1 | /* deflate.c -- compress data using the deflation algorithm |
2 | * Copyright (C) 1995-2023 Jean-loup Gailly and Mark Adler |
3 | * For conditions of distribution and use, see copyright notice in zlib.h |
4 | */ |
5 | |
6 | /* |
7 | * ALGORITHM |
8 | * |
9 | * The "deflation" process depends on being able to identify portions |
10 | * of the input text which are identical to earlier input (within a |
11 | * sliding window trailing behind the input currently being processed). |
12 | * |
13 | * The most straightforward technique turns out to be the fastest for |
14 | * most input files: try all possible matches and select the longest. |
15 | * The key feature of this algorithm is that insertions into the string |
16 | * dictionary are very simple and thus fast, and deletions are avoided |
17 | * completely. Insertions are performed at each input character, whereas |
18 | * string matches are performed only when the previous match ends. So it |
19 | * is preferable to spend more time in matches to allow very fast string |
20 | * insertions and avoid deletions. The matching algorithm for small |
21 | * strings is inspired from that of Rabin & Karp. A brute force approach |
22 | * is used to find longer strings when a small match has been found. |
23 | * A similar algorithm is used in comic (by Jan-Mark Wams) and freeze |
24 | * (by Leonid Broukhis). |
25 | * A previous version of this file used a more sophisticated algorithm |
26 | * (by Fiala and Greene) which is guaranteed to run in linear amortized |
27 | * time, but has a larger average cost, uses more memory and is patented. |
28 | * However the F&G algorithm may be faster for some highly redundant |
29 | * files if the parameter max_chain_length (described below) is too large. |
30 | * |
31 | * ACKNOWLEDGEMENTS |
32 | * |
33 | * The idea of lazy evaluation of matches is due to Jan-Mark Wams, and |
34 | * I found it in 'freeze' written by Leonid Broukhis. |
35 | * Thanks to many people for bug reports and testing. |
36 | * |
37 | * REFERENCES |
38 | * |
39 | * Deutsch, L.P.,"DEFLATE Compressed Data Format Specification". |
40 | * Available in http://tools.ietf.org/html/rfc1951 |
41 | * |
42 | * A description of the Rabin and Karp algorithm is given in the book |
43 | * "Algorithms" by R. Sedgewick, Addison-Wesley, p252. |
44 | * |
45 | * Fiala,E.R., and Greene,D.H. |
46 | * Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595 |
47 | * |
48 | */ |
49 | |
50 | /* @(#) $Id$ */ |
51 | |
52 | #include "deflate.h" |
53 | |
54 | const char deflate_copyright[] = |
55 | " deflate 1.3 Copyright 1995-2023 Jean-loup Gailly and Mark Adler " ; |
56 | /* |
57 | If you use the zlib library in a product, an acknowledgment is welcome |
58 | in the documentation of your product. If for some reason you cannot |
59 | include such an acknowledgment, I would appreciate that you keep this |
60 | copyright string in the executable of your product. |
61 | */ |
62 | |
63 | typedef enum { |
64 | need_more, /* block not completed, need more input or more output */ |
65 | block_done, /* block flush performed */ |
66 | finish_started, /* finish started, need only more output at next deflate */ |
67 | finish_done /* finish done, accept no more input or output */ |
68 | } block_state; |
69 | |
70 | typedef block_state (*compress_func)(deflate_state *s, int flush); |
71 | /* Compression function. Returns the block state after the call. */ |
72 | |
73 | local block_state deflate_stored(deflate_state *s, int flush); |
74 | local block_state deflate_fast(deflate_state *s, int flush); |
75 | #ifndef FASTEST |
76 | local block_state deflate_slow(deflate_state *s, int flush); |
77 | #endif |
78 | local block_state deflate_rle(deflate_state *s, int flush); |
79 | local block_state deflate_huff(deflate_state *s, int flush); |
80 | |
81 | /* =========================================================================== |
82 | * Local data |
83 | */ |
84 | |
85 | #define NIL 0 |
86 | /* Tail of hash chains */ |
87 | |
88 | #ifndef TOO_FAR |
89 | # define TOO_FAR 4096 |
90 | #endif |
91 | /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */ |
92 | |
93 | /* Values for max_lazy_match, good_match and max_chain_length, depending on |
94 | * the desired pack level (0..9). The values given below have been tuned to |
95 | * exclude worst case performance for pathological files. Better values may be |
96 | * found for specific files. |
97 | */ |
98 | typedef struct config_s { |
99 | ush good_length; /* reduce lazy search above this match length */ |
100 | ush max_lazy; /* do not perform lazy search above this match length */ |
101 | ush nice_length; /* quit search above this match length */ |
102 | ush max_chain; |
103 | compress_func func; |
104 | } config; |
105 | |
106 | #ifdef FASTEST |
107 | local const config configuration_table[2] = { |
108 | /* good lazy nice chain */ |
109 | /* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */ |
110 | /* 1 */ {4, 4, 8, 4, deflate_fast}}; /* max speed, no lazy matches */ |
111 | #else |
112 | local const config configuration_table[10] = { |
113 | /* good lazy nice chain */ |
114 | /* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */ |
115 | /* 1 */ {4, 4, 8, 4, deflate_fast}, /* max speed, no lazy matches */ |
116 | /* 2 */ {4, 5, 16, 8, deflate_fast}, |
117 | /* 3 */ {4, 6, 32, 32, deflate_fast}, |
118 | |
119 | /* 4 */ {4, 4, 16, 16, deflate_slow}, /* lazy matches */ |
120 | /* 5 */ {8, 16, 32, 32, deflate_slow}, |
121 | /* 6 */ {8, 16, 128, 128, deflate_slow}, |
122 | /* 7 */ {8, 32, 128, 256, deflate_slow}, |
123 | /* 8 */ {32, 128, 258, 1024, deflate_slow}, |
124 | /* 9 */ {32, 258, 258, 4096, deflate_slow}}; /* max compression */ |
125 | #endif |
126 | |
127 | /* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4 |
128 | * For deflate_fast() (levels <= 3) good is ignored and lazy has a different |
129 | * meaning. |
130 | */ |
131 | |
132 | /* rank Z_BLOCK between Z_NO_FLUSH and Z_PARTIAL_FLUSH */ |
133 | #define RANK(f) (((f) * 2) - ((f) > 4 ? 9 : 0)) |
134 | |
135 | /* =========================================================================== |
136 | * Update a hash value with the given input byte |
137 | * IN assertion: all calls to UPDATE_HASH are made with consecutive input |
138 | * characters, so that a running hash key can be computed from the previous |
139 | * key instead of complete recalculation each time. |
140 | */ |
141 | #define UPDATE_HASH(s,h,c) (h = (((h) << s->hash_shift) ^ (c)) & s->hash_mask) |
142 | |
143 | |
144 | /* =========================================================================== |
145 | * Insert string str in the dictionary and set match_head to the previous head |
146 | * of the hash chain (the most recent string with same hash key). Return |
147 | * the previous length of the hash chain. |
148 | * If this file is compiled with -DFASTEST, the compression level is forced |
149 | * to 1, and no hash chains are maintained. |
150 | * IN assertion: all calls to INSERT_STRING are made with consecutive input |
151 | * characters and the first MIN_MATCH bytes of str are valid (except for |
152 | * the last MIN_MATCH-1 bytes of the input file). |
153 | */ |
154 | #ifdef FASTEST |
155 | #define INSERT_STRING(s, str, match_head) \ |
156 | (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \ |
157 | match_head = s->head[s->ins_h], \ |
158 | s->head[s->ins_h] = (Pos)(str)) |
159 | #else |
160 | #define INSERT_STRING(s, str, match_head) \ |
161 | (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \ |
162 | match_head = s->prev[(str) & s->w_mask] = s->head[s->ins_h], \ |
163 | s->head[s->ins_h] = (Pos)(str)) |
164 | #endif |
165 | |
166 | /* =========================================================================== |
167 | * Initialize the hash table (avoiding 64K overflow for 16 bit systems). |
168 | * prev[] will be initialized on the fly. |
169 | */ |
170 | #define CLEAR_HASH(s) \ |
171 | do { \ |
172 | s->head[s->hash_size - 1] = NIL; \ |
173 | zmemzero((Bytef *)s->head, \ |
174 | (unsigned)(s->hash_size - 1)*sizeof(*s->head)); \ |
175 | } while (0) |
176 | |
177 | /* =========================================================================== |
178 | * Slide the hash table when sliding the window down (could be avoided with 32 |
179 | * bit values at the expense of memory usage). We slide even when level == 0 to |
180 | * keep the hash table consistent if we switch back to level > 0 later. |
181 | */ |
182 | #if defined(__has_feature) |
183 | # if __has_feature(memory_sanitizer) |
184 | __attribute__((no_sanitize("memory" ))) |
185 | # endif |
186 | #endif |
187 | local void slide_hash(deflate_state *s) { |
188 | unsigned n, m; |
189 | Posf *p; |
190 | uInt wsize = s->w_size; |
191 | |
192 | n = s->hash_size; |
193 | p = &s->head[n]; |
194 | do { |
195 | m = *--p; |
196 | *p = (Pos)(m >= wsize ? m - wsize : NIL); |
197 | } while (--n); |
198 | n = wsize; |
199 | #ifndef FASTEST |
200 | p = &s->prev[n]; |
201 | do { |
202 | m = *--p; |
203 | *p = (Pos)(m >= wsize ? m - wsize : NIL); |
204 | /* If n is not on any hash chain, prev[n] is garbage but |
205 | * its value will never be used. |
206 | */ |
207 | } while (--n); |
208 | #endif |
209 | } |
210 | |
211 | /* =========================================================================== |
212 | * Read a new buffer from the current input stream, update the adler32 |
213 | * and total number of bytes read. All deflate() input goes through |
214 | * this function so some applications may wish to modify it to avoid |
215 | * allocating a large strm->next_in buffer and copying from it. |
216 | * (See also flush_pending()). |
217 | */ |
218 | local unsigned read_buf(z_streamp strm, Bytef *buf, unsigned size) { |
219 | unsigned len = strm->avail_in; |
220 | |
221 | if (len > size) len = size; |
222 | if (len == 0) return 0; |
223 | |
224 | strm->avail_in -= len; |
225 | |
226 | zmemcpy(buf, strm->next_in, len); |
227 | if (strm->state->wrap == 1) { |
228 | strm->adler = adler32(strm->adler, buf, len); |
229 | } |
230 | #ifdef GZIP |
231 | else if (strm->state->wrap == 2) { |
232 | strm->adler = crc32(strm->adler, buf, len); |
233 | } |
234 | #endif |
235 | strm->next_in += len; |
236 | strm->total_in += len; |
237 | |
238 | return len; |
239 | } |
240 | |
241 | /* =========================================================================== |
242 | * Fill the window when the lookahead becomes insufficient. |
243 | * Updates strstart and lookahead. |
244 | * |
245 | * IN assertion: lookahead < MIN_LOOKAHEAD |
246 | * OUT assertions: strstart <= window_size-MIN_LOOKAHEAD |
247 | * At least one byte has been read, or avail_in == 0; reads are |
248 | * performed for at least two bytes (required for the zip translate_eol |
249 | * option -- not supported here). |
250 | */ |
251 | local void fill_window(deflate_state *s) { |
252 | unsigned n; |
253 | unsigned more; /* Amount of free space at the end of the window. */ |
254 | uInt wsize = s->w_size; |
255 | |
256 | Assert(s->lookahead < MIN_LOOKAHEAD, "already enough lookahead" ); |
257 | |
258 | do { |
259 | more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart); |
260 | |
261 | /* Deal with !@#$% 64K limit: */ |
262 | if (sizeof(int) <= 2) { |
263 | if (more == 0 && s->strstart == 0 && s->lookahead == 0) { |
264 | more = wsize; |
265 | |
266 | } else if (more == (unsigned)(-1)) { |
267 | /* Very unlikely, but possible on 16 bit machine if |
268 | * strstart == 0 && lookahead == 1 (input done a byte at time) |
269 | */ |
270 | more--; |
271 | } |
272 | } |
273 | |
274 | /* If the window is almost full and there is insufficient lookahead, |
275 | * move the upper half to the lower one to make room in the upper half. |
276 | */ |
277 | if (s->strstart >= wsize + MAX_DIST(s)) { |
278 | |
279 | zmemcpy(s->window, s->window + wsize, (unsigned)wsize - more); |
280 | s->match_start -= wsize; |
281 | s->strstart -= wsize; /* we now have strstart >= MAX_DIST */ |
282 | s->block_start -= (long) wsize; |
283 | if (s->insert > s->strstart) |
284 | s->insert = s->strstart; |
285 | slide_hash(s); |
286 | more += wsize; |
287 | } |
288 | if (s->strm->avail_in == 0) break; |
289 | |
290 | /* If there was no sliding: |
291 | * strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 && |
292 | * more == window_size - lookahead - strstart |
293 | * => more >= window_size - (MIN_LOOKAHEAD-1 + WSIZE + MAX_DIST-1) |
294 | * => more >= window_size - 2*WSIZE + 2 |
295 | * In the BIG_MEM or MMAP case (not yet supported), |
296 | * window_size == input_size + MIN_LOOKAHEAD && |
297 | * strstart + s->lookahead <= input_size => more >= MIN_LOOKAHEAD. |
298 | * Otherwise, window_size == 2*WSIZE so more >= 2. |
299 | * If there was sliding, more >= WSIZE. So in all cases, more >= 2. |
300 | */ |
301 | Assert(more >= 2, "more < 2" ); |
302 | |
303 | n = read_buf(s->strm, s->window + s->strstart + s->lookahead, more); |
304 | s->lookahead += n; |
305 | |
306 | /* Initialize the hash value now that we have some input: */ |
307 | if (s->lookahead + s->insert >= MIN_MATCH) { |
308 | uInt str = s->strstart - s->insert; |
309 | s->ins_h = s->window[str]; |
310 | UPDATE_HASH(s, s->ins_h, s->window[str + 1]); |
311 | #if MIN_MATCH != 3 |
312 | Call UPDATE_HASH() MIN_MATCH-3 more times |
313 | #endif |
314 | while (s->insert) { |
315 | UPDATE_HASH(s, s->ins_h, s->window[str + MIN_MATCH-1]); |
316 | #ifndef FASTEST |
317 | s->prev[str & s->w_mask] = s->head[s->ins_h]; |
318 | #endif |
319 | s->head[s->ins_h] = (Pos)str; |
320 | str++; |
321 | s->insert--; |
322 | if (s->lookahead + s->insert < MIN_MATCH) |
323 | break; |
324 | } |
325 | } |
326 | /* If the whole input has less than MIN_MATCH bytes, ins_h is garbage, |
327 | * but this is not important since only literal bytes will be emitted. |
328 | */ |
329 | |
330 | } while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0); |
331 | |
332 | /* If the WIN_INIT bytes after the end of the current data have never been |
333 | * written, then zero those bytes in order to avoid memory check reports of |
334 | * the use of uninitialized (or uninitialised as Julian writes) bytes by |
335 | * the longest match routines. Update the high water mark for the next |
336 | * time through here. WIN_INIT is set to MAX_MATCH since the longest match |
337 | * routines allow scanning to strstart + MAX_MATCH, ignoring lookahead. |
338 | */ |
339 | if (s->high_water < s->window_size) { |
340 | ulg curr = s->strstart + (ulg)(s->lookahead); |
341 | ulg init; |
342 | |
343 | if (s->high_water < curr) { |
344 | /* Previous high water mark below current data -- zero WIN_INIT |
345 | * bytes or up to end of window, whichever is less. |
346 | */ |
347 | init = s->window_size - curr; |
348 | if (init > WIN_INIT) |
349 | init = WIN_INIT; |
350 | zmemzero(s->window + curr, (unsigned)init); |
351 | s->high_water = curr + init; |
352 | } |
353 | else if (s->high_water < (ulg)curr + WIN_INIT) { |
354 | /* High water mark at or above current data, but below current data |
355 | * plus WIN_INIT -- zero out to current data plus WIN_INIT, or up |
356 | * to end of window, whichever is less. |
357 | */ |
358 | init = (ulg)curr + WIN_INIT - s->high_water; |
359 | if (init > s->window_size - s->high_water) |
360 | init = s->window_size - s->high_water; |
361 | zmemzero(s->window + s->high_water, (unsigned)init); |
362 | s->high_water += init; |
363 | } |
364 | } |
365 | |
366 | Assert((ulg)s->strstart <= s->window_size - MIN_LOOKAHEAD, |
367 | "not enough room for search" ); |
368 | } |
369 | |
370 | /* ========================================================================= */ |
371 | int ZEXPORT deflateInit_(z_streamp strm, int level, const char *version, |
372 | int stream_size) { |
373 | return deflateInit2_(strm, level, Z_DEFLATED, MAX_WBITS, DEF_MEM_LEVEL, |
374 | Z_DEFAULT_STRATEGY, version, stream_size); |
375 | /* To do: ignore strm->next_in if we use it as window */ |
376 | } |
377 | |
378 | /* ========================================================================= */ |
379 | int ZEXPORT deflateInit2_(z_streamp strm, int level, int method, |
380 | int windowBits, int memLevel, int strategy, |
381 | const char *version, int stream_size) { |
382 | deflate_state *s; |
383 | int wrap = 1; |
384 | static const char my_version[] = ZLIB_VERSION; |
385 | |
386 | if (version == Z_NULL || version[0] != my_version[0] || |
387 | stream_size != sizeof(z_stream)) { |
388 | return Z_VERSION_ERROR; |
389 | } |
390 | if (strm == Z_NULL) return Z_STREAM_ERROR; |
391 | |
392 | strm->msg = Z_NULL; |
393 | if (strm->zalloc == (alloc_func)0) { |
394 | #ifdef Z_SOLO |
395 | return Z_STREAM_ERROR; |
396 | #else |
397 | strm->zalloc = zcalloc; |
398 | strm->opaque = (voidpf)0; |
399 | #endif |
400 | } |
401 | if (strm->zfree == (free_func)0) |
402 | #ifdef Z_SOLO |
403 | return Z_STREAM_ERROR; |
404 | #else |
405 | strm->zfree = zcfree; |
406 | #endif |
407 | |
408 | #ifdef FASTEST |
409 | if (level != 0) level = 1; |
410 | #else |
411 | if (level == Z_DEFAULT_COMPRESSION) level = 6; |
412 | #endif |
413 | |
414 | if (windowBits < 0) { /* suppress zlib wrapper */ |
415 | wrap = 0; |
416 | if (windowBits < -15) |
417 | return Z_STREAM_ERROR; |
418 | windowBits = -windowBits; |
419 | } |
420 | #ifdef GZIP |
421 | else if (windowBits > 15) { |
422 | wrap = 2; /* write gzip wrapper instead */ |
423 | windowBits -= 16; |
424 | } |
425 | #endif |
426 | if (memLevel < 1 || memLevel > MAX_MEM_LEVEL || method != Z_DEFLATED || |
427 | windowBits < 8 || windowBits > 15 || level < 0 || level > 9 || |
428 | strategy < 0 || strategy > Z_FIXED || (windowBits == 8 && wrap != 1)) { |
429 | return Z_STREAM_ERROR; |
430 | } |
431 | if (windowBits == 8) windowBits = 9; /* until 256-byte window bug fixed */ |
432 | s = (deflate_state *) ZALLOC(strm, 1, sizeof(deflate_state)); |
433 | if (s == Z_NULL) return Z_MEM_ERROR; |
434 | strm->state = (struct internal_state FAR *)s; |
435 | s->strm = strm; |
436 | s->status = INIT_STATE; /* to pass state test in deflateReset() */ |
437 | |
438 | s->wrap = wrap; |
439 | s->gzhead = Z_NULL; |
440 | s->w_bits = (uInt)windowBits; |
441 | s->w_size = 1 << s->w_bits; |
442 | s->w_mask = s->w_size - 1; |
443 | |
444 | s->hash_bits = (uInt)memLevel + 7; |
445 | s->hash_size = 1 << s->hash_bits; |
446 | s->hash_mask = s->hash_size - 1; |
447 | s->hash_shift = ((s->hash_bits + MIN_MATCH-1) / MIN_MATCH); |
448 | |
449 | s->window = (Bytef *) ZALLOC(strm, s->w_size, 2*sizeof(Byte)); |
450 | s->prev = (Posf *) ZALLOC(strm, s->w_size, sizeof(Pos)); |
451 | s->head = (Posf *) ZALLOC(strm, s->hash_size, sizeof(Pos)); |
452 | |
453 | s->high_water = 0; /* nothing written to s->window yet */ |
454 | |
455 | s->lit_bufsize = 1 << (memLevel + 6); /* 16K elements by default */ |
456 | |
457 | /* We overlay pending_buf and sym_buf. This works since the average size |
458 | * for length/distance pairs over any compressed block is assured to be 31 |
459 | * bits or less. |
460 | * |
461 | * Analysis: The longest fixed codes are a length code of 8 bits plus 5 |
462 | * extra bits, for lengths 131 to 257. The longest fixed distance codes are |
463 | * 5 bits plus 13 extra bits, for distances 16385 to 32768. The longest |
464 | * possible fixed-codes length/distance pair is then 31 bits total. |
465 | * |
466 | * sym_buf starts one-fourth of the way into pending_buf. So there are |
467 | * three bytes in sym_buf for every four bytes in pending_buf. Each symbol |
468 | * in sym_buf is three bytes -- two for the distance and one for the |
469 | * literal/length. As each symbol is consumed, the pointer to the next |
470 | * sym_buf value to read moves forward three bytes. From that symbol, up to |
471 | * 31 bits are written to pending_buf. The closest the written pending_buf |
472 | * bits gets to the next sym_buf symbol to read is just before the last |
473 | * code is written. At that time, 31*(n - 2) bits have been written, just |
474 | * after 24*(n - 2) bits have been consumed from sym_buf. sym_buf starts at |
475 | * 8*n bits into pending_buf. (Note that the symbol buffer fills when n - 1 |
476 | * symbols are written.) The closest the writing gets to what is unread is |
477 | * then n + 14 bits. Here n is lit_bufsize, which is 16384 by default, and |
478 | * can range from 128 to 32768. |
479 | * |
480 | * Therefore, at a minimum, there are 142 bits of space between what is |
481 | * written and what is read in the overlain buffers, so the symbols cannot |
482 | * be overwritten by the compressed data. That space is actually 139 bits, |
483 | * due to the three-bit fixed-code block header. |
484 | * |
485 | * That covers the case where either Z_FIXED is specified, forcing fixed |
486 | * codes, or when the use of fixed codes is chosen, because that choice |
487 | * results in a smaller compressed block than dynamic codes. That latter |
488 | * condition then assures that the above analysis also covers all dynamic |
489 | * blocks. A dynamic-code block will only be chosen to be emitted if it has |
490 | * fewer bits than a fixed-code block would for the same set of symbols. |
491 | * Therefore its average symbol length is assured to be less than 31. So |
492 | * the compressed data for a dynamic block also cannot overwrite the |
493 | * symbols from which it is being constructed. |
494 | */ |
495 | |
496 | s->pending_buf = (uchf *) ZALLOC(strm, s->lit_bufsize, 4); |
497 | s->pending_buf_size = (ulg)s->lit_bufsize * 4; |
498 | |
499 | if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL || |
500 | s->pending_buf == Z_NULL) { |
501 | s->status = FINISH_STATE; |
502 | strm->msg = ERR_MSG(Z_MEM_ERROR); |
503 | deflateEnd (strm); |
504 | return Z_MEM_ERROR; |
505 | } |
506 | s->sym_buf = s->pending_buf + s->lit_bufsize; |
507 | s->sym_end = (s->lit_bufsize - 1) * 3; |
508 | /* We avoid equality with lit_bufsize*3 because of wraparound at 64K |
509 | * on 16 bit machines and because stored blocks are restricted to |
510 | * 64K-1 bytes. |
511 | */ |
512 | |
513 | s->level = level; |
514 | s->strategy = strategy; |
515 | s->method = (Byte)method; |
516 | |
517 | return deflateReset(strm); |
518 | } |
519 | |
520 | /* ========================================================================= |
521 | * Check for a valid deflate stream state. Return 0 if ok, 1 if not. |
522 | */ |
523 | local int deflateStateCheck(z_streamp strm) { |
524 | deflate_state *s; |
525 | if (strm == Z_NULL || |
526 | strm->zalloc == (alloc_func)0 || strm->zfree == (free_func)0) |
527 | return 1; |
528 | s = strm->state; |
529 | if (s == Z_NULL || s->strm != strm || (s->status != INIT_STATE && |
530 | #ifdef GZIP |
531 | s->status != GZIP_STATE && |
532 | #endif |
533 | s->status != EXTRA_STATE && |
534 | s->status != NAME_STATE && |
535 | s->status != COMMENT_STATE && |
536 | s->status != HCRC_STATE && |
537 | s->status != BUSY_STATE && |
538 | s->status != FINISH_STATE)) |
539 | return 1; |
540 | return 0; |
541 | } |
542 | |
543 | /* ========================================================================= */ |
544 | int ZEXPORT deflateSetDictionary(z_streamp strm, const Bytef *dictionary, |
545 | uInt dictLength) { |
546 | deflate_state *s; |
547 | uInt str, n; |
548 | int wrap; |
549 | unsigned avail; |
550 | z_const unsigned char *next; |
551 | |
552 | if (deflateStateCheck(strm) || dictionary == Z_NULL) |
553 | return Z_STREAM_ERROR; |
554 | s = strm->state; |
555 | wrap = s->wrap; |
556 | if (wrap == 2 || (wrap == 1 && s->status != INIT_STATE) || s->lookahead) |
557 | return Z_STREAM_ERROR; |
558 | |
559 | /* when using zlib wrappers, compute Adler-32 for provided dictionary */ |
560 | if (wrap == 1) |
561 | strm->adler = adler32(strm->adler, dictionary, dictLength); |
562 | s->wrap = 0; /* avoid computing Adler-32 in read_buf */ |
563 | |
564 | /* if dictionary would fill window, just replace the history */ |
565 | if (dictLength >= s->w_size) { |
566 | if (wrap == 0) { /* already empty otherwise */ |
567 | CLEAR_HASH(s); |
568 | s->strstart = 0; |
569 | s->block_start = 0L; |
570 | s->insert = 0; |
571 | } |
572 | dictionary += dictLength - s->w_size; /* use the tail */ |
573 | dictLength = s->w_size; |
574 | } |
575 | |
576 | /* insert dictionary into window and hash */ |
577 | avail = strm->avail_in; |
578 | next = strm->next_in; |
579 | strm->avail_in = dictLength; |
580 | strm->next_in = (z_const Bytef *)dictionary; |
581 | fill_window(s); |
582 | while (s->lookahead >= MIN_MATCH) { |
583 | str = s->strstart; |
584 | n = s->lookahead - (MIN_MATCH-1); |
585 | do { |
586 | UPDATE_HASH(s, s->ins_h, s->window[str + MIN_MATCH-1]); |
587 | #ifndef FASTEST |
588 | s->prev[str & s->w_mask] = s->head[s->ins_h]; |
589 | #endif |
590 | s->head[s->ins_h] = (Pos)str; |
591 | str++; |
592 | } while (--n); |
593 | s->strstart = str; |
594 | s->lookahead = MIN_MATCH-1; |
595 | fill_window(s); |
596 | } |
597 | s->strstart += s->lookahead; |
598 | s->block_start = (long)s->strstart; |
599 | s->insert = s->lookahead; |
600 | s->lookahead = 0; |
601 | s->match_length = s->prev_length = MIN_MATCH-1; |
602 | s->match_available = 0; |
603 | strm->next_in = next; |
604 | strm->avail_in = avail; |
605 | s->wrap = wrap; |
606 | return Z_OK; |
607 | } |
608 | |
609 | /* ========================================================================= */ |
610 | int ZEXPORT deflateGetDictionary(z_streamp strm, Bytef *dictionary, |
611 | uInt *dictLength) { |
612 | deflate_state *s; |
613 | uInt len; |
614 | |
615 | if (deflateStateCheck(strm)) |
616 | return Z_STREAM_ERROR; |
617 | s = strm->state; |
618 | len = s->strstart + s->lookahead; |
619 | if (len > s->w_size) |
620 | len = s->w_size; |
621 | if (dictionary != Z_NULL && len) |
622 | zmemcpy(dictionary, s->window + s->strstart + s->lookahead - len, len); |
623 | if (dictLength != Z_NULL) |
624 | *dictLength = len; |
625 | return Z_OK; |
626 | } |
627 | |
628 | /* ========================================================================= */ |
629 | int ZEXPORT deflateResetKeep(z_streamp strm) { |
630 | deflate_state *s; |
631 | |
632 | if (deflateStateCheck(strm)) { |
633 | return Z_STREAM_ERROR; |
634 | } |
635 | |
636 | strm->total_in = strm->total_out = 0; |
637 | strm->msg = Z_NULL; /* use zfree if we ever allocate msg dynamically */ |
638 | strm->data_type = Z_UNKNOWN; |
639 | |
640 | s = (deflate_state *)strm->state; |
641 | s->pending = 0; |
642 | s->pending_out = s->pending_buf; |
643 | |
644 | if (s->wrap < 0) { |
645 | s->wrap = -s->wrap; /* was made negative by deflate(..., Z_FINISH); */ |
646 | } |
647 | s->status = |
648 | #ifdef GZIP |
649 | s->wrap == 2 ? GZIP_STATE : |
650 | #endif |
651 | INIT_STATE; |
652 | strm->adler = |
653 | #ifdef GZIP |
654 | s->wrap == 2 ? crc32(0L, Z_NULL, 0) : |
655 | #endif |
656 | adler32(0L, Z_NULL, 0); |
657 | s->last_flush = -2; |
658 | |
659 | _tr_init(s); |
660 | |
661 | return Z_OK; |
662 | } |
663 | |
664 | /* =========================================================================== |
665 | * Initialize the "longest match" routines for a new zlib stream |
666 | */ |
667 | local void lm_init(deflate_state *s) { |
668 | s->window_size = (ulg)2L*s->w_size; |
669 | |
670 | CLEAR_HASH(s); |
671 | |
672 | /* Set the default configuration parameters: |
673 | */ |
674 | s->max_lazy_match = configuration_table[s->level].max_lazy; |
675 | s->good_match = configuration_table[s->level].good_length; |
676 | s->nice_match = configuration_table[s->level].nice_length; |
677 | s->max_chain_length = configuration_table[s->level].max_chain; |
678 | |
679 | s->strstart = 0; |
680 | s->block_start = 0L; |
681 | s->lookahead = 0; |
682 | s->insert = 0; |
683 | s->match_length = s->prev_length = MIN_MATCH-1; |
684 | s->match_available = 0; |
685 | s->ins_h = 0; |
686 | } |
687 | |
688 | /* ========================================================================= */ |
689 | int ZEXPORT deflateReset(z_streamp strm) { |
690 | int ret; |
691 | |
692 | ret = deflateResetKeep(strm); |
693 | if (ret == Z_OK) |
694 | lm_init(strm->state); |
695 | return ret; |
696 | } |
697 | |
698 | /* ========================================================================= */ |
699 | int ZEXPORT (z_streamp strm, gz_headerp head) { |
700 | if (deflateStateCheck(strm) || strm->state->wrap != 2) |
701 | return Z_STREAM_ERROR; |
702 | strm->state->gzhead = head; |
703 | return Z_OK; |
704 | } |
705 | |
706 | /* ========================================================================= */ |
707 | int ZEXPORT deflatePending(z_streamp strm, unsigned *pending, int *bits) { |
708 | if (deflateStateCheck(strm)) return Z_STREAM_ERROR; |
709 | if (pending != Z_NULL) |
710 | *pending = strm->state->pending; |
711 | if (bits != Z_NULL) |
712 | *bits = strm->state->bi_valid; |
713 | return Z_OK; |
714 | } |
715 | |
716 | /* ========================================================================= */ |
717 | int ZEXPORT deflatePrime(z_streamp strm, int bits, int value) { |
718 | deflate_state *s; |
719 | int put; |
720 | |
721 | if (deflateStateCheck(strm)) return Z_STREAM_ERROR; |
722 | s = strm->state; |
723 | if (bits < 0 || bits > 16 || |
724 | s->sym_buf < s->pending_out + ((Buf_size + 7) >> 3)) |
725 | return Z_BUF_ERROR; |
726 | do { |
727 | put = Buf_size - s->bi_valid; |
728 | if (put > bits) |
729 | put = bits; |
730 | s->bi_buf |= (ush)((value & ((1 << put) - 1)) << s->bi_valid); |
731 | s->bi_valid += put; |
732 | _tr_flush_bits(s); |
733 | value >>= put; |
734 | bits -= put; |
735 | } while (bits); |
736 | return Z_OK; |
737 | } |
738 | |
739 | /* ========================================================================= */ |
740 | int ZEXPORT deflateParams(z_streamp strm, int level, int strategy) { |
741 | deflate_state *s; |
742 | compress_func func; |
743 | |
744 | if (deflateStateCheck(strm)) return Z_STREAM_ERROR; |
745 | s = strm->state; |
746 | |
747 | #ifdef FASTEST |
748 | if (level != 0) level = 1; |
749 | #else |
750 | if (level == Z_DEFAULT_COMPRESSION) level = 6; |
751 | #endif |
752 | if (level < 0 || level > 9 || strategy < 0 || strategy > Z_FIXED) { |
753 | return Z_STREAM_ERROR; |
754 | } |
755 | func = configuration_table[s->level].func; |
756 | |
757 | if ((strategy != s->strategy || func != configuration_table[level].func) && |
758 | s->last_flush != -2) { |
759 | /* Flush the last buffer: */ |
760 | int err = deflate(strm, Z_BLOCK); |
761 | if (err == Z_STREAM_ERROR) |
762 | return err; |
763 | if (strm->avail_in || (s->strstart - s->block_start) + s->lookahead) |
764 | return Z_BUF_ERROR; |
765 | } |
766 | if (s->level != level) { |
767 | if (s->level == 0 && s->matches != 0) { |
768 | if (s->matches == 1) |
769 | slide_hash(s); |
770 | else |
771 | CLEAR_HASH(s); |
772 | s->matches = 0; |
773 | } |
774 | s->level = level; |
775 | s->max_lazy_match = configuration_table[level].max_lazy; |
776 | s->good_match = configuration_table[level].good_length; |
777 | s->nice_match = configuration_table[level].nice_length; |
778 | s->max_chain_length = configuration_table[level].max_chain; |
779 | } |
780 | s->strategy = strategy; |
781 | return Z_OK; |
782 | } |
783 | |
784 | /* ========================================================================= */ |
785 | int ZEXPORT deflateTune(z_streamp strm, int good_length, int max_lazy, |
786 | int nice_length, int max_chain) { |
787 | deflate_state *s; |
788 | |
789 | if (deflateStateCheck(strm)) return Z_STREAM_ERROR; |
790 | s = strm->state; |
791 | s->good_match = (uInt)good_length; |
792 | s->max_lazy_match = (uInt)max_lazy; |
793 | s->nice_match = nice_length; |
794 | s->max_chain_length = (uInt)max_chain; |
795 | return Z_OK; |
796 | } |
797 | |
798 | /* ========================================================================= |
799 | * For the default windowBits of 15 and memLevel of 8, this function returns a |
800 | * close to exact, as well as small, upper bound on the compressed size. This |
801 | * is an expansion of ~0.03%, plus a small constant. |
802 | * |
803 | * For any setting other than those defaults for windowBits and memLevel, one |
804 | * of two worst case bounds is returned. This is at most an expansion of ~4% or |
805 | * ~13%, plus a small constant. |
806 | * |
807 | * Both the 0.03% and 4% derive from the overhead of stored blocks. The first |
808 | * one is for stored blocks of 16383 bytes (memLevel == 8), whereas the second |
809 | * is for stored blocks of 127 bytes (the worst case memLevel == 1). The |
810 | * expansion results from five bytes of header for each stored block. |
811 | * |
812 | * The larger expansion of 13% results from a window size less than or equal to |
813 | * the symbols buffer size (windowBits <= memLevel + 7). In that case some of |
814 | * the data being compressed may have slid out of the sliding window, impeding |
815 | * a stored block from being emitted. Then the only choice is a fixed or |
816 | * dynamic block, where a fixed block limits the maximum expansion to 9 bits |
817 | * per 8-bit byte, plus 10 bits for every block. The smallest block size for |
818 | * which this can occur is 255 (memLevel == 2). |
819 | * |
820 | * Shifts are used to approximate divisions, for speed. |
821 | */ |
822 | uLong ZEXPORT deflateBound(z_streamp strm, uLong sourceLen) { |
823 | deflate_state *s; |
824 | uLong fixedlen, storelen, wraplen; |
825 | |
826 | /* upper bound for fixed blocks with 9-bit literals and length 255 |
827 | (memLevel == 2, which is the lowest that may not use stored blocks) -- |
828 | ~13% overhead plus a small constant */ |
829 | fixedlen = sourceLen + (sourceLen >> 3) + (sourceLen >> 8) + |
830 | (sourceLen >> 9) + 4; |
831 | |
832 | /* upper bound for stored blocks with length 127 (memLevel == 1) -- |
833 | ~4% overhead plus a small constant */ |
834 | storelen = sourceLen + (sourceLen >> 5) + (sourceLen >> 7) + |
835 | (sourceLen >> 11) + 7; |
836 | |
837 | /* if can't get parameters, return larger bound plus a zlib wrapper */ |
838 | if (deflateStateCheck(strm)) |
839 | return (fixedlen > storelen ? fixedlen : storelen) + 6; |
840 | |
841 | /* compute wrapper length */ |
842 | s = strm->state; |
843 | switch (s->wrap) { |
844 | case 0: /* raw deflate */ |
845 | wraplen = 0; |
846 | break; |
847 | case 1: /* zlib wrapper */ |
848 | wraplen = 6 + (s->strstart ? 4 : 0); |
849 | break; |
850 | #ifdef GZIP |
851 | case 2: /* gzip wrapper */ |
852 | wraplen = 18; |
853 | if (s->gzhead != Z_NULL) { /* user-supplied gzip header */ |
854 | Bytef *str; |
855 | if (s->gzhead->extra != Z_NULL) |
856 | wraplen += 2 + s->gzhead->extra_len; |
857 | str = s->gzhead->name; |
858 | if (str != Z_NULL) |
859 | do { |
860 | wraplen++; |
861 | } while (*str++); |
862 | str = s->gzhead->comment; |
863 | if (str != Z_NULL) |
864 | do { |
865 | wraplen++; |
866 | } while (*str++); |
867 | if (s->gzhead->hcrc) |
868 | wraplen += 2; |
869 | } |
870 | break; |
871 | #endif |
872 | default: /* for compiler happiness */ |
873 | wraplen = 6; |
874 | } |
875 | |
876 | /* if not default parameters, return one of the conservative bounds */ |
877 | if (s->w_bits != 15 || s->hash_bits != 8 + 7) |
878 | return (s->w_bits <= s->hash_bits && s->level ? fixedlen : storelen) + |
879 | wraplen; |
880 | |
881 | /* default settings: return tight bound for that case -- ~0.03% overhead |
882 | plus a small constant */ |
883 | return sourceLen + (sourceLen >> 12) + (sourceLen >> 14) + |
884 | (sourceLen >> 25) + 13 - 6 + wraplen; |
885 | } |
886 | |
887 | /* ========================================================================= |
888 | * Put a short in the pending buffer. The 16-bit value is put in MSB order. |
889 | * IN assertion: the stream state is correct and there is enough room in |
890 | * pending_buf. |
891 | */ |
892 | local void putShortMSB(deflate_state *s, uInt b) { |
893 | put_byte(s, (Byte)(b >> 8)); |
894 | put_byte(s, (Byte)(b & 0xff)); |
895 | } |
896 | |
897 | /* ========================================================================= |
898 | * Flush as much pending output as possible. All deflate() output, except for |
899 | * some deflate_stored() output, goes through this function so some |
900 | * applications may wish to modify it to avoid allocating a large |
901 | * strm->next_out buffer and copying into it. (See also read_buf()). |
902 | */ |
903 | local void flush_pending(z_streamp strm) { |
904 | unsigned len; |
905 | deflate_state *s = strm->state; |
906 | |
907 | _tr_flush_bits(s); |
908 | len = s->pending; |
909 | if (len > strm->avail_out) len = strm->avail_out; |
910 | if (len == 0) return; |
911 | |
912 | zmemcpy(strm->next_out, s->pending_out, len); |
913 | strm->next_out += len; |
914 | s->pending_out += len; |
915 | strm->total_out += len; |
916 | strm->avail_out -= len; |
917 | s->pending -= len; |
918 | if (s->pending == 0) { |
919 | s->pending_out = s->pending_buf; |
920 | } |
921 | } |
922 | |
923 | /* =========================================================================== |
924 | * Update the header CRC with the bytes s->pending_buf[beg..s->pending - 1]. |
925 | */ |
926 | #define HCRC_UPDATE(beg) \ |
927 | do { \ |
928 | if (s->gzhead->hcrc && s->pending > (beg)) \ |
929 | strm->adler = crc32(strm->adler, s->pending_buf + (beg), \ |
930 | s->pending - (beg)); \ |
931 | } while (0) |
932 | |
933 | /* ========================================================================= */ |
934 | int ZEXPORT deflate(z_streamp strm, int flush) { |
935 | int old_flush; /* value of flush param for previous deflate call */ |
936 | deflate_state *s; |
937 | |
938 | if (deflateStateCheck(strm) || flush > Z_BLOCK || flush < 0) { |
939 | return Z_STREAM_ERROR; |
940 | } |
941 | s = strm->state; |
942 | |
943 | if (strm->next_out == Z_NULL || |
944 | (strm->avail_in != 0 && strm->next_in == Z_NULL) || |
945 | (s->status == FINISH_STATE && flush != Z_FINISH)) { |
946 | ERR_RETURN(strm, Z_STREAM_ERROR); |
947 | } |
948 | if (strm->avail_out == 0) ERR_RETURN(strm, Z_BUF_ERROR); |
949 | |
950 | old_flush = s->last_flush; |
951 | s->last_flush = flush; |
952 | |
953 | /* Flush as much pending output as possible */ |
954 | if (s->pending != 0) { |
955 | flush_pending(strm); |
956 | if (strm->avail_out == 0) { |
957 | /* Since avail_out is 0, deflate will be called again with |
958 | * more output space, but possibly with both pending and |
959 | * avail_in equal to zero. There won't be anything to do, |
960 | * but this is not an error situation so make sure we |
961 | * return OK instead of BUF_ERROR at next call of deflate: |
962 | */ |
963 | s->last_flush = -1; |
964 | return Z_OK; |
965 | } |
966 | |
967 | /* Make sure there is something to do and avoid duplicate consecutive |
968 | * flushes. For repeated and useless calls with Z_FINISH, we keep |
969 | * returning Z_STREAM_END instead of Z_BUF_ERROR. |
970 | */ |
971 | } else if (strm->avail_in == 0 && RANK(flush) <= RANK(old_flush) && |
972 | flush != Z_FINISH) { |
973 | ERR_RETURN(strm, Z_BUF_ERROR); |
974 | } |
975 | |
976 | /* User must not provide more input after the first FINISH: */ |
977 | if (s->status == FINISH_STATE && strm->avail_in != 0) { |
978 | ERR_RETURN(strm, Z_BUF_ERROR); |
979 | } |
980 | |
981 | /* Write the header */ |
982 | if (s->status == INIT_STATE && s->wrap == 0) |
983 | s->status = BUSY_STATE; |
984 | if (s->status == INIT_STATE) { |
985 | /* zlib header */ |
986 | uInt = (Z_DEFLATED + ((s->w_bits - 8) << 4)) << 8; |
987 | uInt level_flags; |
988 | |
989 | if (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2) |
990 | level_flags = 0; |
991 | else if (s->level < 6) |
992 | level_flags = 1; |
993 | else if (s->level == 6) |
994 | level_flags = 2; |
995 | else |
996 | level_flags = 3; |
997 | header |= (level_flags << 6); |
998 | if (s->strstart != 0) header |= PRESET_DICT; |
999 | header += 31 - (header % 31); |
1000 | |
1001 | putShortMSB(s, header); |
1002 | |
1003 | /* Save the adler32 of the preset dictionary: */ |
1004 | if (s->strstart != 0) { |
1005 | putShortMSB(s, (uInt)(strm->adler >> 16)); |
1006 | putShortMSB(s, (uInt)(strm->adler & 0xffff)); |
1007 | } |
1008 | strm->adler = adler32(0L, Z_NULL, 0); |
1009 | s->status = BUSY_STATE; |
1010 | |
1011 | /* Compression must start with an empty pending buffer */ |
1012 | flush_pending(strm); |
1013 | if (s->pending != 0) { |
1014 | s->last_flush = -1; |
1015 | return Z_OK; |
1016 | } |
1017 | } |
1018 | #ifdef GZIP |
1019 | if (s->status == GZIP_STATE) { |
1020 | /* gzip header */ |
1021 | strm->adler = crc32(0L, Z_NULL, 0); |
1022 | put_byte(s, 31); |
1023 | put_byte(s, 139); |
1024 | put_byte(s, 8); |
1025 | if (s->gzhead == Z_NULL) { |
1026 | put_byte(s, 0); |
1027 | put_byte(s, 0); |
1028 | put_byte(s, 0); |
1029 | put_byte(s, 0); |
1030 | put_byte(s, 0); |
1031 | put_byte(s, s->level == 9 ? 2 : |
1032 | (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ? |
1033 | 4 : 0)); |
1034 | put_byte(s, OS_CODE); |
1035 | s->status = BUSY_STATE; |
1036 | |
1037 | /* Compression must start with an empty pending buffer */ |
1038 | flush_pending(strm); |
1039 | if (s->pending != 0) { |
1040 | s->last_flush = -1; |
1041 | return Z_OK; |
1042 | } |
1043 | } |
1044 | else { |
1045 | put_byte(s, (s->gzhead->text ? 1 : 0) + |
1046 | (s->gzhead->hcrc ? 2 : 0) + |
1047 | (s->gzhead->extra == Z_NULL ? 0 : 4) + |
1048 | (s->gzhead->name == Z_NULL ? 0 : 8) + |
1049 | (s->gzhead->comment == Z_NULL ? 0 : 16) |
1050 | ); |
1051 | put_byte(s, (Byte)(s->gzhead->time & 0xff)); |
1052 | put_byte(s, (Byte)((s->gzhead->time >> 8) & 0xff)); |
1053 | put_byte(s, (Byte)((s->gzhead->time >> 16) & 0xff)); |
1054 | put_byte(s, (Byte)((s->gzhead->time >> 24) & 0xff)); |
1055 | put_byte(s, s->level == 9 ? 2 : |
1056 | (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ? |
1057 | 4 : 0)); |
1058 | put_byte(s, s->gzhead->os & 0xff); |
1059 | if (s->gzhead->extra != Z_NULL) { |
1060 | put_byte(s, s->gzhead->extra_len & 0xff); |
1061 | put_byte(s, (s->gzhead->extra_len >> 8) & 0xff); |
1062 | } |
1063 | if (s->gzhead->hcrc) |
1064 | strm->adler = crc32(strm->adler, s->pending_buf, |
1065 | s->pending); |
1066 | s->gzindex = 0; |
1067 | s->status = EXTRA_STATE; |
1068 | } |
1069 | } |
1070 | if (s->status == EXTRA_STATE) { |
1071 | if (s->gzhead->extra != Z_NULL) { |
1072 | ulg beg = s->pending; /* start of bytes to update crc */ |
1073 | uInt left = (s->gzhead->extra_len & 0xffff) - s->gzindex; |
1074 | while (s->pending + left > s->pending_buf_size) { |
1075 | uInt copy = s->pending_buf_size - s->pending; |
1076 | zmemcpy(s->pending_buf + s->pending, |
1077 | s->gzhead->extra + s->gzindex, copy); |
1078 | s->pending = s->pending_buf_size; |
1079 | HCRC_UPDATE(beg); |
1080 | s->gzindex += copy; |
1081 | flush_pending(strm); |
1082 | if (s->pending != 0) { |
1083 | s->last_flush = -1; |
1084 | return Z_OK; |
1085 | } |
1086 | beg = 0; |
1087 | left -= copy; |
1088 | } |
1089 | zmemcpy(s->pending_buf + s->pending, |
1090 | s->gzhead->extra + s->gzindex, left); |
1091 | s->pending += left; |
1092 | HCRC_UPDATE(beg); |
1093 | s->gzindex = 0; |
1094 | } |
1095 | s->status = NAME_STATE; |
1096 | } |
1097 | if (s->status == NAME_STATE) { |
1098 | if (s->gzhead->name != Z_NULL) { |
1099 | ulg beg = s->pending; /* start of bytes to update crc */ |
1100 | int val; |
1101 | do { |
1102 | if (s->pending == s->pending_buf_size) { |
1103 | HCRC_UPDATE(beg); |
1104 | flush_pending(strm); |
1105 | if (s->pending != 0) { |
1106 | s->last_flush = -1; |
1107 | return Z_OK; |
1108 | } |
1109 | beg = 0; |
1110 | } |
1111 | val = s->gzhead->name[s->gzindex++]; |
1112 | put_byte(s, val); |
1113 | } while (val != 0); |
1114 | HCRC_UPDATE(beg); |
1115 | s->gzindex = 0; |
1116 | } |
1117 | s->status = COMMENT_STATE; |
1118 | } |
1119 | if (s->status == COMMENT_STATE) { |
1120 | if (s->gzhead->comment != Z_NULL) { |
1121 | ulg beg = s->pending; /* start of bytes to update crc */ |
1122 | int val; |
1123 | do { |
1124 | if (s->pending == s->pending_buf_size) { |
1125 | HCRC_UPDATE(beg); |
1126 | flush_pending(strm); |
1127 | if (s->pending != 0) { |
1128 | s->last_flush = -1; |
1129 | return Z_OK; |
1130 | } |
1131 | beg = 0; |
1132 | } |
1133 | val = s->gzhead->comment[s->gzindex++]; |
1134 | put_byte(s, val); |
1135 | } while (val != 0); |
1136 | HCRC_UPDATE(beg); |
1137 | } |
1138 | s->status = HCRC_STATE; |
1139 | } |
1140 | if (s->status == HCRC_STATE) { |
1141 | if (s->gzhead->hcrc) { |
1142 | if (s->pending + 2 > s->pending_buf_size) { |
1143 | flush_pending(strm); |
1144 | if (s->pending != 0) { |
1145 | s->last_flush = -1; |
1146 | return Z_OK; |
1147 | } |
1148 | } |
1149 | put_byte(s, (Byte)(strm->adler & 0xff)); |
1150 | put_byte(s, (Byte)((strm->adler >> 8) & 0xff)); |
1151 | strm->adler = crc32(0L, Z_NULL, 0); |
1152 | } |
1153 | s->status = BUSY_STATE; |
1154 | |
1155 | /* Compression must start with an empty pending buffer */ |
1156 | flush_pending(strm); |
1157 | if (s->pending != 0) { |
1158 | s->last_flush = -1; |
1159 | return Z_OK; |
1160 | } |
1161 | } |
1162 | #endif |
1163 | |
1164 | /* Start a new block or continue the current one. |
1165 | */ |
1166 | if (strm->avail_in != 0 || s->lookahead != 0 || |
1167 | (flush != Z_NO_FLUSH && s->status != FINISH_STATE)) { |
1168 | block_state bstate; |
1169 | |
1170 | bstate = s->level == 0 ? deflate_stored(s, flush) : |
1171 | s->strategy == Z_HUFFMAN_ONLY ? deflate_huff(s, flush) : |
1172 | s->strategy == Z_RLE ? deflate_rle(s, flush) : |
1173 | (*(configuration_table[s->level].func))(s, flush); |
1174 | |
1175 | if (bstate == finish_started || bstate == finish_done) { |
1176 | s->status = FINISH_STATE; |
1177 | } |
1178 | if (bstate == need_more || bstate == finish_started) { |
1179 | if (strm->avail_out == 0) { |
1180 | s->last_flush = -1; /* avoid BUF_ERROR next call, see above */ |
1181 | } |
1182 | return Z_OK; |
1183 | /* If flush != Z_NO_FLUSH && avail_out == 0, the next call |
1184 | * of deflate should use the same flush parameter to make sure |
1185 | * that the flush is complete. So we don't have to output an |
1186 | * empty block here, this will be done at next call. This also |
1187 | * ensures that for a very small output buffer, we emit at most |
1188 | * one empty block. |
1189 | */ |
1190 | } |
1191 | if (bstate == block_done) { |
1192 | if (flush == Z_PARTIAL_FLUSH) { |
1193 | _tr_align(s); |
1194 | } else if (flush != Z_BLOCK) { /* FULL_FLUSH or SYNC_FLUSH */ |
1195 | _tr_stored_block(s, (char*)0, 0L, 0); |
1196 | /* For a full flush, this empty block will be recognized |
1197 | * as a special marker by inflate_sync(). |
1198 | */ |
1199 | if (flush == Z_FULL_FLUSH) { |
1200 | CLEAR_HASH(s); /* forget history */ |
1201 | if (s->lookahead == 0) { |
1202 | s->strstart = 0; |
1203 | s->block_start = 0L; |
1204 | s->insert = 0; |
1205 | } |
1206 | } |
1207 | } |
1208 | flush_pending(strm); |
1209 | if (strm->avail_out == 0) { |
1210 | s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */ |
1211 | return Z_OK; |
1212 | } |
1213 | } |
1214 | } |
1215 | |
1216 | if (flush != Z_FINISH) return Z_OK; |
1217 | if (s->wrap <= 0) return Z_STREAM_END; |
1218 | |
1219 | /* Write the trailer */ |
1220 | #ifdef GZIP |
1221 | if (s->wrap == 2) { |
1222 | put_byte(s, (Byte)(strm->adler & 0xff)); |
1223 | put_byte(s, (Byte)((strm->adler >> 8) & 0xff)); |
1224 | put_byte(s, (Byte)((strm->adler >> 16) & 0xff)); |
1225 | put_byte(s, (Byte)((strm->adler >> 24) & 0xff)); |
1226 | put_byte(s, (Byte)(strm->total_in & 0xff)); |
1227 | put_byte(s, (Byte)((strm->total_in >> 8) & 0xff)); |
1228 | put_byte(s, (Byte)((strm->total_in >> 16) & 0xff)); |
1229 | put_byte(s, (Byte)((strm->total_in >> 24) & 0xff)); |
1230 | } |
1231 | else |
1232 | #endif |
1233 | { |
1234 | putShortMSB(s, (uInt)(strm->adler >> 16)); |
1235 | putShortMSB(s, (uInt)(strm->adler & 0xffff)); |
1236 | } |
1237 | flush_pending(strm); |
1238 | /* If avail_out is zero, the application will call deflate again |
1239 | * to flush the rest. |
1240 | */ |
1241 | if (s->wrap > 0) s->wrap = -s->wrap; /* write the trailer only once! */ |
1242 | return s->pending != 0 ? Z_OK : Z_STREAM_END; |
1243 | } |
1244 | |
1245 | /* ========================================================================= */ |
1246 | int ZEXPORT deflateEnd(z_streamp strm) { |
1247 | int status; |
1248 | |
1249 | if (deflateStateCheck(strm)) return Z_STREAM_ERROR; |
1250 | |
1251 | status = strm->state->status; |
1252 | |
1253 | /* Deallocate in reverse order of allocations: */ |
1254 | TRY_FREE(strm, strm->state->pending_buf); |
1255 | TRY_FREE(strm, strm->state->head); |
1256 | TRY_FREE(strm, strm->state->prev); |
1257 | TRY_FREE(strm, strm->state->window); |
1258 | |
1259 | ZFREE(strm, strm->state); |
1260 | strm->state = Z_NULL; |
1261 | |
1262 | return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK; |
1263 | } |
1264 | |
1265 | /* ========================================================================= |
1266 | * Copy the source state to the destination state. |
1267 | * To simplify the source, this is not supported for 16-bit MSDOS (which |
1268 | * doesn't have enough memory anyway to duplicate compression states). |
1269 | */ |
1270 | int ZEXPORT deflateCopy(z_streamp dest, z_streamp source) { |
1271 | #ifdef MAXSEG_64K |
1272 | (void)dest; |
1273 | (void)source; |
1274 | return Z_STREAM_ERROR; |
1275 | #else |
1276 | deflate_state *ds; |
1277 | deflate_state *ss; |
1278 | |
1279 | |
1280 | if (deflateStateCheck(source) || dest == Z_NULL) { |
1281 | return Z_STREAM_ERROR; |
1282 | } |
1283 | |
1284 | ss = source->state; |
1285 | |
1286 | zmemcpy((voidpf)dest, (voidpf)source, sizeof(z_stream)); |
1287 | |
1288 | ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state)); |
1289 | if (ds == Z_NULL) return Z_MEM_ERROR; |
1290 | dest->state = (struct internal_state FAR *) ds; |
1291 | zmemcpy((voidpf)ds, (voidpf)ss, sizeof(deflate_state)); |
1292 | ds->strm = dest; |
1293 | |
1294 | ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte)); |
1295 | ds->prev = (Posf *) ZALLOC(dest, ds->w_size, sizeof(Pos)); |
1296 | ds->head = (Posf *) ZALLOC(dest, ds->hash_size, sizeof(Pos)); |
1297 | ds->pending_buf = (uchf *) ZALLOC(dest, ds->lit_bufsize, 4); |
1298 | |
1299 | if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL || |
1300 | ds->pending_buf == Z_NULL) { |
1301 | deflateEnd (dest); |
1302 | return Z_MEM_ERROR; |
1303 | } |
1304 | /* following zmemcpy do not work for 16-bit MSDOS */ |
1305 | zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte)); |
1306 | zmemcpy((voidpf)ds->prev, (voidpf)ss->prev, ds->w_size * sizeof(Pos)); |
1307 | zmemcpy((voidpf)ds->head, (voidpf)ss->head, ds->hash_size * sizeof(Pos)); |
1308 | zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size); |
1309 | |
1310 | ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf); |
1311 | ds->sym_buf = ds->pending_buf + ds->lit_bufsize; |
1312 | |
1313 | ds->l_desc.dyn_tree = ds->dyn_ltree; |
1314 | ds->d_desc.dyn_tree = ds->dyn_dtree; |
1315 | ds->bl_desc.dyn_tree = ds->bl_tree; |
1316 | |
1317 | return Z_OK; |
1318 | #endif /* MAXSEG_64K */ |
1319 | } |
1320 | |
1321 | #ifndef FASTEST |
1322 | /* =========================================================================== |
1323 | * Set match_start to the longest match starting at the given string and |
1324 | * return its length. Matches shorter or equal to prev_length are discarded, |
1325 | * in which case the result is equal to prev_length and match_start is |
1326 | * garbage. |
1327 | * IN assertions: cur_match is the head of the hash chain for the current |
1328 | * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1 |
1329 | * OUT assertion: the match length is not greater than s->lookahead. |
1330 | */ |
1331 | local uInt longest_match(deflate_state *s, IPos cur_match) { |
1332 | unsigned chain_length = s->max_chain_length;/* max hash chain length */ |
1333 | register Bytef *scan = s->window + s->strstart; /* current string */ |
1334 | register Bytef *match; /* matched string */ |
1335 | register int len; /* length of current match */ |
1336 | int best_len = (int)s->prev_length; /* best match length so far */ |
1337 | int nice_match = s->nice_match; /* stop if match long enough */ |
1338 | IPos limit = s->strstart > (IPos)MAX_DIST(s) ? |
1339 | s->strstart - (IPos)MAX_DIST(s) : NIL; |
1340 | /* Stop when cur_match becomes <= limit. To simplify the code, |
1341 | * we prevent matches with the string of window index 0. |
1342 | */ |
1343 | Posf *prev = s->prev; |
1344 | uInt wmask = s->w_mask; |
1345 | |
1346 | #ifdef UNALIGNED_OK |
1347 | /* Compare two bytes at a time. Note: this is not always beneficial. |
1348 | * Try with and without -DUNALIGNED_OK to check. |
1349 | */ |
1350 | register Bytef *strend = s->window + s->strstart + MAX_MATCH - 1; |
1351 | register ush scan_start = *(ushf*)scan; |
1352 | register ush scan_end = *(ushf*)(scan + best_len - 1); |
1353 | #else |
1354 | register Bytef *strend = s->window + s->strstart + MAX_MATCH; |
1355 | register Byte scan_end1 = scan[best_len - 1]; |
1356 | register Byte scan_end = scan[best_len]; |
1357 | #endif |
1358 | |
1359 | /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16. |
1360 | * It is easy to get rid of this optimization if necessary. |
1361 | */ |
1362 | Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever" ); |
1363 | |
1364 | /* Do not waste too much time if we already have a good match: */ |
1365 | if (s->prev_length >= s->good_match) { |
1366 | chain_length >>= 2; |
1367 | } |
1368 | /* Do not look for matches beyond the end of the input. This is necessary |
1369 | * to make deflate deterministic. |
1370 | */ |
1371 | if ((uInt)nice_match > s->lookahead) nice_match = (int)s->lookahead; |
1372 | |
1373 | Assert((ulg)s->strstart <= s->window_size - MIN_LOOKAHEAD, |
1374 | "need lookahead" ); |
1375 | |
1376 | do { |
1377 | Assert(cur_match < s->strstart, "no future" ); |
1378 | match = s->window + cur_match; |
1379 | |
1380 | /* Skip to next match if the match length cannot increase |
1381 | * or if the match length is less than 2. Note that the checks below |
1382 | * for insufficient lookahead only occur occasionally for performance |
1383 | * reasons. Therefore uninitialized memory will be accessed, and |
1384 | * conditional jumps will be made that depend on those values. |
1385 | * However the length of the match is limited to the lookahead, so |
1386 | * the output of deflate is not affected by the uninitialized values. |
1387 | */ |
1388 | #if (defined(UNALIGNED_OK) && MAX_MATCH == 258) |
1389 | /* This code assumes sizeof(unsigned short) == 2. Do not use |
1390 | * UNALIGNED_OK if your compiler uses a different size. |
1391 | */ |
1392 | if (*(ushf*)(match + best_len - 1) != scan_end || |
1393 | *(ushf*)match != scan_start) continue; |
1394 | |
1395 | /* It is not necessary to compare scan[2] and match[2] since they are |
1396 | * always equal when the other bytes match, given that the hash keys |
1397 | * are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at |
1398 | * strstart + 3, + 5, up to strstart + 257. We check for insufficient |
1399 | * lookahead only every 4th comparison; the 128th check will be made |
1400 | * at strstart + 257. If MAX_MATCH-2 is not a multiple of 8, it is |
1401 | * necessary to put more guard bytes at the end of the window, or |
1402 | * to check more often for insufficient lookahead. |
1403 | */ |
1404 | Assert(scan[2] == match[2], "scan[2]?" ); |
1405 | scan++, match++; |
1406 | do { |
1407 | } while (*(ushf*)(scan += 2) == *(ushf*)(match += 2) && |
1408 | *(ushf*)(scan += 2) == *(ushf*)(match += 2) && |
1409 | *(ushf*)(scan += 2) == *(ushf*)(match += 2) && |
1410 | *(ushf*)(scan += 2) == *(ushf*)(match += 2) && |
1411 | scan < strend); |
1412 | /* The funny "do {}" generates better code on most compilers */ |
1413 | |
1414 | /* Here, scan <= window + strstart + 257 */ |
1415 | Assert(scan <= s->window + (unsigned)(s->window_size - 1), |
1416 | "wild scan" ); |
1417 | if (*scan == *match) scan++; |
1418 | |
1419 | len = (MAX_MATCH - 1) - (int)(strend - scan); |
1420 | scan = strend - (MAX_MATCH-1); |
1421 | |
1422 | #else /* UNALIGNED_OK */ |
1423 | |
1424 | if (match[best_len] != scan_end || |
1425 | match[best_len - 1] != scan_end1 || |
1426 | *match != *scan || |
1427 | *++match != scan[1]) continue; |
1428 | |
1429 | /* The check at best_len - 1 can be removed because it will be made |
1430 | * again later. (This heuristic is not always a win.) |
1431 | * It is not necessary to compare scan[2] and match[2] since they |
1432 | * are always equal when the other bytes match, given that |
1433 | * the hash keys are equal and that HASH_BITS >= 8. |
1434 | */ |
1435 | scan += 2, match++; |
1436 | Assert(*scan == *match, "match[2]?" ); |
1437 | |
1438 | /* We check for insufficient lookahead only every 8th comparison; |
1439 | * the 256th check will be made at strstart + 258. |
1440 | */ |
1441 | do { |
1442 | } while (*++scan == *++match && *++scan == *++match && |
1443 | *++scan == *++match && *++scan == *++match && |
1444 | *++scan == *++match && *++scan == *++match && |
1445 | *++scan == *++match && *++scan == *++match && |
1446 | scan < strend); |
1447 | |
1448 | Assert(scan <= s->window + (unsigned)(s->window_size - 1), |
1449 | "wild scan" ); |
1450 | |
1451 | len = MAX_MATCH - (int)(strend - scan); |
1452 | scan = strend - MAX_MATCH; |
1453 | |
1454 | #endif /* UNALIGNED_OK */ |
1455 | |
1456 | if (len > best_len) { |
1457 | s->match_start = cur_match; |
1458 | best_len = len; |
1459 | if (len >= nice_match) break; |
1460 | #ifdef UNALIGNED_OK |
1461 | scan_end = *(ushf*)(scan + best_len - 1); |
1462 | #else |
1463 | scan_end1 = scan[best_len - 1]; |
1464 | scan_end = scan[best_len]; |
1465 | #endif |
1466 | } |
1467 | } while ((cur_match = prev[cur_match & wmask]) > limit |
1468 | && --chain_length != 0); |
1469 | |
1470 | if ((uInt)best_len <= s->lookahead) return (uInt)best_len; |
1471 | return s->lookahead; |
1472 | } |
1473 | |
1474 | #else /* FASTEST */ |
1475 | |
1476 | /* --------------------------------------------------------------------------- |
1477 | * Optimized version for FASTEST only |
1478 | */ |
1479 | local uInt longest_match(deflate_state *s, IPos cur_match) { |
1480 | register Bytef *scan = s->window + s->strstart; /* current string */ |
1481 | register Bytef *match; /* matched string */ |
1482 | register int len; /* length of current match */ |
1483 | register Bytef *strend = s->window + s->strstart + MAX_MATCH; |
1484 | |
1485 | /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16. |
1486 | * It is easy to get rid of this optimization if necessary. |
1487 | */ |
1488 | Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever" ); |
1489 | |
1490 | Assert((ulg)s->strstart <= s->window_size - MIN_LOOKAHEAD, |
1491 | "need lookahead" ); |
1492 | |
1493 | Assert(cur_match < s->strstart, "no future" ); |
1494 | |
1495 | match = s->window + cur_match; |
1496 | |
1497 | /* Return failure if the match length is less than 2: |
1498 | */ |
1499 | if (match[0] != scan[0] || match[1] != scan[1]) return MIN_MATCH-1; |
1500 | |
1501 | /* The check at best_len - 1 can be removed because it will be made |
1502 | * again later. (This heuristic is not always a win.) |
1503 | * It is not necessary to compare scan[2] and match[2] since they |
1504 | * are always equal when the other bytes match, given that |
1505 | * the hash keys are equal and that HASH_BITS >= 8. |
1506 | */ |
1507 | scan += 2, match += 2; |
1508 | Assert(*scan == *match, "match[2]?" ); |
1509 | |
1510 | /* We check for insufficient lookahead only every 8th comparison; |
1511 | * the 256th check will be made at strstart + 258. |
1512 | */ |
1513 | do { |
1514 | } while (*++scan == *++match && *++scan == *++match && |
1515 | *++scan == *++match && *++scan == *++match && |
1516 | *++scan == *++match && *++scan == *++match && |
1517 | *++scan == *++match && *++scan == *++match && |
1518 | scan < strend); |
1519 | |
1520 | Assert(scan <= s->window + (unsigned)(s->window_size - 1), "wild scan" ); |
1521 | |
1522 | len = MAX_MATCH - (int)(strend - scan); |
1523 | |
1524 | if (len < MIN_MATCH) return MIN_MATCH - 1; |
1525 | |
1526 | s->match_start = cur_match; |
1527 | return (uInt)len <= s->lookahead ? (uInt)len : s->lookahead; |
1528 | } |
1529 | |
1530 | #endif /* FASTEST */ |
1531 | |
1532 | #ifdef ZLIB_DEBUG |
1533 | |
1534 | #define EQUAL 0 |
1535 | /* result of memcmp for equal strings */ |
1536 | |
1537 | /* =========================================================================== |
1538 | * Check that the match at match_start is indeed a match. |
1539 | */ |
1540 | local void check_match(deflate_state *s, IPos start, IPos match, int length) { |
1541 | /* check that the match is indeed a match */ |
1542 | if (zmemcmp(s->window + match, |
1543 | s->window + start, length) != EQUAL) { |
1544 | fprintf(stderr, " start %u, match %u, length %d\n" , |
1545 | start, match, length); |
1546 | do { |
1547 | fprintf(stderr, "%c%c" , s->window[match++], s->window[start++]); |
1548 | } while (--length != 0); |
1549 | z_error("invalid match" ); |
1550 | } |
1551 | if (z_verbose > 1) { |
1552 | fprintf(stderr,"\\[%d,%d]" , start - match, length); |
1553 | do { putc(s->window[start++], stderr); } while (--length != 0); |
1554 | } |
1555 | } |
1556 | #else |
1557 | # define check_match(s, start, match, length) |
1558 | #endif /* ZLIB_DEBUG */ |
1559 | |
1560 | /* =========================================================================== |
1561 | * Flush the current block, with given end-of-file flag. |
1562 | * IN assertion: strstart is set to the end of the current match. |
1563 | */ |
1564 | #define FLUSH_BLOCK_ONLY(s, last) { \ |
1565 | _tr_flush_block(s, (s->block_start >= 0L ? \ |
1566 | (charf *)&s->window[(unsigned)s->block_start] : \ |
1567 | (charf *)Z_NULL), \ |
1568 | (ulg)((long)s->strstart - s->block_start), \ |
1569 | (last)); \ |
1570 | s->block_start = s->strstart; \ |
1571 | flush_pending(s->strm); \ |
1572 | Tracev((stderr,"[FLUSH]")); \ |
1573 | } |
1574 | |
1575 | /* Same but force premature exit if necessary. */ |
1576 | #define FLUSH_BLOCK(s, last) { \ |
1577 | FLUSH_BLOCK_ONLY(s, last); \ |
1578 | if (s->strm->avail_out == 0) return (last) ? finish_started : need_more; \ |
1579 | } |
1580 | |
1581 | /* Maximum stored block length in deflate format (not including header). */ |
1582 | #define MAX_STORED 65535 |
1583 | |
1584 | /* Minimum of a and b. */ |
1585 | #define MIN(a, b) ((a) > (b) ? (b) : (a)) |
1586 | |
1587 | /* =========================================================================== |
1588 | * Copy without compression as much as possible from the input stream, return |
1589 | * the current block state. |
1590 | * |
1591 | * In case deflateParams() is used to later switch to a non-zero compression |
1592 | * level, s->matches (otherwise unused when storing) keeps track of the number |
1593 | * of hash table slides to perform. If s->matches is 1, then one hash table |
1594 | * slide will be done when switching. If s->matches is 2, the maximum value |
1595 | * allowed here, then the hash table will be cleared, since two or more slides |
1596 | * is the same as a clear. |
1597 | * |
1598 | * deflate_stored() is written to minimize the number of times an input byte is |
1599 | * copied. It is most efficient with large input and output buffers, which |
1600 | * maximizes the opportunities to have a single copy from next_in to next_out. |
1601 | */ |
1602 | local block_state deflate_stored(deflate_state *s, int flush) { |
1603 | /* Smallest worthy block size when not flushing or finishing. By default |
1604 | * this is 32K. This can be as small as 507 bytes for memLevel == 1. For |
1605 | * large input and output buffers, the stored block size will be larger. |
1606 | */ |
1607 | unsigned min_block = MIN(s->pending_buf_size - 5, s->w_size); |
1608 | |
1609 | /* Copy as many min_block or larger stored blocks directly to next_out as |
1610 | * possible. If flushing, copy the remaining available input to next_out as |
1611 | * stored blocks, if there is enough space. |
1612 | */ |
1613 | unsigned len, left, have, last = 0; |
1614 | unsigned used = s->strm->avail_in; |
1615 | do { |
1616 | /* Set len to the maximum size block that we can copy directly with the |
1617 | * available input data and output space. Set left to how much of that |
1618 | * would be copied from what's left in the window. |
1619 | */ |
1620 | len = MAX_STORED; /* maximum deflate stored block length */ |
1621 | have = (s->bi_valid + 42) >> 3; /* number of header bytes */ |
1622 | if (s->strm->avail_out < have) /* need room for header */ |
1623 | break; |
1624 | /* maximum stored block length that will fit in avail_out: */ |
1625 | have = s->strm->avail_out - have; |
1626 | left = s->strstart - s->block_start; /* bytes left in window */ |
1627 | if (len > (ulg)left + s->strm->avail_in) |
1628 | len = left + s->strm->avail_in; /* limit len to the input */ |
1629 | if (len > have) |
1630 | len = have; /* limit len to the output */ |
1631 | |
1632 | /* If the stored block would be less than min_block in length, or if |
1633 | * unable to copy all of the available input when flushing, then try |
1634 | * copying to the window and the pending buffer instead. Also don't |
1635 | * write an empty block when flushing -- deflate() does that. |
1636 | */ |
1637 | if (len < min_block && ((len == 0 && flush != Z_FINISH) || |
1638 | flush == Z_NO_FLUSH || |
1639 | len != left + s->strm->avail_in)) |
1640 | break; |
1641 | |
1642 | /* Make a dummy stored block in pending to get the header bytes, |
1643 | * including any pending bits. This also updates the debugging counts. |
1644 | */ |
1645 | last = flush == Z_FINISH && len == left + s->strm->avail_in ? 1 : 0; |
1646 | _tr_stored_block(s, (char *)0, 0L, last); |
1647 | |
1648 | /* Replace the lengths in the dummy stored block with len. */ |
1649 | s->pending_buf[s->pending - 4] = len; |
1650 | s->pending_buf[s->pending - 3] = len >> 8; |
1651 | s->pending_buf[s->pending - 2] = ~len; |
1652 | s->pending_buf[s->pending - 1] = ~len >> 8; |
1653 | |
1654 | /* Write the stored block header bytes. */ |
1655 | flush_pending(s->strm); |
1656 | |
1657 | #ifdef ZLIB_DEBUG |
1658 | /* Update debugging counts for the data about to be copied. */ |
1659 | s->compressed_len += len << 3; |
1660 | s->bits_sent += len << 3; |
1661 | #endif |
1662 | |
1663 | /* Copy uncompressed bytes from the window to next_out. */ |
1664 | if (left) { |
1665 | if (left > len) |
1666 | left = len; |
1667 | zmemcpy(s->strm->next_out, s->window + s->block_start, left); |
1668 | s->strm->next_out += left; |
1669 | s->strm->avail_out -= left; |
1670 | s->strm->total_out += left; |
1671 | s->block_start += left; |
1672 | len -= left; |
1673 | } |
1674 | |
1675 | /* Copy uncompressed bytes directly from next_in to next_out, updating |
1676 | * the check value. |
1677 | */ |
1678 | if (len) { |
1679 | read_buf(s->strm, s->strm->next_out, len); |
1680 | s->strm->next_out += len; |
1681 | s->strm->avail_out -= len; |
1682 | s->strm->total_out += len; |
1683 | } |
1684 | } while (last == 0); |
1685 | |
1686 | /* Update the sliding window with the last s->w_size bytes of the copied |
1687 | * data, or append all of the copied data to the existing window if less |
1688 | * than s->w_size bytes were copied. Also update the number of bytes to |
1689 | * insert in the hash tables, in the event that deflateParams() switches to |
1690 | * a non-zero compression level. |
1691 | */ |
1692 | used -= s->strm->avail_in; /* number of input bytes directly copied */ |
1693 | if (used) { |
1694 | /* If any input was used, then no unused input remains in the window, |
1695 | * therefore s->block_start == s->strstart. |
1696 | */ |
1697 | if (used >= s->w_size) { /* supplant the previous history */ |
1698 | s->matches = 2; /* clear hash */ |
1699 | zmemcpy(s->window, s->strm->next_in - s->w_size, s->w_size); |
1700 | s->strstart = s->w_size; |
1701 | s->insert = s->strstart; |
1702 | } |
1703 | else { |
1704 | if (s->window_size - s->strstart <= used) { |
1705 | /* Slide the window down. */ |
1706 | s->strstart -= s->w_size; |
1707 | zmemcpy(s->window, s->window + s->w_size, s->strstart); |
1708 | if (s->matches < 2) |
1709 | s->matches++; /* add a pending slide_hash() */ |
1710 | if (s->insert > s->strstart) |
1711 | s->insert = s->strstart; |
1712 | } |
1713 | zmemcpy(s->window + s->strstart, s->strm->next_in - used, used); |
1714 | s->strstart += used; |
1715 | s->insert += MIN(used, s->w_size - s->insert); |
1716 | } |
1717 | s->block_start = s->strstart; |
1718 | } |
1719 | if (s->high_water < s->strstart) |
1720 | s->high_water = s->strstart; |
1721 | |
1722 | /* If the last block was written to next_out, then done. */ |
1723 | if (last) |
1724 | return finish_done; |
1725 | |
1726 | /* If flushing and all input has been consumed, then done. */ |
1727 | if (flush != Z_NO_FLUSH && flush != Z_FINISH && |
1728 | s->strm->avail_in == 0 && (long)s->strstart == s->block_start) |
1729 | return block_done; |
1730 | |
1731 | /* Fill the window with any remaining input. */ |
1732 | have = s->window_size - s->strstart; |
1733 | if (s->strm->avail_in > have && s->block_start >= (long)s->w_size) { |
1734 | /* Slide the window down. */ |
1735 | s->block_start -= s->w_size; |
1736 | s->strstart -= s->w_size; |
1737 | zmemcpy(s->window, s->window + s->w_size, s->strstart); |
1738 | if (s->matches < 2) |
1739 | s->matches++; /* add a pending slide_hash() */ |
1740 | have += s->w_size; /* more space now */ |
1741 | if (s->insert > s->strstart) |
1742 | s->insert = s->strstart; |
1743 | } |
1744 | if (have > s->strm->avail_in) |
1745 | have = s->strm->avail_in; |
1746 | if (have) { |
1747 | read_buf(s->strm, s->window + s->strstart, have); |
1748 | s->strstart += have; |
1749 | s->insert += MIN(have, s->w_size - s->insert); |
1750 | } |
1751 | if (s->high_water < s->strstart) |
1752 | s->high_water = s->strstart; |
1753 | |
1754 | /* There was not enough avail_out to write a complete worthy or flushed |
1755 | * stored block to next_out. Write a stored block to pending instead, if we |
1756 | * have enough input for a worthy block, or if flushing and there is enough |
1757 | * room for the remaining input as a stored block in the pending buffer. |
1758 | */ |
1759 | have = (s->bi_valid + 42) >> 3; /* number of header bytes */ |
1760 | /* maximum stored block length that will fit in pending: */ |
1761 | have = MIN(s->pending_buf_size - have, MAX_STORED); |
1762 | min_block = MIN(have, s->w_size); |
1763 | left = s->strstart - s->block_start; |
1764 | if (left >= min_block || |
1765 | ((left || flush == Z_FINISH) && flush != Z_NO_FLUSH && |
1766 | s->strm->avail_in == 0 && left <= have)) { |
1767 | len = MIN(left, have); |
1768 | last = flush == Z_FINISH && s->strm->avail_in == 0 && |
1769 | len == left ? 1 : 0; |
1770 | _tr_stored_block(s, (charf *)s->window + s->block_start, len, last); |
1771 | s->block_start += len; |
1772 | flush_pending(s->strm); |
1773 | } |
1774 | |
1775 | /* We've done all we can with the available input and output. */ |
1776 | return last ? finish_started : need_more; |
1777 | } |
1778 | |
1779 | /* =========================================================================== |
1780 | * Compress as much as possible from the input stream, return the current |
1781 | * block state. |
1782 | * This function does not perform lazy evaluation of matches and inserts |
1783 | * new strings in the dictionary only for unmatched strings or for short |
1784 | * matches. It is used only for the fast compression options. |
1785 | */ |
1786 | local block_state deflate_fast(deflate_state *s, int flush) { |
1787 | IPos hash_head; /* head of the hash chain */ |
1788 | int bflush; /* set if current block must be flushed */ |
1789 | |
1790 | for (;;) { |
1791 | /* Make sure that we always have enough lookahead, except |
1792 | * at the end of the input file. We need MAX_MATCH bytes |
1793 | * for the next match, plus MIN_MATCH bytes to insert the |
1794 | * string following the next match. |
1795 | */ |
1796 | if (s->lookahead < MIN_LOOKAHEAD) { |
1797 | fill_window(s); |
1798 | if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) { |
1799 | return need_more; |
1800 | } |
1801 | if (s->lookahead == 0) break; /* flush the current block */ |
1802 | } |
1803 | |
1804 | /* Insert the string window[strstart .. strstart + 2] in the |
1805 | * dictionary, and set hash_head to the head of the hash chain: |
1806 | */ |
1807 | hash_head = NIL; |
1808 | if (s->lookahead >= MIN_MATCH) { |
1809 | INSERT_STRING(s, s->strstart, hash_head); |
1810 | } |
1811 | |
1812 | /* Find the longest match, discarding those <= prev_length. |
1813 | * At this point we have always match_length < MIN_MATCH |
1814 | */ |
1815 | if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) { |
1816 | /* To simplify the code, we prevent matches with the string |
1817 | * of window index 0 (in particular we have to avoid a match |
1818 | * of the string with itself at the start of the input file). |
1819 | */ |
1820 | s->match_length = longest_match (s, hash_head); |
1821 | /* longest_match() sets match_start */ |
1822 | } |
1823 | if (s->match_length >= MIN_MATCH) { |
1824 | check_match(s, s->strstart, s->match_start, s->match_length); |
1825 | |
1826 | _tr_tally_dist(s, s->strstart - s->match_start, |
1827 | s->match_length - MIN_MATCH, bflush); |
1828 | |
1829 | s->lookahead -= s->match_length; |
1830 | |
1831 | /* Insert new strings in the hash table only if the match length |
1832 | * is not too large. This saves time but degrades compression. |
1833 | */ |
1834 | #ifndef FASTEST |
1835 | if (s->match_length <= s->max_insert_length && |
1836 | s->lookahead >= MIN_MATCH) { |
1837 | s->match_length--; /* string at strstart already in table */ |
1838 | do { |
1839 | s->strstart++; |
1840 | INSERT_STRING(s, s->strstart, hash_head); |
1841 | /* strstart never exceeds WSIZE-MAX_MATCH, so there are |
1842 | * always MIN_MATCH bytes ahead. |
1843 | */ |
1844 | } while (--s->match_length != 0); |
1845 | s->strstart++; |
1846 | } else |
1847 | #endif |
1848 | { |
1849 | s->strstart += s->match_length; |
1850 | s->match_length = 0; |
1851 | s->ins_h = s->window[s->strstart]; |
1852 | UPDATE_HASH(s, s->ins_h, s->window[s->strstart + 1]); |
1853 | #if MIN_MATCH != 3 |
1854 | Call UPDATE_HASH() MIN_MATCH-3 more times |
1855 | #endif |
1856 | /* If lookahead < MIN_MATCH, ins_h is garbage, but it does not |
1857 | * matter since it will be recomputed at next deflate call. |
1858 | */ |
1859 | } |
1860 | } else { |
1861 | /* No match, output a literal byte */ |
1862 | Tracevv((stderr,"%c" , s->window[s->strstart])); |
1863 | _tr_tally_lit(s, s->window[s->strstart], bflush); |
1864 | s->lookahead--; |
1865 | s->strstart++; |
1866 | } |
1867 | if (bflush) FLUSH_BLOCK(s, 0); |
1868 | } |
1869 | s->insert = s->strstart < MIN_MATCH-1 ? s->strstart : MIN_MATCH-1; |
1870 | if (flush == Z_FINISH) { |
1871 | FLUSH_BLOCK(s, 1); |
1872 | return finish_done; |
1873 | } |
1874 | if (s->sym_next) |
1875 | FLUSH_BLOCK(s, 0); |
1876 | return block_done; |
1877 | } |
1878 | |
1879 | #ifndef FASTEST |
1880 | /* =========================================================================== |
1881 | * Same as above, but achieves better compression. We use a lazy |
1882 | * evaluation for matches: a match is finally adopted only if there is |
1883 | * no better match at the next window position. |
1884 | */ |
1885 | local block_state deflate_slow(deflate_state *s, int flush) { |
1886 | IPos hash_head; /* head of hash chain */ |
1887 | int bflush; /* set if current block must be flushed */ |
1888 | |
1889 | /* Process the input block. */ |
1890 | for (;;) { |
1891 | /* Make sure that we always have enough lookahead, except |
1892 | * at the end of the input file. We need MAX_MATCH bytes |
1893 | * for the next match, plus MIN_MATCH bytes to insert the |
1894 | * string following the next match. |
1895 | */ |
1896 | if (s->lookahead < MIN_LOOKAHEAD) { |
1897 | fill_window(s); |
1898 | if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) { |
1899 | return need_more; |
1900 | } |
1901 | if (s->lookahead == 0) break; /* flush the current block */ |
1902 | } |
1903 | |
1904 | /* Insert the string window[strstart .. strstart + 2] in the |
1905 | * dictionary, and set hash_head to the head of the hash chain: |
1906 | */ |
1907 | hash_head = NIL; |
1908 | if (s->lookahead >= MIN_MATCH) { |
1909 | INSERT_STRING(s, s->strstart, hash_head); |
1910 | } |
1911 | |
1912 | /* Find the longest match, discarding those <= prev_length. |
1913 | */ |
1914 | s->prev_length = s->match_length, s->prev_match = s->match_start; |
1915 | s->match_length = MIN_MATCH-1; |
1916 | |
1917 | if (hash_head != NIL && s->prev_length < s->max_lazy_match && |
1918 | s->strstart - hash_head <= MAX_DIST(s)) { |
1919 | /* To simplify the code, we prevent matches with the string |
1920 | * of window index 0 (in particular we have to avoid a match |
1921 | * of the string with itself at the start of the input file). |
1922 | */ |
1923 | s->match_length = longest_match (s, hash_head); |
1924 | /* longest_match() sets match_start */ |
1925 | |
1926 | if (s->match_length <= 5 && (s->strategy == Z_FILTERED |
1927 | #if TOO_FAR <= 32767 |
1928 | || (s->match_length == MIN_MATCH && |
1929 | s->strstart - s->match_start > TOO_FAR) |
1930 | #endif |
1931 | )) { |
1932 | |
1933 | /* If prev_match is also MIN_MATCH, match_start is garbage |
1934 | * but we will ignore the current match anyway. |
1935 | */ |
1936 | s->match_length = MIN_MATCH-1; |
1937 | } |
1938 | } |
1939 | /* If there was a match at the previous step and the current |
1940 | * match is not better, output the previous match: |
1941 | */ |
1942 | if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) { |
1943 | uInt max_insert = s->strstart + s->lookahead - MIN_MATCH; |
1944 | /* Do not insert strings in hash table beyond this. */ |
1945 | |
1946 | check_match(s, s->strstart - 1, s->prev_match, s->prev_length); |
1947 | |
1948 | _tr_tally_dist(s, s->strstart - 1 - s->prev_match, |
1949 | s->prev_length - MIN_MATCH, bflush); |
1950 | |
1951 | /* Insert in hash table all strings up to the end of the match. |
1952 | * strstart - 1 and strstart are already inserted. If there is not |
1953 | * enough lookahead, the last two strings are not inserted in |
1954 | * the hash table. |
1955 | */ |
1956 | s->lookahead -= s->prev_length - 1; |
1957 | s->prev_length -= 2; |
1958 | do { |
1959 | if (++s->strstart <= max_insert) { |
1960 | INSERT_STRING(s, s->strstart, hash_head); |
1961 | } |
1962 | } while (--s->prev_length != 0); |
1963 | s->match_available = 0; |
1964 | s->match_length = MIN_MATCH-1; |
1965 | s->strstart++; |
1966 | |
1967 | if (bflush) FLUSH_BLOCK(s, 0); |
1968 | |
1969 | } else if (s->match_available) { |
1970 | /* If there was no match at the previous position, output a |
1971 | * single literal. If there was a match but the current match |
1972 | * is longer, truncate the previous match to a single literal. |
1973 | */ |
1974 | Tracevv((stderr,"%c" , s->window[s->strstart - 1])); |
1975 | _tr_tally_lit(s, s->window[s->strstart - 1], bflush); |
1976 | if (bflush) { |
1977 | FLUSH_BLOCK_ONLY(s, 0); |
1978 | } |
1979 | s->strstart++; |
1980 | s->lookahead--; |
1981 | if (s->strm->avail_out == 0) return need_more; |
1982 | } else { |
1983 | /* There is no previous match to compare with, wait for |
1984 | * the next step to decide. |
1985 | */ |
1986 | s->match_available = 1; |
1987 | s->strstart++; |
1988 | s->lookahead--; |
1989 | } |
1990 | } |
1991 | Assert (flush != Z_NO_FLUSH, "no flush?" ); |
1992 | if (s->match_available) { |
1993 | Tracevv((stderr,"%c" , s->window[s->strstart - 1])); |
1994 | _tr_tally_lit(s, s->window[s->strstart - 1], bflush); |
1995 | s->match_available = 0; |
1996 | } |
1997 | s->insert = s->strstart < MIN_MATCH-1 ? s->strstart : MIN_MATCH-1; |
1998 | if (flush == Z_FINISH) { |
1999 | FLUSH_BLOCK(s, 1); |
2000 | return finish_done; |
2001 | } |
2002 | if (s->sym_next) |
2003 | FLUSH_BLOCK(s, 0); |
2004 | return block_done; |
2005 | } |
2006 | #endif /* FASTEST */ |
2007 | |
2008 | /* =========================================================================== |
2009 | * For Z_RLE, simply look for runs of bytes, generate matches only of distance |
2010 | * one. Do not maintain a hash table. (It will be regenerated if this run of |
2011 | * deflate switches away from Z_RLE.) |
2012 | */ |
2013 | local block_state deflate_rle(deflate_state *s, int flush) { |
2014 | int bflush; /* set if current block must be flushed */ |
2015 | uInt prev; /* byte at distance one to match */ |
2016 | Bytef *scan, *strend; /* scan goes up to strend for length of run */ |
2017 | |
2018 | for (;;) { |
2019 | /* Make sure that we always have enough lookahead, except |
2020 | * at the end of the input file. We need MAX_MATCH bytes |
2021 | * for the longest run, plus one for the unrolled loop. |
2022 | */ |
2023 | if (s->lookahead <= MAX_MATCH) { |
2024 | fill_window(s); |
2025 | if (s->lookahead <= MAX_MATCH && flush == Z_NO_FLUSH) { |
2026 | return need_more; |
2027 | } |
2028 | if (s->lookahead == 0) break; /* flush the current block */ |
2029 | } |
2030 | |
2031 | /* See how many times the previous byte repeats */ |
2032 | s->match_length = 0; |
2033 | if (s->lookahead >= MIN_MATCH && s->strstart > 0) { |
2034 | scan = s->window + s->strstart - 1; |
2035 | prev = *scan; |
2036 | if (prev == *++scan && prev == *++scan && prev == *++scan) { |
2037 | strend = s->window + s->strstart + MAX_MATCH; |
2038 | do { |
2039 | } while (prev == *++scan && prev == *++scan && |
2040 | prev == *++scan && prev == *++scan && |
2041 | prev == *++scan && prev == *++scan && |
2042 | prev == *++scan && prev == *++scan && |
2043 | scan < strend); |
2044 | s->match_length = MAX_MATCH - (uInt)(strend - scan); |
2045 | if (s->match_length > s->lookahead) |
2046 | s->match_length = s->lookahead; |
2047 | } |
2048 | Assert(scan <= s->window + (uInt)(s->window_size - 1), |
2049 | "wild scan" ); |
2050 | } |
2051 | |
2052 | /* Emit match if have run of MIN_MATCH or longer, else emit literal */ |
2053 | if (s->match_length >= MIN_MATCH) { |
2054 | check_match(s, s->strstart, s->strstart - 1, s->match_length); |
2055 | |
2056 | _tr_tally_dist(s, 1, s->match_length - MIN_MATCH, bflush); |
2057 | |
2058 | s->lookahead -= s->match_length; |
2059 | s->strstart += s->match_length; |
2060 | s->match_length = 0; |
2061 | } else { |
2062 | /* No match, output a literal byte */ |
2063 | Tracevv((stderr,"%c" , s->window[s->strstart])); |
2064 | _tr_tally_lit(s, s->window[s->strstart], bflush); |
2065 | s->lookahead--; |
2066 | s->strstart++; |
2067 | } |
2068 | if (bflush) FLUSH_BLOCK(s, 0); |
2069 | } |
2070 | s->insert = 0; |
2071 | if (flush == Z_FINISH) { |
2072 | FLUSH_BLOCK(s, 1); |
2073 | return finish_done; |
2074 | } |
2075 | if (s->sym_next) |
2076 | FLUSH_BLOCK(s, 0); |
2077 | return block_done; |
2078 | } |
2079 | |
2080 | /* =========================================================================== |
2081 | * For Z_HUFFMAN_ONLY, do not look for matches. Do not maintain a hash table. |
2082 | * (It will be regenerated if this run of deflate switches away from Huffman.) |
2083 | */ |
2084 | local block_state deflate_huff(deflate_state *s, int flush) { |
2085 | int bflush; /* set if current block must be flushed */ |
2086 | |
2087 | for (;;) { |
2088 | /* Make sure that we have a literal to write. */ |
2089 | if (s->lookahead == 0) { |
2090 | fill_window(s); |
2091 | if (s->lookahead == 0) { |
2092 | if (flush == Z_NO_FLUSH) |
2093 | return need_more; |
2094 | break; /* flush the current block */ |
2095 | } |
2096 | } |
2097 | |
2098 | /* Output a literal byte */ |
2099 | s->match_length = 0; |
2100 | Tracevv((stderr,"%c" , s->window[s->strstart])); |
2101 | _tr_tally_lit(s, s->window[s->strstart], bflush); |
2102 | s->lookahead--; |
2103 | s->strstart++; |
2104 | if (bflush) FLUSH_BLOCK(s, 0); |
2105 | } |
2106 | s->insert = 0; |
2107 | if (flush == Z_FINISH) { |
2108 | FLUSH_BLOCK(s, 1); |
2109 | return finish_done; |
2110 | } |
2111 | if (s->sym_next) |
2112 | FLUSH_BLOCK(s, 0); |
2113 | return block_done; |
2114 | } |
2115 | |