nsz Git - cparser/blob - lexer.c

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