« back to all changes in this revision
Viewing changes to deflate.c
- browse files at revision 14
- compare with another revision
- download diff
- download tarball
- view history from revision 14
- Committer: Bazaar Package Importer
- Author(s): Matthias Klose
- Date: 2007-07-09 17:25:35 UTC
- mfrom: (1.2.1 upstream)
- Revision ID: james.westby@ubuntu.com-20070709172535-fywqps3hjk0bvi12
Tags: 1:1.2.3.3.dfsg-5ubuntu1
* Merge with Debian; remaining changes:
- Install multilib into /usr/lib32 on amd64.
- Install multilib into /usr/lib32 on amd64.
- files added:
- ChangeLog
- amiga
- as400
- contrib
- contrib/ada
- contrib/asm586
- contrib/asm686
- contrib/blast
- contrib/delphi
- contrib/dotzlib
- contrib/dotzlib/DotZLib
- contrib/infback9
- contrib/inflate86
- contrib/iostream
- contrib/iostream2
- contrib/iostream3
- contrib/masm686
- contrib/masmx64
- contrib/masmx86
- contrib/minizip
- contrib/pascal
- contrib/puff
- contrib/testzlib
- contrib/untgz
- contrib/vstudio
- contrib/vstudio/vc7
- contrib/vstudio/vc8
- doc
- examples
- msdos
- old
- old/os2
- projects
- projects/visualc6
- qnx
- watcom
- win32
- files removed:
- debian/README.Debian
- upstream
- upstream/tarballs
- files modified:
- debian/changelog
added
removed
deflate.c
1
/* deflate.c -- compress data using the deflation algorithm
2
* Copyright (C) 1995-2006 Jean-loup Gailly.
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://www.ietf.org/rfc/rfc1951.txt
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 1.2.3.3 Copyright 1995-2006 Jean-loup Gailly ";
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 void fill_window OF((deflate_state *s));
77
local block_state deflate_stored OF((deflate_state *s, int flush));
78
local block_state deflate_fast OF((deflate_state *s, int flush));
79
#ifndef FASTEST
80
local block_state deflate_slow OF((deflate_state *s, int flush));
81
#endif
82
local void lm_init OF((deflate_state *s));
83
local void putShortMSB OF((deflate_state *s, uInt b));
84
local void flush_pending OF((z_streamp strm));
85
local int read_buf OF((z_streamp strm, Bytef *buf, unsigned size));
86
#ifndef FASTEST
87
#ifdef ASMV
88
void match_init OF((void)); /* asm code initialization */
89
uInt longest_match OF((deflate_state *s, IPos cur_match));
90
#else
91
local uInt longest_match OF((deflate_state *s, IPos cur_match));
92
#endif
93
#endif
94
local uInt longest_match_fast OF((deflate_state *s, IPos cur_match));
95
96
#ifdef DEBUG
97
local void check_match OF((deflate_state *s, IPos start, IPos match,
98
int length));
99
#endif
100
101
/* ===========================================================================
102
* Local data
103
*/
104
105
#define NIL 0
106
/* Tail of hash chains */
107
108
#ifndef TOO_FAR
109
# define TOO_FAR 4096
110
#endif
111
/* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
112
113
#define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
114
/* Minimum amount of lookahead, except at the end of the input file.
115
* See deflate.c for comments about the MIN_MATCH+1.
116
*/
117
118
/* Values for max_lazy_match, good_match and max_chain_length, depending on
119
* the desired pack level (0..9). The values given below have been tuned to
120
* exclude worst case performance for pathological files. Better values may be
121
* found for specific files.
122
*/
123
typedef struct config_s {
124
ush good_length; /* reduce lazy search above this match length */
125
ush max_lazy; /* do not perform lazy search above this match length */
126
ush nice_length; /* quit search above this match length */
127
ush max_chain;
128
compress_func func;
129
} config;
130
131
#ifdef FASTEST
132
local const config configuration_table[2] = {
133
/* good lazy nice chain */
134
/* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */
135
/* 1 */ {4, 4, 8, 4, deflate_fast}}; /* max speed, no lazy matches */
136
#else
137
local const config configuration_table[10] = {
138
/* good lazy nice chain */
139
/* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */
140
/* 1 */ {4, 4, 8, 4, deflate_fast}, /* max speed, no lazy matches */
141
/* 2 */ {4, 5, 16, 8, deflate_fast},
142
/* 3 */ {4, 6, 32, 32, deflate_fast},
143
144
/* 4 */ {4, 4, 16, 16, deflate_slow}, /* lazy matches */
145
/* 5 */ {8, 16, 32, 32, deflate_slow},
146
/* 6 */ {8, 16, 128, 128, deflate_slow},
147
/* 7 */ {8, 32, 128, 256, deflate_slow},
148
/* 8 */ {32, 128, 258, 1024, deflate_slow},
149
/* 9 */ {32, 258, 258, 4096, deflate_slow}}; /* max compression */
150
#endif
151
152
/* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
153
* For deflate_fast() (levels <= 3) good is ignored and lazy has a different
154
* meaning.
155
*/
156
157
#define EQUAL 0
158
/* result of memcmp for equal strings */
159
160
#ifndef NO_DUMMY_DECL
161
struct static_tree_desc_s {int dummy;}; /* for buggy compilers */
162
#endif
163
164
/* ===========================================================================
165
* Update a hash value with the given input byte
166
* IN assertion: all calls to to UPDATE_HASH are made with consecutive
167
* input characters, so that a running hash key can be computed from the
168
* previous key instead of complete recalculation each time.
169
*/
170
#define UPDATE_HASH(s,h,c) (h = (((h)<<s->hash_shift) ^ (c)) & s->hash_mask)
171
172
173
/* ===========================================================================
174
* Insert string str in the dictionary and set match_head to the previous head
175
* of the hash chain (the most recent string with same hash key). Return
176
* the previous length of the hash chain.
177
* If this file is compiled with -DFASTEST, the compression level is forced
178
* to 1, and no hash chains are maintained.
179
* IN assertion: all calls to to INSERT_STRING are made with consecutive
180
* input characters and the first MIN_MATCH bytes of str are valid
181
* (except for the last MIN_MATCH-1 bytes of the input file).
182
*/
183
#ifdef FASTEST
184
#define INSERT_STRING(s, str, match_head) \
185
(UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
186
match_head = s->head[s->ins_h], \
187
s->head[s->ins_h] = (Pos)(str))
188
#else
189
#define INSERT_STRING(s, str, match_head) \
190
(UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
191
match_head = s->prev[(str) & s->w_mask] = s->head[s->ins_h], \
192
s->head[s->ins_h] = (Pos)(str))
193
#endif
194
195
/* ===========================================================================
196
* Initialize the hash table (avoiding 64K overflow for 16 bit systems).
197
* prev[] will be initialized on the fly.
198
*/
199
#define CLEAR_HASH(s) \
200
s->head[s->hash_size-1] = NIL; \
201
zmemzero((Bytef *)s->head, (unsigned)(s->hash_size-1)*sizeof(*s->head));
202
203
/* ========================================================================= */
204
int ZEXPORT deflateInit_(strm, level, version, stream_size)
205
z_streamp strm;
206
int level;
207
const char *version;
208
int stream_size;
209
{
210
return deflateInit2_(strm, level, Z_DEFLATED, MAX_WBITS, DEF_MEM_LEVEL,
211
Z_DEFAULT_STRATEGY, version, stream_size);
212
/* To do: ignore strm->next_in if we use it as window */
213
}
214
215
/* ========================================================================= */
216
int ZEXPORT deflateInit2_(strm, level, method, windowBits, memLevel, strategy,
217
version, stream_size)
218
z_streamp strm;
219
int level;
220
int method;
221
int windowBits;
222
int memLevel;
223
int strategy;
224
const char *version;
225
int stream_size;
226
{
227
deflate_state *s;
228
int wrap = 1;
229
static const char my_version[] = ZLIB_VERSION;
230
231
ushf *overlay;
232
/* We overlay pending_buf and d_buf+l_buf. This works since the average
233
* output size for (length,distance) codes is <= 24 bits.
234
*/
235
236
if (version == Z_NULL || version[0] != my_version[0] ||
237
stream_size != sizeof(z_stream)) {
238
return Z_VERSION_ERROR;
239
}
240
if (strm == Z_NULL) return Z_STREAM_ERROR;
241
242
strm->msg = Z_NULL;
243
if (strm->zalloc == (alloc_func)0) {
244
strm->zalloc = zcalloc;
245
strm->opaque = (voidpf)0;
246
}
247
if (strm->zfree == (free_func)0) strm->zfree = zcfree;
248
249
#ifdef FASTEST
250
if (level != 0) level = 1;
251
#else
252
if (level == Z_DEFAULT_COMPRESSION) level = 6;
253
#endif
254
255
if (windowBits < 0) { /* suppress zlib wrapper */
256
wrap = 0;
257
windowBits = -windowBits;
258
}
259
#ifdef GZIP
260
else if (windowBits > 15) {
261
wrap = 2; /* write gzip wrapper instead */
262
windowBits -= 16;
263
}
264
#endif
265
if (memLevel < 1 || memLevel > MAX_MEM_LEVEL || method != Z_DEFLATED ||
266
windowBits < 8 || windowBits > 15 || level < 0 || level > 9 ||
267
strategy < 0 || strategy > Z_FIXED) {
268
return Z_STREAM_ERROR;
269
}
270
if (windowBits == 8) windowBits = 9; /* until 256-byte window bug fixed */
271
s = (deflate_state *) ZALLOC(strm, 1, sizeof(deflate_state));
272
if (s == Z_NULL) return Z_MEM_ERROR;
273
strm->state = (struct internal_state FAR *)s;
274
s->strm = strm;
275
276
s->wrap = wrap;
277
s->gzhead = Z_NULL;
278
s->w_bits = windowBits;
279
s->w_size = 1 << s->w_bits;
280
s->w_mask = s->w_size - 1;
281
282
s->hash_bits = memLevel + 7;
283
s->hash_size = 1 << s->hash_bits;
284
s->hash_mask = s->hash_size - 1;
285
s->hash_shift = ((s->hash_bits+MIN_MATCH-1)/MIN_MATCH);
286
287
s->window = (Bytef *) ZALLOC(strm, s->w_size, 2*sizeof(Byte));
288
s->prev = (Posf *) ZALLOC(strm, s->w_size, sizeof(Pos));
289
s->head = (Posf *) ZALLOC(strm, s->hash_size, sizeof(Pos));
290
291
s->lit_bufsize = 1 << (memLevel + 6); /* 16K elements by default */
292
293
overlay = (ushf *) ZALLOC(strm, s->lit_bufsize, sizeof(ush)+2);
294
s->pending_buf = (uchf *) overlay;
295
s->pending_buf_size = (ulg)s->lit_bufsize * (sizeof(ush)+2L);
296
297
if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL ||
298
s->pending_buf == Z_NULL) {
299
s->status = FINISH_STATE;
300
strm->msg = (char*)ERR_MSG(Z_MEM_ERROR);
301
deflateEnd (strm);
302
return Z_MEM_ERROR;
303
}
304
s->d_buf = overlay + s->lit_bufsize/sizeof(ush);
305
s->l_buf = s->pending_buf + (1+sizeof(ush))*s->lit_bufsize;
306
307
s->level = level;
308
s->strategy = strategy;
309
s->method = (Byte)method;
310
311
return deflateReset(strm);
312
}
313
314
/* ========================================================================= */
315
int ZEXPORT deflateSetDictionary (strm, dictionary, dictLength)
316
z_streamp strm;
317
const Bytef *dictionary;
318
uInt dictLength;
319
{
320
deflate_state *s;
321
uInt length = dictLength;
322
uInt n;
323
IPos hash_head = 0;
324
325
if (strm == Z_NULL || strm->state == Z_NULL || dictionary == Z_NULL ||
326
strm->state->wrap == 2 ||
327
(strm->state->wrap == 1 && strm->state->status != INIT_STATE))
328
return Z_STREAM_ERROR;
329
330
s = strm->state;
331
if (s->wrap)
332
strm->adler = adler32(strm->adler, dictionary, dictLength);
333
334
if (length < MIN_MATCH) return Z_OK;
335
if (length > MAX_DIST(s)) {
336
length = MAX_DIST(s);
337
dictionary += dictLength - length; /* use the tail of the dictionary */
338
}
339
zmemcpy(s->window, dictionary, length);
340
s->strstart = length;
341
s->block_start = (long)length;
342
343
/* Insert all strings in the hash table (except for the last two bytes).
344
* s->lookahead stays null, so s->ins_h will be recomputed at the next
345
* call of fill_window.
346
*/
347
s->ins_h = s->window[0];
348
UPDATE_HASH(s, s->ins_h, s->window[1]);
349
for (n = 0; n <= length - MIN_MATCH; n++) {
350
INSERT_STRING(s, n, hash_head);
351
}
352
if (hash_head) hash_head = 0; /* to make compiler happy */
353
return Z_OK;
354
}
355
356
/* ========================================================================= */
357
int ZEXPORT deflateReset (strm)
358
z_streamp strm;
359
{
360
deflate_state *s;
361
362
if (strm == Z_NULL || strm->state == Z_NULL ||
363
strm->zalloc == (alloc_func)0 || strm->zfree == (free_func)0) {
364
return Z_STREAM_ERROR;
365
}
366
367
strm->total_in = strm->total_out = 0;
368
strm->msg = Z_NULL; /* use zfree if we ever allocate msg dynamically */
369
strm->data_type = Z_UNKNOWN;
370
371
s = (deflate_state *)strm->state;
372
s->pending = 0;
373
s->pending_out = s->pending_buf;
374
375
if (s->wrap < 0) {
376
s->wrap = -s->wrap; /* was made negative by deflate(..., Z_FINISH); */
377
}
378
s->status = s->wrap ? INIT_STATE : BUSY_STATE;
379
strm->adler =
380
#ifdef GZIP
381
s->wrap == 2 ? crc32(0L, Z_NULL, 0) :
382
#endif
383
adler32(0L, Z_NULL, 0);
384
s->last_flush = Z_NO_FLUSH;
385
386
_tr_init(s);
387
lm_init(s);
388
389
return Z_OK;
390
}
391
392
/* ========================================================================= */
393
int ZEXPORT deflateSetHeader (strm, head)
394
z_streamp strm;
395
gz_headerp head;
396
{
397
if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
398
if (strm->state->wrap != 2) return Z_STREAM_ERROR;
399
strm->state->gzhead = head;
400
return Z_OK;
401
}
402
403
/* ========================================================================= */
404
int ZEXPORT deflatePrime (strm, bits, value)
405
z_streamp strm;
406
int bits;
407
int value;
408
{
409
if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
410
strm->state->bi_valid = bits;
411
strm->state->bi_buf = (ush)(value & ((1 << bits) - 1));
412
return Z_OK;
413
}
414
415
/* ========================================================================= */
416
int ZEXPORT deflateParams(strm, level, strategy)
417
z_streamp strm;
418
int level;
419
int strategy;
420
{
421
deflate_state *s;
422
compress_func func;
423
int err = Z_OK;
424
425
if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
426
s = strm->state;
427
428
#ifdef FASTEST
429
if (level != 0) level = 1;
430
#else
431
if (level == Z_DEFAULT_COMPRESSION) level = 6;
432
#endif
433
if (level < 0 || level > 9 || strategy < 0 || strategy > Z_FIXED) {
434
return Z_STREAM_ERROR;
435
}
436
func = configuration_table[s->level].func;
437
438
if (func != configuration_table[level].func && strm->total_in != 0) {
439
/* Flush the last buffer: */
440
err = deflate(strm, Z_PARTIAL_FLUSH);
441
}
442
if (s->level != level) {
443
s->level = level;
444
s->max_lazy_match = configuration_table[level].max_lazy;
445
s->good_match = configuration_table[level].good_length;
446
s->nice_match = configuration_table[level].nice_length;
447
s->max_chain_length = configuration_table[level].max_chain;
448
}
449
s->strategy = strategy;
450
return err;
451
}
452
453
/* ========================================================================= */
454
int ZEXPORT deflateTune(strm, good_length, max_lazy, nice_length, max_chain)
455
z_streamp strm;
456
int good_length;
457
int max_lazy;
458
int nice_length;
459
int max_chain;
460
{
461
deflate_state *s;
462
463
if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
464
s = strm->state;
465
s->good_match = good_length;
466
s->max_lazy_match = max_lazy;
467
s->nice_match = nice_length;
468
s->max_chain_length = max_chain;
469
return Z_OK;
470
}
471
472
/* =========================================================================
473
* For the default windowBits of 15 and memLevel of 8, this function returns
474
* a close to exact, as well as small, upper bound on the compressed size.
475
* They are coded as constants here for a reason--if the #define's are
476
* changed, then this function needs to be changed as well. The return
477
* value for 15 and 8 only works for those exact settings.
478
*
479
* For any setting other than those defaults for windowBits and memLevel,
480
* the value returned is a conservative worst case for the maximum expansion
481
* resulting from using fixed blocks instead of stored blocks, which deflate
482
* can emit on compressed data for some combinations of the parameters.
483
*
484
* This function could be more sophisticated to provide closer upper bounds for
485
* every combination of windowBits and memLevel. But even the conservative
486
* upper bound of about 14% expansion does not seem onerous for output buffer
487
* allocation.
488
*/
489
uLong ZEXPORT deflateBound(strm, sourceLen)
490
z_streamp strm;
491
uLong sourceLen;
492
{
493
deflate_state *s;
494
uLong complen, wraplen;
495
Bytef *str;
496
497
/* conservative upper bound for compressed data */
498
complen = sourceLen +
499
((sourceLen + 7) >> 3) + ((sourceLen + 63) >> 6) + 5;
500
501
/* if can't get parameters, return conservative bound plus zlib wrapper */
502
if (strm == Z_NULL || strm->state == Z_NULL)
503
return complen + 6;
504
505
/* compute wrapper length */
506
s = strm->state;
507
switch (s->wrap) {
508
case 0: /* raw deflate */
509
wraplen = 0;
510
break;
511
case 1: /* zlib wrapper */
512
wraplen = 6 + (s->strstart ? 4 : 0);
513
break;
514
case 2: /* gzip wrapper */
515
wraplen = 18;
516
if (s->gzhead != NULL) { /* user-supplied gzip header */
517
if (s->gzhead->extra != NULL)
518
wraplen += 2 + s->gzhead->extra_len;
519
str = s->gzhead->name;
520
if (str != NULL)
521
do {
522
wraplen++;
523
} while (*str++);
524
str = s->gzhead->comment;
525
if (str != NULL)
526
do {
527
wraplen++;
528
} while (*str++);
529
if (s->gzhead->hcrc)
530
wraplen += 2;
531
}
532
break;
533
default: /* for compiler happiness */
534
wraplen = 6;
535
}
536
537
/* if not default parameters, return conservative bound */
538
if (s->w_bits != 15 || s->hash_bits != 8 + 7)
539
return complen + wraplen;
540
541
/* default settings: return tight bound for that case */
542
return compressBound(sourceLen) - 6 + wraplen;
543
}
544
545
/* =========================================================================
546
* Put a short in the pending buffer. The 16-bit value is put in MSB order.
547
* IN assertion: the stream state is correct and there is enough room in
548
* pending_buf.
549
*/
550
local void putShortMSB (s, b)
551
deflate_state *s;
552
uInt b;
553
{
554
put_byte(s, (Byte)(b >> 8));
555
put_byte(s, (Byte)(b & 0xff));
556
}
557
558
/* =========================================================================
559
* Flush as much pending output as possible. All deflate() output goes
560
* through this function so some applications may wish to modify it
561
* to avoid allocating a large strm->next_out buffer and copying into it.
562
* (See also read_buf()).
563
*/
564
local void flush_pending(strm)
565
z_streamp strm;
566
{
567
unsigned len = strm->state->pending;
568
569
if (len > strm->avail_out) len = strm->avail_out;
570
if (len == 0) return;
571
572
zmemcpy(strm->next_out, strm->state->pending_out, len);
573
strm->next_out += len;
574
strm->state->pending_out += len;
575
strm->total_out += len;
576
strm->avail_out -= len;
577
strm->state->pending -= len;
578
if (strm->state->pending == 0) {
579
strm->state->pending_out = strm->state->pending_buf;
580
}
581
}
582
583
/* ========================================================================= */
584
int ZEXPORT deflate (strm, flush)
585
z_streamp strm;
586
int flush;
587
{
588
int old_flush; /* value of flush param for previous deflate call */
589
deflate_state *s;
590
591
if (strm == Z_NULL || strm->state == Z_NULL ||
592
flush > Z_FINISH || flush < 0) {
593
return Z_STREAM_ERROR;
594
}
595
s = strm->state;
596
597
if (strm->next_out == Z_NULL ||
598
(strm->next_in == Z_NULL && strm->avail_in != 0) ||
599
(s->status == FINISH_STATE && flush != Z_FINISH)) {
600
ERR_RETURN(strm, Z_STREAM_ERROR);
601
}
602
if (strm->avail_out == 0) ERR_RETURN(strm, Z_BUF_ERROR);
603
604
s->strm = strm; /* just in case */
605
old_flush = s->last_flush;
606
s->last_flush = flush;
607
608
/* Write the header */
609
if (s->status == INIT_STATE) {
610
#ifdef GZIP
611
if (s->wrap == 2) {
612
strm->adler = crc32(0L, Z_NULL, 0);
613
put_byte(s, 31);
614
put_byte(s, 139);
615
put_byte(s, 8);
616
if (s->gzhead == NULL) {
617
put_byte(s, 0);
618
put_byte(s, 0);
619
put_byte(s, 0);
620
put_byte(s, 0);
621
put_byte(s, 0);
622
put_byte(s, s->level == 9 ? 2 :
623
(s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
624
4 : 0));
625
put_byte(s, OS_CODE);
626
s->status = BUSY_STATE;
627
}
628
else {
629
put_byte(s, (s->gzhead->text ? 1 : 0) +
630
(s->gzhead->hcrc ? 2 : 0) +
631
(s->gzhead->extra == Z_NULL ? 0 : 4) +
632
(s->gzhead->name == Z_NULL ? 0 : 8) +
633
(s->gzhead->comment == Z_NULL ? 0 : 16)
634
);
635
put_byte(s, (Byte)(s->gzhead->time & 0xff));
636
put_byte(s, (Byte)((s->gzhead->time >> 8) & 0xff));
637
put_byte(s, (Byte)((s->gzhead->time >> 16) & 0xff));
638
put_byte(s, (Byte)((s->gzhead->time >> 24) & 0xff));
639
put_byte(s, s->level == 9 ? 2 :
640
(s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
641
4 : 0));
642
put_byte(s, s->gzhead->os & 0xff);
643
if (s->gzhead->extra != NULL) {
644
put_byte(s, s->gzhead->extra_len & 0xff);
645
put_byte(s, (s->gzhead->extra_len >> 8) & 0xff);
646
}
647
if (s->gzhead->hcrc)
648
strm->adler = crc32(strm->adler, s->pending_buf,
649
s->pending);
650
s->gzindex = 0;
651
s->status = EXTRA_STATE;
652
}
653
}
654
else
655
#endif
656
{
657
uInt header = (Z_DEFLATED + ((s->w_bits-8)<<4)) << 8;
658
uInt level_flags;
659
660
if (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2)
661
level_flags = 0;
662
else if (s->level < 6)
663
level_flags = 1;
664
else if (s->level == 6)
665
level_flags = 2;
666
else
667
level_flags = 3;
668
header |= (level_flags << 6);
669
if (s->strstart != 0) header |= PRESET_DICT;
670
header += 31 - (header % 31);
671
672
s->status = BUSY_STATE;
673
putShortMSB(s, header);
674
675
/* Save the adler32 of the preset dictionary: */
676
if (s->strstart != 0) {
677
putShortMSB(s, (uInt)(strm->adler >> 16));
678
putShortMSB(s, (uInt)(strm->adler & 0xffff));
679
}
680
strm->adler = adler32(0L, Z_NULL, 0);
681
}
682
}
683
#ifdef GZIP
684
if (s->status == EXTRA_STATE) {
685
if (s->gzhead->extra != NULL) {
686
uInt beg = s->pending; /* start of bytes to update crc */
687
688
while (s->gzindex < (s->gzhead->extra_len & 0xffff)) {
689
if (s->pending == s->pending_buf_size) {
690
if (s->gzhead->hcrc && s->pending > beg)
691
strm->adler = crc32(strm->adler, s->pending_buf + beg,
692
s->pending - beg);
693
flush_pending(strm);
694
beg = s->pending;
695
if (s->pending == s->pending_buf_size)
696
break;
697
}
698
put_byte(s, s->gzhead->extra[s->gzindex]);
699
s->gzindex++;
700
}
701
if (s->gzhead->hcrc && s->pending > beg)
702
strm->adler = crc32(strm->adler, s->pending_buf + beg,
703
s->pending - beg);
704
if (s->gzindex == s->gzhead->extra_len) {
705
s->gzindex = 0;
706
s->status = NAME_STATE;
707
}
708
}
709
else
710
s->status = NAME_STATE;
711
}
712
if (s->status == NAME_STATE) {
713
if (s->gzhead->name != NULL) {
714
uInt beg = s->pending; /* start of bytes to update crc */
715
int val;
716
717
do {
718
if (s->pending == s->pending_buf_size) {
719
if (s->gzhead->hcrc && s->pending > beg)
720
strm->adler = crc32(strm->adler, s->pending_buf + beg,
721
s->pending - beg);
722
flush_pending(strm);
723
beg = s->pending;
724
if (s->pending == s->pending_buf_size) {
725
val = 1;
726
break;
727
}
728
}
729
val = s->gzhead->name[s->gzindex++];
730
put_byte(s, val);
731
} while (val != 0);
732
if (s->gzhead->hcrc && s->pending > beg)
733
strm->adler = crc32(strm->adler, s->pending_buf + beg,
734
s->pending - beg);
735
if (val == 0) {
736
s->gzindex = 0;
737
s->status = COMMENT_STATE;
738
}
739
}
740
else
741
s->status = COMMENT_STATE;
742
}
743
if (s->status == COMMENT_STATE) {
744
if (s->gzhead->comment != NULL) {
745
uInt beg = s->pending; /* start of bytes to update crc */
746
int val;
747
748
do {
749
if (s->pending == s->pending_buf_size) {
750
if (s->gzhead->hcrc && s->pending > beg)
751
strm->adler = crc32(strm->adler, s->pending_buf + beg,
752
s->pending - beg);
753
flush_pending(strm);
754
beg = s->pending;
755
if (s->pending == s->pending_buf_size) {
756
val = 1;
757
break;
758
}
759
}
760
val = s->gzhead->comment[s->gzindex++];
761
put_byte(s, val);
762
} while (val != 0);
763
if (s->gzhead->hcrc && s->pending > beg)
764
strm->adler = crc32(strm->adler, s->pending_buf + beg,
765
s->pending - beg);
766
if (val == 0)
767
s->status = HCRC_STATE;
768
}
769
else
770
s->status = HCRC_STATE;
771
}
772
if (s->status == HCRC_STATE) {
773
if (s->gzhead->hcrc) {
774
if (s->pending + 2 > s->pending_buf_size)
775
flush_pending(strm);
776
if (s->pending + 2 <= s->pending_buf_size) {
777
put_byte(s, (Byte)(strm->adler & 0xff));
778
put_byte(s, (Byte)((strm->adler >> 8) & 0xff));
779
strm->adler = crc32(0L, Z_NULL, 0);
780
s->status = BUSY_STATE;
781
}
782
}
783
else
784
s->status = BUSY_STATE;
785
}
786
#endif
787
788
/* Flush as much pending output as possible */
789
if (s->pending != 0) {
790
flush_pending(strm);
791
if (strm->avail_out == 0) {
792
/* Since avail_out is 0, deflate will be called again with
793
* more output space, but possibly with both pending and
794
* avail_in equal to zero. There won't be anything to do,
795
* but this is not an error situation so make sure we
796
* return OK instead of BUF_ERROR at next call of deflate:
797
*/
798
s->last_flush = -1;
799
return Z_OK;
800
}
801
802
/* Make sure there is something to do and avoid duplicate consecutive
803
* flushes. For repeated and useless calls with Z_FINISH, we keep
804
* returning Z_STREAM_END instead of Z_BUF_ERROR.
805
*/
806
} else if (strm->avail_in == 0 && flush <= old_flush &&
807
flush != Z_FINISH) {
808
ERR_RETURN(strm, Z_BUF_ERROR);
809
}
810
811
/* User must not provide more input after the first FINISH: */
812
if (s->status == FINISH_STATE && strm->avail_in != 0) {
813
ERR_RETURN(strm, Z_BUF_ERROR);
814
}
815
816
/* Start a new block or continue the current one.
817
*/
818
if (strm->avail_in != 0 || s->lookahead != 0 ||
819
(flush != Z_NO_FLUSH && s->status != FINISH_STATE)) {
820
block_state bstate;
821
822
bstate = (*(configuration_table[s->level].func))(s, flush);
823
824
if (bstate == finish_started || bstate == finish_done) {
825
s->status = FINISH_STATE;
826
}
827
if (bstate == need_more || bstate == finish_started) {
828
if (strm->avail_out == 0) {
829
s->last_flush = -1; /* avoid BUF_ERROR next call, see above */
830
}
831
return Z_OK;
832
/* If flush != Z_NO_FLUSH && avail_out == 0, the next call
833
* of deflate should use the same flush parameter to make sure
834
* that the flush is complete. So we don't have to output an
835
* empty block here, this will be done at next call. This also
836
* ensures that for a very small output buffer, we emit at most
837
* one empty block.
838
*/
839
}
840
if (bstate == block_done) {
841
if (flush == Z_PARTIAL_FLUSH) {
842
_tr_align(s);
843
} else { /* FULL_FLUSH or SYNC_FLUSH */
844
_tr_stored_block(s, (char*)0, 0L, 0);
845
/* For a full flush, this empty block will be recognized
846
* as a special marker by inflate_sync().
847
*/
848
if (flush == Z_FULL_FLUSH) {
849
CLEAR_HASH(s); /* forget history */
850
}
851
}
852
flush_pending(strm);
853
if (strm->avail_out == 0) {
854
s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */
855
return Z_OK;
856
}
857
}
858
}
859
Assert(strm->avail_out > 0, "bug2");
860
861
if (flush != Z_FINISH) return Z_OK;
862
if (s->wrap <= 0) return Z_STREAM_END;
863
864
/* Write the trailer */
865
#ifdef GZIP
866
if (s->wrap == 2) {
867
put_byte(s, (Byte)(strm->adler & 0xff));
868
put_byte(s, (Byte)((strm->adler >> 8) & 0xff));
869
put_byte(s, (Byte)((strm->adler >> 16) & 0xff));
870
put_byte(s, (Byte)((strm->adler >> 24) & 0xff));
871
put_byte(s, (Byte)(strm->total_in & 0xff));
872
put_byte(s, (Byte)((strm->total_in >> 8) & 0xff));
873
put_byte(s, (Byte)((strm->total_in >> 16) & 0xff));
874
put_byte(s, (Byte)((strm->total_in >> 24) & 0xff));
875
}
876
else
877
#endif
878
{
879
putShortMSB(s, (uInt)(strm->adler >> 16));
880
putShortMSB(s, (uInt)(strm->adler & 0xffff));
881
}
882
flush_pending(strm);
883
/* If avail_out is zero, the application will call deflate again
884
* to flush the rest.
885
*/
886
if (s->wrap > 0) s->wrap = -s->wrap; /* write the trailer only once! */
887
return s->pending != 0 ? Z_OK : Z_STREAM_END;
888
}
889
890
/* ========================================================================= */
891
int ZEXPORT deflateEnd (strm)
892
z_streamp strm;
893
{
894
int status;
895
896
if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
897
898
status = strm->state->status;
899
if (status != INIT_STATE &&
900
status != EXTRA_STATE &&
901
status != NAME_STATE &&
902
status != COMMENT_STATE &&
903
status != HCRC_STATE &&
904
status != BUSY_STATE &&
905
status != FINISH_STATE) {
906
return Z_STREAM_ERROR;
907
}
908
909
/* Deallocate in reverse order of allocations: */
910
TRY_FREE(strm, strm->state->pending_buf);
911
TRY_FREE(strm, strm->state->head);
912
TRY_FREE(strm, strm->state->prev);
913
TRY_FREE(strm, strm->state->window);
914
915
ZFREE(strm, strm->state);
916
strm->state = Z_NULL;
917
918
return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK;
919
}
920
921
/* =========================================================================
922
* Copy the source state to the destination state.
923
* To simplify the source, this is not supported for 16-bit MSDOS (which
924
* doesn't have enough memory anyway to duplicate compression states).
925
*/
926
int ZEXPORT deflateCopy (dest, source)
927
z_streamp dest;
928
z_streamp source;
929
{
930
#ifdef MAXSEG_64K
931
return Z_STREAM_ERROR;
932
#else
933
deflate_state *ds;
934
deflate_state *ss;
935
ushf *overlay;
936
937
938
if (source == Z_NULL || dest == Z_NULL || source->state == Z_NULL) {
939
return Z_STREAM_ERROR;
940
}
941
942
ss = source->state;
943
944
zmemcpy(dest, source, sizeof(z_stream));
945
946
ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state));
947
if (ds == Z_NULL) return Z_MEM_ERROR;
948
dest->state = (struct internal_state FAR *) ds;
949
zmemcpy(ds, ss, sizeof(deflate_state));
950
ds->strm = dest;
951
952
ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte));
953
ds->prev = (Posf *) ZALLOC(dest, ds->w_size, sizeof(Pos));
954
ds->head = (Posf *) ZALLOC(dest, ds->hash_size, sizeof(Pos));
955
overlay = (ushf *) ZALLOC(dest, ds->lit_bufsize, sizeof(ush)+2);
956
ds->pending_buf = (uchf *) overlay;
957
958
if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL ||
959
ds->pending_buf == Z_NULL) {
960
deflateEnd (dest);
961
return Z_MEM_ERROR;
962
}
963
/* following zmemcpy do not work for 16-bit MSDOS */
964
zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte));
965
zmemcpy(ds->prev, ss->prev, ds->w_size * sizeof(Pos));
966
zmemcpy(ds->head, ss->head, ds->hash_size * sizeof(Pos));
967
zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size);
968
969
ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf);
970
ds->d_buf = overlay + ds->lit_bufsize/sizeof(ush);
971
ds->l_buf = ds->pending_buf + (1+sizeof(ush))*ds->lit_bufsize;
972
973
ds->l_desc.dyn_tree = ds->dyn_ltree;
974
ds->d_desc.dyn_tree = ds->dyn_dtree;
975
ds->bl_desc.dyn_tree = ds->bl_tree;
976
977
return Z_OK;
978
#endif /* MAXSEG_64K */
979
}
980
981
/* ===========================================================================
982
* Read a new buffer from the current input stream, update the adler32
983
* and total number of bytes read. All deflate() input goes through
984
* this function so some applications may wish to modify it to avoid
985
* allocating a large strm->next_in buffer and copying from it.
986
* (See also flush_pending()).
987
*/
988
local int read_buf(strm, buf, size)
989
z_streamp strm;
990
Bytef *buf;
991
unsigned size;
992
{
993
unsigned len = strm->avail_in;
994
995
if (len > size) len = size;
996
if (len == 0) return 0;
997
998
strm->avail_in -= len;
999
1000
if (strm->state->wrap == 1) {
1001
strm->adler = adler32(strm->adler, strm->next_in, len);
1002
}
1003
#ifdef GZIP
1004
else if (strm->state->wrap == 2) {
1005
strm->adler = crc32(strm->adler, strm->next_in, len);
1006
}
1007
#endif
1008
zmemcpy(buf, strm->next_in, len);
1009
strm->next_in += len;
1010
strm->total_in += len;
1011
1012
return (int)len;
1013
}
1014
1015
/* ===========================================================================
1016
* Initialize the "longest match" routines for a new zlib stream
1017
*/
1018
local void lm_init (s)
1019
deflate_state *s;
1020
{
1021
s->window_size = (ulg)2L*s->w_size;
1022
1023
CLEAR_HASH(s);
1024
1025
/* Set the default configuration parameters:
1026
*/
1027
s->max_lazy_match = configuration_table[s->level].max_lazy;
1028
s->good_match = configuration_table[s->level].good_length;
1029
s->nice_match = configuration_table[s->level].nice_length;
1030
s->max_chain_length = configuration_table[s->level].max_chain;
1031
1032
s->strstart = 0;
1033
s->block_start = 0L;
1034
s->lookahead = 0;
1035
s->match_length = s->prev_length = MIN_MATCH-1;
1036
s->match_available = 0;
1037
s->ins_h = 0;
1038
#ifndef FASTEST
1039
#ifdef ASMV
1040
match_init(); /* initialize the asm code */
1041
#endif
1042
#endif
1043
}
1044
1045
#ifndef FASTEST
1046
/* ===========================================================================
1047
* Set match_start to the longest match starting at the given string and
1048
* return its length. Matches shorter or equal to prev_length are discarded,
1049
* in which case the result is equal to prev_length and match_start is
1050
* garbage.
1051
* IN assertions: cur_match is the head of the hash chain for the current
1052
* string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
1053
* OUT assertion: the match length is not greater than s->lookahead.
1054
*/
1055
#ifndef ASMV
1056
/* For 80x86 and 680x0, an optimized version will be provided in match.asm or
1057
* match.S. The code will be functionally equivalent.
1058
*/
1059
local uInt longest_match(s, cur_match)
1060
deflate_state *s;
1061
IPos cur_match; /* current match */
1062
{
1063
unsigned chain_length = s->max_chain_length;/* max hash chain length */
1064
register Bytef *scan = s->window + s->strstart; /* current string */
1065
register Bytef *match; /* matched string */
1066
register int len; /* length of current match */
1067
int best_len = s->prev_length; /* best match length so far */
1068
int nice_match = s->nice_match; /* stop if match long enough */
1069
IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
1070
s->strstart - (IPos)MAX_DIST(s) : NIL;
1071
/* Stop when cur_match becomes <= limit. To simplify the code,
1072
* we prevent matches with the string of window index 0.
1073
*/
1074
Posf *prev = s->prev;
1075
uInt wmask = s->w_mask;
1076
1077
#ifdef UNALIGNED_OK
1078
/* Compare two bytes at a time. Note: this is not always beneficial.
1079
* Try with and without -DUNALIGNED_OK to check.
1080
*/
1081
register Bytef *strend = s->window + s->strstart + MAX_MATCH - 1;
1082
register ush scan_start = *(ushf*)scan;
1083
register ush scan_end = *(ushf*)(scan+best_len-1);
1084
#else
1085
register Bytef *strend = s->window + s->strstart + MAX_MATCH;
1086
register Byte scan_end1 = scan[best_len-1];
1087
register Byte scan_end = scan[best_len];
1088
#endif
1089
1090
/* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
1091
* It is easy to get rid of this optimization if necessary.
1092
*/
1093
Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
1094
1095
/* Do not waste too much time if we already have a good match: */
1096
if (s->prev_length >= s->good_match) {
1097
chain_length >>= 2;
1098
}
1099
/* Do not look for matches beyond the end of the input. This is necessary
1100
* to make deflate deterministic.
1101
*/
1102
if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead;
1103
1104
Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
1105
1106
do {
1107
Assert(cur_match < s->strstart, "no future");
1108
match = s->window + cur_match;
1109
1110
/* Skip to next match if the match length cannot increase
1111
* or if the match length is less than 2. Note that the checks below
1112
* for insufficient lookahead only occur occasionally for performance
1113
* reasons. Therefore uninitialized memory will be accessed, and
1114
* conditional jumps will be made that depend on those values.
1115
* However the length of the match is limited to the lookahead, so
1116
* the output of deflate is not affected by the uninitialized values.
1117
*/
1118
#if (defined(UNALIGNED_OK) && MAX_MATCH == 258)
1119
/* This code assumes sizeof(unsigned short) == 2. Do not use
1120
* UNALIGNED_OK if your compiler uses a different size.
1121
*/
1122
if (*(ushf*)(match+best_len-1) != scan_end ||
1123
*(ushf*)match != scan_start) continue;
1124
1125
/* It is not necessary to compare scan[2] and match[2] since they are
1126
* always equal when the other bytes match, given that the hash keys
1127
* are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at
1128
* strstart+3, +5, ... up to strstart+257. We check for insufficient
1129
* lookahead only every 4th comparison; the 128th check will be made
1130
* at strstart+257. If MAX_MATCH-2 is not a multiple of 8, it is
1131
* necessary to put more guard bytes at the end of the window, or
1132
* to check more often for insufficient lookahead.
1133
*/
1134
Assert(scan[2] == match[2], "scan[2]?");
1135
scan++, match++;
1136
do {
1137
} while (*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
1138
*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
1139
*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
1140
*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
1141
scan < strend);
1142
/* The funny "do {}" generates better code on most compilers */
1143
1144
/* Here, scan <= window+strstart+257 */
1145
Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
1146
if (*scan == *match) scan++;
1147
1148
len = (MAX_MATCH - 1) - (int)(strend-scan);
1149
scan = strend - (MAX_MATCH-1);
1150
1151
#else /* UNALIGNED_OK */
1152
1153
if (match[best_len] != scan_end ||
1154
match[best_len-1] != scan_end1 ||
1155
*match != *scan ||
1156
*++match != scan[1]) continue;
1157
1158
/* The check at best_len-1 can be removed because it will be made
1159
* again later. (This heuristic is not always a win.)
1160
* It is not necessary to compare scan[2] and match[2] since they
1161
* are always equal when the other bytes match, given that
1162
* the hash keys are equal and that HASH_BITS >= 8.
1163
*/
1164
scan += 2, match++;
1165
Assert(*scan == *match, "match[2]?");
1166
1167
/* We check for insufficient lookahead only every 8th comparison;
1168
* the 256th check will be made at strstart+258.
1169
*/
1170
do {
1171
} while (*++scan == *++match && *++scan == *++match &&
1172
*++scan == *++match && *++scan == *++match &&
1173
*++scan == *++match && *++scan == *++match &&
1174
*++scan == *++match && *++scan == *++match &&
1175
scan < strend);
1176
1177
Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
1178
1179
len = MAX_MATCH - (int)(strend - scan);
1180
scan = strend - MAX_MATCH;
1181
1182
#endif /* UNALIGNED_OK */
1183
1184
if (len > best_len) {
1185
s->match_start = cur_match;
1186
best_len = len;
1187
if (len >= nice_match) break;
1188
#ifdef UNALIGNED_OK
1189
scan_end = *(ushf*)(scan+best_len-1);
1190
#else
1191
scan_end1 = scan[best_len-1];
1192
scan_end = scan[best_len];
1193
#endif
1194
}
1195
} while ((cur_match = prev[cur_match & wmask]) > limit
1196
&& --chain_length != 0);
1197
1198
if ((uInt)best_len <= s->lookahead) return (uInt)best_len;
1199
return s->lookahead;
1200
}
1201
#endif /* ASMV */
1202
#endif /* FASTEST */
1203
1204
/* ---------------------------------------------------------------------------
1205
* Optimized version for level == 1 or strategy == Z_RLE only
1206
*/
1207
local uInt longest_match_fast(s, cur_match)
1208
deflate_state *s;
1209
IPos cur_match; /* current match */
1210
{
1211
register Bytef *scan = s->window + s->strstart; /* current string */
1212
register Bytef *match; /* matched string */
1213
register int len; /* length of current match */
1214
register Bytef *strend = s->window + s->strstart + MAX_MATCH;
1215
1216
/* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
1217
* It is easy to get rid of this optimization if necessary.
1218
*/
1219
Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
1220
1221
Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
1222
1223
Assert(cur_match < s->strstart, "no future");
1224
1225
match = s->window + cur_match;
1226
1227
/* Return failure if the match length is less than 2:
1228
*/
1229
if (match[0] != scan[0] || match[1] != scan[1]) return MIN_MATCH-1;
1230
1231
/* The check at best_len-1 can be removed because it will be made
1232
* again later. (This heuristic is not always a win.)
1233
* It is not necessary to compare scan[2] and match[2] since they
1234
* are always equal when the other bytes match, given that
1235
* the hash keys are equal and that HASH_BITS >= 8.
1236
*/
1237
scan += 2, match += 2;
1238
Assert(*scan == *match, "match[2]?");
1239
1240
/* We check for insufficient lookahead only every 8th comparison;
1241
* the 256th check will be made at strstart+258.
1242
*/
1243
do {
1244
} while (*++scan == *++match && *++scan == *++match &&
1245
*++scan == *++match && *++scan == *++match &&
1246
*++scan == *++match && *++scan == *++match &&
1247
*++scan == *++match && *++scan == *++match &&
1248
scan < strend);
1249
1250
Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
1251
1252
len = MAX_MATCH - (int)(strend - scan);
1253
1254
if (len < MIN_MATCH) return MIN_MATCH - 1;
1255
1256
s->match_start = cur_match;
1257
return (uInt)len <= s->lookahead ? (uInt)len : s->lookahead;
1258
}
1259
1260
#ifdef DEBUG
1261
/* ===========================================================================
1262
* Check that the match at match_start is indeed a match.
1263
*/
1264
local void check_match(s, start, match, length)
1265
deflate_state *s;
1266
IPos start, match;
1267
int length;
1268
{
1269
/* check that the match is indeed a match */
1270
if (zmemcmp(s->window + match,
1271
s->window + start, length) != EQUAL) {
1272
fprintf(stderr, " start %u, match %u, length %d\n",
1273
start, match, length);
1274
do {
1275
fprintf(stderr, "%c%c", s->window[match++], s->window[start++]);
1276
} while (--length != 0);
1277
z_error("invalid match");
1278
}
1279
if (z_verbose > 1) {
1280
fprintf(stderr,"\\[%d,%d]", start-match, length);
1281
do { putc(s->window[start++], stderr); } while (--length != 0);
1282
}
1283
}
1284
#else
1285
# define check_match(s, start, match, length)
1286
#endif /* DEBUG */
1287
1288
/* ===========================================================================
1289
* Fill the window when the lookahead becomes insufficient.
1290
* Updates strstart and lookahead.
1291
*
1292
* IN assertion: lookahead < MIN_LOOKAHEAD
1293
* OUT assertions: strstart <= window_size-MIN_LOOKAHEAD
1294
* At least one byte has been read, or avail_in == 0; reads are
1295
* performed for at least two bytes (required for the zip translate_eol
1296
* option -- not supported here).
1297
*/
1298
local void fill_window(s)
1299
deflate_state *s;
1300
{
1301
register unsigned n, m;
1302
register Posf *p;
1303
unsigned more; /* Amount of free space at the end of the window. */
1304
uInt wsize = s->w_size;
1305
1306
do {
1307
more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart);
1308
1309
/* Deal with !@#$% 64K limit: */
1310
if (sizeof(int) <= 2) {
1311
if (more == 0 && s->strstart == 0 && s->lookahead == 0) {
1312
more = wsize;
1313
1314
} else if (more == (unsigned)(-1)) {
1315
/* Very unlikely, but possible on 16 bit machine if
1316
* strstart == 0 && lookahead == 1 (input done a byte at time)
1317
*/
1318
more--;
1319
}
1320
}
1321
1322
/* If the window is almost full and there is insufficient lookahead,
1323
* move the upper half to the lower one to make room in the upper half.
1324
*/
1325
if (s->strstart >= wsize+MAX_DIST(s)) {
1326
1327
zmemcpy(s->window, s->window+wsize, (unsigned)wsize);
1328
s->match_start -= wsize;
1329
s->strstart -= wsize; /* we now have strstart >= MAX_DIST */
1330
s->block_start -= (long) wsize;
1331
1332
/* Slide the hash table (could be avoided with 32 bit values
1333
at the expense of memory usage). We slide even when level == 0
1334
to keep the hash table consistent if we switch back to level > 0
1335
later. (Using level 0 permanently is not an optimal usage of
1336
zlib, so we don't care about this pathological case.)
1337
*/
1338
/* %%% avoid this when Z_RLE */
1339
n = s->hash_size;
1340
p = &s->head[n];
1341
do {
1342
m = *--p;
1343
*p = (Pos)(m >= wsize ? m-wsize : NIL);
1344
} while (--n);
1345
1346
n = wsize;
1347
#ifndef FASTEST
1348
p = &s->prev[n];
1349
do {
1350
m = *--p;
1351
*p = (Pos)(m >= wsize ? m-wsize : NIL);
1352
/* If n is not on any hash chain, prev[n] is garbage but
1353
* its value will never be used.
1354
*/
1355
} while (--n);
1356
#endif
1357
more += wsize;
1358
}
1359
if (s->strm->avail_in == 0) return;
1360
1361
/* If there was no sliding:
1362
* strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 &&
1363
* more == window_size - lookahead - strstart
1364
* => more >= window_size - (MIN_LOOKAHEAD-1 + WSIZE + MAX_DIST-1)
1365
* => more >= window_size - 2*WSIZE + 2
1366
* In the BIG_MEM or MMAP case (not yet supported),
1367
* window_size == input_size + MIN_LOOKAHEAD &&
1368
* strstart + s->lookahead <= input_size => more >= MIN_LOOKAHEAD.
1369
* Otherwise, window_size == 2*WSIZE so more >= 2.
1370
* If there was sliding, more >= WSIZE. So in all cases, more >= 2.
1371
*/
1372
Assert(more >= 2, "more < 2");
1373
1374
n = read_buf(s->strm, s->window + s->strstart + s->lookahead, more);
1375
s->lookahead += n;
1376
1377
/* Initialize the hash value now that we have some input: */
1378
if (s->lookahead >= MIN_MATCH) {
1379
s->ins_h = s->window[s->strstart];
1380
UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
1381
#if MIN_MATCH != 3
1382
Call UPDATE_HASH() MIN_MATCH-3 more times
1383
#endif
1384
}
1385
/* If the whole input has less than MIN_MATCH bytes, ins_h is garbage,
1386
* but this is not important since only literal bytes will be emitted.
1387
*/
1388
1389
} while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0);
1390
}
1391
1392
/* ===========================================================================
1393
* Flush the current block, with given end-of-file flag.
1394
* IN assertion: strstart is set to the end of the current match.
1395
*/
1396
#define FLUSH_BLOCK_ONLY(s, eof) { \
1397
_tr_flush_block(s, (s->block_start >= 0L ? \
1398
(charf *)&s->window[(unsigned)s->block_start] : \
1399
(charf *)Z_NULL), \
1400
(ulg)((long)s->strstart - s->block_start), \
1401
(eof)); \
1402
s->block_start = s->strstart; \
1403
flush_pending(s->strm); \
1404
Tracev((stderr,"[FLUSH]")); \
1405
}
1406
1407
/* Same but force premature exit if necessary. */
1408
#define FLUSH_BLOCK(s, eof) { \
1409
FLUSH_BLOCK_ONLY(s, eof); \
1410
if (s->strm->avail_out == 0) return (eof) ? finish_started : need_more; \
1411
}
1412
1413
/* ===========================================================================
1414
* Copy without compression as much as possible from the input stream, return
1415
* the current block state.
1416
* This function does not insert new strings in the dictionary since
1417
* uncompressible data is probably not useful. This function is used
1418
* only for the level=0 compression option.
1419
* NOTE: this function should be optimized to avoid extra copying from
1420
* window to pending_buf.
1421
*/
1422
local block_state deflate_stored(s, flush)
1423
deflate_state *s;
1424
int flush;
1425
{
1426
/* Stored blocks are limited to 0xffff bytes, pending_buf is limited
1427
* to pending_buf_size, and each stored block has a 5 byte header:
1428
*/
1429
ulg max_block_size = 0xffff;
1430
ulg max_start;
1431
1432
if (max_block_size > s->pending_buf_size - 5) {
1433
max_block_size = s->pending_buf_size - 5;
1434
}
1435
1436
/* Copy as much as possible from input to output: */
1437
for (;;) {
1438
/* Fill the window as much as possible: */
1439
if (s->lookahead <= 1) {
1440
1441
Assert(s->strstart < s->w_size+MAX_DIST(s) ||
1442
s->block_start >= (long)s->w_size, "slide too late");
1443
1444
fill_window(s);
1445
if (s->lookahead == 0 && flush == Z_NO_FLUSH) return need_more;
1446
1447
if (s->lookahead == 0) break; /* flush the current block */
1448
}
1449
Assert(s->block_start >= 0L, "block gone");
1450
1451
s->strstart += s->lookahead;
1452
s->lookahead = 0;
1453
1454
/* Emit a stored block if pending_buf will be full: */
1455
max_start = s->block_start + max_block_size;
1456
if (s->strstart == 0 || (ulg)s->strstart >= max_start) {
1457
/* strstart == 0 is possible when wraparound on 16-bit machine */
1458
s->lookahead = (uInt)(s->strstart - max_start);
1459
s->strstart = (uInt)max_start;
1460
FLUSH_BLOCK(s, 0);
1461
}
1462
/* Flush if we may have to slide, otherwise block_start may become
1463
* negative and the data will be gone:
1464
*/
1465
if (s->strstart - (uInt)s->block_start >= MAX_DIST(s)) {
1466
FLUSH_BLOCK(s, 0);
1467
}
1468
}
1469
FLUSH_BLOCK(s, flush == Z_FINISH);
1470
return flush == Z_FINISH ? finish_done : block_done;
1471
}
1472
1473
/* ===========================================================================
1474
* Compress as much as possible from the input stream, return the current
1475
* block state.
1476
* This function does not perform lazy evaluation of matches and inserts
1477
* new strings in the dictionary only for unmatched strings or for short
1478
* matches. It is used only for the fast compression options.
1479
*/
1480
local block_state deflate_fast(s, flush)
1481
deflate_state *s;
1482
int flush;
1483
{
1484
IPos hash_head = NIL; /* head of the hash chain */
1485
int bflush; /* set if current block must be flushed */
1486
1487
for (;;) {
1488
/* Make sure that we always have enough lookahead, except
1489
* at the end of the input file. We need MAX_MATCH bytes
1490
* for the next match, plus MIN_MATCH bytes to insert the
1491
* string following the next match.
1492
*/
1493
if (s->lookahead < MIN_LOOKAHEAD) {
1494
fill_window(s);
1495
if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
1496
return need_more;
1497
}
1498
if (s->lookahead == 0) break; /* flush the current block */
1499
}
1500
1501
/* Insert the string window[strstart .. strstart+2] in the
1502
* dictionary, and set hash_head to the head of the hash chain:
1503
*/
1504
if (s->lookahead >= MIN_MATCH) {
1505
INSERT_STRING(s, s->strstart, hash_head);
1506
}
1507
1508
/* Find the longest match, discarding those <= prev_length.
1509
* At this point we have always match_length < MIN_MATCH
1510
*/
1511
if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) {
1512
/* To simplify the code, we prevent matches with the string
1513
* of window index 0 (in particular we have to avoid a match
1514
* of the string with itself at the start of the input file).
1515
*/
1516
#ifdef FASTEST
1517
if ((s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) ||
1518
(s->strategy == Z_RLE && s->strstart - hash_head == 1)) {
1519
s->match_length = longest_match_fast (s, hash_head);
1520
}
1521
#else
1522
if (s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) {
1523
s->match_length = longest_match (s, hash_head);
1524
} else if (s->strategy == Z_RLE && s->strstart - hash_head == 1) {
1525
s->match_length = longest_match_fast (s, hash_head);
1526
}
1527
#endif
1528
/* longest_match() or longest_match_fast() sets match_start */
1529
}
1530
if (s->match_length >= MIN_MATCH) {
1531
check_match(s, s->strstart, s->match_start, s->match_length);
1532
1533
_tr_tally_dist(s, s->strstart - s->match_start,
1534
s->match_length - MIN_MATCH, bflush);
1535
1536
s->lookahead -= s->match_length;
1537
1538
/* Insert new strings in the hash table only if the match length
1539
* is not too large. This saves time but degrades compression.
1540
*/
1541
#ifndef FASTEST
1542
if (s->match_length <= s->max_insert_length &&
1543
s->lookahead >= MIN_MATCH) {
1544
s->match_length--; /* string at strstart already in table */
1545
do {
1546
s->strstart++;
1547
INSERT_STRING(s, s->strstart, hash_head);
1548
/* strstart never exceeds WSIZE-MAX_MATCH, so there are
1549
* always MIN_MATCH bytes ahead.
1550
*/
1551
} while (--s->match_length != 0);
1552
s->strstart++;
1553
} else
1554
#endif
1555
{
1556
s->strstart += s->match_length;
1557
s->match_length = 0;
1558
s->ins_h = s->window[s->strstart];
1559
UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
1560
#if MIN_MATCH != 3
1561
Call UPDATE_HASH() MIN_MATCH-3 more times
1562
#endif
1563
/* If lookahead < MIN_MATCH, ins_h is garbage, but it does not
1564
* matter since it will be recomputed at next deflate call.
1565
*/
1566
}
1567
} else {
1568
/* No match, output a literal byte */
1569
Tracevv((stderr,"%c", s->window[s->strstart]));
1570
_tr_tally_lit (s, s->window[s->strstart], bflush);
1571
s->lookahead--;
1572
s->strstart++;
1573
}
1574
if (bflush) FLUSH_BLOCK(s, 0);
1575
}
1576
FLUSH_BLOCK(s, flush == Z_FINISH);
1577
return flush == Z_FINISH ? finish_done : block_done;
1578
}
1579
1580
#ifndef FASTEST
1581
/* ===========================================================================
1582
* Same as above, but achieves better compression. We use a lazy
1583
* evaluation for matches: a match is finally adopted only if there is
1584
* no better match at the next window position.
1585
*/
1586
local block_state deflate_slow(s, flush)
1587
deflate_state *s;
1588
int flush;
1589
{
1590
IPos hash_head = NIL; /* head of hash chain */
1591
int bflush; /* set if current block must be flushed */
1592
1593
/* Process the input block. */
1594
for (;;) {
1595
/* Make sure that we always have enough lookahead, except
1596
* at the end of the input file. We need MAX_MATCH bytes
1597
* for the next match, plus MIN_MATCH bytes to insert the
1598
* string following the next match.
1599
*/
1600
if (s->lookahead < MIN_LOOKAHEAD) {
1601
fill_window(s);
1602
if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
1603
return need_more;
1604
}
1605
if (s->lookahead == 0) break; /* flush the current block */
1606
}
1607
1608
/* Insert the string window[strstart .. strstart+2] in the
1609
* dictionary, and set hash_head to the head of the hash chain:
1610
*/
1611
if (s->lookahead >= MIN_MATCH) {
1612
INSERT_STRING(s, s->strstart, hash_head);
1613
}
1614
1615
/* Find the longest match, discarding those <= prev_length.
1616
*/
1617
s->prev_length = s->match_length, s->prev_match = s->match_start;
1618
s->match_length = MIN_MATCH-1;
1619
1620
if (hash_head != NIL && s->prev_length < s->max_lazy_match &&
1621
s->strstart - hash_head <= MAX_DIST(s)) {
1622
/* To simplify the code, we prevent matches with the string
1623
* of window index 0 (in particular we have to avoid a match
1624
* of the string with itself at the start of the input file).
1625
*/
1626
if (s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) {
1627
s->match_length = longest_match (s, hash_head);
1628
} else if (s->strategy == Z_RLE && s->strstart - hash_head == 1) {
1629
s->match_length = longest_match_fast (s, hash_head);
1630
}
1631
/* longest_match() or longest_match_fast() sets match_start */
1632
1633
if (s->match_length <= 5 && (s->strategy == Z_FILTERED
1634
#if TOO_FAR <= 32767
1635
|| (s->match_length == MIN_MATCH &&
1636
s->strstart - s->match_start > TOO_FAR)
1637
#endif
1638
)) {
1639
1640
/* If prev_match is also MIN_MATCH, match_start is garbage
1641
* but we will ignore the current match anyway.
1642
*/
1643
s->match_length = MIN_MATCH-1;
1644
}
1645
}
1646
/* If there was a match at the previous step and the current
1647
* match is not better, output the previous match:
1648
*/
1649
if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) {
1650
uInt max_insert = s->strstart + s->lookahead - MIN_MATCH;
1651
/* Do not insert strings in hash table beyond this. */
1652
1653
check_match(s, s->strstart-1, s->prev_match, s->prev_length);
1654
1655
_tr_tally_dist(s, s->strstart -1 - s->prev_match,
1656
s->prev_length - MIN_MATCH, bflush);
1657
1658
/* Insert in hash table all strings up to the end of the match.
1659
* strstart-1 and strstart are already inserted. If there is not
1660
* enough lookahead, the last two strings are not inserted in
1661
* the hash table.
1662
*/
1663
s->lookahead -= s->prev_length-1;
1664
s->prev_length -= 2;
1665
do {
1666
if (++s->strstart <= max_insert) {
1667
INSERT_STRING(s, s->strstart, hash_head);
1668
}
1669
} while (--s->prev_length != 0);
1670
s->match_available = 0;
1671
s->match_length = MIN_MATCH-1;
1672
s->strstart++;
1673
1674
if (bflush) FLUSH_BLOCK(s, 0);
1675
1676
} else if (s->match_available) {
1677
/* If there was no match at the previous position, output a
1678
* single literal. If there was a match but the current match
1679
* is longer, truncate the previous match to a single literal.
1680
*/
1681
Tracevv((stderr,"%c", s->window[s->strstart-1]));
1682
_tr_tally_lit(s, s->window[s->strstart-1], bflush);
1683
if (bflush) {
1684
FLUSH_BLOCK_ONLY(s, 0);
1685
}
1686
s->strstart++;
1687
s->lookahead--;
1688
if (s->strm->avail_out == 0) return need_more;
1689
} else {
1690
/* There is no previous match to compare with, wait for
1691
* the next step to decide.
1692
*/
1693
s->match_available = 1;
1694
s->strstart++;
1695
s->lookahead--;
1696
}
1697
}
1698
Assert (flush != Z_NO_FLUSH, "no flush?");
1699
if (s->match_available) {
1700
Tracevv((stderr,"%c", s->window[s->strstart-1]));
1701
_tr_tally_lit(s, s->window[s->strstart-1], bflush);
1702
s->match_available = 0;
1703
}
1704
FLUSH_BLOCK(s, flush == Z_FINISH);
1705
return flush == Z_FINISH ? finish_done : block_done;
1706
}
1707
#endif /* FASTEST */
1708
1709
#if 0
1710
/* ===========================================================================
1711
* For Z_RLE, simply look for runs of bytes, generate matches only of distance
1712
* one. Do not maintain a hash table. (It will be regenerated if this run of
1713
* deflate switches away from Z_RLE.)
1714
*/
1715
local block_state deflate_rle(s, flush)
1716
deflate_state *s;
1717
int flush;
1718
{
1719
int bflush; /* set if current block must be flushed */
1720
uInt run; /* length of run */
1721
uInt max; /* maximum length of run */
1722
uInt prev; /* byte at distance one to match */
1723
Bytef *scan; /* scan for end of run */
1724
1725
for (;;) {
1726
/* Make sure that we always have enough lookahead, except
1727
* at the end of the input file. We need MAX_MATCH bytes
1728
* for the longest encodable run.
1729
*/
1730
if (s->lookahead < MAX_MATCH) {
1731
fill_window(s);
1732
if (s->lookahead < MAX_MATCH && flush == Z_NO_FLUSH) {
1733
return need_more;
1734
}
1735
if (s->lookahead == 0) break; /* flush the current block */
1736
}
1737
1738
/* See how many times the previous byte repeats */
1739
run = 0;
1740
if (s->strstart > 0) { /* if there is a previous byte, that is */
1741
max = s->lookahead < MAX_MATCH ? s->lookahead : MAX_MATCH;
1742
scan = s->window + s->strstart - 1;
1743
prev = *scan++;
1744
do {
1745
if (*scan++ != prev)
1746
break;
1747
} while (++run < max);
1748
}
1749
1750
/* Emit match if have run of MIN_MATCH or longer, else emit literal */
1751
if (run >= MIN_MATCH) {
1752
check_match(s, s->strstart, s->strstart - 1, run);
1753
_tr_tally_dist(s, 1, run - MIN_MATCH, bflush);
1754
s->lookahead -= run;
1755
s->strstart += run;
1756
} else {
1757
/* No match, output a literal byte */
1758
Tracevv((stderr,"%c", s->window[s->strstart]));
1759
_tr_tally_lit (s, s->window[s->strstart], bflush);
1760
s->lookahead--;
1761
s->strstart++;
1762
}
1763
if (bflush) FLUSH_BLOCK(s, 0);
1764
}
1765
FLUSH_BLOCK(s, flush == Z_FINISH);
1766
return flush == Z_FINISH ? finish_done : block_done;
1767
}
1768
#endif
Loggerhead is a web-based interface for Breezy
Version: 2.0.1