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  *  Licensed under the Open Software License version 1.1
   9  */
  10 #include <stdio.h>
  11 #include <stdlib.h>
  12 #include <stdarg.h>
  13 #include <stddef.h>
  14 #include <string.h>
  15 #include <ctype.h>
  16 #include <unistd.h>
  17
  18 #include "lib.h"
  19 #include "allocate.h"
  20 #include "token.h"
  21 #include "symbol.h"
  22
  23 #define EOF (-1)
  24
  25 int input_stream_nr = 0;
  26 struct stream *input_streams;
  27 static int input_streams_allocated;
  28 unsigned int tabstop = 8;
  29 int no_lineno = 0;
  30
  31 #define BUFSIZE (8192)
  32
  33 typedef struct {
  34         int fd, offset, size;
  35         int pos, line, nr;
  36         int newline, whitespace;
  37         struct token **tokenlist;
  38         struct token *token;
  39         unsigned char *buffer;
  40 } stream_t;
  41
  42 const char *stream_name(int stream)
  43 {
  44         if (stream < 0 || stream > input_stream_nr)
  45                 return "<bad stream>";
  46         return input_streams[stream].name;
  47 }
  48
  49 static struct position stream_pos(stream_t *stream)
  50 {
  51         struct position pos;
  52         pos.type = 0;
  53         pos.stream = stream->nr;
  54         pos.newline = stream->newline;
  55         pos.whitespace = stream->whitespace;
  56         pos.pos = stream->pos;
  57
  58         pos.line = stream->line;
  59         if (no_lineno)
  60                 pos.line = 123456;
  61
  62         pos.noexpand = 0;
  63         return pos;
  64 }
  65
  66 const char *show_special(int val)
  67 {
  68         static char buffer[4];
  69
  70         buffer[0] = val;
  71         buffer[1] = 0;
  72         if (val >= SPECIAL_BASE)
  73                 strcpy(buffer, (char *) combinations[val - SPECIAL_BASE]);
  74         return buffer;
  75 }
  76
  77 const char *show_ident(const struct ident *ident)
  78 {
  79         static char buffer[256];
  80         if (!ident)
  81                 return "<noident>";
  82         sprintf(buffer, "%.*s", ident->len, ident->name);
  83         return buffer;
  84 }
  85
  86 static char *charstr(char *ptr, unsigned char c, unsigned char escape, unsigned char next)
  87 {
  88         if (isprint(c)) {
  89                 if (c == escape || c == '\\')
  90                         *ptr++ = '\\';
  91                 *ptr++ = c;
  92                 return ptr;
  93         }
  94         *ptr++ = '\\';
  95         switch (c) {
  96         case '\n':
  97                 *ptr++ = 'n';
  98                 return ptr;
  99         case '\t':
 100                 *ptr++ = 't';
 101                 return ptr;
 102         }
 103         if (!isdigit(next))
 104                 return ptr + sprintf(ptr, "%o", c);
 105
 106         return ptr + sprintf(ptr, "%03o", c);
 107 }
 108
 109 const char *show_string(const struct string *string)
 110 {
 111         static char buffer[4 * MAX_STRING + 3];
 112         char *ptr;
 113         int i;
 114
 115         if (!string->length)
 116                 return "<bad_string>";
 117         ptr = buffer;
 118         *ptr++ = '"';
 119         for (i = 0; i < string->length-1; i++) {
 120                 const char *p = string->data + i;
 121                 ptr = charstr(ptr, p[0], '"', p[1]);
 122         }
 123         *ptr++ = '"';
 124         *ptr = '\0';
 125         return buffer;
 126 }
 127
 128 const char *show_token(const struct token *token)
 129 {
 130         static char buffer[256];
 131
 132         if (!token)
 133                 return "<no token>";
 134         switch (token_type(token)) {
 135         case TOKEN_ERROR:
 136                 return "syntax error";
 137
 138         case TOKEN_EOF:
 139                 return "end-of-input";
 140
 141         case TOKEN_IDENT:
 142                 return show_ident(token->ident);
 143
 144         case TOKEN_STRING:
 145                 return show_string(token->string);
 146
 147         case TOKEN_NUMBER:
 148                 return token->number;
 149
 150         case TOKEN_SPECIAL:
 151                 return show_special(token->special);
 152
 153         case TOKEN_CHAR: {
 154                 char *ptr = buffer;
 155                 int c = token->character;
 156                 *ptr++ = '\'';
 157                 ptr = charstr(ptr, c, '\'', 0);
 158                 *ptr++ = '\'';
 159                 *ptr++ = '\0';
 160                 return buffer;
 161         }
 162
 163         case TOKEN_STREAMBEGIN:
 164                 sprintf(buffer, "<beginning of '%s'>", stream_name(token->pos.stream));
 165                 return buffer;
 166
 167         case TOKEN_STREAMEND:
 168                 sprintf(buffer, "<end of '%s'>", stream_name(token->pos.stream));
 169                 return buffer;
 170
 171         case TOKEN_UNTAINT:
 172                 sprintf(buffer, "<untaint>");
 173                 return buffer;
 174
 175         case TOKEN_ARG_COUNT:
 176                 sprintf(buffer, "<argcnt>");
 177                 return buffer;
 178
 179         default:
 180                 sprintf(buffer, "unhandled token type '%d' ", token_type(token));
 181                 return buffer;
 182         }
 183 }
 184
 185 int init_stream(const char *name, int fd, const char **next_path)
 186 {
 187         int stream = input_stream_nr;
 188         struct stream *current;
 189
 190         if (stream >= input_streams_allocated) {
 191                 int newalloc = stream * 4 / 3 + 10;
 192                 input_streams = realloc(input_streams, newalloc * sizeof(struct stream));
 193                 if (!input_streams)
 194                         die("Unable to allocate more streams space");
 195                 input_streams_allocated = newalloc;
 196         }
 197         current = input_streams + stream;
 198         memset(current, 0, sizeof(*current));
 199         current->name = name;
 200         current->fd = fd;
 201         current->next_path = next_path;
 202         current->path = NULL;
 203         current->constant = CONSTANT_FILE_MAYBE;
 204         input_stream_nr = stream+1;
 205         return stream;
 206 }
 207
 208 static struct token * alloc_token(stream_t *stream)
 209 {
 210         struct token *token = __alloc_token(0);
 211         token->pos = stream_pos(stream);
 212         return token;
 213 }
 214
 215 /*
 216  *  Argh...  That was surprisingly messy - handling '\r' complicates the
 217  *  things a _lot_.
 218  */
 219 static int nextchar_slow(stream_t *stream)
 220 {
 221         int offset = stream->offset;
 222         int size = stream->size;
 223         int c;
 224         int spliced = 0, had_cr, had_backslash, complain;
 225
 226 restart:
 227         had_cr = had_backslash = complain = 0;
 228
 229 repeat:
 230         if (offset >= size) {
 231                 if (stream->fd < 0)
 232                         goto got_eof;
 233                 size = read(stream->fd, stream->buffer, BUFSIZE);
 234                 if (size <= 0)
 235                         goto got_eof;
 236                 stream->size = size;
 237                 stream->offset = offset = 0;
 238         }
 239
 240         c = stream->buffer[offset++];
 241
 242         if (had_cr && c != '\n')
 243                 complain = 1;
 244
 245         if (c == '\r') {
 246                 had_cr = 1;
 247                 goto repeat;
 248         }
 249
 250         stream->pos += (c == '\t') ? (tabstop - stream->pos % tabstop) : 1;
 251
 252         if (c == '\n') {
 253                 stream->line++;
 254                 stream->pos = 0;
 255         }
 256
 257         if (!had_backslash) {
 258                 if (c == '\\') {
 259                         had_backslash = 1;
 260                         goto repeat;
 261                 }
 262                 if (c == '\n')
 263                         stream->newline = 1;
 264         } else {
 265                 if (c == '\n') {
 266                         if (complain)
 267                                 warning(stream_pos(stream), "non-ASCII data stream");
 268                         spliced = 1;
 269                         goto restart;
 270                 }
 271                 stream->pos--;
 272                 offset--;
 273                 c = '\\';
 274         }
 275
 276 out:
 277         stream->offset = offset;
 278         if (complain)
 279                 warning(stream_pos(stream), "non-ASCII data stream");
 280
 281         return c;
 282
 283 got_eof:
 284         if (had_backslash) {
 285                 c = '\\';
 286                 goto out;
 287         }
 288         if (stream->pos)
 289                 warning(stream_pos(stream), "no newline at end of file");
 290         else if (had_cr)
 291                 warning(stream_pos(stream), "non-ASCII data stream");
 292         else if (spliced)
 293                 warning(stream_pos(stream), "backslash-newline at end of file");
 294         return EOF;
 295 }
 296
 297 /*
 298  *  We want that as light as possible while covering all normal cases.
 299  *  Slow path (including the logics with line-splicing and EOF sanity
 300  *  checks) is in nextchar_slow().
 301  */
 302 static inline int nextchar(stream_t *stream)
 303 {
 304         int offset = stream->offset;
 305
 306         if (offset < stream->size) {
 307                 int c = stream->buffer[offset++];
 308                 static const char special[256] = {
 309                         ['\t'] = 1, ['\r'] = 1, ['\n'] = 1, ['\\'] = 1
 310                 };
 311                 if (!special[c]) {
 312                         stream->offset = offset;
 313                         stream->pos++;
 314                         return c;
 315                 }
 316         }
 317         return nextchar_slow(stream);
 318 }
 319
 320 struct token eof_token_entry;
 321
 322 static struct token *mark_eof(stream_t *stream)
 323 {
 324         struct token *end;
 325
 326         end = alloc_token(stream);
 327         token_type(end) = TOKEN_STREAMEND;
 328         end->pos.newline = 1;
 329
 330         eof_token_entry.next = &eof_token_entry;
 331         eof_token_entry.pos.newline = 1;
 332
 333         end->next =  &eof_token_entry;
 334         *stream->tokenlist = end;
 335         stream->tokenlist = NULL;
 336         return end;
 337 }
 338
 339 static void add_token(stream_t *stream)
 340 {
 341         struct token *token = stream->token;
 342
 343         stream->token = NULL;
 344         token->next = NULL;
 345         *stream->tokenlist = token;
 346         stream->tokenlist = &token->next;
 347 }
 348
 349 static void drop_token(stream_t *stream)
 350 {
 351         stream->newline |= stream->token->pos.newline;
 352         stream->whitespace |= stream->token->pos.whitespace;
 353         stream->token = NULL;
 354 }
 355
 356 enum {
 357         Letter = 1,
 358         Digit = 2,
 359         Hex = 4,
 360         Exp = 8,
 361         Dot = 16,
 362         ValidSecond = 32,
 363 };
 364
 365 static const long cclass[257] = {
 366         ['0' + 1 ... '9' + 1] = Digit | Hex,
 367         ['A' + 1 ... 'D' + 1] = Letter | Hex,
 368         ['E' + 1] = Letter | Hex | Exp,
 369         ['F' + 1] = Letter | Hex,
 370         ['G' + 1 ... 'O' + 1] = Letter,
 371         ['P' + 1] = Letter | Exp,
 372         ['Q' + 1 ... 'Z' + 1] = Letter,
 373         ['a' + 1 ... 'd' + 1] = Letter | Hex,
 374         ['e' + 1] = Letter | Hex | Exp,
 375         ['f' + 1] = Letter | Hex,
 376         ['g' + 1 ... 'o' + 1] = Letter,
 377         ['p' + 1] = Letter | Exp,
 378         ['q' + 1 ... 'z' + 1] = Letter,
 379         ['_' + 1] = Letter,
 380         ['.' + 1] = Dot | ValidSecond,
 381         ['=' + 1] = ValidSecond,
 382         ['+' + 1] = ValidSecond,
 383         ['-' + 1] = ValidSecond,
 384         ['>' + 1] = ValidSecond,
 385         ['<' + 1] = ValidSecond,
 386         ['&' + 1] = ValidSecond,
 387         ['|' + 1] = ValidSecond,
 388         ['#' + 1] = ValidSecond,
 389 };
 390
 391 /*
 392  * pp-number:
 393  *      digit
 394  *      . digit
 395  *      pp-number digit
 396  *      pp-number identifier-nodigit
 397  *      pp-number e sign
 398  *      pp-number E sign
 399  *      pp-number p sign
 400  *      pp-number P sign
 401  *      pp-number .
 402  */
 403 static int get_one_number(int c, int next, stream_t *stream)
 404 {
 405         struct token *token;
 406         static char buffer[4095];
 407         char *p = buffer, *buf, *buffer_end = buffer + sizeof (buffer);
 408         int len;
 409
 410         *p++ = c;
 411         for (;;) {
 412                 long class =  cclass[next + 1];
 413                 if (!(class & (Dot | Digit | Letter)))
 414                         break;
 415                 if (p != buffer_end)
 416                         *p++ = next;
 417                 next = nextchar(stream);
 418                 if (class & Exp) {
 419                         if (next == '-' || next == '+') {
 420                                 if (p != buffer_end)
 421                                         *p++ = next;
 422                                 next = nextchar(stream);
 423                         }
 424                 }
 425         }
 426
 427         if (p == buffer_end) {
 428                 sparse_error(stream_pos(stream), "number token exceeds %td characters",
 429                       buffer_end - buffer);
 430                 // Pretend we saw just "1".
 431                 buffer[0] = '1';
 432                 p = buffer + 1;
 433         }
 434
 435         *p++ = 0;
 436         len = p - buffer;
 437         buf = __alloc_bytes(len);
 438         memcpy(buf, buffer, len);
 439
 440         token = stream->token;
 441         token_type(token) = TOKEN_NUMBER;
 442         token->number = buf;
 443         add_token(stream);
 444
 445         return next;
 446 }
 447
 448 static int escapechar(int first, int type, stream_t *stream, int *valp)
 449 {
 450         int next, value;
 451
 452         next = nextchar(stream);
 453         value = first;
 454
 455         if (first == '\n')
 456                 warning(stream_pos(stream), "Newline in string or character constant");
 457
 458         if (first == '\\' && next != EOF) {
 459                 value = next;
 460                 next = nextchar(stream);
 461                 if (value != type) {
 462                         switch (value) {
 463                         case 'a':
 464                                 value = '\a';
 465                                 break;
 466                         case 'b':
 467                                 value = '\b';
 468                                 break;
 469                         case 't':
 470                                 value = '\t';
 471                                 break;
 472                         case 'n':
 473                                 value = '\n';
 474                                 break;
 475                         case 'v':
 476                                 value = '\v';
 477                                 break;
 478                         case 'f':
 479                                 value = '\f';
 480                                 break;
 481                         case 'r':
 482                                 value = '\r';
 483                                 break;
 484                         case 'e':
 485                                 value = '\e';
 486                                 break;
 487                         case '\\':
 488                                 break;
 489                         case '?':
 490                                 break;
 491                         case '\'':
 492                                 break;
 493                         case '"':
 494                                 break;
 495                         case '\n':
 496                                 warning(stream_pos(stream), "Newline in string or character constant");
 497                                 break;
 498                         case '0'...'7': {
 499                                 int nr = 2;
 500                                 value -= '0';
 501                                 while (next >= '0' && next <= '9') {
 502                                         value = (value << 3) + (next-'0');
 503                                         next = nextchar(stream);
 504                                         if (!--nr)
 505                                                 break;
 506                                 }
 507                                 value &= 0xff;
 508                                 break;
 509                         }
 510                         case 'x': {
 511                                 int hex = hexval(next);
 512                                 if (hex < 16) {
 513                                         value = hex;
 514                                         next = nextchar(stream);
 515                                         while ((hex = hexval(next)) < 16) {
 516                                                 value = (value << 4) + hex;
 517                                                 next = nextchar(stream);
 518                                         }
 519                                         value &= 0xff;
 520                                         break;
 521                                 }
 522                         }
 523                         /* Fall through */
 524                         default:
 525                                 warning(stream_pos(stream), "Unknown escape '%c'", value);
 526                         }
 527                 }
 528                 /* Mark it as escaped */
 529                 value |= 0x100;
 530         }
 531         *valp = value;
 532         return next;
 533 }
 534
 535 static int get_char_token(int next, stream_t *stream)
 536 {
 537         int value;
 538         struct token *token;
 539
 540         next = escapechar(next, '\'', stream, &value);
 541         if (value == '\'' || next != '\'') {
 542                 sparse_error(stream_pos(stream), "Bad character constant");
 543                 drop_token(stream);
 544                 return next;
 545         }
 546
 547         token = stream->token;
 548         token_type(token) = TOKEN_CHAR;
 549         token->character = value & 0xff;
 550
 551         add_token(stream);
 552         return nextchar(stream);
 553 }
 554
 555 static int get_string_token(int next, stream_t *stream)
 556 {
 557         static char buffer[MAX_STRING];
 558         struct string *string;
 559         struct token *token;
 560         int len = 0;
 561
 562         for (;;) {
 563                 int val;
 564                 next = escapechar(next, '"', stream, &val);
 565                 if (val == '"')
 566                         break;
 567                 if (next == EOF) {
 568                         warning(stream_pos(stream), "End of file in middle of string");
 569                         return next;
 570                 }
 571                 if (len < MAX_STRING)
 572                         buffer[len] = val;
 573                 len++;
 574         }
 575
 576         if (len > MAX_STRING) {
 577                 warning(stream_pos(stream), "string too long (%d bytes, %d bytes max)", len, MAX_STRING);
 578                 len = MAX_STRING;
 579         }
 580
 581         string = __alloc_string(len+1);
 582         memcpy(string->data, buffer, len);
 583         string->data[len] = '\0';
 584         string->length = len+1;
 585
 586         /* Pass it on.. */
 587         token = stream->token;
 588         token_type(token) = TOKEN_STRING;
 589         token->string = string;
 590         add_token(stream);
 591
 592         return next;
 593 }
 594
 595 static int drop_stream_eoln(stream_t *stream)
 596 {
 597         drop_token(stream);
 598         for (;;) {
 599                 switch (nextchar(stream)) {
 600                 case EOF:
 601                         return EOF;
 602                 case '\n':
 603                         return nextchar(stream);
 604                 }
 605         }
 606 }
 607
 608 static int drop_stream_comment(stream_t *stream)
 609 {
 610         int newline;
 611         int next;
 612         drop_token(stream);
 613         newline = stream->newline;
 614
 615         next = nextchar(stream);
 616         for (;;) {
 617                 int curr = next;
 618                 if (curr == EOF) {
 619                         warning(stream_pos(stream), "End of file in the middle of a comment");
 620                         return curr;
 621                 }
 622                 next = nextchar(stream);
 623                 if (curr == '*' && next == '/')
 624                         break;
 625         }
 626         stream->newline = newline;
 627         return nextchar(stream);
 628 }
 629
 630 unsigned char combinations[][4] = COMBINATION_STRINGS;
 631
 632 #define NR_COMBINATIONS (SPECIAL_ARG_SEPARATOR - SPECIAL_BASE)
 633
 634 /* hash function for two-character punctuators - all give unique values */
 635 #define special_hash(c0, c1) (((c0*8+c1*2)+((c0*8+c1*2)>>5))&31)
 636
 637 /*
 638  * note that we won't get false positives - special_hash(0,0) is 0 and
 639  * entry 0 is filled (by +=), so all the missing ones are OK.
 640  */
 641 static unsigned char hash_results[32][2] = {
 642 #define RES(c0, c1) [special_hash(c0, c1)] = {c0, c1}
 643         RES('+', '='), /* 00 */
 644         RES('/', '='), /* 01 */
 645         RES('^', '='), /* 05 */
 646         RES('&', '&'), /* 07 */
 647         RES('#', '#'), /* 08 */
 648         RES('<', '<'), /* 0a */
 649         RES('<', '='), /* 0c */
 650         RES('!', '='), /* 0e */
 651         RES('%', '='), /* 0f */
 652         RES('-', '-'), /* 10 */
 653         RES('-', '='), /* 11 */
 654         RES('-', '>'), /* 13 */
 655         RES('=', '='), /* 15 */
 656         RES('&', '='), /* 17 */
 657         RES('*', '='), /* 18 */
 658         RES('.', '.'), /* 1a */
 659         RES('+', '+'), /* 1b */
 660         RES('|', '='), /* 1c */
 661         RES('>', '='), /* 1d */
 662         RES('|', '|'), /* 1e */
 663         RES('>', '>')  /* 1f */
 664 #undef RES
 665 };
 666 static int code[32] = {
 667 #define CODE(c0, c1, value) [special_hash(c0, c1)] = value
 668         CODE('+', '=', SPECIAL_ADD_ASSIGN), /* 00 */
 669         CODE('/', '=', SPECIAL_DIV_ASSIGN), /* 01 */
 670         CODE('^', '=', SPECIAL_XOR_ASSIGN), /* 05 */
 671         CODE('&', '&', SPECIAL_LOGICAL_AND), /* 07 */
 672         CODE('#', '#', SPECIAL_HASHHASH), /* 08 */
 673         CODE('<', '<', SPECIAL_LEFTSHIFT), /* 0a */
 674         CODE('<', '=', SPECIAL_LTE), /* 0c */
 675         CODE('!', '=', SPECIAL_NOTEQUAL), /* 0e */
 676         CODE('%', '=', SPECIAL_MOD_ASSIGN), /* 0f */
 677         CODE('-', '-', SPECIAL_DECREMENT), /* 10 */
 678         CODE('-', '=', SPECIAL_SUB_ASSIGN), /* 11 */
 679         CODE('-', '>', SPECIAL_DEREFERENCE), /* 13 */
 680         CODE('=', '=', SPECIAL_EQUAL), /* 15 */
 681         CODE('&', '=', SPECIAL_AND_ASSIGN), /* 17 */
 682         CODE('*', '=', SPECIAL_MUL_ASSIGN), /* 18 */
 683         CODE('.', '.', SPECIAL_DOTDOT), /* 1a */
 684         CODE('+', '+', SPECIAL_INCREMENT), /* 1b */
 685         CODE('|', '=', SPECIAL_OR_ASSIGN), /* 1c */
 686         CODE('>', '=', SPECIAL_GTE), /* 1d */
 687         CODE('|', '|', SPECIAL_LOGICAL_OR), /* 1e */
 688         CODE('>', '>', SPECIAL_RIGHTSHIFT)  /* 1f */
 689 #undef CODE
 690 };
 691
 692 static int get_one_special(int c, stream_t *stream)
 693 {
 694         struct token *token;
 695         int next, value, i;
 696
 697         next = nextchar(stream);
 698
 699         /*
 700          * Check for numbers, strings, character constants, and comments
 701          */
 702         switch (c) {
 703         case '.':
 704                 if (next >= '0' && next <= '9')
 705                         return get_one_number(c, next, stream);
 706                 break;
 707         case '"':
 708                 return get_string_token(next, stream);
 709         case '\'':
 710                 return get_char_token(next, stream);
 711         case '/':
 712                 if (next == '/')
 713                         return drop_stream_eoln(stream);
 714                 if (next == '*')
 715                         return drop_stream_comment(stream);
 716         }
 717
 718         /*
 719          * Check for combinations
 720          */
 721         value = c;
 722         if (cclass[next + 1] & ValidSecond) {
 723                 i = special_hash(c, next);
 724                 if (hash_results[i][0] == c && hash_results[i][1] == next) {
 725                         value = code[i];
 726                         next = nextchar(stream);
 727                         if (value >= SPECIAL_LEFTSHIFT &&
 728                             next == "==."[value - SPECIAL_LEFTSHIFT]) {
 729                                 value += 3;
 730                                 next = nextchar(stream);
 731                         }
 732                 }
 733         }
 734
 735         /* Pass it on.. */
 736         token = stream->token;
 737         token_type(token) = TOKEN_SPECIAL;
 738         token->special = value;
 739         add_token(stream);
 740         return next;
 741 }
 742
 743 #define IDENT_HASH_BITS (13)
 744 #define IDENT_HASH_SIZE (1<<IDENT_HASH_BITS)
 745 #define IDENT_HASH_MASK (IDENT_HASH_SIZE-1)
 746
 747 #define ident_hash_init(c)              (c)
 748 #define ident_hash_add(oldhash,c)       ((oldhash)*11 + (c))
 749 #define ident_hash_end(hash)            ((((hash) >> IDENT_HASH_BITS) + (hash)) & IDENT_HASH_MASK)
 750
 751 static struct ident *hash_table[IDENT_HASH_SIZE];
 752 static int ident_hit, ident_miss, idents;
 753
 754 void show_identifier_stats(void)
 755 {
 756         int i;
 757         int distribution[100];
 758
 759         fprintf(stderr, "identifiers: %d hits, %d misses\n",
 760                 ident_hit, ident_miss);
 761
 762         for (i = 0; i < 100; i++)
 763                 distribution[i] = 0;
 764
 765         for (i = 0; i < IDENT_HASH_SIZE; i++) {
 766                 struct ident * ident = hash_table[i];
 767                 int count = 0;
 768
 769                 while (ident) {
 770                         count++;
 771                         ident = ident->next;
 772                 }
 773                 if (count > 99)
 774                         count = 99;
 775                 distribution[count]++;
 776         }
 777
 778         for (i = 0; i < 100; i++) {
 779                 if (distribution[i])
 780                         fprintf(stderr, "%2d: %d buckets\n", i, distribution[i]);
 781         }
 782 }
 783
 784 static struct ident *alloc_ident(const char *name, int len)
 785 {
 786         struct ident *ident = __alloc_ident(len);
 787         ident->symbols = NULL;
 788         ident->len = len;
 789         ident->tainted = 0;
 790         memcpy(ident->name, name, len);
 791         return ident;
 792 }
 793
 794 static struct ident * insert_hash(struct ident *ident, unsigned long hash)
 795 {
 796         ident->next = hash_table[hash];
 797         hash_table[hash] = ident;
 798         ident_miss++;
 799         return ident;
 800 }
 801
 802 static struct ident *create_hashed_ident(const char *name, int len, unsigned long hash)
 803 {
 804         struct ident *ident;
 805         struct ident **p;
 806
 807         p = &hash_table[hash];
 808         while ((ident = *p) != NULL) {
 809                 if (ident->len == (unsigned char) len) {
 810                         if (strncmp(name, ident->name, len) != 0)
 811                                 goto next;
 812
 813                         ident_hit++;
 814                         return ident;
 815                 }
 816 next:
 817                 //misses++;
 818                 p = &ident->next;
 819         }
 820         ident = alloc_ident(name, len);
 821         *p = ident;
 822         ident->next = NULL;
 823         ident_miss++;
 824         idents++;
 825         return ident;
 826 }
 827
 828 static unsigned long hash_name(const char *name, int len)
 829 {
 830         unsigned long hash;
 831         const unsigned char *p = (const unsigned char *)name;
 832
 833         hash = ident_hash_init(*p++);
 834         while (--len) {
 835                 unsigned int i = *p++;
 836                 hash = ident_hash_add(hash, i);
 837         }
 838         return ident_hash_end(hash);
 839 }
 840
 841 struct ident *hash_ident(struct ident *ident)
 842 {
 843         return insert_hash(ident, hash_name(ident->name, ident->len));
 844 }
 845
 846 struct ident *built_in_ident(const char *name)
 847 {
 848         int len = strlen(name);
 849         return create_hashed_ident(name, len, hash_name(name, len));
 850 }
 851
 852 struct token *built_in_token(int stream, const char *name)
 853 {
 854         struct token *token;
 855
 856         token = __alloc_token(0);
 857         token->pos.stream = stream;
 858         token_type(token) = TOKEN_IDENT;
 859         token->ident = built_in_ident(name);
 860         return token;
 861 }
 862
 863 static int get_one_identifier(int c, stream_t *stream)
 864 {
 865         struct token *token;
 866         struct ident *ident;
 867         unsigned long hash;
 868         char buf[256];
 869         int len = 1;
 870         int next;
 871
 872         hash = ident_hash_init(c);
 873         buf[0] = c;
 874         for (;;) {
 875                 next = nextchar(stream);
 876                 if (!(cclass[next + 1] & (Letter | Digit)))
 877                         break;
 878                 if (len >= sizeof(buf))
 879                         break;
 880                 hash = ident_hash_add(hash, next);
 881                 buf[len] = next;
 882                 len++;
 883         };
 884         hash = ident_hash_end(hash);
 885
 886         ident = create_hashed_ident(buf, len, hash);
 887
 888         /* Pass it on.. */
 889         token = stream->token;
 890         token_type(token) = TOKEN_IDENT;
 891         token->ident = ident;
 892         add_token(stream);
 893         return next;
 894 }
 895
 896 static int get_one_token(int c, stream_t *stream)
 897 {
 898         long class = cclass[c + 1];
 899         if (class & Digit)
 900                 return get_one_number(c, nextchar(stream), stream);
 901         if (class & Letter)
 902                 return get_one_identifier(c, stream);
 903         return get_one_special(c, stream);
 904 }
 905
 906 static struct token *setup_stream(stream_t *stream, int idx, int fd,
 907         unsigned char *buf, unsigned int buf_size)
 908 {
 909         struct token *begin;
 910
 911         stream->nr = idx;
 912         stream->line = 1;
 913         stream->newline = 1;
 914         stream->whitespace = 0;
 915         stream->pos = 0;
 916
 917         stream->token = NULL;
 918         stream->fd = fd;
 919         stream->offset = 0;
 920         stream->size = buf_size;
 921         stream->buffer = buf;
 922
 923         begin = alloc_token(stream);
 924         token_type(begin) = TOKEN_STREAMBEGIN;
 925         stream->tokenlist = &begin->next;
 926         return begin;
 927 }
 928
 929 static struct token *tokenize_stream(stream_t *stream)
 930 {
 931         int c = nextchar(stream);
 932         while (c != EOF) {
 933                 if (!isspace(c)) {
 934                         struct token *token = alloc_token(stream);
 935                         stream->token = token;
 936                         stream->newline = 0;
 937                         stream->whitespace = 0;
 938                         c = get_one_token(c, stream);
 939                         continue;
 940                 }
 941                 stream->whitespace = 1;
 942                 c = nextchar(stream);
 943         }
 944         return mark_eof(stream);
 945 }
 946
 947 struct token * tokenize_buffer(void *buffer, unsigned long size, struct token **endtoken)
 948 {
 949         stream_t stream;
 950         struct token *begin;
 951
 952         begin = setup_stream(&stream, 0, -1, buffer, size);
 953         *endtoken = tokenize_stream(&stream);
 954         return begin;
 955 }
 956
 957 struct token * tokenize(const char *name, int fd, struct token *endtoken, const char **next_path)
 958 {
 959         struct token *begin, *end;
 960         stream_t stream;
 961         unsigned char buffer[BUFSIZE];
 962         int idx;
 963
 964         idx = init_stream(name, fd, next_path);
 965         if (idx < 0) {
 966                 // info(endtoken->pos, "File %s is const", name);
 967                 return endtoken;
 968         }
 969
 970         begin = setup_stream(&stream, idx, fd, buffer, 0);
 971         end = tokenize_stream(&stream);
 972         if (endtoken)
 973                 end->next = endtoken;
 974         return begin;
 975 }