tokenize.c

   1 /*
   2  * This is a really stupid C tokenizer. It doesn't do any include
   3  * files or anything complex at all. That's the preprocessor.
   4  *
   5  * Copyright (C) 2003 Transmeta Corp.
   6  *               2003 Linus Torvalds
   7  *
   8  * Permission is hereby granted, free of charge, to any person obtaining a copy
   9  * of this software and associated documentation files (the "Software"), to deal
  10  * in the Software without restriction, including without limitation the rights
  11  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  12  * copies of the Software, and to permit persons to whom the Software is
  13  * furnished to do so, subject to the following conditions:
  14  *
  15  * The above copyright notice and this permission notice shall be included in
  16  * all copies or substantial portions of the Software.
  17  *
  18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  19  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  20  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
  21  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  22  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  23  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  24  * THE SOFTWARE.
  25  */
  26 #include <stdio.h>
  27 #include <stdlib.h>
  28 #include <stdarg.h>
  29 #include <stddef.h>
  30 #include <string.h>
  31 #include <ctype.h>
  32 #include <unistd.h>
  33 #include <stdint.h>
  34
  35 #include "lib.h"
  36 #include "allocate.h"
  37 #include "token.h"
  38 #include "symbol.h"
  39
  40 #define EOF (-1)
  41
  42 int input_stream_nr = 0;
  43 struct stream *input_streams;
  44 static int input_streams_allocated;
  45 unsigned int tabstop = 8;
  46
  47 #define BUFSIZE (8192)
  48
  49 typedef struct {
  50         int fd, offset, size;
  51         int pos, line, nr;
  52         int newline, whitespace;
  53         struct token **tokenlist;
  54         struct token *token;
  55         unsigned char *buffer;
  56 } stream_t;
  57
  58 const char *stream_name(int stream)
  59 {
  60         if (stream < 0 || stream > input_stream_nr)
  61                 return "<bad stream>";
  62         return input_streams[stream].name;
  63 }
  64
  65 static struct position stream_pos(stream_t *stream)
  66 {
  67         struct position pos;
  68         pos.type = 0;
  69         pos.stream = stream->nr;
  70         pos.newline = stream->newline;
  71         pos.whitespace = stream->whitespace;
  72         pos.pos = stream->pos;
  73         pos.line = stream->line;
  74         pos.noexpand = 0;
  75         return pos;
  76 }
  77
  78 const char *show_special(int val)
  79 {
  80         static char buffer[4];
  81
  82         buffer[0] = val;
  83         buffer[1] = 0;
  84         if (val >= SPECIAL_BASE)
  85                 strcpy(buffer, (char *) combinations[val - SPECIAL_BASE]);
  86         return buffer;
  87 }
  88
  89 const char *show_ident(const struct ident *ident)
  90 {
  91         static char buffer[256];
  92         if (!ident)
  93                 return "<noident>";
  94         sprintf(buffer, "%.*s", ident->len, ident->name);
  95         return buffer;
  96 }
  97
  98 static char *charstr(char *ptr, unsigned char c, unsigned char escape, unsigned char next)
  99 {
 100         if (isprint(c)) {
 101                 if (c == escape || c == '\\')
 102                         *ptr++ = '\\';
 103                 *ptr++ = c;
 104                 return ptr;
 105         }
 106         *ptr++ = '\\';
 107         switch (c) {
 108         case '\n':
 109                 *ptr++ = 'n';
 110                 return ptr;
 111         case '\t':
 112                 *ptr++ = 't';
 113                 return ptr;
 114         }
 115         if (!isdigit(next))
 116                 return ptr + sprintf(ptr, "%o", c);
 117
 118         return ptr + sprintf(ptr, "%03o", c);
 119 }
 120
 121 const char *show_string(const struct string *string)
 122 {
 123         static char buffer[4 * MAX_STRING + 3];
 124         char *ptr;
 125         int i;
 126
 127         if (!string->length)
 128                 return "<bad_string>";
 129         ptr = buffer;
 130         *ptr++ = '"';
 131         for (i = 0; i < string->length-1; i++) {
 132                 const char *p = string->data + i;
 133                 ptr = charstr(ptr, p[0], '"', p[1]);
 134         }
 135         *ptr++ = '"';
 136         *ptr = '\0';
 137         return buffer;
 138 }
 139
 140 static const char *show_char(const char *s, size_t len, char prefix, char delim)
 141 {
 142         static char buffer[MAX_STRING + 4];
 143         char *p = buffer;
 144         if (prefix)
 145                 *p++ = prefix;
 146         *p++ = delim;
 147         memcpy(p, s, len);
 148         p += len;
 149         *p++ = delim;
 150         *p++ = '\0';
 151         return buffer;
 152 }
 153
 154 static const char *quote_char(const char *s, size_t len, char prefix, char delim)
 155 {
 156         static char buffer[2*MAX_STRING + 6];
 157         size_t i;
 158         char *p = buffer;
 159         if (prefix)
 160                 *p++ = prefix;
 161         if (delim == '"')
 162                 *p++ = '\\';
 163         *p++ = delim;
 164         for (i = 0; i < len; i++) {
 165                 if (s[i] == '"' || s[i] == '\\')
 166                         *p++ = '\\';
 167                 *p++ = s[i];
 168         }
 169         if (delim == '"')
 170                 *p++ = '\\';
 171         *p++ = delim;
 172         *p++ = '\0';
 173         return buffer;
 174 }
 175
 176 const char *show_token(const struct token *token)
 177 {
 178         static char buffer[256];
 179
 180         if (!token)
 181                 return "<no token>";
 182         switch (token_type(token)) {
 183         case TOKEN_ERROR:
 184                 return "syntax error";
 185
 186         case TOKEN_EOF:
 187                 return "end-of-input";
 188
 189         case TOKEN_IDENT:
 190                 return show_ident(token->ident);
 191
 192         case TOKEN_NUMBER:
 193                 return token->number;
 194
 195         case TOKEN_SPECIAL:
 196                 return show_special(token->special);
 197
 198         case TOKEN_CHAR:
 199                 return show_char(token->string->data,
 200                         token->string->length - 1, 0, '\'');
 201         case TOKEN_CHAR_EMBEDDED_0 ... TOKEN_CHAR_EMBEDDED_3:
 202                 return show_char(token->embedded,
 203                         token_type(token) - TOKEN_CHAR, 0, '\'');
 204         case TOKEN_WIDE_CHAR:
 205                 return show_char(token->string->data,
 206                         token->string->length - 1, 'L', '\'');
 207         case TOKEN_WIDE_CHAR_EMBEDDED_0 ... TOKEN_WIDE_CHAR_EMBEDDED_3:
 208                 return show_char(token->embedded,
 209                         token_type(token) - TOKEN_WIDE_CHAR, 'L', '\'');
 210         case TOKEN_STRING:
 211                 return show_char(token->string->data,
 212                         token->string->length - 1, 0, '"');
 213         case TOKEN_WIDE_STRING:
 214                 return show_char(token->string->data,
 215                         token->string->length - 1, 'L', '"');
 216
 217         case TOKEN_STREAMBEGIN:
 218                 sprintf(buffer, "<beginning of '%s'>", stream_name(token->pos.stream));
 219                 return buffer;
 220
 221         case TOKEN_STREAMEND:
 222                 sprintf(buffer, "<end of '%s'>", stream_name(token->pos.stream));
 223                 return buffer;
 224
 225         case TOKEN_UNTAINT:
 226                 sprintf(buffer, "<untaint>");
 227                 return buffer;
 228
 229         case TOKEN_ARG_COUNT:
 230                 sprintf(buffer, "<argcnt>");
 231                 return buffer;
 232
 233         default:
 234                 sprintf(buffer, "unhandled token type '%d' ", token_type(token));
 235                 return buffer;
 236         }
 237 }
 238
 239 const char *quote_token(const struct token *token)
 240 {
 241         static char buffer[256];
 242
 243         switch (token_type(token)) {
 244         case TOKEN_ERROR:
 245                 return "syntax error";
 246
 247         case TOKEN_IDENT:
 248                 return show_ident(token->ident);
 249
 250         case TOKEN_NUMBER:
 251                 return token->number;
 252
 253         case TOKEN_SPECIAL:
 254                 return show_special(token->special);
 255
 256         case TOKEN_CHAR:
 257                 return quote_char(token->string->data,
 258                         token->string->length - 1, 0, '\'');
 259         case TOKEN_CHAR_EMBEDDED_0 ... TOKEN_CHAR_EMBEDDED_3:
 260                 return quote_char(token->embedded,
 261                         token_type(token) - TOKEN_CHAR, 0, '\'');
 262         case TOKEN_WIDE_CHAR:
 263                 return quote_char(token->string->data,
 264                         token->string->length - 1, 'L', '\'');
 265         case TOKEN_WIDE_CHAR_EMBEDDED_0 ... TOKEN_WIDE_CHAR_EMBEDDED_3:
 266                 return quote_char(token->embedded,
 267                         token_type(token) - TOKEN_WIDE_CHAR, 'L', '\'');
 268         case TOKEN_STRING:
 269                 return quote_char(token->string->data,
 270                         token->string->length - 1, 0, '"');
 271         case TOKEN_WIDE_STRING:
 272                 return quote_char(token->string->data,
 273                         token->string->length - 1, 'L', '"');
 274         default:
 275                 sprintf(buffer, "unhandled token type '%d' ", token_type(token));
 276                 return buffer;
 277         }
 278 }
 279
 280 #define HASHED_INPUT_BITS (6)
 281 #define HASHED_INPUT (1 << HASHED_INPUT_BITS)
 282 #define HASH_PRIME 0x9e370001UL
 283
 284 static int input_stream_hashes[HASHED_INPUT] = { [0 ... HASHED_INPUT-1] = -1 };
 285
 286 int *hash_stream(const char *name)
 287 {
 288         uint32_t hash = 0;
 289         unsigned char c;
 290
 291         while ((c = *name++) != 0)
 292                 hash = (hash + (c << 4) + (c >> 4)) * 11;
 293
 294         hash *= HASH_PRIME;
 295         hash >>= 32 - HASHED_INPUT_BITS;
 296         return input_stream_hashes + hash;
 297 }
 298
 299 int init_stream(const char *name, int fd, const char **next_path)
 300 {
 301         int stream = input_stream_nr, *hash;
 302         struct stream *current;
 303
 304         if (stream >= input_streams_allocated) {
 305                 int newalloc = stream * 4 / 3 + 10;
 306                 input_streams = realloc(input_streams, newalloc * sizeof(struct stream));
 307                 if (!input_streams)
 308                         die("Unable to allocate more streams space");
 309                 input_streams_allocated = newalloc;
 310         }
 311         current = input_streams + stream;
 312         memset(current, 0, sizeof(*current));
 313         current->name = name;
 314         current->fd = fd;
 315         current->next_path = next_path;
 316         current->path = NULL;
 317         current->constant = CONSTANT_FILE_MAYBE;
 318         input_stream_nr = stream+1;
 319         hash = hash_stream(name);
 320         current->next_stream = *hash;
 321         *hash = stream;
 322         return stream;
 323 }
 324
 325 static struct token * alloc_token(stream_t *stream)
 326 {
 327         struct token *token = __alloc_token(0);
 328         token->pos = stream_pos(stream);
 329         return token;
 330 }
 331
 332 /*
 333  *  Argh...  That was surprisingly messy - handling '\r' complicates the
 334  *  things a _lot_.
 335  */
 336 static int nextchar_slow(stream_t *stream)
 337 {
 338         int offset = stream->offset;
 339         int size = stream->size;
 340         int c;
 341         int spliced = 0, had_cr, had_backslash;
 342
 343 restart:
 344         had_cr = had_backslash = 0;
 345
 346 repeat:
 347         if (offset >= size) {
 348                 if (stream->fd < 0)
 349                         goto got_eof;
 350                 size = read(stream->fd, stream->buffer, BUFSIZE);
 351                 if (size <= 0)
 352                         goto got_eof;
 353                 stream->size = size;
 354                 stream->offset = offset = 0;
 355         }
 356
 357         c = stream->buffer[offset++];
 358         if (had_cr)
 359                 goto check_lf;
 360
 361         if (c == '\r') {
 362                 had_cr = 1;
 363                 goto repeat;
 364         }
 365
 366 norm:
 367         if (!had_backslash) {
 368                 switch (c) {
 369                 case '\t':
 370                         stream->pos += tabstop - stream->pos % tabstop;
 371                         break;
 372                 case '\n':
 373                         stream->line++;
 374                         stream->pos = 0;
 375                         stream->newline = 1;
 376                         break;
 377                 case '\\':
 378                         had_backslash = 1;
 379                         stream->pos++;
 380                         goto repeat;
 381                 default:
 382                         stream->pos++;
 383                 }
 384         } else {
 385                 if (c == '\n') {
 386                         stream->line++;
 387                         stream->pos = 0;
 388                         spliced = 1;
 389                         goto restart;
 390                 }
 391                 offset--;
 392                 c = '\\';
 393         }
 394 out:
 395         stream->offset = offset;
 396
 397         return c;
 398
 399 check_lf:
 400         if (c != '\n')
 401                 offset--;
 402         c = '\n';
 403         goto norm;
 404
 405 got_eof:
 406         if (had_backslash) {
 407                 c = '\\';
 408                 goto out;
 409         }
 410         if (stream->pos)
 411                 warning(stream_pos(stream), "no newline at end of file");
 412         else if (spliced)
 413                 warning(stream_pos(stream), "backslash-newline at end of file");
 414         return EOF;
 415 }
 416
 417 /*
 418  *  We want that as light as possible while covering all normal cases.
 419  *  Slow path (including the logics with line-splicing and EOF sanity
 420  *  checks) is in nextchar_slow().
 421  */
 422 static inline int nextchar(stream_t *stream)
 423 {
 424         int offset = stream->offset;
 425
 426         if (offset < stream->size) {
 427                 int c = stream->buffer[offset++];
 428                 static const char special[256] = {
 429                         ['\t'] = 1, ['\r'] = 1, ['\n'] = 1, ['\\'] = 1
 430                 };
 431                 if (!special[c]) {
 432                         stream->offset = offset;
 433                         stream->pos++;
 434                         return c;
 435                 }
 436         }
 437         return nextchar_slow(stream);
 438 }
 439
 440 struct token eof_token_entry;
 441
 442 static struct token *mark_eof(stream_t *stream)
 443 {
 444         struct token *end;
 445
 446         end = alloc_token(stream);
 447         token_type(end) = TOKEN_STREAMEND;
 448         end->pos.newline = 1;
 449
 450         eof_token_entry.next = &eof_token_entry;
 451         eof_token_entry.pos.newline = 1;
 452
 453         end->next =  &eof_token_entry;
 454         *stream->tokenlist = end;
 455         stream->tokenlist = NULL;
 456         return end;
 457 }
 458
 459 static void add_token(stream_t *stream)
 460 {
 461         struct token *token = stream->token;
 462
 463         stream->token = NULL;
 464         token->next = NULL;
 465         *stream->tokenlist = token;
 466         stream->tokenlist = &token->next;
 467 }
 468
 469 static void drop_token(stream_t *stream)
 470 {
 471         stream->newline |= stream->token->pos.newline;
 472         stream->whitespace |= stream->token->pos.whitespace;
 473         stream->token = NULL;
 474 }
 475
 476 enum {
 477         Letter = 1,
 478         Digit = 2,
 479         Hex = 4,
 480         Exp = 8,
 481         Dot = 16,
 482         ValidSecond = 32,
 483         Quote = 64,
 484 };
 485
 486 static const long cclass[257] = {
 487         ['0' + 1 ... '7' + 1] = Digit | Hex,    /* \<octal> */
 488         ['8' + 1 ... '9' + 1] = Digit | Hex,
 489         ['A' + 1 ... 'D' + 1] = Letter | Hex,
 490         ['E' + 1] = Letter | Hex | Exp, /* E<exp> */
 491         ['F' + 1] = Letter | Hex,
 492         ['G' + 1 ... 'O' + 1] = Letter,
 493         ['P' + 1] = Letter | Exp,       /* P<exp> */
 494         ['Q' + 1 ... 'Z' + 1] = Letter,
 495         ['a' + 1 ... 'b' + 1] = Letter | Hex, /* \a, \b */
 496         ['c' + 1 ... 'd' + 1] = Letter | Hex,
 497         ['e' + 1] = Letter | Hex | Exp,/* \e, e<exp> */
 498         ['f' + 1] = Letter | Hex,       /* \f */
 499         ['g' + 1 ... 'm' + 1] = Letter,
 500         ['n' + 1] = Letter,     /* \n */
 501         ['o' + 1] = Letter,
 502         ['p' + 1] = Letter | Exp,       /* p<exp> */
 503         ['q' + 1] = Letter,
 504         ['r' + 1] = Letter,     /* \r */
 505         ['s' + 1] = Letter,
 506         ['t' + 1] = Letter,     /* \t */
 507         ['u' + 1] = Letter,
 508         ['v' + 1] = Letter,     /* \v */
 509         ['w' + 1] = Letter,
 510         ['x' + 1] = Letter,     /* \x<hex> */
 511         ['y' + 1 ... 'z' + 1] = Letter,
 512         ['_' + 1] = Letter,
 513         ['.' + 1] = Dot | ValidSecond,
 514         ['=' + 1] = ValidSecond,
 515         ['+' + 1] = ValidSecond,
 516         ['-' + 1] = ValidSecond,
 517         ['>' + 1] = ValidSecond,
 518         ['<' + 1] = ValidSecond,
 519         ['&' + 1] = ValidSecond,
 520         ['|' + 1] = ValidSecond,
 521         ['#' + 1] = ValidSecond,
 522         ['\'' + 1] = Quote,
 523         ['"' + 1] = Quote,
 524 };
 525
 526 /*
 527  * pp-number:
 528  *      digit
 529  *      . digit
 530  *      pp-number digit
 531  *      pp-number identifier-nodigit
 532  *      pp-number e sign
 533  *      pp-number E sign
 534  *      pp-number p sign
 535  *      pp-number P sign
 536  *      pp-number .
 537  */
 538 static int get_one_number(int c, int next, stream_t *stream)
 539 {
 540         struct token *token;
 541         static char buffer[4095];
 542         char *p = buffer, *buf, *buffer_end = buffer + sizeof (buffer);
 543         int len;
 544
 545         *p++ = c;
 546         for (;;) {
 547                 long class =  cclass[next + 1];
 548                 if (!(class & (Dot | Digit | Letter)))
 549                         break;
 550                 if (p != buffer_end)
 551                         *p++ = next;
 552                 next = nextchar(stream);
 553                 if (class & Exp) {
 554                         if (next == '-' || next == '+') {
 555                                 if (p != buffer_end)
 556                                         *p++ = next;
 557                                 next = nextchar(stream);
 558                         }
 559                 }
 560         }
 561
 562         if (p == buffer_end) {
 563                 sparse_error(stream_pos(stream), "number token exceeds %td characters",
 564                       buffer_end - buffer);
 565                 // Pretend we saw just "1".
 566                 buffer[0] = '1';
 567                 p = buffer + 1;
 568         }
 569
 570         *p++ = 0;
 571         len = p - buffer;
 572         buf = __alloc_bytes(len);
 573         memcpy(buf, buffer, len);
 574
 575         token = stream->token;
 576         token_type(token) = TOKEN_NUMBER;
 577         token->number = buf;
 578         add_token(stream);
 579
 580         return next;
 581 }
 582
 583 static int eat_string(int next, stream_t *stream, enum token_type type)
 584 {
 585         static char buffer[MAX_STRING];
 586         struct string *string;
 587         struct token *token = stream->token;
 588         int len = 0;
 589         int escape;
 590         int want_hex = 0;
 591         char delim = type < TOKEN_STRING ? '\'' : '"';
 592
 593         for (escape = 0; escape || next != delim; next = nextchar(stream)) {
 594                 if (len < MAX_STRING)
 595                         buffer[len] = next;
 596                 len++;
 597                 if (next == '\n') {
 598                         warning(stream_pos(stream),
 599                                 "Newline in string or character constant");
 600                         if (delim == '\'') /* assume it's lost ' */
 601                                 break;
 602                 }
 603                 if (next == EOF) {
 604                         warning(stream_pos(stream),
 605                                 "End of file in middle of string");
 606                         return next;
 607                 }
 608                 if (!escape) {
 609                         if (want_hex && !(cclass[next + 1] & Hex))
 610                                 warning(stream_pos(stream),
 611                                         "\\x used with no following hex digits");
 612                         want_hex = 0;
 613                         escape = next == '\\';
 614                 } else {
 615                         escape = 0;
 616                         want_hex = next == 'x';
 617                 }
 618         }
 619         if (want_hex)
 620                 warning(stream_pos(stream),
 621                         "\\x used with no following hex digits");
 622         if (len > MAX_STRING) {
 623                 warning(stream_pos(stream), "string too long (%d bytes, %d bytes max)", len, MAX_STRING);
 624                 len = MAX_STRING;
 625         }
 626         if (delim == '\'' && len <= 4) {
 627                 if (len == 0) {
 628                         sparse_error(stream_pos(stream),
 629                                 "empty character constant");
 630                         return nextchar(stream);
 631                 }
 632                 token_type(token) = type + len;
 633                 memset(buffer + len, '\0', 4 - len);
 634                 memcpy(token->embedded, buffer, 4);
 635         } else {
 636                 token_type(token) = type;
 637                 string = __alloc_string(len+1);
 638                 memcpy(string->data, buffer, len);
 639                 string->data[len] = '\0';
 640                 string->length = len+1;
 641                 token->string = string;
 642         }
 643
 644         /* Pass it on.. */
 645         token = stream->token;
 646         add_token(stream);
 647         return nextchar(stream);
 648 }
 649
 650 static int drop_stream_eoln(stream_t *stream)
 651 {
 652         drop_token(stream);
 653         for (;;) {
 654                 switch (nextchar(stream)) {
 655                 case EOF:
 656                         return EOF;
 657                 case '\n':
 658                         return nextchar(stream);
 659                 }
 660         }
 661 }
 662
 663 static int drop_stream_comment(stream_t *stream)
 664 {
 665         int newline;
 666         int next;
 667         drop_token(stream);
 668         newline = stream->newline;
 669
 670         next = nextchar(stream);
 671         for (;;) {
 672                 int curr = next;
 673                 if (curr == EOF) {
 674                         warning(stream_pos(stream), "End of file in the middle of a comment");
 675                         return curr;
 676                 }
 677                 next = nextchar(stream);
 678                 if (curr == '*' && next == '/')
 679                         break;
 680         }
 681         stream->newline = newline;
 682         return nextchar(stream);
 683 }
 684
 685 unsigned char combinations[][4] = COMBINATION_STRINGS;
 686
 687 #define NR_COMBINATIONS (SPECIAL_ARG_SEPARATOR - SPECIAL_BASE)
 688
 689 /* hash function for two-character punctuators - all give unique values */
 690 #define special_hash(c0, c1) (((c0*8+c1*2)+((c0*8+c1*2)>>5))&31)
 691
 692 /*
 693  * note that we won't get false positives - special_hash(0,0) is 0 and
 694  * entry 0 is filled (by +=), so all the missing ones are OK.
 695  */
 696 static unsigned char hash_results[32][2] = {
 697 #define RES(c0, c1) [special_hash(c0, c1)] = {c0, c1}
 698         RES('+', '='), /* 00 */
 699         RES('/', '='), /* 01 */
 700         RES('^', '='), /* 05 */
 701         RES('&', '&'), /* 07 */
 702         RES('#', '#'), /* 08 */
 703         RES('<', '<'), /* 0a */
 704         RES('<', '='), /* 0c */
 705         RES('!', '='), /* 0e */
 706         RES('%', '='), /* 0f */
 707         RES('-', '-'), /* 10 */
 708         RES('-', '='), /* 11 */
 709         RES('-', '>'), /* 13 */
 710         RES('=', '='), /* 15 */
 711         RES('&', '='), /* 17 */
 712         RES('*', '='), /* 18 */
 713         RES('.', '.'), /* 1a */
 714         RES('+', '+'), /* 1b */
 715         RES('|', '='), /* 1c */
 716         RES('>', '='), /* 1d */
 717         RES('|', '|'), /* 1e */
 718         RES('>', '>')  /* 1f */
 719 #undef RES
 720 };
 721 static int code[32] = {
 722 #define CODE(c0, c1, value) [special_hash(c0, c1)] = value
 723         CODE('+', '=', SPECIAL_ADD_ASSIGN), /* 00 */
 724         CODE('/', '=', SPECIAL_DIV_ASSIGN), /* 01 */
 725         CODE('^', '=', SPECIAL_XOR_ASSIGN), /* 05 */
 726         CODE('&', '&', SPECIAL_LOGICAL_AND), /* 07 */
 727         CODE('#', '#', SPECIAL_HASHHASH), /* 08 */
 728         CODE('<', '<', SPECIAL_LEFTSHIFT), /* 0a */
 729         CODE('<', '=', SPECIAL_LTE), /* 0c */
 730         CODE('!', '=', SPECIAL_NOTEQUAL), /* 0e */
 731         CODE('%', '=', SPECIAL_MOD_ASSIGN), /* 0f */
 732         CODE('-', '-', SPECIAL_DECREMENT), /* 10 */
 733         CODE('-', '=', SPECIAL_SUB_ASSIGN), /* 11 */
 734         CODE('-', '>', SPECIAL_DEREFERENCE), /* 13 */
 735         CODE('=', '=', SPECIAL_EQUAL), /* 15 */
 736         CODE('&', '=', SPECIAL_AND_ASSIGN), /* 17 */
 737         CODE('*', '=', SPECIAL_MUL_ASSIGN), /* 18 */
 738         CODE('.', '.', SPECIAL_DOTDOT), /* 1a */
 739         CODE('+', '+', SPECIAL_INCREMENT), /* 1b */
 740         CODE('|', '=', SPECIAL_OR_ASSIGN), /* 1c */
 741         CODE('>', '=', SPECIAL_GTE), /* 1d */
 742         CODE('|', '|', SPECIAL_LOGICAL_OR), /* 1e */
 743         CODE('>', '>', SPECIAL_RIGHTSHIFT)  /* 1f */
 744 #undef CODE
 745 };
 746
 747 static int get_one_special(int c, stream_t *stream)
 748 {
 749         struct token *token;
 750         int next, value, i;
 751
 752         next = nextchar(stream);
 753
 754         /*
 755          * Check for numbers, strings, character constants, and comments
 756          */
 757         switch (c) {
 758         case '.':
 759                 if (next >= '0' && next <= '9')
 760                         return get_one_number(c, next, stream);
 761                 break;
 762         case '"':
 763                 return eat_string(next, stream, TOKEN_STRING);
 764         case '\'':
 765                 return eat_string(next, stream, TOKEN_CHAR);
 766         case '/':
 767                 if (next == '/')
 768                         return drop_stream_eoln(stream);
 769                 if (next == '*')
 770                         return drop_stream_comment(stream);
 771         }
 772
 773         /*
 774          * Check for combinations
 775          */
 776         value = c;
 777         if (cclass[next + 1] & ValidSecond) {
 778                 i = special_hash(c, next);
 779                 if (hash_results[i][0] == c && hash_results[i][1] == next) {
 780                         value = code[i];
 781                         next = nextchar(stream);
 782                         if (value >= SPECIAL_LEFTSHIFT &&
 783                             next == "==."[value - SPECIAL_LEFTSHIFT]) {
 784                                 value += 3;
 785                                 next = nextchar(stream);
 786                         }
 787                 }
 788         }
 789
 790         /* Pass it on.. */
 791         token = stream->token;
 792         token_type(token) = TOKEN_SPECIAL;
 793         token->special = value;
 794         add_token(stream);
 795         return next;
 796 }
 797
 798 #define IDENT_HASH_BITS (13)
 799 #define IDENT_HASH_SIZE (1<<IDENT_HASH_BITS)
 800 #define IDENT_HASH_MASK (IDENT_HASH_SIZE-1)
 801
 802 #define ident_hash_init(c)              (c)
 803 #define ident_hash_add(oldhash,c)       ((oldhash)*11 + (c))
 804 #define ident_hash_end(hash)            ((((hash) >> IDENT_HASH_BITS) + (hash)) & IDENT_HASH_MASK)
 805
 806 static struct ident *hash_table[IDENT_HASH_SIZE];
 807 static int ident_hit, ident_miss, idents;
 808
 809 void show_identifier_stats(void)
 810 {
 811         int i;
 812         int distribution[100];
 813
 814         fprintf(stderr, "identifiers: %d hits, %d misses\n",
 815                 ident_hit, ident_miss);
 816
 817         for (i = 0; i < 100; i++)
 818                 distribution[i] = 0;
 819
 820         for (i = 0; i < IDENT_HASH_SIZE; i++) {
 821                 struct ident * ident = hash_table[i];
 822                 int count = 0;
 823
 824                 while (ident) {
 825                         count++;
 826                         ident = ident->next;
 827                 }
 828                 if (count > 99)
 829                         count = 99;
 830                 distribution[count]++;
 831         }
 832
 833         for (i = 0; i < 100; i++) {
 834                 if (distribution[i])
 835                         fprintf(stderr, "%2d: %d buckets\n", i, distribution[i]);
 836         }
 837 }
 838
 839 static struct ident *alloc_ident(const char *name, int len)
 840 {
 841         struct ident *ident = __alloc_ident(len);
 842         ident->symbols = NULL;
 843         ident->len = len;
 844         ident->tainted = 0;
 845         memcpy(ident->name, name, len);
 846         return ident;
 847 }
 848
 849 static struct ident * insert_hash(struct ident *ident, unsigned long hash)
 850 {
 851         ident->next = hash_table[hash];
 852         hash_table[hash] = ident;
 853         ident_miss++;
 854         return ident;
 855 }
 856
 857 static struct ident *create_hashed_ident(const char *name, int len, unsigned long hash)
 858 {
 859         struct ident *ident;
 860         struct ident **p;
 861
 862         p = &hash_table[hash];
 863         while ((ident = *p) != NULL) {
 864                 if (ident->len == (unsigned char) len) {
 865                         if (strncmp(name, ident->name, len) != 0)
 866                                 goto next;
 867
 868                         ident_hit++;
 869                         return ident;
 870                 }
 871 next:
 872                 //misses++;
 873                 p = &ident->next;
 874         }
 875         ident = alloc_ident(name, len);
 876         *p = ident;
 877         ident->next = NULL;
 878         ident_miss++;
 879         idents++;
 880         return ident;
 881 }
 882
 883 static unsigned long hash_name(const char *name, int len)
 884 {
 885         unsigned long hash;
 886         const unsigned char *p = (const unsigned char *)name;
 887
 888         hash = ident_hash_init(*p++);
 889         while (--len) {
 890                 unsigned int i = *p++;
 891                 hash = ident_hash_add(hash, i);
 892         }
 893         return ident_hash_end(hash);
 894 }
 895
 896 struct ident *hash_ident(struct ident *ident)
 897 {
 898         return insert_hash(ident, hash_name(ident->name, ident->len));
 899 }
 900
 901 struct ident *built_in_ident(const char *name)
 902 {
 903         int len = strlen(name);
 904         return create_hashed_ident(name, len, hash_name(name, len));
 905 }
 906
 907 struct token *built_in_token(int stream, struct ident *ident)
 908 {
 909         struct token *token;
 910
 911         token = __alloc_token(0);
 912         token->pos.stream = stream;
 913         token_type(token) = TOKEN_IDENT;
 914         token->ident = ident;
 915         return token;
 916 }
 917
 918 static int get_one_identifier(int c, stream_t *stream)
 919 {
 920         struct token *token;
 921         struct ident *ident;
 922         unsigned long hash;
 923         char buf[256];
 924         int len = 1;
 925         int next;
 926
 927         hash = ident_hash_init(c);
 928         buf[0] = c;
 929         for (;;) {
 930                 next = nextchar(stream);
 931                 if (!(cclass[next + 1] & (Letter | Digit)))
 932                         break;
 933                 if (len >= sizeof(buf))
 934                         break;
 935                 hash = ident_hash_add(hash, next);
 936                 buf[len] = next;
 937                 len++;
 938         };
 939         if (cclass[next + 1] & Quote) {
 940                 if (len == 1 && buf[0] == 'L') {
 941                         if (next == '\'')
 942                                 return eat_string(nextchar(stream), stream,
 943                                                         TOKEN_WIDE_CHAR);
 944                         else
 945                                 return eat_string(nextchar(stream), stream,
 946                                                         TOKEN_WIDE_STRING);
 947                 }
 948         }
 949         hash = ident_hash_end(hash);
 950         ident = create_hashed_ident(buf, len, hash);
 951
 952         /* Pass it on.. */
 953         token = stream->token;
 954         token_type(token) = TOKEN_IDENT;
 955         token->ident = ident;
 956         add_token(stream);
 957         return next;
 958 }
 959
 960 static int get_one_token(int c, stream_t *stream)
 961 {
 962         long class = cclass[c + 1];
 963         if (class & Digit)
 964                 return get_one_number(c, nextchar(stream), stream);
 965         if (class & Letter)
 966                 return get_one_identifier(c, stream);
 967         return get_one_special(c, stream);
 968 }
 969
 970 static struct token *setup_stream(stream_t *stream, int idx, int fd,
 971         unsigned char *buf, unsigned int buf_size)
 972 {
 973         struct token *begin;
 974
 975         stream->nr = idx;
 976         stream->line = 1;
 977         stream->newline = 1;
 978         stream->whitespace = 0;
 979         stream->pos = 0;
 980
 981         stream->token = NULL;
 982         stream->fd = fd;
 983         stream->offset = 0;
 984         stream->size = buf_size;
 985         stream->buffer = buf;
 986
 987         begin = alloc_token(stream);
 988         token_type(begin) = TOKEN_STREAMBEGIN;
 989         stream->tokenlist = &begin->next;
 990         return begin;
 991 }
 992
 993 static struct token *tokenize_stream(stream_t *stream)
 994 {
 995         int c = nextchar(stream);
 996         while (c != EOF) {
 997                 if (!isspace(c)) {
 998                         struct token *token = alloc_token(stream);
 999                         stream->token = token;
1000                         stream->newline = 0;
1001                         stream->whitespace = 0;
1002                         c = get_one_token(c, stream);
1003                         continue;
1004                 }
1005                 stream->whitespace = 1;
1006                 c = nextchar(stream);
1007         }
1008         return mark_eof(stream);
1009 }
1010
1011 struct token * tokenize_buffer(void *buffer, unsigned long size, struct token **endtoken)
1012 {
1013         stream_t stream;
1014         struct token *begin;
1015
1016         begin = setup_stream(&stream, 0, -1, buffer, size);
1017         *endtoken = tokenize_stream(&stream);
1018         return begin;
1019 }
1020
1021 struct token * tokenize(const char *name, int fd, struct token *endtoken, const char **next_path)
1022 {
1023         struct token *begin, *end;
1024         stream_t stream;
1025         unsigned char buffer[BUFSIZE];
1026         int idx;
1027
1028         idx = init_stream(name, fd, next_path);
1029         if (idx < 0) {
1030                 // info(endtoken->pos, "File %s is const", name);
1031                 return endtoken;
1032         }
1033
1034         begin = setup_stream(&stream, idx, fd, buffer, 0);
1035         end = tokenize_stream(&stream);
1036         if (endtoken)
1037                 end->next = endtoken;
1038         return begin;
1039 }