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              error_on_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 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 NEWLINE \
 222         '\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                 goto newline; \
 232                 newline // Let it look like an ordinary case label.
 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         case NEWLINE:
 242                 return;
 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 sym_make_string(string_encoding_t const enc)
 455 {
 456         obstack_1grow(&symbol_obstack, '\0');
 457         size_t      const len    = obstack_object_size(&symbol_obstack) - 1;
 458         char       *const string = obstack_finish(&symbol_obstack);
 459         char const *const result = identify_string(string);
 460         return (string_t){ result, len, enc };
 461 }
 462
 463 static void parse_string(utf32 const delimiter, token_kind_t const kind,
 464                          string_encoding_t const enc,
 465                          char const *const context)
 466 {
 467         const unsigned start_linenr = input.position.lineno;
 468
 469         eat(delimiter);
 470
 471         while (true) {
 472                 switch (input.c) {
 473                 case '\\': {
 474                         if (resolve_escape_sequences) {
 475                                 utf32 const tc = parse_escape_sequence();
 476                                 if (enc == STRING_ENCODING_CHAR) {
 477                                         if (tc >= 0x100) {
 478                                                 warningf(WARN_OTHER, &pp_token.base.source_position, "escape sequence out of range");
 479                                         }
 480                                         obstack_1grow(&symbol_obstack, tc);
 481                                 } else {
 482                                         obstack_grow_utf8(&symbol_obstack, tc);
 483                                 }
 484                         } else {
 485                                 obstack_1grow(&symbol_obstack, (char)input.c);
 486                                 next_char();
 487                                 obstack_1grow(&symbol_obstack, (char)input.c);
 488                                 next_char();
 489                         }
 490                         break;
 491                 }
 492
 493                 case NEWLINE:
 494                         errorf(&pp_token.base.source_position, "newline while parsing %s", context);
 495                         break;
 496
 497                 case EOF: {
 498                         source_position_t source_position;
 499                         source_position.input_name = pp_token.base.source_position.input_name;
 500                         source_position.lineno     = start_linenr;
 501                         errorf(&source_position, "EOF while parsing %s", context);
 502                         goto end_of_string;
 503                 }
 504
 505                 default:
 506                         if (input.c == delimiter) {
 507                                 next_char();
 508                                 goto end_of_string;
 509                         } else {
 510                                 obstack_grow_utf8(&symbol_obstack, input.c);
 511                                 next_char();
 512                                 break;
 513                         }
 514                 }
 515         }
 516
 517 end_of_string:
 518         pp_token.kind           = kind;
 519         pp_token.literal.string = sym_make_string(enc);
 520 }
 521
 522 static void parse_string_literal(string_encoding_t const enc)
 523 {
 524         parse_string('"', T_STRING_LITERAL, enc, "string literal");
 525 }
 526
 527 static void parse_character_constant(string_encoding_t const enc)
 528 {
 529         parse_string('\'', T_CHARACTER_CONSTANT, enc, "character constant");
 530         if (pp_token.literal.string.size == 0) {
 531                 parse_error("empty character constant");
 532         }
 533 }
 534
 535 #define SYMBOL_CASES_WITHOUT_E_P \
 536              'a': \
 537         case 'b': \
 538         case 'c': \
 539         case 'd': \
 540         case 'f': \
 541         case 'g': \
 542         case 'h': \
 543         case 'i': \
 544         case 'j': \
 545         case 'k': \
 546         case 'l': \
 547         case 'm': \
 548         case 'n': \
 549         case 'o': \
 550         case 'q': \
 551         case 'r': \
 552         case 's': \
 553         case 't': \
 554         case 'u': \
 555         case 'v': \
 556         case 'w': \
 557         case 'x': \
 558         case 'y': \
 559         case 'z': \
 560         case 'A': \
 561         case 'B': \
 562         case 'C': \
 563         case 'D': \
 564         case 'F': \
 565         case 'G': \
 566         case 'H': \
 567         case 'I': \
 568         case 'J': \
 569         case 'K': \
 570         case 'L': \
 571         case 'M': \
 572         case 'N': \
 573         case 'O': \
 574         case 'Q': \
 575         case 'R': \
 576         case 'S': \
 577         case 'T': \
 578         case 'U': \
 579         case 'V': \
 580         case 'W': \
 581         case 'X': \
 582         case 'Y': \
 583         case 'Z': \
 584         case '_'
 585
 586 #define SYMBOL_CASES \
 587              SYMBOL_CASES_WITHOUT_E_P: \
 588         case 'e': \
 589         case 'p': \
 590         case 'E': \
 591         case 'P'
 592
 593 #define DIGIT_CASES \
 594              '0':  \
 595         case '1':  \
 596         case '2':  \
 597         case '3':  \
 598         case '4':  \
 599         case '5':  \
 600         case '6':  \
 601         case '7':  \
 602         case '8':  \
 603         case '9'
 604
 605 /**
 606  * returns next final token from a preprocessor macro expansion
 607  */
 608 static void expand_next(void)
 609 {
 610         assert(current_expansion != NULL);
 611
 612         pp_definition_t *definition = current_expansion;
 613
 614 restart:
 615         if (definition->list_len == 0
 616                         || definition->expand_pos >= definition->list_len) {
 617                 /* we're finished with the current macro, move up 1 level in the
 618                  * expansion stack */
 619                 pp_definition_t *parent = definition->parent_expansion;
 620                 definition->parent_expansion = NULL;
 621                 definition->is_expanding     = false;
 622
 623                 /* it was the outermost expansion, parse normal pptoken */
 624                 if (parent == NULL) {
 625                         current_expansion = NULL;
 626                         next_preprocessing_token();
 627                         return;
 628                 }
 629                 definition        = parent;
 630                 current_expansion = definition;
 631                 goto restart;
 632         }
 633         pp_token = definition->token_list[definition->expand_pos];
 634         pp_token.base.source_position = expansion_pos;
 635         ++definition->expand_pos;
 636
 637         if (pp_token.kind != T_IDENTIFIER)
 638                 return;
 639
 640         /* if it was an identifier then we might need to expand again */
 641         pp_definition_t *const symbol_definition = pp_token.base.symbol->pp_definition;
 642         if (symbol_definition != NULL && !symbol_definition->is_expanding) {
 643                 symbol_definition->parent_expansion = definition;
 644                 symbol_definition->expand_pos       = 0;
 645                 symbol_definition->is_expanding     = true;
 646                 definition                          = symbol_definition;
 647                 current_expansion                   = definition;
 648                 goto restart;
 649         }
 650 }
 651
 652 static void skip_line_comment(void)
 653 {
 654         while (true) {
 655                 switch (input.c) {
 656                 case EOF:
 657                         return;
 658
 659                 case '\r':
 660                 case '\n':
 661                         return;
 662
 663                 default:
 664                         next_char();
 665                         break;
 666                 }
 667         }
 668 }
 669
 670 static void skip_multiline_comment(void)
 671 {
 672         unsigned start_linenr = input.position.lineno;
 673         while (true) {
 674                 switch (input.c) {
 675                 case '/':
 676                         next_char();
 677                         if (input.c == '*') {
 678                                 /* TODO: nested comment, warn here */
 679                         }
 680                         break;
 681                 case '*':
 682                         next_char();
 683                         if (input.c == '/') {
 684                                 if (input.position.lineno != input.output_line)
 685                                         info.whitespace = input.position.colno;
 686                                 next_char();
 687                                 return;
 688                         }
 689                         break;
 690
 691                 case NEWLINE:
 692                         break;
 693
 694                 case EOF: {
 695                         source_position_t source_position;
 696                         source_position.input_name = pp_token.base.source_position.input_name;
 697                         source_position.lineno     = start_linenr;
 698                         errorf(&source_position, "at end of file while looking for comment end");
 699                         return;
 700                 }
 701
 702                 default:
 703                         next_char();
 704                         break;
 705                 }
 706         }
 707 }
 708
 709 static void skip_whitespace(void)
 710 {
 711         while (true) {
 712                 switch (input.c) {
 713                 case ' ':
 714                 case '\t':
 715                         next_char();
 716                         continue;
 717
 718                 case NEWLINE:
 719                         info.at_line_begin = true;
 720                         return;
 721
 722                 case '/':
 723                         next_char();
 724                         if (input.c == '/') {
 725                                 next_char();
 726                                 skip_line_comment();
 727                                 continue;
 728                         } else if (input.c == '*') {
 729                                 next_char();
 730                                 skip_multiline_comment();
 731                                 continue;
 732                         } else {
 733                                 put_back(input.c);
 734                                 input.c = '/';
 735                         }
 736                         return;
 737                 default:
 738                         return;
 739                 }
 740         }
 741 }
 742
 743 static inline void eat_pp(pp_token_kind_t const kind)
 744 {
 745         assert(pp_token.base.symbol->pp_ID == kind);
 746         (void) kind;
 747         next_preprocessing_token();
 748 }
 749
 750 static inline void eat_token(token_kind_t const kind)
 751 {
 752         assert(pp_token.kind == kind);
 753         (void)kind;
 754         next_preprocessing_token();
 755 }
 756
 757 static void parse_symbol(void)
 758 {
 759         obstack_1grow(&symbol_obstack, (char) input.c);
 760         next_char();
 761
 762         while (true) {
 763                 switch (input.c) {
 764                 case DIGIT_CASES:
 765                 case SYMBOL_CASES:
 766                         obstack_1grow(&symbol_obstack, (char) input.c);
 767                         next_char();
 768                         break;
 769
 770                 default:
 771                         goto end_symbol;
 772                 }
 773         }
 774
 775 end_symbol:
 776         obstack_1grow(&symbol_obstack, '\0');
 777         char *string = obstack_finish(&symbol_obstack);
 778
 779         /* might be a wide string or character constant ( L"string"/L'c' ) */
 780         if (input.c == '"' && string[0] == 'L' && string[1] == '\0') {
 781                 obstack_free(&symbol_obstack, string);
 782                 parse_string_literal(STRING_ENCODING_WIDE);
 783                 return;
 784         } else if (input.c == '\'' && string[0] == 'L' && string[1] == '\0') {
 785                 obstack_free(&symbol_obstack, string);
 786                 parse_character_constant(STRING_ENCODING_WIDE);
 787                 return;
 788         }
 789
 790         symbol_t *symbol = symbol_table_insert(string);
 791
 792         pp_token.kind        = symbol->ID;
 793         pp_token.base.symbol = symbol;
 794
 795         /* we can free the memory from symbol obstack if we already had an entry in
 796          * the symbol table */
 797         if (symbol->string != string) {
 798                 obstack_free(&symbol_obstack, string);
 799         }
 800 }
 801
 802 static void parse_number(void)
 803 {
 804         obstack_1grow(&symbol_obstack, (char) input.c);
 805         next_char();
 806
 807         while (true) {
 808                 switch (input.c) {
 809                 case '.':
 810                 case DIGIT_CASES:
 811                 case SYMBOL_CASES_WITHOUT_E_P:
 812                         obstack_1grow(&symbol_obstack, (char) input.c);
 813                         next_char();
 814                         break;
 815
 816                 case 'e':
 817                 case 'p':
 818                 case 'E':
 819                 case 'P':
 820                         obstack_1grow(&symbol_obstack, (char) input.c);
 821                         next_char();
 822                         if (input.c == '+' || input.c == '-') {
 823                                 obstack_1grow(&symbol_obstack, (char) input.c);
 824                                 next_char();
 825                         }
 826                         break;
 827
 828                 default:
 829                         goto end_number;
 830                 }
 831         }
 832
 833 end_number:
 834         pp_token.kind           = T_NUMBER;
 835         pp_token.literal.string = sym_make_string(STRING_ENCODING_CHAR);
 836 }
 837
 838 #define MAYBE_PROLOG \
 839         next_char(); \
 840         switch (input.c) {
 841
 842 #define MAYBE(ch, kind) \
 843         case ch: \
 844                 next_char(); \
 845                 pp_token.kind = kind; \
 846                 return;
 847
 848 #define ELSE_CODE(code) \
 849         default: \
 850                 code \
 851                 return; \
 852         }
 853
 854 #define ELSE(kind) ELSE_CODE(pp_token.kind = kind;)
 855
 856 static void next_preprocessing_token(void)
 857 {
 858         if (current_expansion != NULL) {
 859                 expand_next();
 860                 return;
 861         }
 862
 863         info.at_line_begin  = false;
 864         info.had_whitespace = false;
 865 restart:
 866         pp_token.base.source_position = input.position;
 867         pp_token.base.symbol          = NULL;
 868
 869         switch (input.c) {
 870         case ' ':
 871         case '\t':
 872                 ++info.whitespace;
 873                 info.had_whitespace = true;
 874                 next_char();
 875                 goto restart;
 876
 877         case NEWLINE:
 878                 info.at_line_begin = true;
 879                 info.had_whitespace = true;
 880                 goto restart;
 881
 882         case SYMBOL_CASES:
 883                 parse_symbol();
 884                 return;
 885
 886         case DIGIT_CASES:
 887                 parse_number();
 888                 return;
 889
 890         case '"':
 891                 parse_string_literal(STRING_ENCODING_CHAR);
 892                 return;
 893
 894         case '\'':
 895                 parse_character_constant(STRING_ENCODING_CHAR);
 896                 return;
 897
 898         case '.':
 899                 MAYBE_PROLOG
 900                         case '0':
 901                         case '1':
 902                         case '2':
 903                         case '3':
 904                         case '4':
 905                         case '5':
 906                         case '6':
 907                         case '7':
 908                         case '8':
 909                         case '9':
 910                                 put_back(input.c);
 911                                 input.c = '.';
 912                                 parse_number();
 913                                 return;
 914
 915                         case '.':
 916                                 MAYBE_PROLOG
 917                                 MAYBE('.', T_DOTDOTDOT)
 918                                 ELSE_CODE(
 919                                         put_back(input.c);
 920                                         input.c = '.';
 921                                         pp_token.kind = '.';
 922                                 )
 923                 ELSE('.')
 924         case '&':
 925                 MAYBE_PROLOG
 926                 MAYBE('&', T_ANDAND)
 927                 MAYBE('=', T_ANDEQUAL)
 928                 ELSE('&')
 929         case '*':
 930                 MAYBE_PROLOG
 931                 MAYBE('=', T_ASTERISKEQUAL)
 932                 ELSE('*')
 933         case '+':
 934                 MAYBE_PROLOG
 935                 MAYBE('+', T_PLUSPLUS)
 936                 MAYBE('=', T_PLUSEQUAL)
 937                 ELSE('+')
 938         case '-':
 939                 MAYBE_PROLOG
 940                 MAYBE('>', T_MINUSGREATER)
 941                 MAYBE('-', T_MINUSMINUS)
 942                 MAYBE('=', T_MINUSEQUAL)
 943                 ELSE('-')
 944         case '!':
 945                 MAYBE_PROLOG
 946                 MAYBE('=', T_EXCLAMATIONMARKEQUAL)
 947                 ELSE('!')
 948         case '/':
 949                 MAYBE_PROLOG
 950                 MAYBE('=', T_SLASHEQUAL)
 951                 case '*':
 952                         next_char();
 953                         info.had_whitespace = true;
 954                         skip_multiline_comment();
 955                         goto restart;
 956                 case '/':
 957                         next_char();
 958                         info.had_whitespace = true;
 959                         skip_line_comment();
 960                         goto restart;
 961                 ELSE('/')
 962         case '%':
 963                 MAYBE_PROLOG
 964                 MAYBE('>', '}')
 965                 MAYBE('=', T_PERCENTEQUAL)
 966                 case ':':
 967                         MAYBE_PROLOG
 968                         case '%':
 969                                 MAYBE_PROLOG
 970                                 MAYBE(':', T_HASHHASH)
 971                                 ELSE_CODE(
 972                                         put_back(input.c);
 973                                         input.c = '%';
 974                                         pp_token.kind = '#';
 975                                 )
 976                         ELSE('#')
 977                 ELSE('%')
 978         case '<':
 979                 MAYBE_PROLOG
 980                 MAYBE(':', '[')
 981                 MAYBE('%', '{')
 982                 MAYBE('=', T_LESSEQUAL)
 983                 case '<':
 984                         MAYBE_PROLOG
 985                         MAYBE('=', T_LESSLESSEQUAL)
 986                         ELSE(T_LESSLESS)
 987                 ELSE('<')
 988         case '>':
 989                 MAYBE_PROLOG
 990                 MAYBE('=', T_GREATEREQUAL)
 991                 case '>':
 992                         MAYBE_PROLOG
 993                         MAYBE('=', T_GREATERGREATEREQUAL)
 994                         ELSE(T_GREATERGREATER)
 995                 ELSE('>')
 996         case '^':
 997                 MAYBE_PROLOG
 998                 MAYBE('=', T_CARETEQUAL)
 999                 ELSE('^')
1000         case '|':
1001                 MAYBE_PROLOG
1002                 MAYBE('=', T_PIPEEQUAL)
1003                 MAYBE('|', T_PIPEPIPE)
1004                 ELSE('|')
1005         case ':':
1006                 MAYBE_PROLOG
1007                 MAYBE('>', ']')
1008                 ELSE(':')
1009         case '=':
1010                 MAYBE_PROLOG
1011                 MAYBE('=', T_EQUALEQUAL)
1012                 ELSE('=')
1013         case '#':
1014                 MAYBE_PROLOG
1015                 MAYBE('#', T_HASHHASH)
1016                 ELSE_CODE(
1017                         pp_token.kind = '#';
1018                 )
1019
1020         case '?':
1021         case '[':
1022         case ']':
1023         case '(':
1024         case ')':
1025         case '{':
1026         case '}':
1027         case '~':
1028         case ';':
1029         case ',':
1030                 pp_token.kind = input.c;
1031                 next_char();
1032                 return;
1033
1034         case EOF:
1035                 if (input_stack != NULL) {
1036                         close_input();
1037                         pop_restore_input();
1038                         fputc('\n', out);
1039                         print_line_directive(&input.position, "2");
1040                         goto restart;
1041                 } else {
1042                         pp_token.base.source_position.lineno++;
1043                         info.at_line_begin = true;
1044                         pp_token.kind      = T_EOF;
1045                 }
1046                 return;
1047
1048         default:
1049                 if (error_on_unknown_chars) {
1050                         errorf(&pp_token.base.source_position,
1051                                "unknown character '%lc' found\n", input.c);
1052                         next_char();
1053                         goto restart;
1054                 } else {
1055                         assert(obstack_object_size(&symbol_obstack) == 0);
1056                         obstack_grow_utf8(&symbol_obstack, input.c);
1057                         obstack_1grow(&symbol_obstack, '\0');
1058                         char     *const string = obstack_finish(&symbol_obstack);
1059                         symbol_t *const symbol = symbol_table_insert(string);
1060                         if (symbol->string != string)
1061                                 obstack_free(&symbol_obstack, string);
1062
1063                         pp_token.kind        = T_UNKNOWN_CHAR;
1064                         pp_token.base.symbol = symbol;
1065                         next_char();
1066                         return;
1067                 }
1068         }
1069 }
1070
1071 static void print_quoted_string(const char *const string)
1072 {
1073         fputc('"', out);
1074         for (const char *c = string; *c != 0; ++c) {
1075                 switch (*c) {
1076                 case '"': fputs("\\\"", out); break;
1077                 case '\\':  fputs("\\\\", out); break;
1078                 case '\a':  fputs("\\a", out); break;
1079                 case '\b':  fputs("\\b", out); break;
1080                 case '\f':  fputs("\\f", out); break;
1081                 case '\n':  fputs("\\n", out); break;
1082                 case '\r':  fputs("\\r", out); break;
1083                 case '\t':  fputs("\\t", out); break;
1084                 case '\v':  fputs("\\v", out); break;
1085                 case '\?':  fputs("\\?", out); break;
1086                 default:
1087                         if (!isprint(*c)) {
1088                                 fprintf(out, "\\%03o", (unsigned)*c);
1089                                 break;
1090                         }
1091                         fputc(*c, out);
1092                         break;
1093                 }
1094         }
1095         fputc('"', out);
1096 }
1097
1098 static void print_line_directive(const source_position_t *pos, const char *add)
1099 {
1100         fprintf(out, "# %u ", pos->lineno);
1101         print_quoted_string(pos->input_name);
1102         if (add != NULL) {
1103                 fputc(' ', out);
1104                 fputs(add, out);
1105         }
1106
1107         printed_input_name = pos->input_name;
1108         input.output_line  = pos->lineno-1;
1109 }
1110
1111 static bool emit_newlines(void)
1112 {
1113         unsigned delta = pp_token.base.source_position.lineno - input.output_line;
1114         if (delta == 0)
1115                 return false;
1116
1117         if (delta >= 9) {
1118                 fputc('\n', out);
1119                 print_line_directive(&pp_token.base.source_position, NULL);
1120                 fputc('\n', out);
1121         } else {
1122                 for (unsigned i = 0; i < delta; ++i) {
1123                         fputc('\n', out);
1124                 }
1125         }
1126         input.output_line = pp_token.base.source_position.lineno;
1127
1128         for (unsigned i = 0; i < info.whitespace; ++i)
1129                 fputc(' ', out);
1130
1131         return true;
1132 }
1133
1134 static void emit_pp_token(void)
1135 {
1136         if (skip_mode)
1137                 return;
1138
1139         if (!emit_newlines() &&
1140             (info.had_whitespace || tokens_would_paste(last_token, pp_token.kind)))
1141                 fputc(' ', out);
1142
1143         switch (pp_token.kind) {
1144         case T_NUMBER:
1145                 fputs(pp_token.literal.string.begin, out);
1146                 break;
1147
1148         case T_STRING_LITERAL:
1149                 fputs(get_string_encoding_prefix(pp_token.literal.string.encoding), out);
1150                 fputc('"', out);
1151                 fputs(pp_token.literal.string.begin, out);
1152                 fputc('"', out);
1153                 break;
1154
1155         case T_CHARACTER_CONSTANT:
1156                 fputs(get_string_encoding_prefix(pp_token.literal.string.encoding), out);
1157                 fputc('\'', out);
1158                 fputs(pp_token.literal.string.begin, out);
1159                 fputc('\'', out);
1160                 break;
1161
1162         default:
1163                 if (pp_token.base.symbol) {
1164                         fputs(pp_token.base.symbol->string, out);
1165                 } else {
1166                         print_token_kind(out, pp_token.kind);
1167                 }
1168                 break;
1169         }
1170         last_token = pp_token.kind;
1171 }
1172
1173 static void eat_pp_directive(void)
1174 {
1175         while (!info.at_line_begin) {
1176                 next_preprocessing_token();
1177         }
1178 }
1179
1180 static bool strings_equal(const string_t *string1, const string_t *string2)
1181 {
1182         size_t size = string1->size;
1183         if (size != string2->size)
1184                 return false;
1185
1186         const char *c1 = string1->begin;
1187         const char *c2 = string2->begin;
1188         for (size_t i = 0; i < size; ++i, ++c1, ++c2) {
1189                 if (*c1 != *c2)
1190                         return false;
1191         }
1192         return true;
1193 }
1194
1195 static bool pp_tokens_equal(const token_t *token1, const token_t *token2)
1196 {
1197         if (token1->kind != token2->kind)
1198                 return false;
1199
1200         switch (token1->kind) {
1201         case T_IDENTIFIER:
1202                 return token1->base.symbol == token2->base.symbol;
1203
1204         case T_NUMBER:
1205         case T_CHARACTER_CONSTANT:
1206         case T_STRING_LITERAL:
1207                 return strings_equal(&token1->literal.string, &token2->literal.string);
1208
1209         default:
1210                 return true;
1211         }
1212 }
1213
1214 static bool pp_definitions_equal(const pp_definition_t *definition1,
1215                                  const pp_definition_t *definition2)
1216 {
1217         if (definition1->list_len != definition2->list_len)
1218                 return false;
1219
1220         size_t         len = definition1->list_len;
1221         const token_t *t1  = definition1->token_list;
1222         const token_t *t2  = definition2->token_list;
1223         for (size_t i = 0; i < len; ++i, ++t1, ++t2) {
1224                 if (!pp_tokens_equal(t1, t2))
1225                         return false;
1226         }
1227         return true;
1228 }
1229
1230 static void parse_define_directive(void)
1231 {
1232         eat_pp(TP_define);
1233         if (skip_mode) {
1234                 eat_pp_directive();
1235                 return;
1236         }
1237
1238         assert(obstack_object_size(&pp_obstack) == 0);
1239
1240         if (pp_token.kind != T_IDENTIFIER || info.at_line_begin) {
1241                 errorf(&pp_token.base.source_position,
1242                        "expected identifier after #define, got %K", &pp_token);
1243                 goto error_out;
1244         }
1245         symbol_t *const symbol = pp_token.base.symbol;
1246
1247         pp_definition_t *new_definition
1248                 = obstack_alloc(&pp_obstack, sizeof(new_definition[0]));
1249         memset(new_definition, 0, sizeof(new_definition[0]));
1250         new_definition->source_position = input.position;
1251
1252         /* this is probably the only place where spaces are significant in the
1253          * lexer (except for the fact that they separate tokens). #define b(x)
1254          * is something else than #define b (x) */
1255         if (input.c == '(') {
1256                 eat_token(T_IDENTIFIER);
1257                 eat_token('(');
1258
1259                 while (true) {
1260                         switch (pp_token.kind) {
1261                         case T_DOTDOTDOT:
1262                                 new_definition->is_variadic = true;
1263                                 eat_token(T_DOTDOTDOT);
1264                                 if (pp_token.kind != ')') {
1265                                         errorf(&input.position,
1266                                                         "'...' not at end of macro argument list");
1267                                         goto error_out;
1268                                 }
1269                                 break;
1270
1271                         case T_IDENTIFIER:
1272                                 obstack_ptr_grow(&pp_obstack, pp_token.base.symbol);
1273                                 eat_token(T_IDENTIFIER);
1274
1275                                 if (pp_token.kind == ',') {
1276                                         eat_token(',');
1277                                         break;
1278                                 }
1279
1280                                 if (pp_token.kind != ')') {
1281                                         errorf(&pp_token.base.source_position,
1282                                                "expected ',' or ')' after identifier, got %K",
1283                                                &pp_token);
1284                                         goto error_out;
1285                                 }
1286                                 break;
1287
1288                         case ')':
1289                                 eat_token(')');
1290                                 goto finish_argument_list;
1291
1292                         default:
1293                                 errorf(&pp_token.base.source_position,
1294                                        "expected identifier, '...' or ')' in #define argument list, got %K",
1295                                        &pp_token);
1296                                 goto error_out;
1297                         }
1298                 }
1299
1300         finish_argument_list:
1301                 new_definition->has_parameters = true;
1302                 new_definition->n_parameters
1303                         = obstack_object_size(&pp_obstack) / sizeof(new_definition->parameters[0]);
1304                 new_definition->parameters = obstack_finish(&pp_obstack);
1305         } else {
1306                 eat_token(T_IDENTIFIER);
1307         }
1308
1309         /* construct a new pp_definition on the obstack */
1310         assert(obstack_object_size(&pp_obstack) == 0);
1311         size_t list_len = 0;
1312         while (!info.at_line_begin) {
1313                 obstack_grow(&pp_obstack, &pp_token, sizeof(pp_token));
1314                 ++list_len;
1315                 next_preprocessing_token();
1316         }
1317
1318         new_definition->list_len   = list_len;
1319         new_definition->token_list = obstack_finish(&pp_obstack);
1320
1321         pp_definition_t *old_definition = symbol->pp_definition;
1322         if (old_definition != NULL) {
1323                 if (!pp_definitions_equal(old_definition, new_definition)) {
1324                         warningf(WARN_OTHER, &input.position, "multiple definition of macro '%Y' (first defined %P)", symbol, &old_definition->source_position);
1325                 } else {
1326                         /* reuse the old definition */
1327                         obstack_free(&pp_obstack, new_definition);
1328                         new_definition = old_definition;
1329                 }
1330         }
1331
1332         symbol->pp_definition = new_definition;
1333         return;
1334
1335 error_out:
1336         if (obstack_object_size(&pp_obstack) > 0) {
1337                 char *ptr = obstack_finish(&pp_obstack);
1338                 obstack_free(&pp_obstack, ptr);
1339         }
1340         eat_pp_directive();
1341 }
1342
1343 static void parse_undef_directive(void)
1344 {
1345         eat_pp(TP_undef);
1346         if (skip_mode) {
1347                 eat_pp_directive();
1348                 return;
1349         }
1350
1351         if (pp_token.kind != T_IDENTIFIER) {
1352                 errorf(&input.position,
1353                        "expected identifier after #undef, got %K", &pp_token);
1354                 eat_pp_directive();
1355                 return;
1356         }
1357
1358         pp_token.base.symbol->pp_definition = NULL;
1359         eat_token(T_IDENTIFIER);
1360
1361         if (!info.at_line_begin) {
1362                 warningf(WARN_OTHER, &input.position, "extra tokens at end of #undef directive");
1363         }
1364         eat_pp_directive();
1365 }
1366
1367 /** behind an #include we can have the special headername lexems.
1368  * They're only allowed behind an #include so they're not recognized
1369  * by the normal next_preprocessing_token. We handle them as a special
1370  * exception here */
1371 static void parse_headername(void)
1372 {
1373         const source_position_t start_position = input.position;
1374         string_t                string         = { NULL, 0, STRING_ENCODING_CHAR };
1375         assert(obstack_object_size(&symbol_obstack) == 0);
1376
1377         if (info.at_line_begin) {
1378                 parse_error("expected headername after #include");
1379                 goto finish_error;
1380         }
1381
1382         /* check wether we have a "... or <... headername */
1383         switch (input.c) {
1384         {
1385                 utf32 delimiter;
1386         case '<': delimiter = '>'; goto parse_name;
1387         case '"': delimiter = '"'; goto parse_name;
1388 parse_name:
1389                 next_char();
1390                 while (true) {
1391                         switch (input.c) {
1392                         case NEWLINE:
1393                         case EOF:
1394                                 errorf(&pp_token.base.source_position, "header name without closing '%c'", (char)delimiter);
1395                                 goto finish_error;
1396
1397                         default:
1398                                 if (input.c == delimiter) {
1399                                         next_char();
1400                                         goto finished_headername;
1401                                 } else {
1402                                         obstack_1grow(&symbol_obstack, (char)input.c);
1403                                         next_char();
1404                                 }
1405                                 break;
1406                         }
1407                 }
1408                 /* we should never be here */
1409         }
1410
1411         default:
1412                 /* TODO: do normal pp_token parsing and concatenate results */
1413                 panic("pp_token concat include not implemented yet");
1414         }
1415
1416 finished_headername:
1417         string = sym_make_string(STRING_ENCODING_CHAR);
1418
1419 finish_error:
1420         pp_token.base.source_position = start_position;
1421         pp_token.kind                 = T_HEADERNAME;
1422         pp_token.literal.string       = string;
1423 }
1424
1425 static bool do_include(bool system_include, const char *headername)
1426 {
1427         size_t headername_len = strlen(headername);
1428         if (!system_include) {
1429                 /* put dirname of current input on obstack */
1430                 const char *filename   = input.position.input_name;
1431                 const char *last_slash = strrchr(filename, '/');
1432                 if (last_slash != NULL) {
1433                         size_t len = last_slash - filename;
1434                         obstack_grow(&symbol_obstack, filename, len + 1);
1435                         obstack_grow0(&symbol_obstack, headername, headername_len);
1436                         char *complete_path = obstack_finish(&symbol_obstack);
1437                         headername = identify_string(complete_path);
1438                 }
1439
1440                 FILE *file = fopen(headername, "r");
1441                 if (file != NULL) {
1442                         switch_input(file, headername);
1443                         return true;
1444                 }
1445         }
1446
1447         assert(obstack_object_size(&symbol_obstack) == 0);
1448         /* check searchpath */
1449         for (searchpath_entry_t *entry = searchpath; entry != NULL;
1450              entry = entry->next) {
1451             const char *path = entry->path;
1452             size_t      len  = strlen(path);
1453                 obstack_grow(&symbol_obstack, path, len);
1454                 if (path[len-1] != '/')
1455                         obstack_1grow(&symbol_obstack, '/');
1456                 obstack_grow(&symbol_obstack, headername, headername_len+1);
1457
1458                 char *complete_path = obstack_finish(&symbol_obstack);
1459                 FILE *file          = fopen(complete_path, "r");
1460                 if (file != NULL) {
1461                         const char *filename = identify_string(complete_path);
1462                         switch_input(file, filename);
1463                         return true;
1464                 } else {
1465                         obstack_free(&symbol_obstack, complete_path);
1466                 }
1467         }
1468
1469         return false;
1470 }
1471
1472 /* read till next newline character, only for parse_include_directive(),
1473  * use eat_pp_directive() in all other cases */
1474 static void skip_till_newline(void)
1475 {
1476         /* skip till newline */
1477         while (true) {
1478                 switch (input.c) {
1479                 case NEWLINE:
1480                 case EOF:
1481                         return;
1482                 }
1483                 next_char();
1484         }
1485 }
1486
1487 static void parse_include_directive(void)
1488 {
1489         if (skip_mode) {
1490                 eat_pp_directive();
1491                 return;
1492         }
1493
1494         /* don't eat the TP_include here!
1495          * we need an alternative parsing for the next token */
1496         skip_whitespace();
1497         bool system_include = input.c == '<';
1498         parse_headername();
1499         string_t headername = pp_token.literal.string;
1500         if (headername.begin == NULL) {
1501                 eat_pp_directive();
1502                 return;
1503         }
1504
1505         skip_whitespace();
1506         if (!info.at_line_begin) {
1507                 warningf(WARN_OTHER, &pp_token.base.source_position,
1508                          "extra tokens at end of #include directive");
1509                 skip_till_newline();
1510         }
1511
1512         if (n_inputs > INCLUDE_LIMIT) {
1513                 errorf(&pp_token.base.source_position, "#include nested too deeply");
1514                 /* eat \n or EOF */
1515                 next_preprocessing_token();
1516                 return;
1517         }
1518
1519         /* switch inputs */
1520         emit_newlines();
1521         push_input();
1522         bool res = do_include(system_include, pp_token.literal.string.begin);
1523         if (!res) {
1524                 errorf(&pp_token.base.source_position, "failed including '%S': %s", &pp_token.literal, strerror(errno));
1525                 pop_restore_input();
1526         }
1527 }
1528
1529 static pp_conditional_t *push_conditional(void)
1530 {
1531         pp_conditional_t *conditional
1532                 = obstack_alloc(&pp_obstack, sizeof(*conditional));
1533         memset(conditional, 0, sizeof(*conditional));
1534
1535         conditional->parent = conditional_stack;
1536         conditional_stack   = conditional;
1537
1538         return conditional;
1539 }
1540
1541 static void pop_conditional(void)
1542 {
1543         assert(conditional_stack != NULL);
1544         conditional_stack = conditional_stack->parent;
1545 }
1546
1547 static void check_unclosed_conditionals(void)
1548 {
1549         while (conditional_stack != NULL) {
1550                 pp_conditional_t *conditional = conditional_stack;
1551
1552                 if (conditional->in_else) {
1553                         errorf(&conditional->source_position, "unterminated #else");
1554                 } else {
1555                         errorf(&conditional->source_position, "unterminated condition");
1556                 }
1557                 pop_conditional();
1558         }
1559 }
1560
1561 static void parse_ifdef_ifndef_directive(bool const is_ifdef)
1562 {
1563         bool condition;
1564         eat_pp(is_ifdef ? TP_ifdef : TP_ifndef);
1565
1566         if (skip_mode) {
1567                 eat_pp_directive();
1568                 pp_conditional_t *conditional = push_conditional();
1569                 conditional->source_position  = pp_token.base.source_position;
1570                 conditional->skip             = true;
1571                 return;
1572         }
1573
1574         if (pp_token.kind != T_IDENTIFIER || info.at_line_begin) {
1575                 errorf(&pp_token.base.source_position,
1576                        "expected identifier after #%s, got %K",
1577                        is_ifdef ? "ifdef" : "ifndef", &pp_token);
1578                 eat_pp_directive();
1579
1580                 /* just take the true case in the hope to avoid further errors */
1581                 condition = true;
1582         } else {
1583                 /* evaluate wether we are in true or false case */
1584                 condition = (bool)pp_token.base.symbol->pp_definition == is_ifdef;
1585                 eat_token(T_IDENTIFIER);
1586
1587                 if (!info.at_line_begin) {
1588                         errorf(&pp_token.base.source_position,
1589                                "extra tokens at end of #%s",
1590                                is_ifdef ? "ifdef" : "ifndef");
1591                         eat_pp_directive();
1592                 }
1593         }
1594
1595         pp_conditional_t *conditional = push_conditional();
1596         conditional->source_position  = pp_token.base.source_position;
1597         conditional->condition        = condition;
1598
1599         if (!condition) {
1600                 skip_mode = true;
1601         }
1602 }
1603
1604 static void parse_else_directive(void)
1605 {
1606         eat_pp(TP_else);
1607
1608         if (!info.at_line_begin) {
1609                 if (!skip_mode) {
1610                         warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #else");
1611                 }
1612                 eat_pp_directive();
1613         }
1614
1615         pp_conditional_t *conditional = conditional_stack;
1616         if (conditional == NULL) {
1617                 errorf(&pp_token.base.source_position, "#else without prior #if");
1618                 return;
1619         }
1620
1621         if (conditional->in_else) {
1622                 errorf(&pp_token.base.source_position,
1623                        "#else after #else (condition started %P)",
1624                        &conditional->source_position);
1625                 skip_mode = true;
1626                 return;
1627         }
1628
1629         conditional->in_else = true;
1630         if (!conditional->skip) {
1631                 skip_mode = conditional->condition;
1632         }
1633         conditional->source_position = pp_token.base.source_position;
1634 }
1635
1636 static void parse_endif_directive(void)
1637 {
1638         eat_pp(TP_endif);
1639
1640         if (!info.at_line_begin) {
1641                 if (!skip_mode) {
1642                         warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #endif");
1643                 }
1644                 eat_pp_directive();
1645         }
1646
1647         pp_conditional_t *conditional = conditional_stack;
1648         if (conditional == NULL) {
1649                 errorf(&pp_token.base.source_position, "#endif without prior #if");
1650                 return;
1651         }
1652
1653         if (!conditional->skip) {
1654                 skip_mode = false;
1655         }
1656         pop_conditional();
1657 }
1658
1659 static void parse_preprocessing_directive(void)
1660 {
1661         eat_token('#');
1662
1663         if (info.at_line_begin) {
1664                 /* empty directive */
1665                 return;
1666         }
1667
1668         if (pp_token.base.symbol) {
1669                 switch (pp_token.base.symbol->pp_ID) {
1670                 case TP_define:  parse_define_directive();            break;
1671                 case TP_else:    parse_else_directive();              break;
1672                 case TP_endif:   parse_endif_directive();             break;
1673                 case TP_ifdef:   parse_ifdef_ifndef_directive(true);  break;
1674                 case TP_ifndef:  parse_ifdef_ifndef_directive(false); break;
1675                 case TP_include: parse_include_directive();           break;
1676                 case TP_undef:   parse_undef_directive();             break;
1677                 default:         goto skip;
1678                 }
1679         } else {
1680 skip:
1681                 if (!skip_mode) {
1682                         errorf(&pp_token.base.source_position, "invalid preprocessing directive #%K", &pp_token);
1683                 }
1684                 eat_pp_directive();
1685         }
1686
1687         assert(info.at_line_begin);
1688 }
1689
1690 static void prepend_include_path(const char *path)
1691 {
1692         searchpath_entry_t *entry = OALLOCZ(&config_obstack, searchpath_entry_t);
1693         entry->path = path;
1694         entry->next = searchpath;
1695         searchpath  = entry;
1696 }
1697
1698 static void setup_include_path(void)
1699 {
1700         /* built-in paths */
1701         prepend_include_path("/usr/include");
1702
1703         /* parse environment variable */
1704         const char *cpath = getenv("CPATH");
1705         if (cpath != NULL && *cpath != '\0') {
1706                 const char *begin = cpath;
1707                 const char *c;
1708                 do {
1709                         c = begin;
1710                         while (*c != '\0' && *c != ':')
1711                                 ++c;
1712
1713                         size_t len = c-begin;
1714                         if (len == 0) {
1715                                 /* for gcc compatibility (Matze: I would expect that
1716                                  * nothing happens for an empty entry...) */
1717                                 prepend_include_path(".");
1718                         } else {
1719                                 char *string = obstack_alloc(&config_obstack, len+1);
1720                                 memcpy(string, begin, len);
1721                                 string[len] = '\0';
1722
1723                                 prepend_include_path(string);
1724                         }
1725
1726                         begin = c+1;
1727                         /* skip : */
1728                         if (*begin == ':')
1729                                 ++begin;
1730                 } while(*c != '\0');
1731         }
1732 }
1733
1734 int pptest_main(int argc, char **argv);
1735 int pptest_main(int argc, char **argv)
1736 {
1737         init_symbol_table();
1738         init_tokens();
1739
1740         obstack_init(&config_obstack);
1741         obstack_init(&pp_obstack);
1742         obstack_init(&input_obstack);
1743         strset_init(&stringset);
1744
1745         error_on_unknown_chars = false;
1746
1747         setup_include_path();
1748
1749         /* simplistic commandline parser */
1750         const char *filename = NULL;
1751         const char *output = NULL;
1752         for (int i = 1; i < argc; ++i) {
1753                 const char *opt = argv[i];
1754                 if (streq(opt, "-I")) {
1755                         prepend_include_path(argv[++i]);
1756                         continue;
1757                 } else if (streq(opt, "-E")) {
1758                         /* ignore */
1759                 } else if (streq(opt, "-o")) {
1760                         output = argv[++i];
1761                         continue;
1762                 } else if (opt[0] == '-') {
1763                         fprintf(stderr, "Unknown option '%s'\n", opt);
1764                 } else {
1765                         if (filename != NULL)
1766                                 fprintf(stderr, "Multiple inputs not supported\n");
1767                         filename = argv[i];
1768                 }
1769         }
1770         if (filename == NULL) {
1771                 fprintf(stderr, "No input specified\n");
1772                 return 1;
1773         }
1774
1775         if (output == NULL) {
1776                 out = stdout;
1777         } else {
1778                 out = fopen(output, "w");
1779                 if (out == NULL) {
1780                         fprintf(stderr, "Couldn't open output '%s'\n", output);
1781                         return 1;
1782                 }
1783         }
1784
1785         /* just here for gcc compatibility */
1786         fprintf(out, "# 1 \"%s\"\n", filename);
1787         fprintf(out, "# 1 \"<built-in>\"\n");
1788         fprintf(out, "# 1 \"<command-line>\"\n");
1789
1790         FILE *file = fopen(filename, "r");
1791         if (file == NULL) {
1792                 fprintf(stderr, "Couldn't open input '%s'\n", filename);
1793                 return 1;
1794         }
1795         switch_input(file, filename);
1796
1797         while (true) {
1798                 if (pp_token.kind == '#' && info.at_line_begin) {
1799                         parse_preprocessing_directive();
1800                         continue;
1801                 } else if (pp_token.kind == T_EOF) {
1802                         goto end_of_main_loop;
1803                 } else if (pp_token.kind == T_IDENTIFIER) {
1804                         symbol_t        *const symbol        = pp_token.base.symbol;
1805                         pp_definition_t *const pp_definition = symbol->pp_definition;
1806                         if (pp_definition != NULL && !pp_definition->is_expanding) {
1807                                 expansion_pos = pp_token.base.source_position;
1808                                 if (pp_definition->has_parameters) {
1809                                         source_position_t position = pp_token.base.source_position;
1810                                         add_token_info_t old_info = info;
1811                                         eat_token(T_IDENTIFIER);
1812                                         add_token_info_t new_info = info;
1813
1814                                         /* no opening brace -> no expansion */
1815                                         if (pp_token.kind == '(') {
1816                                                 eat_token('(');
1817
1818                                                 /* parse arguments (TODO) */
1819                                                 while (pp_token.kind != T_EOF && pp_token.kind != ')')
1820                                                         next_preprocessing_token();
1821                                         } else {
1822                                                 token_t next_token = pp_token;
1823                                                 /* restore identifier token */
1824                                                 pp_token.kind                 = T_IDENTIFIER;
1825                                                 pp_token.base.symbol          = symbol;
1826                                                 pp_token.base.source_position = position;
1827                                                 info = old_info;
1828                                                 emit_pp_token();
1829
1830                                                 info = new_info;
1831                                                 pp_token = next_token;
1832                                                 continue;
1833                                         }
1834                                         info = old_info;
1835                                 }
1836                                 pp_definition->expand_pos   = 0;
1837                                 pp_definition->is_expanding = true;
1838                                 current_expansion           = pp_definition;
1839                                 expand_next();
1840                                 continue;
1841                         }
1842                 }
1843
1844                 emit_pp_token();
1845                 next_preprocessing_token();
1846         }
1847 end_of_main_loop:
1848
1849         fputc('\n', out);
1850         check_unclosed_conditionals();
1851         close_input();
1852         if (out != stdout)
1853                 fclose(out);
1854
1855         obstack_free(&input_obstack, NULL);
1856         obstack_free(&pp_obstack, NULL);
1857         obstack_free(&config_obstack, NULL);
1858
1859         strset_destroy(&stringset);
1860
1861         exit_tokens();
1862         exit_symbol_table();
1863
1864         return 0;
1865 }