xref: /freebsd/sys/contrib/zlib/deflate.c (revision 1719886f)
1 /* deflate.c -- compress data using the deflation algorithm
2  * Copyright (C) 1995-2024 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 1.3.1 Copyright 1995-2024 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 #if defined(Z_SOLO) && !defined(_KERNEL)
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 #if defined(Z_SOLO) && !defined(_KERNEL)
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, LIT_BUFS);
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 #ifdef LIT_MEM
507     s->d_buf = (ushf *)(s->pending_buf + (s->lit_bufsize << 1));
508     s->l_buf = s->pending_buf + (s->lit_bufsize << 2);
509     s->sym_end = s->lit_bufsize - 1;
510 #else
511     s->sym_buf = s->pending_buf + s->lit_bufsize;
512     s->sym_end = (s->lit_bufsize - 1) * 3;
513 #endif
514     /* We avoid equality with lit_bufsize*3 because of wraparound at 64K
515      * on 16 bit machines and because stored blocks are restricted to
516      * 64K-1 bytes.
517      */
518 
519     s->level = level;
520     s->strategy = strategy;
521     s->method = (Byte)method;
522 
523     return deflateReset(strm);
524 }
525 
526 /* =========================================================================
527  * Check for a valid deflate stream state. Return 0 if ok, 1 if not.
528  */
529 local int deflateStateCheck(z_streamp strm) {
530     deflate_state *s;
531     if (strm == Z_NULL ||
532         strm->zalloc == (alloc_func)0 || strm->zfree == (free_func)0)
533         return 1;
534     s = strm->state;
535     if (s == Z_NULL || s->strm != strm || (s->status != INIT_STATE &&
536 #ifdef GZIP
537                                            s->status != GZIP_STATE &&
538 #endif
539                                            s->status != EXTRA_STATE &&
540                                            s->status != NAME_STATE &&
541                                            s->status != COMMENT_STATE &&
542                                            s->status != HCRC_STATE &&
543                                            s->status != BUSY_STATE &&
544                                            s->status != FINISH_STATE))
545         return 1;
546     return 0;
547 }
548 
549 /* ========================================================================= */
550 int ZEXPORT deflateSetDictionary(z_streamp strm, const Bytef *dictionary,
551                                  uInt  dictLength) {
552     deflate_state *s;
553     uInt str, n;
554     int wrap;
555     unsigned avail;
556     z_const unsigned char *next;
557 
558     if (deflateStateCheck(strm) || dictionary == Z_NULL)
559         return Z_STREAM_ERROR;
560     s = strm->state;
561     wrap = s->wrap;
562     if (wrap == 2 || (wrap == 1 && s->status != INIT_STATE) || s->lookahead)
563         return Z_STREAM_ERROR;
564 
565     /* when using zlib wrappers, compute Adler-32 for provided dictionary */
566     if (wrap == 1)
567         strm->adler = adler32(strm->adler, dictionary, dictLength);
568     s->wrap = 0;                    /* avoid computing Adler-32 in read_buf */
569 
570     /* if dictionary would fill window, just replace the history */
571     if (dictLength >= s->w_size) {
572         if (wrap == 0) {            /* already empty otherwise */
573             CLEAR_HASH(s);
574             s->strstart = 0;
575             s->block_start = 0L;
576             s->insert = 0;
577         }
578         dictionary += dictLength - s->w_size;  /* use the tail */
579         dictLength = s->w_size;
580     }
581 
582     /* insert dictionary into window and hash */
583     avail = strm->avail_in;
584     next = strm->next_in;
585     strm->avail_in = dictLength;
586     strm->next_in = (z_const Bytef *)dictionary;
587     fill_window(s);
588     while (s->lookahead >= MIN_MATCH) {
589         str = s->strstart;
590         n = s->lookahead - (MIN_MATCH-1);
591         do {
592             UPDATE_HASH(s, s->ins_h, s->window[str + MIN_MATCH-1]);
593 #ifndef FASTEST
594             s->prev[str & s->w_mask] = s->head[s->ins_h];
595 #endif
596             s->head[s->ins_h] = (Pos)str;
597             str++;
598         } while (--n);
599         s->strstart = str;
600         s->lookahead = MIN_MATCH-1;
601         fill_window(s);
602     }
603     s->strstart += s->lookahead;
604     s->block_start = (long)s->strstart;
605     s->insert = s->lookahead;
606     s->lookahead = 0;
607     s->match_length = s->prev_length = MIN_MATCH-1;
608     s->match_available = 0;
609     strm->next_in = next;
610     strm->avail_in = avail;
611     s->wrap = wrap;
612     return Z_OK;
613 }
614 
615 /* ========================================================================= */
616 int ZEXPORT deflateGetDictionary(z_streamp strm, Bytef *dictionary,
617                                  uInt *dictLength) {
618     deflate_state *s;
619     uInt len;
620 
621     if (deflateStateCheck(strm))
622         return Z_STREAM_ERROR;
623     s = strm->state;
624     len = s->strstart + s->lookahead;
625     if (len > s->w_size)
626         len = s->w_size;
627     if (dictionary != Z_NULL && len)
628         zmemcpy(dictionary, s->window + s->strstart + s->lookahead - len, len);
629     if (dictLength != Z_NULL)
630         *dictLength = len;
631     return Z_OK;
632 }
633 
634 /* ========================================================================= */
635 int ZEXPORT deflateResetKeep(z_streamp strm) {
636     deflate_state *s;
637 
638     if (deflateStateCheck(strm)) {
639         return Z_STREAM_ERROR;
640     }
641 
642     strm->total_in = strm->total_out = 0;
643     strm->msg = Z_NULL; /* use zfree if we ever allocate msg dynamically */
644     strm->data_type = Z_UNKNOWN;
645 
646     s = (deflate_state *)strm->state;
647     s->pending = 0;
648     s->pending_out = s->pending_buf;
649 
650     if (s->wrap < 0) {
651         s->wrap = -s->wrap; /* was made negative by deflate(..., Z_FINISH); */
652     }
653     s->status =
654 #ifdef GZIP
655         s->wrap == 2 ? GZIP_STATE :
656 #endif
657         INIT_STATE;
658     strm->adler =
659 #ifdef GZIP
660         s->wrap == 2 ? crc32(0L, Z_NULL, 0) :
661 #endif
662         adler32(0L, Z_NULL, 0);
663     s->last_flush = -2;
664 
665     _tr_init(s);
666 
667     return Z_OK;
668 }
669 
670 /* ===========================================================================
671  * Initialize the "longest match" routines for a new zlib stream
672  */
673 local void lm_init(deflate_state *s) {
674     s->window_size = (ulg)2L*s->w_size;
675 
676     CLEAR_HASH(s);
677 
678     /* Set the default configuration parameters:
679      */
680     s->max_lazy_match   = configuration_table[s->level].max_lazy;
681     s->good_match       = configuration_table[s->level].good_length;
682     s->nice_match       = configuration_table[s->level].nice_length;
683     s->max_chain_length = configuration_table[s->level].max_chain;
684 
685     s->strstart = 0;
686     s->block_start = 0L;
687     s->lookahead = 0;
688     s->insert = 0;
689     s->match_length = s->prev_length = MIN_MATCH-1;
690     s->match_available = 0;
691     s->ins_h = 0;
692 }
693 
694 /* ========================================================================= */
695 int ZEXPORT deflateReset(z_streamp strm) {
696     int ret;
697 
698     ret = deflateResetKeep(strm);
699     if (ret == Z_OK)
700         lm_init(strm->state);
701     return ret;
702 }
703 
704 /* ========================================================================= */
705 int ZEXPORT deflateSetHeader(z_streamp strm, gz_headerp head) {
706     if (deflateStateCheck(strm) || strm->state->wrap != 2)
707         return Z_STREAM_ERROR;
708     strm->state->gzhead = head;
709     return Z_OK;
710 }
711 
712 /* ========================================================================= */
713 int ZEXPORT deflatePending(z_streamp strm, unsigned *pending, int *bits) {
714     if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
715     if (pending != Z_NULL)
716         *pending = strm->state->pending;
717     if (bits != Z_NULL)
718         *bits = strm->state->bi_valid;
719     return Z_OK;
720 }
721 
722 /* ========================================================================= */
723 int ZEXPORT deflatePrime(z_streamp strm, int bits, int value) {
724     deflate_state *s;
725     int put;
726 
727     if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
728     s = strm->state;
729 #ifdef LIT_MEM
730     if (bits < 0 || bits > 16 ||
731         (uchf *)s->d_buf < s->pending_out + ((Buf_size + 7) >> 3))
732         return Z_BUF_ERROR;
733 #else
734     if (bits < 0 || bits > 16 ||
735         s->sym_buf < s->pending_out + ((Buf_size + 7) >> 3))
736         return Z_BUF_ERROR;
737 #endif
738     do {
739         put = Buf_size - s->bi_valid;
740         if (put > bits)
741             put = bits;
742         s->bi_buf |= (ush)((value & ((1 << put) - 1)) << s->bi_valid);
743         s->bi_valid += put;
744         _tr_flush_bits(s);
745         value >>= put;
746         bits -= put;
747     } while (bits);
748     return Z_OK;
749 }
750 
751 /* ========================================================================= */
752 int ZEXPORT deflateParams(z_streamp strm, int level, int strategy) {
753     deflate_state *s;
754     compress_func func;
755 
756     if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
757     s = strm->state;
758 
759 #ifdef FASTEST
760     if (level != 0) level = 1;
761 #else
762     if (level == Z_DEFAULT_COMPRESSION) level = 6;
763 #endif
764     if (level < 0 || level > 9 || strategy < 0 || strategy > Z_FIXED) {
765         return Z_STREAM_ERROR;
766     }
767     func = configuration_table[s->level].func;
768 
769     if ((strategy != s->strategy || func != configuration_table[level].func) &&
770         s->last_flush != -2) {
771         /* Flush the last buffer: */
772         int err = deflate(strm, Z_BLOCK);
773         if (err == Z_STREAM_ERROR)
774             return err;
775         if (strm->avail_in || (s->strstart - s->block_start) + s->lookahead)
776             return Z_BUF_ERROR;
777     }
778     if (s->level != level) {
779         if (s->level == 0 && s->matches != 0) {
780             if (s->matches == 1)
781                 slide_hash(s);
782             else
783                 CLEAR_HASH(s);
784             s->matches = 0;
785         }
786         s->level = level;
787         s->max_lazy_match   = configuration_table[level].max_lazy;
788         s->good_match       = configuration_table[level].good_length;
789         s->nice_match       = configuration_table[level].nice_length;
790         s->max_chain_length = configuration_table[level].max_chain;
791     }
792     s->strategy = strategy;
793     return Z_OK;
794 }
795 
796 /* ========================================================================= */
797 int ZEXPORT deflateTune(z_streamp strm, int good_length, int max_lazy,
798                         int nice_length, int max_chain) {
799     deflate_state *s;
800 
801     if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
802     s = strm->state;
803     s->good_match = (uInt)good_length;
804     s->max_lazy_match = (uInt)max_lazy;
805     s->nice_match = nice_length;
806     s->max_chain_length = (uInt)max_chain;
807     return Z_OK;
808 }
809 
810 /* =========================================================================
811  * For the default windowBits of 15 and memLevel of 8, this function returns a
812  * close to exact, as well as small, upper bound on the compressed size. This
813  * is an expansion of ~0.03%, plus a small constant.
814  *
815  * For any setting other than those defaults for windowBits and memLevel, one
816  * of two worst case bounds is returned. This is at most an expansion of ~4% or
817  * ~13%, plus a small constant.
818  *
819  * Both the 0.03% and 4% derive from the overhead of stored blocks. The first
820  * one is for stored blocks of 16383 bytes (memLevel == 8), whereas the second
821  * is for stored blocks of 127 bytes (the worst case memLevel == 1). The
822  * expansion results from five bytes of header for each stored block.
823  *
824  * The larger expansion of 13% results from a window size less than or equal to
825  * the symbols buffer size (windowBits <= memLevel + 7). In that case some of
826  * the data being compressed may have slid out of the sliding window, impeding
827  * a stored block from being emitted. Then the only choice is a fixed or
828  * dynamic block, where a fixed block limits the maximum expansion to 9 bits
829  * per 8-bit byte, plus 10 bits for every block. The smallest block size for
830  * which this can occur is 255 (memLevel == 2).
831  *
832  * Shifts are used to approximate divisions, for speed.
833  */
834 uLong ZEXPORT deflateBound(z_streamp strm, uLong sourceLen) {
835     deflate_state *s;
836     uLong fixedlen, storelen, wraplen;
837 
838     /* upper bound for fixed blocks with 9-bit literals and length 255
839        (memLevel == 2, which is the lowest that may not use stored blocks) --
840        ~13% overhead plus a small constant */
841     fixedlen = sourceLen + (sourceLen >> 3) + (sourceLen >> 8) +
842                (sourceLen >> 9) + 4;
843 
844     /* upper bound for stored blocks with length 127 (memLevel == 1) --
845        ~4% overhead plus a small constant */
846     storelen = sourceLen + (sourceLen >> 5) + (sourceLen >> 7) +
847                (sourceLen >> 11) + 7;
848 
849     /* if can't get parameters, return larger bound plus a zlib wrapper */
850     if (deflateStateCheck(strm))
851         return (fixedlen > storelen ? fixedlen : storelen) + 6;
852 
853     /* compute wrapper length */
854     s = strm->state;
855     switch (s->wrap) {
856     case 0:                                 /* raw deflate */
857         wraplen = 0;
858         break;
859     case 1:                                 /* zlib wrapper */
860         wraplen = 6 + (s->strstart ? 4 : 0);
861         break;
862 #ifdef GZIP
863     case 2:                                 /* gzip wrapper */
864         wraplen = 18;
865         if (s->gzhead != Z_NULL) {          /* user-supplied gzip header */
866             Bytef *str;
867             if (s->gzhead->extra != Z_NULL)
868                 wraplen += 2 + s->gzhead->extra_len;
869             str = s->gzhead->name;
870             if (str != Z_NULL)
871                 do {
872                     wraplen++;
873                 } while (*str++);
874             str = s->gzhead->comment;
875             if (str != Z_NULL)
876                 do {
877                     wraplen++;
878                 } while (*str++);
879             if (s->gzhead->hcrc)
880                 wraplen += 2;
881         }
882         break;
883 #endif
884     default:                                /* for compiler happiness */
885         wraplen = 6;
886     }
887 
888     /* if not default parameters, return one of the conservative bounds */
889     if (s->w_bits != 15 || s->hash_bits != 8 + 7)
890         return (s->w_bits <= s->hash_bits && s->level ? fixedlen : storelen) +
891                wraplen;
892 
893     /* default settings: return tight bound for that case -- ~0.03% overhead
894        plus a small constant */
895     return sourceLen + (sourceLen >> 12) + (sourceLen >> 14) +
896            (sourceLen >> 25) + 13 - 6 + wraplen;
897 }
898 
899 /* =========================================================================
900  * Put a short in the pending buffer. The 16-bit value is put in MSB order.
901  * IN assertion: the stream state is correct and there is enough room in
902  * pending_buf.
903  */
904 local void putShortMSB(deflate_state *s, uInt b) {
905     put_byte(s, (Byte)(b >> 8));
906     put_byte(s, (Byte)(b & 0xff));
907 }
908 
909 /* =========================================================================
910  * Flush as much pending output as possible. All deflate() output, except for
911  * some deflate_stored() output, goes through this function so some
912  * applications may wish to modify it to avoid allocating a large
913  * strm->next_out buffer and copying into it. (See also read_buf()).
914  */
915 local void flush_pending(z_streamp strm) {
916     unsigned len;
917     deflate_state *s = strm->state;
918 
919     _tr_flush_bits(s);
920     len = s->pending;
921     if (len > strm->avail_out) len = strm->avail_out;
922     if (len == 0) return;
923 
924     zmemcpy(strm->next_out, s->pending_out, len);
925     strm->next_out  += len;
926     s->pending_out  += len;
927     strm->total_out += len;
928     strm->avail_out -= len;
929     s->pending      -= len;
930     if (s->pending == 0) {
931         s->pending_out = s->pending_buf;
932     }
933 }
934 
935 /* ===========================================================================
936  * Update the header CRC with the bytes s->pending_buf[beg..s->pending - 1].
937  */
938 #define HCRC_UPDATE(beg) \
939     do { \
940         if (s->gzhead->hcrc && s->pending > (beg)) \
941             strm->adler = crc32(strm->adler, s->pending_buf + (beg), \
942                                 s->pending - (beg)); \
943     } while (0)
944 
945 /* ========================================================================= */
946 int ZEXPORT deflate(z_streamp strm, int flush) {
947     int old_flush; /* value of flush param for previous deflate call */
948     deflate_state *s;
949 
950     if (deflateStateCheck(strm) || flush > Z_BLOCK || flush < 0) {
951         return Z_STREAM_ERROR;
952     }
953     s = strm->state;
954 
955     if (strm->next_out == Z_NULL ||
956         (strm->avail_in != 0 && strm->next_in == Z_NULL) ||
957         (s->status == FINISH_STATE && flush != Z_FINISH)) {
958         ERR_RETURN(strm, Z_STREAM_ERROR);
959     }
960     if (strm->avail_out == 0) ERR_RETURN(strm, Z_BUF_ERROR);
961 
962     old_flush = s->last_flush;
963     s->last_flush = flush;
964 
965     /* Flush as much pending output as possible */
966     if (s->pending != 0) {
967         flush_pending(strm);
968         if (strm->avail_out == 0) {
969             /* Since avail_out is 0, deflate will be called again with
970              * more output space, but possibly with both pending and
971              * avail_in equal to zero. There won't be anything to do,
972              * but this is not an error situation so make sure we
973              * return OK instead of BUF_ERROR at next call of deflate:
974              */
975             s->last_flush = -1;
976             return Z_OK;
977         }
978 
979     /* Make sure there is something to do and avoid duplicate consecutive
980      * flushes. For repeated and useless calls with Z_FINISH, we keep
981      * returning Z_STREAM_END instead of Z_BUF_ERROR.
982      */
983     } else if (strm->avail_in == 0 && RANK(flush) <= RANK(old_flush) &&
984                flush != Z_FINISH) {
985         ERR_RETURN(strm, Z_BUF_ERROR);
986     }
987 
988     /* User must not provide more input after the first FINISH: */
989     if (s->status == FINISH_STATE && strm->avail_in != 0) {
990         ERR_RETURN(strm, Z_BUF_ERROR);
991     }
992 
993     /* Write the header */
994     if (s->status == INIT_STATE && s->wrap == 0)
995         s->status = BUSY_STATE;
996     if (s->status == INIT_STATE) {
997         /* zlib header */
998         uInt header = (Z_DEFLATED + ((s->w_bits - 8) << 4)) << 8;
999         uInt level_flags;
1000 
1001         if (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2)
1002             level_flags = 0;
1003         else if (s->level < 6)
1004             level_flags = 1;
1005         else if (s->level == 6)
1006             level_flags = 2;
1007         else
1008             level_flags = 3;
1009         header |= (level_flags << 6);
1010         if (s->strstart != 0) header |= PRESET_DICT;
1011         header += 31 - (header % 31);
1012 
1013         putShortMSB(s, header);
1014 
1015         /* Save the adler32 of the preset dictionary: */
1016         if (s->strstart != 0) {
1017             putShortMSB(s, (uInt)(strm->adler >> 16));
1018             putShortMSB(s, (uInt)(strm->adler & 0xffff));
1019         }
1020         strm->adler = adler32(0L, Z_NULL, 0);
1021         s->status = BUSY_STATE;
1022 
1023         /* Compression must start with an empty pending buffer */
1024         flush_pending(strm);
1025         if (s->pending != 0) {
1026             s->last_flush = -1;
1027             return Z_OK;
1028         }
1029     }
1030 #ifdef GZIP
1031     if (s->status == GZIP_STATE) {
1032         /* gzip header */
1033         strm->adler = crc32(0L, Z_NULL, 0);
1034         put_byte(s, 31);
1035         put_byte(s, 139);
1036         put_byte(s, 8);
1037         if (s->gzhead == Z_NULL) {
1038             put_byte(s, 0);
1039             put_byte(s, 0);
1040             put_byte(s, 0);
1041             put_byte(s, 0);
1042             put_byte(s, 0);
1043             put_byte(s, s->level == 9 ? 2 :
1044                      (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
1045                       4 : 0));
1046             put_byte(s, OS_CODE);
1047             s->status = BUSY_STATE;
1048 
1049             /* Compression must start with an empty pending buffer */
1050             flush_pending(strm);
1051             if (s->pending != 0) {
1052                 s->last_flush = -1;
1053                 return Z_OK;
1054             }
1055         }
1056         else {
1057             put_byte(s, (s->gzhead->text ? 1 : 0) +
1058                      (s->gzhead->hcrc ? 2 : 0) +
1059                      (s->gzhead->extra == Z_NULL ? 0 : 4) +
1060                      (s->gzhead->name == Z_NULL ? 0 : 8) +
1061                      (s->gzhead->comment == Z_NULL ? 0 : 16)
1062                      );
1063             put_byte(s, (Byte)(s->gzhead->time & 0xff));
1064             put_byte(s, (Byte)((s->gzhead->time >> 8) & 0xff));
1065             put_byte(s, (Byte)((s->gzhead->time >> 16) & 0xff));
1066             put_byte(s, (Byte)((s->gzhead->time >> 24) & 0xff));
1067             put_byte(s, s->level == 9 ? 2 :
1068                      (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
1069                       4 : 0));
1070             put_byte(s, s->gzhead->os & 0xff);
1071             if (s->gzhead->extra != Z_NULL) {
1072                 put_byte(s, s->gzhead->extra_len & 0xff);
1073                 put_byte(s, (s->gzhead->extra_len >> 8) & 0xff);
1074             }
1075             if (s->gzhead->hcrc)
1076                 strm->adler = crc32(strm->adler, s->pending_buf,
1077                                     s->pending);
1078             s->gzindex = 0;
1079             s->status = EXTRA_STATE;
1080         }
1081     }
1082     if (s->status == EXTRA_STATE) {
1083         if (s->gzhead->extra != Z_NULL) {
1084             ulg beg = s->pending;   /* start of bytes to update crc */
1085             uInt left = (s->gzhead->extra_len & 0xffff) - s->gzindex;
1086             while (s->pending + left > s->pending_buf_size) {
1087                 uInt copy = s->pending_buf_size - s->pending;
1088                 zmemcpy(s->pending_buf + s->pending,
1089                         s->gzhead->extra + s->gzindex, copy);
1090                 s->pending = s->pending_buf_size;
1091                 HCRC_UPDATE(beg);
1092                 s->gzindex += copy;
1093                 flush_pending(strm);
1094                 if (s->pending != 0) {
1095                     s->last_flush = -1;
1096                     return Z_OK;
1097                 }
1098                 beg = 0;
1099                 left -= copy;
1100             }
1101             zmemcpy(s->pending_buf + s->pending,
1102                     s->gzhead->extra + s->gzindex, left);
1103             s->pending += left;
1104             HCRC_UPDATE(beg);
1105             s->gzindex = 0;
1106         }
1107         s->status = NAME_STATE;
1108     }
1109     if (s->status == NAME_STATE) {
1110         if (s->gzhead->name != Z_NULL) {
1111             ulg beg = s->pending;   /* start of bytes to update crc */
1112             int val;
1113             do {
1114                 if (s->pending == s->pending_buf_size) {
1115                     HCRC_UPDATE(beg);
1116                     flush_pending(strm);
1117                     if (s->pending != 0) {
1118                         s->last_flush = -1;
1119                         return Z_OK;
1120                     }
1121                     beg = 0;
1122                 }
1123                 val = s->gzhead->name[s->gzindex++];
1124                 put_byte(s, val);
1125             } while (val != 0);
1126             HCRC_UPDATE(beg);
1127             s->gzindex = 0;
1128         }
1129         s->status = COMMENT_STATE;
1130     }
1131     if (s->status == COMMENT_STATE) {
1132         if (s->gzhead->comment != Z_NULL) {
1133             ulg beg = s->pending;   /* start of bytes to update crc */
1134             int val;
1135             do {
1136                 if (s->pending == s->pending_buf_size) {
1137                     HCRC_UPDATE(beg);
1138                     flush_pending(strm);
1139                     if (s->pending != 0) {
1140                         s->last_flush = -1;
1141                         return Z_OK;
1142                     }
1143                     beg = 0;
1144                 }
1145                 val = s->gzhead->comment[s->gzindex++];
1146                 put_byte(s, val);
1147             } while (val != 0);
1148             HCRC_UPDATE(beg);
1149         }
1150         s->status = HCRC_STATE;
1151     }
1152     if (s->status == HCRC_STATE) {
1153         if (s->gzhead->hcrc) {
1154             if (s->pending + 2 > s->pending_buf_size) {
1155                 flush_pending(strm);
1156                 if (s->pending != 0) {
1157                     s->last_flush = -1;
1158                     return Z_OK;
1159                 }
1160             }
1161             put_byte(s, (Byte)(strm->adler & 0xff));
1162             put_byte(s, (Byte)((strm->adler >> 8) & 0xff));
1163             strm->adler = crc32(0L, Z_NULL, 0);
1164         }
1165         s->status = BUSY_STATE;
1166 
1167         /* Compression must start with an empty pending buffer */
1168         flush_pending(strm);
1169         if (s->pending != 0) {
1170             s->last_flush = -1;
1171             return Z_OK;
1172         }
1173     }
1174 #endif
1175 
1176     /* Start a new block or continue the current one.
1177      */
1178     if (strm->avail_in != 0 || s->lookahead != 0 ||
1179         (flush != Z_NO_FLUSH && s->status != FINISH_STATE)) {
1180         block_state bstate;
1181 
1182         bstate = s->level == 0 ? deflate_stored(s, flush) :
1183                  s->strategy == Z_HUFFMAN_ONLY ? deflate_huff(s, flush) :
1184                  s->strategy == Z_RLE ? deflate_rle(s, flush) :
1185                  (*(configuration_table[s->level].func))(s, flush);
1186 
1187         if (bstate == finish_started || bstate == finish_done) {
1188             s->status = FINISH_STATE;
1189         }
1190         if (bstate == need_more || bstate == finish_started) {
1191             if (strm->avail_out == 0) {
1192                 s->last_flush = -1; /* avoid BUF_ERROR next call, see above */
1193             }
1194             return Z_OK;
1195             /* If flush != Z_NO_FLUSH && avail_out == 0, the next call
1196              * of deflate should use the same flush parameter to make sure
1197              * that the flush is complete. So we don't have to output an
1198              * empty block here, this will be done at next call. This also
1199              * ensures that for a very small output buffer, we emit at most
1200              * one empty block.
1201              */
1202         }
1203         if (bstate == block_done) {
1204             if (flush == Z_PARTIAL_FLUSH) {
1205                 _tr_align(s);
1206             } else if (flush != Z_BLOCK) { /* FULL_FLUSH or SYNC_FLUSH */
1207                 _tr_stored_block(s, (char*)0, 0L, 0);
1208                 /* For a full flush, this empty block will be recognized
1209                  * as a special marker by inflate_sync().
1210                  */
1211                 if (flush == Z_FULL_FLUSH) {
1212                     CLEAR_HASH(s);             /* forget history */
1213                     if (s->lookahead == 0) {
1214                         s->strstart = 0;
1215                         s->block_start = 0L;
1216                         s->insert = 0;
1217                     }
1218                 }
1219             }
1220             flush_pending(strm);
1221             if (strm->avail_out == 0) {
1222               s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */
1223               return Z_OK;
1224             }
1225         }
1226     }
1227 
1228     if (flush != Z_FINISH) return Z_OK;
1229     if (s->wrap <= 0) return Z_STREAM_END;
1230 
1231     /* Write the trailer */
1232 #ifdef GZIP
1233     if (s->wrap == 2) {
1234         put_byte(s, (Byte)(strm->adler & 0xff));
1235         put_byte(s, (Byte)((strm->adler >> 8) & 0xff));
1236         put_byte(s, (Byte)((strm->adler >> 16) & 0xff));
1237         put_byte(s, (Byte)((strm->adler >> 24) & 0xff));
1238         put_byte(s, (Byte)(strm->total_in & 0xff));
1239         put_byte(s, (Byte)((strm->total_in >> 8) & 0xff));
1240         put_byte(s, (Byte)((strm->total_in >> 16) & 0xff));
1241         put_byte(s, (Byte)((strm->total_in >> 24) & 0xff));
1242     }
1243     else
1244 #endif
1245     {
1246         putShortMSB(s, (uInt)(strm->adler >> 16));
1247         putShortMSB(s, (uInt)(strm->adler & 0xffff));
1248     }
1249     flush_pending(strm);
1250     /* If avail_out is zero, the application will call deflate again
1251      * to flush the rest.
1252      */
1253     if (s->wrap > 0) s->wrap = -s->wrap; /* write the trailer only once! */
1254     return s->pending != 0 ? Z_OK : Z_STREAM_END;
1255 }
1256 
1257 /* ========================================================================= */
1258 int ZEXPORT deflateEnd(z_streamp strm) {
1259     int status;
1260 
1261     if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
1262 
1263     status = strm->state->status;
1264 
1265     /* Deallocate in reverse order of allocations: */
1266     TRY_FREE(strm, strm->state->pending_buf);
1267     TRY_FREE(strm, strm->state->head);
1268     TRY_FREE(strm, strm->state->prev);
1269     TRY_FREE(strm, strm->state->window);
1270 
1271     ZFREE(strm, strm->state);
1272     strm->state = Z_NULL;
1273 
1274     return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK;
1275 }
1276 
1277 /* =========================================================================
1278  * Copy the source state to the destination state.
1279  * To simplify the source, this is not supported for 16-bit MSDOS (which
1280  * doesn't have enough memory anyway to duplicate compression states).
1281  */
1282 int ZEXPORT deflateCopy(z_streamp dest, z_streamp source) {
1283 #ifdef MAXSEG_64K
1284     (void)dest;
1285     (void)source;
1286     return Z_STREAM_ERROR;
1287 #else
1288     deflate_state *ds;
1289     deflate_state *ss;
1290 
1291 
1292     if (deflateStateCheck(source) || dest == Z_NULL) {
1293         return Z_STREAM_ERROR;
1294     }
1295 
1296     ss = source->state;
1297 
1298     zmemcpy((voidpf)dest, (voidpf)source, sizeof(z_stream));
1299 
1300     ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state));
1301     if (ds == Z_NULL) return Z_MEM_ERROR;
1302     dest->state = (struct internal_state FAR *) ds;
1303     zmemcpy((voidpf)ds, (voidpf)ss, sizeof(deflate_state));
1304     ds->strm = dest;
1305 
1306     ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte));
1307     ds->prev   = (Posf *)  ZALLOC(dest, ds->w_size, sizeof(Pos));
1308     ds->head   = (Posf *)  ZALLOC(dest, ds->hash_size, sizeof(Pos));
1309     ds->pending_buf = (uchf *) ZALLOC(dest, ds->lit_bufsize, LIT_BUFS);
1310 
1311     if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL ||
1312         ds->pending_buf == Z_NULL) {
1313         deflateEnd (dest);
1314         return Z_MEM_ERROR;
1315     }
1316     /* following zmemcpy do not work for 16-bit MSDOS */
1317     zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte));
1318     zmemcpy((voidpf)ds->prev, (voidpf)ss->prev, ds->w_size * sizeof(Pos));
1319     zmemcpy((voidpf)ds->head, (voidpf)ss->head, ds->hash_size * sizeof(Pos));
1320     zmemcpy(ds->pending_buf, ss->pending_buf, ds->lit_bufsize * LIT_BUFS);
1321 
1322     ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf);
1323 #ifdef LIT_MEM
1324     ds->d_buf = (ushf *)(ds->pending_buf + (ds->lit_bufsize << 1));
1325     ds->l_buf = ds->pending_buf + (ds->lit_bufsize << 2);
1326 #else
1327     ds->sym_buf = ds->pending_buf + ds->lit_bufsize;
1328 #endif
1329 
1330     ds->l_desc.dyn_tree = ds->dyn_ltree;
1331     ds->d_desc.dyn_tree = ds->dyn_dtree;
1332     ds->bl_desc.dyn_tree = ds->bl_tree;
1333 
1334     return Z_OK;
1335 #endif /* MAXSEG_64K */
1336 }
1337 
1338 #ifndef FASTEST
1339 /* ===========================================================================
1340  * Set match_start to the longest match starting at the given string and
1341  * return its length. Matches shorter or equal to prev_length are discarded,
1342  * in which case the result is equal to prev_length and match_start is
1343  * garbage.
1344  * IN assertions: cur_match is the head of the hash chain for the current
1345  *   string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
1346  * OUT assertion: the match length is not greater than s->lookahead.
1347  */
1348 local uInt longest_match(deflate_state *s, IPos cur_match) {
1349     unsigned chain_length = s->max_chain_length;/* max hash chain length */
1350     register Bytef *scan = s->window + s->strstart; /* current string */
1351     register Bytef *match;                      /* matched string */
1352     register int len;                           /* length of current match */
1353     int best_len = (int)s->prev_length;         /* best match length so far */
1354     int nice_match = s->nice_match;             /* stop if match long enough */
1355     IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
1356         s->strstart - (IPos)MAX_DIST(s) : NIL;
1357     /* Stop when cur_match becomes <= limit. To simplify the code,
1358      * we prevent matches with the string of window index 0.
1359      */
1360     Posf *prev = s->prev;
1361     uInt wmask = s->w_mask;
1362 
1363 #ifdef UNALIGNED_OK
1364     /* Compare two bytes at a time. Note: this is not always beneficial.
1365      * Try with and without -DUNALIGNED_OK to check.
1366      */
1367     register Bytef *strend = s->window + s->strstart + MAX_MATCH - 1;
1368     register ush scan_start = *(ushf*)scan;
1369     register ush scan_end   = *(ushf*)(scan + best_len - 1);
1370 #else
1371     register Bytef *strend = s->window + s->strstart + MAX_MATCH;
1372     register Byte scan_end1  = scan[best_len - 1];
1373     register Byte scan_end   = scan[best_len];
1374 #endif
1375 
1376     /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
1377      * It is easy to get rid of this optimization if necessary.
1378      */
1379     Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
1380 
1381     /* Do not waste too much time if we already have a good match: */
1382     if (s->prev_length >= s->good_match) {
1383         chain_length >>= 2;
1384     }
1385     /* Do not look for matches beyond the end of the input. This is necessary
1386      * to make deflate deterministic.
1387      */
1388     if ((uInt)nice_match > s->lookahead) nice_match = (int)s->lookahead;
1389 
1390     Assert((ulg)s->strstart <= s->window_size - MIN_LOOKAHEAD,
1391            "need lookahead");
1392 
1393     do {
1394         Assert(cur_match < s->strstart, "no future");
1395         match = s->window + cur_match;
1396 
1397         /* Skip to next match if the match length cannot increase
1398          * or if the match length is less than 2.  Note that the checks below
1399          * for insufficient lookahead only occur occasionally for performance
1400          * reasons.  Therefore uninitialized memory will be accessed, and
1401          * conditional jumps will be made that depend on those values.
1402          * However the length of the match is limited to the lookahead, so
1403          * the output of deflate is not affected by the uninitialized values.
1404          */
1405 #if (defined(UNALIGNED_OK) && MAX_MATCH == 258)
1406         /* This code assumes sizeof(unsigned short) == 2. Do not use
1407          * UNALIGNED_OK if your compiler uses a different size.
1408          */
1409         if (*(ushf*)(match + best_len - 1) != scan_end ||
1410             *(ushf*)match != scan_start) continue;
1411 
1412         /* It is not necessary to compare scan[2] and match[2] since they are
1413          * always equal when the other bytes match, given that the hash keys
1414          * are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at
1415          * strstart + 3, + 5, up to strstart + 257. We check for insufficient
1416          * lookahead only every 4th comparison; the 128th check will be made
1417          * at strstart + 257. If MAX_MATCH-2 is not a multiple of 8, it is
1418          * necessary to put more guard bytes at the end of the window, or
1419          * to check more often for insufficient lookahead.
1420          */
1421         Assert(scan[2] == match[2], "scan[2]?");
1422         scan++, match++;
1423         do {
1424         } while (*(ushf*)(scan += 2) == *(ushf*)(match += 2) &&
1425                  *(ushf*)(scan += 2) == *(ushf*)(match += 2) &&
1426                  *(ushf*)(scan += 2) == *(ushf*)(match += 2) &&
1427                  *(ushf*)(scan += 2) == *(ushf*)(match += 2) &&
1428                  scan < strend);
1429         /* The funny "do {}" generates better code on most compilers */
1430 
1431         /* Here, scan <= window + strstart + 257 */
1432         Assert(scan <= s->window + (unsigned)(s->window_size - 1),
1433                "wild scan");
1434         if (*scan == *match) scan++;
1435 
1436         len = (MAX_MATCH - 1) - (int)(strend - scan);
1437         scan = strend - (MAX_MATCH-1);
1438 
1439 #else /* UNALIGNED_OK */
1440 
1441         if (match[best_len]     != scan_end  ||
1442             match[best_len - 1] != scan_end1 ||
1443             *match              != *scan     ||
1444             *++match            != scan[1])      continue;
1445 
1446         /* The check at best_len - 1 can be removed because it will be made
1447          * again later. (This heuristic is not always a win.)
1448          * It is not necessary to compare scan[2] and match[2] since they
1449          * are always equal when the other bytes match, given that
1450          * the hash keys are equal and that HASH_BITS >= 8.
1451          */
1452         scan += 2, match++;
1453         Assert(*scan == *match, "match[2]?");
1454 
1455         /* We check for insufficient lookahead only every 8th comparison;
1456          * the 256th check will be made at strstart + 258.
1457          */
1458         do {
1459         } while (*++scan == *++match && *++scan == *++match &&
1460                  *++scan == *++match && *++scan == *++match &&
1461                  *++scan == *++match && *++scan == *++match &&
1462                  *++scan == *++match && *++scan == *++match &&
1463                  scan < strend);
1464 
1465         Assert(scan <= s->window + (unsigned)(s->window_size - 1),
1466                "wild scan");
1467 
1468         len = MAX_MATCH - (int)(strend - scan);
1469         scan = strend - MAX_MATCH;
1470 
1471 #endif /* UNALIGNED_OK */
1472 
1473         if (len > best_len) {
1474             s->match_start = cur_match;
1475             best_len = len;
1476             if (len >= nice_match) break;
1477 #ifdef UNALIGNED_OK
1478             scan_end = *(ushf*)(scan + best_len - 1);
1479 #else
1480             scan_end1  = scan[best_len - 1];
1481             scan_end   = scan[best_len];
1482 #endif
1483         }
1484     } while ((cur_match = prev[cur_match & wmask]) > limit
1485              && --chain_length != 0);
1486 
1487     if ((uInt)best_len <= s->lookahead) return (uInt)best_len;
1488     return s->lookahead;
1489 }
1490 
1491 #else /* FASTEST */
1492 
1493 /* ---------------------------------------------------------------------------
1494  * Optimized version for FASTEST only
1495  */
1496 local uInt longest_match(deflate_state *s, IPos cur_match) {
1497     register Bytef *scan = s->window + s->strstart; /* current string */
1498     register Bytef *match;                       /* matched string */
1499     register int len;                           /* length of current match */
1500     register Bytef *strend = s->window + s->strstart + MAX_MATCH;
1501 
1502     /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
1503      * It is easy to get rid of this optimization if necessary.
1504      */
1505     Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
1506 
1507     Assert((ulg)s->strstart <= s->window_size - MIN_LOOKAHEAD,
1508            "need lookahead");
1509 
1510     Assert(cur_match < s->strstart, "no future");
1511 
1512     match = s->window + cur_match;
1513 
1514     /* Return failure if the match length is less than 2:
1515      */
1516     if (match[0] != scan[0] || match[1] != scan[1]) return MIN_MATCH-1;
1517 
1518     /* The check at best_len - 1 can be removed because it will be made
1519      * again later. (This heuristic is not always a win.)
1520      * It is not necessary to compare scan[2] and match[2] since they
1521      * are always equal when the other bytes match, given that
1522      * the hash keys are equal and that HASH_BITS >= 8.
1523      */
1524     scan += 2, match += 2;
1525     Assert(*scan == *match, "match[2]?");
1526 
1527     /* We check for insufficient lookahead only every 8th comparison;
1528      * the 256th check will be made at strstart + 258.
1529      */
1530     do {
1531     } while (*++scan == *++match && *++scan == *++match &&
1532              *++scan == *++match && *++scan == *++match &&
1533              *++scan == *++match && *++scan == *++match &&
1534              *++scan == *++match && *++scan == *++match &&
1535              scan < strend);
1536 
1537     Assert(scan <= s->window + (unsigned)(s->window_size - 1), "wild scan");
1538 
1539     len = MAX_MATCH - (int)(strend - scan);
1540 
1541     if (len < MIN_MATCH) return MIN_MATCH - 1;
1542 
1543     s->match_start = cur_match;
1544     return (uInt)len <= s->lookahead ? (uInt)len : s->lookahead;
1545 }
1546 
1547 #endif /* FASTEST */
1548 
1549 #ifdef ZLIB_DEBUG
1550 
1551 #define EQUAL 0
1552 /* result of memcmp for equal strings */
1553 
1554 /* ===========================================================================
1555  * Check that the match at match_start is indeed a match.
1556  */
1557 local void check_match(deflate_state *s, IPos start, IPos match, int length) {
1558     /* check that the match is indeed a match */
1559     Bytef *back = s->window + (int)match, *here = s->window + start;
1560     IPos len = length;
1561     if (match == (IPos)-1) {
1562         /* match starts one byte before the current window -- just compare the
1563            subsequent length-1 bytes */
1564         back++;
1565         here++;
1566         len--;
1567     }
1568     if (zmemcmp(back, here, len) != EQUAL) {
1569         fprintf(stderr, " start %u, match %d, length %d\n",
1570                 start, (int)match, length);
1571         do {
1572             fprintf(stderr, "(%02x %02x)", *back++, *here++);
1573         } while (--len != 0);
1574         z_error("invalid match");
1575     }
1576     if (z_verbose > 1) {
1577         fprintf(stderr,"\\[%d,%d]", start - match, length);
1578         do { putc(s->window[start++], stderr); } while (--length != 0);
1579     }
1580 }
1581 #else
1582 #  define check_match(s, start, match, length)
1583 #endif /* ZLIB_DEBUG */
1584 
1585 /* ===========================================================================
1586  * Flush the current block, with given end-of-file flag.
1587  * IN assertion: strstart is set to the end of the current match.
1588  */
1589 #define FLUSH_BLOCK_ONLY(s, last) { \
1590    _tr_flush_block(s, (s->block_start >= 0L ? \
1591                    (charf *)&s->window[(unsigned)s->block_start] : \
1592                    (charf *)Z_NULL), \
1593                 (ulg)((long)s->strstart - s->block_start), \
1594                 (last)); \
1595    s->block_start = s->strstart; \
1596    flush_pending(s->strm); \
1597    Tracev((stderr,"[FLUSH]")); \
1598 }
1599 
1600 /* Same but force premature exit if necessary. */
1601 #define FLUSH_BLOCK(s, last) { \
1602    FLUSH_BLOCK_ONLY(s, last); \
1603    if (s->strm->avail_out == 0) return (last) ? finish_started : need_more; \
1604 }
1605 
1606 /* Maximum stored block length in deflate format (not including header). */
1607 #define MAX_STORED 65535
1608 
1609 #if !defined(MIN)
1610 /* Minimum of a and b. */
1611 #define MIN(a, b) ((a) > (b) ? (b) : (a))
1612 #endif
1613 
1614 /* ===========================================================================
1615  * Copy without compression as much as possible from the input stream, return
1616  * the current block state.
1617  *
1618  * In case deflateParams() is used to later switch to a non-zero compression
1619  * level, s->matches (otherwise unused when storing) keeps track of the number
1620  * of hash table slides to perform. If s->matches is 1, then one hash table
1621  * slide will be done when switching. If s->matches is 2, the maximum value
1622  * allowed here, then the hash table will be cleared, since two or more slides
1623  * is the same as a clear.
1624  *
1625  * deflate_stored() is written to minimize the number of times an input byte is
1626  * copied. It is most efficient with large input and output buffers, which
1627  * maximizes the opportunities to have a single copy from next_in to next_out.
1628  */
1629 local block_state deflate_stored(deflate_state *s, int flush) {
1630     /* Smallest worthy block size when not flushing or finishing. By default
1631      * this is 32K. This can be as small as 507 bytes for memLevel == 1. For
1632      * large input and output buffers, the stored block size will be larger.
1633      */
1634     unsigned min_block = MIN(s->pending_buf_size - 5, s->w_size);
1635 
1636     /* Copy as many min_block or larger stored blocks directly to next_out as
1637      * possible. If flushing, copy the remaining available input to next_out as
1638      * stored blocks, if there is enough space.
1639      */
1640     unsigned len, left, have, last = 0;
1641     unsigned used = s->strm->avail_in;
1642     do {
1643         /* Set len to the maximum size block that we can copy directly with the
1644          * available input data and output space. Set left to how much of that
1645          * would be copied from what's left in the window.
1646          */
1647         len = MAX_STORED;       /* maximum deflate stored block length */
1648         have = (s->bi_valid + 42) >> 3;         /* number of header bytes */
1649         if (s->strm->avail_out < have)          /* need room for header */
1650             break;
1651             /* maximum stored block length that will fit in avail_out: */
1652         have = s->strm->avail_out - have;
1653         left = s->strstart - s->block_start;    /* bytes left in window */
1654         if (len > (ulg)left + s->strm->avail_in)
1655             len = left + s->strm->avail_in;     /* limit len to the input */
1656         if (len > have)
1657             len = have;                         /* limit len to the output */
1658 
1659         /* If the stored block would be less than min_block in length, or if
1660          * unable to copy all of the available input when flushing, then try
1661          * copying to the window and the pending buffer instead. Also don't
1662          * write an empty block when flushing -- deflate() does that.
1663          */
1664         if (len < min_block && ((len == 0 && flush != Z_FINISH) ||
1665                                 flush == Z_NO_FLUSH ||
1666                                 len != left + s->strm->avail_in))
1667             break;
1668 
1669         /* Make a dummy stored block in pending to get the header bytes,
1670          * including any pending bits. This also updates the debugging counts.
1671          */
1672         last = flush == Z_FINISH && len == left + s->strm->avail_in ? 1 : 0;
1673         _tr_stored_block(s, (char *)0, 0L, last);
1674 
1675         /* Replace the lengths in the dummy stored block with len. */
1676         s->pending_buf[s->pending - 4] = len;
1677         s->pending_buf[s->pending - 3] = len >> 8;
1678         s->pending_buf[s->pending - 2] = ~len;
1679         s->pending_buf[s->pending - 1] = ~len >> 8;
1680 
1681         /* Write the stored block header bytes. */
1682         flush_pending(s->strm);
1683 
1684 #ifdef ZLIB_DEBUG
1685         /* Update debugging counts for the data about to be copied. */
1686         s->compressed_len += len << 3;
1687         s->bits_sent += len << 3;
1688 #endif
1689 
1690         /* Copy uncompressed bytes from the window to next_out. */
1691         if (left) {
1692             if (left > len)
1693                 left = len;
1694             zmemcpy(s->strm->next_out, s->window + s->block_start, left);
1695             s->strm->next_out += left;
1696             s->strm->avail_out -= left;
1697             s->strm->total_out += left;
1698             s->block_start += left;
1699             len -= left;
1700         }
1701 
1702         /* Copy uncompressed bytes directly from next_in to next_out, updating
1703          * the check value.
1704          */
1705         if (len) {
1706             read_buf(s->strm, s->strm->next_out, len);
1707             s->strm->next_out += len;
1708             s->strm->avail_out -= len;
1709             s->strm->total_out += len;
1710         }
1711     } while (last == 0);
1712 
1713     /* Update the sliding window with the last s->w_size bytes of the copied
1714      * data, or append all of the copied data to the existing window if less
1715      * than s->w_size bytes were copied. Also update the number of bytes to
1716      * insert in the hash tables, in the event that deflateParams() switches to
1717      * a non-zero compression level.
1718      */
1719     used -= s->strm->avail_in;      /* number of input bytes directly copied */
1720     if (used) {
1721         /* If any input was used, then no unused input remains in the window,
1722          * therefore s->block_start == s->strstart.
1723          */
1724         if (used >= s->w_size) {    /* supplant the previous history */
1725             s->matches = 2;         /* clear hash */
1726             zmemcpy(s->window, s->strm->next_in - s->w_size, s->w_size);
1727             s->strstart = s->w_size;
1728             s->insert = s->strstart;
1729         }
1730         else {
1731             if (s->window_size - s->strstart <= used) {
1732                 /* Slide the window down. */
1733                 s->strstart -= s->w_size;
1734                 zmemcpy(s->window, s->window + s->w_size, s->strstart);
1735                 if (s->matches < 2)
1736                     s->matches++;   /* add a pending slide_hash() */
1737                 if (s->insert > s->strstart)
1738                     s->insert = s->strstart;
1739             }
1740             zmemcpy(s->window + s->strstart, s->strm->next_in - used, used);
1741             s->strstart += used;
1742             s->insert += MIN(used, s->w_size - s->insert);
1743         }
1744         s->block_start = s->strstart;
1745     }
1746     if (s->high_water < s->strstart)
1747         s->high_water = s->strstart;
1748 
1749     /* If the last block was written to next_out, then done. */
1750     if (last)
1751         return finish_done;
1752 
1753     /* If flushing and all input has been consumed, then done. */
1754     if (flush != Z_NO_FLUSH && flush != Z_FINISH &&
1755         s->strm->avail_in == 0 && (long)s->strstart == s->block_start)
1756         return block_done;
1757 
1758     /* Fill the window with any remaining input. */
1759     have = s->window_size - s->strstart;
1760     if (s->strm->avail_in > have && s->block_start >= (long)s->w_size) {
1761         /* Slide the window down. */
1762         s->block_start -= s->w_size;
1763         s->strstart -= s->w_size;
1764         zmemcpy(s->window, s->window + s->w_size, s->strstart);
1765         if (s->matches < 2)
1766             s->matches++;           /* add a pending slide_hash() */
1767         have += s->w_size;          /* more space now */
1768         if (s->insert > s->strstart)
1769             s->insert = s->strstart;
1770     }
1771     if (have > s->strm->avail_in)
1772         have = s->strm->avail_in;
1773     if (have) {
1774         read_buf(s->strm, s->window + s->strstart, have);
1775         s->strstart += have;
1776         s->insert += MIN(have, s->w_size - s->insert);
1777     }
1778     if (s->high_water < s->strstart)
1779         s->high_water = s->strstart;
1780 
1781     /* There was not enough avail_out to write a complete worthy or flushed
1782      * stored block to next_out. Write a stored block to pending instead, if we
1783      * have enough input for a worthy block, or if flushing and there is enough
1784      * room for the remaining input as a stored block in the pending buffer.
1785      */
1786     have = (s->bi_valid + 42) >> 3;         /* number of header bytes */
1787         /* maximum stored block length that will fit in pending: */
1788     have = MIN(s->pending_buf_size - have, MAX_STORED);
1789     min_block = MIN(have, s->w_size);
1790     left = s->strstart - s->block_start;
1791     if (left >= min_block ||
1792         ((left || flush == Z_FINISH) && flush != Z_NO_FLUSH &&
1793          s->strm->avail_in == 0 && left <= have)) {
1794         len = MIN(left, have);
1795         last = flush == Z_FINISH && s->strm->avail_in == 0 &&
1796                len == left ? 1 : 0;
1797         _tr_stored_block(s, (charf *)s->window + s->block_start, len, last);
1798         s->block_start += len;
1799         flush_pending(s->strm);
1800     }
1801 
1802     /* We've done all we can with the available input and output. */
1803     return last ? finish_started : need_more;
1804 }
1805 
1806 /* ===========================================================================
1807  * Compress as much as possible from the input stream, return the current
1808  * block state.
1809  * This function does not perform lazy evaluation of matches and inserts
1810  * new strings in the dictionary only for unmatched strings or for short
1811  * matches. It is used only for the fast compression options.
1812  */
1813 local block_state deflate_fast(deflate_state *s, int flush) {
1814     IPos hash_head;       /* head of the hash chain */
1815     int bflush;           /* set if current block must be flushed */
1816 
1817     for (;;) {
1818         /* Make sure that we always have enough lookahead, except
1819          * at the end of the input file. We need MAX_MATCH bytes
1820          * for the next match, plus MIN_MATCH bytes to insert the
1821          * string following the next match.
1822          */
1823         if (s->lookahead < MIN_LOOKAHEAD) {
1824             fill_window(s);
1825             if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
1826                 return need_more;
1827             }
1828             if (s->lookahead == 0) break; /* flush the current block */
1829         }
1830 
1831         /* Insert the string window[strstart .. strstart + 2] in the
1832          * dictionary, and set hash_head to the head of the hash chain:
1833          */
1834         hash_head = NIL;
1835         if (s->lookahead >= MIN_MATCH) {
1836             INSERT_STRING(s, s->strstart, hash_head);
1837         }
1838 
1839         /* Find the longest match, discarding those <= prev_length.
1840          * At this point we have always match_length < MIN_MATCH
1841          */
1842         if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) {
1843             /* To simplify the code, we prevent matches with the string
1844              * of window index 0 (in particular we have to avoid a match
1845              * of the string with itself at the start of the input file).
1846              */
1847             s->match_length = longest_match (s, hash_head);
1848             /* longest_match() sets match_start */
1849         }
1850         if (s->match_length >= MIN_MATCH) {
1851             check_match(s, s->strstart, s->match_start, s->match_length);
1852 
1853             _tr_tally_dist(s, s->strstart - s->match_start,
1854                            s->match_length - MIN_MATCH, bflush);
1855 
1856             s->lookahead -= s->match_length;
1857 
1858             /* Insert new strings in the hash table only if the match length
1859              * is not too large. This saves time but degrades compression.
1860              */
1861 #ifndef FASTEST
1862             if (s->match_length <= s->max_insert_length &&
1863                 s->lookahead >= MIN_MATCH) {
1864                 s->match_length--; /* string at strstart already in table */
1865                 do {
1866                     s->strstart++;
1867                     INSERT_STRING(s, s->strstart, hash_head);
1868                     /* strstart never exceeds WSIZE-MAX_MATCH, so there are
1869                      * always MIN_MATCH bytes ahead.
1870                      */
1871                 } while (--s->match_length != 0);
1872                 s->strstart++;
1873             } else
1874 #endif
1875             {
1876                 s->strstart += s->match_length;
1877                 s->match_length = 0;
1878                 s->ins_h = s->window[s->strstart];
1879                 UPDATE_HASH(s, s->ins_h, s->window[s->strstart + 1]);
1880 #if MIN_MATCH != 3
1881                 Call UPDATE_HASH() MIN_MATCH-3 more times
1882 #endif
1883                 /* If lookahead < MIN_MATCH, ins_h is garbage, but it does not
1884                  * matter since it will be recomputed at next deflate call.
1885                  */
1886             }
1887         } else {
1888             /* No match, output a literal byte */
1889             Tracevv((stderr,"%c", s->window[s->strstart]));
1890             _tr_tally_lit(s, s->window[s->strstart], bflush);
1891             s->lookahead--;
1892             s->strstart++;
1893         }
1894         if (bflush) FLUSH_BLOCK(s, 0);
1895     }
1896     s->insert = s->strstart < MIN_MATCH-1 ? s->strstart : MIN_MATCH-1;
1897     if (flush == Z_FINISH) {
1898         FLUSH_BLOCK(s, 1);
1899         return finish_done;
1900     }
1901     if (s->sym_next)
1902         FLUSH_BLOCK(s, 0);
1903     return block_done;
1904 }
1905 
1906 #ifndef FASTEST
1907 /* ===========================================================================
1908  * Same as above, but achieves better compression. We use a lazy
1909  * evaluation for matches: a match is finally adopted only if there is
1910  * no better match at the next window position.
1911  */
1912 local block_state deflate_slow(deflate_state *s, int flush) {
1913     IPos hash_head;          /* head of hash chain */
1914     int bflush;              /* set if current block must be flushed */
1915 
1916     /* Process the input block. */
1917     for (;;) {
1918         /* Make sure that we always have enough lookahead, except
1919          * at the end of the input file. We need MAX_MATCH bytes
1920          * for the next match, plus MIN_MATCH bytes to insert the
1921          * string following the next match.
1922          */
1923         if (s->lookahead < MIN_LOOKAHEAD) {
1924             fill_window(s);
1925             if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
1926                 return need_more;
1927             }
1928             if (s->lookahead == 0) break; /* flush the current block */
1929         }
1930 
1931         /* Insert the string window[strstart .. strstart + 2] in the
1932          * dictionary, and set hash_head to the head of the hash chain:
1933          */
1934         hash_head = NIL;
1935         if (s->lookahead >= MIN_MATCH) {
1936             INSERT_STRING(s, s->strstart, hash_head);
1937         }
1938 
1939         /* Find the longest match, discarding those <= prev_length.
1940          */
1941         s->prev_length = s->match_length, s->prev_match = s->match_start;
1942         s->match_length = MIN_MATCH-1;
1943 
1944         if (hash_head != NIL && s->prev_length < s->max_lazy_match &&
1945             s->strstart - hash_head <= MAX_DIST(s)) {
1946             /* To simplify the code, we prevent matches with the string
1947              * of window index 0 (in particular we have to avoid a match
1948              * of the string with itself at the start of the input file).
1949              */
1950             s->match_length = longest_match (s, hash_head);
1951             /* longest_match() sets match_start */
1952 
1953             if (s->match_length <= 5 && (s->strategy == Z_FILTERED
1954 #if TOO_FAR <= 32767
1955                 || (s->match_length == MIN_MATCH &&
1956                     s->strstart - s->match_start > TOO_FAR)
1957 #endif
1958                 )) {
1959 
1960                 /* If prev_match is also MIN_MATCH, match_start is garbage
1961                  * but we will ignore the current match anyway.
1962                  */
1963                 s->match_length = MIN_MATCH-1;
1964             }
1965         }
1966         /* If there was a match at the previous step and the current
1967          * match is not better, output the previous match:
1968          */
1969         if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) {
1970             uInt max_insert = s->strstart + s->lookahead - MIN_MATCH;
1971             /* Do not insert strings in hash table beyond this. */
1972 
1973             check_match(s, s->strstart - 1, s->prev_match, s->prev_length);
1974 
1975             _tr_tally_dist(s, s->strstart - 1 - s->prev_match,
1976                            s->prev_length - MIN_MATCH, bflush);
1977 
1978             /* Insert in hash table all strings up to the end of the match.
1979              * strstart - 1 and strstart are already inserted. If there is not
1980              * enough lookahead, the last two strings are not inserted in
1981              * the hash table.
1982              */
1983             s->lookahead -= s->prev_length - 1;
1984             s->prev_length -= 2;
1985             do {
1986                 if (++s->strstart <= max_insert) {
1987                     INSERT_STRING(s, s->strstart, hash_head);
1988                 }
1989             } while (--s->prev_length != 0);
1990             s->match_available = 0;
1991             s->match_length = MIN_MATCH-1;
1992             s->strstart++;
1993 
1994             if (bflush) FLUSH_BLOCK(s, 0);
1995 
1996         } else if (s->match_available) {
1997             /* If there was no match at the previous position, output a
1998              * single literal. If there was a match but the current match
1999              * is longer, truncate the previous match to a single literal.
2000              */
2001             Tracevv((stderr,"%c", s->window[s->strstart - 1]));
2002             _tr_tally_lit(s, s->window[s->strstart - 1], bflush);
2003             if (bflush) {
2004                 FLUSH_BLOCK_ONLY(s, 0);
2005             }
2006             s->strstart++;
2007             s->lookahead--;
2008             if (s->strm->avail_out == 0) return need_more;
2009         } else {
2010             /* There is no previous match to compare with, wait for
2011              * the next step to decide.
2012              */
2013             s->match_available = 1;
2014             s->strstart++;
2015             s->lookahead--;
2016         }
2017     }
2018     Assert (flush != Z_NO_FLUSH, "no flush?");
2019     if (s->match_available) {
2020         Tracevv((stderr,"%c", s->window[s->strstart - 1]));
2021         _tr_tally_lit(s, s->window[s->strstart - 1], bflush);
2022         s->match_available = 0;
2023     }
2024     s->insert = s->strstart < MIN_MATCH-1 ? s->strstart : MIN_MATCH-1;
2025     if (flush == Z_FINISH) {
2026         FLUSH_BLOCK(s, 1);
2027         return finish_done;
2028     }
2029     if (s->sym_next)
2030         FLUSH_BLOCK(s, 0);
2031     return block_done;
2032 }
2033 #endif /* FASTEST */
2034 
2035 /* ===========================================================================
2036  * For Z_RLE, simply look for runs of bytes, generate matches only of distance
2037  * one.  Do not maintain a hash table.  (It will be regenerated if this run of
2038  * deflate switches away from Z_RLE.)
2039  */
2040 local block_state deflate_rle(deflate_state *s, int flush) {
2041     int bflush;             /* set if current block must be flushed */
2042     uInt prev;              /* byte at distance one to match */
2043     Bytef *scan, *strend;   /* scan goes up to strend for length of run */
2044 
2045     for (;;) {
2046         /* Make sure that we always have enough lookahead, except
2047          * at the end of the input file. We need MAX_MATCH bytes
2048          * for the longest run, plus one for the unrolled loop.
2049          */
2050         if (s->lookahead <= MAX_MATCH) {
2051             fill_window(s);
2052             if (s->lookahead <= MAX_MATCH && flush == Z_NO_FLUSH) {
2053                 return need_more;
2054             }
2055             if (s->lookahead == 0) break; /* flush the current block */
2056         }
2057 
2058         /* See how many times the previous byte repeats */
2059         s->match_length = 0;
2060         if (s->lookahead >= MIN_MATCH && s->strstart > 0) {
2061             scan = s->window + s->strstart - 1;
2062             prev = *scan;
2063             if (prev == *++scan && prev == *++scan && prev == *++scan) {
2064                 strend = s->window + s->strstart + MAX_MATCH;
2065                 do {
2066                 } while (prev == *++scan && prev == *++scan &&
2067                          prev == *++scan && prev == *++scan &&
2068                          prev == *++scan && prev == *++scan &&
2069                          prev == *++scan && prev == *++scan &&
2070                          scan < strend);
2071                 s->match_length = MAX_MATCH - (uInt)(strend - scan);
2072                 if (s->match_length > s->lookahead)
2073                     s->match_length = s->lookahead;
2074             }
2075             Assert(scan <= s->window + (uInt)(s->window_size - 1),
2076                    "wild scan");
2077         }
2078 
2079         /* Emit match if have run of MIN_MATCH or longer, else emit literal */
2080         if (s->match_length >= MIN_MATCH) {
2081             check_match(s, s->strstart, s->strstart - 1, s->match_length);
2082 
2083             _tr_tally_dist(s, 1, s->match_length - MIN_MATCH, bflush);
2084 
2085             s->lookahead -= s->match_length;
2086             s->strstart += s->match_length;
2087             s->match_length = 0;
2088         } else {
2089             /* No match, output a literal byte */
2090             Tracevv((stderr,"%c", s->window[s->strstart]));
2091             _tr_tally_lit(s, s->window[s->strstart], bflush);
2092             s->lookahead--;
2093             s->strstart++;
2094         }
2095         if (bflush) FLUSH_BLOCK(s, 0);
2096     }
2097     s->insert = 0;
2098     if (flush == Z_FINISH) {
2099         FLUSH_BLOCK(s, 1);
2100         return finish_done;
2101     }
2102     if (s->sym_next)
2103         FLUSH_BLOCK(s, 0);
2104     return block_done;
2105 }
2106 
2107 /* ===========================================================================
2108  * For Z_HUFFMAN_ONLY, do not look for matches.  Do not maintain a hash table.
2109  * (It will be regenerated if this run of deflate switches away from Huffman.)
2110  */
2111 local block_state deflate_huff(deflate_state *s, int flush) {
2112     int bflush;             /* set if current block must be flushed */
2113 
2114     for (;;) {
2115         /* Make sure that we have a literal to write. */
2116         if (s->lookahead == 0) {
2117             fill_window(s);
2118             if (s->lookahead == 0) {
2119                 if (flush == Z_NO_FLUSH)
2120                     return need_more;
2121                 break;      /* flush the current block */
2122             }
2123         }
2124 
2125         /* Output a literal byte */
2126         s->match_length = 0;
2127         Tracevv((stderr,"%c", s->window[s->strstart]));
2128         _tr_tally_lit(s, s->window[s->strstart], bflush);
2129         s->lookahead--;
2130         s->strstart++;
2131         if (bflush) FLUSH_BLOCK(s, 0);
2132     }
2133     s->insert = 0;
2134     if (flush == Z_FINISH) {
2135         FLUSH_BLOCK(s, 1);
2136         return finish_done;
2137     }
2138     if (s->sym_next)
2139         FLUSH_BLOCK(s, 0);
2140     return block_done;
2141 }
2142