nsz Git - cparser/blob - lexer.c

   1 /*
   2  * This file is part of cparser.
   3  * Copyright (C) 2007-2009 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 "input.h"
  23 #include "diagnostic.h"
  24 #include "lexer.h"
  25 #include "symbol_t.h"
  26 #include "token_t.h"
  27 #include "symbol_table_t.h"
  28 #include "adt/error.h"
  29 #include "adt/strset.h"
  30 #include "adt/util.h"
  31 #include "types.h"
  32 #include "type_t.h"
  33 #include "target_architecture.h"
  34 #include "parser.h"
  35 #include "warning.h"
  36 #include "lang_features.h"
  37
  38 #include <assert.h>
  39 #include <errno.h>
  40 #include <string.h>
  41 #include <stdbool.h>
  42 #include <ctype.h>
  43
  44 #ifndef _WIN32
  45 #include <strings.h>
  46 #endif
  47
  48 #define MAX_PUTBACK 16    // 3 would be enough, but 16 gives a nicer alignment
  49 #define BUF_SIZE    1024
  50
  51 static input_t           *input;
  52 static utf32              input_buf[BUF_SIZE + MAX_PUTBACK];
  53 static const utf32       *bufpos;
  54 static const utf32       *bufend;
  55 static utf32              c;
  56 static source_position_t  lexer_pos;
  57 token_t                   lexer_token;
  58 static symbol_t          *symbol_L;
  59 static strset_t           stringset;
  60 static char              *encoding;
  61 bool                      allow_dollar_in_symbol = true;
  62
  63 /**
  64  * Prints a parse error message at the current token.
  65  *
  66  * @param msg   the error message
  67  */
  68 static void parse_error(const char *msg)
  69 {
  70         errorf(&lexer_pos, "%s", msg);
  71 }
  72
  73 /**
  74  * Prints an internal error message at the current token.
  75  *
  76  * @param msg   the error message
  77  */
  78 static NORETURN internal_error(const char *msg)
  79 {
  80         internal_errorf(&lexer_pos, "%s", msg);
  81 }
  82
  83 static inline void next_real_char(void)
  84 {
  85         assert(bufpos <= bufend);
  86         if (bufpos >= bufend) {
  87                 size_t n = decode(input, input_buf+MAX_PUTBACK, BUF_SIZE);
  88                 if (n == 0) {
  89                         c = EOF;
  90                         return;
  91                 }
  92                 bufpos = input_buf + MAX_PUTBACK;
  93                 bufend = bufpos + n;
  94         }
  95         c = *bufpos++;
  96         ++lexer_pos.colno;
  97 }
  98
  99 /**
 100  * Put a character back into the buffer.
 101  *
 102  * @param pc  the character to put back
 103  */
 104 static inline void put_back(utf32 const pc)
 105 {
 106         *(--bufpos - input_buf + input_buf) = pc;
 107         --lexer_pos.colno;
 108 }
 109
 110 static inline void next_char(void);
 111
 112 #define MATCH_NEWLINE(code)  \
 113         case '\r':               \
 114                 next_char();         \
 115                 if (c == '\n') {     \
 116         case '\n':               \
 117                         next_char();     \
 118                 }                    \
 119                 lexer_pos.lineno++;  \
 120                 lexer_pos.colno = 1; \
 121                 code
 122
 123 #define eat(c_type) (assert(c == c_type), next_char())
 124
 125 static void maybe_concat_lines(void)
 126 {
 127         eat('\\');
 128
 129         switch (c) {
 130         MATCH_NEWLINE(return;)
 131
 132         default:
 133                 break;
 134         }
 135
 136         put_back(c);
 137         c = '\\';
 138 }
 139
 140 /**
 141  * Set c to the next input character, ie.
 142  * after expanding trigraphs.
 143  */
 144 static inline void next_char(void)
 145 {
 146         next_real_char();
 147
 148         /* filter trigraphs */
 149         if (UNLIKELY(c == '\\')) {
 150                 maybe_concat_lines();
 151                 return;
 152         }
 153
 154         if (LIKELY(c != '?'))
 155                 return;
 156
 157         next_real_char();
 158         if (LIKELY(c != '?')) {
 159                 put_back(c);
 160                 c = '?';
 161                 return;
 162         }
 163
 164         next_real_char();
 165         switch (c) {
 166         case '=': c = '#'; break;
 167         case '(': c = '['; break;
 168         case '/': c = '\\'; maybe_concat_lines(); break;
 169         case ')': c = ']'; break;
 170         case '\'': c = '^'; break;
 171         case '<': c = '{'; break;
 172         case '!': c = '|'; break;
 173         case '>': c = '}'; break;
 174         case '-': c = '~'; break;
 175         default:
 176                 put_back(c);
 177                 put_back('?');
 178                 c = '?';
 179                 break;
 180         }
 181 }
 182
 183 #define SYMBOL_CHARS  \
 184         case '$': if (!allow_dollar_in_symbol) goto dollar_sign; \
 185         case 'a':         \
 186         case 'b':         \
 187         case 'c':         \
 188         case 'd':         \
 189         case 'e':         \
 190         case 'f':         \
 191         case 'g':         \
 192         case 'h':         \
 193         case 'i':         \
 194         case 'j':         \
 195         case 'k':         \
 196         case 'l':         \
 197         case 'm':         \
 198         case 'n':         \
 199         case 'o':         \
 200         case 'p':         \
 201         case 'q':         \
 202         case 'r':         \
 203         case 's':         \
 204         case 't':         \
 205         case 'u':         \
 206         case 'v':         \
 207         case 'w':         \
 208         case 'x':         \
 209         case 'y':         \
 210         case 'z':         \
 211         case 'A':         \
 212         case 'B':         \
 213         case 'C':         \
 214         case 'D':         \
 215         case 'E':         \
 216         case 'F':         \
 217         case 'G':         \
 218         case 'H':         \
 219         case 'I':         \
 220         case 'J':         \
 221         case 'K':         \
 222         case 'L':         \
 223         case 'M':         \
 224         case 'N':         \
 225         case 'O':         \
 226         case 'P':         \
 227         case 'Q':         \
 228         case 'R':         \
 229         case 'S':         \
 230         case 'T':         \
 231         case 'U':         \
 232         case 'V':         \
 233         case 'W':         \
 234         case 'X':         \
 235         case 'Y':         \
 236         case 'Z':         \
 237         case '_':
 238
 239 #define DIGITS        \
 240         case '0':         \
 241         case '1':         \
 242         case '2':         \
 243         case '3':         \
 244         case '4':         \
 245         case '5':         \
 246         case '6':         \
 247         case '7':         \
 248         case '8':         \
 249         case '9':
 250
 251 /**
 252  * Read a symbol from the input and build
 253  * the lexer_token.
 254  */
 255 static void parse_symbol(void)
 256 {
 257         obstack_1grow(&symbol_obstack, (char) c);
 258         next_char();
 259
 260         while (true) {
 261                 switch (c) {
 262                 DIGITS
 263                 SYMBOL_CHARS
 264                         obstack_1grow(&symbol_obstack, (char) c);
 265                         next_char();
 266                         break;
 267
 268                 default:
 269 dollar_sign:
 270                         goto end_symbol;
 271                 }
 272         }
 273
 274 end_symbol:
 275         obstack_1grow(&symbol_obstack, '\0');
 276
 277         char     *string = obstack_finish(&symbol_obstack);
 278         symbol_t *symbol = symbol_table_insert(string);
 279
 280         lexer_token.kind              = symbol->ID;
 281         lexer_token.identifier.symbol = symbol;
 282
 283         if (symbol->string != string) {
 284                 obstack_free(&symbol_obstack, string);
 285         }
 286 }
 287
 288 static string_t identify_string(char *string, size_t len)
 289 {
 290         /* TODO hash */
 291 #if 0
 292         const char *result = strset_insert(&stringset, concat);
 293         if (result != concat) {
 294                 obstack_free(&symbol_obstack, concat);
 295         }
 296 #else
 297         const char *result = string;
 298 #endif
 299         return (string_t) {result, len};
 300 }
 301
 302 /**
 303  * parse suffixes like 'LU' or 'f' after numbers
 304  */
 305 static void parse_number_suffix(void)
 306 {
 307         assert(obstack_object_size(&symbol_obstack) == 0);
 308         while (true) {
 309                 switch (c) {
 310                 SYMBOL_CHARS
 311                         obstack_1grow(&symbol_obstack, (char) c);
 312                         next_char();
 313                         break;
 314                 default:
 315                 dollar_sign:
 316                         goto finish_suffix;
 317                 }
 318         }
 319 finish_suffix:
 320         if (obstack_object_size(&symbol_obstack) == 0) {
 321                 lexer_token.number.suffix.begin = NULL;
 322                 lexer_token.number.suffix.size  = 0;
 323                 return;
 324         }
 325
 326         obstack_1grow(&symbol_obstack, '\0');
 327         size_t    size   = obstack_object_size(&symbol_obstack);
 328         char     *string = obstack_finish(&symbol_obstack);
 329
 330         lexer_token.number.suffix = identify_string(string, size);
 331 }
 332
 333 /**
 334  * Parses a hex number including hex floats and set the
 335  * lexer_token.
 336  */
 337 static void parse_number_hex(void)
 338 {
 339         bool is_float   = false;
 340         bool has_digits = false;
 341
 342         assert(obstack_object_size(&symbol_obstack) == 0);
 343         while (isxdigit(c)) {
 344                 has_digits = true;
 345                 obstack_1grow(&symbol_obstack, (char) c);
 346                 next_char();
 347         }
 348
 349         if (c == '.') {
 350                 is_float = true;
 351                 obstack_1grow(&symbol_obstack, (char) c);
 352                 next_char();
 353
 354                 while (isxdigit(c)) {
 355                         has_digits = true;
 356                         obstack_1grow(&symbol_obstack, (char) c);
 357                         next_char();
 358                 }
 359         }
 360         if (c == 'p' || c == 'P') {
 361                 is_float = true;
 362                 obstack_1grow(&symbol_obstack, (char) c);
 363                 next_char();
 364
 365                 if (c == '-' || c == '+') {
 366                         obstack_1grow(&symbol_obstack, (char) c);
 367                         next_char();
 368                 }
 369
 370                 while (isxdigit(c)) {
 371                         obstack_1grow(&symbol_obstack, (char) c);
 372                         next_char();
 373                 }
 374         } else if (is_float) {
 375                 errorf(&lexer_token.base.source_position,
 376                        "hexadecimal floatingpoint constant requires an exponent");
 377         }
 378         obstack_1grow(&symbol_obstack, '\0');
 379
 380         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 381         char   *string = obstack_finish(&symbol_obstack);
 382         lexer_token.number.number = identify_string(string, size);
 383
 384         lexer_token.kind    =
 385                 is_float ? T_FLOATINGPOINT_HEXADECIMAL : T_INTEGER_HEXADECIMAL;
 386
 387         if (!has_digits) {
 388                 errorf(&lexer_token.base.source_position,
 389                        "invalid number literal '0x%S'", &lexer_token.number.number);
 390                 lexer_token.number.number.begin = "0";
 391                 lexer_token.number.number.size  = 1;
 392         }
 393
 394         parse_number_suffix();
 395 }
 396
 397 /**
 398  * Returns true if the given char is a octal digit.
 399  *
 400  * @param char  the character to check
 401  */
 402 static bool is_octal_digit(utf32 chr)
 403 {
 404         return '0' <= chr && chr <= '7';
 405 }
 406
 407 /**
 408  * Parses a number and sets the lexer_token.
 409  */
 410 static void parse_number(void)
 411 {
 412         bool is_float   = false;
 413         bool has_digits = false;
 414
 415         assert(obstack_object_size(&symbol_obstack) == 0);
 416         if (c == '0') {
 417                 next_char();
 418                 if (c == 'x' || c == 'X') {
 419                         next_char();
 420                         parse_number_hex();
 421                         return;
 422                 } else {
 423                         has_digits = true;
 424                 }
 425                 obstack_1grow(&symbol_obstack, '0');
 426         }
 427
 428         while (isdigit(c)) {
 429                 has_digits = true;
 430                 obstack_1grow(&symbol_obstack, (char) c);
 431                 next_char();
 432         }
 433
 434         if (c == '.') {
 435                 is_float = true;
 436                 obstack_1grow(&symbol_obstack, '.');
 437                 next_char();
 438
 439                 while (isdigit(c)) {
 440                         has_digits = true;
 441                         obstack_1grow(&symbol_obstack, (char) c);
 442                         next_char();
 443                 }
 444         }
 445         if (c == 'e' || c == 'E') {
 446                 is_float = true;
 447                 obstack_1grow(&symbol_obstack, 'e');
 448                 next_char();
 449
 450                 if (c == '-' || c == '+') {
 451                         obstack_1grow(&symbol_obstack, (char) c);
 452                         next_char();
 453                 }
 454
 455                 while (isdigit(c)) {
 456                         obstack_1grow(&symbol_obstack, (char) c);
 457                         next_char();
 458                 }
 459         }
 460
 461         obstack_1grow(&symbol_obstack, '\0');
 462         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 463         char   *string = obstack_finish(&symbol_obstack);
 464         lexer_token.number.number = identify_string(string, size);
 465
 466         /* is it an octal number? */
 467         if (is_float) {
 468                 lexer_token.kind = T_FLOATINGPOINT;
 469         } else if (string[0] == '0') {
 470                 lexer_token.kind = T_INTEGER_OCTAL;
 471
 472                 /* check for invalid octal digits */
 473                 for (size_t i= 0; i < size; ++i) {
 474                         char t = string[i];
 475                         if (t >= '8')
 476                                 errorf(&lexer_token.base.source_position,
 477                                        "invalid digit '%c' in octal number", t);
 478                 }
 479         } else {
 480                 lexer_token.kind = T_INTEGER;
 481         }
 482
 483         if (!has_digits) {
 484                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'",
 485                        &lexer_token.number.number);
 486         }
 487
 488         parse_number_suffix();
 489 }
 490
 491 /**
 492  * Returns the value of a digit.
 493  * The only portable way to do it ...
 494  */
 495 static int digit_value(utf32 const digit)
 496 {
 497         switch (digit) {
 498         case '0': return 0;
 499         case '1': return 1;
 500         case '2': return 2;
 501         case '3': return 3;
 502         case '4': return 4;
 503         case '5': return 5;
 504         case '6': return 6;
 505         case '7': return 7;
 506         case '8': return 8;
 507         case '9': return 9;
 508         case 'a':
 509         case 'A': return 10;
 510         case 'b':
 511         case 'B': return 11;
 512         case 'c':
 513         case 'C': return 12;
 514         case 'd':
 515         case 'D': return 13;
 516         case 'e':
 517         case 'E': return 14;
 518         case 'f':
 519         case 'F': return 15;
 520         default:
 521                 internal_error("wrong character given");
 522         }
 523 }
 524
 525 /**
 526  * Parses an octal character sequence.
 527  *
 528  * @param first_digit  the already read first digit
 529  */
 530 static utf32 parse_octal_sequence(utf32 const first_digit)
 531 {
 532         assert(is_octal_digit(first_digit));
 533         utf32 value = digit_value(first_digit);
 534         if (!is_octal_digit(c)) return value;
 535         value = 8 * value + digit_value(c);
 536         next_char();
 537         if (!is_octal_digit(c)) return value;
 538         value = 8 * value + digit_value(c);
 539         next_char();
 540         return value;
 541 }
 542
 543 /**
 544  * Parses a hex character sequence.
 545  */
 546 static utf32 parse_hex_sequence(void)
 547 {
 548         utf32 value = 0;
 549         while (isxdigit(c)) {
 550                 value = 16 * value + digit_value(c);
 551                 next_char();
 552         }
 553         return value;
 554 }
 555
 556 /**
 557  * Parse an escape sequence.
 558  */
 559 static utf32 parse_escape_sequence(void)
 560 {
 561         eat('\\');
 562
 563         utf32 const ec = c;
 564         next_char();
 565
 566         switch (ec) {
 567         case '"':  return '"';
 568         case '\'': return '\'';
 569         case '\\': return '\\';
 570         case '?': return '\?';
 571         case 'a': return '\a';
 572         case 'b': return '\b';
 573         case 'f': return '\f';
 574         case 'n': return '\n';
 575         case 'r': return '\r';
 576         case 't': return '\t';
 577         case 'v': return '\v';
 578         case 'x':
 579                 return parse_hex_sequence();
 580         case '0':
 581         case '1':
 582         case '2':
 583         case '3':
 584         case '4':
 585         case '5':
 586         case '6':
 587         case '7':
 588                 return parse_octal_sequence(ec);
 589         case EOF:
 590                 parse_error("reached end of file while parsing escape sequence");
 591                 return EOF;
 592         /* \E is not documented, but handled, by GCC.  It is acceptable according
 593          * to §6.11.4, whereas \e is not. */
 594         case 'E':
 595         case 'e':
 596                 if (c_mode & _GNUC)
 597                         return 27;   /* hopefully 27 is ALWAYS the code for ESCAPE */
 598                 break;
 599         case 'u':
 600         case 'U':
 601                 parse_error("universal character parsing not implemented yet");
 602                 return EOF;
 603         default:
 604                 break;
 605         }
 606         /* §6.4.4.4:8 footnote 64 */
 607         parse_error("unknown escape sequence");
 608         return EOF;
 609 }
 610
 611 /**
 612  * Concatenate two strings.
 613  */
 614 string_t concat_strings(const string_t *const s1, const string_t *const s2)
 615 {
 616         const size_t len1 = s1->size - 1;
 617         const size_t len2 = s2->size - 1;
 618
 619         char *const concat = obstack_alloc(&symbol_obstack, len1 + len2 + 1);
 620         memcpy(concat, s1->begin, len1);
 621         memcpy(concat + len1, s2->begin, len2 + 1);
 622
 623         return identify_string(concat, len1 + len2 + 1);
 624 }
 625
 626 string_t make_string(const char *string)
 627 {
 628         size_t      len   = strlen(string) + 1;
 629         char *const space = obstack_alloc(&symbol_obstack, len);
 630         memcpy(space, string, len);
 631
 632         return identify_string(space, len);
 633 }
 634
 635 static void grow_symbol(utf32 const tc)
 636 {
 637         struct obstack *const o  = &symbol_obstack;
 638         if (tc < 0x80U) {
 639                 obstack_1grow(o, tc);
 640         } else if (tc < 0x800) {
 641                 obstack_1grow(o, 0xC0 | (tc >> 6));
 642                 obstack_1grow(o, 0x80 | (tc & 0x3F));
 643         } else if (tc < 0x10000) {
 644                 obstack_1grow(o, 0xE0 | ( tc >> 12));
 645                 obstack_1grow(o, 0x80 | ((tc >>  6) & 0x3F));
 646                 obstack_1grow(o, 0x80 | ( tc        & 0x3F));
 647         } else {
 648                 obstack_1grow(o, 0xF0 | ( tc >> 18));
 649                 obstack_1grow(o, 0x80 | ((tc >> 12) & 0x3F));
 650                 obstack_1grow(o, 0x80 | ((tc >>  6) & 0x3F));
 651                 obstack_1grow(o, 0x80 | ( tc        & 0x3F));
 652         }
 653 }
 654
 655 /**
 656  * Parse a string literal and set lexer_token.
 657  */
 658 static void parse_string_literal(void)
 659 {
 660         eat('"');
 661
 662         while (true) {
 663                 switch (c) {
 664                 case '\\': {
 665                         utf32 const tc = parse_escape_sequence();
 666                         if (tc >= 0x100) {
 667                                 warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 668                         }
 669                         obstack_1grow(&symbol_obstack, tc);
 670                         break;
 671                 }
 672
 673                 case EOF: {
 674                         errorf(&lexer_token.base.source_position, "string has no end");
 675                         lexer_token.kind = T_ERROR;
 676                         return;
 677                 }
 678
 679                 case '"':
 680                         next_char();
 681                         goto end_of_string;
 682
 683                 default:
 684                         grow_symbol(c);
 685                         next_char();
 686                         break;
 687                 }
 688         }
 689
 690 end_of_string:
 691
 692         /* TODO: concatenate multiple strings separated by whitespace... */
 693
 694         /* add finishing 0 to the string */
 695         obstack_1grow(&symbol_obstack, '\0');
 696         const size_t  size   = (size_t)obstack_object_size(&symbol_obstack);
 697         char         *string = obstack_finish(&symbol_obstack);
 698
 699         lexer_token.kind          = T_STRING_LITERAL;
 700         lexer_token.string.string = identify_string(string, size);
 701 }
 702
 703 /**
 704  * Parse a wide character constant and set lexer_token.
 705  */
 706 static void parse_wide_character_constant(void)
 707 {
 708         eat('\'');
 709
 710         while (true) {
 711                 switch (c) {
 712                 case '\\': {
 713                         const utf32 tc = parse_escape_sequence();
 714                         grow_symbol(tc);
 715                         break;
 716                 }
 717
 718                 MATCH_NEWLINE(
 719                         parse_error("newline while parsing character constant");
 720                         break;
 721                 )
 722
 723                 case '\'':
 724                         next_char();
 725                         goto end_of_wide_char_constant;
 726
 727                 case EOF: {
 728                         errorf(&lexer_token.base.source_position,
 729                                "EOF while parsing character constant");
 730                         lexer_token.kind = T_ERROR;
 731                         return;
 732                 }
 733
 734                 default:
 735                         grow_symbol(c);
 736                         next_char();
 737                         break;
 738                 }
 739         }
 740
 741 end_of_wide_char_constant:;
 742         obstack_1grow(&symbol_obstack, '\0');
 743         size_t  size   = (size_t) obstack_object_size(&symbol_obstack) - 1;
 744         char   *string = obstack_finish(&symbol_obstack);
 745
 746         lexer_token.kind          = T_WIDE_CHARACTER_CONSTANT;
 747         lexer_token.string.string = identify_string(string, size);
 748
 749         if (size == 0) {
 750                 errorf(&lexer_token.base.source_position, "empty character constant");
 751         }
 752 }
 753
 754 /**
 755  * Parse a wide string literal and set lexer_token.
 756  */
 757 static void parse_wide_string_literal(void)
 758 {
 759         parse_string_literal();
 760         if (lexer_token.kind == T_STRING_LITERAL)
 761                 lexer_token.kind = T_WIDE_STRING_LITERAL;
 762 }
 763
 764 /**
 765  * Parse a character constant and set lexer_token.
 766  */
 767 static void parse_character_constant(void)
 768 {
 769         eat('\'');
 770
 771         while (true) {
 772                 switch (c) {
 773                 case '\\': {
 774                         utf32 const tc = parse_escape_sequence();
 775                         if (tc >= 0x100) {
 776                                 warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 777                         }
 778                         obstack_1grow(&symbol_obstack, tc);
 779                         break;
 780                 }
 781
 782                 MATCH_NEWLINE(
 783                         parse_error("newline while parsing character constant");
 784                         break;
 785                 )
 786
 787                 case '\'':
 788                         next_char();
 789                         goto end_of_char_constant;
 790
 791                 case EOF: {
 792                         errorf(&lexer_token.base.source_position,
 793                                "EOF while parsing character constant");
 794                         lexer_token.kind = T_ERROR;
 795                         return;
 796                 }
 797
 798                 default:
 799                         grow_symbol(c);
 800                         next_char();
 801                         break;
 802
 803                 }
 804         }
 805
 806 end_of_char_constant:;
 807         obstack_1grow(&symbol_obstack, '\0');
 808         const size_t        size   = (size_t)obstack_object_size(&symbol_obstack)-1;
 809         char         *const string = obstack_finish(&symbol_obstack);
 810
 811         lexer_token.kind          = T_CHARACTER_CONSTANT;
 812         lexer_token.string.string = identify_string(string, size);
 813
 814         if (size == 0) {
 815                 errorf(&lexer_token.base.source_position, "empty character constant");
 816         }
 817 }
 818
 819 /**
 820  * Skip a multiline comment.
 821  */
 822 static void skip_multiline_comment(void)
 823 {
 824         while (true) {
 825                 switch (c) {
 826                 case '/':
 827                         next_char();
 828                         if (c == '*') {
 829                                 /* nested comment, warn here */
 830                                 warningf(WARN_COMMENT, &lexer_pos, "'/*' within comment");
 831                         }
 832                         break;
 833                 case '*':
 834                         next_char();
 835                         if (c == '/') {
 836                                 next_char();
 837                                 return;
 838                         }
 839                         break;
 840
 841                 MATCH_NEWLINE(break;)
 842
 843                 case EOF: {
 844                         errorf(&lexer_token.base.source_position,
 845                                "at end of file while looking for comment end");
 846                         return;
 847                 }
 848
 849                 default:
 850                         next_char();
 851                         break;
 852                 }
 853         }
 854 }
 855
 856 /**
 857  * Skip a single line comment.
 858  */
 859 static void skip_line_comment(void)
 860 {
 861         while (true) {
 862                 switch (c) {
 863                 case EOF:
 864                         return;
 865
 866                 case '\n':
 867                 case '\r':
 868                         return;
 869
 870                 case '\\':
 871                         next_char();
 872                         if (c == '\n' || c == '\r') {
 873                                 warningf(WARN_COMMENT, &lexer_pos, "multi-line comment");
 874                                 return;
 875                         }
 876                         break;
 877
 878                 default:
 879                         next_char();
 880                         break;
 881                 }
 882         }
 883 }
 884
 885 /** The current preprocessor token. */
 886 static token_t pp_token;
 887
 888 /**
 889  * Read the next preprocessor token.
 890  */
 891 static inline void next_pp_token(void)
 892 {
 893         lexer_next_preprocessing_token();
 894         pp_token = lexer_token;
 895 }
 896
 897 /**
 898  * Eat all preprocessor tokens until newline.
 899  */
 900 static void eat_until_newline(void)
 901 {
 902         while (pp_token.kind != '\n' && pp_token.kind != T_EOF) {
 903                 next_pp_token();
 904         }
 905 }
 906
 907 /**
 908  * Handle the define directive.
 909  */
 910 static void define_directive(void)
 911 {
 912         lexer_next_preprocessing_token();
 913         if (lexer_token.kind != T_IDENTIFIER) {
 914                 parse_error("expected identifier after #define\n");
 915                 eat_until_newline();
 916         }
 917 }
 918
 919 /**
 920  * Handle the ifdef directive.
 921  */
 922 static void ifdef_directive(int is_ifndef)
 923 {
 924         (void) is_ifndef;
 925         lexer_next_preprocessing_token();
 926         //expect_identifier();
 927         //extect_newline();
 928 }
 929
 930 /**
 931  * Handle the endif directive.
 932  */
 933 static void endif_directive(void)
 934 {
 935         //expect_newline();
 936 }
 937
 938 /**
 939  * Parse the line directive.
 940  */
 941 static void parse_line_directive(void)
 942 {
 943         if (pp_token.kind != T_INTEGER) {
 944                 parse_error("expected integer");
 945         } else {
 946                 /* use offset -1 as this is about the next line */
 947                 lexer_pos.lineno = atoi(pp_token.number.number.begin) - 1;
 948                 next_pp_token();
 949         }
 950         if (pp_token.kind == T_STRING_LITERAL) {
 951                 lexer_pos.input_name = pp_token.string.string.begin;
 952                 next_pp_token();
 953         }
 954
 955         eat_until_newline();
 956 }
 957
 958 /**
 959  * STDC pragmas.
 960  */
 961 typedef enum stdc_pragma_kind_t {
 962         STDC_UNKNOWN,
 963         STDC_FP_CONTRACT,
 964         STDC_FENV_ACCESS,
 965         STDC_CX_LIMITED_RANGE
 966 } stdc_pragma_kind_t;
 967
 968 /**
 969  * STDC pragma values.
 970  */
 971 typedef enum stdc_pragma_value_kind_t {
 972         STDC_VALUE_UNKNOWN,
 973         STDC_VALUE_ON,
 974         STDC_VALUE_OFF,
 975         STDC_VALUE_DEFAULT
 976 } stdc_pragma_value_kind_t;
 977
 978 /**
 979  * Parse a pragma directive.
 980  */
 981 static void parse_pragma(void)
 982 {
 983         bool unknown_pragma = true;
 984
 985         next_pp_token();
 986         if (pp_token.kind != T_IDENTIFIER) {
 987                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
 988                          "expected identifier after #pragma");
 989                 eat_until_newline();
 990                 return;
 991         }
 992
 993         symbol_t *symbol = pp_token.identifier.symbol;
 994         if (symbol->pp_ID == TP_STDC) {
 995                 stdc_pragma_kind_t kind = STDC_UNKNOWN;
 996                 /* a STDC pragma */
 997                 if (c_mode & _C99) {
 998                         next_pp_token();
 999
1000                         switch (pp_token.identifier.symbol->pp_ID) {
1001                         case TP_FP_CONTRACT:
1002                                 kind = STDC_FP_CONTRACT;
1003                                 break;
1004                         case TP_FENV_ACCESS:
1005                                 kind = STDC_FENV_ACCESS;
1006                                 break;
1007                         case TP_CX_LIMITED_RANGE:
1008                                 kind = STDC_CX_LIMITED_RANGE;
1009                                 break;
1010                         default:
1011                                 break;
1012                         }
1013                         if (kind != STDC_UNKNOWN) {
1014                                 stdc_pragma_value_kind_t value = STDC_VALUE_UNKNOWN;
1015                                 next_pp_token();
1016                                 switch (pp_token.identifier.symbol->pp_ID) {
1017                                 case TP_ON:
1018                                         value = STDC_VALUE_ON;
1019                                         break;
1020                                 case TP_OFF:
1021                                         value = STDC_VALUE_OFF;
1022                                         break;
1023                                 case TP_DEFAULT:
1024                                         value = STDC_VALUE_DEFAULT;
1025                                         break;
1026                                 default:
1027                                         break;
1028                                 }
1029                                 if (value != STDC_VALUE_UNKNOWN) {
1030                                         unknown_pragma = false;
1031                                 } else {
1032                                         errorf(&pp_token.base.source_position,
1033                                                "bad STDC pragma argument");
1034                                 }
1035                         }
1036                 }
1037         } else {
1038                 unknown_pragma = true;
1039         }
1040         eat_until_newline();
1041         if (unknown_pragma) {
1042                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
1043                          "encountered unknown #pragma");
1044         }
1045 }
1046
1047 /**
1048  * Parse a preprocessor non-null directive.
1049  */
1050 static void parse_preprocessor_identifier(void)
1051 {
1052         assert(pp_token.kind == T_IDENTIFIER);
1053         symbol_t *symbol = pp_token.identifier.symbol;
1054
1055         switch (symbol->pp_ID) {
1056         case TP_include:
1057                 printf("include - enable header name parsing!\n");
1058                 break;
1059         case TP_define:
1060                 define_directive();
1061                 break;
1062         case TP_ifdef:
1063                 ifdef_directive(0);
1064                 break;
1065         case TP_ifndef:
1066                 ifdef_directive(1);
1067                 break;
1068         case TP_endif:
1069                 endif_directive();
1070                 break;
1071         case TP_line:
1072                 next_pp_token();
1073                 parse_line_directive();
1074                 break;
1075         case TP_if:
1076         case TP_else:
1077         case TP_elif:
1078         case TP_undef:
1079         case TP_error:
1080                 /* TODO; output the rest of the line */
1081                 parse_error("#error directive: ");
1082                 break;
1083         case TP_pragma:
1084                 parse_pragma();
1085                 break;
1086         }
1087 }
1088
1089 /**
1090  * Parse a preprocessor directive.
1091  */
1092 static void parse_preprocessor_directive(void)
1093 {
1094         next_pp_token();
1095
1096         switch (pp_token.kind) {
1097         case T_IDENTIFIER:
1098                 parse_preprocessor_identifier();
1099                 break;
1100         case T_INTEGER:
1101                 parse_line_directive();
1102                 break;
1103         case '\n':
1104                 /* NULL directive, see §6.10.7 */
1105                 break;
1106         default:
1107                 parse_error("invalid preprocessor directive");
1108                 eat_until_newline();
1109                 break;
1110         }
1111 }
1112
1113 #define MAYBE_PROLOG                                       \
1114                         next_char();                                   \
1115                         while (true) {                                 \
1116                                 switch (c) {
1117
1118 #define MAYBE(ch, set_type)                                \
1119                                 case ch:                                   \
1120                                         next_char();                           \
1121                                         lexer_token.kind = set_type;           \
1122                                         return;
1123
1124 /* must use this as last thing */
1125 #define MAYBE_MODE(ch, set_type, mode)                     \
1126                                 case ch:                                   \
1127                                         if (c_mode & mode) {                   \
1128                                                 next_char();                       \
1129                                                 lexer_token.kind = set_type;       \
1130                                                 return;                            \
1131                                         }                                      \
1132                                         /* fallthrough */
1133
1134 #define ELSE_CODE(code)                                    \
1135                                 default:                                   \
1136                                         code                                   \
1137                                         return;                                \
1138                                 }                                          \
1139                         } /* end of while (true) */                    \
1140
1141 #define ELSE(set_type)                                     \
1142                 ELSE_CODE(                                         \
1143                         lexer_token.kind = set_type;                   \
1144                 )
1145
1146 void lexer_next_preprocessing_token(void)
1147 {
1148         while (true) {
1149                 lexer_token.base.source_position = lexer_pos;
1150
1151                 switch (c) {
1152                 case ' ':
1153                 case '\t':
1154                         next_char();
1155                         break;
1156
1157                 MATCH_NEWLINE(
1158                         lexer_token.kind = '\n';
1159                         return;
1160                 )
1161
1162                 SYMBOL_CHARS
1163                         parse_symbol();
1164                         /* might be a wide string ( L"string" ) */
1165                         if (lexer_token.identifier.symbol == symbol_L) {
1166                                 switch (c) {
1167                                         case '"':  parse_wide_string_literal();     break;
1168                                         case '\'': parse_wide_character_constant(); break;
1169                                 }
1170                         }
1171                         return;
1172
1173                 DIGITS
1174                         parse_number();
1175                         return;
1176
1177                 case '"':
1178                         parse_string_literal();
1179                         return;
1180
1181                 case '\'':
1182                         parse_character_constant();
1183                         return;
1184
1185                 case '.':
1186                         MAYBE_PROLOG
1187                                 DIGITS
1188                                         put_back(c);
1189                                         c = '.';
1190                                         parse_number();
1191                                         return;
1192
1193                                 case '.':
1194                                         MAYBE_PROLOG
1195                                         MAYBE('.', T_DOTDOTDOT)
1196                                         ELSE_CODE(
1197                                                 put_back(c);
1198                                                 c = '.';
1199                                                 lexer_token.kind = '.';
1200                                         )
1201                         ELSE('.')
1202                 case '&':
1203                         MAYBE_PROLOG
1204                         MAYBE('&', T_ANDAND)
1205                         MAYBE('=', T_ANDEQUAL)
1206                         ELSE('&')
1207                 case '*':
1208                         MAYBE_PROLOG
1209                         MAYBE('=', T_ASTERISKEQUAL)
1210                         ELSE('*')
1211                 case '+':
1212                         MAYBE_PROLOG
1213                         MAYBE('+', T_PLUSPLUS)
1214                         MAYBE('=', T_PLUSEQUAL)
1215                         ELSE('+')
1216                 case '-':
1217                         MAYBE_PROLOG
1218                         MAYBE('>', T_MINUSGREATER)
1219                         MAYBE('-', T_MINUSMINUS)
1220                         MAYBE('=', T_MINUSEQUAL)
1221                         ELSE('-')
1222                 case '!':
1223                         MAYBE_PROLOG
1224                         MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1225                         ELSE('!')
1226                 case '/':
1227                         MAYBE_PROLOG
1228                         MAYBE('=', T_SLASHEQUAL)
1229                                 case '*':
1230                                         next_char();
1231                                         skip_multiline_comment();
1232                                         lexer_next_preprocessing_token();
1233                                         return;
1234                                 case '/':
1235                                         next_char();
1236                                         skip_line_comment();
1237                                         lexer_next_preprocessing_token();
1238                                         return;
1239                         ELSE('/')
1240                 case '%':
1241                         MAYBE_PROLOG
1242                         MAYBE('>', '}')
1243                         MAYBE('=', T_PERCENTEQUAL)
1244                                 case ':':
1245                                         MAYBE_PROLOG
1246                                                 case '%':
1247                                                         MAYBE_PROLOG
1248                                                         MAYBE(':', T_HASHHASH)
1249                                                         ELSE_CODE(
1250                                                                 put_back(c);
1251                                                                 c = '%';
1252                                                                 lexer_token.kind = '#';
1253                                                         )
1254                                         ELSE('#')
1255                         ELSE('%')
1256                 case '<':
1257                         MAYBE_PROLOG
1258                         MAYBE(':', '[')
1259                         MAYBE('%', '{')
1260                         MAYBE('=', T_LESSEQUAL)
1261                                 case '<':
1262                                         MAYBE_PROLOG
1263                                         MAYBE('=', T_LESSLESSEQUAL)
1264                                         ELSE(T_LESSLESS)
1265                         ELSE('<')
1266                 case '>':
1267                         MAYBE_PROLOG
1268                         MAYBE('=', T_GREATEREQUAL)
1269                                 case '>':
1270                                         MAYBE_PROLOG
1271                                         MAYBE('=', T_GREATERGREATEREQUAL)
1272                                         ELSE(T_GREATERGREATER)
1273                         ELSE('>')
1274                 case '^':
1275                         MAYBE_PROLOG
1276                         MAYBE('=', T_CARETEQUAL)
1277                         ELSE('^')
1278                 case '|':
1279                         MAYBE_PROLOG
1280                         MAYBE('=', T_PIPEEQUAL)
1281                         MAYBE('|', T_PIPEPIPE)
1282                         ELSE('|')
1283                 case ':':
1284                         MAYBE_PROLOG
1285                         MAYBE('>', ']')
1286                         MAYBE_MODE(':', T_COLONCOLON, _CXX)
1287                         ELSE(':')
1288                 case '=':
1289                         MAYBE_PROLOG
1290                         MAYBE('=', T_EQUALEQUAL)
1291                         ELSE('=')
1292                 case '#':
1293                         MAYBE_PROLOG
1294                         MAYBE('#', T_HASHHASH)
1295                         ELSE('#')
1296
1297                 case '?':
1298                 case '[':
1299                 case ']':
1300                 case '(':
1301                 case ')':
1302                 case '{':
1303                 case '}':
1304                 case '~':
1305                 case ';':
1306                 case ',':
1307                 case '\\':
1308                         lexer_token.kind = c;
1309                         next_char();
1310                         return;
1311
1312                 case EOF:
1313                         lexer_token.kind = T_EOF;
1314                         return;
1315
1316                 default:
1317 dollar_sign:
1318                         errorf(&lexer_pos, "unknown character '%c' found", c);
1319                         next_char();
1320                         lexer_token.kind = T_ERROR;
1321                         return;
1322                 }
1323         }
1324 }
1325
1326 void lexer_next_token(void)
1327 {
1328         lexer_next_preprocessing_token();
1329
1330         while (lexer_token.kind == '\n') {
1331 newline_found:
1332                 lexer_next_preprocessing_token();
1333         }
1334
1335         if (lexer_token.kind == '#') {
1336                 parse_preprocessor_directive();
1337                 goto newline_found;
1338         }
1339 }
1340
1341 void init_lexer(void)
1342 {
1343         strset_init(&stringset);
1344         symbol_L = symbol_table_insert("L");
1345 }
1346
1347 static void input_error(unsigned delta_lines, unsigned delta_cols,
1348                         const char *message)
1349 {
1350         lexer_pos.lineno += delta_lines;
1351         lexer_pos.colno  += delta_cols;
1352         errorf(&lexer_pos, "%s", message);
1353 }
1354
1355 void select_input_encoding(char const* new_encoding)
1356 {
1357         if (encoding != NULL)
1358                 xfree(encoding);
1359         encoding = xstrdup(new_encoding);
1360 }
1361
1362 void lexer_open_stream(FILE *stream, const char *input_name)
1363 {
1364         if (input != NULL) {
1365                 input_free(input);
1366                 input = NULL;
1367         }
1368
1369         lexer_pos.lineno     = 0;
1370         lexer_pos.colno      = 0;
1371         lexer_pos.input_name = input_name;
1372
1373         set_input_error_callback(input_error);
1374         input  = input_from_stream(stream, encoding);
1375         bufpos = NULL;
1376         bufend = NULL;
1377
1378         /* place a virtual \n at the beginning so the lexer knows that we're
1379          * at the beginning of a line */
1380         c = '\n';
1381 }
1382
1383 void exit_lexer(void)
1384 {
1385         if (input != NULL) {
1386                 input_free(input);
1387                 input = NULL;
1388         }
1389         strset_destroy(&stringset);
1390 }
1391
1392 static __attribute__((unused))
1393 void dbg_pos(const source_position_t source_position)
1394 {
1395         fprintf(stdout, "%s:%u:%u\n", source_position.input_name,
1396                 source_position.lineno, source_position.colno);
1397         fflush(stdout);
1398 }