tccpp.c

   1 /*
   2  *  TCC - Tiny C Compiler
   3  *
   4  *  Copyright (c) 2001-2004 Fabrice Bellard
   5  *
   6  * This library is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2 of the License, or (at your option) any later version.
  10  *
  11  * This library is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with this library; if not, write to the Free Software
  18  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  19  */
  20
  21 #include "tcc.h"
  22
  23 /********************************************************/
  24 /* global variables */
  25
  26 ST_DATA int tok_flags;
  27 ST_DATA int parse_flags;
  28
  29 ST_DATA struct BufferedFile *file;
  30 ST_DATA int ch, tok;
  31 ST_DATA CValue tokc;
  32 ST_DATA const int *macro_ptr;
  33 ST_DATA CString tokcstr; /* current parsed string, if any */
  34
  35 /* display benchmark infos */
  36 ST_DATA int total_lines;
  37 ST_DATA int total_bytes;
  38 ST_DATA int tok_ident;
  39 ST_DATA TokenSym **table_ident;
  40
  41 /* ------------------------------------------------------------------------- */
  42
  43 static TokenSym *hash_ident[TOK_HASH_SIZE];
  44 static char token_buf[STRING_MAX_SIZE + 1];
  45 static unsigned char isidnum_table[256 - CH_EOF];
  46 /* isidnum_table flags: */
  47 #define IS_SPC 1
  48 #define IS_ID  2
  49 #define IS_NUM 4
  50
  51 static TokenString *macro_stack;
  52
  53 static const char tcc_keywords[] =
  54 #define DEF(id, str) str "\0"
  55 #include "tcctok.h"
  56 #undef DEF
  57 ;
  58
  59 /* WARNING: the content of this string encodes token numbers */
  60 static const unsigned char tok_two_chars[] =
  61 /* outdated -- gr
  62     "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
  63     "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
  64 */{
  65     '<','=', TOK_LE,
  66     '>','=', TOK_GE,
  67     '!','=', TOK_NE,
  68     '&','&', TOK_LAND,
  69     '|','|', TOK_LOR,
  70     '+','+', TOK_INC,
  71     '-','-', TOK_DEC,
  72     '=','=', TOK_EQ,
  73     '<','<', TOK_SHL,
  74     '>','>', TOK_SAR,
  75     '+','=', TOK_A_ADD,
  76     '-','=', TOK_A_SUB,
  77     '*','=', TOK_A_MUL,
  78     '/','=', TOK_A_DIV,
  79     '%','=', TOK_A_MOD,
  80     '&','=', TOK_A_AND,
  81     '^','=', TOK_A_XOR,
  82     '|','=', TOK_A_OR,
  83     '-','>', TOK_ARROW,
  84     '.','.', 0xa8, // C++ token ?
  85     '#','#', TOK_TWOSHARPS,
  86     0
  87 };
  88
  89 static void next_nomacro_spc(void);
  90
  91 ST_FUNC void skip(int c)
  92 {
  93     if (tok != c)
  94         tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
  95     next();
  96 }
  97
  98 ST_FUNC void expect(const char *msg)
  99 {
 100     tcc_error("%s expected", msg);
 101 }
 102
 103 ST_FUNC void begin_macro(TokenString *str, int alloc)
 104 {
 105     str->alloc = alloc;
 106     str->prev = macro_stack;
 107     str->prev_ptr = macro_ptr;
 108     macro_ptr = str->str;
 109     macro_stack = str;
 110 }
 111
 112 ST_FUNC void end_macro(void)
 113 {
 114     TokenString *str = macro_stack;
 115     macro_stack = str->prev;
 116     macro_ptr = str->prev_ptr;
 117     if (str->alloc == 2) {
 118         str->alloc = 3; /* just mark as finished */
 119     } else {
 120         tok_str_free(str->str);
 121         if (str->alloc == 1)
 122             tcc_free(str);
 123     }
 124 }
 125
 126 /* ------------------------------------------------------------------------- */
 127 /* CString handling */
 128 static void cstr_realloc(CString *cstr, int new_size)
 129 {
 130     int size;
 131     void *data;
 132
 133     size = cstr->size_allocated;
 134     if (size == 0)
 135         size = 8; /* no need to allocate a too small first string */
 136     while (size < new_size)
 137         size = size * 2;
 138     data = tcc_realloc(cstr->data_allocated, size);
 139     cstr->data_allocated = data;
 140     cstr->size_allocated = size;
 141     cstr->data = data;
 142 }
 143
 144 /* add a byte */
 145 ST_FUNC void cstr_ccat(CString *cstr, int ch)
 146 {
 147     int size;
 148     size = cstr->size + 1;
 149     if (size > cstr->size_allocated)
 150         cstr_realloc(cstr, size);
 151     ((unsigned char *)cstr->data)[size - 1] = ch;
 152     cstr->size = size;
 153 }
 154
 155 ST_FUNC void cstr_cat(CString *cstr, const char *str)
 156 {
 157     int c;
 158     for(;;) {
 159         c = *str;
 160         if (c == '\0')
 161             break;
 162         cstr_ccat(cstr, c);
 163         str++;
 164     }
 165 }
 166
 167 /* add a wide char */
 168 ST_FUNC void cstr_wccat(CString *cstr, int ch)
 169 {
 170     int size;
 171     size = cstr->size + sizeof(nwchar_t);
 172     if (size > cstr->size_allocated)
 173         cstr_realloc(cstr, size);
 174     *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
 175     cstr->size = size;
 176 }
 177
 178 ST_FUNC void cstr_new(CString *cstr)
 179 {
 180     memset(cstr, 0, sizeof(CString));
 181 }
 182
 183 /* free string and reset it to NULL */
 184 ST_FUNC void cstr_free(CString *cstr)
 185 {
 186     tcc_free(cstr->data_allocated);
 187     cstr_new(cstr);
 188 }
 189
 190 /* reset string to empty */
 191 ST_FUNC void cstr_reset(CString *cstr)
 192 {
 193     cstr->size = 0;
 194 }
 195
 196 /* XXX: unicode ? */
 197 static void add_char(CString *cstr, int c)
 198 {
 199     if (c == '\'' || c == '\"' || c == '\\') {
 200         /* XXX: could be more precise if char or string */
 201         cstr_ccat(cstr, '\\');
 202     }
 203     if (c >= 32 && c <= 126) {
 204         cstr_ccat(cstr, c);
 205     } else {
 206         cstr_ccat(cstr, '\\');
 207         if (c == '\n') {
 208             cstr_ccat(cstr, 'n');
 209         } else {
 210             cstr_ccat(cstr, '0' + ((c >> 6) & 7));
 211             cstr_ccat(cstr, '0' + ((c >> 3) & 7));
 212             cstr_ccat(cstr, '0' + (c & 7));
 213         }
 214     }
 215 }
 216
 217 /* ------------------------------------------------------------------------- */
 218 /* allocate a new token */
 219 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
 220 {
 221     TokenSym *ts, **ptable;
 222     int i;
 223
 224     if (tok_ident >= SYM_FIRST_ANOM)
 225         tcc_error("memory full (symbols)");
 226
 227     /* expand token table if needed */
 228     i = tok_ident - TOK_IDENT;
 229     if ((i % TOK_ALLOC_INCR) == 0) {
 230         ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
 231         table_ident = ptable;
 232     }
 233
 234     ts = tcc_malloc(sizeof(TokenSym) + len);
 235     table_ident[i] = ts;
 236     ts->tok = tok_ident++;
 237     ts->sym_define = NULL;
 238     ts->sym_label = NULL;
 239     ts->sym_struct = NULL;
 240     ts->sym_identifier = NULL;
 241     ts->len = len;
 242     ts->hash_next = NULL;
 243     memcpy(ts->str, str, len);
 244     ts->str[len] = '\0';
 245     *pts = ts;
 246     return ts;
 247 }
 248
 249 #define TOK_HASH_INIT 1
 250 #define TOK_HASH_FUNC(h, c) ((h) * 263 + (c))
 251
 252 /* find a token and add it if not found */
 253 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
 254 {
 255     TokenSym *ts, **pts;
 256     int i;
 257     unsigned int h;
 258
 259     h = TOK_HASH_INIT;
 260     for(i=0;i<len;i++)
 261         h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
 262     h &= (TOK_HASH_SIZE - 1);
 263
 264     pts = &hash_ident[h];
 265     for(;;) {
 266         ts = *pts;
 267         if (!ts)
 268             break;
 269         if (ts->len == len && !memcmp(ts->str, str, len))
 270             return ts;
 271         pts = &(ts->hash_next);
 272     }
 273     return tok_alloc_new(pts, str, len);
 274 }
 275
 276 /* XXX: buffer overflow */
 277 /* XXX: float tokens */
 278 ST_FUNC const char *get_tok_str(int v, CValue *cv)
 279 {
 280     static char buf[STRING_MAX_SIZE + 1];
 281     static CString cstr_buf;
 282     CString *cstr;
 283     char *p;
 284     int i, len;
 285
 286     /* NOTE: to go faster, we give a fixed buffer for small strings */
 287     cstr_reset(&cstr_buf);
 288     cstr_buf.data = buf;
 289     cstr_buf.size_allocated = sizeof(buf);
 290     p = buf;
 291
 292     switch(v) {
 293     case TOK_CINT:
 294     case TOK_CUINT:
 295         /* XXX: not quite exact, but only useful for testing */
 296         sprintf(p, "%llu", (unsigned long long)cv->i);
 297         break;
 298     case TOK_CLLONG:
 299     case TOK_CULLONG:
 300         /* XXX: not quite exact, but only useful for testing  */
 301 #ifdef _WIN32
 302         sprintf(p, "%u", (unsigned)cv->i);
 303 #else
 304         sprintf(p, "%llu", (unsigned long long)cv->i);
 305 #endif
 306         break;
 307     case TOK_LCHAR:
 308         cstr_ccat(&cstr_buf, 'L');
 309     case TOK_CCHAR:
 310         cstr_ccat(&cstr_buf, '\'');
 311         add_char(&cstr_buf, cv->i);
 312         cstr_ccat(&cstr_buf, '\'');
 313         cstr_ccat(&cstr_buf, '\0');
 314         break;
 315     case TOK_PPNUM:
 316     case TOK_PPSTR:
 317         return (char*)cv->cstr->data;
 318     case TOK_LSTR:
 319         cstr_ccat(&cstr_buf, 'L');
 320     case TOK_STR:
 321         cstr = cv->cstr;
 322         cstr_ccat(&cstr_buf, '\"');
 323         if (v == TOK_STR) {
 324             len = cstr->size - 1;
 325             for(i=0;i<len;i++)
 326                 add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 327         } else {
 328             len = (cstr->size / sizeof(nwchar_t)) - 1;
 329             for(i=0;i<len;i++)
 330                 add_char(&cstr_buf, ((nwchar_t *)cstr->data)[i]);
 331         }
 332         cstr_ccat(&cstr_buf, '\"');
 333         cstr_ccat(&cstr_buf, '\0');
 334         break;
 335
 336     case TOK_CFLOAT:
 337         cstr_cat(&cstr_buf, "<float>");
 338         break;
 339     case TOK_CDOUBLE:
 340         cstr_cat(&cstr_buf, "<double>");
 341         break;
 342     case TOK_CLDOUBLE:
 343         cstr_cat(&cstr_buf, "<long double>");
 344         break;
 345     case TOK_LINENUM:
 346         cstr_cat(&cstr_buf, "<linenumber>");
 347         break;
 348         //return NULL; /* should not happen */
 349
 350     /* above tokens have value, the ones below don't */
 351
 352     case TOK_LT:
 353         v = '<';
 354         goto addv;
 355     case TOK_GT:
 356         v = '>';
 357         goto addv;
 358     case TOK_DOTS:
 359         return strcpy(p, "...");
 360     case TOK_A_SHL:
 361         return strcpy(p, "<<=");
 362     case TOK_A_SAR:
 363         return strcpy(p, ">>=");
 364     default:
 365         if (v < TOK_IDENT) {
 366             /* search in two bytes table */
 367             const unsigned char *q = tok_two_chars;
 368             while (*q) {
 369                 if (q[2] == v) {
 370                     *p++ = q[0];
 371                     *p++ = q[1];
 372                     *p = '\0';
 373                     return buf;
 374                 }
 375                 q += 3;
 376             }
 377         if (v >= 127) {
 378             sprintf(buf, "<%02x>", v);
 379             return buf;
 380         }
 381         addv:
 382             *p++ = v;
 383             *p = '\0';
 384         } else if (v < tok_ident) {
 385             return table_ident[v - TOK_IDENT]->str;
 386         } else if (v >= SYM_FIRST_ANOM) {
 387             /* special name for anonymous symbol */
 388             sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
 389         } else {
 390             /* should never happen */
 391             return NULL;
 392         }
 393         break;
 394     }
 395     return cstr_buf.data;
 396 }
 397
 398 /* return the current character, handling end of block if necessary
 399    (but not stray) */
 400 ST_FUNC int handle_eob(void)
 401 {
 402     BufferedFile *bf = file;
 403     int len;
 404
 405     /* only tries to read if really end of buffer */
 406     if (bf->buf_ptr >= bf->buf_end) {
 407         if (bf->fd != -1) {
 408 #if defined(PARSE_DEBUG)
 409             len = 1;
 410 #else
 411             len = IO_BUF_SIZE;
 412 #endif
 413             len = read(bf->fd, bf->buffer, len);
 414             if (len < 0)
 415                 len = 0;
 416         } else {
 417             len = 0;
 418         }
 419         total_bytes += len;
 420         bf->buf_ptr = bf->buffer;
 421         bf->buf_end = bf->buffer + len;
 422         *bf->buf_end = CH_EOB;
 423     }
 424     if (bf->buf_ptr < bf->buf_end) {
 425         return bf->buf_ptr[0];
 426     } else {
 427         bf->buf_ptr = bf->buf_end;
 428         return CH_EOF;
 429     }
 430 }
 431
 432 /* read next char from current input file and handle end of input buffer */
 433 ST_INLN void inp(void)
 434 {
 435     ch = *(++(file->buf_ptr));
 436     /* end of buffer/file handling */
 437     if (ch == CH_EOB)
 438         ch = handle_eob();
 439 }
 440
 441 /* handle '\[\r]\n' */
 442 static int handle_stray_noerror(void)
 443 {
 444     while (ch == '\\') {
 445         inp();
 446         if (ch == '\n') {
 447             file->line_num++;
 448             inp();
 449         } else if (ch == '\r') {
 450             inp();
 451             if (ch != '\n')
 452                 goto fail;
 453             file->line_num++;
 454             inp();
 455         } else {
 456         fail:
 457             return 1;
 458         }
 459     }
 460     return 0;
 461 }
 462
 463 static void handle_stray(void)
 464 {
 465     if (handle_stray_noerror())
 466         tcc_error("stray '\\' in program");
 467 }
 468
 469 /* skip the stray and handle the \\n case. Output an error if
 470    incorrect char after the stray */
 471 static int handle_stray1(uint8_t *p)
 472 {
 473     int c;
 474
 475     file->buf_ptr = p;
 476     if (p >= file->buf_end) {
 477         c = handle_eob();
 478         if (c != '\\')
 479             return c;
 480         p = file->buf_ptr;
 481     }
 482     ch = *p;
 483     if (handle_stray_noerror()) {
 484         if (!(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
 485             tcc_error("stray '\\' in program");
 486         *--file->buf_ptr = '\\';
 487     }
 488     p = file->buf_ptr;
 489     c = *p;
 490     return c;
 491 }
 492
 493 /* handle just the EOB case, but not stray */
 494 #define PEEKC_EOB(c, p)\
 495 {\
 496     p++;\
 497     c = *p;\
 498     if (c == '\\') {\
 499         file->buf_ptr = p;\
 500         c = handle_eob();\
 501         p = file->buf_ptr;\
 502     }\
 503 }
 504
 505 /* handle the complicated stray case */
 506 #define PEEKC(c, p)\
 507 {\
 508     p++;\
 509     c = *p;\
 510     if (c == '\\') {\
 511         c = handle_stray1(p);\
 512         p = file->buf_ptr;\
 513     }\
 514 }
 515
 516 /* input with '\[\r]\n' handling. Note that this function cannot
 517    handle other characters after '\', so you cannot call it inside
 518    strings or comments */
 519 ST_FUNC void minp(void)
 520 {
 521     inp();
 522     if (ch == '\\')
 523         handle_stray();
 524 }
 525
 526
 527 /* single line C++ comments */
 528 static uint8_t *parse_line_comment(uint8_t *p)
 529 {
 530     int c;
 531
 532     p++;
 533     for(;;) {
 534         c = *p;
 535     redo:
 536         if (c == '\n' || c == CH_EOF) {
 537             break;
 538         } else if (c == '\\') {
 539             file->buf_ptr = p;
 540             c = handle_eob();
 541             p = file->buf_ptr;
 542             if (c == '\\') {
 543                 PEEKC_EOB(c, p);
 544                 if (c == '\n') {
 545                     file->line_num++;
 546                     PEEKC_EOB(c, p);
 547                 } else if (c == '\r') {
 548                     PEEKC_EOB(c, p);
 549                     if (c == '\n') {
 550                         file->line_num++;
 551                         PEEKC_EOB(c, p);
 552                     }
 553                 }
 554             } else {
 555                 goto redo;
 556             }
 557         } else {
 558             p++;
 559         }
 560     }
 561     return p;
 562 }
 563
 564 /* C comments */
 565 ST_FUNC uint8_t *parse_comment(uint8_t *p)
 566 {
 567     int c;
 568
 569     p++;
 570     for(;;) {
 571         /* fast skip loop */
 572         for(;;) {
 573             c = *p;
 574             if (c == '\n' || c == '*' || c == '\\')
 575                 break;
 576             p++;
 577             c = *p;
 578             if (c == '\n' || c == '*' || c == '\\')
 579                 break;
 580             p++;
 581         }
 582         /* now we can handle all the cases */
 583         if (c == '\n') {
 584             file->line_num++;
 585             p++;
 586         } else if (c == '*') {
 587             p++;
 588             for(;;) {
 589                 c = *p;
 590                 if (c == '*') {
 591                     p++;
 592                 } else if (c == '/') {
 593                     goto end_of_comment;
 594                 } else if (c == '\\') {
 595                     file->buf_ptr = p;
 596                     c = handle_eob();
 597                     p = file->buf_ptr;
 598                     if (c == CH_EOF)
 599                         tcc_error("unexpected end of file in comment");
 600                     if (c == '\\') {
 601                         /* skip '\[\r]\n', otherwise just skip the stray */
 602                         while (c == '\\') {
 603                             PEEKC_EOB(c, p);
 604                             if (c == '\n') {
 605                                 file->line_num++;
 606                                 PEEKC_EOB(c, p);
 607                             } else if (c == '\r') {
 608                                 PEEKC_EOB(c, p);
 609                                 if (c == '\n') {
 610                                     file->line_num++;
 611                                     PEEKC_EOB(c, p);
 612                                 }
 613                             } else {
 614                                 goto after_star;
 615                             }
 616                         }
 617                     }
 618                 } else {
 619                     break;
 620                 }
 621             }
 622         after_star: ;
 623         } else {
 624             /* stray, eob or eof */
 625             file->buf_ptr = p;
 626             c = handle_eob();
 627             p = file->buf_ptr;
 628             if (c == CH_EOF) {
 629                 tcc_error("unexpected end of file in comment");
 630             } else if (c == '\\') {
 631                 p++;
 632             }
 633         }
 634     }
 635  end_of_comment:
 636     p++;
 637     return p;
 638 }
 639
 640 #define cinp minp
 641
 642 static inline void skip_spaces(void)
 643 {
 644     while (isidnum_table[ch - CH_EOF] & IS_SPC)
 645         cinp();
 646 }
 647
 648 static inline int check_space(int t, int *spc)
 649 {
 650     if (t < 256 && (isidnum_table[t - CH_EOF] & IS_SPC)) {
 651         if (*spc)
 652             return 1;
 653         *spc = 1;
 654     } else
 655         *spc = 0;
 656     return 0;
 657 }
 658
 659 /* parse a string without interpreting escapes */
 660 static uint8_t *parse_pp_string(uint8_t *p,
 661                                 int sep, CString *str)
 662 {
 663     int c;
 664     p++;
 665     for(;;) {
 666         c = *p;
 667         if (c == sep) {
 668             break;
 669         } else if (c == '\\') {
 670             file->buf_ptr = p;
 671             c = handle_eob();
 672             p = file->buf_ptr;
 673             if (c == CH_EOF) {
 674             unterminated_string:
 675                 /* XXX: indicate line number of start of string */
 676                 tcc_error("missing terminating %c character", sep);
 677             } else if (c == '\\') {
 678                 /* escape : just skip \[\r]\n */
 679                 PEEKC_EOB(c, p);
 680                 if (c == '\n') {
 681                     file->line_num++;
 682                     p++;
 683                 } else if (c == '\r') {
 684                     PEEKC_EOB(c, p);
 685                     if (c != '\n')
 686                         expect("'\n' after '\r'");
 687                     file->line_num++;
 688                     p++;
 689                 } else if (c == CH_EOF) {
 690                     goto unterminated_string;
 691                 } else {
 692                     if (str) {
 693                         cstr_ccat(str, '\\');
 694                         cstr_ccat(str, c);
 695                     }
 696                     p++;
 697                 }
 698             }
 699         } else if (c == '\n') {
 700             file->line_num++;
 701             goto add_char;
 702         } else if (c == '\r') {
 703             PEEKC_EOB(c, p);
 704             if (c != '\n') {
 705                 if (str)
 706                     cstr_ccat(str, '\r');
 707             } else {
 708                 file->line_num++;
 709                 goto add_char;
 710             }
 711         } else {
 712         add_char:
 713             if (str)
 714                 cstr_ccat(str, c);
 715             p++;
 716         }
 717     }
 718     p++;
 719     return p;
 720 }
 721
 722 /* skip block of text until #else, #elif or #endif. skip also pairs of
 723    #if/#endif */
 724 static void preprocess_skip(void)
 725 {
 726     int a, start_of_line, c, in_warn_or_error;
 727     uint8_t *p;
 728
 729     p = file->buf_ptr;
 730     a = 0;
 731 redo_start:
 732     start_of_line = 1;
 733     in_warn_or_error = 0;
 734     for(;;) {
 735     redo_no_start:
 736         c = *p;
 737         switch(c) {
 738         case ' ':
 739         case '\t':
 740         case '\f':
 741         case '\v':
 742         case '\r':
 743             p++;
 744             goto redo_no_start;
 745         case '\n':
 746             file->line_num++;
 747             p++;
 748             goto redo_start;
 749         case '\\':
 750             file->buf_ptr = p;
 751             c = handle_eob();
 752             if (c == CH_EOF) {
 753                 expect("#endif");
 754             } else if (c == '\\') {
 755                 ch = file->buf_ptr[0];
 756                 handle_stray_noerror();
 757             }
 758             p = file->buf_ptr;
 759             goto redo_no_start;
 760         /* skip strings */
 761         case '\"':
 762         case '\'':
 763             if (in_warn_or_error)
 764                 goto _default;
 765             p = parse_pp_string(p, c, NULL);
 766             break;
 767         /* skip comments */
 768         case '/':
 769             if (in_warn_or_error)
 770                 goto _default;
 771             file->buf_ptr = p;
 772             ch = *p;
 773             minp();
 774             p = file->buf_ptr;
 775             if (ch == '*') {
 776                 p = parse_comment(p);
 777             } else if (ch == '/') {
 778                 p = parse_line_comment(p);
 779             }
 780             break;
 781         case '#':
 782             p++;
 783             if (start_of_line) {
 784                 file->buf_ptr = p;
 785                 next_nomacro();
 786                 p = file->buf_ptr;
 787                 if (a == 0 &&
 788                     (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
 789                     goto the_end;
 790                 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
 791                     a++;
 792                 else if (tok == TOK_ENDIF)
 793                     a--;
 794                 else if( tok == TOK_ERROR || tok == TOK_WARNING)
 795                     in_warn_or_error = 1;
 796                 else if (tok == TOK_LINEFEED)
 797                     goto redo_start;
 798             } else if (parse_flags & PARSE_FLAG_ASM_FILE)
 799                 p = parse_line_comment(p);
 800             break;
 801 _default:
 802         default:
 803             p++;
 804             break;
 805         }
 806         start_of_line = 0;
 807     }
 808  the_end: ;
 809     file->buf_ptr = p;
 810 }
 811
 812 /* ParseState handling */
 813
 814 /* XXX: currently, no include file info is stored. Thus, we cannot display
 815    accurate messages if the function or data definition spans multiple
 816    files */
 817
 818 /* save current parse state in 's' */
 819 ST_FUNC void save_parse_state(ParseState *s)
 820 {
 821     s->line_num = file->line_num;
 822     s->macro_ptr = macro_ptr;
 823     s->tok = tok;
 824     s->tokc = tokc;
 825 }
 826
 827 /* restore parse state from 's' */
 828 ST_FUNC void restore_parse_state(ParseState *s)
 829 {
 830     file->line_num = s->line_num;
 831     macro_ptr = s->macro_ptr;
 832     tok = s->tok;
 833     tokc = s->tokc;
 834 }
 835
 836 /* return the number of additional 'ints' necessary to store the
 837    token */
 838 static inline int tok_size(const int *p)
 839 {
 840     switch(*p) {
 841         /* 4 bytes */
 842     case TOK_CINT:
 843     case TOK_CUINT:
 844     case TOK_CCHAR:
 845     case TOK_LCHAR:
 846     case TOK_CFLOAT:
 847     case TOK_LINENUM:
 848         return 1 + 1;
 849     case TOK_STR:
 850     case TOK_LSTR:
 851     case TOK_PPNUM:
 852     case TOK_PPSTR:
 853         return 1 + ((sizeof(CString) + ((CString *)(p+1))->size + 3) >> 2);
 854     case TOK_CDOUBLE:
 855     case TOK_CLLONG:
 856     case TOK_CULLONG:
 857         return 1 + 2;
 858     case TOK_CLDOUBLE:
 859         return 1 + LDOUBLE_SIZE / 4;
 860     default:
 861         return 1 + 0;
 862     }
 863 }
 864
 865 /* token string handling */
 866
 867 ST_INLN void tok_str_new(TokenString *s)
 868 {
 869     s->str = NULL;
 870     s->len = 0;
 871     s->allocated_len = 0;
 872     s->last_line_num = -1;
 873 }
 874
 875 ST_FUNC void tok_str_free(int *str)
 876 {
 877     tcc_free(str);
 878 }
 879
 880 static int *tok_str_realloc(TokenString *s)
 881 {
 882     int *str, len;
 883
 884     if (s->allocated_len == 0) {
 885         len = 8;
 886     } else {
 887         len = s->allocated_len * 2;
 888     }
 889     str = tcc_realloc(s->str, len * sizeof(int));
 890     s->allocated_len = len;
 891     s->str = str;
 892     return str;
 893 }
 894
 895 ST_FUNC void tok_str_add(TokenString *s, int t)
 896 {
 897     int len, *str;
 898
 899     len = s->len;
 900     str = s->str;
 901     if (len >= s->allocated_len)
 902         str = tok_str_realloc(s);
 903     str[len++] = t;
 904     s->len = len;
 905 }
 906
 907 static void tok_str_add2(TokenString *s, int t, CValue *cv)
 908 {
 909     int len, *str;
 910
 911     len = s->len;
 912     str = s->str;
 913
 914     /* allocate space for worst case */
 915     if (len + TOK_MAX_SIZE > s->allocated_len)
 916         str = tok_str_realloc(s);
 917     str[len++] = t;
 918     switch(t) {
 919     case TOK_CINT:
 920     case TOK_CUINT:
 921     case TOK_CCHAR:
 922     case TOK_LCHAR:
 923     case TOK_CFLOAT:
 924     case TOK_LINENUM:
 925         str[len++] = cv->tab[0];
 926         break;
 927     case TOK_PPNUM:
 928     case TOK_PPSTR:
 929     case TOK_STR:
 930     case TOK_LSTR:
 931         {
 932             int nb_words;
 933             CString cstr;
 934
 935             nb_words = (sizeof(CString) + cv->cstr->size + 3) >> 2;
 936             while ((len + nb_words) > s->allocated_len)
 937                 str = tok_str_realloc(s);
 938             /* XXX: Insert the CString into the int array.
 939                It may end up incorrectly aligned. */
 940             cstr.data = 0;
 941             cstr.size = cv->cstr->size;
 942             cstr.data_allocated = 0;
 943             cstr.size_allocated = cstr.size;
 944             memcpy(str + len, &cstr, sizeof(CString));
 945             memcpy((char *)(str + len) + sizeof(CString),
 946                    cv->cstr->data, cstr.size);
 947             len += nb_words;
 948         }
 949         break;
 950     case TOK_CDOUBLE:
 951     case TOK_CLLONG:
 952     case TOK_CULLONG:
 953 #if LDOUBLE_SIZE == 8
 954     case TOK_CLDOUBLE:
 955 #endif
 956         str[len++] = cv->tab[0];
 957         str[len++] = cv->tab[1];
 958         break;
 959 #if LDOUBLE_SIZE == 12
 960     case TOK_CLDOUBLE:
 961         str[len++] = cv->tab[0];
 962         str[len++] = cv->tab[1];
 963         str[len++] = cv->tab[2];
 964 #elif LDOUBLE_SIZE == 16
 965     case TOK_CLDOUBLE:
 966         str[len++] = cv->tab[0];
 967         str[len++] = cv->tab[1];
 968         str[len++] = cv->tab[2];
 969         str[len++] = cv->tab[3];
 970 #elif LDOUBLE_SIZE != 8
 971 #error add long double size support
 972 #endif
 973         break;
 974     default:
 975         break;
 976     }
 977     s->len = len;
 978 }
 979
 980 /* add the current parse token in token string 's' */
 981 ST_FUNC void tok_str_add_tok(TokenString *s)
 982 {
 983     CValue cval;
 984
 985     /* save line number info */
 986     if (file->line_num != s->last_line_num) {
 987         s->last_line_num = file->line_num;
 988         cval.i = s->last_line_num;
 989         tok_str_add2(s, TOK_LINENUM, &cval);
 990     }
 991     tok_str_add2(s, tok, &tokc);
 992 }
 993
 994 /* get a token from an integer array and increment pointer
 995    accordingly. we code it as a macro to avoid pointer aliasing. */
 996 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
 997 {
 998     const int *p = *pp;
 999     int n, *tab;
1000
1001     tab = cv->tab;
1002     switch(*t = *p++) {
1003     case TOK_CINT:
1004     case TOK_CUINT:
1005     case TOK_CCHAR:
1006     case TOK_LCHAR:
1007     case TOK_CFLOAT:
1008     case TOK_LINENUM:
1009         tab[0] = *p++;
1010         break;
1011     case TOK_STR:
1012     case TOK_LSTR:
1013     case TOK_PPNUM:
1014     case TOK_PPSTR:
1015         /* XXX: Illegal cast: the pointer p may not be correctly aligned! */
1016         cv->cstr = (CString *)p;
1017         cv->cstr->data = (char *)p + sizeof(CString);
1018         p += (sizeof(CString) + cv->cstr->size + 3) >> 2;
1019         break;
1020     case TOK_CDOUBLE:
1021     case TOK_CLLONG:
1022     case TOK_CULLONG:
1023         n = 2;
1024         goto copy;
1025     case TOK_CLDOUBLE:
1026 #if LDOUBLE_SIZE == 16
1027         n = 4;
1028 #elif LDOUBLE_SIZE == 12
1029         n = 3;
1030 #elif LDOUBLE_SIZE == 8
1031         n = 2;
1032 #else
1033 # error add long double size support
1034 #endif
1035     copy:
1036         do
1037             *tab++ = *p++;
1038         while (--n);
1039         break;
1040     default:
1041         break;
1042     }
1043     *pp = p;
1044 }
1045
1046 /* Calling this function is expensive, but it is not possible
1047    to read a token string backwards. */
1048 static int tok_last(const int *str0, const int *str1)
1049 {
1050     const int *str = str0;
1051     int tok = 0;
1052     CValue cval;
1053
1054     while (str < str1)
1055         TOK_GET(&tok, &str, &cval);
1056     return tok;
1057 }
1058
1059 static int macro_is_equal(const int *a, const int *b)
1060 {
1061     char buf[STRING_MAX_SIZE + 1];
1062     CValue cv;
1063     int t;
1064     while (*a && *b) {
1065         TOK_GET(&t, &a, &cv);
1066         pstrcpy(buf, sizeof buf, get_tok_str(t, &cv));
1067         TOK_GET(&t, &b, &cv);
1068         if (strcmp(buf, get_tok_str(t, &cv)))
1069             return 0;
1070     }
1071     return !(*a || *b);
1072 }
1073
1074 static void pp_line(TCCState *s1, BufferedFile *f, int level)
1075 {
1076     int d = f->line_num - f->line_ref;
1077     if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_NONE
1078         || (level == 0 && f->line_ref && d < 8))
1079     {
1080         while (d > 0)
1081             fputs("\n", s1->ppfp), --d;
1082     }
1083     else
1084     if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_STD) {
1085         fprintf(s1->ppfp, "#line %d \"%s\"\n", f->line_num, f->filename);
1086     }
1087     else {
1088         fprintf(s1->ppfp, "# %d \"%s\"%s\n", f->line_num, f->filename,
1089             level > 0 ? " 1" : level < 0 ? " 2" : "");
1090     }
1091     f->line_ref = f->line_num;
1092 }
1093
1094 static void tok_print(const char *msg, const int *str)
1095 {
1096     FILE *pr = tcc_state->ppfp;
1097     int t;
1098     CValue cval;
1099
1100     fprintf(pr, "%s ", msg);
1101     while (str) {
1102         TOK_GET(&t, &str, &cval);
1103         if (!t)
1104             break;
1105         fprintf(pr,"%s", get_tok_str(t, &cval));
1106     }
1107     fprintf(pr, "\n");
1108 }
1109
1110 static int define_print_prepared(Sym *s)
1111 {
1112     if (!s || !tcc_state->ppfp || tcc_state->dflag == 0)
1113         return 0;
1114
1115     if (s->v < TOK_IDENT || s->v >= tok_ident)
1116         return 0;
1117
1118     if (file) {
1119         file->line_num--;
1120         pp_line(tcc_state, file, 0);
1121         file->line_ref = ++file->line_num;
1122     }
1123     return 1;
1124 }
1125
1126 static void define_print(int v)
1127 {
1128     FILE *pr = tcc_state->ppfp;
1129     Sym *s, *a;
1130
1131     s = define_find(v);
1132     if (define_print_prepared(s) == 0)
1133         return;
1134
1135     fprintf(pr, "// #define %s", get_tok_str(v, NULL));
1136     if (s->type.t == MACRO_FUNC) {
1137         a = s->next;
1138         fprintf(pr,"(");
1139         if (a)
1140             for (;;) {
1141                 fprintf(pr,"%s", get_tok_str(a->v & ~SYM_FIELD, NULL));
1142                 if (!(a = a->next))
1143                     break;
1144                 fprintf(pr,",");
1145             }
1146         fprintf(pr,")");
1147     }
1148     tok_print("", s->d);
1149 }
1150
1151 static void undef_print(int v)
1152 {
1153     FILE *pr = tcc_state->ppfp;
1154     Sym *s;
1155
1156     s = define_find(v);
1157     if (define_print_prepared(s) == 0)
1158         return;
1159
1160     fprintf(pr, "// #undef %s\n", get_tok_str(s->v, NULL));
1161 }
1162
1163 ST_FUNC void print_defines(void)
1164 {
1165     Sym *top = define_stack;
1166     while (top) {
1167         define_print(top->v);
1168         top = top->prev;
1169     }
1170 }
1171
1172 /* defines handling */
1173 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1174 {
1175     Sym *s;
1176
1177     s = define_find(v);
1178     if (s && !macro_is_equal(s->d, str))
1179         tcc_warning("%s redefined", get_tok_str(v, NULL));
1180
1181     s = sym_push2(&define_stack, v, macro_type, 0);
1182     s->d = str;
1183     s->next = first_arg;
1184     table_ident[v - TOK_IDENT]->sym_define = s;
1185 }
1186
1187 /* undefined a define symbol. Its name is just set to zero */
1188 ST_FUNC void define_undef(Sym *s)
1189 {
1190     int v = s->v;
1191     undef_print(v);
1192     if (v >= TOK_IDENT && v < tok_ident)
1193         table_ident[v - TOK_IDENT]->sym_define = NULL;
1194 }
1195
1196 ST_INLN Sym *define_find(int v)
1197 {
1198     v -= TOK_IDENT;
1199     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1200         return NULL;
1201     return table_ident[v]->sym_define;
1202 }
1203
1204 /* free define stack until top reaches 'b' */
1205 ST_FUNC void free_defines(Sym *b)
1206 {
1207     Sym *top, *top1;
1208     int v;
1209
1210     top = define_stack;
1211     while (top != b) {
1212         top1 = top->prev;
1213         /* do not free args or predefined defines */
1214         if (top->d)
1215             tok_str_free(top->d);
1216         v = top->v;
1217         if (v >= TOK_IDENT && v < tok_ident)
1218             table_ident[v - TOK_IDENT]->sym_define = NULL;
1219         sym_free(top);
1220         top = top1;
1221     }
1222     define_stack = b;
1223 }
1224
1225 /* label lookup */
1226 ST_FUNC Sym *label_find(int v)
1227 {
1228     v -= TOK_IDENT;
1229     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1230         return NULL;
1231     return table_ident[v]->sym_label;
1232 }
1233
1234 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1235 {
1236     Sym *s, **ps;
1237     s = sym_push2(ptop, v, 0, 0);
1238     s->r = flags;
1239     ps = &table_ident[v - TOK_IDENT]->sym_label;
1240     if (ptop == &global_label_stack) {
1241         /* modify the top most local identifier, so that
1242            sym_identifier will point to 's' when popped */
1243         while (*ps != NULL)
1244             ps = &(*ps)->prev_tok;
1245     }
1246     s->prev_tok = *ps;
1247     *ps = s;
1248     return s;
1249 }
1250
1251 /* pop labels until element last is reached. Look if any labels are
1252    undefined. Define symbols if '&&label' was used. */
1253 ST_FUNC void label_pop(Sym **ptop, Sym *slast)
1254 {
1255     Sym *s, *s1;
1256     for(s = *ptop; s != slast; s = s1) {
1257         s1 = s->prev;
1258         if (s->r == LABEL_DECLARED) {
1259             tcc_warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1260         } else if (s->r == LABEL_FORWARD) {
1261                 tcc_error("label '%s' used but not defined",
1262                       get_tok_str(s->v, NULL));
1263         } else {
1264             if (s->c) {
1265                 /* define corresponding symbol. A size of
1266                    1 is put. */
1267                 put_extern_sym(s, cur_text_section, s->jnext, 1);
1268             }
1269         }
1270         /* remove label */
1271         table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1272         sym_free(s);
1273     }
1274     *ptop = slast;
1275 }
1276
1277 /* eval an expression for #if/#elif */
1278 static int expr_preprocess(void)
1279 {
1280     int c, t;
1281     TokenString str;
1282
1283     tok_str_new(&str);
1284     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1285         next(); /* do macro subst */
1286         if (tok == TOK_DEFINED) {
1287             next_nomacro();
1288             t = tok;
1289             if (t == '(')
1290                 next_nomacro();
1291             c = define_find(tok) != 0;
1292             if (t == '(')
1293                 next_nomacro();
1294             tok = TOK_CINT;
1295             tokc.i = c;
1296         } else if (tok >= TOK_IDENT) {
1297             /* if undefined macro */
1298             tok = TOK_CINT;
1299             tokc.i = 0;
1300         }
1301         tok_str_add_tok(&str);
1302     }
1303     tok_str_add(&str, -1); /* simulate end of file */
1304     tok_str_add(&str, 0);
1305     /* now evaluate C constant expression */
1306     begin_macro(&str, 0);
1307     next();
1308     c = expr_const();
1309     end_macro();
1310     return c != 0;
1311 }
1312
1313
1314 /* parse after #define */
1315 ST_FUNC void parse_define(void)
1316 {
1317     Sym *s, *first, **ps;
1318     int v, t, varg, is_vaargs, spc;
1319     int saved_parse_flags = parse_flags;
1320     TokenString str;
1321
1322     v = tok;
1323     if (v < TOK_IDENT)
1324         tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1325     /* XXX: should check if same macro (ANSI) */
1326     first = NULL;
1327     t = MACRO_OBJ;
1328     /* '(' must be just after macro definition for MACRO_FUNC */
1329     parse_flags |= PARSE_FLAG_SPACES;
1330     next_nomacro_spc();
1331     if (tok == '(') {
1332         next_nomacro();
1333         ps = &first;
1334         if (tok != ')') for (;;) {
1335             varg = tok;
1336             next_nomacro();
1337             is_vaargs = 0;
1338             if (varg == TOK_DOTS) {
1339                 varg = TOK___VA_ARGS__;
1340                 is_vaargs = 1;
1341             } else if (tok == TOK_DOTS && gnu_ext) {
1342                 is_vaargs = 1;
1343                 next_nomacro();
1344             }
1345             if (varg < TOK_IDENT)
1346         bad_list:
1347                 tcc_error("bad macro parameter list");
1348             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1349             *ps = s;
1350             ps = &s->next;
1351             if (tok == ')')
1352                 break;
1353             if (tok != ',' || is_vaargs)
1354                 goto bad_list;
1355             next_nomacro();
1356         }
1357         next_nomacro_spc();
1358         t = MACRO_FUNC;
1359     }
1360     tok_str_new(&str);
1361     spc = 2;
1362     parse_flags |= PARSE_FLAG_ACCEPT_STRAYS | PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED;
1363     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1364         /* remove spaces around ## and after '#' */
1365         if (TOK_TWOSHARPS == tok) {
1366             if (2 == spc)
1367                 goto bad_twosharp;
1368             if (1 == spc)
1369                 --str.len;
1370             spc = 3;
1371         } else if ('#' == tok) {
1372             spc = 4;
1373         } else if (check_space(tok, &spc)) {
1374             goto skip;
1375         }
1376         tok_str_add2(&str, tok, &tokc);
1377     skip:
1378         next_nomacro_spc();
1379     }
1380
1381     parse_flags = saved_parse_flags;
1382     if (spc == 1)
1383         --str.len; /* remove trailing space */
1384     tok_str_add(&str, 0);
1385     if (3 == spc)
1386 bad_twosharp:
1387         tcc_error("'##' cannot appear at either end of macro");
1388     define_push(v, t, str.str, first);
1389     define_print(v);
1390 }
1391
1392 static inline int hash_cached_include(const char *filename)
1393 {
1394     const unsigned char *s;
1395     unsigned int h;
1396
1397     h = TOK_HASH_INIT;
1398     s = (unsigned char *) filename;
1399     while (*s) {
1400         h = TOK_HASH_FUNC(h, *s);
1401         s++;
1402     }
1403     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1404     return h;
1405 }
1406
1407 static CachedInclude *search_cached_include(TCCState *s1, const char *filename)
1408 {
1409     CachedInclude *e;
1410     int i, h;
1411     h = hash_cached_include(filename);
1412     i = s1->cached_includes_hash[h];
1413     for(;;) {
1414         if (i == 0)
1415             break;
1416         e = s1->cached_includes[i - 1];
1417         if (0 == PATHCMP(e->filename, filename))
1418             return e;
1419         i = e->hash_next;
1420     }
1421     return NULL;
1422 }
1423
1424 static inline void add_cached_include(TCCState *s1, const char *filename, int ifndef_macro)
1425 {
1426     CachedInclude *e;
1427     int h;
1428
1429     if (search_cached_include(s1, filename))
1430         return;
1431 #ifdef INC_DEBUG
1432     printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1433 #endif
1434     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1435     strcpy(e->filename, filename);
1436     e->ifndef_macro = ifndef_macro;
1437     dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1438     /* add in hash table */
1439     h = hash_cached_include(filename);
1440     e->hash_next = s1->cached_includes_hash[h];
1441     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1442 }
1443
1444 static void pragma_parse(TCCState *s1)
1445 {
1446     next_nomacro();
1447     if (tok == TOK_push_macro || tok == TOK_pop_macro) {
1448         int t = tok, v;
1449         Sym *s;
1450
1451         if (next(), tok != '(')
1452             goto pragma_err;
1453         if (next(), tok != TOK_STR)
1454             goto pragma_err;
1455         v = tok_alloc(tokc.cstr->data, tokc.cstr->size - 1)->tok;
1456         if (next(), tok != ')')
1457             goto pragma_err;
1458         if (t == TOK_push_macro) {
1459             while (NULL == (s = define_find(v)))
1460                 define_push(v, 0, NULL, NULL);
1461             s->type.ref = s; /* set push boundary */
1462         } else {
1463             for (s = define_stack; s; s = s->prev)
1464                 if (s->v == v && s->type.ref == s) {
1465                     s->type.ref = NULL;
1466                     break;
1467                 }
1468         }
1469         if (s)
1470             table_ident[v - TOK_IDENT]->sym_define = s->d ? s : NULL;
1471         else
1472             tcc_warning("unbalanced #pragma pop_macro");
1473
1474     } else if (tok == TOK_once) {
1475         add_cached_include(s1, file->filename, TOK_once);
1476
1477     } else if (s1->ppfp) {
1478         /* tcc -E: keep pragmas below unchanged */
1479         unget_tok(' ');
1480         unget_tok(TOK_PRAGMA);
1481         unget_tok('#');
1482         unget_tok(TOK_LINEFEED);
1483
1484     } else if (tok == TOK_pack) {
1485         /* This may be:
1486            #pragma pack(1) // set
1487            #pragma pack() // reset to default
1488            #pragma pack(push,1) // push & set
1489            #pragma pack(pop) // restore previous */
1490         next();
1491         skip('(');
1492         if (tok == TOK_ASM_pop) {
1493             next();
1494             if (s1->pack_stack_ptr <= s1->pack_stack) {
1495             stk_error:
1496                 tcc_error("out of pack stack");
1497             }
1498             s1->pack_stack_ptr--;
1499         } else {
1500             int val = 0;
1501             if (tok != ')') {
1502                 if (tok == TOK_ASM_push) {
1503                     next();
1504                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1505                         goto stk_error;
1506                     s1->pack_stack_ptr++;
1507                     skip(',');
1508                 }
1509                 if (tok != TOK_CINT)
1510                     goto pragma_err;
1511                 val = tokc.i;
1512                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1513                     goto pragma_err;
1514                 next();
1515             }
1516             *s1->pack_stack_ptr = val;
1517         }
1518         if (tok != ')')
1519             goto pragma_err;
1520
1521     } else if (tok == TOK_comment) {
1522         char *file;
1523         next();
1524         skip('(');
1525         if (tok != TOK_lib)
1526             goto pragma_warn;
1527         next();
1528         skip(',');
1529         if (tok != TOK_STR)
1530             goto pragma_err;
1531         file = tcc_strdup((char *)tokc.cstr->data);
1532         dynarray_add((void ***)&s1->pragma_libs, &s1->nb_pragma_libs, file);
1533         next();
1534         if (tok != ')')
1535             goto pragma_err;
1536     } else {
1537 pragma_warn:
1538         if (s1->warn_unsupported)
1539             tcc_warning("#pragma %s is ignored", get_tok_str(tok, &tokc));
1540     }
1541     return;
1542
1543 pragma_err:
1544     tcc_error("malformed #pragma directive");
1545     return;
1546 }
1547
1548 /* is_bof is true if first non space token at beginning of file */
1549 ST_FUNC void preprocess(int is_bof)
1550 {
1551     TCCState *s1 = tcc_state;
1552     int i, c, n, saved_parse_flags;
1553     char buf[1024], *q;
1554     Sym *s;
1555
1556     saved_parse_flags = parse_flags;
1557     parse_flags = PARSE_FLAG_PREPROCESS
1558         | PARSE_FLAG_TOK_NUM
1559         | PARSE_FLAG_TOK_STR
1560         | PARSE_FLAG_LINEFEED
1561         | (parse_flags & PARSE_FLAG_ASM_FILE)
1562         ;
1563
1564     next_nomacro();
1565  redo:
1566     switch(tok) {
1567     case TOK_DEFINE:
1568         next_nomacro();
1569         parse_define();
1570         break;
1571     case TOK_UNDEF:
1572         next_nomacro();
1573         s = define_find(tok);
1574         /* undefine symbol by putting an invalid name */
1575         if (s)
1576             define_undef(s);
1577         break;
1578     case TOK_INCLUDE:
1579     case TOK_INCLUDE_NEXT:
1580         ch = file->buf_ptr[0];
1581         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1582         skip_spaces();
1583         if (ch == '<') {
1584             c = '>';
1585             goto read_name;
1586         } else if (ch == '\"') {
1587             c = ch;
1588         read_name:
1589             inp();
1590             q = buf;
1591             while (ch != c && ch != '\n' && ch != CH_EOF) {
1592                 if ((q - buf) < sizeof(buf) - 1)
1593                     *q++ = ch;
1594                 if (ch == '\\') {
1595                     if (handle_stray_noerror() == 0)
1596                         --q;
1597                 } else
1598                     inp();
1599             }
1600             *q = '\0';
1601             minp();
1602 #if 0
1603             /* eat all spaces and comments after include */
1604             /* XXX: slightly incorrect */
1605             while (ch1 != '\n' && ch1 != CH_EOF)
1606                 inp();
1607 #endif
1608         } else {
1609             /* computed #include : either we have only strings or
1610                we have anything enclosed in '<>' */
1611             next();
1612             buf[0] = '\0';
1613             if (tok == TOK_STR) {
1614                 while (tok != TOK_LINEFEED) {
1615                     if (tok != TOK_STR) {
1616                     include_syntax:
1617                         tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1618                     }
1619                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1620                     next();
1621                 }
1622                 c = '\"';
1623             } else {
1624                 int len;
1625                 while (tok != TOK_LINEFEED) {
1626                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1627                     next();
1628                 }
1629                 len = strlen(buf);
1630                 /* check syntax and remove '<>' */
1631                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1632                     goto include_syntax;
1633                 memmove(buf, buf + 1, len - 2);
1634                 buf[len - 2] = '\0';
1635                 c = '>';
1636             }
1637         }
1638
1639         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1640             tcc_error("#include recursion too deep");
1641         /* store current file in stack, but increment stack later below */
1642         *s1->include_stack_ptr = file;
1643         i = tok == TOK_INCLUDE_NEXT ? file->include_next_index : 0;
1644         n = 2 + s1->nb_include_paths + s1->nb_sysinclude_paths;
1645         for (; i < n; ++i) {
1646             char buf1[sizeof file->filename];
1647             CachedInclude *e;
1648             const char *path;
1649
1650             if (i == 0) {
1651                 /* check absolute include path */
1652                 if (!IS_ABSPATH(buf))
1653                     continue;
1654                 buf1[0] = 0;
1655
1656             } else if (i == 1) {
1657                 /* search in current dir if "header.h" */
1658                 if (c != '\"')
1659                     continue;
1660                 path = file->filename;
1661                 pstrncpy(buf1, path, tcc_basename(path) - path);
1662
1663             } else {
1664                 /* search in all the include paths */
1665                 int j = i - 2, k = j - s1->nb_include_paths;
1666                 path = k < 0 ? s1->include_paths[j] : s1->sysinclude_paths[k];
1667                 pstrcpy(buf1, sizeof(buf1), path);
1668                 pstrcat(buf1, sizeof(buf1), "/");
1669             }
1670
1671             pstrcat(buf1, sizeof(buf1), buf);
1672             e = search_cached_include(s1, buf1);
1673             if (e && (define_find(e->ifndef_macro) || e->ifndef_macro == TOK_once)) {
1674                 /* no need to parse the include because the 'ifndef macro'
1675                    is defined */
1676 #ifdef INC_DEBUG
1677                 printf("%s: skipping cached %s\n", file->filename, buf1);
1678 #endif
1679                 goto include_done;
1680             }
1681
1682             if (tcc_open(s1, buf1) < 0)
1683                 continue;
1684
1685             file->include_next_index = i + 1;
1686 #ifdef INC_DEBUG
1687             printf("%s: including %s\n", file->prev->filename, file->filename);
1688 #endif
1689             /* update target deps */
1690             dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1691                     tcc_strdup(buf1));
1692             /* push current file in stack */
1693             ++s1->include_stack_ptr;
1694             /* add include file debug info */
1695             if (s1->do_debug)
1696                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1697             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1698             ch = file->buf_ptr[0];
1699             goto the_end;
1700         }
1701         tcc_error("include file '%s' not found", buf);
1702 include_done:
1703         break;
1704     case TOK_IFNDEF:
1705         c = 1;
1706         goto do_ifdef;
1707     case TOK_IF:
1708         c = expr_preprocess();
1709         goto do_if;
1710     case TOK_IFDEF:
1711         c = 0;
1712     do_ifdef:
1713         next_nomacro();
1714         if (tok < TOK_IDENT)
1715             tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1716         if (is_bof) {
1717             if (c) {
1718 #ifdef INC_DEBUG
1719                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1720 #endif
1721                 file->ifndef_macro = tok;
1722             }
1723         }
1724         c = (define_find(tok) != 0) ^ c;
1725     do_if:
1726         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1727             tcc_error("memory full (ifdef)");
1728         *s1->ifdef_stack_ptr++ = c;
1729         goto test_skip;
1730     case TOK_ELSE:
1731         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1732             tcc_error("#else without matching #if");
1733         if (s1->ifdef_stack_ptr[-1] & 2)
1734             tcc_error("#else after #else");
1735         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1736         goto test_else;
1737     case TOK_ELIF:
1738         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1739             tcc_error("#elif without matching #if");
1740         c = s1->ifdef_stack_ptr[-1];
1741         if (c > 1)
1742             tcc_error("#elif after #else");
1743         /* last #if/#elif expression was true: we skip */
1744         if (c == 1)
1745             goto skip;
1746         c = expr_preprocess();
1747         s1->ifdef_stack_ptr[-1] = c;
1748     test_else:
1749         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1750             file->ifndef_macro = 0;
1751     test_skip:
1752         if (!(c & 1)) {
1753         skip:
1754             preprocess_skip();
1755             is_bof = 0;
1756             goto redo;
1757         }
1758         break;
1759     case TOK_ENDIF:
1760         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1761             tcc_error("#endif without matching #if");
1762         s1->ifdef_stack_ptr--;
1763         /* '#ifndef macro' was at the start of file. Now we check if
1764            an '#endif' is exactly at the end of file */
1765         if (file->ifndef_macro &&
1766             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1767             file->ifndef_macro_saved = file->ifndef_macro;
1768             /* need to set to zero to avoid false matches if another
1769                #ifndef at middle of file */
1770             file->ifndef_macro = 0;
1771             while (tok != TOK_LINEFEED)
1772                 next_nomacro();
1773             tok_flags |= TOK_FLAG_ENDIF;
1774             goto the_end;
1775         }
1776         break;
1777     case TOK_PPNUM:
1778         n = strtoul((char*)tokc.cstr->data, &q, 10);
1779         goto _line_num;
1780     case TOK_LINE:
1781         next();
1782         if (tok != TOK_CINT)
1783 _line_err:
1784             tcc_error("wrong #line format");
1785         n = tokc.i;
1786 _line_num:
1787         next();
1788         if (tok != TOK_LINEFEED) {
1789             if (tok == TOK_STR)
1790                 pstrcpy(file->filename, sizeof(file->filename), (char *)tokc.cstr->data);
1791             else if (parse_flags & PARSE_FLAG_ASM_FILE)
1792                 break;
1793             else
1794                 goto _line_err;
1795             --n;
1796         }
1797         if (file->fd > 0)
1798             total_lines += file->line_num - n;
1799         file->line_num = n;
1800         if (s1->do_debug)
1801             put_stabs(file->filename, N_BINCL, 0, 0, 0);
1802         break;
1803     case TOK_ERROR:
1804     case TOK_WARNING:
1805         c = tok;
1806         ch = file->buf_ptr[0];
1807         skip_spaces();
1808         q = buf;
1809         while (ch != '\n' && ch != CH_EOF) {
1810             if ((q - buf) < sizeof(buf) - 1)
1811                 *q++ = ch;
1812             if (ch == '\\') {
1813                 if (handle_stray_noerror() == 0)
1814                     --q;
1815             } else
1816                 inp();
1817         }
1818         *q = '\0';
1819         if (c == TOK_ERROR)
1820             tcc_error("#error %s", buf);
1821         else
1822             tcc_warning("#warning %s", buf);
1823         break;
1824     case TOK_PRAGMA:
1825         pragma_parse(s1);
1826         break;
1827     case TOK_LINEFEED:
1828         goto the_end;
1829     default:
1830         /* ignore gas line comment in an 'S' file. */
1831         if (saved_parse_flags & PARSE_FLAG_ASM_FILE)
1832             goto ignore;
1833         if (tok == '!' && is_bof)
1834             /* '!' is ignored at beginning to allow C scripts. */
1835             goto ignore;
1836         tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1837     ignore:
1838         file->buf_ptr = parse_line_comment(file->buf_ptr);
1839         goto the_end;
1840     }
1841     /* ignore other preprocess commands or #! for C scripts */
1842     while (tok != TOK_LINEFEED)
1843         next_nomacro();
1844  the_end:
1845     parse_flags = saved_parse_flags;
1846 }
1847
1848 /* evaluate escape codes in a string. */
1849 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1850 {
1851     int c, n;
1852     const uint8_t *p;
1853
1854     p = buf;
1855     for(;;) {
1856         c = *p;
1857         if (c == '\0')
1858             break;
1859         if (c == '\\') {
1860             p++;
1861             /* escape */
1862             c = *p;
1863             switch(c) {
1864             case '0': case '1': case '2': case '3':
1865             case '4': case '5': case '6': case '7':
1866                 /* at most three octal digits */
1867                 n = c - '0';
1868                 p++;
1869                 c = *p;
1870                 if (isoct(c)) {
1871                     n = n * 8 + c - '0';
1872                     p++;
1873                     c = *p;
1874                     if (isoct(c)) {
1875                         n = n * 8 + c - '0';
1876                         p++;
1877                     }
1878                 }
1879                 c = n;
1880                 goto add_char_nonext;
1881             case 'x':
1882             case 'u':
1883             case 'U':
1884                 p++;
1885                 n = 0;
1886                 for(;;) {
1887                     c = *p;
1888                     if (c >= 'a' && c <= 'f')
1889                         c = c - 'a' + 10;
1890                     else if (c >= 'A' && c <= 'F')
1891                         c = c - 'A' + 10;
1892                     else if (isnum(c))
1893                         c = c - '0';
1894                     else
1895                         break;
1896                     n = n * 16 + c;
1897                     p++;
1898                 }
1899                 c = n;
1900                 goto add_char_nonext;
1901             case 'a':
1902                 c = '\a';
1903                 break;
1904             case 'b':
1905                 c = '\b';
1906                 break;
1907             case 'f':
1908                 c = '\f';
1909                 break;
1910             case 'n':
1911                 c = '\n';
1912                 break;
1913             case 'r':
1914                 c = '\r';
1915                 break;
1916             case 't':
1917                 c = '\t';
1918                 break;
1919             case 'v':
1920                 c = '\v';
1921                 break;
1922             case 'e':
1923                 if (!gnu_ext)
1924                     goto invalid_escape;
1925                 c = 27;
1926                 break;
1927             case '\'':
1928             case '\"':
1929             case '\\':
1930             case '?':
1931                 break;
1932             default:
1933             invalid_escape:
1934                 if (c >= '!' && c <= '~')
1935                     tcc_warning("unknown escape sequence: \'\\%c\'", c);
1936                 else
1937                     tcc_warning("unknown escape sequence: \'\\x%x\'", c);
1938                 break;
1939             }
1940         }
1941         p++;
1942     add_char_nonext:
1943         if (!is_long)
1944             cstr_ccat(outstr, c);
1945         else
1946             cstr_wccat(outstr, c);
1947     }
1948     /* add a trailing '\0' */
1949     if (!is_long)
1950         cstr_ccat(outstr, '\0');
1951     else
1952         cstr_wccat(outstr, '\0');
1953 }
1954
1955 void parse_string(const char *s, int len)
1956 {
1957     uint8_t buf[1000], *p = buf;
1958     int is_long, sep;
1959
1960     if ((is_long = *s == 'L'))
1961         ++s, --len;
1962     sep = *s++;
1963     len -= 2;
1964     if (len >= sizeof buf)
1965         p = tcc_malloc(len + 1);
1966     memcpy(p, s, len);
1967     p[len] = 0;
1968
1969     cstr_reset(&tokcstr);
1970     parse_escape_string(&tokcstr, p, is_long);
1971     if (p != buf)
1972         tcc_free(p);
1973
1974     if (sep == '\'') {
1975         int char_size;
1976         /* XXX: make it portable */
1977         if (!is_long)
1978             char_size = 1;
1979         else
1980             char_size = sizeof(nwchar_t);
1981         if (tokcstr.size <= char_size)
1982             tcc_error("empty character constant");
1983         if (tokcstr.size > 2 * char_size)
1984             tcc_warning("multi-character character constant");
1985         if (!is_long) {
1986             tokc.i = *(int8_t *)tokcstr.data;
1987             tok = TOK_CCHAR;
1988         } else {
1989             tokc.i = *(nwchar_t *)tokcstr.data;
1990             tok = TOK_LCHAR;
1991         }
1992     } else {
1993         tokc.cstr = &tokcstr;
1994         if (!is_long)
1995             tok = TOK_STR;
1996         else
1997             tok = TOK_LSTR;
1998     }
1999 }
2000
2001 /* we use 64 bit numbers */
2002 #define BN_SIZE 2
2003
2004 /* bn = (bn << shift) | or_val */
2005 static void bn_lshift(unsigned int *bn, int shift, int or_val)
2006 {
2007     int i;
2008     unsigned int v;
2009     for(i=0;i<BN_SIZE;i++) {
2010         v = bn[i];
2011         bn[i] = (v << shift) | or_val;
2012         or_val = v >> (32 - shift);
2013     }
2014 }
2015
2016 static void bn_zero(unsigned int *bn)
2017 {
2018     int i;
2019     for(i=0;i<BN_SIZE;i++) {
2020         bn[i] = 0;
2021     }
2022 }
2023
2024 /* parse number in null terminated string 'p' and return it in the
2025    current token */
2026 static void parse_number(const char *p)
2027 {
2028     int b, t, shift, frac_bits, s, exp_val, ch;
2029     char *q;
2030     unsigned int bn[BN_SIZE];
2031     double d;
2032
2033     /* number */
2034     q = token_buf;
2035     ch = *p++;
2036     t = ch;
2037     ch = *p++;
2038     *q++ = t;
2039     b = 10;
2040     if (t == '.') {
2041         goto float_frac_parse;
2042     } else if (t == '0') {
2043         if (ch == 'x' || ch == 'X') {
2044             q--;
2045             ch = *p++;
2046             b = 16;
2047         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
2048             q--;
2049             ch = *p++;
2050             b = 2;
2051         }
2052     }
2053     /* parse all digits. cannot check octal numbers at this stage
2054        because of floating point constants */
2055     while (1) {
2056         if (ch >= 'a' && ch <= 'f')
2057             t = ch - 'a' + 10;
2058         else if (ch >= 'A' && ch <= 'F')
2059             t = ch - 'A' + 10;
2060         else if (isnum(ch))
2061             t = ch - '0';
2062         else
2063             break;
2064         if (t >= b)
2065             break;
2066         if (q >= token_buf + STRING_MAX_SIZE) {
2067         num_too_long:
2068             tcc_error("number too long");
2069         }
2070         *q++ = ch;
2071         ch = *p++;
2072     }
2073     if (ch == '.' ||
2074         ((ch == 'e' || ch == 'E') && b == 10) ||
2075         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
2076         if (b != 10) {
2077             /* NOTE: strtox should support that for hexa numbers, but
2078                non ISOC99 libcs do not support it, so we prefer to do
2079                it by hand */
2080             /* hexadecimal or binary floats */
2081             /* XXX: handle overflows */
2082             *q = '\0';
2083             if (b == 16)
2084                 shift = 4;
2085             else
2086                 shift = 1;
2087             bn_zero(bn);
2088             q = token_buf;
2089             while (1) {
2090                 t = *q++;
2091                 if (t == '\0') {
2092                     break;
2093                 } else if (t >= 'a') {
2094                     t = t - 'a' + 10;
2095                 } else if (t >= 'A') {
2096                     t = t - 'A' + 10;
2097                 } else {
2098                     t = t - '0';
2099                 }
2100                 bn_lshift(bn, shift, t);
2101             }
2102             frac_bits = 0;
2103             if (ch == '.') {
2104                 ch = *p++;
2105                 while (1) {
2106                     t = ch;
2107                     if (t >= 'a' && t <= 'f') {
2108                         t = t - 'a' + 10;
2109                     } else if (t >= 'A' && t <= 'F') {
2110                         t = t - 'A' + 10;
2111                     } else if (t >= '0' && t <= '9') {
2112                         t = t - '0';
2113                     } else {
2114                         break;
2115                     }
2116                     if (t >= b)
2117                         tcc_error("invalid digit");
2118                     bn_lshift(bn, shift, t);
2119                     frac_bits += shift;
2120                     ch = *p++;
2121                 }
2122             }
2123             if (ch != 'p' && ch != 'P')
2124                 expect("exponent");
2125             ch = *p++;
2126             s = 1;
2127             exp_val = 0;
2128             if (ch == '+') {
2129                 ch = *p++;
2130             } else if (ch == '-') {
2131                 s = -1;
2132                 ch = *p++;
2133             }
2134             if (ch < '0' || ch > '9')
2135                 expect("exponent digits");
2136             while (ch >= '0' && ch <= '9') {
2137                 exp_val = exp_val * 10 + ch - '0';
2138                 ch = *p++;
2139             }
2140             exp_val = exp_val * s;
2141
2142             /* now we can generate the number */
2143             /* XXX: should patch directly float number */
2144             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
2145             d = ldexp(d, exp_val - frac_bits);
2146             t = toup(ch);
2147             if (t == 'F') {
2148                 ch = *p++;
2149                 tok = TOK_CFLOAT;
2150                 /* float : should handle overflow */
2151                 tokc.f = (float)d;
2152             } else if (t == 'L') {
2153                 ch = *p++;
2154 #ifdef TCC_TARGET_PE
2155                 tok = TOK_CDOUBLE;
2156                 tokc.d = d;
2157 #else
2158                 tok = TOK_CLDOUBLE;
2159                 /* XXX: not large enough */
2160                 tokc.ld = (long double)d;
2161 #endif
2162             } else {
2163                 tok = TOK_CDOUBLE;
2164                 tokc.d = d;
2165             }
2166         } else {
2167             /* decimal floats */
2168             if (ch == '.') {
2169                 if (q >= token_buf + STRING_MAX_SIZE)
2170                     goto num_too_long;
2171                 *q++ = ch;
2172                 ch = *p++;
2173             float_frac_parse:
2174                 while (ch >= '0' && ch <= '9') {
2175                     if (q >= token_buf + STRING_MAX_SIZE)
2176                         goto num_too_long;
2177                     *q++ = ch;
2178                     ch = *p++;
2179                 }
2180             }
2181             if (ch == 'e' || ch == 'E') {
2182                 if (q >= token_buf + STRING_MAX_SIZE)
2183                     goto num_too_long;
2184                 *q++ = ch;
2185                 ch = *p++;
2186                 if (ch == '-' || ch == '+') {
2187                     if (q >= token_buf + STRING_MAX_SIZE)
2188                         goto num_too_long;
2189                     *q++ = ch;
2190                     ch = *p++;
2191                 }
2192                 if (ch < '0' || ch > '9')
2193                     expect("exponent digits");
2194                 while (ch >= '0' && ch <= '9') {
2195                     if (q >= token_buf + STRING_MAX_SIZE)
2196                         goto num_too_long;
2197                     *q++ = ch;
2198                     ch = *p++;
2199                 }
2200             }
2201             *q = '\0';
2202             t = toup(ch);
2203             errno = 0;
2204             if (t == 'F') {
2205                 ch = *p++;
2206                 tok = TOK_CFLOAT;
2207                 tokc.f = strtof(token_buf, NULL);
2208             } else if (t == 'L') {
2209                 ch = *p++;
2210 #ifdef TCC_TARGET_PE
2211                 tok = TOK_CDOUBLE;
2212                 tokc.d = strtod(token_buf, NULL);
2213 #else
2214                 tok = TOK_CLDOUBLE;
2215                 tokc.ld = strtold(token_buf, NULL);
2216 #endif
2217             } else {
2218                 tok = TOK_CDOUBLE;
2219                 tokc.d = strtod(token_buf, NULL);
2220             }
2221         }
2222     } else {
2223         unsigned long long n, n1;
2224         int lcount, ucount, must_64bit;
2225         const char *p1;
2226
2227         /* integer number */
2228         *q = '\0';
2229         q = token_buf;
2230         if (b == 10 && *q == '0') {
2231             b = 8;
2232             q++;
2233         }
2234         n = 0;
2235         while(1) {
2236             t = *q++;
2237             /* no need for checks except for base 10 / 8 errors */
2238             if (t == '\0')
2239                 break;
2240             else if (t >= 'a')
2241                 t = t - 'a' + 10;
2242             else if (t >= 'A')
2243                 t = t - 'A' + 10;
2244             else
2245                 t = t - '0';
2246             if (t >= b)
2247                 tcc_error("invalid digit");
2248             n1 = n;
2249             n = n * b + t;
2250             /* detect overflow */
2251             /* XXX: this test is not reliable */
2252             if (n < n1)
2253                 tcc_error("integer constant overflow");
2254         }
2255
2256         /* Determine the characteristics (unsigned and/or 64bit) the type of
2257            the constant must have according to the constant suffix(es) */
2258         lcount = ucount = must_64bit = 0;
2259         p1 = p;
2260         for(;;) {
2261             t = toup(ch);
2262             if (t == 'L') {
2263                 if (lcount >= 2)
2264                     tcc_error("three 'l's in integer constant");
2265                 if (lcount && *(p - 1) != ch)
2266                     tcc_error("incorrect integer suffix: %s", p1);
2267                 lcount++;
2268 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
2269                 if (lcount == 2)
2270 #endif
2271                     must_64bit = 1;
2272                 ch = *p++;
2273             } else if (t == 'U') {
2274                 if (ucount >= 1)
2275                     tcc_error("two 'u's in integer constant");
2276                 ucount++;
2277                 ch = *p++;
2278             } else {
2279                 break;
2280             }
2281         }
2282
2283         /* Whether 64 bits are needed to hold the constant's value */
2284         if (n & 0xffffffff00000000LL || must_64bit) {
2285             tok = TOK_CLLONG;
2286             n1 = n >> 32;
2287         } else {
2288             tok = TOK_CINT;
2289             n1 = n;
2290         }
2291
2292         /* Whether type must be unsigned to hold the constant's value */
2293         if (ucount || ((n1 >> 31) && (b != 10))) {
2294             if (tok == TOK_CLLONG)
2295                 tok = TOK_CULLONG;
2296             else
2297                 tok = TOK_CUINT;
2298         /* If decimal and no unsigned suffix, bump to 64 bits or throw error */
2299         } else if (n1 >> 31) {
2300             if (tok == TOK_CINT)
2301                 tok = TOK_CLLONG;
2302             else
2303                 tcc_error("integer constant overflow");
2304         }
2305
2306         if (tok == TOK_CINT || tok == TOK_CUINT)
2307             tokc.i = n;
2308         else
2309             tokc.i = n;
2310     }
2311     if (ch)
2312         tcc_error("invalid number\n");
2313 }
2314
2315
2316 #define PARSE2(c1, tok1, c2, tok2)              \
2317     case c1:                                    \
2318         PEEKC(c, p);                            \
2319         if (c == c2) {                          \
2320             p++;                                \
2321             tok = tok2;                         \
2322         } else {                                \
2323             tok = tok1;                         \
2324         }                                       \
2325         break;
2326
2327 /* return next token without macro substitution */
2328 static inline void next_nomacro1(void)
2329 {
2330     int t, c, is_long;
2331     TokenSym *ts;
2332     uint8_t *p, *p1;
2333     unsigned int h;
2334
2335     p = file->buf_ptr;
2336  redo_no_start:
2337     c = *p;
2338     switch(c) {
2339     case ' ':
2340     case '\t':
2341         tok = c;
2342         p++;
2343         if (parse_flags & PARSE_FLAG_SPACES)
2344             goto keep_tok_flags;
2345         while (isidnum_table[*p - CH_EOF] & IS_SPC)
2346             ++p;
2347         goto redo_no_start;
2348     case '\f':
2349     case '\v':
2350     case '\r':
2351         p++;
2352         goto redo_no_start;
2353     case '\\':
2354         /* first look if it is in fact an end of buffer */
2355         c = handle_stray1(p);
2356         p = file->buf_ptr;
2357         if (c == '\\')
2358             goto parse_simple;
2359         if (c != CH_EOF)
2360             goto redo_no_start;
2361         {
2362             TCCState *s1 = tcc_state;
2363             if ((parse_flags & PARSE_FLAG_LINEFEED)
2364                 && !(tok_flags & TOK_FLAG_EOF)) {
2365                 tok_flags |= TOK_FLAG_EOF;
2366                 tok = TOK_LINEFEED;
2367                 goto keep_tok_flags;
2368             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2369                 tok = TOK_EOF;
2370             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2371                 tcc_error("missing #endif");
2372             } else if (s1->include_stack_ptr == s1->include_stack) {
2373                 /* no include left : end of file. */
2374                 tok = TOK_EOF;
2375             } else {
2376                 tok_flags &= ~TOK_FLAG_EOF;
2377                 /* pop include file */
2378
2379                 /* test if previous '#endif' was after a #ifdef at
2380                    start of file */
2381                 if (tok_flags & TOK_FLAG_ENDIF) {
2382 #ifdef INC_DEBUG
2383                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2384 #endif
2385                     add_cached_include(s1, file->filename, file->ifndef_macro_saved);
2386                     tok_flags &= ~TOK_FLAG_ENDIF;
2387                 }
2388
2389                 /* add end of include file debug info */
2390                 if (tcc_state->do_debug) {
2391                     put_stabd(N_EINCL, 0, 0);
2392                 }
2393                 /* pop include stack */
2394                 tcc_close();
2395                 s1->include_stack_ptr--;
2396                 p = file->buf_ptr;
2397                 goto redo_no_start;
2398             }
2399         }
2400         break;
2401
2402     case '\n':
2403         file->line_num++;
2404         tok_flags |= TOK_FLAG_BOL;
2405         p++;
2406 maybe_newline:
2407         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2408             goto redo_no_start;
2409         tok = TOK_LINEFEED;
2410         goto keep_tok_flags;
2411
2412     case '#':
2413         /* XXX: simplify */
2414         PEEKC(c, p);
2415         if ((tok_flags & TOK_FLAG_BOL) &&
2416             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2417             file->buf_ptr = p;
2418             preprocess(tok_flags & TOK_FLAG_BOF);
2419             p = file->buf_ptr;
2420             goto maybe_newline;
2421         } else {
2422             if (c == '#') {
2423                 p++;
2424                 tok = TOK_TWOSHARPS;
2425             } else {
2426                 if (parse_flags & PARSE_FLAG_ASM_FILE) {
2427                     p = parse_line_comment(p - 1);
2428                     goto redo_no_start;
2429                 } else {
2430                     tok = '#';
2431                 }
2432             }
2433         }
2434         break;
2435
2436     /* dollar is allowed to start identifiers when not parsing asm */
2437     case '$':
2438         if (!(isidnum_table[c - CH_EOF] & IS_ID)
2439          || (parse_flags & PARSE_FLAG_ASM_FILE))
2440             goto parse_simple;
2441
2442     case 'a': case 'b': case 'c': case 'd':
2443     case 'e': case 'f': case 'g': case 'h':
2444     case 'i': case 'j': case 'k': case 'l':
2445     case 'm': case 'n': case 'o': case 'p':
2446     case 'q': case 'r': case 's': case 't':
2447     case 'u': case 'v': case 'w': case 'x':
2448     case 'y': case 'z':
2449     case 'A': case 'B': case 'C': case 'D':
2450     case 'E': case 'F': case 'G': case 'H':
2451     case 'I': case 'J': case 'K':
2452     case 'M': case 'N': case 'O': case 'P':
2453     case 'Q': case 'R': case 'S': case 'T':
2454     case 'U': case 'V': case 'W': case 'X':
2455     case 'Y': case 'Z':
2456     case '_':
2457     parse_ident_fast:
2458         p1 = p;
2459         h = TOK_HASH_INIT;
2460         h = TOK_HASH_FUNC(h, c);
2461         while (c = *++p, isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2462             h = TOK_HASH_FUNC(h, c);
2463         if (c != '\\') {
2464             TokenSym **pts;
2465             int len;
2466
2467             /* fast case : no stray found, so we have the full token
2468                and we have already hashed it */
2469             len = p - p1;
2470             h &= (TOK_HASH_SIZE - 1);
2471             pts = &hash_ident[h];
2472             for(;;) {
2473                 ts = *pts;
2474                 if (!ts)
2475                     break;
2476                 if (ts->len == len && !memcmp(ts->str, p1, len))
2477                     goto token_found;
2478                 pts = &(ts->hash_next);
2479             }
2480             ts = tok_alloc_new(pts, (char *) p1, len);
2481         token_found: ;
2482         } else {
2483             /* slower case */
2484             cstr_reset(&tokcstr);
2485
2486             while (p1 < p) {
2487                 cstr_ccat(&tokcstr, *p1);
2488                 p1++;
2489             }
2490             p--;
2491             PEEKC(c, p);
2492         parse_ident_slow:
2493             while (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM)) {
2494                 cstr_ccat(&tokcstr, c);
2495                 PEEKC(c, p);
2496             }
2497             ts = tok_alloc(tokcstr.data, tokcstr.size);
2498         }
2499         tok = ts->tok;
2500         break;
2501     case 'L':
2502         t = p[1];
2503         if (t != '\\' && t != '\'' && t != '\"') {
2504             /* fast case */
2505             goto parse_ident_fast;
2506         } else {
2507             PEEKC(c, p);
2508             if (c == '\'' || c == '\"') {
2509                 is_long = 1;
2510                 goto str_const;
2511             } else {
2512                 cstr_reset(&tokcstr);
2513                 cstr_ccat(&tokcstr, 'L');
2514                 goto parse_ident_slow;
2515             }
2516         }
2517         break;
2518
2519     case '0': case '1': case '2': case '3':
2520     case '4': case '5': case '6': case '7':
2521     case '8': case '9':
2522         cstr_reset(&tokcstr);
2523         /* after the first digit, accept digits, alpha, '.' or sign if
2524            prefixed by 'eEpP' */
2525     parse_num:
2526         for(;;) {
2527             t = c;
2528             cstr_ccat(&tokcstr, c);
2529             PEEKC(c, p);
2530             if (!((isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2531                   || c == '.'
2532                   || ((c == '+' || c == '-')
2533                       && (t == 'e' || t == 'E' || t == 'p' || t == 'P')
2534                       )))
2535                 break;
2536         }
2537         /* We add a trailing '\0' to ease parsing */
2538         cstr_ccat(&tokcstr, '\0');
2539         tokc.cstr = &tokcstr;
2540         tok = TOK_PPNUM;
2541         break;
2542
2543     case '.':
2544         /* special dot handling because it can also start a number */
2545         PEEKC(c, p);
2546         if (isnum(c)) {
2547             cstr_reset(&tokcstr);
2548             cstr_ccat(&tokcstr, '.');
2549             goto parse_num;
2550         } else if (c == '.') {
2551             PEEKC(c, p);
2552             if (c == '.') {
2553                 p++;
2554                 tok = TOK_DOTS;
2555             } else {
2556                 *--p = '.'; /* may underflow into file->unget[] */
2557                 tok = '.';
2558             }
2559         } else {
2560             tok = '.';
2561         }
2562         break;
2563     case '\'':
2564     case '\"':
2565         is_long = 0;
2566     str_const:
2567         cstr_reset(&tokcstr);
2568         if (is_long)
2569             cstr_ccat(&tokcstr, 'L');
2570         cstr_ccat(&tokcstr, c);
2571         p = parse_pp_string(p, c, &tokcstr);
2572         cstr_ccat(&tokcstr, c);
2573         cstr_ccat(&tokcstr, '\0');
2574         tokc.cstr = &tokcstr;
2575         tok = TOK_PPSTR;
2576         break;
2577
2578     case '<':
2579         PEEKC(c, p);
2580         if (c == '=') {
2581             p++;
2582             tok = TOK_LE;
2583         } else if (c == '<') {
2584             PEEKC(c, p);
2585             if (c == '=') {
2586                 p++;
2587                 tok = TOK_A_SHL;
2588             } else {
2589                 tok = TOK_SHL;
2590             }
2591         } else {
2592             tok = TOK_LT;
2593         }
2594         break;
2595     case '>':
2596         PEEKC(c, p);
2597         if (c == '=') {
2598             p++;
2599             tok = TOK_GE;
2600         } else if (c == '>') {
2601             PEEKC(c, p);
2602             if (c == '=') {
2603                 p++;
2604                 tok = TOK_A_SAR;
2605             } else {
2606                 tok = TOK_SAR;
2607             }
2608         } else {
2609             tok = TOK_GT;
2610         }
2611         break;
2612
2613     case '&':
2614         PEEKC(c, p);
2615         if (c == '&') {
2616             p++;
2617             tok = TOK_LAND;
2618         } else if (c == '=') {
2619             p++;
2620             tok = TOK_A_AND;
2621         } else {
2622             tok = '&';
2623         }
2624         break;
2625
2626     case '|':
2627         PEEKC(c, p);
2628         if (c == '|') {
2629             p++;
2630             tok = TOK_LOR;
2631         } else if (c == '=') {
2632             p++;
2633             tok = TOK_A_OR;
2634         } else {
2635             tok = '|';
2636         }
2637         break;
2638
2639     case '+':
2640         PEEKC(c, p);
2641         if (c == '+') {
2642             p++;
2643             tok = TOK_INC;
2644         } else if (c == '=') {
2645             p++;
2646             tok = TOK_A_ADD;
2647         } else {
2648             tok = '+';
2649         }
2650         break;
2651
2652     case '-':
2653         PEEKC(c, p);
2654         if (c == '-') {
2655             p++;
2656             tok = TOK_DEC;
2657         } else if (c == '=') {
2658             p++;
2659             tok = TOK_A_SUB;
2660         } else if (c == '>') {
2661             p++;
2662             tok = TOK_ARROW;
2663         } else {
2664             tok = '-';
2665         }
2666         break;
2667
2668     PARSE2('!', '!', '=', TOK_NE)
2669     PARSE2('=', '=', '=', TOK_EQ)
2670     PARSE2('*', '*', '=', TOK_A_MUL)
2671     PARSE2('%', '%', '=', TOK_A_MOD)
2672     PARSE2('^', '^', '=', TOK_A_XOR)
2673
2674         /* comments or operator */
2675     case '/':
2676         PEEKC(c, p);
2677         if (c == '*') {
2678             p = parse_comment(p);
2679             /* comments replaced by a blank */
2680             tok = ' ';
2681             goto keep_tok_flags;
2682         } else if (c == '/') {
2683             p = parse_line_comment(p);
2684             tok = ' ';
2685             goto keep_tok_flags;
2686         } else if (c == '=') {
2687             p++;
2688             tok = TOK_A_DIV;
2689         } else {
2690             tok = '/';
2691         }
2692         break;
2693
2694         /* simple tokens */
2695     case '(':
2696     case ')':
2697     case '[':
2698     case ']':
2699     case '{':
2700     case '}':
2701     case ',':
2702     case ';':
2703     case ':':
2704     case '?':
2705     case '~':
2706     case '@': /* only used in assembler */
2707     parse_simple:
2708         tok = c;
2709         p++;
2710         break;
2711     default:
2712         tcc_error("unrecognized character \\x%02x", c);
2713         break;
2714     }
2715     tok_flags = 0;
2716 keep_tok_flags:
2717     file->buf_ptr = p;
2718 #if defined(PARSE_DEBUG)
2719     printf("token = %s\n", get_tok_str(tok, &tokc));
2720 #endif
2721 }
2722
2723 /* return next token without macro substitution. Can read input from
2724    macro_ptr buffer */
2725 static void next_nomacro_spc(void)
2726 {
2727     if (macro_ptr) {
2728     redo:
2729         tok = *macro_ptr;
2730         if (tok) {
2731             TOK_GET(&tok, &macro_ptr, &tokc);
2732             if (tok == TOK_LINENUM) {
2733                 file->line_num = tokc.i;
2734                 goto redo;
2735             }
2736         }
2737     } else {
2738         next_nomacro1();
2739     }
2740 }
2741
2742 ST_FUNC void next_nomacro(void)
2743 {
2744     do {
2745         next_nomacro_spc();
2746     } while (tok < 256 && (isidnum_table[tok - CH_EOF] & IS_SPC));
2747 }
2748
2749
2750 static void macro_subst(
2751     TokenString *tok_str,
2752     Sym **nested_list,
2753     const int *macro_str,
2754     int can_read_stream
2755     );
2756
2757 /* substitute arguments in replacement lists in macro_str by the values in
2758    args (field d) and return allocated string */
2759 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2760 {
2761     int t, t0, t1, spc;
2762     const int *st;
2763     Sym *s;
2764     CValue cval;
2765     TokenString str;
2766     CString cstr;
2767
2768     tok_str_new(&str);
2769     t0 = t1 = 0;
2770     while(1) {
2771         TOK_GET(&t, &macro_str, &cval);
2772         if (!t)
2773             break;
2774         if (t == '#') {
2775             /* stringize */
2776             TOK_GET(&t, &macro_str, &cval);
2777             if (!t)
2778                 goto bad_stringy;
2779             s = sym_find2(args, t);
2780             if (s) {
2781                 cstr_new(&cstr);
2782                 cstr_ccat(&cstr, '\"');
2783                 st = s->d;
2784                 spc = 0;
2785                 while (*st) {
2786                     TOK_GET(&t, &st, &cval);
2787                     if (t != TOK_PLCHLDR
2788                      && t != TOK_NOSUBST
2789                      && 0 == check_space(t, &spc)) {
2790                         const char *s = get_tok_str(t, &cval);
2791                         while (*s) {
2792                             if (t == TOK_PPSTR && *s != '\'')
2793                                 add_char(&cstr, *s);
2794                             else
2795                                 cstr_ccat(&cstr, *s);
2796                             ++s;
2797                         }
2798                     }
2799                 }
2800                 cstr.size -= spc;
2801                 cstr_ccat(&cstr, '\"');
2802                 cstr_ccat(&cstr, '\0');
2803 #ifdef PP_DEBUG
2804                 printf("\nstringize: <%s>\n", (char *)cstr.data);
2805 #endif
2806                 /* add string */
2807                 cval.cstr = &cstr;
2808                 tok_str_add2(&str, TOK_PPSTR, &cval);
2809                 cstr_free(cval.cstr);
2810             } else {
2811         bad_stringy:
2812                 expect("macro parameter after '#'");
2813             }
2814         } else if (t >= TOK_IDENT) {
2815             s = sym_find2(args, t);
2816             if (s) {
2817                 int l0 = str.len;
2818                 st = s->d;
2819                 /* if '##' is present before or after, no arg substitution */
2820                 if (*macro_str == TOK_TWOSHARPS || t1 == TOK_TWOSHARPS) {
2821                     /* special case for var arg macros : ## eats the ','
2822                        if empty VA_ARGS variable. */
2823                     if (t1 == TOK_TWOSHARPS && t0 == ',' && gnu_ext && s->type.t) {
2824                         if (*st == 0) {
2825                             /* suppress ',' '##' */
2826                             str.len -= 2;
2827                         } else {
2828                             /* suppress '##' and add variable */
2829                             str.len--;
2830                             goto add_var;
2831                         }
2832                     } else {
2833                         for(;;) {
2834                             int t1;
2835                             TOK_GET(&t1, &st, &cval);
2836                             if (!t1)
2837                                 break;
2838                             tok_str_add2(&str, t1, &cval);
2839                         }
2840                     }
2841
2842                 } else {
2843             add_var:
2844                     /* NOTE: the stream cannot be read when macro
2845                        substituing an argument */
2846                     macro_subst(&str, nested_list, st, 0);
2847                 }
2848                 if (str.len == l0) /* exanded to empty string */
2849                     tok_str_add(&str, TOK_PLCHLDR);
2850             } else {
2851                 tok_str_add(&str, t);
2852             }
2853         } else {
2854             tok_str_add2(&str, t, &cval);
2855         }
2856         t0 = t1, t1 = t;
2857     }
2858     tok_str_add(&str, 0);
2859     return str.str;
2860 }
2861
2862 static char const ab_month_name[12][4] =
2863 {
2864     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2865     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2866 };
2867
2868 /* peek or read [ws_str == NULL] next token from function macro call,
2869    walking up macro levels up to the file if necessary */
2870 static int next_argstream(Sym **nested_list, int can_read_stream, TokenString *ws_str)
2871 {
2872     int t;
2873     const int *p;
2874     Sym *sa;
2875
2876     for (;;) {
2877         if (macro_ptr) {
2878             p = macro_ptr, t = *p;
2879             if (ws_str) {
2880                 while (is_space(t) || TOK_LINEFEED == t)
2881                     tok_str_add(ws_str, t), t = *++p;
2882             }
2883             if (t == 0 && can_read_stream) {
2884                 end_macro();
2885                 /* also, end of scope for nested defined symbol */
2886                 sa = *nested_list;
2887                 while (sa && sa->v == -1)
2888                     sa = sa->prev;
2889                 if (sa)
2890                     sa->v = -1;
2891                 continue;
2892             }
2893         } else {
2894             ch = handle_eob();
2895             if (ws_str) {
2896                 while (is_space(ch) || ch == '\n' || ch == '/') {
2897                     if (ch == '/') {
2898                         int c;
2899                         uint8_t *p = file->buf_ptr;
2900                         PEEKC(c, p);
2901                         if (c == '*') {
2902                             p = parse_comment(p);
2903                             file->buf_ptr = p - 1;
2904                         } else if (c == '/') {
2905                             p = parse_line_comment(p);
2906                             file->buf_ptr = p - 1;
2907                         } else
2908                             break;
2909                         ch = ' ';
2910                     }
2911                     tok_str_add(ws_str, ch);
2912                     cinp();
2913                 }
2914             }
2915             t = ch;
2916         }
2917
2918         if (ws_str)
2919             return t;
2920         next_nomacro_spc();
2921         return tok;
2922     }
2923 }
2924
2925 /* do macro substitution of current token with macro 's' and add
2926    result to (tok_str,tok_len). 'nested_list' is the list of all
2927    macros we got inside to avoid recursing. Return non zero if no
2928    substitution needs to be done */
2929 static int macro_subst_tok(
2930     TokenString *tok_str,
2931     Sym **nested_list,
2932     Sym *s,
2933     int can_read_stream)
2934 {
2935     Sym *args, *sa, *sa1;
2936     int parlevel, *mstr, t, t1, spc;
2937     TokenString str;
2938     char *cstrval;
2939     CValue cval;
2940     CString cstr;
2941     char buf[32];
2942
2943     /* if symbol is a macro, prepare substitution */
2944     /* special macros */
2945     if (tok == TOK___LINE__) {
2946         snprintf(buf, sizeof(buf), "%d", file->line_num);
2947         cstrval = buf;
2948         t1 = TOK_PPNUM;
2949         goto add_cstr1;
2950     } else if (tok == TOK___FILE__) {
2951         cstrval = file->filename;
2952         goto add_cstr;
2953     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2954         time_t ti;
2955         struct tm *tm;
2956
2957         time(&ti);
2958         tm = localtime(&ti);
2959         if (tok == TOK___DATE__) {
2960             snprintf(buf, sizeof(buf), "%s %2d %d",
2961                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2962         } else {
2963             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2964                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2965         }
2966         cstrval = buf;
2967     add_cstr:
2968         t1 = TOK_STR;
2969     add_cstr1:
2970         cstr_new(&cstr);
2971         cstr_cat(&cstr, cstrval);
2972         cstr_ccat(&cstr, '\0');
2973         cval.cstr = &cstr;
2974         tok_str_add2(tok_str, t1, &cval);
2975         cstr_free(&cstr);
2976     } else {
2977         int saved_parse_flags = parse_flags;
2978
2979         mstr = s->d;
2980         if (s->type.t == MACRO_FUNC) {
2981             /* whitespace between macro name and argument list */
2982             TokenString ws_str;
2983             tok_str_new(&ws_str);
2984
2985             spc = 0;
2986             parse_flags |= PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED
2987                 | PARSE_FLAG_ACCEPT_STRAYS;
2988
2989             /* get next token from argument stream */
2990             t = next_argstream(nested_list, can_read_stream, &ws_str);
2991             if (t != '(') {
2992                 /* not a macro substitution after all, restore the
2993                  * macro token plus all whitespace we've read.
2994                  * whitespace is intentionally not merged to preserve
2995                  * newlines. */
2996                 parse_flags = saved_parse_flags;
2997                 tok_str_add(tok_str, tok);
2998                 if (parse_flags & PARSE_FLAG_SPACES) {
2999                     int i;
3000                     for (i = 0; i < ws_str.len; i++)
3001                         tok_str_add(tok_str, ws_str.str[i]);
3002                 }
3003                 tok_str_free(ws_str.str);
3004                 return 0;
3005             } else {
3006                 tok_str_free(ws_str.str);
3007             }
3008             next_nomacro(); /* eat '(' */
3009
3010             /* argument macro */
3011             args = NULL;
3012             sa = s->next;
3013             /* NOTE: empty args are allowed, except if no args */
3014             for(;;) {
3015                 do {
3016                     next_argstream(nested_list, can_read_stream, NULL);
3017                 } while (is_space(tok) || TOK_LINEFEED == tok);
3018     empty_arg:
3019                 /* handle '()' case */
3020                 if (!args && !sa && tok == ')')
3021                     break;
3022                 if (!sa)
3023                     tcc_error("macro '%s' used with too many args",
3024                           get_tok_str(s->v, 0));
3025                 tok_str_new(&str);
3026                 parlevel = spc = 0;
3027                 /* NOTE: non zero sa->t indicates VA_ARGS */
3028                 while ((parlevel > 0 ||
3029                         (tok != ')' &&
3030                          (tok != ',' || sa->type.t)))) {
3031                     if (tok == TOK_EOF || tok == 0)
3032                         break;
3033                     if (tok == '(')
3034                         parlevel++;
3035                     else if (tok == ')')
3036                         parlevel--;
3037                     if (tok == TOK_LINEFEED)
3038                         tok = ' ';
3039                     if (!check_space(tok, &spc))
3040                         tok_str_add2(&str, tok, &tokc);
3041                     next_argstream(nested_list, can_read_stream, NULL);
3042                 }
3043                 if (parlevel)
3044                     expect(")");
3045                 str.len -= spc;
3046                 tok_str_add(&str, 0);
3047                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
3048                 sa1->d = str.str;
3049                 sa = sa->next;
3050                 if (tok == ')') {
3051                     /* special case for gcc var args: add an empty
3052                        var arg argument if it is omitted */
3053                     if (sa && sa->type.t && gnu_ext)
3054                         goto empty_arg;
3055                     break;
3056                 }
3057                 if (tok != ',')
3058                     expect(",");
3059             }
3060             if (sa) {
3061                 tcc_error("macro '%s' used with too few args",
3062                       get_tok_str(s->v, 0));
3063             }
3064
3065             parse_flags = saved_parse_flags;
3066
3067             /* now subst each arg */
3068             mstr = macro_arg_subst(nested_list, mstr, args);
3069             /* free memory */
3070             sa = args;
3071             while (sa) {
3072                 sa1 = sa->prev;
3073                 tok_str_free(sa->d);
3074                 sym_free(sa);
3075                 sa = sa1;
3076             }
3077         }
3078
3079         sym_push2(nested_list, s->v, 0, 0);
3080         parse_flags = saved_parse_flags;
3081         macro_subst(tok_str, nested_list, mstr, can_read_stream);
3082
3083         /* pop nested defined symbol */
3084         sa1 = *nested_list;
3085         *nested_list = sa1->prev;
3086         sym_free(sa1);
3087         if (mstr != s->d)
3088             tok_str_free(mstr);
3089     }
3090     return 0;
3091 }
3092
3093 int paste_tokens(int t1, CValue *v1, int t2, CValue *v2)
3094 {
3095     CString cstr;
3096     int n;
3097
3098     cstr_new(&cstr);
3099     if (t1 != TOK_PLCHLDR)
3100         cstr_cat(&cstr, get_tok_str(t1, v1));
3101     n = cstr.size;
3102     if (t2 != TOK_PLCHLDR)
3103         cstr_cat(&cstr, get_tok_str(t2, v2));
3104     cstr_ccat(&cstr, '\0');
3105
3106     tcc_open_bf(tcc_state, ":paste:", cstr.size);
3107     memcpy(file->buffer, cstr.data, cstr.size);
3108     for (;;) {
3109         next_nomacro1();
3110         if (0 == *file->buf_ptr)
3111             break;
3112         if (is_space(tok))
3113             continue;
3114         tcc_warning("pasting <%.*s> and <%s> does not give a valid preprocessing token",
3115             n, cstr.data, (char*)cstr.data + n);
3116         break;
3117     }
3118     tcc_close();
3119
3120     //printf("paste <%s>\n", (char*)cstr.data);
3121     cstr_free(&cstr);
3122     return 0;
3123 }
3124
3125 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
3126    return the resulting string (which must be freed). */
3127 static inline int *macro_twosharps(const int *ptr0)
3128 {
3129     int t;
3130     CValue cval;
3131     TokenString macro_str1;
3132     int start_of_nosubsts = -1;
3133     const int *ptr;
3134
3135     /* we search the first '##' */
3136     for (ptr = ptr0;;) {
3137         TOK_GET(&t, &ptr, &cval);
3138         if (t == TOK_TWOSHARPS)
3139             break;
3140         if (t == 0)
3141             return NULL;
3142     }
3143
3144     tok_str_new(&macro_str1);
3145
3146     //tok_print(" $$$", ptr0);
3147     for (ptr = ptr0;;) {
3148         TOK_GET(&t, &ptr, &cval);
3149         if (t == 0)
3150             break;
3151         if (t == TOK_TWOSHARPS)
3152             continue;
3153         while (*ptr == TOK_TWOSHARPS) {
3154             int t1; CValue cv1;
3155             /* given 'a##b', remove nosubsts preceding 'a' */
3156             if (start_of_nosubsts >= 0)
3157                 macro_str1.len = start_of_nosubsts;
3158             /* given 'a##b', remove nosubsts preceding 'b' */
3159             while ((t1 = *++ptr) == TOK_NOSUBST)
3160                 ;
3161             if (t1 && t1 != TOK_TWOSHARPS
3162                 && t1 != ':') /* 'a##:' don't build a new token */
3163             {
3164                 TOK_GET(&t1, &ptr, &cv1);
3165                 if (t != TOK_PLCHLDR || t1 != TOK_PLCHLDR) {
3166                     paste_tokens(t, &cval, t1, &cv1);
3167                     t = tok, cval = tokc;
3168                 }
3169             }
3170         }
3171         if (t == TOK_NOSUBST) {
3172             if (start_of_nosubsts < 0)
3173                 start_of_nosubsts = macro_str1.len;
3174         } else {
3175             start_of_nosubsts = -1;
3176         }
3177         tok_str_add2(&macro_str1, t, &cval);
3178     }
3179     tok_str_add(&macro_str1, 0);
3180     //tok_print(" ###", macro_str1.str);
3181     return macro_str1.str;
3182 }
3183
3184 /* do macro substitution of macro_str and add result to
3185    (tok_str,tok_len). 'nested_list' is the list of all macros we got
3186    inside to avoid recursing. */
3187 static void macro_subst(
3188     TokenString *tok_str,
3189     Sym **nested_list,
3190     const int *macro_str,
3191     int can_read_stream
3192     )
3193 {
3194     Sym *s;
3195     const int *ptr;
3196     int t, spc, nosubst;
3197     CValue cval;
3198     int *macro_str1 = NULL;
3199
3200     /* first scan for '##' operator handling */
3201     ptr = macro_str;
3202     spc = nosubst = 0;
3203
3204     /* first scan for '##' operator handling */
3205     if (can_read_stream) {
3206         macro_str1 = macro_twosharps(ptr);
3207         if (macro_str1)
3208             ptr = macro_str1;
3209     }
3210
3211     while (1) {
3212         TOK_GET(&t, &ptr, &cval);
3213         if (t == 0)
3214             break;
3215
3216         if (t >= TOK_IDENT && 0 == nosubst) {
3217             s = define_find(t);
3218             if (s == NULL)
3219                 goto no_subst;
3220
3221             /* if nested substitution, do nothing */
3222             if (sym_find2(*nested_list, t)) {
3223                 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
3224                 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3225                 goto no_subst;
3226             }
3227
3228             {
3229                 TokenString str;
3230                 str.str = (int*)ptr;
3231                 begin_macro(&str, 2);
3232
3233                 tok = t;
3234                 macro_subst_tok(tok_str, nested_list, s, can_read_stream);
3235
3236                 if (str.alloc == 3) {
3237                     /* already finished by reading function macro arguments */
3238                     break;
3239                 }
3240
3241                 ptr = macro_ptr;
3242                 end_macro ();
3243             }
3244
3245             spc = (tok_str->len &&
3246                    is_space(tok_last(tok_str->str,
3247                                      tok_str->str + tok_str->len)));
3248
3249         } else {
3250
3251             if (t == '\\' && !(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
3252                 tcc_error("stray '\\' in program");
3253
3254 no_subst:
3255             if (!check_space(t, &spc))
3256                 tok_str_add2(tok_str, t, &cval);
3257             nosubst = 0;
3258             if (t == TOK_NOSUBST)
3259                 nosubst = 1;
3260         }
3261     }
3262     if (macro_str1)
3263         tok_str_free(macro_str1);
3264
3265 }
3266
3267 /* return next token with macro substitution */
3268 ST_FUNC void next(void)
3269 {
3270  redo:
3271     if (parse_flags & PARSE_FLAG_SPACES)
3272         next_nomacro_spc();
3273     else
3274         next_nomacro();
3275
3276     if (macro_ptr) {
3277         if (tok == TOK_NOSUBST || tok == TOK_PLCHLDR) {
3278         /* discard preprocessor markers */
3279             goto redo;
3280         } else if (tok == 0) {
3281             /* end of macro or unget token string */
3282             end_macro();
3283             goto redo;
3284         }
3285     } else if (tok >= TOK_IDENT && (parse_flags & PARSE_FLAG_PREPROCESS)) {
3286         Sym *s;
3287         /* if reading from file, try to substitute macros */
3288         s = define_find(tok);
3289         if (s) {
3290             static TokenString str; /* using static string for speed */
3291             Sym *nested_list = NULL;
3292             tok_str_new(&str);
3293             nested_list = NULL;
3294             macro_subst_tok(&str, &nested_list, s, 1);
3295             tok_str_add(&str, 0);
3296             begin_macro(&str, 0);
3297             goto redo;
3298         }
3299     }
3300     /* convert preprocessor tokens into C tokens */
3301     if (tok == TOK_PPNUM) {
3302         if  (parse_flags & PARSE_FLAG_TOK_NUM)
3303             parse_number((char *)tokc.cstr->data);
3304     } else if (tok == TOK_PPSTR) {
3305         if (parse_flags & PARSE_FLAG_TOK_STR)
3306             parse_string((char *)tokc.cstr->data, tokc.cstr->size - 1);
3307     }
3308 }
3309
3310 /* push back current token and set current token to 'last_tok'. Only
3311    identifier case handled for labels. */
3312 ST_INLN void unget_tok(int last_tok)
3313 {
3314     TokenString *str = tcc_malloc(sizeof *str);
3315     tok_str_new(str);
3316     tok_str_add2(str, tok, &tokc);
3317     tok_str_add(str, 0);
3318     begin_macro(str, 1);
3319     tok = last_tok;
3320 }
3321
3322 /* better than nothing, but needs extension to handle '-E' option
3323    correctly too */
3324 ST_FUNC void preprocess_init(TCCState *s1)
3325 {
3326     s1->include_stack_ptr = s1->include_stack;
3327     /* XXX: move that before to avoid having to initialize
3328        file->ifdef_stack_ptr ? */
3329     s1->ifdef_stack_ptr = s1->ifdef_stack;
3330     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3331
3332     pvtop = vtop = vstack - 1;
3333     s1->pack_stack[0] = 0;
3334     s1->pack_stack_ptr = s1->pack_stack;
3335
3336     isidnum_table['$' - CH_EOF] =
3337         tcc_state->dollars_in_identifiers ? IS_ID : 0;
3338 }
3339
3340 ST_FUNC void preprocess_new(void)
3341 {
3342     int i, c;
3343     const char *p, *r;
3344
3345     /* init isid table */
3346     for(i = CH_EOF; i<256; i++)
3347         isidnum_table[i - CH_EOF]
3348             = is_space(i) ? IS_SPC
3349             : isid(i) ? IS_ID
3350             : isnum(i) ? IS_NUM
3351             : 0;
3352
3353     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3354
3355     tok_ident = TOK_IDENT;
3356     p = tcc_keywords;
3357     while (*p) {
3358         r = p;
3359         for(;;) {
3360             c = *r++;
3361             if (c == '\0')
3362                 break;
3363         }
3364         tok_alloc(p, r - p - 1);
3365         p = r;
3366     }
3367 }
3368
3369 ST_FUNC void preprocess_delete(void)
3370 {
3371     int i, n;
3372
3373     /* free -D and compiler defines */
3374     free_defines(NULL);
3375
3376     /* cleanup from error/setjmp */
3377     while (macro_stack)
3378         end_macro();
3379     macro_ptr = NULL;
3380
3381     /* free tokens */
3382     n = tok_ident - TOK_IDENT;
3383     for(i = 0; i < n; i++)
3384         tcc_free(table_ident[i]);
3385     tcc_free(table_ident);
3386     table_ident = NULL;
3387 }
3388
3389 /* Preprocess the current file */
3390 ST_FUNC int tcc_preprocess(TCCState *s1)
3391 {
3392     BufferedFile **iptr;
3393     int token_seen, spcs, level;
3394
3395     preprocess_init(s1);
3396     ch = file->buf_ptr[0];
3397     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3398     parse_flags = PARSE_FLAG_PREPROCESS
3399                 | (parse_flags & PARSE_FLAG_ASM_FILE)
3400                 | PARSE_FLAG_LINEFEED
3401                 | PARSE_FLAG_SPACES
3402                 | PARSE_FLAG_ACCEPT_STRAYS
3403                 ;
3404
3405 #ifdef PP_BENCH
3406     do next(); while (tok != TOK_EOF); return 0;
3407 #endif
3408
3409     token_seen = spcs = 0;
3410     pp_line(s1, file, 0);
3411
3412     for (;;) {
3413         iptr = s1->include_stack_ptr;
3414         next();
3415         if (tok == TOK_EOF)
3416             break;
3417         level = s1->include_stack_ptr - iptr;
3418         if (level) {
3419             if (level > 0)
3420                 pp_line(s1, *iptr, 0);
3421             pp_line(s1, file, level);
3422         }
3423
3424         if (0 == token_seen) {
3425             if (tok == ' ') {
3426                 ++spcs;
3427                 continue;
3428             }
3429             if (tok == TOK_LINEFEED) {
3430                 spcs = 0;
3431                 continue;
3432             }
3433             pp_line(s1, file, 0);
3434             while (spcs)
3435                 fputs(" ", s1->ppfp), --spcs;
3436             token_seen = 1;
3437
3438         } else if (tok == TOK_LINEFEED) {
3439             ++file->line_ref;
3440             token_seen = 0;
3441         }
3442
3443         fputs(get_tok_str(tok, &tokc), s1->ppfp);
3444     }
3445
3446     return 0;
3447 }