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             }
 750             break;
 751 _default:
 752         default:
 753             p++;
 754             break;
 755         }
 756         start_of_line = 0;
 757     }
 758  the_end: ;
 759     file->buf_ptr = p;
 760 }
 761
 762 /* ParseState handling */
 763
 764 /* XXX: currently, no include file info is stored. Thus, we cannot display
 765    accurate messages if the function or data definition spans multiple
 766    files */
 767
 768 /* save current parse state in 's' */
 769 ST_FUNC void save_parse_state(ParseState *s)
 770 {
 771     s->line_num = file->line_num;
 772     s->macro_ptr = macro_ptr;
 773     s->tok = tok;
 774     s->tokc = tokc;
 775 }
 776
 777 /* restore parse state from 's' */
 778 ST_FUNC void restore_parse_state(ParseState *s)
 779 {
 780     file->line_num = s->line_num;
 781     macro_ptr = s->macro_ptr;
 782     tok = s->tok;
 783     tokc = s->tokc;
 784 }
 785
 786 /* return the number of additional 'ints' necessary to store the
 787    token */
 788 static inline int tok_ext_size(int t)
 789 {
 790     switch(t) {
 791         /* 4 bytes */
 792     case TOK_CINT:
 793     case TOK_CUINT:
 794     case TOK_CCHAR:
 795     case TOK_LCHAR:
 796     case TOK_CFLOAT:
 797     case TOK_LINENUM:
 798         return 1;
 799     case TOK_STR:
 800     case TOK_LSTR:
 801     case TOK_PPNUM:
 802         error("unsupported token");
 803         return 1;
 804     case TOK_CDOUBLE:
 805     case TOK_CLLONG:
 806     case TOK_CULLONG:
 807         return 2;
 808     case TOK_CLDOUBLE:
 809         return LDOUBLE_SIZE / 4;
 810     default:
 811         return 0;
 812     }
 813 }
 814
 815 /* token string handling */
 816
 817 ST_INLN void tok_str_new(TokenString *s)
 818 {
 819     s->str = NULL;
 820     s->len = 0;
 821     s->allocated_len = 0;
 822     s->last_line_num = -1;
 823 }
 824
 825 ST_FUNC void tok_str_free(int *str)
 826 {
 827     tcc_free(str);
 828 }
 829
 830 static int *tok_str_realloc(TokenString *s)
 831 {
 832     int *str, len;
 833
 834     if (s->allocated_len == 0) {
 835         len = 8;
 836     } else {
 837         len = s->allocated_len * 2;
 838     }
 839     str = tcc_realloc(s->str, len * sizeof(int));
 840     if (!str)
 841         error("memory full");
 842     s->allocated_len = len;
 843     s->str = str;
 844     return str;
 845 }
 846
 847 ST_FUNC void tok_str_add(TokenString *s, int t)
 848 {
 849     int len, *str;
 850
 851     len = s->len;
 852     str = s->str;
 853     if (len >= s->allocated_len)
 854         str = tok_str_realloc(s);
 855     str[len++] = t;
 856     s->len = len;
 857 }
 858
 859 static void tok_str_add2(TokenString *s, int t, CValue *cv)
 860 {
 861     int len, *str;
 862
 863     len = s->len;
 864     str = s->str;
 865
 866     /* allocate space for worst case */
 867     if (len + TOK_MAX_SIZE > s->allocated_len)
 868         str = tok_str_realloc(s);
 869     str[len++] = t;
 870     switch(t) {
 871     case TOK_CINT:
 872     case TOK_CUINT:
 873     case TOK_CCHAR:
 874     case TOK_LCHAR:
 875     case TOK_CFLOAT:
 876     case TOK_LINENUM:
 877         str[len++] = cv->tab[0];
 878         break;
 879     case TOK_PPNUM:
 880     case TOK_STR:
 881     case TOK_LSTR:
 882         {
 883             int nb_words;
 884             CString *cstr;
 885
 886             nb_words = (sizeof(CString) + cv->cstr->size + 3) >> 2;
 887             while ((len + nb_words) > s->allocated_len)
 888                 str = tok_str_realloc(s);
 889             cstr = (CString *)(str + len);
 890             cstr->data = NULL;
 891             cstr->size = cv->cstr->size;
 892             cstr->data_allocated = NULL;
 893             cstr->size_allocated = cstr->size;
 894             memcpy((char *)cstr + sizeof(CString),
 895                    cv->cstr->data, cstr->size);
 896             len += nb_words;
 897         }
 898         break;
 899     case TOK_CDOUBLE:
 900     case TOK_CLLONG:
 901     case TOK_CULLONG:
 902 #if LDOUBLE_SIZE == 8
 903     case TOK_CLDOUBLE:
 904 #endif
 905         str[len++] = cv->tab[0];
 906         str[len++] = cv->tab[1];
 907         break;
 908 #if LDOUBLE_SIZE == 12
 909     case TOK_CLDOUBLE:
 910         str[len++] = cv->tab[0];
 911         str[len++] = cv->tab[1];
 912         str[len++] = cv->tab[2];
 913 #elif LDOUBLE_SIZE == 16
 914     case TOK_CLDOUBLE:
 915         str[len++] = cv->tab[0];
 916         str[len++] = cv->tab[1];
 917         str[len++] = cv->tab[2];
 918         str[len++] = cv->tab[3];
 919 #elif LDOUBLE_SIZE != 8
 920 #error add long double size support
 921 #endif
 922         break;
 923     default:
 924         break;
 925     }
 926     s->len = len;
 927 }
 928
 929 /* add the current parse token in token string 's' */
 930 ST_FUNC void tok_str_add_tok(TokenString *s)
 931 {
 932     CValue cval;
 933
 934     /* save line number info */
 935     if (file->line_num != s->last_line_num) {
 936         s->last_line_num = file->line_num;
 937         cval.i = s->last_line_num;
 938         tok_str_add2(s, TOK_LINENUM, &cval);
 939     }
 940     tok_str_add2(s, tok, &tokc);
 941 }
 942
 943 /* get a token from an integer array and increment pointer
 944    accordingly. we code it as a macro to avoid pointer aliasing. */
 945 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
 946 {
 947     const int *p = *pp;
 948     int n, *tab;
 949
 950     tab = cv->tab;
 951     switch(*t = *p++) {
 952     case TOK_CINT:
 953     case TOK_CUINT:
 954     case TOK_CCHAR:
 955     case TOK_LCHAR:
 956     case TOK_CFLOAT:
 957     case TOK_LINENUM:
 958         tab[0] = *p++;
 959         break;
 960     case TOK_STR:
 961     case TOK_LSTR:
 962     case TOK_PPNUM:
 963         cv->cstr = (CString *)p;
 964         cv->cstr->data = (char *)p + sizeof(CString);
 965         p += (sizeof(CString) + cv->cstr->size + 3) >> 2;
 966         break;
 967     case TOK_CDOUBLE:
 968     case TOK_CLLONG:
 969     case TOK_CULLONG:
 970         n = 2;
 971         goto copy;
 972     case TOK_CLDOUBLE:
 973 #if LDOUBLE_SIZE == 16
 974         n = 4;
 975 #elif LDOUBLE_SIZE == 12
 976         n = 3;
 977 #elif LDOUBLE_SIZE == 8
 978         n = 2;
 979 #else
 980 # error add long double size support
 981 #endif
 982     copy:
 983         do
 984             *tab++ = *p++;
 985         while (--n);
 986         break;
 987     default:
 988         break;
 989     }
 990     *pp = p;
 991 }
 992
 993 static int macro_is_equal(const int *a, const int *b)
 994 {
 995     char buf[STRING_MAX_SIZE + 1];
 996     CValue cv;
 997     int t;
 998     while (*a && *b) {
 999         TOK_GET(&t, &a, &cv);
1000         pstrcpy(buf, sizeof buf, get_tok_str(t, &cv));
1001         TOK_GET(&t, &b, &cv);
1002         if (strcmp(buf, get_tok_str(t, &cv)))
1003             return 0;
1004     }
1005     return !(*a || *b);
1006 }
1007
1008 /* defines handling */
1009 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1010 {
1011     Sym *s;
1012
1013     s = define_find(v);
1014     if (s && !macro_is_equal(s->d, str))
1015         warning("%s redefined", get_tok_str(v, NULL));
1016
1017     s = sym_push2(&define_stack, v, macro_type, 0);
1018     s->d = str;
1019     s->next = first_arg;
1020     table_ident[v - TOK_IDENT]->sym_define = s;
1021 }
1022
1023 /* undefined a define symbol. Its name is just set to zero */
1024 ST_FUNC void define_undef(Sym *s)
1025 {
1026     int v;
1027     v = s->v;
1028     if (v >= TOK_IDENT && v < tok_ident)
1029         table_ident[v - TOK_IDENT]->sym_define = NULL;
1030     s->v = 0;
1031 }
1032
1033 ST_INLN Sym *define_find(int v)
1034 {
1035     v -= TOK_IDENT;
1036     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1037         return NULL;
1038     return table_ident[v]->sym_define;
1039 }
1040
1041 /* free define stack until top reaches 'b' */
1042 ST_FUNC void free_defines(Sym *b)
1043 {
1044     Sym *top, *top1;
1045     int v;
1046
1047     top = define_stack;
1048     while (top != b) {
1049         top1 = top->prev;
1050         /* do not free args or predefined defines */
1051         if (top->d)
1052             tok_str_free(top->d);
1053         v = top->v;
1054         if (v >= TOK_IDENT && v < tok_ident)
1055             table_ident[v - TOK_IDENT]->sym_define = NULL;
1056         sym_free(top);
1057         top = top1;
1058     }
1059     define_stack = b;
1060 }
1061
1062 /* label lookup */
1063 ST_FUNC Sym *label_find(int v)
1064 {
1065     v -= TOK_IDENT;
1066     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1067         return NULL;
1068     return table_ident[v]->sym_label;
1069 }
1070
1071 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1072 {
1073     Sym *s, **ps;
1074     s = sym_push2(ptop, v, 0, 0);
1075     s->r = flags;
1076     ps = &table_ident[v - TOK_IDENT]->sym_label;
1077     if (ptop == &global_label_stack) {
1078         /* modify the top most local identifier, so that
1079            sym_identifier will point to 's' when popped */
1080         while (*ps != NULL)
1081             ps = &(*ps)->prev_tok;
1082     }
1083     s->prev_tok = *ps;
1084     *ps = s;
1085     return s;
1086 }
1087
1088 /* pop labels until element last is reached. Look if any labels are
1089    undefined. Define symbols if '&&label' was used. */
1090 ST_FUNC void label_pop(Sym **ptop, Sym *slast)
1091 {
1092     Sym *s, *s1;
1093     for(s = *ptop; s != slast; s = s1) {
1094         s1 = s->prev;
1095         if (s->r == LABEL_DECLARED) {
1096             warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1097         } else if (s->r == LABEL_FORWARD) {
1098                 error("label '%s' used but not defined",
1099                       get_tok_str(s->v, NULL));
1100         } else {
1101             if (s->c) {
1102                 /* define corresponding symbol. A size of
1103                    1 is put. */
1104                 put_extern_sym(s, cur_text_section, s->jnext, 1);
1105             }
1106         }
1107         /* remove label */
1108         table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1109         sym_free(s);
1110     }
1111     *ptop = slast;
1112 }
1113
1114 /* eval an expression for #if/#elif */
1115 static int expr_preprocess(void)
1116 {
1117     int c, t;
1118     TokenString str;
1119
1120     tok_str_new(&str);
1121     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1122         next(); /* do macro subst */
1123         if (tok == TOK_DEFINED) {
1124             next_nomacro();
1125             t = tok;
1126             if (t == '(')
1127                 next_nomacro();
1128             c = define_find(tok) != 0;
1129             if (t == '(')
1130                 next_nomacro();
1131             tok = TOK_CINT;
1132             tokc.i = c;
1133         } else if (tok >= TOK_IDENT) {
1134             /* if undefined macro */
1135             tok = TOK_CINT;
1136             tokc.i = 0;
1137         }
1138         tok_str_add_tok(&str);
1139     }
1140     tok_str_add(&str, -1); /* simulate end of file */
1141     tok_str_add(&str, 0);
1142     /* now evaluate C constant expression */
1143     macro_ptr = str.str;
1144     next();
1145     c = expr_const();
1146     macro_ptr = NULL;
1147     tok_str_free(str.str);
1148     return c != 0;
1149 }
1150
1151 #if defined(PARSE_DEBUG) || defined(PP_DEBUG)
1152 static void tok_print(int *str)
1153 {
1154     int t;
1155     CValue cval;
1156
1157     printf("<");
1158     while (1) {
1159         TOK_GET(&t, &str, &cval);
1160         if (!t)
1161             break;
1162         printf("%s", get_tok_str(t, &cval));
1163     }
1164     printf(">\n");
1165 }
1166 #endif
1167
1168 /* parse after #define */
1169 ST_FUNC void parse_define(void)
1170 {
1171     Sym *s, *first, **ps;
1172     int v, t, varg, is_vaargs, spc;
1173     TokenString str;
1174
1175     v = tok;
1176     if (v < TOK_IDENT)
1177         error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1178     /* XXX: should check if same macro (ANSI) */
1179     first = NULL;
1180     t = MACRO_OBJ;
1181     /* '(' must be just after macro definition for MACRO_FUNC */
1182     next_nomacro_spc();
1183     if (tok == '(') {
1184         next_nomacro();
1185         ps = &first;
1186         while (tok != ')') {
1187             varg = tok;
1188             next_nomacro();
1189             is_vaargs = 0;
1190             if (varg == TOK_DOTS) {
1191                 varg = TOK___VA_ARGS__;
1192                 is_vaargs = 1;
1193             } else if (tok == TOK_DOTS && gnu_ext) {
1194                 is_vaargs = 1;
1195                 next_nomacro();
1196             }
1197             if (varg < TOK_IDENT)
1198                 error("badly punctuated parameter list");
1199             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1200             *ps = s;
1201             ps = &s->next;
1202             if (tok != ',')
1203                 break;
1204             next_nomacro();
1205         }
1206         if (tok == ')')
1207             next_nomacro_spc();
1208         t = MACRO_FUNC;
1209     }
1210     tok_str_new(&str);
1211     spc = 2;
1212     /* EOF testing necessary for '-D' handling */
1213     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1214         /* remove spaces around ## and after '#' */
1215         if (TOK_TWOSHARPS == tok) {
1216             if (1 == spc)
1217                 --str.len;
1218             spc = 2;
1219         } else if ('#' == tok) {
1220             spc = 2;
1221         } else if (check_space(tok, &spc)) {
1222             goto skip;
1223         }
1224         tok_str_add2(&str, tok, &tokc);
1225     skip:
1226         next_nomacro_spc();
1227     }
1228     if (spc == 1)
1229         --str.len; /* remove trailing space */
1230     tok_str_add(&str, 0);
1231 #ifdef PP_DEBUG
1232     printf("define %s %d: ", get_tok_str(v, NULL), t);
1233     tok_print(str.str);
1234 #endif
1235     define_push(v, t, str.str, first);
1236 }
1237
1238 static inline int hash_cached_include(int type, const char *filename)
1239 {
1240     const unsigned char *s;
1241     unsigned int h;
1242
1243     h = TOK_HASH_INIT;
1244     h = TOK_HASH_FUNC(h, type);
1245     s = filename;
1246     while (*s) {
1247         h = TOK_HASH_FUNC(h, *s);
1248         s++;
1249     }
1250     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1251     return h;
1252 }
1253
1254 /* XXX: use a token or a hash table to accelerate matching ? */
1255 static CachedInclude *search_cached_include(TCCState *s1,
1256                                             int type, const char *filename)
1257 {
1258     CachedInclude *e;
1259     int i, h;
1260     h = hash_cached_include(type, filename);
1261     i = s1->cached_includes_hash[h];
1262     for(;;) {
1263         if (i == 0)
1264             break;
1265         e = s1->cached_includes[i - 1];
1266         if (e->type == type && !PATHCMP(e->filename, filename))
1267             return e;
1268         i = e->hash_next;
1269     }
1270     return NULL;
1271 }
1272
1273 static inline void add_cached_include(TCCState *s1, int type,
1274                                       const char *filename, int ifndef_macro)
1275 {
1276     CachedInclude *e;
1277     int h;
1278
1279     if (search_cached_include(s1, type, filename))
1280         return;
1281 #ifdef INC_DEBUG
1282     printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1283 #endif
1284     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1285     if (!e)
1286         return;
1287     e->type = type;
1288     strcpy(e->filename, filename);
1289     e->ifndef_macro = ifndef_macro;
1290     dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1291     /* add in hash table */
1292     h = hash_cached_include(type, filename);
1293     e->hash_next = s1->cached_includes_hash[h];
1294     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1295 }
1296
1297 static void pragma_parse(TCCState *s1)
1298 {
1299     int val;
1300
1301     next();
1302     if (tok == TOK_pack) {
1303         /*
1304           This may be:
1305           #pragma pack(1) // set
1306           #pragma pack() // reset to default
1307           #pragma pack(push,1) // push & set
1308           #pragma pack(pop) // restore previous
1309         */
1310         next();
1311         skip('(');
1312         if (tok == TOK_ASM_pop) {
1313             next();
1314             if (s1->pack_stack_ptr <= s1->pack_stack) {
1315             stk_error:
1316                 error("out of pack stack");
1317             }
1318             s1->pack_stack_ptr--;
1319         } else {
1320             val = 0;
1321             if (tok != ')') {
1322                 if (tok == TOK_ASM_push) {
1323                     next();
1324                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1325                         goto stk_error;
1326                     s1->pack_stack_ptr++;
1327                     skip(',');
1328                 }
1329                 if (tok != TOK_CINT) {
1330                 pack_error:
1331                     error("invalid pack pragma");
1332                 }
1333                 val = tokc.i;
1334                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1335                     goto pack_error;
1336                 next();
1337             }
1338             *s1->pack_stack_ptr = val;
1339             skip(')');
1340         }
1341     }
1342 }
1343
1344 /* is_bof is true if first non space token at beginning of file */
1345 ST_FUNC void preprocess(int is_bof)
1346 {
1347     TCCState *s1 = tcc_state;
1348     int i, c, n, saved_parse_flags;
1349     char buf[1024], *q;
1350     Sym *s;
1351
1352     saved_parse_flags = parse_flags;
1353     parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM |
1354         PARSE_FLAG_LINEFEED;
1355     next_nomacro();
1356  redo:
1357     switch(tok) {
1358     case TOK_DEFINE:
1359         next_nomacro();
1360         parse_define();
1361         break;
1362     case TOK_UNDEF:
1363         next_nomacro();
1364         s = define_find(tok);
1365         /* undefine symbol by putting an invalid name */
1366         if (s)
1367             define_undef(s);
1368         break;
1369     case TOK_INCLUDE:
1370     case TOK_INCLUDE_NEXT:
1371         ch = file->buf_ptr[0];
1372         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1373         skip_spaces();
1374         if (ch == '<') {
1375             c = '>';
1376             goto read_name;
1377         } else if (ch == '\"') {
1378             c = ch;
1379         read_name:
1380             inp();
1381             q = buf;
1382             while (ch != c && ch != '\n' && ch != CH_EOF) {
1383                 if ((q - buf) < sizeof(buf) - 1)
1384                     *q++ = ch;
1385                 if (ch == '\\') {
1386                     if (handle_stray_noerror() == 0)
1387                         --q;
1388                 } else
1389                     inp();
1390             }
1391             *q = '\0';
1392             minp();
1393 #if 0
1394             /* eat all spaces and comments after include */
1395             /* XXX: slightly incorrect */
1396             while (ch1 != '\n' && ch1 != CH_EOF)
1397                 inp();
1398 #endif
1399         } else {
1400             /* computed #include : either we have only strings or
1401                we have anything enclosed in '<>' */
1402             next();
1403             buf[0] = '\0';
1404             if (tok == TOK_STR) {
1405                 while (tok != TOK_LINEFEED) {
1406                     if (tok != TOK_STR) {
1407                     include_syntax:
1408                         error("'#include' expects \"FILENAME\" or <FILENAME>");
1409                     }
1410                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1411                     next();
1412                 }
1413                 c = '\"';
1414             } else {
1415                 int len;
1416                 while (tok != TOK_LINEFEED) {
1417                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1418                     next();
1419                 }
1420                 len = strlen(buf);
1421                 /* check syntax and remove '<>' */
1422                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1423                     goto include_syntax;
1424                 memmove(buf, buf + 1, len - 2);
1425                 buf[len - 2] = '\0';
1426                 c = '>';
1427             }
1428         }
1429
1430         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1431             error("#include recursion too deep");
1432
1433         n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1434         for (i = -2; i < n; ++i) {
1435             char buf1[sizeof file->filename];
1436             CachedInclude *e;
1437             const char *path;
1438             int size, fd;
1439
1440             if (i == -2) {
1441                 /* check absolute include path */
1442                 if (!IS_ABSPATH(buf))
1443                     continue;
1444                 buf1[0] = 0;
1445
1446             } else if (i == -1) {
1447                 /* search in current dir if "header.h" */
1448                 if (c != '\"')
1449                     continue;
1450                 size = tcc_basename(file->filename) - file->filename;
1451                 memcpy(buf1, file->filename, size);
1452                 buf1[size] = '\0';
1453
1454             } else {
1455                 /* search in all the include paths */
1456                 if (i < s1->nb_include_paths)
1457                     path = s1->include_paths[i];
1458                 else
1459                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1460                 pstrcpy(buf1, sizeof(buf1), path);
1461                 pstrcat(buf1, sizeof(buf1), "/");
1462             }
1463
1464             pstrcat(buf1, sizeof(buf1), buf);
1465
1466             e = search_cached_include(s1, c, buf1);
1467             if (e && define_find(e->ifndef_macro)) {
1468                 /* no need to parse the include because the 'ifndef macro'
1469                    is defined */
1470 #ifdef INC_DEBUG
1471                 printf("%s: skipping %s\n", file->filename, buf);
1472 #endif
1473                 fd = 0;
1474             } else {
1475                 fd = tcc_open(s1, buf1);
1476                 if (fd < 0)
1477                     continue;
1478             }
1479
1480             if (tok == TOK_INCLUDE_NEXT) {
1481                 tok = TOK_INCLUDE;
1482                 if (fd)
1483                     tcc_close();
1484                 continue;
1485             }
1486
1487             if (0 == fd)
1488                 goto include_done;
1489
1490 #ifdef INC_DEBUG
1491             printf("%s: including %s\n", file->filename, buf1);
1492 #endif
1493             /* update target deps */
1494             dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1495                     tcc_strdup(buf1));
1496            /* XXX: fix current line init */
1497            /* push current file in stack */
1498             *s1->include_stack_ptr++ = file->prev;
1499             file->inc_type = c;
1500             pstrcpy(file->inc_filename, sizeof(file->inc_filename), buf1);
1501             /* add include file debug info */
1502             if (s1->do_debug)
1503                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1504             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1505             ch = file->buf_ptr[0];
1506             goto the_end;
1507         }
1508         error("include file '%s' not found", buf);
1509 include_done:
1510         break;
1511     case TOK_IFNDEF:
1512         c = 1;
1513         goto do_ifdef;
1514     case TOK_IF:
1515         c = expr_preprocess();
1516         goto do_if;
1517     case TOK_IFDEF:
1518         c = 0;
1519     do_ifdef:
1520         next_nomacro();
1521         if (tok < TOK_IDENT)
1522             error("invalid argument for '#if%sdef'", c ? "n" : "");
1523         if (is_bof) {
1524             if (c) {
1525 #ifdef INC_DEBUG
1526                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1527 #endif
1528                 file->ifndef_macro = tok;
1529             }
1530         }
1531         c = (define_find(tok) != 0) ^ c;
1532     do_if:
1533         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1534             error("memory full");
1535         *s1->ifdef_stack_ptr++ = c;
1536         goto test_skip;
1537     case TOK_ELSE:
1538         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1539             error("#else without matching #if");
1540         if (s1->ifdef_stack_ptr[-1] & 2)
1541             error("#else after #else");
1542         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1543         goto test_else;
1544     case TOK_ELIF:
1545         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1546             error("#elif without matching #if");
1547         c = s1->ifdef_stack_ptr[-1];
1548         if (c > 1)
1549             error("#elif after #else");
1550         /* last #if/#elif expression was true: we skip */
1551         if (c == 1)
1552             goto skip;
1553         c = expr_preprocess();
1554         s1->ifdef_stack_ptr[-1] = c;
1555     test_else:
1556         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1557             file->ifndef_macro = 0;
1558     test_skip:
1559         if (!(c & 1)) {
1560         skip:
1561             preprocess_skip();
1562             is_bof = 0;
1563             goto redo;
1564         }
1565         break;
1566     case TOK_ENDIF:
1567         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1568             error("#endif without matching #if");
1569         s1->ifdef_stack_ptr--;
1570         /* '#ifndef macro' was at the start of file. Now we check if
1571            an '#endif' is exactly at the end of file */
1572         if (file->ifndef_macro &&
1573             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1574             file->ifndef_macro_saved = file->ifndef_macro;
1575             /* need to set to zero to avoid false matches if another
1576                #ifndef at middle of file */
1577             file->ifndef_macro = 0;
1578             while (tok != TOK_LINEFEED)
1579                 next_nomacro();
1580             tok_flags |= TOK_FLAG_ENDIF;
1581             goto the_end;
1582         }
1583         break;
1584     case TOK_LINE:
1585         next();
1586         if (tok != TOK_CINT)
1587             error("#line");
1588         file->line_num = tokc.i - 1; /* the line number will be incremented after */
1589         next();
1590         if (tok != TOK_LINEFEED) {
1591             if (tok != TOK_STR)
1592                 error("#line");
1593             pstrcpy(file->filename, sizeof(file->filename),
1594                     (char *)tokc.cstr->data);
1595         }
1596         break;
1597     case TOK_ERROR:
1598     case TOK_WARNING:
1599         c = tok;
1600         ch = file->buf_ptr[0];
1601         skip_spaces();
1602         q = buf;
1603         while (ch != '\n' && ch != CH_EOF) {
1604             if ((q - buf) < sizeof(buf) - 1)
1605                 *q++ = ch;
1606             if (ch == '\\') {
1607                 if (handle_stray_noerror() == 0)
1608                     --q;
1609             } else
1610                 inp();
1611         }
1612         *q = '\0';
1613         if (c == TOK_ERROR)
1614             error("#error %s", buf);
1615         else
1616             warning("#warning %s", buf);
1617         break;
1618     case TOK_PRAGMA:
1619         pragma_parse(s1);
1620         break;
1621     default:
1622         if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_PPNUM) {
1623             /* '!' is ignored to allow C scripts. numbers are ignored
1624                to emulate cpp behaviour */
1625         } else {
1626             if (!(saved_parse_flags & PARSE_FLAG_ASM_COMMENTS))
1627                 warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1628             else {
1629                 /* this is a gas line comment in an 'S' file. */
1630                 file->buf_ptr = parse_line_comment(file->buf_ptr);
1631                 goto the_end;
1632             }
1633         }
1634         break;
1635     }
1636     /* ignore other preprocess commands or #! for C scripts */
1637     while (tok != TOK_LINEFEED)
1638         next_nomacro();
1639  the_end:
1640     parse_flags = saved_parse_flags;
1641 }
1642
1643 /* evaluate escape codes in a string. */
1644 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1645 {
1646     int c, n;
1647     const uint8_t *p;
1648
1649     p = buf;
1650     for(;;) {
1651         c = *p;
1652         if (c == '\0')
1653             break;
1654         if (c == '\\') {
1655             p++;
1656             /* escape */
1657             c = *p;
1658             switch(c) {
1659             case '0': case '1': case '2': case '3':
1660             case '4': case '5': case '6': case '7':
1661                 /* at most three octal digits */
1662                 n = c - '0';
1663                 p++;
1664                 c = *p;
1665                 if (isoct(c)) {
1666                     n = n * 8 + c - '0';
1667                     p++;
1668                     c = *p;
1669                     if (isoct(c)) {
1670                         n = n * 8 + c - '0';
1671                         p++;
1672                     }
1673                 }
1674                 c = n;
1675                 goto add_char_nonext;
1676             case 'x':
1677             case 'u':
1678             case 'U':
1679                 p++;
1680                 n = 0;
1681                 for(;;) {
1682                     c = *p;
1683                     if (c >= 'a' && c <= 'f')
1684                         c = c - 'a' + 10;
1685                     else if (c >= 'A' && c <= 'F')
1686                         c = c - 'A' + 10;
1687                     else if (isnum(c))
1688                         c = c - '0';
1689                     else
1690                         break;
1691                     n = n * 16 + c;
1692                     p++;
1693                 }
1694                 c = n;
1695                 goto add_char_nonext;
1696             case 'a':
1697                 c = '\a';
1698                 break;
1699             case 'b':
1700                 c = '\b';
1701                 break;
1702             case 'f':
1703                 c = '\f';
1704                 break;
1705             case 'n':
1706                 c = '\n';
1707                 break;
1708             case 'r':
1709                 c = '\r';
1710                 break;
1711             case 't':
1712                 c = '\t';
1713                 break;
1714             case 'v':
1715                 c = '\v';
1716                 break;
1717             case 'e':
1718                 if (!gnu_ext)
1719                     goto invalid_escape;
1720                 c = 27;
1721                 break;
1722             case '\'':
1723             case '\"':
1724             case '\\':
1725             case '?':
1726                 break;
1727             default:
1728             invalid_escape:
1729                 if (c >= '!' && c <= '~')
1730                     warning("unknown escape sequence: \'\\%c\'", c);
1731                 else
1732                     warning("unknown escape sequence: \'\\x%x\'", c);
1733                 break;
1734             }
1735         }
1736         p++;
1737     add_char_nonext:
1738         if (!is_long)
1739             cstr_ccat(outstr, c);
1740         else
1741             cstr_wccat(outstr, c);
1742     }
1743     /* add a trailing '\0' */
1744     if (!is_long)
1745         cstr_ccat(outstr, '\0');
1746     else
1747         cstr_wccat(outstr, '\0');
1748 }
1749
1750 /* we use 64 bit numbers */
1751 #define BN_SIZE 2
1752
1753 /* bn = (bn << shift) | or_val */
1754 static void bn_lshift(unsigned int *bn, int shift, int or_val)
1755 {
1756     int i;
1757     unsigned int v;
1758     for(i=0;i<BN_SIZE;i++) {
1759         v = bn[i];
1760         bn[i] = (v << shift) | or_val;
1761         or_val = v >> (32 - shift);
1762     }
1763 }
1764
1765 static void bn_zero(unsigned int *bn)
1766 {
1767     int i;
1768     for(i=0;i<BN_SIZE;i++) {
1769         bn[i] = 0;
1770     }
1771 }
1772
1773 /* parse number in null terminated string 'p' and return it in the
1774    current token */
1775 static void parse_number(const char *p)
1776 {
1777     int b, t, shift, frac_bits, s, exp_val, ch;
1778     char *q;
1779     unsigned int bn[BN_SIZE];
1780     double d;
1781
1782     /* number */
1783     q = token_buf;
1784     ch = *p++;
1785     t = ch;
1786     ch = *p++;
1787     *q++ = t;
1788     b = 10;
1789     if (t == '.') {
1790         goto float_frac_parse;
1791     } else if (t == '0') {
1792         if (ch == 'x' || ch == 'X') {
1793             q--;
1794             ch = *p++;
1795             b = 16;
1796         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
1797             q--;
1798             ch = *p++;
1799             b = 2;
1800         }
1801     }
1802     /* parse all digits. cannot check octal numbers at this stage
1803        because of floating point constants */
1804     while (1) {
1805         if (ch >= 'a' && ch <= 'f')
1806             t = ch - 'a' + 10;
1807         else if (ch >= 'A' && ch <= 'F')
1808             t = ch - 'A' + 10;
1809         else if (isnum(ch))
1810             t = ch - '0';
1811         else
1812             break;
1813         if (t >= b)
1814             break;
1815         if (q >= token_buf + STRING_MAX_SIZE) {
1816         num_too_long:
1817             error("number too long");
1818         }
1819         *q++ = ch;
1820         ch = *p++;
1821     }
1822     if (ch == '.' ||
1823         ((ch == 'e' || ch == 'E') && b == 10) ||
1824         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
1825         if (b != 10) {
1826             /* NOTE: strtox should support that for hexa numbers, but
1827                non ISOC99 libcs do not support it, so we prefer to do
1828                it by hand */
1829             /* hexadecimal or binary floats */
1830             /* XXX: handle overflows */
1831             *q = '\0';
1832             if (b == 16)
1833                 shift = 4;
1834             else
1835                 shift = 2;
1836             bn_zero(bn);
1837             q = token_buf;
1838             while (1) {
1839                 t = *q++;
1840                 if (t == '\0') {
1841                     break;
1842                 } else if (t >= 'a') {
1843                     t = t - 'a' + 10;
1844                 } else if (t >= 'A') {
1845                     t = t - 'A' + 10;
1846                 } else {
1847                     t = t - '0';
1848                 }
1849                 bn_lshift(bn, shift, t);
1850             }
1851             frac_bits = 0;
1852             if (ch == '.') {
1853                 ch = *p++;
1854                 while (1) {
1855                     t = ch;
1856                     if (t >= 'a' && t <= 'f') {
1857                         t = t - 'a' + 10;
1858                     } else if (t >= 'A' && t <= 'F') {
1859                         t = t - 'A' + 10;
1860                     } else if (t >= '0' && t <= '9') {
1861                         t = t - '0';
1862                     } else {
1863                         break;
1864                     }
1865                     if (t >= b)
1866                         error("invalid digit");
1867                     bn_lshift(bn, shift, t);
1868                     frac_bits += shift;
1869                     ch = *p++;
1870                 }
1871             }
1872             if (ch != 'p' && ch != 'P')
1873                 expect("exponent");
1874             ch = *p++;
1875             s = 1;
1876             exp_val = 0;
1877             if (ch == '+') {
1878                 ch = *p++;
1879             } else if (ch == '-') {
1880                 s = -1;
1881                 ch = *p++;
1882             }
1883             if (ch < '0' || ch > '9')
1884                 expect("exponent digits");
1885             while (ch >= '0' && ch <= '9') {
1886                 exp_val = exp_val * 10 + ch - '0';
1887                 ch = *p++;
1888             }
1889             exp_val = exp_val * s;
1890
1891             /* now we can generate the number */
1892             /* XXX: should patch directly float number */
1893             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
1894             d = ldexp(d, exp_val - frac_bits);
1895             t = toup(ch);
1896             if (t == 'F') {
1897                 ch = *p++;
1898                 tok = TOK_CFLOAT;
1899                 /* float : should handle overflow */
1900                 tokc.f = (float)d;
1901             } else if (t == 'L') {
1902                 ch = *p++;
1903 #ifdef TCC_TARGET_PE
1904                 tok = TOK_CDOUBLE;
1905                 tokc.d = d;
1906 #else
1907                 tok = TOK_CLDOUBLE;
1908                 /* XXX: not large enough */
1909                 tokc.ld = (long double)d;
1910 #endif
1911             } else {
1912                 tok = TOK_CDOUBLE;
1913                 tokc.d = d;
1914             }
1915         } else {
1916             /* decimal floats */
1917             if (ch == '.') {
1918                 if (q >= token_buf + STRING_MAX_SIZE)
1919                     goto num_too_long;
1920                 *q++ = ch;
1921                 ch = *p++;
1922             float_frac_parse:
1923                 while (ch >= '0' && ch <= '9') {
1924                     if (q >= token_buf + STRING_MAX_SIZE)
1925                         goto num_too_long;
1926                     *q++ = ch;
1927                     ch = *p++;
1928                 }
1929             }
1930             if (ch == 'e' || ch == 'E') {
1931                 if (q >= token_buf + STRING_MAX_SIZE)
1932                     goto num_too_long;
1933                 *q++ = ch;
1934                 ch = *p++;
1935                 if (ch == '-' || ch == '+') {
1936                     if (q >= token_buf + STRING_MAX_SIZE)
1937                         goto num_too_long;
1938                     *q++ = ch;
1939                     ch = *p++;
1940                 }
1941                 if (ch < '0' || ch > '9')
1942                     expect("exponent digits");
1943                 while (ch >= '0' && ch <= '9') {
1944                     if (q >= token_buf + STRING_MAX_SIZE)
1945                         goto num_too_long;
1946                     *q++ = ch;
1947                     ch = *p++;
1948                 }
1949             }
1950             *q = '\0';
1951             t = toup(ch);
1952             errno = 0;
1953             if (t == 'F') {
1954                 ch = *p++;
1955                 tok = TOK_CFLOAT;
1956                 tokc.f = strtof(token_buf, NULL);
1957             } else if (t == 'L') {
1958                 ch = *p++;
1959 #ifdef TCC_TARGET_PE
1960                 tok = TOK_CDOUBLE;
1961                 tokc.d = strtod(token_buf, NULL);
1962 #else
1963                 tok = TOK_CLDOUBLE;
1964                 tokc.ld = strtold(token_buf, NULL);
1965 #endif
1966             } else {
1967                 tok = TOK_CDOUBLE;
1968                 tokc.d = strtod(token_buf, NULL);
1969             }
1970         }
1971     } else {
1972         unsigned long long n, n1;
1973         int lcount, ucount;
1974
1975         /* integer number */
1976         *q = '\0';
1977         q = token_buf;
1978         if (b == 10 && *q == '0') {
1979             b = 8;
1980             q++;
1981         }
1982         n = 0;
1983         while(1) {
1984             t = *q++;
1985             /* no need for checks except for base 10 / 8 errors */
1986             if (t == '\0') {
1987                 break;
1988             } else if (t >= 'a') {
1989                 t = t - 'a' + 10;
1990             } else if (t >= 'A') {
1991                 t = t - 'A' + 10;
1992             } else {
1993                 t = t - '0';
1994                 if (t >= b)
1995                     error("invalid digit");
1996             }
1997             n1 = n;
1998             n = n * b + t;
1999             /* detect overflow */
2000             /* XXX: this test is not reliable */
2001             if (n < n1)
2002                 error("integer constant overflow");
2003         }
2004
2005         /* XXX: not exactly ANSI compliant */
2006         if ((n & 0xffffffff00000000LL) != 0) {
2007             if ((n >> 63) != 0)
2008                 tok = TOK_CULLONG;
2009             else
2010                 tok = TOK_CLLONG;
2011         } else if (n > 0x7fffffff) {
2012             tok = TOK_CUINT;
2013         } else {
2014             tok = TOK_CINT;
2015         }
2016         lcount = 0;
2017         ucount = 0;
2018         for(;;) {
2019             t = toup(ch);
2020             if (t == 'L') {
2021                 if (lcount >= 2)
2022                     error("three 'l's in integer constant");
2023                 lcount++;
2024                 if (lcount == 2) {
2025                     if (tok == TOK_CINT)
2026                         tok = TOK_CLLONG;
2027                     else if (tok == TOK_CUINT)
2028                         tok = TOK_CULLONG;
2029                 }
2030                 ch = *p++;
2031             } else if (t == 'U') {
2032                 if (ucount >= 1)
2033                     error("two 'u's in integer constant");
2034                 ucount++;
2035                 if (tok == TOK_CINT)
2036                     tok = TOK_CUINT;
2037                 else if (tok == TOK_CLLONG)
2038                     tok = TOK_CULLONG;
2039                 ch = *p++;
2040             } else {
2041                 break;
2042             }
2043         }
2044         if (tok == TOK_CINT || tok == TOK_CUINT)
2045             tokc.ui = n;
2046         else
2047             tokc.ull = n;
2048     }
2049     if (ch)
2050         error("invalid number\n");
2051 }
2052
2053
2054 #define PARSE2(c1, tok1, c2, tok2)              \
2055     case c1:                                    \
2056         PEEKC(c, p);                            \
2057         if (c == c2) {                          \
2058             p++;                                \
2059             tok = tok2;                         \
2060         } else {                                \
2061             tok = tok1;                         \
2062         }                                       \
2063         break;
2064
2065 /* return next token without macro substitution */
2066 static inline void next_nomacro1(void)
2067 {
2068     int t, c, is_long;
2069     TokenSym *ts;
2070     uint8_t *p, *p1;
2071     unsigned int h;
2072
2073     p = file->buf_ptr;
2074  redo_no_start:
2075     c = *p;
2076     switch(c) {
2077     case ' ':
2078     case '\t':
2079         tok = c;
2080         p++;
2081         goto keep_tok_flags;
2082     case '\f':
2083     case '\v':
2084     case '\r':
2085         p++;
2086         goto redo_no_start;
2087     case '\\':
2088         /* first look if it is in fact an end of buffer */
2089         if (p >= file->buf_end) {
2090             file->buf_ptr = p;
2091             handle_eob();
2092             p = file->buf_ptr;
2093             if (p >= file->buf_end)
2094                 goto parse_eof;
2095             else
2096                 goto redo_no_start;
2097         } else {
2098             file->buf_ptr = p;
2099             ch = *p;
2100             handle_stray();
2101             p = file->buf_ptr;
2102             goto redo_no_start;
2103         }
2104     parse_eof:
2105         {
2106             TCCState *s1 = tcc_state;
2107             if ((parse_flags & PARSE_FLAG_LINEFEED)
2108                 && !(tok_flags & TOK_FLAG_EOF)) {
2109                 tok_flags |= TOK_FLAG_EOF;
2110                 tok = TOK_LINEFEED;
2111                 goto keep_tok_flags;
2112             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2113                 tok = TOK_EOF;
2114             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2115                 error("missing #endif");
2116             } else if (s1->include_stack_ptr == s1->include_stack) {
2117                 /* no include left : end of file. */
2118                 tok = TOK_EOF;
2119             } else {
2120                 tok_flags &= ~TOK_FLAG_EOF;
2121                 /* pop include file */
2122
2123                 /* test if previous '#endif' was after a #ifdef at
2124                    start of file */
2125                 if (tok_flags & TOK_FLAG_ENDIF) {
2126 #ifdef INC_DEBUG
2127                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2128 #endif
2129                     add_cached_include(s1, file->inc_type, file->inc_filename,
2130                                        file->ifndef_macro_saved);
2131                 }
2132
2133                 /* add end of include file debug info */
2134                 if (tcc_state->do_debug) {
2135                     put_stabd(N_EINCL, 0, 0);
2136                 }
2137                 /* pop include stack */
2138                 tcc_close();
2139                 s1->include_stack_ptr--;
2140                 p = file->buf_ptr;
2141                 goto redo_no_start;
2142             }
2143         }
2144         break;
2145
2146     case '\n':
2147         file->line_num++;
2148         tok_flags |= TOK_FLAG_BOL;
2149         p++;
2150 maybe_newline:
2151         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2152             goto redo_no_start;
2153         tok = TOK_LINEFEED;
2154         goto keep_tok_flags;
2155
2156     case '#':
2157         /* XXX: simplify */
2158         PEEKC(c, p);
2159         if ((tok_flags & TOK_FLAG_BOL) &&
2160             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2161             file->buf_ptr = p;
2162             preprocess(tok_flags & TOK_FLAG_BOF);
2163             p = file->buf_ptr;
2164             goto maybe_newline;
2165         } else {
2166             if (c == '#') {
2167                 p++;
2168                 tok = TOK_TWOSHARPS;
2169             } else {
2170                 if (parse_flags & PARSE_FLAG_ASM_COMMENTS) {
2171                     p = parse_line_comment(p - 1);
2172                     goto redo_no_start;
2173                 } else {
2174                     tok = '#';
2175                 }
2176             }
2177         }
2178         break;
2179
2180     case 'a': case 'b': case 'c': case 'd':
2181     case 'e': case 'f': case 'g': case 'h':
2182     case 'i': case 'j': case 'k': case 'l':
2183     case 'm': case 'n': case 'o': case 'p':
2184     case 'q': case 'r': case 's': case 't':
2185     case 'u': case 'v': case 'w': case 'x':
2186     case 'y': case 'z':
2187     case 'A': case 'B': case 'C': case 'D':
2188     case 'E': case 'F': case 'G': case 'H':
2189     case 'I': case 'J': case 'K':
2190     case 'M': case 'N': case 'O': case 'P':
2191     case 'Q': case 'R': case 'S': case 'T':
2192     case 'U': case 'V': case 'W': case 'X':
2193     case 'Y': case 'Z':
2194     case '_':
2195     parse_ident_fast:
2196         p1 = p;
2197         h = TOK_HASH_INIT;
2198         h = TOK_HASH_FUNC(h, c);
2199         p++;
2200         for(;;) {
2201             c = *p;
2202             if (!isidnum_table[c-CH_EOF])
2203                 break;
2204             h = TOK_HASH_FUNC(h, c);
2205             p++;
2206         }
2207         if (c != '\\') {
2208             TokenSym **pts;
2209             int len;
2210
2211             /* fast case : no stray found, so we have the full token
2212                and we have already hashed it */
2213             len = p - p1;
2214             h &= (TOK_HASH_SIZE - 1);
2215             pts = &hash_ident[h];
2216             for(;;) {
2217                 ts = *pts;
2218                 if (!ts)
2219                     break;
2220                 if (ts->len == len && !memcmp(ts->str, p1, len))
2221                     goto token_found;
2222                 pts = &(ts->hash_next);
2223             }
2224             ts = tok_alloc_new(pts, p1, len);
2225         token_found: ;
2226         } else {
2227             /* slower case */
2228             cstr_reset(&tokcstr);
2229
2230             while (p1 < p) {
2231                 cstr_ccat(&tokcstr, *p1);
2232                 p1++;
2233             }
2234             p--;
2235             PEEKC(c, p);
2236         parse_ident_slow:
2237             while (isidnum_table[c-CH_EOF]) {
2238                 cstr_ccat(&tokcstr, c);
2239                 PEEKC(c, p);
2240             }
2241             ts = tok_alloc(tokcstr.data, tokcstr.size);
2242         }
2243         tok = ts->tok;
2244         break;
2245     case 'L':
2246         t = p[1];
2247         if (t != '\\' && t != '\'' && t != '\"') {
2248             /* fast case */
2249             goto parse_ident_fast;
2250         } else {
2251             PEEKC(c, p);
2252             if (c == '\'' || c == '\"') {
2253                 is_long = 1;
2254                 goto str_const;
2255             } else {
2256                 cstr_reset(&tokcstr);
2257                 cstr_ccat(&tokcstr, 'L');
2258                 goto parse_ident_slow;
2259             }
2260         }
2261         break;
2262     case '0': case '1': case '2': case '3':
2263     case '4': case '5': case '6': case '7':
2264     case '8': case '9':
2265
2266         cstr_reset(&tokcstr);
2267         /* after the first digit, accept digits, alpha, '.' or sign if
2268            prefixed by 'eEpP' */
2269     parse_num:
2270         for(;;) {
2271             t = c;
2272             cstr_ccat(&tokcstr, c);
2273             PEEKC(c, p);
2274             if (!(isnum(c) || isid(c) || c == '.' ||
2275                   ((c == '+' || c == '-') &&
2276                    (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2277                 break;
2278         }
2279         /* We add a trailing '\0' to ease parsing */
2280         cstr_ccat(&tokcstr, '\0');
2281         tokc.cstr = &tokcstr;
2282         tok = TOK_PPNUM;
2283         break;
2284     case '.':
2285         /* special dot handling because it can also start a number */
2286         PEEKC(c, p);
2287         if (isnum(c)) {
2288             cstr_reset(&tokcstr);
2289             cstr_ccat(&tokcstr, '.');
2290             goto parse_num;
2291         } else if (c == '.') {
2292             PEEKC(c, p);
2293             if (c != '.')
2294                 expect("'.'");
2295             PEEKC(c, p);
2296             tok = TOK_DOTS;
2297         } else {
2298             tok = '.';
2299         }
2300         break;
2301     case '\'':
2302     case '\"':
2303         is_long = 0;
2304     str_const:
2305         {
2306             CString str;
2307             int sep;
2308
2309             sep = c;
2310
2311             /* parse the string */
2312             cstr_new(&str);
2313             p = parse_pp_string(p, sep, &str);
2314             cstr_ccat(&str, '\0');
2315
2316             /* eval the escape (should be done as TOK_PPNUM) */
2317             cstr_reset(&tokcstr);
2318             parse_escape_string(&tokcstr, str.data, is_long);
2319             cstr_free(&str);
2320
2321             if (sep == '\'') {
2322                 int char_size;
2323                 /* XXX: make it portable */
2324                 if (!is_long)
2325                     char_size = 1;
2326                 else
2327                     char_size = sizeof(nwchar_t);
2328                 if (tokcstr.size <= char_size)
2329                     error("empty character constant");
2330                 if (tokcstr.size > 2 * char_size)
2331                     warning("multi-character character constant");
2332                 if (!is_long) {
2333                     tokc.i = *(int8_t *)tokcstr.data;
2334                     tok = TOK_CCHAR;
2335                 } else {
2336                     tokc.i = *(nwchar_t *)tokcstr.data;
2337                     tok = TOK_LCHAR;
2338                 }
2339             } else {
2340                 tokc.cstr = &tokcstr;
2341                 if (!is_long)
2342                     tok = TOK_STR;
2343                 else
2344                     tok = TOK_LSTR;
2345             }
2346         }
2347         break;
2348
2349     case '<':
2350         PEEKC(c, p);
2351         if (c == '=') {
2352             p++;
2353             tok = TOK_LE;
2354         } else if (c == '<') {
2355             PEEKC(c, p);
2356             if (c == '=') {
2357                 p++;
2358                 tok = TOK_A_SHL;
2359             } else {
2360                 tok = TOK_SHL;
2361             }
2362         } else {
2363             tok = TOK_LT;
2364         }
2365         break;
2366
2367     case '>':
2368         PEEKC(c, p);
2369         if (c == '=') {
2370             p++;
2371             tok = TOK_GE;
2372         } else if (c == '>') {
2373             PEEKC(c, p);
2374             if (c == '=') {
2375                 p++;
2376                 tok = TOK_A_SAR;
2377             } else {
2378                 tok = TOK_SAR;
2379             }
2380         } else {
2381             tok = TOK_GT;
2382         }
2383         break;
2384
2385     case '&':
2386         PEEKC(c, p);
2387         if (c == '&') {
2388             p++;
2389             tok = TOK_LAND;
2390         } else if (c == '=') {
2391             p++;
2392             tok = TOK_A_AND;
2393         } else {
2394             tok = '&';
2395         }
2396         break;
2397
2398     case '|':
2399         PEEKC(c, p);
2400         if (c == '|') {
2401             p++;
2402             tok = TOK_LOR;
2403         } else if (c == '=') {
2404             p++;
2405             tok = TOK_A_OR;
2406         } else {
2407             tok = '|';
2408         }
2409         break;
2410
2411     case '+':
2412         PEEKC(c, p);
2413         if (c == '+') {
2414             p++;
2415             tok = TOK_INC;
2416         } else if (c == '=') {
2417             p++;
2418             tok = TOK_A_ADD;
2419         } else {
2420             tok = '+';
2421         }
2422         break;
2423
2424     case '-':
2425         PEEKC(c, p);
2426         if (c == '-') {
2427             p++;
2428             tok = TOK_DEC;
2429         } else if (c == '=') {
2430             p++;
2431             tok = TOK_A_SUB;
2432         } else if (c == '>') {
2433             p++;
2434             tok = TOK_ARROW;
2435         } else {
2436             tok = '-';
2437         }
2438         break;
2439
2440     PARSE2('!', '!', '=', TOK_NE)
2441     PARSE2('=', '=', '=', TOK_EQ)
2442     PARSE2('*', '*', '=', TOK_A_MUL)
2443     PARSE2('%', '%', '=', TOK_A_MOD)
2444     PARSE2('^', '^', '=', TOK_A_XOR)
2445
2446         /* comments or operator */
2447     case '/':
2448         PEEKC(c, p);
2449         if (c == '*') {
2450             p = parse_comment(p);
2451             /* comments replaced by a blank */
2452             tok = ' ';
2453             goto keep_tok_flags;
2454         } else if (c == '/') {
2455             p = parse_line_comment(p);
2456             tok = ' ';
2457             goto keep_tok_flags;
2458         } else if (c == '=') {
2459             p++;
2460             tok = TOK_A_DIV;
2461         } else {
2462             tok = '/';
2463         }
2464         break;
2465
2466         /* simple tokens */
2467     case '(':
2468     case ')':
2469     case '[':
2470     case ']':
2471     case '{':
2472     case '}':
2473     case ',':
2474     case ';':
2475     case ':':
2476     case '?':
2477     case '~':
2478     case '$': /* only used in assembler */
2479     case '@': /* dito */
2480         tok = c;
2481         p++;
2482         break;
2483     default:
2484         error("unrecognized character \\x%02x", c);
2485         break;
2486     }
2487     tok_flags = 0;
2488 keep_tok_flags:
2489     file->buf_ptr = p;
2490 #if defined(PARSE_DEBUG)
2491     printf("token = %s\n", get_tok_str(tok, &tokc));
2492 #endif
2493 }
2494
2495 /* return next token without macro substitution. Can read input from
2496    macro_ptr buffer */
2497 static void next_nomacro_spc(void)
2498 {
2499     if (macro_ptr) {
2500     redo:
2501         tok = *macro_ptr;
2502         if (tok) {
2503             TOK_GET(&tok, &macro_ptr, &tokc);
2504             if (tok == TOK_LINENUM) {
2505                 file->line_num = tokc.i;
2506                 goto redo;
2507             }
2508         }
2509     } else {
2510         next_nomacro1();
2511     }
2512 }
2513
2514 ST_FUNC void next_nomacro(void)
2515 {
2516     do {
2517         next_nomacro_spc();
2518     } while (is_space(tok));
2519 }
2520
2521 /* substitute args in macro_str and return allocated string */
2522 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2523 {
2524     int last_tok, t, spc;
2525     const int *st;
2526     Sym *s;
2527     CValue cval;
2528     TokenString str;
2529     CString cstr;
2530
2531     tok_str_new(&str);
2532     last_tok = 0;
2533     while(1) {
2534         TOK_GET(&t, &macro_str, &cval);
2535         if (!t)
2536             break;
2537         if (t == '#') {
2538             /* stringize */
2539             TOK_GET(&t, &macro_str, &cval);
2540             if (!t)
2541                 break;
2542             s = sym_find2(args, t);
2543             if (s) {
2544                 cstr_new(&cstr);
2545                 st = s->d;
2546                 spc = 0;
2547                 while (*st) {
2548                     TOK_GET(&t, &st, &cval);
2549                     if (!check_space(t, &spc))
2550                         cstr_cat(&cstr, get_tok_str(t, &cval));
2551                 }
2552                 cstr.size -= spc;
2553                 cstr_ccat(&cstr, '\0');
2554 #ifdef PP_DEBUG
2555                 printf("stringize: %s\n", (char *)cstr.data);
2556 #endif
2557                 /* add string */
2558                 cval.cstr = &cstr;
2559                 tok_str_add2(&str, TOK_STR, &cval);
2560                 cstr_free(&cstr);
2561             } else {
2562                 tok_str_add2(&str, t, &cval);
2563             }
2564         } else if (t >= TOK_IDENT) {
2565             s = sym_find2(args, t);
2566             if (s) {
2567                 st = s->d;
2568                 /* if '##' is present before or after, no arg substitution */
2569                 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2570                     /* special case for var arg macros : ## eats the
2571                        ',' if empty VA_ARGS variable. */
2572                     /* XXX: test of the ',' is not 100%
2573                        reliable. should fix it to avoid security
2574                        problems */
2575                     if (gnu_ext && s->type.t &&
2576                         last_tok == TOK_TWOSHARPS &&
2577                         str.len >= 2 && str.str[str.len - 2] == ',') {
2578                         if (*st == 0) {
2579                             /* suppress ',' '##' */
2580                             str.len -= 2;
2581                         } else {
2582                             /* suppress '##' and add variable */
2583                             str.len--;
2584                             goto add_var;
2585                         }
2586                     } else {
2587                         int t1;
2588                     add_var:
2589                         for(;;) {
2590                             TOK_GET(&t1, &st, &cval);
2591                             if (!t1)
2592                                 break;
2593                             tok_str_add2(&str, t1, &cval);
2594                         }
2595                     }
2596                 } else {
2597                     /* NOTE: the stream cannot be read when macro
2598                        substituing an argument */
2599                     macro_subst(&str, nested_list, st, NULL);
2600                 }
2601             } else {
2602                 tok_str_add(&str, t);
2603             }
2604         } else {
2605             tok_str_add2(&str, t, &cval);
2606         }
2607         last_tok = t;
2608     }
2609     tok_str_add(&str, 0);
2610     return str.str;
2611 }
2612
2613 static char const ab_month_name[12][4] =
2614 {
2615     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2616     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2617 };
2618
2619 /* do macro substitution of current token with macro 's' and add
2620    result to (tok_str,tok_len). 'nested_list' is the list of all
2621    macros we got inside to avoid recursing. Return non zero if no
2622    substitution needs to be done */
2623 static int macro_subst_tok(TokenString *tok_str,
2624                            Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2625 {
2626     Sym *args, *sa, *sa1;
2627     int mstr_allocated, parlevel, *mstr, t, t1, spc;
2628     const int *p;
2629     TokenString str;
2630     char *cstrval;
2631     CValue cval;
2632     CString cstr;
2633     char buf[32];
2634
2635     /* if symbol is a macro, prepare substitution */
2636     /* special macros */
2637     if (tok == TOK___LINE__) {
2638         snprintf(buf, sizeof(buf), "%d", file->line_num);
2639         cstrval = buf;
2640         t1 = TOK_PPNUM;
2641         goto add_cstr1;
2642     } else if (tok == TOK___FILE__) {
2643         cstrval = file->filename;
2644         goto add_cstr;
2645     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2646         time_t ti;
2647         struct tm *tm;
2648
2649         time(&ti);
2650         tm = localtime(&ti);
2651         if (tok == TOK___DATE__) {
2652             snprintf(buf, sizeof(buf), "%s %2d %d",
2653                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2654         } else {
2655             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2656                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2657         }
2658         cstrval = buf;
2659     add_cstr:
2660         t1 = TOK_STR;
2661     add_cstr1:
2662         cstr_new(&cstr);
2663         cstr_cat(&cstr, cstrval);
2664         cstr_ccat(&cstr, '\0');
2665         cval.cstr = &cstr;
2666         tok_str_add2(tok_str, t1, &cval);
2667         cstr_free(&cstr);
2668     } else {
2669         mstr = s->d;
2670         mstr_allocated = 0;
2671         if (s->type.t == MACRO_FUNC) {
2672             /* NOTE: we do not use next_nomacro to avoid eating the
2673                next token. XXX: find better solution */
2674         redo:
2675             if (macro_ptr) {
2676                 p = macro_ptr;
2677                 while (is_space(t = *p) || TOK_LINEFEED == t)
2678                     ++p;
2679                 if (t == 0 && can_read_stream) {
2680                     /* end of macro stream: we must look at the token
2681                        after in the file */
2682                     struct macro_level *ml = *can_read_stream;
2683                     macro_ptr = NULL;
2684                     if (ml)
2685                     {
2686                         macro_ptr = ml->p;
2687                         ml->p = NULL;
2688                         *can_read_stream = ml -> prev;
2689                     }
2690                     goto redo;
2691                 }
2692             } else {
2693                 /* XXX: incorrect with comments */
2694                 ch = file->buf_ptr[0];
2695                 while (is_space(ch) || ch == '\n')
2696                     cinp();
2697                 t = ch;
2698             }
2699             if (t != '(') /* no macro subst */
2700                 return -1;
2701
2702             /* argument macro */
2703             next_nomacro();
2704             next_nomacro();
2705             args = NULL;
2706             sa = s->next;
2707             /* NOTE: empty args are allowed, except if no args */
2708             for(;;) {
2709                 /* handle '()' case */
2710                 if (!args && !sa && tok == ')')
2711                     break;
2712                 if (!sa)
2713                     error("macro '%s' used with too many args",
2714                           get_tok_str(s->v, 0));
2715                 tok_str_new(&str);
2716                 parlevel = spc = 0;
2717                 /* NOTE: non zero sa->t indicates VA_ARGS */
2718                 while ((parlevel > 0 ||
2719                         (tok != ')' &&
2720                          (tok != ',' || sa->type.t))) &&
2721                        tok != -1) {
2722                     if (tok == '(')
2723                         parlevel++;
2724                     else if (tok == ')')
2725                         parlevel--;
2726                     if (tok == TOK_LINEFEED)
2727                         tok = ' ';
2728                     if (!check_space(tok, &spc))
2729                         tok_str_add2(&str, tok, &tokc);
2730                     next_nomacro_spc();
2731                 }
2732                 str.len -= spc;
2733                 tok_str_add(&str, 0);
2734                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
2735                 sa1->d = str.str;
2736                 sa = sa->next;
2737                 if (tok == ')') {
2738                     /* special case for gcc var args: add an empty
2739                        var arg argument if it is omitted */
2740                     if (sa && sa->type.t && gnu_ext)
2741                         continue;
2742                     else
2743                         break;
2744                 }
2745                 if (tok != ',')
2746                     expect(",");
2747                 next_nomacro();
2748             }
2749             if (sa) {
2750                 error("macro '%s' used with too few args",
2751                       get_tok_str(s->v, 0));
2752             }
2753
2754             /* now subst each arg */
2755             mstr = macro_arg_subst(nested_list, mstr, args);
2756             /* free memory */
2757             sa = args;
2758             while (sa) {
2759                 sa1 = sa->prev;
2760                 tok_str_free(sa->d);
2761                 sym_free(sa);
2762                 sa = sa1;
2763             }
2764             mstr_allocated = 1;
2765         }
2766         sym_push2(nested_list, s->v, 0, 0);
2767         macro_subst(tok_str, nested_list, mstr, can_read_stream);
2768         /* pop nested defined symbol */
2769         sa1 = *nested_list;
2770         *nested_list = sa1->prev;
2771         sym_free(sa1);
2772         if (mstr_allocated)
2773             tok_str_free(mstr);
2774     }
2775     return 0;
2776 }
2777
2778 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2779    return the resulting string (which must be freed). */
2780 static inline int *macro_twosharps(const int *macro_str)
2781 {
2782     const int *ptr;
2783     int t;
2784     CValue cval;
2785     TokenString macro_str1;
2786     CString cstr;
2787     int n;
2788
2789     /* we search the first '##' */
2790     for(ptr = macro_str;;) {
2791         TOK_GET(&t, &ptr, &cval);
2792         if (t == TOK_TWOSHARPS)
2793             break;
2794         /* nothing more to do if end of string */
2795         if (t == 0)
2796             return NULL;
2797     }
2798
2799     /* we saw '##', so we need more processing to handle it */
2800     tok_str_new(&macro_str1);
2801     for(ptr = macro_str;;) {
2802         TOK_GET(&tok, &ptr, &tokc);
2803         if (tok == 0)
2804             break;
2805         if (tok == TOK_TWOSHARPS)
2806             continue;
2807         while (*ptr == TOK_TWOSHARPS) {
2808             do { t = *++ptr; } while (t == TOK_NOSUBST);
2809
2810             if (t && t != TOK_TWOSHARPS) {
2811                 TOK_GET(&t, &ptr, &cval);
2812
2813                 /* We concatenate the two tokens */
2814                 cstr_new(&cstr);
2815                 cstr_cat(&cstr, get_tok_str(tok, &tokc));
2816                 n = cstr.size;
2817                 cstr_cat(&cstr, get_tok_str(t, &cval));
2818                 cstr_ccat(&cstr, '\0');
2819
2820                 tcc_open_bf(tcc_state, "<paste>", cstr.size);
2821                 memcpy(file->buffer, cstr.data, cstr.size);
2822                 for (;;) {
2823                     next_nomacro1();
2824                     if (0 == *file->buf_ptr)
2825                         break;
2826                     tok_str_add2(&macro_str1, tok, &tokc);
2827                     warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
2828                         n, cstr.data, (char*)cstr.data + n);
2829                 }
2830                 tcc_close();
2831                 cstr_reset(&cstr);
2832             }
2833         }
2834         tok_str_add2(&macro_str1, tok, &tokc);
2835     }
2836     tok_str_add(&macro_str1, 0);
2837     return macro_str1.str;
2838 }
2839
2840
2841 /* do macro substitution of macro_str and add result to
2842    (tok_str,tok_len). 'nested_list' is the list of all macros we got
2843    inside to avoid recursing. */
2844 static void macro_subst(TokenString *tok_str, Sym **nested_list,
2845                         const int *macro_str, struct macro_level ** can_read_stream)
2846 {
2847     Sym *s;
2848     int *macro_str1;
2849     const int *ptr;
2850     int t, ret, spc;
2851     CValue cval;
2852     struct macro_level ml;
2853     int force_blank;
2854
2855     /* first scan for '##' operator handling */
2856     ptr = macro_str;
2857     macro_str1 = macro_twosharps(ptr);
2858
2859     if (macro_str1)
2860         ptr = macro_str1;
2861     spc = 0;
2862     force_blank = 0;
2863
2864     while (1) {
2865         /* NOTE: ptr == NULL can only happen if tokens are read from
2866            file stream due to a macro function call */
2867         if (ptr == NULL)
2868             break;
2869         TOK_GET(&t, &ptr, &cval);
2870         if (t == 0)
2871             break;
2872         if (t == TOK_NOSUBST) {
2873             /* following token has already been subst'd. just copy it on */
2874             tok_str_add2(tok_str, TOK_NOSUBST, NULL);
2875             TOK_GET(&t, &ptr, &cval);
2876             goto no_subst;
2877         }
2878         s = define_find(t);
2879         if (s != NULL) {
2880             /* if nested substitution, do nothing */
2881             if (sym_find2(*nested_list, t)) {
2882                 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
2883                 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
2884                 goto no_subst;
2885             }
2886             ml.p = macro_ptr;
2887             if (can_read_stream)
2888                 ml.prev = *can_read_stream, *can_read_stream = &ml;
2889             macro_ptr = (int *)ptr;
2890             tok = t;
2891             ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
2892             ptr = (int *)macro_ptr;
2893             macro_ptr = ml.p;
2894             if (can_read_stream && *can_read_stream == &ml)
2895                 *can_read_stream = ml.prev;
2896             if (ret != 0)
2897                 goto no_subst;
2898             if (parse_flags & PARSE_FLAG_SPACES)
2899                 force_blank = 1;
2900         } else {
2901         no_subst:
2902             if (force_blank) {
2903                 tok_str_add(tok_str, ' ');
2904                 spc = 1;
2905                 force_blank = 0;
2906             }
2907             if (!check_space(t, &spc))
2908                 tok_str_add2(tok_str, t, &cval);
2909         }
2910     }
2911     if (macro_str1)
2912         tok_str_free(macro_str1);
2913 }
2914
2915 /* return next token with macro substitution */
2916 ST_FUNC void next(void)
2917 {
2918     Sym *nested_list, *s;
2919     TokenString str;
2920     struct macro_level *ml;
2921
2922  redo:
2923     if (parse_flags & PARSE_FLAG_SPACES)
2924         next_nomacro_spc();
2925     else
2926         next_nomacro();
2927     if (!macro_ptr) {
2928         /* if not reading from macro substituted string, then try
2929            to substitute macros */
2930         if (tok >= TOK_IDENT &&
2931             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2932             s = define_find(tok);
2933             if (s) {
2934                 /* we have a macro: we try to substitute */
2935                 tok_str_new(&str);
2936                 nested_list = NULL;
2937                 ml = NULL;
2938                 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
2939                     /* substitution done, NOTE: maybe empty */
2940                     tok_str_add(&str, 0);
2941                     macro_ptr = str.str;
2942                     macro_ptr_allocated = str.str;
2943                     goto redo;
2944                 }
2945             }
2946         }
2947     } else {
2948         if (tok == 0) {
2949             /* end of macro or end of unget buffer */
2950             if (unget_buffer_enabled) {
2951                 macro_ptr = unget_saved_macro_ptr;
2952                 unget_buffer_enabled = 0;
2953             } else {
2954                 /* end of macro string: free it */
2955                 tok_str_free(macro_ptr_allocated);
2956                 macro_ptr_allocated = NULL;
2957                 macro_ptr = NULL;
2958             }
2959             goto redo;
2960         } else if (tok == TOK_NOSUBST) {
2961             /* discard preprocessor's nosubst markers */
2962             goto redo;
2963         }
2964     }
2965
2966     /* convert preprocessor tokens into C tokens */
2967     if (tok == TOK_PPNUM &&
2968         (parse_flags & PARSE_FLAG_TOK_NUM)) {
2969         parse_number((char *)tokc.cstr->data);
2970     }
2971 }
2972
2973 /* push back current token and set current token to 'last_tok'. Only
2974    identifier case handled for labels. */
2975 ST_INLN void unget_tok(int last_tok)
2976 {
2977     int i, n;
2978     int *q;
2979     unget_saved_macro_ptr = macro_ptr;
2980     unget_buffer_enabled = 1;
2981     q = unget_saved_buffer;
2982     macro_ptr = q;
2983     *q++ = tok;
2984     n = tok_ext_size(tok) - 1;
2985     for(i=0;i<n;i++)
2986         *q++ = tokc.tab[i];
2987     *q = 0; /* end of token string */
2988     tok = last_tok;
2989 }
2990
2991
2992 /* better than nothing, but needs extension to handle '-E' option
2993    correctly too */
2994 ST_FUNC void preprocess_init(TCCState *s1)
2995 {
2996     s1->include_stack_ptr = s1->include_stack;
2997     /* XXX: move that before to avoid having to initialize
2998        file->ifdef_stack_ptr ? */
2999     s1->ifdef_stack_ptr = s1->ifdef_stack;
3000     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3001
3002     /* XXX: not ANSI compliant: bound checking says error */
3003     vtop = vstack - 1;
3004     s1->pack_stack[0] = 0;
3005     s1->pack_stack_ptr = s1->pack_stack;
3006 }
3007
3008 ST_FUNC void preprocess_new()
3009 {
3010     int i, c;
3011     const char *p, *r;
3012     TokenSym *ts;
3013
3014     /* init isid table */
3015     for(i=CH_EOF;i<256;i++)
3016         isidnum_table[i-CH_EOF] = isid(i) || isnum(i);
3017
3018     /* add all tokens */
3019     table_ident = NULL;
3020     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3021
3022     tok_ident = TOK_IDENT;
3023     p = tcc_keywords;
3024     while (*p) {
3025         r = p;
3026         for(;;) {
3027             c = *r++;
3028             if (c == '\0')
3029                 break;
3030         }
3031         ts = tok_alloc(p, r - p - 1);
3032         p = r;
3033     }
3034 }
3035
3036 /* Preprocess the current file */
3037 ST_FUNC int tcc_preprocess(TCCState *s1)
3038 {
3039     Sym *define_start;
3040
3041     BufferedFile *file_ref, **iptr, **iptr_new;
3042     int token_seen, line_ref, d;
3043     const char *s;
3044
3045     preprocess_init(s1);
3046     define_start = define_stack;
3047     ch = file->buf_ptr[0];
3048     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3049     parse_flags = PARSE_FLAG_ASM_COMMENTS | PARSE_FLAG_PREPROCESS |
3050         PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
3051     token_seen = 0;
3052     line_ref = 0;
3053     file_ref = NULL;
3054     iptr = s1->include_stack_ptr;
3055
3056     for (;;) {
3057         next();
3058         if (tok == TOK_EOF) {
3059             break;
3060         } else if (file != file_ref) {
3061             goto print_line;
3062         } else if (tok == TOK_LINEFEED) {
3063             if (!token_seen)
3064                 continue;
3065             ++line_ref;
3066             token_seen = 0;
3067         } else if (!token_seen) {
3068             d = file->line_num - line_ref;
3069             if (file != file_ref || d < 0 || d >= 8) {
3070 print_line:
3071                 iptr_new = s1->include_stack_ptr;
3072                 s = iptr_new > iptr ? " 1"
3073                   : iptr_new < iptr ? " 2"
3074                   : iptr_new > s1->include_stack ? " 3"
3075                   : ""
3076                   ;
3077                 iptr = iptr_new;
3078                 fprintf(s1->outfile, "# %d \"%s\"%s\n", file->line_num, file->filename, s);
3079             } else {
3080                 while (d)
3081                     fputs("\n", s1->outfile), --d;
3082             }
3083             line_ref = (file_ref = file)->line_num;
3084             token_seen = tok != TOK_LINEFEED;
3085             if (!token_seen)
3086                 continue;
3087         }
3088         fputs(get_tok_str(tok, &tokc), s1->outfile);
3089     }
3090     free_defines(define_start);
3091     return 0;
3092 }