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