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              skip_mode;
  95 static FILE             *out;
  96 static struct obstack    pp_obstack;
  97 static struct obstack    config_obstack;
  98 static const char       *printed_input_name = NULL;
  99 static source_position_t expansion_pos;
 100 static pp_definition_t  *current_expansion  = NULL;
 101 static strset_t          stringset;
 102 static preprocessor_token_kind_t last_token;
 103
 104 static searchpath_entry_t *searchpath;
 105
 106 static add_token_info_t  info;
 107
 108 static inline void next_char(void);
 109 static void next_preprocessing_token(void);
 110 static void print_line_directive(const source_position_t *pos, const char *add);
 111
 112 static void switch_input(FILE *file, const char *filename)
 113 {
 114         input.file                = file;
 115         input.input               = input_from_stream(file, NULL);
 116         input.bufend              = NULL;
 117         input.bufpos              = NULL;
 118         input.output_line         = 0;
 119         input.position.input_name = filename;
 120         input.position.lineno     = 1;
 121
 122         /* indicate that we're at a new input */
 123         print_line_directive(&input.position, input_stack != NULL ? "1" : NULL);
 124
 125         /* place a virtual '\n' so we realize we're at line begin */
 126         input.position.lineno = 0;
 127         input.c               = '\n';
 128         next_preprocessing_token();
 129 }
 130
 131 static void close_input(void)
 132 {
 133         input_free(input.input);
 134         assert(input.file != NULL);
 135
 136         fclose(input.file);
 137         input.input  = NULL;
 138         input.file   = NULL;
 139         input.bufend = NULL;
 140         input.bufpos = NULL;
 141         input.c      = EOF;
 142 }
 143
 144 static void push_input(void)
 145 {
 146         pp_input_t *saved_input
 147                 = obstack_alloc(&input_obstack, sizeof(*saved_input));
 148
 149         memcpy(saved_input, &input, sizeof(*saved_input));
 150
 151         /* adjust buffer positions */
 152         if (input.bufpos != NULL)
 153                 saved_input->bufpos = saved_input->buf + (input.bufpos - input.buf);
 154         if (input.bufend != NULL)
 155                 saved_input->bufend = saved_input->buf + (input.bufend - input.buf);
 156
 157         saved_input->parent = input_stack;
 158         input_stack         = saved_input;
 159         ++n_inputs;
 160 }
 161
 162 static void pop_restore_input(void)
 163 {
 164         assert(n_inputs > 0);
 165         assert(input_stack != NULL);
 166
 167         pp_input_t *saved_input = input_stack;
 168
 169         memcpy(&input, saved_input, sizeof(input));
 170         input.parent = NULL;
 171
 172         /* adjust buffer positions */
 173         if (saved_input->bufpos != NULL)
 174                 input.bufpos = input.buf + (saved_input->bufpos - saved_input->buf);
 175         if (saved_input->bufend != NULL)
 176                 input.bufend = input.buf + (saved_input->bufend - saved_input->buf);
 177
 178         input_stack = saved_input->parent;
 179         obstack_free(&input_obstack, saved_input);
 180         --n_inputs;
 181 }
 182
 183 /**
 184  * Prints a parse error message at the current token.
 185  *
 186  * @param msg   the error message
 187  */
 188 static void parse_error(const char *msg)
 189 {
 190         errorf(&pp_token.base.source_position,  "%s", msg);
 191 }
 192
 193 static inline void next_real_char(void)
 194 {
 195         assert(input.bufpos <= input.bufend);
 196         if (input.bufpos >= input.bufend) {
 197                 size_t const n = decode(input.input, input.buf + MAX_PUTBACK, lengthof(input.buf) - MAX_PUTBACK);
 198                 if (n == 0) {
 199                         input.c = EOF;
 200                         return;
 201                 }
 202                 input.bufpos = input.buf + MAX_PUTBACK;
 203                 input.bufend = input.bufpos + n;
 204         }
 205         input.c = *input.bufpos++;
 206         ++input.position.colno;
 207 }
 208
 209 /**
 210  * Put a character back into the buffer.
 211  *
 212  * @param pc  the character to put back
 213  */
 214 static inline void put_back(utf32 const pc)
 215 {
 216         assert(input.bufpos > input.buf);
 217         *(--input.bufpos - input.buf + input.buf) = (char) pc;
 218         --input.position.colno;
 219 }
 220
 221 #define MATCH_NEWLINE(code)                   \
 222         case '\r':                                \
 223                 next_char();                          \
 224                 if (input.c == '\n') {                \
 225         case '\n':                                \
 226                         next_char();                      \
 227                 }                                     \
 228                 info.whitespace = 0;                  \
 229                 ++input.position.lineno;              \
 230                 input.position.colno = 1;             \
 231                 code
 232
 233 #define eat(c_type) (assert(input.c == c_type), next_char())
 234
 235 static void maybe_concat_lines(void)
 236 {
 237         eat('\\');
 238
 239         switch (input.c) {
 240         MATCH_NEWLINE(
 241                 return;
 242         )
 243
 244         default:
 245                 break;
 246         }
 247
 248         put_back(input.c);
 249         input.c = '\\';
 250 }
 251
 252 /**
 253  * Set c to the next input character, ie.
 254  * after expanding trigraphs.
 255  */
 256 static inline void next_char(void)
 257 {
 258         next_real_char();
 259
 260         /* filter trigraphs and concatenated lines */
 261         if (UNLIKELY(input.c == '\\')) {
 262                 maybe_concat_lines();
 263                 goto end_of_next_char;
 264         }
 265
 266         if (LIKELY(input.c != '?'))
 267                 goto end_of_next_char;
 268
 269         next_real_char();
 270         if (LIKELY(input.c != '?')) {
 271                 put_back(input.c);
 272                 input.c = '?';
 273                 goto end_of_next_char;
 274         }
 275
 276         next_real_char();
 277         switch (input.c) {
 278         case '=': input.c = '#'; break;
 279         case '(': input.c = '['; break;
 280         case '/': input.c = '\\'; maybe_concat_lines(); break;
 281         case ')': input.c = ']'; 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         default:
 288                 put_back(input.c);
 289                 put_back('?');
 290                 input.c = '?';
 291                 break;
 292         }
 293
 294 end_of_next_char:;
 295 #ifdef DEBUG_CHARS
 296         printf("nchar '%c'\n", input.c);
 297 #endif
 298 }
 299
 300
 301
 302 /**
 303  * Returns true if the given char is a octal digit.
 304  *
 305  * @param char  the character to check
 306  */
 307 static inline bool is_octal_digit(int chr)
 308 {
 309         switch (chr) {
 310         case '0':
 311         case '1':
 312         case '2':
 313         case '3':
 314         case '4':
 315         case '5':
 316         case '6':
 317         case '7':
 318                 return true;
 319         default:
 320                 return false;
 321         }
 322 }
 323
 324 /**
 325  * Returns the value of a digit.
 326  * The only portable way to do it ...
 327  */
 328 static int digit_value(int digit)
 329 {
 330         switch (digit) {
 331         case '0': return 0;
 332         case '1': return 1;
 333         case '2': return 2;
 334         case '3': return 3;
 335         case '4': return 4;
 336         case '5': return 5;
 337         case '6': return 6;
 338         case '7': return 7;
 339         case '8': return 8;
 340         case '9': return 9;
 341         case 'a':
 342         case 'A': return 10;
 343         case 'b':
 344         case 'B': return 11;
 345         case 'c':
 346         case 'C': return 12;
 347         case 'd':
 348         case 'D': return 13;
 349         case 'e':
 350         case 'E': return 14;
 351         case 'f':
 352         case 'F': return 15;
 353         default:
 354                 panic("wrong character given");
 355         }
 356 }
 357
 358 /**
 359  * Parses an octal character sequence.
 360  *
 361  * @param first_digit  the already read first digit
 362  */
 363 static utf32 parse_octal_sequence(const utf32 first_digit)
 364 {
 365         assert(is_octal_digit(first_digit));
 366         utf32 value = digit_value(first_digit);
 367         if (!is_octal_digit(input.c)) return value;
 368         value = 8 * value + digit_value(input.c);
 369         next_char();
 370         if (!is_octal_digit(input.c)) return value;
 371         value = 8 * value + digit_value(input.c);
 372         next_char();
 373         return value;
 374
 375 }
 376
 377 /**
 378  * Parses a hex character sequence.
 379  */
 380 static utf32 parse_hex_sequence(void)
 381 {
 382         utf32 value = 0;
 383         while (isxdigit(input.c)) {
 384                 value = 16 * value + digit_value(input.c);
 385                 next_char();
 386         }
 387         return value;
 388 }
 389
 390 /**
 391  * Parse an escape sequence.
 392  */
 393 static utf32 parse_escape_sequence(void)
 394 {
 395         eat('\\');
 396
 397         utf32 const ec = input.c;
 398         next_char();
 399
 400         switch (ec) {
 401         case '"':  return '"';
 402         case '\'': return '\'';
 403         case '\\': return '\\';
 404         case '?': return '\?';
 405         case 'a': return '\a';
 406         case 'b': return '\b';
 407         case 'f': return '\f';
 408         case 'n': return '\n';
 409         case 'r': return '\r';
 410         case 't': return '\t';
 411         case 'v': return '\v';
 412         case 'x':
 413                 return parse_hex_sequence();
 414         case '0':
 415         case '1':
 416         case '2':
 417         case '3':
 418         case '4':
 419         case '5':
 420         case '6':
 421         case '7':
 422                 return parse_octal_sequence(ec);
 423         case EOF:
 424                 parse_error("reached end of file while parsing escape sequence");
 425                 return EOF;
 426         /* \E is not documented, but handled, by GCC.  It is acceptable according
 427          * to §6.11.4, whereas \e is not. */
 428         case 'E':
 429         case 'e':
 430                 if (c_mode & _GNUC)
 431                         return 27;   /* hopefully 27 is ALWAYS the code for ESCAPE */
 432                 break;
 433         case 'u':
 434         case 'U':
 435                 parse_error("universal character parsing not implemented yet");
 436                 return EOF;
 437         default:
 438                 break;
 439         }
 440         /* §6.4.4.4:8 footnote 64 */
 441         parse_error("unknown escape sequence");
 442         return EOF;
 443 }
 444
 445 static const char *identify_string(char *string)
 446 {
 447         const char *result = strset_insert(&stringset, string);
 448         if (result != string) {
 449                 obstack_free(&symbol_obstack, string);
 450         }
 451         return result;
 452 }
 453
 454 static string_t make_string(char *string, size_t len)
 455 {
 456         const char *result = identify_string(string);
 457         return (string_t) {result, len};
 458 }
 459
 460 static void parse_string_literal(string_encoding_t const enc)
 461 {
 462         const unsigned start_linenr = input.position.lineno;
 463
 464         eat('"');
 465
 466         while (true) {
 467                 switch (input.c) {
 468                 case '\\': {
 469                         utf32 tc;
 470                         if (resolve_escape_sequences) {
 471                                 tc = parse_escape_sequence();
 472                                 obstack_1grow(&symbol_obstack, (char) tc);
 473                         } else {
 474                                 obstack_1grow(&symbol_obstack, (char) input.c);
 475                                 next_char();
 476                                 obstack_1grow(&symbol_obstack, (char) input.c);
 477                                 next_char();
 478                         }
 479                         break;
 480                 }
 481
 482                 case EOF: {
 483                         source_position_t source_position;
 484                         source_position.input_name = pp_token.base.source_position.input_name;
 485                         source_position.lineno     = start_linenr;
 486                         errorf(&source_position, "string has no end");
 487                         goto end_of_string;
 488                 }
 489
 490                 case '"':
 491                         next_char();
 492                         goto end_of_string;
 493
 494                 default:
 495                         obstack_grow_symbol(&symbol_obstack, input.c);
 496                         next_char();
 497                         break;
 498                 }
 499         }
 500
 501 end_of_string:
 502         /* add finishing 0 to the string */
 503         obstack_1grow(&symbol_obstack, '\0');
 504         const size_t size   = (size_t)obstack_object_size(&symbol_obstack);
 505         char *const  string = obstack_finish(&symbol_obstack);
 506
 507         pp_token.kind            = TP_STRING_LITERAL;
 508         pp_token.string.encoding = enc;
 509         pp_token.string.string   = make_string(string, size);
 510 }
 511
 512 static void parse_wide_character_constant(void)
 513 {
 514         eat('\'');
 515
 516         while (true) {
 517                 switch (input.c) {
 518                 case '\\': {
 519                         const utf32 tc = parse_escape_sequence();
 520                         obstack_grow_symbol(&symbol_obstack, tc);
 521                         break;
 522                 }
 523
 524                 MATCH_NEWLINE(
 525                         parse_error("newline while parsing character constant");
 526                         break;
 527                 )
 528
 529                 case '\'':
 530                         next_char();
 531                         goto end_of_wide_char_constant;
 532
 533                 case EOF:
 534                         parse_error("EOF while parsing character constant");
 535                         goto end_of_wide_char_constant;
 536
 537                 default:
 538                         obstack_grow_symbol(&symbol_obstack, input.c);
 539                         next_char();
 540                         break;
 541                 }
 542         }
 543
 544 end_of_wide_char_constant:
 545         obstack_1grow(&symbol_obstack, '\0');
 546         size_t  size = (size_t) obstack_object_size(&symbol_obstack)-1;
 547         char   *string = obstack_finish(&symbol_obstack);
 548         pp_token.kind          = TP_WIDE_CHARACTER_CONSTANT;
 549         pp_token.string.string = make_string(string, size);
 550
 551         if (size == 0) {
 552                 parse_error("empty character constant");
 553         }
 554 }
 555
 556 static void parse_character_constant(void)
 557 {
 558         const unsigned start_linenr = input.position.lineno;
 559
 560         eat('\'');
 561
 562         int tc;
 563         while (true) {
 564                 switch (input.c) {
 565                 case '\\':
 566                         tc = parse_escape_sequence();
 567                         obstack_1grow(&symbol_obstack, (char) tc);
 568                         break;
 569
 570                 MATCH_NEWLINE(
 571                         parse_error("newline while parsing character constant");
 572                         break;
 573                 )
 574
 575                 case EOF: {
 576                         source_position_t source_position;
 577                         source_position.input_name = pp_token.base.source_position.input_name;
 578                         source_position.lineno     = start_linenr;
 579                         errorf(&source_position, "EOF while parsing character constant");
 580                         goto end_of_char_constant;
 581                 }
 582
 583                 case '\'':
 584                         next_char();
 585                         goto end_of_char_constant;
 586
 587                 default:
 588                         obstack_1grow(&symbol_obstack, (char) input.c);
 589                         next_char();
 590                         break;
 591
 592                 }
 593         }
 594
 595 end_of_char_constant:;
 596         obstack_1grow(&symbol_obstack, '\0');
 597         const size_t size   = (size_t)obstack_object_size(&symbol_obstack);
 598         char *const  string = obstack_finish(&symbol_obstack);
 599
 600         pp_token.kind          = TP_CHARACTER_CONSTANT;
 601         pp_token.string.string = make_string(string, size);
 602
 603         if (size == 0) {
 604                 parse_error("empty character constant");
 605         }
 606 }
 607
 608 #define SYMBOL_CHARS_WITHOUT_E_P \
 609         case 'a': \
 610         case 'b': \
 611         case 'c': \
 612         case 'd': \
 613         case 'f': \
 614         case 'g': \
 615         case 'h': \
 616         case 'i': \
 617         case 'j': \
 618         case 'k': \
 619         case 'l': \
 620         case 'm': \
 621         case 'n': \
 622         case 'o': \
 623         case 'q': \
 624         case 'r': \
 625         case 's': \
 626         case 't': \
 627         case 'u': \
 628         case 'v': \
 629         case 'w': \
 630         case 'x': \
 631         case 'y': \
 632         case 'z': \
 633         case 'A': \
 634         case 'B': \
 635         case 'C': \
 636         case 'D': \
 637         case 'F': \
 638         case 'G': \
 639         case 'H': \
 640         case 'I': \
 641         case 'J': \
 642         case 'K': \
 643         case 'L': \
 644         case 'M': \
 645         case 'N': \
 646         case 'O': \
 647         case 'Q': \
 648         case 'R': \
 649         case 'S': \
 650         case 'T': \
 651         case 'U': \
 652         case 'V': \
 653         case 'W': \
 654         case 'X': \
 655         case 'Y': \
 656         case 'Z': \
 657         case '_':
 658
 659 #define SYMBOL_CHARS \
 660         SYMBOL_CHARS_WITHOUT_E_P \
 661         case 'e': \
 662         case 'p': \
 663         case 'E': \
 664         case 'P':
 665
 666 #define DIGITS \
 667         case '0':  \
 668         case '1':  \
 669         case '2':  \
 670         case '3':  \
 671         case '4':  \
 672         case '5':  \
 673         case '6':  \
 674         case '7':  \
 675         case '8':  \
 676         case '9':
 677
 678 /**
 679  * returns next final token from a preprocessor macro expansion
 680  */
 681 static void expand_next(void)
 682 {
 683         assert(current_expansion != NULL);
 684
 685         pp_definition_t *definition = current_expansion;
 686
 687 restart:
 688         if (definition->list_len == 0
 689                         || definition->expand_pos >= definition->list_len) {
 690                 /* we're finished with the current macro, move up 1 level in the
 691                  * expansion stack */
 692                 pp_definition_t *parent = definition->parent_expansion;
 693                 definition->parent_expansion = NULL;
 694                 definition->is_expanding     = false;
 695
 696                 /* it was the outermost expansion, parse normal pptoken */
 697                 if (parent == NULL) {
 698                         current_expansion = NULL;
 699                         next_preprocessing_token();
 700                         return;
 701                 }
 702                 definition        = parent;
 703                 current_expansion = definition;
 704                 goto restart;
 705         }
 706         pp_token = definition->token_list[definition->expand_pos];
 707         pp_token.base.source_position = expansion_pos;
 708         ++definition->expand_pos;
 709
 710         if (pp_token.kind != TP_IDENTIFIER)
 711                 return;
 712
 713         /* if it was an identifier then we might need to expand again */
 714         pp_definition_t *const symbol_definition = pp_token.base.symbol->pp_definition;
 715         if (symbol_definition != NULL && !symbol_definition->is_expanding) {
 716                 symbol_definition->parent_expansion = definition;
 717                 symbol_definition->expand_pos       = 0;
 718                 symbol_definition->is_expanding     = true;
 719                 definition                          = symbol_definition;
 720                 current_expansion                   = definition;
 721                 goto restart;
 722         }
 723 }
 724
 725 static void skip_line_comment(void)
 726 {
 727         while (true) {
 728                 switch (input.c) {
 729                 case EOF:
 730                         return;
 731
 732                 case '\r':
 733                 case '\n':
 734                         return;
 735
 736                 default:
 737                         next_char();
 738                         break;
 739                 }
 740         }
 741 }
 742
 743 static void skip_multiline_comment(void)
 744 {
 745         unsigned start_linenr = input.position.lineno;
 746         while (true) {
 747                 switch (input.c) {
 748                 case '/':
 749                         next_char();
 750                         if (input.c == '*') {
 751                                 /* TODO: nested comment, warn here */
 752                         }
 753                         break;
 754                 case '*':
 755                         next_char();
 756                         if (input.c == '/') {
 757                                 if (input.position.lineno != input.output_line)
 758                                         info.whitespace = input.position.colno;
 759                                 next_char();
 760                                 return;
 761                         }
 762                         break;
 763
 764                 MATCH_NEWLINE(
 765                         break;
 766                 )
 767
 768                 case EOF: {
 769                         source_position_t source_position;
 770                         source_position.input_name = pp_token.base.source_position.input_name;
 771                         source_position.lineno     = start_linenr;
 772                         errorf(&source_position, "at end of file while looking for comment end");
 773                         return;
 774                 }
 775
 776                 default:
 777                         next_char();
 778                         break;
 779                 }
 780         }
 781 }
 782
 783 static void skip_whitespace(void)
 784 {
 785         while (true) {
 786                 switch (input.c) {
 787                 case ' ':
 788                 case '\t':
 789                         next_char();
 790                         continue;
 791
 792                 MATCH_NEWLINE(
 793                         info.at_line_begin = true;
 794                         return;
 795                 )
 796
 797                 case '/':
 798                         next_char();
 799                         if (input.c == '/') {
 800                                 next_char();
 801                                 skip_line_comment();
 802                                 continue;
 803                         } else if (input.c == '*') {
 804                                 next_char();
 805                                 skip_multiline_comment();
 806                                 continue;
 807                         } else {
 808                                 put_back(input.c);
 809                                 input.c = '/';
 810                         }
 811                         return;
 812                 default:
 813                         return;
 814                 }
 815         }
 816 }
 817
 818 static void eat_pp(preprocessor_token_kind_t const type)
 819 {
 820         (void) type;
 821         assert(pp_token.kind == type);
 822         next_preprocessing_token();
 823 }
 824
 825 static void parse_symbol(void)
 826 {
 827         obstack_1grow(&symbol_obstack, (char) input.c);
 828         next_char();
 829
 830         while (true) {
 831                 switch (input.c) {
 832                 DIGITS
 833                 SYMBOL_CHARS
 834                         obstack_1grow(&symbol_obstack, (char) input.c);
 835                         next_char();
 836                         break;
 837
 838                 default:
 839                         goto end_symbol;
 840                 }
 841         }
 842
 843 end_symbol:
 844         obstack_1grow(&symbol_obstack, '\0');
 845         char *string = obstack_finish(&symbol_obstack);
 846
 847         /* might be a wide string or character constant ( L"string"/L'c' ) */
 848         if (input.c == '"' && string[0] == 'L' && string[1] == '\0') {
 849                 obstack_free(&symbol_obstack, string);
 850                 parse_string_literal(STRING_ENCODING_WIDE);
 851                 return;
 852         } else if (input.c == '\'' && string[0] == 'L' && string[1] == '\0') {
 853                 obstack_free(&symbol_obstack, string);
 854                 parse_wide_character_constant();
 855                 return;
 856         }
 857
 858         symbol_t *symbol = symbol_table_insert(string);
 859
 860         pp_token.kind        = symbol->pp_ID;
 861         pp_token.base.symbol = symbol;
 862
 863         /* we can free the memory from symbol obstack if we already had an entry in
 864          * the symbol table */
 865         if (symbol->string != string) {
 866                 obstack_free(&symbol_obstack, string);
 867         }
 868 }
 869
 870 static void parse_number(void)
 871 {
 872         obstack_1grow(&symbol_obstack, (char) input.c);
 873         next_char();
 874
 875         while (true) {
 876                 switch (input.c) {
 877                 case '.':
 878                 DIGITS
 879                 SYMBOL_CHARS_WITHOUT_E_P
 880                         obstack_1grow(&symbol_obstack, (char) input.c);
 881                         next_char();
 882                         break;
 883
 884                 case 'e':
 885                 case 'p':
 886                 case 'E':
 887                 case 'P':
 888                         obstack_1grow(&symbol_obstack, (char) input.c);
 889                         next_char();
 890                         if (input.c == '+' || input.c == '-') {
 891                                 obstack_1grow(&symbol_obstack, (char) input.c);
 892                                 next_char();
 893                         }
 894                         break;
 895
 896                 default:
 897                         goto end_number;
 898                 }
 899         }
 900
 901 end_number:
 902         obstack_1grow(&symbol_obstack, '\0');
 903         size_t  size   = obstack_object_size(&symbol_obstack);
 904         char   *string = obstack_finish(&symbol_obstack);
 905
 906         pp_token.kind          = TP_NUMBER;
 907         pp_token.number.number = make_string(string, size);
 908 }
 909
 910
 911 #define MAYBE_PROLOG                                       \
 912                         next_char();                                   \
 913                         while (true) {                                 \
 914                                 switch (input.c) {
 915
 916 #define MAYBE(ch, set_type)                                \
 917                                 case ch:                                   \
 918                                         next_char();                           \
 919                                         pp_token.kind = set_type;              \
 920                                         return;
 921
 922 #define ELSE_CODE(code)                                    \
 923                                 default:                                   \
 924                                         code                                   \
 925                                         return;                                \
 926                                 }                                          \
 927                         }
 928
 929 #define ELSE(set_type)                                     \
 930                 ELSE_CODE(                                         \
 931                         pp_token.kind = set_type;                      \
 932                 )
 933
 934 static void next_preprocessing_token(void)
 935 {
 936         if (current_expansion != NULL) {
 937                 expand_next();
 938                 return;
 939         }
 940
 941         info.at_line_begin  = false;
 942         info.had_whitespace = false;
 943 restart:
 944         pp_token.base.source_position = input.position;
 945         pp_token.base.symbol          = NULL;
 946
 947         switch (input.c) {
 948         case ' ':
 949         case '\t':
 950                 ++info.whitespace;
 951                 info.had_whitespace = true;
 952                 next_char();
 953                 goto restart;
 954
 955         MATCH_NEWLINE(
 956                 info.at_line_begin = true;
 957                 info.had_whitespace = true;
 958                 goto restart;
 959         )
 960
 961         SYMBOL_CHARS
 962                 parse_symbol();
 963                 return;
 964
 965         DIGITS
 966                 parse_number();
 967                 return;
 968
 969         case '"':
 970                 parse_string_literal(STRING_ENCODING_CHAR);
 971                 return;
 972
 973         case '\'':
 974                 parse_character_constant();
 975                 return;
 976
 977         case '.':
 978                 MAYBE_PROLOG
 979                         case '0':
 980                         case '1':
 981                         case '2':
 982                         case '3':
 983                         case '4':
 984                         case '5':
 985                         case '6':
 986                         case '7':
 987                         case '8':
 988                         case '9':
 989                                 put_back(input.c);
 990                                 input.c = '.';
 991                                 parse_number();
 992                                 return;
 993
 994                         case '.':
 995                                 MAYBE_PROLOG
 996                                 MAYBE('.', TP_DOTDOTDOT)
 997                                 ELSE_CODE(
 998                                         put_back(input.c);
 999                                         input.c = '.';
1000                                         pp_token.kind = '.';
1001                                 )
1002                 ELSE('.')
1003         case '&':
1004                 MAYBE_PROLOG
1005                 MAYBE('&', TP_ANDAND)
1006                 MAYBE('=', TP_ANDEQUAL)
1007                 ELSE('&')
1008         case '*':
1009                 MAYBE_PROLOG
1010                 MAYBE('=', TP_ASTERISKEQUAL)
1011                 ELSE('*')
1012         case '+':
1013                 MAYBE_PROLOG
1014                 MAYBE('+', TP_PLUSPLUS)
1015                 MAYBE('=', TP_PLUSEQUAL)
1016                 ELSE('+')
1017         case '-':
1018                 MAYBE_PROLOG
1019                 MAYBE('>', TP_MINUSGREATER)
1020                 MAYBE('-', TP_MINUSMINUS)
1021                 MAYBE('=', TP_MINUSEQUAL)
1022                 ELSE('-')
1023         case '!':
1024                 MAYBE_PROLOG
1025                 MAYBE('=', TP_EXCLAMATIONMARKEQUAL)
1026                 ELSE('!')
1027         case '/':
1028                 MAYBE_PROLOG
1029                 MAYBE('=', TP_SLASHEQUAL)
1030                         case '*':
1031                                 next_char();
1032                                 info.had_whitespace = true;
1033                                 skip_multiline_comment();
1034                                 goto restart;
1035                         case '/':
1036                                 next_char();
1037                                 info.had_whitespace = true;
1038                                 skip_line_comment();
1039                                 goto restart;
1040                 ELSE('/')
1041         case '%':
1042                 MAYBE_PROLOG
1043                 MAYBE('>', '}')
1044                 MAYBE('=', TP_PERCENTEQUAL)
1045                         case ':':
1046                                 MAYBE_PROLOG
1047                                         case '%':
1048                                                 MAYBE_PROLOG
1049                                                 MAYBE(':', TP_HASHHASH)
1050                                                 ELSE_CODE(
1051                                                         put_back(input.c);
1052                                                         input.c = '%';
1053                                                         pp_token.kind = '#';
1054                                                 )
1055                                 ELSE('#')
1056                 ELSE('%')
1057         case '<':
1058                 MAYBE_PROLOG
1059                 MAYBE(':', '[')
1060                 MAYBE('%', '{')
1061                 MAYBE('=', TP_LESSEQUAL)
1062                         case '<':
1063                                 MAYBE_PROLOG
1064                                 MAYBE('=', TP_LESSLESSEQUAL)
1065                                 ELSE(TP_LESSLESS)
1066                 ELSE('<')
1067         case '>':
1068                 MAYBE_PROLOG
1069                 MAYBE('=', TP_GREATEREQUAL)
1070                         case '>':
1071                                 MAYBE_PROLOG
1072                                 MAYBE('=', TP_GREATERGREATEREQUAL)
1073                                 ELSE(TP_GREATERGREATER)
1074                 ELSE('>')
1075         case '^':
1076                 MAYBE_PROLOG
1077                 MAYBE('=', TP_CARETEQUAL)
1078                 ELSE('^')
1079         case '|':
1080                 MAYBE_PROLOG
1081                 MAYBE('=', TP_PIPEEQUAL)
1082                 MAYBE('|', TP_PIPEPIPE)
1083                 ELSE('|')
1084         case ':':
1085                 MAYBE_PROLOG
1086                 MAYBE('>', ']')
1087                 ELSE(':')
1088         case '=':
1089                 MAYBE_PROLOG
1090                 MAYBE('=', TP_EQUALEQUAL)
1091                 ELSE('=')
1092         case '#':
1093                 MAYBE_PROLOG
1094                 MAYBE('#', TP_HASHHASH)
1095                 ELSE_CODE(
1096                         pp_token.kind = '#';
1097                 )
1098
1099         case '?':
1100         case '[':
1101         case ']':
1102         case '(':
1103         case ')':
1104         case '{':
1105         case '}':
1106         case '~':
1107         case ';':
1108         case ',':
1109         case '\\':
1110                 pp_token.kind = input.c;
1111                 next_char();
1112                 return;
1113
1114         case EOF:
1115                 if (input_stack != NULL) {
1116                         close_input();
1117                         pop_restore_input();
1118                         fputc('\n', out);
1119                         print_line_directive(&input.position, "2");
1120                         goto restart;
1121                 } else {
1122                         pp_token.base.source_position.lineno++;
1123                         info.at_line_begin = true;
1124                         pp_token.kind = TP_EOF;
1125                 }
1126                 return;
1127
1128         default:
1129                 next_char();
1130                 if (!ignore_unknown_chars) {
1131                         errorf(&pp_token.base.source_position,
1132                                "unknown character '%c' found\n", input.c);
1133                         goto restart;
1134                 } else {
1135                         pp_token.kind = input.c;
1136                         return;
1137                 }
1138         }
1139 }
1140
1141 static void print_quoted_string(const char *const string)
1142 {
1143         fputc('"', out);
1144         for (const char *c = string; *c != 0; ++c) {
1145                 switch (*c) {
1146                 case '"': fputs("\\\"", out); break;
1147                 case '\\':  fputs("\\\\", out); break;
1148                 case '\a':  fputs("\\a", out); break;
1149                 case '\b':  fputs("\\b", out); break;
1150                 case '\f':  fputs("\\f", out); break;
1151                 case '\n':  fputs("\\n", out); break;
1152                 case '\r':  fputs("\\r", out); break;
1153                 case '\t':  fputs("\\t", out); break;
1154                 case '\v':  fputs("\\v", out); break;
1155                 case '\?':  fputs("\\?", out); break;
1156                 default:
1157                         if (!isprint(*c)) {
1158                                 fprintf(out, "\\%03o", (unsigned)*c);
1159                                 break;
1160                         }
1161                         fputc(*c, out);
1162                         break;
1163                 }
1164         }
1165         fputc('"', out);
1166 }
1167
1168 static void print_line_directive(const source_position_t *pos, const char *add)
1169 {
1170         fprintf(out, "# %u ", pos->lineno);
1171         print_quoted_string(pos->input_name);
1172         if (add != NULL) {
1173                 fputc(' ', out);
1174                 fputs(add, out);
1175         }
1176
1177         printed_input_name = pos->input_name;
1178         input.output_line  = pos->lineno-1;
1179 }
1180
1181 static bool emit_newlines(void)
1182 {
1183         unsigned delta = pp_token.base.source_position.lineno - input.output_line;
1184         if (delta == 0)
1185                 return false;
1186
1187         if (delta >= 9) {
1188                 fputc('\n', out);
1189                 print_line_directive(&pp_token.base.source_position, NULL);
1190                 fputc('\n', out);
1191         } else {
1192                 for (unsigned i = 0; i < delta; ++i) {
1193                         fputc('\n', out);
1194                 }
1195         }
1196         input.output_line = pp_token.base.source_position.lineno;
1197
1198         for (unsigned i = 0; i < info.whitespace; ++i)
1199                 fputc(' ', out);
1200
1201         return true;
1202 }
1203
1204 static void emit_pp_token(void)
1205 {
1206         if (skip_mode)
1207                 return;
1208
1209         if (!emit_newlines() &&
1210             (info.had_whitespace || tokens_would_paste(last_token, pp_token.kind)))
1211                 fputc(' ', out);
1212
1213         switch (pp_token.kind) {
1214         case TP_IDENTIFIER:
1215                 fputs(pp_token.base.symbol->string, out);
1216                 break;
1217         case TP_NUMBER:
1218                 fputs(pp_token.number.number.begin, out);
1219                 break;
1220
1221         case TP_STRING_LITERAL:
1222                 fputs(get_string_encoding_prefix(pp_token.string.encoding), out);
1223                 fputc('"', out);
1224                 fputs(pp_token.string.string.begin, out);
1225                 fputc('"', out);
1226                 break;
1227         case TP_WIDE_CHARACTER_CONSTANT:
1228                 fputc('L', out);
1229         case TP_CHARACTER_CONSTANT:
1230                 fputc('\'', out);
1231                 fputs(pp_token.string.string.begin, out);
1232                 fputc('\'', out);
1233                 break;
1234         default:
1235                 print_pp_token_kind(out, pp_token.kind);
1236                 break;
1237         }
1238         last_token = pp_token.kind;
1239 }
1240
1241 static void eat_pp_directive(void)
1242 {
1243         while (!info.at_line_begin) {
1244                 next_preprocessing_token();
1245         }
1246 }
1247
1248 static bool strings_equal(const string_t *string1, const string_t *string2)
1249 {
1250         size_t size = string1->size;
1251         if (size != string2->size)
1252                 return false;
1253
1254         const char *c1 = string1->begin;
1255         const char *c2 = string2->begin;
1256         for (size_t i = 0; i < size; ++i, ++c1, ++c2) {
1257                 if (*c1 != *c2)
1258                         return false;
1259         }
1260         return true;
1261 }
1262
1263 static bool pp_tokens_equal(const token_t *token1, const token_t *token2)
1264 {
1265         if (token1->kind != token2->kind)
1266                 return false;
1267
1268         switch (token1->kind) {
1269         case TP_IDENTIFIER:
1270                 return token1->base.symbol == token2->base.symbol;
1271
1272         case TP_NUMBER:
1273         case TP_CHARACTER_CONSTANT:
1274         case TP_STRING_LITERAL:
1275                 return strings_equal(&token1->string.string, &token2->string.string);
1276
1277         default:
1278                 return true;
1279         }
1280 }
1281
1282 static bool pp_definitions_equal(const pp_definition_t *definition1,
1283                                  const pp_definition_t *definition2)
1284 {
1285         if (definition1->list_len != definition2->list_len)
1286                 return false;
1287
1288         size_t         len = definition1->list_len;
1289         const token_t *t1  = definition1->token_list;
1290         const token_t *t2  = definition2->token_list;
1291         for (size_t i = 0; i < len; ++i, ++t1, ++t2) {
1292                 if (!pp_tokens_equal(t1, t2))
1293                         return false;
1294         }
1295         return true;
1296 }
1297
1298 static void parse_define_directive(void)
1299 {
1300         eat_pp(TP_define);
1301         assert(obstack_object_size(&pp_obstack) == 0);
1302
1303         if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1304                 errorf(&pp_token.base.source_position,
1305                        "expected identifier after #define, got '%t'", &pp_token);
1306                 goto error_out;
1307         }
1308         symbol_t *const symbol = pp_token.base.symbol;
1309
1310         pp_definition_t *new_definition
1311                 = obstack_alloc(&pp_obstack, sizeof(new_definition[0]));
1312         memset(new_definition, 0, sizeof(new_definition[0]));
1313         new_definition->source_position = input.position;
1314
1315         /* this is probably the only place where spaces are significant in the
1316          * lexer (except for the fact that they separate tokens). #define b(x)
1317          * is something else than #define b (x) */
1318         if (input.c == '(') {
1319                 /* eat the '(' */
1320                 next_preprocessing_token();
1321                 /* get next token after '(' */
1322                 next_preprocessing_token();
1323
1324                 while (true) {
1325                         switch (pp_token.kind) {
1326                         case TP_DOTDOTDOT:
1327                                 new_definition->is_variadic = true;
1328                                 next_preprocessing_token();
1329                                 if (pp_token.kind != ')') {
1330                                         errorf(&input.position,
1331                                                         "'...' not at end of macro argument list");
1332                                         goto error_out;
1333                                 }
1334                                 break;
1335                         case TP_IDENTIFIER:
1336                                 obstack_ptr_grow(&pp_obstack, pp_token.base.symbol);
1337                                 next_preprocessing_token();
1338
1339                                 if (pp_token.kind == ',') {
1340                                         next_preprocessing_token();
1341                                         break;
1342                                 }
1343
1344                                 if (pp_token.kind != ')') {
1345                                         errorf(&pp_token.base.source_position,
1346                                                "expected ',' or ')' after identifier, got '%t'",
1347                                                &pp_token);
1348                                         goto error_out;
1349                                 }
1350                                 break;
1351                         case ')':
1352                                 next_preprocessing_token();
1353                                 goto finish_argument_list;
1354                         default:
1355                                 errorf(&pp_token.base.source_position,
1356                                        "expected identifier, '...' or ')' in #define argument list, got '%t'",
1357                                        &pp_token);
1358                                 goto error_out;
1359                         }
1360                 }
1361
1362         finish_argument_list:
1363                 new_definition->has_parameters = true;
1364                 new_definition->n_parameters
1365                         = obstack_object_size(&pp_obstack) / sizeof(new_definition->parameters[0]);
1366                 new_definition->parameters = obstack_finish(&pp_obstack);
1367         } else {
1368                 next_preprocessing_token();
1369         }
1370
1371         /* construct a new pp_definition on the obstack */
1372         assert(obstack_object_size(&pp_obstack) == 0);
1373         size_t list_len = 0;
1374         while (!info.at_line_begin) {
1375                 obstack_grow(&pp_obstack, &pp_token, sizeof(pp_token));
1376                 ++list_len;
1377                 next_preprocessing_token();
1378         }
1379
1380         new_definition->list_len   = list_len;
1381         new_definition->token_list = obstack_finish(&pp_obstack);
1382
1383         pp_definition_t *old_definition = symbol->pp_definition;
1384         if (old_definition != NULL) {
1385                 if (!pp_definitions_equal(old_definition, new_definition)) {
1386                         warningf(WARN_OTHER, &input.position, "multiple definition of macro '%Y' (first defined %P)", symbol, &old_definition->source_position);
1387                 } else {
1388                         /* reuse the old definition */
1389                         obstack_free(&pp_obstack, new_definition);
1390                         new_definition = old_definition;
1391                 }
1392         }
1393
1394         symbol->pp_definition = new_definition;
1395         return;
1396
1397 error_out:
1398         if (obstack_object_size(&pp_obstack) > 0) {
1399                 char *ptr = obstack_finish(&pp_obstack);
1400                 obstack_free(&pp_obstack, ptr);
1401         }
1402         eat_pp_directive();
1403 }
1404
1405 static void parse_undef_directive(void)
1406 {
1407         eat_pp(TP_undef);
1408
1409         if (pp_token.kind != TP_IDENTIFIER) {
1410                 errorf(&input.position,
1411                        "expected identifier after #undef, got '%t'", &pp_token);
1412                 eat_pp_directive();
1413                 return;
1414         }
1415
1416         pp_token.base.symbol->pp_definition = NULL;
1417         next_preprocessing_token();
1418
1419         if (!info.at_line_begin) {
1420                 warningf(WARN_OTHER, &input.position, "extra tokens at end of #undef directive");
1421         }
1422         eat_pp_directive();
1423 }
1424
1425 static void parse_headername(void)
1426 {
1427         const source_position_t start_position = input.position;
1428         string_t                string         = {NULL, 0};
1429         assert(obstack_object_size(&symbol_obstack) == 0);
1430
1431         /* behind an #include we can have the special headername lexems.
1432          * They're only allowed behind an #include so they're not recognized
1433          * by the normal next_preprocessing_token. We handle them as a special
1434          * exception here */
1435         if (info.at_line_begin) {
1436                 parse_error("expected headername after #include");
1437                 goto finish_error;
1438         }
1439
1440         /* check wether we have a "... or <... headername */
1441         switch (input.c) {
1442         case '<':
1443                 next_char();
1444                 while (true) {
1445                         switch (input.c) {
1446                         case EOF:
1447                                 /* fallthrough */
1448                         MATCH_NEWLINE(
1449                                 parse_error("header name without closing '>'");
1450                                 goto finish_error;
1451                         )
1452                         case '>':
1453                                 next_char();
1454                                 goto finished_headername;
1455                         }
1456                         obstack_1grow(&symbol_obstack, (char) input.c);
1457                         next_char();
1458                 }
1459                 /* we should never be here */
1460
1461         case '"':
1462                 next_char();
1463                 while (true) {
1464                         switch (input.c) {
1465                         case EOF:
1466                                 /* fallthrough */
1467                         MATCH_NEWLINE(
1468                                 parse_error("header name without closing '>'");
1469                                 goto finish_error;
1470                         )
1471                         case '"':
1472                                 next_char();
1473                                 goto finished_headername;
1474                         }
1475                         obstack_1grow(&symbol_obstack, (char) input.c);
1476                         next_char();
1477                 }
1478                 /* we should never be here */
1479
1480         default:
1481                 /* TODO: do normal pp_token parsing and concatenate results */
1482                 panic("pp_token concat include not implemented yet");
1483         }
1484
1485 finished_headername:
1486         obstack_1grow(&symbol_obstack, '\0');
1487         const size_t size       = (size_t)obstack_object_size(&symbol_obstack);
1488         char *const  headername = obstack_finish(&symbol_obstack);
1489         string                  = make_string(headername, size);
1490
1491 finish_error:
1492         pp_token.base.source_position = start_position;
1493         pp_token.kind                 = TP_HEADERNAME;
1494         pp_token.string.string        = string;
1495 }
1496
1497 static bool do_include(bool system_include, const char *headername)
1498 {
1499         size_t headername_len = strlen(headername);
1500         if (!system_include) {
1501                 /* put dirname of current input on obstack */
1502                 const char *filename   = input.position.input_name;
1503                 const char *last_slash = strrchr(filename, '/');
1504                 if (last_slash != NULL) {
1505                         size_t len = last_slash - filename;
1506                         obstack_grow(&symbol_obstack, filename, len + 1);
1507                         obstack_grow0(&symbol_obstack, headername, headername_len);
1508                         char *complete_path = obstack_finish(&symbol_obstack);
1509                         headername = identify_string(complete_path);
1510                 }
1511
1512                 FILE *file = fopen(headername, "r");
1513                 if (file != NULL) {
1514                         switch_input(file, headername);
1515                         return true;
1516                 }
1517         }
1518
1519         assert(obstack_object_size(&symbol_obstack) == 0);
1520         /* check searchpath */
1521         for (searchpath_entry_t *entry = searchpath; entry != NULL;
1522              entry = entry->next) {
1523             const char *path = entry->path;
1524             size_t      len  = strlen(path);
1525                 obstack_grow(&symbol_obstack, path, len);
1526                 if (path[len-1] != '/')
1527                         obstack_1grow(&symbol_obstack, '/');
1528                 obstack_grow(&symbol_obstack, headername, headername_len+1);
1529
1530                 char *complete_path = obstack_finish(&symbol_obstack);
1531                 FILE *file          = fopen(complete_path, "r");
1532                 if (file != NULL) {
1533                         const char *filename = identify_string(complete_path);
1534                         switch_input(file, filename);
1535                         return true;
1536                 } else {
1537                         obstack_free(&symbol_obstack, complete_path);
1538                 }
1539         }
1540
1541         return false;
1542 }
1543
1544 /* read till next newline character, only for parse_include_directive(),
1545  * use eat_pp_directive() in all other cases */
1546 static void skip_till_newline(void)
1547 {
1548         /* skip till newline */
1549         while (true) {
1550                 switch (input.c) {
1551                 MATCH_NEWLINE(
1552                         return;
1553                 )
1554                 case EOF:
1555                         return;
1556                 }
1557                 next_char();
1558         }
1559 }
1560
1561 static bool parse_include_directive(void)
1562 {
1563         /* don't eat the TP_include here!
1564          * we need an alternative parsing for the next token */
1565         skip_whitespace();
1566         bool system_include = input.c == '<';
1567         parse_headername();
1568         string_t headername = pp_token.string.string;
1569         if (headername.begin == NULL) {
1570                 eat_pp_directive();
1571                 return false;
1572         }
1573
1574         skip_whitespace();
1575         if (!info.at_line_begin) {
1576                 warningf(WARN_OTHER, &pp_token.base.source_position,
1577                          "extra tokens at end of #include directive");
1578                 skip_till_newline();
1579         }
1580
1581         if (n_inputs > INCLUDE_LIMIT) {
1582                 errorf(&pp_token.base.source_position, "#include nested too deeply");
1583                 /* eat \n or EOF */
1584                 next_preprocessing_token();
1585                 return false;
1586         }
1587
1588         /* switch inputs */
1589         emit_newlines();
1590         push_input();
1591         bool res = do_include(system_include, pp_token.string.string.begin);
1592         if (!res) {
1593                 errorf(&pp_token.base.source_position,
1594                        "failed including '%S': %s", pp_token.string, strerror(errno));
1595                 pop_restore_input();
1596                 return false;
1597         }
1598
1599         return true;
1600 }
1601
1602 static pp_conditional_t *push_conditional(void)
1603 {
1604         pp_conditional_t *conditional
1605                 = obstack_alloc(&pp_obstack, sizeof(*conditional));
1606         memset(conditional, 0, sizeof(*conditional));
1607
1608         conditional->parent = conditional_stack;
1609         conditional_stack   = conditional;
1610
1611         return conditional;
1612 }
1613
1614 static void pop_conditional(void)
1615 {
1616         assert(conditional_stack != NULL);
1617         conditional_stack = conditional_stack->parent;
1618 }
1619
1620 static void check_unclosed_conditionals(void)
1621 {
1622         while (conditional_stack != NULL) {
1623                 pp_conditional_t *conditional = conditional_stack;
1624
1625                 if (conditional->in_else) {
1626                         errorf(&conditional->source_position, "unterminated #else");
1627                 } else {
1628                         errorf(&conditional->source_position, "unterminated condition");
1629                 }
1630                 pop_conditional();
1631         }
1632 }
1633
1634 static void parse_ifdef_ifndef_directive(void)
1635 {
1636         bool is_ifndef = (pp_token.kind == TP_ifndef);
1637         bool condition;
1638         next_preprocessing_token();
1639
1640         if (skip_mode) {
1641                 eat_pp_directive();
1642                 pp_conditional_t *conditional = push_conditional();
1643                 conditional->source_position  = pp_token.base.source_position;
1644                 conditional->skip             = true;
1645                 return;
1646         }
1647
1648         if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1649                 errorf(&pp_token.base.source_position,
1650                        "expected identifier after #%s, got '%t'",
1651                        is_ifndef ? "ifndef" : "ifdef", &pp_token);
1652                 eat_pp_directive();
1653
1654                 /* just take the true case in the hope to avoid further errors */
1655                 condition = true;
1656         } else {
1657                 /* evaluate wether we are in true or false case */
1658                 condition = !pp_token.base.symbol->pp_definition == is_ifndef;
1659
1660                 next_preprocessing_token();
1661
1662                 if (!info.at_line_begin) {
1663                         errorf(&pp_token.base.source_position,
1664                                "extra tokens at end of #%s",
1665                                is_ifndef ? "ifndef" : "ifdef");
1666                         eat_pp_directive();
1667                 }
1668         }
1669
1670         pp_conditional_t *conditional = push_conditional();
1671         conditional->source_position  = pp_token.base.source_position;
1672         conditional->condition        = condition;
1673
1674         if (!condition) {
1675                 skip_mode = true;
1676         }
1677 }
1678
1679 static void parse_else_directive(void)
1680 {
1681         eat_pp(TP_else);
1682
1683         if (!info.at_line_begin) {
1684                 if (!skip_mode) {
1685                         warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #else");
1686                 }
1687                 eat_pp_directive();
1688         }
1689
1690         pp_conditional_t *conditional = conditional_stack;
1691         if (conditional == NULL) {
1692                 errorf(&pp_token.base.source_position, "#else without prior #if");
1693                 return;
1694         }
1695
1696         if (conditional->in_else) {
1697                 errorf(&pp_token.base.source_position,
1698                        "#else after #else (condition started %P)",
1699                        conditional->source_position);
1700                 skip_mode = true;
1701                 return;
1702         }
1703
1704         conditional->in_else = true;
1705         if (!conditional->skip) {
1706                 skip_mode = conditional->condition;
1707         }
1708         conditional->source_position = pp_token.base.source_position;
1709 }
1710
1711 static void parse_endif_directive(void)
1712 {
1713         eat_pp(TP_endif);
1714
1715         if (!info.at_line_begin) {
1716                 if (!skip_mode) {
1717                         warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #endif");
1718                 }
1719                 eat_pp_directive();
1720         }
1721
1722         pp_conditional_t *conditional = conditional_stack;
1723         if (conditional == NULL) {
1724                 errorf(&pp_token.base.source_position, "#endif without prior #if");
1725                 return;
1726         }
1727
1728         if (!conditional->skip) {
1729                 skip_mode = false;
1730         }
1731         pop_conditional();
1732 }
1733
1734 static void parse_preprocessing_directive(void)
1735 {
1736         eat_pp('#');
1737
1738         if (info.at_line_begin) {
1739                 /* empty directive */
1740                 return;
1741         }
1742
1743         if (skip_mode) {
1744                 switch (pp_token.kind) {
1745                 case TP_ifdef:
1746                 case TP_ifndef:
1747                         parse_ifdef_ifndef_directive();
1748                         break;
1749                 case TP_else:
1750                         parse_else_directive();
1751                         break;
1752                 case TP_endif:
1753                         parse_endif_directive();
1754                         break;
1755                 default:
1756                         eat_pp_directive();
1757                         break;
1758                 }
1759         } else {
1760                 switch (pp_token.kind) {
1761                 case TP_define:
1762                         parse_define_directive();
1763                         break;
1764                 case TP_undef:
1765                         parse_undef_directive();
1766                         break;
1767                 case TP_ifdef:
1768                 case TP_ifndef:
1769                         parse_ifdef_ifndef_directive();
1770                         break;
1771                 case TP_else:
1772                         parse_else_directive();
1773                         break;
1774                 case TP_endif:
1775                         parse_endif_directive();
1776                         break;
1777                 case TP_include:
1778                         parse_include_directive();
1779                         break;
1780                 default:
1781                         if (info.at_line_begin) {
1782                                 /* the nop directive "#" */
1783                                 break;
1784                         }
1785                         errorf(&pp_token.base.source_position,
1786                                    "invalid preprocessing directive #%t", &pp_token);
1787                         eat_pp_directive();
1788                         break;
1789                 }
1790         }
1791
1792         assert(info.at_line_begin);
1793 }
1794
1795 static void prepend_include_path(const char *path)
1796 {
1797         searchpath_entry_t *entry = OALLOCZ(&config_obstack, searchpath_entry_t);
1798         entry->path = path;
1799         entry->next = searchpath;
1800         searchpath  = entry;
1801 }
1802
1803 static void setup_include_path(void)
1804 {
1805         /* built-in paths */
1806         prepend_include_path("/usr/include");
1807
1808         /* parse environment variable */
1809         const char *cpath = getenv("CPATH");
1810         if (cpath != NULL && *cpath != '\0') {
1811                 const char *begin = cpath;
1812                 const char *c;
1813                 do {
1814                         c = begin;
1815                         while (*c != '\0' && *c != ':')
1816                                 ++c;
1817
1818                         size_t len = c-begin;
1819                         if (len == 0) {
1820                                 /* for gcc compatibility (Matze: I would expect that
1821                                  * nothing happens for an empty entry...) */
1822                                 prepend_include_path(".");
1823                         } else {
1824                                 char *string = obstack_alloc(&config_obstack, len+1);
1825                                 memcpy(string, begin, len);
1826                                 string[len] = '\0';
1827
1828                                 prepend_include_path(string);
1829                         }
1830
1831                         begin = c+1;
1832                         /* skip : */
1833                         if (*begin == ':')
1834                                 ++begin;
1835                 } while(*c != '\0');
1836         }
1837 }
1838
1839 int pptest_main(int argc, char **argv);
1840 int pptest_main(int argc, char **argv)
1841 {
1842         init_symbol_table();
1843         init_tokens();
1844
1845         obstack_init(&config_obstack);
1846         obstack_init(&pp_obstack);
1847         obstack_init(&input_obstack);
1848         strset_init(&stringset);
1849
1850         setup_include_path();
1851
1852         /* simplistic commandline parser */
1853         const char *filename = NULL;
1854         const char *output = NULL;
1855         for (int i = 1; i < argc; ++i) {
1856                 const char *opt = argv[i];
1857                 if (streq(opt, "-I")) {
1858                         prepend_include_path(argv[++i]);
1859                         continue;
1860                 } else if (streq(opt, "-E")) {
1861                         /* ignore */
1862                 } else if (streq(opt, "-o")) {
1863                         output = argv[++i];
1864                         continue;
1865                 } else if (opt[0] == '-') {
1866                         fprintf(stderr, "Unknown option '%s'\n", opt);
1867                 } else {
1868                         if (filename != NULL)
1869                                 fprintf(stderr, "Multiple inputs not supported\n");
1870                         filename = argv[i];
1871                 }
1872         }
1873         if (filename == NULL) {
1874                 fprintf(stderr, "No input specified\n");
1875                 return 1;
1876         }
1877
1878         if (output == NULL) {
1879                 out = stdout;
1880         } else {
1881                 out = fopen(output, "w");
1882                 if (out == NULL) {
1883                         fprintf(stderr, "Couldn't open output '%s'\n", output);
1884                         return 1;
1885                 }
1886         }
1887
1888         /* just here for gcc compatibility */
1889         fprintf(out, "# 1 \"%s\"\n", filename);
1890         fprintf(out, "# 1 \"<built-in>\"\n");
1891         fprintf(out, "# 1 \"<command-line>\"\n");
1892
1893         FILE *file = fopen(filename, "r");
1894         if (file == NULL) {
1895                 fprintf(stderr, "Couldn't open input '%s'\n", filename);
1896                 return 1;
1897         }
1898         switch_input(file, filename);
1899
1900         while (true) {
1901                 if (pp_token.kind == '#' && info.at_line_begin) {
1902                         parse_preprocessing_directive();
1903                         continue;
1904                 } else if (pp_token.kind == TP_EOF) {
1905                         goto end_of_main_loop;
1906                 } else if (pp_token.kind == TP_IDENTIFIER) {
1907                         symbol_t        *const symbol        = pp_token.base.symbol;
1908                         pp_definition_t *const pp_definition = symbol->pp_definition;
1909                         if (pp_definition != NULL && !pp_definition->is_expanding) {
1910                                 expansion_pos = pp_token.base.source_position;
1911                                 if (pp_definition->has_parameters) {
1912                                         source_position_t position = pp_token.base.source_position;
1913                                         add_token_info_t old_info = info;
1914                                         next_preprocessing_token();
1915                                         add_token_info_t new_info = info;
1916
1917                                         /* no opening brace -> no expansion */
1918                                         if (pp_token.kind == '(') {
1919                                                 eat_pp('(');
1920
1921                                                 /* parse arguments (TODO) */
1922                                                 while (pp_token.kind != TP_EOF && pp_token.kind != ')')
1923                                                         next_preprocessing_token();
1924                                         } else {
1925                                                 token_t next_token = pp_token;
1926                                                 /* restore identifier token */
1927                                                 pp_token.kind                 = TP_IDENTIFIER;
1928                                                 pp_token.base.symbol          = symbol;
1929                                                 pp_token.base.source_position = position;
1930                                                 info = old_info;
1931                                                 emit_pp_token();
1932
1933                                                 info = new_info;
1934                                                 pp_token = next_token;
1935                                                 continue;
1936                                         }
1937                                         info = old_info;
1938                                 }
1939                                 pp_definition->expand_pos   = 0;
1940                                 pp_definition->is_expanding = true;
1941                                 current_expansion           = pp_definition;
1942                                 expand_next();
1943                                 continue;
1944                         }
1945                 }
1946
1947                 emit_pp_token();
1948                 next_preprocessing_token();
1949         }
1950 end_of_main_loop:
1951
1952         fputc('\n', out);
1953         check_unclosed_conditionals();
1954         close_input();
1955         if (out != stdout)
1956                 fclose(out);
1957
1958         obstack_free(&input_obstack, NULL);
1959         obstack_free(&pp_obstack, NULL);
1960         obstack_free(&config_obstack, NULL);
1961
1962         strset_destroy(&stringset);
1963
1964         exit_tokens();
1965         exit_symbol_table();
1966
1967         return 0;
1968 }