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  \
 185         case '$': if (!allow_dollar_in_symbol) goto dollar_sign; \
 186         case 'a':         \
 187         case 'b':         \
 188         case 'c':         \
 189         case 'd':         \
 190         case 'e':         \
 191         case 'f':         \
 192         case 'g':         \
 193         case 'h':         \
 194         case 'i':         \
 195         case 'j':         \
 196         case 'k':         \
 197         case 'l':         \
 198         case 'm':         \
 199         case 'n':         \
 200         case 'o':         \
 201         case 'p':         \
 202         case 'q':         \
 203         case 'r':         \
 204         case 's':         \
 205         case 't':         \
 206         case 'u':         \
 207         case 'v':         \
 208         case 'w':         \
 209         case 'x':         \
 210         case 'y':         \
 211         case 'z':         \
 212         case 'A':         \
 213         case 'B':         \
 214         case 'C':         \
 215         case 'D':         \
 216         case 'E':         \
 217         case 'F':         \
 218         case 'G':         \
 219         case 'H':         \
 220         case 'I':         \
 221         case 'J':         \
 222         case 'K':         \
 223         case 'L':         \
 224         case 'M':         \
 225         case 'N':         \
 226         case 'O':         \
 227         case 'P':         \
 228         case 'Q':         \
 229         case 'R':         \
 230         case 'S':         \
 231         case 'T':         \
 232         case 'U':         \
 233         case 'V':         \
 234         case 'W':         \
 235         case 'X':         \
 236         case 'Y':         \
 237         case 'Z':         \
 238         case '_':
 239
 240 #define DIGITS        \
 241         case '0':         \
 242         case '1':         \
 243         case '2':         \
 244         case '3':         \
 245         case '4':         \
 246         case '5':         \
 247         case '6':         \
 248         case '7':         \
 249         case '8':         \
 250         case '9':
 251
 252 static bool is_universal_char_valid(utf32 const v)
 253 {
 254         /* C11 §6.4.3:2 */
 255         if (v < 0xA0U && v != 0x24 && v != 0x40 && v != 0x60)
 256                 return false;
 257         if (0xD800 <= v && v <= 0xDFFF)
 258                 return false;
 259         return true;
 260 }
 261
 262 static int digit_value(utf32 digit);
 263
 264 static utf32 parse_universal_char(unsigned const n_digits)
 265 {
 266         utf32 v = 0;
 267         for (unsigned k = n_digits; k != 0; --k) {
 268                 if (isxdigit(c)) {
 269                         v = 16 * v + digit_value(c);
 270                         next_char();
 271                 } else {
 272                         errorf(&lexer_pos, "short universal character name, expected %u more digits", k);
 273                         break;
 274                 }
 275         }
 276         if (!is_universal_char_valid(v)) {
 277                 errorf(&lexer_pos, "\\%c%0*X is not a valid universal character name", n_digits == 4 ? 'u' : 'U', (int)n_digits, v);
 278         }
 279         return v;
 280 }
 281
 282 static bool is_universal_char_valid_identifier(utf32 const v)
 283 {
 284         /* C11 Annex D.1 */
 285         if (                v == 0x000A8) return true;
 286         if (                v == 0x000AA) return true;
 287         if (                v == 0x000AD) return true;
 288         if (                v == 0x000AF) return true;
 289         if (0x000B2 <= v && v <= 0x000B5) return true;
 290         if (0x000B7 <= v && v <= 0x000BA) return true;
 291         if (0x000BC <= v && v <= 0x000BE) return true;
 292         if (0x000C0 <= v && v <= 0x000D6) return true;
 293         if (0x000D8 <= v && v <= 0x000F6) return true;
 294         if (0x000F8 <= v && v <= 0x000FF) return true;
 295         if (0x00100 <= v && v <= 0x0167F) return true;
 296         if (0x01681 <= v && v <= 0x0180D) return true;
 297         if (0x0180F <= v && v <= 0x01FFF) return true;
 298         if (0x0200B <= v && v <= 0x0200D) return true;
 299         if (0x0202A <= v && v <= 0x0202E) return true;
 300         if (0x0203F <= v && v <= 0x02040) return true;
 301         if (                v == 0x02054) return true;
 302         if (0x02060 <= v && v <= 0x0206F) return true;
 303         if (0x02070 <= v && v <= 0x0218F) return true;
 304         if (0x02460 <= v && v <= 0x024FF) return true;
 305         if (0x02776 <= v && v <= 0x02793) return true;
 306         if (0x02C00 <= v && v <= 0x02DFF) return true;
 307         if (0x02E80 <= v && v <= 0x02FFF) return true;
 308         if (0x03004 <= v && v <= 0x03007) return true;
 309         if (0x03021 <= v && v <= 0x0302F) return true;
 310         if (0x03031 <= v && v <= 0x0303F) return true;
 311         if (0x03040 <= v && v <= 0x0D7FF) return true;
 312         if (0x0F900 <= v && v <= 0x0FD3D) return true;
 313         if (0x0FD40 <= v && v <= 0x0FDCF) return true;
 314         if (0x0FDF0 <= v && v <= 0x0FE44) return true;
 315         if (0x0FE47 <= v && v <= 0x0FFFD) return true;
 316         if (0x10000 <= v && v <= 0x1FFFD) return true;
 317         if (0x20000 <= v && v <= 0x2FFFD) return true;
 318         if (0x30000 <= v && v <= 0x3FFFD) return true;
 319         if (0x40000 <= v && v <= 0x4FFFD) return true;
 320         if (0x50000 <= v && v <= 0x5FFFD) return true;
 321         if (0x60000 <= v && v <= 0x6FFFD) return true;
 322         if (0x70000 <= v && v <= 0x7FFFD) return true;
 323         if (0x80000 <= v && v <= 0x8FFFD) return true;
 324         if (0x90000 <= v && v <= 0x9FFFD) return true;
 325         if (0xA0000 <= v && v <= 0xAFFFD) return true;
 326         if (0xB0000 <= v && v <= 0xBFFFD) return true;
 327         if (0xC0000 <= v && v <= 0xCFFFD) return true;
 328         if (0xD0000 <= v && v <= 0xDFFFD) return true;
 329         if (0xE0000 <= v && v <= 0xEFFFD) return true;
 330         return false;
 331 }
 332
 333 static bool is_universal_char_valid_identifier_start(utf32 const v)
 334 {
 335         /* C11 Annex D.2 */
 336         if (0x0300 <= v && v <= 0x036F) return false;
 337         if (0x1DC0 <= v && v <= 0x1DFF) return false;
 338         if (0x20D0 <= v && v <= 0x20FF) return false;
 339         if (0xFE20 <= v && v <= 0xFE2F) return false;
 340         return true;
 341 }
 342
 343 /**
 344  * Read a symbol from the input and build
 345  * the lexer_token.
 346  */
 347 static void parse_symbol(void)
 348 {
 349         while (true) {
 350                 switch (c) {
 351                 DIGITS
 352                 SYMBOL_CHARS
 353                         obstack_1grow(&symbol_obstack, (char) c);
 354                         next_char();
 355                         break;
 356
 357                 case '\\':
 358                         next_char();
 359                         switch (c) {
 360                         {
 361                                 unsigned n;
 362                         case 'U': n = 8; goto universal;
 363                         case 'u': n = 4; goto universal;
 364 universal:
 365                                 next_char();
 366                                 utf32 const v = parse_universal_char(n);
 367                                 if (!is_universal_char_valid_identifier(v)) {
 368                                         if (is_universal_char_valid(v)) {
 369                                                 errorf(&lexer_pos, "universal character \\%c%0*X is not valid in an identifier", n == 4 ? 'u' : 'U', (int)n, v);
 370                                         }
 371                                 } else if (obstack_object_size(&symbol_obstack) == 0 && !is_universal_char_valid_identifier_start(v)) {
 372                                         errorf(&lexer_pos, "universal character \\%c%0*X is not valid as start of an identifier", n == 4 ? 'u' : 'U', (int)n, v);
 373                                 } else {
 374                                         obstack_grow_symbol(&symbol_obstack, v);
 375                                 }
 376                                 break;
 377                         }
 378
 379                         default:
 380                                 put_back(c);
 381                                 c = '\\';
 382                                 goto end_symbol;
 383                         }
 384
 385                 default:
 386 dollar_sign:
 387                         goto end_symbol;
 388                 }
 389         }
 390
 391 end_symbol:
 392         obstack_1grow(&symbol_obstack, '\0');
 393
 394         char     *string = obstack_finish(&symbol_obstack);
 395         symbol_t *symbol = symbol_table_insert(string);
 396
 397         lexer_token.kind        = symbol->ID;
 398         lexer_token.base.symbol = symbol;
 399
 400         if (symbol->string != string) {
 401                 obstack_free(&symbol_obstack, string);
 402         }
 403 }
 404
 405 static string_t identify_string(char *string, size_t len)
 406 {
 407         /* TODO hash */
 408 #if 0
 409         const char *result = strset_insert(&stringset, concat);
 410         if (result != concat) {
 411                 obstack_free(&symbol_obstack, concat);
 412         }
 413 #else
 414         const char *result = string;
 415 #endif
 416         return (string_t) {result, len};
 417 }
 418
 419 /**
 420  * parse suffixes like 'LU' or 'f' after numbers
 421  */
 422 static void parse_number_suffix(void)
 423 {
 424         assert(obstack_object_size(&symbol_obstack) == 0);
 425         while (true) {
 426                 switch (c) {
 427                 SYMBOL_CHARS
 428                         obstack_1grow(&symbol_obstack, (char) c);
 429                         next_char();
 430                         break;
 431                 default:
 432                 dollar_sign:
 433                         goto finish_suffix;
 434                 }
 435         }
 436 finish_suffix:
 437         if (obstack_object_size(&symbol_obstack) == 0) {
 438                 lexer_token.number.suffix.begin = NULL;
 439                 lexer_token.number.suffix.size  = 0;
 440                 return;
 441         }
 442
 443         obstack_1grow(&symbol_obstack, '\0');
 444         size_t size   = obstack_object_size(&symbol_obstack) - 1;
 445         char  *string = obstack_finish(&symbol_obstack);
 446
 447         lexer_token.number.suffix = identify_string(string, size);
 448 }
 449
 450 static void parse_exponent(void)
 451 {
 452         if (c == '-' || c == '+') {
 453                 obstack_1grow(&symbol_obstack, (char)c);
 454                 next_char();
 455         }
 456
 457         if (isdigit(c)) {
 458                 do {
 459                         obstack_1grow(&symbol_obstack, (char)c);
 460                         next_char();
 461                 } while (isdigit(c));
 462         } else {
 463                 errorf(&lexer_token.base.source_position, "exponent has no digits");
 464         }
 465 }
 466
 467 /**
 468  * Parses a hex number including hex floats and set the
 469  * lexer_token.
 470  */
 471 static void parse_number_hex(void)
 472 {
 473         bool is_float   = false;
 474         bool has_digits = false;
 475
 476         while (isxdigit(c)) {
 477                 has_digits = true;
 478                 obstack_1grow(&symbol_obstack, (char) c);
 479                 next_char();
 480         }
 481
 482         if (c == '.') {
 483                 is_float = true;
 484                 obstack_1grow(&symbol_obstack, (char) c);
 485                 next_char();
 486
 487                 while (isxdigit(c)) {
 488                         has_digits = true;
 489                         obstack_1grow(&symbol_obstack, (char) c);
 490                         next_char();
 491                 }
 492         }
 493         if (c == 'p' || c == 'P') {
 494                 is_float = true;
 495                 obstack_1grow(&symbol_obstack, (char) c);
 496                 next_char();
 497                 parse_exponent();
 498         } else if (is_float) {
 499                 errorf(&lexer_token.base.source_position,
 500                        "hexadecimal floatingpoint constant requires an exponent");
 501         }
 502         obstack_1grow(&symbol_obstack, '\0');
 503
 504         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 505         char   *string = obstack_finish(&symbol_obstack);
 506         lexer_token.number.number = identify_string(string, size);
 507
 508         lexer_token.kind = is_float ? T_FLOATINGPOINT : T_INTEGER;
 509
 510         if (!has_digits) {
 511                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'", &lexer_token.number.number);
 512                 lexer_token.number.number.begin = "0";
 513                 lexer_token.number.number.size  = 1;
 514         }
 515
 516         parse_number_suffix();
 517 }
 518
 519 static void parse_number_bin(void)
 520 {
 521         bool has_digits = false;
 522
 523         while (c == '0' || c == '1') {
 524                 has_digits = true;
 525                 obstack_1grow(&symbol_obstack, (char)c);
 526                 next_char();
 527         }
 528         obstack_1grow(&symbol_obstack, '\0');
 529
 530         size_t  const size   = obstack_object_size(&symbol_obstack) - 1;
 531         char   *const string = obstack_finish(&symbol_obstack);
 532         lexer_token.number.number = identify_string(string, size);
 533         lexer_token.kind          = T_INTEGER;
 534
 535         if (!has_digits) {
 536                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'", &lexer_token.number.number);
 537                 lexer_token.number.number.begin = "0";
 538                 lexer_token.number.number.size  = 1;
 539         }
 540
 541         parse_number_suffix();
 542 }
 543
 544 /**
 545  * Returns true if the given char is a octal digit.
 546  *
 547  * @param char  the character to check
 548  */
 549 static bool is_octal_digit(utf32 chr)
 550 {
 551         return '0' <= chr && chr <= '7';
 552 }
 553
 554 /**
 555  * Parses a number and sets the lexer_token.
 556  */
 557 static void parse_number(void)
 558 {
 559         bool is_float   = false;
 560         bool has_digits = false;
 561
 562         assert(obstack_object_size(&symbol_obstack) == 0);
 563         if (c == '0') {
 564                 obstack_1grow(&symbol_obstack, (char)c);
 565                 next_char();
 566                 if (c == 'x' || c == 'X') {
 567                         obstack_1grow(&symbol_obstack, (char)c);
 568                         next_char();
 569                         parse_number_hex();
 570                         return;
 571                 } else if (c == 'b' || c == 'B') {
 572                         /* GCC extension: binary constant 0x[bB][01]+.  */
 573                         obstack_1grow(&symbol_obstack, (char)c);
 574                         next_char();
 575                         parse_number_bin();
 576                         return;
 577                 }
 578                 has_digits = true;
 579         }
 580
 581         while (isdigit(c)) {
 582                 has_digits = true;
 583                 obstack_1grow(&symbol_obstack, (char) c);
 584                 next_char();
 585         }
 586
 587         if (c == '.') {
 588                 is_float = true;
 589                 obstack_1grow(&symbol_obstack, '.');
 590                 next_char();
 591
 592                 while (isdigit(c)) {
 593                         has_digits = true;
 594                         obstack_1grow(&symbol_obstack, (char) c);
 595                         next_char();
 596                 }
 597         }
 598         if (c == 'e' || c == 'E') {
 599                 is_float = true;
 600                 obstack_1grow(&symbol_obstack, 'e');
 601                 next_char();
 602                 parse_exponent();
 603         }
 604
 605         obstack_1grow(&symbol_obstack, '\0');
 606         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 607         char   *string = obstack_finish(&symbol_obstack);
 608         lexer_token.number.number = identify_string(string, size);
 609
 610         if (is_float) {
 611                 lexer_token.kind = T_FLOATINGPOINT;
 612         } else {
 613                 lexer_token.kind = T_INTEGER;
 614
 615                 if (string[0] == '0') {
 616                         /* check for invalid octal digits */
 617                         for (size_t i= 0; i < size; ++i) {
 618                                 char t = string[i];
 619                                 if (t >= '8')
 620                                         errorf(&lexer_token.base.source_position, "invalid digit '%c' in octal number", t);
 621                         }
 622                 }
 623         }
 624
 625         if (!has_digits) {
 626                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'",
 627                        &lexer_token.number.number);
 628         }
 629
 630         parse_number_suffix();
 631 }
 632
 633 /**
 634  * Returns the value of a digit.
 635  * The only portable way to do it ...
 636  */
 637 static int digit_value(utf32 const digit)
 638 {
 639         switch (digit) {
 640         case '0': return 0;
 641         case '1': return 1;
 642         case '2': return 2;
 643         case '3': return 3;
 644         case '4': return 4;
 645         case '5': return 5;
 646         case '6': return 6;
 647         case '7': return 7;
 648         case '8': return 8;
 649         case '9': return 9;
 650         case 'a':
 651         case 'A': return 10;
 652         case 'b':
 653         case 'B': return 11;
 654         case 'c':
 655         case 'C': return 12;
 656         case 'd':
 657         case 'D': return 13;
 658         case 'e':
 659         case 'E': return 14;
 660         case 'f':
 661         case 'F': return 15;
 662         default:
 663                 internal_error("wrong character given");
 664         }
 665 }
 666
 667 /**
 668  * Parses an octal character sequence.
 669  *
 670  * @param first_digit  the already read first digit
 671  */
 672 static utf32 parse_octal_sequence(utf32 const first_digit)
 673 {
 674         assert(is_octal_digit(first_digit));
 675         utf32 value = digit_value(first_digit);
 676         if (!is_octal_digit(c)) return value;
 677         value = 8 * value + digit_value(c);
 678         next_char();
 679         if (!is_octal_digit(c)) return value;
 680         value = 8 * value + digit_value(c);
 681         next_char();
 682         return value;
 683 }
 684
 685 /**
 686  * Parses a hex character sequence.
 687  */
 688 static utf32 parse_hex_sequence(void)
 689 {
 690         utf32 value = 0;
 691         while (isxdigit(c)) {
 692                 value = 16 * value + digit_value(c);
 693                 next_char();
 694         }
 695         return value;
 696 }
 697
 698 /**
 699  * Parse an escape sequence.
 700  */
 701 static utf32 parse_escape_sequence(void)
 702 {
 703         eat('\\');
 704
 705         utf32 const ec = c;
 706         next_char();
 707
 708         switch (ec) {
 709         case '"':  return '"';
 710         case '\'': return '\'';
 711         case '\\': return '\\';
 712         case '?': return '\?';
 713         case 'a': return '\a';
 714         case 'b': return '\b';
 715         case 'f': return '\f';
 716         case 'n': return '\n';
 717         case 'r': return '\r';
 718         case 't': return '\t';
 719         case 'v': return '\v';
 720         case 'x':
 721                 return parse_hex_sequence();
 722         case '0':
 723         case '1':
 724         case '2':
 725         case '3':
 726         case '4':
 727         case '5':
 728         case '6':
 729         case '7':
 730                 return parse_octal_sequence(ec);
 731         case EOF:
 732                 parse_error("reached end of file while parsing escape sequence");
 733                 return EOF;
 734         /* \E is not documented, but handled, by GCC.  It is acceptable according
 735          * to §6.11.4, whereas \e is not. */
 736         case 'E':
 737         case 'e':
 738                 if (c_mode & _GNUC)
 739                         return 27;   /* hopefully 27 is ALWAYS the code for ESCAPE */
 740                 break;
 741
 742         case 'U': return parse_universal_char(8);
 743         case 'u': return parse_universal_char(4);
 744
 745         default:
 746                 break;
 747         }
 748         /* §6.4.4.4:8 footnote 64 */
 749         parse_error("unknown escape sequence");
 750         return EOF;
 751 }
 752
 753 string_t make_string(const char *string)
 754 {
 755         size_t      len   = strlen(string) + 1;
 756         char *const space = obstack_alloc(&symbol_obstack, len);
 757         memcpy(space, string, len);
 758
 759         return identify_string(space, len);
 760 }
 761
 762 static void parse_string(utf32 const delim, token_kind_t const kind, string_encoding_t const enc, char const *const context)
 763 {
 764         eat(delim);
 765
 766         while (true) {
 767                 switch (c) {
 768                 case '\\': {
 769                         utf32 const tc = parse_escape_sequence();
 770                         if (enc == STRING_ENCODING_CHAR) {
 771                                 if (tc >= 0x100) {
 772                                         warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 773                                 }
 774                                 obstack_1grow(&symbol_obstack, tc);
 775                         } else {
 776                                 obstack_grow_symbol(&symbol_obstack, tc);
 777                         }
 778                         break;
 779                 }
 780
 781                 case NEWLINE:
 782                         errorf(&lexer_pos, "newline while parsing %s", context);
 783                         break;
 784
 785                 case EOF:
 786                         errorf(&lexer_token.base.source_position, "EOF while parsing %s", context);
 787                         goto end_of_string;
 788
 789                 default:
 790                         if (c == delim) {
 791                                 next_char();
 792                                 goto end_of_string;
 793                         } else {
 794                                 obstack_grow_symbol(&symbol_obstack, c);
 795                                 next_char();
 796                                 break;
 797                         }
 798                 }
 799         }
 800
 801 end_of_string:
 802         obstack_1grow(&symbol_obstack, '\0');
 803         size_t const size   = obstack_object_size(&symbol_obstack) - 1;
 804         char  *const string = obstack_finish(&symbol_obstack);
 805
 806         lexer_token.kind            = kind;
 807         lexer_token.string.encoding = enc;
 808         lexer_token.string.string   = identify_string(string, size);
 809 }
 810
 811 /**
 812  * Parse a string literal and set lexer_token.
 813  */
 814 static void parse_string_literal(string_encoding_t const enc)
 815 {
 816         parse_string('"', T_STRING_LITERAL, enc, "string literal");
 817 }
 818
 819 /**
 820  * Parse a character constant and set lexer_token.
 821  */
 822 static void parse_character_constant(string_encoding_t const enc)
 823 {
 824         parse_string('\'', T_CHARACTER_CONSTANT, enc, "character constant");
 825         if (lexer_token.string.string.size == 0) {
 826                 errorf(&lexer_token.base.source_position, "empty character constant");
 827         }
 828 }
 829
 830 /**
 831  * Skip a multiline comment.
 832  */
 833 static void skip_multiline_comment(void)
 834 {
 835         while (true) {
 836                 switch (c) {
 837                 case '/':
 838                         next_char();
 839                         if (c == '*') {
 840                                 /* nested comment, warn here */
 841                                 warningf(WARN_COMMENT, &lexer_pos, "'/*' within comment");
 842                         }
 843                         break;
 844                 case '*':
 845                         next_char();
 846                         if (c == '/') {
 847                                 next_char();
 848                                 return;
 849                         }
 850                         break;
 851
 852                 case NEWLINE:
 853                         break;
 854
 855                 case EOF: {
 856                         errorf(&lexer_token.base.source_position,
 857                                "at end of file while looking for comment end");
 858                         return;
 859                 }
 860
 861                 default:
 862                         next_char();
 863                         break;
 864                 }
 865         }
 866 }
 867
 868 /**
 869  * Skip a single line comment.
 870  */
 871 static void skip_line_comment(void)
 872 {
 873         while (true) {
 874                 switch (c) {
 875                 case EOF:
 876                         return;
 877
 878                 case '\n':
 879                 case '\r':
 880                         return;
 881
 882                 case '\\':
 883                         next_char();
 884                         if (c == '\n' || c == '\r') {
 885                                 warningf(WARN_COMMENT, &lexer_pos, "multi-line comment");
 886                                 return;
 887                         }
 888                         break;
 889
 890                 default:
 891                         next_char();
 892                         break;
 893                 }
 894         }
 895 }
 896
 897 /** The current preprocessor token. */
 898 static token_t pp_token;
 899
 900 /**
 901  * Read the next preprocessor token.
 902  */
 903 static inline void next_pp_token(void)
 904 {
 905         lexer_next_preprocessing_token();
 906         pp_token = lexer_token;
 907 }
 908
 909 /**
 910  * Eat all preprocessor tokens until newline.
 911  */
 912 static void eat_until_newline(void)
 913 {
 914         while (pp_token.kind != '\n' && pp_token.kind != T_EOF) {
 915                 next_pp_token();
 916         }
 917 }
 918
 919 /**
 920  * Parse the line directive.
 921  */
 922 static void parse_line_directive(void)
 923 {
 924         if (pp_token.kind != T_INTEGER) {
 925                 parse_error("expected integer");
 926         } else {
 927                 /* use offset -1 as this is about the next line */
 928                 lexer_pos.lineno = atoi(pp_token.number.number.begin) - 1;
 929                 next_pp_token();
 930         }
 931         if (pp_token.kind == T_STRING_LITERAL && pp_token.string.encoding == STRING_ENCODING_CHAR) {
 932                 lexer_pos.input_name = pp_token.string.string.begin;
 933                 lexer_pos.is_system_header = false;
 934                 next_pp_token();
 935
 936                 /* attempt to parse numeric flags as outputted by gcc preprocessor */
 937                 while (pp_token.kind == T_INTEGER) {
 938                         /* flags:
 939                          * 1 - indicates start of a new file
 940                          * 2 - indicates return from a file
 941                          * 3 - indicates system header
 942                          * 4 - indicates implicit extern "C" in C++ mode
 943                          *
 944                          * currently we're only interested in "3"
 945                          */
 946                         if (streq(pp_token.number.number.begin, "3")) {
 947                                 lexer_pos.is_system_header = true;
 948                         }
 949                         next_pp_token();
 950                 }
 951         }
 952
 953         eat_until_newline();
 954 }
 955
 956 /**
 957  * STDC pragmas.
 958  */
 959 typedef enum stdc_pragma_kind_t {
 960         STDC_UNKNOWN,
 961         STDC_FP_CONTRACT,
 962         STDC_FENV_ACCESS,
 963         STDC_CX_LIMITED_RANGE
 964 } stdc_pragma_kind_t;
 965
 966 /**
 967  * STDC pragma values.
 968  */
 969 typedef enum stdc_pragma_value_kind_t {
 970         STDC_VALUE_UNKNOWN,
 971         STDC_VALUE_ON,
 972         STDC_VALUE_OFF,
 973         STDC_VALUE_DEFAULT
 974 } stdc_pragma_value_kind_t;
 975
 976 /**
 977  * Parse a pragma directive.
 978  */
 979 static void parse_pragma(void)
 980 {
 981         next_pp_token();
 982         if (pp_token.kind != T_IDENTIFIER) {
 983                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
 984                          "expected identifier after #pragma");
 985                 eat_until_newline();
 986                 return;
 987         }
 988
 989         stdc_pragma_kind_t kind = STDC_UNKNOWN;
 990         if (pp_token.base.symbol->pp_ID == TP_STDC && c_mode & _C99) {
 991                 /* a STDC pragma */
 992                 next_pp_token();
 993
 994                 switch (pp_token.base.symbol->pp_ID) {
 995                 case TP_FP_CONTRACT:      kind = STDC_FP_CONTRACT;      break;
 996                 case TP_FENV_ACCESS:      kind = STDC_FENV_ACCESS;      break;
 997                 case TP_CX_LIMITED_RANGE: kind = STDC_CX_LIMITED_RANGE; break;
 998                 default:                  break;
 999                 }
1000                 if (kind != STDC_UNKNOWN) {
1001                         next_pp_token();
1002                         stdc_pragma_value_kind_t value;
1003                         switch (pp_token.base.symbol->pp_ID) {
1004                         case TP_ON:      value = STDC_VALUE_ON;      break;
1005                         case TP_OFF:     value = STDC_VALUE_OFF;     break;
1006                         case TP_DEFAULT: value = STDC_VALUE_DEFAULT; break;
1007                         default:         value = STDC_VALUE_UNKNOWN; break;
1008                         }
1009                         if (value == STDC_VALUE_UNKNOWN) {
1010                                 kind = STDC_UNKNOWN;
1011                                 errorf(&pp_token.base.source_position, "bad STDC pragma argument");
1012                         }
1013                 }
1014         }
1015         eat_until_newline();
1016         if (kind == STDC_UNKNOWN) {
1017                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
1018                          "encountered unknown #pragma");
1019         }
1020 }
1021
1022 /**
1023  * Parse a preprocessor non-null directive.
1024  */
1025 static void parse_preprocessor_identifier(void)
1026 {
1027         assert(pp_token.kind == T_IDENTIFIER);
1028         switch (pp_token.base.symbol->pp_ID) {
1029         case TP_line:
1030                 next_pp_token();
1031                 parse_line_directive();
1032                 break;
1033         case TP_pragma:
1034                 parse_pragma();
1035                 break;
1036         case TP_error:
1037                 /* TODO; output the rest of the line */
1038                 parse_error("#error directive");
1039                 break;
1040         }
1041 }
1042
1043 /**
1044  * Parse a preprocessor directive.
1045  */
1046 static void parse_preprocessor_directive(void)
1047 {
1048         next_pp_token();
1049
1050         switch (pp_token.kind) {
1051         case T_IDENTIFIER:
1052                 parse_preprocessor_identifier();
1053                 break;
1054         case T_INTEGER:
1055                 parse_line_directive();
1056                 break;
1057         case '\n':
1058                 /* NULL directive, see §6.10.7 */
1059                 break;
1060         default:
1061                 parse_error("invalid preprocessor directive");
1062                 eat_until_newline();
1063                 break;
1064         }
1065 }
1066
1067 #define MAYBE_PROLOG                                       \
1068                         next_char();                                   \
1069                         while (true) {                                 \
1070                                 switch (c) {
1071
1072 #define MAYBE(ch, set_type)                                \
1073                                 case ch:                                   \
1074                                         next_char();                           \
1075                                         lexer_token.kind = set_type;           \
1076                                         return;
1077
1078 /* must use this as last thing */
1079 #define MAYBE_MODE(ch, set_type, mode)                     \
1080                                 case ch:                                   \
1081                                         if (c_mode & mode) {                   \
1082                                                 next_char();                       \
1083                                                 lexer_token.kind = set_type;       \
1084                                                 return;                            \
1085                                         }                                      \
1086                                         /* fallthrough */
1087
1088 #define ELSE_CODE(code)                                    \
1089                                 default:                                   \
1090                                         code                                   \
1091                                         return;                                \
1092                                 }                                          \
1093                         } /* end of while (true) */                    \
1094
1095 #define ELSE(set_type)                                     \
1096                 ELSE_CODE(                                         \
1097                         lexer_token.kind = set_type;                   \
1098                 )
1099
1100 void lexer_next_preprocessing_token(void)
1101 {
1102         while (true) {
1103                 lexer_token.base.source_position = lexer_pos;
1104                 lexer_token.base.symbol          = NULL;
1105
1106                 switch (c) {
1107                 case ' ':
1108                 case '\t':
1109                         next_char();
1110                         break;
1111
1112                 case NEWLINE:
1113                         lexer_token.kind = '\n';
1114                         return;
1115
1116                 SYMBOL_CHARS {
1117                         parse_symbol();
1118                         /* might be a wide string ( L"string" ) */
1119                         string_encoding_t const enc = STRING_ENCODING_WIDE;
1120                         if (lexer_token.base.symbol == symbol_L) {
1121                                 switch (c) {
1122                                 case '"':  parse_string_literal(enc);     break;
1123                                 case '\'': parse_character_constant(enc); break;
1124                                 }
1125                         }
1126                         return;
1127                 }
1128
1129                 DIGITS
1130                         parse_number();
1131                         return;
1132
1133                 case '"':
1134                         parse_string_literal(STRING_ENCODING_CHAR);
1135                         return;
1136
1137                 case '\'':
1138                         parse_character_constant(STRING_ENCODING_CHAR);
1139                         return;
1140
1141                 case '.':
1142                         MAYBE_PROLOG
1143                                 DIGITS
1144                                         put_back(c);
1145                                         c = '.';
1146                                         parse_number();
1147                                         return;
1148
1149                                 case '.':
1150                                         MAYBE_PROLOG
1151                                         MAYBE('.', T_DOTDOTDOT)
1152                                         ELSE_CODE(
1153                                                 put_back(c);
1154                                                 c = '.';
1155                                                 lexer_token.kind = '.';
1156                                         )
1157                         ELSE('.')
1158                 case '&':
1159                         MAYBE_PROLOG
1160                         MAYBE('&', T_ANDAND)
1161                         MAYBE('=', T_ANDEQUAL)
1162                         ELSE('&')
1163                 case '*':
1164                         MAYBE_PROLOG
1165                         MAYBE('=', T_ASTERISKEQUAL)
1166                         ELSE('*')
1167                 case '+':
1168                         MAYBE_PROLOG
1169                         MAYBE('+', T_PLUSPLUS)
1170                         MAYBE('=', T_PLUSEQUAL)
1171                         ELSE('+')
1172                 case '-':
1173                         MAYBE_PROLOG
1174                         MAYBE('>', T_MINUSGREATER)
1175                         MAYBE('-', T_MINUSMINUS)
1176                         MAYBE('=', T_MINUSEQUAL)
1177                         ELSE('-')
1178                 case '!':
1179                         MAYBE_PROLOG
1180                         MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1181                         ELSE('!')
1182                 case '/':
1183                         MAYBE_PROLOG
1184                         MAYBE('=', T_SLASHEQUAL)
1185                                 case '*':
1186                                         next_char();
1187                                         skip_multiline_comment();
1188                                         lexer_next_preprocessing_token();
1189                                         return;
1190                                 case '/':
1191                                         next_char();
1192                                         skip_line_comment();
1193                                         lexer_next_preprocessing_token();
1194                                         return;
1195                         ELSE('/')
1196                 case '%':
1197                         MAYBE_PROLOG
1198                         MAYBE('>', '}')
1199                         MAYBE('=', T_PERCENTEQUAL)
1200                                 case ':':
1201                                         MAYBE_PROLOG
1202                                                 case '%':
1203                                                         MAYBE_PROLOG
1204                                                         MAYBE(':', T_HASHHASH)
1205                                                         ELSE_CODE(
1206                                                                 put_back(c);
1207                                                                 c = '%';
1208                                                                 lexer_token.kind = '#';
1209                                                         )
1210                                         ELSE('#')
1211                         ELSE('%')
1212                 case '<':
1213                         MAYBE_PROLOG
1214                         MAYBE(':', '[')
1215                         MAYBE('%', '{')
1216                         MAYBE('=', T_LESSEQUAL)
1217                                 case '<':
1218                                         MAYBE_PROLOG
1219                                         MAYBE('=', T_LESSLESSEQUAL)
1220                                         ELSE(T_LESSLESS)
1221                         ELSE('<')
1222                 case '>':
1223                         MAYBE_PROLOG
1224                         MAYBE('=', T_GREATEREQUAL)
1225                                 case '>':
1226                                         MAYBE_PROLOG
1227                                         MAYBE('=', T_GREATERGREATEREQUAL)
1228                                         ELSE(T_GREATERGREATER)
1229                         ELSE('>')
1230                 case '^':
1231                         MAYBE_PROLOG
1232                         MAYBE('=', T_CARETEQUAL)
1233                         ELSE('^')
1234                 case '|':
1235                         MAYBE_PROLOG
1236                         MAYBE('=', T_PIPEEQUAL)
1237                         MAYBE('|', T_PIPEPIPE)
1238                         ELSE('|')
1239                 case ':':
1240                         MAYBE_PROLOG
1241                         MAYBE('>', ']')
1242                         MAYBE_MODE(':', T_COLONCOLON, _CXX)
1243                         ELSE(':')
1244                 case '=':
1245                         MAYBE_PROLOG
1246                         MAYBE('=', T_EQUALEQUAL)
1247                         ELSE('=')
1248                 case '#':
1249                         MAYBE_PROLOG
1250                         MAYBE('#', T_HASHHASH)
1251                         ELSE('#')
1252
1253                 case '\\':
1254                         next_char();
1255                         if (c == 'U' || c == 'u') {
1256                                 put_back(c);
1257                                 c = '\\';
1258                                 parse_symbol();
1259                         } else {
1260                                 lexer_token.kind = '\\';
1261                         }
1262                         return;
1263
1264                 case '?':
1265                 case '[':
1266                 case ']':
1267                 case '(':
1268                 case ')':
1269                 case '{':
1270                 case '}':
1271                 case '~':
1272                 case ';':
1273                 case ',':
1274                         lexer_token.kind = c;
1275                         next_char();
1276                         return;
1277
1278                 case EOF:
1279                         lexer_token.kind = T_EOF;
1280                         return;
1281
1282                 default:
1283 dollar_sign:
1284                         errorf(&lexer_pos, "unknown character '%c' found", c);
1285                         next_char();
1286                         break;
1287                 }
1288         }
1289 }
1290
1291 void lexer_next_token(void)
1292 {
1293         lexer_next_preprocessing_token();
1294
1295         while (lexer_token.kind == '\n') {
1296 newline_found:
1297                 lexer_next_preprocessing_token();
1298         }
1299
1300         if (lexer_token.kind == '#') {
1301                 parse_preprocessor_directive();
1302                 goto newline_found;
1303         }
1304 }
1305
1306 void init_lexer(void)
1307 {
1308         strset_init(&stringset);
1309         symbol_L = symbol_table_insert("L");
1310 }
1311
1312 static void input_error(unsigned delta_lines, unsigned delta_cols,
1313                         const char *message)
1314 {
1315         lexer_pos.lineno += delta_lines;
1316         lexer_pos.colno  += delta_cols;
1317         errorf(&lexer_pos, "%s", message);
1318 }
1319
1320 void lexer_switch_input(input_t *new_input, const char *input_name)
1321 {
1322         lexer_pos.lineno     = 0;
1323         lexer_pos.colno      = 0;
1324         lexer_pos.input_name = input_name;
1325
1326         set_input_error_callback(input_error);
1327         input  = new_input;
1328         bufpos = NULL;
1329         bufend = NULL;
1330
1331         /* place a virtual \n at the beginning so the lexer knows that we're
1332          * at the beginning of a line */
1333         c = '\n';
1334 }
1335
1336 void exit_lexer(void)
1337 {
1338         strset_destroy(&stringset);
1339 }
1340
1341 static __attribute__((unused))
1342 void dbg_pos(const source_position_t source_position)
1343 {
1344         fprintf(stdout, "%s:%u:%u\n", source_position.input_name,
1345                 source_position.lineno, (unsigned)source_position.colno);
1346         fflush(stdout);
1347 }