nsz Git - cparser/blob - preprocessor.c

   1 #include <config.h>
   2
   3 #include <assert.h>
   4 #include <errno.h>
   5 #include <string.h>
   6 #include <stdbool.h>
   7 #include <ctype.h>
   8
   9 #include "token_t.h"
  10 #include "symbol_t.h"
  11 #include "adt/util.h"
  12 #include "adt/error.h"
  13 #include "adt/strutil.h"
  14 #include "adt/strset.h"
  15 #include "lang_features.h"
  16 #include "diagnostic.h"
  17 #include "string_rep.h"
  18 #include "input.h"
  19
  20 #define MAX_PUTBACK 3
  21 #define INCLUDE_LIMIT 199  /* 199 is for gcc "compatibility" */
  22
  23 struct pp_argument_t {
  24         size_t   list_len;
  25         token_t *token_list;
  26 };
  27
  28 struct pp_definition_t {
  29         symbol_t          *symbol;
  30         source_position_t  source_position;
  31         pp_definition_t   *parent_expansion;
  32         size_t             expand_pos;
  33         bool               is_variadic    : 1;
  34         bool               is_expanding   : 1;
  35         bool               has_parameters : 1;
  36         size_t             n_parameters;
  37         symbol_t          *parameters;
  38
  39         /* replacement */
  40         size_t             list_len;
  41         token_t           *token_list;
  42
  43 };
  44
  45 typedef struct pp_conditional_t pp_conditional_t;
  46 struct pp_conditional_t {
  47         source_position_t  source_position;
  48         bool               condition;
  49         bool               in_else;
  50         bool               skip; /**< conditional in skip mode (then+else gets skipped) */
  51         pp_conditional_t  *parent;
  52 };
  53
  54 typedef struct pp_input_t pp_input_t;
  55 struct pp_input_t {
  56         FILE              *file;
  57         input_t           *input;
  58         utf32              c;
  59         utf32              buf[1024+MAX_PUTBACK];
  60         const utf32       *bufend;
  61         const utf32       *bufpos;
  62         source_position_t  position;
  63         pp_input_t        *parent;
  64         unsigned           output_line;
  65 };
  66
  67 /** additional info about the current token */
  68 typedef struct add_token_info_t {
  69         /** whitespace from beginning of line to the token */
  70         unsigned whitespace;
  71         /** there has been any whitespace before the token */
  72         bool     had_whitespace;
  73         /** the token is at the beginning of the line */
  74         bool     at_line_begin;
  75 } add_token_info_t;
  76
  77 typedef struct searchpath_entry_t searchpath_entry_t;
  78 struct searchpath_entry_t {
  79         const char         *path;
  80         searchpath_entry_t *next;
  81 };
  82
  83 static pp_input_t      input;
  84
  85 static pp_input_t     *input_stack;
  86 static unsigned        n_inputs;
  87 static struct obstack  input_obstack;
  88
  89 static pp_conditional_t *conditional_stack;
  90
  91 static token_t           pp_token;
  92 static bool              resolve_escape_sequences = false;
  93 static bool              ignore_unknown_chars     = true;
  94 static bool              in_pp_directive;
  95 static bool              skip_mode;
  96 static FILE             *out;
  97 static struct obstack    pp_obstack;
  98 static struct obstack    config_obstack;
  99 static const char       *printed_input_name = NULL;
 100 static source_position_t expansion_pos;
 101 static pp_definition_t  *current_expansion  = NULL;
 102 static strset_t          stringset;
 103 static preprocessor_token_kind_t last_token = TP_ERROR;
 104
 105 static searchpath_entry_t *searchpath;
 106
 107 static add_token_info_t  info;
 108
 109 static inline void next_char(void);
 110 static void next_preprocessing_token(void);
 111 static void print_line_directive(const source_position_t *pos, const char *add);
 112
 113 static void switch_input(FILE *file, const char *filename)
 114 {
 115         input.file                = file;
 116         input.input               = input_from_stream(file, NULL);
 117         input.bufend              = NULL;
 118         input.bufpos              = NULL;
 119         input.output_line         = 0;
 120         input.position.input_name = filename;
 121         input.position.lineno     = 1;
 122
 123         /* indicate that we're at a new input */
 124         print_line_directive(&input.position, input_stack != NULL ? "1" : NULL);
 125
 126         /* place a virtual '\n' so we realize we're at line begin */
 127         input.position.lineno = 0;
 128         input.c               = '\n';
 129         next_preprocessing_token();
 130 }
 131
 132 static void close_input(void)
 133 {
 134         input_free(input.input);
 135         assert(input.file != NULL);
 136
 137         fclose(input.file);
 138         input.input  = NULL;
 139         input.file   = NULL;
 140         input.bufend = NULL;
 141         input.bufpos = NULL;
 142         input.c      = EOF;
 143 }
 144
 145 static void push_input(void)
 146 {
 147         pp_input_t *saved_input
 148                 = obstack_alloc(&input_obstack, sizeof(*saved_input));
 149
 150         memcpy(saved_input, &input, sizeof(*saved_input));
 151
 152         /* adjust buffer positions */
 153         if (input.bufpos != NULL)
 154                 saved_input->bufpos = saved_input->buf + (input.bufpos - input.buf);
 155         if (input.bufend != NULL)
 156                 saved_input->bufend = saved_input->buf + (input.bufend - input.buf);
 157
 158         saved_input->parent = input_stack;
 159         input_stack         = saved_input;
 160         ++n_inputs;
 161 }
 162
 163 static void pop_restore_input(void)
 164 {
 165         assert(n_inputs > 0);
 166         assert(input_stack != NULL);
 167
 168         pp_input_t *saved_input = input_stack;
 169
 170         memcpy(&input, saved_input, sizeof(input));
 171         input.parent = NULL;
 172
 173         /* adjust buffer positions */
 174         if (saved_input->bufpos != NULL)
 175                 input.bufpos = input.buf + (saved_input->bufpos - saved_input->buf);
 176         if (saved_input->bufend != NULL)
 177                 input.bufend = input.buf + (saved_input->bufend - saved_input->buf);
 178
 179         input_stack = saved_input->parent;
 180         obstack_free(&input_obstack, saved_input);
 181         --n_inputs;
 182 }
 183
 184 /**
 185  * Prints a parse error message at the current token.
 186  *
 187  * @param msg   the error message
 188  */
 189 static void parse_error(const char *msg)
 190 {
 191         errorf(&pp_token.base.source_position,  "%s", msg);
 192 }
 193
 194 static inline void next_real_char(void)
 195 {
 196         assert(input.bufpos <= input.bufend);
 197         if (input.bufpos >= input.bufend) {
 198                 size_t n = decode(input.input, input.buf + MAX_PUTBACK,
 199                                   sizeof(input.buf)/sizeof(input.buf[0]) - MAX_PUTBACK);
 200                 if (n == 0) {
 201                         input.c = EOF;
 202                         return;
 203                 }
 204                 input.bufpos = input.buf + MAX_PUTBACK;
 205                 input.bufend = input.bufpos + n;
 206         }
 207         input.c = *input.bufpos++;
 208         ++input.position.colno;
 209 }
 210
 211 /**
 212  * Put a character back into the buffer.
 213  *
 214  * @param pc  the character to put back
 215  */
 216 static inline void put_back(utf32 const pc)
 217 {
 218         assert(input.bufpos > input.buf);
 219         *(--input.bufpos - input.buf + input.buf) = (char) pc;
 220         --input.position.colno;
 221 }
 222
 223 #define MATCH_NEWLINE(code)                   \
 224         case '\r':                                \
 225                 next_char();                          \
 226                 if (input.c == '\n') {                \
 227         case '\n':                                \
 228                         next_char();                      \
 229                 }                                     \
 230                 info.whitespace = 0;                  \
 231                 ++input.position.lineno;              \
 232                 input.position.colno = 1;             \
 233                 code
 234
 235 #define eat(c_type) (assert(input.c == c_type), next_char())
 236
 237 static void maybe_concat_lines(void)
 238 {
 239         eat('\\');
 240
 241         switch (input.c) {
 242         MATCH_NEWLINE(
 243                 return;
 244         )
 245
 246         default:
 247                 break;
 248         }
 249
 250         put_back(input.c);
 251         input.c = '\\';
 252 }
 253
 254 /**
 255  * Set c to the next input character, ie.
 256  * after expanding trigraphs.
 257  */
 258 static inline void next_char(void)
 259 {
 260         next_real_char();
 261
 262         /* filter trigraphs and concatenated lines */
 263         if (UNLIKELY(input.c == '\\')) {
 264                 maybe_concat_lines();
 265                 goto end_of_next_char;
 266         }
 267
 268         if (LIKELY(input.c != '?'))
 269                 goto end_of_next_char;
 270
 271         next_real_char();
 272         if (LIKELY(input.c != '?')) {
 273                 put_back(input.c);
 274                 input.c = '?';
 275                 goto end_of_next_char;
 276         }
 277
 278         next_real_char();
 279         switch (input.c) {
 280         case '=': input.c = '#'; break;
 281         case '(': input.c = '['; break;
 282         case '/': input.c = '\\'; maybe_concat_lines(); break;
 283         case ')': input.c = ']'; break;
 284         case '\'': input.c = '^'; break;
 285         case '<': input.c = '{'; break;
 286         case '!': input.c = '|'; break;
 287         case '>': input.c = '}'; break;
 288         case '-': input.c = '~'; break;
 289         default:
 290                 put_back(input.c);
 291                 put_back('?');
 292                 input.c = '?';
 293                 break;
 294         }
 295
 296 end_of_next_char:;
 297 #ifdef DEBUG_CHARS
 298         printf("nchar '%c'\n", input.c);
 299 #endif
 300 }
 301
 302
 303
 304 /**
 305  * Returns true if the given char is a octal digit.
 306  *
 307  * @param char  the character to check
 308  */
 309 static inline bool is_octal_digit(int chr)
 310 {
 311         switch (chr) {
 312         case '0':
 313         case '1':
 314         case '2':
 315         case '3':
 316         case '4':
 317         case '5':
 318         case '6':
 319         case '7':
 320                 return true;
 321         default:
 322                 return false;
 323         }
 324 }
 325
 326 /**
 327  * Returns the value of a digit.
 328  * The only portable way to do it ...
 329  */
 330 static int digit_value(int digit)
 331 {
 332         switch (digit) {
 333         case '0': return 0;
 334         case '1': return 1;
 335         case '2': return 2;
 336         case '3': return 3;
 337         case '4': return 4;
 338         case '5': return 5;
 339         case '6': return 6;
 340         case '7': return 7;
 341         case '8': return 8;
 342         case '9': return 9;
 343         case 'a':
 344         case 'A': return 10;
 345         case 'b':
 346         case 'B': return 11;
 347         case 'c':
 348         case 'C': return 12;
 349         case 'd':
 350         case 'D': return 13;
 351         case 'e':
 352         case 'E': return 14;
 353         case 'f':
 354         case 'F': return 15;
 355         default:
 356                 panic("wrong character given");
 357         }
 358 }
 359
 360 /**
 361  * Parses an octal character sequence.
 362  *
 363  * @param first_digit  the already read first digit
 364  */
 365 static int parse_octal_sequence(const int first_digit)
 366 {
 367         assert(is_octal_digit(first_digit));
 368         int value = digit_value(first_digit);
 369         if (!is_octal_digit(input.c)) return value;
 370         value = 8 * value + digit_value(input.c);
 371         next_char();
 372         if (!is_octal_digit(input.c)) return value;
 373         value = 8 * value + digit_value(input.c);
 374         next_char();
 375
 376         if (char_is_signed) {
 377                 return (signed char) value;
 378         } else {
 379                 return (unsigned char) value;
 380         }
 381 }
 382
 383 /**
 384  * Parses a hex character sequence.
 385  */
 386 static int parse_hex_sequence(void)
 387 {
 388         int value = 0;
 389         while (isxdigit(input.c)) {
 390                 value = 16 * value + digit_value(input.c);
 391                 next_char();
 392         }
 393
 394         if (char_is_signed) {
 395                 return (signed char) value;
 396         } else {
 397                 return (unsigned char) value;
 398         }
 399 }
 400
 401 /**
 402  * Parse an escape sequence.
 403  */
 404 static int parse_escape_sequence(void)
 405 {
 406         eat('\\');
 407
 408         int ec = input.c;
 409         next_char();
 410
 411         switch (ec) {
 412         case '"':  return '"';
 413         case '\'': return '\'';
 414         case '\\': return '\\';
 415         case '?': return '\?';
 416         case 'a': return '\a';
 417         case 'b': return '\b';
 418         case 'f': return '\f';
 419         case 'n': return '\n';
 420         case 'r': return '\r';
 421         case 't': return '\t';
 422         case 'v': return '\v';
 423         case 'x':
 424                 return parse_hex_sequence();
 425         case '0':
 426         case '1':
 427         case '2':
 428         case '3':
 429         case '4':
 430         case '5':
 431         case '6':
 432         case '7':
 433                 return parse_octal_sequence(ec);
 434         case EOF:
 435                 parse_error("reached end of file while parsing escape sequence");
 436                 return EOF;
 437         default:
 438                 parse_error("unknown escape sequence");
 439                 return EOF;
 440         }
 441 }
 442
 443 static void grow_symbol(utf32 const tc)
 444 {
 445         struct obstack *const o  = &symbol_obstack;
 446         if (tc < 0x80U) {
 447                 obstack_1grow(o, tc);
 448         } else if (tc < 0x800) {
 449                 obstack_1grow(o, 0xC0 | (tc >> 6));
 450                 obstack_1grow(o, 0x80 | (tc & 0x3F));
 451         } else if (tc < 0x10000) {
 452                 obstack_1grow(o, 0xE0 | ( tc >> 12));
 453                 obstack_1grow(o, 0x80 | ((tc >>  6) & 0x3F));
 454                 obstack_1grow(o, 0x80 | ( tc        & 0x3F));
 455         } else {
 456                 obstack_1grow(o, 0xF0 | ( tc >> 18));
 457                 obstack_1grow(o, 0x80 | ((tc >> 12) & 0x3F));
 458                 obstack_1grow(o, 0x80 | ((tc >>  6) & 0x3F));
 459                 obstack_1grow(o, 0x80 | ( tc        & 0x3F));
 460         }
 461 }
 462
 463 static const char *identify_string(char *string)
 464 {
 465         const char *result = strset_insert(&stringset, string);
 466         if (result != string) {
 467                 obstack_free(&symbol_obstack, string);
 468         }
 469         return result;
 470 }
 471
 472 static string_t make_string(char *string, size_t len)
 473 {
 474         const char *result = identify_string(string);
 475         return (string_t) {result, len};
 476 }
 477
 478 static void parse_string_literal(void)
 479 {
 480         const unsigned start_linenr = input.position.lineno;
 481
 482         eat('"');
 483
 484         while (true) {
 485                 switch (input.c) {
 486                 case '\\': {
 487                         utf32 tc;
 488                         if (resolve_escape_sequences) {
 489                                 tc = parse_escape_sequence();
 490                                 obstack_1grow(&symbol_obstack, (char) tc);
 491                         } else {
 492                                 obstack_1grow(&symbol_obstack, (char) input.c);
 493                                 next_char();
 494                                 obstack_1grow(&symbol_obstack, (char) input.c);
 495                                 next_char();
 496                         }
 497                         break;
 498                 }
 499
 500                 case EOF: {
 501                         source_position_t source_position;
 502                         source_position.input_name = pp_token.base.source_position.input_name;
 503                         source_position.lineno     = start_linenr;
 504                         errorf(&source_position, "string has no end");
 505                         pp_token.kind = TP_ERROR;
 506                         return;
 507                 }
 508
 509                 case '"':
 510                         next_char();
 511                         goto end_of_string;
 512
 513                 default:
 514                         grow_symbol(input.c);
 515                         next_char();
 516                         break;
 517                 }
 518         }
 519
 520 end_of_string:
 521         /* add finishing 0 to the string */
 522         obstack_1grow(&symbol_obstack, '\0');
 523         const size_t size   = (size_t)obstack_object_size(&symbol_obstack);
 524         char *const  string = obstack_finish(&symbol_obstack);
 525
 526         pp_token.kind          = TP_STRING_LITERAL;
 527         pp_token.string.string = make_string(string, size);
 528 }
 529
 530 /**
 531  * Parse a wide string literal and set lexer_token.
 532  */
 533 static void parse_wide_string_literal(void)
 534 {
 535         parse_string_literal();
 536         if (pp_token.kind == TP_STRING_LITERAL)
 537                 pp_token.kind = TP_WIDE_STRING_LITERAL;
 538 }
 539
 540 static void parse_wide_character_constant(void)
 541 {
 542         eat('\'');
 543
 544         while (true) {
 545                 switch (input.c) {
 546                 case '\\': {
 547                         const utf32 tc = parse_escape_sequence();
 548                         grow_symbol(tc);
 549                         break;
 550                 }
 551
 552                 MATCH_NEWLINE(
 553                         parse_error("newline while parsing character constant");
 554                         break;
 555                 )
 556
 557                 case '\'':
 558                         next_char();
 559                         goto end_of_wide_char_constant;
 560
 561                 case EOF:
 562                         parse_error("EOF while parsing character constant");
 563                         pp_token.kind = TP_ERROR;
 564                         return;
 565
 566                 default:
 567                         grow_symbol(input.c);
 568                         next_char();
 569                         break;
 570                 }
 571         }
 572
 573 end_of_wide_char_constant:
 574         obstack_1grow(&symbol_obstack, '\0');
 575         size_t  size = (size_t) obstack_object_size(&symbol_obstack)-1;
 576         char   *string = obstack_finish(&symbol_obstack);
 577         pp_token.kind          = TP_WIDE_CHARACTER_CONSTANT;
 578         pp_token.string.string = make_string(string, size);
 579
 580         if (size == 0) {
 581                 parse_error("empty character constant");
 582         }
 583 }
 584
 585 static void parse_character_constant(void)
 586 {
 587         const unsigned start_linenr = input.position.lineno;
 588
 589         eat('\'');
 590
 591         int tc;
 592         while (true) {
 593                 switch (input.c) {
 594                 case '\\':
 595                         tc = parse_escape_sequence();
 596                         obstack_1grow(&symbol_obstack, (char) tc);
 597                         break;
 598
 599                 MATCH_NEWLINE(
 600                         parse_error("newline while parsing character constant");
 601                         break;
 602                 )
 603
 604                 case EOF: {
 605                         source_position_t source_position;
 606                         source_position.input_name = pp_token.base.source_position.input_name;
 607                         source_position.lineno     = start_linenr;
 608                         errorf(&source_position, "EOF while parsing character constant");
 609                         pp_token.kind = TP_ERROR;
 610                         return;
 611                 }
 612
 613                 case '\'':
 614                         next_char();
 615                         goto end_of_char_constant;
 616
 617                 default:
 618                         obstack_1grow(&symbol_obstack, (char) input.c);
 619                         next_char();
 620                         break;
 621
 622                 }
 623         }
 624
 625 end_of_char_constant:;
 626         obstack_1grow(&symbol_obstack, '\0');
 627         const size_t size   = (size_t)obstack_object_size(&symbol_obstack);
 628         char *const  string = obstack_finish(&symbol_obstack);
 629
 630         pp_token.kind          = TP_CHARACTER_CONSTANT;
 631         pp_token.string.string = make_string(string, size);
 632
 633         if (size == 0) {
 634                 parse_error("empty character constant");
 635         }
 636 }
 637
 638 #define SYMBOL_CHARS_WITHOUT_E_P \
 639         case 'a': \
 640         case 'b': \
 641         case 'c': \
 642         case 'd': \
 643         case 'f': \
 644         case 'g': \
 645         case 'h': \
 646         case 'i': \
 647         case 'j': \
 648         case 'k': \
 649         case 'l': \
 650         case 'm': \
 651         case 'n': \
 652         case 'o': \
 653         case 'q': \
 654         case 'r': \
 655         case 's': \
 656         case 't': \
 657         case 'u': \
 658         case 'v': \
 659         case 'w': \
 660         case 'x': \
 661         case 'y': \
 662         case 'z': \
 663         case 'A': \
 664         case 'B': \
 665         case 'C': \
 666         case 'D': \
 667         case 'F': \
 668         case 'G': \
 669         case 'H': \
 670         case 'I': \
 671         case 'J': \
 672         case 'K': \
 673         case 'L': \
 674         case 'M': \
 675         case 'N': \
 676         case 'O': \
 677         case 'Q': \
 678         case 'R': \
 679         case 'S': \
 680         case 'T': \
 681         case 'U': \
 682         case 'V': \
 683         case 'W': \
 684         case 'X': \
 685         case 'Y': \
 686         case 'Z': \
 687         case '_':
 688
 689 #define SYMBOL_CHARS \
 690         SYMBOL_CHARS_WITHOUT_E_P \
 691         case 'e': \
 692         case 'p': \
 693         case 'E': \
 694         case 'P':
 695
 696 #define DIGITS \
 697         case '0':  \
 698         case '1':  \
 699         case '2':  \
 700         case '3':  \
 701         case '4':  \
 702         case '5':  \
 703         case '6':  \
 704         case '7':  \
 705         case '8':  \
 706         case '9':
 707
 708 /**
 709  * returns next final token from a preprocessor macro expansion
 710  */
 711 static void expand_next(void)
 712 {
 713         assert(current_expansion != NULL);
 714
 715         pp_definition_t *definition = current_expansion;
 716
 717 restart:
 718         if (definition->list_len == 0
 719                         || definition->expand_pos >= definition->list_len) {
 720                 /* we're finished with the current macro, move up 1 level in the
 721                  * expansion stack */
 722                 pp_definition_t *parent = definition->parent_expansion;
 723                 definition->parent_expansion = NULL;
 724                 definition->is_expanding     = false;
 725
 726                 /* it was the outermost expansion, parse normal pptoken */
 727                 if (parent == NULL) {
 728                         current_expansion = NULL;
 729                         next_preprocessing_token();
 730                         return;
 731                 }
 732                 definition        = parent;
 733                 current_expansion = definition;
 734                 goto restart;
 735         }
 736         pp_token = definition->token_list[definition->expand_pos];
 737         pp_token.base.source_position = expansion_pos;
 738         ++definition->expand_pos;
 739
 740         if (pp_token.kind != TP_IDENTIFIER)
 741                 return;
 742
 743         /* if it was an identifier then we might need to expand again */
 744         pp_definition_t *symbol_definition = pp_token.identifier.symbol->pp_definition;
 745         if (symbol_definition != NULL && !symbol_definition->is_expanding) {
 746                 symbol_definition->parent_expansion = definition;
 747                 symbol_definition->expand_pos       = 0;
 748                 symbol_definition->is_expanding     = true;
 749                 definition                          = symbol_definition;
 750                 current_expansion                   = definition;
 751                 goto restart;
 752         }
 753 }
 754
 755 static void skip_line_comment(void)
 756 {
 757         while (true) {
 758                 switch (input.c) {
 759                 case EOF:
 760                         return;
 761
 762                 case '\r':
 763                 case '\n':
 764                         return;
 765
 766                 default:
 767                         next_char();
 768                         break;
 769                 }
 770         }
 771 }
 772
 773 static void skip_multiline_comment(void)
 774 {
 775         unsigned start_linenr = input.position.lineno;
 776         while (true) {
 777                 switch (input.c) {
 778                 case '/':
 779                         next_char();
 780                         if (input.c == '*') {
 781                                 /* TODO: nested comment, warn here */
 782                         }
 783                         break;
 784                 case '*':
 785                         next_char();
 786                         if (input.c == '/') {
 787                                 next_char();
 788                                 info.whitespace += input.position.colno-1;
 789                                 return;
 790                         }
 791                         break;
 792
 793                 MATCH_NEWLINE(
 794                         info.at_line_begin |= !in_pp_directive;
 795                         break;
 796                 )
 797
 798                 case EOF: {
 799                         source_position_t source_position;
 800                         source_position.input_name = pp_token.base.source_position.input_name;
 801                         source_position.lineno     = start_linenr;
 802                         errorf(&source_position, "at end of file while looking for comment end");
 803                         return;
 804                 }
 805
 806                 default:
 807                         next_char();
 808                         break;
 809                 }
 810         }
 811 }
 812
 813 static void skip_whitespace(void)
 814 {
 815         while (true) {
 816                 switch (input.c) {
 817                 case ' ':
 818                 case '\t':
 819                         next_char();
 820                         continue;
 821
 822                 MATCH_NEWLINE(
 823                         info.at_line_begin = true;
 824                         return;
 825                 )
 826
 827                 case '/':
 828                         next_char();
 829                         if (input.c == '/') {
 830                                 next_char();
 831                                 skip_line_comment();
 832                                 continue;
 833                         } else if (input.c == '*') {
 834                                 next_char();
 835                                 skip_multiline_comment();
 836                                 continue;
 837                         } else {
 838                                 put_back(input.c);
 839                                 input.c = '/';
 840                         }
 841                         return;
 842                 default:
 843                         return;
 844                 }
 845         }
 846 }
 847
 848 static void eat_pp(int type)
 849 {
 850         (void) type;
 851         assert(pp_token.kind == type);
 852         next_preprocessing_token();
 853 }
 854
 855 static void parse_symbol(void)
 856 {
 857         obstack_1grow(&symbol_obstack, (char) input.c);
 858         next_char();
 859
 860         while (true) {
 861                 switch (input.c) {
 862                 DIGITS
 863                 SYMBOL_CHARS
 864                         obstack_1grow(&symbol_obstack, (char) input.c);
 865                         next_char();
 866                         break;
 867
 868                 default:
 869                         goto end_symbol;
 870                 }
 871         }
 872
 873 end_symbol:
 874         obstack_1grow(&symbol_obstack, '\0');
 875         char *string = obstack_finish(&symbol_obstack);
 876
 877         /* might be a wide string or character constant ( L"string"/L'c' ) */
 878         if (input.c == '"' && string[0] == 'L' && string[1] == '\0') {
 879                 obstack_free(&symbol_obstack, string);
 880                 parse_wide_string_literal();
 881                 return;
 882         } else if (input.c == '\'' && string[0] == 'L' && string[1] == '\0') {
 883                 obstack_free(&symbol_obstack, string);
 884                 parse_wide_character_constant();
 885                 return;
 886         }
 887
 888         symbol_t *symbol = symbol_table_insert(string);
 889
 890         pp_token.kind              = symbol->pp_ID;
 891         pp_token.identifier.symbol = symbol;
 892
 893         /* we can free the memory from symbol obstack if we already had an entry in
 894          * the symbol table */
 895         if (symbol->string != string) {
 896                 obstack_free(&symbol_obstack, string);
 897         }
 898 }
 899
 900 static void parse_number(void)
 901 {
 902         obstack_1grow(&symbol_obstack, (char) input.c);
 903         next_char();
 904
 905         while (true) {
 906                 switch (input.c) {
 907                 case '.':
 908                 DIGITS
 909                 SYMBOL_CHARS_WITHOUT_E_P
 910                         obstack_1grow(&symbol_obstack, (char) input.c);
 911                         next_char();
 912                         break;
 913
 914                 case 'e':
 915                 case 'p':
 916                 case 'E':
 917                 case 'P':
 918                         obstack_1grow(&symbol_obstack, (char) input.c);
 919                         next_char();
 920                         if (input.c == '+' || input.c == '-') {
 921                                 obstack_1grow(&symbol_obstack, (char) input.c);
 922                                 next_char();
 923                         }
 924                         break;
 925
 926                 default:
 927                         goto end_number;
 928                 }
 929         }
 930
 931 end_number:
 932         obstack_1grow(&symbol_obstack, '\0');
 933         size_t  size   = obstack_object_size(&symbol_obstack);
 934         char   *string = obstack_finish(&symbol_obstack);
 935
 936         pp_token.kind          = TP_NUMBER;
 937         pp_token.number.number = make_string(string, size);
 938 }
 939
 940
 941 #define MAYBE_PROLOG                                       \
 942                         next_char();                                   \
 943                         while (true) {                                 \
 944                                 switch (input.c) {
 945
 946 #define MAYBE(ch, set_type)                                \
 947                                 case ch:                                   \
 948                                         next_char();                           \
 949                                         pp_token.kind = set_type;              \
 950                                         return;
 951
 952 #define ELSE_CODE(code)                                    \
 953                                 default:                                   \
 954                                         code                                   \
 955                                         return;                                \
 956                                 }                                          \
 957                         }
 958
 959 #define ELSE(set_type)                                     \
 960                 ELSE_CODE(                                         \
 961                         pp_token.kind = set_type;                      \
 962                 )
 963
 964 static void next_preprocessing_token(void)
 965 {
 966         if (current_expansion != NULL) {
 967                 expand_next();
 968                 return;
 969         }
 970
 971         info.at_line_begin  = false;
 972         info.had_whitespace = false;
 973 restart:
 974         pp_token.base.source_position = input.position;
 975         switch (input.c) {
 976         case ' ':
 977         case '\t':
 978                 ++info.whitespace;
 979                 info.had_whitespace = true;
 980                 next_char();
 981                 goto restart;
 982
 983         MATCH_NEWLINE(
 984                 info.at_line_begin = true;
 985                 info.had_whitespace = true;
 986                 goto restart;
 987         )
 988
 989         SYMBOL_CHARS
 990                 parse_symbol();
 991                 return;
 992
 993         DIGITS
 994                 parse_number();
 995                 return;
 996
 997         case '"':
 998                 parse_string_literal();
 999                 return;
1000
1001         case '\'':
1002                 parse_character_constant();
1003                 return;
1004
1005         case '.':
1006                 MAYBE_PROLOG
1007                         case '0':
1008                         case '1':
1009                         case '2':
1010                         case '3':
1011                         case '4':
1012                         case '5':
1013                         case '6':
1014                         case '7':
1015                         case '8':
1016                         case '9':
1017                                 put_back(input.c);
1018                                 input.c = '.';
1019                                 parse_number();
1020                                 return;
1021
1022                         case '.':
1023                                 MAYBE_PROLOG
1024                                 MAYBE('.', TP_DOTDOTDOT)
1025                                 ELSE_CODE(
1026                                         put_back(input.c);
1027                                         input.c = '.';
1028                                         pp_token.kind = '.';
1029                                 )
1030                 ELSE('.')
1031         case '&':
1032                 MAYBE_PROLOG
1033                 MAYBE('&', TP_ANDAND)
1034                 MAYBE('=', TP_ANDEQUAL)
1035                 ELSE('&')
1036         case '*':
1037                 MAYBE_PROLOG
1038                 MAYBE('=', TP_ASTERISKEQUAL)
1039                 ELSE('*')
1040         case '+':
1041                 MAYBE_PROLOG
1042                 MAYBE('+', TP_PLUSPLUS)
1043                 MAYBE('=', TP_PLUSEQUAL)
1044                 ELSE('+')
1045         case '-':
1046                 MAYBE_PROLOG
1047                 MAYBE('>', TP_MINUSGREATER)
1048                 MAYBE('-', TP_MINUSMINUS)
1049                 MAYBE('=', TP_MINUSEQUAL)
1050                 ELSE('-')
1051         case '!':
1052                 MAYBE_PROLOG
1053                 MAYBE('=', TP_EXCLAMATIONMARKEQUAL)
1054                 ELSE('!')
1055         case '/':
1056                 MAYBE_PROLOG
1057                 MAYBE('=', TP_SLASHEQUAL)
1058                         case '*':
1059                                 next_char();
1060                                 info.had_whitespace = true;
1061                                 skip_multiline_comment();
1062                                 goto restart;
1063                         case '/':
1064                                 next_char();
1065                                 info.had_whitespace = true;
1066                                 skip_line_comment();
1067                                 goto restart;
1068                 ELSE('/')
1069         case '%':
1070                 MAYBE_PROLOG
1071                 MAYBE('>', '}')
1072                 MAYBE('=', TP_PERCENTEQUAL)
1073                         case ':':
1074                                 MAYBE_PROLOG
1075                                         case '%':
1076                                                 MAYBE_PROLOG
1077                                                 MAYBE(':', TP_HASHHASH)
1078                                                 ELSE_CODE(
1079                                                         put_back(input.c);
1080                                                         input.c = '%';
1081                                                         pp_token.kind = '#';
1082                                                 )
1083                                 ELSE('#')
1084                 ELSE('%')
1085         case '<':
1086                 MAYBE_PROLOG
1087                 MAYBE(':', '[')
1088                 MAYBE('%', '{')
1089                 MAYBE('=', TP_LESSEQUAL)
1090                         case '<':
1091                                 MAYBE_PROLOG
1092                                 MAYBE('=', TP_LESSLESSEQUAL)
1093                                 ELSE(TP_LESSLESS)
1094                 ELSE('<')
1095         case '>':
1096                 MAYBE_PROLOG
1097                 MAYBE('=', TP_GREATEREQUAL)
1098                         case '>':
1099                                 MAYBE_PROLOG
1100                                 MAYBE('=', TP_GREATERGREATEREQUAL)
1101                                 ELSE(TP_GREATERGREATER)
1102                 ELSE('>')
1103         case '^':
1104                 MAYBE_PROLOG
1105                 MAYBE('=', TP_CARETEQUAL)
1106                 ELSE('^')
1107         case '|':
1108                 MAYBE_PROLOG
1109                 MAYBE('=', TP_PIPEEQUAL)
1110                 MAYBE('|', TP_PIPEPIPE)
1111                 ELSE('|')
1112         case ':':
1113                 MAYBE_PROLOG
1114                 MAYBE('>', ']')
1115                 ELSE(':')
1116         case '=':
1117                 MAYBE_PROLOG
1118                 MAYBE('=', TP_EQUALEQUAL)
1119                 ELSE('=')
1120         case '#':
1121                 MAYBE_PROLOG
1122                 MAYBE('#', TP_HASHHASH)
1123                 ELSE_CODE(
1124                         pp_token.kind = '#';
1125                 )
1126
1127         case '?':
1128         case '[':
1129         case ']':
1130         case '(':
1131         case ')':
1132         case '{':
1133         case '}':
1134         case '~':
1135         case ';':
1136         case ',':
1137         case '\\':
1138                 pp_token.kind = input.c;
1139                 next_char();
1140                 return;
1141
1142         case EOF:
1143                 if (input_stack != NULL) {
1144                         close_input();
1145                         pop_restore_input();
1146                         fputc('\n', out);
1147                         print_line_directive(&input.position, "2");
1148                         goto restart;
1149                 } else {
1150                         pp_token.base.source_position.lineno++;
1151                         info.at_line_begin = true;
1152                         pp_token.kind = TP_EOF;
1153                 }
1154                 return;
1155
1156         default:
1157                 next_char();
1158                 if (!ignore_unknown_chars) {
1159                         errorf(&pp_token.base.source_position,
1160                                "unknown character '%c' found\n", input.c);
1161                         pp_token.kind = TP_ERROR;
1162                 } else {
1163                         pp_token.kind = input.c;
1164                 }
1165                 return;
1166         }
1167 }
1168
1169 static void print_quoted_string(const char *const string)
1170 {
1171         fputc('"', out);
1172         for (const char *c = string; *c != 0; ++c) {
1173                 switch (*c) {
1174                 case '"': fputs("\\\"", out); break;
1175                 case '\\':  fputs("\\\\", out); break;
1176                 case '\a':  fputs("\\a", out); break;
1177                 case '\b':  fputs("\\b", out); break;
1178                 case '\f':  fputs("\\f", out); break;
1179                 case '\n':  fputs("\\n", out); break;
1180                 case '\r':  fputs("\\r", out); break;
1181                 case '\t':  fputs("\\t", out); break;
1182                 case '\v':  fputs("\\v", out); break;
1183                 case '\?':  fputs("\\?", out); break;
1184                 default:
1185                         if (!isprint(*c)) {
1186                                 fprintf(out, "\\%03o", (unsigned)*c);
1187                                 break;
1188                         }
1189                         fputc(*c, out);
1190                         break;
1191                 }
1192         }
1193         fputc('"', out);
1194 }
1195
1196 static void print_line_directive(const source_position_t *pos, const char *add)
1197 {
1198         fprintf(out, "# %u ", pos->lineno);
1199         print_quoted_string(pos->input_name);
1200         if (add != NULL) {
1201                 fputc(' ', out);
1202                 fputs(add, out);
1203         }
1204
1205         printed_input_name = pos->input_name;
1206         input.output_line  = pos->lineno-1;
1207 }
1208
1209 static void emit_newlines(void)
1210 {
1211         unsigned delta = pp_token.base.source_position.lineno - input.output_line;
1212
1213         if (delta >= 9) {
1214                 fputc('\n', out);
1215                 print_line_directive(&pp_token.base.source_position, NULL);
1216                 fputc('\n', out);
1217         } else {
1218                 for (unsigned i = 0; i < delta; ++i) {
1219                         fputc('\n', out);
1220                 }
1221         }
1222         input.output_line = pp_token.base.source_position.lineno;
1223 }
1224
1225 static void emit_pp_token(void)
1226 {
1227         if (skip_mode)
1228                 return;
1229
1230         if (info.at_line_begin) {
1231                 emit_newlines();
1232
1233                 for (unsigned i = 0; i < info.whitespace; ++i)
1234                         fputc(' ', out);
1235
1236         } else if (info.had_whitespace ||
1237                            tokens_would_paste(last_token, pp_token.kind)) {
1238                 fputc(' ', out);
1239         }
1240
1241         switch (pp_token.kind) {
1242         case TP_IDENTIFIER:
1243                 fputs(pp_token.identifier.symbol->string, out);
1244                 break;
1245         case TP_NUMBER:
1246                 fputs(pp_token.number.number.begin, out);
1247                 break;
1248         case TP_WIDE_STRING_LITERAL:
1249                 fputc('L', out);
1250         case TP_STRING_LITERAL:
1251                 fputc('"', out);
1252                 fputs(pp_token.string.string.begin, out);
1253                 fputc('"', out);
1254                 break;
1255         case TP_WIDE_CHARACTER_CONSTANT:
1256                 fputc('L', out);
1257         case TP_CHARACTER_CONSTANT:
1258                 fputc('\'', out);
1259                 fputs(pp_token.string.string.begin, out);
1260                 fputc('\'', out);
1261                 break;
1262         default:
1263                 print_pp_token_kind(out, pp_token.kind);
1264                 break;
1265         }
1266         last_token = pp_token.kind;
1267 }
1268
1269 static void eat_pp_directive(void)
1270 {
1271         while (!info.at_line_begin) {
1272                 next_preprocessing_token();
1273         }
1274 }
1275
1276 static bool strings_equal(const string_t *string1, const string_t *string2)
1277 {
1278         size_t size = string1->size;
1279         if (size != string2->size)
1280                 return false;
1281
1282         const char *c1 = string1->begin;
1283         const char *c2 = string2->begin;
1284         for (size_t i = 0; i < size; ++i, ++c1, ++c2) {
1285                 if (*c1 != *c2)
1286                         return false;
1287         }
1288         return true;
1289 }
1290
1291 static bool pp_tokens_equal(const token_t *token1, const token_t *token2)
1292 {
1293         if (token1->kind != token2->kind)
1294                 return false;
1295
1296         switch (token1->kind) {
1297         case TP_IDENTIFIER:
1298                 return token1->identifier.symbol == token2->identifier.symbol;
1299         case TP_NUMBER:
1300         case TP_CHARACTER_CONSTANT:
1301         case TP_STRING_LITERAL:
1302                 return strings_equal(&token1->string.string, &token2->string.string);
1303
1304         default:
1305                 return true;
1306         }
1307 }
1308
1309 static bool pp_definitions_equal(const pp_definition_t *definition1,
1310                                  const pp_definition_t *definition2)
1311 {
1312         if (definition1->list_len != definition2->list_len)
1313                 return false;
1314
1315         size_t         len = definition1->list_len;
1316         const token_t *t1  = definition1->token_list;
1317         const token_t *t2  = definition2->token_list;
1318         for (size_t i = 0; i < len; ++i, ++t1, ++t2) {
1319                 if (!pp_tokens_equal(t1, t2))
1320                         return false;
1321         }
1322         return true;
1323 }
1324
1325 static void parse_define_directive(void)
1326 {
1327         eat_pp(TP_define);
1328         assert(obstack_object_size(&pp_obstack) == 0);
1329
1330         if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1331                 errorf(&pp_token.base.source_position,
1332                        "expected identifier after #define, got '%t'", &pp_token);
1333                 goto error_out;
1334         }
1335         symbol_t *symbol = pp_token.identifier.symbol;
1336
1337         pp_definition_t *new_definition
1338                 = obstack_alloc(&pp_obstack, sizeof(new_definition[0]));
1339         memset(new_definition, 0, sizeof(new_definition[0]));
1340         new_definition->source_position = input.position;
1341
1342         /* this is probably the only place where spaces are significant in the
1343          * lexer (except for the fact that they separate tokens). #define b(x)
1344          * is something else than #define b (x) */
1345         if (input.c == '(') {
1346                 /* eat the '(' */
1347                 next_preprocessing_token();
1348                 /* get next token after '(' */
1349                 next_preprocessing_token();
1350
1351                 while (true) {
1352                         switch (pp_token.kind) {
1353                         case TP_DOTDOTDOT:
1354                                 new_definition->is_variadic = true;
1355                                 next_preprocessing_token();
1356                                 if (pp_token.kind != ')') {
1357                                         errorf(&input.position,
1358                                                         "'...' not at end of macro argument list");
1359                                         goto error_out;
1360                                 }
1361                                 break;
1362                         case TP_IDENTIFIER:
1363                                 obstack_ptr_grow(&pp_obstack, pp_token.identifier.symbol);
1364                                 next_preprocessing_token();
1365
1366                                 if (pp_token.kind == ',') {
1367                                         next_preprocessing_token();
1368                                         break;
1369                                 }
1370
1371                                 if (pp_token.kind != ')') {
1372                                         errorf(&pp_token.base.source_position,
1373                                                "expected ',' or ')' after identifier, got '%t'",
1374                                                &pp_token);
1375                                         goto error_out;
1376                                 }
1377                                 break;
1378                         case ')':
1379                                 next_preprocessing_token();
1380                                 goto finish_argument_list;
1381                         default:
1382                                 errorf(&pp_token.base.source_position,
1383                                        "expected identifier, '...' or ')' in #define argument list, got '%t'",
1384                                        &pp_token);
1385                                 goto error_out;
1386                         }
1387                 }
1388
1389         finish_argument_list:
1390                 new_definition->has_parameters = true;
1391                 new_definition->n_parameters
1392                         = obstack_object_size(&pp_obstack) / sizeof(new_definition->parameters[0]);
1393                 new_definition->parameters = obstack_finish(&pp_obstack);
1394         } else {
1395                 next_preprocessing_token();
1396         }
1397
1398         /* construct a new pp_definition on the obstack */
1399         assert(obstack_object_size(&pp_obstack) == 0);
1400         size_t list_len = 0;
1401         while (!info.at_line_begin) {
1402                 obstack_grow(&pp_obstack, &pp_token, sizeof(pp_token));
1403                 ++list_len;
1404                 next_preprocessing_token();
1405         }
1406
1407         new_definition->list_len   = list_len;
1408         new_definition->token_list = obstack_finish(&pp_obstack);
1409
1410         pp_definition_t *old_definition = symbol->pp_definition;
1411         if (old_definition != NULL) {
1412                 if (!pp_definitions_equal(old_definition, new_definition)) {
1413                         warningf(WARN_OTHER, &input.position, "multiple definition of macro '%Y' (first defined %P)", symbol, &old_definition->source_position);
1414                 } else {
1415                         /* reuse the old definition */
1416                         obstack_free(&pp_obstack, new_definition);
1417                         new_definition = old_definition;
1418                 }
1419         }
1420
1421         symbol->pp_definition = new_definition;
1422         return;
1423
1424 error_out:
1425         if (obstack_object_size(&pp_obstack) > 0) {
1426                 char *ptr = obstack_finish(&pp_obstack);
1427                 obstack_free(&pp_obstack, ptr);
1428         }
1429         eat_pp_directive();
1430 }
1431
1432 static void parse_undef_directive(void)
1433 {
1434         eat_pp(TP_undef);
1435
1436         if (pp_token.kind != TP_IDENTIFIER) {
1437                 errorf(&input.position,
1438                        "expected identifier after #undef, got '%t'", &pp_token);
1439                 eat_pp_directive();
1440                 return;
1441         }
1442
1443         symbol_t *symbol = pp_token.identifier.symbol;
1444         symbol->pp_definition = NULL;
1445         next_preprocessing_token();
1446
1447         if (!info.at_line_begin) {
1448                 warningf(WARN_OTHER, &input.position, "extra tokens at end of #undef directive");
1449         }
1450         eat_pp_directive();
1451 }
1452
1453 static const char *parse_headername(void)
1454 {
1455         /* behind an #include we can have the special headername lexems.
1456          * They're only allowed behind an #include so they're not recognized
1457          * by the normal next_preprocessing_token. We handle them as a special
1458          * exception here */
1459         if (info.at_line_begin) {
1460                 parse_error("expected headername after #include");
1461                 return NULL;
1462         }
1463
1464         assert(obstack_object_size(&symbol_obstack) == 0);
1465
1466         /* check wether we have a "... or <... headername */
1467         switch (input.c) {
1468         case '<':
1469                 next_char();
1470                 while (true) {
1471                         switch (input.c) {
1472                         case EOF:
1473                                 /* fallthrough */
1474                         MATCH_NEWLINE(
1475                                 parse_error("header name without closing '>'");
1476                                 return NULL;
1477                         )
1478                         case '>':
1479                                 next_char();
1480                                 goto finished_headername;
1481                         }
1482                         obstack_1grow(&symbol_obstack, (char) input.c);
1483                         next_char();
1484                 }
1485                 /* we should never be here */
1486
1487         case '"':
1488                 next_char();
1489                 while (true) {
1490                         switch (input.c) {
1491                         case EOF:
1492                                 /* fallthrough */
1493                         MATCH_NEWLINE(
1494                                 parse_error("header name without closing '>'");
1495                                 return NULL;
1496                         )
1497                         case '"':
1498                                 next_char();
1499                                 goto finished_headername;
1500                         }
1501                         obstack_1grow(&symbol_obstack, (char) input.c);
1502                         next_char();
1503                 }
1504                 /* we should never be here */
1505
1506         default:
1507                 /* TODO: do normal pp_token parsing and concatenate results */
1508                 panic("pp_token concat include not implemented yet");
1509         }
1510
1511 finished_headername:
1512         obstack_1grow(&symbol_obstack, '\0');
1513         char *headername = obstack_finish(&symbol_obstack);
1514
1515         /* TODO: iterate search-path to find the file */
1516
1517         skip_whitespace();
1518
1519         return identify_string(headername);
1520 }
1521
1522 static bool do_include(bool system_include, const char *headername)
1523 {
1524         if (!system_include) {
1525                 /* for "bla" includes first try current dir
1526                  * TODO: this isn't correct, should be the directory of the source file
1527                  */
1528                 FILE *file = fopen(headername, "r");
1529                 if (file != NULL) {
1530                         switch_input(file, headername);
1531                         return true;
1532                 }
1533         }
1534
1535         size_t headername_len = strlen(headername);
1536         assert(obstack_object_size(&pp_obstack) == 0);
1537         /* check searchpath */
1538         for (searchpath_entry_t *entry = searchpath; entry != NULL;
1539              entry = entry->next) {
1540             const char *path = entry->path;
1541             size_t      len  = strlen(path);
1542                 obstack_grow(&pp_obstack, path, len);
1543                 if (path[len-1] != '/')
1544                         obstack_1grow(&pp_obstack, '/');
1545                 obstack_grow(&pp_obstack, headername, headername_len+1);
1546
1547                 char *complete_path = obstack_finish(&pp_obstack);
1548                 FILE *file          = fopen(complete_path, "r");
1549                 if (file != NULL) {
1550                         const char *filename = identify_string(complete_path);
1551                         switch_input(file, filename);
1552                         return true;
1553                 }
1554                 obstack_free(&pp_obstack, complete_path);
1555         }
1556
1557         return false;
1558 }
1559
1560 static bool parse_include_directive(void)
1561 {
1562         /* don't eat the TP_include here!
1563          * we need an alternative parsing for the next token */
1564         skip_whitespace();
1565         bool system_include = input.c == '<';
1566         const char *headername = parse_headername();
1567         if (headername == NULL) {
1568                 eat_pp_directive();
1569                 return false;
1570         }
1571
1572         if (!info.at_line_begin) {
1573                 warningf(WARN_OTHER, &pp_token.base.source_position,
1574                          "extra tokens at end of #include directive");
1575                 eat_pp_directive();
1576         }
1577
1578         if (n_inputs > INCLUDE_LIMIT) {
1579                 errorf(&pp_token.base.source_position, "#include nested too deeply");
1580                 /* eat \n or EOF */
1581                 next_preprocessing_token();
1582                 return false;
1583         }
1584
1585         /* we have to reenable space counting and macro expansion here,
1586          * because it is still disabled in directive parsing,
1587          * but we will trigger a preprocessing token reading of the new file
1588          * now and need expansions/space counting */
1589         in_pp_directive = false;
1590
1591         /* switch inputs */
1592         emit_newlines();
1593         push_input();
1594         bool res = do_include(system_include, headername);
1595         if (!res) {
1596                 errorf(&pp_token.base.source_position,
1597                        "failed including '%s': %s", headername, strerror(errno));
1598                 pop_restore_input();
1599                 return false;
1600         }
1601
1602         return true;
1603 }
1604
1605 static pp_conditional_t *push_conditional(void)
1606 {
1607         pp_conditional_t *conditional
1608                 = obstack_alloc(&pp_obstack, sizeof(*conditional));
1609         memset(conditional, 0, sizeof(*conditional));
1610
1611         conditional->parent = conditional_stack;
1612         conditional_stack   = conditional;
1613
1614         return conditional;
1615 }
1616
1617 static void pop_conditional(void)
1618 {
1619         assert(conditional_stack != NULL);
1620         conditional_stack = conditional_stack->parent;
1621 }
1622
1623 static void check_unclosed_conditionals(void)
1624 {
1625         while (conditional_stack != NULL) {
1626                 pp_conditional_t *conditional = conditional_stack;
1627
1628                 if (conditional->in_else) {
1629                         errorf(&conditional->source_position, "unterminated #else");
1630                 } else {
1631                         errorf(&conditional->source_position, "unterminated condition");
1632                 }
1633                 pop_conditional();
1634         }
1635 }
1636
1637 static void parse_ifdef_ifndef_directive(void)
1638 {
1639         bool is_ifndef = (pp_token.kind == TP_ifndef);
1640         bool condition;
1641         next_preprocessing_token();
1642
1643         if (skip_mode) {
1644                 eat_pp_directive();
1645                 pp_conditional_t *conditional = push_conditional();
1646                 conditional->source_position  = pp_token.base.source_position;
1647                 conditional->skip             = true;
1648                 return;
1649         }
1650
1651         if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1652                 errorf(&pp_token.base.source_position,
1653                        "expected identifier after #%s, got '%t'",
1654                        is_ifndef ? "ifndef" : "ifdef", &pp_token);
1655                 eat_pp_directive();
1656
1657                 /* just take the true case in the hope to avoid further errors */
1658                 condition = true;
1659         } else {
1660                 symbol_t        *symbol        = pp_token.identifier.symbol;
1661                 pp_definition_t *pp_definition = symbol->pp_definition;
1662                 next_preprocessing_token();
1663
1664                 if (!info.at_line_begin) {
1665                         errorf(&pp_token.base.source_position,
1666                                "extra tokens at end of #%s",
1667                                is_ifndef ? "ifndef" : "ifdef");
1668                         eat_pp_directive();
1669                 }
1670
1671                 /* evaluate wether we are in true or false case */
1672                 condition = is_ifndef ? pp_definition == NULL : pp_definition != NULL;
1673         }
1674
1675         pp_conditional_t *conditional = push_conditional();
1676         conditional->source_position  = pp_token.base.source_position;
1677         conditional->condition        = condition;
1678
1679         if (!condition) {
1680                 skip_mode = true;
1681         }
1682 }
1683
1684 static void parse_else_directive(void)
1685 {
1686         eat_pp(TP_else);
1687
1688         if (!info.at_line_begin) {
1689                 if (!skip_mode) {
1690                         warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #else");
1691                 }
1692                 eat_pp_directive();
1693         }
1694
1695         pp_conditional_t *conditional = conditional_stack;
1696         if (conditional == NULL) {
1697                 errorf(&pp_token.base.source_position, "#else without prior #if");
1698                 return;
1699         }
1700
1701         if (conditional->in_else) {
1702                 errorf(&pp_token.base.source_position,
1703                        "#else after #else (condition started %P)",
1704                        conditional->source_position);
1705                 skip_mode = true;
1706                 return;
1707         }
1708
1709         conditional->in_else = true;
1710         if (!conditional->skip) {
1711                 skip_mode = conditional->condition;
1712         }
1713         conditional->source_position = pp_token.base.source_position;
1714 }
1715
1716 static void parse_endif_directive(void)
1717 {
1718         eat_pp(TP_endif);
1719
1720         if (!info.at_line_begin) {
1721                 if (!skip_mode) {
1722                         warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #endif");
1723                 }
1724                 eat_pp_directive();
1725         }
1726
1727         pp_conditional_t *conditional = conditional_stack;
1728         if (conditional == NULL) {
1729                 errorf(&pp_token.base.source_position, "#endif without prior #if");
1730                 return;
1731         }
1732
1733         if (!conditional->skip) {
1734                 skip_mode = false;
1735         }
1736         pop_conditional();
1737 }
1738
1739 static void parse_preprocessing_directive(void)
1740 {
1741         in_pp_directive = true;
1742         eat_pp('#');
1743
1744         if (skip_mode) {
1745                 switch (pp_token.kind) {
1746                 case TP_ifdef:
1747                 case TP_ifndef:
1748                         parse_ifdef_ifndef_directive();
1749                         break;
1750                 case TP_else:
1751                         parse_else_directive();
1752                         break;
1753                 case TP_endif:
1754                         parse_endif_directive();
1755                         break;
1756                 default:
1757                         eat_pp_directive();
1758                         break;
1759                 }
1760         } else {
1761                 switch (pp_token.kind) {
1762                 case TP_define:
1763                         parse_define_directive();
1764                         break;
1765                 case TP_undef:
1766                         parse_undef_directive();
1767                         break;
1768                 case TP_ifdef:
1769                 case TP_ifndef:
1770                         parse_ifdef_ifndef_directive();
1771                         break;
1772                 case TP_else:
1773                         parse_else_directive();
1774                         break;
1775                 case TP_endif:
1776                         parse_endif_directive();
1777                         break;
1778                 case TP_include:
1779                         parse_include_directive();
1780                         break;
1781                 default:
1782                         if (info.at_line_begin) {
1783                                 /* the nop directive "#" */
1784                                 break;
1785                         }
1786                         errorf(&pp_token.base.source_position,
1787                                    "invalid preprocessing directive #%t", &pp_token);
1788                         eat_pp_directive();
1789                         break;
1790                 }
1791         }
1792
1793         in_pp_directive = false;
1794         assert(info.at_line_begin);
1795 }
1796
1797 static void prepend_include_path(const char *path)
1798 {
1799         searchpath_entry_t *entry = OALLOCZ(&config_obstack, searchpath_entry_t);
1800         entry->path = path;
1801         entry->next = searchpath;
1802         searchpath  = entry;
1803 }
1804
1805 static void setup_include_path(void)
1806 {
1807         /* built-in paths */
1808         prepend_include_path("/usr/include");
1809
1810         /* parse environment variable */
1811         const char *cpath = getenv("CPATH");
1812         if (cpath != NULL && *cpath != '\0') {
1813                 const char *begin = cpath;
1814                 const char *c;
1815                 do {
1816                         c = begin;
1817                         while (*c != '\0' && *c != ':')
1818                                 ++c;
1819
1820                         size_t len = c-begin;
1821                         if (len == 0) {
1822                                 /* for gcc compatibility (Matze: I would expect that
1823                                  * nothing happens for an empty entry...) */
1824                                 prepend_include_path(".");
1825                         } else {
1826                                 char *string = obstack_alloc(&config_obstack, len+1);
1827                                 memcpy(string, begin, len);
1828                                 string[len] = '\0';
1829
1830                                 prepend_include_path(string);
1831                         }
1832
1833                         begin = c+1;
1834                         /* skip : */
1835                         if (*begin == ':')
1836                                 ++begin;
1837                 } while(*c != '\0');
1838         }
1839 }
1840
1841 int pptest_main(int argc, char **argv);
1842 int pptest_main(int argc, char **argv)
1843 {
1844         init_symbol_table();
1845         init_tokens();
1846
1847         obstack_init(&config_obstack);
1848         obstack_init(&pp_obstack);
1849         obstack_init(&input_obstack);
1850         strset_init(&stringset);
1851
1852         setup_include_path();
1853
1854         /* simplistic commandline parser */
1855         const char *filename = NULL;
1856         for (int i = 1; i < argc; ++i) {
1857                 const char *opt = argv[i];
1858                 if (streq(opt, "-I")) {
1859                         prepend_include_path(argv[++i]);
1860                         continue;
1861                 } else if (streq(opt, "-E")) {
1862                         /* ignore */
1863                 } else if (opt[0] == '-') {
1864                         fprintf(stderr, "Unknown option '%s'\n", opt);
1865                 } else {
1866                         if (filename != NULL)
1867                                 fprintf(stderr, "Multiple inputs not supported\n");
1868                         filename = argv[i];
1869                 }
1870         }
1871         if (filename == NULL) {
1872                 fprintf(stderr, "No input specified\n");
1873                 return 1;
1874         }
1875
1876         out = stdout;
1877
1878         /* just here for gcc compatibility */
1879         fprintf(out, "# 1 \"%s\"\n", filename);
1880         fprintf(out, "# 1 \"<built-in>\"\n");
1881         fprintf(out, "# 1 \"<command-line>\"\n");
1882
1883         FILE *file = fopen(filename, "r");
1884         if (file == NULL) {
1885                 fprintf(stderr, "Couldn't open input '%s'\n", filename);
1886                 return 1;
1887         }
1888         switch_input(file, filename);
1889
1890         while (true) {
1891                 if (pp_token.kind == '#' && info.at_line_begin) {
1892                         parse_preprocessing_directive();
1893                         continue;
1894                 } else if (pp_token.kind == TP_EOF) {
1895                         goto end_of_main_loop;
1896                 } else if (pp_token.kind == TP_IDENTIFIER && !in_pp_directive) {
1897                         symbol_t *symbol = pp_token.identifier.symbol;
1898                         pp_definition_t *pp_definition = symbol->pp_definition;
1899                         if (pp_definition != NULL && !pp_definition->is_expanding) {
1900                                 expansion_pos = pp_token.base.source_position;
1901                                 if (pp_definition->has_parameters) {
1902                                         source_position_t position = pp_token.base.source_position;
1903                                         add_token_info_t old_info = info;
1904                                         next_preprocessing_token();
1905                                         add_token_info_t new_info = info;
1906
1907                                         /* no opening brace -> no expansion */
1908                                         if (pp_token.kind == '(') {
1909                                                 eat_pp('(');
1910
1911                                                 /* parse arguments (TODO) */
1912                                                 while (pp_token.kind != TP_EOF && pp_token.kind != ')')
1913                                                         next_preprocessing_token();
1914                                         } else {
1915                                                 token_t next_token = pp_token;
1916                                                 /* restore identifier token */
1917                                                 pp_token.kind                 = TP_IDENTIFIER;
1918                                                 pp_token.identifier.symbol    = symbol;
1919                                                 pp_token.base.source_position = position;
1920                                                 info = old_info;
1921                                                 emit_pp_token();
1922
1923                                                 info = new_info;
1924                                                 pp_token = next_token;
1925                                                 continue;
1926                                         }
1927                                         info = old_info;
1928                                 }
1929                                 pp_definition->expand_pos   = 0;
1930                                 pp_definition->is_expanding = true;
1931                                 current_expansion           = pp_definition;
1932                                 expand_next();
1933                                 continue;
1934                         }
1935                 }
1936
1937                 emit_pp_token();
1938                 next_preprocessing_token();
1939         }
1940 end_of_main_loop:
1941
1942         fputc('\n', out);
1943         check_unclosed_conditionals();
1944         close_input();
1945
1946         obstack_free(&input_obstack, NULL);
1947         obstack_free(&pp_obstack, NULL);
1948         obstack_free(&config_obstack, NULL);
1949
1950         strset_destroy(&stringset);
1951
1952         exit_tokens();
1953         exit_symbol_table();
1954
1955         return 0;
1956 }