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 "input.h"
  23 #include "diagnostic.h"
  24 #include "lexer.h"
  25 #include "symbol_t.h"
  26 #include "token_t.h"
  27 #include "symbol_table_t.h"
  28 #include "adt/error.h"
  29 #include "adt/strset.h"
  30 #include "adt/util.h"
  31 #include "types.h"
  32 #include "type_t.h"
  33 #include "target_architecture.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 static char              *encoding;
  61 bool                      allow_dollar_in_symbol = true;
  62
  63 /**
  64  * Prints a parse error message at the current token.
  65  *
  66  * @param msg   the error message
  67  */
  68 static void parse_error(const char *msg)
  69 {
  70         errorf(&lexer_pos, "%s", msg);
  71 }
  72
  73 /**
  74  * Prints an internal error message at the current token.
  75  *
  76  * @param msg   the error message
  77  */
  78 static NORETURN internal_error(const char *msg)
  79 {
  80         internal_errorf(&lexer_pos, "%s", msg);
  81 }
  82
  83 static inline void next_real_char(void)
  84 {
  85         assert(bufpos <= bufend);
  86         if (bufpos >= bufend) {
  87                 size_t n = decode(input, input_buf+MAX_PUTBACK, BUF_SIZE);
  88                 if (n == 0) {
  89                         c = EOF;
  90                         return;
  91                 }
  92                 bufpos = input_buf + MAX_PUTBACK;
  93                 bufend = bufpos + n;
  94         }
  95         c = *bufpos++;
  96         ++lexer_pos.colno;
  97 }
  98
  99 /**
 100  * Put a character back into the buffer.
 101  *
 102  * @param pc  the character to put back
 103  */
 104 static inline void put_back(utf32 const pc)
 105 {
 106         *(--bufpos - input_buf + input_buf) = pc;
 107         --lexer_pos.colno;
 108 }
 109
 110 static inline void next_char(void);
 111
 112 #define MATCH_NEWLINE(code)  \
 113         case '\r':               \
 114                 next_char();         \
 115                 if (c == '\n') {     \
 116         case '\n':               \
 117                         next_char();     \
 118                 }                    \
 119                 lexer_pos.lineno++;  \
 120                 lexer_pos.colno = 1; \
 121                 code
 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         MATCH_NEWLINE(return;)
 131
 132         default:
 133                 break;
 134         }
 135
 136         put_back(c);
 137         c = '\\';
 138 }
 139
 140 /**
 141  * Set c to the next input character, ie.
 142  * after expanding trigraphs.
 143  */
 144 static inline void next_char(void)
 145 {
 146         next_real_char();
 147
 148         /* filter trigraphs */
 149         if (UNLIKELY(c == '\\')) {
 150                 maybe_concat_lines();
 151                 return;
 152         }
 153
 154         if (LIKELY(c != '?'))
 155                 return;
 156
 157         next_real_char();
 158         if (LIKELY(c != '?')) {
 159                 put_back(c);
 160                 c = '?';
 161                 return;
 162         }
 163
 164         next_real_char();
 165         switch (c) {
 166         case '=': c = '#'; break;
 167         case '(': c = '['; break;
 168         case '/': c = '\\'; maybe_concat_lines(); break;
 169         case ')': c = ']'; break;
 170         case '\'': c = '^'; break;
 171         case '<': c = '{'; break;
 172         case '!': c = '|'; break;
 173         case '>': c = '}'; break;
 174         case '-': c = '~'; break;
 175         default:
 176                 put_back(c);
 177                 put_back('?');
 178                 c = '?';
 179                 break;
 180         }
 181 }
 182
 183 #define SYMBOL_CHARS  \
 184         case '$': if (!allow_dollar_in_symbol) goto dollar_sign; \
 185         case 'a':         \
 186         case 'b':         \
 187         case 'c':         \
 188         case 'd':         \
 189         case 'e':         \
 190         case 'f':         \
 191         case 'g':         \
 192         case 'h':         \
 193         case 'i':         \
 194         case 'j':         \
 195         case 'k':         \
 196         case 'l':         \
 197         case 'm':         \
 198         case 'n':         \
 199         case 'o':         \
 200         case 'p':         \
 201         case 'q':         \
 202         case 'r':         \
 203         case 's':         \
 204         case 't':         \
 205         case 'u':         \
 206         case 'v':         \
 207         case 'w':         \
 208         case 'x':         \
 209         case 'y':         \
 210         case 'z':         \
 211         case 'A':         \
 212         case 'B':         \
 213         case 'C':         \
 214         case 'D':         \
 215         case 'E':         \
 216         case 'F':         \
 217         case 'G':         \
 218         case 'H':         \
 219         case 'I':         \
 220         case 'J':         \
 221         case 'K':         \
 222         case 'L':         \
 223         case 'M':         \
 224         case 'N':         \
 225         case 'O':         \
 226         case 'P':         \
 227         case 'Q':         \
 228         case 'R':         \
 229         case 'S':         \
 230         case 'T':         \
 231         case 'U':         \
 232         case 'V':         \
 233         case 'W':         \
 234         case 'X':         \
 235         case 'Y':         \
 236         case 'Z':         \
 237         case '_':
 238
 239 #define DIGITS        \
 240         case '0':         \
 241         case '1':         \
 242         case '2':         \
 243         case '3':         \
 244         case '4':         \
 245         case '5':         \
 246         case '6':         \
 247         case '7':         \
 248         case '8':         \
 249         case '9':
 250
 251 /**
 252  * Read a symbol from the input and build
 253  * the lexer_token.
 254  */
 255 static void parse_symbol(void)
 256 {
 257         obstack_1grow(&symbol_obstack, (char) c);
 258         next_char();
 259
 260         while (true) {
 261                 switch (c) {
 262                 DIGITS
 263                 SYMBOL_CHARS
 264                         obstack_1grow(&symbol_obstack, (char) c);
 265                         next_char();
 266                         break;
 267
 268                 default:
 269 dollar_sign:
 270                         goto end_symbol;
 271                 }
 272         }
 273
 274 end_symbol:
 275         obstack_1grow(&symbol_obstack, '\0');
 276
 277         char     *string = obstack_finish(&symbol_obstack);
 278         symbol_t *symbol = symbol_table_insert(string);
 279
 280         lexer_token.type   = symbol->ID;
 281         lexer_token.symbol = symbol;
 282
 283         if (symbol->string != string) {
 284                 obstack_free(&symbol_obstack, string);
 285         }
 286 }
 287
 288 /**
 289  * parse suffixes like 'LU' or 'f' after numbers
 290  */
 291 static void parse_number_suffix(void)
 292 {
 293         assert(obstack_object_size(&symbol_obstack) == 0);
 294         while (true) {
 295                 switch (c) {
 296                 SYMBOL_CHARS
 297                         obstack_1grow(&symbol_obstack, (char) c);
 298                         next_char();
 299                         break;
 300                 default:
 301                 dollar_sign:
 302                         goto finish_suffix;
 303                 }
 304         }
 305 finish_suffix:
 306         if (obstack_object_size(&symbol_obstack) == 0) {
 307                 lexer_token.symbol = NULL;
 308                 return;
 309         }
 310
 311         obstack_1grow(&symbol_obstack, '\0');
 312         char     *string = obstack_finish(&symbol_obstack);
 313         symbol_t *symbol = symbol_table_insert(string);
 314
 315         if (symbol->string != string) {
 316                 obstack_free(&symbol_obstack, string);
 317         }
 318         lexer_token.symbol = symbol;
 319 }
 320
 321 static string_t identify_string(char *string, size_t len)
 322 {
 323         /* TODO hash */
 324 #if 0
 325         const char *result = strset_insert(&stringset, concat);
 326         if (result != concat) {
 327                 obstack_free(&symbol_obstack, concat);
 328         }
 329 #else
 330         const char *result = string;
 331 #endif
 332         return (string_t) {result, len};
 333 }
 334
 335 /**
 336  * Parses a hex number including hex floats and set the
 337  * lexer_token.
 338  */
 339 static void parse_number_hex(void)
 340 {
 341         bool is_float   = false;
 342         bool has_digits = false;
 343
 344         assert(obstack_object_size(&symbol_obstack) == 0);
 345         while (isxdigit(c)) {
 346                 has_digits = true;
 347                 obstack_1grow(&symbol_obstack, (char) c);
 348                 next_char();
 349         }
 350
 351         if (c == '.') {
 352                 is_float = true;
 353                 obstack_1grow(&symbol_obstack, (char) c);
 354                 next_char();
 355
 356                 while (isxdigit(c)) {
 357                         has_digits = true;
 358                         obstack_1grow(&symbol_obstack, (char) c);
 359                         next_char();
 360                 }
 361         }
 362         if (c == 'p' || c == 'P') {
 363                 is_float = true;
 364                 obstack_1grow(&symbol_obstack, (char) c);
 365                 next_char();
 366
 367                 if (c == '-' || c == '+') {
 368                         obstack_1grow(&symbol_obstack, (char) c);
 369                         next_char();
 370                 }
 371
 372                 while (isxdigit(c)) {
 373                         obstack_1grow(&symbol_obstack, (char) c);
 374                         next_char();
 375                 }
 376         } else if (is_float) {
 377                 errorf(&lexer_token.source_position,
 378                        "hexadecimal floatingpoint constant requires an exponent");
 379         }
 380         obstack_1grow(&symbol_obstack, '\0');
 381
 382         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 383         char   *string = obstack_finish(&symbol_obstack);
 384         lexer_token.literal = identify_string(string, size);
 385
 386         lexer_token.type    =
 387                 is_float ? T_FLOATINGPOINT_HEXADECIMAL : T_INTEGER_HEXADECIMAL;
 388
 389         if (!has_digits) {
 390                 errorf(&lexer_token.source_position, "invalid number literal '0x%S'",
 391                        &lexer_token.literal);
 392                 lexer_token.literal.begin = "0";
 393                 lexer_token.literal.size  = 1;
 394         }
 395
 396         parse_number_suffix();
 397 }
 398
 399 /**
 400  * Returns true if the given char is a octal digit.
 401  *
 402  * @param char  the character to check
 403  */
 404 static bool is_octal_digit(utf32 chr)
 405 {
 406         return '0' <= chr && chr <= '7';
 407 }
 408
 409 /**
 410  * Parses a number and sets the lexer_token.
 411  */
 412 static void parse_number(void)
 413 {
 414         bool is_float   = false;
 415         bool has_digits = false;
 416
 417         assert(obstack_object_size(&symbol_obstack) == 0);
 418         if (c == '0') {
 419                 next_char();
 420                 if (c == 'x' || c == 'X') {
 421                         next_char();
 422                         parse_number_hex();
 423                         return;
 424                 } else {
 425                         has_digits = true;
 426                 }
 427                 obstack_1grow(&symbol_obstack, '0');
 428         }
 429
 430         while (isdigit(c)) {
 431                 has_digits = true;
 432                 obstack_1grow(&symbol_obstack, (char) c);
 433                 next_char();
 434         }
 435
 436         if (c == '.') {
 437                 is_float = true;
 438                 obstack_1grow(&symbol_obstack, '.');
 439                 next_char();
 440
 441                 while (isdigit(c)) {
 442                         has_digits = true;
 443                         obstack_1grow(&symbol_obstack, (char) c);
 444                         next_char();
 445                 }
 446         }
 447         if (c == 'e' || c == 'E') {
 448                 is_float = true;
 449                 obstack_1grow(&symbol_obstack, 'e');
 450                 next_char();
 451
 452                 if (c == '-' || c == '+') {
 453                         obstack_1grow(&symbol_obstack, (char) c);
 454                         next_char();
 455                 }
 456
 457                 while (isdigit(c)) {
 458                         obstack_1grow(&symbol_obstack, (char) c);
 459                         next_char();
 460                 }
 461         }
 462
 463         obstack_1grow(&symbol_obstack, '\0');
 464         size_t  size   = obstack_object_size(&symbol_obstack) - 1;
 465         char   *string = obstack_finish(&symbol_obstack);
 466         lexer_token.literal = identify_string(string, size);
 467
 468         /* is it an octal number? */
 469         if (is_float) {
 470                 lexer_token.type = T_FLOATINGPOINT;
 471         } else if (string[0] == '0') {
 472                 lexer_token.type = T_INTEGER_OCTAL;
 473
 474                 /* check for invalid octal digits */
 475                 for (size_t i= 0; i < size; ++i) {
 476                         char t = string[i];
 477                         if (t >= '8')
 478                                 errorf(&lexer_token.source_position,
 479                                        "invalid digit '%c' in octal number", t);
 480                 }
 481         } else {
 482                 lexer_token.type = T_INTEGER;
 483         }
 484
 485         if (!has_digits) {
 486                 errorf(&lexer_token.source_position, "invalid number literal '%S'",
 487                        &lexer_token.literal);
 488         }
 489
 490         parse_number_suffix();
 491 }
 492
 493 /**
 494  * Returns the value of a digit.
 495  * The only portable way to do it ...
 496  */
 497 static int digit_value(utf32 const digit)
 498 {
 499         switch (digit) {
 500         case '0': return 0;
 501         case '1': return 1;
 502         case '2': return 2;
 503         case '3': return 3;
 504         case '4': return 4;
 505         case '5': return 5;
 506         case '6': return 6;
 507         case '7': return 7;
 508         case '8': return 8;
 509         case '9': return 9;
 510         case 'a':
 511         case 'A': return 10;
 512         case 'b':
 513         case 'B': return 11;
 514         case 'c':
 515         case 'C': return 12;
 516         case 'd':
 517         case 'D': return 13;
 518         case 'e':
 519         case 'E': return 14;
 520         case 'f':
 521         case 'F': return 15;
 522         default:
 523                 internal_error("wrong character given");
 524         }
 525 }
 526
 527 /**
 528  * Parses an octal character sequence.
 529  *
 530  * @param first_digit  the already read first digit
 531  */
 532 static utf32 parse_octal_sequence(utf32 const first_digit)
 533 {
 534         assert(is_octal_digit(first_digit));
 535         utf32 value = digit_value(first_digit);
 536         if (!is_octal_digit(c)) return value;
 537         value = 8 * value + digit_value(c);
 538         next_char();
 539         if (!is_octal_digit(c)) return value;
 540         value = 8 * value + digit_value(c);
 541         next_char();
 542         return value;
 543 }
 544
 545 /**
 546  * Parses a hex character sequence.
 547  */
 548 static utf32 parse_hex_sequence(void)
 549 {
 550         utf32 value = 0;
 551         while (isxdigit(c)) {
 552                 value = 16 * value + digit_value(c);
 553                 next_char();
 554         }
 555         return value;
 556 }
 557
 558 /**
 559  * Parse an escape sequence.
 560  */
 561 static utf32 parse_escape_sequence(void)
 562 {
 563         eat('\\');
 564
 565         utf32 const ec = c;
 566         next_char();
 567
 568         switch (ec) {
 569         case '"':  return '"';
 570         case '\'': return '\'';
 571         case '\\': return '\\';
 572         case '?': return '\?';
 573         case 'a': return '\a';
 574         case 'b': return '\b';
 575         case 'f': return '\f';
 576         case 'n': return '\n';
 577         case 'r': return '\r';
 578         case 't': return '\t';
 579         case 'v': return '\v';
 580         case 'x':
 581                 return parse_hex_sequence();
 582         case '0':
 583         case '1':
 584         case '2':
 585         case '3':
 586         case '4':
 587         case '5':
 588         case '6':
 589         case '7':
 590                 return parse_octal_sequence(ec);
 591         case EOF:
 592                 parse_error("reached end of file while parsing escape sequence");
 593                 return EOF;
 594         /* \E is not documented, but handled, by GCC.  It is acceptable according
 595          * to §6.11.4, whereas \e is not. */
 596         case 'E':
 597         case 'e':
 598                 if (c_mode & _GNUC)
 599                         return 27;   /* hopefully 27 is ALWAYS the code for ESCAPE */
 600                 break;
 601         case 'u':
 602         case 'U':
 603                 parse_error("universal character parsing not implemented yet");
 604                 return EOF;
 605         default:
 606                 break;
 607         }
 608         /* §6.4.4.4:8 footnote 64 */
 609         parse_error("unknown escape sequence");
 610         return EOF;
 611 }
 612
 613 /**
 614  * Concatenate two strings.
 615  */
 616 string_t concat_strings(const string_t *const s1, const string_t *const s2)
 617 {
 618         const size_t len1 = s1->size - 1;
 619         const size_t len2 = s2->size - 1;
 620
 621         char *const concat = obstack_alloc(&symbol_obstack, len1 + len2 + 1);
 622         memcpy(concat, s1->begin, len1);
 623         memcpy(concat + len1, s2->begin, len2 + 1);
 624
 625         return identify_string(concat, len1 + len2 + 1);
 626 }
 627
 628 string_t make_string(const char *string)
 629 {
 630         size_t      len   = strlen(string) + 1;
 631         char *const space = obstack_alloc(&symbol_obstack, len);
 632         memcpy(space, string, len);
 633
 634         return identify_string(space, len);
 635 }
 636
 637 static void grow_symbol(utf32 const tc)
 638 {
 639         struct obstack *const o  = &symbol_obstack;
 640         if (tc < 0x80U) {
 641                 obstack_1grow(o, tc);
 642         } else if (tc < 0x800) {
 643                 obstack_1grow(o, 0xC0 | (tc >> 6));
 644                 obstack_1grow(o, 0x80 | (tc & 0x3F));
 645         } else if (tc < 0x10000) {
 646                 obstack_1grow(o, 0xE0 | ( tc >> 12));
 647                 obstack_1grow(o, 0x80 | ((tc >>  6) & 0x3F));
 648                 obstack_1grow(o, 0x80 | ( tc        & 0x3F));
 649         } else {
 650                 obstack_1grow(o, 0xF0 | ( tc >> 18));
 651                 obstack_1grow(o, 0x80 | ((tc >> 12) & 0x3F));
 652                 obstack_1grow(o, 0x80 | ((tc >>  6) & 0x3F));
 653                 obstack_1grow(o, 0x80 | ( tc        & 0x3F));
 654         }
 655 }
 656
 657 /**
 658  * Parse a string literal and set lexer_token.
 659  */
 660 static void parse_string_literal(void)
 661 {
 662         eat('"');
 663
 664         while (true) {
 665                 switch (c) {
 666                 case '\\': {
 667                         utf32 const tc = parse_escape_sequence();
 668                         if (tc >= 0x100) {
 669                                 warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 670                         }
 671                         obstack_1grow(&symbol_obstack, tc);
 672                         break;
 673                 }
 674
 675                 case EOF: {
 676                         errorf(&lexer_token.source_position, "string has no end");
 677                         lexer_token.type = T_ERROR;
 678                         return;
 679                 }
 680
 681                 case '"':
 682                         next_char();
 683                         goto end_of_string;
 684
 685                 default:
 686                         grow_symbol(c);
 687                         next_char();
 688                         break;
 689                 }
 690         }
 691
 692 end_of_string:
 693
 694         /* TODO: concatenate multiple strings separated by whitespace... */
 695
 696         /* add finishing 0 to the string */
 697         obstack_1grow(&symbol_obstack, '\0');
 698         const size_t  size   = (size_t)obstack_object_size(&symbol_obstack);
 699         char         *string = obstack_finish(&symbol_obstack);
 700
 701         lexer_token.type    = T_STRING_LITERAL;
 702         lexer_token.literal = identify_string(string, size);
 703 }
 704
 705 /**
 706  * Parse a wide character constant and set lexer_token.
 707  */
 708 static void parse_wide_character_constant(void)
 709 {
 710         eat('\'');
 711
 712         while (true) {
 713                 switch (c) {
 714                 case '\\': {
 715                         const utf32 tc = parse_escape_sequence();
 716                         grow_symbol(tc);
 717                         break;
 718                 }
 719
 720                 MATCH_NEWLINE(
 721                         parse_error("newline while parsing character constant");
 722                         break;
 723                 )
 724
 725                 case '\'':
 726                         next_char();
 727                         goto end_of_wide_char_constant;
 728
 729                 case EOF: {
 730                         errorf(&lexer_token.source_position, "EOF while parsing character constant");
 731                         lexer_token.type = T_ERROR;
 732                         return;
 733                 }
 734
 735                 default:
 736                         grow_symbol(c);
 737                         next_char();
 738                         break;
 739                 }
 740         }
 741
 742 end_of_wide_char_constant:;
 743         obstack_1grow(&symbol_obstack, '\0');
 744         size_t  size   = (size_t) obstack_object_size(&symbol_obstack) - 1;
 745         char   *string = obstack_finish(&symbol_obstack);
 746
 747         lexer_token.type     = T_WIDE_CHARACTER_CONSTANT;
 748         lexer_token.literal  = identify_string(string, size);
 749
 750         if (size == 0) {
 751                 errorf(&lexer_token.source_position, "empty character constant");
 752         }
 753 }
 754
 755 /**
 756  * Parse a wide string literal and set lexer_token.
 757  */
 758 static void parse_wide_string_literal(void)
 759 {
 760         parse_string_literal();
 761         if (lexer_token.type == T_STRING_LITERAL)
 762                 lexer_token.type = T_WIDE_STRING_LITERAL;
 763 }
 764
 765 /**
 766  * Parse a character constant and set lexer_token.
 767  */
 768 static void parse_character_constant(void)
 769 {
 770         eat('\'');
 771
 772         while (true) {
 773                 switch (c) {
 774                 case '\\': {
 775                         utf32 const tc = parse_escape_sequence();
 776                         if (tc >= 0x100) {
 777                                 warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
 778                         }
 779                         obstack_1grow(&symbol_obstack, tc);
 780                         break;
 781                 }
 782
 783                 MATCH_NEWLINE(
 784                         parse_error("newline while parsing character constant");
 785                         break;
 786                 )
 787
 788                 case '\'':
 789                         next_char();
 790                         goto end_of_char_constant;
 791
 792                 case EOF: {
 793                         errorf(&lexer_token.source_position, "EOF while parsing character constant");
 794                         lexer_token.type = T_ERROR;
 795                         return;
 796                 }
 797
 798                 default:
 799                         grow_symbol(c);
 800                         next_char();
 801                         break;
 802
 803                 }
 804         }
 805
 806 end_of_char_constant:;
 807         obstack_1grow(&symbol_obstack, '\0');
 808         const size_t        size   = (size_t)obstack_object_size(&symbol_obstack)-1;
 809         char         *const string = obstack_finish(&symbol_obstack);
 810
 811         lexer_token.type    = T_CHARACTER_CONSTANT;
 812         lexer_token.literal = identify_string(string, size);
 813
 814         if (size == 0) {
 815                 errorf(&lexer_token.source_position, "empty character constant");
 816         }
 817 }
 818
 819 /**
 820  * Skip a multiline comment.
 821  */
 822 static void skip_multiline_comment(void)
 823 {
 824         while (true) {
 825                 switch (c) {
 826                 case '/':
 827                         next_char();
 828                         if (c == '*') {
 829                                 /* nested comment, warn here */
 830                                 warningf(WARN_COMMENT, &lexer_pos, "'/*' within comment");
 831                         }
 832                         break;
 833                 case '*':
 834                         next_char();
 835                         if (c == '/') {
 836                                 next_char();
 837                                 return;
 838                         }
 839                         break;
 840
 841                 MATCH_NEWLINE(break;)
 842
 843                 case EOF: {
 844                         errorf(&lexer_token.source_position, "at end of file while looking for comment end");
 845                         return;
 846                 }
 847
 848                 default:
 849                         next_char();
 850                         break;
 851                 }
 852         }
 853 }
 854
 855 /**
 856  * Skip a single line comment.
 857  */
 858 static void skip_line_comment(void)
 859 {
 860         while (true) {
 861                 switch (c) {
 862                 case EOF:
 863                         return;
 864
 865                 case '\n':
 866                 case '\r':
 867                         return;
 868
 869                 case '\\':
 870                         next_char();
 871                         if (c == '\n' || c == '\r') {
 872                                 warningf(WARN_COMMENT, &lexer_pos, "multi-line comment");
 873                                 return;
 874                         }
 875                         break;
 876
 877                 default:
 878                         next_char();
 879                         break;
 880                 }
 881         }
 882 }
 883
 884 /** The current preprocessor token. */
 885 static token_t pp_token;
 886
 887 /**
 888  * Read the next preprocessor token.
 889  */
 890 static inline void next_pp_token(void)
 891 {
 892         lexer_next_preprocessing_token();
 893         pp_token = lexer_token;
 894 }
 895
 896 /**
 897  * Eat all preprocessor tokens until newline.
 898  */
 899 static void eat_until_newline(void)
 900 {
 901         while (pp_token.type != '\n' && pp_token.type != T_EOF) {
 902                 next_pp_token();
 903         }
 904 }
 905
 906 /**
 907  * Handle the define directive.
 908  */
 909 static void define_directive(void)
 910 {
 911         lexer_next_preprocessing_token();
 912         if (lexer_token.type != T_IDENTIFIER) {
 913                 parse_error("expected identifier after #define\n");
 914                 eat_until_newline();
 915         }
 916 }
 917
 918 /**
 919  * Handle the ifdef directive.
 920  */
 921 static void ifdef_directive(int is_ifndef)
 922 {
 923         (void) is_ifndef;
 924         lexer_next_preprocessing_token();
 925         //expect_identifier();
 926         //extect_newline();
 927 }
 928
 929 /**
 930  * Handle the endif directive.
 931  */
 932 static void endif_directive(void)
 933 {
 934         //expect_newline();
 935 }
 936
 937 /**
 938  * Parse the line directive.
 939  */
 940 static void parse_line_directive(void)
 941 {
 942         if (pp_token.type != T_INTEGER) {
 943                 parse_error("expected integer");
 944         } else {
 945                 /* use offset -1 as this is about the next line */
 946                 lexer_pos.lineno = atoi(pp_token.literal.begin) - 1;
 947                 next_pp_token();
 948         }
 949         if (pp_token.type == T_STRING_LITERAL) {
 950                 lexer_pos.input_name = pp_token.literal.begin;
 951                 next_pp_token();
 952         }
 953
 954         eat_until_newline();
 955 }
 956
 957 /**
 958  * STDC pragmas.
 959  */
 960 typedef enum stdc_pragma_kind_t {
 961         STDC_UNKNOWN,
 962         STDC_FP_CONTRACT,
 963         STDC_FENV_ACCESS,
 964         STDC_CX_LIMITED_RANGE
 965 } stdc_pragma_kind_t;
 966
 967 /**
 968  * STDC pragma values.
 969  */
 970 typedef enum stdc_pragma_value_kind_t {
 971         STDC_VALUE_UNKNOWN,
 972         STDC_VALUE_ON,
 973         STDC_VALUE_OFF,
 974         STDC_VALUE_DEFAULT
 975 } stdc_pragma_value_kind_t;
 976
 977 /**
 978  * Parse a pragma directive.
 979  */
 980 static void parse_pragma(void)
 981 {
 982         bool unknown_pragma = true;
 983
 984         next_pp_token();
 985         if (pp_token.symbol->pp_ID == TP_STDC) {
 986                 stdc_pragma_kind_t kind = STDC_UNKNOWN;
 987                 /* a STDC pragma */
 988                 if (c_mode & _C99) {
 989                         next_pp_token();
 990
 991                         switch (pp_token.symbol->pp_ID) {
 992                         case TP_FP_CONTRACT:
 993                                 kind = STDC_FP_CONTRACT;
 994                                 break;
 995                         case TP_FENV_ACCESS:
 996                                 kind = STDC_FENV_ACCESS;
 997                                 break;
 998                         case TP_CX_LIMITED_RANGE:
 999                                 kind = STDC_CX_LIMITED_RANGE;
1000                                 break;
1001                         default:
1002                                 break;
1003                         }
1004                         if (kind != STDC_UNKNOWN) {
1005                                 stdc_pragma_value_kind_t value = STDC_VALUE_UNKNOWN;
1006                                 next_pp_token();
1007                                 switch (pp_token.symbol->pp_ID) {
1008                                 case TP_ON:
1009                                         value = STDC_VALUE_ON;
1010                                         break;
1011                                 case TP_OFF:
1012                                         value = STDC_VALUE_OFF;
1013                                         break;
1014                                 case TP_DEFAULT:
1015                                         value = STDC_VALUE_DEFAULT;
1016                                         break;
1017                                 default:
1018                                         break;
1019                                 }
1020                                 if (value != STDC_VALUE_UNKNOWN) {
1021                                         unknown_pragma = false;
1022                                 } else {
1023                                         errorf(&pp_token.source_position, "bad STDC pragma argument");
1024                                 }
1025                         }
1026                 }
1027         } else {
1028                 unknown_pragma = true;
1029         }
1030         eat_until_newline();
1031         if (unknown_pragma) {
1032                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.source_position, "encountered unknown #pragma");
1033         }
1034 }
1035
1036 /**
1037  * Parse a preprocessor non-null directive.
1038  */
1039 static void parse_preprocessor_identifier(void)
1040 {
1041         assert(pp_token.type == T_IDENTIFIER);
1042         symbol_t *symbol = pp_token.symbol;
1043
1044         switch (symbol->pp_ID) {
1045         case TP_include:
1046                 printf("include - enable header name parsing!\n");
1047                 break;
1048         case TP_define:
1049                 define_directive();
1050                 break;
1051         case TP_ifdef:
1052                 ifdef_directive(0);
1053                 break;
1054         case TP_ifndef:
1055                 ifdef_directive(1);
1056                 break;
1057         case TP_endif:
1058                 endif_directive();
1059                 break;
1060         case TP_line:
1061                 next_pp_token();
1062                 parse_line_directive();
1063                 break;
1064         case TP_if:
1065         case TP_else:
1066         case TP_elif:
1067         case TP_undef:
1068         case TP_error:
1069                 /* TODO; output the rest of the line */
1070                 parse_error("#error directive: ");
1071                 break;
1072         case TP_pragma:
1073                 parse_pragma();
1074                 break;
1075         }
1076 }
1077
1078 /**
1079  * Parse a preprocessor directive.
1080  */
1081 static void parse_preprocessor_directive(void)
1082 {
1083         next_pp_token();
1084
1085         switch (pp_token.type) {
1086         case T_IDENTIFIER:
1087                 parse_preprocessor_identifier();
1088                 break;
1089         case T_INTEGER:
1090                 parse_line_directive();
1091                 break;
1092         case '\n':
1093                 /* NULL directive, see §6.10.7 */
1094                 break;
1095         default:
1096                 parse_error("invalid preprocessor directive");
1097                 eat_until_newline();
1098                 break;
1099         }
1100 }
1101
1102 #define MAYBE_PROLOG                                       \
1103                         next_char();                                   \
1104                         while (true) {                                 \
1105                                 switch (c) {
1106
1107 #define MAYBE(ch, set_type)                                \
1108                                 case ch:                                   \
1109                                         next_char();                           \
1110                                         lexer_token.type = set_type;           \
1111                                         return;
1112
1113 /* must use this as last thing */
1114 #define MAYBE_MODE(ch, set_type, mode)                     \
1115                                 case ch:                                   \
1116                                         if (c_mode & mode) {                   \
1117                                                 next_char();                       \
1118                                                 lexer_token.type = set_type;       \
1119                                                 return;                            \
1120                                         }                                      \
1121                                         /* fallthrough */
1122
1123 #define ELSE_CODE(code)                                    \
1124                                 default:                                   \
1125                                         code                                   \
1126                                         return;                                \
1127                                 }                                          \
1128                         } /* end of while (true) */                    \
1129
1130 #define ELSE(set_type)                                     \
1131                 ELSE_CODE(                                         \
1132                         lexer_token.type = set_type;                   \
1133                 )
1134
1135 void lexer_next_preprocessing_token(void)
1136 {
1137         while (true) {
1138                 lexer_token.source_position = lexer_pos;
1139
1140                 switch (c) {
1141                 case ' ':
1142                 case '\t':
1143                         next_char();
1144                         break;
1145
1146                 MATCH_NEWLINE(
1147                         lexer_token.type = '\n';
1148                         return;
1149                 )
1150
1151                 SYMBOL_CHARS
1152                         parse_symbol();
1153                         /* might be a wide string ( L"string" ) */
1154                         if (lexer_token.symbol == symbol_L) {
1155                                 switch (c) {
1156                                         case '"':  parse_wide_string_literal();     break;
1157                                         case '\'': parse_wide_character_constant(); break;
1158                                 }
1159                         }
1160                         return;
1161
1162                 DIGITS
1163                         parse_number();
1164                         return;
1165
1166                 case '"':
1167                         parse_string_literal();
1168                         return;
1169
1170                 case '\'':
1171                         parse_character_constant();
1172                         return;
1173
1174                 case '.':
1175                         MAYBE_PROLOG
1176                                 DIGITS
1177                                         put_back(c);
1178                                         c = '.';
1179                                         parse_number();
1180                                         return;
1181
1182                                 case '.':
1183                                         MAYBE_PROLOG
1184                                         MAYBE('.', T_DOTDOTDOT)
1185                                         ELSE_CODE(
1186                                                 put_back(c);
1187                                                 c = '.';
1188                                                 lexer_token.type = '.';
1189                                         )
1190                         ELSE('.')
1191                 case '&':
1192                         MAYBE_PROLOG
1193                         MAYBE('&', T_ANDAND)
1194                         MAYBE('=', T_ANDEQUAL)
1195                         ELSE('&')
1196                 case '*':
1197                         MAYBE_PROLOG
1198                         MAYBE('=', T_ASTERISKEQUAL)
1199                         ELSE('*')
1200                 case '+':
1201                         MAYBE_PROLOG
1202                         MAYBE('+', T_PLUSPLUS)
1203                         MAYBE('=', T_PLUSEQUAL)
1204                         ELSE('+')
1205                 case '-':
1206                         MAYBE_PROLOG
1207                         MAYBE('>', T_MINUSGREATER)
1208                         MAYBE('-', T_MINUSMINUS)
1209                         MAYBE('=', T_MINUSEQUAL)
1210                         ELSE('-')
1211                 case '!':
1212                         MAYBE_PROLOG
1213                         MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1214                         ELSE('!')
1215                 case '/':
1216                         MAYBE_PROLOG
1217                         MAYBE('=', T_SLASHEQUAL)
1218                                 case '*':
1219                                         next_char();
1220                                         skip_multiline_comment();
1221                                         lexer_next_preprocessing_token();
1222                                         return;
1223                                 case '/':
1224                                         next_char();
1225                                         skip_line_comment();
1226                                         lexer_next_preprocessing_token();
1227                                         return;
1228                         ELSE('/')
1229                 case '%':
1230                         MAYBE_PROLOG
1231                         MAYBE('>', '}')
1232                         MAYBE('=', T_PERCENTEQUAL)
1233                                 case ':':
1234                                         MAYBE_PROLOG
1235                                                 case '%':
1236                                                         MAYBE_PROLOG
1237                                                         MAYBE(':', T_HASHHASH)
1238                                                         ELSE_CODE(
1239                                                                 put_back(c);
1240                                                                 c = '%';
1241                                                                 lexer_token.type = '#';
1242                                                         )
1243                                         ELSE('#')
1244                         ELSE('%')
1245                 case '<':
1246                         MAYBE_PROLOG
1247                         MAYBE(':', '[')
1248                         MAYBE('%', '{')
1249                         MAYBE('=', T_LESSEQUAL)
1250                                 case '<':
1251                                         MAYBE_PROLOG
1252                                         MAYBE('=', T_LESSLESSEQUAL)
1253                                         ELSE(T_LESSLESS)
1254                         ELSE('<')
1255                 case '>':
1256                         MAYBE_PROLOG
1257                         MAYBE('=', T_GREATEREQUAL)
1258                                 case '>':
1259                                         MAYBE_PROLOG
1260                                         MAYBE('=', T_GREATERGREATEREQUAL)
1261                                         ELSE(T_GREATERGREATER)
1262                         ELSE('>')
1263                 case '^':
1264                         MAYBE_PROLOG
1265                         MAYBE('=', T_CARETEQUAL)
1266                         ELSE('^')
1267                 case '|':
1268                         MAYBE_PROLOG
1269                         MAYBE('=', T_PIPEEQUAL)
1270                         MAYBE('|', T_PIPEPIPE)
1271                         ELSE('|')
1272                 case ':':
1273                         MAYBE_PROLOG
1274                         MAYBE('>', ']')
1275                         MAYBE_MODE(':', T_COLONCOLON, _CXX)
1276                         ELSE(':')
1277                 case '=':
1278                         MAYBE_PROLOG
1279                         MAYBE('=', T_EQUALEQUAL)
1280                         ELSE('=')
1281                 case '#':
1282                         MAYBE_PROLOG
1283                         MAYBE('#', T_HASHHASH)
1284                         ELSE('#')
1285
1286                 case '?':
1287                 case '[':
1288                 case ']':
1289                 case '(':
1290                 case ')':
1291                 case '{':
1292                 case '}':
1293                 case '~':
1294                 case ';':
1295                 case ',':
1296                 case '\\':
1297                         lexer_token.type = c;
1298                         next_char();
1299                         return;
1300
1301                 case EOF:
1302                         lexer_token.type = T_EOF;
1303                         return;
1304
1305                 default:
1306 dollar_sign:
1307                         errorf(&lexer_pos, "unknown character '%c' found", c);
1308                         next_char();
1309                         lexer_token.type = T_ERROR;
1310                         return;
1311                 }
1312         }
1313 }
1314
1315 void lexer_next_token(void)
1316 {
1317         lexer_next_preprocessing_token();
1318
1319         while (lexer_token.type == '\n') {
1320 newline_found:
1321                 lexer_next_preprocessing_token();
1322         }
1323
1324         if (lexer_token.type == '#') {
1325                 parse_preprocessor_directive();
1326                 goto newline_found;
1327         }
1328 }
1329
1330 void init_lexer(void)
1331 {
1332         strset_init(&stringset);
1333         symbol_L = symbol_table_insert("L");
1334 }
1335
1336 static void input_error(unsigned delta_lines, unsigned delta_cols,
1337                         const char *message)
1338 {
1339         lexer_pos.lineno += delta_lines;
1340         lexer_pos.colno  += delta_cols;
1341         errorf(&lexer_pos, "%s", message);
1342 }
1343
1344 void select_input_encoding(char const* new_encoding)
1345 {
1346         if (encoding != NULL)
1347                 xfree(encoding);
1348         encoding = xstrdup(new_encoding);
1349 }
1350
1351 void lexer_open_stream(FILE *stream, const char *input_name)
1352 {
1353         if (input != NULL) {
1354                 input_free(input);
1355                 input = NULL;
1356         }
1357
1358         lexer_pos.lineno     = 0;
1359         lexer_pos.colno      = 0;
1360         lexer_pos.input_name = input_name;
1361
1362         set_input_error_callback(input_error);
1363         input  = input_from_stream(stream, encoding);
1364         bufpos = NULL;
1365         bufend = NULL;
1366
1367         /* place a virtual \n at the beginning so the lexer knows that we're
1368          * at the beginning of a line */
1369         c = '\n';
1370 }
1371
1372 void exit_lexer(void)
1373 {
1374         if (input != NULL) {
1375                 input_free(input);
1376                 input = NULL;
1377         }
1378         strset_destroy(&stringset);
1379 }
1380
1381 static __attribute__((unused))
1382 void dbg_pos(const source_position_t source_position)
1383 {
1384         fprintf(stdout, "%s:%u:%u\n", source_position.input_name,
1385                 source_position.lineno, source_position.colno);
1386         fflush(stdout);
1387 }