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  60 /* outdated -- gr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     } else if (tok == TOK_comment) {
1424         if (s1->ms_extensions) {
1425             next();
1426             skip('(');
1427             if (tok == TOK_lib) {
1428                 next();
1429                 skip(',');
1430                 if (tok != TOK_STR) {
1431                     tcc_error("invalid library specification");
1432                 } else {
1433                     int len = strlen((char *)tokc.cstr->data);
1434                     char *file = tcc_malloc(len + 4); /* filetype, "-l", and \0 at the end */
1435                     file[0] = TCC_FILETYPE_BINARY;
1436                     file[1] = '-';
1437                     file[2] = 'l';
1438                     strcpy(&file[3],(char *)tokc.cstr->data);
1439                     dynarray_add((void ***)&s1->files, &s1->nb_files, file);
1440                 }
1441                 next();
1442                 tok = TOK_LINEFEED;
1443             } else {
1444                 tcc_warning("unknown specifier '%s' in #pragma comment", get_tok_str(tok, &tokc));
1445             }
1446         } else {
1447             tcc_warning("#pragma comment(lib) is ignored");
1448         }
1449     } else if (tok == TOK_once) {
1450         add_cached_include(s1, file->filename, TOK_once);
1451     } else {
1452         tcc_warning("unknown #pragma %s", get_tok_str(tok, &tokc));
1453     }
1454 }
1455
1456 /* is_bof is true if first non space token at beginning of file */
1457 ST_FUNC void preprocess(int is_bof)
1458 {
1459     TCCState *s1 = tcc_state;
1460     int i, c, n, saved_parse_flags;
1461     char buf[1024], *q;
1462     Sym *s;
1463
1464     saved_parse_flags = parse_flags;
1465     parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM |
1466         PARSE_FLAG_LINEFEED;
1467     parse_flags |= (saved_parse_flags & (PARSE_FLAG_ASM_FILE | PARSE_FLAG_ASM_COMMENTS));
1468     next_nomacro();
1469  redo:
1470     switch(tok) {
1471     case TOK_DEFINE:
1472         next_nomacro();
1473         parse_define();
1474         break;
1475     case TOK_UNDEF:
1476         next_nomacro();
1477         s = define_find(tok);
1478         /* undefine symbol by putting an invalid name */
1479         if (s)
1480             define_undef(s);
1481         break;
1482     case TOK_INCLUDE:
1483     case TOK_INCLUDE_NEXT:
1484         ch = file->buf_ptr[0];
1485         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1486         skip_spaces();
1487         if (ch == '<') {
1488             c = '>';
1489             goto read_name;
1490         } else if (ch == '\"') {
1491             c = ch;
1492         read_name:
1493             inp();
1494             q = buf;
1495             while (ch != c && ch != '\n' && ch != CH_EOF) {
1496                 if ((q - buf) < sizeof(buf) - 1)
1497                     *q++ = ch;
1498                 if (ch == '\\') {
1499                     if (handle_stray_noerror() == 0)
1500                         --q;
1501                 } else
1502                     inp();
1503             }
1504             *q = '\0';
1505             minp();
1506 #if 0
1507             /* eat all spaces and comments after include */
1508             /* XXX: slightly incorrect */
1509             while (ch1 != '\n' && ch1 != CH_EOF)
1510                 inp();
1511 #endif
1512         } else {
1513             /* computed #include : either we have only strings or
1514                we have anything enclosed in '<>' */
1515             next();
1516             buf[0] = '\0';
1517             if (tok == TOK_STR) {
1518                 while (tok != TOK_LINEFEED) {
1519                     if (tok != TOK_STR) {
1520                     include_syntax:
1521                         tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1522                     }
1523                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1524                     next();
1525                 }
1526                 c = '\"';
1527             } else {
1528                 int len;
1529                 while (tok != TOK_LINEFEED) {
1530                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1531                     next();
1532                 }
1533                 len = strlen(buf);
1534                 /* check syntax and remove '<>' */
1535                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1536                     goto include_syntax;
1537                 memmove(buf, buf + 1, len - 2);
1538                 buf[len - 2] = '\0';
1539                 c = '>';
1540             }
1541         }
1542
1543         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1544             tcc_error("#include recursion too deep");
1545         /* store current file in stack, but increment stack later below */
1546         *s1->include_stack_ptr = file;
1547
1548         n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1549         for (i = -2; i < n; ++i) {
1550             char buf1[sizeof file->filename];
1551             CachedInclude *e;
1552             BufferedFile **f;
1553             const char *path;
1554
1555             if (i == -2) {
1556                 /* check absolute include path */
1557                 if (!IS_ABSPATH(buf))
1558                     continue;
1559                 buf1[0] = 0;
1560                 i = n; /* force end loop */
1561
1562             } else if (i == -1) {
1563                 /* search in current dir if "header.h" */
1564                 if (c != '\"')
1565                     continue;
1566                 path = file->filename;
1567                 pstrncpy(buf1, path, tcc_basename(path) - path);
1568
1569             } else {
1570                 /* search in all the include paths */
1571                 if (i < s1->nb_include_paths)
1572                     path = s1->include_paths[i];
1573                 else
1574                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1575                 pstrcpy(buf1, sizeof(buf1), path);
1576                 pstrcat(buf1, sizeof(buf1), "/");
1577             }
1578
1579             pstrcat(buf1, sizeof(buf1), buf);
1580
1581             if (tok == TOK_INCLUDE_NEXT)
1582                 for (f = s1->include_stack_ptr; f >= s1->include_stack; --f)
1583                     if (0 == PATHCMP((*f)->filename, buf1)) {
1584 #ifdef INC_DEBUG
1585                         printf("%s: #include_next skipping %s\n", file->filename, buf1);
1586 #endif
1587                         goto include_trynext;
1588                     }
1589
1590             e = search_cached_include(s1, buf1);
1591             if (e && (define_find(e->ifndef_macro) || e->ifndef_macro == TOK_once)) {
1592                 /* no need to parse the include because the 'ifndef macro'
1593                    is defined */
1594 #ifdef INC_DEBUG
1595                 printf("%s: skipping cached %s\n", file->filename, buf1);
1596 #endif
1597                 goto include_done;
1598             }
1599
1600             if (tcc_open(s1, buf1) < 0)
1601 include_trynext:
1602                 continue;
1603
1604 #ifdef INC_DEBUG
1605             printf("%s: including %s\n", file->prev->filename, file->filename);
1606 #endif
1607             /* update target deps */
1608             dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1609                     tcc_strdup(buf1));
1610             /* push current file in stack */
1611             ++s1->include_stack_ptr;
1612             /* add include file debug info */
1613             if (s1->do_debug)
1614                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1615             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1616             ch = file->buf_ptr[0];
1617             goto the_end;
1618         }
1619         tcc_error("include file '%s' not found", buf);
1620 include_done:
1621         break;
1622     case TOK_IFNDEF:
1623         c = 1;
1624         goto do_ifdef;
1625     case TOK_IF:
1626         c = expr_preprocess();
1627         goto do_if;
1628     case TOK_IFDEF:
1629         c = 0;
1630     do_ifdef:
1631         next_nomacro();
1632         if (tok < TOK_IDENT)
1633             tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1634         if (is_bof) {
1635             if (c) {
1636 #ifdef INC_DEBUG
1637                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1638 #endif
1639                 file->ifndef_macro = tok;
1640             }
1641         }
1642         c = (define_find(tok) != 0) ^ c;
1643     do_if:
1644         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1645             tcc_error("memory full (ifdef)");
1646         *s1->ifdef_stack_ptr++ = c;
1647         goto test_skip;
1648     case TOK_ELSE:
1649         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1650             tcc_error("#else without matching #if");
1651         if (s1->ifdef_stack_ptr[-1] & 2)
1652             tcc_error("#else after #else");
1653         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1654         goto test_else;
1655     case TOK_ELIF:
1656         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1657             tcc_error("#elif without matching #if");
1658         c = s1->ifdef_stack_ptr[-1];
1659         if (c > 1)
1660             tcc_error("#elif after #else");
1661         /* last #if/#elif expression was true: we skip */
1662         if (c == 1)
1663             goto skip;
1664         c = expr_preprocess();
1665         s1->ifdef_stack_ptr[-1] = c;
1666     test_else:
1667         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1668             file->ifndef_macro = 0;
1669     test_skip:
1670         if (!(c & 1)) {
1671         skip:
1672             preprocess_skip();
1673             is_bof = 0;
1674             goto redo;
1675         }
1676         break;
1677     case TOK_ENDIF:
1678         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1679             tcc_error("#endif without matching #if");
1680         s1->ifdef_stack_ptr--;
1681         /* '#ifndef macro' was at the start of file. Now we check if
1682            an '#endif' is exactly at the end of file */
1683         if (file->ifndef_macro &&
1684             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1685             file->ifndef_macro_saved = file->ifndef_macro;
1686             /* need to set to zero to avoid false matches if another
1687                #ifndef at middle of file */
1688             file->ifndef_macro = 0;
1689             while (tok != TOK_LINEFEED)
1690                 next_nomacro();
1691             tok_flags |= TOK_FLAG_ENDIF;
1692             goto the_end;
1693         }
1694         break;
1695     case TOK_LINE:
1696         next();
1697         if (tok != TOK_CINT)
1698             tcc_error("A #line format is wrong");
1699     case TOK_PPNUM:
1700         if (tok != TOK_CINT) {
1701             char *p = tokc.cstr->data;
1702             tokc.i = strtoul(p, (char **)&p, 10);
1703         }
1704         i = file->line_num;
1705         file->line_num = tokc.i - 1;
1706         next();
1707         if (tok != TOK_LINEFEED) {
1708             if (tok != TOK_STR) {
1709                 if ((parse_flags & PARSE_FLAG_ASM_COMMENTS) == 0) {
1710                     file->line_num = i;
1711                     tcc_error("#line format is wrong");
1712                 }
1713                 break;
1714             }
1715             pstrcpy(file->filename, sizeof(file->filename),
1716                     (char *)tokc.cstr->data);
1717         }
1718         if (s1->do_debug)
1719             put_stabs(file->filename, N_BINCL, 0, 0, 0);
1720         break;
1721     case TOK_ERROR:
1722     case TOK_WARNING:
1723         c = tok;
1724         ch = file->buf_ptr[0];
1725         skip_spaces();
1726         q = buf;
1727         while (ch != '\n' && ch != CH_EOF) {
1728             if ((q - buf) < sizeof(buf) - 1)
1729                 *q++ = ch;
1730             if (ch == '\\') {
1731                 if (handle_stray_noerror() == 0)
1732                     --q;
1733             } else
1734                 inp();
1735         }
1736         *q = '\0';
1737         if (c == TOK_ERROR)
1738             tcc_error("#error %s", buf);
1739         else
1740             tcc_warning("#warning %s", buf);
1741         break;
1742     case TOK_PRAGMA:
1743         pragma_parse(s1);
1744         break;
1745     default:
1746         if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_PPNUM) {
1747             /* '!' is ignored to allow C scripts. numbers are ignored
1748                to emulate cpp behaviour */
1749         } else {
1750             if (!(parse_flags & PARSE_FLAG_ASM_COMMENTS))
1751                 tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1752             else {
1753                 /* this is a gas line comment in an 'S' file. */
1754                 file->buf_ptr = parse_line_comment(file->buf_ptr);
1755                 goto the_end;
1756             }
1757         }
1758         break;
1759     }
1760     /* ignore other preprocess commands or #! for C scripts */
1761     while (tok != TOK_LINEFEED)
1762         next_nomacro();
1763  the_end:
1764     parse_flags = saved_parse_flags;
1765 }
1766
1767 /* evaluate escape codes in a string. */
1768 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1769 {
1770     int c, n;
1771     const uint8_t *p;
1772
1773     p = buf;
1774     for(;;) {
1775         c = *p;
1776         if (c == '\0')
1777             break;
1778         if (c == '\\') {
1779             p++;
1780             /* escape */
1781             c = *p;
1782             switch(c) {
1783             case '0': case '1': case '2': case '3':
1784             case '4': case '5': case '6': case '7':
1785                 /* at most three octal digits */
1786                 n = c - '0';
1787                 p++;
1788                 c = *p;
1789                 if (isoct(c)) {
1790                     n = n * 8 + c - '0';
1791                     p++;
1792                     c = *p;
1793                     if (isoct(c)) {
1794                         n = n * 8 + c - '0';
1795                         p++;
1796                     }
1797                 }
1798                 c = n;
1799                 goto add_char_nonext;
1800             case 'x':
1801             case 'u':
1802             case 'U':
1803                 p++;
1804                 n = 0;
1805                 for(;;) {
1806                     c = *p;
1807                     if (c >= 'a' && c <= 'f')
1808                         c = c - 'a' + 10;
1809                     else if (c >= 'A' && c <= 'F')
1810                         c = c - 'A' + 10;
1811                     else if (isnum(c))
1812                         c = c - '0';
1813                     else
1814                         break;
1815                     n = n * 16 + c;
1816                     p++;
1817                 }
1818                 c = n;
1819                 goto add_char_nonext;
1820             case 'a':
1821                 c = '\a';
1822                 break;
1823             case 'b':
1824                 c = '\b';
1825                 break;
1826             case 'f':
1827                 c = '\f';
1828                 break;
1829             case 'n':
1830                 c = '\n';
1831                 break;
1832             case 'r':
1833                 c = '\r';
1834                 break;
1835             case 't':
1836                 c = '\t';
1837                 break;
1838             case 'v':
1839                 c = '\v';
1840                 break;
1841             case 'e':
1842                 if (!gnu_ext)
1843                     goto invalid_escape;
1844                 c = 27;
1845                 break;
1846             case '\'':
1847             case '\"':
1848             case '\\':
1849             case '?':
1850                 break;
1851             default:
1852             invalid_escape:
1853                 if (c >= '!' && c <= '~')
1854                     tcc_warning("unknown escape sequence: \'\\%c\'", c);
1855                 else
1856                     tcc_warning("unknown escape sequence: \'\\x%x\'", c);
1857                 break;
1858             }
1859         }
1860         p++;
1861     add_char_nonext:
1862         if (!is_long)
1863             cstr_ccat(outstr, c);
1864         else
1865             cstr_wccat(outstr, c);
1866     }
1867     /* add a trailing '\0' */
1868     if (!is_long)
1869         cstr_ccat(outstr, '\0');
1870     else
1871         cstr_wccat(outstr, '\0');
1872 }
1873
1874 /* we use 64 bit numbers */
1875 #define BN_SIZE 2
1876
1877 /* bn = (bn << shift) | or_val */
1878 static void bn_lshift(unsigned int *bn, int shift, int or_val)
1879 {
1880     int i;
1881     unsigned int v;
1882     for(i=0;i<BN_SIZE;i++) {
1883         v = bn[i];
1884         bn[i] = (v << shift) | or_val;
1885         or_val = v >> (32 - shift);
1886     }
1887 }
1888
1889 static void bn_zero(unsigned int *bn)
1890 {
1891     int i;
1892     for(i=0;i<BN_SIZE;i++) {
1893         bn[i] = 0;
1894     }
1895 }
1896
1897 /* parse number in null terminated string 'p' and return it in the
1898    current token */
1899 static void parse_number(const char *p)
1900 {
1901     int b, t, shift, frac_bits, s, exp_val, ch;
1902     char *q;
1903     unsigned int bn[BN_SIZE];
1904     double d;
1905
1906     /* number */
1907     q = token_buf;
1908     ch = *p++;
1909     t = ch;
1910     ch = *p++;
1911     *q++ = t;
1912     b = 10;
1913     if (t == '.') {
1914         goto float_frac_parse;
1915     } else if (t == '0') {
1916         if (ch == 'x' || ch == 'X') {
1917             q--;
1918             ch = *p++;
1919             b = 16;
1920         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
1921             q--;
1922             ch = *p++;
1923             b = 2;
1924         }
1925     }
1926     /* parse all digits. cannot check octal numbers at this stage
1927        because of floating point constants */
1928     while (1) {
1929         if (ch >= 'a' && ch <= 'f')
1930             t = ch - 'a' + 10;
1931         else if (ch >= 'A' && ch <= 'F')
1932             t = ch - 'A' + 10;
1933         else if (isnum(ch))
1934             t = ch - '0';
1935         else
1936             break;
1937         if (t >= b)
1938             break;
1939         if (q >= token_buf + STRING_MAX_SIZE) {
1940         num_too_long:
1941             tcc_error("number too long");
1942         }
1943         *q++ = ch;
1944         ch = *p++;
1945     }
1946     if (ch == '.' ||
1947         ((ch == 'e' || ch == 'E') && b == 10) ||
1948         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
1949         if (b != 10) {
1950             /* NOTE: strtox should support that for hexa numbers, but
1951                non ISOC99 libcs do not support it, so we prefer to do
1952                it by hand */
1953             /* hexadecimal or binary floats */
1954             /* XXX: handle overflows */
1955             *q = '\0';
1956             if (b == 16)
1957                 shift = 4;
1958             else
1959                 shift = 1;
1960             bn_zero(bn);
1961             q = token_buf;
1962             while (1) {
1963                 t = *q++;
1964                 if (t == '\0') {
1965                     break;
1966                 } else if (t >= 'a') {
1967                     t = t - 'a' + 10;
1968                 } else if (t >= 'A') {
1969                     t = t - 'A' + 10;
1970                 } else {
1971                     t = t - '0';
1972                 }
1973                 bn_lshift(bn, shift, t);
1974             }
1975             frac_bits = 0;
1976             if (ch == '.') {
1977                 ch = *p++;
1978                 while (1) {
1979                     t = ch;
1980                     if (t >= 'a' && t <= 'f') {
1981                         t = t - 'a' + 10;
1982                     } else if (t >= 'A' && t <= 'F') {
1983                         t = t - 'A' + 10;
1984                     } else if (t >= '0' && t <= '9') {
1985                         t = t - '0';
1986                     } else {
1987                         break;
1988                     }
1989                     if (t >= b)
1990                         tcc_error("invalid digit");
1991                     bn_lshift(bn, shift, t);
1992                     frac_bits += shift;
1993                     ch = *p++;
1994                 }
1995             }
1996             if (ch != 'p' && ch != 'P')
1997                 expect("exponent");
1998             ch = *p++;
1999             s = 1;
2000             exp_val = 0;
2001             if (ch == '+') {
2002                 ch = *p++;
2003             } else if (ch == '-') {
2004                 s = -1;
2005                 ch = *p++;
2006             }
2007             if (ch < '0' || ch > '9')
2008                 expect("exponent digits");
2009             while (ch >= '0' && ch <= '9') {
2010                 exp_val = exp_val * 10 + ch - '0';
2011                 ch = *p++;
2012             }
2013             exp_val = exp_val * s;
2014
2015             /* now we can generate the number */
2016             /* XXX: should patch directly float number */
2017             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
2018             d = ldexp(d, exp_val - frac_bits);
2019             t = toup(ch);
2020             if (t == 'F') {
2021                 ch = *p++;
2022                 tok = TOK_CFLOAT;
2023                 /* float : should handle overflow */
2024                 tokc.f = (float)d;
2025             } else if (t == 'L') {
2026                 ch = *p++;
2027 #ifdef TCC_TARGET_PE
2028                 tok = TOK_CDOUBLE;
2029                 tokc.d = d;
2030 #else
2031                 tok = TOK_CLDOUBLE;
2032                 /* XXX: not large enough */
2033                 tokc.ld = (long double)d;
2034 #endif
2035             } else {
2036                 tok = TOK_CDOUBLE;
2037                 tokc.d = d;
2038             }
2039         } else {
2040             /* decimal floats */
2041             if (ch == '.') {
2042                 if (q >= token_buf + STRING_MAX_SIZE)
2043                     goto num_too_long;
2044                 *q++ = ch;
2045                 ch = *p++;
2046             float_frac_parse:
2047                 while (ch >= '0' && ch <= '9') {
2048                     if (q >= token_buf + STRING_MAX_SIZE)
2049                         goto num_too_long;
2050                     *q++ = ch;
2051                     ch = *p++;
2052                 }
2053             }
2054             if (ch == 'e' || ch == 'E') {
2055                 if (q >= token_buf + STRING_MAX_SIZE)
2056                     goto num_too_long;
2057                 *q++ = ch;
2058                 ch = *p++;
2059                 if (ch == '-' || ch == '+') {
2060                     if (q >= token_buf + STRING_MAX_SIZE)
2061                         goto num_too_long;
2062                     *q++ = ch;
2063                     ch = *p++;
2064                 }
2065                 if (ch < '0' || ch > '9')
2066                     expect("exponent digits");
2067                 while (ch >= '0' && ch <= '9') {
2068                     if (q >= token_buf + STRING_MAX_SIZE)
2069                         goto num_too_long;
2070                     *q++ = ch;
2071                     ch = *p++;
2072                 }
2073             }
2074             *q = '\0';
2075             t = toup(ch);
2076             errno = 0;
2077             if (t == 'F') {
2078                 ch = *p++;
2079                 tok = TOK_CFLOAT;
2080                 tokc.f = strtof(token_buf, NULL);
2081             } else if (t == 'L') {
2082                 ch = *p++;
2083 #ifdef TCC_TARGET_PE
2084                 tok = TOK_CDOUBLE;
2085                 tokc.d = strtod(token_buf, NULL);
2086 #else
2087                 tok = TOK_CLDOUBLE;
2088                 tokc.ld = strtold(token_buf, NULL);
2089 #endif
2090             } else {
2091                 tok = TOK_CDOUBLE;
2092                 tokc.d = strtod(token_buf, NULL);
2093             }
2094         }
2095     } else {
2096         unsigned long long n, n1;
2097         int lcount, ucount, must_64bit;
2098         const char *p1;
2099
2100         /* integer number */
2101         *q = '\0';
2102         q = token_buf;
2103         if (b == 10 && *q == '0') {
2104             b = 8;
2105             q++;
2106         }
2107         n = 0;
2108         while(1) {
2109             t = *q++;
2110             /* no need for checks except for base 10 / 8 errors */
2111             if (t == '\0')
2112                 break;
2113             else if (t >= 'a')
2114                 t = t - 'a' + 10;
2115             else if (t >= 'A')
2116                 t = t - 'A' + 10;
2117             else
2118                 t = t - '0';
2119             if (t >= b)
2120                 tcc_error("invalid digit");
2121             n1 = n;
2122             n = n * b + t;
2123             /* detect overflow */
2124             /* XXX: this test is not reliable */
2125             if (n < n1)
2126                 tcc_error("integer constant overflow");
2127         }
2128
2129         /* Determine the characteristics (unsigned and/or 64bit) the type of
2130            the constant must have according to the constant suffix(es) */
2131         lcount = ucount = must_64bit = 0;
2132         p1 = p;
2133         for(;;) {
2134             t = toup(ch);
2135             if (t == 'L') {
2136                 if (lcount >= 2)
2137                     tcc_error("three 'l's in integer constant");
2138                 if (lcount && *(p - 1) != ch)
2139                     tcc_error("incorrect integer suffix: %s", p1);
2140                 lcount++;
2141 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
2142                 if (lcount == 2)
2143 #endif
2144                     must_64bit = 1;
2145                 ch = *p++;
2146             } else if (t == 'U') {
2147                 if (ucount >= 1)
2148                     tcc_error("two 'u's in integer constant");
2149                 ucount++;
2150                 ch = *p++;
2151             } else {
2152                 break;
2153             }
2154         }
2155
2156         /* Whether 64 bits are needed to hold the constant's value */
2157         if (n & 0xffffffff00000000LL || must_64bit) {
2158             tok = TOK_CLLONG;
2159             n1 = n >> 32;
2160         } else {
2161             tok = TOK_CINT;
2162             n1 = n;
2163         }
2164
2165         /* Whether type must be unsigned to hold the constant's value */
2166         if (ucount || ((n1 >> 31) && (b != 10))) {
2167             if (tok == TOK_CLLONG)
2168                 tok = TOK_CULLONG;
2169             else
2170                 tok = TOK_CUINT;
2171         /* If decimal and no unsigned suffix, bump to 64 bits or throw error */
2172         } else if (n1 >> 31) {
2173             if (tok == TOK_CINT)
2174                 tok = TOK_CLLONG;
2175             else
2176                 tcc_error("integer constant overflow");
2177         }
2178
2179         if (tok == TOK_CINT || tok == TOK_CUINT)
2180             tokc.ui = n;
2181         else
2182             tokc.ull = n;
2183     }
2184     if (ch)
2185         tcc_error("invalid number\n");
2186 }
2187
2188
2189 #define PARSE2(c1, tok1, c2, tok2)              \
2190     case c1:                                    \
2191         PEEKC(c, p);                            \
2192         if (c == c2) {                          \
2193             p++;                                \
2194             tok = tok2;                         \
2195         } else {                                \
2196             tok = tok1;                         \
2197         }                                       \
2198         break;
2199
2200 /* return next token without macro substitution */
2201 static inline void next_nomacro1(void)
2202 {
2203     int t, c, is_long;
2204     TokenSym *ts;
2205     uint8_t *p, *p1;
2206     unsigned int h;
2207
2208     p = file->buf_ptr;
2209  redo_no_start:
2210     c = *p;
2211     switch(c) {
2212     case ' ':
2213     case '\t':
2214         tok = c;
2215         p++;
2216         goto keep_tok_flags;
2217     case '\f':
2218     case '\v':
2219     case '\r':
2220         p++;
2221         goto redo_no_start;
2222     case '\\':
2223         /* first look if it is in fact an end of buffer */
2224         if (p >= file->buf_end) {
2225             file->buf_ptr = p;
2226             handle_eob();
2227             p = file->buf_ptr;
2228             if (p >= file->buf_end)
2229                 goto parse_eof;
2230             else
2231                 goto redo_no_start;
2232         } else {
2233             file->buf_ptr = p;
2234             ch = *p;
2235             handle_stray();
2236             p = file->buf_ptr;
2237             goto redo_no_start;
2238         }
2239     parse_eof:
2240         {
2241             TCCState *s1 = tcc_state;
2242             if ((parse_flags & PARSE_FLAG_LINEFEED)
2243                 && !(tok_flags & TOK_FLAG_EOF)) {
2244                 tok_flags |= TOK_FLAG_EOF;
2245                 tok = TOK_LINEFEED;
2246                 goto keep_tok_flags;
2247             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2248                 tok = TOK_EOF;
2249             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2250                 tcc_error("missing #endif");
2251             } else if (s1->include_stack_ptr == s1->include_stack) {
2252                 /* no include left : end of file. */
2253                 tok = TOK_EOF;
2254             } else {
2255                 tok_flags &= ~TOK_FLAG_EOF;
2256                 /* pop include file */
2257
2258                 /* test if previous '#endif' was after a #ifdef at
2259                    start of file */
2260                 if (tok_flags & TOK_FLAG_ENDIF) {
2261 #ifdef INC_DEBUG
2262                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2263 #endif
2264                     add_cached_include(s1, file->filename, file->ifndef_macro_saved);
2265                     tok_flags &= ~TOK_FLAG_ENDIF;
2266                 }
2267
2268                 /* add end of include file debug info */
2269                 if (tcc_state->do_debug) {
2270                     put_stabd(N_EINCL, 0, 0);
2271                 }
2272                 /* pop include stack */
2273                 tcc_close();
2274                 s1->include_stack_ptr--;
2275                 p = file->buf_ptr;
2276                 goto redo_no_start;
2277             }
2278         }
2279         break;
2280
2281     case '\n':
2282         file->line_num++;
2283         tok_flags |= TOK_FLAG_BOL;
2284         p++;
2285 maybe_newline:
2286         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2287             goto redo_no_start;
2288         tok = TOK_LINEFEED;
2289         goto keep_tok_flags;
2290
2291     case '#':
2292         /* XXX: simplify */
2293         PEEKC(c, p);
2294         if (is_space(c) && (parse_flags & PARSE_FLAG_ASM_FILE)) {
2295             p = parse_line_comment(p);
2296             goto redo_no_start;
2297         }
2298         else
2299         if ((tok_flags & TOK_FLAG_BOL) &&
2300             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2301             file->buf_ptr = p;
2302             preprocess(tok_flags & TOK_FLAG_BOF);
2303             p = file->buf_ptr;
2304             goto maybe_newline;
2305         } else {
2306             if (c == '#') {
2307                 p++;
2308                 tok = TOK_TWOSHARPS;
2309             } else {
2310                 if (parse_flags & PARSE_FLAG_ASM_COMMENTS) {
2311                     p = parse_line_comment(p - 1);
2312                     goto redo_no_start;
2313                 } else {
2314                     tok = '#';
2315                 }
2316             }
2317         }
2318         break;
2319
2320     /* dollar is allowed to start identifiers when not parsing asm */
2321     case '$':
2322         if (!tcc_state->dollars_in_identifiers
2323         || (parse_flags & PARSE_FLAG_ASM_FILE)) goto parse_simple;
2324
2325     case 'a': case 'b': case 'c': case 'd':
2326     case 'e': case 'f': case 'g': case 'h':
2327     case 'i': case 'j': case 'k': case 'l':
2328     case 'm': case 'n': case 'o': case 'p':
2329     case 'q': case 'r': case 's': case 't':
2330     case 'u': case 'v': case 'w': case 'x':
2331     case 'y': case 'z':
2332     case 'A': case 'B': case 'C': case 'D':
2333     case 'E': case 'F': case 'G': case 'H':
2334     case 'I': case 'J': case 'K':
2335     case 'M': case 'N': case 'O': case 'P':
2336     case 'Q': case 'R': case 'S': case 'T':
2337     case 'U': case 'V': case 'W': case 'X':
2338     case 'Y': case 'Z':
2339     case '_':
2340     parse_ident_fast:
2341         p1 = p;
2342         h = TOK_HASH_INIT;
2343         h = TOK_HASH_FUNC(h, c);
2344         p++;
2345         for(;;) {
2346             c = *p;
2347             if (!isidnum_table[c-CH_EOF]
2348             && (tcc_state->dollars_in_identifiers ? (c != '$') : 1))
2349                 break;
2350             h = TOK_HASH_FUNC(h, c);
2351             p++;
2352         }
2353         if (c != '\\') {
2354             TokenSym **pts;
2355             int len;
2356
2357             /* fast case : no stray found, so we have the full token
2358                and we have already hashed it */
2359             len = p - p1;
2360             h &= (TOK_HASH_SIZE - 1);
2361             pts = &hash_ident[h];
2362             for(;;) {
2363                 ts = *pts;
2364                 if (!ts)
2365                     break;
2366                 if (ts->len == len && !memcmp(ts->str, p1, len))
2367                     goto token_found;
2368                 pts = &(ts->hash_next);
2369             }
2370             ts = tok_alloc_new(pts, (char *) p1, len);
2371         token_found: ;
2372         } else {
2373             /* slower case */
2374             cstr_reset(&tokcstr);
2375
2376             while (p1 < p) {
2377                 cstr_ccat(&tokcstr, *p1);
2378                 p1++;
2379             }
2380             p--;
2381             PEEKC(c, p);
2382         parse_ident_slow:
2383             while (isidnum_table[c-CH_EOF]
2384             || (tcc_state->dollars_in_identifiers ? (c == '$') : 0)) {
2385                 cstr_ccat(&tokcstr, c);
2386                 PEEKC(c, p);
2387             }
2388             ts = tok_alloc(tokcstr.data, tokcstr.size);
2389         }
2390         tok = ts->tok;
2391         break;
2392     case 'L':
2393         t = p[1];
2394         if (t != '\\' && t != '\'' && t != '\"') {
2395             /* fast case */
2396             goto parse_ident_fast;
2397         } else {
2398             PEEKC(c, p);
2399             if (c == '\'' || c == '\"') {
2400                 is_long = 1;
2401                 goto str_const;
2402             } else {
2403                 cstr_reset(&tokcstr);
2404                 cstr_ccat(&tokcstr, 'L');
2405                 goto parse_ident_slow;
2406             }
2407         }
2408         break;
2409     case '0': case '1': case '2': case '3':
2410     case '4': case '5': case '6': case '7':
2411     case '8': case '9':
2412
2413         cstr_reset(&tokcstr);
2414         /* after the first digit, accept digits, alpha, '.' or sign if
2415            prefixed by 'eEpP' */
2416     parse_num:
2417         for(;;) {
2418             t = c;
2419             cstr_ccat(&tokcstr, c);
2420             PEEKC(c, p);
2421             if (!(isnum(c) || isid(c) || c == '.' ||
2422                   ((c == '+' || c == '-') &&
2423                    (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2424                 break;
2425         }
2426         /* We add a trailing '\0' to ease parsing */
2427         cstr_ccat(&tokcstr, '\0');
2428         tokc.cstr = &tokcstr;
2429         tok = TOK_PPNUM;
2430         break;
2431     case '.':
2432         /* special dot handling because it can also start a number */
2433         PEEKC(c, p);
2434         if (isnum(c)) {
2435             cstr_reset(&tokcstr);
2436             cstr_ccat(&tokcstr, '.');
2437             goto parse_num;
2438         } else if (c == '.') {
2439             PEEKC(c, p);
2440             if (c != '.') {
2441                 if ((parse_flags & PARSE_FLAG_ASM_COMMENTS) == 0)
2442                     expect("'.'");
2443                 tok = '.';
2444                 break;
2445             }
2446             PEEKC(c, p);
2447             tok = TOK_DOTS;
2448         } else {
2449             tok = '.';
2450         }
2451         break;
2452     case '\'':
2453     case '\"':
2454         is_long = 0;
2455     str_const:
2456         {
2457             CString str;
2458             int sep;
2459
2460             sep = c;
2461
2462             /* parse the string */
2463             cstr_new(&str);
2464             p = parse_pp_string(p, sep, &str);
2465             cstr_ccat(&str, '\0');
2466
2467             /* eval the escape (should be done as TOK_PPNUM) */
2468             cstr_reset(&tokcstr);
2469             parse_escape_string(&tokcstr, str.data, is_long);
2470             cstr_free(&str);
2471
2472             if (sep == '\'') {
2473                 int char_size;
2474                 /* XXX: make it portable */
2475                 if (!is_long)
2476                     char_size = 1;
2477                 else
2478                     char_size = sizeof(nwchar_t);
2479                 if (tokcstr.size <= char_size)
2480                     tcc_error("empty character constant");
2481                 if (tokcstr.size > 2 * char_size)
2482                     tcc_warning("multi-character character constant");
2483                 if (!is_long) {
2484                     tokc.i = *(int8_t *)tokcstr.data;
2485                     tok = TOK_CCHAR;
2486                 } else {
2487                     tokc.i = *(nwchar_t *)tokcstr.data;
2488                     tok = TOK_LCHAR;
2489                 }
2490             } else {
2491                 tokc.cstr = &tokcstr;
2492                 if (!is_long)
2493                     tok = TOK_STR;
2494                 else
2495                     tok = TOK_LSTR;
2496             }
2497         }
2498         break;
2499
2500     case '<':
2501         PEEKC(c, p);
2502         if (c == '=') {
2503             p++;
2504             tok = TOK_LE;
2505         } else if (c == '<') {
2506             PEEKC(c, p);
2507             if (c == '=') {
2508                 p++;
2509                 tok = TOK_A_SHL;
2510             } else {
2511                 tok = TOK_SHL;
2512             }
2513         } else {
2514             tok = TOK_LT;
2515         }
2516         break;
2517
2518     case '>':
2519         PEEKC(c, p);
2520         if (c == '=') {
2521             p++;
2522             tok = TOK_GE;
2523         } else if (c == '>') {
2524             PEEKC(c, p);
2525             if (c == '=') {
2526                 p++;
2527                 tok = TOK_A_SAR;
2528             } else {
2529                 tok = TOK_SAR;
2530             }
2531         } else {
2532             tok = TOK_GT;
2533         }
2534         break;
2535
2536     case '&':
2537         PEEKC(c, p);
2538         if (c == '&') {
2539             p++;
2540             tok = TOK_LAND;
2541         } else if (c == '=') {
2542             p++;
2543             tok = TOK_A_AND;
2544         } else {
2545             tok = '&';
2546         }
2547         break;
2548
2549     case '|':
2550         PEEKC(c, p);
2551         if (c == '|') {
2552             p++;
2553             tok = TOK_LOR;
2554         } else if (c == '=') {
2555             p++;
2556             tok = TOK_A_OR;
2557         } else {
2558             tok = '|';
2559         }
2560         break;
2561
2562     case '+':
2563         PEEKC(c, p);
2564         if (c == '+') {
2565             p++;
2566             tok = TOK_INC;
2567         } else if (c == '=') {
2568             p++;
2569             tok = TOK_A_ADD;
2570         } else {
2571             tok = '+';
2572         }
2573         break;
2574
2575     case '-':
2576         PEEKC(c, p);
2577         if (c == '-') {
2578             p++;
2579             tok = TOK_DEC;
2580         } else if (c == '=') {
2581             p++;
2582             tok = TOK_A_SUB;
2583         } else if (c == '>') {
2584             p++;
2585             tok = TOK_ARROW;
2586         } else {
2587             tok = '-';
2588         }
2589         break;
2590
2591     PARSE2('!', '!', '=', TOK_NE)
2592     PARSE2('=', '=', '=', TOK_EQ)
2593     PARSE2('*', '*', '=', TOK_A_MUL)
2594     PARSE2('%', '%', '=', TOK_A_MOD)
2595     PARSE2('^', '^', '=', TOK_A_XOR)
2596
2597         /* comments or operator */
2598     case '/':
2599         PEEKC(c, p);
2600         if (c == '*') {
2601             p = parse_comment(p);
2602             /* comments replaced by a blank */
2603             tok = ' ';
2604             goto keep_tok_flags;
2605         } else if (c == '/') {
2606             p = parse_line_comment(p);
2607             tok = ' ';
2608             goto keep_tok_flags;
2609         } else if (c == '=') {
2610             p++;
2611             tok = TOK_A_DIV;
2612         } else {
2613             tok = '/';
2614         }
2615         break;
2616
2617         /* simple tokens */
2618     case '(':
2619     case ')':
2620     case '[':
2621     case ']':
2622     case '{':
2623     case '}':
2624     case ',':
2625     case ';':
2626     case ':':
2627     case '?':
2628     case '~':
2629     case '@': /* only used in assembler */
2630     parse_simple:
2631         tok = c;
2632         p++;
2633         break;
2634     default:
2635         if ((parse_flags & PARSE_FLAG_ASM_FILE) == 0)
2636             tcc_error("unrecognized character \\x%02x", c);
2637         else {
2638             tok = ' ';
2639             p++;
2640         }
2641         break;
2642     }
2643     tok_flags = 0;
2644 keep_tok_flags:
2645     file->buf_ptr = p;
2646 #if defined(PARSE_DEBUG)
2647     printf("token = %s\n", get_tok_str(tok, &tokc));
2648 #endif
2649 }
2650
2651 /* return next token without macro substitution. Can read input from
2652    macro_ptr buffer */
2653 static void next_nomacro_spc(void)
2654 {
2655     if (macro_ptr) {
2656     redo:
2657         tok = *macro_ptr;
2658         if (tok) {
2659             TOK_GET(&tok, &macro_ptr, &tokc);
2660             if (tok == TOK_LINENUM) {
2661                 file->line_num = tokc.i;
2662                 goto redo;
2663             }
2664         }
2665     } else {
2666         next_nomacro1();
2667     }
2668 }
2669
2670 ST_FUNC void next_nomacro(void)
2671 {
2672     do {
2673         next_nomacro_spc();
2674     } while (is_space(tok));
2675 }
2676
2677 /* substitute arguments in replacement lists in macro_str by the values in
2678    args (field d) and return allocated string */
2679 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2680 {
2681     int last_tok, t, spc;
2682     const int *st;
2683     Sym *s;
2684     CValue cval;
2685     TokenString str;
2686     CString cstr;
2687
2688     tok_str_new(&str);
2689     last_tok = 0;
2690     while(1) {
2691         TOK_GET(&t, &macro_str, &cval);
2692         if (!t)
2693             break;
2694         if (t == '#') {
2695             /* stringize */
2696             TOK_GET(&t, &macro_str, &cval);
2697             if (!t)
2698                 break;
2699             s = sym_find2(args, t);
2700             if (s) {
2701                 cstr_new(&cstr);
2702                 st = s->d;
2703                 spc = 0;
2704                 while (*st) {
2705                     TOK_GET(&t, &st, &cval);
2706                     if (!check_space(t, &spc))
2707                         cstr_cat(&cstr, get_tok_str(t, &cval));
2708                 }
2709                 cstr.size -= spc;
2710                 cstr_ccat(&cstr, '\0');
2711 #ifdef PP_DEBUG
2712                 printf("stringize: %s\n", (char *)cstr.data);
2713 #endif
2714                 /* add string */
2715                 cval.cstr = &cstr;
2716                 tok_str_add2(&str, TOK_STR, &cval);
2717                 cstr_free(&cstr);
2718             } else {
2719                 tok_str_add2(&str, t, &cval);
2720             }
2721         } else if (t >= TOK_IDENT) {
2722             s = sym_find2(args, t);
2723             if (s) {
2724                 st = s->d;
2725                 /* if '##' is present before or after, no arg substitution */
2726                 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2727                     /* special case for var arg macros : ## eats the
2728                        ',' if empty VA_ARGS variable. */
2729                     /* XXX: test of the ',' is not 100%
2730                        reliable. should fix it to avoid security
2731                        problems */
2732                     if (gnu_ext && s->type.t &&
2733                         last_tok == TOK_TWOSHARPS &&
2734                         str.len >= 2 && str.str[str.len - 2] == ',') {
2735                         if (*st == TOK_PLCHLDR) {
2736                             /* suppress ',' '##' */
2737                             str.len -= 2;
2738                         } else {
2739                             /* suppress '##' and add variable */
2740                             str.len--;
2741                             goto add_var;
2742                         }
2743                     } else {
2744                         int t1;
2745                     add_var:
2746                         for(;;) {
2747                             TOK_GET(&t1, &st, &cval);
2748                             if (!t1)
2749                                 break;
2750                             tok_str_add2(&str, t1, &cval);
2751                         }
2752                     }
2753                 } else if (*st != TOK_PLCHLDR) {
2754                     /* NOTE: the stream cannot be read when macro
2755                        substituing an argument */
2756                     macro_subst(&str, nested_list, st, NULL);
2757                 }
2758             } else {
2759                 tok_str_add(&str, t);
2760             }
2761         } else {
2762             tok_str_add2(&str, t, &cval);
2763         }
2764         last_tok = t;
2765     }
2766     tok_str_add(&str, 0);
2767     return str.str;
2768 }
2769
2770 static char const ab_month_name[12][4] =
2771 {
2772     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2773     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2774 };
2775
2776 /* do macro substitution of current token with macro 's' and add
2777    result to (tok_str,tok_len). 'nested_list' is the list of all
2778    macros we got inside to avoid recursing. Return non zero if no
2779    substitution needs to be done */
2780 static int macro_subst_tok(TokenString *tok_str,
2781                            Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2782 {
2783     Sym *args, *sa, *sa1;
2784     int mstr_allocated, parlevel, *mstr, t, t1, spc;
2785     const int *p;
2786     TokenString str;
2787     char *cstrval;
2788     CValue cval;
2789     CString cstr;
2790     char buf[32];
2791
2792     /* if symbol is a macro, prepare substitution */
2793     /* special macros */
2794     if (tok == TOK___LINE__) {
2795         snprintf(buf, sizeof(buf), "%d", file->line_num);
2796         cstrval = buf;
2797         t1 = TOK_PPNUM;
2798         goto add_cstr1;
2799     } else if (tok == TOK___FILE__) {
2800         cstrval = file->filename;
2801         goto add_cstr;
2802     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2803         time_t ti;
2804         struct tm *tm;
2805
2806         time(&ti);
2807         tm = localtime(&ti);
2808         if (tok == TOK___DATE__) {
2809             snprintf(buf, sizeof(buf), "%s %2d %d",
2810                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2811         } else {
2812             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2813                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2814         }
2815         cstrval = buf;
2816     add_cstr:
2817         t1 = TOK_STR;
2818     add_cstr1:
2819         cstr_new(&cstr);
2820         cstr_cat(&cstr, cstrval);
2821         cstr_ccat(&cstr, '\0');
2822         cval.cstr = &cstr;
2823         tok_str_add2(tok_str, t1, &cval);
2824         cstr_free(&cstr);
2825     } else {
2826         mstr = s->d;
2827         mstr_allocated = 0;
2828         if (s->type.t == MACRO_FUNC) {
2829             /* NOTE: we do not use next_nomacro to avoid eating the
2830                next token. XXX: find better solution */
2831         redo:
2832             if (macro_ptr) {
2833                 p = macro_ptr;
2834                 while (is_space(t = *p) || TOK_LINEFEED == t)
2835                     ++p;
2836                 if (t == 0 && can_read_stream) {
2837                     /* end of macro stream: we must look at the token
2838                        after in the file */
2839                     struct macro_level *ml = *can_read_stream;
2840                     macro_ptr = NULL;
2841                     if (ml)
2842                     {
2843                         macro_ptr = ml->p;
2844                         ml->p = NULL;
2845                         *can_read_stream = ml -> prev;
2846                     }
2847                     /* also, end of scope for nested defined symbol */
2848                     (*nested_list)->v = -1;
2849                     goto redo;
2850                 }
2851             } else {
2852                 ch = file->buf_ptr[0];
2853                 while (is_space(ch) || ch == '\n' || ch == '/')
2854                   {
2855                     if (ch == '/')
2856                       {
2857                         int c;
2858                         uint8_t *p = file->buf_ptr;
2859                         PEEKC(c, p);
2860                         if (c == '*') {
2861                             p = parse_comment(p);
2862                             file->buf_ptr = p - 1;
2863                         } else if (c == '/') {
2864                             p = parse_line_comment(p);
2865                             file->buf_ptr = p - 1;
2866                         } else
2867                           break;
2868                       }
2869                     cinp();
2870                   }
2871                 t = ch;
2872             }
2873             if (t != '(') /* no macro subst */
2874                 return -1;
2875
2876             /* argument macro */
2877             next_nomacro();
2878             next_nomacro();
2879             args = NULL;
2880             sa = s->next;
2881             /* NOTE: empty args are allowed, except if no args */
2882             for(;;) {
2883                 /* handle '()' case */
2884                 if (!args && !sa && tok == ')')
2885                     break;
2886                 if (!sa)
2887                     tcc_error("macro '%s' used with too many args",
2888                           get_tok_str(s->v, 0));
2889                 tok_str_new(&str);
2890                 parlevel = spc = 0;
2891                 /* NOTE: non zero sa->t indicates VA_ARGS */
2892                 while ((parlevel > 0 ||
2893                         (tok != ')' &&
2894                          (tok != ',' || sa->type.t))) &&
2895                        tok != -1) {
2896                     if (tok == '(')
2897                         parlevel++;
2898                     else if (tok == ')')
2899                         parlevel--;
2900                     if (tok == TOK_LINEFEED)
2901                         tok = ' ';
2902                     if (!check_space(tok, &spc))
2903                         tok_str_add2(&str, tok, &tokc);
2904                     next_nomacro_spc();
2905                 }
2906                 if (!str.len)
2907                     tok_str_add(&str, TOK_PLCHLDR);
2908                 str.len -= spc;
2909                 tok_str_add(&str, 0);
2910                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
2911                 sa1->d = str.str;
2912                 sa = sa->next;
2913                 if (tok == ')') {
2914                     /* special case for gcc var args: add an empty
2915                        var arg argument if it is omitted */
2916                     if (sa && sa->type.t && gnu_ext)
2917                         continue;
2918                     else
2919                         break;
2920                 }
2921                 if (tok != ',')
2922                     expect(",");
2923                 next_nomacro();
2924             }
2925             if (sa) {
2926                 tcc_error("macro '%s' used with too few args",
2927                       get_tok_str(s->v, 0));
2928             }
2929
2930             /* now subst each arg */
2931             mstr = macro_arg_subst(nested_list, mstr, args);
2932             /* free memory */
2933             sa = args;
2934             while (sa) {
2935                 sa1 = sa->prev;
2936                 tok_str_free(sa->d);
2937                 sym_free(sa);
2938                 sa = sa1;
2939             }
2940             mstr_allocated = 1;
2941         }
2942         sym_push2(nested_list, s->v, 0, 0);
2943         macro_subst(tok_str, nested_list, mstr, can_read_stream);
2944         /* pop nested defined symbol */
2945         sa1 = *nested_list;
2946         *nested_list = sa1->prev;
2947         sym_free(sa1);
2948         if (mstr_allocated)
2949             tok_str_free(mstr);
2950     }
2951     return 0;
2952 }
2953
2954 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2955    return the resulting string (which must be freed). */
2956 static inline int *macro_twosharps(const int *macro_str)
2957 {
2958     const int *ptr;
2959     int t;
2960     TokenString macro_str1;
2961     CString cstr;
2962     int n, start_of_nosubsts;
2963
2964     /* we search the first '##' */
2965     for(ptr = macro_str;;) {
2966         CValue cval;
2967         TOK_GET(&t, &ptr, &cval);
2968         if (t == TOK_TWOSHARPS)
2969             break;
2970         /* nothing more to do if end of string */
2971         if (t == 0)
2972             return NULL;
2973     }
2974
2975     /* we saw '##', so we need more processing to handle it */
2976     start_of_nosubsts = -1;
2977     tok_str_new(&macro_str1);
2978     for(ptr = macro_str;;) {
2979         TOK_GET(&tok, &ptr, &tokc);
2980         if (tok == 0)
2981             break;
2982         if (tok == TOK_TWOSHARPS)
2983             continue;
2984         if (tok == TOK_NOSUBST && start_of_nosubsts < 0)
2985             start_of_nosubsts = macro_str1.len;
2986         while (*ptr == TOK_TWOSHARPS) {
2987             /* given 'a##b', remove nosubsts preceding 'a' */
2988             if (start_of_nosubsts >= 0)
2989                 macro_str1.len = start_of_nosubsts;
2990             /* given 'a##b', skip '##' */
2991             t = *++ptr;
2992             /* given 'a##b', remove nosubsts preceding 'b' */
2993             while (t == TOK_NOSUBST)
2994                 t = *++ptr;
2995             if (t && t != TOK_TWOSHARPS) {
2996                 CValue cval;
2997                 TOK_GET(&t, &ptr, &cval);
2998                 /* We concatenate the two tokens */
2999                 cstr_new(&cstr);
3000                 if (tok != TOK_PLCHLDR)
3001                     cstr_cat(&cstr, get_tok_str(tok, &tokc));
3002                 n = cstr.size;
3003                 if (t != TOK_PLCHLDR || tok == TOK_PLCHLDR)
3004                     cstr_cat(&cstr, get_tok_str(t, &cval));
3005                 cstr_ccat(&cstr, '\0');
3006
3007                 tcc_open_bf(tcc_state, ":paste:", cstr.size);
3008                 memcpy(file->buffer, cstr.data, cstr.size);
3009                 for (;;) {
3010                     next_nomacro1();
3011                     if (0 == *file->buf_ptr)
3012                         break;
3013                     tok_str_add2(&macro_str1, tok, &tokc);
3014                     tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
3015                         n, cstr.data, (char*)cstr.data + n);
3016                 }
3017                 tcc_close();
3018                 cstr_free(&cstr);
3019             }
3020         }
3021         if (tok != TOK_NOSUBST) {
3022             tok_str_add2(&macro_str1, tok, &tokc);
3023             tok = ' ';
3024             start_of_nosubsts = -1;
3025         }
3026         tok_str_add2(&macro_str1, tok, &tokc);
3027     }
3028     tok_str_add(&macro_str1, 0);
3029     return macro_str1.str;
3030 }
3031
3032
3033 /* do macro substitution of macro_str and add result to
3034    (tok_str,tok_len). 'nested_list' is the list of all macros we got
3035    inside to avoid recursing. */
3036 static void macro_subst(TokenString *tok_str, Sym **nested_list,
3037                         const int *macro_str, struct macro_level ** can_read_stream)
3038 {
3039     Sym *s;
3040     int *macro_str1;
3041     const int *ptr;
3042     int t, ret, spc;
3043     CValue cval;
3044     struct macro_level ml;
3045     int force_blank;
3046
3047     /* first scan for '##' operator handling */
3048     ptr = macro_str;
3049     macro_str1 = macro_twosharps(ptr);
3050
3051     if (macro_str1)
3052         ptr = macro_str1;
3053     spc = 0;
3054     force_blank = 0;
3055
3056     while (1) {
3057         /* NOTE: ptr == NULL can only happen if tokens are read from
3058            file stream due to a macro function call */
3059         if (ptr == NULL)
3060             break;
3061         TOK_GET(&t, &ptr, &cval);
3062         if (t == 0)
3063             break;
3064         if (t == TOK_NOSUBST) {
3065             /* following token has already been subst'd. just copy it on */
3066             tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3067             TOK_GET(&t, &ptr, &cval);
3068             goto no_subst;
3069         }
3070         s = define_find(t);
3071         if (s != NULL) {
3072             /* if nested substitution, do nothing */
3073             if (sym_find2(*nested_list, t)) {
3074                 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
3075                 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3076                 goto no_subst;
3077             }
3078             ml.p = macro_ptr;
3079             if (can_read_stream)
3080                 ml.prev = *can_read_stream, *can_read_stream = &ml;
3081             macro_ptr = (int *)ptr;
3082             tok = t;
3083             ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
3084             ptr = (int *)macro_ptr;
3085             macro_ptr = ml.p;
3086             if (can_read_stream && *can_read_stream == &ml)
3087                 *can_read_stream = ml.prev;
3088             if (ret != 0)
3089                 goto no_subst;
3090             if (parse_flags & PARSE_FLAG_SPACES)
3091                 force_blank = 1;
3092         } else {
3093         no_subst:
3094             if (force_blank) {
3095                 tok_str_add(tok_str, ' ');
3096                 spc = 1;
3097                 force_blank = 0;
3098             }
3099             if (!check_space(t, &spc))
3100                 tok_str_add2(tok_str, t, &cval);
3101         }
3102     }
3103     if (macro_str1)
3104         tok_str_free(macro_str1);
3105 }
3106
3107 /* return next token with macro substitution */
3108 ST_FUNC void next(void)
3109 {
3110     Sym *nested_list, *s;
3111     TokenString str;
3112     struct macro_level *ml;
3113
3114  redo:
3115     if (parse_flags & PARSE_FLAG_SPACES)
3116         next_nomacro_spc();
3117     else
3118         next_nomacro();
3119     if (!macro_ptr) {
3120         /* if not reading from macro substituted string, then try
3121            to substitute macros */
3122         if (tok >= TOK_IDENT &&
3123             (parse_flags & PARSE_FLAG_PREPROCESS)) {
3124             s = define_find(tok);
3125             if (s) {
3126                 /* we have a macro: we try to substitute */
3127                 tok_str_new(&str);
3128                 nested_list = NULL;
3129                 ml = NULL;
3130                 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
3131                     /* substitution done, NOTE: maybe empty */
3132                     tok_str_add(&str, 0);
3133                     macro_ptr = str.str;
3134                     macro_ptr_allocated = str.str;
3135                     goto redo;
3136                 }
3137             }
3138         }
3139     } else {
3140         if (tok == 0) {
3141             /* end of macro or end of unget buffer */
3142             if (unget_buffer_enabled) {
3143                 macro_ptr = unget_saved_macro_ptr;
3144                 unget_buffer_enabled = 0;
3145             } else {
3146                 /* end of macro string: free it */
3147                 tok_str_free(macro_ptr_allocated);
3148                 macro_ptr_allocated = NULL;
3149                 macro_ptr = NULL;
3150             }
3151             goto redo;
3152         } else if (tok == TOK_NOSUBST) {
3153             /* discard preprocessor's nosubst markers */
3154             goto redo;
3155         }
3156     }
3157
3158     /* convert preprocessor tokens into C tokens */
3159     if (tok == TOK_PPNUM &&
3160         (parse_flags & PARSE_FLAG_TOK_NUM)) {
3161         parse_number((char *)tokc.cstr->data);
3162     }
3163 }
3164
3165 /* push back current token and set current token to 'last_tok'. Only
3166    identifier case handled for labels. */
3167 ST_INLN void unget_tok(int last_tok)
3168 {
3169     int i, n;
3170     int *q;
3171     if (unget_buffer_enabled)
3172       {
3173         /* assert(macro_ptr == unget_saved_buffer + 1);
3174            assert(*macro_ptr == 0);  */
3175       }
3176     else
3177       {
3178         unget_saved_macro_ptr = macro_ptr;
3179         unget_buffer_enabled = 1;
3180       }
3181     q = unget_saved_buffer;
3182     macro_ptr = q;
3183     *q++ = tok;
3184     n = tok_ext_size(tok) - 1;
3185     for(i=0;i<n;i++)
3186         *q++ = tokc.tab[i];
3187     *q = 0; /* end of token string */
3188     tok = last_tok;
3189 }
3190
3191
3192 /* better than nothing, but needs extension to handle '-E' option
3193    correctly too */
3194 ST_FUNC void preprocess_init(TCCState *s1)
3195 {
3196     s1->include_stack_ptr = s1->include_stack;
3197     /* XXX: move that before to avoid having to initialize
3198        file->ifdef_stack_ptr ? */
3199     s1->ifdef_stack_ptr = s1->ifdef_stack;
3200     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3201
3202     vtop = vstack - 1;
3203     s1->pack_stack[0] = 0;
3204     s1->pack_stack_ptr = s1->pack_stack;
3205 }
3206
3207 ST_FUNC void preprocess_new(void)
3208 {
3209     int i, c;
3210     const char *p, *r;
3211
3212     /* init isid table */
3213
3214     for(i=CH_EOF;i<256;i++)
3215         isidnum_table[i-CH_EOF] = isid(i) || isnum(i);
3216
3217     /* add all tokens */
3218     if (table_ident) {
3219         tcc_free (table_ident);
3220         table_ident = NULL;
3221     }
3222     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3223
3224     tok_ident = TOK_IDENT;
3225     p = tcc_keywords;
3226     while (*p) {
3227         r = p;
3228         for(;;) {
3229             c = *r++;
3230             if (c == '\0')
3231                 break;
3232         }
3233         tok_alloc(p, r - p - 1);
3234         p = r;
3235     }
3236 }
3237
3238 static void line_macro_output(BufferedFile *f, const char *s, TCCState *s1)
3239 {
3240     switch (s1->Pflag) {
3241         case LINE_MACRO_OUTPUT_FORMAT_STD:
3242             /* "tcc -E -P1" case */
3243             fprintf(s1->ppfp, "# line %d \"%s\"\n", f->line_num, f->filename);
3244             break;
3245
3246         case LINE_MACRO_OUTPUT_FORMAT_NONE:
3247             /* "tcc -E -P" case: don't output a line directive */
3248             break;
3249
3250         case LINE_MACRO_OUTPUT_FORMAT_GCC:
3251         default:
3252             /* "tcc -E" case: a gcc standard by default */
3253             fprintf(s1->ppfp, "# %d \"%s\"%s\n", f->line_num, f->filename, s);
3254             break;
3255     }
3256 }
3257
3258 /* Preprocess the current file */
3259 ST_FUNC int tcc_preprocess(TCCState *s1)
3260 {
3261     BufferedFile *file_ref, **iptr, **iptr_new;
3262     int token_seen, d;
3263     const char *s;
3264
3265     preprocess_init(s1);
3266     ch = file->buf_ptr[0];
3267     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3268     parse_flags = (parse_flags & PARSE_FLAG_ASM_FILE);
3269     parse_flags |= PARSE_FLAG_ASM_COMMENTS | PARSE_FLAG_PREPROCESS |
3270         PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
3271     token_seen = 0;
3272     file->line_ref = 0;
3273     file_ref = NULL;
3274     iptr = s1->include_stack_ptr;
3275
3276     for (;;) {
3277         next();
3278         if (tok == TOK_EOF) {
3279             break;
3280         } else if (file != file_ref) {
3281             if (file_ref)
3282                 line_macro_output(file_ref, "", s1);
3283             goto print_line;
3284         } else if (tok == TOK_LINEFEED) {
3285             if (!token_seen)
3286                 continue;
3287             file->line_ref++;
3288             token_seen = 0;
3289         } else if (!token_seen) {
3290             d = file->line_num - file->line_ref;
3291             if (file != file_ref || d >= 8) {
3292 print_line:
3293                 s = "";
3294                 if (tcc_state->Pflag == LINE_MACRO_OUTPUT_FORMAT_GCC) {
3295                     iptr_new = s1->include_stack_ptr;
3296                     s = iptr_new > iptr ? " 1"
3297                     : iptr_new < iptr ? " 2"
3298                     : iptr_new > s1->include_stack ? " 3"
3299                     : ""
3300                     ;
3301                 }
3302                 line_macro_output(file, s, s1);
3303             } else {
3304                 while (d > 0)
3305                     fputs("\n", s1->ppfp), --d;
3306             }
3307             file->line_ref = (file_ref = file)->line_num;
3308             token_seen = tok != TOK_LINEFEED;
3309             if (!token_seen)
3310                 continue;
3311         }
3312         fputs(get_tok_str(tok, &tokc), s1->ppfp);
3313     }
3314     return 0;
3315 }