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 ST_DATA int parse_flags;
  28
  29 ST_DATA struct BufferedFile *file;
  30 ST_DATA int ch, tok;
  31 ST_DATA CValue tokc;
  32 ST_DATA const int *macro_ptr;
  33 ST_DATA CString tokcstr; /* current parsed string, if any */
  34
  35 /* display benchmark infos */
  36 ST_DATA int total_lines;
  37 ST_DATA int total_bytes;
  38 ST_DATA int tok_ident;
  39 ST_DATA TokenSym **table_ident;
  40
  41 /* ------------------------------------------------------------------------- */
  42
  43 static int *macro_ptr_allocated;
  44 static const int *unget_saved_macro_ptr;
  45 static int unget_saved_buffer[TOK_MAX_SIZE + 1];
  46 static int unget_buffer_enabled;
  47 static TokenSym *hash_ident[TOK_HASH_SIZE];
  48 static char token_buf[STRING_MAX_SIZE + 1];
  49 /* true if isid(c) || isnum(c) */
  50 static unsigned char isidnum_table[256-CH_EOF];
  51
  52 static const char tcc_keywords[] =
  53 #define DEF(id, str) str "\0"
  54 #include "tcctok.h"
  55 #undef DEF
  56 ;
  57
  58 /* WARNING: the content of this string encodes token numbers */
  59 static const unsigned char tok_two_chars[] =
  60 /* outdated -- gr
  61     "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
  62     "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
  63 */{
  64     '<','=', TOK_LE,
  65     '>','=', TOK_GE,
  66     '!','=', TOK_NE,
  67     '&','&', TOK_LAND,
  68     '|','|', TOK_LOR,
  69     '+','+', TOK_INC,
  70     '-','-', TOK_DEC,
  71     '=','=', TOK_EQ,
  72     '<','<', TOK_SHL,
  73     '>','>', TOK_SAR,
  74     '+','=', TOK_A_ADD,
  75     '-','=', TOK_A_SUB,
  76     '*','=', TOK_A_MUL,
  77     '/','=', TOK_A_DIV,
  78     '%','=', TOK_A_MOD,
  79     '&','=', TOK_A_AND,
  80     '^','=', TOK_A_XOR,
  81     '|','=', TOK_A_OR,
  82     '-','>', TOK_ARROW,
  83     '.','.', 0xa8, // C++ token ?
  84     '#','#', TOK_TWOSHARPS,
  85     0
  86 };
  87
  88 struct macro_level {
  89     struct macro_level *prev;
  90     const int *p;
  91 };
  92
  93 static void next_nomacro_spc(void);
  94 static void macro_subst(
  95     TokenString *tok_str,
  96     Sym **nested_list,
  97     const int *macro_str,
  98     struct macro_level **can_read_stream
  99     );
 100
 101 ST_FUNC void skip(int c)
 102 {
 103     if (tok != c)
 104         tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
 105     next();
 106 }
 107
 108 ST_FUNC void expect(const char *msg)
 109 {
 110     tcc_error("%s expected", msg);
 111 }
 112
 113 /* ------------------------------------------------------------------------- */
 114 /* CString handling */
 115 static void cstr_realloc(CString *cstr, int new_size)
 116 {
 117     int size;
 118     void *data;
 119
 120     size = cstr->size_allocated;
 121     if (size == 0)
 122         size = 8; /* no need to allocate a too small first string */
 123     while (size < new_size)
 124         size = size * 2;
 125     data = tcc_realloc(cstr->data_allocated, size);
 126     cstr->data_allocated = data;
 127     cstr->size_allocated = size;
 128     cstr->data = data;
 129 }
 130
 131 /* add a byte */
 132 ST_FUNC void cstr_ccat(CString *cstr, int ch)
 133 {
 134     int size;
 135     size = cstr->size + 1;
 136     if (size > cstr->size_allocated)
 137         cstr_realloc(cstr, size);
 138     ((unsigned char *)cstr->data)[size - 1] = ch;
 139     cstr->size = size;
 140 }
 141
 142 ST_FUNC void cstr_cat(CString *cstr, const char *str)
 143 {
 144     int c;
 145     for(;;) {
 146         c = *str;
 147         if (c == '\0')
 148             break;
 149         cstr_ccat(cstr, c);
 150         str++;
 151     }
 152 }
 153
 154 /* add a wide char */
 155 ST_FUNC void cstr_wccat(CString *cstr, int ch)
 156 {
 157     int size;
 158     size = cstr->size + sizeof(nwchar_t);
 159     if (size > cstr->size_allocated)
 160         cstr_realloc(cstr, size);
 161     *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
 162     cstr->size = size;
 163 }
 164
 165 ST_FUNC void cstr_new(CString *cstr)
 166 {
 167     memset(cstr, 0, sizeof(CString));
 168 }
 169
 170 /* free string and reset it to NULL */
 171 ST_FUNC void cstr_free(CString *cstr)
 172 {
 173     tcc_free(cstr->data_allocated);
 174     cstr_new(cstr);
 175 }
 176
 177 /* reset string to empty */
 178 ST_FUNC void cstr_reset(CString *cstr)
 179 {
 180     cstr->size = 0;
 181 }
 182
 183 /* XXX: unicode ? */
 184 static void add_char(CString *cstr, int c)
 185 {
 186     if (c == '\'' || c == '\"' || c == '\\') {
 187         /* XXX: could be more precise if char or string */
 188         cstr_ccat(cstr, '\\');
 189     }
 190     if (c >= 32 && c <= 126) {
 191         cstr_ccat(cstr, c);
 192     } else {
 193         cstr_ccat(cstr, '\\');
 194         if (c == '\n') {
 195             cstr_ccat(cstr, 'n');
 196         } else {
 197             cstr_ccat(cstr, '0' + ((c >> 6) & 7));
 198             cstr_ccat(cstr, '0' + ((c >> 3) & 7));
 199             cstr_ccat(cstr, '0' + (c & 7));
 200         }
 201     }
 202 }
 203
 204 /* ------------------------------------------------------------------------- */
 205 /* allocate a new token */
 206 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
 207 {
 208     TokenSym *ts, **ptable;
 209     int i;
 210
 211     if (tok_ident >= SYM_FIRST_ANOM)
 212         tcc_error("memory full (symbols)");
 213
 214     /* expand token table if needed */
 215     i = tok_ident - TOK_IDENT;
 216     if ((i % TOK_ALLOC_INCR) == 0) {
 217         ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
 218         table_ident = ptable;
 219     }
 220
 221     ts = tcc_malloc(sizeof(TokenSym) + len);
 222     table_ident[i] = ts;
 223     ts->tok = tok_ident++;
 224     ts->sym_define = NULL;
 225     ts->sym_label = NULL;
 226     ts->sym_struct = NULL;
 227     ts->sym_identifier = NULL;
 228     ts->len = len;
 229     ts->hash_next = NULL;
 230     memcpy(ts->str, str, len);
 231     ts->str[len] = '\0';
 232     *pts = ts;
 233     return ts;
 234 }
 235
 236 #define TOK_HASH_INIT 1
 237 #define TOK_HASH_FUNC(h, c) ((h) * 263 + (c))
 238
 239 /* find a token and add it if not found */
 240 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
 241 {
 242     TokenSym *ts, **pts;
 243     int i;
 244     unsigned int h;
 245
 246     h = TOK_HASH_INIT;
 247     for(i=0;i<len;i++)
 248         h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
 249     h &= (TOK_HASH_SIZE - 1);
 250
 251     pts = &hash_ident[h];
 252     for(;;) {
 253         ts = *pts;
 254         if (!ts)
 255             break;
 256         if (ts->len == len && !memcmp(ts->str, str, len))
 257             return ts;
 258         pts = &(ts->hash_next);
 259     }
 260     return tok_alloc_new(pts, str, len);
 261 }
 262
 263 /* XXX: buffer overflow */
 264 /* XXX: float tokens */
 265 ST_FUNC char *get_tok_str(int v, CValue *cv)
 266 {
 267     static char buf[STRING_MAX_SIZE + 1];
 268     static CString cstr_buf;
 269     CString *cstr;
 270     char *p;
 271     int i, len;
 272
 273     /* NOTE: to go faster, we give a fixed buffer for small strings */
 274     cstr_reset(&cstr_buf);
 275     cstr_buf.data = buf;
 276     cstr_buf.size_allocated = sizeof(buf);
 277     p = buf;
 278
 279 /* just an explanation, should never happen:
 280     if (v <= TOK_LINENUM && v >= TOK_CINT && cv == NULL)
 281         tcc_error("internal error: get_tok_str"); */
 282
 283     switch(v) {
 284     case TOK_CINT:
 285     case TOK_CUINT:
 286         /* XXX: not quite exact, but only useful for testing */
 287         sprintf(p, "%u", cv->ui);
 288         break;
 289     case TOK_CLLONG:
 290     case TOK_CULLONG:
 291         /* XXX: not quite exact, but only useful for testing  */
 292 #ifdef _WIN32
 293         sprintf(p, "%u", (unsigned)cv->ull);
 294 #else
 295         sprintf(p, "%llu", cv->ull);
 296 #endif
 297         break;
 298     case TOK_LCHAR:
 299         cstr_ccat(&cstr_buf, 'L');
 300     case TOK_CCHAR:
 301         cstr_ccat(&cstr_buf, '\'');
 302         add_char(&cstr_buf, cv->i);
 303         cstr_ccat(&cstr_buf, '\'');
 304         cstr_ccat(&cstr_buf, '\0');
 305         break;
 306     case TOK_PPNUM:
 307         cstr = cv->cstr;
 308         len = cstr->size - 1;
 309         for(i=0;i<len;i++)
 310             add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 311         cstr_ccat(&cstr_buf, '\0');
 312         break;
 313     case TOK_LSTR:
 314         cstr_ccat(&cstr_buf, 'L');
 315     case TOK_STR:
 316         cstr = cv->cstr;
 317         cstr_ccat(&cstr_buf, '\"');
 318         if (v == TOK_STR) {
 319             len = cstr->size - 1;
 320             for(i=0;i<len;i++)
 321                 add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 322         } else {
 323             len = (cstr->size / sizeof(nwchar_t)) - 1;
 324             for(i=0;i<len;i++)
 325                 add_char(&cstr_buf, ((nwchar_t *)cstr->data)[i]);
 326         }
 327         cstr_ccat(&cstr_buf, '\"');
 328         cstr_ccat(&cstr_buf, '\0');
 329         break;
 330
 331     case TOK_CFLOAT:
 332     case TOK_CDOUBLE:
 333     case TOK_CLDOUBLE:
 334     case TOK_LINENUM:
 335         return NULL; /* should not happen */
 336
 337     /* above tokens have value, the ones below don't */
 338
 339     case TOK_LT:
 340         v = '<';
 341         goto addv;
 342     case TOK_GT:
 343         v = '>';
 344         goto addv;
 345     case TOK_DOTS:
 346         return strcpy(p, "...");
 347     case TOK_A_SHL:
 348         return strcpy(p, "<<=");
 349     case TOK_A_SAR:
 350         return strcpy(p, ">>=");
 351     default:
 352         if (v < TOK_IDENT) {
 353             /* search in two bytes table */
 354             const unsigned char *q = tok_two_chars;
 355             while (*q) {
 356                 if (q[2] == v) {
 357                     *p++ = q[0];
 358                     *p++ = q[1];
 359                     *p = '\0';
 360                     return buf;
 361                 }
 362                 q += 3;
 363             }
 364         addv:
 365             *p++ = v;
 366             *p = '\0';
 367         } else if (v < tok_ident) {
 368             return table_ident[v - TOK_IDENT]->str;
 369         } else if (v >= SYM_FIRST_ANOM) {
 370             /* special name for anonymous symbol */
 371             sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
 372         } else {
 373             /* should never happen */
 374             return NULL;
 375         }
 376         break;
 377     }
 378     return cstr_buf.data;
 379 }
 380
 381 /* fill input buffer and peek next char */
 382 static int tcc_peekc_slow(BufferedFile *bf)
 383 {
 384     int len;
 385     /* only tries to read if really end of buffer */
 386     if (bf->buf_ptr >= bf->buf_end) {
 387         if (bf->fd != -1) {
 388 #if defined(PARSE_DEBUG)
 389             len = 8;
 390 #else
 391             len = IO_BUF_SIZE;
 392 #endif
 393             len = read(bf->fd, bf->buffer, len);
 394             if (len < 0)
 395                 len = 0;
 396         } else {
 397             len = 0;
 398         }
 399         total_bytes += len;
 400         bf->buf_ptr = bf->buffer;
 401         bf->buf_end = bf->buffer + len;
 402         *bf->buf_end = CH_EOB;
 403     }
 404     if (bf->buf_ptr < bf->buf_end) {
 405         return bf->buf_ptr[0];
 406     } else {
 407         bf->buf_ptr = bf->buf_end;
 408         return CH_EOF;
 409     }
 410 }
 411
 412 /* return the current character, handling end of block if necessary
 413    (but not stray) */
 414 ST_FUNC int handle_eob(void)
 415 {
 416     return tcc_peekc_slow(file);
 417 }
 418
 419 /* read next char from current input file and handle end of input buffer */
 420 ST_INLN void inp(void)
 421 {
 422     ch = *(++(file->buf_ptr));
 423     /* end of buffer/file handling */
 424     if (ch == CH_EOB)
 425         ch = handle_eob();
 426 }
 427
 428 /* handle '\[\r]\n' */
 429 static int handle_stray_noerror(void)
 430 {
 431     while (ch == '\\') {
 432         inp();
 433         if (ch == '\n') {
 434             file->line_num++;
 435             inp();
 436         } else if (ch == '\r') {
 437             inp();
 438             if (ch != '\n')
 439                 goto fail;
 440             file->line_num++;
 441             inp();
 442         } else {
 443         fail:
 444             return 1;
 445         }
 446     }
 447     return 0;
 448 }
 449
 450 static void handle_stray(void)
 451 {
 452     if (handle_stray_noerror())
 453         tcc_error("stray '\\' in program");
 454 }
 455
 456 /* skip the stray and handle the \\n case. Output an error if
 457    incorrect char after the stray */
 458 static int handle_stray1(uint8_t *p)
 459 {
 460     int c;
 461
 462     if (p >= file->buf_end) {
 463         file->buf_ptr = p;
 464         c = handle_eob();
 465         p = file->buf_ptr;
 466         if (c == '\\')
 467             goto parse_stray;
 468     } else {
 469     parse_stray:
 470         file->buf_ptr = p;
 471         ch = *p;
 472         handle_stray();
 473         p = file->buf_ptr;
 474         c = *p;
 475     }
 476     return c;
 477 }
 478
 479 /* handle just the EOB case, but not stray */
 480 #define PEEKC_EOB(c, p)\
 481 {\
 482     p++;\
 483     c = *p;\
 484     if (c == '\\') {\
 485         file->buf_ptr = p;\
 486         c = handle_eob();\
 487         p = file->buf_ptr;\
 488     }\
 489 }
 490
 491 /* handle the complicated stray case */
 492 #define PEEKC(c, p)\
 493 {\
 494     p++;\
 495     c = *p;\
 496     if (c == '\\') {\
 497         c = handle_stray1(p);\
 498         p = file->buf_ptr;\
 499     }\
 500 }
 501
 502 /* input with '\[\r]\n' handling. Note that this function cannot
 503    handle other characters after '\', so you cannot call it inside
 504    strings or comments */
 505 ST_FUNC void minp(void)
 506 {
 507     inp();
 508     if (ch == '\\')
 509         handle_stray();
 510 }
 511
 512
 513 /* single line C++ comments */
 514 static uint8_t *parse_line_comment(uint8_t *p)
 515 {
 516     int c;
 517
 518     p++;
 519     for(;;) {
 520         c = *p;
 521     redo:
 522         if (c == '\n' || c == CH_EOF) {
 523             break;
 524         } else if (c == '\\') {
 525             file->buf_ptr = p;
 526             c = handle_eob();
 527             p = file->buf_ptr;
 528             if (c == '\\') {
 529                 PEEKC_EOB(c, p);
 530                 if (c == '\n') {
 531                     file->line_num++;
 532                     PEEKC_EOB(c, p);
 533                 } else if (c == '\r') {
 534                     PEEKC_EOB(c, p);
 535                     if (c == '\n') {
 536                         file->line_num++;
 537                         PEEKC_EOB(c, p);
 538                     }
 539                 }
 540             } else {
 541                 goto redo;
 542             }
 543         } else {
 544             p++;
 545         }
 546     }
 547     return p;
 548 }
 549
 550 /* C comments */
 551 ST_FUNC uint8_t *parse_comment(uint8_t *p)
 552 {
 553     int c;
 554
 555     p++;
 556     for(;;) {
 557         /* fast skip loop */
 558         for(;;) {
 559             c = *p;
 560             if (c == '\n' || c == '*' || c == '\\')
 561                 break;
 562             p++;
 563             c = *p;
 564             if (c == '\n' || c == '*' || c == '\\')
 565                 break;
 566             p++;
 567         }
 568         /* now we can handle all the cases */
 569         if (c == '\n') {
 570             file->line_num++;
 571             p++;
 572         } else if (c == '*') {
 573             p++;
 574             for(;;) {
 575                 c = *p;
 576                 if (c == '*') {
 577                     p++;
 578                 } else if (c == '/') {
 579                     goto end_of_comment;
 580                 } else if (c == '\\') {
 581                     file->buf_ptr = p;
 582                     c = handle_eob();
 583                     p = file->buf_ptr;
 584                     if (c == '\\') {
 585                         /* skip '\[\r]\n', otherwise just skip the stray */
 586                         while (c == '\\') {
 587                             PEEKC_EOB(c, p);
 588                             if (c == '\n') {
 589                                 file->line_num++;
 590                                 PEEKC_EOB(c, p);
 591                             } else if (c == '\r') {
 592                                 PEEKC_EOB(c, p);
 593                                 if (c == '\n') {
 594                                     file->line_num++;
 595                                     PEEKC_EOB(c, p);
 596                                 }
 597                             } else {
 598                                 goto after_star;
 599                             }
 600                         }
 601                     }
 602                 } else {
 603                     break;
 604                 }
 605             }
 606         after_star: ;
 607         } else {
 608             /* stray, eob or eof */
 609             file->buf_ptr = p;
 610             c = handle_eob();
 611             p = file->buf_ptr;
 612             if (c == CH_EOF) {
 613                 tcc_error("unexpected end of file in comment");
 614             } else if (c == '\\') {
 615                 p++;
 616             }
 617         }
 618     }
 619  end_of_comment:
 620     p++;
 621     return p;
 622 }
 623
 624 #define cinp minp
 625
 626 static inline void skip_spaces(void)
 627 {
 628     while (is_space(ch))
 629         cinp();
 630 }
 631
 632 static inline int check_space(int t, int *spc)
 633 {
 634     if (is_space(t)) {
 635         if (*spc)
 636             return 1;
 637         *spc = 1;
 638     } else
 639         *spc = 0;
 640     return 0;
 641 }
 642
 643 /* parse a string without interpreting escapes */
 644 static uint8_t *parse_pp_string(uint8_t *p,
 645                                 int sep, CString *str)
 646 {
 647     int c;
 648     p++;
 649     for(;;) {
 650         c = *p;
 651         if (c == sep) {
 652             break;
 653         } else if (c == '\\') {
 654             file->buf_ptr = p;
 655             c = handle_eob();
 656             p = file->buf_ptr;
 657             if (c == CH_EOF) {
 658             unterminated_string:
 659                 /* XXX: indicate line number of start of string */
 660                 tcc_error("missing terminating %c character", sep);
 661             } else if (c == '\\') {
 662                 /* escape : just skip \[\r]\n */
 663                 PEEKC_EOB(c, p);
 664                 if (c == '\n') {
 665                     file->line_num++;
 666                     p++;
 667                 } else if (c == '\r') {
 668                     PEEKC_EOB(c, p);
 669                     if (c != '\n')
 670                         expect("'\n' after '\r'");
 671                     file->line_num++;
 672                     p++;
 673                 } else if (c == CH_EOF) {
 674                     goto unterminated_string;
 675                 } else {
 676                     if (str) {
 677                         cstr_ccat(str, '\\');
 678                         cstr_ccat(str, c);
 679                     }
 680                     p++;
 681                 }
 682             }
 683         } else if (c == '\n') {
 684             file->line_num++;
 685             goto add_char;
 686         } else if (c == '\r') {
 687             PEEKC_EOB(c, p);
 688             if (c != '\n') {
 689                 if (str)
 690                     cstr_ccat(str, '\r');
 691             } else {
 692                 file->line_num++;
 693                 goto add_char;
 694             }
 695         } else {
 696         add_char:
 697             if (str)
 698                 cstr_ccat(str, c);
 699             p++;
 700         }
 701     }
 702     p++;
 703     return p;
 704 }
 705
 706 /* skip block of text until #else, #elif or #endif. skip also pairs of
 707    #if/#endif */
 708 static void preprocess_skip(void)
 709 {
 710     int a, start_of_line, c, in_warn_or_error;
 711     uint8_t *p;
 712
 713     p = file->buf_ptr;
 714     a = 0;
 715 redo_start:
 716     start_of_line = 1;
 717     in_warn_or_error = 0;
 718     for(;;) {
 719     redo_no_start:
 720         c = *p;
 721         switch(c) {
 722         case ' ':
 723         case '\t':
 724         case '\f':
 725         case '\v':
 726         case '\r':
 727             p++;
 728             goto redo_no_start;
 729         case '\n':
 730             file->line_num++;
 731             p++;
 732             goto redo_start;
 733         case '\\':
 734             file->buf_ptr = p;
 735             c = handle_eob();
 736             if (c == CH_EOF) {
 737                 expect("#endif");
 738             } else if (c == '\\') {
 739                 ch = file->buf_ptr[0];
 740                 handle_stray_noerror();
 741             }
 742             p = file->buf_ptr;
 743             goto redo_no_start;
 744         /* skip strings */
 745         case '\"':
 746         case '\'':
 747             if (in_warn_or_error)
 748                 goto _default;
 749             p = parse_pp_string(p, c, NULL);
 750             break;
 751         /* skip comments */
 752         case '/':
 753             if (in_warn_or_error)
 754                 goto _default;
 755             file->buf_ptr = p;
 756             ch = *p;
 757             minp();
 758             p = file->buf_ptr;
 759             if (ch == '*') {
 760                 p = parse_comment(p);
 761             } else if (ch == '/') {
 762                 p = parse_line_comment(p);
 763             }
 764             break;
 765         case '#':
 766             p++;
 767             if (start_of_line) {
 768                 file->buf_ptr = p;
 769                 next_nomacro();
 770                 p = file->buf_ptr;
 771                 if (a == 0 &&
 772                     (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
 773                     goto the_end;
 774                 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
 775                     a++;
 776                 else if (tok == TOK_ENDIF)
 777                     a--;
 778                 else if( tok == TOK_ERROR || tok == TOK_WARNING)
 779                     in_warn_or_error = 1;
 780                 else if (tok == TOK_LINEFEED)
 781                     goto redo_start;
 782                 else if (parse_flags & PARSE_FLAG_ASM_COMMENTS)
 783                     p = parse_line_comment(p);
 784             }
 785             else if (parse_flags & PARSE_FLAG_ASM_FILE)
 786                 p = parse_line_comment(p);
 787             break;
 788 _default:
 789         default:
 790             p++;
 791             break;
 792         }
 793         start_of_line = 0;
 794     }
 795  the_end: ;
 796     file->buf_ptr = p;
 797 }
 798
 799 /* ParseState handling */
 800
 801 /* XXX: currently, no include file info is stored. Thus, we cannot display
 802    accurate messages if the function or data definition spans multiple
 803    files */
 804
 805 /* save current parse state in 's' */
 806 ST_FUNC void save_parse_state(ParseState *s)
 807 {
 808     s->line_num = file->line_num;
 809     s->macro_ptr = macro_ptr;
 810     s->tok = tok;
 811     s->tokc = tokc;
 812 }
 813
 814 /* restore parse state from 's' */
 815 ST_FUNC void restore_parse_state(ParseState *s)
 816 {
 817     file->line_num = s->line_num;
 818     macro_ptr = s->macro_ptr;
 819     tok = s->tok;
 820     tokc = s->tokc;
 821 }
 822
 823 /* return the number of additional 'ints' necessary to store the
 824    token */
 825 static inline int tok_ext_size(int t)
 826 {
 827     switch(t) {
 828         /* 4 bytes */
 829     case TOK_CINT:
 830     case TOK_CUINT:
 831     case TOK_CCHAR:
 832     case TOK_LCHAR:
 833     case TOK_CFLOAT:
 834     case TOK_LINENUM:
 835         return 1;
 836     case TOK_STR:
 837     case TOK_LSTR:
 838     case TOK_PPNUM:
 839         tcc_error("unsupported token");
 840         return 1;
 841     case TOK_CDOUBLE:
 842     case TOK_CLLONG:
 843     case TOK_CULLONG:
 844         return 2;
 845     case TOK_CLDOUBLE:
 846         return LDOUBLE_SIZE / 4;
 847     default:
 848         return 0;
 849     }
 850 }
 851
 852 /* token string handling */
 853
 854 ST_INLN void tok_str_new(TokenString *s)
 855 {
 856     s->str = NULL;
 857     s->len = 0;
 858     s->allocated_len = 0;
 859     s->last_line_num = -1;
 860 }
 861
 862 ST_FUNC void tok_str_free(int *str)
 863 {
 864     tcc_free(str);
 865 }
 866
 867 static int *tok_str_realloc(TokenString *s)
 868 {
 869     int *str, len;
 870
 871     if (s->allocated_len == 0) {
 872         len = 8;
 873     } else {
 874         len = s->allocated_len * 2;
 875     }
 876     str = tcc_realloc(s->str, len * sizeof(int));
 877     s->allocated_len = len;
 878     s->str = str;
 879     return str;
 880 }
 881
 882 ST_FUNC void tok_str_add(TokenString *s, int t)
 883 {
 884     int len, *str;
 885
 886     len = s->len;
 887     str = s->str;
 888     if (len >= s->allocated_len)
 889         str = tok_str_realloc(s);
 890     str[len++] = t;
 891     s->len = len;
 892 }
 893
 894 static void tok_str_add2(TokenString *s, int t, CValue *cv)
 895 {
 896     int len, *str;
 897
 898     len = s->len;
 899     str = s->str;
 900
 901     /* allocate space for worst case */
 902     if (len + TOK_MAX_SIZE > s->allocated_len)
 903         str = tok_str_realloc(s);
 904     str[len++] = t;
 905     switch(t) {
 906     case TOK_CINT:
 907     case TOK_CUINT:
 908     case TOK_CCHAR:
 909     case TOK_LCHAR:
 910     case TOK_CFLOAT:
 911     case TOK_LINENUM:
 912         str[len++] = cv->tab[0];
 913         break;
 914     case TOK_PPNUM:
 915     case TOK_STR:
 916     case TOK_LSTR:
 917         {
 918             int nb_words;
 919             CString *cstr;
 920
 921             nb_words = (sizeof(CString) + cv->cstr->size + 3) >> 2;
 922             while ((len + nb_words) > s->allocated_len)
 923                 str = tok_str_realloc(s);
 924             cstr = (CString *)(str + len);
 925             cstr->data = NULL;
 926             cstr->size = cv->cstr->size;
 927             cstr->data_allocated = NULL;
 928             cstr->size_allocated = cstr->size;
 929             memcpy((char *)cstr + sizeof(CString),
 930                    cv->cstr->data, cstr->size);
 931             len += nb_words;
 932         }
 933         break;
 934     case TOK_CDOUBLE:
 935     case TOK_CLLONG:
 936     case TOK_CULLONG:
 937 #if LDOUBLE_SIZE == 8
 938     case TOK_CLDOUBLE:
 939 #endif
 940         str[len++] = cv->tab[0];
 941         str[len++] = cv->tab[1];
 942         break;
 943 #if LDOUBLE_SIZE == 12
 944     case TOK_CLDOUBLE:
 945         str[len++] = cv->tab[0];
 946         str[len++] = cv->tab[1];
 947         str[len++] = cv->tab[2];
 948 #elif LDOUBLE_SIZE == 16
 949     case TOK_CLDOUBLE:
 950         str[len++] = cv->tab[0];
 951         str[len++] = cv->tab[1];
 952         str[len++] = cv->tab[2];
 953         str[len++] = cv->tab[3];
 954 #elif LDOUBLE_SIZE != 8
 955 #error add long double size support
 956 #endif
 957         break;
 958     default:
 959         break;
 960     }
 961     s->len = len;
 962 }
 963
 964 /* add the current parse token in token string 's' */
 965 ST_FUNC void tok_str_add_tok(TokenString *s)
 966 {
 967     CValue cval;
 968
 969     /* save line number info */
 970     if (file->line_num != s->last_line_num) {
 971         s->last_line_num = file->line_num;
 972         cval.i = s->last_line_num;
 973         tok_str_add2(s, TOK_LINENUM, &cval);
 974     }
 975     tok_str_add2(s, tok, &tokc);
 976 }
 977
 978 /* get a token from an integer array and increment pointer
 979    accordingly. we code it as a macro to avoid pointer aliasing. */
 980 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
 981 {
 982     const int *p = *pp;
 983     int n, *tab;
 984
 985     tab = cv->tab;
 986     switch(*t = *p++) {
 987     case TOK_CINT:
 988     case TOK_CUINT:
 989     case TOK_CCHAR:
 990     case TOK_LCHAR:
 991     case TOK_CFLOAT:
 992     case TOK_LINENUM:
 993         tab[0] = *p++;
 994         break;
 995     case TOK_STR:
 996     case TOK_LSTR:
 997     case TOK_PPNUM:
 998         cv->cstr = (CString *)p;
 999         cv->cstr->data = (char *)p + sizeof(CString);
1000         p += (sizeof(CString) + cv->cstr->size + 3) >> 2;
1001         break;
1002     case TOK_CDOUBLE:
1003     case TOK_CLLONG:
1004     case TOK_CULLONG:
1005         n = 2;
1006         goto copy;
1007     case TOK_CLDOUBLE:
1008 #if LDOUBLE_SIZE == 16
1009         n = 4;
1010 #elif LDOUBLE_SIZE == 12
1011         n = 3;
1012 #elif LDOUBLE_SIZE == 8
1013         n = 2;
1014 #else
1015 # error add long double size support
1016 #endif
1017     copy:
1018         do
1019             *tab++ = *p++;
1020         while (--n);
1021         break;
1022     default:
1023         break;
1024     }
1025     *pp = p;
1026 }
1027
1028 static int macro_is_equal(const int *a, const int *b)
1029 {
1030     char buf[STRING_MAX_SIZE + 1];
1031     CValue cv;
1032     int t;
1033     while (*a && *b) {
1034         TOK_GET(&t, &a, &cv);
1035         pstrcpy(buf, sizeof buf, get_tok_str(t, &cv));
1036         TOK_GET(&t, &b, &cv);
1037         if (strcmp(buf, get_tok_str(t, &cv)))
1038             return 0;
1039     }
1040     return !(*a || *b);
1041 }
1042
1043 static void define_print(Sym *s, int is_undef)
1044 {
1045     int c, t;
1046     CValue cval;
1047     const int *str;
1048     Sym *arg;
1049
1050     if (tcc_state->dflag == 0 || !s || !tcc_state->ppfp)
1051         return;
1052
1053     if (file) {
1054         c = file->line_num - file->line_ref - 1;
1055         if (c > 0) {
1056             while (c--)
1057                 fputs("\n", tcc_state->ppfp);
1058             file->line_ref = file->line_num;
1059         }
1060     }
1061
1062     if (is_undef) {
1063         fprintf(tcc_state->ppfp, "// #undef %s\n", get_tok_str(s->v, NULL));
1064         return;
1065     }
1066
1067     fprintf(tcc_state->ppfp, "// #define %s", get_tok_str(s->v, NULL));
1068     arg = s->next;
1069     if (arg) {
1070         char *sep = "(";
1071         while (arg) {
1072             fprintf(tcc_state->ppfp, "%s%s", sep, get_tok_str(arg->v & ~SYM_FIELD, NULL));
1073             sep = ",";
1074             arg = arg->next;
1075         }
1076         fprintf(tcc_state->ppfp, ")");
1077     }
1078
1079     str = s->d;
1080     if (str)
1081         fprintf(tcc_state->ppfp, " ");
1082
1083     while (str) {
1084         TOK_GET(&t, &str, &cval);
1085         if (!t)
1086             break;
1087         fprintf(tcc_state->ppfp, "%s", get_tok_str(t, &cval));
1088     }
1089     fprintf(tcc_state->ppfp, "\n");
1090 }
1091
1092 /* defines handling */
1093 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1094 {
1095     Sym *s;
1096
1097     s = define_find(v);
1098     if (s && !macro_is_equal(s->d, str))
1099         tcc_warning("%s redefined", get_tok_str(v, NULL));
1100
1101     s = sym_push2(&define_stack, v, macro_type, 0);
1102     s->d = str;
1103     s->next = first_arg;
1104     table_ident[v - TOK_IDENT]->sym_define = s;
1105     define_print(s, 0);
1106 }
1107
1108 /* undefined a define symbol. Its name is just set to zero */
1109 ST_FUNC void define_undef(Sym *s)
1110 {
1111     int v;
1112     v = s->v;
1113     if (v >= TOK_IDENT && v < tok_ident) {
1114         define_print(s, 1);
1115         table_ident[v - TOK_IDENT]->sym_define = NULL;
1116     }
1117     s->v = 0;
1118 }
1119
1120 ST_INLN Sym *define_find(int v)
1121 {
1122     v -= TOK_IDENT;
1123     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1124         return NULL;
1125     return table_ident[v]->sym_define;
1126 }
1127
1128 /* free define stack until top reaches 'b' */
1129 ST_FUNC void free_defines(Sym *b)
1130 {
1131     Sym *top, *top1;
1132     int v;
1133
1134     top = define_stack;
1135     while (top != b) {
1136         top1 = top->prev;
1137         /* do not free args or predefined defines */
1138         if (top->d)
1139             tok_str_free(top->d);
1140         v = top->v;
1141         if (v >= TOK_IDENT && v < tok_ident)
1142             table_ident[v - TOK_IDENT]->sym_define = NULL;
1143         sym_free(top);
1144         top = top1;
1145     }
1146     define_stack = b;
1147 }
1148
1149 ST_FUNC void print_defines(void)
1150 {
1151     Sym *top, *s;
1152     int v;
1153
1154     top = define_stack;
1155     while (top) {
1156         v = top->v;
1157         if (v >= TOK_IDENT && v < tok_ident) {
1158             s = table_ident[v - TOK_IDENT]->sym_define;
1159             define_print(s, 0);
1160         }
1161         top = top->prev;
1162     }
1163 }
1164
1165 /* label lookup */
1166 ST_FUNC Sym *label_find(int v)
1167 {
1168     v -= TOK_IDENT;
1169     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1170         return NULL;
1171     return table_ident[v]->sym_label;
1172 }
1173
1174 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1175 {
1176     Sym *s, **ps;
1177     s = sym_push2(ptop, v, 0, 0);
1178     s->r = flags;
1179     ps = &table_ident[v - TOK_IDENT]->sym_label;
1180     if (ptop == &global_label_stack) {
1181         /* modify the top most local identifier, so that
1182            sym_identifier will point to 's' when popped */
1183         while (*ps != NULL)
1184             ps = &(*ps)->prev_tok;
1185     }
1186     s->prev_tok = *ps;
1187     *ps = s;
1188     return s;
1189 }
1190
1191 /* pop labels until element last is reached. Look if any labels are
1192    undefined. Define symbols if '&&label' was used. */
1193 ST_FUNC void label_pop(Sym **ptop, Sym *slast)
1194 {
1195     Sym *s, *s1;
1196     for(s = *ptop; s != slast; s = s1) {
1197         s1 = s->prev;
1198         if (s->r == LABEL_DECLARED) {
1199             tcc_warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1200         } else if (s->r == LABEL_FORWARD) {
1201                 tcc_error("label '%s' used but not defined",
1202                       get_tok_str(s->v, NULL));
1203         } else {
1204             if (s->c) {
1205                 /* define corresponding symbol. A size of
1206                    1 is put. */
1207                 put_extern_sym(s, cur_text_section, s->jnext, 1);
1208             }
1209         }
1210         /* remove label */
1211         table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1212         sym_free(s);
1213     }
1214     *ptop = slast;
1215 }
1216
1217 /* eval an expression for #if/#elif */
1218 static int expr_preprocess(void)
1219 {
1220     int c, t;
1221     TokenString str;
1222
1223     tok_str_new(&str);
1224     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1225         next(); /* do macro subst */
1226         if (tok == TOK_DEFINED) {
1227             next_nomacro();
1228             t = tok;
1229             if (t == '(')
1230                 next_nomacro();
1231             c = define_find(tok) != 0;
1232             if (t == '(')
1233                 next_nomacro();
1234             tok = TOK_CINT;
1235             tokc.i = c;
1236         } else if (tok >= TOK_IDENT) {
1237             /* if undefined macro */
1238             tok = TOK_CINT;
1239             tokc.i = 0;
1240         }
1241         tok_str_add_tok(&str);
1242     }
1243     tok_str_add(&str, -1); /* simulate end of file */
1244     tok_str_add(&str, 0);
1245     /* now evaluate C constant expression */
1246     macro_ptr = str.str;
1247     next();
1248     c = expr_const();
1249     macro_ptr = NULL;
1250     tok_str_free(str.str);
1251     return c != 0;
1252 }
1253
1254 /* parse after #define */
1255 ST_FUNC void parse_define(void)
1256 {
1257     Sym *s, *first, **ps;
1258     int v, t, varg, is_vaargs, spc, ptok, macro_list_start;
1259     TokenString str;
1260
1261     v = tok;
1262     if (v < TOK_IDENT)
1263         tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1264     /* XXX: should check if same macro (ANSI) */
1265     first = NULL;
1266     t = MACRO_OBJ;
1267     /* '(' must be just after macro definition for MACRO_FUNC */
1268     next_nomacro_spc();
1269     if (tok == '(') {
1270         next_nomacro();
1271         ps = &first;
1272         while (tok != ')') {
1273             varg = tok;
1274             next_nomacro();
1275             is_vaargs = 0;
1276             if (varg == TOK_DOTS) {
1277                 varg = TOK___VA_ARGS__;
1278                 is_vaargs = 1;
1279             } else if (tok == TOK_DOTS && gnu_ext) {
1280                 is_vaargs = 1;
1281                 next_nomacro();
1282             }
1283             if (varg < TOK_IDENT)
1284                 tcc_error( "\'%s\' may not appear in parameter list", get_tok_str(varg, NULL));
1285             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1286             *ps = s;
1287             ps = &s->next;
1288             if (tok != ',')
1289                 continue;
1290             next_nomacro();
1291         }
1292         next_nomacro_spc();
1293         t = MACRO_FUNC;
1294     }
1295     tok_str_new(&str);
1296     spc = 2;
1297     /* EOF testing necessary for '-D' handling */
1298     ptok = 0;
1299     macro_list_start = 1;
1300     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1301         if (!macro_list_start && spc == 2 && tok == TOK_TWOSHARPS)
1302             tcc_error("'##' invalid at start of macro");
1303         ptok = tok;
1304         /* remove spaces around ## and after '#' */
1305         if (TOK_TWOSHARPS == tok) {
1306             if (1 == spc)
1307                 --str.len;
1308             spc = 2;
1309         } else if ('#' == tok) {
1310             spc = 2;
1311         } else if (check_space(tok, &spc)) {
1312             goto skip;
1313         }
1314         tok_str_add2(&str, tok, &tokc);
1315     skip:
1316         next_nomacro_spc();
1317         macro_list_start = 0;
1318     }
1319     if (ptok == TOK_TWOSHARPS)
1320         tcc_error("'##' invalid at end of macro");
1321     if (spc == 1)
1322         --str.len; /* remove trailing space */
1323     tok_str_add(&str, 0);
1324     define_push(v, t, str.str, first);
1325 }
1326
1327 static inline int hash_cached_include(const char *filename)
1328 {
1329     const unsigned char *s;
1330     unsigned int h;
1331
1332     h = TOK_HASH_INIT;
1333     s = (unsigned char *) filename;
1334     while (*s) {
1335         h = TOK_HASH_FUNC(h, *s);
1336         s++;
1337     }
1338     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1339     return h;
1340 }
1341
1342 static CachedInclude *search_cached_include(TCCState *s1, const char *filename)
1343 {
1344     CachedInclude *e;
1345     int i, h;
1346     h = hash_cached_include(filename);
1347     i = s1->cached_includes_hash[h];
1348     for(;;) {
1349         if (i == 0)
1350             break;
1351         e = s1->cached_includes[i - 1];
1352         if (0 == PATHCMP(e->filename, filename))
1353             return e;
1354         i = e->hash_next;
1355     }
1356     return NULL;
1357 }
1358
1359 static inline void add_cached_include(TCCState *s1, const char *filename, int ifndef_macro)
1360 {
1361     CachedInclude *e;
1362     int h;
1363
1364     if (search_cached_include(s1, filename))
1365         return;
1366 #ifdef INC_DEBUG
1367     printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1368 #endif
1369     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1370     strcpy(e->filename, filename);
1371     e->ifndef_macro = ifndef_macro;
1372     dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1373     /* add in hash table */
1374     h = hash_cached_include(filename);
1375     e->hash_next = s1->cached_includes_hash[h];
1376     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1377 }
1378
1379 static void pragma_parse(TCCState *s1)
1380 {
1381     int val;
1382
1383     next();
1384     if (tok == TOK_pack) {
1385         /*
1386           This may be:
1387           #pragma pack(1) // set
1388           #pragma pack() // reset to default
1389           #pragma pack(push,1) // push & set
1390           #pragma pack(pop) // restore previous
1391         */
1392         next();
1393         skip('(');
1394         if (tok == TOK_ASM_pop) {
1395             next();
1396             if (s1->pack_stack_ptr <= s1->pack_stack) {
1397             stk_error:
1398                 tcc_error("out of pack stack");
1399             }
1400             s1->pack_stack_ptr--;
1401         } else {
1402             val = 0;
1403             if (tok != ')') {
1404                 if (tok == TOK_ASM_push) {
1405                     next();
1406                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1407                         goto stk_error;
1408                     s1->pack_stack_ptr++;
1409                     skip(',');
1410                 }
1411                 if (tok != TOK_CINT) {
1412                 pack_error:
1413                     tcc_error("invalid pack pragma");
1414                 }
1415                 val = tokc.i;
1416                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1417                     goto pack_error;
1418                 next();
1419             }
1420             *s1->pack_stack_ptr = val;
1421             skip(')');
1422         }
1423     }
1424 }
1425
1426 /* is_bof is true if first non space token at beginning of file */
1427 ST_FUNC void preprocess(int is_bof)
1428 {
1429     TCCState *s1 = tcc_state;
1430     int i, c, n, saved_parse_flags;
1431     char buf[1024], *q;
1432     Sym *s;
1433
1434     saved_parse_flags = parse_flags;
1435     parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM |
1436         PARSE_FLAG_LINEFEED;
1437     parse_flags |= (saved_parse_flags & (PARSE_FLAG_ASM_FILE | PARSE_FLAG_ASM_COMMENTS));
1438     next_nomacro();
1439  redo:
1440     switch(tok) {
1441     case TOK_DEFINE:
1442         next_nomacro();
1443         parse_define();
1444         break;
1445     case TOK_UNDEF:
1446         next_nomacro();
1447         s = define_find(tok);
1448         /* undefine symbol by putting an invalid name */
1449         if (s)
1450             define_undef(s);
1451         break;
1452     case TOK_INCLUDE:
1453     case TOK_INCLUDE_NEXT:
1454         ch = file->buf_ptr[0];
1455         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1456         skip_spaces();
1457         if (ch == '<') {
1458             c = '>';
1459             goto read_name;
1460         } else if (ch == '\"') {
1461             c = ch;
1462         read_name:
1463             inp();
1464             q = buf;
1465             while (ch != c && ch != '\n' && ch != CH_EOF) {
1466                 if ((q - buf) < sizeof(buf) - 1)
1467                     *q++ = ch;
1468                 if (ch == '\\') {
1469                     if (handle_stray_noerror() == 0)
1470                         --q;
1471                 } else
1472                     inp();
1473             }
1474             *q = '\0';
1475             minp();
1476 #if 0
1477             /* eat all spaces and comments after include */
1478             /* XXX: slightly incorrect */
1479             while (ch1 != '\n' && ch1 != CH_EOF)
1480                 inp();
1481 #endif
1482         } else {
1483             /* computed #include : either we have only strings or
1484                we have anything enclosed in '<>' */
1485             next();
1486             buf[0] = '\0';
1487             if (tok == TOK_STR) {
1488                 while (tok != TOK_LINEFEED) {
1489                     if (tok != TOK_STR) {
1490                     include_syntax:
1491                         tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1492                     }
1493                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1494                     next();
1495                 }
1496                 c = '\"';
1497             } else {
1498                 int len;
1499                 while (tok != TOK_LINEFEED) {
1500                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1501                     next();
1502                 }
1503                 len = strlen(buf);
1504                 /* check syntax and remove '<>' */
1505                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1506                     goto include_syntax;
1507                 memmove(buf, buf + 1, len - 2);
1508                 buf[len - 2] = '\0';
1509                 c = '>';
1510             }
1511         }
1512
1513         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1514             tcc_error("#include recursion too deep");
1515         /* store current file in stack, but increment stack later below */
1516         *s1->include_stack_ptr = file;
1517
1518         n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1519         for (i = -2; i < n; ++i) {
1520             char buf1[sizeof file->filename];
1521             CachedInclude *e;
1522             BufferedFile **f;
1523             const char *path;
1524
1525             if (i == -2) {
1526                 /* check absolute include path */
1527                 if (!IS_ABSPATH(buf))
1528                     continue;
1529                 buf1[0] = 0;
1530                 i = n; /* force end loop */
1531
1532             } else if (i == -1) {
1533                 /* search in current dir if "header.h" */
1534                 if (c != '\"')
1535                     continue;
1536                 path = file->filename;
1537                 pstrncpy(buf1, path, tcc_basename(path) - path);
1538
1539             } else {
1540                 /* search in all the include paths */
1541                 if (i < s1->nb_include_paths)
1542                     path = s1->include_paths[i];
1543                 else
1544                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1545                 pstrcpy(buf1, sizeof(buf1), path);
1546                 pstrcat(buf1, sizeof(buf1), "/");
1547             }
1548
1549             pstrcat(buf1, sizeof(buf1), buf);
1550
1551             if (tok == TOK_INCLUDE_NEXT)
1552                 for (f = s1->include_stack_ptr; f >= s1->include_stack; --f)
1553                     if (0 == PATHCMP((*f)->filename, buf1)) {
1554 #ifdef INC_DEBUG
1555                         printf("%s: #include_next skipping %s\n", file->filename, buf1);
1556 #endif
1557                         goto include_trynext;
1558                     }
1559
1560             e = search_cached_include(s1, buf1);
1561             if (e && define_find(e->ifndef_macro)) {
1562                 /* no need to parse the include because the 'ifndef macro'
1563                    is defined */
1564 #ifdef INC_DEBUG
1565                 printf("%s: skipping cached %s\n", file->filename, buf1);
1566 #endif
1567                 goto include_done;
1568             }
1569
1570             if (tcc_open(s1, buf1) < 0)
1571 include_trynext:
1572                 continue;
1573
1574 #ifdef INC_DEBUG
1575             printf("%s: including %s\n", file->prev->filename, file->filename);
1576 #endif
1577             /* update target deps */
1578             dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1579                     tcc_strdup(buf1));
1580             /* push current file in stack */
1581             ++s1->include_stack_ptr;
1582             /* add include file debug info */
1583             if (s1->do_debug)
1584                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1585             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1586             ch = file->buf_ptr[0];
1587             goto the_end;
1588         }
1589         tcc_error("include file '%s' not found", buf);
1590 include_done:
1591         break;
1592     case TOK_IFNDEF:
1593         c = 1;
1594         goto do_ifdef;
1595     case TOK_IF:
1596         c = expr_preprocess();
1597         goto do_if;
1598     case TOK_IFDEF:
1599         c = 0;
1600     do_ifdef:
1601         next_nomacro();
1602         if (tok < TOK_IDENT)
1603             tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1604         if (is_bof) {
1605             if (c) {
1606 #ifdef INC_DEBUG
1607                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1608 #endif
1609                 file->ifndef_macro = tok;
1610             }
1611         }
1612         c = (define_find(tok) != 0) ^ c;
1613     do_if:
1614         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1615             tcc_error("memory full (ifdef)");
1616         *s1->ifdef_stack_ptr++ = c;
1617         goto test_skip;
1618     case TOK_ELSE:
1619         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1620             tcc_error("#else without matching #if");
1621         if (s1->ifdef_stack_ptr[-1] & 2)
1622             tcc_error("#else after #else");
1623         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1624         goto test_else;
1625     case TOK_ELIF:
1626         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1627             tcc_error("#elif without matching #if");
1628         c = s1->ifdef_stack_ptr[-1];
1629         if (c > 1)
1630             tcc_error("#elif after #else");
1631         /* last #if/#elif expression was true: we skip */
1632         if (c == 1)
1633             goto skip;
1634         c = expr_preprocess();
1635         s1->ifdef_stack_ptr[-1] = c;
1636     test_else:
1637         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1638             file->ifndef_macro = 0;
1639     test_skip:
1640         if (!(c & 1)) {
1641         skip:
1642             preprocess_skip();
1643             is_bof = 0;
1644             goto redo;
1645         }
1646         break;
1647     case TOK_ENDIF:
1648         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1649             tcc_error("#endif without matching #if");
1650         s1->ifdef_stack_ptr--;
1651         /* '#ifndef macro' was at the start of file. Now we check if
1652            an '#endif' is exactly at the end of file */
1653         if (file->ifndef_macro &&
1654             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1655             file->ifndef_macro_saved = file->ifndef_macro;
1656             /* need to set to zero to avoid false matches if another
1657                #ifndef at middle of file */
1658             file->ifndef_macro = 0;
1659             while (tok != TOK_LINEFEED)
1660                 next_nomacro();
1661             tok_flags |= TOK_FLAG_ENDIF;
1662             goto the_end;
1663         }
1664         break;
1665     case TOK_LINE:
1666         next();
1667         if (tok != TOK_CINT)
1668             tcc_error("A #line format is wrong");
1669     case TOK_PPNUM:
1670         if (tok != TOK_CINT) {
1671             char *p = tokc.cstr->data;
1672             tokc.i = strtoul(p, (char **)&p, 10);
1673         }
1674         i = file->line_num;
1675         file->line_num = tokc.i - 1;
1676         next();
1677         if (tok != TOK_LINEFEED) {
1678             if (tok != TOK_STR) {
1679                 if ((parse_flags & PARSE_FLAG_ASM_COMMENTS) == 0) {
1680                     file->line_num = i;
1681                     tcc_error("#line format is wrong");
1682                 }
1683                 break;
1684             }
1685             pstrcpy(file->filename, sizeof(file->filename),
1686                     (char *)tokc.cstr->data);
1687         }
1688         if (s1->do_debug)
1689             put_stabs(file->filename, N_BINCL, 0, 0, 0);
1690         break;
1691     case TOK_ERROR:
1692     case TOK_WARNING:
1693         c = tok;
1694         ch = file->buf_ptr[0];
1695         skip_spaces();
1696         q = buf;
1697         while (ch != '\n' && ch != CH_EOF) {
1698             if ((q - buf) < sizeof(buf) - 1)
1699                 *q++ = ch;
1700             if (ch == '\\') {
1701                 if (handle_stray_noerror() == 0)
1702                     --q;
1703             } else
1704                 inp();
1705         }
1706         *q = '\0';
1707         if (c == TOK_ERROR)
1708             tcc_error("#error %s", buf);
1709         else
1710             tcc_warning("#warning %s", buf);
1711         break;
1712     case TOK_PRAGMA:
1713         pragma_parse(s1);
1714         break;
1715     default:
1716         if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_PPNUM) {
1717             /* '!' is ignored to allow C scripts. numbers are ignored
1718                to emulate cpp behaviour */
1719         } else {
1720             if (!(parse_flags & PARSE_FLAG_ASM_COMMENTS))
1721                 tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1722             else {
1723                 /* this is a gas line comment in an 'S' file. */
1724                 file->buf_ptr = parse_line_comment(file->buf_ptr);
1725                 goto the_end;
1726             }
1727         }
1728         break;
1729     }
1730     /* ignore other preprocess commands or #! for C scripts */
1731     while (tok != TOK_LINEFEED)
1732         next_nomacro();
1733  the_end:
1734     parse_flags = saved_parse_flags;
1735 }
1736
1737 /* evaluate escape codes in a string. */
1738 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1739 {
1740     int c, n;
1741     const uint8_t *p;
1742
1743     p = buf;
1744     for(;;) {
1745         c = *p;
1746         if (c == '\0')
1747             break;
1748         if (c == '\\') {
1749             p++;
1750             /* escape */
1751             c = *p;
1752             switch(c) {
1753             case '0': case '1': case '2': case '3':
1754             case '4': case '5': case '6': case '7':
1755                 /* at most three octal digits */
1756                 n = c - '0';
1757                 p++;
1758                 c = *p;
1759                 if (isoct(c)) {
1760                     n = n * 8 + c - '0';
1761                     p++;
1762                     c = *p;
1763                     if (isoct(c)) {
1764                         n = n * 8 + c - '0';
1765                         p++;
1766                     }
1767                 }
1768                 c = n;
1769                 goto add_char_nonext;
1770             case 'x':
1771             case 'u':
1772             case 'U':
1773                 p++;
1774                 n = 0;
1775                 for(;;) {
1776                     c = *p;
1777                     if (c >= 'a' && c <= 'f')
1778                         c = c - 'a' + 10;
1779                     else if (c >= 'A' && c <= 'F')
1780                         c = c - 'A' + 10;
1781                     else if (isnum(c))
1782                         c = c - '0';
1783                     else
1784                         break;
1785                     n = n * 16 + c;
1786                     p++;
1787                 }
1788                 c = n;
1789                 goto add_char_nonext;
1790             case 'a':
1791                 c = '\a';
1792                 break;
1793             case 'b':
1794                 c = '\b';
1795                 break;
1796             case 'f':
1797                 c = '\f';
1798                 break;
1799             case 'n':
1800                 c = '\n';
1801                 break;
1802             case 'r':
1803                 c = '\r';
1804                 break;
1805             case 't':
1806                 c = '\t';
1807                 break;
1808             case 'v':
1809                 c = '\v';
1810                 break;
1811             case 'e':
1812                 if (!gnu_ext)
1813                     goto invalid_escape;
1814                 c = 27;
1815                 break;
1816             case '\'':
1817             case '\"':
1818             case '\\':
1819             case '?':
1820                 break;
1821             default:
1822             invalid_escape:
1823                 if (c >= '!' && c <= '~')
1824                     tcc_warning("unknown escape sequence: \'\\%c\'", c);
1825                 else
1826                     tcc_warning("unknown escape sequence: \'\\x%x\'", c);
1827                 break;
1828             }
1829         }
1830         p++;
1831     add_char_nonext:
1832         if (!is_long)
1833             cstr_ccat(outstr, c);
1834         else
1835             cstr_wccat(outstr, c);
1836     }
1837     /* add a trailing '\0' */
1838     if (!is_long)
1839         cstr_ccat(outstr, '\0');
1840     else
1841         cstr_wccat(outstr, '\0');
1842 }
1843
1844 /* we use 64 bit numbers */
1845 #define BN_SIZE 2
1846
1847 /* bn = (bn << shift) | or_val */
1848 static void bn_lshift(unsigned int *bn, int shift, int or_val)
1849 {
1850     int i;
1851     unsigned int v;
1852     for(i=0;i<BN_SIZE;i++) {
1853         v = bn[i];
1854         bn[i] = (v << shift) | or_val;
1855         or_val = v >> (32 - shift);
1856     }
1857 }
1858
1859 static void bn_zero(unsigned int *bn)
1860 {
1861     int i;
1862     for(i=0;i<BN_SIZE;i++) {
1863         bn[i] = 0;
1864     }
1865 }
1866
1867 /* parse number in null terminated string 'p' and return it in the
1868    current token */
1869 static void parse_number(const char *p)
1870 {
1871     int b, t, shift, frac_bits, s, exp_val, ch;
1872     char *q;
1873     unsigned int bn[BN_SIZE];
1874     double d;
1875
1876     /* number */
1877     q = token_buf;
1878     ch = *p++;
1879     t = ch;
1880     ch = *p++;
1881     *q++ = t;
1882     b = 10;
1883     if (t == '.') {
1884         goto float_frac_parse;
1885     } else if (t == '0') {
1886         if (ch == 'x' || ch == 'X') {
1887             q--;
1888             ch = *p++;
1889             b = 16;
1890         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
1891             q--;
1892             ch = *p++;
1893             b = 2;
1894         }
1895     }
1896     /* parse all digits. cannot check octal numbers at this stage
1897        because of floating point constants */
1898     while (1) {
1899         if (ch >= 'a' && ch <= 'f')
1900             t = ch - 'a' + 10;
1901         else if (ch >= 'A' && ch <= 'F')
1902             t = ch - 'A' + 10;
1903         else if (isnum(ch))
1904             t = ch - '0';
1905         else
1906             break;
1907         if (t >= b)
1908             break;
1909         if (q >= token_buf + STRING_MAX_SIZE) {
1910         num_too_long:
1911             tcc_error("number too long");
1912         }
1913         *q++ = ch;
1914         ch = *p++;
1915     }
1916     if (ch == '.' ||
1917         ((ch == 'e' || ch == 'E') && b == 10) ||
1918         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
1919         if (b != 10) {
1920             /* NOTE: strtox should support that for hexa numbers, but
1921                non ISOC99 libcs do not support it, so we prefer to do
1922                it by hand */
1923             /* hexadecimal or binary floats */
1924             /* XXX: handle overflows */
1925             *q = '\0';
1926             if (b == 16)
1927                 shift = 4;
1928             else
1929                 shift = 1;
1930             bn_zero(bn);
1931             q = token_buf;
1932             while (1) {
1933                 t = *q++;
1934                 if (t == '\0') {
1935                     break;
1936                 } else if (t >= 'a') {
1937                     t = t - 'a' + 10;
1938                 } else if (t >= 'A') {
1939                     t = t - 'A' + 10;
1940                 } else {
1941                     t = t - '0';
1942                 }
1943                 bn_lshift(bn, shift, t);
1944             }
1945             frac_bits = 0;
1946             if (ch == '.') {
1947                 ch = *p++;
1948                 while (1) {
1949                     t = ch;
1950                     if (t >= 'a' && t <= 'f') {
1951                         t = t - 'a' + 10;
1952                     } else if (t >= 'A' && t <= 'F') {
1953                         t = t - 'A' + 10;
1954                     } else if (t >= '0' && t <= '9') {
1955                         t = t - '0';
1956                     } else {
1957                         break;
1958                     }
1959                     if (t >= b)
1960                         tcc_error("invalid digit");
1961                     bn_lshift(bn, shift, t);
1962                     frac_bits += shift;
1963                     ch = *p++;
1964                 }
1965             }
1966             if (ch != 'p' && ch != 'P')
1967                 expect("exponent");
1968             ch = *p++;
1969             s = 1;
1970             exp_val = 0;
1971             if (ch == '+') {
1972                 ch = *p++;
1973             } else if (ch == '-') {
1974                 s = -1;
1975                 ch = *p++;
1976             }
1977             if (ch < '0' || ch > '9')
1978                 expect("exponent digits");
1979             while (ch >= '0' && ch <= '9') {
1980                 exp_val = exp_val * 10 + ch - '0';
1981                 ch = *p++;
1982             }
1983             exp_val = exp_val * s;
1984
1985             /* now we can generate the number */
1986             /* XXX: should patch directly float number */
1987             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
1988             d = ldexp(d, exp_val - frac_bits);
1989             t = toup(ch);
1990             if (t == 'F') {
1991                 ch = *p++;
1992                 tok = TOK_CFLOAT;
1993                 /* float : should handle overflow */
1994                 tokc.f = (float)d;
1995             } else if (t == 'L') {
1996                 ch = *p++;
1997 #ifdef TCC_TARGET_PE
1998                 tok = TOK_CDOUBLE;
1999                 tokc.d = d;
2000 #else
2001                 tok = TOK_CLDOUBLE;
2002                 /* XXX: not large enough */
2003                 tokc.ld = (long double)d;
2004 #endif
2005             } else {
2006                 tok = TOK_CDOUBLE;
2007                 tokc.d = d;
2008             }
2009         } else {
2010             /* decimal floats */
2011             if (ch == '.') {
2012                 if (q >= token_buf + STRING_MAX_SIZE)
2013                     goto num_too_long;
2014                 *q++ = ch;
2015                 ch = *p++;
2016             float_frac_parse:
2017                 while (ch >= '0' && ch <= '9') {
2018                     if (q >= token_buf + STRING_MAX_SIZE)
2019                         goto num_too_long;
2020                     *q++ = ch;
2021                     ch = *p++;
2022                 }
2023             }
2024             if (ch == 'e' || ch == 'E') {
2025                 if (q >= token_buf + STRING_MAX_SIZE)
2026                     goto num_too_long;
2027                 *q++ = ch;
2028                 ch = *p++;
2029                 if (ch == '-' || ch == '+') {
2030                     if (q >= token_buf + STRING_MAX_SIZE)
2031                         goto num_too_long;
2032                     *q++ = ch;
2033                     ch = *p++;
2034                 }
2035                 if (ch < '0' || ch > '9')
2036                     expect("exponent digits");
2037                 while (ch >= '0' && ch <= '9') {
2038                     if (q >= token_buf + STRING_MAX_SIZE)
2039                         goto num_too_long;
2040                     *q++ = ch;
2041                     ch = *p++;
2042                 }
2043             }
2044             *q = '\0';
2045             t = toup(ch);
2046             errno = 0;
2047             if (t == 'F') {
2048                 ch = *p++;
2049                 tok = TOK_CFLOAT;
2050                 tokc.f = strtof(token_buf, NULL);
2051             } else if (t == 'L') {
2052                 ch = *p++;
2053 #ifdef TCC_TARGET_PE
2054                 tok = TOK_CDOUBLE;
2055                 tokc.d = strtod(token_buf, NULL);
2056 #else
2057                 tok = TOK_CLDOUBLE;
2058                 tokc.ld = strtold(token_buf, NULL);
2059 #endif
2060             } else {
2061                 tok = TOK_CDOUBLE;
2062                 tokc.d = strtod(token_buf, NULL);
2063             }
2064         }
2065     } else {
2066         unsigned long long n, n1;
2067         int lcount, ucount, must_64bit;
2068         const char *p1;
2069
2070         /* integer number */
2071         *q = '\0';
2072         q = token_buf;
2073         if (b == 10 && *q == '0') {
2074             b = 8;
2075             q++;
2076         }
2077         n = 0;
2078         while(1) {
2079             t = *q++;
2080             /* no need for checks except for base 10 / 8 errors */
2081             if (t == '\0')
2082                 break;
2083             else if (t >= 'a')
2084                 t = t - 'a' + 10;
2085             else if (t >= 'A')
2086                 t = t - 'A' + 10;
2087             else
2088                 t = t - '0';
2089             if (t >= b)
2090                 tcc_error("invalid digit");
2091             n1 = n;
2092             n = n * b + t;
2093             /* detect overflow */
2094             /* XXX: this test is not reliable */
2095             if (n < n1)
2096                 tcc_error("integer constant overflow");
2097         }
2098
2099         /* Determine the characteristics (unsigned and/or 64bit) the type of
2100            the constant must have according to the constant suffix(es) */
2101         lcount = ucount = must_64bit = 0;
2102         p1 = p;
2103         for(;;) {
2104             t = toup(ch);
2105             if (t == 'L') {
2106                 if (lcount >= 2)
2107                     tcc_error("three 'l's in integer constant");
2108                 if (lcount && *(p - 1) != ch)
2109                     tcc_error("incorrect integer suffix: %s", p1);
2110                 lcount++;
2111 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
2112                 if (lcount == 2)
2113 #endif
2114                     must_64bit = 1;
2115                 ch = *p++;
2116             } else if (t == 'U') {
2117                 if (ucount >= 1)
2118                     tcc_error("two 'u's in integer constant");
2119                 ucount++;
2120                 ch = *p++;
2121             } else {
2122                 break;
2123             }
2124         }
2125
2126         /* Whether 64 bits are needed to hold the constant's value */
2127         if (n & 0xffffffff00000000LL || must_64bit) {
2128             tok = TOK_CLLONG;
2129             n1 = n >> 32;
2130         } else {
2131             tok = TOK_CINT;
2132             n1 = n;
2133         }
2134
2135         /* Whether type must be unsigned to hold the constant's value */
2136         if (ucount || ((n1 >> 31) && (b != 10))) {
2137             if (tok == TOK_CLLONG)
2138                 tok = TOK_CULLONG;
2139             else
2140                 tok = TOK_CUINT;
2141         /* If decimal and no unsigned suffix, bump to 64 bits or throw error */
2142         } else if (n1 >> 31) {
2143             if (tok == TOK_CINT)
2144                 tok = TOK_CLLONG;
2145             else
2146                 tcc_error("integer constant overflow");
2147         }
2148
2149         if (tok == TOK_CINT || tok == TOK_CUINT)
2150             tokc.ui = n;
2151         else
2152             tokc.ull = n;
2153     }
2154     if (ch)
2155         tcc_error("invalid number\n");
2156 }
2157
2158
2159 #define PARSE2(c1, tok1, c2, tok2)              \
2160     case c1:                                    \
2161         PEEKC(c, p);                            \
2162         if (c == c2) {                          \
2163             p++;                                \
2164             tok = tok2;                         \
2165         } else {                                \
2166             tok = tok1;                         \
2167         }                                       \
2168         break;
2169
2170 /* return next token without macro substitution */
2171 static inline void next_nomacro1(void)
2172 {
2173     int t, c, is_long;
2174     TokenSym *ts;
2175     uint8_t *p, *p1;
2176     unsigned int h;
2177
2178     p = file->buf_ptr;
2179  redo_no_start:
2180     c = *p;
2181     switch(c) {
2182     case ' ':
2183     case '\t':
2184         tok = c;
2185         p++;
2186         goto keep_tok_flags;
2187     case '\f':
2188     case '\v':
2189     case '\r':
2190         p++;
2191         goto redo_no_start;
2192     case '\\':
2193         /* first look if it is in fact an end of buffer */
2194         if (p >= file->buf_end) {
2195             file->buf_ptr = p;
2196             handle_eob();
2197             p = file->buf_ptr;
2198             if (p >= file->buf_end)
2199                 goto parse_eof;
2200             else
2201                 goto redo_no_start;
2202         } else {
2203             file->buf_ptr = p;
2204             ch = *p;
2205             handle_stray();
2206             p = file->buf_ptr;
2207             goto redo_no_start;
2208         }
2209     parse_eof:
2210         {
2211             TCCState *s1 = tcc_state;
2212             if ((parse_flags & PARSE_FLAG_LINEFEED)
2213                 && !(tok_flags & TOK_FLAG_EOF)) {
2214                 tok_flags |= TOK_FLAG_EOF;
2215                 tok = TOK_LINEFEED;
2216                 goto keep_tok_flags;
2217             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2218                 tok = TOK_EOF;
2219             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2220                 tcc_error("missing #endif");
2221             } else if (s1->include_stack_ptr == s1->include_stack) {
2222                 /* no include left : end of file. */
2223                 tok = TOK_EOF;
2224             } else {
2225                 tok_flags &= ~TOK_FLAG_EOF;
2226                 /* pop include file */
2227
2228                 /* test if previous '#endif' was after a #ifdef at
2229                    start of file */
2230                 if (tok_flags & TOK_FLAG_ENDIF) {
2231 #ifdef INC_DEBUG
2232                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2233 #endif
2234                     add_cached_include(s1, file->filename, file->ifndef_macro_saved);
2235                     tok_flags &= ~TOK_FLAG_ENDIF;
2236                 }
2237
2238                 /* add end of include file debug info */
2239                 if (tcc_state->do_debug) {
2240                     put_stabd(N_EINCL, 0, 0);
2241                 }
2242                 /* pop include stack */
2243                 tcc_close();
2244                 s1->include_stack_ptr--;
2245                 p = file->buf_ptr;
2246                 goto redo_no_start;
2247             }
2248         }
2249         break;
2250
2251     case '\n':
2252         file->line_num++;
2253         tok_flags |= TOK_FLAG_BOL;
2254         p++;
2255 maybe_newline:
2256         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2257             goto redo_no_start;
2258         tok = TOK_LINEFEED;
2259         goto keep_tok_flags;
2260
2261     case '#':
2262         /* XXX: simplify */
2263         PEEKC(c, p);
2264         if (is_space(c) && (parse_flags & PARSE_FLAG_ASM_FILE)) {
2265             p = parse_line_comment(p);
2266             goto redo_no_start;
2267         }
2268         else
2269         if ((tok_flags & TOK_FLAG_BOL) &&
2270             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2271             file->buf_ptr = p;
2272             preprocess(tok_flags & TOK_FLAG_BOF);
2273             p = file->buf_ptr;
2274             goto maybe_newline;
2275         } else {
2276             if (c == '#') {
2277                 p++;
2278                 tok = TOK_TWOSHARPS;
2279             } else {
2280                 if (parse_flags & PARSE_FLAG_ASM_COMMENTS) {
2281                     p = parse_line_comment(p - 1);
2282                     goto redo_no_start;
2283                 } else {
2284                     tok = '#';
2285                 }
2286             }
2287         }
2288         break;
2289
2290     /* treat $ as allowed char in indentifier  */
2291     case '$': if (!tcc_state->dollars_in_identifiers) goto parse_simple;
2292
2293     case 'a': case 'b': case 'c': case 'd':
2294     case 'e': case 'f': case 'g': case 'h':
2295     case 'i': case 'j': case 'k': case 'l':
2296     case 'm': case 'n': case 'o': case 'p':
2297     case 'q': case 'r': case 's': case 't':
2298     case 'u': case 'v': case 'w': case 'x':
2299     case 'y': case 'z':
2300     case 'A': case 'B': case 'C': case 'D':
2301     case 'E': case 'F': case 'G': case 'H':
2302     case 'I': case 'J': case 'K':
2303     case 'M': case 'N': case 'O': case 'P':
2304     case 'Q': case 'R': case 'S': case 'T':
2305     case 'U': case 'V': case 'W': case 'X':
2306     case 'Y': case 'Z':
2307     case '_':
2308     parse_ident_fast:
2309         p1 = p;
2310         h = TOK_HASH_INIT;
2311         h = TOK_HASH_FUNC(h, c);
2312         p++;
2313         for(;;) {
2314             c = *p;
2315             if (!isidnum_table[c-CH_EOF])
2316                 break;
2317             h = TOK_HASH_FUNC(h, c);
2318             p++;
2319         }
2320         if (c != '\\') {
2321             TokenSym **pts;
2322             int len;
2323
2324             /* fast case : no stray found, so we have the full token
2325                and we have already hashed it */
2326             len = p - p1;
2327             h &= (TOK_HASH_SIZE - 1);
2328             pts = &hash_ident[h];
2329             for(;;) {
2330                 ts = *pts;
2331                 if (!ts)
2332                     break;
2333                 if (ts->len == len && !memcmp(ts->str, p1, len))
2334                     goto token_found;
2335                 pts = &(ts->hash_next);
2336             }
2337             ts = tok_alloc_new(pts, (char *) p1, len);
2338         token_found: ;
2339         } else {
2340             /* slower case */
2341             cstr_reset(&tokcstr);
2342
2343             while (p1 < p) {
2344                 cstr_ccat(&tokcstr, *p1);
2345                 p1++;
2346             }
2347             p--;
2348             PEEKC(c, p);
2349         parse_ident_slow:
2350             while (isidnum_table[c-CH_EOF]) {
2351                 cstr_ccat(&tokcstr, c);
2352                 PEEKC(c, p);
2353             }
2354             ts = tok_alloc(tokcstr.data, tokcstr.size);
2355         }
2356         tok = ts->tok;
2357         break;
2358     case 'L':
2359         t = p[1];
2360         if (t != '\\' && t != '\'' && t != '\"') {
2361             /* fast case */
2362             goto parse_ident_fast;
2363         } else {
2364             PEEKC(c, p);
2365             if (c == '\'' || c == '\"') {
2366                 is_long = 1;
2367                 goto str_const;
2368             } else {
2369                 cstr_reset(&tokcstr);
2370                 cstr_ccat(&tokcstr, 'L');
2371                 goto parse_ident_slow;
2372             }
2373         }
2374         break;
2375     case '0': case '1': case '2': case '3':
2376     case '4': case '5': case '6': case '7':
2377     case '8': case '9':
2378
2379         cstr_reset(&tokcstr);
2380         /* after the first digit, accept digits, alpha, '.' or sign if
2381            prefixed by 'eEpP' */
2382     parse_num:
2383         for(;;) {
2384             t = c;
2385             cstr_ccat(&tokcstr, c);
2386             PEEKC(c, p);
2387             if (!(isnum(c) || isid(c) || c == '.' ||
2388                   ((c == '+' || c == '-') &&
2389                    (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2390                 break;
2391         }
2392         /* We add a trailing '\0' to ease parsing */
2393         cstr_ccat(&tokcstr, '\0');
2394         tokc.cstr = &tokcstr;
2395         tok = TOK_PPNUM;
2396         break;
2397     case '.':
2398         /* special dot handling because it can also start a number */
2399         PEEKC(c, p);
2400         if (isnum(c)) {
2401             cstr_reset(&tokcstr);
2402             cstr_ccat(&tokcstr, '.');
2403             goto parse_num;
2404         } else if (c == '.') {
2405             PEEKC(c, p);
2406             if (c != '.') {
2407                 if ((parse_flags & PARSE_FLAG_ASM_COMMENTS) == 0)
2408                     expect("'.'");
2409                 tok = '.';
2410                 break;
2411             }
2412             PEEKC(c, p);
2413             tok = TOK_DOTS;
2414         } else {
2415             tok = '.';
2416         }
2417         break;
2418     case '\'':
2419     case '\"':
2420         is_long = 0;
2421     str_const:
2422         {
2423             CString str;
2424             int sep;
2425
2426             sep = c;
2427
2428             /* parse the string */
2429             cstr_new(&str);
2430             p = parse_pp_string(p, sep, &str);
2431             cstr_ccat(&str, '\0');
2432
2433             /* eval the escape (should be done as TOK_PPNUM) */
2434             cstr_reset(&tokcstr);
2435             parse_escape_string(&tokcstr, str.data, is_long);
2436             cstr_free(&str);
2437
2438             if (sep == '\'') {
2439                 int char_size;
2440                 /* XXX: make it portable */
2441                 if (!is_long)
2442                     char_size = 1;
2443                 else
2444                     char_size = sizeof(nwchar_t);
2445                 if (tokcstr.size <= char_size)
2446                     tcc_error("empty character constant");
2447                 if (tokcstr.size > 2 * char_size)
2448                     tcc_warning("multi-character character constant");
2449                 if (!is_long) {
2450                     tokc.i = *(int8_t *)tokcstr.data;
2451                     tok = TOK_CCHAR;
2452                 } else {
2453                     tokc.i = *(nwchar_t *)tokcstr.data;
2454                     tok = TOK_LCHAR;
2455                 }
2456             } else {
2457                 tokc.cstr = &tokcstr;
2458                 if (!is_long)
2459                     tok = TOK_STR;
2460                 else
2461                     tok = TOK_LSTR;
2462             }
2463         }
2464         break;
2465
2466     case '<':
2467         PEEKC(c, p);
2468         if (c == '=') {
2469             p++;
2470             tok = TOK_LE;
2471         } else if (c == '<') {
2472             PEEKC(c, p);
2473             if (c == '=') {
2474                 p++;
2475                 tok = TOK_A_SHL;
2476             } else {
2477                 tok = TOK_SHL;
2478             }
2479         } else {
2480             tok = TOK_LT;
2481         }
2482         break;
2483
2484     case '>':
2485         PEEKC(c, p);
2486         if (c == '=') {
2487             p++;
2488             tok = TOK_GE;
2489         } else if (c == '>') {
2490             PEEKC(c, p);
2491             if (c == '=') {
2492                 p++;
2493                 tok = TOK_A_SAR;
2494             } else {
2495                 tok = TOK_SAR;
2496             }
2497         } else {
2498             tok = TOK_GT;
2499         }
2500         break;
2501
2502     case '&':
2503         PEEKC(c, p);
2504         if (c == '&') {
2505             p++;
2506             tok = TOK_LAND;
2507         } else if (c == '=') {
2508             p++;
2509             tok = TOK_A_AND;
2510         } else {
2511             tok = '&';
2512         }
2513         break;
2514
2515     case '|':
2516         PEEKC(c, p);
2517         if (c == '|') {
2518             p++;
2519             tok = TOK_LOR;
2520         } else if (c == '=') {
2521             p++;
2522             tok = TOK_A_OR;
2523         } else {
2524             tok = '|';
2525         }
2526         break;
2527
2528     case '+':
2529         PEEKC(c, p);
2530         if (c == '+') {
2531             p++;
2532             tok = TOK_INC;
2533         } else if (c == '=') {
2534             p++;
2535             tok = TOK_A_ADD;
2536         } else {
2537             tok = '+';
2538         }
2539         break;
2540
2541     case '-':
2542         PEEKC(c, p);
2543         if (c == '-') {
2544             p++;
2545             tok = TOK_DEC;
2546         } else if (c == '=') {
2547             p++;
2548             tok = TOK_A_SUB;
2549         } else if (c == '>') {
2550             p++;
2551             tok = TOK_ARROW;
2552         } else {
2553             tok = '-';
2554         }
2555         break;
2556
2557     PARSE2('!', '!', '=', TOK_NE)
2558     PARSE2('=', '=', '=', TOK_EQ)
2559     PARSE2('*', '*', '=', TOK_A_MUL)
2560     PARSE2('%', '%', '=', TOK_A_MOD)
2561     PARSE2('^', '^', '=', TOK_A_XOR)
2562
2563         /* comments or operator */
2564     case '/':
2565         PEEKC(c, p);
2566         if (c == '*') {
2567             p = parse_comment(p);
2568             /* comments replaced by a blank */
2569             tok = ' ';
2570             goto keep_tok_flags;
2571         } else if (c == '/') {
2572             p = parse_line_comment(p);
2573             tok = ' ';
2574             goto keep_tok_flags;
2575         } else if (c == '=') {
2576             p++;
2577             tok = TOK_A_DIV;
2578         } else {
2579             tok = '/';
2580         }
2581         break;
2582
2583         /* simple tokens */
2584     case '(':
2585     case ')':
2586     case '[':
2587     case ']':
2588     case '{':
2589     case '}':
2590     case ',':
2591     case ';':
2592     case ':':
2593     case '?':
2594     case '~':
2595     case '@': /* only used in assembler */
2596     parse_simple:
2597         tok = c;
2598         p++;
2599         break;
2600     default:
2601         if ((parse_flags & PARSE_FLAG_ASM_FILE) == 0)
2602             tcc_error("unrecognized character \\x%02x", c);
2603         else {
2604             tok = ' ';
2605             p++;
2606         }
2607         break;
2608     }
2609     tok_flags = 0;
2610 keep_tok_flags:
2611     file->buf_ptr = p;
2612 #if defined(PARSE_DEBUG)
2613     printf("token = %s\n", get_tok_str(tok, &tokc));
2614 #endif
2615 }
2616
2617 /* return next token without macro substitution. Can read input from
2618    macro_ptr buffer */
2619 static void next_nomacro_spc(void)
2620 {
2621     if (macro_ptr) {
2622     redo:
2623         tok = *macro_ptr;
2624         if (tok) {
2625             TOK_GET(&tok, &macro_ptr, &tokc);
2626             if (tok == TOK_LINENUM) {
2627                 file->line_num = tokc.i;
2628                 goto redo;
2629             }
2630         }
2631     } else {
2632         next_nomacro1();
2633     }
2634 }
2635
2636 ST_FUNC void next_nomacro(void)
2637 {
2638     do {
2639         next_nomacro_spc();
2640     } while (is_space(tok));
2641 }
2642
2643 /* substitute arguments in replacement lists in macro_str by the values in
2644    args (field d) and return allocated string */
2645 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2646 {
2647     int last_tok, t, spc;
2648     const int *st;
2649     Sym *s;
2650     CValue cval;
2651     TokenString str;
2652     CString cstr;
2653
2654     tok_str_new(&str);
2655     last_tok = 0;
2656     while(1) {
2657         TOK_GET(&t, &macro_str, &cval);
2658         if (!t)
2659             break;
2660         if (t == '#') {
2661             /* stringize */
2662             TOK_GET(&t, &macro_str, &cval);
2663             if (!t)
2664                 break;
2665             s = sym_find2(args, t);
2666             if (s) {
2667                 cstr_new(&cstr);
2668                 st = s->d;
2669                 spc = 0;
2670                 while (*st) {
2671                     TOK_GET(&t, &st, &cval);
2672                     if (!check_space(t, &spc))
2673                         cstr_cat(&cstr, get_tok_str(t, &cval));
2674                 }
2675                 cstr.size -= spc;
2676                 cstr_ccat(&cstr, '\0');
2677 #ifdef PP_DEBUG
2678                 printf("stringize: %s\n", (char *)cstr.data);
2679 #endif
2680                 /* add string */
2681                 cval.cstr = &cstr;
2682                 tok_str_add2(&str, TOK_STR, &cval);
2683                 cstr_free(&cstr);
2684             } else {
2685                 tok_str_add2(&str, t, &cval);
2686             }
2687         } else if (t >= TOK_IDENT) {
2688             s = sym_find2(args, t);
2689             if (s) {
2690                 st = s->d;
2691                 /* if '##' is present before or after, no arg substitution */
2692                 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2693                     /* special case for var arg macros : ## eats the
2694                        ',' if empty VA_ARGS variable. */
2695                     /* XXX: test of the ',' is not 100%
2696                        reliable. should fix it to avoid security
2697                        problems */
2698                     if (gnu_ext && s->type.t &&
2699                         last_tok == TOK_TWOSHARPS &&
2700                         str.len >= 2 && str.str[str.len - 2] == ',') {
2701                         if (*st == TOK_PLCHLDR) {
2702                             /* suppress ',' '##' */
2703                             str.len -= 2;
2704                         } else {
2705                             /* suppress '##' and add variable */
2706                             str.len--;
2707                             goto add_var;
2708                         }
2709                     } else {
2710                         int t1;
2711                     add_var:
2712                         for(;;) {
2713                             TOK_GET(&t1, &st, &cval);
2714                             if (!t1)
2715                                 break;
2716                             tok_str_add2(&str, t1, &cval);
2717                         }
2718                     }
2719                 } else if (*st != TOK_PLCHLDR) {
2720                     /* NOTE: the stream cannot be read when macro
2721                        substituing an argument */
2722                     macro_subst(&str, nested_list, st, NULL);
2723                 }
2724             } else {
2725                 tok_str_add(&str, t);
2726             }
2727         } else {
2728             tok_str_add2(&str, t, &cval);
2729         }
2730         last_tok = t;
2731     }
2732     tok_str_add(&str, 0);
2733     return str.str;
2734 }
2735
2736 static char const ab_month_name[12][4] =
2737 {
2738     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2739     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2740 };
2741
2742 /* do macro substitution of current token with macro 's' and add
2743    result to (tok_str,tok_len). 'nested_list' is the list of all
2744    macros we got inside to avoid recursing. Return non zero if no
2745    substitution needs to be done */
2746 static int macro_subst_tok(TokenString *tok_str,
2747                            Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2748 {
2749     Sym *args, *sa, *sa1;
2750     int mstr_allocated, parlevel, *mstr, t, t1, spc;
2751     const int *p;
2752     TokenString str;
2753     char *cstrval;
2754     CValue cval;
2755     CString cstr;
2756     char buf[32];
2757
2758     /* if symbol is a macro, prepare substitution */
2759     /* special macros */
2760     if (tok == TOK___LINE__) {
2761         snprintf(buf, sizeof(buf), "%d", file->line_num);
2762         cstrval = buf;
2763         t1 = TOK_PPNUM;
2764         goto add_cstr1;
2765     } else if (tok == TOK___FILE__) {
2766         cstrval = file->filename;
2767         goto add_cstr;
2768     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2769         time_t ti;
2770         struct tm *tm;
2771
2772         time(&ti);
2773         tm = localtime(&ti);
2774         if (tok == TOK___DATE__) {
2775             snprintf(buf, sizeof(buf), "%s %2d %d",
2776                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2777         } else {
2778             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2779                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2780         }
2781         cstrval = buf;
2782     add_cstr:
2783         t1 = TOK_STR;
2784     add_cstr1:
2785         cstr_new(&cstr);
2786         cstr_cat(&cstr, cstrval);
2787         cstr_ccat(&cstr, '\0');
2788         cval.cstr = &cstr;
2789         tok_str_add2(tok_str, t1, &cval);
2790         cstr_free(&cstr);
2791     } else {
2792         mstr = s->d;
2793         mstr_allocated = 0;
2794         if (s->type.t == MACRO_FUNC) {
2795             /* NOTE: we do not use next_nomacro to avoid eating the
2796                next token. XXX: find better solution */
2797         redo:
2798             if (macro_ptr) {
2799                 p = macro_ptr;
2800                 while (is_space(t = *p) || TOK_LINEFEED == t)
2801                     ++p;
2802                 if (t == 0 && can_read_stream) {
2803                     /* end of macro stream: we must look at the token
2804                        after in the file */
2805                     struct macro_level *ml = *can_read_stream;
2806                     macro_ptr = NULL;
2807                     if (ml)
2808                     {
2809                         macro_ptr = ml->p;
2810                         ml->p = NULL;
2811                         *can_read_stream = ml -> prev;
2812                     }
2813                     /* also, end of scope for nested defined symbol */
2814                     (*nested_list)->v = -1;
2815                     goto redo;
2816                 }
2817             } else {
2818                 ch = file->buf_ptr[0];
2819                 while (is_space(ch) || ch == '\n' || ch == '/')
2820                   {
2821                     if (ch == '/')
2822                       {
2823                         int c;
2824                         uint8_t *p = file->buf_ptr;
2825                         PEEKC(c, p);
2826                         if (c == '*') {
2827                             p = parse_comment(p);
2828                             file->buf_ptr = p - 1;
2829                         } else if (c == '/') {
2830                             p = parse_line_comment(p);
2831                             file->buf_ptr = p - 1;
2832                         } else
2833                           break;
2834                       }
2835                     cinp();
2836                   }
2837                 t = ch;
2838             }
2839             if (t != '(') /* no macro subst */
2840                 return -1;
2841
2842             /* argument macro */
2843             next_nomacro();
2844             next_nomacro();
2845             args = NULL;
2846             sa = s->next;
2847             /* NOTE: empty args are allowed, except if no args */
2848             for(;;) {
2849                 /* handle '()' case */
2850                 if (!args && !sa && tok == ')')
2851                     break;
2852                 if (!sa)
2853                     tcc_error("macro '%s' used with too many args",
2854                           get_tok_str(s->v, 0));
2855                 tok_str_new(&str);
2856                 parlevel = spc = 0;
2857                 /* NOTE: non zero sa->t indicates VA_ARGS */
2858                 while ((parlevel > 0 ||
2859                         (tok != ')' &&
2860                          (tok != ',' || sa->type.t))) &&
2861                        tok != -1) {
2862                     if (tok == '(')
2863                         parlevel++;
2864                     else if (tok == ')')
2865                         parlevel--;
2866                     if (tok == TOK_LINEFEED)
2867                         tok = ' ';
2868                     if (!check_space(tok, &spc))
2869                         tok_str_add2(&str, tok, &tokc);
2870                     next_nomacro_spc();
2871                 }
2872                 if (!str.len)
2873                     tok_str_add(&str, TOK_PLCHLDR);
2874                 str.len -= spc;
2875                 tok_str_add(&str, 0);
2876                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
2877                 sa1->d = str.str;
2878                 sa = sa->next;
2879                 if (tok == ')') {
2880                     /* special case for gcc var args: add an empty
2881                        var arg argument if it is omitted */
2882                     if (sa && sa->type.t && gnu_ext)
2883                         continue;
2884                     else
2885                         break;
2886                 }
2887                 if (tok != ',')
2888                     expect(",");
2889                 next_nomacro();
2890             }
2891             if (sa) {
2892                 tcc_error("macro '%s' used with too few args",
2893                       get_tok_str(s->v, 0));
2894             }
2895
2896             /* now subst each arg */
2897             mstr = macro_arg_subst(nested_list, mstr, args);
2898             /* free memory */
2899             sa = args;
2900             while (sa) {
2901                 sa1 = sa->prev;
2902                 tok_str_free(sa->d);
2903                 sym_free(sa);
2904                 sa = sa1;
2905             }
2906             mstr_allocated = 1;
2907         }
2908         sym_push2(nested_list, s->v, 0, 0);
2909         macro_subst(tok_str, nested_list, mstr, can_read_stream);
2910         /* pop nested defined symbol */
2911         sa1 = *nested_list;
2912         *nested_list = sa1->prev;
2913         sym_free(sa1);
2914         if (mstr_allocated)
2915             tok_str_free(mstr);
2916     }
2917     return 0;
2918 }
2919
2920 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2921    return the resulting string (which must be freed). */
2922 static inline int *macro_twosharps(const int *macro_str)
2923 {
2924     const int *ptr;
2925     int t;
2926     TokenString macro_str1;
2927     CString cstr;
2928     int n, start_of_nosubsts;
2929
2930     /* we search the first '##' */
2931     for(ptr = macro_str;;) {
2932         CValue cval;
2933         TOK_GET(&t, &ptr, &cval);
2934         if (t == TOK_TWOSHARPS)
2935             break;
2936         /* nothing more to do if end of string */
2937         if (t == 0)
2938             return NULL;
2939     }
2940
2941     /* we saw '##', so we need more processing to handle it */
2942     start_of_nosubsts = -1;
2943     tok_str_new(&macro_str1);
2944     for(ptr = macro_str;;) {
2945         TOK_GET(&tok, &ptr, &tokc);
2946         if (tok == 0)
2947             break;
2948         if (tok == TOK_TWOSHARPS)
2949             continue;
2950         if (tok == TOK_NOSUBST && start_of_nosubsts < 0)
2951             start_of_nosubsts = macro_str1.len;
2952         while (*ptr == TOK_TWOSHARPS) {
2953             /* given 'a##b', remove nosubsts preceding 'a' */
2954             if (start_of_nosubsts >= 0)
2955                 macro_str1.len = start_of_nosubsts;
2956             /* given 'a##b', skip '##' */
2957             t = *++ptr;
2958             /* given 'a##b', remove nosubsts preceding 'b' */
2959             while (t == TOK_NOSUBST)
2960                 t = *++ptr;
2961             if (t && t != TOK_TWOSHARPS) {
2962                 CValue cval;
2963                 TOK_GET(&t, &ptr, &cval);
2964                 /* We concatenate the two tokens */
2965                 cstr_new(&cstr);
2966                 if (tok != TOK_PLCHLDR)
2967                     cstr_cat(&cstr, get_tok_str(tok, &tokc));
2968                 n = cstr.size;
2969                 if (t != TOK_PLCHLDR || tok == TOK_PLCHLDR)
2970                     cstr_cat(&cstr, get_tok_str(t, &cval));
2971                 cstr_ccat(&cstr, '\0');
2972
2973                 tcc_open_bf(tcc_state, ":paste:", cstr.size);
2974                 memcpy(file->buffer, cstr.data, cstr.size);
2975                 for (;;) {
2976                     next_nomacro1();
2977                     if (0 == *file->buf_ptr)
2978                         break;
2979                     tok_str_add2(&macro_str1, tok, &tokc);
2980                     tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
2981                         n, cstr.data, (char*)cstr.data + n);
2982                 }
2983                 tcc_close();
2984                 cstr_free(&cstr);
2985             }
2986         }
2987         if (tok != TOK_NOSUBST) {
2988             tok_str_add2(&macro_str1, tok, &tokc);
2989             tok = ' ';
2990             start_of_nosubsts = -1;
2991         }
2992         tok_str_add2(&macro_str1, tok, &tokc);
2993     }
2994     tok_str_add(&macro_str1, 0);
2995     return macro_str1.str;
2996 }
2997
2998
2999 /* do macro substitution of macro_str and add result to
3000    (tok_str,tok_len). 'nested_list' is the list of all macros we got
3001    inside to avoid recursing. */
3002 static void macro_subst(TokenString *tok_str, Sym **nested_list,
3003                         const int *macro_str, struct macro_level ** can_read_stream)
3004 {
3005     Sym *s;
3006     int *macro_str1;
3007     const int *ptr;
3008     int t, ret, spc;
3009     CValue cval;
3010     struct macro_level ml;
3011     int force_blank;
3012
3013     /* first scan for '##' operator handling */
3014     ptr = macro_str;
3015     macro_str1 = macro_twosharps(ptr);
3016
3017     if (macro_str1)
3018         ptr = macro_str1;
3019     spc = 0;
3020     force_blank = 0;
3021
3022     while (1) {
3023         /* NOTE: ptr == NULL can only happen if tokens are read from
3024            file stream due to a macro function call */
3025         if (ptr == NULL)
3026             break;
3027         TOK_GET(&t, &ptr, &cval);
3028         if (t == 0)
3029             break;
3030         if (t == TOK_NOSUBST) {
3031             /* following token has already been subst'd. just copy it on */
3032             tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3033             TOK_GET(&t, &ptr, &cval);
3034             goto no_subst;
3035         }
3036         s = define_find(t);
3037         if (s != NULL) {
3038             /* if nested substitution, do nothing */
3039             if (sym_find2(*nested_list, t)) {
3040                 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
3041                 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3042                 goto no_subst;
3043             }
3044             ml.p = macro_ptr;
3045             if (can_read_stream)
3046                 ml.prev = *can_read_stream, *can_read_stream = &ml;
3047             macro_ptr = (int *)ptr;
3048             tok = t;
3049             ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
3050             ptr = (int *)macro_ptr;
3051             macro_ptr = ml.p;
3052             if (can_read_stream && *can_read_stream == &ml)
3053                 *can_read_stream = ml.prev;
3054             if (ret != 0)
3055                 goto no_subst;
3056             if (parse_flags & PARSE_FLAG_SPACES)
3057                 force_blank = 1;
3058         } else {
3059         no_subst:
3060             if (force_blank) {
3061                 tok_str_add(tok_str, ' ');
3062                 spc = 1;
3063                 force_blank = 0;
3064             }
3065             if (!check_space(t, &spc))
3066                 tok_str_add2(tok_str, t, &cval);
3067         }
3068     }
3069     if (macro_str1)
3070         tok_str_free(macro_str1);
3071 }
3072
3073 /* return next token with macro substitution */
3074 ST_FUNC void next(void)
3075 {
3076     Sym *nested_list, *s;
3077     TokenString str;
3078     struct macro_level *ml;
3079
3080  redo:
3081     if (parse_flags & PARSE_FLAG_SPACES)
3082         next_nomacro_spc();
3083     else
3084         next_nomacro();
3085     if (!macro_ptr) {
3086         /* if not reading from macro substituted string, then try
3087            to substitute macros */
3088         if (tok >= TOK_IDENT &&
3089             (parse_flags & PARSE_FLAG_PREPROCESS)) {
3090             s = define_find(tok);
3091             if (s) {
3092                 /* we have a macro: we try to substitute */
3093                 tok_str_new(&str);
3094                 nested_list = NULL;
3095                 ml = NULL;
3096                 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
3097                     /* substitution done, NOTE: maybe empty */
3098                     tok_str_add(&str, 0);
3099                     macro_ptr = str.str;
3100                     macro_ptr_allocated = str.str;
3101                     goto redo;
3102                 }
3103             }
3104         }
3105     } else {
3106         if (tok == 0) {
3107             /* end of macro or end of unget buffer */
3108             if (unget_buffer_enabled) {
3109                 macro_ptr = unget_saved_macro_ptr;
3110                 unget_buffer_enabled = 0;
3111             } else {
3112                 /* end of macro string: free it */
3113                 tok_str_free(macro_ptr_allocated);
3114                 macro_ptr_allocated = NULL;
3115                 macro_ptr = NULL;
3116             }
3117             goto redo;
3118         } else if (tok == TOK_NOSUBST) {
3119             /* discard preprocessor's nosubst markers */
3120             goto redo;
3121         }
3122     }
3123
3124     /* convert preprocessor tokens into C tokens */
3125     if (tok == TOK_PPNUM &&
3126         (parse_flags & PARSE_FLAG_TOK_NUM)) {
3127         parse_number((char *)tokc.cstr->data);
3128     }
3129 }
3130
3131 /* push back current token and set current token to 'last_tok'. Only
3132    identifier case handled for labels. */
3133 ST_INLN void unget_tok(int last_tok)
3134 {
3135     int i, n;
3136     int *q;
3137     if (unget_buffer_enabled)
3138       {
3139         /* assert(macro_ptr == unget_saved_buffer + 1);
3140            assert(*macro_ptr == 0);  */
3141       }
3142     else
3143       {
3144         unget_saved_macro_ptr = macro_ptr;
3145         unget_buffer_enabled = 1;
3146       }
3147     q = unget_saved_buffer;
3148     macro_ptr = q;
3149     *q++ = tok;
3150     n = tok_ext_size(tok) - 1;
3151     for(i=0;i<n;i++)
3152         *q++ = tokc.tab[i];
3153     *q = 0; /* end of token string */
3154     tok = last_tok;
3155 }
3156
3157
3158 /* better than nothing, but needs extension to handle '-E' option
3159    correctly too */
3160 ST_FUNC void preprocess_init(TCCState *s1)
3161 {
3162     s1->include_stack_ptr = s1->include_stack;
3163     /* XXX: move that before to avoid having to initialize
3164        file->ifdef_stack_ptr ? */
3165     s1->ifdef_stack_ptr = s1->ifdef_stack;
3166     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3167
3168     vtop = vstack - 1;
3169     s1->pack_stack[0] = 0;
3170     s1->pack_stack_ptr = s1->pack_stack;
3171 }
3172
3173 ST_FUNC void preprocess_new(void)
3174 {
3175     int i, c;
3176     const char *p, *r;
3177
3178     /* init isid table */
3179     for(i=CH_EOF;i<256;i++)
3180         isidnum_table[i-CH_EOF] = (isid(i) || isnum(i) ||
3181             (tcc_state->dollars_in_identifiers ? i == '$' : 0));
3182
3183     /* add all tokens */
3184     if (table_ident) {
3185         tcc_free (table_ident);
3186         table_ident = NULL;
3187     }
3188     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3189
3190     tok_ident = TOK_IDENT;
3191     p = tcc_keywords;
3192     while (*p) {
3193         r = p;
3194         for(;;) {
3195             c = *r++;
3196             if (c == '\0')
3197                 break;
3198         }
3199         tok_alloc(p, r - p - 1);
3200         p = r;
3201     }
3202 }
3203
3204 static void line_macro_output(BufferedFile *f, const char *s, TCCState *s1)
3205 {
3206     switch (s1->Pflag) {
3207         case LINE_MACRO_OUTPUT_FORMAT_STD:
3208             /* "tcc -E -P1" case */
3209             fprintf(s1->ppfp, "# line %d \"%s\"\n", f->line_num, f->filename);
3210             break;
3211
3212         case LINE_MACRO_OUTPUT_FORMAT_NONE:
3213             /* "tcc -E -P" case: don't output a line directive */
3214             break;
3215
3216         case LINE_MACRO_OUTPUT_FORMAT_GCC:
3217         default:
3218             /* "tcc -E" case: a gcc standard by default */
3219             fprintf(s1->ppfp, "# %d \"%s\"%s\n", f->line_num, f->filename, s);
3220             break;
3221     }
3222 }
3223
3224 /* Preprocess the current file */
3225 ST_FUNC int tcc_preprocess(TCCState *s1)
3226 {
3227     BufferedFile *file_ref, **iptr, **iptr_new;
3228     int token_seen, d;
3229     const char *s;
3230
3231     preprocess_init(s1);
3232     ch = file->buf_ptr[0];
3233     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3234     parse_flags = (parse_flags & PARSE_FLAG_ASM_FILE);
3235     parse_flags |= PARSE_FLAG_ASM_COMMENTS | PARSE_FLAG_PREPROCESS |
3236         PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
3237     token_seen = 0;
3238     file->line_ref = 0;
3239     file_ref = NULL;
3240     iptr = s1->include_stack_ptr;
3241
3242     for (;;) {
3243         next();
3244         if (tok == TOK_EOF) {
3245             break;
3246         } else if (file != file_ref) {
3247             if (file_ref)
3248                 line_macro_output(file_ref, "", s1);
3249             goto print_line;
3250         } else if (tok == TOK_LINEFEED) {
3251             if (!token_seen)
3252                 continue;
3253             file->line_ref++;
3254             token_seen = 0;
3255         } else if (!token_seen) {
3256             d = file->line_num - file->line_ref;
3257             if (file != file_ref || d >= 8) {
3258 print_line:
3259                 s = "";
3260                 if (tcc_state->Pflag == LINE_MACRO_OUTPUT_FORMAT_GCC) {
3261                     iptr_new = s1->include_stack_ptr;
3262                     s = iptr_new > iptr ? " 1"
3263                     : iptr_new < iptr ? " 2"
3264                     : iptr_new > s1->include_stack ? " 3"
3265                     : ""
3266                     ;
3267                 }
3268                 line_macro_output(file, s, s1);
3269             } else {
3270                 while (d > 0)
3271                     fputs("\n", s1->ppfp), --d;
3272             }
3273             file->line_ref = (file_ref = file)->line_num;
3274             token_seen = tok != TOK_LINEFEED;
3275             if (!token_seen)
3276                 continue;
3277         }
3278         fputs(get_tok_str(tok, &tokc), s1->ppfp);
3279     }
3280     return 0;
3281 }