1 /*
2 * Copyright (c) 2008-2016 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28 /* deflate.c -- compress data using the deflation algorithm
29 * Copyright (C) 1995-2005 Jean-loup Gailly.
30 * For conditions of distribution and use, see copyright notice in zlib.h
31 */
32
33 /*
34 * ALGORITHM
35 *
36 * The "deflation" process depends on being able to identify portions
37 * of the input text which are identical to earlier input (within a
38 * sliding window trailing behind the input currently being processed).
39 *
40 * The most straightforward technique turns out to be the fastest for
41 * most input files: try all possible matches and select the longest.
42 * The key feature of this algorithm is that insertions into the string
43 * dictionary are very simple and thus fast, and deletions are avoided
44 * completely. Insertions are performed at each input character, whereas
45 * string matches are performed only when the previous match ends. So it
46 * is preferable to spend more time in matches to allow very fast string
47 * insertions and avoid deletions. The matching algorithm for small
48 * strings is inspired from that of Rabin & Karp. A brute force approach
49 * is used to find longer strings when a small match has been found.
50 * A similar algorithm is used in comic (by Jan-Mark Wams) and freeze
51 * (by Leonid Broukhis).
52 * A previous version of this file used a more sophisticated algorithm
53 * (by Fiala and Greene) which is guaranteed to run in linear amortized
54 * time, but has a larger average cost, uses more memory and is patented.
55 * However the F&G algorithm may be faster for some highly redundant
56 * files if the parameter max_chain_length (described below) is too large.
57 *
58 * ACKNOWLEDGEMENTS
59 *
60 * The idea of lazy evaluation of matches is due to Jan-Mark Wams, and
61 * I found it in 'freeze' written by Leonid Broukhis.
62 * Thanks to many people for bug reports and testing.
63 *
64 * REFERENCES
65 *
66 * Deutsch, L.P.,"DEFLATE Compressed Data Format Specification".
67 * Available in http://www.ietf.org/rfc/rfc1951.txt
68 *
69 * A description of the Rabin and Karp algorithm is given in the book
70 * "Algorithms" by R. Sedgewick, Addison-Wesley, p252.
71 *
72 * Fiala,E.R., and Greene,D.H.
73 * Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595
74 *
75 */
76
77 /* @(#) $Id$ */
78
79 #include "deflate.h"
80
81 const char deflate_copyright[] =
82 " deflate 1.2.3 Copyright 1995-2005 Jean-loup Gailly ";
83 /*
84 If you use the zlib library in a product, an acknowledgment is welcome
85 in the documentation of your product. If for some reason you cannot
86 include such an acknowledgment, I would appreciate that you keep this
87 copyright string in the executable of your product.
88 */
89
90 /* ===========================================================================
91 * Function prototypes.
92 */
93 typedef enum {
94 need_more, /* block not completed, need more input or more output */
95 block_done, /* block flush performed */
96 finish_started, /* finish started, need only more output at next deflate */
97 finish_done /* finish done, accept no more input or output */
98 } block_state;
99
100 typedef block_state (*compress_func) OF((deflate_state *s, int flush));
101 /* Compression function. Returns the block state after the call. */
102
103 local void fill_window OF((deflate_state *s));
104 local block_state deflate_stored OF((deflate_state *s, int flush));
105 local block_state deflate_fast OF((deflate_state *s, int flush));
106 #ifndef FASTEST
107 local block_state deflate_slow OF((deflate_state *s, int flush));
108 #endif
109 local void lm_init OF((deflate_state *s));
110 local void putShortMSB OF((deflate_state *s, uInt b));
111 local void flush_pending OF((z_streamp strm));
112 local int read_buf OF((z_streamp strm, Bytef *buf, unsigned size));
113 #ifndef FASTEST
114 #ifdef ASMV
115 void match_init OF((void)); /* asm code initialization */
116 uInt longest_match OF((deflate_state *s, IPos cur_match));
117 #else
118 local uInt longest_match OF((deflate_state *s, IPos cur_match));
119 #endif
120 #endif
121 local uInt longest_match_fast OF((deflate_state *s, IPos cur_match));
122
123 #ifdef DEBUG
124 local void check_match OF((deflate_state *s, IPos start, IPos match,
125 int length));
126 #endif
127
128 /* ===========================================================================
129 * Local data
130 */
131
132 #define NIL 0
133 /* Tail of hash chains */
134
135 #ifndef TOO_FAR
136 # define TOO_FAR 4096
137 #endif
138 /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
139
140 #define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
141 /* Minimum amount of lookahead, except at the end of the input file.
142 * See deflate.c for comments about the MIN_MATCH+1.
143 */
144
145 /* Values for max_lazy_match, good_match and max_chain_length, depending on
146 * the desired pack level (0..9). The values given below have been tuned to
147 * exclude worst case performance for pathological files. Better values may be
148 * found for specific files.
149 */
150 typedef struct config_s {
151 ush good_length; /* reduce lazy search above this match length */
152 ush max_lazy; /* do not perform lazy search above this match length */
153 ush nice_length; /* quit search above this match length */
154 ush max_chain;
155 compress_func func;
156 } config;
157
158 #ifdef FASTEST
159 local const config configuration_table[2] = {
160 /* good lazy nice chain */
161 /* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */
162 /* 1 */ {4, 4, 8, 4, deflate_fast}}; /* max speed, no lazy matches */
163 #else
164 local const config configuration_table[10] = {
165 /* good lazy nice chain */
166 /* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */
167 /* 1 */ {4, 4, 8, 4, deflate_fast}, /* max speed, no lazy matches */
168 /* 2 */ {4, 5, 16, 8, deflate_fast},
169 /* 3 */ {4, 6, 32, 32, deflate_fast},
170
171 /* 4 */ {4, 4, 16, 16, deflate_slow}, /* lazy matches */
172 /* 5 */ {8, 16, 32, 32, deflate_slow},
173 /* 6 */ {8, 16, 128, 128, deflate_slow},
174 /* 7 */ {8, 32, 128, 256, deflate_slow},
175 /* 8 */ {32, 128, 258, 1024, deflate_slow},
176 /* 9 */ {32, 258, 258, 4096, deflate_slow}}; /* max compression */
177 #endif
178
179 /* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
180 * For deflate_fast() (levels <= 3) good is ignored and lazy has a different
181 * meaning.
182 */
183
184 #define EQUAL 0
185 /* result of memcmp for equal strings */
186
187 #ifndef NO_DUMMY_DECL
188 struct static_tree_desc_s {int dummy;}; /* for buggy compilers */
189 #endif
190
191 /* ===========================================================================
192 * Update a hash value with the given input byte
193 * IN assertion: all calls to to UPDATE_HASH are made with consecutive
194 * input characters, so that a running hash key can be computed from the
195 * previous key instead of complete recalculation each time.
196 */
197 #define UPDATE_HASH(s,h,c) (h = (((h)<<s->hash_shift) ^ (c)) & s->hash_mask)
198
199
200 /* ===========================================================================
201 * Insert string str in the dictionary and set match_head to the previous head
202 * of the hash chain (the most recent string with same hash key). Return
203 * the previous length of the hash chain.
204 * If this file is compiled with -DFASTEST, the compression level is forced
205 * to 1, and no hash chains are maintained.
206 * IN assertion: all calls to to INSERT_STRING are made with consecutive
207 * input characters and the first MIN_MATCH bytes of str are valid
208 * (except for the last MIN_MATCH-1 bytes of the input file).
209 */
210 #ifdef FASTEST
211 #define INSERT_STRING(s, str, match_head) \
212 (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
213 match_head = s->head[s->ins_h], \
214 s->head[s->ins_h] = (Pos)(str))
215 #else
216 #define INSERT_STRING(s, str, match_head) \
217 (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
218 match_head = s->prev[(str) & s->w_mask] = s->head[s->ins_h], \
219 s->head[s->ins_h] = (Pos)(str))
220 #endif
221
222 /* ===========================================================================
223 * Initialize the hash table (avoiding 64K overflow for 16 bit systems).
224 * prev[] will be initialized on the fly.
225 */
226 #define CLEAR_HASH(s) \
227 s->head[s->hash_size-1] = NIL; \
228 zmemzero((Bytef *)s->head, (unsigned)(s->hash_size-1)*sizeof(*s->head));
229
230 /* ========================================================================= */
231 int ZEXPORT
deflateInit_(z_streamp strm,int level,const char * version,int stream_size)232 deflateInit_(z_streamp strm, int level, const char *version, 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_(z_streamp strm,int level,int method,int windowBits,int memLevel,int strategy,const char * version,int stream_size)241 deflateInit2_(z_streamp strm, int level, int method, int windowBits,
242 int memLevel, int strategy, const char *version,
243 int stream_size)
244 {
245 deflate_state *s;
246 int wrap = 1;
247 static const char my_version[] = ZLIB_VERSION;
248
249 ushf *overlay;
250 /* We overlay pending_buf and d_buf+l_buf. This works since the average
251 * output size for (length,distance) codes is <= 24 bits.
252 */
253
254 if (version == Z_NULL || version[0] != my_version[0] ||
255 stream_size != sizeof(z_stream)) {
256 return Z_VERSION_ERROR;
257 }
258 if (strm == Z_NULL) return Z_STREAM_ERROR;
259
260 strm->msg = Z_NULL;
261 #ifndef NO_ZCFUNCS
262 if (strm->zalloc == (alloc_func)0) {
263 strm->zalloc = zcalloc;
264 strm->opaque = (voidpf)0;
265 }
266 if (strm->zfree == (free_func)0) strm->zfree = zcfree;
267 #endif /* NO_ZCFUNCS */
268
269 #ifdef FASTEST
270 if (level != 0) level = 1;
271 #else
272 if (level == Z_DEFAULT_COMPRESSION) level = 6;
273 #endif
274
275 if (windowBits < 0) { /* suppress zlib wrapper */
276 wrap = 0;
277 windowBits = -windowBits;
278 }
279 #ifdef GZIP
280 else if (windowBits > 15) {
281 wrap = 2; /* write gzip wrapper instead */
282 windowBits -= 16;
283 }
284 #endif
285 if (memLevel < 1 || memLevel > MAX_MEM_LEVEL || method != Z_DEFLATED ||
286 windowBits < 8 || windowBits > 15 || level < 0 || level > 9 ||
287 strategy < 0 || strategy > Z_FIXED) {
288 return Z_STREAM_ERROR;
289 }
290 if (windowBits == 8) windowBits = 9; /* until 256-byte window bug fixed */
291 s = (deflate_state *) ZALLOC(strm, 1, sizeof(deflate_state));
292 if (s == Z_NULL) return Z_MEM_ERROR;
293 strm->state = (struct internal_state FAR *)s;
294 s->strm = strm;
295
296 s->wrap = wrap;
297 s->gzhead = Z_NULL;
298 s->w_bits = windowBits;
299 s->w_size = 1 << s->w_bits;
300 s->w_mask = s->w_size - 1;
301
302 s->hash_bits = memLevel + 7;
303 s->hash_size = 1 << s->hash_bits;
304 s->hash_mask = s->hash_size - 1;
305 s->hash_shift = ((s->hash_bits+MIN_MATCH-1)/MIN_MATCH);
306
307 s->window = (Bytef *) ZALLOC(strm, s->w_size, 2*sizeof(Byte));
308 s->prev = (Posf *) ZALLOC(strm, s->w_size, sizeof(Pos));
309 s->head = (Posf *) ZALLOC(strm, s->hash_size, sizeof(Pos));
310
311 s->lit_bufsize = 1 << (memLevel + 6); /* 16K elements by default */
312
313 overlay = (ushf *) ZALLOC(strm, s->lit_bufsize, sizeof(ush)+2);
314 s->pending_buf = (uchf *) overlay;
315 s->pending_buf_size = (ulg)s->lit_bufsize * (sizeof(ush)+2L);
316
317 if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL ||
318 s->pending_buf == Z_NULL) {
319 s->status = FINISH_STATE;
320 strm->msg = (char*)ERR_MSG(Z_MEM_ERROR);
321 deflateEnd (strm);
322 return Z_MEM_ERROR;
323 }
324 s->d_buf = overlay + s->lit_bufsize/sizeof(ush);
325 s->l_buf = s->pending_buf + (1+sizeof(ush))*s->lit_bufsize;
326
327 s->level = level;
328 s->strategy = strategy;
329 s->method = (Byte)method;
330
331 return deflateReset(strm);
332 }
333
334 /* ========================================================================= */
335 int ZEXPORT
deflateSetDictionary(z_streamp strm,const Bytef * dictionary,uInt dictLength)336 deflateSetDictionary(z_streamp strm, const Bytef *dictionary, uInt dictLength)
337 {
338 deflate_state *s;
339 uInt length = dictLength;
340 uInt n;
341 IPos hash_head = 0;
342
343 if (strm == Z_NULL || strm->state == Z_NULL || dictionary == Z_NULL ||
344 strm->state->wrap == 2 ||
345 (strm->state->wrap == 1 && strm->state->status != INIT_STATE))
346 return Z_STREAM_ERROR;
347
348 s = strm->state;
349 if (s->wrap)
350 strm->adler = adler32(strm->adler, dictionary, dictLength);
351
352 if (length < MIN_MATCH) return Z_OK;
353 if (length > MAX_DIST(s)) {
354 length = MAX_DIST(s);
355 dictionary += dictLength - length; /* use the tail of the dictionary */
356 }
357 zmemcpy(s->window, dictionary, length);
358 s->strstart = length;
359 s->block_start = (long)length;
360
361 /* Insert all strings in the hash table (except for the last two bytes).
362 * s->lookahead stays null, so s->ins_h will be recomputed at the next
363 * call of fill_window.
364 */
365 s->ins_h = s->window[0];
366 UPDATE_HASH(s, s->ins_h, s->window[1]);
367 for (n = 0; n <= length - MIN_MATCH; n++) {
368 INSERT_STRING(s, n, hash_head);
369 }
370 if (hash_head) hash_head = 0; /* to make compiler happy */
371 return Z_OK;
372 }
373
374 /* ========================================================================= */
375
376 ZEXTERN int ZEXPORT
deflateResetWithIO(z_streamp strm,z_input_func zinput,z_output_func zoutput)377 deflateResetWithIO(z_streamp strm, z_input_func zinput, z_output_func zoutput)
378 {
379 int zerr;
380
381 zerr = deflateReset(strm);
382 if (Z_OK != zerr) return (zerr);
383 strm->state->zinput = zinput;
384 strm->state->zoutput = zoutput;
385 return Z_OK;
386 }
387
388 /* ========================================================================= */
389
390 int ZEXPORT
deflateReset(z_streamp strm)391 deflateReset(z_streamp strm)
392 {
393 deflate_state *s;
394
395 if (strm == Z_NULL || strm->state == Z_NULL ||
396 strm->zalloc == (alloc_func)0 || strm->zfree == (free_func)0) {
397 return Z_STREAM_ERROR;
398 }
399
400 strm->total_in = strm->total_out = 0;
401 strm->msg = Z_NULL; /* use zfree if we ever allocate msg dynamically */
402 strm->data_type = Z_UNKNOWN;
403
404 s = (deflate_state *)strm->state;
405 s->pending = 0;
406 s->pending_out = s->pending_buf;
407 s->zinput = &read_buf;
408 s->zoutput = NULL;
409
410 if (s->wrap < 0) {
411 s->wrap = -s->wrap; /* was made negative by deflate(..., Z_FINISH); */
412 }
413 s->status = s->wrap ? INIT_STATE : BUSY_STATE;
414 strm->adler =
415 #ifdef GZIP
416 s->wrap == 2 ? z_crc32(0L, Z_NULL, 0) :
417 #endif
418 adler32(0L, Z_NULL, 0);
419 s->last_flush = Z_NO_FLUSH;
420
421 _tr_init(s);
422 lm_init(s);
423
424 return Z_OK;
425 }
426
427 /* ========================================================================= */
428 int ZEXPORT
deflateSetHeader(z_streamp strm,gz_headerp head)429 deflateSetHeader(z_streamp strm, gz_headerp head)
430 {
431 if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
432 if (strm->state->wrap != 2) return Z_STREAM_ERROR;
433 strm->state->gzhead = head;
434 return Z_OK;
435 }
436
437 /* ========================================================================= */
438 int ZEXPORT
deflatePrime(z_streamp strm,int bits,int value)439 deflatePrime(z_streamp strm, int bits, int value)
440 {
441 if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
442 strm->state->bi_valid = bits;
443 strm->state->bi_buf = (ush)(value & ((1 << bits) - 1));
444 return Z_OK;
445 }
446
447 /* ========================================================================= */
448 int ZEXPORT
deflateParams(z_streamp strm,int level,int strategy)449 deflateParams(z_streamp strm, int level, int strategy)
450 {
451 deflate_state *s;
452 compress_func func;
453 int err = Z_OK;
454
455 if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
456 s = strm->state;
457
458 #ifdef FASTEST
459 if (level != 0) level = 1;
460 #else
461 if (level == Z_DEFAULT_COMPRESSION) level = 6;
462 #endif
463 if (level < 0 || level > 9 || strategy < 0 || strategy > Z_FIXED) {
464 return Z_STREAM_ERROR;
465 }
466 func = configuration_table[s->level].func;
467
468 if (func != configuration_table[level].func && strm->total_in != 0) {
469 /* Flush the last buffer: */
470 err = deflate(strm, Z_PARTIAL_FLUSH);
471 }
472 if (s->level != level) {
473 s->level = level;
474 s->max_lazy_match = configuration_table[level].max_lazy;
475 s->good_match = configuration_table[level].good_length;
476 s->nice_match = configuration_table[level].nice_length;
477 s->max_chain_length = configuration_table[level].max_chain;
478 }
479 s->strategy = strategy;
480 return err;
481 }
482
483 /* ========================================================================= */
484 int ZEXPORT
deflateTune(z_streamp strm,int good_length,int max_lazy,int nice_length,int max_chain)485 deflateTune(z_streamp strm, int good_length, int max_lazy, int nice_length,
486 int max_chain)
487 {
488 deflate_state *s;
489
490 if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
491 s = strm->state;
492 s->good_match = good_length;
493 s->max_lazy_match = max_lazy;
494 s->nice_match = nice_length;
495 s->max_chain_length = max_chain;
496 return Z_OK;
497 }
498
499 /* =========================================================================
500 * For the default windowBits of 15 and memLevel of 8, this function returns
501 * a close to exact, as well as small, upper bound on the compressed size.
502 * They are coded as constants here for a reason--if the #define's are
503 * changed, then this function needs to be changed as well. The return
504 * value for 15 and 8 only works for those exact settings.
505 *
506 * For any setting other than those defaults for windowBits and memLevel,
507 * the value returned is a conservative worst case for the maximum expansion
508 * resulting from using fixed blocks instead of stored blocks, which deflate
509 * can emit on compressed data for some combinations of the parameters.
510 *
511 * This function could be more sophisticated to provide closer upper bounds
512 * for every combination of windowBits and memLevel, as well as wrap.
513 * But even the conservative upper bound of about 14% expansion does not
514 * seem onerous for output buffer allocation.
515 */
516 uLong ZEXPORT
deflateBound(z_streamp strm,uLong sourceLen)517 deflateBound(z_streamp strm, uLong sourceLen)
518 {
519 deflate_state *s;
520 uLong destLen;
521
522 /* conservative upper bound */
523 destLen = sourceLen +
524 ((sourceLen + 7) >> 3) + ((sourceLen + 63) >> 6) + 11;
525
526 /* if can't get parameters, return conservative bound */
527 if (strm == Z_NULL || strm->state == Z_NULL)
528 return destLen;
529
530 /* if not default parameters, return conservative bound */
531 s = strm->state;
532 if (s->w_bits != 15 || s->hash_bits != 8 + 7)
533 return destLen;
534
535 /* default settings: return tight bound for that case */
536 return compressBound(sourceLen);
537 }
538
539 /* =========================================================================
540 * Put a short in the pending buffer. The 16-bit value is put in MSB order.
541 * IN assertion: the stream state is correct and there is enough room in
542 * pending_buf.
543 */
544 local void
putShortMSB(deflate_state * s,uInt b)545 putShortMSB(deflate_state *s, uInt b)
546 {
547 put_byte(s, (Byte)(b >> 8));
548 put_byte(s, (Byte)(b & 0xff));
549 }
550
551 /* =========================================================================
552 * Flush as much pending output as possible. All deflate() output goes
553 * through this function so some applications may wish to modify it
554 * to avoid allocating a large strm->next_out buffer and copying into it.
555 * (See also read_buf()).
556 */
557 local void
flush_pending(z_streamp strm)558 flush_pending(z_streamp strm)
559 {
560 unsigned len = strm->state->pending;
561
562 if (strm->state->zoutput) {
563 len = (*strm->state->zoutput)(strm, strm->state->pending_out, len);
564 } else {
565 if (len > strm->avail_out) len = strm->avail_out;
566 if (len == 0) return;
567 zmemcpy(strm->next_out, strm->state->pending_out, len);
568 strm->next_out += len;
569 strm->avail_out -= len;
570 }
571
572 strm->state->pending_out += len;
573 strm->total_out += len;
574 strm->state->pending -= len;
575 if (strm->state->pending == 0) {
576 strm->state->pending_out = strm->state->pending_buf;
577 }
578 }
579
580 /* ========================================================================= */
581 int ZEXPORT
deflate(z_streamp strm,int flush)582 deflate(z_streamp strm, int flush)
583 {
584 int old_flush; /* value of flush param for previous deflate call */
585 deflate_state *s;
586
587 if (strm == Z_NULL || strm->state == Z_NULL ||
588 flush > Z_FINISH || flush < 0) {
589 return Z_STREAM_ERROR;
590 }
591 s = strm->state;
592
593 if (strm->next_out == Z_NULL ||
594 (strm->next_in == Z_NULL && strm->avail_in != 0) ||
595 (s->status == FINISH_STATE && flush != Z_FINISH)) {
596 ERR_RETURN(strm, Z_STREAM_ERROR);
597 }
598 if (strm->avail_out == 0) ERR_RETURN(strm, Z_BUF_ERROR);
599
600 s->strm = strm; /* just in case */
601 old_flush = s->last_flush;
602 s->last_flush = flush;
603
604 /* Write the header */
605 if (s->status == INIT_STATE) {
606 #ifdef GZIP
607 if (s->wrap == 2) {
608 strm->adler = z_crc32(0L, Z_NULL, 0);
609 put_byte(s, 31);
610 put_byte(s, 139);
611 put_byte(s, 8);
612 if (s->gzhead == NULL) {
613 put_byte(s, 0);
614 put_byte(s, 0);
615 put_byte(s, 0);
616 put_byte(s, 0);
617 put_byte(s, 0);
618 put_byte(s, s->level == 9 ? 2 :
619 (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
620 4 : 0));
621 put_byte(s, OS_CODE);
622 s->status = BUSY_STATE;
623 }
624 else {
625 put_byte(s, (s->gzhead->text ? 1 : 0) +
626 (s->gzhead->hcrc ? 2 : 0) +
627 (s->gzhead->extra == Z_NULL ? 0 : 4) +
628 (s->gzhead->name == Z_NULL ? 0 : 8) +
629 (s->gzhead->comment == Z_NULL ? 0 : 16)
630 );
631 put_byte(s, (Byte)(s->gzhead->time & 0xff));
632 put_byte(s, (Byte)((s->gzhead->time >> 8) & 0xff));
633 put_byte(s, (Byte)((s->gzhead->time >> 16) & 0xff));
634 put_byte(s, (Byte)((s->gzhead->time >> 24) & 0xff));
635 put_byte(s, s->level == 9 ? 2 :
636 (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
637 4 : 0));
638 put_byte(s, s->gzhead->os & 0xff);
639 if (s->gzhead->extra != NULL) {
640 put_byte(s, s->gzhead->extra_len & 0xff);
641 put_byte(s, (s->gzhead->extra_len >> 8) & 0xff);
642 }
643 if (s->gzhead->hcrc)
644 strm->adler = z_crc32(strm->adler, s->pending_buf,
645 s->pending);
646 s->gzindex = 0;
647 s->status = EXTRA_STATE;
648 }
649 }
650 else
651 #endif
652 {
653 uInt header = (Z_DEFLATED + ((s->w_bits-8)<<4)) << 8;
654 uInt level_flags;
655
656 if (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2)
657 level_flags = 0;
658 else if (s->level < 6)
659 level_flags = 1;
660 else if (s->level == 6)
661 level_flags = 2;
662 else
663 level_flags = 3;
664 header |= (level_flags << 6);
665 if (s->strstart != 0) header |= PRESET_DICT;
666 header += 31 - (header % 31);
667
668 s->status = BUSY_STATE;
669 putShortMSB(s, header);
670
671 /* Save the adler32 of the preset dictionary: */
672 if (s->strstart != 0) {
673 putShortMSB(s, (uInt)(strm->adler >> 16));
674 putShortMSB(s, (uInt)(strm->adler & 0xffff));
675 }
676 strm->adler = adler32(0L, Z_NULL, 0);
677 }
678 }
679 #ifdef GZIP
680 if (s->status == EXTRA_STATE) {
681 if (s->gzhead->extra != NULL) {
682 uInt beg = s->pending; /* start of bytes to update crc */
683
684 while (s->gzindex < (s->gzhead->extra_len & 0xffff)) {
685 if (s->pending == s->pending_buf_size) {
686 if (s->gzhead->hcrc && s->pending > beg)
687 strm->adler = z_crc32(strm->adler, s->pending_buf + beg,
688 s->pending - beg);
689 flush_pending(strm);
690 beg = s->pending;
691 if (s->pending == s->pending_buf_size)
692 break;
693 }
694 put_byte(s, s->gzhead->extra[s->gzindex]);
695 s->gzindex++;
696 }
697 if (s->gzhead->hcrc && s->pending > beg)
698 strm->adler = z_crc32(strm->adler, s->pending_buf + beg,
699 s->pending - beg);
700 if (s->gzindex == s->gzhead->extra_len) {
701 s->gzindex = 0;
702 s->status = NAME_STATE;
703 }
704 }
705 else
706 s->status = NAME_STATE;
707 }
708 if (s->status == NAME_STATE) {
709 if (s->gzhead->name != NULL) {
710 uInt beg = s->pending; /* start of bytes to update crc */
711 Bytef val;
712
713 do {
714 if (s->pending == s->pending_buf_size) {
715 if (s->gzhead->hcrc && s->pending > beg)
716 strm->adler = z_crc32(strm->adler, s->pending_buf + beg,
717 s->pending - beg);
718 flush_pending(strm);
719 beg = s->pending;
720 if (s->pending == s->pending_buf_size) {
721 val = 1;
722 break;
723 }
724 }
725 val = s->gzhead->name[s->gzindex++];
726 put_byte(s, val);
727 } while (val != 0);
728 if (s->gzhead->hcrc && s->pending > beg)
729 strm->adler = z_crc32(strm->adler, s->pending_buf + beg,
730 s->pending - beg);
731 if (val == 0) {
732 s->gzindex = 0;
733 s->status = COMMENT_STATE;
734 }
735 }
736 else
737 s->status = COMMENT_STATE;
738 }
739 if (s->status == COMMENT_STATE) {
740 if (s->gzhead->comment != NULL) {
741 uInt beg = s->pending; /* start of bytes to update crc */
742 Bytef val;
743
744 do {
745 if (s->pending == s->pending_buf_size) {
746 if (s->gzhead->hcrc && s->pending > beg)
747 strm->adler = z_crc32(strm->adler, s->pending_buf + beg,
748 s->pending - beg);
749 flush_pending(strm);
750 beg = s->pending;
751 if (s->pending == s->pending_buf_size) {
752 val = 1;
753 break;
754 }
755 }
756 val = s->gzhead->comment[s->gzindex++];
757 put_byte(s, val);
758 } while (val != 0);
759 if (s->gzhead->hcrc && s->pending > beg)
760 strm->adler = z_crc32(strm->adler, s->pending_buf + beg,
761 s->pending - beg);
762 if (val == 0)
763 s->status = HCRC_STATE;
764 }
765 else
766 s->status = HCRC_STATE;
767 }
768 if (s->status == HCRC_STATE) {
769 if (s->gzhead->hcrc) {
770 if (s->pending + 2 > s->pending_buf_size)
771 flush_pending(strm);
772 if (s->pending + 2 <= s->pending_buf_size) {
773 put_byte(s, (Byte)(strm->adler & 0xff));
774 put_byte(s, (Byte)((strm->adler >> 8) & 0xff));
775 strm->adler = z_crc32(0L, Z_NULL, 0);
776 s->status = BUSY_STATE;
777 }
778 }
779 else
780 s->status = BUSY_STATE;
781 }
782 #endif
783
784 /* Flush as much pending output as possible */
785 if (s->pending != 0) {
786 flush_pending(strm);
787 if (strm->avail_out == 0) {
788 /* Since avail_out is 0, deflate will be called again with
789 * more output space, but possibly with both pending and
790 * avail_in equal to zero. There won't be anything to do,
791 * but this is not an error situation so make sure we
792 * return OK instead of BUF_ERROR at next call of deflate:
793 */
794 s->last_flush = -1;
795 return Z_OK;
796 }
797
798 /* Make sure there is something to do and avoid duplicate consecutive
799 * flushes. For repeated and useless calls with Z_FINISH, we keep
800 * returning Z_STREAM_END instead of Z_BUF_ERROR.
801 */
802 } else if (strm->avail_in == 0 && flush <= old_flush &&
803 flush != Z_FINISH) {
804 ERR_RETURN(strm, Z_BUF_ERROR);
805 }
806
807 /* User must not provide more input after the first FINISH: */
808 if (s->status == FINISH_STATE && strm->avail_in != 0) {
809 ERR_RETURN(strm, Z_BUF_ERROR);
810 }
811
812 /* Start a new block or continue the current one.
813 */
814 if (strm->avail_in != 0 || s->lookahead != 0 ||
815 (flush != Z_NO_FLUSH && s->status != FINISH_STATE)) {
816 block_state bstate;
817
818 bstate = (*(configuration_table[s->level].func))(s, flush);
819
820 if (bstate == finish_started || bstate == finish_done) {
821 s->status = FINISH_STATE;
822 }
823 if (bstate == need_more || bstate == finish_started) {
824 if (strm->avail_out == 0) {
825 s->last_flush = -1; /* avoid BUF_ERROR next call, see above */
826 }
827 return Z_OK;
828 /* If flush != Z_NO_FLUSH && avail_out == 0, the next call
829 * of deflate should use the same flush parameter to make sure
830 * that the flush is complete. So we don't have to output an
831 * empty block here, this will be done at next call. This also
832 * ensures that for a very small output buffer, we emit at most
833 * one empty block.
834 */
835 }
836 if (bstate == block_done) {
837 if (flush == Z_PARTIAL_FLUSH) {
838 _tr_align(s);
839 } else { /* FULL_FLUSH or SYNC_FLUSH */
840 _tr_stored_block(s, (char*)0, 0L, 0);
841 /* For a full flush, this empty block will be recognized
842 * as a special marker by inflate_sync().
843 */
844 if (flush == Z_FULL_FLUSH) {
845 CLEAR_HASH(s); /* forget history */
846 }
847 }
848 flush_pending(strm);
849 if (strm->avail_out == 0) {
850 s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */
851 return Z_OK;
852 }
853 }
854 }
855 Assert(strm->avail_out > 0, "bug2");
856
857 if (flush != Z_FINISH) return Z_OK;
858 if (s->wrap <= 0) return Z_STREAM_END;
859
860 /* Write the trailer */
861 #ifdef GZIP
862 if (s->wrap == 2) {
863 put_byte(s, (Byte)(strm->adler & 0xff));
864 put_byte(s, (Byte)((strm->adler >> 8) & 0xff));
865 put_byte(s, (Byte)((strm->adler >> 16) & 0xff));
866 put_byte(s, (Byte)((strm->adler >> 24) & 0xff));
867 put_byte(s, (Byte)(strm->total_in & 0xff));
868 put_byte(s, (Byte)((strm->total_in >> 8) & 0xff));
869 put_byte(s, (Byte)((strm->total_in >> 16) & 0xff));
870 put_byte(s, (Byte)((strm->total_in >> 24) & 0xff));
871 }
872 else
873 #endif
874 {
875 putShortMSB(s, (uInt)(strm->adler >> 16));
876 putShortMSB(s, (uInt)(strm->adler & 0xffff));
877 }
878 flush_pending(strm);
879 /* If avail_out is zero, the application will call deflate again
880 * to flush the rest.
881 */
882 if (s->wrap > 0) s->wrap = -s->wrap; /* write the trailer only once! */
883 return s->pending != 0 ? Z_OK : Z_STREAM_END;
884 }
885
886 /* ========================================================================= */
887 int ZEXPORT
deflateEnd(z_streamp strm)888 deflateEnd(z_streamp strm)
889 {
890 int status;
891
892 if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
893
894 status = strm->state->status;
895 if (status != INIT_STATE &&
896 status != EXTRA_STATE &&
897 status != NAME_STATE &&
898 status != COMMENT_STATE &&
899 status != HCRC_STATE &&
900 status != BUSY_STATE &&
901 status != FINISH_STATE) {
902 return Z_STREAM_ERROR;
903 }
904
905 /* Deallocate in reverse order of allocations: */
906 TRY_FREE(strm, strm->state->pending_buf);
907 TRY_FREE(strm, strm->state->head);
908 TRY_FREE(strm, strm->state->prev);
909 TRY_FREE(strm, strm->state->window);
910
911 ZFREE(strm, strm->state);
912 strm->state = Z_NULL;
913
914 return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK;
915 }
916
917 /* =========================================================================
918 * Copy the source state to the destination state.
919 * To simplify the source, this is not supported for 16-bit MSDOS (which
920 * doesn't have enough memory anyway to duplicate compression states).
921 */
922 int ZEXPORT
deflateCopy(z_streamp dest,z_streamp source)923 deflateCopy(z_streamp dest, z_streamp source)
924 {
925 #ifdef MAXSEG_64K
926 return Z_STREAM_ERROR;
927 #else
928 deflate_state *ds;
929 deflate_state *ss;
930 ushf *overlay;
931
932
933 if (source == Z_NULL || dest == Z_NULL || source->state == Z_NULL) {
934 return Z_STREAM_ERROR;
935 }
936
937 ss = source->state;
938
939 zmemcpy(dest, source, sizeof(z_stream));
940
941 ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state));
942 if (ds == Z_NULL) return Z_MEM_ERROR;
943 dest->state = (struct internal_state FAR *) ds;
944 zmemcpy(ds, ss, sizeof(deflate_state));
945 ds->strm = dest;
946
947 ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte));
948 ds->prev = (Posf *) ZALLOC(dest, ds->w_size, sizeof(Pos));
949 ds->head = (Posf *) ZALLOC(dest, ds->hash_size, sizeof(Pos));
950 overlay = (ushf *) ZALLOC(dest, ds->lit_bufsize, sizeof(ush)+2);
951 ds->pending_buf = (uchf *) overlay;
952
953 if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL ||
954 ds->pending_buf == Z_NULL) {
955 deflateEnd (dest);
956 return Z_MEM_ERROR;
957 }
958 /* following zmemcpy do not work for 16-bit MSDOS */
959 zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte));
960 zmemcpy(ds->prev, ss->prev, ds->w_size * sizeof(Pos));
961 zmemcpy(ds->head, ss->head, ds->hash_size * sizeof(Pos));
962 zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size);
963
964 ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf);
965 ds->d_buf = overlay + ds->lit_bufsize/sizeof(ush);
966 ds->l_buf = ds->pending_buf + (1+sizeof(ush))*ds->lit_bufsize;
967
968 ds->l_desc.dyn_tree = ds->dyn_ltree;
969 ds->d_desc.dyn_tree = ds->dyn_dtree;
970 ds->bl_desc.dyn_tree = ds->bl_tree;
971
972 return Z_OK;
973 #endif /* MAXSEG_64K */
974 }
975
976 /* ===========================================================================
977 * Read a new buffer from the current input stream, update the adler32
978 * and total number of bytes read. All deflate() input goes through
979 * this function so some applications may wish to modify it to avoid
980 * allocating a large strm->next_in buffer and copying from it.
981 * (See also flush_pending()).
982 */
983 local int
read_buf(z_streamp strm,Bytef * buf,unsigned size)984 read_buf(z_streamp strm, Bytef *buf, unsigned size)
985 {
986 unsigned len = strm->avail_in;
987
988 if (len > size) len = size;
989 if (len == 0) return 0;
990
991 strm->avail_in -= len;
992
993 if (strm->state->wrap == 1) {
994 strm->adler = adler32(strm->adler, strm->next_in, len);
995 }
996 #ifdef GZIP
997 else if (strm->state->wrap == 2) {
998 strm->adler = z_crc32(strm->adler, strm->next_in, len);
999 }
1000 #endif
1001 zmemcpy(buf, strm->next_in, len);
1002 strm->next_in += len;
1003 strm->total_in += len;
1004
1005 return (int)len;
1006 }
1007
1008 /* ===========================================================================
1009 * Initialize the "longest match" routines for a new zlib stream
1010 */
1011 local void
lm_init(deflate_state * s)1012 lm_init(deflate_state *s)
1013 {
1014 s->window_size = (ulg)2L*s->w_size;
1015
1016 CLEAR_HASH(s);
1017
1018 /* Set the default configuration parameters:
1019 */
1020 s->max_lazy_match = configuration_table[s->level].max_lazy;
1021 s->good_match = configuration_table[s->level].good_length;
1022 s->nice_match = configuration_table[s->level].nice_length;
1023 s->max_chain_length = configuration_table[s->level].max_chain;
1024
1025 s->strstart = 0;
1026 s->block_start = 0L;
1027 s->lookahead = 0;
1028 s->match_length = s->prev_length = MIN_MATCH-1;
1029 s->match_available = 0;
1030 s->ins_h = 0;
1031 #ifndef FASTEST
1032 #ifdef ASMV
1033 match_init(); /* initialize the asm code */
1034 #endif
1035 #endif
1036 }
1037
1038 #ifndef FASTEST
1039 /* ===========================================================================
1040 * Set match_start to the longest match starting at the given string and
1041 * return its length. Matches shorter or equal to prev_length are discarded,
1042 * in which case the result is equal to prev_length and match_start is
1043 * garbage.
1044 * IN assertions: cur_match is the head of the hash chain for the current
1045 * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
1046 * OUT assertion: the match length is not greater than s->lookahead.
1047 */
1048 #ifndef ASMV
1049 /* For 80x86 and 680x0, an optimized version will be provided in match.asm or
1050 * match.S. The code will be functionally equivalent.
1051 * @param cur_match current match
1052 */
1053 local uInt
longest_match(deflate_state * s,IPos cur_match)1054 longest_match(deflate_state *s, IPos cur_match)
1055 {
1056 unsigned chain_length = s->max_chain_length;/* max hash chain length */
1057 Bytef *scan = s->window + s->strstart; /* current string */
1058 Bytef *match; /* matched string */
1059 int len; /* length of current match */
1060 int best_len = s->prev_length; /* best match length so far */
1061 int nice_match = s->nice_match; /* stop if match long enough */
1062 IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
1063 s->strstart - (IPos)MAX_DIST(s) : NIL;
1064 /* Stop when cur_match becomes <= limit. To simplify the code,
1065 * we prevent matches with the string of window index 0.
1066 */
1067 Posf *prev = s->prev;
1068 uInt wmask = s->w_mask;
1069
1070 #ifdef UNALIGNED_OK
1071 /* Compare two bytes at a time. Note: this is not always beneficial.
1072 * Try with and without -DUNALIGNED_OK to check.
1073 */
1074 Bytef *strend = s->window + s->strstart + MAX_MATCH - 1;
1075 ush scan_start = *(ushf*)scan;
1076 ush scan_end = *(ushf*)(scan+best_len-1);
1077 #else
1078 Bytef *strend = s->window + s->strstart + MAX_MATCH;
1079 Byte scan_end1 = scan[best_len-1];
1080 Byte scan_end = scan[best_len];
1081 #endif
1082
1083 /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
1084 * It is easy to get rid of this optimization if necessary.
1085 */
1086 Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
1087
1088 /* Do not waste too much time if we already have a good match: */
1089 if (s->prev_length >= s->good_match) {
1090 chain_length >>= 2;
1091 }
1092 /* Do not look for matches beyond the end of the input. This is necessary
1093 * to make deflate deterministic.
1094 */
1095 if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead;
1096
1097 Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
1098
1099 do {
1100 Assert(cur_match < s->strstart, "no future");
1101 match = s->window + cur_match;
1102
1103 /* Skip to next match if the match length cannot increase
1104 * or if the match length is less than 2. Note that the checks below
1105 * for insufficient lookahead only occur occasionally for performance
1106 * reasons. Therefore uninitialized memory will be accessed, and
1107 * conditional jumps will be made that depend on those values.
1108 * However the length of the match is limited to the lookahead, so
1109 * the output of deflate is not affected by the uninitialized values.
1110 */
1111 #if (defined(UNALIGNED_OK) && MAX_MATCH == 258)
1112 /* This code assumes sizeof(unsigned short) == 2. Do not use
1113 * UNALIGNED_OK if your compiler uses a different size.
1114 */
1115 if (*(ushf*)(match+best_len-1) != scan_end ||
1116 *(ushf*)match != scan_start) continue;
1117
1118 /* It is not necessary to compare scan[2] and match[2] since they are
1119 * always equal when the other bytes match, given that the hash keys
1120 * are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at
1121 * strstart+3, +5, ... up to strstart+257. We check for insufficient
1122 * lookahead only every 4th comparison; the 128th check will be made
1123 * at strstart+257. If MAX_MATCH-2 is not a multiple of 8, it is
1124 * necessary to put more guard bytes at the end of the window, or
1125 * to check more often for insufficient lookahead.
1126 */
1127 Assert(scan[2] == match[2], "scan[2]?");
1128 scan++, match++;
1129 do {
1130 } while (*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
1131 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
1132 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
1133 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
1134 scan < strend);
1135 /* The funny "do {}" generates better code on most compilers */
1136
1137 /* Here, scan <= window+strstart+257 */
1138 Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
1139 if (*scan == *match) scan++;
1140
1141 len = (MAX_MATCH - 1) - (int)(strend-scan);
1142 scan = strend - (MAX_MATCH-1);
1143
1144 #else /* UNALIGNED_OK */
1145
1146 if (match[best_len] != scan_end ||
1147 match[best_len-1] != scan_end1 ||
1148 *match != *scan ||
1149 *++match != scan[1]) continue;
1150
1151 /* The check at best_len-1 can be removed because it will be made
1152 * again later. (This heuristic is not always a win.)
1153 * It is not necessary to compare scan[2] and match[2] since they
1154 * are always equal when the other bytes match, given that
1155 * the hash keys are equal and that HASH_BITS >= 8.
1156 */
1157 scan += 2;
1158 match++;
1159 Assert(*scan == *match, "match[2]?");
1160
1161 /* We check for insufficient lookahead only every 8th comparison;
1162 * the 256th check will be made at strstart+258.
1163 */
1164 do {
1165 } while (*++scan == *++match && *++scan == *++match &&
1166 *++scan == *++match && *++scan == *++match &&
1167 *++scan == *++match && *++scan == *++match &&
1168 *++scan == *++match && *++scan == *++match &&
1169 scan < strend);
1170
1171 Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
1172
1173 len = MAX_MATCH - (int)(strend - scan);
1174 scan = strend - MAX_MATCH;
1175
1176 #endif /* UNALIGNED_OK */
1177
1178 if (len > best_len) {
1179 s->match_start = cur_match;
1180 best_len = len;
1181 if (len >= nice_match) break;
1182 #ifdef UNALIGNED_OK
1183 scan_end = *(ushf*)(scan+best_len-1);
1184 #else
1185 scan_end1 = scan[best_len-1];
1186 scan_end = scan[best_len];
1187 #endif
1188 }
1189 } while ((cur_match = prev[cur_match & wmask]) > limit
1190 && --chain_length != 0);
1191
1192 if ((uInt)best_len <= s->lookahead) return (uInt)best_len;
1193 return s->lookahead;
1194 }
1195 #endif /* ASMV */
1196 #endif /* FASTEST */
1197
1198 /* ---------------------------------------------------------------------------
1199 * Optimized version for level == 1 or strategy == Z_RLE only
1200 * @param cur_match current match
1201 */
1202 local uInt
longest_match_fast(deflate_state * s,IPos cur_match)1203 longest_match_fast(deflate_state *s, IPos cur_match)
1204 {
1205 Bytef *scan = s->window + s->strstart; /* current string */
1206 Bytef *match; /* matched string */
1207 int len; /* length of current match */
1208 Bytef *strend = s->window + s->strstart + MAX_MATCH;
1209
1210 /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
1211 * It is easy to get rid of this optimization if necessary.
1212 */
1213 Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
1214
1215 Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
1216
1217 Assert(cur_match < s->strstart, "no future");
1218
1219 match = s->window + cur_match;
1220
1221 /* Return failure if the match length is less than 2:
1222 */
1223 if (match[0] != scan[0] || match[1] != scan[1]) return MIN_MATCH-1;
1224
1225 /* The check at best_len-1 can be removed because it will be made
1226 * again later. (This heuristic is not always a win.)
1227 * It is not necessary to compare scan[2] and match[2] since they
1228 * are always equal when the other bytes match, given that
1229 * the hash keys are equal and that HASH_BITS >= 8.
1230 */
1231 scan += 2;
1232 match += 2;
1233 Assert(*scan == *match, "match[2]?");
1234
1235 /* We check for insufficient lookahead only every 8th comparison;
1236 * the 256th check will be made at strstart+258.
1237 */
1238 do {
1239 } while (*++scan == *++match && *++scan == *++match &&
1240 *++scan == *++match && *++scan == *++match &&
1241 *++scan == *++match && *++scan == *++match &&
1242 *++scan == *++match && *++scan == *++match &&
1243 scan < strend);
1244
1245 Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
1246
1247 len = MAX_MATCH - (int)(strend - scan);
1248
1249 if (len < MIN_MATCH) return MIN_MATCH - 1;
1250
1251 s->match_start = cur_match;
1252 return (uInt)len <= s->lookahead ? (uInt)len : s->lookahead;
1253 }
1254
1255 #ifdef DEBUG
1256 /* ===========================================================================
1257 * Check that the match at match_start is indeed a match.
1258 */
1259 local void
check_match(deflate_state * s,IPos start,IPos match,int length)1260 check_match(deflate_state *s, IPos start, IPos match, int length)
1261 {
1262 /* check that the match is indeed a match */
1263 if (zmemcmp(s->window + match,
1264 s->window + start, length) != EQUAL) {
1265 fprintf(stderr, " start %u, match %u, length %d\n",
1266 start, match, length);
1267 do {
1268 fprintf(stderr, "%c%c", s->window[match++], s->window[start++]);
1269 } while (--length != 0);
1270 z_error("invalid match");
1271 }
1272 if (z_verbose > 1) {
1273 fprintf(stderr,"\\[%d,%d]", start-match, length);
1274 do { putc(s->window[start++], stderr); } while (--length != 0);
1275 }
1276 }
1277 #else
1278 # define check_match(s, start, match, length)
1279 #endif /* DEBUG */
1280
1281 /* ===========================================================================
1282 * Fill the window when the lookahead becomes insufficient.
1283 * Updates strstart and lookahead.
1284 *
1285 * IN assertion: lookahead < MIN_LOOKAHEAD
1286 * OUT assertions: strstart <= window_size-MIN_LOOKAHEAD
1287 * At least one byte has been read, or avail_in == 0; reads are
1288 * performed for at least two bytes (required for the zip translate_eol
1289 * option -- not supported here).
1290 */
1291 local void
fill_window(deflate_state * s)1292 fill_window(deflate_state *s)
1293 {
1294 unsigned n, m;
1295 Posf *p;
1296 unsigned more; /* Amount of free space at the end of the window. */
1297 uInt wsize = s->w_size;
1298
1299 do {
1300 more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart);
1301
1302 /* Deal with !@#$% 64K limit: */
1303 if (sizeof(int) <= 2) {
1304 if (more == 0 && s->strstart == 0 && s->lookahead == 0) {
1305 more = wsize;
1306
1307 } else if (more == (unsigned)(-1)) {
1308 /* Very unlikely, but possible on 16 bit machine if
1309 * strstart == 0 && lookahead == 1 (input done a byte at time)
1310 */
1311 more--;
1312 }
1313 }
1314
1315 /* If the window is almost full and there is insufficient lookahead,
1316 * move the upper half to the lower one to make room in the upper half.
1317 */
1318 if (s->strstart >= wsize+MAX_DIST(s)) {
1319
1320 zmemcpy(s->window, s->window+wsize, (unsigned)wsize);
1321 s->match_start -= wsize;
1322 s->strstart -= wsize; /* we now have strstart >= MAX_DIST */
1323 s->block_start -= (long) wsize;
1324
1325 /* Slide the hash table (could be avoided with 32 bit values
1326 at the expense of memory usage). We slide even when level == 0
1327 to keep the hash table consistent if we switch back to level > 0
1328 later. (Using level 0 permanently is not an optimal usage of
1329 zlib, so we don't care about this pathological case.)
1330 */
1331 /* %%% avoid this when Z_RLE */
1332 n = s->hash_size;
1333 p = &s->head[n];
1334 do {
1335 m = *--p;
1336 *p = (Pos)(m >= wsize ? m-wsize : NIL);
1337 } while (--n);
1338
1339 n = wsize;
1340 #ifndef FASTEST
1341 p = &s->prev[n];
1342 do {
1343 m = *--p;
1344 *p = (Pos)(m >= wsize ? m-wsize : NIL);
1345 /* If n is not on any hash chain, prev[n] is garbage but
1346 * its value will never be used.
1347 */
1348 } while (--n);
1349 #endif
1350 more += wsize;
1351 }
1352 if (s->strm->avail_in == 0) return;
1353
1354 /* If there was no sliding:
1355 * strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 &&
1356 * more == window_size - lookahead - strstart
1357 * => more >= window_size - (MIN_LOOKAHEAD-1 + WSIZE + MAX_DIST-1)
1358 * => more >= window_size - 2*WSIZE + 2
1359 * In the BIG_MEM or MMAP case (not yet supported),
1360 * window_size == input_size + MIN_LOOKAHEAD &&
1361 * strstart + s->lookahead <= input_size => more >= MIN_LOOKAHEAD.
1362 * Otherwise, window_size == 2*WSIZE so more >= 2.
1363 * If there was sliding, more >= WSIZE. So in all cases, more >= 2.
1364 */
1365 Assert(more >= 2, "more < 2");
1366
1367 n = (*s->zinput)(s->strm, s->window + s->strstart + s->lookahead, more);
1368 s->lookahead += n;
1369
1370 /* Initialize the hash value now that we have some input: */
1371 if (s->lookahead >= MIN_MATCH) {
1372 s->ins_h = s->window[s->strstart];
1373 UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
1374 #if MIN_MATCH != 3
1375 Call UPDATE_HASH() MIN_MATCH-3 more times
1376 #endif
1377 }
1378 /* If the whole input has less than MIN_MATCH bytes, ins_h is garbage,
1379 * but this is not important since only literal bytes will be emitted.
1380 */
1381
1382 } while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0);
1383 }
1384
1385 /* ===========================================================================
1386 * Flush the current block, with given end-of-file flag.
1387 * IN assertion: strstart is set to the end of the current match.
1388 */
1389 #define FLUSH_BLOCK_ONLY(s, eof) { \
1390 _tr_flush_block(s, (s->block_start >= 0L ? \
1391 (charf *)&s->window[(unsigned)s->block_start] : \
1392 (charf *)Z_NULL), \
1393 (ulg)((long)s->strstart - s->block_start), \
1394 (eof)); \
1395 s->block_start = s->strstart; \
1396 flush_pending(s->strm); \
1397 Tracev((stderr,"[FLUSH]")); \
1398 }
1399
1400 /* Same but force premature exit if necessary. */
1401 #define FLUSH_BLOCK(s, eof) { \
1402 FLUSH_BLOCK_ONLY(s, eof); \
1403 if (s->strm->avail_out == 0) return (eof) ? finish_started : need_more; \
1404 }
1405
1406 /* ===========================================================================
1407 * Copy without compression as much as possible from the input stream, return
1408 * the current block state.
1409 * This function does not insert new strings in the dictionary since
1410 * uncompressible data is probably not useful. This function is used
1411 * only for the level=0 compression option.
1412 * NOTE: this function should be optimized to avoid extra copying from
1413 * window to pending_buf.
1414 */
1415 local block_state
deflate_stored(deflate_state * s,int flush)1416 deflate_stored(deflate_state *s, int flush)
1417 {
1418 /* Stored blocks are limited to 0xffff bytes, pending_buf is limited
1419 * to pending_buf_size, and each stored block has a 5 byte header:
1420 */
1421 ulg max_block_size = 0xffff;
1422 ulg max_start;
1423
1424 if (max_block_size > s->pending_buf_size - 5) {
1425 max_block_size = s->pending_buf_size - 5;
1426 }
1427
1428 /* Copy as much as possible from input to output: */
1429 for (;;) {
1430 /* Fill the window as much as possible: */
1431 if (s->lookahead <= 1) {
1432
1433 Assert(s->strstart < s->w_size+MAX_DIST(s) ||
1434 s->block_start >= (long)s->w_size, "slide too late");
1435
1436 fill_window(s);
1437 if (s->lookahead == 0 && flush == Z_NO_FLUSH) return need_more;
1438
1439 if (s->lookahead == 0) break; /* flush the current block */
1440 }
1441 Assert(s->block_start >= 0L, "block gone");
1442
1443 s->strstart += s->lookahead;
1444 s->lookahead = 0;
1445
1446 /* Emit a stored block if pending_buf will be full: */
1447 max_start = s->block_start + max_block_size;
1448 if (s->strstart == 0 || (ulg)s->strstart >= max_start) {
1449 /* strstart == 0 is possible when wraparound on 16-bit machine */
1450 s->lookahead = (uInt)(s->strstart - max_start);
1451 s->strstart = (uInt)max_start;
1452 FLUSH_BLOCK(s, 0);
1453 }
1454 /* Flush if we may have to slide, otherwise block_start may become
1455 * negative and the data will be gone:
1456 */
1457 if (s->strstart - (uInt)s->block_start >= MAX_DIST(s)) {
1458 FLUSH_BLOCK(s, 0);
1459 }
1460 }
1461 FLUSH_BLOCK(s, flush == Z_FINISH);
1462 return flush == Z_FINISH ? finish_done : block_done;
1463 }
1464
1465 /* ===========================================================================
1466 * Compress as much as possible from the input stream, return the current
1467 * block state.
1468 * This function does not perform lazy evaluation of matches and inserts
1469 * new strings in the dictionary only for unmatched strings or for short
1470 * matches. It is used only for the fast compression options.
1471 */
1472 local block_state
deflate_fast(deflate_state * s,int flush)1473 deflate_fast(deflate_state *s, int flush)
1474 {
1475 IPos hash_head = NIL; /* head of the hash chain */
1476 int bflush; /* set if current block must be flushed */
1477
1478 for (;;) {
1479 /* Make sure that we always have enough lookahead, except
1480 * at the end of the input file. We need MAX_MATCH bytes
1481 * for the next match, plus MIN_MATCH bytes to insert the
1482 * string following the next match.
1483 */
1484 if (s->lookahead < MIN_LOOKAHEAD) {
1485 fill_window(s);
1486 if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
1487 return need_more;
1488 }
1489 if (s->lookahead == 0) break; /* flush the current block */
1490 }
1491
1492 /* Insert the string window[strstart .. strstart+2] in the
1493 * dictionary, and set hash_head to the head of the hash chain:
1494 */
1495 if (s->lookahead >= MIN_MATCH) {
1496 INSERT_STRING(s, s->strstart, hash_head);
1497 }
1498
1499 /* Find the longest match, discarding those <= prev_length.
1500 * At this point we have always match_length < MIN_MATCH
1501 */
1502 if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) {
1503 /* To simplify the code, we prevent matches with the string
1504 * of window index 0 (in particular we have to avoid a match
1505 * of the string with itself at the start of the input file).
1506 */
1507 #ifdef FASTEST
1508 if ((s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) ||
1509 (s->strategy == Z_RLE && s->strstart - hash_head == 1)) {
1510 s->match_length = longest_match_fast (s, hash_head);
1511 }
1512 #else
1513 if (s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) {
1514 s->match_length = longest_match (s, hash_head);
1515 } else if (s->strategy == Z_RLE && s->strstart - hash_head == 1) {
1516 s->match_length = longest_match_fast (s, hash_head);
1517 }
1518 #endif
1519 /* longest_match() or longest_match_fast() sets match_start */
1520 }
1521 if (s->match_length >= MIN_MATCH) {
1522 check_match(s, s->strstart, s->match_start, s->match_length);
1523
1524 _tr_tally_dist(s, s->strstart - s->match_start,
1525 s->match_length - MIN_MATCH, bflush);
1526
1527 s->lookahead -= s->match_length;
1528
1529 /* Insert new strings in the hash table only if the match length
1530 * is not too large. This saves time but degrades compression.
1531 */
1532 #ifndef FASTEST
1533 if (s->match_length <= s->max_insert_length &&
1534 s->lookahead >= MIN_MATCH) {
1535 s->match_length--; /* string at strstart already in table */
1536 do {
1537 s->strstart++;
1538 INSERT_STRING(s, s->strstart, hash_head);
1539 /* strstart never exceeds WSIZE-MAX_MATCH, so there are
1540 * always MIN_MATCH bytes ahead.
1541 */
1542 } while (--s->match_length != 0);
1543 s->strstart++;
1544 } else
1545 #endif
1546 {
1547 s->strstart += s->match_length;
1548 s->match_length = 0;
1549 s->ins_h = s->window[s->strstart];
1550 UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
1551 #if MIN_MATCH != 3
1552 Call UPDATE_HASH() MIN_MATCH-3 more times
1553 #endif
1554 /* If lookahead < MIN_MATCH, ins_h is garbage, but it does not
1555 * matter since it will be recomputed at next deflate call.
1556 */
1557 }
1558 } else {
1559 /* No match, output a literal byte */
1560 Tracevv((stderr,"%c", s->window[s->strstart]));
1561 _tr_tally_lit (s, s->window[s->strstart], bflush);
1562 s->lookahead--;
1563 s->strstart++;
1564 }
1565 if (bflush) FLUSH_BLOCK(s, 0);
1566 }
1567 FLUSH_BLOCK(s, flush == Z_FINISH);
1568 return flush == Z_FINISH ? finish_done : block_done;
1569 }
1570
1571 #ifndef FASTEST
1572 /* ===========================================================================
1573 * Same as above, but achieves better compression. We use a lazy
1574 * evaluation for matches: a match is finally adopted only if there is
1575 * no better match at the next window position.
1576 */
1577 local block_state
deflate_slow(deflate_state * s,int flush)1578 deflate_slow(deflate_state *s, int flush)
1579 {
1580 IPos hash_head = NIL; /* head of hash chain */
1581 int bflush; /* set if current block must be flushed */
1582
1583 /* Process the input block. */
1584 for (;;) {
1585 /* Make sure that we always have enough lookahead, except
1586 * at the end of the input file. We need MAX_MATCH bytes
1587 * for the next match, plus MIN_MATCH bytes to insert the
1588 * string following the next match.
1589 */
1590 if (s->lookahead < MIN_LOOKAHEAD) {
1591 fill_window(s);
1592 if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
1593 return need_more;
1594 }
1595 if (s->lookahead == 0) break; /* flush the current block */
1596 }
1597
1598 /* Insert the string window[strstart .. strstart+2] in the
1599 * dictionary, and set hash_head to the head of the hash chain:
1600 */
1601 if (s->lookahead >= MIN_MATCH) {
1602 INSERT_STRING(s, s->strstart, hash_head);
1603 }
1604
1605 /* Find the longest match, discarding those <= prev_length.
1606 */
1607 s->prev_length = s->match_length;
1608 s->prev_match = s->match_start;
1609 s->match_length = MIN_MATCH-1;
1610
1611 if (hash_head != NIL && s->prev_length < s->max_lazy_match &&
1612 s->strstart - hash_head <= MAX_DIST(s)) {
1613 /* To simplify the code, we prevent matches with the string
1614 * of window index 0 (in particular we have to avoid a match
1615 * of the string with itself at the start of the input file).
1616 */
1617 if (s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) {
1618 s->match_length = longest_match (s, hash_head);
1619 } else if (s->strategy == Z_RLE && s->strstart - hash_head == 1) {
1620 s->match_length = longest_match_fast (s, hash_head);
1621 }
1622 /* longest_match() or longest_match_fast() sets match_start */
1623
1624 if (s->match_length <= 5 && (s->strategy == Z_FILTERED
1625 #if TOO_FAR <= 32767
1626 || (s->match_length == MIN_MATCH &&
1627 s->strstart - s->match_start > TOO_FAR)
1628 #endif
1629 )) {
1630
1631 /* If prev_match is also MIN_MATCH, match_start is garbage
1632 * but we will ignore the current match anyway.
1633 */
1634 s->match_length = MIN_MATCH-1;
1635 }
1636 }
1637 /* If there was a match at the previous step and the current
1638 * match is not better, output the previous match:
1639 */
1640 if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) {
1641 uInt max_insert = s->strstart + s->lookahead - MIN_MATCH;
1642 /* Do not insert strings in hash table beyond this. */
1643
1644 check_match(s, s->strstart-1, s->prev_match, s->prev_length);
1645
1646 _tr_tally_dist(s, s->strstart -1 - s->prev_match,
1647 s->prev_length - MIN_MATCH, bflush);
1648
1649 /* Insert in hash table all strings up to the end of the match.
1650 * strstart-1 and strstart are already inserted. If there is not
1651 * enough lookahead, the last two strings are not inserted in
1652 * the hash table.
1653 */
1654 s->lookahead -= s->prev_length-1;
1655 s->prev_length -= 2;
1656 do {
1657 if (++s->strstart <= max_insert) {
1658 INSERT_STRING(s, s->strstart, hash_head);
1659 }
1660 } while (--s->prev_length != 0);
1661 s->match_available = 0;
1662 s->match_length = MIN_MATCH-1;
1663 s->strstart++;
1664
1665 if (bflush) FLUSH_BLOCK(s, 0);
1666
1667 } else if (s->match_available) {
1668 /* If there was no match at the previous position, output a
1669 * single literal. If there was a match but the current match
1670 * is longer, truncate the previous match to a single literal.
1671 */
1672 Tracevv((stderr,"%c", s->window[s->strstart-1]));
1673 _tr_tally_lit(s, s->window[s->strstart-1], bflush);
1674 if (bflush) {
1675 FLUSH_BLOCK_ONLY(s, 0);
1676 }
1677 s->strstart++;
1678 s->lookahead--;
1679 if (s->strm->avail_out == 0) return need_more;
1680 } else {
1681 /* There is no previous match to compare with, wait for
1682 * the next step to decide.
1683 */
1684 s->match_available = 1;
1685 s->strstart++;
1686 s->lookahead--;
1687 }
1688 }
1689 Assert (flush != Z_NO_FLUSH, "no flush?");
1690 if (s->match_available) {
1691 Tracevv((stderr,"%c", s->window[s->strstart-1]));
1692 _tr_tally_lit(s, s->window[s->strstart-1], bflush);
1693 s->match_available = 0;
1694 }
1695 FLUSH_BLOCK(s, flush == Z_FINISH);
1696 return flush == Z_FINISH ? finish_done : block_done;
1697 }
1698 #endif /* FASTEST */
1699
1700 #if 0
1701 /* ===========================================================================
1702 * For Z_RLE, simply look for runs of bytes, generate matches only of distance
1703 * one. Do not maintain a hash table. (It will be regenerated if this run of
1704 * deflate switches away from Z_RLE.)
1705 */
1706 local block_state
1707 deflate_rle(deflate_state *s, int flush)
1708 {
1709 int bflush; /* set if current block must be flushed */
1710 uInt run; /* length of run */
1711 uInt max; /* maximum length of run */
1712 uInt prev; /* byte at distance one to match */
1713 Bytef *scan; /* scan for end of run */
1714
1715 for (;;) {
1716 /* Make sure that we always have enough lookahead, except
1717 * at the end of the input file. We need MAX_MATCH bytes
1718 * for the longest encodable run.
1719 */
1720 if (s->lookahead < MAX_MATCH) {
1721 fill_window(s);
1722 if (s->lookahead < MAX_MATCH && flush == Z_NO_FLUSH) {
1723 return need_more;
1724 }
1725 if (s->lookahead == 0) break; /* flush the current block */
1726 }
1727
1728 /* See how many times the previous byte repeats */
1729 run = 0;
1730 if (s->strstart > 0) { /* if there is a previous byte, that is */
1731 max = s->lookahead < MAX_MATCH ? s->lookahead : MAX_MATCH;
1732 scan = s->window + s->strstart - 1;
1733 prev = *scan++;
1734 do {
1735 if (*scan++ != prev)
1736 break;
1737 } while (++run < max);
1738 }
1739
1740 /* Emit match if have run of MIN_MATCH or longer, else emit literal */
1741 if (run >= MIN_MATCH) {
1742 check_match(s, s->strstart, s->strstart - 1, run);
1743 _tr_tally_dist(s, 1, run - MIN_MATCH, bflush);
1744 s->lookahead -= run;
1745 s->strstart += run;
1746 } else {
1747 /* No match, output a literal byte */
1748 Tracevv((stderr,"%c", s->window[s->strstart]));
1749 _tr_tally_lit (s, s->window[s->strstart], bflush);
1750 s->lookahead--;
1751 s->strstart++;
1752 }
1753 if (bflush) FLUSH_BLOCK(s, 0);
1754 }
1755 FLUSH_BLOCK(s, flush == Z_FINISH);
1756 return flush == Z_FINISH ? finish_done : block_done;
1757 }
1758 #endif
1759
1760 #if XNU_KERNEL_PRIVATE
1761
1762 uLong
zlib_deflate_memory_size(int wbits,int memlevel)1763 zlib_deflate_memory_size(int wbits, int memlevel)
1764 {
1765 return (31 + sizeof(deflate_state) + (1 << (wbits + 2)) + (1 << (memlevel + 9)));
1766 }
1767
1768 #endif /* XNU_KERNEL_PRIVATE */
1769