Back to index

plt-scheme  4.2.1
deflate.c
Go to the documentation of this file.
00001 /* deflate.c -- compress data using the deflation algorithm
00002  * Copyright (C) 1995-2002 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 ftp://ds.internic.net/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: deflate.c,v 1.1 2004/03/01 20:54:46 cozmic Exp $ */
00051 
00052 #include "deflate.h"
00053 
00054 const char deflate_copyright[] =
00055    " deflate 1.1.4 Copyright 1995-2002 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 local block_state deflate_slow   OF((deflate_state *s, int flush));
00080 local void lm_init        OF((deflate_state *s));
00081 local void putShortMSB    OF((deflate_state *s, uInt b));
00082 local void flush_pending  OF((z_streamp strm));
00083 local int read_buf        OF((z_streamp strm, Bytef *buf, unsigned size));
00084 #ifdef ASMV
00085       void match_init OF((void)); /* asm code initialization */
00086       uInt longest_match  OF((deflate_state *s, IPos cur_match));
00087 #else
00088 local uInt longest_match  OF((deflate_state *s, IPos cur_match));
00089 #endif
00090 
00091 #ifdef DEBUG
00092 local  void check_match OF((deflate_state *s, IPos start, IPos match,
00093                             int length));
00094 #endif
00095 
00096 /* ===========================================================================
00097  * Local data
00098  */
00099 
00100 #define NIL 0
00101 /* Tail of hash chains */
00102 
00103 #ifndef TOO_FAR
00104 #  define TOO_FAR 4096
00105 #endif
00106 /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
00107 
00108 #define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
00109 /* Minimum amount of lookahead, except at the end of the input file.
00110  * See deflate.c for comments about the MIN_MATCH+1.
00111  */
00112 
00113 /* Values for max_lazy_match, good_match and max_chain_length, depending on
00114  * the desired pack level (0..9). The values given below have been tuned to
00115  * exclude worst case performance for pathological files. Better values may be
00116  * found for specific files.
00117  */
00118 typedef struct config_s {
00119    ush good_length; /* reduce lazy search above this match length */
00120    ush max_lazy;    /* do not perform lazy search above this match length */
00121    ush nice_length; /* quit search above this match length */
00122    ush max_chain;
00123    compress_func func;
00124 } config;
00125 
00126 local const config configuration_table[10] = {
00127 /*      good lazy nice chain */
00128 /* 0 */ {0,    0,  0,    0, deflate_stored},  /* store only */
00129 /* 1 */ {4,    4,  8,    4, deflate_fast}, /* maximum speed, no lazy matches */
00130 /* 2 */ {4,    5, 16,    8, deflate_fast},
00131 /* 3 */ {4,    6, 32,   32, deflate_fast},
00132 
00133 /* 4 */ {4,    4, 16,   16, deflate_slow},  /* lazy matches */
00134 /* 5 */ {8,   16, 32,   32, deflate_slow},
00135 /* 6 */ {8,   16, 128, 128, deflate_slow},
00136 /* 7 */ {8,   32, 128, 256, deflate_slow},
00137 /* 8 */ {32, 128, 258, 1024, deflate_slow},
00138 /* 9 */ {32, 258, 258, 4096, deflate_slow}}; /* maximum compression */
00139 
00140 /* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
00141  * For deflate_fast() (levels <= 3) good is ignored and lazy has a different
00142  * meaning.
00143  */
00144 
00145 #define EQUAL 0
00146 /* result of memcmp for equal strings */
00147 
00148 struct static_tree_desc_s {int dummy;}; /* for buggy compilers */
00149 
00150 /* ===========================================================================
00151  * Update a hash value with the given input byte
00152  * IN  assertion: all calls to to UPDATE_HASH are made with consecutive
00153  *    input characters, so that a running hash key can be computed from the
00154  *    previous key instead of complete recalculation each time.
00155  */
00156 #define UPDATE_HASH(s,h,c) (h = (((h)<<s->hash_shift) ^ (c)) & s->hash_mask)
00157 
00158 
00159 /* ===========================================================================
00160  * Insert string str in the dictionary and set match_head to the previous head
00161  * of the hash chain (the most recent string with same hash key). Return
00162  * the previous length of the hash chain.
00163  * If this file is compiled with -DFASTEST, the compression level is forced
00164  * to 1, and no hash chains are maintained.
00165  * IN  assertion: all calls to to INSERT_STRING are made with consecutive
00166  *    input characters and the first MIN_MATCH bytes of str are valid
00167  *    (except for the last MIN_MATCH-1 bytes of the input file).
00168  */
00169 #ifdef FASTEST
00170 #define INSERT_STRING(s, str, match_head) \
00171    (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
00172     match_head = s->head[s->ins_h], \
00173     s->head[s->ins_h] = (Pos)(str))
00174 #else
00175 #define INSERT_STRING(s, str, match_head) \
00176    (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
00177     s->prev[(str) & s->w_mask] = match_head = s->head[s->ins_h], \
00178     s->head[s->ins_h] = (Pos)(str))
00179 #endif
00180 
00181 /* ===========================================================================
00182  * Initialize the hash table (avoiding 64K overflow for 16 bit systems).
00183  * prev[] will be initialized on the fly.
00184  */
00185 #define CLEAR_HASH(s) \
00186     s->head[s->hash_size-1] = NIL; \
00187     zmemzero((Bytef *)s->head, (unsigned)(s->hash_size-1)*sizeof(*s->head));
00188 
00189 /* ========================================================================= */
00190 int ZEXPORT deflateInit_(strm, level, version, stream_size)
00191     z_streamp strm;
00192     int level;
00193     const char *version;
00194     int stream_size;
00195 {
00196     return deflateInit2_(strm, level, Z_DEFLATED, MAX_WBITS, DEF_MEM_LEVEL,
00197                       Z_DEFAULT_STRATEGY, version, stream_size);
00198     /* To do: ignore strm->next_in if we use it as window */
00199 }
00200 
00201 /* ========================================================================= */
00202 int ZEXPORT deflateInit2_(strm, level, method, windowBits, memLevel, strategy,
00203                 version, stream_size)
00204     z_streamp strm;
00205     int  level;
00206     int  method;
00207     int  windowBits;
00208     int  memLevel;
00209     int  strategy;
00210     const char *version;
00211     int stream_size;
00212 {
00213     deflate_state *s;
00214     int noheader = 0;
00215     static const char* my_version = ZLIB_VERSION;
00216 
00217     ushf *overlay;
00218     /* We overlay pending_buf and d_buf+l_buf. This works since the average
00219      * output size for (length,distance) codes is <= 24 bits.
00220      */
00221 
00222     if (version == Z_NULL || version[0] != my_version[0] ||
00223         stream_size != sizeof(z_stream)) {
00224        return Z_VERSION_ERROR;
00225     }
00226     if (strm == Z_NULL) return Z_STREAM_ERROR;
00227 
00228     strm->msg = Z_NULL;
00229     if (strm->zalloc == Z_NULL) {
00230        strm->zalloc = zcalloc;
00231        strm->opaque = (voidpf)0;
00232     }
00233     if (strm->zfree == Z_NULL) strm->zfree = zcfree;
00234 
00235     if (level == Z_DEFAULT_COMPRESSION) level = 6;
00236 #ifdef FASTEST
00237     level = 1;
00238 #endif
00239 
00240     if (windowBits < 0) { /* undocumented feature: suppress zlib header */
00241         noheader = 1;
00242         windowBits = -windowBits;
00243     }
00244     if (memLevel < 1 || memLevel > MAX_MEM_LEVEL || method != Z_DEFLATED ||
00245         windowBits < 9 || windowBits > 15 || level < 0 || level > 9 ||
00246        strategy < 0 || strategy > Z_HUFFMAN_ONLY) {
00247         return Z_STREAM_ERROR;
00248     }
00249     s = (deflate_state *) ZALLOC(strm, 1, sizeof(deflate_state));
00250     if (s == Z_NULL) return Z_MEM_ERROR;
00251     strm->state = (struct internal_state FAR *)s;
00252     s->strm = strm;
00253 
00254     s->noheader = noheader;
00255     s->w_bits = windowBits;
00256     s->w_size = 1 << s->w_bits;
00257     s->w_mask = s->w_size - 1;
00258 
00259     s->hash_bits = memLevel + 7;
00260     s->hash_size = 1 << s->hash_bits;
00261     s->hash_mask = s->hash_size - 1;
00262     s->hash_shift =  ((s->hash_bits+MIN_MATCH-1)/MIN_MATCH);
00263 
00264     s->window = (Bytef *) ZALLOC(strm, s->w_size, 2*sizeof(Byte));
00265     s->prev   = (Posf *)  ZALLOC(strm, s->w_size, sizeof(Pos));
00266     s->head   = (Posf *)  ZALLOC(strm, s->hash_size, sizeof(Pos));
00267 
00268     s->lit_bufsize = 1 << (memLevel + 6); /* 16K elements by default */
00269 
00270     overlay = (ushf *) ZALLOC(strm, s->lit_bufsize, sizeof(ush)+2);
00271     s->pending_buf = (uchf *) overlay;
00272     s->pending_buf_size = (ulg)s->lit_bufsize * (sizeof(ush)+2L);
00273 
00274     if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL ||
00275         s->pending_buf == Z_NULL) {
00276         strm->msg = (char*)ERR_MSG(Z_MEM_ERROR);
00277         deflateEnd (strm);
00278         return Z_MEM_ERROR;
00279     }
00280     s->d_buf = overlay + s->lit_bufsize/sizeof(ush);
00281     s->l_buf = s->pending_buf + (1+sizeof(ush))*s->lit_bufsize;
00282 
00283     s->level = level;
00284     s->strategy = strategy;
00285     s->method = (Byte)method;
00286 
00287     return deflateReset(strm);
00288 }
00289 
00290 /* ========================================================================= */
00291 int ZEXPORT deflateSetDictionary (strm, dictionary, dictLength)
00292     z_streamp strm;
00293     const Bytef *dictionary;
00294     uInt  dictLength;
00295 {
00296     deflate_state *s;
00297     uInt length = dictLength;
00298     uInt n;
00299     IPos hash_head = 0;
00300 
00301     if (strm == Z_NULL || strm->state == Z_NULL || dictionary == Z_NULL ||
00302         strm->state->status != INIT_STATE) return Z_STREAM_ERROR;
00303 
00304     s = strm->state;
00305     strm->adler = adler32(strm->adler, dictionary, dictLength);
00306 
00307     if (length < MIN_MATCH) return Z_OK;
00308     if (length > MAX_DIST(s)) {
00309        length = MAX_DIST(s);
00310 #ifndef USE_DICT_HEAD
00311        dictionary += dictLength - length; /* use the tail of the dictionary */
00312 #endif
00313     }
00314     zmemcpy(s->window, dictionary, length);
00315     s->strstart = length;
00316     s->block_start = (long)length;
00317 
00318     /* Insert all strings in the hash table (except for the last two bytes).
00319      * s->lookahead stays null, so s->ins_h will be recomputed at the next
00320      * call of fill_window.
00321      */
00322     s->ins_h = s->window[0];
00323     UPDATE_HASH(s, s->ins_h, s->window[1]);
00324     for (n = 0; n <= length - MIN_MATCH; n++) {
00325        INSERT_STRING(s, n, hash_head);
00326     }
00327     if (hash_head) hash_head = 0;  /* to make compiler happy */
00328     return Z_OK;
00329 }
00330 
00331 /* ========================================================================= */
00332 int ZEXPORT deflateReset (strm)
00333     z_streamp strm;
00334 {
00335     deflate_state *s;
00336     
00337     if (strm == Z_NULL || strm->state == Z_NULL ||
00338         strm->zalloc == Z_NULL || strm->zfree == Z_NULL) return Z_STREAM_ERROR;
00339 
00340     strm->total_in = strm->total_out = 0;
00341     strm->msg = Z_NULL; /* use zfree if we ever allocate msg dynamically */
00342     strm->data_type = Z_UNKNOWN;
00343 
00344     s = (deflate_state *)strm->state;
00345     s->pending = 0;
00346     s->pending_out = s->pending_buf;
00347 
00348     if (s->noheader < 0) {
00349         s->noheader = 0; /* was set to -1 by deflate(..., Z_FINISH); */
00350     }
00351     s->status = s->noheader ? BUSY_STATE : INIT_STATE;
00352     strm->adler = 1;
00353     s->last_flush = Z_NO_FLUSH;
00354 
00355     _tr_init(s);
00356     lm_init(s);
00357 
00358     return Z_OK;
00359 }
00360 
00361 /* ========================================================================= */
00362 int ZEXPORT deflateParams(strm, level, strategy)
00363     z_streamp strm;
00364     int level;
00365     int strategy;
00366 {
00367     deflate_state *s;
00368     compress_func func;
00369     int err = Z_OK;
00370 
00371     if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
00372     s = strm->state;
00373 
00374     if (level == Z_DEFAULT_COMPRESSION) {
00375        level = 6;
00376     }
00377     if (level < 0 || level > 9 || strategy < 0 || strategy > Z_HUFFMAN_ONLY) {
00378        return Z_STREAM_ERROR;
00379     }
00380     func = configuration_table[s->level].func;
00381 
00382     if (func != configuration_table[level].func && strm->total_in != 0) {
00383        /* Flush the last buffer: */
00384        err = deflate(strm, Z_PARTIAL_FLUSH);
00385     }
00386     if (s->level != level) {
00387        s->level = level;
00388        s->max_lazy_match   = configuration_table[level].max_lazy;
00389        s->good_match       = configuration_table[level].good_length;
00390        s->nice_match       = configuration_table[level].nice_length;
00391        s->max_chain_length = configuration_table[level].max_chain;
00392     }
00393     s->strategy = strategy;
00394     return err;
00395 }
00396 
00397 /* =========================================================================
00398  * Put a short in the pending buffer. The 16-bit value is put in MSB order.
00399  * IN assertion: the stream state is correct and there is enough room in
00400  * pending_buf.
00401  */
00402 local void putShortMSB (s, b)
00403     deflate_state *s;
00404     uInt b;
00405 {
00406     put_byte(s, (Byte)(b >> 8));
00407     put_byte(s, (Byte)(b & 0xff));
00408 }   
00409 
00410 /* =========================================================================
00411  * Flush as much pending output as possible. All deflate() output goes
00412  * through this function so some applications may wish to modify it
00413  * to avoid allocating a large strm->next_out buffer and copying into it.
00414  * (See also read_buf()).
00415  */
00416 local void flush_pending(strm)
00417     z_streamp strm;
00418 {
00419     unsigned len = strm->state->pending;
00420 
00421     if (len > strm->avail_out) len = strm->avail_out;
00422     if (len == 0) return;
00423 
00424     zmemcpy(strm->next_out, strm->state->pending_out, len);
00425     strm->next_out  += len;
00426     strm->state->pending_out  += len;
00427     strm->total_out += len;
00428     strm->avail_out  -= len;
00429     strm->state->pending -= len;
00430     if (strm->state->pending == 0) {
00431         strm->state->pending_out = strm->state->pending_buf;
00432     }
00433 }
00434 
00435 /* ========================================================================= */
00436 int ZEXPORT deflate (strm, flush)
00437     z_streamp strm;
00438     int flush;
00439 {
00440     int old_flush; /* value of flush param for previous deflate call */
00441     deflate_state *s;
00442 
00443     if (strm == Z_NULL || strm->state == Z_NULL ||
00444        flush > Z_FINISH || flush < 0) {
00445         return Z_STREAM_ERROR;
00446     }
00447     s = strm->state;
00448 
00449     if (strm->next_out == Z_NULL ||
00450         (strm->next_in == Z_NULL && strm->avail_in != 0) ||
00451        (s->status == FINISH_STATE && flush != Z_FINISH)) {
00452         ERR_RETURN(strm, Z_STREAM_ERROR);
00453     }
00454     if (strm->avail_out == 0) ERR_RETURN(strm, Z_BUF_ERROR);
00455 
00456     s->strm = strm; /* just in case */
00457     old_flush = s->last_flush;
00458     s->last_flush = flush;
00459 
00460     /* Write the zlib header */
00461     if (s->status == INIT_STATE) {
00462 
00463         uInt header = (Z_DEFLATED + ((s->w_bits-8)<<4)) << 8;
00464         uInt level_flags = (s->level-1) >> 1;
00465 
00466         if (level_flags > 3) level_flags = 3;
00467         header |= (level_flags << 6);
00468        if (s->strstart != 0) header |= PRESET_DICT;
00469         header += 31 - (header % 31);
00470 
00471         s->status = BUSY_STATE;
00472         putShortMSB(s, header);
00473 
00474        /* Save the adler32 of the preset dictionary: */
00475        if (s->strstart != 0) {
00476            putShortMSB(s, (uInt)(strm->adler >> 16));
00477            putShortMSB(s, (uInt)(strm->adler & 0xffff));
00478        }
00479        strm->adler = 1L;
00480     }
00481 
00482     /* Flush as much pending output as possible */
00483     if (s->pending != 0) {
00484         flush_pending(strm);
00485         if (strm->avail_out == 0) {
00486            /* Since avail_out is 0, deflate will be called again with
00487             * more output space, but possibly with both pending and
00488             * avail_in equal to zero. There won't be anything to do,
00489             * but this is not an error situation so make sure we
00490             * return OK instead of BUF_ERROR at next call of deflate:
00491              */
00492            s->last_flush = -1;
00493            return Z_OK;
00494        }
00495 
00496     /* Make sure there is something to do and avoid duplicate consecutive
00497      * flushes. For repeated and useless calls with Z_FINISH, we keep
00498      * returning Z_STREAM_END instead of Z_BUFF_ERROR.
00499      */
00500     } else if (strm->avail_in == 0 && flush <= old_flush &&
00501               flush != Z_FINISH) {
00502         ERR_RETURN(strm, Z_BUF_ERROR);
00503     }
00504 
00505     /* User must not provide more input after the first FINISH: */
00506     if (s->status == FINISH_STATE && strm->avail_in != 0) {
00507         ERR_RETURN(strm, Z_BUF_ERROR);
00508     }
00509 
00510     /* Start a new block or continue the current one.
00511      */
00512     if (strm->avail_in != 0 || s->lookahead != 0 ||
00513         (flush != Z_NO_FLUSH && s->status != FINISH_STATE)) {
00514         block_state bstate;
00515 
00516        bstate = (*(configuration_table[s->level].func))(s, flush);
00517 
00518         if (bstate == finish_started || bstate == finish_done) {
00519             s->status = FINISH_STATE;
00520         }
00521         if (bstate == need_more || bstate == finish_started) {
00522            if (strm->avail_out == 0) {
00523                s->last_flush = -1; /* avoid BUF_ERROR next call, see above */
00524            }
00525            return Z_OK;
00526            /* If flush != Z_NO_FLUSH && avail_out == 0, the next call
00527             * of deflate should use the same flush parameter to make sure
00528             * that the flush is complete. So we don't have to output an
00529             * empty block here, this will be done at next call. This also
00530             * ensures that for a very small output buffer, we emit at most
00531             * one empty block.
00532             */
00533        }
00534         if (bstate == block_done) {
00535             if (flush == Z_PARTIAL_FLUSH) {
00536                 _tr_align(s);
00537             } else { /* FULL_FLUSH or SYNC_FLUSH */
00538                 _tr_stored_block(s, (char*)0, 0L, 0);
00539                 /* For a full flush, this empty block will be recognized
00540                  * as a special marker by inflate_sync().
00541                  */
00542                 if (flush == Z_FULL_FLUSH) {
00543                     CLEAR_HASH(s);             /* forget history */
00544                 }
00545             }
00546             flush_pending(strm);
00547            if (strm->avail_out == 0) {
00548              s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */
00549              return Z_OK;
00550            }
00551         }
00552     }
00553     Assert(strm->avail_out > 0, "bug2");
00554 
00555     if (flush != Z_FINISH) return Z_OK;
00556     if (s->noheader) return Z_STREAM_END;
00557 
00558     /* Write the zlib trailer (adler32) */
00559     putShortMSB(s, (uInt)(strm->adler >> 16));
00560     putShortMSB(s, (uInt)(strm->adler & 0xffff));
00561     flush_pending(strm);
00562     /* If avail_out is zero, the application will call deflate again
00563      * to flush the rest.
00564      */
00565     s->noheader = -1; /* write the trailer only once! */
00566     return s->pending != 0 ? Z_OK : Z_STREAM_END;
00567 }
00568 
00569 /* ========================================================================= */
00570 int ZEXPORT deflateEnd (strm)
00571     z_streamp strm;
00572 {
00573     int status;
00574 
00575     if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
00576 
00577     status = strm->state->status;
00578     if (status != INIT_STATE && status != BUSY_STATE &&
00579        status != FINISH_STATE) {
00580       return Z_STREAM_ERROR;
00581     }
00582 
00583     /* Deallocate in reverse order of allocations: */
00584     TRY_FREE(strm, strm->state->pending_buf);
00585     TRY_FREE(strm, strm->state->head);
00586     TRY_FREE(strm, strm->state->prev);
00587     TRY_FREE(strm, strm->state->window);
00588 
00589     ZFREE(strm, strm->state);
00590     strm->state = Z_NULL;
00591 
00592     return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK;
00593 }
00594 
00595 /* =========================================================================
00596  * Copy the source state to the destination state.
00597  * To simplify the source, this is not supported for 16-bit MSDOS (which
00598  * doesn't have enough memory anyway to duplicate compression states).
00599  */
00600 int ZEXPORT deflateCopy (dest, source)
00601     z_streamp dest;
00602     z_streamp source;
00603 {
00604 #ifdef MAXSEG_64K
00605     return Z_STREAM_ERROR;
00606 #else
00607     deflate_state *ds;
00608     deflate_state *ss;
00609     ushf *overlay;
00610 
00611 
00612     if (source == Z_NULL || dest == Z_NULL || source->state == Z_NULL) {
00613         return Z_STREAM_ERROR;
00614     }
00615 
00616     ss = source->state;
00617 
00618     *dest = *source;
00619 
00620     ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state));
00621     if (ds == Z_NULL) return Z_MEM_ERROR;
00622     dest->state = (struct internal_state FAR *) ds;
00623     *ds = *ss;
00624     ds->strm = dest;
00625 
00626     ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte));
00627     ds->prev   = (Posf *)  ZALLOC(dest, ds->w_size, sizeof(Pos));
00628     ds->head   = (Posf *)  ZALLOC(dest, ds->hash_size, sizeof(Pos));
00629     overlay = (ushf *) ZALLOC(dest, ds->lit_bufsize, sizeof(ush)+2);
00630     ds->pending_buf = (uchf *) overlay;
00631 
00632     if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL ||
00633         ds->pending_buf == Z_NULL) {
00634         deflateEnd (dest);
00635         return Z_MEM_ERROR;
00636     }
00637     /* following zmemcpy do not work for 16-bit MSDOS */
00638     zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte));
00639     zmemcpy(ds->prev, ss->prev, ds->w_size * sizeof(Pos));
00640     zmemcpy(ds->head, ss->head, ds->hash_size * sizeof(Pos));
00641     zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size);
00642 
00643     ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf);
00644     ds->d_buf = overlay + ds->lit_bufsize/sizeof(ush);
00645     ds->l_buf = ds->pending_buf + (1+sizeof(ush))*ds->lit_bufsize;
00646 
00647     ds->l_desc.dyn_tree = ds->dyn_ltree;
00648     ds->d_desc.dyn_tree = ds->dyn_dtree;
00649     ds->bl_desc.dyn_tree = ds->bl_tree;
00650 
00651     return Z_OK;
00652 #endif
00653 }
00654 
00655 /* ===========================================================================
00656  * Read a new buffer from the current input stream, update the adler32
00657  * and total number of bytes read.  All deflate() input goes through
00658  * this function so some applications may wish to modify it to avoid
00659  * allocating a large strm->next_in buffer and copying from it.
00660  * (See also flush_pending()).
00661  */
00662 local int read_buf(strm, buf, size)
00663     z_streamp strm;
00664     Bytef *buf;
00665     unsigned size;
00666 {
00667     unsigned len = strm->avail_in;
00668 
00669     if (len > size) len = size;
00670     if (len == 0) return 0;
00671 
00672     strm->avail_in  -= len;
00673 
00674     if (!strm->state->noheader) {
00675         strm->adler = adler32(strm->adler, strm->next_in, len);
00676     }
00677     zmemcpy(buf, strm->next_in, len);
00678     strm->next_in  += len;
00679     strm->total_in += len;
00680 
00681     return (int)len;
00682 }
00683 
00684 /* ===========================================================================
00685  * Initialize the "longest match" routines for a new zlib stream
00686  */
00687 local void lm_init (s)
00688     deflate_state *s;
00689 {
00690     s->window_size = (ulg)2L*s->w_size;
00691 
00692     CLEAR_HASH(s);
00693 
00694     /* Set the default configuration parameters:
00695      */
00696     s->max_lazy_match   = configuration_table[s->level].max_lazy;
00697     s->good_match       = configuration_table[s->level].good_length;
00698     s->nice_match       = configuration_table[s->level].nice_length;
00699     s->max_chain_length = configuration_table[s->level].max_chain;
00700 
00701     s->strstart = 0;
00702     s->block_start = 0L;
00703     s->lookahead = 0;
00704     s->match_length = s->prev_length = MIN_MATCH-1;
00705     s->match_available = 0;
00706     s->ins_h = 0;
00707 #ifdef ASMV
00708     match_init(); /* initialize the asm code */
00709 #endif
00710 }
00711 
00712 /* ===========================================================================
00713  * Set match_start to the longest match starting at the given string and
00714  * return its length. Matches shorter or equal to prev_length are discarded,
00715  * in which case the result is equal to prev_length and match_start is
00716  * garbage.
00717  * IN assertions: cur_match is the head of the hash chain for the current
00718  *   string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
00719  * OUT assertion: the match length is not greater than s->lookahead.
00720  */
00721 #ifndef ASMV
00722 /* For 80x86 and 680x0, an optimized version will be provided in match.asm or
00723  * match.S. The code will be functionally equivalent.
00724  */
00725 #ifndef FASTEST
00726 local uInt longest_match(s, cur_match)
00727     deflate_state *s;
00728     IPos cur_match;                             /* current match */
00729 {
00730     unsigned chain_length = s->max_chain_length;/* max hash chain length */
00731     register Bytef *scan = s->window + s->strstart; /* current string */
00732     register Bytef *match;                       /* matched string */
00733     register int len;                           /* length of current match */
00734     int best_len = s->prev_length;              /* best match length so far */
00735     int nice_match = s->nice_match;             /* stop if match long enough */
00736     IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
00737         s->strstart - (IPos)MAX_DIST(s) : NIL;
00738     /* Stop when cur_match becomes <= limit. To simplify the code,
00739      * we prevent matches with the string of window index 0.
00740      */
00741     Posf *prev = s->prev;
00742     uInt wmask = s->w_mask;
00743 
00744 #ifdef UNALIGNED_OK
00745     /* Compare two bytes at a time. Note: this is not always beneficial.
00746      * Try with and without -DUNALIGNED_OK to check.
00747      */
00748     register Bytef *strend = s->window + s->strstart + MAX_MATCH - 1;
00749     register ush scan_start = *(ushf*)scan;
00750     register ush scan_end   = *(ushf*)(scan+best_len-1);
00751 #else
00752     register Bytef *strend = s->window + s->strstart + MAX_MATCH;
00753     register Byte scan_end1  = scan[best_len-1];
00754     register Byte scan_end   = scan[best_len];
00755 #endif
00756 
00757     /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
00758      * It is easy to get rid of this optimization if necessary.
00759      */
00760     Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
00761 
00762     /* Do not waste too much time if we already have a good match: */
00763     if (s->prev_length >= s->good_match) {
00764         chain_length >>= 2;
00765     }
00766     /* Do not look for matches beyond the end of the input. This is necessary
00767      * to make deflate deterministic.
00768      */
00769     if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead;
00770 
00771     Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
00772 
00773     do {
00774         Assert(cur_match < s->strstart, "no future");
00775         match = s->window + cur_match;
00776 
00777         /* Skip to next match if the match length cannot increase
00778          * or if the match length is less than 2:
00779          */
00780 #if (defined(UNALIGNED_OK) && MAX_MATCH == 258)
00781         /* This code assumes sizeof(unsigned short) == 2. Do not use
00782          * UNALIGNED_OK if your compiler uses a different size.
00783          */
00784         if (*(ushf*)(match+best_len-1) != scan_end ||
00785             *(ushf*)match != scan_start) continue;
00786 
00787         /* It is not necessary to compare scan[2] and match[2] since they are
00788          * always equal when the other bytes match, given that the hash keys
00789          * are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at
00790          * strstart+3, +5, ... up to strstart+257. We check for insufficient
00791          * lookahead only every 4th comparison; the 128th check will be made
00792          * at strstart+257. If MAX_MATCH-2 is not a multiple of 8, it is
00793          * necessary to put more guard bytes at the end of the window, or
00794          * to check more often for insufficient lookahead.
00795          */
00796         Assert(scan[2] == match[2], "scan[2]?");
00797         scan++, match++;
00798         do {
00799         } while (*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
00800                  *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
00801                  *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
00802                  *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
00803                  scan < strend);
00804         /* The funny "do {}" generates better code on most compilers */
00805 
00806         /* Here, scan <= window+strstart+257 */
00807         Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
00808         if (*scan == *match) scan++;
00809 
00810         len = (MAX_MATCH - 1) - (int)(strend-scan);
00811         scan = strend - (MAX_MATCH-1);
00812 
00813 #else /* UNALIGNED_OK */
00814 
00815         if (match[best_len]   != scan_end  ||
00816             match[best_len-1] != scan_end1 ||
00817             *match            != *scan     ||
00818             *++match          != scan[1])      continue;
00819 
00820         /* The check at best_len-1 can be removed because it will be made
00821          * again later. (This heuristic is not always a win.)
00822          * It is not necessary to compare scan[2] and match[2] since they
00823          * are always equal when the other bytes match, given that
00824          * the hash keys are equal and that HASH_BITS >= 8.
00825          */
00826         scan += 2, match++;
00827         Assert(*scan == *match, "match[2]?");
00828 
00829         /* We check for insufficient lookahead only every 8th comparison;
00830          * the 256th check will be made at strstart+258.
00831          */
00832         do {
00833         } while (*++scan == *++match && *++scan == *++match &&
00834                  *++scan == *++match && *++scan == *++match &&
00835                  *++scan == *++match && *++scan == *++match &&
00836                  *++scan == *++match && *++scan == *++match &&
00837                  scan < strend);
00838 
00839         Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
00840 
00841         len = MAX_MATCH - (int)(strend - scan);
00842         scan = strend - MAX_MATCH;
00843 
00844 #endif /* UNALIGNED_OK */
00845 
00846         if (len > best_len) {
00847             s->match_start = cur_match;
00848             best_len = len;
00849             if (len >= nice_match) break;
00850 #ifdef UNALIGNED_OK
00851             scan_end = *(ushf*)(scan+best_len-1);
00852 #else
00853             scan_end1  = scan[best_len-1];
00854             scan_end   = scan[best_len];
00855 #endif
00856         }
00857     } while ((cur_match = prev[cur_match & wmask]) > limit
00858              && --chain_length != 0);
00859 
00860     if ((uInt)best_len <= s->lookahead) return (uInt)best_len;
00861     return s->lookahead;
00862 }
00863 
00864 #else /* FASTEST */
00865 /* ---------------------------------------------------------------------------
00866  * Optimized version for level == 1 only
00867  */
00868 local uInt longest_match(s, cur_match)
00869     deflate_state *s;
00870     IPos cur_match;                             /* current match */
00871 {
00872     register Bytef *scan = s->window + s->strstart; /* current string */
00873     register Bytef *match;                       /* matched string */
00874     register int len;                           /* length of current match */
00875     register Bytef *strend = s->window + s->strstart + MAX_MATCH;
00876 
00877     /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
00878      * It is easy to get rid of this optimization if necessary.
00879      */
00880     Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
00881 
00882     Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
00883 
00884     Assert(cur_match < s->strstart, "no future");
00885 
00886     match = s->window + cur_match;
00887 
00888     /* Return failure if the match length is less than 2:
00889      */
00890     if (match[0] != scan[0] || match[1] != scan[1]) return MIN_MATCH-1;
00891 
00892     /* The check at best_len-1 can be removed because it will be made
00893      * again later. (This heuristic is not always a win.)
00894      * It is not necessary to compare scan[2] and match[2] since they
00895      * are always equal when the other bytes match, given that
00896      * the hash keys are equal and that HASH_BITS >= 8.
00897      */
00898     scan += 2, match += 2;
00899     Assert(*scan == *match, "match[2]?");
00900 
00901     /* We check for insufficient lookahead only every 8th comparison;
00902      * the 256th check will be made at strstart+258.
00903      */
00904     do {
00905     } while (*++scan == *++match && *++scan == *++match &&
00906             *++scan == *++match && *++scan == *++match &&
00907             *++scan == *++match && *++scan == *++match &&
00908             *++scan == *++match && *++scan == *++match &&
00909             scan < strend);
00910 
00911     Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
00912 
00913     len = MAX_MATCH - (int)(strend - scan);
00914 
00915     if (len < MIN_MATCH) return MIN_MATCH - 1;
00916 
00917     s->match_start = cur_match;
00918     return len <= s->lookahead ? len : s->lookahead;
00919 }
00920 #endif /* FASTEST */
00921 #endif /* ASMV */
00922 
00923 #ifdef DEBUG
00924 /* ===========================================================================
00925  * Check that the match at match_start is indeed a match.
00926  */
00927 local void check_match(s, start, match, length)
00928     deflate_state *s;
00929     IPos start, match;
00930     int length;
00931 {
00932     /* check that the match is indeed a match */
00933     if (zmemcmp(s->window + match,
00934                 s->window + start, length) != EQUAL) {
00935         fprintf(stderr, " start %u, match %u, length %d\n",
00936               start, match, length);
00937         do {
00938            fprintf(stderr, "%c%c", s->window[match++], s->window[start++]);
00939        } while (--length != 0);
00940         z_error("invalid match");
00941     }
00942     if (z_verbose > 1) {
00943         fprintf(stderr,"\\[%d,%d]", start-match, length);
00944         do { putc(s->window[start++], stderr); } while (--length != 0);
00945     }
00946 }
00947 #else
00948 #  define check_match(s, start, match, length)
00949 #endif
00950 
00951 /* ===========================================================================
00952  * Fill the window when the lookahead becomes insufficient.
00953  * Updates strstart and lookahead.
00954  *
00955  * IN assertion: lookahead < MIN_LOOKAHEAD
00956  * OUT assertions: strstart <= window_size-MIN_LOOKAHEAD
00957  *    At least one byte has been read, or avail_in == 0; reads are
00958  *    performed for at least two bytes (required for the zip translate_eol
00959  *    option -- not supported here).
00960  */
00961 local void fill_window(s)
00962     deflate_state *s;
00963 {
00964     register unsigned n, m;
00965     register Posf *p;
00966     unsigned more;    /* Amount of free space at the end of the window. */
00967     uInt wsize = s->w_size;
00968 
00969     do {
00970         more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart);
00971 
00972         /* Deal with !@#$% 64K limit: */
00973         if (more == 0 && s->strstart == 0 && s->lookahead == 0) {
00974             more = wsize;
00975 
00976         } else if (more == (unsigned)(-1)) {
00977             /* Very unlikely, but possible on 16 bit machine if strstart == 0
00978              * and lookahead == 1 (input done one byte at time)
00979              */
00980             more--;
00981 
00982         /* If the window is almost full and there is insufficient lookahead,
00983          * move the upper half to the lower one to make room in the upper half.
00984          */
00985         } else if (s->strstart >= wsize+MAX_DIST(s)) {
00986 
00987             zmemcpy(s->window, s->window+wsize, (unsigned)wsize);
00988             s->match_start -= wsize;
00989             s->strstart    -= wsize; /* we now have strstart >= MAX_DIST */
00990             s->block_start -= (long) wsize;
00991 
00992             /* Slide the hash table (could be avoided with 32 bit values
00993                at the expense of memory usage). We slide even when level == 0
00994                to keep the hash table consistent if we switch back to level > 0
00995                later. (Using level 0 permanently is not an optimal usage of
00996                zlib, so we don't care about this pathological case.)
00997              */
00998            n = s->hash_size;
00999            p = &s->head[n];
01000            do {
01001               m = *--p;
01002               *p = (Pos)(m >= wsize ? m-wsize : NIL);
01003            } while (--n);
01004 
01005            n = wsize;
01006 #ifndef FASTEST
01007            p = &s->prev[n];
01008            do {
01009               m = *--p;
01010               *p = (Pos)(m >= wsize ? m-wsize : NIL);
01011               /* If n is not on any hash chain, prev[n] is garbage but
01012                * its value will never be used.
01013                */
01014            } while (--n);
01015 #endif
01016             more += wsize;
01017         }
01018         if (s->strm->avail_in == 0) return;
01019 
01020         /* If there was no sliding:
01021          *    strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 &&
01022          *    more == window_size - lookahead - strstart
01023          * => more >= window_size - (MIN_LOOKAHEAD-1 + WSIZE + MAX_DIST-1)
01024          * => more >= window_size - 2*WSIZE + 2
01025          * In the BIG_MEM or MMAP case (not yet supported),
01026          *   window_size == input_size + MIN_LOOKAHEAD  &&
01027          *   strstart + s->lookahead <= input_size => more >= MIN_LOOKAHEAD.
01028          * Otherwise, window_size == 2*WSIZE so more >= 2.
01029          * If there was sliding, more >= WSIZE. So in all cases, more >= 2.
01030          */
01031         Assert(more >= 2, "more < 2");
01032 
01033         n = read_buf(s->strm, s->window + s->strstart + s->lookahead, more);
01034         s->lookahead += n;
01035 
01036         /* Initialize the hash value now that we have some input: */
01037         if (s->lookahead >= MIN_MATCH) {
01038             s->ins_h = s->window[s->strstart];
01039             UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
01040 #if MIN_MATCH != 3
01041             Call UPDATE_HASH() MIN_MATCH-3 more times
01042 #endif
01043         }
01044         /* If the whole input has less than MIN_MATCH bytes, ins_h is garbage,
01045          * but this is not important since only literal bytes will be emitted.
01046          */
01047 
01048     } while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0);
01049 }
01050 
01051 /* ===========================================================================
01052  * Flush the current block, with given end-of-file flag.
01053  * IN assertion: strstart is set to the end of the current match.
01054  */
01055 #define FLUSH_BLOCK_ONLY(s, eof) { \
01056    _tr_flush_block(s, (s->block_start >= 0L ? \
01057                    (charf *)&s->window[(unsigned)s->block_start] : \
01058                    (charf *)Z_NULL), \
01059               (ulg)((long)s->strstart - s->block_start), \
01060               (eof)); \
01061    s->block_start = s->strstart; \
01062    flush_pending(s->strm); \
01063    Tracev((stderr,"[FLUSH]")); \
01064 }
01065 
01066 /* Same but force premature exit if necessary. */
01067 #define FLUSH_BLOCK(s, eof) { \
01068    FLUSH_BLOCK_ONLY(s, eof); \
01069    if (s->strm->avail_out == 0) return (eof) ? finish_started : need_more; \
01070 }
01071 
01072 /* ===========================================================================
01073  * Copy without compression as much as possible from the input stream, return
01074  * the current block state.
01075  * This function does not insert new strings in the dictionary since
01076  * uncompressible data is probably not useful. This function is used
01077  * only for the level=0 compression option.
01078  * NOTE: this function should be optimized to avoid extra copying from
01079  * window to pending_buf.
01080  */
01081 local block_state deflate_stored(s, flush)
01082     deflate_state *s;
01083     int flush;
01084 {
01085     /* Stored blocks are limited to 0xffff bytes, pending_buf is limited
01086      * to pending_buf_size, and each stored block has a 5 byte header:
01087      */
01088     ulg max_block_size = 0xffff;
01089     ulg max_start;
01090 
01091     if (max_block_size > s->pending_buf_size - 5) {
01092         max_block_size = s->pending_buf_size - 5;
01093     }
01094 
01095     /* Copy as much as possible from input to output: */
01096     for (;;) {
01097         /* Fill the window as much as possible: */
01098         if (s->lookahead <= 1) {
01099 
01100             Assert(s->strstart < s->w_size+MAX_DIST(s) ||
01101                  s->block_start >= (long)s->w_size, "slide too late");
01102 
01103             fill_window(s);
01104             if (s->lookahead == 0 && flush == Z_NO_FLUSH) return need_more;
01105 
01106             if (s->lookahead == 0) break; /* flush the current block */
01107         }
01108        Assert(s->block_start >= 0L, "block gone");
01109 
01110        s->strstart += s->lookahead;
01111        s->lookahead = 0;
01112 
01113        /* Emit a stored block if pending_buf will be full: */
01114        max_start = s->block_start + max_block_size;
01115         if (s->strstart == 0 || (ulg)s->strstart >= max_start) {
01116            /* strstart == 0 is possible when wraparound on 16-bit machine */
01117            s->lookahead = (uInt)(s->strstart - max_start);
01118            s->strstart = (uInt)max_start;
01119             FLUSH_BLOCK(s, 0);
01120        }
01121        /* Flush if we may have to slide, otherwise block_start may become
01122          * negative and the data will be gone:
01123          */
01124         if (s->strstart - (uInt)s->block_start >= MAX_DIST(s)) {
01125             FLUSH_BLOCK(s, 0);
01126        }
01127     }
01128     FLUSH_BLOCK(s, flush == Z_FINISH);
01129     return flush == Z_FINISH ? finish_done : block_done;
01130 }
01131 
01132 /* ===========================================================================
01133  * Compress as much as possible from the input stream, return the current
01134  * block state.
01135  * This function does not perform lazy evaluation of matches and inserts
01136  * new strings in the dictionary only for unmatched strings or for short
01137  * matches. It is used only for the fast compression options.
01138  */
01139 local block_state deflate_fast(s, flush)
01140     deflate_state *s;
01141     int flush;
01142 {
01143     IPos hash_head = NIL; /* head of the hash chain */
01144     int bflush;           /* set if current block must be flushed */
01145 
01146     for (;;) {
01147         /* Make sure that we always have enough lookahead, except
01148          * at the end of the input file. We need MAX_MATCH bytes
01149          * for the next match, plus MIN_MATCH bytes to insert the
01150          * string following the next match.
01151          */
01152         if (s->lookahead < MIN_LOOKAHEAD) {
01153             fill_window(s);
01154             if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
01155                return need_more;
01156            }
01157             if (s->lookahead == 0) break; /* flush the current block */
01158         }
01159 
01160         /* Insert the string window[strstart .. strstart+2] in the
01161          * dictionary, and set hash_head to the head of the hash chain:
01162          */
01163         if (s->lookahead >= MIN_MATCH) {
01164             INSERT_STRING(s, s->strstart, hash_head);
01165         }
01166 
01167         /* Find the longest match, discarding those <= prev_length.
01168          * At this point we have always match_length < MIN_MATCH
01169          */
01170         if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) {
01171             /* To simplify the code, we prevent matches with the string
01172              * of window index 0 (in particular we have to avoid a match
01173              * of the string with itself at the start of the input file).
01174              */
01175             if (s->strategy != Z_HUFFMAN_ONLY) {
01176                 s->match_length = longest_match (s, hash_head);
01177             }
01178             /* longest_match() sets match_start */
01179         }
01180         if (s->match_length >= MIN_MATCH) {
01181             check_match(s, s->strstart, s->match_start, s->match_length);
01182 
01183             _tr_tally_dist(s, s->strstart - s->match_start,
01184                            s->match_length - MIN_MATCH, bflush);
01185 
01186             s->lookahead -= s->match_length;
01187 
01188             /* Insert new strings in the hash table only if the match length
01189              * is not too large. This saves time but degrades compression.
01190              */
01191 #ifndef FASTEST
01192             if (s->match_length <= s->max_insert_length &&
01193                 s->lookahead >= MIN_MATCH) {
01194                 s->match_length--; /* string at strstart already in hash table */
01195                 do {
01196                     s->strstart++;
01197                     INSERT_STRING(s, s->strstart, hash_head);
01198                     /* strstart never exceeds WSIZE-MAX_MATCH, so there are
01199                      * always MIN_MATCH bytes ahead.
01200                      */
01201                 } while (--s->match_length != 0);
01202                 s->strstart++; 
01203             } else
01204 #endif
01205            {
01206                 s->strstart += s->match_length;
01207                 s->match_length = 0;
01208                 s->ins_h = s->window[s->strstart];
01209                 UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
01210 #if MIN_MATCH != 3
01211                 Call UPDATE_HASH() MIN_MATCH-3 more times
01212 #endif
01213                 /* If lookahead < MIN_MATCH, ins_h is garbage, but it does not
01214                  * matter since it will be recomputed at next deflate call.
01215                  */
01216             }
01217         } else {
01218             /* No match, output a literal byte */
01219             Tracevv((stderr,"%c", s->window[s->strstart]));
01220             _tr_tally_lit (s, s->window[s->strstart], bflush);
01221             s->lookahead--;
01222             s->strstart++; 
01223         }
01224         if (bflush) FLUSH_BLOCK(s, 0);
01225     }
01226     FLUSH_BLOCK(s, flush == Z_FINISH);
01227     return flush == Z_FINISH ? finish_done : block_done;
01228 }
01229 
01230 /* ===========================================================================
01231  * Same as above, but achieves better compression. We use a lazy
01232  * evaluation for matches: a match is finally adopted only if there is
01233  * no better match at the next window position.
01234  */
01235 local block_state deflate_slow(s, flush)
01236     deflate_state *s;
01237     int flush;
01238 {
01239     IPos hash_head = NIL;    /* head of hash chain */
01240     int bflush;              /* set if current block must be flushed */
01241 
01242     /* Process the input block. */
01243     for (;;) {
01244         /* Make sure that we always have enough lookahead, except
01245          * at the end of the input file. We need MAX_MATCH bytes
01246          * for the next match, plus MIN_MATCH bytes to insert the
01247          * string following the next match.
01248          */
01249         if (s->lookahead < MIN_LOOKAHEAD) {
01250             fill_window(s);
01251             if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
01252                return need_more;
01253            }
01254             if (s->lookahead == 0) break; /* flush the current block */
01255         }
01256 
01257         /* Insert the string window[strstart .. strstart+2] in the
01258          * dictionary, and set hash_head to the head of the hash chain:
01259          */
01260         if (s->lookahead >= MIN_MATCH) {
01261             INSERT_STRING(s, s->strstart, hash_head);
01262         }
01263 
01264         /* Find the longest match, discarding those <= prev_length.
01265          */
01266         s->prev_length = s->match_length, s->prev_match = s->match_start;
01267         s->match_length = MIN_MATCH-1;
01268 
01269         if (hash_head != NIL && s->prev_length < s->max_lazy_match &&
01270             s->strstart - hash_head <= MAX_DIST(s)) {
01271             /* To simplify the code, we prevent matches with the string
01272              * of window index 0 (in particular we have to avoid a match
01273              * of the string with itself at the start of the input file).
01274              */
01275             if (s->strategy != Z_HUFFMAN_ONLY) {
01276                 s->match_length = longest_match (s, hash_head);
01277             }
01278             /* longest_match() sets match_start */
01279 
01280             if (s->match_length <= 5 && (s->strategy == Z_FILTERED ||
01281                  (s->match_length == MIN_MATCH &&
01282                   s->strstart - s->match_start > TOO_FAR))) {
01283 
01284                 /* If prev_match is also MIN_MATCH, match_start is garbage
01285                  * but we will ignore the current match anyway.
01286                  */
01287                 s->match_length = MIN_MATCH-1;
01288             }
01289         }
01290         /* If there was a match at the previous step and the current
01291          * match is not better, output the previous match:
01292          */
01293         if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) {
01294             uInt max_insert = s->strstart + s->lookahead - MIN_MATCH;
01295             /* Do not insert strings in hash table beyond this. */
01296 
01297             check_match(s, s->strstart-1, s->prev_match, s->prev_length);
01298 
01299             _tr_tally_dist(s, s->strstart -1 - s->prev_match,
01300                         s->prev_length - MIN_MATCH, bflush);
01301 
01302             /* Insert in hash table all strings up to the end of the match.
01303              * strstart-1 and strstart are already inserted. If there is not
01304              * enough lookahead, the last two strings are not inserted in
01305              * the hash table.
01306              */
01307             s->lookahead -= s->prev_length-1;
01308             s->prev_length -= 2;
01309             do {
01310                 if (++s->strstart <= max_insert) {
01311                     INSERT_STRING(s, s->strstart, hash_head);
01312                 }
01313             } while (--s->prev_length != 0);
01314             s->match_available = 0;
01315             s->match_length = MIN_MATCH-1;
01316             s->strstart++;
01317 
01318             if (bflush) FLUSH_BLOCK(s, 0);
01319 
01320         } else if (s->match_available) {
01321             /* If there was no match at the previous position, output a
01322              * single literal. If there was a match but the current match
01323              * is longer, truncate the previous match to a single literal.
01324              */
01325             Tracevv((stderr,"%c", s->window[s->strstart-1]));
01326            _tr_tally_lit(s, s->window[s->strstart-1], bflush);
01327            if (bflush) {
01328                 FLUSH_BLOCK_ONLY(s, 0);
01329             }
01330             s->strstart++;
01331             s->lookahead--;
01332             if (s->strm->avail_out == 0) return need_more;
01333         } else {
01334             /* There is no previous match to compare with, wait for
01335              * the next step to decide.
01336              */
01337             s->match_available = 1;
01338             s->strstart++;
01339             s->lookahead--;
01340         }
01341     }
01342     Assert (flush != Z_NO_FLUSH, "no flush?");
01343     if (s->match_available) {
01344         Tracevv((stderr,"%c", s->window[s->strstart-1]));
01345         _tr_tally_lit(s, s->window[s->strstart-1], bflush);
01346         s->match_available = 0;
01347     }
01348     FLUSH_BLOCK(s, flush == Z_FINISH);
01349     return flush == Z_FINISH ? finish_done : block_done;
01350 }