tccpp.c

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  61 /* outdated -- gr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