1 /* |
|
2 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. |
|
3 * |
|
4 * This code is free software; you can redistribute it and/or modify it |
|
5 * under the terms of the GNU General Public License version 2 only, as |
|
6 * published by the Free Software Foundation. Oracle designates this |
|
7 * particular file as subject to the "Classpath" exception as provided |
|
8 * by Oracle in the LICENSE file that accompanied this code. |
|
9 * |
|
10 * This code is distributed in the hope that it will be useful, but WITHOUT |
|
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or |
|
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
|
13 * version 2 for more details (a copy is included in the LICENSE file that |
|
14 * accompanied this code). |
|
15 * |
|
16 * You should have received a copy of the GNU General Public License version |
|
17 * 2 along with this work; if not, write to the Free Software Foundation, |
|
18 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. |
|
19 * |
|
20 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA |
|
21 * or visit www.oracle.com if you need additional information or have any |
|
22 * questions. |
|
23 */ |
|
24 |
|
25 /* deflate.c -- compress data using the deflation algorithm |
|
26 * Copyright (C) 1995-2010 Jean-loup Gailly and Mark Adler |
|
27 * For conditions of distribution and use, see copyright notice in zlib.h |
|
28 */ |
|
29 |
|
30 /* |
|
31 * ALGORITHM |
|
32 * |
|
33 * The "deflation" process depends on being able to identify portions |
|
34 * of the input text which are identical to earlier input (within a |
|
35 * sliding window trailing behind the input currently being processed). |
|
36 * |
|
37 * The most straightforward technique turns out to be the fastest for |
|
38 * most input files: try all possible matches and select the longest. |
|
39 * The key feature of this algorithm is that insertions into the string |
|
40 * dictionary are very simple and thus fast, and deletions are avoided |
|
41 * completely. Insertions are performed at each input character, whereas |
|
42 * string matches are performed only when the previous match ends. So it |
|
43 * is preferable to spend more time in matches to allow very fast string |
|
44 * insertions and avoid deletions. The matching algorithm for small |
|
45 * strings is inspired from that of Rabin & Karp. A brute force approach |
|
46 * is used to find longer strings when a small match has been found. |
|
47 * A similar algorithm is used in comic (by Jan-Mark Wams) and freeze |
|
48 * (by Leonid Broukhis). |
|
49 * A previous version of this file used a more sophisticated algorithm |
|
50 * (by Fiala and Greene) which is guaranteed to run in linear amortized |
|
51 * time, but has a larger average cost, uses more memory and is patented. |
|
52 * However the F&G algorithm may be faster for some highly redundant |
|
53 * files if the parameter max_chain_length (described below) is too large. |
|
54 * |
|
55 * ACKNOWLEDGEMENTS |
|
56 * |
|
57 * The idea of lazy evaluation of matches is due to Jan-Mark Wams, and |
|
58 * I found it in 'freeze' written by Leonid Broukhis. |
|
59 * Thanks to many people for bug reports and testing. |
|
60 * |
|
61 * REFERENCES |
|
62 * |
|
63 * Deutsch, L.P.,"DEFLATE Compressed Data Format Specification". |
|
64 * Available in http://www.ietf.org/rfc/rfc1951.txt |
|
65 * |
|
66 * A description of the Rabin and Karp algorithm is given in the book |
|
67 * "Algorithms" by R. Sedgewick, Addison-Wesley, p252. |
|
68 * |
|
69 * Fiala,E.R., and Greene,D.H. |
|
70 * Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595 |
|
71 * |
|
72 */ |
|
73 |
|
74 /* @(#) $Id$ */ |
|
75 |
|
76 #include "deflate.h" |
|
77 |
|
78 const char deflate_copyright[] = |
|
79 " deflate 1.2.5 Copyright 1995-2010 Jean-loup Gailly and Mark Adler "; |
|
80 /* |
|
81 If you use the zlib library in a product, an acknowledgment is welcome |
|
82 in the documentation of your product. If for some reason you cannot |
|
83 include such an acknowledgment, I would appreciate that you keep this |
|
84 copyright string in the executable of your product. |
|
85 */ |
|
86 |
|
87 /* =========================================================================== |
|
88 * Function prototypes. |
|
89 */ |
|
90 typedef enum { |
|
91 need_more, /* block not completed, need more input or more output */ |
|
92 block_done, /* block flush performed */ |
|
93 finish_started, /* finish started, need only more output at next deflate */ |
|
94 finish_done /* finish done, accept no more input or output */ |
|
95 } block_state; |
|
96 |
|
97 typedef block_state (*compress_func) OF((deflate_state *s, int flush)); |
|
98 /* Compression function. Returns the block state after the call. */ |
|
99 |
|
100 local void fill_window OF((deflate_state *s)); |
|
101 local block_state deflate_stored OF((deflate_state *s, int flush)); |
|
102 local block_state deflate_fast OF((deflate_state *s, int flush)); |
|
103 #ifndef FASTEST |
|
104 local block_state deflate_slow OF((deflate_state *s, int flush)); |
|
105 #endif |
|
106 local block_state deflate_rle OF((deflate_state *s, int flush)); |
|
107 local block_state deflate_huff OF((deflate_state *s, int flush)); |
|
108 local void lm_init OF((deflate_state *s)); |
|
109 local void putShortMSB OF((deflate_state *s, uInt b)); |
|
110 local void flush_pending OF((z_streamp strm)); |
|
111 local int read_buf OF((z_streamp strm, Bytef *buf, unsigned size)); |
|
112 #ifdef ASMV |
|
113 void match_init OF((void)); /* asm code initialization */ |
|
114 uInt longest_match OF((deflate_state *s, IPos cur_match)); |
|
115 #else |
|
116 local uInt longest_match OF((deflate_state *s, IPos cur_match)); |
|
117 #endif |
|
118 |
|
119 #ifdef DEBUG |
|
120 local void check_match OF((deflate_state *s, IPos start, IPos match, |
|
121 int length)); |
|
122 #endif |
|
123 |
|
124 /* =========================================================================== |
|
125 * Local data |
|
126 */ |
|
127 |
|
128 #define NIL 0 |
|
129 /* Tail of hash chains */ |
|
130 |
|
131 #ifndef TOO_FAR |
|
132 # define TOO_FAR 4096 |
|
133 #endif |
|
134 /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */ |
|
135 |
|
136 /* Values for max_lazy_match, good_match and max_chain_length, depending on |
|
137 * the desired pack level (0..9). The values given below have been tuned to |
|
138 * exclude worst case performance for pathological files. Better values may be |
|
139 * found for specific files. |
|
140 */ |
|
141 typedef struct config_s { |
|
142 ush good_length; /* reduce lazy search above this match length */ |
|
143 ush max_lazy; /* do not perform lazy search above this match length */ |
|
144 ush nice_length; /* quit search above this match length */ |
|
145 ush max_chain; |
|
146 compress_func func; |
|
147 } config; |
|
148 |
|
149 #ifdef FASTEST |
|
150 local const config configuration_table[2] = { |
|
151 /* good lazy nice chain */ |
|
152 /* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */ |
|
153 /* 1 */ {4, 4, 8, 4, deflate_fast}}; /* max speed, no lazy matches */ |
|
154 #else |
|
155 local const config configuration_table[10] = { |
|
156 /* good lazy nice chain */ |
|
157 /* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */ |
|
158 /* 1 */ {4, 4, 8, 4, deflate_fast}, /* max speed, no lazy matches */ |
|
159 /* 2 */ {4, 5, 16, 8, deflate_fast}, |
|
160 /* 3 */ {4, 6, 32, 32, deflate_fast}, |
|
161 |
|
162 /* 4 */ {4, 4, 16, 16, deflate_slow}, /* lazy matches */ |
|
163 /* 5 */ {8, 16, 32, 32, deflate_slow}, |
|
164 /* 6 */ {8, 16, 128, 128, deflate_slow}, |
|
165 /* 7 */ {8, 32, 128, 256, deflate_slow}, |
|
166 /* 8 */ {32, 128, 258, 1024, deflate_slow}, |
|
167 /* 9 */ {32, 258, 258, 4096, deflate_slow}}; /* max compression */ |
|
168 #endif |
|
169 |
|
170 /* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4 |
|
171 * For deflate_fast() (levels <= 3) good is ignored and lazy has a different |
|
172 * meaning. |
|
173 */ |
|
174 |
|
175 #define EQUAL 0 |
|
176 /* result of memcmp for equal strings */ |
|
177 |
|
178 #ifndef NO_DUMMY_DECL |
|
179 struct static_tree_desc_s {int dummy;}; /* for buggy compilers */ |
|
180 #endif |
|
181 |
|
182 /* =========================================================================== |
|
183 * Update a hash value with the given input byte |
|
184 * IN assertion: all calls to to UPDATE_HASH are made with consecutive |
|
185 * input characters, so that a running hash key can be computed from the |
|
186 * previous key instead of complete recalculation each time. |
|
187 */ |
|
188 #define UPDATE_HASH(s,h,c) (h = (((h)<<s->hash_shift) ^ (c)) & s->hash_mask) |
|
189 |
|
190 |
|
191 /* =========================================================================== |
|
192 * Insert string str in the dictionary and set match_head to the previous head |
|
193 * of the hash chain (the most recent string with same hash key). Return |
|
194 * the previous length of the hash chain. |
|
195 * If this file is compiled with -DFASTEST, the compression level is forced |
|
196 * to 1, and no hash chains are maintained. |
|
197 * IN assertion: all calls to to INSERT_STRING are made with consecutive |
|
198 * input characters and the first MIN_MATCH bytes of str are valid |
|
199 * (except for the last MIN_MATCH-1 bytes of the input file). |
|
200 */ |
|
201 #ifdef FASTEST |
|
202 #define INSERT_STRING(s, str, match_head) \ |
|
203 (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \ |
|
204 match_head = s->head[s->ins_h], \ |
|
205 s->head[s->ins_h] = (Pos)(str)) |
|
206 #else |
|
207 #define INSERT_STRING(s, str, match_head) \ |
|
208 (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \ |
|
209 match_head = s->prev[(str) & s->w_mask] = s->head[s->ins_h], \ |
|
210 s->head[s->ins_h] = (Pos)(str)) |
|
211 #endif |
|
212 |
|
213 /* =========================================================================== |
|
214 * Initialize the hash table (avoiding 64K overflow for 16 bit systems). |
|
215 * prev[] will be initialized on the fly. |
|
216 */ |
|
217 #define CLEAR_HASH(s) \ |
|
218 s->head[s->hash_size-1] = NIL; \ |
|
219 zmemzero((Bytef *)s->head, (unsigned)(s->hash_size-1)*sizeof(*s->head)); |
|
220 |
|
221 /* ========================================================================= */ |
|
222 int ZEXPORT deflateInit_(strm, level, version, stream_size) |
|
223 z_streamp strm; |
|
224 int level; |
|
225 const char *version; |
|
226 int stream_size; |
|
227 { |
|
228 return deflateInit2_(strm, level, Z_DEFLATED, MAX_WBITS, DEF_MEM_LEVEL, |
|
229 Z_DEFAULT_STRATEGY, version, stream_size); |
|
230 /* To do: ignore strm->next_in if we use it as window */ |
|
231 } |
|
232 |
|
233 /* ========================================================================= */ |
|
234 int ZEXPORT deflateInit2_(strm, level, method, windowBits, memLevel, strategy, |
|
235 version, stream_size) |
|
236 z_streamp strm; |
|
237 int level; |
|
238 int method; |
|
239 int windowBits; |
|
240 int memLevel; |
|
241 int strategy; |
|
242 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 if (strm->zalloc == (alloc_func)0) { |
|
262 strm->zalloc = zcalloc; |
|
263 strm->opaque = (voidpf)0; |
|
264 } |
|
265 if (strm->zfree == (free_func)0) strm->zfree = zcfree; |
|
266 |
|
267 #ifdef FASTEST |
|
268 if (level != 0) level = 1; |
|
269 #else |
|
270 if (level == Z_DEFAULT_COMPRESSION) level = 6; |
|
271 #endif |
|
272 |
|
273 if (windowBits < 0) { /* suppress zlib wrapper */ |
|
274 wrap = 0; |
|
275 windowBits = -windowBits; |
|
276 } |
|
277 #ifdef GZIP |
|
278 else if (windowBits > 15) { |
|
279 wrap = 2; /* write gzip wrapper instead */ |
|
280 windowBits -= 16; |
|
281 } |
|
282 #endif |
|
283 if (memLevel < 1 || memLevel > MAX_MEM_LEVEL || method != Z_DEFLATED || |
|
284 windowBits < 8 || windowBits > 15 || level < 0 || level > 9 || |
|
285 strategy < 0 || strategy > Z_FIXED) { |
|
286 return Z_STREAM_ERROR; |
|
287 } |
|
288 if (windowBits == 8) windowBits = 9; /* until 256-byte window bug fixed */ |
|
289 s = (deflate_state *) ZALLOC(strm, 1, sizeof(deflate_state)); |
|
290 if (s == Z_NULL) return Z_MEM_ERROR; |
|
291 strm->state = (struct internal_state FAR *)s; |
|
292 s->strm = strm; |
|
293 |
|
294 s->wrap = wrap; |
|
295 s->gzhead = Z_NULL; |
|
296 s->w_bits = windowBits; |
|
297 s->w_size = 1 << s->w_bits; |
|
298 s->w_mask = s->w_size - 1; |
|
299 |
|
300 s->hash_bits = memLevel + 7; |
|
301 s->hash_size = 1 << s->hash_bits; |
|
302 s->hash_mask = s->hash_size - 1; |
|
303 s->hash_shift = ((s->hash_bits+MIN_MATCH-1)/MIN_MATCH); |
|
304 |
|
305 s->window = (Bytef *) ZALLOC(strm, s->w_size, 2*sizeof(Byte)); |
|
306 s->prev = (Posf *) ZALLOC(strm, s->w_size, sizeof(Pos)); |
|
307 s->head = (Posf *) ZALLOC(strm, s->hash_size, sizeof(Pos)); |
|
308 |
|
309 s->high_water = 0; /* nothing written to s->window yet */ |
|
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 (strm, dictionary, dictLength) |
|
336 z_streamp strm; |
|
337 const Bytef *dictionary; |
|
338 uInt dictLength; |
|
339 { |
|
340 deflate_state *s; |
|
341 uInt length = dictLength; |
|
342 uInt n; |
|
343 IPos hash_head = 0; |
|
344 |
|
345 if (strm == Z_NULL || strm->state == Z_NULL || dictionary == Z_NULL || |
|
346 strm->state->wrap == 2 || |
|
347 (strm->state->wrap == 1 && strm->state->status != INIT_STATE)) |
|
348 return Z_STREAM_ERROR; |
|
349 |
|
350 s = strm->state; |
|
351 if (s->wrap) |
|
352 strm->adler = adler32(strm->adler, dictionary, dictLength); |
|
353 |
|
354 if (length < MIN_MATCH) return Z_OK; |
|
355 if (length > s->w_size) { |
|
356 length = s->w_size; |
|
357 dictionary += dictLength - length; /* use the tail of the dictionary */ |
|
358 } |
|
359 zmemcpy(s->window, dictionary, length); |
|
360 s->strstart = length; |
|
361 s->block_start = (long)length; |
|
362 |
|
363 /* Insert all strings in the hash table (except for the last two bytes). |
|
364 * s->lookahead stays null, so s->ins_h will be recomputed at the next |
|
365 * call of fill_window. |
|
366 */ |
|
367 s->ins_h = s->window[0]; |
|
368 UPDATE_HASH(s, s->ins_h, s->window[1]); |
|
369 for (n = 0; n <= length - MIN_MATCH; n++) { |
|
370 INSERT_STRING(s, n, hash_head); |
|
371 } |
|
372 if (hash_head) hash_head = 0; /* to make compiler happy */ |
|
373 return Z_OK; |
|
374 } |
|
375 |
|
376 /* ========================================================================= */ |
|
377 int ZEXPORT deflateReset (strm) |
|
378 z_streamp strm; |
|
379 { |
|
380 deflate_state *s; |
|
381 |
|
382 if (strm == Z_NULL || strm->state == Z_NULL || |
|
383 strm->zalloc == (alloc_func)0 || strm->zfree == (free_func)0) { |
|
384 return Z_STREAM_ERROR; |
|
385 } |
|
386 |
|
387 strm->total_in = strm->total_out = 0; |
|
388 strm->msg = Z_NULL; /* use zfree if we ever allocate msg dynamically */ |
|
389 strm->data_type = Z_UNKNOWN; |
|
390 |
|
391 s = (deflate_state *)strm->state; |
|
392 s->pending = 0; |
|
393 s->pending_out = s->pending_buf; |
|
394 |
|
395 if (s->wrap < 0) { |
|
396 s->wrap = -s->wrap; /* was made negative by deflate(..., Z_FINISH); */ |
|
397 } |
|
398 s->status = s->wrap ? INIT_STATE : BUSY_STATE; |
|
399 strm->adler = |
|
400 #ifdef GZIP |
|
401 s->wrap == 2 ? crc32(0L, Z_NULL, 0) : |
|
402 #endif |
|
403 adler32(0L, Z_NULL, 0); |
|
404 s->last_flush = Z_NO_FLUSH; |
|
405 |
|
406 _tr_init(s); |
|
407 lm_init(s); |
|
408 |
|
409 return Z_OK; |
|
410 } |
|
411 |
|
412 /* ========================================================================= */ |
|
413 int ZEXPORT deflateSetHeader (strm, head) |
|
414 z_streamp strm; |
|
415 gz_headerp head; |
|
416 { |
|
417 if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; |
|
418 if (strm->state->wrap != 2) return Z_STREAM_ERROR; |
|
419 strm->state->gzhead = head; |
|
420 return Z_OK; |
|
421 } |
|
422 |
|
423 /* ========================================================================= */ |
|
424 int ZEXPORT deflatePrime (strm, bits, value) |
|
425 z_streamp strm; |
|
426 int bits; |
|
427 int value; |
|
428 { |
|
429 if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; |
|
430 strm->state->bi_valid = bits; |
|
431 strm->state->bi_buf = (ush)(value & ((1 << bits) - 1)); |
|
432 return Z_OK; |
|
433 } |
|
434 |
|
435 /* ========================================================================= */ |
|
436 int ZEXPORT deflateParams(strm, level, strategy) |
|
437 z_streamp strm; |
|
438 int level; |
|
439 int strategy; |
|
440 { |
|
441 deflate_state *s; |
|
442 compress_func func; |
|
443 int err = Z_OK; |
|
444 |
|
445 if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; |
|
446 s = strm->state; |
|
447 |
|
448 #ifdef FASTEST |
|
449 if (level != 0) level = 1; |
|
450 #else |
|
451 if (level == Z_DEFAULT_COMPRESSION) level = 6; |
|
452 #endif |
|
453 if (level < 0 || level > 9 || strategy < 0 || strategy > Z_FIXED) { |
|
454 return Z_STREAM_ERROR; |
|
455 } |
|
456 func = configuration_table[s->level].func; |
|
457 |
|
458 if ((strategy != s->strategy || func != configuration_table[level].func) && |
|
459 strm->total_in != 0) { |
|
460 /* Flush the last buffer: */ |
|
461 err = deflate(strm, Z_BLOCK); |
|
462 } |
|
463 if (s->level != level) { |
|
464 s->level = level; |
|
465 s->max_lazy_match = configuration_table[level].max_lazy; |
|
466 s->good_match = configuration_table[level].good_length; |
|
467 s->nice_match = configuration_table[level].nice_length; |
|
468 s->max_chain_length = configuration_table[level].max_chain; |
|
469 } |
|
470 s->strategy = strategy; |
|
471 return err; |
|
472 } |
|
473 |
|
474 /* ========================================================================= */ |
|
475 int ZEXPORT deflateTune(strm, good_length, max_lazy, nice_length, max_chain) |
|
476 z_streamp strm; |
|
477 int good_length; |
|
478 int max_lazy; |
|
479 int nice_length; |
|
480 int max_chain; |
|
481 { |
|
482 deflate_state *s; |
|
483 |
|
484 if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; |
|
485 s = strm->state; |
|
486 s->good_match = good_length; |
|
487 s->max_lazy_match = max_lazy; |
|
488 s->nice_match = nice_length; |
|
489 s->max_chain_length = max_chain; |
|
490 return Z_OK; |
|
491 } |
|
492 |
|
493 /* ========================================================================= |
|
494 * For the default windowBits of 15 and memLevel of 8, this function returns |
|
495 * a close to exact, as well as small, upper bound on the compressed size. |
|
496 * They are coded as constants here for a reason--if the #define's are |
|
497 * changed, then this function needs to be changed as well. The return |
|
498 * value for 15 and 8 only works for those exact settings. |
|
499 * |
|
500 * For any setting other than those defaults for windowBits and memLevel, |
|
501 * the value returned is a conservative worst case for the maximum expansion |
|
502 * resulting from using fixed blocks instead of stored blocks, which deflate |
|
503 * can emit on compressed data for some combinations of the parameters. |
|
504 * |
|
505 * This function could be more sophisticated to provide closer upper bounds for |
|
506 * every combination of windowBits and memLevel. But even the conservative |
|
507 * upper bound of about 14% expansion does not seem onerous for output buffer |
|
508 * allocation. |
|
509 */ |
|
510 uLong ZEXPORT deflateBound(strm, sourceLen) |
|
511 z_streamp strm; |
|
512 uLong sourceLen; |
|
513 { |
|
514 deflate_state *s; |
|
515 uLong complen, wraplen; |
|
516 Bytef *str; |
|
517 |
|
518 /* conservative upper bound for compressed data */ |
|
519 complen = sourceLen + |
|
520 ((sourceLen + 7) >> 3) + ((sourceLen + 63) >> 6) + 5; |
|
521 |
|
522 /* if can't get parameters, return conservative bound plus zlib wrapper */ |
|
523 if (strm == Z_NULL || strm->state == Z_NULL) |
|
524 return complen + 6; |
|
525 |
|
526 /* compute wrapper length */ |
|
527 s = strm->state; |
|
528 switch (s->wrap) { |
|
529 case 0: /* raw deflate */ |
|
530 wraplen = 0; |
|
531 break; |
|
532 case 1: /* zlib wrapper */ |
|
533 wraplen = 6 + (s->strstart ? 4 : 0); |
|
534 break; |
|
535 case 2: /* gzip wrapper */ |
|
536 wraplen = 18; |
|
537 if (s->gzhead != Z_NULL) { /* user-supplied gzip header */ |
|
538 if (s->gzhead->extra != Z_NULL) |
|
539 wraplen += 2 + s->gzhead->extra_len; |
|
540 str = s->gzhead->name; |
|
541 if (str != Z_NULL) |
|
542 do { |
|
543 wraplen++; |
|
544 } while (*str++); |
|
545 str = s->gzhead->comment; |
|
546 if (str != Z_NULL) |
|
547 do { |
|
548 wraplen++; |
|
549 } while (*str++); |
|
550 if (s->gzhead->hcrc) |
|
551 wraplen += 2; |
|
552 } |
|
553 break; |
|
554 default: /* for compiler happiness */ |
|
555 wraplen = 6; |
|
556 } |
|
557 |
|
558 /* if not default parameters, return conservative bound */ |
|
559 if (s->w_bits != 15 || s->hash_bits != 8 + 7) |
|
560 return complen + wraplen; |
|
561 |
|
562 /* default settings: return tight bound for that case */ |
|
563 return sourceLen + (sourceLen >> 12) + (sourceLen >> 14) + |
|
564 (sourceLen >> 25) + 13 - 6 + wraplen; |
|
565 } |
|
566 |
|
567 /* ========================================================================= |
|
568 * Put a short in the pending buffer. The 16-bit value is put in MSB order. |
|
569 * IN assertion: the stream state is correct and there is enough room in |
|
570 * pending_buf. |
|
571 */ |
|
572 local void putShortMSB (s, b) |
|
573 deflate_state *s; |
|
574 uInt b; |
|
575 { |
|
576 put_byte(s, (Byte)(b >> 8)); |
|
577 put_byte(s, (Byte)(b & 0xff)); |
|
578 } |
|
579 |
|
580 /* ========================================================================= |
|
581 * Flush as much pending output as possible. All deflate() output goes |
|
582 * through this function so some applications may wish to modify it |
|
583 * to avoid allocating a large strm->next_out buffer and copying into it. |
|
584 * (See also read_buf()). |
|
585 */ |
|
586 local void flush_pending(strm) |
|
587 z_streamp strm; |
|
588 { |
|
589 unsigned len = strm->state->pending; |
|
590 |
|
591 if (len > strm->avail_out) len = strm->avail_out; |
|
592 if (len == 0) return; |
|
593 |
|
594 zmemcpy(strm->next_out, strm->state->pending_out, len); |
|
595 strm->next_out += len; |
|
596 strm->state->pending_out += len; |
|
597 strm->total_out += len; |
|
598 strm->avail_out -= len; |
|
599 strm->state->pending -= len; |
|
600 if (strm->state->pending == 0) { |
|
601 strm->state->pending_out = strm->state->pending_buf; |
|
602 } |
|
603 } |
|
604 |
|
605 /* ========================================================================= */ |
|
606 int ZEXPORT deflate (strm, flush) |
|
607 z_streamp strm; |
|
608 int flush; |
|
609 { |
|
610 int old_flush; /* value of flush param for previous deflate call */ |
|
611 deflate_state *s; |
|
612 |
|
613 if (strm == Z_NULL || strm->state == Z_NULL || |
|
614 flush > Z_BLOCK || flush < 0) { |
|
615 return Z_STREAM_ERROR; |
|
616 } |
|
617 s = strm->state; |
|
618 |
|
619 if (strm->next_out == Z_NULL || |
|
620 (strm->next_in == Z_NULL && strm->avail_in != 0) || |
|
621 (s->status == FINISH_STATE && flush != Z_FINISH)) { |
|
622 ERR_RETURN(strm, Z_STREAM_ERROR); |
|
623 } |
|
624 if (strm->avail_out == 0) ERR_RETURN(strm, Z_BUF_ERROR); |
|
625 |
|
626 s->strm = strm; /* just in case */ |
|
627 old_flush = s->last_flush; |
|
628 s->last_flush = flush; |
|
629 |
|
630 /* Write the header */ |
|
631 if (s->status == INIT_STATE) { |
|
632 #ifdef GZIP |
|
633 if (s->wrap == 2) { |
|
634 strm->adler = crc32(0L, Z_NULL, 0); |
|
635 put_byte(s, 31); |
|
636 put_byte(s, 139); |
|
637 put_byte(s, 8); |
|
638 if (s->gzhead == Z_NULL) { |
|
639 put_byte(s, 0); |
|
640 put_byte(s, 0); |
|
641 put_byte(s, 0); |
|
642 put_byte(s, 0); |
|
643 put_byte(s, 0); |
|
644 put_byte(s, s->level == 9 ? 2 : |
|
645 (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ? |
|
646 4 : 0)); |
|
647 put_byte(s, OS_CODE); |
|
648 s->status = BUSY_STATE; |
|
649 } |
|
650 else { |
|
651 put_byte(s, (s->gzhead->text ? 1 : 0) + |
|
652 (s->gzhead->hcrc ? 2 : 0) + |
|
653 (s->gzhead->extra == Z_NULL ? 0 : 4) + |
|
654 (s->gzhead->name == Z_NULL ? 0 : 8) + |
|
655 (s->gzhead->comment == Z_NULL ? 0 : 16) |
|
656 ); |
|
657 put_byte(s, (Byte)(s->gzhead->time & 0xff)); |
|
658 put_byte(s, (Byte)((s->gzhead->time >> 8) & 0xff)); |
|
659 put_byte(s, (Byte)((s->gzhead->time >> 16) & 0xff)); |
|
660 put_byte(s, (Byte)((s->gzhead->time >> 24) & 0xff)); |
|
661 put_byte(s, s->level == 9 ? 2 : |
|
662 (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ? |
|
663 4 : 0)); |
|
664 put_byte(s, s->gzhead->os & 0xff); |
|
665 if (s->gzhead->extra != Z_NULL) { |
|
666 put_byte(s, s->gzhead->extra_len & 0xff); |
|
667 put_byte(s, (s->gzhead->extra_len >> 8) & 0xff); |
|
668 } |
|
669 if (s->gzhead->hcrc) |
|
670 strm->adler = crc32(strm->adler, s->pending_buf, |
|
671 s->pending); |
|
672 s->gzindex = 0; |
|
673 s->status = EXTRA_STATE; |
|
674 } |
|
675 } |
|
676 else |
|
677 #endif |
|
678 { |
|
679 uInt header = (Z_DEFLATED + ((s->w_bits-8)<<4)) << 8; |
|
680 uInt level_flags; |
|
681 |
|
682 if (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2) |
|
683 level_flags = 0; |
|
684 else if (s->level < 6) |
|
685 level_flags = 1; |
|
686 else if (s->level == 6) |
|
687 level_flags = 2; |
|
688 else |
|
689 level_flags = 3; |
|
690 header |= (level_flags << 6); |
|
691 if (s->strstart != 0) header |= PRESET_DICT; |
|
692 header += 31 - (header % 31); |
|
693 |
|
694 s->status = BUSY_STATE; |
|
695 putShortMSB(s, header); |
|
696 |
|
697 /* Save the adler32 of the preset dictionary: */ |
|
698 if (s->strstart != 0) { |
|
699 putShortMSB(s, (uInt)(strm->adler >> 16)); |
|
700 putShortMSB(s, (uInt)(strm->adler & 0xffff)); |
|
701 } |
|
702 strm->adler = adler32(0L, Z_NULL, 0); |
|
703 } |
|
704 } |
|
705 #ifdef GZIP |
|
706 if (s->status == EXTRA_STATE) { |
|
707 if (s->gzhead->extra != Z_NULL) { |
|
708 uInt beg = s->pending; /* start of bytes to update crc */ |
|
709 |
|
710 while (s->gzindex < (s->gzhead->extra_len & 0xffff)) { |
|
711 if (s->pending == s->pending_buf_size) { |
|
712 if (s->gzhead->hcrc && s->pending > beg) |
|
713 strm->adler = crc32(strm->adler, s->pending_buf + beg, |
|
714 s->pending - beg); |
|
715 flush_pending(strm); |
|
716 beg = s->pending; |
|
717 if (s->pending == s->pending_buf_size) |
|
718 break; |
|
719 } |
|
720 put_byte(s, s->gzhead->extra[s->gzindex]); |
|
721 s->gzindex++; |
|
722 } |
|
723 if (s->gzhead->hcrc && s->pending > beg) |
|
724 strm->adler = crc32(strm->adler, s->pending_buf + beg, |
|
725 s->pending - beg); |
|
726 if (s->gzindex == s->gzhead->extra_len) { |
|
727 s->gzindex = 0; |
|
728 s->status = NAME_STATE; |
|
729 } |
|
730 } |
|
731 else |
|
732 s->status = NAME_STATE; |
|
733 } |
|
734 if (s->status == NAME_STATE) { |
|
735 if (s->gzhead->name != Z_NULL) { |
|
736 uInt beg = s->pending; /* start of bytes to update crc */ |
|
737 int val; |
|
738 |
|
739 do { |
|
740 if (s->pending == s->pending_buf_size) { |
|
741 if (s->gzhead->hcrc && s->pending > beg) |
|
742 strm->adler = crc32(strm->adler, s->pending_buf + beg, |
|
743 s->pending - beg); |
|
744 flush_pending(strm); |
|
745 beg = s->pending; |
|
746 if (s->pending == s->pending_buf_size) { |
|
747 val = 1; |
|
748 break; |
|
749 } |
|
750 } |
|
751 val = s->gzhead->name[s->gzindex++]; |
|
752 put_byte(s, val); |
|
753 } while (val != 0); |
|
754 if (s->gzhead->hcrc && s->pending > beg) |
|
755 strm->adler = crc32(strm->adler, s->pending_buf + beg, |
|
756 s->pending - beg); |
|
757 if (val == 0) { |
|
758 s->gzindex = 0; |
|
759 s->status = COMMENT_STATE; |
|
760 } |
|
761 } |
|
762 else |
|
763 s->status = COMMENT_STATE; |
|
764 } |
|
765 if (s->status == COMMENT_STATE) { |
|
766 if (s->gzhead->comment != Z_NULL) { |
|
767 uInt beg = s->pending; /* start of bytes to update crc */ |
|
768 int val; |
|
769 |
|
770 do { |
|
771 if (s->pending == s->pending_buf_size) { |
|
772 if (s->gzhead->hcrc && s->pending > beg) |
|
773 strm->adler = crc32(strm->adler, s->pending_buf + beg, |
|
774 s->pending - beg); |
|
775 flush_pending(strm); |
|
776 beg = s->pending; |
|
777 if (s->pending == s->pending_buf_size) { |
|
778 val = 1; |
|
779 break; |
|
780 } |
|
781 } |
|
782 val = s->gzhead->comment[s->gzindex++]; |
|
783 put_byte(s, val); |
|
784 } while (val != 0); |
|
785 if (s->gzhead->hcrc && s->pending > beg) |
|
786 strm->adler = crc32(strm->adler, s->pending_buf + beg, |
|
787 s->pending - beg); |
|
788 if (val == 0) |
|
789 s->status = HCRC_STATE; |
|
790 } |
|
791 else |
|
792 s->status = HCRC_STATE; |
|
793 } |
|
794 if (s->status == HCRC_STATE) { |
|
795 if (s->gzhead->hcrc) { |
|
796 if (s->pending + 2 > s->pending_buf_size) |
|
797 flush_pending(strm); |
|
798 if (s->pending + 2 <= s->pending_buf_size) { |
|
799 put_byte(s, (Byte)(strm->adler & 0xff)); |
|
800 put_byte(s, (Byte)((strm->adler >> 8) & 0xff)); |
|
801 strm->adler = crc32(0L, Z_NULL, 0); |
|
802 s->status = BUSY_STATE; |
|
803 } |
|
804 } |
|
805 else |
|
806 s->status = BUSY_STATE; |
|
807 } |
|
808 #endif |
|
809 |
|
810 /* Flush as much pending output as possible */ |
|
811 if (s->pending != 0) { |
|
812 flush_pending(strm); |
|
813 if (strm->avail_out == 0) { |
|
814 /* Since avail_out is 0, deflate will be called again with |
|
815 * more output space, but possibly with both pending and |
|
816 * avail_in equal to zero. There won't be anything to do, |
|
817 * but this is not an error situation so make sure we |
|
818 * return OK instead of BUF_ERROR at next call of deflate: |
|
819 */ |
|
820 s->last_flush = -1; |
|
821 return Z_OK; |
|
822 } |
|
823 |
|
824 /* Make sure there is something to do and avoid duplicate consecutive |
|
825 * flushes. For repeated and useless calls with Z_FINISH, we keep |
|
826 * returning Z_STREAM_END instead of Z_BUF_ERROR. |
|
827 */ |
|
828 } else if (strm->avail_in == 0 && flush <= old_flush && |
|
829 flush != Z_FINISH) { |
|
830 ERR_RETURN(strm, Z_BUF_ERROR); |
|
831 } |
|
832 |
|
833 /* User must not provide more input after the first FINISH: */ |
|
834 if (s->status == FINISH_STATE && strm->avail_in != 0) { |
|
835 ERR_RETURN(strm, Z_BUF_ERROR); |
|
836 } |
|
837 |
|
838 /* Start a new block or continue the current one. |
|
839 */ |
|
840 if (strm->avail_in != 0 || s->lookahead != 0 || |
|
841 (flush != Z_NO_FLUSH && s->status != FINISH_STATE)) { |
|
842 block_state bstate; |
|
843 |
|
844 bstate = s->strategy == Z_HUFFMAN_ONLY ? deflate_huff(s, flush) : |
|
845 (s->strategy == Z_RLE ? deflate_rle(s, flush) : |
|
846 (*(configuration_table[s->level].func))(s, flush)); |
|
847 |
|
848 if (bstate == finish_started || bstate == finish_done) { |
|
849 s->status = FINISH_STATE; |
|
850 } |
|
851 if (bstate == need_more || bstate == finish_started) { |
|
852 if (strm->avail_out == 0) { |
|
853 s->last_flush = -1; /* avoid BUF_ERROR next call, see above */ |
|
854 } |
|
855 return Z_OK; |
|
856 /* If flush != Z_NO_FLUSH && avail_out == 0, the next call |
|
857 * of deflate should use the same flush parameter to make sure |
|
858 * that the flush is complete. So we don't have to output an |
|
859 * empty block here, this will be done at next call. This also |
|
860 * ensures that for a very small output buffer, we emit at most |
|
861 * one empty block. |
|
862 */ |
|
863 } |
|
864 if (bstate == block_done) { |
|
865 if (flush == Z_PARTIAL_FLUSH) { |
|
866 _tr_align(s); |
|
867 } else if (flush != Z_BLOCK) { /* FULL_FLUSH or SYNC_FLUSH */ |
|
868 _tr_stored_block(s, (char*)0, 0L, 0); |
|
869 /* For a full flush, this empty block will be recognized |
|
870 * as a special marker by inflate_sync(). |
|
871 */ |
|
872 if (flush == Z_FULL_FLUSH) { |
|
873 CLEAR_HASH(s); /* forget history */ |
|
874 if (s->lookahead == 0) { |
|
875 s->strstart = 0; |
|
876 s->block_start = 0L; |
|
877 } |
|
878 } |
|
879 } |
|
880 flush_pending(strm); |
|
881 if (strm->avail_out == 0) { |
|
882 s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */ |
|
883 return Z_OK; |
|
884 } |
|
885 } |
|
886 } |
|
887 Assert(strm->avail_out > 0, "bug2"); |
|
888 |
|
889 if (flush != Z_FINISH) return Z_OK; |
|
890 if (s->wrap <= 0) return Z_STREAM_END; |
|
891 |
|
892 /* Write the trailer */ |
|
893 #ifdef GZIP |
|
894 if (s->wrap == 2) { |
|
895 put_byte(s, (Byte)(strm->adler & 0xff)); |
|
896 put_byte(s, (Byte)((strm->adler >> 8) & 0xff)); |
|
897 put_byte(s, (Byte)((strm->adler >> 16) & 0xff)); |
|
898 put_byte(s, (Byte)((strm->adler >> 24) & 0xff)); |
|
899 put_byte(s, (Byte)(strm->total_in & 0xff)); |
|
900 put_byte(s, (Byte)((strm->total_in >> 8) & 0xff)); |
|
901 put_byte(s, (Byte)((strm->total_in >> 16) & 0xff)); |
|
902 put_byte(s, (Byte)((strm->total_in >> 24) & 0xff)); |
|
903 } |
|
904 else |
|
905 #endif |
|
906 { |
|
907 putShortMSB(s, (uInt)(strm->adler >> 16)); |
|
908 putShortMSB(s, (uInt)(strm->adler & 0xffff)); |
|
909 } |
|
910 flush_pending(strm); |
|
911 /* If avail_out is zero, the application will call deflate again |
|
912 * to flush the rest. |
|
913 */ |
|
914 if (s->wrap > 0) s->wrap = -s->wrap; /* write the trailer only once! */ |
|
915 return s->pending != 0 ? Z_OK : Z_STREAM_END; |
|
916 } |
|
917 |
|
918 /* ========================================================================= */ |
|
919 int ZEXPORT deflateEnd (strm) |
|
920 z_streamp strm; |
|
921 { |
|
922 int status; |
|
923 |
|
924 if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; |
|
925 |
|
926 status = strm->state->status; |
|
927 if (status != INIT_STATE && |
|
928 status != EXTRA_STATE && |
|
929 status != NAME_STATE && |
|
930 status != COMMENT_STATE && |
|
931 status != HCRC_STATE && |
|
932 status != BUSY_STATE && |
|
933 status != FINISH_STATE) { |
|
934 return Z_STREAM_ERROR; |
|
935 } |
|
936 |
|
937 /* Deallocate in reverse order of allocations: */ |
|
938 TRY_FREE(strm, strm->state->pending_buf); |
|
939 TRY_FREE(strm, strm->state->head); |
|
940 TRY_FREE(strm, strm->state->prev); |
|
941 TRY_FREE(strm, strm->state->window); |
|
942 |
|
943 ZFREE(strm, strm->state); |
|
944 strm->state = Z_NULL; |
|
945 |
|
946 return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK; |
|
947 } |
|
948 |
|
949 /* ========================================================================= |
|
950 * Copy the source state to the destination state. |
|
951 * To simplify the source, this is not supported for 16-bit MSDOS (which |
|
952 * doesn't have enough memory anyway to duplicate compression states). |
|
953 */ |
|
954 int ZEXPORT deflateCopy (dest, source) |
|
955 z_streamp dest; |
|
956 z_streamp source; |
|
957 { |
|
958 #ifdef MAXSEG_64K |
|
959 return Z_STREAM_ERROR; |
|
960 #else |
|
961 deflate_state *ds; |
|
962 deflate_state *ss; |
|
963 ushf *overlay; |
|
964 |
|
965 |
|
966 if (source == Z_NULL || dest == Z_NULL || source->state == Z_NULL) { |
|
967 return Z_STREAM_ERROR; |
|
968 } |
|
969 |
|
970 ss = source->state; |
|
971 |
|
972 zmemcpy(dest, source, sizeof(z_stream)); |
|
973 |
|
974 ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state)); |
|
975 if (ds == Z_NULL) return Z_MEM_ERROR; |
|
976 dest->state = (struct internal_state FAR *) ds; |
|
977 zmemcpy(ds, ss, sizeof(deflate_state)); |
|
978 ds->strm = dest; |
|
979 |
|
980 ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte)); |
|
981 ds->prev = (Posf *) ZALLOC(dest, ds->w_size, sizeof(Pos)); |
|
982 ds->head = (Posf *) ZALLOC(dest, ds->hash_size, sizeof(Pos)); |
|
983 overlay = (ushf *) ZALLOC(dest, ds->lit_bufsize, sizeof(ush)+2); |
|
984 ds->pending_buf = (uchf *) overlay; |
|
985 |
|
986 if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL || |
|
987 ds->pending_buf == Z_NULL) { |
|
988 deflateEnd (dest); |
|
989 return Z_MEM_ERROR; |
|
990 } |
|
991 /* following zmemcpy do not work for 16-bit MSDOS */ |
|
992 zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte)); |
|
993 zmemcpy(ds->prev, ss->prev, ds->w_size * sizeof(Pos)); |
|
994 zmemcpy(ds->head, ss->head, ds->hash_size * sizeof(Pos)); |
|
995 zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size); |
|
996 |
|
997 ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf); |
|
998 ds->d_buf = overlay + ds->lit_bufsize/sizeof(ush); |
|
999 ds->l_buf = ds->pending_buf + (1+sizeof(ush))*ds->lit_bufsize; |
|
1000 |
|
1001 ds->l_desc.dyn_tree = ds->dyn_ltree; |
|
1002 ds->d_desc.dyn_tree = ds->dyn_dtree; |
|
1003 ds->bl_desc.dyn_tree = ds->bl_tree; |
|
1004 |
|
1005 return Z_OK; |
|
1006 #endif /* MAXSEG_64K */ |
|
1007 } |
|
1008 |
|
1009 /* =========================================================================== |
|
1010 * Read a new buffer from the current input stream, update the adler32 |
|
1011 * and total number of bytes read. All deflate() input goes through |
|
1012 * this function so some applications may wish to modify it to avoid |
|
1013 * allocating a large strm->next_in buffer and copying from it. |
|
1014 * (See also flush_pending()). |
|
1015 */ |
|
1016 local int read_buf(strm, buf, size) |
|
1017 z_streamp strm; |
|
1018 Bytef *buf; |
|
1019 unsigned size; |
|
1020 { |
|
1021 unsigned len = strm->avail_in; |
|
1022 |
|
1023 if (len > size) len = size; |
|
1024 if (len == 0) return 0; |
|
1025 |
|
1026 strm->avail_in -= len; |
|
1027 |
|
1028 if (strm->state->wrap == 1) { |
|
1029 strm->adler = adler32(strm->adler, strm->next_in, len); |
|
1030 } |
|
1031 #ifdef GZIP |
|
1032 else if (strm->state->wrap == 2) { |
|
1033 strm->adler = crc32(strm->adler, strm->next_in, len); |
|
1034 } |
|
1035 #endif |
|
1036 zmemcpy(buf, strm->next_in, len); |
|
1037 strm->next_in += len; |
|
1038 strm->total_in += len; |
|
1039 |
|
1040 return (int)len; |
|
1041 } |
|
1042 |
|
1043 /* =========================================================================== |
|
1044 * Initialize the "longest match" routines for a new zlib stream |
|
1045 */ |
|
1046 local void lm_init (s) |
|
1047 deflate_state *s; |
|
1048 { |
|
1049 s->window_size = (ulg)2L*s->w_size; |
|
1050 |
|
1051 CLEAR_HASH(s); |
|
1052 |
|
1053 /* Set the default configuration parameters: |
|
1054 */ |
|
1055 s->max_lazy_match = configuration_table[s->level].max_lazy; |
|
1056 s->good_match = configuration_table[s->level].good_length; |
|
1057 s->nice_match = configuration_table[s->level].nice_length; |
|
1058 s->max_chain_length = configuration_table[s->level].max_chain; |
|
1059 |
|
1060 s->strstart = 0; |
|
1061 s->block_start = 0L; |
|
1062 s->lookahead = 0; |
|
1063 s->match_length = s->prev_length = MIN_MATCH-1; |
|
1064 s->match_available = 0; |
|
1065 s->ins_h = 0; |
|
1066 #ifndef FASTEST |
|
1067 #ifdef ASMV |
|
1068 match_init(); /* initialize the asm code */ |
|
1069 #endif |
|
1070 #endif |
|
1071 } |
|
1072 |
|
1073 #ifndef FASTEST |
|
1074 /* =========================================================================== |
|
1075 * Set match_start to the longest match starting at the given string and |
|
1076 * return its length. Matches shorter or equal to prev_length are discarded, |
|
1077 * in which case the result is equal to prev_length and match_start is |
|
1078 * garbage. |
|
1079 * IN assertions: cur_match is the head of the hash chain for the current |
|
1080 * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1 |
|
1081 * OUT assertion: the match length is not greater than s->lookahead. |
|
1082 */ |
|
1083 #ifndef ASMV |
|
1084 /* For 80x86 and 680x0, an optimized version will be provided in match.asm or |
|
1085 * match.S. The code will be functionally equivalent. |
|
1086 */ |
|
1087 local uInt longest_match(s, cur_match) |
|
1088 deflate_state *s; |
|
1089 IPos cur_match; /* current match */ |
|
1090 { |
|
1091 unsigned chain_length = s->max_chain_length;/* max hash chain length */ |
|
1092 register Bytef *scan = s->window + s->strstart; /* current string */ |
|
1093 register Bytef *match; /* matched string */ |
|
1094 register int len; /* length of current match */ |
|
1095 int best_len = s->prev_length; /* best match length so far */ |
|
1096 int nice_match = s->nice_match; /* stop if match long enough */ |
|
1097 IPos limit = s->strstart > (IPos)MAX_DIST(s) ? |
|
1098 s->strstart - (IPos)MAX_DIST(s) : NIL; |
|
1099 /* Stop when cur_match becomes <= limit. To simplify the code, |
|
1100 * we prevent matches with the string of window index 0. |
|
1101 */ |
|
1102 Posf *prev = s->prev; |
|
1103 uInt wmask = s->w_mask; |
|
1104 |
|
1105 #ifdef UNALIGNED_OK |
|
1106 /* Compare two bytes at a time. Note: this is not always beneficial. |
|
1107 * Try with and without -DUNALIGNED_OK to check. |
|
1108 */ |
|
1109 register Bytef *strend = s->window + s->strstart + MAX_MATCH - 1; |
|
1110 register ush scan_start = *(ushf*)scan; |
|
1111 register ush scan_end = *(ushf*)(scan+best_len-1); |
|
1112 #else |
|
1113 register Bytef *strend = s->window + s->strstart + MAX_MATCH; |
|
1114 register Byte scan_end1 = scan[best_len-1]; |
|
1115 register Byte scan_end = scan[best_len]; |
|
1116 #endif |
|
1117 |
|
1118 /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16. |
|
1119 * It is easy to get rid of this optimization if necessary. |
|
1120 */ |
|
1121 Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever"); |
|
1122 |
|
1123 /* Do not waste too much time if we already have a good match: */ |
|
1124 if (s->prev_length >= s->good_match) { |
|
1125 chain_length >>= 2; |
|
1126 } |
|
1127 /* Do not look for matches beyond the end of the input. This is necessary |
|
1128 * to make deflate deterministic. |
|
1129 */ |
|
1130 if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead; |
|
1131 |
|
1132 Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead"); |
|
1133 |
|
1134 do { |
|
1135 Assert(cur_match < s->strstart, "no future"); |
|
1136 match = s->window + cur_match; |
|
1137 |
|
1138 /* Skip to next match if the match length cannot increase |
|
1139 * or if the match length is less than 2. Note that the checks below |
|
1140 * for insufficient lookahead only occur occasionally for performance |
|
1141 * reasons. Therefore uninitialized memory will be accessed, and |
|
1142 * conditional jumps will be made that depend on those values. |
|
1143 * However the length of the match is limited to the lookahead, so |
|
1144 * the output of deflate is not affected by the uninitialized values. |
|
1145 */ |
|
1146 #if (defined(UNALIGNED_OK) && MAX_MATCH == 258) |
|
1147 /* This code assumes sizeof(unsigned short) == 2. Do not use |
|
1148 * UNALIGNED_OK if your compiler uses a different size. |
|
1149 */ |
|
1150 if (*(ushf*)(match+best_len-1) != scan_end || |
|
1151 *(ushf*)match != scan_start) continue; |
|
1152 |
|
1153 /* It is not necessary to compare scan[2] and match[2] since they are |
|
1154 * always equal when the other bytes match, given that the hash keys |
|
1155 * are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at |
|
1156 * strstart+3, +5, ... up to strstart+257. We check for insufficient |
|
1157 * lookahead only every 4th comparison; the 128th check will be made |
|
1158 * at strstart+257. If MAX_MATCH-2 is not a multiple of 8, it is |
|
1159 * necessary to put more guard bytes at the end of the window, or |
|
1160 * to check more often for insufficient lookahead. |
|
1161 */ |
|
1162 Assert(scan[2] == match[2], "scan[2]?"); |
|
1163 scan++, match++; |
|
1164 do { |
|
1165 } while (*(ushf*)(scan+=2) == *(ushf*)(match+=2) && |
|
1166 *(ushf*)(scan+=2) == *(ushf*)(match+=2) && |
|
1167 *(ushf*)(scan+=2) == *(ushf*)(match+=2) && |
|
1168 *(ushf*)(scan+=2) == *(ushf*)(match+=2) && |
|
1169 scan < strend); |
|
1170 /* The funny "do {}" generates better code on most compilers */ |
|
1171 |
|
1172 /* Here, scan <= window+strstart+257 */ |
|
1173 Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan"); |
|
1174 if (*scan == *match) scan++; |
|
1175 |
|
1176 len = (MAX_MATCH - 1) - (int)(strend-scan); |
|
1177 scan = strend - (MAX_MATCH-1); |
|
1178 |
|
1179 #else /* UNALIGNED_OK */ |
|
1180 |
|
1181 if (match[best_len] != scan_end || |
|
1182 match[best_len-1] != scan_end1 || |
|
1183 *match != *scan || |
|
1184 *++match != scan[1]) continue; |
|
1185 |
|
1186 /* The check at best_len-1 can be removed because it will be made |
|
1187 * again later. (This heuristic is not always a win.) |
|
1188 * It is not necessary to compare scan[2] and match[2] since they |
|
1189 * are always equal when the other bytes match, given that |
|
1190 * the hash keys are equal and that HASH_BITS >= 8. |
|
1191 */ |
|
1192 scan += 2, match++; |
|
1193 Assert(*scan == *match, "match[2]?"); |
|
1194 |
|
1195 /* We check for insufficient lookahead only every 8th comparison; |
|
1196 * the 256th check will be made at strstart+258. |
|
1197 */ |
|
1198 do { |
|
1199 } while (*++scan == *++match && *++scan == *++match && |
|
1200 *++scan == *++match && *++scan == *++match && |
|
1201 *++scan == *++match && *++scan == *++match && |
|
1202 *++scan == *++match && *++scan == *++match && |
|
1203 scan < strend); |
|
1204 |
|
1205 Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan"); |
|
1206 |
|
1207 len = MAX_MATCH - (int)(strend - scan); |
|
1208 scan = strend - MAX_MATCH; |
|
1209 |
|
1210 #endif /* UNALIGNED_OK */ |
|
1211 |
|
1212 if (len > best_len) { |
|
1213 s->match_start = cur_match; |
|
1214 best_len = len; |
|
1215 if (len >= nice_match) break; |
|
1216 #ifdef UNALIGNED_OK |
|
1217 scan_end = *(ushf*)(scan+best_len-1); |
|
1218 #else |
|
1219 scan_end1 = scan[best_len-1]; |
|
1220 scan_end = scan[best_len]; |
|
1221 #endif |
|
1222 } |
|
1223 } while ((cur_match = prev[cur_match & wmask]) > limit |
|
1224 && --chain_length != 0); |
|
1225 |
|
1226 if ((uInt)best_len <= s->lookahead) return (uInt)best_len; |
|
1227 return s->lookahead; |
|
1228 } |
|
1229 #endif /* ASMV */ |
|
1230 |
|
1231 #else /* FASTEST */ |
|
1232 |
|
1233 /* --------------------------------------------------------------------------- |
|
1234 * Optimized version for FASTEST only |
|
1235 */ |
|
1236 local uInt longest_match(s, cur_match) |
|
1237 deflate_state *s; |
|
1238 IPos cur_match; /* current match */ |
|
1239 { |
|
1240 register Bytef *scan = s->window + s->strstart; /* current string */ |
|
1241 register Bytef *match; /* matched string */ |
|
1242 register int len; /* length of current match */ |
|
1243 register Bytef *strend = s->window + s->strstart + MAX_MATCH; |
|
1244 |
|
1245 /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16. |
|
1246 * It is easy to get rid of this optimization if necessary. |
|
1247 */ |
|
1248 Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever"); |
|
1249 |
|
1250 Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead"); |
|
1251 |
|
1252 Assert(cur_match < s->strstart, "no future"); |
|
1253 |
|
1254 match = s->window + cur_match; |
|
1255 |
|
1256 /* Return failure if the match length is less than 2: |
|
1257 */ |
|
1258 if (match[0] != scan[0] || match[1] != scan[1]) return MIN_MATCH-1; |
|
1259 |
|
1260 /* The check at best_len-1 can be removed because it will be made |
|
1261 * again later. (This heuristic is not always a win.) |
|
1262 * It is not necessary to compare scan[2] and match[2] since they |
|
1263 * are always equal when the other bytes match, given that |
|
1264 * the hash keys are equal and that HASH_BITS >= 8. |
|
1265 */ |
|
1266 scan += 2, match += 2; |
|
1267 Assert(*scan == *match, "match[2]?"); |
|
1268 |
|
1269 /* We check for insufficient lookahead only every 8th comparison; |
|
1270 * the 256th check will be made at strstart+258. |
|
1271 */ |
|
1272 do { |
|
1273 } while (*++scan == *++match && *++scan == *++match && |
|
1274 *++scan == *++match && *++scan == *++match && |
|
1275 *++scan == *++match && *++scan == *++match && |
|
1276 *++scan == *++match && *++scan == *++match && |
|
1277 scan < strend); |
|
1278 |
|
1279 Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan"); |
|
1280 |
|
1281 len = MAX_MATCH - (int)(strend - scan); |
|
1282 |
|
1283 if (len < MIN_MATCH) return MIN_MATCH - 1; |
|
1284 |
|
1285 s->match_start = cur_match; |
|
1286 return (uInt)len <= s->lookahead ? (uInt)len : s->lookahead; |
|
1287 } |
|
1288 |
|
1289 #endif /* FASTEST */ |
|
1290 |
|
1291 #ifdef DEBUG |
|
1292 /* =========================================================================== |
|
1293 * Check that the match at match_start is indeed a match. |
|
1294 */ |
|
1295 local void check_match(s, start, match, length) |
|
1296 deflate_state *s; |
|
1297 IPos start, match; |
|
1298 int length; |
|
1299 { |
|
1300 /* check that the match is indeed a match */ |
|
1301 if (zmemcmp(s->window + match, |
|
1302 s->window + start, length) != EQUAL) { |
|
1303 fprintf(stderr, " start %u, match %u, length %d\n", |
|
1304 start, match, length); |
|
1305 do { |
|
1306 fprintf(stderr, "%c%c", s->window[match++], s->window[start++]); |
|
1307 } while (--length != 0); |
|
1308 z_error("invalid match"); |
|
1309 } |
|
1310 if (z_verbose > 1) { |
|
1311 fprintf(stderr,"\\[%d,%d]", start-match, length); |
|
1312 do { putc(s->window[start++], stderr); } while (--length != 0); |
|
1313 } |
|
1314 } |
|
1315 #else |
|
1316 # define check_match(s, start, match, length) |
|
1317 #endif /* DEBUG */ |
|
1318 |
|
1319 /* =========================================================================== |
|
1320 * Fill the window when the lookahead becomes insufficient. |
|
1321 * Updates strstart and lookahead. |
|
1322 * |
|
1323 * IN assertion: lookahead < MIN_LOOKAHEAD |
|
1324 * OUT assertions: strstart <= window_size-MIN_LOOKAHEAD |
|
1325 * At least one byte has been read, or avail_in == 0; reads are |
|
1326 * performed for at least two bytes (required for the zip translate_eol |
|
1327 * option -- not supported here). |
|
1328 */ |
|
1329 local void fill_window(s) |
|
1330 deflate_state *s; |
|
1331 { |
|
1332 register unsigned n, m; |
|
1333 register Posf *p; |
|
1334 unsigned more; /* Amount of free space at the end of the window. */ |
|
1335 uInt wsize = s->w_size; |
|
1336 |
|
1337 do { |
|
1338 more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart); |
|
1339 |
|
1340 /* Deal with !@#$% 64K limit: */ |
|
1341 if (sizeof(int) <= 2) { |
|
1342 if (more == 0 && s->strstart == 0 && s->lookahead == 0) { |
|
1343 more = wsize; |
|
1344 |
|
1345 } else if (more == (unsigned)(-1)) { |
|
1346 /* Very unlikely, but possible on 16 bit machine if |
|
1347 * strstart == 0 && lookahead == 1 (input done a byte at time) |
|
1348 */ |
|
1349 more--; |
|
1350 } |
|
1351 } |
|
1352 |
|
1353 /* If the window is almost full and there is insufficient lookahead, |
|
1354 * move the upper half to the lower one to make room in the upper half. |
|
1355 */ |
|
1356 if (s->strstart >= wsize+MAX_DIST(s)) { |
|
1357 |
|
1358 zmemcpy(s->window, s->window+wsize, (unsigned)wsize); |
|
1359 s->match_start -= wsize; |
|
1360 s->strstart -= wsize; /* we now have strstart >= MAX_DIST */ |
|
1361 s->block_start -= (long) wsize; |
|
1362 |
|
1363 /* Slide the hash table (could be avoided with 32 bit values |
|
1364 at the expense of memory usage). We slide even when level == 0 |
|
1365 to keep the hash table consistent if we switch back to level > 0 |
|
1366 later. (Using level 0 permanently is not an optimal usage of |
|
1367 zlib, so we don't care about this pathological case.) |
|
1368 */ |
|
1369 n = s->hash_size; |
|
1370 p = &s->head[n]; |
|
1371 do { |
|
1372 m = *--p; |
|
1373 *p = (Pos)(m >= wsize ? m-wsize : NIL); |
|
1374 } while (--n); |
|
1375 |
|
1376 n = wsize; |
|
1377 #ifndef FASTEST |
|
1378 p = &s->prev[n]; |
|
1379 do { |
|
1380 m = *--p; |
|
1381 *p = (Pos)(m >= wsize ? m-wsize : NIL); |
|
1382 /* If n is not on any hash chain, prev[n] is garbage but |
|
1383 * its value will never be used. |
|
1384 */ |
|
1385 } while (--n); |
|
1386 #endif |
|
1387 more += wsize; |
|
1388 } |
|
1389 if (s->strm->avail_in == 0) return; |
|
1390 |
|
1391 /* If there was no sliding: |
|
1392 * strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 && |
|
1393 * more == window_size - lookahead - strstart |
|
1394 * => more >= window_size - (MIN_LOOKAHEAD-1 + WSIZE + MAX_DIST-1) |
|
1395 * => more >= window_size - 2*WSIZE + 2 |
|
1396 * In the BIG_MEM or MMAP case (not yet supported), |
|
1397 * window_size == input_size + MIN_LOOKAHEAD && |
|
1398 * strstart + s->lookahead <= input_size => more >= MIN_LOOKAHEAD. |
|
1399 * Otherwise, window_size == 2*WSIZE so more >= 2. |
|
1400 * If there was sliding, more >= WSIZE. So in all cases, more >= 2. |
|
1401 */ |
|
1402 Assert(more >= 2, "more < 2"); |
|
1403 |
|
1404 n = read_buf(s->strm, s->window + s->strstart + s->lookahead, more); |
|
1405 s->lookahead += n; |
|
1406 |
|
1407 /* Initialize the hash value now that we have some input: */ |
|
1408 if (s->lookahead >= MIN_MATCH) { |
|
1409 s->ins_h = s->window[s->strstart]; |
|
1410 UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]); |
|
1411 #if MIN_MATCH != 3 |
|
1412 Call UPDATE_HASH() MIN_MATCH-3 more times |
|
1413 #endif |
|
1414 } |
|
1415 /* If the whole input has less than MIN_MATCH bytes, ins_h is garbage, |
|
1416 * but this is not important since only literal bytes will be emitted. |
|
1417 */ |
|
1418 |
|
1419 } while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0); |
|
1420 |
|
1421 /* If the WIN_INIT bytes after the end of the current data have never been |
|
1422 * written, then zero those bytes in order to avoid memory check reports of |
|
1423 * the use of uninitialized (or uninitialised as Julian writes) bytes by |
|
1424 * the longest match routines. Update the high water mark for the next |
|
1425 * time through here. WIN_INIT is set to MAX_MATCH since the longest match |
|
1426 * routines allow scanning to strstart + MAX_MATCH, ignoring lookahead. |
|
1427 */ |
|
1428 if (s->high_water < s->window_size) { |
|
1429 ulg curr = s->strstart + (ulg)(s->lookahead); |
|
1430 ulg init; |
|
1431 |
|
1432 if (s->high_water < curr) { |
|
1433 /* Previous high water mark below current data -- zero WIN_INIT |
|
1434 * bytes or up to end of window, whichever is less. |
|
1435 */ |
|
1436 init = s->window_size - curr; |
|
1437 if (init > WIN_INIT) |
|
1438 init = WIN_INIT; |
|
1439 zmemzero(s->window + curr, (unsigned)init); |
|
1440 s->high_water = curr + init; |
|
1441 } |
|
1442 else if (s->high_water < (ulg)curr + WIN_INIT) { |
|
1443 /* High water mark at or above current data, but below current data |
|
1444 * plus WIN_INIT -- zero out to current data plus WIN_INIT, or up |
|
1445 * to end of window, whichever is less. |
|
1446 */ |
|
1447 init = (ulg)curr + WIN_INIT - s->high_water; |
|
1448 if (init > s->window_size - s->high_water) |
|
1449 init = s->window_size - s->high_water; |
|
1450 zmemzero(s->window + s->high_water, (unsigned)init); |
|
1451 s->high_water += init; |
|
1452 } |
|
1453 } |
|
1454 } |
|
1455 |
|
1456 /* =========================================================================== |
|
1457 * Flush the current block, with given end-of-file flag. |
|
1458 * IN assertion: strstart is set to the end of the current match. |
|
1459 */ |
|
1460 #define FLUSH_BLOCK_ONLY(s, last) { \ |
|
1461 _tr_flush_block(s, (s->block_start >= 0L ? \ |
|
1462 (charf *)&s->window[(unsigned)s->block_start] : \ |
|
1463 (charf *)Z_NULL), \ |
|
1464 (ulg)((long)s->strstart - s->block_start), \ |
|
1465 (last)); \ |
|
1466 s->block_start = s->strstart; \ |
|
1467 flush_pending(s->strm); \ |
|
1468 Tracev((stderr,"[FLUSH]")); \ |
|
1469 } |
|
1470 |
|
1471 /* Same but force premature exit if necessary. */ |
|
1472 #define FLUSH_BLOCK(s, last) { \ |
|
1473 FLUSH_BLOCK_ONLY(s, last); \ |
|
1474 if (s->strm->avail_out == 0) return (last) ? finish_started : need_more; \ |
|
1475 } |
|
1476 |
|
1477 /* =========================================================================== |
|
1478 * Copy without compression as much as possible from the input stream, return |
|
1479 * the current block state. |
|
1480 * This function does not insert new strings in the dictionary since |
|
1481 * uncompressible data is probably not useful. This function is used |
|
1482 * only for the level=0 compression option. |
|
1483 * NOTE: this function should be optimized to avoid extra copying from |
|
1484 * window to pending_buf. |
|
1485 */ |
|
1486 local block_state deflate_stored(s, flush) |
|
1487 deflate_state *s; |
|
1488 int flush; |
|
1489 { |
|
1490 /* Stored blocks are limited to 0xffff bytes, pending_buf is limited |
|
1491 * to pending_buf_size, and each stored block has a 5 byte header: |
|
1492 */ |
|
1493 ulg max_block_size = 0xffff; |
|
1494 ulg max_start; |
|
1495 |
|
1496 if (max_block_size > s->pending_buf_size - 5) { |
|
1497 max_block_size = s->pending_buf_size - 5; |
|
1498 } |
|
1499 |
|
1500 /* Copy as much as possible from input to output: */ |
|
1501 for (;;) { |
|
1502 /* Fill the window as much as possible: */ |
|
1503 if (s->lookahead <= 1) { |
|
1504 |
|
1505 Assert(s->strstart < s->w_size+MAX_DIST(s) || |
|
1506 s->block_start >= (long)s->w_size, "slide too late"); |
|
1507 |
|
1508 fill_window(s); |
|
1509 if (s->lookahead == 0 && flush == Z_NO_FLUSH) return need_more; |
|
1510 |
|
1511 if (s->lookahead == 0) break; /* flush the current block */ |
|
1512 } |
|
1513 Assert(s->block_start >= 0L, "block gone"); |
|
1514 |
|
1515 s->strstart += s->lookahead; |
|
1516 s->lookahead = 0; |
|
1517 |
|
1518 /* Emit a stored block if pending_buf will be full: */ |
|
1519 max_start = s->block_start + max_block_size; |
|
1520 if (s->strstart == 0 || (ulg)s->strstart >= max_start) { |
|
1521 /* strstart == 0 is possible when wraparound on 16-bit machine */ |
|
1522 s->lookahead = (uInt)(s->strstart - max_start); |
|
1523 s->strstart = (uInt)max_start; |
|
1524 FLUSH_BLOCK(s, 0); |
|
1525 } |
|
1526 /* Flush if we may have to slide, otherwise block_start may become |
|
1527 * negative and the data will be gone: |
|
1528 */ |
|
1529 if (s->strstart - (uInt)s->block_start >= MAX_DIST(s)) { |
|
1530 FLUSH_BLOCK(s, 0); |
|
1531 } |
|
1532 } |
|
1533 FLUSH_BLOCK(s, flush == Z_FINISH); |
|
1534 return flush == Z_FINISH ? finish_done : block_done; |
|
1535 } |
|
1536 |
|
1537 /* =========================================================================== |
|
1538 * Compress as much as possible from the input stream, return the current |
|
1539 * block state. |
|
1540 * This function does not perform lazy evaluation of matches and inserts |
|
1541 * new strings in the dictionary only for unmatched strings or for short |
|
1542 * matches. It is used only for the fast compression options. |
|
1543 */ |
|
1544 local block_state deflate_fast(s, flush) |
|
1545 deflate_state *s; |
|
1546 int flush; |
|
1547 { |
|
1548 IPos hash_head; /* head of the hash chain */ |
|
1549 int bflush; /* set if current block must be flushed */ |
|
1550 |
|
1551 for (;;) { |
|
1552 /* Make sure that we always have enough lookahead, except |
|
1553 * at the end of the input file. We need MAX_MATCH bytes |
|
1554 * for the next match, plus MIN_MATCH bytes to insert the |
|
1555 * string following the next match. |
|
1556 */ |
|
1557 if (s->lookahead < MIN_LOOKAHEAD) { |
|
1558 fill_window(s); |
|
1559 if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) { |
|
1560 return need_more; |
|
1561 } |
|
1562 if (s->lookahead == 0) break; /* flush the current block */ |
|
1563 } |
|
1564 |
|
1565 /* Insert the string window[strstart .. strstart+2] in the |
|
1566 * dictionary, and set hash_head to the head of the hash chain: |
|
1567 */ |
|
1568 hash_head = NIL; |
|
1569 if (s->lookahead >= MIN_MATCH) { |
|
1570 INSERT_STRING(s, s->strstart, hash_head); |
|
1571 } |
|
1572 |
|
1573 /* Find the longest match, discarding those <= prev_length. |
|
1574 * At this point we have always match_length < MIN_MATCH |
|
1575 */ |
|
1576 if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) { |
|
1577 /* To simplify the code, we prevent matches with the string |
|
1578 * of window index 0 (in particular we have to avoid a match |
|
1579 * of the string with itself at the start of the input file). |
|
1580 */ |
|
1581 s->match_length = longest_match (s, hash_head); |
|
1582 /* longest_match() sets match_start */ |
|
1583 } |
|
1584 if (s->match_length >= MIN_MATCH) { |
|
1585 check_match(s, s->strstart, s->match_start, s->match_length); |
|
1586 |
|
1587 _tr_tally_dist(s, s->strstart - s->match_start, |
|
1588 s->match_length - MIN_MATCH, bflush); |
|
1589 |
|
1590 s->lookahead -= s->match_length; |
|
1591 |
|
1592 /* Insert new strings in the hash table only if the match length |
|
1593 * is not too large. This saves time but degrades compression. |
|
1594 */ |
|
1595 #ifndef FASTEST |
|
1596 if (s->match_length <= s->max_insert_length && |
|
1597 s->lookahead >= MIN_MATCH) { |
|
1598 s->match_length--; /* string at strstart already in table */ |
|
1599 do { |
|
1600 s->strstart++; |
|
1601 INSERT_STRING(s, s->strstart, hash_head); |
|
1602 /* strstart never exceeds WSIZE-MAX_MATCH, so there are |
|
1603 * always MIN_MATCH bytes ahead. |
|
1604 */ |
|
1605 } while (--s->match_length != 0); |
|
1606 s->strstart++; |
|
1607 } else |
|
1608 #endif |
|
1609 { |
|
1610 s->strstart += s->match_length; |
|
1611 s->match_length = 0; |
|
1612 s->ins_h = s->window[s->strstart]; |
|
1613 UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]); |
|
1614 #if MIN_MATCH != 3 |
|
1615 Call UPDATE_HASH() MIN_MATCH-3 more times |
|
1616 #endif |
|
1617 /* If lookahead < MIN_MATCH, ins_h is garbage, but it does not |
|
1618 * matter since it will be recomputed at next deflate call. |
|
1619 */ |
|
1620 } |
|
1621 } else { |
|
1622 /* No match, output a literal byte */ |
|
1623 Tracevv((stderr,"%c", s->window[s->strstart])); |
|
1624 _tr_tally_lit (s, s->window[s->strstart], bflush); |
|
1625 s->lookahead--; |
|
1626 s->strstart++; |
|
1627 } |
|
1628 if (bflush) FLUSH_BLOCK(s, 0); |
|
1629 } |
|
1630 FLUSH_BLOCK(s, flush == Z_FINISH); |
|
1631 return flush == Z_FINISH ? finish_done : block_done; |
|
1632 } |
|
1633 |
|
1634 #ifndef FASTEST |
|
1635 /* =========================================================================== |
|
1636 * Same as above, but achieves better compression. We use a lazy |
|
1637 * evaluation for matches: a match is finally adopted only if there is |
|
1638 * no better match at the next window position. |
|
1639 */ |
|
1640 local block_state deflate_slow(s, flush) |
|
1641 deflate_state *s; |
|
1642 int flush; |
|
1643 { |
|
1644 IPos hash_head; /* head of hash chain */ |
|
1645 int bflush; /* set if current block must be flushed */ |
|
1646 |
|
1647 /* Process the input block. */ |
|
1648 for (;;) { |
|
1649 /* Make sure that we always have enough lookahead, except |
|
1650 * at the end of the input file. We need MAX_MATCH bytes |
|
1651 * for the next match, plus MIN_MATCH bytes to insert the |
|
1652 * string following the next match. |
|
1653 */ |
|
1654 if (s->lookahead < MIN_LOOKAHEAD) { |
|
1655 fill_window(s); |
|
1656 if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) { |
|
1657 return need_more; |
|
1658 } |
|
1659 if (s->lookahead == 0) break; /* flush the current block */ |
|
1660 } |
|
1661 |
|
1662 /* Insert the string window[strstart .. strstart+2] in the |
|
1663 * dictionary, and set hash_head to the head of the hash chain: |
|
1664 */ |
|
1665 hash_head = NIL; |
|
1666 if (s->lookahead >= MIN_MATCH) { |
|
1667 INSERT_STRING(s, s->strstart, hash_head); |
|
1668 } |
|
1669 |
|
1670 /* Find the longest match, discarding those <= prev_length. |
|
1671 */ |
|
1672 s->prev_length = s->match_length, s->prev_match = s->match_start; |
|
1673 s->match_length = MIN_MATCH-1; |
|
1674 |
|
1675 if (hash_head != NIL && s->prev_length < s->max_lazy_match && |
|
1676 s->strstart - hash_head <= MAX_DIST(s)) { |
|
1677 /* To simplify the code, we prevent matches with the string |
|
1678 * of window index 0 (in particular we have to avoid a match |
|
1679 * of the string with itself at the start of the input file). |
|
1680 */ |
|
1681 s->match_length = longest_match (s, hash_head); |
|
1682 /* longest_match() sets match_start */ |
|
1683 |
|
1684 if (s->match_length <= 5 && (s->strategy == Z_FILTERED |
|
1685 #if TOO_FAR <= 32767 |
|
1686 || (s->match_length == MIN_MATCH && |
|
1687 s->strstart - s->match_start > TOO_FAR) |
|
1688 #endif |
|
1689 )) { |
|
1690 |
|
1691 /* If prev_match is also MIN_MATCH, match_start is garbage |
|
1692 * but we will ignore the current match anyway. |
|
1693 */ |
|
1694 s->match_length = MIN_MATCH-1; |
|
1695 } |
|
1696 } |
|
1697 /* If there was a match at the previous step and the current |
|
1698 * match is not better, output the previous match: |
|
1699 */ |
|
1700 if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) { |
|
1701 uInt max_insert = s->strstart + s->lookahead - MIN_MATCH; |
|
1702 /* Do not insert strings in hash table beyond this. */ |
|
1703 |
|
1704 check_match(s, s->strstart-1, s->prev_match, s->prev_length); |
|
1705 |
|
1706 _tr_tally_dist(s, s->strstart -1 - s->prev_match, |
|
1707 s->prev_length - MIN_MATCH, bflush); |
|
1708 |
|
1709 /* Insert in hash table all strings up to the end of the match. |
|
1710 * strstart-1 and strstart are already inserted. If there is not |
|
1711 * enough lookahead, the last two strings are not inserted in |
|
1712 * the hash table. |
|
1713 */ |
|
1714 s->lookahead -= s->prev_length-1; |
|
1715 s->prev_length -= 2; |
|
1716 do { |
|
1717 if (++s->strstart <= max_insert) { |
|
1718 INSERT_STRING(s, s->strstart, hash_head); |
|
1719 } |
|
1720 } while (--s->prev_length != 0); |
|
1721 s->match_available = 0; |
|
1722 s->match_length = MIN_MATCH-1; |
|
1723 s->strstart++; |
|
1724 |
|
1725 if (bflush) FLUSH_BLOCK(s, 0); |
|
1726 |
|
1727 } else if (s->match_available) { |
|
1728 /* If there was no match at the previous position, output a |
|
1729 * single literal. If there was a match but the current match |
|
1730 * is longer, truncate the previous match to a single literal. |
|
1731 */ |
|
1732 Tracevv((stderr,"%c", s->window[s->strstart-1])); |
|
1733 _tr_tally_lit(s, s->window[s->strstart-1], bflush); |
|
1734 if (bflush) { |
|
1735 FLUSH_BLOCK_ONLY(s, 0); |
|
1736 } |
|
1737 s->strstart++; |
|
1738 s->lookahead--; |
|
1739 if (s->strm->avail_out == 0) return need_more; |
|
1740 } else { |
|
1741 /* There is no previous match to compare with, wait for |
|
1742 * the next step to decide. |
|
1743 */ |
|
1744 s->match_available = 1; |
|
1745 s->strstart++; |
|
1746 s->lookahead--; |
|
1747 } |
|
1748 } |
|
1749 Assert (flush != Z_NO_FLUSH, "no flush?"); |
|
1750 if (s->match_available) { |
|
1751 Tracevv((stderr,"%c", s->window[s->strstart-1])); |
|
1752 _tr_tally_lit(s, s->window[s->strstart-1], bflush); |
|
1753 s->match_available = 0; |
|
1754 } |
|
1755 FLUSH_BLOCK(s, flush == Z_FINISH); |
|
1756 return flush == Z_FINISH ? finish_done : block_done; |
|
1757 } |
|
1758 #endif /* FASTEST */ |
|
1759 |
|
1760 /* =========================================================================== |
|
1761 * For Z_RLE, simply look for runs of bytes, generate matches only of distance |
|
1762 * one. Do not maintain a hash table. (It will be regenerated if this run of |
|
1763 * deflate switches away from Z_RLE.) |
|
1764 */ |
|
1765 local block_state deflate_rle(s, flush) |
|
1766 deflate_state *s; |
|
1767 int flush; |
|
1768 { |
|
1769 int bflush; /* set if current block must be flushed */ |
|
1770 uInt prev; /* byte at distance one to match */ |
|
1771 Bytef *scan, *strend; /* scan goes up to strend for length of run */ |
|
1772 |
|
1773 for (;;) { |
|
1774 /* Make sure that we always have enough lookahead, except |
|
1775 * at the end of the input file. We need MAX_MATCH bytes |
|
1776 * for the longest encodable run. |
|
1777 */ |
|
1778 if (s->lookahead < MAX_MATCH) { |
|
1779 fill_window(s); |
|
1780 if (s->lookahead < MAX_MATCH && flush == Z_NO_FLUSH) { |
|
1781 return need_more; |
|
1782 } |
|
1783 if (s->lookahead == 0) break; /* flush the current block */ |
|
1784 } |
|
1785 |
|
1786 /* See how many times the previous byte repeats */ |
|
1787 s->match_length = 0; |
|
1788 if (s->lookahead >= MIN_MATCH && s->strstart > 0) { |
|
1789 scan = s->window + s->strstart - 1; |
|
1790 prev = *scan; |
|
1791 if (prev == *++scan && prev == *++scan && prev == *++scan) { |
|
1792 strend = s->window + s->strstart + MAX_MATCH; |
|
1793 do { |
|
1794 } while (prev == *++scan && prev == *++scan && |
|
1795 prev == *++scan && prev == *++scan && |
|
1796 prev == *++scan && prev == *++scan && |
|
1797 prev == *++scan && prev == *++scan && |
|
1798 scan < strend); |
|
1799 s->match_length = MAX_MATCH - (int)(strend - scan); |
|
1800 if (s->match_length > s->lookahead) |
|
1801 s->match_length = s->lookahead; |
|
1802 } |
|
1803 } |
|
1804 |
|
1805 /* Emit match if have run of MIN_MATCH or longer, else emit literal */ |
|
1806 if (s->match_length >= MIN_MATCH) { |
|
1807 check_match(s, s->strstart, s->strstart - 1, s->match_length); |
|
1808 |
|
1809 _tr_tally_dist(s, 1, s->match_length - MIN_MATCH, bflush); |
|
1810 |
|
1811 s->lookahead -= s->match_length; |
|
1812 s->strstart += s->match_length; |
|
1813 s->match_length = 0; |
|
1814 } else { |
|
1815 /* No match, output a literal byte */ |
|
1816 Tracevv((stderr,"%c", s->window[s->strstart])); |
|
1817 _tr_tally_lit (s, s->window[s->strstart], bflush); |
|
1818 s->lookahead--; |
|
1819 s->strstart++; |
|
1820 } |
|
1821 if (bflush) FLUSH_BLOCK(s, 0); |
|
1822 } |
|
1823 FLUSH_BLOCK(s, flush == Z_FINISH); |
|
1824 return flush == Z_FINISH ? finish_done : block_done; |
|
1825 } |
|
1826 |
|
1827 /* =========================================================================== |
|
1828 * For Z_HUFFMAN_ONLY, do not look for matches. Do not maintain a hash table. |
|
1829 * (It will be regenerated if this run of deflate switches away from Huffman.) |
|
1830 */ |
|
1831 local block_state deflate_huff(s, flush) |
|
1832 deflate_state *s; |
|
1833 int flush; |
|
1834 { |
|
1835 int bflush; /* set if current block must be flushed */ |
|
1836 |
|
1837 for (;;) { |
|
1838 /* Make sure that we have a literal to write. */ |
|
1839 if (s->lookahead == 0) { |
|
1840 fill_window(s); |
|
1841 if (s->lookahead == 0) { |
|
1842 if (flush == Z_NO_FLUSH) |
|
1843 return need_more; |
|
1844 break; /* flush the current block */ |
|
1845 } |
|
1846 } |
|
1847 |
|
1848 /* Output a literal byte */ |
|
1849 s->match_length = 0; |
|
1850 Tracevv((stderr,"%c", s->window[s->strstart])); |
|
1851 _tr_tally_lit (s, s->window[s->strstart], bflush); |
|
1852 s->lookahead--; |
|
1853 s->strstart++; |
|
1854 if (bflush) FLUSH_BLOCK(s, 0); |
|
1855 } |
|
1856 FLUSH_BLOCK(s, flush == Z_FINISH); |
|
1857 return flush == Z_FINISH ? finish_done : block_done; |
|
1858 } |
|