tccpp.c

   1 /*
   2  *  TCC - Tiny C Compiler
   3  *
   4  *  Copyright (c) 2001-2004 Fabrice Bellard
   5  *
   6  * This library is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2 of the License, or (at your option) any later version.
  10  *
  11  * This library is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with this library; if not, write to the Free Software
  18  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  19  */
  20
  21 #include "tcc.h"
  22
  23 /********************************************************/
  24 /* global variables */
  25
  26 ST_DATA int tok_flags;
  27 /* additional informations about token */
  28 #define TOK_FLAG_BOL   0x0001 /* beginning of line before */
  29 #define TOK_FLAG_BOF   0x0002 /* beginning of file before */
  30 #define TOK_FLAG_ENDIF 0x0004 /* a endif was found matching starting #ifdef */
  31 #define TOK_FLAG_EOF   0x0008 /* end of file */
  32
  33 ST_DATA int parse_flags;
  34 #define PARSE_FLAG_PREPROCESS 0x0001 /* activate preprocessing */
  35 #define PARSE_FLAG_TOK_NUM    0x0002 /* return numbers instead of TOK_PPNUM */
  36 #define PARSE_FLAG_LINEFEED   0x0004 /* line feed is returned as a
  37                                         token. line feed is also
  38                                         returned at eof */
  39 #define PARSE_FLAG_ASM_COMMENTS 0x0008 /* '#' can be used for line comment */
  40 #define PARSE_FLAG_SPACES     0x0010 /* next() returns space tokens (for -E) */
  41
  42 ST_DATA struct BufferedFile *file;
  43 ST_DATA int ch, tok;
  44 ST_DATA CValue tokc;
  45 ST_DATA const int *macro_ptr;
  46 ST_DATA CString tokcstr; /* current parsed string, if any */
  47
  48 /* display benchmark infos */
  49 ST_DATA int total_lines;
  50 ST_DATA int total_bytes;
  51 ST_DATA int tok_ident;
  52 ST_DATA TokenSym **table_ident;
  53
  54 /* ------------------------------------------------------------------------- */
  55
  56 static int *macro_ptr_allocated;
  57 static const int *unget_saved_macro_ptr;
  58 static int unget_saved_buffer[TOK_MAX_SIZE + 1];
  59 static int unget_buffer_enabled;
  60 static TokenSym *hash_ident[TOK_HASH_SIZE];
  61 static char token_buf[STRING_MAX_SIZE + 1];
  62 /* true if isid(c) || isnum(c) */
  63 static unsigned char isidnum_table[256-CH_EOF];
  64
  65 static const char tcc_keywords[] =
  66 #define DEF(id, str) str "\0"
  67 #include "tcctok.h"
  68 #undef DEF
  69 ;
  70
  71 /* WARNING: the content of this string encodes token numbers */
  72 static const unsigned char tok_two_chars[] =
  73     "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
  74     "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
  75
  76 struct macro_level {
  77     struct macro_level *prev;
  78     const int *p;
  79 };
  80
  81 ST_FUNC void next_nomacro(void);
  82 static void next_nomacro_spc(void);
  83 static void macro_subst(
  84     TokenString *tok_str,
  85     Sym **nested_list,
  86     const int *macro_str,
  87     struct macro_level **can_read_stream
  88     );
  89
  90 ST_FUNC void skip(int c)
  91 {
  92     if (tok != c)
  93         error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
  94     next();
  95 }
  96
  97 /* ------------------------------------------------------------------------- */
  98 /* CString handling */
  99 static void cstr_realloc(CString *cstr, int new_size)
 100 {
 101     int size;
 102     void *data;
 103
 104     size = cstr->size_allocated;
 105     if (size == 0)
 106         size = 8; /* no need to allocate a too small first string */
 107     while (size < new_size)
 108         size = size * 2;
 109     data = tcc_realloc(cstr->data_allocated, size);
 110     if (!data)
 111         error("memory full");
 112     cstr->data_allocated = data;
 113     cstr->size_allocated = size;
 114     cstr->data = data;
 115 }
 116
 117 /* add a byte */
 118 ST_INLN void cstr_ccat(CString *cstr, int ch)
 119 {
 120     int size;
 121     size = cstr->size + 1;
 122     if (size > cstr->size_allocated)
 123         cstr_realloc(cstr, size);
 124     ((unsigned char *)cstr->data)[size - 1] = ch;
 125     cstr->size = size;
 126 }
 127
 128 ST_FUNC void cstr_cat(CString *cstr, const char *str)
 129 {
 130     int c;
 131     for(;;) {
 132         c = *str;
 133         if (c == '\0')
 134             break;
 135         cstr_ccat(cstr, c);
 136         str++;
 137     }
 138 }
 139
 140 /* add a wide char */
 141 ST_FUNC void cstr_wccat(CString *cstr, int ch)
 142 {
 143     int size;
 144     size = cstr->size + sizeof(nwchar_t);
 145     if (size > cstr->size_allocated)
 146         cstr_realloc(cstr, size);
 147     *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
 148     cstr->size = size;
 149 }
 150
 151 ST_FUNC void cstr_new(CString *cstr)
 152 {
 153     memset(cstr, 0, sizeof(CString));
 154 }
 155
 156 /* free string and reset it to NULL */
 157 ST_FUNC void cstr_free(CString *cstr)
 158 {
 159     tcc_free(cstr->data_allocated);
 160     cstr_new(cstr);
 161 }
 162
 163 /* XXX: unicode ? */
 164 ST_FUNC void add_char(CString *cstr, int c)
 165 {
 166     if (c == '\'' || c == '\"' || c == '\\') {
 167         /* XXX: could be more precise if char or string */
 168         cstr_ccat(cstr, '\\');
 169     }
 170     if (c >= 32 && c <= 126) {
 171         cstr_ccat(cstr, c);
 172     } else {
 173         cstr_ccat(cstr, '\\');
 174         if (c == '\n') {
 175             cstr_ccat(cstr, 'n');
 176         } else {
 177             cstr_ccat(cstr, '0' + ((c >> 6) & 7));
 178             cstr_ccat(cstr, '0' + ((c >> 3) & 7));
 179             cstr_ccat(cstr, '0' + (c & 7));
 180         }
 181     }
 182 }
 183
 184 /* ------------------------------------------------------------------------- */
 185 /* allocate a new token */
 186 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
 187 {
 188     TokenSym *ts, **ptable;
 189     int i;
 190
 191     if (tok_ident >= SYM_FIRST_ANOM)
 192         error("memory full");
 193
 194     /* expand token table if needed */
 195     i = tok_ident - TOK_IDENT;
 196     if ((i % TOK_ALLOC_INCR) == 0) {
 197         ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
 198         if (!ptable)
 199             error("memory full");
 200         table_ident = ptable;
 201     }
 202
 203     ts = tcc_malloc(sizeof(TokenSym) + len);
 204     table_ident[i] = ts;
 205     ts->tok = tok_ident++;
 206     ts->sym_define = NULL;
 207     ts->sym_label = NULL;
 208     ts->sym_struct = NULL;
 209     ts->sym_identifier = NULL;
 210     ts->len = len;
 211     ts->hash_next = NULL;
 212     memcpy(ts->str, str, len);
 213     ts->str[len] = '\0';
 214     *pts = ts;
 215     return ts;
 216 }
 217
 218 #define TOK_HASH_INIT 1
 219 #define TOK_HASH_FUNC(h, c) ((h) * 263 + (c))
 220
 221 /* find a token and add it if not found */
 222 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
 223 {
 224     TokenSym *ts, **pts;
 225     int i;
 226     unsigned int h;
 227
 228     h = TOK_HASH_INIT;
 229     for(i=0;i<len;i++)
 230         h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
 231     h &= (TOK_HASH_SIZE - 1);
 232
 233     pts = &hash_ident[h];
 234     for(;;) {
 235         ts = *pts;
 236         if (!ts)
 237             break;
 238         if (ts->len == len && !memcmp(ts->str, str, len))
 239             return ts;
 240         pts = &(ts->hash_next);
 241     }
 242     return tok_alloc_new(pts, str, len);
 243 }
 244
 245 /* XXX: buffer overflow */
 246 /* XXX: float tokens */
 247 ST_FUNC char *get_tok_str(int v, CValue *cv)
 248 {
 249     static char buf[STRING_MAX_SIZE + 1];
 250     static CString cstr_buf;
 251     CString *cstr;
 252     char *p;
 253     int i, len;
 254
 255     /* NOTE: to go faster, we give a fixed buffer for small strings */
 256     cstr_reset(&cstr_buf);
 257     cstr_buf.data = buf;
 258     cstr_buf.size_allocated = sizeof(buf);
 259     p = buf;
 260
 261     switch(v) {
 262     case TOK_CINT:
 263     case TOK_CUINT:
 264         /* XXX: not quite exact, but only useful for testing */
 265         sprintf(p, "%u", cv->ui);
 266         break;
 267     case TOK_CLLONG:
 268     case TOK_CULLONG:
 269         /* XXX: not quite exact, but only useful for testing  */
 270 #ifdef _WIN32
 271         sprintf(p, "%u", (unsigned)cv->ull);
 272 #else
 273         sprintf(p, "%Lu", cv->ull);
 274 #endif
 275         break;
 276     case TOK_LCHAR:
 277         cstr_ccat(&cstr_buf, 'L');
 278     case TOK_CCHAR:
 279         cstr_ccat(&cstr_buf, '\'');
 280         add_char(&cstr_buf, cv->i);
 281         cstr_ccat(&cstr_buf, '\'');
 282         cstr_ccat(&cstr_buf, '\0');
 283         break;
 284     case TOK_PPNUM:
 285         cstr = cv->cstr;
 286         len = cstr->size - 1;
 287         for(i=0;i<len;i++)
 288             add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 289         cstr_ccat(&cstr_buf, '\0');
 290         break;
 291     case TOK_LSTR:
 292         cstr_ccat(&cstr_buf, 'L');
 293     case TOK_STR:
 294         cstr = cv->cstr;
 295         cstr_ccat(&cstr_buf, '\"');
 296         if (v == TOK_STR) {
 297             len = cstr->size - 1;
 298             for(i=0;i<len;i++)
 299                 add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 300         } else {
 301             len = (cstr->size / sizeof(nwchar_t)) - 1;
 302             for(i=0;i<len;i++)
 303                 add_char(&cstr_buf, ((nwchar_t *)cstr->data)[i]);
 304         }
 305         cstr_ccat(&cstr_buf, '\"');
 306         cstr_ccat(&cstr_buf, '\0');
 307         break;
 308     case TOK_LT:
 309         v = '<';
 310         goto addv;
 311     case TOK_GT:
 312         v = '>';
 313         goto addv;
 314     case TOK_DOTS:
 315         return strcpy(p, "...");
 316     case TOK_A_SHL:
 317         return strcpy(p, "<<=");
 318     case TOK_A_SAR:
 319         return strcpy(p, ">>=");
 320     default:
 321         if (v < TOK_IDENT) {
 322             /* search in two bytes table */
 323             const unsigned char *q = tok_two_chars;
 324             while (*q) {
 325                 if (q[2] == v) {
 326                     *p++ = q[0];
 327                     *p++ = q[1];
 328                     *p = '\0';
 329                     return buf;
 330                 }
 331                 q += 3;
 332             }
 333         addv:
 334             *p++ = v;
 335             *p = '\0';
 336         } else if (v < tok_ident) {
 337             return table_ident[v - TOK_IDENT]->str;
 338         } else if (v >= SYM_FIRST_ANOM) {
 339             /* special name for anonymous symbol */
 340             sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
 341         } else {
 342             /* should never happen */
 343             return NULL;
 344         }
 345         break;
 346     }
 347     return cstr_buf.data;
 348 }
 349
 350 /* fill input buffer and peek next char */
 351 static int tcc_peekc_slow(BufferedFile *bf)
 352 {
 353     int len;
 354     /* only tries to read if really end of buffer */
 355     if (bf->buf_ptr >= bf->buf_end) {
 356         if (bf->fd != -1) {
 357 #if defined(PARSE_DEBUG)
 358             len = 8;
 359 #else
 360             len = IO_BUF_SIZE;
 361 #endif
 362             len = read(bf->fd, bf->buffer, len);
 363             if (len < 0)
 364                 len = 0;
 365         } else {
 366             len = 0;
 367         }
 368         total_bytes += len;
 369         bf->buf_ptr = bf->buffer;
 370         bf->buf_end = bf->buffer + len;
 371         *bf->buf_end = CH_EOB;
 372     }
 373     if (bf->buf_ptr < bf->buf_end) {
 374         return bf->buf_ptr[0];
 375     } else {
 376         bf->buf_ptr = bf->buf_end;
 377         return CH_EOF;
 378     }
 379 }
 380
 381 /* return the current character, handling end of block if necessary
 382    (but not stray) */
 383 ST_FUNC int handle_eob(void)
 384 {
 385     return tcc_peekc_slow(file);
 386 }
 387
 388 /* read next char from current input file and handle end of input buffer */
 389 ST_INLN void inp(void)
 390 {
 391     ch = *(++(file->buf_ptr));
 392     /* end of buffer/file handling */
 393     if (ch == CH_EOB)
 394         ch = handle_eob();
 395 }
 396
 397 /* handle '\[\r]\n' */
 398 static int handle_stray_noerror(void)
 399 {
 400     while (ch == '\\') {
 401         inp();
 402         if (ch == '\n') {
 403             file->line_num++;
 404             inp();
 405         } else if (ch == '\r') {
 406             inp();
 407             if (ch != '\n')
 408                 goto fail;
 409             file->line_num++;
 410             inp();
 411         } else {
 412         fail:
 413             return 1;
 414         }
 415     }
 416     return 0;
 417 }
 418
 419 static void handle_stray(void)
 420 {
 421     if (handle_stray_noerror())
 422         error("stray '\\' in program");
 423 }
 424
 425 /* skip the stray and handle the \\n case. Output an error if
 426    incorrect char after the stray */
 427 static int handle_stray1(uint8_t *p)
 428 {
 429     int c;
 430
 431     if (p >= file->buf_end) {
 432         file->buf_ptr = p;
 433         c = handle_eob();
 434         p = file->buf_ptr;
 435         if (c == '\\')
 436             goto parse_stray;
 437     } else {
 438     parse_stray:
 439         file->buf_ptr = p;
 440         ch = *p;
 441         handle_stray();
 442         p = file->buf_ptr;
 443         c = *p;
 444     }
 445     return c;
 446 }
 447
 448 /* handle just the EOB case, but not stray */
 449 #define PEEKC_EOB(c, p)\
 450 {\
 451     p++;\
 452     c = *p;\
 453     if (c == '\\') {\
 454         file->buf_ptr = p;\
 455         c = handle_eob();\
 456         p = file->buf_ptr;\
 457     }\
 458 }
 459
 460 /* handle the complicated stray case */
 461 #define PEEKC(c, p)\
 462 {\
 463     p++;\
 464     c = *p;\
 465     if (c == '\\') {\
 466         c = handle_stray1(p);\
 467         p = file->buf_ptr;\
 468     }\
 469 }
 470
 471 /* input with '\[\r]\n' handling. Note that this function cannot
 472    handle other characters after '\', so you cannot call it inside
 473    strings or comments */
 474 ST_FUNC void minp(void)
 475 {
 476     inp();
 477     if (ch == '\\')
 478         handle_stray();
 479 }
 480
 481
 482 /* single line C++ comments */
 483 static uint8_t *parse_line_comment(uint8_t *p)
 484 {
 485     int c;
 486
 487     p++;
 488     for(;;) {
 489         c = *p;
 490     redo:
 491         if (c == '\n' || c == CH_EOF) {
 492             break;
 493         } else if (c == '\\') {
 494             file->buf_ptr = p;
 495             c = handle_eob();
 496             p = file->buf_ptr;
 497             if (c == '\\') {
 498                 PEEKC_EOB(c, p);
 499                 if (c == '\n') {
 500                     file->line_num++;
 501                     PEEKC_EOB(c, p);
 502                 } else if (c == '\r') {
 503                     PEEKC_EOB(c, p);
 504                     if (c == '\n') {
 505                         file->line_num++;
 506                         PEEKC_EOB(c, p);
 507                     }
 508                 }
 509             } else {
 510                 goto redo;
 511             }
 512         } else {
 513             p++;
 514         }
 515     }
 516     return p;
 517 }
 518
 519 /* C comments */
 520 ST_FUNC uint8_t *parse_comment(uint8_t *p)
 521 {
 522     int c;
 523
 524     p++;
 525     for(;;) {
 526         /* fast skip loop */
 527         for(;;) {
 528             c = *p;
 529             if (c == '\n' || c == '*' || c == '\\')
 530                 break;
 531             p++;
 532             c = *p;
 533             if (c == '\n' || c == '*' || c == '\\')
 534                 break;
 535             p++;
 536         }
 537         /* now we can handle all the cases */
 538         if (c == '\n') {
 539             file->line_num++;
 540             p++;
 541         } else if (c == '*') {
 542             p++;
 543             for(;;) {
 544                 c = *p;
 545                 if (c == '*') {
 546                     p++;
 547                 } else if (c == '/') {
 548                     goto end_of_comment;
 549                 } else if (c == '\\') {
 550                     file->buf_ptr = p;
 551                     c = handle_eob();
 552                     p = file->buf_ptr;
 553                     if (c == '\\') {
 554                         /* skip '\[\r]\n', otherwise just skip the stray */
 555                         while (c == '\\') {
 556                             PEEKC_EOB(c, p);
 557                             if (c == '\n') {
 558                                 file->line_num++;
 559                                 PEEKC_EOB(c, p);
 560                             } else if (c == '\r') {
 561                                 PEEKC_EOB(c, p);
 562                                 if (c == '\n') {
 563                                     file->line_num++;
 564                                     PEEKC_EOB(c, p);
 565                                 }
 566                             } else {
 567                                 goto after_star;
 568                             }
 569                         }
 570                     }
 571                 } else {
 572                     break;
 573                 }
 574             }
 575         after_star: ;
 576         } else {
 577             /* stray, eob or eof */
 578             file->buf_ptr = p;
 579             c = handle_eob();
 580             p = file->buf_ptr;
 581             if (c == CH_EOF) {
 582                 error("unexpected end of file in comment");
 583             } else if (c == '\\') {
 584                 p++;
 585             }
 586         }
 587     }
 588  end_of_comment:
 589     p++;
 590     return p;
 591 }
 592
 593 #define cinp minp
 594
 595 static inline void skip_spaces(void)
 596 {
 597     while (is_space(ch))
 598         cinp();
 599 }
 600
 601 static inline int check_space(int t, int *spc)
 602 {
 603     if (is_space(t)) {
 604         if (*spc)
 605             return 1;
 606         *spc = 1;
 607     } else
 608         *spc = 0;
 609     return 0;
 610 }
 611
 612 /* parse a string without interpreting escapes */
 613 static uint8_t *parse_pp_string(uint8_t *p,
 614                                 int sep, CString *str)
 615 {
 616     int c;
 617     p++;
 618     for(;;) {
 619         c = *p;
 620         if (c == sep) {
 621             break;
 622         } else if (c == '\\') {
 623             file->buf_ptr = p;
 624             c = handle_eob();
 625             p = file->buf_ptr;
 626             if (c == CH_EOF) {
 627             unterminated_string:
 628                 /* XXX: indicate line number of start of string */
 629                 error("missing terminating %c character", sep);
 630             } else if (c == '\\') {
 631                 /* escape : just skip \[\r]\n */
 632                 PEEKC_EOB(c, p);
 633                 if (c == '\n') {
 634                     file->line_num++;
 635                     p++;
 636                 } else if (c == '\r') {
 637                     PEEKC_EOB(c, p);
 638                     if (c != '\n')
 639                         expect("'\n' after '\r'");
 640                     file->line_num++;
 641                     p++;
 642                 } else if (c == CH_EOF) {
 643                     goto unterminated_string;
 644                 } else {
 645                     if (str) {
 646                         cstr_ccat(str, '\\');
 647                         cstr_ccat(str, c);
 648                     }
 649                     p++;
 650                 }
 651             }
 652         } else if (c == '\n') {
 653             file->line_num++;
 654             goto add_char;
 655         } else if (c == '\r') {
 656             PEEKC_EOB(c, p);
 657             if (c != '\n') {
 658                 if (str)
 659                     cstr_ccat(str, '\r');
 660             } else {
 661                 file->line_num++;
 662                 goto add_char;
 663             }
 664         } else {
 665         add_char:
 666             if (str)
 667                 cstr_ccat(str, c);
 668             p++;
 669         }
 670     }
 671     p++;
 672     return p;
 673 }
 674
 675 /* skip block of text until #else, #elif or #endif. skip also pairs of
 676    #if/#endif */
 677 static void preprocess_skip(void)
 678 {
 679     int a, start_of_line, c, in_warn_or_error;
 680     uint8_t *p;
 681
 682     p = file->buf_ptr;
 683     a = 0;
 684 redo_start:
 685     start_of_line = 1;
 686     in_warn_or_error = 0;
 687     for(;;) {
 688     redo_no_start:
 689         c = *p;
 690         switch(c) {
 691         case ' ':
 692         case '\t':
 693         case '\f':
 694         case '\v':
 695         case '\r':
 696             p++;
 697             goto redo_no_start;
 698         case '\n':
 699             file->line_num++;
 700             p++;
 701             goto redo_start;
 702         case '\\':
 703             file->buf_ptr = p;
 704             c = handle_eob();
 705             if (c == CH_EOF) {
 706                 expect("#endif");
 707             } else if (c == '\\') {
 708                 ch = file->buf_ptr[0];
 709                 handle_stray_noerror();
 710             }
 711             p = file->buf_ptr;
 712             goto redo_no_start;
 713         /* skip strings */
 714         case '\"':
 715         case '\'':
 716             if (in_warn_or_error)
 717                 goto _default;
 718             p = parse_pp_string(p, c, NULL);
 719             break;
 720         /* skip comments */
 721         case '/':
 722             if (in_warn_or_error)
 723                 goto _default;
 724             file->buf_ptr = p;
 725             ch = *p;
 726             minp();
 727             p = file->buf_ptr;
 728             if (ch == '*') {
 729                 p = parse_comment(p);
 730             } else if (ch == '/') {
 731                 p = parse_line_comment(p);
 732             }
 733             break;
 734         case '#':
 735             p++;
 736             if (start_of_line) {
 737                 file->buf_ptr = p;
 738                 next_nomacro();
 739                 p = file->buf_ptr;
 740                 if (a == 0 &&
 741                     (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
 742                     goto the_end;
 743                 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
 744                     a++;
 745                 else if (tok == TOK_ENDIF)
 746                     a--;
 747                 else if( tok == TOK_ERROR || tok == TOK_WARNING)
 748                     in_warn_or_error = 1;
 749                 else if (tok == TOK_LINEFEED)
 750                     goto redo_start;
 751             }
 752             break;
 753 _default:
 754         default:
 755             p++;
 756             break;
 757         }
 758         start_of_line = 0;
 759     }
 760  the_end: ;
 761     file->buf_ptr = p;
 762 }
 763
 764 /* ParseState handling */
 765
 766 /* XXX: currently, no include file info is stored. Thus, we cannot display
 767    accurate messages if the function or data definition spans multiple
 768    files */
 769
 770 /* save current parse state in 's' */
 771 ST_FUNC void save_parse_state(ParseState *s)
 772 {
 773     s->line_num = file->line_num;
 774     s->macro_ptr = macro_ptr;
 775     s->tok = tok;
 776     s->tokc = tokc;
 777 }
 778
 779 /* restore parse state from 's' */
 780 ST_FUNC void restore_parse_state(ParseState *s)
 781 {
 782     file->line_num = s->line_num;
 783     macro_ptr = s->macro_ptr;
 784     tok = s->tok;
 785     tokc = s->tokc;
 786 }
 787
 788 /* return the number of additional 'ints' necessary to store the
 789    token */
 790 static inline int tok_ext_size(int t)
 791 {
 792     switch(t) {
 793         /* 4 bytes */
 794     case TOK_CINT:
 795     case TOK_CUINT:
 796     case TOK_CCHAR:
 797     case TOK_LCHAR:
 798     case TOK_CFLOAT:
 799     case TOK_LINENUM:
 800         return 1;
 801     case TOK_STR:
 802     case TOK_LSTR:
 803     case TOK_PPNUM:
 804         error("unsupported token");
 805         return 1;
 806     case TOK_CDOUBLE:
 807     case TOK_CLLONG:
 808     case TOK_CULLONG:
 809         return 2;
 810     case TOK_CLDOUBLE:
 811         return LDOUBLE_SIZE / 4;
 812     default:
 813         return 0;
 814     }
 815 }
 816
 817 /* token string handling */
 818
 819 ST_INLN void tok_str_new(TokenString *s)
 820 {
 821     s->str = NULL;
 822     s->len = 0;
 823     s->allocated_len = 0;
 824     s->last_line_num = -1;
 825 }
 826
 827 ST_FUNC void tok_str_free(int *str)
 828 {
 829     tcc_free(str);
 830 }
 831
 832 static int *tok_str_realloc(TokenString *s)
 833 {
 834     int *str, len;
 835
 836     if (s->allocated_len == 0) {
 837         len = 8;
 838     } else {
 839         len = s->allocated_len * 2;
 840     }
 841     str = tcc_realloc(s->str, len * sizeof(int));
 842     if (!str)
 843         error("memory full");
 844     s->allocated_len = len;
 845     s->str = str;
 846     return str;
 847 }
 848
 849 ST_FUNC void tok_str_add(TokenString *s, int t)
 850 {
 851     int len, *str;
 852
 853     len = s->len;
 854     str = s->str;
 855     if (len >= s->allocated_len)
 856         str = tok_str_realloc(s);
 857     str[len++] = t;
 858     s->len = len;
 859 }
 860
 861 static void tok_str_add2(TokenString *s, int t, CValue *cv)
 862 {
 863     int len, *str;
 864
 865     len = s->len;
 866     str = s->str;
 867
 868     /* allocate space for worst case */
 869     if (len + TOK_MAX_SIZE > s->allocated_len)
 870         str = tok_str_realloc(s);
 871     str[len++] = t;
 872     switch(t) {
 873     case TOK_CINT:
 874     case TOK_CUINT:
 875     case TOK_CCHAR:
 876     case TOK_LCHAR:
 877     case TOK_CFLOAT:
 878     case TOK_LINENUM:
 879         str[len++] = cv->tab[0];
 880         break;
 881     case TOK_PPNUM:
 882     case TOK_STR:
 883     case TOK_LSTR:
 884         {
 885             int nb_words;
 886             CString *cstr;
 887
 888             nb_words = (sizeof(CString) + cv->cstr->size + 3) >> 2;
 889             while ((len + nb_words) > s->allocated_len)
 890                 str = tok_str_realloc(s);
 891             cstr = (CString *)(str + len);
 892             cstr->data = NULL;
 893             cstr->size = cv->cstr->size;
 894             cstr->data_allocated = NULL;
 895             cstr->size_allocated = cstr->size;
 896             memcpy((char *)cstr + sizeof(CString),
 897                    cv->cstr->data, cstr->size);
 898             len += nb_words;
 899         }
 900         break;
 901     case TOK_CDOUBLE:
 902     case TOK_CLLONG:
 903     case TOK_CULLONG:
 904 #if LDOUBLE_SIZE == 8
 905     case TOK_CLDOUBLE:
 906 #endif
 907         str[len++] = cv->tab[0];
 908         str[len++] = cv->tab[1];
 909         break;
 910 #if LDOUBLE_SIZE == 12
 911     case TOK_CLDOUBLE:
 912         str[len++] = cv->tab[0];
 913         str[len++] = cv->tab[1];
 914         str[len++] = cv->tab[2];
 915 #elif LDOUBLE_SIZE == 16
 916     case TOK_CLDOUBLE:
 917         str[len++] = cv->tab[0];
 918         str[len++] = cv->tab[1];
 919         str[len++] = cv->tab[2];
 920         str[len++] = cv->tab[3];
 921 #elif LDOUBLE_SIZE != 8
 922 #error add long double size support
 923 #endif
 924         break;
 925     default:
 926         break;
 927     }
 928     s->len = len;
 929 }
 930
 931 /* add the current parse token in token string 's' */
 932 ST_FUNC void tok_str_add_tok(TokenString *s)
 933 {
 934     CValue cval;
 935
 936     /* save line number info */
 937     if (file->line_num != s->last_line_num) {
 938         s->last_line_num = file->line_num;
 939         cval.i = s->last_line_num;
 940         tok_str_add2(s, TOK_LINENUM, &cval);
 941     }
 942     tok_str_add2(s, tok, &tokc);
 943 }
 944
 945 /* get a token from an integer array and increment pointer
 946    accordingly. we code it as a macro to avoid pointer aliasing. */
 947 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
 948 {
 949     const int *p = *pp;
 950     int n, *tab;
 951
 952     tab = cv->tab;
 953     switch(*t = *p++) {
 954     case TOK_CINT:
 955     case TOK_CUINT:
 956     case TOK_CCHAR:
 957     case TOK_LCHAR:
 958     case TOK_CFLOAT:
 959     case TOK_LINENUM:
 960         tab[0] = *p++;
 961         break;
 962     case TOK_STR:
 963     case TOK_LSTR:
 964     case TOK_PPNUM:
 965         cv->cstr = (CString *)p;
 966         cv->cstr->data = (char *)p + sizeof(CString);
 967         p += (sizeof(CString) + cv->cstr->size + 3) >> 2;
 968         break;
 969     case TOK_CDOUBLE:
 970     case TOK_CLLONG:
 971     case TOK_CULLONG:
 972         n = 2;
 973         goto copy;
 974     case TOK_CLDOUBLE:
 975 #if LDOUBLE_SIZE == 16
 976         n = 4;
 977 #elif LDOUBLE_SIZE == 12
 978         n = 3;
 979 #elif LDOUBLE_SIZE == 8
 980         n = 2;
 981 #else
 982 # error add long double size support
 983 #endif
 984     copy:
 985         do
 986             *tab++ = *p++;
 987         while (--n);
 988         break;
 989     default:
 990         break;
 991     }
 992     *pp = p;
 993 }
 994
 995 static int macro_is_equal(const int *a, const int *b)
 996 {
 997     char buf[STRING_MAX_SIZE + 1];
 998     CValue cv;
 999     int t;
1000     while (*a && *b) {
1001         TOK_GET(&t, &a, &cv);
1002         pstrcpy(buf, sizeof buf, get_tok_str(t, &cv));
1003         TOK_GET(&t, &b, &cv);
1004         if (strcmp(buf, get_tok_str(t, &cv)))
1005             return 0;
1006     }
1007     return !(*a || *b);
1008 }
1009
1010 /* defines handling */
1011 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1012 {
1013     Sym *s;
1014
1015     s = define_find(v);
1016     if (s && !macro_is_equal(s->d, str))
1017         warning("%s redefined", get_tok_str(v, NULL));
1018
1019     s = sym_push2(&define_stack, v, macro_type, 0);
1020     s->d = str;
1021     s->next = first_arg;
1022     table_ident[v - TOK_IDENT]->sym_define = s;
1023 }
1024
1025 /* undefined a define symbol. Its name is just set to zero */
1026 ST_FUNC void define_undef(Sym *s)
1027 {
1028     int v;
1029     v = s->v;
1030     if (v >= TOK_IDENT && v < tok_ident)
1031         table_ident[v - TOK_IDENT]->sym_define = NULL;
1032     s->v = 0;
1033 }
1034
1035 ST_INLN Sym *define_find(int v)
1036 {
1037     v -= TOK_IDENT;
1038     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1039         return NULL;
1040     return table_ident[v]->sym_define;
1041 }
1042
1043 /* free define stack until top reaches 'b' */
1044 ST_FUNC void free_defines(Sym *b)
1045 {
1046     Sym *top, *top1;
1047     int v;
1048
1049     top = define_stack;
1050     while (top != b) {
1051         top1 = top->prev;
1052         /* do not free args or predefined defines */
1053         if (top->d)
1054             tok_str_free(top->d);
1055         v = top->v;
1056         if (v >= TOK_IDENT && v < tok_ident)
1057             table_ident[v - TOK_IDENT]->sym_define = NULL;
1058         sym_free(top);
1059         top = top1;
1060     }
1061     define_stack = b;
1062 }
1063
1064 /* label lookup */
1065 ST_FUNC Sym *label_find(int v)
1066 {
1067     v -= TOK_IDENT;
1068     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1069         return NULL;
1070     return table_ident[v]->sym_label;
1071 }
1072
1073 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1074 {
1075     Sym *s, **ps;
1076     s = sym_push2(ptop, v, 0, 0);
1077     s->r = flags;
1078     ps = &table_ident[v - TOK_IDENT]->sym_label;
1079     if (ptop == &global_label_stack) {
1080         /* modify the top most local identifier, so that
1081            sym_identifier will point to 's' when popped */
1082         while (*ps != NULL)
1083             ps = &(*ps)->prev_tok;
1084     }
1085     s->prev_tok = *ps;
1086     *ps = s;
1087     return s;
1088 }
1089
1090 /* pop labels until element last is reached. Look if any labels are
1091    undefined. Define symbols if '&&label' was used. */
1092 ST_FUNC void label_pop(Sym **ptop, Sym *slast)
1093 {
1094     Sym *s, *s1;
1095     for(s = *ptop; s != slast; s = s1) {
1096         s1 = s->prev;
1097         if (s->r == LABEL_DECLARED) {
1098             warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1099         } else if (s->r == LABEL_FORWARD) {
1100                 error("label '%s' used but not defined",
1101                       get_tok_str(s->v, NULL));
1102         } else {
1103             if (s->c) {
1104                 /* define corresponding symbol. A size of
1105                    1 is put. */
1106                 put_extern_sym(s, cur_text_section, s->jnext, 1);
1107             }
1108         }
1109         /* remove label */
1110         table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1111         sym_free(s);
1112     }
1113     *ptop = slast;
1114 }
1115
1116 /* eval an expression for #if/#elif */
1117 static int expr_preprocess(void)
1118 {
1119     int c, t;
1120     TokenString str;
1121
1122     tok_str_new(&str);
1123     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1124         next(); /* do macro subst */
1125         if (tok == TOK_DEFINED) {
1126             next_nomacro();
1127             t = tok;
1128             if (t == '(')
1129                 next_nomacro();
1130             c = define_find(tok) != 0;
1131             if (t == '(')
1132                 next_nomacro();
1133             tok = TOK_CINT;
1134             tokc.i = c;
1135         } else if (tok >= TOK_IDENT) {
1136             /* if undefined macro */
1137             tok = TOK_CINT;
1138             tokc.i = 0;
1139         }
1140         tok_str_add_tok(&str);
1141     }
1142     tok_str_add(&str, -1); /* simulate end of file */
1143     tok_str_add(&str, 0);
1144     /* now evaluate C constant expression */
1145     macro_ptr = str.str;
1146     next();
1147     c = expr_const();
1148     macro_ptr = NULL;
1149     tok_str_free(str.str);
1150     return c != 0;
1151 }
1152
1153 #if defined(PARSE_DEBUG) || defined(PP_DEBUG)
1154 static void tok_print(int *str)
1155 {
1156     int t;
1157     CValue cval;
1158
1159     printf("<");
1160     while (1) {
1161         TOK_GET(&t, &str, &cval);
1162         if (!t)
1163             break;
1164         printf("%s", get_tok_str(t, &cval));
1165     }
1166     printf(">\n");
1167 }
1168 #endif
1169
1170 /* parse after #define */
1171 ST_FUNC void parse_define(void)
1172 {
1173     Sym *s, *first, **ps;
1174     int v, t, varg, is_vaargs, spc;
1175     TokenString str;
1176
1177     v = tok;
1178     if (v < TOK_IDENT)
1179         error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1180     /* XXX: should check if same macro (ANSI) */
1181     first = NULL;
1182     t = MACRO_OBJ;
1183     /* '(' must be just after macro definition for MACRO_FUNC */
1184     next_nomacro_spc();
1185     if (tok == '(') {
1186         next_nomacro();
1187         ps = &first;
1188         while (tok != ')') {
1189             varg = tok;
1190             next_nomacro();
1191             is_vaargs = 0;
1192             if (varg == TOK_DOTS) {
1193                 varg = TOK___VA_ARGS__;
1194                 is_vaargs = 1;
1195             } else if (tok == TOK_DOTS && gnu_ext) {
1196                 is_vaargs = 1;
1197                 next_nomacro();
1198             }
1199             if (varg < TOK_IDENT)
1200                 error("badly punctuated parameter list");
1201             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1202             *ps = s;
1203             ps = &s->next;
1204             if (tok != ',')
1205                 break;
1206             next_nomacro();
1207         }
1208         if (tok == ')')
1209             next_nomacro_spc();
1210         t = MACRO_FUNC;
1211     }
1212     tok_str_new(&str);
1213     spc = 2;
1214     /* EOF testing necessary for '-D' handling */
1215     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1216         /* remove spaces around ## and after '#' */
1217         if (TOK_TWOSHARPS == tok) {
1218             if (1 == spc)
1219                 --str.len;
1220             spc = 2;
1221         } else if ('#' == tok) {
1222             spc = 2;
1223         } else if (check_space(tok, &spc)) {
1224             goto skip;
1225         }
1226         tok_str_add2(&str, tok, &tokc);
1227     skip:
1228         next_nomacro_spc();
1229     }
1230     if (spc == 1)
1231         --str.len; /* remove trailing space */
1232     tok_str_add(&str, 0);
1233 #ifdef PP_DEBUG
1234     printf("define %s %d: ", get_tok_str(v, NULL), t);
1235     tok_print(str.str);
1236 #endif
1237     define_push(v, t, str.str, first);
1238 }
1239
1240 static inline int hash_cached_include(int type, const char *filename)
1241 {
1242     const unsigned char *s;
1243     unsigned int h;
1244
1245     h = TOK_HASH_INIT;
1246     h = TOK_HASH_FUNC(h, type);
1247     s = filename;
1248     while (*s) {
1249         h = TOK_HASH_FUNC(h, *s);
1250         s++;
1251     }
1252     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1253     return h;
1254 }
1255
1256 /* XXX: use a token or a hash table to accelerate matching ? */
1257 static CachedInclude *search_cached_include(TCCState *s1,
1258                                             int type, const char *filename)
1259 {
1260     CachedInclude *e;
1261     int i, h;
1262     h = hash_cached_include(type, filename);
1263     i = s1->cached_includes_hash[h];
1264     for(;;) {
1265         if (i == 0)
1266             break;
1267         e = s1->cached_includes[i - 1];
1268         if (e->type == type && !PATHCMP(e->filename, filename))
1269             return e;
1270         i = e->hash_next;
1271     }
1272     return NULL;
1273 }
1274
1275 static inline void add_cached_include(TCCState *s1, int type,
1276                                       const char *filename, int ifndef_macro)
1277 {
1278     CachedInclude *e;
1279     int h;
1280
1281     if (search_cached_include(s1, type, filename))
1282         return;
1283 #ifdef INC_DEBUG
1284     printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1285 #endif
1286     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1287     if (!e)
1288         return;
1289     e->type = type;
1290     strcpy(e->filename, filename);
1291     e->ifndef_macro = ifndef_macro;
1292     dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1293     /* add in hash table */
1294     h = hash_cached_include(type, filename);
1295     e->hash_next = s1->cached_includes_hash[h];
1296     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1297 }
1298
1299 static void pragma_parse(TCCState *s1)
1300 {
1301     int val;
1302
1303     next();
1304     if (tok == TOK_pack) {
1305         /*
1306           This may be:
1307           #pragma pack(1) // set
1308           #pragma pack() // reset to default
1309           #pragma pack(push,1) // push & set
1310           #pragma pack(pop) // restore previous
1311         */
1312         next();
1313         skip('(');
1314         if (tok == TOK_ASM_pop) {
1315             next();
1316             if (s1->pack_stack_ptr <= s1->pack_stack) {
1317             stk_error:
1318                 error("out of pack stack");
1319             }
1320             s1->pack_stack_ptr--;
1321         } else {
1322             val = 0;
1323             if (tok != ')') {
1324                 if (tok == TOK_ASM_push) {
1325                     next();
1326                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1327                         goto stk_error;
1328                     s1->pack_stack_ptr++;
1329                     skip(',');
1330                 }
1331                 if (tok != TOK_CINT) {
1332                 pack_error:
1333                     error("invalid pack pragma");
1334                 }
1335                 val = tokc.i;
1336                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1337                     goto pack_error;
1338                 next();
1339             }
1340             *s1->pack_stack_ptr = val;
1341             skip(')');
1342         }
1343     }
1344 }
1345
1346 /* is_bof is true if first non space token at beginning of file */
1347 ST_FUNC void preprocess(int is_bof)
1348 {
1349     TCCState *s1 = tcc_state;
1350     int i, c, n, saved_parse_flags;
1351     char buf[1024], *q;
1352     Sym *s;
1353
1354     saved_parse_flags = parse_flags;
1355     parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM |
1356         PARSE_FLAG_LINEFEED;
1357     next_nomacro();
1358  redo:
1359     switch(tok) {
1360     case TOK_DEFINE:
1361         next_nomacro();
1362         parse_define();
1363         break;
1364     case TOK_UNDEF:
1365         next_nomacro();
1366         s = define_find(tok);
1367         /* undefine symbol by putting an invalid name */
1368         if (s)
1369             define_undef(s);
1370         break;
1371     case TOK_INCLUDE:
1372     case TOK_INCLUDE_NEXT:
1373         ch = file->buf_ptr[0];
1374         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1375         skip_spaces();
1376         if (ch == '<') {
1377             c = '>';
1378             goto read_name;
1379         } else if (ch == '\"') {
1380             c = ch;
1381         read_name:
1382             inp();
1383             q = buf;
1384             while (ch != c && ch != '\n' && ch != CH_EOF) {
1385                 if ((q - buf) < sizeof(buf) - 1)
1386                     *q++ = ch;
1387                 if (ch == '\\') {
1388                     if (handle_stray_noerror() == 0)
1389                         --q;
1390                 } else
1391                     inp();
1392             }
1393             *q = '\0';
1394             minp();
1395 #if 0
1396             /* eat all spaces and comments after include */
1397             /* XXX: slightly incorrect */
1398             while (ch1 != '\n' && ch1 != CH_EOF)
1399                 inp();
1400 #endif
1401         } else {
1402             /* computed #include : either we have only strings or
1403                we have anything enclosed in '<>' */
1404             next();
1405             buf[0] = '\0';
1406             if (tok == TOK_STR) {
1407                 while (tok != TOK_LINEFEED) {
1408                     if (tok != TOK_STR) {
1409                     include_syntax:
1410                         error("'#include' expects \"FILENAME\" or <FILENAME>");
1411                     }
1412                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1413                     next();
1414                 }
1415                 c = '\"';
1416             } else {
1417                 int len;
1418                 while (tok != TOK_LINEFEED) {
1419                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1420                     next();
1421                 }
1422                 len = strlen(buf);
1423                 /* check syntax and remove '<>' */
1424                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1425                     goto include_syntax;
1426                 memmove(buf, buf + 1, len - 2);
1427                 buf[len - 2] = '\0';
1428                 c = '>';
1429             }
1430         }
1431
1432         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1433             error("#include recursion too deep");
1434
1435         n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1436         for (i = -2; i < n; ++i) {
1437             char buf1[sizeof file->filename];
1438             CachedInclude *e;
1439             const char *path;
1440             int size, fd;
1441
1442             if (i == -2) {
1443                 /* check absolute include path */
1444                 if (!IS_ABSPATH(buf))
1445                     continue;
1446                 buf1[0] = 0;
1447
1448             } else if (i == -1) {
1449                 /* search in current dir if "header.h" */
1450                 if (c != '\"')
1451                     continue;
1452                 size = tcc_basename(file->filename) - file->filename;
1453                 memcpy(buf1, file->filename, size);
1454                 buf1[size] = '\0';
1455
1456             } else {
1457                 /* search in all the include paths */
1458                 if (i < s1->nb_include_paths)
1459                     path = s1->include_paths[i];
1460                 else
1461                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1462                 pstrcpy(buf1, sizeof(buf1), path);
1463                 pstrcat(buf1, sizeof(buf1), "/");
1464             }
1465
1466             pstrcat(buf1, sizeof(buf1), buf);
1467
1468             e = search_cached_include(s1, c, buf1);
1469             if (e && define_find(e->ifndef_macro)) {
1470                 /* no need to parse the include because the 'ifndef macro'
1471                    is defined */
1472 #ifdef INC_DEBUG
1473                 printf("%s: skipping %s\n", file->filename, buf);
1474 #endif
1475                 fd = 0;
1476             } else {
1477                 fd = tcc_open(s1, buf1);
1478                 if (fd < 0)
1479                     continue;
1480             }
1481
1482             if (tok == TOK_INCLUDE_NEXT) {
1483                 tok = TOK_INCLUDE;
1484                 if (fd)
1485                     tcc_close();
1486                 continue;
1487             }
1488
1489             if (0 == fd)
1490                 goto include_done;
1491
1492 #ifdef INC_DEBUG
1493             printf("%s: including %s\n", file->filename, buf1);
1494 #endif
1495             /* update target deps */
1496             dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1497                     tcc_strdup(buf1));
1498            /* XXX: fix current line init */
1499            /* push current file in stack */
1500             *s1->include_stack_ptr++ = file->prev;
1501             file->inc_type = c;
1502             pstrcpy(file->inc_filename, sizeof(file->inc_filename), buf1);
1503             /* add include file debug info */
1504             if (s1->do_debug)
1505                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1506             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1507             ch = file->buf_ptr[0];
1508             goto the_end;
1509         }
1510         error("include file '%s' not found", buf);
1511 include_done:
1512         break;
1513     case TOK_IFNDEF:
1514         c = 1;
1515         goto do_ifdef;
1516     case TOK_IF:
1517         c = expr_preprocess();
1518         goto do_if;
1519     case TOK_IFDEF:
1520         c = 0;
1521     do_ifdef:
1522         next_nomacro();
1523         if (tok < TOK_IDENT)
1524             error("invalid argument for '#if%sdef'", c ? "n" : "");
1525         if (is_bof) {
1526             if (c) {
1527 #ifdef INC_DEBUG
1528                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1529 #endif
1530                 file->ifndef_macro = tok;
1531             }
1532         }
1533         c = (define_find(tok) != 0) ^ c;
1534     do_if:
1535         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1536             error("memory full");
1537         *s1->ifdef_stack_ptr++ = c;
1538         goto test_skip;
1539     case TOK_ELSE:
1540         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1541             error("#else without matching #if");
1542         if (s1->ifdef_stack_ptr[-1] & 2)
1543             error("#else after #else");
1544         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1545         goto test_else;
1546     case TOK_ELIF:
1547         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1548             error("#elif without matching #if");
1549         c = s1->ifdef_stack_ptr[-1];
1550         if (c > 1)
1551             error("#elif after #else");
1552         /* last #if/#elif expression was true: we skip */
1553         if (c == 1)
1554             goto skip;
1555         c = expr_preprocess();
1556         s1->ifdef_stack_ptr[-1] = c;
1557     test_else:
1558         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1559             file->ifndef_macro = 0;
1560     test_skip:
1561         if (!(c & 1)) {
1562         skip:
1563             preprocess_skip();
1564             is_bof = 0;
1565             goto redo;
1566         }
1567         break;
1568     case TOK_ENDIF:
1569         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1570             error("#endif without matching #if");
1571         s1->ifdef_stack_ptr--;
1572         /* '#ifndef macro' was at the start of file. Now we check if
1573            an '#endif' is exactly at the end of file */
1574         if (file->ifndef_macro &&
1575             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1576             file->ifndef_macro_saved = file->ifndef_macro;
1577             /* need to set to zero to avoid false matches if another
1578                #ifndef at middle of file */
1579             file->ifndef_macro = 0;
1580             while (tok != TOK_LINEFEED)
1581                 next_nomacro();
1582             tok_flags |= TOK_FLAG_ENDIF;
1583             goto the_end;
1584         }
1585         break;
1586     case TOK_LINE:
1587         next();
1588         if (tok != TOK_CINT)
1589             error("#line");
1590         file->line_num = tokc.i - 1; /* the line number will be incremented after */
1591         next();
1592         if (tok != TOK_LINEFEED) {
1593             if (tok != TOK_STR)
1594                 error("#line");
1595             pstrcpy(file->filename, sizeof(file->filename),
1596                     (char *)tokc.cstr->data);
1597         }
1598         break;
1599     case TOK_ERROR:
1600     case TOK_WARNING:
1601         c = tok;
1602         ch = file->buf_ptr[0];
1603         skip_spaces();
1604         q = buf;
1605         while (ch != '\n' && ch != CH_EOF) {
1606             if ((q - buf) < sizeof(buf) - 1)
1607                 *q++ = ch;
1608             if (ch == '\\') {
1609                 if (handle_stray_noerror() == 0)
1610                     --q;
1611             } else
1612                 inp();
1613         }
1614         *q = '\0';
1615         if (c == TOK_ERROR)
1616             error("#error %s", buf);
1617         else
1618             warning("#warning %s", buf);
1619         break;
1620     case TOK_PRAGMA:
1621         pragma_parse(s1);
1622         break;
1623     default:
1624         if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_PPNUM) {
1625             /* '!' is ignored to allow C scripts. numbers are ignored
1626                to emulate cpp behaviour */
1627         } else {
1628             if (!(saved_parse_flags & PARSE_FLAG_ASM_COMMENTS))
1629                 warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1630             else {
1631                 /* this is a gas line comment in an 'S' file. */
1632                 file->buf_ptr = parse_line_comment(file->buf_ptr);
1633                 goto the_end;
1634             }
1635         }
1636         break;
1637     }
1638     /* ignore other preprocess commands or #! for C scripts */
1639     while (tok != TOK_LINEFEED)
1640         next_nomacro();
1641  the_end:
1642     parse_flags = saved_parse_flags;
1643 }
1644
1645 /* evaluate escape codes in a string. */
1646 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1647 {
1648     int c, n;
1649     const uint8_t *p;
1650
1651     p = buf;
1652     for(;;) {
1653         c = *p;
1654         if (c == '\0')
1655             break;
1656         if (c == '\\') {
1657             p++;
1658             /* escape */
1659             c = *p;
1660             switch(c) {
1661             case '0': case '1': case '2': case '3':
1662             case '4': case '5': case '6': case '7':
1663                 /* at most three octal digits */
1664                 n = c - '0';
1665                 p++;
1666                 c = *p;
1667                 if (isoct(c)) {
1668                     n = n * 8 + c - '0';
1669                     p++;
1670                     c = *p;
1671                     if (isoct(c)) {
1672                         n = n * 8 + c - '0';
1673                         p++;
1674                     }
1675                 }
1676                 c = n;
1677                 goto add_char_nonext;
1678             case 'x':
1679             case 'u':
1680             case 'U':
1681                 p++;
1682                 n = 0;
1683                 for(;;) {
1684                     c = *p;
1685                     if (c >= 'a' && c <= 'f')
1686                         c = c - 'a' + 10;
1687                     else if (c >= 'A' && c <= 'F')
1688                         c = c - 'A' + 10;
1689                     else if (isnum(c))
1690                         c = c - '0';
1691                     else
1692                         break;
1693                     n = n * 16 + c;
1694                     p++;
1695                 }
1696                 c = n;
1697                 goto add_char_nonext;
1698             case 'a':
1699                 c = '\a';
1700                 break;
1701             case 'b':
1702                 c = '\b';
1703                 break;
1704             case 'f':
1705                 c = '\f';
1706                 break;
1707             case 'n':
1708                 c = '\n';
1709                 break;
1710             case 'r':
1711                 c = '\r';
1712                 break;
1713             case 't':
1714                 c = '\t';
1715                 break;
1716             case 'v':
1717                 c = '\v';
1718                 break;
1719             case 'e':
1720                 if (!gnu_ext)
1721                     goto invalid_escape;
1722                 c = 27;
1723                 break;
1724             case '\'':
1725             case '\"':
1726             case '\\':
1727             case '?':
1728                 break;
1729             default:
1730             invalid_escape:
1731                 if (c >= '!' && c <= '~')
1732                     warning("unknown escape sequence: \'\\%c\'", c);
1733                 else
1734                     warning("unknown escape sequence: \'\\x%x\'", c);
1735                 break;
1736             }
1737         }
1738         p++;
1739     add_char_nonext:
1740         if (!is_long)
1741             cstr_ccat(outstr, c);
1742         else
1743             cstr_wccat(outstr, c);
1744     }
1745     /* add a trailing '\0' */
1746     if (!is_long)
1747         cstr_ccat(outstr, '\0');
1748     else
1749         cstr_wccat(outstr, '\0');
1750 }
1751
1752 /* we use 64 bit numbers */
1753 #define BN_SIZE 2
1754
1755 /* bn = (bn << shift) | or_val */
1756 static void bn_lshift(unsigned int *bn, int shift, int or_val)
1757 {
1758     int i;
1759     unsigned int v;
1760     for(i=0;i<BN_SIZE;i++) {
1761         v = bn[i];
1762         bn[i] = (v << shift) | or_val;
1763         or_val = v >> (32 - shift);
1764     }
1765 }
1766
1767 static void bn_zero(unsigned int *bn)
1768 {
1769     int i;
1770     for(i=0;i<BN_SIZE;i++) {
1771         bn[i] = 0;
1772     }
1773 }
1774
1775 /* parse number in null terminated string 'p' and return it in the
1776    current token */
1777 static void parse_number(const char *p)
1778 {
1779     int b, t, shift, frac_bits, s, exp_val, ch;
1780     char *q;
1781     unsigned int bn[BN_SIZE];
1782     double d;
1783
1784     /* number */
1785     q = token_buf;
1786     ch = *p++;
1787     t = ch;
1788     ch = *p++;
1789     *q++ = t;
1790     b = 10;
1791     if (t == '.') {
1792         goto float_frac_parse;
1793     } else if (t == '0') {
1794         if (ch == 'x' || ch == 'X') {
1795             q--;
1796             ch = *p++;
1797             b = 16;
1798         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
1799             q--;
1800             ch = *p++;
1801             b = 2;
1802         }
1803     }
1804     /* parse all digits. cannot check octal numbers at this stage
1805        because of floating point constants */
1806     while (1) {
1807         if (ch >= 'a' && ch <= 'f')
1808             t = ch - 'a' + 10;
1809         else if (ch >= 'A' && ch <= 'F')
1810             t = ch - 'A' + 10;
1811         else if (isnum(ch))
1812             t = ch - '0';
1813         else
1814             break;
1815         if (t >= b)
1816             break;
1817         if (q >= token_buf + STRING_MAX_SIZE) {
1818         num_too_long:
1819             error("number too long");
1820         }
1821         *q++ = ch;
1822         ch = *p++;
1823     }
1824     if (ch == '.' ||
1825         ((ch == 'e' || ch == 'E') && b == 10) ||
1826         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
1827         if (b != 10) {
1828             /* NOTE: strtox should support that for hexa numbers, but
1829                non ISOC99 libcs do not support it, so we prefer to do
1830                it by hand */
1831             /* hexadecimal or binary floats */
1832             /* XXX: handle overflows */
1833             *q = '\0';
1834             if (b == 16)
1835                 shift = 4;
1836             else
1837                 shift = 2;
1838             bn_zero(bn);
1839             q = token_buf;
1840             while (1) {
1841                 t = *q++;
1842                 if (t == '\0') {
1843                     break;
1844                 } else if (t >= 'a') {
1845                     t = t - 'a' + 10;
1846                 } else if (t >= 'A') {
1847                     t = t - 'A' + 10;
1848                 } else {
1849                     t = t - '0';
1850                 }
1851                 bn_lshift(bn, shift, t);
1852             }
1853             frac_bits = 0;
1854             if (ch == '.') {
1855                 ch = *p++;
1856                 while (1) {
1857                     t = ch;
1858                     if (t >= 'a' && t <= 'f') {
1859                         t = t - 'a' + 10;
1860                     } else if (t >= 'A' && t <= 'F') {
1861                         t = t - 'A' + 10;
1862                     } else if (t >= '0' && t <= '9') {
1863                         t = t - '0';
1864                     } else {
1865                         break;
1866                     }
1867                     if (t >= b)
1868                         error("invalid digit");
1869                     bn_lshift(bn, shift, t);
1870                     frac_bits += shift;
1871                     ch = *p++;
1872                 }
1873             }
1874             if (ch != 'p' && ch != 'P')
1875                 expect("exponent");
1876             ch = *p++;
1877             s = 1;
1878             exp_val = 0;
1879             if (ch == '+') {
1880                 ch = *p++;
1881             } else if (ch == '-') {
1882                 s = -1;
1883                 ch = *p++;
1884             }
1885             if (ch < '0' || ch > '9')
1886                 expect("exponent digits");
1887             while (ch >= '0' && ch <= '9') {
1888                 exp_val = exp_val * 10 + ch - '0';
1889                 ch = *p++;
1890             }
1891             exp_val = exp_val * s;
1892
1893             /* now we can generate the number */
1894             /* XXX: should patch directly float number */
1895             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
1896             d = ldexp(d, exp_val - frac_bits);
1897             t = toup(ch);
1898             if (t == 'F') {
1899                 ch = *p++;
1900                 tok = TOK_CFLOAT;
1901                 /* float : should handle overflow */
1902                 tokc.f = (float)d;
1903             } else if (t == 'L') {
1904                 ch = *p++;
1905 #ifdef TCC_TARGET_PE
1906                 tok = TOK_CDOUBLE;
1907                 tokc.d = d;
1908 #else
1909                 tok = TOK_CLDOUBLE;
1910                 /* XXX: not large enough */
1911                 tokc.ld = (long double)d;
1912 #endif
1913             } else {
1914                 tok = TOK_CDOUBLE;
1915                 tokc.d = d;
1916             }
1917         } else {
1918             /* decimal floats */
1919             if (ch == '.') {
1920                 if (q >= token_buf + STRING_MAX_SIZE)
1921                     goto num_too_long;
1922                 *q++ = ch;
1923                 ch = *p++;
1924             float_frac_parse:
1925                 while (ch >= '0' && ch <= '9') {
1926                     if (q >= token_buf + STRING_MAX_SIZE)
1927                         goto num_too_long;
1928                     *q++ = ch;
1929                     ch = *p++;
1930                 }
1931             }
1932             if (ch == 'e' || ch == 'E') {
1933                 if (q >= token_buf + STRING_MAX_SIZE)
1934                     goto num_too_long;
1935                 *q++ = ch;
1936                 ch = *p++;
1937                 if (ch == '-' || ch == '+') {
1938                     if (q >= token_buf + STRING_MAX_SIZE)
1939                         goto num_too_long;
1940                     *q++ = ch;
1941                     ch = *p++;
1942                 }
1943                 if (ch < '0' || ch > '9')
1944                     expect("exponent digits");
1945                 while (ch >= '0' && ch <= '9') {
1946                     if (q >= token_buf + STRING_MAX_SIZE)
1947                         goto num_too_long;
1948                     *q++ = ch;
1949                     ch = *p++;
1950                 }
1951             }
1952             *q = '\0';
1953             t = toup(ch);
1954             errno = 0;
1955             if (t == 'F') {
1956                 ch = *p++;
1957                 tok = TOK_CFLOAT;
1958                 tokc.f = strtof(token_buf, NULL);
1959             } else if (t == 'L') {
1960                 ch = *p++;
1961 #ifdef TCC_TARGET_PE
1962                 tok = TOK_CDOUBLE;
1963                 tokc.d = strtod(token_buf, NULL);
1964 #else
1965                 tok = TOK_CLDOUBLE;
1966                 tokc.ld = strtold(token_buf, NULL);
1967 #endif
1968             } else {
1969                 tok = TOK_CDOUBLE;
1970                 tokc.d = strtod(token_buf, NULL);
1971             }
1972         }
1973     } else {
1974         unsigned long long n, n1;
1975         int lcount, ucount;
1976
1977         /* integer number */
1978         *q = '\0';
1979         q = token_buf;
1980         if (b == 10 && *q == '0') {
1981             b = 8;
1982             q++;
1983         }
1984         n = 0;
1985         while(1) {
1986             t = *q++;
1987             /* no need for checks except for base 10 / 8 errors */
1988             if (t == '\0') {
1989                 break;
1990             } else if (t >= 'a') {
1991                 t = t - 'a' + 10;
1992             } else if (t >= 'A') {
1993                 t = t - 'A' + 10;
1994             } else {
1995                 t = t - '0';
1996                 if (t >= b)
1997                     error("invalid digit");
1998             }
1999             n1 = n;
2000             n = n * b + t;
2001             /* detect overflow */
2002             /* XXX: this test is not reliable */
2003             if (n < n1)
2004                 error("integer constant overflow");
2005         }
2006
2007         /* XXX: not exactly ANSI compliant */
2008         if ((n & 0xffffffff00000000LL) != 0) {
2009             if ((n >> 63) != 0)
2010                 tok = TOK_CULLONG;
2011             else
2012                 tok = TOK_CLLONG;
2013         } else if (n > 0x7fffffff) {
2014             tok = TOK_CUINT;
2015         } else {
2016             tok = TOK_CINT;
2017         }
2018         lcount = 0;
2019         ucount = 0;
2020         for(;;) {
2021             t = toup(ch);
2022             if (t == 'L') {
2023                 if (lcount >= 2)
2024                     error("three 'l's in integer constant");
2025                 lcount++;
2026                 if (lcount == 2) {
2027                     if (tok == TOK_CINT)
2028                         tok = TOK_CLLONG;
2029                     else if (tok == TOK_CUINT)
2030                         tok = TOK_CULLONG;
2031                 }
2032                 ch = *p++;
2033             } else if (t == 'U') {
2034                 if (ucount >= 1)
2035                     error("two 'u's in integer constant");
2036                 ucount++;
2037                 if (tok == TOK_CINT)
2038                     tok = TOK_CUINT;
2039                 else if (tok == TOK_CLLONG)
2040                     tok = TOK_CULLONG;
2041                 ch = *p++;
2042             } else {
2043                 break;
2044             }
2045         }
2046         if (tok == TOK_CINT || tok == TOK_CUINT)
2047             tokc.ui = n;
2048         else
2049             tokc.ull = n;
2050     }
2051     if (ch)
2052         error("invalid number\n");
2053 }
2054
2055
2056 #define PARSE2(c1, tok1, c2, tok2)              \
2057     case c1:                                    \
2058         PEEKC(c, p);                            \
2059         if (c == c2) {                          \
2060             p++;                                \
2061             tok = tok2;                         \
2062         } else {                                \
2063             tok = tok1;                         \
2064         }                                       \
2065         break;
2066
2067 /* return next token without macro substitution */
2068 static inline void next_nomacro1(void)
2069 {
2070     int t, c, is_long;
2071     TokenSym *ts;
2072     uint8_t *p, *p1;
2073     unsigned int h;
2074
2075     p = file->buf_ptr;
2076  redo_no_start:
2077     c = *p;
2078     switch(c) {
2079     case ' ':
2080     case '\t':
2081         tok = c;
2082         p++;
2083         goto keep_tok_flags;
2084     case '\f':
2085     case '\v':
2086     case '\r':
2087         p++;
2088         goto redo_no_start;
2089     case '\\':
2090         /* first look if it is in fact an end of buffer */
2091         if (p >= file->buf_end) {
2092             file->buf_ptr = p;
2093             handle_eob();
2094             p = file->buf_ptr;
2095             if (p >= file->buf_end)
2096                 goto parse_eof;
2097             else
2098                 goto redo_no_start;
2099         } else {
2100             file->buf_ptr = p;
2101             ch = *p;
2102             handle_stray();
2103             p = file->buf_ptr;
2104             goto redo_no_start;
2105         }
2106     parse_eof:
2107         {
2108             TCCState *s1 = tcc_state;
2109             if ((parse_flags & PARSE_FLAG_LINEFEED)
2110                 && !(tok_flags & TOK_FLAG_EOF)) {
2111                 tok_flags |= TOK_FLAG_EOF;
2112                 tok = TOK_LINEFEED;
2113                 goto keep_tok_flags;
2114             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2115                 tok = TOK_EOF;
2116             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2117                 error("missing #endif");
2118             } else if (s1->include_stack_ptr == s1->include_stack) {
2119                 /* no include left : end of file. */
2120                 tok = TOK_EOF;
2121             } else {
2122                 tok_flags &= ~TOK_FLAG_EOF;
2123                 /* pop include file */
2124
2125                 /* test if previous '#endif' was after a #ifdef at
2126                    start of file */
2127                 if (tok_flags & TOK_FLAG_ENDIF) {
2128 #ifdef INC_DEBUG
2129                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2130 #endif
2131                     add_cached_include(s1, file->inc_type, file->inc_filename,
2132                                        file->ifndef_macro_saved);
2133                 }
2134
2135                 /* add end of include file debug info */
2136                 if (tcc_state->do_debug) {
2137                     put_stabd(N_EINCL, 0, 0);
2138                 }
2139                 /* pop include stack */
2140                 tcc_close();
2141                 s1->include_stack_ptr--;
2142                 p = file->buf_ptr;
2143                 goto redo_no_start;
2144             }
2145         }
2146         break;
2147
2148     case '\n':
2149         file->line_num++;
2150         tok_flags |= TOK_FLAG_BOL;
2151         p++;
2152 maybe_newline:
2153         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2154             goto redo_no_start;
2155         tok = TOK_LINEFEED;
2156         goto keep_tok_flags;
2157
2158     case '#':
2159         /* XXX: simplify */
2160         PEEKC(c, p);
2161         if ((tok_flags & TOK_FLAG_BOL) &&
2162             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2163             file->buf_ptr = p;
2164             preprocess(tok_flags & TOK_FLAG_BOF);
2165             p = file->buf_ptr;
2166             goto maybe_newline;
2167         } else {
2168             if (c == '#') {
2169                 p++;
2170                 tok = TOK_TWOSHARPS;
2171             } else {
2172                 if (parse_flags & PARSE_FLAG_ASM_COMMENTS) {
2173                     p = parse_line_comment(p - 1);
2174                     goto redo_no_start;
2175                 } else {
2176                     tok = '#';
2177                 }
2178             }
2179         }
2180         break;
2181
2182     case 'a': case 'b': case 'c': case 'd':
2183     case 'e': case 'f': case 'g': case 'h':
2184     case 'i': case 'j': case 'k': case 'l':
2185     case 'm': case 'n': case 'o': case 'p':
2186     case 'q': case 'r': case 's': case 't':
2187     case 'u': case 'v': case 'w': case 'x':
2188     case 'y': case 'z':
2189     case 'A': case 'B': case 'C': case 'D':
2190     case 'E': case 'F': case 'G': case 'H':
2191     case 'I': case 'J': case 'K':
2192     case 'M': case 'N': case 'O': case 'P':
2193     case 'Q': case 'R': case 'S': case 'T':
2194     case 'U': case 'V': case 'W': case 'X':
2195     case 'Y': case 'Z':
2196     case '_':
2197     parse_ident_fast:
2198         p1 = p;
2199         h = TOK_HASH_INIT;
2200         h = TOK_HASH_FUNC(h, c);
2201         p++;
2202         for(;;) {
2203             c = *p;
2204             if (!isidnum_table[c-CH_EOF])
2205                 break;
2206             h = TOK_HASH_FUNC(h, c);
2207             p++;
2208         }
2209         if (c != '\\') {
2210             TokenSym **pts;
2211             int len;
2212
2213             /* fast case : no stray found, so we have the full token
2214                and we have already hashed it */
2215             len = p - p1;
2216             h &= (TOK_HASH_SIZE - 1);
2217             pts = &hash_ident[h];
2218             for(;;) {
2219                 ts = *pts;
2220                 if (!ts)
2221                     break;
2222                 if (ts->len == len && !memcmp(ts->str, p1, len))
2223                     goto token_found;
2224                 pts = &(ts->hash_next);
2225             }
2226             ts = tok_alloc_new(pts, p1, len);
2227         token_found: ;
2228         } else {
2229             /* slower case */
2230             cstr_reset(&tokcstr);
2231
2232             while (p1 < p) {
2233                 cstr_ccat(&tokcstr, *p1);
2234                 p1++;
2235             }
2236             p--;
2237             PEEKC(c, p);
2238         parse_ident_slow:
2239             while (isidnum_table[c-CH_EOF]) {
2240                 cstr_ccat(&tokcstr, c);
2241                 PEEKC(c, p);
2242             }
2243             ts = tok_alloc(tokcstr.data, tokcstr.size);
2244         }
2245         tok = ts->tok;
2246         break;
2247     case 'L':
2248         t = p[1];
2249         if (t != '\\' && t != '\'' && t != '\"') {
2250             /* fast case */
2251             goto parse_ident_fast;
2252         } else {
2253             PEEKC(c, p);
2254             if (c == '\'' || c == '\"') {
2255                 is_long = 1;
2256                 goto str_const;
2257             } else {
2258                 cstr_reset(&tokcstr);
2259                 cstr_ccat(&tokcstr, 'L');
2260                 goto parse_ident_slow;
2261             }
2262         }
2263         break;
2264     case '0': case '1': case '2': case '3':
2265     case '4': case '5': case '6': case '7':
2266     case '8': case '9':
2267
2268         cstr_reset(&tokcstr);
2269         /* after the first digit, accept digits, alpha, '.' or sign if
2270            prefixed by 'eEpP' */
2271     parse_num:
2272         for(;;) {
2273             t = c;
2274             cstr_ccat(&tokcstr, c);
2275             PEEKC(c, p);
2276             if (!(isnum(c) || isid(c) || c == '.' ||
2277                   ((c == '+' || c == '-') &&
2278                    (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2279                 break;
2280         }
2281         /* We add a trailing '\0' to ease parsing */
2282         cstr_ccat(&tokcstr, '\0');
2283         tokc.cstr = &tokcstr;
2284         tok = TOK_PPNUM;
2285         break;
2286     case '.':
2287         /* special dot handling because it can also start a number */
2288         PEEKC(c, p);
2289         if (isnum(c)) {
2290             cstr_reset(&tokcstr);
2291             cstr_ccat(&tokcstr, '.');
2292             goto parse_num;
2293         } else if (c == '.') {
2294             PEEKC(c, p);
2295             if (c != '.')
2296                 expect("'.'");
2297             PEEKC(c, p);
2298             tok = TOK_DOTS;
2299         } else {
2300             tok = '.';
2301         }
2302         break;
2303     case '\'':
2304     case '\"':
2305         is_long = 0;
2306     str_const:
2307         {
2308             CString str;
2309             int sep;
2310
2311             sep = c;
2312
2313             /* parse the string */
2314             cstr_new(&str);
2315             p = parse_pp_string(p, sep, &str);
2316             cstr_ccat(&str, '\0');
2317
2318             /* eval the escape (should be done as TOK_PPNUM) */
2319             cstr_reset(&tokcstr);
2320             parse_escape_string(&tokcstr, str.data, is_long);
2321             cstr_free(&str);
2322
2323             if (sep == '\'') {
2324                 int char_size;
2325                 /* XXX: make it portable */
2326                 if (!is_long)
2327                     char_size = 1;
2328                 else
2329                     char_size = sizeof(nwchar_t);
2330                 if (tokcstr.size <= char_size)
2331                     error("empty character constant");
2332                 if (tokcstr.size > 2 * char_size)
2333                     warning("multi-character character constant");
2334                 if (!is_long) {
2335                     tokc.i = *(int8_t *)tokcstr.data;
2336                     tok = TOK_CCHAR;
2337                 } else {
2338                     tokc.i = *(nwchar_t *)tokcstr.data;
2339                     tok = TOK_LCHAR;
2340                 }
2341             } else {
2342                 tokc.cstr = &tokcstr;
2343                 if (!is_long)
2344                     tok = TOK_STR;
2345                 else
2346                     tok = TOK_LSTR;
2347             }
2348         }
2349         break;
2350
2351     case '<':
2352         PEEKC(c, p);
2353         if (c == '=') {
2354             p++;
2355             tok = TOK_LE;
2356         } else if (c == '<') {
2357             PEEKC(c, p);
2358             if (c == '=') {
2359                 p++;
2360                 tok = TOK_A_SHL;
2361             } else {
2362                 tok = TOK_SHL;
2363             }
2364         } else {
2365             tok = TOK_LT;
2366         }
2367         break;
2368
2369     case '>':
2370         PEEKC(c, p);
2371         if (c == '=') {
2372             p++;
2373             tok = TOK_GE;
2374         } else if (c == '>') {
2375             PEEKC(c, p);
2376             if (c == '=') {
2377                 p++;
2378                 tok = TOK_A_SAR;
2379             } else {
2380                 tok = TOK_SAR;
2381             }
2382         } else {
2383             tok = TOK_GT;
2384         }
2385         break;
2386
2387     case '&':
2388         PEEKC(c, p);
2389         if (c == '&') {
2390             p++;
2391             tok = TOK_LAND;
2392         } else if (c == '=') {
2393             p++;
2394             tok = TOK_A_AND;
2395         } else {
2396             tok = '&';
2397         }
2398         break;
2399
2400     case '|':
2401         PEEKC(c, p);
2402         if (c == '|') {
2403             p++;
2404             tok = TOK_LOR;
2405         } else if (c == '=') {
2406             p++;
2407             tok = TOK_A_OR;
2408         } else {
2409             tok = '|';
2410         }
2411         break;
2412
2413     case '+':
2414         PEEKC(c, p);
2415         if (c == '+') {
2416             p++;
2417             tok = TOK_INC;
2418         } else if (c == '=') {
2419             p++;
2420             tok = TOK_A_ADD;
2421         } else {
2422             tok = '+';
2423         }
2424         break;
2425
2426     case '-':
2427         PEEKC(c, p);
2428         if (c == '-') {
2429             p++;
2430             tok = TOK_DEC;
2431         } else if (c == '=') {
2432             p++;
2433             tok = TOK_A_SUB;
2434         } else if (c == '>') {
2435             p++;
2436             tok = TOK_ARROW;
2437         } else {
2438             tok = '-';
2439         }
2440         break;
2441
2442     PARSE2('!', '!', '=', TOK_NE)
2443     PARSE2('=', '=', '=', TOK_EQ)
2444     PARSE2('*', '*', '=', TOK_A_MUL)
2445     PARSE2('%', '%', '=', TOK_A_MOD)
2446     PARSE2('^', '^', '=', TOK_A_XOR)
2447
2448         /* comments or operator */
2449     case '/':
2450         PEEKC(c, p);
2451         if (c == '*') {
2452             p = parse_comment(p);
2453             /* comments replaced by a blank */
2454             tok = ' ';
2455             goto keep_tok_flags;
2456         } else if (c == '/') {
2457             p = parse_line_comment(p);
2458             tok = ' ';
2459             goto keep_tok_flags;
2460         } else if (c == '=') {
2461             p++;
2462             tok = TOK_A_DIV;
2463         } else {
2464             tok = '/';
2465         }
2466         break;
2467
2468         /* simple tokens */
2469     case '(':
2470     case ')':
2471     case '[':
2472     case ']':
2473     case '{':
2474     case '}':
2475     case ',':
2476     case ';':
2477     case ':':
2478     case '?':
2479     case '~':
2480     case '$': /* only used in assembler */
2481     case '@': /* dito */
2482         tok = c;
2483         p++;
2484         break;
2485     default:
2486         error("unrecognized character \\x%02x", c);
2487         break;
2488     }
2489     tok_flags = 0;
2490 keep_tok_flags:
2491     file->buf_ptr = p;
2492 #if defined(PARSE_DEBUG)
2493     printf("token = %s\n", get_tok_str(tok, &tokc));
2494 #endif
2495 }
2496
2497 /* return next token without macro substitution. Can read input from
2498    macro_ptr buffer */
2499 static void next_nomacro_spc(void)
2500 {
2501     if (macro_ptr) {
2502     redo:
2503         tok = *macro_ptr;
2504         if (tok) {
2505             TOK_GET(&tok, &macro_ptr, &tokc);
2506             if (tok == TOK_LINENUM) {
2507                 file->line_num = tokc.i;
2508                 goto redo;
2509             }
2510         }
2511     } else {
2512         next_nomacro1();
2513     }
2514 }
2515
2516 ST_FUNC void next_nomacro(void)
2517 {
2518     do {
2519         next_nomacro_spc();
2520     } while (is_space(tok));
2521 }
2522
2523 /* substitute args in macro_str and return allocated string */
2524 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2525 {
2526     int last_tok, t, spc;
2527     const int *st;
2528     Sym *s;
2529     CValue cval;
2530     TokenString str;
2531     CString cstr;
2532
2533     tok_str_new(&str);
2534     last_tok = 0;
2535     while(1) {
2536         TOK_GET(&t, &macro_str, &cval);
2537         if (!t)
2538             break;
2539         if (t == '#') {
2540             /* stringize */
2541             TOK_GET(&t, &macro_str, &cval);
2542             if (!t)
2543                 break;
2544             s = sym_find2(args, t);
2545             if (s) {
2546                 cstr_new(&cstr);
2547                 st = s->d;
2548                 spc = 0;
2549                 while (*st) {
2550                     TOK_GET(&t, &st, &cval);
2551                     if (!check_space(t, &spc))
2552                         cstr_cat(&cstr, get_tok_str(t, &cval));
2553                 }
2554                 cstr.size -= spc;
2555                 cstr_ccat(&cstr, '\0');
2556 #ifdef PP_DEBUG
2557                 printf("stringize: %s\n", (char *)cstr.data);
2558 #endif
2559                 /* add string */
2560                 cval.cstr = &cstr;
2561                 tok_str_add2(&str, TOK_STR, &cval);
2562                 cstr_free(&cstr);
2563             } else {
2564                 tok_str_add2(&str, t, &cval);
2565             }
2566         } else if (t >= TOK_IDENT) {
2567             s = sym_find2(args, t);
2568             if (s) {
2569                 st = s->d;
2570                 /* if '##' is present before or after, no arg substitution */
2571                 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2572                     /* special case for var arg macros : ## eats the
2573                        ',' if empty VA_ARGS variable. */
2574                     /* XXX: test of the ',' is not 100%
2575                        reliable. should fix it to avoid security
2576                        problems */
2577                     if (gnu_ext && s->type.t &&
2578                         last_tok == TOK_TWOSHARPS &&
2579                         str.len >= 2 && str.str[str.len - 2] == ',') {
2580                         if (*st == 0) {
2581                             /* suppress ',' '##' */
2582                             str.len -= 2;
2583                         } else {
2584                             /* suppress '##' and add variable */
2585                             str.len--;
2586                             goto add_var;
2587                         }
2588                     } else {
2589                         int t1;
2590                     add_var:
2591                         for(;;) {
2592                             TOK_GET(&t1, &st, &cval);
2593                             if (!t1)
2594                                 break;
2595                             tok_str_add2(&str, t1, &cval);
2596                         }
2597                     }
2598                 } else {
2599                     /* NOTE: the stream cannot be read when macro
2600                        substituing an argument */
2601                     macro_subst(&str, nested_list, st, NULL);
2602                 }
2603             } else {
2604                 tok_str_add(&str, t);
2605             }
2606         } else {
2607             tok_str_add2(&str, t, &cval);
2608         }
2609         last_tok = t;
2610     }
2611     tok_str_add(&str, 0);
2612     return str.str;
2613 }
2614
2615 static char const ab_month_name[12][4] =
2616 {
2617     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2618     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2619 };
2620
2621 /* do macro substitution of current token with macro 's' and add
2622    result to (tok_str,tok_len). 'nested_list' is the list of all
2623    macros we got inside to avoid recursing. Return non zero if no
2624    substitution needs to be done */
2625 static int macro_subst_tok(TokenString *tok_str,
2626                            Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2627 {
2628     Sym *args, *sa, *sa1;
2629     int mstr_allocated, parlevel, *mstr, t, t1, spc;
2630     const int *p;
2631     TokenString str;
2632     char *cstrval;
2633     CValue cval;
2634     CString cstr;
2635     char buf[32];
2636
2637     /* if symbol is a macro, prepare substitution */
2638     /* special macros */
2639     if (tok == TOK___LINE__) {
2640         snprintf(buf, sizeof(buf), "%d", file->line_num);
2641         cstrval = buf;
2642         t1 = TOK_PPNUM;
2643         goto add_cstr1;
2644     } else if (tok == TOK___FILE__) {
2645         cstrval = file->filename;
2646         goto add_cstr;
2647     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2648         time_t ti;
2649         struct tm *tm;
2650
2651         time(&ti);
2652         tm = localtime(&ti);
2653         if (tok == TOK___DATE__) {
2654             snprintf(buf, sizeof(buf), "%s %2d %d",
2655                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2656         } else {
2657             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2658                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2659         }
2660         cstrval = buf;
2661     add_cstr:
2662         t1 = TOK_STR;
2663     add_cstr1:
2664         cstr_new(&cstr);
2665         cstr_cat(&cstr, cstrval);
2666         cstr_ccat(&cstr, '\0');
2667         cval.cstr = &cstr;
2668         tok_str_add2(tok_str, t1, &cval);
2669         cstr_free(&cstr);
2670     } else {
2671         mstr = s->d;
2672         mstr_allocated = 0;
2673         if (s->type.t == MACRO_FUNC) {
2674             /* NOTE: we do not use next_nomacro to avoid eating the
2675                next token. XXX: find better solution */
2676         redo:
2677             if (macro_ptr) {
2678                 p = macro_ptr;
2679                 while (is_space(t = *p) || TOK_LINEFEED == t)
2680                     ++p;
2681                 if (t == 0 && can_read_stream) {
2682                     /* end of macro stream: we must look at the token
2683                        after in the file */
2684                     struct macro_level *ml = *can_read_stream;
2685                     macro_ptr = NULL;
2686                     if (ml)
2687                     {
2688                         macro_ptr = ml->p;
2689                         ml->p = NULL;
2690                         *can_read_stream = ml -> prev;
2691                     }
2692                     goto redo;
2693                 }
2694             } else {
2695                 /* XXX: incorrect with comments */
2696                 ch = file->buf_ptr[0];
2697                 while (is_space(ch) || ch == '\n')
2698                     cinp();
2699                 t = ch;
2700             }
2701             if (t != '(') /* no macro subst */
2702                 return -1;
2703
2704             /* argument macro */
2705             next_nomacro();
2706             next_nomacro();
2707             args = NULL;
2708             sa = s->next;
2709             /* NOTE: empty args are allowed, except if no args */
2710             for(;;) {
2711                 /* handle '()' case */
2712                 if (!args && !sa && tok == ')')
2713                     break;
2714                 if (!sa)
2715                     error("macro '%s' used with too many args",
2716                           get_tok_str(s->v, 0));
2717                 tok_str_new(&str);
2718                 parlevel = spc = 0;
2719                 /* NOTE: non zero sa->t indicates VA_ARGS */
2720                 while ((parlevel > 0 ||
2721                         (tok != ')' &&
2722                          (tok != ',' || sa->type.t))) &&
2723                        tok != -1) {
2724                     if (tok == '(')
2725                         parlevel++;
2726                     else if (tok == ')')
2727                         parlevel--;
2728                     if (tok == TOK_LINEFEED)
2729                         tok = ' ';
2730                     if (!check_space(tok, &spc))
2731                         tok_str_add2(&str, tok, &tokc);
2732                     next_nomacro_spc();
2733                 }
2734                 str.len -= spc;
2735                 tok_str_add(&str, 0);
2736                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
2737                 sa1->d = str.str;
2738                 sa = sa->next;
2739                 if (tok == ')') {
2740                     /* special case for gcc var args: add an empty
2741                        var arg argument if it is omitted */
2742                     if (sa && sa->type.t && gnu_ext)
2743                         continue;
2744                     else
2745                         break;
2746                 }
2747                 if (tok != ',')
2748                     expect(",");
2749                 next_nomacro();
2750             }
2751             if (sa) {
2752                 error("macro '%s' used with too few args",
2753                       get_tok_str(s->v, 0));
2754             }
2755
2756             /* now subst each arg */
2757             mstr = macro_arg_subst(nested_list, mstr, args);
2758             /* free memory */
2759             sa = args;
2760             while (sa) {
2761                 sa1 = sa->prev;
2762                 tok_str_free(sa->d);
2763                 sym_free(sa);
2764                 sa = sa1;
2765             }
2766             mstr_allocated = 1;
2767         }
2768         sym_push2(nested_list, s->v, 0, 0);
2769         macro_subst(tok_str, nested_list, mstr, can_read_stream);
2770         /* pop nested defined symbol */
2771         sa1 = *nested_list;
2772         *nested_list = sa1->prev;
2773         sym_free(sa1);
2774         if (mstr_allocated)
2775             tok_str_free(mstr);
2776     }
2777     return 0;
2778 }
2779
2780 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2781    return the resulting string (which must be freed). */
2782 static inline int *macro_twosharps(const int *macro_str)
2783 {
2784     const int *ptr;
2785     int t;
2786     CValue cval;
2787     TokenString macro_str1;
2788     CString cstr;
2789     int n;
2790
2791     /* we search the first '##' */
2792     for(ptr = macro_str;;) {
2793         TOK_GET(&t, &ptr, &cval);
2794         if (t == TOK_TWOSHARPS)
2795             break;
2796         /* nothing more to do if end of string */
2797         if (t == 0)
2798             return NULL;
2799     }
2800
2801     /* we saw '##', so we need more processing to handle it */
2802     tok_str_new(&macro_str1);
2803     for(ptr = macro_str;;) {
2804         TOK_GET(&tok, &ptr, &tokc);
2805         if (tok == 0)
2806             break;
2807         if (tok == TOK_TWOSHARPS)
2808             continue;
2809         while (*ptr == TOK_TWOSHARPS) {
2810             do { t = *++ptr; } while (t == TOK_NOSUBST);
2811
2812             if (t && t != TOK_TWOSHARPS) {
2813                 TOK_GET(&t, &ptr, &cval);
2814
2815                 /* We concatenate the two tokens */
2816                 cstr_new(&cstr);
2817                 cstr_cat(&cstr, get_tok_str(tok, &tokc));
2818                 n = cstr.size;
2819                 cstr_cat(&cstr, get_tok_str(t, &cval));
2820                 cstr_ccat(&cstr, '\0');
2821
2822                 tcc_open_bf(tcc_state, "<paste>", cstr.size);
2823                 memcpy(file->buffer, cstr.data, cstr.size);
2824                 for (;;) {
2825                     next_nomacro1();
2826                     if (0 == *file->buf_ptr)
2827                         break;
2828                     tok_str_add2(&macro_str1, tok, &tokc);
2829                     warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
2830                         n, cstr.data, (char*)cstr.data + n);
2831                 }
2832                 tcc_close();
2833                 cstr_reset(&cstr);
2834             }
2835         }
2836         tok_str_add2(&macro_str1, tok, &tokc);
2837     }
2838     tok_str_add(&macro_str1, 0);
2839     return macro_str1.str;
2840 }
2841
2842
2843 /* do macro substitution of macro_str and add result to
2844    (tok_str,tok_len). 'nested_list' is the list of all macros we got
2845    inside to avoid recursing. */
2846 static void macro_subst(TokenString *tok_str, Sym **nested_list,
2847                         const int *macro_str, struct macro_level ** can_read_stream)
2848 {
2849     Sym *s;
2850     int *macro_str1;
2851     const int *ptr;
2852     int t, ret, spc;
2853     CValue cval;
2854     struct macro_level ml;
2855     int force_blank;
2856
2857     /* first scan for '##' operator handling */
2858     ptr = macro_str;
2859     macro_str1 = macro_twosharps(ptr);
2860
2861     if (macro_str1)
2862         ptr = macro_str1;
2863     spc = 0;
2864     force_blank = 0;
2865
2866     while (1) {
2867         /* NOTE: ptr == NULL can only happen if tokens are read from
2868            file stream due to a macro function call */
2869         if (ptr == NULL)
2870             break;
2871         TOK_GET(&t, &ptr, &cval);
2872         if (t == 0)
2873             break;
2874         if (t == TOK_NOSUBST) {
2875             /* following token has already been subst'd. just copy it on */
2876             tok_str_add2(tok_str, TOK_NOSUBST, NULL);
2877             TOK_GET(&t, &ptr, &cval);
2878             goto no_subst;
2879         }
2880         s = define_find(t);
2881         if (s != NULL) {
2882             /* if nested substitution, do nothing */
2883             if (sym_find2(*nested_list, t)) {
2884                 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
2885                 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
2886                 goto no_subst;
2887             }
2888             ml.p = macro_ptr;
2889             if (can_read_stream)
2890                 ml.prev = *can_read_stream, *can_read_stream = &ml;
2891             macro_ptr = (int *)ptr;
2892             tok = t;
2893             ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
2894             ptr = (int *)macro_ptr;
2895             macro_ptr = ml.p;
2896             if (can_read_stream && *can_read_stream == &ml)
2897                 *can_read_stream = ml.prev;
2898             if (ret != 0)
2899                 goto no_subst;
2900             if (parse_flags & PARSE_FLAG_SPACES)
2901                 force_blank = 1;
2902         } else {
2903         no_subst:
2904             if (force_blank) {
2905                 tok_str_add(tok_str, ' ');
2906                 spc = 1;
2907                 force_blank = 0;
2908             }
2909             if (!check_space(t, &spc))
2910                 tok_str_add2(tok_str, t, &cval);
2911         }
2912     }
2913     if (macro_str1)
2914         tok_str_free(macro_str1);
2915 }
2916
2917 /* return next token with macro substitution */
2918 ST_FUNC void next(void)
2919 {
2920     Sym *nested_list, *s;
2921     TokenString str;
2922     struct macro_level *ml;
2923
2924  redo:
2925     if (parse_flags & PARSE_FLAG_SPACES)
2926         next_nomacro_spc();
2927     else
2928         next_nomacro();
2929     if (!macro_ptr) {
2930         /* if not reading from macro substituted string, then try
2931            to substitute macros */
2932         if (tok >= TOK_IDENT &&
2933             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2934             s = define_find(tok);
2935             if (s) {
2936                 /* we have a macro: we try to substitute */
2937                 tok_str_new(&str);
2938                 nested_list = NULL;
2939                 ml = NULL;
2940                 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
2941                     /* substitution done, NOTE: maybe empty */
2942                     tok_str_add(&str, 0);
2943                     macro_ptr = str.str;
2944                     macro_ptr_allocated = str.str;
2945                     goto redo;
2946                 }
2947             }
2948         }
2949     } else {
2950         if (tok == 0) {
2951             /* end of macro or end of unget buffer */
2952             if (unget_buffer_enabled) {
2953                 macro_ptr = unget_saved_macro_ptr;
2954                 unget_buffer_enabled = 0;
2955             } else {
2956                 /* end of macro string: free it */
2957                 tok_str_free(macro_ptr_allocated);
2958                 macro_ptr_allocated = NULL;
2959                 macro_ptr = NULL;
2960             }
2961             goto redo;
2962         } else if (tok == TOK_NOSUBST) {
2963             /* discard preprocessor's nosubst markers */
2964             goto redo;
2965         }
2966     }
2967
2968     /* convert preprocessor tokens into C tokens */
2969     if (tok == TOK_PPNUM &&
2970         (parse_flags & PARSE_FLAG_TOK_NUM)) {
2971         parse_number((char *)tokc.cstr->data);
2972     }
2973 }
2974
2975 /* push back current token and set current token to 'last_tok'. Only
2976    identifier case handled for labels. */
2977 ST_INLN void unget_tok(int last_tok)
2978 {
2979     int i, n;
2980     int *q;
2981     unget_saved_macro_ptr = macro_ptr;
2982     unget_buffer_enabled = 1;
2983     q = unget_saved_buffer;
2984     macro_ptr = q;
2985     *q++ = tok;
2986     n = tok_ext_size(tok) - 1;
2987     for(i=0;i<n;i++)
2988         *q++ = tokc.tab[i];
2989     *q = 0; /* end of token string */
2990     tok = last_tok;
2991 }
2992
2993
2994 /* better than nothing, but needs extension to handle '-E' option
2995    correctly too */
2996 ST_FUNC void preprocess_init(TCCState *s1)
2997 {
2998     s1->include_stack_ptr = s1->include_stack;
2999     /* XXX: move that before to avoid having to initialize
3000        file->ifdef_stack_ptr ? */
3001     s1->ifdef_stack_ptr = s1->ifdef_stack;
3002     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3003
3004     /* XXX: not ANSI compliant: bound checking says error */
3005     vtop = vstack - 1;
3006     s1->pack_stack[0] = 0;
3007     s1->pack_stack_ptr = s1->pack_stack;
3008 }
3009
3010 ST_FUNC void preprocess_new()
3011 {
3012     int i, c;
3013     const char *p, *r;
3014     TokenSym *ts;
3015
3016     /* init isid table */
3017     for(i=CH_EOF;i<256;i++)
3018         isidnum_table[i-CH_EOF] = isid(i) || isnum(i);
3019
3020     /* add all tokens */
3021     table_ident = NULL;
3022     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3023
3024     tok_ident = TOK_IDENT;
3025     p = tcc_keywords;
3026     while (*p) {
3027         r = p;
3028         for(;;) {
3029             c = *r++;
3030             if (c == '\0')
3031                 break;
3032         }
3033         ts = tok_alloc(p, r - p - 1);
3034         p = r;
3035     }
3036 }
3037
3038 /* Preprocess the current file */
3039 ST_FUNC int tcc_preprocess(TCCState *s1)
3040 {
3041     Sym *define_start;
3042
3043     BufferedFile *file_ref, **iptr, **iptr_new;
3044     int token_seen, line_ref, d;
3045     const char *s;
3046
3047     preprocess_init(s1);
3048     define_start = define_stack;
3049     ch = file->buf_ptr[0];
3050     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3051     parse_flags = PARSE_FLAG_ASM_COMMENTS | PARSE_FLAG_PREPROCESS |
3052         PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
3053     token_seen = 0;
3054     line_ref = 0;
3055     file_ref = NULL;
3056     iptr = s1->include_stack_ptr;
3057
3058     for (;;) {
3059         next();
3060         if (tok == TOK_EOF) {
3061             break;
3062         } else if (file != file_ref) {
3063             goto print_line;
3064         } else if (tok == TOK_LINEFEED) {
3065             if (!token_seen)
3066                 continue;
3067             ++line_ref;
3068             token_seen = 0;
3069         } else if (!token_seen) {
3070             d = file->line_num - line_ref;
3071             if (file != file_ref || d < 0 || d >= 8) {
3072 print_line:
3073                 iptr_new = s1->include_stack_ptr;
3074                 s = iptr_new > iptr ? " 1"
3075                   : iptr_new < iptr ? " 2"
3076                   : iptr_new > s1->include_stack ? " 3"
3077                   : ""
3078                   ;
3079                 iptr = iptr_new;
3080                 fprintf(s1->outfile, "# %d \"%s\"%s\n", file->line_num, file->filename, s);
3081             } else {
3082                 while (d)
3083                     fputs("\n", s1->outfile), --d;
3084             }
3085             line_ref = (file_ref = file)->line_num;
3086             token_seen = tok != TOK_LINEFEED;
3087             if (!token_seen)
3088                 continue;
3089         }
3090         fputs(get_tok_str(tok, &tokc), s1->outfile);
3091     }
3092     free_defines(define_start);
3093     return 0;
3094 }