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  67 /* outdated -- gr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