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