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
  22 static const char tcc_keywords[] =
  23 #define DEF(id, str) str "\0"
  24 #include "tcctok.h"
  25 #undef DEF
  26 ;
  27
  28 /* WARNING: the content of this string encodes token numbers */
  29 static char tok_two_chars[] = "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
  30
  31 /* true if isid(c) || isnum(c) */
  32 static unsigned char isidnum_table[256-CH_EOF];
  33
  34
  35 struct macro_level {
  36     struct macro_level *prev;
  37     int *p;
  38 };
  39
  40 static void next_nomacro(void);
  41 static void next_nomacro_spc(void);
  42 static void macro_subst(TokenString *tok_str, Sym **nested_list,
  43                         const int *macro_str, struct macro_level **can_read_stream);
  44
  45
  46 /* allocate a new token */
  47 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
  48 {
  49     TokenSym *ts, **ptable;
  50     int i;
  51
  52     if (tok_ident >= SYM_FIRST_ANOM)
  53         error("memory full");
  54
  55     /* expand token table if needed */
  56     i = tok_ident - TOK_IDENT;
  57     if ((i % TOK_ALLOC_INCR) == 0) {
  58         ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
  59         if (!ptable)
  60             error("memory full");
  61         table_ident = ptable;
  62     }
  63
  64     ts = tcc_malloc(sizeof(TokenSym) + len);
  65     table_ident[i] = ts;
  66     ts->tok = tok_ident++;
  67     ts->sym_define = NULL;
  68     ts->sym_label = NULL;
  69     ts->sym_struct = NULL;
  70     ts->sym_identifier = NULL;
  71     ts->len = len;
  72     ts->hash_next = NULL;
  73     memcpy(ts->str, str, len);
  74     ts->str[len] = '\0';
  75     *pts = ts;
  76     return ts;
  77 }
  78
  79 #define TOK_HASH_INIT 1
  80 #define TOK_HASH_FUNC(h, c) ((h) * 263 + (c))
  81
  82 /* find a token and add it if not found */
  83 static TokenSym *tok_alloc(const char *str, int len)
  84 {
  85     TokenSym *ts, **pts;
  86     int i;
  87     unsigned int h;
  88
  89     h = TOK_HASH_INIT;
  90     for(i=0;i<len;i++)
  91         h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
  92     h &= (TOK_HASH_SIZE - 1);
  93
  94     pts = &hash_ident[h];
  95     for(;;) {
  96         ts = *pts;
  97         if (!ts)
  98             break;
  99         if (ts->len == len && !memcmp(ts->str, str, len))
 100             return ts;
 101         pts = &(ts->hash_next);
 102     }
 103     return tok_alloc_new(pts, str, len);
 104 }
 105
 106 /* XXX: buffer overflow */
 107 /* XXX: float tokens */
 108 char *get_tok_str(int v, CValue *cv)
 109 {
 110     static char buf[STRING_MAX_SIZE + 1];
 111     static CString cstr_buf;
 112     CString *cstr;
 113     unsigned char *q;
 114     char *p;
 115     int i, len;
 116
 117     /* NOTE: to go faster, we give a fixed buffer for small strings */
 118     cstr_reset(&cstr_buf);
 119     cstr_buf.data = buf;
 120     cstr_buf.size_allocated = sizeof(buf);
 121     p = buf;
 122
 123     switch(v) {
 124     case TOK_CINT:
 125     case TOK_CUINT:
 126         /* XXX: not quite exact, but only useful for testing */
 127         sprintf(p, "%u", cv->ui);
 128         break;
 129     case TOK_CLLONG:
 130     case TOK_CULLONG:
 131         /* XXX: not quite exact, but only useful for testing  */
 132         sprintf(p, "%Lu", cv->ull);
 133         break;
 134     case TOK_LCHAR:
 135         cstr_ccat(&cstr_buf, 'L');
 136     case TOK_CCHAR:
 137         cstr_ccat(&cstr_buf, '\'');
 138         add_char(&cstr_buf, cv->i);
 139         cstr_ccat(&cstr_buf, '\'');
 140         cstr_ccat(&cstr_buf, '\0');
 141         break;
 142     case TOK_PPNUM:
 143         cstr = cv->cstr;
 144         len = cstr->size - 1;
 145         for(i=0;i<len;i++)
 146             add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 147         cstr_ccat(&cstr_buf, '\0');
 148         break;
 149     case TOK_LSTR:
 150         cstr_ccat(&cstr_buf, 'L');
 151     case TOK_STR:
 152         cstr = cv->cstr;
 153         cstr_ccat(&cstr_buf, '\"');
 154         if (v == TOK_STR) {
 155             len = cstr->size - 1;
 156             for(i=0;i<len;i++)
 157                 add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 158         } else {
 159             len = (cstr->size / sizeof(nwchar_t)) - 1;
 160             for(i=0;i<len;i++)
 161                 add_char(&cstr_buf, ((nwchar_t *)cstr->data)[i]);
 162         }
 163         cstr_ccat(&cstr_buf, '\"');
 164         cstr_ccat(&cstr_buf, '\0');
 165         break;
 166     case TOK_LT:
 167         v = '<';
 168         goto addv;
 169     case TOK_GT:
 170         v = '>';
 171         goto addv;
 172     case TOK_DOTS:
 173         return strcpy(p, "...");
 174     case TOK_A_SHL:
 175         return strcpy(p, "<<=");
 176     case TOK_A_SAR:
 177         return strcpy(p, ">>=");
 178     default:
 179         if (v < TOK_IDENT) {
 180             /* search in two bytes table */
 181             q = tok_two_chars;
 182             while (*q) {
 183                 if (q[2] == v) {
 184                     *p++ = q[0];
 185                     *p++ = q[1];
 186                     *p = '\0';
 187                     return buf;
 188                 }
 189                 q += 3;
 190             }
 191         addv:
 192             *p++ = v;
 193             *p = '\0';
 194         } else if (v < tok_ident) {
 195             return table_ident[v - TOK_IDENT]->str;
 196         } else if (v >= SYM_FIRST_ANOM) {
 197             /* special name for anonymous symbol */
 198             sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
 199         } else {
 200             /* should never happen */
 201             return NULL;
 202         }
 203         break;
 204     }
 205     return cstr_buf.data;
 206 }
 207
 208 /* fill input buffer and peek next char */
 209 static int tcc_peekc_slow(BufferedFile *bf)
 210 {
 211     int len;
 212     /* only tries to read if really end of buffer */
 213     if (bf->buf_ptr >= bf->buf_end) {
 214         if (bf->fd != -1) {
 215 #if defined(PARSE_DEBUG)
 216             len = 8;
 217 #else
 218             len = IO_BUF_SIZE;
 219 #endif
 220             len = read(bf->fd, bf->buffer, len);
 221             if (len < 0)
 222                 len = 0;
 223         } else {
 224             len = 0;
 225         }
 226         total_bytes += len;
 227         bf->buf_ptr = bf->buffer;
 228         bf->buf_end = bf->buffer + len;
 229         *bf->buf_end = CH_EOB;
 230     }
 231     if (bf->buf_ptr < bf->buf_end) {
 232         return bf->buf_ptr[0];
 233     } else {
 234         bf->buf_ptr = bf->buf_end;
 235         return CH_EOF;
 236     }
 237 }
 238
 239 /* return the current character, handling end of block if necessary
 240    (but not stray) */
 241 static int handle_eob(void)
 242 {
 243     return tcc_peekc_slow(file);
 244 }
 245
 246 /* read next char from current input file and handle end of input buffer */
 247 static inline void inp(void)
 248 {
 249     ch = *(++(file->buf_ptr));
 250     /* end of buffer/file handling */
 251     if (ch == CH_EOB)
 252         ch = handle_eob();
 253 }
 254
 255 /* handle '\[\r]\n' */
 256 static int handle_stray_noerror(void)
 257 {
 258     while (ch == '\\') {
 259         inp();
 260         if (ch == '\n') {
 261             file->line_num++;
 262             inp();
 263         } else if (ch == '\r') {
 264             inp();
 265             if (ch != '\n')
 266                 goto fail;
 267             file->line_num++;
 268             inp();
 269         } else {
 270         fail:
 271             return 1;
 272         }
 273     }
 274     return 0;
 275 }
 276
 277 static void handle_stray(void)
 278 {
 279     if (handle_stray_noerror())
 280         error("stray '\\' in program");
 281 }
 282
 283 /* skip the stray and handle the \\n case. Output an error if
 284    incorrect char after the stray */
 285 static int handle_stray1(uint8_t *p)
 286 {
 287     int c;
 288
 289     if (p >= file->buf_end) {
 290         file->buf_ptr = p;
 291         c = handle_eob();
 292         p = file->buf_ptr;
 293         if (c == '\\')
 294             goto parse_stray;
 295     } else {
 296     parse_stray:
 297         file->buf_ptr = p;
 298         ch = *p;
 299         handle_stray();
 300         p = file->buf_ptr;
 301         c = *p;
 302     }
 303     return c;
 304 }
 305
 306 /* handle just the EOB case, but not stray */
 307 #define PEEKC_EOB(c, p)\
 308 {\
 309     p++;\
 310     c = *p;\
 311     if (c == '\\') {\
 312         file->buf_ptr = p;\
 313         c = handle_eob();\
 314         p = file->buf_ptr;\
 315     }\
 316 }
 317
 318 /* handle the complicated stray case */
 319 #define PEEKC(c, p)\
 320 {\
 321     p++;\
 322     c = *p;\
 323     if (c == '\\') {\
 324         c = handle_stray1(p);\
 325         p = file->buf_ptr;\
 326     }\
 327 }
 328
 329 /* input with '\[\r]\n' handling. Note that this function cannot
 330    handle other characters after '\', so you cannot call it inside
 331    strings or comments */
 332 static void minp(void)
 333 {
 334     inp();
 335     if (ch == '\\')
 336         handle_stray();
 337 }
 338
 339
 340 /* single line C++ comments */
 341 static uint8_t *parse_line_comment(uint8_t *p)
 342 {
 343     int c;
 344
 345     p++;
 346     for(;;) {
 347         c = *p;
 348     redo:
 349         if (c == '\n' || c == CH_EOF) {
 350             break;
 351         } else if (c == '\\') {
 352             file->buf_ptr = p;
 353             c = handle_eob();
 354             p = file->buf_ptr;
 355             if (c == '\\') {
 356                 PEEKC_EOB(c, p);
 357                 if (c == '\n') {
 358                     file->line_num++;
 359                     PEEKC_EOB(c, p);
 360                 } else if (c == '\r') {
 361                     PEEKC_EOB(c, p);
 362                     if (c == '\n') {
 363                         file->line_num++;
 364                         PEEKC_EOB(c, p);
 365                     }
 366                 }
 367             } else {
 368                 goto redo;
 369             }
 370         } else {
 371             p++;
 372         }
 373     }
 374     return p;
 375 }
 376
 377 /* C comments */
 378 static uint8_t *parse_comment(uint8_t *p)
 379 {
 380     int c;
 381
 382     p++;
 383     for(;;) {
 384         /* fast skip loop */
 385         for(;;) {
 386             c = *p;
 387             if (c == '\n' || c == '*' || c == '\\')
 388                 break;
 389             p++;
 390             c = *p;
 391             if (c == '\n' || c == '*' || c == '\\')
 392                 break;
 393             p++;
 394         }
 395         /* now we can handle all the cases */
 396         if (c == '\n') {
 397             file->line_num++;
 398             p++;
 399         } else if (c == '*') {
 400             p++;
 401             for(;;) {
 402                 c = *p;
 403                 if (c == '*') {
 404                     p++;
 405                 } else if (c == '/') {
 406                     goto end_of_comment;
 407                 } else if (c == '\\') {
 408                     file->buf_ptr = p;
 409                     c = handle_eob();
 410                     p = file->buf_ptr;
 411                     if (c == '\\') {
 412                         /* skip '\[\r]\n', otherwise just skip the stray */
 413                         while (c == '\\') {
 414                             PEEKC_EOB(c, p);
 415                             if (c == '\n') {
 416                                 file->line_num++;
 417                                 PEEKC_EOB(c, p);
 418                             } else if (c == '\r') {
 419                                 PEEKC_EOB(c, p);
 420                                 if (c == '\n') {
 421                                     file->line_num++;
 422                                     PEEKC_EOB(c, p);
 423                                 }
 424                             } else {
 425                                 goto after_star;
 426                             }
 427                         }
 428                     }
 429                 } else {
 430                     break;
 431                 }
 432             }
 433         after_star: ;
 434         } else {
 435             /* stray, eob or eof */
 436             file->buf_ptr = p;
 437             c = handle_eob();
 438             p = file->buf_ptr;
 439             if (c == CH_EOF) {
 440                 error("unexpected end of file in comment");
 441             } else if (c == '\\') {
 442                 p++;
 443             }
 444         }
 445     }
 446  end_of_comment:
 447     p++;
 448     return p;
 449 }
 450
 451 #define cinp minp
 452
 453 static inline void skip_spaces(void)
 454 {
 455     while (is_space(ch))
 456         cinp();
 457 }
 458
 459 static inline int check_space(int t, int *spc)
 460 {
 461     if (is_space(t)) {
 462         if (*spc)
 463             return 1;
 464         *spc = 1;
 465     } else
 466         *spc = 0;
 467     return 0;
 468 }
 469
 470 /* parse a string without interpreting escapes */
 471 static uint8_t *parse_pp_string(uint8_t *p,
 472                                 int sep, CString *str)
 473 {
 474     int c;
 475     p++;
 476     for(;;) {
 477         c = *p;
 478         if (c == sep) {
 479             break;
 480         } else if (c == '\\') {
 481             file->buf_ptr = p;
 482             c = handle_eob();
 483             p = file->buf_ptr;
 484             if (c == CH_EOF) {
 485             unterminated_string:
 486                 /* XXX: indicate line number of start of string */
 487                 error("missing terminating %c character", sep);
 488             } else if (c == '\\') {
 489                 /* escape : just skip \[\r]\n */
 490                 PEEKC_EOB(c, p);
 491                 if (c == '\n') {
 492                     file->line_num++;
 493                     p++;
 494                 } else if (c == '\r') {
 495                     PEEKC_EOB(c, p);
 496                     if (c != '\n')
 497                         expect("'\n' after '\r'");
 498                     file->line_num++;
 499                     p++;
 500                 } else if (c == CH_EOF) {
 501                     goto unterminated_string;
 502                 } else {
 503                     if (str) {
 504                         cstr_ccat(str, '\\');
 505                         cstr_ccat(str, c);
 506                     }
 507                     p++;
 508                 }
 509             }
 510         } else if (c == '\n') {
 511             file->line_num++;
 512             goto add_char;
 513         } else if (c == '\r') {
 514             PEEKC_EOB(c, p);
 515             if (c != '\n') {
 516                 if (str)
 517                     cstr_ccat(str, '\r');
 518             } else {
 519                 file->line_num++;
 520                 goto add_char;
 521             }
 522         } else {
 523         add_char:
 524             if (str)
 525                 cstr_ccat(str, c);
 526             p++;
 527         }
 528     }
 529     p++;
 530     return p;
 531 }
 532
 533 /* skip block of text until #else, #elif or #endif. skip also pairs of
 534    #if/#endif */
 535 void preprocess_skip(void)
 536 {
 537     int a, start_of_line, c, in_warn_or_error;
 538     uint8_t *p;
 539
 540     p = file->buf_ptr;
 541     a = 0;
 542 redo_start:
 543     start_of_line = 1;
 544     in_warn_or_error = 0;
 545     for(;;) {
 546     redo_no_start:
 547         c = *p;
 548         switch(c) {
 549         case ' ':
 550         case '\t':
 551         case '\f':
 552         case '\v':
 553         case '\r':
 554             p++;
 555             goto redo_no_start;
 556         case '\n':
 557             file->line_num++;
 558             p++;
 559             goto redo_start;
 560         case '\\':
 561             file->buf_ptr = p;
 562             c = handle_eob();
 563             if (c == CH_EOF) {
 564                 expect("#endif");
 565             } else if (c == '\\') {
 566                 ch = file->buf_ptr[0];
 567                 handle_stray_noerror();
 568             }
 569             p = file->buf_ptr;
 570             goto redo_no_start;
 571         /* skip strings */
 572         case '\"':
 573         case '\'':
 574             if (in_warn_or_error)
 575                 goto _default;
 576             p = parse_pp_string(p, c, NULL);
 577             break;
 578         /* skip comments */
 579         case '/':
 580             if (in_warn_or_error)
 581                 goto _default;
 582             file->buf_ptr = p;
 583             ch = *p;
 584             minp();
 585             p = file->buf_ptr;
 586             if (ch == '*') {
 587                 p = parse_comment(p);
 588             } else if (ch == '/') {
 589                 p = parse_line_comment(p);
 590             }
 591             break;
 592         case '#':
 593             p++;
 594             if (start_of_line) {
 595                 file->buf_ptr = p;
 596                 next_nomacro();
 597                 p = file->buf_ptr;
 598                 if (a == 0 &&
 599                     (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
 600                     goto the_end;
 601                 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
 602                     a++;
 603                 else if (tok == TOK_ENDIF)
 604                     a--;
 605                 else if( tok == TOK_ERROR || tok == TOK_WARNING)
 606                     in_warn_or_error = 1;
 607             }
 608             break;
 609 _default:
 610         default:
 611             p++;
 612             break;
 613         }
 614         start_of_line = 0;
 615     }
 616  the_end: ;
 617     file->buf_ptr = p;
 618 }
 619
 620 /* ParseState handling */
 621
 622 /* XXX: currently, no include file info is stored. Thus, we cannot display
 623    accurate messages if the function or data definition spans multiple
 624    files */
 625
 626 /* save current parse state in 's' */
 627 void save_parse_state(ParseState *s)
 628 {
 629     s->line_num = file->line_num;
 630     s->macro_ptr = macro_ptr;
 631     s->tok = tok;
 632     s->tokc = tokc;
 633 }
 634
 635 /* restore parse state from 's' */
 636 void restore_parse_state(ParseState *s)
 637 {
 638     file->line_num = s->line_num;
 639     macro_ptr = s->macro_ptr;
 640     tok = s->tok;
 641     tokc = s->tokc;
 642 }
 643
 644 /* return the number of additional 'ints' necessary to store the
 645    token */
 646 static inline int tok_ext_size(int t)
 647 {
 648     switch(t) {
 649         /* 4 bytes */
 650     case TOK_CINT:
 651     case TOK_CUINT:
 652     case TOK_CCHAR:
 653     case TOK_LCHAR:
 654     case TOK_CFLOAT:
 655     case TOK_LINENUM:
 656         return 1;
 657     case TOK_STR:
 658     case TOK_LSTR:
 659     case TOK_PPNUM:
 660         error("unsupported token");
 661         return 1;
 662     case TOK_CDOUBLE:
 663     case TOK_CLLONG:
 664     case TOK_CULLONG:
 665         return 2;
 666     case TOK_CLDOUBLE:
 667         return LDOUBLE_SIZE / 4;
 668     default:
 669         return 0;
 670     }
 671 }
 672
 673 /* token string handling */
 674
 675 static inline void tok_str_new(TokenString *s)
 676 {
 677     s->str = NULL;
 678     s->len = 0;
 679     s->allocated_len = 0;
 680     s->last_line_num = -1;
 681 }
 682
 683 static void tok_str_free(int *str)
 684 {
 685     tcc_free(str);
 686 }
 687
 688 static int *tok_str_realloc(TokenString *s)
 689 {
 690     int *str, len;
 691
 692     if (s->allocated_len == 0) {
 693         len = 8;
 694     } else {
 695         len = s->allocated_len * 2;
 696     }
 697     str = tcc_realloc(s->str, len * sizeof(int));
 698     if (!str)
 699         error("memory full");
 700     s->allocated_len = len;
 701     s->str = str;
 702     return str;
 703 }
 704
 705 static void tok_str_add(TokenString *s, int t)
 706 {
 707     int len, *str;
 708
 709     len = s->len;
 710     str = s->str;
 711     if (len >= s->allocated_len)
 712         str = tok_str_realloc(s);
 713     str[len++] = t;
 714     s->len = len;
 715 }
 716
 717 static void tok_str_add2(TokenString *s, int t, CValue *cv)
 718 {
 719     int len, *str;
 720
 721     len = s->len;
 722     str = s->str;
 723
 724     /* allocate space for worst case */
 725     if (len + TOK_MAX_SIZE > s->allocated_len)
 726         str = tok_str_realloc(s);
 727     str[len++] = t;
 728     switch(t) {
 729     case TOK_CINT:
 730     case TOK_CUINT:
 731     case TOK_CCHAR:
 732     case TOK_LCHAR:
 733     case TOK_CFLOAT:
 734     case TOK_LINENUM:
 735         str[len++] = cv->tab[0];
 736         break;
 737     case TOK_PPNUM:
 738     case TOK_STR:
 739     case TOK_LSTR:
 740         {
 741             int nb_words;
 742             CString *cstr;
 743
 744             nb_words = (sizeof(CString) + cv->cstr->size + 3) >> 2;
 745             while ((len + nb_words) > s->allocated_len)
 746                 str = tok_str_realloc(s);
 747             cstr = (CString *)(str + len);
 748             cstr->data = NULL;
 749             cstr->size = cv->cstr->size;
 750             cstr->data_allocated = NULL;
 751             cstr->size_allocated = cstr->size;
 752             memcpy((char *)cstr + sizeof(CString),
 753                    cv->cstr->data, cstr->size);
 754             len += nb_words;
 755         }
 756         break;
 757     case TOK_CDOUBLE:
 758     case TOK_CLLONG:
 759     case TOK_CULLONG:
 760 #if LDOUBLE_SIZE == 8
 761     case TOK_CLDOUBLE:
 762 #endif
 763         str[len++] = cv->tab[0];
 764         str[len++] = cv->tab[1];
 765         break;
 766 #if LDOUBLE_SIZE == 12
 767     case TOK_CLDOUBLE:
 768         str[len++] = cv->tab[0];
 769         str[len++] = cv->tab[1];
 770         str[len++] = cv->tab[2];
 771 #elif LDOUBLE_SIZE == 16
 772     case TOK_CLDOUBLE:
 773         str[len++] = cv->tab[0];
 774         str[len++] = cv->tab[1];
 775         str[len++] = cv->tab[2];
 776         str[len++] = cv->tab[3];
 777 #elif LDOUBLE_SIZE != 8
 778 #error add long double size support
 779 #endif
 780         break;
 781     default:
 782         break;
 783     }
 784     s->len = len;
 785 }
 786
 787 /* add the current parse token in token string 's' */
 788 static void tok_str_add_tok(TokenString *s)
 789 {
 790     CValue cval;
 791
 792     /* save line number info */
 793     if (file->line_num != s->last_line_num) {
 794         s->last_line_num = file->line_num;
 795         cval.i = s->last_line_num;
 796         tok_str_add2(s, TOK_LINENUM, &cval);
 797     }
 798     tok_str_add2(s, tok, &tokc);
 799 }
 800
 801 #if LDOUBLE_SIZE == 16
 802 #define LDOUBLE_GET(p, cv)                      \
 803         cv.tab[0] = p[0];                       \
 804         cv.tab[1] = p[1];                       \
 805         cv.tab[2] = p[2];                       \
 806         cv.tab[3] = p[3];
 807 #elif LDOUBLE_SIZE == 12
 808 #define LDOUBLE_GET(p, cv)                      \
 809         cv.tab[0] = p[0];                       \
 810         cv.tab[1] = p[1];                       \
 811         cv.tab[2] = p[2];
 812 #elif LDOUBLE_SIZE == 8
 813 #define LDOUBLE_GET(p, cv)                      \
 814         cv.tab[0] = p[0];                       \
 815         cv.tab[1] = p[1];
 816 #else
 817 #error add long double size support
 818 #endif
 819
 820
 821 /* get a token from an integer array and increment pointer
 822    accordingly. we code it as a macro to avoid pointer aliasing. */
 823 #define TOK_GET(t, p, cv)                       \
 824 {                                               \
 825     t = *p++;                                   \
 826     switch(t) {                                 \
 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         cv.tab[0] = *p++;                       \
 834         break;                                  \
 835     case TOK_STR:                               \
 836     case TOK_LSTR:                              \
 837     case TOK_PPNUM:                             \
 838         cv.cstr = (CString *)p;                 \
 839         cv.cstr->data = (char *)p + sizeof(CString);\
 840         p += (sizeof(CString) + cv.cstr->size + 3) >> 2;\
 841         break;                                  \
 842     case TOK_CDOUBLE:                           \
 843     case TOK_CLLONG:                            \
 844     case TOK_CULLONG:                           \
 845         cv.tab[0] = p[0];                       \
 846         cv.tab[1] = p[1];                       \
 847         p += 2;                                 \
 848         break;                                  \
 849     case TOK_CLDOUBLE:                          \
 850         LDOUBLE_GET(p, cv);                     \
 851         p += LDOUBLE_SIZE / 4;                  \
 852         break;                                  \
 853     default:                                    \
 854         break;                                  \
 855     }                                           \
 856 }
 857
 858 /* defines handling */
 859 static inline void define_push(int v, int macro_type, int *str, Sym *first_arg)
 860 {
 861     Sym *s;
 862
 863     s = sym_push2(&define_stack, v, macro_type, 0);
 864     s->d = str;
 865     s->next = first_arg;
 866     table_ident[v - TOK_IDENT]->sym_define = s;
 867 }
 868
 869 /* undefined a define symbol. Its name is just set to zero */
 870 static void define_undef(Sym *s)
 871 {
 872     int v;
 873     v = s->v;
 874     if (v >= TOK_IDENT && v < tok_ident)
 875         table_ident[v - TOK_IDENT]->sym_define = NULL;
 876     s->v = 0;
 877 }
 878
 879 static inline Sym *define_find(int v)
 880 {
 881     v -= TOK_IDENT;
 882     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
 883         return NULL;
 884     return table_ident[v]->sym_define;
 885 }
 886
 887 /* free define stack until top reaches 'b' */
 888 static void free_defines(Sym *b)
 889 {
 890     Sym *top, *top1;
 891     int v;
 892
 893     top = define_stack;
 894     while (top != b) {
 895         top1 = top->prev;
 896         /* do not free args or predefined defines */
 897         if (top->d)
 898             tok_str_free(top->d);
 899         v = top->v;
 900         if (v >= TOK_IDENT && v < tok_ident)
 901             table_ident[v - TOK_IDENT]->sym_define = NULL;
 902         sym_free(top);
 903         top = top1;
 904     }
 905     define_stack = b;
 906 }
 907
 908 /* label lookup */
 909 static Sym *label_find(int v)
 910 {
 911     v -= TOK_IDENT;
 912     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
 913         return NULL;
 914     return table_ident[v]->sym_label;
 915 }
 916
 917 static Sym *label_push(Sym **ptop, int v, int flags)
 918 {
 919     Sym *s, **ps;
 920     s = sym_push2(ptop, v, 0, 0);
 921     s->r = flags;
 922     ps = &table_ident[v - TOK_IDENT]->sym_label;
 923     if (ptop == &global_label_stack) {
 924         /* modify the top most local identifier, so that
 925            sym_identifier will point to 's' when popped */
 926         while (*ps != NULL)
 927             ps = &(*ps)->prev_tok;
 928     }
 929     s->prev_tok = *ps;
 930     *ps = s;
 931     return s;
 932 }
 933
 934 /* pop labels until element last is reached. Look if any labels are
 935    undefined. Define symbols if '&&label' was used. */
 936 static void label_pop(Sym **ptop, Sym *slast)
 937 {
 938     Sym *s, *s1;
 939     for(s = *ptop; s != slast; s = s1) {
 940         s1 = s->prev;
 941         if (s->r == LABEL_DECLARED) {
 942             warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
 943         } else if (s->r == LABEL_FORWARD) {
 944                 error("label '%s' used but not defined",
 945                       get_tok_str(s->v, NULL));
 946         } else {
 947             if (s->c) {
 948                 /* define corresponding symbol. A size of
 949                    1 is put. */
 950                 put_extern_sym(s, cur_text_section, (long)s->next, 1);
 951             }
 952         }
 953         /* remove label */
 954         table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
 955         sym_free(s);
 956     }
 957     *ptop = slast;
 958 }
 959
 960 /* eval an expression for #if/#elif */
 961 static int expr_preprocess(void)
 962 {
 963     int c, t;
 964     TokenString str;
 965
 966     tok_str_new(&str);
 967     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
 968         next(); /* do macro subst */
 969         if (tok == TOK_DEFINED) {
 970             next_nomacro();
 971             t = tok;
 972             if (t == '(')
 973                 next_nomacro();
 974             c = define_find(tok) != 0;
 975             if (t == '(')
 976                 next_nomacro();
 977             tok = TOK_CINT;
 978             tokc.i = c;
 979         } else if (tok >= TOK_IDENT) {
 980             /* if undefined macro */
 981             tok = TOK_CINT;
 982             tokc.i = 0;
 983         }
 984         tok_str_add_tok(&str);
 985     }
 986     tok_str_add(&str, -1); /* simulate end of file */
 987     tok_str_add(&str, 0);
 988     /* now evaluate C constant expression */
 989     macro_ptr = str.str;
 990     next();
 991     c = expr_const();
 992     macro_ptr = NULL;
 993     tok_str_free(str.str);
 994     return c != 0;
 995 }
 996
 997 #if defined(PARSE_DEBUG) || defined(PP_DEBUG)
 998 static void tok_print(int *str)
 999 {
1000     int t;
1001     CValue cval;
1002
1003     printf("<");
1004     while (1) {
1005         TOK_GET(t, str, cval);
1006         if (!t)
1007             break;
1008         printf("%s", get_tok_str(t, &cval));
1009     }
1010     printf(">\n");
1011 }
1012 #endif
1013
1014 /* parse after #define */
1015 static void parse_define(void)
1016 {
1017     Sym *s, *first, **ps;
1018     int v, t, varg, is_vaargs, spc;
1019     TokenString str;
1020
1021     v = tok;
1022     if (v < TOK_IDENT)
1023         error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1024     /* XXX: should check if same macro (ANSI) */
1025     first = NULL;
1026     t = MACRO_OBJ;
1027     /* '(' must be just after macro definition for MACRO_FUNC */
1028     next_nomacro_spc();
1029     if (tok == '(') {
1030         next_nomacro();
1031         ps = &first;
1032         while (tok != ')') {
1033             varg = tok;
1034             next_nomacro();
1035             is_vaargs = 0;
1036             if (varg == TOK_DOTS) {
1037                 varg = TOK___VA_ARGS__;
1038                 is_vaargs = 1;
1039             } else if (tok == TOK_DOTS && gnu_ext) {
1040                 is_vaargs = 1;
1041                 next_nomacro();
1042             }
1043             if (varg < TOK_IDENT)
1044                 error("badly punctuated parameter list");
1045             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1046             *ps = s;
1047             ps = &s->next;
1048             if (tok != ',')
1049                 break;
1050             next_nomacro();
1051         }
1052         if (tok == ')')
1053             next_nomacro_spc();
1054         t = MACRO_FUNC;
1055     }
1056     tok_str_new(&str);
1057     spc = 2;
1058     /* EOF testing necessary for '-D' handling */
1059     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1060         /* remove spaces around ## and after '#' */
1061         if (TOK_TWOSHARPS == tok) {
1062             if (1 == spc)
1063                 --str.len;
1064             spc = 2;
1065         } else if ('#' == tok) {
1066             spc = 2;
1067         } else if (check_space(tok, &spc)) {
1068             goto skip;
1069         }
1070         tok_str_add2(&str, tok, &tokc);
1071     skip:
1072         next_nomacro_spc();
1073     }
1074     if (spc == 1)
1075         --str.len; /* remove trailing space */
1076     tok_str_add(&str, 0);
1077 #ifdef PP_DEBUG
1078     printf("define %s %d: ", get_tok_str(v, NULL), t);
1079     tok_print(str.str);
1080 #endif
1081     define_push(v, t, str.str, first);
1082 }
1083
1084 static inline int hash_cached_include(int type, const char *filename)
1085 {
1086     const unsigned char *s;
1087     unsigned int h;
1088
1089     h = TOK_HASH_INIT;
1090     h = TOK_HASH_FUNC(h, type);
1091     s = filename;
1092     while (*s) {
1093         h = TOK_HASH_FUNC(h, *s);
1094         s++;
1095     }
1096     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1097     return h;
1098 }
1099
1100 /* XXX: use a token or a hash table to accelerate matching ? */
1101 static CachedInclude *search_cached_include(TCCState *s1,
1102                                             int type, const char *filename)
1103 {
1104     CachedInclude *e;
1105     int i, h;
1106     h = hash_cached_include(type, filename);
1107     i = s1->cached_includes_hash[h];
1108     for(;;) {
1109         if (i == 0)
1110             break;
1111         e = s1->cached_includes[i - 1];
1112         if (e->type == type && !PATHCMP(e->filename, filename))
1113             return e;
1114         i = e->hash_next;
1115     }
1116     return NULL;
1117 }
1118
1119 static inline void add_cached_include(TCCState *s1, int type,
1120                                       const char *filename, int ifndef_macro)
1121 {
1122     CachedInclude *e;
1123     int h;
1124
1125     if (search_cached_include(s1, type, filename))
1126         return;
1127 #ifdef INC_DEBUG
1128     printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1129 #endif
1130     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1131     if (!e)
1132         return;
1133     e->type = type;
1134     strcpy(e->filename, filename);
1135     e->ifndef_macro = ifndef_macro;
1136     dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1137     /* add in hash table */
1138     h = hash_cached_include(type, filename);
1139     e->hash_next = s1->cached_includes_hash[h];
1140     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1141 }
1142
1143 static void pragma_parse(TCCState *s1)
1144 {
1145     int val;
1146
1147     next();
1148     if (tok == TOK_pack) {
1149         /*
1150           This may be:
1151           #pragma pack(1) // set
1152           #pragma pack() // reset to default
1153           #pragma pack(push,1) // push & set
1154           #pragma pack(pop) // restore previous
1155         */
1156         next();
1157         skip('(');
1158         if (tok == TOK_ASM_pop) {
1159             next();
1160             if (s1->pack_stack_ptr <= s1->pack_stack) {
1161             stk_error:
1162                 error("out of pack stack");
1163             }
1164             s1->pack_stack_ptr--;
1165         } else {
1166             val = 0;
1167             if (tok != ')') {
1168                 if (tok == TOK_ASM_push) {
1169                     next();
1170                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1171                         goto stk_error;
1172                     s1->pack_stack_ptr++;
1173                     skip(',');
1174                 }
1175                 if (tok != TOK_CINT) {
1176                 pack_error:
1177                     error("invalid pack pragma");
1178                 }
1179                 val = tokc.i;
1180                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1181                     goto pack_error;
1182                 next();
1183             }
1184             *s1->pack_stack_ptr = val;
1185             skip(')');
1186         }
1187     }
1188 }
1189
1190 /* is_bof is true if first non space token at beginning of file */
1191 static void preprocess(int is_bof)
1192 {
1193     TCCState *s1 = tcc_state;
1194     int i, c, n, saved_parse_flags;
1195     char buf[1024], *q;
1196     Sym *s;
1197
1198     saved_parse_flags = parse_flags;
1199     parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM |
1200         PARSE_FLAG_LINEFEED;
1201     next_nomacro();
1202  redo:
1203     switch(tok) {
1204     case TOK_DEFINE:
1205         next_nomacro();
1206         parse_define();
1207         break;
1208     case TOK_UNDEF:
1209         next_nomacro();
1210         s = define_find(tok);
1211         /* undefine symbol by putting an invalid name */
1212         if (s)
1213             define_undef(s);
1214         break;
1215     case TOK_INCLUDE:
1216     case TOK_INCLUDE_NEXT:
1217         ch = file->buf_ptr[0];
1218         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1219         skip_spaces();
1220         if (ch == '<') {
1221             c = '>';
1222             goto read_name;
1223         } else if (ch == '\"') {
1224             c = ch;
1225         read_name:
1226             inp();
1227             q = buf;
1228             while (ch != c && ch != '\n' && ch != CH_EOF) {
1229                 if ((q - buf) < sizeof(buf) - 1)
1230                     *q++ = ch;
1231                 if (ch == '\\') {
1232                     if (handle_stray_noerror() == 0)
1233                         --q;
1234                 } else
1235                     inp();
1236             }
1237             *q = '\0';
1238             minp();
1239 #if 0
1240             /* eat all spaces and comments after include */
1241             /* XXX: slightly incorrect */
1242             while (ch1 != '\n' && ch1 != CH_EOF)
1243                 inp();
1244 #endif
1245         } else {
1246             /* computed #include : either we have only strings or
1247                we have anything enclosed in '<>' */
1248             next();
1249             buf[0] = '\0';
1250             if (tok == TOK_STR) {
1251                 while (tok != TOK_LINEFEED) {
1252                     if (tok != TOK_STR) {
1253                     include_syntax:
1254                         error("'#include' expects \"FILENAME\" or <FILENAME>");
1255                     }
1256                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1257                     next();
1258                 }
1259                 c = '\"';
1260             } else {
1261                 int len;
1262                 while (tok != TOK_LINEFEED) {
1263                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1264                     next();
1265                 }
1266                 len = strlen(buf);
1267                 /* check syntax and remove '<>' */
1268                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1269                     goto include_syntax;
1270                 memmove(buf, buf + 1, len - 2);
1271                 buf[len - 2] = '\0';
1272                 c = '>';
1273             }
1274         }
1275
1276         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1277             error("#include recursion too deep");
1278
1279         n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1280         for (i = -2; i < n; ++i) {
1281             char buf1[sizeof file->filename];
1282             BufferedFile *f;
1283             CachedInclude *e;
1284             const char *path;
1285             int size;
1286
1287             if (i == -2) {
1288                 /* check absolute include path */
1289                 if (!IS_ABSPATH(buf))
1290                     continue;
1291                 buf1[0] = 0;
1292
1293             } else if (i == -1) {
1294                 /* search in current dir if "header.h" */
1295                 if (c != '\"')
1296                     continue;
1297                 size = tcc_basename(file->filename) - file->filename;
1298                 memcpy(buf1, file->filename, size);
1299                 buf1[size] = '\0';
1300
1301             } else {
1302                 /* search in all the include paths */
1303                 if (i < s1->nb_include_paths)
1304                     path = s1->include_paths[i];
1305                 else
1306                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1307                 pstrcpy(buf1, sizeof(buf1), path);
1308                 pstrcat(buf1, sizeof(buf1), "/");
1309             }
1310
1311             pstrcat(buf1, sizeof(buf1), buf);
1312
1313             e = search_cached_include(s1, c, buf1);
1314             if (e && define_find(e->ifndef_macro)) {
1315                 /* no need to parse the include because the 'ifndef macro'
1316                    is defined */
1317 #ifdef INC_DEBUG
1318                 printf("%s: skipping %s\n", file->filename, buf);
1319 #endif
1320                 f = NULL;
1321             }  else {
1322                 f = tcc_open(s1, buf1);
1323                 if (!f)
1324                     continue;
1325             }
1326
1327             if (tok == TOK_INCLUDE_NEXT) {
1328                 tok = TOK_INCLUDE;
1329                 if (f)
1330                     tcc_close(f);
1331                 continue;
1332             }
1333
1334             if (!f)
1335                 goto include_done;
1336
1337 #ifdef INC_DEBUG
1338             printf("%s: including %s\n", file->filename, buf1);
1339 #endif
1340
1341            /* XXX: fix current line init */
1342            /* push current file in stack */
1343             *s1->include_stack_ptr++ = file;
1344             f->inc_type = c;
1345             pstrcpy(f->inc_filename, sizeof(f->inc_filename), buf1);
1346             file = f;
1347             /* add include file debug info */
1348             if (tcc_state->do_debug) {
1349                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1350             }
1351             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1352             ch = file->buf_ptr[0];
1353             goto the_end;
1354         }
1355         error("include file '%s' not found", buf);
1356 include_done:
1357         break;
1358     case TOK_IFNDEF:
1359         c = 1;
1360         goto do_ifdef;
1361     case TOK_IF:
1362         c = expr_preprocess();
1363         goto do_if;
1364     case TOK_IFDEF:
1365         c = 0;
1366     do_ifdef:
1367         next_nomacro();
1368         if (tok < TOK_IDENT)
1369             error("invalid argument for '#if%sdef'", c ? "n" : "");
1370         if (is_bof) {
1371             if (c) {
1372 #ifdef INC_DEBUG
1373                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1374 #endif
1375                 file->ifndef_macro = tok;
1376             }
1377         }
1378         c = (define_find(tok) != 0) ^ c;
1379     do_if:
1380         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1381             error("memory full");
1382         *s1->ifdef_stack_ptr++ = c;
1383         goto test_skip;
1384     case TOK_ELSE:
1385         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1386             error("#else without matching #if");
1387         if (s1->ifdef_stack_ptr[-1] & 2)
1388             error("#else after #else");
1389         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1390         goto test_skip;
1391     case TOK_ELIF:
1392         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1393             error("#elif without matching #if");
1394         c = s1->ifdef_stack_ptr[-1];
1395         if (c > 1)
1396             error("#elif after #else");
1397         /* last #if/#elif expression was true: we skip */
1398         if (c == 1)
1399             goto skip;
1400         c = expr_preprocess();
1401         s1->ifdef_stack_ptr[-1] = c;
1402     test_skip:
1403         if (!(c & 1)) {
1404         skip:
1405             preprocess_skip();
1406             is_bof = 0;
1407             goto redo;
1408         }
1409         break;
1410     case TOK_ENDIF:
1411         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1412             error("#endif without matching #if");
1413         s1->ifdef_stack_ptr--;
1414         /* '#ifndef macro' was at the start of file. Now we check if
1415            an '#endif' is exactly at the end of file */
1416         if (file->ifndef_macro &&
1417             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1418             file->ifndef_macro_saved = file->ifndef_macro;
1419             /* need to set to zero to avoid false matches if another
1420                #ifndef at middle of file */
1421             file->ifndef_macro = 0;
1422             while (tok != TOK_LINEFEED)
1423                 next_nomacro();
1424             tok_flags |= TOK_FLAG_ENDIF;
1425             goto the_end;
1426         }
1427         break;
1428     case TOK_LINE:
1429         next();
1430         if (tok != TOK_CINT)
1431             error("#line");
1432         file->line_num = tokc.i - 1; /* the line number will be incremented after */
1433         next();
1434         if (tok != TOK_LINEFEED) {
1435             if (tok != TOK_STR)
1436                 error("#line");
1437             pstrcpy(file->filename, sizeof(file->filename),
1438                     (char *)tokc.cstr->data);
1439         }
1440         break;
1441     case TOK_ERROR:
1442     case TOK_WARNING:
1443         c = tok;
1444         ch = file->buf_ptr[0];
1445         skip_spaces();
1446         q = buf;
1447         while (ch != '\n' && ch != CH_EOF) {
1448             if ((q - buf) < sizeof(buf) - 1)
1449                 *q++ = ch;
1450             if (ch == '\\') {
1451                 if (handle_stray_noerror() == 0)
1452                     --q;
1453             } else
1454                 inp();
1455         }
1456         *q = '\0';
1457         if (c == TOK_ERROR)
1458             error("#error %s", buf);
1459         else
1460             warning("#warning %s", buf);
1461         break;
1462     case TOK_PRAGMA:
1463         pragma_parse(s1);
1464         break;
1465     default:
1466         if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_CINT) {
1467             /* '!' is ignored to allow C scripts. numbers are ignored
1468                to emulate cpp behaviour */
1469         } else {
1470             if (!(saved_parse_flags & PARSE_FLAG_ASM_COMMENTS))
1471                 warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1472         }
1473         break;
1474     }
1475     /* ignore other preprocess commands or #! for C scripts */
1476     while (tok != TOK_LINEFEED)
1477         next_nomacro();
1478  the_end:
1479     parse_flags = saved_parse_flags;
1480 }
1481
1482 /* evaluate escape codes in a string. */
1483 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1484 {
1485     int c, n;
1486     const uint8_t *p;
1487
1488     p = buf;
1489     for(;;) {
1490         c = *p;
1491         if (c == '\0')
1492             break;
1493         if (c == '\\') {
1494             p++;
1495             /* escape */
1496             c = *p;
1497             switch(c) {
1498             case '0': case '1': case '2': case '3':
1499             case '4': case '5': case '6': case '7':
1500                 /* at most three octal digits */
1501                 n = c - '0';
1502                 p++;
1503                 c = *p;
1504                 if (isoct(c)) {
1505                     n = n * 8 + c - '0';
1506                     p++;
1507                     c = *p;
1508                     if (isoct(c)) {
1509                         n = n * 8 + c - '0';
1510                         p++;
1511                     }
1512                 }
1513                 c = n;
1514                 goto add_char_nonext;
1515             case 'x':
1516             case 'u':
1517             case 'U':
1518                 p++;
1519                 n = 0;
1520                 for(;;) {
1521                     c = *p;
1522                     if (c >= 'a' && c <= 'f')
1523                         c = c - 'a' + 10;
1524                     else if (c >= 'A' && c <= 'F')
1525                         c = c - 'A' + 10;
1526                     else if (isnum(c))
1527                         c = c - '0';
1528                     else
1529                         break;
1530                     n = n * 16 + c;
1531                     p++;
1532                 }
1533                 c = n;
1534                 goto add_char_nonext;
1535             case 'a':
1536                 c = '\a';
1537                 break;
1538             case 'b':
1539                 c = '\b';
1540                 break;
1541             case 'f':
1542                 c = '\f';
1543                 break;
1544             case 'n':
1545                 c = '\n';
1546                 break;
1547             case 'r':
1548                 c = '\r';
1549                 break;
1550             case 't':
1551                 c = '\t';
1552                 break;
1553             case 'v':
1554                 c = '\v';
1555                 break;
1556             case 'e':
1557                 if (!gnu_ext)
1558                     goto invalid_escape;
1559                 c = 27;
1560                 break;
1561             case '\'':
1562             case '\"':
1563             case '\\':
1564             case '?':
1565                 break;
1566             default:
1567             invalid_escape:
1568                 if (c >= '!' && c <= '~')
1569                     warning("unknown escape sequence: \'\\%c\'", c);
1570                 else
1571                     warning("unknown escape sequence: \'\\x%x\'", c);
1572                 break;
1573             }
1574         }
1575         p++;
1576     add_char_nonext:
1577         if (!is_long)
1578             cstr_ccat(outstr, c);
1579         else
1580             cstr_wccat(outstr, c);
1581     }
1582     /* add a trailing '\0' */
1583     if (!is_long)
1584         cstr_ccat(outstr, '\0');
1585     else
1586         cstr_wccat(outstr, '\0');
1587 }
1588
1589 /* we use 64 bit numbers */
1590 #define BN_SIZE 2
1591
1592 /* bn = (bn << shift) | or_val */
1593 void bn_lshift(unsigned int *bn, int shift, int or_val)
1594 {
1595     int i;
1596     unsigned int v;
1597     for(i=0;i<BN_SIZE;i++) {
1598         v = bn[i];
1599         bn[i] = (v << shift) | or_val;
1600         or_val = v >> (32 - shift);
1601     }
1602 }
1603
1604 void bn_zero(unsigned int *bn)
1605 {
1606     int i;
1607     for(i=0;i<BN_SIZE;i++) {
1608         bn[i] = 0;
1609     }
1610 }
1611
1612 /* parse number in null terminated string 'p' and return it in the
1613    current token */
1614 void parse_number(const char *p)
1615 {
1616     int b, t, shift, frac_bits, s, exp_val, ch;
1617     char *q;
1618     unsigned int bn[BN_SIZE];
1619     double d;
1620
1621     /* number */
1622     q = token_buf;
1623     ch = *p++;
1624     t = ch;
1625     ch = *p++;
1626     *q++ = t;
1627     b = 10;
1628     if (t == '.') {
1629         goto float_frac_parse;
1630     } else if (t == '0') {
1631         if (ch == 'x' || ch == 'X') {
1632             q--;
1633             ch = *p++;
1634             b = 16;
1635         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
1636             q--;
1637             ch = *p++;
1638             b = 2;
1639         }
1640     }
1641     /* parse all digits. cannot check octal numbers at this stage
1642        because of floating point constants */
1643     while (1) {
1644         if (ch >= 'a' && ch <= 'f')
1645             t = ch - 'a' + 10;
1646         else if (ch >= 'A' && ch <= 'F')
1647             t = ch - 'A' + 10;
1648         else if (isnum(ch))
1649             t = ch - '0';
1650         else
1651             break;
1652         if (t >= b)
1653             break;
1654         if (q >= token_buf + STRING_MAX_SIZE) {
1655         num_too_long:
1656             error("number too long");
1657         }
1658         *q++ = ch;
1659         ch = *p++;
1660     }
1661     if (ch == '.' ||
1662         ((ch == 'e' || ch == 'E') && b == 10) ||
1663         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
1664         if (b != 10) {
1665             /* NOTE: strtox should support that for hexa numbers, but
1666                non ISOC99 libcs do not support it, so we prefer to do
1667                it by hand */
1668             /* hexadecimal or binary floats */
1669             /* XXX: handle overflows */
1670             *q = '\0';
1671             if (b == 16)
1672                 shift = 4;
1673             else
1674                 shift = 2;
1675             bn_zero(bn);
1676             q = token_buf;
1677             while (1) {
1678                 t = *q++;
1679                 if (t == '\0') {
1680                     break;
1681                 } else if (t >= 'a') {
1682                     t = t - 'a' + 10;
1683                 } else if (t >= 'A') {
1684                     t = t - 'A' + 10;
1685                 } else {
1686                     t = t - '0';
1687                 }
1688                 bn_lshift(bn, shift, t);
1689             }
1690             frac_bits = 0;
1691             if (ch == '.') {
1692                 ch = *p++;
1693                 while (1) {
1694                     t = ch;
1695                     if (t >= 'a' && t <= 'f') {
1696                         t = t - 'a' + 10;
1697                     } else if (t >= 'A' && t <= 'F') {
1698                         t = t - 'A' + 10;
1699                     } else if (t >= '0' && t <= '9') {
1700                         t = t - '0';
1701                     } else {
1702                         break;
1703                     }
1704                     if (t >= b)
1705                         error("invalid digit");
1706                     bn_lshift(bn, shift, t);
1707                     frac_bits += shift;
1708                     ch = *p++;
1709                 }
1710             }
1711             if (ch != 'p' && ch != 'P')
1712                 expect("exponent");
1713             ch = *p++;
1714             s = 1;
1715             exp_val = 0;
1716             if (ch == '+') {
1717                 ch = *p++;
1718             } else if (ch == '-') {
1719                 s = -1;
1720                 ch = *p++;
1721             }
1722             if (ch < '0' || ch > '9')
1723                 expect("exponent digits");
1724             while (ch >= '0' && ch <= '9') {
1725                 exp_val = exp_val * 10 + ch - '0';
1726                 ch = *p++;
1727             }
1728             exp_val = exp_val * s;
1729
1730             /* now we can generate the number */
1731             /* XXX: should patch directly float number */
1732             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
1733             d = ldexp(d, exp_val - frac_bits);
1734             t = toup(ch);
1735             if (t == 'F') {
1736                 ch = *p++;
1737                 tok = TOK_CFLOAT;
1738                 /* float : should handle overflow */
1739                 tokc.f = (float)d;
1740             } else if (t == 'L') {
1741                 ch = *p++;
1742                 tok = TOK_CLDOUBLE;
1743                 /* XXX: not large enough */
1744                 tokc.ld = (long double)d;
1745             } else {
1746                 tok = TOK_CDOUBLE;
1747                 tokc.d = d;
1748             }
1749         } else {
1750             /* decimal floats */
1751             if (ch == '.') {
1752                 if (q >= token_buf + STRING_MAX_SIZE)
1753                     goto num_too_long;
1754                 *q++ = ch;
1755                 ch = *p++;
1756             float_frac_parse:
1757                 while (ch >= '0' && ch <= '9') {
1758                     if (q >= token_buf + STRING_MAX_SIZE)
1759                         goto num_too_long;
1760                     *q++ = ch;
1761                     ch = *p++;
1762                 }
1763             }
1764             if (ch == 'e' || ch == 'E') {
1765                 if (q >= token_buf + STRING_MAX_SIZE)
1766                     goto num_too_long;
1767                 *q++ = ch;
1768                 ch = *p++;
1769                 if (ch == '-' || ch == '+') {
1770                     if (q >= token_buf + STRING_MAX_SIZE)
1771                         goto num_too_long;
1772                     *q++ = ch;
1773                     ch = *p++;
1774                 }
1775                 if (ch < '0' || ch > '9')
1776                     expect("exponent digits");
1777                 while (ch >= '0' && ch <= '9') {
1778                     if (q >= token_buf + STRING_MAX_SIZE)
1779                         goto num_too_long;
1780                     *q++ = ch;
1781                     ch = *p++;
1782                 }
1783             }
1784             *q = '\0';
1785             t = toup(ch);
1786             errno = 0;
1787             if (t == 'F') {
1788                 ch = *p++;
1789                 tok = TOK_CFLOAT;
1790                 tokc.f = strtof(token_buf, NULL);
1791             } else if (t == 'L') {
1792                 ch = *p++;
1793                 tok = TOK_CLDOUBLE;
1794                 tokc.ld = strtold(token_buf, NULL);
1795             } else {
1796                 tok = TOK_CDOUBLE;
1797                 tokc.d = strtod(token_buf, NULL);
1798             }
1799         }
1800     } else {
1801         unsigned long long n, n1;
1802         int lcount, ucount;
1803
1804         /* integer number */
1805         *q = '\0';
1806         q = token_buf;
1807         if (b == 10 && *q == '0') {
1808             b = 8;
1809             q++;
1810         }
1811         n = 0;
1812         while(1) {
1813             t = *q++;
1814             /* no need for checks except for base 10 / 8 errors */
1815             if (t == '\0') {
1816                 break;
1817             } else if (t >= 'a') {
1818                 t = t - 'a' + 10;
1819             } else if (t >= 'A') {
1820                 t = t - 'A' + 10;
1821             } else {
1822                 t = t - '0';
1823                 if (t >= b)
1824                     error("invalid digit");
1825             }
1826             n1 = n;
1827             n = n * b + t;
1828             /* detect overflow */
1829             /* XXX: this test is not reliable */
1830             if (n < n1)
1831                 error("integer constant overflow");
1832         }
1833
1834         /* XXX: not exactly ANSI compliant */
1835         if ((n & 0xffffffff00000000LL) != 0) {
1836             if ((n >> 63) != 0)
1837                 tok = TOK_CULLONG;
1838             else
1839                 tok = TOK_CLLONG;
1840         } else if (n > 0x7fffffff) {
1841             tok = TOK_CUINT;
1842         } else {
1843             tok = TOK_CINT;
1844         }
1845         lcount = 0;
1846         ucount = 0;
1847         for(;;) {
1848             t = toup(ch);
1849             if (t == 'L') {
1850                 if (lcount >= 2)
1851                     error("three 'l's in integer constant");
1852                 lcount++;
1853                 if (lcount == 2) {
1854                     if (tok == TOK_CINT)
1855                         tok = TOK_CLLONG;
1856                     else if (tok == TOK_CUINT)
1857                         tok = TOK_CULLONG;
1858                 }
1859                 ch = *p++;
1860             } else if (t == 'U') {
1861                 if (ucount >= 1)
1862                     error("two 'u's in integer constant");
1863                 ucount++;
1864                 if (tok == TOK_CINT)
1865                     tok = TOK_CUINT;
1866                 else if (tok == TOK_CLLONG)
1867                     tok = TOK_CULLONG;
1868                 ch = *p++;
1869             } else {
1870                 break;
1871             }
1872         }
1873         if (tok == TOK_CINT || tok == TOK_CUINT)
1874             tokc.ui = n;
1875         else
1876             tokc.ull = n;
1877     }
1878     if (ch)
1879         error("invalid number\n");
1880 }
1881
1882
1883 #define PARSE2(c1, tok1, c2, tok2)              \
1884     case c1:                                    \
1885         PEEKC(c, p);                            \
1886         if (c == c2) {                          \
1887             p++;                                \
1888             tok = tok2;                         \
1889         } else {                                \
1890             tok = tok1;                         \
1891         }                                       \
1892         break;
1893
1894 /* return next token without macro substitution */
1895 static inline void next_nomacro1(void)
1896 {
1897     int t, c, is_long;
1898     TokenSym *ts;
1899     uint8_t *p, *p1;
1900     unsigned int h;
1901
1902     p = file->buf_ptr;
1903  redo_no_start:
1904     c = *p;
1905     switch(c) {
1906     case ' ':
1907     case '\t':
1908         tok = c;
1909         p++;
1910         goto keep_tok_flags;
1911     case '\f':
1912     case '\v':
1913     case '\r':
1914         p++;
1915         goto redo_no_start;
1916     case '\\':
1917         /* first look if it is in fact an end of buffer */
1918         if (p >= file->buf_end) {
1919             file->buf_ptr = p;
1920             handle_eob();
1921             p = file->buf_ptr;
1922             if (p >= file->buf_end)
1923                 goto parse_eof;
1924             else
1925                 goto redo_no_start;
1926         } else {
1927             file->buf_ptr = p;
1928             ch = *p;
1929             handle_stray();
1930             p = file->buf_ptr;
1931             goto redo_no_start;
1932         }
1933     parse_eof:
1934         {
1935             TCCState *s1 = tcc_state;
1936             if ((parse_flags & PARSE_FLAG_LINEFEED)
1937                 && !(tok_flags & TOK_FLAG_EOF)) {
1938                 tok_flags |= TOK_FLAG_EOF;
1939                 tok = TOK_LINEFEED;
1940                 goto keep_tok_flags;
1941             } else if (s1->include_stack_ptr == s1->include_stack ||
1942                        !(parse_flags & PARSE_FLAG_PREPROCESS)) {
1943                 /* no include left : end of file. */
1944                 tok = TOK_EOF;
1945             } else {
1946                 tok_flags &= ~TOK_FLAG_EOF;
1947                 /* pop include file */
1948
1949                 /* test if previous '#endif' was after a #ifdef at
1950                    start of file */
1951                 if (tok_flags & TOK_FLAG_ENDIF) {
1952 #ifdef INC_DEBUG
1953                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
1954 #endif
1955                     add_cached_include(s1, file->inc_type, file->inc_filename,
1956                                        file->ifndef_macro_saved);
1957                 }
1958
1959                 /* add end of include file debug info */
1960                 if (tcc_state->do_debug) {
1961                     put_stabd(N_EINCL, 0, 0);
1962                 }
1963                 /* pop include stack */
1964                 tcc_close(file);
1965                 s1->include_stack_ptr--;
1966                 file = *s1->include_stack_ptr;
1967                 p = file->buf_ptr;
1968                 goto redo_no_start;
1969             }
1970         }
1971         break;
1972
1973     case '\n':
1974         file->line_num++;
1975         tok_flags |= TOK_FLAG_BOL;
1976         p++;
1977 maybe_newline:
1978         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
1979             goto redo_no_start;
1980         tok = TOK_LINEFEED;
1981         goto keep_tok_flags;
1982
1983     case '#':
1984         /* XXX: simplify */
1985         PEEKC(c, p);
1986         if ((tok_flags & TOK_FLAG_BOL) &&
1987             (parse_flags & PARSE_FLAG_PREPROCESS)) {
1988             file->buf_ptr = p;
1989             preprocess(tok_flags & TOK_FLAG_BOF);
1990             p = file->buf_ptr;
1991             goto maybe_newline;
1992         } else {
1993             if (c == '#') {
1994                 p++;
1995                 tok = TOK_TWOSHARPS;
1996             } else {
1997                 if (parse_flags & PARSE_FLAG_ASM_COMMENTS) {
1998                     p = parse_line_comment(p - 1);
1999                     goto redo_no_start;
2000                 } else {
2001                     tok = '#';
2002                 }
2003             }
2004         }
2005         break;
2006
2007     case 'a': case 'b': case 'c': case 'd':
2008     case 'e': case 'f': case 'g': case 'h':
2009     case 'i': case 'j': case 'k': case 'l':
2010     case 'm': case 'n': case 'o': case 'p':
2011     case 'q': case 'r': case 's': case 't':
2012     case 'u': case 'v': case 'w': case 'x':
2013     case 'y': case 'z':
2014     case 'A': case 'B': case 'C': case 'D':
2015     case 'E': case 'F': case 'G': case 'H':
2016     case 'I': case 'J': case 'K':
2017     case 'M': case 'N': case 'O': case 'P':
2018     case 'Q': case 'R': case 'S': case 'T':
2019     case 'U': case 'V': case 'W': case 'X':
2020     case 'Y': case 'Z':
2021     case '_':
2022     parse_ident_fast:
2023         p1 = p;
2024         h = TOK_HASH_INIT;
2025         h = TOK_HASH_FUNC(h, c);
2026         p++;
2027         for(;;) {
2028             c = *p;
2029             if (!isidnum_table[c-CH_EOF])
2030                 break;
2031             h = TOK_HASH_FUNC(h, c);
2032             p++;
2033         }
2034         if (c != '\\') {
2035             TokenSym **pts;
2036             int len;
2037
2038             /* fast case : no stray found, so we have the full token
2039                and we have already hashed it */
2040             len = p - p1;
2041             h &= (TOK_HASH_SIZE - 1);
2042             pts = &hash_ident[h];
2043             for(;;) {
2044                 ts = *pts;
2045                 if (!ts)
2046                     break;
2047                 if (ts->len == len && !memcmp(ts->str, p1, len))
2048                     goto token_found;
2049                 pts = &(ts->hash_next);
2050             }
2051             ts = tok_alloc_new(pts, p1, len);
2052         token_found: ;
2053         } else {
2054             /* slower case */
2055             cstr_reset(&tokcstr);
2056
2057             while (p1 < p) {
2058                 cstr_ccat(&tokcstr, *p1);
2059                 p1++;
2060             }
2061             p--;
2062             PEEKC(c, p);
2063         parse_ident_slow:
2064             while (isidnum_table[c-CH_EOF]) {
2065                 cstr_ccat(&tokcstr, c);
2066                 PEEKC(c, p);
2067             }
2068             ts = tok_alloc(tokcstr.data, tokcstr.size);
2069         }
2070         tok = ts->tok;
2071         break;
2072     case 'L':
2073         t = p[1];
2074         if (t != '\\' && t != '\'' && t != '\"') {
2075             /* fast case */
2076             goto parse_ident_fast;
2077         } else {
2078             PEEKC(c, p);
2079             if (c == '\'' || c == '\"') {
2080                 is_long = 1;
2081                 goto str_const;
2082             } else {
2083                 cstr_reset(&tokcstr);
2084                 cstr_ccat(&tokcstr, 'L');
2085                 goto parse_ident_slow;
2086             }
2087         }
2088         break;
2089     case '0': case '1': case '2': case '3':
2090     case '4': case '5': case '6': case '7':
2091     case '8': case '9':
2092
2093         cstr_reset(&tokcstr);
2094         /* after the first digit, accept digits, alpha, '.' or sign if
2095            prefixed by 'eEpP' */
2096     parse_num:
2097         for(;;) {
2098             t = c;
2099             cstr_ccat(&tokcstr, c);
2100             PEEKC(c, p);
2101             if (!(isnum(c) || isid(c) || c == '.' ||
2102                   ((c == '+' || c == '-') &&
2103                    (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2104                 break;
2105         }
2106         /* We add a trailing '\0' to ease parsing */
2107         cstr_ccat(&tokcstr, '\0');
2108         tokc.cstr = &tokcstr;
2109         tok = TOK_PPNUM;
2110         break;
2111     case '.':
2112         /* special dot handling because it can also start a number */
2113         PEEKC(c, p);
2114         if (isnum(c)) {
2115             cstr_reset(&tokcstr);
2116             cstr_ccat(&tokcstr, '.');
2117             goto parse_num;
2118         } else if (c == '.') {
2119             PEEKC(c, p);
2120             if (c != '.')
2121                 expect("'.'");
2122             PEEKC(c, p);
2123             tok = TOK_DOTS;
2124         } else {
2125             tok = '.';
2126         }
2127         break;
2128     case '\'':
2129     case '\"':
2130         is_long = 0;
2131     str_const:
2132         {
2133             CString str;
2134             int sep;
2135
2136             sep = c;
2137
2138             /* parse the string */
2139             cstr_new(&str);
2140             p = parse_pp_string(p, sep, &str);
2141             cstr_ccat(&str, '\0');
2142
2143             /* eval the escape (should be done as TOK_PPNUM) */
2144             cstr_reset(&tokcstr);
2145             parse_escape_string(&tokcstr, str.data, is_long);
2146             cstr_free(&str);
2147
2148             if (sep == '\'') {
2149                 int char_size;
2150                 /* XXX: make it portable */
2151                 if (!is_long)
2152                     char_size = 1;
2153                 else
2154                     char_size = sizeof(nwchar_t);
2155                 if (tokcstr.size <= char_size)
2156                     error("empty character constant");
2157                 if (tokcstr.size > 2 * char_size)
2158                     warning("multi-character character constant");
2159                 if (!is_long) {
2160                     tokc.i = *(int8_t *)tokcstr.data;
2161                     tok = TOK_CCHAR;
2162                 } else {
2163                     tokc.i = *(nwchar_t *)tokcstr.data;
2164                     tok = TOK_LCHAR;
2165                 }
2166             } else {
2167                 tokc.cstr = &tokcstr;
2168                 if (!is_long)
2169                     tok = TOK_STR;
2170                 else
2171                     tok = TOK_LSTR;
2172             }
2173         }
2174         break;
2175
2176     case '<':
2177         PEEKC(c, p);
2178         if (c == '=') {
2179             p++;
2180             tok = TOK_LE;
2181         } else if (c == '<') {
2182             PEEKC(c, p);
2183             if (c == '=') {
2184                 p++;
2185                 tok = TOK_A_SHL;
2186             } else {
2187                 tok = TOK_SHL;
2188             }
2189         } else {
2190             tok = TOK_LT;
2191         }
2192         break;
2193
2194     case '>':
2195         PEEKC(c, p);
2196         if (c == '=') {
2197             p++;
2198             tok = TOK_GE;
2199         } else if (c == '>') {
2200             PEEKC(c, p);
2201             if (c == '=') {
2202                 p++;
2203                 tok = TOK_A_SAR;
2204             } else {
2205                 tok = TOK_SAR;
2206             }
2207         } else {
2208             tok = TOK_GT;
2209         }
2210         break;
2211
2212     case '&':
2213         PEEKC(c, p);
2214         if (c == '&') {
2215             p++;
2216             tok = TOK_LAND;
2217         } else if (c == '=') {
2218             p++;
2219             tok = TOK_A_AND;
2220         } else {
2221             tok = '&';
2222         }
2223         break;
2224
2225     case '|':
2226         PEEKC(c, p);
2227         if (c == '|') {
2228             p++;
2229             tok = TOK_LOR;
2230         } else if (c == '=') {
2231             p++;
2232             tok = TOK_A_OR;
2233         } else {
2234             tok = '|';
2235         }
2236         break;
2237
2238     case '+':
2239         PEEKC(c, p);
2240         if (c == '+') {
2241             p++;
2242             tok = TOK_INC;
2243         } else if (c == '=') {
2244             p++;
2245             tok = TOK_A_ADD;
2246         } else {
2247             tok = '+';
2248         }
2249         break;
2250
2251     case '-':
2252         PEEKC(c, p);
2253         if (c == '-') {
2254             p++;
2255             tok = TOK_DEC;
2256         } else if (c == '=') {
2257             p++;
2258             tok = TOK_A_SUB;
2259         } else if (c == '>') {
2260             p++;
2261             tok = TOK_ARROW;
2262         } else {
2263             tok = '-';
2264         }
2265         break;
2266
2267     PARSE2('!', '!', '=', TOK_NE)
2268     PARSE2('=', '=', '=', TOK_EQ)
2269     PARSE2('*', '*', '=', TOK_A_MUL)
2270     PARSE2('%', '%', '=', TOK_A_MOD)
2271     PARSE2('^', '^', '=', TOK_A_XOR)
2272
2273         /* comments or operator */
2274     case '/':
2275         PEEKC(c, p);
2276         if (c == '*') {
2277             p = parse_comment(p);
2278             goto redo_no_start;
2279         } else if (c == '/') {
2280             p = parse_line_comment(p);
2281             goto redo_no_start;
2282         } else if (c == '=') {
2283             p++;
2284             tok = TOK_A_DIV;
2285         } else {
2286             tok = '/';
2287         }
2288         break;
2289
2290         /* simple tokens */
2291     case '(':
2292     case ')':
2293     case '[':
2294     case ']':
2295     case '{':
2296     case '}':
2297     case ',':
2298     case ';':
2299     case ':':
2300     case '?':
2301     case '~':
2302     case '$': /* only used in assembler */
2303     case '@': /* dito */
2304         tok = c;
2305         p++;
2306         break;
2307     default:
2308         error("unrecognized character \\x%02x", c);
2309         break;
2310     }
2311     tok_flags = 0;
2312 keep_tok_flags:
2313     file->buf_ptr = p;
2314 #if defined(PARSE_DEBUG)
2315     printf("token = %s\n", get_tok_str(tok, &tokc));
2316 #endif
2317 }
2318
2319 /* return next token without macro substitution. Can read input from
2320    macro_ptr buffer */
2321 static void next_nomacro_spc(void)
2322 {
2323     if (macro_ptr) {
2324     redo:
2325         tok = *macro_ptr;
2326         if (tok) {
2327             TOK_GET(tok, macro_ptr, tokc);
2328             if (tok == TOK_LINENUM) {
2329                 file->line_num = tokc.i;
2330                 goto redo;
2331             }
2332         }
2333     } else {
2334         next_nomacro1();
2335     }
2336 }
2337
2338 static void next_nomacro(void)
2339 {
2340     do {
2341         next_nomacro_spc();
2342     } while (is_space(tok));
2343 }
2344
2345 /* substitute args in macro_str and return allocated string */
2346 static int *macro_arg_subst(Sym **nested_list, int *macro_str, Sym *args)
2347 {
2348     int *st, last_tok, t, spc;
2349     Sym *s;
2350     CValue cval;
2351     TokenString str;
2352     CString cstr;
2353
2354     tok_str_new(&str);
2355     last_tok = 0;
2356     while(1) {
2357         TOK_GET(t, macro_str, cval);
2358         if (!t)
2359             break;
2360         if (t == '#') {
2361             /* stringize */
2362             TOK_GET(t, macro_str, cval);
2363             if (!t)
2364                 break;
2365             s = sym_find2(args, t);
2366             if (s) {
2367                 cstr_new(&cstr);
2368                 st = s->d;
2369                 spc = 0;
2370                 while (*st) {
2371                     TOK_GET(t, st, cval);
2372                     if (!check_space(t, &spc))
2373                         cstr_cat(&cstr, get_tok_str(t, &cval));
2374                 }
2375                 cstr.size -= spc;
2376                 cstr_ccat(&cstr, '\0');
2377 #ifdef PP_DEBUG
2378                 printf("stringize: %s\n", (char *)cstr.data);
2379 #endif
2380                 /* add string */
2381                 cval.cstr = &cstr;
2382                 tok_str_add2(&str, TOK_STR, &cval);
2383                 cstr_free(&cstr);
2384             } else {
2385                 tok_str_add2(&str, t, &cval);
2386             }
2387         } else if (t >= TOK_IDENT) {
2388             s = sym_find2(args, t);
2389             if (s) {
2390                 st = s->d;
2391                 /* if '##' is present before or after, no arg substitution */
2392                 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2393                     /* special case for var arg macros : ## eats the
2394                        ',' if empty VA_ARGS variable. */
2395                     /* XXX: test of the ',' is not 100%
2396                        reliable. should fix it to avoid security
2397                        problems */
2398                     if (gnu_ext && s->type.t &&
2399                         last_tok == TOK_TWOSHARPS &&
2400                         str.len >= 2 && str.str[str.len - 2] == ',') {
2401                         if (*st == 0) {
2402                             /* suppress ',' '##' */
2403                             str.len -= 2;
2404                         } else {
2405                             /* suppress '##' and add variable */
2406                             str.len--;
2407                             goto add_var;
2408                         }
2409                     } else {
2410                         int t1;
2411                     add_var:
2412                         for(;;) {
2413                             TOK_GET(t1, st, cval);
2414                             if (!t1)
2415                                 break;
2416                             tok_str_add2(&str, t1, &cval);
2417                         }
2418                     }
2419                 } else {
2420                     /* NOTE: the stream cannot be read when macro
2421                        substituing an argument */
2422                     macro_subst(&str, nested_list, st, NULL);
2423                 }
2424             } else {
2425                 tok_str_add(&str, t);
2426             }
2427         } else {
2428             tok_str_add2(&str, t, &cval);
2429         }
2430         last_tok = t;
2431     }
2432     tok_str_add(&str, 0);
2433     return str.str;
2434 }
2435
2436 static char const ab_month_name[12][4] =
2437 {
2438     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2439     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2440 };
2441
2442 /* do macro substitution of current token with macro 's' and add
2443    result to (tok_str,tok_len). 'nested_list' is the list of all
2444    macros we got inside to avoid recursing. Return non zero if no
2445    substitution needs to be done */
2446 static int macro_subst_tok(TokenString *tok_str,
2447                            Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2448 {
2449     Sym *args, *sa, *sa1;
2450     int mstr_allocated, parlevel, *mstr, t, t1, *p, spc;
2451     TokenString str;
2452     char *cstrval;
2453     CValue cval;
2454     CString cstr;
2455     char buf[32];
2456
2457     /* if symbol is a macro, prepare substitution */
2458     /* special macros */
2459     if (tok == TOK___LINE__) {
2460         snprintf(buf, sizeof(buf), "%d", file->line_num);
2461         cstrval = buf;
2462         t1 = TOK_PPNUM;
2463         goto add_cstr1;
2464     } else if (tok == TOK___FILE__) {
2465         cstrval = file->filename;
2466         goto add_cstr;
2467     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2468         time_t ti;
2469         struct tm *tm;
2470
2471         time(&ti);
2472         tm = localtime(&ti);
2473         if (tok == TOK___DATE__) {
2474             snprintf(buf, sizeof(buf), "%s %2d %d",
2475                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2476         } else {
2477             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2478                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2479         }
2480         cstrval = buf;
2481     add_cstr:
2482         t1 = TOK_STR;
2483     add_cstr1:
2484         cstr_new(&cstr);
2485         cstr_cat(&cstr, cstrval);
2486         cstr_ccat(&cstr, '\0');
2487         cval.cstr = &cstr;
2488         tok_str_add2(tok_str, t1, &cval);
2489         cstr_free(&cstr);
2490     } else {
2491         mstr = s->d;
2492         mstr_allocated = 0;
2493         if (s->type.t == MACRO_FUNC) {
2494             /* NOTE: we do not use next_nomacro to avoid eating the
2495                next token. XXX: find better solution */
2496         redo:
2497             if (macro_ptr) {
2498                 p = macro_ptr;
2499                 while (is_space(t = *p) || TOK_LINEFEED == t)
2500                     ++p;
2501                 if (t == 0 && can_read_stream) {
2502                     /* end of macro stream: we must look at the token
2503                        after in the file */
2504                     struct macro_level *ml = *can_read_stream;
2505                     macro_ptr = NULL;
2506                     if (ml)
2507                     {
2508                         macro_ptr = ml->p;
2509                         ml->p = NULL;
2510                         *can_read_stream = ml -> prev;
2511                     }
2512                     goto redo;
2513                 }
2514             } else {
2515                 /* XXX: incorrect with comments */
2516                 ch = file->buf_ptr[0];
2517                 while (is_space(ch) || ch == '\n')
2518                     cinp();
2519                 t = ch;
2520             }
2521             if (t != '(') /* no macro subst */
2522                 return -1;
2523
2524             /* argument macro */
2525             next_nomacro();
2526             next_nomacro();
2527             args = NULL;
2528             sa = s->next;
2529             /* NOTE: empty args are allowed, except if no args */
2530             for(;;) {
2531                 /* handle '()' case */
2532                 if (!args && !sa && tok == ')')
2533                     break;
2534                 if (!sa)
2535                     error("macro '%s' used with too many args",
2536                           get_tok_str(s->v, 0));
2537                 tok_str_new(&str);
2538                 parlevel = spc = 0;
2539                 /* NOTE: non zero sa->t indicates VA_ARGS */
2540                 while ((parlevel > 0 ||
2541                         (tok != ')' &&
2542                          (tok != ',' || sa->type.t))) &&
2543                        tok != -1) {
2544                     if (tok == '(')
2545                         parlevel++;
2546                     else if (tok == ')')
2547                         parlevel--;
2548                     if (tok == TOK_LINEFEED)
2549                         tok = ' ';
2550                     if (!check_space(tok, &spc))
2551                         tok_str_add2(&str, tok, &tokc);
2552                     next_nomacro_spc();
2553                 }
2554                 str.len -= spc;
2555                 tok_str_add(&str, 0);
2556                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
2557                 sa1->d = str.str;
2558                 sa = sa->next;
2559                 if (tok == ')') {
2560                     /* special case for gcc var args: add an empty
2561                        var arg argument if it is omitted */
2562                     if (sa && sa->type.t && gnu_ext)
2563                         continue;
2564                     else
2565                         break;
2566                 }
2567                 if (tok != ',')
2568                     expect(",");
2569                 next_nomacro();
2570             }
2571             if (sa) {
2572                 error("macro '%s' used with too few args",
2573                       get_tok_str(s->v, 0));
2574             }
2575
2576             /* now subst each arg */
2577             mstr = macro_arg_subst(nested_list, mstr, args);
2578             /* free memory */
2579             sa = args;
2580             while (sa) {
2581                 sa1 = sa->prev;
2582                 tok_str_free(sa->d);
2583                 sym_free(sa);
2584                 sa = sa1;
2585             }
2586             mstr_allocated = 1;
2587         }
2588         sym_push2(nested_list, s->v, 0, 0);
2589         macro_subst(tok_str, nested_list, mstr, can_read_stream);
2590         /* pop nested defined symbol */
2591         sa1 = *nested_list;
2592         *nested_list = sa1->prev;
2593         sym_free(sa1);
2594         if (mstr_allocated)
2595             tok_str_free(mstr);
2596     }
2597     return 0;
2598 }
2599
2600 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2601    return the resulting string (which must be freed). */
2602 static inline int *macro_twosharps(const int *macro_str)
2603 {
2604     const int *ptr;
2605     int t;
2606     CValue cval;
2607     TokenString macro_str1;
2608     CString cstr;
2609     char *p;
2610     int n;
2611
2612     /* we search the first '##' */
2613     for(ptr = macro_str;;) {
2614         TOK_GET(t, ptr, cval);
2615         if (t == TOK_TWOSHARPS)
2616             break;
2617         /* nothing more to do if end of string */
2618         if (t == 0)
2619             return NULL;
2620     }
2621
2622     /* we saw '##', so we need more processing to handle it */
2623     tok_str_new(&macro_str1);
2624     for(ptr = macro_str;;) {
2625         TOK_GET(tok, ptr, tokc);
2626         if (tok == 0)
2627             break;
2628         if (tok == TOK_TWOSHARPS)
2629             continue;
2630         while (*ptr == TOK_TWOSHARPS) {
2631             t = *++ptr;
2632             if (t && t != TOK_TWOSHARPS) {
2633                 TOK_GET(t, ptr, cval);
2634
2635                 /* We concatenate the two tokens */
2636                 cstr_new(&cstr);
2637                 cstr_cat(&cstr, get_tok_str(tok, &tokc));
2638                 n = cstr.size;
2639                 cstr_cat(&cstr, get_tok_str(t, &cval));
2640                 cstr_ccat(&cstr, '\0');
2641
2642                 p = file->buf_ptr;
2643                 file->buf_ptr = cstr.data;
2644                 for (;;) {
2645                     next_nomacro1();
2646                     if (0 == *file->buf_ptr)
2647                         break;
2648                     tok_str_add2(&macro_str1, tok, &tokc);
2649
2650                     warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
2651                         n, cstr.data, (char*)cstr.data + n);
2652                 }
2653                 file->buf_ptr = p;
2654                 cstr_reset(&cstr);
2655             }
2656         }
2657         tok_str_add2(&macro_str1, tok, &tokc);
2658     }
2659     tok_str_add(&macro_str1, 0);
2660     return macro_str1.str;
2661 }
2662
2663
2664 /* do macro substitution of macro_str and add result to
2665    (tok_str,tok_len). 'nested_list' is the list of all macros we got
2666    inside to avoid recursing. */
2667 static void macro_subst(TokenString *tok_str, Sym **nested_list,
2668                         const int *macro_str, struct macro_level ** can_read_stream)
2669 {
2670     Sym *s;
2671     int *macro_str1;
2672     const int *ptr;
2673     int t, ret, spc;
2674     CValue cval;
2675     struct macro_level ml;
2676
2677     /* first scan for '##' operator handling */
2678     ptr = macro_str;
2679     macro_str1 = macro_twosharps(ptr);
2680     if (macro_str1)
2681         ptr = macro_str1;
2682     spc = 0;
2683     while (1) {
2684         /* NOTE: ptr == NULL can only happen if tokens are read from
2685            file stream due to a macro function call */
2686         if (ptr == NULL)
2687             break;
2688         TOK_GET(t, ptr, cval);
2689         if (t == 0)
2690             break;
2691         s = define_find(t);
2692         if (s != NULL) {
2693             /* if nested substitution, do nothing */
2694             if (sym_find2(*nested_list, t))
2695                 goto no_subst;
2696             ml.p = macro_ptr;
2697             if (can_read_stream)
2698                 ml.prev = *can_read_stream, *can_read_stream = &ml;
2699             macro_ptr = (int *)ptr;
2700             tok = t;
2701             ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
2702             ptr = (int *)macro_ptr;
2703             macro_ptr = ml.p;
2704             if (can_read_stream && *can_read_stream == &ml)
2705                 *can_read_stream = ml.prev;
2706             if (ret != 0)
2707                 goto no_subst;
2708         } else {
2709         no_subst:
2710             if (!check_space(t, &spc))
2711                 tok_str_add2(tok_str, t, &cval);
2712         }
2713     }
2714     if (macro_str1)
2715         tok_str_free(macro_str1);
2716 }
2717
2718 /* return next token with macro substitution */
2719 static void next(void)
2720 {
2721     Sym *nested_list, *s;
2722     TokenString str;
2723     struct macro_level *ml;
2724
2725  redo:
2726     if (parse_flags & PARSE_FLAG_SPACES)
2727         next_nomacro_spc();
2728     else
2729         next_nomacro();
2730     if (!macro_ptr) {
2731         /* if not reading from macro substituted string, then try
2732            to substitute macros */
2733         if (tok >= TOK_IDENT &&
2734             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2735             s = define_find(tok);
2736             if (s) {
2737                 /* we have a macro: we try to substitute */
2738                 tok_str_new(&str);
2739                 nested_list = NULL;
2740                 ml = NULL;
2741                 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
2742                     /* substitution done, NOTE: maybe empty */
2743                     tok_str_add(&str, 0);
2744                     macro_ptr = str.str;
2745                     macro_ptr_allocated = str.str;
2746                     goto redo;
2747                 }
2748             }
2749         }
2750     } else {
2751         if (tok == 0) {
2752             /* end of macro or end of unget buffer */
2753             if (unget_buffer_enabled) {
2754                 macro_ptr = unget_saved_macro_ptr;
2755                 unget_buffer_enabled = 0;
2756             } else {
2757                 /* end of macro string: free it */
2758                 tok_str_free(macro_ptr_allocated);
2759                 macro_ptr = NULL;
2760             }
2761             goto redo;
2762         }
2763     }
2764
2765     /* convert preprocessor tokens into C tokens */
2766     if (tok == TOK_PPNUM &&
2767         (parse_flags & PARSE_FLAG_TOK_NUM)) {
2768         parse_number((char *)tokc.cstr->data);
2769     }
2770 }
2771
2772 /* push back current token and set current token to 'last_tok'. Only
2773    identifier case handled for labels. */
2774 static inline void unget_tok(int last_tok)
2775 {
2776     int i, n;
2777     int *q;
2778     unget_saved_macro_ptr = macro_ptr;
2779     unget_buffer_enabled = 1;
2780     q = unget_saved_buffer;
2781     macro_ptr = q;
2782     *q++ = tok;
2783     n = tok_ext_size(tok) - 1;
2784     for(i=0;i<n;i++)
2785         *q++ = tokc.tab[i];
2786     *q = 0; /* end of token string */
2787     tok = last_tok;
2788 }
2789
2790
2791 /* better than nothing, but needs extension to handle '-E' option
2792    correctly too */
2793 static void preprocess_init(TCCState *s1)
2794 {
2795     s1->include_stack_ptr = s1->include_stack;
2796     /* XXX: move that before to avoid having to initialize
2797        file->ifdef_stack_ptr ? */
2798     s1->ifdef_stack_ptr = s1->ifdef_stack;
2799     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
2800
2801     /* XXX: not ANSI compliant: bound checking says error */
2802     vtop = vstack - 1;
2803     s1->pack_stack[0] = 0;
2804     s1->pack_stack_ptr = s1->pack_stack;
2805 }
2806
2807 void preprocess_new()
2808 {
2809     int i, c;
2810     const char *p, *r;
2811     TokenSym *ts;
2812
2813     /* init isid table */
2814     for(i=CH_EOF;i<256;i++)
2815         isidnum_table[i-CH_EOF] = isid(i) || isnum(i);
2816
2817     /* add all tokens */
2818     table_ident = NULL;
2819     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
2820
2821     tok_ident = TOK_IDENT;
2822     p = tcc_keywords;
2823     while (*p) {
2824         r = p;
2825         for(;;) {
2826             c = *r++;
2827             if (c == '\0')
2828                 break;
2829         }
2830         ts = tok_alloc(p, r - p - 1);
2831         p = r;
2832     }
2833 }
2834
2835 /* Preprocess the current file */
2836 static int tcc_preprocess(TCCState *s1)
2837 {
2838     Sym *define_start;
2839     BufferedFile *file_ref, **iptr, **iptr_new;
2840     int token_seen, line_ref, d;
2841     const char *s;
2842
2843     preprocess_init(s1);
2844     define_start = define_stack;
2845     ch = file->buf_ptr[0];
2846     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
2847     parse_flags = PARSE_FLAG_ASM_COMMENTS | PARSE_FLAG_PREPROCESS |
2848         PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
2849     token_seen = 0;
2850     line_ref = 0;
2851     file_ref = NULL;
2852
2853     iptr = s1->include_stack_ptr;
2854     for (;;) {
2855         next();
2856         if (tok == TOK_EOF) {
2857             break;
2858         } else if (file != file_ref) {
2859             goto print_line;
2860         } else if (tok == TOK_LINEFEED) {
2861             if (!token_seen)
2862                 continue;
2863             ++line_ref;
2864             token_seen = 0;
2865         } else if (!token_seen) {
2866             d = file->line_num - line_ref;
2867             if (file != file_ref || d < 0 || d >= 8) {
2868 print_line:
2869                 iptr_new = s1->include_stack_ptr;
2870                 s = iptr_new > iptr ? " 1"
2871                   : iptr_new < iptr ? " 2"
2872                   : iptr_new > s1->include_stack ? " 3"
2873                   : ""
2874                   ;
2875                 iptr = iptr_new;
2876                 fprintf(s1->outfile, "# %d \"%s\"%s\n", file->line_num, file->filename, s);
2877             } else {
2878                 while (d)
2879                     fputs("\n", s1->outfile), --d;
2880             }
2881             line_ref = (file_ref = file)->line_num;
2882             token_seen = tok != TOK_LINEFEED;
2883             if (!token_seen)
2884                 continue;
2885         }
2886         fputs(get_tok_str(tok, &tokc), s1->outfile);
2887     }
2888     free_defines(define_start);
2889     return 0;
2890 }
2891