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 TokenSym *hash_ident[TOK_HASH_SIZE];
  44 static char token_buf[STRING_MAX_SIZE + 1];
  45 static unsigned char isidnum_table[256 - CH_EOF];
  46 /* isidnum_table flags: */
  47 #define IS_SPC 1
  48 #define IS_ID  2
  49 #define IS_NUM 4
  50
  51 static TokenString *macro_stack;
  52
  53 static const char tcc_keywords[] =
  54 #define DEF(id, str) str "\0"
  55 #include "tcctok.h"
  56 #undef DEF
  57 ;
  58
  59 /* WARNING: the content of this string encodes token numbers */
  60 static const unsigned char tok_two_chars[] =
  61 /* outdated -- gr
  62     "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
  63     "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
  64 */{
  65     '<','=', TOK_LE,
  66     '>','=', TOK_GE,
  67     '!','=', TOK_NE,
  68     '&','&', TOK_LAND,
  69     '|','|', TOK_LOR,
  70     '+','+', TOK_INC,
  71     '-','-', TOK_DEC,
  72     '=','=', TOK_EQ,
  73     '<','<', TOK_SHL,
  74     '>','>', TOK_SAR,
  75     '+','=', TOK_A_ADD,
  76     '-','=', TOK_A_SUB,
  77     '*','=', TOK_A_MUL,
  78     '/','=', TOK_A_DIV,
  79     '%','=', TOK_A_MOD,
  80     '&','=', TOK_A_AND,
  81     '^','=', TOK_A_XOR,
  82     '|','=', TOK_A_OR,
  83     '-','>', TOK_ARROW,
  84     '.','.', 0xa8, // C++ token ?
  85     '#','#', TOK_TWOSHARPS,
  86     0
  87 };
  88
  89 static void next_nomacro_spc(void);
  90
  91 ST_FUNC void skip(int c)
  92 {
  93     if (tok != c)
  94         tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
  95     next();
  96 }
  97
  98 ST_FUNC void expect(const char *msg)
  99 {
 100     tcc_error("%s expected", msg);
 101 }
 102
 103 ST_FUNC void begin_macro(TokenString *str, int alloc)
 104 {
 105     str->alloc = alloc;
 106     str->prev = macro_stack;
 107     str->prev_ptr = macro_ptr;
 108     macro_ptr = str->str;
 109     macro_stack = str;
 110 }
 111
 112 ST_FUNC void end_macro(void)
 113 {
 114     TokenString *str = macro_stack;
 115     macro_stack = str->prev;
 116     macro_ptr = str->prev_ptr;
 117     if (str->alloc == 2) {
 118         str->alloc = 3; /* just mark as finished */
 119     } else {
 120         tok_str_free(str->str);
 121         if (str->alloc == 1)
 122             tcc_free(str);
 123     }
 124 }
 125
 126 /* ------------------------------------------------------------------------- */
 127 /* CString handling */
 128 static void cstr_realloc(CString *cstr, int new_size)
 129 {
 130     int size;
 131     void *data;
 132
 133     size = cstr->size_allocated;
 134     if (size == 0)
 135         size = 8; /* no need to allocate a too small first string */
 136     while (size < new_size)
 137         size = size * 2;
 138     data = tcc_realloc(cstr->data_allocated, size);
 139     cstr->data_allocated = data;
 140     cstr->size_allocated = size;
 141     cstr->data = data;
 142 }
 143
 144 /* add a byte */
 145 ST_FUNC void cstr_ccat(CString *cstr, int ch)
 146 {
 147     int size;
 148     size = cstr->size + 1;
 149     if (size > cstr->size_allocated)
 150         cstr_realloc(cstr, size);
 151     ((unsigned char *)cstr->data)[size - 1] = ch;
 152     cstr->size = size;
 153 }
 154
 155 ST_FUNC void cstr_cat(CString *cstr, const char *str)
 156 {
 157     int c;
 158     for(;;) {
 159         c = *str;
 160         if (c == '\0')
 161             break;
 162         cstr_ccat(cstr, c);
 163         str++;
 164     }
 165 }
 166
 167 /* add a wide char */
 168 ST_FUNC void cstr_wccat(CString *cstr, int ch)
 169 {
 170     int size;
 171     size = cstr->size + sizeof(nwchar_t);
 172     if (size > cstr->size_allocated)
 173         cstr_realloc(cstr, size);
 174     *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
 175     cstr->size = size;
 176 }
 177
 178 ST_FUNC void cstr_new(CString *cstr)
 179 {
 180     memset(cstr, 0, sizeof(CString));
 181 }
 182
 183 /* free string and reset it to NULL */
 184 ST_FUNC void cstr_free(CString *cstr)
 185 {
 186     tcc_free(cstr->data_allocated);
 187     cstr_new(cstr);
 188 }
 189
 190 /* reset string to empty */
 191 ST_FUNC void cstr_reset(CString *cstr)
 192 {
 193     cstr->size = 0;
 194 }
 195
 196 /* XXX: unicode ? */
 197 static void add_char(CString *cstr, int c)
 198 {
 199     if (c == '\'' || c == '\"' || c == '\\') {
 200         /* XXX: could be more precise if char or string */
 201         cstr_ccat(cstr, '\\');
 202     }
 203     if (c >= 32 && c <= 126) {
 204         cstr_ccat(cstr, c);
 205     } else {
 206         cstr_ccat(cstr, '\\');
 207         if (c == '\n') {
 208             cstr_ccat(cstr, 'n');
 209         } else {
 210             cstr_ccat(cstr, '0' + ((c >> 6) & 7));
 211             cstr_ccat(cstr, '0' + ((c >> 3) & 7));
 212             cstr_ccat(cstr, '0' + (c & 7));
 213         }
 214     }
 215 }
 216
 217 /* ------------------------------------------------------------------------- */
 218 /* allocate a new token */
 219 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
 220 {
 221     TokenSym *ts, **ptable;
 222     int i;
 223
 224     if (tok_ident >= SYM_FIRST_ANOM)
 225         tcc_error("memory full (symbols)");
 226
 227     /* expand token table if needed */
 228     i = tok_ident - TOK_IDENT;
 229     if ((i % TOK_ALLOC_INCR) == 0) {
 230         ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
 231         table_ident = ptable;
 232     }
 233
 234     ts = tcc_malloc(sizeof(TokenSym) + len);
 235     table_ident[i] = ts;
 236     ts->tok = tok_ident++;
 237     ts->sym_define = NULL;
 238     ts->sym_label = NULL;
 239     ts->sym_struct = NULL;
 240     ts->sym_identifier = NULL;
 241     ts->len = len;
 242     ts->hash_next = NULL;
 243     memcpy(ts->str, str, len);
 244     ts->str[len] = '\0';
 245     *pts = ts;
 246     return ts;
 247 }
 248
 249 #define TOK_HASH_INIT 1
 250 #define TOK_HASH_FUNC(h, c) ((h) * 263 + (c))
 251
 252 /* find a token and add it if not found */
 253 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
 254 {
 255     TokenSym *ts, **pts;
 256     int i;
 257     unsigned int h;
 258
 259     h = TOK_HASH_INIT;
 260     for(i=0;i<len;i++)
 261         h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
 262     h &= (TOK_HASH_SIZE - 1);
 263
 264     pts = &hash_ident[h];
 265     for(;;) {
 266         ts = *pts;
 267         if (!ts)
 268             break;
 269         if (ts->len == len && !memcmp(ts->str, str, len))
 270             return ts;
 271         pts = &(ts->hash_next);
 272     }
 273     return tok_alloc_new(pts, str, len);
 274 }
 275
 276 /* XXX: buffer overflow */
 277 /* XXX: float tokens */
 278 ST_FUNC const char *get_tok_str(int v, CValue *cv)
 279 {
 280     static char buf[STRING_MAX_SIZE + 1];
 281     static CString cstr_buf;
 282     CString *cstr;
 283     char *p;
 284     int i, len;
 285
 286     /* NOTE: to go faster, we give a fixed buffer for small strings */
 287     cstr_reset(&cstr_buf);
 288     cstr_buf.data = buf;
 289     cstr_buf.size_allocated = sizeof(buf);
 290     p = buf;
 291
 292     switch(v) {
 293     case TOK_CINT:
 294     case TOK_CUINT:
 295         /* XXX: not quite exact, but only useful for testing */
 296         sprintf(p, "%llu", (unsigned long long)cv->i);
 297         break;
 298     case TOK_CLLONG:
 299     case TOK_CULLONG:
 300         /* XXX: not quite exact, but only useful for testing  */
 301 #ifdef _WIN32
 302         sprintf(p, "%u", (unsigned)cv->i);
 303 #else
 304         sprintf(p, "%llu", (unsigned long long)cv->i);
 305 #endif
 306         break;
 307     case TOK_LCHAR:
 308         cstr_ccat(&cstr_buf, 'L');
 309     case TOK_CCHAR:
 310         cstr_ccat(&cstr_buf, '\'');
 311         add_char(&cstr_buf, cv->i);
 312         cstr_ccat(&cstr_buf, '\'');
 313         cstr_ccat(&cstr_buf, '\0');
 314         break;
 315     case TOK_PPNUM:
 316     case TOK_PPSTR:
 317         return (char*)cv->cstr->data;
 318     case TOK_LSTR:
 319         cstr_ccat(&cstr_buf, 'L');
 320     case TOK_STR:
 321         cstr = cv->cstr;
 322         cstr_ccat(&cstr_buf, '\"');
 323         if (v == TOK_STR) {
 324             len = cstr->size - 1;
 325             for(i=0;i<len;i++)
 326                 add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 327         } else {
 328             len = (cstr->size / sizeof(nwchar_t)) - 1;
 329             for(i=0;i<len;i++)
 330                 add_char(&cstr_buf, ((nwchar_t *)cstr->data)[i]);
 331         }
 332         cstr_ccat(&cstr_buf, '\"');
 333         cstr_ccat(&cstr_buf, '\0');
 334         break;
 335
 336     case TOK_CFLOAT:
 337     case TOK_CDOUBLE:
 338     case TOK_CLDOUBLE:
 339     case TOK_LINENUM:
 340         return NULL; /* should not happen */
 341
 342     /* above tokens have value, the ones below don't */
 343
 344     case TOK_LT:
 345         v = '<';
 346         goto addv;
 347     case TOK_GT:
 348         v = '>';
 349         goto addv;
 350     case TOK_DOTS:
 351         return strcpy(p, "...");
 352     case TOK_A_SHL:
 353         return strcpy(p, "<<=");
 354     case TOK_A_SAR:
 355         return strcpy(p, ">>=");
 356     default:
 357         if (v < TOK_IDENT) {
 358             /* search in two bytes table */
 359             const unsigned char *q = tok_two_chars;
 360             while (*q) {
 361                 if (q[2] == v) {
 362                     *p++ = q[0];
 363                     *p++ = q[1];
 364                     *p = '\0';
 365                     return buf;
 366                 }
 367                 q += 3;
 368             }
 369         if (v >= 127) {
 370             sprintf(buf, "<%02x>", v);
 371             return buf;
 372         }
 373         addv:
 374             *p++ = v;
 375             *p = '\0';
 376         } else if (v < tok_ident) {
 377             return table_ident[v - TOK_IDENT]->str;
 378         } else if (v >= SYM_FIRST_ANOM) {
 379             /* special name for anonymous symbol */
 380             sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
 381         } else {
 382             /* should never happen */
 383             return NULL;
 384         }
 385         break;
 386     }
 387     return cstr_buf.data;
 388 }
 389
 390 /* return the current character, handling end of block if necessary
 391    (but not stray) */
 392 ST_FUNC int handle_eob(void)
 393 {
 394     BufferedFile *bf = file;
 395     int len;
 396
 397     /* only tries to read if really end of buffer */
 398     if (bf->buf_ptr >= bf->buf_end) {
 399         if (bf->fd != -1) {
 400 #if defined(PARSE_DEBUG)
 401             len = 1;
 402 #else
 403             len = IO_BUF_SIZE;
 404 #endif
 405             len = read(bf->fd, bf->buffer, len);
 406             if (len < 0)
 407                 len = 0;
 408         } else {
 409             len = 0;
 410         }
 411         total_bytes += len;
 412         bf->buf_ptr = bf->buffer;
 413         bf->buf_end = bf->buffer + len;
 414         *bf->buf_end = CH_EOB;
 415     }
 416     if (bf->buf_ptr < bf->buf_end) {
 417         return bf->buf_ptr[0];
 418     } else {
 419         bf->buf_ptr = bf->buf_end;
 420         return CH_EOF;
 421     }
 422 }
 423
 424 /* read next char from current input file and handle end of input buffer */
 425 ST_INLN void inp(void)
 426 {
 427     ch = *(++(file->buf_ptr));
 428     /* end of buffer/file handling */
 429     if (ch == CH_EOB)
 430         ch = handle_eob();
 431 }
 432
 433 /* handle '\[\r]\n' */
 434 static int handle_stray_noerror(void)
 435 {
 436     while (ch == '\\') {
 437         inp();
 438         if (ch == '\n') {
 439             file->line_num++;
 440             inp();
 441         } else if (ch == '\r') {
 442             inp();
 443             if (ch != '\n')
 444                 goto fail;
 445             file->line_num++;
 446             inp();
 447         } else {
 448         fail:
 449             return 1;
 450         }
 451     }
 452     return 0;
 453 }
 454
 455 static void handle_stray(void)
 456 {
 457     if (handle_stray_noerror())
 458         tcc_error("stray '\\' in program");
 459 }
 460
 461 /* skip the stray and handle the \\n case. Output an error if
 462    incorrect char after the stray */
 463 static int handle_stray1(uint8_t *p)
 464 {
 465     int c;
 466
 467     file->buf_ptr = p;
 468     if (p >= file->buf_end) {
 469         c = handle_eob();
 470         if (c != '\\')
 471             return c;
 472         p = file->buf_ptr;
 473     }
 474     ch = *p;
 475     if (handle_stray_noerror()) {
 476         if (!(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
 477             tcc_error("stray '\\' in program");
 478         *--file->buf_ptr = '\\';
 479     }
 480     p = file->buf_ptr;
 481     c = *p;
 482     return c;
 483 }
 484
 485 /* handle just the EOB case, but not stray */
 486 #define PEEKC_EOB(c, p)\
 487 {\
 488     p++;\
 489     c = *p;\
 490     if (c == '\\') {\
 491         file->buf_ptr = p;\
 492         c = handle_eob();\
 493         p = file->buf_ptr;\
 494     }\
 495 }
 496
 497 /* handle the complicated stray case */
 498 #define PEEKC(c, p)\
 499 {\
 500     p++;\
 501     c = *p;\
 502     if (c == '\\') {\
 503         c = handle_stray1(p);\
 504         p = file->buf_ptr;\
 505     }\
 506 }
 507
 508 /* input with '\[\r]\n' handling. Note that this function cannot
 509    handle other characters after '\', so you cannot call it inside
 510    strings or comments */
 511 ST_FUNC void minp(void)
 512 {
 513     inp();
 514     if (ch == '\\')
 515         handle_stray();
 516 }
 517
 518
 519 /* single line C++ comments */
 520 static uint8_t *parse_line_comment(uint8_t *p)
 521 {
 522     int c;
 523
 524     p++;
 525     for(;;) {
 526         c = *p;
 527     redo:
 528         if (c == '\n' || c == CH_EOF) {
 529             break;
 530         } else if (c == '\\') {
 531             file->buf_ptr = p;
 532             c = handle_eob();
 533             p = file->buf_ptr;
 534             if (c == '\\') {
 535                 PEEKC_EOB(c, p);
 536                 if (c == '\n') {
 537                     file->line_num++;
 538                     PEEKC_EOB(c, p);
 539                 } else if (c == '\r') {
 540                     PEEKC_EOB(c, p);
 541                     if (c == '\n') {
 542                         file->line_num++;
 543                         PEEKC_EOB(c, p);
 544                     }
 545                 }
 546             } else {
 547                 goto redo;
 548             }
 549         } else {
 550             p++;
 551         }
 552     }
 553     return p;
 554 }
 555
 556 /* C comments */
 557 ST_FUNC uint8_t *parse_comment(uint8_t *p)
 558 {
 559     int c;
 560
 561     p++;
 562     for(;;) {
 563         /* fast skip loop */
 564         for(;;) {
 565             c = *p;
 566             if (c == '\n' || c == '*' || c == '\\')
 567                 break;
 568             p++;
 569             c = *p;
 570             if (c == '\n' || c == '*' || c == '\\')
 571                 break;
 572             p++;
 573         }
 574         /* now we can handle all the cases */
 575         if (c == '\n') {
 576             file->line_num++;
 577             p++;
 578         } else if (c == '*') {
 579             p++;
 580             for(;;) {
 581                 c = *p;
 582                 if (c == '*') {
 583                     p++;
 584                 } else if (c == '/') {
 585                     goto end_of_comment;
 586                 } else if (c == '\\') {
 587                     file->buf_ptr = p;
 588                     c = handle_eob();
 589                     p = file->buf_ptr;
 590                     if (c == CH_EOF)
 591                         tcc_error("unexpected end of file in comment");
 592                     if (c == '\\') {
 593                         /* skip '\[\r]\n', otherwise just skip the stray */
 594                         while (c == '\\') {
 595                             PEEKC_EOB(c, p);
 596                             if (c == '\n') {
 597                                 file->line_num++;
 598                                 PEEKC_EOB(c, p);
 599                             } else if (c == '\r') {
 600                                 PEEKC_EOB(c, p);
 601                                 if (c == '\n') {
 602                                     file->line_num++;
 603                                     PEEKC_EOB(c, p);
 604                                 }
 605                             } else {
 606                                 goto after_star;
 607                             }
 608                         }
 609                     }
 610                 } else {
 611                     break;
 612                 }
 613             }
 614         after_star: ;
 615         } else {
 616             /* stray, eob or eof */
 617             file->buf_ptr = p;
 618             c = handle_eob();
 619             p = file->buf_ptr;
 620             if (c == CH_EOF) {
 621                 tcc_error("unexpected end of file in comment");
 622             } else if (c == '\\') {
 623                 p++;
 624             }
 625         }
 626     }
 627  end_of_comment:
 628     p++;
 629     return p;
 630 }
 631
 632 #define cinp minp
 633
 634 static inline void skip_spaces(void)
 635 {
 636     while (isidnum_table[ch - CH_EOF] & IS_SPC)
 637         cinp();
 638 }
 639
 640 static inline int check_space(int t, int *spc)
 641 {
 642     if (t < 256 && (isidnum_table[t - CH_EOF] & IS_SPC)) {
 643         if (*spc)
 644             return 1;
 645         *spc = 1;
 646     } else
 647         *spc = 0;
 648     return 0;
 649 }
 650
 651 /* parse a string without interpreting escapes */
 652 static uint8_t *parse_pp_string(uint8_t *p,
 653                                 int sep, CString *str)
 654 {
 655     int c;
 656     p++;
 657     for(;;) {
 658         c = *p;
 659         if (c == sep) {
 660             break;
 661         } else if (c == '\\') {
 662             file->buf_ptr = p;
 663             c = handle_eob();
 664             p = file->buf_ptr;
 665             if (c == CH_EOF) {
 666             unterminated_string:
 667                 /* XXX: indicate line number of start of string */
 668                 tcc_error("missing terminating %c character", sep);
 669             } else if (c == '\\') {
 670                 /* escape : just skip \[\r]\n */
 671                 PEEKC_EOB(c, p);
 672                 if (c == '\n') {
 673                     file->line_num++;
 674                     p++;
 675                 } else if (c == '\r') {
 676                     PEEKC_EOB(c, p);
 677                     if (c != '\n')
 678                         expect("'\n' after '\r'");
 679                     file->line_num++;
 680                     p++;
 681                 } else if (c == CH_EOF) {
 682                     goto unterminated_string;
 683                 } else {
 684                     if (str) {
 685                         cstr_ccat(str, '\\');
 686                         cstr_ccat(str, c);
 687                     }
 688                     p++;
 689                 }
 690             }
 691         } else if (c == '\n') {
 692             file->line_num++;
 693             goto add_char;
 694         } else if (c == '\r') {
 695             PEEKC_EOB(c, p);
 696             if (c != '\n') {
 697                 if (str)
 698                     cstr_ccat(str, '\r');
 699             } else {
 700                 file->line_num++;
 701                 goto add_char;
 702             }
 703         } else {
 704         add_char:
 705             if (str)
 706                 cstr_ccat(str, c);
 707             p++;
 708         }
 709     }
 710     p++;
 711     return p;
 712 }
 713
 714 /* skip block of text until #else, #elif or #endif. skip also pairs of
 715    #if/#endif */
 716 static void preprocess_skip(void)
 717 {
 718     int a, start_of_line, c, in_warn_or_error;
 719     uint8_t *p;
 720
 721     p = file->buf_ptr;
 722     a = 0;
 723 redo_start:
 724     start_of_line = 1;
 725     in_warn_or_error = 0;
 726     for(;;) {
 727     redo_no_start:
 728         c = *p;
 729         switch(c) {
 730         case ' ':
 731         case '\t':
 732         case '\f':
 733         case '\v':
 734         case '\r':
 735             p++;
 736             goto redo_no_start;
 737         case '\n':
 738             file->line_num++;
 739             p++;
 740             goto redo_start;
 741         case '\\':
 742             file->buf_ptr = p;
 743             c = handle_eob();
 744             if (c == CH_EOF) {
 745                 expect("#endif");
 746             } else if (c == '\\') {
 747                 ch = file->buf_ptr[0];
 748                 handle_stray_noerror();
 749             }
 750             p = file->buf_ptr;
 751             goto redo_no_start;
 752         /* skip strings */
 753         case '\"':
 754         case '\'':
 755             if (in_warn_or_error)
 756                 goto _default;
 757             p = parse_pp_string(p, c, NULL);
 758             break;
 759         /* skip comments */
 760         case '/':
 761             if (in_warn_or_error)
 762                 goto _default;
 763             file->buf_ptr = p;
 764             ch = *p;
 765             minp();
 766             p = file->buf_ptr;
 767             if (ch == '*') {
 768                 p = parse_comment(p);
 769             } else if (ch == '/') {
 770                 p = parse_line_comment(p);
 771             }
 772             break;
 773         case '#':
 774             p++;
 775             if (start_of_line) {
 776                 file->buf_ptr = p;
 777                 next_nomacro();
 778                 p = file->buf_ptr;
 779                 if (a == 0 &&
 780                     (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
 781                     goto the_end;
 782                 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
 783                     a++;
 784                 else if (tok == TOK_ENDIF)
 785                     a--;
 786                 else if( tok == TOK_ERROR || tok == TOK_WARNING)
 787                     in_warn_or_error = 1;
 788                 else if (tok == TOK_LINEFEED)
 789                     goto redo_start;
 790             } else if (parse_flags & PARSE_FLAG_ASM_FILE)
 791                 p = parse_line_comment(p);
 792             break;
 793 _default:
 794         default:
 795             p++;
 796             break;
 797         }
 798         start_of_line = 0;
 799     }
 800  the_end: ;
 801     file->buf_ptr = p;
 802 }
 803
 804 /* ParseState handling */
 805
 806 /* XXX: currently, no include file info is stored. Thus, we cannot display
 807    accurate messages if the function or data definition spans multiple
 808    files */
 809
 810 /* save current parse state in 's' */
 811 ST_FUNC void save_parse_state(ParseState *s)
 812 {
 813     s->line_num = file->line_num;
 814     s->macro_ptr = macro_ptr;
 815     s->tok = tok;
 816     s->tokc = tokc;
 817 }
 818
 819 /* restore parse state from 's' */
 820 ST_FUNC void restore_parse_state(ParseState *s)
 821 {
 822     file->line_num = s->line_num;
 823     macro_ptr = s->macro_ptr;
 824     tok = s->tok;
 825     tokc = s->tokc;
 826 }
 827
 828 /* return the number of additional 'ints' necessary to store the
 829    token */
 830 static inline int tok_size(const int *p)
 831 {
 832     switch(*p) {
 833         /* 4 bytes */
 834     case TOK_CINT:
 835     case TOK_CUINT:
 836     case TOK_CCHAR:
 837     case TOK_LCHAR:
 838     case TOK_CFLOAT:
 839     case TOK_LINENUM:
 840         return 1 + 1;
 841     case TOK_STR:
 842     case TOK_LSTR:
 843     case TOK_PPNUM:
 844     case TOK_PPSTR:
 845         return 1 + ((sizeof(CString) + ((CString *)(p+1))->size + 3) >> 2);
 846     case TOK_CDOUBLE:
 847     case TOK_CLLONG:
 848     case TOK_CULLONG:
 849         return 1 + 2;
 850     case TOK_CLDOUBLE:
 851         return 1 + LDOUBLE_SIZE / 4;
 852     default:
 853         return 1 + 0;
 854     }
 855 }
 856
 857 /* token string handling */
 858
 859 ST_INLN void tok_str_new(TokenString *s)
 860 {
 861     s->str = NULL;
 862     s->len = 0;
 863     s->allocated_len = 0;
 864     s->last_line_num = -1;
 865 }
 866
 867 ST_FUNC void tok_str_free(int *str)
 868 {
 869     tcc_free(str);
 870 }
 871
 872 static int *tok_str_realloc(TokenString *s)
 873 {
 874     int *str, len;
 875
 876     if (s->allocated_len == 0) {
 877         len = 8;
 878     } else {
 879         len = s->allocated_len * 2;
 880     }
 881     str = tcc_realloc(s->str, len * sizeof(int));
 882     s->allocated_len = len;
 883     s->str = str;
 884     return str;
 885 }
 886
 887 ST_FUNC void tok_str_add(TokenString *s, int t)
 888 {
 889     int len, *str;
 890
 891     len = s->len;
 892     str = s->str;
 893     if (len >= s->allocated_len)
 894         str = tok_str_realloc(s);
 895     str[len++] = t;
 896     s->len = len;
 897 }
 898
 899 static void tok_str_add2(TokenString *s, int t, CValue *cv)
 900 {
 901     int len, *str;
 902
 903     len = s->len;
 904     str = s->str;
 905
 906     /* allocate space for worst case */
 907     if (len + TOK_MAX_SIZE > s->allocated_len)
 908         str = tok_str_realloc(s);
 909     str[len++] = t;
 910     switch(t) {
 911     case TOK_CINT:
 912     case TOK_CUINT:
 913     case TOK_CCHAR:
 914     case TOK_LCHAR:
 915     case TOK_CFLOAT:
 916     case TOK_LINENUM:
 917         str[len++] = cv->tab[0];
 918         break;
 919     case TOK_PPNUM:
 920     case TOK_PPSTR:
 921     case TOK_STR:
 922     case TOK_LSTR:
 923         {
 924             int nb_words;
 925             CString cstr;
 926
 927             nb_words = (sizeof(CString) + cv->cstr->size + 3) >> 2;
 928             while ((len + nb_words) > s->allocated_len)
 929                 str = tok_str_realloc(s);
 930             /* XXX: Insert the CString into the int array.
 931                It may end up incorrectly aligned. */
 932             cstr.data = 0;
 933             cstr.size = cv->cstr->size;
 934             cstr.data_allocated = 0;
 935             cstr.size_allocated = cstr.size;
 936             memcpy(str + len, &cstr, sizeof(CString));
 937             memcpy((char *)(str + len) + sizeof(CString),
 938                    cv->cstr->data, cstr.size);
 939             len += nb_words;
 940         }
 941         break;
 942     case TOK_CDOUBLE:
 943     case TOK_CLLONG:
 944     case TOK_CULLONG:
 945 #if LDOUBLE_SIZE == 8
 946     case TOK_CLDOUBLE:
 947 #endif
 948         str[len++] = cv->tab[0];
 949         str[len++] = cv->tab[1];
 950         break;
 951 #if LDOUBLE_SIZE == 12
 952     case TOK_CLDOUBLE:
 953         str[len++] = cv->tab[0];
 954         str[len++] = cv->tab[1];
 955         str[len++] = cv->tab[2];
 956 #elif LDOUBLE_SIZE == 16
 957     case TOK_CLDOUBLE:
 958         str[len++] = cv->tab[0];
 959         str[len++] = cv->tab[1];
 960         str[len++] = cv->tab[2];
 961         str[len++] = cv->tab[3];
 962 #elif LDOUBLE_SIZE != 8
 963 #error add long double size support
 964 #endif
 965         break;
 966     default:
 967         break;
 968     }
 969     s->len = len;
 970 }
 971
 972 /* add the current parse token in token string 's' */
 973 ST_FUNC void tok_str_add_tok(TokenString *s)
 974 {
 975     CValue cval;
 976
 977     /* save line number info */
 978     if (file->line_num != s->last_line_num) {
 979         s->last_line_num = file->line_num;
 980         cval.i = s->last_line_num;
 981         tok_str_add2(s, TOK_LINENUM, &cval);
 982     }
 983     tok_str_add2(s, tok, &tokc);
 984 }
 985
 986 /* get a token from an integer array and increment pointer
 987    accordingly. we code it as a macro to avoid pointer aliasing. */
 988 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
 989 {
 990     const int *p = *pp;
 991     int n, *tab;
 992
 993     tab = cv->tab;
 994     switch(*t = *p++) {
 995     case TOK_CINT:
 996     case TOK_CUINT:
 997     case TOK_CCHAR:
 998     case TOK_LCHAR:
 999     case TOK_CFLOAT:
1000     case TOK_LINENUM:
1001         tab[0] = *p++;
1002         break;
1003     case TOK_STR:
1004     case TOK_LSTR:
1005     case TOK_PPNUM:
1006     case TOK_PPSTR:
1007         /* XXX: Illegal cast: the pointer p may not be correctly aligned! */
1008         cv->cstr = (CString *)p;
1009         cv->cstr->data = (char *)p + sizeof(CString);
1010         p += (sizeof(CString) + cv->cstr->size + 3) >> 2;
1011         break;
1012     case TOK_CDOUBLE:
1013     case TOK_CLLONG:
1014     case TOK_CULLONG:
1015         n = 2;
1016         goto copy;
1017     case TOK_CLDOUBLE:
1018 #if LDOUBLE_SIZE == 16
1019         n = 4;
1020 #elif LDOUBLE_SIZE == 12
1021         n = 3;
1022 #elif LDOUBLE_SIZE == 8
1023         n = 2;
1024 #else
1025 # error add long double size support
1026 #endif
1027     copy:
1028         do
1029             *tab++ = *p++;
1030         while (--n);
1031         break;
1032     default:
1033         break;
1034     }
1035     *pp = p;
1036 }
1037
1038 /* Calling this function is expensive, but it is not possible
1039    to read a token string backwards. */
1040 static int tok_last(const int *str0, const int *str1)
1041 {
1042     const int *str = str0;
1043     int tok = 0;
1044     CValue cval;
1045
1046     while (str < str1)
1047         TOK_GET(&tok, &str, &cval);
1048     return tok;
1049 }
1050
1051 static int macro_is_equal(const int *a, const int *b)
1052 {
1053     char buf[STRING_MAX_SIZE + 1];
1054     CValue cv;
1055     int t;
1056     while (*a && *b) {
1057         TOK_GET(&t, &a, &cv);
1058         pstrcpy(buf, sizeof buf, get_tok_str(t, &cv));
1059         TOK_GET(&t, &b, &cv);
1060         if (strcmp(buf, get_tok_str(t, &cv)))
1061             return 0;
1062     }
1063     return !(*a || *b);
1064 }
1065
1066 static void pp_line(TCCState *s1, BufferedFile *f, int level)
1067 {
1068     int d = f->line_num - f->line_ref;
1069     if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_NONE
1070         || (level == 0 && f->line_ref && d < 8))
1071     {
1072         while (d > 0)
1073             fputs("\n", s1->ppfp), --d;
1074     }
1075     else
1076     if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_STD) {
1077         fprintf(s1->ppfp, "#line %d \"%s\"\n", f->line_num, f->filename);
1078     }
1079     else {
1080         fprintf(s1->ppfp, "# %d \"%s\"%s\n", f->line_num, f->filename,
1081             level > 0 ? " 1" : level < 0 ? " 2" : "");
1082     }
1083     f->line_ref = f->line_num;
1084 }
1085
1086 static void tok_print(const char *msg, const int *str)
1087 {
1088     FILE *pr = tcc_state->ppfp;
1089     int t;
1090     CValue cval;
1091
1092     fprintf(pr, "%s ", msg);
1093     while (str) {
1094         TOK_GET(&t, &str, &cval);
1095         if (!t)
1096             break;
1097         fprintf(pr,"%s", get_tok_str(t, &cval));
1098     }
1099     fprintf(pr, "\n");
1100 }
1101
1102 static int define_print_prepared(Sym *s)
1103 {
1104     if (!s || !tcc_state->ppfp || tcc_state->dflag == 0)
1105         return 0;
1106
1107     if (s->v < TOK_IDENT || s->v >= tok_ident)
1108         return 0;
1109
1110     if (file) {
1111         file->line_num--;
1112         pp_line(tcc_state, file, 0);
1113         file->line_ref = ++file->line_num;
1114     }
1115     return 1;
1116 }
1117
1118 static void define_print(int v)
1119 {
1120     FILE *pr = tcc_state->ppfp;
1121     Sym *s, *a;
1122
1123     s = define_find(v);
1124     if (define_print_prepared(s) == 0)
1125         return;
1126
1127     fprintf(pr, "// #define %s", get_tok_str(v, NULL));
1128     if (s->type.t == MACRO_FUNC) {
1129         a = s->next;
1130         fprintf(pr,"(");
1131         if (a)
1132             for (;;) {
1133                 fprintf(pr,"%s", get_tok_str(a->v & ~SYM_FIELD, NULL));
1134                 if (!(a = a->next))
1135                     break;
1136                 fprintf(pr,",");
1137             }
1138         fprintf(pr,")");
1139     }
1140     tok_print("", s->d);
1141 }
1142
1143 static void undef_print(int v)
1144 {
1145     FILE *pr = tcc_state->ppfp;
1146     Sym *s;
1147
1148     s = define_find(v);
1149     if (define_print_prepared(s) == 0)
1150         return;
1151
1152     fprintf(pr, "// #undef %s\n", get_tok_str(s->v, NULL));
1153 }
1154
1155 ST_FUNC void print_defines(void)
1156 {
1157     Sym *top = define_stack;
1158     while (top) {
1159         define_print(top->v);
1160         top = top->prev;
1161     }
1162 }
1163
1164 /* defines handling */
1165 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1166 {
1167     Sym *s;
1168
1169     s = define_find(v);
1170     if (s && !macro_is_equal(s->d, str))
1171         tcc_warning("%s redefined", get_tok_str(v, NULL));
1172
1173     s = sym_push2(&define_stack, v, macro_type, 0);
1174     s->d = str;
1175     s->next = first_arg;
1176     table_ident[v - TOK_IDENT]->sym_define = s;
1177 }
1178
1179 /* undefined a define symbol. Its name is just set to zero */
1180 ST_FUNC void define_undef(Sym *s)
1181 {
1182     int v = s->v;
1183     undef_print(v);
1184     if (v >= TOK_IDENT && v < tok_ident)
1185         table_ident[v - TOK_IDENT]->sym_define = NULL;
1186 }
1187
1188 ST_INLN Sym *define_find(int v)
1189 {
1190     v -= TOK_IDENT;
1191     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1192         return NULL;
1193     return table_ident[v]->sym_define;
1194 }
1195
1196 /* free define stack until top reaches 'b' */
1197 ST_FUNC void free_defines(Sym *b)
1198 {
1199     Sym *top, *top1;
1200     int v;
1201
1202     top = define_stack;
1203     while (top != b) {
1204         top1 = top->prev;
1205         /* do not free args or predefined defines */
1206         if (top->d)
1207             tok_str_free(top->d);
1208         v = top->v;
1209         if (v >= TOK_IDENT && v < tok_ident)
1210             table_ident[v - TOK_IDENT]->sym_define = NULL;
1211         sym_free(top);
1212         top = top1;
1213     }
1214     define_stack = b;
1215 }
1216
1217 /* label lookup */
1218 ST_FUNC Sym *label_find(int v)
1219 {
1220     v -= TOK_IDENT;
1221     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1222         return NULL;
1223     return table_ident[v]->sym_label;
1224 }
1225
1226 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1227 {
1228     Sym *s, **ps;
1229     s = sym_push2(ptop, v, 0, 0);
1230     s->r = flags;
1231     ps = &table_ident[v - TOK_IDENT]->sym_label;
1232     if (ptop == &global_label_stack) {
1233         /* modify the top most local identifier, so that
1234            sym_identifier will point to 's' when popped */
1235         while (*ps != NULL)
1236             ps = &(*ps)->prev_tok;
1237     }
1238     s->prev_tok = *ps;
1239     *ps = s;
1240     return s;
1241 }
1242
1243 /* pop labels until element last is reached. Look if any labels are
1244    undefined. Define symbols if '&&label' was used. */
1245 ST_FUNC void label_pop(Sym **ptop, Sym *slast)
1246 {
1247     Sym *s, *s1;
1248     for(s = *ptop; s != slast; s = s1) {
1249         s1 = s->prev;
1250         if (s->r == LABEL_DECLARED) {
1251             tcc_warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1252         } else if (s->r == LABEL_FORWARD) {
1253                 tcc_error("label '%s' used but not defined",
1254                       get_tok_str(s->v, NULL));
1255         } else {
1256             if (s->c) {
1257                 /* define corresponding symbol. A size of
1258                    1 is put. */
1259                 put_extern_sym(s, cur_text_section, s->jnext, 1);
1260             }
1261         }
1262         /* remove label */
1263         table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1264         sym_free(s);
1265     }
1266     *ptop = slast;
1267 }
1268
1269 /* eval an expression for #if/#elif */
1270 static int expr_preprocess(void)
1271 {
1272     int c, t;
1273     TokenString str;
1274
1275     tok_str_new(&str);
1276     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1277         next(); /* do macro subst */
1278         if (tok == TOK_DEFINED) {
1279             next_nomacro();
1280             t = tok;
1281             if (t == '(')
1282                 next_nomacro();
1283             c = define_find(tok) != 0;
1284             if (t == '(')
1285                 next_nomacro();
1286             tok = TOK_CINT;
1287             tokc.i = c;
1288         } else if (tok >= TOK_IDENT) {
1289             /* if undefined macro */
1290             tok = TOK_CINT;
1291             tokc.i = 0;
1292         }
1293         tok_str_add_tok(&str);
1294     }
1295     tok_str_add(&str, -1); /* simulate end of file */
1296     tok_str_add(&str, 0);
1297     /* now evaluate C constant expression */
1298     begin_macro(&str, 0);
1299     next();
1300     c = expr_const();
1301     end_macro();
1302     return c != 0;
1303 }
1304
1305
1306 /* parse after #define */
1307 ST_FUNC void parse_define(void)
1308 {
1309     Sym *s, *first, **ps;
1310     int v, t, varg, is_vaargs, spc;
1311     int saved_parse_flags = parse_flags;
1312     TokenString str;
1313
1314     v = tok;
1315     if (v < TOK_IDENT)
1316         tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1317     /* XXX: should check if same macro (ANSI) */
1318     first = NULL;
1319     t = MACRO_OBJ;
1320     /* '(' must be just after macro definition for MACRO_FUNC */
1321     parse_flags |= PARSE_FLAG_SPACES;
1322     next_nomacro_spc();
1323     if (tok == '(') {
1324         next_nomacro();
1325         ps = &first;
1326         if (tok != ')') for (;;) {
1327             varg = tok;
1328             next_nomacro();
1329             is_vaargs = 0;
1330             if (varg == TOK_DOTS) {
1331                 varg = TOK___VA_ARGS__;
1332                 is_vaargs = 1;
1333             } else if (tok == TOK_DOTS && gnu_ext) {
1334                 is_vaargs = 1;
1335                 next_nomacro();
1336             }
1337             if (varg < TOK_IDENT)
1338         bad_list:
1339                 tcc_error("bad macro parameter list");
1340             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1341             *ps = s;
1342             ps = &s->next;
1343             if (tok == ')')
1344                 break;
1345             if (tok != ',' || is_vaargs)
1346                 goto bad_list;
1347             next_nomacro();
1348         }
1349         next_nomacro_spc();
1350         t = MACRO_FUNC;
1351     }
1352     tok_str_new(&str);
1353     spc = 2;
1354     parse_flags |= PARSE_FLAG_ACCEPT_STRAYS | PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED;
1355     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1356         /* remove spaces around ## and after '#' */
1357         if (TOK_TWOSHARPS == tok) {
1358             if (2 == spc)
1359                 goto bad_twosharp;
1360             if (1 == spc)
1361                 --str.len;
1362             spc = 3;
1363         } else if ('#' == tok) {
1364             spc = 4;
1365         } else if (check_space(tok, &spc)) {
1366             goto skip;
1367         }
1368         tok_str_add2(&str, tok, &tokc);
1369     skip:
1370         next_nomacro_spc();
1371     }
1372
1373     parse_flags = saved_parse_flags;
1374     if (spc == 1)
1375         --str.len; /* remove trailing space */
1376     tok_str_add(&str, 0);
1377     if (3 == spc)
1378 bad_twosharp:
1379         tcc_error("'##' cannot appear at either end of macro");
1380     define_push(v, t, str.str, first);
1381     define_print(v);
1382 }
1383
1384 static inline int hash_cached_include(const char *filename)
1385 {
1386     const unsigned char *s;
1387     unsigned int h;
1388
1389     h = TOK_HASH_INIT;
1390     s = (unsigned char *) filename;
1391     while (*s) {
1392         h = TOK_HASH_FUNC(h, *s);
1393         s++;
1394     }
1395     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1396     return h;
1397 }
1398
1399 static CachedInclude *search_cached_include(TCCState *s1, const char *filename)
1400 {
1401     CachedInclude *e;
1402     int i, h;
1403     h = hash_cached_include(filename);
1404     i = s1->cached_includes_hash[h];
1405     for(;;) {
1406         if (i == 0)
1407             break;
1408         e = s1->cached_includes[i - 1];
1409         if (0 == PATHCMP(e->filename, filename))
1410             return e;
1411         i = e->hash_next;
1412     }
1413     return NULL;
1414 }
1415
1416 static inline void add_cached_include(TCCState *s1, const char *filename, int ifndef_macro)
1417 {
1418     CachedInclude *e;
1419     int h;
1420
1421     if (search_cached_include(s1, filename))
1422         return;
1423 #ifdef INC_DEBUG
1424     printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1425 #endif
1426     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1427     strcpy(e->filename, filename);
1428     e->ifndef_macro = ifndef_macro;
1429     dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1430     /* add in hash table */
1431     h = hash_cached_include(filename);
1432     e->hash_next = s1->cached_includes_hash[h];
1433     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1434 }
1435
1436 static void pragma_parse(TCCState *s1)
1437 {
1438     next_nomacro();
1439     if (tok == TOK_push_macro || tok == TOK_pop_macro) {
1440         int t = tok, v;
1441         Sym *s;
1442
1443         if (next(), tok != '(')
1444             goto pragma_err;
1445         if (next(), tok != TOK_STR)
1446             goto pragma_err;
1447         v = tok_alloc(tokc.cstr->data, tokc.cstr->size - 1)->tok;
1448         if (next(), tok != ')')
1449             goto pragma_err;
1450         if (t == TOK_push_macro) {
1451             while (NULL == (s = define_find(v)))
1452                 define_push(v, 0, NULL, NULL);
1453             s->type.ref = s; /* set push boundary */
1454         } else {
1455             for (s = define_stack; s; s = s->prev)
1456                 if (s->v == v && s->type.ref == s) {
1457                     s->type.ref = NULL;
1458                     break;
1459                 }
1460         }
1461         if (s)
1462             table_ident[v - TOK_IDENT]->sym_define = s->d ? s : NULL;
1463         else
1464             tcc_warning("unbalanced #pragma pop_macro");
1465
1466     } else if (tok == TOK_once) {
1467         add_cached_include(s1, file->filename, TOK_once);
1468
1469     } else if (s1->ppfp) {
1470         /* tcc -E: keep pragmas below unchanged */
1471         unget_tok(' ');
1472         unget_tok(TOK_PRAGMA);
1473         unget_tok('#');
1474         unget_tok(TOK_LINEFEED);
1475
1476     } else if (tok == TOK_pack) {
1477         /* This may be:
1478            #pragma pack(1) // set
1479            #pragma pack() // reset to default
1480            #pragma pack(push,1) // push & set
1481            #pragma pack(pop) // restore previous */
1482         next();
1483         skip('(');
1484         if (tok == TOK_ASM_pop) {
1485             next();
1486             if (s1->pack_stack_ptr <= s1->pack_stack) {
1487             stk_error:
1488                 tcc_error("out of pack stack");
1489             }
1490             s1->pack_stack_ptr--;
1491         } else {
1492             int val = 0;
1493             if (tok != ')') {
1494                 if (tok == TOK_ASM_push) {
1495                     next();
1496                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1497                         goto stk_error;
1498                     s1->pack_stack_ptr++;
1499                     skip(',');
1500                 }
1501                 if (tok != TOK_CINT)
1502                     goto pragma_err;
1503                 val = tokc.i;
1504                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1505                     goto pragma_err;
1506                 next();
1507             }
1508             *s1->pack_stack_ptr = val;
1509         }
1510         if (tok != ')')
1511             goto pragma_err;
1512
1513     } else if (tok == TOK_comment) {
1514         char *file;
1515         next();
1516         skip('(');
1517         if (tok != TOK_lib)
1518             goto pragma_warn;
1519         next();
1520         skip(',');
1521         if (tok != TOK_STR)
1522             goto pragma_err;
1523         file = tcc_strdup((char *)tokc.cstr->data);
1524         dynarray_add((void ***)&s1->pragma_libs, &s1->nb_pragma_libs, file);
1525         next();
1526         if (tok != ')')
1527             goto pragma_err;
1528     } else {
1529 pragma_warn:
1530         if (s1->warn_unsupported)
1531             tcc_warning("#pragma %s is ignored", get_tok_str(tok, &tokc));
1532     }
1533     return;
1534
1535 pragma_err:
1536     tcc_error("malformed #pragma directive");
1537     return;
1538 }
1539
1540 /* is_bof is true if first non space token at beginning of file */
1541 ST_FUNC void preprocess(int is_bof)
1542 {
1543     TCCState *s1 = tcc_state;
1544     int i, c, n, saved_parse_flags;
1545     char buf[1024], *q;
1546     Sym *s;
1547
1548     saved_parse_flags = parse_flags;
1549     parse_flags = PARSE_FLAG_PREPROCESS
1550         | PARSE_FLAG_TOK_NUM
1551         | PARSE_FLAG_TOK_STR
1552         | PARSE_FLAG_LINEFEED
1553         | (parse_flags & PARSE_FLAG_ASM_FILE)
1554         ;
1555
1556     next_nomacro();
1557  redo:
1558     switch(tok) {
1559     case TOK_DEFINE:
1560         next_nomacro();
1561         parse_define();
1562         break;
1563     case TOK_UNDEF:
1564         next_nomacro();
1565         s = define_find(tok);
1566         /* undefine symbol by putting an invalid name */
1567         if (s)
1568             define_undef(s);
1569         break;
1570     case TOK_INCLUDE:
1571     case TOK_INCLUDE_NEXT:
1572         ch = file->buf_ptr[0];
1573         skip_spaces(); /* XXX: incorrect if comments : use next_nomacro with a special mode */
1574         c = 0;
1575         if (ch == '<')
1576             c = '>';
1577         if (ch == '\"')
1578             c = ch;
1579         if (c) {
1580             inp();
1581             q = buf;
1582             while (ch != c && ch != '\n' && ch != CH_EOF) {
1583                 if ((q - buf) < sizeof(buf) - 1)
1584                     *q++ = ch;
1585                 if (ch == '\\') {
1586                     if (handle_stray_noerror() == 0)
1587                         --q;
1588                 } else
1589                     inp();
1590             }
1591             *q = '\0';
1592             minp();
1593         } else {
1594             /* computed #include : either we have only strings or
1595                we have anything enclosed in '<>' */
1596             next();
1597             buf[0] = '\0';
1598             if (tok == TOK_STR) {
1599                 while (tok != TOK_LINEFEED) {
1600                     if (tok != TOK_STR) {
1601                     include_syntax:
1602                         tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1603                     }
1604                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1605                     next();
1606                 }
1607                 c = '\"';
1608             } else {
1609                 int len;
1610                 while (tok != TOK_LINEFEED) {
1611                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1612                     next();
1613                 }
1614                 len = strlen(buf);
1615                 /* check syntax and remove '<>' */
1616                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1617                     goto include_syntax;
1618                 memmove(buf, buf + 1, len - 2);
1619                 buf[len - 2] = '\0';
1620                 c = '>';
1621             }
1622         }
1623
1624         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1625             tcc_error("#include recursion too deep");
1626
1627         i = -2;
1628         if (tok == TOK_INCLUDE_NEXT)
1629             i = file->inc_path_index + 1;
1630
1631         n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1632         for (; i < n; ++i) {
1633             char buf1[sizeof file->filename];
1634             CachedInclude *e;
1635             const char *path;
1636
1637             if (i == -2) {
1638                 /* check absolute include path */
1639                 if (!IS_ABSPATH(buf))
1640                     continue;
1641                 buf1[0] = 0;
1642                 i = n - 1; /* force end loop */
1643
1644             } else if (i == -1) {
1645                 /* search in current dir if "header.h" */
1646                 if (c != '\"')
1647                     continue;
1648                 path = file->filename;
1649                 pstrncpy(buf1, path, tcc_basename(path) - path);
1650
1651             } else {
1652                 /* search in all the include paths */
1653                 if (i < s1->nb_include_paths)
1654                     path = s1->include_paths[i];
1655                 else
1656                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1657                 if (path == 0) continue;
1658                 pstrcpy(buf1, sizeof(buf1), path);
1659                 pstrcat(buf1, sizeof(buf1), "/");
1660             }
1661
1662             pstrcat(buf1, sizeof(buf1), buf);
1663             e = search_cached_include(s1, buf1);
1664             if (e && (define_find(e->ifndef_macro) || e->ifndef_macro == TOK_once))
1665                 break; /* no need to parse the include */
1666
1667             if (tcc_open(s1, buf1) < 0)
1668                 continue;
1669
1670             file->inc_path_index = i;
1671             *(s1->include_stack_ptr++) = file->prev;
1672
1673             /* update target deps */
1674             dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1675                 tcc_strdup(buf1));
1676
1677             /* add include file debug info */
1678             if (s1->do_debug)
1679                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1680
1681             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1682             ch = file->buf_ptr[0];
1683             break;
1684         }
1685         if (i >= n) tcc_error("include file '%s' not found", buf);
1686         goto the_end;
1687     case TOK_IFNDEF:
1688         c = 1;
1689         goto do_ifdef;
1690     case TOK_IF:
1691         c = expr_preprocess();
1692         goto do_if;
1693     case TOK_IFDEF:
1694         c = 0;
1695     do_ifdef:
1696         next_nomacro();
1697         if (tok < TOK_IDENT)
1698             tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1699         if (is_bof) {
1700             if (c) {
1701                 #ifdef INC_DEBUG
1702                     printf("#ifndef %s\n", get_tok_str(tok, NULL));
1703                 #endif
1704                 file->ifndef_macro = tok;
1705             }
1706         }
1707         c = (define_find(tok) != 0) ^ c;
1708     do_if:
1709         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1710             tcc_error("memory full (ifdef)");
1711         *s1->ifdef_stack_ptr++ = c;
1712         goto test_skip;
1713     case TOK_ELSE:
1714         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1715             tcc_error("#else without matching #if");
1716         if (s1->ifdef_stack_ptr[-1] & 2)
1717             tcc_error("#else after #else");
1718         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1719         goto test_else;
1720     case TOK_ELIF:
1721         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1722             tcc_error("#elif without matching #if");
1723         c = s1->ifdef_stack_ptr[-1];
1724         if (c > 1)
1725             tcc_error("#elif after #else");
1726         /* last #if/#elif expression was true: we skip */
1727         if (c == 1)
1728             goto skip;
1729         c = expr_preprocess();
1730         s1->ifdef_stack_ptr[-1] = c;
1731     test_else:
1732         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1733             file->ifndef_macro = 0;
1734     test_skip:
1735         if (!(c & 1)) {
1736         skip:
1737             preprocess_skip();
1738             is_bof = 0;
1739             goto redo;
1740         }
1741         break;
1742     case TOK_ENDIF:
1743         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1744             tcc_error("#endif without matching #if");
1745         s1->ifdef_stack_ptr--;
1746         /* '#ifndef macro' was at the start of file. Now we check if
1747            an '#endif' is exactly at the end of file */
1748         if (file->ifndef_macro &&
1749             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1750             file->ifndef_macro_saved = file->ifndef_macro;
1751             /* need to set to zero to avoid false matches if another
1752                #ifndef at middle of file */
1753             file->ifndef_macro = 0;
1754             while (tok != TOK_LINEFEED)
1755                 next_nomacro();
1756             tok_flags |= TOK_FLAG_ENDIF;
1757             goto the_end;
1758         }
1759         break;
1760     case TOK_PPNUM:
1761         n = strtoul((char*)tokc.cstr->data, &q, 10);
1762         goto _line_num;
1763     case TOK_LINE:
1764         next();
1765         if (tok != TOK_CINT)
1766 _line_err:
1767             tcc_error("wrong #line format");
1768         n = tokc.i;
1769 _line_num:
1770         next();
1771         if (tok != TOK_LINEFEED) {
1772             if (tok == TOK_STR)
1773                 pstrcpy(file->filename, sizeof(file->filename), (char *)tokc.cstr->data);
1774             else if (parse_flags & PARSE_FLAG_ASM_FILE)
1775                 break;
1776             else
1777                 goto _line_err;
1778             --n;
1779         }
1780         if (file->fd > 0)
1781             total_lines += file->line_num - n;
1782         file->line_num = n;
1783         if (s1->do_debug)
1784             put_stabs(file->filename, N_BINCL, 0, 0, 0);
1785         break;
1786     case TOK_ERROR:
1787     case TOK_WARNING:
1788         c = tok;
1789         ch = file->buf_ptr[0];
1790         skip_spaces();
1791         q = buf;
1792         while (ch != '\n' && ch != CH_EOF) {
1793             if ((q - buf) < sizeof(buf) - 1)
1794                 *q++ = ch;
1795             if (ch == '\\') {
1796                 if (handle_stray_noerror() == 0)
1797                     --q;
1798             } else
1799                 inp();
1800         }
1801         *q = '\0';
1802         if (c == TOK_ERROR)
1803             tcc_error("#error %s", buf);
1804         else
1805             tcc_warning("#warning %s", buf);
1806         break;
1807     case TOK_PRAGMA:
1808         pragma_parse(s1);
1809         break;
1810     case TOK_LINEFEED:
1811         goto the_end;
1812     default:
1813         /* ignore gas line comment in an 'S' file. */
1814         if (saved_parse_flags & PARSE_FLAG_ASM_FILE)
1815             goto ignore;
1816         if (tok == '!' && is_bof)
1817             /* '!' is ignored at beginning to allow C scripts. */
1818             goto ignore;
1819         tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1820     ignore:
1821         file->buf_ptr = parse_line_comment(file->buf_ptr);
1822         goto the_end;
1823     }
1824     /* ignore other preprocess commands or #! for C scripts */
1825     while (tok != TOK_LINEFEED)
1826         next_nomacro();
1827  the_end:
1828     parse_flags = saved_parse_flags;
1829 }
1830
1831 /* evaluate escape codes in a string. */
1832 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1833 {
1834     int c, n;
1835     const uint8_t *p;
1836
1837     p = buf;
1838     for(;;) {
1839         c = *p;
1840         if (c == '\0')
1841             break;
1842         if (c == '\\') {
1843             p++;
1844             /* escape */
1845             c = *p;
1846             switch(c) {
1847             case '0': case '1': case '2': case '3':
1848             case '4': case '5': case '6': case '7':
1849                 /* at most three octal digits */
1850                 n = c - '0';
1851                 p++;
1852                 c = *p;
1853                 if (isoct(c)) {
1854                     n = n * 8 + c - '0';
1855                     p++;
1856                     c = *p;
1857                     if (isoct(c)) {
1858                         n = n * 8 + c - '0';
1859                         p++;
1860                     }
1861                 }
1862                 c = n;
1863                 goto add_char_nonext;
1864             case 'x':
1865             case 'u':
1866             case 'U':
1867                 p++;
1868                 n = 0;
1869                 for(;;) {
1870                     c = *p;
1871                     if (c >= 'a' && c <= 'f')
1872                         c = c - 'a' + 10;
1873                     else if (c >= 'A' && c <= 'F')
1874                         c = c - 'A' + 10;
1875                     else if (isnum(c))
1876                         c = c - '0';
1877                     else
1878                         break;
1879                     n = n * 16 + c;
1880                     p++;
1881                 }
1882                 c = n;
1883                 goto add_char_nonext;
1884             case 'a':
1885                 c = '\a';
1886                 break;
1887             case 'b':
1888                 c = '\b';
1889                 break;
1890             case 'f':
1891                 c = '\f';
1892                 break;
1893             case 'n':
1894                 c = '\n';
1895                 break;
1896             case 'r':
1897                 c = '\r';
1898                 break;
1899             case 't':
1900                 c = '\t';
1901                 break;
1902             case 'v':
1903                 c = '\v';
1904                 break;
1905             case 'e':
1906                 if (!gnu_ext)
1907                     goto invalid_escape;
1908                 c = 27;
1909                 break;
1910             case '\'':
1911             case '\"':
1912             case '\\':
1913             case '?':
1914                 break;
1915             default:
1916             invalid_escape:
1917                 if (c >= '!' && c <= '~')
1918                     tcc_warning("unknown escape sequence: \'\\%c\'", c);
1919                 else
1920                     tcc_warning("unknown escape sequence: \'\\x%x\'", c);
1921                 break;
1922             }
1923         }
1924         p++;
1925     add_char_nonext:
1926         if (!is_long)
1927             cstr_ccat(outstr, c);
1928         else
1929             cstr_wccat(outstr, c);
1930     }
1931     /* add a trailing '\0' */
1932     if (!is_long)
1933         cstr_ccat(outstr, '\0');
1934     else
1935         cstr_wccat(outstr, '\0');
1936 }
1937
1938 void parse_string(const char *s, int len)
1939 {
1940     uint8_t buf[1000], *p = buf;
1941     int is_long, sep;
1942
1943     if ((is_long = *s == 'L'))
1944         ++s, --len;
1945     sep = *s++;
1946     len -= 2;
1947     if (len >= sizeof buf)
1948         p = tcc_malloc(len + 1);
1949     memcpy(p, s, len);
1950     p[len] = 0;
1951
1952     cstr_reset(&tokcstr);
1953     parse_escape_string(&tokcstr, p, is_long);
1954     if (p != buf)
1955         tcc_free(p);
1956
1957     if (sep == '\'') {
1958         int char_size;
1959         /* XXX: make it portable */
1960         if (!is_long)
1961             char_size = 1;
1962         else
1963             char_size = sizeof(nwchar_t);
1964         if (tokcstr.size <= char_size)
1965             tcc_error("empty character constant");
1966         if (tokcstr.size > 2 * char_size)
1967             tcc_warning("multi-character character constant");
1968         if (!is_long) {
1969             tokc.i = *(int8_t *)tokcstr.data;
1970             tok = TOK_CCHAR;
1971         } else {
1972             tokc.i = *(nwchar_t *)tokcstr.data;
1973             tok = TOK_LCHAR;
1974         }
1975     } else {
1976         tokc.cstr = &tokcstr;
1977         if (!is_long)
1978             tok = TOK_STR;
1979         else
1980             tok = TOK_LSTR;
1981     }
1982 }
1983
1984 /* we use 64 bit numbers */
1985 #define BN_SIZE 2
1986
1987 /* bn = (bn << shift) | or_val */
1988 static void bn_lshift(unsigned int *bn, int shift, int or_val)
1989 {
1990     int i;
1991     unsigned int v;
1992     for(i=0;i<BN_SIZE;i++) {
1993         v = bn[i];
1994         bn[i] = (v << shift) | or_val;
1995         or_val = v >> (32 - shift);
1996     }
1997 }
1998
1999 static void bn_zero(unsigned int *bn)
2000 {
2001     int i;
2002     for(i=0;i<BN_SIZE;i++) {
2003         bn[i] = 0;
2004     }
2005 }
2006
2007 /* parse number in null terminated string 'p' and return it in the
2008    current token */
2009 static void parse_number(const char *p)
2010 {
2011     int b, t, shift, frac_bits, s, exp_val, ch;
2012     char *q;
2013     unsigned int bn[BN_SIZE];
2014     double d;
2015
2016     /* number */
2017     q = token_buf;
2018     ch = *p++;
2019     t = ch;
2020     ch = *p++;
2021     *q++ = t;
2022     b = 10;
2023     if (t == '.') {
2024         goto float_frac_parse;
2025     } else if (t == '0') {
2026         if (ch == 'x' || ch == 'X') {
2027             q--;
2028             ch = *p++;
2029             b = 16;
2030         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
2031             q--;
2032             ch = *p++;
2033             b = 2;
2034         }
2035     }
2036     /* parse all digits. cannot check octal numbers at this stage
2037        because of floating point constants */
2038     while (1) {
2039         if (ch >= 'a' && ch <= 'f')
2040             t = ch - 'a' + 10;
2041         else if (ch >= 'A' && ch <= 'F')
2042             t = ch - 'A' + 10;
2043         else if (isnum(ch))
2044             t = ch - '0';
2045         else
2046             break;
2047         if (t >= b)
2048             break;
2049         if (q >= token_buf + STRING_MAX_SIZE) {
2050         num_too_long:
2051             tcc_error("number too long");
2052         }
2053         *q++ = ch;
2054         ch = *p++;
2055     }
2056     if (ch == '.' ||
2057         ((ch == 'e' || ch == 'E') && b == 10) ||
2058         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
2059         if (b != 10) {
2060             /* NOTE: strtox should support that for hexa numbers, but
2061                non ISOC99 libcs do not support it, so we prefer to do
2062                it by hand */
2063             /* hexadecimal or binary floats */
2064             /* XXX: handle overflows */
2065             *q = '\0';
2066             if (b == 16)
2067                 shift = 4;
2068             else
2069                 shift = 1;
2070             bn_zero(bn);
2071             q = token_buf;
2072             while (1) {
2073                 t = *q++;
2074                 if (t == '\0') {
2075                     break;
2076                 } else if (t >= 'a') {
2077                     t = t - 'a' + 10;
2078                 } else if (t >= 'A') {
2079                     t = t - 'A' + 10;
2080                 } else {
2081                     t = t - '0';
2082                 }
2083                 bn_lshift(bn, shift, t);
2084             }
2085             frac_bits = 0;
2086             if (ch == '.') {
2087                 ch = *p++;
2088                 while (1) {
2089                     t = ch;
2090                     if (t >= 'a' && t <= 'f') {
2091                         t = t - 'a' + 10;
2092                     } else if (t >= 'A' && t <= 'F') {
2093                         t = t - 'A' + 10;
2094                     } else if (t >= '0' && t <= '9') {
2095                         t = t - '0';
2096                     } else {
2097                         break;
2098                     }
2099                     if (t >= b)
2100                         tcc_error("invalid digit");
2101                     bn_lshift(bn, shift, t);
2102                     frac_bits += shift;
2103                     ch = *p++;
2104                 }
2105             }
2106             if (ch != 'p' && ch != 'P')
2107                 expect("exponent");
2108             ch = *p++;
2109             s = 1;
2110             exp_val = 0;
2111             if (ch == '+') {
2112                 ch = *p++;
2113             } else if (ch == '-') {
2114                 s = -1;
2115                 ch = *p++;
2116             }
2117             if (ch < '0' || ch > '9')
2118                 expect("exponent digits");
2119             while (ch >= '0' && ch <= '9') {
2120                 exp_val = exp_val * 10 + ch - '0';
2121                 ch = *p++;
2122             }
2123             exp_val = exp_val * s;
2124
2125             /* now we can generate the number */
2126             /* XXX: should patch directly float number */
2127             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
2128             d = ldexp(d, exp_val - frac_bits);
2129             t = toup(ch);
2130             if (t == 'F') {
2131                 ch = *p++;
2132                 tok = TOK_CFLOAT;
2133                 /* float : should handle overflow */
2134                 tokc.f = (float)d;
2135             } else if (t == 'L') {
2136                 ch = *p++;
2137 #ifdef TCC_TARGET_PE
2138                 tok = TOK_CDOUBLE;
2139                 tokc.d = d;
2140 #else
2141                 tok = TOK_CLDOUBLE;
2142                 /* XXX: not large enough */
2143                 tokc.ld = (long double)d;
2144 #endif
2145             } else {
2146                 tok = TOK_CDOUBLE;
2147                 tokc.d = d;
2148             }
2149         } else {
2150             /* decimal floats */
2151             if (ch == '.') {
2152                 if (q >= token_buf + STRING_MAX_SIZE)
2153                     goto num_too_long;
2154                 *q++ = ch;
2155                 ch = *p++;
2156             float_frac_parse:
2157                 while (ch >= '0' && ch <= '9') {
2158                     if (q >= token_buf + STRING_MAX_SIZE)
2159                         goto num_too_long;
2160                     *q++ = ch;
2161                     ch = *p++;
2162                 }
2163             }
2164             if (ch == 'e' || ch == 'E') {
2165                 if (q >= token_buf + STRING_MAX_SIZE)
2166                     goto num_too_long;
2167                 *q++ = ch;
2168                 ch = *p++;
2169                 if (ch == '-' || ch == '+') {
2170                     if (q >= token_buf + STRING_MAX_SIZE)
2171                         goto num_too_long;
2172                     *q++ = ch;
2173                     ch = *p++;
2174                 }
2175                 if (ch < '0' || ch > '9')
2176                     expect("exponent digits");
2177                 while (ch >= '0' && ch <= '9') {
2178                     if (q >= token_buf + STRING_MAX_SIZE)
2179                         goto num_too_long;
2180                     *q++ = ch;
2181                     ch = *p++;
2182                 }
2183             }
2184             *q = '\0';
2185             t = toup(ch);
2186             errno = 0;
2187             if (t == 'F') {
2188                 ch = *p++;
2189                 tok = TOK_CFLOAT;
2190                 tokc.f = strtof(token_buf, NULL);
2191             } else if (t == 'L') {
2192                 ch = *p++;
2193 #ifdef TCC_TARGET_PE
2194                 tok = TOK_CDOUBLE;
2195                 tokc.d = strtod(token_buf, NULL);
2196 #else
2197                 tok = TOK_CLDOUBLE;
2198                 tokc.ld = strtold(token_buf, NULL);
2199 #endif
2200             } else {
2201                 tok = TOK_CDOUBLE;
2202                 tokc.d = strtod(token_buf, NULL);
2203             }
2204         }
2205     } else {
2206         unsigned long long n, n1;
2207         int lcount, ucount, must_64bit;
2208         const char *p1;
2209
2210         /* integer number */
2211         *q = '\0';
2212         q = token_buf;
2213         if (b == 10 && *q == '0') {
2214             b = 8;
2215             q++;
2216         }
2217         n = 0;
2218         while(1) {
2219             t = *q++;
2220             /* no need for checks except for base 10 / 8 errors */
2221             if (t == '\0')
2222                 break;
2223             else if (t >= 'a')
2224                 t = t - 'a' + 10;
2225             else if (t >= 'A')
2226                 t = t - 'A' + 10;
2227             else
2228                 t = t - '0';
2229             if (t >= b)
2230                 tcc_error("invalid digit");
2231             n1 = n;
2232             n = n * b + t;
2233             /* detect overflow */
2234             /* XXX: this test is not reliable */
2235             if (n < n1)
2236                 tcc_error("integer constant overflow");
2237         }
2238
2239         /* Determine the characteristics (unsigned and/or 64bit) the type of
2240            the constant must have according to the constant suffix(es) */
2241         lcount = ucount = must_64bit = 0;
2242         p1 = p;
2243         for(;;) {
2244             t = toup(ch);
2245             if (t == 'L') {
2246                 if (lcount >= 2)
2247                     tcc_error("three 'l's in integer constant");
2248                 if (lcount && *(p - 1) != ch)
2249                     tcc_error("incorrect integer suffix: %s", p1);
2250                 lcount++;
2251 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
2252                 if (lcount == 2)
2253 #endif
2254                     must_64bit = 1;
2255                 ch = *p++;
2256             } else if (t == 'U') {
2257                 if (ucount >= 1)
2258                     tcc_error("two 'u's in integer constant");
2259                 ucount++;
2260                 ch = *p++;
2261             } else {
2262                 break;
2263             }
2264         }
2265
2266         /* Whether 64 bits are needed to hold the constant's value */
2267         if (n & 0xffffffff00000000LL || must_64bit) {
2268             tok = TOK_CLLONG;
2269             n1 = n >> 32;
2270         } else {
2271             tok = TOK_CINT;
2272             n1 = n;
2273         }
2274
2275         /* Whether type must be unsigned to hold the constant's value */
2276         if (ucount || ((n1 >> 31) && (b != 10))) {
2277             if (tok == TOK_CLLONG)
2278                 tok = TOK_CULLONG;
2279             else
2280                 tok = TOK_CUINT;
2281         /* If decimal and no unsigned suffix, bump to 64 bits or throw error */
2282         } else if (n1 >> 31) {
2283             if (tok == TOK_CINT)
2284                 tok = TOK_CLLONG;
2285             else
2286                 tcc_error("integer constant overflow");
2287         }
2288
2289         if (tok == TOK_CINT || tok == TOK_CUINT)
2290             tokc.i = n;
2291         else
2292             tokc.i = n;
2293     }
2294     if (ch)
2295         tcc_error("invalid number\n");
2296 }
2297
2298
2299 #define PARSE2(c1, tok1, c2, tok2)              \
2300     case c1:                                    \
2301         PEEKC(c, p);                            \
2302         if (c == c2) {                          \
2303             p++;                                \
2304             tok = tok2;                         \
2305         } else {                                \
2306             tok = tok1;                         \
2307         }                                       \
2308         break;
2309
2310 /* return next token without macro substitution */
2311 static inline void next_nomacro1(void)
2312 {
2313     int t, c, is_long;
2314     TokenSym *ts;
2315     uint8_t *p, *p1;
2316     unsigned int h;
2317
2318     p = file->buf_ptr;
2319  redo_no_start:
2320     c = *p;
2321     switch(c) {
2322     case ' ':
2323     case '\t':
2324         tok = c;
2325         p++;
2326         if (parse_flags & PARSE_FLAG_SPACES)
2327             goto keep_tok_flags;
2328         while (isidnum_table[*p - CH_EOF] & IS_SPC)
2329             ++p;
2330         goto redo_no_start;
2331     case '\f':
2332     case '\v':
2333     case '\r':
2334         p++;
2335         goto redo_no_start;
2336     case '\\':
2337         /* first look if it is in fact an end of buffer */
2338         c = handle_stray1(p);
2339         p = file->buf_ptr;
2340         if (c == '\\')
2341             goto parse_simple;
2342         if (c != CH_EOF)
2343             goto redo_no_start;
2344         {
2345             TCCState *s1 = tcc_state;
2346             if ((parse_flags & PARSE_FLAG_LINEFEED)
2347                 && !(tok_flags & TOK_FLAG_EOF)) {
2348                 tok_flags |= TOK_FLAG_EOF;
2349                 tok = TOK_LINEFEED;
2350                 goto keep_tok_flags;
2351             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2352                 tok = TOK_EOF;
2353             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2354                 tcc_error("missing #endif");
2355             } else if (s1->include_stack_ptr == s1->include_stack) {
2356                 /* no include left : end of file. */
2357                 tok = TOK_EOF;
2358             } else {
2359                 tok_flags &= ~TOK_FLAG_EOF;
2360                 /* pop include file */
2361
2362                 /* test if previous '#endif' was after a #ifdef at
2363                    start of file */
2364                 if (tok_flags & TOK_FLAG_ENDIF) {
2365 #ifdef INC_DEBUG
2366                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2367 #endif
2368                     add_cached_include(s1, file->filename, file->ifndef_macro_saved);
2369                     tok_flags &= ~TOK_FLAG_ENDIF;
2370                 }
2371
2372                 /* add end of include file debug info */
2373                 if (tcc_state->do_debug) {
2374                     put_stabd(N_EINCL, 0, 0);
2375                 }
2376                 /* pop include stack */
2377                 tcc_close();
2378                 s1->include_stack_ptr--;
2379                 p = file->buf_ptr;
2380                 goto redo_no_start;
2381             }
2382         }
2383         break;
2384
2385     case '\n':
2386         file->line_num++;
2387         tok_flags |= TOK_FLAG_BOL;
2388         p++;
2389 maybe_newline:
2390         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2391             goto redo_no_start;
2392         tok = TOK_LINEFEED;
2393         goto keep_tok_flags;
2394
2395     case '#':
2396         /* XXX: simplify */
2397         PEEKC(c, p);
2398         if ((tok_flags & TOK_FLAG_BOL) &&
2399             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2400             file->buf_ptr = p;
2401             preprocess(tok_flags & TOK_FLAG_BOF);
2402             p = file->buf_ptr;
2403             goto maybe_newline;
2404         } else {
2405             if (c == '#') {
2406                 p++;
2407                 tok = TOK_TWOSHARPS;
2408             } else {
2409                 if (parse_flags & PARSE_FLAG_ASM_FILE) {
2410                     p = parse_line_comment(p - 1);
2411                     goto redo_no_start;
2412                 } else {
2413                     tok = '#';
2414                 }
2415             }
2416         }
2417         break;
2418
2419     /* dollar is allowed to start identifiers when not parsing asm */
2420     case '$':
2421         if (!(isidnum_table[c - CH_EOF] & IS_ID)
2422          || (parse_flags & PARSE_FLAG_ASM_FILE))
2423             goto parse_simple;
2424
2425     case 'a': case 'b': case 'c': case 'd':
2426     case 'e': case 'f': case 'g': case 'h':
2427     case 'i': case 'j': case 'k': case 'l':
2428     case 'm': case 'n': case 'o': case 'p':
2429     case 'q': case 'r': case 's': case 't':
2430     case 'u': case 'v': case 'w': case 'x':
2431     case 'y': case 'z':
2432     case 'A': case 'B': case 'C': case 'D':
2433     case 'E': case 'F': case 'G': case 'H':
2434     case 'I': case 'J': case 'K':
2435     case 'M': case 'N': case 'O': case 'P':
2436     case 'Q': case 'R': case 'S': case 'T':
2437     case 'U': case 'V': case 'W': case 'X':
2438     case 'Y': case 'Z':
2439     case '_':
2440     parse_ident_fast:
2441         p1 = p;
2442         h = TOK_HASH_INIT;
2443         h = TOK_HASH_FUNC(h, c);
2444         while (c = *++p, isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2445             h = TOK_HASH_FUNC(h, c);
2446         if (c != '\\') {
2447             TokenSym **pts;
2448             int len;
2449
2450             /* fast case : no stray found, so we have the full token
2451                and we have already hashed it */
2452             len = p - p1;
2453             h &= (TOK_HASH_SIZE - 1);
2454             pts = &hash_ident[h];
2455             for(;;) {
2456                 ts = *pts;
2457                 if (!ts)
2458                     break;
2459                 if (ts->len == len && !memcmp(ts->str, p1, len))
2460                     goto token_found;
2461                 pts = &(ts->hash_next);
2462             }
2463             ts = tok_alloc_new(pts, (char *) p1, len);
2464         token_found: ;
2465         } else {
2466             /* slower case */
2467             cstr_reset(&tokcstr);
2468
2469             while (p1 < p) {
2470                 cstr_ccat(&tokcstr, *p1);
2471                 p1++;
2472             }
2473             p--;
2474             PEEKC(c, p);
2475         parse_ident_slow:
2476             while (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM)) {
2477                 cstr_ccat(&tokcstr, c);
2478                 PEEKC(c, p);
2479             }
2480             ts = tok_alloc(tokcstr.data, tokcstr.size);
2481         }
2482         tok = ts->tok;
2483         break;
2484     case 'L':
2485         t = p[1];
2486         if (t != '\\' && t != '\'' && t != '\"') {
2487             /* fast case */
2488             goto parse_ident_fast;
2489         } else {
2490             PEEKC(c, p);
2491             if (c == '\'' || c == '\"') {
2492                 is_long = 1;
2493                 goto str_const;
2494             } else {
2495                 cstr_reset(&tokcstr);
2496                 cstr_ccat(&tokcstr, 'L');
2497                 goto parse_ident_slow;
2498             }
2499         }
2500         break;
2501
2502     case '0': case '1': case '2': case '3':
2503     case '4': case '5': case '6': case '7':
2504     case '8': case '9':
2505         cstr_reset(&tokcstr);
2506         /* after the first digit, accept digits, alpha, '.' or sign if
2507            prefixed by 'eEpP' */
2508     parse_num:
2509         for(;;) {
2510             t = c;
2511             cstr_ccat(&tokcstr, c);
2512             PEEKC(c, p);
2513             if (!((isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2514                   || c == '.'
2515                   || ((c == '+' || c == '-')
2516                       && (t == 'e' || t == 'E' || t == 'p' || t == 'P')
2517                       )))
2518                 break;
2519         }
2520         /* We add a trailing '\0' to ease parsing */
2521         cstr_ccat(&tokcstr, '\0');
2522         tokc.cstr = &tokcstr;
2523         tok = TOK_PPNUM;
2524         break;
2525
2526     case '.':
2527         /* special dot handling because it can also start a number */
2528         PEEKC(c, p);
2529         if (isnum(c)) {
2530             cstr_reset(&tokcstr);
2531             cstr_ccat(&tokcstr, '.');
2532             goto parse_num;
2533         } else if ((c == '.') && (p[1] == '.')){
2534             PEEKC(c, p);
2535             PEEKC(c, p);
2536             tok = TOK_DOTS;
2537         } else {
2538             tok = '.';
2539         }
2540         break;
2541     case '\'':
2542     case '\"':
2543         is_long = 0;
2544     str_const:
2545         cstr_reset(&tokcstr);
2546         if (is_long)
2547             cstr_ccat(&tokcstr, 'L');
2548         cstr_ccat(&tokcstr, c);
2549         p = parse_pp_string(p, c, &tokcstr);
2550         cstr_ccat(&tokcstr, c);
2551         cstr_ccat(&tokcstr, '\0');
2552         tokc.cstr = &tokcstr;
2553         tok = TOK_PPSTR;
2554         break;
2555
2556     case '<':
2557         PEEKC(c, p);
2558         if (c == '=') {
2559             p++;
2560             tok = TOK_LE;
2561         } else if (c == '<') {
2562             PEEKC(c, p);
2563             if (c == '=') {
2564                 p++;
2565                 tok = TOK_A_SHL;
2566             } else {
2567                 tok = TOK_SHL;
2568             }
2569         } else {
2570             tok = TOK_LT;
2571         }
2572         break;
2573     case '>':
2574         PEEKC(c, p);
2575         if (c == '=') {
2576             p++;
2577             tok = TOK_GE;
2578         } else if (c == '>') {
2579             PEEKC(c, p);
2580             if (c == '=') {
2581                 p++;
2582                 tok = TOK_A_SAR;
2583             } else {
2584                 tok = TOK_SAR;
2585             }
2586         } else {
2587             tok = TOK_GT;
2588         }
2589         break;
2590
2591     case '&':
2592         PEEKC(c, p);
2593         if (c == '&') {
2594             p++;
2595             tok = TOK_LAND;
2596         } else if (c == '=') {
2597             p++;
2598             tok = TOK_A_AND;
2599         } else {
2600             tok = '&';
2601         }
2602         break;
2603
2604     case '|':
2605         PEEKC(c, p);
2606         if (c == '|') {
2607             p++;
2608             tok = TOK_LOR;
2609         } else if (c == '=') {
2610             p++;
2611             tok = TOK_A_OR;
2612         } else {
2613             tok = '|';
2614         }
2615         break;
2616
2617     case '+':
2618         PEEKC(c, p);
2619         if (c == '+') {
2620             p++;
2621             tok = TOK_INC;
2622         } else if (c == '=') {
2623             p++;
2624             tok = TOK_A_ADD;
2625         } else {
2626             tok = '+';
2627         }
2628         break;
2629
2630     case '-':
2631         PEEKC(c, p);
2632         if (c == '-') {
2633             p++;
2634             tok = TOK_DEC;
2635         } else if (c == '=') {
2636             p++;
2637             tok = TOK_A_SUB;
2638         } else if (c == '>') {
2639             p++;
2640             tok = TOK_ARROW;
2641         } else {
2642             tok = '-';
2643         }
2644         break;
2645
2646     PARSE2('!', '!', '=', TOK_NE)
2647     PARSE2('=', '=', '=', TOK_EQ)
2648     PARSE2('*', '*', '=', TOK_A_MUL)
2649     PARSE2('%', '%', '=', TOK_A_MOD)
2650     PARSE2('^', '^', '=', TOK_A_XOR)
2651
2652         /* comments or operator */
2653     case '/':
2654         PEEKC(c, p);
2655         if (c == '*') {
2656             p = parse_comment(p);
2657             /* comments replaced by a blank */
2658             tok = ' ';
2659             goto keep_tok_flags;
2660         } else if (c == '/') {
2661             p = parse_line_comment(p);
2662             tok = ' ';
2663             goto keep_tok_flags;
2664         } else if (c == '=') {
2665             p++;
2666             tok = TOK_A_DIV;
2667         } else {
2668             tok = '/';
2669         }
2670         break;
2671
2672         /* simple tokens */
2673     case '(':
2674     case ')':
2675     case '[':
2676     case ']':
2677     case '{':
2678     case '}':
2679     case ',':
2680     case ';':
2681     case ':':
2682     case '?':
2683     case '~':
2684     case '@': /* only used in assembler */
2685     parse_simple:
2686         tok = c;
2687         p++;
2688         break;
2689     default:
2690         tcc_error("unrecognized character \\x%02x", c);
2691         break;
2692     }
2693     tok_flags = 0;
2694 keep_tok_flags:
2695     file->buf_ptr = p;
2696 #if defined(PARSE_DEBUG)
2697     printf("token = %s\n", get_tok_str(tok, &tokc));
2698 #endif
2699 }
2700
2701 /* return next token without macro substitution. Can read input from
2702    macro_ptr buffer */
2703 static void next_nomacro_spc(void)
2704 {
2705     if (macro_ptr) {
2706     redo:
2707         tok = *macro_ptr;
2708         if (tok) {
2709             TOK_GET(&tok, &macro_ptr, &tokc);
2710             if (tok == TOK_LINENUM) {
2711                 file->line_num = tokc.i;
2712                 goto redo;
2713             }
2714         }
2715     } else {
2716         next_nomacro1();
2717     }
2718 }
2719
2720 ST_FUNC void next_nomacro(void)
2721 {
2722     do {
2723         next_nomacro_spc();
2724     } while (tok < 256 && (isidnum_table[tok - CH_EOF] & IS_SPC));
2725 }
2726
2727
2728 static void macro_subst(
2729     TokenString *tok_str,
2730     Sym **nested_list,
2731     const int *macro_str,
2732     int can_read_stream
2733     );
2734
2735 /* substitute arguments in replacement lists in macro_str by the values in
2736    args (field d) and return allocated string */
2737 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2738 {
2739     int t, t0, t1, spc;
2740     const int *st;
2741     Sym *s;
2742     CValue cval;
2743     TokenString str;
2744     CString cstr;
2745
2746     tok_str_new(&str);
2747     t0 = t1 = 0;
2748     while(1) {
2749         TOK_GET(&t, &macro_str, &cval);
2750         if (!t)
2751             break;
2752         if (t == '#') {
2753             /* stringize */
2754             TOK_GET(&t, &macro_str, &cval);
2755             if (!t)
2756                 goto bad_stringy;
2757             s = sym_find2(args, t);
2758             if (s) {
2759                 cstr_new(&cstr);
2760                 cstr_ccat(&cstr, '\"');
2761                 st = s->d;
2762                 spc = 0;
2763                 while (*st) {
2764                     TOK_GET(&t, &st, &cval);
2765                     if (t != TOK_PLCHLDR
2766                      && t != TOK_NOSUBST
2767                      && 0 == check_space(t, &spc)) {
2768                         const char *s = get_tok_str(t, &cval);
2769                         while (*s) {
2770                             if (t == TOK_PPSTR && *s != '\'')
2771                                 add_char(&cstr, *s);
2772                             else
2773                                 cstr_ccat(&cstr, *s);
2774                             ++s;
2775                         }
2776                     }
2777                 }
2778                 cstr.size -= spc;
2779                 cstr_ccat(&cstr, '\"');
2780                 cstr_ccat(&cstr, '\0');
2781 #ifdef PP_DEBUG
2782                 printf("\nstringize: <%s>\n", (char *)cstr.data);
2783 #endif
2784                 /* add string */
2785                 cval.cstr = &cstr;
2786                 tok_str_add2(&str, TOK_PPSTR, &cval);
2787                 cstr_free(cval.cstr);
2788             } else {
2789         bad_stringy:
2790                 expect("macro parameter after '#'");
2791             }
2792         } else if (t >= TOK_IDENT) {
2793             s = sym_find2(args, t);
2794             if (s) {
2795                 int l0 = str.len;
2796                 st = s->d;
2797                 /* if '##' is present before or after, no arg substitution */
2798                 if (*macro_str == TOK_TWOSHARPS || t1 == TOK_TWOSHARPS) {
2799                     /* special case for var arg macros : ## eats the ','
2800                        if empty VA_ARGS variable. */
2801                     if (t1 == TOK_TWOSHARPS && t0 == ',' && gnu_ext && s->type.t) {
2802                         if (*st == 0) {
2803                             /* suppress ',' '##' */
2804                             str.len -= 2;
2805                         } else {
2806                             /* suppress '##' and add variable */
2807                             str.len--;
2808                             goto add_var;
2809                         }
2810                     } else {
2811                         for(;;) {
2812                             int t1;
2813                             TOK_GET(&t1, &st, &cval);
2814                             if (!t1)
2815                                 break;
2816                             tok_str_add2(&str, t1, &cval);
2817                         }
2818                     }
2819
2820                 } else {
2821             add_var:
2822                     /* NOTE: the stream cannot be read when macro
2823                        substituing an argument */
2824                     macro_subst(&str, nested_list, st, 0);
2825                 }
2826                 if (str.len == l0) /* exanded to empty string */
2827                     tok_str_add(&str, TOK_PLCHLDR);
2828             } else {
2829                 tok_str_add(&str, t);
2830             }
2831         } else {
2832             tok_str_add2(&str, t, &cval);
2833         }
2834         t0 = t1, t1 = t;
2835     }
2836     tok_str_add(&str, 0);
2837     return str.str;
2838 }
2839
2840 static char const ab_month_name[12][4] =
2841 {
2842     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2843     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2844 };
2845
2846 /* peek or read [ws_str == NULL] next token from function macro call,
2847    walking up macro levels up to the file if necessary */
2848 static int next_argstream(Sym **nested_list, int can_read_stream, TokenString *ws_str)
2849 {
2850     int t;
2851     const int *p;
2852     Sym *sa;
2853
2854     for (;;) {
2855         if (macro_ptr) {
2856             p = macro_ptr, t = *p;
2857             if (ws_str) {
2858                 while (is_space(t) || TOK_LINEFEED == t)
2859                     tok_str_add(ws_str, t), t = *++p;
2860             }
2861             if (t == 0 && can_read_stream) {
2862                 end_macro();
2863                 /* also, end of scope for nested defined symbol */
2864                 sa = *nested_list;
2865                 while (sa && sa->v == -1)
2866                     sa = sa->prev;
2867                 if (sa)
2868                     sa->v = -1;
2869                 continue;
2870             }
2871         } else {
2872             ch = handle_eob();
2873             if (ws_str) {
2874                 while (is_space(ch) || ch == '\n' || ch == '/') {
2875                     if (ch == '/') {
2876                         int c;
2877                         uint8_t *p = file->buf_ptr;
2878                         PEEKC(c, p);
2879                         if (c == '*') {
2880                             p = parse_comment(p);
2881                             file->buf_ptr = p - 1;
2882                         } else if (c == '/') {
2883                             p = parse_line_comment(p);
2884                             file->buf_ptr = p - 1;
2885                         } else
2886                             break;
2887                         ch = ' ';
2888                     }
2889                     tok_str_add(ws_str, ch);
2890                     cinp();
2891                 }
2892             }
2893             t = ch;
2894         }
2895
2896         if (ws_str)
2897             return t;
2898         next_nomacro_spc();
2899         return tok;
2900     }
2901 }
2902
2903 /* do macro substitution of current token with macro 's' and add
2904    result to (tok_str,tok_len). 'nested_list' is the list of all
2905    macros we got inside to avoid recursing. Return non zero if no
2906    substitution needs to be done */
2907 static int macro_subst_tok(
2908     TokenString *tok_str,
2909     Sym **nested_list,
2910     Sym *s,
2911     int can_read_stream)
2912 {
2913     Sym *args, *sa, *sa1;
2914     int parlevel, *mstr, t, t1, spc;
2915     TokenString str;
2916     char *cstrval;
2917     CValue cval;
2918     CString cstr;
2919     char buf[32];
2920
2921     /* if symbol is a macro, prepare substitution */
2922     /* special macros */
2923     if (tok == TOK___LINE__) {
2924         snprintf(buf, sizeof(buf), "%d", file->line_num);
2925         cstrval = buf;
2926         t1 = TOK_PPNUM;
2927         goto add_cstr1;
2928     } else if (tok == TOK___FILE__) {
2929         cstrval = file->filename;
2930         goto add_cstr;
2931     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2932         time_t ti;
2933         struct tm *tm;
2934
2935         time(&ti);
2936         tm = localtime(&ti);
2937         if (tok == TOK___DATE__) {
2938             snprintf(buf, sizeof(buf), "%s %2d %d",
2939                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2940         } else {
2941             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2942                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2943         }
2944         cstrval = buf;
2945     add_cstr:
2946         t1 = TOK_STR;
2947     add_cstr1:
2948         cstr_new(&cstr);
2949         cstr_cat(&cstr, cstrval);
2950         cstr_ccat(&cstr, '\0');
2951         cval.cstr = &cstr;
2952         tok_str_add2(tok_str, t1, &cval);
2953         cstr_free(&cstr);
2954     } else {
2955         int saved_parse_flags = parse_flags;
2956
2957         mstr = s->d;
2958         if (s->type.t == MACRO_FUNC) {
2959             /* whitespace between macro name and argument list */
2960             TokenString ws_str;
2961             tok_str_new(&ws_str);
2962
2963             spc = 0;
2964             parse_flags |= PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED
2965                 | PARSE_FLAG_ACCEPT_STRAYS;
2966
2967             /* get next token from argument stream */
2968             t = next_argstream(nested_list, can_read_stream, &ws_str);
2969             if (t != '(') {
2970                 /* not a macro substitution after all, restore the
2971                  * macro token plus all whitespace we've read.
2972                  * whitespace is intentionally not merged to preserve
2973                  * newlines. */
2974                 parse_flags = saved_parse_flags;
2975                 tok_str_add(tok_str, tok);
2976                 if (parse_flags & PARSE_FLAG_SPACES) {
2977                     int i;
2978                     for (i = 0; i < ws_str.len; i++)
2979                         tok_str_add(tok_str, ws_str.str[i]);
2980                 }
2981                 tok_str_free(ws_str.str);
2982                 return 0;
2983             } else {
2984                 tok_str_free(ws_str.str);
2985             }
2986             next_nomacro(); /* eat '(' */
2987
2988             /* argument macro */
2989             args = NULL;
2990             sa = s->next;
2991             /* NOTE: empty args are allowed, except if no args */
2992             for(;;) {
2993                 do {
2994                     next_argstream(nested_list, can_read_stream, NULL);
2995                 } while (is_space(tok) || TOK_LINEFEED == tok);
2996     empty_arg:
2997                 /* handle '()' case */
2998                 if (!args && !sa && tok == ')')
2999                     break;
3000                 if (!sa)
3001                     tcc_error("macro '%s' used with too many args",
3002                           get_tok_str(s->v, 0));
3003                 tok_str_new(&str);
3004                 parlevel = spc = 0;
3005                 /* NOTE: non zero sa->t indicates VA_ARGS */
3006                 while ((parlevel > 0 ||
3007                         (tok != ')' &&
3008                          (tok != ',' || sa->type.t)))) {
3009                     if (tok == TOK_EOF || tok == 0)
3010                         break;
3011                     if (tok == '(')
3012                         parlevel++;
3013                     else if (tok == ')')
3014                         parlevel--;
3015                     if (tok == TOK_LINEFEED)
3016                         tok = ' ';
3017                     if (!check_space(tok, &spc))
3018                         tok_str_add2(&str, tok, &tokc);
3019                     next_argstream(nested_list, can_read_stream, NULL);
3020                 }
3021                 if (parlevel)
3022                     expect(")");
3023                 str.len -= spc;
3024                 tok_str_add(&str, 0);
3025                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
3026                 sa1->d = str.str;
3027                 sa = sa->next;
3028                 if (tok == ')') {
3029                     /* special case for gcc var args: add an empty
3030                        var arg argument if it is omitted */
3031                     if (sa && sa->type.t && gnu_ext)
3032                         goto empty_arg;
3033                     break;
3034                 }
3035                 if (tok != ',')
3036                     expect(",");
3037             }
3038             if (sa) {
3039                 tcc_error("macro '%s' used with too few args",
3040                       get_tok_str(s->v, 0));
3041             }
3042
3043             parse_flags = saved_parse_flags;
3044
3045             /* now subst each arg */
3046             mstr = macro_arg_subst(nested_list, mstr, args);
3047             /* free memory */
3048             sa = args;
3049             while (sa) {
3050                 sa1 = sa->prev;
3051                 tok_str_free(sa->d);
3052                 sym_free(sa);
3053                 sa = sa1;
3054             }
3055         }
3056
3057         sym_push2(nested_list, s->v, 0, 0);
3058         parse_flags = saved_parse_flags;
3059         macro_subst(tok_str, nested_list, mstr, can_read_stream);
3060
3061         /* pop nested defined symbol */
3062         sa1 = *nested_list;
3063         *nested_list = sa1->prev;
3064         sym_free(sa1);
3065         if (mstr != s->d)
3066             tok_str_free(mstr);
3067     }
3068     return 0;
3069 }
3070
3071 int paste_tokens(int t1, CValue *v1, int t2, CValue *v2)
3072 {
3073     CString cstr;
3074     int n;
3075
3076     cstr_new(&cstr);
3077     if (t1 != TOK_PLCHLDR)
3078         cstr_cat(&cstr, get_tok_str(t1, v1));
3079     n = cstr.size;
3080     if (t2 != TOK_PLCHLDR)
3081         cstr_cat(&cstr, get_tok_str(t2, v2));
3082     cstr_ccat(&cstr, '\0');
3083
3084     tcc_open_bf(tcc_state, ":paste:", cstr.size);
3085     memcpy(file->buffer, cstr.data, cstr.size);
3086     for (;;) {
3087         next_nomacro1();
3088         if (0 == *file->buf_ptr)
3089             break;
3090         if (is_space(tok))
3091             continue;
3092         tcc_warning("pasting <%.*s> and <%s> does not give a valid preprocessing token",
3093             n, cstr.data, (char*)cstr.data + n);
3094         break;
3095     }
3096     tcc_close();
3097
3098     //printf("paste <%s>\n", (char*)cstr.data);
3099     cstr_free(&cstr);
3100     return 0;
3101 }
3102
3103 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
3104    return the resulting string (which must be freed). */
3105 static inline int *macro_twosharps(const int *ptr0)
3106 {
3107     int t;
3108     CValue cval;
3109     TokenString macro_str1;
3110     int start_of_nosubsts = -1;
3111     const int *ptr;
3112
3113     /* we search the first '##' */
3114     for (ptr = ptr0;;) {
3115         TOK_GET(&t, &ptr, &cval);
3116         if (t == TOK_TWOSHARPS)
3117             break;
3118         if (t == 0)
3119             return NULL;
3120     }
3121
3122     tok_str_new(&macro_str1);
3123
3124     //tok_print(" $$$", ptr0);
3125     for (ptr = ptr0;;) {
3126         TOK_GET(&t, &ptr, &cval);
3127         if (t == 0)
3128             break;
3129         if (t == TOK_TWOSHARPS)
3130             continue;
3131         while (*ptr == TOK_TWOSHARPS) {
3132             int t1; CValue cv1;
3133             /* given 'a##b', remove nosubsts preceding 'a' */
3134             if (start_of_nosubsts >= 0)
3135                 macro_str1.len = start_of_nosubsts;
3136             /* given 'a##b', remove nosubsts preceding 'b' */
3137             while ((t1 = *++ptr) == TOK_NOSUBST)
3138                 ;
3139             if (t1 && t1 != TOK_TWOSHARPS
3140                 && t1 != ':') /* 'a##:' don't build a new token */
3141             {
3142                 TOK_GET(&t1, &ptr, &cv1);
3143                 if (t != TOK_PLCHLDR || t1 != TOK_PLCHLDR) {
3144                     paste_tokens(t, &cval, t1, &cv1);
3145                     t = tok, cval = tokc;
3146                 }
3147             }
3148         }
3149         if (t == TOK_NOSUBST) {
3150             if (start_of_nosubsts < 0)
3151                 start_of_nosubsts = macro_str1.len;
3152         } else {
3153             start_of_nosubsts = -1;
3154         }
3155         tok_str_add2(&macro_str1, t, &cval);
3156     }
3157     tok_str_add(&macro_str1, 0);
3158     //tok_print(" ###", macro_str1.str);
3159     return macro_str1.str;
3160 }
3161
3162 /* do macro substitution of macro_str and add result to
3163    (tok_str,tok_len). 'nested_list' is the list of all macros we got
3164    inside to avoid recursing. */
3165 static void macro_subst(
3166     TokenString *tok_str,
3167     Sym **nested_list,
3168     const int *macro_str,
3169     int can_read_stream
3170     )
3171 {
3172     Sym *s;
3173     const int *ptr;
3174     int t, spc, nosubst;
3175     CValue cval;
3176     int *macro_str1 = NULL;
3177
3178     /* first scan for '##' operator handling */
3179     ptr = macro_str;
3180     spc = nosubst = 0;
3181
3182     /* first scan for '##' operator handling */
3183     if (can_read_stream) {
3184         macro_str1 = macro_twosharps(ptr);
3185         if (macro_str1)
3186             ptr = macro_str1;
3187     }
3188
3189     while (1) {
3190         TOK_GET(&t, &ptr, &cval);
3191         if (t == 0)
3192             break;
3193
3194         if (t >= TOK_IDENT && 0 == nosubst) {
3195             s = define_find(t);
3196             if (s == NULL)
3197                 goto no_subst;
3198
3199             /* if nested substitution, do nothing */
3200             if (sym_find2(*nested_list, t)) {
3201                 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
3202                 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3203                 goto no_subst;
3204             }
3205
3206             {
3207                 TokenString str;
3208                 str.str = (int*)ptr;
3209                 begin_macro(&str, 2);
3210
3211                 tok = t;
3212                 macro_subst_tok(tok_str, nested_list, s, can_read_stream);
3213
3214                 if (str.alloc == 3) {
3215                     /* already finished by reading function macro arguments */
3216                     break;
3217                 }
3218
3219                 ptr = macro_ptr;
3220                 end_macro ();
3221             }
3222
3223             spc = (tok_str->len &&
3224                    is_space(tok_last(tok_str->str,
3225                                      tok_str->str + tok_str->len)));
3226
3227         } else {
3228
3229             if (t == '\\' && !(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
3230                 tcc_error("stray '\\' in program");
3231
3232 no_subst:
3233             if (!check_space(t, &spc))
3234                 tok_str_add2(tok_str, t, &cval);
3235             nosubst = 0;
3236             if (t == TOK_NOSUBST)
3237                 nosubst = 1;
3238         }
3239     }
3240     if (macro_str1)
3241         tok_str_free(macro_str1);
3242
3243 }
3244
3245 /* return next token with macro substitution */
3246 ST_FUNC void next(void)
3247 {
3248  redo:
3249     if (parse_flags & PARSE_FLAG_SPACES)
3250         next_nomacro_spc();
3251     else
3252         next_nomacro();
3253
3254     if (macro_ptr) {
3255         if (tok == TOK_NOSUBST || tok == TOK_PLCHLDR) {
3256         /* discard preprocessor markers */
3257             goto redo;
3258         } else if (tok == 0) {
3259             /* end of macro or unget token string */
3260             end_macro();
3261             goto redo;
3262         }
3263     } else if (tok >= TOK_IDENT && (parse_flags & PARSE_FLAG_PREPROCESS)) {
3264         Sym *s;
3265         /* if reading from file, try to substitute macros */
3266         s = define_find(tok);
3267         if (s) {
3268             static TokenString str; /* using static string for speed */
3269             Sym *nested_list = NULL;
3270             tok_str_new(&str);
3271             nested_list = NULL;
3272             macro_subst_tok(&str, &nested_list, s, 1);
3273             tok_str_add(&str, 0);
3274             begin_macro(&str, 0);
3275             goto redo;
3276         }
3277     }
3278     /* convert preprocessor tokens into C tokens */
3279     if (tok == TOK_PPNUM) {
3280         if  (parse_flags & PARSE_FLAG_TOK_NUM)
3281             parse_number((char *)tokc.cstr->data);
3282     } else if (tok == TOK_PPSTR) {
3283         if (parse_flags & PARSE_FLAG_TOK_STR)
3284             parse_string((char *)tokc.cstr->data, tokc.cstr->size - 1);
3285     }
3286 }
3287
3288 /* push back current token and set current token to 'last_tok'. Only
3289    identifier case handled for labels. */
3290 ST_INLN void unget_tok(int last_tok)
3291 {
3292     TokenString *str = tcc_malloc(sizeof *str);
3293     tok_str_new(str);
3294     tok_str_add2(str, tok, &tokc);
3295     tok_str_add(str, 0);
3296     begin_macro(str, 1);
3297     tok = last_tok;
3298 }
3299
3300 /* better than nothing, but needs extension to handle '-E' option
3301    correctly too */
3302 ST_FUNC void preprocess_init(TCCState *s1)
3303 {
3304     s1->include_stack_ptr = s1->include_stack;
3305     /* XXX: move that before to avoid having to initialize
3306        file->ifdef_stack_ptr ? */
3307     s1->ifdef_stack_ptr = s1->ifdef_stack;
3308     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3309
3310     pvtop = vtop = vstack - 1;
3311     s1->pack_stack[0] = 0;
3312     s1->pack_stack_ptr = s1->pack_stack;
3313
3314     isidnum_table['$' - CH_EOF] =
3315         tcc_state->dollars_in_identifiers ? IS_ID : 0;
3316 }
3317
3318 ST_FUNC void preprocess_new(void)
3319 {
3320     int i, c;
3321     const char *p, *r;
3322
3323     /* init isid table */
3324     for(i = CH_EOF; i<256; i++)
3325         isidnum_table[i - CH_EOF]
3326             = is_space(i) ? IS_SPC
3327             : isid(i) ? IS_ID
3328             : isnum(i) ? IS_NUM
3329             : 0;
3330
3331     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3332
3333     tok_ident = TOK_IDENT;
3334     p = tcc_keywords;
3335     while (*p) {
3336         r = p;
3337         for(;;) {
3338             c = *r++;
3339             if (c == '\0')
3340                 break;
3341         }
3342         tok_alloc(p, r - p - 1);
3343         p = r;
3344     }
3345 }
3346
3347 ST_FUNC void preprocess_delete(void)
3348 {
3349     int i, n;
3350
3351     /* free -D and compiler defines */
3352     free_defines(NULL);
3353
3354     /* cleanup from error/setjmp */
3355     while (macro_stack)
3356         end_macro();
3357     macro_ptr = NULL;
3358
3359     /* free tokens */
3360     n = tok_ident - TOK_IDENT;
3361     for(i = 0; i < n; i++)
3362         tcc_free(table_ident[i]);
3363     tcc_free(table_ident);
3364     table_ident = NULL;
3365 }
3366
3367 /* Preprocess the current file */
3368 ST_FUNC int tcc_preprocess(TCCState *s1)
3369 {
3370     BufferedFile **iptr;
3371     int token_seen, spcs, level;
3372
3373     preprocess_init(s1);
3374     ch = file->buf_ptr[0];
3375     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3376     parse_flags = PARSE_FLAG_PREPROCESS
3377                 | (parse_flags & PARSE_FLAG_ASM_FILE)
3378                 | PARSE_FLAG_LINEFEED
3379                 | PARSE_FLAG_SPACES
3380                 | PARSE_FLAG_ACCEPT_STRAYS
3381                 ;
3382
3383 #ifdef PP_BENCH
3384     do next(); while (tok != TOK_EOF); return 0;
3385 #endif
3386
3387     token_seen = spcs = 0;
3388     pp_line(s1, file, 0);
3389
3390     for (;;) {
3391         iptr = s1->include_stack_ptr;
3392         next();
3393         if (tok == TOK_EOF)
3394             break;
3395         level = s1->include_stack_ptr - iptr;
3396         if (level) {
3397             if (level > 0)
3398                 pp_line(s1, *iptr, 0);
3399             pp_line(s1, file, level);
3400         }
3401
3402         if (0 == token_seen) {
3403             if (tok == ' ') {
3404                 ++spcs;
3405                 continue;
3406             }
3407             if (tok == TOK_LINEFEED) {
3408                 spcs = 0;
3409                 continue;
3410             }
3411             pp_line(s1, file, 0);
3412             while (spcs)
3413                 fputs(" ", s1->ppfp), --spcs;
3414             token_seen = 1;
3415
3416         } else if (tok == TOK_LINEFEED) {
3417             ++file->line_ref;
3418             token_seen = 0;
3419         }
3420
3421         fputs(get_tok_str(tok, &tokc), s1->ppfp);
3422     }
3423
3424     return 0;
3425 }