tccpp.c

   1 /*
   2  *  TCC - Tiny C Compiler
   3  *
   4  *  Copyright (c) 2001-2004 Fabrice Bellard
   5  *
   6  * This library is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2 of the License, or (at your option) any later version.
  10  *
  11  * This library is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with this library; if not, write to the Free Software
  18  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  19  */
  20
  21 #include "tcc.h"
  22
  23 /********************************************************/
  24 /* global variables */
  25
  26 ST_DATA int tok_flags;
  27 ST_DATA int parse_flags;
  28
  29 ST_DATA struct BufferedFile *file;
  30 ST_DATA int ch, tok;
  31 ST_DATA CValue tokc;
  32 ST_DATA const int *macro_ptr;
  33 ST_DATA CString tokcstr; /* current parsed string, if any */
  34
  35 /* display benchmark infos */
  36 ST_DATA int total_lines;
  37 ST_DATA int total_bytes;
  38 ST_DATA int tok_ident;
  39 ST_DATA TokenSym **table_ident;
  40
  41 /* ------------------------------------------------------------------------- */
  42
  43 static int *macro_ptr_allocated;
  44 static const int *unget_saved_macro_ptr;
  45 static int unget_saved_buffer[TOK_MAX_SIZE + 1];
  46 static int unget_buffer_enabled;
  47 static TokenSym *hash_ident[TOK_HASH_SIZE];
  48 static char token_buf[STRING_MAX_SIZE + 1];
  49 /* true if isid(c) || isnum(c) */
  50 static unsigned char isidnum_table[256-CH_EOF];
  51
  52 static const char tcc_keywords[] =
  53 #define DEF(id, str) str "\0"
  54 #include "tcctok.h"
  55 #undef DEF
  56 ;
  57
  58 /* WARNING: the content of this string encodes token numbers */
  59 static const unsigned char tok_two_chars[] =
  60 /* outdated -- gr
  61     "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
  62     "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
  63 */{
  64     '<','=', TOK_LE,
  65     '>','=', TOK_GE,
  66     '!','=', TOK_NE,
  67     '&','&', TOK_LAND,
  68     '|','|', TOK_LOR,
  69     '+','+', TOK_INC,
  70     '-','-', TOK_DEC,
  71     '=','=', TOK_EQ,
  72     '<','<', TOK_SHL,
  73     '>','>', TOK_SAR,
  74     '+','=', TOK_A_ADD,
  75     '-','=', TOK_A_SUB,
  76     '*','=', TOK_A_MUL,
  77     '/','=', TOK_A_DIV,
  78     '%','=', TOK_A_MOD,
  79     '&','=', TOK_A_AND,
  80     '^','=', TOK_A_XOR,
  81     '|','=', TOK_A_OR,
  82     '-','>', TOK_ARROW,
  83     '.','.', 0xa8, // C++ token ?
  84     '#','#', TOK_TWOSHARPS,
  85     '\\','#', TOK_QHASH,
  86     0
  87 };
  88
  89 struct macro_level {
  90     struct macro_level *prev;
  91     const int *p;
  92 };
  93
  94 static void next_nomacro_spc(void);
  95 static void macro_subst(
  96     TokenString *tok_str,
  97     Sym **nested_list,
  98     const int *macro_str,
  99     struct macro_level **can_read_stream
 100     );
 101
 102 ST_FUNC void skip(int c)
 103 {
 104     if (tok != c)
 105         tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
 106     next();
 107 }
 108
 109 ST_FUNC void expect(const char *msg)
 110 {
 111     tcc_error("%s expected", msg);
 112 }
 113
 114 /* ------------------------------------------------------------------------- */
 115 /* CString handling */
 116 static void cstr_realloc(CString *cstr, int new_size)
 117 {
 118     int size;
 119     void *data;
 120
 121     size = cstr->size_allocated;
 122     if (size == 0)
 123         size = 8; /* no need to allocate a too small first string */
 124     while (size < new_size)
 125         size = size * 2;
 126     data = tcc_realloc(cstr->data_allocated, size);
 127     cstr->data_allocated = data;
 128     cstr->size_allocated = size;
 129     cstr->data = data;
 130 }
 131
 132 /* add a byte */
 133 ST_FUNC void cstr_ccat(CString *cstr, int ch)
 134 {
 135     int size;
 136     size = cstr->size + 1;
 137     if (size > cstr->size_allocated)
 138         cstr_realloc(cstr, size);
 139     ((unsigned char *)cstr->data)[size - 1] = ch;
 140     cstr->size = size;
 141 }
 142
 143 ST_FUNC void cstr_cat(CString *cstr, const char *str)
 144 {
 145     int c;
 146     for(;;) {
 147         c = *str;
 148         if (c == '\0')
 149             break;
 150         cstr_ccat(cstr, c);
 151         str++;
 152     }
 153 }
 154
 155 /* add a wide char */
 156 ST_FUNC void cstr_wccat(CString *cstr, int ch)
 157 {
 158     int size;
 159     size = cstr->size + sizeof(nwchar_t);
 160     if (size > cstr->size_allocated)
 161         cstr_realloc(cstr, size);
 162     *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
 163     cstr->size = size;
 164 }
 165
 166 ST_FUNC void cstr_new(CString *cstr)
 167 {
 168     memset(cstr, 0, sizeof(CString));
 169 }
 170
 171 /* free string and reset it to NULL */
 172 ST_FUNC void cstr_free(CString *cstr)
 173 {
 174     tcc_free(cstr->data_allocated);
 175     cstr_new(cstr);
 176 }
 177
 178 /* reset string to empty */
 179 ST_FUNC void cstr_reset(CString *cstr)
 180 {
 181     cstr->size = 0;
 182 }
 183
 184 /* XXX: unicode ? */
 185 static void add_char(CString *cstr, int c)
 186 {
 187     if (c == '\'' || c == '\"' || c == '\\') {
 188         /* XXX: could be more precise if char or string */
 189         cstr_ccat(cstr, '\\');
 190     }
 191     if (c >= 32 && c <= 126) {
 192         cstr_ccat(cstr, c);
 193     } else {
 194         cstr_ccat(cstr, '\\');
 195         if (c == '\n') {
 196             cstr_ccat(cstr, 'n');
 197         } else {
 198             cstr_ccat(cstr, '0' + ((c >> 6) & 7));
 199             cstr_ccat(cstr, '0' + ((c >> 3) & 7));
 200             cstr_ccat(cstr, '0' + (c & 7));
 201         }
 202     }
 203 }
 204
 205 /* ------------------------------------------------------------------------- */
 206 /* allocate a new token */
 207 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
 208 {
 209     TokenSym *ts, **ptable;
 210     int i;
 211
 212     if (tok_ident >= SYM_FIRST_ANOM)
 213         tcc_error("memory full (symbols)");
 214
 215     /* expand token table if needed */
 216     i = tok_ident - TOK_IDENT;
 217     if ((i % TOK_ALLOC_INCR) == 0) {
 218         ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
 219         table_ident = ptable;
 220     }
 221
 222     ts = tcc_malloc(sizeof(TokenSym) + len);
 223     table_ident[i] = ts;
 224     ts->tok = tok_ident++;
 225     ts->sym_define = NULL;
 226     ts->sym_label = NULL;
 227     ts->sym_struct = NULL;
 228     ts->sym_identifier = NULL;
 229     ts->len = len;
 230     ts->hash_next = NULL;
 231     memcpy(ts->str, str, len);
 232     ts->str[len] = '\0';
 233     *pts = ts;
 234     return ts;
 235 }
 236
 237 #define TOK_HASH_INIT 1
 238 #define TOK_HASH_FUNC(h, c) ((h) * 263 + (c))
 239
 240 /* find a token and add it if not found */
 241 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
 242 {
 243     TokenSym *ts, **pts;
 244     int i;
 245     unsigned int h;
 246
 247     h = TOK_HASH_INIT;
 248     for(i=0;i<len;i++)
 249         h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
 250     h &= (TOK_HASH_SIZE - 1);
 251
 252     pts = &hash_ident[h];
 253     for(;;) {
 254         ts = *pts;
 255         if (!ts)
 256             break;
 257         if (ts->len == len && !memcmp(ts->str, str, len))
 258             return ts;
 259         pts = &(ts->hash_next);
 260     }
 261     return tok_alloc_new(pts, str, len);
 262 }
 263
 264 /* XXX: buffer overflow */
 265 /* XXX: float tokens */
 266 ST_FUNC char *get_tok_str(int v, CValue *cv)
 267 {
 268     static char buf[STRING_MAX_SIZE + 1];
 269     static CString cstr_buf;
 270     CString *cstr;
 271     char *p;
 272     int i, len;
 273
 274     /* NOTE: to go faster, we give a fixed buffer for small strings */
 275     cstr_reset(&cstr_buf);
 276     cstr_buf.data = buf;
 277     cstr_buf.size_allocated = sizeof(buf);
 278     p = buf;
 279
 280 /* just an explanation, should never happen:
 281     if (v <= TOK_LINENUM && v >= TOK_CINT && cv == NULL)
 282         tcc_error("internal error: get_tok_str"); */
 283
 284     switch(v) {
 285     case TOK_CINT:
 286     case TOK_CUINT:
 287         /* XXX: not quite exact, but only useful for testing */
 288         sprintf(p, "%u", cv->ui);
 289         break;
 290     case TOK_CLLONG:
 291     case TOK_CULLONG:
 292         /* XXX: not quite exact, but only useful for testing  */
 293 #ifdef _WIN32
 294         sprintf(p, "%u", (unsigned)cv->ull);
 295 #else
 296         sprintf(p, "%llu", cv->ull);
 297 #endif
 298         break;
 299     case TOK_LCHAR:
 300         cstr_ccat(&cstr_buf, 'L');
 301     case TOK_CCHAR:
 302         cstr_ccat(&cstr_buf, '\'');
 303         add_char(&cstr_buf, cv->i);
 304         cstr_ccat(&cstr_buf, '\'');
 305         cstr_ccat(&cstr_buf, '\0');
 306         break;
 307     case TOK_PPNUM:
 308         cstr = cv->cstr;
 309         len = cstr->size - 1;
 310         for(i=0;i<len;i++)
 311             add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 312         cstr_ccat(&cstr_buf, '\0');
 313         break;
 314     case TOK_LSTR:
 315         cstr_ccat(&cstr_buf, 'L');
 316     case TOK_STR:
 317         cstr = cv->cstr;
 318         cstr_ccat(&cstr_buf, '\"');
 319         if (v == TOK_STR) {
 320             len = cstr->size - 1;
 321             for(i=0;i<len;i++)
 322                 add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 323         } else {
 324             len = (cstr->size / sizeof(nwchar_t)) - 1;
 325             for(i=0;i<len;i++)
 326                 add_char(&cstr_buf, ((nwchar_t *)cstr->data)[i]);
 327         }
 328         cstr_ccat(&cstr_buf, '\"');
 329         cstr_ccat(&cstr_buf, '\0');
 330         break;
 331
 332     case TOK_CFLOAT:
 333     case TOK_CDOUBLE:
 334     case TOK_CLDOUBLE:
 335     case TOK_LINENUM:
 336         return NULL; /* should not happen */
 337
 338     /* above tokens have value, the ones below don't */
 339
 340     case TOK_LT:
 341         v = '<';
 342         goto addv;
 343     case TOK_GT:
 344         v = '>';
 345         goto addv;
 346     case TOK_DOTS:
 347         return strcpy(p, "...");
 348     case TOK_A_SHL:
 349         return strcpy(p, "<<=");
 350     case TOK_A_SAR:
 351         return strcpy(p, ">>=");
 352     default:
 353         if (v < TOK_IDENT) {
 354             /* search in two bytes table */
 355             const unsigned char *q = tok_two_chars;
 356             while (*q) {
 357                 if (q[2] == v) {
 358                     *p++ = q[0];
 359                     *p++ = q[1];
 360                     *p = '\0';
 361                     return buf;
 362                 }
 363                 q += 3;
 364             }
 365         addv:
 366             *p++ = v;
 367             *p = '\0';
 368         } else if (v < tok_ident) {
 369             return table_ident[v - TOK_IDENT]->str;
 370         } else if (v >= SYM_FIRST_ANOM) {
 371             /* special name for anonymous symbol */
 372             sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
 373         } else {
 374             /* should never happen */
 375             return NULL;
 376         }
 377         break;
 378     }
 379     return cstr_buf.data;
 380 }
 381
 382 /* fill input buffer and peek next char */
 383 static int tcc_peekc_slow(BufferedFile *bf)
 384 {
 385     int len;
 386     /* only tries to read if really end of buffer */
 387     if (bf->buf_ptr >= bf->buf_end) {
 388         if (bf->fd != -1) {
 389 #if defined(PARSE_DEBUG)
 390             len = 1;
 391 #else
 392             len = IO_BUF_SIZE;
 393 #endif
 394             len = read(bf->fd, bf->buffer, len);
 395             if (len < 0)
 396                 len = 0;
 397         } else {
 398             len = 0;
 399         }
 400         total_bytes += len;
 401         bf->buf_ptr = bf->buffer;
 402         bf->buf_end = bf->buffer + len;
 403         *bf->buf_end = CH_EOB;
 404     }
 405     if (bf->buf_ptr < bf->buf_end) {
 406         return bf->buf_ptr[0];
 407     } else {
 408         bf->buf_ptr = bf->buf_end;
 409         return CH_EOF;
 410     }
 411 }
 412
 413 /* return the current character, handling end of block if necessary
 414    (but not stray) */
 415 ST_FUNC int handle_eob(void)
 416 {
 417     return tcc_peekc_slow(file);
 418 }
 419
 420 /* read next char from current input file and handle end of input buffer */
 421 ST_INLN void inp(void)
 422 {
 423     ch = *(++(file->buf_ptr));
 424     /* end of buffer/file handling */
 425     if (ch == CH_EOB)
 426         ch = handle_eob();
 427 }
 428
 429 /* handle '\[\r]\n' */
 430 static int handle_stray_noerror(void)
 431 {
 432     while (ch == '\\') {
 433         inp();
 434         if (ch == '\n') {
 435             file->line_num++;
 436             inp();
 437         } else if (ch == '\r') {
 438             inp();
 439             if (ch != '\n')
 440                 goto fail;
 441             file->line_num++;
 442             inp();
 443         } else if (ch == '#') {
 444             ch = TOK_QHASH;
 445         } else {
 446         fail:
 447             return 1;
 448         }
 449     }
 450     return 0;
 451 }
 452
 453 static void handle_stray(void)
 454 {
 455     if (handle_stray_noerror())
 456         tcc_error("stray '\\' in program");
 457 }
 458
 459 /* skip the stray and handle the \\n case. Output an error if
 460    incorrect char after the stray */
 461 static int handle_stray1(uint8_t *p)
 462 {
 463     int c;
 464
 465     if (p >= file->buf_end) {
 466         file->buf_ptr = p;
 467         c = handle_eob();
 468         p = file->buf_ptr;
 469         if (c == '\\')
 470             goto parse_stray;
 471     } else {
 472     parse_stray:
 473         file->buf_ptr = p;
 474         ch = *p;
 475         handle_stray();
 476         p = file->buf_ptr;
 477         c = *p;
 478     }
 479     return c;
 480 }
 481
 482 /* handle just the EOB case, but not stray */
 483 #define PEEKC_EOB(c, p)\
 484 {\
 485     p++;\
 486     c = *p;\
 487     if (c == '\\') {\
 488         file->buf_ptr = p;\
 489         c = handle_eob();\
 490         p = file->buf_ptr;\
 491     }\
 492 }
 493
 494 /* handle the complicated stray case */
 495 #define PEEKC(c, p)\
 496 {\
 497     p++;\
 498     c = *p;\
 499     if (c == '\\') {\
 500         c = handle_stray1(p);\
 501         p = file->buf_ptr;\
 502     }\
 503 }
 504
 505 /* input with '\[\r]\n' handling. Note that this function cannot
 506    handle other characters after '\', so you cannot call it inside
 507    strings or comments */
 508 ST_FUNC void minp(void)
 509 {
 510     inp();
 511     if (ch == '\\')
 512         handle_stray();
 513 }
 514
 515
 516 /* single line C++ comments */
 517 static uint8_t *parse_line_comment(uint8_t *p)
 518 {
 519     int c;
 520
 521     p++;
 522     for(;;) {
 523         c = *p;
 524     redo:
 525         if (c == '\n' || c == CH_EOF) {
 526             break;
 527         } else if (c == '\\') {
 528             file->buf_ptr = p;
 529             c = handle_eob();
 530             p = file->buf_ptr;
 531             if (c == '\\') {
 532                 PEEKC_EOB(c, p);
 533                 if (c == '\n') {
 534                     file->line_num++;
 535                     PEEKC_EOB(c, p);
 536                 } else if (c == '\r') {
 537                     PEEKC_EOB(c, p);
 538                     if (c == '\n') {
 539                         file->line_num++;
 540                         PEEKC_EOB(c, p);
 541                     }
 542                 }
 543             } else {
 544                 goto redo;
 545             }
 546         } else {
 547             p++;
 548         }
 549     }
 550     return p;
 551 }
 552
 553 /* C comments */
 554 ST_FUNC uint8_t *parse_comment(uint8_t *p)
 555 {
 556     int c;
 557
 558     p++;
 559     for(;;) {
 560         /* fast skip loop */
 561         for(;;) {
 562             c = *p;
 563             if (c == '\n' || c == '*' || c == '\\')
 564                 break;
 565             p++;
 566             c = *p;
 567             if (c == '\n' || c == '*' || c == '\\')
 568                 break;
 569             p++;
 570         }
 571         /* now we can handle all the cases */
 572         if (c == '\n') {
 573             file->line_num++;
 574             p++;
 575         } else if (c == '*') {
 576             p++;
 577             for(;;) {
 578                 c = *p;
 579                 if (c == '*') {
 580                     p++;
 581                 } else if (c == '/') {
 582                     goto end_of_comment;
 583                 } else if (c == '\\') {
 584                     file->buf_ptr = p;
 585                     c = handle_eob();
 586                     p = file->buf_ptr;
 587                     if (c == '\\') {
 588                         /* skip '\[\r]\n', otherwise just skip the stray */
 589                         while (c == '\\') {
 590                             PEEKC_EOB(c, p);
 591                             if (c == '\n') {
 592                                 file->line_num++;
 593                                 PEEKC_EOB(c, p);
 594                             } else if (c == '\r') {
 595                                 PEEKC_EOB(c, p);
 596                                 if (c == '\n') {
 597                                     file->line_num++;
 598                                     PEEKC_EOB(c, p);
 599                                 }
 600                             } else {
 601                                 goto after_star;
 602                             }
 603                         }
 604                     }
 605                 } else {
 606                     break;
 607                 }
 608             }
 609         after_star: ;
 610         } else {
 611             /* stray, eob or eof */
 612             file->buf_ptr = p;
 613             c = handle_eob();
 614             p = file->buf_ptr;
 615             if (c == CH_EOF) {
 616                 tcc_error("unexpected end of file in comment");
 617             } else if (c == '\\') {
 618                 p++;
 619             }
 620         }
 621     }
 622  end_of_comment:
 623     p++;
 624     return p;
 625 }
 626
 627 #define cinp minp
 628
 629 static inline void skip_spaces(void)
 630 {
 631     while (is_space(ch))
 632         cinp();
 633 }
 634
 635 static inline int check_space(int t, int *spc)
 636 {
 637     if (is_space(t)) {
 638         if (*spc)
 639             return 1;
 640         *spc = 1;
 641     } else
 642         *spc = 0;
 643     return 0;
 644 }
 645
 646 /* parse a string without interpreting escapes */
 647 static uint8_t *parse_pp_string(uint8_t *p,
 648                                 int sep, CString *str)
 649 {
 650     int c;
 651     p++;
 652     for(;;) {
 653         c = *p;
 654         if (c == sep) {
 655             break;
 656         } else if (c == '\\') {
 657             file->buf_ptr = p;
 658             c = handle_eob();
 659             p = file->buf_ptr;
 660             if (c == CH_EOF) {
 661             unterminated_string:
 662                 /* XXX: indicate line number of start of string */
 663                 tcc_error("missing terminating %c character", sep);
 664             } else if (c == '\\') {
 665                 /* escape : just skip \[\r]\n */
 666                 PEEKC_EOB(c, p);
 667                 if (c == '\n') {
 668                     file->line_num++;
 669                     p++;
 670                 } else if (c == '\r') {
 671                     PEEKC_EOB(c, p);
 672                     if (c != '\n')
 673                         expect("'\n' after '\r'");
 674                     file->line_num++;
 675                     p++;
 676                 } else if (c == CH_EOF) {
 677                     goto unterminated_string;
 678                 } else {
 679                     if (str) {
 680                         cstr_ccat(str, '\\');
 681                         cstr_ccat(str, c);
 682                     }
 683                     p++;
 684                 }
 685             }
 686         } else if (c == '\n') {
 687             file->line_num++;
 688             goto add_char;
 689         } else if (c == '\r') {
 690             PEEKC_EOB(c, p);
 691             if (c != '\n') {
 692                 if (str)
 693                     cstr_ccat(str, '\r');
 694             } else {
 695                 file->line_num++;
 696                 goto add_char;
 697             }
 698         } else {
 699         add_char:
 700             if (str)
 701                 cstr_ccat(str, c);
 702             p++;
 703         }
 704     }
 705     p++;
 706     return p;
 707 }
 708
 709 /* skip block of text until #else, #elif or #endif. skip also pairs of
 710    #if/#endif */
 711 static void preprocess_skip(void)
 712 {
 713     int a, start_of_line, c, in_warn_or_error;
 714     uint8_t *p;
 715
 716     p = file->buf_ptr;
 717     a = 0;
 718 redo_start:
 719     start_of_line = 1;
 720     in_warn_or_error = 0;
 721     for(;;) {
 722     redo_no_start:
 723         c = *p;
 724         switch(c) {
 725         case ' ':
 726         case '\t':
 727         case '\f':
 728         case '\v':
 729         case '\r':
 730             p++;
 731             goto redo_no_start;
 732         case '\n':
 733             file->line_num++;
 734             p++;
 735             goto redo_start;
 736         case '\\':
 737             file->buf_ptr = p;
 738             c = handle_eob();
 739             if (c == CH_EOF) {
 740                 expect("#endif");
 741             } else if (c == '\\') {
 742                 ch = file->buf_ptr[0];
 743                 handle_stray_noerror();
 744             }
 745             p = file->buf_ptr;
 746             goto redo_no_start;
 747         /* skip strings */
 748         case '\"':
 749         case '\'':
 750             if (in_warn_or_error)
 751                 goto _default;
 752             p = parse_pp_string(p, c, NULL);
 753             break;
 754         /* skip comments */
 755         case '/':
 756             if (in_warn_or_error)
 757                 goto _default;
 758             file->buf_ptr = p;
 759             ch = *p;
 760             minp();
 761             p = file->buf_ptr;
 762             if (ch == '*') {
 763                 p = parse_comment(p);
 764             } else if (ch == '/') {
 765                 p = parse_line_comment(p);
 766             }
 767             break;
 768         case '#':
 769             p++;
 770             if (start_of_line) {
 771                 file->buf_ptr = p;
 772                 next_nomacro();
 773                 p = file->buf_ptr;
 774                 if (a == 0 &&
 775                     (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
 776                     goto the_end;
 777                 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
 778                     a++;
 779                 else if (tok == TOK_ENDIF)
 780                     a--;
 781                 else if( tok == TOK_ERROR || tok == TOK_WARNING)
 782                     in_warn_or_error = 1;
 783                 else if (tok == TOK_LINEFEED)
 784                     goto redo_start;
 785                 else if (parse_flags & PARSE_FLAG_ASM_FILE)
 786                     p = parse_line_comment(p);
 787             }
 788             else if (parse_flags & PARSE_FLAG_ASM_FILE)
 789                 p = parse_line_comment(p);
 790             break;
 791 _default:
 792         default:
 793             p++;
 794             break;
 795         }
 796         start_of_line = 0;
 797     }
 798  the_end: ;
 799     file->buf_ptr = p;
 800 }
 801
 802 /* ParseState handling */
 803
 804 /* XXX: currently, no include file info is stored. Thus, we cannot display
 805    accurate messages if the function or data definition spans multiple
 806    files */
 807
 808 /* save current parse state in 's' */
 809 ST_FUNC void save_parse_state(ParseState *s)
 810 {
 811     s->line_num = file->line_num;
 812     s->macro_ptr = macro_ptr;
 813     s->tok = tok;
 814     s->tokc = tokc;
 815 }
 816
 817 /* restore parse state from 's' */
 818 ST_FUNC void restore_parse_state(ParseState *s)
 819 {
 820     file->line_num = s->line_num;
 821     macro_ptr = s->macro_ptr;
 822     tok = s->tok;
 823     tokc = s->tokc;
 824 }
 825
 826 /* return the number of additional 'ints' necessary to store the
 827    token */
 828 static inline int tok_ext_size(int t)
 829 {
 830     switch(t) {
 831         /* 4 bytes */
 832     case TOK_CINT:
 833     case TOK_CUINT:
 834     case TOK_CCHAR:
 835     case TOK_LCHAR:
 836     case TOK_CFLOAT:
 837     case TOK_LINENUM:
 838         return 1;
 839     case TOK_STR:
 840     case TOK_LSTR:
 841     case TOK_PPNUM:
 842         tcc_error("unsupported token");
 843         return 1;
 844     case TOK_CDOUBLE:
 845     case TOK_CLLONG:
 846     case TOK_CULLONG:
 847         return 2;
 848     case TOK_CLDOUBLE:
 849         return LDOUBLE_SIZE / 4;
 850     default:
 851         return 0;
 852     }
 853 }
 854
 855 /* token string handling */
 856
 857 ST_INLN void tok_str_new(TokenString *s)
 858 {
 859     s->str = NULL;
 860     s->len = 0;
 861     s->allocated_len = 0;
 862     s->last_line_num = -1;
 863 }
 864
 865 ST_FUNC void tok_str_free(int *str)
 866 {
 867     tcc_free(str);
 868 }
 869
 870 static int *tok_str_realloc(TokenString *s)
 871 {
 872     int *str, len;
 873
 874     if (s->allocated_len == 0) {
 875         len = 8;
 876     } else {
 877         len = s->allocated_len * 2;
 878     }
 879     str = tcc_realloc(s->str, len * sizeof(int));
 880     s->allocated_len = len;
 881     s->str = str;
 882     return str;
 883 }
 884
 885 ST_FUNC void tok_str_add(TokenString *s, int t)
 886 {
 887     int len, *str;
 888
 889     len = s->len;
 890     str = s->str;
 891     if (len >= s->allocated_len)
 892         str = tok_str_realloc(s);
 893     str[len++] = t;
 894     s->len = len;
 895 }
 896
 897 static void tok_str_add2(TokenString *s, int t, CValue *cv)
 898 {
 899     int len, *str;
 900
 901     len = s->len;
 902     str = s->str;
 903
 904     /* allocate space for worst case */
 905     if (len + TOK_MAX_SIZE > s->allocated_len)
 906         str = tok_str_realloc(s);
 907     str[len++] = t;
 908     switch(t) {
 909     case TOK_CINT:
 910     case TOK_CUINT:
 911     case TOK_CCHAR:
 912     case TOK_LCHAR:
 913     case TOK_CFLOAT:
 914     case TOK_LINENUM:
 915         str[len++] = cv->tab[0];
 916         break;
 917     case TOK_PPNUM:
 918     case TOK_STR:
 919     case TOK_LSTR:
 920         {
 921             int nb_words;
 922             CString *cstr;
 923
 924             nb_words = (sizeof(CString) + cv->cstr->size + 3) >> 2;
 925             while ((len + nb_words) > s->allocated_len)
 926                 str = tok_str_realloc(s);
 927             cstr = (CString *)(str + len);
 928             cstr->data = NULL;
 929             cstr->size = cv->cstr->size;
 930             cstr->data_allocated = NULL;
 931             cstr->size_allocated = cstr->size;
 932             memcpy((char *)cstr + sizeof(CString),
 933                    cv->cstr->data, cstr->size);
 934             len += nb_words;
 935         }
 936         break;
 937     case TOK_CDOUBLE:
 938     case TOK_CLLONG:
 939     case TOK_CULLONG:
 940 #if LDOUBLE_SIZE == 8
 941     case TOK_CLDOUBLE:
 942 #endif
 943         str[len++] = cv->tab[0];
 944         str[len++] = cv->tab[1];
 945         break;
 946 #if LDOUBLE_SIZE == 12
 947     case TOK_CLDOUBLE:
 948         str[len++] = cv->tab[0];
 949         str[len++] = cv->tab[1];
 950         str[len++] = cv->tab[2];
 951 #elif LDOUBLE_SIZE == 16
 952     case TOK_CLDOUBLE:
 953         str[len++] = cv->tab[0];
 954         str[len++] = cv->tab[1];
 955         str[len++] = cv->tab[2];
 956         str[len++] = cv->tab[3];
 957 #elif LDOUBLE_SIZE != 8
 958 #error add long double size support
 959 #endif
 960         break;
 961     default:
 962         break;
 963     }
 964     s->len = len;
 965 }
 966
 967 /* add the current parse token in token string 's' */
 968 ST_FUNC void tok_str_add_tok(TokenString *s)
 969 {
 970     CValue cval;
 971
 972     /* save line number info */
 973     if (file->line_num != s->last_line_num) {
 974         s->last_line_num = file->line_num;
 975         cval.i = s->last_line_num;
 976         tok_str_add2(s, TOK_LINENUM, &cval);
 977     }
 978     tok_str_add2(s, tok, &tokc);
 979 }
 980
 981 /* get a token from an integer array and increment pointer
 982    accordingly. we code it as a macro to avoid pointer aliasing. */
 983 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
 984 {
 985     const int *p = *pp;
 986     int n, *tab;
 987
 988     tab = cv->tab;
 989     switch(*t = *p++) {
 990     case TOK_CINT:
 991     case TOK_CUINT:
 992     case TOK_CCHAR:
 993     case TOK_LCHAR:
 994     case TOK_CFLOAT:
 995     case TOK_LINENUM:
 996         tab[0] = *p++;
 997         break;
 998     case TOK_STR:
 999     case TOK_LSTR:
1000     case TOK_PPNUM:
1001         cv->cstr = (CString *)p;
1002         cv->cstr->data = (char *)p + sizeof(CString);
1003         p += (sizeof(CString) + cv->cstr->size + 3) >> 2;
1004         break;
1005     case TOK_CDOUBLE:
1006     case TOK_CLLONG:
1007     case TOK_CULLONG:
1008         n = 2;
1009         goto copy;
1010     case TOK_CLDOUBLE:
1011 #if LDOUBLE_SIZE == 16
1012         n = 4;
1013 #elif LDOUBLE_SIZE == 12
1014         n = 3;
1015 #elif LDOUBLE_SIZE == 8
1016         n = 2;
1017 #else
1018 # error add long double size support
1019 #endif
1020     copy:
1021         do
1022             *tab++ = *p++;
1023         while (--n);
1024         break;
1025     default:
1026         break;
1027     }
1028     *pp = p;
1029 }
1030
1031 static int macro_is_equal(const int *a, const int *b)
1032 {
1033     char buf[STRING_MAX_SIZE + 1];
1034     CValue cv;
1035     int t;
1036     while (*a && *b) {
1037         TOK_GET(&t, &a, &cv);
1038         pstrcpy(buf, sizeof buf, get_tok_str(t, &cv));
1039         TOK_GET(&t, &b, &cv);
1040         if (strcmp(buf, get_tok_str(t, &cv)))
1041             return 0;
1042     }
1043     return !(*a || *b);
1044 }
1045
1046 static void define_print(Sym *s, int is_undef)
1047 {
1048     int c, t;
1049     CValue cval;
1050     const int *str;
1051     Sym *arg;
1052
1053     if (tcc_state->dflag == 0 || !s || !tcc_state->ppfp)
1054         return;
1055
1056     if (file) {
1057         c = file->line_num - file->line_ref - 1;
1058         if (c > 0) {
1059             while (c--)
1060                 fputs("\n", tcc_state->ppfp);
1061             file->line_ref = file->line_num;
1062         }
1063     }
1064
1065     if (is_undef) {
1066         fprintf(tcc_state->ppfp, "// #undef %s", get_tok_str(s->v, NULL));
1067         return;
1068     }
1069
1070     fprintf(tcc_state->ppfp, "// #define %s", get_tok_str(s->v, NULL));
1071     arg = s->next;
1072     if (arg) {
1073         char *sep = "(";
1074         while (arg) {
1075             fprintf(tcc_state->ppfp, "%s%s", sep, get_tok_str(arg->v & ~SYM_FIELD, NULL));
1076             sep = ",";
1077             arg = arg->next;
1078         }
1079         fprintf(tcc_state->ppfp, ")");
1080     }
1081
1082     str = s->d;
1083     if (str)
1084         fprintf(tcc_state->ppfp, " ");
1085
1086     while (str) {
1087         TOK_GET(&t, &str, &cval);
1088         if (!t)
1089             break;
1090         fprintf(tcc_state->ppfp, "%s", get_tok_str(t, &cval));
1091     }
1092 }
1093
1094 /* defines handling */
1095 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1096 {
1097     Sym *s;
1098
1099     s = define_find(v);
1100     if (s && !macro_is_equal(s->d, str))
1101         tcc_warning("%s redefined", get_tok_str(v, NULL));
1102
1103     s = sym_push2(&define_stack, v, macro_type, 0);
1104     s->d = str;
1105     s->next = first_arg;
1106     table_ident[v - TOK_IDENT]->sym_define = s;
1107     define_print(s, 0);
1108 }
1109
1110 /* undefined a define symbol. Its name is just set to zero */
1111 ST_FUNC void define_undef(Sym *s)
1112 {
1113     int v;
1114     v = s->v;
1115     if (v >= TOK_IDENT && v < tok_ident) {
1116         define_print(s, 1);
1117         table_ident[v - TOK_IDENT]->sym_define = NULL;
1118     }
1119 }
1120
1121 ST_INLN Sym *define_find(int v)
1122 {
1123     v -= TOK_IDENT;
1124     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1125         return NULL;
1126     return table_ident[v]->sym_define;
1127 }
1128
1129 /* free define stack until top reaches 'b' */
1130 ST_FUNC void free_defines(Sym *b)
1131 {
1132     Sym *top, *top1;
1133     int v;
1134
1135     top = define_stack;
1136     while (top != b) {
1137         top1 = top->prev;
1138         /* do not free args or predefined defines */
1139         if (top->d)
1140             tok_str_free(top->d);
1141         v = top->v;
1142         if (v >= TOK_IDENT && v < tok_ident)
1143             table_ident[v - TOK_IDENT]->sym_define = NULL;
1144         sym_free(top);
1145         top = top1;
1146     }
1147     define_stack = b;
1148 }
1149
1150 ST_FUNC void print_defines(void)
1151 {
1152     Sym *top, *s;
1153     int v;
1154
1155     top = define_stack;
1156     while (top) {
1157         v = top->v;
1158         if (v >= TOK_IDENT && v < tok_ident) {
1159             s = table_ident[v - TOK_IDENT]->sym_define;
1160             define_print(s, 0);
1161         }
1162         top = top->prev;
1163     }
1164 }
1165
1166 /* label lookup */
1167 ST_FUNC Sym *label_find(int v)
1168 {
1169     v -= TOK_IDENT;
1170     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1171         return NULL;
1172     return table_ident[v]->sym_label;
1173 }
1174
1175 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1176 {
1177     Sym *s, **ps;
1178     s = sym_push2(ptop, v, 0, 0);
1179     s->r = flags;
1180     ps = &table_ident[v - TOK_IDENT]->sym_label;
1181     if (ptop == &global_label_stack) {
1182         /* modify the top most local identifier, so that
1183            sym_identifier will point to 's' when popped */
1184         while (*ps != NULL)
1185             ps = &(*ps)->prev_tok;
1186     }
1187     s->prev_tok = *ps;
1188     *ps = s;
1189     return s;
1190 }
1191
1192 /* pop labels until element last is reached. Look if any labels are
1193    undefined. Define symbols if '&&label' was used. */
1194 ST_FUNC void label_pop(Sym **ptop, Sym *slast)
1195 {
1196     Sym *s, *s1;
1197     for(s = *ptop; s != slast; s = s1) {
1198         s1 = s->prev;
1199         if (s->r == LABEL_DECLARED) {
1200             tcc_warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1201         } else if (s->r == LABEL_FORWARD) {
1202                 tcc_error("label '%s' used but not defined",
1203                       get_tok_str(s->v, NULL));
1204         } else {
1205             if (s->c) {
1206                 /* define corresponding symbol. A size of
1207                    1 is put. */
1208                 put_extern_sym(s, cur_text_section, s->jnext, 1);
1209             }
1210         }
1211         /* remove label */
1212         table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1213         sym_free(s);
1214     }
1215     *ptop = slast;
1216 }
1217
1218 /* eval an expression for #if/#elif */
1219 static int expr_preprocess(void)
1220 {
1221     int c, t;
1222     TokenString str;
1223
1224     tok_str_new(&str);
1225     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1226         next(); /* do macro subst */
1227         if (tok == TOK_DEFINED) {
1228             next_nomacro();
1229             t = tok;
1230             if (t == '(')
1231                 next_nomacro();
1232             c = define_find(tok) != 0;
1233             if (t == '(')
1234                 next_nomacro();
1235             tok = TOK_CINT;
1236             tokc.i = c;
1237         } else if (tok >= TOK_IDENT) {
1238             /* if undefined macro */
1239             tok = TOK_CINT;
1240             tokc.i = 0;
1241         }
1242         tok_str_add_tok(&str);
1243     }
1244     tok_str_add(&str, -1); /* simulate end of file */
1245     tok_str_add(&str, 0);
1246     /* now evaluate C constant expression */
1247     macro_ptr = str.str;
1248     next();
1249     c = expr_const();
1250     macro_ptr = NULL;
1251     tok_str_free(str.str);
1252     return c != 0;
1253 }
1254
1255 /* parse after #define */
1256 ST_FUNC void parse_define(void)
1257 {
1258     Sym *s, *first, **ps;
1259     int v, t, varg, is_vaargs, spc, ptok, macro_list_start;
1260     TokenString str;
1261
1262     v = tok;
1263     if (v < TOK_IDENT)
1264         tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1265     /* XXX: should check if same macro (ANSI) */
1266     first = NULL;
1267     t = MACRO_OBJ;
1268     /* '(' must be just after macro definition for MACRO_FUNC */
1269     next_nomacro_spc();
1270     if (tok == '(') {
1271         next_nomacro();
1272         ps = &first;
1273         while (tok != ')') {
1274             varg = tok;
1275             next_nomacro();
1276             is_vaargs = 0;
1277             if (varg == TOK_DOTS) {
1278                 varg = TOK___VA_ARGS__;
1279                 is_vaargs = 1;
1280             } else if (tok == TOK_DOTS && gnu_ext) {
1281                 is_vaargs = 1;
1282                 next_nomacro();
1283             }
1284             if (varg < TOK_IDENT)
1285                 tcc_error( "\'%s\' may not appear in parameter list", get_tok_str(varg, NULL));
1286             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1287             *ps = s;
1288             ps = &s->next;
1289             if (tok != ',')
1290                 continue;
1291             next_nomacro();
1292         }
1293         next_nomacro_spc();
1294         t = MACRO_FUNC;
1295     }
1296     tok_str_new(&str);
1297     spc = 2;
1298     /* EOF testing necessary for '-D' handling */
1299     ptok = 0;
1300     macro_list_start = 1;
1301     int old_parse_flags = parse_flags;
1302     parse_flags |= PARSE_FLAG_ACCEPT_STRAYS;
1303     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1304         if (!macro_list_start && spc == 2 && tok == TOK_TWOSHARPS)
1305             tcc_error("'##' invalid at start of macro");
1306         ptok = tok;
1307         /* remove spaces around ## and after '#' */
1308         if (TOK_TWOSHARPS == tok) {
1309             if (1 == spc)
1310                 --str.len;
1311             spc = 2;
1312         } else if ('#' == tok) {
1313             spc = 3;
1314         } else if (check_space(tok, &spc)) {
1315             goto skip;
1316         }
1317         tok_str_add2(&str, tok, &tokc);
1318     skip: {
1319         next_nomacro_spc();
1320         macro_list_start = 0;
1321     }
1322     }
1323     parse_flags = old_parse_flags;
1324     if (ptok == TOK_TWOSHARPS)
1325         tcc_error("'##' invalid at end of macro");
1326     if (spc == 1)
1327         --str.len; /* remove trailing space */
1328     tok_str_add(&str, 0);
1329     define_push(v, t, str.str, first);
1330 }
1331
1332 static inline int hash_cached_include(const char *filename)
1333 {
1334     const unsigned char *s;
1335     unsigned int h;
1336
1337     h = TOK_HASH_INIT;
1338     s = (unsigned char *) filename;
1339     while (*s) {
1340         h = TOK_HASH_FUNC(h, *s);
1341         s++;
1342     }
1343     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1344     return h;
1345 }
1346
1347 static CachedInclude *search_cached_include(TCCState *s1, const char *filename)
1348 {
1349     CachedInclude *e;
1350     int i, h;
1351     h = hash_cached_include(filename);
1352     i = s1->cached_includes_hash[h];
1353     for(;;) {
1354         if (i == 0)
1355             break;
1356         e = s1->cached_includes[i - 1];
1357         if (0 == PATHCMP(e->filename, filename))
1358             return e;
1359         i = e->hash_next;
1360     }
1361     return NULL;
1362 }
1363
1364 static inline void add_cached_include(TCCState *s1, const char *filename, int ifndef_macro)
1365 {
1366     CachedInclude *e;
1367     int h;
1368
1369     if (search_cached_include(s1, filename))
1370         return;
1371 #ifdef INC_DEBUG
1372     printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1373 #endif
1374     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1375     strcpy(e->filename, filename);
1376     e->ifndef_macro = ifndef_macro;
1377     dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1378     /* add in hash table */
1379     h = hash_cached_include(filename);
1380     e->hash_next = s1->cached_includes_hash[h];
1381     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1382 }
1383
1384 static void pragma_parse(TCCState *s1)
1385 {
1386     next_nomacro();
1387     if (tok == TOK_pack) {
1388         /*
1389           This may be:
1390           #pragma pack(1) // set
1391           #pragma pack() // reset to default
1392           #pragma pack(push,1) // push & set
1393           #pragma pack(pop) // restore previous
1394         */
1395         next();
1396         skip('(');
1397         if (tok == TOK_ASM_pop) {
1398             next();
1399             if (s1->pack_stack_ptr <= s1->pack_stack) {
1400             stk_error:
1401                 tcc_error("out of pack stack");
1402             }
1403             s1->pack_stack_ptr--;
1404         } else {
1405             int val = 0;
1406             if (tok != ')') {
1407                 if (tok == TOK_ASM_push) {
1408                     next();
1409                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1410                         goto stk_error;
1411                     s1->pack_stack_ptr++;
1412                     skip(',');
1413                 }
1414                 if (tok != TOK_CINT)
1415                     goto pragma_err;
1416                 val = tokc.i;
1417                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1418                     goto pragma_err;
1419                 next();
1420             }
1421             *s1->pack_stack_ptr = val;
1422         }
1423         if (tok != ')')
1424             goto pragma_err;
1425     } else if (tok == TOK_comment) {
1426         if (s1->ms_extensions) {
1427             next();
1428             skip('(');
1429             if (tok == TOK_lib) {
1430                 next();
1431                 skip(',');
1432                 if (tok != TOK_STR) {
1433                     tcc_error("invalid library specification");
1434                 } else {
1435                     int len = strlen((char *)tokc.cstr->data);
1436                     char *file = tcc_malloc(len + 4); /* filetype, "-l", and \0 at the end */
1437                     file[0] = TCC_FILETYPE_BINARY;
1438                     file[1] = '-';
1439                     file[2] = 'l';
1440                     strcpy(&file[3],(char *)tokc.cstr->data);
1441                     dynarray_add((void ***)&s1->files, &s1->nb_files, file);
1442                 }
1443                 next();
1444                 tok = TOK_LINEFEED;
1445             } else {
1446                 tcc_warning("unknown specifier '%s' in #pragma comment", get_tok_str(tok, &tokc));
1447             }
1448         } else {
1449             tcc_warning("#pragma comment(lib) is ignored");
1450         }
1451     } else if (tok == TOK_push_macro || tok == TOK_pop_macro) {
1452         int t = tok, v;
1453         Sym *s;
1454
1455         if (next_nomacro(), tok != '(')
1456             goto pragma_err;
1457         if (next_nomacro(), tok != TOK_STR)
1458             goto pragma_err;
1459         v = tok_alloc(tokc.cstr->data, tokc.cstr->size - 1)->tok;
1460         if (next_nomacro(), tok != ')')
1461             goto pragma_err;
1462         if (t == TOK_push_macro) {
1463             while (NULL == (s = define_find(v)))
1464                 define_push(v, 0, NULL, NULL);
1465             s->type.ref = s; /* set push boundary */
1466         } else {
1467             for (s = define_stack; s; s = s->prev)
1468                 if (s->v == v && s->type.ref == s) {
1469                     s->type.ref = NULL;
1470                     break;
1471                 }
1472         }
1473         if (s)
1474             table_ident[v - TOK_IDENT]->sym_define = s->d ? s : NULL;
1475         else
1476             tcc_warning("unbalanced #pragma pop_macro");
1477
1478         /* print info when tcc is called with "-E -dD" switches */
1479         if (s1->dflag && s1->ppfp) {
1480             if (file) {
1481                 int c = file->line_num - file->line_ref - 1;
1482                 if (c > 0) {
1483                     while (c--)
1484                     fputs("\n", tcc_state->ppfp);
1485                     file->line_ref = file->line_num;
1486                 }
1487             }
1488             fprintf(s1->ppfp, "// #pragma %s_macro(\"%s\")",
1489                 (t == TOK_push_macro) ? "push" : "pop",
1490                 get_tok_str(v, NULL));
1491         }
1492     } else if (tok == TOK_once) {
1493         add_cached_include(s1, file->filename, TOK_once);
1494     } else {
1495         tcc_warning("unknown #pragma %s", get_tok_str(tok, &tokc));
1496     }
1497     return;
1498 pragma_err:
1499     tcc_error("malformed #pragma directive");
1500 }
1501
1502 /* is_bof is true if first non space token at beginning of file */
1503 ST_FUNC void preprocess(int is_bof)
1504 {
1505     TCCState *s1 = tcc_state;
1506     int i, c, n, saved_parse_flags;
1507     char buf[1024], *q;
1508     Sym *s;
1509
1510     saved_parse_flags = parse_flags;
1511     parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM |
1512         PARSE_FLAG_LINEFEED;
1513     parse_flags |= (saved_parse_flags & PARSE_FLAG_ASM_FILE);
1514     next_nomacro();
1515  redo:
1516     switch(tok) {
1517     case TOK_DEFINE:
1518         next_nomacro();
1519         parse_define();
1520         break;
1521     case TOK_UNDEF:
1522         next_nomacro();
1523         s = define_find(tok);
1524         /* undefine symbol by putting an invalid name */
1525         if (s)
1526             define_undef(s);
1527         break;
1528     case TOK_INCLUDE:
1529     case TOK_INCLUDE_NEXT:
1530         ch = file->buf_ptr[0];
1531         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1532         skip_spaces();
1533         if (ch == '<') {
1534             c = '>';
1535             goto read_name;
1536         } else if (ch == '\"') {
1537             c = ch;
1538         read_name:
1539             inp();
1540             q = buf;
1541             while (ch != c && ch != '\n' && ch != CH_EOF) {
1542                 if ((q - buf) < sizeof(buf) - 1)
1543                     *q++ = ch;
1544                 if (ch == '\\') {
1545                     if (handle_stray_noerror() == 0)
1546                         --q;
1547                 } else
1548                     inp();
1549             }
1550             *q = '\0';
1551             minp();
1552 #if 0
1553             /* eat all spaces and comments after include */
1554             /* XXX: slightly incorrect */
1555             while (ch1 != '\n' && ch1 != CH_EOF)
1556                 inp();
1557 #endif
1558         } else {
1559             /* computed #include : either we have only strings or
1560                we have anything enclosed in '<>' */
1561             next();
1562             buf[0] = '\0';
1563             if (tok == TOK_STR) {
1564                 while (tok != TOK_LINEFEED) {
1565                     if (tok != TOK_STR) {
1566                     include_syntax:
1567                         tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1568                     }
1569                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1570                     next();
1571                 }
1572                 c = '\"';
1573             } else {
1574                 int len;
1575                 while (tok != TOK_LINEFEED) {
1576                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1577                     next();
1578                 }
1579                 len = strlen(buf);
1580                 /* check syntax and remove '<>' */
1581                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1582                     goto include_syntax;
1583                 memmove(buf, buf + 1, len - 2);
1584                 buf[len - 2] = '\0';
1585                 c = '>';
1586             }
1587         }
1588
1589         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1590             tcc_error("#include recursion too deep");
1591         /* store current file in stack, but increment stack later below */
1592         *s1->include_stack_ptr = file;
1593
1594         n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1595         for (i = -2; i < n; ++i) {
1596             char buf1[sizeof file->filename];
1597             CachedInclude *e;
1598             BufferedFile **f;
1599             const char *path;
1600
1601             if (i == -2) {
1602                 /* check absolute include path */
1603                 if (!IS_ABSPATH(buf))
1604                     continue;
1605                 buf1[0] = 0;
1606                 i = n; /* force end loop */
1607
1608             } else if (i == -1) {
1609                 /* search in current dir if "header.h" */
1610                 if (c != '\"')
1611                     continue;
1612                 path = file->filename;
1613                 pstrncpy(buf1, path, tcc_basename(path) - path);
1614
1615             } else {
1616                 /* search in all the include paths */
1617                 if (i < s1->nb_include_paths)
1618                     path = s1->include_paths[i];
1619                 else
1620                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1621                 pstrcpy(buf1, sizeof(buf1), path);
1622                 pstrcat(buf1, sizeof(buf1), "/");
1623             }
1624
1625             pstrcat(buf1, sizeof(buf1), buf);
1626
1627             if (tok == TOK_INCLUDE_NEXT)
1628                 for (f = s1->include_stack_ptr; f >= s1->include_stack; --f)
1629                     if (0 == PATHCMP((*f)->filename, buf1)) {
1630 #ifdef INC_DEBUG
1631                         printf("%s: #include_next skipping %s\n", file->filename, buf1);
1632 #endif
1633                         goto include_trynext;
1634                     }
1635
1636             e = search_cached_include(s1, buf1);
1637             if (e && (define_find(e->ifndef_macro) || e->ifndef_macro == TOK_once)) {
1638                 /* no need to parse the include because the 'ifndef macro'
1639                    is defined */
1640 #ifdef INC_DEBUG
1641                 printf("%s: skipping cached %s\n", file->filename, buf1);
1642 #endif
1643                 goto include_done;
1644             }
1645
1646             if (tcc_open(s1, buf1) < 0)
1647 include_trynext:
1648                 continue;
1649
1650 #ifdef INC_DEBUG
1651             printf("%s: including %s\n", file->prev->filename, file->filename);
1652 #endif
1653             /* update target deps */
1654             dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1655                     tcc_strdup(buf1));
1656             /* push current file in stack */
1657             ++s1->include_stack_ptr;
1658             /* add include file debug info */
1659             if (s1->do_debug)
1660                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1661             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1662             ch = file->buf_ptr[0];
1663             goto the_end;
1664         }
1665         tcc_error("include file '%s' not found", buf);
1666 include_done:
1667         break;
1668     case TOK_IFNDEF:
1669         c = 1;
1670         goto do_ifdef;
1671     case TOK_IF:
1672         c = expr_preprocess();
1673         goto do_if;
1674     case TOK_IFDEF:
1675         c = 0;
1676     do_ifdef:
1677         next_nomacro();
1678         if (tok < TOK_IDENT)
1679             tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1680         if (is_bof) {
1681             if (c) {
1682 #ifdef INC_DEBUG
1683                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1684 #endif
1685                 file->ifndef_macro = tok;
1686             }
1687         }
1688         c = (define_find(tok) != 0) ^ c;
1689     do_if:
1690         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1691             tcc_error("memory full (ifdef)");
1692         *s1->ifdef_stack_ptr++ = c;
1693         goto test_skip;
1694     case TOK_ELSE:
1695         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1696             tcc_error("#else without matching #if");
1697         if (s1->ifdef_stack_ptr[-1] & 2)
1698             tcc_error("#else after #else");
1699         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1700         goto test_else;
1701     case TOK_ELIF:
1702         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1703             tcc_error("#elif without matching #if");
1704         c = s1->ifdef_stack_ptr[-1];
1705         if (c > 1)
1706             tcc_error("#elif after #else");
1707         /* last #if/#elif expression was true: we skip */
1708         if (c == 1)
1709             goto skip;
1710         c = expr_preprocess();
1711         s1->ifdef_stack_ptr[-1] = c;
1712     test_else:
1713         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1714             file->ifndef_macro = 0;
1715     test_skip:
1716         if (!(c & 1)) {
1717         skip:
1718             preprocess_skip();
1719             is_bof = 0;
1720             goto redo;
1721         }
1722         break;
1723     case TOK_ENDIF:
1724         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1725             tcc_error("#endif without matching #if");
1726         s1->ifdef_stack_ptr--;
1727         /* '#ifndef macro' was at the start of file. Now we check if
1728            an '#endif' is exactly at the end of file */
1729         if (file->ifndef_macro &&
1730             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1731             file->ifndef_macro_saved = file->ifndef_macro;
1732             /* need to set to zero to avoid false matches if another
1733                #ifndef at middle of file */
1734             file->ifndef_macro = 0;
1735             while (tok != TOK_LINEFEED)
1736                 next_nomacro();
1737             tok_flags |= TOK_FLAG_ENDIF;
1738             goto the_end;
1739         }
1740         break;
1741     case TOK_LINE:
1742         next();
1743         if (tok != TOK_CINT)
1744             tcc_error("A #line format is wrong");
1745     case TOK_PPNUM:
1746         if (tok != TOK_CINT) {
1747             char *p = tokc.cstr->data;
1748             tokc.i = strtoul(p, (char **)&p, 10);
1749         }
1750         i = file->line_num;
1751         file->line_num = tokc.i - 1;
1752         next();
1753         if (tok != TOK_LINEFEED) {
1754             if (tok != TOK_STR) {
1755                 if ((parse_flags & PARSE_FLAG_ASM_FILE) == 0) {
1756                     file->line_num = i;
1757                     tcc_error("#line format is wrong");
1758                 }
1759                 break;
1760             }
1761             pstrcpy(file->filename, sizeof(file->filename),
1762                     (char *)tokc.cstr->data);
1763         }
1764         if (s1->do_debug)
1765             put_stabs(file->filename, N_BINCL, 0, 0, 0);
1766         break;
1767     case TOK_ERROR:
1768     case TOK_WARNING:
1769         c = tok;
1770         ch = file->buf_ptr[0];
1771         skip_spaces();
1772         q = buf;
1773         while (ch != '\n' && ch != CH_EOF) {
1774             if ((q - buf) < sizeof(buf) - 1)
1775                 *q++ = ch;
1776             if (ch == '\\') {
1777                 if (handle_stray_noerror() == 0)
1778                     --q;
1779             } else
1780                 inp();
1781         }
1782         *q = '\0';
1783         if (c == TOK_ERROR)
1784             tcc_error("#error %s", buf);
1785         else
1786             tcc_warning("#warning %s", buf);
1787         break;
1788     case TOK_PRAGMA:
1789         pragma_parse(s1);
1790         break;
1791     default:
1792         if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_PPNUM) {
1793             /* '!' is ignored to allow C scripts. numbers are ignored
1794                to emulate cpp behaviour */
1795         } else {
1796             if (!(parse_flags & PARSE_FLAG_ASM_FILE))
1797                 tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1798             else {
1799                 /* this is a gas line comment in an 'S' file. */
1800                 file->buf_ptr = parse_line_comment(file->buf_ptr);
1801                 goto the_end;
1802             }
1803         }
1804         break;
1805     }
1806     /* ignore other preprocess commands or #! for C scripts */
1807     while (tok != TOK_LINEFEED)
1808         next_nomacro();
1809  the_end:
1810     parse_flags = saved_parse_flags;
1811 }
1812
1813 /* evaluate escape codes in a string. */
1814 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1815 {
1816     int c, n;
1817     const uint8_t *p;
1818
1819     p = buf;
1820     for(;;) {
1821         c = *p;
1822         if (c == '\0')
1823             break;
1824         if (c == '\\') {
1825             p++;
1826             /* escape */
1827             c = *p;
1828             switch(c) {
1829             case '0': case '1': case '2': case '3':
1830             case '4': case '5': case '6': case '7':
1831                 /* at most three octal digits */
1832                 n = c - '0';
1833                 p++;
1834                 c = *p;
1835                 if (isoct(c)) {
1836                     n = n * 8 + c - '0';
1837                     p++;
1838                     c = *p;
1839                     if (isoct(c)) {
1840                         n = n * 8 + c - '0';
1841                         p++;
1842                     }
1843                 }
1844                 c = n;
1845                 goto add_char_nonext;
1846             case 'x':
1847             case 'u':
1848             case 'U':
1849                 p++;
1850                 n = 0;
1851                 for(;;) {
1852                     c = *p;
1853                     if (c >= 'a' && c <= 'f')
1854                         c = c - 'a' + 10;
1855                     else if (c >= 'A' && c <= 'F')
1856                         c = c - 'A' + 10;
1857                     else if (isnum(c))
1858                         c = c - '0';
1859                     else
1860                         break;
1861                     n = n * 16 + c;
1862                     p++;
1863                 }
1864                 c = n;
1865                 goto add_char_nonext;
1866             case 'a':
1867                 c = '\a';
1868                 break;
1869             case 'b':
1870                 c = '\b';
1871                 break;
1872             case 'f':
1873                 c = '\f';
1874                 break;
1875             case 'n':
1876                 c = '\n';
1877                 break;
1878             case 'r':
1879                 c = '\r';
1880                 break;
1881             case 't':
1882                 c = '\t';
1883                 break;
1884             case 'v':
1885                 c = '\v';
1886                 break;
1887             case 'e':
1888                 if (!gnu_ext)
1889                     goto invalid_escape;
1890                 c = 27;
1891                 break;
1892             case '\'':
1893             case '\"':
1894             case '\\':
1895             case '?':
1896                 break;
1897             default:
1898             invalid_escape:
1899                 if (c >= '!' && c <= '~')
1900                     tcc_warning("unknown escape sequence: \'\\%c\'", c);
1901                 else
1902                     tcc_warning("unknown escape sequence: \'\\x%x\'", c);
1903                 break;
1904             }
1905         }
1906         p++;
1907     add_char_nonext:
1908         if (!is_long)
1909             cstr_ccat(outstr, c);
1910         else
1911             cstr_wccat(outstr, c);
1912     }
1913     /* add a trailing '\0' */
1914     if (!is_long)
1915         cstr_ccat(outstr, '\0');
1916     else
1917         cstr_wccat(outstr, '\0');
1918 }
1919
1920 /* we use 64 bit numbers */
1921 #define BN_SIZE 2
1922
1923 /* bn = (bn << shift) | or_val */
1924 static void bn_lshift(unsigned int *bn, int shift, int or_val)
1925 {
1926     int i;
1927     unsigned int v;
1928     for(i=0;i<BN_SIZE;i++) {
1929         v = bn[i];
1930         bn[i] = (v << shift) | or_val;
1931         or_val = v >> (32 - shift);
1932     }
1933 }
1934
1935 static void bn_zero(unsigned int *bn)
1936 {
1937     int i;
1938     for(i=0;i<BN_SIZE;i++) {
1939         bn[i] = 0;
1940     }
1941 }
1942
1943 /* parse number in null terminated string 'p' and return it in the
1944    current token */
1945 static void parse_number(const char *p)
1946 {
1947     int b, t, shift, frac_bits, s, exp_val, ch;
1948     char *q;
1949     unsigned int bn[BN_SIZE];
1950     double d;
1951
1952     /* number */
1953     q = token_buf;
1954     ch = *p++;
1955     t = ch;
1956     ch = *p++;
1957     *q++ = t;
1958     b = 10;
1959     if (t == '.') {
1960         goto float_frac_parse;
1961     } else if (t == '0') {
1962         if (ch == 'x' || ch == 'X') {
1963             q--;
1964             ch = *p++;
1965             b = 16;
1966         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
1967             q--;
1968             ch = *p++;
1969             b = 2;
1970         }
1971     }
1972     /* parse all digits. cannot check octal numbers at this stage
1973        because of floating point constants */
1974     while (1) {
1975         if (ch >= 'a' && ch <= 'f')
1976             t = ch - 'a' + 10;
1977         else if (ch >= 'A' && ch <= 'F')
1978             t = ch - 'A' + 10;
1979         else if (isnum(ch))
1980             t = ch - '0';
1981         else
1982             break;
1983         if (t >= b)
1984             break;
1985         if (q >= token_buf + STRING_MAX_SIZE) {
1986         num_too_long:
1987             tcc_error("number too long");
1988         }
1989         *q++ = ch;
1990         ch = *p++;
1991     }
1992     if (ch == '.' ||
1993         ((ch == 'e' || ch == 'E') && b == 10) ||
1994         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
1995         if (b != 10) {
1996             /* NOTE: strtox should support that for hexa numbers, but
1997                non ISOC99 libcs do not support it, so we prefer to do
1998                it by hand */
1999             /* hexadecimal or binary floats */
2000             /* XXX: handle overflows */
2001             *q = '\0';
2002             if (b == 16)
2003                 shift = 4;
2004             else
2005                 shift = 1;
2006             bn_zero(bn);
2007             q = token_buf;
2008             while (1) {
2009                 t = *q++;
2010                 if (t == '\0') {
2011                     break;
2012                 } else if (t >= 'a') {
2013                     t = t - 'a' + 10;
2014                 } else if (t >= 'A') {
2015                     t = t - 'A' + 10;
2016                 } else {
2017                     t = t - '0';
2018                 }
2019                 bn_lshift(bn, shift, t);
2020             }
2021             frac_bits = 0;
2022             if (ch == '.') {
2023                 ch = *p++;
2024                 while (1) {
2025                     t = ch;
2026                     if (t >= 'a' && t <= 'f') {
2027                         t = t - 'a' + 10;
2028                     } else if (t >= 'A' && t <= 'F') {
2029                         t = t - 'A' + 10;
2030                     } else if (t >= '0' && t <= '9') {
2031                         t = t - '0';
2032                     } else {
2033                         break;
2034                     }
2035                     if (t >= b)
2036                         tcc_error("invalid digit");
2037                     bn_lshift(bn, shift, t);
2038                     frac_bits += shift;
2039                     ch = *p++;
2040                 }
2041             }
2042             if (ch != 'p' && ch != 'P')
2043                 expect("exponent");
2044             ch = *p++;
2045             s = 1;
2046             exp_val = 0;
2047             if (ch == '+') {
2048                 ch = *p++;
2049             } else if (ch == '-') {
2050                 s = -1;
2051                 ch = *p++;
2052             }
2053             if (ch < '0' || ch > '9')
2054                 expect("exponent digits");
2055             while (ch >= '0' && ch <= '9') {
2056                 exp_val = exp_val * 10 + ch - '0';
2057                 ch = *p++;
2058             }
2059             exp_val = exp_val * s;
2060
2061             /* now we can generate the number */
2062             /* XXX: should patch directly float number */
2063             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
2064             d = ldexp(d, exp_val - frac_bits);
2065             t = toup(ch);
2066             if (t == 'F') {
2067                 ch = *p++;
2068                 tok = TOK_CFLOAT;
2069                 /* float : should handle overflow */
2070                 tokc.f = (float)d;
2071             } else if (t == 'L') {
2072                 ch = *p++;
2073 #ifdef TCC_TARGET_PE
2074                 tok = TOK_CDOUBLE;
2075                 tokc.d = d;
2076 #else
2077                 tok = TOK_CLDOUBLE;
2078                 /* XXX: not large enough */
2079                 tokc.ld = (long double)d;
2080 #endif
2081             } else {
2082                 tok = TOK_CDOUBLE;
2083                 tokc.d = d;
2084             }
2085         } else {
2086             /* decimal floats */
2087             if (ch == '.') {
2088                 if (q >= token_buf + STRING_MAX_SIZE)
2089                     goto num_too_long;
2090                 *q++ = ch;
2091                 ch = *p++;
2092             float_frac_parse:
2093                 while (ch >= '0' && ch <= '9') {
2094                     if (q >= token_buf + STRING_MAX_SIZE)
2095                         goto num_too_long;
2096                     *q++ = ch;
2097                     ch = *p++;
2098                 }
2099             }
2100             if (ch == 'e' || ch == 'E') {
2101                 if (q >= token_buf + STRING_MAX_SIZE)
2102                     goto num_too_long;
2103                 *q++ = ch;
2104                 ch = *p++;
2105                 if (ch == '-' || ch == '+') {
2106                     if (q >= token_buf + STRING_MAX_SIZE)
2107                         goto num_too_long;
2108                     *q++ = ch;
2109                     ch = *p++;
2110                 }
2111                 if (ch < '0' || ch > '9')
2112                     expect("exponent digits");
2113                 while (ch >= '0' && ch <= '9') {
2114                     if (q >= token_buf + STRING_MAX_SIZE)
2115                         goto num_too_long;
2116                     *q++ = ch;
2117                     ch = *p++;
2118                 }
2119             }
2120             *q = '\0';
2121             t = toup(ch);
2122             errno = 0;
2123             if (t == 'F') {
2124                 ch = *p++;
2125                 tok = TOK_CFLOAT;
2126                 tokc.f = strtof(token_buf, NULL);
2127             } else if (t == 'L') {
2128                 ch = *p++;
2129 #ifdef TCC_TARGET_PE
2130                 tok = TOK_CDOUBLE;
2131                 tokc.d = strtod(token_buf, NULL);
2132 #else
2133                 tok = TOK_CLDOUBLE;
2134                 tokc.ld = strtold(token_buf, NULL);
2135 #endif
2136             } else {
2137                 tok = TOK_CDOUBLE;
2138                 tokc.d = strtod(token_buf, NULL);
2139             }
2140         }
2141     } else {
2142         unsigned long long n, n1;
2143         int lcount, ucount, must_64bit;
2144         const char *p1;
2145
2146         /* integer number */
2147         *q = '\0';
2148         q = token_buf;
2149         if (b == 10 && *q == '0') {
2150             b = 8;
2151             q++;
2152         }
2153         n = 0;
2154         while(1) {
2155             t = *q++;
2156             /* no need for checks except for base 10 / 8 errors */
2157             if (t == '\0')
2158                 break;
2159             else if (t >= 'a')
2160                 t = t - 'a' + 10;
2161             else if (t >= 'A')
2162                 t = t - 'A' + 10;
2163             else
2164                 t = t - '0';
2165             if (t >= b)
2166                 tcc_error("invalid digit");
2167             n1 = n;
2168             n = n * b + t;
2169             /* detect overflow */
2170             /* XXX: this test is not reliable */
2171             if (n < n1)
2172                 tcc_error("integer constant overflow");
2173         }
2174
2175         /* Determine the characteristics (unsigned and/or 64bit) the type of
2176            the constant must have according to the constant suffix(es) */
2177         lcount = ucount = must_64bit = 0;
2178         p1 = p;
2179         for(;;) {
2180             t = toup(ch);
2181             if (t == 'L') {
2182                 if (lcount >= 2)
2183                     tcc_error("three 'l's in integer constant");
2184                 if (lcount && *(p - 1) != ch)
2185                     tcc_error("incorrect integer suffix: %s", p1);
2186                 lcount++;
2187 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
2188                 if (lcount == 2)
2189 #endif
2190                     must_64bit = 1;
2191                 ch = *p++;
2192             } else if (t == 'U') {
2193                 if (ucount >= 1)
2194                     tcc_error("two 'u's in integer constant");
2195                 ucount++;
2196                 ch = *p++;
2197             } else {
2198                 break;
2199             }
2200         }
2201
2202         /* Whether 64 bits are needed to hold the constant's value */
2203         if (n & 0xffffffff00000000LL || must_64bit) {
2204             tok = TOK_CLLONG;
2205             n1 = n >> 32;
2206         } else {
2207             tok = TOK_CINT;
2208             n1 = n;
2209         }
2210
2211         /* Whether type must be unsigned to hold the constant's value */
2212         if (ucount || ((n1 >> 31) && (b != 10))) {
2213             if (tok == TOK_CLLONG)
2214                 tok = TOK_CULLONG;
2215             else
2216                 tok = TOK_CUINT;
2217         /* If decimal and no unsigned suffix, bump to 64 bits or throw error */
2218         } else if (n1 >> 31) {
2219             if (tok == TOK_CINT)
2220                 tok = TOK_CLLONG;
2221             else
2222                 tcc_error("integer constant overflow");
2223         }
2224
2225         if (tok == TOK_CINT || tok == TOK_CUINT)
2226             tokc.ui = n;
2227         else
2228             tokc.ull = n;
2229     }
2230     if (ch)
2231         tcc_error("invalid number\n");
2232 }
2233
2234
2235 #define PARSE2(c1, tok1, c2, tok2)              \
2236     case c1:                                    \
2237         PEEKC(c, p);                            \
2238         if (c == c2) {                          \
2239             p++;                                \
2240             tok = tok2;                         \
2241         } else {                                \
2242             tok = tok1;                         \
2243         }                                       \
2244         break;
2245
2246 /* return next token without macro substitution */
2247 static inline void next_nomacro1(void)
2248 {
2249     int t, c, is_long;
2250     TokenSym *ts;
2251     uint8_t *p, *p1;
2252     unsigned int h;
2253
2254     p = file->buf_ptr;
2255  redo_no_start:
2256     c = *p;
2257  redo_no_start_2:
2258     switch(c) {
2259     case ' ':
2260     case '\t':
2261         tok = c;
2262         p++;
2263         goto keep_tok_flags;
2264     case '\f':
2265     case '\v':
2266     case '\r':
2267         p++;
2268         goto redo_no_start;
2269     case '\\':
2270         /* first look if it is in fact an end of buffer */
2271         if (p >= file->buf_end) {
2272             file->buf_ptr = p;
2273             handle_eob();
2274             p = file->buf_ptr;
2275             if (p >= file->buf_end)
2276                 goto parse_eof;
2277             else
2278                 goto redo_no_start;
2279         } else {
2280             file->buf_ptr = p;
2281             ch = *p;
2282             if (parse_flags & PARSE_FLAG_ACCEPT_STRAYS) {
2283                 if (handle_stray_noerror() != 0) {
2284                     goto parse_simple;
2285                 }
2286             } else {
2287                 handle_stray();
2288             }
2289             p = file->buf_ptr;
2290             c = ch;
2291             goto redo_no_start_2;
2292         }
2293     parse_eof:
2294         {
2295             TCCState *s1 = tcc_state;
2296             if ((parse_flags & PARSE_FLAG_LINEFEED)
2297                 && !(tok_flags & TOK_FLAG_EOF)) {
2298                 tok_flags |= TOK_FLAG_EOF;
2299                 tok = TOK_LINEFEED;
2300                 goto keep_tok_flags;
2301             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2302                 tok = TOK_EOF;
2303             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2304                 tcc_error("missing #endif");
2305             } else if (s1->include_stack_ptr == s1->include_stack) {
2306                 /* no include left : end of file. */
2307                 tok = TOK_EOF;
2308             } else {
2309                 tok_flags &= ~TOK_FLAG_EOF;
2310                 /* pop include file */
2311
2312                 /* test if previous '#endif' was after a #ifdef at
2313                    start of file */
2314                 if (tok_flags & TOK_FLAG_ENDIF) {
2315 #ifdef INC_DEBUG
2316                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2317 #endif
2318                     add_cached_include(s1, file->filename, file->ifndef_macro_saved);
2319                     tok_flags &= ~TOK_FLAG_ENDIF;
2320                 }
2321
2322                 /* add end of include file debug info */
2323                 if (tcc_state->do_debug) {
2324                     put_stabd(N_EINCL, 0, 0);
2325                 }
2326                 /* pop include stack */
2327                 tcc_close();
2328                 s1->include_stack_ptr--;
2329                 p = file->buf_ptr;
2330                 goto redo_no_start;
2331             }
2332         }
2333         break;
2334
2335     case '\n':
2336         file->line_num++;
2337         tok_flags |= TOK_FLAG_BOL;
2338         p++;
2339 maybe_newline:
2340         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2341             goto redo_no_start;
2342         tok = TOK_LINEFEED;
2343         goto keep_tok_flags;
2344
2345     case '#':
2346         /* XXX: simplify */
2347         PEEKC(c, p);
2348         if (is_space(c) && (parse_flags & PARSE_FLAG_ASM_FILE)) {
2349             p = parse_line_comment(p);
2350             goto redo_no_start;
2351         }
2352         else
2353         if ((tok_flags & TOK_FLAG_BOL) &&
2354             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2355             file->buf_ptr = p;
2356             preprocess(tok_flags & TOK_FLAG_BOF);
2357             p = file->buf_ptr;
2358             goto maybe_newline;
2359         } else {
2360             if (c == '#') {
2361                 p++;
2362                 tok = TOK_TWOSHARPS;
2363             } else {
2364                 if (parse_flags & PARSE_FLAG_ASM_FILE) {
2365                     p = parse_line_comment(p - 1);
2366                     goto redo_no_start;
2367                 } else {
2368                     tok = '#';
2369                 }
2370             }
2371         }
2372         break;
2373
2374     /* dollar is allowed to start identifiers when not parsing asm */
2375     case '$':
2376         if (!tcc_state->dollars_in_identifiers
2377         || (parse_flags & PARSE_FLAG_ASM_FILE)) goto parse_simple;
2378
2379     case 'a': case 'b': case 'c': case 'd':
2380     case 'e': case 'f': case 'g': case 'h':
2381     case 'i': case 'j': case 'k': case 'l':
2382     case 'm': case 'n': case 'o': case 'p':
2383     case 'q': case 'r': case 's': case 't':
2384     case 'u': case 'v': case 'w': case 'x':
2385     case 'y': case 'z':
2386     case 'A': case 'B': case 'C': case 'D':
2387     case 'E': case 'F': case 'G': case 'H':
2388     case 'I': case 'J': case 'K':
2389     case 'M': case 'N': case 'O': case 'P':
2390     case 'Q': case 'R': case 'S': case 'T':
2391     case 'U': case 'V': case 'W': case 'X':
2392     case 'Y': case 'Z':
2393     case '_':
2394     parse_ident_fast:
2395         p1 = p;
2396         h = TOK_HASH_INIT;
2397         h = TOK_HASH_FUNC(h, c);
2398         p++;
2399         for(;;) {
2400             c = *p;
2401             if (!isidnum_table[c-CH_EOF]
2402             && (tcc_state->dollars_in_identifiers ? (c != '$') : 1))
2403                 break;
2404             h = TOK_HASH_FUNC(h, c);
2405             p++;
2406         }
2407         if (c != '\\') {
2408             TokenSym **pts;
2409             int len;
2410
2411             /* fast case : no stray found, so we have the full token
2412                and we have already hashed it */
2413             len = p - p1;
2414             h &= (TOK_HASH_SIZE - 1);
2415             pts = &hash_ident[h];
2416             for(;;) {
2417                 ts = *pts;
2418                 if (!ts)
2419                     break;
2420                 if (ts->len == len && !memcmp(ts->str, p1, len))
2421                     goto token_found;
2422                 pts = &(ts->hash_next);
2423             }
2424             ts = tok_alloc_new(pts, (char *) p1, len);
2425         token_found: ;
2426         } else {
2427             /* slower case */
2428             cstr_reset(&tokcstr);
2429
2430             while (p1 < p) {
2431                 cstr_ccat(&tokcstr, *p1);
2432                 p1++;
2433             }
2434             p--;
2435             PEEKC(c, p);
2436         parse_ident_slow:
2437             while (isidnum_table[c-CH_EOF]
2438             || (tcc_state->dollars_in_identifiers ? (c == '$') : 0)) {
2439                 cstr_ccat(&tokcstr, c);
2440                 PEEKC(c, p);
2441             }
2442             ts = tok_alloc(tokcstr.data, tokcstr.size);
2443         }
2444         tok = ts->tok;
2445         break;
2446     case 'L':
2447         t = p[1];
2448         if (t != '\\' && t != '\'' && t != '\"') {
2449             /* fast case */
2450             goto parse_ident_fast;
2451         } else {
2452             PEEKC(c, p);
2453             if (c == '\'' || c == '\"') {
2454                 is_long = 1;
2455                 goto str_const;
2456             } else {
2457                 cstr_reset(&tokcstr);
2458                 cstr_ccat(&tokcstr, 'L');
2459                 goto parse_ident_slow;
2460             }
2461         }
2462         break;
2463     case '0': case '1': case '2': case '3':
2464     case '4': case '5': case '6': case '7':
2465     case '8': case '9':
2466
2467         cstr_reset(&tokcstr);
2468         /* after the first digit, accept digits, alpha, '.' or sign if
2469            prefixed by 'eEpP' */
2470     parse_num:
2471         for(;;) {
2472             t = c;
2473             cstr_ccat(&tokcstr, c);
2474             PEEKC(c, p);
2475             if (!(isnum(c) || isid(c) || c == '.' ||
2476                   ((c == '+' || c == '-') &&
2477                    (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2478                 break;
2479         }
2480         /* We add a trailing '\0' to ease parsing */
2481         cstr_ccat(&tokcstr, '\0');
2482         tokc.cstr = &tokcstr;
2483         tok = TOK_PPNUM;
2484         break;
2485     case '.':
2486         /* special dot handling because it can also start a number */
2487         PEEKC(c, p);
2488         if (isnum(c)) {
2489             cstr_reset(&tokcstr);
2490             cstr_ccat(&tokcstr, '.');
2491             goto parse_num;
2492         } else if (c == '.') {
2493             PEEKC(c, p);
2494             if (c != '.') {
2495                 if ((parse_flags & PARSE_FLAG_ASM_FILE) == 0)
2496                     expect("'.'");
2497                 tok = '.';
2498                 break;
2499             }
2500             PEEKC(c, p);
2501             tok = TOK_DOTS;
2502         } else {
2503             tok = '.';
2504         }
2505         break;
2506     case '\'':
2507     case '\"':
2508         is_long = 0;
2509     str_const:
2510         {
2511             CString str;
2512             int sep;
2513
2514             sep = c;
2515
2516             /* parse the string */
2517             cstr_new(&str);
2518             p = parse_pp_string(p, sep, &str);
2519             cstr_ccat(&str, '\0');
2520
2521             /* eval the escape (should be done as TOK_PPNUM) */
2522             cstr_reset(&tokcstr);
2523             parse_escape_string(&tokcstr, str.data, is_long);
2524             cstr_free(&str);
2525
2526             if (sep == '\'') {
2527                 int char_size;
2528                 /* XXX: make it portable */
2529                 if (!is_long)
2530                     char_size = 1;
2531                 else
2532                     char_size = sizeof(nwchar_t);
2533                 if (tokcstr.size <= char_size)
2534                     tcc_error("empty character constant");
2535                 if (tokcstr.size > 2 * char_size)
2536                     tcc_warning("multi-character character constant");
2537                 if (!is_long) {
2538                     tokc.i = *(int8_t *)tokcstr.data;
2539                     tok = TOK_CCHAR;
2540                 } else {
2541                     tokc.i = *(nwchar_t *)tokcstr.data;
2542                     tok = TOK_LCHAR;
2543                 }
2544             } else {
2545                 tokc.cstr = &tokcstr;
2546                 if (!is_long)
2547                     tok = TOK_STR;
2548                 else
2549                     tok = TOK_LSTR;
2550             }
2551         }
2552         break;
2553
2554     case '<':
2555         PEEKC(c, p);
2556         if (c == '=') {
2557             p++;
2558             tok = TOK_LE;
2559         } else if (c == '<') {
2560             PEEKC(c, p);
2561             if (c == '=') {
2562                 p++;
2563                 tok = TOK_A_SHL;
2564             } else {
2565                 tok = TOK_SHL;
2566             }
2567         } else {
2568             tok = TOK_LT;
2569         }
2570         break;
2571
2572     case '>':
2573         PEEKC(c, p);
2574         if (c == '=') {
2575             p++;
2576             tok = TOK_GE;
2577         } else if (c == '>') {
2578             PEEKC(c, p);
2579             if (c == '=') {
2580                 p++;
2581                 tok = TOK_A_SAR;
2582             } else {
2583                 tok = TOK_SAR;
2584             }
2585         } else {
2586             tok = TOK_GT;
2587         }
2588         break;
2589
2590     case '&':
2591         PEEKC(c, p);
2592         if (c == '&') {
2593             p++;
2594             tok = TOK_LAND;
2595         } else if (c == '=') {
2596             p++;
2597             tok = TOK_A_AND;
2598         } else {
2599             tok = '&';
2600         }
2601         break;
2602
2603     case '|':
2604         PEEKC(c, p);
2605         if (c == '|') {
2606             p++;
2607             tok = TOK_LOR;
2608         } else if (c == '=') {
2609             p++;
2610             tok = TOK_A_OR;
2611         } else {
2612             tok = '|';
2613         }
2614         break;
2615
2616     case '+':
2617         PEEKC(c, p);
2618         if (c == '+') {
2619             p++;
2620             tok = TOK_INC;
2621         } else if (c == '=') {
2622             p++;
2623             tok = TOK_A_ADD;
2624         } else {
2625             tok = '+';
2626         }
2627         break;
2628
2629     case '-':
2630         PEEKC(c, p);
2631         if (c == '-') {
2632             p++;
2633             tok = TOK_DEC;
2634         } else if (c == '=') {
2635             p++;
2636             tok = TOK_A_SUB;
2637         } else if (c == '>') {
2638             p++;
2639             tok = TOK_ARROW;
2640         } else {
2641             tok = '-';
2642         }
2643         break;
2644
2645     PARSE2('!', '!', '=', TOK_NE)
2646     PARSE2('=', '=', '=', TOK_EQ)
2647     PARSE2('*', '*', '=', TOK_A_MUL)
2648     PARSE2('%', '%', '=', TOK_A_MOD)
2649     PARSE2('^', '^', '=', TOK_A_XOR)
2650
2651         /* comments or operator */
2652     case '/':
2653         PEEKC(c, p);
2654         if (c == '*') {
2655             p = parse_comment(p);
2656             /* comments replaced by a blank */
2657             tok = ' ';
2658             goto keep_tok_flags;
2659         } else if (c == '/') {
2660             p = parse_line_comment(p);
2661             tok = ' ';
2662             goto keep_tok_flags;
2663         } else if (c == '=') {
2664             p++;
2665             tok = TOK_A_DIV;
2666         } else {
2667             tok = '/';
2668         }
2669         break;
2670
2671         /* simple tokens */
2672     case '(':
2673     case ')':
2674     case '[':
2675     case ']':
2676     case '{':
2677     case '}':
2678     case ',':
2679     case ';':
2680     case ':':
2681     case '?':
2682     case '~':
2683     case '@': /* only used in assembler */
2684     case TOK_QHASH:
2685     parse_simple:
2686         tok = c;
2687         p++;
2688         break;
2689     default:
2690         if ((parse_flags & PARSE_FLAG_ASM_FILE) == 0)
2691             tcc_error("unrecognized character \\x%02x", c);
2692         else {
2693             tok = ' ';
2694             p++;
2695         }
2696         break;
2697     }
2698     tok_flags = 0;
2699 keep_tok_flags:
2700     file->buf_ptr = p;
2701 #if defined(PARSE_DEBUG)
2702     printf("token = %s\n", get_tok_str(tok, &tokc));
2703 #endif
2704 }
2705
2706 /* return next token without macro substitution. Can read input from
2707    macro_ptr buffer */
2708 static void next_nomacro_spc(void)
2709 {
2710     if (macro_ptr) {
2711     redo:
2712         tok = *macro_ptr;
2713         if (tok) {
2714             TOK_GET(&tok, &macro_ptr, &tokc);
2715             if (tok == TOK_LINENUM) {
2716                 file->line_num = tokc.i;
2717                 goto redo;
2718             }
2719         }
2720     } else {
2721         next_nomacro1();
2722     }
2723 }
2724
2725 ST_FUNC void next_nomacro(void)
2726 {
2727     do {
2728         next_nomacro_spc();
2729     } while (is_space(tok));
2730 }
2731
2732 /* substitute arguments in replacement lists in macro_str by the values in
2733    args (field d) and return allocated string */
2734 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2735 {
2736     int last_tok, t, spc;
2737     const int *st;
2738     Sym *s;
2739     CValue cval;
2740     TokenString str;
2741     CString cstr;
2742     CString cstr2;
2743
2744     tok_str_new(&str);
2745     last_tok = 0;
2746     while(1) {
2747         TOK_GET(&t, &macro_str, &cval);
2748         if (!t)
2749             break;
2750         if (t == '#') {
2751             /* stringize */
2752             TOK_GET(&t, &macro_str, &cval);
2753             if (!t)
2754                 break;
2755             s = sym_find2(args, t);
2756             if (s) {
2757                 cstr_new(&cstr);
2758                 st = s->d;
2759                 spc = 0;
2760                 while (*st) {
2761                     TOK_GET(&t, &st, &cval);
2762                     if (t != TOK_PLCHLDR && !check_space(t, &spc))
2763                         cstr_cat(&cstr, get_tok_str(t, &cval));
2764                 }
2765                 cstr.size -= spc;
2766                 cstr_ccat(&cstr, '\0');
2767                 /* add string */
2768                 cstr_new(&cstr2);
2769                 /* emulate GCC behaviour and parse escapes in the token string */
2770                 parse_escape_string(&cstr2, cstr.data, 0);
2771                 cstr_free(&cstr);
2772                 cval.cstr = &cstr2;
2773                 tok_str_add2(&str, TOK_STR, &cval);
2774                 cstr_free(cval.cstr);
2775             } else {
2776                 tok_str_add2(&str, t, &cval);
2777             }
2778         } else if (t >= TOK_IDENT) {
2779             s = sym_find2(args, t);
2780             if (s) {
2781                 st = s->d;
2782                 /* if '##' is present before or after, no arg substitution */
2783                 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2784                     /* special case for var arg macros : ## eats the
2785                        ',' if empty VA_ARGS variable. */
2786                     /* XXX: test of the ',' is not 100%
2787                        reliable. should fix it to avoid security
2788                        problems */
2789                     if (gnu_ext && s->type.t &&
2790                         last_tok == TOK_TWOSHARPS &&
2791                         str.len >= 2 && str.str[str.len - 2] == ',') {
2792                         str.len -= 2;
2793                         tok_str_add(&str, TOK_GNUCOMMA);
2794                         str.str[str.len] = 0; // not actually needed.
2795                     }
2796
2797                     for(;;) {
2798                         int t1;
2799                         TOK_GET(&t1, &st, &cval);
2800                         if (!t1)
2801                             break;
2802                         tok_str_add2(&str, t1, &cval);
2803                     }
2804                 } else {
2805                     /* NOTE: the stream cannot be read when macro
2806                        substituing an argument */
2807                     macro_subst(&str, nested_list, st, NULL);
2808                 }
2809             } else {
2810                 tok_str_add(&str, t);
2811             }
2812         } else {
2813             tok_str_add2(&str, t, &cval);
2814         }
2815         last_tok = t;
2816     }
2817     if (str.len == 0)
2818         tok_str_add(&str, TOK_PLCHLDR);
2819     tok_str_add(&str, 0);
2820     return str.str;
2821 }
2822
2823 static char const ab_month_name[12][4] =
2824 {
2825     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2826     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2827 };
2828
2829 /* do macro substitution of current token with macro 's' and add
2830    result to (tok_str,tok_len). 'nested_list' is the list of all
2831    macros we got inside to avoid recursing. Return non zero if no
2832    substitution needs to be done */
2833 static int macro_subst_tok(TokenString *tok_str,
2834                            Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2835 {
2836     Sym *args, *sa, *sa1;
2837     int mstr_allocated, parlevel, *mstr, t, t1, spc;
2838     const int *p;
2839     TokenString str;
2840     char *cstrval;
2841     CValue cval;
2842     CString cstr;
2843     char buf[32];
2844
2845     /* if symbol is a macro, prepare substitution */
2846     /* special macros */
2847     if (tok == TOK___LINE__) {
2848         snprintf(buf, sizeof(buf), "%d", file->line_num);
2849         cstrval = buf;
2850         t1 = TOK_PPNUM;
2851         goto add_cstr1;
2852     } else if (tok == TOK___FILE__) {
2853         cstrval = file->filename;
2854         goto add_cstr;
2855     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2856         time_t ti;
2857         struct tm *tm;
2858
2859         time(&ti);
2860         tm = localtime(&ti);
2861         if (tok == TOK___DATE__) {
2862             snprintf(buf, sizeof(buf), "%s %2d %d",
2863                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2864         } else {
2865             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2866                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2867         }
2868         cstrval = buf;
2869     add_cstr:
2870         t1 = TOK_STR;
2871     add_cstr1:
2872         cstr_new(&cstr);
2873         cstr_cat(&cstr, cstrval);
2874         cstr_ccat(&cstr, '\0');
2875         cval.cstr = &cstr;
2876         tok_str_add2(tok_str, t1, &cval);
2877         cstr_free(&cstr);
2878     } else {
2879         int mtok = tok;
2880         int old_parse_flags = parse_flags;
2881         parse_flags |= PARSE_FLAG_ACCEPT_STRAYS;
2882
2883         mstr = s->d;
2884         mstr_allocated = 0;
2885         if (s->type.t == MACRO_FUNC) {
2886             /* NOTE: we do not use next_nomacro to avoid eating the
2887                next token. XXX: find better solution */
2888         redo:
2889             if (macro_ptr) {
2890                 p = macro_ptr;
2891                 while (is_space(t = *p) || TOK_LINEFEED == t)
2892                     ++p;
2893                 if (t == 0 && can_read_stream) {
2894                     /* end of macro stream: we must look at the token
2895                        after in the file */
2896                     struct macro_level *ml = *can_read_stream;
2897                     macro_ptr = NULL;
2898                     if (ml)
2899                     {
2900                         macro_ptr = ml->p;
2901                         ml->p = NULL;
2902                         *can_read_stream = ml -> prev;
2903                     }
2904                     /* also, end of scope for nested defined symbol */
2905                     (*nested_list)->v = -1;
2906                     goto redo;
2907                 }
2908             } else {
2909                 ch = tcc_peekc_slow(file);
2910                 while (is_space(ch) || ch == '\n' || ch == '/')
2911                   {
2912                     if (ch == '/')
2913                       {
2914                         int c;
2915                         uint8_t *p = file->buf_ptr;
2916                         PEEKC(c, p);
2917                         if (c == '*') {
2918                             p = parse_comment(p);
2919                             file->buf_ptr = p - 1;
2920                         } else if (c == '/') {
2921                             p = parse_line_comment(p);
2922                             file->buf_ptr = p - 1;
2923                         } else
2924                           break;
2925                       }
2926                     cinp();
2927                   }
2928                 t = ch;
2929             }
2930             if (t != '(') /* no macro subst */
2931                 return -1;
2932
2933             /* argument macro */
2934             next_nomacro();
2935             next_nomacro();
2936             args = NULL;
2937             sa = s->next;
2938             /* NOTE: empty args are allowed, except if no args */
2939             for(;;) {
2940                 /* handle '()' case */
2941                 if (!args && !sa && tok == ')')
2942                     break;
2943                 if (!sa)
2944                     tcc_error("macro '%s' used with too many args",
2945                           get_tok_str(s->v, 0));
2946                 tok_str_new(&str);
2947                 parlevel = spc = 0;
2948                 /* NOTE: non zero sa->t indicates VA_ARGS */
2949                 while ((parlevel > 0 ||
2950                         (tok != ')' &&
2951                          (tok != ',' || sa->type.t))) &&
2952                        tok != -1 &&
2953                        tok != 0) {
2954                     if (tok == '(')
2955                         parlevel++;
2956                     else if (tok == ')')
2957                         parlevel--;
2958                     if (tok == TOK_LINEFEED)
2959                         tok = ' ';
2960                     if (tok == TOK_EOF || tok == 0)
2961                         break;
2962                     if (!check_space(tok, &spc))
2963                         tok_str_add2(&str, tok, &tokc);
2964                     next_nomacro_spc();
2965                 }
2966                 if (parlevel)
2967                     return -1;
2968                 str.len -= spc;
2969                 if (str.len == 0)
2970                     tok_str_add(&str, TOK_PLCHLDR);
2971                 tok_str_add(&str, 0);
2972                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
2973                 sa1->d = str.str;
2974                 sa = sa->next;
2975                 if (tok == ')') {
2976                     /* special case for gcc var args: add an empty
2977                        var arg argument if it is omitted */
2978                     if (sa && sa->type.t && gnu_ext)
2979                         continue;
2980                     else
2981                         break;
2982                 }
2983                 /*
2984                  * #define f(a) f(a)
2985                  * #define h f(r
2986                  * h 5)
2987                  */
2988                 if (tok != ',') {
2989                     /* Argh! Not a macro invocation after all, at this
2990                      * point, so put everything back onto mstr that's
2991                      * been skipped since we saw the '(' )*/
2992                     tok_str_new(&str);
2993                     tok_str_add(&str, mtok);
2994                     tok_str_add(&str, '(');
2995                     for (sa = s->next; sa; sa = sa->next) {
2996                         int *p = sa->d;
2997                         while (p && *p) {
2998                             tok_str_add(&str, *p);
2999                             p++;
3000                         }
3001                         mstr = str.str;
3002                         /* leak memory */;
3003                         mstr_allocated = 0;
3004                         goto free_memory;
3005                     }
3006                 }
3007                 next_nomacro();
3008             }
3009             if (sa) {
3010                 tcc_error("macro '%s' used with too few args",
3011                       get_tok_str(s->v, 0));
3012             }
3013
3014             /* now subst each arg */
3015             mstr = macro_arg_subst(nested_list, mstr, args);
3016         free_memory:
3017             /* free memory */
3018             sa = args;
3019             while (sa) {
3020                 sa1 = sa->prev;
3021                 tok_str_free(sa->d);
3022                 sym_free(sa);
3023                 sa = sa1;
3024             }
3025             mstr_allocated = 1;
3026         }
3027         sym_push2(nested_list, s->v, 0, 0);
3028         parse_flags = old_parse_flags;
3029         macro_subst(tok_str, nested_list, mstr, can_read_stream);
3030         /* pop nested defined symbol */
3031         sa1 = *nested_list;
3032         *nested_list = sa1->prev;
3033         sym_free(sa1);
3034         if (mstr_allocated)
3035             tok_str_free(mstr);
3036     }
3037     return 0;
3038 }
3039
3040 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
3041    return the resulting string (which must be freed). */
3042 static inline int *macro_twosharps(const int *macro_str)
3043 {
3044     const int *ptr;
3045     int t;
3046     int last_tok = -1;
3047     TokenString macro_str1;
3048     CString cstr;
3049     int n, start_of_nosubsts;
3050
3051     /* we search the first '##' */
3052     for(ptr = macro_str;;) {
3053         CValue cval;
3054         TOK_GET(&t, &ptr, &cval);
3055         if (t == TOK_TWOSHARPS)
3056             break;
3057         /* nothing more to do if end of string */
3058         if (t == 0)
3059             return NULL;
3060     }
3061
3062     /* we saw '##', so we need more processing to handle it */
3063     start_of_nosubsts = -1;
3064     tok_str_new(&macro_str1);
3065     for(ptr = macro_str;;) {
3066         TOK_GET(&tok, &ptr, &tokc);
3067         if (tok == 0)
3068             break;
3069         if (tok == TOK_NOSUBST && start_of_nosubsts < 0)
3070             start_of_nosubsts = macro_str1.len;
3071         while (*ptr == TOK_TWOSHARPS) {
3072             /* given 'a##b', remove nosubsts preceding 'a' */
3073             if (start_of_nosubsts >= 0)
3074                 macro_str1.len = start_of_nosubsts;
3075             /* given 'a##b', skip '##' */
3076             t = *++ptr;
3077             /* given 'a##b', remove nosubsts preceding 'b' */
3078             while (t == TOK_NOSUBST)
3079                 t = *++ptr;
3080             if (t && t != TOK_TWOSHARPS) {
3081                 CValue cval;
3082                 TOK_GET(&t, &ptr, &cval);
3083                 /* We concatenate the two tokens */
3084                 cstr_new(&cstr);
3085                 if (tok != TOK_PLCHLDR)
3086                     cstr_cat(&cstr, get_tok_str(tok, &tokc));
3087                 n = cstr.size;
3088                 if (t != TOK_PLCHLDR)
3089                     cstr_cat(&cstr, get_tok_str(t, &cval));
3090                 cstr_ccat(&cstr, '\0');
3091
3092                 tcc_open_bf(tcc_state, ":paste:", cstr.size);
3093                 memcpy(file->buffer, cstr.data, cstr.size);
3094                 for (;;) {
3095                     if (0 == *file->buf_ptr)
3096                         break;
3097                     next_nomacro1();
3098                     if (0 == *file->buf_ptr)
3099                         break;
3100                     tok_str_add2(&macro_str1, tok, &tokc);
3101                     tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
3102                         n, cstr.data, (char*)cstr.data + n);
3103                 }
3104                 tcc_close();
3105                 cstr_free(&cstr);
3106             }
3107             if (tok == TOK_TWOSHARPS) {
3108                 /* two sharps twosharped together tokenize to two
3109                  * sharp tokens, not a twosharp token. */
3110                 /* That's fun to say, but is it actually true? GCC
3111                  * stringifies #define a # ## # ## # to "## #" (and a
3112                  * warning), while we produce "###" (no warning) */
3113                 tok_str_add(&macro_str1, '#');
3114                 tok = '#';
3115             }
3116         }
3117         if (tok != TOK_NOSUBST) {
3118             tok_str_add2(&macro_str1, tok, &tokc);
3119             tok = ' ';
3120             start_of_nosubsts = -1;
3121         } else
3122             tok_str_add2(&macro_str1, tok, &tokc);
3123     }
3124     tok_str_add(&macro_str1, 0);
3125     return macro_str1.str;
3126 }
3127
3128
3129 /* do macro substitution of macro_str and add result to
3130    (tok_str,tok_len). 'nested_list' is the list of all macros we got
3131    inside to avoid recursing. */
3132 static void macro_subst(TokenString *tok_str, Sym **nested_list,
3133                         const int *macro_str, struct macro_level ** can_read_stream)
3134 {
3135     Sym *s;
3136     int *macro_str1;
3137     const int *ptr;
3138     int t, ret, spc;
3139     CValue cval;
3140     struct macro_level ml;
3141     int force_blank;
3142     int gnucomma_index = -1;
3143
3144     /* first scan for '##' operator handling */
3145     ptr = macro_str;
3146     macro_str1 = macro_twosharps(ptr);
3147
3148     if (macro_str1)
3149         ptr = macro_str1;
3150     spc = 0;
3151     force_blank = 0;
3152
3153     while (1) {
3154         /* NOTE: ptr == NULL can only happen if tokens are read from
3155            file stream due to a macro function call */
3156         if (ptr == NULL)
3157             break;
3158         TOK_GET(&t, &ptr, &cval);
3159         if (t == 0)
3160             break;
3161         if (t == '\\' && !(parse_flags & PARSE_FLAG_ACCEPT_STRAYS)) {
3162             tcc_error("stray '\\' in program");
3163         }
3164         if (t == TOK_NOSUBST) {
3165             /* following token has already been subst'd. just copy it on */
3166             tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3167             TOK_GET(&t, &ptr, &cval);
3168             goto no_subst;
3169         }
3170         if (t == TOK_GNUCOMMA) {
3171             if (gnucomma_index != -1)
3172                 tcc_error("two GNU commas in the same macro");
3173             gnucomma_index = tok_str->len;
3174             tok_str_add(tok_str, ',');
3175             TOK_GET(&t, &ptr, &cval);
3176         }
3177         s = define_find(t);
3178         if (s != NULL) {
3179             int old_length = tok_str->len;
3180             /* if nested substitution, do nothing */
3181             if (sym_find2(*nested_list, t)) {
3182                 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
3183                 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3184                 goto no_subst;
3185             }
3186             ml.p = macro_ptr;
3187             if (can_read_stream)
3188                 ml.prev = *can_read_stream, *can_read_stream = &ml;
3189             macro_ptr = (int *)ptr;
3190             tok = t;
3191             ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
3192             spc = 0;
3193             if (tok_str->len && tok_str->str[tok_str->len-1] == ' ')
3194                 spc = 1;
3195             ptr = (int *)macro_ptr;
3196             macro_ptr = ml.p;
3197             if (can_read_stream && *can_read_stream == &ml)
3198                 *can_read_stream = ml.prev;
3199             if (ret != 0) {
3200                 if (!check_space(t, &spc))
3201                     tok_str_add2(tok_str, t, &cval);
3202                 tok_str_add(tok_str, ' ');
3203                 spc = 1;
3204                 force_blank = 0;
3205             }
3206             if (ret == 0 && old_length == tok_str->len)
3207                 tok_str_add(tok_str, TOK_PLCHLDR);
3208             if (parse_flags & PARSE_FLAG_SPACES)
3209                 force_blank = 0; //1;
3210         } else {
3211         no_subst:
3212             if (force_blank) {
3213                 tok_str_add(tok_str, ' ');
3214                 spc = 1;
3215                 force_blank = 0;
3216             }
3217             if (!check_space(t, &spc))
3218                 tok_str_add2(tok_str, t, &cval);
3219         }
3220         if (gnucomma_index != -1 && tok_str->len >= gnucomma_index+2) {
3221             if (tok_str->str[gnucomma_index+1] == TOK_PLCHLDR)
3222                 tok_str->len -= 2;
3223             gnucomma_index = -1;
3224         }
3225         if (tok_str->len && tok_str->str[tok_str->len-1] == TOK_PLCHLDR)
3226             tok_str->len--;
3227     }
3228     if (macro_str1)
3229         tok_str_free(macro_str1);
3230 }
3231
3232 /* return next token with macro substitution */
3233 ST_FUNC void next(void)
3234 {
3235     Sym *nested_list, *s;
3236     TokenString str;
3237     struct macro_level *ml;
3238
3239  redo:
3240     if (parse_flags & PARSE_FLAG_SPACES)
3241         next_nomacro_spc();
3242     else
3243         next_nomacro();
3244     if (!macro_ptr) {
3245         /* if not reading from macro substituted string, then try
3246            to substitute macros */
3247         if (tok >= TOK_IDENT &&
3248             (parse_flags & PARSE_FLAG_PREPROCESS)) {
3249             s = define_find(tok);
3250             if (s) {
3251                 /* we have a macro: we try to substitute */
3252                 tok_str_new(&str);
3253                 nested_list = NULL;
3254                 ml = NULL;
3255                 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
3256                     /* substitution done, NOTE: maybe empty */
3257                     tok_str_add(&str, 0);
3258                     macro_ptr = str.str;
3259                     macro_ptr_allocated = str.str;
3260                     goto redo;
3261                 }
3262                 if (str.len && str.str[str.len-1] == TOK_PLCHLDR)
3263                     str.len--;
3264             }
3265         }
3266     } else {
3267         if (tok == 0) {
3268             /* end of macro or end of unget buffer */
3269             if (unget_buffer_enabled) {
3270                 macro_ptr = unget_saved_macro_ptr;
3271                 unget_buffer_enabled = 0;
3272             } else {
3273                 /* end of macro string: free it */
3274                 tok_str_free(macro_ptr_allocated);
3275                 macro_ptr_allocated = NULL;
3276                 macro_ptr = NULL;
3277             }
3278             goto redo;
3279         } else if (tok == TOK_NOSUBST) {
3280             /* discard preprocessor's nosubst markers */
3281             goto redo;
3282         }
3283     }
3284
3285     /* convert preprocessor tokens into C tokens */
3286     if (tok == TOK_PPNUM &&
3287         (parse_flags & PARSE_FLAG_TOK_NUM)) {
3288         parse_number((char *)tokc.cstr->data);
3289     }
3290 }
3291
3292 /* push back current token and set current token to 'last_tok'. Only
3293    identifier case handled for labels. */
3294 ST_INLN void unget_tok(int last_tok)
3295 {
3296     int i, n;
3297     int *q;
3298     if (unget_buffer_enabled)
3299       {
3300         /* assert(macro_ptr == unget_saved_buffer + 1);
3301            assert(*macro_ptr == 0);  */
3302       }
3303     else
3304       {
3305         unget_saved_macro_ptr = macro_ptr;
3306         unget_buffer_enabled = 1;
3307       }
3308     q = unget_saved_buffer;
3309     macro_ptr = q;
3310     *q++ = tok;
3311     n = tok_ext_size(tok) - 1;
3312     for(i=0;i<n;i++)
3313         *q++ = tokc.tab[i];
3314     *q = 0; /* end of token string */
3315     tok = last_tok;
3316 }
3317
3318
3319 /* better than nothing, but needs extension to handle '-E' option
3320    correctly too */
3321 ST_FUNC void preprocess_init(TCCState *s1)
3322 {
3323     s1->include_stack_ptr = s1->include_stack;
3324     /* XXX: move that before to avoid having to initialize
3325        file->ifdef_stack_ptr ? */
3326     s1->ifdef_stack_ptr = s1->ifdef_stack;
3327     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3328
3329     vtop = vstack - 1;
3330     s1->pack_stack[0] = 0;
3331     s1->pack_stack_ptr = s1->pack_stack;
3332 }
3333
3334 ST_FUNC void preprocess_new(void)
3335 {
3336     int i, c;
3337     const char *p, *r;
3338
3339     /* init isid table */
3340
3341     for(i=CH_EOF;i<256;i++)
3342         isidnum_table[i-CH_EOF] = isid(i) || isnum(i);
3343
3344     /* add all tokens */
3345     if (table_ident) {
3346         tcc_free (table_ident);
3347         table_ident = NULL;
3348     }
3349     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3350
3351     tok_ident = TOK_IDENT;
3352     p = tcc_keywords;
3353     while (*p) {
3354         r = p;
3355         for(;;) {
3356             c = *r++;
3357             if (c == '\0')
3358                 break;
3359         }
3360         tok_alloc(p, r - p - 1);
3361         p = r;
3362     }
3363 }
3364
3365 static void line_macro_output(BufferedFile *f, const char *s, TCCState *s1)
3366 {
3367     switch (s1->Pflag) {
3368         case LINE_MACRO_OUTPUT_FORMAT_STD:
3369             /* "tcc -E -P1" case */
3370             fprintf(s1->ppfp, "# line %d \"%s\"\n", f->line_num, f->filename);
3371             break;
3372
3373         case LINE_MACRO_OUTPUT_FORMAT_NONE:
3374             /* "tcc -E -P" case: don't output a line directive */
3375             break;
3376
3377         case LINE_MACRO_OUTPUT_FORMAT_GCC:
3378         default:
3379             /* "tcc -E" case: a gcc standard by default */
3380             fprintf(s1->ppfp, "# %d \"%s\"%s\n", f->line_num, f->filename, s);
3381             break;
3382     }
3383 }
3384
3385 /* Preprocess the current file */
3386 ST_FUNC int tcc_preprocess(TCCState *s1)
3387 {
3388     BufferedFile *file_ref, **iptr, **iptr_new;
3389     int token_seen, d;
3390     const char *s;
3391
3392     preprocess_init(s1);
3393     ch = file->buf_ptr[0];
3394     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3395     parse_flags &= PARSE_FLAG_ASM_FILE;
3396     parse_flags |= PARSE_FLAG_PREPROCESS | PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
3397     token_seen = 0;
3398     file->line_ref = 0;
3399     file_ref = NULL;
3400     iptr = s1->include_stack_ptr;
3401
3402     for (;;) {
3403         next();
3404         if (tok == TOK_EOF) {
3405             break;
3406         } else if (file != file_ref) {
3407             if (file_ref)
3408                 line_macro_output(file_ref, "", s1);
3409             goto print_line;
3410         } else if (tok == TOK_LINEFEED) {
3411             if (!token_seen)
3412                 continue;
3413             file->line_ref++;
3414             token_seen = 0;
3415         } else if (!token_seen) {
3416             d = file->line_num - file->line_ref;
3417             if (file != file_ref || d >= 8) {
3418 print_line:
3419                 s = "";
3420                 if (tcc_state->Pflag == LINE_MACRO_OUTPUT_FORMAT_GCC) {
3421                     iptr_new = s1->include_stack_ptr;
3422                     s = iptr_new > iptr ? " 1"
3423                     : iptr_new < iptr ? " 2"
3424                     : iptr_new > s1->include_stack ? " 3"
3425                     : ""
3426                     ;
3427                 }
3428                 line_macro_output(file, s, s1);
3429             } else {
3430                 while (d > 0)
3431                     fputs("\n", s1->ppfp), --d;
3432             }
3433             file->line_ref = (file_ref = file)->line_num;
3434             token_seen = tok != TOK_LINEFEED;
3435             if (!token_seen)
3436                 continue;
3437         }
3438         fputs(get_tok_str(tok, &tokc), s1->ppfp);
3439     }
3440     return 0;
3441 }