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  61 /* outdated -- gr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, "%u", cv->ui);
 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->ull);
 303 #else
 304         sprintf(p, "%llu", cv->ull);
 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             cstr = (CString *)(str + len);
 931             cstr->data = NULL;
 932             cstr->size = cv->cstr->size;
 933             cstr->data_allocated = NULL;
 934             cstr->size_allocated = cstr->size;
 935             memcpy((char *)cstr + sizeof(CString),
 936                    cv->cstr->data, cstr->size);
 937             len += nb_words;
 938         }
 939         break;
 940     case TOK_CDOUBLE:
 941     case TOK_CLLONG:
 942     case TOK_CULLONG:
 943 #if LDOUBLE_SIZE == 8
 944     case TOK_CLDOUBLE:
 945 #endif
 946         str[len++] = cv->tab[0];
 947         str[len++] = cv->tab[1];
 948         break;
 949 #if LDOUBLE_SIZE == 12
 950     case TOK_CLDOUBLE:
 951         str[len++] = cv->tab[0];
 952         str[len++] = cv->tab[1];
 953         str[len++] = cv->tab[2];
 954 #elif LDOUBLE_SIZE == 16
 955     case TOK_CLDOUBLE:
 956         str[len++] = cv->tab[0];
 957         str[len++] = cv->tab[1];
 958         str[len++] = cv->tab[2];
 959         str[len++] = cv->tab[3];
 960 #elif LDOUBLE_SIZE != 8
 961 #error add long double size support
 962 #endif
 963         break;
 964     default:
 965         break;
 966     }
 967     s->len = len;
 968 }
 969
 970 /* add the current parse token in token string 's' */
 971 ST_FUNC void tok_str_add_tok(TokenString *s)
 972 {
 973     CValue cval;
 974
 975     /* save line number info */
 976     if (file->line_num != s->last_line_num) {
 977         s->last_line_num = file->line_num;
 978         cval.i = s->last_line_num;
 979         tok_str_add2(s, TOK_LINENUM, &cval);
 980     }
 981     tok_str_add2(s, tok, &tokc);
 982 }
 983
 984 /* get a token from an integer array and increment pointer
 985    accordingly. we code it as a macro to avoid pointer aliasing. */
 986 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
 987 {
 988     const int *p = *pp;
 989     int n, *tab;
 990
 991     tab = cv->tab;
 992     switch(*t = *p++) {
 993     case TOK_CINT:
 994     case TOK_CUINT:
 995     case TOK_CCHAR:
 996     case TOK_LCHAR:
 997     case TOK_CFLOAT:
 998     case TOK_LINENUM:
 999         tab[0] = *p++;
1000         break;
1001     case TOK_STR:
1002     case TOK_LSTR:
1003     case TOK_PPNUM:
1004     case TOK_PPSTR:
1005         cv->cstr = (CString *)p;
1006         cv->cstr->data = (char *)p + sizeof(CString);
1007         p += (sizeof(CString) + cv->cstr->size + 3) >> 2;
1008         break;
1009     case TOK_CDOUBLE:
1010     case TOK_CLLONG:
1011     case TOK_CULLONG:
1012         n = 2;
1013         goto copy;
1014     case TOK_CLDOUBLE:
1015 #if LDOUBLE_SIZE == 16
1016         n = 4;
1017 #elif LDOUBLE_SIZE == 12
1018         n = 3;
1019 #elif LDOUBLE_SIZE == 8
1020         n = 2;
1021 #else
1022 # error add long double size support
1023 #endif
1024     copy:
1025         do
1026             *tab++ = *p++;
1027         while (--n);
1028         break;
1029     default:
1030         break;
1031     }
1032     *pp = p;
1033 }
1034
1035 static int macro_is_equal(const int *a, const int *b)
1036 {
1037     char buf[STRING_MAX_SIZE + 1];
1038     CValue cv;
1039     int t;
1040     while (*a && *b) {
1041         TOK_GET(&t, &a, &cv);
1042         pstrcpy(buf, sizeof buf, get_tok_str(t, &cv));
1043         TOK_GET(&t, &b, &cv);
1044         if (strcmp(buf, get_tok_str(t, &cv)))
1045             return 0;
1046     }
1047     return !(*a || *b);
1048 }
1049
1050 static void pp_line(TCCState *s1, BufferedFile *f, int level)
1051 {
1052     int d = f->line_num - f->line_ref;
1053     if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_NONE
1054         || (level == 0 && f->line_ref && d < 8))
1055     {
1056         while (d > 0)
1057             fputs("\n", s1->ppfp), --d;
1058     }
1059     else
1060     if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_STD) {
1061         fprintf(s1->ppfp, "#line %d \"%s\"\n", f->line_num, f->filename);
1062     }
1063     else {
1064         fprintf(s1->ppfp, "# %d \"%s\"%s\n", f->line_num, f->filename,
1065             level > 0 ? " 1" : level < 0 ? " 2" : "");
1066     }
1067     f->line_ref = f->line_num;
1068 }
1069
1070 static void tok_print(const char *msg, const int *str)
1071 {
1072     FILE *pr = tcc_state->ppfp;
1073     int t;
1074     CValue cval;
1075
1076     fprintf(pr, "%s ", msg);
1077     while (str) {
1078         TOK_GET(&t, &str, &cval);
1079         if (!t)
1080             break;
1081         fprintf(pr,"%s", get_tok_str(t, &cval));
1082     }
1083     fprintf(pr, "\n");
1084 }
1085
1086 static int define_print_prepared(Sym *s)
1087 {
1088     if (!s || !tcc_state->ppfp || tcc_state->dflag == 0)
1089         return 0;
1090
1091     if (s->v < TOK_IDENT || s->v >= tok_ident)
1092         return 0;
1093
1094     if (file) {
1095         file->line_num--;
1096         pp_line(tcc_state, file, 0);
1097         file->line_ref = ++file->line_num;
1098     }
1099     return 1;
1100 }
1101
1102 static void define_print(int v)
1103 {
1104     FILE *pr = tcc_state->ppfp;
1105     Sym *s, *a;
1106
1107     s = define_find(v);
1108     if (define_print_prepared(s) == 0)
1109         return;
1110
1111     fprintf(pr, "// #define %s", get_tok_str(v, NULL));
1112     if (s->type.t == MACRO_FUNC) {
1113         a = s->next;
1114         fprintf(pr,"(");
1115         if (a)
1116             for (;;) {
1117                 fprintf(pr,"%s", get_tok_str(a->v & ~SYM_FIELD, NULL));
1118                 if (!(a = a->next))
1119                     break;
1120                 fprintf(pr,",");
1121             }
1122         fprintf(pr,")");
1123     }
1124     tok_print("", s->d);
1125 }
1126
1127 static void undef_print(int v)
1128 {
1129     FILE *pr = tcc_state->ppfp;
1130     Sym *s;
1131
1132     s = define_find(v);
1133     if (define_print_prepared(s) == 0)
1134         return;
1135
1136     fprintf(pr, "// #undef %s\n", get_tok_str(s->v, NULL));
1137 }
1138
1139 ST_FUNC void print_defines(void)
1140 {
1141     Sym *top = define_stack;
1142     while (top) {
1143         define_print(top->v);
1144         top = top->prev;
1145     }
1146 }
1147
1148 /* defines handling */
1149 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1150 {
1151     Sym *s;
1152
1153     s = define_find(v);
1154     if (s && !macro_is_equal(s->d, str))
1155         tcc_warning("%s redefined", get_tok_str(v, NULL));
1156
1157     s = sym_push2(&define_stack, v, macro_type, 0);
1158     s->d = str;
1159     s->next = first_arg;
1160     table_ident[v - TOK_IDENT]->sym_define = s;
1161 }
1162
1163 /* undefined a define symbol. Its name is just set to zero */
1164 ST_FUNC void define_undef(Sym *s)
1165 {
1166     int v = s->v;
1167     undef_print(v);
1168     if (v >= TOK_IDENT && v < tok_ident)
1169         table_ident[v - TOK_IDENT]->sym_define = NULL;
1170 }
1171
1172 ST_INLN Sym *define_find(int v)
1173 {
1174     v -= TOK_IDENT;
1175     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1176         return NULL;
1177     return table_ident[v]->sym_define;
1178 }
1179
1180 /* free define stack until top reaches 'b' */
1181 ST_FUNC void free_defines(Sym *b)
1182 {
1183     Sym *top, *top1;
1184     int v;
1185
1186     top = define_stack;
1187     while (top != b) {
1188         top1 = top->prev;
1189         /* do not free args or predefined defines */
1190         if (top->d)
1191             tok_str_free(top->d);
1192         v = top->v;
1193         if (v >= TOK_IDENT && v < tok_ident)
1194             table_ident[v - TOK_IDENT]->sym_define = NULL;
1195         sym_free(top);
1196         top = top1;
1197     }
1198     define_stack = b;
1199 }
1200
1201 /* label lookup */
1202 ST_FUNC Sym *label_find(int v)
1203 {
1204     v -= TOK_IDENT;
1205     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1206         return NULL;
1207     return table_ident[v]->sym_label;
1208 }
1209
1210 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1211 {
1212     Sym *s, **ps;
1213     s = sym_push2(ptop, v, 0, 0);
1214     s->r = flags;
1215     ps = &table_ident[v - TOK_IDENT]->sym_label;
1216     if (ptop == &global_label_stack) {
1217         /* modify the top most local identifier, so that
1218            sym_identifier will point to 's' when popped */
1219         while (*ps != NULL)
1220             ps = &(*ps)->prev_tok;
1221     }
1222     s->prev_tok = *ps;
1223     *ps = s;
1224     return s;
1225 }
1226
1227 /* pop labels until element last is reached. Look if any labels are
1228    undefined. Define symbols if '&&label' was used. */
1229 ST_FUNC void label_pop(Sym **ptop, Sym *slast)
1230 {
1231     Sym *s, *s1;
1232     for(s = *ptop; s != slast; s = s1) {
1233         s1 = s->prev;
1234         if (s->r == LABEL_DECLARED) {
1235             tcc_warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1236         } else if (s->r == LABEL_FORWARD) {
1237                 tcc_error("label '%s' used but not defined",
1238                       get_tok_str(s->v, NULL));
1239         } else {
1240             if (s->c) {
1241                 /* define corresponding symbol. A size of
1242                    1 is put. */
1243                 put_extern_sym(s, cur_text_section, s->jnext, 1);
1244             }
1245         }
1246         /* remove label */
1247         table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1248         sym_free(s);
1249     }
1250     *ptop = slast;
1251 }
1252
1253 /* eval an expression for #if/#elif */
1254 static int expr_preprocess(void)
1255 {
1256     int c, t;
1257     TokenString str;
1258
1259     tok_str_new(&str);
1260     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1261         next(); /* do macro subst */
1262         if (tok == TOK_DEFINED) {
1263             next_nomacro();
1264             t = tok;
1265             if (t == '(')
1266                 next_nomacro();
1267             c = define_find(tok) != 0;
1268             if (t == '(')
1269                 next_nomacro();
1270             tok = TOK_CINT;
1271             tokc.i = c;
1272         } else if (tok >= TOK_IDENT) {
1273             /* if undefined macro */
1274             tok = TOK_CINT;
1275             tokc.i = 0;
1276         }
1277         tok_str_add_tok(&str);
1278     }
1279     tok_str_add(&str, -1); /* simulate end of file */
1280     tok_str_add(&str, 0);
1281     /* now evaluate C constant expression */
1282     begin_macro(&str, 0);
1283     next();
1284     c = expr_const();
1285     end_macro();
1286     return c != 0;
1287 }
1288
1289
1290 /* parse after #define */
1291 ST_FUNC void parse_define(void)
1292 {
1293     Sym *s, *first, **ps;
1294     int v, t, varg, is_vaargs, spc;
1295     int saved_parse_flags = parse_flags;
1296     TokenString str;
1297
1298     v = tok;
1299     if (v < TOK_IDENT)
1300         tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1301     /* XXX: should check if same macro (ANSI) */
1302     first = NULL;
1303     t = MACRO_OBJ;
1304     /* '(' must be just after macro definition for MACRO_FUNC */
1305     parse_flags |= PARSE_FLAG_SPACES;
1306     next_nomacro_spc();
1307     if (tok == '(') {
1308         next_nomacro();
1309         ps = &first;
1310         if (tok != ')') for (;;) {
1311             varg = tok;
1312             next_nomacro();
1313             is_vaargs = 0;
1314             if (varg == TOK_DOTS) {
1315                 varg = TOK___VA_ARGS__;
1316                 is_vaargs = 1;
1317             } else if (tok == TOK_DOTS && gnu_ext) {
1318                 is_vaargs = 1;
1319                 next_nomacro();
1320             }
1321             if (varg < TOK_IDENT)
1322         bad_list:
1323                 tcc_error("bad macro parameter list");
1324             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1325             *ps = s;
1326             ps = &s->next;
1327             if (tok == ')')
1328                 break;
1329             if (tok != ',' || is_vaargs)
1330                 goto bad_list;
1331             next_nomacro();
1332         }
1333         next_nomacro_spc();
1334         t = MACRO_FUNC;
1335     }
1336     tok_str_new(&str);
1337     spc = 2;
1338     parse_flags |= PARSE_FLAG_ACCEPT_STRAYS | PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED;
1339     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1340         /* remove spaces around ## and after '#' */
1341         if (TOK_TWOSHARPS == tok) {
1342             if (2 == spc)
1343                 goto bad_twosharp;
1344             if (1 == spc)
1345                 --str.len;
1346             spc = 3;
1347         } else if ('#' == tok) {
1348             spc = 4;
1349         } else if (check_space(tok, &spc)) {
1350             goto skip;
1351         }
1352         tok_str_add2(&str, tok, &tokc);
1353     skip:
1354         next_nomacro_spc();
1355     }
1356
1357     parse_flags = saved_parse_flags;
1358     if (spc == 1)
1359         --str.len; /* remove trailing space */
1360     tok_str_add(&str, 0);
1361     if (3 == spc)
1362 bad_twosharp:
1363         tcc_error("'##' cannot appear at either end of macro");
1364     define_push(v, t, str.str, first);
1365     define_print(v);
1366 }
1367
1368 static inline int hash_cached_include(const char *filename)
1369 {
1370     const unsigned char *s;
1371     unsigned int h;
1372
1373     h = TOK_HASH_INIT;
1374     s = (unsigned char *) filename;
1375     while (*s) {
1376         h = TOK_HASH_FUNC(h, *s);
1377         s++;
1378     }
1379     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1380     return h;
1381 }
1382
1383 static CachedInclude *search_cached_include(TCCState *s1, const char *filename)
1384 {
1385     CachedInclude *e;
1386     int i, h;
1387     h = hash_cached_include(filename);
1388     i = s1->cached_includes_hash[h];
1389     for(;;) {
1390         if (i == 0)
1391             break;
1392         e = s1->cached_includes[i - 1];
1393         if (0 == PATHCMP(e->filename, filename))
1394             return e;
1395         i = e->hash_next;
1396     }
1397     return NULL;
1398 }
1399
1400 static inline void add_cached_include(TCCState *s1, const char *filename, int ifndef_macro)
1401 {
1402     CachedInclude *e;
1403     int h;
1404
1405     if (search_cached_include(s1, filename))
1406         return;
1407 #ifdef INC_DEBUG
1408     printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1409 #endif
1410     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1411     strcpy(e->filename, filename);
1412     e->ifndef_macro = ifndef_macro;
1413     dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1414     /* add in hash table */
1415     h = hash_cached_include(filename);
1416     e->hash_next = s1->cached_includes_hash[h];
1417     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1418 }
1419
1420 static void pragma_parse(TCCState *s1)
1421 {
1422     next_nomacro();
1423     if (tok == TOK_push_macro || tok == TOK_pop_macro) {
1424         int t = tok, v;
1425         Sym *s;
1426
1427         if (next(), tok != '(')
1428             goto pragma_err;
1429         if (next(), tok != TOK_STR)
1430             goto pragma_err;
1431         v = tok_alloc(tokc.cstr->data, tokc.cstr->size - 1)->tok;
1432         if (next(), tok != ')')
1433             goto pragma_err;
1434         if (t == TOK_push_macro) {
1435             while (NULL == (s = define_find(v)))
1436                 define_push(v, 0, NULL, NULL);
1437             s->type.ref = s; /* set push boundary */
1438         } else {
1439             for (s = define_stack; s; s = s->prev)
1440                 if (s->v == v && s->type.ref == s) {
1441                     s->type.ref = NULL;
1442                     break;
1443                 }
1444         }
1445         if (s)
1446             table_ident[v - TOK_IDENT]->sym_define = s->d ? s : NULL;
1447         else
1448             tcc_warning("unbalanced #pragma pop_macro");
1449
1450     } else if (tok == TOK_once) {
1451         add_cached_include(s1, file->filename, TOK_once);
1452
1453     } else if (s1->ppfp) {
1454         /* tcc -E: keep pragmas below unchanged */
1455         unget_tok(' ');
1456         unget_tok(TOK_PRAGMA);
1457         unget_tok('#');
1458         unget_tok(TOK_LINEFEED);
1459
1460     } else if (tok == TOK_pack) {
1461         /* This may be:
1462            #pragma pack(1) // set
1463            #pragma pack() // reset to default
1464            #pragma pack(push,1) // push & set
1465            #pragma pack(pop) // restore previous */
1466         next();
1467         skip('(');
1468         if (tok == TOK_ASM_pop) {
1469             next();
1470             if (s1->pack_stack_ptr <= s1->pack_stack) {
1471             stk_error:
1472                 tcc_error("out of pack stack");
1473             }
1474             s1->pack_stack_ptr--;
1475         } else {
1476             int val = 0;
1477             if (tok != ')') {
1478                 if (tok == TOK_ASM_push) {
1479                     next();
1480                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1481                         goto stk_error;
1482                     s1->pack_stack_ptr++;
1483                     skip(',');
1484                 }
1485                 if (tok != TOK_CINT)
1486                     goto pragma_err;
1487                 val = tokc.i;
1488                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1489                     goto pragma_err;
1490                 next();
1491             }
1492             *s1->pack_stack_ptr = val;
1493         }
1494         if (tok != ')')
1495             goto pragma_err;
1496
1497     } else if (tok == TOK_comment) {
1498         char *file;
1499         next();
1500         skip('(');
1501         if (tok != TOK_lib)
1502             goto pragma_warn;
1503         next();
1504         skip(',');
1505         if (tok != TOK_STR)
1506             goto pragma_err;
1507         file = tcc_strdup((char *)tokc.cstr->data);
1508         dynarray_add((void ***)&s1->pragma_libs, &s1->nb_pragma_libs, file);
1509         next();
1510         if (tok != ')')
1511             goto pragma_err;
1512     } else {
1513 pragma_warn:
1514         if (s1->warn_unsupported)
1515             tcc_warning("#pragma %s is ignored", get_tok_str(tok, &tokc));
1516     }
1517     return;
1518
1519 pragma_err:
1520     tcc_error("malformed #pragma directive");
1521     return;
1522 }
1523
1524 /* is_bof is true if first non space token at beginning of file */
1525 ST_FUNC void preprocess(int is_bof)
1526 {
1527     TCCState *s1 = tcc_state;
1528     int i, c, n, saved_parse_flags;
1529     char buf[1024], *q;
1530     Sym *s;
1531     char *p3 = 0;
1532
1533     saved_parse_flags = parse_flags;
1534     parse_flags = PARSE_FLAG_PREPROCESS
1535         | PARSE_FLAG_TOK_NUM
1536         | PARSE_FLAG_TOK_STR
1537         | PARSE_FLAG_LINEFEED
1538         | (parse_flags & PARSE_FLAG_ASM_FILE)
1539         ;
1540
1541     next_nomacro();
1542  redo:
1543     switch(tok) {
1544     case TOK_DEFINE:
1545         next_nomacro();
1546         parse_define();
1547         break;
1548     case TOK_UNDEF:
1549         next_nomacro();
1550         s = define_find(tok);
1551         /* undefine symbol by putting an invalid name */
1552         if (s)
1553             define_undef(s);
1554         break;
1555     case TOK_INCLUDE:
1556     case TOK_INCLUDE_NEXT:
1557         ch = file->buf_ptr[0];
1558         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1559         skip_spaces();
1560         if (ch == '<') {
1561             c = '>';
1562             goto read_name;
1563         } else if (ch == '\"') {
1564             c = ch;
1565         read_name:
1566             inp();
1567             q = buf;
1568             while (ch != c && ch != '\n' && ch != CH_EOF) {
1569                 if ((q - buf) < sizeof(buf) - 1)
1570                     *q++ = ch;
1571                 if (ch == '\\') {
1572                     if (handle_stray_noerror() == 0)
1573                         --q;
1574                 } else
1575                     inp();
1576             }
1577             *q = '\0';
1578             minp();
1579 #if 0
1580             /* eat all spaces and comments after include */
1581             /* XXX: slightly incorrect */
1582             while (ch1 != '\n' && ch1 != CH_EOF)
1583                 inp();
1584 #endif
1585         } else {
1586             /* computed #include : either we have only strings or
1587                we have anything enclosed in '<>' */
1588             next();
1589             buf[0] = '\0';
1590             if (tok == TOK_STR) {
1591                 while (tok != TOK_LINEFEED) {
1592                     if (tok != TOK_STR) {
1593                     include_syntax:
1594                         tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1595                     }
1596                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1597                     next();
1598                 }
1599                 c = '\"';
1600             } else {
1601                 int len;
1602                 while (tok != TOK_LINEFEED) {
1603                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1604                     next();
1605                 }
1606                 len = strlen(buf);
1607                 /* check syntax and remove '<>' */
1608                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1609                     goto include_syntax;
1610                 memmove(buf, buf + 1, len - 2);
1611                 buf[len - 2] = '\0';
1612                 c = '>';
1613             }
1614         }
1615
1616         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1617             tcc_error("#include recursion too deep");
1618         /* store current file in stack, but increment stack later below */
1619         *s1->include_stack_ptr = file;
1620
1621         #ifdef INC_DEBUG
1622             if (tok == TOK_INCLUDE_NEXT)
1623                 printf("%s (1) include_next file %s\n", file->filename, buf);
1624         #endif
1625
1626         n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1627         for (i = -2; i < n; ++i) {
1628             char buf1[sizeof file->filename];
1629             CachedInclude *e;
1630             BufferedFile **f;
1631             const char *path;
1632
1633             if (i == -2) {
1634                 /* check absolute include path */
1635                 if (!IS_ABSPATH(buf))
1636                     continue;
1637                 buf1[0] = 0;
1638                 i = n; /* force end loop */
1639
1640             } else if (i == -1) {
1641                 /* search in current dir if "header.h" */
1642                 if (c != '\"')
1643                     continue;
1644                 path = file->filename;
1645                 pstrncpy(buf1, path, tcc_basename(path) - path);
1646
1647             } else {
1648                 /* search in all the include paths */
1649                 if (i < s1->nb_include_paths)
1650                     path = s1->include_paths[i];
1651                 else
1652                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1653                 pstrcpy(buf1, sizeof(buf1), path);
1654                 pstrcat(buf1, sizeof(buf1), "/");
1655             }
1656
1657             #ifdef INC_DEBUG
1658             if (tok == TOK_INCLUDE_NEXT)
1659                 printf("%s (2) include_next path <%s> name <%s>\n", file->filename, buf1, buf);
1660             #endif
1661
1662             pstrcat(buf1, sizeof(buf1), buf);
1663
1664             if (tok == TOK_INCLUDE_NEXT) {
1665                 char *p1 = buf1;
1666                 if ((p1[0] == '.') && IS_DIRSEP(p1[1])) p1 += 2;
1667
1668                 if (p3) {
1669                     if (0 == PATHCMP(p1, p3)) {
1670                         #ifdef INC_DEBUG
1671                         printf("%s (3) include_next skipping <%s>\n", file->filename, p1);
1672                         #endif
1673                         goto include_trynext;
1674                     }
1675                 }
1676                 else {
1677                     for (f = s1->include_stack_ptr; f >= s1->include_stack; --f) {
1678                         char *p2 = (*f)->filename;
1679                         if ((p2[0] == '.') && IS_DIRSEP(p2[1])) p2 += 2;
1680                         if (0 == PATHCMP(p2, p1)) {
1681                             p3 = p2;
1682                             break;
1683                         }
1684                     }
1685                     #ifdef INC_DEBUG
1686                         printf("%s: (4) include_next skipping <%s> (p3=%p)\n", file->filename, buf1, p3);
1687                     #endif
1688                     goto include_trynext;
1689                 }
1690                 #ifdef INC_DEBUG
1691                 printf("%s (5) include_next considering <%s>\n", file->filename, buf1);
1692                 #endif
1693             }
1694
1695             e = search_cached_include(s1, buf1);
1696             if (e && (define_find(e->ifndef_macro) || e->ifndef_macro == TOK_once)) {
1697                 /* no need to parse the include because the 'ifndef macro'
1698                    is defined */
1699
1700                 #ifdef INC_DEBUG
1701                 printf("%s: skipping cached <%s>\n", file->filename, buf1);
1702                 #endif
1703                 goto include_done;
1704             }
1705
1706             if (tcc_open(s1, buf1) < 0) {
1707                 #ifdef INC_DEBUG
1708                 printf("%s: include open failed <%s>\n", file->filename, buf1);
1709                 #endif
1710 include_trynext:
1711                 continue;
1712             }
1713
1714             #ifdef INC_DEBUG
1715             printf("%s: including <%s>\n", file->prev->filename, file->filename);
1716             #endif
1717
1718             /* update target deps */
1719             dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1720                     tcc_strdup(buf1));
1721             /* push current file in stack */
1722             ++s1->include_stack_ptr;
1723             /* add include file debug info */
1724             if (s1->do_debug)
1725                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1726             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1727             ch = file->buf_ptr[0];
1728             goto the_end;
1729         }
1730         tcc_error("include file '%s' not found", buf);
1731 include_done:
1732         break;
1733     case TOK_IFNDEF:
1734         c = 1;
1735         goto do_ifdef;
1736     case TOK_IF:
1737         c = expr_preprocess();
1738         goto do_if;
1739     case TOK_IFDEF:
1740         c = 0;
1741     do_ifdef:
1742         next_nomacro();
1743         if (tok < TOK_IDENT)
1744             tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1745         if (is_bof) {
1746             if (c) {
1747                 #ifdef INC_DEBUG
1748                     printf("#ifndef %s\n", get_tok_str(tok, NULL));
1749                 #endif
1750                 file->ifndef_macro = tok;
1751             }
1752         }
1753         c = (define_find(tok) != 0) ^ c;
1754     do_if:
1755         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1756             tcc_error("memory full (ifdef)");
1757         *s1->ifdef_stack_ptr++ = c;
1758         goto test_skip;
1759     case TOK_ELSE:
1760         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1761             tcc_error("#else without matching #if");
1762         if (s1->ifdef_stack_ptr[-1] & 2)
1763             tcc_error("#else after #else");
1764         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1765         goto test_else;
1766     case TOK_ELIF:
1767         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1768             tcc_error("#elif without matching #if");
1769         c = s1->ifdef_stack_ptr[-1];
1770         if (c > 1)
1771             tcc_error("#elif after #else");
1772         /* last #if/#elif expression was true: we skip */
1773         if (c == 1)
1774             goto skip;
1775         c = expr_preprocess();
1776         s1->ifdef_stack_ptr[-1] = c;
1777     test_else:
1778         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1779             file->ifndef_macro = 0;
1780     test_skip:
1781         if (!(c & 1)) {
1782         skip:
1783             preprocess_skip();
1784             is_bof = 0;
1785             goto redo;
1786         }
1787         break;
1788     case TOK_ENDIF:
1789         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1790             tcc_error("#endif without matching #if");
1791         s1->ifdef_stack_ptr--;
1792         /* '#ifndef macro' was at the start of file. Now we check if
1793            an '#endif' is exactly at the end of file */
1794         if (file->ifndef_macro &&
1795             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1796             file->ifndef_macro_saved = file->ifndef_macro;
1797             /* need to set to zero to avoid false matches if another
1798                #ifndef at middle of file */
1799             file->ifndef_macro = 0;
1800             while (tok != TOK_LINEFEED)
1801                 next_nomacro();
1802             tok_flags |= TOK_FLAG_ENDIF;
1803             goto the_end;
1804         }
1805         break;
1806     case TOK_PPNUM:
1807         n = strtoul((char*)tokc.cstr->data, &q, 10);
1808         goto _line_num;
1809     case TOK_LINE:
1810         next();
1811         if (tok != TOK_CINT)
1812 _line_err:
1813             tcc_error("wrong #line format");
1814         n = tokc.i;
1815 _line_num:
1816         next();
1817         if (tok != TOK_LINEFEED) {
1818             if (tok == TOK_STR)
1819                 pstrcpy(file->filename, sizeof(file->filename), (char *)tokc.cstr->data);
1820             else if (parse_flags & PARSE_FLAG_ASM_FILE)
1821                 break;
1822             else
1823                 goto _line_err;
1824             --n;
1825         }
1826         if (file->fd > 0)
1827             total_lines += file->line_num - n;
1828         file->line_num = n;
1829         if (s1->do_debug)
1830             put_stabs(file->filename, N_BINCL, 0, 0, 0);
1831         break;
1832     case TOK_ERROR:
1833     case TOK_WARNING:
1834         c = tok;
1835         ch = file->buf_ptr[0];
1836         skip_spaces();
1837         q = buf;
1838         while (ch != '\n' && ch != CH_EOF) {
1839             if ((q - buf) < sizeof(buf) - 1)
1840                 *q++ = ch;
1841             if (ch == '\\') {
1842                 if (handle_stray_noerror() == 0)
1843                     --q;
1844             } else
1845                 inp();
1846         }
1847         *q = '\0';
1848         if (c == TOK_ERROR)
1849             tcc_error("#error %s", buf);
1850         else
1851             tcc_warning("#warning %s", buf);
1852         break;
1853     case TOK_PRAGMA:
1854         pragma_parse(s1);
1855         break;
1856     case TOK_LINEFEED:
1857         goto the_end;
1858     default:
1859         /* ignore gas line comment in an 'S' file. */
1860         if (saved_parse_flags & PARSE_FLAG_ASM_FILE)
1861             goto ignore;
1862         if (tok == '!' && is_bof)
1863             /* '!' is ignored at beginning to allow C scripts. */
1864             goto ignore;
1865         tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1866     ignore:
1867         file->buf_ptr = parse_line_comment(file->buf_ptr);
1868         goto the_end;
1869     }
1870     /* ignore other preprocess commands or #! for C scripts */
1871     while (tok != TOK_LINEFEED)
1872         next_nomacro();
1873  the_end:
1874     parse_flags = saved_parse_flags;
1875 }
1876
1877 /* evaluate escape codes in a string. */
1878 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1879 {
1880     int c, n;
1881     const uint8_t *p;
1882
1883     p = buf;
1884     for(;;) {
1885         c = *p;
1886         if (c == '\0')
1887             break;
1888         if (c == '\\') {
1889             p++;
1890             /* escape */
1891             c = *p;
1892             switch(c) {
1893             case '0': case '1': case '2': case '3':
1894             case '4': case '5': case '6': case '7':
1895                 /* at most three octal digits */
1896                 n = c - '0';
1897                 p++;
1898                 c = *p;
1899                 if (isoct(c)) {
1900                     n = n * 8 + c - '0';
1901                     p++;
1902                     c = *p;
1903                     if (isoct(c)) {
1904                         n = n * 8 + c - '0';
1905                         p++;
1906                     }
1907                 }
1908                 c = n;
1909                 goto add_char_nonext;
1910             case 'x':
1911             case 'u':
1912             case 'U':
1913                 p++;
1914                 n = 0;
1915                 for(;;) {
1916                     c = *p;
1917                     if (c >= 'a' && c <= 'f')
1918                         c = c - 'a' + 10;
1919                     else if (c >= 'A' && c <= 'F')
1920                         c = c - 'A' + 10;
1921                     else if (isnum(c))
1922                         c = c - '0';
1923                     else
1924                         break;
1925                     n = n * 16 + c;
1926                     p++;
1927                 }
1928                 c = n;
1929                 goto add_char_nonext;
1930             case 'a':
1931                 c = '\a';
1932                 break;
1933             case 'b':
1934                 c = '\b';
1935                 break;
1936             case 'f':
1937                 c = '\f';
1938                 break;
1939             case 'n':
1940                 c = '\n';
1941                 break;
1942             case 'r':
1943                 c = '\r';
1944                 break;
1945             case 't':
1946                 c = '\t';
1947                 break;
1948             case 'v':
1949                 c = '\v';
1950                 break;
1951             case 'e':
1952                 if (!gnu_ext)
1953                     goto invalid_escape;
1954                 c = 27;
1955                 break;
1956             case '\'':
1957             case '\"':
1958             case '\\':
1959             case '?':
1960                 break;
1961             default:
1962             invalid_escape:
1963                 if (c >= '!' && c <= '~')
1964                     tcc_warning("unknown escape sequence: \'\\%c\'", c);
1965                 else
1966                     tcc_warning("unknown escape sequence: \'\\x%x\'", c);
1967                 break;
1968             }
1969         }
1970         p++;
1971     add_char_nonext:
1972         if (!is_long)
1973             cstr_ccat(outstr, c);
1974         else
1975             cstr_wccat(outstr, c);
1976     }
1977     /* add a trailing '\0' */
1978     if (!is_long)
1979         cstr_ccat(outstr, '\0');
1980     else
1981         cstr_wccat(outstr, '\0');
1982 }
1983
1984 void parse_string(const char *s, int len)
1985 {
1986     uint8_t buf[1000], *p = buf;
1987     int is_long, sep;
1988
1989     if ((is_long = *s == 'L'))
1990         ++s, --len;
1991     sep = *s++;
1992     len -= 2;
1993     if (len >= sizeof buf)
1994         p = tcc_malloc(len + 1);
1995     memcpy(p, s, len);
1996     p[len] = 0;
1997
1998     cstr_reset(&tokcstr);
1999     parse_escape_string(&tokcstr, p, is_long);
2000     if (p != buf)
2001         tcc_free(p);
2002
2003     if (sep == '\'') {
2004         int char_size;
2005         /* XXX: make it portable */
2006         if (!is_long)
2007             char_size = 1;
2008         else
2009             char_size = sizeof(nwchar_t);
2010         if (tokcstr.size <= char_size)
2011             tcc_error("empty character constant");
2012         if (tokcstr.size > 2 * char_size)
2013             tcc_warning("multi-character character constant");
2014         if (!is_long) {
2015             tokc.i = *(int8_t *)tokcstr.data;
2016             tok = TOK_CCHAR;
2017         } else {
2018             tokc.i = *(nwchar_t *)tokcstr.data;
2019             tok = TOK_LCHAR;
2020         }
2021     } else {
2022         tokc.cstr = &tokcstr;
2023         if (!is_long)
2024             tok = TOK_STR;
2025         else
2026             tok = TOK_LSTR;
2027     }
2028 }
2029
2030 /* we use 64 bit numbers */
2031 #define BN_SIZE 2
2032
2033 /* bn = (bn << shift) | or_val */
2034 static void bn_lshift(unsigned int *bn, int shift, int or_val)
2035 {
2036     int i;
2037     unsigned int v;
2038     for(i=0;i<BN_SIZE;i++) {
2039         v = bn[i];
2040         bn[i] = (v << shift) | or_val;
2041         or_val = v >> (32 - shift);
2042     }
2043 }
2044
2045 static void bn_zero(unsigned int *bn)
2046 {
2047     int i;
2048     for(i=0;i<BN_SIZE;i++) {
2049         bn[i] = 0;
2050     }
2051 }
2052
2053 /* parse number in null terminated string 'p' and return it in the
2054    current token */
2055 static void parse_number(const char *p)
2056 {
2057     int b, t, shift, frac_bits, s, exp_val, ch;
2058     char *q;
2059     unsigned int bn[BN_SIZE];
2060     double d;
2061
2062     /* number */
2063     q = token_buf;
2064     ch = *p++;
2065     t = ch;
2066     ch = *p++;
2067     *q++ = t;
2068     b = 10;
2069     if (t == '.') {
2070         goto float_frac_parse;
2071     } else if (t == '0') {
2072         if (ch == 'x' || ch == 'X') {
2073             q--;
2074             ch = *p++;
2075             b = 16;
2076         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
2077             q--;
2078             ch = *p++;
2079             b = 2;
2080         }
2081     }
2082     /* parse all digits. cannot check octal numbers at this stage
2083        because of floating point constants */
2084     while (1) {
2085         if (ch >= 'a' && ch <= 'f')
2086             t = ch - 'a' + 10;
2087         else if (ch >= 'A' && ch <= 'F')
2088             t = ch - 'A' + 10;
2089         else if (isnum(ch))
2090             t = ch - '0';
2091         else
2092             break;
2093         if (t >= b)
2094             break;
2095         if (q >= token_buf + STRING_MAX_SIZE) {
2096         num_too_long:
2097             tcc_error("number too long");
2098         }
2099         *q++ = ch;
2100         ch = *p++;
2101     }
2102     if (ch == '.' ||
2103         ((ch == 'e' || ch == 'E') && b == 10) ||
2104         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
2105         if (b != 10) {
2106             /* NOTE: strtox should support that for hexa numbers, but
2107                non ISOC99 libcs do not support it, so we prefer to do
2108                it by hand */
2109             /* hexadecimal or binary floats */
2110             /* XXX: handle overflows */
2111             *q = '\0';
2112             if (b == 16)
2113                 shift = 4;
2114             else
2115                 shift = 1;
2116             bn_zero(bn);
2117             q = token_buf;
2118             while (1) {
2119                 t = *q++;
2120                 if (t == '\0') {
2121                     break;
2122                 } else if (t >= 'a') {
2123                     t = t - 'a' + 10;
2124                 } else if (t >= 'A') {
2125                     t = t - 'A' + 10;
2126                 } else {
2127                     t = t - '0';
2128                 }
2129                 bn_lshift(bn, shift, t);
2130             }
2131             frac_bits = 0;
2132             if (ch == '.') {
2133                 ch = *p++;
2134                 while (1) {
2135                     t = ch;
2136                     if (t >= 'a' && t <= 'f') {
2137                         t = t - 'a' + 10;
2138                     } else if (t >= 'A' && t <= 'F') {
2139                         t = t - 'A' + 10;
2140                     } else if (t >= '0' && t <= '9') {
2141                         t = t - '0';
2142                     } else {
2143                         break;
2144                     }
2145                     if (t >= b)
2146                         tcc_error("invalid digit");
2147                     bn_lshift(bn, shift, t);
2148                     frac_bits += shift;
2149                     ch = *p++;
2150                 }
2151             }
2152             if (ch != 'p' && ch != 'P')
2153                 expect("exponent");
2154             ch = *p++;
2155             s = 1;
2156             exp_val = 0;
2157             if (ch == '+') {
2158                 ch = *p++;
2159             } else if (ch == '-') {
2160                 s = -1;
2161                 ch = *p++;
2162             }
2163             if (ch < '0' || ch > '9')
2164                 expect("exponent digits");
2165             while (ch >= '0' && ch <= '9') {
2166                 exp_val = exp_val * 10 + ch - '0';
2167                 ch = *p++;
2168             }
2169             exp_val = exp_val * s;
2170
2171             /* now we can generate the number */
2172             /* XXX: should patch directly float number */
2173             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
2174             d = ldexp(d, exp_val - frac_bits);
2175             t = toup(ch);
2176             if (t == 'F') {
2177                 ch = *p++;
2178                 tok = TOK_CFLOAT;
2179                 /* float : should handle overflow */
2180                 tokc.f = (float)d;
2181             } else if (t == 'L') {
2182                 ch = *p++;
2183 #ifdef TCC_TARGET_PE
2184                 tok = TOK_CDOUBLE;
2185                 tokc.d = d;
2186 #else
2187                 tok = TOK_CLDOUBLE;
2188                 /* XXX: not large enough */
2189                 tokc.ld = (long double)d;
2190 #endif
2191             } else {
2192                 tok = TOK_CDOUBLE;
2193                 tokc.d = d;
2194             }
2195         } else {
2196             /* decimal floats */
2197             if (ch == '.') {
2198                 if (q >= token_buf + STRING_MAX_SIZE)
2199                     goto num_too_long;
2200                 *q++ = ch;
2201                 ch = *p++;
2202             float_frac_parse:
2203                 while (ch >= '0' && ch <= '9') {
2204                     if (q >= token_buf + STRING_MAX_SIZE)
2205                         goto num_too_long;
2206                     *q++ = ch;
2207                     ch = *p++;
2208                 }
2209             }
2210             if (ch == 'e' || ch == 'E') {
2211                 if (q >= token_buf + STRING_MAX_SIZE)
2212                     goto num_too_long;
2213                 *q++ = ch;
2214                 ch = *p++;
2215                 if (ch == '-' || ch == '+') {
2216                     if (q >= token_buf + STRING_MAX_SIZE)
2217                         goto num_too_long;
2218                     *q++ = ch;
2219                     ch = *p++;
2220                 }
2221                 if (ch < '0' || ch > '9')
2222                     expect("exponent digits");
2223                 while (ch >= '0' && ch <= '9') {
2224                     if (q >= token_buf + STRING_MAX_SIZE)
2225                         goto num_too_long;
2226                     *q++ = ch;
2227                     ch = *p++;
2228                 }
2229             }
2230             *q = '\0';
2231             t = toup(ch);
2232             errno = 0;
2233             if (t == 'F') {
2234                 ch = *p++;
2235                 tok = TOK_CFLOAT;
2236                 tokc.f = strtof(token_buf, NULL);
2237             } else if (t == 'L') {
2238                 ch = *p++;
2239 #ifdef TCC_TARGET_PE
2240                 tok = TOK_CDOUBLE;
2241                 tokc.d = strtod(token_buf, NULL);
2242 #else
2243                 tok = TOK_CLDOUBLE;
2244                 tokc.ld = strtold(token_buf, NULL);
2245 #endif
2246             } else {
2247                 tok = TOK_CDOUBLE;
2248                 tokc.d = strtod(token_buf, NULL);
2249             }
2250         }
2251     } else {
2252         unsigned long long n, n1;
2253         int lcount, ucount, must_64bit;
2254         const char *p1;
2255
2256         /* integer number */
2257         *q = '\0';
2258         q = token_buf;
2259         if (b == 10 && *q == '0') {
2260             b = 8;
2261             q++;
2262         }
2263         n = 0;
2264         while(1) {
2265             t = *q++;
2266             /* no need for checks except for base 10 / 8 errors */
2267             if (t == '\0')
2268                 break;
2269             else if (t >= 'a')
2270                 t = t - 'a' + 10;
2271             else if (t >= 'A')
2272                 t = t - 'A' + 10;
2273             else
2274                 t = t - '0';
2275             if (t >= b)
2276                 tcc_error("invalid digit");
2277             n1 = n;
2278             n = n * b + t;
2279             /* detect overflow */
2280             /* XXX: this test is not reliable */
2281             if (n < n1)
2282                 tcc_error("integer constant overflow");
2283         }
2284
2285         /* Determine the characteristics (unsigned and/or 64bit) the type of
2286            the constant must have according to the constant suffix(es) */
2287         lcount = ucount = must_64bit = 0;
2288         p1 = p;
2289         for(;;) {
2290             t = toup(ch);
2291             if (t == 'L') {
2292                 if (lcount >= 2)
2293                     tcc_error("three 'l's in integer constant");
2294                 if (lcount && *(p - 1) != ch)
2295                     tcc_error("incorrect integer suffix: %s", p1);
2296                 lcount++;
2297 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
2298                 if (lcount == 2)
2299 #endif
2300                     must_64bit = 1;
2301                 ch = *p++;
2302             } else if (t == 'U') {
2303                 if (ucount >= 1)
2304                     tcc_error("two 'u's in integer constant");
2305                 ucount++;
2306                 ch = *p++;
2307             } else {
2308                 break;
2309             }
2310         }
2311
2312         /* Whether 64 bits are needed to hold the constant's value */
2313         if (n & 0xffffffff00000000LL || must_64bit) {
2314             tok = TOK_CLLONG;
2315             n1 = n >> 32;
2316         } else {
2317             tok = TOK_CINT;
2318             n1 = n;
2319         }
2320
2321         /* Whether type must be unsigned to hold the constant's value */
2322         if (ucount || ((n1 >> 31) && (b != 10))) {
2323             if (tok == TOK_CLLONG)
2324                 tok = TOK_CULLONG;
2325             else
2326                 tok = TOK_CUINT;
2327         /* If decimal and no unsigned suffix, bump to 64 bits or throw error */
2328         } else if (n1 >> 31) {
2329             if (tok == TOK_CINT)
2330                 tok = TOK_CLLONG;
2331             else
2332                 tcc_error("integer constant overflow");
2333         }
2334
2335         if (tok == TOK_CINT || tok == TOK_CUINT)
2336             tokc.ui = n;
2337         else
2338             tokc.ull = n;
2339     }
2340     if (ch)
2341         tcc_error("invalid number\n");
2342 }
2343
2344
2345 #define PARSE2(c1, tok1, c2, tok2)              \
2346     case c1:                                    \
2347         PEEKC(c, p);                            \
2348         if (c == c2) {                          \
2349             p++;                                \
2350             tok = tok2;                         \
2351         } else {                                \
2352             tok = tok1;                         \
2353         }                                       \
2354         break;
2355
2356 /* return next token without macro substitution */
2357 static inline void next_nomacro1(void)
2358 {
2359     int t, c, is_long;
2360     TokenSym *ts;
2361     uint8_t *p, *p1;
2362     unsigned int h;
2363
2364     p = file->buf_ptr;
2365  redo_no_start:
2366     c = *p;
2367     switch(c) {
2368     case ' ':
2369     case '\t':
2370         tok = c;
2371         p++;
2372         if (parse_flags & PARSE_FLAG_SPACES)
2373             goto keep_tok_flags;
2374         while (isidnum_table[*p - CH_EOF] & IS_SPC)
2375             ++p;
2376         goto redo_no_start;
2377     case '\f':
2378     case '\v':
2379     case '\r':
2380         p++;
2381         goto redo_no_start;
2382     case '\\':
2383         /* first look if it is in fact an end of buffer */
2384         c = handle_stray1(p);
2385         p = file->buf_ptr;
2386         if (c == '\\')
2387             goto parse_simple;
2388         if (c != CH_EOF)
2389             goto redo_no_start;
2390         {
2391             TCCState *s1 = tcc_state;
2392             if ((parse_flags & PARSE_FLAG_LINEFEED)
2393                 && !(tok_flags & TOK_FLAG_EOF)) {
2394                 tok_flags |= TOK_FLAG_EOF;
2395                 tok = TOK_LINEFEED;
2396                 goto keep_tok_flags;
2397             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2398                 tok = TOK_EOF;
2399             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2400                 tcc_error("missing #endif");
2401             } else if (s1->include_stack_ptr == s1->include_stack) {
2402                 /* no include left : end of file. */
2403                 tok = TOK_EOF;
2404             } else {
2405                 tok_flags &= ~TOK_FLAG_EOF;
2406                 /* pop include file */
2407
2408                 /* test if previous '#endif' was after a #ifdef at
2409                    start of file */
2410                 if (tok_flags & TOK_FLAG_ENDIF) {
2411 #ifdef INC_DEBUG
2412                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2413 #endif
2414                     add_cached_include(s1, file->filename, file->ifndef_macro_saved);
2415                     tok_flags &= ~TOK_FLAG_ENDIF;
2416                 }
2417
2418                 /* add end of include file debug info */
2419                 if (tcc_state->do_debug) {
2420                     put_stabd(N_EINCL, 0, 0);
2421                 }
2422                 /* pop include stack */
2423                 tcc_close();
2424                 s1->include_stack_ptr--;
2425                 p = file->buf_ptr;
2426                 goto redo_no_start;
2427             }
2428         }
2429         break;
2430
2431     case '\n':
2432         file->line_num++;
2433         tok_flags |= TOK_FLAG_BOL;
2434         p++;
2435 maybe_newline:
2436         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2437             goto redo_no_start;
2438         tok = TOK_LINEFEED;
2439         goto keep_tok_flags;
2440
2441     case '#':
2442         /* XXX: simplify */
2443         PEEKC(c, p);
2444         if ((tok_flags & TOK_FLAG_BOL) &&
2445             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2446             file->buf_ptr = p;
2447             preprocess(tok_flags & TOK_FLAG_BOF);
2448             p = file->buf_ptr;
2449             goto maybe_newline;
2450         } else {
2451             if (c == '#') {
2452                 p++;
2453                 tok = TOK_TWOSHARPS;
2454             } else {
2455                 if (parse_flags & PARSE_FLAG_ASM_FILE) {
2456                     p = parse_line_comment(p - 1);
2457                     goto redo_no_start;
2458                 } else {
2459                     tok = '#';
2460                 }
2461             }
2462         }
2463         break;
2464
2465     /* dollar is allowed to start identifiers when not parsing asm */
2466     case '$':
2467         if (!(isidnum_table[c - CH_EOF] & IS_ID)
2468          || (parse_flags & PARSE_FLAG_ASM_FILE))
2469             goto parse_simple;
2470
2471     case 'a': case 'b': case 'c': case 'd':
2472     case 'e': case 'f': case 'g': case 'h':
2473     case 'i': case 'j': case 'k': case 'l':
2474     case 'm': case 'n': case 'o': case 'p':
2475     case 'q': case 'r': case 's': case 't':
2476     case 'u': case 'v': case 'w': case 'x':
2477     case 'y': case 'z':
2478     case 'A': case 'B': case 'C': case 'D':
2479     case 'E': case 'F': case 'G': case 'H':
2480     case 'I': case 'J': case 'K':
2481     case 'M': case 'N': case 'O': case 'P':
2482     case 'Q': case 'R': case 'S': case 'T':
2483     case 'U': case 'V': case 'W': case 'X':
2484     case 'Y': case 'Z':
2485     case '_':
2486     parse_ident_fast:
2487         p1 = p;
2488         h = TOK_HASH_INIT;
2489         h = TOK_HASH_FUNC(h, c);
2490         while (c = *++p, isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2491             h = TOK_HASH_FUNC(h, c);
2492         if (c != '\\') {
2493             TokenSym **pts;
2494             int len;
2495
2496             /* fast case : no stray found, so we have the full token
2497                and we have already hashed it */
2498             len = p - p1;
2499             h &= (TOK_HASH_SIZE - 1);
2500             pts = &hash_ident[h];
2501             for(;;) {
2502                 ts = *pts;
2503                 if (!ts)
2504                     break;
2505                 if (ts->len == len && !memcmp(ts->str, p1, len))
2506                     goto token_found;
2507                 pts = &(ts->hash_next);
2508             }
2509             ts = tok_alloc_new(pts, (char *) p1, len);
2510         token_found: ;
2511         } else {
2512             /* slower case */
2513             cstr_reset(&tokcstr);
2514
2515             while (p1 < p) {
2516                 cstr_ccat(&tokcstr, *p1);
2517                 p1++;
2518             }
2519             p--;
2520             PEEKC(c, p);
2521         parse_ident_slow:
2522             while (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM)) {
2523                 cstr_ccat(&tokcstr, c);
2524                 PEEKC(c, p);
2525             }
2526             ts = tok_alloc(tokcstr.data, tokcstr.size);
2527         }
2528         tok = ts->tok;
2529         break;
2530     case 'L':
2531         t = p[1];
2532         if (t != '\\' && t != '\'' && t != '\"') {
2533             /* fast case */
2534             goto parse_ident_fast;
2535         } else {
2536             PEEKC(c, p);
2537             if (c == '\'' || c == '\"') {
2538                 is_long = 1;
2539                 goto str_const;
2540             } else {
2541                 cstr_reset(&tokcstr);
2542                 cstr_ccat(&tokcstr, 'L');
2543                 goto parse_ident_slow;
2544             }
2545         }
2546         break;
2547
2548     case '0': case '1': case '2': case '3':
2549     case '4': case '5': case '6': case '7':
2550     case '8': case '9':
2551         cstr_reset(&tokcstr);
2552         /* after the first digit, accept digits, alpha, '.' or sign if
2553            prefixed by 'eEpP' */
2554     parse_num:
2555         for(;;) {
2556             t = c;
2557             cstr_ccat(&tokcstr, c);
2558             PEEKC(c, p);
2559             if (!((isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2560                   || c == '.'
2561                   || ((c == '+' || c == '-')
2562                       && (t == 'e' || t == 'E' || t == 'p' || t == 'P')
2563                       )))
2564                 break;
2565         }
2566         /* We add a trailing '\0' to ease parsing */
2567         cstr_ccat(&tokcstr, '\0');
2568         tokc.cstr = &tokcstr;
2569         tok = TOK_PPNUM;
2570         break;
2571
2572     case '.':
2573         /* special dot handling because it can also start a number */
2574         PEEKC(c, p);
2575         if (isnum(c)) {
2576             cstr_reset(&tokcstr);
2577             cstr_ccat(&tokcstr, '.');
2578             goto parse_num;
2579         } else if ((c == '.') && (p[1] == '.')){
2580             PEEKC(c, p);
2581             PEEKC(c, p);
2582             tok = TOK_DOTS;
2583         } else {
2584             tok = '.';
2585         }
2586         break;
2587     case '\'':
2588     case '\"':
2589         is_long = 0;
2590     str_const:
2591         cstr_reset(&tokcstr);
2592         if (is_long)
2593             cstr_ccat(&tokcstr, 'L');
2594         cstr_ccat(&tokcstr, c);
2595         p = parse_pp_string(p, c, &tokcstr);
2596         cstr_ccat(&tokcstr, c);
2597         cstr_ccat(&tokcstr, '\0');
2598         tokc.cstr = &tokcstr;
2599         tok = TOK_PPSTR;
2600         break;
2601
2602     case '<':
2603         PEEKC(c, p);
2604         if (c == '=') {
2605             p++;
2606             tok = TOK_LE;
2607         } else if (c == '<') {
2608             PEEKC(c, p);
2609             if (c == '=') {
2610                 p++;
2611                 tok = TOK_A_SHL;
2612             } else {
2613                 tok = TOK_SHL;
2614             }
2615         } else {
2616             tok = TOK_LT;
2617         }
2618         break;
2619     case '>':
2620         PEEKC(c, p);
2621         if (c == '=') {
2622             p++;
2623             tok = TOK_GE;
2624         } else if (c == '>') {
2625             PEEKC(c, p);
2626             if (c == '=') {
2627                 p++;
2628                 tok = TOK_A_SAR;
2629             } else {
2630                 tok = TOK_SAR;
2631             }
2632         } else {
2633             tok = TOK_GT;
2634         }
2635         break;
2636
2637     case '&':
2638         PEEKC(c, p);
2639         if (c == '&') {
2640             p++;
2641             tok = TOK_LAND;
2642         } else if (c == '=') {
2643             p++;
2644             tok = TOK_A_AND;
2645         } else {
2646             tok = '&';
2647         }
2648         break;
2649
2650     case '|':
2651         PEEKC(c, p);
2652         if (c == '|') {
2653             p++;
2654             tok = TOK_LOR;
2655         } else if (c == '=') {
2656             p++;
2657             tok = TOK_A_OR;
2658         } else {
2659             tok = '|';
2660         }
2661         break;
2662
2663     case '+':
2664         PEEKC(c, p);
2665         if (c == '+') {
2666             p++;
2667             tok = TOK_INC;
2668         } else if (c == '=') {
2669             p++;
2670             tok = TOK_A_ADD;
2671         } else {
2672             tok = '+';
2673         }
2674         break;
2675
2676     case '-':
2677         PEEKC(c, p);
2678         if (c == '-') {
2679             p++;
2680             tok = TOK_DEC;
2681         } else if (c == '=') {
2682             p++;
2683             tok = TOK_A_SUB;
2684         } else if (c == '>') {
2685             p++;
2686             tok = TOK_ARROW;
2687         } else {
2688             tok = '-';
2689         }
2690         break;
2691
2692     PARSE2('!', '!', '=', TOK_NE)
2693     PARSE2('=', '=', '=', TOK_EQ)
2694     PARSE2('*', '*', '=', TOK_A_MUL)
2695     PARSE2('%', '%', '=', TOK_A_MOD)
2696     PARSE2('^', '^', '=', TOK_A_XOR)
2697
2698         /* comments or operator */
2699     case '/':
2700         PEEKC(c, p);
2701         if (c == '*') {
2702             p = parse_comment(p);
2703             /* comments replaced by a blank */
2704             tok = ' ';
2705             goto keep_tok_flags;
2706         } else if (c == '/') {
2707             p = parse_line_comment(p);
2708             tok = ' ';
2709             goto keep_tok_flags;
2710         } else if (c == '=') {
2711             p++;
2712             tok = TOK_A_DIV;
2713         } else {
2714             tok = '/';
2715         }
2716         break;
2717
2718         /* simple tokens */
2719     case '(':
2720     case ')':
2721     case '[':
2722     case ']':
2723     case '{':
2724     case '}':
2725     case ',':
2726     case ';':
2727     case ':':
2728     case '?':
2729     case '~':
2730     case '@': /* only used in assembler */
2731     parse_simple:
2732         tok = c;
2733         p++;
2734         break;
2735     default:
2736         tcc_error("unrecognized character \\x%02x", c);
2737         break;
2738     }
2739     tok_flags = 0;
2740 keep_tok_flags:
2741     file->buf_ptr = p;
2742 #if defined(PARSE_DEBUG)
2743     printf("token = %s\n", get_tok_str(tok, &tokc));
2744 #endif
2745 }
2746
2747 /* return next token without macro substitution. Can read input from
2748    macro_ptr buffer */
2749 static void next_nomacro_spc(void)
2750 {
2751     if (macro_ptr) {
2752     redo:
2753         tok = *macro_ptr;
2754         if (tok) {
2755             TOK_GET(&tok, &macro_ptr, &tokc);
2756             if (tok == TOK_LINENUM) {
2757                 file->line_num = tokc.i;
2758                 goto redo;
2759             }
2760         }
2761     } else {
2762         next_nomacro1();
2763     }
2764 }
2765
2766 ST_FUNC void next_nomacro(void)
2767 {
2768     do {
2769         next_nomacro_spc();
2770     } while (tok < 256 && (isidnum_table[tok - CH_EOF] & IS_SPC));
2771 }
2772
2773
2774 static void macro_subst(
2775     TokenString *tok_str,
2776     Sym **nested_list,
2777     const int *macro_str,
2778     int can_read_stream
2779     );
2780
2781 /* substitute arguments in replacement lists in macro_str by the values in
2782    args (field d) and return allocated string */
2783 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2784 {
2785     int t, t0, t1, spc;
2786     const int *st;
2787     Sym *s;
2788     CValue cval;
2789     TokenString str;
2790     CString cstr;
2791
2792     tok_str_new(&str);
2793     t0 = t1 = 0;
2794     while(1) {
2795         TOK_GET(&t, &macro_str, &cval);
2796         if (!t)
2797             break;
2798         if (t == '#') {
2799             /* stringize */
2800             TOK_GET(&t, &macro_str, &cval);
2801             if (!t)
2802                 goto bad_stringy;
2803             s = sym_find2(args, t);
2804             if (s) {
2805                 cstr_new(&cstr);
2806                 cstr_ccat(&cstr, '\"');
2807                 st = s->d;
2808                 spc = 0;
2809                 while (*st) {
2810                     TOK_GET(&t, &st, &cval);
2811                     if (t != TOK_PLCHLDR
2812                      && t != TOK_NOSUBST
2813                      && 0 == check_space(t, &spc)) {
2814                         const char *s = get_tok_str(t, &cval);
2815                         while (*s) {
2816                             if (t == TOK_PPSTR && *s != '\'')
2817                                 add_char(&cstr, *s);
2818                             else
2819                                 cstr_ccat(&cstr, *s);
2820                             ++s;
2821                         }
2822                     }
2823                 }
2824                 cstr.size -= spc;
2825                 cstr_ccat(&cstr, '\"');
2826                 cstr_ccat(&cstr, '\0');
2827 #ifdef PP_DEBUG
2828                 printf("\nstringize: <%s>\n", (char *)cstr.data);
2829 #endif
2830                 /* add string */
2831                 cval.cstr = &cstr;
2832                 tok_str_add2(&str, TOK_PPSTR, &cval);
2833                 cstr_free(cval.cstr);
2834             } else {
2835         bad_stringy:
2836                 expect("macro parameter after '#'");
2837             }
2838         } else if (t >= TOK_IDENT) {
2839             s = sym_find2(args, t);
2840             if (s) {
2841                 int l0 = str.len;
2842                 st = s->d;
2843                 /* if '##' is present before or after, no arg substitution */
2844                 if (*macro_str == TOK_TWOSHARPS || t1 == TOK_TWOSHARPS) {
2845                     /* special case for var arg macros : ## eats the ','
2846                        if empty VA_ARGS variable. */
2847                     if (t1 == TOK_TWOSHARPS && t0 == ',' && gnu_ext && s->type.t) {
2848                         if (*st == 0) {
2849                             /* suppress ',' '##' */
2850                             str.len -= 2;
2851                         } else {
2852                             /* suppress '##' and add variable */
2853                             str.len--;
2854                             goto add_var;
2855                         }
2856                     } else {
2857                         for(;;) {
2858                             int t1;
2859                             TOK_GET(&t1, &st, &cval);
2860                             if (!t1)
2861                                 break;
2862                             tok_str_add2(&str, t1, &cval);
2863                         }
2864                     }
2865
2866                 } else {
2867             add_var:
2868                     /* NOTE: the stream cannot be read when macro
2869                        substituing an argument */
2870                     macro_subst(&str, nested_list, st, 0);
2871                 }
2872                 if (str.len == l0) /* exanded to empty string */
2873                     tok_str_add(&str, TOK_PLCHLDR);
2874             } else {
2875                 tok_str_add(&str, t);
2876             }
2877         } else {
2878             tok_str_add2(&str, t, &cval);
2879         }
2880         t0 = t1, t1 = t;
2881     }
2882     tok_str_add(&str, 0);
2883     return str.str;
2884 }
2885
2886 static char const ab_month_name[12][4] =
2887 {
2888     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2889     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2890 };
2891
2892 /* peek or read [ws_str == NULL] next token from function macro call,
2893    walking up macro levels up to the file if necessary */
2894 static int next_argstream(Sym **nested_list, int can_read_stream, TokenString *ws_str)
2895 {
2896     int t;
2897     const int *p;
2898     Sym *sa;
2899
2900     for (;;) {
2901         if (macro_ptr) {
2902             p = macro_ptr, t = *p;
2903             if (ws_str) {
2904                 while (is_space(t) || TOK_LINEFEED == t)
2905                     tok_str_add(ws_str, t), t = *++p;
2906             }
2907             if (t == 0 && can_read_stream) {
2908                 end_macro();
2909                 /* also, end of scope for nested defined symbol */
2910                 sa = *nested_list;
2911                 while (sa && sa->v == -1)
2912                     sa = sa->prev;
2913                 if (sa)
2914                     sa->v = -1;
2915                 continue;
2916             }
2917         } else {
2918             ch = handle_eob();
2919             if (ws_str) {
2920                 while (is_space(ch) || ch == '\n' || ch == '/') {
2921                     if (ch == '/') {
2922                         int c;
2923                         uint8_t *p = file->buf_ptr;
2924                         PEEKC(c, p);
2925                         if (c == '*') {
2926                             p = parse_comment(p);
2927                             file->buf_ptr = p - 1;
2928                         } else if (c == '/') {
2929                             p = parse_line_comment(p);
2930                             file->buf_ptr = p - 1;
2931                         } else
2932                             break;
2933                         ch = ' ';
2934                     }
2935                     tok_str_add(ws_str, ch);
2936                     cinp();
2937                 }
2938             }
2939             t = ch;
2940         }
2941
2942         if (ws_str)
2943             return t;
2944         next_nomacro_spc();
2945         return tok;
2946     }
2947 }
2948
2949 /* do macro substitution of current token with macro 's' and add
2950    result to (tok_str,tok_len). 'nested_list' is the list of all
2951    macros we got inside to avoid recursing. Return non zero if no
2952    substitution needs to be done */
2953 static int macro_subst_tok(
2954     TokenString *tok_str,
2955     Sym **nested_list,
2956     Sym *s,
2957     int can_read_stream)
2958 {
2959     Sym *args, *sa, *sa1;
2960     int parlevel, *mstr, t, t1, spc;
2961     TokenString str;
2962     char *cstrval;
2963     CValue cval;
2964     CString cstr;
2965     char buf[32];
2966
2967     /* if symbol is a macro, prepare substitution */
2968     /* special macros */
2969     if (tok == TOK___LINE__) {
2970         snprintf(buf, sizeof(buf), "%d", file->line_num);
2971         cstrval = buf;
2972         t1 = TOK_PPNUM;
2973         goto add_cstr1;
2974     } else if (tok == TOK___FILE__) {
2975         cstrval = file->filename;
2976         goto add_cstr;
2977     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2978         time_t ti;
2979         struct tm *tm;
2980
2981         time(&ti);
2982         tm = localtime(&ti);
2983         if (tok == TOK___DATE__) {
2984             snprintf(buf, sizeof(buf), "%s %2d %d",
2985                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2986         } else {
2987             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2988                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2989         }
2990         cstrval = buf;
2991     add_cstr:
2992         t1 = TOK_STR;
2993     add_cstr1:
2994         cstr_new(&cstr);
2995         cstr_cat(&cstr, cstrval);
2996         cstr_ccat(&cstr, '\0');
2997         cval.cstr = &cstr;
2998         tok_str_add2(tok_str, t1, &cval);
2999         cstr_free(&cstr);
3000     } else {
3001         int saved_parse_flags = parse_flags;
3002
3003         mstr = s->d;
3004         if (s->type.t == MACRO_FUNC) {
3005             /* whitespace between macro name and argument list */
3006             TokenString ws_str;
3007             tok_str_new(&ws_str);
3008
3009             spc = 0;
3010             parse_flags |= PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED
3011                 | PARSE_FLAG_ACCEPT_STRAYS;
3012
3013             /* get next token from argument stream */
3014             t = next_argstream(nested_list, can_read_stream, &ws_str);
3015             if (t != '(') {
3016                 /* not a macro substitution after all, restore the
3017                  * macro token plus all whitespace we've read.
3018                  * whitespace is intentionally not merged to preserve
3019                  * newlines. */
3020                 parse_flags = saved_parse_flags;
3021                 tok_str_add(tok_str, tok);
3022                 if (parse_flags & PARSE_FLAG_SPACES) {
3023                     int i;
3024                     for (i = 0; i < ws_str.len; i++)
3025                         tok_str_add(tok_str, ws_str.str[i]);
3026                 }
3027                 tok_str_free(ws_str.str);
3028                 return 0;
3029             } else {
3030                 tok_str_free(ws_str.str);
3031             }
3032             next_nomacro(); /* eat '(' */
3033
3034             /* argument macro */
3035             args = NULL;
3036             sa = s->next;
3037             /* NOTE: empty args are allowed, except if no args */
3038             for(;;) {
3039                 do {
3040                     next_argstream(nested_list, can_read_stream, NULL);
3041                 } while (is_space(tok) || TOK_LINEFEED == tok);
3042     empty_arg:
3043                 /* handle '()' case */
3044                 if (!args && !sa && tok == ')')
3045                     break;
3046                 if (!sa)
3047                     tcc_error("macro '%s' used with too many args",
3048                           get_tok_str(s->v, 0));
3049                 tok_str_new(&str);
3050                 parlevel = spc = 0;
3051                 /* NOTE: non zero sa->t indicates VA_ARGS */
3052                 while ((parlevel > 0 ||
3053                         (tok != ')' &&
3054                          (tok != ',' || sa->type.t)))) {
3055                     if (tok == TOK_EOF || tok == 0)
3056                         break;
3057                     if (tok == '(')
3058                         parlevel++;
3059                     else if (tok == ')')
3060                         parlevel--;
3061                     if (tok == TOK_LINEFEED)
3062                         tok = ' ';
3063                     if (!check_space(tok, &spc))
3064                         tok_str_add2(&str, tok, &tokc);
3065                     next_argstream(nested_list, can_read_stream, NULL);
3066                 }
3067                 if (parlevel)
3068                     expect(")");
3069                 str.len -= spc;
3070                 tok_str_add(&str, 0);
3071                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
3072                 sa1->d = str.str;
3073                 sa = sa->next;
3074                 if (tok == ')') {
3075                     /* special case for gcc var args: add an empty
3076                        var arg argument if it is omitted */
3077                     if (sa && sa->type.t && gnu_ext)
3078                         goto empty_arg;
3079                     break;
3080                 }
3081                 if (tok != ',')
3082                     expect(",");
3083             }
3084             if (sa) {
3085                 tcc_error("macro '%s' used with too few args",
3086                       get_tok_str(s->v, 0));
3087             }
3088
3089             parse_flags = saved_parse_flags;
3090
3091             /* now subst each arg */
3092             mstr = macro_arg_subst(nested_list, mstr, args);
3093             /* free memory */
3094             sa = args;
3095             while (sa) {
3096                 sa1 = sa->prev;
3097                 tok_str_free(sa->d);
3098                 sym_free(sa);
3099                 sa = sa1;
3100             }
3101         }
3102
3103         sym_push2(nested_list, s->v, 0, 0);
3104         parse_flags = saved_parse_flags;
3105         macro_subst(tok_str, nested_list, mstr, can_read_stream);
3106
3107         /* pop nested defined symbol */
3108         sa1 = *nested_list;
3109         *nested_list = sa1->prev;
3110         sym_free(sa1);
3111         if (mstr != s->d)
3112             tok_str_free(mstr);
3113     }
3114     return 0;
3115 }
3116
3117 int paste_tokens(int t1, CValue *v1, int t2, CValue *v2)
3118 {
3119     CString cstr;
3120     int n;
3121
3122     cstr_new(&cstr);
3123     if (t1 != TOK_PLCHLDR)
3124         cstr_cat(&cstr, get_tok_str(t1, v1));
3125     n = cstr.size;
3126     if (t2 != TOK_PLCHLDR)
3127         cstr_cat(&cstr, get_tok_str(t2, v2));
3128     cstr_ccat(&cstr, '\0');
3129
3130     tcc_open_bf(tcc_state, ":paste:", cstr.size);
3131     memcpy(file->buffer, cstr.data, cstr.size);
3132     for (;;) {
3133         next_nomacro1();
3134         if (0 == *file->buf_ptr)
3135             break;
3136         if (is_space(tok))
3137             continue;
3138         tcc_warning("pasting <%.*s> and <%s> does not give a valid preprocessing token",
3139             n, cstr.data, (char*)cstr.data + n);
3140         break;
3141     }
3142     tcc_close();
3143
3144     //printf("paste <%s>\n", (char*)cstr.data);
3145     cstr_free(&cstr);
3146     return 0;
3147 }
3148
3149 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
3150    return the resulting string (which must be freed). */
3151 static inline int *macro_twosharps(const int *ptr0)
3152 {
3153     int t;
3154     CValue cval;
3155     TokenString macro_str1;
3156     int start_of_nosubsts = -1;
3157     const int *ptr;
3158
3159     /* we search the first '##' */
3160     for (ptr = ptr0;;) {
3161         TOK_GET(&t, &ptr, &cval);
3162         if (t == TOK_TWOSHARPS)
3163             break;
3164         if (t == 0)
3165             return NULL;
3166     }
3167
3168     tok_str_new(&macro_str1);
3169
3170     //tok_print(" $$$", ptr0);
3171     for (ptr = ptr0;;) {
3172         TOK_GET(&t, &ptr, &cval);
3173         if (t == 0)
3174             break;
3175         if (t == TOK_TWOSHARPS)
3176             continue;
3177         while (*ptr == TOK_TWOSHARPS) {
3178             int t1; CValue cv1;
3179             /* given 'a##b', remove nosubsts preceding 'a' */
3180             if (start_of_nosubsts >= 0)
3181                 macro_str1.len = start_of_nosubsts;
3182             /* given 'a##b', remove nosubsts preceding 'b' */
3183             while ((t1 = *++ptr) == TOK_NOSUBST)
3184                 ;
3185             if (t1 && t1 != TOK_TWOSHARPS
3186                 && t1 != ':') /* 'a##:' don't build a new token */
3187             {
3188                 TOK_GET(&t1, &ptr, &cv1);
3189                 if (t != TOK_PLCHLDR || t1 != TOK_PLCHLDR) {
3190                     paste_tokens(t, &cval, t1, &cv1);
3191                     t = tok, cval = tokc;
3192                 }
3193             }
3194         }
3195         if (t == TOK_NOSUBST) {
3196             if (start_of_nosubsts < 0)
3197                 start_of_nosubsts = macro_str1.len;
3198         } else {
3199             start_of_nosubsts = -1;
3200         }
3201         tok_str_add2(&macro_str1, t, &cval);
3202     }
3203     tok_str_add(&macro_str1, 0);
3204     //tok_print(" ###", macro_str1.str);
3205     return macro_str1.str;
3206 }
3207
3208 /* do macro substitution of macro_str and add result to
3209    (tok_str,tok_len). 'nested_list' is the list of all macros we got
3210    inside to avoid recursing. */
3211 static void macro_subst(
3212     TokenString *tok_str,
3213     Sym **nested_list,
3214     const int *macro_str,
3215     int can_read_stream
3216     )
3217 {
3218     Sym *s;
3219     const int *ptr;
3220     int t, spc, nosubst;
3221     CValue cval;
3222     int *macro_str1 = NULL;
3223
3224     /* first scan for '##' operator handling */
3225     ptr = macro_str;
3226     spc = nosubst = 0;
3227
3228     /* first scan for '##' operator handling */
3229     if (can_read_stream) {
3230         macro_str1 = macro_twosharps(ptr);
3231         if (macro_str1)
3232             ptr = macro_str1;
3233     }
3234
3235     while (1) {
3236         TOK_GET(&t, &ptr, &cval);
3237         if (t == 0)
3238             break;
3239
3240         if (t >= TOK_IDENT && 0 == nosubst) {
3241             s = define_find(t);
3242             if (s == NULL)
3243                 goto no_subst;
3244
3245             /* if nested substitution, do nothing */
3246             if (sym_find2(*nested_list, t)) {
3247                 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
3248                 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3249                 goto no_subst;
3250             }
3251
3252             {
3253                 TokenString str;
3254                 str.str = (int*)ptr;
3255                 begin_macro(&str, 2);
3256
3257                 tok = t;
3258                 macro_subst_tok(tok_str, nested_list, s, can_read_stream);
3259
3260                 if (str.alloc == 3) {
3261                     /* already finished by reading function macro arguments */
3262                     break;
3263                 }
3264
3265                 ptr = macro_ptr;
3266                 end_macro ();
3267             }
3268
3269             spc = tok_str->len && is_space(tok_str->str[tok_str->len-1]);
3270
3271         } else {
3272
3273             if (t == '\\' && !(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
3274                 tcc_error("stray '\\' in program");
3275
3276 no_subst:
3277             if (!check_space(t, &spc))
3278                 tok_str_add2(tok_str, t, &cval);
3279             nosubst = 0;
3280             if (t == TOK_NOSUBST)
3281                 nosubst = 1;
3282         }
3283     }
3284     if (macro_str1)
3285         tok_str_free(macro_str1);
3286
3287 }
3288
3289 /* return next token with macro substitution */
3290 ST_FUNC void next(void)
3291 {
3292  redo:
3293     if (parse_flags & PARSE_FLAG_SPACES)
3294         next_nomacro_spc();
3295     else
3296         next_nomacro();
3297
3298     if (macro_ptr) {
3299         if (tok == TOK_NOSUBST || tok == TOK_PLCHLDR) {
3300         /* discard preprocessor markers */
3301             goto redo;
3302         } else if (tok == 0) {
3303             /* end of macro or unget token string */
3304             end_macro();
3305             goto redo;
3306         }
3307     } else if (tok >= TOK_IDENT && (parse_flags & PARSE_FLAG_PREPROCESS)) {
3308         Sym *s;
3309         /* if reading from file, try to substitute macros */
3310         s = define_find(tok);
3311         if (s) {
3312             static TokenString str; /* using static string for speed */
3313             Sym *nested_list = NULL;
3314             tok_str_new(&str);
3315             nested_list = NULL;
3316             macro_subst_tok(&str, &nested_list, s, 1);
3317             tok_str_add(&str, 0);
3318             begin_macro(&str, 0);
3319             goto redo;
3320         }
3321     }
3322     /* convert preprocessor tokens into C tokens */
3323     if (tok == TOK_PPNUM) {
3324         if  (parse_flags & PARSE_FLAG_TOK_NUM)
3325             parse_number((char *)tokc.cstr->data);
3326     } else if (tok == TOK_PPSTR) {
3327         if (parse_flags & PARSE_FLAG_TOK_STR)
3328             parse_string((char *)tokc.cstr->data, tokc.cstr->size - 1);
3329     }
3330 }
3331
3332 /* push back current token and set current token to 'last_tok'. Only
3333    identifier case handled for labels. */
3334 ST_INLN void unget_tok(int last_tok)
3335 {
3336     TokenString *str = tcc_malloc(sizeof *str);
3337     tok_str_new(str);
3338     tok_str_add2(str, tok, &tokc);
3339     tok_str_add(str, 0);
3340     begin_macro(str, 1);
3341     tok = last_tok;
3342 }
3343
3344 /* better than nothing, but needs extension to handle '-E' option
3345    correctly too */
3346 ST_FUNC void preprocess_init(TCCState *s1)
3347 {
3348     s1->include_stack_ptr = s1->include_stack;
3349     /* XXX: move that before to avoid having to initialize
3350        file->ifdef_stack_ptr ? */
3351     s1->ifdef_stack_ptr = s1->ifdef_stack;
3352     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3353
3354     pvtop = vtop = vstack - 1;
3355     s1->pack_stack[0] = 0;
3356     s1->pack_stack_ptr = s1->pack_stack;
3357
3358     isidnum_table['$' - CH_EOF] =
3359         tcc_state->dollars_in_identifiers ? IS_ID : 0;
3360 }
3361
3362 ST_FUNC void preprocess_new(void)
3363 {
3364     int i, c;
3365     const char *p, *r;
3366
3367     /* init isid table */
3368     for(i = CH_EOF; i<256; i++)
3369         isidnum_table[i - CH_EOF]
3370             = is_space(i) ? IS_SPC
3371             : isid(i) ? IS_ID
3372             : isnum(i) ? IS_NUM
3373             : 0;
3374
3375     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3376
3377     tok_ident = TOK_IDENT;
3378     p = tcc_keywords;
3379     while (*p) {
3380         r = p;
3381         for(;;) {
3382             c = *r++;
3383             if (c == '\0')
3384                 break;
3385         }
3386         tok_alloc(p, r - p - 1);
3387         p = r;
3388     }
3389 }
3390
3391 ST_FUNC void preprocess_delete(void)
3392 {
3393     int i, n;
3394
3395     /* free -D and compiler defines */
3396     free_defines(NULL);
3397
3398     /* cleanup from error/setjmp */
3399     while (macro_stack)
3400         end_macro();
3401     macro_ptr = NULL;
3402
3403     /* free tokens */
3404     n = tok_ident - TOK_IDENT;
3405     for(i = 0; i < n; i++)
3406         tcc_free(table_ident[i]);
3407     tcc_free(table_ident);
3408     table_ident = NULL;
3409 }
3410
3411 /* Preprocess the current file */
3412 ST_FUNC int tcc_preprocess(TCCState *s1)
3413 {
3414     BufferedFile **iptr;
3415     int token_seen, spcs, level;
3416
3417     preprocess_init(s1);
3418     ch = file->buf_ptr[0];
3419     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3420     parse_flags = PARSE_FLAG_PREPROCESS
3421                 | (parse_flags & PARSE_FLAG_ASM_FILE)
3422                 | PARSE_FLAG_LINEFEED
3423                 | PARSE_FLAG_SPACES
3424                 | PARSE_FLAG_ACCEPT_STRAYS
3425                 ;
3426
3427 #ifdef PP_BENCH
3428     do next(); while (tok != TOK_EOF); return 0;
3429 #endif
3430
3431     token_seen = spcs = 0;
3432     pp_line(s1, file, 0);
3433
3434     for (;;) {
3435         iptr = s1->include_stack_ptr;
3436         next();
3437         if (tok == TOK_EOF)
3438             break;
3439         level = s1->include_stack_ptr - iptr;
3440         if (level) {
3441             if (level > 0)
3442                 pp_line(s1, *iptr, 0);
3443             pp_line(s1, file, level);
3444         }
3445
3446         if (0 == token_seen) {
3447             if (tok == ' ') {
3448                 ++spcs;
3449                 continue;
3450             }
3451             if (tok == TOK_LINEFEED) {
3452                 spcs = 0;
3453                 continue;
3454             }
3455             pp_line(s1, file, 0);
3456             while (spcs)
3457                 fputs(" ", s1->ppfp), --spcs;
3458             token_seen = 1;
3459
3460         } else if (tok == TOK_LINEFEED) {
3461             ++file->line_ref;
3462             token_seen = 0;
3463         }
3464
3465         fputs(get_tok_str(tok, &tokc), s1->ppfp);
3466     }
3467
3468     return 0;
3469 }