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, &tokc));
  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             CachedInclude *e;
1437             const char *path;
1438             int size, fd;
1439
1440             if (i == -2) {
1441                 /* check absolute include path */
1442                 if (!IS_ABSPATH(buf))
1443                     continue;
1444                 buf1[0] = 0;
1445
1446             } else if (i == -1) {
1447                 /* search in current dir if "header.h" */
1448                 if (c != '\"')
1449                     continue;
1450                 size = tcc_basename(file->filename) - file->filename;
1451                 memcpy(buf1, file->filename, size);
1452                 buf1[size] = '\0';
1453
1454             } else {
1455                 /* search in all the include paths */
1456                 if (i < s1->nb_include_paths)
1457                     path = s1->include_paths[i];
1458                 else
1459                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1460                 pstrcpy(buf1, sizeof(buf1), path);
1461                 pstrcat(buf1, sizeof(buf1), "/");
1462             }
1463
1464             pstrcat(buf1, sizeof(buf1), buf);
1465
1466             e = search_cached_include(s1, c, buf1);
1467             if (e && define_find(e->ifndef_macro)) {
1468                 /* no need to parse the include because the 'ifndef macro'
1469                    is defined */
1470 #ifdef INC_DEBUG
1471                 printf("%s: skipping %s\n", file->filename, buf);
1472 #endif
1473                 fd = 0;
1474             } else {
1475                 fd = tcc_open(s1, buf1);
1476                 if (fd < 0)
1477                     continue;
1478             }
1479
1480             if (tok == TOK_INCLUDE_NEXT) {
1481                 tok = TOK_INCLUDE;
1482                 if (fd)
1483                     tcc_close();
1484                 continue;
1485             }
1486
1487             if (0 == fd)
1488                 goto include_done;
1489
1490 #ifdef INC_DEBUG
1491             printf("%s: including %s\n", file->filename, buf1);
1492 #endif
1493             /* update target deps */
1494             dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1495                     tcc_strdup(buf1));
1496            /* XXX: fix current line init */
1497            /* push current file in stack */
1498             *s1->include_stack_ptr++ = file->prev;
1499             file->inc_type = c;
1500             pstrcpy(file->inc_filename, sizeof(file->inc_filename), buf1);
1501             /* add include file debug info */
1502             if (s1->do_debug)
1503                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1504             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1505             ch = file->buf_ptr[0];
1506             goto the_end;
1507         }
1508         error("include file '%s' not found", buf);
1509 include_done:
1510         break;
1511     case TOK_IFNDEF:
1512         c = 1;
1513         goto do_ifdef;
1514     case TOK_IF:
1515         c = expr_preprocess();
1516         goto do_if;
1517     case TOK_IFDEF:
1518         c = 0;
1519     do_ifdef:
1520         next_nomacro();
1521         if (tok < TOK_IDENT)
1522             error("invalid argument for '#if%sdef'", c ? "n" : "");
1523         if (is_bof) {
1524             if (c) {
1525 #ifdef INC_DEBUG
1526                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1527 #endif
1528                 file->ifndef_macro = tok;
1529             }
1530         }
1531         c = (define_find(tok) != 0) ^ c;
1532     do_if:
1533         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1534             error("memory full");
1535         *s1->ifdef_stack_ptr++ = c;
1536         goto test_skip;
1537     case TOK_ELSE:
1538         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1539             error("#else without matching #if");
1540         if (s1->ifdef_stack_ptr[-1] & 2)
1541             error("#else after #else");
1542         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1543         goto test_else;
1544     case TOK_ELIF:
1545         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1546             error("#elif without matching #if");
1547         c = s1->ifdef_stack_ptr[-1];
1548         if (c > 1)
1549             error("#elif after #else");
1550         /* last #if/#elif expression was true: we skip */
1551         if (c == 1)
1552             goto skip;
1553         c = expr_preprocess();
1554         s1->ifdef_stack_ptr[-1] = c;
1555     test_else:
1556         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1557             file->ifndef_macro = 0;
1558     test_skip:
1559         if (!(c & 1)) {
1560         skip:
1561             preprocess_skip();
1562             is_bof = 0;
1563             goto redo;
1564         }
1565         break;
1566     case TOK_ENDIF:
1567         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1568             error("#endif without matching #if");
1569         s1->ifdef_stack_ptr--;
1570         /* '#ifndef macro' was at the start of file. Now we check if
1571            an '#endif' is exactly at the end of file */
1572         if (file->ifndef_macro &&
1573             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1574             file->ifndef_macro_saved = file->ifndef_macro;
1575             /* need to set to zero to avoid false matches if another
1576                #ifndef at middle of file */
1577             file->ifndef_macro = 0;
1578             while (tok != TOK_LINEFEED)
1579                 next_nomacro();
1580             tok_flags |= TOK_FLAG_ENDIF;
1581             goto the_end;
1582         }
1583         break;
1584     case TOK_LINE:
1585         next();
1586         if (tok != TOK_CINT)
1587             error("#line");
1588         file->line_num = tokc.i - 1; /* the line number will be incremented after */
1589         next();
1590         if (tok != TOK_LINEFEED) {
1591             if (tok != TOK_STR)
1592                 error("#line");
1593             pstrcpy(file->filename, sizeof(file->filename),
1594                     (char *)tokc.cstr->data);
1595         }
1596         break;
1597     case TOK_ERROR:
1598     case TOK_WARNING:
1599         c = tok;
1600         ch = file->buf_ptr[0];
1601         skip_spaces();
1602         q = buf;
1603         while (ch != '\n' && ch != CH_EOF) {
1604             if ((q - buf) < sizeof(buf) - 1)
1605                 *q++ = ch;
1606             if (ch == '\\') {
1607                 if (handle_stray_noerror() == 0)
1608                     --q;
1609             } else
1610                 inp();
1611         }
1612         *q = '\0';
1613         if (c == TOK_ERROR)
1614             error("#error %s", buf);
1615         else
1616             warning("#warning %s", buf);
1617         break;
1618     case TOK_PRAGMA:
1619         pragma_parse(s1);
1620         break;
1621     default:
1622         if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_PPNUM) {
1623             /* '!' is ignored to allow C scripts. numbers are ignored
1624                to emulate cpp behaviour */
1625         } else {
1626             if (!(saved_parse_flags & PARSE_FLAG_ASM_COMMENTS))
1627                 warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1628         }
1629         break;
1630     }
1631     /* ignore other preprocess commands or #! for C scripts */
1632     while (tok != TOK_LINEFEED)
1633         next_nomacro();
1634  the_end:
1635     parse_flags = saved_parse_flags;
1636 }
1637
1638 /* evaluate escape codes in a string. */
1639 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1640 {
1641     int c, n;
1642     const uint8_t *p;
1643
1644     p = buf;
1645     for(;;) {
1646         c = *p;
1647         if (c == '\0')
1648             break;
1649         if (c == '\\') {
1650             p++;
1651             /* escape */
1652             c = *p;
1653             switch(c) {
1654             case '0': case '1': case '2': case '3':
1655             case '4': case '5': case '6': case '7':
1656                 /* at most three octal digits */
1657                 n = c - '0';
1658                 p++;
1659                 c = *p;
1660                 if (isoct(c)) {
1661                     n = n * 8 + c - '0';
1662                     p++;
1663                     c = *p;
1664                     if (isoct(c)) {
1665                         n = n * 8 + c - '0';
1666                         p++;
1667                     }
1668                 }
1669                 c = n;
1670                 goto add_char_nonext;
1671             case 'x':
1672             case 'u':
1673             case 'U':
1674                 p++;
1675                 n = 0;
1676                 for(;;) {
1677                     c = *p;
1678                     if (c >= 'a' && c <= 'f')
1679                         c = c - 'a' + 10;
1680                     else if (c >= 'A' && c <= 'F')
1681                         c = c - 'A' + 10;
1682                     else if (isnum(c))
1683                         c = c - '0';
1684                     else
1685                         break;
1686                     n = n * 16 + c;
1687                     p++;
1688                 }
1689                 c = n;
1690                 goto add_char_nonext;
1691             case 'a':
1692                 c = '\a';
1693                 break;
1694             case 'b':
1695                 c = '\b';
1696                 break;
1697             case 'f':
1698                 c = '\f';
1699                 break;
1700             case 'n':
1701                 c = '\n';
1702                 break;
1703             case 'r':
1704                 c = '\r';
1705                 break;
1706             case 't':
1707                 c = '\t';
1708                 break;
1709             case 'v':
1710                 c = '\v';
1711                 break;
1712             case 'e':
1713                 if (!gnu_ext)
1714                     goto invalid_escape;
1715                 c = 27;
1716                 break;
1717             case '\'':
1718             case '\"':
1719             case '\\':
1720             case '?':
1721                 break;
1722             default:
1723             invalid_escape:
1724                 if (c >= '!' && c <= '~')
1725                     warning("unknown escape sequence: \'\\%c\'", c);
1726                 else
1727                     warning("unknown escape sequence: \'\\x%x\'", c);
1728                 break;
1729             }
1730         }
1731         p++;
1732     add_char_nonext:
1733         if (!is_long)
1734             cstr_ccat(outstr, c);
1735         else
1736             cstr_wccat(outstr, c);
1737     }
1738     /* add a trailing '\0' */
1739     if (!is_long)
1740         cstr_ccat(outstr, '\0');
1741     else
1742         cstr_wccat(outstr, '\0');
1743 }
1744
1745 /* we use 64 bit numbers */
1746 #define BN_SIZE 2
1747
1748 /* bn = (bn << shift) | or_val */
1749 static void bn_lshift(unsigned int *bn, int shift, int or_val)
1750 {
1751     int i;
1752     unsigned int v;
1753     for(i=0;i<BN_SIZE;i++) {
1754         v = bn[i];
1755         bn[i] = (v << shift) | or_val;
1756         or_val = v >> (32 - shift);
1757     }
1758 }
1759
1760 static void bn_zero(unsigned int *bn)
1761 {
1762     int i;
1763     for(i=0;i<BN_SIZE;i++) {
1764         bn[i] = 0;
1765     }
1766 }
1767
1768 /* parse number in null terminated string 'p' and return it in the
1769    current token */
1770 static void parse_number(const char *p)
1771 {
1772     int b, t, shift, frac_bits, s, exp_val, ch;
1773     char *q;
1774     unsigned int bn[BN_SIZE];
1775     double d;
1776
1777     /* number */
1778     q = token_buf;
1779     ch = *p++;
1780     t = ch;
1781     ch = *p++;
1782     *q++ = t;
1783     b = 10;
1784     if (t == '.') {
1785         goto float_frac_parse;
1786     } else if (t == '0') {
1787         if (ch == 'x' || ch == 'X') {
1788             q--;
1789             ch = *p++;
1790             b = 16;
1791         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
1792             q--;
1793             ch = *p++;
1794             b = 2;
1795         }
1796     }
1797     /* parse all digits. cannot check octal numbers at this stage
1798        because of floating point constants */
1799     while (1) {
1800         if (ch >= 'a' && ch <= 'f')
1801             t = ch - 'a' + 10;
1802         else if (ch >= 'A' && ch <= 'F')
1803             t = ch - 'A' + 10;
1804         else if (isnum(ch))
1805             t = ch - '0';
1806         else
1807             break;
1808         if (t >= b)
1809             break;
1810         if (q >= token_buf + STRING_MAX_SIZE) {
1811         num_too_long:
1812             error("number too long");
1813         }
1814         *q++ = ch;
1815         ch = *p++;
1816     }
1817     if (ch == '.' ||
1818         ((ch == 'e' || ch == 'E') && b == 10) ||
1819         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
1820         if (b != 10) {
1821             /* NOTE: strtox should support that for hexa numbers, but
1822                non ISOC99 libcs do not support it, so we prefer to do
1823                it by hand */
1824             /* hexadecimal or binary floats */
1825             /* XXX: handle overflows */
1826             *q = '\0';
1827             if (b == 16)
1828                 shift = 4;
1829             else
1830                 shift = 2;
1831             bn_zero(bn);
1832             q = token_buf;
1833             while (1) {
1834                 t = *q++;
1835                 if (t == '\0') {
1836                     break;
1837                 } else if (t >= 'a') {
1838                     t = t - 'a' + 10;
1839                 } else if (t >= 'A') {
1840                     t = t - 'A' + 10;
1841                 } else {
1842                     t = t - '0';
1843                 }
1844                 bn_lshift(bn, shift, t);
1845             }
1846             frac_bits = 0;
1847             if (ch == '.') {
1848                 ch = *p++;
1849                 while (1) {
1850                     t = ch;
1851                     if (t >= 'a' && t <= 'f') {
1852                         t = t - 'a' + 10;
1853                     } else if (t >= 'A' && t <= 'F') {
1854                         t = t - 'A' + 10;
1855                     } else if (t >= '0' && t <= '9') {
1856                         t = t - '0';
1857                     } else {
1858                         break;
1859                     }
1860                     if (t >= b)
1861                         error("invalid digit");
1862                     bn_lshift(bn, shift, t);
1863                     frac_bits += shift;
1864                     ch = *p++;
1865                 }
1866             }
1867             if (ch != 'p' && ch != 'P')
1868                 expect("exponent");
1869             ch = *p++;
1870             s = 1;
1871             exp_val = 0;
1872             if (ch == '+') {
1873                 ch = *p++;
1874             } else if (ch == '-') {
1875                 s = -1;
1876                 ch = *p++;
1877             }
1878             if (ch < '0' || ch > '9')
1879                 expect("exponent digits");
1880             while (ch >= '0' && ch <= '9') {
1881                 exp_val = exp_val * 10 + ch - '0';
1882                 ch = *p++;
1883             }
1884             exp_val = exp_val * s;
1885
1886             /* now we can generate the number */
1887             /* XXX: should patch directly float number */
1888             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
1889             d = ldexp(d, exp_val - frac_bits);
1890             t = toup(ch);
1891             if (t == 'F') {
1892                 ch = *p++;
1893                 tok = TOK_CFLOAT;
1894                 /* float : should handle overflow */
1895                 tokc.f = (float)d;
1896             } else if (t == 'L') {
1897                 ch = *p++;
1898 #ifdef TCC_TARGET_PE
1899                 tok = TOK_CDOUBLE;
1900                 tokc.d = d;
1901 #else
1902                 tok = TOK_CLDOUBLE;
1903                 /* XXX: not large enough */
1904                 tokc.ld = (long double)d;
1905 #endif
1906             } else {
1907                 tok = TOK_CDOUBLE;
1908                 tokc.d = d;
1909             }
1910         } else {
1911             /* decimal floats */
1912             if (ch == '.') {
1913                 if (q >= token_buf + STRING_MAX_SIZE)
1914                     goto num_too_long;
1915                 *q++ = ch;
1916                 ch = *p++;
1917             float_frac_parse:
1918                 while (ch >= '0' && ch <= '9') {
1919                     if (q >= token_buf + STRING_MAX_SIZE)
1920                         goto num_too_long;
1921                     *q++ = ch;
1922                     ch = *p++;
1923                 }
1924             }
1925             if (ch == 'e' || ch == 'E') {
1926                 if (q >= token_buf + STRING_MAX_SIZE)
1927                     goto num_too_long;
1928                 *q++ = ch;
1929                 ch = *p++;
1930                 if (ch == '-' || ch == '+') {
1931                     if (q >= token_buf + STRING_MAX_SIZE)
1932                         goto num_too_long;
1933                     *q++ = ch;
1934                     ch = *p++;
1935                 }
1936                 if (ch < '0' || ch > '9')
1937                     expect("exponent digits");
1938                 while (ch >= '0' && ch <= '9') {
1939                     if (q >= token_buf + STRING_MAX_SIZE)
1940                         goto num_too_long;
1941                     *q++ = ch;
1942                     ch = *p++;
1943                 }
1944             }
1945             *q = '\0';
1946             t = toup(ch);
1947             errno = 0;
1948             if (t == 'F') {
1949                 ch = *p++;
1950                 tok = TOK_CFLOAT;
1951                 tokc.f = strtof(token_buf, NULL);
1952             } else if (t == 'L') {
1953                 ch = *p++;
1954 #ifdef TCC_TARGET_PE
1955                 tok = TOK_CDOUBLE;
1956                 tokc.d = strtod(token_buf, NULL);
1957 #else
1958                 tok = TOK_CLDOUBLE;
1959                 tokc.ld = strtold(token_buf, NULL);
1960 #endif
1961             } else {
1962                 tok = TOK_CDOUBLE;
1963                 tokc.d = strtod(token_buf, NULL);
1964             }
1965         }
1966     } else {
1967         unsigned long long n, n1;
1968         int lcount, ucount;
1969
1970         /* integer number */
1971         *q = '\0';
1972         q = token_buf;
1973         if (b == 10 && *q == '0') {
1974             b = 8;
1975             q++;
1976         }
1977         n = 0;
1978         while(1) {
1979             t = *q++;
1980             /* no need for checks except for base 10 / 8 errors */
1981             if (t == '\0') {
1982                 break;
1983             } else if (t >= 'a') {
1984                 t = t - 'a' + 10;
1985             } else if (t >= 'A') {
1986                 t = t - 'A' + 10;
1987             } else {
1988                 t = t - '0';
1989                 if (t >= b)
1990                     error("invalid digit");
1991             }
1992             n1 = n;
1993             n = n * b + t;
1994             /* detect overflow */
1995             /* XXX: this test is not reliable */
1996             if (n < n1)
1997                 error("integer constant overflow");
1998         }
1999
2000         /* XXX: not exactly ANSI compliant */
2001         if ((n & 0xffffffff00000000LL) != 0) {
2002             if ((n >> 63) != 0)
2003                 tok = TOK_CULLONG;
2004             else
2005                 tok = TOK_CLLONG;
2006         } else if (n > 0x7fffffff) {
2007             tok = TOK_CUINT;
2008         } else {
2009             tok = TOK_CINT;
2010         }
2011         lcount = 0;
2012         ucount = 0;
2013         for(;;) {
2014             t = toup(ch);
2015             if (t == 'L') {
2016                 if (lcount >= 2)
2017                     error("three 'l's in integer constant");
2018                 lcount++;
2019                 if (lcount == 2) {
2020                     if (tok == TOK_CINT)
2021                         tok = TOK_CLLONG;
2022                     else if (tok == TOK_CUINT)
2023                         tok = TOK_CULLONG;
2024                 }
2025                 ch = *p++;
2026             } else if (t == 'U') {
2027                 if (ucount >= 1)
2028                     error("two 'u's in integer constant");
2029                 ucount++;
2030                 if (tok == TOK_CINT)
2031                     tok = TOK_CUINT;
2032                 else if (tok == TOK_CLLONG)
2033                     tok = TOK_CULLONG;
2034                 ch = *p++;
2035             } else {
2036                 break;
2037             }
2038         }
2039         if (tok == TOK_CINT || tok == TOK_CUINT)
2040             tokc.ui = n;
2041         else
2042             tokc.ull = n;
2043     }
2044     if (ch)
2045         error("invalid number\n");
2046 }
2047
2048
2049 #define PARSE2(c1, tok1, c2, tok2)              \
2050     case c1:                                    \
2051         PEEKC(c, p);                            \
2052         if (c == c2) {                          \
2053             p++;                                \
2054             tok = tok2;                         \
2055         } else {                                \
2056             tok = tok1;                         \
2057         }                                       \
2058         break;
2059
2060 /* return next token without macro substitution */
2061 static inline void next_nomacro1(void)
2062 {
2063     int t, c, is_long;
2064     TokenSym *ts;
2065     uint8_t *p, *p1;
2066     unsigned int h;
2067
2068     p = file->buf_ptr;
2069  redo_no_start:
2070     c = *p;
2071     switch(c) {
2072     case ' ':
2073     case '\t':
2074         tok = c;
2075         p++;
2076         goto keep_tok_flags;
2077     case '\f':
2078     case '\v':
2079     case '\r':
2080         p++;
2081         goto redo_no_start;
2082     case '\\':
2083         /* first look if it is in fact an end of buffer */
2084         if (p >= file->buf_end) {
2085             file->buf_ptr = p;
2086             handle_eob();
2087             p = file->buf_ptr;
2088             if (p >= file->buf_end)
2089                 goto parse_eof;
2090             else
2091                 goto redo_no_start;
2092         } else {
2093             file->buf_ptr = p;
2094             ch = *p;
2095             handle_stray();
2096             p = file->buf_ptr;
2097             goto redo_no_start;
2098         }
2099     parse_eof:
2100         {
2101             TCCState *s1 = tcc_state;
2102             if ((parse_flags & PARSE_FLAG_LINEFEED)
2103                 && !(tok_flags & TOK_FLAG_EOF)) {
2104                 tok_flags |= TOK_FLAG_EOF;
2105                 tok = TOK_LINEFEED;
2106                 goto keep_tok_flags;
2107             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2108                 tok = TOK_EOF;
2109             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2110                 error("missing #endif");
2111             } else if (s1->include_stack_ptr == s1->include_stack) {
2112                 /* no include left : end of file. */
2113                 tok = TOK_EOF;
2114             } else {
2115                 tok_flags &= ~TOK_FLAG_EOF;
2116                 /* pop include file */
2117
2118                 /* test if previous '#endif' was after a #ifdef at
2119                    start of file */
2120                 if (tok_flags & TOK_FLAG_ENDIF) {
2121 #ifdef INC_DEBUG
2122                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2123 #endif
2124                     add_cached_include(s1, file->inc_type, file->inc_filename,
2125                                        file->ifndef_macro_saved);
2126                 }
2127
2128                 /* add end of include file debug info */
2129                 if (tcc_state->do_debug) {
2130                     put_stabd(N_EINCL, 0, 0);
2131                 }
2132                 /* pop include stack */
2133                 tcc_close();
2134                 s1->include_stack_ptr--;
2135                 p = file->buf_ptr;
2136                 goto redo_no_start;
2137             }
2138         }
2139         break;
2140
2141     case '\n':
2142         file->line_num++;
2143         tok_flags |= TOK_FLAG_BOL;
2144         p++;
2145 maybe_newline:
2146         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2147             goto redo_no_start;
2148         tok = TOK_LINEFEED;
2149         goto keep_tok_flags;
2150
2151     case '#':
2152         /* XXX: simplify */
2153         PEEKC(c, p);
2154         if ((tok_flags & TOK_FLAG_BOL) &&
2155             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2156             file->buf_ptr = p;
2157             preprocess(tok_flags & TOK_FLAG_BOF);
2158             p = file->buf_ptr;
2159             goto maybe_newline;
2160         } else {
2161             if (c == '#') {
2162                 p++;
2163                 tok = TOK_TWOSHARPS;
2164             } else {
2165                 if (parse_flags & PARSE_FLAG_ASM_COMMENTS) {
2166                     p = parse_line_comment(p - 1);
2167                     goto redo_no_start;
2168                 } else {
2169                     tok = '#';
2170                 }
2171             }
2172         }
2173         break;
2174
2175     case 'a': case 'b': case 'c': case 'd':
2176     case 'e': case 'f': case 'g': case 'h':
2177     case 'i': case 'j': case 'k': case 'l':
2178     case 'm': case 'n': case 'o': case 'p':
2179     case 'q': case 'r': case 's': case 't':
2180     case 'u': case 'v': case 'w': case 'x':
2181     case 'y': case 'z':
2182     case 'A': case 'B': case 'C': case 'D':
2183     case 'E': case 'F': case 'G': case 'H':
2184     case 'I': case 'J': case 'K':
2185     case 'M': case 'N': case 'O': case 'P':
2186     case 'Q': case 'R': case 'S': case 'T':
2187     case 'U': case 'V': case 'W': case 'X':
2188     case 'Y': case 'Z':
2189     case '_':
2190     parse_ident_fast:
2191         p1 = p;
2192         h = TOK_HASH_INIT;
2193         h = TOK_HASH_FUNC(h, c);
2194         p++;
2195         for(;;) {
2196             c = *p;
2197             if (!isidnum_table[c-CH_EOF])
2198                 break;
2199             h = TOK_HASH_FUNC(h, c);
2200             p++;
2201         }
2202         if (c != '\\') {
2203             TokenSym **pts;
2204             int len;
2205
2206             /* fast case : no stray found, so we have the full token
2207                and we have already hashed it */
2208             len = p - p1;
2209             h &= (TOK_HASH_SIZE - 1);
2210             pts = &hash_ident[h];
2211             for(;;) {
2212                 ts = *pts;
2213                 if (!ts)
2214                     break;
2215                 if (ts->len == len && !memcmp(ts->str, p1, len))
2216                     goto token_found;
2217                 pts = &(ts->hash_next);
2218             }
2219             ts = tok_alloc_new(pts, p1, len);
2220         token_found: ;
2221         } else {
2222             /* slower case */
2223             cstr_reset(&tokcstr);
2224
2225             while (p1 < p) {
2226                 cstr_ccat(&tokcstr, *p1);
2227                 p1++;
2228             }
2229             p--;
2230             PEEKC(c, p);
2231         parse_ident_slow:
2232             while (isidnum_table[c-CH_EOF]) {
2233                 cstr_ccat(&tokcstr, c);
2234                 PEEKC(c, p);
2235             }
2236             ts = tok_alloc(tokcstr.data, tokcstr.size);
2237         }
2238         tok = ts->tok;
2239         break;
2240     case 'L':
2241         t = p[1];
2242         if (t != '\\' && t != '\'' && t != '\"') {
2243             /* fast case */
2244             goto parse_ident_fast;
2245         } else {
2246             PEEKC(c, p);
2247             if (c == '\'' || c == '\"') {
2248                 is_long = 1;
2249                 goto str_const;
2250             } else {
2251                 cstr_reset(&tokcstr);
2252                 cstr_ccat(&tokcstr, 'L');
2253                 goto parse_ident_slow;
2254             }
2255         }
2256         break;
2257     case '0': case '1': case '2': case '3':
2258     case '4': case '5': case '6': case '7':
2259     case '8': case '9':
2260
2261         cstr_reset(&tokcstr);
2262         /* after the first digit, accept digits, alpha, '.' or sign if
2263            prefixed by 'eEpP' */
2264     parse_num:
2265         for(;;) {
2266             t = c;
2267             cstr_ccat(&tokcstr, c);
2268             PEEKC(c, p);
2269             if (!(isnum(c) || isid(c) || c == '.' ||
2270                   ((c == '+' || c == '-') &&
2271                    (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2272                 break;
2273         }
2274         /* We add a trailing '\0' to ease parsing */
2275         cstr_ccat(&tokcstr, '\0');
2276         tokc.cstr = &tokcstr;
2277         tok = TOK_PPNUM;
2278         break;
2279     case '.':
2280         /* special dot handling because it can also start a number */
2281         PEEKC(c, p);
2282         if (isnum(c)) {
2283             cstr_reset(&tokcstr);
2284             cstr_ccat(&tokcstr, '.');
2285             goto parse_num;
2286         } else if (c == '.') {
2287             PEEKC(c, p);
2288             if (c != '.')
2289                 expect("'.'");
2290             PEEKC(c, p);
2291             tok = TOK_DOTS;
2292         } else {
2293             tok = '.';
2294         }
2295         break;
2296     case '\'':
2297     case '\"':
2298         is_long = 0;
2299     str_const:
2300         {
2301             CString str;
2302             int sep;
2303
2304             sep = c;
2305
2306             /* parse the string */
2307             cstr_new(&str);
2308             p = parse_pp_string(p, sep, &str);
2309             cstr_ccat(&str, '\0');
2310
2311             /* eval the escape (should be done as TOK_PPNUM) */
2312             cstr_reset(&tokcstr);
2313             parse_escape_string(&tokcstr, str.data, is_long);
2314             cstr_free(&str);
2315
2316             if (sep == '\'') {
2317                 int char_size;
2318                 /* XXX: make it portable */
2319                 if (!is_long)
2320                     char_size = 1;
2321                 else
2322                     char_size = sizeof(nwchar_t);
2323                 if (tokcstr.size <= char_size)
2324                     error("empty character constant");
2325                 if (tokcstr.size > 2 * char_size)
2326                     warning("multi-character character constant");
2327                 if (!is_long) {
2328                     tokc.i = *(int8_t *)tokcstr.data;
2329                     tok = TOK_CCHAR;
2330                 } else {
2331                     tokc.i = *(nwchar_t *)tokcstr.data;
2332                     tok = TOK_LCHAR;
2333                 }
2334             } else {
2335                 tokc.cstr = &tokcstr;
2336                 if (!is_long)
2337                     tok = TOK_STR;
2338                 else
2339                     tok = TOK_LSTR;
2340             }
2341         }
2342         break;
2343
2344     case '<':
2345         PEEKC(c, p);
2346         if (c == '=') {
2347             p++;
2348             tok = TOK_LE;
2349         } else if (c == '<') {
2350             PEEKC(c, p);
2351             if (c == '=') {
2352                 p++;
2353                 tok = TOK_A_SHL;
2354             } else {
2355                 tok = TOK_SHL;
2356             }
2357         } else {
2358             tok = TOK_LT;
2359         }
2360         break;
2361
2362     case '>':
2363         PEEKC(c, p);
2364         if (c == '=') {
2365             p++;
2366             tok = TOK_GE;
2367         } else if (c == '>') {
2368             PEEKC(c, p);
2369             if (c == '=') {
2370                 p++;
2371                 tok = TOK_A_SAR;
2372             } else {
2373                 tok = TOK_SAR;
2374             }
2375         } else {
2376             tok = TOK_GT;
2377         }
2378         break;
2379
2380     case '&':
2381         PEEKC(c, p);
2382         if (c == '&') {
2383             p++;
2384             tok = TOK_LAND;
2385         } else if (c == '=') {
2386             p++;
2387             tok = TOK_A_AND;
2388         } else {
2389             tok = '&';
2390         }
2391         break;
2392
2393     case '|':
2394         PEEKC(c, p);
2395         if (c == '|') {
2396             p++;
2397             tok = TOK_LOR;
2398         } else if (c == '=') {
2399             p++;
2400             tok = TOK_A_OR;
2401         } else {
2402             tok = '|';
2403         }
2404         break;
2405
2406     case '+':
2407         PEEKC(c, p);
2408         if (c == '+') {
2409             p++;
2410             tok = TOK_INC;
2411         } else if (c == '=') {
2412             p++;
2413             tok = TOK_A_ADD;
2414         } else {
2415             tok = '+';
2416         }
2417         break;
2418
2419     case '-':
2420         PEEKC(c, p);
2421         if (c == '-') {
2422             p++;
2423             tok = TOK_DEC;
2424         } else if (c == '=') {
2425             p++;
2426             tok = TOK_A_SUB;
2427         } else if (c == '>') {
2428             p++;
2429             tok = TOK_ARROW;
2430         } else {
2431             tok = '-';
2432         }
2433         break;
2434
2435     PARSE2('!', '!', '=', TOK_NE)
2436     PARSE2('=', '=', '=', TOK_EQ)
2437     PARSE2('*', '*', '=', TOK_A_MUL)
2438     PARSE2('%', '%', '=', TOK_A_MOD)
2439     PARSE2('^', '^', '=', TOK_A_XOR)
2440
2441         /* comments or operator */
2442     case '/':
2443         PEEKC(c, p);
2444         if (c == '*') {
2445             p = parse_comment(p);
2446             goto redo_no_start;
2447         } else if (c == '/') {
2448             p = parse_line_comment(p);
2449             goto redo_no_start;
2450         } else if (c == '=') {
2451             p++;
2452             tok = TOK_A_DIV;
2453         } else {
2454             tok = '/';
2455         }
2456         break;
2457
2458         /* simple tokens */
2459     case '(':
2460     case ')':
2461     case '[':
2462     case ']':
2463     case '{':
2464     case '}':
2465     case ',':
2466     case ';':
2467     case ':':
2468     case '?':
2469     case '~':
2470     case '$': /* only used in assembler */
2471     case '@': /* dito */
2472         tok = c;
2473         p++;
2474         break;
2475     default:
2476         error("unrecognized character \\x%02x", c);
2477         break;
2478     }
2479     tok_flags = 0;
2480 keep_tok_flags:
2481     file->buf_ptr = p;
2482 #if defined(PARSE_DEBUG)
2483     printf("token = %s\n", get_tok_str(tok, &tokc));
2484 #endif
2485 }
2486
2487 /* return next token without macro substitution. Can read input from
2488    macro_ptr buffer */
2489 static void next_nomacro_spc(void)
2490 {
2491     if (macro_ptr) {
2492     redo:
2493         tok = *macro_ptr;
2494         if (tok) {
2495             TOK_GET(&tok, &macro_ptr, &tokc);
2496             if (tok == TOK_LINENUM) {
2497                 file->line_num = tokc.i;
2498                 goto redo;
2499             }
2500         }
2501     } else {
2502         next_nomacro1();
2503     }
2504 }
2505
2506 ST_FUNC void next_nomacro(void)
2507 {
2508     do {
2509         next_nomacro_spc();
2510     } while (is_space(tok));
2511 }
2512
2513 /* substitute args in macro_str and return allocated string */
2514 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2515 {
2516     int last_tok, t, spc;
2517     const int *st;
2518     Sym *s;
2519     CValue cval;
2520     TokenString str;
2521     CString cstr;
2522
2523     tok_str_new(&str);
2524     last_tok = 0;
2525     while(1) {
2526         TOK_GET(&t, &macro_str, &cval);
2527         if (!t)
2528             break;
2529         if (t == '#') {
2530             /* stringize */
2531             TOK_GET(&t, &macro_str, &cval);
2532             if (!t)
2533                 break;
2534             s = sym_find2(args, t);
2535             if (s) {
2536                 cstr_new(&cstr);
2537                 st = s->d;
2538                 spc = 0;
2539                 while (*st) {
2540                     TOK_GET(&t, &st, &cval);
2541                     if (!check_space(t, &spc))
2542                         cstr_cat(&cstr, get_tok_str(t, &cval));
2543                 }
2544                 cstr.size -= spc;
2545                 cstr_ccat(&cstr, '\0');
2546 #ifdef PP_DEBUG
2547                 printf("stringize: %s\n", (char *)cstr.data);
2548 #endif
2549                 /* add string */
2550                 cval.cstr = &cstr;
2551                 tok_str_add2(&str, TOK_STR, &cval);
2552                 cstr_free(&cstr);
2553             } else {
2554                 tok_str_add2(&str, t, &cval);
2555             }
2556         } else if (t >= TOK_IDENT) {
2557             s = sym_find2(args, t);
2558             if (s) {
2559                 st = s->d;
2560                 /* if '##' is present before or after, no arg substitution */
2561                 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2562                     /* special case for var arg macros : ## eats the
2563                        ',' if empty VA_ARGS variable. */
2564                     /* XXX: test of the ',' is not 100%
2565                        reliable. should fix it to avoid security
2566                        problems */
2567                     if (gnu_ext && s->type.t &&
2568                         last_tok == TOK_TWOSHARPS &&
2569                         str.len >= 2 && str.str[str.len - 2] == ',') {
2570                         if (*st == 0) {
2571                             /* suppress ',' '##' */
2572                             str.len -= 2;
2573                         } else {
2574                             /* suppress '##' and add variable */
2575                             str.len--;
2576                             goto add_var;
2577                         }
2578                     } else {
2579                         int t1;
2580                     add_var:
2581                         for(;;) {
2582                             TOK_GET(&t1, &st, &cval);
2583                             if (!t1)
2584                                 break;
2585                             tok_str_add2(&str, t1, &cval);
2586                         }
2587                     }
2588                 } else {
2589                     /* NOTE: the stream cannot be read when macro
2590                        substituing an argument */
2591                     macro_subst(&str, nested_list, st, NULL);
2592                 }
2593             } else {
2594                 tok_str_add(&str, t);
2595             }
2596         } else {
2597             tok_str_add2(&str, t, &cval);
2598         }
2599         last_tok = t;
2600     }
2601     tok_str_add(&str, 0);
2602     return str.str;
2603 }
2604
2605 static char const ab_month_name[12][4] =
2606 {
2607     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2608     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2609 };
2610
2611 /* do macro substitution of current token with macro 's' and add
2612    result to (tok_str,tok_len). 'nested_list' is the list of all
2613    macros we got inside to avoid recursing. Return non zero if no
2614    substitution needs to be done */
2615 static int macro_subst_tok(TokenString *tok_str,
2616                            Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2617 {
2618     Sym *args, *sa, *sa1;
2619     int mstr_allocated, parlevel, *mstr, t, t1, spc;
2620     const int *p;
2621     TokenString str;
2622     char *cstrval;
2623     CValue cval;
2624     CString cstr;
2625     char buf[32];
2626
2627     /* if symbol is a macro, prepare substitution */
2628     /* special macros */
2629     if (tok == TOK___LINE__) {
2630         snprintf(buf, sizeof(buf), "%d", file->line_num);
2631         cstrval = buf;
2632         t1 = TOK_PPNUM;
2633         goto add_cstr1;
2634     } else if (tok == TOK___FILE__) {
2635         cstrval = file->filename;
2636         goto add_cstr;
2637     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2638         time_t ti;
2639         struct tm *tm;
2640
2641         time(&ti);
2642         tm = localtime(&ti);
2643         if (tok == TOK___DATE__) {
2644             snprintf(buf, sizeof(buf), "%s %2d %d",
2645                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2646         } else {
2647             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2648                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2649         }
2650         cstrval = buf;
2651     add_cstr:
2652         t1 = TOK_STR;
2653     add_cstr1:
2654         cstr_new(&cstr);
2655         cstr_cat(&cstr, cstrval);
2656         cstr_ccat(&cstr, '\0');
2657         cval.cstr = &cstr;
2658         tok_str_add2(tok_str, t1, &cval);
2659         cstr_free(&cstr);
2660     } else {
2661         mstr = s->d;
2662         mstr_allocated = 0;
2663         if (s->type.t == MACRO_FUNC) {
2664             /* NOTE: we do not use next_nomacro to avoid eating the
2665                next token. XXX: find better solution */
2666         redo:
2667             if (macro_ptr) {
2668                 p = macro_ptr;
2669                 while (is_space(t = *p) || TOK_LINEFEED == t)
2670                     ++p;
2671                 if (t == 0 && can_read_stream) {
2672                     /* end of macro stream: we must look at the token
2673                        after in the file */
2674                     struct macro_level *ml = *can_read_stream;
2675                     macro_ptr = NULL;
2676                     if (ml)
2677                     {
2678                         macro_ptr = ml->p;
2679                         ml->p = NULL;
2680                         *can_read_stream = ml -> prev;
2681                     }
2682                     goto redo;
2683                 }
2684             } else {
2685                 /* XXX: incorrect with comments */
2686                 ch = file->buf_ptr[0];
2687                 while (is_space(ch) || ch == '\n')
2688                     cinp();
2689                 t = ch;
2690             }
2691             if (t != '(') /* no macro subst */
2692                 return -1;
2693
2694             /* argument macro */
2695             next_nomacro();
2696             next_nomacro();
2697             args = NULL;
2698             sa = s->next;
2699             /* NOTE: empty args are allowed, except if no args */
2700             for(;;) {
2701                 /* handle '()' case */
2702                 if (!args && !sa && tok == ')')
2703                     break;
2704                 if (!sa)
2705                     error("macro '%s' used with too many args",
2706                           get_tok_str(s->v, 0));
2707                 tok_str_new(&str);
2708                 parlevel = spc = 0;
2709                 /* NOTE: non zero sa->t indicates VA_ARGS */
2710                 while ((parlevel > 0 ||
2711                         (tok != ')' &&
2712                          (tok != ',' || sa->type.t))) &&
2713                        tok != -1) {
2714                     if (tok == '(')
2715                         parlevel++;
2716                     else if (tok == ')')
2717                         parlevel--;
2718                     if (tok == TOK_LINEFEED)
2719                         tok = ' ';
2720                     if (!check_space(tok, &spc))
2721                         tok_str_add2(&str, tok, &tokc);
2722                     next_nomacro_spc();
2723                 }
2724                 str.len -= spc;
2725                 tok_str_add(&str, 0);
2726                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
2727                 sa1->d = str.str;
2728                 sa = sa->next;
2729                 if (tok == ')') {
2730                     /* special case for gcc var args: add an empty
2731                        var arg argument if it is omitted */
2732                     if (sa && sa->type.t && gnu_ext)
2733                         continue;
2734                     else
2735                         break;
2736                 }
2737                 if (tok != ',')
2738                     expect(",");
2739                 next_nomacro();
2740             }
2741             if (sa) {
2742                 error("macro '%s' used with too few args",
2743                       get_tok_str(s->v, 0));
2744             }
2745
2746             /* now subst each arg */
2747             mstr = macro_arg_subst(nested_list, mstr, args);
2748             /* free memory */
2749             sa = args;
2750             while (sa) {
2751                 sa1 = sa->prev;
2752                 tok_str_free(sa->d);
2753                 sym_free(sa);
2754                 sa = sa1;
2755             }
2756             mstr_allocated = 1;
2757         }
2758         sym_push2(nested_list, s->v, 0, 0);
2759         macro_subst(tok_str, nested_list, mstr, can_read_stream);
2760         /* pop nested defined symbol */
2761         sa1 = *nested_list;
2762         *nested_list = sa1->prev;
2763         sym_free(sa1);
2764         if (mstr_allocated)
2765             tok_str_free(mstr);
2766     }
2767     return 0;
2768 }
2769
2770 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2771    return the resulting string (which must be freed). */
2772 static inline int *macro_twosharps(const int *macro_str)
2773 {
2774     const int *ptr;
2775     int t;
2776     CValue cval;
2777     TokenString macro_str1;
2778     CString cstr;
2779     int n;
2780
2781     /* we search the first '##' */
2782     for(ptr = macro_str;;) {
2783         TOK_GET(&t, &ptr, &cval);
2784         if (t == TOK_TWOSHARPS)
2785             break;
2786         /* nothing more to do if end of string */
2787         if (t == 0)
2788             return NULL;
2789     }
2790
2791     /* we saw '##', so we need more processing to handle it */
2792     tok_str_new(&macro_str1);
2793     for(ptr = macro_str;;) {
2794         TOK_GET(&tok, &ptr, &tokc);
2795         if (tok == 0)
2796             break;
2797         if (tok == TOK_TWOSHARPS)
2798             continue;
2799         while (*ptr == TOK_TWOSHARPS) {
2800             do { t = *++ptr; } while (t == TOK_NOSUBST);
2801
2802             if (t && t != TOK_TWOSHARPS) {
2803                 TOK_GET(&t, &ptr, &cval);
2804
2805                 /* We concatenate the two tokens */
2806                 cstr_new(&cstr);
2807                 cstr_cat(&cstr, get_tok_str(tok, &tokc));
2808                 n = cstr.size;
2809                 cstr_cat(&cstr, get_tok_str(t, &cval));
2810                 cstr_ccat(&cstr, '\0');
2811
2812                 tcc_open_bf(tcc_state, "<paste>", cstr.size);
2813                 memcpy(file->buffer, cstr.data, cstr.size);
2814                 for (;;) {
2815                     next_nomacro1();
2816                     if (0 == *file->buf_ptr)
2817                         break;
2818                     tok_str_add2(&macro_str1, tok, &tokc);
2819                     warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
2820                         n, cstr.data, (char*)cstr.data + n);
2821                 }
2822                 tcc_close();
2823                 cstr_reset(&cstr);
2824             }
2825         }
2826         tok_str_add2(&macro_str1, tok, &tokc);
2827     }
2828     tok_str_add(&macro_str1, 0);
2829     return macro_str1.str;
2830 }
2831
2832
2833 /* do macro substitution of macro_str and add result to
2834    (tok_str,tok_len). 'nested_list' is the list of all macros we got
2835    inside to avoid recursing. */
2836 static void macro_subst(TokenString *tok_str, Sym **nested_list,
2837                         const int *macro_str, struct macro_level ** can_read_stream)
2838 {
2839     Sym *s;
2840     int *macro_str1;
2841     const int *ptr;
2842     int t, ret, spc;
2843     CValue cval;
2844     struct macro_level ml;
2845
2846     /* first scan for '##' operator handling */
2847     ptr = macro_str;
2848     macro_str1 = macro_twosharps(ptr);
2849     if (macro_str1)
2850         ptr = macro_str1;
2851     spc = 0;
2852     while (1) {
2853         /* NOTE: ptr == NULL can only happen if tokens are read from
2854            file stream due to a macro function call */
2855         if (ptr == NULL)
2856             break;
2857         TOK_GET(&t, &ptr, &cval);
2858         if (t == 0)
2859             break;
2860         if (t == TOK_NOSUBST) {
2861             /* following token has already been subst'd. just copy it on */
2862             tok_str_add2(tok_str, TOK_NOSUBST, NULL);
2863             TOK_GET(&t, &ptr, &cval);
2864             goto no_subst;
2865         }
2866         s = define_find(t);
2867         if (s != NULL) {
2868             /* if nested substitution, do nothing */
2869             if (sym_find2(*nested_list, t)) {
2870                 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
2871                 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
2872                 goto no_subst;
2873             }
2874             ml.p = macro_ptr;
2875             if (can_read_stream)
2876                 ml.prev = *can_read_stream, *can_read_stream = &ml;
2877             macro_ptr = (int *)ptr;
2878             tok = t;
2879             ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
2880             ptr = (int *)macro_ptr;
2881             macro_ptr = ml.p;
2882             if (can_read_stream && *can_read_stream == &ml)
2883                 *can_read_stream = ml.prev;
2884             if (ret != 0)
2885                 goto no_subst;
2886         } else {
2887         no_subst:
2888             if (!check_space(t, &spc))
2889                 tok_str_add2(tok_str, t, &cval);
2890         }
2891     }
2892     if (macro_str1)
2893         tok_str_free(macro_str1);
2894 }
2895
2896 /* return next token with macro substitution */
2897 ST_FUNC void next(void)
2898 {
2899     Sym *nested_list, *s;
2900     TokenString str;
2901     struct macro_level *ml;
2902
2903  redo:
2904     if (parse_flags & PARSE_FLAG_SPACES)
2905         next_nomacro_spc();
2906     else
2907         next_nomacro();
2908     if (!macro_ptr) {
2909         /* if not reading from macro substituted string, then try
2910            to substitute macros */
2911         if (tok >= TOK_IDENT &&
2912             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2913             s = define_find(tok);
2914             if (s) {
2915                 /* we have a macro: we try to substitute */
2916                 tok_str_new(&str);
2917                 nested_list = NULL;
2918                 ml = NULL;
2919                 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
2920                     /* substitution done, NOTE: maybe empty */
2921                     tok_str_add(&str, 0);
2922                     macro_ptr = str.str;
2923                     macro_ptr_allocated = str.str;
2924                     goto redo;
2925                 }
2926             }
2927         }
2928     } else {
2929         if (tok == 0) {
2930             /* end of macro or end of unget buffer */
2931             if (unget_buffer_enabled) {
2932                 macro_ptr = unget_saved_macro_ptr;
2933                 unget_buffer_enabled = 0;
2934             } else {
2935                 /* end of macro string: free it */
2936                 tok_str_free(macro_ptr_allocated);
2937                 macro_ptr_allocated = NULL;
2938                 macro_ptr = NULL;
2939             }
2940             goto redo;
2941         } else if (tok == TOK_NOSUBST) {
2942             /* discard preprocessor's nosubst markers */
2943             goto redo;
2944         }
2945     }
2946
2947     /* convert preprocessor tokens into C tokens */
2948     if (tok == TOK_PPNUM &&
2949         (parse_flags & PARSE_FLAG_TOK_NUM)) {
2950         parse_number((char *)tokc.cstr->data);
2951     }
2952 }
2953
2954 /* push back current token and set current token to 'last_tok'. Only
2955    identifier case handled for labels. */
2956 ST_INLN void unget_tok(int last_tok)
2957 {
2958     int i, n;
2959     int *q;
2960     unget_saved_macro_ptr = macro_ptr;
2961     unget_buffer_enabled = 1;
2962     q = unget_saved_buffer;
2963     macro_ptr = q;
2964     *q++ = tok;
2965     n = tok_ext_size(tok) - 1;
2966     for(i=0;i<n;i++)
2967         *q++ = tokc.tab[i];
2968     *q = 0; /* end of token string */
2969     tok = last_tok;
2970 }
2971
2972
2973 /* better than nothing, but needs extension to handle '-E' option
2974    correctly too */
2975 ST_FUNC void preprocess_init(TCCState *s1)
2976 {
2977     s1->include_stack_ptr = s1->include_stack;
2978     /* XXX: move that before to avoid having to initialize
2979        file->ifdef_stack_ptr ? */
2980     s1->ifdef_stack_ptr = s1->ifdef_stack;
2981     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
2982
2983     /* XXX: not ANSI compliant: bound checking says error */
2984     vtop = vstack - 1;
2985     s1->pack_stack[0] = 0;
2986     s1->pack_stack_ptr = s1->pack_stack;
2987 }
2988
2989 ST_FUNC void preprocess_new()
2990 {
2991     int i, c;
2992     const char *p, *r;
2993     TokenSym *ts;
2994
2995     /* init isid table */
2996     for(i=CH_EOF;i<256;i++)
2997         isidnum_table[i-CH_EOF] = isid(i) || isnum(i);
2998
2999     /* add all tokens */
3000     table_ident = NULL;
3001     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3002
3003     tok_ident = TOK_IDENT;
3004     p = tcc_keywords;
3005     while (*p) {
3006         r = p;
3007         for(;;) {
3008             c = *r++;
3009             if (c == '\0')
3010                 break;
3011         }
3012         ts = tok_alloc(p, r - p - 1);
3013         p = r;
3014     }
3015 }
3016
3017 /* Preprocess the current file */
3018 ST_FUNC int tcc_preprocess(TCCState *s1)
3019 {
3020     Sym *define_start;
3021
3022     BufferedFile *file_ref, **iptr, **iptr_new;
3023     int token_seen, line_ref, d;
3024     const char *s;
3025
3026     preprocess_init(s1);
3027     define_start = define_stack;
3028     ch = file->buf_ptr[0];
3029     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3030     parse_flags = PARSE_FLAG_ASM_COMMENTS | PARSE_FLAG_PREPROCESS |
3031         PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
3032     token_seen = 0;
3033     line_ref = 0;
3034     file_ref = NULL;
3035     iptr = s1->include_stack_ptr;
3036
3037     for (;;) {
3038         next();
3039         if (tok == TOK_EOF) {
3040             break;
3041         } else if (file != file_ref) {
3042             goto print_line;
3043         } else if (tok == TOK_LINEFEED) {
3044             if (!token_seen)
3045                 continue;
3046             ++line_ref;
3047             token_seen = 0;
3048         } else if (!token_seen) {
3049             d = file->line_num - line_ref;
3050             if (file != file_ref || d < 0 || d >= 8) {
3051 print_line:
3052                 iptr_new = s1->include_stack_ptr;
3053                 s = iptr_new > iptr ? " 1"
3054                   : iptr_new < iptr ? " 2"
3055                   : iptr_new > s1->include_stack ? " 3"
3056                   : ""
3057                   ;
3058                 iptr = iptr_new;
3059                 fprintf(s1->outfile, "# %d \"%s\"%s\n", file->line_num, file->filename, s);
3060             } else {
3061                 while (d)
3062                     fputs("\n", s1->outfile), --d;
3063             }
3064             line_ref = (file_ref = file)->line_num;
3065             token_seen = tok != TOK_LINEFEED;
3066             if (!token_seen)
3067                 continue;
3068         }
3069         fputs(get_tok_str(tok, &tokc), s1->outfile);
3070     }
3071     free_defines(define_start);
3072     return 0;
3073 }