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 ST_DATA int parse_flags;
  28
  29 ST_DATA struct BufferedFile *file;
  30 ST_DATA int ch, tok;
  31 ST_DATA CValue tokc;
  32 ST_DATA const int *macro_ptr;
  33 ST_DATA CString tokcstr; /* current parsed string, if any */
  34
  35 /* display benchmark infos */
  36 ST_DATA int total_lines;
  37 ST_DATA int total_bytes;
  38 ST_DATA int tok_ident;
  39 ST_DATA TokenSym **table_ident;
  40
  41 /* ------------------------------------------------------------------------- */
  42
  43 static int *macro_ptr_allocated;
  44 static const int *unget_saved_macro_ptr;
  45 static int unget_saved_buffer[TOK_MAX_SIZE + 1];
  46 static int unget_buffer_enabled;
  47 static TokenSym *hash_ident[TOK_HASH_SIZE];
  48 static char token_buf[STRING_MAX_SIZE + 1];
  49 /* true if isid(c) || isnum(c) */
  50 static unsigned char isidnum_table[256-CH_EOF];
  51
  52 static const char tcc_keywords[] =
  53 #define DEF(id, str) str "\0"
  54 #include "tcctok.h"
  55 #undef DEF
  56 ;
  57
  58 /* WARNING: the content of this string encodes token numbers */
  59 static const unsigned char tok_two_chars[] =
  60 /* outdated -- gr
  61     "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
  62     "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
  63 */{
  64     '<','=', TOK_LE,
  65     '>','=', TOK_GE,
  66     '!','=', TOK_NE,
  67     '&','&', TOK_LAND,
  68     '|','|', TOK_LOR,
  69     '+','+', TOK_INC,
  70     '-','-', TOK_DEC,
  71     '=','=', TOK_EQ,
  72     '<','<', TOK_SHL,
  73     '>','>', TOK_SAR,
  74     '+','=', TOK_A_ADD,
  75     '-','=', TOK_A_SUB,
  76     '*','=', TOK_A_MUL,
  77     '/','=', TOK_A_DIV,
  78     '%','=', TOK_A_MOD,
  79     '&','=', TOK_A_AND,
  80     '^','=', TOK_A_XOR,
  81     '|','=', TOK_A_OR,
  82     '-','>', TOK_ARROW,
  83     '.','.', 0xa8, // C++ token ?
  84     '#','#', TOK_TWOSHARPS,
  85     0
  86 };
  87
  88 struct macro_level {
  89     struct macro_level *prev;
  90     const int *p;
  91 };
  92
  93 static void next_nomacro_spc(void);
  94 static void macro_subst(
  95     TokenString *tok_str,
  96     Sym **nested_list,
  97     const int *macro_str,
  98     struct macro_level **can_read_stream
  99     );
 100
 101 ST_FUNC void skip(int c)
 102 {
 103     if (tok != c)
 104         tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
 105     next();
 106 }
 107
 108 ST_FUNC void expect(const char *msg)
 109 {
 110     tcc_error("%s expected", msg);
 111 }
 112
 113 /* ------------------------------------------------------------------------- */
 114 /* CString handling */
 115 static void cstr_realloc(CString *cstr, int new_size)
 116 {
 117     int size;
 118     void *data;
 119
 120     size = cstr->size_allocated;
 121     if (size == 0)
 122         size = 8; /* no need to allocate a too small first string */
 123     while (size < new_size)
 124         size = size * 2;
 125     data = tcc_realloc(cstr->data_allocated, size);
 126     cstr->data_allocated = data;
 127     cstr->size_allocated = size;
 128     cstr->data = data;
 129 }
 130
 131 /* add a byte */
 132 ST_FUNC void cstr_ccat(CString *cstr, int ch)
 133 {
 134     int size;
 135     size = cstr->size + 1;
 136     if (size > cstr->size_allocated)
 137         cstr_realloc(cstr, size);
 138     ((unsigned char *)cstr->data)[size - 1] = ch;
 139     cstr->size = size;
 140 }
 141
 142 ST_FUNC void cstr_cat(CString *cstr, const char *str)
 143 {
 144     int c;
 145     for(;;) {
 146         c = *str;
 147         if (c == '\0')
 148             break;
 149         cstr_ccat(cstr, c);
 150         str++;
 151     }
 152 }
 153
 154 /* add a wide char */
 155 ST_FUNC void cstr_wccat(CString *cstr, int ch)
 156 {
 157     int size;
 158     size = cstr->size + sizeof(nwchar_t);
 159     if (size > cstr->size_allocated)
 160         cstr_realloc(cstr, size);
 161     *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
 162     cstr->size = size;
 163 }
 164
 165 ST_FUNC void cstr_new(CString *cstr)
 166 {
 167     memset(cstr, 0, sizeof(CString));
 168 }
 169
 170 /* free string and reset it to NULL */
 171 ST_FUNC void cstr_free(CString *cstr)
 172 {
 173     tcc_free(cstr->data_allocated);
 174     cstr_new(cstr);
 175 }
 176
 177 /* reset string to empty */
 178 ST_FUNC void cstr_reset(CString *cstr)
 179 {
 180     cstr->size = 0;
 181 }
 182
 183 /* XXX: unicode ? */
 184 static void add_char(CString *cstr, int c)
 185 {
 186     if (c == '\'' || c == '\"' || c == '\\') {
 187         /* XXX: could be more precise if char or string */
 188         cstr_ccat(cstr, '\\');
 189     }
 190     if (c >= 32 && c <= 126) {
 191         cstr_ccat(cstr, c);
 192     } else {
 193         cstr_ccat(cstr, '\\');
 194         if (c == '\n') {
 195             cstr_ccat(cstr, 'n');
 196         } else {
 197             cstr_ccat(cstr, '0' + ((c >> 6) & 7));
 198             cstr_ccat(cstr, '0' + ((c >> 3) & 7));
 199             cstr_ccat(cstr, '0' + (c & 7));
 200         }
 201     }
 202 }
 203
 204 /* ------------------------------------------------------------------------- */
 205 /* allocate a new token */
 206 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
 207 {
 208     TokenSym *ts, **ptable;
 209     int i;
 210
 211     if (tok_ident >= SYM_FIRST_ANOM)
 212         tcc_error("memory full (symbols)");
 213
 214     /* expand token table if needed */
 215     i = tok_ident - TOK_IDENT;
 216     if ((i % TOK_ALLOC_INCR) == 0) {
 217         ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
 218         table_ident = ptable;
 219     }
 220
 221     ts = tcc_malloc(sizeof(TokenSym) + len);
 222     table_ident[i] = ts;
 223     ts->tok = tok_ident++;
 224     ts->sym_define = NULL;
 225     ts->sym_label = NULL;
 226     ts->sym_struct = NULL;
 227     ts->sym_identifier = NULL;
 228     ts->len = len;
 229     ts->hash_next = NULL;
 230     memcpy(ts->str, str, len);
 231     ts->str[len] = '\0';
 232     *pts = ts;
 233     return ts;
 234 }
 235
 236 #define TOK_HASH_INIT 1
 237 #define TOK_HASH_FUNC(h, c) ((h) * 263 + (c))
 238
 239 /* find a token and add it if not found */
 240 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
 241 {
 242     TokenSym *ts, **pts;
 243     int i;
 244     unsigned int h;
 245
 246     h = TOK_HASH_INIT;
 247     for(i=0;i<len;i++)
 248         h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
 249     h &= (TOK_HASH_SIZE - 1);
 250
 251     pts = &hash_ident[h];
 252     for(;;) {
 253         ts = *pts;
 254         if (!ts)
 255             break;
 256         if (ts->len == len && !memcmp(ts->str, str, len))
 257             return ts;
 258         pts = &(ts->hash_next);
 259     }
 260     return tok_alloc_new(pts, str, len);
 261 }
 262
 263 /* XXX: buffer overflow */
 264 /* XXX: float tokens */
 265 ST_FUNC char *get_tok_str(int v, CValue *cv)
 266 {
 267     static char buf[STRING_MAX_SIZE + 1];
 268     static CString cstr_buf;
 269     CString *cstr;
 270     char *p;
 271     int i, len;
 272
 273     /* NOTE: to go faster, we give a fixed buffer for small strings */
 274     cstr_reset(&cstr_buf);
 275     cstr_buf.data = buf;
 276     cstr_buf.size_allocated = sizeof(buf);
 277     p = buf;
 278
 279 /* just an explanation, should never happen:
 280     if (v <= TOK_LINENUM && v >= TOK_CINT && cv == NULL)
 281         tcc_error("internal error: get_tok_str"); */
 282
 283     switch(v) {
 284     case TOK_CINT:
 285     case TOK_CUINT:
 286         /* XXX: not quite exact, but only useful for testing */
 287         sprintf(p, "%u", cv->ui);
 288         break;
 289     case TOK_CLLONG:
 290     case TOK_CULLONG:
 291         /* XXX: not quite exact, but only useful for testing  */
 292 #ifdef _WIN32
 293         sprintf(p, "%u", (unsigned)cv->ull);
 294 #else
 295         sprintf(p, "%llu", cv->ull);
 296 #endif
 297         break;
 298     case TOK_LCHAR:
 299         cstr_ccat(&cstr_buf, 'L');
 300     case TOK_CCHAR:
 301         cstr_ccat(&cstr_buf, '\'');
 302         add_char(&cstr_buf, cv->i);
 303         cstr_ccat(&cstr_buf, '\'');
 304         cstr_ccat(&cstr_buf, '\0');
 305         break;
 306     case TOK_PPNUM:
 307         cstr = cv->cstr;
 308         len = cstr->size - 1;
 309         for(i=0;i<len;i++)
 310             add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 311         cstr_ccat(&cstr_buf, '\0');
 312         break;
 313     case TOK_LSTR:
 314         cstr_ccat(&cstr_buf, 'L');
 315     case TOK_STR:
 316         cstr = cv->cstr;
 317         cstr_ccat(&cstr_buf, '\"');
 318         if (v == TOK_STR) {
 319             len = cstr->size - 1;
 320             for(i=0;i<len;i++)
 321                 add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 322         } else {
 323             len = (cstr->size / sizeof(nwchar_t)) - 1;
 324             for(i=0;i<len;i++)
 325                 add_char(&cstr_buf, ((nwchar_t *)cstr->data)[i]);
 326         }
 327         cstr_ccat(&cstr_buf, '\"');
 328         cstr_ccat(&cstr_buf, '\0');
 329         break;
 330
 331     case TOK_CFLOAT:
 332     case TOK_CDOUBLE:
 333     case TOK_CLDOUBLE:
 334     case TOK_LINENUM:
 335         return NULL; /* should not happen */
 336
 337     /* above tokens have value, the ones below don't */
 338
 339     case TOK_LT:
 340         v = '<';
 341         goto addv;
 342     case TOK_GT:
 343         v = '>';
 344         goto addv;
 345     case TOK_DOTS:
 346         return strcpy(p, "...");
 347     case TOK_A_SHL:
 348         return strcpy(p, "<<=");
 349     case TOK_A_SAR:
 350         return strcpy(p, ">>=");
 351     default:
 352         if (v < TOK_IDENT) {
 353             /* search in two bytes table */
 354             const unsigned char *q = tok_two_chars;
 355             while (*q) {
 356                 if (q[2] == v) {
 357                     *p++ = q[0];
 358                     *p++ = q[1];
 359                     *p = '\0';
 360                     return buf;
 361                 }
 362                 q += 3;
 363             }
 364         addv:
 365             *p++ = v;
 366             *p = '\0';
 367         } else if (v < tok_ident) {
 368             return table_ident[v - TOK_IDENT]->str;
 369         } else if (v >= SYM_FIRST_ANOM) {
 370             /* special name for anonymous symbol */
 371             sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
 372         } else {
 373             /* should never happen */
 374             return NULL;
 375         }
 376         break;
 377     }
 378     return cstr_buf.data;
 379 }
 380
 381 /* fill input buffer and peek next char */
 382 static int tcc_peekc_slow(BufferedFile *bf)
 383 {
 384     int len;
 385     /* only tries to read if really end of buffer */
 386     if (bf->buf_ptr >= bf->buf_end) {
 387         if (bf->fd != -1) {
 388 #if defined(PARSE_DEBUG)
 389             len = 8;
 390 #else
 391             len = IO_BUF_SIZE;
 392 #endif
 393             len = read(bf->fd, bf->buffer, len);
 394             if (len < 0)
 395                 len = 0;
 396         } else {
 397             len = 0;
 398         }
 399         total_bytes += len;
 400         bf->buf_ptr = bf->buffer;
 401         bf->buf_end = bf->buffer + len;
 402         *bf->buf_end = CH_EOB;
 403     }
 404     if (bf->buf_ptr < bf->buf_end) {
 405         return bf->buf_ptr[0];
 406     } else {
 407         bf->buf_ptr = bf->buf_end;
 408         return CH_EOF;
 409     }
 410 }
 411
 412 /* return the current character, handling end of block if necessary
 413    (but not stray) */
 414 ST_FUNC int handle_eob(void)
 415 {
 416     return tcc_peekc_slow(file);
 417 }
 418
 419 /* read next char from current input file and handle end of input buffer */
 420 ST_INLN void inp(void)
 421 {
 422     ch = *(++(file->buf_ptr));
 423     /* end of buffer/file handling */
 424     if (ch == CH_EOB)
 425         ch = handle_eob();
 426 }
 427
 428 /* handle '\[\r]\n' */
 429 static int handle_stray_noerror(void)
 430 {
 431     while (ch == '\\') {
 432         inp();
 433         if (ch == '\n') {
 434             file->line_num++;
 435             inp();
 436         } else if (ch == '\r') {
 437             inp();
 438             if (ch != '\n')
 439                 goto fail;
 440             file->line_num++;
 441             inp();
 442         } else {
 443         fail:
 444             return 1;
 445         }
 446     }
 447     return 0;
 448 }
 449
 450 static void handle_stray(void)
 451 {
 452     if (handle_stray_noerror())
 453         tcc_error("stray '\\' in program");
 454 }
 455
 456 /* skip the stray and handle the \\n case. Output an error if
 457    incorrect char after the stray */
 458 static int handle_stray1(uint8_t *p)
 459 {
 460     int c;
 461
 462     if (p >= file->buf_end) {
 463         file->buf_ptr = p;
 464         c = handle_eob();
 465         p = file->buf_ptr;
 466         if (c == '\\')
 467             goto parse_stray;
 468     } else {
 469     parse_stray:
 470         file->buf_ptr = p;
 471         ch = *p;
 472         handle_stray();
 473         p = file->buf_ptr;
 474         c = *p;
 475     }
 476     return c;
 477 }
 478
 479 /* handle just the EOB case, but not stray */
 480 #define PEEKC_EOB(c, p)\
 481 {\
 482     p++;\
 483     c = *p;\
 484     if (c == '\\') {\
 485         file->buf_ptr = p;\
 486         c = handle_eob();\
 487         p = file->buf_ptr;\
 488     }\
 489 }
 490
 491 /* handle the complicated stray case */
 492 #define PEEKC(c, p)\
 493 {\
 494     p++;\
 495     c = *p;\
 496     if (c == '\\') {\
 497         c = handle_stray1(p);\
 498         p = file->buf_ptr;\
 499     }\
 500 }
 501
 502 /* input with '\[\r]\n' handling. Note that this function cannot
 503    handle other characters after '\', so you cannot call it inside
 504    strings or comments */
 505 ST_FUNC void minp(void)
 506 {
 507     inp();
 508     if (ch == '\\')
 509         handle_stray();
 510 }
 511
 512
 513 /* single line C++ comments */
 514 static uint8_t *parse_line_comment(uint8_t *p)
 515 {
 516     int c;
 517
 518     p++;
 519     for(;;) {
 520         c = *p;
 521     redo:
 522         if (c == '\n' || c == CH_EOF) {
 523             break;
 524         } else if (c == '\\') {
 525             file->buf_ptr = p;
 526             c = handle_eob();
 527             p = file->buf_ptr;
 528             if (c == '\\') {
 529                 PEEKC_EOB(c, p);
 530                 if (c == '\n') {
 531                     file->line_num++;
 532                     PEEKC_EOB(c, p);
 533                 } else if (c == '\r') {
 534                     PEEKC_EOB(c, p);
 535                     if (c == '\n') {
 536                         file->line_num++;
 537                         PEEKC_EOB(c, p);
 538                     }
 539                 }
 540             } else {
 541                 goto redo;
 542             }
 543         } else {
 544             p++;
 545         }
 546     }
 547     return p;
 548 }
 549
 550 /* C comments */
 551 ST_FUNC uint8_t *parse_comment(uint8_t *p)
 552 {
 553     int c;
 554
 555     p++;
 556     for(;;) {
 557         /* fast skip loop */
 558         for(;;) {
 559             c = *p;
 560             if (c == '\n' || c == '*' || c == '\\')
 561                 break;
 562             p++;
 563             c = *p;
 564             if (c == '\n' || c == '*' || c == '\\')
 565                 break;
 566             p++;
 567         }
 568         /* now we can handle all the cases */
 569         if (c == '\n') {
 570             file->line_num++;
 571             p++;
 572         } else if (c == '*') {
 573             p++;
 574             for(;;) {
 575                 c = *p;
 576                 if (c == '*') {
 577                     p++;
 578                 } else if (c == '/') {
 579                     goto end_of_comment;
 580                 } else if (c == '\\') {
 581                     file->buf_ptr = p;
 582                     c = handle_eob();
 583                     p = file->buf_ptr;
 584                     if (c == '\\') {
 585                         /* skip '\[\r]\n', otherwise just skip the stray */
 586                         while (c == '\\') {
 587                             PEEKC_EOB(c, p);
 588                             if (c == '\n') {
 589                                 file->line_num++;
 590                                 PEEKC_EOB(c, p);
 591                             } else if (c == '\r') {
 592                                 PEEKC_EOB(c, p);
 593                                 if (c == '\n') {
 594                                     file->line_num++;
 595                                     PEEKC_EOB(c, p);
 596                                 }
 597                             } else {
 598                                 goto after_star;
 599                             }
 600                         }
 601                     }
 602                 } else {
 603                     break;
 604                 }
 605             }
 606         after_star: ;
 607         } else {
 608             /* stray, eob or eof */
 609             file->buf_ptr = p;
 610             c = handle_eob();
 611             p = file->buf_ptr;
 612             if (c == CH_EOF) {
 613                 tcc_error("unexpected end of file in comment");
 614             } else if (c == '\\') {
 615                 p++;
 616             }
 617         }
 618     }
 619  end_of_comment:
 620     p++;
 621     return p;
 622 }
 623
 624 #define cinp minp
 625
 626 static inline void skip_spaces(void)
 627 {
 628     while (is_space(ch))
 629         cinp();
 630 }
 631
 632 static inline int check_space(int t, int *spc)
 633 {
 634     if (is_space(t)) {
 635         if (*spc)
 636             return 1;
 637         *spc = 1;
 638     } else
 639         *spc = 0;
 640     return 0;
 641 }
 642
 643 /* parse a string without interpreting escapes */
 644 static uint8_t *parse_pp_string(uint8_t *p,
 645                                 int sep, CString *str)
 646 {
 647     int c;
 648     p++;
 649     for(;;) {
 650         c = *p;
 651         if (c == sep) {
 652             break;
 653         } else if (c == '\\') {
 654             file->buf_ptr = p;
 655             c = handle_eob();
 656             p = file->buf_ptr;
 657             if (c == CH_EOF) {
 658             unterminated_string:
 659                 /* XXX: indicate line number of start of string */
 660                 tcc_error("missing terminating %c character", sep);
 661             } else if (c == '\\') {
 662                 /* escape : just skip \[\r]\n */
 663                 PEEKC_EOB(c, p);
 664                 if (c == '\n') {
 665                     file->line_num++;
 666                     p++;
 667                 } else if (c == '\r') {
 668                     PEEKC_EOB(c, p);
 669                     if (c != '\n')
 670                         expect("'\n' after '\r'");
 671                     file->line_num++;
 672                     p++;
 673                 } else if (c == CH_EOF) {
 674                     goto unterminated_string;
 675                 } else {
 676                     if (str) {
 677                         cstr_ccat(str, '\\');
 678                         cstr_ccat(str, c);
 679                     }
 680                     p++;
 681                 }
 682             }
 683         } else if (c == '\n') {
 684             file->line_num++;
 685             goto add_char;
 686         } else if (c == '\r') {
 687             PEEKC_EOB(c, p);
 688             if (c != '\n') {
 689                 if (str)
 690                     cstr_ccat(str, '\r');
 691             } else {
 692                 file->line_num++;
 693                 goto add_char;
 694             }
 695         } else {
 696         add_char:
 697             if (str)
 698                 cstr_ccat(str, c);
 699             p++;
 700         }
 701     }
 702     p++;
 703     return p;
 704 }
 705
 706 /* skip block of text until #else, #elif or #endif. skip also pairs of
 707    #if/#endif */
 708 static void preprocess_skip(void)
 709 {
 710     int a, start_of_line, c, in_warn_or_error;
 711     uint8_t *p;
 712
 713     p = file->buf_ptr;
 714     a = 0;
 715 redo_start:
 716     start_of_line = 1;
 717     in_warn_or_error = 0;
 718     for(;;) {
 719     redo_no_start:
 720         c = *p;
 721         switch(c) {
 722         case ' ':
 723         case '\t':
 724         case '\f':
 725         case '\v':
 726         case '\r':
 727             p++;
 728             goto redo_no_start;
 729         case '\n':
 730             file->line_num++;
 731             p++;
 732             goto redo_start;
 733         case '\\':
 734             file->buf_ptr = p;
 735             c = handle_eob();
 736             if (c == CH_EOF) {
 737                 expect("#endif");
 738             } else if (c == '\\') {
 739                 ch = file->buf_ptr[0];
 740                 handle_stray_noerror();
 741             }
 742             p = file->buf_ptr;
 743             goto redo_no_start;
 744         /* skip strings */
 745         case '\"':
 746         case '\'':
 747             if (in_warn_or_error)
 748                 goto _default;
 749             p = parse_pp_string(p, c, NULL);
 750             break;
 751         /* skip comments */
 752         case '/':
 753             if (in_warn_or_error)
 754                 goto _default;
 755             file->buf_ptr = p;
 756             ch = *p;
 757             minp();
 758             p = file->buf_ptr;
 759             if (ch == '*') {
 760                 p = parse_comment(p);
 761             } else if (ch == '/') {
 762                 p = parse_line_comment(p);
 763             }
 764             break;
 765         case '#':
 766             p++;
 767             if (start_of_line) {
 768                 file->buf_ptr = p;
 769                 next_nomacro();
 770                 p = file->buf_ptr;
 771                 if (a == 0 &&
 772                     (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
 773                     goto the_end;
 774                 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
 775                     a++;
 776                 else if (tok == TOK_ENDIF)
 777                     a--;
 778                 else if( tok == TOK_ERROR || tok == TOK_WARNING)
 779                     in_warn_or_error = 1;
 780                 else if (tok == TOK_LINEFEED)
 781                     goto redo_start;
 782                 else if (parse_flags & PARSE_FLAG_ASM_COMMENTS)
 783                     p = parse_line_comment(p);
 784             }
 785             else if (parse_flags & PARSE_FLAG_ASM_FILE)
 786                 p = parse_line_comment(p);
 787             break;
 788 _default:
 789         default:
 790             p++;
 791             break;
 792         }
 793         start_of_line = 0;
 794     }
 795  the_end: ;
 796     file->buf_ptr = p;
 797 }
 798
 799 /* ParseState handling */
 800
 801 /* XXX: currently, no include file info is stored. Thus, we cannot display
 802    accurate messages if the function or data definition spans multiple
 803    files */
 804
 805 /* save current parse state in 's' */
 806 ST_FUNC void save_parse_state(ParseState *s)
 807 {
 808     s->line_num = file->line_num;
 809     s->macro_ptr = macro_ptr;
 810     s->tok = tok;
 811     s->tokc = tokc;
 812 }
 813
 814 /* restore parse state from 's' */
 815 ST_FUNC void restore_parse_state(ParseState *s)
 816 {
 817     file->line_num = s->line_num;
 818     macro_ptr = s->macro_ptr;
 819     tok = s->tok;
 820     tokc = s->tokc;
 821 }
 822
 823 /* return the number of additional 'ints' necessary to store the
 824    token */
 825 static inline int tok_ext_size(int t)
 826 {
 827     switch(t) {
 828         /* 4 bytes */
 829     case TOK_CINT:
 830     case TOK_CUINT:
 831     case TOK_CCHAR:
 832     case TOK_LCHAR:
 833     case TOK_CFLOAT:
 834     case TOK_LINENUM:
 835         return 1;
 836     case TOK_STR:
 837     case TOK_LSTR:
 838     case TOK_PPNUM:
 839         tcc_error("unsupported token");
 840         return 1;
 841     case TOK_CDOUBLE:
 842     case TOK_CLLONG:
 843     case TOK_CULLONG:
 844         return 2;
 845     case TOK_CLDOUBLE:
 846         return LDOUBLE_SIZE / 4;
 847     default:
 848         return 0;
 849     }
 850 }
 851
 852 /* token string handling */
 853
 854 ST_INLN void tok_str_new(TokenString *s)
 855 {
 856     s->str = NULL;
 857     s->len = 0;
 858     s->allocated_len = 0;
 859     s->last_line_num = -1;
 860 }
 861
 862 ST_FUNC void tok_str_free(int *str)
 863 {
 864     tcc_free(str);
 865 }
 866
 867 static int *tok_str_realloc(TokenString *s)
 868 {
 869     int *str, len;
 870
 871     if (s->allocated_len == 0) {
 872         len = 8;
 873     } else {
 874         len = s->allocated_len * 2;
 875     }
 876     str = tcc_realloc(s->str, len * sizeof(int));
 877     s->allocated_len = len;
 878     s->str = str;
 879     return str;
 880 }
 881
 882 ST_FUNC void tok_str_add(TokenString *s, int t)
 883 {
 884     int len, *str;
 885
 886     len = s->len;
 887     str = s->str;
 888     if (len >= s->allocated_len)
 889         str = tok_str_realloc(s);
 890     str[len++] = t;
 891     s->len = len;
 892 }
 893
 894 static void tok_str_add2(TokenString *s, int t, CValue *cv)
 895 {
 896     int len, *str;
 897
 898     len = s->len;
 899     str = s->str;
 900
 901     /* allocate space for worst case */
 902     if (len + TOK_MAX_SIZE > s->allocated_len)
 903         str = tok_str_realloc(s);
 904     str[len++] = t;
 905     switch(t) {
 906     case TOK_CINT:
 907     case TOK_CUINT:
 908     case TOK_CCHAR:
 909     case TOK_LCHAR:
 910     case TOK_CFLOAT:
 911     case TOK_LINENUM:
 912         str[len++] = cv->tab[0];
 913         break;
 914     case TOK_PPNUM:
 915     case TOK_STR:
 916     case TOK_LSTR:
 917         {
 918             int nb_words;
 919             CString *cstr;
 920
 921             nb_words = (sizeof(CString) + cv->cstr->size + 3) >> 2;
 922             while ((len + nb_words) > s->allocated_len)
 923                 str = tok_str_realloc(s);
 924             cstr = (CString *)(str + len);
 925             cstr->data = NULL;
 926             cstr->size = cv->cstr->size;
 927             cstr->data_allocated = NULL;
 928             cstr->size_allocated = cstr->size;
 929             memcpy((char *)cstr + sizeof(CString),
 930                    cv->cstr->data, cstr->size);
 931             len += nb_words;
 932         }
 933         break;
 934     case TOK_CDOUBLE:
 935     case TOK_CLLONG:
 936     case TOK_CULLONG:
 937 #if LDOUBLE_SIZE == 8
 938     case TOK_CLDOUBLE:
 939 #endif
 940         str[len++] = cv->tab[0];
 941         str[len++] = cv->tab[1];
 942         break;
 943 #if LDOUBLE_SIZE == 12
 944     case TOK_CLDOUBLE:
 945         str[len++] = cv->tab[0];
 946         str[len++] = cv->tab[1];
 947         str[len++] = cv->tab[2];
 948 #elif LDOUBLE_SIZE == 16
 949     case TOK_CLDOUBLE:
 950         str[len++] = cv->tab[0];
 951         str[len++] = cv->tab[1];
 952         str[len++] = cv->tab[2];
 953         str[len++] = cv->tab[3];
 954 #elif LDOUBLE_SIZE != 8
 955 #error add long double size support
 956 #endif
 957         break;
 958     default:
 959         break;
 960     }
 961     s->len = len;
 962 }
 963
 964 /* add the current parse token in token string 's' */
 965 ST_FUNC void tok_str_add_tok(TokenString *s)
 966 {
 967     CValue cval;
 968
 969     /* save line number info */
 970     if (file->line_num != s->last_line_num) {
 971         s->last_line_num = file->line_num;
 972         cval.i = s->last_line_num;
 973         tok_str_add2(s, TOK_LINENUM, &cval);
 974     }
 975     tok_str_add2(s, tok, &tokc);
 976 }
 977
 978 /* get a token from an integer array and increment pointer
 979    accordingly. we code it as a macro to avoid pointer aliasing. */
 980 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
 981 {
 982     const int *p = *pp;
 983     int n, *tab;
 984
 985     tab = cv->tab;
 986     switch(*t = *p++) {
 987     case TOK_CINT:
 988     case TOK_CUINT:
 989     case TOK_CCHAR:
 990     case TOK_LCHAR:
 991     case TOK_CFLOAT:
 992     case TOK_LINENUM:
 993         tab[0] = *p++;
 994         break;
 995     case TOK_STR:
 996     case TOK_LSTR:
 997     case TOK_PPNUM:
 998         cv->cstr = (CString *)p;
 999         cv->cstr->data = (char *)p + sizeof(CString);
1000         p += (sizeof(CString) + cv->cstr->size + 3) >> 2;
1001         break;
1002     case TOK_CDOUBLE:
1003     case TOK_CLLONG:
1004     case TOK_CULLONG:
1005         n = 2;
1006         goto copy;
1007     case TOK_CLDOUBLE:
1008 #if LDOUBLE_SIZE == 16
1009         n = 4;
1010 #elif LDOUBLE_SIZE == 12
1011         n = 3;
1012 #elif LDOUBLE_SIZE == 8
1013         n = 2;
1014 #else
1015 # error add long double size support
1016 #endif
1017     copy:
1018         do
1019             *tab++ = *p++;
1020         while (--n);
1021         break;
1022     default:
1023         break;
1024     }
1025     *pp = p;
1026 }
1027
1028 static int macro_is_equal(const int *a, const int *b)
1029 {
1030     char buf[STRING_MAX_SIZE + 1];
1031     CValue cv;
1032     int t;
1033     while (*a && *b) {
1034         TOK_GET(&t, &a, &cv);
1035         pstrcpy(buf, sizeof buf, get_tok_str(t, &cv));
1036         TOK_GET(&t, &b, &cv);
1037         if (strcmp(buf, get_tok_str(t, &cv)))
1038             return 0;
1039     }
1040     return !(*a || *b);
1041 }
1042
1043 static void define_print(Sym *s, int is_undef)
1044 {
1045     int c, t;
1046     CValue cval;
1047     const int *str;
1048     Sym *arg;
1049
1050     if (tcc_state->dflag == 0 || !s || !tcc_state->ppfp)
1051         return;
1052
1053     if (file) {
1054         c = file->line_num - file->line_ref - 1;
1055         if (c > 0) {
1056             while (c--)
1057                 fputs("\n", tcc_state->ppfp);
1058             file->line_ref = file->line_num;
1059         }
1060     }
1061
1062     if (is_undef) {
1063         fprintf(tcc_state->ppfp, "// #undef %s", get_tok_str(s->v, NULL));
1064         return;
1065     }
1066
1067     fprintf(tcc_state->ppfp, "// #define %s", get_tok_str(s->v, NULL));
1068     arg = s->next;
1069     if (arg) {
1070         char *sep = "(";
1071         while (arg) {
1072             fprintf(tcc_state->ppfp, "%s%s", sep, get_tok_str(arg->v & ~SYM_FIELD, NULL));
1073             sep = ",";
1074             arg = arg->next;
1075         }
1076         fprintf(tcc_state->ppfp, ")");
1077     }
1078
1079     str = s->d;
1080     if (str)
1081         fprintf(tcc_state->ppfp, " ");
1082
1083     while (str) {
1084         TOK_GET(&t, &str, &cval);
1085         if (!t)
1086             break;
1087         fprintf(tcc_state->ppfp, "%s", get_tok_str(t, &cval));
1088     }
1089 }
1090
1091 /* defines handling */
1092 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1093 {
1094     Sym *s;
1095
1096     s = define_find(v);
1097     if (s && !macro_is_equal(s->d, str))
1098         tcc_warning("%s redefined", get_tok_str(v, NULL));
1099
1100     s = sym_push2(&define_stack, v, macro_type, 0);
1101     s->d = str;
1102     s->next = first_arg;
1103     table_ident[v - TOK_IDENT]->sym_define = s;
1104     define_print(s, 0);
1105 }
1106
1107 /* undefined a define symbol. Its name is just set to zero */
1108 ST_FUNC void define_undef(Sym *s)
1109 {
1110     int v;
1111     v = s->v;
1112     if (v >= TOK_IDENT && v < tok_ident) {
1113         define_print(s, 1);
1114         table_ident[v - TOK_IDENT]->sym_define = NULL;
1115     }
1116 }
1117
1118 ST_INLN Sym *define_find(int v)
1119 {
1120     v -= TOK_IDENT;
1121     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1122         return NULL;
1123     return table_ident[v]->sym_define;
1124 }
1125
1126 /* free define stack until top reaches 'b' */
1127 ST_FUNC void free_defines(Sym *b)
1128 {
1129     Sym *top, *top1;
1130     int v;
1131
1132     top = define_stack;
1133     while (top != b) {
1134         top1 = top->prev;
1135         /* do not free args or predefined defines */
1136         if (top->d)
1137             tok_str_free(top->d);
1138         v = top->v;
1139         if (v >= TOK_IDENT && v < tok_ident)
1140             table_ident[v - TOK_IDENT]->sym_define = NULL;
1141         sym_free(top);
1142         top = top1;
1143     }
1144     define_stack = b;
1145 }
1146
1147 ST_FUNC void print_defines(void)
1148 {
1149     Sym *top, *s;
1150     int v;
1151
1152     top = define_stack;
1153     while (top) {
1154         v = top->v;
1155         if (v >= TOK_IDENT && v < tok_ident) {
1156             s = table_ident[v - TOK_IDENT]->sym_define;
1157             define_print(s, 0);
1158         }
1159         top = top->prev;
1160     }
1161 }
1162
1163 /* label lookup */
1164 ST_FUNC Sym *label_find(int v)
1165 {
1166     v -= TOK_IDENT;
1167     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1168         return NULL;
1169     return table_ident[v]->sym_label;
1170 }
1171
1172 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1173 {
1174     Sym *s, **ps;
1175     s = sym_push2(ptop, v, 0, 0);
1176     s->r = flags;
1177     ps = &table_ident[v - TOK_IDENT]->sym_label;
1178     if (ptop == &global_label_stack) {
1179         /* modify the top most local identifier, so that
1180            sym_identifier will point to 's' when popped */
1181         while (*ps != NULL)
1182             ps = &(*ps)->prev_tok;
1183     }
1184     s->prev_tok = *ps;
1185     *ps = s;
1186     return s;
1187 }
1188
1189 /* pop labels until element last is reached. Look if any labels are
1190    undefined. Define symbols if '&&label' was used. */
1191 ST_FUNC void label_pop(Sym **ptop, Sym *slast)
1192 {
1193     Sym *s, *s1;
1194     for(s = *ptop; s != slast; s = s1) {
1195         s1 = s->prev;
1196         if (s->r == LABEL_DECLARED) {
1197             tcc_warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1198         } else if (s->r == LABEL_FORWARD) {
1199                 tcc_error("label '%s' used but not defined",
1200                       get_tok_str(s->v, NULL));
1201         } else {
1202             if (s->c) {
1203                 /* define corresponding symbol. A size of
1204                    1 is put. */
1205                 put_extern_sym(s, cur_text_section, s->jnext, 1);
1206             }
1207         }
1208         /* remove label */
1209         table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1210         sym_free(s);
1211     }
1212     *ptop = slast;
1213 }
1214
1215 /* eval an expression for #if/#elif */
1216 static int expr_preprocess(void)
1217 {
1218     int c, t;
1219     TokenString str;
1220
1221     tok_str_new(&str);
1222     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1223         next(); /* do macro subst */
1224         if (tok == TOK_DEFINED) {
1225             next_nomacro();
1226             t = tok;
1227             if (t == '(')
1228                 next_nomacro();
1229             c = define_find(tok) != 0;
1230             if (t == '(')
1231                 next_nomacro();
1232             tok = TOK_CINT;
1233             tokc.i = c;
1234         } else if (tok >= TOK_IDENT) {
1235             /* if undefined macro */
1236             tok = TOK_CINT;
1237             tokc.i = 0;
1238         }
1239         tok_str_add_tok(&str);
1240     }
1241     tok_str_add(&str, -1); /* simulate end of file */
1242     tok_str_add(&str, 0);
1243     /* now evaluate C constant expression */
1244     macro_ptr = str.str;
1245     next();
1246     c = expr_const();
1247     macro_ptr = NULL;
1248     tok_str_free(str.str);
1249     return c != 0;
1250 }
1251
1252 /* parse after #define */
1253 ST_FUNC void parse_define(void)
1254 {
1255     Sym *s, *first, **ps;
1256     int v, t, varg, is_vaargs, spc, ptok, macro_list_start;
1257     TokenString str;
1258
1259     v = tok;
1260     if (v < TOK_IDENT)
1261         tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1262     /* XXX: should check if same macro (ANSI) */
1263     first = NULL;
1264     t = MACRO_OBJ;
1265     /* '(' must be just after macro definition for MACRO_FUNC */
1266     next_nomacro_spc();
1267     if (tok == '(') {
1268         next_nomacro();
1269         ps = &first;
1270         while (tok != ')') {
1271             varg = tok;
1272             next_nomacro();
1273             is_vaargs = 0;
1274             if (varg == TOK_DOTS) {
1275                 varg = TOK___VA_ARGS__;
1276                 is_vaargs = 1;
1277             } else if (tok == TOK_DOTS && gnu_ext) {
1278                 is_vaargs = 1;
1279                 next_nomacro();
1280             }
1281             if (varg < TOK_IDENT)
1282                 tcc_error( "\'%s\' may not appear in parameter list", get_tok_str(varg, NULL));
1283             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1284             *ps = s;
1285             ps = &s->next;
1286             if (tok != ',')
1287                 continue;
1288             next_nomacro();
1289         }
1290         next_nomacro_spc();
1291         t = MACRO_FUNC;
1292     }
1293     tok_str_new(&str);
1294     spc = 2;
1295     /* EOF testing necessary for '-D' handling */
1296     ptok = 0;
1297     macro_list_start = 1;
1298     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1299         if (!macro_list_start && spc == 2 && tok == TOK_TWOSHARPS)
1300             tcc_error("'##' invalid at start of macro");
1301         ptok = tok;
1302         /* remove spaces around ## and after '#' */
1303         if (TOK_TWOSHARPS == tok) {
1304             if (1 == spc)
1305                 --str.len;
1306             spc = 2;
1307         } else if ('#' == tok) {
1308             spc = 2;
1309         } else if (check_space(tok, &spc)) {
1310             goto skip;
1311         }
1312         tok_str_add2(&str, tok, &tokc);
1313     skip:
1314         next_nomacro_spc();
1315         macro_list_start = 0;
1316     }
1317     if (ptok == TOK_TWOSHARPS)
1318         tcc_error("'##' invalid at end of macro");
1319     if (spc == 1)
1320         --str.len; /* remove trailing space */
1321     tok_str_add(&str, 0);
1322     define_push(v, t, str.str, first);
1323 }
1324
1325 static inline int hash_cached_include(const char *filename)
1326 {
1327     const unsigned char *s;
1328     unsigned int h;
1329
1330     h = TOK_HASH_INIT;
1331     s = (unsigned char *) filename;
1332     while (*s) {
1333         h = TOK_HASH_FUNC(h, *s);
1334         s++;
1335     }
1336     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1337     return h;
1338 }
1339
1340 static CachedInclude *search_cached_include(TCCState *s1, const char *filename)
1341 {
1342     CachedInclude *e;
1343     int i, h;
1344     h = hash_cached_include(filename);
1345     i = s1->cached_includes_hash[h];
1346     for(;;) {
1347         if (i == 0)
1348             break;
1349         e = s1->cached_includes[i - 1];
1350         if (0 == PATHCMP(e->filename, filename))
1351             return e;
1352         i = e->hash_next;
1353     }
1354     return NULL;
1355 }
1356
1357 static inline void add_cached_include(TCCState *s1, const char *filename, int ifndef_macro)
1358 {
1359     CachedInclude *e;
1360     int h;
1361
1362     if (search_cached_include(s1, filename))
1363         return;
1364 #ifdef INC_DEBUG
1365     printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1366 #endif
1367     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1368     strcpy(e->filename, filename);
1369     e->ifndef_macro = ifndef_macro;
1370     dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1371     /* add in hash table */
1372     h = hash_cached_include(filename);
1373     e->hash_next = s1->cached_includes_hash[h];
1374     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1375 }
1376
1377 static void pragma_parse(TCCState *s1)
1378 {
1379     next_nomacro();
1380     if (tok == TOK_pack) {
1381         /*
1382           This may be:
1383           #pragma pack(1) // set
1384           #pragma pack() // reset to default
1385           #pragma pack(push,1) // push & set
1386           #pragma pack(pop) // restore previous
1387         */
1388         next();
1389         skip('(');
1390         if (tok == TOK_ASM_pop) {
1391             next();
1392             if (s1->pack_stack_ptr <= s1->pack_stack) {
1393             stk_error:
1394                 tcc_error("out of pack stack");
1395             }
1396             s1->pack_stack_ptr--;
1397         } else {
1398             int val = 0;
1399             if (tok != ')') {
1400                 if (tok == TOK_ASM_push) {
1401                     next();
1402                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1403                         goto stk_error;
1404                     s1->pack_stack_ptr++;
1405                     skip(',');
1406                 }
1407                 if (tok != TOK_CINT)
1408                     goto pragma_err;
1409                 val = tokc.i;
1410                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1411                     goto pragma_err;
1412                 next();
1413             }
1414             *s1->pack_stack_ptr = val;
1415         }
1416         if (tok != ')')
1417             goto pragma_err;
1418     } else if (tok == TOK_comment) {
1419         if (s1->ms_extensions) {
1420             next();
1421             skip('(');
1422             if (tok == TOK_lib) {
1423                 next();
1424                 skip(',');
1425                 if (tok != TOK_STR) {
1426                     tcc_error("invalid library specification");
1427                 } else {
1428                     int len = strlen((char *)tokc.cstr->data);
1429                     char *file = tcc_malloc(len + 4); /* filetype, "-l", and \0 at the end */
1430                     file[0] = TCC_FILETYPE_BINARY;
1431                     file[1] = '-';
1432                     file[2] = 'l';
1433                     strcpy(&file[3],(char *)tokc.cstr->data);
1434                     dynarray_add((void ***)&s1->files, &s1->nb_files, file);
1435                 }
1436                 next();
1437                 tok = TOK_LINEFEED;
1438             } else {
1439                 tcc_warning("unknown specifier '%s' in #pragma comment", get_tok_str(tok, &tokc));
1440             }
1441         } else {
1442             tcc_warning("#pragma comment(lib) is ignored");
1443         }
1444     } else if (tok == TOK_push_macro || tok == TOK_pop_macro) {
1445         int t = tok, v;
1446         Sym *s;
1447
1448         if (next_nomacro(), tok != '(')
1449             goto pragma_err;
1450         if (next_nomacro(), tok != TOK_STR)
1451             goto pragma_err;
1452         v = tok_alloc(tokc.cstr->data, tokc.cstr->size - 1)->tok;
1453         if (next_nomacro(), tok != ')')
1454             goto pragma_err;
1455         if (t == TOK_push_macro) {
1456             while (NULL == (s = define_find(v)))
1457                 define_push(v, 0, NULL, NULL);
1458             s->type.ref = s; /* set push boundary */
1459         } else {
1460             for (s = define_stack; s; s = s->prev)
1461                 if (s->v == v && s->type.ref == s) {
1462                     s->type.ref = NULL;
1463                     break;
1464                 }
1465         }
1466         if (s)
1467             table_ident[v - TOK_IDENT]->sym_define = s->d ? s : NULL;
1468         else
1469             tcc_warning("unbalanced #pragma pop_macro");
1470
1471         /* print info when tcc is called with "-E -dD" switches */
1472         if (s1->dflag && s1->ppfp) {
1473             if (file) {
1474                 int c = file->line_num - file->line_ref - 1;
1475                 if (c > 0) {
1476                     while (c--)
1477                     fputs("\n", tcc_state->ppfp);
1478                     file->line_ref = file->line_num;
1479                 }
1480             }
1481             fprintf(s1->ppfp, "// #pragma %s_macro(\"%s\")",
1482                 (t == TOK_push_macro) ? "push" : "pop",
1483                 get_tok_str(v, NULL));
1484         }
1485     } else if (tok == TOK_once) {
1486         add_cached_include(s1, file->filename, TOK_once);
1487     } else {
1488         tcc_warning("unknown #pragma %s", get_tok_str(tok, &tokc));
1489     }
1490     return;
1491 pragma_err:
1492     tcc_error("malformed #pragma directive");
1493 }
1494
1495 /* is_bof is true if first non space token at beginning of file */
1496 ST_FUNC void preprocess(int is_bof)
1497 {
1498     TCCState *s1 = tcc_state;
1499     int i, c, n, saved_parse_flags;
1500     char buf[1024], *q;
1501     Sym *s;
1502
1503     saved_parse_flags = parse_flags;
1504     parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM |
1505         PARSE_FLAG_LINEFEED;
1506     parse_flags |= (saved_parse_flags & (PARSE_FLAG_ASM_FILE | PARSE_FLAG_ASM_COMMENTS));
1507     next_nomacro();
1508  redo:
1509     switch(tok) {
1510     case TOK_DEFINE:
1511         next_nomacro();
1512         parse_define();
1513         break;
1514     case TOK_UNDEF:
1515         next_nomacro();
1516         s = define_find(tok);
1517         /* undefine symbol by putting an invalid name */
1518         if (s)
1519             define_undef(s);
1520         break;
1521     case TOK_INCLUDE:
1522     case TOK_INCLUDE_NEXT:
1523         ch = file->buf_ptr[0];
1524         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1525         skip_spaces();
1526         if (ch == '<') {
1527             c = '>';
1528             goto read_name;
1529         } else if (ch == '\"') {
1530             c = ch;
1531         read_name:
1532             inp();
1533             q = buf;
1534             while (ch != c && ch != '\n' && ch != CH_EOF) {
1535                 if ((q - buf) < sizeof(buf) - 1)
1536                     *q++ = ch;
1537                 if (ch == '\\') {
1538                     if (handle_stray_noerror() == 0)
1539                         --q;
1540                 } else
1541                     inp();
1542             }
1543             *q = '\0';
1544             minp();
1545 #if 0
1546             /* eat all spaces and comments after include */
1547             /* XXX: slightly incorrect */
1548             while (ch1 != '\n' && ch1 != CH_EOF)
1549                 inp();
1550 #endif
1551         } else {
1552             /* computed #include : either we have only strings or
1553                we have anything enclosed in '<>' */
1554             next();
1555             buf[0] = '\0';
1556             if (tok == TOK_STR) {
1557                 while (tok != TOK_LINEFEED) {
1558                     if (tok != TOK_STR) {
1559                     include_syntax:
1560                         tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1561                     }
1562                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1563                     next();
1564                 }
1565                 c = '\"';
1566             } else {
1567                 int len;
1568                 while (tok != TOK_LINEFEED) {
1569                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1570                     next();
1571                 }
1572                 len = strlen(buf);
1573                 /* check syntax and remove '<>' */
1574                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1575                     goto include_syntax;
1576                 memmove(buf, buf + 1, len - 2);
1577                 buf[len - 2] = '\0';
1578                 c = '>';
1579             }
1580         }
1581
1582         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1583             tcc_error("#include recursion too deep");
1584         /* store current file in stack, but increment stack later below */
1585         *s1->include_stack_ptr = file;
1586
1587         n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1588         for (i = -2; i < n; ++i) {
1589             char buf1[sizeof file->filename];
1590             CachedInclude *e;
1591             BufferedFile **f;
1592             const char *path;
1593
1594             if (i == -2) {
1595                 /* check absolute include path */
1596                 if (!IS_ABSPATH(buf))
1597                     continue;
1598                 buf1[0] = 0;
1599                 i = n; /* force end loop */
1600
1601             } else if (i == -1) {
1602                 /* search in current dir if "header.h" */
1603                 if (c != '\"')
1604                     continue;
1605                 path = file->filename;
1606                 pstrncpy(buf1, path, tcc_basename(path) - path);
1607
1608             } else {
1609                 /* search in all the include paths */
1610                 if (i < s1->nb_include_paths)
1611                     path = s1->include_paths[i];
1612                 else
1613                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1614                 pstrcpy(buf1, sizeof(buf1), path);
1615                 pstrcat(buf1, sizeof(buf1), "/");
1616             }
1617
1618             pstrcat(buf1, sizeof(buf1), buf);
1619
1620             if (tok == TOK_INCLUDE_NEXT)
1621                 for (f = s1->include_stack_ptr; f >= s1->include_stack; --f)
1622                     if (0 == PATHCMP((*f)->filename, buf1)) {
1623 #ifdef INC_DEBUG
1624                         printf("%s: #include_next skipping %s\n", file->filename, buf1);
1625 #endif
1626                         goto include_trynext;
1627                     }
1628
1629             e = search_cached_include(s1, buf1);
1630             if (e && (define_find(e->ifndef_macro) || e->ifndef_macro == TOK_once)) {
1631                 /* no need to parse the include because the 'ifndef macro'
1632                    is defined */
1633 #ifdef INC_DEBUG
1634                 printf("%s: skipping cached %s\n", file->filename, buf1);
1635 #endif
1636                 goto include_done;
1637             }
1638
1639             if (tcc_open(s1, buf1) < 0)
1640 include_trynext:
1641                 continue;
1642
1643 #ifdef INC_DEBUG
1644             printf("%s: including %s\n", file->prev->filename, file->filename);
1645 #endif
1646             /* update target deps */
1647             dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1648                     tcc_strdup(buf1));
1649             /* push current file in stack */
1650             ++s1->include_stack_ptr;
1651             /* add include file debug info */
1652             if (s1->do_debug)
1653                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1654             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1655             ch = file->buf_ptr[0];
1656             goto the_end;
1657         }
1658         tcc_error("include file '%s' not found", buf);
1659 include_done:
1660         break;
1661     case TOK_IFNDEF:
1662         c = 1;
1663         goto do_ifdef;
1664     case TOK_IF:
1665         c = expr_preprocess();
1666         goto do_if;
1667     case TOK_IFDEF:
1668         c = 0;
1669     do_ifdef:
1670         next_nomacro();
1671         if (tok < TOK_IDENT)
1672             tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1673         if (is_bof) {
1674             if (c) {
1675 #ifdef INC_DEBUG
1676                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1677 #endif
1678                 file->ifndef_macro = tok;
1679             }
1680         }
1681         c = (define_find(tok) != 0) ^ c;
1682     do_if:
1683         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1684             tcc_error("memory full (ifdef)");
1685         *s1->ifdef_stack_ptr++ = c;
1686         goto test_skip;
1687     case TOK_ELSE:
1688         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1689             tcc_error("#else without matching #if");
1690         if (s1->ifdef_stack_ptr[-1] & 2)
1691             tcc_error("#else after #else");
1692         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1693         goto test_else;
1694     case TOK_ELIF:
1695         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1696             tcc_error("#elif without matching #if");
1697         c = s1->ifdef_stack_ptr[-1];
1698         if (c > 1)
1699             tcc_error("#elif after #else");
1700         /* last #if/#elif expression was true: we skip */
1701         if (c == 1)
1702             goto skip;
1703         c = expr_preprocess();
1704         s1->ifdef_stack_ptr[-1] = c;
1705     test_else:
1706         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1707             file->ifndef_macro = 0;
1708     test_skip:
1709         if (!(c & 1)) {
1710         skip:
1711             preprocess_skip();
1712             is_bof = 0;
1713             goto redo;
1714         }
1715         break;
1716     case TOK_ENDIF:
1717         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1718             tcc_error("#endif without matching #if");
1719         s1->ifdef_stack_ptr--;
1720         /* '#ifndef macro' was at the start of file. Now we check if
1721            an '#endif' is exactly at the end of file */
1722         if (file->ifndef_macro &&
1723             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1724             file->ifndef_macro_saved = file->ifndef_macro;
1725             /* need to set to zero to avoid false matches if another
1726                #ifndef at middle of file */
1727             file->ifndef_macro = 0;
1728             while (tok != TOK_LINEFEED)
1729                 next_nomacro();
1730             tok_flags |= TOK_FLAG_ENDIF;
1731             goto the_end;
1732         }
1733         break;
1734     case TOK_LINE:
1735         next();
1736         if (tok != TOK_CINT)
1737             tcc_error("A #line format is wrong");
1738     case TOK_PPNUM:
1739         if (tok != TOK_CINT) {
1740             char *p = tokc.cstr->data;
1741             tokc.i = strtoul(p, (char **)&p, 10);
1742         }
1743         i = file->line_num;
1744         file->line_num = tokc.i - 1;
1745         next();
1746         if (tok != TOK_LINEFEED) {
1747             if (tok != TOK_STR) {
1748                 if ((parse_flags & PARSE_FLAG_ASM_COMMENTS) == 0) {
1749                     file->line_num = i;
1750                     tcc_error("#line format is wrong");
1751                 }
1752                 break;
1753             }
1754             pstrcpy(file->filename, sizeof(file->filename),
1755                     (char *)tokc.cstr->data);
1756         }
1757         if (s1->do_debug)
1758             put_stabs(file->filename, N_BINCL, 0, 0, 0);
1759         break;
1760     case TOK_ERROR:
1761     case TOK_WARNING:
1762         c = tok;
1763         ch = file->buf_ptr[0];
1764         skip_spaces();
1765         q = buf;
1766         while (ch != '\n' && ch != CH_EOF) {
1767             if ((q - buf) < sizeof(buf) - 1)
1768                 *q++ = ch;
1769             if (ch == '\\') {
1770                 if (handle_stray_noerror() == 0)
1771                     --q;
1772             } else
1773                 inp();
1774         }
1775         *q = '\0';
1776         if (c == TOK_ERROR)
1777             tcc_error("#error %s", buf);
1778         else
1779             tcc_warning("#warning %s", buf);
1780         break;
1781     case TOK_PRAGMA:
1782         pragma_parse(s1);
1783         break;
1784     default:
1785         if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_PPNUM) {
1786             /* '!' is ignored to allow C scripts. numbers are ignored
1787                to emulate cpp behaviour */
1788         } else {
1789             if (!(parse_flags & PARSE_FLAG_ASM_COMMENTS))
1790                 tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1791             else {
1792                 /* this is a gas line comment in an 'S' file. */
1793                 file->buf_ptr = parse_line_comment(file->buf_ptr);
1794                 goto the_end;
1795             }
1796         }
1797         break;
1798     }
1799     /* ignore other preprocess commands or #! for C scripts */
1800     while (tok != TOK_LINEFEED)
1801         next_nomacro();
1802  the_end:
1803     parse_flags = saved_parse_flags;
1804 }
1805
1806 /* evaluate escape codes in a string. */
1807 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1808 {
1809     int c, n;
1810     const uint8_t *p;
1811
1812     p = buf;
1813     for(;;) {
1814         c = *p;
1815         if (c == '\0')
1816             break;
1817         if (c == '\\') {
1818             p++;
1819             /* escape */
1820             c = *p;
1821             switch(c) {
1822             case '0': case '1': case '2': case '3':
1823             case '4': case '5': case '6': case '7':
1824                 /* at most three octal digits */
1825                 n = c - '0';
1826                 p++;
1827                 c = *p;
1828                 if (isoct(c)) {
1829                     n = n * 8 + c - '0';
1830                     p++;
1831                     c = *p;
1832                     if (isoct(c)) {
1833                         n = n * 8 + c - '0';
1834                         p++;
1835                     }
1836                 }
1837                 c = n;
1838                 goto add_char_nonext;
1839             case 'x':
1840             case 'u':
1841             case 'U':
1842                 p++;
1843                 n = 0;
1844                 for(;;) {
1845                     c = *p;
1846                     if (c >= 'a' && c <= 'f')
1847                         c = c - 'a' + 10;
1848                     else if (c >= 'A' && c <= 'F')
1849                         c = c - 'A' + 10;
1850                     else if (isnum(c))
1851                         c = c - '0';
1852                     else
1853                         break;
1854                     n = n * 16 + c;
1855                     p++;
1856                 }
1857                 c = n;
1858                 goto add_char_nonext;
1859             case 'a':
1860                 c = '\a';
1861                 break;
1862             case 'b':
1863                 c = '\b';
1864                 break;
1865             case 'f':
1866                 c = '\f';
1867                 break;
1868             case 'n':
1869                 c = '\n';
1870                 break;
1871             case 'r':
1872                 c = '\r';
1873                 break;
1874             case 't':
1875                 c = '\t';
1876                 break;
1877             case 'v':
1878                 c = '\v';
1879                 break;
1880             case 'e':
1881                 if (!gnu_ext)
1882                     goto invalid_escape;
1883                 c = 27;
1884                 break;
1885             case '\'':
1886             case '\"':
1887             case '\\':
1888             case '?':
1889                 break;
1890             default:
1891             invalid_escape:
1892                 if (c >= '!' && c <= '~')
1893                     tcc_warning("unknown escape sequence: \'\\%c\'", c);
1894                 else
1895                     tcc_warning("unknown escape sequence: \'\\x%x\'", c);
1896                 break;
1897             }
1898         }
1899         p++;
1900     add_char_nonext:
1901         if (!is_long)
1902             cstr_ccat(outstr, c);
1903         else
1904             cstr_wccat(outstr, c);
1905     }
1906     /* add a trailing '\0' */
1907     if (!is_long)
1908         cstr_ccat(outstr, '\0');
1909     else
1910         cstr_wccat(outstr, '\0');
1911 }
1912
1913 /* we use 64 bit numbers */
1914 #define BN_SIZE 2
1915
1916 /* bn = (bn << shift) | or_val */
1917 static void bn_lshift(unsigned int *bn, int shift, int or_val)
1918 {
1919     int i;
1920     unsigned int v;
1921     for(i=0;i<BN_SIZE;i++) {
1922         v = bn[i];
1923         bn[i] = (v << shift) | or_val;
1924         or_val = v >> (32 - shift);
1925     }
1926 }
1927
1928 static void bn_zero(unsigned int *bn)
1929 {
1930     int i;
1931     for(i=0;i<BN_SIZE;i++) {
1932         bn[i] = 0;
1933     }
1934 }
1935
1936 /* parse number in null terminated string 'p' and return it in the
1937    current token */
1938 static void parse_number(const char *p)
1939 {
1940     int b, t, shift, frac_bits, s, exp_val, ch;
1941     char *q;
1942     unsigned int bn[BN_SIZE];
1943     double d;
1944
1945     /* number */
1946     q = token_buf;
1947     ch = *p++;
1948     t = ch;
1949     ch = *p++;
1950     *q++ = t;
1951     b = 10;
1952     if (t == '.') {
1953         goto float_frac_parse;
1954     } else if (t == '0') {
1955         if (ch == 'x' || ch == 'X') {
1956             q--;
1957             ch = *p++;
1958             b = 16;
1959         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
1960             q--;
1961             ch = *p++;
1962             b = 2;
1963         }
1964     }
1965     /* parse all digits. cannot check octal numbers at this stage
1966        because of floating point constants */
1967     while (1) {
1968         if (ch >= 'a' && ch <= 'f')
1969             t = ch - 'a' + 10;
1970         else if (ch >= 'A' && ch <= 'F')
1971             t = ch - 'A' + 10;
1972         else if (isnum(ch))
1973             t = ch - '0';
1974         else
1975             break;
1976         if (t >= b)
1977             break;
1978         if (q >= token_buf + STRING_MAX_SIZE) {
1979         num_too_long:
1980             tcc_error("number too long");
1981         }
1982         *q++ = ch;
1983         ch = *p++;
1984     }
1985     if (ch == '.' ||
1986         ((ch == 'e' || ch == 'E') && b == 10) ||
1987         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
1988         if (b != 10) {
1989             /* NOTE: strtox should support that for hexa numbers, but
1990                non ISOC99 libcs do not support it, so we prefer to do
1991                it by hand */
1992             /* hexadecimal or binary floats */
1993             /* XXX: handle overflows */
1994             *q = '\0';
1995             if (b == 16)
1996                 shift = 4;
1997             else
1998                 shift = 1;
1999             bn_zero(bn);
2000             q = token_buf;
2001             while (1) {
2002                 t = *q++;
2003                 if (t == '\0') {
2004                     break;
2005                 } else if (t >= 'a') {
2006                     t = t - 'a' + 10;
2007                 } else if (t >= 'A') {
2008                     t = t - 'A' + 10;
2009                 } else {
2010                     t = t - '0';
2011                 }
2012                 bn_lshift(bn, shift, t);
2013             }
2014             frac_bits = 0;
2015             if (ch == '.') {
2016                 ch = *p++;
2017                 while (1) {
2018                     t = ch;
2019                     if (t >= 'a' && t <= 'f') {
2020                         t = t - 'a' + 10;
2021                     } else if (t >= 'A' && t <= 'F') {
2022                         t = t - 'A' + 10;
2023                     } else if (t >= '0' && t <= '9') {
2024                         t = t - '0';
2025                     } else {
2026                         break;
2027                     }
2028                     if (t >= b)
2029                         tcc_error("invalid digit");
2030                     bn_lshift(bn, shift, t);
2031                     frac_bits += shift;
2032                     ch = *p++;
2033                 }
2034             }
2035             if (ch != 'p' && ch != 'P')
2036                 expect("exponent");
2037             ch = *p++;
2038             s = 1;
2039             exp_val = 0;
2040             if (ch == '+') {
2041                 ch = *p++;
2042             } else if (ch == '-') {
2043                 s = -1;
2044                 ch = *p++;
2045             }
2046             if (ch < '0' || ch > '9')
2047                 expect("exponent digits");
2048             while (ch >= '0' && ch <= '9') {
2049                 exp_val = exp_val * 10 + ch - '0';
2050                 ch = *p++;
2051             }
2052             exp_val = exp_val * s;
2053
2054             /* now we can generate the number */
2055             /* XXX: should patch directly float number */
2056             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
2057             d = ldexp(d, exp_val - frac_bits);
2058             t = toup(ch);
2059             if (t == 'F') {
2060                 ch = *p++;
2061                 tok = TOK_CFLOAT;
2062                 /* float : should handle overflow */
2063                 tokc.f = (float)d;
2064             } else if (t == 'L') {
2065                 ch = *p++;
2066 #ifdef TCC_TARGET_PE
2067                 tok = TOK_CDOUBLE;
2068                 tokc.d = d;
2069 #else
2070                 tok = TOK_CLDOUBLE;
2071                 /* XXX: not large enough */
2072                 tokc.ld = (long double)d;
2073 #endif
2074             } else {
2075                 tok = TOK_CDOUBLE;
2076                 tokc.d = d;
2077             }
2078         } else {
2079             /* decimal floats */
2080             if (ch == '.') {
2081                 if (q >= token_buf + STRING_MAX_SIZE)
2082                     goto num_too_long;
2083                 *q++ = ch;
2084                 ch = *p++;
2085             float_frac_parse:
2086                 while (ch >= '0' && ch <= '9') {
2087                     if (q >= token_buf + STRING_MAX_SIZE)
2088                         goto num_too_long;
2089                     *q++ = ch;
2090                     ch = *p++;
2091                 }
2092             }
2093             if (ch == 'e' || ch == 'E') {
2094                 if (q >= token_buf + STRING_MAX_SIZE)
2095                     goto num_too_long;
2096                 *q++ = ch;
2097                 ch = *p++;
2098                 if (ch == '-' || ch == '+') {
2099                     if (q >= token_buf + STRING_MAX_SIZE)
2100                         goto num_too_long;
2101                     *q++ = ch;
2102                     ch = *p++;
2103                 }
2104                 if (ch < '0' || ch > '9')
2105                     expect("exponent digits");
2106                 while (ch >= '0' && ch <= '9') {
2107                     if (q >= token_buf + STRING_MAX_SIZE)
2108                         goto num_too_long;
2109                     *q++ = ch;
2110                     ch = *p++;
2111                 }
2112             }
2113             *q = '\0';
2114             t = toup(ch);
2115             errno = 0;
2116             if (t == 'F') {
2117                 ch = *p++;
2118                 tok = TOK_CFLOAT;
2119                 tokc.f = strtof(token_buf, NULL);
2120             } else if (t == 'L') {
2121                 ch = *p++;
2122 #ifdef TCC_TARGET_PE
2123                 tok = TOK_CDOUBLE;
2124                 tokc.d = strtod(token_buf, NULL);
2125 #else
2126                 tok = TOK_CLDOUBLE;
2127                 tokc.ld = strtold(token_buf, NULL);
2128 #endif
2129             } else {
2130                 tok = TOK_CDOUBLE;
2131                 tokc.d = strtod(token_buf, NULL);
2132             }
2133         }
2134     } else {
2135         unsigned long long n, n1;
2136         int lcount, ucount, must_64bit;
2137         const char *p1;
2138
2139         /* integer number */
2140         *q = '\0';
2141         q = token_buf;
2142         if (b == 10 && *q == '0') {
2143             b = 8;
2144             q++;
2145         }
2146         n = 0;
2147         while(1) {
2148             t = *q++;
2149             /* no need for checks except for base 10 / 8 errors */
2150             if (t == '\0')
2151                 break;
2152             else if (t >= 'a')
2153                 t = t - 'a' + 10;
2154             else if (t >= 'A')
2155                 t = t - 'A' + 10;
2156             else
2157                 t = t - '0';
2158             if (t >= b)
2159                 tcc_error("invalid digit");
2160             n1 = n;
2161             n = n * b + t;
2162             /* detect overflow */
2163             /* XXX: this test is not reliable */
2164             if (n < n1)
2165                 tcc_error("integer constant overflow");
2166         }
2167
2168         /* Determine the characteristics (unsigned and/or 64bit) the type of
2169            the constant must have according to the constant suffix(es) */
2170         lcount = ucount = must_64bit = 0;
2171         p1 = p;
2172         for(;;) {
2173             t = toup(ch);
2174             if (t == 'L') {
2175                 if (lcount >= 2)
2176                     tcc_error("three 'l's in integer constant");
2177                 if (lcount && *(p - 1) != ch)
2178                     tcc_error("incorrect integer suffix: %s", p1);
2179                 lcount++;
2180 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
2181                 if (lcount == 2)
2182 #endif
2183                     must_64bit = 1;
2184                 ch = *p++;
2185             } else if (t == 'U') {
2186                 if (ucount >= 1)
2187                     tcc_error("two 'u's in integer constant");
2188                 ucount++;
2189                 ch = *p++;
2190             } else {
2191                 break;
2192             }
2193         }
2194
2195         /* Whether 64 bits are needed to hold the constant's value */
2196         if (n & 0xffffffff00000000LL || must_64bit) {
2197             tok = TOK_CLLONG;
2198             n1 = n >> 32;
2199         } else {
2200             tok = TOK_CINT;
2201             n1 = n;
2202         }
2203
2204         /* Whether type must be unsigned to hold the constant's value */
2205         if (ucount || ((n1 >> 31) && (b != 10))) {
2206             if (tok == TOK_CLLONG)
2207                 tok = TOK_CULLONG;
2208             else
2209                 tok = TOK_CUINT;
2210         /* If decimal and no unsigned suffix, bump to 64 bits or throw error */
2211         } else if (n1 >> 31) {
2212             if (tok == TOK_CINT)
2213                 tok = TOK_CLLONG;
2214             else
2215                 tcc_error("integer constant overflow");
2216         }
2217
2218         if (tok == TOK_CINT || tok == TOK_CUINT)
2219             tokc.ui = n;
2220         else
2221             tokc.ull = n;
2222     }
2223     if (ch)
2224         tcc_error("invalid number\n");
2225 }
2226
2227
2228 #define PARSE2(c1, tok1, c2, tok2)              \
2229     case c1:                                    \
2230         PEEKC(c, p);                            \
2231         if (c == c2) {                          \
2232             p++;                                \
2233             tok = tok2;                         \
2234         } else {                                \
2235             tok = tok1;                         \
2236         }                                       \
2237         break;
2238
2239 /* return next token without macro substitution */
2240 static inline void next_nomacro1(void)
2241 {
2242     int t, c, is_long;
2243     TokenSym *ts;
2244     uint8_t *p, *p1;
2245     unsigned int h;
2246
2247     p = file->buf_ptr;
2248  redo_no_start:
2249     c = *p;
2250     switch(c) {
2251     case ' ':
2252     case '\t':
2253         tok = c;
2254         p++;
2255         goto keep_tok_flags;
2256     case '\f':
2257     case '\v':
2258     case '\r':
2259         p++;
2260         goto redo_no_start;
2261     case '\\':
2262         /* first look if it is in fact an end of buffer */
2263         if (p >= file->buf_end) {
2264             file->buf_ptr = p;
2265             handle_eob();
2266             p = file->buf_ptr;
2267             if (p >= file->buf_end)
2268                 goto parse_eof;
2269             else
2270                 goto redo_no_start;
2271         } else {
2272             file->buf_ptr = p;
2273             ch = *p;
2274             handle_stray();
2275             p = file->buf_ptr;
2276             goto redo_no_start;
2277         }
2278     parse_eof:
2279         {
2280             TCCState *s1 = tcc_state;
2281             if ((parse_flags & PARSE_FLAG_LINEFEED)
2282                 && !(tok_flags & TOK_FLAG_EOF)) {
2283                 tok_flags |= TOK_FLAG_EOF;
2284                 tok = TOK_LINEFEED;
2285                 goto keep_tok_flags;
2286             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2287                 tok = TOK_EOF;
2288             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2289                 tcc_error("missing #endif");
2290             } else if (s1->include_stack_ptr == s1->include_stack) {
2291                 /* no include left : end of file. */
2292                 tok = TOK_EOF;
2293             } else {
2294                 tok_flags &= ~TOK_FLAG_EOF;
2295                 /* pop include file */
2296
2297                 /* test if previous '#endif' was after a #ifdef at
2298                    start of file */
2299                 if (tok_flags & TOK_FLAG_ENDIF) {
2300 #ifdef INC_DEBUG
2301                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2302 #endif
2303                     add_cached_include(s1, file->filename, file->ifndef_macro_saved);
2304                     tok_flags &= ~TOK_FLAG_ENDIF;
2305                 }
2306
2307                 /* add end of include file debug info */
2308                 if (tcc_state->do_debug) {
2309                     put_stabd(N_EINCL, 0, 0);
2310                 }
2311                 /* pop include stack */
2312                 tcc_close();
2313                 s1->include_stack_ptr--;
2314                 p = file->buf_ptr;
2315                 goto redo_no_start;
2316             }
2317         }
2318         break;
2319
2320     case '\n':
2321         file->line_num++;
2322         tok_flags |= TOK_FLAG_BOL;
2323         p++;
2324 maybe_newline:
2325         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2326             goto redo_no_start;
2327         tok = TOK_LINEFEED;
2328         goto keep_tok_flags;
2329
2330     case '#':
2331         /* XXX: simplify */
2332         PEEKC(c, p);
2333         if (is_space(c) && (parse_flags & PARSE_FLAG_ASM_FILE)) {
2334             p = parse_line_comment(p);
2335             goto redo_no_start;
2336         }
2337         else
2338         if ((tok_flags & TOK_FLAG_BOL) &&
2339             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2340             file->buf_ptr = p;
2341             preprocess(tok_flags & TOK_FLAG_BOF);
2342             p = file->buf_ptr;
2343             goto maybe_newline;
2344         } else {
2345             if (c == '#') {
2346                 p++;
2347                 tok = TOK_TWOSHARPS;
2348             } else {
2349                 if (parse_flags & PARSE_FLAG_ASM_COMMENTS) {
2350                     p = parse_line_comment(p - 1);
2351                     goto redo_no_start;
2352                 } else {
2353                     tok = '#';
2354                 }
2355             }
2356         }
2357         break;
2358
2359     /* dollar is allowed to start identifiers when not parsing asm */
2360     case '$':
2361         if (!tcc_state->dollars_in_identifiers
2362         || (parse_flags & PARSE_FLAG_ASM_FILE)) goto parse_simple;
2363
2364     case 'a': case 'b': case 'c': case 'd':
2365     case 'e': case 'f': case 'g': case 'h':
2366     case 'i': case 'j': case 'k': case 'l':
2367     case 'm': case 'n': case 'o': case 'p':
2368     case 'q': case 'r': case 's': case 't':
2369     case 'u': case 'v': case 'w': case 'x':
2370     case 'y': case 'z':
2371     case 'A': case 'B': case 'C': case 'D':
2372     case 'E': case 'F': case 'G': case 'H':
2373     case 'I': case 'J': case 'K':
2374     case 'M': case 'N': case 'O': case 'P':
2375     case 'Q': case 'R': case 'S': case 'T':
2376     case 'U': case 'V': case 'W': case 'X':
2377     case 'Y': case 'Z':
2378     case '_':
2379     parse_ident_fast:
2380         p1 = p;
2381         h = TOK_HASH_INIT;
2382         h = TOK_HASH_FUNC(h, c);
2383         p++;
2384         for(;;) {
2385             c = *p;
2386             if (!isidnum_table[c-CH_EOF]
2387             && (tcc_state->dollars_in_identifiers ? (c != '$') : 1))
2388                 break;
2389             h = TOK_HASH_FUNC(h, c);
2390             p++;
2391         }
2392         if (c != '\\') {
2393             TokenSym **pts;
2394             int len;
2395
2396             /* fast case : no stray found, so we have the full token
2397                and we have already hashed it */
2398             len = p - p1;
2399             h &= (TOK_HASH_SIZE - 1);
2400             pts = &hash_ident[h];
2401             for(;;) {
2402                 ts = *pts;
2403                 if (!ts)
2404                     break;
2405                 if (ts->len == len && !memcmp(ts->str, p1, len))
2406                     goto token_found;
2407                 pts = &(ts->hash_next);
2408             }
2409             ts = tok_alloc_new(pts, (char *) p1, len);
2410         token_found: ;
2411         } else {
2412             /* slower case */
2413             cstr_reset(&tokcstr);
2414
2415             while (p1 < p) {
2416                 cstr_ccat(&tokcstr, *p1);
2417                 p1++;
2418             }
2419             p--;
2420             PEEKC(c, p);
2421         parse_ident_slow:
2422             while (isidnum_table[c-CH_EOF]
2423             || (tcc_state->dollars_in_identifiers ? (c == '$') : 0)) {
2424                 cstr_ccat(&tokcstr, c);
2425                 PEEKC(c, p);
2426             }
2427             ts = tok_alloc(tokcstr.data, tokcstr.size);
2428         }
2429         tok = ts->tok;
2430         break;
2431     case 'L':
2432         t = p[1];
2433         if (t != '\\' && t != '\'' && t != '\"') {
2434             /* fast case */
2435             goto parse_ident_fast;
2436         } else {
2437             PEEKC(c, p);
2438             if (c == '\'' || c == '\"') {
2439                 is_long = 1;
2440                 goto str_const;
2441             } else {
2442                 cstr_reset(&tokcstr);
2443                 cstr_ccat(&tokcstr, 'L');
2444                 goto parse_ident_slow;
2445             }
2446         }
2447         break;
2448     case '0': case '1': case '2': case '3':
2449     case '4': case '5': case '6': case '7':
2450     case '8': case '9':
2451
2452         cstr_reset(&tokcstr);
2453         /* after the first digit, accept digits, alpha, '.' or sign if
2454            prefixed by 'eEpP' */
2455     parse_num:
2456         for(;;) {
2457             t = c;
2458             cstr_ccat(&tokcstr, c);
2459             PEEKC(c, p);
2460             if (!(isnum(c) || isid(c) || c == '.' ||
2461                   ((c == '+' || c == '-') &&
2462                    (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2463                 break;
2464         }
2465         /* We add a trailing '\0' to ease parsing */
2466         cstr_ccat(&tokcstr, '\0');
2467         tokc.cstr = &tokcstr;
2468         tok = TOK_PPNUM;
2469         break;
2470     case '.':
2471         /* special dot handling because it can also start a number */
2472         PEEKC(c, p);
2473         if (isnum(c)) {
2474             cstr_reset(&tokcstr);
2475             cstr_ccat(&tokcstr, '.');
2476             goto parse_num;
2477         } else if (c == '.') {
2478             PEEKC(c, p);
2479             if (c != '.') {
2480                 if ((parse_flags & PARSE_FLAG_ASM_COMMENTS) == 0)
2481                     expect("'.'");
2482                 tok = '.';
2483                 break;
2484             }
2485             PEEKC(c, p);
2486             tok = TOK_DOTS;
2487         } else {
2488             tok = '.';
2489         }
2490         break;
2491     case '\'':
2492     case '\"':
2493         is_long = 0;
2494     str_const:
2495         {
2496             CString str;
2497             int sep;
2498
2499             sep = c;
2500
2501             /* parse the string */
2502             cstr_new(&str);
2503             p = parse_pp_string(p, sep, &str);
2504             cstr_ccat(&str, '\0');
2505
2506             /* eval the escape (should be done as TOK_PPNUM) */
2507             cstr_reset(&tokcstr);
2508             parse_escape_string(&tokcstr, str.data, is_long);
2509             cstr_free(&str);
2510
2511             if (sep == '\'') {
2512                 int char_size;
2513                 /* XXX: make it portable */
2514                 if (!is_long)
2515                     char_size = 1;
2516                 else
2517                     char_size = sizeof(nwchar_t);
2518                 if (tokcstr.size <= char_size)
2519                     tcc_error("empty character constant");
2520                 if (tokcstr.size > 2 * char_size)
2521                     tcc_warning("multi-character character constant");
2522                 if (!is_long) {
2523                     tokc.i = *(int8_t *)tokcstr.data;
2524                     tok = TOK_CCHAR;
2525                 } else {
2526                     tokc.i = *(nwchar_t *)tokcstr.data;
2527                     tok = TOK_LCHAR;
2528                 }
2529             } else {
2530                 tokc.cstr = &tokcstr;
2531                 if (!is_long)
2532                     tok = TOK_STR;
2533                 else
2534                     tok = TOK_LSTR;
2535             }
2536         }
2537         break;
2538
2539     case '<':
2540         PEEKC(c, p);
2541         if (c == '=') {
2542             p++;
2543             tok = TOK_LE;
2544         } else if (c == '<') {
2545             PEEKC(c, p);
2546             if (c == '=') {
2547                 p++;
2548                 tok = TOK_A_SHL;
2549             } else {
2550                 tok = TOK_SHL;
2551             }
2552         } else {
2553             tok = TOK_LT;
2554         }
2555         break;
2556
2557     case '>':
2558         PEEKC(c, p);
2559         if (c == '=') {
2560             p++;
2561             tok = TOK_GE;
2562         } else if (c == '>') {
2563             PEEKC(c, p);
2564             if (c == '=') {
2565                 p++;
2566                 tok = TOK_A_SAR;
2567             } else {
2568                 tok = TOK_SAR;
2569             }
2570         } else {
2571             tok = TOK_GT;
2572         }
2573         break;
2574
2575     case '&':
2576         PEEKC(c, p);
2577         if (c == '&') {
2578             p++;
2579             tok = TOK_LAND;
2580         } else if (c == '=') {
2581             p++;
2582             tok = TOK_A_AND;
2583         } else {
2584             tok = '&';
2585         }
2586         break;
2587
2588     case '|':
2589         PEEKC(c, p);
2590         if (c == '|') {
2591             p++;
2592             tok = TOK_LOR;
2593         } else if (c == '=') {
2594             p++;
2595             tok = TOK_A_OR;
2596         } else {
2597             tok = '|';
2598         }
2599         break;
2600
2601     case '+':
2602         PEEKC(c, p);
2603         if (c == '+') {
2604             p++;
2605             tok = TOK_INC;
2606         } else if (c == '=') {
2607             p++;
2608             tok = TOK_A_ADD;
2609         } else {
2610             tok = '+';
2611         }
2612         break;
2613
2614     case '-':
2615         PEEKC(c, p);
2616         if (c == '-') {
2617             p++;
2618             tok = TOK_DEC;
2619         } else if (c == '=') {
2620             p++;
2621             tok = TOK_A_SUB;
2622         } else if (c == '>') {
2623             p++;
2624             tok = TOK_ARROW;
2625         } else {
2626             tok = '-';
2627         }
2628         break;
2629
2630     PARSE2('!', '!', '=', TOK_NE)
2631     PARSE2('=', '=', '=', TOK_EQ)
2632     PARSE2('*', '*', '=', TOK_A_MUL)
2633     PARSE2('%', '%', '=', TOK_A_MOD)
2634     PARSE2('^', '^', '=', TOK_A_XOR)
2635
2636         /* comments or operator */
2637     case '/':
2638         PEEKC(c, p);
2639         if (c == '*') {
2640             p = parse_comment(p);
2641             /* comments replaced by a blank */
2642             tok = ' ';
2643             goto keep_tok_flags;
2644         } else if (c == '/') {
2645             p = parse_line_comment(p);
2646             tok = ' ';
2647             goto keep_tok_flags;
2648         } else if (c == '=') {
2649             p++;
2650             tok = TOK_A_DIV;
2651         } else {
2652             tok = '/';
2653         }
2654         break;
2655
2656         /* simple tokens */
2657     case '(':
2658     case ')':
2659     case '[':
2660     case ']':
2661     case '{':
2662     case '}':
2663     case ',':
2664     case ';':
2665     case ':':
2666     case '?':
2667     case '~':
2668     case '@': /* only used in assembler */
2669     parse_simple:
2670         tok = c;
2671         p++;
2672         break;
2673     default:
2674         if ((parse_flags & PARSE_FLAG_ASM_FILE) == 0)
2675             tcc_error("unrecognized character \\x%02x", c);
2676         else {
2677             tok = ' ';
2678             p++;
2679         }
2680         break;
2681     }
2682     tok_flags = 0;
2683 keep_tok_flags:
2684     file->buf_ptr = p;
2685 #if defined(PARSE_DEBUG)
2686     printf("token = %s\n", get_tok_str(tok, &tokc));
2687 #endif
2688 }
2689
2690 /* return next token without macro substitution. Can read input from
2691    macro_ptr buffer */
2692 static void next_nomacro_spc(void)
2693 {
2694     if (macro_ptr) {
2695     redo:
2696         tok = *macro_ptr;
2697         if (tok) {
2698             TOK_GET(&tok, &macro_ptr, &tokc);
2699             if (tok == TOK_LINENUM) {
2700                 file->line_num = tokc.i;
2701                 goto redo;
2702             }
2703         }
2704     } else {
2705         next_nomacro1();
2706     }
2707 }
2708
2709 ST_FUNC void next_nomacro(void)
2710 {
2711     do {
2712         next_nomacro_spc();
2713     } while (is_space(tok));
2714 }
2715
2716 /* substitute arguments in replacement lists in macro_str by the values in
2717    args (field d) and return allocated string */
2718 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2719 {
2720     int last_tok, t, spc;
2721     const int *st;
2722     Sym *s;
2723     CValue cval;
2724     TokenString str;
2725     CString cstr;
2726
2727     tok_str_new(&str);
2728     last_tok = 0;
2729     while(1) {
2730         TOK_GET(&t, &macro_str, &cval);
2731         if (!t)
2732             break;
2733         if (t == '#') {
2734             /* stringize */
2735             TOK_GET(&t, &macro_str, &cval);
2736             if (!t)
2737                 break;
2738             s = sym_find2(args, t);
2739             if (s) {
2740                 cstr_new(&cstr);
2741                 st = s->d;
2742                 spc = 0;
2743                 while (*st) {
2744                     TOK_GET(&t, &st, &cval);
2745                     if (!check_space(t, &spc))
2746                         cstr_cat(&cstr, get_tok_str(t, &cval));
2747                 }
2748                 cstr.size -= spc;
2749                 cstr_ccat(&cstr, '\0');
2750 #ifdef PP_DEBUG
2751                 printf("stringize: %s\n", (char *)cstr.data);
2752 #endif
2753                 /* add string */
2754                 cval.cstr = &cstr;
2755                 tok_str_add2(&str, TOK_STR, &cval);
2756                 cstr_free(&cstr);
2757             } else {
2758                 tok_str_add2(&str, t, &cval);
2759             }
2760         } else if (t >= TOK_IDENT) {
2761             s = sym_find2(args, t);
2762             if (s) {
2763                 st = s->d;
2764                 /* if '##' is present before or after, no arg substitution */
2765                 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2766                     /* special case for var arg macros : ## eats the
2767                        ',' if empty VA_ARGS variable. */
2768                     /* XXX: test of the ',' is not 100%
2769                        reliable. should fix it to avoid security
2770                        problems */
2771                     if (gnu_ext && s->type.t &&
2772                         last_tok == TOK_TWOSHARPS &&
2773                         str.len >= 2 && str.str[str.len - 2] == ',') {
2774                         if (*st == TOK_PLCHLDR) {
2775                             /* suppress ',' '##' */
2776                             str.len -= 2;
2777                         } else {
2778                             /* suppress '##' and add variable */
2779                             str.len--;
2780                             goto add_var;
2781                         }
2782                     } else {
2783                         int t1;
2784                     add_var:
2785                         for(;;) {
2786                             TOK_GET(&t1, &st, &cval);
2787                             if (!t1)
2788                                 break;
2789                             tok_str_add2(&str, t1, &cval);
2790                         }
2791                     }
2792                 } else if (*st != TOK_PLCHLDR) {
2793                     /* NOTE: the stream cannot be read when macro
2794                        substituing an argument */
2795                     macro_subst(&str, nested_list, st, NULL);
2796                 }
2797             } else {
2798                 tok_str_add(&str, t);
2799             }
2800         } else {
2801             tok_str_add2(&str, t, &cval);
2802         }
2803         last_tok = t;
2804     }
2805     tok_str_add(&str, 0);
2806     return str.str;
2807 }
2808
2809 static char const ab_month_name[12][4] =
2810 {
2811     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2812     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2813 };
2814
2815 /* do macro substitution of current token with macro 's' and add
2816    result to (tok_str,tok_len). 'nested_list' is the list of all
2817    macros we got inside to avoid recursing. Return non zero if no
2818    substitution needs to be done */
2819 static int macro_subst_tok(TokenString *tok_str,
2820                            Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2821 {
2822     Sym *args, *sa, *sa1;
2823     int mstr_allocated, parlevel, *mstr, t, t1, spc;
2824     const int *p;
2825     TokenString str;
2826     char *cstrval;
2827     CValue cval;
2828     CString cstr;
2829     char buf[32];
2830
2831     /* if symbol is a macro, prepare substitution */
2832     /* special macros */
2833     if (tok == TOK___LINE__) {
2834         snprintf(buf, sizeof(buf), "%d", file->line_num);
2835         cstrval = buf;
2836         t1 = TOK_PPNUM;
2837         goto add_cstr1;
2838     } else if (tok == TOK___FILE__) {
2839         cstrval = file->filename;
2840         goto add_cstr;
2841     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2842         time_t ti;
2843         struct tm *tm;
2844
2845         time(&ti);
2846         tm = localtime(&ti);
2847         if (tok == TOK___DATE__) {
2848             snprintf(buf, sizeof(buf), "%s %2d %d",
2849                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2850         } else {
2851             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2852                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2853         }
2854         cstrval = buf;
2855     add_cstr:
2856         t1 = TOK_STR;
2857     add_cstr1:
2858         cstr_new(&cstr);
2859         cstr_cat(&cstr, cstrval);
2860         cstr_ccat(&cstr, '\0');
2861         cval.cstr = &cstr;
2862         tok_str_add2(tok_str, t1, &cval);
2863         cstr_free(&cstr);
2864     } else {
2865         mstr = s->d;
2866         mstr_allocated = 0;
2867         if (s->type.t == MACRO_FUNC) {
2868             /* NOTE: we do not use next_nomacro to avoid eating the
2869                next token. XXX: find better solution */
2870         redo:
2871             if (macro_ptr) {
2872                 p = macro_ptr;
2873                 while (is_space(t = *p) || TOK_LINEFEED == t)
2874                     ++p;
2875                 if (t == 0 && can_read_stream) {
2876                     /* end of macro stream: we must look at the token
2877                        after in the file */
2878                     struct macro_level *ml = *can_read_stream;
2879                     macro_ptr = NULL;
2880                     if (ml)
2881                     {
2882                         macro_ptr = ml->p;
2883                         ml->p = NULL;
2884                         *can_read_stream = ml -> prev;
2885                     }
2886                     /* also, end of scope for nested defined symbol */
2887                     (*nested_list)->v = -1;
2888                     goto redo;
2889                 }
2890             } else {
2891                 ch = file->buf_ptr[0];
2892                 while (is_space(ch) || ch == '\n' || ch == '/')
2893                   {
2894                     if (ch == '/')
2895                       {
2896                         int c;
2897                         uint8_t *p = file->buf_ptr;
2898                         PEEKC(c, p);
2899                         if (c == '*') {
2900                             p = parse_comment(p);
2901                             file->buf_ptr = p - 1;
2902                         } else if (c == '/') {
2903                             p = parse_line_comment(p);
2904                             file->buf_ptr = p - 1;
2905                         } else
2906                           break;
2907                       }
2908                     cinp();
2909                   }
2910                 t = ch;
2911             }
2912             if (t != '(') /* no macro subst */
2913                 return -1;
2914
2915             /* argument macro */
2916             next_nomacro();
2917             next_nomacro();
2918             args = NULL;
2919             sa = s->next;
2920             /* NOTE: empty args are allowed, except if no args */
2921             for(;;) {
2922                 /* handle '()' case */
2923                 if (!args && !sa && tok == ')')
2924                     break;
2925                 if (!sa)
2926                     tcc_error("macro '%s' used with too many args",
2927                           get_tok_str(s->v, 0));
2928                 tok_str_new(&str);
2929                 parlevel = spc = 0;
2930                 /* NOTE: non zero sa->t indicates VA_ARGS */
2931                 while ((parlevel > 0 ||
2932                         (tok != ')' &&
2933                          (tok != ',' || sa->type.t))) &&
2934                        tok != -1) {
2935                     if (tok == '(')
2936                         parlevel++;
2937                     else if (tok == ')')
2938                         parlevel--;
2939                     if (tok == TOK_LINEFEED)
2940                         tok = ' ';
2941                     if (!check_space(tok, &spc))
2942                         tok_str_add2(&str, tok, &tokc);
2943                     next_nomacro_spc();
2944                 }
2945                 if (!str.len)
2946                     tok_str_add(&str, TOK_PLCHLDR);
2947                 str.len -= spc;
2948                 tok_str_add(&str, 0);
2949                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
2950                 sa1->d = str.str;
2951                 sa = sa->next;
2952                 if (tok == ')') {
2953                     /* special case for gcc var args: add an empty
2954                        var arg argument if it is omitted */
2955                     if (sa && sa->type.t && gnu_ext)
2956                         continue;
2957                     else
2958                         break;
2959                 }
2960                 if (tok != ',')
2961                     expect(",");
2962                 next_nomacro();
2963             }
2964             if (sa) {
2965                 tcc_error("macro '%s' used with too few args",
2966                       get_tok_str(s->v, 0));
2967             }
2968
2969             /* now subst each arg */
2970             mstr = macro_arg_subst(nested_list, mstr, args);
2971             /* free memory */
2972             sa = args;
2973             while (sa) {
2974                 sa1 = sa->prev;
2975                 tok_str_free(sa->d);
2976                 sym_free(sa);
2977                 sa = sa1;
2978             }
2979             mstr_allocated = 1;
2980         }
2981         sym_push2(nested_list, s->v, 0, 0);
2982         macro_subst(tok_str, nested_list, mstr, can_read_stream);
2983         /* pop nested defined symbol */
2984         sa1 = *nested_list;
2985         *nested_list = sa1->prev;
2986         sym_free(sa1);
2987         if (mstr_allocated)
2988             tok_str_free(mstr);
2989     }
2990     return 0;
2991 }
2992
2993 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2994    return the resulting string (which must be freed). */
2995 static inline int *macro_twosharps(const int *macro_str)
2996 {
2997     const int *ptr;
2998     int t;
2999     TokenString macro_str1;
3000     CString cstr;
3001     int n, start_of_nosubsts;
3002
3003     /* we search the first '##' */
3004     for(ptr = macro_str;;) {
3005         CValue cval;
3006         TOK_GET(&t, &ptr, &cval);
3007         if (t == TOK_TWOSHARPS)
3008             break;
3009         /* nothing more to do if end of string */
3010         if (t == 0)
3011             return NULL;
3012     }
3013
3014     /* we saw '##', so we need more processing to handle it */
3015     start_of_nosubsts = -1;
3016     tok_str_new(&macro_str1);
3017     for(ptr = macro_str;;) {
3018         TOK_GET(&tok, &ptr, &tokc);
3019         if (tok == 0)
3020             break;
3021         if (tok == TOK_TWOSHARPS)
3022             continue;
3023         if (tok == TOK_NOSUBST && start_of_nosubsts < 0)
3024             start_of_nosubsts = macro_str1.len;
3025         while (*ptr == TOK_TWOSHARPS) {
3026             /* given 'a##b', remove nosubsts preceding 'a' */
3027             if (start_of_nosubsts >= 0)
3028                 macro_str1.len = start_of_nosubsts;
3029             /* given 'a##b', skip '##' */
3030             t = *++ptr;
3031             /* given 'a##b', remove nosubsts preceding 'b' */
3032             while (t == TOK_NOSUBST)
3033                 t = *++ptr;
3034             if (t && t != TOK_TWOSHARPS) {
3035                 CValue cval;
3036                 TOK_GET(&t, &ptr, &cval);
3037                 /* We concatenate the two tokens */
3038                 cstr_new(&cstr);
3039                 if (tok != TOK_PLCHLDR)
3040                     cstr_cat(&cstr, get_tok_str(tok, &tokc));
3041                 n = cstr.size;
3042                 if (t != TOK_PLCHLDR || tok == TOK_PLCHLDR)
3043                     cstr_cat(&cstr, get_tok_str(t, &cval));
3044                 cstr_ccat(&cstr, '\0');
3045
3046                 tcc_open_bf(tcc_state, ":paste:", cstr.size);
3047                 memcpy(file->buffer, cstr.data, cstr.size);
3048                 for (;;) {
3049                     next_nomacro1();
3050                     if (0 == *file->buf_ptr)
3051                         break;
3052                     tok_str_add2(&macro_str1, tok, &tokc);
3053                     tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
3054                         n, cstr.data, (char*)cstr.data + n);
3055                 }
3056                 tcc_close();
3057                 cstr_free(&cstr);
3058             }
3059         }
3060         if (tok != TOK_NOSUBST) {
3061             tok_str_add2(&macro_str1, tok, &tokc);
3062             tok = ' ';
3063             start_of_nosubsts = -1;
3064         }
3065         tok_str_add2(&macro_str1, tok, &tokc);
3066     }
3067     tok_str_add(&macro_str1, 0);
3068     return macro_str1.str;
3069 }
3070
3071
3072 /* do macro substitution of macro_str and add result to
3073    (tok_str,tok_len). 'nested_list' is the list of all macros we got
3074    inside to avoid recursing. */
3075 static void macro_subst(TokenString *tok_str, Sym **nested_list,
3076                         const int *macro_str, struct macro_level ** can_read_stream)
3077 {
3078     Sym *s;
3079     int *macro_str1;
3080     const int *ptr;
3081     int t, ret, spc;
3082     CValue cval;
3083     struct macro_level ml;
3084     int force_blank;
3085
3086     /* first scan for '##' operator handling */
3087     ptr = macro_str;
3088     macro_str1 = macro_twosharps(ptr);
3089
3090     if (macro_str1)
3091         ptr = macro_str1;
3092     spc = 0;
3093     force_blank = 0;
3094
3095     while (1) {
3096         /* NOTE: ptr == NULL can only happen if tokens are read from
3097            file stream due to a macro function call */
3098         if (ptr == NULL)
3099             break;
3100         TOK_GET(&t, &ptr, &cval);
3101         if (t == 0)
3102             break;
3103         if (t == TOK_NOSUBST) {
3104             /* following token has already been subst'd. just copy it on */
3105             tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3106             TOK_GET(&t, &ptr, &cval);
3107             goto no_subst;
3108         }
3109         s = define_find(t);
3110         if (s != NULL) {
3111             /* if nested substitution, do nothing */
3112             if (sym_find2(*nested_list, t)) {
3113                 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
3114                 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3115                 goto no_subst;
3116             }
3117             ml.p = macro_ptr;
3118             if (can_read_stream)
3119                 ml.prev = *can_read_stream, *can_read_stream = &ml;
3120             macro_ptr = (int *)ptr;
3121             tok = t;
3122             ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
3123             ptr = (int *)macro_ptr;
3124             macro_ptr = ml.p;
3125             if (can_read_stream && *can_read_stream == &ml)
3126                 *can_read_stream = ml.prev;
3127             if (ret != 0)
3128                 goto no_subst;
3129             if (parse_flags & PARSE_FLAG_SPACES)
3130                 force_blank = 1;
3131         } else {
3132         no_subst:
3133             if (force_blank) {
3134                 tok_str_add(tok_str, ' ');
3135                 spc = 1;
3136                 force_blank = 0;
3137             }
3138             if (!check_space(t, &spc))
3139                 tok_str_add2(tok_str, t, &cval);
3140         }
3141     }
3142     if (macro_str1)
3143         tok_str_free(macro_str1);
3144 }
3145
3146 /* return next token with macro substitution */
3147 ST_FUNC void next(void)
3148 {
3149     Sym *nested_list, *s;
3150     TokenString str;
3151     struct macro_level *ml;
3152
3153  redo:
3154     if (parse_flags & PARSE_FLAG_SPACES)
3155         next_nomacro_spc();
3156     else
3157         next_nomacro();
3158     if (!macro_ptr) {
3159         /* if not reading from macro substituted string, then try
3160            to substitute macros */
3161         if (tok >= TOK_IDENT &&
3162             (parse_flags & PARSE_FLAG_PREPROCESS)) {
3163             s = define_find(tok);
3164             if (s) {
3165                 /* we have a macro: we try to substitute */
3166                 tok_str_new(&str);
3167                 nested_list = NULL;
3168                 ml = NULL;
3169                 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
3170                     /* substitution done, NOTE: maybe empty */
3171                     tok_str_add(&str, 0);
3172                     macro_ptr = str.str;
3173                     macro_ptr_allocated = str.str;
3174                     goto redo;
3175                 }
3176             }
3177         }
3178     } else {
3179         if (tok == 0) {
3180             /* end of macro or end of unget buffer */
3181             if (unget_buffer_enabled) {
3182                 macro_ptr = unget_saved_macro_ptr;
3183                 unget_buffer_enabled = 0;
3184             } else {
3185                 /* end of macro string: free it */
3186                 tok_str_free(macro_ptr_allocated);
3187                 macro_ptr_allocated = NULL;
3188                 macro_ptr = NULL;
3189             }
3190             goto redo;
3191         } else if (tok == TOK_NOSUBST) {
3192             /* discard preprocessor's nosubst markers */
3193             goto redo;
3194         }
3195     }
3196
3197     /* convert preprocessor tokens into C tokens */
3198     if (tok == TOK_PPNUM &&
3199         (parse_flags & PARSE_FLAG_TOK_NUM)) {
3200         parse_number((char *)tokc.cstr->data);
3201     }
3202 }
3203
3204 /* push back current token and set current token to 'last_tok'. Only
3205    identifier case handled for labels. */
3206 ST_INLN void unget_tok(int last_tok)
3207 {
3208     int i, n;
3209     int *q;
3210     if (unget_buffer_enabled)
3211       {
3212         /* assert(macro_ptr == unget_saved_buffer + 1);
3213            assert(*macro_ptr == 0);  */
3214       }
3215     else
3216       {
3217         unget_saved_macro_ptr = macro_ptr;
3218         unget_buffer_enabled = 1;
3219       }
3220     q = unget_saved_buffer;
3221     macro_ptr = q;
3222     *q++ = tok;
3223     n = tok_ext_size(tok) - 1;
3224     for(i=0;i<n;i++)
3225         *q++ = tokc.tab[i];
3226     *q = 0; /* end of token string */
3227     tok = last_tok;
3228 }
3229
3230
3231 /* better than nothing, but needs extension to handle '-E' option
3232    correctly too */
3233 ST_FUNC void preprocess_init(TCCState *s1)
3234 {
3235     s1->include_stack_ptr = s1->include_stack;
3236     /* XXX: move that before to avoid having to initialize
3237        file->ifdef_stack_ptr ? */
3238     s1->ifdef_stack_ptr = s1->ifdef_stack;
3239     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3240
3241     vtop = vstack - 1;
3242     s1->pack_stack[0] = 0;
3243     s1->pack_stack_ptr = s1->pack_stack;
3244 }
3245
3246 ST_FUNC void preprocess_new(void)
3247 {
3248     int i, c;
3249     const char *p, *r;
3250
3251     /* init isid table */
3252
3253     for(i=CH_EOF;i<256;i++)
3254         isidnum_table[i-CH_EOF] = isid(i) || isnum(i);
3255
3256     /* add all tokens */
3257     if (table_ident) {
3258         tcc_free (table_ident);
3259         table_ident = NULL;
3260     }
3261     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3262
3263     tok_ident = TOK_IDENT;
3264     p = tcc_keywords;
3265     while (*p) {
3266         r = p;
3267         for(;;) {
3268             c = *r++;
3269             if (c == '\0')
3270                 break;
3271         }
3272         tok_alloc(p, r - p - 1);
3273         p = r;
3274     }
3275 }
3276
3277 static void line_macro_output(BufferedFile *f, const char *s, TCCState *s1)
3278 {
3279     switch (s1->Pflag) {
3280         case LINE_MACRO_OUTPUT_FORMAT_STD:
3281             /* "tcc -E -P1" case */
3282             fprintf(s1->ppfp, "# line %d \"%s\"\n", f->line_num, f->filename);
3283             break;
3284
3285         case LINE_MACRO_OUTPUT_FORMAT_NONE:
3286             /* "tcc -E -P" case: don't output a line directive */
3287             break;
3288
3289         case LINE_MACRO_OUTPUT_FORMAT_GCC:
3290         default:
3291             /* "tcc -E" case: a gcc standard by default */
3292             fprintf(s1->ppfp, "# %d \"%s\"%s\n", f->line_num, f->filename, s);
3293             break;
3294     }
3295 }
3296
3297 /* Preprocess the current file */
3298 ST_FUNC int tcc_preprocess(TCCState *s1)
3299 {
3300     BufferedFile *file_ref, **iptr, **iptr_new;
3301     int token_seen, d;
3302     const char *s;
3303
3304     preprocess_init(s1);
3305     ch = file->buf_ptr[0];
3306     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3307     parse_flags &= (PARSE_FLAG_ASM_FILE | PARSE_FLAG_ASM_COMMENTS);
3308     parse_flags |= PARSE_FLAG_PREPROCESS | PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
3309     token_seen = 0;
3310     file->line_ref = 0;
3311     file_ref = NULL;
3312     iptr = s1->include_stack_ptr;
3313
3314     for (;;) {
3315         next();
3316         if (tok == TOK_EOF) {
3317             break;
3318         } else if (file != file_ref) {
3319             if (file_ref)
3320                 line_macro_output(file_ref, "", s1);
3321             goto print_line;
3322         } else if (tok == TOK_LINEFEED) {
3323             if (!token_seen)
3324                 continue;
3325             file->line_ref++;
3326             token_seen = 0;
3327         } else if (!token_seen) {
3328             d = file->line_num - file->line_ref;
3329             if (file != file_ref || d >= 8) {
3330 print_line:
3331                 s = "";
3332                 if (tcc_state->Pflag == LINE_MACRO_OUTPUT_FORMAT_GCC) {
3333                     iptr_new = s1->include_stack_ptr;
3334                     s = iptr_new > iptr ? " 1"
3335                     : iptr_new < iptr ? " 2"
3336                     : iptr_new > s1->include_stack ? " 3"
3337                     : ""
3338                     ;
3339                 }
3340                 line_macro_output(file, s, s1);
3341             } else {
3342                 while (d > 0)
3343                     fputs("\n", s1->ppfp), --d;
3344             }
3345             file->line_ref = (file_ref = file)->line_num;
3346             token_seen = tok != TOK_LINEFEED;
3347             if (!token_seen)
3348                 continue;
3349         }
3350         fputs(get_tok_str(tok, &tokc), s1->ppfp);
3351     }
3352     return 0;
3353 }