1099                         next_char();
1100                         break;
1101                 }
1102                 }
1103         }
1104
1105 end_of_string:;
1106
1107         /* TODO: concatenate multiple strings separated by whitespace... */
1108
1109         /* add finishing 0 to the string */
1110         wchar_rep_t nul = L'\0';
1111         obstack_grow(&symbol_obstack, &nul, sizeof(nul));
1112         const size_t             size   = (size_t)obstack_object_size(&symbol_obstack) / sizeof(wchar_rep_t);
1113         const wchar_rep_t *const string = obstack_finish(&symbol_obstack);
1114
1115 #if 0 /* TODO hash */
1116         /* check if there is already a copy of the string */
1117         const wchar_rep_t *const result = strset_insert(&stringset, string);
1118         if(result != string) {
1119                 obstack_free(&symbol_obstack, string);
1120         }
1121 #else
1122         const wchar_rep_t *const result = string;
1123 #endif
1124
1125         lexer_token.type                = T_WIDE_STRING_LITERAL;
1126         lexer_token.v.wide_string.begin = result;
1127         lexer_token.v.wide_string.size  = size;
1128 }
1129
1130 /**
1131  * Parse a character constant and set lexer_token.
1132  */
1133 static void parse_character_constant(void)
1134 {
1135         const unsigned start_linenr = lexer_token.source_position.linenr;
1136
1137         eat('\'');
1138
1139         while(1) {
1140                 switch(c) {
1141                 case '\\': {
1142                         int tc = parse_escape_sequence();
1143                         obstack_1grow(&symbol_obstack, (char) tc);
1144                         break;
1145                 }
1146
1147                 MATCH_NEWLINE(
1148                         parse_error("newline while parsing character constant");
1149                         break;
1150                 )
1151
1152                 case '\'':
1153                         next_char();
1154                         goto end_of_char_constant;
1155
1156                 case EOF: {
1157                         source_position_t source_position;
1158                         source_position.input_name = lexer_token.source_position.input_name;
1159                         source_position.linenr     = start_linenr;
1160                         errorf(&source_position, "EOF while parsing character constant");
1161                         lexer_token.type = T_ERROR;
1162                         return;
1163                 }
1164
1165                 default:
1166                         obstack_1grow(&symbol_obstack, (char) c);
1167                         next_char();
1168                         break;
1169
1170                 }
1171         }
1172
1173 end_of_char_constant:;
1174         const size_t      size   = (size_t)obstack_object_size(&symbol_obstack);
1175         const char *const string = obstack_finish(&symbol_obstack);
1176
1177         lexer_token.type           = T_CHARACTER_CONSTANT;
1178         lexer_token.v.string.begin = string;
1179         lexer_token.v.string.size  = size;
1180         lexer_token.datatype       = type_int;
1181 }
1182
1183 /**
1184  * Skip a multiline comment.
1185  */
1186 static void skip_multiline_comment(void)
1187 {
1188         unsigned start_linenr = lexer_token.source_position.linenr;
1189
1190         while(1) {
1191                 switch(c) {
1192                 case '/':
1193                         next_char();
1194                         if (c == '*') {
1195                                 /* nested comment, warn here */
1196                                 if (warning.comment) {
1197                                         warningf(&lexer_token.source_position, "'/*' within comment");
1198                                 }
1199                         }
1200                         break;
1201                 case '*':
1202                         next_char();
1203                         if(c == '/') {
1204                                 next_char();
1205                                 return;
1206                         }
1207                         break;
1208
1209                 MATCH_NEWLINE(break;)
1210
1211                 case EOF: {
1212                         source_position_t source_position;
1213                         source_position.input_name = lexer_token.source_position.input_name;
1214                         source_position.linenr     = start_linenr;
1215                         errorf(&source_position, "at end of file while looking for comment end");
1216                         return;
1217                 }
1218
1219                 default:
1220                         next_char();
1221                         break;
1222                 }
1223         }
1224 }
1225
1226 /**
1227  * Skip a single line comment.
1228  */
1229 static void skip_line_comment(void)
1230 {
1231         while(1) {
1232                 switch(c) {
1233                 case EOF:
1234                         return;
1235
1236                 case '\n':
1237                 case '\r':
1238                         return;
1239
1240                 case '\\':
1241                         next_char();
1242                         if (c == '\n' || c == '\r') {
1243                                 if (warning.comment)
1244                                         warningf(&lexer_token.source_position, "multi-line comment");
1245                                 return;
1246                         }
1247                         break;
1248
1249                 default:
1250                         next_char();
1251                         break;
1252                 }
1253         }
1254 }
1255
1256 /** The current preprocessor token. */
1257 static token_t pp_token;
1258
1259 /**
1260  * Read the next preprocessor token.
1261  */
1262 static inline void next_pp_token(void)
1263 {
1264         lexer_next_preprocessing_token();
1265         pp_token = lexer_token;
1266 }
1267
1268 /**
1269  * Eat all preprocessor tokens until newline.
1270  */
1271 static void eat_until_newline(void)
1272 {
1273         while(pp_token.type != '\n' && pp_token.type != T_EOF) {
1274                 next_pp_token();
1275         }
1276 }
1277
1278 /**
1279  * Handle the define directive.
1280  */
1281 static void define_directive(void)
1282 {
1283         lexer_next_preprocessing_token();
1284         if(lexer_token.type != T_IDENTIFIER) {
1285                 parse_error("expected identifier after #define\n");
1286                 eat_until_newline();
1287         }
1288 }
1289
1290 /**
1291  * Handle the ifdef directive.
1292  */
1293 static void ifdef_directive(int is_ifndef)
1294 {
1295         (void) is_ifndef;
1296         lexer_next_preprocessing_token();
1297         //expect_identifier();
1298         //extect_newline();
1299 }
1300
1301 /**
1302  * Handle the endif directive.
1303  */
1304 static void endif_directive(void)
1305 {
1306         //expect_newline();
1307 }
1308
1309 /**
1310  * Parse the line directive.
1311  */
1312 static void parse_line_directive(void)
1313 {
1314         if(pp_token.type != T_INTEGER) {
1315                 parse_error("expected integer");
1316         } else {
1317                 lexer_token.source_position.linenr = (unsigned int)(pp_token.v.intvalue - 1);
1318                 next_pp_token();
1319         }
1320         if(pp_token.type == T_STRING_LITERAL) {
1321                 lexer_token.source_position.input_name = pp_token.v.string.begin;
1322                 next_pp_token();
1323         }
1324
1325         eat_until_newline();
1326 }
1327
1328 /**
1329  * STDC pragmas.
1330  */
1331 typedef enum stdc_pragma_kind_t {
1332         STDC_UNKNOWN,
1333         STDC_FP_CONTRACT,
1334         STDC_FENV_ACCESS,
1335         STDC_CX_LIMITED_RANGE
1336 } stdc_pragma_kind_t;
1337
1338 /**
1339  * STDC pragma values.
1340  */
1341 typedef enum stdc_pragma_value_kind_t {
1342         STDC_VALUE_UNKNOWN,
1343         STDC_VALUE_ON,
1344         STDC_VALUE_OFF,
1345         STDC_VALUE_DEFAULT
1346 } stdc_pragma_value_kind_t;
1347
1348 /**
1349  * Parse a pragma directive.
1350  */
1351 static void parse_pragma(void) {
1352         bool unknown_pragma = true;
1353
1354         next_pp_token();
1355         if (pp_token.v.symbol->pp_ID == TP_STDC) {
1356                 stdc_pragma_kind_t kind = STDC_UNKNOWN;
1357                 /* a STDC pragma */
1358                 if (c_mode & _C99) {
1359                         next_pp_token();
1360
1361                         switch (pp_token.v.symbol->pp_ID) {
1362                         case TP_FP_CONTRACT:
1363                                 kind = STDC_FP_CONTRACT;
1364                                 break;
1365                         case TP_FENV_ACCESS:
1366                                 kind = STDC_FENV_ACCESS;
1367                                 break;
1368                         case TP_CX_LIMITED_RANGE:
1369                                 kind = STDC_CX_LIMITED_RANGE;
1370                                 break;
1371                         default:
1372                                 break;
1373                         }
1374                         if (kind != STDC_UNKNOWN) {
1375                                 stdc_pragma_value_kind_t value = STDC_VALUE_UNKNOWN;
1376                                 next_pp_token();
1377                                 switch (pp_token.v.symbol->pp_ID) {
1378                                 case TP_ON:
1379                                         value = STDC_VALUE_ON;
1380                                         break;
1381                                 case TP_OFF:
1382                                         value = STDC_VALUE_OFF;
1383                                         break;
1384                                 case TP_DEFAULT:
1385                                         value = STDC_VALUE_DEFAULT;
1386                                         break;
1387                                 default:
1388                                         break;
1389                                 }
1390                                 if (value != STDC_VALUE_UNKNOWN) {
1391                                         unknown_pragma = false;
1392                                 } else {
1393                                         errorf(&pp_token.source_position, "bad STDC pragma argument");
1394                                 }
1395                         }
1396                 }
1397         } else {
1398                 unknown_pragma = true;
1399         }
1400         eat_until_newline();
1401         if (unknown_pragma && warning.unknown_pragmas) {
1402                 warningf(&pp_token.source_position, "encountered unknown #pragma");
1403         }
1404 }
1405
1406 /**
1407  * Parse a preprocessor non-null directive.
1408  */
1409 static void parse_preprocessor_identifier(void)
1410 {
1411         assert(pp_token.type == T_IDENTIFIER);
1412         symbol_t *symbol = pp_token.v.symbol;
1413
1414         switch(symbol->pp_ID) {
1415         case TP_include:
1416                 printf("include - enable header name parsing!\n");
1417                 break;
1418         case TP_define:
1419                 define_directive();
1420                 break;
1421         case TP_ifdef:
1422                 ifdef_directive(0);
1423                 break;
1424         case TP_ifndef:
1425                 ifdef_directive(1);
1426                 break;
1427         case TP_endif:
1428                 endif_directive();
1429                 break;
1430         case TP_line:
1431                 next_pp_token();
1432                 parse_line_directive();
1433                 break;
1434         case TP_if:
1435         case TP_else:
1436         case TP_elif:
1437         case TP_undef:
1438         case TP_error:
1439                 /* TODO; output the rest of the line */
1440                 parse_error("#error directive: ");
1441                 break;
1442         case TP_pragma:
1443                 parse_pragma();
1444                 break;
1445         }
1446 }
1447
1448 /**
1449  * Parse a preprocessor directive.
1450  */
1451 static void parse_preprocessor_directive(void)
1452 {
1453         next_pp_token();
1454
1455         switch(pp_token.type) {
1456         case T_IDENTIFIER:
1457                 parse_preprocessor_identifier();
1458                 break;
1459         case T_INTEGER:
1460                 parse_line_directive();
1461                 break;
1462         case '\n':
1463                 /* NULL directive, see § 6.10.7 */
1464                 break;
1465         default:
1466                 parse_error("invalid preprocessor directive");
1467                 eat_until_newline();
1468                 break;
1469         }
1470 }
1471
1472 #define MAYBE_PROLOG                                       \
1473                         next_char();                                   \
1474                         while(1) {                                     \
1475                                 switch(c) {
1476
1477 #define MAYBE(ch, set_type)                                \
1478                                 case ch:                                   \
1479                                         next_char();                           \
1480                                         lexer_token.type = set_type;           \
1481                                         return;
1482
1483 #define ELSE_CODE(code)                                    \
1484                                 default:                                   \
1485                                         code                                   \
1486                                 }                                          \
1487                         } /* end of while(1) */                        \
1488                         break;
1489
1490 #define ELSE(set_type)                                     \
1491                 ELSE_CODE(                                         \
1492                         lexer_token.type = set_type;                   \
1493                         return;                                        \
1494                 )
1495
1496 void lexer_next_preprocessing_token(void)
1497 {
1498         while(1) {
1499                 switch(c) {
1500                 case ' ':
1501                 case '\t':
1502                         next_char();
1503                         break;
1504
1505                 MATCH_NEWLINE(
1506                         lexer_token.type = '\n';
1507                         return;
1508                 )
1509
1510                 SYMBOL_CHARS
1511                         parse_symbol();
1512                         /* might be a wide string ( L"string" ) */
1513                         if(lexer_token.type == T_IDENTIFIER &&
1514                             lexer_token.v.symbol == symbol_L) {
1515                             if(c == '"') {
1516                                         parse_wide_string_literal();
1517                                 } else if(c == '\'') {
1518                                         parse_wide_character_constant();
1519                                 }
1520                         }
1521                         return;
1522
1523                 DIGITS
1524                         parse_number();
1525                         return;
1526
1527                 case '"':
1528                         parse_string_literal();
1529                         return;
1530
1531                 case '\'':
1532                         parse_character_constant();
1533                         return;
1534
1535                 case '.':
1536                         MAYBE_PROLOG
1537                                 DIGITS
1538                                         put_back(c);
1539                                         c = '.';
1540                                         parse_number_dec();
1541                                         return;
1542
1543                                 case '.':
1544                                         MAYBE_PROLOG
1545                                         MAYBE('.', T_DOTDOTDOT)
1546                                         ELSE_CODE(
1547                                                 put_back(c);
1548                                                 c = '.';
1549                                                 lexer_token.type = '.';
1550                                                 return;
1551                                         )
1552                         ELSE('.')
1553                 case '&':
1554                         MAYBE_PROLOG
1555                         MAYBE('&', T_ANDAND)
1556                         MAYBE('=', T_ANDEQUAL)
1557                         ELSE('&')
1558                 case '*':
1559                         MAYBE_PROLOG
1560                         MAYBE('=', T_ASTERISKEQUAL)
1561                         ELSE('*')
1562                 case '+':
1563                         MAYBE_PROLOG
1564                         MAYBE('+', T_PLUSPLUS)
1565                         MAYBE('=', T_PLUSEQUAL)
1566                         ELSE('+')
1567                 case '-':
1568                         MAYBE_PROLOG
1569                         MAYBE('>', T_MINUSGREATER)
1570                         MAYBE('-', T_MINUSMINUS)
1571                         MAYBE('=', T_MINUSEQUAL)
1572                         ELSE('-')
1573                 case '!':
1574                         MAYBE_PROLOG
1575                         MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1576                         ELSE('!')
1577                 case '/':
1578                         MAYBE_PROLOG
1579                         MAYBE('=', T_SLASHEQUAL)
1580                                 case '*':
1581                                         next_char();
1582                                         skip_multiline_comment();
1583                                         lexer_next_preprocessing_token();
1584                                         return;
1585                                 case '/':
1586                                         next_char();
1587                                         skip_line_comment();
1588                                         lexer_next_preprocessing_token();
1589                                         return;
1590                         ELSE('/')
1591                 case '%':
1592                         MAYBE_PROLOG
1593                         MAYBE('>', '}')
1594                         MAYBE('=', T_PERCENTEQUAL)
1595                                 case ':':
1596                                         MAYBE_PROLOG
1597                                                 case '%':
1598                                                         MAYBE_PROLOG
1599                                                         MAYBE(':', T_HASHHASH)
1600                                                         ELSE_CODE(
1601                                                                 put_back(c);
1602                                                                 c = '%';
1603                                                                 lexer_token.type = '#';
1604                                                                 return;
1605                                                         )
1606                                         ELSE('#')
1607                         ELSE('%')
1608                 case '<':
1609                         MAYBE_PROLOG
1610                         MAYBE(':', '[')
1611                         MAYBE('%', '{')
1612                         MAYBE('=', T_LESSEQUAL)
1613                                 case '<':
1614                                         MAYBE_PROLOG
1615                                         MAYBE('=', T_LESSLESSEQUAL)
1616                                         ELSE(T_LESSLESS)
1617                         ELSE('<')
1618                 case '>':
1619                         MAYBE_PROLOG
1620                         MAYBE('=', T_GREATEREQUAL)
1621                                 case '>':
1622                                         MAYBE_PROLOG
1623                                         MAYBE('=', T_GREATERGREATEREQUAL)
1624                                         ELSE(T_GREATERGREATER)
1625                         ELSE('>')
1626                 case '^':
1627                         MAYBE_PROLOG
1628                         MAYBE('=', T_CARETEQUAL)
1629                         ELSE('^')
1630                 case '|':
1631                         MAYBE_PROLOG
1632                         MAYBE('=', T_PIPEEQUAL)
1633                         MAYBE('|', T_PIPEPIPE)
1634                         ELSE('|')
1635                 case ':':
1636                         MAYBE_PROLOG
1637                         MAYBE('>', ']')
1638                         ELSE(':')
1639                 case '=':
1640                         MAYBE_PROLOG
1641                         MAYBE('=', T_EQUALEQUAL)
1642                         ELSE('=')
1643                 case '#':
1644                         MAYBE_PROLOG
1645                         MAYBE('#', T_HASHHASH)
1646                         ELSE('#')
1647
1648                 case '?':
1649                 case '[':
1650                 case ']':
1651                 case '(':
1652                 case ')':
1653                 case '{':
1654                 case '}':
1655                 case '~':
1656                 case ';':
1657                 case ',':
1658                 case '\\':
1659                         lexer_token.type = c;
1660                         next_char();
1661                         return;
1662
1663                 case EOF:
1664                         lexer_token.type = T_EOF;
1665                         return;
1666
1667                 default:
1668 dollar_sign:
1669                         errorf(&lexer_token.source_position, "unknown character '%c' found", c);
1670                         next_char();
1671                         lexer_token.type = T_ERROR;
1672                         return;
1673                 }
1674         }
1675 }
1676
1677 void lexer_next_token(void)
1678 {
1679         lexer_next_preprocessing_token();
1680
1681         while (lexer_token.type == '\n') {
1682 newline_found:
1683                 lexer_next_preprocessing_token();
1684         }
1685
1686         if (lexer_token.type == '#') {
1687                 parse_preprocessor_directive();
1688                 goto newline_found;
1689         }
1690 }
1691
1692 void init_lexer(void)
1693 {
1694         strset_init(&stringset);
1695         symbol_L = symbol_table_insert("L");
1696 }
1697
1698 void lexer_open_stream(FILE *stream, const char *input_name)
1699 {
1700         input                                  = stream;
1701         lexer_token.source_position.linenr     = 0;
1702         lexer_token.source_position.input_name = input_name;
1703
1704         bufpos = NULL;
1705         bufend = NULL;
1706
1707         /* place a virtual \n at the beginning so the lexer knows that we're
1708          * at the beginning of a line */
1709         c = '\n';
1710 }
1711
1712 void lexer_open_buffer(const char *buffer, size_t len, const char *input_name)
1713 {
1714         input                                  = NULL;
1715         lexer_token.source_position.linenr     = 0;
1716         lexer_token.source_position.input_name = input_name;
1717
1718         bufpos = buffer;
1719         bufend = buffer + len;
1720
1721         /* place a virtual \n at the beginning so the lexer knows that we're
1722          * at the beginning of a line */
1723         c = '\n';
1724 }
1725
1726 void exit_lexer(void)
1727 {
1728         strset_destroy(&stringset);
1729 }
1730
1731 static __attribute__((unused))
1732 void dbg_pos(const source_position_t source_position)
1733 {
1734         fprintf(stdout, "%s:%u\n", source_position.input_name,
1735                 source_position.linenr);
1736         fflush(stdout);
1737 }