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 NEWLINE  \
 112         '\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                 goto newline; \
 121                 newline // Let it look like an ordinary case label.
 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         case NEWLINE:
 131                 return;
 132
 133         default:
 134                 break;
 135         }
 136
 137         put_back(c);
 138         c = '\\';
 139 }
 140
 141 /**
 142  * Set c to the next input character, ie.
 143  * after expanding trigraphs.
 144  */
 145 static inline void next_char(void)
 146 {
 147         next_real_char();
 148
 149         /* filter trigraphs */
 150         if (UNLIKELY(c == '\\')) {
 151                 maybe_concat_lines();
 152                 return;
 153         }
 154
 155         if (LIKELY(c != '?'))
 156                 return;
 157
 158         next_real_char();
 159         if (LIKELY(c != '?')) {
 160                 put_back(c);
 161                 c = '?';
 162                 return;
 163         }
 164
 165         next_real_char();
 166         switch (c) {
 167         case '=': c = '#'; break;
 168         case '(': c = '['; break;
 169         case '/': c = '\\'; maybe_concat_lines(); break;
 170         case ')': c = ']'; break;
 171         case '\'': c = '^'; break;
 172         case '<': c = '{'; break;
 173         case '!': c = '|'; break;
 174         case '>': c = '}'; break;
 175         case '-': c = '~'; break;
 176         default:
 177                 put_back(c);
 178                 put_back('?');
 179                 c = '?';
 180                 break;
 181         }
 182 }
 183
 184 #define SYMBOL_CHARS  \
 185         case '$': if (!allow_dollar_in_symbol) goto dollar_sign; \
 186         case 'a':         \
 187         case 'b':         \
 188         case 'c':         \
 189         case 'd':         \
 190         case 'e':         \
 191         case 'f':         \
 192         case 'g':         \
 193         case 'h':         \
 194         case 'i':         \
 195         case 'j':         \
 196         case 'k':         \
 197         case 'l':         \
 198         case 'm':         \
 199         case 'n':         \
 200         case 'o':         \
 201         case 'p':         \
 202         case 'q':         \
 203         case 'r':         \
 204         case 's':         \
 205         case 't':         \
 206         case 'u':         \
 207         case 'v':         \
 208         case 'w':         \
 209         case 'x':         \
 210         case 'y':         \
 211         case 'z':         \
 212         case 'A':         \
 213         case 'B':         \
 214         case 'C':         \
 215         case 'D':         \
 216         case 'E':         \
 217         case 'F':         \
 218         case 'G':         \
 219         case 'H':         \
 220         case 'I':         \
 221         case 'J':         \
 222         case 'K':         \
 223         case 'L':         \
 224         case 'M':         \
 225         case 'N':         \
 226         case 'O':         \
 227         case 'P':         \
 228         case 'Q':         \
 229         case 'R':         \
 230         case 'S':         \
 231         case 'T':         \
 232         case 'U':         \
 233         case 'V':         \
 234         case 'W':         \
 235         case 'X':         \
 236         case 'Y':         \
 237         case 'Z':         \
 238         case '_':
 239
 240 #define DIGITS        \
 241         case '0':         \
 242         case '1':         \
 243         case '2':         \
 244         case '3':         \
 245         case '4':         \
 246         case '5':         \
 247         case '6':         \
 248         case '7':         \
 249         case '8':         \
 250         case '9':
 251
 252 /**
 253  * Read a symbol from the input and build
 254  * the lexer_token.
 255  */
 256 static void parse_symbol(void)
 257 {
 258         obstack_1grow(&symbol_obstack, (char) c);
 259         next_char();
 260
 261         while (true) {
 262                 switch (c) {
 263                 DIGITS
 264                 SYMBOL_CHARS
 265                         obstack_1grow(&symbol_obstack, (char) c);
 266                         next_char();
 267                         break;
 268
 269                 default:
 270 dollar_sign:
 271                         goto end_symbol;
 272                 }
 273         }
 274
 275 end_symbol:
 276         obstack_1grow(&symbol_obstack, '\0');
 277
 278         char     *string = obstack_finish(&symbol_obstack);
 279         symbol_t *symbol = symbol_table_insert(string);
 280
 281         lexer_token.kind        = symbol->ID;
 282         lexer_token.base.symbol = symbol;
 283
 284         if (symbol->string != string) {
 285                 obstack_free(&symbol_obstack, string);
 286         }
 287 }
 288
 289 static string_t identify_string(char *string, size_t len)
 290 {
 291         /* TODO hash */
 292 #if 0
 293         const char *result = strset_insert(&stringset, concat);
 294         if (result != concat) {
 295                 obstack_free(&symbol_obstack, concat);
 296         }
 297 #else
 298         const char *result = string;
 299 #endif
 300         return (string_t) {result, len};
 301 }
 302
 303 /**
 304  * parse suffixes like 'LU' or 'f' after numbers
 305  */
 306 static void parse_number_suffix(void)
 307 {
 308         assert(obstack_object_size(&symbol_obstack) == 0);
 309         while (true) {
 310                 switch (c) {
 311                 SYMBOL_CHARS
 312                         obstack_1grow(&symbol_obstack, (char) c);
 313                         next_char();
 314                         break;
 315                 default:
 316                 dollar_sign:
 317                         goto finish_suffix;
 318                 }
 319         }
 320 finish_suffix:
 321         if (obstack_object_size(&symbol_obstack) == 0) {
 322                 lexer_token.number.suffix.begin = NULL;
 323                 lexer_token.number.suffix.size  = 0;
 324                 return;
 325         }
 326
 327         obstack_1grow(&symbol_obstack, '\0');
 328         size_t size   = obstack_object_size(&symbol_obstack) - 1;
 329         char  *string = obstack_finish(&symbol_obstack);
 330
 331         lexer_token.number.suffix = identify_string(string, size);
 332 }
 333
 334 static void parse_exponent(void)
 335 {
 336         if (c == '-' || c == '+') {
 337                 obstack_1grow(&symbol_obstack, (char)c);
 338                 next_char();
 339         }
 340
 341         if (isdigit(c)) {
 342                 do {
 343                         obstack_1grow(&symbol_obstack, (char)c);
 344                         next_char();
 345                 } while (isdigit(c));
 346         } else {
 347                 errorf(&lexer_token.base.source_position, "exponent has no digits");
 348         }
 349 }
 350
 351 /**
 352  * Parses a hex number including hex floats and set the
 353  * lexer_token.
 354  */
 355 static void parse_number_hex(void)
 356 {
 357         bool is_float   = false;
 358         bool has_digits = false;
 359
 360         while (isxdigit(c)) {
 361                 has_digits = true;
 362                 obstack_1grow(&symbol_obstack, (char) c);
 363                 next_char();
 364         }
 365
 366         if (c == '.') {
 367                 is_float = true;
 368                 obstack_1grow(&symbol_obstack, (char) c);
 369                 next_char();
 370
 371                 while (isxdigit(c)) {
 372                         has_digits = true;
 373                         obstack_1grow(&symbol_obstack, (char) c);
 374                         next_char();
 375                 }
 376         }
 377         if (c == 'p' || c == 'P') {
 378                 is_float = true;
 379                 obstack_1grow(&symbol_obstack, (char) c);
 380                 next_char();
 381                 parse_exponent();
 382         } else if (is_float) {
 383                 errorf(&lexer_token.base.source_position,
 384                        "hexadecimal floatingpoint constant requires an exponent");
 385         }
 386         obstack_1grow(&symbol_obstack, '\0');
 387
 388         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 389         char   *string = obstack_finish(&symbol_obstack);
 390         lexer_token.number.number = identify_string(string, size);
 391
 392         lexer_token.kind = is_float ? T_FLOATINGPOINT : T_INTEGER;
 393
 394         if (!has_digits) {
 395                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'", &lexer_token.number.number);
 396                 lexer_token.number.number.begin = "0";
 397                 lexer_token.number.number.size  = 1;
 398         }
 399
 400         parse_number_suffix();
 401 }
 402
 403 static void parse_number_bin(void)
 404 {
 405         bool has_digits = false;
 406
 407         while (c == '0' || c == '1') {
 408                 has_digits = true;
 409                 obstack_1grow(&symbol_obstack, (char)c);
 410                 next_char();
 411         }
 412         obstack_1grow(&symbol_obstack, '\0');
 413
 414         size_t  const size   = obstack_object_size(&symbol_obstack) - 1;
 415         char   *const string = obstack_finish(&symbol_obstack);
 416         lexer_token.number.number = identify_string(string, size);
 417         lexer_token.kind          = T_INTEGER;
 418
 419         if (!has_digits) {
 420                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'", &lexer_token.number.number);
 421                 lexer_token.number.number.begin = "0";
 422                 lexer_token.number.number.size  = 1;
 423         }
 424
 425         parse_number_suffix();
 426 }
 427
 428 /**
 429  * Returns true if the given char is a octal digit.
 430  *
 431  * @param char  the character to check
 432  */
 433 static bool is_octal_digit(utf32 chr)
 434 {
 435         return '0' <= chr && chr <= '7';
 436 }
 437
 438 /**
 439  * Parses a number and sets the lexer_token.
 440  */
 441 static void parse_number(void)
 442 {
 443         bool is_float   = false;
 444         bool has_digits = false;
 445
 446         assert(obstack_object_size(&symbol_obstack) == 0);
 447         if (c == '0') {
 448                 obstack_1grow(&symbol_obstack, (char)c);
 449                 next_char();
 450                 if (c == 'x' || c == 'X') {
 451                         obstack_1grow(&symbol_obstack, (char)c);
 452                         next_char();
 453                         parse_number_hex();
 454                         return;
 455                 } else if (c == 'b' || c == 'B') {
 456                         /* GCC extension: binary constant 0x[bB][01]+.  */
 457                         obstack_1grow(&symbol_obstack, (char)c);
 458                         next_char();
 459                         parse_number_bin();
 460                         return;
 461                 }
 462                 has_digits = true;
 463         }
 464
 465         while (isdigit(c)) {
 466                 has_digits = true;
 467                 obstack_1grow(&symbol_obstack, (char) c);
 468                 next_char();
 469         }
 470
 471         if (c == '.') {
 472                 is_float = true;
 473                 obstack_1grow(&symbol_obstack, '.');
 474                 next_char();
 475
 476                 while (isdigit(c)) {
 477                         has_digits = true;
 478                         obstack_1grow(&symbol_obstack, (char) c);
 479                         next_char();
 480                 }
 481         }
 482         if (c == 'e' || c == 'E') {
 483                 is_float = true;
 484                 obstack_1grow(&symbol_obstack, 'e');
 485                 next_char();
 486                 parse_exponent();
 487         }
 488
 489         obstack_1grow(&symbol_obstack, '\0');
 490         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 491         char   *string = obstack_finish(&symbol_obstack);
 492         lexer_token.number.number = identify_string(string, size);
 493
 494         if (is_float) {
 495                 lexer_token.kind = T_FLOATINGPOINT;
 496         } else {
 497                 lexer_token.kind = T_INTEGER;
 498
 499                 if (string[0] == '0') {
 500                         /* check for invalid octal digits */
 501                         for (size_t i= 0; i < size; ++i) {
 502                                 char t = string[i];
 503                                 if (t >= '8')
 504                                         errorf(&lexer_token.base.source_position, "invalid digit '%c' in octal number", t);
 505                         }
 506                 }
 507         }
 508
 509         if (!has_digits) {
 510                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'",
 511                        &lexer_token.number.number);
 512         }
 513
 514         parse_number_suffix();
 515 }
 516
 517 /**
 518  * Returns the value of a digit.
 519  * The only portable way to do it ...
 520  */
 521 static int digit_value(utf32 const digit)
 522 {
 523         switch (digit) {
 524         case '0': return 0;
 525         case '1': return 1;
 526         case '2': return 2;
 527         case '3': return 3;
 528         case '4': return 4;
 529         case '5': return 5;
 530         case '6': return 6;
 531         case '7': return 7;
 532         case '8': return 8;
 533         case '9': return 9;
 534         case 'a':
 535         case 'A': return 10;
 536         case 'b':
 537         case 'B': return 11;
 538         case 'c':
 539         case 'C': return 12;
 540         case 'd':
 541         case 'D': return 13;
 542         case 'e':
 543         case 'E': return 14;
 544         case 'f':
 545         case 'F': return 15;
 546         default:
 547                 internal_error("wrong character given");
 548         }
 549 }
 550
 551 /**
 552  * Parses an octal character sequence.
 553  *
 554  * @param first_digit  the already read first digit
 555  */
 556 static utf32 parse_octal_sequence(utf32 const first_digit)
 557 {
 558         assert(is_octal_digit(first_digit));
 559         utf32 value = digit_value(first_digit);
 560         if (!is_octal_digit(c)) return value;
 561         value = 8 * value + digit_value(c);
 562         next_char();
 563         if (!is_octal_digit(c)) return value;
 564         value = 8 * value + digit_value(c);
 565         next_char();
 566         return value;
 567 }
 568
 569 /**
 570  * Parses a hex character sequence.
 571  */
 572 static utf32 parse_hex_sequence(void)
 573 {
 574         utf32 value = 0;
 575         while (isxdigit(c)) {
 576                 value = 16 * value + digit_value(c);
 577                 next_char();
 578         }
 579         return value;
 580 }
 581
 582 /**
 583  * Parse an escape sequence.
 584  */
 585 static utf32 parse_escape_sequence(void)
 586 {
 587         eat('\\');
 588
 589         utf32 const ec = c;
 590         next_char();
 591
 592         switch (ec) {
 593         case '"':  return '"';
 594         case '\'': return '\'';
 595         case '\\': return '\\';
 596         case '?': return '\?';
 597         case 'a': return '\a';
 598         case 'b': return '\b';
 599         case 'f': return '\f';
 600         case 'n': return '\n';
 601         case 'r': return '\r';
 602         case 't': return '\t';
 603         case 'v': return '\v';
 604         case 'x':
 605                 return parse_hex_sequence();
 606         case '0':
 607         case '1':
 608         case '2':
 609         case '3':
 610         case '4':
 611         case '5':
 612         case '6':
 613         case '7':
 614                 return parse_octal_sequence(ec);
 615         case EOF:
 616                 parse_error("reached end of file while parsing escape sequence");
 617                 return EOF;
 618         /* \E is not documented, but handled, by GCC.  It is acceptable according
 619          * to §6.11.4, whereas \e is not. */
 620         case 'E':
 621         case 'e':
 622                 if (c_mode & _GNUC)
 623                         return 27;   /* hopefully 27 is ALWAYS the code for ESCAPE */
 624                 break;
 625         case 'u':
 626         case 'U':
 627                 parse_error("universal character parsing not implemented yet");
 628                 return EOF;
 629         default:
 630                 break;
 631         }
 632         /* §6.4.4.4:8 footnote 64 */
 633         parse_error("unknown escape sequence");
 634         return EOF;
 635 }
 636
 637 string_t make_string(const char *string)
 638 {
 639         size_t      len   = strlen(string) + 1;
 640         char *const space = obstack_alloc(&symbol_obstack, len);
 641         memcpy(space, string, len);
 642
 643         return identify_string(space, len);
 644 }
 645
 646 static void parse_string(utf32 const delim, token_kind_t const kind, string_encoding_t const enc, char const *const context)
 647 {
 648         eat(delim);
 649
 650         while (true) {
 651                 switch (c) {
 652                 case '\\': {
 653                         utf32 const tc = parse_escape_sequence();
 654                         if (enc == STRING_ENCODING_CHAR) {
 655                                 if (tc >= 0x100) {
 656                                         warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 657                                 }
 658                                 obstack_1grow(&symbol_obstack, tc);
 659                         } else {
 660                                 obstack_grow_symbol(&symbol_obstack, tc);
 661                         }
 662                         break;
 663                 }
 664
 665                 case NEWLINE:
 666                         errorf(&lexer_pos, "newline while parsing %s", context);
 667                         break;
 668
 669                 case EOF:
 670                         errorf(&lexer_token.base.source_position, "EOF while parsing %s", context);
 671                         goto end_of_string;
 672
 673                 default:
 674                         if (c == delim) {
 675                                 next_char();
 676                                 goto end_of_string;
 677                         } else {
 678                                 obstack_grow_symbol(&symbol_obstack, c);
 679                                 next_char();
 680                                 break;
 681                         }
 682                 }
 683         }
 684
 685 end_of_string:
 686         obstack_1grow(&symbol_obstack, '\0');
 687         size_t const size   = obstack_object_size(&symbol_obstack) - 1;
 688         char  *const string = obstack_finish(&symbol_obstack);
 689
 690         lexer_token.kind            = kind;
 691         lexer_token.string.encoding = enc;
 692         lexer_token.string.string   = identify_string(string, size);
 693 }
 694
 695 /**
 696  * Parse a string literal and set lexer_token.
 697  */
 698 static void parse_string_literal(string_encoding_t const enc)
 699 {
 700         parse_string('"', T_STRING_LITERAL, enc, "string literal");
 701 }
 702
 703 /**
 704  * Parse a character constant and set lexer_token.
 705  */
 706 static void parse_character_constant(string_encoding_t const enc)
 707 {
 708         parse_string('\'', T_CHARACTER_CONSTANT, enc, "character constant");
 709         if (lexer_token.string.string.size == 0) {
 710                 errorf(&lexer_token.base.source_position, "empty character constant");
 711         }
 712 }
 713
 714 /**
 715  * Skip a multiline comment.
 716  */
 717 static void skip_multiline_comment(void)
 718 {
 719         while (true) {
 720                 switch (c) {
 721                 case '/':
 722                         next_char();
 723                         if (c == '*') {
 724                                 /* nested comment, warn here */
 725                                 warningf(WARN_COMMENT, &lexer_pos, "'/*' within comment");
 726                         }
 727                         break;
 728                 case '*':
 729                         next_char();
 730                         if (c == '/') {
 731                                 next_char();
 732                                 return;
 733                         }
 734                         break;
 735
 736                 case NEWLINE:
 737                         break;
 738
 739                 case EOF: {
 740                         errorf(&lexer_token.base.source_position,
 741                                "at end of file while looking for comment end");
 742                         return;
 743                 }
 744
 745                 default:
 746                         next_char();
 747                         break;
 748                 }
 749         }
 750 }
 751
 752 /**
 753  * Skip a single line comment.
 754  */
 755 static void skip_line_comment(void)
 756 {
 757         while (true) {
 758                 switch (c) {
 759                 case EOF:
 760                         return;
 761
 762                 case '\n':
 763                 case '\r':
 764                         return;
 765
 766                 case '\\':
 767                         next_char();
 768                         if (c == '\n' || c == '\r') {
 769                                 warningf(WARN_COMMENT, &lexer_pos, "multi-line comment");
 770                                 return;
 771                         }
 772                         break;
 773
 774                 default:
 775                         next_char();
 776                         break;
 777                 }
 778         }
 779 }
 780
 781 /** The current preprocessor token. */
 782 static token_t pp_token;
 783
 784 /**
 785  * Read the next preprocessor token.
 786  */
 787 static inline void next_pp_token(void)
 788 {
 789         lexer_next_preprocessing_token();
 790         pp_token = lexer_token;
 791 }
 792
 793 /**
 794  * Eat all preprocessor tokens until newline.
 795  */
 796 static void eat_until_newline(void)
 797 {
 798         while (pp_token.kind != '\n' && pp_token.kind != T_EOF) {
 799                 next_pp_token();
 800         }
 801 }
 802
 803 /**
 804  * Parse the line directive.
 805  */
 806 static void parse_line_directive(void)
 807 {
 808         if (pp_token.kind != T_INTEGER) {
 809                 parse_error("expected integer");
 810         } else {
 811                 /* use offset -1 as this is about the next line */
 812                 lexer_pos.lineno = atoi(pp_token.number.number.begin) - 1;
 813                 next_pp_token();
 814         }
 815         if (pp_token.kind == T_STRING_LITERAL && pp_token.string.encoding == STRING_ENCODING_CHAR) {
 816                 lexer_pos.input_name = pp_token.string.string.begin;
 817                 lexer_pos.is_system_header = false;
 818                 next_pp_token();
 819
 820                 /* attempt to parse numeric flags as outputted by gcc preprocessor */
 821                 while (pp_token.kind == T_INTEGER) {
 822                         /* flags:
 823                          * 1 - indicates start of a new file
 824                          * 2 - indicates return from a file
 825                          * 3 - indicates system header
 826                          * 4 - indicates implicit extern "C" in C++ mode
 827                          *
 828                          * currently we're only interested in "3"
 829                          */
 830                         if (streq(pp_token.number.number.begin, "3")) {
 831                                 lexer_pos.is_system_header = true;
 832                         }
 833                         next_pp_token();
 834                 }
 835         }
 836
 837         eat_until_newline();
 838 }
 839
 840 /**
 841  * STDC pragmas.
 842  */
 843 typedef enum stdc_pragma_kind_t {
 844         STDC_UNKNOWN,
 845         STDC_FP_CONTRACT,
 846         STDC_FENV_ACCESS,
 847         STDC_CX_LIMITED_RANGE
 848 } stdc_pragma_kind_t;
 849
 850 /**
 851  * STDC pragma values.
 852  */
 853 typedef enum stdc_pragma_value_kind_t {
 854         STDC_VALUE_UNKNOWN,
 855         STDC_VALUE_ON,
 856         STDC_VALUE_OFF,
 857         STDC_VALUE_DEFAULT
 858 } stdc_pragma_value_kind_t;
 859
 860 /**
 861  * Parse a pragma directive.
 862  */
 863 static void parse_pragma(void)
 864 {
 865         next_pp_token();
 866         if (pp_token.kind != T_IDENTIFIER) {
 867                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
 868                          "expected identifier after #pragma");
 869                 eat_until_newline();
 870                 return;
 871         }
 872
 873         stdc_pragma_kind_t kind = STDC_UNKNOWN;
 874         if (pp_token.base.symbol->pp_ID == TP_STDC && c_mode & _C99) {
 875                 /* a STDC pragma */
 876                 next_pp_token();
 877
 878                 switch (pp_token.base.symbol->pp_ID) {
 879                 case TP_FP_CONTRACT:      kind = STDC_FP_CONTRACT;      break;
 880                 case TP_FENV_ACCESS:      kind = STDC_FENV_ACCESS;      break;
 881                 case TP_CX_LIMITED_RANGE: kind = STDC_CX_LIMITED_RANGE; break;
 882                 default:                  break;
 883                 }
 884                 if (kind != STDC_UNKNOWN) {
 885                         next_pp_token();
 886                         stdc_pragma_value_kind_t value;
 887                         switch (pp_token.base.symbol->pp_ID) {
 888                         case TP_ON:      value = STDC_VALUE_ON;      break;
 889                         case TP_OFF:     value = STDC_VALUE_OFF;     break;
 890                         case TP_DEFAULT: value = STDC_VALUE_DEFAULT; break;
 891                         default:         value = STDC_VALUE_UNKNOWN; break;
 892                         }
 893                         if (value == STDC_VALUE_UNKNOWN) {
 894                                 kind = STDC_UNKNOWN;
 895                                 errorf(&pp_token.base.source_position, "bad STDC pragma argument");
 896                         }
 897                 }
 898         }
 899         eat_until_newline();
 900         if (kind == STDC_UNKNOWN) {
 901                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
 902                          "encountered unknown #pragma");
 903         }
 904 }
 905
 906 /**
 907  * Parse a preprocessor non-null directive.
 908  */
 909 static void parse_preprocessor_identifier(void)
 910 {
 911         assert(pp_token.kind == T_IDENTIFIER);
 912         switch (pp_token.base.symbol->pp_ID) {
 913         case TP_line:
 914                 next_pp_token();
 915                 parse_line_directive();
 916                 break;
 917         case TP_pragma:
 918                 parse_pragma();
 919                 break;
 920         case TP_error:
 921                 /* TODO; output the rest of the line */
 922                 parse_error("#error directive");
 923                 break;
 924         }
 925 }
 926
 927 /**
 928  * Parse a preprocessor directive.
 929  */
 930 static void parse_preprocessor_directive(void)
 931 {
 932         next_pp_token();
 933
 934         switch (pp_token.kind) {
 935         case T_IDENTIFIER:
 936                 parse_preprocessor_identifier();
 937                 break;
 938         case T_INTEGER:
 939                 parse_line_directive();
 940                 break;
 941         case '\n':
 942                 /* NULL directive, see §6.10.7 */
 943                 break;
 944         default:
 945                 parse_error("invalid preprocessor directive");
 946                 eat_until_newline();
 947                 break;
 948         }
 949 }
 950
 951 #define MAYBE_PROLOG                                       \
 952                         next_char();                                   \
 953                         while (true) {                                 \
 954                                 switch (c) {
 955
 956 #define MAYBE(ch, set_type)                                \
 957                                 case ch:                                   \
 958                                         next_char();                           \
 959                                         lexer_token.kind = set_type;           \
 960                                         return;
 961
 962 /* must use this as last thing */
 963 #define MAYBE_MODE(ch, set_type, mode)                     \
 964                                 case ch:                                   \
 965                                         if (c_mode & mode) {                   \
 966                                                 next_char();                       \
 967                                                 lexer_token.kind = set_type;       \
 968                                                 return;                            \
 969                                         }                                      \
 970                                         /* fallthrough */
 971
 972 #define ELSE_CODE(code)                                    \
 973                                 default:                                   \
 974                                         code                                   \
 975                                         return;                                \
 976                                 }                                          \
 977                         } /* end of while (true) */                    \
 978
 979 #define ELSE(set_type)                                     \
 980                 ELSE_CODE(                                         \
 981                         lexer_token.kind = set_type;                   \
 982                 )
 983
 984 void lexer_next_preprocessing_token(void)
 985 {
 986         while (true) {
 987                 lexer_token.base.source_position = lexer_pos;
 988                 lexer_token.base.symbol          = NULL;
 989
 990                 switch (c) {
 991                 case ' ':
 992                 case '\t':
 993                         next_char();
 994                         break;
 995
 996                 case NEWLINE:
 997                         lexer_token.kind = '\n';
 998                         return;
 999
1000                 SYMBOL_CHARS {
1001                         parse_symbol();
1002                         /* might be a wide string ( L"string" ) */
1003                         string_encoding_t const enc = STRING_ENCODING_WIDE;
1004                         if (lexer_token.base.symbol == symbol_L) {
1005                                 switch (c) {
1006                                 case '"':  parse_string_literal(enc);     break;
1007                                 case '\'': parse_character_constant(enc); break;
1008                                 }
1009                         }
1010                         return;
1011                 }
1012
1013                 DIGITS
1014                         parse_number();
1015                         return;
1016
1017                 case '"':
1018                         parse_string_literal(STRING_ENCODING_CHAR);
1019                         return;
1020
1021                 case '\'':
1022                         parse_character_constant(STRING_ENCODING_CHAR);
1023                         return;
1024
1025                 case '.':
1026                         MAYBE_PROLOG
1027                                 DIGITS
1028                                         put_back(c);
1029                                         c = '.';
1030                                         parse_number();
1031                                         return;
1032
1033                                 case '.':
1034                                         MAYBE_PROLOG
1035                                         MAYBE('.', T_DOTDOTDOT)
1036                                         ELSE_CODE(
1037                                                 put_back(c);
1038                                                 c = '.';
1039                                                 lexer_token.kind = '.';
1040                                         )
1041                         ELSE('.')
1042                 case '&':
1043                         MAYBE_PROLOG
1044                         MAYBE('&', T_ANDAND)
1045                         MAYBE('=', T_ANDEQUAL)
1046                         ELSE('&')
1047                 case '*':
1048                         MAYBE_PROLOG
1049                         MAYBE('=', T_ASTERISKEQUAL)
1050                         ELSE('*')
1051                 case '+':
1052                         MAYBE_PROLOG
1053                         MAYBE('+', T_PLUSPLUS)
1054                         MAYBE('=', T_PLUSEQUAL)
1055                         ELSE('+')
1056                 case '-':
1057                         MAYBE_PROLOG
1058                         MAYBE('>', T_MINUSGREATER)
1059                         MAYBE('-', T_MINUSMINUS)
1060                         MAYBE('=', T_MINUSEQUAL)
1061                         ELSE('-')
1062                 case '!':
1063                         MAYBE_PROLOG
1064                         MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1065                         ELSE('!')
1066                 case '/':
1067                         MAYBE_PROLOG
1068                         MAYBE('=', T_SLASHEQUAL)
1069                                 case '*':
1070                                         next_char();
1071                                         skip_multiline_comment();
1072                                         lexer_next_preprocessing_token();
1073                                         return;
1074                                 case '/':
1075                                         next_char();
1076                                         skip_line_comment();
1077                                         lexer_next_preprocessing_token();
1078                                         return;
1079                         ELSE('/')
1080                 case '%':
1081                         MAYBE_PROLOG
1082                         MAYBE('>', '}')
1083                         MAYBE('=', T_PERCENTEQUAL)
1084                                 case ':':
1085                                         MAYBE_PROLOG
1086                                                 case '%':
1087                                                         MAYBE_PROLOG
1088                                                         MAYBE(':', T_HASHHASH)
1089                                                         ELSE_CODE(
1090                                                                 put_back(c);
1091                                                                 c = '%';
1092                                                                 lexer_token.kind = '#';
1093                                                         )
1094                                         ELSE('#')
1095                         ELSE('%')
1096                 case '<':
1097                         MAYBE_PROLOG
1098                         MAYBE(':', '[')
1099                         MAYBE('%', '{')
1100                         MAYBE('=', T_LESSEQUAL)
1101                                 case '<':
1102                                         MAYBE_PROLOG
1103                                         MAYBE('=', T_LESSLESSEQUAL)
1104                                         ELSE(T_LESSLESS)
1105                         ELSE('<')
1106                 case '>':
1107                         MAYBE_PROLOG
1108                         MAYBE('=', T_GREATEREQUAL)
1109                                 case '>':
1110                                         MAYBE_PROLOG
1111                                         MAYBE('=', T_GREATERGREATEREQUAL)
1112                                         ELSE(T_GREATERGREATER)
1113                         ELSE('>')
1114                 case '^':
1115                         MAYBE_PROLOG
1116                         MAYBE('=', T_CARETEQUAL)
1117                         ELSE('^')
1118                 case '|':
1119                         MAYBE_PROLOG
1120                         MAYBE('=', T_PIPEEQUAL)
1121                         MAYBE('|', T_PIPEPIPE)
1122                         ELSE('|')
1123                 case ':':
1124                         MAYBE_PROLOG
1125                         MAYBE('>', ']')
1126                         MAYBE_MODE(':', T_COLONCOLON, _CXX)
1127                         ELSE(':')
1128                 case '=':
1129                         MAYBE_PROLOG
1130                         MAYBE('=', T_EQUALEQUAL)
1131                         ELSE('=')
1132                 case '#':
1133                         MAYBE_PROLOG
1134                         MAYBE('#', T_HASHHASH)
1135                         ELSE('#')
1136
1137                 case '?':
1138                 case '[':
1139                 case ']':
1140                 case '(':
1141                 case ')':
1142                 case '{':
1143                 case '}':
1144                 case '~':
1145                 case ';':
1146                 case ',':
1147                 case '\\':
1148                         lexer_token.kind = c;
1149                         next_char();
1150                         return;
1151
1152                 case EOF:
1153                         lexer_token.kind = T_EOF;
1154                         return;
1155
1156                 default:
1157 dollar_sign:
1158                         errorf(&lexer_pos, "unknown character '%c' found", c);
1159                         next_char();
1160                         break;
1161                 }
1162         }
1163 }
1164
1165 void lexer_next_token(void)
1166 {
1167         lexer_next_preprocessing_token();
1168
1169         while (lexer_token.kind == '\n') {
1170 newline_found:
1171                 lexer_next_preprocessing_token();
1172         }
1173
1174         if (lexer_token.kind == '#') {
1175                 parse_preprocessor_directive();
1176                 goto newline_found;
1177         }
1178 }
1179
1180 void init_lexer(void)
1181 {
1182         strset_init(&stringset);
1183         symbol_L = symbol_table_insert("L");
1184 }
1185
1186 static void input_error(unsigned delta_lines, unsigned delta_cols,
1187                         const char *message)
1188 {
1189         lexer_pos.lineno += delta_lines;
1190         lexer_pos.colno  += delta_cols;
1191         errorf(&lexer_pos, "%s", message);
1192 }
1193
1194 void lexer_switch_input(input_t *new_input, const char *input_name)
1195 {
1196         lexer_pos.lineno     = 0;
1197         lexer_pos.colno      = 0;
1198         lexer_pos.input_name = input_name;
1199
1200         set_input_error_callback(input_error);
1201         input  = new_input;
1202         bufpos = NULL;
1203         bufend = NULL;
1204
1205         /* place a virtual \n at the beginning so the lexer knows that we're
1206          * at the beginning of a line */
1207         c = '\n';
1208 }
1209
1210 void exit_lexer(void)
1211 {
1212         strset_destroy(&stringset);
1213 }
1214
1215 static __attribute__((unused))
1216 void dbg_pos(const source_position_t source_position)
1217 {
1218         fprintf(stdout, "%s:%u:%u\n", source_position.input_name,
1219                 source_position.lineno, (unsigned)source_position.colno);
1220         fflush(stdout);
1221 }