Commit | Line | Data |
---|---|---|
25f0751f | 1 | /* deflate.c -- compress data using the deflation algorithm |
b864a746 | 2 | * Copyright (C) 1995-2022 Jean-loup Gailly and Mark Adler |
25f0751f PM |
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". | |
7c8257d9 | 40 | * Available in http://tools.ietf.org/html/rfc1951 |
25f0751f PM |
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[] = | |
a98bbd4e | 55 | " deflate 1.2.13 Copyright 1995-2022 Jean-loup Gailly and Mark Adler "; |
25f0751f PM |
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 | ||
6b93e4bc PM |
76 | local int deflateStateCheck OF((z_streamp strm)); |
77 | local void slide_hash OF((deflate_state *s)); | |
25f0751f PM |
78 | local void fill_window OF((deflate_state *s)); |
79 | local block_state deflate_stored OF((deflate_state *s, int flush)); | |
80 | local block_state deflate_fast OF((deflate_state *s, int flush)); | |
81 | #ifndef FASTEST | |
82 | local block_state deflate_slow OF((deflate_state *s, int flush)); | |
83 | #endif | |
f02c02ec CBW |
84 | local block_state deflate_rle OF((deflate_state *s, int flush)); |
85 | local block_state deflate_huff OF((deflate_state *s, int flush)); | |
25f0751f PM |
86 | local void lm_init OF((deflate_state *s)); |
87 | local void putShortMSB OF((deflate_state *s, uInt b)); | |
88 | local void flush_pending OF((z_streamp strm)); | |
6b93e4bc | 89 | local unsigned read_buf OF((z_streamp strm, Bytef *buf, unsigned size)); |
25f0751f | 90 | local uInt longest_match OF((deflate_state *s, IPos cur_match)); |
25f0751f | 91 | |
6b93e4bc | 92 | #ifdef ZLIB_DEBUG |
25f0751f PM |
93 | local void check_match OF((deflate_state *s, IPos start, IPos match, |
94 | int length)); | |
95 | #endif | |
96 | ||
97 | /* =========================================================================== | |
98 | * Local data | |
99 | */ | |
100 | ||
101 | #define NIL 0 | |
102 | /* Tail of hash chains */ | |
103 | ||
104 | #ifndef TOO_FAR | |
105 | # define TOO_FAR 4096 | |
106 | #endif | |
107 | /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */ | |
108 | ||
25f0751f PM |
109 | /* Values for max_lazy_match, good_match and max_chain_length, depending on |
110 | * the desired pack level (0..9). The values given below have been tuned to | |
111 | * exclude worst case performance for pathological files. Better values may be | |
112 | * found for specific files. | |
113 | */ | |
114 | typedef struct config_s { | |
115 | ush good_length; /* reduce lazy search above this match length */ | |
116 | ush max_lazy; /* do not perform lazy search above this match length */ | |
117 | ush nice_length; /* quit search above this match length */ | |
118 | ush max_chain; | |
119 | compress_func func; | |
120 | } config; | |
121 | ||
122 | #ifdef FASTEST | |
123 | local const config configuration_table[2] = { | |
124 | /* good lazy nice chain */ | |
125 | /* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */ | |
126 | /* 1 */ {4, 4, 8, 4, deflate_fast}}; /* max speed, no lazy matches */ | |
127 | #else | |
128 | local const config configuration_table[10] = { | |
129 | /* good lazy nice chain */ | |
130 | /* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */ | |
131 | /* 1 */ {4, 4, 8, 4, deflate_fast}, /* max speed, no lazy matches */ | |
132 | /* 2 */ {4, 5, 16, 8, deflate_fast}, | |
133 | /* 3 */ {4, 6, 32, 32, deflate_fast}, | |
134 | ||
135 | /* 4 */ {4, 4, 16, 16, deflate_slow}, /* lazy matches */ | |
136 | /* 5 */ {8, 16, 32, 32, deflate_slow}, | |
137 | /* 6 */ {8, 16, 128, 128, deflate_slow}, | |
138 | /* 7 */ {8, 32, 128, 256, deflate_slow}, | |
139 | /* 8 */ {32, 128, 258, 1024, deflate_slow}, | |
140 | /* 9 */ {32, 258, 258, 4096, deflate_slow}}; /* max compression */ | |
141 | #endif | |
142 | ||
143 | /* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4 | |
144 | * For deflate_fast() (levels <= 3) good is ignored and lazy has a different | |
145 | * meaning. | |
146 | */ | |
147 | ||
7c8257d9 | 148 | /* rank Z_BLOCK between Z_NO_FLUSH and Z_PARTIAL_FLUSH */ |
2b91859c | 149 | #define RANK(f) (((f) * 2) - ((f) > 4 ? 9 : 0)) |
7c8257d9 | 150 | |
25f0751f PM |
151 | /* =========================================================================== |
152 | * Update a hash value with the given input byte | |
6b93e4bc PM |
153 | * IN assertion: all calls to UPDATE_HASH are made with consecutive input |
154 | * characters, so that a running hash key can be computed from the previous | |
155 | * key instead of complete recalculation each time. | |
25f0751f | 156 | */ |
a98bbd4e | 157 | #define UPDATE_HASH(s,h,c) (h = (((h) << s->hash_shift) ^ (c)) & s->hash_mask) |
25f0751f PM |
158 | |
159 | ||
160 | /* =========================================================================== | |
161 | * Insert string str in the dictionary and set match_head to the previous head | |
162 | * of the hash chain (the most recent string with same hash key). Return | |
163 | * the previous length of the hash chain. | |
164 | * If this file is compiled with -DFASTEST, the compression level is forced | |
165 | * to 1, and no hash chains are maintained. | |
6b93e4bc PM |
166 | * IN assertion: all calls to INSERT_STRING are made with consecutive input |
167 | * characters and the first MIN_MATCH bytes of str are valid (except for | |
168 | * the last MIN_MATCH-1 bytes of the input file). | |
25f0751f PM |
169 | */ |
170 | #ifdef FASTEST | |
171 | #define INSERT_STRING(s, str, match_head) \ | |
172 | (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \ | |
173 | match_head = s->head[s->ins_h], \ | |
174 | s->head[s->ins_h] = (Pos)(str)) | |
175 | #else | |
176 | #define INSERT_STRING(s, str, match_head) \ | |
177 | (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \ | |
178 | match_head = s->prev[(str) & s->w_mask] = s->head[s->ins_h], \ | |
179 | s->head[s->ins_h] = (Pos)(str)) | |
180 | #endif | |
181 | ||
182 | /* =========================================================================== | |
183 | * Initialize the hash table (avoiding 64K overflow for 16 bit systems). | |
184 | * prev[] will be initialized on the fly. | |
185 | */ | |
186 | #define CLEAR_HASH(s) \ | |
f4a99837 | 187 | do { \ |
a98bbd4e | 188 | s->head[s->hash_size - 1] = NIL; \ |
f4a99837 | 189 | zmemzero((Bytef *)s->head, \ |
a98bbd4e | 190 | (unsigned)(s->hash_size - 1)*sizeof(*s->head)); \ |
f4a99837 | 191 | } while (0) |
25f0751f | 192 | |
6b93e4bc PM |
193 | /* =========================================================================== |
194 | * Slide the hash table when sliding the window down (could be avoided with 32 | |
195 | * bit values at the expense of memory usage). We slide even when level == 0 to | |
196 | * keep the hash table consistent if we switch back to level > 0 later. | |
197 | */ | |
198 | local void slide_hash( | |
199 | deflate_state *s) | |
200 | { | |
201 | unsigned n, m; | |
202 | Posf *p; | |
203 | uInt wsize = s->w_size; | |
204 | ||
205 | n = s->hash_size; | |
206 | p = &s->head[n]; | |
207 | do { | |
208 | m = *--p; | |
209 | *p = (Pos)(m >= wsize ? m - wsize : NIL); | |
210 | } while (--n); | |
211 | n = wsize; | |
212 | #ifndef FASTEST | |
213 | p = &s->prev[n]; | |
214 | do { | |
215 | m = *--p; | |
216 | *p = (Pos)(m >= wsize ? m - wsize : NIL); | |
217 | /* If n is not on any hash chain, prev[n] is garbage but | |
218 | * its value will never be used. | |
219 | */ | |
220 | } while (--n); | |
221 | #endif | |
222 | } | |
223 | ||
25f0751f | 224 | /* ========================================================================= */ |
a7026383 CBW |
225 | int ZEXPORT deflateInit_( |
226 | z_streamp strm, | |
227 | int level, | |
228 | const char *version, | |
229 | int stream_size) | |
25f0751f PM |
230 | { |
231 | return deflateInit2_(strm, level, Z_DEFLATED, MAX_WBITS, DEF_MEM_LEVEL, | |
232 | Z_DEFAULT_STRATEGY, version, stream_size); | |
233 | /* To do: ignore strm->next_in if we use it as window */ | |
234 | } | |
235 | ||
236 | /* ========================================================================= */ | |
a7026383 CBW |
237 | int ZEXPORT deflateInit2_( |
238 | z_streamp strm, | |
239 | int level, | |
240 | int method, | |
241 | int windowBits, | |
242 | int memLevel, | |
243 | int strategy, | |
244 | const char *version, | |
245 | int stream_size) | |
25f0751f PM |
246 | { |
247 | deflate_state *s; | |
248 | int wrap = 1; | |
249 | static const char my_version[] = ZLIB_VERSION; | |
250 | ||
25f0751f PM |
251 | if (version == Z_NULL || version[0] != my_version[0] || |
252 | stream_size != sizeof(z_stream)) { | |
253 | return Z_VERSION_ERROR; | |
254 | } | |
255 | if (strm == Z_NULL) return Z_STREAM_ERROR; | |
256 | ||
257 | strm->msg = Z_NULL; | |
258 | if (strm->zalloc == (alloc_func)0) { | |
7c8257d9 CBW |
259 | #ifdef Z_SOLO |
260 | return Z_STREAM_ERROR; | |
261 | #else | |
25f0751f PM |
262 | strm->zalloc = zcalloc; |
263 | strm->opaque = (voidpf)0; | |
7c8257d9 | 264 | #endif |
25f0751f | 265 | } |
7c8257d9 CBW |
266 | if (strm->zfree == (free_func)0) |
267 | #ifdef Z_SOLO | |
268 | return Z_STREAM_ERROR; | |
269 | #else | |
270 | strm->zfree = zcfree; | |
271 | #endif | |
25f0751f PM |
272 | |
273 | #ifdef FASTEST | |
274 | if (level != 0) level = 1; | |
275 | #else | |
276 | if (level == Z_DEFAULT_COMPRESSION) level = 6; | |
277 | #endif | |
278 | ||
279 | if (windowBits < 0) { /* suppress zlib wrapper */ | |
280 | wrap = 0; | |
a98bbd4e YO |
281 | if (windowBits < -15) |
282 | return Z_STREAM_ERROR; | |
25f0751f PM |
283 | windowBits = -windowBits; |
284 | } | |
285 | #ifdef GZIP | |
286 | else if (windowBits > 15) { | |
287 | wrap = 2; /* write gzip wrapper instead */ | |
288 | windowBits -= 16; | |
289 | } | |
290 | #endif | |
291 | if (memLevel < 1 || memLevel > MAX_MEM_LEVEL || method != Z_DEFLATED || | |
292 | windowBits < 8 || windowBits > 15 || level < 0 || level > 9 || | |
6b93e4bc | 293 | strategy < 0 || strategy > Z_FIXED || (windowBits == 8 && wrap != 1)) { |
25f0751f PM |
294 | return Z_STREAM_ERROR; |
295 | } | |
296 | if (windowBits == 8) windowBits = 9; /* until 256-byte window bug fixed */ | |
297 | s = (deflate_state *) ZALLOC(strm, 1, sizeof(deflate_state)); | |
298 | if (s == Z_NULL) return Z_MEM_ERROR; | |
299 | strm->state = (struct internal_state FAR *)s; | |
300 | s->strm = strm; | |
6b93e4bc | 301 | s->status = INIT_STATE; /* to pass state test in deflateReset() */ |
25f0751f PM |
302 | |
303 | s->wrap = wrap; | |
304 | s->gzhead = Z_NULL; | |
6b93e4bc | 305 | s->w_bits = (uInt)windowBits; |
25f0751f PM |
306 | s->w_size = 1 << s->w_bits; |
307 | s->w_mask = s->w_size - 1; | |
308 | ||
6b93e4bc | 309 | s->hash_bits = (uInt)memLevel + 7; |
25f0751f PM |
310 | s->hash_size = 1 << s->hash_bits; |
311 | s->hash_mask = s->hash_size - 1; | |
a98bbd4e | 312 | s->hash_shift = ((s->hash_bits + MIN_MATCH-1) / MIN_MATCH); |
25f0751f PM |
313 | |
314 | s->window = (Bytef *) ZALLOC(strm, s->w_size, 2*sizeof(Byte)); | |
315 | s->prev = (Posf *) ZALLOC(strm, s->w_size, sizeof(Pos)); | |
316 | s->head = (Posf *) ZALLOC(strm, s->hash_size, sizeof(Pos)); | |
317 | ||
f02c02ec CBW |
318 | s->high_water = 0; /* nothing written to s->window yet */ |
319 | ||
25f0751f PM |
320 | s->lit_bufsize = 1 << (memLevel + 6); /* 16K elements by default */ |
321 | ||
b864a746 PM |
322 | /* We overlay pending_buf and sym_buf. This works since the average size |
323 | * for length/distance pairs over any compressed block is assured to be 31 | |
324 | * bits or less. | |
325 | * | |
326 | * Analysis: The longest fixed codes are a length code of 8 bits plus 5 | |
327 | * extra bits, for lengths 131 to 257. The longest fixed distance codes are | |
328 | * 5 bits plus 13 extra bits, for distances 16385 to 32768. The longest | |
329 | * possible fixed-codes length/distance pair is then 31 bits total. | |
330 | * | |
331 | * sym_buf starts one-fourth of the way into pending_buf. So there are | |
332 | * three bytes in sym_buf for every four bytes in pending_buf. Each symbol | |
333 | * in sym_buf is three bytes -- two for the distance and one for the | |
334 | * literal/length. As each symbol is consumed, the pointer to the next | |
335 | * sym_buf value to read moves forward three bytes. From that symbol, up to | |
336 | * 31 bits are written to pending_buf. The closest the written pending_buf | |
337 | * bits gets to the next sym_buf symbol to read is just before the last | |
a98bbd4e YO |
338 | * code is written. At that time, 31*(n - 2) bits have been written, just |
339 | * after 24*(n - 2) bits have been consumed from sym_buf. sym_buf starts at | |
340 | * 8*n bits into pending_buf. (Note that the symbol buffer fills when n - 1 | |
b864a746 | 341 | * symbols are written.) The closest the writing gets to what is unread is |
a98bbd4e | 342 | * then n + 14 bits. Here n is lit_bufsize, which is 16384 by default, and |
b864a746 PM |
343 | * can range from 128 to 32768. |
344 | * | |
345 | * Therefore, at a minimum, there are 142 bits of space between what is | |
346 | * written and what is read in the overlain buffers, so the symbols cannot | |
347 | * be overwritten by the compressed data. That space is actually 139 bits, | |
348 | * due to the three-bit fixed-code block header. | |
349 | * | |
350 | * That covers the case where either Z_FIXED is specified, forcing fixed | |
351 | * codes, or when the use of fixed codes is chosen, because that choice | |
352 | * results in a smaller compressed block than dynamic codes. That latter | |
353 | * condition then assures that the above analysis also covers all dynamic | |
354 | * blocks. A dynamic-code block will only be chosen to be emitted if it has | |
355 | * fewer bits than a fixed-code block would for the same set of symbols. | |
356 | * Therefore its average symbol length is assured to be less than 31. So | |
357 | * the compressed data for a dynamic block also cannot overwrite the | |
358 | * symbols from which it is being constructed. | |
359 | */ | |
360 | ||
361 | s->pending_buf = (uchf *) ZALLOC(strm, s->lit_bufsize, 4); | |
362 | s->pending_buf_size = (ulg)s->lit_bufsize * 4; | |
25f0751f PM |
363 | |
364 | if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL || | |
365 | s->pending_buf == Z_NULL) { | |
366 | s->status = FINISH_STATE; | |
1cae2293 | 367 | strm->msg = ERR_MSG(Z_MEM_ERROR); |
25f0751f PM |
368 | deflateEnd (strm); |
369 | return Z_MEM_ERROR; | |
370 | } | |
b864a746 PM |
371 | s->sym_buf = s->pending_buf + s->lit_bufsize; |
372 | s->sym_end = (s->lit_bufsize - 1) * 3; | |
373 | /* We avoid equality with lit_bufsize*3 because of wraparound at 64K | |
374 | * on 16 bit machines and because stored blocks are restricted to | |
375 | * 64K-1 bytes. | |
376 | */ | |
25f0751f PM |
377 | |
378 | s->level = level; | |
379 | s->strategy = strategy; | |
380 | s->method = (Byte)method; | |
381 | ||
382 | return deflateReset(strm); | |
383 | } | |
384 | ||
6b93e4bc PM |
385 | /* ========================================================================= |
386 | * Check for a valid deflate stream state. Return 0 if ok, 1 if not. | |
387 | */ | |
388 | local int deflateStateCheck ( | |
389 | z_streamp strm) | |
390 | { | |
391 | deflate_state *s; | |
392 | if (strm == Z_NULL || | |
393 | strm->zalloc == (alloc_func)0 || strm->zfree == (free_func)0) | |
394 | return 1; | |
395 | s = strm->state; | |
396 | if (s == Z_NULL || s->strm != strm || (s->status != INIT_STATE && | |
397 | #ifdef GZIP | |
398 | s->status != GZIP_STATE && | |
399 | #endif | |
400 | s->status != EXTRA_STATE && | |
401 | s->status != NAME_STATE && | |
402 | s->status != COMMENT_STATE && | |
403 | s->status != HCRC_STATE && | |
404 | s->status != BUSY_STATE && | |
405 | s->status != FINISH_STATE)) | |
406 | return 1; | |
407 | return 0; | |
408 | } | |
409 | ||
25f0751f | 410 | /* ========================================================================= */ |
a7026383 CBW |
411 | int ZEXPORT deflateSetDictionary ( |
412 | z_streamp strm, | |
413 | const Bytef *dictionary, | |
414 | uInt dictLength) | |
25f0751f PM |
415 | { |
416 | deflate_state *s; | |
7c8257d9 CBW |
417 | uInt str, n; |
418 | int wrap; | |
419 | unsigned avail; | |
1cae2293 | 420 | z_const unsigned char *next; |
25f0751f | 421 | |
6b93e4bc | 422 | if (deflateStateCheck(strm) || dictionary == Z_NULL) |
14c55418 | 423 | return Z_STREAM_ERROR; |
30faa565 | 424 | s = strm->state; |
7c8257d9 CBW |
425 | wrap = s->wrap; |
426 | if (wrap == 2 || (wrap == 1 && s->status != INIT_STATE) || s->lookahead) | |
427 | return Z_STREAM_ERROR; | |
30faa565 | 428 | |
7c8257d9 CBW |
429 | /* when using zlib wrappers, compute Adler-32 for provided dictionary */ |
430 | if (wrap == 1) | |
431 | strm->adler = adler32(strm->adler, dictionary, dictLength); | |
432 | s->wrap = 0; /* avoid computing Adler-32 in read_buf */ | |
433 | ||
434 | /* if dictionary would fill window, just replace the history */ | |
435 | if (dictLength >= s->w_size) { | |
436 | if (wrap == 0) { /* already empty otherwise */ | |
437 | CLEAR_HASH(s); | |
438 | s->strstart = 0; | |
439 | s->block_start = 0L; | |
440 | s->insert = 0; | |
441 | } | |
442 | dictionary += dictLength - s->w_size; /* use the tail */ | |
443 | dictLength = s->w_size; | |
25f0751f | 444 | } |
25f0751f | 445 | |
7c8257d9 CBW |
446 | /* insert dictionary into window and hash */ |
447 | avail = strm->avail_in; | |
448 | next = strm->next_in; | |
449 | strm->avail_in = dictLength; | |
1cae2293 | 450 | strm->next_in = (z_const Bytef *)dictionary; |
7c8257d9 CBW |
451 | fill_window(s); |
452 | while (s->lookahead >= MIN_MATCH) { | |
453 | str = s->strstart; | |
454 | n = s->lookahead - (MIN_MATCH-1); | |
455 | do { | |
456 | UPDATE_HASH(s, s->ins_h, s->window[str + MIN_MATCH-1]); | |
457 | #ifndef FASTEST | |
458 | s->prev[str & s->w_mask] = s->head[s->ins_h]; | |
459 | #endif | |
460 | s->head[s->ins_h] = (Pos)str; | |
461 | str++; | |
462 | } while (--n); | |
463 | s->strstart = str; | |
464 | s->lookahead = MIN_MATCH-1; | |
465 | fill_window(s); | |
25f0751f | 466 | } |
7c8257d9 CBW |
467 | s->strstart += s->lookahead; |
468 | s->block_start = (long)s->strstart; | |
469 | s->insert = s->lookahead; | |
470 | s->lookahead = 0; | |
471 | s->match_length = s->prev_length = MIN_MATCH-1; | |
472 | s->match_available = 0; | |
473 | strm->next_in = next; | |
474 | strm->avail_in = avail; | |
475 | s->wrap = wrap; | |
25f0751f PM |
476 | return Z_OK; |
477 | } | |
478 | ||
479 | /* ========================================================================= */ | |
6b93e4bc PM |
480 | int ZEXPORT deflateGetDictionary ( |
481 | z_streamp strm, | |
482 | Bytef *dictionary, | |
483 | uInt *dictLength) | |
484 | { | |
485 | deflate_state *s; | |
486 | uInt len; | |
487 | ||
488 | if (deflateStateCheck(strm)) | |
489 | return Z_STREAM_ERROR; | |
490 | s = strm->state; | |
491 | len = s->strstart + s->lookahead; | |
492 | if (len > s->w_size) | |
493 | len = s->w_size; | |
494 | if (dictionary != Z_NULL && len) | |
495 | zmemcpy(dictionary, s->window + s->strstart + s->lookahead - len, len); | |
496 | if (dictLength != Z_NULL) | |
497 | *dictLength = len; | |
498 | return Z_OK; | |
499 | } | |
500 | ||
501 | /* ========================================================================= */ | |
7c8257d9 | 502 | int ZEXPORT deflateResetKeep ( |
a7026383 | 503 | z_streamp strm) |
25f0751f PM |
504 | { |
505 | deflate_state *s; | |
506 | ||
6b93e4bc | 507 | if (deflateStateCheck(strm)) { |
25f0751f PM |
508 | return Z_STREAM_ERROR; |
509 | } | |
510 | ||
511 | strm->total_in = strm->total_out = 0; | |
512 | strm->msg = Z_NULL; /* use zfree if we ever allocate msg dynamically */ | |
513 | strm->data_type = Z_UNKNOWN; | |
514 | ||
515 | s = (deflate_state *)strm->state; | |
516 | s->pending = 0; | |
517 | s->pending_out = s->pending_buf; | |
518 | ||
519 | if (s->wrap < 0) { | |
520 | s->wrap = -s->wrap; /* was made negative by deflate(..., Z_FINISH); */ | |
521 | } | |
6b93e4bc PM |
522 | s->status = |
523 | #ifdef GZIP | |
524 | s->wrap == 2 ? GZIP_STATE : | |
525 | #endif | |
b864a746 | 526 | INIT_STATE; |
25f0751f PM |
527 | strm->adler = |
528 | #ifdef GZIP | |
529 | s->wrap == 2 ? crc32(0L, Z_NULL, 0) : | |
530 | #endif | |
531 | adler32(0L, Z_NULL, 0); | |
b864a746 | 532 | s->last_flush = -2; |
25f0751f PM |
533 | |
534 | _tr_init(s); | |
25f0751f PM |
535 | |
536 | return Z_OK; | |
537 | } | |
538 | ||
539 | /* ========================================================================= */ | |
7c8257d9 CBW |
540 | int ZEXPORT deflateReset ( |
541 | z_streamp strm) | |
542 | { | |
543 | int ret; | |
544 | ||
545 | ret = deflateResetKeep(strm); | |
546 | if (ret == Z_OK) | |
547 | lm_init(strm->state); | |
548 | return ret; | |
549 | } | |
550 | ||
551 | /* ========================================================================= */ | |
a7026383 CBW |
552 | int ZEXPORT deflateSetHeader ( |
553 | z_streamp strm, | |
554 | gz_headerp head) | |
25f0751f | 555 | { |
6b93e4bc PM |
556 | if (deflateStateCheck(strm) || strm->state->wrap != 2) |
557 | return Z_STREAM_ERROR; | |
25f0751f PM |
558 | strm->state->gzhead = head; |
559 | return Z_OK; | |
560 | } | |
561 | ||
562 | /* ========================================================================= */ | |
7c8257d9 CBW |
563 | int ZEXPORT deflatePending ( |
564 | z_streamp strm, | |
565 | unsigned *pending, | |
566 | int *bits) | |
567 | { | |
6b93e4bc | 568 | if (deflateStateCheck(strm)) return Z_STREAM_ERROR; |
7c8257d9 CBW |
569 | if (pending != Z_NULL) |
570 | *pending = strm->state->pending; | |
571 | if (bits != Z_NULL) | |
572 | *bits = strm->state->bi_valid; | |
573 | return Z_OK; | |
574 | } | |
575 | ||
576 | /* ========================================================================= */ | |
a7026383 CBW |
577 | int ZEXPORT deflatePrime ( |
578 | z_streamp strm, | |
579 | int bits, | |
580 | int value) | |
14c55418 | 581 | { |
7c8257d9 CBW |
582 | deflate_state *s; |
583 | int put; | |
584 | ||
6b93e4bc | 585 | if (deflateStateCheck(strm)) return Z_STREAM_ERROR; |
7c8257d9 | 586 | s = strm->state; |
b864a746 PM |
587 | if (bits < 0 || bits > 16 || |
588 | s->sym_buf < s->pending_out + ((Buf_size + 7) >> 3)) | |
7c8257d9 CBW |
589 | return Z_BUF_ERROR; |
590 | do { | |
591 | put = Buf_size - s->bi_valid; | |
592 | if (put > bits) | |
593 | put = bits; | |
594 | s->bi_buf |= (ush)((value & ((1 << put) - 1)) << s->bi_valid); | |
595 | s->bi_valid += put; | |
596 | _tr_flush_bits(s); | |
597 | value >>= put; | |
598 | bits -= put; | |
599 | } while (bits); | |
14c55418 A |
600 | return Z_OK; |
601 | } | |
602 | ||
603 | /* ========================================================================= */ | |
a7026383 CBW |
604 | int ZEXPORT deflateParams( |
605 | z_streamp strm, | |
606 | int level, | |
607 | int strategy) | |
25f0751f PM |
608 | { |
609 | deflate_state *s; | |
610 | compress_func func; | |
25f0751f | 611 | |
6b93e4bc | 612 | if (deflateStateCheck(strm)) return Z_STREAM_ERROR; |
25f0751f PM |
613 | s = strm->state; |
614 | ||
615 | #ifdef FASTEST | |
616 | if (level != 0) level = 1; | |
617 | #else | |
618 | if (level == Z_DEFAULT_COMPRESSION) level = 6; | |
619 | #endif | |
620 | if (level < 0 || level > 9 || strategy < 0 || strategy > Z_FIXED) { | |
621 | return Z_STREAM_ERROR; | |
622 | } | |
623 | func = configuration_table[s->level].func; | |
624 | ||
f02c02ec | 625 | if ((strategy != s->strategy || func != configuration_table[level].func) && |
b864a746 | 626 | s->last_flush != -2) { |
25f0751f | 627 | /* Flush the last buffer: */ |
6b93e4bc PM |
628 | int err = deflate(strm, Z_BLOCK); |
629 | if (err == Z_STREAM_ERROR) | |
630 | return err; | |
b864a746 | 631 | if (strm->avail_in || (s->strstart - s->block_start) + s->lookahead) |
6b93e4bc | 632 | return Z_BUF_ERROR; |
25f0751f PM |
633 | } |
634 | if (s->level != level) { | |
6b93e4bc PM |
635 | if (s->level == 0 && s->matches != 0) { |
636 | if (s->matches == 1) | |
637 | slide_hash(s); | |
638 | else | |
639 | CLEAR_HASH(s); | |
640 | s->matches = 0; | |
641 | } | |
25f0751f PM |
642 | s->level = level; |
643 | s->max_lazy_match = configuration_table[level].max_lazy; | |
644 | s->good_match = configuration_table[level].good_length; | |
645 | s->nice_match = configuration_table[level].nice_length; | |
646 | s->max_chain_length = configuration_table[level].max_chain; | |
647 | } | |
648 | s->strategy = strategy; | |
6b93e4bc | 649 | return Z_OK; |
25f0751f PM |
650 | } |
651 | ||
652 | /* ========================================================================= */ | |
a7026383 CBW |
653 | int ZEXPORT deflateTune( |
654 | z_streamp strm, | |
655 | int good_length, | |
656 | int max_lazy, | |
657 | int nice_length, | |
658 | int max_chain) | |
25f0751f PM |
659 | { |
660 | deflate_state *s; | |
661 | ||
6b93e4bc | 662 | if (deflateStateCheck(strm)) return Z_STREAM_ERROR; |
25f0751f | 663 | s = strm->state; |
6b93e4bc PM |
664 | s->good_match = (uInt)good_length; |
665 | s->max_lazy_match = (uInt)max_lazy; | |
25f0751f | 666 | s->nice_match = nice_length; |
6b93e4bc | 667 | s->max_chain_length = (uInt)max_chain; |
25f0751f PM |
668 | return Z_OK; |
669 | } | |
670 | ||
671 | /* ========================================================================= | |
a98bbd4e YO |
672 | * For the default windowBits of 15 and memLevel of 8, this function returns a |
673 | * close to exact, as well as small, upper bound on the compressed size. This | |
674 | * is an expansion of ~0.03%, plus a small constant. | |
675 | * | |
676 | * For any setting other than those defaults for windowBits and memLevel, one | |
677 | * of two worst case bounds is returned. This is at most an expansion of ~4% or | |
678 | * ~13%, plus a small constant. | |
25f0751f | 679 | * |
a98bbd4e YO |
680 | * Both the 0.03% and 4% derive from the overhead of stored blocks. The first |
681 | * one is for stored blocks of 16383 bytes (memLevel == 8), whereas the second | |
682 | * is for stored blocks of 127 bytes (the worst case memLevel == 1). The | |
683 | * expansion results from five bytes of header for each stored block. | |
25f0751f | 684 | * |
a98bbd4e YO |
685 | * The larger expansion of 13% results from a window size less than or equal to |
686 | * the symbols buffer size (windowBits <= memLevel + 7). In that case some of | |
687 | * the data being compressed may have slid out of the sliding window, impeding | |
688 | * a stored block from being emitted. Then the only choice is a fixed or | |
689 | * dynamic block, where a fixed block limits the maximum expansion to 9 bits | |
690 | * per 8-bit byte, plus 10 bits for every block. The smallest block size for | |
691 | * which this can occur is 255 (memLevel == 2). | |
692 | * | |
693 | * Shifts are used to approximate divisions, for speed. | |
25f0751f | 694 | */ |
a7026383 CBW |
695 | uLong ZEXPORT deflateBound( |
696 | z_streamp strm, | |
697 | uLong sourceLen) | |
25f0751f PM |
698 | { |
699 | deflate_state *s; | |
a98bbd4e YO |
700 | uLong fixedlen, storelen, wraplen; |
701 | ||
702 | /* upper bound for fixed blocks with 9-bit literals and length 255 | |
703 | (memLevel == 2, which is the lowest that may not use stored blocks) -- | |
704 | ~13% overhead plus a small constant */ | |
705 | fixedlen = sourceLen + (sourceLen >> 3) + (sourceLen >> 8) + | |
706 | (sourceLen >> 9) + 4; | |
25f0751f | 707 | |
a98bbd4e YO |
708 | /* upper bound for stored blocks with length 127 (memLevel == 1) -- |
709 | ~4% overhead plus a small constant */ | |
710 | storelen = sourceLen + (sourceLen >> 5) + (sourceLen >> 7) + | |
711 | (sourceLen >> 11) + 7; | |
25f0751f | 712 | |
a98bbd4e | 713 | /* if can't get parameters, return larger bound plus a zlib wrapper */ |
6b93e4bc | 714 | if (deflateStateCheck(strm)) |
a98bbd4e | 715 | return (fixedlen > storelen ? fixedlen : storelen) + 6; |
25f0751f | 716 | |
f02c02ec | 717 | /* compute wrapper length */ |
25f0751f | 718 | s = strm->state; |
f02c02ec CBW |
719 | switch (s->wrap) { |
720 | case 0: /* raw deflate */ | |
721 | wraplen = 0; | |
722 | break; | |
723 | case 1: /* zlib wrapper */ | |
724 | wraplen = 6 + (s->strstart ? 4 : 0); | |
725 | break; | |
6b93e4bc | 726 | #ifdef GZIP |
f02c02ec CBW |
727 | case 2: /* gzip wrapper */ |
728 | wraplen = 18; | |
729 | if (s->gzhead != Z_NULL) { /* user-supplied gzip header */ | |
6b93e4bc | 730 | Bytef *str; |
f02c02ec CBW |
731 | if (s->gzhead->extra != Z_NULL) |
732 | wraplen += 2 + s->gzhead->extra_len; | |
733 | str = s->gzhead->name; | |
734 | if (str != Z_NULL) | |
735 | do { | |
736 | wraplen++; | |
737 | } while (*str++); | |
738 | str = s->gzhead->comment; | |
739 | if (str != Z_NULL) | |
740 | do { | |
741 | wraplen++; | |
742 | } while (*str++); | |
743 | if (s->gzhead->hcrc) | |
744 | wraplen += 2; | |
745 | } | |
746 | break; | |
6b93e4bc | 747 | #endif |
f02c02ec CBW |
748 | default: /* for compiler happiness */ |
749 | wraplen = 6; | |
750 | } | |
751 | ||
a98bbd4e | 752 | /* if not default parameters, return one of the conservative bounds */ |
25f0751f | 753 | if (s->w_bits != 15 || s->hash_bits != 8 + 7) |
a98bbd4e | 754 | return (s->w_bits <= s->hash_bits ? fixedlen : storelen) + wraplen; |
25f0751f | 755 | |
a98bbd4e YO |
756 | /* default settings: return tight bound for that case -- ~0.03% overhead |
757 | plus a small constant */ | |
f02c02ec CBW |
758 | return sourceLen + (sourceLen >> 12) + (sourceLen >> 14) + |
759 | (sourceLen >> 25) + 13 - 6 + wraplen; | |
25f0751f PM |
760 | } |
761 | ||
762 | /* ========================================================================= | |
763 | * Put a short in the pending buffer. The 16-bit value is put in MSB order. | |
764 | * IN assertion: the stream state is correct and there is enough room in | |
765 | * pending_buf. | |
766 | */ | |
a7026383 CBW |
767 | local void putShortMSB ( |
768 | deflate_state *s, | |
769 | uInt b) | |
25f0751f PM |
770 | { |
771 | put_byte(s, (Byte)(b >> 8)); | |
772 | put_byte(s, (Byte)(b & 0xff)); | |
773 | } | |
774 | ||
775 | /* ========================================================================= | |
6b93e4bc PM |
776 | * Flush as much pending output as possible. All deflate() output, except for |
777 | * some deflate_stored() output, goes through this function so some | |
778 | * applications may wish to modify it to avoid allocating a large | |
779 | * strm->next_out buffer and copying into it. (See also read_buf()). | |
25f0751f | 780 | */ |
a7026383 CBW |
781 | local void flush_pending( |
782 | z_streamp strm) | |
25f0751f | 783 | { |
7c8257d9 CBW |
784 | unsigned len; |
785 | deflate_state *s = strm->state; | |
25f0751f | 786 | |
7c8257d9 CBW |
787 | _tr_flush_bits(s); |
788 | len = s->pending; | |
25f0751f PM |
789 | if (len > strm->avail_out) len = strm->avail_out; |
790 | if (len == 0) return; | |
791 | ||
7c8257d9 | 792 | zmemcpy(strm->next_out, s->pending_out, len); |
25f0751f | 793 | strm->next_out += len; |
7c8257d9 | 794 | s->pending_out += len; |
25f0751f | 795 | strm->total_out += len; |
6b93e4bc PM |
796 | strm->avail_out -= len; |
797 | s->pending -= len; | |
7c8257d9 CBW |
798 | if (s->pending == 0) { |
799 | s->pending_out = s->pending_buf; | |
25f0751f PM |
800 | } |
801 | } | |
802 | ||
6b93e4bc PM |
803 | /* =========================================================================== |
804 | * Update the header CRC with the bytes s->pending_buf[beg..s->pending - 1]. | |
805 | */ | |
806 | #define HCRC_UPDATE(beg) \ | |
807 | do { \ | |
808 | if (s->gzhead->hcrc && s->pending > (beg)) \ | |
809 | strm->adler = crc32(strm->adler, s->pending_buf + (beg), \ | |
810 | s->pending - (beg)); \ | |
811 | } while (0) | |
812 | ||
25f0751f | 813 | /* ========================================================================= */ |
a7026383 CBW |
814 | int ZEXPORT deflate ( |
815 | z_streamp strm, | |
816 | int flush) | |
25f0751f PM |
817 | { |
818 | int old_flush; /* value of flush param for previous deflate call */ | |
819 | deflate_state *s; | |
820 | ||
6b93e4bc | 821 | if (deflateStateCheck(strm) || flush > Z_BLOCK || flush < 0) { |
25f0751f PM |
822 | return Z_STREAM_ERROR; |
823 | } | |
824 | s = strm->state; | |
825 | ||
826 | if (strm->next_out == Z_NULL || | |
6b93e4bc | 827 | (strm->avail_in != 0 && strm->next_in == Z_NULL) || |
25f0751f PM |
828 | (s->status == FINISH_STATE && flush != Z_FINISH)) { |
829 | ERR_RETURN(strm, Z_STREAM_ERROR); | |
830 | } | |
831 | if (strm->avail_out == 0) ERR_RETURN(strm, Z_BUF_ERROR); | |
832 | ||
25f0751f PM |
833 | old_flush = s->last_flush; |
834 | s->last_flush = flush; | |
835 | ||
6b93e4bc PM |
836 | /* Flush as much pending output as possible */ |
837 | if (s->pending != 0) { | |
838 | flush_pending(strm); | |
839 | if (strm->avail_out == 0) { | |
840 | /* Since avail_out is 0, deflate will be called again with | |
841 | * more output space, but possibly with both pending and | |
842 | * avail_in equal to zero. There won't be anything to do, | |
843 | * but this is not an error situation so make sure we | |
844 | * return OK instead of BUF_ERROR at next call of deflate: | |
845 | */ | |
846 | s->last_flush = -1; | |
847 | return Z_OK; | |
848 | } | |
849 | ||
850 | /* Make sure there is something to do and avoid duplicate consecutive | |
851 | * flushes. For repeated and useless calls with Z_FINISH, we keep | |
852 | * returning Z_STREAM_END instead of Z_BUF_ERROR. | |
853 | */ | |
854 | } else if (strm->avail_in == 0 && RANK(flush) <= RANK(old_flush) && | |
855 | flush != Z_FINISH) { | |
856 | ERR_RETURN(strm, Z_BUF_ERROR); | |
857 | } | |
858 | ||
859 | /* User must not provide more input after the first FINISH: */ | |
860 | if (s->status == FINISH_STATE && strm->avail_in != 0) { | |
861 | ERR_RETURN(strm, Z_BUF_ERROR); | |
862 | } | |
863 | ||
25f0751f | 864 | /* Write the header */ |
b864a746 PM |
865 | if (s->status == INIT_STATE && s->wrap == 0) |
866 | s->status = BUSY_STATE; | |
25f0751f | 867 | if (s->status == INIT_STATE) { |
6b93e4bc | 868 | /* zlib header */ |
a98bbd4e | 869 | uInt header = (Z_DEFLATED + ((s->w_bits - 8) << 4)) << 8; |
6b93e4bc PM |
870 | uInt level_flags; |
871 | ||
872 | if (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2) | |
873 | level_flags = 0; | |
874 | else if (s->level < 6) | |
875 | level_flags = 1; | |
876 | else if (s->level == 6) | |
877 | level_flags = 2; | |
25f0751f | 878 | else |
6b93e4bc PM |
879 | level_flags = 3; |
880 | header |= (level_flags << 6); | |
881 | if (s->strstart != 0) header |= PRESET_DICT; | |
882 | header += 31 - (header % 31); | |
883 | ||
884 | putShortMSB(s, header); | |
25f0751f | 885 | |
6b93e4bc PM |
886 | /* Save the adler32 of the preset dictionary: */ |
887 | if (s->strstart != 0) { | |
888 | putShortMSB(s, (uInt)(strm->adler >> 16)); | |
889 | putShortMSB(s, (uInt)(strm->adler & 0xffff)); | |
890 | } | |
891 | strm->adler = adler32(0L, Z_NULL, 0); | |
892 | s->status = BUSY_STATE; | |
893 | ||
894 | /* Compression must start with an empty pending buffer */ | |
895 | flush_pending(strm); | |
896 | if (s->pending != 0) { | |
897 | s->last_flush = -1; | |
898 | return Z_OK; | |
899 | } | |
900 | } | |
901 | #ifdef GZIP | |
902 | if (s->status == GZIP_STATE) { | |
903 | /* gzip header */ | |
904 | strm->adler = crc32(0L, Z_NULL, 0); | |
905 | put_byte(s, 31); | |
906 | put_byte(s, 139); | |
907 | put_byte(s, 8); | |
908 | if (s->gzhead == Z_NULL) { | |
909 | put_byte(s, 0); | |
910 | put_byte(s, 0); | |
911 | put_byte(s, 0); | |
912 | put_byte(s, 0); | |
913 | put_byte(s, 0); | |
914 | put_byte(s, s->level == 9 ? 2 : | |
915 | (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ? | |
916 | 4 : 0)); | |
917 | put_byte(s, OS_CODE); | |
25f0751f | 918 | s->status = BUSY_STATE; |
25f0751f | 919 | |
6b93e4bc PM |
920 | /* Compression must start with an empty pending buffer */ |
921 | flush_pending(strm); | |
922 | if (s->pending != 0) { | |
923 | s->last_flush = -1; | |
924 | return Z_OK; | |
925 | } | |
926 | } | |
927 | else { | |
928 | put_byte(s, (s->gzhead->text ? 1 : 0) + | |
929 | (s->gzhead->hcrc ? 2 : 0) + | |
930 | (s->gzhead->extra == Z_NULL ? 0 : 4) + | |
931 | (s->gzhead->name == Z_NULL ? 0 : 8) + | |
932 | (s->gzhead->comment == Z_NULL ? 0 : 16) | |
933 | ); | |
934 | put_byte(s, (Byte)(s->gzhead->time & 0xff)); | |
935 | put_byte(s, (Byte)((s->gzhead->time >> 8) & 0xff)); | |
936 | put_byte(s, (Byte)((s->gzhead->time >> 16) & 0xff)); | |
937 | put_byte(s, (Byte)((s->gzhead->time >> 24) & 0xff)); | |
938 | put_byte(s, s->level == 9 ? 2 : | |
939 | (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ? | |
940 | 4 : 0)); | |
941 | put_byte(s, s->gzhead->os & 0xff); | |
942 | if (s->gzhead->extra != Z_NULL) { | |
943 | put_byte(s, s->gzhead->extra_len & 0xff); | |
944 | put_byte(s, (s->gzhead->extra_len >> 8) & 0xff); | |
25f0751f | 945 | } |
6b93e4bc PM |
946 | if (s->gzhead->hcrc) |
947 | strm->adler = crc32(strm->adler, s->pending_buf, | |
948 | s->pending); | |
949 | s->gzindex = 0; | |
950 | s->status = EXTRA_STATE; | |
25f0751f PM |
951 | } |
952 | } | |
25f0751f | 953 | if (s->status == EXTRA_STATE) { |
f02c02ec | 954 | if (s->gzhead->extra != Z_NULL) { |
6b93e4bc PM |
955 | ulg beg = s->pending; /* start of bytes to update crc */ |
956 | uInt left = (s->gzhead->extra_len & 0xffff) - s->gzindex; | |
957 | while (s->pending + left > s->pending_buf_size) { | |
958 | uInt copy = s->pending_buf_size - s->pending; | |
959 | zmemcpy(s->pending_buf + s->pending, | |
960 | s->gzhead->extra + s->gzindex, copy); | |
961 | s->pending = s->pending_buf_size; | |
962 | HCRC_UPDATE(beg); | |
963 | s->gzindex += copy; | |
964 | flush_pending(strm); | |
965 | if (s->pending != 0) { | |
966 | s->last_flush = -1; | |
967 | return Z_OK; | |
25f0751f | 968 | } |
6b93e4bc PM |
969 | beg = 0; |
970 | left -= copy; | |
25f0751f | 971 | } |
6b93e4bc PM |
972 | zmemcpy(s->pending_buf + s->pending, |
973 | s->gzhead->extra + s->gzindex, left); | |
974 | s->pending += left; | |
975 | HCRC_UPDATE(beg); | |
976 | s->gzindex = 0; | |
25f0751f | 977 | } |
6b93e4bc | 978 | s->status = NAME_STATE; |
25f0751f PM |
979 | } |
980 | if (s->status == NAME_STATE) { | |
f02c02ec | 981 | if (s->gzhead->name != Z_NULL) { |
6b93e4bc | 982 | ulg beg = s->pending; /* start of bytes to update crc */ |
25f0751f | 983 | int val; |
25f0751f PM |
984 | do { |
985 | if (s->pending == s->pending_buf_size) { | |
6b93e4bc | 986 | HCRC_UPDATE(beg); |
25f0751f | 987 | flush_pending(strm); |
6b93e4bc PM |
988 | if (s->pending != 0) { |
989 | s->last_flush = -1; | |
990 | return Z_OK; | |
25f0751f | 991 | } |
6b93e4bc | 992 | beg = 0; |
25f0751f PM |
993 | } |
994 | val = s->gzhead->name[s->gzindex++]; | |
995 | put_byte(s, val); | |
996 | } while (val != 0); | |
6b93e4bc PM |
997 | HCRC_UPDATE(beg); |
998 | s->gzindex = 0; | |
25f0751f | 999 | } |
6b93e4bc | 1000 | s->status = COMMENT_STATE; |
25f0751f PM |
1001 | } |
1002 | if (s->status == COMMENT_STATE) { | |
f02c02ec | 1003 | if (s->gzhead->comment != Z_NULL) { |
6b93e4bc | 1004 | ulg beg = s->pending; /* start of bytes to update crc */ |
25f0751f | 1005 | int val; |
25f0751f PM |
1006 | do { |
1007 | if (s->pending == s->pending_buf_size) { | |
6b93e4bc | 1008 | HCRC_UPDATE(beg); |
25f0751f | 1009 | flush_pending(strm); |
6b93e4bc PM |
1010 | if (s->pending != 0) { |
1011 | s->last_flush = -1; | |
1012 | return Z_OK; | |
25f0751f | 1013 | } |
6b93e4bc | 1014 | beg = 0; |
25f0751f PM |
1015 | } |
1016 | val = s->gzhead->comment[s->gzindex++]; | |
1017 | put_byte(s, val); | |
1018 | } while (val != 0); | |
6b93e4bc | 1019 | HCRC_UPDATE(beg); |
25f0751f | 1020 | } |
6b93e4bc | 1021 | s->status = HCRC_STATE; |
25f0751f PM |
1022 | } |
1023 | if (s->status == HCRC_STATE) { | |
1024 | if (s->gzhead->hcrc) { | |
6b93e4bc | 1025 | if (s->pending + 2 > s->pending_buf_size) { |
25f0751f | 1026 | flush_pending(strm); |
6b93e4bc PM |
1027 | if (s->pending != 0) { |
1028 | s->last_flush = -1; | |
1029 | return Z_OK; | |
1030 | } | |
25f0751f | 1031 | } |
6b93e4bc PM |
1032 | put_byte(s, (Byte)(strm->adler & 0xff)); |
1033 | put_byte(s, (Byte)((strm->adler >> 8) & 0xff)); | |
1034 | strm->adler = crc32(0L, Z_NULL, 0); | |
25f0751f | 1035 | } |
6b93e4bc | 1036 | s->status = BUSY_STATE; |
25f0751f | 1037 | |
6b93e4bc | 1038 | /* Compression must start with an empty pending buffer */ |
25f0751f | 1039 | flush_pending(strm); |
6b93e4bc | 1040 | if (s->pending != 0) { |
25f0751f PM |
1041 | s->last_flush = -1; |
1042 | return Z_OK; | |
1043 | } | |
25f0751f | 1044 | } |
6b93e4bc | 1045 | #endif |
25f0751f PM |
1046 | |
1047 | /* Start a new block or continue the current one. | |
1048 | */ | |
1049 | if (strm->avail_in != 0 || s->lookahead != 0 || | |
1050 | (flush != Z_NO_FLUSH && s->status != FINISH_STATE)) { | |
1051 | block_state bstate; | |
1052 | ||
6b93e4bc PM |
1053 | bstate = s->level == 0 ? deflate_stored(s, flush) : |
1054 | s->strategy == Z_HUFFMAN_ONLY ? deflate_huff(s, flush) : | |
1055 | s->strategy == Z_RLE ? deflate_rle(s, flush) : | |
1056 | (*(configuration_table[s->level].func))(s, flush); | |
25f0751f PM |
1057 | |
1058 | if (bstate == finish_started || bstate == finish_done) { | |
1059 | s->status = FINISH_STATE; | |
1060 | } | |
1061 | if (bstate == need_more || bstate == finish_started) { | |
1062 | if (strm->avail_out == 0) { | |
1063 | s->last_flush = -1; /* avoid BUF_ERROR next call, see above */ | |
1064 | } | |
1065 | return Z_OK; | |
1066 | /* If flush != Z_NO_FLUSH && avail_out == 0, the next call | |
1067 | * of deflate should use the same flush parameter to make sure | |
1068 | * that the flush is complete. So we don't have to output an | |
1069 | * empty block here, this will be done at next call. This also | |
1070 | * ensures that for a very small output buffer, we emit at most | |
1071 | * one empty block. | |
1072 | */ | |
1073 | } | |
1074 | if (bstate == block_done) { | |
1075 | if (flush == Z_PARTIAL_FLUSH) { | |
1076 | _tr_align(s); | |
f02c02ec | 1077 | } else if (flush != Z_BLOCK) { /* FULL_FLUSH or SYNC_FLUSH */ |
25f0751f PM |
1078 | _tr_stored_block(s, (char*)0, 0L, 0); |
1079 | /* For a full flush, this empty block will be recognized | |
1080 | * as a special marker by inflate_sync(). | |
1081 | */ | |
1082 | if (flush == Z_FULL_FLUSH) { | |
1083 | CLEAR_HASH(s); /* forget history */ | |
f02c02ec CBW |
1084 | if (s->lookahead == 0) { |
1085 | s->strstart = 0; | |
1086 | s->block_start = 0L; | |
7c8257d9 | 1087 | s->insert = 0; |
f02c02ec | 1088 | } |
25f0751f PM |
1089 | } |
1090 | } | |
1091 | flush_pending(strm); | |
1092 | if (strm->avail_out == 0) { | |
1093 | s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */ | |
1094 | return Z_OK; | |
1095 | } | |
1096 | } | |
1097 | } | |
25f0751f PM |
1098 | |
1099 | if (flush != Z_FINISH) return Z_OK; | |
1100 | if (s->wrap <= 0) return Z_STREAM_END; | |
1101 | ||
1102 | /* Write the trailer */ | |
1103 | #ifdef GZIP | |
1104 | if (s->wrap == 2) { | |
1105 | put_byte(s, (Byte)(strm->adler & 0xff)); | |
1106 | put_byte(s, (Byte)((strm->adler >> 8) & 0xff)); | |
1107 | put_byte(s, (Byte)((strm->adler >> 16) & 0xff)); | |
1108 | put_byte(s, (Byte)((strm->adler >> 24) & 0xff)); | |
1109 | put_byte(s, (Byte)(strm->total_in & 0xff)); | |
1110 | put_byte(s, (Byte)((strm->total_in >> 8) & 0xff)); | |
1111 | put_byte(s, (Byte)((strm->total_in >> 16) & 0xff)); | |
1112 | put_byte(s, (Byte)((strm->total_in >> 24) & 0xff)); | |
1113 | } | |
1114 | else | |
1115 | #endif | |
1116 | { | |
1117 | putShortMSB(s, (uInt)(strm->adler >> 16)); | |
1118 | putShortMSB(s, (uInt)(strm->adler & 0xffff)); | |
1119 | } | |
1120 | flush_pending(strm); | |
1121 | /* If avail_out is zero, the application will call deflate again | |
1122 | * to flush the rest. | |
1123 | */ | |
1124 | if (s->wrap > 0) s->wrap = -s->wrap; /* write the trailer only once! */ | |
1125 | return s->pending != 0 ? Z_OK : Z_STREAM_END; | |
1126 | } | |
1127 | ||
1128 | /* ========================================================================= */ | |
a7026383 CBW |
1129 | int ZEXPORT deflateEnd ( |
1130 | z_streamp strm) | |
25f0751f PM |
1131 | { |
1132 | int status; | |
1133 | ||
6b93e4bc | 1134 | if (deflateStateCheck(strm)) return Z_STREAM_ERROR; |
25f0751f PM |
1135 | |
1136 | status = strm->state->status; | |
25f0751f PM |
1137 | |
1138 | /* Deallocate in reverse order of allocations: */ | |
1139 | TRY_FREE(strm, strm->state->pending_buf); | |
1140 | TRY_FREE(strm, strm->state->head); | |
1141 | TRY_FREE(strm, strm->state->prev); | |
1142 | TRY_FREE(strm, strm->state->window); | |
1143 | ||
1144 | ZFREE(strm, strm->state); | |
1145 | strm->state = Z_NULL; | |
1146 | ||
1147 | return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK; | |
1148 | } | |
1149 | ||
1150 | /* ========================================================================= | |
1151 | * Copy the source state to the destination state. | |
1152 | * To simplify the source, this is not supported for 16-bit MSDOS (which | |
1153 | * doesn't have enough memory anyway to duplicate compression states). | |
1154 | */ | |
a7026383 CBW |
1155 | int ZEXPORT deflateCopy ( |
1156 | z_streamp dest, | |
1157 | z_streamp source) | |
25f0751f PM |
1158 | { |
1159 | #ifdef MAXSEG_64K | |
1160 | return Z_STREAM_ERROR; | |
1161 | #else | |
1162 | deflate_state *ds; | |
1163 | deflate_state *ss; | |
25f0751f PM |
1164 | |
1165 | ||
6b93e4bc | 1166 | if (deflateStateCheck(source) || dest == Z_NULL) { |
25f0751f PM |
1167 | return Z_STREAM_ERROR; |
1168 | } | |
1169 | ||
1170 | ss = source->state; | |
1171 | ||
7c8257d9 | 1172 | zmemcpy((Bytef*)dest, (Bytef*)source, sizeof(z_stream)); |
25f0751f PM |
1173 | |
1174 | ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state)); | |
1175 | if (ds == Z_NULL) return Z_MEM_ERROR; | |
1176 | dest->state = (struct internal_state FAR *) ds; | |
7c8257d9 | 1177 | zmemcpy((Bytef*)ds, (Bytef*)ss, sizeof(deflate_state)); |
25f0751f PM |
1178 | ds->strm = dest; |
1179 | ||
1180 | ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte)); | |
1181 | ds->prev = (Posf *) ZALLOC(dest, ds->w_size, sizeof(Pos)); | |
1182 | ds->head = (Posf *) ZALLOC(dest, ds->hash_size, sizeof(Pos)); | |
b864a746 | 1183 | ds->pending_buf = (uchf *) ZALLOC(dest, ds->lit_bufsize, 4); |
25f0751f PM |
1184 | |
1185 | if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL || | |
1186 | ds->pending_buf == Z_NULL) { | |
1187 | deflateEnd (dest); | |
1188 | return Z_MEM_ERROR; | |
1189 | } | |
1190 | /* following zmemcpy do not work for 16-bit MSDOS */ | |
1191 | zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte)); | |
7c8257d9 CBW |
1192 | zmemcpy((Bytef*)ds->prev, (Bytef*)ss->prev, ds->w_size * sizeof(Pos)); |
1193 | zmemcpy((Bytef*)ds->head, (Bytef*)ss->head, ds->hash_size * sizeof(Pos)); | |
25f0751f PM |
1194 | zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size); |
1195 | ||
1196 | ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf); | |
b864a746 | 1197 | ds->sym_buf = ds->pending_buf + ds->lit_bufsize; |
25f0751f PM |
1198 | |
1199 | ds->l_desc.dyn_tree = ds->dyn_ltree; | |
1200 | ds->d_desc.dyn_tree = ds->dyn_dtree; | |
1201 | ds->bl_desc.dyn_tree = ds->bl_tree; | |
1202 | ||
1203 | return Z_OK; | |
1204 | #endif /* MAXSEG_64K */ | |
1205 | } | |
1206 | ||
1207 | /* =========================================================================== | |
1208 | * Read a new buffer from the current input stream, update the adler32 | |
1209 | * and total number of bytes read. All deflate() input goes through | |
1210 | * this function so some applications may wish to modify it to avoid | |
1211 | * allocating a large strm->next_in buffer and copying from it. | |
1212 | * (See also flush_pending()). | |
1213 | */ | |
6b93e4bc | 1214 | local unsigned read_buf( |
a7026383 CBW |
1215 | z_streamp strm, |
1216 | Bytef *buf, | |
1217 | unsigned size) | |
25f0751f PM |
1218 | { |
1219 | unsigned len = strm->avail_in; | |
1220 | ||
1221 | if (len > size) len = size; | |
1222 | if (len == 0) return 0; | |
1223 | ||
1224 | strm->avail_in -= len; | |
1225 | ||
7c8257d9 | 1226 | zmemcpy(buf, strm->next_in, len); |
25f0751f | 1227 | if (strm->state->wrap == 1) { |
7c8257d9 | 1228 | strm->adler = adler32(strm->adler, buf, len); |
25f0751f PM |
1229 | } |
1230 | #ifdef GZIP | |
1231 | else if (strm->state->wrap == 2) { | |
7c8257d9 | 1232 | strm->adler = crc32(strm->adler, buf, len); |
25f0751f PM |
1233 | } |
1234 | #endif | |
25f0751f PM |
1235 | strm->next_in += len; |
1236 | strm->total_in += len; | |
1237 | ||
6b93e4bc | 1238 | return len; |
25f0751f PM |
1239 | } |
1240 | ||
1241 | /* =========================================================================== | |
1242 | * Initialize the "longest match" routines for a new zlib stream | |
1243 | */ | |
a7026383 CBW |
1244 | local void lm_init ( |
1245 | deflate_state *s) | |
25f0751f PM |
1246 | { |
1247 | s->window_size = (ulg)2L*s->w_size; | |
1248 | ||
1249 | CLEAR_HASH(s); | |
1250 | ||
1251 | /* Set the default configuration parameters: | |
1252 | */ | |
1253 | s->max_lazy_match = configuration_table[s->level].max_lazy; | |
1254 | s->good_match = configuration_table[s->level].good_length; | |
1255 | s->nice_match = configuration_table[s->level].nice_length; | |
1256 | s->max_chain_length = configuration_table[s->level].max_chain; | |
1257 | ||
1258 | s->strstart = 0; | |
1259 | s->block_start = 0L; | |
1260 | s->lookahead = 0; | |
7c8257d9 | 1261 | s->insert = 0; |
25f0751f PM |
1262 | s->match_length = s->prev_length = MIN_MATCH-1; |
1263 | s->match_available = 0; | |
1264 | s->ins_h = 0; | |
25f0751f PM |
1265 | } |
1266 | ||
1267 | #ifndef FASTEST | |
1268 | /* =========================================================================== | |
1269 | * Set match_start to the longest match starting at the given string and | |
1270 | * return its length. Matches shorter or equal to prev_length are discarded, | |
1271 | * in which case the result is equal to prev_length and match_start is | |
1272 | * garbage. | |
1273 | * IN assertions: cur_match is the head of the hash chain for the current | |
1274 | * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1 | |
1275 | * OUT assertion: the match length is not greater than s->lookahead. | |
1276 | */ | |
a7026383 CBW |
1277 | local uInt longest_match( |
1278 | deflate_state *s, | |
1279 | IPos cur_match) | |
25f0751f PM |
1280 | { |
1281 | unsigned chain_length = s->max_chain_length;/* max hash chain length */ | |
492d6b6e TC |
1282 | Bytef *scan = s->window + s->strstart; /* current string */ |
1283 | Bytef *match; /* matched string */ | |
1284 | int len; /* length of current match */ | |
6b93e4bc | 1285 | int best_len = (int)s->prev_length; /* best match length so far */ |
25f0751f PM |
1286 | int nice_match = s->nice_match; /* stop if match long enough */ |
1287 | IPos limit = s->strstart > (IPos)MAX_DIST(s) ? | |
1288 | s->strstart - (IPos)MAX_DIST(s) : NIL; | |
1289 | /* Stop when cur_match becomes <= limit. To simplify the code, | |
1290 | * we prevent matches with the string of window index 0. | |
1291 | */ | |
1292 | Posf *prev = s->prev; | |
1293 | uInt wmask = s->w_mask; | |
1294 | ||
1295 | #ifdef UNALIGNED_OK | |
1296 | /* Compare two bytes at a time. Note: this is not always beneficial. | |
1297 | * Try with and without -DUNALIGNED_OK to check. | |
1298 | */ | |
492d6b6e TC |
1299 | Bytef *strend = s->window + s->strstart + MAX_MATCH - 1; |
1300 | ush scan_start = *(ushf*)scan; | |
1301 | ush scan_end = *(ushf*)(scan + best_len - 1); | |
25f0751f | 1302 | #else |
492d6b6e TC |
1303 | Bytef *strend = s->window + s->strstart + MAX_MATCH; |
1304 | Byte scan_end1 = scan[best_len - 1]; | |
1305 | Byte scan_end = scan[best_len]; | |
25f0751f PM |
1306 | #endif |
1307 | ||
1308 | /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16. | |
1309 | * It is easy to get rid of this optimization if necessary. | |
1310 | */ | |
1311 | Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever"); | |
1312 | ||
1313 | /* Do not waste too much time if we already have a good match: */ | |
1314 | if (s->prev_length >= s->good_match) { | |
1315 | chain_length >>= 2; | |
1316 | } | |
1317 | /* Do not look for matches beyond the end of the input. This is necessary | |
1318 | * to make deflate deterministic. | |
1319 | */ | |
6b93e4bc | 1320 | if ((uInt)nice_match > s->lookahead) nice_match = (int)s->lookahead; |
25f0751f | 1321 | |
a98bbd4e YO |
1322 | Assert((ulg)s->strstart <= s->window_size - MIN_LOOKAHEAD, |
1323 | "need lookahead"); | |
25f0751f PM |
1324 | |
1325 | do { | |
1326 | Assert(cur_match < s->strstart, "no future"); | |
1327 | match = s->window + cur_match; | |
1328 | ||
1329 | /* Skip to next match if the match length cannot increase | |
1330 | * or if the match length is less than 2. Note that the checks below | |
1331 | * for insufficient lookahead only occur occasionally for performance | |
1332 | * reasons. Therefore uninitialized memory will be accessed, and | |
1333 | * conditional jumps will be made that depend on those values. | |
1334 | * However the length of the match is limited to the lookahead, so | |
1335 | * the output of deflate is not affected by the uninitialized values. | |
1336 | */ | |
1337 | #if (defined(UNALIGNED_OK) && MAX_MATCH == 258) | |
1338 | /* This code assumes sizeof(unsigned short) == 2. Do not use | |
1339 | * UNALIGNED_OK if your compiler uses a different size. | |
1340 | */ | |
a98bbd4e | 1341 | if (*(ushf*)(match + best_len - 1) != scan_end || |
25f0751f PM |
1342 | *(ushf*)match != scan_start) continue; |
1343 | ||
1344 | /* It is not necessary to compare scan[2] and match[2] since they are | |
1345 | * always equal when the other bytes match, given that the hash keys | |
1346 | * are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at | |
a98bbd4e | 1347 | * strstart + 3, + 5, up to strstart + 257. We check for insufficient |
25f0751f | 1348 | * lookahead only every 4th comparison; the 128th check will be made |
a98bbd4e | 1349 | * at strstart + 257. If MAX_MATCH-2 is not a multiple of 8, it is |
25f0751f PM |
1350 | * necessary to put more guard bytes at the end of the window, or |
1351 | * to check more often for insufficient lookahead. | |
1352 | */ | |
1353 | Assert(scan[2] == match[2], "scan[2]?"); | |
1354 | scan++, match++; | |
1355 | do { | |
a98bbd4e YO |
1356 | } while (*(ushf*)(scan += 2) == *(ushf*)(match += 2) && |
1357 | *(ushf*)(scan += 2) == *(ushf*)(match += 2) && | |
1358 | *(ushf*)(scan += 2) == *(ushf*)(match += 2) && | |
1359 | *(ushf*)(scan += 2) == *(ushf*)(match += 2) && | |
25f0751f PM |
1360 | scan < strend); |
1361 | /* The funny "do {}" generates better code on most compilers */ | |
1362 | ||
a98bbd4e YO |
1363 | /* Here, scan <= window + strstart + 257 */ |
1364 | Assert(scan <= s->window + (unsigned)(s->window_size - 1), | |
1365 | "wild scan"); | |
25f0751f PM |
1366 | if (*scan == *match) scan++; |
1367 | ||
a98bbd4e | 1368 | len = (MAX_MATCH - 1) - (int)(strend - scan); |
25f0751f PM |
1369 | scan = strend - (MAX_MATCH-1); |
1370 | ||
1371 | #else /* UNALIGNED_OK */ | |
1372 | ||
a98bbd4e YO |
1373 | if (match[best_len] != scan_end || |
1374 | match[best_len - 1] != scan_end1 || | |
1375 | *match != *scan || | |
1376 | *++match != scan[1]) continue; | |
25f0751f | 1377 | |
a98bbd4e | 1378 | /* The check at best_len - 1 can be removed because it will be made |
25f0751f PM |
1379 | * again later. (This heuristic is not always a win.) |
1380 | * It is not necessary to compare scan[2] and match[2] since they | |
1381 | * are always equal when the other bytes match, given that | |
1382 | * the hash keys are equal and that HASH_BITS >= 8. | |
1383 | */ | |
1384 | scan += 2, match++; | |
1385 | Assert(*scan == *match, "match[2]?"); | |
1386 | ||
1387 | /* We check for insufficient lookahead only every 8th comparison; | |
a98bbd4e | 1388 | * the 256th check will be made at strstart + 258. |
25f0751f PM |
1389 | */ |
1390 | do { | |
1391 | } while (*++scan == *++match && *++scan == *++match && | |
1392 | *++scan == *++match && *++scan == *++match && | |
1393 | *++scan == *++match && *++scan == *++match && | |
1394 | *++scan == *++match && *++scan == *++match && | |
1395 | scan < strend); | |
1396 | ||
a98bbd4e YO |
1397 | Assert(scan <= s->window + (unsigned)(s->window_size - 1), |
1398 | "wild scan"); | |
25f0751f PM |
1399 | |
1400 | len = MAX_MATCH - (int)(strend - scan); | |
1401 | scan = strend - MAX_MATCH; | |
1402 | ||
1403 | #endif /* UNALIGNED_OK */ | |
1404 | ||
1405 | if (len > best_len) { | |
1406 | s->match_start = cur_match; | |
1407 | best_len = len; | |
1408 | if (len >= nice_match) break; | |
1409 | #ifdef UNALIGNED_OK | |
a98bbd4e | 1410 | scan_end = *(ushf*)(scan + best_len - 1); |
25f0751f | 1411 | #else |
a98bbd4e | 1412 | scan_end1 = scan[best_len - 1]; |
25f0751f PM |
1413 | scan_end = scan[best_len]; |
1414 | #endif | |
1415 | } | |
1416 | } while ((cur_match = prev[cur_match & wmask]) > limit | |
1417 | && --chain_length != 0); | |
1418 | ||
1419 | if ((uInt)best_len <= s->lookahead) return (uInt)best_len; | |
1420 | return s->lookahead; | |
1421 | } | |
f02c02ec CBW |
1422 | |
1423 | #else /* FASTEST */ | |
25f0751f PM |
1424 | |
1425 | /* --------------------------------------------------------------------------- | |
f02c02ec | 1426 | * Optimized version for FASTEST only |
25f0751f | 1427 | */ |
a7026383 CBW |
1428 | local uInt longest_match( |
1429 | deflate_state *s, | |
1430 | IPos cur_match) | |
25f0751f | 1431 | { |
492d6b6e TC |
1432 | Bytef *scan = s->window + s->strstart; /* current string */ |
1433 | Bytef *match; /* matched string */ | |
1434 | int len; /* length of current match */ | |
1435 | Bytef *strend = s->window + s->strstart + MAX_MATCH; | |
25f0751f PM |
1436 | |
1437 | /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16. | |
1438 | * It is easy to get rid of this optimization if necessary. | |
1439 | */ | |
1440 | Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever"); | |
1441 | ||
a98bbd4e YO |
1442 | Assert((ulg)s->strstart <= s->window_size - MIN_LOOKAHEAD, |
1443 | "need lookahead"); | |
25f0751f PM |
1444 | |
1445 | Assert(cur_match < s->strstart, "no future"); | |
1446 | ||
1447 | match = s->window + cur_match; | |
1448 | ||
1449 | /* Return failure if the match length is less than 2: | |
1450 | */ | |
1451 | if (match[0] != scan[0] || match[1] != scan[1]) return MIN_MATCH-1; | |
1452 | ||
a98bbd4e | 1453 | /* The check at best_len - 1 can be removed because it will be made |
25f0751f PM |
1454 | * again later. (This heuristic is not always a win.) |
1455 | * It is not necessary to compare scan[2] and match[2] since they | |
1456 | * are always equal when the other bytes match, given that | |
1457 | * the hash keys are equal and that HASH_BITS >= 8. | |
1458 | */ | |
1459 | scan += 2, match += 2; | |
1460 | Assert(*scan == *match, "match[2]?"); | |
1461 | ||
1462 | /* We check for insufficient lookahead only every 8th comparison; | |
a98bbd4e | 1463 | * the 256th check will be made at strstart + 258. |
25f0751f PM |
1464 | */ |
1465 | do { | |
1466 | } while (*++scan == *++match && *++scan == *++match && | |
1467 | *++scan == *++match && *++scan == *++match && | |
1468 | *++scan == *++match && *++scan == *++match && | |
1469 | *++scan == *++match && *++scan == *++match && | |
1470 | scan < strend); | |
1471 | ||
a98bbd4e | 1472 | Assert(scan <= s->window + (unsigned)(s->window_size - 1), "wild scan"); |
25f0751f PM |
1473 | |
1474 | len = MAX_MATCH - (int)(strend - scan); | |
1475 | ||
1476 | if (len < MIN_MATCH) return MIN_MATCH - 1; | |
1477 | ||
1478 | s->match_start = cur_match; | |
1479 | return (uInt)len <= s->lookahead ? (uInt)len : s->lookahead; | |
1480 | } | |
1481 | ||
f02c02ec CBW |
1482 | #endif /* FASTEST */ |
1483 | ||
6b93e4bc PM |
1484 | #ifdef ZLIB_DEBUG |
1485 | ||
1486 | #define EQUAL 0 | |
1487 | /* result of memcmp for equal strings */ | |
1488 | ||
25f0751f PM |
1489 | /* =========================================================================== |
1490 | * Check that the match at match_start is indeed a match. | |
1491 | */ | |
a7026383 CBW |
1492 | local void check_match( |
1493 | deflate_state *s, | |
1494 | IPos start, | |
1495 | IPos match, | |
1496 | int length) | |
25f0751f PM |
1497 | { |
1498 | /* check that the match is indeed a match */ | |
1499 | if (zmemcmp(s->window + match, | |
1500 | s->window + start, length) != EQUAL) { | |
1501 | fprintf(stderr, " start %u, match %u, length %d\n", | |
1502 | start, match, length); | |
1503 | do { | |
1504 | fprintf(stderr, "%c%c", s->window[match++], s->window[start++]); | |
1505 | } while (--length != 0); | |
1506 | z_error("invalid match"); | |
1507 | } | |
1508 | if (z_verbose > 1) { | |
a98bbd4e | 1509 | fprintf(stderr,"\\[%d,%d]", start - match, length); |
25f0751f PM |
1510 | do { putc(s->window[start++], stderr); } while (--length != 0); |
1511 | } | |
1512 | } | |
1513 | #else | |
1514 | # define check_match(s, start, match, length) | |
6b93e4bc | 1515 | #endif /* ZLIB_DEBUG */ |
25f0751f PM |
1516 | |
1517 | /* =========================================================================== | |
1518 | * Fill the window when the lookahead becomes insufficient. | |
1519 | * Updates strstart and lookahead. | |
1520 | * | |
1521 | * IN assertion: lookahead < MIN_LOOKAHEAD | |
1522 | * OUT assertions: strstart <= window_size-MIN_LOOKAHEAD | |
1523 | * At least one byte has been read, or avail_in == 0; reads are | |
1524 | * performed for at least two bytes (required for the zip translate_eol | |
1525 | * option -- not supported here). | |
1526 | */ | |
a7026383 CBW |
1527 | local void fill_window( |
1528 | deflate_state *s) | |
25f0751f | 1529 | { |
6b93e4bc | 1530 | unsigned n; |
25f0751f PM |
1531 | unsigned more; /* Amount of free space at the end of the window. */ |
1532 | uInt wsize = s->w_size; | |
1533 | ||
7c8257d9 CBW |
1534 | Assert(s->lookahead < MIN_LOOKAHEAD, "already enough lookahead"); |
1535 | ||
25f0751f PM |
1536 | do { |
1537 | more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart); | |
1538 | ||
1539 | /* Deal with !@#$% 64K limit: */ | |
1540 | if (sizeof(int) <= 2) { | |
1541 | if (more == 0 && s->strstart == 0 && s->lookahead == 0) { | |
1542 | more = wsize; | |
1543 | ||
1544 | } else if (more == (unsigned)(-1)) { | |
1545 | /* Very unlikely, but possible on 16 bit machine if | |
1546 | * strstart == 0 && lookahead == 1 (input done a byte at time) | |
1547 | */ | |
1548 | more--; | |
1549 | } | |
1550 | } | |
1551 | ||
1552 | /* If the window is almost full and there is insufficient lookahead, | |
1553 | * move the upper half to the lower one to make room in the upper half. | |
1554 | */ | |
a98bbd4e | 1555 | if (s->strstart >= wsize + MAX_DIST(s)) { |
25f0751f | 1556 | |
a98bbd4e | 1557 | zmemcpy(s->window, s->window + wsize, (unsigned)wsize - more); |
25f0751f PM |
1558 | s->match_start -= wsize; |
1559 | s->strstart -= wsize; /* we now have strstart >= MAX_DIST */ | |
1560 | s->block_start -= (long) wsize; | |
b864a746 PM |
1561 | if (s->insert > s->strstart) |
1562 | s->insert = s->strstart; | |
6b93e4bc | 1563 | slide_hash(s); |
25f0751f PM |
1564 | more += wsize; |
1565 | } | |
7c8257d9 | 1566 | if (s->strm->avail_in == 0) break; |
25f0751f PM |
1567 | |
1568 | /* If there was no sliding: | |
1569 | * strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 && | |
1570 | * more == window_size - lookahead - strstart | |
1571 | * => more >= window_size - (MIN_LOOKAHEAD-1 + WSIZE + MAX_DIST-1) | |
1572 | * => more >= window_size - 2*WSIZE + 2 | |
1573 | * In the BIG_MEM or MMAP case (not yet supported), | |
1574 | * window_size == input_size + MIN_LOOKAHEAD && | |
1575 | * strstart + s->lookahead <= input_size => more >= MIN_LOOKAHEAD. | |
1576 | * Otherwise, window_size == 2*WSIZE so more >= 2. | |
1577 | * If there was sliding, more >= WSIZE. So in all cases, more >= 2. | |
1578 | */ | |
1579 | Assert(more >= 2, "more < 2"); | |
1580 | ||
1581 | n = read_buf(s->strm, s->window + s->strstart + s->lookahead, more); | |
1582 | s->lookahead += n; | |
1583 | ||
1584 | /* Initialize the hash value now that we have some input: */ | |
7c8257d9 CBW |
1585 | if (s->lookahead + s->insert >= MIN_MATCH) { |
1586 | uInt str = s->strstart - s->insert; | |
1587 | s->ins_h = s->window[str]; | |
1588 | UPDATE_HASH(s, s->ins_h, s->window[str + 1]); | |
25f0751f PM |
1589 | #if MIN_MATCH != 3 |
1590 | Call UPDATE_HASH() MIN_MATCH-3 more times | |
1591 | #endif | |
7c8257d9 CBW |
1592 | while (s->insert) { |
1593 | UPDATE_HASH(s, s->ins_h, s->window[str + MIN_MATCH-1]); | |
1594 | #ifndef FASTEST | |
1595 | s->prev[str & s->w_mask] = s->head[s->ins_h]; | |
1596 | #endif | |
1597 | s->head[s->ins_h] = (Pos)str; | |
1598 | str++; | |
1599 | s->insert--; | |
1600 | if (s->lookahead + s->insert < MIN_MATCH) | |
1601 | break; | |
1602 | } | |
25f0751f PM |
1603 | } |
1604 | /* If the whole input has less than MIN_MATCH bytes, ins_h is garbage, | |
1605 | * but this is not important since only literal bytes will be emitted. | |
1606 | */ | |
1607 | ||
1608 | } while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0); | |
f02c02ec CBW |
1609 | |
1610 | /* If the WIN_INIT bytes after the end of the current data have never been | |
1611 | * written, then zero those bytes in order to avoid memory check reports of | |
1612 | * the use of uninitialized (or uninitialised as Julian writes) bytes by | |
1613 | * the longest match routines. Update the high water mark for the next | |
1614 | * time through here. WIN_INIT is set to MAX_MATCH since the longest match | |
1615 | * routines allow scanning to strstart + MAX_MATCH, ignoring lookahead. | |
1616 | */ | |
1617 | if (s->high_water < s->window_size) { | |
1618 | ulg curr = s->strstart + (ulg)(s->lookahead); | |
1619 | ulg init; | |
1620 | ||
1621 | if (s->high_water < curr) { | |
1622 | /* Previous high water mark below current data -- zero WIN_INIT | |
1623 | * bytes or up to end of window, whichever is less. | |
1624 | */ | |
1625 | init = s->window_size - curr; | |
1626 | if (init > WIN_INIT) | |
1627 | init = WIN_INIT; | |
1628 | zmemzero(s->window + curr, (unsigned)init); | |
1629 | s->high_water = curr + init; | |
1630 | } | |
1631 | else if (s->high_water < (ulg)curr + WIN_INIT) { | |
1632 | /* High water mark at or above current data, but below current data | |
1633 | * plus WIN_INIT -- zero out to current data plus WIN_INIT, or up | |
1634 | * to end of window, whichever is less. | |
1635 | */ | |
1636 | init = (ulg)curr + WIN_INIT - s->high_water; | |
1637 | if (init > s->window_size - s->high_water) | |
1638 | init = s->window_size - s->high_water; | |
1639 | zmemzero(s->window + s->high_water, (unsigned)init); | |
1640 | s->high_water += init; | |
1641 | } | |
1642 | } | |
7c8257d9 CBW |
1643 | |
1644 | Assert((ulg)s->strstart <= s->window_size - MIN_LOOKAHEAD, | |
1645 | "not enough room for search"); | |
25f0751f PM |
1646 | } |
1647 | ||
1648 | /* =========================================================================== | |
1649 | * Flush the current block, with given end-of-file flag. | |
1650 | * IN assertion: strstart is set to the end of the current match. | |
1651 | */ | |
f02c02ec | 1652 | #define FLUSH_BLOCK_ONLY(s, last) { \ |
25f0751f PM |
1653 | _tr_flush_block(s, (s->block_start >= 0L ? \ |
1654 | (charf *)&s->window[(unsigned)s->block_start] : \ | |
1655 | (charf *)Z_NULL), \ | |
1656 | (ulg)((long)s->strstart - s->block_start), \ | |
f02c02ec | 1657 | (last)); \ |
25f0751f PM |
1658 | s->block_start = s->strstart; \ |
1659 | flush_pending(s->strm); \ | |
1660 | Tracev((stderr,"[FLUSH]")); \ | |
1661 | } | |
1662 | ||
1663 | /* Same but force premature exit if necessary. */ | |
f02c02ec CBW |
1664 | #define FLUSH_BLOCK(s, last) { \ |
1665 | FLUSH_BLOCK_ONLY(s, last); \ | |
1666 | if (s->strm->avail_out == 0) return (last) ? finish_started : need_more; \ | |
25f0751f PM |
1667 | } |
1668 | ||
6b93e4bc PM |
1669 | /* Maximum stored block length in deflate format (not including header). */ |
1670 | #define MAX_STORED 65535 | |
1671 | ||
1672 | /* Minimum of a and b. */ | |
1673 | #define MIN(a, b) ((a) > (b) ? (b) : (a)) | |
1674 | ||
25f0751f PM |
1675 | /* =========================================================================== |
1676 | * Copy without compression as much as possible from the input stream, return | |
1677 | * the current block state. | |
6b93e4bc PM |
1678 | * |
1679 | * In case deflateParams() is used to later switch to a non-zero compression | |
1680 | * level, s->matches (otherwise unused when storing) keeps track of the number | |
1681 | * of hash table slides to perform. If s->matches is 1, then one hash table | |
1682 | * slide will be done when switching. If s->matches is 2, the maximum value | |
1683 | * allowed here, then the hash table will be cleared, since two or more slides | |
1684 | * is the same as a clear. | |
1685 | * | |
1686 | * deflate_stored() is written to minimize the number of times an input byte is | |
1687 | * copied. It is most efficient with large input and output buffers, which | |
a98bbd4e | 1688 | * maximizes the opportunities to have a single copy from next_in to next_out. |
25f0751f | 1689 | */ |
a7026383 CBW |
1690 | local block_state deflate_stored( |
1691 | deflate_state *s, | |
1692 | int flush) | |
25f0751f | 1693 | { |
6b93e4bc PM |
1694 | /* Smallest worthy block size when not flushing or finishing. By default |
1695 | * this is 32K. This can be as small as 507 bytes for memLevel == 1. For | |
1696 | * large input and output buffers, the stored block size will be larger. | |
25f0751f | 1697 | */ |
6b93e4bc | 1698 | unsigned min_block = MIN(s->pending_buf_size - 5, s->w_size); |
25f0751f | 1699 | |
6b93e4bc PM |
1700 | /* Copy as many min_block or larger stored blocks directly to next_out as |
1701 | * possible. If flushing, copy the remaining available input to next_out as | |
1702 | * stored blocks, if there is enough space. | |
1703 | */ | |
1704 | unsigned len, left, have, last = 0; | |
1705 | unsigned used = s->strm->avail_in; | |
1706 | do { | |
1707 | /* Set len to the maximum size block that we can copy directly with the | |
1708 | * available input data and output space. Set left to how much of that | |
1709 | * would be copied from what's left in the window. | |
1710 | */ | |
1711 | len = MAX_STORED; /* maximum deflate stored block length */ | |
1712 | have = (s->bi_valid + 42) >> 3; /* number of header bytes */ | |
1713 | if (s->strm->avail_out < have) /* need room for header */ | |
1714 | break; | |
1715 | /* maximum stored block length that will fit in avail_out: */ | |
1716 | have = s->strm->avail_out - have; | |
1717 | left = s->strstart - s->block_start; /* bytes left in window */ | |
1718 | if (len > (ulg)left + s->strm->avail_in) | |
1719 | len = left + s->strm->avail_in; /* limit len to the input */ | |
1720 | if (len > have) | |
1721 | len = have; /* limit len to the output */ | |
1722 | ||
1723 | /* If the stored block would be less than min_block in length, or if | |
1724 | * unable to copy all of the available input when flushing, then try | |
1725 | * copying to the window and the pending buffer instead. Also don't | |
1726 | * write an empty block when flushing -- deflate() does that. | |
1727 | */ | |
1728 | if (len < min_block && ((len == 0 && flush != Z_FINISH) || | |
1729 | flush == Z_NO_FLUSH || | |
1730 | len != left + s->strm->avail_in)) | |
1731 | break; | |
25f0751f | 1732 | |
6b93e4bc PM |
1733 | /* Make a dummy stored block in pending to get the header bytes, |
1734 | * including any pending bits. This also updates the debugging counts. | |
1735 | */ | |
1736 | last = flush == Z_FINISH && len == left + s->strm->avail_in ? 1 : 0; | |
1737 | _tr_stored_block(s, (char *)0, 0L, last); | |
1738 | ||
1739 | /* Replace the lengths in the dummy stored block with len. */ | |
1740 | s->pending_buf[s->pending - 4] = len; | |
1741 | s->pending_buf[s->pending - 3] = len >> 8; | |
1742 | s->pending_buf[s->pending - 2] = ~len; | |
1743 | s->pending_buf[s->pending - 1] = ~len >> 8; | |
1744 | ||
1745 | /* Write the stored block header bytes. */ | |
1746 | flush_pending(s->strm); | |
1747 | ||
1748 | #ifdef ZLIB_DEBUG | |
1749 | /* Update debugging counts for the data about to be copied. */ | |
1750 | s->compressed_len += len << 3; | |
1751 | s->bits_sent += len << 3; | |
1752 | #endif | |
25f0751f | 1753 | |
6b93e4bc PM |
1754 | /* Copy uncompressed bytes from the window to next_out. */ |
1755 | if (left) { | |
1756 | if (left > len) | |
1757 | left = len; | |
1758 | zmemcpy(s->strm->next_out, s->window + s->block_start, left); | |
1759 | s->strm->next_out += left; | |
1760 | s->strm->avail_out -= left; | |
1761 | s->strm->total_out += left; | |
1762 | s->block_start += left; | |
1763 | len -= left; | |
25f0751f | 1764 | } |
6b93e4bc PM |
1765 | |
1766 | /* Copy uncompressed bytes directly from next_in to next_out, updating | |
1767 | * the check value. | |
1768 | */ | |
1769 | if (len) { | |
1770 | read_buf(s->strm, s->strm->next_out, len); | |
1771 | s->strm->next_out += len; | |
1772 | s->strm->avail_out -= len; | |
1773 | s->strm->total_out += len; | |
25f0751f | 1774 | } |
6b93e4bc PM |
1775 | } while (last == 0); |
1776 | ||
1777 | /* Update the sliding window with the last s->w_size bytes of the copied | |
1778 | * data, or append all of the copied data to the existing window if less | |
1779 | * than s->w_size bytes were copied. Also update the number of bytes to | |
1780 | * insert in the hash tables, in the event that deflateParams() switches to | |
1781 | * a non-zero compression level. | |
1782 | */ | |
1783 | used -= s->strm->avail_in; /* number of input bytes directly copied */ | |
1784 | if (used) { | |
1785 | /* If any input was used, then no unused input remains in the window, | |
1786 | * therefore s->block_start == s->strstart. | |
25f0751f | 1787 | */ |
6b93e4bc PM |
1788 | if (used >= s->w_size) { /* supplant the previous history */ |
1789 | s->matches = 2; /* clear hash */ | |
1790 | zmemcpy(s->window, s->strm->next_in - s->w_size, s->w_size); | |
1791 | s->strstart = s->w_size; | |
b864a746 | 1792 | s->insert = s->strstart; |
25f0751f | 1793 | } |
6b93e4bc PM |
1794 | else { |
1795 | if (s->window_size - s->strstart <= used) { | |
1796 | /* Slide the window down. */ | |
1797 | s->strstart -= s->w_size; | |
1798 | zmemcpy(s->window, s->window + s->w_size, s->strstart); | |
1799 | if (s->matches < 2) | |
1800 | s->matches++; /* add a pending slide_hash() */ | |
b864a746 PM |
1801 | if (s->insert > s->strstart) |
1802 | s->insert = s->strstart; | |
6b93e4bc PM |
1803 | } |
1804 | zmemcpy(s->window + s->strstart, s->strm->next_in - used, used); | |
1805 | s->strstart += used; | |
b864a746 | 1806 | s->insert += MIN(used, s->w_size - s->insert); |
6b93e4bc PM |
1807 | } |
1808 | s->block_start = s->strstart; | |
25f0751f | 1809 | } |
6b93e4bc PM |
1810 | if (s->high_water < s->strstart) |
1811 | s->high_water = s->strstart; | |
1812 | ||
1813 | /* If the last block was written to next_out, then done. */ | |
1814 | if (last) | |
7c8257d9 | 1815 | return finish_done; |
6b93e4bc PM |
1816 | |
1817 | /* If flushing and all input has been consumed, then done. */ | |
1818 | if (flush != Z_NO_FLUSH && flush != Z_FINISH && | |
1819 | s->strm->avail_in == 0 && (long)s->strstart == s->block_start) | |
1820 | return block_done; | |
1821 | ||
1822 | /* Fill the window with any remaining input. */ | |
b864a746 | 1823 | have = s->window_size - s->strstart; |
6b93e4bc PM |
1824 | if (s->strm->avail_in > have && s->block_start >= (long)s->w_size) { |
1825 | /* Slide the window down. */ | |
1826 | s->block_start -= s->w_size; | |
1827 | s->strstart -= s->w_size; | |
1828 | zmemcpy(s->window, s->window + s->w_size, s->strstart); | |
1829 | if (s->matches < 2) | |
1830 | s->matches++; /* add a pending slide_hash() */ | |
1831 | have += s->w_size; /* more space now */ | |
b864a746 PM |
1832 | if (s->insert > s->strstart) |
1833 | s->insert = s->strstart; | |
7c8257d9 | 1834 | } |
6b93e4bc PM |
1835 | if (have > s->strm->avail_in) |
1836 | have = s->strm->avail_in; | |
1837 | if (have) { | |
1838 | read_buf(s->strm, s->window + s->strstart, have); | |
1839 | s->strstart += have; | |
b864a746 | 1840 | s->insert += MIN(have, s->w_size - s->insert); |
6b93e4bc PM |
1841 | } |
1842 | if (s->high_water < s->strstart) | |
1843 | s->high_water = s->strstart; | |
1844 | ||
1845 | /* There was not enough avail_out to write a complete worthy or flushed | |
1846 | * stored block to next_out. Write a stored block to pending instead, if we | |
1847 | * have enough input for a worthy block, or if flushing and there is enough | |
1848 | * room for the remaining input as a stored block in the pending buffer. | |
1849 | */ | |
1850 | have = (s->bi_valid + 42) >> 3; /* number of header bytes */ | |
1851 | /* maximum stored block length that will fit in pending: */ | |
1852 | have = MIN(s->pending_buf_size - have, MAX_STORED); | |
1853 | min_block = MIN(have, s->w_size); | |
1854 | left = s->strstart - s->block_start; | |
1855 | if (left >= min_block || | |
1856 | ((left || flush == Z_FINISH) && flush != Z_NO_FLUSH && | |
1857 | s->strm->avail_in == 0 && left <= have)) { | |
1858 | len = MIN(left, have); | |
1859 | last = flush == Z_FINISH && s->strm->avail_in == 0 && | |
1860 | len == left ? 1 : 0; | |
1861 | _tr_stored_block(s, (charf *)s->window + s->block_start, len, last); | |
1862 | s->block_start += len; | |
1863 | flush_pending(s->strm); | |
1864 | } | |
1865 | ||
1866 | /* We've done all we can with the available input and output. */ | |
1867 | return last ? finish_started : need_more; | |
25f0751f PM |
1868 | } |
1869 | ||
1870 | /* =========================================================================== | |
1871 | * Compress as much as possible from the input stream, return the current | |
1872 | * block state. | |
1873 | * This function does not perform lazy evaluation of matches and inserts | |
1874 | * new strings in the dictionary only for unmatched strings or for short | |
1875 | * matches. It is used only for the fast compression options. | |
1876 | */ | |
a7026383 CBW |
1877 | local block_state deflate_fast( |
1878 | deflate_state *s, | |
1879 | int flush) | |
25f0751f | 1880 | { |
f02c02ec | 1881 | IPos hash_head; /* head of the hash chain */ |
25f0751f PM |
1882 | int bflush; /* set if current block must be flushed */ |
1883 | ||
1884 | for (;;) { | |
1885 | /* Make sure that we always have enough lookahead, except | |
1886 | * at the end of the input file. We need MAX_MATCH bytes | |
1887 | * for the next match, plus MIN_MATCH bytes to insert the | |
1888 | * string following the next match. | |
1889 | */ | |
1890 | if (s->lookahead < MIN_LOOKAHEAD) { | |
1891 | fill_window(s); | |
1892 | if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) { | |
1893 | return need_more; | |
1894 | } | |
1895 | if (s->lookahead == 0) break; /* flush the current block */ | |
1896 | } | |
1897 | ||
a98bbd4e | 1898 | /* Insert the string window[strstart .. strstart + 2] in the |
25f0751f PM |
1899 | * dictionary, and set hash_head to the head of the hash chain: |
1900 | */ | |
f02c02ec | 1901 | hash_head = NIL; |
25f0751f PM |
1902 | if (s->lookahead >= MIN_MATCH) { |
1903 | INSERT_STRING(s, s->strstart, hash_head); | |
1904 | } | |
1905 | ||
1906 | /* Find the longest match, discarding those <= prev_length. | |
1907 | * At this point we have always match_length < MIN_MATCH | |
1908 | */ | |
1909 | if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) { | |
1910 | /* To simplify the code, we prevent matches with the string | |
1911 | * of window index 0 (in particular we have to avoid a match | |
1912 | * of the string with itself at the start of the input file). | |
1913 | */ | |
f02c02ec CBW |
1914 | s->match_length = longest_match (s, hash_head); |
1915 | /* longest_match() sets match_start */ | |
25f0751f PM |
1916 | } |
1917 | if (s->match_length >= MIN_MATCH) { | |
1918 | check_match(s, s->strstart, s->match_start, s->match_length); | |
1919 | ||
1920 | _tr_tally_dist(s, s->strstart - s->match_start, | |
1921 | s->match_length - MIN_MATCH, bflush); | |
1922 | ||
1923 | s->lookahead -= s->match_length; | |
1924 | ||
1925 | /* Insert new strings in the hash table only if the match length | |
1926 | * is not too large. This saves time but degrades compression. | |
1927 | */ | |
1928 | #ifndef FASTEST | |
1929 | if (s->match_length <= s->max_insert_length && | |
1930 | s->lookahead >= MIN_MATCH) { | |
1931 | s->match_length--; /* string at strstart already in table */ | |
1932 | do { | |
1933 | s->strstart++; | |
1934 | INSERT_STRING(s, s->strstart, hash_head); | |
1935 | /* strstart never exceeds WSIZE-MAX_MATCH, so there are | |
1936 | * always MIN_MATCH bytes ahead. | |
1937 | */ | |
1938 | } while (--s->match_length != 0); | |
1939 | s->strstart++; | |
1940 | } else | |
1941 | #endif | |
1942 | { | |
1943 | s->strstart += s->match_length; | |
1944 | s->match_length = 0; | |
1945 | s->ins_h = s->window[s->strstart]; | |
a98bbd4e | 1946 | UPDATE_HASH(s, s->ins_h, s->window[s->strstart + 1]); |
25f0751f PM |
1947 | #if MIN_MATCH != 3 |
1948 | Call UPDATE_HASH() MIN_MATCH-3 more times | |
1949 | #endif | |
1950 | /* If lookahead < MIN_MATCH, ins_h is garbage, but it does not | |
1951 | * matter since it will be recomputed at next deflate call. | |
1952 | */ | |
1953 | } | |
1954 | } else { | |
1955 | /* No match, output a literal byte */ | |
1956 | Tracevv((stderr,"%c", s->window[s->strstart])); | |
a98bbd4e | 1957 | _tr_tally_lit(s, s->window[s->strstart], bflush); |
25f0751f PM |
1958 | s->lookahead--; |
1959 | s->strstart++; | |
1960 | } | |
1961 | if (bflush) FLUSH_BLOCK(s, 0); | |
1962 | } | |
7c8257d9 CBW |
1963 | s->insert = s->strstart < MIN_MATCH-1 ? s->strstart : MIN_MATCH-1; |
1964 | if (flush == Z_FINISH) { | |
1965 | FLUSH_BLOCK(s, 1); | |
1966 | return finish_done; | |
1967 | } | |
b864a746 | 1968 | if (s->sym_next) |
7c8257d9 CBW |
1969 | FLUSH_BLOCK(s, 0); |
1970 | return block_done; | |
25f0751f PM |
1971 | } |
1972 | ||
1973 | #ifndef FASTEST | |
1974 | /* =========================================================================== | |
1975 | * Same as above, but achieves better compression. We use a lazy | |
1976 | * evaluation for matches: a match is finally adopted only if there is | |
1977 | * no better match at the next window position. | |
1978 | */ | |
a7026383 CBW |
1979 | local block_state deflate_slow( |
1980 | deflate_state *s, | |
1981 | int flush) | |
25f0751f | 1982 | { |
f02c02ec | 1983 | IPos hash_head; /* head of hash chain */ |
25f0751f PM |
1984 | int bflush; /* set if current block must be flushed */ |
1985 | ||
1986 | /* Process the input block. */ | |
1987 | for (;;) { | |
1988 | /* Make sure that we always have enough lookahead, except | |
1989 | * at the end of the input file. We need MAX_MATCH bytes | |
1990 | * for the next match, plus MIN_MATCH bytes to insert the | |
1991 | * string following the next match. | |
1992 | */ | |
1993 | if (s->lookahead < MIN_LOOKAHEAD) { | |
1994 | fill_window(s); | |
1995 | if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) { | |
1996 | return need_more; | |
1997 | } | |
1998 | if (s->lookahead == 0) break; /* flush the current block */ | |
1999 | } | |
2000 | ||
a98bbd4e | 2001 | /* Insert the string window[strstart .. strstart + 2] in the |
25f0751f PM |
2002 | * dictionary, and set hash_head to the head of the hash chain: |
2003 | */ | |
f02c02ec | 2004 | hash_head = NIL; |
25f0751f PM |
2005 | if (s->lookahead >= MIN_MATCH) { |
2006 | INSERT_STRING(s, s->strstart, hash_head); | |
2007 | } | |
2008 | ||
2009 | /* Find the longest match, discarding those <= prev_length. | |
2010 | */ | |
2011 | s->prev_length = s->match_length, s->prev_match = s->match_start; | |
2012 | s->match_length = MIN_MATCH-1; | |
2013 | ||
2014 | if (hash_head != NIL && s->prev_length < s->max_lazy_match && | |
2015 | s->strstart - hash_head <= MAX_DIST(s)) { | |
2016 | /* To simplify the code, we prevent matches with the string | |
2017 | * of window index 0 (in particular we have to avoid a match | |
2018 | * of the string with itself at the start of the input file). | |
2019 | */ | |
f02c02ec CBW |
2020 | s->match_length = longest_match (s, hash_head); |
2021 | /* longest_match() sets match_start */ | |
25f0751f PM |
2022 | |
2023 | if (s->match_length <= 5 && (s->strategy == Z_FILTERED | |
2024 | #if TOO_FAR <= 32767 | |
2025 | || (s->match_length == MIN_MATCH && | |
2026 | s->strstart - s->match_start > TOO_FAR) | |
2027 | #endif | |
2028 | )) { | |
2029 | ||
2030 | /* If prev_match is also MIN_MATCH, match_start is garbage | |
2031 | * but we will ignore the current match anyway. | |
2032 | */ | |
2033 | s->match_length = MIN_MATCH-1; | |
2034 | } | |
2035 | } | |
2036 | /* If there was a match at the previous step and the current | |
2037 | * match is not better, output the previous match: | |
2038 | */ | |
2039 | if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) { | |
2040 | uInt max_insert = s->strstart + s->lookahead - MIN_MATCH; | |
2041 | /* Do not insert strings in hash table beyond this. */ | |
2042 | ||
a98bbd4e | 2043 | check_match(s, s->strstart - 1, s->prev_match, s->prev_length); |
25f0751f | 2044 | |
a98bbd4e | 2045 | _tr_tally_dist(s, s->strstart - 1 - s->prev_match, |
25f0751f PM |
2046 | s->prev_length - MIN_MATCH, bflush); |
2047 | ||
2048 | /* Insert in hash table all strings up to the end of the match. | |
a98bbd4e | 2049 | * strstart - 1 and strstart are already inserted. If there is not |
25f0751f PM |
2050 | * enough lookahead, the last two strings are not inserted in |
2051 | * the hash table. | |
2052 | */ | |
a98bbd4e | 2053 | s->lookahead -= s->prev_length - 1; |
25f0751f PM |
2054 | s->prev_length -= 2; |
2055 | do { | |
2056 | if (++s->strstart <= max_insert) { | |
2057 | INSERT_STRING(s, s->strstart, hash_head); | |
2058 | } | |
2059 | } while (--s->prev_length != 0); | |
2060 | s->match_available = 0; | |
2061 | s->match_length = MIN_MATCH-1; | |
2062 | s->strstart++; | |
2063 | ||
2064 | if (bflush) FLUSH_BLOCK(s, 0); | |
2065 | ||
2066 | } else if (s->match_available) { | |
2067 | /* If there was no match at the previous position, output a | |
2068 | * single literal. If there was a match but the current match | |
2069 | * is longer, truncate the previous match to a single literal. | |
2070 | */ | |
a98bbd4e YO |
2071 | Tracevv((stderr,"%c", s->window[s->strstart - 1])); |
2072 | _tr_tally_lit(s, s->window[s->strstart - 1], bflush); | |
25f0751f PM |
2073 | if (bflush) { |
2074 | FLUSH_BLOCK_ONLY(s, 0); | |
2075 | } | |
2076 | s->strstart++; | |
2077 | s->lookahead--; | |
2078 | if (s->strm->avail_out == 0) return need_more; | |
2079 | } else { | |
2080 | /* There is no previous match to compare with, wait for | |
2081 | * the next step to decide. | |
2082 | */ | |
2083 | s->match_available = 1; | |
2084 | s->strstart++; | |
2085 | s->lookahead--; | |
2086 | } | |
2087 | } | |
2088 | Assert (flush != Z_NO_FLUSH, "no flush?"); | |
2089 | if (s->match_available) { | |
a98bbd4e YO |
2090 | Tracevv((stderr,"%c", s->window[s->strstart - 1])); |
2091 | _tr_tally_lit(s, s->window[s->strstart - 1], bflush); | |
25f0751f PM |
2092 | s->match_available = 0; |
2093 | } | |
7c8257d9 CBW |
2094 | s->insert = s->strstart < MIN_MATCH-1 ? s->strstart : MIN_MATCH-1; |
2095 | if (flush == Z_FINISH) { | |
2096 | FLUSH_BLOCK(s, 1); | |
2097 | return finish_done; | |
2098 | } | |
b864a746 | 2099 | if (s->sym_next) |
7c8257d9 CBW |
2100 | FLUSH_BLOCK(s, 0); |
2101 | return block_done; | |
25f0751f PM |
2102 | } |
2103 | #endif /* FASTEST */ | |
2104 | ||
25f0751f PM |
2105 | /* =========================================================================== |
2106 | * For Z_RLE, simply look for runs of bytes, generate matches only of distance | |
2107 | * one. Do not maintain a hash table. (It will be regenerated if this run of | |
2108 | * deflate switches away from Z_RLE.) | |
2109 | */ | |
a7026383 CBW |
2110 | local block_state deflate_rle( |
2111 | deflate_state *s, | |
2112 | int flush) | |
25f0751f | 2113 | { |
f02c02ec CBW |
2114 | int bflush; /* set if current block must be flushed */ |
2115 | uInt prev; /* byte at distance one to match */ | |
2116 | Bytef *scan, *strend; /* scan goes up to strend for length of run */ | |
25f0751f PM |
2117 | |
2118 | for (;;) { | |
2119 | /* Make sure that we always have enough lookahead, except | |
2120 | * at the end of the input file. We need MAX_MATCH bytes | |
7c8257d9 | 2121 | * for the longest run, plus one for the unrolled loop. |
25f0751f | 2122 | */ |
7c8257d9 | 2123 | if (s->lookahead <= MAX_MATCH) { |
25f0751f | 2124 | fill_window(s); |
7c8257d9 | 2125 | if (s->lookahead <= MAX_MATCH && flush == Z_NO_FLUSH) { |
25f0751f PM |
2126 | return need_more; |
2127 | } | |
2128 | if (s->lookahead == 0) break; /* flush the current block */ | |
2129 | } | |
2130 | ||
2131 | /* See how many times the previous byte repeats */ | |
f02c02ec CBW |
2132 | s->match_length = 0; |
2133 | if (s->lookahead >= MIN_MATCH && s->strstart > 0) { | |
25f0751f | 2134 | scan = s->window + s->strstart - 1; |
f02c02ec CBW |
2135 | prev = *scan; |
2136 | if (prev == *++scan && prev == *++scan && prev == *++scan) { | |
2137 | strend = s->window + s->strstart + MAX_MATCH; | |
2138 | do { | |
2139 | } while (prev == *++scan && prev == *++scan && | |
2140 | prev == *++scan && prev == *++scan && | |
2141 | prev == *++scan && prev == *++scan && | |
2142 | prev == *++scan && prev == *++scan && | |
2143 | scan < strend); | |
6b93e4bc | 2144 | s->match_length = MAX_MATCH - (uInt)(strend - scan); |
f02c02ec CBW |
2145 | if (s->match_length > s->lookahead) |
2146 | s->match_length = s->lookahead; | |
2147 | } | |
a98bbd4e YO |
2148 | Assert(scan <= s->window + (uInt)(s->window_size - 1), |
2149 | "wild scan"); | |
25f0751f PM |
2150 | } |
2151 | ||
2152 | /* Emit match if have run of MIN_MATCH or longer, else emit literal */ | |
f02c02ec CBW |
2153 | if (s->match_length >= MIN_MATCH) { |
2154 | check_match(s, s->strstart, s->strstart - 1, s->match_length); | |
2155 | ||
2156 | _tr_tally_dist(s, 1, s->match_length - MIN_MATCH, bflush); | |
2157 | ||
2158 | s->lookahead -= s->match_length; | |
2159 | s->strstart += s->match_length; | |
2160 | s->match_length = 0; | |
25f0751f PM |
2161 | } else { |
2162 | /* No match, output a literal byte */ | |
2163 | Tracevv((stderr,"%c", s->window[s->strstart])); | |
a98bbd4e | 2164 | _tr_tally_lit(s, s->window[s->strstart], bflush); |
25f0751f PM |
2165 | s->lookahead--; |
2166 | s->strstart++; | |
2167 | } | |
2168 | if (bflush) FLUSH_BLOCK(s, 0); | |
2169 | } | |
7c8257d9 CBW |
2170 | s->insert = 0; |
2171 | if (flush == Z_FINISH) { | |
2172 | FLUSH_BLOCK(s, 1); | |
2173 | return finish_done; | |
2174 | } | |
b864a746 | 2175 | if (s->sym_next) |
7c8257d9 CBW |
2176 | FLUSH_BLOCK(s, 0); |
2177 | return block_done; | |
25f0751f | 2178 | } |
f02c02ec CBW |
2179 | |
2180 | /* =========================================================================== | |
2181 | * For Z_HUFFMAN_ONLY, do not look for matches. Do not maintain a hash table. | |
2182 | * (It will be regenerated if this run of deflate switches away from Huffman.) | |
2183 | */ | |
a7026383 CBW |
2184 | local block_state deflate_huff( |
2185 | deflate_state *s, | |
2186 | int flush) | |
f02c02ec CBW |
2187 | { |
2188 | int bflush; /* set if current block must be flushed */ | |
2189 | ||
2190 | for (;;) { | |
2191 | /* Make sure that we have a literal to write. */ | |
2192 | if (s->lookahead == 0) { | |
2193 | fill_window(s); | |
2194 | if (s->lookahead == 0) { | |
2195 | if (flush == Z_NO_FLUSH) | |
2196 | return need_more; | |
2197 | break; /* flush the current block */ | |
2198 | } | |
2199 | } | |
2200 | ||
2201 | /* Output a literal byte */ | |
2202 | s->match_length = 0; | |
2203 | Tracevv((stderr,"%c", s->window[s->strstart])); | |
a98bbd4e | 2204 | _tr_tally_lit(s, s->window[s->strstart], bflush); |
f02c02ec CBW |
2205 | s->lookahead--; |
2206 | s->strstart++; | |
2207 | if (bflush) FLUSH_BLOCK(s, 0); | |
2208 | } | |
7c8257d9 CBW |
2209 | s->insert = 0; |
2210 | if (flush == Z_FINISH) { | |
2211 | FLUSH_BLOCK(s, 1); | |
2212 | return finish_done; | |
2213 | } | |
b864a746 | 2214 | if (s->sym_next) |
7c8257d9 CBW |
2215 | FLUSH_BLOCK(s, 0); |
2216 | return block_done; | |
f02c02ec | 2217 | } |