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