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 /**
 333  * Parses a hex number including hex floats and set the
 334  * lexer_token.
 335  */
 336 static void parse_number_hex(void)
 337 {
 338         bool is_float   = false;
 339         bool has_digits = false;
 340
 341         assert(obstack_object_size(&symbol_obstack) == 0);
 342         while (isxdigit(c)) {
 343                 has_digits = true;
 344                 obstack_1grow(&symbol_obstack, (char) c);
 345                 next_char();
 346         }
 347
 348         if (c == '.') {
 349                 is_float = true;
 350                 obstack_1grow(&symbol_obstack, (char) c);
 351                 next_char();
 352
 353                 while (isxdigit(c)) {
 354                         has_digits = true;
 355                         obstack_1grow(&symbol_obstack, (char) c);
 356                         next_char();
 357                 }
 358         }
 359         if (c == 'p' || c == 'P') {
 360                 is_float = true;
 361                 obstack_1grow(&symbol_obstack, (char) c);
 362                 next_char();
 363
 364                 if (c == '-' || c == '+') {
 365                         obstack_1grow(&symbol_obstack, (char) c);
 366                         next_char();
 367                 }
 368
 369                 while (isxdigit(c)) {
 370                         obstack_1grow(&symbol_obstack, (char) c);
 371                         next_char();
 372                 }
 373         } else if (is_float) {
 374                 errorf(&lexer_token.base.source_position,
 375                        "hexadecimal floatingpoint constant requires an exponent");
 376         }
 377         obstack_1grow(&symbol_obstack, '\0');
 378
 379         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 380         char   *string = obstack_finish(&symbol_obstack);
 381         lexer_token.number.number = identify_string(string, size);
 382
 383         lexer_token.kind    =
 384                 is_float ? T_FLOATINGPOINT_HEXADECIMAL : T_INTEGER_HEXADECIMAL;
 385
 386         if (!has_digits) {
 387                 errorf(&lexer_token.base.source_position,
 388                        "invalid number literal '0x%S'", &lexer_token.number.number);
 389                 lexer_token.number.number.begin = "0";
 390                 lexer_token.number.number.size  = 1;
 391         }
 392
 393         parse_number_suffix();
 394 }
 395
 396 /**
 397  * Returns true if the given char is a octal digit.
 398  *
 399  * @param char  the character to check
 400  */
 401 static bool is_octal_digit(utf32 chr)
 402 {
 403         return '0' <= chr && chr <= '7';
 404 }
 405
 406 /**
 407  * Parses a number and sets the lexer_token.
 408  */
 409 static void parse_number(void)
 410 {
 411         bool is_float   = false;
 412         bool has_digits = false;
 413
 414         assert(obstack_object_size(&symbol_obstack) == 0);
 415         if (c == '0') {
 416                 next_char();
 417                 if (c == 'x' || c == 'X') {
 418                         next_char();
 419                         parse_number_hex();
 420                         return;
 421                 } else {
 422                         has_digits = true;
 423                 }
 424                 obstack_1grow(&symbol_obstack, '0');
 425         }
 426
 427         while (isdigit(c)) {
 428                 has_digits = true;
 429                 obstack_1grow(&symbol_obstack, (char) c);
 430                 next_char();
 431         }
 432
 433         if (c == '.') {
 434                 is_float = true;
 435                 obstack_1grow(&symbol_obstack, '.');
 436                 next_char();
 437
 438                 while (isdigit(c)) {
 439                         has_digits = true;
 440                         obstack_1grow(&symbol_obstack, (char) c);
 441                         next_char();
 442                 }
 443         }
 444         if (c == 'e' || c == 'E') {
 445                 is_float = true;
 446                 obstack_1grow(&symbol_obstack, 'e');
 447                 next_char();
 448
 449                 if (c == '-' || c == '+') {
 450                         obstack_1grow(&symbol_obstack, (char) c);
 451                         next_char();
 452                 }
 453
 454                 while (isdigit(c)) {
 455                         obstack_1grow(&symbol_obstack, (char) c);
 456                         next_char();
 457                 }
 458         }
 459
 460         obstack_1grow(&symbol_obstack, '\0');
 461         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 462         char   *string = obstack_finish(&symbol_obstack);
 463         lexer_token.number.number = identify_string(string, size);
 464
 465         /* is it an octal number? */
 466         if (is_float) {
 467                 lexer_token.kind = T_FLOATINGPOINT;
 468         } else if (string[0] == '0') {
 469                 lexer_token.kind = T_INTEGER_OCTAL;
 470
 471                 /* check for invalid octal digits */
 472                 for (size_t i= 0; i < size; ++i) {
 473                         char t = string[i];
 474                         if (t >= '8')
 475                                 errorf(&lexer_token.base.source_position,
 476                                        "invalid digit '%c' in octal number", t);
 477                 }
 478         } else {
 479                 lexer_token.kind = T_INTEGER;
 480         }
 481
 482         if (!has_digits) {
 483                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'",
 484                        &lexer_token.number.number);
 485         }
 486
 487         parse_number_suffix();
 488 }
 489
 490 /**
 491  * Returns the value of a digit.
 492  * The only portable way to do it ...
 493  */
 494 static int digit_value(utf32 const digit)
 495 {
 496         switch (digit) {
 497         case '0': return 0;
 498         case '1': return 1;
 499         case '2': return 2;
 500         case '3': return 3;
 501         case '4': return 4;
 502         case '5': return 5;
 503         case '6': return 6;
 504         case '7': return 7;
 505         case '8': return 8;
 506         case '9': return 9;
 507         case 'a':
 508         case 'A': return 10;
 509         case 'b':
 510         case 'B': return 11;
 511         case 'c':
 512         case 'C': return 12;
 513         case 'd':
 514         case 'D': return 13;
 515         case 'e':
 516         case 'E': return 14;
 517         case 'f':
 518         case 'F': return 15;
 519         default:
 520                 internal_error("wrong character given");
 521         }
 522 }
 523
 524 /**
 525  * Parses an octal character sequence.
 526  *
 527  * @param first_digit  the already read first digit
 528  */
 529 static utf32 parse_octal_sequence(utf32 const first_digit)
 530 {
 531         assert(is_octal_digit(first_digit));
 532         utf32 value = digit_value(first_digit);
 533         if (!is_octal_digit(c)) return value;
 534         value = 8 * value + digit_value(c);
 535         next_char();
 536         if (!is_octal_digit(c)) return value;
 537         value = 8 * value + digit_value(c);
 538         next_char();
 539         return value;
 540 }
 541
 542 /**
 543  * Parses a hex character sequence.
 544  */
 545 static utf32 parse_hex_sequence(void)
 546 {
 547         utf32 value = 0;
 548         while (isxdigit(c)) {
 549                 value = 16 * value + digit_value(c);
 550                 next_char();
 551         }
 552         return value;
 553 }
 554
 555 /**
 556  * Parse an escape sequence.
 557  */
 558 static utf32 parse_escape_sequence(void)
 559 {
 560         eat('\\');
 561
 562         utf32 const ec = c;
 563         next_char();
 564
 565         switch (ec) {
 566         case '"':  return '"';
 567         case '\'': return '\'';
 568         case '\\': return '\\';
 569         case '?': return '\?';
 570         case 'a': return '\a';
 571         case 'b': return '\b';
 572         case 'f': return '\f';
 573         case 'n': return '\n';
 574         case 'r': return '\r';
 575         case 't': return '\t';
 576         case 'v': return '\v';
 577         case 'x':
 578                 return parse_hex_sequence();
 579         case '0':
 580         case '1':
 581         case '2':
 582         case '3':
 583         case '4':
 584         case '5':
 585         case '6':
 586         case '7':
 587                 return parse_octal_sequence(ec);
 588         case EOF:
 589                 parse_error("reached end of file while parsing escape sequence");
 590                 return EOF;
 591         /* \E is not documented, but handled, by GCC.  It is acceptable according
 592          * to §6.11.4, whereas \e is not. */
 593         case 'E':
 594         case 'e':
 595                 if (c_mode & _GNUC)
 596                         return 27;   /* hopefully 27 is ALWAYS the code for ESCAPE */
 597                 break;
 598         case 'u':
 599         case 'U':
 600                 parse_error("universal character parsing not implemented yet");
 601                 return EOF;
 602         default:
 603                 break;
 604         }
 605         /* §6.4.4.4:8 footnote 64 */
 606         parse_error("unknown escape sequence");
 607         return EOF;
 608 }
 609
 610 /**
 611  * Concatenate two strings.
 612  */
 613 string_t concat_strings(const string_t *const s1, const string_t *const s2)
 614 {
 615         const size_t len1 = s1->size - 1;
 616         const size_t len2 = s2->size - 1;
 617
 618         char *const concat = obstack_alloc(&symbol_obstack, len1 + len2 + 1);
 619         memcpy(concat, s1->begin, len1);
 620         memcpy(concat + len1, s2->begin, len2 + 1);
 621
 622         return identify_string(concat, len1 + len2 + 1);
 623 }
 624
 625 string_t make_string(const char *string)
 626 {
 627         size_t      len   = strlen(string) + 1;
 628         char *const space = obstack_alloc(&symbol_obstack, len);
 629         memcpy(space, string, len);
 630
 631         return identify_string(space, len);
 632 }
 633
 634 /**
 635  * Parse a string literal and set lexer_token.
 636  */
 637 static void parse_string_literal(void)
 638 {
 639         eat('"');
 640
 641         while (true) {
 642                 switch (c) {
 643                 case '\\': {
 644                         utf32 const tc = parse_escape_sequence();
 645                         if (tc >= 0x100) {
 646                                 warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 647                         }
 648                         obstack_1grow(&symbol_obstack, tc);
 649                         break;
 650                 }
 651
 652                 case EOF:
 653                         errorf(&lexer_token.base.source_position, "string has no end");
 654                         goto end_of_string;
 655
 656                 case '"':
 657                         next_char();
 658                         goto end_of_string;
 659
 660                 default:
 661                         obstack_grow_symbol(&symbol_obstack, c);
 662                         next_char();
 663                         break;
 664                 }
 665         }
 666
 667 end_of_string:
 668
 669         /* TODO: concatenate multiple strings separated by whitespace... */
 670
 671         /* add finishing 0 to the string */
 672         obstack_1grow(&symbol_obstack, '\0');
 673         const size_t  size   = (size_t)obstack_object_size(&symbol_obstack);
 674         char         *string = obstack_finish(&symbol_obstack);
 675
 676         lexer_token.kind          = T_STRING_LITERAL;
 677         lexer_token.string.string = identify_string(string, size);
 678 }
 679
 680 /**
 681  * Parse a wide character constant and set lexer_token.
 682  */
 683 static void parse_wide_character_constant(void)
 684 {
 685         eat('\'');
 686
 687         while (true) {
 688                 switch (c) {
 689                 case '\\': {
 690                         const utf32 tc = parse_escape_sequence();
 691                         obstack_grow_symbol(&symbol_obstack, tc);
 692                         break;
 693                 }
 694
 695                 MATCH_NEWLINE(
 696                         parse_error("newline while parsing character constant");
 697                         break;
 698                 )
 699
 700                 case '\'':
 701                         next_char();
 702                         goto end_of_wide_char_constant;
 703
 704                 case EOF:
 705                         errorf(&lexer_token.base.source_position, "EOF while parsing character constant");
 706                         goto end_of_wide_char_constant;
 707
 708                 default:
 709                         obstack_grow_symbol(&symbol_obstack, c);
 710                         next_char();
 711                         break;
 712                 }
 713         }
 714
 715 end_of_wide_char_constant:;
 716         obstack_1grow(&symbol_obstack, '\0');
 717         size_t  size   = (size_t) obstack_object_size(&symbol_obstack) - 1;
 718         char   *string = obstack_finish(&symbol_obstack);
 719
 720         lexer_token.kind          = T_WIDE_CHARACTER_CONSTANT;
 721         lexer_token.string.string = identify_string(string, size);
 722
 723         if (size == 0) {
 724                 errorf(&lexer_token.base.source_position, "empty character constant");
 725         }
 726 }
 727
 728 /**
 729  * Parse a wide string literal and set lexer_token.
 730  */
 731 static void parse_wide_string_literal(void)
 732 {
 733         parse_string_literal();
 734         if (lexer_token.kind == T_STRING_LITERAL)
 735                 lexer_token.kind = T_WIDE_STRING_LITERAL;
 736 }
 737
 738 /**
 739  * Parse a character constant and set lexer_token.
 740  */
 741 static void parse_character_constant(void)
 742 {
 743         eat('\'');
 744
 745         while (true) {
 746                 switch (c) {
 747                 case '\\': {
 748                         utf32 const tc = parse_escape_sequence();
 749                         if (tc >= 0x100) {
 750                                 warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 751                         }
 752                         obstack_1grow(&symbol_obstack, tc);
 753                         break;
 754                 }
 755
 756                 MATCH_NEWLINE(
 757                         parse_error("newline while parsing character constant");
 758                         break;
 759                 )
 760
 761                 case '\'':
 762                         next_char();
 763                         goto end_of_char_constant;
 764
 765                 case EOF:
 766                         errorf(&lexer_token.base.source_position, "EOF while parsing character constant");
 767                         goto end_of_char_constant;
 768
 769                 default:
 770                         obstack_grow_symbol(&symbol_obstack, c);
 771                         next_char();
 772                         break;
 773
 774                 }
 775         }
 776
 777 end_of_char_constant:;
 778         obstack_1grow(&symbol_obstack, '\0');
 779         const size_t        size   = (size_t)obstack_object_size(&symbol_obstack)-1;
 780         char         *const string = obstack_finish(&symbol_obstack);
 781
 782         lexer_token.kind          = T_CHARACTER_CONSTANT;
 783         lexer_token.string.string = identify_string(string, size);
 784
 785         if (size == 0) {
 786                 errorf(&lexer_token.base.source_position, "empty character constant");
 787         }
 788 }
 789
 790 /**
 791  * Skip a multiline comment.
 792  */
 793 static void skip_multiline_comment(void)
 794 {
 795         while (true) {
 796                 switch (c) {
 797                 case '/':
 798                         next_char();
 799                         if (c == '*') {
 800                                 /* nested comment, warn here */
 801                                 warningf(WARN_COMMENT, &lexer_pos, "'/*' within comment");
 802                         }
 803                         break;
 804                 case '*':
 805                         next_char();
 806                         if (c == '/') {
 807                                 next_char();
 808                                 return;
 809                         }
 810                         break;
 811
 812                 MATCH_NEWLINE(break;)
 813
 814                 case EOF: {
 815                         errorf(&lexer_token.base.source_position,
 816                                "at end of file while looking for comment end");
 817                         return;
 818                 }
 819
 820                 default:
 821                         next_char();
 822                         break;
 823                 }
 824         }
 825 }
 826
 827 /**
 828  * Skip a single line comment.
 829  */
 830 static void skip_line_comment(void)
 831 {
 832         while (true) {
 833                 switch (c) {
 834                 case EOF:
 835                         return;
 836
 837                 case '\n':
 838                 case '\r':
 839                         return;
 840
 841                 case '\\':
 842                         next_char();
 843                         if (c == '\n' || c == '\r') {
 844                                 warningf(WARN_COMMENT, &lexer_pos, "multi-line comment");
 845                                 return;
 846                         }
 847                         break;
 848
 849                 default:
 850                         next_char();
 851                         break;
 852                 }
 853         }
 854 }
 855
 856 /** The current preprocessor token. */
 857 static token_t pp_token;
 858
 859 /**
 860  * Read the next preprocessor token.
 861  */
 862 static inline void next_pp_token(void)
 863 {
 864         lexer_next_preprocessing_token();
 865         pp_token = lexer_token;
 866 }
 867
 868 /**
 869  * Eat all preprocessor tokens until newline.
 870  */
 871 static void eat_until_newline(void)
 872 {
 873         while (pp_token.kind != '\n' && pp_token.kind != T_EOF) {
 874                 next_pp_token();
 875         }
 876 }
 877
 878 /**
 879  * Parse the line directive.
 880  */
 881 static void parse_line_directive(void)
 882 {
 883         if (pp_token.kind != T_INTEGER) {
 884                 parse_error("expected integer");
 885         } else {
 886                 /* use offset -1 as this is about the next line */
 887                 lexer_pos.lineno = atoi(pp_token.number.number.begin) - 1;
 888                 next_pp_token();
 889         }
 890         if (pp_token.kind == T_STRING_LITERAL) {
 891                 lexer_pos.input_name = pp_token.string.string.begin;
 892                 lexer_pos.is_system_header = false;
 893                 next_pp_token();
 894
 895                 /* attempt to parse numeric flags as outputted by gcc preprocessor */
 896                 while (pp_token.kind == T_INTEGER) {
 897                         /* flags:
 898                          * 1 - indicates start of a new file
 899                          * 2 - indicates return from a file
 900                          * 3 - indicates system header
 901                          * 4 - indicates implicit extern "C" in C++ mode
 902                          *
 903                          * currently we're only interested in "3"
 904                          */
 905                         if (streq(pp_token.number.number.begin, "3")) {
 906                                 lexer_pos.is_system_header = true;
 907                         }
 908                         next_pp_token();
 909                 }
 910         }
 911
 912         eat_until_newline();
 913 }
 914
 915 /**
 916  * STDC pragmas.
 917  */
 918 typedef enum stdc_pragma_kind_t {
 919         STDC_UNKNOWN,
 920         STDC_FP_CONTRACT,
 921         STDC_FENV_ACCESS,
 922         STDC_CX_LIMITED_RANGE
 923 } stdc_pragma_kind_t;
 924
 925 /**
 926  * STDC pragma values.
 927  */
 928 typedef enum stdc_pragma_value_kind_t {
 929         STDC_VALUE_UNKNOWN,
 930         STDC_VALUE_ON,
 931         STDC_VALUE_OFF,
 932         STDC_VALUE_DEFAULT
 933 } stdc_pragma_value_kind_t;
 934
 935 /**
 936  * Parse a pragma directive.
 937  */
 938 static void parse_pragma(void)
 939 {
 940         bool unknown_pragma = true;
 941
 942         next_pp_token();
 943         if (pp_token.kind != T_IDENTIFIER) {
 944                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
 945                          "expected identifier after #pragma");
 946                 eat_until_newline();
 947                 return;
 948         }
 949
 950         symbol_t *symbol = pp_token.identifier.symbol;
 951         if (symbol->pp_ID == TP_STDC) {
 952                 stdc_pragma_kind_t kind = STDC_UNKNOWN;
 953                 /* a STDC pragma */
 954                 if (c_mode & _C99) {
 955                         next_pp_token();
 956
 957                         switch (pp_token.identifier.symbol->pp_ID) {
 958                         case TP_FP_CONTRACT:
 959                                 kind = STDC_FP_CONTRACT;
 960                                 break;
 961                         case TP_FENV_ACCESS:
 962                                 kind = STDC_FENV_ACCESS;
 963                                 break;
 964                         case TP_CX_LIMITED_RANGE:
 965                                 kind = STDC_CX_LIMITED_RANGE;
 966                                 break;
 967                         default:
 968                                 break;
 969                         }
 970                         if (kind != STDC_UNKNOWN) {
 971                                 stdc_pragma_value_kind_t value = STDC_VALUE_UNKNOWN;
 972                                 next_pp_token();
 973                                 switch (pp_token.identifier.symbol->pp_ID) {
 974                                 case TP_ON:
 975                                         value = STDC_VALUE_ON;
 976                                         break;
 977                                 case TP_OFF:
 978                                         value = STDC_VALUE_OFF;
 979                                         break;
 980                                 case TP_DEFAULT:
 981                                         value = STDC_VALUE_DEFAULT;
 982                                         break;
 983                                 default:
 984                                         break;
 985                                 }
 986                                 if (value != STDC_VALUE_UNKNOWN) {
 987                                         unknown_pragma = false;
 988                                 } else {
 989                                         errorf(&pp_token.base.source_position,
 990                                                "bad STDC pragma argument");
 991                                 }
 992                         }
 993                 }
 994         } else {
 995                 unknown_pragma = true;
 996         }
 997         eat_until_newline();
 998         if (unknown_pragma) {
 999                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
1000                          "encountered unknown #pragma");
1001         }
1002 }
1003
1004 /**
1005  * Parse a preprocessor non-null directive.
1006  */
1007 static void parse_preprocessor_identifier(void)
1008 {
1009         assert(pp_token.kind == T_IDENTIFIER);
1010         symbol_t *symbol = pp_token.identifier.symbol;
1011
1012         switch (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
1089                 switch (c) {
1090                 case ' ':
1091                 case '\t':
1092                         next_char();
1093                         break;
1094
1095                 MATCH_NEWLINE(
1096                         lexer_token.kind = '\n';
1097                         return;
1098                 )
1099
1100                 SYMBOL_CHARS
1101                         parse_symbol();
1102                         /* might be a wide string ( L"string" ) */
1103                         if (lexer_token.identifier.symbol == symbol_L) {
1104                                 switch (c) {
1105                                         case '"':  parse_wide_string_literal();     break;
1106                                         case '\'': parse_wide_character_constant(); break;
1107                                 }
1108                         }
1109                         return;
1110
1111                 DIGITS
1112                         parse_number();
1113                         return;
1114
1115                 case '"':
1116                         parse_string_literal();
1117                         return;
1118
1119                 case '\'':
1120                         parse_character_constant();
1121                         return;
1122
1123                 case '.':
1124                         MAYBE_PROLOG
1125                                 DIGITS
1126                                         put_back(c);
1127                                         c = '.';
1128                                         parse_number();
1129                                         return;
1130
1131                                 case '.':
1132                                         MAYBE_PROLOG
1133                                         MAYBE('.', T_DOTDOTDOT)
1134                                         ELSE_CODE(
1135                                                 put_back(c);
1136                                                 c = '.';
1137                                                 lexer_token.kind = '.';
1138                                         )
1139                         ELSE('.')
1140                 case '&':
1141                         MAYBE_PROLOG
1142                         MAYBE('&', T_ANDAND)
1143                         MAYBE('=', T_ANDEQUAL)
1144                         ELSE('&')
1145                 case '*':
1146                         MAYBE_PROLOG
1147                         MAYBE('=', T_ASTERISKEQUAL)
1148                         ELSE('*')
1149                 case '+':
1150                         MAYBE_PROLOG
1151                         MAYBE('+', T_PLUSPLUS)
1152                         MAYBE('=', T_PLUSEQUAL)
1153                         ELSE('+')
1154                 case '-':
1155                         MAYBE_PROLOG
1156                         MAYBE('>', T_MINUSGREATER)
1157                         MAYBE('-', T_MINUSMINUS)
1158                         MAYBE('=', T_MINUSEQUAL)
1159                         ELSE('-')
1160                 case '!':
1161                         MAYBE_PROLOG
1162                         MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1163                         ELSE('!')
1164                 case '/':
1165                         MAYBE_PROLOG
1166                         MAYBE('=', T_SLASHEQUAL)
1167                                 case '*':
1168                                         next_char();
1169                                         skip_multiline_comment();
1170                                         lexer_next_preprocessing_token();
1171                                         return;
1172                                 case '/':
1173                                         next_char();
1174                                         skip_line_comment();
1175                                         lexer_next_preprocessing_token();
1176                                         return;
1177                         ELSE('/')
1178                 case '%':
1179                         MAYBE_PROLOG
1180                         MAYBE('>', '}')
1181                         MAYBE('=', T_PERCENTEQUAL)
1182                                 case ':':
1183                                         MAYBE_PROLOG
1184                                                 case '%':
1185                                                         MAYBE_PROLOG
1186                                                         MAYBE(':', T_HASHHASH)
1187                                                         ELSE_CODE(
1188                                                                 put_back(c);
1189                                                                 c = '%';
1190                                                                 lexer_token.kind = '#';
1191                                                         )
1192                                         ELSE('#')
1193                         ELSE('%')
1194                 case '<':
1195                         MAYBE_PROLOG
1196                         MAYBE(':', '[')
1197                         MAYBE('%', '{')
1198                         MAYBE('=', T_LESSEQUAL)
1199                                 case '<':
1200                                         MAYBE_PROLOG
1201                                         MAYBE('=', T_LESSLESSEQUAL)
1202                                         ELSE(T_LESSLESS)
1203                         ELSE('<')
1204                 case '>':
1205                         MAYBE_PROLOG
1206                         MAYBE('=', T_GREATEREQUAL)
1207                                 case '>':
1208                                         MAYBE_PROLOG
1209                                         MAYBE('=', T_GREATERGREATEREQUAL)
1210                                         ELSE(T_GREATERGREATER)
1211                         ELSE('>')
1212                 case '^':
1213                         MAYBE_PROLOG
1214                         MAYBE('=', T_CARETEQUAL)
1215                         ELSE('^')
1216                 case '|':
1217                         MAYBE_PROLOG
1218                         MAYBE('=', T_PIPEEQUAL)
1219                         MAYBE('|', T_PIPEPIPE)
1220                         ELSE('|')
1221                 case ':':
1222                         MAYBE_PROLOG
1223                         MAYBE('>', ']')
1224                         MAYBE_MODE(':', T_COLONCOLON, _CXX)
1225                         ELSE(':')
1226                 case '=':
1227                         MAYBE_PROLOG
1228                         MAYBE('=', T_EQUALEQUAL)
1229                         ELSE('=')
1230                 case '#':
1231                         MAYBE_PROLOG
1232                         MAYBE('#', T_HASHHASH)
1233                         ELSE('#')
1234
1235                 case '?':
1236                 case '[':
1237                 case ']':
1238                 case '(':
1239                 case ')':
1240                 case '{':
1241                 case '}':
1242                 case '~':
1243                 case ';':
1244                 case ',':
1245                 case '\\':
1246                         lexer_token.kind = c;
1247                         next_char();
1248                         return;
1249
1250                 case EOF:
1251                         lexer_token.kind = T_EOF;
1252                         return;
1253
1254                 default:
1255 dollar_sign:
1256                         errorf(&lexer_pos, "unknown character '%c' found", c);
1257                         next_char();
1258                         break;
1259                 }
1260         }
1261 }
1262
1263 void lexer_next_token(void)
1264 {
1265         lexer_next_preprocessing_token();
1266
1267         while (lexer_token.kind == '\n') {
1268 newline_found:
1269                 lexer_next_preprocessing_token();
1270         }
1271
1272         if (lexer_token.kind == '#') {
1273                 parse_preprocessor_directive();
1274                 goto newline_found;
1275         }
1276 }
1277
1278 void init_lexer(void)
1279 {
1280         strset_init(&stringset);
1281         symbol_L = symbol_table_insert("L");
1282 }
1283
1284 static void input_error(unsigned delta_lines, unsigned delta_cols,
1285                         const char *message)
1286 {
1287         lexer_pos.lineno += delta_lines;
1288         lexer_pos.colno  += delta_cols;
1289         errorf(&lexer_pos, "%s", message);
1290 }
1291
1292 void lexer_switch_input(input_t *new_input, const char *input_name)
1293 {
1294         lexer_pos.lineno     = 0;
1295         lexer_pos.colno      = 0;
1296         lexer_pos.input_name = input_name;
1297
1298         set_input_error_callback(input_error);
1299         input  = new_input;
1300         bufpos = NULL;
1301         bufend = NULL;
1302
1303         /* place a virtual \n at the beginning so the lexer knows that we're
1304          * at the beginning of a line */
1305         c = '\n';
1306 }
1307
1308 void exit_lexer(void)
1309 {
1310         strset_destroy(&stringset);
1311 }
1312
1313 static __attribute__((unused))
1314 void dbg_pos(const source_position_t source_position)
1315 {
1316         fprintf(stdout, "%s:%u:%u\n", source_position.input_name,
1317                 source_position.lineno, (unsigned)source_position.colno);
1318         fflush(stdout);
1319 }