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 sym_make_string(string_encoding_t const enc)
 406 {
 407         obstack_1grow(&symbol_obstack, '\0');
 408         size_t const len    = obstack_object_size(&symbol_obstack) - 1;
 409         char  *const string = obstack_finish(&symbol_obstack);
 410
 411         /* TODO hash */
 412 #if 0
 413         const char *result = strset_insert(&stringset, concat);
 414         if (result != concat) {
 415                 obstack_free(&symbol_obstack, concat);
 416         }
 417 #else
 418         const char *result = string;
 419 #endif
 420         return (string_t){ result, len, enc };
 421 }
 422
 423 /**
 424  * parse suffixes like 'LU' or 'f' after numbers
 425  */
 426 static void parse_number_suffix(void)
 427 {
 428         assert(obstack_object_size(&symbol_obstack) == 0);
 429         while (true) {
 430                 switch (c) {
 431                 SYMBOL_CHARS
 432                         obstack_1grow(&symbol_obstack, (char) c);
 433                         next_char();
 434                         break;
 435                 default:
 436                 dollar_sign:
 437                         goto finish_suffix;
 438                 }
 439         }
 440 finish_suffix:
 441         if (obstack_object_size(&symbol_obstack) == 0) {
 442                 lexer_token.number.suffix.begin = NULL;
 443                 lexer_token.number.suffix.size  = 0;
 444                 return;
 445         }
 446
 447         lexer_token.number.suffix = sym_make_string(STRING_ENCODING_CHAR);
 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
 503         lexer_token.number.number = sym_make_string(STRING_ENCODING_CHAR);
 504
 505         lexer_token.kind = is_float ? T_FLOATINGPOINT : T_INTEGER;
 506
 507         if (!has_digits) {
 508                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'", &lexer_token.number.number);
 509                 lexer_token.number.number.begin = "0";
 510                 lexer_token.number.number.size  = 1;
 511         }
 512
 513         parse_number_suffix();
 514 }
 515
 516 static void parse_number_bin(void)
 517 {
 518         bool has_digits = false;
 519
 520         while (c == '0' || c == '1') {
 521                 has_digits = true;
 522                 obstack_1grow(&symbol_obstack, (char)c);
 523                 next_char();
 524         }
 525
 526         lexer_token.number.number = sym_make_string(STRING_ENCODING_CHAR);
 527         lexer_token.kind          = T_INTEGER;
 528
 529         if (!has_digits) {
 530                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'", &lexer_token.number.number);
 531                 lexer_token.number.number.begin = "0";
 532                 lexer_token.number.number.size  = 1;
 533         }
 534
 535         parse_number_suffix();
 536 }
 537
 538 /**
 539  * Returns true if the given char is a octal digit.
 540  *
 541  * @param char  the character to check
 542  */
 543 static bool is_octal_digit(utf32 chr)
 544 {
 545         return '0' <= chr && chr <= '7';
 546 }
 547
 548 /**
 549  * Parses a number and sets the lexer_token.
 550  */
 551 static void parse_number(void)
 552 {
 553         bool is_float   = false;
 554         bool has_digits = false;
 555
 556         assert(obstack_object_size(&symbol_obstack) == 0);
 557         if (c == '0') {
 558                 obstack_1grow(&symbol_obstack, (char)c);
 559                 next_char();
 560                 if (c == 'x' || c == 'X') {
 561                         obstack_1grow(&symbol_obstack, (char)c);
 562                         next_char();
 563                         parse_number_hex();
 564                         return;
 565                 } else if (c == 'b' || c == 'B') {
 566                         /* GCC extension: binary constant 0x[bB][01]+.  */
 567                         obstack_1grow(&symbol_obstack, (char)c);
 568                         next_char();
 569                         parse_number_bin();
 570                         return;
 571                 }
 572                 has_digits = true;
 573         }
 574
 575         while (isdigit(c)) {
 576                 has_digits = true;
 577                 obstack_1grow(&symbol_obstack, (char) c);
 578                 next_char();
 579         }
 580
 581         if (c == '.') {
 582                 is_float = true;
 583                 obstack_1grow(&symbol_obstack, '.');
 584                 next_char();
 585
 586                 while (isdigit(c)) {
 587                         has_digits = true;
 588                         obstack_1grow(&symbol_obstack, (char) c);
 589                         next_char();
 590                 }
 591         }
 592         if (c == 'e' || c == 'E') {
 593                 is_float = true;
 594                 obstack_1grow(&symbol_obstack, 'e');
 595                 next_char();
 596                 parse_exponent();
 597         }
 598
 599         lexer_token.number.number = sym_make_string(STRING_ENCODING_CHAR);
 600
 601         if (is_float) {
 602                 lexer_token.kind = T_FLOATINGPOINT;
 603         } else {
 604                 lexer_token.kind = T_INTEGER;
 605
 606                 if (lexer_token.number.number.begin[0] == '0') {
 607                         /* check for invalid octal digits */
 608                         for (size_t i= 0; i < lexer_token.number.number.size; ++i) {
 609                                 char t = lexer_token.number.number.begin[i];
 610                                 if (t >= '8')
 611                                         errorf(&lexer_token.base.source_position, "invalid digit '%c' in octal number", t);
 612                         }
 613                 }
 614         }
 615
 616         if (!has_digits) {
 617                 errorf(&lexer_token.base.source_position, "invalid number literal '%S'",
 618                        &lexer_token.number.number);
 619         }
 620
 621         parse_number_suffix();
 622 }
 623
 624 /**
 625  * Returns the value of a digit.
 626  * The only portable way to do it ...
 627  */
 628 static int digit_value(utf32 const digit)
 629 {
 630         switch (digit) {
 631         case '0': return 0;
 632         case '1': return 1;
 633         case '2': return 2;
 634         case '3': return 3;
 635         case '4': return 4;
 636         case '5': return 5;
 637         case '6': return 6;
 638         case '7': return 7;
 639         case '8': return 8;
 640         case '9': return 9;
 641         case 'a':
 642         case 'A': return 10;
 643         case 'b':
 644         case 'B': return 11;
 645         case 'c':
 646         case 'C': return 12;
 647         case 'd':
 648         case 'D': return 13;
 649         case 'e':
 650         case 'E': return 14;
 651         case 'f':
 652         case 'F': return 15;
 653         default:
 654                 internal_error("wrong character given");
 655         }
 656 }
 657
 658 /**
 659  * Parses an octal character sequence.
 660  *
 661  * @param first_digit  the already read first digit
 662  */
 663 static utf32 parse_octal_sequence(utf32 const first_digit)
 664 {
 665         assert(is_octal_digit(first_digit));
 666         utf32 value = digit_value(first_digit);
 667         if (!is_octal_digit(c)) return value;
 668         value = 8 * value + digit_value(c);
 669         next_char();
 670         if (!is_octal_digit(c)) return value;
 671         value = 8 * value + digit_value(c);
 672         next_char();
 673         return value;
 674 }
 675
 676 /**
 677  * Parses a hex character sequence.
 678  */
 679 static utf32 parse_hex_sequence(void)
 680 {
 681         utf32 value = 0;
 682         while (isxdigit(c)) {
 683                 value = 16 * value + digit_value(c);
 684                 next_char();
 685         }
 686         return value;
 687 }
 688
 689 /**
 690  * Parse an escape sequence.
 691  */
 692 static utf32 parse_escape_sequence(void)
 693 {
 694         eat('\\');
 695
 696         utf32 const ec = c;
 697         next_char();
 698
 699         switch (ec) {
 700         case '"':  return '"';
 701         case '\'': return '\'';
 702         case '\\': return '\\';
 703         case '?': return '\?';
 704         case 'a': return '\a';
 705         case 'b': return '\b';
 706         case 'f': return '\f';
 707         case 'n': return '\n';
 708         case 'r': return '\r';
 709         case 't': return '\t';
 710         case 'v': return '\v';
 711         case 'x':
 712                 return parse_hex_sequence();
 713         case '0':
 714         case '1':
 715         case '2':
 716         case '3':
 717         case '4':
 718         case '5':
 719         case '6':
 720         case '7':
 721                 return parse_octal_sequence(ec);
 722         case EOF:
 723                 parse_error("reached end of file while parsing escape sequence");
 724                 return EOF;
 725         /* \E is not documented, but handled, by GCC.  It is acceptable according
 726          * to §6.11.4, whereas \e is not. */
 727         case 'E':
 728         case 'e':
 729                 if (c_mode & _GNUC)
 730                         return 27;   /* hopefully 27 is ALWAYS the code for ESCAPE */
 731                 break;
 732
 733         case 'U': return parse_universal_char(8);
 734         case 'u': return parse_universal_char(4);
 735
 736         default:
 737                 break;
 738         }
 739         /* §6.4.4.4:8 footnote 64 */
 740         parse_error("unknown escape sequence");
 741         return EOF;
 742 }
 743
 744 string_t make_string(const char *string)
 745 {
 746         obstack_grow(&symbol_obstack, string, strlen(string));
 747         return sym_make_string(STRING_ENCODING_CHAR);
 748 }
 749
 750 static void parse_string(utf32 const delim, token_kind_t const kind, string_encoding_t const enc, char const *const context)
 751 {
 752         eat(delim);
 753
 754         while (true) {
 755                 switch (c) {
 756                 case '\\': {
 757                         utf32 const tc = parse_escape_sequence();
 758                         if (enc == STRING_ENCODING_CHAR) {
 759                                 if (tc >= 0x100) {
 760                                         warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 761                                 }
 762                                 obstack_1grow(&symbol_obstack, tc);
 763                         } else {
 764                                 obstack_grow_symbol(&symbol_obstack, tc);
 765                         }
 766                         break;
 767                 }
 768
 769                 case NEWLINE:
 770                         errorf(&lexer_pos, "newline while parsing %s", context);
 771                         break;
 772
 773                 case EOF:
 774                         errorf(&lexer_token.base.source_position, "EOF while parsing %s", context);
 775                         goto end_of_string;
 776
 777                 default:
 778                         if (c == delim) {
 779                                 next_char();
 780                                 goto end_of_string;
 781                         } else {
 782                                 obstack_grow_symbol(&symbol_obstack, c);
 783                                 next_char();
 784                                 break;
 785                         }
 786                 }
 787         }
 788
 789 end_of_string:
 790         lexer_token.kind          = kind;
 791         lexer_token.string.string = sym_make_string(enc);
 792 }
 793
 794 /**
 795  * Parse a string literal and set lexer_token.
 796  */
 797 static void parse_string_literal(string_encoding_t const enc)
 798 {
 799         parse_string('"', T_STRING_LITERAL, enc, "string literal");
 800 }
 801
 802 /**
 803  * Parse a character constant and set lexer_token.
 804  */
 805 static void parse_character_constant(string_encoding_t const enc)
 806 {
 807         parse_string('\'', T_CHARACTER_CONSTANT, enc, "character constant");
 808         if (lexer_token.string.string.size == 0) {
 809                 errorf(&lexer_token.base.source_position, "empty character constant");
 810         }
 811 }
 812
 813 /**
 814  * Skip a multiline comment.
 815  */
 816 static void skip_multiline_comment(void)
 817 {
 818         while (true) {
 819                 switch (c) {
 820                 case '/':
 821                         next_char();
 822                         if (c == '*') {
 823                                 /* nested comment, warn here */
 824                                 warningf(WARN_COMMENT, &lexer_pos, "'/*' within comment");
 825                         }
 826                         break;
 827                 case '*':
 828                         next_char();
 829                         if (c == '/') {
 830                                 next_char();
 831                                 return;
 832                         }
 833                         break;
 834
 835                 case NEWLINE:
 836                         break;
 837
 838                 case EOF: {
 839                         errorf(&lexer_token.base.source_position,
 840                                "at end of file while looking for comment end");
 841                         return;
 842                 }
 843
 844                 default:
 845                         next_char();
 846                         break;
 847                 }
 848         }
 849 }
 850
 851 /**
 852  * Skip a single line comment.
 853  */
 854 static void skip_line_comment(void)
 855 {
 856         while (true) {
 857                 switch (c) {
 858                 case EOF:
 859                         return;
 860
 861                 case '\n':
 862                 case '\r':
 863                         return;
 864
 865                 case '\\':
 866                         next_char();
 867                         if (c == '\n' || c == '\r') {
 868                                 warningf(WARN_COMMENT, &lexer_pos, "multi-line comment");
 869                                 return;
 870                         }
 871                         break;
 872
 873                 default:
 874                         next_char();
 875                         break;
 876                 }
 877         }
 878 }
 879
 880 /** The current preprocessor token. */
 881 static token_t pp_token;
 882
 883 /**
 884  * Read the next preprocessor token.
 885  */
 886 static inline void next_pp_token(void)
 887 {
 888         lexer_next_preprocessing_token();
 889         pp_token = lexer_token;
 890 }
 891
 892 /**
 893  * Eat all preprocessor tokens until newline.
 894  */
 895 static void eat_until_newline(void)
 896 {
 897         while (pp_token.kind != '\n' && pp_token.kind != T_EOF) {
 898                 next_pp_token();
 899         }
 900 }
 901
 902 /**
 903  * Parse the line directive.
 904  */
 905 static void parse_line_directive(void)
 906 {
 907         if (pp_token.kind != T_INTEGER) {
 908                 parse_error("expected integer");
 909         } else {
 910                 /* use offset -1 as this is about the next line */
 911                 lexer_pos.lineno = atoi(pp_token.number.number.begin) - 1;
 912                 next_pp_token();
 913         }
 914         if (pp_token.kind == T_STRING_LITERAL && pp_token.string.string.encoding == STRING_ENCODING_CHAR) {
 915                 lexer_pos.input_name       = pp_token.string.string.begin;
 916                 lexer_pos.is_system_header = false;
 917                 next_pp_token();
 918
 919                 /* attempt to parse numeric flags as outputted by gcc preprocessor */
 920                 while (pp_token.kind == T_INTEGER) {
 921                         /* flags:
 922                          * 1 - indicates start of a new file
 923                          * 2 - indicates return from a file
 924                          * 3 - indicates system header
 925                          * 4 - indicates implicit extern "C" in C++ mode
 926                          *
 927                          * currently we're only interested in "3"
 928                          */
 929                         if (streq(pp_token.number.number.begin, "3")) {
 930                                 lexer_pos.is_system_header = true;
 931                         }
 932                         next_pp_token();
 933                 }
 934         }
 935
 936         eat_until_newline();
 937 }
 938
 939 /**
 940  * STDC pragmas.
 941  */
 942 typedef enum stdc_pragma_kind_t {
 943         STDC_UNKNOWN,
 944         STDC_FP_CONTRACT,
 945         STDC_FENV_ACCESS,
 946         STDC_CX_LIMITED_RANGE
 947 } stdc_pragma_kind_t;
 948
 949 /**
 950  * STDC pragma values.
 951  */
 952 typedef enum stdc_pragma_value_kind_t {
 953         STDC_VALUE_UNKNOWN,
 954         STDC_VALUE_ON,
 955         STDC_VALUE_OFF,
 956         STDC_VALUE_DEFAULT
 957 } stdc_pragma_value_kind_t;
 958
 959 /**
 960  * Parse a pragma directive.
 961  */
 962 static void parse_pragma(void)
 963 {
 964         next_pp_token();
 965         if (pp_token.kind != T_IDENTIFIER) {
 966                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
 967                          "expected identifier after #pragma");
 968                 eat_until_newline();
 969                 return;
 970         }
 971
 972         stdc_pragma_kind_t kind = STDC_UNKNOWN;
 973         if (pp_token.base.symbol->pp_ID == TP_STDC && c_mode & _C99) {
 974                 /* a STDC pragma */
 975                 next_pp_token();
 976
 977                 switch (pp_token.base.symbol->pp_ID) {
 978                 case TP_FP_CONTRACT:      kind = STDC_FP_CONTRACT;      break;
 979                 case TP_FENV_ACCESS:      kind = STDC_FENV_ACCESS;      break;
 980                 case TP_CX_LIMITED_RANGE: kind = STDC_CX_LIMITED_RANGE; break;
 981                 default:                  break;
 982                 }
 983                 if (kind != STDC_UNKNOWN) {
 984                         next_pp_token();
 985                         stdc_pragma_value_kind_t value;
 986                         switch (pp_token.base.symbol->pp_ID) {
 987                         case TP_ON:      value = STDC_VALUE_ON;      break;
 988                         case TP_OFF:     value = STDC_VALUE_OFF;     break;
 989                         case TP_DEFAULT: value = STDC_VALUE_DEFAULT; break;
 990                         default:         value = STDC_VALUE_UNKNOWN; break;
 991                         }
 992                         if (value == STDC_VALUE_UNKNOWN) {
 993                                 kind = STDC_UNKNOWN;
 994                                 errorf(&pp_token.base.source_position, "bad STDC pragma argument");
 995                         }
 996                 }
 997         }
 998         eat_until_newline();
 999         if (kind == STDC_UNKNOWN) {
1000                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
1001                          "encountered unknown #pragma");
1002         }
1003 }
1004
1005 /**
1006  * Parse a preprocessor non-null directive.
1007  */
1008 static void parse_preprocessor_identifier(void)
1009 {
1010         assert(pp_token.kind == T_IDENTIFIER);
1011         switch (pp_token.base.symbol->pp_ID) {
1012         case TP_line:
1013                 next_pp_token();
1014                 parse_line_directive();
1015                 break;
1016         case TP_pragma:
1017                 parse_pragma();
1018                 break;
1019         case TP_error:
1020                 /* TODO; output the rest of the line */
1021                 parse_error("#error directive");
1022                 break;
1023         }
1024 }
1025
1026 /**
1027  * Parse a preprocessor directive.
1028  */
1029 static void parse_preprocessor_directive(void)
1030 {
1031         next_pp_token();
1032
1033         switch (pp_token.kind) {
1034         case T_IDENTIFIER:
1035                 parse_preprocessor_identifier();
1036                 break;
1037         case T_INTEGER:
1038                 parse_line_directive();
1039                 break;
1040         case '\n':
1041                 /* NULL directive, see §6.10.7 */
1042                 break;
1043         default:
1044                 parse_error("invalid preprocessor directive");
1045                 eat_until_newline();
1046                 break;
1047         }
1048 }
1049
1050 #define MAYBE_PROLOG                                       \
1051                         next_char();                                   \
1052                         while (true) {                                 \
1053                                 switch (c) {
1054
1055 #define MAYBE(ch, set_type)                                \
1056                                 case ch:                                   \
1057                                         next_char();                           \
1058                                         lexer_token.kind = set_type;           \
1059                                         return;
1060
1061 /* must use this as last thing */
1062 #define MAYBE_MODE(ch, set_type, mode)                     \
1063                                 case ch:                                   \
1064                                         if (c_mode & mode) {                   \
1065                                                 next_char();                       \
1066                                                 lexer_token.kind = set_type;       \
1067                                                 return;                            \
1068                                         }                                      \
1069                                         /* fallthrough */
1070
1071 #define ELSE_CODE(code)                                    \
1072                                 default:                                   \
1073                                         code                                   \
1074                                         return;                                \
1075                                 }                                          \
1076                         } /* end of while (true) */                    \
1077
1078 #define ELSE(set_type)                                     \
1079                 ELSE_CODE(                                         \
1080                         lexer_token.kind = set_type;                   \
1081                 )
1082
1083 void lexer_next_preprocessing_token(void)
1084 {
1085         while (true) {
1086                 lexer_token.base.source_position = lexer_pos;
1087                 lexer_token.base.symbol          = NULL;
1088
1089                 switch (c) {
1090                 case ' ':
1091                 case '\t':
1092                         next_char();
1093                         break;
1094
1095                 case NEWLINE:
1096                         lexer_token.kind = '\n';
1097                         return;
1098
1099                 SYMBOL_CHARS {
1100                         parse_symbol();
1101                         /* might be a wide string ( L"string" ) */
1102                         string_encoding_t const enc = STRING_ENCODING_WIDE;
1103                         if (lexer_token.base.symbol == symbol_L) {
1104                                 switch (c) {
1105                                 case '"':  parse_string_literal(enc);     break;
1106                                 case '\'': parse_character_constant(enc); break;
1107                                 }
1108                         }
1109                         return;
1110                 }
1111
1112                 DIGITS
1113                         parse_number();
1114                         return;
1115
1116                 case '"':
1117                         parse_string_literal(STRING_ENCODING_CHAR);
1118                         return;
1119
1120                 case '\'':
1121                         parse_character_constant(STRING_ENCODING_CHAR);
1122                         return;
1123
1124                 case '.':
1125                         MAYBE_PROLOG
1126                                 DIGITS
1127                                         put_back(c);
1128                                         c = '.';
1129                                         parse_number();
1130                                         return;
1131
1132                                 case '.':
1133                                         MAYBE_PROLOG
1134                                         MAYBE('.', T_DOTDOTDOT)
1135                                         ELSE_CODE(
1136                                                 put_back(c);
1137                                                 c = '.';
1138                                                 lexer_token.kind = '.';
1139                                         )
1140                         ELSE('.')
1141                 case '&':
1142                         MAYBE_PROLOG
1143                         MAYBE('&', T_ANDAND)
1144                         MAYBE('=', T_ANDEQUAL)
1145                         ELSE('&')
1146                 case '*':
1147                         MAYBE_PROLOG
1148                         MAYBE('=', T_ASTERISKEQUAL)
1149                         ELSE('*')
1150                 case '+':
1151                         MAYBE_PROLOG
1152                         MAYBE('+', T_PLUSPLUS)
1153                         MAYBE('=', T_PLUSEQUAL)
1154                         ELSE('+')
1155                 case '-':
1156                         MAYBE_PROLOG
1157                         MAYBE('>', T_MINUSGREATER)
1158                         MAYBE('-', T_MINUSMINUS)
1159                         MAYBE('=', T_MINUSEQUAL)
1160                         ELSE('-')
1161                 case '!':
1162                         MAYBE_PROLOG
1163                         MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1164                         ELSE('!')
1165                 case '/':
1166                         MAYBE_PROLOG
1167                         MAYBE('=', T_SLASHEQUAL)
1168                                 case '*':
1169                                         next_char();
1170                                         skip_multiline_comment();
1171                                         lexer_next_preprocessing_token();
1172                                         return;
1173                                 case '/':
1174                                         next_char();
1175                                         skip_line_comment();
1176                                         lexer_next_preprocessing_token();
1177                                         return;
1178                         ELSE('/')
1179                 case '%':
1180                         MAYBE_PROLOG
1181                         MAYBE('>', '}')
1182                         MAYBE('=', T_PERCENTEQUAL)
1183                                 case ':':
1184                                         MAYBE_PROLOG
1185                                                 case '%':
1186                                                         MAYBE_PROLOG
1187                                                         MAYBE(':', T_HASHHASH)
1188                                                         ELSE_CODE(
1189                                                                 put_back(c);
1190                                                                 c = '%';
1191                                                                 lexer_token.kind = '#';
1192                                                         )
1193                                         ELSE('#')
1194                         ELSE('%')
1195                 case '<':
1196                         MAYBE_PROLOG
1197                         MAYBE(':', '[')
1198                         MAYBE('%', '{')
1199                         MAYBE('=', T_LESSEQUAL)
1200                                 case '<':
1201                                         MAYBE_PROLOG
1202                                         MAYBE('=', T_LESSLESSEQUAL)
1203                                         ELSE(T_LESSLESS)
1204                         ELSE('<')
1205                 case '>':
1206                         MAYBE_PROLOG
1207                         MAYBE('=', T_GREATEREQUAL)
1208                                 case '>':
1209                                         MAYBE_PROLOG
1210                                         MAYBE('=', T_GREATERGREATEREQUAL)
1211                                         ELSE(T_GREATERGREATER)
1212                         ELSE('>')
1213                 case '^':
1214                         MAYBE_PROLOG
1215                         MAYBE('=', T_CARETEQUAL)
1216                         ELSE('^')
1217                 case '|':
1218                         MAYBE_PROLOG
1219                         MAYBE('=', T_PIPEEQUAL)
1220                         MAYBE('|', T_PIPEPIPE)
1221                         ELSE('|')
1222                 case ':':
1223                         MAYBE_PROLOG
1224                         MAYBE('>', ']')
1225                         MAYBE_MODE(':', T_COLONCOLON, _CXX)
1226                         ELSE(':')
1227                 case '=':
1228                         MAYBE_PROLOG
1229                         MAYBE('=', T_EQUALEQUAL)
1230                         ELSE('=')
1231                 case '#':
1232                         MAYBE_PROLOG
1233                         MAYBE('#', T_HASHHASH)
1234                         ELSE('#')
1235
1236                 case '\\':
1237                         next_char();
1238                         if (c == 'U' || c == 'u') {
1239                                 put_back(c);
1240                                 c = '\\';
1241                                 parse_symbol();
1242                         } else {
1243                                 lexer_token.kind = '\\';
1244                         }
1245                         return;
1246
1247                 case '?':
1248                 case '[':
1249                 case ']':
1250                 case '(':
1251                 case ')':
1252                 case '{':
1253                 case '}':
1254                 case '~':
1255                 case ';':
1256                 case ',':
1257                         lexer_token.kind = c;
1258                         next_char();
1259                         return;
1260
1261                 case EOF:
1262                         lexer_token.kind = T_EOF;
1263                         return;
1264
1265                 default:
1266 dollar_sign:
1267                         errorf(&lexer_pos, "unknown character '%c' found", c);
1268                         next_char();
1269                         break;
1270                 }
1271         }
1272 }
1273
1274 void lexer_next_token(void)
1275 {
1276         lexer_next_preprocessing_token();
1277
1278         while (lexer_token.kind == '\n') {
1279 newline_found:
1280                 lexer_next_preprocessing_token();
1281         }
1282
1283         if (lexer_token.kind == '#') {
1284                 parse_preprocessor_directive();
1285                 goto newline_found;
1286         }
1287 }
1288
1289 void init_lexer(void)
1290 {
1291         strset_init(&stringset);
1292         symbol_L = symbol_table_insert("L");
1293 }
1294
1295 static void input_error(unsigned delta_lines, unsigned delta_cols,
1296                         const char *message)
1297 {
1298         lexer_pos.lineno += delta_lines;
1299         lexer_pos.colno  += delta_cols;
1300         errorf(&lexer_pos, "%s", message);
1301 }
1302
1303 void lexer_switch_input(input_t *new_input, const char *input_name)
1304 {
1305         lexer_pos.lineno     = 0;
1306         lexer_pos.colno      = 0;
1307         lexer_pos.input_name = input_name;
1308
1309         set_input_error_callback(input_error);
1310         input  = new_input;
1311         bufpos = NULL;
1312         bufend = NULL;
1313
1314         /* place a virtual \n at the beginning so the lexer knows that we're
1315          * at the beginning of a line */
1316         c = '\n';
1317 }
1318
1319 void exit_lexer(void)
1320 {
1321         strset_destroy(&stringset);
1322 }
1323
1324 static __attribute__((unused))
1325 void dbg_pos(const source_position_t source_position)
1326 {
1327         fprintf(stdout, "%s:%u:%u\n", source_position.input_name,
1328                 source_position.lineno, (unsigned)source_position.colno);
1329         fflush(stdout);
1330 }