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.base.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 static void parse_number_bin(void)
 402 {
 403         bool has_digits = false;
 404
 405         while (c == '0' || c == '1') {
 406                 has_digits = true;
 407                 obstack_1grow(&symbol_obstack, (char)c);
 408                 next_char();
 409         }
 410         obstack_1grow(&symbol_obstack, '\0');
 411
 412         size_t  const size   = obstack_object_size(&symbol_obstack) - 1;
 413         char   *const string = obstack_finish(&symbol_obstack);
 414         lexer_token.number.number = identify_string(string, size);
 415         lexer_token.kind          = T_INTEGER;
 416
 417         if (!has_digits) {
 418                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'", &lexer_token.number.number);
 419                 lexer_token.number.number.begin = "0";
 420                 lexer_token.number.number.size  = 1;
 421         }
 422
 423         parse_number_suffix();
 424 }
 425
 426 /**
 427  * Returns true if the given char is a octal digit.
 428  *
 429  * @param char  the character to check
 430  */
 431 static bool is_octal_digit(utf32 chr)
 432 {
 433         return '0' <= chr && chr <= '7';
 434 }
 435
 436 /**
 437  * Parses a number and sets the lexer_token.
 438  */
 439 static void parse_number(void)
 440 {
 441         bool is_float   = false;
 442         bool has_digits = false;
 443
 444         assert(obstack_object_size(&symbol_obstack) == 0);
 445         if (c == '0') {
 446                 obstack_1grow(&symbol_obstack, (char)c);
 447                 next_char();
 448                 if (c == 'x' || c == 'X') {
 449                         obstack_1grow(&symbol_obstack, (char)c);
 450                         next_char();
 451                         parse_number_hex();
 452                         return;
 453                 } else if (c == 'b' || c == 'B') {
 454                         /* GCC extension: binary constant 0x[bB][01]+.  */
 455                         obstack_1grow(&symbol_obstack, (char)c);
 456                         next_char();
 457                         parse_number_bin();
 458                         return;
 459                 }
 460                 has_digits = true;
 461         }
 462
 463         while (isdigit(c)) {
 464                 has_digits = true;
 465                 obstack_1grow(&symbol_obstack, (char) c);
 466                 next_char();
 467         }
 468
 469         if (c == '.') {
 470                 is_float = true;
 471                 obstack_1grow(&symbol_obstack, '.');
 472                 next_char();
 473
 474                 while (isdigit(c)) {
 475                         has_digits = true;
 476                         obstack_1grow(&symbol_obstack, (char) c);
 477                         next_char();
 478                 }
 479         }
 480         if (c == 'e' || c == 'E') {
 481                 is_float = true;
 482                 obstack_1grow(&symbol_obstack, 'e');
 483                 next_char();
 484                 parse_exponent();
 485         }
 486
 487         obstack_1grow(&symbol_obstack, '\0');
 488         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 489         char   *string = obstack_finish(&symbol_obstack);
 490         lexer_token.number.number = identify_string(string, size);
 491
 492         if (is_float) {
 493                 lexer_token.kind = T_FLOATINGPOINT;
 494         } else {
 495                 lexer_token.kind = T_INTEGER;
 496
 497                 if (string[0] == '0') {
 498                         /* check for invalid octal digits */
 499                         for (size_t i= 0; i < size; ++i) {
 500                                 char t = string[i];
 501                                 if (t >= '8')
 502                                         errorf(&lexer_token.base.source_position, "invalid digit '%c' in octal number", t);
 503                         }
 504                 }
 505         }
 506
 507         if (!has_digits) {
 508                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'",
 509                        &lexer_token.number.number);
 510         }
 511
 512         parse_number_suffix();
 513 }
 514
 515 /**
 516  * Returns the value of a digit.
 517  * The only portable way to do it ...
 518  */
 519 static int digit_value(utf32 const digit)
 520 {
 521         switch (digit) {
 522         case '0': return 0;
 523         case '1': return 1;
 524         case '2': return 2;
 525         case '3': return 3;
 526         case '4': return 4;
 527         case '5': return 5;
 528         case '6': return 6;
 529         case '7': return 7;
 530         case '8': return 8;
 531         case '9': return 9;
 532         case 'a':
 533         case 'A': return 10;
 534         case 'b':
 535         case 'B': return 11;
 536         case 'c':
 537         case 'C': return 12;
 538         case 'd':
 539         case 'D': return 13;
 540         case 'e':
 541         case 'E': return 14;
 542         case 'f':
 543         case 'F': return 15;
 544         default:
 545                 internal_error("wrong character given");
 546         }
 547 }
 548
 549 /**
 550  * Parses an octal character sequence.
 551  *
 552  * @param first_digit  the already read first digit
 553  */
 554 static utf32 parse_octal_sequence(utf32 const first_digit)
 555 {
 556         assert(is_octal_digit(first_digit));
 557         utf32 value = digit_value(first_digit);
 558         if (!is_octal_digit(c)) return value;
 559         value = 8 * value + digit_value(c);
 560         next_char();
 561         if (!is_octal_digit(c)) return value;
 562         value = 8 * value + digit_value(c);
 563         next_char();
 564         return value;
 565 }
 566
 567 /**
 568  * Parses a hex character sequence.
 569  */
 570 static utf32 parse_hex_sequence(void)
 571 {
 572         utf32 value = 0;
 573         while (isxdigit(c)) {
 574                 value = 16 * value + digit_value(c);
 575                 next_char();
 576         }
 577         return value;
 578 }
 579
 580 /**
 581  * Parse an escape sequence.
 582  */
 583 static utf32 parse_escape_sequence(void)
 584 {
 585         eat('\\');
 586
 587         utf32 const ec = c;
 588         next_char();
 589
 590         switch (ec) {
 591         case '"':  return '"';
 592         case '\'': return '\'';
 593         case '\\': return '\\';
 594         case '?': return '\?';
 595         case 'a': return '\a';
 596         case 'b': return '\b';
 597         case 'f': return '\f';
 598         case 'n': return '\n';
 599         case 'r': return '\r';
 600         case 't': return '\t';
 601         case 'v': return '\v';
 602         case 'x':
 603                 return parse_hex_sequence();
 604         case '0':
 605         case '1':
 606         case '2':
 607         case '3':
 608         case '4':
 609         case '5':
 610         case '6':
 611         case '7':
 612                 return parse_octal_sequence(ec);
 613         case EOF:
 614                 parse_error("reached end of file while parsing escape sequence");
 615                 return EOF;
 616         /* \E is not documented, but handled, by GCC.  It is acceptable according
 617          * to §6.11.4, whereas \e is not. */
 618         case 'E':
 619         case 'e':
 620                 if (c_mode & _GNUC)
 621                         return 27;   /* hopefully 27 is ALWAYS the code for ESCAPE */
 622                 break;
 623         case 'u':
 624         case 'U':
 625                 parse_error("universal character parsing not implemented yet");
 626                 return EOF;
 627         default:
 628                 break;
 629         }
 630         /* §6.4.4.4:8 footnote 64 */
 631         parse_error("unknown escape sequence");
 632         return EOF;
 633 }
 634
 635 string_t make_string(const char *string)
 636 {
 637         size_t      len   = strlen(string) + 1;
 638         char *const space = obstack_alloc(&symbol_obstack, len);
 639         memcpy(space, string, len);
 640
 641         return identify_string(space, len);
 642 }
 643
 644 static void parse_string(utf32 const delim, token_kind_t const kind, string_encoding_t const enc, char const *const context)
 645 {
 646         eat(delim);
 647
 648         while (true) {
 649                 switch (c) {
 650                 case '\\': {
 651                         utf32 const tc = parse_escape_sequence();
 652                         if (enc == STRING_ENCODING_CHAR) {
 653                                 if (tc >= 0x100) {
 654                                         warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 655                                 }
 656                                 obstack_1grow(&symbol_obstack, tc);
 657                         } else {
 658                                 obstack_grow_symbol(&symbol_obstack, tc);
 659                         }
 660                         break;
 661                 }
 662
 663                 MATCH_NEWLINE(
 664                         errorf(&lexer_pos, "newline while parsing %s", context);
 665                         break;
 666                 )
 667
 668                 case EOF:
 669                         errorf(&lexer_token.base.source_position, "EOF while parsing %s", context);
 670                         goto end_of_string;
 671
 672                 default:
 673                         if (c == delim) {
 674                                 next_char();
 675                                 goto end_of_string;
 676                         } else {
 677                                 obstack_grow_symbol(&symbol_obstack, c);
 678                                 next_char();
 679                                 break;
 680                         }
 681                 }
 682         }
 683
 684 end_of_string:
 685         obstack_1grow(&symbol_obstack, '\0');
 686         size_t const size   = obstack_object_size(&symbol_obstack) - 1;
 687         char  *const string = obstack_finish(&symbol_obstack);
 688
 689         lexer_token.kind            = kind;
 690         lexer_token.string.encoding = enc;
 691         lexer_token.string.string   = identify_string(string, size);
 692 }
 693
 694 /**
 695  * Parse a string literal and set lexer_token.
 696  */
 697 static void parse_string_literal(string_encoding_t const enc)
 698 {
 699         parse_string('"', T_STRING_LITERAL, enc, "string literal");
 700 }
 701
 702 /**
 703  * Parse a character constant and set lexer_token.
 704  */
 705 static void parse_character_constant(string_encoding_t const enc)
 706 {
 707         parse_string('\'', T_CHARACTER_CONSTANT, enc, "character constant");
 708         if (lexer_token.string.string.size == 0) {
 709                 errorf(&lexer_token.base.source_position, "empty character constant");
 710         }
 711 }
 712
 713 /**
 714  * Skip a multiline comment.
 715  */
 716 static void skip_multiline_comment(void)
 717 {
 718         while (true) {
 719                 switch (c) {
 720                 case '/':
 721                         next_char();
 722                         if (c == '*') {
 723                                 /* nested comment, warn here */
 724                                 warningf(WARN_COMMENT, &lexer_pos, "'/*' within comment");
 725                         }
 726                         break;
 727                 case '*':
 728                         next_char();
 729                         if (c == '/') {
 730                                 next_char();
 731                                 return;
 732                         }
 733                         break;
 734
 735                 MATCH_NEWLINE(break;)
 736
 737                 case EOF: {
 738                         errorf(&lexer_token.base.source_position,
 739                                "at end of file while looking for comment end");
 740                         return;
 741                 }
 742
 743                 default:
 744                         next_char();
 745                         break;
 746                 }
 747         }
 748 }
 749
 750 /**
 751  * Skip a single line comment.
 752  */
 753 static void skip_line_comment(void)
 754 {
 755         while (true) {
 756                 switch (c) {
 757                 case EOF:
 758                         return;
 759
 760                 case '\n':
 761                 case '\r':
 762                         return;
 763
 764                 case '\\':
 765                         next_char();
 766                         if (c == '\n' || c == '\r') {
 767                                 warningf(WARN_COMMENT, &lexer_pos, "multi-line comment");
 768                                 return;
 769                         }
 770                         break;
 771
 772                 default:
 773                         next_char();
 774                         break;
 775                 }
 776         }
 777 }
 778
 779 /** The current preprocessor token. */
 780 static token_t pp_token;
 781
 782 /**
 783  * Read the next preprocessor token.
 784  */
 785 static inline void next_pp_token(void)
 786 {
 787         lexer_next_preprocessing_token();
 788         pp_token = lexer_token;
 789 }
 790
 791 /**
 792  * Eat all preprocessor tokens until newline.
 793  */
 794 static void eat_until_newline(void)
 795 {
 796         while (pp_token.kind != '\n' && pp_token.kind != T_EOF) {
 797                 next_pp_token();
 798         }
 799 }
 800
 801 /**
 802  * Parse the line directive.
 803  */
 804 static void parse_line_directive(void)
 805 {
 806         if (pp_token.kind != T_INTEGER) {
 807                 parse_error("expected integer");
 808         } else {
 809                 /* use offset -1 as this is about the next line */
 810                 lexer_pos.lineno = atoi(pp_token.number.number.begin) - 1;
 811                 next_pp_token();
 812         }
 813         if (pp_token.kind == T_STRING_LITERAL && pp_token.string.encoding == STRING_ENCODING_CHAR) {
 814                 lexer_pos.input_name = pp_token.string.string.begin;
 815                 lexer_pos.is_system_header = false;
 816                 next_pp_token();
 817
 818                 /* attempt to parse numeric flags as outputted by gcc preprocessor */
 819                 while (pp_token.kind == T_INTEGER) {
 820                         /* flags:
 821                          * 1 - indicates start of a new file
 822                          * 2 - indicates return from a file
 823                          * 3 - indicates system header
 824                          * 4 - indicates implicit extern "C" in C++ mode
 825                          *
 826                          * currently we're only interested in "3"
 827                          */
 828                         if (streq(pp_token.number.number.begin, "3")) {
 829                                 lexer_pos.is_system_header = true;
 830                         }
 831                         next_pp_token();
 832                 }
 833         }
 834
 835         eat_until_newline();
 836 }
 837
 838 /**
 839  * STDC pragmas.
 840  */
 841 typedef enum stdc_pragma_kind_t {
 842         STDC_UNKNOWN,
 843         STDC_FP_CONTRACT,
 844         STDC_FENV_ACCESS,
 845         STDC_CX_LIMITED_RANGE
 846 } stdc_pragma_kind_t;
 847
 848 /**
 849  * STDC pragma values.
 850  */
 851 typedef enum stdc_pragma_value_kind_t {
 852         STDC_VALUE_UNKNOWN,
 853         STDC_VALUE_ON,
 854         STDC_VALUE_OFF,
 855         STDC_VALUE_DEFAULT
 856 } stdc_pragma_value_kind_t;
 857
 858 /**
 859  * Parse a pragma directive.
 860  */
 861 static void parse_pragma(void)
 862 {
 863         next_pp_token();
 864         if (pp_token.kind != T_IDENTIFIER) {
 865                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
 866                          "expected identifier after #pragma");
 867                 eat_until_newline();
 868                 return;
 869         }
 870
 871         stdc_pragma_kind_t kind = STDC_UNKNOWN;
 872         if (pp_token.base.symbol->pp_ID == TP_STDC && c_mode & _C99) {
 873                 /* a STDC pragma */
 874                 next_pp_token();
 875
 876                 switch (pp_token.base.symbol->pp_ID) {
 877                 case TP_FP_CONTRACT:      kind = STDC_FP_CONTRACT;      break;
 878                 case TP_FENV_ACCESS:      kind = STDC_FENV_ACCESS;      break;
 879                 case TP_CX_LIMITED_RANGE: kind = STDC_CX_LIMITED_RANGE; break;
 880                 default:                  break;
 881                 }
 882                 if (kind != STDC_UNKNOWN) {
 883                         next_pp_token();
 884                         stdc_pragma_value_kind_t value;
 885                         switch (pp_token.base.symbol->pp_ID) {
 886                         case TP_ON:      value = STDC_VALUE_ON;      break;
 887                         case TP_OFF:     value = STDC_VALUE_OFF;     break;
 888                         case TP_DEFAULT: value = STDC_VALUE_DEFAULT; break;
 889                         default:         value = STDC_VALUE_UNKNOWN; break;
 890                         }
 891                         if (value == STDC_VALUE_UNKNOWN) {
 892                                 kind = STDC_UNKNOWN;
 893                                 errorf(&pp_token.base.source_position, "bad STDC pragma argument");
 894                         }
 895                 }
 896         }
 897         eat_until_newline();
 898         if (kind == STDC_UNKNOWN) {
 899                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
 900                          "encountered unknown #pragma");
 901         }
 902 }
 903
 904 /**
 905  * Parse a preprocessor non-null directive.
 906  */
 907 static void parse_preprocessor_identifier(void)
 908 {
 909         assert(pp_token.kind == T_IDENTIFIER);
 910         switch (pp_token.base.symbol->pp_ID) {
 911         case TP_line:
 912                 next_pp_token();
 913                 parse_line_directive();
 914                 break;
 915         case TP_pragma:
 916                 parse_pragma();
 917                 break;
 918         case TP_error:
 919                 /* TODO; output the rest of the line */
 920                 parse_error("#error directive");
 921                 break;
 922         }
 923 }
 924
 925 /**
 926  * Parse a preprocessor directive.
 927  */
 928 static void parse_preprocessor_directive(void)
 929 {
 930         next_pp_token();
 931
 932         switch (pp_token.kind) {
 933         case T_IDENTIFIER:
 934                 parse_preprocessor_identifier();
 935                 break;
 936         case T_INTEGER:
 937                 parse_line_directive();
 938                 break;
 939         case '\n':
 940                 /* NULL directive, see §6.10.7 */
 941                 break;
 942         default:
 943                 parse_error("invalid preprocessor directive");
 944                 eat_until_newline();
 945                 break;
 946         }
 947 }
 948
 949 #define MAYBE_PROLOG                                       \
 950                         next_char();                                   \
 951                         while (true) {                                 \
 952                                 switch (c) {
 953
 954 #define MAYBE(ch, set_type)                                \
 955                                 case ch:                                   \
 956                                         next_char();                           \
 957                                         lexer_token.kind = set_type;           \
 958                                         return;
 959
 960 /* must use this as last thing */
 961 #define MAYBE_MODE(ch, set_type, mode)                     \
 962                                 case ch:                                   \
 963                                         if (c_mode & mode) {                   \
 964                                                 next_char();                       \
 965                                                 lexer_token.kind = set_type;       \
 966                                                 return;                            \
 967                                         }                                      \
 968                                         /* fallthrough */
 969
 970 #define ELSE_CODE(code)                                    \
 971                                 default:                                   \
 972                                         code                                   \
 973                                         return;                                \
 974                                 }                                          \
 975                         } /* end of while (true) */                    \
 976
 977 #define ELSE(set_type)                                     \
 978                 ELSE_CODE(                                         \
 979                         lexer_token.kind = set_type;                   \
 980                 )
 981
 982 void lexer_next_preprocessing_token(void)
 983 {
 984         while (true) {
 985                 lexer_token.base.source_position = lexer_pos;
 986                 lexer_token.base.symbol          = NULL;
 987
 988                 switch (c) {
 989                 case ' ':
 990                 case '\t':
 991                         next_char();
 992                         break;
 993
 994                 MATCH_NEWLINE(
 995                         lexer_token.kind = '\n';
 996                         return;
 997                 )
 998
 999                 SYMBOL_CHARS {
1000                         parse_symbol();
1001                         /* might be a wide string ( L"string" ) */
1002                         string_encoding_t const enc = STRING_ENCODING_WIDE;
1003                         if (lexer_token.base.symbol == symbol_L) {
1004                                 switch (c) {
1005                                 case '"':  parse_string_literal(enc);     break;
1006                                 case '\'': parse_character_constant(enc); break;
1007                                 }
1008                         }
1009                         return;
1010                 }
1011
1012                 DIGITS
1013                         parse_number();
1014                         return;
1015
1016                 case '"':
1017                         parse_string_literal(STRING_ENCODING_CHAR);
1018                         return;
1019
1020                 case '\'':
1021                         parse_character_constant(STRING_ENCODING_CHAR);
1022                         return;
1023
1024                 case '.':
1025                         MAYBE_PROLOG
1026                                 DIGITS
1027                                         put_back(c);
1028                                         c = '.';
1029                                         parse_number();
1030                                         return;
1031
1032                                 case '.':
1033                                         MAYBE_PROLOG
1034                                         MAYBE('.', T_DOTDOTDOT)
1035                                         ELSE_CODE(
1036                                                 put_back(c);
1037                                                 c = '.';
1038                                                 lexer_token.kind = '.';
1039                                         )
1040                         ELSE('.')
1041                 case '&':
1042                         MAYBE_PROLOG
1043                         MAYBE('&', T_ANDAND)
1044                         MAYBE('=', T_ANDEQUAL)
1045                         ELSE('&')
1046                 case '*':
1047                         MAYBE_PROLOG
1048                         MAYBE('=', T_ASTERISKEQUAL)
1049                         ELSE('*')
1050                 case '+':
1051                         MAYBE_PROLOG
1052                         MAYBE('+', T_PLUSPLUS)
1053                         MAYBE('=', T_PLUSEQUAL)
1054                         ELSE('+')
1055                 case '-':
1056                         MAYBE_PROLOG
1057                         MAYBE('>', T_MINUSGREATER)
1058                         MAYBE('-', T_MINUSMINUS)
1059                         MAYBE('=', T_MINUSEQUAL)
1060                         ELSE('-')
1061                 case '!':
1062                         MAYBE_PROLOG
1063                         MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1064                         ELSE('!')
1065                 case '/':
1066                         MAYBE_PROLOG
1067                         MAYBE('=', T_SLASHEQUAL)
1068                                 case '*':
1069                                         next_char();
1070                                         skip_multiline_comment();
1071                                         lexer_next_preprocessing_token();
1072                                         return;
1073                                 case '/':
1074                                         next_char();
1075                                         skip_line_comment();
1076                                         lexer_next_preprocessing_token();
1077                                         return;
1078                         ELSE('/')
1079                 case '%':
1080                         MAYBE_PROLOG
1081                         MAYBE('>', '}')
1082                         MAYBE('=', T_PERCENTEQUAL)
1083                                 case ':':
1084                                         MAYBE_PROLOG
1085                                                 case '%':
1086                                                         MAYBE_PROLOG
1087                                                         MAYBE(':', T_HASHHASH)
1088                                                         ELSE_CODE(
1089                                                                 put_back(c);
1090                                                                 c = '%';
1091                                                                 lexer_token.kind = '#';
1092                                                         )
1093                                         ELSE('#')
1094                         ELSE('%')
1095                 case '<':
1096                         MAYBE_PROLOG
1097                         MAYBE(':', '[')
1098                         MAYBE('%', '{')
1099                         MAYBE('=', T_LESSEQUAL)
1100                                 case '<':
1101                                         MAYBE_PROLOG
1102                                         MAYBE('=', T_LESSLESSEQUAL)
1103                                         ELSE(T_LESSLESS)
1104                         ELSE('<')
1105                 case '>':
1106                         MAYBE_PROLOG
1107                         MAYBE('=', T_GREATEREQUAL)
1108                                 case '>':
1109                                         MAYBE_PROLOG
1110                                         MAYBE('=', T_GREATERGREATEREQUAL)
1111                                         ELSE(T_GREATERGREATER)
1112                         ELSE('>')
1113                 case '^':
1114                         MAYBE_PROLOG
1115                         MAYBE('=', T_CARETEQUAL)
1116                         ELSE('^')
1117                 case '|':
1118                         MAYBE_PROLOG
1119                         MAYBE('=', T_PIPEEQUAL)
1120                         MAYBE('|', T_PIPEPIPE)
1121                         ELSE('|')
1122                 case ':':
1123                         MAYBE_PROLOG
1124                         MAYBE('>', ']')
1125                         MAYBE_MODE(':', T_COLONCOLON, _CXX)
1126                         ELSE(':')
1127                 case '=':
1128                         MAYBE_PROLOG
1129                         MAYBE('=', T_EQUALEQUAL)
1130                         ELSE('=')
1131                 case '#':
1132                         MAYBE_PROLOG
1133                         MAYBE('#', T_HASHHASH)
1134                         ELSE('#')
1135
1136                 case '?':
1137                 case '[':
1138                 case ']':
1139                 case '(':
1140                 case ')':
1141                 case '{':
1142                 case '}':
1143                 case '~':
1144                 case ';':
1145                 case ',':
1146                 case '\\':
1147                         lexer_token.kind = c;
1148                         next_char();
1149                         return;
1150
1151                 case EOF:
1152                         lexer_token.kind = T_EOF;
1153                         return;
1154
1155                 default:
1156 dollar_sign:
1157                         errorf(&lexer_pos, "unknown character '%c' found", c);
1158                         next_char();
1159                         break;
1160                 }
1161         }
1162 }
1163
1164 void lexer_next_token(void)
1165 {
1166         lexer_next_preprocessing_token();
1167
1168         while (lexer_token.kind == '\n') {
1169 newline_found:
1170                 lexer_next_preprocessing_token();
1171         }
1172
1173         if (lexer_token.kind == '#') {
1174                 parse_preprocessor_directive();
1175                 goto newline_found;
1176         }
1177 }
1178
1179 void init_lexer(void)
1180 {
1181         strset_init(&stringset);
1182         symbol_L = symbol_table_insert("L");
1183 }
1184
1185 static void input_error(unsigned delta_lines, unsigned delta_cols,
1186                         const char *message)
1187 {
1188         lexer_pos.lineno += delta_lines;
1189         lexer_pos.colno  += delta_cols;
1190         errorf(&lexer_pos, "%s", message);
1191 }
1192
1193 void lexer_switch_input(input_t *new_input, const char *input_name)
1194 {
1195         lexer_pos.lineno     = 0;
1196         lexer_pos.colno      = 0;
1197         lexer_pos.input_name = input_name;
1198
1199         set_input_error_callback(input_error);
1200         input  = new_input;
1201         bufpos = NULL;
1202         bufend = NULL;
1203
1204         /* place a virtual \n at the beginning so the lexer knows that we're
1205          * at the beginning of a line */
1206         c = '\n';
1207 }
1208
1209 void exit_lexer(void)
1210 {
1211         strset_destroy(&stringset);
1212 }
1213
1214 static __attribute__((unused))
1215 void dbg_pos(const source_position_t source_position)
1216 {
1217         fprintf(stdout, "%s:%u:%u\n", source_position.input_name,
1218                 source_position.lineno, (unsigned)source_position.colno);
1219         fflush(stdout);
1220 }