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;
 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 const n = decode(input.input, input.buf + MAX_PUTBACK, lengthof(input.buf) - MAX_PUTBACK);
 199                 if (n == 0) {
 200                         input.c = EOF;
 201                         return;
 202                 }
 203                 input.bufpos = input.buf + MAX_PUTBACK;
 204                 input.bufend = input.bufpos + n;
 205         }
 206         input.c = *input.bufpos++;
 207         ++input.position.colno;
 208 }
 209
 210 /**
 211  * Put a character back into the buffer.
 212  *
 213  * @param pc  the character to put back
 214  */
 215 static inline void put_back(utf32 const pc)
 216 {
 217         assert(input.bufpos > input.buf);
 218         *(--input.bufpos - input.buf + input.buf) = (char) pc;
 219         --input.position.colno;
 220 }
 221
 222 #define MATCH_NEWLINE(code)                   \
 223         case '\r':                                \
 224                 next_char();                          \
 225                 if (input.c == '\n') {                \
 226         case '\n':                                \
 227                         next_char();                      \
 228                 }                                     \
 229                 info.whitespace = 0;                  \
 230                 ++input.position.lineno;              \
 231                 input.position.colno = 1;             \
 232                 code
 233
 234 #define eat(c_type) (assert(input.c == c_type), next_char())
 235
 236 static void maybe_concat_lines(void)
 237 {
 238         eat('\\');
 239
 240         switch (input.c) {
 241         MATCH_NEWLINE(
 242                 return;
 243         )
 244
 245         default:
 246                 break;
 247         }
 248
 249         put_back(input.c);
 250         input.c = '\\';
 251 }
 252
 253 /**
 254  * Set c to the next input character, ie.
 255  * after expanding trigraphs.
 256  */
 257 static inline void next_char(void)
 258 {
 259         next_real_char();
 260
 261         /* filter trigraphs and concatenated lines */
 262         if (UNLIKELY(input.c == '\\')) {
 263                 maybe_concat_lines();
 264                 goto end_of_next_char;
 265         }
 266
 267         if (LIKELY(input.c != '?'))
 268                 goto end_of_next_char;
 269
 270         next_real_char();
 271         if (LIKELY(input.c != '?')) {
 272                 put_back(input.c);
 273                 input.c = '?';
 274                 goto end_of_next_char;
 275         }
 276
 277         next_real_char();
 278         switch (input.c) {
 279         case '=': input.c = '#'; break;
 280         case '(': input.c = '['; break;
 281         case '/': input.c = '\\'; maybe_concat_lines(); break;
 282         case ')': input.c = ']'; 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         default:
 289                 put_back(input.c);
 290                 put_back('?');
 291                 input.c = '?';
 292                 break;
 293         }
 294
 295 end_of_next_char:;
 296 #ifdef DEBUG_CHARS
 297         printf("nchar '%c'\n", input.c);
 298 #endif
 299 }
 300
 301
 302
 303 /**
 304  * Returns true if the given char is a octal digit.
 305  *
 306  * @param char  the character to check
 307  */
 308 static inline bool is_octal_digit(int chr)
 309 {
 310         switch (chr) {
 311         case '0':
 312         case '1':
 313         case '2':
 314         case '3':
 315         case '4':
 316         case '5':
 317         case '6':
 318         case '7':
 319                 return true;
 320         default:
 321                 return false;
 322         }
 323 }
 324
 325 /**
 326  * Returns the value of a digit.
 327  * The only portable way to do it ...
 328  */
 329 static int digit_value(int digit)
 330 {
 331         switch (digit) {
 332         case '0': return 0;
 333         case '1': return 1;
 334         case '2': return 2;
 335         case '3': return 3;
 336         case '4': return 4;
 337         case '5': return 5;
 338         case '6': return 6;
 339         case '7': return 7;
 340         case '8': return 8;
 341         case '9': return 9;
 342         case 'a':
 343         case 'A': return 10;
 344         case 'b':
 345         case 'B': return 11;
 346         case 'c':
 347         case 'C': return 12;
 348         case 'd':
 349         case 'D': return 13;
 350         case 'e':
 351         case 'E': return 14;
 352         case 'f':
 353         case 'F': return 15;
 354         default:
 355                 panic("wrong character given");
 356         }
 357 }
 358
 359 /**
 360  * Parses an octal character sequence.
 361  *
 362  * @param first_digit  the already read first digit
 363  */
 364 static utf32 parse_octal_sequence(const utf32 first_digit)
 365 {
 366         assert(is_octal_digit(first_digit));
 367         utf32 value = digit_value(first_digit);
 368         if (!is_octal_digit(input.c)) return value;
 369         value = 8 * value + digit_value(input.c);
 370         next_char();
 371         if (!is_octal_digit(input.c)) return value;
 372         value = 8 * value + digit_value(input.c);
 373         next_char();
 374         return value;
 375
 376 }
 377
 378 /**
 379  * Parses a hex character sequence.
 380  */
 381 static utf32 parse_hex_sequence(void)
 382 {
 383         utf32 value = 0;
 384         while (isxdigit(input.c)) {
 385                 value = 16 * value + digit_value(input.c);
 386                 next_char();
 387         }
 388         return value;
 389 }
 390
 391 /**
 392  * Parse an escape sequence.
 393  */
 394 static utf32 parse_escape_sequence(void)
 395 {
 396         eat('\\');
 397
 398         utf32 const ec = input.c;
 399         next_char();
 400
 401         switch (ec) {
 402         case '"':  return '"';
 403         case '\'': return '\'';
 404         case '\\': return '\\';
 405         case '?': return '\?';
 406         case 'a': return '\a';
 407         case 'b': return '\b';
 408         case 'f': return '\f';
 409         case 'n': return '\n';
 410         case 'r': return '\r';
 411         case 't': return '\t';
 412         case 'v': return '\v';
 413         case 'x':
 414                 return parse_hex_sequence();
 415         case '0':
 416         case '1':
 417         case '2':
 418         case '3':
 419         case '4':
 420         case '5':
 421         case '6':
 422         case '7':
 423                 return parse_octal_sequence(ec);
 424         case EOF:
 425                 parse_error("reached end of file while parsing escape sequence");
 426                 return EOF;
 427         /* \E is not documented, but handled, by GCC.  It is acceptable according
 428          * to §6.11.4, whereas \e is not. */
 429         case 'E':
 430         case 'e':
 431                 if (c_mode & _GNUC)
 432                         return 27;   /* hopefully 27 is ALWAYS the code for ESCAPE */
 433                 break;
 434         case 'u':
 435         case 'U':
 436                 parse_error("universal character parsing not implemented yet");
 437                 return EOF;
 438         default:
 439                 break;
 440         }
 441         /* §6.4.4.4:8 footnote 64 */
 442         parse_error("unknown escape sequence");
 443         return EOF;
 444 }
 445
 446 static const char *identify_string(char *string)
 447 {
 448         const char *result = strset_insert(&stringset, string);
 449         if (result != string) {
 450                 obstack_free(&symbol_obstack, string);
 451         }
 452         return result;
 453 }
 454
 455 static string_t make_string(char *string, size_t len)
 456 {
 457         const char *result = identify_string(string);
 458         return (string_t) {result, len};
 459 }
 460
 461 static void parse_string_literal(void)
 462 {
 463         const unsigned start_linenr = input.position.lineno;
 464
 465         eat('"');
 466
 467         while (true) {
 468                 switch (input.c) {
 469                 case '\\': {
 470                         utf32 tc;
 471                         if (resolve_escape_sequences) {
 472                                 tc = parse_escape_sequence();
 473                                 obstack_1grow(&symbol_obstack, (char) tc);
 474                         } else {
 475                                 obstack_1grow(&symbol_obstack, (char) input.c);
 476                                 next_char();
 477                                 obstack_1grow(&symbol_obstack, (char) input.c);
 478                                 next_char();
 479                         }
 480                         break;
 481                 }
 482
 483                 case EOF: {
 484                         source_position_t source_position;
 485                         source_position.input_name = pp_token.base.source_position.input_name;
 486                         source_position.lineno     = start_linenr;
 487                         errorf(&source_position, "string has no end");
 488                         goto end_of_string;
 489                 }
 490
 491                 case '"':
 492                         next_char();
 493                         goto end_of_string;
 494
 495                 default:
 496                         obstack_grow_symbol(&symbol_obstack, input.c);
 497                         next_char();
 498                         break;
 499                 }
 500         }
 501
 502 end_of_string:
 503         /* add finishing 0 to the string */
 504         obstack_1grow(&symbol_obstack, '\0');
 505         const size_t size   = (size_t)obstack_object_size(&symbol_obstack);
 506         char *const  string = obstack_finish(&symbol_obstack);
 507
 508         pp_token.kind          = TP_STRING_LITERAL;
 509         pp_token.string.string = make_string(string, size);
 510 }
 511
 512 /**
 513  * Parse a wide string literal and set lexer_token.
 514  */
 515 static void parse_wide_string_literal(void)
 516 {
 517         parse_string_literal();
 518         if (pp_token.kind == TP_STRING_LITERAL)
 519                 pp_token.kind = TP_WIDE_STRING_LITERAL;
 520 }
 521
 522 static void parse_wide_character_constant(void)
 523 {
 524         eat('\'');
 525
 526         while (true) {
 527                 switch (input.c) {
 528                 case '\\': {
 529                         const utf32 tc = parse_escape_sequence();
 530                         obstack_grow_symbol(&symbol_obstack, tc);
 531                         break;
 532                 }
 533
 534                 MATCH_NEWLINE(
 535                         parse_error("newline while parsing character constant");
 536                         break;
 537                 )
 538
 539                 case '\'':
 540                         next_char();
 541                         goto end_of_wide_char_constant;
 542
 543                 case EOF:
 544                         parse_error("EOF while parsing character constant");
 545                         goto end_of_wide_char_constant;
 546
 547                 default:
 548                         obstack_grow_symbol(&symbol_obstack, input.c);
 549                         next_char();
 550                         break;
 551                 }
 552         }
 553
 554 end_of_wide_char_constant:
 555         obstack_1grow(&symbol_obstack, '\0');
 556         size_t  size = (size_t) obstack_object_size(&symbol_obstack)-1;
 557         char   *string = obstack_finish(&symbol_obstack);
 558         pp_token.kind          = TP_WIDE_CHARACTER_CONSTANT;
 559         pp_token.string.string = make_string(string, size);
 560
 561         if (size == 0) {
 562                 parse_error("empty character constant");
 563         }
 564 }
 565
 566 static void parse_character_constant(void)
 567 {
 568         const unsigned start_linenr = input.position.lineno;
 569
 570         eat('\'');
 571
 572         int tc;
 573         while (true) {
 574                 switch (input.c) {
 575                 case '\\':
 576                         tc = parse_escape_sequence();
 577                         obstack_1grow(&symbol_obstack, (char) tc);
 578                         break;
 579
 580                 MATCH_NEWLINE(
 581                         parse_error("newline while parsing character constant");
 582                         break;
 583                 )
 584
 585                 case EOF: {
 586                         source_position_t source_position;
 587                         source_position.input_name = pp_token.base.source_position.input_name;
 588                         source_position.lineno     = start_linenr;
 589                         errorf(&source_position, "EOF while parsing character constant");
 590                         goto end_of_char_constant;
 591                 }
 592
 593                 case '\'':
 594                         next_char();
 595                         goto end_of_char_constant;
 596
 597                 default:
 598                         obstack_1grow(&symbol_obstack, (char) input.c);
 599                         next_char();
 600                         break;
 601
 602                 }
 603         }
 604
 605 end_of_char_constant:;
 606         obstack_1grow(&symbol_obstack, '\0');
 607         const size_t size   = (size_t)obstack_object_size(&symbol_obstack);
 608         char *const  string = obstack_finish(&symbol_obstack);
 609
 610         pp_token.kind          = TP_CHARACTER_CONSTANT;
 611         pp_token.string.string = make_string(string, size);
 612
 613         if (size == 0) {
 614                 parse_error("empty character constant");
 615         }
 616 }
 617
 618 #define SYMBOL_CHARS_WITHOUT_E_P \
 619         case 'a': \
 620         case 'b': \
 621         case 'c': \
 622         case 'd': \
 623         case 'f': \
 624         case 'g': \
 625         case 'h': \
 626         case 'i': \
 627         case 'j': \
 628         case 'k': \
 629         case 'l': \
 630         case 'm': \
 631         case 'n': \
 632         case 'o': \
 633         case 'q': \
 634         case 'r': \
 635         case 's': \
 636         case 't': \
 637         case 'u': \
 638         case 'v': \
 639         case 'w': \
 640         case 'x': \
 641         case 'y': \
 642         case 'z': \
 643         case 'A': \
 644         case 'B': \
 645         case 'C': \
 646         case 'D': \
 647         case 'F': \
 648         case 'G': \
 649         case 'H': \
 650         case 'I': \
 651         case 'J': \
 652         case 'K': \
 653         case 'L': \
 654         case 'M': \
 655         case 'N': \
 656         case 'O': \
 657         case 'Q': \
 658         case 'R': \
 659         case 'S': \
 660         case 'T': \
 661         case 'U': \
 662         case 'V': \
 663         case 'W': \
 664         case 'X': \
 665         case 'Y': \
 666         case 'Z': \
 667         case '_':
 668
 669 #define SYMBOL_CHARS \
 670         SYMBOL_CHARS_WITHOUT_E_P \
 671         case 'e': \
 672         case 'p': \
 673         case 'E': \
 674         case 'P':
 675
 676 #define DIGITS \
 677         case '0':  \
 678         case '1':  \
 679         case '2':  \
 680         case '3':  \
 681         case '4':  \
 682         case '5':  \
 683         case '6':  \
 684         case '7':  \
 685         case '8':  \
 686         case '9':
 687
 688 /**
 689  * returns next final token from a preprocessor macro expansion
 690  */
 691 static void expand_next(void)
 692 {
 693         assert(current_expansion != NULL);
 694
 695         pp_definition_t *definition = current_expansion;
 696
 697 restart:
 698         if (definition->list_len == 0
 699                         || definition->expand_pos >= definition->list_len) {
 700                 /* we're finished with the current macro, move up 1 level in the
 701                  * expansion stack */
 702                 pp_definition_t *parent = definition->parent_expansion;
 703                 definition->parent_expansion = NULL;
 704                 definition->is_expanding     = false;
 705
 706                 /* it was the outermost expansion, parse normal pptoken */
 707                 if (parent == NULL) {
 708                         current_expansion = NULL;
 709                         next_preprocessing_token();
 710                         return;
 711                 }
 712                 definition        = parent;
 713                 current_expansion = definition;
 714                 goto restart;
 715         }
 716         pp_token = definition->token_list[definition->expand_pos];
 717         pp_token.base.source_position = expansion_pos;
 718         ++definition->expand_pos;
 719
 720         if (pp_token.kind != TP_IDENTIFIER)
 721                 return;
 722
 723         /* if it was an identifier then we might need to expand again */
 724         pp_definition_t *const symbol_definition = pp_token.base.symbol->pp_definition;
 725         if (symbol_definition != NULL && !symbol_definition->is_expanding) {
 726                 symbol_definition->parent_expansion = definition;
 727                 symbol_definition->expand_pos       = 0;
 728                 symbol_definition->is_expanding     = true;
 729                 definition                          = symbol_definition;
 730                 current_expansion                   = definition;
 731                 goto restart;
 732         }
 733 }
 734
 735 static void skip_line_comment(void)
 736 {
 737         while (true) {
 738                 switch (input.c) {
 739                 case EOF:
 740                         return;
 741
 742                 case '\r':
 743                 case '\n':
 744                         return;
 745
 746                 default:
 747                         next_char();
 748                         break;
 749                 }
 750         }
 751 }
 752
 753 static void skip_multiline_comment(void)
 754 {
 755         unsigned start_linenr = input.position.lineno;
 756         while (true) {
 757                 switch (input.c) {
 758                 case '/':
 759                         next_char();
 760                         if (input.c == '*') {
 761                                 /* TODO: nested comment, warn here */
 762                         }
 763                         break;
 764                 case '*':
 765                         next_char();
 766                         if (input.c == '/') {
 767                                 next_char();
 768                                 info.whitespace += input.position.colno-1;
 769                                 return;
 770                         }
 771                         break;
 772
 773                 MATCH_NEWLINE(
 774                         info.at_line_begin |= !in_pp_directive;
 775                         break;
 776                 )
 777
 778                 case EOF: {
 779                         source_position_t source_position;
 780                         source_position.input_name = pp_token.base.source_position.input_name;
 781                         source_position.lineno     = start_linenr;
 782                         errorf(&source_position, "at end of file while looking for comment end");
 783                         return;
 784                 }
 785
 786                 default:
 787                         next_char();
 788                         break;
 789                 }
 790         }
 791 }
 792
 793 static void skip_whitespace(void)
 794 {
 795         while (true) {
 796                 switch (input.c) {
 797                 case ' ':
 798                 case '\t':
 799                         next_char();
 800                         continue;
 801
 802                 MATCH_NEWLINE(
 803                         info.at_line_begin = true;
 804                         return;
 805                 )
 806
 807                 case '/':
 808                         next_char();
 809                         if (input.c == '/') {
 810                                 next_char();
 811                                 skip_line_comment();
 812                                 continue;
 813                         } else if (input.c == '*') {
 814                                 next_char();
 815                                 skip_multiline_comment();
 816                                 continue;
 817                         } else {
 818                                 put_back(input.c);
 819                                 input.c = '/';
 820                         }
 821                         return;
 822                 default:
 823                         return;
 824                 }
 825         }
 826 }
 827
 828 static void eat_pp(preprocessor_token_kind_t const type)
 829 {
 830         (void) type;
 831         assert(pp_token.kind == type);
 832         next_preprocessing_token();
 833 }
 834
 835 static void parse_symbol(void)
 836 {
 837         obstack_1grow(&symbol_obstack, (char) input.c);
 838         next_char();
 839
 840         while (true) {
 841                 switch (input.c) {
 842                 DIGITS
 843                 SYMBOL_CHARS
 844                         obstack_1grow(&symbol_obstack, (char) input.c);
 845                         next_char();
 846                         break;
 847
 848                 default:
 849                         goto end_symbol;
 850                 }
 851         }
 852
 853 end_symbol:
 854         obstack_1grow(&symbol_obstack, '\0');
 855         char *string = obstack_finish(&symbol_obstack);
 856
 857         /* might be a wide string or character constant ( L"string"/L'c' ) */
 858         if (input.c == '"' && string[0] == 'L' && string[1] == '\0') {
 859                 obstack_free(&symbol_obstack, string);
 860                 parse_wide_string_literal();
 861                 return;
 862         } else if (input.c == '\'' && string[0] == 'L' && string[1] == '\0') {
 863                 obstack_free(&symbol_obstack, string);
 864                 parse_wide_character_constant();
 865                 return;
 866         }
 867
 868         symbol_t *symbol = symbol_table_insert(string);
 869
 870         pp_token.kind        = symbol->pp_ID;
 871         pp_token.base.symbol = symbol;
 872
 873         /* we can free the memory from symbol obstack if we already had an entry in
 874          * the symbol table */
 875         if (symbol->string != string) {
 876                 obstack_free(&symbol_obstack, string);
 877         }
 878 }
 879
 880 static void parse_number(void)
 881 {
 882         obstack_1grow(&symbol_obstack, (char) input.c);
 883         next_char();
 884
 885         while (true) {
 886                 switch (input.c) {
 887                 case '.':
 888                 DIGITS
 889                 SYMBOL_CHARS_WITHOUT_E_P
 890                         obstack_1grow(&symbol_obstack, (char) input.c);
 891                         next_char();
 892                         break;
 893
 894                 case 'e':
 895                 case 'p':
 896                 case 'E':
 897                 case 'P':
 898                         obstack_1grow(&symbol_obstack, (char) input.c);
 899                         next_char();
 900                         if (input.c == '+' || input.c == '-') {
 901                                 obstack_1grow(&symbol_obstack, (char) input.c);
 902                                 next_char();
 903                         }
 904                         break;
 905
 906                 default:
 907                         goto end_number;
 908                 }
 909         }
 910
 911 end_number:
 912         obstack_1grow(&symbol_obstack, '\0');
 913         size_t  size   = obstack_object_size(&symbol_obstack);
 914         char   *string = obstack_finish(&symbol_obstack);
 915
 916         pp_token.kind          = TP_NUMBER;
 917         pp_token.number.number = make_string(string, size);
 918 }
 919
 920
 921 #define MAYBE_PROLOG                                       \
 922                         next_char();                                   \
 923                         while (true) {                                 \
 924                                 switch (input.c) {
 925
 926 #define MAYBE(ch, set_type)                                \
 927                                 case ch:                                   \
 928                                         next_char();                           \
 929                                         pp_token.kind = set_type;              \
 930                                         return;
 931
 932 #define ELSE_CODE(code)                                    \
 933                                 default:                                   \
 934                                         code                                   \
 935                                         return;                                \
 936                                 }                                          \
 937                         }
 938
 939 #define ELSE(set_type)                                     \
 940                 ELSE_CODE(                                         \
 941                         pp_token.kind = set_type;                      \
 942                 )
 943
 944 static void next_preprocessing_token(void)
 945 {
 946         if (current_expansion != NULL) {
 947                 expand_next();
 948                 return;
 949         }
 950
 951         info.at_line_begin  = false;
 952         info.had_whitespace = false;
 953 restart:
 954         pp_token.base.source_position = input.position;
 955         pp_token.base.symbol          = NULL;
 956
 957         switch (input.c) {
 958         case ' ':
 959         case '\t':
 960                 ++info.whitespace;
 961                 info.had_whitespace = true;
 962                 next_char();
 963                 goto restart;
 964
 965         MATCH_NEWLINE(
 966                 info.at_line_begin = true;
 967                 info.had_whitespace = true;
 968                 goto restart;
 969         )
 970
 971         SYMBOL_CHARS
 972                 parse_symbol();
 973                 return;
 974
 975         DIGITS
 976                 parse_number();
 977                 return;
 978
 979         case '"':
 980                 parse_string_literal();
 981                 return;
 982
 983         case '\'':
 984                 parse_character_constant();
 985                 return;
 986
 987         case '.':
 988                 MAYBE_PROLOG
 989                         case '0':
 990                         case '1':
 991                         case '2':
 992                         case '3':
 993                         case '4':
 994                         case '5':
 995                         case '6':
 996                         case '7':
 997                         case '8':
 998                         case '9':
 999                                 put_back(input.c);
1000                                 input.c = '.';
1001                                 parse_number();
1002                                 return;
1003
1004                         case '.':
1005                                 MAYBE_PROLOG
1006                                 MAYBE('.', TP_DOTDOTDOT)
1007                                 ELSE_CODE(
1008                                         put_back(input.c);
1009                                         input.c = '.';
1010                                         pp_token.kind = '.';
1011                                 )
1012                 ELSE('.')
1013         case '&':
1014                 MAYBE_PROLOG
1015                 MAYBE('&', TP_ANDAND)
1016                 MAYBE('=', TP_ANDEQUAL)
1017                 ELSE('&')
1018         case '*':
1019                 MAYBE_PROLOG
1020                 MAYBE('=', TP_ASTERISKEQUAL)
1021                 ELSE('*')
1022         case '+':
1023                 MAYBE_PROLOG
1024                 MAYBE('+', TP_PLUSPLUS)
1025                 MAYBE('=', TP_PLUSEQUAL)
1026                 ELSE('+')
1027         case '-':
1028                 MAYBE_PROLOG
1029                 MAYBE('>', TP_MINUSGREATER)
1030                 MAYBE('-', TP_MINUSMINUS)
1031                 MAYBE('=', TP_MINUSEQUAL)
1032                 ELSE('-')
1033         case '!':
1034                 MAYBE_PROLOG
1035                 MAYBE('=', TP_EXCLAMATIONMARKEQUAL)
1036                 ELSE('!')
1037         case '/':
1038                 MAYBE_PROLOG
1039                 MAYBE('=', TP_SLASHEQUAL)
1040                         case '*':
1041                                 next_char();
1042                                 info.had_whitespace = true;
1043                                 skip_multiline_comment();
1044                                 goto restart;
1045                         case '/':
1046                                 next_char();
1047                                 info.had_whitespace = true;
1048                                 skip_line_comment();
1049                                 goto restart;
1050                 ELSE('/')
1051         case '%':
1052                 MAYBE_PROLOG
1053                 MAYBE('>', '}')
1054                 MAYBE('=', TP_PERCENTEQUAL)
1055                         case ':':
1056                                 MAYBE_PROLOG
1057                                         case '%':
1058                                                 MAYBE_PROLOG
1059                                                 MAYBE(':', TP_HASHHASH)
1060                                                 ELSE_CODE(
1061                                                         put_back(input.c);
1062                                                         input.c = '%';
1063                                                         pp_token.kind = '#';
1064                                                 )
1065                                 ELSE('#')
1066                 ELSE('%')
1067         case '<':
1068                 MAYBE_PROLOG
1069                 MAYBE(':', '[')
1070                 MAYBE('%', '{')
1071                 MAYBE('=', TP_LESSEQUAL)
1072                         case '<':
1073                                 MAYBE_PROLOG
1074                                 MAYBE('=', TP_LESSLESSEQUAL)
1075                                 ELSE(TP_LESSLESS)
1076                 ELSE('<')
1077         case '>':
1078                 MAYBE_PROLOG
1079                 MAYBE('=', TP_GREATEREQUAL)
1080                         case '>':
1081                                 MAYBE_PROLOG
1082                                 MAYBE('=', TP_GREATERGREATEREQUAL)
1083                                 ELSE(TP_GREATERGREATER)
1084                 ELSE('>')
1085         case '^':
1086                 MAYBE_PROLOG
1087                 MAYBE('=', TP_CARETEQUAL)
1088                 ELSE('^')
1089         case '|':
1090                 MAYBE_PROLOG
1091                 MAYBE('=', TP_PIPEEQUAL)
1092                 MAYBE('|', TP_PIPEPIPE)
1093                 ELSE('|')
1094         case ':':
1095                 MAYBE_PROLOG
1096                 MAYBE('>', ']')
1097                 ELSE(':')
1098         case '=':
1099                 MAYBE_PROLOG
1100                 MAYBE('=', TP_EQUALEQUAL)
1101                 ELSE('=')
1102         case '#':
1103                 MAYBE_PROLOG
1104                 MAYBE('#', TP_HASHHASH)
1105                 ELSE_CODE(
1106                         pp_token.kind = '#';
1107                 )
1108
1109         case '?':
1110         case '[':
1111         case ']':
1112         case '(':
1113         case ')':
1114         case '{':
1115         case '}':
1116         case '~':
1117         case ';':
1118         case ',':
1119         case '\\':
1120                 pp_token.kind = input.c;
1121                 next_char();
1122                 return;
1123
1124         case EOF:
1125                 if (input_stack != NULL) {
1126                         close_input();
1127                         pop_restore_input();
1128                         fputc('\n', out);
1129                         print_line_directive(&input.position, "2");
1130                         goto restart;
1131                 } else {
1132                         pp_token.base.source_position.lineno++;
1133                         info.at_line_begin = true;
1134                         pp_token.kind = TP_EOF;
1135                 }
1136                 return;
1137
1138         default:
1139                 next_char();
1140                 if (!ignore_unknown_chars) {
1141                         errorf(&pp_token.base.source_position,
1142                                "unknown character '%c' found\n", input.c);
1143                         goto restart;
1144                 } else {
1145                         pp_token.kind = input.c;
1146                         return;
1147                 }
1148         }
1149 }
1150
1151 static void print_quoted_string(const char *const string)
1152 {
1153         fputc('"', out);
1154         for (const char *c = string; *c != 0; ++c) {
1155                 switch (*c) {
1156                 case '"': fputs("\\\"", out); break;
1157                 case '\\':  fputs("\\\\", out); break;
1158                 case '\a':  fputs("\\a", out); break;
1159                 case '\b':  fputs("\\b", out); break;
1160                 case '\f':  fputs("\\f", out); break;
1161                 case '\n':  fputs("\\n", out); break;
1162                 case '\r':  fputs("\\r", out); break;
1163                 case '\t':  fputs("\\t", out); break;
1164                 case '\v':  fputs("\\v", out); break;
1165                 case '\?':  fputs("\\?", out); break;
1166                 default:
1167                         if (!isprint(*c)) {
1168                                 fprintf(out, "\\%03o", (unsigned)*c);
1169                                 break;
1170                         }
1171                         fputc(*c, out);
1172                         break;
1173                 }
1174         }
1175         fputc('"', out);
1176 }
1177
1178 static void print_line_directive(const source_position_t *pos, const char *add)
1179 {
1180         fprintf(out, "# %u ", pos->lineno);
1181         print_quoted_string(pos->input_name);
1182         if (add != NULL) {
1183                 fputc(' ', out);
1184                 fputs(add, out);
1185         }
1186
1187         printed_input_name = pos->input_name;
1188         input.output_line  = pos->lineno-1;
1189 }
1190
1191 static void emit_newlines(void)
1192 {
1193         unsigned delta = pp_token.base.source_position.lineno - input.output_line;
1194
1195         if (delta >= 9) {
1196                 fputc('\n', out);
1197                 print_line_directive(&pp_token.base.source_position, NULL);
1198                 fputc('\n', out);
1199         } else {
1200                 for (unsigned i = 0; i < delta; ++i) {
1201                         fputc('\n', out);
1202                 }
1203         }
1204         input.output_line = pp_token.base.source_position.lineno;
1205 }
1206
1207 static void emit_pp_token(void)
1208 {
1209         if (skip_mode)
1210                 return;
1211
1212         if (info.at_line_begin) {
1213                 emit_newlines();
1214
1215                 for (unsigned i = 0; i < info.whitespace; ++i)
1216                         fputc(' ', out);
1217
1218         } else if (info.had_whitespace ||
1219                            tokens_would_paste(last_token, pp_token.kind)) {
1220                 fputc(' ', out);
1221         }
1222
1223         switch (pp_token.kind) {
1224         case TP_IDENTIFIER:
1225                 fputs(pp_token.base.symbol->string, out);
1226                 break;
1227         case TP_NUMBER:
1228                 fputs(pp_token.number.number.begin, out);
1229                 break;
1230         case TP_WIDE_STRING_LITERAL:
1231                 fputc('L', out);
1232         case TP_STRING_LITERAL:
1233                 fputc('"', out);
1234                 fputs(pp_token.string.string.begin, out);
1235                 fputc('"', out);
1236                 break;
1237         case TP_WIDE_CHARACTER_CONSTANT:
1238                 fputc('L', out);
1239         case TP_CHARACTER_CONSTANT:
1240                 fputc('\'', out);
1241                 fputs(pp_token.string.string.begin, out);
1242                 fputc('\'', out);
1243                 break;
1244         default:
1245                 print_pp_token_kind(out, pp_token.kind);
1246                 break;
1247         }
1248         last_token = pp_token.kind;
1249 }
1250
1251 static void eat_pp_directive(void)
1252 {
1253         while (!info.at_line_begin) {
1254                 next_preprocessing_token();
1255         }
1256 }
1257
1258 static bool strings_equal(const string_t *string1, const string_t *string2)
1259 {
1260         size_t size = string1->size;
1261         if (size != string2->size)
1262                 return false;
1263
1264         const char *c1 = string1->begin;
1265         const char *c2 = string2->begin;
1266         for (size_t i = 0; i < size; ++i, ++c1, ++c2) {
1267                 if (*c1 != *c2)
1268                         return false;
1269         }
1270         return true;
1271 }
1272
1273 static bool pp_tokens_equal(const token_t *token1, const token_t *token2)
1274 {
1275         if (token1->kind != token2->kind)
1276                 return false;
1277
1278         switch (token1->kind) {
1279         case TP_IDENTIFIER:
1280                 return token1->base.symbol == token2->base.symbol;
1281
1282         case TP_NUMBER:
1283         case TP_CHARACTER_CONSTANT:
1284         case TP_STRING_LITERAL:
1285                 return strings_equal(&token1->string.string, &token2->string.string);
1286
1287         default:
1288                 return true;
1289         }
1290 }
1291
1292 static bool pp_definitions_equal(const pp_definition_t *definition1,
1293                                  const pp_definition_t *definition2)
1294 {
1295         if (definition1->list_len != definition2->list_len)
1296                 return false;
1297
1298         size_t         len = definition1->list_len;
1299         const token_t *t1  = definition1->token_list;
1300         const token_t *t2  = definition2->token_list;
1301         for (size_t i = 0; i < len; ++i, ++t1, ++t2) {
1302                 if (!pp_tokens_equal(t1, t2))
1303                         return false;
1304         }
1305         return true;
1306 }
1307
1308 static void parse_define_directive(void)
1309 {
1310         eat_pp(TP_define);
1311         assert(obstack_object_size(&pp_obstack) == 0);
1312
1313         if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1314                 errorf(&pp_token.base.source_position,
1315                        "expected identifier after #define, got '%t'", &pp_token);
1316                 goto error_out;
1317         }
1318         symbol_t *const symbol = pp_token.base.symbol;
1319
1320         pp_definition_t *new_definition
1321                 = obstack_alloc(&pp_obstack, sizeof(new_definition[0]));
1322         memset(new_definition, 0, sizeof(new_definition[0]));
1323         new_definition->source_position = input.position;
1324
1325         /* this is probably the only place where spaces are significant in the
1326          * lexer (except for the fact that they separate tokens). #define b(x)
1327          * is something else than #define b (x) */
1328         if (input.c == '(') {
1329                 /* eat the '(' */
1330                 next_preprocessing_token();
1331                 /* get next token after '(' */
1332                 next_preprocessing_token();
1333
1334                 while (true) {
1335                         switch (pp_token.kind) {
1336                         case TP_DOTDOTDOT:
1337                                 new_definition->is_variadic = true;
1338                                 next_preprocessing_token();
1339                                 if (pp_token.kind != ')') {
1340                                         errorf(&input.position,
1341                                                         "'...' not at end of macro argument list");
1342                                         goto error_out;
1343                                 }
1344                                 break;
1345                         case TP_IDENTIFIER:
1346                                 obstack_ptr_grow(&pp_obstack, pp_token.base.symbol);
1347                                 next_preprocessing_token();
1348
1349                                 if (pp_token.kind == ',') {
1350                                         next_preprocessing_token();
1351                                         break;
1352                                 }
1353
1354                                 if (pp_token.kind != ')') {
1355                                         errorf(&pp_token.base.source_position,
1356                                                "expected ',' or ')' after identifier, got '%t'",
1357                                                &pp_token);
1358                                         goto error_out;
1359                                 }
1360                                 break;
1361                         case ')':
1362                                 next_preprocessing_token();
1363                                 goto finish_argument_list;
1364                         default:
1365                                 errorf(&pp_token.base.source_position,
1366                                        "expected identifier, '...' or ')' in #define argument list, got '%t'",
1367                                        &pp_token);
1368                                 goto error_out;
1369                         }
1370                 }
1371
1372         finish_argument_list:
1373                 new_definition->has_parameters = true;
1374                 new_definition->n_parameters
1375                         = obstack_object_size(&pp_obstack) / sizeof(new_definition->parameters[0]);
1376                 new_definition->parameters = obstack_finish(&pp_obstack);
1377         } else {
1378                 next_preprocessing_token();
1379         }
1380
1381         /* construct a new pp_definition on the obstack */
1382         assert(obstack_object_size(&pp_obstack) == 0);
1383         size_t list_len = 0;
1384         while (!info.at_line_begin) {
1385                 obstack_grow(&pp_obstack, &pp_token, sizeof(pp_token));
1386                 ++list_len;
1387                 next_preprocessing_token();
1388         }
1389
1390         new_definition->list_len   = list_len;
1391         new_definition->token_list = obstack_finish(&pp_obstack);
1392
1393         pp_definition_t *old_definition = symbol->pp_definition;
1394         if (old_definition != NULL) {
1395                 if (!pp_definitions_equal(old_definition, new_definition)) {
1396                         warningf(WARN_OTHER, &input.position, "multiple definition of macro '%Y' (first defined %P)", symbol, &old_definition->source_position);
1397                 } else {
1398                         /* reuse the old definition */
1399                         obstack_free(&pp_obstack, new_definition);
1400                         new_definition = old_definition;
1401                 }
1402         }
1403
1404         symbol->pp_definition = new_definition;
1405         return;
1406
1407 error_out:
1408         if (obstack_object_size(&pp_obstack) > 0) {
1409                 char *ptr = obstack_finish(&pp_obstack);
1410                 obstack_free(&pp_obstack, ptr);
1411         }
1412         eat_pp_directive();
1413 }
1414
1415 static void parse_undef_directive(void)
1416 {
1417         eat_pp(TP_undef);
1418
1419         if (pp_token.kind != TP_IDENTIFIER) {
1420                 errorf(&input.position,
1421                        "expected identifier after #undef, got '%t'", &pp_token);
1422                 eat_pp_directive();
1423                 return;
1424         }
1425
1426         pp_token.base.symbol->pp_definition = NULL;
1427         next_preprocessing_token();
1428
1429         if (!info.at_line_begin) {
1430                 warningf(WARN_OTHER, &input.position, "extra tokens at end of #undef directive");
1431         }
1432         eat_pp_directive();
1433 }
1434
1435 static const char *parse_headername(void)
1436 {
1437         /* behind an #include we can have the special headername lexems.
1438          * They're only allowed behind an #include so they're not recognized
1439          * by the normal next_preprocessing_token. We handle them as a special
1440          * exception here */
1441         if (info.at_line_begin) {
1442                 parse_error("expected headername after #include");
1443                 return NULL;
1444         }
1445
1446         assert(obstack_object_size(&symbol_obstack) == 0);
1447
1448         /* check wether we have a "... or <... headername */
1449         switch (input.c) {
1450         case '<':
1451                 next_char();
1452                 while (true) {
1453                         switch (input.c) {
1454                         case EOF:
1455                                 /* fallthrough */
1456                         MATCH_NEWLINE(
1457                                 parse_error("header name without closing '>'");
1458                                 return NULL;
1459                         )
1460                         case '>':
1461                                 next_char();
1462                                 goto finished_headername;
1463                         }
1464                         obstack_1grow(&symbol_obstack, (char) input.c);
1465                         next_char();
1466                 }
1467                 /* we should never be here */
1468
1469         case '"':
1470                 next_char();
1471                 while (true) {
1472                         switch (input.c) {
1473                         case EOF:
1474                                 /* fallthrough */
1475                         MATCH_NEWLINE(
1476                                 parse_error("header name without closing '>'");
1477                                 return NULL;
1478                         )
1479                         case '"':
1480                                 next_char();
1481                                 goto finished_headername;
1482                         }
1483                         obstack_1grow(&symbol_obstack, (char) input.c);
1484                         next_char();
1485                 }
1486                 /* we should never be here */
1487
1488         default:
1489                 /* TODO: do normal pp_token parsing and concatenate results */
1490                 panic("pp_token concat include not implemented yet");
1491         }
1492
1493 finished_headername:
1494         obstack_1grow(&symbol_obstack, '\0');
1495         char *headername = obstack_finish(&symbol_obstack);
1496
1497         /* TODO: iterate search-path to find the file */
1498
1499         skip_whitespace();
1500
1501         return identify_string(headername);
1502 }
1503
1504 static bool do_include(bool system_include, const char *headername)
1505 {
1506         if (!system_include) {
1507                 /* for "bla" includes first try current dir
1508                  * TODO: this isn't correct, should be the directory of the source file
1509                  */
1510                 FILE *file = fopen(headername, "r");
1511                 if (file != NULL) {
1512                         switch_input(file, headername);
1513                         return true;
1514                 }
1515         }
1516
1517         size_t headername_len = strlen(headername);
1518         assert(obstack_object_size(&pp_obstack) == 0);
1519         /* check searchpath */
1520         for (searchpath_entry_t *entry = searchpath; entry != NULL;
1521              entry = entry->next) {
1522             const char *path = entry->path;
1523             size_t      len  = strlen(path);
1524                 obstack_grow(&pp_obstack, path, len);
1525                 if (path[len-1] != '/')
1526                         obstack_1grow(&pp_obstack, '/');
1527                 obstack_grow(&pp_obstack, headername, headername_len+1);
1528
1529                 char *complete_path = obstack_finish(&pp_obstack);
1530                 FILE *file          = fopen(complete_path, "r");
1531                 if (file != NULL) {
1532                         const char *filename = identify_string(complete_path);
1533                         switch_input(file, filename);
1534                         return true;
1535                 }
1536                 obstack_free(&pp_obstack, complete_path);
1537         }
1538
1539         return false;
1540 }
1541
1542 static bool parse_include_directive(void)
1543 {
1544         /* don't eat the TP_include here!
1545          * we need an alternative parsing for the next token */
1546         skip_whitespace();
1547         bool system_include = input.c == '<';
1548         const char *headername = parse_headername();
1549         if (headername == NULL) {
1550                 eat_pp_directive();
1551                 return false;
1552         }
1553
1554         if (!info.at_line_begin) {
1555                 warningf(WARN_OTHER, &pp_token.base.source_position,
1556                          "extra tokens at end of #include directive");
1557                 eat_pp_directive();
1558         }
1559
1560         if (n_inputs > INCLUDE_LIMIT) {
1561                 errorf(&pp_token.base.source_position, "#include nested too deeply");
1562                 /* eat \n or EOF */
1563                 next_preprocessing_token();
1564                 return false;
1565         }
1566
1567         /* we have to reenable space counting and macro expansion here,
1568          * because it is still disabled in directive parsing,
1569          * but we will trigger a preprocessing token reading of the new file
1570          * now and need expansions/space counting */
1571         in_pp_directive = false;
1572
1573         /* switch inputs */
1574         emit_newlines();
1575         push_input();
1576         bool res = do_include(system_include, headername);
1577         if (!res) {
1578                 errorf(&pp_token.base.source_position,
1579                        "failed including '%s': %s", headername, strerror(errno));
1580                 pop_restore_input();
1581                 return false;
1582         }
1583
1584         return true;
1585 }
1586
1587 static pp_conditional_t *push_conditional(void)
1588 {
1589         pp_conditional_t *conditional
1590                 = obstack_alloc(&pp_obstack, sizeof(*conditional));
1591         memset(conditional, 0, sizeof(*conditional));
1592
1593         conditional->parent = conditional_stack;
1594         conditional_stack   = conditional;
1595
1596         return conditional;
1597 }
1598
1599 static void pop_conditional(void)
1600 {
1601         assert(conditional_stack != NULL);
1602         conditional_stack = conditional_stack->parent;
1603 }
1604
1605 static void check_unclosed_conditionals(void)
1606 {
1607         while (conditional_stack != NULL) {
1608                 pp_conditional_t *conditional = conditional_stack;
1609
1610                 if (conditional->in_else) {
1611                         errorf(&conditional->source_position, "unterminated #else");
1612                 } else {
1613                         errorf(&conditional->source_position, "unterminated condition");
1614                 }
1615                 pop_conditional();
1616         }
1617 }
1618
1619 static void parse_ifdef_ifndef_directive(void)
1620 {
1621         bool is_ifndef = (pp_token.kind == TP_ifndef);
1622         bool condition;
1623         next_preprocessing_token();
1624
1625         if (skip_mode) {
1626                 eat_pp_directive();
1627                 pp_conditional_t *conditional = push_conditional();
1628                 conditional->source_position  = pp_token.base.source_position;
1629                 conditional->skip             = true;
1630                 return;
1631         }
1632
1633         if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1634                 errorf(&pp_token.base.source_position,
1635                        "expected identifier after #%s, got '%t'",
1636                        is_ifndef ? "ifndef" : "ifdef", &pp_token);
1637                 eat_pp_directive();
1638
1639                 /* just take the true case in the hope to avoid further errors */
1640                 condition = true;
1641         } else {
1642                 /* evaluate wether we are in true or false case */
1643                 condition = !pp_token.base.symbol->pp_definition == is_ifndef;
1644
1645                 next_preprocessing_token();
1646
1647                 if (!info.at_line_begin) {
1648                         errorf(&pp_token.base.source_position,
1649                                "extra tokens at end of #%s",
1650                                is_ifndef ? "ifndef" : "ifdef");
1651                         eat_pp_directive();
1652                 }
1653         }
1654
1655         pp_conditional_t *conditional = push_conditional();
1656         conditional->source_position  = pp_token.base.source_position;
1657         conditional->condition        = condition;
1658
1659         if (!condition) {
1660                 skip_mode = true;
1661         }
1662 }
1663
1664 static void parse_else_directive(void)
1665 {
1666         eat_pp(TP_else);
1667
1668         if (!info.at_line_begin) {
1669                 if (!skip_mode) {
1670                         warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #else");
1671                 }
1672                 eat_pp_directive();
1673         }
1674
1675         pp_conditional_t *conditional = conditional_stack;
1676         if (conditional == NULL) {
1677                 errorf(&pp_token.base.source_position, "#else without prior #if");
1678                 return;
1679         }
1680
1681         if (conditional->in_else) {
1682                 errorf(&pp_token.base.source_position,
1683                        "#else after #else (condition started %P)",
1684                        conditional->source_position);
1685                 skip_mode = true;
1686                 return;
1687         }
1688
1689         conditional->in_else = true;
1690         if (!conditional->skip) {
1691                 skip_mode = conditional->condition;
1692         }
1693         conditional->source_position = pp_token.base.source_position;
1694 }
1695
1696 static void parse_endif_directive(void)
1697 {
1698         eat_pp(TP_endif);
1699
1700         if (!info.at_line_begin) {
1701                 if (!skip_mode) {
1702                         warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #endif");
1703                 }
1704                 eat_pp_directive();
1705         }
1706
1707         pp_conditional_t *conditional = conditional_stack;
1708         if (conditional == NULL) {
1709                 errorf(&pp_token.base.source_position, "#endif without prior #if");
1710                 return;
1711         }
1712
1713         if (!conditional->skip) {
1714                 skip_mode = false;
1715         }
1716         pop_conditional();
1717 }
1718
1719 static void parse_preprocessing_directive(void)
1720 {
1721         in_pp_directive = true;
1722         eat_pp('#');
1723
1724         if (skip_mode) {
1725                 switch (pp_token.kind) {
1726                 case TP_ifdef:
1727                 case TP_ifndef:
1728                         parse_ifdef_ifndef_directive();
1729                         break;
1730                 case TP_else:
1731                         parse_else_directive();
1732                         break;
1733                 case TP_endif:
1734                         parse_endif_directive();
1735                         break;
1736                 default:
1737                         eat_pp_directive();
1738                         break;
1739                 }
1740         } else {
1741                 switch (pp_token.kind) {
1742                 case TP_define:
1743                         parse_define_directive();
1744                         break;
1745                 case TP_undef:
1746                         parse_undef_directive();
1747                         break;
1748                 case TP_ifdef:
1749                 case TP_ifndef:
1750                         parse_ifdef_ifndef_directive();
1751                         break;
1752                 case TP_else:
1753                         parse_else_directive();
1754                         break;
1755                 case TP_endif:
1756                         parse_endif_directive();
1757                         break;
1758                 case TP_include:
1759                         parse_include_directive();
1760                         break;
1761                 default:
1762                         if (info.at_line_begin) {
1763                                 /* the nop directive "#" */
1764                                 break;
1765                         }
1766                         errorf(&pp_token.base.source_position,
1767                                    "invalid preprocessing directive #%t", &pp_token);
1768                         eat_pp_directive();
1769                         break;
1770                 }
1771         }
1772
1773         in_pp_directive = false;
1774         assert(info.at_line_begin);
1775 }
1776
1777 static void prepend_include_path(const char *path)
1778 {
1779         searchpath_entry_t *entry = OALLOCZ(&config_obstack, searchpath_entry_t);
1780         entry->path = path;
1781         entry->next = searchpath;
1782         searchpath  = entry;
1783 }
1784
1785 static void setup_include_path(void)
1786 {
1787         /* built-in paths */
1788         prepend_include_path("/usr/include");
1789
1790         /* parse environment variable */
1791         const char *cpath = getenv("CPATH");
1792         if (cpath != NULL && *cpath != '\0') {
1793                 const char *begin = cpath;
1794                 const char *c;
1795                 do {
1796                         c = begin;
1797                         while (*c != '\0' && *c != ':')
1798                                 ++c;
1799
1800                         size_t len = c-begin;
1801                         if (len == 0) {
1802                                 /* for gcc compatibility (Matze: I would expect that
1803                                  * nothing happens for an empty entry...) */
1804                                 prepend_include_path(".");
1805                         } else {
1806                                 char *string = obstack_alloc(&config_obstack, len+1);
1807                                 memcpy(string, begin, len);
1808                                 string[len] = '\0';
1809
1810                                 prepend_include_path(string);
1811                         }
1812
1813                         begin = c+1;
1814                         /* skip : */
1815                         if (*begin == ':')
1816                                 ++begin;
1817                 } while(*c != '\0');
1818         }
1819 }
1820
1821 int pptest_main(int argc, char **argv);
1822 int pptest_main(int argc, char **argv)
1823 {
1824         init_symbol_table();
1825         init_tokens();
1826
1827         obstack_init(&config_obstack);
1828         obstack_init(&pp_obstack);
1829         obstack_init(&input_obstack);
1830         strset_init(&stringset);
1831
1832         setup_include_path();
1833
1834         /* simplistic commandline parser */
1835         const char *filename = NULL;
1836         const char *output = NULL;
1837         for (int i = 1; i < argc; ++i) {
1838                 const char *opt = argv[i];
1839                 if (streq(opt, "-I")) {
1840                         prepend_include_path(argv[++i]);
1841                         continue;
1842                 } else if (streq(opt, "-E")) {
1843                         /* ignore */
1844                 } else if (streq(opt, "-o")) {
1845                         output = argv[++i];
1846                         continue;
1847                 } else if (opt[0] == '-') {
1848                         fprintf(stderr, "Unknown option '%s'\n", opt);
1849                 } else {
1850                         if (filename != NULL)
1851                                 fprintf(stderr, "Multiple inputs not supported\n");
1852                         filename = argv[i];
1853                 }
1854         }
1855         if (filename == NULL) {
1856                 fprintf(stderr, "No input specified\n");
1857                 return 1;
1858         }
1859
1860         if (output == NULL) {
1861                 out = stdout;
1862         } else {
1863                 out = fopen(output, "w");
1864                 if (out == NULL) {
1865                         fprintf(stderr, "Couldn't open output '%s'\n", output);
1866                         return 1;
1867                 }
1868         }
1869
1870         /* just here for gcc compatibility */
1871         fprintf(out, "# 1 \"%s\"\n", filename);
1872         fprintf(out, "# 1 \"<built-in>\"\n");
1873         fprintf(out, "# 1 \"<command-line>\"\n");
1874
1875         FILE *file = fopen(filename, "r");
1876         if (file == NULL) {
1877                 fprintf(stderr, "Couldn't open input '%s'\n", filename);
1878                 return 1;
1879         }
1880         switch_input(file, filename);
1881
1882         while (true) {
1883                 if (pp_token.kind == '#' && info.at_line_begin) {
1884                         parse_preprocessing_directive();
1885                         continue;
1886                 } else if (pp_token.kind == TP_EOF) {
1887                         goto end_of_main_loop;
1888                 } else if (pp_token.kind == TP_IDENTIFIER && !in_pp_directive) {
1889                         symbol_t        *const symbol        = pp_token.base.symbol;
1890                         pp_definition_t *const pp_definition = symbol->pp_definition;
1891                         if (pp_definition != NULL && !pp_definition->is_expanding) {
1892                                 expansion_pos = pp_token.base.source_position;
1893                                 if (pp_definition->has_parameters) {
1894                                         source_position_t position = pp_token.base.source_position;
1895                                         add_token_info_t old_info = info;
1896                                         next_preprocessing_token();
1897                                         add_token_info_t new_info = info;
1898
1899                                         /* no opening brace -> no expansion */
1900                                         if (pp_token.kind == '(') {
1901                                                 eat_pp('(');
1902
1903                                                 /* parse arguments (TODO) */
1904                                                 while (pp_token.kind != TP_EOF && pp_token.kind != ')')
1905                                                         next_preprocessing_token();
1906                                         } else {
1907                                                 token_t next_token = pp_token;
1908                                                 /* restore identifier token */
1909                                                 pp_token.kind                 = TP_IDENTIFIER;
1910                                                 pp_token.base.symbol          = symbol;
1911                                                 pp_token.base.source_position = position;
1912                                                 info = old_info;
1913                                                 emit_pp_token();
1914
1915                                                 info = new_info;
1916                                                 pp_token = next_token;
1917                                                 continue;
1918                                         }
1919                                         info = old_info;
1920                                 }
1921                                 pp_definition->expand_pos   = 0;
1922                                 pp_definition->is_expanding = true;
1923                                 current_expansion           = pp_definition;
1924                                 expand_next();
1925                                 continue;
1926                         }
1927                 }
1928
1929                 emit_pp_token();
1930                 next_preprocessing_token();
1931         }
1932 end_of_main_loop:
1933
1934         fputc('\n', out);
1935         check_unclosed_conditionals();
1936         close_input();
1937         if (out != stdout)
1938                 fclose(out);
1939
1940         obstack_free(&input_obstack, NULL);
1941         obstack_free(&pp_obstack, NULL);
1942         obstack_free(&config_obstack, NULL);
1943
1944         strset_destroy(&stringset);
1945
1946         exit_tokens();
1947         exit_symbol_table();
1948
1949         return 0;
1950 }