1/* deflate.c -- compress data using the deflation algorithm
2 * Copyright (C) 1995-2010 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://www.ietf.org/rfc/rfc1951.txt
41 *
42 *      A description of the Rabin and Karp algorithm is given in the book
43 *         "Algorithms" by R. Sedgewick, Addison-Wesley, p252.
44 *
45 *      Fiala,E.R., and Greene,D.H.
46 *         Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595
47 *
48 */
49
50/* @(#) $Id$ */
51
52#include "deflate.h"
53
54const char deflate_copyright[] =
55   " deflate 1.2.5 Copyright 1995-2010 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/* ===========================================================================
64 *  Function prototypes.
65 */
66typedef enum {
67    need_more,      /* block not completed, need more input or more output */
68    block_done,     /* block flush performed */
69    finish_started, /* finish started, need only more output at next deflate */
70    finish_done     /* finish done, accept no more input or output */
71} block_state;
72
73typedef block_state (*compress_func) OF((deflate_state *s, int flush));
74/* Compression function. Returns the block state after the call. */
75
76local void fill_window    OF((deflate_state *s));
77local block_state deflate_stored OF((deflate_state *s, int flush));
78local block_state deflate_fast   OF((deflate_state *s, int flush));
79#ifndef FASTEST
80local block_state deflate_slow   OF((deflate_state *s, int flush));
81#endif
82local block_state deflate_rle    OF((deflate_state *s, int flush));
83local block_state deflate_huff   OF((deflate_state *s, int flush));
84local void lm_init        OF((deflate_state *s));
85local void putShortMSB    OF((deflate_state *s, uInt b));
86local void flush_pending  OF((z_streamp strm));
87local int read_buf        OF((z_streamp strm, Bytef *buf, unsigned size));
88#ifdef ASMV
89      void match_init OF((void)); /* asm code initialization */
90      uInt longest_match  OF((deflate_state *s, IPos cur_match));
91#else
92local uInt longest_match  OF((deflate_state *s, IPos cur_match));
93#endif
94
95#ifdef DEBUG
96local  void check_match OF((deflate_state *s, IPos start, IPos match,
97                            int length));
98#endif
99
100/* ===========================================================================
101 * Local data
102 */
103
104#define NIL 0
105/* Tail of hash chains */
106
107#ifndef TOO_FAR
108#  define TOO_FAR 4096
109#endif
110/* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
111
112/* Values for max_lazy_match, good_match and max_chain_length, depending on
113 * the desired pack level (0..9). The values given below have been tuned to
114 * exclude worst case performance for pathological files. Better values may be
115 * found for specific files.
116 */
117typedef struct config_s {
118   ush good_length; /* reduce lazy search above this match length */
119   ush max_lazy;    /* do not perform lazy search above this match length */
120   ush nice_length; /* quit search above this match length */
121   ush max_chain;
122   compress_func func;
123} config;
124
125#ifdef FASTEST
126local const config configuration_table[2] = {
127/*      good lazy nice chain */
128/* 0 */ {0,    0,  0,    0, deflate_stored},  /* store only */
129/* 1 */ {4,    4,  8,    4, deflate_fast}}; /* max speed, no lazy matches */
130#else
131local const config configuration_table[10] = {
132/*      good lazy nice chain */
133/* 0 */ {0,    0,  0,    0, deflate_stored},  /* store only */
134/* 1 */ {4,    4,  8,    4, deflate_fast}, /* max speed, no lazy matches */
135/* 2 */ {4,    5, 16,    8, deflate_fast},
136/* 3 */ {4,    6, 32,   32, deflate_fast},
137
138/* 4 */ {4,    4, 16,   16, deflate_slow},  /* lazy matches */
139/* 5 */ {8,   16, 32,   32, deflate_slow},
140/* 6 */ {8,   16, 128, 128, deflate_slow},
141/* 7 */ {8,   32, 128, 256, deflate_slow},
142/* 8 */ {32, 128, 258, 1024, deflate_slow},
143/* 9 */ {32, 258, 258, 4096, deflate_slow}}; /* max compression */
144#endif
145
146/* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
147 * For deflate_fast() (levels <= 3) good is ignored and lazy has a different
148 * meaning.
149 */
150
151#define EQUAL 0
152/* result of memcmp for equal strings */
153
154#ifndef NO_DUMMY_DECL
155struct static_tree_desc_s {int dummy;}; /* for buggy compilers */
156#endif
157
158/* ===========================================================================
159 * Update a hash value with the given input byte
160 * IN  assertion: all calls to to UPDATE_HASH are made with consecutive
161 *    input characters, so that a running hash key can be computed from the
162 *    previous key instead of complete recalculation each time.
163 */
164#define UPDATE_HASH(s,h,c) (h = (((h)<<s->hash_shift) ^ (c)) & s->hash_mask)
165
166
167/* ===========================================================================
168 * Insert string str in the dictionary and set match_head to the previous head
169 * of the hash chain (the most recent string with same hash key). Return
170 * the previous length of the hash chain.
171 * If this file is compiled with -DFASTEST, the compression level is forced
172 * to 1, and no hash chains are maintained.
173 * IN  assertion: all calls to to INSERT_STRING are made with consecutive
174 *    input characters and the first MIN_MATCH bytes of str are valid
175 *    (except for the last MIN_MATCH-1 bytes of the input file).
176 */
177#ifdef FASTEST
178#define INSERT_STRING(s, str, match_head) \
179   (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
180    match_head = s->head[s->ins_h], \
181    s->head[s->ins_h] = (Pos)(str))
182#else
183#define INSERT_STRING(s, str, match_head) \
184   (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
185    match_head = s->prev[(str) & s->w_mask] = s->head[s->ins_h], \
186    s->head[s->ins_h] = (Pos)(str))
187#endif
188
189/* ===========================================================================
190 * Initialize the hash table (avoiding 64K overflow for 16 bit systems).
191 * prev[] will be initialized on the fly.
192 */
193#define CLEAR_HASH(s) \
194    s->head[s->hash_size-1] = NIL; \
195    zmemzero((Bytef *)s->head, (unsigned)(s->hash_size-1)*sizeof(*s->head));
196
197/* ========================================================================= */
198int ZEXPORT deflateInit_(strm, level, version, stream_size)
199    z_streamp strm;
200    int level;
201    const char *version;
202    int stream_size;
203{
204    return deflateInit2_(strm, level, Z_DEFLATED, MAX_WBITS, DEF_MEM_LEVEL,
205                         Z_DEFAULT_STRATEGY, version, stream_size);
206    /* To do: ignore strm->next_in if we use it as window */
207}
208
209/* ========================================================================= */
210int ZEXPORT deflateInit2_(strm, level, method, windowBits, memLevel, strategy,
211                  version, stream_size)
212    z_streamp strm;
213    int  level;
214    int  method;
215    int  windowBits;
216    int  memLevel;
217    int  strategy;
218    const char *version;
219    int stream_size;
220{
221    deflate_state *s;
222    int wrap = 1;
223    static const char my_version[] = ZLIB_VERSION;
224
225    ushf *overlay;
226    /* We overlay pending_buf and d_buf+l_buf. This works since the average
227     * output size for (length,distance) codes is <= 24 bits.
228     */
229
230    if (version == Z_NULL || version[0] != my_version[0] ||
231        stream_size != sizeof(z_stream)) {
232        return Z_VERSION_ERROR;
233    }
234    if (strm == Z_NULL) return Z_STREAM_ERROR;
235
236    strm->msg = Z_NULL;
237    if (strm->zalloc == (alloc_func)0) {
238        strm->zalloc = zcalloc;
239        strm->opaque = (voidpf)0;
240    }
241    if (strm->zfree == (free_func)0) strm->zfree = zcfree;
242
243#ifdef FASTEST
244    if (level != 0) level = 1;
245#else
246    if (level == Z_DEFAULT_COMPRESSION) level = 6;
247#endif
248
249    if (windowBits < 0) { /* suppress zlib wrapper */
250        wrap = 0;
251        windowBits = -windowBits;
252    }
253#ifdef GZIP
254    else if (windowBits > 15) {
255        wrap = 2;       /* write gzip wrapper instead */
256        windowBits -= 16;
257    }
258#endif
259    if (memLevel < 1 || memLevel > MAX_MEM_LEVEL || method != Z_DEFLATED ||
260        windowBits < 8 || windowBits > 15 || level < 0 || level > 9 ||
261        strategy < 0 || strategy > Z_FIXED) {
262        return Z_STREAM_ERROR;
263    }
264    if (windowBits == 8) windowBits = 9;  /* until 256-byte window bug fixed */
265    s = (deflate_state *) ZALLOC(strm, 1, sizeof(deflate_state));
266    if (s == Z_NULL) return Z_MEM_ERROR;
267    strm->state = (struct internal_state FAR *)s;
268    s->strm = strm;
269
270    s->wrap = wrap;
271    s->gzhead = Z_NULL;
272    s->w_bits = windowBits;
273    s->w_size = 1 << s->w_bits;
274    s->w_mask = s->w_size - 1;
275
276    s->hash_bits = memLevel + 7;
277    s->hash_size = 1 << s->hash_bits;
278    s->hash_mask = s->hash_size - 1;
279    s->hash_shift =  ((s->hash_bits+MIN_MATCH-1)/MIN_MATCH);
280
281    s->window = (Bytef *) ZALLOC(strm, s->w_size, 2*sizeof(Byte));
282    s->prev   = (Posf *)  ZALLOC(strm, s->w_size, sizeof(Pos));
283    s->head   = (Posf *)  ZALLOC(strm, s->hash_size, sizeof(Pos));
284
285    s->high_water = 0;      /* nothing written to s->window yet */
286
287    s->lit_bufsize = 1 << (memLevel + 6); /* 16K elements by default */
288
289    overlay = (ushf *) ZALLOC(strm, s->lit_bufsize, sizeof(ush)+2);
290    s->pending_buf = (uchf *) overlay;
291    s->pending_buf_size = (ulg)s->lit_bufsize * (sizeof(ush)+2L);
292
293    if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL ||
294        s->pending_buf == Z_NULL) {
295        s->status = FINISH_STATE;
296        strm->msg = (char*)ERR_MSG(Z_MEM_ERROR);
297        deflateEnd (strm);
298        return Z_MEM_ERROR;
299    }
300    s->d_buf = overlay + s->lit_bufsize/sizeof(ush);
301    s->l_buf = s->pending_buf + (1+sizeof(ush))*s->lit_bufsize;
302
303    s->level = level;
304    s->strategy = strategy;
305    s->method = (Byte)method;
306
307    return deflateReset(strm);
308}
309
310/* ========================================================================= */
311int ZEXPORT deflateSetDictionary (strm, dictionary, dictLength)
312    z_streamp strm;
313    const Bytef *dictionary;
314    uInt  dictLength;
315{
316    deflate_state *s;
317    uInt length = dictLength;
318    uInt n;
319    IPos hash_head = 0;
320
321    if (strm == Z_NULL || strm->state == Z_NULL || dictionary == Z_NULL ||
322        strm->state->wrap == 2 ||
323        (strm->state->wrap == 1 && strm->state->status != INIT_STATE))
324        return Z_STREAM_ERROR;
325
326    s = strm->state;
327    if (s->wrap)
328        strm->adler = adler32(strm->adler, dictionary, dictLength);
329
330    if (length < MIN_MATCH) return Z_OK;
331    if (length > s->w_size) {
332        length = s->w_size;
333        dictionary += dictLength - length; /* use the tail of the dictionary */
334    }
335    zmemcpy(s->window, dictionary, length);
336    s->strstart = length;
337    s->block_start = (long)length;
338
339    /* Insert all strings in the hash table (except for the last two bytes).
340     * s->lookahead stays null, so s->ins_h will be recomputed at the next
341     * call of fill_window.
342     */
343    s->ins_h = s->window[0];
344    UPDATE_HASH(s, s->ins_h, s->window[1]);
345    for (n = 0; n <= length - MIN_MATCH; n++) {
346        INSERT_STRING(s, n, hash_head);
347    }
348    if (hash_head) hash_head = 0;  /* to make compiler happy */
349    return Z_OK;
350}
351
352/* ========================================================================= */
353int ZEXPORT deflateReset (strm)
354    z_streamp strm;
355{
356    deflate_state *s;
357
358    if (strm == Z_NULL || strm->state == Z_NULL ||
359        strm->zalloc == (alloc_func)0 || strm->zfree == (free_func)0) {
360        return Z_STREAM_ERROR;
361    }
362
363    strm->total_in = strm->total_out = 0;
364    strm->msg = Z_NULL; /* use zfree if we ever allocate msg dynamically */
365    strm->data_type = Z_UNKNOWN;
366
367    s = (deflate_state *)strm->state;
368    s->pending = 0;
369    s->pending_out = s->pending_buf;
370
371    if (s->wrap < 0) {
372        s->wrap = -s->wrap; /* was made negative by deflate(..., Z_FINISH); */
373    }
374    s->status = s->wrap ? INIT_STATE : BUSY_STATE;
375    strm->adler =
376#ifdef GZIP
377        s->wrap == 2 ? crc32(0L, Z_NULL, 0) :
378#endif
379        adler32(0L, Z_NULL, 0);
380    s->last_flush = Z_NO_FLUSH;
381
382    _tr_init(s);
383    lm_init(s);
384
385    return Z_OK;
386}
387
388/* ========================================================================= */
389int ZEXPORT deflateSetHeader (strm, head)
390    z_streamp strm;
391    gz_headerp head;
392{
393    if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
394    if (strm->state->wrap != 2) return Z_STREAM_ERROR;
395    strm->state->gzhead = head;
396    return Z_OK;
397}
398
399/* ========================================================================= */
400int ZEXPORT deflatePrime (strm, bits, value)
401    z_streamp strm;
402    int bits;
403    int value;
404{
405    if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
406    strm->state->bi_valid = bits;
407    strm->state->bi_buf = (ush)(value & ((1 << bits) - 1));
408    return Z_OK;
409}
410
411/* ========================================================================= */
412int ZEXPORT deflateParams(strm, level, strategy)
413    z_streamp strm;
414    int level;
415    int strategy;
416{
417    deflate_state *s;
418    compress_func func;
419    int err = Z_OK;
420
421    if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
422    s = strm->state;
423
424#ifdef FASTEST
425    if (level != 0) level = 1;
426#else
427    if (level == Z_DEFAULT_COMPRESSION) level = 6;
428#endif
429    if (level < 0 || level > 9 || strategy < 0 || strategy > Z_FIXED) {
430        return Z_STREAM_ERROR;
431    }
432    func = configuration_table[s->level].func;
433
434    if ((strategy != s->strategy || func != configuration_table[level].func) &&
435        strm->total_in != 0) {
436        /* Flush the last buffer: */
437        err = deflate(strm, Z_BLOCK);
438    }
439    if (s->level != level) {
440        s->level = level;
441        s->max_lazy_match   = configuration_table[level].max_lazy;
442        s->good_match       = configuration_table[level].good_length;
443        s->nice_match       = configuration_table[level].nice_length;
444        s->max_chain_length = configuration_table[level].max_chain;
445    }
446    s->strategy = strategy;
447    return err;
448}
449
450/* ========================================================================= */
451int ZEXPORT deflateTune(strm, good_length, max_lazy, nice_length, max_chain)
452    z_streamp strm;
453    int good_length;
454    int max_lazy;
455    int nice_length;
456    int max_chain;
457{
458    deflate_state *s;
459
460    if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
461    s = strm->state;
462    s->good_match = good_length;
463    s->max_lazy_match = max_lazy;
464    s->nice_match = nice_length;
465    s->max_chain_length = max_chain;
466    return Z_OK;
467}
468
469/* =========================================================================
470 * For the default windowBits of 15 and memLevel of 8, this function returns
471 * a close to exact, as well as small, upper bound on the compressed size.
472 * They are coded as constants here for a reason--if the #define's are
473 * changed, then this function needs to be changed as well.  The return
474 * value for 15 and 8 only works for those exact settings.
475 *
476 * For any setting other than those defaults for windowBits and memLevel,
477 * the value returned is a conservative worst case for the maximum expansion
478 * resulting from using fixed blocks instead of stored blocks, which deflate
479 * can emit on compressed data for some combinations of the parameters.
480 *
481 * This function could be more sophisticated to provide closer upper bounds for
482 * every combination of windowBits and memLevel.  But even the conservative
483 * upper bound of about 14% expansion does not seem onerous for output buffer
484 * allocation.
485 */
486uLong ZEXPORT deflateBound(strm, sourceLen)
487    z_streamp strm;
488    uLong sourceLen;
489{
490    deflate_state *s;
491    uLong complen, wraplen;
492    Bytef *str;
493
494    /* conservative upper bound for compressed data */
495    complen = sourceLen +
496              ((sourceLen + 7) >> 3) + ((sourceLen + 63) >> 6) + 5;
497
498    /* if can't get parameters, return conservative bound plus zlib wrapper */
499    if (strm == Z_NULL || strm->state == Z_NULL)
500        return complen + 6;
501
502    /* compute wrapper length */
503    s = strm->state;
504    switch (s->wrap) {
505    case 0:                                 /* raw deflate */
506        wraplen = 0;
507        break;
508    case 1:                                 /* zlib wrapper */
509        wraplen = 6 + (s->strstart ? 4 : 0);
510        break;
511    case 2:                                 /* gzip wrapper */
512        wraplen = 18;
513        if (s->gzhead != Z_NULL) {          /* user-supplied gzip header */
514            if (s->gzhead->extra != Z_NULL)
515                wraplen += 2 + s->gzhead->extra_len;
516            str = s->gzhead->name;
517            if (str != Z_NULL)
518                do {
519                    wraplen++;
520                } while (*str++);
521            str = s->gzhead->comment;
522            if (str != Z_NULL)
523                do {
524                    wraplen++;
525                } while (*str++);
526            if (s->gzhead->hcrc)
527                wraplen += 2;
528        }
529        break;
530    default:                                /* for compiler happiness */
531        wraplen = 6;
532    }
533
534    /* if not default parameters, return conservative bound */
535    if (s->w_bits != 15 || s->hash_bits != 8 + 7)
536        return complen + wraplen;
537
538    /* default settings: return tight bound for that case */
539    return sourceLen + (sourceLen >> 12) + (sourceLen >> 14) +
540           (sourceLen >> 25) + 13 - 6 + wraplen;
541}
542
543/* =========================================================================
544 * Put a short in the pending buffer. The 16-bit value is put in MSB order.
545 * IN assertion: the stream state is correct and there is enough room in
546 * pending_buf.
547 */
548local void putShortMSB (s, b)
549    deflate_state *s;
550    uInt b;
551{
552    put_byte(s, (Byte)(b >> 8));
553    put_byte(s, (Byte)(b & 0xff));
554}
555
556/* =========================================================================
557 * Flush as much pending output as possible. All deflate() output goes
558 * through this function so some applications may wish to modify it
559 * to avoid allocating a large strm->next_out buffer and copying into it.
560 * (See also read_buf()).
561 */
562local void flush_pending(strm)
563    z_streamp strm;
564{
565    unsigned len = strm->state->pending;
566
567    if (len > strm->avail_out) len = strm->avail_out;
568    if (len == 0) return;
569
570    zmemcpy(strm->next_out, strm->state->pending_out, len);
571    strm->next_out  += len;
572    strm->state->pending_out  += len;
573    strm->total_out += len;
574    strm->avail_out  -= len;
575    strm->state->pending -= len;
576    if (strm->state->pending == 0) {
577        strm->state->pending_out = strm->state->pending_buf;
578    }
579}
580
581/* ========================================================================= */
582int ZEXPORT deflate (strm, flush)
583    z_streamp strm;
584    int flush;
585{
586    int old_flush; /* value of flush param for previous deflate call */
587    deflate_state *s;
588
589    if (strm == Z_NULL || strm->state == Z_NULL ||
590        flush > Z_BLOCK || flush < 0) {
591        return Z_STREAM_ERROR;
592    }
593    s = strm->state;
594
595    if (strm->next_out == Z_NULL ||
596        (strm->next_in == Z_NULL && strm->avail_in != 0) ||
597        (s->status == FINISH_STATE && flush != Z_FINISH)) {
598        ERR_RETURN(strm, Z_STREAM_ERROR);
599    }
600    if (strm->avail_out == 0) ERR_RETURN(strm, Z_BUF_ERROR);
601
602    s->strm = strm; /* just in case */
603    old_flush = s->last_flush;
604    s->last_flush = flush;
605
606    /* Write the header */
607    if (s->status == INIT_STATE) {
608#ifdef GZIP
609        if (s->wrap == 2) {
610            strm->adler = crc32(0L, Z_NULL, 0);
611            put_byte(s, 31);
612            put_byte(s, 139);
613            put_byte(s, 8);
614            if (s->gzhead == Z_NULL) {
615                put_byte(s, 0);
616                put_byte(s, 0);
617                put_byte(s, 0);
618                put_byte(s, 0);
619                put_byte(s, 0);
620                put_byte(s, s->level == 9 ? 2 :
621                            (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
622                             4 : 0));
623                put_byte(s, OS_CODE);
624                s->status = BUSY_STATE;
625            }
626            else {
627                put_byte(s, (s->gzhead->text ? 1 : 0) +
628                            (s->gzhead->hcrc ? 2 : 0) +
629                            (s->gzhead->extra == Z_NULL ? 0 : 4) +
630                            (s->gzhead->name == Z_NULL ? 0 : 8) +
631                            (s->gzhead->comment == Z_NULL ? 0 : 16)
632                        );
633                put_byte(s, (Byte)(s->gzhead->time & 0xff));
634                put_byte(s, (Byte)((s->gzhead->time >> 8) & 0xff));
635                put_byte(s, (Byte)((s->gzhead->time >> 16) & 0xff));
636                put_byte(s, (Byte)((s->gzhead->time >> 24) & 0xff));
637                put_byte(s, s->level == 9 ? 2 :
638                            (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
639                             4 : 0));
640                put_byte(s, s->gzhead->os & 0xff);
641                if (s->gzhead->extra != Z_NULL) {
642                    put_byte(s, s->gzhead->extra_len & 0xff);
643                    put_byte(s, (s->gzhead->extra_len >> 8) & 0xff);
644                }
645                if (s->gzhead->hcrc)
646                    strm->adler = crc32(strm->adler, s->pending_buf,
647                                        s->pending);
648                s->gzindex = 0;
649                s->status = EXTRA_STATE;
650            }
651        }
652        else
653#endif
654        {
655            uInt header = (Z_DEFLATED + ((s->w_bits-8)<<4)) << 8;
656            uInt level_flags;
657
658            if (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2)
659                level_flags = 0;
660            else if (s->level < 6)
661                level_flags = 1;
662            else if (s->level == 6)
663                level_flags = 2;
664            else
665                level_flags = 3;
666            header |= (level_flags << 6);
667            if (s->strstart != 0) header |= PRESET_DICT;
668            header += 31 - (header % 31);
669
670            s->status = BUSY_STATE;
671            putShortMSB(s, header);
672
673            /* Save the adler32 of the preset dictionary: */
674            if (s->strstart != 0) {
675                putShortMSB(s, (uInt)(strm->adler >> 16));
676                putShortMSB(s, (uInt)(strm->adler & 0xffff));
677            }
678            strm->adler = adler32(0L, Z_NULL, 0);
679        }
680    }
681#ifdef GZIP
682    if (s->status == EXTRA_STATE) {
683        if (s->gzhead->extra != Z_NULL) {
684            uInt beg = s->pending;  /* start of bytes to update crc */
685
686            while (s->gzindex < (s->gzhead->extra_len & 0xffff)) {
687                if (s->pending == s->pending_buf_size) {
688                    if (s->gzhead->hcrc && s->pending > beg)
689                        strm->adler = crc32(strm->adler, s->pending_buf + beg,
690                                            s->pending - beg);
691                    flush_pending(strm);
692                    beg = s->pending;
693                    if (s->pending == s->pending_buf_size)
694                        break;
695                }
696                put_byte(s, s->gzhead->extra[s->gzindex]);
697                s->gzindex++;
698            }
699            if (s->gzhead->hcrc && s->pending > beg)
700                strm->adler = crc32(strm->adler, s->pending_buf + beg,
701                                    s->pending - beg);
702            if (s->gzindex == s->gzhead->extra_len) {
703                s->gzindex = 0;
704                s->status = NAME_STATE;
705            }
706        }
707        else
708            s->status = NAME_STATE;
709    }
710    if (s->status == NAME_STATE) {
711        if (s->gzhead->name != Z_NULL) {
712            uInt beg = s->pending;  /* start of bytes to update crc */
713            int val;
714
715            do {
716                if (s->pending == s->pending_buf_size) {
717                    if (s->gzhead->hcrc && s->pending > beg)
718                        strm->adler = crc32(strm->adler, s->pending_buf + beg,
719                                            s->pending - beg);
720                    flush_pending(strm);
721                    beg = s->pending;
722                    if (s->pending == s->pending_buf_size) {
723                        val = 1;
724                        break;
725                    }
726                }
727                val = s->gzhead->name[s->gzindex++];
728                put_byte(s, val);
729            } while (val != 0);
730            if (s->gzhead->hcrc && s->pending > beg)
731                strm->adler = crc32(strm->adler, s->pending_buf + beg,
732                                    s->pending - beg);
733            if (val == 0) {
734                s->gzindex = 0;
735                s->status = COMMENT_STATE;
736            }
737        }
738        else
739            s->status = COMMENT_STATE;
740    }
741    if (s->status == COMMENT_STATE) {
742        if (s->gzhead->comment != Z_NULL) {
743            uInt beg = s->pending;  /* start of bytes to update crc */
744            int val;
745
746            do {
747                if (s->pending == s->pending_buf_size) {
748                    if (s->gzhead->hcrc && s->pending > beg)
749                        strm->adler = crc32(strm->adler, s->pending_buf + beg,
750                                            s->pending - beg);
751                    flush_pending(strm);
752                    beg = s->pending;
753                    if (s->pending == s->pending_buf_size) {
754                        val = 1;
755                        break;
756                    }
757                }
758                val = s->gzhead->comment[s->gzindex++];
759                put_byte(s, val);
760            } while (val != 0);
761            if (s->gzhead->hcrc && s->pending > beg)
762                strm->adler = crc32(strm->adler, s->pending_buf + beg,
763                                    s->pending - beg);
764            if (val == 0)
765                s->status = HCRC_STATE;
766        }
767        else
768            s->status = HCRC_STATE;
769    }
770    if (s->status == HCRC_STATE) {
771        if (s->gzhead->hcrc) {
772            if (s->pending + 2 > s->pending_buf_size)
773                flush_pending(strm);
774            if (s->pending + 2 <= s->pending_buf_size) {
775                put_byte(s, (Byte)(strm->adler & 0xff));
776                put_byte(s, (Byte)((strm->adler >> 8) & 0xff));
777                strm->adler = crc32(0L, Z_NULL, 0);
778                s->status = BUSY_STATE;
779            }
780        }
781        else
782            s->status = BUSY_STATE;
783    }
784#endif
785
786    /* Flush as much pending output as possible */
787    if (s->pending != 0) {
788        flush_pending(strm);
789        if (strm->avail_out == 0) {
790            /* Since avail_out is 0, deflate will be called again with
791             * more output space, but possibly with both pending and
792             * avail_in equal to zero. There won't be anything to do,
793             * but this is not an error situation so make sure we
794             * return OK instead of BUF_ERROR at next call of deflate:
795             */
796            s->last_flush = -1;
797            return Z_OK;
798        }
799
800    /* Make sure there is something to do and avoid duplicate consecutive
801     * flushes. For repeated and useless calls with Z_FINISH, we keep
802     * returning Z_STREAM_END instead of Z_BUF_ERROR.
803     */
804    } else if (strm->avail_in == 0 && flush <= old_flush &&
805               flush != Z_FINISH) {
806        ERR_RETURN(strm, Z_BUF_ERROR);
807    }
808
809    /* User must not provide more input after the first FINISH: */
810    if (s->status == FINISH_STATE && strm->avail_in != 0) {
811        ERR_RETURN(strm, Z_BUF_ERROR);
812    }
813
814    /* Start a new block or continue the current one.
815     */
816    if (strm->avail_in != 0 || s->lookahead != 0 ||
817        (flush != Z_NO_FLUSH && s->status != FINISH_STATE)) {
818        block_state bstate;
819
820        bstate = s->strategy == Z_HUFFMAN_ONLY ? deflate_huff(s, flush) :
821                    (s->strategy == Z_RLE ? deflate_rle(s, flush) :
822                        (*(configuration_table[s->level].func))(s, flush));
823
824        if (bstate == finish_started || bstate == finish_done) {
825            s->status = FINISH_STATE;
826        }
827        if (bstate == need_more || bstate == finish_started) {
828            if (strm->avail_out == 0) {
829                s->last_flush = -1; /* avoid BUF_ERROR next call, see above */
830            }
831            return Z_OK;
832            /* If flush != Z_NO_FLUSH && avail_out == 0, the next call
833             * of deflate should use the same flush parameter to make sure
834             * that the flush is complete. So we don't have to output an
835             * empty block here, this will be done at next call. This also
836             * ensures that for a very small output buffer, we emit at most
837             * one empty block.
838             */
839        }
840        if (bstate == block_done) {
841            if (flush == Z_PARTIAL_FLUSH) {
842                _tr_align(s);
843            } else if (flush != Z_BLOCK) { /* FULL_FLUSH or SYNC_FLUSH */
844                _tr_stored_block(s, (char*)0, 0L, 0);
845                /* For a full flush, this empty block will be recognized
846                 * as a special marker by inflate_sync().
847                 */
848                if (flush == Z_FULL_FLUSH) {
849                    CLEAR_HASH(s);             /* forget history */
850                    if (s->lookahead == 0) {
851                        s->strstart = 0;
852                        s->block_start = 0L;
853                    }
854                }
855            }
856            flush_pending(strm);
857            if (strm->avail_out == 0) {
858              s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */
859              return Z_OK;
860            }
861        }
862    }
863    Assert(strm->avail_out > 0, "bug2");
864
865    if (flush != Z_FINISH) return Z_OK;
866    if (s->wrap <= 0) return Z_STREAM_END;
867
868    /* Write the trailer */
869#ifdef GZIP
870    if (s->wrap == 2) {
871        put_byte(s, (Byte)(strm->adler & 0xff));
872        put_byte(s, (Byte)((strm->adler >> 8) & 0xff));
873        put_byte(s, (Byte)((strm->adler >> 16) & 0xff));
874        put_byte(s, (Byte)((strm->adler >> 24) & 0xff));
875        put_byte(s, (Byte)(strm->total_in & 0xff));
876        put_byte(s, (Byte)((strm->total_in >> 8) & 0xff));
877        put_byte(s, (Byte)((strm->total_in >> 16) & 0xff));
878        put_byte(s, (Byte)((strm->total_in >> 24) & 0xff));
879    }
880    else
881#endif
882    {
883        putShortMSB(s, (uInt)(strm->adler >> 16));
884        putShortMSB(s, (uInt)(strm->adler & 0xffff));
885    }
886    flush_pending(strm);
887    /* If avail_out is zero, the application will call deflate again
888     * to flush the rest.
889     */
890    if (s->wrap > 0) s->wrap = -s->wrap; /* write the trailer only once! */
891    return s->pending != 0 ? Z_OK : Z_STREAM_END;
892}
893
894/* ========================================================================= */
895int ZEXPORT deflateEnd (strm)
896    z_streamp strm;
897{
898    int status;
899
900    if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
901
902    status = strm->state->status;
903    if (status != INIT_STATE &&
904        status != EXTRA_STATE &&
905        status != NAME_STATE &&
906        status != COMMENT_STATE &&
907        status != HCRC_STATE &&
908        status != BUSY_STATE &&
909        status != FINISH_STATE) {
910      return Z_STREAM_ERROR;
911    }
912
913    /* Deallocate in reverse order of allocations: */
914    TRY_FREE(strm, strm->state->pending_buf);
915    TRY_FREE(strm, strm->state->head);
916    TRY_FREE(strm, strm->state->prev);
917    TRY_FREE(strm, strm->state->window);
918
919    ZFREE(strm, strm->state);
920    strm->state = Z_NULL;
921
922    return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK;
923}
924
925/* =========================================================================
926 * Copy the source state to the destination state.
927 * To simplify the source, this is not supported for 16-bit MSDOS (which
928 * doesn't have enough memory anyway to duplicate compression states).
929 */
930int ZEXPORT deflateCopy (dest, source)
931    z_streamp dest;
932    z_streamp source;
933{
934#ifdef MAXSEG_64K
935    return Z_STREAM_ERROR;
936#else
937    deflate_state *ds;
938    deflate_state *ss;
939    ushf *overlay;
940
941
942    if (source == Z_NULL || dest == Z_NULL || source->state == Z_NULL) {
943        return Z_STREAM_ERROR;
944    }
945
946    ss = source->state;
947
948    zmemcpy(dest, source, sizeof(z_stream));
949
950    ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state));
951    if (ds == Z_NULL) return Z_MEM_ERROR;
952    dest->state = (struct internal_state FAR *) ds;
953    zmemcpy(ds, ss, sizeof(deflate_state));
954    ds->strm = dest;
955
956    ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte));
957    ds->prev   = (Posf *)  ZALLOC(dest, ds->w_size, sizeof(Pos));
958    ds->head   = (Posf *)  ZALLOC(dest, ds->hash_size, sizeof(Pos));
959    overlay = (ushf *) ZALLOC(dest, ds->lit_bufsize, sizeof(ush)+2);
960    ds->pending_buf = (uchf *) overlay;
961
962    if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL ||
963        ds->pending_buf == Z_NULL) {
964        deflateEnd (dest);
965        return Z_MEM_ERROR;
966    }
967    /* following zmemcpy do not work for 16-bit MSDOS */
968    zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte));
969    zmemcpy(ds->prev, ss->prev, ds->w_size * sizeof(Pos));
970    zmemcpy(ds->head, ss->head, ds->hash_size * sizeof(Pos));
971    zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size);
972
973    ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf);
974    ds->d_buf = overlay + ds->lit_bufsize/sizeof(ush);
975    ds->l_buf = ds->pending_buf + (1+sizeof(ush))*ds->lit_bufsize;
976
977    ds->l_desc.dyn_tree = ds->dyn_ltree;
978    ds->d_desc.dyn_tree = ds->dyn_dtree;
979    ds->bl_desc.dyn_tree = ds->bl_tree;
980
981    return Z_OK;
982#endif /* MAXSEG_64K */
983}
984
985/* ===========================================================================
986 * Read a new buffer from the current input stream, update the adler32
987 * and total number of bytes read.  All deflate() input goes through
988 * this function so some applications may wish to modify it to avoid
989 * allocating a large strm->next_in buffer and copying from it.
990 * (See also flush_pending()).
991 */
992local int read_buf(strm, buf, size)
993    z_streamp strm;
994    Bytef *buf;
995    unsigned size;
996{
997    unsigned len = strm->avail_in;
998
999    if (len > size) len = size;
1000    if (len == 0) return 0;
1001
1002    strm->avail_in  -= len;
1003
1004    if (strm->state->wrap == 1) {
1005        strm->adler = adler32(strm->adler, strm->next_in, len);
1006    }
1007#ifdef GZIP
1008    else if (strm->state->wrap == 2) {
1009        strm->adler = crc32(strm->adler, strm->next_in, len);
1010    }
1011#endif
1012    zmemcpy(buf, strm->next_in, len);
1013    strm->next_in  += len;
1014    strm->total_in += len;
1015
1016    return (int)len;
1017}
1018
1019/* ===========================================================================
1020 * Initialize the "longest match" routines for a new zlib stream
1021 */
1022local void lm_init (s)
1023    deflate_state *s;
1024{
1025    s->window_size = (ulg)2L*s->w_size;
1026
1027    CLEAR_HASH(s);
1028
1029    /* Set the default configuration parameters:
1030     */
1031    s->max_lazy_match   = configuration_table[s->level].max_lazy;
1032    s->good_match       = configuration_table[s->level].good_length;
1033    s->nice_match       = configuration_table[s->level].nice_length;
1034    s->max_chain_length = configuration_table[s->level].max_chain;
1035
1036    s->strstart = 0;
1037    s->block_start = 0L;
1038    s->lookahead = 0;
1039    s->match_length = s->prev_length = MIN_MATCH-1;
1040    s->match_available = 0;
1041    s->ins_h = 0;
1042#ifndef FASTEST
1043#ifdef ASMV
1044    match_init(); /* initialize the asm code */
1045#endif
1046#endif
1047}
1048
1049#ifndef FASTEST
1050/* ===========================================================================
1051 * Set match_start to the longest match starting at the given string and
1052 * return its length. Matches shorter or equal to prev_length are discarded,
1053 * in which case the result is equal to prev_length and match_start is
1054 * garbage.
1055 * IN assertions: cur_match is the head of the hash chain for the current
1056 *   string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
1057 * OUT assertion: the match length is not greater than s->lookahead.
1058 */
1059#ifndef ASMV
1060/* For 80x86 and 680x0, an optimized version will be provided in match.asm or
1061 * match.S. The code will be functionally equivalent.
1062 */
1063local uInt longest_match(s, cur_match)
1064    deflate_state *s;
1065    IPos cur_match;                             /* current match */
1066{
1067    unsigned chain_length = s->max_chain_length;/* max hash chain length */
1068    register Bytef *scan = s->window + s->strstart; /* current string */
1069    register Bytef *match;                       /* matched string */
1070    register int len;                           /* length of current match */
1071    int best_len = s->prev_length;              /* best match length so far */
1072    int nice_match = s->nice_match;             /* stop if match long enough */
1073    IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
1074        s->strstart - (IPos)MAX_DIST(s) : NIL;
1075    /* Stop when cur_match becomes <= limit. To simplify the code,
1076     * we prevent matches with the string of window index 0.
1077     */
1078    Posf *prev = s->prev;
1079    uInt wmask = s->w_mask;
1080
1081#ifdef UNALIGNED_OK
1082    /* Compare two bytes at a time. Note: this is not always beneficial.
1083     * Try with and without -DUNALIGNED_OK to check.
1084     */
1085    register Bytef *strend = s->window + s->strstart + MAX_MATCH - 1;
1086    register ush scan_start = *(ushf*)scan;
1087    register ush scan_end   = *(ushf*)(scan+best_len-1);
1088#else
1089    register Bytef *strend = s->window + s->strstart + MAX_MATCH;
1090    register Byte scan_end1  = scan[best_len-1];
1091    register Byte scan_end   = scan[best_len];
1092#endif
1093
1094    /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
1095     * It is easy to get rid of this optimization if necessary.
1096     */
1097    Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
1098
1099    /* Do not waste too much time if we already have a good match: */
1100    if (s->prev_length >= s->good_match) {
1101        chain_length >>= 2;
1102    }
1103    /* Do not look for matches beyond the end of the input. This is necessary
1104     * to make deflate deterministic.
1105     */
1106    if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead;
1107
1108    Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
1109
1110    do {
1111        Assert(cur_match < s->strstart, "no future");
1112        match = s->window + cur_match;
1113
1114        /* Skip to next match if the match length cannot increase
1115         * or if the match length is less than 2.  Note that the checks below
1116         * for insufficient lookahead only occur occasionally for performance
1117         * reasons.  Therefore uninitialized memory will be accessed, and
1118         * conditional jumps will be made that depend on those values.
1119         * However the length of the match is limited to the lookahead, so
1120         * the output of deflate is not affected by the uninitialized values.
1121         */
1122#if (defined(UNALIGNED_OK) && MAX_MATCH == 258)
1123        /* This code assumes sizeof(unsigned short) == 2. Do not use
1124         * UNALIGNED_OK if your compiler uses a different size.
1125         */
1126        if (*(ushf*)(match+best_len-1) != scan_end ||
1127            *(ushf*)match != scan_start) continue;
1128
1129        /* It is not necessary to compare scan[2] and match[2] since they are
1130         * always equal when the other bytes match, given that the hash keys
1131         * are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at
1132         * strstart+3, +5, ... up to strstart+257. We check for insufficient
1133         * lookahead only every 4th comparison; the 128th check will be made
1134         * at strstart+257. If MAX_MATCH-2 is not a multiple of 8, it is
1135         * necessary to put more guard bytes at the end of the window, or
1136         * to check more often for insufficient lookahead.
1137         */
1138        Assert(scan[2] == match[2], "scan[2]?");
1139        scan++, match++;
1140        do {
1141        } while (*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
1142                 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
1143                 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
1144                 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
1145                 scan < strend);
1146        /* The funny "do {}" generates better code on most compilers */
1147
1148        /* Here, scan <= window+strstart+257 */
1149        Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
1150        if (*scan == *match) scan++;
1151
1152        len = (MAX_MATCH - 1) - (int)(strend-scan);
1153        scan = strend - (MAX_MATCH-1);
1154
1155#else /* UNALIGNED_OK */
1156
1157        if (match[best_len]   != scan_end  ||
1158            match[best_len-1] != scan_end1 ||
1159            *match            != *scan     ||
1160            *++match          != scan[1])      continue;
1161
1162        /* The check at best_len-1 can be removed because it will be made
1163         * again later. (This heuristic is not always a win.)
1164         * It is not necessary to compare scan[2] and match[2] since they
1165         * are always equal when the other bytes match, given that
1166         * the hash keys are equal and that HASH_BITS >= 8.
1167         */
1168        scan += 2, match++;
1169        Assert(*scan == *match, "match[2]?");
1170
1171        /* We check for insufficient lookahead only every 8th comparison;
1172         * the 256th check will be made at strstart+258.
1173         */
1174        do {
1175        } while (*++scan == *++match && *++scan == *++match &&
1176                 *++scan == *++match && *++scan == *++match &&
1177                 *++scan == *++match && *++scan == *++match &&
1178                 *++scan == *++match && *++scan == *++match &&
1179                 scan < strend);
1180
1181        Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
1182
1183        len = MAX_MATCH - (int)(strend - scan);
1184        scan = strend - MAX_MATCH;
1185
1186#endif /* UNALIGNED_OK */
1187
1188        if (len > best_len) {
1189            s->match_start = cur_match;
1190            best_len = len;
1191            if (len >= nice_match) break;
1192#ifdef UNALIGNED_OK
1193            scan_end = *(ushf*)(scan+best_len-1);
1194#else
1195            scan_end1  = scan[best_len-1];
1196            scan_end   = scan[best_len];
1197#endif
1198        }
1199    } while ((cur_match = prev[cur_match & wmask]) > limit
1200             && --chain_length != 0);
1201
1202    if ((uInt)best_len <= s->lookahead) return (uInt)best_len;
1203    return s->lookahead;
1204}
1205#endif /* ASMV */
1206
1207#else /* FASTEST */
1208
1209/* ---------------------------------------------------------------------------
1210 * Optimized version for FASTEST only
1211 */
1212local uInt longest_match(s, cur_match)
1213    deflate_state *s;
1214    IPos cur_match;                             /* current match */
1215{
1216    register Bytef *scan = s->window + s->strstart; /* current string */
1217    register Bytef *match;                       /* matched string */
1218    register int len;                           /* length of current match */
1219    register Bytef *strend = s->window + s->strstart + MAX_MATCH;
1220
1221    /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
1222     * It is easy to get rid of this optimization if necessary.
1223     */
1224    Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
1225
1226    Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
1227
1228    Assert(cur_match < s->strstart, "no future");
1229
1230    match = s->window + cur_match;
1231
1232    /* Return failure if the match length is less than 2:
1233     */
1234    if (match[0] != scan[0] || match[1] != scan[1]) return MIN_MATCH-1;
1235
1236    /* The check at best_len-1 can be removed because it will be made
1237     * again later. (This heuristic is not always a win.)
1238     * It is not necessary to compare scan[2] and match[2] since they
1239     * are always equal when the other bytes match, given that
1240     * the hash keys are equal and that HASH_BITS >= 8.
1241     */
1242    scan += 2, match += 2;
1243    Assert(*scan == *match, "match[2]?");
1244
1245    /* We check for insufficient lookahead only every 8th comparison;
1246     * the 256th check will be made at strstart+258.
1247     */
1248    do {
1249    } while (*++scan == *++match && *++scan == *++match &&
1250             *++scan == *++match && *++scan == *++match &&
1251             *++scan == *++match && *++scan == *++match &&
1252             *++scan == *++match && *++scan == *++match &&
1253             scan < strend);
1254
1255    Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
1256
1257    len = MAX_MATCH - (int)(strend - scan);
1258
1259    if (len < MIN_MATCH) return MIN_MATCH - 1;
1260
1261    s->match_start = cur_match;
1262    return (uInt)len <= s->lookahead ? (uInt)len : s->lookahead;
1263}
1264
1265#endif /* FASTEST */
1266
1267#ifdef DEBUG
1268/* ===========================================================================
1269 * Check that the match at match_start is indeed a match.
1270 */
1271local void check_match(s, start, match, length)
1272    deflate_state *s;
1273    IPos start, match;
1274    int length;
1275{
1276    /* check that the match is indeed a match */
1277    if (zmemcmp(s->window + match,
1278                s->window + start, length) != EQUAL) {
1279        fprintf(stderr, " start %u, match %u, length %d\n",
1280                start, match, length);
1281        do {
1282            fprintf(stderr, "%c%c", s->window[match++], s->window[start++]);
1283        } while (--length != 0);
1284        z_error("invalid match");
1285    }
1286    if (z_verbose > 1) {
1287        fprintf(stderr,"\\[%d,%d]", start-match, length);
1288        do { putc(s->window[start++], stderr); } while (--length != 0);
1289    }
1290}
1291#else
1292#  define check_match(s, start, match, length)
1293#endif /* DEBUG */
1294
1295/* ===========================================================================
1296 * Fill the window when the lookahead becomes insufficient.
1297 * Updates strstart and lookahead.
1298 *
1299 * IN assertion: lookahead < MIN_LOOKAHEAD
1300 * OUT assertions: strstart <= window_size-MIN_LOOKAHEAD
1301 *    At least one byte has been read, or avail_in == 0; reads are
1302 *    performed for at least two bytes (required for the zip translate_eol
1303 *    option -- not supported here).
1304 */
1305local void fill_window(s)
1306    deflate_state *s;
1307{
1308    register unsigned n, m;
1309    register Posf *p;
1310    unsigned more;    /* Amount of free space at the end of the window. */
1311    uInt wsize = s->w_size;
1312
1313    do {
1314        more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart);
1315
1316        /* Deal with !@#$% 64K limit: */
1317        if (sizeof(int) <= 2) {
1318            if (more == 0 && s->strstart == 0 && s->lookahead == 0) {
1319                more = wsize;
1320
1321            } else if (more == (unsigned)(-1)) {
1322                /* Very unlikely, but possible on 16 bit machine if
1323                 * strstart == 0 && lookahead == 1 (input done a byte at time)
1324                 */
1325                more--;
1326            }
1327        }
1328
1329        /* If the window is almost full and there is insufficient lookahead,
1330         * move the upper half to the lower one to make room in the upper half.
1331         */
1332        if (s->strstart >= wsize+MAX_DIST(s)) {
1333
1334            zmemcpy(s->window, s->window+wsize, (unsigned)wsize);
1335            s->match_start -= wsize;
1336            s->strstart    -= wsize; /* we now have strstart >= MAX_DIST */
1337            s->block_start -= (long) wsize;
1338
1339            /* Slide the hash table (could be avoided with 32 bit values
1340               at the expense of memory usage). We slide even when level == 0
1341               to keep the hash table consistent if we switch back to level > 0
1342               later. (Using level 0 permanently is not an optimal usage of
1343               zlib, so we don't care about this pathological case.)
1344             */
1345            n = s->hash_size;
1346            p = &s->head[n];
1347            do {
1348                m = *--p;
1349                *p = (Pos)(m >= wsize ? m-wsize : NIL);
1350            } while (--n);
1351
1352            n = wsize;
1353#ifndef FASTEST
1354            p = &s->prev[n];
1355            do {
1356                m = *--p;
1357                *p = (Pos)(m >= wsize ? m-wsize : NIL);
1358                /* If n is not on any hash chain, prev[n] is garbage but
1359                 * its value will never be used.
1360                 */
1361            } while (--n);
1362#endif
1363            more += wsize;
1364        }
1365        if (s->strm->avail_in == 0) return;
1366
1367        /* If there was no sliding:
1368         *    strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 &&
1369         *    more == window_size - lookahead - strstart
1370         * => more >= window_size - (MIN_LOOKAHEAD-1 + WSIZE + MAX_DIST-1)
1371         * => more >= window_size - 2*WSIZE + 2
1372         * In the BIG_MEM or MMAP case (not yet supported),
1373         *   window_size == input_size + MIN_LOOKAHEAD  &&
1374         *   strstart + s->lookahead <= input_size => more >= MIN_LOOKAHEAD.
1375         * Otherwise, window_size == 2*WSIZE so more >= 2.
1376         * If there was sliding, more >= WSIZE. So in all cases, more >= 2.
1377         */
1378        Assert(more >= 2, "more < 2");
1379
1380        n = read_buf(s->strm, s->window + s->strstart + s->lookahead, more);
1381        s->lookahead += n;
1382
1383        /* Initialize the hash value now that we have some input: */
1384        if (s->lookahead >= MIN_MATCH) {
1385            s->ins_h = s->window[s->strstart];
1386            UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
1387#if MIN_MATCH != 3
1388            Call UPDATE_HASH() MIN_MATCH-3 more times
1389#endif
1390        }
1391        /* If the whole input has less than MIN_MATCH bytes, ins_h is garbage,
1392         * but this is not important since only literal bytes will be emitted.
1393         */
1394
1395    } while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0);
1396
1397    /* If the WIN_INIT bytes after the end of the current data have never been
1398     * written, then zero those bytes in order to avoid memory check reports of
1399     * the use of uninitialized (or uninitialised as Julian writes) bytes by
1400     * the longest match routines.  Update the high water mark for the next
1401     * time through here.  WIN_INIT is set to MAX_MATCH since the longest match
1402     * routines allow scanning to strstart + MAX_MATCH, ignoring lookahead.
1403     */
1404    if (s->high_water < s->window_size) {
1405        ulg curr = s->strstart + (ulg)(s->lookahead);
1406        ulg init;
1407
1408        if (s->high_water < curr) {
1409            /* Previous high water mark below current data -- zero WIN_INIT
1410             * bytes or up to end of window, whichever is less.
1411             */
1412            init = s->window_size - curr;
1413            if (init > WIN_INIT)
1414                init = WIN_INIT;
1415            zmemzero(s->window + curr, (unsigned)init);
1416            s->high_water = curr + init;
1417        }
1418        else if (s->high_water < (ulg)curr + WIN_INIT) {
1419            /* High water mark at or above current data, but below current data
1420             * plus WIN_INIT -- zero out to current data plus WIN_INIT, or up
1421             * to end of window, whichever is less.
1422             */
1423            init = (ulg)curr + WIN_INIT - s->high_water;
1424            if (init > s->window_size - s->high_water)
1425                init = s->window_size - s->high_water;
1426            zmemzero(s->window + s->high_water, (unsigned)init);
1427            s->high_water += init;
1428        }
1429    }
1430}
1431
1432/* ===========================================================================
1433 * Flush the current block, with given end-of-file flag.
1434 * IN assertion: strstart is set to the end of the current match.
1435 */
1436#define FLUSH_BLOCK_ONLY(s, last) { \
1437   _tr_flush_block(s, (s->block_start >= 0L ? \
1438                   (charf *)&s->window[(unsigned)s->block_start] : \
1439                   (charf *)Z_NULL), \
1440                (ulg)((long)s->strstart - s->block_start), \
1441                (last)); \
1442   s->block_start = s->strstart; \
1443   flush_pending(s->strm); \
1444   Tracev((stderr,"[FLUSH]")); \
1445}
1446
1447/* Same but force premature exit if necessary. */
1448#define FLUSH_BLOCK(s, last) { \
1449   FLUSH_BLOCK_ONLY(s, last); \
1450   if (s->strm->avail_out == 0) return (last) ? finish_started : need_more; \
1451}
1452
1453/* ===========================================================================
1454 * Copy without compression as much as possible from the input stream, return
1455 * the current block state.
1456 * This function does not insert new strings in the dictionary since
1457 * uncompressible data is probably not useful. This function is used
1458 * only for the level=0 compression option.
1459 * NOTE: this function should be optimized to avoid extra copying from
1460 * window to pending_buf.
1461 */
1462local block_state deflate_stored(s, flush)
1463    deflate_state *s;
1464    int flush;
1465{
1466    /* Stored blocks are limited to 0xffff bytes, pending_buf is limited
1467     * to pending_buf_size, and each stored block has a 5 byte header:
1468     */
1469    ulg max_block_size = 0xffff;
1470    ulg max_start;
1471
1472    if (max_block_size > s->pending_buf_size - 5) {
1473        max_block_size = s->pending_buf_size - 5;
1474    }
1475
1476    /* Copy as much as possible from input to output: */
1477    for (;;) {
1478        /* Fill the window as much as possible: */
1479        if (s->lookahead <= 1) {
1480
1481            Assert(s->strstart < s->w_size+MAX_DIST(s) ||
1482                   s->block_start >= (long)s->w_size, "slide too late");
1483
1484            fill_window(s);
1485            if (s->lookahead == 0 && flush == Z_NO_FLUSH) return need_more;
1486
1487            if (s->lookahead == 0) break; /* flush the current block */
1488        }
1489        Assert(s->block_start >= 0L, "block gone");
1490
1491        s->strstart += s->lookahead;
1492        s->lookahead = 0;
1493
1494        /* Emit a stored block if pending_buf will be full: */
1495        max_start = s->block_start + max_block_size;
1496        if (s->strstart == 0 || (ulg)s->strstart >= max_start) {
1497            /* strstart == 0 is possible when wraparound on 16-bit machine */
1498            s->lookahead = (uInt)(s->strstart - max_start);
1499            s->strstart = (uInt)max_start;
1500            FLUSH_BLOCK(s, 0);
1501        }
1502        /* Flush if we may have to slide, otherwise block_start may become
1503         * negative and the data will be gone:
1504         */
1505        if (s->strstart - (uInt)s->block_start >= MAX_DIST(s)) {
1506            FLUSH_BLOCK(s, 0);
1507        }
1508    }
1509    FLUSH_BLOCK(s, flush == Z_FINISH);
1510    return flush == Z_FINISH ? finish_done : block_done;
1511}
1512
1513/* ===========================================================================
1514 * Compress as much as possible from the input stream, return the current
1515 * block state.
1516 * This function does not perform lazy evaluation of matches and inserts
1517 * new strings in the dictionary only for unmatched strings or for short
1518 * matches. It is used only for the fast compression options.
1519 */
1520local block_state deflate_fast(s, flush)
1521    deflate_state *s;
1522    int flush;
1523{
1524    IPos hash_head;       /* head of the hash chain */
1525    int bflush;           /* set if current block must be flushed */
1526
1527    for (;;) {
1528        /* Make sure that we always have enough lookahead, except
1529         * at the end of the input file. We need MAX_MATCH bytes
1530         * for the next match, plus MIN_MATCH bytes to insert the
1531         * string following the next match.
1532         */
1533        if (s->lookahead < MIN_LOOKAHEAD) {
1534            fill_window(s);
1535            if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
1536                return need_more;
1537            }
1538            if (s->lookahead == 0) break; /* flush the current block */
1539        }
1540
1541        /* Insert the string window[strstart .. strstart+2] in the
1542         * dictionary, and set hash_head to the head of the hash chain:
1543         */
1544        hash_head = NIL;
1545        if (s->lookahead >= MIN_MATCH) {
1546            INSERT_STRING(s, s->strstart, hash_head);
1547        }
1548
1549        /* Find the longest match, discarding those <= prev_length.
1550         * At this point we have always match_length < MIN_MATCH
1551         */
1552        if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) {
1553            /* To simplify the code, we prevent matches with the string
1554             * of window index 0 (in particular we have to avoid a match
1555             * of the string with itself at the start of the input file).
1556             */
1557            s->match_length = longest_match (s, hash_head);
1558            /* longest_match() sets match_start */
1559        }
1560        if (s->match_length >= MIN_MATCH) {
1561            check_match(s, s->strstart, s->match_start, s->match_length);
1562
1563            _tr_tally_dist(s, s->strstart - s->match_start,
1564                           s->match_length - MIN_MATCH, bflush);
1565
1566            s->lookahead -= s->match_length;
1567
1568            /* Insert new strings in the hash table only if the match length
1569             * is not too large. This saves time but degrades compression.
1570             */
1571#ifndef FASTEST
1572            if (s->match_length <= s->max_insert_length &&
1573                s->lookahead >= MIN_MATCH) {
1574                s->match_length--; /* string at strstart already in table */
1575                do {
1576                    s->strstart++;
1577                    INSERT_STRING(s, s->strstart, hash_head);
1578                    /* strstart never exceeds WSIZE-MAX_MATCH, so there are
1579                     * always MIN_MATCH bytes ahead.
1580                     */
1581                } while (--s->match_length != 0);
1582                s->strstart++;
1583            } else
1584#endif
1585            {
1586                s->strstart += s->match_length;
1587                s->match_length = 0;
1588                s->ins_h = s->window[s->strstart];
1589                UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
1590#if MIN_MATCH != 3
1591                Call UPDATE_HASH() MIN_MATCH-3 more times
1592#endif
1593                /* If lookahead < MIN_MATCH, ins_h is garbage, but it does not
1594                 * matter since it will be recomputed at next deflate call.
1595                 */
1596            }
1597        } else {
1598            /* No match, output a literal byte */
1599            Tracevv((stderr,"%c", s->window[s->strstart]));
1600            _tr_tally_lit (s, s->window[s->strstart], bflush);
1601            s->lookahead--;
1602            s->strstart++;
1603        }
1604        if (bflush) FLUSH_BLOCK(s, 0);
1605    }
1606    FLUSH_BLOCK(s, flush == Z_FINISH);
1607    return flush == Z_FINISH ? finish_done : block_done;
1608}
1609
1610#ifndef FASTEST
1611/* ===========================================================================
1612 * Same as above, but achieves better compression. We use a lazy
1613 * evaluation for matches: a match is finally adopted only if there is
1614 * no better match at the next window position.
1615 */
1616local block_state deflate_slow(s, flush)
1617    deflate_state *s;
1618    int flush;
1619{
1620    IPos hash_head;          /* head of hash chain */
1621    int bflush;              /* set if current block must be flushed */
1622
1623    /* Process the input block. */
1624    for (;;) {
1625        /* Make sure that we always have enough lookahead, except
1626         * at the end of the input file. We need MAX_MATCH bytes
1627         * for the next match, plus MIN_MATCH bytes to insert the
1628         * string following the next match.
1629         */
1630        if (s->lookahead < MIN_LOOKAHEAD) {
1631            fill_window(s);
1632            if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
1633                return need_more;
1634            }
1635            if (s->lookahead == 0) break; /* flush the current block */
1636        }
1637
1638        /* Insert the string window[strstart .. strstart+2] in the
1639         * dictionary, and set hash_head to the head of the hash chain:
1640         */
1641        hash_head = NIL;
1642        if (s->lookahead >= MIN_MATCH) {
1643            INSERT_STRING(s, s->strstart, hash_head);
1644        }
1645
1646        /* Find the longest match, discarding those <= prev_length.
1647         */
1648        s->prev_length = s->match_length, s->prev_match = s->match_start;
1649        s->match_length = MIN_MATCH-1;
1650
1651        if (hash_head != NIL && s->prev_length < s->max_lazy_match &&
1652            s->strstart - hash_head <= MAX_DIST(s)) {
1653            /* To simplify the code, we prevent matches with the string
1654             * of window index 0 (in particular we have to avoid a match
1655             * of the string with itself at the start of the input file).
1656             */
1657            s->match_length = longest_match (s, hash_head);
1658            /* longest_match() sets match_start */
1659
1660            if (s->match_length <= 5 && (s->strategy == Z_FILTERED
1661#if TOO_FAR <= 32767
1662                || (s->match_length == MIN_MATCH &&
1663                    s->strstart - s->match_start > TOO_FAR)
1664#endif
1665                )) {
1666
1667                /* If prev_match is also MIN_MATCH, match_start is garbage
1668                 * but we will ignore the current match anyway.
1669                 */
1670                s->match_length = MIN_MATCH-1;
1671            }
1672        }
1673        /* If there was a match at the previous step and the current
1674         * match is not better, output the previous match:
1675         */
1676        if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) {
1677            uInt max_insert = s->strstart + s->lookahead - MIN_MATCH;
1678            /* Do not insert strings in hash table beyond this. */
1679
1680            check_match(s, s->strstart-1, s->prev_match, s->prev_length);
1681
1682            _tr_tally_dist(s, s->strstart -1 - s->prev_match,
1683                           s->prev_length - MIN_MATCH, bflush);
1684
1685            /* Insert in hash table all strings up to the end of the match.
1686             * strstart-1 and strstart are already inserted. If there is not
1687             * enough lookahead, the last two strings are not inserted in
1688             * the hash table.
1689             */
1690            s->lookahead -= s->prev_length-1;
1691            s->prev_length -= 2;
1692            do {
1693                if (++s->strstart <= max_insert) {
1694                    INSERT_STRING(s, s->strstart, hash_head);
1695                }
1696            } while (--s->prev_length != 0);
1697            s->match_available = 0;
1698            s->match_length = MIN_MATCH-1;
1699            s->strstart++;
1700
1701            if (bflush) FLUSH_BLOCK(s, 0);
1702
1703        } else if (s->match_available) {
1704            /* If there was no match at the previous position, output a
1705             * single literal. If there was a match but the current match
1706             * is longer, truncate the previous match to a single literal.
1707             */
1708            Tracevv((stderr,"%c", s->window[s->strstart-1]));
1709            _tr_tally_lit(s, s->window[s->strstart-1], bflush);
1710            if (bflush) {
1711                FLUSH_BLOCK_ONLY(s, 0);
1712            }
1713            s->strstart++;
1714            s->lookahead--;
1715            if (s->strm->avail_out == 0) return need_more;
1716        } else {
1717            /* There is no previous match to compare with, wait for
1718             * the next step to decide.
1719             */
1720            s->match_available = 1;
1721            s->strstart++;
1722            s->lookahead--;
1723        }
1724    }
1725    Assert (flush != Z_NO_FLUSH, "no flush?");
1726    if (s->match_available) {
1727        Tracevv((stderr,"%c", s->window[s->strstart-1]));
1728        _tr_tally_lit(s, s->window[s->strstart-1], bflush);
1729        s->match_available = 0;
1730    }
1731    FLUSH_BLOCK(s, flush == Z_FINISH);
1732    return flush == Z_FINISH ? finish_done : block_done;
1733}
1734#endif /* FASTEST */
1735
1736/* ===========================================================================
1737 * For Z_RLE, simply look for runs of bytes, generate matches only of distance
1738 * one.  Do not maintain a hash table.  (It will be regenerated if this run of
1739 * deflate switches away from Z_RLE.)
1740 */
1741local block_state deflate_rle(s, flush)
1742    deflate_state *s;
1743    int flush;
1744{
1745    int bflush;             /* set if current block must be flushed */
1746    uInt prev;              /* byte at distance one to match */
1747    Bytef *scan, *strend;   /* scan goes up to strend for length of run */
1748
1749    for (;;) {
1750        /* Make sure that we always have enough lookahead, except
1751         * at the end of the input file. We need MAX_MATCH bytes
1752         * for the longest encodable run.
1753         */
1754        if (s->lookahead < MAX_MATCH) {
1755            fill_window(s);
1756            if (s->lookahead < MAX_MATCH && flush == Z_NO_FLUSH) {
1757                return need_more;
1758            }
1759            if (s->lookahead == 0) break; /* flush the current block */
1760        }
1761
1762        /* See how many times the previous byte repeats */
1763        s->match_length = 0;
1764        if (s->lookahead >= MIN_MATCH && s->strstart > 0) {
1765            scan = s->window + s->strstart - 1;
1766            prev = *scan;
1767            if (prev == *++scan && prev == *++scan && prev == *++scan) {
1768                strend = s->window + s->strstart + MAX_MATCH;
1769                do {
1770                } while (prev == *++scan && prev == *++scan &&
1771                         prev == *++scan && prev == *++scan &&
1772                         prev == *++scan && prev == *++scan &&
1773                         prev == *++scan && prev == *++scan &&
1774                         scan < strend);
1775                s->match_length = MAX_MATCH - (int)(strend - scan);
1776                if (s->match_length > s->lookahead)
1777                    s->match_length = s->lookahead;
1778            }
1779        }
1780
1781        /* Emit match if have run of MIN_MATCH or longer, else emit literal */
1782        if (s->match_length >= MIN_MATCH) {
1783            check_match(s, s->strstart, s->strstart - 1, s->match_length);
1784
1785            _tr_tally_dist(s, 1, s->match_length - MIN_MATCH, bflush);
1786
1787            s->lookahead -= s->match_length;
1788            s->strstart += s->match_length;
1789            s->match_length = 0;
1790        } else {
1791            /* No match, output a literal byte */
1792            Tracevv((stderr,"%c", s->window[s->strstart]));
1793            _tr_tally_lit (s, s->window[s->strstart], bflush);
1794            s->lookahead--;
1795            s->strstart++;
1796        }
1797        if (bflush) FLUSH_BLOCK(s, 0);
1798    }
1799    FLUSH_BLOCK(s, flush == Z_FINISH);
1800    return flush == Z_FINISH ? finish_done : block_done;
1801}
1802
1803/* ===========================================================================
1804 * For Z_HUFFMAN_ONLY, do not look for matches.  Do not maintain a hash table.
1805 * (It will be regenerated if this run of deflate switches away from Huffman.)
1806 */
1807local block_state deflate_huff(s, flush)
1808    deflate_state *s;
1809    int flush;
1810{
1811    int bflush;             /* set if current block must be flushed */
1812
1813    for (;;) {
1814        /* Make sure that we have a literal to write. */
1815        if (s->lookahead == 0) {
1816            fill_window(s);
1817            if (s->lookahead == 0) {
1818                if (flush == Z_NO_FLUSH)
1819                    return need_more;
1820                break;      /* flush the current block */
1821            }
1822        }
1823
1824        /* Output a literal byte */
1825        s->match_length = 0;
1826        Tracevv((stderr,"%c", s->window[s->strstart]));
1827        _tr_tally_lit (s, s->window[s->strstart], bflush);
1828        s->lookahead--;
1829        s->strstart++;
1830        if (bflush) FLUSH_BLOCK(s, 0);
1831    }
1832    FLUSH_BLOCK(s, flush == Z_FINISH);
1833    return flush == Z_FINISH ? finish_done : block_done;
1834}
1835