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 /* additional informations about token */
  28 #define TOK_FLAG_BOL   0x0001 /* beginning of line before */
  29 #define TOK_FLAG_BOF   0x0002 /* beginning of file before */
  30 #define TOK_FLAG_ENDIF 0x0004 /* a endif was found matching starting #ifdef */
  31 #define TOK_FLAG_EOF   0x0008 /* end of file */
  32
  33 ST_DATA int parse_flags;
  34 #define PARSE_FLAG_PREPROCESS 0x0001 /* activate preprocessing */
  35 #define PARSE_FLAG_TOK_NUM    0x0002 /* return numbers instead of TOK_PPNUM */
  36 #define PARSE_FLAG_LINEFEED   0x0004 /* line feed is returned as a
  37                                         token. line feed is also
  38                                         returned at eof */
  39 #define PARSE_FLAG_ASM_COMMENTS 0x0008 /* '#' can be used for line comment */
  40 #define PARSE_FLAG_SPACES     0x0010 /* next() returns space tokens (for -E) */
  41
  42 ST_DATA struct BufferedFile *file;
  43 ST_DATA int ch, tok;
  44 ST_DATA CValue tokc;
  45 ST_DATA const int *macro_ptr;
  46 ST_DATA CString tokcstr; /* current parsed string, if any */
  47
  48 /* display benchmark infos */
  49 ST_DATA int total_lines;
  50 ST_DATA int total_bytes;
  51 ST_DATA int tok_ident;
  52 ST_DATA TokenSym **table_ident;
  53
  54 /* ------------------------------------------------------------------------- */
  55
  56 static int *macro_ptr_allocated;
  57 static const int *unget_saved_macro_ptr;
  58 static int unget_saved_buffer[TOK_MAX_SIZE + 1];
  59 static int unget_buffer_enabled;
  60 static TokenSym *hash_ident[TOK_HASH_SIZE];
  61 static char token_buf[STRING_MAX_SIZE + 1];
  62 /* true if isid(c) || isnum(c) */
  63 static unsigned char isidnum_table[256-CH_EOF];
  64
  65 static const char tcc_keywords[] =
  66 #define DEF(id, str) str "\0"
  67 #include "tcctok.h"
  68 #undef DEF
  69 ;
  70
  71 /* WARNING: the content of this string encodes token numbers */
  72 static const unsigned char tok_two_chars[] =
  73     "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
  74     "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
  75
  76 struct macro_level {
  77     struct macro_level *prev;
  78     const int *p;
  79 };
  80
  81 ST_FUNC void next_nomacro(void);
  82 static void next_nomacro_spc(void);
  83 static void macro_subst(
  84     TokenString *tok_str,
  85     Sym **nested_list,
  86     const int *macro_str,
  87     struct macro_level **can_read_stream
  88     );
  89
  90 ST_FUNC void skip(int c)
  91 {
  92     if (tok != c)
  93         error("'%c' expected (got \"%s\")", c, get_tok_str(tok, NULL));
  94     next();
  95 }
  96
  97 /* ------------------------------------------------------------------------- */
  98 /* CString handling */
  99 static void cstr_realloc(CString *cstr, int new_size)
 100 {
 101     int size;
 102     void *data;
 103
 104     size = cstr->size_allocated;
 105     if (size == 0)
 106         size = 8; /* no need to allocate a too small first string */
 107     while (size < new_size)
 108         size = size * 2;
 109     data = tcc_realloc(cstr->data_allocated, size);
 110     if (!data)
 111         error("memory full");
 112     cstr->data_allocated = data;
 113     cstr->size_allocated = size;
 114     cstr->data = data;
 115 }
 116
 117 /* add a byte */
 118 ST_INLN void cstr_ccat(CString *cstr, int ch)
 119 {
 120     int size;
 121     size = cstr->size + 1;
 122     if (size > cstr->size_allocated)
 123         cstr_realloc(cstr, size);
 124     ((unsigned char *)cstr->data)[size - 1] = ch;
 125     cstr->size = size;
 126 }
 127
 128 ST_FUNC void cstr_cat(CString *cstr, const char *str)
 129 {
 130     int c;
 131     for(;;) {
 132         c = *str;
 133         if (c == '\0')
 134             break;
 135         cstr_ccat(cstr, c);
 136         str++;
 137     }
 138 }
 139
 140 /* add a wide char */
 141 ST_FUNC void cstr_wccat(CString *cstr, int ch)
 142 {
 143     int size;
 144     size = cstr->size + sizeof(nwchar_t);
 145     if (size > cstr->size_allocated)
 146         cstr_realloc(cstr, size);
 147     *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
 148     cstr->size = size;
 149 }
 150
 151 ST_FUNC void cstr_new(CString *cstr)
 152 {
 153     memset(cstr, 0, sizeof(CString));
 154 }
 155
 156 /* free string and reset it to NULL */
 157 ST_FUNC void cstr_free(CString *cstr)
 158 {
 159     tcc_free(cstr->data_allocated);
 160     cstr_new(cstr);
 161 }
 162
 163 /* XXX: unicode ? */
 164 ST_FUNC void add_char(CString *cstr, int c)
 165 {
 166     if (c == '\'' || c == '\"' || c == '\\') {
 167         /* XXX: could be more precise if char or string */
 168         cstr_ccat(cstr, '\\');
 169     }
 170     if (c >= 32 && c <= 126) {
 171         cstr_ccat(cstr, c);
 172     } else {
 173         cstr_ccat(cstr, '\\');
 174         if (c == '\n') {
 175             cstr_ccat(cstr, 'n');
 176         } else {
 177             cstr_ccat(cstr, '0' + ((c >> 6) & 7));
 178             cstr_ccat(cstr, '0' + ((c >> 3) & 7));
 179             cstr_ccat(cstr, '0' + (c & 7));
 180         }
 181     }
 182 }
 183
 184 /* ------------------------------------------------------------------------- */
 185 /* allocate a new token */
 186 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
 187 {
 188     TokenSym *ts, **ptable;
 189     int i;
 190
 191     if (tok_ident >= SYM_FIRST_ANOM)
 192         error("memory full");
 193
 194     /* expand token table if needed */
 195     i = tok_ident - TOK_IDENT;
 196     if ((i % TOK_ALLOC_INCR) == 0) {
 197         ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
 198         if (!ptable)
 199             error("memory full");
 200         table_ident = ptable;
 201     }
 202
 203     ts = tcc_malloc(sizeof(TokenSym) + len);
 204     table_ident[i] = ts;
 205     ts->tok = tok_ident++;
 206     ts->sym_define = NULL;
 207     ts->sym_label = NULL;
 208     ts->sym_struct = NULL;
 209     ts->sym_identifier = NULL;
 210     ts->len = len;
 211     ts->hash_next = NULL;
 212     memcpy(ts->str, str, len);
 213     ts->str[len] = '\0';
 214     *pts = ts;
 215     return ts;
 216 }
 217
 218 #define TOK_HASH_INIT 1
 219 #define TOK_HASH_FUNC(h, c) ((h) * 263 + (c))
 220
 221 /* find a token and add it if not found */
 222 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
 223 {
 224     TokenSym *ts, **pts;
 225     int i;
 226     unsigned int h;
 227
 228     h = TOK_HASH_INIT;
 229     for(i=0;i<len;i++)
 230         h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
 231     h &= (TOK_HASH_SIZE - 1);
 232
 233     pts = &hash_ident[h];
 234     for(;;) {
 235         ts = *pts;
 236         if (!ts)
 237             break;
 238         if (ts->len == len && !memcmp(ts->str, str, len))
 239             return ts;
 240         pts = &(ts->hash_next);
 241     }
 242     return tok_alloc_new(pts, str, len);
 243 }
 244
 245 /* XXX: buffer overflow */
 246 /* XXX: float tokens */
 247 ST_FUNC char *get_tok_str(int v, CValue *cv)
 248 {
 249     static char buf[STRING_MAX_SIZE + 1];
 250     static CString cstr_buf;
 251     CString *cstr;
 252     char *p;
 253     int i, len;
 254
 255     /* NOTE: to go faster, we give a fixed buffer for small strings */
 256     cstr_reset(&cstr_buf);
 257     cstr_buf.data = buf;
 258     cstr_buf.size_allocated = sizeof(buf);
 259     p = buf;
 260
 261     switch(v) {
 262     case TOK_CINT:
 263     case TOK_CUINT:
 264         /* XXX: not quite exact, but only useful for testing */
 265         sprintf(p, "%u", cv->ui);
 266         break;
 267     case TOK_CLLONG:
 268     case TOK_CULLONG:
 269         /* XXX: not quite exact, but only useful for testing  */
 270 #ifdef _WIN32
 271         sprintf(p, "%u", (unsigned)cv->ull);
 272 #else
 273         sprintf(p, "%Lu", cv->ull);
 274 #endif
 275         break;
 276     case TOK_LCHAR:
 277         cstr_ccat(&cstr_buf, 'L');
 278     case TOK_CCHAR:
 279         cstr_ccat(&cstr_buf, '\'');
 280         add_char(&cstr_buf, cv->i);
 281         cstr_ccat(&cstr_buf, '\'');
 282         cstr_ccat(&cstr_buf, '\0');
 283         break;
 284     case TOK_PPNUM:
 285         cstr = cv->cstr;
 286         len = cstr->size - 1;
 287         for(i=0;i<len;i++)
 288             add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 289         cstr_ccat(&cstr_buf, '\0');
 290         break;
 291     case TOK_LSTR:
 292         cstr_ccat(&cstr_buf, 'L');
 293     case TOK_STR:
 294         cstr = cv->cstr;
 295         cstr_ccat(&cstr_buf, '\"');
 296         if (v == TOK_STR) {
 297             len = cstr->size - 1;
 298             for(i=0;i<len;i++)
 299                 add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 300         } else {
 301             len = (cstr->size / sizeof(nwchar_t)) - 1;
 302             for(i=0;i<len;i++)
 303                 add_char(&cstr_buf, ((nwchar_t *)cstr->data)[i]);
 304         }
 305         cstr_ccat(&cstr_buf, '\"');
 306         cstr_ccat(&cstr_buf, '\0');
 307         break;
 308     case TOK_LT:
 309         v = '<';
 310         goto addv;
 311     case TOK_GT:
 312         v = '>';
 313         goto addv;
 314     case TOK_DOTS:
 315         return strcpy(p, "...");
 316     case TOK_A_SHL:
 317         return strcpy(p, "<<=");
 318     case TOK_A_SAR:
 319         return strcpy(p, ">>=");
 320     default:
 321         if (v < TOK_IDENT) {
 322             /* search in two bytes table */
 323             const unsigned char *q = tok_two_chars;
 324             while (*q) {
 325                 if (q[2] == v) {
 326                     *p++ = q[0];
 327                     *p++ = q[1];
 328                     *p = '\0';
 329                     return buf;
 330                 }
 331                 q += 3;
 332             }
 333         addv:
 334             *p++ = v;
 335             *p = '\0';
 336         } else if (v < tok_ident) {
 337             return table_ident[v - TOK_IDENT]->str;
 338         } else if (v >= SYM_FIRST_ANOM) {
 339             /* special name for anonymous symbol */
 340             sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
 341         } else {
 342             /* should never happen */
 343             return NULL;
 344         }
 345         break;
 346     }
 347     return cstr_buf.data;
 348 }
 349
 350 /* fill input buffer and peek next char */
 351 static int tcc_peekc_slow(BufferedFile *bf)
 352 {
 353     int len;
 354     /* only tries to read if really end of buffer */
 355     if (bf->buf_ptr >= bf->buf_end) {
 356         if (bf->fd != -1) {
 357 #if defined(PARSE_DEBUG)
 358             len = 8;
 359 #else
 360             len = IO_BUF_SIZE;
 361 #endif
 362             len = read(bf->fd, bf->buffer, len);
 363             if (len < 0)
 364                 len = 0;
 365         } else {
 366             len = 0;
 367         }
 368         total_bytes += len;
 369         bf->buf_ptr = bf->buffer;
 370         bf->buf_end = bf->buffer + len;
 371         *bf->buf_end = CH_EOB;
 372     }
 373     if (bf->buf_ptr < bf->buf_end) {
 374         return bf->buf_ptr[0];
 375     } else {
 376         bf->buf_ptr = bf->buf_end;
 377         return CH_EOF;
 378     }
 379 }
 380
 381 /* return the current character, handling end of block if necessary
 382    (but not stray) */
 383 ST_FUNC int handle_eob(void)
 384 {
 385     return tcc_peekc_slow(file);
 386 }
 387
 388 /* read next char from current input file and handle end of input buffer */
 389 ST_INLN void inp(void)
 390 {
 391     ch = *(++(file->buf_ptr));
 392     /* end of buffer/file handling */
 393     if (ch == CH_EOB)
 394         ch = handle_eob();
 395 }
 396
 397 /* handle '\[\r]\n' */
 398 static int handle_stray_noerror(void)
 399 {
 400     while (ch == '\\') {
 401         inp();
 402         if (ch == '\n') {
 403             file->line_num++;
 404             inp();
 405         } else if (ch == '\r') {
 406             inp();
 407             if (ch != '\n')
 408                 goto fail;
 409             file->line_num++;
 410             inp();
 411         } else {
 412         fail:
 413             return 1;
 414         }
 415     }
 416     return 0;
 417 }
 418
 419 static void handle_stray(void)
 420 {
 421     if (handle_stray_noerror())
 422         error("stray '\\' in program");
 423 }
 424
 425 /* skip the stray and handle the \\n case. Output an error if
 426    incorrect char after the stray */
 427 static int handle_stray1(uint8_t *p)
 428 {
 429     int c;
 430
 431     if (p >= file->buf_end) {
 432         file->buf_ptr = p;
 433         c = handle_eob();
 434         p = file->buf_ptr;
 435         if (c == '\\')
 436             goto parse_stray;
 437     } else {
 438     parse_stray:
 439         file->buf_ptr = p;
 440         ch = *p;
 441         handle_stray();
 442         p = file->buf_ptr;
 443         c = *p;
 444     }
 445     return c;
 446 }
 447
 448 /* handle just the EOB case, but not stray */
 449 #define PEEKC_EOB(c, p)\
 450 {\
 451     p++;\
 452     c = *p;\
 453     if (c == '\\') {\
 454         file->buf_ptr = p;\
 455         c = handle_eob();\
 456         p = file->buf_ptr;\
 457     }\
 458 }
 459
 460 /* handle the complicated stray case */
 461 #define PEEKC(c, p)\
 462 {\
 463     p++;\
 464     c = *p;\
 465     if (c == '\\') {\
 466         c = handle_stray1(p);\
 467         p = file->buf_ptr;\
 468     }\
 469 }
 470
 471 /* input with '\[\r]\n' handling. Note that this function cannot
 472    handle other characters after '\', so you cannot call it inside
 473    strings or comments */
 474 ST_FUNC void minp(void)
 475 {
 476     inp();
 477     if (ch == '\\')
 478         handle_stray();
 479 }
 480
 481
 482 /* single line C++ comments */
 483 static uint8_t *parse_line_comment(uint8_t *p)
 484 {
 485     int c;
 486
 487     p++;
 488     for(;;) {
 489         c = *p;
 490     redo:
 491         if (c == '\n' || c == CH_EOF) {
 492             break;
 493         } else if (c == '\\') {
 494             file->buf_ptr = p;
 495             c = handle_eob();
 496             p = file->buf_ptr;
 497             if (c == '\\') {
 498                 PEEKC_EOB(c, p);
 499                 if (c == '\n') {
 500                     file->line_num++;
 501                     PEEKC_EOB(c, p);
 502                 } else if (c == '\r') {
 503                     PEEKC_EOB(c, p);
 504                     if (c == '\n') {
 505                         file->line_num++;
 506                         PEEKC_EOB(c, p);
 507                     }
 508                 }
 509             } else {
 510                 goto redo;
 511             }
 512         } else {
 513             p++;
 514         }
 515     }
 516     return p;
 517 }
 518
 519 /* C comments */
 520 ST_FUNC uint8_t *parse_comment(uint8_t *p)
 521 {
 522     int c;
 523
 524     p++;
 525     for(;;) {
 526         /* fast skip loop */
 527         for(;;) {
 528             c = *p;
 529             if (c == '\n' || c == '*' || c == '\\')
 530                 break;
 531             p++;
 532             c = *p;
 533             if (c == '\n' || c == '*' || c == '\\')
 534                 break;
 535             p++;
 536         }
 537         /* now we can handle all the cases */
 538         if (c == '\n') {
 539             file->line_num++;
 540             p++;
 541         } else if (c == '*') {
 542             p++;
 543             for(;;) {
 544                 c = *p;
 545                 if (c == '*') {
 546                     p++;
 547                 } else if (c == '/') {
 548                     goto end_of_comment;
 549                 } else if (c == '\\') {
 550                     file->buf_ptr = p;
 551                     c = handle_eob();
 552                     p = file->buf_ptr;
 553                     if (c == '\\') {
 554                         /* skip '\[\r]\n', otherwise just skip the stray */
 555                         while (c == '\\') {
 556                             PEEKC_EOB(c, p);
 557                             if (c == '\n') {
 558                                 file->line_num++;
 559                                 PEEKC_EOB(c, p);
 560                             } else if (c == '\r') {
 561                                 PEEKC_EOB(c, p);
 562                                 if (c == '\n') {
 563                                     file->line_num++;
 564                                     PEEKC_EOB(c, p);
 565                                 }
 566                             } else {
 567                                 goto after_star;
 568                             }
 569                         }
 570                     }
 571                 } else {
 572                     break;
 573                 }
 574             }
 575         after_star: ;
 576         } else {
 577             /* stray, eob or eof */
 578             file->buf_ptr = p;
 579             c = handle_eob();
 580             p = file->buf_ptr;
 581             if (c == CH_EOF) {
 582                 error("unexpected end of file in comment");
 583             } else if (c == '\\') {
 584                 p++;
 585             }
 586         }
 587     }
 588  end_of_comment:
 589     p++;
 590     return p;
 591 }
 592
 593 #define cinp minp
 594
 595 static inline void skip_spaces(void)
 596 {
 597     while (is_space(ch))
 598         cinp();
 599 }
 600
 601 static inline int check_space(int t, int *spc)
 602 {
 603     if (is_space(t)) {
 604         if (*spc)
 605             return 1;
 606         *spc = 1;
 607     } else
 608         *spc = 0;
 609     return 0;
 610 }
 611
 612 /* parse a string without interpreting escapes */
 613 static uint8_t *parse_pp_string(uint8_t *p,
 614                                 int sep, CString *str)
 615 {
 616     int c;
 617     p++;
 618     for(;;) {
 619         c = *p;
 620         if (c == sep) {
 621             break;
 622         } else if (c == '\\') {
 623             file->buf_ptr = p;
 624             c = handle_eob();
 625             p = file->buf_ptr;
 626             if (c == CH_EOF) {
 627             unterminated_string:
 628                 /* XXX: indicate line number of start of string */
 629                 error("missing terminating %c character", sep);
 630             } else if (c == '\\') {
 631                 /* escape : just skip \[\r]\n */
 632                 PEEKC_EOB(c, p);
 633                 if (c == '\n') {
 634                     file->line_num++;
 635                     p++;
 636                 } else if (c == '\r') {
 637                     PEEKC_EOB(c, p);
 638                     if (c != '\n')
 639                         expect("'\n' after '\r'");
 640                     file->line_num++;
 641                     p++;
 642                 } else if (c == CH_EOF) {
 643                     goto unterminated_string;
 644                 } else {
 645                     if (str) {
 646                         cstr_ccat(str, '\\');
 647                         cstr_ccat(str, c);
 648                     }
 649                     p++;
 650                 }
 651             }
 652         } else if (c == '\n') {
 653             file->line_num++;
 654             goto add_char;
 655         } else if (c == '\r') {
 656             PEEKC_EOB(c, p);
 657             if (c != '\n') {
 658                 if (str)
 659                     cstr_ccat(str, '\r');
 660             } else {
 661                 file->line_num++;
 662                 goto add_char;
 663             }
 664         } else {
 665         add_char:
 666             if (str)
 667                 cstr_ccat(str, c);
 668             p++;
 669         }
 670     }
 671     p++;
 672     return p;
 673 }
 674
 675 /* skip block of text until #else, #elif or #endif. skip also pairs of
 676    #if/#endif */
 677 static void preprocess_skip(void)
 678 {
 679     int a, start_of_line, c, in_warn_or_error;
 680     uint8_t *p;
 681
 682     p = file->buf_ptr;
 683     a = 0;
 684 redo_start:
 685     start_of_line = 1;
 686     in_warn_or_error = 0;
 687     for(;;) {
 688     redo_no_start:
 689         c = *p;
 690         switch(c) {
 691         case ' ':
 692         case '\t':
 693         case '\f':
 694         case '\v':
 695         case '\r':
 696             p++;
 697             goto redo_no_start;
 698         case '\n':
 699             file->line_num++;
 700             p++;
 701             goto redo_start;
 702         case '\\':
 703             file->buf_ptr = p;
 704             c = handle_eob();
 705             if (c == CH_EOF) {
 706                 expect("#endif");
 707             } else if (c == '\\') {
 708                 ch = file->buf_ptr[0];
 709                 handle_stray_noerror();
 710             }
 711             p = file->buf_ptr;
 712             goto redo_no_start;
 713         /* skip strings */
 714         case '\"':
 715         case '\'':
 716             if (in_warn_or_error)
 717                 goto _default;
 718             p = parse_pp_string(p, c, NULL);
 719             break;
 720         /* skip comments */
 721         case '/':
 722             if (in_warn_or_error)
 723                 goto _default;
 724             file->buf_ptr = p;
 725             ch = *p;
 726             minp();
 727             p = file->buf_ptr;
 728             if (ch == '*') {
 729                 p = parse_comment(p);
 730             } else if (ch == '/') {
 731                 p = parse_line_comment(p);
 732             }
 733             break;
 734         case '#':
 735             p++;
 736             if (start_of_line) {
 737                 file->buf_ptr = p;
 738                 next_nomacro();
 739                 p = file->buf_ptr;
 740                 if (a == 0 &&
 741                     (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
 742                     goto the_end;
 743                 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
 744                     a++;
 745                 else if (tok == TOK_ENDIF)
 746                     a--;
 747                 else if( tok == TOK_ERROR || tok == TOK_WARNING)
 748                     in_warn_or_error = 1;
 749             }
 750             break;
 751 _default:
 752         default:
 753             p++;
 754             break;
 755         }
 756         start_of_line = 0;
 757     }
 758  the_end: ;
 759     file->buf_ptr = p;
 760 }
 761
 762 /* ParseState handling */
 763
 764 /* XXX: currently, no include file info is stored. Thus, we cannot display
 765    accurate messages if the function or data definition spans multiple
 766    files */
 767
 768 /* save current parse state in 's' */
 769 ST_FUNC void save_parse_state(ParseState *s)
 770 {
 771     s->line_num = file->line_num;
 772     s->macro_ptr = macro_ptr;
 773     s->tok = tok;
 774     s->tokc = tokc;
 775 }
 776
 777 /* restore parse state from 's' */
 778 ST_FUNC void restore_parse_state(ParseState *s)
 779 {
 780     file->line_num = s->line_num;
 781     macro_ptr = s->macro_ptr;
 782     tok = s->tok;
 783     tokc = s->tokc;
 784 }
 785
 786 /* return the number of additional 'ints' necessary to store the
 787    token */
 788 static inline int tok_ext_size(int t)
 789 {
 790     switch(t) {
 791         /* 4 bytes */
 792     case TOK_CINT:
 793     case TOK_CUINT:
 794     case TOK_CCHAR:
 795     case TOK_LCHAR:
 796     case TOK_CFLOAT:
 797     case TOK_LINENUM:
 798         return 1;
 799     case TOK_STR:
 800     case TOK_LSTR:
 801     case TOK_PPNUM:
 802         error("unsupported token");
 803         return 1;
 804     case TOK_CDOUBLE:
 805     case TOK_CLLONG:
 806     case TOK_CULLONG:
 807         return 2;
 808     case TOK_CLDOUBLE:
 809         return LDOUBLE_SIZE / 4;
 810     default:
 811         return 0;
 812     }
 813 }
 814
 815 /* token string handling */
 816
 817 ST_INLN void tok_str_new(TokenString *s)
 818 {
 819     s->str = NULL;
 820     s->len = 0;
 821     s->allocated_len = 0;
 822     s->last_line_num = -1;
 823 }
 824
 825 ST_FUNC void tok_str_free(int *str)
 826 {
 827     tcc_free(str);
 828 }
 829
 830 static int *tok_str_realloc(TokenString *s)
 831 {
 832     int *str, len;
 833
 834     if (s->allocated_len == 0) {
 835         len = 8;
 836     } else {
 837         len = s->allocated_len * 2;
 838     }
 839     str = tcc_realloc(s->str, len * sizeof(int));
 840     if (!str)
 841         error("memory full");
 842     s->allocated_len = len;
 843     s->str = str;
 844     return str;
 845 }
 846
 847 ST_FUNC void tok_str_add(TokenString *s, int t)
 848 {
 849     int len, *str;
 850
 851     len = s->len;
 852     str = s->str;
 853     if (len >= s->allocated_len)
 854         str = tok_str_realloc(s);
 855     str[len++] = t;
 856     s->len = len;
 857 }
 858
 859 static void tok_str_add2(TokenString *s, int t, CValue *cv)
 860 {
 861     int len, *str;
 862
 863     len = s->len;
 864     str = s->str;
 865
 866     /* allocate space for worst case */
 867     if (len + TOK_MAX_SIZE > s->allocated_len)
 868         str = tok_str_realloc(s);
 869     str[len++] = t;
 870     switch(t) {
 871     case TOK_CINT:
 872     case TOK_CUINT:
 873     case TOK_CCHAR:
 874     case TOK_LCHAR:
 875     case TOK_CFLOAT:
 876     case TOK_LINENUM:
 877         str[len++] = cv->tab[0];
 878         break;
 879     case TOK_PPNUM:
 880     case TOK_STR:
 881     case TOK_LSTR:
 882         {
 883             int nb_words;
 884             CString *cstr;
 885
 886             nb_words = (sizeof(CString) + cv->cstr->size + 3) >> 2;
 887             while ((len + nb_words) > s->allocated_len)
 888                 str = tok_str_realloc(s);
 889             cstr = (CString *)(str + len);
 890             cstr->data = NULL;
 891             cstr->size = cv->cstr->size;
 892             cstr->data_allocated = NULL;
 893             cstr->size_allocated = cstr->size;
 894             memcpy((char *)cstr + sizeof(CString),
 895                    cv->cstr->data, cstr->size);
 896             len += nb_words;
 897         }
 898         break;
 899     case TOK_CDOUBLE:
 900     case TOK_CLLONG:
 901     case TOK_CULLONG:
 902 #if LDOUBLE_SIZE == 8
 903     case TOK_CLDOUBLE:
 904 #endif
 905         str[len++] = cv->tab[0];
 906         str[len++] = cv->tab[1];
 907         break;
 908 #if LDOUBLE_SIZE == 12
 909     case TOK_CLDOUBLE:
 910         str[len++] = cv->tab[0];
 911         str[len++] = cv->tab[1];
 912         str[len++] = cv->tab[2];
 913 #elif LDOUBLE_SIZE == 16
 914     case TOK_CLDOUBLE:
 915         str[len++] = cv->tab[0];
 916         str[len++] = cv->tab[1];
 917         str[len++] = cv->tab[2];
 918         str[len++] = cv->tab[3];
 919 #elif LDOUBLE_SIZE != 8
 920 #error add long double size support
 921 #endif
 922         break;
 923     default:
 924         break;
 925     }
 926     s->len = len;
 927 }
 928
 929 /* add the current parse token in token string 's' */
 930 ST_FUNC void tok_str_add_tok(TokenString *s)
 931 {
 932     CValue cval;
 933
 934     /* save line number info */
 935     if (file->line_num != s->last_line_num) {
 936         s->last_line_num = file->line_num;
 937         cval.i = s->last_line_num;
 938         tok_str_add2(s, TOK_LINENUM, &cval);
 939     }
 940     tok_str_add2(s, tok, &tokc);
 941 }
 942
 943 /* get a token from an integer array and increment pointer
 944    accordingly. we code it as a macro to avoid pointer aliasing. */
 945 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
 946 {
 947     const int *p = *pp;
 948     int n, *tab;
 949
 950     tab = cv->tab;
 951     switch(*t = *p++) {
 952     case TOK_CINT:
 953     case TOK_CUINT:
 954     case TOK_CCHAR:
 955     case TOK_LCHAR:
 956     case TOK_CFLOAT:
 957     case TOK_LINENUM:
 958         tab[0] = *p++;
 959         break;
 960     case TOK_STR:
 961     case TOK_LSTR:
 962     case TOK_PPNUM:
 963         cv->cstr = (CString *)p;
 964         cv->cstr->data = (char *)p + sizeof(CString);
 965         p += (sizeof(CString) + cv->cstr->size + 3) >> 2;
 966         break;
 967     case TOK_CDOUBLE:
 968     case TOK_CLLONG:
 969     case TOK_CULLONG:
 970         n = 2;
 971         goto copy;
 972     case TOK_CLDOUBLE:
 973 #if LDOUBLE_SIZE == 16
 974         n = 4;
 975 #elif LDOUBLE_SIZE == 12
 976         n = 3;
 977 #elif LDOUBLE_SIZE == 8
 978         n = 2;
 979 #else
 980 # error add long double size support
 981 #endif
 982     copy:
 983         do
 984             *tab++ = *p++;
 985         while (--n);
 986         break;
 987     default:
 988         break;
 989     }
 990     *pp = p;
 991 }
 992
 993 static int macro_is_equal(const int *a, const int *b)
 994 {
 995     char buf[STRING_MAX_SIZE + 1];
 996     CValue cv;
 997     int t;
 998     while (*a && *b) {
 999         TOK_GET(&t, &a, &cv);
1000         pstrcpy(buf, sizeof buf, get_tok_str(t, &cv));
1001         TOK_GET(&t, &b, &cv);
1002         if (strcmp(buf, get_tok_str(t, &cv)))
1003             return 0;
1004     }
1005     return !(*a || *b);
1006 }
1007
1008 /* defines handling */
1009 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1010 {
1011     Sym *s;
1012
1013     s = define_find(v);
1014     if (s && !macro_is_equal(s->d, str))
1015         warning("%s redefined", get_tok_str(v, NULL));
1016
1017     s = sym_push2(&define_stack, v, macro_type, 0);
1018     s->d = str;
1019     s->next = first_arg;
1020     table_ident[v - TOK_IDENT]->sym_define = s;
1021 }
1022
1023 /* undefined a define symbol. Its name is just set to zero */
1024 ST_FUNC void define_undef(Sym *s)
1025 {
1026     int v;
1027     v = s->v;
1028     if (v >= TOK_IDENT && v < tok_ident)
1029         table_ident[v - TOK_IDENT]->sym_define = NULL;
1030     s->v = 0;
1031 }
1032
1033 ST_INLN Sym *define_find(int v)
1034 {
1035     v -= TOK_IDENT;
1036     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1037         return NULL;
1038     return table_ident[v]->sym_define;
1039 }
1040
1041 /* free define stack until top reaches 'b' */
1042 ST_FUNC void free_defines(Sym *b)
1043 {
1044     Sym *top, *top1;
1045     int v;
1046
1047     top = define_stack;
1048     while (top != b) {
1049         top1 = top->prev;
1050         /* do not free args or predefined defines */
1051         if (top->d)
1052             tok_str_free(top->d);
1053         v = top->v;
1054         if (v >= TOK_IDENT && v < tok_ident)
1055             table_ident[v - TOK_IDENT]->sym_define = NULL;
1056         sym_free(top);
1057         top = top1;
1058     }
1059     define_stack = b;
1060 }
1061
1062 /* label lookup */
1063 ST_FUNC Sym *label_find(int v)
1064 {
1065     v -= TOK_IDENT;
1066     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1067         return NULL;
1068     return table_ident[v]->sym_label;
1069 }
1070
1071 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1072 {
1073     Sym *s, **ps;
1074     s = sym_push2(ptop, v, 0, 0);
1075     s->r = flags;
1076     ps = &table_ident[v - TOK_IDENT]->sym_label;
1077     if (ptop == &global_label_stack) {
1078         /* modify the top most local identifier, so that
1079            sym_identifier will point to 's' when popped */
1080         while (*ps != NULL)
1081             ps = &(*ps)->prev_tok;
1082     }
1083     s->prev_tok = *ps;
1084     *ps = s;
1085     return s;
1086 }
1087
1088 /* pop labels until element last is reached. Look if any labels are
1089    undefined. Define symbols if '&&label' was used. */
1090 ST_FUNC void label_pop(Sym **ptop, Sym *slast)
1091 {
1092     Sym *s, *s1;
1093     for(s = *ptop; s != slast; s = s1) {
1094         s1 = s->prev;
1095         if (s->r == LABEL_DECLARED) {
1096             warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1097         } else if (s->r == LABEL_FORWARD) {
1098                 error("label '%s' used but not defined",
1099                       get_tok_str(s->v, NULL));
1100         } else {
1101             if (s->c) {
1102                 /* define corresponding symbol. A size of
1103                    1 is put. */
1104                 put_extern_sym(s, cur_text_section, s->jnext, 1);
1105             }
1106         }
1107         /* remove label */
1108         table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1109         sym_free(s);
1110     }
1111     *ptop = slast;
1112 }
1113
1114 /* eval an expression for #if/#elif */
1115 static int expr_preprocess(void)
1116 {
1117     int c, t;
1118     TokenString str;
1119
1120     tok_str_new(&str);
1121     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1122         next(); /* do macro subst */
1123         if (tok == TOK_DEFINED) {
1124             next_nomacro();
1125             t = tok;
1126             if (t == '(')
1127                 next_nomacro();
1128             c = define_find(tok) != 0;
1129             if (t == '(')
1130                 next_nomacro();
1131             tok = TOK_CINT;
1132             tokc.i = c;
1133         } else if (tok >= TOK_IDENT) {
1134             /* if undefined macro */
1135             tok = TOK_CINT;
1136             tokc.i = 0;
1137         }
1138         tok_str_add_tok(&str);
1139     }
1140     tok_str_add(&str, -1); /* simulate end of file */
1141     tok_str_add(&str, 0);
1142     /* now evaluate C constant expression */
1143     macro_ptr = str.str;
1144     next();
1145     c = expr_const();
1146     macro_ptr = NULL;
1147     tok_str_free(str.str);
1148     return c != 0;
1149 }
1150
1151 #if defined(PARSE_DEBUG) || defined(PP_DEBUG)
1152 static void tok_print(int *str)
1153 {
1154     int t;
1155     CValue cval;
1156
1157     printf("<");
1158     while (1) {
1159         TOK_GET(&t, &str, &cval);
1160         if (!t)
1161             break;
1162         printf("%s", get_tok_str(t, &cval));
1163     }
1164     printf(">\n");
1165 }
1166 #endif
1167
1168 /* parse after #define */
1169 ST_FUNC void parse_define(void)
1170 {
1171     Sym *s, *first, **ps;
1172     int v, t, varg, is_vaargs, spc;
1173     TokenString str;
1174
1175     v = tok;
1176     if (v < TOK_IDENT)
1177         error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1178     /* XXX: should check if same macro (ANSI) */
1179     first = NULL;
1180     t = MACRO_OBJ;
1181     /* '(' must be just after macro definition for MACRO_FUNC */
1182     next_nomacro_spc();
1183     if (tok == '(') {
1184         next_nomacro();
1185         ps = &first;
1186         while (tok != ')') {
1187             varg = tok;
1188             next_nomacro();
1189             is_vaargs = 0;
1190             if (varg == TOK_DOTS) {
1191                 varg = TOK___VA_ARGS__;
1192                 is_vaargs = 1;
1193             } else if (tok == TOK_DOTS && gnu_ext) {
1194                 is_vaargs = 1;
1195                 next_nomacro();
1196             }
1197             if (varg < TOK_IDENT)
1198                 error("badly punctuated parameter list");
1199             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1200             *ps = s;
1201             ps = &s->next;
1202             if (tok != ',')
1203                 break;
1204             next_nomacro();
1205         }
1206         if (tok == ')')
1207             next_nomacro_spc();
1208         t = MACRO_FUNC;
1209     }
1210     tok_str_new(&str);
1211     spc = 2;
1212     /* EOF testing necessary for '-D' handling */
1213     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1214         /* remove spaces around ## and after '#' */
1215         if (TOK_TWOSHARPS == tok) {
1216             if (1 == spc)
1217                 --str.len;
1218             spc = 2;
1219         } else if ('#' == tok) {
1220             spc = 2;
1221         } else if (check_space(tok, &spc)) {
1222             goto skip;
1223         }
1224         tok_str_add2(&str, tok, &tokc);
1225     skip:
1226         next_nomacro_spc();
1227     }
1228     if (spc == 1)
1229         --str.len; /* remove trailing space */
1230     tok_str_add(&str, 0);
1231 #ifdef PP_DEBUG
1232     printf("define %s %d: ", get_tok_str(v, NULL), t);
1233     tok_print(str.str);
1234 #endif
1235     define_push(v, t, str.str, first);
1236 }
1237
1238 static inline int hash_cached_include(int type, const char *filename)
1239 {
1240     const unsigned char *s;
1241     unsigned int h;
1242
1243     h = TOK_HASH_INIT;
1244     h = TOK_HASH_FUNC(h, type);
1245     s = filename;
1246     while (*s) {
1247         h = TOK_HASH_FUNC(h, *s);
1248         s++;
1249     }
1250     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1251     return h;
1252 }
1253
1254 /* XXX: use a token or a hash table to accelerate matching ? */
1255 static CachedInclude *search_cached_include(TCCState *s1,
1256                                             int type, const char *filename)
1257 {
1258     CachedInclude *e;
1259     int i, h;
1260     h = hash_cached_include(type, filename);
1261     i = s1->cached_includes_hash[h];
1262     for(;;) {
1263         if (i == 0)
1264             break;
1265         e = s1->cached_includes[i - 1];
1266         if (e->type == type && !PATHCMP(e->filename, filename))
1267             return e;
1268         i = e->hash_next;
1269     }
1270     return NULL;
1271 }
1272
1273 static inline void add_cached_include(TCCState *s1, int type,
1274                                       const char *filename, int ifndef_macro)
1275 {
1276     CachedInclude *e;
1277     int h;
1278
1279     if (search_cached_include(s1, type, filename))
1280         return;
1281 #ifdef INC_DEBUG
1282     printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1283 #endif
1284     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1285     if (!e)
1286         return;
1287     e->type = type;
1288     strcpy(e->filename, filename);
1289     e->ifndef_macro = ifndef_macro;
1290     dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1291     /* add in hash table */
1292     h = hash_cached_include(type, filename);
1293     e->hash_next = s1->cached_includes_hash[h];
1294     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1295 }
1296
1297 static void pragma_parse(TCCState *s1)
1298 {
1299     int val;
1300
1301     next();
1302     if (tok == TOK_pack) {
1303         /*
1304           This may be:
1305           #pragma pack(1) // set
1306           #pragma pack() // reset to default
1307           #pragma pack(push,1) // push & set
1308           #pragma pack(pop) // restore previous
1309         */
1310         next();
1311         skip('(');
1312         if (tok == TOK_ASM_pop) {
1313             next();
1314             if (s1->pack_stack_ptr <= s1->pack_stack) {
1315             stk_error:
1316                 error("out of pack stack");
1317             }
1318             s1->pack_stack_ptr--;
1319         } else {
1320             val = 0;
1321             if (tok != ')') {
1322                 if (tok == TOK_ASM_push) {
1323                     next();
1324                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1325                         goto stk_error;
1326                     s1->pack_stack_ptr++;
1327                     skip(',');
1328                 }
1329                 if (tok != TOK_CINT) {
1330                 pack_error:
1331                     error("invalid pack pragma");
1332                 }
1333                 val = tokc.i;
1334                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1335                     goto pack_error;
1336                 next();
1337             }
1338             *s1->pack_stack_ptr = val;
1339             skip(')');
1340         }
1341     }
1342 }
1343
1344 /* is_bof is true if first non space token at beginning of file */
1345 ST_FUNC void preprocess(int is_bof)
1346 {
1347     TCCState *s1 = tcc_state;
1348     int i, c, n, saved_parse_flags;
1349     char buf[1024], *q;
1350     Sym *s;
1351
1352     saved_parse_flags = parse_flags;
1353     parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM |
1354         PARSE_FLAG_LINEFEED;
1355     next_nomacro();
1356  redo:
1357     switch(tok) {
1358     case TOK_DEFINE:
1359         next_nomacro();
1360         parse_define();
1361         break;
1362     case TOK_UNDEF:
1363         next_nomacro();
1364         s = define_find(tok);
1365         /* undefine symbol by putting an invalid name */
1366         if (s)
1367             define_undef(s);
1368         break;
1369     case TOK_INCLUDE:
1370     case TOK_INCLUDE_NEXT:
1371         ch = file->buf_ptr[0];
1372         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1373         skip_spaces();
1374         if (ch == '<') {
1375             c = '>';
1376             goto read_name;
1377         } else if (ch == '\"') {
1378             c = ch;
1379         read_name:
1380             inp();
1381             q = buf;
1382             while (ch != c && ch != '\n' && ch != CH_EOF) {
1383                 if ((q - buf) < sizeof(buf) - 1)
1384                     *q++ = ch;
1385                 if (ch == '\\') {
1386                     if (handle_stray_noerror() == 0)
1387                         --q;
1388                 } else
1389                     inp();
1390             }
1391             *q = '\0';
1392             minp();
1393 #if 0
1394             /* eat all spaces and comments after include */
1395             /* XXX: slightly incorrect */
1396             while (ch1 != '\n' && ch1 != CH_EOF)
1397                 inp();
1398 #endif
1399         } else {
1400             /* computed #include : either we have only strings or
1401                we have anything enclosed in '<>' */
1402             next();
1403             buf[0] = '\0';
1404             if (tok == TOK_STR) {
1405                 while (tok != TOK_LINEFEED) {
1406                     if (tok != TOK_STR) {
1407                     include_syntax:
1408                         error("'#include' expects \"FILENAME\" or <FILENAME>");
1409                     }
1410                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1411                     next();
1412                 }
1413                 c = '\"';
1414             } else {
1415                 int len;
1416                 while (tok != TOK_LINEFEED) {
1417                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1418                     next();
1419                 }
1420                 len = strlen(buf);
1421                 /* check syntax and remove '<>' */
1422                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1423                     goto include_syntax;
1424                 memmove(buf, buf + 1, len - 2);
1425                 buf[len - 2] = '\0';
1426                 c = '>';
1427             }
1428         }
1429
1430         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1431             error("#include recursion too deep");
1432
1433         n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1434         for (i = -2; i < n; ++i) {
1435             char buf1[sizeof file->filename];
1436             BufferedFile *f;
1437             CachedInclude *e;
1438             const char *path;
1439             int size;
1440
1441             if (i == -2) {
1442                 /* check absolute include path */
1443                 if (!IS_ABSPATH(buf))
1444                     continue;
1445                 buf1[0] = 0;
1446
1447             } else if (i == -1) {
1448                 /* search in current dir if "header.h" */
1449                 if (c != '\"')
1450                     continue;
1451                 size = tcc_basename(file->filename) - file->filename;
1452                 memcpy(buf1, file->filename, size);
1453                 buf1[size] = '\0';
1454
1455             } else {
1456                 /* search in all the include paths */
1457                 if (i < s1->nb_include_paths)
1458                     path = s1->include_paths[i];
1459                 else
1460                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1461                 pstrcpy(buf1, sizeof(buf1), path);
1462                 pstrcat(buf1, sizeof(buf1), "/");
1463             }
1464
1465             pstrcat(buf1, sizeof(buf1), buf);
1466
1467             e = search_cached_include(s1, c, buf1);
1468             if (e && define_find(e->ifndef_macro)) {
1469                 /* no need to parse the include because the 'ifndef macro'
1470                    is defined */
1471 #ifdef INC_DEBUG
1472                 printf("%s: skipping %s\n", file->filename, buf);
1473 #endif
1474                 f = NULL;
1475             }  else {
1476                 f = tcc_open(s1, buf1);
1477                 if (!f)
1478                     continue;
1479             }
1480
1481             if (tok == TOK_INCLUDE_NEXT) {
1482                 tok = TOK_INCLUDE;
1483                 if (f)
1484                     tcc_close(f);
1485                 continue;
1486             }
1487
1488             if (!f)
1489                 goto include_done;
1490
1491 #ifdef INC_DEBUG
1492             printf("%s: including %s\n", file->filename, buf1);
1493 #endif
1494
1495            /* XXX: fix current line init */
1496            /* push current file in stack */
1497             *s1->include_stack_ptr++ = file;
1498             f->inc_type = c;
1499             pstrcpy(f->inc_filename, sizeof(f->inc_filename), buf1);
1500             file = f;
1501             /* add include file debug info */
1502             if (s1->do_debug) {
1503                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1504             }
1505             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1506             ch = file->buf_ptr[0];
1507             goto the_end;
1508         }
1509         error("include file '%s' not found", buf);
1510 include_done:
1511         break;
1512     case TOK_IFNDEF:
1513         c = 1;
1514         goto do_ifdef;
1515     case TOK_IF:
1516         c = expr_preprocess();
1517         goto do_if;
1518     case TOK_IFDEF:
1519         c = 0;
1520     do_ifdef:
1521         next_nomacro();
1522         if (tok < TOK_IDENT)
1523             error("invalid argument for '#if%sdef'", c ? "n" : "");
1524         if (is_bof) {
1525             if (c) {
1526 #ifdef INC_DEBUG
1527                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1528 #endif
1529                 file->ifndef_macro = tok;
1530             }
1531         }
1532         c = (define_find(tok) != 0) ^ c;
1533     do_if:
1534         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1535             error("memory full");
1536         *s1->ifdef_stack_ptr++ = c;
1537         goto test_skip;
1538     case TOK_ELSE:
1539         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1540             error("#else without matching #if");
1541         if (s1->ifdef_stack_ptr[-1] & 2)
1542             error("#else after #else");
1543         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1544         goto test_else;
1545     case TOK_ELIF:
1546         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1547             error("#elif without matching #if");
1548         c = s1->ifdef_stack_ptr[-1];
1549         if (c > 1)
1550             error("#elif after #else");
1551         /* last #if/#elif expression was true: we skip */
1552         if (c == 1)
1553             goto skip;
1554         c = expr_preprocess();
1555         s1->ifdef_stack_ptr[-1] = c;
1556     test_else:
1557         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1558             file->ifndef_macro = 0;
1559     test_skip:
1560         if (!(c & 1)) {
1561         skip:
1562             preprocess_skip();
1563             is_bof = 0;
1564             goto redo;
1565         }
1566         break;
1567     case TOK_ENDIF:
1568         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1569             error("#endif without matching #if");
1570         s1->ifdef_stack_ptr--;
1571         /* '#ifndef macro' was at the start of file. Now we check if
1572            an '#endif' is exactly at the end of file */
1573         if (file->ifndef_macro &&
1574             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1575             file->ifndef_macro_saved = file->ifndef_macro;
1576             /* need to set to zero to avoid false matches if another
1577                #ifndef at middle of file */
1578             file->ifndef_macro = 0;
1579             while (tok != TOK_LINEFEED)
1580                 next_nomacro();
1581             tok_flags |= TOK_FLAG_ENDIF;
1582             goto the_end;
1583         }
1584         break;
1585     case TOK_LINE:
1586         next();
1587         if (tok != TOK_CINT)
1588             error("#line");
1589         file->line_num = tokc.i - 1; /* the line number will be incremented after */
1590         next();
1591         if (tok != TOK_LINEFEED) {
1592             if (tok != TOK_STR)
1593                 error("#line");
1594             pstrcpy(file->filename, sizeof(file->filename),
1595                     (char *)tokc.cstr->data);
1596         }
1597         break;
1598     case TOK_ERROR:
1599     case TOK_WARNING:
1600         c = tok;
1601         ch = file->buf_ptr[0];
1602         skip_spaces();
1603         q = buf;
1604         while (ch != '\n' && ch != CH_EOF) {
1605             if ((q - buf) < sizeof(buf) - 1)
1606                 *q++ = ch;
1607             if (ch == '\\') {
1608                 if (handle_stray_noerror() == 0)
1609                     --q;
1610             } else
1611                 inp();
1612         }
1613         *q = '\0';
1614         if (c == TOK_ERROR)
1615             error("#error %s", buf);
1616         else
1617             warning("#warning %s", buf);
1618         break;
1619     case TOK_PRAGMA:
1620         pragma_parse(s1);
1621         break;
1622     default:
1623         if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_PPNUM) {
1624             /* '!' is ignored to allow C scripts. numbers are ignored
1625                to emulate cpp behaviour */
1626         } else {
1627             if (!(saved_parse_flags & PARSE_FLAG_ASM_COMMENTS))
1628                 warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1629         }
1630         break;
1631     }
1632     /* ignore other preprocess commands or #! for C scripts */
1633     while (tok != TOK_LINEFEED)
1634         next_nomacro();
1635  the_end:
1636     parse_flags = saved_parse_flags;
1637 }
1638
1639 /* evaluate escape codes in a string. */
1640 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1641 {
1642     int c, n;
1643     const uint8_t *p;
1644
1645     p = buf;
1646     for(;;) {
1647         c = *p;
1648         if (c == '\0')
1649             break;
1650         if (c == '\\') {
1651             p++;
1652             /* escape */
1653             c = *p;
1654             switch(c) {
1655             case '0': case '1': case '2': case '3':
1656             case '4': case '5': case '6': case '7':
1657                 /* at most three octal digits */
1658                 n = c - '0';
1659                 p++;
1660                 c = *p;
1661                 if (isoct(c)) {
1662                     n = n * 8 + c - '0';
1663                     p++;
1664                     c = *p;
1665                     if (isoct(c)) {
1666                         n = n * 8 + c - '0';
1667                         p++;
1668                     }
1669                 }
1670                 c = n;
1671                 goto add_char_nonext;
1672             case 'x':
1673             case 'u':
1674             case 'U':
1675                 p++;
1676                 n = 0;
1677                 for(;;) {
1678                     c = *p;
1679                     if (c >= 'a' && c <= 'f')
1680                         c = c - 'a' + 10;
1681                     else if (c >= 'A' && c <= 'F')
1682                         c = c - 'A' + 10;
1683                     else if (isnum(c))
1684                         c = c - '0';
1685                     else
1686                         break;
1687                     n = n * 16 + c;
1688                     p++;
1689                 }
1690                 c = n;
1691                 goto add_char_nonext;
1692             case 'a':
1693                 c = '\a';
1694                 break;
1695             case 'b':
1696                 c = '\b';
1697                 break;
1698             case 'f':
1699                 c = '\f';
1700                 break;
1701             case 'n':
1702                 c = '\n';
1703                 break;
1704             case 'r':
1705                 c = '\r';
1706                 break;
1707             case 't':
1708                 c = '\t';
1709                 break;
1710             case 'v':
1711                 c = '\v';
1712                 break;
1713             case 'e':
1714                 if (!gnu_ext)
1715                     goto invalid_escape;
1716                 c = 27;
1717                 break;
1718             case '\'':
1719             case '\"':
1720             case '\\':
1721             case '?':
1722                 break;
1723             default:
1724             invalid_escape:
1725                 if (c >= '!' && c <= '~')
1726                     warning("unknown escape sequence: \'\\%c\'", c);
1727                 else
1728                     warning("unknown escape sequence: \'\\x%x\'", c);
1729                 break;
1730             }
1731         }
1732         p++;
1733     add_char_nonext:
1734         if (!is_long)
1735             cstr_ccat(outstr, c);
1736         else
1737             cstr_wccat(outstr, c);
1738     }
1739     /* add a trailing '\0' */
1740     if (!is_long)
1741         cstr_ccat(outstr, '\0');
1742     else
1743         cstr_wccat(outstr, '\0');
1744 }
1745
1746 /* we use 64 bit numbers */
1747 #define BN_SIZE 2
1748
1749 /* bn = (bn << shift) | or_val */
1750 static void bn_lshift(unsigned int *bn, int shift, int or_val)
1751 {
1752     int i;
1753     unsigned int v;
1754     for(i=0;i<BN_SIZE;i++) {
1755         v = bn[i];
1756         bn[i] = (v << shift) | or_val;
1757         or_val = v >> (32 - shift);
1758     }
1759 }
1760
1761 static void bn_zero(unsigned int *bn)
1762 {
1763     int i;
1764     for(i=0;i<BN_SIZE;i++) {
1765         bn[i] = 0;
1766     }
1767 }
1768
1769 /* parse number in null terminated string 'p' and return it in the
1770    current token */
1771 static void parse_number(const char *p)
1772 {
1773     int b, t, shift, frac_bits, s, exp_val, ch;
1774     char *q;
1775     unsigned int bn[BN_SIZE];
1776     double d;
1777
1778     /* number */
1779     q = token_buf;
1780     ch = *p++;
1781     t = ch;
1782     ch = *p++;
1783     *q++ = t;
1784     b = 10;
1785     if (t == '.') {
1786         goto float_frac_parse;
1787     } else if (t == '0') {
1788         if (ch == 'x' || ch == 'X') {
1789             q--;
1790             ch = *p++;
1791             b = 16;
1792         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
1793             q--;
1794             ch = *p++;
1795             b = 2;
1796         }
1797     }
1798     /* parse all digits. cannot check octal numbers at this stage
1799        because of floating point constants */
1800     while (1) {
1801         if (ch >= 'a' && ch <= 'f')
1802             t = ch - 'a' + 10;
1803         else if (ch >= 'A' && ch <= 'F')
1804             t = ch - 'A' + 10;
1805         else if (isnum(ch))
1806             t = ch - '0';
1807         else
1808             break;
1809         if (t >= b)
1810             break;
1811         if (q >= token_buf + STRING_MAX_SIZE) {
1812         num_too_long:
1813             error("number too long");
1814         }
1815         *q++ = ch;
1816         ch = *p++;
1817     }
1818     if (ch == '.' ||
1819         ((ch == 'e' || ch == 'E') && b == 10) ||
1820         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
1821         if (b != 10) {
1822             /* NOTE: strtox should support that for hexa numbers, but
1823                non ISOC99 libcs do not support it, so we prefer to do
1824                it by hand */
1825             /* hexadecimal or binary floats */
1826             /* XXX: handle overflows */
1827             *q = '\0';
1828             if (b == 16)
1829                 shift = 4;
1830             else
1831                 shift = 2;
1832             bn_zero(bn);
1833             q = token_buf;
1834             while (1) {
1835                 t = *q++;
1836                 if (t == '\0') {
1837                     break;
1838                 } else if (t >= 'a') {
1839                     t = t - 'a' + 10;
1840                 } else if (t >= 'A') {
1841                     t = t - 'A' + 10;
1842                 } else {
1843                     t = t - '0';
1844                 }
1845                 bn_lshift(bn, shift, t);
1846             }
1847             frac_bits = 0;
1848             if (ch == '.') {
1849                 ch = *p++;
1850                 while (1) {
1851                     t = ch;
1852                     if (t >= 'a' && t <= 'f') {
1853                         t = t - 'a' + 10;
1854                     } else if (t >= 'A' && t <= 'F') {
1855                         t = t - 'A' + 10;
1856                     } else if (t >= '0' && t <= '9') {
1857                         t = t - '0';
1858                     } else {
1859                         break;
1860                     }
1861                     if (t >= b)
1862                         error("invalid digit");
1863                     bn_lshift(bn, shift, t);
1864                     frac_bits += shift;
1865                     ch = *p++;
1866                 }
1867             }
1868             if (ch != 'p' && ch != 'P')
1869                 expect("exponent");
1870             ch = *p++;
1871             s = 1;
1872             exp_val = 0;
1873             if (ch == '+') {
1874                 ch = *p++;
1875             } else if (ch == '-') {
1876                 s = -1;
1877                 ch = *p++;
1878             }
1879             if (ch < '0' || ch > '9')
1880                 expect("exponent digits");
1881             while (ch >= '0' && ch <= '9') {
1882                 exp_val = exp_val * 10 + ch - '0';
1883                 ch = *p++;
1884             }
1885             exp_val = exp_val * s;
1886
1887             /* now we can generate the number */
1888             /* XXX: should patch directly float number */
1889             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
1890             d = ldexp(d, exp_val - frac_bits);
1891             t = toup(ch);
1892             if (t == 'F') {
1893                 ch = *p++;
1894                 tok = TOK_CFLOAT;
1895                 /* float : should handle overflow */
1896                 tokc.f = (float)d;
1897             } else if (t == 'L') {
1898                 ch = *p++;
1899 #ifdef TCC_TARGET_PE
1900                 tok = TOK_CDOUBLE;
1901                 tokc.d = d;
1902 #else
1903                 tok = TOK_CLDOUBLE;
1904                 /* XXX: not large enough */
1905                 tokc.ld = (long double)d;
1906 #endif
1907             } else {
1908                 tok = TOK_CDOUBLE;
1909                 tokc.d = d;
1910             }
1911         } else {
1912             /* decimal floats */
1913             if (ch == '.') {
1914                 if (q >= token_buf + STRING_MAX_SIZE)
1915                     goto num_too_long;
1916                 *q++ = ch;
1917                 ch = *p++;
1918             float_frac_parse:
1919                 while (ch >= '0' && ch <= '9') {
1920                     if (q >= token_buf + STRING_MAX_SIZE)
1921                         goto num_too_long;
1922                     *q++ = ch;
1923                     ch = *p++;
1924                 }
1925             }
1926             if (ch == 'e' || ch == 'E') {
1927                 if (q >= token_buf + STRING_MAX_SIZE)
1928                     goto num_too_long;
1929                 *q++ = ch;
1930                 ch = *p++;
1931                 if (ch == '-' || ch == '+') {
1932                     if (q >= token_buf + STRING_MAX_SIZE)
1933                         goto num_too_long;
1934                     *q++ = ch;
1935                     ch = *p++;
1936                 }
1937                 if (ch < '0' || ch > '9')
1938                     expect("exponent digits");
1939                 while (ch >= '0' && ch <= '9') {
1940                     if (q >= token_buf + STRING_MAX_SIZE)
1941                         goto num_too_long;
1942                     *q++ = ch;
1943                     ch = *p++;
1944                 }
1945             }
1946             *q = '\0';
1947             t = toup(ch);
1948             errno = 0;
1949             if (t == 'F') {
1950                 ch = *p++;
1951                 tok = TOK_CFLOAT;
1952                 tokc.f = strtof(token_buf, NULL);
1953             } else if (t == 'L') {
1954                 ch = *p++;
1955 #ifdef TCC_TARGET_PE
1956                 tok = TOK_CDOUBLE;
1957                 tokc.d = strtod(token_buf, NULL);
1958 #else
1959                 tok = TOK_CLDOUBLE;
1960                 tokc.ld = strtold(token_buf, NULL);
1961 #endif
1962             } else {
1963                 tok = TOK_CDOUBLE;
1964                 tokc.d = strtod(token_buf, NULL);
1965             }
1966         }
1967     } else {
1968         unsigned long long n, n1;
1969         int lcount, ucount;
1970
1971         /* integer number */
1972         *q = '\0';
1973         q = token_buf;
1974         if (b == 10 && *q == '0') {
1975             b = 8;
1976             q++;
1977         }
1978         n = 0;
1979         while(1) {
1980             t = *q++;
1981             /* no need for checks except for base 10 / 8 errors */
1982             if (t == '\0') {
1983                 break;
1984             } else if (t >= 'a') {
1985                 t = t - 'a' + 10;
1986             } else if (t >= 'A') {
1987                 t = t - 'A' + 10;
1988             } else {
1989                 t = t - '0';
1990                 if (t >= b)
1991                     error("invalid digit");
1992             }
1993             n1 = n;
1994             n = n * b + t;
1995             /* detect overflow */
1996             /* XXX: this test is not reliable */
1997             if (n < n1)
1998                 error("integer constant overflow");
1999         }
2000
2001         /* XXX: not exactly ANSI compliant */
2002         if ((n & 0xffffffff00000000LL) != 0) {
2003             if ((n >> 63) != 0)
2004                 tok = TOK_CULLONG;
2005             else
2006                 tok = TOK_CLLONG;
2007         } else if (n > 0x7fffffff) {
2008             tok = TOK_CUINT;
2009         } else {
2010             tok = TOK_CINT;
2011         }
2012         lcount = 0;
2013         ucount = 0;
2014         for(;;) {
2015             t = toup(ch);
2016             if (t == 'L') {
2017                 if (lcount >= 2)
2018                     error("three 'l's in integer constant");
2019                 lcount++;
2020                 if (lcount == 2) {
2021                     if (tok == TOK_CINT)
2022                         tok = TOK_CLLONG;
2023                     else if (tok == TOK_CUINT)
2024                         tok = TOK_CULLONG;
2025                 }
2026                 ch = *p++;
2027             } else if (t == 'U') {
2028                 if (ucount >= 1)
2029                     error("two 'u's in integer constant");
2030                 ucount++;
2031                 if (tok == TOK_CINT)
2032                     tok = TOK_CUINT;
2033                 else if (tok == TOK_CLLONG)
2034                     tok = TOK_CULLONG;
2035                 ch = *p++;
2036             } else {
2037                 break;
2038             }
2039         }
2040         if (tok == TOK_CINT || tok == TOK_CUINT)
2041             tokc.ui = n;
2042         else
2043             tokc.ull = n;
2044     }
2045     if (ch)
2046         error("invalid number\n");
2047 }
2048
2049
2050 #define PARSE2(c1, tok1, c2, tok2)              \
2051     case c1:                                    \
2052         PEEKC(c, p);                            \
2053         if (c == c2) {                          \
2054             p++;                                \
2055             tok = tok2;                         \
2056         } else {                                \
2057             tok = tok1;                         \
2058         }                                       \
2059         break;
2060
2061 /* return next token without macro substitution */
2062 static inline void next_nomacro1(void)
2063 {
2064     int t, c, is_long;
2065     TokenSym *ts;
2066     uint8_t *p, *p1;
2067     unsigned int h;
2068
2069     p = file->buf_ptr;
2070  redo_no_start:
2071     c = *p;
2072     switch(c) {
2073     case ' ':
2074     case '\t':
2075         tok = c;
2076         p++;
2077         goto keep_tok_flags;
2078     case '\f':
2079     case '\v':
2080     case '\r':
2081         p++;
2082         goto redo_no_start;
2083     case '\\':
2084         /* first look if it is in fact an end of buffer */
2085         if (p >= file->buf_end) {
2086             file->buf_ptr = p;
2087             handle_eob();
2088             p = file->buf_ptr;
2089             if (p >= file->buf_end)
2090                 goto parse_eof;
2091             else
2092                 goto redo_no_start;
2093         } else {
2094             file->buf_ptr = p;
2095             ch = *p;
2096             handle_stray();
2097             p = file->buf_ptr;
2098             goto redo_no_start;
2099         }
2100     parse_eof:
2101         {
2102             TCCState *s1 = tcc_state;
2103             if ((parse_flags & PARSE_FLAG_LINEFEED)
2104                 && !(tok_flags & TOK_FLAG_EOF)) {
2105                 tok_flags |= TOK_FLAG_EOF;
2106                 tok = TOK_LINEFEED;
2107                 goto keep_tok_flags;
2108             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2109                 tok = TOK_EOF;
2110             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2111                 error("missing #endif");
2112             } else if (s1->include_stack_ptr == s1->include_stack) {
2113                 /* no include left : end of file. */
2114                 tok = TOK_EOF;
2115             } else {
2116                 tok_flags &= ~TOK_FLAG_EOF;
2117                 /* pop include file */
2118
2119                 /* test if previous '#endif' was after a #ifdef at
2120                    start of file */
2121                 if (tok_flags & TOK_FLAG_ENDIF) {
2122 #ifdef INC_DEBUG
2123                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2124 #endif
2125                     add_cached_include(s1, file->inc_type, file->inc_filename,
2126                                        file->ifndef_macro_saved);
2127                 }
2128
2129                 /* add end of include file debug info */
2130                 if (tcc_state->do_debug) {
2131                     put_stabd(N_EINCL, 0, 0);
2132                 }
2133                 /* pop include stack */
2134                 tcc_close(file);
2135                 s1->include_stack_ptr--;
2136                 file = *s1->include_stack_ptr;
2137                 p = file->buf_ptr;
2138                 goto redo_no_start;
2139             }
2140         }
2141         break;
2142
2143     case '\n':
2144         file->line_num++;
2145         tok_flags |= TOK_FLAG_BOL;
2146         p++;
2147 maybe_newline:
2148         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2149             goto redo_no_start;
2150         tok = TOK_LINEFEED;
2151         goto keep_tok_flags;
2152
2153     case '#':
2154         /* XXX: simplify */
2155         PEEKC(c, p);
2156         if ((tok_flags & TOK_FLAG_BOL) &&
2157             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2158             file->buf_ptr = p;
2159             preprocess(tok_flags & TOK_FLAG_BOF);
2160             p = file->buf_ptr;
2161             goto maybe_newline;
2162         } else {
2163             if (c == '#') {
2164                 p++;
2165                 tok = TOK_TWOSHARPS;
2166             } else {
2167                 if (parse_flags & PARSE_FLAG_ASM_COMMENTS) {
2168                     p = parse_line_comment(p - 1);
2169                     goto redo_no_start;
2170                 } else {
2171                     tok = '#';
2172                 }
2173             }
2174         }
2175         break;
2176
2177     case 'a': case 'b': case 'c': case 'd':
2178     case 'e': case 'f': case 'g': case 'h':
2179     case 'i': case 'j': case 'k': case 'l':
2180     case 'm': case 'n': case 'o': case 'p':
2181     case 'q': case 'r': case 's': case 't':
2182     case 'u': case 'v': case 'w': case 'x':
2183     case 'y': case 'z':
2184     case 'A': case 'B': case 'C': case 'D':
2185     case 'E': case 'F': case 'G': case 'H':
2186     case 'I': case 'J': case 'K':
2187     case 'M': case 'N': case 'O': case 'P':
2188     case 'Q': case 'R': case 'S': case 'T':
2189     case 'U': case 'V': case 'W': case 'X':
2190     case 'Y': case 'Z':
2191     case '_':
2192     parse_ident_fast:
2193         p1 = p;
2194         h = TOK_HASH_INIT;
2195         h = TOK_HASH_FUNC(h, c);
2196         p++;
2197         for(;;) {
2198             c = *p;
2199             if (!isidnum_table[c-CH_EOF])
2200                 break;
2201             h = TOK_HASH_FUNC(h, c);
2202             p++;
2203         }
2204         if (c != '\\') {
2205             TokenSym **pts;
2206             int len;
2207
2208             /* fast case : no stray found, so we have the full token
2209                and we have already hashed it */
2210             len = p - p1;
2211             h &= (TOK_HASH_SIZE - 1);
2212             pts = &hash_ident[h];
2213             for(;;) {
2214                 ts = *pts;
2215                 if (!ts)
2216                     break;
2217                 if (ts->len == len && !memcmp(ts->str, p1, len))
2218                     goto token_found;
2219                 pts = &(ts->hash_next);
2220             }
2221             ts = tok_alloc_new(pts, p1, len);
2222         token_found: ;
2223         } else {
2224             /* slower case */
2225             cstr_reset(&tokcstr);
2226
2227             while (p1 < p) {
2228                 cstr_ccat(&tokcstr, *p1);
2229                 p1++;
2230             }
2231             p--;
2232             PEEKC(c, p);
2233         parse_ident_slow:
2234             while (isidnum_table[c-CH_EOF]) {
2235                 cstr_ccat(&tokcstr, c);
2236                 PEEKC(c, p);
2237             }
2238             ts = tok_alloc(tokcstr.data, tokcstr.size);
2239         }
2240         tok = ts->tok;
2241         break;
2242     case 'L':
2243         t = p[1];
2244         if (t != '\\' && t != '\'' && t != '\"') {
2245             /* fast case */
2246             goto parse_ident_fast;
2247         } else {
2248             PEEKC(c, p);
2249             if (c == '\'' || c == '\"') {
2250                 is_long = 1;
2251                 goto str_const;
2252             } else {
2253                 cstr_reset(&tokcstr);
2254                 cstr_ccat(&tokcstr, 'L');
2255                 goto parse_ident_slow;
2256             }
2257         }
2258         break;
2259     case '0': case '1': case '2': case '3':
2260     case '4': case '5': case '6': case '7':
2261     case '8': case '9':
2262
2263         cstr_reset(&tokcstr);
2264         /* after the first digit, accept digits, alpha, '.' or sign if
2265            prefixed by 'eEpP' */
2266     parse_num:
2267         for(;;) {
2268             t = c;
2269             cstr_ccat(&tokcstr, c);
2270             PEEKC(c, p);
2271             if (!(isnum(c) || isid(c) || c == '.' ||
2272                   ((c == '+' || c == '-') &&
2273                    (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2274                 break;
2275         }
2276         /* We add a trailing '\0' to ease parsing */
2277         cstr_ccat(&tokcstr, '\0');
2278         tokc.cstr = &tokcstr;
2279         tok = TOK_PPNUM;
2280         break;
2281     case '.':
2282         /* special dot handling because it can also start a number */
2283         PEEKC(c, p);
2284         if (isnum(c)) {
2285             cstr_reset(&tokcstr);
2286             cstr_ccat(&tokcstr, '.');
2287             goto parse_num;
2288         } else if (c == '.') {
2289             PEEKC(c, p);
2290             if (c != '.')
2291                 expect("'.'");
2292             PEEKC(c, p);
2293             tok = TOK_DOTS;
2294         } else {
2295             tok = '.';
2296         }
2297         break;
2298     case '\'':
2299     case '\"':
2300         is_long = 0;
2301     str_const:
2302         {
2303             CString str;
2304             int sep;
2305
2306             sep = c;
2307
2308             /* parse the string */
2309             cstr_new(&str);
2310             p = parse_pp_string(p, sep, &str);
2311             cstr_ccat(&str, '\0');
2312
2313             /* eval the escape (should be done as TOK_PPNUM) */
2314             cstr_reset(&tokcstr);
2315             parse_escape_string(&tokcstr, str.data, is_long);
2316             cstr_free(&str);
2317
2318             if (sep == '\'') {
2319                 int char_size;
2320                 /* XXX: make it portable */
2321                 if (!is_long)
2322                     char_size = 1;
2323                 else
2324                     char_size = sizeof(nwchar_t);
2325                 if (tokcstr.size <= char_size)
2326                     error("empty character constant");
2327                 if (tokcstr.size > 2 * char_size)
2328                     warning("multi-character character constant");
2329                 if (!is_long) {
2330                     tokc.i = *(int8_t *)tokcstr.data;
2331                     tok = TOK_CCHAR;
2332                 } else {
2333                     tokc.i = *(nwchar_t *)tokcstr.data;
2334                     tok = TOK_LCHAR;
2335                 }
2336             } else {
2337                 tokc.cstr = &tokcstr;
2338                 if (!is_long)
2339                     tok = TOK_STR;
2340                 else
2341                     tok = TOK_LSTR;
2342             }
2343         }
2344         break;
2345
2346     case '<':
2347         PEEKC(c, p);
2348         if (c == '=') {
2349             p++;
2350             tok = TOK_LE;
2351         } else if (c == '<') {
2352             PEEKC(c, p);
2353             if (c == '=') {
2354                 p++;
2355                 tok = TOK_A_SHL;
2356             } else {
2357                 tok = TOK_SHL;
2358             }
2359         } else {
2360             tok = TOK_LT;
2361         }
2362         break;
2363
2364     case '>':
2365         PEEKC(c, p);
2366         if (c == '=') {
2367             p++;
2368             tok = TOK_GE;
2369         } else if (c == '>') {
2370             PEEKC(c, p);
2371             if (c == '=') {
2372                 p++;
2373                 tok = TOK_A_SAR;
2374             } else {
2375                 tok = TOK_SAR;
2376             }
2377         } else {
2378             tok = TOK_GT;
2379         }
2380         break;
2381
2382     case '&':
2383         PEEKC(c, p);
2384         if (c == '&') {
2385             p++;
2386             tok = TOK_LAND;
2387         } else if (c == '=') {
2388             p++;
2389             tok = TOK_A_AND;
2390         } else {
2391             tok = '&';
2392         }
2393         break;
2394
2395     case '|':
2396         PEEKC(c, p);
2397         if (c == '|') {
2398             p++;
2399             tok = TOK_LOR;
2400         } else if (c == '=') {
2401             p++;
2402             tok = TOK_A_OR;
2403         } else {
2404             tok = '|';
2405         }
2406         break;
2407
2408     case '+':
2409         PEEKC(c, p);
2410         if (c == '+') {
2411             p++;
2412             tok = TOK_INC;
2413         } else if (c == '=') {
2414             p++;
2415             tok = TOK_A_ADD;
2416         } else {
2417             tok = '+';
2418         }
2419         break;
2420
2421     case '-':
2422         PEEKC(c, p);
2423         if (c == '-') {
2424             p++;
2425             tok = TOK_DEC;
2426         } else if (c == '=') {
2427             p++;
2428             tok = TOK_A_SUB;
2429         } else if (c == '>') {
2430             p++;
2431             tok = TOK_ARROW;
2432         } else {
2433             tok = '-';
2434         }
2435         break;
2436
2437     PARSE2('!', '!', '=', TOK_NE)
2438     PARSE2('=', '=', '=', TOK_EQ)
2439     PARSE2('*', '*', '=', TOK_A_MUL)
2440     PARSE2('%', '%', '=', TOK_A_MOD)
2441     PARSE2('^', '^', '=', TOK_A_XOR)
2442
2443         /* comments or operator */
2444     case '/':
2445         PEEKC(c, p);
2446         if (c == '*') {
2447             p = parse_comment(p);
2448             goto redo_no_start;
2449         } else if (c == '/') {
2450             p = parse_line_comment(p);
2451             goto redo_no_start;
2452         } else if (c == '=') {
2453             p++;
2454             tok = TOK_A_DIV;
2455         } else {
2456             tok = '/';
2457         }
2458         break;
2459
2460         /* simple tokens */
2461     case '(':
2462     case ')':
2463     case '[':
2464     case ']':
2465     case '{':
2466     case '}':
2467     case ',':
2468     case ';':
2469     case ':':
2470     case '?':
2471     case '~':
2472     case '$': /* only used in assembler */
2473     case '@': /* dito */
2474         tok = c;
2475         p++;
2476         break;
2477     default:
2478         error("unrecognized character \\x%02x", c);
2479         break;
2480     }
2481     tok_flags = 0;
2482 keep_tok_flags:
2483     file->buf_ptr = p;
2484 #if defined(PARSE_DEBUG)
2485     printf("token = %s\n", get_tok_str(tok, &tokc));
2486 #endif
2487 }
2488
2489 /* return next token without macro substitution. Can read input from
2490    macro_ptr buffer */
2491 static void next_nomacro_spc(void)
2492 {
2493     if (macro_ptr) {
2494     redo:
2495         tok = *macro_ptr;
2496         if (tok) {
2497             TOK_GET(&tok, &macro_ptr, &tokc);
2498             if (tok == TOK_LINENUM) {
2499                 file->line_num = tokc.i;
2500                 goto redo;
2501             }
2502         }
2503     } else {
2504         next_nomacro1();
2505     }
2506 }
2507
2508 ST_FUNC void next_nomacro(void)
2509 {
2510     do {
2511         next_nomacro_spc();
2512     } while (is_space(tok));
2513 }
2514
2515 /* substitute args in macro_str and return allocated string */
2516 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2517 {
2518     int last_tok, t, spc;
2519     const int *st;
2520     Sym *s;
2521     CValue cval;
2522     TokenString str;
2523     CString cstr;
2524
2525     tok_str_new(&str);
2526     last_tok = 0;
2527     while(1) {
2528         TOK_GET(&t, &macro_str, &cval);
2529         if (!t)
2530             break;
2531         if (t == '#') {
2532             /* stringize */
2533             TOK_GET(&t, &macro_str, &cval);
2534             if (!t)
2535                 break;
2536             s = sym_find2(args, t);
2537             if (s) {
2538                 cstr_new(&cstr);
2539                 st = s->d;
2540                 spc = 0;
2541                 while (*st) {
2542                     TOK_GET(&t, &st, &cval);
2543                     if (!check_space(t, &spc))
2544                         cstr_cat(&cstr, get_tok_str(t, &cval));
2545                 }
2546                 cstr.size -= spc;
2547                 cstr_ccat(&cstr, '\0');
2548 #ifdef PP_DEBUG
2549                 printf("stringize: %s\n", (char *)cstr.data);
2550 #endif
2551                 /* add string */
2552                 cval.cstr = &cstr;
2553                 tok_str_add2(&str, TOK_STR, &cval);
2554                 cstr_free(&cstr);
2555             } else {
2556                 tok_str_add2(&str, t, &cval);
2557             }
2558         } else if (t >= TOK_IDENT) {
2559             s = sym_find2(args, t);
2560             if (s) {
2561                 st = s->d;
2562                 /* if '##' is present before or after, no arg substitution */
2563                 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2564                     /* special case for var arg macros : ## eats the
2565                        ',' if empty VA_ARGS variable. */
2566                     /* XXX: test of the ',' is not 100%
2567                        reliable. should fix it to avoid security
2568                        problems */
2569                     if (gnu_ext && s->type.t &&
2570                         last_tok == TOK_TWOSHARPS &&
2571                         str.len >= 2 && str.str[str.len - 2] == ',') {
2572                         if (*st == 0) {
2573                             /* suppress ',' '##' */
2574                             str.len -= 2;
2575                         } else {
2576                             /* suppress '##' and add variable */
2577                             str.len--;
2578                             goto add_var;
2579                         }
2580                     } else {
2581                         int t1;
2582                     add_var:
2583                         for(;;) {
2584                             TOK_GET(&t1, &st, &cval);
2585                             if (!t1)
2586                                 break;
2587                             tok_str_add2(&str, t1, &cval);
2588                         }
2589                     }
2590                 } else {
2591                     /* NOTE: the stream cannot be read when macro
2592                        substituing an argument */
2593                     macro_subst(&str, nested_list, st, NULL);
2594                 }
2595             } else {
2596                 tok_str_add(&str, t);
2597             }
2598         } else {
2599             tok_str_add2(&str, t, &cval);
2600         }
2601         last_tok = t;
2602     }
2603     tok_str_add(&str, 0);
2604     return str.str;
2605 }
2606
2607 static char const ab_month_name[12][4] =
2608 {
2609     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2610     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2611 };
2612
2613 /* do macro substitution of current token with macro 's' and add
2614    result to (tok_str,tok_len). 'nested_list' is the list of all
2615    macros we got inside to avoid recursing. Return non zero if no
2616    substitution needs to be done */
2617 static int macro_subst_tok(TokenString *tok_str,
2618                            Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2619 {
2620     Sym *args, *sa, *sa1;
2621     int mstr_allocated, parlevel, *mstr, t, t1, spc;
2622     const int *p;
2623     TokenString str;
2624     char *cstrval;
2625     CValue cval;
2626     CString cstr;
2627     char buf[32];
2628
2629     /* if symbol is a macro, prepare substitution */
2630     /* special macros */
2631     if (tok == TOK___LINE__) {
2632         snprintf(buf, sizeof(buf), "%d", file->line_num);
2633         cstrval = buf;
2634         t1 = TOK_PPNUM;
2635         goto add_cstr1;
2636     } else if (tok == TOK___FILE__) {
2637         cstrval = file->filename;
2638         goto add_cstr;
2639     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2640         time_t ti;
2641         struct tm *tm;
2642
2643         time(&ti);
2644         tm = localtime(&ti);
2645         if (tok == TOK___DATE__) {
2646             snprintf(buf, sizeof(buf), "%s %2d %d",
2647                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2648         } else {
2649             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2650                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2651         }
2652         cstrval = buf;
2653     add_cstr:
2654         t1 = TOK_STR;
2655     add_cstr1:
2656         cstr_new(&cstr);
2657         cstr_cat(&cstr, cstrval);
2658         cstr_ccat(&cstr, '\0');
2659         cval.cstr = &cstr;
2660         tok_str_add2(tok_str, t1, &cval);
2661         cstr_free(&cstr);
2662     } else {
2663         mstr = s->d;
2664         mstr_allocated = 0;
2665         if (s->type.t == MACRO_FUNC) {
2666             /* NOTE: we do not use next_nomacro to avoid eating the
2667                next token. XXX: find better solution */
2668         redo:
2669             if (macro_ptr) {
2670                 p = macro_ptr;
2671                 while (is_space(t = *p) || TOK_LINEFEED == t)
2672                     ++p;
2673                 if (t == 0 && can_read_stream) {
2674                     /* end of macro stream: we must look at the token
2675                        after in the file */
2676                     struct macro_level *ml = *can_read_stream;
2677                     macro_ptr = NULL;
2678                     if (ml)
2679                     {
2680                         macro_ptr = ml->p;
2681                         ml->p = NULL;
2682                         *can_read_stream = ml -> prev;
2683                     }
2684                     goto redo;
2685                 }
2686             } else {
2687                 /* XXX: incorrect with comments */
2688                 ch = file->buf_ptr[0];
2689                 while (is_space(ch) || ch == '\n')
2690                     cinp();
2691                 t = ch;
2692             }
2693             if (t != '(') /* no macro subst */
2694                 return -1;
2695
2696             /* argument macro */
2697             next_nomacro();
2698             next_nomacro();
2699             args = NULL;
2700             sa = s->next;
2701             /* NOTE: empty args are allowed, except if no args */
2702             for(;;) {
2703                 /* handle '()' case */
2704                 if (!args && !sa && tok == ')')
2705                     break;
2706                 if (!sa)
2707                     error("macro '%s' used with too many args",
2708                           get_tok_str(s->v, 0));
2709                 tok_str_new(&str);
2710                 parlevel = spc = 0;
2711                 /* NOTE: non zero sa->t indicates VA_ARGS */
2712                 while ((parlevel > 0 ||
2713                         (tok != ')' &&
2714                          (tok != ',' || sa->type.t))) &&
2715                        tok != -1) {
2716                     if (tok == '(')
2717                         parlevel++;
2718                     else if (tok == ')')
2719                         parlevel--;
2720                     if (tok == TOK_LINEFEED)
2721                         tok = ' ';
2722                     if (!check_space(tok, &spc))
2723                         tok_str_add2(&str, tok, &tokc);
2724                     next_nomacro_spc();
2725                 }
2726                 str.len -= spc;
2727                 tok_str_add(&str, 0);
2728                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
2729                 sa1->d = str.str;
2730                 sa = sa->next;
2731                 if (tok == ')') {
2732                     /* special case for gcc var args: add an empty
2733                        var arg argument if it is omitted */
2734                     if (sa && sa->type.t && gnu_ext)
2735                         continue;
2736                     else
2737                         break;
2738                 }
2739                 if (tok != ',')
2740                     expect(",");
2741                 next_nomacro();
2742             }
2743             if (sa) {
2744                 error("macro '%s' used with too few args",
2745                       get_tok_str(s->v, 0));
2746             }
2747
2748             /* now subst each arg */
2749             mstr = macro_arg_subst(nested_list, mstr, args);
2750             /* free memory */
2751             sa = args;
2752             while (sa) {
2753                 sa1 = sa->prev;
2754                 tok_str_free(sa->d);
2755                 sym_free(sa);
2756                 sa = sa1;
2757             }
2758             mstr_allocated = 1;
2759         }
2760         sym_push2(nested_list, s->v, 0, 0);
2761         macro_subst(tok_str, nested_list, mstr, can_read_stream);
2762         /* pop nested defined symbol */
2763         sa1 = *nested_list;
2764         *nested_list = sa1->prev;
2765         sym_free(sa1);
2766         if (mstr_allocated)
2767             tok_str_free(mstr);
2768     }
2769     return 0;
2770 }
2771
2772 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2773    return the resulting string (which must be freed). */
2774 static inline int *macro_twosharps(const int *macro_str)
2775 {
2776     const int *ptr;
2777     int t;
2778     CValue cval;
2779     TokenString macro_str1;
2780     CString cstr;
2781     char *p;
2782     int n;
2783
2784     /* we search the first '##' */
2785     for(ptr = macro_str;;) {
2786         TOK_GET(&t, &ptr, &cval);
2787         if (t == TOK_TWOSHARPS)
2788             break;
2789         /* nothing more to do if end of string */
2790         if (t == 0)
2791             return NULL;
2792     }
2793
2794     /* we saw '##', so we need more processing to handle it */
2795     tok_str_new(&macro_str1);
2796     for(ptr = macro_str;;) {
2797         TOK_GET(&tok, &ptr, &tokc);
2798         if (tok == 0)
2799             break;
2800         if (tok == TOK_TWOSHARPS)
2801             continue;
2802         while (*ptr == TOK_TWOSHARPS) {
2803             t = *++ptr;
2804             if (t && t != TOK_TWOSHARPS) {
2805                 TOK_GET(&t, &ptr, &cval);
2806
2807                 /* We concatenate the two tokens */
2808                 cstr_new(&cstr);
2809                 cstr_cat(&cstr, get_tok_str(tok, &tokc));
2810                 n = cstr.size;
2811                 cstr_cat(&cstr, get_tok_str(t, &cval));
2812                 cstr_ccat(&cstr, '\0');
2813
2814                 p = file->buf_ptr;
2815                 file->buf_ptr = cstr.data;
2816                 for (;;) {
2817                     next_nomacro1();
2818                     if (0 == *file->buf_ptr)
2819                         break;
2820                     tok_str_add2(&macro_str1, tok, &tokc);
2821
2822                     warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
2823                         n, cstr.data, (char*)cstr.data + n);
2824                 }
2825                 file->buf_ptr = p;
2826                 cstr_reset(&cstr);
2827             }
2828         }
2829         tok_str_add2(&macro_str1, tok, &tokc);
2830     }
2831     tok_str_add(&macro_str1, 0);
2832     return macro_str1.str;
2833 }
2834
2835
2836 /* do macro substitution of macro_str and add result to
2837    (tok_str,tok_len). 'nested_list' is the list of all macros we got
2838    inside to avoid recursing. */
2839 static void macro_subst(TokenString *tok_str, Sym **nested_list,
2840                         const int *macro_str, struct macro_level ** can_read_stream)
2841 {
2842     Sym *s;
2843     int *macro_str1;
2844     const int *ptr;
2845     int t, ret, spc;
2846     CValue cval;
2847     struct macro_level ml;
2848
2849     /* first scan for '##' operator handling */
2850     ptr = macro_str;
2851     macro_str1 = macro_twosharps(ptr);
2852     if (macro_str1)
2853         ptr = macro_str1;
2854     spc = 0;
2855     while (1) {
2856         /* NOTE: ptr == NULL can only happen if tokens are read from
2857            file stream due to a macro function call */
2858         if (ptr == NULL)
2859             break;
2860         TOK_GET(&t, &ptr, &cval);
2861         if (t == 0)
2862             break;
2863         s = define_find(t);
2864         if (s != NULL) {
2865             /* if nested substitution, do nothing */
2866             if (sym_find2(*nested_list, t))
2867                 goto no_subst;
2868             ml.p = macro_ptr;
2869             if (can_read_stream)
2870                 ml.prev = *can_read_stream, *can_read_stream = &ml;
2871             macro_ptr = (int *)ptr;
2872             tok = t;
2873             ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
2874             ptr = (int *)macro_ptr;
2875             macro_ptr = ml.p;
2876             if (can_read_stream && *can_read_stream == &ml)
2877                 *can_read_stream = ml.prev;
2878             if (ret != 0)
2879                 goto no_subst;
2880         } else {
2881         no_subst:
2882             if (!check_space(t, &spc))
2883                 tok_str_add2(tok_str, t, &cval);
2884         }
2885     }
2886     if (macro_str1)
2887         tok_str_free(macro_str1);
2888 }
2889
2890 /* return next token with macro substitution */
2891 ST_FUNC void next(void)
2892 {
2893     Sym *nested_list, *s;
2894     TokenString str;
2895     struct macro_level *ml;
2896
2897  redo:
2898     if (parse_flags & PARSE_FLAG_SPACES)
2899         next_nomacro_spc();
2900     else
2901         next_nomacro();
2902     if (!macro_ptr) {
2903         /* if not reading from macro substituted string, then try
2904            to substitute macros */
2905         if (tok >= TOK_IDENT &&
2906             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2907             s = define_find(tok);
2908             if (s) {
2909                 /* we have a macro: we try to substitute */
2910                 tok_str_new(&str);
2911                 nested_list = NULL;
2912                 ml = NULL;
2913                 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
2914                     /* substitution done, NOTE: maybe empty */
2915                     tok_str_add(&str, 0);
2916                     macro_ptr = str.str;
2917                     macro_ptr_allocated = str.str;
2918                     goto redo;
2919                 }
2920             }
2921         }
2922     } else {
2923         if (tok == 0) {
2924             /* end of macro or end of unget buffer */
2925             if (unget_buffer_enabled) {
2926                 macro_ptr = unget_saved_macro_ptr;
2927                 unget_buffer_enabled = 0;
2928             } else {
2929                 /* end of macro string: free it */
2930                 tok_str_free(macro_ptr_allocated);
2931                 macro_ptr_allocated = NULL;
2932                 macro_ptr = NULL;
2933             }
2934             goto redo;
2935         }
2936     }
2937
2938     /* convert preprocessor tokens into C tokens */
2939     if (tok == TOK_PPNUM &&
2940         (parse_flags & PARSE_FLAG_TOK_NUM)) {
2941         parse_number((char *)tokc.cstr->data);
2942     }
2943 }
2944
2945 /* push back current token and set current token to 'last_tok'. Only
2946    identifier case handled for labels. */
2947 ST_INLN void unget_tok(int last_tok)
2948 {
2949     int i, n;
2950     int *q;
2951     unget_saved_macro_ptr = macro_ptr;
2952     unget_buffer_enabled = 1;
2953     q = unget_saved_buffer;
2954     macro_ptr = q;
2955     *q++ = tok;
2956     n = tok_ext_size(tok) - 1;
2957     for(i=0;i<n;i++)
2958         *q++ = tokc.tab[i];
2959     *q = 0; /* end of token string */
2960     tok = last_tok;
2961 }
2962
2963
2964 /* better than nothing, but needs extension to handle '-E' option
2965    correctly too */
2966 ST_FUNC void preprocess_init(TCCState *s1)
2967 {
2968     s1->include_stack_ptr = s1->include_stack;
2969     /* XXX: move that before to avoid having to initialize
2970        file->ifdef_stack_ptr ? */
2971     s1->ifdef_stack_ptr = s1->ifdef_stack;
2972     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
2973
2974     /* XXX: not ANSI compliant: bound checking says error */
2975     vtop = vstack - 1;
2976     s1->pack_stack[0] = 0;
2977     s1->pack_stack_ptr = s1->pack_stack;
2978 }
2979
2980 ST_FUNC void preprocess_new()
2981 {
2982     int i, c;
2983     const char *p, *r;
2984     TokenSym *ts;
2985
2986     /* init isid table */
2987     for(i=CH_EOF;i<256;i++)
2988         isidnum_table[i-CH_EOF] = isid(i) || isnum(i);
2989
2990     /* add all tokens */
2991     table_ident = NULL;
2992     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
2993
2994     tok_ident = TOK_IDENT;
2995     p = tcc_keywords;
2996     while (*p) {
2997         r = p;
2998         for(;;) {
2999             c = *r++;
3000             if (c == '\0')
3001                 break;
3002         }
3003         ts = tok_alloc(p, r - p - 1);
3004         p = r;
3005     }
3006 }
3007
3008 /* Preprocess the current file */
3009 ST_FUNC int tcc_preprocess(TCCState *s1)
3010 {
3011     Sym *define_start;
3012
3013     BufferedFile *file_ref, **iptr, **iptr_new;
3014     int token_seen, line_ref, d;
3015     const char *s;
3016
3017     preprocess_init(s1);
3018     define_start = define_stack;
3019     ch = file->buf_ptr[0];
3020     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3021     parse_flags = PARSE_FLAG_ASM_COMMENTS | PARSE_FLAG_PREPROCESS |
3022         PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
3023     token_seen = 0;
3024     line_ref = 0;
3025     file_ref = NULL;
3026     iptr = s1->include_stack_ptr;
3027
3028     for (;;) {
3029         next();
3030         if (tok == TOK_EOF) {
3031             break;
3032         } else if (file != file_ref) {
3033             goto print_line;
3034         } else if (tok == TOK_LINEFEED) {
3035             if (!token_seen)
3036                 continue;
3037             ++line_ref;
3038             token_seen = 0;
3039         } else if (!token_seen) {
3040             d = file->line_num - line_ref;
3041             if (file != file_ref || d < 0 || d >= 8) {
3042 print_line:
3043                 iptr_new = s1->include_stack_ptr;
3044                 s = iptr_new > iptr ? " 1"
3045                   : iptr_new < iptr ? " 2"
3046                   : iptr_new > s1->include_stack ? " 3"
3047                   : ""
3048                   ;
3049                 iptr = iptr_new;
3050                 fprintf(s1->outfile, "# %d \"%s\"%s\n", file->line_num, file->filename, s);
3051             } else {
3052                 while (d)
3053                     fputs("\n", s1->outfile), --d;
3054             }
3055             line_ref = (file_ref = file)->line_num;
3056             token_seen = tok != TOK_LINEFEED;
3057             if (!token_seen)
3058                 continue;
3059         }
3060         fputs(get_tok_str(tok, &tokc), s1->outfile);
3061     }
3062     free_defines(define_start);
3063     return 0;
3064 }