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 NEWLINE  \
 112         '\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                 goto newline; \
 121                 newline // Let it look like an ordinary case label.
 122
 123 #define eat(c_type) (assert(c == c_type), next_char())
 124
 125 static void maybe_concat_lines(void)
 126 {
 127         eat('\\');
 128
 129         switch (c) {
 130         case NEWLINE:
 131                 return;
 132
 133         default:
 134                 break;
 135         }
 136
 137         put_back(c);
 138         c = '\\';
 139 }
 140
 141 /**
 142  * Set c to the next input character, ie.
 143  * after expanding trigraphs.
 144  */
 145 static inline void next_char(void)
 146 {
 147         next_real_char();
 148
 149         /* filter trigraphs */
 150         if (UNLIKELY(c == '\\')) {
 151                 maybe_concat_lines();
 152                 return;
 153         }
 154
 155         if (LIKELY(c != '?'))
 156                 return;
 157
 158         next_real_char();
 159         if (LIKELY(c != '?')) {
 160                 put_back(c);
 161                 c = '?';
 162                 return;
 163         }
 164
 165         next_real_char();
 166         switch (c) {
 167         case '=': c = '#'; break;
 168         case '(': c = '['; break;
 169         case '/': c = '\\'; maybe_concat_lines(); break;
 170         case ')': c = ']'; break;
 171         case '\'': c = '^'; break;
 172         case '<': c = '{'; break;
 173         case '!': c = '|'; break;
 174         case '>': c = '}'; break;
 175         case '-': c = '~'; break;
 176         default:
 177                 put_back(c);
 178                 put_back('?');
 179                 c = '?';
 180                 break;
 181         }
 182 }
 183
 184 #define SYMBOL_CHARS_WITHOUT_E_P \
 185              '$': if (!allow_dollar_in_symbol) goto dollar_sign; \
 186         case 'a':         \
 187         case 'b':         \
 188         case 'c':         \
 189         case 'd':         \
 190         case 'f':         \
 191         case 'g':         \
 192         case 'h':         \
 193         case 'i':         \
 194         case 'j':         \
 195         case 'k':         \
 196         case 'l':         \
 197         case 'm':         \
 198         case 'n':         \
 199         case 'o':         \
 200         case '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 'F':         \
 215         case 'G':         \
 216         case 'H':         \
 217         case 'I':         \
 218         case 'J':         \
 219         case 'K':         \
 220         case 'L':         \
 221         case 'M':         \
 222         case 'N':         \
 223         case 'O':         \
 224         case 'Q':         \
 225         case 'R':         \
 226         case 'S':         \
 227         case 'T':         \
 228         case 'U':         \
 229         case 'V':         \
 230         case 'W':         \
 231         case 'X':         \
 232         case 'Y':         \
 233         case 'Z':         \
 234         case '_'
 235
 236 #define SYMBOL_CHARS_E_P \
 237              'E': \
 238         case 'P': \
 239         case 'e': \
 240         case 'p'
 241
 242 #define SYMBOL_CHARS  \
 243              SYMBOL_CHARS_WITHOUT_E_P: \
 244         case SYMBOL_CHARS_E_P
 245
 246 #define DIGITS        \
 247              '0':         \
 248         case '1':         \
 249         case '2':         \
 250         case '3':         \
 251         case '4':         \
 252         case '5':         \
 253         case '6':         \
 254         case '7':         \
 255         case '8':         \
 256         case '9'
 257
 258 static bool is_universal_char_valid(utf32 const v)
 259 {
 260         /* C11 §6.4.3:2 */
 261         if (v < 0xA0U && v != 0x24 && v != 0x40 && v != 0x60)
 262                 return false;
 263         if (0xD800 <= v && v <= 0xDFFF)
 264                 return false;
 265         return true;
 266 }
 267
 268 static int digit_value(utf32 digit);
 269
 270 static utf32 parse_universal_char(unsigned const n_digits)
 271 {
 272         utf32 v = 0;
 273         for (unsigned k = n_digits; k != 0; --k) {
 274                 if (isxdigit(c)) {
 275                         v = 16 * v + digit_value(c);
 276                         next_char();
 277                 } else {
 278                         errorf(&lexer_pos, "short universal character name, expected %u more digits", k);
 279                         break;
 280                 }
 281         }
 282         if (!is_universal_char_valid(v)) {
 283                 errorf(&lexer_pos, "\\%c%0*X is not a valid universal character name", n_digits == 4 ? 'u' : 'U', (int)n_digits, v);
 284         }
 285         return v;
 286 }
 287
 288 static bool is_universal_char_valid_identifier(utf32 const v)
 289 {
 290         /* C11 Annex D.1 */
 291         if (                v == 0x000A8) return true;
 292         if (                v == 0x000AA) return true;
 293         if (                v == 0x000AD) return true;
 294         if (                v == 0x000AF) return true;
 295         if (0x000B2 <= v && v <= 0x000B5) return true;
 296         if (0x000B7 <= v && v <= 0x000BA) return true;
 297         if (0x000BC <= v && v <= 0x000BE) return true;
 298         if (0x000C0 <= v && v <= 0x000D6) return true;
 299         if (0x000D8 <= v && v <= 0x000F6) return true;
 300         if (0x000F8 <= v && v <= 0x000FF) return true;
 301         if (0x00100 <= v && v <= 0x0167F) return true;
 302         if (0x01681 <= v && v <= 0x0180D) return true;
 303         if (0x0180F <= v && v <= 0x01FFF) return true;
 304         if (0x0200B <= v && v <= 0x0200D) return true;
 305         if (0x0202A <= v && v <= 0x0202E) return true;
 306         if (0x0203F <= v && v <= 0x02040) return true;
 307         if (                v == 0x02054) return true;
 308         if (0x02060 <= v && v <= 0x0206F) return true;
 309         if (0x02070 <= v && v <= 0x0218F) return true;
 310         if (0x02460 <= v && v <= 0x024FF) return true;
 311         if (0x02776 <= v && v <= 0x02793) return true;
 312         if (0x02C00 <= v && v <= 0x02DFF) return true;
 313         if (0x02E80 <= v && v <= 0x02FFF) return true;
 314         if (0x03004 <= v && v <= 0x03007) return true;
 315         if (0x03021 <= v && v <= 0x0302F) return true;
 316         if (0x03031 <= v && v <= 0x0303F) return true;
 317         if (0x03040 <= v && v <= 0x0D7FF) return true;
 318         if (0x0F900 <= v && v <= 0x0FD3D) return true;
 319         if (0x0FD40 <= v && v <= 0x0FDCF) return true;
 320         if (0x0FDF0 <= v && v <= 0x0FE44) return true;
 321         if (0x0FE47 <= v && v <= 0x0FFFD) return true;
 322         if (0x10000 <= v && v <= 0x1FFFD) return true;
 323         if (0x20000 <= v && v <= 0x2FFFD) return true;
 324         if (0x30000 <= v && v <= 0x3FFFD) return true;
 325         if (0x40000 <= v && v <= 0x4FFFD) return true;
 326         if (0x50000 <= v && v <= 0x5FFFD) return true;
 327         if (0x60000 <= v && v <= 0x6FFFD) return true;
 328         if (0x70000 <= v && v <= 0x7FFFD) return true;
 329         if (0x80000 <= v && v <= 0x8FFFD) return true;
 330         if (0x90000 <= v && v <= 0x9FFFD) return true;
 331         if (0xA0000 <= v && v <= 0xAFFFD) return true;
 332         if (0xB0000 <= v && v <= 0xBFFFD) return true;
 333         if (0xC0000 <= v && v <= 0xCFFFD) return true;
 334         if (0xD0000 <= v && v <= 0xDFFFD) return true;
 335         if (0xE0000 <= v && v <= 0xEFFFD) return true;
 336         return false;
 337 }
 338
 339 static bool is_universal_char_valid_identifier_start(utf32 const v)
 340 {
 341         /* C11 Annex D.2 */
 342         if (0x0300 <= v && v <= 0x036F) return false;
 343         if (0x1DC0 <= v && v <= 0x1DFF) return false;
 344         if (0x20D0 <= v && v <= 0x20FF) return false;
 345         if (0xFE20 <= v && v <= 0xFE2F) return false;
 346         return true;
 347 }
 348
 349 /**
 350  * Read a symbol from the input and build
 351  * the lexer_token.
 352  */
 353 static void parse_symbol(void)
 354 {
 355         while (true) {
 356                 switch (c) {
 357                 case DIGITS:
 358                 case SYMBOL_CHARS:
 359                         obstack_1grow(&symbol_obstack, (char) c);
 360                         next_char();
 361                         break;
 362
 363                 case '\\':
 364                         next_char();
 365                         switch (c) {
 366                         {
 367                                 unsigned n;
 368                         case 'U': n = 8; goto universal;
 369                         case 'u': n = 4; goto universal;
 370 universal:
 371                                 next_char();
 372                                 utf32 const v = parse_universal_char(n);
 373                                 if (!is_universal_char_valid_identifier(v)) {
 374                                         if (is_universal_char_valid(v)) {
 375                                                 errorf(&lexer_pos, "universal character \\%c%0*X is not valid in an identifier", n == 4 ? 'u' : 'U', (int)n, v);
 376                                         }
 377                                 } else if (obstack_object_size(&symbol_obstack) == 0 && !is_universal_char_valid_identifier_start(v)) {
 378                                         errorf(&lexer_pos, "universal character \\%c%0*X is not valid as start of an identifier", n == 4 ? 'u' : 'U', (int)n, v);
 379                                 } else {
 380                                         obstack_grow_utf8(&symbol_obstack, v);
 381                                 }
 382                                 break;
 383                         }
 384
 385                         default:
 386                                 put_back(c);
 387                                 c = '\\';
 388                                 goto end_symbol;
 389                         }
 390
 391                 default:
 392 dollar_sign:
 393                         goto end_symbol;
 394                 }
 395         }
 396
 397 end_symbol:
 398         obstack_1grow(&symbol_obstack, '\0');
 399
 400         char     *string = obstack_finish(&symbol_obstack);
 401         symbol_t *symbol = symbol_table_insert(string);
 402
 403         lexer_token.kind        = symbol->ID;
 404         lexer_token.base.symbol = symbol;
 405
 406         if (symbol->string != string) {
 407                 obstack_free(&symbol_obstack, string);
 408         }
 409 }
 410
 411 static string_t sym_make_string(string_encoding_t const enc)
 412 {
 413         obstack_1grow(&symbol_obstack, '\0');
 414         size_t const len    = obstack_object_size(&symbol_obstack) - 1;
 415         char  *const string = obstack_finish(&symbol_obstack);
 416
 417         /* TODO hash */
 418 #if 0
 419         const char *result = strset_insert(&stringset, concat);
 420         if (result != concat) {
 421                 obstack_free(&symbol_obstack, concat);
 422         }
 423 #else
 424         const char *result = string;
 425 #endif
 426         return (string_t){ result, len, enc };
 427 }
 428
 429 static void parse_pp_number(void)
 430 {
 431         for (;;) {
 432                 switch (c) {
 433                 case SYMBOL_CHARS_E_P:
 434                         obstack_1grow(&symbol_obstack, (char)c);
 435                         next_char();
 436                         if (c == '+' || c == '-') {
 437                 case '.':
 438                 case DIGITS:
 439                 case SYMBOL_CHARS_WITHOUT_E_P:
 440                                 obstack_1grow(&symbol_obstack, (char)c);
 441                                 next_char();
 442                         }
 443                         break;
 444
 445                 default:
 446 dollar_sign:
 447                         lexer_token.kind           = T_NUMBER;
 448                         lexer_token.literal.string = sym_make_string(STRING_ENCODING_CHAR);
 449                         return;
 450                 }
 451         }
 452 }
 453
 454 /**
 455  * Returns true if the given char is a octal digit.
 456  *
 457  * @param char  the character to check
 458  */
 459 static bool is_octal_digit(utf32 chr)
 460 {
 461         return '0' <= chr && chr <= '7';
 462 }
 463
 464 /**
 465  * Returns the value of a digit.
 466  * The only portable way to do it ...
 467  */
 468 static int digit_value(utf32 const digit)
 469 {
 470         switch (digit) {
 471         case '0': return 0;
 472         case '1': return 1;
 473         case '2': return 2;
 474         case '3': return 3;
 475         case '4': return 4;
 476         case '5': return 5;
 477         case '6': return 6;
 478         case '7': return 7;
 479         case '8': return 8;
 480         case '9': return 9;
 481         case 'a':
 482         case 'A': return 10;
 483         case 'b':
 484         case 'B': return 11;
 485         case 'c':
 486         case 'C': return 12;
 487         case 'd':
 488         case 'D': return 13;
 489         case 'e':
 490         case 'E': return 14;
 491         case 'f':
 492         case 'F': return 15;
 493         default:
 494                 internal_error("wrong character given");
 495         }
 496 }
 497
 498 /**
 499  * Parses an octal character sequence.
 500  *
 501  * @param first_digit  the already read first digit
 502  */
 503 static utf32 parse_octal_sequence(utf32 const first_digit)
 504 {
 505         assert(is_octal_digit(first_digit));
 506         utf32 value = digit_value(first_digit);
 507         if (!is_octal_digit(c)) return value;
 508         value = 8 * value + digit_value(c);
 509         next_char();
 510         if (!is_octal_digit(c)) return value;
 511         value = 8 * value + digit_value(c);
 512         next_char();
 513         return value;
 514 }
 515
 516 /**
 517  * Parses a hex character sequence.
 518  */
 519 static utf32 parse_hex_sequence(void)
 520 {
 521         utf32 value = 0;
 522         while (isxdigit(c)) {
 523                 value = 16 * value + digit_value(c);
 524                 next_char();
 525         }
 526         return value;
 527 }
 528
 529 /**
 530  * Parse an escape sequence.
 531  */
 532 static utf32 parse_escape_sequence(void)
 533 {
 534         eat('\\');
 535
 536         utf32 const ec = c;
 537         next_char();
 538
 539         switch (ec) {
 540         case '"':  return '"';
 541         case '\'': return '\'';
 542         case '\\': return '\\';
 543         case '?': return '\?';
 544         case 'a': return '\a';
 545         case 'b': return '\b';
 546         case 'f': return '\f';
 547         case 'n': return '\n';
 548         case 'r': return '\r';
 549         case 't': return '\t';
 550         case 'v': return '\v';
 551         case 'x':
 552                 return parse_hex_sequence();
 553         case '0':
 554         case '1':
 555         case '2':
 556         case '3':
 557         case '4':
 558         case '5':
 559         case '6':
 560         case '7':
 561                 return parse_octal_sequence(ec);
 562         case EOF:
 563                 parse_error("reached end of file while parsing escape sequence");
 564                 return EOF;
 565         /* \E is not documented, but handled, by GCC.  It is acceptable according
 566          * to §6.11.4, whereas \e is not. */
 567         case 'E':
 568         case 'e':
 569                 if (c_mode & _GNUC)
 570                         return 27;   /* hopefully 27 is ALWAYS the code for ESCAPE */
 571                 break;
 572
 573         case 'U': return parse_universal_char(8);
 574         case 'u': return parse_universal_char(4);
 575
 576         default:
 577                 break;
 578         }
 579         /* §6.4.4.4:8 footnote 64 */
 580         parse_error("unknown escape sequence");
 581         return EOF;
 582 }
 583
 584 string_t make_string(const char *string)
 585 {
 586         obstack_grow(&symbol_obstack, string, strlen(string));
 587         return sym_make_string(STRING_ENCODING_CHAR);
 588 }
 589
 590 static void parse_string(utf32 const delim, token_kind_t const kind, string_encoding_t const enc, char const *const context)
 591 {
 592         eat(delim);
 593
 594         while (true) {
 595                 switch (c) {
 596                 case '\\': {
 597                         utf32 const tc = parse_escape_sequence();
 598                         if (enc == STRING_ENCODING_CHAR) {
 599                                 if (tc >= 0x100) {
 600                                         warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 601                                 }
 602                                 obstack_1grow(&symbol_obstack, tc);
 603                         } else {
 604                                 obstack_grow_utf8(&symbol_obstack, tc);
 605                         }
 606                         break;
 607                 }
 608
 609                 case NEWLINE:
 610                         errorf(&lexer_pos, "newline while parsing %s", context);
 611                         break;
 612
 613                 case EOF:
 614                         errorf(&lexer_token.base.source_position, "EOF while parsing %s", context);
 615                         goto end_of_string;
 616
 617                 default:
 618                         if (c == delim) {
 619                                 next_char();
 620                                 goto end_of_string;
 621                         } else {
 622                                 obstack_grow_utf8(&symbol_obstack, c);
 623                                 next_char();
 624                                 break;
 625                         }
 626                 }
 627         }
 628
 629 end_of_string:
 630         lexer_token.kind           = kind;
 631         lexer_token.literal.string = sym_make_string(enc);
 632 }
 633
 634 /**
 635  * Parse a string literal and set lexer_token.
 636  */
 637 static void parse_string_literal(string_encoding_t const enc)
 638 {
 639         parse_string('"', T_STRING_LITERAL, enc, "string literal");
 640 }
 641
 642 /**
 643  * Parse a character constant and set lexer_token.
 644  */
 645 static void parse_character_constant(string_encoding_t const enc)
 646 {
 647         parse_string('\'', T_CHARACTER_CONSTANT, enc, "character constant");
 648         if (lexer_token.literal.string.size == 0) {
 649                 errorf(&lexer_token.base.source_position, "empty character constant");
 650         }
 651 }
 652
 653 /**
 654  * Skip a multiline comment.
 655  */
 656 static void skip_multiline_comment(void)
 657 {
 658         while (true) {
 659                 switch (c) {
 660                 case '/':
 661                         next_char();
 662                         if (c == '*') {
 663                                 /* nested comment, warn here */
 664                                 warningf(WARN_COMMENT, &lexer_pos, "'/*' within comment");
 665                         }
 666                         break;
 667                 case '*':
 668                         next_char();
 669                         if (c == '/') {
 670                                 next_char();
 671                                 return;
 672                         }
 673                         break;
 674
 675                 case NEWLINE:
 676                         break;
 677
 678                 case EOF: {
 679                         errorf(&lexer_token.base.source_position,
 680                                "at end of file while looking for comment end");
 681                         return;
 682                 }
 683
 684                 default:
 685                         next_char();
 686                         break;
 687                 }
 688         }
 689 }
 690
 691 /**
 692  * Skip a single line comment.
 693  */
 694 static void skip_line_comment(void)
 695 {
 696         while (true) {
 697                 switch (c) {
 698                 case EOF:
 699                         return;
 700
 701                 case '\n':
 702                 case '\r':
 703                         return;
 704
 705                 case '\\':
 706                         next_char();
 707                         if (c == '\n' || c == '\r') {
 708                                 warningf(WARN_COMMENT, &lexer_pos, "multi-line comment");
 709                                 return;
 710                         }
 711                         break;
 712
 713                 default:
 714                         next_char();
 715                         break;
 716                 }
 717         }
 718 }
 719
 720 /** The current preprocessor token. */
 721 static token_t pp_token;
 722
 723 /**
 724  * Read the next preprocessor token.
 725  */
 726 static inline void next_pp_token(void)
 727 {
 728         lexer_next_preprocessing_token();
 729         pp_token = lexer_token;
 730 }
 731
 732 /**
 733  * Eat all preprocessor tokens until newline.
 734  */
 735 static void eat_until_newline(void)
 736 {
 737         while (pp_token.kind != '\n' && pp_token.kind != T_EOF) {
 738                 next_pp_token();
 739         }
 740 }
 741
 742 /**
 743  * Parse the line directive.
 744  */
 745 static void parse_line_directive(void)
 746 {
 747         if (pp_token.kind != T_NUMBER) {
 748                 parse_error("expected integer");
 749         } else {
 750                 /* use offset -1 as this is about the next line */
 751                 char      *end;
 752                 long const line = strtol(pp_token.literal.string.begin, &end, 0);
 753                 if (*end == '\0') {
 754                         lexer_pos.lineno = line - 1;
 755                 } else {
 756                         errorf(&lexer_pos, "'%S' is not a valid line number", &pp_token.literal.string);
 757                 }
 758                 next_pp_token();
 759         }
 760         if (pp_token.kind == T_STRING_LITERAL && pp_token.literal.string.encoding == STRING_ENCODING_CHAR) {
 761                 lexer_pos.input_name       = pp_token.literal.string.begin;
 762                 lexer_pos.is_system_header = false;
 763                 next_pp_token();
 764
 765                 /* attempt to parse numeric flags as outputted by gcc preprocessor */
 766                 while (pp_token.kind == T_NUMBER) {
 767                         /* flags:
 768                          * 1 - indicates start of a new file
 769                          * 2 - indicates return from a file
 770                          * 3 - indicates system header
 771                          * 4 - indicates implicit extern "C" in C++ mode
 772                          *
 773                          * currently we're only interested in "3"
 774                          */
 775                         if (streq(pp_token.literal.string.begin, "3")) {
 776                                 lexer_pos.is_system_header = true;
 777                         }
 778                         next_pp_token();
 779                 }
 780         }
 781
 782         eat_until_newline();
 783 }
 784
 785 /**
 786  * STDC pragmas.
 787  */
 788 typedef enum stdc_pragma_kind_t {
 789         STDC_UNKNOWN,
 790         STDC_FP_CONTRACT,
 791         STDC_FENV_ACCESS,
 792         STDC_CX_LIMITED_RANGE
 793 } stdc_pragma_kind_t;
 794
 795 /**
 796  * STDC pragma values.
 797  */
 798 typedef enum stdc_pragma_value_kind_t {
 799         STDC_VALUE_UNKNOWN,
 800         STDC_VALUE_ON,
 801         STDC_VALUE_OFF,
 802         STDC_VALUE_DEFAULT
 803 } stdc_pragma_value_kind_t;
 804
 805 /**
 806  * Parse a pragma directive.
 807  */
 808 static void parse_pragma(void)
 809 {
 810         next_pp_token();
 811         if (pp_token.kind != T_IDENTIFIER) {
 812                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
 813                          "expected identifier after #pragma");
 814                 eat_until_newline();
 815                 return;
 816         }
 817
 818         stdc_pragma_kind_t kind = STDC_UNKNOWN;
 819         if (pp_token.base.symbol->pp_ID == TP_STDC && c_mode & _C99) {
 820                 /* a STDC pragma */
 821                 next_pp_token();
 822
 823                 switch (pp_token.base.symbol->pp_ID) {
 824                 case TP_FP_CONTRACT:      kind = STDC_FP_CONTRACT;      break;
 825                 case TP_FENV_ACCESS:      kind = STDC_FENV_ACCESS;      break;
 826                 case TP_CX_LIMITED_RANGE: kind = STDC_CX_LIMITED_RANGE; break;
 827                 default:                  break;
 828                 }
 829                 if (kind != STDC_UNKNOWN) {
 830                         next_pp_token();
 831                         stdc_pragma_value_kind_t value;
 832                         switch (pp_token.base.symbol->pp_ID) {
 833                         case TP_ON:      value = STDC_VALUE_ON;      break;
 834                         case TP_OFF:     value = STDC_VALUE_OFF;     break;
 835                         case TP_DEFAULT: value = STDC_VALUE_DEFAULT; break;
 836                         default:         value = STDC_VALUE_UNKNOWN; break;
 837                         }
 838                         if (value == STDC_VALUE_UNKNOWN) {
 839                                 kind = STDC_UNKNOWN;
 840                                 errorf(&pp_token.base.source_position, "bad STDC pragma argument");
 841                         }
 842                 }
 843         }
 844         eat_until_newline();
 845         if (kind == STDC_UNKNOWN) {
 846                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
 847                          "encountered unknown #pragma");
 848         }
 849 }
 850
 851 /**
 852  * Parse a preprocessor non-null directive.
 853  */
 854 static void parse_preprocessor_identifier(void)
 855 {
 856         assert(pp_token.kind == T_IDENTIFIER);
 857         switch (pp_token.base.symbol->pp_ID) {
 858         case TP_line:
 859                 next_pp_token();
 860                 parse_line_directive();
 861                 break;
 862         case TP_pragma:
 863                 parse_pragma();
 864                 break;
 865         case TP_error:
 866                 /* TODO; output the rest of the line */
 867                 parse_error("#error directive");
 868                 break;
 869         }
 870 }
 871
 872 /**
 873  * Parse a preprocessor directive.
 874  */
 875 static void parse_preprocessor_directive(void)
 876 {
 877         next_pp_token();
 878
 879         switch (pp_token.kind) {
 880         case T_IDENTIFIER:
 881                 parse_preprocessor_identifier();
 882                 break;
 883         case T_NUMBER:
 884                 parse_line_directive();
 885                 break;
 886         case '\n':
 887                 /* NULL directive, see §6.10.7 */
 888                 break;
 889         default:
 890                 parse_error("invalid preprocessor directive");
 891                 eat_until_newline();
 892                 break;
 893         }
 894 }
 895
 896 #define MAYBE_PROLOG                                       \
 897                         next_char();                                   \
 898                         while (true) {                                 \
 899                                 switch (c) {
 900
 901 #define MAYBE(ch, set_type)                                \
 902                                 case ch:                                   \
 903                                         next_char();                           \
 904                                         lexer_token.kind = set_type;           \
 905                                         return;
 906
 907 /* must use this as last thing */
 908 #define MAYBE_MODE(ch, set_type, mode)                     \
 909                                 case ch:                                   \
 910                                         if (c_mode & mode) {                   \
 911                                                 next_char();                       \
 912                                                 lexer_token.kind = set_type;       \
 913                                                 return;                            \
 914                                         }                                      \
 915                                         /* fallthrough */
 916
 917 #define ELSE_CODE(code)                                    \
 918                                 default:                                   \
 919                                         code                                   \
 920                                         return;                                \
 921                                 }                                          \
 922                         } /* end of while (true) */                    \
 923
 924 #define ELSE(set_type)                                     \
 925                 ELSE_CODE(                                         \
 926                         lexer_token.kind = set_type;                   \
 927                 )
 928
 929 void lexer_next_preprocessing_token(void)
 930 {
 931         while (true) {
 932                 lexer_token.base.source_position = lexer_pos;
 933                 lexer_token.base.symbol          = NULL;
 934
 935                 switch (c) {
 936                 case ' ':
 937                 case '\t':
 938                         next_char();
 939                         break;
 940
 941                 case NEWLINE:
 942                         lexer_token.kind = '\n';
 943                         return;
 944
 945                 case SYMBOL_CHARS: {
 946                         parse_symbol();
 947                         /* might be a wide string ( L"string" ) */
 948                         string_encoding_t const enc = STRING_ENCODING_WIDE;
 949                         if (lexer_token.base.symbol == symbol_L) {
 950                                 switch (c) {
 951                                 case '"':  parse_string_literal(enc);     break;
 952                                 case '\'': parse_character_constant(enc); break;
 953                                 }
 954                         }
 955                         return;
 956                 }
 957
 958                 case DIGITS:
 959                         parse_pp_number();
 960                         return;
 961
 962                 case '"':
 963                         parse_string_literal(STRING_ENCODING_CHAR);
 964                         return;
 965
 966                 case '\'':
 967                         parse_character_constant(STRING_ENCODING_CHAR);
 968                         return;
 969
 970                 case '.':
 971                         MAYBE_PROLOG
 972                                 case DIGITS:
 973                                         put_back(c);
 974                                         c = '.';
 975                                         parse_pp_number();
 976                                         return;
 977
 978                                 case '.':
 979                                         MAYBE_PROLOG
 980                                         MAYBE('.', T_DOTDOTDOT)
 981                                         ELSE_CODE(
 982                                                 put_back(c);
 983                                                 c = '.';
 984                                                 lexer_token.kind = '.';
 985                                         )
 986                         ELSE('.')
 987                 case '&':
 988                         MAYBE_PROLOG
 989                         MAYBE('&', T_ANDAND)
 990                         MAYBE('=', T_ANDEQUAL)
 991                         ELSE('&')
 992                 case '*':
 993                         MAYBE_PROLOG
 994                         MAYBE('=', T_ASTERISKEQUAL)
 995                         ELSE('*')
 996                 case '+':
 997                         MAYBE_PROLOG
 998                         MAYBE('+', T_PLUSPLUS)
 999                         MAYBE('=', T_PLUSEQUAL)
1000                         ELSE('+')
1001                 case '-':
1002                         MAYBE_PROLOG
1003                         MAYBE('>', T_MINUSGREATER)
1004                         MAYBE('-', T_MINUSMINUS)
1005                         MAYBE('=', T_MINUSEQUAL)
1006                         ELSE('-')
1007                 case '!':
1008                         MAYBE_PROLOG
1009                         MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1010                         ELSE('!')
1011                 case '/':
1012                         MAYBE_PROLOG
1013                         MAYBE('=', T_SLASHEQUAL)
1014                                 case '*':
1015                                         next_char();
1016                                         skip_multiline_comment();
1017                                         lexer_next_preprocessing_token();
1018                                         return;
1019                                 case '/':
1020                                         next_char();
1021                                         skip_line_comment();
1022                                         lexer_next_preprocessing_token();
1023                                         return;
1024                         ELSE('/')
1025                 case '%':
1026                         MAYBE_PROLOG
1027                         MAYBE('>', '}')
1028                         MAYBE('=', T_PERCENTEQUAL)
1029                                 case ':':
1030                                         MAYBE_PROLOG
1031                                                 case '%':
1032                                                         MAYBE_PROLOG
1033                                                         MAYBE(':', T_HASHHASH)
1034                                                         ELSE_CODE(
1035                                                                 put_back(c);
1036                                                                 c = '%';
1037                                                                 lexer_token.kind = '#';
1038                                                         )
1039                                         ELSE('#')
1040                         ELSE('%')
1041                 case '<':
1042                         MAYBE_PROLOG
1043                         MAYBE(':', '[')
1044                         MAYBE('%', '{')
1045                         MAYBE('=', T_LESSEQUAL)
1046                                 case '<':
1047                                         MAYBE_PROLOG
1048                                         MAYBE('=', T_LESSLESSEQUAL)
1049                                         ELSE(T_LESSLESS)
1050                         ELSE('<')
1051                 case '>':
1052                         MAYBE_PROLOG
1053                         MAYBE('=', T_GREATEREQUAL)
1054                                 case '>':
1055                                         MAYBE_PROLOG
1056                                         MAYBE('=', T_GREATERGREATEREQUAL)
1057                                         ELSE(T_GREATERGREATER)
1058                         ELSE('>')
1059                 case '^':
1060                         MAYBE_PROLOG
1061                         MAYBE('=', T_CARETEQUAL)
1062                         ELSE('^')
1063                 case '|':
1064                         MAYBE_PROLOG
1065                         MAYBE('=', T_PIPEEQUAL)
1066                         MAYBE('|', T_PIPEPIPE)
1067                         ELSE('|')
1068                 case ':':
1069                         MAYBE_PROLOG
1070                         MAYBE('>', ']')
1071                         MAYBE_MODE(':', T_COLONCOLON, _CXX)
1072                         ELSE(':')
1073                 case '=':
1074                         MAYBE_PROLOG
1075                         MAYBE('=', T_EQUALEQUAL)
1076                         ELSE('=')
1077                 case '#':
1078                         MAYBE_PROLOG
1079                         MAYBE('#', T_HASHHASH)
1080                         ELSE('#')
1081
1082                 case '\\':
1083                         next_char();
1084                         if (c == 'U' || c == 'u') {
1085                                 put_back(c);
1086                                 c = '\\';
1087                                 parse_symbol();
1088                         } else {
1089                                 lexer_token.kind = '\\';
1090                         }
1091                         return;
1092
1093                 case '?':
1094                 case '[':
1095                 case ']':
1096                 case '(':
1097                 case ')':
1098                 case '{':
1099                 case '}':
1100                 case '~':
1101                 case ';':
1102                 case ',':
1103                         lexer_token.kind = c;
1104                         next_char();
1105                         return;
1106
1107                 case EOF:
1108                         lexer_token.kind = T_EOF;
1109                         return;
1110
1111                 default:
1112 dollar_sign:
1113                         errorf(&lexer_pos, "unknown character '%lc' found", c);
1114                         next_char();
1115                         break;
1116                 }
1117         }
1118 }
1119
1120 void lexer_next_token(void)
1121 {
1122         lexer_next_preprocessing_token();
1123
1124         while (lexer_token.kind == '\n') {
1125 newline_found:
1126                 lexer_next_preprocessing_token();
1127         }
1128
1129         if (lexer_token.kind == '#') {
1130                 parse_preprocessor_directive();
1131                 goto newline_found;
1132         }
1133 }
1134
1135 void init_lexer(void)
1136 {
1137         strset_init(&stringset);
1138         symbol_L = symbol_table_insert("L");
1139 }
1140
1141 static void input_error(unsigned delta_lines, unsigned delta_cols,
1142                         const char *message)
1143 {
1144         lexer_pos.lineno += delta_lines;
1145         lexer_pos.colno  += delta_cols;
1146         errorf(&lexer_pos, "%s", message);
1147 }
1148
1149 void lexer_switch_input(input_t *new_input, const char *input_name)
1150 {
1151         lexer_pos.lineno     = 0;
1152         lexer_pos.colno      = 0;
1153         lexer_pos.input_name = input_name;
1154
1155         set_input_error_callback(input_error);
1156         input  = new_input;
1157         bufpos = NULL;
1158         bufend = NULL;
1159
1160         /* place a virtual \n at the beginning so the lexer knows that we're
1161          * at the beginning of a line */
1162         c = '\n';
1163 }
1164
1165 void exit_lexer(void)
1166 {
1167         strset_destroy(&stringset);
1168 }
1169
1170 static __attribute__((unused))
1171 void dbg_pos(const source_position_t source_position)
1172 {
1173         fprintf(stdout, "%s:%u:%u\n", source_position.input_name,
1174                 source_position.lineno, (unsigned)source_position.colno);
1175         fflush(stdout);
1176 }