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         assert(obstack_object_size(&symbol_obstack) == 0);
 359         while (isxdigit(c)) {
 360                 has_digits = true;
 361                 obstack_1grow(&symbol_obstack, (char) c);
 362                 next_char();
 363         }
 364
 365         if (c == '.') {
 366                 is_float = true;
 367                 obstack_1grow(&symbol_obstack, (char) c);
 368                 next_char();
 369
 370                 while (isxdigit(c)) {
 371                         has_digits = true;
 372                         obstack_1grow(&symbol_obstack, (char) c);
 373                         next_char();
 374                 }
 375         }
 376         if (c == 'p' || c == 'P') {
 377                 is_float = true;
 378                 obstack_1grow(&symbol_obstack, (char) c);
 379                 next_char();
 380                 parse_exponent();
 381         } else if (is_float) {
 382                 errorf(&lexer_token.base.source_position,
 383                        "hexadecimal floatingpoint constant requires an exponent");
 384         }
 385         obstack_1grow(&symbol_obstack, '\0');
 386
 387         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 388         char   *string = obstack_finish(&symbol_obstack);
 389         lexer_token.number.number = identify_string(string, size);
 390
 391         lexer_token.kind    =
 392                 is_float ? T_FLOATINGPOINT_HEXADECIMAL : T_INTEGER_HEXADECIMAL;
 393
 394         if (!has_digits) {
 395                 errorf(&lexer_token.base.source_position,
 396                        "invalid number literal '0x%S'", &lexer_token.number.number);
 397                 lexer_token.number.number.begin = "0";
 398                 lexer_token.number.number.size  = 1;
 399         }
 400
 401         parse_number_suffix();
 402 }
 403
 404 /**
 405  * Returns true if the given char is a octal digit.
 406  *
 407  * @param char  the character to check
 408  */
 409 static bool is_octal_digit(utf32 chr)
 410 {
 411         return '0' <= chr && chr <= '7';
 412 }
 413
 414 /**
 415  * Parses a number and sets the lexer_token.
 416  */
 417 static void parse_number(void)
 418 {
 419         bool is_float   = false;
 420         bool has_digits = false;
 421
 422         assert(obstack_object_size(&symbol_obstack) == 0);
 423         if (c == '0') {
 424                 next_char();
 425                 if (c == 'x' || c == 'X') {
 426                         next_char();
 427                         parse_number_hex();
 428                         return;
 429                 } else {
 430                         has_digits = true;
 431                 }
 432                 obstack_1grow(&symbol_obstack, '0');
 433         }
 434
 435         while (isdigit(c)) {
 436                 has_digits = true;
 437                 obstack_1grow(&symbol_obstack, (char) c);
 438                 next_char();
 439         }
 440
 441         if (c == '.') {
 442                 is_float = true;
 443                 obstack_1grow(&symbol_obstack, '.');
 444                 next_char();
 445
 446                 while (isdigit(c)) {
 447                         has_digits = true;
 448                         obstack_1grow(&symbol_obstack, (char) c);
 449                         next_char();
 450                 }
 451         }
 452         if (c == 'e' || c == 'E') {
 453                 is_float = true;
 454                 obstack_1grow(&symbol_obstack, 'e');
 455                 next_char();
 456                 parse_exponent();
 457         }
 458
 459         obstack_1grow(&symbol_obstack, '\0');
 460         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 461         char   *string = obstack_finish(&symbol_obstack);
 462         lexer_token.number.number = identify_string(string, size);
 463
 464         /* is it an octal number? */
 465         if (is_float) {
 466                 lexer_token.kind = T_FLOATINGPOINT;
 467         } else if (string[0] == '0') {
 468                 lexer_token.kind = T_INTEGER_OCTAL;
 469
 470                 /* check for invalid octal digits */
 471                 for (size_t i= 0; i < size; ++i) {
 472                         char t = string[i];
 473                         if (t >= '8')
 474                                 errorf(&lexer_token.base.source_position,
 475                                        "invalid digit '%c' in octal number", t);
 476                 }
 477         } else {
 478                 lexer_token.kind = T_INTEGER;
 479         }
 480
 481         if (!has_digits) {
 482                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'",
 483                        &lexer_token.number.number);
 484         }
 485
 486         parse_number_suffix();
 487 }
 488
 489 /**
 490  * Returns the value of a digit.
 491  * The only portable way to do it ...
 492  */
 493 static int digit_value(utf32 const digit)
 494 {
 495         switch (digit) {
 496         case '0': return 0;
 497         case '1': return 1;
 498         case '2': return 2;
 499         case '3': return 3;
 500         case '4': return 4;
 501         case '5': return 5;
 502         case '6': return 6;
 503         case '7': return 7;
 504         case '8': return 8;
 505         case '9': return 9;
 506         case 'a':
 507         case 'A': return 10;
 508         case 'b':
 509         case 'B': return 11;
 510         case 'c':
 511         case 'C': return 12;
 512         case 'd':
 513         case 'D': return 13;
 514         case 'e':
 515         case 'E': return 14;
 516         case 'f':
 517         case 'F': return 15;
 518         default:
 519                 internal_error("wrong character given");
 520         }
 521 }
 522
 523 /**
 524  * Parses an octal character sequence.
 525  *
 526  * @param first_digit  the already read first digit
 527  */
 528 static utf32 parse_octal_sequence(utf32 const first_digit)
 529 {
 530         assert(is_octal_digit(first_digit));
 531         utf32 value = digit_value(first_digit);
 532         if (!is_octal_digit(c)) return value;
 533         value = 8 * value + digit_value(c);
 534         next_char();
 535         if (!is_octal_digit(c)) return value;
 536         value = 8 * value + digit_value(c);
 537         next_char();
 538         return value;
 539 }
 540
 541 /**
 542  * Parses a hex character sequence.
 543  */
 544 static utf32 parse_hex_sequence(void)
 545 {
 546         utf32 value = 0;
 547         while (isxdigit(c)) {
 548                 value = 16 * value + digit_value(c);
 549                 next_char();
 550         }
 551         return value;
 552 }
 553
 554 /**
 555  * Parse an escape sequence.
 556  */
 557 static utf32 parse_escape_sequence(void)
 558 {
 559         eat('\\');
 560
 561         utf32 const ec = c;
 562         next_char();
 563
 564         switch (ec) {
 565         case '"':  return '"';
 566         case '\'': return '\'';
 567         case '\\': return '\\';
 568         case '?': return '\?';
 569         case 'a': return '\a';
 570         case 'b': return '\b';
 571         case 'f': return '\f';
 572         case 'n': return '\n';
 573         case 'r': return '\r';
 574         case 't': return '\t';
 575         case 'v': return '\v';
 576         case 'x':
 577                 return parse_hex_sequence();
 578         case '0':
 579         case '1':
 580         case '2':
 581         case '3':
 582         case '4':
 583         case '5':
 584         case '6':
 585         case '7':
 586                 return parse_octal_sequence(ec);
 587         case EOF:
 588                 parse_error("reached end of file while parsing escape sequence");
 589                 return EOF;
 590         /* \E is not documented, but handled, by GCC.  It is acceptable according
 591          * to §6.11.4, whereas \e is not. */
 592         case 'E':
 593         case 'e':
 594                 if (c_mode & _GNUC)
 595                         return 27;   /* hopefully 27 is ALWAYS the code for ESCAPE */
 596                 break;
 597         case 'u':
 598         case 'U':
 599                 parse_error("universal character parsing not implemented yet");
 600                 return EOF;
 601         default:
 602                 break;
 603         }
 604         /* §6.4.4.4:8 footnote 64 */
 605         parse_error("unknown escape sequence");
 606         return EOF;
 607 }
 608
 609 /**
 610  * Concatenate two strings.
 611  */
 612 string_t concat_strings(const string_t *const s1, const string_t *const s2)
 613 {
 614         const size_t len1 = s1->size - 1;
 615         const size_t len2 = s2->size - 1;
 616
 617         char *const concat = obstack_alloc(&symbol_obstack, len1 + len2 + 1);
 618         memcpy(concat, s1->begin, len1);
 619         memcpy(concat + len1, s2->begin, len2 + 1);
 620
 621         return identify_string(concat, len1 + len2 + 1);
 622 }
 623
 624 string_t make_string(const char *string)
 625 {
 626         size_t      len   = strlen(string) + 1;
 627         char *const space = obstack_alloc(&symbol_obstack, len);
 628         memcpy(space, string, len);
 629
 630         return identify_string(space, len);
 631 }
 632
 633 /**
 634  * Parse a string literal and set lexer_token.
 635  */
 636 static void parse_string_literal(void)
 637 {
 638         eat('"');
 639
 640         while (true) {
 641                 switch (c) {
 642                 case '\\': {
 643                         utf32 const tc = parse_escape_sequence();
 644                         if (tc >= 0x100) {
 645                                 warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 646                         }
 647                         obstack_1grow(&symbol_obstack, tc);
 648                         break;
 649                 }
 650
 651                 case EOF:
 652                         errorf(&lexer_token.base.source_position, "string has no end");
 653                         goto end_of_string;
 654
 655                 case '"':
 656                         next_char();
 657                         goto end_of_string;
 658
 659                 default:
 660                         obstack_grow_symbol(&symbol_obstack, c);
 661                         next_char();
 662                         break;
 663                 }
 664         }
 665
 666 end_of_string:
 667
 668         /* TODO: concatenate multiple strings separated by whitespace... */
 669
 670         /* add finishing 0 to the string */
 671         obstack_1grow(&symbol_obstack, '\0');
 672         const size_t  size   = (size_t)obstack_object_size(&symbol_obstack);
 673         char         *string = obstack_finish(&symbol_obstack);
 674
 675         lexer_token.kind          = T_STRING_LITERAL;
 676         lexer_token.string.string = identify_string(string, size);
 677 }
 678
 679 /**
 680  * Parse a wide character constant and set lexer_token.
 681  */
 682 static void parse_wide_character_constant(void)
 683 {
 684         eat('\'');
 685
 686         while (true) {
 687                 switch (c) {
 688                 case '\\': {
 689                         const utf32 tc = parse_escape_sequence();
 690                         obstack_grow_symbol(&symbol_obstack, tc);
 691                         break;
 692                 }
 693
 694                 MATCH_NEWLINE(
 695                         parse_error("newline while parsing character constant");
 696                         break;
 697                 )
 698
 699                 case '\'':
 700                         next_char();
 701                         goto end_of_wide_char_constant;
 702
 703                 case EOF:
 704                         errorf(&lexer_token.base.source_position, "EOF while parsing character constant");
 705                         goto end_of_wide_char_constant;
 706
 707                 default:
 708                         obstack_grow_symbol(&symbol_obstack, c);
 709                         next_char();
 710                         break;
 711                 }
 712         }
 713
 714 end_of_wide_char_constant:;
 715         obstack_1grow(&symbol_obstack, '\0');
 716         size_t  size   = (size_t) obstack_object_size(&symbol_obstack) - 1;
 717         char   *string = obstack_finish(&symbol_obstack);
 718
 719         lexer_token.kind          = T_WIDE_CHARACTER_CONSTANT;
 720         lexer_token.string.string = identify_string(string, size);
 721
 722         if (size == 0) {
 723                 errorf(&lexer_token.base.source_position, "empty character constant");
 724         }
 725 }
 726
 727 /**
 728  * Parse a wide string literal and set lexer_token.
 729  */
 730 static void parse_wide_string_literal(void)
 731 {
 732         parse_string_literal();
 733         if (lexer_token.kind == T_STRING_LITERAL)
 734                 lexer_token.kind = T_WIDE_STRING_LITERAL;
 735 }
 736
 737 /**
 738  * Parse a character constant and set lexer_token.
 739  */
 740 static void parse_character_constant(void)
 741 {
 742         eat('\'');
 743
 744         while (true) {
 745                 switch (c) {
 746                 case '\\': {
 747                         utf32 const tc = parse_escape_sequence();
 748                         if (tc >= 0x100) {
 749                                 warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 750                         }
 751                         obstack_1grow(&symbol_obstack, tc);
 752                         break;
 753                 }
 754
 755                 MATCH_NEWLINE(
 756                         parse_error("newline while parsing character constant");
 757                         break;
 758                 )
 759
 760                 case '\'':
 761                         next_char();
 762                         goto end_of_char_constant;
 763
 764                 case EOF:
 765                         errorf(&lexer_token.base.source_position, "EOF while parsing character constant");
 766                         goto end_of_char_constant;
 767
 768                 default:
 769                         obstack_grow_symbol(&symbol_obstack, c);
 770                         next_char();
 771                         break;
 772
 773                 }
 774         }
 775
 776 end_of_char_constant:;
 777         obstack_1grow(&symbol_obstack, '\0');
 778         const size_t        size   = (size_t)obstack_object_size(&symbol_obstack)-1;
 779         char         *const string = obstack_finish(&symbol_obstack);
 780
 781         lexer_token.kind          = T_CHARACTER_CONSTANT;
 782         lexer_token.string.string = identify_string(string, size);
 783
 784         if (size == 0) {
 785                 errorf(&lexer_token.base.source_position, "empty character constant");
 786         }
 787 }
 788
 789 /**
 790  * Skip a multiline comment.
 791  */
 792 static void skip_multiline_comment(void)
 793 {
 794         while (true) {
 795                 switch (c) {
 796                 case '/':
 797                         next_char();
 798                         if (c == '*') {
 799                                 /* nested comment, warn here */
 800                                 warningf(WARN_COMMENT, &lexer_pos, "'/*' within comment");
 801                         }
 802                         break;
 803                 case '*':
 804                         next_char();
 805                         if (c == '/') {
 806                                 next_char();
 807                                 return;
 808                         }
 809                         break;
 810
 811                 MATCH_NEWLINE(break;)
 812
 813                 case EOF: {
 814                         errorf(&lexer_token.base.source_position,
 815                                "at end of file while looking for comment end");
 816                         return;
 817                 }
 818
 819                 default:
 820                         next_char();
 821                         break;
 822                 }
 823         }
 824 }
 825
 826 /**
 827  * Skip a single line comment.
 828  */
 829 static void skip_line_comment(void)
 830 {
 831         while (true) {
 832                 switch (c) {
 833                 case EOF:
 834                         return;
 835
 836                 case '\n':
 837                 case '\r':
 838                         return;
 839
 840                 case '\\':
 841                         next_char();
 842                         if (c == '\n' || c == '\r') {
 843                                 warningf(WARN_COMMENT, &lexer_pos, "multi-line comment");
 844                                 return;
 845                         }
 846                         break;
 847
 848                 default:
 849                         next_char();
 850                         break;
 851                 }
 852         }
 853 }
 854
 855 /** The current preprocessor token. */
 856 static token_t pp_token;
 857
 858 /**
 859  * Read the next preprocessor token.
 860  */
 861 static inline void next_pp_token(void)
 862 {
 863         lexer_next_preprocessing_token();
 864         pp_token = lexer_token;
 865 }
 866
 867 /**
 868  * Eat all preprocessor tokens until newline.
 869  */
 870 static void eat_until_newline(void)
 871 {
 872         while (pp_token.kind != '\n' && pp_token.kind != T_EOF) {
 873                 next_pp_token();
 874         }
 875 }
 876
 877 /**
 878  * Parse the line directive.
 879  */
 880 static void parse_line_directive(void)
 881 {
 882         if (pp_token.kind != T_INTEGER) {
 883                 parse_error("expected integer");
 884         } else {
 885                 /* use offset -1 as this is about the next line */
 886                 lexer_pos.lineno = atoi(pp_token.number.number.begin) - 1;
 887                 next_pp_token();
 888         }
 889         if (pp_token.kind == T_STRING_LITERAL) {
 890                 lexer_pos.input_name = pp_token.string.string.begin;
 891                 lexer_pos.is_system_header = false;
 892                 next_pp_token();
 893
 894                 /* attempt to parse numeric flags as outputted by gcc preprocessor */
 895                 while (pp_token.kind == T_INTEGER) {
 896                         /* flags:
 897                          * 1 - indicates start of a new file
 898                          * 2 - indicates return from a file
 899                          * 3 - indicates system header
 900                          * 4 - indicates implicit extern "C" in C++ mode
 901                          *
 902                          * currently we're only interested in "3"
 903                          */
 904                         if (streq(pp_token.number.number.begin, "3")) {
 905                                 lexer_pos.is_system_header = true;
 906                         }
 907                         next_pp_token();
 908                 }
 909         }
 910
 911         eat_until_newline();
 912 }
 913
 914 /**
 915  * STDC pragmas.
 916  */
 917 typedef enum stdc_pragma_kind_t {
 918         STDC_UNKNOWN,
 919         STDC_FP_CONTRACT,
 920         STDC_FENV_ACCESS,
 921         STDC_CX_LIMITED_RANGE
 922 } stdc_pragma_kind_t;
 923
 924 /**
 925  * STDC pragma values.
 926  */
 927 typedef enum stdc_pragma_value_kind_t {
 928         STDC_VALUE_UNKNOWN,
 929         STDC_VALUE_ON,
 930         STDC_VALUE_OFF,
 931         STDC_VALUE_DEFAULT
 932 } stdc_pragma_value_kind_t;
 933
 934 /**
 935  * Parse a pragma directive.
 936  */
 937 static void parse_pragma(void)
 938 {
 939         bool unknown_pragma = true;
 940
 941         next_pp_token();
 942         if (pp_token.kind != T_IDENTIFIER) {
 943                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
 944                          "expected identifier after #pragma");
 945                 eat_until_newline();
 946                 return;
 947         }
 948
 949         symbol_t *symbol = pp_token.identifier.symbol;
 950         if (symbol->pp_ID == TP_STDC) {
 951                 stdc_pragma_kind_t kind = STDC_UNKNOWN;
 952                 /* a STDC pragma */
 953                 if (c_mode & _C99) {
 954                         next_pp_token();
 955
 956                         switch (pp_token.identifier.symbol->pp_ID) {
 957                         case TP_FP_CONTRACT:
 958                                 kind = STDC_FP_CONTRACT;
 959                                 break;
 960                         case TP_FENV_ACCESS:
 961                                 kind = STDC_FENV_ACCESS;
 962                                 break;
 963                         case TP_CX_LIMITED_RANGE:
 964                                 kind = STDC_CX_LIMITED_RANGE;
 965                                 break;
 966                         default:
 967                                 break;
 968                         }
 969                         if (kind != STDC_UNKNOWN) {
 970                                 stdc_pragma_value_kind_t value = STDC_VALUE_UNKNOWN;
 971                                 next_pp_token();
 972                                 switch (pp_token.identifier.symbol->pp_ID) {
 973                                 case TP_ON:
 974                                         value = STDC_VALUE_ON;
 975                                         break;
 976                                 case TP_OFF:
 977                                         value = STDC_VALUE_OFF;
 978                                         break;
 979                                 case TP_DEFAULT:
 980                                         value = STDC_VALUE_DEFAULT;
 981                                         break;
 982                                 default:
 983                                         break;
 984                                 }
 985                                 if (value != STDC_VALUE_UNKNOWN) {
 986                                         unknown_pragma = false;
 987                                 } else {
 988                                         errorf(&pp_token.base.source_position,
 989                                                "bad STDC pragma argument");
 990                                 }
 991                         }
 992                 }
 993         } else {
 994                 unknown_pragma = true;
 995         }
 996         eat_until_newline();
 997         if (unknown_pragma) {
 998                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
 999                          "encountered unknown #pragma");
1000         }
1001 }
1002
1003 /**
1004  * Parse a preprocessor non-null directive.
1005  */
1006 static void parse_preprocessor_identifier(void)
1007 {
1008         assert(pp_token.kind == T_IDENTIFIER);
1009         symbol_t *symbol = pp_token.identifier.symbol;
1010
1011         switch (symbol->pp_ID) {
1012         case TP_line:
1013                 next_pp_token();
1014                 parse_line_directive();
1015                 break;
1016         case TP_pragma:
1017                 parse_pragma();
1018                 break;
1019         case TP_error:
1020                 /* TODO; output the rest of the line */
1021                 parse_error("#error directive");
1022                 break;
1023         }
1024 }
1025
1026 /**
1027  * Parse a preprocessor directive.
1028  */
1029 static void parse_preprocessor_directive(void)
1030 {
1031         next_pp_token();
1032
1033         switch (pp_token.kind) {
1034         case T_IDENTIFIER:
1035                 parse_preprocessor_identifier();
1036                 break;
1037         case T_INTEGER:
1038                 parse_line_directive();
1039                 break;
1040         case '\n':
1041                 /* NULL directive, see §6.10.7 */
1042                 break;
1043         default:
1044                 parse_error("invalid preprocessor directive");
1045                 eat_until_newline();
1046                 break;
1047         }
1048 }
1049
1050 #define MAYBE_PROLOG                                       \
1051                         next_char();                                   \
1052                         while (true) {                                 \
1053                                 switch (c) {
1054
1055 #define MAYBE(ch, set_type)                                \
1056                                 case ch:                                   \
1057                                         next_char();                           \
1058                                         lexer_token.kind = set_type;           \
1059                                         return;
1060
1061 /* must use this as last thing */
1062 #define MAYBE_MODE(ch, set_type, mode)                     \
1063                                 case ch:                                   \
1064                                         if (c_mode & mode) {                   \
1065                                                 next_char();                       \
1066                                                 lexer_token.kind = set_type;       \
1067                                                 return;                            \
1068                                         }                                      \
1069                                         /* fallthrough */
1070
1071 #define ELSE_CODE(code)                                    \
1072                                 default:                                   \
1073                                         code                                   \
1074                                         return;                                \
1075                                 }                                          \
1076                         } /* end of while (true) */                    \
1077
1078 #define ELSE(set_type)                                     \
1079                 ELSE_CODE(                                         \
1080                         lexer_token.kind = set_type;                   \
1081                 )
1082
1083 void lexer_next_preprocessing_token(void)
1084 {
1085         while (true) {
1086                 lexer_token.base.source_position = lexer_pos;
1087
1088                 switch (c) {
1089                 case ' ':
1090                 case '\t':
1091                         next_char();
1092                         break;
1093
1094                 MATCH_NEWLINE(
1095                         lexer_token.kind = '\n';
1096                         return;
1097                 )
1098
1099                 SYMBOL_CHARS
1100                         parse_symbol();
1101                         /* might be a wide string ( L"string" ) */
1102                         if (lexer_token.identifier.symbol == symbol_L) {
1103                                 switch (c) {
1104                                         case '"':  parse_wide_string_literal();     break;
1105                                         case '\'': parse_wide_character_constant(); break;
1106                                 }
1107                         }
1108                         return;
1109
1110                 DIGITS
1111                         parse_number();
1112                         return;
1113
1114                 case '"':
1115                         parse_string_literal();
1116                         return;
1117
1118                 case '\'':
1119                         parse_character_constant();
1120                         return;
1121
1122                 case '.':
1123                         MAYBE_PROLOG
1124                                 DIGITS
1125                                         put_back(c);
1126                                         c = '.';
1127                                         parse_number();
1128                                         return;
1129
1130                                 case '.':
1131                                         MAYBE_PROLOG
1132                                         MAYBE('.', T_DOTDOTDOT)
1133                                         ELSE_CODE(
1134                                                 put_back(c);
1135                                                 c = '.';
1136                                                 lexer_token.kind = '.';
1137                                         )
1138                         ELSE('.')
1139                 case '&':
1140                         MAYBE_PROLOG
1141                         MAYBE('&', T_ANDAND)
1142                         MAYBE('=', T_ANDEQUAL)
1143                         ELSE('&')
1144                 case '*':
1145                         MAYBE_PROLOG
1146                         MAYBE('=', T_ASTERISKEQUAL)
1147                         ELSE('*')
1148                 case '+':
1149                         MAYBE_PROLOG
1150                         MAYBE('+', T_PLUSPLUS)
1151                         MAYBE('=', T_PLUSEQUAL)
1152                         ELSE('+')
1153                 case '-':
1154                         MAYBE_PROLOG
1155                         MAYBE('>', T_MINUSGREATER)
1156                         MAYBE('-', T_MINUSMINUS)
1157                         MAYBE('=', T_MINUSEQUAL)
1158                         ELSE('-')
1159                 case '!':
1160                         MAYBE_PROLOG
1161                         MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1162                         ELSE('!')
1163                 case '/':
1164                         MAYBE_PROLOG
1165                         MAYBE('=', T_SLASHEQUAL)
1166                                 case '*':
1167                                         next_char();
1168                                         skip_multiline_comment();
1169                                         lexer_next_preprocessing_token();
1170                                         return;
1171                                 case '/':
1172                                         next_char();
1173                                         skip_line_comment();
1174                                         lexer_next_preprocessing_token();
1175                                         return;
1176                         ELSE('/')
1177                 case '%':
1178                         MAYBE_PROLOG
1179                         MAYBE('>', '}')
1180                         MAYBE('=', T_PERCENTEQUAL)
1181                                 case ':':
1182                                         MAYBE_PROLOG
1183                                                 case '%':
1184                                                         MAYBE_PROLOG
1185                                                         MAYBE(':', T_HASHHASH)
1186                                                         ELSE_CODE(
1187                                                                 put_back(c);
1188                                                                 c = '%';
1189                                                                 lexer_token.kind = '#';
1190                                                         )
1191                                         ELSE('#')
1192                         ELSE('%')
1193                 case '<':
1194                         MAYBE_PROLOG
1195                         MAYBE(':', '[')
1196                         MAYBE('%', '{')
1197                         MAYBE('=', T_LESSEQUAL)
1198                                 case '<':
1199                                         MAYBE_PROLOG
1200                                         MAYBE('=', T_LESSLESSEQUAL)
1201                                         ELSE(T_LESSLESS)
1202                         ELSE('<')
1203                 case '>':
1204                         MAYBE_PROLOG
1205                         MAYBE('=', T_GREATEREQUAL)
1206                                 case '>':
1207                                         MAYBE_PROLOG
1208                                         MAYBE('=', T_GREATERGREATEREQUAL)
1209                                         ELSE(T_GREATERGREATER)
1210                         ELSE('>')
1211                 case '^':
1212                         MAYBE_PROLOG
1213                         MAYBE('=', T_CARETEQUAL)
1214                         ELSE('^')
1215                 case '|':
1216                         MAYBE_PROLOG
1217                         MAYBE('=', T_PIPEEQUAL)
1218                         MAYBE('|', T_PIPEPIPE)
1219                         ELSE('|')
1220                 case ':':
1221                         MAYBE_PROLOG
1222                         MAYBE('>', ']')
1223                         MAYBE_MODE(':', T_COLONCOLON, _CXX)
1224                         ELSE(':')
1225                 case '=':
1226                         MAYBE_PROLOG
1227                         MAYBE('=', T_EQUALEQUAL)
1228                         ELSE('=')
1229                 case '#':
1230                         MAYBE_PROLOG
1231                         MAYBE('#', T_HASHHASH)
1232                         ELSE('#')
1233
1234                 case '?':
1235                 case '[':
1236                 case ']':
1237                 case '(':
1238                 case ')':
1239                 case '{':
1240                 case '}':
1241                 case '~':
1242                 case ';':
1243                 case ',':
1244                 case '\\':
1245                         lexer_token.kind = c;
1246                         next_char();
1247                         return;
1248
1249                 case EOF:
1250                         lexer_token.kind = T_EOF;
1251                         return;
1252
1253                 default:
1254 dollar_sign:
1255                         errorf(&lexer_pos, "unknown character '%c' found", c);
1256                         next_char();
1257                         break;
1258                 }
1259         }
1260 }
1261
1262 void lexer_next_token(void)
1263 {
1264         lexer_next_preprocessing_token();
1265
1266         while (lexer_token.kind == '\n') {
1267 newline_found:
1268                 lexer_next_preprocessing_token();
1269         }
1270
1271         if (lexer_token.kind == '#') {
1272                 parse_preprocessor_directive();
1273                 goto newline_found;
1274         }
1275 }
1276
1277 void init_lexer(void)
1278 {
1279         strset_init(&stringset);
1280         symbol_L = symbol_table_insert("L");
1281 }
1282
1283 static void input_error(unsigned delta_lines, unsigned delta_cols,
1284                         const char *message)
1285 {
1286         lexer_pos.lineno += delta_lines;
1287         lexer_pos.colno  += delta_cols;
1288         errorf(&lexer_pos, "%s", message);
1289 }
1290
1291 void lexer_switch_input(input_t *new_input, const char *input_name)
1292 {
1293         lexer_pos.lineno     = 0;
1294         lexer_pos.colno      = 0;
1295         lexer_pos.input_name = input_name;
1296
1297         set_input_error_callback(input_error);
1298         input  = new_input;
1299         bufpos = NULL;
1300         bufend = NULL;
1301
1302         /* place a virtual \n at the beginning so the lexer knows that we're
1303          * at the beginning of a line */
1304         c = '\n';
1305 }
1306
1307 void exit_lexer(void)
1308 {
1309         strset_destroy(&stringset);
1310 }
1311
1312 static __attribute__((unused))
1313 void dbg_pos(const source_position_t source_position)
1314 {
1315         fprintf(stdout, "%s:%u:%u\n", source_position.input_name,
1316                 source_position.lineno, (unsigned)source_position.colno);
1317         fflush(stdout);
1318 }