deflate.c
Go to the documentation of this file.
00001 /* deflate.c -- compress data using the deflation algorithm
00002  * Copyright (C) 1995-2005 Jean-loup Gailly.
00003  * For conditions of distribution and use, see copyright notice in zlib.h
00004  */
00005 
00006 /*
00007  *  ALGORITHM
00008  *
00009  *      The "deflation" process depends on being able to identify portions
00010  *      of the input text which are identical to earlier input (within a
00011  *      sliding window trailing behind the input currently being processed).
00012  *
00013  *      The most straightforward technique turns out to be the fastest for
00014  *      most input files: try all possible matches and select the longest.
00015  *      The key feature of this algorithm is that insertions into the string
00016  *      dictionary are very simple and thus fast, and deletions are avoided
00017  *      completely. Insertions are performed at each input character, whereas
00018  *      string matches are performed only when the previous match ends. So it
00019  *      is preferable to spend more time in matches to allow very fast string
00020  *      insertions and avoid deletions. The matching algorithm for small
00021  *      strings is inspired from that of Rabin & Karp. A brute force approach
00022  *      is used to find longer strings when a small match has been found.
00023  *      A similar algorithm is used in comic (by Jan-Mark Wams) and freeze
00024  *      (by Leonid Broukhis).
00025  *         A previous version of this file used a more sophisticated algorithm
00026  *      (by Fiala and Greene) which is guaranteed to run in linear amortized
00027  *      time, but has a larger average cost, uses more memory and is patented.
00028  *      However the F&G algorithm may be faster for some highly redundant
00029  *      files if the parameter max_chain_length (described below) is too large.
00030  *
00031  *  ACKNOWLEDGEMENTS
00032  *
00033  *      The idea of lazy evaluation of matches is due to Jan-Mark Wams, and
00034  *      I found it in 'freeze' written by Leonid Broukhis.
00035  *      Thanks to many people for bug reports and testing.
00036  *
00037  *  REFERENCES
00038  *
00039  *      Deutsch, L.P.,"DEFLATE Compressed Data Format Specification".
00040  *      Available in http://www.ietf.org/rfc/rfc1951.txt
00041  *
00042  *      A description of the Rabin and Karp algorithm is given in the book
00043  *         "Algorithms" by R. Sedgewick, Addison-Wesley, p252.
00044  *
00045  *      Fiala,E.R., and Greene,D.H.
00046  *         Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595
00047  *
00048  */
00049 
00050 /* @(#) $Id$ */
00051 
00052 #include "deflate.h"
00053 
00054 const char deflate_copyright[] =
00055    " deflate 1.2.3 Copyright 1995-2005 Jean-loup Gailly ";
00056 /*
00057   If you use the zlib library in a product, an acknowledgment is welcome
00058   in the documentation of your product. If for some reason you cannot
00059   include such an acknowledgment, I would appreciate that you keep this
00060   copyright string in the executable of your product.
00061  */
00062 
00063 /* ===========================================================================
00064  *  Function prototypes.
00065  */
00066 typedef enum {
00067     need_more,      /* block not completed, need more input or more output */
00068     block_done,     /* block flush performed */
00069     finish_started, /* finish started, need only more output at next deflate */
00070     finish_done     /* finish done, accept no more input or output */
00071 } block_state;
00072 
00073 typedef block_state (*compress_func) OF((deflate_state *s, int flush));
00074 /* Compression function. Returns the block state after the call. */
00075 
00076 local void fill_window    OF((deflate_state *s));
00077 local block_state deflate_stored OF((deflate_state *s, int flush));
00078 local block_state deflate_fast   OF((deflate_state *s, int flush));
00079 #ifndef FASTEST
00080 local block_state deflate_slow   OF((deflate_state *s, int flush));
00081 #endif
00082 local void lm_init        OF((deflate_state *s));
00083 local void putShortMSB    OF((deflate_state *s, uInt b));
00084 local void flush_pending  OF((z_streamp strm));
00085 local int read_buf        OF((z_streamp strm, Bytef *buf, unsigned size));
00086 #ifndef FASTEST
00087 #ifdef ASMV
00088       void match_init OF((void)); /* asm code initialization */
00089       uInt longest_match  OF((deflate_state *s, IPos cur_match));
00090 #else
00091 local uInt longest_match  OF((deflate_state *s, IPos cur_match));
00092 #endif
00093 #endif
00094 local uInt longest_match_fast OF((deflate_state *s, IPos cur_match));
00095 
00096 #ifdef DEBUG
00097 local  void check_match OF((deflate_state *s, IPos start, IPos match,
00098                             int length));
00099 #endif
00100 
00101 /* ===========================================================================
00102  * Local data
00103  */
00104 
00105 #define NIL 0
00106 /* Tail of hash chains */
00107 
00108 #ifndef TOO_FAR
00109 #  define TOO_FAR 4096
00110 #endif
00111 /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
00112 
00113 #define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
00114 /* Minimum amount of lookahead, except at the end of the input file.
00115  * See deflate.c for comments about the MIN_MATCH+1.
00116  */
00117 
00118 /* Values for max_lazy_match, good_match and max_chain_length, depending on
00119  * the desired pack level (0..9). The values given below have been tuned to
00120  * exclude worst case performance for pathological files. Better values may be
00121  * found for specific files.
00122  */
00123 typedef struct config_s {
00124    ush good_length; /* reduce lazy search above this match length */
00125    ush max_lazy;    /* do not perform lazy search above this match length */
00126    ush nice_length; /* quit search above this match length */
00127    ush max_chain;
00128    compress_func func;
00129 } config;
00130 
00131 #ifdef FASTEST
00132 local const config configuration_table[2] = {
00133 /*      good lazy nice chain */
00134 /* 0 */ {0,    0,  0,    0, deflate_stored},  /* store only */
00135 /* 1 */ {4,    4,  8,    4, deflate_fast}}; /* max speed, no lazy matches */
00136 #else
00137 local const config configuration_table[10] = {
00138 /*      good lazy nice chain */
00139 /* 0 */ {0,    0,  0,    0, deflate_stored},  /* store only */
00140 /* 1 */ {4,    4,  8,    4, deflate_fast}, /* max speed, no lazy matches */
00141 /* 2 */ {4,    5, 16,    8, deflate_fast},
00142 /* 3 */ {4,    6, 32,   32, deflate_fast},
00143 
00144 /* 4 */ {4,    4, 16,   16, deflate_slow},  /* lazy matches */
00145 /* 5 */ {8,   16, 32,   32, deflate_slow},
00146 /* 6 */ {8,   16, 128, 128, deflate_slow},
00147 /* 7 */ {8,   32, 128, 256, deflate_slow},
00148 /* 8 */ {32, 128, 258, 1024, deflate_slow},
00149 /* 9 */ {32, 258, 258, 4096, deflate_slow}}; /* max compression */
00150 #endif
00151 
00152 /* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
00153  * For deflate_fast() (levels <= 3) good is ignored and lazy has a different
00154  * meaning.
00155  */
00156 
00157 #define EQUAL 0
00158 /* result of memcmp for equal strings */
00159 
00160 #ifndef NO_DUMMY_DECL
00161 struct static_tree_desc_s {int dummy;}; /* for buggy compilers */
00162 #endif
00163 
00164 /* ===========================================================================
00165  * Update a hash value with the given input byte
00166  * IN  assertion: all calls to to UPDATE_HASH are made with consecutive
00167  *    input characters, so that a running hash key can be computed from the
00168  *    previous key instead of complete recalculation each time.
00169  */
00170 #define UPDATE_HASH(s,h,c) (h = (((h)<<s->hash_shift) ^ (c)) & s->hash_mask)
00171 
00172 
00173 /* ===========================================================================
00174  * Insert string str in the dictionary and set match_head to the previous head
00175  * of the hash chain (the most recent string with same hash key). Return
00176  * the previous length of the hash chain.
00177  * If this file is compiled with -DFASTEST, the compression level is forced
00178  * to 1, and no hash chains are maintained.
00179  * IN  assertion: all calls to to INSERT_STRING are made with consecutive
00180  *    input characters and the first MIN_MATCH bytes of str are valid
00181  *    (except for the last MIN_MATCH-1 bytes of the input file).
00182  */
00183 #ifdef FASTEST
00184 #define INSERT_STRING(s, str, match_head) \
00185    (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
00186     match_head = s->head[s->ins_h], \
00187     s->head[s->ins_h] = (Pos)(str))
00188 #else
00189 #define INSERT_STRING(s, str, match_head) \
00190    (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
00191     match_head = s->prev[(str) & s->w_mask] = s->head[s->ins_h], \
00192     s->head[s->ins_h] = (Pos)(str))
00193 #endif
00194 
00195 /* ===========================================================================
00196  * Initialize the hash table (avoiding 64K overflow for 16 bit systems).
00197  * prev[] will be initialized on the fly.
00198  */
00199 #define CLEAR_HASH(s) \
00200     s->head[s->hash_size-1] = NIL; \
00201     zmemzero((Bytef *)s->head, (unsigned)(s->hash_size-1)*sizeof(*s->head));
00202 
00203 /* ========================================================================= */
00204 int ZEXPORT deflateInit_(strm, level, version, stream_size)
00205     z_streamp strm;
00206     int level;
00207     const char *version;
00208     int stream_size;
00209 {
00210     return deflateInit2_(strm, level, Z_DEFLATED, MAX_WBITS, DEF_MEM_LEVEL,
00211                          Z_DEFAULT_STRATEGY, version, stream_size);
00212     /* To do: ignore strm->next_in if we use it as window */
00213 }
00214 
00215 /* ========================================================================= */
00216 int ZEXPORT deflateInit2_(strm, level, method, windowBits, memLevel, strategy,
00217                   version, stream_size)
00218     z_streamp strm;
00219     int  level;
00220     int  method;
00221     int  windowBits;
00222     int  memLevel;
00223     int  strategy;
00224     const char *version;
00225     int stream_size;
00226 {
00227     deflate_state *s;
00228     int wrap = 1;
00229     static const char my_version[] = ZLIB_VERSION;
00230 
00231     ushf *overlay;
00232     /* We overlay pending_buf and d_buf+l_buf. This works since the average
00233      * output size for (length,distance) codes is <= 24 bits.
00234      */
00235 
00236     if (version == Z_NULL || version[0] != my_version[0] ||
00237         stream_size != sizeof(z_stream)) {
00238         return Z_VERSION_ERROR;
00239     }
00240     if (strm == Z_NULL) return Z_STREAM_ERROR;
00241 
00242     strm->msg = Z_NULL;
00243     if (strm->zalloc == (alloc_func)0) {
00244         strm->zalloc = zcalloc;
00245         strm->opaque = (voidpf)0;
00246     }
00247     if (strm->zfree == (free_func)0) strm->zfree = zcfree;
00248 
00249 #ifdef FASTEST
00250     if (level != 0) level = 1;
00251 #else
00252     if (level == Z_DEFAULT_COMPRESSION) level = 6;
00253 #endif
00254 
00255     if (windowBits < 0) { /* suppress zlib wrapper */
00256         wrap = 0;
00257         windowBits = -windowBits;
00258     }
00259 #ifdef GZIP
00260     else if (windowBits > 15) {
00261         wrap = 2;       /* write gzip wrapper instead */
00262         windowBits -= 16;
00263     }
00264 #endif
00265     if (memLevel < 1 || memLevel > MAX_MEM_LEVEL || method != Z_DEFLATED ||
00266         windowBits < 8 || windowBits > 15 || level < 0 || level > 9 ||
00267         strategy < 0 || strategy > Z_FIXED) {
00268         return Z_STREAM_ERROR;
00269     }
00270     if (windowBits == 8) windowBits = 9;  /* until 256-byte window bug fixed */
00271     s = (deflate_state *) ZALLOC(strm, 1, sizeof(deflate_state));
00272     if (s == Z_NULL) return Z_MEM_ERROR;
00273     strm->state = (struct internal_state FAR *)s;
00274     s->strm = strm;
00275 
00276     s->wrap = wrap;
00277     s->gzhead = Z_NULL;
00278     s->w_bits = windowBits;
00279     s->w_size = 1 << s->w_bits;
00280     s->w_mask = s->w_size - 1;
00281 
00282     s->hash_bits = memLevel + 7;
00283     s->hash_size = 1 << s->hash_bits;
00284     s->hash_mask = s->hash_size - 1;
00285     s->hash_shift =  ((s->hash_bits+MIN_MATCH-1)/MIN_MATCH);
00286 
00287     s->window = (Bytef *) ZALLOC(strm, s->w_size, 2*sizeof(Byte));
00288     s->prev   = (Posf *)  ZALLOC(strm, s->w_size, sizeof(Pos));
00289     s->head   = (Posf *)  ZALLOC(strm, s->hash_size, sizeof(Pos));
00290 
00291     s->lit_bufsize = 1 << (memLevel + 6); /* 16K elements by default */
00292 
00293     overlay = (ushf *) ZALLOC(strm, s->lit_bufsize, sizeof(ush)+2);
00294     s->pending_buf = (uchf *) overlay;
00295     s->pending_buf_size = (ulg)s->lit_bufsize * (sizeof(ush)+2L);
00296 
00297     if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL ||
00298         s->pending_buf == Z_NULL) {
00299         s->status = FINISH_STATE;
00300         strm->msg = (char*)ERR_MSG(Z_MEM_ERROR);
00301         deflateEnd (strm);
00302         return Z_MEM_ERROR;
00303     }
00304     s->d_buf = overlay + s->lit_bufsize/sizeof(ush);
00305     s->l_buf = s->pending_buf + (1+sizeof(ush))*s->lit_bufsize;
00306 
00307     s->level = level;
00308     s->strategy = strategy;
00309     s->method = (Byte)method;
00310 
00311     return deflateReset(strm);
00312 }
00313 
00314 /* ========================================================================= */
00315 int ZEXPORT deflateSetDictionary (strm, dictionary, dictLength)
00316     z_streamp strm;
00317     const Bytef *dictionary;
00318     uInt  dictLength;
00319 {
00320     deflate_state *s;
00321     uInt length = dictLength;
00322     uInt n;
00323     IPos hash_head = 0;
00324 
00325     if (strm == Z_NULL || strm->state == Z_NULL || dictionary == Z_NULL ||
00326         strm->state->wrap == 2 ||
00327         (strm->state->wrap == 1 && strm->state->status != INIT_STATE))
00328         return Z_STREAM_ERROR;
00329 
00330     s = strm->state;
00331     if (s->wrap)
00332         strm->adler = adler32(strm->adler, dictionary, dictLength);
00333 
00334     if (length < MIN_MATCH) return Z_OK;
00335     if (length > MAX_DIST(s)) {
00336         length = MAX_DIST(s);
00337         dictionary += dictLength - length; /* use the tail of the dictionary */
00338     }
00339     zmemcpy(s->window, dictionary, length);
00340     s->strstart = length;
00341     s->block_start = (long)length;
00342 
00343     /* Insert all strings in the hash table (except for the last two bytes).
00344      * s->lookahead stays null, so s->ins_h will be recomputed at the next
00345      * call of fill_window.
00346      */
00347     s->ins_h = s->window[0];
00348     UPDATE_HASH(s, s->ins_h, s->window[1]);
00349     for (n = 0; n <= length - MIN_MATCH; n++) {
00350         INSERT_STRING(s, n, hash_head);
00351     }
00352     if (hash_head) hash_head = 0;  /* to make compiler happy */
00353     return Z_OK;
00354 }
00355 
00356 /* ========================================================================= */
00357 int ZEXPORT deflateReset (strm)
00358     z_streamp strm;
00359 {
00360     deflate_state *s;
00361 
00362     if (strm == Z_NULL || strm->state == Z_NULL ||
00363         strm->zalloc == (alloc_func)0 || strm->zfree == (free_func)0) {
00364         return Z_STREAM_ERROR;
00365     }
00366 
00367     strm->total_in = strm->total_out = 0;
00368     strm->msg = Z_NULL; /* use zfree if we ever allocate msg dynamically */
00369     strm->data_type = Z_UNKNOWN;
00370 
00371     s = (deflate_state *)strm->state;
00372     s->pending = 0;
00373     s->pending_out = s->pending_buf;
00374 
00375     if (s->wrap < 0) {
00376         s->wrap = -s->wrap; /* was made negative by deflate(..., Z_FINISH); */
00377     }
00378     s->status = s->wrap ? INIT_STATE : BUSY_STATE;
00379     strm->adler =
00380 #ifdef GZIP
00381         s->wrap == 2 ? crc32(0L, Z_NULL, 0) :
00382 #endif
00383         adler32(0L, Z_NULL, 0);
00384     s->last_flush = Z_NO_FLUSH;
00385 
00386     _tr_init(s);
00387     lm_init(s);
00388 
00389     return Z_OK;
00390 }
00391 
00392 /* ========================================================================= */
00393 int ZEXPORT deflateSetHeader (strm, head)
00394     z_streamp strm;
00395     gz_headerp head;
00396 {
00397     if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
00398     if (strm->state->wrap != 2) return Z_STREAM_ERROR;
00399     strm->state->gzhead = head;
00400     return Z_OK;
00401 }
00402 
00403 /* ========================================================================= */
00404 int ZEXPORT deflatePrime (strm, bits, value)
00405     z_streamp strm;
00406     int bits;
00407     int value;
00408 {
00409     if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
00410     strm->state->bi_valid = bits;
00411     strm->state->bi_buf = (ush)(value & ((1 << bits) - 1));
00412     return Z_OK;
00413 }
00414 
00415 /* ========================================================================= */
00416 int ZEXPORT deflateParams(strm, level, strategy)
00417     z_streamp strm;
00418     int level;
00419     int strategy;
00420 {
00421     deflate_state *s;
00422     compress_func func;
00423     int err = Z_OK;
00424 
00425     if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
00426     s = strm->state;
00427 
00428 #ifdef FASTEST
00429     if (level != 0) level = 1;
00430 #else
00431     if (level == Z_DEFAULT_COMPRESSION) level = 6;
00432 #endif
00433     if (level < 0 || level > 9 || strategy < 0 || strategy > Z_FIXED) {
00434         return Z_STREAM_ERROR;
00435     }
00436     func = configuration_table[s->level].func;
00437 
00438     if (func != configuration_table[level].func && strm->total_in != 0) {
00439         /* Flush the last buffer: */
00440         err = deflate(strm, Z_PARTIAL_FLUSH);
00441     }
00442     if (s->level != level) {
00443         s->level = level;
00444         s->max_lazy_match   = configuration_table[level].max_lazy;
00445         s->good_match       = configuration_table[level].good_length;
00446         s->nice_match       = configuration_table[level].nice_length;
00447         s->max_chain_length = configuration_table[level].max_chain;
00448     }
00449     s->strategy = strategy;
00450     return err;
00451 }
00452 
00453 /* ========================================================================= */
00454 int ZEXPORT deflateTune(strm, good_length, max_lazy, nice_length, max_chain)
00455     z_streamp strm;
00456     int good_length;
00457     int max_lazy;
00458     int nice_length;
00459     int max_chain;
00460 {
00461     deflate_state *s;
00462 
00463     if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
00464     s = strm->state;
00465     s->good_match = good_length;
00466     s->max_lazy_match = max_lazy;
00467     s->nice_match = nice_length;
00468     s->max_chain_length = max_chain;
00469     return Z_OK;
00470 }
00471 
00472 /* =========================================================================
00473  * For the default windowBits of 15 and memLevel of 8, this function returns
00474  * a close to exact, as well as small, upper bound on the compressed size.
00475  * They are coded as constants here for a reason--if the #define's are
00476  * changed, then this function needs to be changed as well.  The return
00477  * value for 15 and 8 only works for those exact settings.
00478  *
00479  * For any setting other than those defaults for windowBits and memLevel,
00480  * the value returned is a conservative worst case for the maximum expansion
00481  * resulting from using fixed blocks instead of stored blocks, which deflate
00482  * can emit on compressed data for some combinations of the parameters.
00483  *
00484  * This function could be more sophisticated to provide closer upper bounds
00485  * for every combination of windowBits and memLevel, as well as wrap.
00486  * But even the conservative upper bound of about 14% expansion does not
00487  * seem onerous for output buffer allocation.
00488  */
00489 uLong ZEXPORT deflateBound(strm, sourceLen)
00490     z_streamp strm;
00491     uLong sourceLen;
00492 {
00493     deflate_state *s;
00494     uLong destLen;
00495 
00496     /* conservative upper bound */
00497     destLen = sourceLen +
00498               ((sourceLen + 7) >> 3) + ((sourceLen + 63) >> 6) + 11;
00499 
00500     /* if can't get parameters, return conservative bound */
00501     if (strm == Z_NULL || strm->state == Z_NULL)
00502         return destLen;
00503 
00504     /* if not default parameters, return conservative bound */
00505     s = strm->state;
00506     if (s->w_bits != 15 || s->hash_bits != 8 + 7)
00507         return destLen;
00508 
00509     /* default settings: return tight bound for that case */
00510     return compressBound(sourceLen);
00511 }
00512 
00513 /* =========================================================================
00514  * Put a short in the pending buffer. The 16-bit value is put in MSB order.
00515  * IN assertion: the stream state is correct and there is enough room in
00516  * pending_buf.
00517  */
00518 local void putShortMSB (s, b)
00519     deflate_state *s;
00520     uInt b;
00521 {
00522     put_byte(s, (Byte)(b >> 8));
00523     put_byte(s, (Byte)(b & 0xff));
00524 }
00525 
00526 /* =========================================================================
00527  * Flush as much pending output as possible. All deflate() output goes
00528  * through this function so some applications may wish to modify it
00529  * to avoid allocating a large strm->next_out buffer and copying into it.
00530  * (See also read_buf()).
00531  */
00532 local void flush_pending(strm)
00533     z_streamp strm;
00534 {
00535     unsigned len = strm->state->pending;
00536 
00537     if (len > strm->avail_out) len = strm->avail_out;
00538     if (len == 0) return;
00539 
00540     zmemcpy(strm->next_out, strm->state->pending_out, len);
00541     strm->next_out  += len;
00542     strm->state->pending_out  += len;
00543     strm->total_out += len;
00544     strm->avail_out  -= len;
00545     strm->state->pending -= len;
00546     if (strm->state->pending == 0) {
00547         strm->state->pending_out = strm->state->pending_buf;
00548     }
00549 }
00550 
00551 /* ========================================================================= */
00552 int ZEXPORT deflate (strm, flush)
00553     z_streamp strm;
00554     int flush;
00555 {
00556     int old_flush; /* value of flush param for previous deflate call */
00557     deflate_state *s;
00558 
00559     if (strm == Z_NULL || strm->state == Z_NULL ||
00560         flush > Z_FINISH || flush < 0) {
00561         return Z_STREAM_ERROR;
00562     }
00563     s = strm->state;
00564 
00565     if (strm->next_out == Z_NULL ||
00566         (strm->next_in == Z_NULL && strm->avail_in != 0) ||
00567         (s->status == FINISH_STATE && flush != Z_FINISH)) {
00568         ERR_RETURN(strm, Z_STREAM_ERROR);
00569     }
00570     if (strm->avail_out == 0) ERR_RETURN(strm, Z_BUF_ERROR);
00571 
00572     s->strm = strm; /* just in case */
00573     old_flush = s->last_flush;
00574     s->last_flush = flush;
00575 
00576     /* Write the header */
00577     if (s->status == INIT_STATE) {
00578 #ifdef GZIP
00579         if (s->wrap == 2) {
00580             strm->adler = crc32(0L, Z_NULL, 0);
00581             put_byte(s, 31);
00582             put_byte(s, 139);
00583             put_byte(s, 8);
00584             if (s->gzhead == NULL) {
00585                 put_byte(s, 0);
00586                 put_byte(s, 0);
00587                 put_byte(s, 0);
00588                 put_byte(s, 0);
00589                 put_byte(s, 0);
00590                 put_byte(s, s->level == 9 ? 2 :
00591                             (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
00592                              4 : 0));
00593                 put_byte(s, OS_CODE);
00594                 s->status = BUSY_STATE;
00595             }
00596             else {
00597                 put_byte(s, (s->gzhead->text ? 1 : 0) +
00598                             (s->gzhead->hcrc ? 2 : 0) +
00599                             (s->gzhead->extra == Z_NULL ? 0 : 4) +
00600                             (s->gzhead->name == Z_NULL ? 0 : 8) +
00601                             (s->gzhead->comment == Z_NULL ? 0 : 16)
00602                         );
00603                 put_byte(s, (Byte)(s->gzhead->time & 0xff));
00604                 put_byte(s, (Byte)((s->gzhead->time >> 8) & 0xff));
00605                 put_byte(s, (Byte)((s->gzhead->time >> 16) & 0xff));
00606                 put_byte(s, (Byte)((s->gzhead->time >> 24) & 0xff));
00607                 put_byte(s, s->level == 9 ? 2 :
00608                             (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
00609                              4 : 0));
00610                 put_byte(s, s->gzhead->os & 0xff);
00611                 if (s->gzhead->extra != NULL) {
00612                     put_byte(s, s->gzhead->extra_len & 0xff);
00613                     put_byte(s, (s->gzhead->extra_len >> 8) & 0xff);
00614                 }
00615                 if (s->gzhead->hcrc)
00616                     strm->adler = crc32(strm->adler, s->pending_buf,
00617                                         s->pending);
00618                 s->gzindex = 0;
00619                 s->status = EXTRA_STATE;
00620             }
00621         }
00622         else
00623 #endif
00624         {
00625             uInt header = (Z_DEFLATED + ((s->w_bits-8)<<4)) << 8;
00626             uInt level_flags;
00627 
00628             if (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2)
00629                 level_flags = 0;
00630             else if (s->level < 6)
00631                 level_flags = 1;
00632             else if (s->level == 6)
00633                 level_flags = 2;
00634             else
00635                 level_flags = 3;
00636             header |= (level_flags << 6);
00637             if (s->strstart != 0) header |= PRESET_DICT;
00638             header += 31 - (header % 31);
00639 
00640             s->status = BUSY_STATE;
00641             putShortMSB(s, header);
00642 
00643             /* Save the adler32 of the preset dictionary: */
00644             if (s->strstart != 0) {
00645                 putShortMSB(s, (uInt)(strm->adler >> 16));
00646                 putShortMSB(s, (uInt)(strm->adler & 0xffff));
00647             }
00648             strm->adler = adler32(0L, Z_NULL, 0);
00649         }
00650     }
00651 #ifdef GZIP
00652     if (s->status == EXTRA_STATE) {
00653         if (s->gzhead->extra != NULL) {
00654             uInt beg = s->pending;  /* start of bytes to update crc */
00655 
00656             while (s->gzindex < (s->gzhead->extra_len & 0xffff)) {
00657                 if (s->pending == s->pending_buf_size) {
00658                     if (s->gzhead->hcrc && s->pending > beg)
00659                         strm->adler = crc32(strm->adler, s->pending_buf + beg,
00660                                             s->pending - beg);
00661                     flush_pending(strm);
00662                     beg = s->pending;
00663                     if (s->pending == s->pending_buf_size)
00664                         break;
00665                 }
00666                 put_byte(s, s->gzhead->extra[s->gzindex]);
00667                 s->gzindex++;
00668             }
00669             if (s->gzhead->hcrc && s->pending > beg)
00670                 strm->adler = crc32(strm->adler, s->pending_buf + beg,
00671                                     s->pending - beg);
00672             if (s->gzindex == s->gzhead->extra_len) {
00673                 s->gzindex = 0;
00674                 s->status = NAME_STATE;
00675             }
00676         }
00677         else
00678             s->status = NAME_STATE;
00679     }
00680     if (s->status == NAME_STATE) {
00681         if (s->gzhead->name != NULL) {
00682             uInt beg = s->pending;  /* start of bytes to update crc */
00683             int val;
00684 
00685             do {
00686                 if (s->pending == s->pending_buf_size) {
00687                     if (s->gzhead->hcrc && s->pending > beg)
00688                         strm->adler = crc32(strm->adler, s->pending_buf + beg,
00689                                             s->pending - beg);
00690                     flush_pending(strm);
00691                     beg = s->pending;
00692                     if (s->pending == s->pending_buf_size) {
00693                         val = 1;
00694                         break;
00695                     }
00696                 }
00697                 val = s->gzhead->name[s->gzindex++];
00698                 put_byte(s, val);
00699             } while (val != 0);
00700             if (s->gzhead->hcrc && s->pending > beg)
00701                 strm->adler = crc32(strm->adler, s->pending_buf + beg,
00702                                     s->pending - beg);
00703             if (val == 0) {
00704                 s->gzindex = 0;
00705                 s->status = COMMENT_STATE;
00706             }
00707         }
00708         else
00709             s->status = COMMENT_STATE;
00710     }
00711     if (s->status == COMMENT_STATE) {
00712         if (s->gzhead->comment != NULL) {
00713             uInt beg = s->pending;  /* start of bytes to update crc */
00714             int val;
00715 
00716             do {
00717                 if (s->pending == s->pending_buf_size) {
00718                     if (s->gzhead->hcrc && s->pending > beg)
00719                         strm->adler = crc32(strm->adler, s->pending_buf + beg,
00720                                             s->pending - beg);
00721                     flush_pending(strm);
00722                     beg = s->pending;
00723                     if (s->pending == s->pending_buf_size) {
00724                         val = 1;
00725                         break;
00726                     }
00727                 }
00728                 val = s->gzhead->comment[s->gzindex++];
00729                 put_byte(s, val);
00730             } while (val != 0);
00731             if (s->gzhead->hcrc && s->pending > beg)
00732                 strm->adler = crc32(strm->adler, s->pending_buf + beg,
00733                                     s->pending - beg);
00734             if (val == 0)
00735                 s->status = HCRC_STATE;
00736         }
00737         else
00738             s->status = HCRC_STATE;
00739     }
00740     if (s->status == HCRC_STATE) {
00741         if (s->gzhead->hcrc) {
00742             if (s->pending + 2 > s->pending_buf_size)
00743                 flush_pending(strm);
00744             if (s->pending + 2 <= s->pending_buf_size) {
00745                 put_byte(s, (Byte)(strm->adler & 0xff));
00746                 put_byte(s, (Byte)((strm->adler >> 8) & 0xff));
00747                 strm->adler = crc32(0L, Z_NULL, 0);
00748                 s->status = BUSY_STATE;
00749             }
00750         }
00751         else
00752             s->status = BUSY_STATE;
00753     }
00754 #endif
00755 
00756     /* Flush as much pending output as possible */
00757     if (s->pending != 0) {
00758         flush_pending(strm);
00759         if (strm->avail_out == 0) {
00760             /* Since avail_out is 0, deflate will be called again with
00761              * more output space, but possibly with both pending and
00762              * avail_in equal to zero. There won't be anything to do,
00763              * but this is not an error situation so make sure we
00764              * return OK instead of BUF_ERROR at next call of deflate:
00765              */
00766             s->last_flush = -1;
00767             return Z_OK;
00768         }
00769 
00770     /* Make sure there is something to do and avoid duplicate consecutive
00771      * flushes. For repeated and useless calls with Z_FINISH, we keep
00772      * returning Z_STREAM_END instead of Z_BUF_ERROR.
00773      */
00774     } else if (strm->avail_in == 0 && flush <= old_flush &&
00775                flush != Z_FINISH) {
00776         ERR_RETURN(strm, Z_BUF_ERROR);
00777     }
00778 
00779     /* User must not provide more input after the first FINISH: */
00780     if (s->status == FINISH_STATE && strm->avail_in != 0) {
00781         ERR_RETURN(strm, Z_BUF_ERROR);
00782     }
00783 
00784     /* Start a new block or continue the current one.
00785      */
00786     if (strm->avail_in != 0 || s->lookahead != 0 ||
00787         (flush != Z_NO_FLUSH && s->status != FINISH_STATE)) {
00788         block_state bstate;
00789 
00790         bstate = (*(configuration_table[s->level].func))(s, flush);
00791 
00792         if (bstate == finish_started || bstate == finish_done) {
00793             s->status = FINISH_STATE;
00794         }
00795         if (bstate == need_more || bstate == finish_started) {
00796             if (strm->avail_out == 0) {
00797                 s->last_flush = -1; /* avoid BUF_ERROR next call, see above */
00798             }
00799             return Z_OK;
00800             /* If flush != Z_NO_FLUSH && avail_out == 0, the next call
00801              * of deflate should use the same flush parameter to make sure
00802              * that the flush is complete. So we don't have to output an
00803              * empty block here, this will be done at next call. This also
00804              * ensures that for a very small output buffer, we emit at most
00805              * one empty block.
00806              */
00807         }
00808         if (bstate == block_done) {
00809             if (flush == Z_PARTIAL_FLUSH) {
00810                 _tr_align(s);
00811             } else { /* FULL_FLUSH or SYNC_FLUSH */
00812                 _tr_stored_block(s, (char*)0, 0L, 0);
00813                 /* For a full flush, this empty block will be recognized
00814                  * as a special marker by inflate_sync().
00815                  */
00816                 if (flush == Z_FULL_FLUSH) {
00817                     CLEAR_HASH(s);             /* forget history */
00818                 }
00819             }
00820             flush_pending(strm);
00821             if (strm->avail_out == 0) {
00822               s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */
00823               return Z_OK;
00824             }
00825         }
00826     }
00827     Assert(strm->avail_out > 0, "bug2");
00828 
00829     if (flush != Z_FINISH) return Z_OK;
00830     if (s->wrap <= 0) return Z_STREAM_END;
00831 
00832     /* Write the trailer */
00833 #ifdef GZIP
00834     if (s->wrap == 2) {
00835         put_byte(s, (Byte)(strm->adler & 0xff));
00836         put_byte(s, (Byte)((strm->adler >> 8) & 0xff));
00837         put_byte(s, (Byte)((strm->adler >> 16) & 0xff));
00838         put_byte(s, (Byte)((strm->adler >> 24) & 0xff));
00839         put_byte(s, (Byte)(strm->total_in & 0xff));
00840         put_byte(s, (Byte)((strm->total_in >> 8) & 0xff));
00841         put_byte(s, (Byte)((strm->total_in >> 16) & 0xff));
00842         put_byte(s, (Byte)((strm->total_in >> 24) & 0xff));
00843     }
00844     else
00845 #endif
00846     {
00847         putShortMSB(s, (uInt)(strm->adler >> 16));
00848         putShortMSB(s, (uInt)(strm->adler & 0xffff));
00849     }
00850     flush_pending(strm);
00851     /* If avail_out is zero, the application will call deflate again
00852      * to flush the rest.
00853      */
00854     if (s->wrap > 0) s->wrap = -s->wrap; /* write the trailer only once! */
00855     return s->pending != 0 ? Z_OK : Z_STREAM_END;
00856 }
00857 
00858 /* ========================================================================= */
00859 int ZEXPORT deflateEnd (strm)
00860     z_streamp strm;
00861 {
00862     int status;
00863 
00864     if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
00865 
00866     status = strm->state->status;
00867     if (status != INIT_STATE &&
00868         status != EXTRA_STATE &&
00869         status != NAME_STATE &&
00870         status != COMMENT_STATE &&
00871         status != HCRC_STATE &&
00872         status != BUSY_STATE &&
00873         status != FINISH_STATE) {
00874       return Z_STREAM_ERROR;
00875     }
00876 
00877     /* Deallocate in reverse order of allocations: */
00878     TRY_FREE(strm, strm->state->pending_buf);
00879     TRY_FREE(strm, strm->state->head);
00880     TRY_FREE(strm, strm->state->prev);
00881     TRY_FREE(strm, strm->state->window);
00882 
00883     ZFREE(strm, strm->state);
00884     strm->state = Z_NULL;
00885 
00886     return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK;
00887 }
00888 
00889 /* =========================================================================
00890  * Copy the source state to the destination state.
00891  * To simplify the source, this is not supported for 16-bit MSDOS (which
00892  * doesn't have enough memory anyway to duplicate compression states).
00893  */
00894 int ZEXPORT deflateCopy (dest, source)
00895     z_streamp dest;
00896     z_streamp source;
00897 {
00898 #ifdef MAXSEG_64K
00899     return Z_STREAM_ERROR;
00900 #else
00901     deflate_state *ds;
00902     deflate_state *ss;
00903     ushf *overlay;
00904 
00905 
00906     if (source == Z_NULL || dest == Z_NULL || source->state == Z_NULL) {
00907         return Z_STREAM_ERROR;
00908     }
00909 
00910     ss = source->state;
00911 
00912     zmemcpy(dest, source, sizeof(z_stream));
00913 
00914     ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state));
00915     if (ds == Z_NULL) return Z_MEM_ERROR;
00916     dest->state = (struct internal_state FAR *) ds;
00917     zmemcpy(ds, ss, sizeof(deflate_state));
00918     ds->strm = dest;
00919 
00920     ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte));
00921     ds->prev   = (Posf *)  ZALLOC(dest, ds->w_size, sizeof(Pos));
00922     ds->head   = (Posf *)  ZALLOC(dest, ds->hash_size, sizeof(Pos));
00923     overlay = (ushf *) ZALLOC(dest, ds->lit_bufsize, sizeof(ush)+2);
00924     ds->pending_buf = (uchf *) overlay;
00925 
00926     if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL ||
00927         ds->pending_buf == Z_NULL) {
00928         deflateEnd (dest);
00929         return Z_MEM_ERROR;
00930     }
00931     /* following zmemcpy do not work for 16-bit MSDOS */
00932     zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte));
00933     zmemcpy(ds->prev, ss->prev, ds->w_size * sizeof(Pos));
00934     zmemcpy(ds->head, ss->head, ds->hash_size * sizeof(Pos));
00935     zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size);
00936 
00937     ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf);
00938     ds->d_buf = overlay + ds->lit_bufsize/sizeof(ush);
00939     ds->l_buf = ds->pending_buf + (1+sizeof(ush))*ds->lit_bufsize;
00940 
00941     ds->l_desc.dyn_tree = ds->dyn_ltree;
00942     ds->d_desc.dyn_tree = ds->dyn_dtree;
00943     ds->bl_desc.dyn_tree = ds->bl_tree;
00944 
00945     return Z_OK;
00946 #endif /* MAXSEG_64K */
00947 }
00948 
00949 /* ===========================================================================
00950  * Read a new buffer from the current input stream, update the adler32
00951  * and total number of bytes read.  All deflate() input goes through
00952  * this function so some applications may wish to modify it to avoid
00953  * allocating a large strm->next_in buffer and copying from it.
00954  * (See also flush_pending()).
00955  */
00956 local int read_buf(strm, buf, size)
00957     z_streamp strm;
00958     Bytef *buf;
00959     unsigned size;
00960 {
00961     unsigned len = strm->avail_in;
00962 
00963     if (len > size) len = size;
00964     if (len == 0) return 0;
00965 
00966     strm->avail_in  -= len;
00967 
00968     if (strm->state->wrap == 1) {
00969         strm->adler = adler32(strm->adler, strm->next_in, len);
00970     }
00971 #ifdef GZIP
00972     else if (strm->state->wrap == 2) {
00973         strm->adler = crc32(strm->adler, strm->next_in, len);
00974     }
00975 #endif
00976     zmemcpy(buf, strm->next_in, len);
00977     strm->next_in  += len;
00978     strm->total_in += len;
00979 
00980     return (int)len;
00981 }
00982 
00983 /* ===========================================================================
00984  * Initialize the "longest match" routines for a new zlib stream
00985  */
00986 local void lm_init (s)
00987     deflate_state *s;
00988 {
00989     s->window_size = (ulg)2L*s->w_size;
00990 
00991     CLEAR_HASH(s);
00992 
00993     /* Set the default configuration parameters:
00994      */
00995     s->max_lazy_match   = configuration_table[s->level].max_lazy;
00996     s->good_match       = configuration_table[s->level].good_length;
00997     s->nice_match       = configuration_table[s->level].nice_length;
00998     s->max_chain_length = configuration_table[s->level].max_chain;
00999 
01000     s->strstart = 0;
01001     s->block_start = 0L;
01002     s->lookahead = 0;
01003     s->match_length = s->prev_length = MIN_MATCH-1;
01004     s->match_available = 0;
01005     s->ins_h = 0;
01006 #ifndef FASTEST
01007 #ifdef ASMV
01008     match_init(); /* initialize the asm code */
01009 #endif
01010 #endif
01011 }
01012 
01013 #ifndef FASTEST
01014 /* ===========================================================================
01015  * Set match_start to the longest match starting at the given string and
01016  * return its length. Matches shorter or equal to prev_length are discarded,
01017  * in which case the result is equal to prev_length and match_start is
01018  * garbage.
01019  * IN assertions: cur_match is the head of the hash chain for the current
01020  *   string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
01021  * OUT assertion: the match length is not greater than s->lookahead.
01022  */
01023 #ifndef ASMV
01024 /* For 80x86 and 680x0, an optimized version will be provided in match.asm or
01025  * match.S. The code will be functionally equivalent.
01026  */
01027 local uInt longest_match(s, cur_match)
01028     deflate_state *s;
01029     IPos cur_match;                             /* current match */
01030 {
01031     unsigned chain_length = s->max_chain_length;/* max hash chain length */
01032     register Bytef *scan = s->window + s->strstart; /* current string */
01033     register Bytef *match;                       /* matched string */
01034     register int len;                           /* length of current match */
01035     int best_len = s->prev_length;              /* best match length so far */
01036     int nice_match = s->nice_match;             /* stop if match long enough */
01037     IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
01038         s->strstart - (IPos)MAX_DIST(s) : NIL;
01039     /* Stop when cur_match becomes <= limit. To simplify the code,
01040      * we prevent matches with the string of window index 0.
01041      */
01042     Posf *prev = s->prev;
01043     uInt wmask = s->w_mask;
01044 
01045 #ifdef UNALIGNED_OK
01046     /* Compare two bytes at a time. Note: this is not always beneficial.
01047      * Try with and without -DUNALIGNED_OK to check.
01048      */
01049     register Bytef *strend = s->window + s->strstart + MAX_MATCH - 1;
01050     register ush scan_start = *(ushf*)scan;
01051     register ush scan_end   = *(ushf*)(scan+best_len-1);
01052 #else
01053     register Bytef *strend = s->window + s->strstart + MAX_MATCH;
01054     register Byte scan_end1  = scan[best_len-1];
01055     register Byte scan_end   = scan[best_len];
01056 #endif
01057 
01058     /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
01059      * It is easy to get rid of this optimization if necessary.
01060      */
01061     Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
01062 
01063     /* Do not waste too much time if we already have a good match: */
01064     if (s->prev_length >= s->good_match) {
01065         chain_length >>= 2;
01066     }
01067     /* Do not look for matches beyond the end of the input. This is necessary
01068      * to make deflate deterministic.
01069      */
01070     if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead;
01071 
01072     Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
01073 
01074     do {
01075         Assert(cur_match < s->strstart, "no future");
01076         match = s->window + cur_match;
01077 
01078         /* Skip to next match if the match length cannot increase
01079          * or if the match length is less than 2.  Note that the checks below
01080          * for insufficient lookahead only occur occasionally for performance
01081          * reasons.  Therefore uninitialized memory will be accessed, and
01082          * conditional jumps will be made that depend on those values.
01083          * However the length of the match is limited to the lookahead, so
01084          * the output of deflate is not affected by the uninitialized values.
01085          */
01086 #if (defined(UNALIGNED_OK) && MAX_MATCH == 258)
01087         /* This code assumes sizeof(unsigned short) == 2. Do not use
01088          * UNALIGNED_OK if your compiler uses a different size.
01089          */
01090         if (*(ushf*)(match+best_len-1) != scan_end ||
01091             *(ushf*)match != scan_start) continue;
01092 
01093         /* It is not necessary to compare scan[2] and match[2] since they are
01094          * always equal when the other bytes match, given that the hash keys
01095          * are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at
01096          * strstart+3, +5, ... up to strstart+257. We check for insufficient
01097          * lookahead only every 4th comparison; the 128th check will be made
01098          * at strstart+257. If MAX_MATCH-2 is not a multiple of 8, it is
01099          * necessary to put more guard bytes at the end of the window, or
01100          * to check more often for insufficient lookahead.
01101          */
01102         Assert(scan[2] == match[2], "scan[2]?");
01103         scan++, match++;
01104         do {
01105         } while (*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
01106                  *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
01107                  *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
01108                  *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
01109                  scan < strend);
01110         /* The funny "do {}" generates better code on most compilers */
01111 
01112         /* Here, scan <= window+strstart+257 */
01113         Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
01114         if (*scan == *match) scan++;
01115 
01116         len = (MAX_MATCH - 1) - (int)(strend-scan);
01117         scan = strend - (MAX_MATCH-1);
01118 
01119 #else /* UNALIGNED_OK */
01120 
01121         if (match[best_len]   != scan_end  ||
01122             match[best_len-1] != scan_end1 ||
01123             *match            != *scan     ||
01124             *++match          != scan[1])      continue;
01125 
01126         /* The check at best_len-1 can be removed because it will be made
01127          * again later. (This heuristic is not always a win.)
01128          * It is not necessary to compare scan[2] and match[2] since they
01129          * are always equal when the other bytes match, given that
01130          * the hash keys are equal and that HASH_BITS >= 8.
01131          */
01132         scan += 2, match++;
01133         Assert(*scan == *match, "match[2]?");
01134 
01135         /* We check for insufficient lookahead only every 8th comparison;
01136          * the 256th check will be made at strstart+258.
01137          */
01138         do {
01139         } while (*++scan == *++match && *++scan == *++match &&
01140                  *++scan == *++match && *++scan == *++match &&
01141                  *++scan == *++match && *++scan == *++match &&
01142                  *++scan == *++match && *++scan == *++match &&
01143                  scan < strend);
01144 
01145         Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
01146 
01147         len = MAX_MATCH - (int)(strend - scan);
01148         scan = strend - MAX_MATCH;
01149 
01150 #endif /* UNALIGNED_OK */
01151 
01152         if (len > best_len) {
01153             s->match_start = cur_match;
01154             best_len = len;
01155             if (len >= nice_match) break;
01156 #ifdef UNALIGNED_OK
01157             scan_end = *(ushf*)(scan+best_len-1);
01158 #else
01159             scan_end1  = scan[best_len-1];
01160             scan_end   = scan[best_len];
01161 #endif
01162         }
01163     } while ((cur_match = prev[cur_match & wmask]) > limit
01164              && --chain_length != 0);
01165 
01166     if ((uInt)best_len <= s->lookahead) return (uInt)best_len;
01167     return s->lookahead;
01168 }
01169 #endif /* ASMV */
01170 #endif /* FASTEST */
01171 
01172 /* ---------------------------------------------------------------------------
01173  * Optimized version for level == 1 or strategy == Z_RLE only
01174  */
01175 local uInt longest_match_fast(s, cur_match)
01176     deflate_state *s;
01177     IPos cur_match;                             /* current match */
01178 {
01179     register Bytef *scan = s->window + s->strstart; /* current string */
01180     register Bytef *match;                       /* matched string */
01181     register int len;                           /* length of current match */
01182     register Bytef *strend = s->window + s->strstart + MAX_MATCH;
01183 
01184     /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
01185      * It is easy to get rid of this optimization if necessary.
01186      */
01187     Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
01188 
01189     Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
01190 
01191     Assert(cur_match < s->strstart, "no future");
01192 
01193     match = s->window + cur_match;
01194 
01195     /* Return failure if the match length is less than 2:
01196      */
01197     if (match[0] != scan[0] || match[1] != scan[1]) return MIN_MATCH-1;
01198 
01199     /* The check at best_len-1 can be removed because it will be made
01200      * again later. (This heuristic is not always a win.)
01201      * It is not necessary to compare scan[2] and match[2] since they
01202      * are always equal when the other bytes match, given that
01203      * the hash keys are equal and that HASH_BITS >= 8.
01204      */
01205     scan += 2, match += 2;
01206     Assert(*scan == *match, "match[2]?");
01207 
01208     /* We check for insufficient lookahead only every 8th comparison;
01209      * the 256th check will be made at strstart+258.
01210      */
01211     do {
01212     } while (*++scan == *++match && *++scan == *++match &&
01213              *++scan == *++match && *++scan == *++match &&
01214              *++scan == *++match && *++scan == *++match &&
01215              *++scan == *++match && *++scan == *++match &&
01216              scan < strend);
01217 
01218     Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
01219 
01220     len = MAX_MATCH - (int)(strend - scan);
01221 
01222     if (len < MIN_MATCH) return MIN_MATCH - 1;
01223 
01224     s->match_start = cur_match;
01225     return (uInt)len <= s->lookahead ? (uInt)len : s->lookahead;
01226 }
01227 
01228 #ifdef DEBUG
01229 /* ===========================================================================
01230  * Check that the match at match_start is indeed a match.
01231  */
01232 local void check_match(s, start, match, length)
01233     deflate_state *s;
01234     IPos start, match;
01235     int length;
01236 {
01237     /* check that the match is indeed a match */
01238     if (zmemcmp(s->window + match,
01239                 s->window + start, length) != EQUAL) {
01240         fprintf(stderr, " start %u, match %u, length %d\n",
01241                 start, match, length);
01242         do {
01243             fprintf(stderr, "%c%c", s->window[match++], s->window[start++]);
01244         } while (--length != 0);
01245         z_error("invalid match");
01246     }
01247     if (z_verbose > 1) {
01248         fprintf(stderr,"\\[%d,%d]", start-match, length);
01249         do { putc(s->window[start++], stderr); } while (--length != 0);
01250     }
01251 }
01252 #else
01253 #  define check_match(s, start, match, length)
01254 #endif /* DEBUG */
01255 
01256 /* ===========================================================================
01257  * Fill the window when the lookahead becomes insufficient.
01258  * Updates strstart and lookahead.
01259  *
01260  * IN assertion: lookahead < MIN_LOOKAHEAD
01261  * OUT assertions: strstart <= window_size-MIN_LOOKAHEAD
01262  *    At least one byte has been read, or avail_in == 0; reads are
01263  *    performed for at least two bytes (required for the zip translate_eol
01264  *    option -- not supported here).
01265  */
01266 local void fill_window(s)
01267     deflate_state *s;
01268 {
01269     register unsigned n, m;
01270     register Posf *p;
01271     unsigned more;    /* Amount of free space at the end of the window. */
01272     uInt wsize = s->w_size;
01273 
01274     do {
01275         more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart);
01276 
01277         /* Deal with !@#$% 64K limit: */
01278         if (sizeof(int) <= 2) {
01279             if (more == 0 && s->strstart == 0 && s->lookahead == 0) {
01280                 more = wsize;
01281 
01282             } else if (more == (unsigned)(-1)) {
01283                 /* Very unlikely, but possible on 16 bit machine if
01284                  * strstart == 0 && lookahead == 1 (input done a byte at time)
01285                  */
01286                 more--;
01287             }
01288         }
01289 
01290         /* If the window is almost full and there is insufficient lookahead,
01291          * move the upper half to the lower one to make room in the upper half.
01292          */
01293         if (s->strstart >= wsize+MAX_DIST(s)) {
01294 
01295             zmemcpy(s->window, s->window+wsize, (unsigned)wsize);
01296             s->match_start -= wsize;
01297             s->strstart    -= wsize; /* we now have strstart >= MAX_DIST */
01298             s->block_start -= (long) wsize;
01299 
01300             /* Slide the hash table (could be avoided with 32 bit values
01301                at the expense of memory usage). We slide even when level == 0
01302                to keep the hash table consistent if we switch back to level > 0
01303                later. (Using level 0 permanently is not an optimal usage of
01304                zlib, so we don't care about this pathological case.)
01305              */
01306             /* %%% avoid this when Z_RLE */
01307             n = s->hash_size;
01308             p = &s->head[n];
01309             do {
01310                 m = *--p;
01311                 *p = (Pos)(m >= wsize ? m-wsize : NIL);
01312             } while (--n);
01313 
01314             n = wsize;
01315 #ifndef FASTEST
01316             p = &s->prev[n];
01317             do {
01318                 m = *--p;
01319                 *p = (Pos)(m >= wsize ? m-wsize : NIL);
01320                 /* If n is not on any hash chain, prev[n] is garbage but
01321                  * its value will never be used.
01322                  */
01323             } while (--n);
01324 #endif
01325             more += wsize;
01326         }
01327         if (s->strm->avail_in == 0) return;
01328 
01329         /* If there was no sliding:
01330          *    strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 &&
01331          *    more == window_size - lookahead - strstart
01332          * => more >= window_size - (MIN_LOOKAHEAD-1 + WSIZE + MAX_DIST-1)
01333          * => more >= window_size - 2*WSIZE + 2
01334          * In the BIG_MEM or MMAP case (not yet supported),
01335          *   window_size == input_size + MIN_LOOKAHEAD  &&
01336          *   strstart + s->lookahead <= input_size => more >= MIN_LOOKAHEAD.
01337          * Otherwise, window_size == 2*WSIZE so more >= 2.
01338          * If there was sliding, more >= WSIZE. So in all cases, more >= 2.
01339          */
01340         Assert(more >= 2, "more < 2");
01341 
01342         n = read_buf(s->strm, s->window + s->strstart + s->lookahead, more);
01343         s->lookahead += n;
01344 
01345         /* Initialize the hash value now that we have some input: */
01346         if (s->lookahead >= MIN_MATCH) {
01347             s->ins_h = s->window[s->strstart];
01348             UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
01349 #if MIN_MATCH != 3
01350             Call UPDATE_HASH() MIN_MATCH-3 more times
01351 #endif
01352         }
01353         /* If the whole input has less than MIN_MATCH bytes, ins_h is garbage,
01354          * but this is not important since only literal bytes will be emitted.
01355          */
01356 
01357     } while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0);
01358 }
01359 
01360 /* ===========================================================================
01361  * Flush the current block, with given end-of-file flag.
01362  * IN assertion: strstart is set to the end of the current match.
01363  */
01364 #define FLUSH_BLOCK_ONLY(s, eof) { \
01365    _tr_flush_block(s, (s->block_start >= 0L ? \
01366                    (charf *)&s->window[(unsigned)s->block_start] : \
01367                    (charf *)Z_NULL), \
01368                 (ulg)((long)s->strstart - s->block_start), \
01369                 (eof)); \
01370    s->block_start = s->strstart; \
01371    flush_pending(s->strm); \
01372    Tracev((stderr,"[FLUSH]")); \
01373 }
01374 
01375 /* Same but force premature exit if necessary. */
01376 #define FLUSH_BLOCK(s, eof) { \
01377    FLUSH_BLOCK_ONLY(s, eof); \
01378    if (s->strm->avail_out == 0) return (eof) ? finish_started : need_more; \
01379 }
01380 
01381 /* ===========================================================================
01382  * Copy without compression as much as possible from the input stream, return
01383  * the current block state.
01384  * This function does not insert new strings in the dictionary since
01385  * uncompressible data is probably not useful. This function is used
01386  * only for the level=0 compression option.
01387  * NOTE: this function should be optimized to avoid extra copying from
01388  * window to pending_buf.
01389  */
01390 local block_state deflate_stored(s, flush)
01391     deflate_state *s;
01392     int flush;
01393 {
01394     /* Stored blocks are limited to 0xffff bytes, pending_buf is limited
01395      * to pending_buf_size, and each stored block has a 5 byte header:
01396      */
01397     ulg max_block_size = 0xffff;
01398     ulg max_start;
01399 
01400     if (max_block_size > s->pending_buf_size - 5) {
01401         max_block_size = s->pending_buf_size - 5;
01402     }
01403 
01404     /* Copy as much as possible from input to output: */
01405     for (;;) {
01406         /* Fill the window as much as possible: */
01407         if (s->lookahead <= 1) {
01408 
01409             Assert(s->strstart < s->w_size+MAX_DIST(s) ||
01410                    s->block_start >= (long)s->w_size, "slide too late");
01411 
01412             fill_window(s);
01413             if (s->lookahead == 0 && flush == Z_NO_FLUSH) return need_more;
01414 
01415             if (s->lookahead == 0) break; /* flush the current block */
01416         }
01417         Assert(s->block_start >= 0L, "block gone");
01418 
01419         s->strstart += s->lookahead;
01420         s->lookahead = 0;
01421 
01422         /* Emit a stored block if pending_buf will be full: */
01423         max_start = s->block_start + max_block_size;
01424         if (s->strstart == 0 || (ulg)s->strstart >= max_start) {
01425             /* strstart == 0 is possible when wraparound on 16-bit machine */
01426             s->lookahead = (uInt)(s->strstart - max_start);
01427             s->strstart = (uInt)max_start;
01428             FLUSH_BLOCK(s, 0);
01429         }
01430         /* Flush if we may have to slide, otherwise block_start may become
01431          * negative and the data will be gone:
01432          */
01433         if (s->strstart - (uInt)s->block_start >= MAX_DIST(s)) {
01434             FLUSH_BLOCK(s, 0);
01435         }
01436     }
01437     FLUSH_BLOCK(s, flush == Z_FINISH);
01438     return flush == Z_FINISH ? finish_done : block_done;
01439 }
01440 
01441 /* ===========================================================================
01442  * Compress as much as possible from the input stream, return the current
01443  * block state.
01444  * This function does not perform lazy evaluation of matches and inserts
01445  * new strings in the dictionary only for unmatched strings or for short
01446  * matches. It is used only for the fast compression options.
01447  */
01448 local block_state deflate_fast(s, flush)
01449     deflate_state *s;
01450     int flush;
01451 {
01452     IPos hash_head = NIL; /* head of the hash chain */
01453     int bflush;           /* set if current block must be flushed */
01454 
01455     for (;;) {
01456         /* Make sure that we always have enough lookahead, except
01457          * at the end of the input file. We need MAX_MATCH bytes
01458          * for the next match, plus MIN_MATCH bytes to insert the
01459          * string following the next match.
01460          */
01461         if (s->lookahead < MIN_LOOKAHEAD) {
01462             fill_window(s);
01463             if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
01464                 return need_more;
01465             }
01466             if (s->lookahead == 0) break; /* flush the current block */
01467         }
01468 
01469         /* Insert the string window[strstart .. strstart+2] in the
01470          * dictionary, and set hash_head to the head of the hash chain:
01471          */
01472         if (s->lookahead >= MIN_MATCH) {
01473             INSERT_STRING(s, s->strstart, hash_head);
01474         }
01475 
01476         /* Find the longest match, discarding those <= prev_length.
01477          * At this point we have always match_length < MIN_MATCH
01478          */
01479         if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) {
01480             /* To simplify the code, we prevent matches with the string
01481              * of window index 0 (in particular we have to avoid a match
01482              * of the string with itself at the start of the input file).
01483              */
01484 #ifdef FASTEST
01485             if ((s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) ||
01486                 (s->strategy == Z_RLE && s->strstart - hash_head == 1)) {
01487                 s->match_length = longest_match_fast (s, hash_head);
01488             }
01489 #else
01490             if (s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) {
01491                 s->match_length = longest_match (s, hash_head);
01492             } else if (s->strategy == Z_RLE && s->strstart - hash_head == 1) {
01493                 s->match_length = longest_match_fast (s, hash_head);
01494             }
01495 #endif
01496             /* longest_match() or longest_match_fast() sets match_start */
01497         }
01498         if (s->match_length >= MIN_MATCH) {
01499             check_match(s, s->strstart, s->match_start, s->match_length);
01500 
01501             _tr_tally_dist(s, s->strstart - s->match_start,
01502                            s->match_length - MIN_MATCH, bflush);
01503 
01504             s->lookahead -= s->match_length;
01505 
01506             /* Insert new strings in the hash table only if the match length
01507              * is not too large. This saves time but degrades compression.
01508              */
01509 #ifndef FASTEST
01510             if (s->match_length <= s->max_insert_length &&
01511                 s->lookahead >= MIN_MATCH) {
01512                 s->match_length--; /* string at strstart already in table */
01513                 do {
01514                     s->strstart++;
01515                     INSERT_STRING(s, s->strstart, hash_head);
01516                     /* strstart never exceeds WSIZE-MAX_MATCH, so there are
01517                      * always MIN_MATCH bytes ahead.
01518                      */
01519                 } while (--s->match_length != 0);
01520                 s->strstart++;
01521             } else
01522 #endif
01523             {
01524                 s->strstart += s->match_length;
01525                 s->match_length = 0;
01526                 s->ins_h = s->window[s->strstart];
01527                 UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
01528 #if MIN_MATCH != 3
01529                 Call UPDATE_HASH() MIN_MATCH-3 more times
01530 #endif
01531                 /* If lookahead < MIN_MATCH, ins_h is garbage, but it does not
01532                  * matter since it will be recomputed at next deflate call.
01533                  */
01534             }
01535         } else {
01536             /* No match, output a literal byte */
01537             Tracevv((stderr,"%c", s->window[s->strstart]));
01538             _tr_tally_lit (s, s->window[s->strstart], bflush);
01539             s->lookahead--;
01540             s->strstart++;
01541         }
01542         if (bflush) FLUSH_BLOCK(s, 0);
01543     }
01544     FLUSH_BLOCK(s, flush == Z_FINISH);
01545     return flush == Z_FINISH ? finish_done : block_done;
01546 }
01547 
01548 #ifndef FASTEST
01549 /* ===========================================================================
01550  * Same as above, but achieves better compression. We use a lazy
01551  * evaluation for matches: a match is finally adopted only if there is
01552  * no better match at the next window position.
01553  */
01554 local block_state deflate_slow(s, flush)
01555     deflate_state *s;
01556     int flush;
01557 {
01558     IPos hash_head = NIL;    /* head of hash chain */
01559     int bflush;              /* set if current block must be flushed */
01560 
01561     /* Process the input block. */
01562     for (;;) {
01563         /* Make sure that we always have enough lookahead, except
01564          * at the end of the input file. We need MAX_MATCH bytes
01565          * for the next match, plus MIN_MATCH bytes to insert the
01566          * string following the next match.
01567          */
01568         if (s->lookahead < MIN_LOOKAHEAD) {
01569             fill_window(s);
01570             if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
01571                 return need_more;
01572             }
01573             if (s->lookahead == 0) break; /* flush the current block */
01574         }
01575 
01576         /* Insert the string window[strstart .. strstart+2] in the
01577          * dictionary, and set hash_head to the head of the hash chain:
01578          */
01579         if (s->lookahead >= MIN_MATCH) {
01580             INSERT_STRING(s, s->strstart, hash_head);
01581         }
01582 
01583         /* Find the longest match, discarding those <= prev_length.
01584          */
01585         s->prev_length = s->match_length, s->prev_match = s->match_start;
01586         s->match_length = MIN_MATCH-1;
01587 
01588         if (hash_head != NIL && s->prev_length < s->max_lazy_match &&
01589             s->strstart - hash_head <= MAX_DIST(s)) {
01590             /* To simplify the code, we prevent matches with the string
01591              * of window index 0 (in particular we have to avoid a match
01592              * of the string with itself at the start of the input file).
01593              */
01594             if (s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) {
01595                 s->match_length = longest_match (s, hash_head);
01596             } else if (s->strategy == Z_RLE && s->strstart - hash_head == 1) {
01597                 s->match_length = longest_match_fast (s, hash_head);
01598             }
01599             /* longest_match() or longest_match_fast() sets match_start */
01600 
01601             if (s->match_length <= 5 && (s->strategy == Z_FILTERED
01602 #if TOO_FAR <= 32767
01603                 || (s->match_length == MIN_MATCH &&
01604                     s->strstart - s->match_start > TOO_FAR)
01605 #endif
01606                 )) {
01607 
01608                 /* If prev_match is also MIN_MATCH, match_start is garbage
01609                  * but we will ignore the current match anyway.
01610                  */
01611                 s->match_length = MIN_MATCH-1;
01612             }
01613         }
01614         /* If there was a match at the previous step and the current
01615          * match is not better, output the previous match:
01616          */
01617         if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) {
01618             uInt max_insert = s->strstart + s->lookahead - MIN_MATCH;
01619             /* Do not insert strings in hash table beyond this. */
01620 
01621             check_match(s, s->strstart-1, s->prev_match, s->prev_length);
01622 
01623             _tr_tally_dist(s, s->strstart -1 - s->prev_match,
01624                            s->prev_length - MIN_MATCH, bflush);
01625 
01626             /* Insert in hash table all strings up to the end of the match.
01627              * strstart-1 and strstart are already inserted. If there is not
01628              * enough lookahead, the last two strings are not inserted in
01629              * the hash table.
01630              */
01631             s->lookahead -= s->prev_length-1;
01632             s->prev_length -= 2;
01633             do {
01634                 if (++s->strstart <= max_insert) {
01635                     INSERT_STRING(s, s->strstart, hash_head);
01636                 }
01637             } while (--s->prev_length != 0);
01638             s->match_available = 0;
01639             s->match_length = MIN_MATCH-1;
01640             s->strstart++;
01641 
01642             if (bflush) FLUSH_BLOCK(s, 0);
01643 
01644         } else if (s->match_available) {
01645             /* If there was no match at the previous position, output a
01646              * single literal. If there was a match but the current match
01647              * is longer, truncate the previous match to a single literal.
01648              */
01649             Tracevv((stderr,"%c", s->window[s->strstart-1]));
01650             _tr_tally_lit(s, s->window[s->strstart-1], bflush);
01651             if (bflush) {
01652                 FLUSH_BLOCK_ONLY(s, 0);
01653             }
01654             s->strstart++;
01655             s->lookahead--;
01656             if (s->strm->avail_out == 0) return need_more;
01657         } else {
01658             /* There is no previous match to compare with, wait for
01659              * the next step to decide.
01660              */
01661             s->match_available = 1;
01662             s->strstart++;
01663             s->lookahead--;
01664         }
01665     }
01666     Assert (flush != Z_NO_FLUSH, "no flush?");
01667     if (s->match_available) {
01668         Tracevv((stderr,"%c", s->window[s->strstart-1]));
01669         _tr_tally_lit(s, s->window[s->strstart-1], bflush);
01670         s->match_available = 0;
01671     }
01672     FLUSH_BLOCK(s, flush == Z_FINISH);
01673     return flush == Z_FINISH ? finish_done : block_done;
01674 }
01675 #endif /* FASTEST */
01676 
01677 #if 0
01678 /* ===========================================================================
01679  * For Z_RLE, simply look for runs of bytes, generate matches only of distance
01680  * one.  Do not maintain a hash table.  (It will be regenerated if this run of
01681  * deflate switches away from Z_RLE.)
01682  */
01683 local block_state deflate_rle(s, flush)
01684     deflate_state *s;
01685     int flush;
01686 {
01687     int bflush;         /* set if current block must be flushed */
01688     uInt run;           /* length of run */
01689     uInt max;           /* maximum length of run */
01690     uInt prev;          /* byte at distance one to match */
01691     Bytef *scan;        /* scan for end of run */
01692 
01693     for (;;) {
01694         /* Make sure that we always have enough lookahead, except
01695          * at the end of the input file. We need MAX_MATCH bytes
01696          * for the longest encodable run.
01697          */
01698         if (s->lookahead < MAX_MATCH) {
01699             fill_window(s);
01700             if (s->lookahead < MAX_MATCH && flush == Z_NO_FLUSH) {
01701                 return need_more;
01702             }
01703             if (s->lookahead == 0) break; /* flush the current block */
01704         }
01705 
01706         /* See how many times the previous byte repeats */
01707         run = 0;
01708         if (s->strstart > 0) {      /* if there is a previous byte, that is */
01709             max = s->lookahead < MAX_MATCH ? s->lookahead : MAX_MATCH;
01710             scan = s->window + s->strstart - 1;
01711             prev = *scan++;
01712             do {
01713                 if (*scan++ != prev)
01714                     break;
01715             } while (++run < max);
01716         }
01717 
01718         /* Emit match if have run of MIN_MATCH or longer, else emit literal */
01719         if (run >= MIN_MATCH) {
01720             check_match(s, s->strstart, s->strstart - 1, run);
01721             _tr_tally_dist(s, 1, run - MIN_MATCH, bflush);
01722             s->lookahead -= run;
01723             s->strstart += run;
01724         } else {
01725             /* No match, output a literal byte */
01726             Tracevv((stderr,"%c", s->window[s->strstart]));
01727             _tr_tally_lit (s, s->window[s->strstart], bflush);
01728             s->lookahead--;
01729             s->strstart++;
01730         }
01731         if (bflush) FLUSH_BLOCK(s, 0);
01732     }
01733     FLUSH_BLOCK(s, flush == Z_FINISH);
01734     return flush == Z_FINISH ? finish_done : block_done;
01735 }
01736 #endif


openhrp3
Author(s): AIST, General Robotix Inc., Nakamura Lab of Dept. of Mechano Informatics at University of Tokyo
autogenerated on Sun Apr 2 2017 03:43:53