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 "preprocessor.h"
  10 #include "token_t.h"
  11 #include "symbol_t.h"
  12 #include "adt/util.h"
  13 #include "adt/error.h"
  14 #include "adt/strutil.h"
  15 #include "adt/strset.h"
  16 #include "lang_features.h"
  17 #include "diagnostic.h"
  18 #include "string_rep.h"
  19 #include "input.h"
  20
  21 #define MAX_PUTBACK 3
  22 #define INCLUDE_LIMIT 199  /* 199 is for gcc "compatibility" */
  23
  24 typedef struct saved_token_t {
  25         token_t token;
  26         bool    had_whitespace;
  27 } saved_token_t;
  28
  29 typedef struct whitespace_info_t {
  30         /** current token had whitespace in front of it */
  31         bool     had_whitespace;
  32         /** current token is at the beginning of a line.
  33          * => a "#" at line begin starts a preprocessing directive. */
  34         bool     at_line_begin;
  35         /** number of spaces before the first token in a line */
  36         unsigned whitespace_at_line_begin;
  37 } whitespace_info_t;
  38
  39 struct pp_definition_t {
  40         symbol_t          *symbol;
  41         source_position_t  source_position;
  42         pp_definition_t   *parent_expansion;
  43         size_t             expand_pos;
  44         whitespace_info_t  expand_info;
  45         bool               is_variadic    : 1;
  46         bool               is_expanding   : 1;
  47         bool               has_parameters : 1;
  48         bool               is_parameter   : 1;
  49         pp_definition_t   *function_definition;
  50         size_t             n_parameters;
  51         pp_definition_t   *parameters;
  52
  53         /* replacement */
  54         size_t             list_len;
  55         saved_token_t     *token_list;
  56 };
  57
  58 typedef struct pp_conditional_t pp_conditional_t;
  59 struct pp_conditional_t {
  60         source_position_t  source_position;
  61         bool               condition;
  62         bool               in_else;
  63         /** conditional in skip mode (then+else gets skipped) */
  64         bool               skip;
  65         pp_conditional_t  *parent;
  66 };
  67
  68 typedef struct pp_input_t pp_input_t;
  69 struct pp_input_t {
  70         FILE               *file;
  71         input_t            *input;
  72         utf32               c;
  73         utf32               buf[1024+MAX_PUTBACK];
  74         const utf32        *bufend;
  75         const utf32        *bufpos;
  76         source_position_t   position;
  77         pp_input_t         *parent;
  78         unsigned            output_line;
  79         searchpath_entry_t *path;
  80 };
  81
  82 struct searchpath_entry_t {
  83         const char         *path;
  84         searchpath_entry_t *next;
  85 };
  86
  87 static pp_input_t      input;
  88
  89 static pp_input_t     *input_stack;
  90 static unsigned        n_inputs;
  91 static struct obstack  input_obstack;
  92
  93 static pp_conditional_t *conditional_stack;
  94
  95 token_t                  pp_token;
  96 bool                     allow_dollar_in_symbol   = true;
  97 static bool              resolve_escape_sequences = true;
  98 static bool              error_on_unknown_chars   = true;
  99 static bool              skip_mode;
 100 static FILE             *out;
 101 static struct obstack    pp_obstack;
 102 static struct obstack    config_obstack;
 103 static const char       *printed_input_name = NULL;
 104 static source_position_t expansion_pos;
 105 static pp_definition_t  *current_expansion  = NULL;
 106 static pp_definition_t  *current_call       = NULL;
 107 static pp_definition_t  *current_argument   = NULL;
 108 static pp_definition_t  *argument_expanding = NULL;
 109 static unsigned          argument_brace_count;
 110 static strset_t          stringset;
 111 static token_kind_t      last_token;
 112
 113 static searchpath_entry_t *searchpath;
 114
 115 static whitespace_info_t next_info; /* valid if had_whitespace is true */
 116 static whitespace_info_t info;
 117
 118 static inline void next_char(void);
 119 static void next_input_token(void);
 120 static void print_line_directive(const source_position_t *pos, const char *add);
 121
 122 static symbol_t *symbol_colongreater;
 123 static symbol_t *symbol_lesscolon;
 124 static symbol_t *symbol_lesspercent;
 125 static symbol_t *symbol_percentcolon;
 126 static symbol_t *symbol_percentcolonpercentcolon;
 127 static symbol_t *symbol_percentgreater;
 128
 129 static void init_symbols(void)
 130 {
 131         symbol_colongreater             = symbol_table_insert(":>");
 132         symbol_lesscolon                = symbol_table_insert("<:");
 133         symbol_lesspercent              = symbol_table_insert("<%");
 134         symbol_percentcolon             = symbol_table_insert("%:");
 135         symbol_percentcolonpercentcolon = symbol_table_insert("%:%:");
 136         symbol_percentgreater           = symbol_table_insert("%>");
 137 }
 138
 139 void switch_pp_input(FILE *const file, char const *const filename, searchpath_entry_t *const path)
 140 {
 141         input.file                = file;
 142         input.input               = input_from_stream(file, NULL);
 143         input.bufend              = NULL;
 144         input.bufpos              = NULL;
 145         input.output_line         = 0;
 146         input.position.input_name = filename;
 147         input.position.lineno     = 1;
 148         input.path                = path;
 149
 150         /* indicate that we're at a new input */
 151         print_line_directive(&input.position, input_stack != NULL ? "1" : NULL);
 152
 153         /* place a virtual '\n' so we realize we're at line begin */
 154         input.position.lineno = 0;
 155         input.c               = '\n';
 156 }
 157
 158 FILE *close_pp_input(void)
 159 {
 160         input_free(input.input);
 161
 162         FILE* const file = input.file;
 163         assert(file);
 164
 165         input.input  = NULL;
 166         input.file   = NULL;
 167         input.bufend = NULL;
 168         input.bufpos = NULL;
 169         input.c      = EOF;
 170
 171         return file;
 172 }
 173
 174 static void push_input(void)
 175 {
 176         pp_input_t *const saved_input = obstack_copy(&input_obstack, &input, sizeof(input));
 177
 178         /* adjust buffer positions */
 179         if (input.bufpos != NULL)
 180                 saved_input->bufpos = saved_input->buf + (input.bufpos - input.buf);
 181         if (input.bufend != NULL)
 182                 saved_input->bufend = saved_input->buf + (input.bufend - input.buf);
 183
 184         saved_input->parent = input_stack;
 185         input_stack         = saved_input;
 186         ++n_inputs;
 187 }
 188
 189 static void pop_restore_input(void)
 190 {
 191         assert(n_inputs > 0);
 192         assert(input_stack != NULL);
 193
 194         pp_input_t *saved_input = input_stack;
 195
 196         memcpy(&input, saved_input, sizeof(input));
 197         input.parent = NULL;
 198
 199         /* adjust buffer positions */
 200         if (saved_input->bufpos != NULL)
 201                 input.bufpos = input.buf + (saved_input->bufpos - saved_input->buf);
 202         if (saved_input->bufend != NULL)
 203                 input.bufend = input.buf + (saved_input->bufend - saved_input->buf);
 204
 205         input_stack = saved_input->parent;
 206         obstack_free(&input_obstack, saved_input);
 207         --n_inputs;
 208 }
 209
 210 /**
 211  * Prints a parse error message at the current token.
 212  *
 213  * @param msg   the error message
 214  */
 215 static void parse_error(const char *msg)
 216 {
 217         errorf(&pp_token.base.source_position,  "%s", msg);
 218 }
 219
 220 static inline void next_real_char(void)
 221 {
 222         assert(input.bufpos <= input.bufend);
 223         if (input.bufpos >= input.bufend) {
 224                 size_t const n = decode(input.input, input.buf + MAX_PUTBACK, lengthof(input.buf) - MAX_PUTBACK);
 225                 if (n == 0) {
 226                         input.c = EOF;
 227                         return;
 228                 }
 229                 input.bufpos = input.buf + MAX_PUTBACK;
 230                 input.bufend = input.bufpos + n;
 231         }
 232         input.c = *input.bufpos++;
 233         ++input.position.colno;
 234 }
 235
 236 /**
 237  * Put a character back into the buffer.
 238  *
 239  * @param pc  the character to put back
 240  */
 241 static inline void put_back(utf32 const pc)
 242 {
 243         assert(input.bufpos > input.buf);
 244         *(--input.bufpos - input.buf + input.buf) = (char) pc;
 245         --input.position.colno;
 246 }
 247
 248 #define NEWLINE \
 249         '\r': \
 250                 next_char(); \
 251                 if (input.c == '\n') { \
 252         case '\n': \
 253                         next_char(); \
 254                 } \
 255                 ++input.position.lineno; \
 256                 input.position.colno = 1; \
 257                 goto newline; \
 258                 newline // Let it look like an ordinary case label.
 259
 260 #define eat(c_type) (assert(input.c == c_type), next_char())
 261
 262 static void maybe_concat_lines(void)
 263 {
 264         eat('\\');
 265
 266         switch (input.c) {
 267         case NEWLINE:
 268                 info.whitespace_at_line_begin = 0;
 269                 return;
 270
 271         default:
 272                 break;
 273         }
 274
 275         put_back(input.c);
 276         input.c = '\\';
 277 }
 278
 279 /**
 280  * Set c to the next input character, ie.
 281  * after expanding trigraphs.
 282  */
 283 static inline void next_char(void)
 284 {
 285         next_real_char();
 286
 287         /* filter trigraphs and concatenated lines */
 288         if (UNLIKELY(input.c == '\\')) {
 289                 maybe_concat_lines();
 290                 goto end_of_next_char;
 291         }
 292
 293         if (LIKELY(input.c != '?'))
 294                 goto end_of_next_char;
 295
 296         next_real_char();
 297         if (LIKELY(input.c != '?')) {
 298                 put_back(input.c);
 299                 input.c = '?';
 300                 goto end_of_next_char;
 301         }
 302
 303         next_real_char();
 304         switch (input.c) {
 305         case '=': input.c = '#'; break;
 306         case '(': input.c = '['; break;
 307         case '/': input.c = '\\'; maybe_concat_lines(); break;
 308         case ')': input.c = ']'; break;
 309         case '\'': input.c = '^'; break;
 310         case '<': input.c = '{'; break;
 311         case '!': input.c = '|'; break;
 312         case '>': input.c = '}'; break;
 313         case '-': input.c = '~'; break;
 314         default:
 315                 put_back(input.c);
 316                 put_back('?');
 317                 input.c = '?';
 318                 break;
 319         }
 320
 321 end_of_next_char:;
 322 #ifdef DEBUG_CHARS
 323         printf("nchar '%c'\n", input.c);
 324 #endif
 325 }
 326
 327
 328
 329 /**
 330  * Returns true if the given char is a octal digit.
 331  *
 332  * @param char  the character to check
 333  */
 334 static inline bool is_octal_digit(int chr)
 335 {
 336         switch (chr) {
 337         case '0':
 338         case '1':
 339         case '2':
 340         case '3':
 341         case '4':
 342         case '5':
 343         case '6':
 344         case '7':
 345                 return true;
 346         default:
 347                 return false;
 348         }
 349 }
 350
 351 /**
 352  * Returns the value of a digit.
 353  * The only portable way to do it ...
 354  */
 355 static int digit_value(int digit)
 356 {
 357         switch (digit) {
 358         case '0': return 0;
 359         case '1': return 1;
 360         case '2': return 2;
 361         case '3': return 3;
 362         case '4': return 4;
 363         case '5': return 5;
 364         case '6': return 6;
 365         case '7': return 7;
 366         case '8': return 8;
 367         case '9': return 9;
 368         case 'a':
 369         case 'A': return 10;
 370         case 'b':
 371         case 'B': return 11;
 372         case 'c':
 373         case 'C': return 12;
 374         case 'd':
 375         case 'D': return 13;
 376         case 'e':
 377         case 'E': return 14;
 378         case 'f':
 379         case 'F': return 15;
 380         default:
 381                 panic("wrong character given");
 382         }
 383 }
 384
 385 /**
 386  * Parses an octal character sequence.
 387  *
 388  * @param first_digit  the already read first digit
 389  */
 390 static utf32 parse_octal_sequence(const utf32 first_digit)
 391 {
 392         assert(is_octal_digit(first_digit));
 393         utf32 value = digit_value(first_digit);
 394         if (!is_octal_digit(input.c)) return value;
 395         value = 8 * value + digit_value(input.c);
 396         next_char();
 397         if (!is_octal_digit(input.c)) return value;
 398         value = 8 * value + digit_value(input.c);
 399         next_char();
 400         return value;
 401
 402 }
 403
 404 /**
 405  * Parses a hex character sequence.
 406  */
 407 static utf32 parse_hex_sequence(void)
 408 {
 409         utf32 value = 0;
 410         while (isxdigit(input.c)) {
 411                 value = 16 * value + digit_value(input.c);
 412                 next_char();
 413         }
 414         return value;
 415 }
 416
 417 static bool is_universal_char_valid(utf32 const v)
 418 {
 419         /* C11 §6.4.3:2 */
 420         if (v < 0xA0U && v != 0x24 && v != 0x40 && v != 0x60)
 421                 return false;
 422         if (0xD800 <= v && v <= 0xDFFF)
 423                 return false;
 424         return true;
 425 }
 426
 427 static utf32 parse_universal_char(unsigned const n_digits)
 428 {
 429         utf32 v = 0;
 430         for (unsigned k = n_digits; k != 0; --k) {
 431                 if (isxdigit(input.c)) {
 432                         v = 16 * v + digit_value(input.c);
 433                         if (!resolve_escape_sequences)
 434                                 obstack_1grow(&symbol_obstack, input.c);
 435                         next_char();
 436                 } else {
 437                         errorf(&input.position,
 438                                "short universal character name, expected %u more digits",
 439                                    k);
 440                         break;
 441                 }
 442         }
 443         if (!is_universal_char_valid(v)) {
 444                 errorf(&input.position,
 445                        "\\%c%0*X is not a valid universal character name",
 446                        n_digits == 4 ? 'u' : 'U', (int)n_digits, v);
 447         }
 448         return v;
 449 }
 450
 451 static bool is_universal_char_valid_identifier(utf32 const v)
 452 {
 453         /* C11 Annex D.1 */
 454         if (                v == 0x000A8) return true;
 455         if (                v == 0x000AA) return true;
 456         if (                v == 0x000AD) return true;
 457         if (                v == 0x000AF) return true;
 458         if (0x000B2 <= v && v <= 0x000B5) return true;
 459         if (0x000B7 <= v && v <= 0x000BA) return true;
 460         if (0x000BC <= v && v <= 0x000BE) return true;
 461         if (0x000C0 <= v && v <= 0x000D6) return true;
 462         if (0x000D8 <= v && v <= 0x000F6) return true;
 463         if (0x000F8 <= v && v <= 0x000FF) return true;
 464         if (0x00100 <= v && v <= 0x0167F) return true;
 465         if (0x01681 <= v && v <= 0x0180D) return true;
 466         if (0x0180F <= v && v <= 0x01FFF) return true;
 467         if (0x0200B <= v && v <= 0x0200D) return true;
 468         if (0x0202A <= v && v <= 0x0202E) return true;
 469         if (0x0203F <= v && v <= 0x02040) return true;
 470         if (                v == 0x02054) return true;
 471         if (0x02060 <= v && v <= 0x0206F) return true;
 472         if (0x02070 <= v && v <= 0x0218F) return true;
 473         if (0x02460 <= v && v <= 0x024FF) return true;
 474         if (0x02776 <= v && v <= 0x02793) return true;
 475         if (0x02C00 <= v && v <= 0x02DFF) return true;
 476         if (0x02E80 <= v && v <= 0x02FFF) return true;
 477         if (0x03004 <= v && v <= 0x03007) return true;
 478         if (0x03021 <= v && v <= 0x0302F) return true;
 479         if (0x03031 <= v && v <= 0x0303F) return true;
 480         if (0x03040 <= v && v <= 0x0D7FF) return true;
 481         if (0x0F900 <= v && v <= 0x0FD3D) return true;
 482         if (0x0FD40 <= v && v <= 0x0FDCF) return true;
 483         if (0x0FDF0 <= v && v <= 0x0FE44) return true;
 484         if (0x0FE47 <= v && v <= 0x0FFFD) return true;
 485         if (0x10000 <= v && v <= 0x1FFFD) return true;
 486         if (0x20000 <= v && v <= 0x2FFFD) return true;
 487         if (0x30000 <= v && v <= 0x3FFFD) return true;
 488         if (0x40000 <= v && v <= 0x4FFFD) return true;
 489         if (0x50000 <= v && v <= 0x5FFFD) return true;
 490         if (0x60000 <= v && v <= 0x6FFFD) return true;
 491         if (0x70000 <= v && v <= 0x7FFFD) return true;
 492         if (0x80000 <= v && v <= 0x8FFFD) return true;
 493         if (0x90000 <= v && v <= 0x9FFFD) return true;
 494         if (0xA0000 <= v && v <= 0xAFFFD) return true;
 495         if (0xB0000 <= v && v <= 0xBFFFD) return true;
 496         if (0xC0000 <= v && v <= 0xCFFFD) return true;
 497         if (0xD0000 <= v && v <= 0xDFFFD) return true;
 498         if (0xE0000 <= v && v <= 0xEFFFD) return true;
 499         return false;
 500 }
 501
 502 static bool is_universal_char_valid_identifier_start(utf32 const v)
 503 {
 504         /* C11 Annex D.2 */
 505         if (0x0300 <= v && v <= 0x036F) return false;
 506         if (0x1DC0 <= v && v <= 0x1DFF) return false;
 507         if (0x20D0 <= v && v <= 0x20FF) return false;
 508         if (0xFE20 <= v && v <= 0xFE2F) return false;
 509         return true;
 510 }
 511
 512 /**
 513  * Parse an escape sequence.
 514  */
 515 static utf32 parse_escape_sequence(void)
 516 {
 517         eat('\\');
 518
 519         utf32 const ec = input.c;
 520         next_char();
 521
 522         switch (ec) {
 523         case '"':  return '"';
 524         case '\'': return '\'';
 525         case '\\': return '\\';
 526         case '?': return '\?';
 527         case 'a': return '\a';
 528         case 'b': return '\b';
 529         case 'f': return '\f';
 530         case 'n': return '\n';
 531         case 'r': return '\r';
 532         case 't': return '\t';
 533         case 'v': return '\v';
 534         case 'x':
 535                 return parse_hex_sequence();
 536         case '0':
 537         case '1':
 538         case '2':
 539         case '3':
 540         case '4':
 541         case '5':
 542         case '6':
 543         case '7':
 544                 return parse_octal_sequence(ec);
 545         case EOF:
 546                 parse_error("reached end of file while parsing escape sequence");
 547                 return EOF;
 548         /* \E is not documented, but handled, by GCC.  It is acceptable according
 549          * to §6.11.4, whereas \e is not. */
 550         case 'E':
 551         case 'e':
 552                 if (c_mode & _GNUC)
 553                         return 27;   /* hopefully 27 is ALWAYS the code for ESCAPE */
 554                 break;
 555
 556         case 'U': return parse_universal_char(8);
 557         case 'u': return parse_universal_char(4);
 558
 559         default:
 560                 break;
 561         }
 562         /* §6.4.4.4:8 footnote 64 */
 563         parse_error("unknown escape sequence");
 564         return EOF;
 565 }
 566
 567 static const char *identify_string(char *string)
 568 {
 569         const char *result = strset_insert(&stringset, string);
 570         if (result != string) {
 571                 obstack_free(&symbol_obstack, string);
 572         }
 573         return result;
 574 }
 575
 576 static string_t sym_make_string(string_encoding_t const enc)
 577 {
 578         obstack_1grow(&symbol_obstack, '\0');
 579         size_t      const len    = obstack_object_size(&symbol_obstack) - 1;
 580         char       *const string = obstack_finish(&symbol_obstack);
 581         char const *const result = identify_string(string);
 582         return (string_t){ result, len, enc };
 583 }
 584
 585 string_t make_string(char const *const string)
 586 {
 587         obstack_grow(&symbol_obstack, string, strlen(string));
 588         return sym_make_string(STRING_ENCODING_CHAR);
 589 }
 590
 591 static void parse_string(utf32 const delimiter, token_kind_t const kind,
 592                          string_encoding_t const enc,
 593                          char const *const context)
 594 {
 595         const unsigned start_linenr = input.position.lineno;
 596
 597         eat(delimiter);
 598
 599         while (true) {
 600                 switch (input.c) {
 601                 case '\\': {
 602                         if (resolve_escape_sequences) {
 603                                 utf32 const tc = parse_escape_sequence();
 604                                 if (enc == STRING_ENCODING_CHAR) {
 605                                         if (tc >= 0x100) {
 606                                                 warningf(WARN_OTHER, &pp_token.base.source_position, "escape sequence out of range");
 607                                         }
 608                                         obstack_1grow(&symbol_obstack, tc);
 609                                 } else {
 610                                         obstack_grow_utf8(&symbol_obstack, tc);
 611                                 }
 612                         } else {
 613                                 obstack_1grow(&symbol_obstack, (char)input.c);
 614                                 next_char();
 615                                 obstack_1grow(&symbol_obstack, (char)input.c);
 616                                 next_char();
 617                         }
 618                         break;
 619                 }
 620
 621                 case NEWLINE:
 622                         errorf(&pp_token.base.source_position, "newline while parsing %s", context);
 623                         break;
 624
 625                 case EOF: {
 626                         source_position_t source_position;
 627                         source_position.input_name = pp_token.base.source_position.input_name;
 628                         source_position.lineno     = start_linenr;
 629                         errorf(&source_position, "EOF while parsing %s", context);
 630                         goto end_of_string;
 631                 }
 632
 633                 default:
 634                         if (input.c == delimiter) {
 635                                 next_char();
 636                                 goto end_of_string;
 637                         } else {
 638                                 obstack_grow_utf8(&symbol_obstack, input.c);
 639                                 next_char();
 640                                 break;
 641                         }
 642                 }
 643         }
 644
 645 end_of_string:
 646         pp_token.kind           = kind;
 647         pp_token.literal.string = sym_make_string(enc);
 648 }
 649
 650 static void parse_string_literal(string_encoding_t const enc)
 651 {
 652         parse_string('"', T_STRING_LITERAL, enc, "string literal");
 653 }
 654
 655 static void parse_character_constant(string_encoding_t const enc)
 656 {
 657         parse_string('\'', T_CHARACTER_CONSTANT, enc, "character constant");
 658         if (pp_token.literal.string.size == 0) {
 659                 parse_error("empty character constant");
 660         }
 661 }
 662
 663 #define SYMBOL_CASES_WITHOUT_E_P \
 664              '$': if (!allow_dollar_in_symbol) goto dollar_sign; \
 665         case 'a': \
 666         case 'b': \
 667         case 'c': \
 668         case 'd': \
 669         case 'f': \
 670         case 'g': \
 671         case 'h': \
 672         case 'i': \
 673         case 'j': \
 674         case 'k': \
 675         case 'l': \
 676         case 'm': \
 677         case 'n': \
 678         case 'o': \
 679         case 'q': \
 680         case 'r': \
 681         case 's': \
 682         case 't': \
 683         case 'u': \
 684         case 'v': \
 685         case 'w': \
 686         case 'x': \
 687         case 'y': \
 688         case 'z': \
 689         case 'A': \
 690         case 'B': \
 691         case 'C': \
 692         case 'D': \
 693         case 'F': \
 694         case 'G': \
 695         case 'H': \
 696         case 'I': \
 697         case 'J': \
 698         case 'K': \
 699         case 'L': \
 700         case 'M': \
 701         case 'N': \
 702         case 'O': \
 703         case 'Q': \
 704         case 'R': \
 705         case 'S': \
 706         case 'T': \
 707         case 'U': \
 708         case 'V': \
 709         case 'W': \
 710         case 'X': \
 711         case 'Y': \
 712         case 'Z': \
 713         case '_'
 714
 715 #define SYMBOL_CASES \
 716              SYMBOL_CASES_WITHOUT_E_P: \
 717         case 'e': \
 718         case 'p': \
 719         case 'E': \
 720         case 'P'
 721
 722 #define DIGIT_CASES \
 723              '0':  \
 724         case '1':  \
 725         case '2':  \
 726         case '3':  \
 727         case '4':  \
 728         case '5':  \
 729         case '6':  \
 730         case '7':  \
 731         case '8':  \
 732         case '9'
 733
 734 static void start_expanding(pp_definition_t *definition)
 735 {
 736         definition->parent_expansion = current_expansion;
 737         definition->expand_pos       = 0;
 738         definition->is_expanding     = true;
 739         if (definition->list_len > 0) {
 740                 definition->token_list[0].had_whitespace
 741                         = info.had_whitespace;
 742         }
 743         current_expansion = definition;
 744 }
 745
 746 static void finished_expanding(pp_definition_t *definition)
 747 {
 748         assert(definition->is_expanding);
 749         pp_definition_t *parent = definition->parent_expansion;
 750         definition->parent_expansion = NULL;
 751         definition->is_expanding     = false;
 752
 753         /* stop further expanding once we expanded a parameter used in a
 754          * sub macro-call */
 755         if (definition == argument_expanding)
 756                 argument_expanding = NULL;
 757
 758         assert(current_expansion == definition);
 759         current_expansion = parent;
 760 }
 761
 762 static inline void set_punctuator(token_kind_t const kind)
 763 {
 764         pp_token.kind        = kind;
 765         pp_token.base.symbol = token_symbols[kind];
 766 }
 767
 768 static inline void set_digraph(token_kind_t const kind, symbol_t *const symbol)
 769 {
 770         pp_token.kind        = kind;
 771         pp_token.base.symbol = symbol;
 772 }
 773
 774 /**
 775  * returns next final token from a preprocessor macro expansion
 776  */
 777 static bool expand_next(void)
 778 {
 779         if (current_expansion == NULL)
 780                 return false;
 781
 782 restart:;
 783         size_t pos = current_expansion->expand_pos;
 784         if (pos >= current_expansion->list_len) {
 785                 finished_expanding(current_expansion);
 786                 /* it was the outermost expansion, parse pptoken normally */
 787                 if (current_expansion == NULL) {
 788                         return false;
 789                 }
 790                 goto restart;
 791         }
 792         const saved_token_t *saved = &current_expansion->token_list[pos++];
 793         pp_token = saved->token;
 794
 795         if (current_expansion->expand_pos > 0)
 796                 info.had_whitespace = saved->had_whitespace;
 797         pp_token.base.source_position = expansion_pos;
 798         ++current_expansion->expand_pos;
 799
 800         return true;
 801 }
 802
 803 /**
 804  * Returns the next token kind found when continuing the current expansions
 805  * without starting new sub-expansions.
 806  */
 807 static token_kind_t peek_expansion(void)
 808 {
 809         for (pp_definition_t *e = current_expansion; e; e = e->parent_expansion) {
 810                 if (e->expand_pos < e->list_len)
 811                         return e->token_list[e->expand_pos].token.kind;
 812         }
 813         return T_EOF;
 814 }
 815
 816 static void skip_line_comment(void)
 817 {
 818         info.had_whitespace = true;
 819         while (true) {
 820                 switch (input.c) {
 821                 case EOF:
 822                         return;
 823
 824                 case '\r':
 825                 case '\n':
 826                         return;
 827
 828                 default:
 829                         next_char();
 830                         break;
 831                 }
 832         }
 833 }
 834
 835 static void skip_multiline_comment(void)
 836 {
 837         info.had_whitespace = true;
 838
 839         unsigned start_linenr = input.position.lineno;
 840         while (true) {
 841                 switch (input.c) {
 842                 case '/':
 843                         next_char();
 844                         if (input.c == '*') {
 845                                 /* TODO: nested comment, warn here */
 846                         }
 847                         break;
 848                 case '*':
 849                         next_char();
 850                         if (input.c == '/') {
 851                                 if (input.position.lineno != input.output_line)
 852                                         info.whitespace_at_line_begin = input.position.colno;
 853                                 next_char();
 854                                 return;
 855                         }
 856                         break;
 857
 858                 case NEWLINE:
 859                         break;
 860
 861                 case EOF: {
 862                         source_position_t source_position;
 863                         source_position.input_name = pp_token.base.source_position.input_name;
 864                         source_position.lineno     = start_linenr;
 865                         errorf(&source_position, "at end of file while looking for comment end");
 866                         return;
 867                 }
 868
 869                 default:
 870                         next_char();
 871                         break;
 872                 }
 873         }
 874 }
 875
 876 static bool skip_till_newline(bool stop_at_non_whitespace)
 877 {
 878         bool res = false;
 879         while (true) {
 880                 switch (input.c) {
 881                 case ' ':
 882                 case '\t':
 883                         next_char();
 884                         continue;
 885
 886                 case '/':
 887                         next_char();
 888                         if (input.c == '/') {
 889                                 next_char();
 890                                 skip_line_comment();
 891                                 continue;
 892                         } else if (input.c == '*') {
 893                                 next_char();
 894                                 skip_multiline_comment();
 895                                 continue;
 896                         } else {
 897                                 put_back(input.c);
 898                                 input.c = '/';
 899                         }
 900                         return true;
 901
 902                 case NEWLINE:
 903                         return res;
 904
 905                 default:
 906                         if (stop_at_non_whitespace)
 907                                 return false;
 908                         res = true;
 909                         next_char();
 910                         continue;
 911                 }
 912         }
 913 }
 914
 915 static void skip_whitespace(void)
 916 {
 917         while (true) {
 918                 switch (input.c) {
 919                 case ' ':
 920                 case '\t':
 921                         ++info.whitespace_at_line_begin;
 922                         info.had_whitespace = true;
 923                         next_char();
 924                         continue;
 925
 926                 case NEWLINE:
 927                         info.at_line_begin  = true;
 928                         info.had_whitespace = true;
 929                         info.whitespace_at_line_begin = 0;
 930                         continue;
 931
 932                 case '/':
 933                         next_char();
 934                         if (input.c == '/') {
 935                                 next_char();
 936                                 skip_line_comment();
 937                                 continue;
 938                         } else if (input.c == '*') {
 939                                 next_char();
 940                                 skip_multiline_comment();
 941                                 continue;
 942                         } else {
 943                                 put_back(input.c);
 944                                 input.c = '/';
 945                         }
 946                         return;
 947
 948                 default:
 949                         return;
 950                 }
 951         }
 952 }
 953
 954 static inline void eat_pp(pp_token_kind_t const kind)
 955 {
 956         assert(pp_token.base.symbol->pp_ID == kind);
 957         (void) kind;
 958         next_input_token();
 959 }
 960
 961 static inline void eat_token(token_kind_t const kind)
 962 {
 963         assert(pp_token.kind == kind);
 964         (void)kind;
 965         next_input_token();
 966 }
 967
 968 static void parse_symbol(void)
 969 {
 970         assert(obstack_object_size(&symbol_obstack) == 0);
 971         while (true) {
 972                 switch (input.c) {
 973                 case DIGIT_CASES:
 974                 case SYMBOL_CASES:
 975                         obstack_1grow(&symbol_obstack, (char) input.c);
 976                         next_char();
 977                         break;
 978
 979                 case '\\':
 980                         next_char();
 981                         switch (input.c) {
 982                         {
 983                                 unsigned n;
 984                         case 'U': n = 8; goto universal;
 985                         case 'u': n = 4; goto universal;
 986 universal:
 987                                 if (!resolve_escape_sequences) {
 988                                         obstack_1grow(&symbol_obstack, '\\');
 989                                         obstack_1grow(&symbol_obstack, input.c);
 990                                 }
 991                                 next_char();
 992                                 utf32 const v = parse_universal_char(n);
 993                                 if (!is_universal_char_valid_identifier(v)) {
 994                                         if (is_universal_char_valid(v)) {
 995                                                 errorf(&input.position,
 996                                                            "universal character \\%c%0*X is not valid in an identifier",
 997                                                            n == 4 ? 'u' : 'U', (int)n, v);
 998                                         }
 999                                 } else if (obstack_object_size(&symbol_obstack) == 0 && !is_universal_char_valid_identifier_start(v)) {
1000                                         errorf(&input.position,
1001                                                    "universal character \\%c%0*X is not valid as start of an identifier",
1002                                                    n == 4 ? 'u' : 'U', (int)n, v);
1003                                 } else if (resolve_escape_sequences) {
1004                                         obstack_grow_utf8(&symbol_obstack, v);
1005                                 }
1006                                 break;
1007                         }
1008
1009                         default:
1010                                 put_back(input.c);
1011                                 input.c = '\\';
1012                                 goto end_symbol;
1013                         }
1014
1015                 default:
1016 dollar_sign:
1017                         goto end_symbol;
1018                 }
1019         }
1020
1021 end_symbol:
1022         obstack_1grow(&symbol_obstack, '\0');
1023         char *string = obstack_finish(&symbol_obstack);
1024
1025         /* might be a wide string or character constant ( L"string"/L'c' ) */
1026         if (input.c == '"' && string[0] == 'L' && string[1] == '\0') {
1027                 obstack_free(&symbol_obstack, string);
1028                 parse_string_literal(STRING_ENCODING_WIDE);
1029                 return;
1030         } else if (input.c == '\'' && string[0] == 'L' && string[1] == '\0') {
1031                 obstack_free(&symbol_obstack, string);
1032                 parse_character_constant(STRING_ENCODING_WIDE);
1033                 return;
1034         }
1035
1036         symbol_t *symbol = symbol_table_insert(string);
1037
1038         pp_token.kind        = symbol->ID;
1039         pp_token.base.symbol = symbol;
1040
1041         /* we can free the memory from symbol obstack if we already had an entry in
1042          * the symbol table */
1043         if (symbol->string != string) {
1044                 obstack_free(&symbol_obstack, string);
1045         }
1046 }
1047
1048 static void parse_number(void)
1049 {
1050         obstack_1grow(&symbol_obstack, (char) input.c);
1051         next_char();
1052
1053         while (true) {
1054                 switch (input.c) {
1055                 case '.':
1056                 case DIGIT_CASES:
1057                 case SYMBOL_CASES_WITHOUT_E_P:
1058                         obstack_1grow(&symbol_obstack, (char) input.c);
1059                         next_char();
1060                         break;
1061
1062                 case 'e':
1063                 case 'p':
1064                 case 'E':
1065                 case 'P':
1066                         obstack_1grow(&symbol_obstack, (char) input.c);
1067                         next_char();
1068                         if (input.c == '+' || input.c == '-') {
1069                                 obstack_1grow(&symbol_obstack, (char) input.c);
1070                                 next_char();
1071                         }
1072                         break;
1073
1074                 default:
1075 dollar_sign:
1076                         goto end_number;
1077                 }
1078         }
1079
1080 end_number:
1081         pp_token.kind           = T_NUMBER;
1082         pp_token.literal.string = sym_make_string(STRING_ENCODING_CHAR);
1083 }
1084
1085 #define MAYBE_PROLOG \
1086         next_char(); \
1087         switch (input.c) {
1088
1089 #define MAYBE(ch, kind) \
1090         case ch: \
1091                 next_char(); \
1092                 set_punctuator(kind); \
1093                 return;
1094
1095 #define MAYBE_DIGRAPH(ch, kind, symbol) \
1096         case ch: \
1097                 next_char(); \
1098                 set_digraph(kind, symbol); \
1099                 return;
1100
1101 #define ELSE_CODE(code) \
1102         default: \
1103                 code \
1104         }
1105
1106 #define ELSE(kind) ELSE_CODE(set_punctuator(kind); return;)
1107
1108 /** identifies and returns the next preprocessing token contained in the
1109  * input stream. No macro expansion is performed. */
1110 static void next_input_token(void)
1111 {
1112         if (next_info.had_whitespace) {
1113                 info = next_info;
1114                 next_info.had_whitespace = false;
1115         } else {
1116                 info.at_line_begin  = false;
1117                 info.had_whitespace = false;
1118         }
1119 restart:
1120         pp_token.base.source_position = input.position;
1121         pp_token.base.symbol          = NULL;
1122
1123         switch (input.c) {
1124         case ' ':
1125         case '\t':
1126                 info.whitespace_at_line_begin++;
1127                 info.had_whitespace = true;
1128                 next_char();
1129                 goto restart;
1130
1131         case NEWLINE:
1132                 info.at_line_begin            = true;
1133                 info.had_whitespace           = true;
1134                 info.whitespace_at_line_begin = 0;
1135                 goto restart;
1136
1137         case SYMBOL_CASES:
1138                 parse_symbol();
1139                 return;
1140
1141         case DIGIT_CASES:
1142                 parse_number();
1143                 return;
1144
1145         case '"':
1146                 parse_string_literal(STRING_ENCODING_CHAR);
1147                 return;
1148
1149         case '\'':
1150                 parse_character_constant(STRING_ENCODING_CHAR);
1151                 return;
1152
1153         case '.':
1154                 MAYBE_PROLOG
1155                         case '0':
1156                         case '1':
1157                         case '2':
1158                         case '3':
1159                         case '4':
1160                         case '5':
1161                         case '6':
1162                         case '7':
1163                         case '8':
1164                         case '9':
1165                                 put_back(input.c);
1166                                 input.c = '.';
1167                                 parse_number();
1168                                 return;
1169
1170                         case '.':
1171                                 MAYBE_PROLOG
1172                                 MAYBE('.', T_DOTDOTDOT)
1173                                 ELSE_CODE(
1174                                         put_back(input.c);
1175                                         input.c = '.';
1176                                         set_punctuator('.');
1177                                         return;
1178                                 )
1179                 ELSE('.')
1180         case '&':
1181                 MAYBE_PROLOG
1182                 MAYBE('&', T_ANDAND)
1183                 MAYBE('=', T_ANDEQUAL)
1184                 ELSE('&')
1185         case '*':
1186                 MAYBE_PROLOG
1187                 MAYBE('=', T_ASTERISKEQUAL)
1188                 ELSE('*')
1189         case '+':
1190                 MAYBE_PROLOG
1191                 MAYBE('+', T_PLUSPLUS)
1192                 MAYBE('=', T_PLUSEQUAL)
1193                 ELSE('+')
1194         case '-':
1195                 MAYBE_PROLOG
1196                 MAYBE('>', T_MINUSGREATER)
1197                 MAYBE('-', T_MINUSMINUS)
1198                 MAYBE('=', T_MINUSEQUAL)
1199                 ELSE('-')
1200         case '!':
1201                 MAYBE_PROLOG
1202                 MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1203                 ELSE('!')
1204         case '/':
1205                 MAYBE_PROLOG
1206                 MAYBE('=', T_SLASHEQUAL)
1207                 case '*':
1208                         next_char();
1209                         skip_multiline_comment();
1210                         goto restart;
1211                 case '/':
1212                         next_char();
1213                         skip_line_comment();
1214                         goto restart;
1215                 ELSE('/')
1216         case '%':
1217                 MAYBE_PROLOG
1218                 MAYBE_DIGRAPH('>', '}', symbol_percentgreater)
1219                 MAYBE('=', T_PERCENTEQUAL)
1220                 case ':':
1221                         MAYBE_PROLOG
1222                         case '%':
1223                                 MAYBE_PROLOG
1224                                 MAYBE_DIGRAPH(':', T_HASHHASH, symbol_percentcolonpercentcolon)
1225                                 ELSE_CODE(
1226                                         put_back(input.c);
1227                                         input.c = '%';
1228                                         goto digraph_percentcolon;
1229                                 )
1230                         ELSE_CODE(
1231 digraph_percentcolon:
1232                                 set_digraph('#', symbol_percentcolon);
1233                                 return;
1234                         )
1235                 ELSE('%')
1236         case '<':
1237                 MAYBE_PROLOG
1238                 MAYBE_DIGRAPH(':', '[', symbol_lesscolon)
1239                 MAYBE_DIGRAPH('%', '{', symbol_lesspercent)
1240                 MAYBE('=', T_LESSEQUAL)
1241                 case '<':
1242                         MAYBE_PROLOG
1243                         MAYBE('=', T_LESSLESSEQUAL)
1244                         ELSE(T_LESSLESS)
1245                 ELSE('<')
1246         case '>':
1247                 MAYBE_PROLOG
1248                 MAYBE('=', T_GREATEREQUAL)
1249                 case '>':
1250                         MAYBE_PROLOG
1251                         MAYBE('=', T_GREATERGREATEREQUAL)
1252                         ELSE(T_GREATERGREATER)
1253                 ELSE('>')
1254         case '^':
1255                 MAYBE_PROLOG
1256                 MAYBE('=', T_CARETEQUAL)
1257                 ELSE('^')
1258         case '|':
1259                 MAYBE_PROLOG
1260                 MAYBE('=', T_PIPEEQUAL)
1261                 MAYBE('|', T_PIPEPIPE)
1262                 ELSE('|')
1263         case ':':
1264                 MAYBE_PROLOG
1265                 MAYBE_DIGRAPH('>', ']', symbol_colongreater)
1266                 case ':':
1267                         if (c_mode & _CXX) {
1268                                 next_char();
1269                                 set_punctuator(T_COLONCOLON);
1270                                 return;
1271                         }
1272                         /* FALLTHROUGH */
1273                 ELSE(':')
1274         case '=':
1275                 MAYBE_PROLOG
1276                 MAYBE('=', T_EQUALEQUAL)
1277                 ELSE('=')
1278         case '#':
1279                 MAYBE_PROLOG
1280                 MAYBE('#', T_HASHHASH)
1281                 ELSE('#')
1282
1283         case '?':
1284         case '[':
1285         case ']':
1286         case '(':
1287         case ')':
1288         case '{':
1289         case '}':
1290         case '~':
1291         case ';':
1292         case ',':
1293                 set_punctuator(input.c);
1294                 next_char();
1295                 return;
1296
1297         case EOF:
1298                 if (input_stack != NULL) {
1299                         fclose(close_pp_input());
1300                         pop_restore_input();
1301                         fputc('\n', out);
1302                         if (input.c == (utf32)EOF)
1303                                 --input.position.lineno;
1304                         print_line_directive(&input.position, "2");
1305                         goto restart;
1306                 } else {
1307                         info.at_line_begin = true;
1308                         set_punctuator(T_EOF);
1309                 }
1310                 return;
1311
1312         case '\\':
1313                 next_char();
1314                 int next_c = input.c;
1315                 put_back(input.c);
1316                 input.c = '\\';
1317                 if (next_c == 'U' || next_c == 'u') {
1318                         parse_symbol();
1319                         return;
1320                 }
1321                 /* FALLTHROUGH */
1322         default:
1323 dollar_sign:
1324                 if (error_on_unknown_chars) {
1325                         errorf(&pp_token.base.source_position,
1326                                "unknown character '%lc' found\n", input.c);
1327                         next_char();
1328                         goto restart;
1329                 } else {
1330                         assert(obstack_object_size(&symbol_obstack) == 0);
1331                         obstack_grow_utf8(&symbol_obstack, input.c);
1332                         obstack_1grow(&symbol_obstack, '\0');
1333                         char     *const string = obstack_finish(&symbol_obstack);
1334                         symbol_t *const symbol = symbol_table_insert(string);
1335                         if (symbol->string != string)
1336                                 obstack_free(&symbol_obstack, string);
1337
1338                         pp_token.kind        = T_UNKNOWN_CHAR;
1339                         pp_token.base.symbol = symbol;
1340                         next_char();
1341                         return;
1342                 }
1343         }
1344 }
1345
1346 static void print_quoted_string(const char *const string)
1347 {
1348         fputc('"', out);
1349         for (const char *c = string; *c != 0; ++c) {
1350                 switch (*c) {
1351                 case '"': fputs("\\\"", out); break;
1352                 case '\\':  fputs("\\\\", out); break;
1353                 case '\a':  fputs("\\a", out); break;
1354                 case '\b':  fputs("\\b", out); break;
1355                 case '\f':  fputs("\\f", out); break;
1356                 case '\n':  fputs("\\n", out); break;
1357                 case '\r':  fputs("\\r", out); break;
1358                 case '\t':  fputs("\\t", out); break;
1359                 case '\v':  fputs("\\v", out); break;
1360                 case '\?':  fputs("\\?", out); break;
1361                 default:
1362                         if (!isprint(*c)) {
1363                                 fprintf(out, "\\%03o", (unsigned)*c);
1364                                 break;
1365                         }
1366                         fputc(*c, out);
1367                         break;
1368                 }
1369         }
1370         fputc('"', out);
1371 }
1372
1373 static void print_line_directive(const source_position_t *pos, const char *add)
1374 {
1375         if (!out)
1376                 return;
1377
1378         fprintf(out, "# %u ", pos->lineno);
1379         print_quoted_string(pos->input_name);
1380         if (add != NULL) {
1381                 fputc(' ', out);
1382                 fputs(add, out);
1383         }
1384         if (pos->is_system_header) {
1385                 fputs(" 3", out);
1386         }
1387
1388         printed_input_name = pos->input_name;
1389         input.output_line  = pos->lineno-1;
1390 }
1391
1392 static bool emit_newlines(void)
1393 {
1394         unsigned delta = pp_token.base.source_position.lineno - input.output_line;
1395         if (delta == 0)
1396                 return false;
1397
1398         if (delta >= 9) {
1399                 fputc('\n', out);
1400                 print_line_directive(&pp_token.base.source_position, NULL);
1401                 fputc('\n', out);
1402         } else {
1403                 for (unsigned i = 0; i < delta; ++i) {
1404                         fputc('\n', out);
1405                 }
1406         }
1407         input.output_line = pp_token.base.source_position.lineno;
1408
1409         for (unsigned i = 0; i < info.whitespace_at_line_begin; ++i)
1410                 fputc(' ', out);
1411
1412         return true;
1413 }
1414
1415 void set_preprocessor_output(FILE *output)
1416 {
1417         out = output;
1418         if (out != NULL) {
1419                 error_on_unknown_chars   = false;
1420                 resolve_escape_sequences = false;
1421         } else {
1422                 error_on_unknown_chars   = true;
1423                 resolve_escape_sequences = true;
1424         }
1425 }
1426
1427 void emit_pp_token(void)
1428 {
1429         if (!emit_newlines() &&
1430             (info.had_whitespace || tokens_would_paste(last_token, pp_token.kind)))
1431                 fputc(' ', out);
1432
1433         switch (pp_token.kind) {
1434         case T_NUMBER:
1435                 fputs(pp_token.literal.string.begin, out);
1436                 break;
1437
1438         case T_STRING_LITERAL:
1439                 fputs(get_string_encoding_prefix(pp_token.literal.string.encoding), out);
1440                 fputc('"', out);
1441                 fputs(pp_token.literal.string.begin, out);
1442                 fputc('"', out);
1443                 break;
1444
1445         case T_CHARACTER_CONSTANT:
1446                 fputs(get_string_encoding_prefix(pp_token.literal.string.encoding), out);
1447                 fputc('\'', out);
1448                 fputs(pp_token.literal.string.begin, out);
1449                 fputc('\'', out);
1450                 break;
1451
1452         case T_MACRO_PARAMETER:
1453                 panic("macro parameter not expanded");
1454
1455         default:
1456                 fputs(pp_token.base.symbol->string, out);
1457                 break;
1458         }
1459         last_token = pp_token.kind;
1460 }
1461
1462 static void eat_pp_directive(void)
1463 {
1464         while (!info.at_line_begin) {
1465                 next_input_token();
1466         }
1467 }
1468
1469 static bool strings_equal(const string_t *string1, const string_t *string2)
1470 {
1471         size_t size = string1->size;
1472         if (size != string2->size)
1473                 return false;
1474
1475         const char *c1 = string1->begin;
1476         const char *c2 = string2->begin;
1477         for (size_t i = 0; i < size; ++i, ++c1, ++c2) {
1478                 if (*c1 != *c2)
1479                         return false;
1480         }
1481         return true;
1482 }
1483
1484 static bool pp_tokens_equal(const token_t *token1, const token_t *token2)
1485 {
1486         if (token1->kind != token2->kind)
1487                 return false;
1488
1489         switch (token1->kind) {
1490         case T_NUMBER:
1491         case T_CHARACTER_CONSTANT:
1492         case T_STRING_LITERAL:
1493                 return strings_equal(&token1->literal.string, &token2->literal.string);
1494
1495         case T_MACRO_PARAMETER:
1496                 return token1->macro_parameter.def->symbol
1497                     == token2->macro_parameter.def->symbol;
1498
1499         default:
1500                 return token1->base.symbol == token2->base.symbol;
1501         }
1502 }
1503
1504 static bool pp_definitions_equal(const pp_definition_t *definition1,
1505                                  const pp_definition_t *definition2)
1506 {
1507         if (definition1->list_len != definition2->list_len)
1508                 return false;
1509
1510         size_t               len = definition1->list_len;
1511         const saved_token_t *t1  = definition1->token_list;
1512         const saved_token_t *t2  = definition2->token_list;
1513         for (size_t i = 0; i < len; ++i, ++t1, ++t2) {
1514                 if (!pp_tokens_equal(&t1->token, &t2->token))
1515                         return false;
1516         }
1517         return true;
1518 }
1519
1520 static bool is_defineable_token(char const *const context)
1521 {
1522         if (info.at_line_begin) {
1523                 errorf(&pp_token.base.source_position, "unexpected end of line after %s", context);
1524         }
1525
1526         symbol_t *const symbol = pp_token.base.symbol;
1527         if (!symbol)
1528                 goto no_ident;
1529
1530         if (pp_token.kind != T_IDENTIFIER) {
1531                 switch (symbol->string[0]) {
1532                 case SYMBOL_CASES:
1533 dollar_sign:
1534                         break;
1535
1536                 default:
1537 no_ident:
1538                         errorf(&pp_token.base.source_position, "expected identifier after %s, got %K", context, &pp_token);
1539                         return false;
1540                 }
1541         }
1542
1543         /* TODO turn this into a flag in pp_def. */
1544         switch (symbol->pp_ID) {
1545         /* §6.10.8:4 */
1546         case TP_defined:
1547                 errorf(&pp_token.base.source_position, "%K cannot be used as macro name in %s", &pp_token, context);
1548                 return false;
1549
1550         default:
1551                 return true;
1552         }
1553 }
1554
1555 static void parse_define_directive(void)
1556 {
1557         eat_pp(TP_define);
1558         if (skip_mode) {
1559                 eat_pp_directive();
1560                 return;
1561         }
1562
1563         assert(obstack_object_size(&pp_obstack) == 0);
1564
1565         if (!is_defineable_token("#define"))
1566                 goto error_out;
1567         symbol_t *const symbol = pp_token.base.symbol;
1568
1569         pp_definition_t *new_definition
1570                 = obstack_alloc(&pp_obstack, sizeof(new_definition[0]));
1571         memset(new_definition, 0, sizeof(new_definition[0]));
1572         new_definition->symbol          = symbol;
1573         new_definition->source_position = input.position;
1574
1575         /* this is probably the only place where spaces are significant in the
1576          * lexer (except for the fact that they separate tokens). #define b(x)
1577          * is something else than #define b (x) */
1578         if (input.c == '(') {
1579                 next_input_token();
1580                 eat_token('(');
1581
1582                 while (true) {
1583                         switch (pp_token.kind) {
1584                         case T_DOTDOTDOT:
1585                                 new_definition->is_variadic = true;
1586                                 eat_token(T_DOTDOTDOT);
1587                                 if (pp_token.kind != ')') {
1588                                         errorf(&input.position,
1589                                                         "'...' not at end of macro argument list");
1590                                         goto error_out;
1591                                 }
1592                                 break;
1593
1594                         case T_IDENTIFIER: {
1595                                 pp_definition_t parameter;
1596                                 memset(&parameter, 0, sizeof(parameter));
1597                                 parameter.source_position = pp_token.base.source_position;
1598                                 parameter.symbol          = pp_token.base.symbol;
1599                                 parameter.is_parameter    = true;
1600                                 obstack_grow(&pp_obstack, &parameter, sizeof(parameter));
1601                                 eat_token(T_IDENTIFIER);
1602
1603                                 if (pp_token.kind == ',') {
1604                                         eat_token(',');
1605                                         break;
1606                                 }
1607
1608                                 if (pp_token.kind != ')') {
1609                                         errorf(&pp_token.base.source_position,
1610                                                "expected ',' or ')' after identifier, got %K",
1611                                                &pp_token);
1612                                         goto error_out;
1613                                 }
1614                                 break;
1615                         }
1616
1617                         case ')':
1618                                 eat_token(')');
1619                                 goto finish_argument_list;
1620
1621                         default:
1622                                 errorf(&pp_token.base.source_position,
1623                                        "expected identifier, '...' or ')' in #define argument list, got %K",
1624                                        &pp_token);
1625                                 goto error_out;
1626                         }
1627                 }
1628
1629         finish_argument_list:
1630                 new_definition->has_parameters = true;
1631                 size_t size = obstack_object_size(&pp_obstack);
1632                 new_definition->n_parameters
1633                         = size / sizeof(new_definition->parameters[0]);
1634                 new_definition->parameters = obstack_finish(&pp_obstack);
1635                 for (size_t i = 0; i < new_definition->n_parameters; ++i) {
1636                         pp_definition_t *param    = &new_definition->parameters[i];
1637                         symbol_t        *symbol   = param->symbol;
1638                         pp_definition_t *previous = symbol->pp_definition;
1639                         if (previous != NULL
1640                             && previous->function_definition == new_definition) {
1641                                 errorf(&param->source_position,
1642                                        "duplicate macro parameter '%Y'", symbol);
1643                                 param->symbol = sym_anonymous;
1644                                 continue;
1645                         }
1646                         param->parent_expansion    = previous;
1647                         param->function_definition = new_definition;
1648                         symbol->pp_definition      = param;
1649                 }
1650         } else {
1651                 next_input_token();
1652         }
1653
1654         /* construct token list */
1655         assert(obstack_object_size(&pp_obstack) == 0);
1656         while (!info.at_line_begin) {
1657                 if (pp_token.kind == T_IDENTIFIER) {
1658                         const symbol_t  *symbol     = pp_token.base.symbol;
1659                         pp_definition_t *definition = symbol->pp_definition;
1660                         if (definition != NULL
1661                             && definition->function_definition == new_definition) {
1662                             pp_token.kind                = T_MACRO_PARAMETER;
1663                             pp_token.macro_parameter.def = definition;
1664                         }
1665                 }
1666                 saved_token_t saved_token;
1667                 saved_token.token = pp_token;
1668                 saved_token.had_whitespace = info.had_whitespace;
1669                 obstack_grow(&pp_obstack, &saved_token, sizeof(saved_token));
1670                 next_input_token();
1671         }
1672
1673         new_definition->list_len   = obstack_object_size(&pp_obstack)
1674                 / sizeof(new_definition->token_list[0]);
1675         new_definition->token_list = obstack_finish(&pp_obstack);
1676
1677         if (new_definition->has_parameters) {
1678                 for (size_t i = 0; i < new_definition->n_parameters; ++i) {
1679                         pp_definition_t *param      = &new_definition->parameters[i];
1680                         symbol_t        *symbol     = param->symbol;
1681                         if (symbol == sym_anonymous)
1682                                 continue;
1683                         assert(symbol->pp_definition == param);
1684                         assert(param->function_definition == new_definition);
1685                         symbol->pp_definition   = param->parent_expansion;
1686                         param->parent_expansion = NULL;
1687                 }
1688         }
1689
1690         pp_definition_t *old_definition = symbol->pp_definition;
1691         if (old_definition != NULL) {
1692                 if (!pp_definitions_equal(old_definition, new_definition)) {
1693                         warningf(WARN_OTHER, &input.position, "multiple definition of macro '%Y' (first defined %P)", symbol, &old_definition->source_position);
1694                 } else {
1695                         /* reuse the old definition */
1696                         obstack_free(&pp_obstack, new_definition);
1697                         new_definition = old_definition;
1698                 }
1699         }
1700
1701         symbol->pp_definition = new_definition;
1702         return;
1703
1704 error_out:
1705         if (obstack_object_size(&pp_obstack) > 0) {
1706                 char *ptr = obstack_finish(&pp_obstack);
1707                 obstack_free(&pp_obstack, ptr);
1708         }
1709         eat_pp_directive();
1710 }
1711
1712 static void parse_undef_directive(void)
1713 {
1714         eat_pp(TP_undef);
1715         if (skip_mode) {
1716                 eat_pp_directive();
1717                 return;
1718         }
1719
1720         if (!is_defineable_token("#undef")) {
1721                 eat_pp_directive();
1722                 return;
1723         }
1724
1725         pp_token.base.symbol->pp_definition = NULL;
1726         next_input_token();
1727
1728         if (!info.at_line_begin) {
1729                 warningf(WARN_OTHER, &input.position, "extra tokens at end of #undef directive");
1730         }
1731         eat_pp_directive();
1732 }
1733
1734 /** behind an #include we can have the special headername lexems.
1735  * They're only allowed behind an #include so they're not recognized
1736  * by the normal next_preprocessing_token. We handle them as a special
1737  * exception here */
1738 static void parse_headername(void)
1739 {
1740         const source_position_t start_position = input.position;
1741         string_t                string         = { NULL, 0, STRING_ENCODING_CHAR };
1742         assert(obstack_object_size(&symbol_obstack) == 0);
1743
1744         if (info.at_line_begin) {
1745                 parse_error("expected headername after #include");
1746                 goto finish_error;
1747         }
1748
1749         /* check wether we have a "... or <... headername */
1750         switch (input.c) {
1751         {
1752                 utf32 delimiter;
1753         case '<': delimiter = '>'; goto parse_name;
1754         case '"': delimiter = '"'; goto parse_name;
1755 parse_name:
1756                 next_char();
1757                 while (true) {
1758                         switch (input.c) {
1759                         case NEWLINE:
1760                         case EOF:
1761                                 errorf(&pp_token.base.source_position, "header name without closing '%c'", (char)delimiter);
1762                                 goto finish_error;
1763
1764                         default:
1765                                 if (input.c == delimiter) {
1766                                         next_char();
1767                                         goto finished_headername;
1768                                 } else {
1769                                         obstack_1grow(&symbol_obstack, (char)input.c);
1770                                         next_char();
1771                                 }
1772                                 break;
1773                         }
1774                 }
1775                 /* we should never be here */
1776         }
1777
1778         default:
1779                 /* TODO: do normal pp_token parsing and concatenate results */
1780                 panic("pp_token concat include not implemented yet");
1781         }
1782
1783 finished_headername:
1784         string = sym_make_string(STRING_ENCODING_CHAR);
1785
1786 finish_error:
1787         pp_token.base.source_position = start_position;
1788         pp_token.kind                 = T_HEADERNAME;
1789         pp_token.literal.string       = string;
1790 }
1791
1792 static bool do_include(bool const system_include, bool const include_next, char const *const headername)
1793 {
1794         size_t const        headername_len = strlen(headername);
1795         searchpath_entry_t *entry;
1796         if (include_next) {
1797                 entry = input.path ? input.path->next : searchpath;
1798         } else {
1799                 if (!system_include) {
1800                         /* put dirname of current input on obstack */
1801                         const char *filename   = input.position.input_name;
1802                         const char *last_slash = strrchr(filename, '/');
1803                         const char *full_name;
1804                         if (last_slash != NULL) {
1805                                 size_t len = last_slash - filename;
1806                                 obstack_grow(&symbol_obstack, filename, len + 1);
1807                                 obstack_grow0(&symbol_obstack, headername, headername_len);
1808                                 char *complete_path = obstack_finish(&symbol_obstack);
1809                                 full_name = identify_string(complete_path);
1810                         } else {
1811                                 full_name = headername;
1812                         }
1813
1814                         FILE *file = fopen(full_name, "r");
1815                         if (file != NULL) {
1816                                 switch_pp_input(file, full_name, NULL);
1817                                 return true;
1818                         }
1819                 }
1820
1821                 entry = searchpath;
1822         }
1823
1824         assert(obstack_object_size(&symbol_obstack) == 0);
1825         /* check searchpath */
1826         for (; entry; entry = entry->next) {
1827             const char *path = entry->path;
1828             size_t      len  = strlen(path);
1829                 obstack_grow(&symbol_obstack, path, len);
1830                 if (path[len-1] != '/')
1831                         obstack_1grow(&symbol_obstack, '/');
1832                 obstack_grow(&symbol_obstack, headername, headername_len+1);
1833
1834                 char *complete_path = obstack_finish(&symbol_obstack);
1835                 FILE *file          = fopen(complete_path, "r");
1836                 if (file != NULL) {
1837                         const char *filename = identify_string(complete_path);
1838                         switch_pp_input(file, filename, entry);
1839                         return true;
1840                 } else {
1841                         obstack_free(&symbol_obstack, complete_path);
1842                 }
1843         }
1844
1845         return false;
1846 }
1847
1848 static void parse_include_directive(bool const include_next)
1849 {
1850         if (skip_mode) {
1851                 eat_pp_directive();
1852                 return;
1853         }
1854
1855         /* don't eat the TP_include here!
1856          * we need an alternative parsing for the next token */
1857         skip_till_newline(true);
1858         bool system_include = input.c == '<';
1859         parse_headername();
1860         string_t headername = pp_token.literal.string;
1861         if (headername.begin == NULL) {
1862                 eat_pp_directive();
1863                 return;
1864         }
1865
1866         bool had_nonwhitespace = skip_till_newline(false);
1867         if (had_nonwhitespace) {
1868                 warningf(WARN_OTHER, &pp_token.base.source_position,
1869                          "extra tokens at end of #include directive");
1870         }
1871
1872         if (n_inputs > INCLUDE_LIMIT) {
1873                 errorf(&pp_token.base.source_position, "#include nested too deeply");
1874                 /* eat \n or EOF */
1875                 next_input_token();
1876                 return;
1877         }
1878
1879         /* switch inputs */
1880         info.whitespace_at_line_begin = 0;
1881         info.had_whitespace           = false;
1882         info.at_line_begin            = true;
1883         emit_newlines();
1884         push_input();
1885         bool res = do_include(system_include, include_next, pp_token.literal.string.begin);
1886         if (res) {
1887                 next_input_token();
1888         } else {
1889                 errorf(&pp_token.base.source_position, "failed including '%S': %s", &pp_token.literal.string, strerror(errno));
1890                 pop_restore_input();
1891         }
1892 }
1893
1894 static pp_conditional_t *push_conditional(void)
1895 {
1896         pp_conditional_t *conditional
1897                 = obstack_alloc(&pp_obstack, sizeof(*conditional));
1898         memset(conditional, 0, sizeof(*conditional));
1899
1900         conditional->parent = conditional_stack;
1901         conditional_stack   = conditional;
1902
1903         return conditional;
1904 }
1905
1906 static void pop_conditional(void)
1907 {
1908         assert(conditional_stack != NULL);
1909         conditional_stack = conditional_stack->parent;
1910 }
1911
1912 void check_unclosed_conditionals(void)
1913 {
1914         while (conditional_stack != NULL) {
1915                 pp_conditional_t *conditional = conditional_stack;
1916
1917                 if (conditional->in_else) {
1918                         errorf(&conditional->source_position, "unterminated #else");
1919                 } else {
1920                         errorf(&conditional->source_position, "unterminated condition");
1921                 }
1922                 pop_conditional();
1923         }
1924 }
1925
1926 static void parse_ifdef_ifndef_directive(bool const is_ifdef)
1927 {
1928         bool condition;
1929         eat_pp(is_ifdef ? TP_ifdef : TP_ifndef);
1930
1931         if (skip_mode) {
1932                 eat_pp_directive();
1933                 pp_conditional_t *conditional = push_conditional();
1934                 conditional->source_position  = pp_token.base.source_position;
1935                 conditional->skip             = true;
1936                 return;
1937         }
1938
1939         if (pp_token.kind != T_IDENTIFIER || info.at_line_begin) {
1940                 errorf(&pp_token.base.source_position,
1941                        "expected identifier after #%s, got %K",
1942                        is_ifdef ? "ifdef" : "ifndef", &pp_token);
1943                 eat_pp_directive();
1944
1945                 /* just take the true case in the hope to avoid further errors */
1946                 condition = true;
1947         } else {
1948                 /* evaluate wether we are in true or false case */
1949                 condition = (bool)pp_token.base.symbol->pp_definition == is_ifdef;
1950                 eat_token(T_IDENTIFIER);
1951
1952                 if (!info.at_line_begin) {
1953                         errorf(&pp_token.base.source_position,
1954                                "extra tokens at end of #%s",
1955                                is_ifdef ? "ifdef" : "ifndef");
1956                         eat_pp_directive();
1957                 }
1958         }
1959
1960         pp_conditional_t *conditional = push_conditional();
1961         conditional->source_position  = pp_token.base.source_position;
1962         conditional->condition        = condition;
1963
1964         if (!condition) {
1965                 skip_mode = true;
1966         }
1967 }
1968
1969 static void parse_else_directive(void)
1970 {
1971         eat_pp(TP_else);
1972
1973         if (!info.at_line_begin) {
1974                 if (!skip_mode) {
1975                         warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #else");
1976                 }
1977                 eat_pp_directive();
1978         }
1979
1980         pp_conditional_t *conditional = conditional_stack;
1981         if (conditional == NULL) {
1982                 errorf(&pp_token.base.source_position, "#else without prior #if");
1983                 return;
1984         }
1985
1986         if (conditional->in_else) {
1987                 errorf(&pp_token.base.source_position,
1988                        "#else after #else (condition started %P)",
1989                        &conditional->source_position);
1990                 skip_mode = true;
1991                 return;
1992         }
1993
1994         conditional->in_else = true;
1995         if (!conditional->skip) {
1996                 skip_mode = conditional->condition;
1997         }
1998         conditional->source_position = pp_token.base.source_position;
1999 }
2000
2001 static void parse_endif_directive(void)
2002 {
2003         eat_pp(TP_endif);
2004
2005         if (!info.at_line_begin) {
2006                 if (!skip_mode) {
2007                         warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #endif");
2008                 }
2009                 eat_pp_directive();
2010         }
2011
2012         pp_conditional_t *conditional = conditional_stack;
2013         if (conditional == NULL) {
2014                 errorf(&pp_token.base.source_position, "#endif without prior #if");
2015                 return;
2016         }
2017
2018         if (!conditional->skip) {
2019                 skip_mode = false;
2020         }
2021         pop_conditional();
2022 }
2023
2024 typedef enum stdc_pragma_kind_t {
2025         STDC_UNKNOWN,
2026         STDC_FP_CONTRACT,
2027         STDC_FENV_ACCESS,
2028         STDC_CX_LIMITED_RANGE
2029 } stdc_pragma_kind_t;
2030
2031 typedef enum stdc_pragma_value_kind_t {
2032         STDC_VALUE_UNKNOWN,
2033         STDC_VALUE_ON,
2034         STDC_VALUE_OFF,
2035         STDC_VALUE_DEFAULT
2036 } stdc_pragma_value_kind_t;
2037
2038 static void parse_pragma_directive(void)
2039 {
2040         eat_pp(TP_pragma);
2041         if (skip_mode) {
2042                 eat_pp_directive();
2043                 return;
2044         }
2045
2046         if (pp_token.kind != T_IDENTIFIER) {
2047                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
2048                          "expected identifier after #pragma");
2049                 eat_pp_directive();
2050                 return;
2051         }
2052
2053         stdc_pragma_kind_t kind = STDC_UNKNOWN;
2054         if (pp_token.base.symbol->pp_ID == TP_STDC && c_mode & _C99) {
2055                 /* a STDC pragma */
2056                 next_input_token();
2057
2058                 switch (pp_token.base.symbol->pp_ID) {
2059                 case TP_FP_CONTRACT:      kind = STDC_FP_CONTRACT;      break;
2060                 case TP_FENV_ACCESS:      kind = STDC_FENV_ACCESS;      break;
2061                 case TP_CX_LIMITED_RANGE: kind = STDC_CX_LIMITED_RANGE; break;
2062                 default:                  break;
2063                 }
2064                 if (kind != STDC_UNKNOWN) {
2065                         next_input_token();
2066                         stdc_pragma_value_kind_t value;
2067                         switch (pp_token.base.symbol->pp_ID) {
2068                         case TP_ON:      value = STDC_VALUE_ON;      break;
2069                         case TP_OFF:     value = STDC_VALUE_OFF;     break;
2070                         case TP_DEFAULT: value = STDC_VALUE_DEFAULT; break;
2071                         default:         value = STDC_VALUE_UNKNOWN; break;
2072                         }
2073                         if (value == STDC_VALUE_UNKNOWN) {
2074                                 kind = STDC_UNKNOWN;
2075                                 errorf(&pp_token.base.source_position, "bad STDC pragma argument");
2076                         }
2077                 }
2078         }
2079         eat_pp_directive();
2080         if (kind == STDC_UNKNOWN) {
2081                 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
2082                          "encountered unknown #pragma");
2083         }
2084 }
2085
2086 static void parse_line_directive(void)
2087 {
2088         if (pp_token.kind != T_NUMBER) {
2089                 if (!skip_mode)
2090                         parse_error("expected integer");
2091         } else {
2092                 char      *end;
2093                 long const line = strtol(pp_token.literal.string.begin, &end, 0);
2094                 if (*end == '\0') {
2095                         /* use offset -1 as this is about the next line */
2096                         input.position.lineno = line - 1;
2097                         /* force output of line */
2098                         input.output_line = input.position.lineno - 20;
2099                 } else {
2100                         if (!skip_mode) {
2101                                 errorf(&input.position, "'%S' is not a valid line number",
2102                                            &pp_token.literal.string);
2103                         }
2104                 }
2105                 next_input_token();
2106                 if (info.at_line_begin)
2107                         return;
2108         }
2109         if (pp_token.kind == T_STRING_LITERAL
2110             && pp_token.literal.string.encoding == STRING_ENCODING_CHAR) {
2111                 input.position.input_name       = pp_token.literal.string.begin;
2112                 input.position.is_system_header = false;
2113                 next_input_token();
2114
2115                 /* attempt to parse numeric flags as outputted by gcc preprocessor */
2116                 while (!info.at_line_begin && pp_token.kind == T_NUMBER) {
2117                         /* flags:
2118                          * 1 - indicates start of a new file
2119                          * 2 - indicates return from a file
2120                          * 3 - indicates system header
2121                          * 4 - indicates implicit extern "C" in C++ mode
2122                          *
2123                          * currently we're only interested in "3"
2124                          */
2125                         if (streq(pp_token.literal.string.begin, "3")) {
2126                                 input.position.is_system_header = true;
2127                         }
2128                         next_input_token();
2129                 }
2130         }
2131
2132         eat_pp_directive();
2133 }
2134
2135 static void parse_error_directive(void)
2136 {
2137         if (skip_mode) {
2138                 eat_pp_directive();
2139                 return;
2140         }
2141
2142         bool const old_resolve_escape_sequences = resolve_escape_sequences;
2143         resolve_escape_sequences = false;
2144
2145         source_position_t const pos = pp_token.base.source_position;
2146         do {
2147                 if (info.had_whitespace && obstack_object_size(&pp_obstack) != 0)
2148                         obstack_1grow(&pp_obstack, ' ');
2149
2150                 switch (pp_token.kind) {
2151                 case T_NUMBER: {
2152                         string_t const *const str = &pp_token.literal.string;
2153                         obstack_grow(&pp_obstack, str->begin, str->size);
2154                         break;
2155                 }
2156
2157                 {
2158                         char delim;
2159                 case T_STRING_LITERAL:     delim =  '"'; goto string;
2160                 case T_CHARACTER_CONSTANT: delim = '\''; goto string;
2161 string:;
2162                         string_t const *const str = &pp_token.literal.string;
2163                         char     const *const enc = get_string_encoding_prefix(str->encoding);
2164                         obstack_printf(&pp_obstack, "%s%c%s%c", enc, delim, str->begin, delim);
2165                         break;
2166                 }
2167
2168                 default: {
2169                         char const *const str = pp_token.base.symbol->string;
2170                         obstack_grow(&pp_obstack, str, strlen(str));
2171                         break;
2172                 }
2173                 }
2174
2175                 next_input_token();
2176         } while (!info.at_line_begin);
2177
2178         resolve_escape_sequences = old_resolve_escape_sequences;
2179
2180         obstack_1grow(&pp_obstack, '\0');
2181         char *const str = obstack_finish(&pp_obstack);
2182         errorf(&pos, "#%s", str);
2183         obstack_free(&pp_obstack, str);
2184 }
2185
2186 static void parse_preprocessing_directive(void)
2187 {
2188         eat_token('#');
2189
2190         if (info.at_line_begin) {
2191                 /* empty directive */
2192                 return;
2193         }
2194
2195         if (pp_token.base.symbol) {
2196                 switch (pp_token.base.symbol->pp_ID) {
2197                 case TP_define:       parse_define_directive();            break;
2198                 case TP_else:         parse_else_directive();              break;
2199                 case TP_endif:        parse_endif_directive();             break;
2200                 case TP_error:        parse_error_directive();             break;
2201                 case TP_ifdef:        parse_ifdef_ifndef_directive(true);  break;
2202                 case TP_ifndef:       parse_ifdef_ifndef_directive(false); break;
2203                 case TP_include:      parse_include_directive(false);      break;
2204                 case TP_include_next: parse_include_directive(true);       break;
2205                 case TP_line:         next_input_token(); goto line_directive;
2206                 case TP_pragma:       parse_pragma_directive();            break;
2207                 case TP_undef:        parse_undef_directive();             break;
2208                 default:              goto skip;
2209                 }
2210         } else if (pp_token.kind == T_NUMBER) {
2211 line_directive:
2212                 parse_line_directive();
2213         } else {
2214 skip:
2215                 if (!skip_mode) {
2216                         errorf(&pp_token.base.source_position, "invalid preprocessing directive #%K", &pp_token);
2217                 }
2218                 eat_pp_directive();
2219         }
2220
2221         assert(info.at_line_begin);
2222 }
2223
2224 static void finish_current_argument(void)
2225 {
2226         if (current_argument == NULL)
2227                 return;
2228         size_t size = obstack_object_size(&pp_obstack);
2229         current_argument->list_len   = size/sizeof(current_argument->token_list[0]);
2230         current_argument->token_list = obstack_finish(&pp_obstack);
2231 }
2232
2233 void next_preprocessing_token(void)
2234 {
2235 restart:
2236         if (!expand_next()) {
2237                 do {
2238                         next_input_token();
2239                         while (pp_token.kind == '#' && info.at_line_begin) {
2240                                 parse_preprocessing_directive();
2241                         }
2242                 } while (skip_mode && pp_token.kind != T_EOF);
2243         }
2244
2245         const token_kind_t kind = pp_token.kind;
2246         if (current_call == NULL || argument_expanding != NULL) {
2247                 symbol_t *const symbol = pp_token.base.symbol;
2248                 if (symbol) {
2249                         if (kind == T_MACRO_PARAMETER) {
2250                                 assert(current_expansion != NULL);
2251                                 start_expanding(pp_token.macro_parameter.def);
2252                                 goto restart;
2253                         }
2254
2255                         pp_definition_t *const pp_definition = symbol->pp_definition;
2256                         if (pp_definition != NULL && !pp_definition->is_expanding) {
2257                                 if (pp_definition->has_parameters) {
2258
2259                                         /* check if next token is a '(' */
2260                                         whitespace_info_t old_info   = info;
2261                                         token_kind_t      next_token = peek_expansion();
2262                                         if (next_token == T_EOF) {
2263                                                 info.at_line_begin  = false;
2264                                                 info.had_whitespace = false;
2265                                                 skip_whitespace();
2266                                                 if (input.c == '(') {
2267                                                         next_token = '(';
2268                                                 }
2269                                         }
2270
2271                                         if (next_token == '(') {
2272                                                 if (current_expansion == NULL)
2273                                                         expansion_pos = pp_token.base.source_position;
2274                                                 next_preprocessing_token();
2275                                                 assert(pp_token.kind == '(');
2276
2277                                                 pp_definition->parent_expansion = current_expansion;
2278                                                 current_call              = pp_definition;
2279                                                 current_call->expand_pos  = 0;
2280                                                 current_call->expand_info = old_info;
2281                                                 if (current_call->n_parameters > 0) {
2282                                                         current_argument = &current_call->parameters[0];
2283                                                         assert(argument_brace_count == 0);
2284                                                 }
2285                                                 goto restart;
2286                                         } else {
2287                                                 /* skip_whitespaces() skipped newlines and whitespace,
2288                                                  * remember results for next token */
2289                                                 next_info = info;
2290                                                 info      = old_info;
2291                                                 return;
2292                                         }
2293                                 } else {
2294                                         if (current_expansion == NULL)
2295                                                 expansion_pos = pp_token.base.source_position;
2296                                         start_expanding(pp_definition);
2297                                         goto restart;
2298                                 }
2299                         }
2300                 }
2301         }
2302
2303         if (current_call != NULL) {
2304                 /* current_call != NULL */
2305                 if (kind == '(') {
2306                         ++argument_brace_count;
2307                 } else if (kind == ')') {
2308                         if (argument_brace_count > 0) {
2309                                 --argument_brace_count;
2310                         } else {
2311                                 finish_current_argument();
2312                                 assert(kind == ')');
2313                                 start_expanding(current_call);
2314                                 info = current_call->expand_info;
2315                                 current_call     = NULL;
2316                                 current_argument = NULL;
2317                                 goto restart;
2318                         }
2319                 } else if (kind == ',' && argument_brace_count == 0) {
2320                         finish_current_argument();
2321                         current_call->expand_pos++;
2322                         if (current_call->expand_pos >= current_call->n_parameters) {
2323                                 errorf(&pp_token.base.source_position,
2324                                            "too many arguments passed for macro '%Y'",
2325                                            current_call->symbol);
2326                                 current_argument = NULL;
2327                         } else {
2328                                 current_argument
2329                                         = &current_call->parameters[current_call->expand_pos];
2330                         }
2331                         goto restart;
2332                 } else if (kind == T_MACRO_PARAMETER) {
2333                         /* parameters have to be fully expanded before being used as
2334                          * parameters for another macro-call */
2335                         assert(current_expansion != NULL);
2336                         pp_definition_t *argument = pp_token.macro_parameter.def;
2337                         argument_expanding = argument;
2338                         start_expanding(argument);
2339                         goto restart;
2340                 } else if (kind == T_EOF) {
2341                         errorf(&expansion_pos,
2342                                "reached end of file while parsing arguments for '%Y'",
2343                                current_call->symbol);
2344                         return;
2345                 }
2346                 if (current_argument != NULL) {
2347                         saved_token_t saved;
2348                         saved.token = pp_token;
2349                         saved.had_whitespace = info.had_whitespace;
2350                         obstack_grow(&pp_obstack, &saved, sizeof(saved));
2351                 }
2352                 goto restart;
2353         }
2354 }
2355
2356
2357 static void prepend_include_path(const char *path)
2358 {
2359         searchpath_entry_t *entry = OALLOCZ(&config_obstack, searchpath_entry_t);
2360         entry->path = path;
2361         entry->next = searchpath;
2362         searchpath  = entry;
2363 }
2364
2365 static void setup_include_path(void)
2366 {
2367         /* built-in paths */
2368         prepend_include_path("/usr/include");
2369
2370         /* parse environment variable */
2371         const char *cpath = getenv("CPATH");
2372         if (cpath != NULL && *cpath != '\0') {
2373                 const char *begin = cpath;
2374                 const char *c;
2375                 do {
2376                         c = begin;
2377                         while (*c != '\0' && *c != ':')
2378                                 ++c;
2379
2380                         size_t len = c-begin;
2381                         if (len == 0) {
2382                                 /* for gcc compatibility (Matze: I would expect that
2383                                  * nothing happens for an empty entry...) */
2384                                 prepend_include_path(".");
2385                         } else {
2386                                 char *const string = obstack_copy0(&config_obstack, begin, len);
2387                                 prepend_include_path(string);
2388                         }
2389
2390                         begin = c+1;
2391                         /* skip : */
2392                         if (*begin == ':')
2393                                 ++begin;
2394                 } while(*c != '\0');
2395         }
2396 }
2397
2398 static void input_error(unsigned const delta_lines, unsigned const delta_cols, char const *const message)
2399 {
2400         source_position_t pos = pp_token.base.source_position;
2401         pos.lineno += delta_lines;
2402         pos.colno  += delta_cols;
2403         errorf(&pos, "%s", message);
2404 }
2405
2406 void init_preprocessor(void)
2407 {
2408         init_symbols();
2409
2410         obstack_init(&config_obstack);
2411         obstack_init(&pp_obstack);
2412         obstack_init(&input_obstack);
2413         strset_init(&stringset);
2414
2415         setup_include_path();
2416
2417         set_input_error_callback(input_error);
2418 }
2419
2420 void exit_preprocessor(void)
2421 {
2422         obstack_free(&input_obstack, NULL);
2423         obstack_free(&pp_obstack, NULL);
2424         obstack_free(&config_obstack, NULL);
2425
2426         strset_destroy(&stringset);
2427 }
2428
2429 int pptest_main(int argc, char **argv);
2430 int pptest_main(int argc, char **argv)
2431 {
2432         init_symbol_table();
2433         init_preprocessor();
2434         init_tokens();
2435
2436         error_on_unknown_chars   = false;
2437         resolve_escape_sequences = false;
2438
2439         /* simplistic commandline parser */
2440         const char *filename = NULL;
2441         const char *output = NULL;
2442         for (int i = 1; i < argc; ++i) {
2443                 const char *opt = argv[i];
2444                 if (streq(opt, "-I")) {
2445                         prepend_include_path(argv[++i]);
2446                         continue;
2447                 } else if (streq(opt, "-E")) {
2448                         /* ignore */
2449                 } else if (streq(opt, "-o")) {
2450                         output = argv[++i];
2451                         continue;
2452                 } else if (opt[0] == '-') {
2453                         fprintf(stderr, "Unknown option '%s'\n", opt);
2454                 } else {
2455                         if (filename != NULL)
2456                                 fprintf(stderr, "Multiple inputs not supported\n");
2457                         filename = argv[i];
2458                 }
2459         }
2460         if (filename == NULL) {
2461                 fprintf(stderr, "No input specified\n");
2462                 return 1;
2463         }
2464
2465         if (output == NULL) {
2466                 out = stdout;
2467         } else {
2468                 out = fopen(output, "w");
2469                 if (out == NULL) {
2470                         fprintf(stderr, "Couldn't open output '%s'\n", output);
2471                         return 1;
2472                 }
2473         }
2474
2475         /* just here for gcc compatibility */
2476         fprintf(out, "# 1 \"%s\"\n", filename);
2477         fprintf(out, "# 1 \"<built-in>\"\n");
2478         fprintf(out, "# 1 \"<command-line>\"\n");
2479
2480         FILE *file = fopen(filename, "r");
2481         if (file == NULL) {
2482                 fprintf(stderr, "Couldn't open input '%s'\n", filename);
2483                 return 1;
2484         }
2485         switch_pp_input(file, filename, NULL);
2486
2487         for (;;) {
2488                 next_preprocessing_token();
2489                 if (pp_token.kind == T_EOF)
2490                         break;
2491                 emit_pp_token();
2492         }
2493
2494         fputc('\n', out);
2495         check_unclosed_conditionals();
2496         fclose(close_pp_input());
2497         if (out != stdout)
2498                 fclose(out);
2499
2500         exit_tokens();
2501         exit_preprocessor();
2502         exit_symbol_table();
2503
2504         return 0;
2505 }