tccpp.c

   1 /*
   2  *  TCC - Tiny C Compiler
   3  *
   4  *  Copyright (c) 2001-2004 Fabrice Bellard
   5  *
   6  * This library is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2 of the License, or (at your option) any later version.
  10  *
  11  * This library is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with this library; if not, write to the Free Software
  18  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  19  */
  20
  21 #include "tcc.h"
  22
  23 /********************************************************/
  24 /* global variables */
  25
  26 ST_DATA int tok_flags;
  27 /* additional informations about token */
  28 #define TOK_FLAG_BOL   0x0001 /* beginning of line before */
  29 #define TOK_FLAG_BOF   0x0002 /* beginning of file before */
  30 #define TOK_FLAG_ENDIF 0x0004 /* a endif was found matching starting #ifdef */
  31 #define TOK_FLAG_EOF   0x0008 /* end of file */
  32
  33 ST_DATA int parse_flags;
  34 #define PARSE_FLAG_PREPROCESS 0x0001 /* activate preprocessing */
  35 #define PARSE_FLAG_TOK_NUM    0x0002 /* return numbers instead of TOK_PPNUM */
  36 #define PARSE_FLAG_LINEFEED   0x0004 /* line feed is returned as a
  37                                         token. line feed is also
  38                                         returned at eof */
  39 #define PARSE_FLAG_ASM_COMMENTS 0x0008 /* '#' can be used for line comment */
  40 #define PARSE_FLAG_SPACES     0x0010 /* next() returns space tokens (for -E) */
  41
  42 ST_DATA struct BufferedFile *file;
  43 ST_DATA int ch, tok;
  44 ST_DATA CValue tokc;
  45 ST_DATA const int *macro_ptr;
  46 ST_DATA CString tokcstr; /* current parsed string, if any */
  47
  48 /* display benchmark infos */
  49 ST_DATA int total_lines;
  50 ST_DATA int total_bytes;
  51 ST_DATA int tok_ident;
  52 ST_DATA TokenSym **table_ident;
  53
  54 /* ------------------------------------------------------------------------- */
  55
  56 static int *macro_ptr_allocated;
  57 static const int *unget_saved_macro_ptr;
  58 static int unget_saved_buffer[TOK_MAX_SIZE + 1];
  59 static int unget_buffer_enabled;
  60 static TokenSym *hash_ident[TOK_HASH_SIZE];
  61 static char token_buf[STRING_MAX_SIZE + 1];
  62 /* true if isid(c) || isnum(c) */
  63 static unsigned char isidnum_table[256-CH_EOF];
  64
  65 static const char tcc_keywords[] =
  66 #define DEF(id, str) str "\0"
  67 #include "tcctok.h"
  68 #undef DEF
  69 ;
  70
  71 /* WARNING: the content of this string encodes token numbers */
  72 static const unsigned char tok_two_chars[] =
  73 /* outdated -- gr
  74     "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
  75     "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
  76 */{
  77     '<','=', TOK_LE,
  78     '>','=', TOK_GE,
  79     '!','=', TOK_NE,
  80     '&','&', TOK_LAND,
  81     '|','|', TOK_LOR,
  82     '+','+', TOK_INC,
  83     '-','-', TOK_DEC,
  84     '=','=', TOK_EQ,
  85     '<','<', TOK_SHL,
  86     '>','>', TOK_SAR,
  87     '+','=', TOK_A_ADD,
  88     '-','=', TOK_A_SUB,
  89     '*','=', TOK_A_MUL,
  90     '/','=', TOK_A_DIV,
  91     '%','=', TOK_A_MOD,
  92     '&','=', TOK_A_AND,
  93     '^','=', TOK_A_XOR,
  94     '|','=', TOK_A_OR,
  95     '-','>', TOK_ARROW,
  96     '.','.', 0xa8, // C++ token ?
  97     '#','#', TOK_TWOSHARPS,
  98     0
  99 };
 100
 101 struct macro_level {
 102     struct macro_level *prev;
 103     const int *p;
 104 };
 105
 106 static void next_nomacro_spc(void);
 107 static void macro_subst(
 108     TokenString *tok_str,
 109     Sym **nested_list,
 110     const int *macro_str,
 111     struct macro_level **can_read_stream
 112     );
 113
 114 ST_FUNC void skip(int c)
 115 {
 116     if (tok != c)
 117         tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
 118     next();
 119 }
 120
 121 ST_FUNC void expect(const char *msg)
 122 {
 123     tcc_error("%s expected", msg);
 124 }
 125
 126 /* ------------------------------------------------------------------------- */
 127 /* CString handling */
 128 static void cstr_realloc(CString *cstr, int new_size)
 129 {
 130     int size;
 131     void *data;
 132
 133     size = cstr->size_allocated;
 134     if (size == 0)
 135         size = 8; /* no need to allocate a too small first string */
 136     while (size < new_size)
 137         size = size * 2;
 138     data = tcc_realloc(cstr->data_allocated, size);
 139     cstr->data_allocated = data;
 140     cstr->size_allocated = size;
 141     cstr->data = data;
 142 }
 143
 144 /* add a byte */
 145 ST_FUNC void cstr_ccat(CString *cstr, int ch)
 146 {
 147     int size;
 148     size = cstr->size + 1;
 149     if (size > cstr->size_allocated)
 150         cstr_realloc(cstr, size);
 151     ((unsigned char *)cstr->data)[size - 1] = ch;
 152     cstr->size = size;
 153 }
 154
 155 ST_FUNC void cstr_cat(CString *cstr, const char *str)
 156 {
 157     int c;
 158     for(;;) {
 159         c = *str;
 160         if (c == '\0')
 161             break;
 162         cstr_ccat(cstr, c);
 163         str++;
 164     }
 165 }
 166
 167 /* add a wide char */
 168 ST_FUNC void cstr_wccat(CString *cstr, int ch)
 169 {
 170     int size;
 171     size = cstr->size + sizeof(nwchar_t);
 172     if (size > cstr->size_allocated)
 173         cstr_realloc(cstr, size);
 174     *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
 175     cstr->size = size;
 176 }
 177
 178 ST_FUNC void cstr_new(CString *cstr)
 179 {
 180     memset(cstr, 0, sizeof(CString));
 181 }
 182
 183 /* free string and reset it to NULL */
 184 ST_FUNC void cstr_free(CString *cstr)
 185 {
 186     tcc_free(cstr->data_allocated);
 187     cstr_new(cstr);
 188 }
 189
 190 /* reset string to empty */
 191 ST_FUNC void cstr_reset(CString *cstr)
 192 {
 193     cstr->size = 0;
 194 }
 195
 196 /* XXX: unicode ? */
 197 static void add_char(CString *cstr, int c)
 198 {
 199     if (c == '\'' || c == '\"' || c == '\\') {
 200         /* XXX: could be more precise if char or string */
 201         cstr_ccat(cstr, '\\');
 202     }
 203     if (c >= 32 && c <= 126) {
 204         cstr_ccat(cstr, c);
 205     } else {
 206         cstr_ccat(cstr, '\\');
 207         if (c == '\n') {
 208             cstr_ccat(cstr, 'n');
 209         } else {
 210             cstr_ccat(cstr, '0' + ((c >> 6) & 7));
 211             cstr_ccat(cstr, '0' + ((c >> 3) & 7));
 212             cstr_ccat(cstr, '0' + (c & 7));
 213         }
 214     }
 215 }
 216
 217 /* ------------------------------------------------------------------------- */
 218 /* allocate a new token */
 219 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
 220 {
 221     TokenSym *ts, **ptable;
 222     int i;
 223
 224     if (tok_ident >= SYM_FIRST_ANOM)
 225         tcc_error("memory full (symbols)");
 226
 227     /* expand token table if needed */
 228     i = tok_ident - TOK_IDENT;
 229     if ((i % TOK_ALLOC_INCR) == 0) {
 230         ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
 231         table_ident = ptable;
 232     }
 233
 234     ts = tcc_malloc(sizeof(TokenSym) + len);
 235     table_ident[i] = ts;
 236     ts->tok = tok_ident++;
 237     ts->sym_define = NULL;
 238     ts->sym_label = NULL;
 239     ts->sym_struct = NULL;
 240     ts->sym_identifier = NULL;
 241     ts->len = len;
 242     ts->hash_next = NULL;
 243     memcpy(ts->str, str, len);
 244     ts->str[len] = '\0';
 245     *pts = ts;
 246     return ts;
 247 }
 248
 249 #define TOK_HASH_INIT 1
 250 #define TOK_HASH_FUNC(h, c) ((h) * 263 + (c))
 251
 252 /* find a token and add it if not found */
 253 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
 254 {
 255     TokenSym *ts, **pts;
 256     int i;
 257     unsigned int h;
 258
 259     h = TOK_HASH_INIT;
 260     for(i=0;i<len;i++)
 261         h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
 262     h &= (TOK_HASH_SIZE - 1);
 263
 264     pts = &hash_ident[h];
 265     for(;;) {
 266         ts = *pts;
 267         if (!ts)
 268             break;
 269         if (ts->len == len && !memcmp(ts->str, str, len))
 270             return ts;
 271         pts = &(ts->hash_next);
 272     }
 273     return tok_alloc_new(pts, str, len);
 274 }
 275
 276 /* XXX: buffer overflow */
 277 /* XXX: float tokens */
 278 ST_FUNC char *get_tok_str(int v, CValue *cv)
 279 {
 280     static char buf[STRING_MAX_SIZE + 1];
 281     static CString cstr_buf;
 282     CString *cstr;
 283     char *p;
 284     int i, len;
 285
 286     /* NOTE: to go faster, we give a fixed buffer for small strings */
 287     cstr_reset(&cstr_buf);
 288     cstr_buf.data = buf;
 289     cstr_buf.size_allocated = sizeof(buf);
 290     p = buf;
 291
 292 /* just an explanation, should never happen:
 293     if (v <= TOK_LINENUM && v >= TOK_CINT && cv == NULL)
 294         tcc_error("internal error: get_tok_str"); */
 295
 296     switch(v) {
 297     case TOK_CINT:
 298     case TOK_CUINT:
 299         /* XXX: not quite exact, but only useful for testing */
 300         sprintf(p, "%u", cv->ui);
 301         break;
 302     case TOK_CLLONG:
 303     case TOK_CULLONG:
 304         /* XXX: not quite exact, but only useful for testing  */
 305 #ifdef _WIN32
 306         sprintf(p, "%u", (unsigned)cv->ull);
 307 #else
 308         sprintf(p, "%llu", cv->ull);
 309 #endif
 310         break;
 311     case TOK_LCHAR:
 312         cstr_ccat(&cstr_buf, 'L');
 313     case TOK_CCHAR:
 314         cstr_ccat(&cstr_buf, '\'');
 315         add_char(&cstr_buf, cv->i);
 316         cstr_ccat(&cstr_buf, '\'');
 317         cstr_ccat(&cstr_buf, '\0');
 318         break;
 319     case TOK_PPNUM:
 320         cstr = cv->cstr;
 321         len = cstr->size - 1;
 322         for(i=0;i<len;i++)
 323             add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 324         cstr_ccat(&cstr_buf, '\0');
 325         break;
 326     case TOK_LSTR:
 327         cstr_ccat(&cstr_buf, 'L');
 328     case TOK_STR:
 329         cstr = cv->cstr;
 330         cstr_ccat(&cstr_buf, '\"');
 331         if (v == TOK_STR) {
 332             len = cstr->size - 1;
 333             for(i=0;i<len;i++)
 334                 add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 335         } else {
 336             len = (cstr->size / sizeof(nwchar_t)) - 1;
 337             for(i=0;i<len;i++)
 338                 add_char(&cstr_buf, ((nwchar_t *)cstr->data)[i]);
 339         }
 340         cstr_ccat(&cstr_buf, '\"');
 341         cstr_ccat(&cstr_buf, '\0');
 342         break;
 343
 344     case TOK_CFLOAT:
 345     case TOK_CDOUBLE:
 346     case TOK_CLDOUBLE:
 347     case TOK_LINENUM:
 348         return NULL; /* should not happen */
 349
 350     /* above tokens have value, the ones below don't */
 351
 352     case TOK_LT:
 353         v = '<';
 354         goto addv;
 355     case TOK_GT:
 356         v = '>';
 357         goto addv;
 358     case TOK_DOTS:
 359         return strcpy(p, "...");
 360     case TOK_A_SHL:
 361         return strcpy(p, "<<=");
 362     case TOK_A_SAR:
 363         return strcpy(p, ">>=");
 364     default:
 365         if (v < TOK_IDENT) {
 366             /* search in two bytes table */
 367             const unsigned char *q = tok_two_chars;
 368             while (*q) {
 369                 if (q[2] == v) {
 370                     *p++ = q[0];
 371                     *p++ = q[1];
 372                     *p = '\0';
 373                     return buf;
 374                 }
 375                 q += 3;
 376             }
 377         addv:
 378             *p++ = v;
 379             *p = '\0';
 380         } else if (v < tok_ident) {
 381             return table_ident[v - TOK_IDENT]->str;
 382         } else if (v >= SYM_FIRST_ANOM) {
 383             /* special name for anonymous symbol */
 384             sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
 385         } else {
 386             /* should never happen */
 387             return NULL;
 388         }
 389         break;
 390     }
 391     return cstr_buf.data;
 392 }
 393
 394 /* fill input buffer and peek next char */
 395 static int tcc_peekc_slow(BufferedFile *bf)
 396 {
 397     int len;
 398     /* only tries to read if really end of buffer */
 399     if (bf->buf_ptr >= bf->buf_end) {
 400         if (bf->fd != -1) {
 401 #if defined(PARSE_DEBUG)
 402             len = 8;
 403 #else
 404             len = IO_BUF_SIZE;
 405 #endif
 406             len = read(bf->fd, bf->buffer, len);
 407             if (len < 0)
 408                 len = 0;
 409         } else {
 410             len = 0;
 411         }
 412         total_bytes += len;
 413         bf->buf_ptr = bf->buffer;
 414         bf->buf_end = bf->buffer + len;
 415         *bf->buf_end = CH_EOB;
 416     }
 417     if (bf->buf_ptr < bf->buf_end) {
 418         return bf->buf_ptr[0];
 419     } else {
 420         bf->buf_ptr = bf->buf_end;
 421         return CH_EOF;
 422     }
 423 }
 424
 425 /* return the current character, handling end of block if necessary
 426    (but not stray) */
 427 ST_FUNC int handle_eob(void)
 428 {
 429     return tcc_peekc_slow(file);
 430 }
 431
 432 /* read next char from current input file and handle end of input buffer */
 433 ST_INLN void inp(void)
 434 {
 435     ch = *(++(file->buf_ptr));
 436     /* end of buffer/file handling */
 437     if (ch == CH_EOB)
 438         ch = handle_eob();
 439 }
 440
 441 /* handle '\[\r]\n' */
 442 static int handle_stray_noerror(void)
 443 {
 444     while (ch == '\\') {
 445         inp();
 446         if (ch == '\n') {
 447             file->line_num++;
 448             inp();
 449         } else if (ch == '\r') {
 450             inp();
 451             if (ch != '\n')
 452                 goto fail;
 453             file->line_num++;
 454             inp();
 455         } else {
 456         fail:
 457             return 1;
 458         }
 459     }
 460     return 0;
 461 }
 462
 463 static void handle_stray(void)
 464 {
 465     if (handle_stray_noerror())
 466         tcc_error("stray '\\' in program");
 467 }
 468
 469 /* skip the stray and handle the \\n case. Output an error if
 470    incorrect char after the stray */
 471 static int handle_stray1(uint8_t *p)
 472 {
 473     int c;
 474
 475     if (p >= file->buf_end) {
 476         file->buf_ptr = p;
 477         c = handle_eob();
 478         p = file->buf_ptr;
 479         if (c == '\\')
 480             goto parse_stray;
 481     } else {
 482     parse_stray:
 483         file->buf_ptr = p;
 484         ch = *p;
 485         handle_stray();
 486         p = file->buf_ptr;
 487         c = *p;
 488     }
 489     return c;
 490 }
 491
 492 /* handle just the EOB case, but not stray */
 493 #define PEEKC_EOB(c, p)\
 494 {\
 495     p++;\
 496     c = *p;\
 497     if (c == '\\') {\
 498         file->buf_ptr = p;\
 499         c = handle_eob();\
 500         p = file->buf_ptr;\
 501     }\
 502 }
 503
 504 /* handle the complicated stray case */
 505 #define PEEKC(c, p)\
 506 {\
 507     p++;\
 508     c = *p;\
 509     if (c == '\\') {\
 510         c = handle_stray1(p);\
 511         p = file->buf_ptr;\
 512     }\
 513 }
 514
 515 /* input with '\[\r]\n' handling. Note that this function cannot
 516    handle other characters after '\', so you cannot call it inside
 517    strings or comments */
 518 ST_FUNC void minp(void)
 519 {
 520     inp();
 521     if (ch == '\\')
 522         handle_stray();
 523 }
 524
 525
 526 /* single line C++ comments */
 527 static uint8_t *parse_line_comment(uint8_t *p)
 528 {
 529     int c;
 530
 531     p++;
 532     for(;;) {
 533         c = *p;
 534     redo:
 535         if (c == '\n' || c == CH_EOF) {
 536             break;
 537         } else if (c == '\\') {
 538             file->buf_ptr = p;
 539             c = handle_eob();
 540             p = file->buf_ptr;
 541             if (c == '\\') {
 542                 PEEKC_EOB(c, p);
 543                 if (c == '\n') {
 544                     file->line_num++;
 545                     PEEKC_EOB(c, p);
 546                 } else if (c == '\r') {
 547                     PEEKC_EOB(c, p);
 548                     if (c == '\n') {
 549                         file->line_num++;
 550                         PEEKC_EOB(c, p);
 551                     }
 552                 }
 553             } else {
 554                 goto redo;
 555             }
 556         } else {
 557             p++;
 558         }
 559     }
 560     return p;
 561 }
 562
 563 /* C comments */
 564 ST_FUNC uint8_t *parse_comment(uint8_t *p)
 565 {
 566     int c;
 567
 568     p++;
 569     for(;;) {
 570         /* fast skip loop */
 571         for(;;) {
 572             c = *p;
 573             if (c == '\n' || c == '*' || c == '\\')
 574                 break;
 575             p++;
 576             c = *p;
 577             if (c == '\n' || c == '*' || c == '\\')
 578                 break;
 579             p++;
 580         }
 581         /* now we can handle all the cases */
 582         if (c == '\n') {
 583             file->line_num++;
 584             p++;
 585         } else if (c == '*') {
 586             p++;
 587             for(;;) {
 588                 c = *p;
 589                 if (c == '*') {
 590                     p++;
 591                 } else if (c == '/') {
 592                     goto end_of_comment;
 593                 } else if (c == '\\') {
 594                     file->buf_ptr = p;
 595                     c = handle_eob();
 596                     p = file->buf_ptr;
 597                     if (c == '\\') {
 598                         /* skip '\[\r]\n', otherwise just skip the stray */
 599                         while (c == '\\') {
 600                             PEEKC_EOB(c, p);
 601                             if (c == '\n') {
 602                                 file->line_num++;
 603                                 PEEKC_EOB(c, p);
 604                             } else if (c == '\r') {
 605                                 PEEKC_EOB(c, p);
 606                                 if (c == '\n') {
 607                                     file->line_num++;
 608                                     PEEKC_EOB(c, p);
 609                                 }
 610                             } else {
 611                                 goto after_star;
 612                             }
 613                         }
 614                     }
 615                 } else {
 616                     break;
 617                 }
 618             }
 619         after_star: ;
 620         } else {
 621             /* stray, eob or eof */
 622             file->buf_ptr = p;
 623             c = handle_eob();
 624             p = file->buf_ptr;
 625             if (c == CH_EOF) {
 626                 tcc_error("unexpected end of file in comment");
 627             } else if (c == '\\') {
 628                 p++;
 629             }
 630         }
 631     }
 632  end_of_comment:
 633     p++;
 634     return p;
 635 }
 636
 637 #define cinp minp
 638
 639 static inline void skip_spaces(void)
 640 {
 641     while (is_space(ch))
 642         cinp();
 643 }
 644
 645 static inline int check_space(int t, int *spc)
 646 {
 647     if (is_space(t)) {
 648         if (*spc)
 649             return 1;
 650         *spc = 1;
 651     } else
 652         *spc = 0;
 653     return 0;
 654 }
 655
 656 /* parse a string without interpreting escapes */
 657 static uint8_t *parse_pp_string(uint8_t *p,
 658                                 int sep, CString *str)
 659 {
 660     int c;
 661     p++;
 662     for(;;) {
 663         c = *p;
 664         if (c == sep) {
 665             break;
 666         } else if (c == '\\') {
 667             file->buf_ptr = p;
 668             c = handle_eob();
 669             p = file->buf_ptr;
 670             if (c == CH_EOF) {
 671             unterminated_string:
 672                 /* XXX: indicate line number of start of string */
 673                 tcc_error("missing terminating %c character", sep);
 674             } else if (c == '\\') {
 675                 /* escape : just skip \[\r]\n */
 676                 PEEKC_EOB(c, p);
 677                 if (c == '\n') {
 678                     file->line_num++;
 679                     p++;
 680                 } else if (c == '\r') {
 681                     PEEKC_EOB(c, p);
 682                     if (c != '\n')
 683                         expect("'\n' after '\r'");
 684                     file->line_num++;
 685                     p++;
 686                 } else if (c == CH_EOF) {
 687                     goto unterminated_string;
 688                 } else {
 689                     if (str) {
 690                         cstr_ccat(str, '\\');
 691                         cstr_ccat(str, c);
 692                     }
 693                     p++;
 694                 }
 695             }
 696         } else if (c == '\n') {
 697             file->line_num++;
 698             goto add_char;
 699         } else if (c == '\r') {
 700             PEEKC_EOB(c, p);
 701             if (c != '\n') {
 702                 if (str)
 703                     cstr_ccat(str, '\r');
 704             } else {
 705                 file->line_num++;
 706                 goto add_char;
 707             }
 708         } else {
 709         add_char:
 710             if (str)
 711                 cstr_ccat(str, c);
 712             p++;
 713         }
 714     }
 715     p++;
 716     return p;
 717 }
 718
 719 /* skip block of text until #else, #elif or #endif. skip also pairs of
 720    #if/#endif */
 721 static void preprocess_skip(void)
 722 {
 723     int a, start_of_line, c, in_warn_or_error;
 724     uint8_t *p;
 725
 726     p = file->buf_ptr;
 727     a = 0;
 728 redo_start:
 729     start_of_line = 1;
 730     in_warn_or_error = 0;
 731     for(;;) {
 732     redo_no_start:
 733         c = *p;
 734         switch(c) {
 735         case ' ':
 736         case '\t':
 737         case '\f':
 738         case '\v':
 739         case '\r':
 740             p++;
 741             goto redo_no_start;
 742         case '\n':
 743             file->line_num++;
 744             p++;
 745             goto redo_start;
 746         case '\\':
 747             file->buf_ptr = p;
 748             c = handle_eob();
 749             if (c == CH_EOF) {
 750                 expect("#endif");
 751             } else if (c == '\\') {
 752                 ch = file->buf_ptr[0];
 753                 handle_stray_noerror();
 754             }
 755             p = file->buf_ptr;
 756             goto redo_no_start;
 757         /* skip strings */
 758         case '\"':
 759         case '\'':
 760             if (in_warn_or_error)
 761                 goto _default;
 762             p = parse_pp_string(p, c, NULL);
 763             break;
 764         /* skip comments */
 765         case '/':
 766             if (in_warn_or_error)
 767                 goto _default;
 768             file->buf_ptr = p;
 769             ch = *p;
 770             minp();
 771             p = file->buf_ptr;
 772             if (ch == '*') {
 773                 p = parse_comment(p);
 774             } else if (ch == '/') {
 775                 p = parse_line_comment(p);
 776             }
 777             break;
 778         case '#':
 779             p++;
 780             if (start_of_line) {
 781                 file->buf_ptr = p;
 782                 next_nomacro();
 783                 p = file->buf_ptr;
 784                 if (a == 0 &&
 785                     (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
 786                     goto the_end;
 787                 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
 788                     a++;
 789                 else if (tok == TOK_ENDIF)
 790                     a--;
 791                 else if( tok == TOK_ERROR || tok == TOK_WARNING)
 792                     in_warn_or_error = 1;
 793                 else if (tok == TOK_LINEFEED)
 794                     goto redo_start;
 795             }
 796             break;
 797 _default:
 798         default:
 799             p++;
 800             break;
 801         }
 802         start_of_line = 0;
 803     }
 804  the_end: ;
 805     file->buf_ptr = p;
 806 }
 807
 808 /* ParseState handling */
 809
 810 /* XXX: currently, no include file info is stored. Thus, we cannot display
 811    accurate messages if the function or data definition spans multiple
 812    files */
 813
 814 /* save current parse state in 's' */
 815 ST_FUNC void save_parse_state(ParseState *s)
 816 {
 817     s->line_num = file->line_num;
 818     s->macro_ptr = macro_ptr;
 819     s->tok = tok;
 820     s->tokc = tokc;
 821 }
 822
 823 /* restore parse state from 's' */
 824 ST_FUNC void restore_parse_state(ParseState *s)
 825 {
 826     file->line_num = s->line_num;
 827     macro_ptr = s->macro_ptr;
 828     tok = s->tok;
 829     tokc = s->tokc;
 830 }
 831
 832 /* return the number of additional 'ints' necessary to store the
 833    token */
 834 static inline int tok_ext_size(int t)
 835 {
 836     switch(t) {
 837         /* 4 bytes */
 838     case TOK_CINT:
 839     case TOK_CUINT:
 840     case TOK_CCHAR:
 841     case TOK_LCHAR:
 842     case TOK_CFLOAT:
 843     case TOK_LINENUM:
 844         return 1;
 845     case TOK_STR:
 846     case TOK_LSTR:
 847     case TOK_PPNUM:
 848         tcc_error("unsupported token");
 849         return 1;
 850     case TOK_CDOUBLE:
 851     case TOK_CLLONG:
 852     case TOK_CULLONG:
 853         return 2;
 854     case TOK_CLDOUBLE:
 855         return LDOUBLE_SIZE / 4;
 856     default:
 857         return 0;
 858     }
 859 }
 860
 861 /* token string handling */
 862
 863 ST_INLN void tok_str_new(TokenString *s)
 864 {
 865     s->str = NULL;
 866     s->len = 0;
 867     s->allocated_len = 0;
 868     s->last_line_num = -1;
 869 }
 870
 871 ST_FUNC void tok_str_free(int *str)
 872 {
 873     tcc_free(str);
 874 }
 875
 876 static int *tok_str_realloc(TokenString *s)
 877 {
 878     int *str, len;
 879
 880     if (s->allocated_len == 0) {
 881         len = 8;
 882     } else {
 883         len = s->allocated_len * 2;
 884     }
 885     str = tcc_realloc(s->str, len * sizeof(int));
 886     s->allocated_len = len;
 887     s->str = str;
 888     return str;
 889 }
 890
 891 ST_FUNC void tok_str_add(TokenString *s, int t)
 892 {
 893     int len, *str;
 894
 895     len = s->len;
 896     str = s->str;
 897     if (len >= s->allocated_len)
 898         str = tok_str_realloc(s);
 899     str[len++] = t;
 900     s->len = len;
 901 }
 902
 903 static void tok_str_add2(TokenString *s, int t, CValue *cv)
 904 {
 905     int len, *str;
 906
 907     len = s->len;
 908     str = s->str;
 909
 910     /* allocate space for worst case */
 911     if (len + TOK_MAX_SIZE > s->allocated_len)
 912         str = tok_str_realloc(s);
 913     str[len++] = t;
 914     switch(t) {
 915     case TOK_CINT:
 916     case TOK_CUINT:
 917     case TOK_CCHAR:
 918     case TOK_LCHAR:
 919     case TOK_CFLOAT:
 920     case TOK_LINENUM:
 921         str[len++] = cv->tab[0];
 922         break;
 923     case TOK_PPNUM:
 924     case TOK_STR:
 925     case TOK_LSTR:
 926         {
 927             int nb_words;
 928             CString *cstr;
 929
 930             nb_words = (sizeof(CString) + cv->cstr->size + 3) >> 2;
 931             while ((len + nb_words) > s->allocated_len)
 932                 str = tok_str_realloc(s);
 933             cstr = (CString *)(str + len);
 934             cstr->data = NULL;
 935             cstr->size = cv->cstr->size;
 936             cstr->data_allocated = NULL;
 937             cstr->size_allocated = cstr->size;
 938             memcpy((char *)cstr + sizeof(CString),
 939                    cv->cstr->data, cstr->size);
 940             len += nb_words;
 941         }
 942         break;
 943     case TOK_CDOUBLE:
 944     case TOK_CLLONG:
 945     case TOK_CULLONG:
 946 #if LDOUBLE_SIZE == 8
 947     case TOK_CLDOUBLE:
 948 #endif
 949         str[len++] = cv->tab[0];
 950         str[len++] = cv->tab[1];
 951         break;
 952 #if LDOUBLE_SIZE == 12
 953     case TOK_CLDOUBLE:
 954         str[len++] = cv->tab[0];
 955         str[len++] = cv->tab[1];
 956         str[len++] = cv->tab[2];
 957 #elif LDOUBLE_SIZE == 16
 958     case TOK_CLDOUBLE:
 959         str[len++] = cv->tab[0];
 960         str[len++] = cv->tab[1];
 961         str[len++] = cv->tab[2];
 962         str[len++] = cv->tab[3];
 963 #elif LDOUBLE_SIZE != 8
 964 #error add long double size support
 965 #endif
 966         break;
 967     default:
 968         break;
 969     }
 970     s->len = len;
 971 }
 972
 973 /* add the current parse token in token string 's' */
 974 ST_FUNC void tok_str_add_tok(TokenString *s)
 975 {
 976     CValue cval;
 977
 978     /* save line number info */
 979     if (file->line_num != s->last_line_num) {
 980         s->last_line_num = file->line_num;
 981         cval.i = s->last_line_num;
 982         tok_str_add2(s, TOK_LINENUM, &cval);
 983     }
 984     tok_str_add2(s, tok, &tokc);
 985 }
 986
 987 /* get a token from an integer array and increment pointer
 988    accordingly. we code it as a macro to avoid pointer aliasing. */
 989 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
 990 {
 991     const int *p = *pp;
 992     int n, *tab;
 993
 994     tab = cv->tab;
 995     switch(*t = *p++) {
 996     case TOK_CINT:
 997     case TOK_CUINT:
 998     case TOK_CCHAR:
 999     case TOK_LCHAR:
1000     case TOK_CFLOAT:
1001     case TOK_LINENUM:
1002         tab[0] = *p++;
1003         break;
1004     case TOK_STR:
1005     case TOK_LSTR:
1006     case TOK_PPNUM:
1007         cv->cstr = (CString *)p;
1008         cv->cstr->data = (char *)p + sizeof(CString);
1009         p += (sizeof(CString) + cv->cstr->size + 3) >> 2;
1010         break;
1011     case TOK_CDOUBLE:
1012     case TOK_CLLONG:
1013     case TOK_CULLONG:
1014         n = 2;
1015         goto copy;
1016     case TOK_CLDOUBLE:
1017 #if LDOUBLE_SIZE == 16
1018         n = 4;
1019 #elif LDOUBLE_SIZE == 12
1020         n = 3;
1021 #elif LDOUBLE_SIZE == 8
1022         n = 2;
1023 #else
1024 # error add long double size support
1025 #endif
1026     copy:
1027         do
1028             *tab++ = *p++;
1029         while (--n);
1030         break;
1031     default:
1032         break;
1033     }
1034     *pp = p;
1035 }
1036
1037 static int macro_is_equal(const int *a, const int *b)
1038 {
1039     char buf[STRING_MAX_SIZE + 1];
1040     CValue cv;
1041     int t;
1042     while (*a && *b) {
1043         TOK_GET(&t, &a, &cv);
1044         pstrcpy(buf, sizeof buf, get_tok_str(t, &cv));
1045         TOK_GET(&t, &b, &cv);
1046         if (strcmp(buf, get_tok_str(t, &cv)))
1047             return 0;
1048     }
1049     return !(*a || *b);
1050 }
1051
1052 /* defines handling */
1053 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1054 {
1055     Sym *s;
1056
1057     s = define_find(v);
1058     if (s && !macro_is_equal(s->d, str))
1059         tcc_warning("%s redefined", get_tok_str(v, NULL));
1060
1061     s = sym_push2(&define_stack, v, macro_type, 0);
1062     s->d = str;
1063     s->next = first_arg;
1064     table_ident[v - TOK_IDENT]->sym_define = s;
1065 }
1066
1067 /* undefined a define symbol. Its name is just set to zero */
1068 ST_FUNC void define_undef(Sym *s)
1069 {
1070     int v;
1071     v = s->v;
1072     if (v >= TOK_IDENT && v < tok_ident)
1073         table_ident[v - TOK_IDENT]->sym_define = NULL;
1074     s->v = 0;
1075 }
1076
1077 ST_INLN Sym *define_find(int v)
1078 {
1079     v -= TOK_IDENT;
1080     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1081         return NULL;
1082     return table_ident[v]->sym_define;
1083 }
1084
1085 /* free define stack until top reaches 'b' */
1086 ST_FUNC void free_defines(Sym *b)
1087 {
1088     Sym *top, *top1;
1089     int v;
1090
1091     top = define_stack;
1092     while (top != b) {
1093         top1 = top->prev;
1094         /* do not free args or predefined defines */
1095         if (top->d)
1096             tok_str_free(top->d);
1097         v = top->v;
1098         if (v >= TOK_IDENT && v < tok_ident)
1099             table_ident[v - TOK_IDENT]->sym_define = NULL;
1100         sym_free(top);
1101         top = top1;
1102     }
1103     define_stack = b;
1104 }
1105
1106 /* label lookup */
1107 ST_FUNC Sym *label_find(int v)
1108 {
1109     v -= TOK_IDENT;
1110     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1111         return NULL;
1112     return table_ident[v]->sym_label;
1113 }
1114
1115 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1116 {
1117     Sym *s, **ps;
1118     s = sym_push2(ptop, v, 0, 0);
1119     s->r = flags;
1120     ps = &table_ident[v - TOK_IDENT]->sym_label;
1121     if (ptop == &global_label_stack) {
1122         /* modify the top most local identifier, so that
1123            sym_identifier will point to 's' when popped */
1124         while (*ps != NULL)
1125             ps = &(*ps)->prev_tok;
1126     }
1127     s->prev_tok = *ps;
1128     *ps = s;
1129     return s;
1130 }
1131
1132 /* pop labels until element last is reached. Look if any labels are
1133    undefined. Define symbols if '&&label' was used. */
1134 ST_FUNC void label_pop(Sym **ptop, Sym *slast)
1135 {
1136     Sym *s, *s1;
1137     for(s = *ptop; s != slast; s = s1) {
1138         s1 = s->prev;
1139         if (s->r == LABEL_DECLARED) {
1140             tcc_warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1141         } else if (s->r == LABEL_FORWARD) {
1142                 tcc_error("label '%s' used but not defined",
1143                       get_tok_str(s->v, NULL));
1144         } else {
1145             if (s->c) {
1146                 /* define corresponding symbol. A size of
1147                    1 is put. */
1148                 put_extern_sym(s, cur_text_section, s->jnext, 1);
1149             }
1150         }
1151         /* remove label */
1152         table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1153         sym_free(s);
1154     }
1155     *ptop = slast;
1156 }
1157
1158 /* eval an expression for #if/#elif */
1159 static int expr_preprocess(void)
1160 {
1161     int c, t;
1162     TokenString str;
1163
1164     tok_str_new(&str);
1165     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1166         next(); /* do macro subst */
1167         if (tok == TOK_DEFINED) {
1168             next_nomacro();
1169             t = tok;
1170             if (t == '(')
1171                 next_nomacro();
1172             c = define_find(tok) != 0;
1173             if (t == '(')
1174                 next_nomacro();
1175             tok = TOK_CINT;
1176             tokc.i = c;
1177         } else if (tok >= TOK_IDENT) {
1178             /* if undefined macro */
1179             tok = TOK_CINT;
1180             tokc.i = 0;
1181         }
1182         tok_str_add_tok(&str);
1183     }
1184     tok_str_add(&str, -1); /* simulate end of file */
1185     tok_str_add(&str, 0);
1186     /* now evaluate C constant expression */
1187     macro_ptr = str.str;
1188     next();
1189     c = expr_const();
1190     macro_ptr = NULL;
1191     tok_str_free(str.str);
1192     return c != 0;
1193 }
1194
1195 #if defined(PARSE_DEBUG) || defined(PP_DEBUG)
1196 static void tok_print(int *str)
1197 {
1198     int t;
1199     CValue cval;
1200
1201     printf("<");
1202     while (1) {
1203         TOK_GET(&t, &str, &cval);
1204         if (!t)
1205             break;
1206         printf("%s", get_tok_str(t, &cval));
1207     }
1208     printf(">\n");
1209 }
1210 #endif
1211
1212 /* parse after #define */
1213 ST_FUNC void parse_define(void)
1214 {
1215     Sym *s, *first, **ps;
1216     int v, t, varg, is_vaargs, spc, ptok, macro_list_start;
1217     TokenString str;
1218
1219     v = tok;
1220     if (v < TOK_IDENT)
1221         tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1222     /* XXX: should check if same macro (ANSI) */
1223     first = NULL;
1224     t = MACRO_OBJ;
1225     /* '(' must be just after macro definition for MACRO_FUNC */
1226     next_nomacro_spc();
1227     if (tok == '(') {
1228         next_nomacro();
1229         ps = &first;
1230         while (tok != ')') {
1231             varg = tok;
1232             next_nomacro();
1233             is_vaargs = 0;
1234             if (varg == TOK_DOTS) {
1235                 varg = TOK___VA_ARGS__;
1236                 is_vaargs = 1;
1237             } else if (tok == TOK_DOTS && gnu_ext) {
1238                 is_vaargs = 1;
1239                 next_nomacro();
1240             }
1241             if (varg < TOK_IDENT)
1242                 tcc_error("badly punctuated parameter list");
1243             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1244             *ps = s;
1245             ps = &s->next;
1246             if (tok != ',')
1247                 break;
1248             next_nomacro();
1249         }
1250         if (tok == ')')
1251             next_nomacro_spc();
1252         t = MACRO_FUNC;
1253     }
1254     tok_str_new(&str);
1255     spc = 2;
1256     /* EOF testing necessary for '-D' handling */
1257     ptok = 0;
1258     macro_list_start = 1;
1259     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1260         if (!macro_list_start && spc == 2 && tok == TOK_TWOSHARPS)
1261             tcc_error("'##' invalid at start of macro");
1262         ptok = tok;
1263         /* remove spaces around ## and after '#' */
1264         if (TOK_TWOSHARPS == tok) {
1265             if (1 == spc)
1266                 --str.len;
1267             spc = 2;
1268         } else if ('#' == tok) {
1269             spc = 2;
1270         } else if (check_space(tok, &spc)) {
1271             goto skip;
1272         }
1273         tok_str_add2(&str, tok, &tokc);
1274     skip:
1275         next_nomacro_spc();
1276         macro_list_start = 0;
1277     }
1278     if (ptok == TOK_TWOSHARPS)
1279         tcc_error("'##' invalid at end of macro");
1280     if (spc == 1)
1281         --str.len; /* remove trailing space */
1282     tok_str_add(&str, 0);
1283 #ifdef PP_DEBUG
1284     printf("define %s %d: ", get_tok_str(v, NULL), t);
1285     tok_print(str.str);
1286 #endif
1287     define_push(v, t, str.str, first);
1288 }
1289
1290 static inline int hash_cached_include(const char *filename)
1291 {
1292     const unsigned char *s;
1293     unsigned int h;
1294
1295     h = TOK_HASH_INIT;
1296     s = (unsigned char *) filename;
1297     while (*s) {
1298         h = TOK_HASH_FUNC(h, *s);
1299         s++;
1300     }
1301     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1302     return h;
1303 }
1304
1305 static CachedInclude *search_cached_include(TCCState *s1, const char *filename)
1306 {
1307     CachedInclude *e;
1308     int i, h;
1309     h = hash_cached_include(filename);
1310     i = s1->cached_includes_hash[h];
1311     for(;;) {
1312         if (i == 0)
1313             break;
1314         e = s1->cached_includes[i - 1];
1315         if (0 == PATHCMP(e->filename, filename))
1316             return e;
1317         i = e->hash_next;
1318     }
1319     return NULL;
1320 }
1321
1322 static inline void add_cached_include(TCCState *s1, const char *filename, int ifndef_macro)
1323 {
1324     CachedInclude *e;
1325     int h;
1326
1327     if (search_cached_include(s1, filename))
1328         return;
1329 #ifdef INC_DEBUG
1330     printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1331 #endif
1332     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1333     strcpy(e->filename, filename);
1334     e->ifndef_macro = ifndef_macro;
1335     dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1336     /* add in hash table */
1337     h = hash_cached_include(filename);
1338     e->hash_next = s1->cached_includes_hash[h];
1339     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1340 }
1341
1342 static void pragma_parse(TCCState *s1)
1343 {
1344     int val;
1345
1346     next();
1347     if (tok == TOK_pack) {
1348         /*
1349           This may be:
1350           #pragma pack(1) // set
1351           #pragma pack() // reset to default
1352           #pragma pack(push,1) // push & set
1353           #pragma pack(pop) // restore previous
1354         */
1355         next();
1356         skip('(');
1357         if (tok == TOK_ASM_pop) {
1358             next();
1359             if (s1->pack_stack_ptr <= s1->pack_stack) {
1360             stk_error:
1361                 tcc_error("out of pack stack");
1362             }
1363             s1->pack_stack_ptr--;
1364         } else {
1365             val = 0;
1366             if (tok != ')') {
1367                 if (tok == TOK_ASM_push) {
1368                     next();
1369                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1370                         goto stk_error;
1371                     s1->pack_stack_ptr++;
1372                     skip(',');
1373                 }
1374                 if (tok != TOK_CINT) {
1375                 pack_error:
1376                     tcc_error("invalid pack pragma");
1377                 }
1378                 val = tokc.i;
1379                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1380                     goto pack_error;
1381                 next();
1382             }
1383             *s1->pack_stack_ptr = val;
1384             skip(')');
1385         }
1386     }
1387 }
1388
1389 /* is_bof is true if first non space token at beginning of file */
1390 ST_FUNC void preprocess(int is_bof)
1391 {
1392     TCCState *s1 = tcc_state;
1393     int i, c, n, saved_parse_flags;
1394     char buf[1024], *q;
1395     Sym *s;
1396
1397     saved_parse_flags = parse_flags;
1398     parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM |
1399         PARSE_FLAG_LINEFEED;
1400     next_nomacro();
1401  redo:
1402     switch(tok) {
1403     case TOK_DEFINE:
1404         next_nomacro();
1405         parse_define();
1406         break;
1407     case TOK_UNDEF:
1408         next_nomacro();
1409         s = define_find(tok);
1410         /* undefine symbol by putting an invalid name */
1411         if (s)
1412             define_undef(s);
1413         break;
1414     case TOK_INCLUDE:
1415     case TOK_INCLUDE_NEXT:
1416         ch = file->buf_ptr[0];
1417         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1418         skip_spaces();
1419         if (ch == '<') {
1420             c = '>';
1421             goto read_name;
1422         } else if (ch == '\"') {
1423             c = ch;
1424         read_name:
1425             inp();
1426             q = buf;
1427             while (ch != c && ch != '\n' && ch != CH_EOF) {
1428                 if ((q - buf) < sizeof(buf) - 1)
1429                     *q++ = ch;
1430                 if (ch == '\\') {
1431                     if (handle_stray_noerror() == 0)
1432                         --q;
1433                 } else
1434                     inp();
1435             }
1436             *q = '\0';
1437             minp();
1438 #if 0
1439             /* eat all spaces and comments after include */
1440             /* XXX: slightly incorrect */
1441             while (ch1 != '\n' && ch1 != CH_EOF)
1442                 inp();
1443 #endif
1444         } else {
1445             /* computed #include : either we have only strings or
1446                we have anything enclosed in '<>' */
1447             next();
1448             buf[0] = '\0';
1449             if (tok == TOK_STR) {
1450                 while (tok != TOK_LINEFEED) {
1451                     if (tok != TOK_STR) {
1452                     include_syntax:
1453                         tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1454                     }
1455                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1456                     next();
1457                 }
1458                 c = '\"';
1459             } else {
1460                 int len;
1461                 while (tok != TOK_LINEFEED) {
1462                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1463                     next();
1464                 }
1465                 len = strlen(buf);
1466                 /* check syntax and remove '<>' */
1467                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1468                     goto include_syntax;
1469                 memmove(buf, buf + 1, len - 2);
1470                 buf[len - 2] = '\0';
1471                 c = '>';
1472             }
1473         }
1474
1475         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1476             tcc_error("#include recursion too deep");
1477         /* store current file in stack, but increment stack later below */
1478         *s1->include_stack_ptr = file;
1479
1480         n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1481         for (i = -2; i < n; ++i) {
1482             char buf1[sizeof file->filename];
1483             CachedInclude *e;
1484             BufferedFile **f;
1485             const char *path;
1486
1487             if (i == -2) {
1488                 /* check absolute include path */
1489                 if (!IS_ABSPATH(buf))
1490                     continue;
1491                 buf1[0] = 0;
1492                 i = n; /* force end loop */
1493
1494             } else if (i == -1) {
1495                 /* search in current dir if "header.h" */
1496                 if (c != '\"')
1497                     continue;
1498                 path = file->filename;
1499                 pstrncpy(buf1, path, tcc_basename(path) - path);
1500
1501             } else {
1502                 /* search in all the include paths */
1503                 if (i < s1->nb_include_paths)
1504                     path = s1->include_paths[i];
1505                 else
1506                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1507                 pstrcpy(buf1, sizeof(buf1), path);
1508                 pstrcat(buf1, sizeof(buf1), "/");
1509             }
1510
1511             pstrcat(buf1, sizeof(buf1), buf);
1512
1513             if (tok == TOK_INCLUDE_NEXT)
1514                 for (f = s1->include_stack_ptr; f >= s1->include_stack; --f)
1515                     if (0 == PATHCMP((*f)->filename, buf1)) {
1516 #ifdef INC_DEBUG
1517                         printf("%s: #include_next skipping %s\n", file->filename, buf1);
1518 #endif
1519                         goto include_trynext;
1520                     }
1521
1522             e = search_cached_include(s1, buf1);
1523             if (e && define_find(e->ifndef_macro)) {
1524                 /* no need to parse the include because the 'ifndef macro'
1525                    is defined */
1526 #ifdef INC_DEBUG
1527                 printf("%s: skipping cached %s\n", file->filename, buf1);
1528 #endif
1529                 goto include_done;
1530             }
1531
1532             if (tcc_open(s1, buf1) < 0)
1533 include_trynext:
1534                 continue;
1535
1536 #ifdef INC_DEBUG
1537             printf("%s: including %s\n", file->prev->filename, file->filename);
1538 #endif
1539             /* update target deps */
1540             dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1541                     tcc_strdup(buf1));
1542             /* push current file in stack */
1543             ++s1->include_stack_ptr;
1544             /* add include file debug info */
1545             if (s1->do_debug)
1546                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1547             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1548             ch = file->buf_ptr[0];
1549             goto the_end;
1550         }
1551         tcc_error("include file '%s' not found", buf);
1552 include_done:
1553         break;
1554     case TOK_IFNDEF:
1555         c = 1;
1556         goto do_ifdef;
1557     case TOK_IF:
1558         c = expr_preprocess();
1559         goto do_if;
1560     case TOK_IFDEF:
1561         c = 0;
1562     do_ifdef:
1563         next_nomacro();
1564         if (tok < TOK_IDENT)
1565             tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1566         if (is_bof) {
1567             if (c) {
1568 #ifdef INC_DEBUG
1569                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1570 #endif
1571                 file->ifndef_macro = tok;
1572             }
1573         }
1574         c = (define_find(tok) != 0) ^ c;
1575     do_if:
1576         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1577             tcc_error("memory full (ifdef)");
1578         *s1->ifdef_stack_ptr++ = c;
1579         goto test_skip;
1580     case TOK_ELSE:
1581         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1582             tcc_error("#else without matching #if");
1583         if (s1->ifdef_stack_ptr[-1] & 2)
1584             tcc_error("#else after #else");
1585         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1586         goto test_else;
1587     case TOK_ELIF:
1588         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1589             tcc_error("#elif without matching #if");
1590         c = s1->ifdef_stack_ptr[-1];
1591         if (c > 1)
1592             tcc_error("#elif after #else");
1593         /* last #if/#elif expression was true: we skip */
1594         if (c == 1)
1595             goto skip;
1596         c = expr_preprocess();
1597         s1->ifdef_stack_ptr[-1] = c;
1598     test_else:
1599         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1600             file->ifndef_macro = 0;
1601     test_skip:
1602         if (!(c & 1)) {
1603         skip:
1604             preprocess_skip();
1605             is_bof = 0;
1606             goto redo;
1607         }
1608         break;
1609     case TOK_ENDIF:
1610         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1611             tcc_error("#endif without matching #if");
1612         s1->ifdef_stack_ptr--;
1613         /* '#ifndef macro' was at the start of file. Now we check if
1614            an '#endif' is exactly at the end of file */
1615         if (file->ifndef_macro &&
1616             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1617             file->ifndef_macro_saved = file->ifndef_macro;
1618             /* need to set to zero to avoid false matches if another
1619                #ifndef at middle of file */
1620             file->ifndef_macro = 0;
1621             while (tok != TOK_LINEFEED)
1622                 next_nomacro();
1623             tok_flags |= TOK_FLAG_ENDIF;
1624             goto the_end;
1625         }
1626         break;
1627     case TOK_LINE:
1628         next();
1629         if (tok != TOK_CINT)
1630             tcc_error("#line");
1631         file->line_num = tokc.i - 1; /* the line number will be incremented after */
1632         next();
1633         if (tok != TOK_LINEFEED) {
1634             if (tok != TOK_STR)
1635                 tcc_error("#line");
1636             pstrcpy(file->filename, sizeof(file->filename),
1637                     (char *)tokc.cstr->data);
1638         }
1639         break;
1640     case TOK_ERROR:
1641     case TOK_WARNING:
1642         c = tok;
1643         ch = file->buf_ptr[0];
1644         skip_spaces();
1645         q = buf;
1646         while (ch != '\n' && ch != CH_EOF) {
1647             if ((q - buf) < sizeof(buf) - 1)
1648                 *q++ = ch;
1649             if (ch == '\\') {
1650                 if (handle_stray_noerror() == 0)
1651                     --q;
1652             } else
1653                 inp();
1654         }
1655         *q = '\0';
1656         if (c == TOK_ERROR)
1657             tcc_error("#error %s", buf);
1658         else
1659             tcc_warning("#warning %s", buf);
1660         break;
1661     case TOK_PRAGMA:
1662         pragma_parse(s1);
1663         break;
1664     default:
1665         if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_PPNUM) {
1666             /* '!' is ignored to allow C scripts. numbers are ignored
1667                to emulate cpp behaviour */
1668         } else {
1669             if (!(saved_parse_flags & PARSE_FLAG_ASM_COMMENTS))
1670                 tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1671             else {
1672                 /* this is a gas line comment in an 'S' file. */
1673                 file->buf_ptr = parse_line_comment(file->buf_ptr);
1674                 goto the_end;
1675             }
1676         }
1677         break;
1678     }
1679     /* ignore other preprocess commands or #! for C scripts */
1680     while (tok != TOK_LINEFEED)
1681         next_nomacro();
1682  the_end:
1683     parse_flags = saved_parse_flags;
1684 }
1685
1686 /* evaluate escape codes in a string. */
1687 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1688 {
1689     int c, n;
1690     const uint8_t *p;
1691
1692     p = buf;
1693     for(;;) {
1694         c = *p;
1695         if (c == '\0')
1696             break;
1697         if (c == '\\') {
1698             p++;
1699             /* escape */
1700             c = *p;
1701             switch(c) {
1702             case '0': case '1': case '2': case '3':
1703             case '4': case '5': case '6': case '7':
1704                 /* at most three octal digits */
1705                 n = c - '0';
1706                 p++;
1707                 c = *p;
1708                 if (isoct(c)) {
1709                     n = n * 8 + c - '0';
1710                     p++;
1711                     c = *p;
1712                     if (isoct(c)) {
1713                         n = n * 8 + c - '0';
1714                         p++;
1715                     }
1716                 }
1717                 c = n;
1718                 goto add_char_nonext;
1719             case 'x':
1720             case 'u':
1721             case 'U':
1722                 p++;
1723                 n = 0;
1724                 for(;;) {
1725                     c = *p;
1726                     if (c >= 'a' && c <= 'f')
1727                         c = c - 'a' + 10;
1728                     else if (c >= 'A' && c <= 'F')
1729                         c = c - 'A' + 10;
1730                     else if (isnum(c))
1731                         c = c - '0';
1732                     else
1733                         break;
1734                     n = n * 16 + c;
1735                     p++;
1736                 }
1737                 c = n;
1738                 goto add_char_nonext;
1739             case 'a':
1740                 c = '\a';
1741                 break;
1742             case 'b':
1743                 c = '\b';
1744                 break;
1745             case 'f':
1746                 c = '\f';
1747                 break;
1748             case 'n':
1749                 c = '\n';
1750                 break;
1751             case 'r':
1752                 c = '\r';
1753                 break;
1754             case 't':
1755                 c = '\t';
1756                 break;
1757             case 'v':
1758                 c = '\v';
1759                 break;
1760             case 'e':
1761                 if (!gnu_ext)
1762                     goto invalid_escape;
1763                 c = 27;
1764                 break;
1765             case '\'':
1766             case '\"':
1767             case '\\':
1768             case '?':
1769                 break;
1770             default:
1771             invalid_escape:
1772                 if (c >= '!' && c <= '~')
1773                     tcc_warning("unknown escape sequence: \'\\%c\'", c);
1774                 else
1775                     tcc_warning("unknown escape sequence: \'\\x%x\'", c);
1776                 break;
1777             }
1778         }
1779         p++;
1780     add_char_nonext:
1781         if (!is_long)
1782             cstr_ccat(outstr, c);
1783         else
1784             cstr_wccat(outstr, c);
1785     }
1786     /* add a trailing '\0' */
1787     if (!is_long)
1788         cstr_ccat(outstr, '\0');
1789     else
1790         cstr_wccat(outstr, '\0');
1791 }
1792
1793 /* we use 64 bit numbers */
1794 #define BN_SIZE 2
1795
1796 /* bn = (bn << shift) | or_val */
1797 static void bn_lshift(unsigned int *bn, int shift, int or_val)
1798 {
1799     int i;
1800     unsigned int v;
1801     for(i=0;i<BN_SIZE;i++) {
1802         v = bn[i];
1803         bn[i] = (v << shift) | or_val;
1804         or_val = v >> (32 - shift);
1805     }
1806 }
1807
1808 static void bn_zero(unsigned int *bn)
1809 {
1810     int i;
1811     for(i=0;i<BN_SIZE;i++) {
1812         bn[i] = 0;
1813     }
1814 }
1815
1816 /* parse number in null terminated string 'p' and return it in the
1817    current token */
1818 static void parse_number(const char *p)
1819 {
1820     int b, t, c;
1821
1822     c = *p++;
1823         t = *p++;
1824     b = 10;
1825         if(c=='.'){
1826                 --p;
1827                 goto float_frac_parse;
1828         }
1829         if(c == '0'){
1830                 if (t == 'x' || t == 'X') {
1831             b = 16;
1832                         c = *p++;
1833         } else if (tcc_ext && (t == 'b' || t == 'B')) {
1834             b = 2;
1835                         c = *p++;
1836         }else{
1837                         --p;
1838                 }
1839         }else
1840                 --p;
1841     if(strchr(p , '.') || (b == 10 && (strchr(p,'e') || strchr(p,'E'))) ||
1842                 ((b == 2 || b == 16)&& (strchr(p,'p') || strchr(p,'P')))){
1843                 long double ld, sh, fb;
1844                 int exp;
1845                 /* NOTE: strtox should support that for hexa numbers, but
1846                         non ISOC99 libcs do not support it, so we prefer to do
1847                         it by hand */
1848                 /* hexadecimal or binary floats */
1849                 /* XXX: handle overflows */
1850 float_frac_parse:
1851                 fb = 1.0L/b;
1852                 sh = b;
1853                 ld = 0.0;
1854
1855                 while(1){
1856                         if (c == '\0')
1857                                 break;
1858                         if (c >= 'a' && c <= 'f')
1859                                 t = c - 'a' + 10;
1860                         else if (c >= 'A' && c <= 'F')
1861                                 t = c - 'A' + 10;
1862                         else if(isnum(c))
1863                                 t = c - '0';
1864                         else
1865                                 break;
1866                         if (t >= b)
1867                                 tcc_error("invalid digit");
1868                         ld = ld * b + t;
1869                         c = *p++;
1870                 }
1871                 if (c == '.'){
1872                         c = *p++;
1873                         sh = fb;
1874                         while (1){
1875                                 if (c == '\0')
1876                                         break;
1877                                 if (c >= 'a' && c <= 'f')
1878                                         t = c - 'a' + 10;
1879                                 else if (c >= 'A' && c <= 'F')
1880                                         t = c - 'A' + 10;
1881                                 else if (isnum(c))
1882                                         t =c - '0';
1883                                 else
1884                                         break;
1885                                 if (t >= b){
1886                                         if(b == 10 && (c == 'e' || c == 'E' || c == 'f' || c == 'F'))
1887                                                 break;
1888                                         tcc_error("invalid digit");
1889                                 }
1890                                 ld += sh*t;
1891                                 sh*=fb;
1892                                 c = *p++;
1893                         }
1894                 }
1895                 if ((b == 16 || b == 2) && c != 'p' && c != 'P')
1896                 expect("exponent");
1897                 if(((c == 'e' || c == 'E') && b == 10) ||
1898                         ((c == 'p' || c == 'P') && (b == 16 || b == 2))){
1899                         c = *p++;
1900                         if(c == '+' || c == '-'){
1901                                 if (c == '-')
1902                                         sh = fb;
1903                                 c = *p++;
1904                         }else
1905                                 sh = b;
1906                         if (!isnum(c))
1907                                 expect("exponent digits");
1908                         exp = 0;
1909                         do{
1910                                 exp = exp * 10 + c - '0';
1911                                 c = *p++;
1912                         }while(isnum(c));
1913                         while (exp != 0){
1914                                 if (exp & 1)
1915                                         ld *= sh;
1916                                 exp >>= 1;
1917                                 sh *= sh;
1918                         }
1919                 }
1920                 t = toup(c);
1921                 if (t == 'F') {
1922                         c = *p++;
1923                         tok = TOK_CFLOAT;
1924                         tokc.f = (float)ld;
1925                 } else if (t == 'L') {
1926                         c = *p++;
1927 #ifdef TCC_TARGET_PE
1928                         tok = TOK_CDOUBLE;
1929                         tokc.d = (double)ld;
1930 #else
1931                         tok = TOK_CLDOUBLE;
1932                         tokc.ld = ld;
1933 #endif
1934                 } else {
1935                         tok = TOK_CDOUBLE;
1936                         tokc.d = (double)ld;
1937                 }
1938     } else {
1939                 uint64_t n = 0, n1;
1940                 int warn = 1;
1941         int lcount, ucount;
1942                 if (b == 10 && c == '0') {
1943             b = 8;
1944         }
1945                 while(1){
1946                         if (c == '\0')
1947                                 break;
1948                         if (c >= 'a' && c <= 'f')
1949                                 t = c - 'a' + 10;
1950                         else if (c >= 'A' && c <= 'F')
1951                                 t = c - 'A' + 10;
1952                         else if(isnum(c))
1953                                 t = c - '0';
1954                         else
1955                                 break;
1956                         if (t >= b)
1957                                 tcc_error("invalid digit");
1958                         n1 = n;
1959                         n = n * b + t;
1960                         if (n < n1 && warn){
1961                                 tcc_warning("integer constant overflow");
1962                                 warn = 0;
1963                         }
1964                         c = *p++;
1965                 }
1966         /* XXX: not exactly ANSI compliant */
1967         if ((n & 0xffffffff00000000LL) != 0) {
1968             if ((n >> 63) != 0)
1969                 tok = TOK_CULLONG;
1970             else
1971                 tok = TOK_CLLONG;
1972         } else if (n > 0x7fffffff) {
1973             tok = TOK_CUINT;
1974         } else {
1975             tok = TOK_CINT;
1976         }
1977         lcount = 0;
1978         ucount = 0;
1979         for(;;) {
1980             t = toup(c);
1981             if (t == 'L') {
1982                 if (lcount >= 2)
1983                     tcc_error("three 'l's in integer constant");
1984                 lcount++;
1985 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
1986                 if (lcount == 2) {
1987 #endif
1988                     if (tok == TOK_CINT)
1989                         tok = TOK_CLLONG;
1990                     else if (tok == TOK_CUINT)
1991                         tok = TOK_CULLONG;
1992 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
1993                 }
1994 #endif
1995                 c = *p++;
1996             } else if (t == 'U') {
1997                 if (ucount >= 1)
1998                     tcc_error("two 'u's in integer constant");
1999                 ucount++;
2000                 if (tok == TOK_CINT)
2001                     tok = TOK_CUINT;
2002                 else if (tok == TOK_CLLONG)
2003                     tok = TOK_CULLONG;
2004                 c = *p++;
2005             } else {
2006                 break;
2007             }
2008         }
2009         if (tok == TOK_CINT || tok == TOK_CUINT)
2010             tokc.ui = n;
2011         else
2012             tokc.ull = n;
2013     }
2014     if (c)
2015         tcc_error("invalid number\n");
2016 }
2017
2018
2019 #define PARSE2(c1, tok1, c2, tok2)              \
2020     case c1:                                    \
2021         PEEKC(c, p);                            \
2022         if (c == c2) {                          \
2023             p++;                                \
2024             tok = tok2;                         \
2025         } else {                                \
2026             tok = tok1;                         \
2027         }                                       \
2028         break;
2029
2030 /* return next token without macro substitution */
2031 static inline void next_nomacro1(void)
2032 {
2033     int t, c, is_long;
2034     TokenSym *ts;
2035     uint8_t *p, *p1;
2036     unsigned int h;
2037
2038     p = file->buf_ptr;
2039  redo_no_start:
2040     c = *p;
2041     switch(c) {
2042     case ' ':
2043     case '\t':
2044         tok = c;
2045         p++;
2046         goto keep_tok_flags;
2047     case '\f':
2048     case '\v':
2049     case '\r':
2050         p++;
2051         goto redo_no_start;
2052     case '\\':
2053         /* first look if it is in fact an end of buffer */
2054         if (p >= file->buf_end) {
2055             file->buf_ptr = p;
2056             handle_eob();
2057             p = file->buf_ptr;
2058             if (p >= file->buf_end)
2059                 goto parse_eof;
2060             else
2061                 goto redo_no_start;
2062         } else {
2063             file->buf_ptr = p;
2064             ch = *p;
2065             handle_stray();
2066             p = file->buf_ptr;
2067             goto redo_no_start;
2068         }
2069     parse_eof:
2070         {
2071             TCCState *s1 = tcc_state;
2072             if ((parse_flags & PARSE_FLAG_LINEFEED)
2073                 && !(tok_flags & TOK_FLAG_EOF)) {
2074                 tok_flags |= TOK_FLAG_EOF;
2075                 tok = TOK_LINEFEED;
2076                 goto keep_tok_flags;
2077             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2078                 tok = TOK_EOF;
2079             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2080                 tcc_error("missing #endif");
2081             } else if (s1->include_stack_ptr == s1->include_stack) {
2082                 /* no include left : end of file. */
2083                 tok = TOK_EOF;
2084             } else {
2085                 tok_flags &= ~TOK_FLAG_EOF;
2086                 /* pop include file */
2087
2088                 /* test if previous '#endif' was after a #ifdef at
2089                    start of file */
2090                 if (tok_flags & TOK_FLAG_ENDIF) {
2091 #ifdef INC_DEBUG
2092                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2093 #endif
2094                     add_cached_include(s1, file->filename, file->ifndef_macro_saved);
2095                     tok_flags &= ~TOK_FLAG_ENDIF;
2096                 }
2097
2098                 /* add end of include file debug info */
2099                 if (tcc_state->do_debug) {
2100                     put_stabd(N_EINCL, 0, 0);
2101                 }
2102                 /* pop include stack */
2103                 tcc_close();
2104                 s1->include_stack_ptr--;
2105                 p = file->buf_ptr;
2106                 goto redo_no_start;
2107             }
2108         }
2109         break;
2110
2111     case '\n':
2112         file->line_num++;
2113         tok_flags |= TOK_FLAG_BOL;
2114         p++;
2115 maybe_newline:
2116         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2117             goto redo_no_start;
2118         tok = TOK_LINEFEED;
2119         goto keep_tok_flags;
2120
2121     case '#':
2122         /* XXX: simplify */
2123         PEEKC(c, p);
2124         if ((tok_flags & TOK_FLAG_BOL) &&
2125             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2126             file->buf_ptr = p;
2127             preprocess(tok_flags & TOK_FLAG_BOF);
2128             p = file->buf_ptr;
2129             goto maybe_newline;
2130         } else {
2131             if (c == '#') {
2132                 p++;
2133                 tok = TOK_TWOSHARPS;
2134             } else {
2135                 if (parse_flags & PARSE_FLAG_ASM_COMMENTS) {
2136                     p = parse_line_comment(p - 1);
2137                     goto redo_no_start;
2138                 } else {
2139                     tok = '#';
2140                 }
2141             }
2142         }
2143         break;
2144
2145     case 'a': case 'b': case 'c': case 'd':
2146     case 'e': case 'f': case 'g': case 'h':
2147     case 'i': case 'j': case 'k': case 'l':
2148     case 'm': case 'n': case 'o': case 'p':
2149     case 'q': case 'r': case 's': case 't':
2150     case 'u': case 'v': case 'w': case 'x':
2151     case 'y': case 'z':
2152     case 'A': case 'B': case 'C': case 'D':
2153     case 'E': case 'F': case 'G': case 'H':
2154     case 'I': case 'J': case 'K':
2155     case 'M': case 'N': case 'O': case 'P':
2156     case 'Q': case 'R': case 'S': case 'T':
2157     case 'U': case 'V': case 'W': case 'X':
2158     case 'Y': case 'Z':
2159     case '_':
2160     parse_ident_fast:
2161         p1 = p;
2162         h = TOK_HASH_INIT;
2163         h = TOK_HASH_FUNC(h, c);
2164         p++;
2165         for(;;) {
2166             c = *p;
2167             if (!isidnum_table[c-CH_EOF])
2168                 break;
2169             h = TOK_HASH_FUNC(h, c);
2170             p++;
2171         }
2172         if (c != '\\') {
2173             TokenSym **pts;
2174             int len;
2175
2176             /* fast case : no stray found, so we have the full token
2177                and we have already hashed it */
2178             len = p - p1;
2179             h &= (TOK_HASH_SIZE - 1);
2180             pts = &hash_ident[h];
2181             for(;;) {
2182                 ts = *pts;
2183                 if (!ts)
2184                     break;
2185                 if (ts->len == len && !memcmp(ts->str, p1, len))
2186                     goto token_found;
2187                 pts = &(ts->hash_next);
2188             }
2189             ts = tok_alloc_new(pts, (char *) p1, len);
2190         token_found: ;
2191         } else {
2192             /* slower case */
2193             cstr_reset(&tokcstr);
2194
2195             while (p1 < p) {
2196                 cstr_ccat(&tokcstr, *p1);
2197                 p1++;
2198             }
2199             p--;
2200             PEEKC(c, p);
2201         parse_ident_slow:
2202             while (isidnum_table[c-CH_EOF]) {
2203                 cstr_ccat(&tokcstr, c);
2204                 PEEKC(c, p);
2205             }
2206             ts = tok_alloc(tokcstr.data, tokcstr.size);
2207         }
2208         tok = ts->tok;
2209         break;
2210     case 'L':
2211         t = p[1];
2212         if (t != '\\' && t != '\'' && t != '\"') {
2213             /* fast case */
2214             goto parse_ident_fast;
2215         } else {
2216             PEEKC(c, p);
2217             if (c == '\'' || c == '\"') {
2218                 is_long = 1;
2219                 goto str_const;
2220             } else {
2221                 cstr_reset(&tokcstr);
2222                 cstr_ccat(&tokcstr, 'L');
2223                 goto parse_ident_slow;
2224             }
2225         }
2226         break;
2227     case '0': case '1': case '2': case '3':
2228     case '4': case '5': case '6': case '7':
2229     case '8': case '9':
2230
2231         cstr_reset(&tokcstr);
2232         /* after the first digit, accept digits, alpha, '.' or sign if
2233            prefixed by 'eEpP' */
2234     parse_num:
2235         for(;;) {
2236             t = c;
2237             cstr_ccat(&tokcstr, c);
2238             PEEKC(c, p);
2239             if (!(isnum(c) || isid(c) || c == '.' ||
2240                   ((c == '+' || c == '-') &&
2241                    (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2242                 break;
2243         }
2244         /* We add a trailing '\0' to ease parsing */
2245         cstr_ccat(&tokcstr, '\0');
2246         tokc.cstr = &tokcstr;
2247         tok = TOK_PPNUM;
2248         break;
2249     case '.':
2250         /* special dot handling because it can also start a number */
2251         PEEKC(c, p);
2252         if (isnum(c)) {
2253             cstr_reset(&tokcstr);
2254             cstr_ccat(&tokcstr, '.');
2255             goto parse_num;
2256         } else if (c == '.') {
2257             PEEKC(c, p);
2258             if (c != '.')
2259                 expect("'.'");
2260             PEEKC(c, p);
2261             tok = TOK_DOTS;
2262         } else {
2263             tok = '.';
2264         }
2265         break;
2266     case '\'':
2267     case '\"':
2268         is_long = 0;
2269     str_const:
2270         {
2271             CString str;
2272             int sep;
2273
2274             sep = c;
2275
2276             /* parse the string */
2277             cstr_new(&str);
2278             p = parse_pp_string(p, sep, &str);
2279             cstr_ccat(&str, '\0');
2280
2281             /* eval the escape (should be done as TOK_PPNUM) */
2282             cstr_reset(&tokcstr);
2283             parse_escape_string(&tokcstr, str.data, is_long);
2284             cstr_free(&str);
2285
2286             if (sep == '\'') {
2287                 int char_size;
2288                 /* XXX: make it portable */
2289                 if (!is_long)
2290                     char_size = 1;
2291                 else
2292                     char_size = sizeof(nwchar_t);
2293                 if (tokcstr.size <= char_size)
2294                     tcc_error("empty character constant");
2295                 if (tokcstr.size > 2 * char_size)
2296                     tcc_warning("multi-character character constant");
2297                 if (!is_long) {
2298                     tokc.i = *(int8_t *)tokcstr.data;
2299                     tok = TOK_CCHAR;
2300                 } else {
2301                     tokc.i = *(nwchar_t *)tokcstr.data;
2302                     tok = TOK_LCHAR;
2303                 }
2304             } else {
2305                 tokc.cstr = &tokcstr;
2306                 if (!is_long)
2307                     tok = TOK_STR;
2308                 else
2309                     tok = TOK_LSTR;
2310             }
2311         }
2312         break;
2313
2314     case '<':
2315         PEEKC(c, p);
2316         if (c == '=') {
2317             p++;
2318             tok = TOK_LE;
2319         } else if (c == '<') {
2320             PEEKC(c, p);
2321             if (c == '=') {
2322                 p++;
2323                 tok = TOK_A_SHL;
2324             } else {
2325                 tok = TOK_SHL;
2326             }
2327         } else {
2328             tok = TOK_LT;
2329         }
2330         break;
2331
2332     case '>':
2333         PEEKC(c, p);
2334         if (c == '=') {
2335             p++;
2336             tok = TOK_GE;
2337         } else if (c == '>') {
2338             PEEKC(c, p);
2339             if (c == '=') {
2340                 p++;
2341                 tok = TOK_A_SAR;
2342             } else {
2343                 tok = TOK_SAR;
2344             }
2345         } else {
2346             tok = TOK_GT;
2347         }
2348         break;
2349
2350     case '&':
2351         PEEKC(c, p);
2352         if (c == '&') {
2353             p++;
2354             tok = TOK_LAND;
2355         } else if (c == '=') {
2356             p++;
2357             tok = TOK_A_AND;
2358         } else {
2359             tok = '&';
2360         }
2361         break;
2362
2363     case '|':
2364         PEEKC(c, p);
2365         if (c == '|') {
2366             p++;
2367             tok = TOK_LOR;
2368         } else if (c == '=') {
2369             p++;
2370             tok = TOK_A_OR;
2371         } else {
2372             tok = '|';
2373         }
2374         break;
2375
2376     case '+':
2377         PEEKC(c, p);
2378         if (c == '+') {
2379             p++;
2380             tok = TOK_INC;
2381         } else if (c == '=') {
2382             p++;
2383             tok = TOK_A_ADD;
2384         } else {
2385             tok = '+';
2386         }
2387         break;
2388
2389     case '-':
2390         PEEKC(c, p);
2391         if (c == '-') {
2392             p++;
2393             tok = TOK_DEC;
2394         } else if (c == '=') {
2395             p++;
2396             tok = TOK_A_SUB;
2397         } else if (c == '>') {
2398             p++;
2399             tok = TOK_ARROW;
2400         } else {
2401             tok = '-';
2402         }
2403         break;
2404
2405     PARSE2('!', '!', '=', TOK_NE)
2406     PARSE2('=', '=', '=', TOK_EQ)
2407     PARSE2('*', '*', '=', TOK_A_MUL)
2408     PARSE2('%', '%', '=', TOK_A_MOD)
2409     PARSE2('^', '^', '=', TOK_A_XOR)
2410
2411         /* comments or operator */
2412     case '/':
2413         PEEKC(c, p);
2414         if (c == '*') {
2415             p = parse_comment(p);
2416             /* comments replaced by a blank */
2417             tok = ' ';
2418             goto keep_tok_flags;
2419         } else if (c == '/') {
2420             p = parse_line_comment(p);
2421             tok = ' ';
2422             goto keep_tok_flags;
2423         } else if (c == '=') {
2424             p++;
2425             tok = TOK_A_DIV;
2426         } else {
2427             tok = '/';
2428         }
2429         break;
2430
2431         /* simple tokens */
2432     case '(':
2433     case ')':
2434     case '[':
2435     case ']':
2436     case '{':
2437     case '}':
2438     case ',':
2439     case ';':
2440     case ':':
2441     case '?':
2442     case '~':
2443     case '$': /* only used in assembler */
2444     case '@': /* dito */
2445         tok = c;
2446         p++;
2447         break;
2448     default:
2449         tcc_error("unrecognized character \\x%02x", c);
2450         break;
2451     }
2452     tok_flags = 0;
2453 keep_tok_flags:
2454     file->buf_ptr = p;
2455 #if defined(PARSE_DEBUG)
2456     printf("token = %s\n", get_tok_str(tok, &tokc));
2457 #endif
2458 }
2459
2460 /* return next token without macro substitution. Can read input from
2461    macro_ptr buffer */
2462 static void next_nomacro_spc(void)
2463 {
2464     if (macro_ptr) {
2465     redo:
2466         tok = *macro_ptr;
2467         if (tok) {
2468             TOK_GET(&tok, &macro_ptr, &tokc);
2469             if (tok == TOK_LINENUM) {
2470                 file->line_num = tokc.i;
2471                 goto redo;
2472             }
2473         }
2474     } else {
2475         next_nomacro1();
2476     }
2477 }
2478
2479 ST_FUNC void next_nomacro(void)
2480 {
2481     do {
2482         next_nomacro_spc();
2483     } while (is_space(tok));
2484 }
2485
2486 /* substitute arguments in replacement lists in macro_str by the values in
2487    args (field d) and return allocated string */
2488 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2489 {
2490     int last_tok, t, spc;
2491     const int *st;
2492     Sym *s;
2493     CValue cval;
2494     TokenString str;
2495     CString cstr;
2496
2497     tok_str_new(&str);
2498     last_tok = 0;
2499     while(1) {
2500         TOK_GET(&t, &macro_str, &cval);
2501         if (!t)
2502             break;
2503         if (t == '#') {
2504             /* stringize */
2505             TOK_GET(&t, &macro_str, &cval);
2506             if (!t)
2507                 break;
2508             s = sym_find2(args, t);
2509             if (s) {
2510                 cstr_new(&cstr);
2511                 st = s->d;
2512                 spc = 0;
2513                 while (*st) {
2514                     TOK_GET(&t, &st, &cval);
2515                     if (!check_space(t, &spc))
2516                         cstr_cat(&cstr, get_tok_str(t, &cval));
2517                 }
2518                 cstr.size -= spc;
2519                 cstr_ccat(&cstr, '\0');
2520 #ifdef PP_DEBUG
2521                 printf("stringize: %s\n", (char *)cstr.data);
2522 #endif
2523                 /* add string */
2524                 cval.cstr = &cstr;
2525                 tok_str_add2(&str, TOK_STR, &cval);
2526                 cstr_free(&cstr);
2527             } else {
2528                 tok_str_add2(&str, t, &cval);
2529             }
2530         } else if (t >= TOK_IDENT) {
2531             s = sym_find2(args, t);
2532             if (s) {
2533                 st = s->d;
2534                 /* if '##' is present before or after, no arg substitution */
2535                 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2536                     /* special case for var arg macros : ## eats the
2537                        ',' if empty VA_ARGS variable. */
2538                     /* XXX: test of the ',' is not 100%
2539                        reliable. should fix it to avoid security
2540                        problems */
2541                     if (gnu_ext && s->type.t &&
2542                         last_tok == TOK_TWOSHARPS &&
2543                         str.len >= 2 && str.str[str.len - 2] == ',') {
2544                         if (*st == TOK_PLCHLDR) {
2545                             /* suppress ',' '##' */
2546                             str.len -= 2;
2547                         } else {
2548                             /* suppress '##' and add variable */
2549                             str.len--;
2550                             goto add_var;
2551                         }
2552                     } else {
2553                         int t1;
2554                     add_var:
2555                         for(;;) {
2556                             TOK_GET(&t1, &st, &cval);
2557                             if (!t1)
2558                                 break;
2559                             tok_str_add2(&str, t1, &cval);
2560                         }
2561                     }
2562                 } else {
2563                     /* NOTE: the stream cannot be read when macro
2564                        substituing an argument */
2565                     macro_subst(&str, nested_list, st, NULL);
2566                 }
2567             } else {
2568                 tok_str_add(&str, t);
2569             }
2570         } else {
2571             tok_str_add2(&str, t, &cval);
2572         }
2573         last_tok = t;
2574     }
2575     tok_str_add(&str, 0);
2576     return str.str;
2577 }
2578
2579 static char const ab_month_name[12][4] =
2580 {
2581     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2582     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2583 };
2584
2585 /* do macro substitution of current token with macro 's' and add
2586    result to (tok_str,tok_len). 'nested_list' is the list of all
2587    macros we got inside to avoid recursing. Return non zero if no
2588    substitution needs to be done */
2589 static int macro_subst_tok(TokenString *tok_str,
2590                            Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2591 {
2592     Sym *args, *sa, *sa1;
2593     int mstr_allocated, parlevel, *mstr, t, t1, spc;
2594     const int *p;
2595     TokenString str;
2596     char *cstrval;
2597     CValue cval;
2598     CString cstr;
2599     char buf[32];
2600
2601     /* if symbol is a macro, prepare substitution */
2602     /* special macros */
2603     if (tok == TOK___LINE__) {
2604         snprintf(buf, sizeof(buf), "%d", file->line_num);
2605         cstrval = buf;
2606         t1 = TOK_PPNUM;
2607         goto add_cstr1;
2608     } else if (tok == TOK___FILE__) {
2609         cstrval = file->filename;
2610         goto add_cstr;
2611     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2612         time_t ti;
2613         struct tm *tm;
2614
2615         time(&ti);
2616         tm = localtime(&ti);
2617         if (tok == TOK___DATE__) {
2618             snprintf(buf, sizeof(buf), "%s %2d %d",
2619                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2620         } else {
2621             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2622                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2623         }
2624         cstrval = buf;
2625     add_cstr:
2626         t1 = TOK_STR;
2627     add_cstr1:
2628         cstr_new(&cstr);
2629         cstr_cat(&cstr, cstrval);
2630         cstr_ccat(&cstr, '\0');
2631         cval.cstr = &cstr;
2632         tok_str_add2(tok_str, t1, &cval);
2633         cstr_free(&cstr);
2634     } else {
2635         mstr = s->d;
2636         mstr_allocated = 0;
2637         if (s->type.t == MACRO_FUNC) {
2638             /* NOTE: we do not use next_nomacro to avoid eating the
2639                next token. XXX: find better solution */
2640         redo:
2641             if (macro_ptr) {
2642                 p = macro_ptr;
2643                 while (is_space(t = *p) || TOK_LINEFEED == t)
2644                     ++p;
2645                 if (t == 0 && can_read_stream) {
2646                     /* end of macro stream: we must look at the token
2647                        after in the file */
2648                     struct macro_level *ml = *can_read_stream;
2649                     macro_ptr = NULL;
2650                     if (ml)
2651                     {
2652                         macro_ptr = ml->p;
2653                         ml->p = NULL;
2654                         *can_read_stream = ml -> prev;
2655                     }
2656                     /* also, end of scope for nested defined symbol */
2657                     (*nested_list)->v = -1;
2658                     goto redo;
2659                 }
2660             } else {
2661                 ch = file->buf_ptr[0];
2662                 while (is_space(ch) || ch == '\n' || ch == '/')
2663                   {
2664                     if (ch == '/')
2665                       {
2666                         int c;
2667                         uint8_t *p = file->buf_ptr;
2668                         PEEKC(c, p);
2669                         if (c == '*') {
2670                             p = parse_comment(p);
2671                             file->buf_ptr = p - 1;
2672                         } else if (c == '/') {
2673                             p = parse_line_comment(p);
2674                             file->buf_ptr = p - 1;
2675                         } else
2676                           break;
2677                       }
2678                     cinp();
2679                   }
2680                 t = ch;
2681             }
2682             if (t != '(') /* no macro subst */
2683                 return -1;
2684
2685             /* argument macro */
2686             next_nomacro();
2687             next_nomacro();
2688             args = NULL;
2689             sa = s->next;
2690             /* NOTE: empty args are allowed, except if no args */
2691             for(;;) {
2692                 /* handle '()' case */
2693                 if (!args && !sa && tok == ')')
2694                     break;
2695                 if (!sa)
2696                     tcc_error("macro '%s' used with too many args",
2697                           get_tok_str(s->v, 0));
2698                 tok_str_new(&str);
2699                 parlevel = spc = 0;
2700                 /* NOTE: non zero sa->t indicates VA_ARGS */
2701                 while ((parlevel > 0 ||
2702                         (tok != ')' &&
2703                          (tok != ',' || sa->type.t))) &&
2704                        tok != -1) {
2705                     if (tok == '(')
2706                         parlevel++;
2707                     else if (tok == ')')
2708                         parlevel--;
2709                     if (tok == TOK_LINEFEED)
2710                         tok = ' ';
2711                     if (!check_space(tok, &spc))
2712                         tok_str_add2(&str, tok, &tokc);
2713                     next_nomacro_spc();
2714                 }
2715                 if (!str.len)
2716                     tok_str_add(&str, TOK_PLCHLDR);
2717                 str.len -= spc;
2718                 tok_str_add(&str, 0);
2719                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
2720                 sa1->d = str.str;
2721                 sa = sa->next;
2722                 if (tok == ')') {
2723                     /* special case for gcc var args: add an empty
2724                        var arg argument if it is omitted */
2725                     if (sa && sa->type.t && gnu_ext)
2726                         continue;
2727                     else
2728                         break;
2729                 }
2730                 if (tok != ',')
2731                     expect(",");
2732                 next_nomacro();
2733             }
2734             if (sa) {
2735                 tcc_error("macro '%s' used with too few args",
2736                       get_tok_str(s->v, 0));
2737             }
2738
2739             /* now subst each arg */
2740             mstr = macro_arg_subst(nested_list, mstr, args);
2741             /* free memory */
2742             sa = args;
2743             while (sa) {
2744                 sa1 = sa->prev;
2745                 tok_str_free(sa->d);
2746                 sym_free(sa);
2747                 sa = sa1;
2748             }
2749             mstr_allocated = 1;
2750         }
2751         sym_push2(nested_list, s->v, 0, 0);
2752         macro_subst(tok_str, nested_list, mstr, can_read_stream);
2753         /* pop nested defined symbol */
2754         sa1 = *nested_list;
2755         *nested_list = sa1->prev;
2756         sym_free(sa1);
2757         if (mstr_allocated)
2758             tok_str_free(mstr);
2759     }
2760     return 0;
2761 }
2762
2763 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2764    return the resulting string (which must be freed). */
2765 static inline int *macro_twosharps(const int *macro_str)
2766 {
2767     const int *ptr;
2768     int t;
2769     TokenString macro_str1;
2770     CString cstr;
2771     int n, start_of_nosubsts;
2772
2773     /* we search the first '##' */
2774     for(ptr = macro_str;;) {
2775         CValue cval;
2776         TOK_GET(&t, &ptr, &cval);
2777         if (t == TOK_TWOSHARPS)
2778             break;
2779         /* nothing more to do if end of string */
2780         if (t == 0)
2781             return NULL;
2782     }
2783
2784     /* we saw '##', so we need more processing to handle it */
2785     start_of_nosubsts = -1;
2786     tok_str_new(&macro_str1);
2787     for(ptr = macro_str;;) {
2788         TOK_GET(&tok, &ptr, &tokc);
2789         if (tok == 0)
2790             break;
2791         if (tok == TOK_TWOSHARPS)
2792             continue;
2793         if (tok == TOK_NOSUBST && start_of_nosubsts < 0)
2794             start_of_nosubsts = macro_str1.len;
2795         while (*ptr == TOK_TWOSHARPS) {
2796             /* given 'a##b', remove nosubsts preceding 'a' */
2797             if (start_of_nosubsts >= 0)
2798                 macro_str1.len = start_of_nosubsts;
2799             /* given 'a##b', skip '##' */
2800             t = *++ptr;
2801             /* given 'a##b', remove nosubsts preceding 'b' */
2802             while (t == TOK_NOSUBST)
2803                 t = *++ptr;
2804             if (t && t != TOK_TWOSHARPS) {
2805                 CValue cval;
2806                 TOK_GET(&t, &ptr, &cval);
2807                 /* We concatenate the two tokens */
2808                 cstr_new(&cstr);
2809                 if (tok != TOK_PLCHLDR)
2810                     cstr_cat(&cstr, get_tok_str(tok, &tokc));
2811                 n = cstr.size;
2812                 if (t != TOK_PLCHLDR || tok == TOK_PLCHLDR)
2813                     cstr_cat(&cstr, get_tok_str(t, &cval));
2814                 cstr_ccat(&cstr, '\0');
2815
2816                 tcc_open_bf(tcc_state, ":paste:", cstr.size);
2817                 memcpy(file->buffer, cstr.data, cstr.size);
2818                 for (;;) {
2819                     next_nomacro1();
2820                     if (0 == *file->buf_ptr)
2821                         break;
2822                     tok_str_add2(&macro_str1, tok, &tokc);
2823                     tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
2824                         n, cstr.data, (char*)cstr.data + n);
2825                 }
2826                 tcc_close();
2827                 cstr_free(&cstr);
2828             }
2829         }
2830         if (tok != TOK_NOSUBST) {
2831             tok_str_add2(&macro_str1, tok, &tokc);
2832             tok = ' ';
2833             start_of_nosubsts = -1;
2834         }
2835         tok_str_add2(&macro_str1, tok, &tokc);
2836     }
2837     tok_str_add(&macro_str1, 0);
2838     return macro_str1.str;
2839 }
2840
2841
2842 /* do macro substitution of macro_str and add result to
2843    (tok_str,tok_len). 'nested_list' is the list of all macros we got
2844    inside to avoid recursing. */
2845 static void macro_subst(TokenString *tok_str, Sym **nested_list,
2846                         const int *macro_str, struct macro_level ** can_read_stream)
2847 {
2848     Sym *s;
2849     int *macro_str1;
2850     const int *ptr;
2851     int t, ret, spc;
2852     CValue cval;
2853     struct macro_level ml;
2854     int force_blank;
2855
2856     /* first scan for '##' operator handling */
2857     ptr = macro_str;
2858     macro_str1 = macro_twosharps(ptr);
2859
2860     if (macro_str1)
2861         ptr = macro_str1;
2862     spc = 0;
2863     force_blank = 0;
2864
2865     while (1) {
2866         /* NOTE: ptr == NULL can only happen if tokens are read from
2867            file stream due to a macro function call */
2868         if (ptr == NULL)
2869             break;
2870         TOK_GET(&t, &ptr, &cval);
2871         if (t == 0)
2872             break;
2873         if (t == TOK_NOSUBST) {
2874             /* following token has already been subst'd. just copy it on */
2875             tok_str_add2(tok_str, TOK_NOSUBST, NULL);
2876             TOK_GET(&t, &ptr, &cval);
2877             goto no_subst;
2878         }
2879         s = define_find(t);
2880         if (s != NULL) {
2881             /* if nested substitution, do nothing */
2882             if (sym_find2(*nested_list, t)) {
2883                 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
2884                 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
2885                 goto no_subst;
2886             }
2887             ml.p = macro_ptr;
2888             if (can_read_stream)
2889                 ml.prev = *can_read_stream, *can_read_stream = &ml;
2890             macro_ptr = (int *)ptr;
2891             tok = t;
2892             ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
2893             ptr = (int *)macro_ptr;
2894             macro_ptr = ml.p;
2895             if (can_read_stream && *can_read_stream == &ml)
2896                 *can_read_stream = ml.prev;
2897             if (ret != 0)
2898                 goto no_subst;
2899             if (parse_flags & PARSE_FLAG_SPACES)
2900                 force_blank = 1;
2901         } else {
2902         no_subst:
2903             if (force_blank) {
2904                 tok_str_add(tok_str, ' ');
2905                 spc = 1;
2906                 force_blank = 0;
2907             }
2908             if (!check_space(t, &spc))
2909                 tok_str_add2(tok_str, t, &cval);
2910         }
2911     }
2912     if (macro_str1)
2913         tok_str_free(macro_str1);
2914 }
2915
2916 /* return next token with macro substitution */
2917 ST_FUNC void next(void)
2918 {
2919     Sym *nested_list, *s;
2920     TokenString str;
2921     struct macro_level *ml;
2922
2923  redo:
2924     if (parse_flags & PARSE_FLAG_SPACES)
2925         next_nomacro_spc();
2926     else
2927         next_nomacro();
2928     if (!macro_ptr) {
2929         /* if not reading from macro substituted string, then try
2930            to substitute macros */
2931         if (tok >= TOK_IDENT &&
2932             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2933             s = define_find(tok);
2934             if (s) {
2935                 /* we have a macro: we try to substitute */
2936                 tok_str_new(&str);
2937                 nested_list = NULL;
2938                 ml = NULL;
2939                 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
2940                     /* substitution done, NOTE: maybe empty */
2941                     tok_str_add(&str, 0);
2942                     macro_ptr = str.str;
2943                     macro_ptr_allocated = str.str;
2944                     goto redo;
2945                 }
2946             }
2947         }
2948     } else {
2949         if (tok == 0) {
2950             /* end of macro or end of unget buffer */
2951             if (unget_buffer_enabled) {
2952                 macro_ptr = unget_saved_macro_ptr;
2953                 unget_buffer_enabled = 0;
2954             } else {
2955                 /* end of macro string: free it */
2956                 tok_str_free(macro_ptr_allocated);
2957                 macro_ptr_allocated = NULL;
2958                 macro_ptr = NULL;
2959             }
2960             goto redo;
2961         } else if (tok == TOK_NOSUBST) {
2962             /* discard preprocessor's nosubst markers */
2963             goto redo;
2964         }
2965     }
2966
2967     /* convert preprocessor tokens into C tokens */
2968     if (tok == TOK_PPNUM &&
2969         (parse_flags & PARSE_FLAG_TOK_NUM)) {
2970         parse_number((char *)tokc.cstr->data);
2971     }
2972 }
2973
2974 /* push back current token and set current token to 'last_tok'. Only
2975    identifier case handled for labels. */
2976 ST_INLN void unget_tok(int last_tok)
2977 {
2978     int i, n;
2979     int *q;
2980     if (unget_buffer_enabled)
2981       {
2982         /* assert(macro_ptr == unget_saved_buffer + 1);
2983            assert(*macro_ptr == 0);  */
2984       }
2985     else
2986       {
2987         unget_saved_macro_ptr = macro_ptr;
2988         unget_buffer_enabled = 1;
2989       }
2990     q = unget_saved_buffer;
2991     macro_ptr = q;
2992     *q++ = tok;
2993     n = tok_ext_size(tok) - 1;
2994     for(i=0;i<n;i++)
2995         *q++ = tokc.tab[i];
2996     *q = 0; /* end of token string */
2997     tok = last_tok;
2998 }
2999
3000
3001 /* better than nothing, but needs extension to handle '-E' option
3002    correctly too */
3003 ST_FUNC void preprocess_init(TCCState *s1)
3004 {
3005     s1->include_stack_ptr = s1->include_stack;
3006     /* XXX: move that before to avoid having to initialize
3007        file->ifdef_stack_ptr ? */
3008     s1->ifdef_stack_ptr = s1->ifdef_stack;
3009     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3010
3011     vtop = vstack - 1;
3012     s1->pack_stack[0] = 0;
3013     s1->pack_stack_ptr = s1->pack_stack;
3014 }
3015
3016 ST_FUNC void preprocess_new(void)
3017 {
3018     int i, c;
3019     const char *p, *r;
3020
3021     /* init isid table */
3022     for(i=CH_EOF;i<256;i++)
3023         isidnum_table[i-CH_EOF] = isid(i) || isnum(i);
3024
3025     /* add all tokens */
3026     table_ident = NULL;
3027     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3028
3029     tok_ident = TOK_IDENT;
3030     p = tcc_keywords;
3031     while (*p) {
3032         r = p;
3033         for(;;) {
3034             c = *r++;
3035             if (c == '\0')
3036                 break;
3037         }
3038         tok_alloc(p, r - p - 1);
3039         p = r;
3040     }
3041 }
3042
3043 /* Preprocess the current file */
3044 ST_FUNC int tcc_preprocess(TCCState *s1)
3045 {
3046     Sym *define_start;
3047
3048     BufferedFile *file_ref, **iptr, **iptr_new;
3049     int token_seen, line_ref, d;
3050     const char *s;
3051
3052     preprocess_init(s1);
3053     define_start = define_stack;
3054     ch = file->buf_ptr[0];
3055     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3056     parse_flags = PARSE_FLAG_ASM_COMMENTS | PARSE_FLAG_PREPROCESS |
3057         PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
3058     token_seen = 0;
3059     line_ref = 0;
3060     file_ref = NULL;
3061     iptr = s1->include_stack_ptr;
3062
3063     for (;;) {
3064         next();
3065         if (tok == TOK_EOF) {
3066             break;
3067         } else if (file != file_ref) {
3068             goto print_line;
3069         } else if (tok == TOK_LINEFEED) {
3070             if (!token_seen)
3071                 continue;
3072             ++line_ref;
3073             token_seen = 0;
3074         } else if (!token_seen) {
3075             d = file->line_num - line_ref;
3076             if (file != file_ref || d < 0 || d >= 8) {
3077 print_line:
3078                 iptr_new = s1->include_stack_ptr;
3079                 s = iptr_new > iptr ? " 1"
3080                   : iptr_new < iptr ? " 2"
3081                   : iptr_new > s1->include_stack ? " 3"
3082                   : ""
3083                   ;
3084                 iptr = iptr_new;
3085                 fprintf(s1->ppfp, "# %d \"%s\"%s\n", file->line_num, file->filename, s);
3086             } else {
3087                 while (d)
3088                     fputs("\n", s1->ppfp), --d;
3089             }
3090             line_ref = (file_ref = file)->line_num;
3091             token_seen = tok != TOK_LINEFEED;
3092             if (!token_seen)
3093                 continue;
3094         }
3095         fputs(get_tok_str(tok, &tokc), s1->ppfp);
3096     }
3097     free_defines(define_start);
3098     return 0;
3099 }