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, (long)str);
 864     s->next = first_arg;
 865     table_ident[v - TOK_IDENT]->sym_define = s;
 866 }
 867
 868 /* undefined a define symbol. Its name is just set to zero */
 869 static void define_undef(Sym *s)
 870 {
 871     int v;
 872     v = s->v;
 873     if (v >= TOK_IDENT && v < tok_ident)
 874         table_ident[v - TOK_IDENT]->sym_define = NULL;
 875     s->v = 0;
 876 }
 877
 878 static inline Sym *define_find(int v)
 879 {
 880     v -= TOK_IDENT;
 881     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
 882         return NULL;
 883     return table_ident[v]->sym_define;
 884 }
 885
 886 /* free define stack until top reaches 'b' */
 887 static void free_defines(Sym *b)
 888 {
 889     Sym *top, *top1;
 890     int v;
 891
 892     top = define_stack;
 893     while (top != b) {
 894         top1 = top->prev;
 895         /* do not free args or predefined defines */
 896         if (top->c)
 897             tok_str_free((int *)top->c);
 898         v = top->v;
 899         if (v >= TOK_IDENT && v < tok_ident)
 900             table_ident[v - TOK_IDENT]->sym_define = NULL;
 901         sym_free(top);
 902         top = top1;
 903     }
 904     define_stack = b;
 905 }
 906
 907 /* label lookup */
 908 static Sym *label_find(int v)
 909 {
 910     v -= TOK_IDENT;
 911     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
 912         return NULL;
 913     return table_ident[v]->sym_label;
 914 }
 915
 916 static Sym *label_push(Sym **ptop, int v, int flags)
 917 {
 918     Sym *s, **ps;
 919     s = sym_push2(ptop, v, 0, 0);
 920     s->r = flags;
 921     ps = &table_ident[v - TOK_IDENT]->sym_label;
 922     if (ptop == &global_label_stack) {
 923         /* modify the top most local identifier, so that
 924            sym_identifier will point to 's' when popped */
 925         while (*ps != NULL)
 926             ps = &(*ps)->prev_tok;
 927     }
 928     s->prev_tok = *ps;
 929     *ps = s;
 930     return s;
 931 }
 932
 933 /* pop labels until element last is reached. Look if any labels are
 934    undefined. Define symbols if '&&label' was used. */
 935 static void label_pop(Sym **ptop, Sym *slast)
 936 {
 937     Sym *s, *s1;
 938     for(s = *ptop; s != slast; s = s1) {
 939         s1 = s->prev;
 940         if (s->r == LABEL_DECLARED) {
 941             warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
 942         } else if (s->r == LABEL_FORWARD) {
 943                 error("label '%s' used but not defined",
 944                       get_tok_str(s->v, NULL));
 945         } else {
 946             if (s->c) {
 947                 /* define corresponding symbol. A size of
 948                    1 is put. */
 949                 put_extern_sym(s, cur_text_section, (long)s->next, 1);
 950             }
 951         }
 952         /* remove label */
 953         table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
 954         sym_free(s);
 955     }
 956     *ptop = slast;
 957 }
 958
 959 /* eval an expression for #if/#elif */
 960 static int expr_preprocess(void)
 961 {
 962     int c, t;
 963     TokenString str;
 964
 965     tok_str_new(&str);
 966     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
 967         next(); /* do macro subst */
 968         if (tok == TOK_DEFINED) {
 969             next_nomacro();
 970             t = tok;
 971             if (t == '(')
 972                 next_nomacro();
 973             c = define_find(tok) != 0;
 974             if (t == '(')
 975                 next_nomacro();
 976             tok = TOK_CINT;
 977             tokc.i = c;
 978         } else if (tok >= TOK_IDENT) {
 979             /* if undefined macro */
 980             tok = TOK_CINT;
 981             tokc.i = 0;
 982         }
 983         tok_str_add_tok(&str);
 984     }
 985     tok_str_add(&str, -1); /* simulate end of file */
 986     tok_str_add(&str, 0);
 987     /* now evaluate C constant expression */
 988     macro_ptr = str.str;
 989     next();
 990     c = expr_const();
 991     macro_ptr = NULL;
 992     tok_str_free(str.str);
 993     return c != 0;
 994 }
 995
 996 #if defined(PARSE_DEBUG) || defined(PP_DEBUG)
 997 static void tok_print(int *str)
 998 {
 999     int t;
1000     CValue cval;
1001
1002     printf("<");
1003     while (1) {
1004         TOK_GET(t, str, cval);
1005         if (!t)
1006             break;
1007         printf("%s", get_tok_str(t, &cval));
1008     }
1009     printf(">\n");
1010 }
1011 #endif
1012
1013 /* parse after #define */
1014 static void parse_define(void)
1015 {
1016     Sym *s, *first, **ps;
1017     int v, t, varg, is_vaargs, spc;
1018     TokenString str;
1019
1020     v = tok;
1021     if (v < TOK_IDENT)
1022         error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1023     /* XXX: should check if same macro (ANSI) */
1024     first = NULL;
1025     t = MACRO_OBJ;
1026     /* '(' must be just after macro definition for MACRO_FUNC */
1027     next_nomacro_spc();
1028     if (tok == '(') {
1029         next_nomacro();
1030         ps = &first;
1031         while (tok != ')') {
1032             varg = tok;
1033             next_nomacro();
1034             is_vaargs = 0;
1035             if (varg == TOK_DOTS) {
1036                 varg = TOK___VA_ARGS__;
1037                 is_vaargs = 1;
1038             } else if (tok == TOK_DOTS && gnu_ext) {
1039                 is_vaargs = 1;
1040                 next_nomacro();
1041             }
1042             if (varg < TOK_IDENT)
1043                 error("badly punctuated parameter list");
1044             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1045             *ps = s;
1046             ps = &s->next;
1047             if (tok != ',')
1048                 break;
1049             next_nomacro();
1050         }
1051         if (tok == ')')
1052             next_nomacro_spc();
1053         t = MACRO_FUNC;
1054     }
1055     tok_str_new(&str);
1056     spc = 2;
1057     /* EOF testing necessary for '-D' handling */
1058     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1059         /* remove spaces around ## and after '#' */
1060         if (TOK_TWOSHARPS == tok) {
1061             if (1 == spc)
1062                 --str.len;
1063             spc = 2;
1064         } else if ('#' == tok) {
1065             spc = 2;
1066         } else if (check_space(tok, &spc)) {
1067             goto skip;
1068         }
1069         tok_str_add2(&str, tok, &tokc);
1070     skip:
1071         next_nomacro_spc();
1072     }
1073     if (spc == 1)
1074         --str.len; /* remove trailing space */
1075     tok_str_add(&str, 0);
1076 #ifdef PP_DEBUG
1077     printf("define %s %d: ", get_tok_str(v, NULL), t);
1078     tok_print(str.str);
1079 #endif
1080     define_push(v, t, str.str, first);
1081 }
1082
1083 static inline int hash_cached_include(int type, const char *filename)
1084 {
1085     const unsigned char *s;
1086     unsigned int h;
1087
1088     h = TOK_HASH_INIT;
1089     h = TOK_HASH_FUNC(h, type);
1090     s = filename;
1091     while (*s) {
1092         h = TOK_HASH_FUNC(h, *s);
1093         s++;
1094     }
1095     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1096     return h;
1097 }
1098
1099 /* XXX: use a token or a hash table to accelerate matching ? */
1100 static CachedInclude *search_cached_include(TCCState *s1,
1101                                             int type, const char *filename)
1102 {
1103     CachedInclude *e;
1104     int i, h;
1105     h = hash_cached_include(type, filename);
1106     i = s1->cached_includes_hash[h];
1107     for(;;) {
1108         if (i == 0)
1109             break;
1110         e = s1->cached_includes[i - 1];
1111         if (e->type == type && !PATHCMP(e->filename, filename))
1112             return e;
1113         i = e->hash_next;
1114     }
1115     return NULL;
1116 }
1117
1118 static inline void add_cached_include(TCCState *s1, int type,
1119                                       const char *filename, int ifndef_macro)
1120 {
1121     CachedInclude *e;
1122     int h;
1123
1124     if (search_cached_include(s1, type, filename))
1125         return;
1126 #ifdef INC_DEBUG
1127     printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1128 #endif
1129     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1130     if (!e)
1131         return;
1132     e->type = type;
1133     strcpy(e->filename, filename);
1134     e->ifndef_macro = ifndef_macro;
1135     dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1136     /* add in hash table */
1137     h = hash_cached_include(type, filename);
1138     e->hash_next = s1->cached_includes_hash[h];
1139     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1140 }
1141
1142 static void pragma_parse(TCCState *s1)
1143 {
1144     int val;
1145
1146     next();
1147     if (tok == TOK_pack) {
1148         /*
1149           This may be:
1150           #pragma pack(1) // set
1151           #pragma pack() // reset to default
1152           #pragma pack(push,1) // push & set
1153           #pragma pack(pop) // restore previous
1154         */
1155         next();
1156         skip('(');
1157         if (tok == TOK_ASM_pop) {
1158             next();
1159             if (s1->pack_stack_ptr <= s1->pack_stack) {
1160             stk_error:
1161                 error("out of pack stack");
1162             }
1163             s1->pack_stack_ptr--;
1164         } else {
1165             val = 0;
1166             if (tok != ')') {
1167                 if (tok == TOK_ASM_push) {
1168                     next();
1169                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1170                         goto stk_error;
1171                     s1->pack_stack_ptr++;
1172                     skip(',');
1173                 }
1174                 if (tok != TOK_CINT) {
1175                 pack_error:
1176                     error("invalid pack pragma");
1177                 }
1178                 val = tokc.i;
1179                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1180                     goto pack_error;
1181                 next();
1182             }
1183             *s1->pack_stack_ptr = val;
1184             skip(')');
1185         }
1186     }
1187 }
1188
1189 /* is_bof is true if first non space token at beginning of file */
1190 static void preprocess(int is_bof)
1191 {
1192     TCCState *s1 = tcc_state;
1193     int size, i, c, n, saved_parse_flags;
1194     char buf[1024], *q;
1195     char buf1[1024];
1196     BufferedFile *f;
1197     Sym *s;
1198     CachedInclude *e;
1199
1200     saved_parse_flags = parse_flags;
1201     parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM |
1202         PARSE_FLAG_LINEFEED;
1203     next_nomacro();
1204  redo:
1205     switch(tok) {
1206     case TOK_DEFINE:
1207         next_nomacro();
1208         parse_define();
1209         break;
1210     case TOK_UNDEF:
1211         next_nomacro();
1212         s = define_find(tok);
1213         /* undefine symbol by putting an invalid name */
1214         if (s)
1215             define_undef(s);
1216         break;
1217     case TOK_INCLUDE:
1218     case TOK_INCLUDE_NEXT:
1219         ch = file->buf_ptr[0];
1220         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1221         skip_spaces();
1222         if (ch == '<') {
1223             c = '>';
1224             goto read_name;
1225         } else if (ch == '\"') {
1226             c = ch;
1227         read_name:
1228             inp();
1229             q = buf;
1230             while (ch != c && ch != '\n' && ch != CH_EOF) {
1231                 if ((q - buf) < sizeof(buf) - 1)
1232                     *q++ = ch;
1233                 if (ch == '\\') {
1234                     if (handle_stray_noerror() == 0)
1235                         --q;
1236                 } else
1237                     inp();
1238             }
1239             *q = '\0';
1240             minp();
1241 #if 0
1242             /* eat all spaces and comments after include */
1243             /* XXX: slightly incorrect */
1244             while (ch1 != '\n' && ch1 != CH_EOF)
1245                 inp();
1246 #endif
1247         } else {
1248             /* computed #include : either we have only strings or
1249                we have anything enclosed in '<>' */
1250             next();
1251             buf[0] = '\0';
1252             if (tok == TOK_STR) {
1253                 while (tok != TOK_LINEFEED) {
1254                     if (tok != TOK_STR) {
1255                     include_syntax:
1256                         error("'#include' expects \"FILENAME\" or <FILENAME>");
1257                     }
1258                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1259                     next();
1260                 }
1261                 c = '\"';
1262             } else {
1263                 int len;
1264                 while (tok != TOK_LINEFEED) {
1265                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1266                     next();
1267                 }
1268                 len = strlen(buf);
1269                 /* check syntax and remove '<>' */
1270                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1271                     goto include_syntax;
1272                 memmove(buf, buf + 1, len - 2);
1273                 buf[len - 2] = '\0';
1274                 c = '>';
1275             }
1276         }
1277
1278         e = search_cached_include(s1, c, buf);
1279         if (e && define_find(e->ifndef_macro)) {
1280             /* no need to parse the include because the 'ifndef macro'
1281                is defined */
1282 #ifdef INC_DEBUG
1283             printf("%s: skipping %s\n", file->filename, buf);
1284 #endif
1285         } else {
1286             if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1287                 error("#include recursion too deep");
1288             /* push current file in stack */
1289             /* XXX: fix current line init */
1290             *s1->include_stack_ptr++ = file;
1291
1292             /* check absolute include path */
1293             if (IS_ABSPATH(buf)) {
1294                 f = tcc_open(s1, buf);
1295                 if (f)
1296                     goto found;
1297             }
1298             if (c == '\"') {
1299                 /* first search in current dir if "header.h" */
1300                 size = tcc_basename(file->filename) - file->filename;
1301                 if (size > sizeof(buf1) - 1)
1302                     size = sizeof(buf1) - 1;
1303                 memcpy(buf1, file->filename, size);
1304                 buf1[size] = '\0';
1305                 pstrcat(buf1, sizeof(buf1), buf);
1306                 f = tcc_open(s1, buf1);
1307                 if (f) {
1308                     if (tok == TOK_INCLUDE_NEXT)
1309                         tok = TOK_INCLUDE;
1310                     else
1311                         goto found;
1312                 }
1313             }
1314             /* now search in all the include paths */
1315             n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1316             for(i = 0; i < n; i++) {
1317                 const char *path;
1318                 if (i < s1->nb_include_paths)
1319                     path = s1->include_paths[i];
1320                 else
1321                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1322                 pstrcpy(buf1, sizeof(buf1), path);
1323                 pstrcat(buf1, sizeof(buf1), "/");
1324                 pstrcat(buf1, sizeof(buf1), buf);
1325                 f = tcc_open(s1, buf1);
1326                 if (f) {
1327                     if (tok == TOK_INCLUDE_NEXT)
1328                         tok = TOK_INCLUDE;
1329                     else
1330                         goto found;
1331                 }
1332             }
1333             --s1->include_stack_ptr;
1334             error("include file '%s' not found", buf);
1335             break;
1336         found:
1337 #ifdef INC_DEBUG
1338             printf("%s: including %s\n", file->filename, buf1);
1339 #endif
1340             f->inc_type = c;
1341             pstrcpy(f->inc_filename, sizeof(f->inc_filename), buf);
1342             file = f;
1343             /* add include file debug info */
1344             if (tcc_state->do_debug) {
1345                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1346             }
1347             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1348             ch = file->buf_ptr[0];
1349             goto the_end;
1350         }
1351         break;
1352     case TOK_IFNDEF:
1353         c = 1;
1354         goto do_ifdef;
1355     case TOK_IF:
1356         c = expr_preprocess();
1357         goto do_if;
1358     case TOK_IFDEF:
1359         c = 0;
1360     do_ifdef:
1361         next_nomacro();
1362         if (tok < TOK_IDENT)
1363             error("invalid argument for '#if%sdef'", c ? "n" : "");
1364         if (is_bof) {
1365             if (c) {
1366 #ifdef INC_DEBUG
1367                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1368 #endif
1369                 file->ifndef_macro = tok;
1370             }
1371         }
1372         c = (define_find(tok) != 0) ^ c;
1373     do_if:
1374         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1375             error("memory full");
1376         *s1->ifdef_stack_ptr++ = c;
1377         goto test_skip;
1378     case TOK_ELSE:
1379         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1380             error("#else without matching #if");
1381         if (s1->ifdef_stack_ptr[-1] & 2)
1382             error("#else after #else");
1383         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1384         goto test_skip;
1385     case TOK_ELIF:
1386         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1387             error("#elif without matching #if");
1388         c = s1->ifdef_stack_ptr[-1];
1389         if (c > 1)
1390             error("#elif after #else");
1391         /* last #if/#elif expression was true: we skip */
1392         if (c == 1)
1393             goto skip;
1394         c = expr_preprocess();
1395         s1->ifdef_stack_ptr[-1] = c;
1396     test_skip:
1397         if (!(c & 1)) {
1398         skip:
1399             preprocess_skip();
1400             is_bof = 0;
1401             goto redo;
1402         }
1403         break;
1404     case TOK_ENDIF:
1405         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1406             error("#endif without matching #if");
1407         s1->ifdef_stack_ptr--;
1408         /* '#ifndef macro' was at the start of file. Now we check if
1409            an '#endif' is exactly at the end of file */
1410         if (file->ifndef_macro &&
1411             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1412             file->ifndef_macro_saved = file->ifndef_macro;
1413             /* need to set to zero to avoid false matches if another
1414                #ifndef at middle of file */
1415             file->ifndef_macro = 0;
1416             while (tok != TOK_LINEFEED)
1417                 next_nomacro();
1418             tok_flags |= TOK_FLAG_ENDIF;
1419             goto the_end;
1420         }
1421         break;
1422     case TOK_LINE:
1423         next();
1424         if (tok != TOK_CINT)
1425             error("#line");
1426         file->line_num = tokc.i - 1; /* the line number will be incremented after */
1427         next();
1428         if (tok != TOK_LINEFEED) {
1429             if (tok != TOK_STR)
1430                 error("#line");
1431             pstrcpy(file->filename, sizeof(file->filename),
1432                     (char *)tokc.cstr->data);
1433         }
1434         break;
1435     case TOK_ERROR:
1436     case TOK_WARNING:
1437         c = tok;
1438         ch = file->buf_ptr[0];
1439         skip_spaces();
1440         q = buf;
1441         while (ch != '\n' && ch != CH_EOF) {
1442             if ((q - buf) < sizeof(buf) - 1)
1443                 *q++ = ch;
1444             if (ch == '\\') {
1445                 if (handle_stray_noerror() == 0)
1446                     --q;
1447             } else
1448                 inp();
1449         }
1450         *q = '\0';
1451         if (c == TOK_ERROR)
1452             error("#error %s", buf);
1453         else
1454             warning("#warning %s", buf);
1455         break;
1456     case TOK_PRAGMA:
1457         pragma_parse(s1);
1458         break;
1459     default:
1460         if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_CINT) {
1461             /* '!' is ignored to allow C scripts. numbers are ignored
1462                to emulate cpp behaviour */
1463         } else {
1464             if (!(saved_parse_flags & PARSE_FLAG_ASM_COMMENTS))
1465                 warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1466         }
1467         break;
1468     }
1469     /* ignore other preprocess commands or #! for C scripts */
1470     while (tok != TOK_LINEFEED)
1471         next_nomacro();
1472  the_end:
1473     parse_flags = saved_parse_flags;
1474 }
1475
1476 /* evaluate escape codes in a string. */
1477 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1478 {
1479     int c, n;
1480     const uint8_t *p;
1481
1482     p = buf;
1483     for(;;) {
1484         c = *p;
1485         if (c == '\0')
1486             break;
1487         if (c == '\\') {
1488             p++;
1489             /* escape */
1490             c = *p;
1491             switch(c) {
1492             case '0': case '1': case '2': case '3':
1493             case '4': case '5': case '6': case '7':
1494                 /* at most three octal digits */
1495                 n = c - '0';
1496                 p++;
1497                 c = *p;
1498                 if (isoct(c)) {
1499                     n = n * 8 + c - '0';
1500                     p++;
1501                     c = *p;
1502                     if (isoct(c)) {
1503                         n = n * 8 + c - '0';
1504                         p++;
1505                     }
1506                 }
1507                 c = n;
1508                 goto add_char_nonext;
1509             case 'x':
1510             case 'u':
1511             case 'U':
1512                 p++;
1513                 n = 0;
1514                 for(;;) {
1515                     c = *p;
1516                     if (c >= 'a' && c <= 'f')
1517                         c = c - 'a' + 10;
1518                     else if (c >= 'A' && c <= 'F')
1519                         c = c - 'A' + 10;
1520                     else if (isnum(c))
1521                         c = c - '0';
1522                     else
1523                         break;
1524                     n = n * 16 + c;
1525                     p++;
1526                 }
1527                 c = n;
1528                 goto add_char_nonext;
1529             case 'a':
1530                 c = '\a';
1531                 break;
1532             case 'b':
1533                 c = '\b';
1534                 break;
1535             case 'f':
1536                 c = '\f';
1537                 break;
1538             case 'n':
1539                 c = '\n';
1540                 break;
1541             case 'r':
1542                 c = '\r';
1543                 break;
1544             case 't':
1545                 c = '\t';
1546                 break;
1547             case 'v':
1548                 c = '\v';
1549                 break;
1550             case 'e':
1551                 if (!gnu_ext)
1552                     goto invalid_escape;
1553                 c = 27;
1554                 break;
1555             case '\'':
1556             case '\"':
1557             case '\\':
1558             case '?':
1559                 break;
1560             default:
1561             invalid_escape:
1562                 if (c >= '!' && c <= '~')
1563                     warning("unknown escape sequence: \'\\%c\'", c);
1564                 else
1565                     warning("unknown escape sequence: \'\\x%x\'", c);
1566                 break;
1567             }
1568         }
1569         p++;
1570     add_char_nonext:
1571         if (!is_long)
1572             cstr_ccat(outstr, c);
1573         else
1574             cstr_wccat(outstr, c);
1575     }
1576     /* add a trailing '\0' */
1577     if (!is_long)
1578         cstr_ccat(outstr, '\0');
1579     else
1580         cstr_wccat(outstr, '\0');
1581 }
1582
1583 /* we use 64 bit numbers */
1584 #define BN_SIZE 2
1585
1586 /* bn = (bn << shift) | or_val */
1587 void bn_lshift(unsigned int *bn, int shift, int or_val)
1588 {
1589     int i;
1590     unsigned int v;
1591     for(i=0;i<BN_SIZE;i++) {
1592         v = bn[i];
1593         bn[i] = (v << shift) | or_val;
1594         or_val = v >> (32 - shift);
1595     }
1596 }
1597
1598 void bn_zero(unsigned int *bn)
1599 {
1600     int i;
1601     for(i=0;i<BN_SIZE;i++) {
1602         bn[i] = 0;
1603     }
1604 }
1605
1606 /* parse number in null terminated string 'p' and return it in the
1607    current token */
1608 void parse_number(const char *p)
1609 {
1610     int b, t, shift, frac_bits, s, exp_val, ch;
1611     char *q;
1612     unsigned int bn[BN_SIZE];
1613     double d;
1614
1615     /* number */
1616     q = token_buf;
1617     ch = *p++;
1618     t = ch;
1619     ch = *p++;
1620     *q++ = t;
1621     b = 10;
1622     if (t == '.') {
1623         goto float_frac_parse;
1624     } else if (t == '0') {
1625         if (ch == 'x' || ch == 'X') {
1626             q--;
1627             ch = *p++;
1628             b = 16;
1629         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
1630             q--;
1631             ch = *p++;
1632             b = 2;
1633         }
1634     }
1635     /* parse all digits. cannot check octal numbers at this stage
1636        because of floating point constants */
1637     while (1) {
1638         if (ch >= 'a' && ch <= 'f')
1639             t = ch - 'a' + 10;
1640         else if (ch >= 'A' && ch <= 'F')
1641             t = ch - 'A' + 10;
1642         else if (isnum(ch))
1643             t = ch - '0';
1644         else
1645             break;
1646         if (t >= b)
1647             break;
1648         if (q >= token_buf + STRING_MAX_SIZE) {
1649         num_too_long:
1650             error("number too long");
1651         }
1652         *q++ = ch;
1653         ch = *p++;
1654     }
1655     if (ch == '.' ||
1656         ((ch == 'e' || ch == 'E') && b == 10) ||
1657         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
1658         if (b != 10) {
1659             /* NOTE: strtox should support that for hexa numbers, but
1660                non ISOC99 libcs do not support it, so we prefer to do
1661                it by hand */
1662             /* hexadecimal or binary floats */
1663             /* XXX: handle overflows */
1664             *q = '\0';
1665             if (b == 16)
1666                 shift = 4;
1667             else
1668                 shift = 2;
1669             bn_zero(bn);
1670             q = token_buf;
1671             while (1) {
1672                 t = *q++;
1673                 if (t == '\0') {
1674                     break;
1675                 } else if (t >= 'a') {
1676                     t = t - 'a' + 10;
1677                 } else if (t >= 'A') {
1678                     t = t - 'A' + 10;
1679                 } else {
1680                     t = t - '0';
1681                 }
1682                 bn_lshift(bn, shift, t);
1683             }
1684             frac_bits = 0;
1685             if (ch == '.') {
1686                 ch = *p++;
1687                 while (1) {
1688                     t = ch;
1689                     if (t >= 'a' && t <= 'f') {
1690                         t = t - 'a' + 10;
1691                     } else if (t >= 'A' && t <= 'F') {
1692                         t = t - 'A' + 10;
1693                     } else if (t >= '0' && t <= '9') {
1694                         t = t - '0';
1695                     } else {
1696                         break;
1697                     }
1698                     if (t >= b)
1699                         error("invalid digit");
1700                     bn_lshift(bn, shift, t);
1701                     frac_bits += shift;
1702                     ch = *p++;
1703                 }
1704             }
1705             if (ch != 'p' && ch != 'P')
1706                 expect("exponent");
1707             ch = *p++;
1708             s = 1;
1709             exp_val = 0;
1710             if (ch == '+') {
1711                 ch = *p++;
1712             } else if (ch == '-') {
1713                 s = -1;
1714                 ch = *p++;
1715             }
1716             if (ch < '0' || ch > '9')
1717                 expect("exponent digits");
1718             while (ch >= '0' && ch <= '9') {
1719                 exp_val = exp_val * 10 + ch - '0';
1720                 ch = *p++;
1721             }
1722             exp_val = exp_val * s;
1723
1724             /* now we can generate the number */
1725             /* XXX: should patch directly float number */
1726             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
1727             d = ldexp(d, exp_val - frac_bits);
1728             t = toup(ch);
1729             if (t == 'F') {
1730                 ch = *p++;
1731                 tok = TOK_CFLOAT;
1732                 /* float : should handle overflow */
1733                 tokc.f = (float)d;
1734             } else if (t == 'L') {
1735                 ch = *p++;
1736                 tok = TOK_CLDOUBLE;
1737                 /* XXX: not large enough */
1738                 tokc.ld = (long double)d;
1739             } else {
1740                 tok = TOK_CDOUBLE;
1741                 tokc.d = d;
1742             }
1743         } else {
1744             /* decimal floats */
1745             if (ch == '.') {
1746                 if (q >= token_buf + STRING_MAX_SIZE)
1747                     goto num_too_long;
1748                 *q++ = ch;
1749                 ch = *p++;
1750             float_frac_parse:
1751                 while (ch >= '0' && ch <= '9') {
1752                     if (q >= token_buf + STRING_MAX_SIZE)
1753                         goto num_too_long;
1754                     *q++ = ch;
1755                     ch = *p++;
1756                 }
1757             }
1758             if (ch == 'e' || ch == 'E') {
1759                 if (q >= token_buf + STRING_MAX_SIZE)
1760                     goto num_too_long;
1761                 *q++ = ch;
1762                 ch = *p++;
1763                 if (ch == '-' || ch == '+') {
1764                     if (q >= token_buf + STRING_MAX_SIZE)
1765                         goto num_too_long;
1766                     *q++ = ch;
1767                     ch = *p++;
1768                 }
1769                 if (ch < '0' || ch > '9')
1770                     expect("exponent digits");
1771                 while (ch >= '0' && ch <= '9') {
1772                     if (q >= token_buf + STRING_MAX_SIZE)
1773                         goto num_too_long;
1774                     *q++ = ch;
1775                     ch = *p++;
1776                 }
1777             }
1778             *q = '\0';
1779             t = toup(ch);
1780             errno = 0;
1781             if (t == 'F') {
1782                 ch = *p++;
1783                 tok = TOK_CFLOAT;
1784                 tokc.f = strtof(token_buf, NULL);
1785             } else if (t == 'L') {
1786                 ch = *p++;
1787                 tok = TOK_CLDOUBLE;
1788                 tokc.ld = strtold(token_buf, NULL);
1789             } else {
1790                 tok = TOK_CDOUBLE;
1791                 tokc.d = strtod(token_buf, NULL);
1792             }
1793         }
1794     } else {
1795         unsigned long long n, n1;
1796         int lcount, ucount;
1797
1798         /* integer number */
1799         *q = '\0';
1800         q = token_buf;
1801         if (b == 10 && *q == '0') {
1802             b = 8;
1803             q++;
1804         }
1805         n = 0;
1806         while(1) {
1807             t = *q++;
1808             /* no need for checks except for base 10 / 8 errors */
1809             if (t == '\0') {
1810                 break;
1811             } else if (t >= 'a') {
1812                 t = t - 'a' + 10;
1813             } else if (t >= 'A') {
1814                 t = t - 'A' + 10;
1815             } else {
1816                 t = t - '0';
1817                 if (t >= b)
1818                     error("invalid digit");
1819             }
1820             n1 = n;
1821             n = n * b + t;
1822             /* detect overflow */
1823             /* XXX: this test is not reliable */
1824             if (n < n1)
1825                 error("integer constant overflow");
1826         }
1827
1828         /* XXX: not exactly ANSI compliant */
1829         if ((n & 0xffffffff00000000LL) != 0) {
1830             if ((n >> 63) != 0)
1831                 tok = TOK_CULLONG;
1832             else
1833                 tok = TOK_CLLONG;
1834         } else if (n > 0x7fffffff) {
1835             tok = TOK_CUINT;
1836         } else {
1837             tok = TOK_CINT;
1838         }
1839         lcount = 0;
1840         ucount = 0;
1841         for(;;) {
1842             t = toup(ch);
1843             if (t == 'L') {
1844                 if (lcount >= 2)
1845                     error("three 'l's in integer constant");
1846                 lcount++;
1847                 if (lcount == 2) {
1848                     if (tok == TOK_CINT)
1849                         tok = TOK_CLLONG;
1850                     else if (tok == TOK_CUINT)
1851                         tok = TOK_CULLONG;
1852                 }
1853                 ch = *p++;
1854             } else if (t == 'U') {
1855                 if (ucount >= 1)
1856                     error("two 'u's in integer constant");
1857                 ucount++;
1858                 if (tok == TOK_CINT)
1859                     tok = TOK_CUINT;
1860                 else if (tok == TOK_CLLONG)
1861                     tok = TOK_CULLONG;
1862                 ch = *p++;
1863             } else {
1864                 break;
1865             }
1866         }
1867         if (tok == TOK_CINT || tok == TOK_CUINT)
1868             tokc.ui = n;
1869         else
1870             tokc.ull = n;
1871     }
1872     if (ch)
1873         error("invalid number\n");
1874 }
1875
1876
1877 #define PARSE2(c1, tok1, c2, tok2)              \
1878     case c1:                                    \
1879         PEEKC(c, p);                            \
1880         if (c == c2) {                          \
1881             p++;                                \
1882             tok = tok2;                         \
1883         } else {                                \
1884             tok = tok1;                         \
1885         }                                       \
1886         break;
1887
1888 /* return next token without macro substitution */
1889 static inline void next_nomacro1(void)
1890 {
1891     int t, c, is_long;
1892     TokenSym *ts;
1893     uint8_t *p, *p1;
1894     unsigned int h;
1895
1896     p = file->buf_ptr;
1897  redo_no_start:
1898     c = *p;
1899     switch(c) {
1900     case ' ':
1901     case '\t':
1902         tok = c;
1903         p++;
1904         goto keep_tok_flags;
1905     case '\f':
1906     case '\v':
1907     case '\r':
1908         p++;
1909         goto redo_no_start;
1910     case '\\':
1911         /* first look if it is in fact an end of buffer */
1912         if (p >= file->buf_end) {
1913             file->buf_ptr = p;
1914             handle_eob();
1915             p = file->buf_ptr;
1916             if (p >= file->buf_end)
1917                 goto parse_eof;
1918             else
1919                 goto redo_no_start;
1920         } else {
1921             file->buf_ptr = p;
1922             ch = *p;
1923             handle_stray();
1924             p = file->buf_ptr;
1925             goto redo_no_start;
1926         }
1927     parse_eof:
1928         {
1929             TCCState *s1 = tcc_state;
1930             if ((parse_flags & PARSE_FLAG_LINEFEED)
1931                 && !(tok_flags & TOK_FLAG_EOF)) {
1932                 tok_flags |= TOK_FLAG_EOF;
1933                 tok = TOK_LINEFEED;
1934                 goto keep_tok_flags;
1935             } else if (s1->include_stack_ptr == s1->include_stack ||
1936                        !(parse_flags & PARSE_FLAG_PREPROCESS)) {
1937                 /* no include left : end of file. */
1938                 tok = TOK_EOF;
1939             } else {
1940                 tok_flags &= ~TOK_FLAG_EOF;
1941                 /* pop include file */
1942
1943                 /* test if previous '#endif' was after a #ifdef at
1944                    start of file */
1945                 if (tok_flags & TOK_FLAG_ENDIF) {
1946 #ifdef INC_DEBUG
1947                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
1948 #endif
1949                     add_cached_include(s1, file->inc_type, file->inc_filename,
1950                                        file->ifndef_macro_saved);
1951                 }
1952
1953                 /* add end of include file debug info */
1954                 if (tcc_state->do_debug) {
1955                     put_stabd(N_EINCL, 0, 0);
1956                 }
1957                 /* pop include stack */
1958                 tcc_close(file);
1959                 s1->include_stack_ptr--;
1960                 file = *s1->include_stack_ptr;
1961                 p = file->buf_ptr;
1962                 goto redo_no_start;
1963             }
1964         }
1965         break;
1966
1967     case '\n':
1968         file->line_num++;
1969         tok_flags |= TOK_FLAG_BOL;
1970         p++;
1971         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
1972             goto redo_no_start;
1973         tok = TOK_LINEFEED;
1974         goto keep_tok_flags;
1975
1976     case '#':
1977         /* XXX: simplify */
1978         PEEKC(c, p);
1979         if ((tok_flags & TOK_FLAG_BOL) &&
1980             (parse_flags & PARSE_FLAG_PREPROCESS)) {
1981             file->buf_ptr = p;
1982             preprocess(tok_flags & TOK_FLAG_BOF);
1983             p = file->buf_ptr;
1984             goto redo_no_start;
1985         } else {
1986             if (c == '#') {
1987                 p++;
1988                 tok = TOK_TWOSHARPS;
1989             } else {
1990                 if (parse_flags & PARSE_FLAG_ASM_COMMENTS) {
1991                     p = parse_line_comment(p - 1);
1992                     goto redo_no_start;
1993                 } else {
1994                     tok = '#';
1995                 }
1996             }
1997         }
1998         break;
1999
2000     case 'a': case 'b': case 'c': case 'd':
2001     case 'e': case 'f': case 'g': case 'h':
2002     case 'i': case 'j': case 'k': case 'l':
2003     case 'm': case 'n': case 'o': case 'p':
2004     case 'q': case 'r': case 's': case 't':
2005     case 'u': case 'v': case 'w': case 'x':
2006     case 'y': case 'z':
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':
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 '_':
2015     parse_ident_fast:
2016         p1 = p;
2017         h = TOK_HASH_INIT;
2018         h = TOK_HASH_FUNC(h, c);
2019         p++;
2020         for(;;) {
2021             c = *p;
2022             if (!isidnum_table[c-CH_EOF])
2023                 break;
2024             h = TOK_HASH_FUNC(h, c);
2025             p++;
2026         }
2027         if (c != '\\') {
2028             TokenSym **pts;
2029             int len;
2030
2031             /* fast case : no stray found, so we have the full token
2032                and we have already hashed it */
2033             len = p - p1;
2034             h &= (TOK_HASH_SIZE - 1);
2035             pts = &hash_ident[h];
2036             for(;;) {
2037                 ts = *pts;
2038                 if (!ts)
2039                     break;
2040                 if (ts->len == len && !memcmp(ts->str, p1, len))
2041                     goto token_found;
2042                 pts = &(ts->hash_next);
2043             }
2044             ts = tok_alloc_new(pts, p1, len);
2045         token_found: ;
2046         } else {
2047             /* slower case */
2048             cstr_reset(&tokcstr);
2049
2050             while (p1 < p) {
2051                 cstr_ccat(&tokcstr, *p1);
2052                 p1++;
2053             }
2054             p--;
2055             PEEKC(c, p);
2056         parse_ident_slow:
2057             while (isidnum_table[c-CH_EOF]) {
2058                 cstr_ccat(&tokcstr, c);
2059                 PEEKC(c, p);
2060             }
2061             ts = tok_alloc(tokcstr.data, tokcstr.size);
2062         }
2063         tok = ts->tok;
2064         break;
2065     case 'L':
2066         t = p[1];
2067         if (t != '\\' && t != '\'' && t != '\"') {
2068             /* fast case */
2069             goto parse_ident_fast;
2070         } else {
2071             PEEKC(c, p);
2072             if (c == '\'' || c == '\"') {
2073                 is_long = 1;
2074                 goto str_const;
2075             } else {
2076                 cstr_reset(&tokcstr);
2077                 cstr_ccat(&tokcstr, 'L');
2078                 goto parse_ident_slow;
2079             }
2080         }
2081         break;
2082     case '0': case '1': case '2': case '3':
2083     case '4': case '5': case '6': case '7':
2084     case '8': case '9':
2085
2086         cstr_reset(&tokcstr);
2087         /* after the first digit, accept digits, alpha, '.' or sign if
2088            prefixed by 'eEpP' */
2089     parse_num:
2090         for(;;) {
2091             t = c;
2092             cstr_ccat(&tokcstr, c);
2093             PEEKC(c, p);
2094             if (!(isnum(c) || isid(c) || c == '.' ||
2095                   ((c == '+' || c == '-') &&
2096                    (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2097                 break;
2098         }
2099         /* We add a trailing '\0' to ease parsing */
2100         cstr_ccat(&tokcstr, '\0');
2101         tokc.cstr = &tokcstr;
2102         tok = TOK_PPNUM;
2103         break;
2104     case '.':
2105         /* special dot handling because it can also start a number */
2106         PEEKC(c, p);
2107         if (isnum(c)) {
2108             cstr_reset(&tokcstr);
2109             cstr_ccat(&tokcstr, '.');
2110             goto parse_num;
2111         } else if (c == '.') {
2112             PEEKC(c, p);
2113             if (c != '.')
2114                 expect("'.'");
2115             PEEKC(c, p);
2116             tok = TOK_DOTS;
2117         } else {
2118             tok = '.';
2119         }
2120         break;
2121     case '\'':
2122     case '\"':
2123         is_long = 0;
2124     str_const:
2125         {
2126             CString str;
2127             int sep;
2128
2129             sep = c;
2130
2131             /* parse the string */
2132             cstr_new(&str);
2133             p = parse_pp_string(p, sep, &str);
2134             cstr_ccat(&str, '\0');
2135
2136             /* eval the escape (should be done as TOK_PPNUM) */
2137             cstr_reset(&tokcstr);
2138             parse_escape_string(&tokcstr, str.data, is_long);
2139             cstr_free(&str);
2140
2141             if (sep == '\'') {
2142                 int char_size;
2143                 /* XXX: make it portable */
2144                 if (!is_long)
2145                     char_size = 1;
2146                 else
2147                     char_size = sizeof(nwchar_t);
2148                 if (tokcstr.size <= char_size)
2149                     error("empty character constant");
2150                 if (tokcstr.size > 2 * char_size)
2151                     warning("multi-character character constant");
2152                 if (!is_long) {
2153                     tokc.i = *(int8_t *)tokcstr.data;
2154                     tok = TOK_CCHAR;
2155                 } else {
2156                     tokc.i = *(nwchar_t *)tokcstr.data;
2157                     tok = TOK_LCHAR;
2158                 }
2159             } else {
2160                 tokc.cstr = &tokcstr;
2161                 if (!is_long)
2162                     tok = TOK_STR;
2163                 else
2164                     tok = TOK_LSTR;
2165             }
2166         }
2167         break;
2168
2169     case '<':
2170         PEEKC(c, p);
2171         if (c == '=') {
2172             p++;
2173             tok = TOK_LE;
2174         } else if (c == '<') {
2175             PEEKC(c, p);
2176             if (c == '=') {
2177                 p++;
2178                 tok = TOK_A_SHL;
2179             } else {
2180                 tok = TOK_SHL;
2181             }
2182         } else {
2183             tok = TOK_LT;
2184         }
2185         break;
2186
2187     case '>':
2188         PEEKC(c, p);
2189         if (c == '=') {
2190             p++;
2191             tok = TOK_GE;
2192         } else if (c == '>') {
2193             PEEKC(c, p);
2194             if (c == '=') {
2195                 p++;
2196                 tok = TOK_A_SAR;
2197             } else {
2198                 tok = TOK_SAR;
2199             }
2200         } else {
2201             tok = TOK_GT;
2202         }
2203         break;
2204
2205     case '&':
2206         PEEKC(c, p);
2207         if (c == '&') {
2208             p++;
2209             tok = TOK_LAND;
2210         } else if (c == '=') {
2211             p++;
2212             tok = TOK_A_AND;
2213         } else {
2214             tok = '&';
2215         }
2216         break;
2217
2218     case '|':
2219         PEEKC(c, p);
2220         if (c == '|') {
2221             p++;
2222             tok = TOK_LOR;
2223         } else if (c == '=') {
2224             p++;
2225             tok = TOK_A_OR;
2226         } else {
2227             tok = '|';
2228         }
2229         break;
2230
2231     case '+':
2232         PEEKC(c, p);
2233         if (c == '+') {
2234             p++;
2235             tok = TOK_INC;
2236         } else if (c == '=') {
2237             p++;
2238             tok = TOK_A_ADD;
2239         } else {
2240             tok = '+';
2241         }
2242         break;
2243
2244     case '-':
2245         PEEKC(c, p);
2246         if (c == '-') {
2247             p++;
2248             tok = TOK_DEC;
2249         } else if (c == '=') {
2250             p++;
2251             tok = TOK_A_SUB;
2252         } else if (c == '>') {
2253             p++;
2254             tok = TOK_ARROW;
2255         } else {
2256             tok = '-';
2257         }
2258         break;
2259
2260     PARSE2('!', '!', '=', TOK_NE)
2261     PARSE2('=', '=', '=', TOK_EQ)
2262     PARSE2('*', '*', '=', TOK_A_MUL)
2263     PARSE2('%', '%', '=', TOK_A_MOD)
2264     PARSE2('^', '^', '=', TOK_A_XOR)
2265
2266         /* comments or operator */
2267     case '/':
2268         PEEKC(c, p);
2269         if (c == '*') {
2270             p = parse_comment(p);
2271             goto redo_no_start;
2272         } else if (c == '/') {
2273             p = parse_line_comment(p);
2274             goto redo_no_start;
2275         } else if (c == '=') {
2276             p++;
2277             tok = TOK_A_DIV;
2278         } else {
2279             tok = '/';
2280         }
2281         break;
2282
2283         /* simple tokens */
2284     case '(':
2285     case ')':
2286     case '[':
2287     case ']':
2288     case '{':
2289     case '}':
2290     case ',':
2291     case ';':
2292     case ':':
2293     case '?':
2294     case '~':
2295     case '$': /* only used in assembler */
2296     case '@': /* dito */
2297         tok = c;
2298         p++;
2299         break;
2300     default:
2301         error("unrecognized character \\x%02x", c);
2302         break;
2303     }
2304     tok_flags = 0;
2305 keep_tok_flags:
2306     file->buf_ptr = p;
2307 #if defined(PARSE_DEBUG)
2308     printf("token = %s\n", get_tok_str(tok, &tokc));
2309 #endif
2310 }
2311
2312 /* return next token without macro substitution. Can read input from
2313    macro_ptr buffer */
2314 static void next_nomacro_spc(void)
2315 {
2316     if (macro_ptr) {
2317     redo:
2318         tok = *macro_ptr;
2319         if (tok) {
2320             TOK_GET(tok, macro_ptr, tokc);
2321             if (tok == TOK_LINENUM) {
2322                 file->line_num = tokc.i;
2323                 goto redo;
2324             }
2325         }
2326     } else {
2327         next_nomacro1();
2328     }
2329 }
2330
2331 static void next_nomacro(void)
2332 {
2333     do {
2334         next_nomacro_spc();
2335     } while (is_space(tok));
2336 }
2337
2338 /* substitute args in macro_str and return allocated string */
2339 static int *macro_arg_subst(Sym **nested_list, int *macro_str, Sym *args)
2340 {
2341     int *st, last_tok, t, spc;
2342     Sym *s;
2343     CValue cval;
2344     TokenString str;
2345     CString cstr;
2346
2347     tok_str_new(&str);
2348     last_tok = 0;
2349     while(1) {
2350         TOK_GET(t, macro_str, cval);
2351         if (!t)
2352             break;
2353         if (t == '#') {
2354             /* stringize */
2355             TOK_GET(t, macro_str, cval);
2356             if (!t)
2357                 break;
2358             s = sym_find2(args, t);
2359             if (s) {
2360                 cstr_new(&cstr);
2361                 st = (int *)s->c;
2362                 spc = 0;
2363                 while (*st) {
2364                     TOK_GET(t, st, cval);
2365                     if (!check_space(t, &spc))
2366                         cstr_cat(&cstr, get_tok_str(t, &cval));
2367                 }
2368                 cstr.size -= spc;
2369                 cstr_ccat(&cstr, '\0');
2370 #ifdef PP_DEBUG
2371                 printf("stringize: %s\n", (char *)cstr.data);
2372 #endif
2373                 /* add string */
2374                 cval.cstr = &cstr;
2375                 tok_str_add2(&str, TOK_STR, &cval);
2376                 cstr_free(&cstr);
2377             } else {
2378                 tok_str_add2(&str, t, &cval);
2379             }
2380         } else if (t >= TOK_IDENT) {
2381             s = sym_find2(args, t);
2382             if (s) {
2383                 st = (int *)s->c;
2384                 /* if '##' is present before or after, no arg substitution */
2385                 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2386                     /* special case for var arg macros : ## eats the
2387                        ',' if empty VA_ARGS variable. */
2388                     /* XXX: test of the ',' is not 100%
2389                        reliable. should fix it to avoid security
2390                        problems */
2391                     if (gnu_ext && s->type.t &&
2392                         last_tok == TOK_TWOSHARPS &&
2393                         str.len >= 2 && str.str[str.len - 2] == ',') {
2394                         if (*st == 0) {
2395                             /* suppress ',' '##' */
2396                             str.len -= 2;
2397                         } else {
2398                             /* suppress '##' and add variable */
2399                             str.len--;
2400                             goto add_var;
2401                         }
2402                     } else {
2403                         int t1;
2404                     add_var:
2405                         for(;;) {
2406                             TOK_GET(t1, st, cval);
2407                             if (!t1)
2408                                 break;
2409                             tok_str_add2(&str, t1, &cval);
2410                         }
2411                     }
2412                 } else {
2413                     /* NOTE: the stream cannot be read when macro
2414                        substituing an argument */
2415                     macro_subst(&str, nested_list, st, NULL);
2416                 }
2417             } else {
2418                 tok_str_add(&str, t);
2419             }
2420         } else {
2421             tok_str_add2(&str, t, &cval);
2422         }
2423         last_tok = t;
2424     }
2425     tok_str_add(&str, 0);
2426     return str.str;
2427 }
2428
2429 static char const ab_month_name[12][4] =
2430 {
2431     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2432     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2433 };
2434
2435 /* do macro substitution of current token with macro 's' and add
2436    result to (tok_str,tok_len). 'nested_list' is the list of all
2437    macros we got inside to avoid recursing. Return non zero if no
2438    substitution needs to be done */
2439 static int macro_subst_tok(TokenString *tok_str,
2440                            Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2441 {
2442     Sym *args, *sa, *sa1;
2443     int mstr_allocated, parlevel, *mstr, t, t1, *p, spc;
2444     TokenString str;
2445     char *cstrval;
2446     CValue cval;
2447     CString cstr;
2448     char buf[32];
2449
2450     /* if symbol is a macro, prepare substitution */
2451     /* special macros */
2452     if (tok == TOK___LINE__) {
2453         snprintf(buf, sizeof(buf), "%d", file->line_num);
2454         cstrval = buf;
2455         t1 = TOK_PPNUM;
2456         goto add_cstr1;
2457     } else if (tok == TOK___FILE__) {
2458         cstrval = file->filename;
2459         goto add_cstr;
2460     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2461         time_t ti;
2462         struct tm *tm;
2463
2464         time(&ti);
2465         tm = localtime(&ti);
2466         if (tok == TOK___DATE__) {
2467             snprintf(buf, sizeof(buf), "%s %2d %d",
2468                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2469         } else {
2470             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2471                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2472         }
2473         cstrval = buf;
2474     add_cstr:
2475         t1 = TOK_STR;
2476     add_cstr1:
2477         cstr_new(&cstr);
2478         cstr_cat(&cstr, cstrval);
2479         cstr_ccat(&cstr, '\0');
2480         cval.cstr = &cstr;
2481         tok_str_add2(tok_str, t1, &cval);
2482         cstr_free(&cstr);
2483     } else {
2484         mstr = (int *)s->c;
2485         mstr_allocated = 0;
2486         if (s->type.t == MACRO_FUNC) {
2487             /* NOTE: we do not use next_nomacro to avoid eating the
2488                next token. XXX: find better solution */
2489         redo:
2490             if (macro_ptr) {
2491                 p = macro_ptr;
2492                 while (is_space(t = *p) || TOK_LINEFEED == t)
2493                     ++p;
2494                 if (t == 0 && can_read_stream) {
2495                     /* end of macro stream: we must look at the token
2496                        after in the file */
2497                     struct macro_level *ml = *can_read_stream;
2498                     macro_ptr = NULL;
2499                     if (ml)
2500                     {
2501                         macro_ptr = ml->p;
2502                         ml->p = NULL;
2503                         *can_read_stream = ml -> prev;
2504                     }
2505                     goto redo;
2506                 }
2507             } else {
2508                 /* XXX: incorrect with comments */
2509                 ch = file->buf_ptr[0];
2510                 while (is_space(ch) || ch == '\n')
2511                     cinp();
2512                 t = ch;
2513             }
2514             if (t != '(') /* no macro subst */
2515                 return -1;
2516
2517             /* argument macro */
2518             next_nomacro();
2519             next_nomacro();
2520             args = NULL;
2521             sa = s->next;
2522             /* NOTE: empty args are allowed, except if no args */
2523             for(;;) {
2524                 /* handle '()' case */
2525                 if (!args && !sa && tok == ')')
2526                     break;
2527                 if (!sa)
2528                     error("macro '%s' used with too many args",
2529                           get_tok_str(s->v, 0));
2530                 tok_str_new(&str);
2531                 parlevel = spc = 0;
2532                 /* NOTE: non zero sa->t indicates VA_ARGS */
2533                 while ((parlevel > 0 ||
2534                         (tok != ')' &&
2535                          (tok != ',' || sa->type.t))) &&
2536                        tok != -1) {
2537                     if (tok == '(')
2538                         parlevel++;
2539                     else if (tok == ')')
2540                         parlevel--;
2541                     if (tok == TOK_LINEFEED)
2542                         tok = ' ';
2543                     if (!check_space(tok, &spc))
2544                         tok_str_add2(&str, tok, &tokc);
2545                     next_nomacro_spc();
2546                 }
2547                 str.len -= spc;
2548                 tok_str_add(&str, 0);
2549                 sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, (long)str.str);
2550                 sa = sa->next;
2551                 if (tok == ')') {
2552                     /* special case for gcc var args: add an empty
2553                        var arg argument if it is omitted */
2554                     if (sa && sa->type.t && gnu_ext)
2555                         continue;
2556                     else
2557                         break;
2558                 }
2559                 if (tok != ',')
2560                     expect(",");
2561                 next_nomacro();
2562             }
2563             if (sa) {
2564                 error("macro '%s' used with too few args",
2565                       get_tok_str(s->v, 0));
2566             }
2567
2568             /* now subst each arg */
2569             mstr = macro_arg_subst(nested_list, mstr, args);
2570             /* free memory */
2571             sa = args;
2572             while (sa) {
2573                 sa1 = sa->prev;
2574                 tok_str_free((int *)sa->c);
2575                 sym_free(sa);
2576                 sa = sa1;
2577             }
2578             mstr_allocated = 1;
2579         }
2580         sym_push2(nested_list, s->v, 0, 0);
2581         macro_subst(tok_str, nested_list, mstr, can_read_stream);
2582         /* pop nested defined symbol */
2583         sa1 = *nested_list;
2584         *nested_list = sa1->prev;
2585         sym_free(sa1);
2586         if (mstr_allocated)
2587             tok_str_free(mstr);
2588     }
2589     return 0;
2590 }
2591
2592 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2593    return the resulting string (which must be freed). */
2594 static inline int *macro_twosharps(const int *macro_str)
2595 {
2596     TokenSym *ts;
2597     const int *ptr, *saved_macro_ptr;
2598     int t;
2599     const char *p1, *p2;
2600     CValue cval;
2601     TokenString macro_str1;
2602     CString cstr;
2603
2604     /* we search the first '##' */
2605     for(ptr = macro_str;;) {
2606         TOK_GET(t, ptr, cval);
2607         if (t == TOK_TWOSHARPS)
2608             break;
2609         /* nothing more to do if end of string */
2610         if (t == 0)
2611             return NULL;
2612     }
2613
2614     /* we saw '##', so we need more processing to handle it */
2615     cstr_new(&cstr);
2616     tok_str_new(&macro_str1);
2617     saved_macro_ptr = macro_ptr;
2618     /* XXX: get rid of the use of macro_ptr here */
2619     macro_ptr = (int *)macro_str;
2620     for(;;) {
2621         next_nomacro_spc();
2622         if (tok == 0)
2623             break;
2624         if (tok == TOK_TWOSHARPS)
2625             continue;
2626         while (*macro_ptr == TOK_TWOSHARPS) {
2627             t = *++macro_ptr;
2628             if (t && t != TOK_TWOSHARPS) {
2629                 TOK_GET(t, macro_ptr, cval);
2630                 /* We concatenate the two tokens if we have an
2631                    identifier or a preprocessing number */
2632                 cstr_reset(&cstr);
2633                 p1 = get_tok_str(tok, &tokc);
2634                 cstr_cat(&cstr, p1);
2635                 p2 = get_tok_str(t, &cval);
2636                 cstr_cat(&cstr, p2);
2637                 cstr_ccat(&cstr, '\0');
2638
2639                 if ((tok >= TOK_IDENT || tok == TOK_PPNUM) &&
2640                     (t >= TOK_IDENT || t == TOK_PPNUM)) {
2641                     if (tok == TOK_PPNUM) {
2642                         /* if number, then create a number token */
2643                         /* NOTE: no need to allocate because
2644                            tok_str_add2() does it */
2645                         cstr_reset(&tokcstr);
2646                         tokcstr = cstr;
2647                         cstr_new(&cstr);
2648                         tokc.cstr = &tokcstr;
2649                     } else {
2650                         /* if identifier, we must do a test to
2651                            validate we have a correct identifier */
2652                         if (t == TOK_PPNUM) {
2653                             const char *p;
2654                             int c;
2655
2656                             p = p2;
2657                             for(;;) {
2658                                 c = *p;
2659                                 if (c == '\0')
2660                                     break;
2661                                 p++;
2662                                 if (!isnum(c) && !isid(c))
2663                                     goto error_pasting;
2664                             }
2665                         }
2666                         ts = tok_alloc(cstr.data, strlen(cstr.data));
2667                         tok = ts->tok; /* modify current token */
2668                     }
2669                 } else {
2670                     const char *str = cstr.data;
2671                     const unsigned char *q;
2672
2673                     /* we look for a valid token */
2674                     /* XXX: do more extensive checks */
2675                     if (!strcmp(str, ">>=")) {
2676                         tok = TOK_A_SAR;
2677                     } else if (!strcmp(str, "<<=")) {
2678                         tok = TOK_A_SHL;
2679                     } else if (strlen(str) == 2) {
2680                         /* search in two bytes table */
2681                         q = tok_two_chars;
2682                         for(;;) {
2683                             if (!*q)
2684                                 goto error_pasting;
2685                             if (q[0] == str[0] && q[1] == str[1])
2686                                 break;
2687                             q += 3;
2688                         }
2689                         tok = q[2];
2690                     } else {
2691                     error_pasting:
2692                         /* NOTE: because get_tok_str use a static buffer,
2693                            we must save it */
2694                         cstr_reset(&cstr);
2695                         p1 = get_tok_str(tok, &tokc);
2696                         cstr_cat(&cstr, p1);
2697                         cstr_ccat(&cstr, '\0');
2698                         p2 = get_tok_str(t, &cval);
2699                         warning("pasting \"%s\" and \"%s\" does not give a valid preprocessing token", cstr.data, p2);
2700                         /* cannot merge tokens: just add them separately */
2701                         tok_str_add2(&macro_str1, tok, &tokc);
2702                         /* XXX: free associated memory ? */
2703                         tok = t;
2704                         tokc = cval;
2705                     }
2706                 }
2707             }
2708         }
2709         tok_str_add2(&macro_str1, tok, &tokc);
2710     }
2711     macro_ptr = (int *)saved_macro_ptr;
2712     cstr_free(&cstr);
2713     tok_str_add(&macro_str1, 0);
2714     return macro_str1.str;
2715 }
2716
2717
2718 /* do macro substitution of macro_str and add result to
2719    (tok_str,tok_len). 'nested_list' is the list of all macros we got
2720    inside to avoid recursing. */
2721 static void macro_subst(TokenString *tok_str, Sym **nested_list,
2722                         const int *macro_str, struct macro_level ** can_read_stream)
2723 {
2724     Sym *s;
2725     int *macro_str1;
2726     const int *ptr;
2727     int t, ret, spc;
2728     CValue cval;
2729     struct macro_level ml;
2730
2731     /* first scan for '##' operator handling */
2732     ptr = macro_str;
2733     macro_str1 = macro_twosharps(ptr);
2734     if (macro_str1)
2735         ptr = macro_str1;
2736     spc = 0;
2737     while (1) {
2738         /* NOTE: ptr == NULL can only happen if tokens are read from
2739            file stream due to a macro function call */
2740         if (ptr == NULL)
2741             break;
2742         TOK_GET(t, ptr, cval);
2743         if (t == 0)
2744             break;
2745         s = define_find(t);
2746         if (s != NULL) {
2747             /* if nested substitution, do nothing */
2748             if (sym_find2(*nested_list, t))
2749                 goto no_subst;
2750             ml.p = macro_ptr;
2751             if (can_read_stream)
2752                 ml.prev = *can_read_stream, *can_read_stream = &ml;
2753             macro_ptr = (int *)ptr;
2754             tok = t;
2755             ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
2756             ptr = (int *)macro_ptr;
2757             macro_ptr = ml.p;
2758             if (can_read_stream && *can_read_stream == &ml)
2759                 *can_read_stream = ml.prev;
2760             if (ret != 0)
2761                 goto no_subst;
2762         } else {
2763         no_subst:
2764             if (!check_space(t, &spc))
2765                 tok_str_add2(tok_str, t, &cval);
2766         }
2767     }
2768     if (macro_str1)
2769         tok_str_free(macro_str1);
2770 }
2771
2772 /* return next token with macro substitution */
2773 static void next(void)
2774 {
2775     Sym *nested_list, *s;
2776     TokenString str;
2777     struct macro_level *ml;
2778
2779  redo:
2780     if (parse_flags & PARSE_FLAG_SPACES)
2781         next_nomacro_spc();
2782     else
2783         next_nomacro();
2784     if (!macro_ptr) {
2785         /* if not reading from macro substituted string, then try
2786            to substitute macros */
2787         if (tok >= TOK_IDENT &&
2788             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2789             s = define_find(tok);
2790             if (s) {
2791                 /* we have a macro: we try to substitute */
2792                 tok_str_new(&str);
2793                 nested_list = NULL;
2794                 ml = NULL;
2795                 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
2796                     /* substitution done, NOTE: maybe empty */
2797                     tok_str_add(&str, 0);
2798                     macro_ptr = str.str;
2799                     macro_ptr_allocated = str.str;
2800                     goto redo;
2801                 }
2802             }
2803         }
2804     } else {
2805         if (tok == 0) {
2806             /* end of macro or end of unget buffer */
2807             if (unget_buffer_enabled) {
2808                 macro_ptr = unget_saved_macro_ptr;
2809                 unget_buffer_enabled = 0;
2810             } else {
2811                 /* end of macro string: free it */
2812                 tok_str_free(macro_ptr_allocated);
2813                 macro_ptr = NULL;
2814             }
2815             goto redo;
2816         }
2817     }
2818
2819     /* convert preprocessor tokens into C tokens */
2820     if (tok == TOK_PPNUM &&
2821         (parse_flags & PARSE_FLAG_TOK_NUM)) {
2822         parse_number((char *)tokc.cstr->data);
2823     }
2824 }
2825
2826 /* push back current token and set current token to 'last_tok'. Only
2827    identifier case handled for labels. */
2828 static inline void unget_tok(int last_tok)
2829 {
2830     int i, n;
2831     int *q;
2832     unget_saved_macro_ptr = macro_ptr;
2833     unget_buffer_enabled = 1;
2834     q = unget_saved_buffer;
2835     macro_ptr = q;
2836     *q++ = tok;
2837     n = tok_ext_size(tok) - 1;
2838     for(i=0;i<n;i++)
2839         *q++ = tokc.tab[i];
2840     *q = 0; /* end of token string */
2841     tok = last_tok;
2842 }
2843
2844
2845 /* better than nothing, but needs extension to handle '-E' option
2846    correctly too */
2847 static void preprocess_init(TCCState *s1)
2848 {
2849     s1->include_stack_ptr = s1->include_stack;
2850     /* XXX: move that before to avoid having to initialize
2851        file->ifdef_stack_ptr ? */
2852     s1->ifdef_stack_ptr = s1->ifdef_stack;
2853     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
2854
2855     /* XXX: not ANSI compliant: bound checking says error */
2856     vtop = vstack - 1;
2857     s1->pack_stack[0] = 0;
2858     s1->pack_stack_ptr = s1->pack_stack;
2859 }
2860
2861 void preprocess_new()
2862 {
2863     int i, c;
2864     const char *p, *r;
2865     TokenSym *ts;
2866
2867     /* init isid table */
2868     for(i=CH_EOF;i<256;i++)
2869         isidnum_table[i-CH_EOF] = isid(i) || isnum(i);
2870
2871     /* add all tokens */
2872     table_ident = NULL;
2873     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
2874
2875     tok_ident = TOK_IDENT;
2876     p = tcc_keywords;
2877     while (*p) {
2878         r = p;
2879         for(;;) {
2880             c = *r++;
2881             if (c == '\0')
2882                 break;
2883         }
2884         ts = tok_alloc(p, r - p - 1);
2885         p = r;
2886     }
2887 }
2888
2889 /* Preprocess the current file */
2890 static int tcc_preprocess(TCCState *s1)
2891 {
2892     Sym *define_start;
2893     BufferedFile *file_ref;
2894     int token_seen, line_ref;
2895
2896     preprocess_init(s1);
2897     define_start = define_stack;
2898     ch = file->buf_ptr[0];
2899     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
2900     parse_flags = PARSE_FLAG_ASM_COMMENTS | PARSE_FLAG_PREPROCESS |
2901         PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
2902     token_seen = 0;
2903     line_ref = 0;
2904     file_ref = NULL;
2905
2906     for (;;) {
2907         next();
2908         if (tok == TOK_EOF) {
2909             break;
2910         } else if (tok == TOK_LINEFEED) {
2911             if (!token_seen)
2912                 continue;
2913             ++line_ref;
2914             token_seen = 0;
2915         } else if (!token_seen) {
2916             int d = file->line_num - line_ref;
2917             if (file != file_ref || d < 0 || d >= 8)
2918                 fprintf(s1->outfile, "# %d \"%s\"\n", file->line_num, file->filename);
2919             else
2920                 while (d)
2921                     fputs("\n", s1->outfile), --d;
2922             line_ref = (file_ref = file)->line_num;
2923             token_seen = 1;
2924         }
2925         fputs(get_tok_str(tok, &tokc), s1->outfile);
2926     }
2927     free_defines(define_start);
2928     return 0;
2929 }
2930