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  60 /* outdated -- gr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