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 #define USING_GLOBALS
  22 #include "tcc.h"
  23
  24 /********************************************************/
  25 /* global variables */
  26
  27 ST_DATA int tok_flags;
  28 ST_DATA int parse_flags;
  29
  30 ST_DATA struct BufferedFile *file;
  31 ST_DATA int ch, tok;
  32 ST_DATA CValue tokc;
  33 ST_DATA const int *macro_ptr;
  34 ST_DATA CString tokcstr; /* current parsed string, if any */
  35
  36 /* display benchmark infos */
  37 ST_DATA int tok_ident;
  38 ST_DATA TokenSym **table_ident;
  39
  40 /* ------------------------------------------------------------------------- */
  41
  42 static TokenSym *hash_ident[TOK_HASH_SIZE];
  43 static char token_buf[STRING_MAX_SIZE + 1];
  44 static CString cstr_buf;
  45 static CString macro_equal_buf;
  46 static TokenString tokstr_buf;
  47 static unsigned char isidnum_table[256 - CH_EOF];
  48 static int pp_debug_tok, pp_debug_symv;
  49 static int pp_once;
  50 static int pp_expr;
  51 static int pp_counter;
  52 static void tok_print(const char *msg, const int *str);
  53
  54 static struct TinyAlloc *toksym_alloc;
  55 static struct TinyAlloc *tokstr_alloc;
  56
  57 static TokenString *macro_stack;
  58
  59 static const char tcc_keywords[] =
  60 #define DEF(id, str) str "\0"
  61 #include "tcctok.h"
  62 #undef DEF
  63 ;
  64
  65 /* WARNING: the content of this string encodes token numbers */
  66 static const unsigned char tok_two_chars[] =
  67 /* outdated -- gr
  68     "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
  69     "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
  70 */{
  71     '<','=', TOK_LE,
  72     '>','=', TOK_GE,
  73     '!','=', TOK_NE,
  74     '&','&', TOK_LAND,
  75     '|','|', TOK_LOR,
  76     '+','+', TOK_INC,
  77     '-','-', TOK_DEC,
  78     '=','=', TOK_EQ,
  79     '<','<', TOK_SHL,
  80     '>','>', TOK_SAR,
  81     '+','=', TOK_A_ADD,
  82     '-','=', TOK_A_SUB,
  83     '*','=', TOK_A_MUL,
  84     '/','=', TOK_A_DIV,
  85     '%','=', TOK_A_MOD,
  86     '&','=', TOK_A_AND,
  87     '^','=', TOK_A_XOR,
  88     '|','=', TOK_A_OR,
  89     '-','>', TOK_ARROW,
  90     '.','.', TOK_TWODOTS,
  91     '#','#', TOK_TWOSHARPS,
  92     '#','#', TOK_PPJOIN,
  93     0
  94 };
  95
  96 static void next_nomacro(void);
  97
  98 ST_FUNC void skip(int c)
  99 {
 100     if (tok != c)
 101         tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
 102     next();
 103 }
 104
 105 ST_FUNC void expect(const char *msg)
 106 {
 107     tcc_error("%s expected", msg);
 108 }
 109
 110 /* ------------------------------------------------------------------------- */
 111 /* Custom allocator for tiny objects */
 112
 113 #define USE_TAL
 114
 115 #ifndef USE_TAL
 116 #define tal_free(al, p) tcc_free(p)
 117 #define tal_realloc(al, p, size) tcc_realloc(p, size)
 118 #define tal_new(a,b,c)
 119 #define tal_delete(a)
 120 #else
 121 #if !defined(MEM_DEBUG)
 122 #define tal_free(al, p) tal_free_impl(al, p)
 123 #define tal_realloc(al, p, size) tal_realloc_impl(&al, p, size)
 124 #define TAL_DEBUG_PARAMS
 125 #else
 126 #define TAL_DEBUG 1
 127 //#define TAL_INFO 1 /* collect and dump allocators stats */
 128 #define tal_free(al, p) tal_free_impl(al, p, __FILE__, __LINE__)
 129 #define tal_realloc(al, p, size) tal_realloc_impl(&al, p, size, __FILE__, __LINE__)
 130 #define TAL_DEBUG_PARAMS , const char *file, int line
 131 #define TAL_DEBUG_FILE_LEN 40
 132 #endif
 133
 134 #define TOKSYM_TAL_SIZE     (768 * 1024) /* allocator for tiny TokenSym in table_ident */
 135 #define TOKSTR_TAL_SIZE     (768 * 1024) /* allocator for tiny TokenString instances */
 136 #define CSTR_TAL_SIZE       (256 * 1024) /* allocator for tiny CString instances */
 137 #define TOKSYM_TAL_LIMIT    256 /* prefer unique limits to distinguish allocators debug msgs */
 138 #define TOKSTR_TAL_LIMIT    128 /* 32 * sizeof(int) */
 139 #define CSTR_TAL_LIMIT      1024
 140
 141 typedef struct TinyAlloc {
 142     unsigned  limit;
 143     unsigned  size;
 144     uint8_t *buffer;
 145     uint8_t *p;
 146     unsigned  nb_allocs;
 147     struct TinyAlloc *next, *top;
 148 #ifdef TAL_INFO
 149     unsigned  nb_peak;
 150     unsigned  nb_total;
 151     unsigned  nb_missed;
 152     uint8_t *peak_p;
 153 #endif
 154 } TinyAlloc;
 155
 156 typedef struct tal_header_t {
 157     unsigned  size;
 158 #ifdef TAL_DEBUG
 159     int     line_num; /* negative line_num used for double free check */
 160     char    file_name[TAL_DEBUG_FILE_LEN + 1];
 161 #endif
 162 } tal_header_t;
 163
 164 /* ------------------------------------------------------------------------- */
 165
 166 static TinyAlloc *tal_new(TinyAlloc **pal, unsigned limit, unsigned size)
 167 {
 168     TinyAlloc *al = tcc_mallocz(sizeof(TinyAlloc));
 169     al->p = al->buffer = tcc_malloc(size);
 170     al->limit = limit;
 171     al->size = size;
 172     if (pal) *pal = al;
 173     return al;
 174 }
 175
 176 static void tal_delete(TinyAlloc *al)
 177 {
 178     TinyAlloc *next;
 179
 180 tail_call:
 181     if (!al)
 182         return;
 183 #ifdef TAL_INFO
 184     fprintf(stderr, "limit=%5d, size=%5g MB, nb_peak=%6d, nb_total=%8d, nb_missed=%6d, usage=%5.1f%%\n",
 185             al->limit, al->size / 1024.0 / 1024.0, al->nb_peak, al->nb_total, al->nb_missed,
 186             (al->peak_p - al->buffer) * 100.0 / al->size);
 187 #endif
 188 #ifdef TAL_DEBUG
 189     if (al->nb_allocs > 0) {
 190         uint8_t *p;
 191         fprintf(stderr, "TAL_DEBUG: memory leak %d chunk(s) (limit= %d)\n",
 192                 al->nb_allocs, al->limit);
 193         p = al->buffer;
 194         while (p < al->p) {
 195             tal_header_t *header = (tal_header_t *)p;
 196             if (header->line_num > 0) {
 197                 fprintf(stderr, "%s:%d: chunk of %d bytes leaked\n",
 198                         header->file_name, header->line_num, header->size);
 199             }
 200             p += header->size + sizeof(tal_header_t);
 201         }
 202 #if MEM_DEBUG-0 == 2
 203         exit(2);
 204 #endif
 205     }
 206 #endif
 207     next = al->next;
 208     tcc_free(al->buffer);
 209     tcc_free(al);
 210     al = next;
 211     goto tail_call;
 212 }
 213
 214 static void tal_free_impl(TinyAlloc *al, void *p TAL_DEBUG_PARAMS)
 215 {
 216     if (!p)
 217         return;
 218 tail_call:
 219     if (al->buffer <= (uint8_t *)p && (uint8_t *)p < al->buffer + al->size) {
 220 #ifdef TAL_DEBUG
 221         tal_header_t *header = (((tal_header_t *)p) - 1);
 222         if (header->line_num < 0) {
 223             fprintf(stderr, "%s:%d: TAL_DEBUG: double frees chunk from\n",
 224                     file, line);
 225             fprintf(stderr, "%s:%d: %d bytes\n",
 226                     header->file_name, (int)-header->line_num, (int)header->size);
 227         } else
 228             header->line_num = -header->line_num;
 229 #endif
 230         al->nb_allocs--;
 231         if (!al->nb_allocs)
 232             al->p = al->buffer;
 233     } else if (al->next) {
 234         al = al->next;
 235         goto tail_call;
 236     }
 237     else
 238         tcc_free(p);
 239 }
 240
 241 static void *tal_realloc_impl(TinyAlloc **pal, void *p, unsigned size TAL_DEBUG_PARAMS)
 242 {
 243     tal_header_t *header;
 244     void *ret;
 245     int is_own;
 246     unsigned adj_size = (size + 3) & -4;
 247     TinyAlloc *al = *pal;
 248
 249 tail_call:
 250     is_own = (al->buffer <= (uint8_t *)p && (uint8_t *)p < al->buffer + al->size);
 251     if ((!p || is_own) && size <= al->limit) {
 252         if (al->p - al->buffer + adj_size + sizeof(tal_header_t) < al->size) {
 253             header = (tal_header_t *)al->p;
 254             header->size = adj_size;
 255 #ifdef TAL_DEBUG
 256             { int ofs = strlen(file) - TAL_DEBUG_FILE_LEN;
 257             strncpy(header->file_name, file + (ofs > 0 ? ofs : 0), TAL_DEBUG_FILE_LEN);
 258             header->file_name[TAL_DEBUG_FILE_LEN] = 0;
 259             header->line_num = line; }
 260 #endif
 261             ret = al->p + sizeof(tal_header_t);
 262             al->p += adj_size + sizeof(tal_header_t);
 263             if (is_own) {
 264                 header = (((tal_header_t *)p) - 1);
 265                 if (p) memcpy(ret, p, header->size);
 266 #ifdef TAL_DEBUG
 267                 header->line_num = -header->line_num;
 268 #endif
 269             } else {
 270                 al->nb_allocs++;
 271             }
 272 #ifdef TAL_INFO
 273             if (al->nb_peak < al->nb_allocs)
 274                 al->nb_peak = al->nb_allocs;
 275             if (al->peak_p < al->p)
 276                 al->peak_p = al->p;
 277             al->nb_total++;
 278 #endif
 279             return ret;
 280         } else if (is_own) {
 281             al->nb_allocs--;
 282             ret = tal_realloc(*pal, 0, size);
 283             header = (((tal_header_t *)p) - 1);
 284             if (p) memcpy(ret, p, header->size);
 285 #ifdef TAL_DEBUG
 286             header->line_num = -header->line_num;
 287 #endif
 288             return ret;
 289         }
 290         if (al->next) {
 291             al = al->next;
 292         } else {
 293             TinyAlloc *bottom = al, *next = al->top ? al->top : al;
 294
 295             al = tal_new(pal, next->limit, next->size * 2);
 296             al->next = next;
 297             bottom->top = al;
 298         }
 299         goto tail_call;
 300     }
 301     if (is_own) {
 302         al->nb_allocs--;
 303         ret = tcc_malloc(size);
 304         header = (((tal_header_t *)p) - 1);
 305         if (p) memcpy(ret, p, header->size);
 306 #ifdef TAL_DEBUG
 307         header->line_num = -header->line_num;
 308 #endif
 309     } else if (al->next) {
 310         al = al->next;
 311         goto tail_call;
 312     } else
 313         ret = tcc_realloc(p, size);
 314 #ifdef TAL_INFO
 315     al->nb_missed++;
 316 #endif
 317     return ret;
 318 }
 319
 320 #endif /* USE_TAL */
 321
 322 /* ------------------------------------------------------------------------- */
 323 /* CString handling */
 324 static void cstr_realloc(CString *cstr, int new_size)
 325 {
 326     int size;
 327
 328     size = cstr->size_allocated;
 329     if (size < 8)
 330         size = 8; /* no need to allocate a too small first string */
 331     while (size < new_size)
 332         size = size * 2;
 333     cstr->data = tcc_realloc(cstr->data, size);
 334     cstr->size_allocated = size;
 335 }
 336
 337 /* add a byte */
 338 ST_INLN void cstr_ccat(CString *cstr, int ch)
 339 {
 340     int size;
 341     size = cstr->size + 1;
 342     if (size > cstr->size_allocated)
 343         cstr_realloc(cstr, size);
 344     ((unsigned char *)cstr->data)[size - 1] = ch;
 345     cstr->size = size;
 346 }
 347
 348 ST_INLN char *unicode_to_utf8 (char *b, uint32_t Uc)
 349 {
 350     if (Uc<0x80) *b++=Uc;
 351     else if (Uc<0x800) *b++=192+Uc/64, *b++=128+Uc%64;
 352     else if (Uc-0xd800u<0x800) return b;
 353     else if (Uc<0x10000) *b++=224+Uc/4096, *b++=128+Uc/64%64, *b++=128+Uc%64;
 354     else if (Uc<0x110000) *b++=240+Uc/262144, *b++=128+Uc/4096%64, *b++=128+Uc/64%64, *b++=128+Uc%64;
 355     return b;
 356 }
 357
 358 /* add a unicode character expanded into utf8 */
 359 ST_INLN void cstr_u8cat(CString *cstr, int ch)
 360 {
 361     char buf[4], *e;
 362     e = unicode_to_utf8(buf, (uint32_t)ch);
 363     cstr_cat(cstr, buf, e - buf);
 364 }
 365
 366 ST_FUNC void cstr_cat(CString *cstr, const char *str, int len)
 367 {
 368     int size;
 369     if (len <= 0)
 370         len = strlen(str) + 1 + len;
 371     size = cstr->size + len;
 372     if (size > cstr->size_allocated)
 373         cstr_realloc(cstr, size);
 374     memmove(((unsigned char *)cstr->data) + cstr->size, str, len);
 375     cstr->size = size;
 376 }
 377
 378 /* add a wide char */
 379 ST_FUNC void cstr_wccat(CString *cstr, int ch)
 380 {
 381     int size;
 382     size = cstr->size + sizeof(nwchar_t);
 383     if (size > cstr->size_allocated)
 384         cstr_realloc(cstr, size);
 385     *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
 386     cstr->size = size;
 387 }
 388
 389 ST_FUNC void cstr_new(CString *cstr)
 390 {
 391     memset(cstr, 0, sizeof(CString));
 392 }
 393
 394 /* free string and reset it to NULL */
 395 ST_FUNC void cstr_free(CString *cstr)
 396 {
 397     tcc_free(cstr->data);
 398     cstr_new(cstr);
 399 }
 400
 401 /* reset string to empty */
 402 ST_FUNC void cstr_reset(CString *cstr)
 403 {
 404     cstr->size = 0;
 405 }
 406
 407 ST_FUNC int cstr_vprintf(CString *cstr, const char *fmt, va_list ap)
 408 {
 409     va_list v;
 410     int len, size = 80;
 411     for (;;) {
 412         size += cstr->size;
 413         if (size > cstr->size_allocated)
 414             cstr_realloc(cstr, size);
 415         size = cstr->size_allocated - cstr->size;
 416         va_copy(v, ap);
 417         len = vsnprintf((char*)cstr->data + cstr->size, size, fmt, v);
 418         va_end(v);
 419         if (len > 0 && len < size)
 420             break;
 421         size *= 2;
 422     }
 423     cstr->size += len;
 424     return len;
 425 }
 426
 427 ST_FUNC int cstr_printf(CString *cstr, const char *fmt, ...)
 428 {
 429     va_list ap; int len;
 430     va_start(ap, fmt);
 431     len = cstr_vprintf(cstr, fmt, ap);
 432     va_end(ap);
 433     return len;
 434 }
 435
 436 /* XXX: unicode ? */
 437 static void add_char(CString *cstr, int c)
 438 {
 439     if (c == '\'' || c == '\"' || c == '\\') {
 440         /* XXX: could be more precise if char or string */
 441         cstr_ccat(cstr, '\\');
 442     }
 443     if (c >= 32 && c <= 126) {
 444         cstr_ccat(cstr, c);
 445     } else {
 446         cstr_ccat(cstr, '\\');
 447         if (c == '\n') {
 448             cstr_ccat(cstr, 'n');
 449         } else {
 450             cstr_ccat(cstr, '0' + ((c >> 6) & 7));
 451             cstr_ccat(cstr, '0' + ((c >> 3) & 7));
 452             cstr_ccat(cstr, '0' + (c & 7));
 453         }
 454     }
 455 }
 456
 457 /* ------------------------------------------------------------------------- */
 458 /* allocate a new token */
 459 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
 460 {
 461     TokenSym *ts, **ptable;
 462     int i;
 463
 464     if (tok_ident >= SYM_FIRST_ANOM)
 465         tcc_error("memory full (symbols)");
 466
 467     /* expand token table if needed */
 468     i = tok_ident - TOK_IDENT;
 469     if ((i % TOK_ALLOC_INCR) == 0) {
 470         ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
 471         table_ident = ptable;
 472     }
 473
 474     ts = tal_realloc(toksym_alloc, 0, sizeof(TokenSym) + len);
 475     table_ident[i] = ts;
 476     ts->tok = tok_ident++;
 477     ts->sym_define = NULL;
 478     ts->sym_label = NULL;
 479     ts->sym_struct = NULL;
 480     ts->sym_identifier = NULL;
 481     ts->len = len;
 482     ts->hash_next = NULL;
 483     memcpy(ts->str, str, len);
 484     ts->str[len] = '\0';
 485     *pts = ts;
 486     return ts;
 487 }
 488
 489 #define TOK_HASH_INIT 1
 490 #define TOK_HASH_FUNC(h, c) ((h) + ((h) << 5) + ((h) >> 27) + (c))
 491
 492
 493 /* find a token and add it if not found */
 494 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
 495 {
 496     TokenSym *ts, **pts;
 497     int i;
 498     unsigned int h;
 499
 500     h = TOK_HASH_INIT;
 501     for(i=0;i<len;i++)
 502         h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
 503     h &= (TOK_HASH_SIZE - 1);
 504
 505     pts = &hash_ident[h];
 506     for(;;) {
 507         ts = *pts;
 508         if (!ts)
 509             break;
 510         if (ts->len == len && !memcmp(ts->str, str, len))
 511             return ts;
 512         pts = &(ts->hash_next);
 513     }
 514     return tok_alloc_new(pts, str, len);
 515 }
 516
 517 ST_FUNC int tok_alloc_const(const char *str)
 518 {
 519     return tok_alloc(str, strlen(str))->tok;
 520 }
 521
 522
 523 /* XXX: buffer overflow */
 524 /* XXX: float tokens */
 525 ST_FUNC const char *get_tok_str(int v, CValue *cv)
 526 {
 527     char *p;
 528     int i, len;
 529
 530     cstr_reset(&cstr_buf);
 531     p = cstr_buf.data;
 532
 533     switch(v) {
 534     case TOK_CINT:
 535     case TOK_CUINT:
 536     case TOK_CLONG:
 537     case TOK_CULONG:
 538     case TOK_CLLONG:
 539     case TOK_CULLONG:
 540         /* XXX: not quite exact, but only useful for testing  */
 541 #ifdef _WIN32
 542         sprintf(p, "%u", (unsigned)cv->i);
 543 #else
 544         sprintf(p, "%llu", (unsigned long long)cv->i);
 545 #endif
 546         break;
 547     case TOK_LCHAR:
 548         cstr_ccat(&cstr_buf, 'L');
 549     case TOK_CCHAR:
 550         cstr_ccat(&cstr_buf, '\'');
 551         add_char(&cstr_buf, cv->i);
 552         cstr_ccat(&cstr_buf, '\'');
 553         cstr_ccat(&cstr_buf, '\0');
 554         break;
 555     case TOK_PPNUM:
 556     case TOK_PPSTR:
 557         return (char*)cv->str.data;
 558     case TOK_LSTR:
 559         cstr_ccat(&cstr_buf, 'L');
 560     case TOK_STR:
 561         cstr_ccat(&cstr_buf, '\"');
 562         if (v == TOK_STR) {
 563             len = cv->str.size - 1;
 564             for(i=0;i<len;i++)
 565                 add_char(&cstr_buf, ((unsigned char *)cv->str.data)[i]);
 566         } else {
 567             len = (cv->str.size / sizeof(nwchar_t)) - 1;
 568             for(i=0;i<len;i++)
 569                 add_char(&cstr_buf, ((nwchar_t *)cv->str.data)[i]);
 570         }
 571         cstr_ccat(&cstr_buf, '\"');
 572         cstr_ccat(&cstr_buf, '\0');
 573         break;
 574
 575     case TOK_CFLOAT:
 576         cstr_cat(&cstr_buf, "<float>", 0);
 577         break;
 578     case TOK_CDOUBLE:
 579         cstr_cat(&cstr_buf, "<double>", 0);
 580         break;
 581     case TOK_CLDOUBLE:
 582         cstr_cat(&cstr_buf, "<long double>", 0);
 583         break;
 584     case TOK_LINENUM:
 585         cstr_cat(&cstr_buf, "<linenumber>", 0);
 586         break;
 587
 588     /* above tokens have value, the ones below don't */
 589     case TOK_LT:
 590         v = '<';
 591         goto addv;
 592     case TOK_GT:
 593         v = '>';
 594         goto addv;
 595     case TOK_DOTS:
 596         return strcpy(p, "...");
 597     case TOK_A_SHL:
 598         return strcpy(p, "<<=");
 599     case TOK_A_SAR:
 600         return strcpy(p, ">>=");
 601     case TOK_EOF:
 602         return strcpy(p, "<eof>");
 603     default:
 604         if (v < TOK_IDENT) {
 605             /* search in two bytes table */
 606             const unsigned char *q = tok_two_chars;
 607             while (*q) {
 608                 if (q[2] == v) {
 609                     *p++ = q[0];
 610                     *p++ = q[1];
 611                     *p = '\0';
 612                     return cstr_buf.data;
 613                 }
 614                 q += 3;
 615             }
 616         if (v >= 127) {
 617             sprintf(cstr_buf.data, "<%02x>", v);
 618             return cstr_buf.data;
 619         }
 620         addv:
 621             *p++ = v;
 622     case 0: /* nameless anonymous symbol */
 623             *p = '\0';
 624         } else if (v < tok_ident) {
 625             return table_ident[v - TOK_IDENT]->str;
 626         } else if (v >= SYM_FIRST_ANOM) {
 627             /* special name for anonymous symbol */
 628             sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
 629         } else {
 630             /* should never happen */
 631             return NULL;
 632         }
 633         break;
 634     }
 635     return cstr_buf.data;
 636 }
 637
 638 /* return the current character, handling end of block if necessary
 639    (but not stray) */
 640 static int handle_eob(void)
 641 {
 642     BufferedFile *bf = file;
 643     int len;
 644
 645     /* only tries to read if really end of buffer */
 646     if (bf->buf_ptr >= bf->buf_end) {
 647         if (bf->fd >= 0) {
 648 #if defined(PARSE_DEBUG)
 649             len = 1;
 650 #else
 651             len = IO_BUF_SIZE;
 652 #endif
 653             len = read(bf->fd, bf->buffer, len);
 654             if (len < 0)
 655                 len = 0;
 656         } else {
 657             len = 0;
 658         }
 659         total_bytes += len;
 660         bf->buf_ptr = bf->buffer;
 661         bf->buf_end = bf->buffer + len;
 662         *bf->buf_end = CH_EOB;
 663     }
 664     if (bf->buf_ptr < bf->buf_end) {
 665         return bf->buf_ptr[0];
 666     } else {
 667         bf->buf_ptr = bf->buf_end;
 668         return CH_EOF;
 669     }
 670 }
 671
 672 /* read next char from current input file and handle end of input buffer */
 673 static inline void inp(void)
 674 {
 675     ch = *(++(file->buf_ptr));
 676     /* end of buffer/file handling */
 677     if (ch == CH_EOB)
 678         ch = handle_eob();
 679 }
 680
 681 /* handle '\[\r]\n' */
 682 static int handle_stray_noerror(void)
 683 {
 684     while (ch == '\\') {
 685         inp();
 686         if (ch == '\n') {
 687             file->line_num++;
 688             inp();
 689         } else if (ch == '\r') {
 690             inp();
 691             if (ch != '\n')
 692                 goto fail;
 693             file->line_num++;
 694             inp();
 695         } else {
 696         fail:
 697             return 1;
 698         }
 699     }
 700     return 0;
 701 }
 702
 703 static void handle_stray(void)
 704 {
 705     if (handle_stray_noerror())
 706         tcc_error("stray '\\' in program");
 707 }
 708
 709 /* skip the stray and handle the \\n case. Output an error if
 710    incorrect char after the stray */
 711 static int handle_stray1(uint8_t *p)
 712 {
 713     int c;
 714
 715     file->buf_ptr = p;
 716     if (p >= file->buf_end) {
 717         c = handle_eob();
 718         if (c != '\\')
 719             return c;
 720         p = file->buf_ptr;
 721     }
 722     ch = *p;
 723     if (handle_stray_noerror()) {
 724         if (!(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
 725             tcc_error("stray '\\' in program");
 726         *--file->buf_ptr = '\\';
 727     }
 728     p = file->buf_ptr;
 729     c = *p;
 730     return c;
 731 }
 732
 733 /* handle just the EOB case, but not stray */
 734 #define PEEKC_EOB(c, p)\
 735 {\
 736     p++;\
 737     c = *p;\
 738     if (c == '\\') {\
 739         file->buf_ptr = p;\
 740         c = handle_eob();\
 741         p = file->buf_ptr;\
 742     }\
 743 }
 744
 745 /* handle the complicated stray case */
 746 #define PEEKC(c, p)\
 747 {\
 748     p++;\
 749     c = *p;\
 750     if (c == '\\') {\
 751         c = handle_stray1(p);\
 752         p = file->buf_ptr;\
 753     }\
 754 }
 755
 756 /* input with '\[\r]\n' handling. Note that this function cannot
 757    handle other characters after '\', so you cannot call it inside
 758    strings or comments */
 759 static void minp(void)
 760 {
 761     inp();
 762     if (ch == '\\')
 763         handle_stray();
 764 }
 765
 766 /* single line C++ comments */
 767 static uint8_t *parse_line_comment(uint8_t *p)
 768 {
 769     int c;
 770
 771     p++;
 772     for(;;) {
 773         c = *p;
 774     redo:
 775         if (c == '\n' || c == CH_EOF) {
 776             break;
 777         } else if (c == '\\') {
 778             file->buf_ptr = p;
 779             c = handle_eob();
 780             p = file->buf_ptr;
 781             if (c == '\\') {
 782                 PEEKC_EOB(c, p);
 783                 if (c == '\n') {
 784                     file->line_num++;
 785                     PEEKC_EOB(c, p);
 786                 } else if (c == '\r') {
 787                     PEEKC_EOB(c, p);
 788                     if (c == '\n') {
 789                         file->line_num++;
 790                         PEEKC_EOB(c, p);
 791                     }
 792                 }
 793             } else {
 794                 goto redo;
 795             }
 796         } else {
 797             p++;
 798         }
 799     }
 800     return p;
 801 }
 802
 803 /* C comments */
 804 static uint8_t *parse_comment(uint8_t *p)
 805 {
 806     int c;
 807
 808     p++;
 809     for(;;) {
 810         /* fast skip loop */
 811         for(;;) {
 812             c = *p;
 813             if (c == '\n' || c == '*' || c == '\\')
 814                 break;
 815             p++;
 816             c = *p;
 817             if (c == '\n' || c == '*' || c == '\\')
 818                 break;
 819             p++;
 820         }
 821         /* now we can handle all the cases */
 822         if (c == '\n') {
 823             file->line_num++;
 824             p++;
 825         } else if (c == '*') {
 826             p++;
 827             for(;;) {
 828                 c = *p;
 829                 if (c == '*') {
 830                     p++;
 831                 } else if (c == '/') {
 832                     goto end_of_comment;
 833                 } else if (c == '\\') {
 834                     file->buf_ptr = p;
 835                     c = handle_eob();
 836                     p = file->buf_ptr;
 837                     if (c == CH_EOF)
 838                         tcc_error("unexpected end of file in comment");
 839                     if (c == '\\') {
 840                         /* skip '\[\r]\n', otherwise just skip the stray */
 841                         while (c == '\\') {
 842                             PEEKC_EOB(c, p);
 843                             if (c == '\n') {
 844                                 file->line_num++;
 845                                 PEEKC_EOB(c, p);
 846                             } else if (c == '\r') {
 847                                 PEEKC_EOB(c, p);
 848                                 if (c == '\n') {
 849                                     file->line_num++;
 850                                     PEEKC_EOB(c, p);
 851                                 }
 852                             } else {
 853                                 goto after_star;
 854                             }
 855                         }
 856                     }
 857                 } else {
 858                     break;
 859                 }
 860             }
 861         after_star: ;
 862         } else {
 863             /* stray, eob or eof */
 864             file->buf_ptr = p;
 865             c = handle_eob();
 866             p = file->buf_ptr;
 867             if (c == CH_EOF) {
 868                 tcc_error("unexpected end of file in comment");
 869             } else if (c == '\\') {
 870                 p++;
 871             }
 872         }
 873     }
 874  end_of_comment:
 875     p++;
 876     return p;
 877 }
 878
 879 ST_FUNC int set_idnum(int c, int val)
 880 {
 881     int prev = isidnum_table[c - CH_EOF];
 882     isidnum_table[c - CH_EOF] = val;
 883     return prev;
 884 }
 885
 886 #define cinp minp
 887
 888 static inline void skip_spaces(void)
 889 {
 890     while (isidnum_table[ch - CH_EOF] & IS_SPC)
 891         cinp();
 892 }
 893
 894 static inline int check_space(int t, int *spc)
 895 {
 896     if (t < 256 && (isidnum_table[t - CH_EOF] & IS_SPC)) {
 897         if (*spc)
 898             return 1;
 899         *spc = 1;
 900     } else
 901         *spc = 0;
 902     return 0;
 903 }
 904
 905 /* parse a string without interpreting escapes */
 906 static uint8_t *parse_pp_string(uint8_t *p,
 907                                 int sep, CString *str)
 908 {
 909     int c;
 910     p++;
 911     for(;;) {
 912         c = *p;
 913         if (c == sep) {
 914             break;
 915         } else if (c == '\\') {
 916             file->buf_ptr = p;
 917             c = handle_eob();
 918             p = file->buf_ptr;
 919             if (c == CH_EOF) {
 920             unterminated_string:
 921                 /* XXX: indicate line number of start of string */
 922                 tcc_error("missing terminating %c character", sep);
 923             } else if (c == '\\') {
 924                 /* escape : just skip \[\r]\n */
 925                 PEEKC_EOB(c, p);
 926                 if (c == '\n') {
 927                     file->line_num++;
 928                     p++;
 929                 } else if (c == '\r') {
 930                     PEEKC_EOB(c, p);
 931                     if (c != '\n')
 932                         expect("'\n' after '\r'");
 933                     file->line_num++;
 934                     p++;
 935                 } else if (c == CH_EOF) {
 936                     goto unterminated_string;
 937                 } else {
 938                     if (str) {
 939                         cstr_ccat(str, '\\');
 940                         cstr_ccat(str, c);
 941                     }
 942                     p++;
 943                 }
 944             }
 945         } else if (c == '\n') {
 946             file->line_num++;
 947             goto add_char;
 948         } else if (c == '\r') {
 949             PEEKC_EOB(c, p);
 950             if (c != '\n') {
 951                 if (str)
 952                     cstr_ccat(str, '\r');
 953             } else {
 954                 file->line_num++;
 955                 goto add_char;
 956             }
 957         } else {
 958         add_char:
 959             if (str)
 960                 cstr_ccat(str, c);
 961             p++;
 962         }
 963     }
 964     p++;
 965     return p;
 966 }
 967
 968 /* skip block of text until #else, #elif or #endif. skip also pairs of
 969    #if/#endif */
 970 static void preprocess_skip(void)
 971 {
 972     int a, start_of_line, c, in_warn_or_error;
 973     uint8_t *p;
 974
 975     p = file->buf_ptr;
 976     a = 0;
 977 redo_start:
 978     start_of_line = 1;
 979     in_warn_or_error = 0;
 980     for(;;) {
 981     redo_no_start:
 982         c = *p;
 983         switch(c) {
 984         case ' ':
 985         case '\t':
 986         case '\f':
 987         case '\v':
 988         case '\r':
 989             p++;
 990             goto redo_no_start;
 991         case '\n':
 992             file->line_num++;
 993             p++;
 994             goto redo_start;
 995         case '\\':
 996             file->buf_ptr = p;
 997             c = handle_eob();
 998             if (c == CH_EOF) {
 999                 expect("#endif");
1000             } else if (c == '\\') {
1001                 ch = file->buf_ptr[0];
1002                 handle_stray_noerror();
1003             }
1004             p = file->buf_ptr;
1005             goto redo_no_start;
1006         /* skip strings */
1007         case '\"':
1008         case '\'':
1009             if (in_warn_or_error)
1010                 goto _default;
1011             p = parse_pp_string(p, c, NULL);
1012             break;
1013         /* skip comments */
1014         case '/':
1015             if (in_warn_or_error)
1016                 goto _default;
1017             file->buf_ptr = p;
1018             ch = *p;
1019             minp();
1020             p = file->buf_ptr;
1021             if (ch == '*') {
1022                 p = parse_comment(p);
1023             } else if (ch == '/') {
1024                 p = parse_line_comment(p);
1025             }
1026             break;
1027         case '#':
1028             p++;
1029             if (start_of_line) {
1030                 file->buf_ptr = p;
1031                 next_nomacro();
1032                 p = file->buf_ptr;
1033                 if (a == 0 &&
1034                     (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
1035                     goto the_end;
1036                 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
1037                     a++;
1038                 else if (tok == TOK_ENDIF)
1039                     a--;
1040                 else if( tok == TOK_ERROR || tok == TOK_WARNING)
1041                     in_warn_or_error = 1;
1042                 else if (tok == TOK_LINEFEED)
1043                     goto redo_start;
1044                 else if (parse_flags & PARSE_FLAG_ASM_FILE)
1045                     p = parse_line_comment(p - 1);
1046             }
1047 #if !defined(TCC_TARGET_ARM)
1048             else if (parse_flags & PARSE_FLAG_ASM_FILE)
1049                 p = parse_line_comment(p - 1);
1050 #else
1051             /* ARM assembly uses '#' for constants */
1052 #endif
1053             break;
1054 _default:
1055         default:
1056             p++;
1057             break;
1058         }
1059         start_of_line = 0;
1060     }
1061  the_end: ;
1062     file->buf_ptr = p;
1063 }
1064
1065 #if 0
1066 /* return the number of additional 'ints' necessary to store the
1067    token */
1068 static inline int tok_size(const int *p)
1069 {
1070     switch(*p) {
1071         /* 4 bytes */
1072     case TOK_CINT:
1073     case TOK_CUINT:
1074     case TOK_CCHAR:
1075     case TOK_LCHAR:
1076     case TOK_CFLOAT:
1077     case TOK_LINENUM:
1078         return 1 + 1;
1079     case TOK_STR:
1080     case TOK_LSTR:
1081     case TOK_PPNUM:
1082     case TOK_PPSTR:
1083         return 1 + ((sizeof(CString) + ((CString *)(p+1))->size + 3) >> 2);
1084     case TOK_CLONG:
1085     case TOK_CULONG:
1086         return 1 + LONG_SIZE / 4;
1087     case TOK_CDOUBLE:
1088     case TOK_CLLONG:
1089     case TOK_CULLONG:
1090         return 1 + 2;
1091     case TOK_CLDOUBLE:
1092         return 1 + LDOUBLE_SIZE / 4;
1093     default:
1094         return 1 + 0;
1095     }
1096 }
1097 #endif
1098
1099 /* token string handling */
1100 ST_INLN void tok_str_new(TokenString *s)
1101 {
1102     s->str = NULL;
1103     s->len = s->lastlen = 0;
1104     s->allocated_len = 0;
1105     s->last_line_num = -1;
1106 }
1107
1108 ST_FUNC TokenString *tok_str_alloc(void)
1109 {
1110     TokenString *str = tal_realloc(tokstr_alloc, 0, sizeof *str);
1111     tok_str_new(str);
1112     return str;
1113 }
1114
1115 ST_FUNC int *tok_str_dup(TokenString *s)
1116 {
1117     int *str;
1118
1119     str = tal_realloc(tokstr_alloc, 0, s->len * sizeof(int));
1120     memcpy(str, s->str, s->len * sizeof(int));
1121     return str;
1122 }
1123
1124 ST_FUNC void tok_str_free_str(int *str)
1125 {
1126     tal_free(tokstr_alloc, str);
1127 }
1128
1129 ST_FUNC void tok_str_free(TokenString *str)
1130 {
1131     tok_str_free_str(str->str);
1132     tal_free(tokstr_alloc, str);
1133 }
1134
1135 ST_FUNC int *tok_str_realloc(TokenString *s, int new_size)
1136 {
1137     int *str, size;
1138
1139     size = s->allocated_len;
1140     if (size < 16)
1141         size = 16;
1142     while (size < new_size)
1143         size = size * 2;
1144     if (size > s->allocated_len) {
1145         str = tal_realloc(tokstr_alloc, s->str, size * sizeof(int));
1146         s->allocated_len = size;
1147         s->str = str;
1148     }
1149     return s->str;
1150 }
1151
1152 ST_FUNC void tok_str_add(TokenString *s, int t)
1153 {
1154     int len, *str;
1155
1156     len = s->len;
1157     str = s->str;
1158     if (len >= s->allocated_len)
1159         str = tok_str_realloc(s, len + 1);
1160     str[len++] = t;
1161     s->len = len;
1162 }
1163
1164 ST_FUNC void begin_macro(TokenString *str, int alloc)
1165 {
1166     str->alloc = alloc;
1167     str->prev = macro_stack;
1168     str->prev_ptr = macro_ptr;
1169     str->save_line_num = file->line_num;
1170     macro_ptr = str->str;
1171     macro_stack = str;
1172 }
1173
1174 ST_FUNC void end_macro(void)
1175 {
1176     TokenString *str = macro_stack;
1177     macro_stack = str->prev;
1178     macro_ptr = str->prev_ptr;
1179     file->line_num = str->save_line_num;
1180     if (str->alloc != 0) {
1181         if (str->alloc == 2)
1182             str->str = NULL; /* don't free */
1183         tok_str_free(str);
1184     }
1185 }
1186
1187 static void tok_str_add2(TokenString *s, int t, CValue *cv)
1188 {
1189     int len, *str;
1190
1191     len = s->lastlen = s->len;
1192     str = s->str;
1193
1194     /* allocate space for worst case */
1195     if (len + TOK_MAX_SIZE >= s->allocated_len)
1196         str = tok_str_realloc(s, len + TOK_MAX_SIZE + 1);
1197     str[len++] = t;
1198     switch(t) {
1199     case TOK_CINT:
1200     case TOK_CUINT:
1201     case TOK_CCHAR:
1202     case TOK_LCHAR:
1203     case TOK_CFLOAT:
1204     case TOK_LINENUM:
1205 #if LONG_SIZE == 4
1206     case TOK_CLONG:
1207     case TOK_CULONG:
1208 #endif
1209         str[len++] = cv->tab[0];
1210         break;
1211     case TOK_PPNUM:
1212     case TOK_PPSTR:
1213     case TOK_STR:
1214     case TOK_LSTR:
1215         {
1216             /* Insert the string into the int array. */
1217             size_t nb_words =
1218                 1 + (cv->str.size + sizeof(int) - 1) / sizeof(int);
1219             if (len + nb_words >= s->allocated_len)
1220                 str = tok_str_realloc(s, len + nb_words + 1);
1221             str[len] = cv->str.size;
1222             memcpy(&str[len + 1], cv->str.data, cv->str.size);
1223             len += nb_words;
1224         }
1225         break;
1226     case TOK_CDOUBLE:
1227     case TOK_CLLONG:
1228     case TOK_CULLONG:
1229 #if LONG_SIZE == 8
1230     case TOK_CLONG:
1231     case TOK_CULONG:
1232 #endif
1233 #if LDOUBLE_SIZE == 8
1234     case TOK_CLDOUBLE:
1235 #endif
1236         str[len++] = cv->tab[0];
1237         str[len++] = cv->tab[1];
1238         break;
1239 #if LDOUBLE_SIZE == 12
1240     case TOK_CLDOUBLE:
1241         str[len++] = cv->tab[0];
1242         str[len++] = cv->tab[1];
1243         str[len++] = cv->tab[2];
1244 #elif LDOUBLE_SIZE == 16
1245     case TOK_CLDOUBLE:
1246         str[len++] = cv->tab[0];
1247         str[len++] = cv->tab[1];
1248         str[len++] = cv->tab[2];
1249         str[len++] = cv->tab[3];
1250 #elif LDOUBLE_SIZE != 8
1251 #error add long double size support
1252 #endif
1253         break;
1254     default:
1255         break;
1256     }
1257     s->len = len;
1258 }
1259
1260 /* add the current parse token in token string 's' */
1261 ST_FUNC void tok_str_add_tok(TokenString *s)
1262 {
1263     CValue cval;
1264
1265     /* save line number info */
1266     if (file->line_num != s->last_line_num) {
1267         s->last_line_num = file->line_num;
1268         cval.i = s->last_line_num;
1269         tok_str_add2(s, TOK_LINENUM, &cval);
1270     }
1271     tok_str_add2(s, tok, &tokc);
1272 }
1273
1274 /* get a token from an integer array and increment pointer. */
1275 static inline void tok_get(int *t, const int **pp, CValue *cv)
1276 {
1277     const int *p = *pp;
1278     int n, *tab;
1279
1280     tab = cv->tab;
1281     switch(*t = *p++) {
1282 #if LONG_SIZE == 4
1283     case TOK_CLONG:
1284 #endif
1285     case TOK_CINT:
1286     case TOK_CCHAR:
1287     case TOK_LCHAR:
1288     case TOK_LINENUM:
1289         cv->i = *p++;
1290         break;
1291 #if LONG_SIZE == 4
1292     case TOK_CULONG:
1293 #endif
1294     case TOK_CUINT:
1295         cv->i = (unsigned)*p++;
1296         break;
1297     case TOK_CFLOAT:
1298         tab[0] = *p++;
1299         break;
1300     case TOK_STR:
1301     case TOK_LSTR:
1302     case TOK_PPNUM:
1303     case TOK_PPSTR:
1304         cv->str.size = *p++;
1305         cv->str.data = p;
1306         p += (cv->str.size + sizeof(int) - 1) / sizeof(int);
1307         break;
1308     case TOK_CDOUBLE:
1309     case TOK_CLLONG:
1310     case TOK_CULLONG:
1311 #if LONG_SIZE == 8
1312     case TOK_CLONG:
1313     case TOK_CULONG:
1314 #endif
1315         n = 2;
1316         goto copy;
1317     case TOK_CLDOUBLE:
1318 #if LDOUBLE_SIZE == 16
1319         n = 4;
1320 #elif LDOUBLE_SIZE == 12
1321         n = 3;
1322 #elif LDOUBLE_SIZE == 8
1323         n = 2;
1324 #else
1325 # error add long double size support
1326 #endif
1327     copy:
1328         do
1329             *tab++ = *p++;
1330         while (--n);
1331         break;
1332     default:
1333         break;
1334     }
1335     *pp = p;
1336 }
1337
1338 #if 0
1339 # define TOK_GET(t,p,c) tok_get(t,p,c)
1340 #else
1341 # define TOK_GET(t,p,c) do { \
1342     int _t = **(p); \
1343     if (TOK_HAS_VALUE(_t)) \
1344         tok_get(t, p, c); \
1345     else \
1346         *(t) = _t, ++*(p); \
1347     } while (0)
1348 #endif
1349
1350 static int macro_is_equal(const int *a, const int *b)
1351 {
1352     CValue cv;
1353     int t;
1354
1355     if (!a || !b)
1356         return 1;
1357
1358     while (*a && *b) {
1359         /* first time preallocate macro_equal_buf, next time only reset position to start */
1360         cstr_reset(&macro_equal_buf);
1361         TOK_GET(&t, &a, &cv);
1362         cstr_cat(&macro_equal_buf, get_tok_str(t, &cv), 0);
1363         TOK_GET(&t, &b, &cv);
1364         if (strcmp(macro_equal_buf.data, get_tok_str(t, &cv)))
1365             return 0;
1366     }
1367     return !(*a || *b);
1368 }
1369
1370 /* defines handling */
1371 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1372 {
1373     Sym *s, *o;
1374
1375     o = define_find(v);
1376     s = sym_push2(&define_stack, v, macro_type, 0);
1377     s->d = str;
1378     s->next = first_arg;
1379     table_ident[v - TOK_IDENT]->sym_define = s;
1380
1381     if (o && !macro_is_equal(o->d, s->d))
1382         tcc_warning("%s redefined", get_tok_str(v, NULL));
1383 }
1384
1385 /* undefined a define symbol. Its name is just set to zero */
1386 ST_FUNC void define_undef(Sym *s)
1387 {
1388     int v = s->v;
1389     if (v >= TOK_IDENT && v < tok_ident)
1390         table_ident[v - TOK_IDENT]->sym_define = NULL;
1391 }
1392
1393 ST_INLN Sym *define_find(int v)
1394 {
1395     v -= TOK_IDENT;
1396     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1397         return NULL;
1398     return table_ident[v]->sym_define;
1399 }
1400
1401 /* free define stack until top reaches 'b' */
1402 ST_FUNC void free_defines(Sym *b)
1403 {
1404     while (define_stack != b) {
1405         Sym *top = define_stack;
1406         define_stack = top->prev;
1407         tok_str_free_str(top->d);
1408         define_undef(top);
1409         sym_free(top);
1410     }
1411 }
1412
1413 /* label lookup */
1414 ST_FUNC Sym *label_find(int v)
1415 {
1416     v -= TOK_IDENT;
1417     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1418         return NULL;
1419     return table_ident[v]->sym_label;
1420 }
1421
1422 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1423 {
1424     Sym *s, **ps;
1425     s = sym_push2(ptop, v, 0, 0);
1426     s->r = flags;
1427     ps = &table_ident[v - TOK_IDENT]->sym_label;
1428     if (ptop == &global_label_stack) {
1429         /* modify the top most local identifier, so that
1430            sym_identifier will point to 's' when popped */
1431         while (*ps != NULL)
1432             ps = &(*ps)->prev_tok;
1433     }
1434     s->prev_tok = *ps;
1435     *ps = s;
1436     return s;
1437 }
1438
1439 /* pop labels until element last is reached. Look if any labels are
1440    undefined. Define symbols if '&&label' was used. */
1441 ST_FUNC void label_pop(Sym **ptop, Sym *slast, int keep)
1442 {
1443     Sym *s, *s1;
1444     for(s = *ptop; s != slast; s = s1) {
1445         s1 = s->prev;
1446         if (s->r == LABEL_DECLARED) {
1447             tcc_warning_c(warn_all)("label '%s' declared but not used", get_tok_str(s->v, NULL));
1448         } else if (s->r == LABEL_FORWARD) {
1449                 tcc_error("label '%s' used but not defined",
1450                       get_tok_str(s->v, NULL));
1451         } else {
1452             if (s->c) {
1453                 /* define corresponding symbol. A size of
1454                    1 is put. */
1455                 put_extern_sym(s, cur_text_section, s->jnext, 1);
1456             }
1457         }
1458         /* remove label */
1459         if (s->r != LABEL_GONE)
1460             table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1461         if (!keep)
1462             sym_free(s);
1463         else
1464             s->r = LABEL_GONE;
1465     }
1466     if (!keep)
1467         *ptop = slast;
1468 }
1469
1470 /* fake the nth "#if defined test_..." for tcc -dt -run */
1471 static void maybe_run_test(TCCState *s)
1472 {
1473     const char *p;
1474     if (s->include_stack_ptr != s->include_stack)
1475         return;
1476     p = get_tok_str(tok, NULL);
1477     if (0 != memcmp(p, "test_", 5))
1478         return;
1479     if (0 != --s->run_test)
1480         return;
1481     fprintf(s->ppfp, &"\n[%s]\n"[!(s->dflag & 32)], p), fflush(s->ppfp);
1482     define_push(tok, MACRO_OBJ, NULL, NULL);
1483 }
1484
1485 /* eval an expression for #if/#elif */
1486 static int expr_preprocess(void)
1487 {
1488     int c, t;
1489     TokenString *str;
1490
1491     str = tok_str_alloc();
1492     pp_expr = 1;
1493     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1494         next(); /* do macro subst */
1495       redo:
1496         if (tok == TOK_DEFINED) {
1497             next_nomacro();
1498             t = tok;
1499             if (t == '(')
1500                 next_nomacro();
1501             if (tok < TOK_IDENT)
1502                 expect("identifier");
1503             if (tcc_state->run_test)
1504                 maybe_run_test(tcc_state);
1505             c = define_find(tok) != 0;
1506             if (t == '(') {
1507                 next_nomacro();
1508                 if (tok != ')')
1509                     expect("')'");
1510             }
1511             tok = TOK_CINT;
1512             tokc.i = c;
1513         } else if (1 && tok == TOK___HAS_INCLUDE) {
1514             next();  /* XXX check if correct to use expansion */
1515             skip('(');
1516             while (tok != ')' && tok != TOK_EOF)
1517               next();
1518             if (tok != ')')
1519               expect("')'");
1520             tok = TOK_CINT;
1521             tokc.i = 0;
1522         } else if (tok >= TOK_IDENT) {
1523             /* if undefined macro, replace with zero, check for func-like */
1524             t = tok;
1525             tok = TOK_CINT;
1526             tokc.i = 0;
1527             tok_str_add_tok(str);
1528             next();
1529             if (tok == '(')
1530                 tcc_error("function-like macro '%s' is not defined",
1531                           get_tok_str(t, NULL));
1532             goto redo;
1533         }
1534         tok_str_add_tok(str);
1535     }
1536     pp_expr = 0;
1537     tok_str_add(str, -1); /* simulate end of file */
1538     tok_str_add(str, 0);
1539     /* now evaluate C constant expression */
1540     begin_macro(str, 1);
1541     next();
1542     c = expr_const();
1543     end_macro();
1544     return c != 0;
1545 }
1546
1547
1548 /* parse after #define */
1549 ST_FUNC void parse_define(void)
1550 {
1551     Sym *s, *first, **ps;
1552     int v, t, varg, is_vaargs, spc;
1553     int saved_parse_flags = parse_flags;
1554
1555     v = tok;
1556     if (v < TOK_IDENT || v == TOK_DEFINED)
1557         tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1558     /* XXX: should check if same macro (ANSI) */
1559     first = NULL;
1560     t = MACRO_OBJ;
1561     /* We have to parse the whole define as if not in asm mode, in particular
1562        no line comment with '#' must be ignored.  Also for function
1563        macros the argument list must be parsed without '.' being an ID
1564        character.  */
1565     parse_flags = ((parse_flags & ~PARSE_FLAG_ASM_FILE) | PARSE_FLAG_SPACES);
1566     /* '(' must be just after macro definition for MACRO_FUNC */
1567     next_nomacro();
1568     parse_flags &= ~PARSE_FLAG_SPACES;
1569     if (tok == '(') {
1570         int dotid = set_idnum('.', 0);
1571         next_nomacro();
1572         ps = &first;
1573         if (tok != ')') for (;;) {
1574             varg = tok;
1575             next_nomacro();
1576             is_vaargs = 0;
1577             if (varg == TOK_DOTS) {
1578                 varg = TOK___VA_ARGS__;
1579                 is_vaargs = 1;
1580             } else if (tok == TOK_DOTS && gnu_ext) {
1581                 is_vaargs = 1;
1582                 next_nomacro();
1583             }
1584             if (varg < TOK_IDENT)
1585         bad_list:
1586                 tcc_error("bad macro parameter list");
1587             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1588             *ps = s;
1589             ps = &s->next;
1590             if (tok == ')')
1591                 break;
1592             if (tok != ',' || is_vaargs)
1593                 goto bad_list;
1594             next_nomacro();
1595         }
1596         parse_flags |= PARSE_FLAG_SPACES;
1597         next_nomacro();
1598         t = MACRO_FUNC;
1599         set_idnum('.', dotid);
1600     }
1601
1602     tokstr_buf.len = 0;
1603     spc = 2;
1604     parse_flags |= PARSE_FLAG_ACCEPT_STRAYS | PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED;
1605     /* The body of a macro definition should be parsed such that identifiers
1606        are parsed like the file mode determines (i.e. with '.' being an
1607        ID character in asm mode).  But '#' should be retained instead of
1608        regarded as line comment leader, so still don't set ASM_FILE
1609        in parse_flags. */
1610     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1611         /* remove spaces around ## and after '#' */
1612         if (TOK_TWOSHARPS == tok) {
1613             if (2 == spc)
1614                 goto bad_twosharp;
1615             if (1 == spc)
1616                 --tokstr_buf.len;
1617             spc = 3;
1618             tok = TOK_PPJOIN;
1619         } else if ('#' == tok) {
1620             spc = 4;
1621         } else if (check_space(tok, &spc)) {
1622             goto skip;
1623         }
1624         tok_str_add2(&tokstr_buf, tok, &tokc);
1625     skip:
1626         next_nomacro();
1627     }
1628
1629     parse_flags = saved_parse_flags;
1630     if (spc == 1)
1631         --tokstr_buf.len; /* remove trailing space */
1632     tok_str_add(&tokstr_buf, 0);
1633     if (3 == spc)
1634 bad_twosharp:
1635         tcc_error("'##' cannot appear at either end of macro");
1636     define_push(v, t, tok_str_dup(&tokstr_buf), first);
1637 }
1638
1639 static CachedInclude *search_cached_include(TCCState *s1, const char *filename, int add)
1640 {
1641     const unsigned char *s;
1642     unsigned int h;
1643     CachedInclude *e;
1644     int i;
1645
1646     h = TOK_HASH_INIT;
1647     s = (unsigned char *) filename;
1648     while (*s) {
1649 #ifdef _WIN32
1650         h = TOK_HASH_FUNC(h, toup(*s));
1651 #else
1652         h = TOK_HASH_FUNC(h, *s);
1653 #endif
1654         s++;
1655     }
1656     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1657
1658     i = s1->cached_includes_hash[h];
1659     for(;;) {
1660         if (i == 0)
1661             break;
1662         e = s1->cached_includes[i - 1];
1663         if (0 == PATHCMP(e->filename, filename))
1664             return e;
1665         i = e->hash_next;
1666     }
1667     if (!add)
1668         return NULL;
1669
1670     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1671     strcpy(e->filename, filename);
1672     e->ifndef_macro = e->once = 0;
1673     dynarray_add(&s1->cached_includes, &s1->nb_cached_includes, e);
1674     /* add in hash table */
1675     e->hash_next = s1->cached_includes_hash[h];
1676     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1677 #ifdef INC_DEBUG
1678     printf("adding cached '%s'\n", filename);
1679 #endif
1680     return e;
1681 }
1682
1683 static void pragma_parse(TCCState *s1)
1684 {
1685     next_nomacro();
1686     if (tok == TOK_push_macro || tok == TOK_pop_macro) {
1687         int t = tok, v;
1688         Sym *s;
1689
1690         if (next(), tok != '(')
1691             goto pragma_err;
1692         if (next(), tok != TOK_STR)
1693             goto pragma_err;
1694         v = tok_alloc(tokc.str.data, tokc.str.size - 1)->tok;
1695         if (next(), tok != ')')
1696             goto pragma_err;
1697         if (t == TOK_push_macro) {
1698             while (NULL == (s = define_find(v)))
1699                 define_push(v, 0, NULL, NULL);
1700             s->type.ref = s; /* set push boundary */
1701         } else {
1702             for (s = define_stack; s; s = s->prev)
1703                 if (s->v == v && s->type.ref == s) {
1704                     s->type.ref = NULL;
1705                     break;
1706                 }
1707         }
1708         if (s)
1709             table_ident[v - TOK_IDENT]->sym_define = s->d ? s : NULL;
1710         else
1711             tcc_warning("unbalanced #pragma pop_macro");
1712         pp_debug_tok = t, pp_debug_symv = v;
1713
1714     } else if (tok == TOK_once) {
1715         search_cached_include(s1, file->filename, 1)->once = pp_once;
1716
1717     } else if (s1->output_type == TCC_OUTPUT_PREPROCESS) {
1718         /* tcc -E: keep pragmas below unchanged */
1719         unget_tok(' ');
1720         unget_tok(TOK_PRAGMA);
1721         unget_tok('#');
1722         unget_tok(TOK_LINEFEED);
1723
1724     } else if (tok == TOK_pack) {
1725         /* This may be:
1726            #pragma pack(1) // set
1727            #pragma pack() // reset to default
1728            #pragma pack(push,1) // push & set
1729            #pragma pack(pop) // restore previous */
1730         next();
1731         skip('(');
1732         if (tok == TOK_ASM_pop) {
1733             next();
1734             if (s1->pack_stack_ptr <= s1->pack_stack) {
1735             stk_error:
1736                 tcc_error("out of pack stack");
1737             }
1738             s1->pack_stack_ptr--;
1739         } else {
1740             int val = 0;
1741             if (tok != ')') {
1742                 if (tok == TOK_ASM_push) {
1743                     next();
1744                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1745                         goto stk_error;
1746                     s1->pack_stack_ptr++;
1747                     skip(',');
1748                 }
1749                 if (tok != TOK_CINT)
1750                     goto pragma_err;
1751                 val = tokc.i;
1752                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1753                     goto pragma_err;
1754                 next();
1755             }
1756             *s1->pack_stack_ptr = val;
1757         }
1758         if (tok != ')')
1759             goto pragma_err;
1760
1761     } else if (tok == TOK_comment) {
1762         char *p; int t;
1763         next();
1764         skip('(');
1765         t = tok;
1766         next();
1767         skip(',');
1768         if (tok != TOK_STR)
1769             goto pragma_err;
1770         p = tcc_strdup((char *)tokc.str.data);
1771         next();
1772         if (tok != ')')
1773             goto pragma_err;
1774         if (t == TOK_lib) {
1775             dynarray_add(&s1->pragma_libs, &s1->nb_pragma_libs, p);
1776         } else {
1777             if (t == TOK_option)
1778                 tcc_set_options(s1, p);
1779             tcc_free(p);
1780         }
1781
1782     } else
1783         tcc_warning_c(warn_unsupported)("#pragma %s ignored", get_tok_str(tok, &tokc));
1784     return;
1785
1786 pragma_err:
1787     tcc_error("malformed #pragma directive");
1788     return;
1789 }
1790
1791 /* is_bof is true if first non space token at beginning of file */
1792 ST_FUNC void preprocess(int is_bof)
1793 {
1794     TCCState *s1 = tcc_state;
1795     int i, c, n, saved_parse_flags;
1796     char buf[1024], *q;
1797     Sym *s;
1798
1799     saved_parse_flags = parse_flags;
1800     parse_flags = PARSE_FLAG_PREPROCESS
1801         | PARSE_FLAG_TOK_NUM
1802         | PARSE_FLAG_TOK_STR
1803         | PARSE_FLAG_LINEFEED
1804         | (parse_flags & PARSE_FLAG_ASM_FILE)
1805         ;
1806
1807     next_nomacro();
1808  redo:
1809     switch(tok) {
1810     case TOK_DEFINE:
1811         pp_debug_tok = tok;
1812         next_nomacro();
1813         pp_debug_symv = tok;
1814         parse_define();
1815         break;
1816     case TOK_UNDEF:
1817         pp_debug_tok = tok;
1818         next_nomacro();
1819         pp_debug_symv = tok;
1820         s = define_find(tok);
1821         /* undefine symbol by putting an invalid name */
1822         if (s)
1823             define_undef(s);
1824         break;
1825     case TOK_INCLUDE:
1826     case TOK_INCLUDE_NEXT:
1827         ch = file->buf_ptr[0];
1828         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1829         skip_spaces();
1830         if (ch == '<') {
1831             c = '>';
1832             goto read_name;
1833         } else if (ch == '\"') {
1834             c = ch;
1835         read_name:
1836             inp();
1837             q = buf;
1838             while (ch != c && ch != '\n' && ch != CH_EOF) {
1839                 if ((q - buf) < sizeof(buf) - 1)
1840                     *q++ = ch;
1841                 if (ch == '\\') {
1842                     if (handle_stray_noerror() == 0)
1843                         --q;
1844                 } else
1845                     inp();
1846             }
1847             *q = '\0';
1848             minp();
1849 #if 0
1850             /* eat all spaces and comments after include */
1851             /* XXX: slightly incorrect */
1852             while (ch1 != '\n' && ch1 != CH_EOF)
1853                 inp();
1854 #endif
1855         } else {
1856             int len;
1857             /* computed #include : concatenate everything up to linefeed,
1858                the result must be one of the two accepted forms.
1859                Don't convert pp-tokens to tokens here.  */
1860             parse_flags = (PARSE_FLAG_PREPROCESS
1861                            | PARSE_FLAG_LINEFEED
1862                            | (parse_flags & PARSE_FLAG_ASM_FILE));
1863             next();
1864             buf[0] = '\0';
1865             while (tok != TOK_LINEFEED) {
1866                 pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1867                 next();
1868             }
1869             len = strlen(buf);
1870             /* check syntax and remove '<>|""' */
1871             if ((len < 2 || ((buf[0] != '"' || buf[len-1] != '"') &&
1872                              (buf[0] != '<' || buf[len-1] != '>'))))
1873                 tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1874             c = buf[len-1];
1875             memmove(buf, buf + 1, len - 2);
1876             buf[len - 2] = '\0';
1877         }
1878
1879         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1880             tcc_error("#include recursion too deep");
1881         i = tok == TOK_INCLUDE_NEXT ? file->include_next_index + 1 : 0;
1882         n = 2 + s1->nb_include_paths + s1->nb_sysinclude_paths;
1883         for (; i < n; ++i) {
1884             char buf1[sizeof file->filename];
1885             CachedInclude *e;
1886             const char *path;
1887
1888             if (i == 0) {
1889                 /* check absolute include path */
1890                 if (!IS_ABSPATH(buf))
1891                     continue;
1892                 buf1[0] = 0;
1893
1894             } else if (i == 1) {
1895                 /* search in file's dir if "header.h" */
1896                 if (c != '\"')
1897                     continue;
1898                 /* https://savannah.nongnu.org/bugs/index.php?50847 */
1899                 path = file->true_filename;
1900                 pstrncpy(buf1, path, tcc_basename(path) - path);
1901
1902             } else {
1903                 /* search in all the include paths */
1904                 int j = i - 2, k = j - s1->nb_include_paths;
1905                 path = k < 0 ? s1->include_paths[j] : s1->sysinclude_paths[k];
1906                 pstrcpy(buf1, sizeof(buf1), path);
1907                 pstrcat(buf1, sizeof(buf1), "/");
1908             }
1909
1910             pstrcat(buf1, sizeof(buf1), buf);
1911             e = search_cached_include(s1, buf1, 0);
1912             if (e && (define_find(e->ifndef_macro) || e->once == pp_once)) {
1913                 /* no need to parse the include because the 'ifndef macro'
1914                    is defined (or had #pragma once) */
1915 #ifdef INC_DEBUG
1916                 printf("%s: skipping cached %s\n", file->filename, buf1);
1917 #endif
1918                 goto include_done;
1919             }
1920
1921             if (tcc_open(s1, buf1) < 0)
1922                 continue;
1923             /* push previous file on stack */
1924             *s1->include_stack_ptr++ = file->prev;
1925             file->include_next_index = i;
1926 #ifdef INC_DEBUG
1927             printf("%s: including %s\n", file->prev->filename, file->filename);
1928 #endif
1929             /* update target deps */
1930             if (s1->gen_deps) {
1931                 BufferedFile *bf = file;
1932                 while (i == 1 && (bf = bf->prev))
1933                     i = bf->include_next_index;
1934                 /* skip system include files */
1935                 if (s1->include_sys_deps || n - i > s1->nb_sysinclude_paths)
1936                     dynarray_add(&s1->target_deps, &s1->nb_target_deps,
1937                         tcc_strdup(buf1));
1938             }
1939             /* add include file debug info */
1940             tcc_debug_bincl(tcc_state);
1941             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1942             ch = file->buf_ptr[0];
1943             goto the_end;
1944         }
1945         tcc_error("include file '%s' not found", buf);
1946 include_done:
1947         break;
1948     case TOK_IFNDEF:
1949         c = 1;
1950         goto do_ifdef;
1951     case TOK_IF:
1952         c = expr_preprocess();
1953         goto do_if;
1954     case TOK_IFDEF:
1955         c = 0;
1956     do_ifdef:
1957         next_nomacro();
1958         if (tok < TOK_IDENT)
1959             tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1960         if (is_bof) {
1961             if (c) {
1962 #ifdef INC_DEBUG
1963                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1964 #endif
1965                 file->ifndef_macro = tok;
1966             }
1967         }
1968         c = (define_find(tok) != 0) ^ c;
1969     do_if:
1970         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1971             tcc_error("memory full (ifdef)");
1972         *s1->ifdef_stack_ptr++ = c;
1973         goto test_skip;
1974     case TOK_ELSE:
1975         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1976             tcc_error("#else without matching #if");
1977         if (s1->ifdef_stack_ptr[-1] & 2)
1978             tcc_error("#else after #else");
1979         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1980         goto test_else;
1981     case TOK_ELIF:
1982         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1983             tcc_error("#elif without matching #if");
1984         c = s1->ifdef_stack_ptr[-1];
1985         if (c > 1)
1986             tcc_error("#elif after #else");
1987         /* last #if/#elif expression was true: we skip */
1988         if (c == 1) {
1989             c = 0;
1990         } else {
1991             c = expr_preprocess();
1992             s1->ifdef_stack_ptr[-1] = c;
1993         }
1994     test_else:
1995         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1996             file->ifndef_macro = 0;
1997     test_skip:
1998         if (!(c & 1)) {
1999             preprocess_skip();
2000             is_bof = 0;
2001             goto redo;
2002         }
2003         break;
2004     case TOK_ENDIF:
2005         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
2006             tcc_error("#endif without matching #if");
2007         s1->ifdef_stack_ptr--;
2008         /* '#ifndef macro' was at the start of file. Now we check if
2009            an '#endif' is exactly at the end of file */
2010         if (file->ifndef_macro &&
2011             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
2012             file->ifndef_macro_saved = file->ifndef_macro;
2013             /* need to set to zero to avoid false matches if another
2014                #ifndef at middle of file */
2015             file->ifndef_macro = 0;
2016             while (tok != TOK_LINEFEED)
2017                 next_nomacro();
2018             tok_flags |= TOK_FLAG_ENDIF;
2019             goto the_end;
2020         }
2021         break;
2022     case TOK_PPNUM:
2023         n = strtoul((char*)tokc.str.data, &q, 10);
2024         goto _line_num;
2025     case TOK_LINE:
2026         next();
2027         if (tok != TOK_CINT)
2028     _line_err:
2029             tcc_error("wrong #line format");
2030         n = tokc.i;
2031     _line_num:
2032         next();
2033         if (tok != TOK_LINEFEED) {
2034             if (tok == TOK_STR) {
2035                 if (file->true_filename == file->filename)
2036                     file->true_filename = tcc_strdup(file->filename);
2037                 /* prepend directory from real file */
2038                 pstrcpy(buf, sizeof buf, file->true_filename);
2039                 *tcc_basename(buf) = 0;
2040                 pstrcat(buf, sizeof buf, (char *)tokc.str.data);
2041                 tcc_debug_putfile(s1, buf);
2042             } else if (parse_flags & PARSE_FLAG_ASM_FILE)
2043                 break;
2044             else
2045                 goto _line_err;
2046             --n;
2047         }
2048         if (file->fd > 0)
2049             total_lines += file->line_num - n;
2050         file->line_num = n;
2051         break;
2052     case TOK_ERROR:
2053     case TOK_WARNING:
2054         c = tok;
2055         ch = file->buf_ptr[0];
2056         skip_spaces();
2057         q = buf;
2058         while (ch != '\n' && ch != CH_EOF) {
2059             if ((q - buf) < sizeof(buf) - 1)
2060                 *q++ = ch;
2061             if (ch == '\\') {
2062                 if (handle_stray_noerror() == 0)
2063                     --q;
2064             } else
2065                 inp();
2066         }
2067         *q = '\0';
2068         if (c == TOK_ERROR)
2069             tcc_error("#error %s", buf);
2070         else
2071             tcc_warning("#warning %s", buf);
2072         break;
2073     case TOK_PRAGMA:
2074         pragma_parse(s1);
2075         break;
2076     case TOK_LINEFEED:
2077         goto the_end;
2078     default:
2079         /* ignore gas line comment in an 'S' file. */
2080         if (saved_parse_flags & PARSE_FLAG_ASM_FILE)
2081             goto ignore;
2082         if (tok == '!' && is_bof)
2083             /* '!' is ignored at beginning to allow C scripts. */
2084             goto ignore;
2085         tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
2086     ignore:
2087         file->buf_ptr = parse_line_comment(file->buf_ptr - 1);
2088         goto the_end;
2089     }
2090     /* ignore other preprocess commands or #! for C scripts */
2091     while (tok != TOK_LINEFEED)
2092         next_nomacro();
2093  the_end:
2094     parse_flags = saved_parse_flags;
2095 }
2096
2097 /* evaluate escape codes in a string. */
2098 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
2099 {
2100     int c, n, i;
2101     const uint8_t *p;
2102
2103     p = buf;
2104     for(;;) {
2105         c = *p;
2106         if (c == '\0')
2107             break;
2108         if (c == '\\') {
2109             p++;
2110             /* escape */
2111             c = *p;
2112             switch(c) {
2113             case '0': case '1': case '2': case '3':
2114             case '4': case '5': case '6': case '7':
2115                 /* at most three octal digits */
2116                 n = c - '0';
2117                 p++;
2118                 c = *p;
2119                 if (isoct(c)) {
2120                     n = n * 8 + c - '0';
2121                     p++;
2122                     c = *p;
2123                     if (isoct(c)) {
2124                         n = n * 8 + c - '0';
2125                         p++;
2126                     }
2127                 }
2128                 c = n;
2129                 goto add_char_nonext;
2130             case 'x': i = 0; goto parse_hex_or_ucn;
2131             case 'u': i = 4; goto parse_hex_or_ucn;
2132             case 'U': i = 8; goto parse_hex_or_ucn;
2133     parse_hex_or_ucn:
2134                 p++;
2135                 n = 0;
2136                 do {
2137                     c = *p;
2138                     if (c >= 'a' && c <= 'f')
2139                         c = c - 'a' + 10;
2140                     else if (c >= 'A' && c <= 'F')
2141                         c = c - 'A' + 10;
2142                     else if (isnum(c))
2143                         c = c - '0';
2144                     else if (i > 0)
2145                         expect("more hex digits in universal-character-name");
2146                     else {
2147                         c = n;
2148                         goto add_char_nonext;
2149                     }
2150                     n = n * 16 + c;
2151                     p++;
2152                 } while (--i);
2153                 cstr_u8cat(outstr, n);
2154                 continue;
2155             case 'a':
2156                 c = '\a';
2157                 break;
2158             case 'b':
2159                 c = '\b';
2160                 break;
2161             case 'f':
2162                 c = '\f';
2163                 break;
2164             case 'n':
2165                 c = '\n';
2166                 break;
2167             case 'r':
2168                 c = '\r';
2169                 break;
2170             case 't':
2171                 c = '\t';
2172                 break;
2173             case 'v':
2174                 c = '\v';
2175                 break;
2176             case 'e':
2177                 if (!gnu_ext)
2178                     goto invalid_escape;
2179                 c = 27;
2180                 break;
2181             case '\'':
2182             case '\"':
2183             case '\\':
2184             case '?':
2185                 break;
2186             default:
2187             invalid_escape:
2188                 if (c >= '!' && c <= '~')
2189                     tcc_warning("unknown escape sequence: \'\\%c\'", c);
2190                 else
2191                     tcc_warning("unknown escape sequence: \'\\x%x\'", c);
2192                 break;
2193             }
2194         } else if (is_long && c >= 0x80) {
2195             /* assume we are processing UTF-8 sequence */
2196             /* reference: The Unicode Standard, Version 10.0, ch3.9 */
2197
2198             int cont; /* count of continuation bytes */
2199             int skip; /* how many bytes should skip when error occurred */
2200             int i;
2201
2202             /* decode leading byte */
2203             if (c < 0xC2) {
2204                     skip = 1; goto invalid_utf8_sequence;
2205             } else if (c <= 0xDF) {
2206                     cont = 1; n = c & 0x1f;
2207             } else if (c <= 0xEF) {
2208                     cont = 2; n = c & 0xf;
2209             } else if (c <= 0xF4) {
2210                     cont = 3; n = c & 0x7;
2211             } else {
2212                     skip = 1; goto invalid_utf8_sequence;
2213             }
2214
2215             /* decode continuation bytes */
2216             for (i = 1; i <= cont; i++) {
2217                 int l = 0x80, h = 0xBF;
2218
2219                 /* adjust limit for second byte */
2220                 if (i == 1) {
2221                     switch (c) {
2222                     case 0xE0: l = 0xA0; break;
2223                     case 0xED: h = 0x9F; break;
2224                     case 0xF0: l = 0x90; break;
2225                     case 0xF4: h = 0x8F; break;
2226                     }
2227                 }
2228
2229                 if (p[i] < l || p[i] > h) {
2230                     skip = i; goto invalid_utf8_sequence;
2231                 }
2232
2233                 n = (n << 6) | (p[i] & 0x3f);
2234             }
2235
2236             /* advance pointer */
2237             p += 1 + cont;
2238             c = n;
2239             goto add_char_nonext;
2240
2241             /* error handling */
2242         invalid_utf8_sequence:
2243             tcc_warning("ill-formed UTF-8 subsequence starting with: \'\\x%x\'", c);
2244             c = 0xFFFD;
2245             p += skip;
2246             goto add_char_nonext;
2247
2248         }
2249         p++;
2250     add_char_nonext:
2251         if (!is_long)
2252             cstr_ccat(outstr, c);
2253         else {
2254 #ifdef TCC_TARGET_PE
2255             /* store as UTF-16 */
2256             if (c < 0x10000) {
2257                 cstr_wccat(outstr, c);
2258             } else {
2259                 c -= 0x10000;
2260                 cstr_wccat(outstr, (c >> 10) + 0xD800);
2261                 cstr_wccat(outstr, (c & 0x3FF) + 0xDC00);
2262             }
2263 #else
2264             cstr_wccat(outstr, c);
2265 #endif
2266         }
2267     }
2268     /* add a trailing '\0' */
2269     if (!is_long)
2270         cstr_ccat(outstr, '\0');
2271     else
2272         cstr_wccat(outstr, '\0');
2273 }
2274
2275 static void parse_string(const char *s, int len)
2276 {
2277     uint8_t buf[1000], *p = buf;
2278     int is_long, sep;
2279
2280     if ((is_long = *s == 'L'))
2281         ++s, --len;
2282     sep = *s++;
2283     len -= 2;
2284     if (len >= sizeof buf)
2285         p = tcc_malloc(len + 1);
2286     memcpy(p, s, len);
2287     p[len] = 0;
2288
2289     cstr_reset(&tokcstr);
2290     parse_escape_string(&tokcstr, p, is_long);
2291     if (p != buf)
2292         tcc_free(p);
2293
2294     if (sep == '\'') {
2295         int char_size, i, n, c;
2296         /* XXX: make it portable */
2297         if (!is_long)
2298             tok = TOK_CCHAR, char_size = 1;
2299         else
2300             tok = TOK_LCHAR, char_size = sizeof(nwchar_t);
2301         n = tokcstr.size / char_size - 1;
2302         if (n < 1)
2303             tcc_error("empty character constant");
2304         if (n > 1)
2305             tcc_warning_c(warn_all)("multi-character character constant");
2306         for (c = i = 0; i < n; ++i) {
2307             if (is_long)
2308                 c = ((nwchar_t *)tokcstr.data)[i];
2309             else
2310                 c = (c << 8) | ((char *)tokcstr.data)[i];
2311         }
2312         tokc.i = c;
2313     } else {
2314         tokc.str.size = tokcstr.size;
2315         tokc.str.data = tokcstr.data;
2316         if (!is_long)
2317             tok = TOK_STR;
2318         else
2319             tok = TOK_LSTR;
2320     }
2321 }
2322
2323 /* we use 64 bit numbers */
2324 #define BN_SIZE 2
2325
2326 /* bn = (bn << shift) | or_val */
2327 static void bn_lshift(unsigned int *bn, int shift, int or_val)
2328 {
2329     int i;
2330     unsigned int v;
2331     for(i=0;i<BN_SIZE;i++) {
2332         v = bn[i];
2333         bn[i] = (v << shift) | or_val;
2334         or_val = v >> (32 - shift);
2335     }
2336 }
2337
2338 static void bn_zero(unsigned int *bn)
2339 {
2340     int i;
2341     for(i=0;i<BN_SIZE;i++) {
2342         bn[i] = 0;
2343     }
2344 }
2345
2346 /* parse number in null terminated string 'p' and return it in the
2347    current token */
2348 static void parse_number(const char *p)
2349 {
2350     int b, t, shift, frac_bits, s, exp_val, ch;
2351     char *q;
2352     unsigned int bn[BN_SIZE];
2353     double d;
2354
2355     /* number */
2356     q = token_buf;
2357     ch = *p++;
2358     t = ch;
2359     ch = *p++;
2360     *q++ = t;
2361     b = 10;
2362     if (t == '.') {
2363         goto float_frac_parse;
2364     } else if (t == '0') {
2365         if (ch == 'x' || ch == 'X') {
2366             q--;
2367             ch = *p++;
2368             b = 16;
2369         } else if (tcc_state->tcc_ext && (ch == 'b' || ch == 'B')) {
2370             q--;
2371             ch = *p++;
2372             b = 2;
2373         }
2374     }
2375     /* parse all digits. cannot check octal numbers at this stage
2376        because of floating point constants */
2377     while (1) {
2378         if (ch >= 'a' && ch <= 'f')
2379             t = ch - 'a' + 10;
2380         else if (ch >= 'A' && ch <= 'F')
2381             t = ch - 'A' + 10;
2382         else if (isnum(ch))
2383             t = ch - '0';
2384         else
2385             break;
2386         if (t >= b)
2387             break;
2388         if (q >= token_buf + STRING_MAX_SIZE) {
2389         num_too_long:
2390             tcc_error("number too long");
2391         }
2392         *q++ = ch;
2393         ch = *p++;
2394     }
2395     if (ch == '.' ||
2396         ((ch == 'e' || ch == 'E') && b == 10) ||
2397         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
2398         if (b != 10) {
2399             /* NOTE: strtox should support that for hexa numbers, but
2400                non ISOC99 libcs do not support it, so we prefer to do
2401                it by hand */
2402             /* hexadecimal or binary floats */
2403             /* XXX: handle overflows */
2404             *q = '\0';
2405             if (b == 16)
2406                 shift = 4;
2407             else
2408                 shift = 1;
2409             bn_zero(bn);
2410             q = token_buf;
2411             while (1) {
2412                 t = *q++;
2413                 if (t == '\0') {
2414                     break;
2415                 } else if (t >= 'a') {
2416                     t = t - 'a' + 10;
2417                 } else if (t >= 'A') {
2418                     t = t - 'A' + 10;
2419                 } else {
2420                     t = t - '0';
2421                 }
2422                 bn_lshift(bn, shift, t);
2423             }
2424             frac_bits = 0;
2425             if (ch == '.') {
2426                 ch = *p++;
2427                 while (1) {
2428                     t = ch;
2429                     if (t >= 'a' && t <= 'f') {
2430                         t = t - 'a' + 10;
2431                     } else if (t >= 'A' && t <= 'F') {
2432                         t = t - 'A' + 10;
2433                     } else if (t >= '0' && t <= '9') {
2434                         t = t - '0';
2435                     } else {
2436                         break;
2437                     }
2438                     if (t >= b)
2439                         tcc_error("invalid digit");
2440                     bn_lshift(bn, shift, t);
2441                     frac_bits += shift;
2442                     ch = *p++;
2443                 }
2444             }
2445             if (ch != 'p' && ch != 'P')
2446                 expect("exponent");
2447             ch = *p++;
2448             s = 1;
2449             exp_val = 0;
2450             if (ch == '+') {
2451                 ch = *p++;
2452             } else if (ch == '-') {
2453                 s = -1;
2454                 ch = *p++;
2455             }
2456             if (ch < '0' || ch > '9')
2457                 expect("exponent digits");
2458             while (ch >= '0' && ch <= '9') {
2459                 exp_val = exp_val * 10 + ch - '0';
2460                 ch = *p++;
2461             }
2462             exp_val = exp_val * s;
2463
2464             /* now we can generate the number */
2465             /* XXX: should patch directly float number */
2466             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
2467             d = ldexp(d, exp_val - frac_bits);
2468             t = toup(ch);
2469             if (t == 'F') {
2470                 ch = *p++;
2471                 tok = TOK_CFLOAT;
2472                 /* float : should handle overflow */
2473                 tokc.f = (float)d;
2474             } else if (t == 'L') {
2475                 ch = *p++;
2476 #ifdef TCC_USING_DOUBLE_FOR_LDOUBLE
2477                 tok = TOK_CDOUBLE;
2478                 tokc.d = d;
2479 #else
2480                 tok = TOK_CLDOUBLE;
2481                 /* XXX: not large enough */
2482                 tokc.ld = (long double)d;
2483 #endif
2484             } else {
2485                 tok = TOK_CDOUBLE;
2486                 tokc.d = d;
2487             }
2488         } else {
2489             /* decimal floats */
2490             if (ch == '.') {
2491                 if (q >= token_buf + STRING_MAX_SIZE)
2492                     goto num_too_long;
2493                 *q++ = ch;
2494                 ch = *p++;
2495             float_frac_parse:
2496                 while (ch >= '0' && ch <= '9') {
2497                     if (q >= token_buf + STRING_MAX_SIZE)
2498                         goto num_too_long;
2499                     *q++ = ch;
2500                     ch = *p++;
2501                 }
2502             }
2503             if (ch == 'e' || ch == 'E') {
2504                 if (q >= token_buf + STRING_MAX_SIZE)
2505                     goto num_too_long;
2506                 *q++ = ch;
2507                 ch = *p++;
2508                 if (ch == '-' || ch == '+') {
2509                     if (q >= token_buf + STRING_MAX_SIZE)
2510                         goto num_too_long;
2511                     *q++ = ch;
2512                     ch = *p++;
2513                 }
2514                 if (ch < '0' || ch > '9')
2515                     expect("exponent digits");
2516                 while (ch >= '0' && ch <= '9') {
2517                     if (q >= token_buf + STRING_MAX_SIZE)
2518                         goto num_too_long;
2519                     *q++ = ch;
2520                     ch = *p++;
2521                 }
2522             }
2523             *q = '\0';
2524             t = toup(ch);
2525             errno = 0;
2526             if (t == 'F') {
2527                 ch = *p++;
2528                 tok = TOK_CFLOAT;
2529                 tokc.f = strtof(token_buf, NULL);
2530             } else if (t == 'L') {
2531                 ch = *p++;
2532 #ifdef TCC_USING_DOUBLE_FOR_LDOUBLE
2533                 tok = TOK_CDOUBLE;
2534                 tokc.d = strtod(token_buf, NULL);
2535 #else
2536                 tok = TOK_CLDOUBLE;
2537                 tokc.ld = strtold(token_buf, NULL);
2538 #endif
2539             } else {
2540                 tok = TOK_CDOUBLE;
2541                 tokc.d = strtod(token_buf, NULL);
2542             }
2543         }
2544     } else {
2545         unsigned long long n, n1;
2546         int lcount, ucount, ov = 0;
2547         const char *p1;
2548
2549         /* integer number */
2550         *q = '\0';
2551         q = token_buf;
2552         if (b == 10 && *q == '0') {
2553             b = 8;
2554             q++;
2555         }
2556         n = 0;
2557         while(1) {
2558             t = *q++;
2559             /* no need for checks except for base 10 / 8 errors */
2560             if (t == '\0')
2561                 break;
2562             else if (t >= 'a')
2563                 t = t - 'a' + 10;
2564             else if (t >= 'A')
2565                 t = t - 'A' + 10;
2566             else
2567                 t = t - '0';
2568             if (t >= b)
2569                 tcc_error("invalid digit");
2570             n1 = n;
2571             n = n * b + t;
2572             /* detect overflow */
2573             if (n1 >= 0x1000000000000000ULL && n / b != n1)
2574                 ov = 1;
2575         }
2576
2577         /* Determine the characteristics (unsigned and/or 64bit) the type of
2578            the constant must have according to the constant suffix(es) */
2579         lcount = ucount = 0;
2580         p1 = p;
2581         for(;;) {
2582             t = toup(ch);
2583             if (t == 'L') {
2584                 if (lcount >= 2)
2585                     tcc_error("three 'l's in integer constant");
2586                 if (lcount && *(p - 1) != ch)
2587                     tcc_error("incorrect integer suffix: %s", p1);
2588                 lcount++;
2589                 ch = *p++;
2590             } else if (t == 'U') {
2591                 if (ucount >= 1)
2592                     tcc_error("two 'u's in integer constant");
2593                 ucount++;
2594                 ch = *p++;
2595             } else {
2596                 break;
2597             }
2598         }
2599
2600         /* Determine if it needs 64 bits and/or unsigned in order to fit */
2601         if (ucount == 0 && b == 10) {
2602             if (lcount <= (LONG_SIZE == 4)) {
2603                 if (n >= 0x80000000U)
2604                     lcount = (LONG_SIZE == 4) + 1;
2605             }
2606             if (n >= 0x8000000000000000ULL)
2607                 ov = 1, ucount = 1;
2608         } else {
2609             if (lcount <= (LONG_SIZE == 4)) {
2610                 if (n >= 0x100000000ULL)
2611                     lcount = (LONG_SIZE == 4) + 1;
2612                 else if (n >= 0x80000000U)
2613                     ucount = 1;
2614             }
2615             if (n >= 0x8000000000000000ULL)
2616                 ucount = 1;
2617         }
2618
2619         if (ov)
2620             tcc_warning("integer constant overflow");
2621
2622         tok = TOK_CINT;
2623         if (lcount) {
2624             tok = TOK_CLONG;
2625             if (lcount == 2)
2626                 tok = TOK_CLLONG;
2627         }
2628         if (ucount)
2629             ++tok; /* TOK_CU... */
2630         tokc.i = n;
2631     }
2632     if (ch)
2633         tcc_error("invalid number");
2634 }
2635
2636
2637 #define PARSE2(c1, tok1, c2, tok2)              \
2638     case c1:                                    \
2639         PEEKC(c, p);                            \
2640         if (c == c2) {                          \
2641             p++;                                \
2642             tok = tok2;                         \
2643         } else {                                \
2644             tok = tok1;                         \
2645         }                                       \
2646         break;
2647
2648 /* return next token without macro substitution */
2649 static inline void next_nomacro1(void)
2650 {
2651     int t, c, is_long, len;
2652     TokenSym *ts;
2653     uint8_t *p, *p1;
2654     unsigned int h;
2655
2656     p = file->buf_ptr;
2657  redo_no_start:
2658     c = *p;
2659     switch(c) {
2660     case ' ':
2661     case '\t':
2662         tok = c;
2663         p++;
2664  maybe_space:
2665         if (parse_flags & PARSE_FLAG_SPACES)
2666             goto keep_tok_flags;
2667         while (isidnum_table[*p - CH_EOF] & IS_SPC)
2668             ++p;
2669         goto redo_no_start;
2670     case '\f':
2671     case '\v':
2672     case '\r':
2673         p++;
2674         goto redo_no_start;
2675     case '\\':
2676         /* first look if it is in fact an end of buffer */
2677         c = handle_stray1(p);
2678         p = file->buf_ptr;
2679         if (c == '\\')
2680             goto parse_simple;
2681         if (c != CH_EOF)
2682             goto redo_no_start;
2683         {
2684             TCCState *s1 = tcc_state;
2685             if ((parse_flags & PARSE_FLAG_LINEFEED)
2686                 && !(tok_flags & TOK_FLAG_EOF)) {
2687                 tok_flags |= TOK_FLAG_EOF;
2688                 tok = TOK_LINEFEED;
2689                 goto keep_tok_flags;
2690             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2691                 tok = TOK_EOF;
2692             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2693                 tcc_error("missing #endif");
2694             } else if (s1->include_stack_ptr == s1->include_stack) {
2695                 /* no include left : end of file. */
2696                 tok = TOK_EOF;
2697             } else {
2698                 tok_flags &= ~TOK_FLAG_EOF;
2699                 /* pop include file */
2700
2701                 /* test if previous '#endif' was after a #ifdef at
2702                    start of file */
2703                 if (tok_flags & TOK_FLAG_ENDIF) {
2704 #ifdef INC_DEBUG
2705                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2706 #endif
2707                     search_cached_include(s1, file->filename, 1)
2708                         ->ifndef_macro = file->ifndef_macro_saved;
2709                     tok_flags &= ~TOK_FLAG_ENDIF;
2710                 }
2711
2712                 /* add end of include file debug info */
2713                 tcc_debug_eincl(tcc_state);
2714                 /* pop include stack */
2715                 tcc_close();
2716                 s1->include_stack_ptr--;
2717                 p = file->buf_ptr;
2718                 if (p == file->buffer)
2719                     tok_flags = TOK_FLAG_BOF|TOK_FLAG_BOL;
2720                 goto redo_no_start;
2721             }
2722         }
2723         break;
2724
2725     case '\n':
2726         file->line_num++;
2727         tok_flags |= TOK_FLAG_BOL;
2728         p++;
2729 maybe_newline:
2730         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2731             goto redo_no_start;
2732         tok = TOK_LINEFEED;
2733         goto keep_tok_flags;
2734
2735     case '#':
2736         /* XXX: simplify */
2737         PEEKC(c, p);
2738         if ((tok_flags & TOK_FLAG_BOL) &&
2739             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2740             file->buf_ptr = p;
2741             preprocess(tok_flags & TOK_FLAG_BOF);
2742             p = file->buf_ptr;
2743             goto maybe_newline;
2744         } else {
2745             if (c == '#') {
2746                 p++;
2747                 tok = TOK_TWOSHARPS;
2748             } else {
2749 #if !defined(TCC_TARGET_ARM)
2750                 if (parse_flags & PARSE_FLAG_ASM_FILE) {
2751                     p = parse_line_comment(p - 1);
2752                     goto redo_no_start;
2753                 } else
2754 #endif
2755                 {
2756                     tok = '#';
2757                 }
2758             }
2759         }
2760         break;
2761
2762     /* dollar is allowed to start identifiers when not parsing asm */
2763     case '$':
2764         if (!(isidnum_table[c - CH_EOF] & IS_ID)
2765          || (parse_flags & PARSE_FLAG_ASM_FILE))
2766             goto parse_simple;
2767
2768     case 'a': case 'b': case 'c': case 'd':
2769     case 'e': case 'f': case 'g': case 'h':
2770     case 'i': case 'j': case 'k': case 'l':
2771     case 'm': case 'n': case 'o': case 'p':
2772     case 'q': case 'r': case 's': case 't':
2773     case 'u': case 'v': case 'w': case 'x':
2774     case 'y': case 'z':
2775     case 'A': case 'B': case 'C': case 'D':
2776     case 'E': case 'F': case 'G': case 'H':
2777     case 'I': case 'J': case 'K':
2778     case 'M': case 'N': case 'O': case 'P':
2779     case 'Q': case 'R': case 'S': case 'T':
2780     case 'U': case 'V': case 'W': case 'X':
2781     case 'Y': case 'Z':
2782     case '_':
2783     parse_ident_fast:
2784         p1 = p;
2785         h = TOK_HASH_INIT;
2786         h = TOK_HASH_FUNC(h, c);
2787         while (c = *++p, isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2788             h = TOK_HASH_FUNC(h, c);
2789         len = p - p1;
2790         if (c != '\\') {
2791             TokenSym **pts;
2792
2793             /* fast case : no stray found, so we have the full token
2794                and we have already hashed it */
2795             h &= (TOK_HASH_SIZE - 1);
2796             pts = &hash_ident[h];
2797             for(;;) {
2798                 ts = *pts;
2799                 if (!ts)
2800                     break;
2801                 if (ts->len == len && !memcmp(ts->str, p1, len))
2802                     goto token_found;
2803                 pts = &(ts->hash_next);
2804             }
2805             ts = tok_alloc_new(pts, (char *) p1, len);
2806         token_found: ;
2807         } else {
2808             /* slower case */
2809             cstr_reset(&tokcstr);
2810             cstr_cat(&tokcstr, (char *) p1, len);
2811             p--;
2812             PEEKC(c, p);
2813         parse_ident_slow:
2814             while (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2815             {
2816                 cstr_ccat(&tokcstr, c);
2817                 PEEKC(c, p);
2818             }
2819             ts = tok_alloc(tokcstr.data, tokcstr.size);
2820         }
2821         tok = ts->tok;
2822         break;
2823     case 'L':
2824         t = p[1];
2825         if (t != '\\' && t != '\'' && t != '\"') {
2826             /* fast case */
2827             goto parse_ident_fast;
2828         } else {
2829             PEEKC(c, p);
2830             if (c == '\'' || c == '\"') {
2831                 is_long = 1;
2832                 goto str_const;
2833             } else {
2834                 cstr_reset(&tokcstr);
2835                 cstr_ccat(&tokcstr, 'L');
2836                 goto parse_ident_slow;
2837             }
2838         }
2839         break;
2840
2841     case '0': case '1': case '2': case '3':
2842     case '4': case '5': case '6': case '7':
2843     case '8': case '9':
2844         t = c;
2845         PEEKC(c, p);
2846         /* after the first digit, accept digits, alpha, '.' or sign if
2847            prefixed by 'eEpP' */
2848     parse_num:
2849         cstr_reset(&tokcstr);
2850         for(;;) {
2851             cstr_ccat(&tokcstr, t);
2852             if (!((isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2853                   || c == '.'
2854                   || ((c == '+' || c == '-')
2855                       && (((t == 'e' || t == 'E')
2856                             && !(parse_flags & PARSE_FLAG_ASM_FILE
2857                                 /* 0xe+1 is 3 tokens in asm */
2858                                 && ((char*)tokcstr.data)[0] == '0'
2859                                 && toup(((char*)tokcstr.data)[1]) == 'X'))
2860                           || t == 'p' || t == 'P'))))
2861                 break;
2862             t = c;
2863             PEEKC(c, p);
2864         }
2865         /* We add a trailing '\0' to ease parsing */
2866         cstr_ccat(&tokcstr, '\0');
2867         tokc.str.size = tokcstr.size;
2868         tokc.str.data = tokcstr.data;
2869         tok = TOK_PPNUM;
2870         break;
2871
2872     case '.':
2873         /* special dot handling because it can also start a number */
2874         PEEKC(c, p);
2875         if (isnum(c)) {
2876             t = '.';
2877             goto parse_num;
2878         } else if ((isidnum_table['.' - CH_EOF] & IS_ID)
2879                    && (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))) {
2880             *--p = c = '.';
2881             goto parse_ident_fast;
2882         } else if (c == '.') {
2883             PEEKC(c, p);
2884             if (c == '.') {
2885                 p++;
2886                 tok = TOK_DOTS;
2887             } else {
2888                 *--p = '.'; /* may underflow into file->unget[] */
2889                 tok = '.';
2890             }
2891         } else {
2892             tok = '.';
2893         }
2894         break;
2895     case '\'':
2896     case '\"':
2897         is_long = 0;
2898     str_const:
2899         cstr_reset(&tokcstr);
2900         if (is_long)
2901             cstr_ccat(&tokcstr, 'L');
2902         cstr_ccat(&tokcstr, c);
2903         p = parse_pp_string(p, c, &tokcstr);
2904         cstr_ccat(&tokcstr, c);
2905         cstr_ccat(&tokcstr, '\0');
2906         tokc.str.size = tokcstr.size;
2907         tokc.str.data = tokcstr.data;
2908         tok = TOK_PPSTR;
2909         break;
2910
2911     case '<':
2912         PEEKC(c, p);
2913         if (c == '=') {
2914             p++;
2915             tok = TOK_LE;
2916         } else if (c == '<') {
2917             PEEKC(c, p);
2918             if (c == '=') {
2919                 p++;
2920                 tok = TOK_A_SHL;
2921             } else {
2922                 tok = TOK_SHL;
2923             }
2924         } else {
2925             tok = TOK_LT;
2926         }
2927         break;
2928     case '>':
2929         PEEKC(c, p);
2930         if (c == '=') {
2931             p++;
2932             tok = TOK_GE;
2933         } else if (c == '>') {
2934             PEEKC(c, p);
2935             if (c == '=') {
2936                 p++;
2937                 tok = TOK_A_SAR;
2938             } else {
2939                 tok = TOK_SAR;
2940             }
2941         } else {
2942             tok = TOK_GT;
2943         }
2944         break;
2945
2946     case '&':
2947         PEEKC(c, p);
2948         if (c == '&') {
2949             p++;
2950             tok = TOK_LAND;
2951         } else if (c == '=') {
2952             p++;
2953             tok = TOK_A_AND;
2954         } else {
2955             tok = '&';
2956         }
2957         break;
2958
2959     case '|':
2960         PEEKC(c, p);
2961         if (c == '|') {
2962             p++;
2963             tok = TOK_LOR;
2964         } else if (c == '=') {
2965             p++;
2966             tok = TOK_A_OR;
2967         } else {
2968             tok = '|';
2969         }
2970         break;
2971
2972     case '+':
2973         PEEKC(c, p);
2974         if (c == '+') {
2975             p++;
2976             tok = TOK_INC;
2977         } else if (c == '=') {
2978             p++;
2979             tok = TOK_A_ADD;
2980         } else {
2981             tok = '+';
2982         }
2983         break;
2984
2985     case '-':
2986         PEEKC(c, p);
2987         if (c == '-') {
2988             p++;
2989             tok = TOK_DEC;
2990         } else if (c == '=') {
2991             p++;
2992             tok = TOK_A_SUB;
2993         } else if (c == '>') {
2994             p++;
2995             tok = TOK_ARROW;
2996         } else {
2997             tok = '-';
2998         }
2999         break;
3000
3001     PARSE2('!', '!', '=', TOK_NE)
3002     PARSE2('=', '=', '=', TOK_EQ)
3003     PARSE2('*', '*', '=', TOK_A_MUL)
3004     PARSE2('%', '%', '=', TOK_A_MOD)
3005     PARSE2('^', '^', '=', TOK_A_XOR)
3006
3007         /* comments or operator */
3008     case '/':
3009         PEEKC(c, p);
3010         if (c == '*') {
3011             p = parse_comment(p);
3012             /* comments replaced by a blank */
3013             tok = ' ';
3014             goto maybe_space;
3015         } else if (c == '/') {
3016             p = parse_line_comment(p);
3017             tok = ' ';
3018             goto maybe_space;
3019         } else if (c == '=') {
3020             p++;
3021             tok = TOK_A_DIV;
3022         } else {
3023             tok = '/';
3024         }
3025         break;
3026
3027         /* simple tokens */
3028     case '(':
3029     case ')':
3030     case '[':
3031     case ']':
3032     case '{':
3033     case '}':
3034     case ',':
3035     case ';':
3036     case ':':
3037     case '?':
3038     case '~':
3039     case '@': /* only used in assembler */
3040     parse_simple:
3041         tok = c;
3042         p++;
3043         break;
3044     default:
3045         if (c >= 0x80 && c <= 0xFF) /* utf8 identifiers */
3046             goto parse_ident_fast;
3047         if (parse_flags & PARSE_FLAG_ASM_FILE)
3048             goto parse_simple;
3049         tcc_error("unrecognized character \\x%02x", c);
3050         break;
3051     }
3052     tok_flags = 0;
3053 keep_tok_flags:
3054     file->buf_ptr = p;
3055 #if defined(PARSE_DEBUG)
3056     printf("token = %d %s\n", tok, get_tok_str(tok, &tokc));
3057 #endif
3058 }
3059
3060 static void macro_subst(
3061     TokenString *tok_str,
3062     Sym **nested_list,
3063     const int *macro_str
3064     );
3065
3066 /* substitute arguments in replacement lists in macro_str by the values in
3067    args (field d) and return allocated string */
3068 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
3069 {
3070     int t, t0, t1, spc;
3071     const int *st;
3072     Sym *s;
3073     CValue cval;
3074     TokenString str;
3075     CString cstr;
3076
3077     tok_str_new(&str);
3078     t0 = t1 = 0;
3079     while(1) {
3080         TOK_GET(&t, &macro_str, &cval);
3081         if (!t)
3082             break;
3083         if (t == '#') {
3084             /* stringize */
3085             TOK_GET(&t, &macro_str, &cval);
3086             if (!t)
3087                 goto bad_stringy;
3088             s = sym_find2(args, t);
3089             if (s) {
3090                 cstr_new(&cstr);
3091                 cstr_ccat(&cstr, '\"');
3092                 st = s->d;
3093                 spc = 0;
3094                 while (*st >= 0) {
3095                     TOK_GET(&t, &st, &cval);
3096                     if (t != TOK_PLCHLDR
3097                      && t != TOK_NOSUBST
3098                      && 0 == check_space(t, &spc)) {
3099                         const char *s = get_tok_str(t, &cval);
3100                         while (*s) {
3101                             if (t == TOK_PPSTR && *s != '\'')
3102                                 add_char(&cstr, *s);
3103                             else
3104                                 cstr_ccat(&cstr, *s);
3105                             ++s;
3106                         }
3107                     }
3108                 }
3109                 cstr.size -= spc;
3110                 cstr_ccat(&cstr, '\"');
3111                 cstr_ccat(&cstr, '\0');
3112 #ifdef PP_DEBUG
3113                 printf("\nstringize: <%s>\n", (char *)cstr.data);
3114 #endif
3115                 /* add string */
3116                 cval.str.size = cstr.size;
3117                 cval.str.data = cstr.data;
3118                 tok_str_add2(&str, TOK_PPSTR, &cval);
3119                 cstr_free(&cstr);
3120             } else {
3121         bad_stringy:
3122                 expect("macro parameter after '#'");
3123             }
3124         } else if (t >= TOK_IDENT) {
3125             s = sym_find2(args, t);
3126             if (s) {
3127                 int l0 = str.len;
3128                 st = s->d;
3129                 /* if '##' is present before or after, no arg substitution */
3130                 if (*macro_str == TOK_PPJOIN || t1 == TOK_PPJOIN) {
3131                     /* special case for var arg macros : ## eats the ','
3132                        if empty VA_ARGS variable. */
3133                     if (t1 == TOK_PPJOIN && t0 == ',' && gnu_ext && s->type.t) {
3134                         if (*st <= 0) {
3135                             /* suppress ',' '##' */
3136                             str.len -= 2;
3137                         } else {
3138                             /* suppress '##' and add variable */
3139                             str.len--;
3140                             goto add_var;
3141                         }
3142                     }
3143                 } else {
3144             add_var:
3145                     if (!s->next) {
3146                         /* Expand arguments tokens and store them.  In most
3147                            cases we could also re-expand each argument if
3148                            used multiple times, but not if the argument
3149                            contains the __COUNTER__ macro.  */
3150                         TokenString str2;
3151                         sym_push2(&s->next, s->v, s->type.t, 0);
3152                         tok_str_new(&str2);
3153                         macro_subst(&str2, nested_list, st);
3154                         tok_str_add(&str2, 0);
3155                         s->next->d = str2.str;
3156                     }
3157                     st = s->next->d;
3158                 }
3159                 for(;;) {
3160                     int t2;
3161                     TOK_GET(&t2, &st, &cval);
3162                     if (t2 <= 0)
3163                         break;
3164                     tok_str_add2(&str, t2, &cval);
3165                 }
3166                 if (str.len == l0) /* expanded to empty string */
3167                     tok_str_add(&str, TOK_PLCHLDR);
3168             } else {
3169                 tok_str_add(&str, t);
3170             }
3171         } else {
3172             tok_str_add2(&str, t, &cval);
3173         }
3174         t0 = t1, t1 = t;
3175     }
3176     tok_str_add(&str, 0);
3177     return str.str;
3178 }
3179
3180 static char const ab_month_name[12][4] =
3181 {
3182     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
3183     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
3184 };
3185
3186 static int paste_tokens(int t1, CValue *v1, int t2, CValue *v2)
3187 {
3188     CString cstr;
3189     int n, ret = 1;
3190
3191     cstr_new(&cstr);
3192     if (t1 != TOK_PLCHLDR)
3193         cstr_cat(&cstr, get_tok_str(t1, v1), -1);
3194     n = cstr.size;
3195     if (t2 != TOK_PLCHLDR)
3196         cstr_cat(&cstr, get_tok_str(t2, v2), -1);
3197     cstr_ccat(&cstr, '\0');
3198
3199     tcc_open_bf(tcc_state, ":paste:", cstr.size);
3200     memcpy(file->buffer, cstr.data, cstr.size);
3201     tok_flags = 0;
3202     for (;;) {
3203         next_nomacro1();
3204         if (0 == *file->buf_ptr)
3205             break;
3206         if (is_space(tok))
3207             continue;
3208         tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid"
3209             " preprocessing token", n, (char *)cstr.data, (char*)cstr.data + n);
3210         ret = 0;
3211         break;
3212     }
3213     tcc_close();
3214     //printf("paste <%s>\n", (char*)cstr.data);
3215     cstr_free(&cstr);
3216     return ret;
3217 }
3218
3219 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
3220    return the resulting string (which must be freed). */
3221 static inline int *macro_twosharps(const int *ptr0)
3222 {
3223     int t;
3224     CValue cval;
3225     TokenString macro_str1;
3226     int start_of_nosubsts = -1;
3227     const int *ptr;
3228
3229     /* we search the first '##' */
3230     for (ptr = ptr0;;) {
3231         TOK_GET(&t, &ptr, &cval);
3232         if (t == TOK_PPJOIN)
3233             break;
3234         if (t == 0)
3235             return NULL;
3236     }
3237
3238     tok_str_new(&macro_str1);
3239
3240     //tok_print(" $$$", ptr0);
3241     for (ptr = ptr0;;) {
3242         TOK_GET(&t, &ptr, &cval);
3243         if (t == 0)
3244             break;
3245         if (t == TOK_PPJOIN)
3246             continue;
3247         while (*ptr == TOK_PPJOIN) {
3248             int t1; CValue cv1;
3249             /* given 'a##b', remove nosubsts preceding 'a' */
3250             if (start_of_nosubsts >= 0)
3251                 macro_str1.len = start_of_nosubsts;
3252             /* given 'a##b', remove nosubsts preceding 'b' */
3253             while ((t1 = *++ptr) == TOK_NOSUBST)
3254                 ;
3255             if (t1 && t1 != TOK_PPJOIN) {
3256                 TOK_GET(&t1, &ptr, &cv1);
3257                 if (t != TOK_PLCHLDR || t1 != TOK_PLCHLDR) {
3258                     if (paste_tokens(t, &cval, t1, &cv1)) {
3259                         t = tok, cval = tokc;
3260                     } else {
3261                         tok_str_add2(&macro_str1, t, &cval);
3262                         t = t1, cval = cv1;
3263                     }
3264                 }
3265             }
3266         }
3267         if (t == TOK_NOSUBST) {
3268             if (start_of_nosubsts < 0)
3269                 start_of_nosubsts = macro_str1.len;
3270         } else {
3271             start_of_nosubsts = -1;
3272         }
3273         tok_str_add2(&macro_str1, t, &cval);
3274     }
3275     tok_str_add(&macro_str1, 0);
3276     //tok_print(" ###", macro_str1.str);
3277     return macro_str1.str;
3278 }
3279
3280 /* peek or read [ws_str == NULL] next token from function macro call,
3281    walking up macro levels up to the file if necessary */
3282 static int next_argstream(Sym **nested_list, TokenString *ws_str)
3283 {
3284     int t;
3285     const int *p;
3286     Sym *sa;
3287
3288     for (;;) {
3289         if (macro_ptr) {
3290             p = macro_ptr, t = *p;
3291             if (ws_str) {
3292                 while (is_space(t) || TOK_LINEFEED == t || TOK_PLCHLDR == t)
3293                     tok_str_add(ws_str, t), t = *++p;
3294             }
3295             if (t == 0) {
3296                 end_macro();
3297                 /* also, end of scope for nested defined symbol */
3298                 sa = *nested_list;
3299                 while (sa && sa->v == 0)
3300                     sa = sa->prev;
3301                 if (sa)
3302                     sa->v = 0;
3303                 continue;
3304             }
3305         } else {
3306             ch = handle_eob();
3307             if (ws_str) {
3308                 while (is_space(ch) || ch == '\n' || ch == '/') {
3309                     if (ch == '/') {
3310                         int c;
3311                         uint8_t *p = file->buf_ptr;
3312                         PEEKC(c, p);
3313                         if (c == '*') {
3314                             p = parse_comment(p);
3315                             file->buf_ptr = p - 1;
3316                         } else if (c == '/') {
3317                             p = parse_line_comment(p);
3318                             file->buf_ptr = p - 1;
3319                         } else
3320                             break;
3321                         ch = ' ';
3322                     }
3323                     if (ch == '\n')
3324                         file->line_num++;
3325                     if (!(ch == '\f' || ch == '\v' || ch == '\r'))
3326                         tok_str_add(ws_str, ch);
3327                     cinp();
3328                 }
3329             }
3330             t = ch;
3331         }
3332
3333         if (ws_str)
3334             return t;
3335         next_nomacro();
3336         return tok;
3337     }
3338 }
3339
3340 /* do macro substitution of current token with macro 's' and add
3341    result to (tok_str,tok_len). 'nested_list' is the list of all
3342    macros we got inside to avoid recursing. Return non zero if no
3343    substitution needs to be done */
3344 static int macro_subst_tok(
3345     TokenString *tok_str,
3346     Sym **nested_list,
3347     Sym *s)
3348 {
3349     Sym *args, *sa, *sa1;
3350     int parlevel, t, t1, spc;
3351     TokenString str;
3352     char *cstrval;
3353     CValue cval;
3354     CString cstr;
3355     char buf[32];
3356
3357     /* if symbol is a macro, prepare substitution */
3358     /* special macros */
3359     if (tok == TOK___LINE__ || tok == TOK___COUNTER__) {
3360         t = tok == TOK___LINE__ ? file->line_num : pp_counter++;
3361         snprintf(buf, sizeof(buf), "%d", t);
3362         cstrval = buf;
3363         t1 = TOK_PPNUM;
3364         goto add_cstr1;
3365     } else if (tok == TOK___FILE__) {
3366         cstrval = file->filename;
3367         goto add_cstr;
3368     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
3369         time_t ti;
3370         struct tm *tm;
3371
3372         time(&ti);
3373         tm = localtime(&ti);
3374         if (tok == TOK___DATE__) {
3375             snprintf(buf, sizeof(buf), "%s %2d %d",
3376                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
3377         } else {
3378             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
3379                      tm->tm_hour, tm->tm_min, tm->tm_sec);
3380         }
3381         cstrval = buf;
3382     add_cstr:
3383         t1 = TOK_STR;
3384     add_cstr1:
3385         cstr_new(&cstr);
3386         cstr_cat(&cstr, cstrval, 0);
3387         cval.str.size = cstr.size;
3388         cval.str.data = cstr.data;
3389         tok_str_add2(tok_str, t1, &cval);
3390         cstr_free(&cstr);
3391     } else if (s->d) {
3392         int saved_parse_flags = parse_flags;
3393         int *joined_str = NULL;
3394         int *mstr = s->d;
3395
3396         if (s->type.t == MACRO_FUNC) {
3397             /* whitespace between macro name and argument list */
3398             TokenString ws_str;
3399             tok_str_new(&ws_str);
3400
3401             spc = 0;
3402             parse_flags |= PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED
3403                 | PARSE_FLAG_ACCEPT_STRAYS;
3404
3405             /* get next token from argument stream */
3406             t = next_argstream(nested_list, &ws_str);
3407             if (t != '(') {
3408                 /* not a macro substitution after all, restore the
3409                  * macro token plus all whitespace we've read.
3410                  * whitespace is intentionally not merged to preserve
3411                  * newlines. */
3412                 parse_flags = saved_parse_flags;
3413                 tok_str_add(tok_str, tok);
3414                 if (parse_flags & PARSE_FLAG_SPACES) {
3415                     int i;
3416                     for (i = 0; i < ws_str.len; i++)
3417                         tok_str_add(tok_str, ws_str.str[i]);
3418                 }
3419                 tok_str_free_str(ws_str.str);
3420                 return 0;
3421             } else {
3422                 tok_str_free_str(ws_str.str);
3423             }
3424             do {
3425                 next_nomacro(); /* eat '(' */
3426             } while (tok == TOK_PLCHLDR || is_space(tok));
3427
3428             /* argument macro */
3429             args = NULL;
3430             sa = s->next;
3431             /* NOTE: empty args are allowed, except if no args */
3432             for(;;) {
3433                 do {
3434                     next_argstream(nested_list, NULL);
3435                 } while (tok == TOK_PLCHLDR || is_space(tok) ||
3436                          TOK_LINEFEED == tok);
3437     empty_arg:
3438                 /* handle '()' case */
3439                 if (!args && !sa && tok == ')')
3440                     break;
3441                 if (!sa)
3442                     tcc_error("macro '%s' used with too many args",
3443                           get_tok_str(s->v, 0));
3444                 tok_str_new(&str);
3445                 parlevel = spc = 0;
3446                 /* NOTE: non zero sa->t indicates VA_ARGS */
3447                 while ((parlevel > 0 ||
3448                         (tok != ')' &&
3449                          (tok != ',' || sa->type.t)))) {
3450                     if (tok == TOK_EOF || tok == 0)
3451                         break;
3452                     if (tok == '(')
3453                         parlevel++;
3454                     else if (tok == ')')
3455                         parlevel--;
3456                     if (tok == TOK_LINEFEED)
3457                         tok = ' ';
3458                     if (!check_space(tok, &spc))
3459                         tok_str_add2(&str, tok, &tokc);
3460                     next_argstream(nested_list, NULL);
3461                 }
3462                 if (parlevel)
3463                     expect(")");
3464                 str.len -= spc;
3465                 tok_str_add(&str, -1);
3466                 tok_str_add(&str, 0);
3467                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
3468                 sa1->d = str.str;
3469                 sa = sa->next;
3470                 if (tok == ')') {
3471                     /* special case for gcc var args: add an empty
3472                        var arg argument if it is omitted */
3473                     if (sa && sa->type.t && gnu_ext)
3474                         goto empty_arg;
3475                     break;
3476                 }
3477                 if (tok != ',')
3478                     expect(",");
3479             }
3480             if (sa) {
3481                 tcc_error("macro '%s' used with too few args",
3482                       get_tok_str(s->v, 0));
3483             }
3484
3485             /* now subst each arg */
3486             mstr = macro_arg_subst(nested_list, mstr, args);
3487             /* free memory */
3488             sa = args;
3489             while (sa) {
3490                 sa1 = sa->prev;
3491                 tok_str_free_str(sa->d);
3492                 if (sa->next) {
3493                     tok_str_free_str(sa->next->d);
3494                     sym_free(sa->next);
3495                 }
3496                 sym_free(sa);
3497                 sa = sa1;
3498             }
3499             parse_flags = saved_parse_flags;
3500         }
3501
3502         sym_push2(nested_list, s->v, 0, 0);
3503         parse_flags = saved_parse_flags;
3504         joined_str = macro_twosharps(mstr);
3505         macro_subst(tok_str, nested_list, joined_str ? joined_str : mstr);
3506
3507         /* pop nested defined symbol */
3508         sa1 = *nested_list;
3509         *nested_list = sa1->prev;
3510         sym_free(sa1);
3511         if (joined_str)
3512             tok_str_free_str(joined_str);
3513         if (mstr != s->d)
3514             tok_str_free_str(mstr);
3515     }
3516     return 0;
3517 }
3518
3519 /* do macro substitution of macro_str and add result to
3520    (tok_str,tok_len). 'nested_list' is the list of all macros we got
3521    inside to avoid recursing. */
3522 static void macro_subst(
3523     TokenString *tok_str,
3524     Sym **nested_list,
3525     const int *macro_str
3526     )
3527 {
3528     Sym *s;
3529     int t, spc, nosubst;
3530     CValue cval;
3531
3532     spc = nosubst = 0;
3533
3534     while (1) {
3535         TOK_GET(&t, &macro_str, &cval);
3536         if (t <= 0)
3537             break;
3538
3539         if (t >= TOK_IDENT && 0 == nosubst) {
3540             s = define_find(t);
3541             if (s == NULL)
3542                 goto no_subst;
3543
3544             /* if nested substitution, do nothing */
3545             if (sym_find2(*nested_list, t)) {
3546                 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
3547                 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3548                 goto no_subst;
3549             }
3550
3551             {
3552                 TokenString *str = tok_str_alloc();
3553                 str->str = (int*)macro_str;
3554                 begin_macro(str, 2);
3555
3556                 tok = t;
3557                 macro_subst_tok(tok_str, nested_list, s);
3558
3559                 if (macro_stack != str) {
3560                     /* already finished by reading function macro arguments */
3561                     break;
3562                 }
3563
3564                 macro_str = macro_ptr;
3565                 end_macro ();
3566             }
3567             if (tok_str->len)
3568                 spc = is_space(t = tok_str->str[tok_str->lastlen]);
3569         } else {
3570 no_subst:
3571             if (!check_space(t, &spc))
3572                 tok_str_add2(tok_str, t, &cval);
3573
3574             if (nosubst) {
3575                 if (nosubst > 1 && (spc || (++nosubst == 3 && t == '(')))
3576                     continue;
3577                 nosubst = 0;
3578             }
3579             if (t == TOK_NOSUBST)
3580                 nosubst = 1;
3581         }
3582         /* GCC supports 'defined' as result of a macro substitution */
3583         if (t == TOK_DEFINED && pp_expr)
3584             nosubst = 2;
3585     }
3586 }
3587
3588 /* return next token without macro substitution. Can read input from
3589    macro_ptr buffer */
3590 static void next_nomacro(void)
3591 {
3592     int t;
3593     if (macro_ptr) {
3594  redo:
3595         t = *macro_ptr;
3596         if (TOK_HAS_VALUE(t)) {
3597             tok_get(&tok, &macro_ptr, &tokc);
3598             if (t == TOK_LINENUM) {
3599                 file->line_num = tokc.i;
3600                 goto redo;
3601             }
3602         } else {
3603             macro_ptr++;
3604             if (t < TOK_IDENT) {
3605                 if (!(parse_flags & PARSE_FLAG_SPACES)
3606                     && (isidnum_table[t - CH_EOF] & IS_SPC))
3607                     goto redo;
3608             }
3609             tok = t;
3610         }
3611     } else {
3612         next_nomacro1();
3613     }
3614 }
3615
3616 /* return next token with macro substitution */
3617 ST_FUNC void next(void)
3618 {
3619     int t;
3620  redo:
3621     next_nomacro();
3622     t = tok;
3623     if (macro_ptr) {
3624         if (!TOK_HAS_VALUE(t)) {
3625             if (t == TOK_NOSUBST || t == TOK_PLCHLDR) {
3626                 /* discard preprocessor markers */
3627                 goto redo;
3628             } else if (t == 0) {
3629                 /* end of macro or unget token string */
3630                 end_macro();
3631                 goto redo;
3632             } else if (t == '\\') {
3633                 if (!(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
3634                     tcc_error("stray '\\' in program");
3635             }
3636             return;
3637         }
3638     } else if (t >= TOK_IDENT && (parse_flags & PARSE_FLAG_PREPROCESS)) {
3639         /* if reading from file, try to substitute macros */
3640         Sym *s = define_find(t);
3641         if (s) {
3642             Sym *nested_list = NULL;
3643             tokstr_buf.len = 0;
3644             macro_subst_tok(&tokstr_buf, &nested_list, s);
3645             tok_str_add(&tokstr_buf, 0);
3646             begin_macro(&tokstr_buf, 0);
3647             goto redo;
3648         }
3649         return;
3650     }
3651     /* convert preprocessor tokens into C tokens */
3652     if (t == TOK_PPNUM) {
3653         if  (parse_flags & PARSE_FLAG_TOK_NUM)
3654             parse_number((char *)tokc.str.data);
3655     } else if (t == TOK_PPSTR) {
3656         if (parse_flags & PARSE_FLAG_TOK_STR)
3657             parse_string((char *)tokc.str.data, tokc.str.size - 1);
3658     }
3659 }
3660
3661 /* push back current token and set current token to 'last_tok'. Only
3662    identifier case handled for labels. */
3663 ST_INLN void unget_tok(int last_tok)
3664 {
3665
3666     TokenString *str = tok_str_alloc();
3667     tok_str_add2(str, tok, &tokc);
3668     tok_str_add(str, 0);
3669     begin_macro(str, 1);
3670     tok = last_tok;
3671 }
3672
3673 /* ------------------------------------------------------------------------- */
3674 /* init preprocessor */
3675
3676 static const char * const target_os_defs =
3677 #ifdef TCC_TARGET_PE
3678     "_WIN32\0"
3679 # if PTR_SIZE == 8
3680     "_WIN64\0"
3681 # endif
3682 #else
3683 # if defined TCC_TARGET_MACHO
3684     "__APPLE__\0"
3685 # elif TARGETOS_FreeBSD
3686     "__FreeBSD__ 12\0"
3687 # elif TARGETOS_FreeBSD_kernel
3688     "__FreeBSD_kernel__\0"
3689 # elif TARGETOS_NetBSD
3690     "__NetBSD__\0"
3691 # elif TARGETOS_OpenBSD
3692     "__OpenBSD__\0"
3693 # else
3694     "__linux__\0"
3695     "__linux\0"
3696 # endif
3697     "__unix__\0"
3698     "__unix\0"
3699 #endif
3700     ;
3701
3702 static void putdef(CString *cs, const char *p)
3703 {
3704     cstr_printf(cs, "#define %s%s\n", p, &" 1"[!!strchr(p, ' ')*2]);
3705 }
3706
3707 static void tcc_predefs(TCCState *s1, CString *cs, int is_asm)
3708 {
3709     int a, b, c;
3710     const char *defs[] = { target_machine_defs, target_os_defs, NULL };
3711     const char *p;
3712
3713     sscanf(TCC_VERSION, "%d.%d.%d", &a, &b, &c);
3714     cstr_printf(cs, "#define __TINYC__ %d\n", a*10000 + b*100 + c);
3715     for (a = 0; defs[a]; ++a)
3716         for (p = defs[a]; *p; p = strchr(p, 0) + 1)
3717             putdef(cs, p);
3718 #ifdef TCC_TARGET_ARM
3719     if (s1->float_abi == ARM_HARD_FLOAT)
3720       putdef(cs, "__ARM_PCS_VFP");
3721 #endif
3722     if (is_asm)
3723       putdef(cs, "__ASSEMBLER__");
3724     if (s1->output_type == TCC_OUTPUT_PREPROCESS)
3725       putdef(cs, "__TCC_PP__");
3726     if (s1->output_type == TCC_OUTPUT_MEMORY)
3727       putdef(cs, "__TCC_RUN__");
3728     if (s1->char_is_unsigned)
3729       putdef(cs, "__CHAR_UNSIGNED__");
3730     if (s1->optimize > 0)
3731       putdef(cs, "__OPTIMIZE__");
3732     if (s1->option_pthread)
3733       putdef(cs, "_REENTRANT");
3734     if (s1->leading_underscore)
3735       putdef(cs, "__leading_underscore");
3736 #ifdef CONFIG_TCC_BCHECK
3737     if (s1->do_bounds_check)
3738       putdef(cs, "__BOUNDS_CHECKING_ON");
3739 #endif
3740     cstr_printf(cs, "#define __SIZEOF_POINTER__ %d\n", PTR_SIZE);
3741     cstr_printf(cs, "#define __SIZEOF_LONG__ %d\n", LONG_SIZE);
3742     if (!is_asm) {
3743       putdef(cs, "__STDC__");
3744       cstr_printf(cs, "#define __STDC_VERSION__ %dL\n", s1->cversion);
3745       cstr_cat(cs,
3746         /* load more predefs and __builtins */
3747 #if CONFIG_TCC_PREDEFS
3748         #include "tccdefs_.h" /* include as strings */
3749 #else
3750         "#include <tccdefs.h>\n" /* load at runtime */
3751 #endif
3752         , -1);
3753     }
3754     cstr_printf(cs, "#define __BASE_FILE__ \"%s\"\n", file->filename);
3755 }
3756
3757 ST_FUNC void preprocess_start(TCCState *s1, int filetype)
3758 {
3759     int is_asm = !!(filetype & (AFF_TYPE_ASM|AFF_TYPE_ASMPP));
3760     CString cstr;
3761
3762     tccpp_new(s1);
3763
3764     s1->include_stack_ptr = s1->include_stack;
3765     s1->ifdef_stack_ptr = s1->ifdef_stack;
3766     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3767     pp_expr = 0;
3768     pp_counter = 0;
3769     pp_debug_tok = pp_debug_symv = 0;
3770     pp_once++;
3771     s1->pack_stack[0] = 0;
3772     s1->pack_stack_ptr = s1->pack_stack;
3773
3774     set_idnum('$', !is_asm && s1->dollars_in_identifiers ? IS_ID : 0);
3775     set_idnum('.', is_asm ? IS_ID : 0);
3776
3777     if (!(filetype & AFF_TYPE_ASM)) {
3778         cstr_new(&cstr);
3779         tcc_predefs(s1, &cstr, is_asm);
3780         if (s1->cmdline_defs.size)
3781           cstr_cat(&cstr, s1->cmdline_defs.data, s1->cmdline_defs.size);
3782         if (s1->cmdline_incl.size)
3783           cstr_cat(&cstr, s1->cmdline_incl.data, s1->cmdline_incl.size);
3784         //printf("%s\n", (char*)cstr.data);
3785         *s1->include_stack_ptr++ = file;
3786         tcc_open_bf(s1, "<command line>", cstr.size);
3787         memcpy(file->buffer, cstr.data, cstr.size);
3788         cstr_free(&cstr);
3789     }
3790
3791     parse_flags = is_asm ? PARSE_FLAG_ASM_FILE : 0;
3792     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3793 }
3794
3795 /* cleanup from error/setjmp */
3796 ST_FUNC void preprocess_end(TCCState *s1)
3797 {
3798     while (macro_stack)
3799         end_macro();
3800     macro_ptr = NULL;
3801     while (file)
3802         tcc_close();
3803     tccpp_delete(s1);
3804 }
3805
3806 ST_FUNC void tccpp_new(TCCState *s)
3807 {
3808     int i, c;
3809     const char *p, *r;
3810
3811     /* init isid table */
3812     for(i = CH_EOF; i<128; i++)
3813         set_idnum(i,
3814             is_space(i) ? IS_SPC
3815             : isid(i) ? IS_ID
3816             : isnum(i) ? IS_NUM
3817             : 0);
3818
3819     for(i = 128; i<256; i++)
3820         set_idnum(i, IS_ID);
3821
3822     /* init allocators */
3823     tal_new(&toksym_alloc, TOKSYM_TAL_LIMIT, TOKSYM_TAL_SIZE);
3824     tal_new(&tokstr_alloc, TOKSTR_TAL_LIMIT, TOKSTR_TAL_SIZE);
3825
3826     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3827     memset(s->cached_includes_hash, 0, sizeof s->cached_includes_hash);
3828
3829     cstr_new(&cstr_buf);
3830     cstr_realloc(&cstr_buf, STRING_MAX_SIZE);
3831     tok_str_new(&tokstr_buf);
3832     tok_str_realloc(&tokstr_buf, TOKSTR_MAX_SIZE);
3833
3834     tok_ident = TOK_IDENT;
3835     p = tcc_keywords;
3836     while (*p) {
3837         r = p;
3838         for(;;) {
3839             c = *r++;
3840             if (c == '\0')
3841                 break;
3842         }
3843         tok_alloc(p, r - p - 1);
3844         p = r;
3845     }
3846
3847     /* we add dummy defines for some special macros to speed up tests
3848        and to have working defined() */
3849     define_push(TOK___LINE__, MACRO_OBJ, NULL, NULL);
3850     define_push(TOK___FILE__, MACRO_OBJ, NULL, NULL);
3851     define_push(TOK___DATE__, MACRO_OBJ, NULL, NULL);
3852     define_push(TOK___TIME__, MACRO_OBJ, NULL, NULL);
3853     define_push(TOK___COUNTER__, MACRO_OBJ, NULL, NULL);
3854 }
3855
3856 ST_FUNC void tccpp_delete(TCCState *s)
3857 {
3858     int i, n;
3859
3860     dynarray_reset(&s->cached_includes, &s->nb_cached_includes);
3861
3862     /* free tokens */
3863     n = tok_ident - TOK_IDENT;
3864     if (n > total_idents)
3865         total_idents = n;
3866     for(i = 0; i < n; i++)
3867         tal_free(toksym_alloc, table_ident[i]);
3868     tcc_free(table_ident);
3869     table_ident = NULL;
3870
3871     /* free static buffers */
3872     cstr_free(&tokcstr);
3873     cstr_free(&cstr_buf);
3874     cstr_free(&macro_equal_buf);
3875     tok_str_free_str(tokstr_buf.str);
3876
3877     /* free allocators */
3878     tal_delete(toksym_alloc);
3879     toksym_alloc = NULL;
3880     tal_delete(tokstr_alloc);
3881     tokstr_alloc = NULL;
3882 }
3883
3884 /* ------------------------------------------------------------------------- */
3885 /* tcc -E [-P[1]] [-dD} support */
3886
3887 static void tok_print(const char *msg, const int *str)
3888 {
3889     FILE *fp;
3890     int t, s = 0;
3891     CValue cval;
3892
3893     fp = tcc_state->ppfp;
3894     fprintf(fp, "%s", msg);
3895     while (str) {
3896         TOK_GET(&t, &str, &cval);
3897         if (!t)
3898             break;
3899         fprintf(fp, &" %s"[s], get_tok_str(t, &cval)), s = 1;
3900     }
3901     fprintf(fp, "\n");
3902 }
3903
3904 static void pp_line(TCCState *s1, BufferedFile *f, int level)
3905 {
3906     int d = f->line_num - f->line_ref;
3907
3908     if (s1->dflag & 4)
3909         return;
3910
3911     if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_NONE) {
3912         ;
3913     } else if (level == 0 && f->line_ref && d < 8) {
3914         while (d > 0)
3915             fputs("\n", s1->ppfp), --d;
3916     } else if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_STD) {
3917         fprintf(s1->ppfp, "#line %d \"%s\"\n", f->line_num, f->filename);
3918     } else {
3919         fprintf(s1->ppfp, "# %d \"%s\"%s\n", f->line_num, f->filename,
3920             level > 0 ? " 1" : level < 0 ? " 2" : "");
3921     }
3922     f->line_ref = f->line_num;
3923 }
3924
3925 static void define_print(TCCState *s1, int v)
3926 {
3927     FILE *fp;
3928     Sym *s;
3929
3930     s = define_find(v);
3931     if (NULL == s || NULL == s->d)
3932         return;
3933
3934     fp = s1->ppfp;
3935     fprintf(fp, "#define %s", get_tok_str(v, NULL));
3936     if (s->type.t == MACRO_FUNC) {
3937         Sym *a = s->next;
3938         fprintf(fp,"(");
3939         if (a)
3940             for (;;) {
3941                 fprintf(fp,"%s", get_tok_str(a->v & ~SYM_FIELD, NULL));
3942                 if (!(a = a->next))
3943                     break;
3944                 fprintf(fp,",");
3945             }
3946         fprintf(fp,")");
3947     }
3948     tok_print("", s->d);
3949 }
3950
3951 static void pp_debug_defines(TCCState *s1)
3952 {
3953     int v, t;
3954     const char *vs;
3955     FILE *fp;
3956
3957     t = pp_debug_tok;
3958     if (t == 0)
3959         return;
3960
3961     file->line_num--;
3962     pp_line(s1, file, 0);
3963     file->line_ref = ++file->line_num;
3964
3965     fp = s1->ppfp;
3966     v = pp_debug_symv;
3967     vs = get_tok_str(v, NULL);
3968     if (t == TOK_DEFINE) {
3969         define_print(s1, v);
3970     } else if (t == TOK_UNDEF) {
3971         fprintf(fp, "#undef %s\n", vs);
3972     } else if (t == TOK_push_macro) {
3973         fprintf(fp, "#pragma push_macro(\"%s\")\n", vs);
3974     } else if (t == TOK_pop_macro) {
3975         fprintf(fp, "#pragma pop_macro(\"%s\")\n", vs);
3976     }
3977     pp_debug_tok = 0;
3978 }
3979
3980 static void pp_debug_builtins(TCCState *s1)
3981 {
3982     int v;
3983     for (v = TOK_IDENT; v < tok_ident; ++v)
3984         define_print(s1, v);
3985 }
3986
3987 /* Add a space between tokens a and b to avoid unwanted textual pasting */
3988 static int pp_need_space(int a, int b)
3989 {
3990     return 'E' == a ? '+' == b || '-' == b
3991         : '+' == a ? TOK_INC == b || '+' == b
3992         : '-' == a ? TOK_DEC == b || '-' == b
3993         : a >= TOK_IDENT ? b >= TOK_IDENT
3994         : a == TOK_PPNUM ? b >= TOK_IDENT
3995         : 0;
3996 }
3997
3998 /* maybe hex like 0x1e */
3999 static int pp_check_he0xE(int t, const char *p)
4000 {
4001     if (t == TOK_PPNUM && toup(strchr(p, 0)[-1]) == 'E')
4002         return 'E';
4003     return t;
4004 }
4005
4006 /* Preprocess the current file */
4007 ST_FUNC int tcc_preprocess(TCCState *s1)
4008 {
4009     BufferedFile **iptr;
4010     int token_seen, spcs, level;
4011     const char *p;
4012     char white[400];
4013
4014     parse_flags = PARSE_FLAG_PREPROCESS
4015                 | (parse_flags & PARSE_FLAG_ASM_FILE)
4016                 | PARSE_FLAG_LINEFEED
4017                 | PARSE_FLAG_SPACES
4018                 | PARSE_FLAG_ACCEPT_STRAYS
4019                 ;
4020     /* Credits to Fabrice Bellard's initial revision to demonstrate its
4021        capability to compile and run itself, provided all numbers are
4022        given as decimals. tcc -E -P10 will do. */
4023     if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_P10)
4024         parse_flags |= PARSE_FLAG_TOK_NUM, s1->Pflag = 1;
4025
4026     if (s1->do_bench) {
4027         /* for PP benchmarks */
4028         do next(); while (tok != TOK_EOF);
4029         return 0;
4030     }
4031
4032     if (s1->dflag & 1) {
4033         pp_debug_builtins(s1);
4034         s1->dflag &= ~1;
4035     }
4036
4037     token_seen = TOK_LINEFEED, spcs = 0, level = 0;
4038     if (file->prev)
4039         pp_line(s1, file->prev, level++);
4040     pp_line(s1, file, level);
4041     for (;;) {
4042         iptr = s1->include_stack_ptr;
4043         next();
4044         if (tok == TOK_EOF)
4045             break;
4046
4047         level = s1->include_stack_ptr - iptr;
4048         if (level) {
4049             if (level > 0)
4050                 pp_line(s1, *iptr, 0);
4051             pp_line(s1, file, level);
4052         }
4053         if (s1->dflag & 7) {
4054             pp_debug_defines(s1);
4055             if (s1->dflag & 4)
4056                 continue;
4057         }
4058
4059         if (is_space(tok)) {
4060             if (spcs < sizeof white - 1)
4061                 white[spcs++] = tok;
4062             continue;
4063         } else if (tok == TOK_LINEFEED) {
4064             spcs = 0;
4065             if (token_seen == TOK_LINEFEED)
4066                 continue;
4067             ++file->line_ref;
4068         } else if (token_seen == TOK_LINEFEED) {
4069             pp_line(s1, file, 0);
4070         } else if (spcs == 0 && pp_need_space(token_seen, tok)) {
4071             white[spcs++] = ' ';
4072         }
4073
4074         white[spcs] = 0, fputs(white, s1->ppfp), spcs = 0;
4075         fputs(p = get_tok_str(tok, &tokc), s1->ppfp);
4076         token_seen = pp_check_he0xE(tok, p);
4077     }
4078     return 0;
4079 }
4080
4081 /* ------------------------------------------------------------------------- */