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(utf32 const delimiter, preprocessor_token_kind_t const kind, string_encoding_t const enc, char const *const context)
 461 {
 462         const unsigned start_linenr = input.position.lineno;
 463
 464         eat(delimiter);
 465
 466         while (true) {
 467                 switch (input.c) {
 468                 case '\\': {
 469                         if (resolve_escape_sequences) {
 470                                 utf32 const tc = parse_escape_sequence();
 471                                 if (enc == STRING_ENCODING_CHAR) {
 472                                         if (tc >= 0x100) {
 473                                                 warningf(WARN_OTHER, &pp_token.base.source_position, "escape sequence out of range");
 474                                         }
 475                                         obstack_1grow(&symbol_obstack, tc);
 476                                 } else {
 477                                         obstack_grow_symbol(&symbol_obstack, tc);
 478                                 }
 479                         } else {
 480                                 obstack_1grow(&symbol_obstack, (char)input.c);
 481                                 next_char();
 482                                 obstack_1grow(&symbol_obstack, (char)input.c);
 483                                 next_char();
 484                         }
 485                         break;
 486                 }
 487
 488                 MATCH_NEWLINE(
 489                         errorf(&pp_token.base.source_position, "newline while parsing %s", context);
 490                         break;
 491                 )
 492
 493                 case EOF: {
 494                         source_position_t source_position;
 495                         source_position.input_name = pp_token.base.source_position.input_name;
 496                         source_position.lineno     = start_linenr;
 497                         errorf(&source_position, "EOF while parsing %s", context);
 498                         goto end_of_string;
 499                 }
 500
 501                 default:
 502                         if (input.c == delimiter) {
 503                                 next_char();
 504                                 goto end_of_string;
 505                         } else {
 506                                 obstack_grow_symbol(&symbol_obstack, input.c);
 507                                 next_char();
 508                                 break;
 509                         }
 510                 }
 511         }
 512
 513 end_of_string:;
 514         obstack_1grow(&symbol_obstack, '\0');
 515         size_t const size   = obstack_object_size(&symbol_obstack) - 1;
 516         char  *const string = obstack_finish(&symbol_obstack);
 517
 518         pp_token.kind            = kind;
 519         pp_token.string.encoding = enc;
 520         pp_token.string.string   = make_string(string, size);
 521 }
 522
 523 static void parse_string_literal(string_encoding_t const enc)
 524 {
 525         parse_string('"', TP_STRING_LITERAL, enc, "string literal");
 526 }
 527
 528 static void parse_character_constant(string_encoding_t const enc)
 529 {
 530         parse_string('\'', TP_CHARACTER_CONSTANT, enc, "character constant");
 531         if (pp_token.string.string.size == 0) {
 532                 parse_error("empty character constant");
 533         }
 534 }
 535
 536 #define SYMBOL_CHARS_WITHOUT_E_P \
 537         case 'a': \
 538         case 'b': \
 539         case 'c': \
 540         case 'd': \
 541         case 'f': \
 542         case 'g': \
 543         case 'h': \
 544         case 'i': \
 545         case 'j': \
 546         case 'k': \
 547         case 'l': \
 548         case 'm': \
 549         case 'n': \
 550         case 'o': \
 551         case 'q': \
 552         case 'r': \
 553         case 's': \
 554         case 't': \
 555         case 'u': \
 556         case 'v': \
 557         case 'w': \
 558         case 'x': \
 559         case 'y': \
 560         case 'z': \
 561         case 'A': \
 562         case 'B': \
 563         case 'C': \
 564         case 'D': \
 565         case 'F': \
 566         case 'G': \
 567         case 'H': \
 568         case 'I': \
 569         case 'J': \
 570         case 'K': \
 571         case 'L': \
 572         case 'M': \
 573         case 'N': \
 574         case 'O': \
 575         case 'Q': \
 576         case 'R': \
 577         case 'S': \
 578         case 'T': \
 579         case 'U': \
 580         case 'V': \
 581         case 'W': \
 582         case 'X': \
 583         case 'Y': \
 584         case 'Z': \
 585         case '_':
 586
 587 #define SYMBOL_CHARS \
 588         SYMBOL_CHARS_WITHOUT_E_P \
 589         case 'e': \
 590         case 'p': \
 591         case 'E': \
 592         case 'P':
 593
 594 #define DIGITS \
 595         case '0':  \
 596         case '1':  \
 597         case '2':  \
 598         case '3':  \
 599         case '4':  \
 600         case '5':  \
 601         case '6':  \
 602         case '7':  \
 603         case '8':  \
 604         case '9':
 605
 606 /**
 607  * returns next final token from a preprocessor macro expansion
 608  */
 609 static void expand_next(void)
 610 {
 611         assert(current_expansion != NULL);
 612
 613         pp_definition_t *definition = current_expansion;
 614
 615 restart:
 616         if (definition->list_len == 0
 617                         || definition->expand_pos >= definition->list_len) {
 618                 /* we're finished with the current macro, move up 1 level in the
 619                  * expansion stack */
 620                 pp_definition_t *parent = definition->parent_expansion;
 621                 definition->parent_expansion = NULL;
 622                 definition->is_expanding     = false;
 623
 624                 /* it was the outermost expansion, parse normal pptoken */
 625                 if (parent == NULL) {
 626                         current_expansion = NULL;
 627                         next_preprocessing_token();
 628                         return;
 629                 }
 630                 definition        = parent;
 631                 current_expansion = definition;
 632                 goto restart;
 633         }
 634         pp_token = definition->token_list[definition->expand_pos];
 635         pp_token.base.source_position = expansion_pos;
 636         ++definition->expand_pos;
 637
 638         if (pp_token.kind != TP_IDENTIFIER)
 639                 return;
 640
 641         /* if it was an identifier then we might need to expand again */
 642         pp_definition_t *const symbol_definition = pp_token.base.symbol->pp_definition;
 643         if (symbol_definition != NULL && !symbol_definition->is_expanding) {
 644                 symbol_definition->parent_expansion = definition;
 645                 symbol_definition->expand_pos       = 0;
 646                 symbol_definition->is_expanding     = true;
 647                 definition                          = symbol_definition;
 648                 current_expansion                   = definition;
 649                 goto restart;
 650         }
 651 }
 652
 653 static void skip_line_comment(void)
 654 {
 655         while (true) {
 656                 switch (input.c) {
 657                 case EOF:
 658                         return;
 659
 660                 case '\r':
 661                 case '\n':
 662                         return;
 663
 664                 default:
 665                         next_char();
 666                         break;
 667                 }
 668         }
 669 }
 670
 671 static void skip_multiline_comment(void)
 672 {
 673         unsigned start_linenr = input.position.lineno;
 674         while (true) {
 675                 switch (input.c) {
 676                 case '/':
 677                         next_char();
 678                         if (input.c == '*') {
 679                                 /* TODO: nested comment, warn here */
 680                         }
 681                         break;
 682                 case '*':
 683                         next_char();
 684                         if (input.c == '/') {
 685                                 if (input.position.lineno != input.output_line)
 686                                         info.whitespace = input.position.colno;
 687                                 next_char();
 688                                 return;
 689                         }
 690                         break;
 691
 692                 MATCH_NEWLINE(
 693                         break;
 694                 )
 695
 696                 case EOF: {
 697                         source_position_t source_position;
 698                         source_position.input_name = pp_token.base.source_position.input_name;
 699                         source_position.lineno     = start_linenr;
 700                         errorf(&source_position, "at end of file while looking for comment end");
 701                         return;
 702                 }
 703
 704                 default:
 705                         next_char();
 706                         break;
 707                 }
 708         }
 709 }
 710
 711 static void skip_whitespace(void)
 712 {
 713         while (true) {
 714                 switch (input.c) {
 715                 case ' ':
 716                 case '\t':
 717                         next_char();
 718                         continue;
 719
 720                 MATCH_NEWLINE(
 721                         info.at_line_begin = true;
 722                         return;
 723                 )
 724
 725                 case '/':
 726                         next_char();
 727                         if (input.c == '/') {
 728                                 next_char();
 729                                 skip_line_comment();
 730                                 continue;
 731                         } else if (input.c == '*') {
 732                                 next_char();
 733                                 skip_multiline_comment();
 734                                 continue;
 735                         } else {
 736                                 put_back(input.c);
 737                                 input.c = '/';
 738                         }
 739                         return;
 740                 default:
 741                         return;
 742                 }
 743         }
 744 }
 745
 746 static void eat_pp(preprocessor_token_kind_t const type)
 747 {
 748         (void) type;
 749         assert(pp_token.kind == type);
 750         next_preprocessing_token();
 751 }
 752
 753 static void parse_symbol(void)
 754 {
 755         obstack_1grow(&symbol_obstack, (char) input.c);
 756         next_char();
 757
 758         while (true) {
 759                 switch (input.c) {
 760                 DIGITS
 761                 SYMBOL_CHARS
 762                         obstack_1grow(&symbol_obstack, (char) input.c);
 763                         next_char();
 764                         break;
 765
 766                 default:
 767                         goto end_symbol;
 768                 }
 769         }
 770
 771 end_symbol:
 772         obstack_1grow(&symbol_obstack, '\0');
 773         char *string = obstack_finish(&symbol_obstack);
 774
 775         /* might be a wide string or character constant ( L"string"/L'c' ) */
 776         if (input.c == '"' && string[0] == 'L' && string[1] == '\0') {
 777                 obstack_free(&symbol_obstack, string);
 778                 parse_string_literal(STRING_ENCODING_WIDE);
 779                 return;
 780         } else if (input.c == '\'' && string[0] == 'L' && string[1] == '\0') {
 781                 obstack_free(&symbol_obstack, string);
 782                 parse_character_constant(STRING_ENCODING_WIDE);
 783                 return;
 784         }
 785
 786         symbol_t *symbol = symbol_table_insert(string);
 787
 788         pp_token.kind        = symbol->pp_ID;
 789         pp_token.base.symbol = symbol;
 790
 791         /* we can free the memory from symbol obstack if we already had an entry in
 792          * the symbol table */
 793         if (symbol->string != string) {
 794                 obstack_free(&symbol_obstack, string);
 795         }
 796 }
 797
 798 static void parse_number(void)
 799 {
 800         obstack_1grow(&symbol_obstack, (char) input.c);
 801         next_char();
 802
 803         while (true) {
 804                 switch (input.c) {
 805                 case '.':
 806                 DIGITS
 807                 SYMBOL_CHARS_WITHOUT_E_P
 808                         obstack_1grow(&symbol_obstack, (char) input.c);
 809                         next_char();
 810                         break;
 811
 812                 case 'e':
 813                 case 'p':
 814                 case 'E':
 815                 case 'P':
 816                         obstack_1grow(&symbol_obstack, (char) input.c);
 817                         next_char();
 818                         if (input.c == '+' || input.c == '-') {
 819                                 obstack_1grow(&symbol_obstack, (char) input.c);
 820                                 next_char();
 821                         }
 822                         break;
 823
 824                 default:
 825                         goto end_number;
 826                 }
 827         }
 828
 829 end_number:
 830         obstack_1grow(&symbol_obstack, '\0');
 831         size_t  size   = obstack_object_size(&symbol_obstack);
 832         char   *string = obstack_finish(&symbol_obstack);
 833
 834         pp_token.kind          = TP_NUMBER;
 835         pp_token.number.number = make_string(string, size);
 836 }
 837
 838
 839 #define MAYBE_PROLOG                                       \
 840                         next_char();                                   \
 841                         while (true) {                                 \
 842                                 switch (input.c) {
 843
 844 #define MAYBE(ch, set_type)                                \
 845                                 case ch:                                   \
 846                                         next_char();                           \
 847                                         pp_token.kind = set_type;              \
 848                                         return;
 849
 850 #define ELSE_CODE(code)                                    \
 851                                 default:                                   \
 852                                         code                                   \
 853                                         return;                                \
 854                                 }                                          \
 855                         }
 856
 857 #define ELSE(set_type)                                     \
 858                 ELSE_CODE(                                         \
 859                         pp_token.kind = set_type;                      \
 860                 )
 861
 862 static void next_preprocessing_token(void)
 863 {
 864         if (current_expansion != NULL) {
 865                 expand_next();
 866                 return;
 867         }
 868
 869         info.at_line_begin  = false;
 870         info.had_whitespace = false;
 871 restart:
 872         pp_token.base.source_position = input.position;
 873         pp_token.base.symbol          = NULL;
 874
 875         switch (input.c) {
 876         case ' ':
 877         case '\t':
 878                 ++info.whitespace;
 879                 info.had_whitespace = true;
 880                 next_char();
 881                 goto restart;
 882
 883         MATCH_NEWLINE(
 884                 info.at_line_begin = true;
 885                 info.had_whitespace = true;
 886                 goto restart;
 887         )
 888
 889         SYMBOL_CHARS
 890                 parse_symbol();
 891                 return;
 892
 893         DIGITS
 894                 parse_number();
 895                 return;
 896
 897         case '"':
 898                 parse_string_literal(STRING_ENCODING_CHAR);
 899                 return;
 900
 901         case '\'':
 902                 parse_character_constant(STRING_ENCODING_CHAR);
 903                 return;
 904
 905         case '.':
 906                 MAYBE_PROLOG
 907                         case '0':
 908                         case '1':
 909                         case '2':
 910                         case '3':
 911                         case '4':
 912                         case '5':
 913                         case '6':
 914                         case '7':
 915                         case '8':
 916                         case '9':
 917                                 put_back(input.c);
 918                                 input.c = '.';
 919                                 parse_number();
 920                                 return;
 921
 922                         case '.':
 923                                 MAYBE_PROLOG
 924                                 MAYBE('.', TP_DOTDOTDOT)
 925                                 ELSE_CODE(
 926                                         put_back(input.c);
 927                                         input.c = '.';
 928                                         pp_token.kind = '.';
 929                                 )
 930                 ELSE('.')
 931         case '&':
 932                 MAYBE_PROLOG
 933                 MAYBE('&', TP_ANDAND)
 934                 MAYBE('=', TP_ANDEQUAL)
 935                 ELSE('&')
 936         case '*':
 937                 MAYBE_PROLOG
 938                 MAYBE('=', TP_ASTERISKEQUAL)
 939                 ELSE('*')
 940         case '+':
 941                 MAYBE_PROLOG
 942                 MAYBE('+', TP_PLUSPLUS)
 943                 MAYBE('=', TP_PLUSEQUAL)
 944                 ELSE('+')
 945         case '-':
 946                 MAYBE_PROLOG
 947                 MAYBE('>', TP_MINUSGREATER)
 948                 MAYBE('-', TP_MINUSMINUS)
 949                 MAYBE('=', TP_MINUSEQUAL)
 950                 ELSE('-')
 951         case '!':
 952                 MAYBE_PROLOG
 953                 MAYBE('=', TP_EXCLAMATIONMARKEQUAL)
 954                 ELSE('!')
 955         case '/':
 956                 MAYBE_PROLOG
 957                 MAYBE('=', TP_SLASHEQUAL)
 958                         case '*':
 959                                 next_char();
 960                                 info.had_whitespace = true;
 961                                 skip_multiline_comment();
 962                                 goto restart;
 963                         case '/':
 964                                 next_char();
 965                                 info.had_whitespace = true;
 966                                 skip_line_comment();
 967                                 goto restart;
 968                 ELSE('/')
 969         case '%':
 970                 MAYBE_PROLOG
 971                 MAYBE('>', '}')
 972                 MAYBE('=', TP_PERCENTEQUAL)
 973                         case ':':
 974                                 MAYBE_PROLOG
 975                                         case '%':
 976                                                 MAYBE_PROLOG
 977                                                 MAYBE(':', TP_HASHHASH)
 978                                                 ELSE_CODE(
 979                                                         put_back(input.c);
 980                                                         input.c = '%';
 981                                                         pp_token.kind = '#';
 982                                                 )
 983                                 ELSE('#')
 984                 ELSE('%')
 985         case '<':
 986                 MAYBE_PROLOG
 987                 MAYBE(':', '[')
 988                 MAYBE('%', '{')
 989                 MAYBE('=', TP_LESSEQUAL)
 990                         case '<':
 991                                 MAYBE_PROLOG
 992                                 MAYBE('=', TP_LESSLESSEQUAL)
 993                                 ELSE(TP_LESSLESS)
 994                 ELSE('<')
 995         case '>':
 996                 MAYBE_PROLOG
 997                 MAYBE('=', TP_GREATEREQUAL)
 998                         case '>':
 999                                 MAYBE_PROLOG
1000                                 MAYBE('=', TP_GREATERGREATEREQUAL)
1001                                 ELSE(TP_GREATERGREATER)
1002                 ELSE('>')
1003         case '^':
1004                 MAYBE_PROLOG
1005                 MAYBE('=', TP_CARETEQUAL)
1006                 ELSE('^')
1007         case '|':
1008                 MAYBE_PROLOG
1009                 MAYBE('=', TP_PIPEEQUAL)
1010                 MAYBE('|', TP_PIPEPIPE)
1011                 ELSE('|')
1012         case ':':
1013                 MAYBE_PROLOG
1014                 MAYBE('>', ']')
1015                 ELSE(':')
1016         case '=':
1017                 MAYBE_PROLOG
1018                 MAYBE('=', TP_EQUALEQUAL)
1019                 ELSE('=')
1020         case '#':
1021                 MAYBE_PROLOG
1022                 MAYBE('#', TP_HASHHASH)
1023                 ELSE_CODE(
1024                         pp_token.kind = '#';
1025                 )
1026
1027         case '?':
1028         case '[':
1029         case ']':
1030         case '(':
1031         case ')':
1032         case '{':
1033         case '}':
1034         case '~':
1035         case ';':
1036         case ',':
1037         case '\\':
1038                 pp_token.kind = input.c;
1039                 next_char();
1040                 return;
1041
1042         case EOF:
1043                 if (input_stack != NULL) {
1044                         close_input();
1045                         pop_restore_input();
1046                         fputc('\n', out);
1047                         print_line_directive(&input.position, "2");
1048                         goto restart;
1049                 } else {
1050                         pp_token.base.source_position.lineno++;
1051                         info.at_line_begin = true;
1052                         pp_token.kind = TP_EOF;
1053                 }
1054                 return;
1055
1056         default:
1057                 next_char();
1058                 if (!ignore_unknown_chars) {
1059                         errorf(&pp_token.base.source_position,
1060                                "unknown character '%c' found\n", input.c);
1061                         goto restart;
1062                 } else {
1063                         pp_token.kind = input.c;
1064                         return;
1065                 }
1066         }
1067 }
1068
1069 static void print_quoted_string(const char *const string)
1070 {
1071         fputc('"', out);
1072         for (const char *c = string; *c != 0; ++c) {
1073                 switch (*c) {
1074                 case '"': fputs("\\\"", out); break;
1075                 case '\\':  fputs("\\\\", out); break;
1076                 case '\a':  fputs("\\a", out); break;
1077                 case '\b':  fputs("\\b", out); break;
1078                 case '\f':  fputs("\\f", out); break;
1079                 case '\n':  fputs("\\n", out); break;
1080                 case '\r':  fputs("\\r", out); break;
1081                 case '\t':  fputs("\\t", out); break;
1082                 case '\v':  fputs("\\v", out); break;
1083                 case '\?':  fputs("\\?", out); break;
1084                 default:
1085                         if (!isprint(*c)) {
1086                                 fprintf(out, "\\%03o", (unsigned)*c);
1087                                 break;
1088                         }
1089                         fputc(*c, out);
1090                         break;
1091                 }
1092         }
1093         fputc('"', out);
1094 }
1095
1096 static void print_line_directive(const source_position_t *pos, const char *add)
1097 {
1098         fprintf(out, "# %u ", pos->lineno);
1099         print_quoted_string(pos->input_name);
1100         if (add != NULL) {
1101                 fputc(' ', out);
1102                 fputs(add, out);
1103         }
1104
1105         printed_input_name = pos->input_name;
1106         input.output_line  = pos->lineno-1;
1107 }
1108
1109 static bool emit_newlines(void)
1110 {
1111         unsigned delta = pp_token.base.source_position.lineno - input.output_line;
1112         if (delta == 0)
1113                 return false;
1114
1115         if (delta >= 9) {
1116                 fputc('\n', out);
1117                 print_line_directive(&pp_token.base.source_position, NULL);
1118                 fputc('\n', out);
1119         } else {
1120                 for (unsigned i = 0; i < delta; ++i) {
1121                         fputc('\n', out);
1122                 }
1123         }
1124         input.output_line = pp_token.base.source_position.lineno;
1125
1126         for (unsigned i = 0; i < info.whitespace; ++i)
1127                 fputc(' ', out);
1128
1129         return true;
1130 }
1131
1132 static void emit_pp_token(void)
1133 {
1134         if (skip_mode)
1135                 return;
1136
1137         if (!emit_newlines() &&
1138             (info.had_whitespace || tokens_would_paste(last_token, pp_token.kind)))
1139                 fputc(' ', out);
1140
1141         switch (pp_token.kind) {
1142         case TP_IDENTIFIER:
1143                 fputs(pp_token.base.symbol->string, out);
1144                 break;
1145         case TP_NUMBER:
1146                 fputs(pp_token.number.number.begin, out);
1147                 break;
1148
1149         case TP_STRING_LITERAL:
1150                 fputs(get_string_encoding_prefix(pp_token.string.encoding), out);
1151                 fputc('"', out);
1152                 fputs(pp_token.string.string.begin, out);
1153                 fputc('"', out);
1154                 break;
1155
1156         case TP_CHARACTER_CONSTANT:
1157                 fputs(get_string_encoding_prefix(pp_token.string.encoding), out);
1158                 fputc('\'', out);
1159                 fputs(pp_token.string.string.begin, out);
1160                 fputc('\'', out);
1161                 break;
1162         default:
1163                 print_pp_token_kind(out, pp_token.kind);
1164                 break;
1165         }
1166         last_token = pp_token.kind;
1167 }
1168
1169 static void eat_pp_directive(void)
1170 {
1171         while (!info.at_line_begin) {
1172                 next_preprocessing_token();
1173         }
1174 }
1175
1176 static bool strings_equal(const string_t *string1, const string_t *string2)
1177 {
1178         size_t size = string1->size;
1179         if (size != string2->size)
1180                 return false;
1181
1182         const char *c1 = string1->begin;
1183         const char *c2 = string2->begin;
1184         for (size_t i = 0; i < size; ++i, ++c1, ++c2) {
1185                 if (*c1 != *c2)
1186                         return false;
1187         }
1188         return true;
1189 }
1190
1191 static bool pp_tokens_equal(const token_t *token1, const token_t *token2)
1192 {
1193         if (token1->kind != token2->kind)
1194                 return false;
1195
1196         switch (token1->kind) {
1197         case TP_IDENTIFIER:
1198                 return token1->base.symbol == token2->base.symbol;
1199
1200         case TP_NUMBER:
1201         case TP_CHARACTER_CONSTANT:
1202         case TP_STRING_LITERAL:
1203                 return strings_equal(&token1->string.string, &token2->string.string);
1204
1205         default:
1206                 return true;
1207         }
1208 }
1209
1210 static bool pp_definitions_equal(const pp_definition_t *definition1,
1211                                  const pp_definition_t *definition2)
1212 {
1213         if (definition1->list_len != definition2->list_len)
1214                 return false;
1215
1216         size_t         len = definition1->list_len;
1217         const token_t *t1  = definition1->token_list;
1218         const token_t *t2  = definition2->token_list;
1219         for (size_t i = 0; i < len; ++i, ++t1, ++t2) {
1220                 if (!pp_tokens_equal(t1, t2))
1221                         return false;
1222         }
1223         return true;
1224 }
1225
1226 static void parse_define_directive(void)
1227 {
1228         eat_pp(TP_define);
1229         assert(obstack_object_size(&pp_obstack) == 0);
1230
1231         if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1232                 errorf(&pp_token.base.source_position,
1233                        "expected identifier after #define, got '%t'", &pp_token);
1234                 goto error_out;
1235         }
1236         symbol_t *const symbol = pp_token.base.symbol;
1237
1238         pp_definition_t *new_definition
1239                 = obstack_alloc(&pp_obstack, sizeof(new_definition[0]));
1240         memset(new_definition, 0, sizeof(new_definition[0]));
1241         new_definition->source_position = input.position;
1242
1243         /* this is probably the only place where spaces are significant in the
1244          * lexer (except for the fact that they separate tokens). #define b(x)
1245          * is something else than #define b (x) */
1246         if (input.c == '(') {
1247                 /* eat the '(' */
1248                 next_preprocessing_token();
1249                 /* get next token after '(' */
1250                 next_preprocessing_token();
1251
1252                 while (true) {
1253                         switch (pp_token.kind) {
1254                         case TP_DOTDOTDOT:
1255                                 new_definition->is_variadic = true;
1256                                 next_preprocessing_token();
1257                                 if (pp_token.kind != ')') {
1258                                         errorf(&input.position,
1259                                                         "'...' not at end of macro argument list");
1260                                         goto error_out;
1261                                 }
1262                                 break;
1263                         case TP_IDENTIFIER:
1264                                 obstack_ptr_grow(&pp_obstack, pp_token.base.symbol);
1265                                 next_preprocessing_token();
1266
1267                                 if (pp_token.kind == ',') {
1268                                         next_preprocessing_token();
1269                                         break;
1270                                 }
1271
1272                                 if (pp_token.kind != ')') {
1273                                         errorf(&pp_token.base.source_position,
1274                                                "expected ',' or ')' after identifier, got '%t'",
1275                                                &pp_token);
1276                                         goto error_out;
1277                                 }
1278                                 break;
1279                         case ')':
1280                                 next_preprocessing_token();
1281                                 goto finish_argument_list;
1282                         default:
1283                                 errorf(&pp_token.base.source_position,
1284                                        "expected identifier, '...' or ')' in #define argument list, got '%t'",
1285                                        &pp_token);
1286                                 goto error_out;
1287                         }
1288                 }
1289
1290         finish_argument_list:
1291                 new_definition->has_parameters = true;
1292                 new_definition->n_parameters
1293                         = obstack_object_size(&pp_obstack) / sizeof(new_definition->parameters[0]);
1294                 new_definition->parameters = obstack_finish(&pp_obstack);
1295         } else {
1296                 next_preprocessing_token();
1297         }
1298
1299         /* construct a new pp_definition on the obstack */
1300         assert(obstack_object_size(&pp_obstack) == 0);
1301         size_t list_len = 0;
1302         while (!info.at_line_begin) {
1303                 obstack_grow(&pp_obstack, &pp_token, sizeof(pp_token));
1304                 ++list_len;
1305                 next_preprocessing_token();
1306         }
1307
1308         new_definition->list_len   = list_len;
1309         new_definition->token_list = obstack_finish(&pp_obstack);
1310
1311         pp_definition_t *old_definition = symbol->pp_definition;
1312         if (old_definition != NULL) {
1313                 if (!pp_definitions_equal(old_definition, new_definition)) {
1314                         warningf(WARN_OTHER, &input.position, "multiple definition of macro '%Y' (first defined %P)", symbol, &old_definition->source_position);
1315                 } else {
1316                         /* reuse the old definition */
1317                         obstack_free(&pp_obstack, new_definition);
1318                         new_definition = old_definition;
1319                 }
1320         }
1321
1322         symbol->pp_definition = new_definition;
1323         return;
1324
1325 error_out:
1326         if (obstack_object_size(&pp_obstack) > 0) {
1327                 char *ptr = obstack_finish(&pp_obstack);
1328                 obstack_free(&pp_obstack, ptr);
1329         }
1330         eat_pp_directive();
1331 }
1332
1333 static void parse_undef_directive(void)
1334 {
1335         eat_pp(TP_undef);
1336
1337         if (pp_token.kind != TP_IDENTIFIER) {
1338                 errorf(&input.position,
1339                        "expected identifier after #undef, got '%t'", &pp_token);
1340                 eat_pp_directive();
1341                 return;
1342         }
1343
1344         pp_token.base.symbol->pp_definition = NULL;
1345         next_preprocessing_token();
1346
1347         if (!info.at_line_begin) {
1348                 warningf(WARN_OTHER, &input.position, "extra tokens at end of #undef directive");
1349         }
1350         eat_pp_directive();
1351 }
1352
1353 static void parse_headername(void)
1354 {
1355         const source_position_t start_position = input.position;
1356         string_t                string         = {NULL, 0};
1357         assert(obstack_object_size(&symbol_obstack) == 0);
1358
1359         /* behind an #include we can have the special headername lexems.
1360          * They're only allowed behind an #include so they're not recognized
1361          * by the normal next_preprocessing_token. We handle them as a special
1362          * exception here */
1363         if (info.at_line_begin) {
1364                 parse_error("expected headername after #include");
1365                 goto finish_error;
1366         }
1367
1368         /* check wether we have a "... or <... headername */
1369         switch (input.c) {
1370         {
1371                 utf32 delimiter;
1372         case '<': delimiter = '>'; goto parse_name;
1373         case '"': delimiter = '"'; goto parse_name;
1374 parse_name:
1375                 next_char();
1376                 while (true) {
1377                         switch (input.c) {
1378                         case EOF:
1379                                 /* fallthrough */
1380                         MATCH_NEWLINE(
1381                                 errorf(&pp_token.base.source_position, "header name without closing '%c'", (char)delimiter);
1382                                 goto finish_error;
1383                         )
1384
1385                         default:
1386                                 if (input.c == delimiter) {
1387                                         next_char();
1388                                         goto finished_headername;
1389                                 } else {
1390                                         obstack_1grow(&symbol_obstack, (char)input.c);
1391                                         next_char();
1392                                 }
1393                                 break;
1394                         }
1395                 }
1396                 /* we should never be here */
1397         }
1398
1399         default:
1400                 /* TODO: do normal pp_token parsing and concatenate results */
1401                 panic("pp_token concat include not implemented yet");
1402         }
1403
1404 finished_headername:
1405         obstack_1grow(&symbol_obstack, '\0');
1406         const size_t size       = (size_t)obstack_object_size(&symbol_obstack);
1407         char *const  headername = obstack_finish(&symbol_obstack);
1408         string                  = make_string(headername, size);
1409
1410 finish_error:
1411         pp_token.base.source_position = start_position;
1412         pp_token.kind                 = TP_HEADERNAME;
1413         pp_token.string.string        = string;
1414 }
1415
1416 static bool do_include(bool system_include, const char *headername)
1417 {
1418         size_t headername_len = strlen(headername);
1419         if (!system_include) {
1420                 /* put dirname of current input on obstack */
1421                 const char *filename   = input.position.input_name;
1422                 const char *last_slash = strrchr(filename, '/');
1423                 if (last_slash != NULL) {
1424                         size_t len = last_slash - filename;
1425                         obstack_grow(&symbol_obstack, filename, len + 1);
1426                         obstack_grow0(&symbol_obstack, headername, headername_len);
1427                         char *complete_path = obstack_finish(&symbol_obstack);
1428                         headername = identify_string(complete_path);
1429                 }
1430
1431                 FILE *file = fopen(headername, "r");
1432                 if (file != NULL) {
1433                         switch_input(file, headername);
1434                         return true;
1435                 }
1436         }
1437
1438         assert(obstack_object_size(&symbol_obstack) == 0);
1439         /* check searchpath */
1440         for (searchpath_entry_t *entry = searchpath; entry != NULL;
1441              entry = entry->next) {
1442             const char *path = entry->path;
1443             size_t      len  = strlen(path);
1444                 obstack_grow(&symbol_obstack, path, len);
1445                 if (path[len-1] != '/')
1446                         obstack_1grow(&symbol_obstack, '/');
1447                 obstack_grow(&symbol_obstack, headername, headername_len+1);
1448
1449                 char *complete_path = obstack_finish(&symbol_obstack);
1450                 FILE *file          = fopen(complete_path, "r");
1451                 if (file != NULL) {
1452                         const char *filename = identify_string(complete_path);
1453                         switch_input(file, filename);
1454                         return true;
1455                 } else {
1456                         obstack_free(&symbol_obstack, complete_path);
1457                 }
1458         }
1459
1460         return false;
1461 }
1462
1463 /* read till next newline character, only for parse_include_directive(),
1464  * use eat_pp_directive() in all other cases */
1465 static void skip_till_newline(void)
1466 {
1467         /* skip till newline */
1468         while (true) {
1469                 switch (input.c) {
1470                 MATCH_NEWLINE(
1471                         return;
1472                 )
1473                 case EOF:
1474                         return;
1475                 }
1476                 next_char();
1477         }
1478 }
1479
1480 static bool parse_include_directive(void)
1481 {
1482         /* don't eat the TP_include here!
1483          * we need an alternative parsing for the next token */
1484         skip_whitespace();
1485         bool system_include = input.c == '<';
1486         parse_headername();
1487         string_t headername = pp_token.string.string;
1488         if (headername.begin == NULL) {
1489                 eat_pp_directive();
1490                 return false;
1491         }
1492
1493         skip_whitespace();
1494         if (!info.at_line_begin) {
1495                 warningf(WARN_OTHER, &pp_token.base.source_position,
1496                          "extra tokens at end of #include directive");
1497                 skip_till_newline();
1498         }
1499
1500         if (n_inputs > INCLUDE_LIMIT) {
1501                 errorf(&pp_token.base.source_position, "#include nested too deeply");
1502                 /* eat \n or EOF */
1503                 next_preprocessing_token();
1504                 return false;
1505         }
1506
1507         /* switch inputs */
1508         emit_newlines();
1509         push_input();
1510         bool res = do_include(system_include, pp_token.string.string.begin);
1511         if (!res) {
1512                 errorf(&pp_token.base.source_position,
1513                        "failed including '%S': %s", pp_token.string, strerror(errno));
1514                 pop_restore_input();
1515                 return false;
1516         }
1517
1518         return true;
1519 }
1520
1521 static pp_conditional_t *push_conditional(void)
1522 {
1523         pp_conditional_t *conditional
1524                 = obstack_alloc(&pp_obstack, sizeof(*conditional));
1525         memset(conditional, 0, sizeof(*conditional));
1526
1527         conditional->parent = conditional_stack;
1528         conditional_stack   = conditional;
1529
1530         return conditional;
1531 }
1532
1533 static void pop_conditional(void)
1534 {
1535         assert(conditional_stack != NULL);
1536         conditional_stack = conditional_stack->parent;
1537 }
1538
1539 static void check_unclosed_conditionals(void)
1540 {
1541         while (conditional_stack != NULL) {
1542                 pp_conditional_t *conditional = conditional_stack;
1543
1544                 if (conditional->in_else) {
1545                         errorf(&conditional->source_position, "unterminated #else");
1546                 } else {
1547                         errorf(&conditional->source_position, "unterminated condition");
1548                 }
1549                 pop_conditional();
1550         }
1551 }
1552
1553 static void parse_ifdef_ifndef_directive(void)
1554 {
1555         bool is_ifndef = (pp_token.kind == TP_ifndef);
1556         bool condition;
1557         next_preprocessing_token();
1558
1559         if (skip_mode) {
1560                 eat_pp_directive();
1561                 pp_conditional_t *conditional = push_conditional();
1562                 conditional->source_position  = pp_token.base.source_position;
1563                 conditional->skip             = true;
1564                 return;
1565         }
1566
1567         if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1568                 errorf(&pp_token.base.source_position,
1569                        "expected identifier after #%s, got '%t'",
1570                        is_ifndef ? "ifndef" : "ifdef", &pp_token);
1571                 eat_pp_directive();
1572
1573                 /* just take the true case in the hope to avoid further errors */
1574                 condition = true;
1575         } else {
1576                 /* evaluate wether we are in true or false case */
1577                 condition = !pp_token.base.symbol->pp_definition == is_ifndef;
1578
1579                 next_preprocessing_token();
1580
1581                 if (!info.at_line_begin) {
1582                         errorf(&pp_token.base.source_position,
1583                                "extra tokens at end of #%s",
1584                                is_ifndef ? "ifndef" : "ifdef");
1585                         eat_pp_directive();
1586                 }
1587         }
1588
1589         pp_conditional_t *conditional = push_conditional();
1590         conditional->source_position  = pp_token.base.source_position;
1591         conditional->condition        = condition;
1592
1593         if (!condition) {
1594                 skip_mode = true;
1595         }
1596 }
1597
1598 static void parse_else_directive(void)
1599 {
1600         eat_pp(TP_else);
1601
1602         if (!info.at_line_begin) {
1603                 if (!skip_mode) {
1604                         warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #else");
1605                 }
1606                 eat_pp_directive();
1607         }
1608
1609         pp_conditional_t *conditional = conditional_stack;
1610         if (conditional == NULL) {
1611                 errorf(&pp_token.base.source_position, "#else without prior #if");
1612                 return;
1613         }
1614
1615         if (conditional->in_else) {
1616                 errorf(&pp_token.base.source_position,
1617                        "#else after #else (condition started %P)",
1618                        conditional->source_position);
1619                 skip_mode = true;
1620                 return;
1621         }
1622
1623         conditional->in_else = true;
1624         if (!conditional->skip) {
1625                 skip_mode = conditional->condition;
1626         }
1627         conditional->source_position = pp_token.base.source_position;
1628 }
1629
1630 static void parse_endif_directive(void)
1631 {
1632         eat_pp(TP_endif);
1633
1634         if (!info.at_line_begin) {
1635                 if (!skip_mode) {
1636                         warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #endif");
1637                 }
1638                 eat_pp_directive();
1639         }
1640
1641         pp_conditional_t *conditional = conditional_stack;
1642         if (conditional == NULL) {
1643                 errorf(&pp_token.base.source_position, "#endif without prior #if");
1644                 return;
1645         }
1646
1647         if (!conditional->skip) {
1648                 skip_mode = false;
1649         }
1650         pop_conditional();
1651 }
1652
1653 static void parse_preprocessing_directive(void)
1654 {
1655         eat_pp('#');
1656
1657         if (info.at_line_begin) {
1658                 /* empty directive */
1659                 return;
1660         }
1661
1662         if (skip_mode) {
1663                 switch (pp_token.kind) {
1664                 case TP_ifdef:
1665                 case TP_ifndef:
1666                         parse_ifdef_ifndef_directive();
1667                         break;
1668                 case TP_else:
1669                         parse_else_directive();
1670                         break;
1671                 case TP_endif:
1672                         parse_endif_directive();
1673                         break;
1674                 default:
1675                         eat_pp_directive();
1676                         break;
1677                 }
1678         } else {
1679                 switch (pp_token.kind) {
1680                 case TP_define:
1681                         parse_define_directive();
1682                         break;
1683                 case TP_undef:
1684                         parse_undef_directive();
1685                         break;
1686                 case TP_ifdef:
1687                 case TP_ifndef:
1688                         parse_ifdef_ifndef_directive();
1689                         break;
1690                 case TP_else:
1691                         parse_else_directive();
1692                         break;
1693                 case TP_endif:
1694                         parse_endif_directive();
1695                         break;
1696                 case TP_include:
1697                         parse_include_directive();
1698                         break;
1699                 default:
1700                         if (info.at_line_begin) {
1701                                 /* the nop directive "#" */
1702                                 break;
1703                         }
1704                         errorf(&pp_token.base.source_position,
1705                                    "invalid preprocessing directive #%t", &pp_token);
1706                         eat_pp_directive();
1707                         break;
1708                 }
1709         }
1710
1711         assert(info.at_line_begin);
1712 }
1713
1714 static void prepend_include_path(const char *path)
1715 {
1716         searchpath_entry_t *entry = OALLOCZ(&config_obstack, searchpath_entry_t);
1717         entry->path = path;
1718         entry->next = searchpath;
1719         searchpath  = entry;
1720 }
1721
1722 static void setup_include_path(void)
1723 {
1724         /* built-in paths */
1725         prepend_include_path("/usr/include");
1726
1727         /* parse environment variable */
1728         const char *cpath = getenv("CPATH");
1729         if (cpath != NULL && *cpath != '\0') {
1730                 const char *begin = cpath;
1731                 const char *c;
1732                 do {
1733                         c = begin;
1734                         while (*c != '\0' && *c != ':')
1735                                 ++c;
1736
1737                         size_t len = c-begin;
1738                         if (len == 0) {
1739                                 /* for gcc compatibility (Matze: I would expect that
1740                                  * nothing happens for an empty entry...) */
1741                                 prepend_include_path(".");
1742                         } else {
1743                                 char *string = obstack_alloc(&config_obstack, len+1);
1744                                 memcpy(string, begin, len);
1745                                 string[len] = '\0';
1746
1747                                 prepend_include_path(string);
1748                         }
1749
1750                         begin = c+1;
1751                         /* skip : */
1752                         if (*begin == ':')
1753                                 ++begin;
1754                 } while(*c != '\0');
1755         }
1756 }
1757
1758 int pptest_main(int argc, char **argv);
1759 int pptest_main(int argc, char **argv)
1760 {
1761         init_symbol_table();
1762         init_tokens();
1763
1764         obstack_init(&config_obstack);
1765         obstack_init(&pp_obstack);
1766         obstack_init(&input_obstack);
1767         strset_init(&stringset);
1768
1769         setup_include_path();
1770
1771         /* simplistic commandline parser */
1772         const char *filename = NULL;
1773         const char *output = NULL;
1774         for (int i = 1; i < argc; ++i) {
1775                 const char *opt = argv[i];
1776                 if (streq(opt, "-I")) {
1777                         prepend_include_path(argv[++i]);
1778                         continue;
1779                 } else if (streq(opt, "-E")) {
1780                         /* ignore */
1781                 } else if (streq(opt, "-o")) {
1782                         output = argv[++i];
1783                         continue;
1784                 } else if (opt[0] == '-') {
1785                         fprintf(stderr, "Unknown option '%s'\n", opt);
1786                 } else {
1787                         if (filename != NULL)
1788                                 fprintf(stderr, "Multiple inputs not supported\n");
1789                         filename = argv[i];
1790                 }
1791         }
1792         if (filename == NULL) {
1793                 fprintf(stderr, "No input specified\n");
1794                 return 1;
1795         }
1796
1797         if (output == NULL) {
1798                 out = stdout;
1799         } else {
1800                 out = fopen(output, "w");
1801                 if (out == NULL) {
1802                         fprintf(stderr, "Couldn't open output '%s'\n", output);
1803                         return 1;
1804                 }
1805         }
1806
1807         /* just here for gcc compatibility */
1808         fprintf(out, "# 1 \"%s\"\n", filename);
1809         fprintf(out, "# 1 \"<built-in>\"\n");
1810         fprintf(out, "# 1 \"<command-line>\"\n");
1811
1812         FILE *file = fopen(filename, "r");
1813         if (file == NULL) {
1814                 fprintf(stderr, "Couldn't open input '%s'\n", filename);
1815                 return 1;
1816         }
1817         switch_input(file, filename);
1818
1819         while (true) {
1820                 if (pp_token.kind == '#' && info.at_line_begin) {
1821                         parse_preprocessing_directive();
1822                         continue;
1823                 } else if (pp_token.kind == TP_EOF) {
1824                         goto end_of_main_loop;
1825                 } else if (pp_token.kind == TP_IDENTIFIER) {
1826                         symbol_t        *const symbol        = pp_token.base.symbol;
1827                         pp_definition_t *const pp_definition = symbol->pp_definition;
1828                         if (pp_definition != NULL && !pp_definition->is_expanding) {
1829                                 expansion_pos = pp_token.base.source_position;
1830                                 if (pp_definition->has_parameters) {
1831                                         source_position_t position = pp_token.base.source_position;
1832                                         add_token_info_t old_info = info;
1833                                         next_preprocessing_token();
1834                                         add_token_info_t new_info = info;
1835
1836                                         /* no opening brace -> no expansion */
1837                                         if (pp_token.kind == '(') {
1838                                                 eat_pp('(');
1839
1840                                                 /* parse arguments (TODO) */
1841                                                 while (pp_token.kind != TP_EOF && pp_token.kind != ')')
1842                                                         next_preprocessing_token();
1843                                         } else {
1844                                                 token_t next_token = pp_token;
1845                                                 /* restore identifier token */
1846                                                 pp_token.kind                 = TP_IDENTIFIER;
1847                                                 pp_token.base.symbol          = symbol;
1848                                                 pp_token.base.source_position = position;
1849                                                 info = old_info;
1850                                                 emit_pp_token();
1851
1852                                                 info = new_info;
1853                                                 pp_token = next_token;
1854                                                 continue;
1855                                         }
1856                                         info = old_info;
1857                                 }
1858                                 pp_definition->expand_pos   = 0;
1859                                 pp_definition->is_expanding = true;
1860                                 current_expansion           = pp_definition;
1861                                 expand_next();
1862                                 continue;
1863                         }
1864                 }
1865
1866                 emit_pp_token();
1867                 next_preprocessing_token();
1868         }
1869 end_of_main_loop:
1870
1871         fputc('\n', out);
1872         check_unclosed_conditionals();
1873         close_input();
1874         if (out != stdout)
1875                 fclose(out);
1876
1877         obstack_free(&input_obstack, NULL);
1878         obstack_free(&pp_obstack, NULL);
1879         obstack_free(&config_obstack, NULL);
1880
1881         strset_destroy(&stringset);
1882
1883         exit_tokens();
1884         exit_symbol_table();
1885
1886         return 0;
1887 }