deflate.c
1 /* deflate.c -- compress data using the deflation algorithm 2 * Copyright (C) 1995-2017 Jean-loup Gailly and Mark Adler 3 * For conditions of distribution and use, see copyright notice in zlib.h 4 */ 5 6 /* 7 * ALGORITHM 8 * 9 * The "deflation" process depends on being able to identify portions 10 * of the input text which are identical to earlier input (within a 11 * sliding window trailing behind the input currently being processed). 12 * 13 * The most straightforward technique turns out to be the fastest for 14 * most input files: try all possible matches and select the longest. 15 * The key feature of this algorithm is that insertions into the string 16 * dictionary are very simple and thus fast, and deletions are avoided 17 * completely. Insertions are performed at each input character, whereas 18 * string matches are performed only when the previous match ends. So it 19 * is preferable to spend more time in matches to allow very fast string 20 * insertions and avoid deletions. The matching algorithm for small 21 * strings is inspired from that of Rabin & Karp. A brute force approach 22 * is used to find longer strings when a small match has been found. 23 * A similar algorithm is used in comic (by Jan-Mark Wams) and freeze 24 * (by Leonid Broukhis). 25 * A previous version of this file used a more sophisticated algorithm 26 * (by Fiala and Greene) which is guaranteed to run in linear amortized 27 * time, but has a larger average cost, uses more memory and is patented. 28 * However the F&G algorithm may be faster for some highly redundant 29 * files if the parameter max_chain_length (described below) is too large. 30 * 31 * ACKNOWLEDGEMENTS 32 * 33 * The idea of lazy evaluation of matches is due to Jan-Mark Wams, and 34 * I found it in 'freeze' written by Leonid Broukhis. 35 * Thanks to many people for bug reports and testing. 36 * 37 * REFERENCES 38 * 39 * Deutsch, L.P.,"DEFLATE Compressed Data Format Specification". 40 * Available in http://tools.ietf.org/html/rfc1951 41 * 42 * A description of the Rabin and Karp algorithm is given in the book 43 * "Algorithms" by R. Sedgewick, Addison-Wesley, p252. 44 * 45 * Fiala,E.R., and Greene,D.H. 46 * Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595 47 * 48 */ 49 50 /* @(#) $Id$ */ 51 52 #include "deflate.h" 53 54 const char deflate_copyright[] = 55 " deflate 1.2.11 Copyright 1995-2017 Jean-loup Gailly and Mark Adler "; 56 /* 57 If you use the zlib library in a product, an acknowledgment is welcome 58 in the documentation of your product. If for some reason you cannot 59 include such an acknowledgment, I would appreciate that you keep this 60 copyright string in the executable of your product. 61 */ 62 63 /* =========================================================================== 64 * Function prototypes. 65 */ 66 typedef enum { 67 need_more, /* block not completed, need more input or more output */ 68 block_done, /* block flush performed */ 69 finish_started, /* finish started, need only more output at next deflate */ 70 finish_done /* finish done, accept no more input or output */ 71 } block_state; 72 73 typedef block_state (*compress_func) OF((deflate_state *s, int flush)); 74 /* Compression function. Returns the block state after the call. */ 75 76 local int deflateStateCheck OF((z_streamp strm)); 77 local void slide_hash OF((deflate_state *s)); 78 local void fill_window OF((deflate_state *s)); 79 local block_state deflate_stored OF((deflate_state *s, int flush)); 80 local block_state deflate_fast OF((deflate_state *s, int flush)); 81 #ifndef FASTEST 82 local block_state deflate_slow OF((deflate_state *s, int flush)); 83 #endif 84 local block_state deflate_rle OF((deflate_state *s, int flush)); 85 local block_state deflate_huff OF((deflate_state *s, int flush)); 86 local void lm_init OF((deflate_state *s)); 87 local void putShortMSB OF((deflate_state *s, uInt b)); 88 local void flush_pending OF((z_streamp strm)); 89 local unsigned read_buf OF((z_streamp strm, Bytef *buf, unsigned size)); 90 #ifdef ASMV 91 # pragma message("Assembler code may have bugs -- use at your own risk") 92 void match_init OF((void)); /* asm code initialization */ 93 uInt longest_match OF((deflate_state *s, IPos cur_match)); 94 #else 95 local uInt longest_match OF((deflate_state *s, IPos cur_match)); 96 #endif 97 98 #ifdef ZLIB_DEBUG 99 local void check_match OF((deflate_state *s, IPos start, IPos match, 100 int length)); 101 #endif 102 103 /* =========================================================================== 104 * Local data 105 */ 106 107 #define NIL 0 108 /* Tail of hash chains */ 109 110 #ifndef TOO_FAR 111 # define TOO_FAR 4096 112 #endif 113 /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */ 114 115 /* Values for max_lazy_match, good_match and max_chain_length, depending on 116 * the desired pack level (0..9). The values given below have been tuned to 117 * exclude worst case performance for pathological files. Better values may be 118 * found for specific files. 119 */ 120 typedef struct config_s { 121 ush good_length; /* reduce lazy search above this match length */ 122 ush max_lazy; /* do not perform lazy search above this match length */ 123 ush nice_length; /* quit search above this match length */ 124 ush max_chain; 125 compress_func func; 126 } config; 127 128 #ifdef FASTEST 129 local const config configuration_table[2] = { 130 /* good lazy nice chain */ 131 /* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */ 132 /* 1 */ {4, 4, 8, 4, deflate_fast}}; /* max speed, no lazy matches */ 133 #else 134 local const config configuration_table[10] = { 135 /* good lazy nice chain */ 136 /* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */ 137 /* 1 */ {4, 4, 8, 4, deflate_fast}, /* max speed, no lazy matches */ 138 /* 2 */ {4, 5, 16, 8, deflate_fast}, 139 /* 3 */ {4, 6, 32, 32, deflate_fast}, 140 141 /* 4 */ {4, 4, 16, 16, deflate_slow}, /* lazy matches */ 142 /* 5 */ {8, 16, 32, 32, deflate_slow}, 143 /* 6 */ {8, 16, 128, 128, deflate_slow}, 144 /* 7 */ {8, 32, 128, 256, deflate_slow}, 145 /* 8 */ {32, 128, 258, 1024, deflate_slow}, 146 /* 9 */ {32, 258, 258, 4096, deflate_slow}}; /* max compression */ 147 #endif 148 149 /* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4 150 * For deflate_fast() (levels <= 3) good is ignored and lazy has a different 151 * meaning. 152 */ 153 154 /* rank Z_BLOCK between Z_NO_FLUSH and Z_PARTIAL_FLUSH */ 155 #define RANK(f) (((f) * 2) - ((f) > 4 ? 9 : 0)) 156 157 /* =========================================================================== 158 * Update a hash value with the given input byte 159 * IN assertion: all calls to UPDATE_HASH are made with consecutive input 160 * characters, so that a running hash key can be computed from the previous 161 * key instead of complete recalculation each time. 162 */ 163 #define UPDATE_HASH(s,h,c) (h = (((h)<<s->hash_shift) ^ (c)) & s->hash_mask) 164 165 166 /* =========================================================================== 167 * Insert string str in the dictionary and set match_head to the previous head 168 * of the hash chain (the most recent string with same hash key). Return 169 * the previous length of the hash chain. 170 * If this file is compiled with -DFASTEST, the compression level is forced 171 * to 1, and no hash chains are maintained. 172 * IN assertion: all calls to INSERT_STRING are made with consecutive input 173 * characters and the first MIN_MATCH bytes of str are valid (except for 174 * the last MIN_MATCH-1 bytes of the input file). 175 */ 176 #ifdef FASTEST 177 #define INSERT_STRING(s, str, match_head) \ 178 (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \ 179 match_head = s->head[s->ins_h], \ 180 s->head[s->ins_h] = (Pos)(str)) 181 #else 182 #define INSERT_STRING(s, str, match_head) \ 183 (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \ 184 match_head = s->prev[(str) & s->w_mask] = s->head[s->ins_h], \ 185 s->head[s->ins_h] = (Pos)(str)) 186 #endif 187 188 /* =========================================================================== 189 * Initialize the hash table (avoiding 64K overflow for 16 bit systems). 190 * prev[] will be initialized on the fly. 191 */ 192 #define CLEAR_HASH(s) \ 193 s->head[s->hash_size-1] = NIL; \ 194 zmemzero((Bytef *)s->head, (unsigned)(s->hash_size-1)*sizeof(*s->head)); 195 196 /* =========================================================================== 197 * Slide the hash table when sliding the window down (could be avoided with 32 198 * bit values at the expense of memory usage). We slide even when level == 0 to 199 * keep the hash table consistent if we switch back to level > 0 later. 200 */ 201 local void slide_hash(s) 202 deflate_state *s; 203 { 204 unsigned n, m; 205 Posf *p; 206 uInt wsize = s->w_size; 207 208 n = s->hash_size; 209 p = &s->head[n]; 210 do { 211 m = *--p; 212 *p = (Pos)(m >= wsize ? m - wsize : NIL); 213 } while (--n); 214 n = wsize; 215 #ifndef FASTEST 216 p = &s->prev[n]; 217 do { 218 m = *--p; 219 *p = (Pos)(m >= wsize ? m - wsize : NIL); 220 /* If n is not on any hash chain, prev[n] is garbage but 221 * its value will never be used. 222 */ 223 } while (--n); 224 #endif 225 } 226 227 /* ========================================================================= */ 228 int ZEXPORT deflateInit_(strm, level, version, stream_size) 229 z_streamp strm; 230 int level; 231 const char *version; 232 int stream_size; 233 { 234 return deflateInit2_(strm, level, Z_DEFLATED, MAX_WBITS, DEF_MEM_LEVEL, 235 Z_DEFAULT_STRATEGY, version, stream_size); 236 /* To do: ignore strm->next_in if we use it as window */ 237 } 238 239 /* ========================================================================= */ 240 int ZEXPORT deflateInit2_(strm, level, method, windowBits, memLevel, strategy, 241 version, stream_size) 242 z_streamp strm; 243 int level; 244 int method; 245 int windowBits; 246 int memLevel; 247 int strategy; 248 const char *version; 249 int stream_size; 250 { 251 deflate_state *s; 252 int wrap = 1; 253 static const char my_version[] = ZLIB_VERSION; 254 255 ushf *overlay; 256 /* We overlay pending_buf and d_buf+l_buf. This works since the average 257 * output size for (length,distance) codes is <= 24 bits. 258 */ 259 260 if (version == Z_NULL || version[0] != my_version[0] || 261 stream_size != sizeof(z_stream)) { 262 return Z_VERSION_ERROR; 263 } 264 if (strm == Z_NULL) return Z_STREAM_ERROR; 265 266 strm->msg = Z_NULL; 267 if (strm->zalloc == (alloc_func)0) { 268 #ifdef Z_SOLO 269 return Z_STREAM_ERROR; 270 #else 271 strm->zalloc = zcalloc; 272 strm->opaque = (voidpf)0; 273 #endif 274 } 275 if (strm->zfree == (free_func)0) 276 #ifdef Z_SOLO 277 return Z_STREAM_ERROR; 278 #else 279 strm->zfree = zcfree; 280 #endif 281 282 #ifdef FASTEST 283 if (level != 0) level = 1; 284 #else 285 if (level == Z_DEFAULT_COMPRESSION) level = 6; 286 #endif 287 288 if (windowBits < 0) { /* suppress zlib wrapper */ 289 wrap = 0; 290 windowBits = -windowBits; 291 } 292 #ifdef GZIP 293 else if (windowBits > 15) { 294 wrap = 2; /* write gzip wrapper instead */ 295 windowBits -= 16; 296 } 297 #endif 298 if (memLevel < 1 || memLevel > MAX_MEM_LEVEL || method != Z_DEFLATED || 299 windowBits < 8 || windowBits > 15 || level < 0 || level > 9 || 300 strategy < 0 || strategy > Z_FIXED || (windowBits == 8 && wrap != 1)) { 301 return Z_STREAM_ERROR; 302 } 303 if (windowBits == 8) windowBits = 9; /* until 256-byte window bug fixed */ 304 s = (deflate_state *) ZALLOC(strm, 1, sizeof(deflate_state)); 305 if (s == Z_NULL) return Z_MEM_ERROR; 306 strm->state = (struct internal_state FAR *)s; 307 s->strm = strm; 308 s->status = INIT_STATE; /* to pass state test in deflateReset() */ 309 310 s->wrap = wrap; 311 s->gzhead = Z_NULL; 312 s->w_bits = (uInt)windowBits; 313 s->w_size = 1 << s->w_bits; 314 s->w_mask = s->w_size - 1; 315 316 s->hash_bits = (uInt)memLevel + 7; 317 s->hash_size = 1 << s->hash_bits; 318 s->hash_mask = s->hash_size - 1; 319 s->hash_shift = ((s->hash_bits+MIN_MATCH-1)/MIN_MATCH); 320 321 s->window = (Bytef *) ZALLOC(strm, s->w_size, 2*sizeof(Byte)); 322 s->prev = (Posf *) ZALLOC(strm, s->w_size, sizeof(Pos)); 323 s->head = (Posf *) ZALLOC(strm, s->hash_size, sizeof(Pos)); 324 325 s->high_water = 0; /* nothing written to s->window yet */ 326 327 s->lit_bufsize = 1 << (memLevel + 6); /* 16K elements by default */ 328 329 overlay = (ushf *) ZALLOC(strm, s->lit_bufsize, sizeof(ush)+2); 330 s->pending_buf = (uchf *) overlay; 331 s->pending_buf_size = (ulg)s->lit_bufsize * (sizeof(ush)+2L); 332 333 if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL || 334 s->pending_buf == Z_NULL) { 335 s->status = FINISH_STATE; 336 strm->msg = ERR_MSG(Z_MEM_ERROR); 337 deflateEnd (strm); 338 return Z_MEM_ERROR; 339 } 340 s->d_buf = overlay + s->lit_bufsize/sizeof(ush); 341 s->l_buf = s->pending_buf + (1+sizeof(ush))*s->lit_bufsize; 342 343 s->level = level; 344 s->strategy = strategy; 345 s->method = (Byte)method; 346 347 return deflateReset(strm); 348 } 349 350 /* ========================================================================= 351 * Check for a valid deflate stream state. Return 0 if ok, 1 if not. 352 */ 353 local int deflateStateCheck (strm) 354 z_streamp strm; 355 { 356 deflate_state *s; 357 if (strm == Z_NULL || 358 strm->zalloc == (alloc_func)0 || strm->zfree == (free_func)0) 359 return 1; 360 s = strm->state; 361 if (s == Z_NULL || s->strm != strm || (s->status != INIT_STATE && 362 #ifdef GZIP 363 s->status != GZIP_STATE && 364 #endif 365 s->status != EXTRA_STATE && 366 s->status != NAME_STATE && 367 s->status != COMMENT_STATE && 368 s->status != HCRC_STATE && 369 s->status != BUSY_STATE && 370 s->status != FINISH_STATE)) 371 return 1; 372 return 0; 373 } 374 375 /* ========================================================================= */ 376 int ZEXPORT deflateSetDictionary (strm, dictionary, dictLength) 377 z_streamp strm; 378 const Bytef *dictionary; 379 uInt dictLength; 380 { 381 deflate_state *s; 382 uInt str, n; 383 int wrap; 384 unsigned avail; 385 z_const unsigned char *next; 386 387 if (deflateStateCheck(strm) || dictionary == Z_NULL) 388 return Z_STREAM_ERROR; 389 s = strm->state; 390 wrap = s->wrap; 391 if (wrap == 2 || (wrap == 1 && s->status != INIT_STATE) || s->lookahead) 392 return Z_STREAM_ERROR; 393 394 /* when using zlib wrappers, compute Adler-32 for provided dictionary */ 395 if (wrap == 1) 396 strm->adler = adler32(strm->adler, dictionary, dictLength); 397 s->wrap = 0; /* avoid computing Adler-32 in read_buf */ 398 399 /* if dictionary would fill window, just replace the history */ 400 if (dictLength >= s->w_size) { 401 if (wrap == 0) { /* already empty otherwise */ 402 CLEAR_HASH(s); 403 s->strstart = 0; 404 s->block_start = 0L; 405 s->insert = 0; 406 } 407 dictionary += dictLength - s->w_size; /* use the tail */ 408 dictLength = s->w_size; 409 } 410 411 /* insert dictionary into window and hash */ 412 avail = strm->avail_in; 413 next = strm->next_in; 414 strm->avail_in = dictLength; 415 strm->next_in = (z_const Bytef *)dictionary; 416 fill_window(s); 417 while (s->lookahead >= MIN_MATCH) { 418 str = s->strstart; 419 n = s->lookahead - (MIN_MATCH-1); 420 do { 421 UPDATE_HASH(s, s->ins_h, s->window[str + MIN_MATCH-1]); 422 #ifndef FASTEST 423 s->prev[str & s->w_mask] = s->head[s->ins_h]; 424 #endif 425 s->head[s->ins_h] = (Pos)str; 426 str++; 427 } while (--n); 428 s->strstart = str; 429 s->lookahead = MIN_MATCH-1; 430 fill_window(s); 431 } 432 s->strstart += s->lookahead; 433 s->block_start = (long)s->strstart; 434 s->insert = s->lookahead; 435 s->lookahead = 0; 436 s->match_length = s->prev_length = MIN_MATCH-1; 437 s->match_available = 0; 438 strm->next_in = next; 439 strm->avail_in = avail; 440 s->wrap = wrap; 441 return Z_OK; 442 } 443 444 /* ========================================================================= */ 445 int ZEXPORT deflateGetDictionary (strm, dictionary, dictLength) 446 z_streamp strm; 447 Bytef *dictionary; 448 uInt *dictLength; 449 { 450 deflate_state *s; 451 uInt len; 452 453 if (deflateStateCheck(strm)) 454 return Z_STREAM_ERROR; 455 s = strm->state; 456 len = s->strstart + s->lookahead; 457 if (len > s->w_size) 458 len = s->w_size; 459 if (dictionary != Z_NULL && len) 460 zmemcpy(dictionary, s->window + s->strstart + s->lookahead - len, len); 461 if (dictLength != Z_NULL) 462 *dictLength = len; 463 return Z_OK; 464 } 465 466 /* ========================================================================= */ 467 int ZEXPORT deflateResetKeep (strm) 468 z_streamp strm; 469 { 470 deflate_state *s; 471 472 if (deflateStateCheck(strm)) { 473 return Z_STREAM_ERROR; 474 } 475 476 strm->total_in = strm->total_out = 0; 477 strm->msg = Z_NULL; /* use zfree if we ever allocate msg dynamically */ 478 strm->data_type = Z_UNKNOWN; 479 480 s = (deflate_state *)strm->state; 481 s->pending = 0; 482 s->pending_out = s->pending_buf; 483 484 if (s->wrap < 0) { 485 s->wrap = -s->wrap; /* was made negative by deflate(..., Z_FINISH); */ 486 } 487 s->status = 488 #ifdef GZIP 489 s->wrap == 2 ? GZIP_STATE : 490 #endif 491 s->wrap ? INIT_STATE : BUSY_STATE; 492 strm->adler = 493 #ifdef GZIP 494 s->wrap == 2 ? crc32(0L, Z_NULL, 0) : 495 #endif 496 adler32(0L, Z_NULL, 0); 497 s->last_flush = Z_NO_FLUSH; 498 499 _tr_init(s); 500 501 return Z_OK; 502 } 503 504 /* ========================================================================= */ 505 int ZEXPORT deflateReset (strm) 506 z_streamp strm; 507 { 508 int ret; 509 510 ret = deflateResetKeep(strm); 511 if (ret == Z_OK) 512 lm_init(strm->state); 513 return ret; 514 } 515 516 /* ========================================================================= */ 517 int ZEXPORT deflateSetHeader (strm, head) 518 z_streamp strm; 519 gz_headerp head; 520 { 521 if (deflateStateCheck(strm) || strm->state->wrap != 2) 522 return Z_STREAM_ERROR; 523 strm->state->gzhead = head; 524 return Z_OK; 525 } 526 527 /* ========================================================================= */ 528 int ZEXPORT deflatePending (strm, pending, bits) 529 unsigned *pending; 530 int *bits; 531 z_streamp strm; 532 { 533 if (deflateStateCheck(strm)) return Z_STREAM_ERROR; 534 if (pending != Z_NULL) 535 *pending = strm->state->pending; 536 if (bits != Z_NULL) 537 *bits = strm->state->bi_valid; 538 return Z_OK; 539 } 540 541 /* ========================================================================= */ 542 int ZEXPORT deflatePrime (strm, bits, value) 543 z_streamp strm; 544 int bits; 545 int value; 546 { 547 deflate_state *s; 548 int put; 549 550 if (deflateStateCheck(strm)) return Z_STREAM_ERROR; 551 s = strm->state; 552 if ((Bytef *)(s->d_buf) < s->pending_out + ((Buf_size + 7) >> 3)) 553 return Z_BUF_ERROR; 554 do { 555 put = Buf_size - s->bi_valid; 556 if (put > bits) 557 put = bits; 558 s->bi_buf |= (ush)((value & ((1 << put) - 1)) << s->bi_valid); 559 s->bi_valid += put; 560 _tr_flush_bits(s); 561 value >>= put; 562 bits -= put; 563 } while (bits); 564 return Z_OK; 565 } 566 567 /* ========================================================================= */ 568 int ZEXPORT deflateParams(strm, level, strategy) 569 z_streamp strm; 570 int level; 571 int strategy; 572 { 573 deflate_state *s; 574 compress_func func; 575 576 if (deflateStateCheck(strm)) return Z_STREAM_ERROR; 577 s = strm->state; 578 579 #ifdef FASTEST 580 if (level != 0) level = 1; 581 #else 582 if (level == Z_DEFAULT_COMPRESSION) level = 6; 583 #endif 584 if (level < 0 || level > 9 || strategy < 0 || strategy > Z_FIXED) { 585 return Z_STREAM_ERROR; 586 } 587 func = configuration_table[s->level].func; 588 589 if ((strategy != s->strategy || func != configuration_table[level].func) && 590 s->high_water) { 591 /* Flush the last buffer: */ 592 int err = deflate(strm, Z_BLOCK); 593 if (err == Z_STREAM_ERROR) 594 return err; 595 if (strm->avail_out == 0) 596 return Z_BUF_ERROR; 597 } 598 if (s->level != level) { 599 if (s->level == 0 && s->matches != 0) { 600 if (s->matches == 1) 601 slide_hash(s); 602 else 603 CLEAR_HASH(s); 604 s->matches = 0; 605 } 606 s->level = level; 607 s->max_lazy_match = configuration_table[level].max_lazy; 608 s->good_match = configuration_table[level].good_length; 609 s->nice_match = configuration_table[level].nice_length; 610 s->max_chain_length = configuration_table[level].max_chain; 611 } 612 s->strategy = strategy; 613 return Z_OK; 614 } 615 616 /* ========================================================================= */ 617 int ZEXPORT deflateTune(strm, good_length, max_lazy, nice_length, max_chain) 618 z_streamp strm; 619 int good_length; 620 int max_lazy; 621 int nice_length; 622 int max_chain; 623 { 624 deflate_state *s; 625 626 if (deflateStateCheck(strm)) return Z_STREAM_ERROR; 627 s = strm->state; 628 s->good_match = (uInt)good_length; 629 s->max_lazy_match = (uInt)max_lazy; 630 s->nice_match = nice_length; 631 s->max_chain_length = (uInt)max_chain; 632 return Z_OK; 633 } 634 635 /* ========================================================================= 636 * For the default windowBits of 15 and memLevel of 8, this function returns 637 * a close to exact, as well as small, upper bound on the compressed size. 638 * They are coded as constants here for a reason--if the #define's are 639 * changed, then this function needs to be changed as well. The return 640 * value for 15 and 8 only works for those exact settings. 641 * 642 * For any setting other than those defaults for windowBits and memLevel, 643 * the value returned is a conservative worst case for the maximum expansion 644 * resulting from using fixed blocks instead of stored blocks, which deflate 645 * can emit on compressed data for some combinations of the parameters. 646 * 647 * This function could be more sophisticated to provide closer upper bounds for 648 * every combination of windowBits and memLevel. But even the conservative 649 * upper bound of about 14% expansion does not seem onerous for output buffer 650 * allocation. 651 */ 652 uLong ZEXPORT deflateBound(strm, sourceLen) 653 z_streamp strm; 654 uLong sourceLen; 655 { 656 deflate_state *s; 657 uLong complen, wraplen; 658 659 /* conservative upper bound for compressed data */ 660 complen = sourceLen + 661 ((sourceLen + 7) >> 3) + ((sourceLen + 63) >> 6) + 5; 662 663 /* if can't get parameters, return conservative bound plus zlib wrapper */ 664 if (deflateStateCheck(strm)) 665 return complen + 6; 666 667 /* compute wrapper length */ 668 s = strm->state; 669 switch (s->wrap) { 670 case 0: /* raw deflate */ 671 wraplen = 0; 672 break; 673 case 1: /* zlib wrapper */ 674 wraplen = 6 + (s->strstart ? 4 : 0); 675 break; 676 #ifdef GZIP 677 case 2: /* gzip wrapper */ 678 wraplen = 18; 679 if (s->gzhead != Z_NULL) { /* user-supplied gzip header */ 680 Bytef *str; 681 if (s->gzhead->extra != Z_NULL) 682 wraplen += 2 + s->gzhead->extra_len; 683 str = s->gzhead->name; 684 if (str != Z_NULL) 685 do { 686 wraplen++; 687 } while (*str++); 688 str = s->gzhead->comment; 689 if (str != Z_NULL) 690 do { 691 wraplen++; 692 } while (*str++); 693 if (s->gzhead->hcrc) 694 wraplen += 2; 695 } 696 break; 697 #endif 698 default: /* for compiler happiness */ 699 wraplen = 6; 700 } 701 702 /* if not default parameters, return conservative bound */ 703 if (s->w_bits != 15 || s->hash_bits != 8 + 7) 704 return complen + wraplen; 705 706 /* default settings: return tight bound for that case */ 707 return sourceLen + (sourceLen >> 12) + (sourceLen >> 14) + 708 (sourceLen >> 25) + 13 - 6 + wraplen; 709 } 710 711 /* ========================================================================= 712 * Put a short in the pending buffer. The 16-bit value is put in MSB order. 713 * IN assertion: the stream state is correct and there is enough room in 714 * pending_buf. 715 */ 716 local void putShortMSB (s, b) 717 deflate_state *s; 718 uInt b; 719 { 720 put_byte(s, (Byte)(b >> 8)); 721 put_byte(s, (Byte)(b & 0xff)); 722 } 723 724 /* ========================================================================= 725 * Flush as much pending output as possible. All deflate() output, except for 726 * some deflate_stored() output, goes through this function so some 727 * applications may wish to modify it to avoid allocating a large 728 * strm->next_out buffer and copying into it. (See also read_buf()). 729 */ 730 local void flush_pending(strm) 731 z_streamp strm; 732 { 733 unsigned len; 734 deflate_state *s = strm->state; 735 736 _tr_flush_bits(s); 737 len = s->pending; 738 if (len > strm->avail_out) len = strm->avail_out; 739 if (len == 0) return; 740 741 zmemcpy(strm->next_out, s->pending_out, len); 742 strm->next_out += len; 743 s->pending_out += len; 744 strm->total_out += len; 745 strm->avail_out -= len; 746 s->pending -= len; 747 if (s->pending == 0) { 748 s->pending_out = s->pending_buf; 749 } 750 } 751 752 /* =========================================================================== 753 * Update the header CRC with the bytes s->pending_buf[beg..s->pending - 1]. 754 */ 755 #define HCRC_UPDATE(beg) \ 756 do { \ 757 if (s->gzhead->hcrc && s->pending > (beg)) \ 758 strm->adler = crc32(strm->adler, s->pending_buf + (beg), \ 759 s->pending - (beg)); \ 760 } while (0) 761 762 /* ========================================================================= */ 763 int ZEXPORT deflate (strm, flush) 764 z_streamp strm; 765 int flush; 766 { 767 int old_flush; /* value of flush param for previous deflate call */ 768 deflate_state *s; 769 770 if (deflateStateCheck(strm) || flush > Z_BLOCK || flush < 0) { 771 return Z_STREAM_ERROR; 772 } 773 s = strm->state; 774 775 if (strm->next_out == Z_NULL || 776 (strm->avail_in != 0 && strm->next_in == Z_NULL) || 777 (s->status == FINISH_STATE && flush != Z_FINISH)) { 778 ERR_RETURN(strm, Z_STREAM_ERROR); 779 } 780 if (strm->avail_out == 0) ERR_RETURN(strm, Z_BUF_ERROR); 781 782 old_flush = s->last_flush; 783 s->last_flush = flush; 784 785 /* Flush as much pending output as possible */ 786 if (s->pending != 0) { 787 flush_pending(strm); 788 if (strm->avail_out == 0) { 789 /* Since avail_out is 0, deflate will be called again with 790 * more output space, but possibly with both pending and 791 * avail_in equal to zero. There won't be anything to do, 792 * but this is not an error situation so make sure we 793 * return OK instead of BUF_ERROR at next call of deflate: 794 */ 795 s->last_flush = -1; 796 return Z_OK; 797 } 798 799 /* Make sure there is something to do and avoid duplicate consecutive 800 * flushes. For repeated and useless calls with Z_FINISH, we keep 801 * returning Z_STREAM_END instead of Z_BUF_ERROR. 802 */ 803 } else if (strm->avail_in == 0 && RANK(flush) <= RANK(old_flush) && 804 flush != Z_FINISH) { 805 ERR_RETURN(strm, Z_BUF_ERROR); 806 } 807 808 /* User must not provide more input after the first FINISH: */ 809 if (s->status == FINISH_STATE && strm->avail_in != 0) { 810 ERR_RETURN(strm, Z_BUF_ERROR); 811 } 812 813 /* Write the header */ 814 if (s->status == INIT_STATE) { 815 /* zlib header */ 816 uInt header = (Z_DEFLATED + ((s->w_bits-8)<<4)) << 8; 817 uInt level_flags; 818 819 if (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2) 820 level_flags = 0; 821 else if (s->level < 6) 822 level_flags = 1; 823 else if (s->level == 6) 824 level_flags = 2; 825 else 826 level_flags = 3; 827 header |= (level_flags << 6); 828 if (s->strstart != 0) header |= PRESET_DICT; 829 header += 31 - (header % 31); 830 831 putShortMSB(s, header); 832 833 /* Save the adler32 of the preset dictionary: */ 834 if (s->strstart != 0) { 835 putShortMSB(s, (uInt)(strm->adler >> 16)); 836 putShortMSB(s, (uInt)(strm->adler & 0xffff)); 837 } 838 strm->adler = adler32(0L, Z_NULL, 0); 839 s->status = BUSY_STATE; 840 841 /* Compression must start with an empty pending buffer */ 842 flush_pending(strm); 843 if (s->pending != 0) { 844 s->last_flush = -1; 845 return Z_OK; 846 } 847 } 848 #ifdef GZIP 849 if (s->status == GZIP_STATE) { 850 /* gzip header */ 851 strm->adler = crc32(0L, Z_NULL, 0); 852 put_byte(s, 31); 853 put_byte(s, 139); 854 put_byte(s, 8); 855 if (s->gzhead == Z_NULL) { 856 put_byte(s, 0); 857 put_byte(s, 0); 858 put_byte(s, 0); 859 put_byte(s, 0); 860 put_byte(s, 0); 861 put_byte(s, s->level == 9 ? 2 : 862 (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ? 863 4 : 0)); 864 put_byte(s, OS_CODE); 865 s->status = BUSY_STATE; 866 867 /* Compression must start with an empty pending buffer */ 868 flush_pending(strm); 869 if (s->pending != 0) { 870 s->last_flush = -1; 871 return Z_OK; 872 } 873 } 874 else { 875 put_byte(s, (s->gzhead->text ? 1 : 0) + 876 (s->gzhead->hcrc ? 2 : 0) + 877 (s->gzhead->extra == Z_NULL ? 0 : 4) + 878 (s->gzhead->name == Z_NULL ? 0 : 8) + 879 (s->gzhead->comment == Z_NULL ? 0 : 16) 880 ); 881 put_byte(s, (Byte)(s->gzhead->time & 0xff)); 882 put_byte(s, (Byte)((s->gzhead->time >> 8) & 0xff)); 883 put_byte(s, (Byte)((s->gzhead->time >> 16) & 0xff)); 884 put_byte(s, (Byte)((s->gzhead->time >> 24) & 0xff)); 885 put_byte(s, s->level == 9 ? 2 : 886 (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ? 887 4 : 0)); 888 put_byte(s, s->gzhead->os & 0xff); 889 if (s->gzhead->extra != Z_NULL) { 890 put_byte(s, s->gzhead->extra_len & 0xff); 891 put_byte(s, (s->gzhead->extra_len >> 8) & 0xff); 892 } 893 if (s->gzhead->hcrc) 894 strm->adler = crc32(strm->adler, s->pending_buf, 895 s->pending); 896 s->gzindex = 0; 897 s->status = EXTRA_STATE; 898 } 899 } 900 if (s->status == EXTRA_STATE) { 901 if (s->gzhead->extra != Z_NULL) { 902 ulg beg = s->pending; /* start of bytes to update crc */ 903 uInt left = (s->gzhead->extra_len & 0xffff) - s->gzindex; 904 while (s->pending + left > s->pending_buf_size) { 905 uInt copy = s->pending_buf_size - s->pending; 906 zmemcpy(s->pending_buf + s->pending, 907 s->gzhead->extra + s->gzindex, copy); 908 s->pending = s->pending_buf_size; 909 HCRC_UPDATE(beg); 910 s->gzindex += copy; 911 flush_pending(strm); 912 if (s->pending != 0) { 913 s->last_flush = -1; 914 return Z_OK; 915 } 916 beg = 0; 917 left -= copy; 918 } 919 zmemcpy(s->pending_buf + s->pending, 920 s->gzhead->extra + s->gzindex, left); 921 s->pending += left; 922 HCRC_UPDATE(beg); 923 s->gzindex = 0; 924 } 925 s->status = NAME_STATE; 926 } 927 if (s->status == NAME_STATE) { 928 if (s->gzhead->name != Z_NULL) { 929 ulg beg = s->pending; /* start of bytes to update crc */ 930 int val; 931 do { 932 if (s->pending == s->pending_buf_size) { 933 HCRC_UPDATE(beg); 934 flush_pending(strm); 935 if (s->pending != 0) { 936 s->last_flush = -1; 937 return Z_OK; 938 } 939 beg = 0; 940 } 941 val = s->gzhead->name[s->gzindex++]; 942 put_byte(s, val); 943 } while (val != 0); 944 HCRC_UPDATE(beg); 945 s->gzindex = 0; 946 } 947 s->status = COMMENT_STATE; 948 } 949 if (s->status == COMMENT_STATE) { 950 if (s->gzhead->comment != Z_NULL) { 951 ulg beg = s->pending; /* start of bytes to update crc */ 952 int val; 953 do { 954 if (s->pending == s->pending_buf_size) { 955 HCRC_UPDATE(beg); 956 flush_pending(strm); 957 if (s->pending != 0) { 958 s->last_flush = -1; 959 return Z_OK; 960 } 961 beg = 0; 962 } 963 val = s->gzhead->comment[s->gzindex++]; 964 put_byte(s, val); 965 } while (val != 0); 966 HCRC_UPDATE(beg); 967 } 968 s->status = HCRC_STATE; 969 } 970 if (s->status == HCRC_STATE) { 971 if (s->gzhead->hcrc) { 972 if (s->pending + 2 > s->pending_buf_size) { 973 flush_pending(strm); 974 if (s->pending != 0) { 975 s->last_flush = -1; 976 return Z_OK; 977 } 978 } 979 put_byte(s, (Byte)(strm->adler & 0xff)); 980 put_byte(s, (Byte)((strm->adler >> 8) & 0xff)); 981 strm->adler = crc32(0L, Z_NULL, 0); 982 } 983 s->status = BUSY_STATE; 984 985 /* Compression must start with an empty pending buffer */ 986 flush_pending(strm); 987 if (s->pending != 0) { 988 s->last_flush = -1; 989 return Z_OK; 990 } 991 } 992 #endif 993 994 /* Start a new block or continue the current one. 995 */ 996 if (strm->avail_in != 0 || s->lookahead != 0 || 997 (flush != Z_NO_FLUSH && s->status != FINISH_STATE)) { 998 block_state bstate; 999 1000 bstate = s->level == 0 ? deflate_stored(s, flush) : 1001 s->strategy == Z_HUFFMAN_ONLY ? deflate_huff(s, flush) : 1002 s->strategy == Z_RLE ? deflate_rle(s, flush) : 1003 (*(configuration_table[s->level].func))(s, flush); 1004 1005 if (bstate == finish_started || bstate == finish_done) { 1006 s->status = FINISH_STATE; 1007 } 1008 if (bstate == need_more || bstate == finish_started) { 1009 if (strm->avail_out == 0) { 1010 s->last_flush = -1; /* avoid BUF_ERROR next call, see above */ 1011 } 1012 return Z_OK; 1013 /* If flush != Z_NO_FLUSH && avail_out == 0, the next call 1014 * of deflate should use the same flush parameter to make sure 1015 * that the flush is complete. So we don't have to output an 1016 * empty block here, this will be done at next call. This also 1017 * ensures that for a very small output buffer, we emit at most 1018 * one empty block. 1019 */ 1020 } 1021 if (bstate == block_done) { 1022 if (flush == Z_PARTIAL_FLUSH) { 1023 _tr_align(s); 1024 } else if (flush != Z_BLOCK) { /* FULL_FLUSH or SYNC_FLUSH */ 1025 _tr_stored_block(s, (char*)0, 0L, 0); 1026 /* For a full flush, this empty block will be recognized 1027 * as a special marker by inflate_sync(). 1028 */ 1029 if (flush == Z_FULL_FLUSH) { 1030 CLEAR_HASH(s); /* forget history */ 1031 if (s->lookahead == 0) { 1032 s->strstart = 0; 1033 s->block_start = 0L; 1034 s->insert = 0; 1035 } 1036 } 1037 } 1038 flush_pending(strm); 1039 if (strm->avail_out == 0) { 1040 s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */ 1041 return Z_OK; 1042 } 1043 } 1044 } 1045 1046 if (flush != Z_FINISH) return Z_OK; 1047 if (s->wrap <= 0) return Z_STREAM_END; 1048 1049 /* Write the trailer */ 1050 #ifdef GZIP 1051 if (s->wrap == 2) { 1052 put_byte(s, (Byte)(strm->adler & 0xff)); 1053 put_byte(s, (Byte)((strm->adler >> 8) & 0xff)); 1054 put_byte(s, (Byte)((strm->adler >> 16) & 0xff)); 1055 put_byte(s, (Byte)((strm->adler >> 24) & 0xff)); 1056 put_byte(s, (Byte)(strm->total_in & 0xff)); 1057 put_byte(s, (Byte)((strm->total_in >> 8) & 0xff)); 1058 put_byte(s, (Byte)((strm->total_in >> 16) & 0xff)); 1059 put_byte(s, (Byte)((strm->total_in >> 24) & 0xff)); 1060 } 1061 else 1062 #endif 1063 { 1064 putShortMSB(s, (uInt)(strm->adler >> 16)); 1065 putShortMSB(s, (uInt)(strm->adler & 0xffff)); 1066 } 1067 flush_pending(strm); 1068 /* If avail_out is zero, the application will call deflate again 1069 * to flush the rest. 1070 */ 1071 if (s->wrap > 0) s->wrap = -s->wrap; /* write the trailer only once! */ 1072 return s->pending != 0 ? Z_OK : Z_STREAM_END; 1073 } 1074 1075 /* ========================================================================= */ 1076 int ZEXPORT deflateEnd (strm) 1077 z_streamp strm; 1078 { 1079 int status; 1080 1081 if (deflateStateCheck(strm)) return Z_STREAM_ERROR; 1082 1083 status = strm->state->status; 1084 1085 /* Deallocate in reverse order of allocations: */ 1086 TRY_FREE(strm, strm->state->pending_buf); 1087 TRY_FREE(strm, strm->state->head); 1088 TRY_FREE(strm, strm->state->prev); 1089 TRY_FREE(strm, strm->state->window); 1090 1091 ZFREE(strm, strm->state); 1092 strm->state = Z_NULL; 1093 1094 return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK; 1095 } 1096 1097 /* ========================================================================= 1098 * Copy the source state to the destination state. 1099 * To simplify the source, this is not supported for 16-bit MSDOS (which 1100 * doesn't have enough memory anyway to duplicate compression states). 1101 */ 1102 int ZEXPORT deflateCopy (dest, source) 1103 z_streamp dest; 1104 z_streamp source; 1105 { 1106 #ifdef MAXSEG_64K 1107 return Z_STREAM_ERROR; 1108 #else 1109 deflate_state *ds; 1110 deflate_state *ss; 1111 ushf *overlay; 1112 1113 1114 if (deflateStateCheck(source) || dest == Z_NULL) { 1115 return Z_STREAM_ERROR; 1116 } 1117 1118 ss = source->state; 1119 1120 zmemcpy((voidpf)dest, (voidpf)source, sizeof(z_stream)); 1121 1122 ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state)); 1123 if (ds == Z_NULL) return Z_MEM_ERROR; 1124 dest->state = (struct internal_state FAR *) ds; 1125 zmemcpy((voidpf)ds, (voidpf)ss, sizeof(deflate_state)); 1126 ds->strm = dest; 1127 1128 ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte)); 1129 ds->prev = (Posf *) ZALLOC(dest, ds->w_size, sizeof(Pos)); 1130 ds->head = (Posf *) ZALLOC(dest, ds->hash_size, sizeof(Pos)); 1131 overlay = (ushf *) ZALLOC(dest, ds->lit_bufsize, sizeof(ush)+2); 1132 ds->pending_buf = (uchf *) overlay; 1133 1134 if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL || 1135 ds->pending_buf == Z_NULL) { 1136 deflateEnd (dest); 1137 return Z_MEM_ERROR; 1138 } 1139 /* following zmemcpy do not work for 16-bit MSDOS */ 1140 zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte)); 1141 zmemcpy((voidpf)ds->prev, (voidpf)ss->prev, ds->w_size * sizeof(Pos)); 1142 zmemcpy((voidpf)ds->head, (voidpf)ss->head, ds->hash_size * sizeof(Pos)); 1143 zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size); 1144 1145 ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf); 1146 ds->d_buf = overlay + ds->lit_bufsize/sizeof(ush); 1147 ds->l_buf = ds->pending_buf + (1+sizeof(ush))*ds->lit_bufsize; 1148 1149 ds->l_desc.dyn_tree = ds->dyn_ltree; 1150 ds->d_desc.dyn_tree = ds->dyn_dtree; 1151 ds->bl_desc.dyn_tree = ds->bl_tree; 1152 1153 return Z_OK; 1154 #endif /* MAXSEG_64K */ 1155 } 1156 1157 /* =========================================================================== 1158 * Read a new buffer from the current input stream, update the adler32 1159 * and total number of bytes read. All deflate() input goes through 1160 * this function so some applications may wish to modify it to avoid 1161 * allocating a large strm->next_in buffer and copying from it. 1162 * (See also flush_pending()). 1163 */ 1164 local unsigned read_buf(strm, buf, size) 1165 z_streamp strm; 1166 Bytef *buf; 1167 unsigned size; 1168 { 1169 unsigned len = strm->avail_in; 1170 1171 if (len > size) len = size; 1172 if (len == 0) return 0; 1173 1174 strm->avail_in -= len; 1175 1176 zmemcpy(buf, strm->next_in, len); 1177 if (strm->state->wrap == 1) { 1178 strm->adler = adler32(strm->adler, buf, len); 1179 } 1180 #ifdef GZIP 1181 else if (strm->state->wrap == 2) { 1182 strm->adler = crc32(strm->adler, buf, len); 1183 } 1184 #endif 1185 strm->next_in += len; 1186 strm->total_in += len; 1187 1188 return len; 1189 } 1190 1191 /* =========================================================================== 1192 * Initialize the "longest match" routines for a new zlib stream 1193 */ 1194 local void lm_init (s) 1195 deflate_state *s; 1196 { 1197 s->window_size = (ulg)2L*s->w_size; 1198 1199 CLEAR_HASH(s); 1200 1201 /* Set the default configuration parameters: 1202 */ 1203 s->max_lazy_match = configuration_table[s->level].max_lazy; 1204 s->good_match = configuration_table[s->level].good_length; 1205 s->nice_match = configuration_table[s->level].nice_length; 1206 s->max_chain_length = configuration_table[s->level].max_chain; 1207 1208 s->strstart = 0; 1209 s->block_start = 0L; 1210 s->lookahead = 0; 1211 s->insert = 0; 1212 s->match_length = s->prev_length = MIN_MATCH-1; 1213 s->match_available = 0; 1214 s->ins_h = 0; 1215 #ifndef FASTEST 1216 #ifdef ASMV 1217 match_init(); /* initialize the asm code */ 1218 #endif 1219 #endif 1220 } 1221 1222 #ifndef FASTEST 1223 /* =========================================================================== 1224 * Set match_start to the longest match starting at the given string and 1225 * return its length. Matches shorter or equal to prev_length are discarded, 1226 * in which case the result is equal to prev_length and match_start is 1227 * garbage. 1228 * IN assertions: cur_match is the head of the hash chain for the current 1229 * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1 1230 * OUT assertion: the match length is not greater than s->lookahead. 1231 */ 1232 #ifndef ASMV 1233 /* For 80x86 and 680x0, an optimized version will be provided in match.asm or 1234 * match.S. The code will be functionally equivalent. 1235 */ 1236 local uInt longest_match(s, cur_match) 1237 deflate_state *s; 1238 IPos cur_match; /* current match */ 1239 { 1240 unsigned chain_length = s->max_chain_length;/* max hash chain length */ 1241 register Bytef *scan = s->window + s->strstart; /* current string */ 1242 register Bytef *match; /* matched string */ 1243 register int len; /* length of current match */ 1244 int best_len = (int)s->prev_length; /* best match length so far */ 1245 int nice_match = s->nice_match; /* stop if match long enough */ 1246 IPos limit = s->strstart > (IPos)MAX_DIST(s) ? 1247 s->strstart - (IPos)MAX_DIST(s) : NIL; 1248 /* Stop when cur_match becomes <= limit. To simplify the code, 1249 * we prevent matches with the string of window index 0. 1250 */ 1251 Posf *prev = s->prev; 1252 uInt wmask = s->w_mask; 1253 1254 #ifdef UNALIGNED_OK 1255 /* Compare two bytes at a time. Note: this is not always beneficial. 1256 * Try with and without -DUNALIGNED_OK to check. 1257 */ 1258 register Bytef *strend = s->window + s->strstart + MAX_MATCH - 1; 1259 register ush scan_start = *(ushf*)scan; 1260 register ush scan_end = *(ushf*)(scan+best_len-1); 1261 #else 1262 register Bytef *strend = s->window + s->strstart + MAX_MATCH; 1263 register Byte scan_end1 = scan[best_len-1]; 1264 register Byte scan_end = scan[best_len]; 1265 #endif 1266 1267 /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16. 1268 * It is easy to get rid of this optimization if necessary. 1269 */ 1270 Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever"); 1271 1272 /* Do not waste too much time if we already have a good match: */ 1273 if (s->prev_length >= s->good_match) { 1274 chain_length >>= 2; 1275 } 1276 /* Do not look for matches beyond the end of the input. This is necessary 1277 * to make deflate deterministic. 1278 */ 1279 if ((uInt)nice_match > s->lookahead) nice_match = (int)s->lookahead; 1280 1281 Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead"); 1282 1283 do { 1284 Assert(cur_match < s->strstart, "no future"); 1285 match = s->window + cur_match; 1286 1287 /* Skip to next match if the match length cannot increase 1288 * or if the match length is less than 2. Note that the checks below 1289 * for insufficient lookahead only occur occasionally for performance 1290 * reasons. Therefore uninitialized memory will be accessed, and 1291 * conditional jumps will be made that depend on those values. 1292 * However the length of the match is limited to the lookahead, so 1293 * the output of deflate is not affected by the uninitialized values. 1294 */ 1295 #if (defined(UNALIGNED_OK) && MAX_MATCH == 258) 1296 /* This code assumes sizeof(unsigned short) == 2. Do not use 1297 * UNALIGNED_OK if your compiler uses a different size. 1298 */ 1299 if (*(ushf*)(match+best_len-1) != scan_end || 1300 *(ushf*)match != scan_start) continue; 1301 1302 /* It is not necessary to compare scan[2] and match[2] since they are 1303 * always equal when the other bytes match, given that the hash keys 1304 * are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at 1305 * strstart+3, +5, ... up to strstart+257. We check for insufficient 1306 * lookahead only every 4th comparison; the 128th check will be made 1307 * at strstart+257. If MAX_MATCH-2 is not a multiple of 8, it is 1308 * necessary to put more guard bytes at the end of the window, or 1309 * to check more often for insufficient lookahead. 1310 */ 1311 Assert(scan[2] == match[2], "scan[2]?"); 1312 scan++, match++; 1313 do { 1314 } while (*(ushf*)(scan+=2) == *(ushf*)(match+=2) && 1315 *(ushf*)(scan+=2) == *(ushf*)(match+=2) && 1316 *(ushf*)(scan+=2) == *(ushf*)(match+=2) && 1317 *(ushf*)(scan+=2) == *(ushf*)(match+=2) && 1318 scan < strend); 1319 /* The funny "do {}" generates better code on most compilers */ 1320 1321 /* Here, scan <= window+strstart+257 */ 1322 Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan"); 1323 if (*scan == *match) scan++; 1324 1325 len = (MAX_MATCH - 1) - (int)(strend-scan); 1326 scan = strend - (MAX_MATCH-1); 1327 1328 #else /* UNALIGNED_OK */ 1329 1330 if (match[best_len] != scan_end || 1331 match[best_len-1] != scan_end1 || 1332 *match != *scan || 1333 *++match != scan[1]) continue; 1334 1335 /* The check at best_len-1 can be removed because it will be made 1336 * again later. (This heuristic is not always a win.) 1337 * It is not necessary to compare scan[2] and match[2] since they 1338 * are always equal when the other bytes match, given that 1339 * the hash keys are equal and that HASH_BITS >= 8. 1340 */ 1341 scan += 2, match++; 1342 Assert(*scan == *match, "match[2]?"); 1343 1344 /* We check for insufficient lookahead only every 8th comparison; 1345 * the 256th check will be made at strstart+258. 1346 */ 1347 do { 1348 } while (*++scan == *++match && *++scan == *++match && 1349 *++scan == *++match && *++scan == *++match && 1350 *++scan == *++match && *++scan == *++match && 1351 *++scan == *++match && *++scan == *++match && 1352 scan < strend); 1353 1354 Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan"); 1355 1356 len = MAX_MATCH - (int)(strend - scan); 1357 scan = strend - MAX_MATCH; 1358 1359 #endif /* UNALIGNED_OK */ 1360 1361 if (len > best_len) { 1362 s->match_start = cur_match; 1363 best_len = len; 1364 if (len >= nice_match) break; 1365 #ifdef UNALIGNED_OK 1366 scan_end = *(ushf*)(scan+best_len-1); 1367 #else 1368 scan_end1 = scan[best_len-1]; 1369 scan_end = scan[best_len]; 1370 #endif 1371 } 1372 } while ((cur_match = prev[cur_match & wmask]) > limit 1373 && --chain_length != 0); 1374 1375 if ((uInt)best_len <= s->lookahead) return (uInt)best_len; 1376 return s->lookahead; 1377 } 1378 #endif /* ASMV */ 1379 1380 #else /* FASTEST */ 1381 1382 /* --------------------------------------------------------------------------- 1383 * Optimized version for FASTEST only 1384 */ 1385 local uInt longest_match(s, cur_match) 1386 deflate_state *s; 1387 IPos cur_match; /* current match */ 1388 { 1389 register Bytef *scan = s->window + s->strstart; /* current string */ 1390 register Bytef *match; /* matched string */ 1391 register int len; /* length of current match */ 1392 register Bytef *strend = s->window + s->strstart + MAX_MATCH; 1393 1394 /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16. 1395 * It is easy to get rid of this optimization if necessary. 1396 */ 1397 Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever"); 1398 1399 Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead"); 1400 1401 Assert(cur_match < s->strstart, "no future"); 1402 1403 match = s->window + cur_match; 1404 1405 /* Return failure if the match length is less than 2: 1406 */ 1407 if (match[0] != scan[0] || match[1] != scan[1]) return MIN_MATCH-1; 1408 1409 /* The check at best_len-1 can be removed because it will be made 1410 * again later. (This heuristic is not always a win.) 1411 * It is not necessary to compare scan[2] and match[2] since they 1412 * are always equal when the other bytes match, given that 1413 * the hash keys are equal and that HASH_BITS >= 8. 1414 */ 1415 scan += 2, match += 2; 1416 Assert(*scan == *match, "match[2]?"); 1417 1418 /* We check for insufficient lookahead only every 8th comparison; 1419 * the 256th check will be made at strstart+258. 1420 */ 1421 do { 1422 } while (*++scan == *++match && *++scan == *++match && 1423 *++scan == *++match && *++scan == *++match && 1424 *++scan == *++match && *++scan == *++match && 1425 *++scan == *++match && *++scan == *++match && 1426 scan < strend); 1427 1428 Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan"); 1429 1430 len = MAX_MATCH - (int)(strend - scan); 1431 1432 if (len < MIN_MATCH) return MIN_MATCH - 1; 1433 1434 s->match_start = cur_match; 1435 return (uInt)len <= s->lookahead ? (uInt)len : s->lookahead; 1436 } 1437 1438 #endif /* FASTEST */ 1439 1440 #ifdef ZLIB_DEBUG 1441 1442 #define EQUAL 0 1443 /* result of memcmp for equal strings */ 1444 1445 /* =========================================================================== 1446 * Check that the match at match_start is indeed a match. 1447 */ 1448 local void check_match(s, start, match, length) 1449 deflate_state *s; 1450 IPos start, match; 1451 int length; 1452 { 1453 /* check that the match is indeed a match */ 1454 if (zmemcmp(s->window + match, 1455 s->window + start, length) != EQUAL) { 1456 fprintf(stderr, " start %u, match %u, length %d\n", 1457 start, match, length); 1458 do { 1459 fprintf(stderr, "%c%c", s->window[match++], s->window[start++]); 1460 } while (--length != 0); 1461 z_error("invalid match"); 1462 } 1463 if (z_verbose > 1) { 1464 fprintf(stderr,"\\[%d,%d]", start-match, length); 1465 do { putc(s->window[start++], stderr); } while (--length != 0); 1466 } 1467 } 1468 #else 1469 # define check_match(s, start, match, length) 1470 #endif /* ZLIB_DEBUG */ 1471 1472 /* =========================================================================== 1473 * Fill the window when the lookahead becomes insufficient. 1474 * Updates strstart and lookahead. 1475 * 1476 * IN assertion: lookahead < MIN_LOOKAHEAD 1477 * OUT assertions: strstart <= window_size-MIN_LOOKAHEAD 1478 * At least one byte has been read, or avail_in == 0; reads are 1479 * performed for at least two bytes (required for the zip translate_eol 1480 * option -- not supported here). 1481 */ 1482 local void fill_window(s) 1483 deflate_state *s; 1484 { 1485 unsigned n; 1486 unsigned more; /* Amount of free space at the end of the window. */ 1487 uInt wsize = s->w_size; 1488 1489 Assert(s->lookahead < MIN_LOOKAHEAD, "already enough lookahead"); 1490 1491 do { 1492 more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart); 1493 1494 /* Deal with !@#$% 64K limit: */ 1495 if (sizeof(int) <= 2) { 1496 if (more == 0 && s->strstart == 0 && s->lookahead == 0) { 1497 more = wsize; 1498 1499 } else if (more == (unsigned)(-1)) { 1500 /* Very unlikely, but possible on 16 bit machine if 1501 * strstart == 0 && lookahead == 1 (input done a byte at time) 1502 */ 1503 more--; 1504 } 1505 } 1506 1507 /* If the window is almost full and there is insufficient lookahead, 1508 * move the upper half to the lower one to make room in the upper half. 1509 */ 1510 if (s->strstart >= wsize+MAX_DIST(s)) { 1511 1512 zmemcpy(s->window, s->window+wsize, (unsigned)wsize - more); 1513 s->match_start -= wsize; 1514 s->strstart -= wsize; /* we now have strstart >= MAX_DIST */ 1515 s->block_start -= (long) wsize; 1516 slide_hash(s); 1517 more += wsize; 1518 } 1519 if (s->strm->avail_in == 0) break; 1520 1521 /* If there was no sliding: 1522 * strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 && 1523 * more == window_size - lookahead - strstart 1524 * => more >= window_size - (MIN_LOOKAHEAD-1 + WSIZE + MAX_DIST-1) 1525 * => more >= window_size - 2*WSIZE + 2 1526 * In the BIG_MEM or MMAP case (not yet supported), 1527 * window_size == input_size + MIN_LOOKAHEAD && 1528 * strstart + s->lookahead <= input_size => more >= MIN_LOOKAHEAD. 1529 * Otherwise, window_size == 2*WSIZE so more >= 2. 1530 * If there was sliding, more >= WSIZE. So in all cases, more >= 2. 1531 */ 1532 Assert(more >= 2, "more < 2"); 1533 1534 n = read_buf(s->strm, s->window + s->strstart + s->lookahead, more); 1535 s->lookahead += n; 1536 1537 /* Initialize the hash value now that we have some input: */ 1538 if (s->lookahead + s->insert >= MIN_MATCH) { 1539 uInt str = s->strstart - s->insert; 1540 s->ins_h = s->window[str]; 1541 UPDATE_HASH(s, s->ins_h, s->window[str + 1]); 1542 #if MIN_MATCH != 3 1543 Call UPDATE_HASH() MIN_MATCH-3 more times 1544 #endif 1545 while (s->insert) { 1546 UPDATE_HASH(s, s->ins_h, s->window[str + MIN_MATCH-1]); 1547 #ifndef FASTEST 1548 s->prev[str & s->w_mask] = s->head[s->ins_h]; 1549 #endif 1550 s->head[s->ins_h] = (Pos)str; 1551 str++; 1552 s->insert--; 1553 if (s->lookahead + s->insert < MIN_MATCH) 1554 break; 1555 } 1556 } 1557 /* If the whole input has less than MIN_MATCH bytes, ins_h is garbage, 1558 * but this is not important since only literal bytes will be emitted. 1559 */ 1560 1561 } while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0); 1562 1563 /* If the WIN_INIT bytes after the end of the current data have never been 1564 * written, then zero those bytes in order to avoid memory check reports of 1565 * the use of uninitialized (or uninitialised as Julian writes) bytes by 1566 * the longest match routines. Update the high water mark for the next 1567 * time through here. WIN_INIT is set to MAX_MATCH since the longest match 1568 * routines allow scanning to strstart + MAX_MATCH, ignoring lookahead. 1569 */ 1570 if (s->high_water < s->window_size) { 1571 ulg curr = s->strstart + (ulg)(s->lookahead); 1572 ulg init; 1573 1574 if (s->high_water < curr) { 1575 /* Previous high water mark below current data -- zero WIN_INIT 1576 * bytes or up to end of window, whichever is less. 1577 */ 1578 init = s->window_size - curr; 1579 if (init > WIN_INIT) 1580 init = WIN_INIT; 1581 zmemzero(s->window + curr, (unsigned)init); 1582 s->high_water = curr + init; 1583 } 1584 else if (s->high_water < (ulg)curr + WIN_INIT) { 1585 /* High water mark at or above current data, but below current data 1586 * plus WIN_INIT -- zero out to current data plus WIN_INIT, or up 1587 * to end of window, whichever is less. 1588 */ 1589 init = (ulg)curr + WIN_INIT - s->high_water; 1590 if (init > s->window_size - s->high_water) 1591 init = s->window_size - s->high_water; 1592 zmemzero(s->window + s->high_water, (unsigned)init); 1593 s->high_water += init; 1594 } 1595 } 1596 1597 Assert((ulg)s->strstart <= s->window_size - MIN_LOOKAHEAD, 1598 "not enough room for search"); 1599 } 1600 1601 /* =========================================================================== 1602 * Flush the current block, with given end-of-file flag. 1603 * IN assertion: strstart is set to the end of the current match. 1604 */ 1605 #define FLUSH_BLOCK_ONLY(s, last) { \ 1606 _tr_flush_block(s, (s->block_start >= 0L ? \ 1607 (charf *)&s->window[(unsigned)s->block_start] : \ 1608 (charf *)Z_NULL), \ 1609 (ulg)((long)s->strstart - s->block_start), \ 1610 (last)); \ 1611 s->block_start = s->strstart; \ 1612 flush_pending(s->strm); \ 1613 Tracev((stderr,"[FLUSH]")); \ 1614 } 1615 1616 /* Same but force premature exit if necessary. */ 1617 #define FLUSH_BLOCK(s, last) { \ 1618 FLUSH_BLOCK_ONLY(s, last); \ 1619 if (s->strm->avail_out == 0) return (last) ? finish_started : need_more; \ 1620 } 1621 1622 /* Maximum stored block length in deflate format (not including header). */ 1623 #define MAX_STORED 65535 1624 1625 /* Minimum of a and b. */ 1626 #define MIN(a, b) ((a) > (b) ? (b) : (a)) 1627 1628 /* =========================================================================== 1629 * Copy without compression as much as possible from the input stream, return 1630 * the current block state. 1631 * 1632 * In case deflateParams() is used to later switch to a non-zero compression 1633 * level, s->matches (otherwise unused when storing) keeps track of the number 1634 * of hash table slides to perform. If s->matches is 1, then one hash table 1635 * slide will be done when switching. If s->matches is 2, the maximum value 1636 * allowed here, then the hash table will be cleared, since two or more slides 1637 * is the same as a clear. 1638 * 1639 * deflate_stored() is written to minimize the number of times an input byte is 1640 * copied. It is most efficient with large input and output buffers, which 1641 * maximizes the opportunites to have a single copy from next_in to next_out. 1642 */ 1643 local block_state deflate_stored(s, flush) 1644 deflate_state *s; 1645 int flush; 1646 { 1647 /* Smallest worthy block size when not flushing or finishing. By default 1648 * this is 32K. This can be as small as 507 bytes for memLevel == 1. For 1649 * large input and output buffers, the stored block size will be larger. 1650 */ 1651 unsigned min_block = MIN(s->pending_buf_size - 5, s->w_size); 1652 1653 /* Copy as many min_block or larger stored blocks directly to next_out as 1654 * possible. If flushing, copy the remaining available input to next_out as 1655 * stored blocks, if there is enough space. 1656 */ 1657 unsigned len, left, have, last = 0; 1658 unsigned used = s->strm->avail_in; 1659 do { 1660 /* Set len to the maximum size block that we can copy directly with the 1661 * available input data and output space. Set left to how much of that 1662 * would be copied from what's left in the window. 1663 */ 1664 len = MAX_STORED; /* maximum deflate stored block length */ 1665 have = (s->bi_valid + 42) >> 3; /* number of header bytes */ 1666 if (s->strm->avail_out < have) /* need room for header */ 1667 break; 1668 /* maximum stored block length that will fit in avail_out: */ 1669 have = s->strm->avail_out - have; 1670 left = s->strstart - s->block_start; /* bytes left in window */ 1671 if (len > (ulg)left + s->strm->avail_in) 1672 len = left + s->strm->avail_in; /* limit len to the input */ 1673 if (len > have) 1674 len = have; /* limit len to the output */ 1675 1676 /* If the stored block would be less than min_block in length, or if 1677 * unable to copy all of the available input when flushing, then try 1678 * copying to the window and the pending buffer instead. Also don't 1679 * write an empty block when flushing -- deflate() does that. 1680 */ 1681 if (len < min_block && ((len == 0 && flush != Z_FINISH) || 1682 flush == Z_NO_FLUSH || 1683 len != left + s->strm->avail_in)) 1684 break; 1685 1686 /* Make a dummy stored block in pending to get the header bytes, 1687 * including any pending bits. This also updates the debugging counts. 1688 */ 1689 last = flush == Z_FINISH && len == left + s->strm->avail_in ? 1 : 0; 1690 _tr_stored_block(s, (char *)0, 0L, last); 1691 1692 /* Replace the lengths in the dummy stored block with len. */ 1693 s->pending_buf[s->pending - 4] = len; 1694 s->pending_buf[s->pending - 3] = len >> 8; 1695 s->pending_buf[s->pending - 2] = ~len; 1696 s->pending_buf[s->pending - 1] = ~len >> 8; 1697 1698 /* Write the stored block header bytes. */ 1699 flush_pending(s->strm); 1700 1701 #ifdef ZLIB_DEBUG 1702 /* Update debugging counts for the data about to be copied. */ 1703 s->compressed_len += len << 3; 1704 s->bits_sent += len << 3; 1705 #endif 1706 1707 /* Copy uncompressed bytes from the window to next_out. */ 1708 if (left) { 1709 if (left > len) 1710 left = len; 1711 zmemcpy(s->strm->next_out, s->window + s->block_start, left); 1712 s->strm->next_out += left; 1713 s->strm->avail_out -= left; 1714 s->strm->total_out += left; 1715 s->block_start += left; 1716 len -= left; 1717 } 1718 1719 /* Copy uncompressed bytes directly from next_in to next_out, updating 1720 * the check value. 1721 */ 1722 if (len) { 1723 read_buf(s->strm, s->strm->next_out, len); 1724 s->strm->next_out += len; 1725 s->strm->avail_out -= len; 1726 s->strm->total_out += len; 1727 } 1728 } while (last == 0); 1729 1730 /* Update the sliding window with the last s->w_size bytes of the copied 1731 * data, or append all of the copied data to the existing window if less 1732 * than s->w_size bytes were copied. Also update the number of bytes to 1733 * insert in the hash tables, in the event that deflateParams() switches to 1734 * a non-zero compression level. 1735 */ 1736 used -= s->strm->avail_in; /* number of input bytes directly copied */ 1737 if (used) { 1738 /* If any input was used, then no unused input remains in the window, 1739 * therefore s->block_start == s->strstart. 1740 */ 1741 if (used >= s->w_size) { /* supplant the previous history */ 1742 s->matches = 2; /* clear hash */ 1743 zmemcpy(s->window, s->strm->next_in - s->w_size, s->w_size); 1744 s->strstart = s->w_size; 1745 } 1746 else { 1747 if (s->window_size - s->strstart <= used) { 1748 /* Slide the window down. */ 1749 s->strstart -= s->w_size; 1750 zmemcpy(s->window, s->window + s->w_size, s->strstart); 1751 if (s->matches < 2) 1752 s->matches++; /* add a pending slide_hash() */ 1753 } 1754 zmemcpy(s->window + s->strstart, s->strm->next_in - used, used); 1755 s->strstart += used; 1756 } 1757 s->block_start = s->strstart; 1758 s->insert += MIN(used, s->w_size - s->insert); 1759 } 1760 if (s->high_water < s->strstart) 1761 s->high_water = s->strstart; 1762 1763 /* If the last block was written to next_out, then done. */ 1764 if (last) 1765 return finish_done; 1766 1767 /* If flushing and all input has been consumed, then done. */ 1768 if (flush != Z_NO_FLUSH && flush != Z_FINISH && 1769 s->strm->avail_in == 0 && (long)s->strstart == s->block_start) 1770 return block_done; 1771 1772 /* Fill the window with any remaining input. */ 1773 have = s->window_size - s->strstart - 1; 1774 if (s->strm->avail_in > have && s->block_start >= (long)s->w_size) { 1775 /* Slide the window down. */ 1776 s->block_start -= s->w_size; 1777 s->strstart -= s->w_size; 1778 zmemcpy(s->window, s->window + s->w_size, s->strstart); 1779 if (s->matches < 2) 1780 s->matches++; /* add a pending slide_hash() */ 1781 have += s->w_size; /* more space now */ 1782 } 1783 if (have > s->strm->avail_in) 1784 have = s->strm->avail_in; 1785 if (have) { 1786 read_buf(s->strm, s->window + s->strstart, have); 1787 s->strstart += have; 1788 } 1789 if (s->high_water < s->strstart) 1790 s->high_water = s->strstart; 1791 1792 /* There was not enough avail_out to write a complete worthy or flushed 1793 * stored block to next_out. Write a stored block to pending instead, if we 1794 * have enough input for a worthy block, or if flushing and there is enough 1795 * room for the remaining input as a stored block in the pending buffer. 1796 */ 1797 have = (s->bi_valid + 42) >> 3; /* number of header bytes */ 1798 /* maximum stored block length that will fit in pending: */ 1799 have = MIN(s->pending_buf_size - have, MAX_STORED); 1800 min_block = MIN(have, s->w_size); 1801 left = s->strstart - s->block_start; 1802 if (left >= min_block || 1803 ((left || flush == Z_FINISH) && flush != Z_NO_FLUSH && 1804 s->strm->avail_in == 0 && left <= have)) { 1805 len = MIN(left, have); 1806 last = flush == Z_FINISH && s->strm->avail_in == 0 && 1807 len == left ? 1 : 0; 1808 _tr_stored_block(s, (charf *)s->window + s->block_start, len, last); 1809 s->block_start += len; 1810 flush_pending(s->strm); 1811 } 1812 1813 /* We've done all we can with the available input and output. */ 1814 return last ? finish_started : need_more; 1815 } 1816 1817 /* =========================================================================== 1818 * Compress as much as possible from the input stream, return the current 1819 * block state. 1820 * This function does not perform lazy evaluation of matches and inserts 1821 * new strings in the dictionary only for unmatched strings or for short 1822 * matches. It is used only for the fast compression options. 1823 */ 1824 local block_state deflate_fast(s, flush) 1825 deflate_state *s; 1826 int flush; 1827 { 1828 IPos hash_head; /* head of the hash chain */ 1829 int bflush; /* set if current block must be flushed */ 1830 1831 for (;;) { 1832 /* Make sure that we always have enough lookahead, except 1833 * at the end of the input file. We need MAX_MATCH bytes 1834 * for the next match, plus MIN_MATCH bytes to insert the 1835 * string following the next match. 1836 */ 1837 if (s->lookahead < MIN_LOOKAHEAD) { 1838 fill_window(s); 1839 if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) { 1840 return need_more; 1841 } 1842 if (s->lookahead == 0) break; /* flush the current block */ 1843 } 1844 1845 /* Insert the string window[strstart .. strstart+2] in the 1846 * dictionary, and set hash_head to the head of the hash chain: 1847 */ 1848 hash_head = NIL; 1849 if (s->lookahead >= MIN_MATCH) { 1850 INSERT_STRING(s, s->strstart, hash_head); 1851 } 1852 1853 /* Find the longest match, discarding those <= prev_length. 1854 * At this point we have always match_length < MIN_MATCH 1855 */ 1856 if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) { 1857 /* To simplify the code, we prevent matches with the string 1858 * of window index 0 (in particular we have to avoid a match 1859 * of the string with itself at the start of the input file). 1860 */ 1861 s->match_length = longest_match (s, hash_head); 1862 /* longest_match() sets match_start */ 1863 } 1864 if (s->match_length >= MIN_MATCH) { 1865 check_match(s, s->strstart, s->match_start, s->match_length); 1866 1867 _tr_tally_dist(s, s->strstart - s->match_start, 1868 s->match_length - MIN_MATCH, bflush); 1869 1870 s->lookahead -= s->match_length; 1871 1872 /* Insert new strings in the hash table only if the match length 1873 * is not too large. This saves time but degrades compression. 1874 */ 1875 #ifndef FASTEST 1876 if (s->match_length <= s->max_insert_length && 1877 s->lookahead >= MIN_MATCH) { 1878 s->match_length--; /* string at strstart already in table */ 1879 do { 1880 s->strstart++; 1881 INSERT_STRING(s, s->strstart, hash_head); 1882 /* strstart never exceeds WSIZE-MAX_MATCH, so there are 1883 * always MIN_MATCH bytes ahead. 1884 */ 1885 } while (--s->match_length != 0); 1886 s->strstart++; 1887 } else 1888 #endif 1889 { 1890 s->strstart += s->match_length; 1891 s->match_length = 0; 1892 s->ins_h = s->window[s->strstart]; 1893 UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]); 1894 #if MIN_MATCH != 3 1895 Call UPDATE_HASH() MIN_MATCH-3 more times 1896 #endif 1897 /* If lookahead < MIN_MATCH, ins_h is garbage, but it does not 1898 * matter since it will be recomputed at next deflate call. 1899 */ 1900 } 1901 } else { 1902 /* No match, output a literal byte */ 1903 Tracevv((stderr,"%c", s->window[s->strstart])); 1904 _tr_tally_lit (s, s->window[s->strstart], bflush); 1905 s->lookahead--; 1906 s->strstart++; 1907 } 1908 if (bflush) FLUSH_BLOCK(s, 0); 1909 } 1910 s->insert = s->strstart < MIN_MATCH-1 ? s->strstart : MIN_MATCH-1; 1911 if (flush == Z_FINISH) { 1912 FLUSH_BLOCK(s, 1); 1913 return finish_done; 1914 } 1915 if (s->last_lit) 1916 FLUSH_BLOCK(s, 0); 1917 return block_done; 1918 } 1919 1920 #ifndef FASTEST 1921 /* =========================================================================== 1922 * Same as above, but achieves better compression. We use a lazy 1923 * evaluation for matches: a match is finally adopted only if there is 1924 * no better match at the next window position. 1925 */ 1926 local block_state deflate_slow(s, flush) 1927 deflate_state *s; 1928 int flush; 1929 { 1930 IPos hash_head; /* head of hash chain */ 1931 int bflush; /* set if current block must be flushed */ 1932 1933 /* Process the input block. */ 1934 for (;;) { 1935 /* Make sure that we always have enough lookahead, except 1936 * at the end of the input file. We need MAX_MATCH bytes 1937 * for the next match, plus MIN_MATCH bytes to insert the 1938 * string following the next match. 1939 */ 1940 if (s->lookahead < MIN_LOOKAHEAD) { 1941 fill_window(s); 1942 if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) { 1943 return need_more; 1944 } 1945 if (s->lookahead == 0) break; /* flush the current block */ 1946 } 1947 1948 /* Insert the string window[strstart .. strstart+2] in the 1949 * dictionary, and set hash_head to the head of the hash chain: 1950 */ 1951 hash_head = NIL; 1952 if (s->lookahead >= MIN_MATCH) { 1953 INSERT_STRING(s, s->strstart, hash_head); 1954 } 1955 1956 /* Find the longest match, discarding those <= prev_length. 1957 */ 1958 s->prev_length = s->match_length, s->prev_match = s->match_start; 1959 s->match_length = MIN_MATCH-1; 1960 1961 if (hash_head != NIL && s->prev_length < s->max_lazy_match && 1962 s->strstart - hash_head <= MAX_DIST(s)) { 1963 /* To simplify the code, we prevent matches with the string 1964 * of window index 0 (in particular we have to avoid a match 1965 * of the string with itself at the start of the input file). 1966 */ 1967 s->match_length = longest_match (s, hash_head); 1968 /* longest_match() sets match_start */ 1969 1970 if (s->match_length <= 5 && (s->strategy == Z_FILTERED 1971 #if TOO_FAR <= 32767 1972 || (s->match_length == MIN_MATCH && 1973 s->strstart - s->match_start > TOO_FAR) 1974 #endif 1975 )) { 1976 1977 /* If prev_match is also MIN_MATCH, match_start is garbage 1978 * but we will ignore the current match anyway. 1979 */ 1980 s->match_length = MIN_MATCH-1; 1981 } 1982 } 1983 /* If there was a match at the previous step and the current 1984 * match is not better, output the previous match: 1985 */ 1986 if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) { 1987 uInt max_insert = s->strstart + s->lookahead - MIN_MATCH; 1988 /* Do not insert strings in hash table beyond this. */ 1989 1990 check_match(s, s->strstart-1, s->prev_match, s->prev_length); 1991 1992 _tr_tally_dist(s, s->strstart -1 - s->prev_match, 1993 s->prev_length - MIN_MATCH, bflush); 1994 1995 /* Insert in hash table all strings up to the end of the match. 1996 * strstart-1 and strstart are already inserted. If there is not 1997 * enough lookahead, the last two strings are not inserted in 1998 * the hash table. 1999 */ 2000 s->lookahead -= s->prev_length-1; 2001 s->prev_length -= 2; 2002 do { 2003 if (++s->strstart <= max_insert) { 2004 INSERT_STRING(s, s->strstart, hash_head); 2005 } 2006 } while (--s->prev_length != 0); 2007 s->match_available = 0; 2008 s->match_length = MIN_MATCH-1; 2009 s->strstart++; 2010 2011 if (bflush) FLUSH_BLOCK(s, 0); 2012 2013 } else if (s->match_available) { 2014 /* If there was no match at the previous position, output a 2015 * single literal. If there was a match but the current match 2016 * is longer, truncate the previous match to a single literal. 2017 */ 2018 Tracevv((stderr,"%c", s->window[s->strstart-1])); 2019 _tr_tally_lit(s, s->window[s->strstart-1], bflush); 2020 if (bflush) { 2021 FLUSH_BLOCK_ONLY(s, 0); 2022 } 2023 s->strstart++; 2024 s->lookahead--; 2025 if (s->strm->avail_out == 0) return need_more; 2026 } else { 2027 /* There is no previous match to compare with, wait for 2028 * the next step to decide. 2029 */ 2030 s->match_available = 1; 2031 s->strstart++; 2032 s->lookahead--; 2033 } 2034 } 2035 Assert (flush != Z_NO_FLUSH, "no flush?"); 2036 if (s->match_available) { 2037 Tracevv((stderr,"%c", s->window[s->strstart-1])); 2038 _tr_tally_lit(s, s->window[s->strstart-1], bflush); 2039 s->match_available = 0; 2040 } 2041 s->insert = s->strstart < MIN_MATCH-1 ? s->strstart : MIN_MATCH-1; 2042 if (flush == Z_FINISH) { 2043 FLUSH_BLOCK(s, 1); 2044 return finish_done; 2045 } 2046 if (s->last_lit) 2047 FLUSH_BLOCK(s, 0); 2048 return block_done; 2049 } 2050 #endif /* FASTEST */ 2051 2052 /* =========================================================================== 2053 * For Z_RLE, simply look for runs of bytes, generate matches only of distance 2054 * one. Do not maintain a hash table. (It will be regenerated if this run of 2055 * deflate switches away from Z_RLE.) 2056 */ 2057 local block_state deflate_rle(s, flush) 2058 deflate_state *s; 2059 int flush; 2060 { 2061 int bflush; /* set if current block must be flushed */ 2062 uInt prev; /* byte at distance one to match */ 2063 Bytef *scan, *strend; /* scan goes up to strend for length of run */ 2064 2065 for (;;) { 2066 /* Make sure that we always have enough lookahead, except 2067 * at the end of the input file. We need MAX_MATCH bytes 2068 * for the longest run, plus one for the unrolled loop. 2069 */ 2070 if (s->lookahead <= MAX_MATCH) { 2071 fill_window(s); 2072 if (s->lookahead <= MAX_MATCH && flush == Z_NO_FLUSH) { 2073 return need_more; 2074 } 2075 if (s->lookahead == 0) break; /* flush the current block */ 2076 } 2077 2078 /* See how many times the previous byte repeats */ 2079 s->match_length = 0; 2080 if (s->lookahead >= MIN_MATCH && s->strstart > 0) { 2081 scan = s->window + s->strstart - 1; 2082 prev = *scan; 2083 if (prev == *++scan && prev == *++scan && prev == *++scan) { 2084 strend = s->window + s->strstart + MAX_MATCH; 2085 do { 2086 } while (prev == *++scan && prev == *++scan && 2087 prev == *++scan && prev == *++scan && 2088 prev == *++scan && prev == *++scan && 2089 prev == *++scan && prev == *++scan && 2090 scan < strend); 2091 s->match_length = MAX_MATCH - (uInt)(strend - scan); 2092 if (s->match_length > s->lookahead) 2093 s->match_length = s->lookahead; 2094 } 2095 Assert(scan <= s->window+(uInt)(s->window_size-1), "wild scan"); 2096 } 2097 2098 /* Emit match if have run of MIN_MATCH or longer, else emit literal */ 2099 if (s->match_length >= MIN_MATCH) { 2100 check_match(s, s->strstart, s->strstart - 1, s->match_length); 2101 2102 _tr_tally_dist(s, 1, s->match_length - MIN_MATCH, bflush); 2103 2104 s->lookahead -= s->match_length; 2105 s->strstart += s->match_length; 2106 s->match_length = 0; 2107 } else { 2108 /* No match, output a literal byte */ 2109 Tracevv((stderr,"%c", s->window[s->strstart])); 2110 _tr_tally_lit (s, s->window[s->strstart], bflush); 2111 s->lookahead--; 2112 s->strstart++; 2113 } 2114 if (bflush) FLUSH_BLOCK(s, 0); 2115 } 2116 s->insert = 0; 2117 if (flush == Z_FINISH) { 2118 FLUSH_BLOCK(s, 1); 2119 return finish_done; 2120 } 2121 if (s->last_lit) 2122 FLUSH_BLOCK(s, 0); 2123 return block_done; 2124 } 2125 2126 /* =========================================================================== 2127 * For Z_HUFFMAN_ONLY, do not look for matches. Do not maintain a hash table. 2128 * (It will be regenerated if this run of deflate switches away from Huffman.) 2129 */ 2130 local block_state deflate_huff(s, flush) 2131 deflate_state *s; 2132 int flush; 2133 { 2134 int bflush; /* set if current block must be flushed */ 2135 2136 for (;;) { 2137 /* Make sure that we have a literal to write. */ 2138 if (s->lookahead == 0) { 2139 fill_window(s); 2140 if (s->lookahead == 0) { 2141 if (flush == Z_NO_FLUSH) 2142 return need_more; 2143 break; /* flush the current block */ 2144 } 2145 } 2146 2147 /* Output a literal byte */ 2148 s->match_length = 0; 2149 Tracevv((stderr,"%c", s->window[s->strstart])); 2150 _tr_tally_lit (s, s->window[s->strstart], bflush); 2151 s->lookahead--; 2152 s->strstart++; 2153 if (bflush) FLUSH_BLOCK(s, 0); 2154 } 2155 s->insert = 0; 2156 if (flush == Z_FINISH) { 2157 FLUSH_BLOCK(s, 1); 2158 return finish_done; 2159 } 2160 if (s->last_lit) 2161 FLUSH_BLOCK(s, 0); 2162 return block_done; 2163 }