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 sym_make_string(string_encoding_t const enc)
 455 {
 456         obstack_1grow(&symbol_obstack, '\0');
 457         size_t      const len    = obstack_object_size(&symbol_obstack) - 1;
 458         char       *const string = obstack_finish(&symbol_obstack);
 459         char const *const result = identify_string(string);
 460         return (string_t){ result, len, enc };
 461 }
 462
 463 static void parse_string(utf32 const delimiter, preprocessor_token_kind_t const kind, string_encoding_t const enc, char const *const context)
 464 {
 465         const unsigned start_linenr = input.position.lineno;
 466
 467         eat(delimiter);
 468
 469         while (true) {
 470                 switch (input.c) {
 471                 case '\\': {
 472                         if (resolve_escape_sequences) {
 473                                 utf32 const tc = parse_escape_sequence();
 474                                 if (enc == STRING_ENCODING_CHAR) {
 475                                         if (tc >= 0x100) {
 476                                                 warningf(WARN_OTHER, &pp_token.base.source_position, "escape sequence out of range");
 477                                         }
 478                                         obstack_1grow(&symbol_obstack, tc);
 479                                 } else {
 480                                         obstack_grow_symbol(&symbol_obstack, tc);
 481                                 }
 482                         } else {
 483                                 obstack_1grow(&symbol_obstack, (char)input.c);
 484                                 next_char();
 485                                 obstack_1grow(&symbol_obstack, (char)input.c);
 486                                 next_char();
 487                         }
 488                         break;
 489                 }
 490
 491                 case NEWLINE:
 492                         errorf(&pp_token.base.source_position, "newline while parsing %s", context);
 493                         break;
 494
 495                 case EOF: {
 496                         source_position_t source_position;
 497                         source_position.input_name = pp_token.base.source_position.input_name;
 498                         source_position.lineno     = start_linenr;
 499                         errorf(&source_position, "EOF while parsing %s", context);
 500                         goto end_of_string;
 501                 }
 502
 503                 default:
 504                         if (input.c == delimiter) {
 505                                 next_char();
 506                                 goto end_of_string;
 507                         } else {
 508                                 obstack_grow_symbol(&symbol_obstack, input.c);
 509                                 next_char();
 510                                 break;
 511                         }
 512                 }
 513         }
 514
 515 end_of_string:
 516         pp_token.kind           = kind;
 517         pp_token.literal.string = sym_make_string(enc);
 518 }
 519
 520 static void parse_string_literal(string_encoding_t const enc)
 521 {
 522         parse_string('"', TP_STRING_LITERAL, enc, "string literal");
 523 }
 524
 525 static void parse_character_constant(string_encoding_t const enc)
 526 {
 527         parse_string('\'', TP_CHARACTER_CONSTANT, enc, "character constant");
 528         if (pp_token.literal.string.size == 0) {
 529                 parse_error("empty character constant");
 530         }
 531 }
 532
 533 #define SYMBOL_CASES_WITHOUT_E_P \
 534              'a': \
 535         case 'b': \
 536         case 'c': \
 537         case 'd': \
 538         case 'f': \
 539         case 'g': \
 540         case 'h': \
 541         case 'i': \
 542         case 'j': \
 543         case 'k': \
 544         case 'l': \
 545         case 'm': \
 546         case 'n': \
 547         case 'o': \
 548         case 'q': \
 549         case 'r': \
 550         case 's': \
 551         case 't': \
 552         case 'u': \
 553         case 'v': \
 554         case 'w': \
 555         case 'x': \
 556         case 'y': \
 557         case 'z': \
 558         case 'A': \
 559         case 'B': \
 560         case 'C': \
 561         case 'D': \
 562         case 'F': \
 563         case 'G': \
 564         case 'H': \
 565         case 'I': \
 566         case 'J': \
 567         case 'K': \
 568         case 'L': \
 569         case 'M': \
 570         case 'N': \
 571         case 'O': \
 572         case 'Q': \
 573         case 'R': \
 574         case 'S': \
 575         case 'T': \
 576         case 'U': \
 577         case 'V': \
 578         case 'W': \
 579         case 'X': \
 580         case 'Y': \
 581         case 'Z': \
 582         case '_'
 583
 584 #define SYMBOL_CASES \
 585              SYMBOL_CASES_WITHOUT_E_P: \
 586         case 'e': \
 587         case 'p': \
 588         case 'E': \
 589         case 'P'
 590
 591 #define DIGIT_CASES \
 592              '0':  \
 593         case '1':  \
 594         case '2':  \
 595         case '3':  \
 596         case '4':  \
 597         case '5':  \
 598         case '6':  \
 599         case '7':  \
 600         case '8':  \
 601         case '9'
 602
 603 /**
 604  * returns next final token from a preprocessor macro expansion
 605  */
 606 static void expand_next(void)
 607 {
 608         assert(current_expansion != NULL);
 609
 610         pp_definition_t *definition = current_expansion;
 611
 612 restart:
 613         if (definition->list_len == 0
 614                         || definition->expand_pos >= definition->list_len) {
 615                 /* we're finished with the current macro, move up 1 level in the
 616                  * expansion stack */
 617                 pp_definition_t *parent = definition->parent_expansion;
 618                 definition->parent_expansion = NULL;
 619                 definition->is_expanding     = false;
 620
 621                 /* it was the outermost expansion, parse normal pptoken */
 622                 if (parent == NULL) {
 623                         current_expansion = NULL;
 624                         next_preprocessing_token();
 625                         return;
 626                 }
 627                 definition        = parent;
 628                 current_expansion = definition;
 629                 goto restart;
 630         }
 631         pp_token = definition->token_list[definition->expand_pos];
 632         pp_token.base.source_position = expansion_pos;
 633         ++definition->expand_pos;
 634
 635         if (pp_token.kind != TP_IDENTIFIER)
 636                 return;
 637
 638         /* if it was an identifier then we might need to expand again */
 639         pp_definition_t *const symbol_definition = pp_token.base.symbol->pp_definition;
 640         if (symbol_definition != NULL && !symbol_definition->is_expanding) {
 641                 symbol_definition->parent_expansion = definition;
 642                 symbol_definition->expand_pos       = 0;
 643                 symbol_definition->is_expanding     = true;
 644                 definition                          = symbol_definition;
 645                 current_expansion                   = definition;
 646                 goto restart;
 647         }
 648 }
 649
 650 static void skip_line_comment(void)
 651 {
 652         while (true) {
 653                 switch (input.c) {
 654                 case EOF:
 655                         return;
 656
 657                 case '\r':
 658                 case '\n':
 659                         return;
 660
 661                 default:
 662                         next_char();
 663                         break;
 664                 }
 665         }
 666 }
 667
 668 static void skip_multiline_comment(void)
 669 {
 670         unsigned start_linenr = input.position.lineno;
 671         while (true) {
 672                 switch (input.c) {
 673                 case '/':
 674                         next_char();
 675                         if (input.c == '*') {
 676                                 /* TODO: nested comment, warn here */
 677                         }
 678                         break;
 679                 case '*':
 680                         next_char();
 681                         if (input.c == '/') {
 682                                 if (input.position.lineno != input.output_line)
 683                                         info.whitespace = input.position.colno;
 684                                 next_char();
 685                                 return;
 686                         }
 687                         break;
 688
 689                 case NEWLINE:
 690                         break;
 691
 692                 case EOF: {
 693                         source_position_t source_position;
 694                         source_position.input_name = pp_token.base.source_position.input_name;
 695                         source_position.lineno     = start_linenr;
 696                         errorf(&source_position, "at end of file while looking for comment end");
 697                         return;
 698                 }
 699
 700                 default:
 701                         next_char();
 702                         break;
 703                 }
 704         }
 705 }
 706
 707 static void skip_whitespace(void)
 708 {
 709         while (true) {
 710                 switch (input.c) {
 711                 case ' ':
 712                 case '\t':
 713                         next_char();
 714                         continue;
 715
 716                 case NEWLINE:
 717                         info.at_line_begin = true;
 718                         return;
 719
 720                 case '/':
 721                         next_char();
 722                         if (input.c == '/') {
 723                                 next_char();
 724                                 skip_line_comment();
 725                                 continue;
 726                         } else if (input.c == '*') {
 727                                 next_char();
 728                                 skip_multiline_comment();
 729                                 continue;
 730                         } else {
 731                                 put_back(input.c);
 732                                 input.c = '/';
 733                         }
 734                         return;
 735                 default:
 736                         return;
 737                 }
 738         }
 739 }
 740
 741 static void eat_pp(preprocessor_token_kind_t const type)
 742 {
 743         (void) type;
 744         assert(pp_token.kind == type);
 745         next_preprocessing_token();
 746 }
 747
 748 static void parse_symbol(void)
 749 {
 750         obstack_1grow(&symbol_obstack, (char) input.c);
 751         next_char();
 752
 753         while (true) {
 754                 switch (input.c) {
 755                 case DIGIT_CASES:
 756                 case SYMBOL_CASES:
 757                         obstack_1grow(&symbol_obstack, (char) input.c);
 758                         next_char();
 759                         break;
 760
 761                 default:
 762                         goto end_symbol;
 763                 }
 764         }
 765
 766 end_symbol:
 767         obstack_1grow(&symbol_obstack, '\0');
 768         char *string = obstack_finish(&symbol_obstack);
 769
 770         /* might be a wide string or character constant ( L"string"/L'c' ) */
 771         if (input.c == '"' && string[0] == 'L' && string[1] == '\0') {
 772                 obstack_free(&symbol_obstack, string);
 773                 parse_string_literal(STRING_ENCODING_WIDE);
 774                 return;
 775         } else if (input.c == '\'' && string[0] == 'L' && string[1] == '\0') {
 776                 obstack_free(&symbol_obstack, string);
 777                 parse_character_constant(STRING_ENCODING_WIDE);
 778                 return;
 779         }
 780
 781         symbol_t *symbol = symbol_table_insert(string);
 782
 783         pp_token.kind        = symbol->pp_ID;
 784         pp_token.base.symbol = symbol;
 785
 786         /* we can free the memory from symbol obstack if we already had an entry in
 787          * the symbol table */
 788         if (symbol->string != string) {
 789                 obstack_free(&symbol_obstack, string);
 790         }
 791 }
 792
 793 static void parse_number(void)
 794 {
 795         obstack_1grow(&symbol_obstack, (char) input.c);
 796         next_char();
 797
 798         while (true) {
 799                 switch (input.c) {
 800                 case '.':
 801                 case DIGIT_CASES:
 802                 case SYMBOL_CASES_WITHOUT_E_P:
 803                         obstack_1grow(&symbol_obstack, (char) input.c);
 804                         next_char();
 805                         break;
 806
 807                 case 'e':
 808                 case 'p':
 809                 case 'E':
 810                 case 'P':
 811                         obstack_1grow(&symbol_obstack, (char) input.c);
 812                         next_char();
 813                         if (input.c == '+' || input.c == '-') {
 814                                 obstack_1grow(&symbol_obstack, (char) input.c);
 815                                 next_char();
 816                         }
 817                         break;
 818
 819                 default:
 820                         goto end_number;
 821                 }
 822         }
 823
 824 end_number:
 825         pp_token.kind           = TP_NUMBER;
 826         pp_token.literal.string = sym_make_string(STRING_ENCODING_CHAR);
 827 }
 828
 829
 830 #define MAYBE_PROLOG                                       \
 831                         next_char();                                   \
 832                         while (true) {                                 \
 833                                 switch (input.c) {
 834
 835 #define MAYBE(ch, set_type)                                \
 836                                 case ch:                                   \
 837                                         next_char();                           \
 838                                         pp_token.kind = set_type;              \
 839                                         return;
 840
 841 #define ELSE_CODE(code)                                    \
 842                                 default:                                   \
 843                                         code                                   \
 844                                         return;                                \
 845                                 }                                          \
 846                         }
 847
 848 #define ELSE(set_type)                                     \
 849                 ELSE_CODE(                                         \
 850                         pp_token.kind = set_type;                      \
 851                 )
 852
 853 static void next_preprocessing_token(void)
 854 {
 855         if (current_expansion != NULL) {
 856                 expand_next();
 857                 return;
 858         }
 859
 860         info.at_line_begin  = false;
 861         info.had_whitespace = false;
 862 restart:
 863         pp_token.base.source_position = input.position;
 864         pp_token.base.symbol          = NULL;
 865
 866         switch (input.c) {
 867         case ' ':
 868         case '\t':
 869                 ++info.whitespace;
 870                 info.had_whitespace = true;
 871                 next_char();
 872                 goto restart;
 873
 874         case NEWLINE:
 875                 info.at_line_begin = true;
 876                 info.had_whitespace = true;
 877                 goto restart;
 878
 879         case SYMBOL_CASES:
 880                 parse_symbol();
 881                 return;
 882
 883         case DIGIT_CASES:
 884                 parse_number();
 885                 return;
 886
 887         case '"':
 888                 parse_string_literal(STRING_ENCODING_CHAR);
 889                 return;
 890
 891         case '\'':
 892                 parse_character_constant(STRING_ENCODING_CHAR);
 893                 return;
 894
 895         case '.':
 896                 MAYBE_PROLOG
 897                         case '0':
 898                         case '1':
 899                         case '2':
 900                         case '3':
 901                         case '4':
 902                         case '5':
 903                         case '6':
 904                         case '7':
 905                         case '8':
 906                         case '9':
 907                                 put_back(input.c);
 908                                 input.c = '.';
 909                                 parse_number();
 910                                 return;
 911
 912                         case '.':
 913                                 MAYBE_PROLOG
 914                                 MAYBE('.', TP_DOTDOTDOT)
 915                                 ELSE_CODE(
 916                                         put_back(input.c);
 917                                         input.c = '.';
 918                                         pp_token.kind = '.';
 919                                 )
 920                 ELSE('.')
 921         case '&':
 922                 MAYBE_PROLOG
 923                 MAYBE('&', TP_ANDAND)
 924                 MAYBE('=', TP_ANDEQUAL)
 925                 ELSE('&')
 926         case '*':
 927                 MAYBE_PROLOG
 928                 MAYBE('=', TP_ASTERISKEQUAL)
 929                 ELSE('*')
 930         case '+':
 931                 MAYBE_PROLOG
 932                 MAYBE('+', TP_PLUSPLUS)
 933                 MAYBE('=', TP_PLUSEQUAL)
 934                 ELSE('+')
 935         case '-':
 936                 MAYBE_PROLOG
 937                 MAYBE('>', TP_MINUSGREATER)
 938                 MAYBE('-', TP_MINUSMINUS)
 939                 MAYBE('=', TP_MINUSEQUAL)
 940                 ELSE('-')
 941         case '!':
 942                 MAYBE_PROLOG
 943                 MAYBE('=', TP_EXCLAMATIONMARKEQUAL)
 944                 ELSE('!')
 945         case '/':
 946                 MAYBE_PROLOG
 947                 MAYBE('=', TP_SLASHEQUAL)
 948                         case '*':
 949                                 next_char();
 950                                 info.had_whitespace = true;
 951                                 skip_multiline_comment();
 952                                 goto restart;
 953                         case '/':
 954                                 next_char();
 955                                 info.had_whitespace = true;
 956                                 skip_line_comment();
 957                                 goto restart;
 958                 ELSE('/')
 959         case '%':
 960                 MAYBE_PROLOG
 961                 MAYBE('>', '}')
 962                 MAYBE('=', TP_PERCENTEQUAL)
 963                         case ':':
 964                                 MAYBE_PROLOG
 965                                         case '%':
 966                                                 MAYBE_PROLOG
 967                                                 MAYBE(':', TP_HASHHASH)
 968                                                 ELSE_CODE(
 969                                                         put_back(input.c);
 970                                                         input.c = '%';
 971                                                         pp_token.kind = '#';
 972                                                 )
 973                                 ELSE('#')
 974                 ELSE('%')
 975         case '<':
 976                 MAYBE_PROLOG
 977                 MAYBE(':', '[')
 978                 MAYBE('%', '{')
 979                 MAYBE('=', TP_LESSEQUAL)
 980                         case '<':
 981                                 MAYBE_PROLOG
 982                                 MAYBE('=', TP_LESSLESSEQUAL)
 983                                 ELSE(TP_LESSLESS)
 984                 ELSE('<')
 985         case '>':
 986                 MAYBE_PROLOG
 987                 MAYBE('=', TP_GREATEREQUAL)
 988                         case '>':
 989                                 MAYBE_PROLOG
 990                                 MAYBE('=', TP_GREATERGREATEREQUAL)
 991                                 ELSE(TP_GREATERGREATER)
 992                 ELSE('>')
 993         case '^':
 994                 MAYBE_PROLOG
 995                 MAYBE('=', TP_CARETEQUAL)
 996                 ELSE('^')
 997         case '|':
 998                 MAYBE_PROLOG
 999                 MAYBE('=', TP_PIPEEQUAL)
1000                 MAYBE('|', TP_PIPEPIPE)
1001                 ELSE('|')
1002         case ':':
1003                 MAYBE_PROLOG
1004                 MAYBE('>', ']')
1005                 ELSE(':')
1006         case '=':
1007                 MAYBE_PROLOG
1008                 MAYBE('=', TP_EQUALEQUAL)
1009                 ELSE('=')
1010         case '#':
1011                 MAYBE_PROLOG
1012                 MAYBE('#', TP_HASHHASH)
1013                 ELSE_CODE(
1014                         pp_token.kind = '#';
1015                 )
1016
1017         case '?':
1018         case '[':
1019         case ']':
1020         case '(':
1021         case ')':
1022         case '{':
1023         case '}':
1024         case '~':
1025         case ';':
1026         case ',':
1027                 pp_token.kind = input.c;
1028                 next_char();
1029                 return;
1030
1031         case EOF:
1032                 if (input_stack != NULL) {
1033                         close_input();
1034                         pop_restore_input();
1035                         fputc('\n', out);
1036                         print_line_directive(&input.position, "2");
1037                         goto restart;
1038                 } else {
1039                         pp_token.base.source_position.lineno++;
1040                         info.at_line_begin = true;
1041                         pp_token.kind = TP_EOF;
1042                 }
1043                 return;
1044
1045         default:
1046                 next_char();
1047                 if (!ignore_unknown_chars) {
1048                         errorf(&pp_token.base.source_position,
1049                                "unknown character '%c' found\n", input.c);
1050                         goto restart;
1051                 } else {
1052                         pp_token.kind = input.c;
1053                         return;
1054                 }
1055         }
1056 }
1057
1058 static void print_quoted_string(const char *const string)
1059 {
1060         fputc('"', out);
1061         for (const char *c = string; *c != 0; ++c) {
1062                 switch (*c) {
1063                 case '"': fputs("\\\"", out); break;
1064                 case '\\':  fputs("\\\\", out); break;
1065                 case '\a':  fputs("\\a", out); break;
1066                 case '\b':  fputs("\\b", out); break;
1067                 case '\f':  fputs("\\f", out); break;
1068                 case '\n':  fputs("\\n", out); break;
1069                 case '\r':  fputs("\\r", out); break;
1070                 case '\t':  fputs("\\t", out); break;
1071                 case '\v':  fputs("\\v", out); break;
1072                 case '\?':  fputs("\\?", out); break;
1073                 default:
1074                         if (!isprint(*c)) {
1075                                 fprintf(out, "\\%03o", (unsigned)*c);
1076                                 break;
1077                         }
1078                         fputc(*c, out);
1079                         break;
1080                 }
1081         }
1082         fputc('"', out);
1083 }
1084
1085 static void print_line_directive(const source_position_t *pos, const char *add)
1086 {
1087         fprintf(out, "# %u ", pos->lineno);
1088         print_quoted_string(pos->input_name);
1089         if (add != NULL) {
1090                 fputc(' ', out);
1091                 fputs(add, out);
1092         }
1093
1094         printed_input_name = pos->input_name;
1095         input.output_line  = pos->lineno-1;
1096 }
1097
1098 static bool emit_newlines(void)
1099 {
1100         unsigned delta = pp_token.base.source_position.lineno - input.output_line;
1101         if (delta == 0)
1102                 return false;
1103
1104         if (delta >= 9) {
1105                 fputc('\n', out);
1106                 print_line_directive(&pp_token.base.source_position, NULL);
1107                 fputc('\n', out);
1108         } else {
1109                 for (unsigned i = 0; i < delta; ++i) {
1110                         fputc('\n', out);
1111                 }
1112         }
1113         input.output_line = pp_token.base.source_position.lineno;
1114
1115         for (unsigned i = 0; i < info.whitespace; ++i)
1116                 fputc(' ', out);
1117
1118         return true;
1119 }
1120
1121 static void emit_pp_token(void)
1122 {
1123         if (skip_mode)
1124                 return;
1125
1126         if (!emit_newlines() &&
1127             (info.had_whitespace || tokens_would_paste(last_token, pp_token.kind)))
1128                 fputc(' ', out);
1129
1130         switch (pp_token.kind) {
1131         case TP_IDENTIFIER:
1132                 fputs(pp_token.base.symbol->string, out);
1133                 break;
1134         case TP_NUMBER:
1135                 fputs(pp_token.literal.string.begin, out);
1136                 break;
1137
1138         case TP_STRING_LITERAL:
1139                 fputs(get_string_encoding_prefix(pp_token.literal.string.encoding), out);
1140                 fputc('"', out);
1141                 fputs(pp_token.literal.string.begin, out);
1142                 fputc('"', out);
1143                 break;
1144
1145         case TP_CHARACTER_CONSTANT:
1146                 fputs(get_string_encoding_prefix(pp_token.literal.string.encoding), out);
1147                 fputc('\'', out);
1148                 fputs(pp_token.literal.string.begin, out);
1149                 fputc('\'', out);
1150                 break;
1151         default:
1152                 print_pp_token_kind(out, pp_token.kind);
1153                 break;
1154         }
1155         last_token = pp_token.kind;
1156 }
1157
1158 static void eat_pp_directive(void)
1159 {
1160         while (!info.at_line_begin) {
1161                 next_preprocessing_token();
1162         }
1163 }
1164
1165 static bool strings_equal(const string_t *string1, const string_t *string2)
1166 {
1167         size_t size = string1->size;
1168         if (size != string2->size)
1169                 return false;
1170
1171         const char *c1 = string1->begin;
1172         const char *c2 = string2->begin;
1173         for (size_t i = 0; i < size; ++i, ++c1, ++c2) {
1174                 if (*c1 != *c2)
1175                         return false;
1176         }
1177         return true;
1178 }
1179
1180 static bool pp_tokens_equal(const token_t *token1, const token_t *token2)
1181 {
1182         if (token1->kind != token2->kind)
1183                 return false;
1184
1185         switch (token1->kind) {
1186         case TP_IDENTIFIER:
1187                 return token1->base.symbol == token2->base.symbol;
1188
1189         case TP_NUMBER:
1190         case TP_CHARACTER_CONSTANT:
1191         case TP_STRING_LITERAL:
1192                 return strings_equal(&token1->literal.string, &token2->literal.string);
1193
1194         default:
1195                 return true;
1196         }
1197 }
1198
1199 static bool pp_definitions_equal(const pp_definition_t *definition1,
1200                                  const pp_definition_t *definition2)
1201 {
1202         if (definition1->list_len != definition2->list_len)
1203                 return false;
1204
1205         size_t         len = definition1->list_len;
1206         const token_t *t1  = definition1->token_list;
1207         const token_t *t2  = definition2->token_list;
1208         for (size_t i = 0; i < len; ++i, ++t1, ++t2) {
1209                 if (!pp_tokens_equal(t1, t2))
1210                         return false;
1211         }
1212         return true;
1213 }
1214
1215 static void parse_define_directive(void)
1216 {
1217         eat_pp(TP_define);
1218         assert(obstack_object_size(&pp_obstack) == 0);
1219
1220         if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1221                 errorf(&pp_token.base.source_position,
1222                        "expected identifier after #define, got '%t'", &pp_token);
1223                 goto error_out;
1224         }
1225         symbol_t *const symbol = pp_token.base.symbol;
1226
1227         pp_definition_t *new_definition
1228                 = obstack_alloc(&pp_obstack, sizeof(new_definition[0]));
1229         memset(new_definition, 0, sizeof(new_definition[0]));
1230         new_definition->source_position = input.position;
1231
1232         /* this is probably the only place where spaces are significant in the
1233          * lexer (except for the fact that they separate tokens). #define b(x)
1234          * is something else than #define b (x) */
1235         if (input.c == '(') {
1236                 /* eat the '(' */
1237                 next_preprocessing_token();
1238                 /* get next token after '(' */
1239                 next_preprocessing_token();
1240
1241                 while (true) {
1242                         switch (pp_token.kind) {
1243                         case TP_DOTDOTDOT:
1244                                 new_definition->is_variadic = true;
1245                                 next_preprocessing_token();
1246                                 if (pp_token.kind != ')') {
1247                                         errorf(&input.position,
1248                                                         "'...' not at end of macro argument list");
1249                                         goto error_out;
1250                                 }
1251                                 break;
1252                         case TP_IDENTIFIER:
1253                                 obstack_ptr_grow(&pp_obstack, pp_token.base.symbol);
1254                                 next_preprocessing_token();
1255
1256                                 if (pp_token.kind == ',') {
1257                                         next_preprocessing_token();
1258                                         break;
1259                                 }
1260
1261                                 if (pp_token.kind != ')') {
1262                                         errorf(&pp_token.base.source_position,
1263                                                "expected ',' or ')' after identifier, got '%t'",
1264                                                &pp_token);
1265                                         goto error_out;
1266                                 }
1267                                 break;
1268                         case ')':
1269                                 next_preprocessing_token();
1270                                 goto finish_argument_list;
1271                         default:
1272                                 errorf(&pp_token.base.source_position,
1273                                        "expected identifier, '...' or ')' in #define argument list, got '%t'",
1274                                        &pp_token);
1275                                 goto error_out;
1276                         }
1277                 }
1278
1279         finish_argument_list:
1280                 new_definition->has_parameters = true;
1281                 new_definition->n_parameters
1282                         = obstack_object_size(&pp_obstack) / sizeof(new_definition->parameters[0]);
1283                 new_definition->parameters = obstack_finish(&pp_obstack);
1284         } else {
1285                 next_preprocessing_token();
1286         }
1287
1288         /* construct a new pp_definition on the obstack */
1289         assert(obstack_object_size(&pp_obstack) == 0);
1290         size_t list_len = 0;
1291         while (!info.at_line_begin) {
1292                 obstack_grow(&pp_obstack, &pp_token, sizeof(pp_token));
1293                 ++list_len;
1294                 next_preprocessing_token();
1295         }
1296
1297         new_definition->list_len   = list_len;
1298         new_definition->token_list = obstack_finish(&pp_obstack);
1299
1300         pp_definition_t *old_definition = symbol->pp_definition;
1301         if (old_definition != NULL) {
1302                 if (!pp_definitions_equal(old_definition, new_definition)) {
1303                         warningf(WARN_OTHER, &input.position, "multiple definition of macro '%Y' (first defined %P)", symbol, &old_definition->source_position);
1304                 } else {
1305                         /* reuse the old definition */
1306                         obstack_free(&pp_obstack, new_definition);
1307                         new_definition = old_definition;
1308                 }
1309         }
1310
1311         symbol->pp_definition = new_definition;
1312         return;
1313
1314 error_out:
1315         if (obstack_object_size(&pp_obstack) > 0) {
1316                 char *ptr = obstack_finish(&pp_obstack);
1317                 obstack_free(&pp_obstack, ptr);
1318         }
1319         eat_pp_directive();
1320 }
1321
1322 static void parse_undef_directive(void)
1323 {
1324         eat_pp(TP_undef);
1325
1326         if (pp_token.kind != TP_IDENTIFIER) {
1327                 errorf(&input.position,
1328                        "expected identifier after #undef, got '%t'", &pp_token);
1329                 eat_pp_directive();
1330                 return;
1331         }
1332
1333         pp_token.base.symbol->pp_definition = NULL;
1334         next_preprocessing_token();
1335
1336         if (!info.at_line_begin) {
1337                 warningf(WARN_OTHER, &input.position, "extra tokens at end of #undef directive");
1338         }
1339         eat_pp_directive();
1340 }
1341
1342 /** behind an #include we can have the special headername lexems.
1343  * They're only allowed behind an #include so they're not recognized
1344  * by the normal next_preprocessing_token. We handle them as a special
1345  * exception here */
1346 static void parse_headername(void)
1347 {
1348         const source_position_t start_position = input.position;
1349         string_t                string         = { NULL, 0, STRING_ENCODING_CHAR };
1350         assert(obstack_object_size(&symbol_obstack) == 0);
1351
1352         if (info.at_line_begin) {
1353                 parse_error("expected headername after #include");
1354                 goto finish_error;
1355         }
1356
1357         /* check wether we have a "... or <... headername */
1358         switch (input.c) {
1359         {
1360                 utf32 delimiter;
1361         case '<': delimiter = '>'; goto parse_name;
1362         case '"': delimiter = '"'; goto parse_name;
1363 parse_name:
1364                 next_char();
1365                 while (true) {
1366                         switch (input.c) {
1367                         case NEWLINE:
1368                         case EOF:
1369                                 errorf(&pp_token.base.source_position, "header name without closing '%c'", (char)delimiter);
1370                                 goto finish_error;
1371
1372                         default:
1373                                 if (input.c == delimiter) {
1374                                         next_char();
1375                                         goto finished_headername;
1376                                 } else {
1377                                         obstack_1grow(&symbol_obstack, (char)input.c);
1378                                         next_char();
1379                                 }
1380                                 break;
1381                         }
1382                 }
1383                 /* we should never be here */
1384         }
1385
1386         default:
1387                 /* TODO: do normal pp_token parsing and concatenate results */
1388                 panic("pp_token concat include not implemented yet");
1389         }
1390
1391 finished_headername:
1392         string = sym_make_string(STRING_ENCODING_CHAR);
1393
1394 finish_error:
1395         pp_token.base.source_position = start_position;
1396         pp_token.kind                 = TP_HEADERNAME;
1397         pp_token.literal.string       = string;
1398 }
1399
1400 static bool do_include(bool system_include, const char *headername)
1401 {
1402         size_t headername_len = strlen(headername);
1403         if (!system_include) {
1404                 /* put dirname of current input on obstack */
1405                 const char *filename   = input.position.input_name;
1406                 const char *last_slash = strrchr(filename, '/');
1407                 if (last_slash != NULL) {
1408                         size_t len = last_slash - filename;
1409                         obstack_grow(&symbol_obstack, filename, len + 1);
1410                         obstack_grow0(&symbol_obstack, headername, headername_len);
1411                         char *complete_path = obstack_finish(&symbol_obstack);
1412                         headername = identify_string(complete_path);
1413                 }
1414
1415                 FILE *file = fopen(headername, "r");
1416                 if (file != NULL) {
1417                         switch_input(file, headername);
1418                         return true;
1419                 }
1420         }
1421
1422         assert(obstack_object_size(&symbol_obstack) == 0);
1423         /* check searchpath */
1424         for (searchpath_entry_t *entry = searchpath; entry != NULL;
1425              entry = entry->next) {
1426             const char *path = entry->path;
1427             size_t      len  = strlen(path);
1428                 obstack_grow(&symbol_obstack, path, len);
1429                 if (path[len-1] != '/')
1430                         obstack_1grow(&symbol_obstack, '/');
1431                 obstack_grow(&symbol_obstack, headername, headername_len+1);
1432
1433                 char *complete_path = obstack_finish(&symbol_obstack);
1434                 FILE *file          = fopen(complete_path, "r");
1435                 if (file != NULL) {
1436                         const char *filename = identify_string(complete_path);
1437                         switch_input(file, filename);
1438                         return true;
1439                 } else {
1440                         obstack_free(&symbol_obstack, complete_path);
1441                 }
1442         }
1443
1444         return false;
1445 }
1446
1447 /* read till next newline character, only for parse_include_directive(),
1448  * use eat_pp_directive() in all other cases */
1449 static void skip_till_newline(void)
1450 {
1451         /* skip till newline */
1452         while (true) {
1453                 switch (input.c) {
1454                 case NEWLINE:
1455                 case EOF:
1456                         return;
1457                 }
1458                 next_char();
1459         }
1460 }
1461
1462 static bool parse_include_directive(void)
1463 {
1464         /* don't eat the TP_include here!
1465          * we need an alternative parsing for the next token */
1466         skip_whitespace();
1467         bool system_include = input.c == '<';
1468         parse_headername();
1469         string_t headername = pp_token.literal.string;
1470         if (headername.begin == NULL) {
1471                 eat_pp_directive();
1472                 return false;
1473         }
1474
1475         skip_whitespace();
1476         if (!info.at_line_begin) {
1477                 warningf(WARN_OTHER, &pp_token.base.source_position,
1478                          "extra tokens at end of #include directive");
1479                 skip_till_newline();
1480         }
1481
1482         if (n_inputs > INCLUDE_LIMIT) {
1483                 errorf(&pp_token.base.source_position, "#include nested too deeply");
1484                 /* eat \n or EOF */
1485                 next_preprocessing_token();
1486                 return false;
1487         }
1488
1489         /* switch inputs */
1490         emit_newlines();
1491         push_input();
1492         bool res = do_include(system_include, pp_token.literal.string.begin);
1493         if (!res) {
1494                 errorf(&pp_token.base.source_position, "failed including '%S': %s", &pp_token.literal, strerror(errno));
1495                 pop_restore_input();
1496                 return false;
1497         }
1498
1499         return true;
1500 }
1501
1502 static pp_conditional_t *push_conditional(void)
1503 {
1504         pp_conditional_t *conditional
1505                 = obstack_alloc(&pp_obstack, sizeof(*conditional));
1506         memset(conditional, 0, sizeof(*conditional));
1507
1508         conditional->parent = conditional_stack;
1509         conditional_stack   = conditional;
1510
1511         return conditional;
1512 }
1513
1514 static void pop_conditional(void)
1515 {
1516         assert(conditional_stack != NULL);
1517         conditional_stack = conditional_stack->parent;
1518 }
1519
1520 static void check_unclosed_conditionals(void)
1521 {
1522         while (conditional_stack != NULL) {
1523                 pp_conditional_t *conditional = conditional_stack;
1524
1525                 if (conditional->in_else) {
1526                         errorf(&conditional->source_position, "unterminated #else");
1527                 } else {
1528                         errorf(&conditional->source_position, "unterminated condition");
1529                 }
1530                 pop_conditional();
1531         }
1532 }
1533
1534 static void parse_ifdef_ifndef_directive(void)
1535 {
1536         bool is_ifndef = (pp_token.kind == TP_ifndef);
1537         bool condition;
1538         next_preprocessing_token();
1539
1540         if (skip_mode) {
1541                 eat_pp_directive();
1542                 pp_conditional_t *conditional = push_conditional();
1543                 conditional->source_position  = pp_token.base.source_position;
1544                 conditional->skip             = true;
1545                 return;
1546         }
1547
1548         if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1549                 errorf(&pp_token.base.source_position,
1550                        "expected identifier after #%s, got '%t'",
1551                        is_ifndef ? "ifndef" : "ifdef", &pp_token);
1552                 eat_pp_directive();
1553
1554                 /* just take the true case in the hope to avoid further errors */
1555                 condition = true;
1556         } else {
1557                 /* evaluate wether we are in true or false case */
1558                 condition = (bool)!pp_token.base.symbol->pp_definition == is_ifndef;
1559
1560                 next_preprocessing_token();
1561
1562                 if (!info.at_line_begin) {
1563                         errorf(&pp_token.base.source_position,
1564                                "extra tokens at end of #%s",
1565                                is_ifndef ? "ifndef" : "ifdef");
1566                         eat_pp_directive();
1567                 }
1568         }
1569
1570         pp_conditional_t *conditional = push_conditional();
1571         conditional->source_position  = pp_token.base.source_position;
1572         conditional->condition        = condition;
1573
1574         if (!condition) {
1575                 skip_mode = true;
1576         }
1577 }
1578
1579 static void parse_else_directive(void)
1580 {
1581         eat_pp(TP_else);
1582
1583         if (!info.at_line_begin) {
1584                 if (!skip_mode) {
1585                         warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #else");
1586                 }
1587                 eat_pp_directive();
1588         }
1589
1590         pp_conditional_t *conditional = conditional_stack;
1591         if (conditional == NULL) {
1592                 errorf(&pp_token.base.source_position, "#else without prior #if");
1593                 return;
1594         }
1595
1596         if (conditional->in_else) {
1597                 errorf(&pp_token.base.source_position,
1598                        "#else after #else (condition started %P)",
1599                        &conditional->source_position);
1600                 skip_mode = true;
1601                 return;
1602         }
1603
1604         conditional->in_else = true;
1605         if (!conditional->skip) {
1606                 skip_mode = conditional->condition;
1607         }
1608         conditional->source_position = pp_token.base.source_position;
1609 }
1610
1611 static void parse_endif_directive(void)
1612 {
1613         eat_pp(TP_endif);
1614
1615         if (!info.at_line_begin) {
1616                 if (!skip_mode) {
1617                         warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #endif");
1618                 }
1619                 eat_pp_directive();
1620         }
1621
1622         pp_conditional_t *conditional = conditional_stack;
1623         if (conditional == NULL) {
1624                 errorf(&pp_token.base.source_position, "#endif without prior #if");
1625                 return;
1626         }
1627
1628         if (!conditional->skip) {
1629                 skip_mode = false;
1630         }
1631         pop_conditional();
1632 }
1633
1634 static void parse_preprocessing_directive(void)
1635 {
1636         eat_pp('#');
1637
1638         if (info.at_line_begin) {
1639                 /* empty directive */
1640                 return;
1641         }
1642
1643         if (skip_mode) {
1644                 switch (pp_token.kind) {
1645                 case TP_ifdef:
1646                 case TP_ifndef:
1647                         parse_ifdef_ifndef_directive();
1648                         break;
1649                 case TP_else:
1650                         parse_else_directive();
1651                         break;
1652                 case TP_endif:
1653                         parse_endif_directive();
1654                         break;
1655                 default:
1656                         eat_pp_directive();
1657                         break;
1658                 }
1659         } else {
1660                 switch (pp_token.kind) {
1661                 case TP_define:
1662                         parse_define_directive();
1663                         break;
1664                 case TP_undef:
1665                         parse_undef_directive();
1666                         break;
1667                 case TP_ifdef:
1668                 case TP_ifndef:
1669                         parse_ifdef_ifndef_directive();
1670                         break;
1671                 case TP_else:
1672                         parse_else_directive();
1673                         break;
1674                 case TP_endif:
1675                         parse_endif_directive();
1676                         break;
1677                 case TP_include:
1678                         parse_include_directive();
1679                         break;
1680                 default:
1681                         if (info.at_line_begin) {
1682                                 /* the nop directive "#" */
1683                                 break;
1684                         }
1685                         errorf(&pp_token.base.source_position,
1686                                    "invalid preprocessing directive #%t", &pp_token);
1687                         eat_pp_directive();
1688                         break;
1689                 }
1690         }
1691
1692         assert(info.at_line_begin);
1693 }
1694
1695 static void prepend_include_path(const char *path)
1696 {
1697         searchpath_entry_t *entry = OALLOCZ(&config_obstack, searchpath_entry_t);
1698         entry->path = path;
1699         entry->next = searchpath;
1700         searchpath  = entry;
1701 }
1702
1703 static void setup_include_path(void)
1704 {
1705         /* built-in paths */
1706         prepend_include_path("/usr/include");
1707
1708         /* parse environment variable */
1709         const char *cpath = getenv("CPATH");
1710         if (cpath != NULL && *cpath != '\0') {
1711                 const char *begin = cpath;
1712                 const char *c;
1713                 do {
1714                         c = begin;
1715                         while (*c != '\0' && *c != ':')
1716                                 ++c;
1717
1718                         size_t len = c-begin;
1719                         if (len == 0) {
1720                                 /* for gcc compatibility (Matze: I would expect that
1721                                  * nothing happens for an empty entry...) */
1722                                 prepend_include_path(".");
1723                         } else {
1724                                 char *string = obstack_alloc(&config_obstack, len+1);
1725                                 memcpy(string, begin, len);
1726                                 string[len] = '\0';
1727
1728                                 prepend_include_path(string);
1729                         }
1730
1731                         begin = c+1;
1732                         /* skip : */
1733                         if (*begin == ':')
1734                                 ++begin;
1735                 } while(*c != '\0');
1736         }
1737 }
1738
1739 int pptest_main(int argc, char **argv);
1740 int pptest_main(int argc, char **argv)
1741 {
1742         init_symbol_table();
1743         init_tokens();
1744
1745         obstack_init(&config_obstack);
1746         obstack_init(&pp_obstack);
1747         obstack_init(&input_obstack);
1748         strset_init(&stringset);
1749
1750         setup_include_path();
1751
1752         /* simplistic commandline parser */
1753         const char *filename = NULL;
1754         const char *output = NULL;
1755         for (int i = 1; i < argc; ++i) {
1756                 const char *opt = argv[i];
1757                 if (streq(opt, "-I")) {
1758                         prepend_include_path(argv[++i]);
1759                         continue;
1760                 } else if (streq(opt, "-E")) {
1761                         /* ignore */
1762                 } else if (streq(opt, "-o")) {
1763                         output = argv[++i];
1764                         continue;
1765                 } else if (opt[0] == '-') {
1766                         fprintf(stderr, "Unknown option '%s'\n", opt);
1767                 } else {
1768                         if (filename != NULL)
1769                                 fprintf(stderr, "Multiple inputs not supported\n");
1770                         filename = argv[i];
1771                 }
1772         }
1773         if (filename == NULL) {
1774                 fprintf(stderr, "No input specified\n");
1775                 return 1;
1776         }
1777
1778         if (output == NULL) {
1779                 out = stdout;
1780         } else {
1781                 out = fopen(output, "w");
1782                 if (out == NULL) {
1783                         fprintf(stderr, "Couldn't open output '%s'\n", output);
1784                         return 1;
1785                 }
1786         }
1787
1788         /* just here for gcc compatibility */
1789         fprintf(out, "# 1 \"%s\"\n", filename);
1790         fprintf(out, "# 1 \"<built-in>\"\n");
1791         fprintf(out, "# 1 \"<command-line>\"\n");
1792
1793         FILE *file = fopen(filename, "r");
1794         if (file == NULL) {
1795                 fprintf(stderr, "Couldn't open input '%s'\n", filename);
1796                 return 1;
1797         }
1798         switch_input(file, filename);
1799
1800         while (true) {
1801                 if (pp_token.kind == '#' && info.at_line_begin) {
1802                         parse_preprocessing_directive();
1803                         continue;
1804                 } else if (pp_token.kind == TP_EOF) {
1805                         goto end_of_main_loop;
1806                 } else if (pp_token.kind == TP_IDENTIFIER) {
1807                         symbol_t        *const symbol        = pp_token.base.symbol;
1808                         pp_definition_t *const pp_definition = symbol->pp_definition;
1809                         if (pp_definition != NULL && !pp_definition->is_expanding) {
1810                                 expansion_pos = pp_token.base.source_position;
1811                                 if (pp_definition->has_parameters) {
1812                                         source_position_t position = pp_token.base.source_position;
1813                                         add_token_info_t old_info = info;
1814                                         next_preprocessing_token();
1815                                         add_token_info_t new_info = info;
1816
1817                                         /* no opening brace -> no expansion */
1818                                         if (pp_token.kind == '(') {
1819                                                 eat_pp('(');
1820
1821                                                 /* parse arguments (TODO) */
1822                                                 while (pp_token.kind != TP_EOF && pp_token.kind != ')')
1823                                                         next_preprocessing_token();
1824                                         } else {
1825                                                 token_t next_token = pp_token;
1826                                                 /* restore identifier token */
1827                                                 pp_token.kind                 = TP_IDENTIFIER;
1828                                                 pp_token.base.symbol          = symbol;
1829                                                 pp_token.base.source_position = position;
1830                                                 info = old_info;
1831                                                 emit_pp_token();
1832
1833                                                 info = new_info;
1834                                                 pp_token = next_token;
1835                                                 continue;
1836                                         }
1837                                         info = old_info;
1838                                 }
1839                                 pp_definition->expand_pos   = 0;
1840                                 pp_definition->is_expanding = true;
1841                                 current_expansion           = pp_definition;
1842                                 expand_next();
1843                                 continue;
1844                         }
1845                 }
1846
1847                 emit_pp_token();
1848                 next_preprocessing_token();
1849         }
1850 end_of_main_loop:
1851
1852         fputc('\n', out);
1853         check_unclosed_conditionals();
1854         close_input();
1855         if (out != stdout)
1856                 fclose(out);
1857
1858         obstack_free(&input_obstack, NULL);
1859         obstack_free(&pp_obstack, NULL);
1860         obstack_free(&config_obstack, NULL);
1861
1862         strset_destroy(&stringset);
1863
1864         exit_tokens();
1865         exit_symbol_table();
1866
1867         return 0;
1868 }