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 /* true if isid(c) || isnum(c) */
  62 static unsigned char isidnum_table[256-CH_EOF];
  63
  64 static const char tcc_keywords[] =
  65 #define DEF(id, str) str "\0"
  66 #include "tcctok.h"
  67 #undef DEF
  68 ;
  69
  70 /* WARNING: the content of this string encodes token numbers */
  71 static const unsigned char tok_two_chars[] =
  72 /* outdated -- gr
  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     '<','=', TOK_LE,
  77     '>','=', TOK_GE,
  78     '!','=', TOK_NE,
  79     '&','&', TOK_LAND,
  80     '|','|', TOK_LOR,
  81     '+','+', TOK_INC,
  82     '-','-', TOK_DEC,
  83     '=','=', TOK_EQ,
  84     '<','<', TOK_SHL,
  85     '>','>', TOK_SAR,
  86     '+','=', TOK_A_ADD,
  87     '-','=', TOK_A_SUB,
  88     '*','=', TOK_A_MUL,
  89     '/','=', TOK_A_DIV,
  90     '%','=', TOK_A_MOD,
  91     '&','=', TOK_A_AND,
  92     '^','=', TOK_A_XOR,
  93     '|','=', TOK_A_OR,
  94     '-','>', TOK_ARROW,
  95     '.','.', 0xa8, // C++ token ?
  96     '#','#', TOK_TWOSHARPS,
  97     0
  98 };
  99
 100 struct macro_level {
 101     struct macro_level *prev;
 102     const int *p;
 103 };
 104
 105 static void next_nomacro_spc(void);
 106 static void macro_subst(
 107     TokenString *tok_str,
 108     Sym **nested_list,
 109     const int *macro_str,
 110     struct macro_level **can_read_stream
 111     );
 112
 113 ST_FUNC void skip(int c)
 114 {
 115     if (tok != c)
 116         tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
 117     next();
 118 }
 119
 120 ST_FUNC void expect(const char *msg)
 121 {
 122     tcc_error("%s expected", msg);
 123 }
 124
 125 /* ------------------------------------------------------------------------- */
 126 /* CString handling */
 127 static void cstr_realloc(CString *cstr, int new_size)
 128 {
 129     int size;
 130     void *data;
 131
 132     size = cstr->size_allocated;
 133     if (size == 0)
 134         size = 8; /* no need to allocate a too small first string */
 135     while (size < new_size)
 136         size = size * 2;
 137     data = tcc_realloc(cstr->data_allocated, size);
 138     cstr->data_allocated = data;
 139     cstr->size_allocated = size;
 140     cstr->data = data;
 141 }
 142
 143 /* add a byte */
 144 ST_FUNC void cstr_ccat(CString *cstr, int ch)
 145 {
 146     int size;
 147     size = cstr->size + 1;
 148     if (size > cstr->size_allocated)
 149         cstr_realloc(cstr, size);
 150     ((unsigned char *)cstr->data)[size - 1] = ch;
 151     cstr->size = size;
 152 }
 153
 154 ST_FUNC void cstr_cat(CString *cstr, const char *str)
 155 {
 156     int c;
 157     for(;;) {
 158         c = *str;
 159         if (c == '\0')
 160             break;
 161         cstr_ccat(cstr, c);
 162         str++;
 163     }
 164 }
 165
 166 /* add a wide char */
 167 ST_FUNC void cstr_wccat(CString *cstr, int ch)
 168 {
 169     int size;
 170     size = cstr->size + sizeof(nwchar_t);
 171     if (size > cstr->size_allocated)
 172         cstr_realloc(cstr, size);
 173     *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
 174     cstr->size = size;
 175 }
 176
 177 ST_FUNC void cstr_new(CString *cstr)
 178 {
 179     memset(cstr, 0, sizeof(CString));
 180 }
 181
 182 /* free string and reset it to NULL */
 183 ST_FUNC void cstr_free(CString *cstr)
 184 {
 185     tcc_free(cstr->data_allocated);
 186     cstr_new(cstr);
 187 }
 188
 189 /* reset string to empty */
 190 ST_FUNC void cstr_reset(CString *cstr)
 191 {
 192     cstr->size = 0;
 193 }
 194
 195 /* XXX: unicode ? */
 196 static void add_char(CString *cstr, int c)
 197 {
 198     if (c == '\'' || c == '\"' || c == '\\') {
 199         /* XXX: could be more precise if char or string */
 200         cstr_ccat(cstr, '\\');
 201     }
 202     if (c >= 32 && c <= 126) {
 203         cstr_ccat(cstr, c);
 204     } else {
 205         cstr_ccat(cstr, '\\');
 206         if (c == '\n') {
 207             cstr_ccat(cstr, 'n');
 208         } else {
 209             cstr_ccat(cstr, '0' + ((c >> 6) & 7));
 210             cstr_ccat(cstr, '0' + ((c >> 3) & 7));
 211             cstr_ccat(cstr, '0' + (c & 7));
 212         }
 213     }
 214 }
 215
 216 /* ------------------------------------------------------------------------- */
 217 /* allocate a new token */
 218 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
 219 {
 220     TokenSym *ts, **ptable;
 221     int i;
 222
 223     if (tok_ident >= SYM_FIRST_ANOM)
 224         tcc_error("memory full (symbols)");
 225
 226     /* expand token table if needed */
 227     i = tok_ident - TOK_IDENT;
 228     if ((i % TOK_ALLOC_INCR) == 0) {
 229         ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
 230         table_ident = ptable;
 231     }
 232
 233     ts = tcc_malloc(sizeof(TokenSym) + len);
 234     table_ident[i] = ts;
 235     ts->tok = tok_ident++;
 236     ts->sym_define = NULL;
 237     ts->sym_label = NULL;
 238     ts->sym_struct = NULL;
 239     ts->sym_identifier = NULL;
 240     ts->len = len;
 241     ts->hash_next = NULL;
 242     memcpy(ts->str, str, len);
 243     ts->str[len] = '\0';
 244     *pts = ts;
 245     return ts;
 246 }
 247
 248 #define TOK_HASH_INIT 1
 249 #define TOK_HASH_FUNC(h, c) ((h) * 263 + (c))
 250
 251 /* find a token and add it if not found */
 252 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
 253 {
 254     TokenSym *ts, **pts;
 255     int i;
 256     unsigned int h;
 257
 258     h = TOK_HASH_INIT;
 259     for(i=0;i<len;i++)
 260         h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
 261     h &= (TOK_HASH_SIZE - 1);
 262
 263     pts = &hash_ident[h];
 264     for(;;) {
 265         ts = *pts;
 266         if (!ts)
 267             break;
 268         if (ts->len == len && !memcmp(ts->str, str, len))
 269             return ts;
 270         pts = &(ts->hash_next);
 271     }
 272     return tok_alloc_new(pts, str, len);
 273 }
 274
 275 /* XXX: buffer overflow */
 276 /* XXX: float tokens */
 277 ST_FUNC char *get_tok_str(int v, CValue *cv)
 278 {
 279     static char buf[STRING_MAX_SIZE + 1];
 280     static CString cstr_buf;
 281     CString *cstr;
 282     char *p;
 283     int i, len;
 284
 285     /* NOTE: to go faster, we give a fixed buffer for small strings */
 286     cstr_reset(&cstr_buf);
 287     cstr_buf.data = buf;
 288     cstr_buf.size_allocated = sizeof(buf);
 289     p = buf;
 290
 291 /* just an explanation, should never happen:
 292     if (v <= TOK_LINENUM && v >= TOK_CINT && cv == NULL)
 293         tcc_error("internal error: get_tok_str"); */
 294
 295     switch(v) {
 296     case TOK_CINT:
 297     case TOK_CUINT:
 298         /* XXX: not quite exact, but only useful for testing */
 299         sprintf(p, "%u", cv->ui);
 300         break;
 301     case TOK_CLLONG:
 302     case TOK_CULLONG:
 303         /* XXX: not quite exact, but only useful for testing  */
 304 #ifdef _WIN32
 305         sprintf(p, "%u", (unsigned)cv->ull);
 306 #else
 307         sprintf(p, "%llu", cv->ull);
 308 #endif
 309         break;
 310     case TOK_LCHAR:
 311         cstr_ccat(&cstr_buf, 'L');
 312     case TOK_CCHAR:
 313         cstr_ccat(&cstr_buf, '\'');
 314         add_char(&cstr_buf, cv->i);
 315         cstr_ccat(&cstr_buf, '\'');
 316         cstr_ccat(&cstr_buf, '\0');
 317         break;
 318     case TOK_PPNUM:
 319         cstr = cv->cstr;
 320         len = cstr->size - 1;
 321         for(i=0;i<len;i++)
 322             add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 323         cstr_ccat(&cstr_buf, '\0');
 324         break;
 325     case TOK_LSTR:
 326         cstr_ccat(&cstr_buf, 'L');
 327     case TOK_STR:
 328         cstr = cv->cstr;
 329         cstr_ccat(&cstr_buf, '\"');
 330         if (v == TOK_STR) {
 331             len = cstr->size - 1;
 332             for(i=0;i<len;i++)
 333                 add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 334         } else {
 335             len = (cstr->size / sizeof(nwchar_t)) - 1;
 336             for(i=0;i<len;i++)
 337                 add_char(&cstr_buf, ((nwchar_t *)cstr->data)[i]);
 338         }
 339         cstr_ccat(&cstr_buf, '\"');
 340         cstr_ccat(&cstr_buf, '\0');
 341         break;
 342
 343     case TOK_CFLOAT:
 344     case TOK_CDOUBLE:
 345     case TOK_CLDOUBLE:
 346     case TOK_LINENUM:
 347         return NULL; /* should not happen */
 348
 349     /* above tokens have value, the ones below don't */
 350
 351     case TOK_LT:
 352         v = '<';
 353         goto addv;
 354     case TOK_GT:
 355         v = '>';
 356         goto addv;
 357     case TOK_DOTS:
 358         return strcpy(p, "...");
 359     case TOK_A_SHL:
 360         return strcpy(p, "<<=");
 361     case TOK_A_SAR:
 362         return strcpy(p, ">>=");
 363     default:
 364         if (v < TOK_IDENT) {
 365             /* search in two bytes table */
 366             const unsigned char *q = tok_two_chars;
 367             while (*q) {
 368                 if (q[2] == v) {
 369                     *p++ = q[0];
 370                     *p++ = q[1];
 371                     *p = '\0';
 372                     return buf;
 373                 }
 374                 q += 3;
 375             }
 376         addv:
 377             *p++ = v;
 378             *p = '\0';
 379         } else if (v < tok_ident) {
 380             return table_ident[v - TOK_IDENT]->str;
 381         } else if (v >= SYM_FIRST_ANOM) {
 382             /* special name for anonymous symbol */
 383             sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
 384         } else {
 385             /* should never happen */
 386             return NULL;
 387         }
 388         break;
 389     }
 390     return cstr_buf.data;
 391 }
 392
 393 /* fill input buffer and peek next char */
 394 static int tcc_peekc_slow(BufferedFile *bf)
 395 {
 396     int len;
 397     /* only tries to read if really end of buffer */
 398     if (bf->buf_ptr >= bf->buf_end) {
 399         if (bf->fd != -1) {
 400 #if defined(PARSE_DEBUG)
 401             len = 8;
 402 #else
 403             len = IO_BUF_SIZE;
 404 #endif
 405             len = read(bf->fd, bf->buffer, len);
 406             if (len < 0)
 407                 len = 0;
 408         } else {
 409             len = 0;
 410         }
 411         total_bytes += len;
 412         bf->buf_ptr = bf->buffer;
 413         bf->buf_end = bf->buffer + len;
 414         *bf->buf_end = CH_EOB;
 415     }
 416     if (bf->buf_ptr < bf->buf_end) {
 417         return bf->buf_ptr[0];
 418     } else {
 419         bf->buf_ptr = bf->buf_end;
 420         return CH_EOF;
 421     }
 422 }
 423
 424 /* return the current character, handling end of block if necessary
 425    (but not stray) */
 426 ST_FUNC int handle_eob(void)
 427 {
 428     return tcc_peekc_slow(file);
 429 }
 430
 431 /* read next char from current input file and handle end of input buffer */
 432 ST_INLN void inp(void)
 433 {
 434     ch = *(++(file->buf_ptr));
 435     /* end of buffer/file handling */
 436     if (ch == CH_EOB)
 437         ch = handle_eob();
 438 }
 439
 440 /* handle '\[\r]\n' */
 441 static int handle_stray_noerror(void)
 442 {
 443     while (ch == '\\') {
 444         inp();
 445         if (ch == '\n') {
 446             file->line_num++;
 447             inp();
 448         } else if (ch == '\r') {
 449             inp();
 450             if (ch != '\n')
 451                 goto fail;
 452             file->line_num++;
 453             inp();
 454         } else {
 455         fail:
 456             return 1;
 457         }
 458     }
 459     return 0;
 460 }
 461
 462 static void handle_stray(void)
 463 {
 464     if (handle_stray_noerror())
 465         tcc_error("stray '\\' in program");
 466 }
 467
 468 /* skip the stray and handle the \\n case. Output an error if
 469    incorrect char after the stray */
 470 static int handle_stray1(uint8_t *p)
 471 {
 472     int c;
 473
 474     if (p >= file->buf_end) {
 475         file->buf_ptr = p;
 476         c = handle_eob();
 477         p = file->buf_ptr;
 478         if (c == '\\')
 479             goto parse_stray;
 480     } else {
 481     parse_stray:
 482         file->buf_ptr = p;
 483         ch = *p;
 484         handle_stray();
 485         p = file->buf_ptr;
 486         c = *p;
 487     }
 488     return c;
 489 }
 490
 491 /* handle just the EOB case, but not stray */
 492 #define PEEKC_EOB(c, p)\
 493 {\
 494     p++;\
 495     c = *p;\
 496     if (c == '\\') {\
 497         file->buf_ptr = p;\
 498         c = handle_eob();\
 499         p = file->buf_ptr;\
 500     }\
 501 }
 502
 503 /* handle the complicated stray case */
 504 #define PEEKC(c, p)\
 505 {\
 506     p++;\
 507     c = *p;\
 508     if (c == '\\') {\
 509         c = handle_stray1(p);\
 510         p = file->buf_ptr;\
 511     }\
 512 }
 513
 514 /* input with '\[\r]\n' handling. Note that this function cannot
 515    handle other characters after '\', so you cannot call it inside
 516    strings or comments */
 517 ST_FUNC void minp(void)
 518 {
 519     inp();
 520     if (ch == '\\')
 521         handle_stray();
 522 }
 523
 524
 525 /* single line C++ comments */
 526 static uint8_t *parse_line_comment(uint8_t *p)
 527 {
 528     int c;
 529
 530     p++;
 531     for(;;) {
 532         c = *p;
 533     redo:
 534         if (c == '\n' || c == CH_EOF) {
 535             break;
 536         } else if (c == '\\') {
 537             file->buf_ptr = p;
 538             c = handle_eob();
 539             p = file->buf_ptr;
 540             if (c == '\\') {
 541                 PEEKC_EOB(c, p);
 542                 if (c == '\n') {
 543                     file->line_num++;
 544                     PEEKC_EOB(c, p);
 545                 } else if (c == '\r') {
 546                     PEEKC_EOB(c, p);
 547                     if (c == '\n') {
 548                         file->line_num++;
 549                         PEEKC_EOB(c, p);
 550                     }
 551                 }
 552             } else {
 553                 goto redo;
 554             }
 555         } else {
 556             p++;
 557         }
 558     }
 559     return p;
 560 }
 561
 562 /* C comments */
 563 ST_FUNC uint8_t *parse_comment(uint8_t *p)
 564 {
 565     int c;
 566
 567     p++;
 568     for(;;) {
 569         /* fast skip loop */
 570         for(;;) {
 571             c = *p;
 572             if (c == '\n' || c == '*' || c == '\\')
 573                 break;
 574             p++;
 575             c = *p;
 576             if (c == '\n' || c == '*' || c == '\\')
 577                 break;
 578             p++;
 579         }
 580         /* now we can handle all the cases */
 581         if (c == '\n') {
 582             file->line_num++;
 583             p++;
 584         } else if (c == '*') {
 585             p++;
 586             for(;;) {
 587                 c = *p;
 588                 if (c == '*') {
 589                     p++;
 590                 } else if (c == '/') {
 591                     goto end_of_comment;
 592                 } else if (c == '\\') {
 593                     file->buf_ptr = p;
 594                     c = handle_eob();
 595                     p = file->buf_ptr;
 596                     if (c == '\\') {
 597                         /* skip '\[\r]\n', otherwise just skip the stray */
 598                         while (c == '\\') {
 599                             PEEKC_EOB(c, p);
 600                             if (c == '\n') {
 601                                 file->line_num++;
 602                                 PEEKC_EOB(c, p);
 603                             } else if (c == '\r') {
 604                                 PEEKC_EOB(c, p);
 605                                 if (c == '\n') {
 606                                     file->line_num++;
 607                                     PEEKC_EOB(c, p);
 608                                 }
 609                             } else {
 610                                 goto after_star;
 611                             }
 612                         }
 613                     }
 614                 } else {
 615                     break;
 616                 }
 617             }
 618         after_star: ;
 619         } else {
 620             /* stray, eob or eof */
 621             file->buf_ptr = p;
 622             c = handle_eob();
 623             p = file->buf_ptr;
 624             if (c == CH_EOF) {
 625                 tcc_error("unexpected end of file in comment");
 626             } else if (c == '\\') {
 627                 p++;
 628             }
 629         }
 630     }
 631  end_of_comment:
 632     p++;
 633     return p;
 634 }
 635
 636 #define cinp minp
 637
 638 static inline void skip_spaces(void)
 639 {
 640     while (is_space(ch))
 641         cinp();
 642 }
 643
 644 static inline int check_space(int t, int *spc)
 645 {
 646     if (is_space(t)) {
 647         if (*spc)
 648             return 1;
 649         *spc = 1;
 650     } else
 651         *spc = 0;
 652     return 0;
 653 }
 654
 655 /* parse a string without interpreting escapes */
 656 static uint8_t *parse_pp_string(uint8_t *p,
 657                                 int sep, CString *str)
 658 {
 659     int c;
 660     p++;
 661     for(;;) {
 662         c = *p;
 663         if (c == sep) {
 664             break;
 665         } else if (c == '\\') {
 666             file->buf_ptr = p;
 667             c = handle_eob();
 668             p = file->buf_ptr;
 669             if (c == CH_EOF) {
 670             unterminated_string:
 671                 /* XXX: indicate line number of start of string */
 672                 tcc_error("missing terminating %c character", sep);
 673             } else if (c == '\\') {
 674                 /* escape : just skip \[\r]\n */
 675                 PEEKC_EOB(c, p);
 676                 if (c == '\n') {
 677                     file->line_num++;
 678                     p++;
 679                 } else if (c == '\r') {
 680                     PEEKC_EOB(c, p);
 681                     if (c != '\n')
 682                         expect("'\n' after '\r'");
 683                     file->line_num++;
 684                     p++;
 685                 } else if (c == CH_EOF) {
 686                     goto unterminated_string;
 687                 } else {
 688                     if (str) {
 689                         cstr_ccat(str, '\\');
 690                         cstr_ccat(str, c);
 691                     }
 692                     p++;
 693                 }
 694             }
 695         } else if (c == '\n') {
 696             file->line_num++;
 697             goto add_char;
 698         } else if (c == '\r') {
 699             PEEKC_EOB(c, p);
 700             if (c != '\n') {
 701                 if (str)
 702                     cstr_ccat(str, '\r');
 703             } else {
 704                 file->line_num++;
 705                 goto add_char;
 706             }
 707         } else {
 708         add_char:
 709             if (str)
 710                 cstr_ccat(str, c);
 711             p++;
 712         }
 713     }
 714     p++;
 715     return p;
 716 }
 717
 718 /* skip block of text until #else, #elif or #endif. skip also pairs of
 719    #if/#endif */
 720 static void preprocess_skip(void)
 721 {
 722     int a, start_of_line, c, in_warn_or_error;
 723     uint8_t *p;
 724
 725     p = file->buf_ptr;
 726     a = 0;
 727 redo_start:
 728     start_of_line = 1;
 729     in_warn_or_error = 0;
 730     for(;;) {
 731     redo_no_start:
 732         c = *p;
 733         switch(c) {
 734         case ' ':
 735         case '\t':
 736         case '\f':
 737         case '\v':
 738         case '\r':
 739             p++;
 740             goto redo_no_start;
 741         case '\n':
 742             file->line_num++;
 743             p++;
 744             goto redo_start;
 745         case '\\':
 746             file->buf_ptr = p;
 747             c = handle_eob();
 748             if (c == CH_EOF) {
 749                 expect("#endif");
 750             } else if (c == '\\') {
 751                 ch = file->buf_ptr[0];
 752                 handle_stray_noerror();
 753             }
 754             p = file->buf_ptr;
 755             goto redo_no_start;
 756         /* skip strings */
 757         case '\"':
 758         case '\'':
 759             if (in_warn_or_error)
 760                 goto _default;
 761             p = parse_pp_string(p, c, NULL);
 762             break;
 763         /* skip comments */
 764         case '/':
 765             if (in_warn_or_error)
 766                 goto _default;
 767             file->buf_ptr = p;
 768             ch = *p;
 769             minp();
 770             p = file->buf_ptr;
 771             if (ch == '*') {
 772                 p = parse_comment(p);
 773             } else if (ch == '/') {
 774                 p = parse_line_comment(p);
 775             }
 776             break;
 777         case '#':
 778             p++;
 779             if (start_of_line) {
 780                 file->buf_ptr = p;
 781                 next_nomacro();
 782                 p = file->buf_ptr;
 783                 if (a == 0 &&
 784                     (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
 785                     goto the_end;
 786                 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
 787                     a++;
 788                 else if (tok == TOK_ENDIF)
 789                     a--;
 790                 else if( tok == TOK_ERROR || tok == TOK_WARNING)
 791                     in_warn_or_error = 1;
 792                 else if (tok == TOK_LINEFEED)
 793                     goto redo_start;
 794             }
 795             break;
 796 _default:
 797         default:
 798             p++;
 799             break;
 800         }
 801         start_of_line = 0;
 802     }
 803  the_end: ;
 804     file->buf_ptr = p;
 805 }
 806
 807 /* ParseState handling */
 808
 809 /* XXX: currently, no include file info is stored. Thus, we cannot display
 810    accurate messages if the function or data definition spans multiple
 811    files */
 812
 813 /* save current parse state in 's' */
 814 ST_FUNC void save_parse_state(ParseState *s)
 815 {
 816     s->line_num = file->line_num;
 817     s->macro_ptr = macro_ptr;
 818     s->tok = tok;
 819     s->tokc = tokc;
 820 }
 821
 822 /* restore parse state from 's' */
 823 ST_FUNC void restore_parse_state(ParseState *s)
 824 {
 825     file->line_num = s->line_num;
 826     macro_ptr = s->macro_ptr;
 827     tok = s->tok;
 828     tokc = s->tokc;
 829 }
 830
 831 /* return the number of additional 'ints' necessary to store the
 832    token */
 833 static inline int tok_ext_size(int t)
 834 {
 835     switch(t) {
 836         /* 4 bytes */
 837     case TOK_CINT:
 838     case TOK_CUINT:
 839     case TOK_CCHAR:
 840     case TOK_LCHAR:
 841     case TOK_CFLOAT:
 842     case TOK_LINENUM:
 843         return 1;
 844     case TOK_STR:
 845     case TOK_LSTR:
 846     case TOK_PPNUM:
 847         tcc_error("unsupported token");
 848         return 1;
 849     case TOK_CDOUBLE:
 850     case TOK_CLLONG:
 851     case TOK_CULLONG:
 852         return 2;
 853     case TOK_CLDOUBLE:
 854         return LDOUBLE_SIZE / 4;
 855     default:
 856         return 0;
 857     }
 858 }
 859
 860 /* token string handling */
 861
 862 ST_INLN void tok_str_new(TokenString *s)
 863 {
 864     s->str = NULL;
 865     s->len = 0;
 866     s->allocated_len = 0;
 867     s->last_line_num = -1;
 868 }
 869
 870 ST_FUNC void tok_str_free(int *str)
 871 {
 872     tcc_free(str);
 873 }
 874
 875 static int *tok_str_realloc(TokenString *s)
 876 {
 877     int *str, len;
 878
 879     if (s->allocated_len == 0) {
 880         len = 8;
 881     } else {
 882         len = s->allocated_len * 2;
 883     }
 884     str = tcc_realloc(s->str, len * sizeof(int));
 885     s->allocated_len = len;
 886     s->str = str;
 887     return str;
 888 }
 889
 890 ST_FUNC void tok_str_add(TokenString *s, int t)
 891 {
 892     int len, *str;
 893
 894     len = s->len;
 895     str = s->str;
 896     if (len >= s->allocated_len)
 897         str = tok_str_realloc(s);
 898     str[len++] = t;
 899     s->len = len;
 900 }
 901
 902 static void tok_str_add2(TokenString *s, int t, CValue *cv)
 903 {
 904     int len, *str;
 905
 906     len = s->len;
 907     str = s->str;
 908
 909     /* allocate space for worst case */
 910     if (len + TOK_MAX_SIZE > s->allocated_len)
 911         str = tok_str_realloc(s);
 912     str[len++] = t;
 913     switch(t) {
 914     case TOK_CINT:
 915     case TOK_CUINT:
 916     case TOK_CCHAR:
 917     case TOK_LCHAR:
 918     case TOK_CFLOAT:
 919     case TOK_LINENUM:
 920         str[len++] = cv->tab[0];
 921         break;
 922     case TOK_PPNUM:
 923     case TOK_STR:
 924     case TOK_LSTR:
 925         {
 926             int nb_words;
 927             CString *cstr;
 928
 929             nb_words = (sizeof(CString) + cv->cstr->size + 3) >> 2;
 930             while ((len + nb_words) > s->allocated_len)
 931                 str = tok_str_realloc(s);
 932             cstr = (CString *)(str + len);
 933             cstr->data = NULL;
 934             cstr->size = cv->cstr->size;
 935             cstr->data_allocated = NULL;
 936             cstr->size_allocated = cstr->size;
 937             memcpy((char *)cstr + sizeof(CString),
 938                    cv->cstr->data, cstr->size);
 939             len += nb_words;
 940         }
 941         break;
 942     case TOK_CDOUBLE:
 943     case TOK_CLLONG:
 944     case TOK_CULLONG:
 945 #if LDOUBLE_SIZE == 8
 946     case TOK_CLDOUBLE:
 947 #endif
 948         str[len++] = cv->tab[0];
 949         str[len++] = cv->tab[1];
 950         break;
 951 #if LDOUBLE_SIZE == 12
 952     case TOK_CLDOUBLE:
 953         str[len++] = cv->tab[0];
 954         str[len++] = cv->tab[1];
 955         str[len++] = cv->tab[2];
 956 #elif LDOUBLE_SIZE == 16
 957     case TOK_CLDOUBLE:
 958         str[len++] = cv->tab[0];
 959         str[len++] = cv->tab[1];
 960         str[len++] = cv->tab[2];
 961         str[len++] = cv->tab[3];
 962 #elif LDOUBLE_SIZE != 8
 963 #error add long double size support
 964 #endif
 965         break;
 966     default:
 967         break;
 968     }
 969     s->len = len;
 970 }
 971
 972 /* add the current parse token in token string 's' */
 973 ST_FUNC void tok_str_add_tok(TokenString *s)
 974 {
 975     CValue cval;
 976
 977     /* save line number info */
 978     if (file->line_num != s->last_line_num) {
 979         s->last_line_num = file->line_num;
 980         cval.i = s->last_line_num;
 981         tok_str_add2(s, TOK_LINENUM, &cval);
 982     }
 983     tok_str_add2(s, tok, &tokc);
 984 }
 985
 986 /* get a token from an integer array and increment pointer
 987    accordingly. we code it as a macro to avoid pointer aliasing. */
 988 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
 989 {
 990     const int *p = *pp;
 991     int n, *tab;
 992
 993     tab = cv->tab;
 994     switch(*t = *p++) {
 995     case TOK_CINT:
 996     case TOK_CUINT:
 997     case TOK_CCHAR:
 998     case TOK_LCHAR:
 999     case TOK_CFLOAT:
1000     case TOK_LINENUM:
1001         tab[0] = *p++;
1002         break;
1003     case TOK_STR:
1004     case TOK_LSTR:
1005     case TOK_PPNUM:
1006         cv->cstr = (CString *)p;
1007         cv->cstr->data = (char *)p + sizeof(CString);
1008         p += (sizeof(CString) + cv->cstr->size + 3) >> 2;
1009         break;
1010     case TOK_CDOUBLE:
1011     case TOK_CLLONG:
1012     case TOK_CULLONG:
1013         n = 2;
1014         goto copy;
1015     case TOK_CLDOUBLE:
1016 #if LDOUBLE_SIZE == 16
1017         n = 4;
1018 #elif LDOUBLE_SIZE == 12
1019         n = 3;
1020 #elif LDOUBLE_SIZE == 8
1021         n = 2;
1022 #else
1023 # error add long double size support
1024 #endif
1025     copy:
1026         do
1027             *tab++ = *p++;
1028         while (--n);
1029         break;
1030     default:
1031         break;
1032     }
1033     *pp = p;
1034 }
1035
1036 static int macro_is_equal(const int *a, const int *b)
1037 {
1038     char buf[STRING_MAX_SIZE + 1];
1039     CValue cv;
1040     int t;
1041     while (*a && *b) {
1042         TOK_GET(&t, &a, &cv);
1043         pstrcpy(buf, sizeof buf, get_tok_str(t, &cv));
1044         TOK_GET(&t, &b, &cv);
1045         if (strcmp(buf, get_tok_str(t, &cv)))
1046             return 0;
1047     }
1048     return !(*a || *b);
1049 }
1050
1051 /* defines handling */
1052 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1053 {
1054     Sym *s;
1055
1056     s = define_find(v);
1057     if (s && !macro_is_equal(s->d, str))
1058         tcc_warning("%s redefined", get_tok_str(v, NULL));
1059
1060     s = sym_push2(&define_stack, v, macro_type, 0);
1061     s->d = str;
1062     s->next = first_arg;
1063     table_ident[v - TOK_IDENT]->sym_define = s;
1064 }
1065
1066 /* undefined a define symbol. Its name is just set to zero */
1067 ST_FUNC void define_undef(Sym *s)
1068 {
1069     int v;
1070     v = s->v;
1071     if (v >= TOK_IDENT && v < tok_ident)
1072         table_ident[v - TOK_IDENT]->sym_define = NULL;
1073     s->v = 0;
1074 }
1075
1076 ST_INLN Sym *define_find(int v)
1077 {
1078     v -= TOK_IDENT;
1079     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1080         return NULL;
1081     return table_ident[v]->sym_define;
1082 }
1083
1084 /* free define stack until top reaches 'b' */
1085 ST_FUNC void free_defines(Sym *b)
1086 {
1087     Sym *top, *top1;
1088     int v;
1089
1090     top = define_stack;
1091     while (top != b) {
1092         top1 = top->prev;
1093         /* do not free args or predefined defines */
1094         if (top->d)
1095             tok_str_free(top->d);
1096         v = top->v;
1097         if (v >= TOK_IDENT && v < tok_ident)
1098             table_ident[v - TOK_IDENT]->sym_define = NULL;
1099         sym_free(top);
1100         top = top1;
1101     }
1102     define_stack = b;
1103 }
1104
1105 /* label lookup */
1106 ST_FUNC Sym *label_find(int v)
1107 {
1108     v -= TOK_IDENT;
1109     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1110         return NULL;
1111     return table_ident[v]->sym_label;
1112 }
1113
1114 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1115 {
1116     Sym *s, **ps;
1117     s = sym_push2(ptop, v, 0, 0);
1118     s->r = flags;
1119     ps = &table_ident[v - TOK_IDENT]->sym_label;
1120     if (ptop == &global_label_stack) {
1121         /* modify the top most local identifier, so that
1122            sym_identifier will point to 's' when popped */
1123         while (*ps != NULL)
1124             ps = &(*ps)->prev_tok;
1125     }
1126     s->prev_tok = *ps;
1127     *ps = s;
1128     return s;
1129 }
1130
1131 /* pop labels until element last is reached. Look if any labels are
1132    undefined. Define symbols if '&&label' was used. */
1133 ST_FUNC void label_pop(Sym **ptop, Sym *slast)
1134 {
1135     Sym *s, *s1;
1136     for(s = *ptop; s != slast; s = s1) {
1137         s1 = s->prev;
1138         if (s->r == LABEL_DECLARED) {
1139             tcc_warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1140         } else if (s->r == LABEL_FORWARD) {
1141                 tcc_error("label '%s' used but not defined",
1142                       get_tok_str(s->v, NULL));
1143         } else {
1144             if (s->c) {
1145                 /* define corresponding symbol. A size of
1146                    1 is put. */
1147                 put_extern_sym(s, cur_text_section, s->jnext, 1);
1148             }
1149         }
1150         /* remove label */
1151         table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1152         sym_free(s);
1153     }
1154     *ptop = slast;
1155 }
1156
1157 /* eval an expression for #if/#elif */
1158 static int expr_preprocess(void)
1159 {
1160     int c, t;
1161     TokenString str;
1162
1163     tok_str_new(&str);
1164     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1165         next(); /* do macro subst */
1166         if (tok == TOK_DEFINED) {
1167             next_nomacro();
1168             t = tok;
1169             if (t == '(')
1170                 next_nomacro();
1171             c = define_find(tok) != 0;
1172             if (t == '(')
1173                 next_nomacro();
1174             tok = TOK_CINT;
1175             tokc.i = c;
1176         } else if (tok >= TOK_IDENT) {
1177             /* if undefined macro */
1178             tok = TOK_CINT;
1179             tokc.i = 0;
1180         }
1181         tok_str_add_tok(&str);
1182     }
1183     tok_str_add(&str, -1); /* simulate end of file */
1184     tok_str_add(&str, 0);
1185     /* now evaluate C constant expression */
1186     macro_ptr = str.str;
1187     next();
1188     c = expr_const();
1189     macro_ptr = NULL;
1190     tok_str_free(str.str);
1191     return c != 0;
1192 }
1193
1194 #if defined(PARSE_DEBUG) || defined(PP_DEBUG)
1195 static void tok_print(int *str)
1196 {
1197     int t;
1198     CValue cval;
1199
1200     printf("<");
1201     while (1) {
1202         TOK_GET(&t, &str, &cval);
1203         if (!t)
1204             break;
1205         printf("%s", get_tok_str(t, &cval));
1206     }
1207     printf(">\n");
1208 }
1209 #endif
1210
1211 /* parse after #define */
1212 ST_FUNC void parse_define(void)
1213 {
1214     Sym *s, *first, **ps;
1215     int v, t, varg, is_vaargs, spc, ptok, macro_list_start;
1216     TokenString str;
1217
1218     v = tok;
1219     if (v < TOK_IDENT)
1220         tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1221     /* XXX: should check if same macro (ANSI) */
1222     first = NULL;
1223     t = MACRO_OBJ;
1224     /* '(' must be just after macro definition for MACRO_FUNC */
1225     next_nomacro_spc();
1226     if (tok == '(') {
1227         next_nomacro();
1228         ps = &first;
1229         while (tok != ')') {
1230             varg = tok;
1231             next_nomacro();
1232             is_vaargs = 0;
1233             if (varg == TOK_DOTS) {
1234                 varg = TOK___VA_ARGS__;
1235                 is_vaargs = 1;
1236             } else if (tok == TOK_DOTS && gnu_ext) {
1237                 is_vaargs = 1;
1238                 next_nomacro();
1239             }
1240             if (varg < TOK_IDENT)
1241                 tcc_error("badly punctuated parameter list");
1242             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1243             *ps = s;
1244             ps = &s->next;
1245             if (tok != ',')
1246                 break;
1247             next_nomacro();
1248         }
1249         if (tok == ')')
1250             next_nomacro_spc();
1251         t = MACRO_FUNC;
1252     }
1253     tok_str_new(&str);
1254     spc = 2;
1255     /* EOF testing necessary for '-D' handling */
1256     ptok = 0;
1257     macro_list_start = 1;
1258     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1259         if (!macro_list_start && spc == 2 && tok == TOK_TWOSHARPS)
1260             tcc_error("'##' invalid at start of macro");
1261         ptok = tok;
1262         /* remove spaces around ## and after '#' */
1263         if (TOK_TWOSHARPS == tok) {
1264             if (1 == spc)
1265                 --str.len;
1266             spc = 2;
1267         } else if ('#' == tok) {
1268             spc = 2;
1269         } else if (check_space(tok, &spc)) {
1270             goto skip;
1271         }
1272         tok_str_add2(&str, tok, &tokc);
1273     skip:
1274         next_nomacro_spc();
1275         macro_list_start = 0;
1276     }
1277     if (ptok == TOK_TWOSHARPS)
1278         tcc_error("'##' invalid at end of macro");
1279     if (spc == 1)
1280         --str.len; /* remove trailing space */
1281     tok_str_add(&str, 0);
1282 #ifdef PP_DEBUG
1283     printf("define %s %d: ", get_tok_str(v, NULL), t);
1284     tok_print(str.str);
1285 #endif
1286     define_push(v, t, str.str, first);
1287 }
1288
1289 static inline int hash_cached_include(const char *filename)
1290 {
1291     const unsigned char *s;
1292     unsigned int h;
1293
1294     h = TOK_HASH_INIT;
1295     s = (unsigned char *) filename;
1296     while (*s) {
1297         h = TOK_HASH_FUNC(h, *s);
1298         s++;
1299     }
1300     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1301     return h;
1302 }
1303
1304 static CachedInclude *search_cached_include(TCCState *s1, const char *filename)
1305 {
1306     CachedInclude *e;
1307     int i, h;
1308     h = hash_cached_include(filename);
1309     i = s1->cached_includes_hash[h];
1310     for(;;) {
1311         if (i == 0)
1312             break;
1313         e = s1->cached_includes[i - 1];
1314         if (0 == PATHCMP(e->filename, filename))
1315             return e;
1316         i = e->hash_next;
1317     }
1318     return NULL;
1319 }
1320
1321 static inline void add_cached_include(TCCState *s1, const char *filename, int ifndef_macro)
1322 {
1323     CachedInclude *e;
1324     int h;
1325
1326     if (search_cached_include(s1, filename))
1327         return;
1328 #ifdef INC_DEBUG
1329     printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1330 #endif
1331     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1332     strcpy(e->filename, filename);
1333     e->ifndef_macro = ifndef_macro;
1334     dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1335     /* add in hash table */
1336     h = hash_cached_include(filename);
1337     e->hash_next = s1->cached_includes_hash[h];
1338     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1339 }
1340
1341 static void pragma_parse(TCCState *s1)
1342 {
1343     int val;
1344
1345     next();
1346     if (tok == TOK_pack) {
1347         /*
1348           This may be:
1349           #pragma pack(1) // set
1350           #pragma pack() // reset to default
1351           #pragma pack(push,1) // push & set
1352           #pragma pack(pop) // restore previous
1353         */
1354         next();
1355         skip('(');
1356         if (tok == TOK_ASM_pop) {
1357             next();
1358             if (s1->pack_stack_ptr <= s1->pack_stack) {
1359             stk_error:
1360                 tcc_error("out of pack stack");
1361             }
1362             s1->pack_stack_ptr--;
1363         } else {
1364             val = 0;
1365             if (tok != ')') {
1366                 if (tok == TOK_ASM_push) {
1367                     next();
1368                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1369                         goto stk_error;
1370                     s1->pack_stack_ptr++;
1371                     skip(',');
1372                 }
1373                 if (tok != TOK_CINT) {
1374                 pack_error:
1375                     tcc_error("invalid pack pragma");
1376                 }
1377                 val = tokc.i;
1378                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1379                     goto pack_error;
1380                 next();
1381             }
1382             *s1->pack_stack_ptr = val;
1383             skip(')');
1384         }
1385     }
1386 }
1387
1388 /* is_bof is true if first non space token at beginning of file */
1389 ST_FUNC void preprocess(int is_bof)
1390 {
1391     TCCState *s1 = tcc_state;
1392     int i, c, n, saved_parse_flags;
1393     char buf[1024], *q;
1394     Sym *s;
1395
1396     saved_parse_flags = parse_flags;
1397     parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM |
1398         PARSE_FLAG_LINEFEED;
1399     next_nomacro();
1400  redo:
1401     switch(tok) {
1402     case TOK_DEFINE:
1403         next_nomacro();
1404         parse_define();
1405         break;
1406     case TOK_UNDEF:
1407         next_nomacro();
1408         s = define_find(tok);
1409         /* undefine symbol by putting an invalid name */
1410         if (s)
1411             define_undef(s);
1412         break;
1413     case TOK_INCLUDE:
1414     case TOK_INCLUDE_NEXT:
1415         ch = file->buf_ptr[0];
1416         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1417         skip_spaces();
1418         if (ch == '<') {
1419             c = '>';
1420             goto read_name;
1421         } else if (ch == '\"') {
1422             c = ch;
1423         read_name:
1424             inp();
1425             q = buf;
1426             while (ch != c && ch != '\n' && ch != CH_EOF) {
1427                 if ((q - buf) < sizeof(buf) - 1)
1428                     *q++ = ch;
1429                 if (ch == '\\') {
1430                     if (handle_stray_noerror() == 0)
1431                         --q;
1432                 } else
1433                     inp();
1434             }
1435             *q = '\0';
1436             minp();
1437 #if 0
1438             /* eat all spaces and comments after include */
1439             /* XXX: slightly incorrect */
1440             while (ch1 != '\n' && ch1 != CH_EOF)
1441                 inp();
1442 #endif
1443         } else {
1444             /* computed #include : either we have only strings or
1445                we have anything enclosed in '<>' */
1446             next();
1447             buf[0] = '\0';
1448             if (tok == TOK_STR) {
1449                 while (tok != TOK_LINEFEED) {
1450                     if (tok != TOK_STR) {
1451                     include_syntax:
1452                         tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1453                     }
1454                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1455                     next();
1456                 }
1457                 c = '\"';
1458             } else {
1459                 int len;
1460                 while (tok != TOK_LINEFEED) {
1461                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1462                     next();
1463                 }
1464                 len = strlen(buf);
1465                 /* check syntax and remove '<>' */
1466                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1467                     goto include_syntax;
1468                 memmove(buf, buf + 1, len - 2);
1469                 buf[len - 2] = '\0';
1470                 c = '>';
1471             }
1472         }
1473
1474         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1475             tcc_error("#include recursion too deep");
1476         /* store current file in stack, but increment stack later below */
1477         *s1->include_stack_ptr = file;
1478
1479         n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1480         for (i = -2; i < n; ++i) {
1481             char buf1[sizeof file->filename];
1482             CachedInclude *e;
1483             BufferedFile **f;
1484             const char *path;
1485
1486             if (i == -2) {
1487                 /* check absolute include path */
1488                 if (!IS_ABSPATH(buf))
1489                     continue;
1490                 buf1[0] = 0;
1491                 i = n; /* force end loop */
1492
1493             } else if (i == -1) {
1494                 /* search in current dir if "header.h" */
1495                 if (c != '\"')
1496                     continue;
1497                 path = file->filename;
1498                 pstrncpy(buf1, path, tcc_basename(path) - path);
1499
1500             } else {
1501                 /* search in all the include paths */
1502                 if (i < s1->nb_include_paths)
1503                     path = s1->include_paths[i];
1504                 else
1505                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1506                 pstrcpy(buf1, sizeof(buf1), path);
1507                 pstrcat(buf1, sizeof(buf1), "/");
1508             }
1509
1510             pstrcat(buf1, sizeof(buf1), buf);
1511
1512             if (tok == TOK_INCLUDE_NEXT)
1513                 for (f = s1->include_stack_ptr; f >= s1->include_stack; --f)
1514                     if (0 == PATHCMP((*f)->filename, buf1)) {
1515 #ifdef INC_DEBUG
1516                         printf("%s: #include_next skipping %s\n", file->filename, buf1);
1517 #endif
1518                         goto include_trynext;
1519                     }
1520
1521             e = search_cached_include(s1, buf1);
1522             if (e && define_find(e->ifndef_macro)) {
1523                 /* no need to parse the include because the 'ifndef macro'
1524                    is defined */
1525 #ifdef INC_DEBUG
1526                 printf("%s: skipping cached %s\n", file->filename, buf1);
1527 #endif
1528                 goto include_done;
1529             }
1530
1531             if (tcc_open(s1, buf1) < 0)
1532 include_trynext:
1533                 continue;
1534
1535 #ifdef INC_DEBUG
1536             printf("%s: including %s\n", file->prev->filename, file->filename);
1537 #endif
1538             /* update target deps */
1539             dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1540                     tcc_strdup(buf1));
1541             /* push current file in stack */
1542             ++s1->include_stack_ptr;
1543             /* add include file debug info */
1544             if (s1->do_debug)
1545                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1546             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1547             ch = file->buf_ptr[0];
1548             goto the_end;
1549         }
1550         tcc_error("include file '%s' not found", buf);
1551 include_done:
1552         break;
1553     case TOK_IFNDEF:
1554         c = 1;
1555         goto do_ifdef;
1556     case TOK_IF:
1557         c = expr_preprocess();
1558         goto do_if;
1559     case TOK_IFDEF:
1560         c = 0;
1561     do_ifdef:
1562         next_nomacro();
1563         if (tok < TOK_IDENT)
1564             tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1565         if (is_bof) {
1566             if (c) {
1567 #ifdef INC_DEBUG
1568                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1569 #endif
1570                 file->ifndef_macro = tok;
1571             }
1572         }
1573         c = (define_find(tok) != 0) ^ c;
1574     do_if:
1575         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1576             tcc_error("memory full (ifdef)");
1577         *s1->ifdef_stack_ptr++ = c;
1578         goto test_skip;
1579     case TOK_ELSE:
1580         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1581             tcc_error("#else without matching #if");
1582         if (s1->ifdef_stack_ptr[-1] & 2)
1583             tcc_error("#else after #else");
1584         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1585         goto test_else;
1586     case TOK_ELIF:
1587         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1588             tcc_error("#elif without matching #if");
1589         c = s1->ifdef_stack_ptr[-1];
1590         if (c > 1)
1591             tcc_error("#elif after #else");
1592         /* last #if/#elif expression was true: we skip */
1593         if (c == 1)
1594             goto skip;
1595         c = expr_preprocess();
1596         s1->ifdef_stack_ptr[-1] = c;
1597     test_else:
1598         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1599             file->ifndef_macro = 0;
1600     test_skip:
1601         if (!(c & 1)) {
1602         skip:
1603             preprocess_skip();
1604             is_bof = 0;
1605             goto redo;
1606         }
1607         break;
1608     case TOK_ENDIF:
1609         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1610             tcc_error("#endif without matching #if");
1611         s1->ifdef_stack_ptr--;
1612         /* '#ifndef macro' was at the start of file. Now we check if
1613            an '#endif' is exactly at the end of file */
1614         if (file->ifndef_macro &&
1615             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1616             file->ifndef_macro_saved = file->ifndef_macro;
1617             /* need to set to zero to avoid false matches if another
1618                #ifndef at middle of file */
1619             file->ifndef_macro = 0;
1620             while (tok != TOK_LINEFEED)
1621                 next_nomacro();
1622             tok_flags |= TOK_FLAG_ENDIF;
1623             goto the_end;
1624         }
1625         break;
1626     case TOK_LINE:
1627         next();
1628         if (tok != TOK_CINT)
1629             tcc_error("#line");
1630         file->line_num = tokc.i - 1; /* the line number will be incremented after */
1631         next();
1632         if (tok != TOK_LINEFEED) {
1633             if (tok != TOK_STR)
1634                 tcc_error("#line");
1635             pstrcpy(file->filename, sizeof(file->filename),
1636                     (char *)tokc.cstr->data);
1637         }
1638         break;
1639     case TOK_ERROR:
1640     case TOK_WARNING:
1641         c = tok;
1642         ch = file->buf_ptr[0];
1643         skip_spaces();
1644         q = buf;
1645         while (ch != '\n' && ch != CH_EOF) {
1646             if ((q - buf) < sizeof(buf) - 1)
1647                 *q++ = ch;
1648             if (ch == '\\') {
1649                 if (handle_stray_noerror() == 0)
1650                     --q;
1651             } else
1652                 inp();
1653         }
1654         *q = '\0';
1655         if (c == TOK_ERROR)
1656             tcc_error("#error %s", buf);
1657         else
1658             tcc_warning("#warning %s", buf);
1659         break;
1660     case TOK_PRAGMA:
1661         pragma_parse(s1);
1662         break;
1663     default:
1664         if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_PPNUM) {
1665             /* '!' is ignored to allow C scripts. numbers are ignored
1666                to emulate cpp behaviour */
1667         } else {
1668             if (!(saved_parse_flags & PARSE_FLAG_ASM_COMMENTS))
1669                 tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1670             else {
1671                 /* this is a gas line comment in an 'S' file. */
1672                 file->buf_ptr = parse_line_comment(file->buf_ptr);
1673                 goto the_end;
1674             }
1675         }
1676         break;
1677     }
1678     /* ignore other preprocess commands or #! for C scripts */
1679     while (tok != TOK_LINEFEED)
1680         next_nomacro();
1681  the_end:
1682     parse_flags = saved_parse_flags;
1683 }
1684
1685 /* evaluate escape codes in a string. */
1686 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1687 {
1688     int c, n;
1689     const uint8_t *p;
1690
1691     p = buf;
1692     for(;;) {
1693         c = *p;
1694         if (c == '\0')
1695             break;
1696         if (c == '\\') {
1697             p++;
1698             /* escape */
1699             c = *p;
1700             switch(c) {
1701             case '0': case '1': case '2': case '3':
1702             case '4': case '5': case '6': case '7':
1703                 /* at most three octal digits */
1704                 n = c - '0';
1705                 p++;
1706                 c = *p;
1707                 if (isoct(c)) {
1708                     n = n * 8 + c - '0';
1709                     p++;
1710                     c = *p;
1711                     if (isoct(c)) {
1712                         n = n * 8 + c - '0';
1713                         p++;
1714                     }
1715                 }
1716                 c = n;
1717                 goto add_char_nonext;
1718             case 'x':
1719             case 'u':
1720             case 'U':
1721                 p++;
1722                 n = 0;
1723                 for(;;) {
1724                     c = *p;
1725                     if (c >= 'a' && c <= 'f')
1726                         c = c - 'a' + 10;
1727                     else if (c >= 'A' && c <= 'F')
1728                         c = c - 'A' + 10;
1729                     else if (isnum(c))
1730                         c = c - '0';
1731                     else
1732                         break;
1733                     n = n * 16 + c;
1734                     p++;
1735                 }
1736                 c = n;
1737                 goto add_char_nonext;
1738             case 'a':
1739                 c = '\a';
1740                 break;
1741             case 'b':
1742                 c = '\b';
1743                 break;
1744             case 'f':
1745                 c = '\f';
1746                 break;
1747             case 'n':
1748                 c = '\n';
1749                 break;
1750             case 'r':
1751                 c = '\r';
1752                 break;
1753             case 't':
1754                 c = '\t';
1755                 break;
1756             case 'v':
1757                 c = '\v';
1758                 break;
1759             case 'e':
1760                 if (!gnu_ext)
1761                     goto invalid_escape;
1762                 c = 27;
1763                 break;
1764             case '\'':
1765             case '\"':
1766             case '\\':
1767             case '?':
1768                 break;
1769             default:
1770             invalid_escape:
1771                 if (c >= '!' && c <= '~')
1772                     tcc_warning("unknown escape sequence: \'\\%c\'", c);
1773                 else
1774                     tcc_warning("unknown escape sequence: \'\\x%x\'", c);
1775                 break;
1776             }
1777         }
1778         p++;
1779     add_char_nonext:
1780         if (!is_long)
1781             cstr_ccat(outstr, c);
1782         else
1783             cstr_wccat(outstr, c);
1784     }
1785     /* add a trailing '\0' */
1786     if (!is_long)
1787         cstr_ccat(outstr, '\0');
1788     else
1789         cstr_wccat(outstr, '\0');
1790 }
1791
1792 /* parse number in null terminated string 'p' and return it in the
1793    current token */
1794 static void parse_number(const char *p)
1795 {
1796     int b, t, c;
1797
1798     c = *p++;
1799         t = *p++;
1800     b = 10;
1801         if(c=='.'){
1802                 --p;
1803                 goto float_frac_parse;
1804         }
1805         if(c == '0'){
1806                 if (t == 'x' || t == 'X') {
1807             b = 16;
1808                         c = *p++;
1809         } else if (tcc_ext && (t == 'b' || t == 'B')) {
1810             b = 2;
1811                         c = *p++;
1812         }else{
1813                         --p;
1814                 }
1815         }else
1816                 --p;
1817     if(strchr(p , '.') || (b == 10 && (strchr(p,'e') || strchr(p,'E'))) ||
1818                 ((b == 2 || b == 16)&& (strchr(p,'p') || strchr(p,'P')))){
1819                 long double ld, sh, fb;
1820                 int exp;
1821                 /* NOTE: strtox should support that for hexa numbers, but
1822                         non ISOC99 libcs do not support it, so we prefer to do
1823                         it by hand */
1824                 /* hexadecimal or binary floats */
1825                 /* XXX: handle overflows */
1826 float_frac_parse:
1827                 fb = 1.0L/b;
1828                 sh = b;
1829                 ld = 0.0;
1830
1831                 while(1){
1832                         if (c == '\0')
1833                                 break;
1834                         if (c >= 'a' && c <= 'f')
1835                                 t = c - 'a' + 10;
1836                         else if (c >= 'A' && c <= 'F')
1837                                 t = c - 'A' + 10;
1838                         else if(isnum(c))
1839                                 t = c - '0';
1840                         else
1841                                 break;
1842                         if (t >= b)
1843                                 tcc_error("invalid digit");
1844                         ld = ld * b + t;
1845                         c = *p++;
1846                 }
1847                 if (c == '.'){
1848                         c = *p++;
1849                         sh = fb;
1850                         while (1){
1851                                 if (c == '\0')
1852                                         break;
1853                                 if (c >= 'a' && c <= 'f')
1854                                         t = c - 'a' + 10;
1855                                 else if (c >= 'A' && c <= 'F')
1856                                         t = c - 'A' + 10;
1857                                 else if (isnum(c))
1858                                         t =c - '0';
1859                                 else
1860                                         break;
1861                                 if (t >= b){
1862                                         if(b == 10 && (c == 'e' || c == 'E' || c == 'f' || c == 'F'))
1863                                                 break;
1864                                         tcc_error("invalid digit");
1865                                 }
1866                                 ld += sh*t;
1867                                 sh*=fb;
1868                                 c = *p++;
1869                         }
1870                 }
1871                 if ((b == 16 || b == 2) && c != 'p' && c != 'P')
1872                 expect("exponent");
1873                 if(((c == 'e' || c == 'E') && b == 10) ||
1874                         ((c == 'p' || c == 'P') && (b == 16 || b == 2))){
1875                         c = *p++;
1876                         if(c == '+' || c == '-'){
1877                                 if (c == '-')
1878                                         sh = fb;
1879                                 c = *p++;
1880                         }else
1881                                 sh = b;
1882                         if (!isnum(c))
1883                                 expect("exponent digits");
1884                         exp = 0;
1885                         do{
1886                                 exp = exp * 10 + c - '0';
1887                                 c = *p++;
1888                         }while(isnum(c));
1889                         while (exp != 0){
1890                                 if (exp & 1)
1891                                         ld *= sh;
1892                                 exp >>= 1;
1893                                 sh *= sh;
1894                         }
1895                 }
1896                 t = toup(c);
1897                 if (t == 'F') {
1898                         c = *p++;
1899                         tok = TOK_CFLOAT;
1900                         tokc.f = (float)ld;
1901                 } else if (t == 'L') {
1902                         c = *p++;
1903 #ifdef TCC_TARGET_PE
1904                         tok = TOK_CDOUBLE;
1905                         tokc.d = (double)ld;
1906 #else
1907                         tok = TOK_CLDOUBLE;
1908                         tokc.ld = ld;
1909 #endif
1910                 } else {
1911                         tok = TOK_CDOUBLE;
1912                         tokc.d = (double)ld;
1913                 }
1914     } else {
1915                 uint64_t n = 0, n1;
1916                 int warn = 1;
1917         int lcount, ucount;
1918                 if (b == 10 && c == '0') {
1919             b = 8;
1920         }
1921                 while(1){
1922                         if (c == '\0')
1923                                 break;
1924                         if (c >= 'a' && c <= 'f')
1925                                 t = c - 'a' + 10;
1926                         else if (c >= 'A' && c <= 'F')
1927                                 t = c - 'A' + 10;
1928                         else if(isnum(c))
1929                                 t = c - '0';
1930                         else
1931                                 break;
1932                         if (t >= b)
1933                                 tcc_error("invalid digit");
1934                         n1 = n;
1935                         n = n * b + t;
1936                         if (n < n1 && warn){
1937                                 tcc_warning("integer constant overflow");
1938                                 warn = 0;
1939                         }
1940                         c = *p++;
1941                 }
1942         /* XXX: not exactly ANSI compliant */
1943         if ((n & 0xffffffff00000000LL) != 0) {
1944             if ((n >> 63) != 0)
1945                 tok = TOK_CULLONG;
1946             else
1947                 tok = TOK_CLLONG;
1948         } else if (n > 0x7fffffff) {
1949             tok = TOK_CUINT;
1950         } else {
1951             tok = TOK_CINT;
1952         }
1953         lcount = 0;
1954         ucount = 0;
1955         for(;;) {
1956             t = toup(c);
1957             if (t == 'L') {
1958                 if (lcount >= 2)
1959                     tcc_error("three 'l's in integer constant");
1960                 lcount++;
1961 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
1962                 if (lcount == 2) {
1963 #endif
1964                     if (tok == TOK_CINT)
1965                         tok = TOK_CLLONG;
1966                     else if (tok == TOK_CUINT)
1967                         tok = TOK_CULLONG;
1968 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
1969                 }
1970 #endif
1971                 c = *p++;
1972             } else if (t == 'U') {
1973                 if (ucount >= 1)
1974                     tcc_error("two 'u's in integer constant");
1975                 ucount++;
1976                 if (tok == TOK_CINT)
1977                     tok = TOK_CUINT;
1978                 else if (tok == TOK_CLLONG)
1979                     tok = TOK_CULLONG;
1980                 c = *p++;
1981             } else {
1982                 break;
1983             }
1984         }
1985         if (tok == TOK_CINT || tok == TOK_CUINT)
1986             tokc.ui = n;
1987         else
1988             tokc.ull = n;
1989     }
1990     if (c)
1991         tcc_error("invalid number\n");
1992 }
1993
1994
1995 #define PARSE2(c1, tok1, c2, tok2)              \
1996     case c1:                                    \
1997         PEEKC(c, p);                            \
1998         if (c == c2) {                          \
1999             p++;                                \
2000             tok = tok2;                         \
2001         } else {                                \
2002             tok = tok1;                         \
2003         }                                       \
2004         break;
2005
2006 /* return next token without macro substitution */
2007 static inline void next_nomacro1(void)
2008 {
2009     int t, c, is_long;
2010     TokenSym *ts;
2011     uint8_t *p, *p1;
2012     unsigned int h;
2013
2014     p = file->buf_ptr;
2015  redo_no_start:
2016     c = *p;
2017     switch(c) {
2018     case ' ':
2019     case '\t':
2020         tok = c;
2021         p++;
2022         goto keep_tok_flags;
2023     case '\f':
2024     case '\v':
2025     case '\r':
2026         p++;
2027         goto redo_no_start;
2028     case '\\':
2029         /* first look if it is in fact an end of buffer */
2030         if (p >= file->buf_end) {
2031             file->buf_ptr = p;
2032             handle_eob();
2033             p = file->buf_ptr;
2034             if (p >= file->buf_end)
2035                 goto parse_eof;
2036             else
2037                 goto redo_no_start;
2038         } else {
2039             file->buf_ptr = p;
2040             ch = *p;
2041             handle_stray();
2042             p = file->buf_ptr;
2043             goto redo_no_start;
2044         }
2045     parse_eof:
2046         {
2047             TCCState *s1 = tcc_state;
2048             if ((parse_flags & PARSE_FLAG_LINEFEED)
2049                 && !(tok_flags & TOK_FLAG_EOF)) {
2050                 tok_flags |= TOK_FLAG_EOF;
2051                 tok = TOK_LINEFEED;
2052                 goto keep_tok_flags;
2053             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2054                 tok = TOK_EOF;
2055             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2056                 tcc_error("missing #endif");
2057             } else if (s1->include_stack_ptr == s1->include_stack) {
2058                 /* no include left : end of file. */
2059                 tok = TOK_EOF;
2060             } else {
2061                 tok_flags &= ~TOK_FLAG_EOF;
2062                 /* pop include file */
2063
2064                 /* test if previous '#endif' was after a #ifdef at
2065                    start of file */
2066                 if (tok_flags & TOK_FLAG_ENDIF) {
2067 #ifdef INC_DEBUG
2068                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2069 #endif
2070                     add_cached_include(s1, file->filename, file->ifndef_macro_saved);
2071                     tok_flags &= ~TOK_FLAG_ENDIF;
2072                 }
2073
2074                 /* add end of include file debug info */
2075                 if (tcc_state->do_debug) {
2076                     put_stabd(N_EINCL, 0, 0);
2077                 }
2078                 /* pop include stack */
2079                 tcc_close();
2080                 s1->include_stack_ptr--;
2081                 p = file->buf_ptr;
2082                 goto redo_no_start;
2083             }
2084         }
2085         break;
2086
2087     case '\n':
2088         file->line_num++;
2089         tok_flags |= TOK_FLAG_BOL;
2090         p++;
2091 maybe_newline:
2092         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2093             goto redo_no_start;
2094         tok = TOK_LINEFEED;
2095         goto keep_tok_flags;
2096
2097     case '#':
2098         /* XXX: simplify */
2099         PEEKC(c, p);
2100         if ((tok_flags & TOK_FLAG_BOL) &&
2101             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2102             file->buf_ptr = p;
2103             preprocess(tok_flags & TOK_FLAG_BOF);
2104             p = file->buf_ptr;
2105             goto maybe_newline;
2106         } else {
2107             if (c == '#') {
2108                 p++;
2109                 tok = TOK_TWOSHARPS;
2110             } else {
2111                 if (parse_flags & PARSE_FLAG_ASM_COMMENTS) {
2112                     p = parse_line_comment(p - 1);
2113                     goto redo_no_start;
2114                 } else {
2115                     tok = '#';
2116                 }
2117             }
2118         }
2119         break;
2120
2121     case 'a': case 'b': case 'c': case 'd':
2122     case 'e': case 'f': case 'g': case 'h':
2123     case 'i': case 'j': case 'k': case 'l':
2124     case 'm': case 'n': case 'o': case 'p':
2125     case 'q': case 'r': case 's': case 't':
2126     case 'u': case 'v': case 'w': case 'x':
2127     case 'y': case 'z':
2128     case 'A': case 'B': case 'C': case 'D':
2129     case 'E': case 'F': case 'G': case 'H':
2130     case 'I': case 'J': case 'K':
2131     case 'M': case 'N': case 'O': case 'P':
2132     case 'Q': case 'R': case 'S': case 'T':
2133     case 'U': case 'V': case 'W': case 'X':
2134     case 'Y': case 'Z':
2135     case '_':
2136     parse_ident_fast:
2137         p1 = p;
2138         h = TOK_HASH_INIT;
2139         h = TOK_HASH_FUNC(h, c);
2140         p++;
2141         for(;;) {
2142             c = *p;
2143             if (!isidnum_table[c-CH_EOF])
2144                 break;
2145             h = TOK_HASH_FUNC(h, c);
2146             p++;
2147         }
2148         if (c != '\\') {
2149             TokenSym **pts;
2150             int len;
2151
2152             /* fast case : no stray found, so we have the full token
2153                and we have already hashed it */
2154             len = p - p1;
2155             h &= (TOK_HASH_SIZE - 1);
2156             pts = &hash_ident[h];
2157             for(;;) {
2158                 ts = *pts;
2159                 if (!ts)
2160                     break;
2161                 if (ts->len == len && !memcmp(ts->str, p1, len))
2162                     goto token_found;
2163                 pts = &(ts->hash_next);
2164             }
2165             ts = tok_alloc_new(pts, (char *) p1, len);
2166         token_found: ;
2167         } else {
2168             /* slower case */
2169             cstr_reset(&tokcstr);
2170
2171             while (p1 < p) {
2172                 cstr_ccat(&tokcstr, *p1);
2173                 p1++;
2174             }
2175             p--;
2176             PEEKC(c, p);
2177         parse_ident_slow:
2178             while (isidnum_table[c-CH_EOF]) {
2179                 cstr_ccat(&tokcstr, c);
2180                 PEEKC(c, p);
2181             }
2182             ts = tok_alloc(tokcstr.data, tokcstr.size);
2183         }
2184         tok = ts->tok;
2185         break;
2186     case 'L':
2187         t = p[1];
2188         if (t != '\\' && t != '\'' && t != '\"') {
2189             /* fast case */
2190             goto parse_ident_fast;
2191         } else {
2192             PEEKC(c, p);
2193             if (c == '\'' || c == '\"') {
2194                 is_long = 1;
2195                 goto str_const;
2196             } else {
2197                 cstr_reset(&tokcstr);
2198                 cstr_ccat(&tokcstr, 'L');
2199                 goto parse_ident_slow;
2200             }
2201         }
2202         break;
2203     case '0': case '1': case '2': case '3':
2204     case '4': case '5': case '6': case '7':
2205     case '8': case '9':
2206
2207         cstr_reset(&tokcstr);
2208         /* after the first digit, accept digits, alpha, '.' or sign if
2209            prefixed by 'eEpP' */
2210     parse_num:
2211         for(;;) {
2212             t = c;
2213             cstr_ccat(&tokcstr, c);
2214             PEEKC(c, p);
2215             if (!(isnum(c) || isid(c) || c == '.' ||
2216                   ((c == '+' || c == '-') &&
2217                    (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2218                 break;
2219         }
2220         /* We add a trailing '\0' to ease parsing */
2221         cstr_ccat(&tokcstr, '\0');
2222         tokc.cstr = &tokcstr;
2223         tok = TOK_PPNUM;
2224         break;
2225     case '.':
2226         /* special dot handling because it can also start a number */
2227         PEEKC(c, p);
2228         if (isnum(c)) {
2229             cstr_reset(&tokcstr);
2230             cstr_ccat(&tokcstr, '.');
2231             goto parse_num;
2232         } else if (c == '.') {
2233             PEEKC(c, p);
2234             if (c != '.')
2235                 expect("'.'");
2236             PEEKC(c, p);
2237             tok = TOK_DOTS;
2238         } else {
2239             tok = '.';
2240         }
2241         break;
2242     case '\'':
2243     case '\"':
2244         is_long = 0;
2245     str_const:
2246         {
2247             CString str;
2248             int sep;
2249
2250             sep = c;
2251
2252             /* parse the string */
2253             cstr_new(&str);
2254             p = parse_pp_string(p, sep, &str);
2255             cstr_ccat(&str, '\0');
2256
2257             /* eval the escape (should be done as TOK_PPNUM) */
2258             cstr_reset(&tokcstr);
2259             parse_escape_string(&tokcstr, str.data, is_long);
2260             cstr_free(&str);
2261
2262             if (sep == '\'') {
2263                 int char_size;
2264                 /* XXX: make it portable */
2265                 if (!is_long)
2266                     char_size = 1;
2267                 else
2268                     char_size = sizeof(nwchar_t);
2269                 if (tokcstr.size <= char_size)
2270                     tcc_error("empty character constant");
2271                 if (tokcstr.size > 2 * char_size)
2272                     tcc_warning("multi-character character constant");
2273                 if (!is_long) {
2274                     tokc.i = *(int8_t *)tokcstr.data;
2275                     tok = TOK_CCHAR;
2276                 } else {
2277                     tokc.i = *(nwchar_t *)tokcstr.data;
2278                     tok = TOK_LCHAR;
2279                 }
2280             } else {
2281                 tokc.cstr = &tokcstr;
2282                 if (!is_long)
2283                     tok = TOK_STR;
2284                 else
2285                     tok = TOK_LSTR;
2286             }
2287         }
2288         break;
2289
2290     case '<':
2291         PEEKC(c, p);
2292         if (c == '=') {
2293             p++;
2294             tok = TOK_LE;
2295         } else if (c == '<') {
2296             PEEKC(c, p);
2297             if (c == '=') {
2298                 p++;
2299                 tok = TOK_A_SHL;
2300             } else {
2301                 tok = TOK_SHL;
2302             }
2303         } else {
2304             tok = TOK_LT;
2305         }
2306         break;
2307
2308     case '>':
2309         PEEKC(c, p);
2310         if (c == '=') {
2311             p++;
2312             tok = TOK_GE;
2313         } else if (c == '>') {
2314             PEEKC(c, p);
2315             if (c == '=') {
2316                 p++;
2317                 tok = TOK_A_SAR;
2318             } else {
2319                 tok = TOK_SAR;
2320             }
2321         } else {
2322             tok = TOK_GT;
2323         }
2324         break;
2325
2326     case '&':
2327         PEEKC(c, p);
2328         if (c == '&') {
2329             p++;
2330             tok = TOK_LAND;
2331         } else if (c == '=') {
2332             p++;
2333             tok = TOK_A_AND;
2334         } else {
2335             tok = '&';
2336         }
2337         break;
2338
2339     case '|':
2340         PEEKC(c, p);
2341         if (c == '|') {
2342             p++;
2343             tok = TOK_LOR;
2344         } else if (c == '=') {
2345             p++;
2346             tok = TOK_A_OR;
2347         } else {
2348             tok = '|';
2349         }
2350         break;
2351
2352     case '+':
2353         PEEKC(c, p);
2354         if (c == '+') {
2355             p++;
2356             tok = TOK_INC;
2357         } else if (c == '=') {
2358             p++;
2359             tok = TOK_A_ADD;
2360         } else {
2361             tok = '+';
2362         }
2363         break;
2364
2365     case '-':
2366         PEEKC(c, p);
2367         if (c == '-') {
2368             p++;
2369             tok = TOK_DEC;
2370         } else if (c == '=') {
2371             p++;
2372             tok = TOK_A_SUB;
2373         } else if (c == '>') {
2374             p++;
2375             tok = TOK_ARROW;
2376         } else {
2377             tok = '-';
2378         }
2379         break;
2380
2381     PARSE2('!', '!', '=', TOK_NE)
2382     PARSE2('=', '=', '=', TOK_EQ)
2383     PARSE2('*', '*', '=', TOK_A_MUL)
2384     PARSE2('%', '%', '=', TOK_A_MOD)
2385     PARSE2('^', '^', '=', TOK_A_XOR)
2386
2387         /* comments or operator */
2388     case '/':
2389         PEEKC(c, p);
2390         if (c == '*') {
2391             p = parse_comment(p);
2392             /* comments replaced by a blank */
2393             tok = ' ';
2394             goto keep_tok_flags;
2395         } else if (c == '/') {
2396             p = parse_line_comment(p);
2397             tok = ' ';
2398             goto keep_tok_flags;
2399         } else if (c == '=') {
2400             p++;
2401             tok = TOK_A_DIV;
2402         } else {
2403             tok = '/';
2404         }
2405         break;
2406
2407         /* simple tokens */
2408     case '(':
2409     case ')':
2410     case '[':
2411     case ']':
2412     case '{':
2413     case '}':
2414     case ',':
2415     case ';':
2416     case ':':
2417     case '?':
2418     case '~':
2419     case '$': /* only used in assembler */
2420     case '@': /* dito */
2421         tok = c;
2422         p++;
2423         break;
2424     default:
2425         tcc_error("unrecognized character \\x%02x", c);
2426         break;
2427     }
2428     tok_flags = 0;
2429 keep_tok_flags:
2430     file->buf_ptr = p;
2431 #if defined(PARSE_DEBUG)
2432     printf("token = %s\n", get_tok_str(tok, &tokc));
2433 #endif
2434 }
2435
2436 /* return next token without macro substitution. Can read input from
2437    macro_ptr buffer */
2438 static void next_nomacro_spc(void)
2439 {
2440     if (macro_ptr) {
2441     redo:
2442         tok = *macro_ptr;
2443         if (tok) {
2444             TOK_GET(&tok, &macro_ptr, &tokc);
2445             if (tok == TOK_LINENUM) {
2446                 file->line_num = tokc.i;
2447                 goto redo;
2448             }
2449         }
2450     } else {
2451         next_nomacro1();
2452     }
2453 }
2454
2455 ST_FUNC void next_nomacro(void)
2456 {
2457     do {
2458         next_nomacro_spc();
2459     } while (is_space(tok));
2460 }
2461
2462 /* substitute arguments in replacement lists in macro_str by the values in
2463    args (field d) and return allocated string */
2464 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2465 {
2466     int last_tok, t, spc;
2467     const int *st;
2468     Sym *s;
2469     CValue cval;
2470     TokenString str;
2471     CString cstr;
2472
2473     tok_str_new(&str);
2474     last_tok = 0;
2475     while(1) {
2476         TOK_GET(&t, &macro_str, &cval);
2477         if (!t)
2478             break;
2479         if (t == '#') {
2480             /* stringize */
2481             TOK_GET(&t, &macro_str, &cval);
2482             if (!t)
2483                 break;
2484             s = sym_find2(args, t);
2485             if (s) {
2486                 cstr_new(&cstr);
2487                 st = s->d;
2488                 spc = 0;
2489                 while (*st) {
2490                     TOK_GET(&t, &st, &cval);
2491                     if (!check_space(t, &spc))
2492                         cstr_cat(&cstr, get_tok_str(t, &cval));
2493                 }
2494                 cstr.size -= spc;
2495                 cstr_ccat(&cstr, '\0');
2496 #ifdef PP_DEBUG
2497                 printf("stringize: %s\n", (char *)cstr.data);
2498 #endif
2499                 /* add string */
2500                 cval.cstr = &cstr;
2501                 tok_str_add2(&str, TOK_STR, &cval);
2502                 cstr_free(&cstr);
2503             } else {
2504                 tok_str_add2(&str, t, &cval);
2505             }
2506         } else if (t >= TOK_IDENT) {
2507             s = sym_find2(args, t);
2508             if (s) {
2509                 st = s->d;
2510                 /* if '##' is present before or after, no arg substitution */
2511                 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2512                     /* special case for var arg macros : ## eats the
2513                        ',' if empty VA_ARGS variable. */
2514                     /* XXX: test of the ',' is not 100%
2515                        reliable. should fix it to avoid security
2516                        problems */
2517                     if (gnu_ext && s->type.t &&
2518                         last_tok == TOK_TWOSHARPS &&
2519                         str.len >= 2 && str.str[str.len - 2] == ',') {
2520                         if (*st == TOK_PLCHLDR) {
2521                             /* suppress ',' '##' */
2522                             str.len -= 2;
2523                         } else {
2524                             /* suppress '##' and add variable */
2525                             str.len--;
2526                             goto add_var;
2527                         }
2528                     } else {
2529                         int t1;
2530                     add_var:
2531                         for(;;) {
2532                             TOK_GET(&t1, &st, &cval);
2533                             if (!t1)
2534                                 break;
2535                             tok_str_add2(&str, t1, &cval);
2536                         }
2537                     }
2538                 } else {
2539                     /* NOTE: the stream cannot be read when macro
2540                        substituing an argument */
2541                     macro_subst(&str, nested_list, st, NULL);
2542                 }
2543             } else {
2544                 tok_str_add(&str, t);
2545             }
2546         } else {
2547             tok_str_add2(&str, t, &cval);
2548         }
2549         last_tok = t;
2550     }
2551     tok_str_add(&str, 0);
2552     return str.str;
2553 }
2554
2555 static char const ab_month_name[12][4] =
2556 {
2557     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2558     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2559 };
2560
2561 /* do macro substitution of current token with macro 's' and add
2562    result to (tok_str,tok_len). 'nested_list' is the list of all
2563    macros we got inside to avoid recursing. Return non zero if no
2564    substitution needs to be done */
2565 static int macro_subst_tok(TokenString *tok_str,
2566                            Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2567 {
2568     Sym *args, *sa, *sa1;
2569     int mstr_allocated, parlevel, *mstr, t, t1, spc;
2570     const int *p;
2571     TokenString str;
2572     char *cstrval;
2573     CValue cval;
2574     CString cstr;
2575     char buf[32];
2576
2577     /* if symbol is a macro, prepare substitution */
2578     /* special macros */
2579     if (tok == TOK___LINE__) {
2580         snprintf(buf, sizeof(buf), "%d", file->line_num);
2581         cstrval = buf;
2582         t1 = TOK_PPNUM;
2583         goto add_cstr1;
2584     } else if (tok == TOK___FILE__) {
2585         cstrval = file->filename;
2586         goto add_cstr;
2587     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2588         time_t ti;
2589         struct tm *tm;
2590
2591         time(&ti);
2592         tm = localtime(&ti);
2593         if (tok == TOK___DATE__) {
2594             snprintf(buf, sizeof(buf), "%s %2d %d",
2595                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2596         } else {
2597             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2598                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2599         }
2600         cstrval = buf;
2601     add_cstr:
2602         t1 = TOK_STR;
2603     add_cstr1:
2604         cstr_new(&cstr);
2605         cstr_cat(&cstr, cstrval);
2606         cstr_ccat(&cstr, '\0');
2607         cval.cstr = &cstr;
2608         tok_str_add2(tok_str, t1, &cval);
2609         cstr_free(&cstr);
2610     } else {
2611         mstr = s->d;
2612         mstr_allocated = 0;
2613         if (s->type.t == MACRO_FUNC) {
2614             /* NOTE: we do not use next_nomacro to avoid eating the
2615                next token. XXX: find better solution */
2616         redo:
2617             if (macro_ptr) {
2618                 p = macro_ptr;
2619                 while (is_space(t = *p) || TOK_LINEFEED == t)
2620                     ++p;
2621                 if (t == 0 && can_read_stream) {
2622                     /* end of macro stream: we must look at the token
2623                        after in the file */
2624                     struct macro_level *ml = *can_read_stream;
2625                     macro_ptr = NULL;
2626                     if (ml)
2627                     {
2628                         macro_ptr = ml->p;
2629                         ml->p = NULL;
2630                         *can_read_stream = ml -> prev;
2631                     }
2632                     /* also, end of scope for nested defined symbol */
2633                     (*nested_list)->v = -1;
2634                     goto redo;
2635                 }
2636             } else {
2637                 ch = file->buf_ptr[0];
2638                 while (is_space(ch) || ch == '\n' || ch == '/')
2639                   {
2640                     if (ch == '/')
2641                       {
2642                         int c;
2643                         uint8_t *p = file->buf_ptr;
2644                         PEEKC(c, p);
2645                         if (c == '*') {
2646                             p = parse_comment(p);
2647                             file->buf_ptr = p - 1;
2648                         } else if (c == '/') {
2649                             p = parse_line_comment(p);
2650                             file->buf_ptr = p - 1;
2651                         } else
2652                           break;
2653                       }
2654                     cinp();
2655                   }
2656                 t = ch;
2657             }
2658             if (t != '(') /* no macro subst */
2659                 return -1;
2660
2661             /* argument macro */
2662             next_nomacro();
2663             next_nomacro();
2664             args = NULL;
2665             sa = s->next;
2666             /* NOTE: empty args are allowed, except if no args */
2667             for(;;) {
2668                 /* handle '()' case */
2669                 if (!args && !sa && tok == ')')
2670                     break;
2671                 if (!sa)
2672                     tcc_error("macro '%s' used with too many args",
2673                           get_tok_str(s->v, 0));
2674                 tok_str_new(&str);
2675                 parlevel = spc = 0;
2676                 /* NOTE: non zero sa->t indicates VA_ARGS */
2677                 while ((parlevel > 0 ||
2678                         (tok != ')' &&
2679                          (tok != ',' || sa->type.t))) &&
2680                        tok != -1) {
2681                     if (tok == '(')
2682                         parlevel++;
2683                     else if (tok == ')')
2684                         parlevel--;
2685                     if (tok == TOK_LINEFEED)
2686                         tok = ' ';
2687                     if (!check_space(tok, &spc))
2688                         tok_str_add2(&str, tok, &tokc);
2689                     next_nomacro_spc();
2690                 }
2691                 if (!str.len)
2692                     tok_str_add(&str, TOK_PLCHLDR);
2693                 str.len -= spc;
2694                 tok_str_add(&str, 0);
2695                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
2696                 sa1->d = str.str;
2697                 sa = sa->next;
2698                 if (tok == ')') {
2699                     /* special case for gcc var args: add an empty
2700                        var arg argument if it is omitted */
2701                     if (sa && sa->type.t && gnu_ext)
2702                         continue;
2703                     else
2704                         break;
2705                 }
2706                 if (tok != ',')
2707                     expect(",");
2708                 next_nomacro();
2709             }
2710             if (sa) {
2711                 tcc_error("macro '%s' used with too few args",
2712                       get_tok_str(s->v, 0));
2713             }
2714
2715             /* now subst each arg */
2716             mstr = macro_arg_subst(nested_list, mstr, args);
2717             /* free memory */
2718             sa = args;
2719             while (sa) {
2720                 sa1 = sa->prev;
2721                 tok_str_free(sa->d);
2722                 sym_free(sa);
2723                 sa = sa1;
2724             }
2725             mstr_allocated = 1;
2726         }
2727         sym_push2(nested_list, s->v, 0, 0);
2728         macro_subst(tok_str, nested_list, mstr, can_read_stream);
2729         /* pop nested defined symbol */
2730         sa1 = *nested_list;
2731         *nested_list = sa1->prev;
2732         sym_free(sa1);
2733         if (mstr_allocated)
2734             tok_str_free(mstr);
2735     }
2736     return 0;
2737 }
2738
2739 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2740    return the resulting string (which must be freed). */
2741 static inline int *macro_twosharps(const int *macro_str)
2742 {
2743     const int *ptr;
2744     int t;
2745     TokenString macro_str1;
2746     CString cstr;
2747     int n, start_of_nosubsts;
2748
2749     /* we search the first '##' */
2750     for(ptr = macro_str;;) {
2751         CValue cval;
2752         TOK_GET(&t, &ptr, &cval);
2753         if (t == TOK_TWOSHARPS)
2754             break;
2755         /* nothing more to do if end of string */
2756         if (t == 0)
2757             return NULL;
2758     }
2759
2760     /* we saw '##', so we need more processing to handle it */
2761     start_of_nosubsts = -1;
2762     tok_str_new(&macro_str1);
2763     for(ptr = macro_str;;) {
2764         TOK_GET(&tok, &ptr, &tokc);
2765         if (tok == 0)
2766             break;
2767         if (tok == TOK_TWOSHARPS)
2768             continue;
2769         if (tok == TOK_NOSUBST && start_of_nosubsts < 0)
2770             start_of_nosubsts = macro_str1.len;
2771         while (*ptr == TOK_TWOSHARPS) {
2772             /* given 'a##b', remove nosubsts preceding 'a' */
2773             if (start_of_nosubsts >= 0)
2774                 macro_str1.len = start_of_nosubsts;
2775             /* given 'a##b', skip '##' */
2776             t = *++ptr;
2777             /* given 'a##b', remove nosubsts preceding 'b' */
2778             while (t == TOK_NOSUBST)
2779                 t = *++ptr;
2780             if (t && t != TOK_TWOSHARPS) {
2781                 CValue cval;
2782                 TOK_GET(&t, &ptr, &cval);
2783                 /* We concatenate the two tokens */
2784                 cstr_new(&cstr);
2785                 if (tok != TOK_PLCHLDR)
2786                     cstr_cat(&cstr, get_tok_str(tok, &tokc));
2787                 n = cstr.size;
2788                 if (t != TOK_PLCHLDR || tok == TOK_PLCHLDR)
2789                     cstr_cat(&cstr, get_tok_str(t, &cval));
2790                 cstr_ccat(&cstr, '\0');
2791
2792                 tcc_open_bf(tcc_state, ":paste:", cstr.size);
2793                 memcpy(file->buffer, cstr.data, cstr.size);
2794                 for (;;) {
2795                     next_nomacro1();
2796                     if (0 == *file->buf_ptr)
2797                         break;
2798                     tok_str_add2(&macro_str1, tok, &tokc);
2799                     tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
2800                         n, cstr.data, (char*)cstr.data + n);
2801                 }
2802                 tcc_close();
2803                 cstr_free(&cstr);
2804             }
2805         }
2806         if (tok != TOK_NOSUBST) {
2807             tok_str_add2(&macro_str1, tok, &tokc);
2808             tok = ' ';
2809             start_of_nosubsts = -1;
2810         }
2811         tok_str_add2(&macro_str1, tok, &tokc);
2812     }
2813     tok_str_add(&macro_str1, 0);
2814     return macro_str1.str;
2815 }
2816
2817
2818 /* do macro substitution of macro_str and add result to
2819    (tok_str,tok_len). 'nested_list' is the list of all macros we got
2820    inside to avoid recursing. */
2821 static void macro_subst(TokenString *tok_str, Sym **nested_list,
2822                         const int *macro_str, struct macro_level ** can_read_stream)
2823 {
2824     Sym *s;
2825     int *macro_str1;
2826     const int *ptr;
2827     int t, ret, spc;
2828     CValue cval;
2829     struct macro_level ml;
2830     int force_blank;
2831
2832     /* first scan for '##' operator handling */
2833     ptr = macro_str;
2834     macro_str1 = macro_twosharps(ptr);
2835
2836     if (macro_str1)
2837         ptr = macro_str1;
2838     spc = 0;
2839     force_blank = 0;
2840
2841     while (1) {
2842         /* NOTE: ptr == NULL can only happen if tokens are read from
2843            file stream due to a macro function call */
2844         if (ptr == NULL)
2845             break;
2846         TOK_GET(&t, &ptr, &cval);
2847         if (t == 0)
2848             break;
2849         if (t == TOK_NOSUBST) {
2850             /* following token has already been subst'd. just copy it on */
2851             tok_str_add2(tok_str, TOK_NOSUBST, NULL);
2852             TOK_GET(&t, &ptr, &cval);
2853             goto no_subst;
2854         }
2855         s = define_find(t);
2856         if (s != NULL) {
2857             /* if nested substitution, do nothing */
2858             if (sym_find2(*nested_list, t)) {
2859                 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
2860                 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
2861                 goto no_subst;
2862             }
2863             ml.p = macro_ptr;
2864             if (can_read_stream)
2865                 ml.prev = *can_read_stream, *can_read_stream = &ml;
2866             macro_ptr = (int *)ptr;
2867             tok = t;
2868             ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
2869             ptr = (int *)macro_ptr;
2870             macro_ptr = ml.p;
2871             if (can_read_stream && *can_read_stream == &ml)
2872                 *can_read_stream = ml.prev;
2873             if (ret != 0)
2874                 goto no_subst;
2875             if (parse_flags & PARSE_FLAG_SPACES)
2876                 force_blank = 1;
2877         } else {
2878         no_subst:
2879             if (force_blank) {
2880                 tok_str_add(tok_str, ' ');
2881                 spc = 1;
2882                 force_blank = 0;
2883             }
2884             if (!check_space(t, &spc))
2885                 tok_str_add2(tok_str, t, &cval);
2886         }
2887     }
2888     if (macro_str1)
2889         tok_str_free(macro_str1);
2890 }
2891
2892 /* return next token with macro substitution */
2893 ST_FUNC void next(void)
2894 {
2895     Sym *nested_list, *s;
2896     TokenString str;
2897     struct macro_level *ml;
2898
2899  redo:
2900     if (parse_flags & PARSE_FLAG_SPACES)
2901         next_nomacro_spc();
2902     else
2903         next_nomacro();
2904     if (!macro_ptr) {
2905         /* if not reading from macro substituted string, then try
2906            to substitute macros */
2907         if (tok >= TOK_IDENT &&
2908             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2909             s = define_find(tok);
2910             if (s) {
2911                 /* we have a macro: we try to substitute */
2912                 tok_str_new(&str);
2913                 nested_list = NULL;
2914                 ml = NULL;
2915                 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
2916                     /* substitution done, NOTE: maybe empty */
2917                     tok_str_add(&str, 0);
2918                     macro_ptr = str.str;
2919                     macro_ptr_allocated = str.str;
2920                     goto redo;
2921                 }
2922             }
2923         }
2924     } else {
2925         if (tok == 0) {
2926             /* end of macro or end of unget buffer */
2927             if (unget_buffer_enabled) {
2928                 macro_ptr = unget_saved_macro_ptr;
2929                 unget_buffer_enabled = 0;
2930             } else {
2931                 /* end of macro string: free it */
2932                 tok_str_free(macro_ptr_allocated);
2933                 macro_ptr_allocated = NULL;
2934                 macro_ptr = NULL;
2935             }
2936             goto redo;
2937         } else if (tok == TOK_NOSUBST) {
2938             /* discard preprocessor's nosubst markers */
2939             goto redo;
2940         }
2941     }
2942
2943     /* convert preprocessor tokens into C tokens */
2944     if (tok == TOK_PPNUM &&
2945         (parse_flags & PARSE_FLAG_TOK_NUM)) {
2946         parse_number((char *)tokc.cstr->data);
2947     }
2948 }
2949
2950 /* push back current token and set current token to 'last_tok'. Only
2951    identifier case handled for labels. */
2952 ST_INLN void unget_tok(int last_tok)
2953 {
2954     int i, n;
2955     int *q;
2956     if (unget_buffer_enabled)
2957       {
2958         /* assert(macro_ptr == unget_saved_buffer + 1);
2959            assert(*macro_ptr == 0);  */
2960       }
2961     else
2962       {
2963         unget_saved_macro_ptr = macro_ptr;
2964         unget_buffer_enabled = 1;
2965       }
2966     q = unget_saved_buffer;
2967     macro_ptr = q;
2968     *q++ = tok;
2969     n = tok_ext_size(tok) - 1;
2970     for(i=0;i<n;i++)
2971         *q++ = tokc.tab[i];
2972     *q = 0; /* end of token string */
2973     tok = last_tok;
2974 }
2975
2976
2977 /* better than nothing, but needs extension to handle '-E' option
2978    correctly too */
2979 ST_FUNC void preprocess_init(TCCState *s1)
2980 {
2981     s1->include_stack_ptr = s1->include_stack;
2982     /* XXX: move that before to avoid having to initialize
2983        file->ifdef_stack_ptr ? */
2984     s1->ifdef_stack_ptr = s1->ifdef_stack;
2985     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
2986
2987     vtop = vstack - 1;
2988     s1->pack_stack[0] = 0;
2989     s1->pack_stack_ptr = s1->pack_stack;
2990 }
2991
2992 ST_FUNC void preprocess_new(void)
2993 {
2994     int i, c;
2995     const char *p, *r;
2996
2997     /* init isid table */
2998     for(i=CH_EOF;i<256;i++)
2999         isidnum_table[i-CH_EOF] = isid(i) || isnum(i);
3000
3001     /* add all tokens */
3002     table_ident = NULL;
3003     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3004
3005     tok_ident = TOK_IDENT;
3006     p = tcc_keywords;
3007     while (*p) {
3008         r = p;
3009         for(;;) {
3010             c = *r++;
3011             if (c == '\0')
3012                 break;
3013         }
3014         tok_alloc(p, r - p - 1);
3015         p = r;
3016     }
3017 }
3018
3019 /* Preprocess the current file */
3020 ST_FUNC int tcc_preprocess(TCCState *s1)
3021 {
3022     Sym *define_start;
3023
3024     BufferedFile *file_ref, **iptr, **iptr_new;
3025     int token_seen, line_ref, d;
3026     const char *s;
3027
3028     preprocess_init(s1);
3029     define_start = define_stack;
3030     ch = file->buf_ptr[0];
3031     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3032     parse_flags = PARSE_FLAG_ASM_COMMENTS | PARSE_FLAG_PREPROCESS |
3033         PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
3034     token_seen = 0;
3035     line_ref = 0;
3036     file_ref = NULL;
3037     iptr = s1->include_stack_ptr;
3038
3039     for (;;) {
3040         next();
3041         if (tok == TOK_EOF) {
3042             break;
3043         } else if (file != file_ref) {
3044             goto print_line;
3045         } else if (tok == TOK_LINEFEED) {
3046             if (!token_seen)
3047                 continue;
3048             ++line_ref;
3049             token_seen = 0;
3050         } else if (!token_seen) {
3051             d = file->line_num - line_ref;
3052             if (file != file_ref || d < 0 || d >= 8) {
3053 print_line:
3054                 iptr_new = s1->include_stack_ptr;
3055                 s = iptr_new > iptr ? " 1"
3056                   : iptr_new < iptr ? " 2"
3057                   : iptr_new > s1->include_stack ? " 3"
3058                   : ""
3059                   ;
3060                 iptr = iptr_new;
3061                 fprintf(s1->ppfp, "# %d \"%s\"%s\n", file->line_num, file->filename, s);
3062             } else {
3063                 while (d)
3064                     fputs("\n", s1->ppfp), --d;
3065             }
3066             line_ref = (file_ref = file)->line_num;
3067             token_seen = tok != TOK_LINEFEED;
3068             if (!token_seen)
3069                 continue;
3070         }
3071         fputs(get_tok_str(tok, &tokc), s1->ppfp);
3072     }
3073     free_defines(define_start);
3074     return 0;
3075 }