nsz Git - cparser/blob - lexer.c

   1 /*
   2  * This file is part of cparser.
   3  * Copyright (C) 2007-2009 Matthias Braun <matze@braunis.de>
   4  *
   5  * This program is free software; you can redistribute it and/or
   6  * modify it under the terms of the GNU General Public License
   7  * as published by the Free Software Foundation; either version 2
   8  * of the License, or (at your option) any later version.
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License
  16  * along with this program; if not, write to the Free Software
  17  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
  18  * 02111-1307, USA.
  19  */
  20 #include <config.h>
  21
  22 #include "adt/strutil.h"
  23 #include "input.h"
  24 #include "diagnostic.h"
  25 #include "lexer.h"
  26 #include "symbol_t.h"
  27 #include "token_t.h"
  28 #include "symbol_table_t.h"
  29 #include "adt/error.h"
  30 #include "adt/strset.h"
  31 #include "adt/util.h"
  32 #include "types.h"
  33 #include "type_t.h"
  34 #include "parser.h"
  35 #include "warning.h"
  36 #include "lang_features.h"
  37
  38 #include <assert.h>
  39 #include <errno.h>
  40 #include <string.h>
  41 #include <stdbool.h>
  42 #include <ctype.h>
  43
  44 #ifndef _WIN32
  45 #include <strings.h>
  46 #endif
  47
  48 #define MAX_PUTBACK 16    // 3 would be enough, but 16 gives a nicer alignment
  49 #define BUF_SIZE    1024
  50
  51 static input_t           *input;
  52 static utf32              input_buf[BUF_SIZE + MAX_PUTBACK];
  53 static const utf32       *bufpos;
  54 static const utf32       *bufend;
  55 static utf32              c;
  56 static source_position_t  lexer_pos;
  57 token_t                   lexer_token;
  58 static symbol_t          *symbol_L;
  59 static strset_t           stringset;
  60 bool                      allow_dollar_in_symbol = true;
  61
  62 /**
  63  * Prints a parse error message at the current token.
  64  *
  65  * @param msg   the error message
  66  */
  67 static void parse_error(const char *msg)
  68 {
  69         errorf(&lexer_pos, "%s", msg);
  70 }
  71
  72 /**
  73  * Prints an internal error message at the current token.
  74  *
  75  * @param msg   the error message
  76  */
  77 static NORETURN internal_error(const char *msg)
  78 {
  79         internal_errorf(&lexer_pos, "%s", msg);
  80 }
  81
  82 static inline void next_real_char(void)
  83 {
  84         assert(bufpos <= bufend);
  85         if (bufpos >= bufend) {
  86                 size_t n = decode(input, input_buf+MAX_PUTBACK, BUF_SIZE);
  87                 if (n == 0) {
  88                         c = EOF;
  89                         return;
  90                 }
  91                 bufpos = input_buf + MAX_PUTBACK;
  92                 bufend = bufpos + n;
  93         }
  94         c = *bufpos++;
  95         ++lexer_pos.colno;
  96 }
  97
  98 /**
  99  * Put a character back into the buffer.
 100  *
 101  * @param pc  the character to put back
 102  */
 103 static inline void put_back(utf32 const pc)
 104 {
 105         *(--bufpos - input_buf + input_buf) = pc;
 106         --lexer_pos.colno;
 107 }
 108
 109 static inline void next_char(void);
 110
 111 #define MATCH_NEWLINE(code)  \
 112         case '\r':               \
 113                 next_char();         \
 114                 if (c == '\n') {     \
 115         case '\n':               \
 116                         next_char();     \
 117                 }                    \
 118                 lexer_pos.lineno++;  \
 119                 lexer_pos.colno = 1; \
 120                 code
 121
 122 #define eat(c_type) (assert(c == c_type), next_char())
 123
 124 static void maybe_concat_lines(void)
 125 {
 126         eat('\\');
 127
 128         switch (c) {
 129         MATCH_NEWLINE(return;)
 130
 131         default:
 132                 break;
 133         }
 134
 135         put_back(c);
 136         c = '\\';
 137 }
 138
 139 /**
 140  * Set c to the next input character, ie.
 141  * after expanding trigraphs.
 142  */
 143 static inline void next_char(void)
 144 {
 145         next_real_char();
 146
 147         /* filter trigraphs */
 148         if (UNLIKELY(c == '\\')) {
 149                 maybe_concat_lines();
 150                 return;
 151         }
 152
 153         if (LIKELY(c != '?'))
 154                 return;
 155
 156         next_real_char();
 157         if (LIKELY(c != '?')) {
 158                 put_back(c);
 159                 c = '?';
 160                 return;
 161         }
 162
 163         next_real_char();
 164         switch (c) {
 165         case '=': c = '#'; break;
 166         case '(': c = '['; break;
 167         case '/': c = '\\'; maybe_concat_lines(); break;
 168         case ')': c = ']'; break;
 169         case '\'': c = '^'; break;
 170         case '<': c = '{'; break;
 171         case '!': c = '|'; break;
 172         case '>': c = '}'; break;
 173         case '-': c = '~'; break;
 174         default:
 175                 put_back(c);
 176                 put_back('?');
 177                 c = '?';
 178                 break;
 179         }
 180 }
 181
 182 #define SYMBOL_CHARS  \
 183         case '$': if (!allow_dollar_in_symbol) goto dollar_sign; \
 184         case 'a':         \
 185         case 'b':         \
 186         case 'c':         \
 187         case 'd':         \
 188         case 'e':         \
 189         case 'f':         \
 190         case 'g':         \
 191         case 'h':         \
 192         case 'i':         \
 193         case 'j':         \
 194         case 'k':         \
 195         case 'l':         \
 196         case 'm':         \
 197         case 'n':         \
 198         case 'o':         \
 199         case 'p':         \
 200         case 'q':         \
 201         case 'r':         \
 202         case 's':         \
 203         case 't':         \
 204         case 'u':         \
 205         case 'v':         \
 206         case 'w':         \
 207         case 'x':         \
 208         case 'y':         \
 209         case 'z':         \
 210         case 'A':         \
 211         case 'B':         \
 212         case 'C':         \
 213         case 'D':         \
 214         case 'E':         \
 215         case 'F':         \
 216         case 'G':         \
 217         case 'H':         \
 218         case 'I':         \
 219         case 'J':         \
 220         case 'K':         \
 221         case 'L':         \
 222         case 'M':         \
 223         case 'N':         \
 224         case 'O':         \
 225         case 'P':         \
 226         case 'Q':         \
 227         case 'R':         \
 228         case 'S':         \
 229         case 'T':         \
 230         case 'U':         \
 231         case 'V':         \
 232         case 'W':         \
 233         case 'X':         \
 234         case 'Y':         \
 235         case 'Z':         \
 236         case '_':
 237
 238 #define DIGITS        \
 239         case '0':         \
 240         case '1':         \
 241         case '2':         \
 242         case '3':         \
 243         case '4':         \
 244         case '5':         \
 245         case '6':         \
 246         case '7':         \
 247         case '8':         \
 248         case '9':
 249
 250 /**
 251  * Read a symbol from the input and build
 252  * the lexer_token.
 253  */
 254 static void parse_symbol(void)
 255 {
 256         obstack_1grow(&symbol_obstack, (char) c);
 257         next_char();
 258
 259         while (true) {
 260                 switch (c) {
 261                 DIGITS
 262                 SYMBOL_CHARS
 263                         obstack_1grow(&symbol_obstack, (char) c);
 264                         next_char();
 265                         break;
 266
 267                 default:
 268 dollar_sign:
 269                         goto end_symbol;
 270                 }
 271         }
 272
 273 end_symbol:
 274         obstack_1grow(&symbol_obstack, '\0');
 275
 276         char     *string = obstack_finish(&symbol_obstack);
 277         symbol_t *symbol = symbol_table_insert(string);
 278
 279         lexer_token.kind        = symbol->ID;
 280         lexer_token.base.symbol = symbol;
 281
 282         if (symbol->string != string) {
 283                 obstack_free(&symbol_obstack, string);
 284         }
 285 }
 286
 287 static string_t identify_string(char *string, size_t len)
 288 {
 289         /* TODO hash */
 290 #if 0
 291         const char *result = strset_insert(&stringset, concat);
 292         if (result != concat) {
 293                 obstack_free(&symbol_obstack, concat);
 294         }
 295 #else
 296         const char *result = string;
 297 #endif
 298         return (string_t) {result, len};
 299 }
 300
 301 /**
 302  * parse suffixes like 'LU' or 'f' after numbers
 303  */
 304 static void parse_number_suffix(void)
 305 {
 306         assert(obstack_object_size(&symbol_obstack) == 0);
 307         while (true) {
 308                 switch (c) {
 309                 SYMBOL_CHARS
 310                         obstack_1grow(&symbol_obstack, (char) c);
 311                         next_char();
 312                         break;
 313                 default:
 314                 dollar_sign:
 315                         goto finish_suffix;
 316                 }
 317         }
 318 finish_suffix:
 319         if (obstack_object_size(&symbol_obstack) == 0) {
 320                 lexer_token.number.suffix.begin = NULL;
 321                 lexer_token.number.suffix.size  = 0;
 322                 return;
 323         }
 324
 325         obstack_1grow(&symbol_obstack, '\0');
 326         size_t size   = obstack_object_size(&symbol_obstack) - 1;
 327         char  *string = obstack_finish(&symbol_obstack);
 328
 329         lexer_token.number.suffix = identify_string(string, size);
 330 }
 331
 332 static void parse_exponent(void)
 333 {
 334         if (c == '-' || c == '+') {
 335                 obstack_1grow(&symbol_obstack, (char)c);
 336                 next_char();
 337         }
 338
 339         if (isdigit(c)) {
 340                 do {
 341                         obstack_1grow(&symbol_obstack, (char)c);
 342                         next_char();
 343                 } while (isdigit(c));
 344         } else {
 345                 errorf(&lexer_token.base.source_position, "exponent has no digits");
 346         }
 347 }
 348
 349 /**
 350  * Parses a hex number including hex floats and set the
 351  * lexer_token.
 352  */
 353 static void parse_number_hex(void)
 354 {
 355         bool is_float   = false;
 356         bool has_digits = false;
 357
 358         while (isxdigit(c)) {
 359                 has_digits = true;
 360                 obstack_1grow(&symbol_obstack, (char) c);
 361                 next_char();
 362         }
 363
 364         if (c == '.') {
 365                 is_float = true;
 366                 obstack_1grow(&symbol_obstack, (char) c);
 367                 next_char();
 368
 369                 while (isxdigit(c)) {
 370                         has_digits = true;
 371                         obstack_1grow(&symbol_obstack, (char) c);
 372                         next_char();
 373                 }
 374         }
 375         if (c == 'p' || c == 'P') {
 376                 is_float = true;
 377                 obstack_1grow(&symbol_obstack, (char) c);
 378                 next_char();
 379                 parse_exponent();
 380         } else if (is_float) {
 381                 errorf(&lexer_token.base.source_position,
 382                        "hexadecimal floatingpoint constant requires an exponent");
 383         }
 384         obstack_1grow(&symbol_obstack, '\0');
 385
 386         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 387         char   *string = obstack_finish(&symbol_obstack);
 388         lexer_token.number.number = identify_string(string, size);
 389
 390         lexer_token.kind = is_float ? T_FLOATINGPOINT : T_INTEGER;
 391
 392         if (!has_digits) {
 393                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'", &lexer_token.number.number);
 394                 lexer_token.number.number.begin = "0";
 395                 lexer_token.number.number.size  = 1;
 396         }
 397
 398         parse_number_suffix();
 399 }
 400
 401 static void parse_number_bin(void)
 402 {
 403         bool has_digits = false;
 404
 405         while (c == '0' || c == '1') {
 406                 has_digits = true;
 407                 obstack_1grow(&symbol_obstack, (char)c);
 408                 next_char();
 409         }
 410         obstack_1grow(&symbol_obstack, '\0');
 411
 412         size_t  const size   = obstack_object_size(&symbol_obstack) - 1;
 413         char   *const string = obstack_finish(&symbol_obstack);
 414         lexer_token.number.number = identify_string(string, size);
 415         lexer_token.kind          = T_INTEGER;
 416
 417         if (!has_digits) {
 418                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'", &lexer_token.number.number);
 419                 lexer_token.number.number.begin = "0";
 420                 lexer_token.number.number.size  = 1;
 421         }
 422
 423         parse_number_suffix();
 424 }
 425
 426 /**
 427  * Returns true if the given char is a octal digit.
 428  *
 429  * @param char  the character to check
 430  */
 431 static bool is_octal_digit(utf32 chr)
 432 {
 433         return '0' <= chr && chr <= '7';
 434 }
 435
 436 /**
 437  * Parses a number and sets the lexer_token.
 438  */
 439 static void parse_number(void)
 440 {
 441         bool is_float   = false;
 442         bool has_digits = false;
 443
 444         assert(obstack_object_size(&symbol_obstack) == 0);
 445         if (c == '0') {
 446                 obstack_1grow(&symbol_obstack, (char)c);
 447                 next_char();
 448                 if (c == 'x' || c == 'X') {
 449                         obstack_1grow(&symbol_obstack, (char)c);
 450                         next_char();
 451                         parse_number_hex();
 452                         return;
 453                 } else if (c == 'b' || c == 'B') {
 454                         /* GCC extension: binary constant 0x[bB][01]+.  */
 455                         obstack_1grow(&symbol_obstack, (char)c);
 456                         next_char();
 457                         parse_number_bin();
 458                         return;
 459                 }
 460                 has_digits = true;
 461         }
 462
 463         while (isdigit(c)) {
 464                 has_digits = true;
 465                 obstack_1grow(&symbol_obstack, (char) c);
 466                 next_char();
 467         }
 468
 469         if (c == '.') {
 470                 is_float = true;
 471                 obstack_1grow(&symbol_obstack, '.');
 472                 next_char();
 473
 474                 while (isdigit(c)) {
 475                         has_digits = true;
 476                         obstack_1grow(&symbol_obstack, (char) c);
 477                         next_char();
 478                 }
 479         }
 480         if (c == 'e' || c == 'E') {
 481                 is_float = true;
 482                 obstack_1grow(&symbol_obstack, 'e');
 483                 next_char();
 484                 parse_exponent();
 485         }
 486
 487         obstack_1grow(&symbol_obstack, '\0');
 488         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 489         char   *string = obstack_finish(&symbol_obstack);
 490         lexer_token.number.number = identify_string(string, size);
 491
 492         if (is_float) {
 493                 lexer_token.kind = T_FLOATINGPOINT;
 494         } else {
 495                 lexer_token.kind = T_INTEGER;
 496
 497                 if (string[0] == '0') {
 498                         /* check for invalid octal digits */
 499                         for (size_t i= 0; i < size; ++i) {
 500                                 char t = string[i];
 501                                 if (t >= '8')
 502                                         errorf(&lexer_token.base.source_position, "invalid digit '%c' in octal number", t);
 503                         }
 504                 }
 505         }
 506
 507         if (!has_digits) {
 508                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'",
 509                        &lexer_token.number.number);
 510         }
 511
 512         parse_number_suffix();
 513 }
 514
 515 /**
 516  * Returns the value of a digit.
 517  * The only portable way to do it ...
 518  */
 519 static int digit_value(utf32 const digit)
 520 {
 521         switch (digit) {
 522         case '0': return 0;
 523         case '1': return 1;
 524         case '2': return 2;
 525         case '3': return 3;
 526         case '4': return 4;
 527         case '5': return 5;
 528         case '6': return 6;
 529         case '7': return 7;
 530         case '8': return 8;
 531         case '9': return 9;
 532         case 'a':
 533         case 'A': return 10;
 534         case 'b':
 535         case 'B': return 11;
 536         case 'c':
 537         case 'C': return 12;
 538         case 'd':
 539         case 'D': return 13;
 540         case 'e':
 541         case 'E': return 14;
 542         case 'f':
 543         case 'F': return 15;
 544         default:
 545                 internal_error("wrong character given");
 546         }
 547 }
 548
 549 /**
 550  * Parses an octal character sequence.
 551  *
 552  * @param first_digit  the already read first digit
 553  */
 554 static utf32 parse_octal_sequence(utf32 const first_digit)
 555 {
 556         assert(is_octal_digit(first_digit));
 557         utf32 value = digit_value(first_digit);
 558         if (!is_octal_digit(c)) return value;
 559         value = 8 * value + digit_value(c);
 560         next_char();
 561         if (!is_octal_digit(c)) return value;
 562         value = 8 * value + digit_value(c);
 563         next_char();
 564         return value;
 565 }
 566
 567 /**
 568  * Parses a hex character sequence.
 569  */
 570 static utf32 parse_hex_sequence(void)
 571 {
 572         utf32 value = 0;
 573         while (isxdigit(c)) {
 574                 value = 16 * value + digit_value(c);
 575                 next_char();
 576         }
 577         return value;
 578 }
 579
 580 /**
 581  * Parse an escape sequence.
 582  */
 583 static utf32 parse_escape_sequence(void)
 584 {
 585         eat('\\');
 586
 587         utf32 const ec = c;
 588         next_char();
 589
 590         switch (ec) {
 591         case '"':  return '"';
 592         case '\'': return '\'';
 593         case '\\': return '\\';
 594         case '?': return '\?';
 595         case 'a': return '\a';
 596         case 'b': return '\b';
 597         case 'f': return '\f';
 598         case 'n': return '\n';
 599         case 'r': return '\r';
 600         case 't': return '\t';
 601         case 'v': return '\v';
 602         case 'x':
 603                 return parse_hex_sequence();
 604         case '0':
 605         case '1':
 606         case '2':
 607         case '3':
 608         case '4':
 609         case '5':
 610         case '6':
 611         case '7':
 612                 return parse_octal_sequence(ec);
 613         case EOF:
 614                 parse_error("reached end of file while parsing escape sequence");
 615                 return EOF;
 616         /* \E is not documented, but handled, by GCC.  It is acceptable according
 617          * to §6.11.4, whereas \e is not. */
 618         case 'E':
 619         case 'e':
 620                 if (c_mode & _GNUC)
 621                         return 27;   /* hopefully 27 is ALWAYS the code for ESCAPE */
 622                 break;
 623         case 'u':
 624         case 'U':
 625                 parse_error("universal character parsing not implemented yet");
 626                 return EOF;
 627         default:
 628                 break;
 629         }
 630         /* §6.4.4.4:8 footnote 64 */
 631         parse_error("unknown escape sequence");
 632         return EOF;
 633 }
 634
 635 /**
 636  * Concatenate two strings.
 637  */
 638 string_t concat_strings(const string_t *const s1, const string_t *const s2)
 639 {
 640         const size_t len1 = s1->size - 1;
 641         const size_t len2 = s2->size - 1;
 642
 643         char *const concat = obstack_alloc(&symbol_obstack, len1 + len2 + 1);
 644         memcpy(concat, s1->begin, len1);
 645         memcpy(concat + len1, s2->begin, len2 + 1);
 646
 647         return identify_string(concat, len1 + len2 + 1);
 648 }
 649
 650 string_t make_string(const char *string)
 651 {
 652         size_t      len   = strlen(string) + 1;
 653         char *const space = obstack_alloc(&symbol_obstack, len);
 654         memcpy(space, string, len);
 655
 656         return identify_string(space, len);
 657 }
 658
 659 /**
 660  * Parse a string literal and set lexer_token.
 661  */
 662 static void parse_string_literal(void)
 663 {
 664         eat('"');
 665
 666         while (true) {
 667                 switch (c) {
 668                 case '\\': {
 669                         utf32 const tc = parse_escape_sequence();
 670                         if (tc >= 0x100) {
 671                                 warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 672                         }
 673                         obstack_1grow(&symbol_obstack, tc);
 674                         break;
 675                 }
 676
 677                 case EOF:
 678                         errorf(&lexer_token.base.source_position, "string has no end");
 679                         goto end_of_string;
 680
 681                 case '"':
 682                         next_char();
 683                         goto end_of_string;
 684
 685                 default:
 686                         obstack_grow_symbol(&symbol_obstack, c);
 687                         next_char();
 688                         break;
 689                 }
 690         }
 691
 692 end_of_string:
 693
 694         /* TODO: concatenate multiple strings separated by whitespace... */
 695
 696         /* add finishing 0 to the string */
 697         obstack_1grow(&symbol_obstack, '\0');
 698         const size_t  size   = (size_t)obstack_object_size(&symbol_obstack);
 699         char         *string = obstack_finish(&symbol_obstack);
 700
 701         lexer_token.kind          = T_STRING_LITERAL;
 702         lexer_token.string.string = identify_string(string, size);
 703 }
 704
 705 /**
 706  * Parse a wide character constant and set lexer_token.
 707  */
 708 static void parse_wide_character_constant(void)
 709 {
 710         eat('\'');
 711
 712         while (true) {
 713                 switch (c) {
 714                 case '\\': {
 715                         const utf32 tc = parse_escape_sequence();
 716                         obstack_grow_symbol(&symbol_obstack, tc);
 717                         break;
 718                 }
 719
 720                 MATCH_NEWLINE(
 721                         parse_error("newline while parsing character constant");
 722                         break;
 723                 )
 724
 725                 case '\'':
 726                         next_char();
 727                         goto end_of_wide_char_constant;
 728
 729                 case EOF:
 730                         errorf(&lexer_token.base.source_position, "EOF while parsing character constant");
 731                         goto end_of_wide_char_constant;
 732
 733                 default:
 734                         obstack_grow_symbol(&symbol_obstack, c);
 735                         next_char();
 736                         break;
 737                 }
 738         }
 739
 740 end_of_wide_char_constant:;
 741         obstack_1grow(&symbol_obstack, '\0');
 742         size_t  size   = (size_t) obstack_object_size(&symbol_obstack) - 1;
 743         char   *string = obstack_finish(&symbol_obstack);
 744
 745         lexer_token.kind          = T_WIDE_CHARACTER_CONSTANT;
 746         lexer_token.string.string = identify_string(string, size);
 747
 748         if (size == 0) {
 749                 errorf(&lexer_token.base.source_position, "empty character constant");
 750         }
 751 }
 752
 753 /**
 754  * Parse a wide string literal and set lexer_token.
 755  */
 756 static void parse_wide_string_literal(void)
 757 {
 758         parse_string_literal();
 759         if (lexer_token.kind == T_STRING_LITERAL)
 760                 lexer_token.kind = T_WIDE_STRING_LITERAL;
 761 }
 762
 763 /**
 764  * Parse a character constant and set lexer_token.
 765  */
 766 static void parse_character_constant(void)
 767 {
 768         eat('\'');
 769
 770         while (true) {
 771                 switch (c) {
 772                 case '\\': {
 773                         utf32 const tc = parse_escape_sequence();
 774                         if (tc >= 0x100) {
 775                                 warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 776                         }
 777                         obstack_1grow(&symbol_obstack, tc);
 778                         break;
 779                 }
 780
 781                 MATCH_NEWLINE(
 782                         parse_error("newline while parsing character constant");
 783                         break;
 784                 )
 785
 786                 case '\'':
 787                         next_char();
 788                         goto end_of_char_constant;
 789
 790                 case EOF:
 791                         errorf(&lexer_token.base.source_position, "EOF while parsing character constant");
 792                         goto end_of_char_constant;
 793
 794                 default:
 795                         obstack_grow_symbol(&symbol_obstack, c);
 796                         next_char();
 797                         break;
 798
 799                 }
 800         }
 801
 802 end_of_char_constant:;
 803         obstack_1grow(&symbol_obstack, '\0');
 804         const size_t        size   = (size_t)obstack_object_size(&symbol_obstack)-1;
 805         char         *const string = obstack_finish(&symbol_obstack);
 806
 807         lexer_token.kind          = T_CHARACTER_CONSTANT;
 808         lexer_token.string.string = identify_string(string, size);
 809
 810         if (size == 0) {
 811                 errorf(&lexer_token.base.source_position, "empty character constant");
 812         }
 813 }
 814
 815 /**
 816  * Skip a multiline comment.
 817  */
 818 static void skip_multiline_comment(void)
 819 {
 820         while (true) {
 821                 switch (c) {
 822                 case '/':
 823                         next_char();
 824                         if (c == '*') {
 825                                 /* nested comment, warn here */
 826                                 warningf(WARN_COMMENT, &lexer_pos, "'/*' within comment");
 827                         }
 828                         break;
 829                 case '*':
 830                         next_char();
 831                         if (c == '/') {
 832                                 next_char();
 833                                 return;
 834                         }
 835                         break;
 836
 837                 MATCH_NEWLINE(break;)
 838
 839                 case EOF: {
 840                         errorf(&lexer_token.base.source_position,
 841                                "at end of file while looking for comment end");
 842                         return;
 843                 }
 844
 845                 default:
 846                         next_char();
 847                         break;
 848                 }
 849         }
 850 }
 851
 852 /**
 853  * Skip a single line comment.
 854  */
 855 static void skip_line_comment(void)
 856 {
 857         while (true) {
 858                 switch (c) {
 859                 case EOF:
 860                         return;
 861
 862                 case '\n':
 863                 case '\r':
 864                         return;
 865
 866                 case '\\':
 867                         next_char();
 868                         if (c == '\n' || c == '\r') {
 869                                 warningf(WARN_COMMENT, &lexer_pos, "multi-line comment");
 870                                 return;
 871                         }
 872                         break;
 873
 874                 default:
 875                         next_char();
 876                         break;
 877                 }
 878         }
 879 }
 880
 881 /** The current preprocessor token. */
 882 static token_t pp_token;
 883
 884 /**
 885  * Read the next preprocessor token.
 886  */
 887 static inline void next_pp_token(void)
 888 {
 889         lexer_next_preprocessing_token();
 890         pp_token = lexer_token;
 891 }
 892
 893 /**
 894  * Eat all preprocessor tokens until newline.
 895  */
 896 static void eat_until_newline(void)
 897 {
 898         while (pp_token.kind != '\n' && pp_token.kind != T_EOF) {
 899                 next_pp_token();
 900         }
 901 }
 902
 903 /**
 904  * Parse the line directive.
 905  */
 906 static void parse_line_directive(void)
 907 {
 908         if (pp_token.kind != T_INTEGER) {
 909                 parse_error("expected integer");
 910         } else {
 911                 /* use offset -1 as this is about the next line */
 912                 lexer_pos.lineno = atoi(pp_token.number.number.begin) - 1;
 913                 next_pp_token();
 914         }
 915         if (pp_token.kind == T_STRING_LITERAL) {
 916                 lexer_pos.input_name = pp_token.string.string.begin;
 917                 lexer_pos.is_system_header = false;
 918                 next_pp_token();
 919
 920                 /* attempt to parse numeric flags as outputted by gcc preprocessor */
 921                 while (pp_token.kind == T_INTEGER) {
 922                         /* flags:
 923                          * 1 - indicates start of a new file
 924                          * 2 - indicates return from a file
 925                          * 3 - indicates system header
 926                          * 4 - indicates implicit extern "C" in C++ mode
 927                          *
 928                          * currently we're only interested in "3"
 929                          */
 930                         if (streq(pp_token.number.number.begin, "3")) {
 931                                 lexer_pos.is_system_header = true;
 932                         }
 933                         next_pp_token();
 934                 }
 935         }
 936
 937         eat_until_newline();
 938 }
 939
 940 /**
 941  * STDC pragmas.
 942  */
 943 typedef enum stdc_pragma_kind_t {
 944         STDC_UNKNOWN,
 945         STDC_FP_CONTRACT,
 946         STDC_FENV_ACCESS,
 947         STDC_CX_LIMITED_RANGE
 948 } stdc_pragma_kind_t;
 949
 950 /**
 951  * STDC pragma values.
 952  */
 953 typedef enum stdc_pragma_value_kind_t {
 954         STDC_VALUE_UNKNOWN,
 955         STDC_VALUE_ON,
 956         STDC_VALUE_OFF,
 957         STDC_VALUE_DEFAULT
 958 } stdc_pragma_value_kind_t;
 959
 960 /**
 961  * Parse a pragma directive.
 962  */
 963 static void parse_pragma(void)
 964 {
 965         next_pp_token();
 966         if (pp_token.kind != T_IDENTIFIER) {
 967                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
 968                          "expected identifier after #pragma");
 969                 eat_until_newline();
 970                 return;
 971         }
 972
 973         stdc_pragma_kind_t kind = STDC_UNKNOWN;
 974         if (pp_token.base.symbol->pp_ID == TP_STDC && c_mode & _C99) {
 975                 /* a STDC pragma */
 976                 next_pp_token();
 977
 978                 switch (pp_token.base.symbol->pp_ID) {
 979                 case TP_FP_CONTRACT:      kind = STDC_FP_CONTRACT;      break;
 980                 case TP_FENV_ACCESS:      kind = STDC_FENV_ACCESS;      break;
 981                 case TP_CX_LIMITED_RANGE: kind = STDC_CX_LIMITED_RANGE; break;
 982                 default:                  break;
 983                 }
 984                 if (kind != STDC_UNKNOWN) {
 985                         next_pp_token();
 986                         stdc_pragma_value_kind_t value;
 987                         switch (pp_token.base.symbol->pp_ID) {
 988                         case TP_ON:      value = STDC_VALUE_ON;      break;
 989                         case TP_OFF:     value = STDC_VALUE_OFF;     break;
 990                         case TP_DEFAULT: value = STDC_VALUE_DEFAULT; break;
 991                         default:         value = STDC_VALUE_UNKNOWN; break;
 992                         }
 993                         if (value == STDC_VALUE_UNKNOWN) {
 994                                 kind = STDC_UNKNOWN;
 995                                 errorf(&pp_token.base.source_position, "bad STDC pragma argument");
 996                         }
 997                 }
 998         }
 999         eat_until_newline();
1000         if (kind == STDC_UNKNOWN) {
1001                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
1002                          "encountered unknown #pragma");
1003         }
1004 }
1005
1006 /**
1007  * Parse a preprocessor non-null directive.
1008  */
1009 static void parse_preprocessor_identifier(void)
1010 {
1011         assert(pp_token.kind == T_IDENTIFIER);
1012         switch (pp_token.base.symbol->pp_ID) {
1013         case TP_line:
1014                 next_pp_token();
1015                 parse_line_directive();
1016                 break;
1017         case TP_pragma:
1018                 parse_pragma();
1019                 break;
1020         case TP_error:
1021                 /* TODO; output the rest of the line */
1022                 parse_error("#error directive");
1023                 break;
1024         }
1025 }
1026
1027 /**
1028  * Parse a preprocessor directive.
1029  */
1030 static void parse_preprocessor_directive(void)
1031 {
1032         next_pp_token();
1033
1034         switch (pp_token.kind) {
1035         case T_IDENTIFIER:
1036                 parse_preprocessor_identifier();
1037                 break;
1038         case T_INTEGER:
1039                 parse_line_directive();
1040                 break;
1041         case '\n':
1042                 /* NULL directive, see §6.10.7 */
1043                 break;
1044         default:
1045                 parse_error("invalid preprocessor directive");
1046                 eat_until_newline();
1047                 break;
1048         }
1049 }
1050
1051 #define MAYBE_PROLOG                                       \
1052                         next_char();                                   \
1053                         while (true) {                                 \
1054                                 switch (c) {
1055
1056 #define MAYBE(ch, set_type)                                \
1057                                 case ch:                                   \
1058                                         next_char();                           \
1059                                         lexer_token.kind = set_type;           \
1060                                         return;
1061
1062 /* must use this as last thing */
1063 #define MAYBE_MODE(ch, set_type, mode)                     \
1064                                 case ch:                                   \
1065                                         if (c_mode & mode) {                   \
1066                                                 next_char();                       \
1067                                                 lexer_token.kind = set_type;       \
1068                                                 return;                            \
1069                                         }                                      \
1070                                         /* fallthrough */
1071
1072 #define ELSE_CODE(code)                                    \
1073                                 default:                                   \
1074                                         code                                   \
1075                                         return;                                \
1076                                 }                                          \
1077                         } /* end of while (true) */                    \
1078
1079 #define ELSE(set_type)                                     \
1080                 ELSE_CODE(                                         \
1081                         lexer_token.kind = set_type;                   \
1082                 )
1083
1084 void lexer_next_preprocessing_token(void)
1085 {
1086         while (true) {
1087                 lexer_token.base.source_position = lexer_pos;
1088                 lexer_token.base.symbol          = NULL;
1089
1090                 switch (c) {
1091                 case ' ':
1092                 case '\t':
1093                         next_char();
1094                         break;
1095
1096                 MATCH_NEWLINE(
1097                         lexer_token.kind = '\n';
1098                         return;
1099                 )
1100
1101                 SYMBOL_CHARS
1102                         parse_symbol();
1103                         /* might be a wide string ( L"string" ) */
1104                         if (lexer_token.base.symbol == symbol_L) {
1105                                 switch (c) {
1106                                         case '"':  parse_wide_string_literal();     break;
1107                                         case '\'': parse_wide_character_constant(); break;
1108                                 }
1109                         }
1110                         return;
1111
1112                 DIGITS
1113                         parse_number();
1114                         return;
1115
1116                 case '"':
1117                         parse_string_literal();
1118                         return;
1119
1120                 case '\'':
1121                         parse_character_constant();
1122                         return;
1123
1124                 case '.':
1125                         MAYBE_PROLOG
1126                                 DIGITS
1127                                         put_back(c);
1128                                         c = '.';
1129                                         parse_number();
1130                                         return;
1131
1132                                 case '.':
1133                                         MAYBE_PROLOG
1134                                         MAYBE('.', T_DOTDOTDOT)
1135                                         ELSE_CODE(
1136                                                 put_back(c);
1137                                                 c = '.';
1138                                                 lexer_token.kind = '.';
1139                                         )
1140                         ELSE('.')
1141                 case '&':
1142                         MAYBE_PROLOG
1143                         MAYBE('&', T_ANDAND)
1144                         MAYBE('=', T_ANDEQUAL)
1145                         ELSE('&')
1146                 case '*':
1147                         MAYBE_PROLOG
1148                         MAYBE('=', T_ASTERISKEQUAL)
1149                         ELSE('*')
1150                 case '+':
1151                         MAYBE_PROLOG
1152                         MAYBE('+', T_PLUSPLUS)
1153                         MAYBE('=', T_PLUSEQUAL)
1154                         ELSE('+')
1155                 case '-':
1156                         MAYBE_PROLOG
1157                         MAYBE('>', T_MINUSGREATER)
1158                         MAYBE('-', T_MINUSMINUS)
1159                         MAYBE('=', T_MINUSEQUAL)
1160                         ELSE('-')
1161                 case '!':
1162                         MAYBE_PROLOG
1163                         MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1164                         ELSE('!')
1165                 case '/':
1166                         MAYBE_PROLOG
1167                         MAYBE('=', T_SLASHEQUAL)
1168                                 case '*':
1169                                         next_char();
1170                                         skip_multiline_comment();
1171                                         lexer_next_preprocessing_token();
1172                                         return;
1173                                 case '/':
1174                                         next_char();
1175                                         skip_line_comment();
1176                                         lexer_next_preprocessing_token();
1177                                         return;
1178                         ELSE('/')
1179                 case '%':
1180                         MAYBE_PROLOG
1181                         MAYBE('>', '}')
1182                         MAYBE('=', T_PERCENTEQUAL)
1183                                 case ':':
1184                                         MAYBE_PROLOG
1185                                                 case '%':
1186                                                         MAYBE_PROLOG
1187                                                         MAYBE(':', T_HASHHASH)
1188                                                         ELSE_CODE(
1189                                                                 put_back(c);
1190                                                                 c = '%';
1191                                                                 lexer_token.kind = '#';
1192                                                         )
1193                                         ELSE('#')
1194                         ELSE('%')
1195                 case '<':
1196                         MAYBE_PROLOG
1197                         MAYBE(':', '[')
1198                         MAYBE('%', '{')
1199                         MAYBE('=', T_LESSEQUAL)
1200                                 case '<':
1201                                         MAYBE_PROLOG
1202                                         MAYBE('=', T_LESSLESSEQUAL)
1203                                         ELSE(T_LESSLESS)
1204                         ELSE('<')
1205                 case '>':
1206                         MAYBE_PROLOG
1207                         MAYBE('=', T_GREATEREQUAL)
1208                                 case '>':
1209                                         MAYBE_PROLOG
1210                                         MAYBE('=', T_GREATERGREATEREQUAL)
1211                                         ELSE(T_GREATERGREATER)
1212                         ELSE('>')
1213                 case '^':
1214                         MAYBE_PROLOG
1215                         MAYBE('=', T_CARETEQUAL)
1216                         ELSE('^')
1217                 case '|':
1218                         MAYBE_PROLOG
1219                         MAYBE('=', T_PIPEEQUAL)
1220                         MAYBE('|', T_PIPEPIPE)
1221                         ELSE('|')
1222                 case ':':
1223                         MAYBE_PROLOG
1224                         MAYBE('>', ']')
1225                         MAYBE_MODE(':', T_COLONCOLON, _CXX)
1226                         ELSE(':')
1227                 case '=':
1228                         MAYBE_PROLOG
1229                         MAYBE('=', T_EQUALEQUAL)
1230                         ELSE('=')
1231                 case '#':
1232                         MAYBE_PROLOG
1233                         MAYBE('#', T_HASHHASH)
1234                         ELSE('#')
1235
1236                 case '?':
1237                 case '[':
1238                 case ']':
1239                 case '(':
1240                 case ')':
1241                 case '{':
1242                 case '}':
1243                 case '~':
1244                 case ';':
1245                 case ',':
1246                 case '\\':
1247                         lexer_token.kind = c;
1248                         next_char();
1249                         return;
1250
1251                 case EOF:
1252                         lexer_token.kind = T_EOF;
1253                         return;
1254
1255                 default:
1256 dollar_sign:
1257                         errorf(&lexer_pos, "unknown character '%c' found", c);
1258                         next_char();
1259                         break;
1260                 }
1261         }
1262 }
1263
1264 void lexer_next_token(void)
1265 {
1266         lexer_next_preprocessing_token();
1267
1268         while (lexer_token.kind == '\n') {
1269 newline_found:
1270                 lexer_next_preprocessing_token();
1271         }
1272
1273         if (lexer_token.kind == '#') {
1274                 parse_preprocessor_directive();
1275                 goto newline_found;
1276         }
1277 }
1278
1279 void init_lexer(void)
1280 {
1281         strset_init(&stringset);
1282         symbol_L = symbol_table_insert("L");
1283 }
1284
1285 static void input_error(unsigned delta_lines, unsigned delta_cols,
1286                         const char *message)
1287 {
1288         lexer_pos.lineno += delta_lines;
1289         lexer_pos.colno  += delta_cols;
1290         errorf(&lexer_pos, "%s", message);
1291 }
1292
1293 void lexer_switch_input(input_t *new_input, const char *input_name)
1294 {
1295         lexer_pos.lineno     = 0;
1296         lexer_pos.colno      = 0;
1297         lexer_pos.input_name = input_name;
1298
1299         set_input_error_callback(input_error);
1300         input  = new_input;
1301         bufpos = NULL;
1302         bufend = NULL;
1303
1304         /* place a virtual \n at the beginning so the lexer knows that we're
1305          * at the beginning of a line */
1306         c = '\n';
1307 }
1308
1309 void exit_lexer(void)
1310 {
1311         strset_destroy(&stringset);
1312 }
1313
1314 static __attribute__((unused))
1315 void dbg_pos(const source_position_t source_position)
1316 {
1317         fprintf(stdout, "%s:%u:%u\n", source_position.input_name,
1318                 source_position.lineno, (unsigned)source_position.colno);
1319         fflush(stdout);
1320 }