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 i, c, n, saved_parse_flags;
1194     char buf[1024], *q;
1195     Sym *s;
1196
1197     saved_parse_flags = parse_flags;
1198     parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM |
1199         PARSE_FLAG_LINEFEED;
1200     next_nomacro();
1201  redo:
1202     switch(tok) {
1203     case TOK_DEFINE:
1204         next_nomacro();
1205         parse_define();
1206         break;
1207     case TOK_UNDEF:
1208         next_nomacro();
1209         s = define_find(tok);
1210         /* undefine symbol by putting an invalid name */
1211         if (s)
1212             define_undef(s);
1213         break;
1214     case TOK_INCLUDE:
1215     case TOK_INCLUDE_NEXT:
1216         ch = file->buf_ptr[0];
1217         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1218         skip_spaces();
1219         if (ch == '<') {
1220             c = '>';
1221             goto read_name;
1222         } else if (ch == '\"') {
1223             c = ch;
1224         read_name:
1225             inp();
1226             q = buf;
1227             while (ch != c && ch != '\n' && ch != CH_EOF) {
1228                 if ((q - buf) < sizeof(buf) - 1)
1229                     *q++ = ch;
1230                 if (ch == '\\') {
1231                     if (handle_stray_noerror() == 0)
1232                         --q;
1233                 } else
1234                     inp();
1235             }
1236             *q = '\0';
1237             minp();
1238 #if 0
1239             /* eat all spaces and comments after include */
1240             /* XXX: slightly incorrect */
1241             while (ch1 != '\n' && ch1 != CH_EOF)
1242                 inp();
1243 #endif
1244         } else {
1245             /* computed #include : either we have only strings or
1246                we have anything enclosed in '<>' */
1247             next();
1248             buf[0] = '\0';
1249             if (tok == TOK_STR) {
1250                 while (tok != TOK_LINEFEED) {
1251                     if (tok != TOK_STR) {
1252                     include_syntax:
1253                         error("'#include' expects \"FILENAME\" or <FILENAME>");
1254                     }
1255                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1256                     next();
1257                 }
1258                 c = '\"';
1259             } else {
1260                 int len;
1261                 while (tok != TOK_LINEFEED) {
1262                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1263                     next();
1264                 }
1265                 len = strlen(buf);
1266                 /* check syntax and remove '<>' */
1267                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1268                     goto include_syntax;
1269                 memmove(buf, buf + 1, len - 2);
1270                 buf[len - 2] = '\0';
1271                 c = '>';
1272             }
1273         }
1274
1275         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1276             error("#include recursion too deep");
1277
1278         n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1279         for (i = -2; i < n; ++i) {
1280             char buf1[sizeof file->filename];
1281             BufferedFile *f;
1282             CachedInclude *e;
1283             const char *path;
1284             int size;
1285
1286             if (i == -2) {
1287                 /* check absolute include path */
1288                 if (!IS_ABSPATH(buf))
1289                     continue;
1290                 buf1[0] = 0;
1291
1292             } else if (i == -1) {
1293                 /* search in current dir if "header.h" */
1294                 if (c != '\"')
1295                     continue;
1296                 size = tcc_basename(file->filename) - file->filename;
1297                 memcpy(buf1, file->filename, size);
1298                 buf1[size] = '\0';
1299
1300             } else {
1301                 /* search in all the include paths */
1302                 if (i < s1->nb_include_paths)
1303                     path = s1->include_paths[i];
1304                 else
1305                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1306                 pstrcpy(buf1, sizeof(buf1), path);
1307                 pstrcat(buf1, sizeof(buf1), "/");
1308             }
1309
1310             pstrcat(buf1, sizeof(buf1), buf);
1311
1312             e = search_cached_include(s1, c, buf1);
1313             if (e && define_find(e->ifndef_macro)) {
1314                 /* no need to parse the include because the 'ifndef macro'
1315                    is defined */
1316 #ifdef INC_DEBUG
1317                 printf("%s: skipping %s\n", file->filename, buf);
1318 #endif
1319                 f = NULL;
1320             }  else {
1321                 f = tcc_open(s1, buf1);
1322                 if (!f)
1323                     continue;
1324             }
1325
1326             if (tok == TOK_INCLUDE_NEXT) {
1327                 tok = TOK_INCLUDE;
1328                 if (f)
1329                     tcc_close(f);
1330                 continue;
1331             }
1332
1333             if (!f)
1334                 goto include_done;
1335
1336 #ifdef INC_DEBUG
1337             printf("%s: including %s\n", file->filename, buf1);
1338 #endif
1339
1340            /* XXX: fix current line init */
1341            /* push current file in stack */
1342             *s1->include_stack_ptr++ = file;
1343             f->inc_type = c;
1344             pstrcpy(f->inc_filename, sizeof(f->inc_filename), buf1);
1345             file = f;
1346             /* add include file debug info */
1347             if (tcc_state->do_debug) {
1348                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1349             }
1350             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1351             ch = file->buf_ptr[0];
1352             goto the_end;
1353         }
1354         error("include file '%s' not found", buf);
1355 include_done:
1356         break;
1357     case TOK_IFNDEF:
1358         c = 1;
1359         goto do_ifdef;
1360     case TOK_IF:
1361         c = expr_preprocess();
1362         goto do_if;
1363     case TOK_IFDEF:
1364         c = 0;
1365     do_ifdef:
1366         next_nomacro();
1367         if (tok < TOK_IDENT)
1368             error("invalid argument for '#if%sdef'", c ? "n" : "");
1369         if (is_bof) {
1370             if (c) {
1371 #ifdef INC_DEBUG
1372                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1373 #endif
1374                 file->ifndef_macro = tok;
1375             }
1376         }
1377         c = (define_find(tok) != 0) ^ c;
1378     do_if:
1379         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1380             error("memory full");
1381         *s1->ifdef_stack_ptr++ = c;
1382         goto test_skip;
1383     case TOK_ELSE:
1384         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1385             error("#else without matching #if");
1386         if (s1->ifdef_stack_ptr[-1] & 2)
1387             error("#else after #else");
1388         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1389         goto test_skip;
1390     case TOK_ELIF:
1391         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1392             error("#elif without matching #if");
1393         c = s1->ifdef_stack_ptr[-1];
1394         if (c > 1)
1395             error("#elif after #else");
1396         /* last #if/#elif expression was true: we skip */
1397         if (c == 1)
1398             goto skip;
1399         c = expr_preprocess();
1400         s1->ifdef_stack_ptr[-1] = c;
1401     test_skip:
1402         if (!(c & 1)) {
1403         skip:
1404             preprocess_skip();
1405             is_bof = 0;
1406             goto redo;
1407         }
1408         break;
1409     case TOK_ENDIF:
1410         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1411             error("#endif without matching #if");
1412         s1->ifdef_stack_ptr--;
1413         /* '#ifndef macro' was at the start of file. Now we check if
1414            an '#endif' is exactly at the end of file */
1415         if (file->ifndef_macro &&
1416             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1417             file->ifndef_macro_saved = file->ifndef_macro;
1418             /* need to set to zero to avoid false matches if another
1419                #ifndef at middle of file */
1420             file->ifndef_macro = 0;
1421             while (tok != TOK_LINEFEED)
1422                 next_nomacro();
1423             tok_flags |= TOK_FLAG_ENDIF;
1424             goto the_end;
1425         }
1426         break;
1427     case TOK_LINE:
1428         next();
1429         if (tok != TOK_CINT)
1430             error("#line");
1431         file->line_num = tokc.i - 1; /* the line number will be incremented after */
1432         next();
1433         if (tok != TOK_LINEFEED) {
1434             if (tok != TOK_STR)
1435                 error("#line");
1436             pstrcpy(file->filename, sizeof(file->filename),
1437                     (char *)tokc.cstr->data);
1438         }
1439         break;
1440     case TOK_ERROR:
1441     case TOK_WARNING:
1442         c = tok;
1443         ch = file->buf_ptr[0];
1444         skip_spaces();
1445         q = buf;
1446         while (ch != '\n' && ch != CH_EOF) {
1447             if ((q - buf) < sizeof(buf) - 1)
1448                 *q++ = ch;
1449             if (ch == '\\') {
1450                 if (handle_stray_noerror() == 0)
1451                     --q;
1452             } else
1453                 inp();
1454         }
1455         *q = '\0';
1456         if (c == TOK_ERROR)
1457             error("#error %s", buf);
1458         else
1459             warning("#warning %s", buf);
1460         break;
1461     case TOK_PRAGMA:
1462         pragma_parse(s1);
1463         break;
1464     default:
1465         if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_CINT) {
1466             /* '!' is ignored to allow C scripts. numbers are ignored
1467                to emulate cpp behaviour */
1468         } else {
1469             if (!(saved_parse_flags & PARSE_FLAG_ASM_COMMENTS))
1470                 warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1471         }
1472         break;
1473     }
1474     /* ignore other preprocess commands or #! for C scripts */
1475     while (tok != TOK_LINEFEED)
1476         next_nomacro();
1477  the_end:
1478     parse_flags = saved_parse_flags;
1479 }
1480
1481 /* evaluate escape codes in a string. */
1482 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1483 {
1484     int c, n;
1485     const uint8_t *p;
1486
1487     p = buf;
1488     for(;;) {
1489         c = *p;
1490         if (c == '\0')
1491             break;
1492         if (c == '\\') {
1493             p++;
1494             /* escape */
1495             c = *p;
1496             switch(c) {
1497             case '0': case '1': case '2': case '3':
1498             case '4': case '5': case '6': case '7':
1499                 /* at most three octal digits */
1500                 n = c - '0';
1501                 p++;
1502                 c = *p;
1503                 if (isoct(c)) {
1504                     n = n * 8 + c - '0';
1505                     p++;
1506                     c = *p;
1507                     if (isoct(c)) {
1508                         n = n * 8 + c - '0';
1509                         p++;
1510                     }
1511                 }
1512                 c = n;
1513                 goto add_char_nonext;
1514             case 'x':
1515             case 'u':
1516             case 'U':
1517                 p++;
1518                 n = 0;
1519                 for(;;) {
1520                     c = *p;
1521                     if (c >= 'a' && c <= 'f')
1522                         c = c - 'a' + 10;
1523                     else if (c >= 'A' && c <= 'F')
1524                         c = c - 'A' + 10;
1525                     else if (isnum(c))
1526                         c = c - '0';
1527                     else
1528                         break;
1529                     n = n * 16 + c;
1530                     p++;
1531                 }
1532                 c = n;
1533                 goto add_char_nonext;
1534             case 'a':
1535                 c = '\a';
1536                 break;
1537             case 'b':
1538                 c = '\b';
1539                 break;
1540             case 'f':
1541                 c = '\f';
1542                 break;
1543             case 'n':
1544                 c = '\n';
1545                 break;
1546             case 'r':
1547                 c = '\r';
1548                 break;
1549             case 't':
1550                 c = '\t';
1551                 break;
1552             case 'v':
1553                 c = '\v';
1554                 break;
1555             case 'e':
1556                 if (!gnu_ext)
1557                     goto invalid_escape;
1558                 c = 27;
1559                 break;
1560             case '\'':
1561             case '\"':
1562             case '\\':
1563             case '?':
1564                 break;
1565             default:
1566             invalid_escape:
1567                 if (c >= '!' && c <= '~')
1568                     warning("unknown escape sequence: \'\\%c\'", c);
1569                 else
1570                     warning("unknown escape sequence: \'\\x%x\'", c);
1571                 break;
1572             }
1573         }
1574         p++;
1575     add_char_nonext:
1576         if (!is_long)
1577             cstr_ccat(outstr, c);
1578         else
1579             cstr_wccat(outstr, c);
1580     }
1581     /* add a trailing '\0' */
1582     if (!is_long)
1583         cstr_ccat(outstr, '\0');
1584     else
1585         cstr_wccat(outstr, '\0');
1586 }
1587
1588 /* we use 64 bit numbers */
1589 #define BN_SIZE 2
1590
1591 /* bn = (bn << shift) | or_val */
1592 void bn_lshift(unsigned int *bn, int shift, int or_val)
1593 {
1594     int i;
1595     unsigned int v;
1596     for(i=0;i<BN_SIZE;i++) {
1597         v = bn[i];
1598         bn[i] = (v << shift) | or_val;
1599         or_val = v >> (32 - shift);
1600     }
1601 }
1602
1603 void bn_zero(unsigned int *bn)
1604 {
1605     int i;
1606     for(i=0;i<BN_SIZE;i++) {
1607         bn[i] = 0;
1608     }
1609 }
1610
1611 /* parse number in null terminated string 'p' and return it in the
1612    current token */
1613 void parse_number(const char *p)
1614 {
1615     int b, t, shift, frac_bits, s, exp_val, ch;
1616     char *q;
1617     unsigned int bn[BN_SIZE];
1618     double d;
1619
1620     /* number */
1621     q = token_buf;
1622     ch = *p++;
1623     t = ch;
1624     ch = *p++;
1625     *q++ = t;
1626     b = 10;
1627     if (t == '.') {
1628         goto float_frac_parse;
1629     } else if (t == '0') {
1630         if (ch == 'x' || ch == 'X') {
1631             q--;
1632             ch = *p++;
1633             b = 16;
1634         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
1635             q--;
1636             ch = *p++;
1637             b = 2;
1638         }
1639     }
1640     /* parse all digits. cannot check octal numbers at this stage
1641        because of floating point constants */
1642     while (1) {
1643         if (ch >= 'a' && ch <= 'f')
1644             t = ch - 'a' + 10;
1645         else if (ch >= 'A' && ch <= 'F')
1646             t = ch - 'A' + 10;
1647         else if (isnum(ch))
1648             t = ch - '0';
1649         else
1650             break;
1651         if (t >= b)
1652             break;
1653         if (q >= token_buf + STRING_MAX_SIZE) {
1654         num_too_long:
1655             error("number too long");
1656         }
1657         *q++ = ch;
1658         ch = *p++;
1659     }
1660     if (ch == '.' ||
1661         ((ch == 'e' || ch == 'E') && b == 10) ||
1662         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
1663         if (b != 10) {
1664             /* NOTE: strtox should support that for hexa numbers, but
1665                non ISOC99 libcs do not support it, so we prefer to do
1666                it by hand */
1667             /* hexadecimal or binary floats */
1668             /* XXX: handle overflows */
1669             *q = '\0';
1670             if (b == 16)
1671                 shift = 4;
1672             else
1673                 shift = 2;
1674             bn_zero(bn);
1675             q = token_buf;
1676             while (1) {
1677                 t = *q++;
1678                 if (t == '\0') {
1679                     break;
1680                 } else if (t >= 'a') {
1681                     t = t - 'a' + 10;
1682                 } else if (t >= 'A') {
1683                     t = t - 'A' + 10;
1684                 } else {
1685                     t = t - '0';
1686                 }
1687                 bn_lshift(bn, shift, t);
1688             }
1689             frac_bits = 0;
1690             if (ch == '.') {
1691                 ch = *p++;
1692                 while (1) {
1693                     t = ch;
1694                     if (t >= 'a' && t <= 'f') {
1695                         t = t - 'a' + 10;
1696                     } else if (t >= 'A' && t <= 'F') {
1697                         t = t - 'A' + 10;
1698                     } else if (t >= '0' && t <= '9') {
1699                         t = t - '0';
1700                     } else {
1701                         break;
1702                     }
1703                     if (t >= b)
1704                         error("invalid digit");
1705                     bn_lshift(bn, shift, t);
1706                     frac_bits += shift;
1707                     ch = *p++;
1708                 }
1709             }
1710             if (ch != 'p' && ch != 'P')
1711                 expect("exponent");
1712             ch = *p++;
1713             s = 1;
1714             exp_val = 0;
1715             if (ch == '+') {
1716                 ch = *p++;
1717             } else if (ch == '-') {
1718                 s = -1;
1719                 ch = *p++;
1720             }
1721             if (ch < '0' || ch > '9')
1722                 expect("exponent digits");
1723             while (ch >= '0' && ch <= '9') {
1724                 exp_val = exp_val * 10 + ch - '0';
1725                 ch = *p++;
1726             }
1727             exp_val = exp_val * s;
1728
1729             /* now we can generate the number */
1730             /* XXX: should patch directly float number */
1731             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
1732             d = ldexp(d, exp_val - frac_bits);
1733             t = toup(ch);
1734             if (t == 'F') {
1735                 ch = *p++;
1736                 tok = TOK_CFLOAT;
1737                 /* float : should handle overflow */
1738                 tokc.f = (float)d;
1739             } else if (t == 'L') {
1740                 ch = *p++;
1741                 tok = TOK_CLDOUBLE;
1742                 /* XXX: not large enough */
1743                 tokc.ld = (long double)d;
1744             } else {
1745                 tok = TOK_CDOUBLE;
1746                 tokc.d = d;
1747             }
1748         } else {
1749             /* decimal floats */
1750             if (ch == '.') {
1751                 if (q >= token_buf + STRING_MAX_SIZE)
1752                     goto num_too_long;
1753                 *q++ = ch;
1754                 ch = *p++;
1755             float_frac_parse:
1756                 while (ch >= '0' && ch <= '9') {
1757                     if (q >= token_buf + STRING_MAX_SIZE)
1758                         goto num_too_long;
1759                     *q++ = ch;
1760                     ch = *p++;
1761                 }
1762             }
1763             if (ch == 'e' || ch == 'E') {
1764                 if (q >= token_buf + STRING_MAX_SIZE)
1765                     goto num_too_long;
1766                 *q++ = ch;
1767                 ch = *p++;
1768                 if (ch == '-' || ch == '+') {
1769                     if (q >= token_buf + STRING_MAX_SIZE)
1770                         goto num_too_long;
1771                     *q++ = ch;
1772                     ch = *p++;
1773                 }
1774                 if (ch < '0' || ch > '9')
1775                     expect("exponent digits");
1776                 while (ch >= '0' && ch <= '9') {
1777                     if (q >= token_buf + STRING_MAX_SIZE)
1778                         goto num_too_long;
1779                     *q++ = ch;
1780                     ch = *p++;
1781                 }
1782             }
1783             *q = '\0';
1784             t = toup(ch);
1785             errno = 0;
1786             if (t == 'F') {
1787                 ch = *p++;
1788                 tok = TOK_CFLOAT;
1789                 tokc.f = strtof(token_buf, NULL);
1790             } else if (t == 'L') {
1791                 ch = *p++;
1792                 tok = TOK_CLDOUBLE;
1793                 tokc.ld = strtold(token_buf, NULL);
1794             } else {
1795                 tok = TOK_CDOUBLE;
1796                 tokc.d = strtod(token_buf, NULL);
1797             }
1798         }
1799     } else {
1800         unsigned long long n, n1;
1801         int lcount, ucount;
1802
1803         /* integer number */
1804         *q = '\0';
1805         q = token_buf;
1806         if (b == 10 && *q == '0') {
1807             b = 8;
1808             q++;
1809         }
1810         n = 0;
1811         while(1) {
1812             t = *q++;
1813             /* no need for checks except for base 10 / 8 errors */
1814             if (t == '\0') {
1815                 break;
1816             } else if (t >= 'a') {
1817                 t = t - 'a' + 10;
1818             } else if (t >= 'A') {
1819                 t = t - 'A' + 10;
1820             } else {
1821                 t = t - '0';
1822                 if (t >= b)
1823                     error("invalid digit");
1824             }
1825             n1 = n;
1826             n = n * b + t;
1827             /* detect overflow */
1828             /* XXX: this test is not reliable */
1829             if (n < n1)
1830                 error("integer constant overflow");
1831         }
1832
1833         /* XXX: not exactly ANSI compliant */
1834         if ((n & 0xffffffff00000000LL) != 0) {
1835             if ((n >> 63) != 0)
1836                 tok = TOK_CULLONG;
1837             else
1838                 tok = TOK_CLLONG;
1839         } else if (n > 0x7fffffff) {
1840             tok = TOK_CUINT;
1841         } else {
1842             tok = TOK_CINT;
1843         }
1844         lcount = 0;
1845         ucount = 0;
1846         for(;;) {
1847             t = toup(ch);
1848             if (t == 'L') {
1849                 if (lcount >= 2)
1850                     error("three 'l's in integer constant");
1851                 lcount++;
1852                 if (lcount == 2) {
1853                     if (tok == TOK_CINT)
1854                         tok = TOK_CLLONG;
1855                     else if (tok == TOK_CUINT)
1856                         tok = TOK_CULLONG;
1857                 }
1858                 ch = *p++;
1859             } else if (t == 'U') {
1860                 if (ucount >= 1)
1861                     error("two 'u's in integer constant");
1862                 ucount++;
1863                 if (tok == TOK_CINT)
1864                     tok = TOK_CUINT;
1865                 else if (tok == TOK_CLLONG)
1866                     tok = TOK_CULLONG;
1867                 ch = *p++;
1868             } else {
1869                 break;
1870             }
1871         }
1872         if (tok == TOK_CINT || tok == TOK_CUINT)
1873             tokc.ui = n;
1874         else
1875             tokc.ull = n;
1876     }
1877     if (ch)
1878         error("invalid number\n");
1879 }
1880
1881
1882 #define PARSE2(c1, tok1, c2, tok2)              \
1883     case c1:                                    \
1884         PEEKC(c, p);                            \
1885         if (c == c2) {                          \
1886             p++;                                \
1887             tok = tok2;                         \
1888         } else {                                \
1889             tok = tok1;                         \
1890         }                                       \
1891         break;
1892
1893 /* return next token without macro substitution */
1894 static inline void next_nomacro1(void)
1895 {
1896     int t, c, is_long;
1897     TokenSym *ts;
1898     uint8_t *p, *p1;
1899     unsigned int h;
1900
1901     p = file->buf_ptr;
1902  redo_no_start:
1903     c = *p;
1904     switch(c) {
1905     case ' ':
1906     case '\t':
1907         tok = c;
1908         p++;
1909         goto keep_tok_flags;
1910     case '\f':
1911     case '\v':
1912     case '\r':
1913         p++;
1914         goto redo_no_start;
1915     case '\\':
1916         /* first look if it is in fact an end of buffer */
1917         if (p >= file->buf_end) {
1918             file->buf_ptr = p;
1919             handle_eob();
1920             p = file->buf_ptr;
1921             if (p >= file->buf_end)
1922                 goto parse_eof;
1923             else
1924                 goto redo_no_start;
1925         } else {
1926             file->buf_ptr = p;
1927             ch = *p;
1928             handle_stray();
1929             p = file->buf_ptr;
1930             goto redo_no_start;
1931         }
1932     parse_eof:
1933         {
1934             TCCState *s1 = tcc_state;
1935             if ((parse_flags & PARSE_FLAG_LINEFEED)
1936                 && !(tok_flags & TOK_FLAG_EOF)) {
1937                 tok_flags |= TOK_FLAG_EOF;
1938                 tok = TOK_LINEFEED;
1939                 goto keep_tok_flags;
1940             } else if (s1->include_stack_ptr == s1->include_stack ||
1941                        !(parse_flags & PARSE_FLAG_PREPROCESS)) {
1942                 /* no include left : end of file. */
1943                 tok = TOK_EOF;
1944             } else {
1945                 tok_flags &= ~TOK_FLAG_EOF;
1946                 /* pop include file */
1947
1948                 /* test if previous '#endif' was after a #ifdef at
1949                    start of file */
1950                 if (tok_flags & TOK_FLAG_ENDIF) {
1951 #ifdef INC_DEBUG
1952                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
1953 #endif
1954                     add_cached_include(s1, file->inc_type, file->inc_filename,
1955                                        file->ifndef_macro_saved);
1956                 }
1957
1958                 /* add end of include file debug info */
1959                 if (tcc_state->do_debug) {
1960                     put_stabd(N_EINCL, 0, 0);
1961                 }
1962                 /* pop include stack */
1963                 tcc_close(file);
1964                 s1->include_stack_ptr--;
1965                 file = *s1->include_stack_ptr;
1966                 p = file->buf_ptr;
1967                 goto redo_no_start;
1968             }
1969         }
1970         break;
1971
1972     case '\n':
1973         file->line_num++;
1974         tok_flags |= TOK_FLAG_BOL;
1975         p++;
1976 maybe_newline:
1977         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
1978             goto redo_no_start;
1979         tok = TOK_LINEFEED;
1980         goto keep_tok_flags;
1981
1982     case '#':
1983         /* XXX: simplify */
1984         PEEKC(c, p);
1985         if ((tok_flags & TOK_FLAG_BOL) &&
1986             (parse_flags & PARSE_FLAG_PREPROCESS)) {
1987             file->buf_ptr = p;
1988             preprocess(tok_flags & TOK_FLAG_BOF);
1989             p = file->buf_ptr;
1990             goto maybe_newline;
1991         } else {
1992             if (c == '#') {
1993                 p++;
1994                 tok = TOK_TWOSHARPS;
1995             } else {
1996                 if (parse_flags & PARSE_FLAG_ASM_COMMENTS) {
1997                     p = parse_line_comment(p - 1);
1998                     goto redo_no_start;
1999                 } else {
2000                     tok = '#';
2001                 }
2002             }
2003         }
2004         break;
2005
2006     case 'a': case 'b': case 'c': case 'd':
2007     case 'e': case 'f': case 'g': case 'h':
2008     case 'i': case 'j': case 'k': case 'l':
2009     case 'm': case 'n': case 'o': case 'p':
2010     case 'q': case 'r': case 's': case 't':
2011     case 'u': case 'v': case 'w': case 'x':
2012     case 'y': case 'z':
2013     case 'A': case 'B': case 'C': case 'D':
2014     case 'E': case 'F': case 'G': case 'H':
2015     case 'I': case 'J': case 'K':
2016     case 'M': case 'N': case 'O': case 'P':
2017     case 'Q': case 'R': case 'S': case 'T':
2018     case 'U': case 'V': case 'W': case 'X':
2019     case 'Y': case 'Z':
2020     case '_':
2021     parse_ident_fast:
2022         p1 = p;
2023         h = TOK_HASH_INIT;
2024         h = TOK_HASH_FUNC(h, c);
2025         p++;
2026         for(;;) {
2027             c = *p;
2028             if (!isidnum_table[c-CH_EOF])
2029                 break;
2030             h = TOK_HASH_FUNC(h, c);
2031             p++;
2032         }
2033         if (c != '\\') {
2034             TokenSym **pts;
2035             int len;
2036
2037             /* fast case : no stray found, so we have the full token
2038                and we have already hashed it */
2039             len = p - p1;
2040             h &= (TOK_HASH_SIZE - 1);
2041             pts = &hash_ident[h];
2042             for(;;) {
2043                 ts = *pts;
2044                 if (!ts)
2045                     break;
2046                 if (ts->len == len && !memcmp(ts->str, p1, len))
2047                     goto token_found;
2048                 pts = &(ts->hash_next);
2049             }
2050             ts = tok_alloc_new(pts, p1, len);
2051         token_found: ;
2052         } else {
2053             /* slower case */
2054             cstr_reset(&tokcstr);
2055
2056             while (p1 < p) {
2057                 cstr_ccat(&tokcstr, *p1);
2058                 p1++;
2059             }
2060             p--;
2061             PEEKC(c, p);
2062         parse_ident_slow:
2063             while (isidnum_table[c-CH_EOF]) {
2064                 cstr_ccat(&tokcstr, c);
2065                 PEEKC(c, p);
2066             }
2067             ts = tok_alloc(tokcstr.data, tokcstr.size);
2068         }
2069         tok = ts->tok;
2070         break;
2071     case 'L':
2072         t = p[1];
2073         if (t != '\\' && t != '\'' && t != '\"') {
2074             /* fast case */
2075             goto parse_ident_fast;
2076         } else {
2077             PEEKC(c, p);
2078             if (c == '\'' || c == '\"') {
2079                 is_long = 1;
2080                 goto str_const;
2081             } else {
2082                 cstr_reset(&tokcstr);
2083                 cstr_ccat(&tokcstr, 'L');
2084                 goto parse_ident_slow;
2085             }
2086         }
2087         break;
2088     case '0': case '1': case '2': case '3':
2089     case '4': case '5': case '6': case '7':
2090     case '8': case '9':
2091
2092         cstr_reset(&tokcstr);
2093         /* after the first digit, accept digits, alpha, '.' or sign if
2094            prefixed by 'eEpP' */
2095     parse_num:
2096         for(;;) {
2097             t = c;
2098             cstr_ccat(&tokcstr, c);
2099             PEEKC(c, p);
2100             if (!(isnum(c) || isid(c) || c == '.' ||
2101                   ((c == '+' || c == '-') &&
2102                    (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2103                 break;
2104         }
2105         /* We add a trailing '\0' to ease parsing */
2106         cstr_ccat(&tokcstr, '\0');
2107         tokc.cstr = &tokcstr;
2108         tok = TOK_PPNUM;
2109         break;
2110     case '.':
2111         /* special dot handling because it can also start a number */
2112         PEEKC(c, p);
2113         if (isnum(c)) {
2114             cstr_reset(&tokcstr);
2115             cstr_ccat(&tokcstr, '.');
2116             goto parse_num;
2117         } else if (c == '.') {
2118             PEEKC(c, p);
2119             if (c != '.')
2120                 expect("'.'");
2121             PEEKC(c, p);
2122             tok = TOK_DOTS;
2123         } else {
2124             tok = '.';
2125         }
2126         break;
2127     case '\'':
2128     case '\"':
2129         is_long = 0;
2130     str_const:
2131         {
2132             CString str;
2133             int sep;
2134
2135             sep = c;
2136
2137             /* parse the string */
2138             cstr_new(&str);
2139             p = parse_pp_string(p, sep, &str);
2140             cstr_ccat(&str, '\0');
2141
2142             /* eval the escape (should be done as TOK_PPNUM) */
2143             cstr_reset(&tokcstr);
2144             parse_escape_string(&tokcstr, str.data, is_long);
2145             cstr_free(&str);
2146
2147             if (sep == '\'') {
2148                 int char_size;
2149                 /* XXX: make it portable */
2150                 if (!is_long)
2151                     char_size = 1;
2152                 else
2153                     char_size = sizeof(nwchar_t);
2154                 if (tokcstr.size <= char_size)
2155                     error("empty character constant");
2156                 if (tokcstr.size > 2 * char_size)
2157                     warning("multi-character character constant");
2158                 if (!is_long) {
2159                     tokc.i = *(int8_t *)tokcstr.data;
2160                     tok = TOK_CCHAR;
2161                 } else {
2162                     tokc.i = *(nwchar_t *)tokcstr.data;
2163                     tok = TOK_LCHAR;
2164                 }
2165             } else {
2166                 tokc.cstr = &tokcstr;
2167                 if (!is_long)
2168                     tok = TOK_STR;
2169                 else
2170                     tok = TOK_LSTR;
2171             }
2172         }
2173         break;
2174
2175     case '<':
2176         PEEKC(c, p);
2177         if (c == '=') {
2178             p++;
2179             tok = TOK_LE;
2180         } else if (c == '<') {
2181             PEEKC(c, p);
2182             if (c == '=') {
2183                 p++;
2184                 tok = TOK_A_SHL;
2185             } else {
2186                 tok = TOK_SHL;
2187             }
2188         } else {
2189             tok = TOK_LT;
2190         }
2191         break;
2192
2193     case '>':
2194         PEEKC(c, p);
2195         if (c == '=') {
2196             p++;
2197             tok = TOK_GE;
2198         } else if (c == '>') {
2199             PEEKC(c, p);
2200             if (c == '=') {
2201                 p++;
2202                 tok = TOK_A_SAR;
2203             } else {
2204                 tok = TOK_SAR;
2205             }
2206         } else {
2207             tok = TOK_GT;
2208         }
2209         break;
2210
2211     case '&':
2212         PEEKC(c, p);
2213         if (c == '&') {
2214             p++;
2215             tok = TOK_LAND;
2216         } else if (c == '=') {
2217             p++;
2218             tok = TOK_A_AND;
2219         } else {
2220             tok = '&';
2221         }
2222         break;
2223
2224     case '|':
2225         PEEKC(c, p);
2226         if (c == '|') {
2227             p++;
2228             tok = TOK_LOR;
2229         } else if (c == '=') {
2230             p++;
2231             tok = TOK_A_OR;
2232         } else {
2233             tok = '|';
2234         }
2235         break;
2236
2237     case '+':
2238         PEEKC(c, p);
2239         if (c == '+') {
2240             p++;
2241             tok = TOK_INC;
2242         } else if (c == '=') {
2243             p++;
2244             tok = TOK_A_ADD;
2245         } else {
2246             tok = '+';
2247         }
2248         break;
2249
2250     case '-':
2251         PEEKC(c, p);
2252         if (c == '-') {
2253             p++;
2254             tok = TOK_DEC;
2255         } else if (c == '=') {
2256             p++;
2257             tok = TOK_A_SUB;
2258         } else if (c == '>') {
2259             p++;
2260             tok = TOK_ARROW;
2261         } else {
2262             tok = '-';
2263         }
2264         break;
2265
2266     PARSE2('!', '!', '=', TOK_NE)
2267     PARSE2('=', '=', '=', TOK_EQ)
2268     PARSE2('*', '*', '=', TOK_A_MUL)
2269     PARSE2('%', '%', '=', TOK_A_MOD)
2270     PARSE2('^', '^', '=', TOK_A_XOR)
2271
2272         /* comments or operator */
2273     case '/':
2274         PEEKC(c, p);
2275         if (c == '*') {
2276             p = parse_comment(p);
2277             goto redo_no_start;
2278         } else if (c == '/') {
2279             p = parse_line_comment(p);
2280             goto redo_no_start;
2281         } else if (c == '=') {
2282             p++;
2283             tok = TOK_A_DIV;
2284         } else {
2285             tok = '/';
2286         }
2287         break;
2288
2289         /* simple tokens */
2290     case '(':
2291     case ')':
2292     case '[':
2293     case ']':
2294     case '{':
2295     case '}':
2296     case ',':
2297     case ';':
2298     case ':':
2299     case '?':
2300     case '~':
2301     case '$': /* only used in assembler */
2302     case '@': /* dito */
2303         tok = c;
2304         p++;
2305         break;
2306     default:
2307         error("unrecognized character \\x%02x", c);
2308         break;
2309     }
2310     tok_flags = 0;
2311 keep_tok_flags:
2312     file->buf_ptr = p;
2313 #if defined(PARSE_DEBUG)
2314     printf("token = %s\n", get_tok_str(tok, &tokc));
2315 #endif
2316 }
2317
2318 /* return next token without macro substitution. Can read input from
2319    macro_ptr buffer */
2320 static void next_nomacro_spc(void)
2321 {
2322     if (macro_ptr) {
2323     redo:
2324         tok = *macro_ptr;
2325         if (tok) {
2326             TOK_GET(tok, macro_ptr, tokc);
2327             if (tok == TOK_LINENUM) {
2328                 file->line_num = tokc.i;
2329                 goto redo;
2330             }
2331         }
2332     } else {
2333         next_nomacro1();
2334     }
2335 }
2336
2337 static void next_nomacro(void)
2338 {
2339     do {
2340         next_nomacro_spc();
2341     } while (is_space(tok));
2342 }
2343
2344 /* substitute args in macro_str and return allocated string */
2345 static int *macro_arg_subst(Sym **nested_list, int *macro_str, Sym *args)
2346 {
2347     int *st, last_tok, t, spc;
2348     Sym *s;
2349     CValue cval;
2350     TokenString str;
2351     CString cstr;
2352
2353     tok_str_new(&str);
2354     last_tok = 0;
2355     while(1) {
2356         TOK_GET(t, macro_str, cval);
2357         if (!t)
2358             break;
2359         if (t == '#') {
2360             /* stringize */
2361             TOK_GET(t, macro_str, cval);
2362             if (!t)
2363                 break;
2364             s = sym_find2(args, t);
2365             if (s) {
2366                 cstr_new(&cstr);
2367                 st = (int *)s->c;
2368                 spc = 0;
2369                 while (*st) {
2370                     TOK_GET(t, st, cval);
2371                     if (!check_space(t, &spc))
2372                         cstr_cat(&cstr, get_tok_str(t, &cval));
2373                 }
2374                 cstr.size -= spc;
2375                 cstr_ccat(&cstr, '\0');
2376 #ifdef PP_DEBUG
2377                 printf("stringize: %s\n", (char *)cstr.data);
2378 #endif
2379                 /* add string */
2380                 cval.cstr = &cstr;
2381                 tok_str_add2(&str, TOK_STR, &cval);
2382                 cstr_free(&cstr);
2383             } else {
2384                 tok_str_add2(&str, t, &cval);
2385             }
2386         } else if (t >= TOK_IDENT) {
2387             s = sym_find2(args, t);
2388             if (s) {
2389                 st = (int *)s->c;
2390                 /* if '##' is present before or after, no arg substitution */
2391                 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2392                     /* special case for var arg macros : ## eats the
2393                        ',' if empty VA_ARGS variable. */
2394                     /* XXX: test of the ',' is not 100%
2395                        reliable. should fix it to avoid security
2396                        problems */
2397                     if (gnu_ext && s->type.t &&
2398                         last_tok == TOK_TWOSHARPS &&
2399                         str.len >= 2 && str.str[str.len - 2] == ',') {
2400                         if (*st == 0) {
2401                             /* suppress ',' '##' */
2402                             str.len -= 2;
2403                         } else {
2404                             /* suppress '##' and add variable */
2405                             str.len--;
2406                             goto add_var;
2407                         }
2408                     } else {
2409                         int t1;
2410                     add_var:
2411                         for(;;) {
2412                             TOK_GET(t1, st, cval);
2413                             if (!t1)
2414                                 break;
2415                             tok_str_add2(&str, t1, &cval);
2416                         }
2417                     }
2418                 } else {
2419                     /* NOTE: the stream cannot be read when macro
2420                        substituing an argument */
2421                     macro_subst(&str, nested_list, st, NULL);
2422                 }
2423             } else {
2424                 tok_str_add(&str, t);
2425             }
2426         } else {
2427             tok_str_add2(&str, t, &cval);
2428         }
2429         last_tok = t;
2430     }
2431     tok_str_add(&str, 0);
2432     return str.str;
2433 }
2434
2435 static char const ab_month_name[12][4] =
2436 {
2437     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2438     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2439 };
2440
2441 /* do macro substitution of current token with macro 's' and add
2442    result to (tok_str,tok_len). 'nested_list' is the list of all
2443    macros we got inside to avoid recursing. Return non zero if no
2444    substitution needs to be done */
2445 static int macro_subst_tok(TokenString *tok_str,
2446                            Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2447 {
2448     Sym *args, *sa, *sa1;
2449     int mstr_allocated, parlevel, *mstr, t, t1, *p, spc;
2450     TokenString str;
2451     char *cstrval;
2452     CValue cval;
2453     CString cstr;
2454     char buf[32];
2455
2456     /* if symbol is a macro, prepare substitution */
2457     /* special macros */
2458     if (tok == TOK___LINE__) {
2459         snprintf(buf, sizeof(buf), "%d", file->line_num);
2460         cstrval = buf;
2461         t1 = TOK_PPNUM;
2462         goto add_cstr1;
2463     } else if (tok == TOK___FILE__) {
2464         cstrval = file->filename;
2465         goto add_cstr;
2466     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2467         time_t ti;
2468         struct tm *tm;
2469
2470         time(&ti);
2471         tm = localtime(&ti);
2472         if (tok == TOK___DATE__) {
2473             snprintf(buf, sizeof(buf), "%s %2d %d",
2474                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2475         } else {
2476             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2477                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2478         }
2479         cstrval = buf;
2480     add_cstr:
2481         t1 = TOK_STR;
2482     add_cstr1:
2483         cstr_new(&cstr);
2484         cstr_cat(&cstr, cstrval);
2485         cstr_ccat(&cstr, '\0');
2486         cval.cstr = &cstr;
2487         tok_str_add2(tok_str, t1, &cval);
2488         cstr_free(&cstr);
2489     } else {
2490         mstr = (int *)s->c;
2491         mstr_allocated = 0;
2492         if (s->type.t == MACRO_FUNC) {
2493             /* NOTE: we do not use next_nomacro to avoid eating the
2494                next token. XXX: find better solution */
2495         redo:
2496             if (macro_ptr) {
2497                 p = macro_ptr;
2498                 while (is_space(t = *p) || TOK_LINEFEED == t)
2499                     ++p;
2500                 if (t == 0 && can_read_stream) {
2501                     /* end of macro stream: we must look at the token
2502                        after in the file */
2503                     struct macro_level *ml = *can_read_stream;
2504                     macro_ptr = NULL;
2505                     if (ml)
2506                     {
2507                         macro_ptr = ml->p;
2508                         ml->p = NULL;
2509                         *can_read_stream = ml -> prev;
2510                     }
2511                     goto redo;
2512                 }
2513             } else {
2514                 /* XXX: incorrect with comments */
2515                 ch = file->buf_ptr[0];
2516                 while (is_space(ch) || ch == '\n')
2517                     cinp();
2518                 t = ch;
2519             }
2520             if (t != '(') /* no macro subst */
2521                 return -1;
2522
2523             /* argument macro */
2524             next_nomacro();
2525             next_nomacro();
2526             args = NULL;
2527             sa = s->next;
2528             /* NOTE: empty args are allowed, except if no args */
2529             for(;;) {
2530                 /* handle '()' case */
2531                 if (!args && !sa && tok == ')')
2532                     break;
2533                 if (!sa)
2534                     error("macro '%s' used with too many args",
2535                           get_tok_str(s->v, 0));
2536                 tok_str_new(&str);
2537                 parlevel = spc = 0;
2538                 /* NOTE: non zero sa->t indicates VA_ARGS */
2539                 while ((parlevel > 0 ||
2540                         (tok != ')' &&
2541                          (tok != ',' || sa->type.t))) &&
2542                        tok != -1) {
2543                     if (tok == '(')
2544                         parlevel++;
2545                     else if (tok == ')')
2546                         parlevel--;
2547                     if (tok == TOK_LINEFEED)
2548                         tok = ' ';
2549                     if (!check_space(tok, &spc))
2550                         tok_str_add2(&str, tok, &tokc);
2551                     next_nomacro_spc();
2552                 }
2553                 str.len -= spc;
2554                 tok_str_add(&str, 0);
2555                 sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, (long)str.str);
2556                 sa = sa->next;
2557                 if (tok == ')') {
2558                     /* special case for gcc var args: add an empty
2559                        var arg argument if it is omitted */
2560                     if (sa && sa->type.t && gnu_ext)
2561                         continue;
2562                     else
2563                         break;
2564                 }
2565                 if (tok != ',')
2566                     expect(",");
2567                 next_nomacro();
2568             }
2569             if (sa) {
2570                 error("macro '%s' used with too few args",
2571                       get_tok_str(s->v, 0));
2572             }
2573
2574             /* now subst each arg */
2575             mstr = macro_arg_subst(nested_list, mstr, args);
2576             /* free memory */
2577             sa = args;
2578             while (sa) {
2579                 sa1 = sa->prev;
2580                 tok_str_free((int *)sa->c);
2581                 sym_free(sa);
2582                 sa = sa1;
2583             }
2584             mstr_allocated = 1;
2585         }
2586         sym_push2(nested_list, s->v, 0, 0);
2587         macro_subst(tok_str, nested_list, mstr, can_read_stream);
2588         /* pop nested defined symbol */
2589         sa1 = *nested_list;
2590         *nested_list = sa1->prev;
2591         sym_free(sa1);
2592         if (mstr_allocated)
2593             tok_str_free(mstr);
2594     }
2595     return 0;
2596 }
2597
2598 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2599    return the resulting string (which must be freed). */
2600 static inline int *macro_twosharps(const int *macro_str)
2601 {
2602     const int *ptr;
2603     int t;
2604     CValue cval;
2605     TokenString macro_str1;
2606     CString cstr;
2607     char *p;
2608     int n;
2609
2610     /* we search the first '##' */
2611     for(ptr = macro_str;;) {
2612         TOK_GET(t, ptr, cval);
2613         if (t == TOK_TWOSHARPS)
2614             break;
2615         /* nothing more to do if end of string */
2616         if (t == 0)
2617             return NULL;
2618     }
2619
2620     /* we saw '##', so we need more processing to handle it */
2621     tok_str_new(&macro_str1);
2622     for(ptr = macro_str;;) {
2623         TOK_GET(tok, ptr, tokc);
2624         if (tok == 0)
2625             break;
2626         if (tok == TOK_TWOSHARPS)
2627             continue;
2628         while (*ptr == TOK_TWOSHARPS) {
2629             t = *++ptr;
2630             if (t && t != TOK_TWOSHARPS) {
2631                 TOK_GET(t, ptr, cval);
2632
2633                 /* We concatenate the two tokens */
2634                 cstr_new(&cstr);
2635                 cstr_cat(&cstr, get_tok_str(tok, &tokc));
2636                 n = cstr.size;
2637                 cstr_cat(&cstr, get_tok_str(t, &cval));
2638                 cstr_ccat(&cstr, '\0');
2639
2640                 p = file->buf_ptr;
2641                 file->buf_ptr = cstr.data;
2642                 for (;;) {
2643                     next_nomacro1();
2644                     if (0 == *file->buf_ptr)
2645                         break;
2646                     tok_str_add2(&macro_str1, tok, &tokc);
2647
2648                     warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
2649                         n, cstr.data, (char*)cstr.data + n);
2650                 }
2651                 file->buf_ptr = p;
2652                 cstr_reset(&cstr);
2653             }
2654         }
2655         tok_str_add2(&macro_str1, tok, &tokc);
2656     }
2657     tok_str_add(&macro_str1, 0);
2658     return macro_str1.str;
2659 }
2660
2661
2662 /* do macro substitution of macro_str and add result to
2663    (tok_str,tok_len). 'nested_list' is the list of all macros we got
2664    inside to avoid recursing. */
2665 static void macro_subst(TokenString *tok_str, Sym **nested_list,
2666                         const int *macro_str, struct macro_level ** can_read_stream)
2667 {
2668     Sym *s;
2669     int *macro_str1;
2670     const int *ptr;
2671     int t, ret, spc;
2672     CValue cval;
2673     struct macro_level ml;
2674
2675     /* first scan for '##' operator handling */
2676     ptr = macro_str;
2677     macro_str1 = macro_twosharps(ptr);
2678     if (macro_str1)
2679         ptr = macro_str1;
2680     spc = 0;
2681     while (1) {
2682         /* NOTE: ptr == NULL can only happen if tokens are read from
2683            file stream due to a macro function call */
2684         if (ptr == NULL)
2685             break;
2686         TOK_GET(t, ptr, cval);
2687         if (t == 0)
2688             break;
2689         s = define_find(t);
2690         if (s != NULL) {
2691             /* if nested substitution, do nothing */
2692             if (sym_find2(*nested_list, t))
2693                 goto no_subst;
2694             ml.p = macro_ptr;
2695             if (can_read_stream)
2696                 ml.prev = *can_read_stream, *can_read_stream = &ml;
2697             macro_ptr = (int *)ptr;
2698             tok = t;
2699             ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
2700             ptr = (int *)macro_ptr;
2701             macro_ptr = ml.p;
2702             if (can_read_stream && *can_read_stream == &ml)
2703                 *can_read_stream = ml.prev;
2704             if (ret != 0)
2705                 goto no_subst;
2706         } else {
2707         no_subst:
2708             if (!check_space(t, &spc))
2709                 tok_str_add2(tok_str, t, &cval);
2710         }
2711     }
2712     if (macro_str1)
2713         tok_str_free(macro_str1);
2714 }
2715
2716 /* return next token with macro substitution */
2717 static void next(void)
2718 {
2719     Sym *nested_list, *s;
2720     TokenString str;
2721     struct macro_level *ml;
2722
2723  redo:
2724     if (parse_flags & PARSE_FLAG_SPACES)
2725         next_nomacro_spc();
2726     else
2727         next_nomacro();
2728     if (!macro_ptr) {
2729         /* if not reading from macro substituted string, then try
2730            to substitute macros */
2731         if (tok >= TOK_IDENT &&
2732             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2733             s = define_find(tok);
2734             if (s) {
2735                 /* we have a macro: we try to substitute */
2736                 tok_str_new(&str);
2737                 nested_list = NULL;
2738                 ml = NULL;
2739                 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
2740                     /* substitution done, NOTE: maybe empty */
2741                     tok_str_add(&str, 0);
2742                     macro_ptr = str.str;
2743                     macro_ptr_allocated = str.str;
2744                     goto redo;
2745                 }
2746             }
2747         }
2748     } else {
2749         if (tok == 0) {
2750             /* end of macro or end of unget buffer */
2751             if (unget_buffer_enabled) {
2752                 macro_ptr = unget_saved_macro_ptr;
2753                 unget_buffer_enabled = 0;
2754             } else {
2755                 /* end of macro string: free it */
2756                 tok_str_free(macro_ptr_allocated);
2757                 macro_ptr = NULL;
2758             }
2759             goto redo;
2760         }
2761     }
2762
2763     /* convert preprocessor tokens into C tokens */
2764     if (tok == TOK_PPNUM &&
2765         (parse_flags & PARSE_FLAG_TOK_NUM)) {
2766         parse_number((char *)tokc.cstr->data);
2767     }
2768 }
2769
2770 /* push back current token and set current token to 'last_tok'. Only
2771    identifier case handled for labels. */
2772 static inline void unget_tok(int last_tok)
2773 {
2774     int i, n;
2775     int *q;
2776     unget_saved_macro_ptr = macro_ptr;
2777     unget_buffer_enabled = 1;
2778     q = unget_saved_buffer;
2779     macro_ptr = q;
2780     *q++ = tok;
2781     n = tok_ext_size(tok) - 1;
2782     for(i=0;i<n;i++)
2783         *q++ = tokc.tab[i];
2784     *q = 0; /* end of token string */
2785     tok = last_tok;
2786 }
2787
2788
2789 /* better than nothing, but needs extension to handle '-E' option
2790    correctly too */
2791 static void preprocess_init(TCCState *s1)
2792 {
2793     s1->include_stack_ptr = s1->include_stack;
2794     /* XXX: move that before to avoid having to initialize
2795        file->ifdef_stack_ptr ? */
2796     s1->ifdef_stack_ptr = s1->ifdef_stack;
2797     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
2798
2799     /* XXX: not ANSI compliant: bound checking says error */
2800     vtop = vstack - 1;
2801     s1->pack_stack[0] = 0;
2802     s1->pack_stack_ptr = s1->pack_stack;
2803 }
2804
2805 void preprocess_new()
2806 {
2807     int i, c;
2808     const char *p, *r;
2809     TokenSym *ts;
2810
2811     /* init isid table */
2812     for(i=CH_EOF;i<256;i++)
2813         isidnum_table[i-CH_EOF] = isid(i) || isnum(i);
2814
2815     /* add all tokens */
2816     table_ident = NULL;
2817     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
2818
2819     tok_ident = TOK_IDENT;
2820     p = tcc_keywords;
2821     while (*p) {
2822         r = p;
2823         for(;;) {
2824             c = *r++;
2825             if (c == '\0')
2826                 break;
2827         }
2828         ts = tok_alloc(p, r - p - 1);
2829         p = r;
2830     }
2831 }
2832
2833 /* Preprocess the current file */
2834 static int tcc_preprocess(TCCState *s1)
2835 {
2836     Sym *define_start;
2837     BufferedFile *file_ref, **iptr, **iptr_new;
2838     int token_seen, line_ref, d;
2839     const char *s;
2840
2841     preprocess_init(s1);
2842     define_start = define_stack;
2843     ch = file->buf_ptr[0];
2844     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
2845     parse_flags = PARSE_FLAG_ASM_COMMENTS | PARSE_FLAG_PREPROCESS |
2846         PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
2847     token_seen = 0;
2848     line_ref = 0;
2849     file_ref = NULL;
2850
2851     iptr = s1->include_stack_ptr;
2852     for (;;) {
2853         next();
2854         if (tok == TOK_EOF) {
2855             break;
2856         } else if (file != file_ref) {
2857             goto print_line;
2858         } else if (tok == TOK_LINEFEED) {
2859             if (!token_seen)
2860                 continue;
2861             ++line_ref;
2862             token_seen = 0;
2863         } else if (!token_seen) {
2864             d = file->line_num - line_ref;
2865             if (file != file_ref || d < 0 || d >= 8) {
2866 print_line:
2867                 iptr_new = s1->include_stack_ptr;
2868                 s = iptr_new > iptr ? " 1"
2869                   : iptr_new < iptr ? " 2"
2870                   : iptr_new > s1->include_stack ? " 3"
2871                   : ""
2872                   ;
2873                 iptr = iptr_new;
2874                 fprintf(s1->outfile, "# %d \"%s\"%s\n", file->line_num, file->filename, s);
2875             } else {
2876                 while (d)
2877                     fputs("\n", s1->outfile), --d;
2878             }
2879             line_ref = (file_ref = file)->line_num;
2880             token_seen = tok != TOK_LINEFEED;
2881             if (!token_seen)
2882                 continue;
2883         }
2884         fputs(get_tok_str(tok, &tokc), s1->outfile);
2885     }
2886     free_defines(define_start);
2887     return 0;
2888 }
2889