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