1 /*
  2  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  3  *
  4  * This code is free software; you can redistribute it and/or modify it
  5  * under the terms of the GNU General Public License version 2 only, as
  6  * published by the Free Software Foundation.  Oracle designates this
  7  * particular file as subject to the "Classpath" exception as provided
  8  * by Oracle in the LICENSE file that accompanied this code.
  9  *
 10  * This code is distributed in the hope that it will be useful, but WITHOUT
 11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 13  * version 2 for more details (a copy is included in the LICENSE file that
 14  * accompanied this code).
 15  *
 16  * You should have received a copy of the GNU General Public License version
 17  * 2 along with this work; if not, write to the Free Software Foundation,
 18  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
 19  *
 20  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 21  * or visit www.oracle.com if you need additional information or have any
 22  * questions.
 23  */
 24 
 25 /* deflate.h -- internal compression state
 26  * Copyright (C) 1995-2024 Jean-loup Gailly
 27  * For conditions of distribution and use, see copyright notice in zlib.h
 28  */
 29 
 30 /* WARNING: this file should *not* be used by applications. It is
 31    part of the implementation of the compression library and is
 32    subject to change. Applications should only use zlib.h.
 33  */
 34 
 35 /* @(#) $Id$ */
 36 
 37 #ifndef DEFLATE_H
 38 #define DEFLATE_H
 39 
 40 #include "zutil.h"
 41 
 42 /* define NO_GZIP when compiling if you want to disable gzip header and
 43    trailer creation by deflate().  NO_GZIP would be used to avoid linking in
 44    the crc code when it is not needed.  For shared libraries, gzip encoding
 45    should be left enabled. */
 46 #ifndef NO_GZIP
 47 #  define GZIP
 48 #endif
 49 
 50 /* define LIT_MEM to slightly increase the speed of deflate (order 1% to 2%) at
 51    the cost of a larger memory footprint */
 52 /* #define LIT_MEM */
 53 
 54 /* ===========================================================================
 55  * Internal compression state.
 56  */
 57 
 58 #define LENGTH_CODES 29
 59 /* number of length codes, not counting the special END_BLOCK code */
 60 
 61 #define LITERALS  256
 62 /* number of literal bytes 0..255 */
 63 
 64 #define L_CODES (LITERALS+1+LENGTH_CODES)
 65 /* number of Literal or Length codes, including the END_BLOCK code */
 66 
 67 #define D_CODES   30
 68 /* number of distance codes */
 69 
 70 #define BL_CODES  19
 71 /* number of codes used to transfer the bit lengths */
 72 
 73 #define HEAP_SIZE (2*L_CODES+1)
 74 /* maximum heap size */
 75 
 76 #define MAX_BITS 15
 77 /* All codes must not exceed MAX_BITS bits */
 78 
 79 #define Buf_size 16
 80 /* size of bit buffer in bi_buf */
 81 
 82 #define INIT_STATE    42    /* zlib header -> BUSY_STATE */
 83 #ifdef GZIP
 84 #  define GZIP_STATE  57    /* gzip header -> BUSY_STATE | EXTRA_STATE */
 85 #endif
 86 #define EXTRA_STATE   69    /* gzip extra block -> NAME_STATE */
 87 #define NAME_STATE    73    /* gzip file name -> COMMENT_STATE */
 88 #define COMMENT_STATE 91    /* gzip comment -> HCRC_STATE */
 89 #define HCRC_STATE   103    /* gzip header CRC -> BUSY_STATE */
 90 #define BUSY_STATE   113    /* deflate -> FINISH_STATE */
 91 #define FINISH_STATE 666    /* stream complete */
 92 /* Stream status */
 93 
 94 
 95 /* Data structure describing a single value and its code string. */
 96 typedef struct ct_data_s {
 97     union {
 98         ush  freq;       /* frequency count */
 99         ush  code;       /* bit string */
100     } fc;
101     union {
102         ush  dad;        /* father node in Huffman tree */
103         ush  len;        /* length of bit string */
104     } dl;
105 } FAR ct_data;
106 
107 #define Freq fc.freq
108 #define Code fc.code
109 #define Dad  dl.dad
110 #define Len  dl.len
111 
112 typedef struct static_tree_desc_s  static_tree_desc;
113 
114 typedef struct tree_desc_s {
115     ct_data *dyn_tree;           /* the dynamic tree */
116     int     max_code;            /* largest code with non zero frequency */
117     const static_tree_desc *stat_desc;  /* the corresponding static tree */
118 } FAR tree_desc;
119 
120 typedef ush Pos;
121 typedef Pos FAR Posf;
122 typedef unsigned IPos;
123 
124 /* A Pos is an index in the character window. We use short instead of int to
125  * save space in the various tables. IPos is used only for parameter passing.
126  */
127 
128 typedef struct internal_state {
129     z_streamp strm;      /* pointer back to this zlib stream */
130     int   status;        /* as the name implies */
131     Bytef *pending_buf;  /* output still pending */
132     ulg   pending_buf_size; /* size of pending_buf */
133     Bytef *pending_out;  /* next pending byte to output to the stream */
134     ulg   pending;       /* nb of bytes in the pending buffer */
135     int   wrap;          /* bit 0 true for zlib, bit 1 true for gzip */
136     gz_headerp  gzhead;  /* gzip header information to write */
137     ulg   gzindex;       /* where in extra, name, or comment */
138     Byte  method;        /* can only be DEFLATED */
139     int   last_flush;    /* value of flush param for previous deflate call */
140 
141                 /* used by deflate.c: */
142 
143     uInt  w_size;        /* LZ77 window size (32K by default) */
144     uInt  w_bits;        /* log2(w_size)  (8..16) */
145     uInt  w_mask;        /* w_size - 1 */
146 
147     Bytef *window;
148     /* Sliding window. Input bytes are read into the second half of the window,
149      * and move to the first half later to keep a dictionary of at least wSize
150      * bytes. With this organization, matches are limited to a distance of
151      * wSize-MAX_MATCH bytes, but this ensures that IO is always
152      * performed with a length multiple of the block size. Also, it limits
153      * the window size to 64K, which is quite useful on MSDOS.
154      * To do: use the user input buffer as sliding window.
155      */
156 
157     ulg window_size;
158     /* Actual size of window: 2*wSize, except when the user input buffer
159      * is directly used as sliding window.
160      */
161 
162     Posf *prev;
163     /* Link to older string with same hash index. To limit the size of this
164      * array to 64K, this link is maintained only for the last 32K strings.
165      * An index in this array is thus a window index modulo 32K.
166      */
167 
168     Posf *head; /* Heads of the hash chains or NIL. */
169 
170     uInt  ins_h;          /* hash index of string to be inserted */
171     uInt  hash_size;      /* number of elements in hash table */
172     uInt  hash_bits;      /* log2(hash_size) */
173     uInt  hash_mask;      /* hash_size-1 */
174 
175     uInt  hash_shift;
176     /* Number of bits by which ins_h must be shifted at each input
177      * step. It must be such that after MIN_MATCH steps, the oldest
178      * byte no longer takes part in the hash key, that is:
179      *   hash_shift * MIN_MATCH >= hash_bits
180      */
181 
182     long block_start;
183     /* Window position at the beginning of the current output block. Gets
184      * negative when the window is moved backwards.
185      */
186 
187     uInt match_length;           /* length of best match */
188     IPos prev_match;             /* previous match */
189     int match_available;         /* set if previous match exists */
190     uInt strstart;               /* start of string to insert */
191     uInt match_start;            /* start of matching string */
192     uInt lookahead;              /* number of valid bytes ahead in window */
193 
194     uInt prev_length;
195     /* Length of the best match at previous step. Matches not greater than this
196      * are discarded. This is used in the lazy match evaluation.
197      */
198 
199     uInt max_chain_length;
200     /* To speed up deflation, hash chains are never searched beyond this
201      * length.  A higher limit improves compression ratio but degrades the
202      * speed.
203      */
204 
205     uInt max_lazy_match;
206     /* Attempt to find a better match only when the current match is strictly
207      * smaller than this value. This mechanism is used only for compression
208      * levels >= 4.
209      */
210 #   define max_insert_length  max_lazy_match
211     /* Insert new strings in the hash table only if the match length is not
212      * greater than this length. This saves time but degrades compression.
213      * max_insert_length is used only for compression levels <= 3.
214      */
215 
216     int level;    /* compression level (1..9) */
217     int strategy; /* favor or force Huffman coding*/
218 
219     uInt good_match;
220     /* Use a faster search when the previous match is longer than this */
221 
222     int nice_match; /* Stop searching when current match exceeds this */
223 
224                 /* used by trees.c: */
225     /* Didn't use ct_data typedef below to suppress compiler warning */
226     struct ct_data_s dyn_ltree[HEAP_SIZE];   /* literal and length tree */
227     struct ct_data_s dyn_dtree[2*D_CODES+1]; /* distance tree */
228     struct ct_data_s bl_tree[2*BL_CODES+1];  /* Huffman tree for bit lengths */
229 
230     struct tree_desc_s l_desc;               /* desc. for literal tree */
231     struct tree_desc_s d_desc;               /* desc. for distance tree */
232     struct tree_desc_s bl_desc;              /* desc. for bit length tree */
233 
234     ush bl_count[MAX_BITS+1];
235     /* number of codes at each bit length for an optimal tree */
236 
237     int heap[2*L_CODES+1];      /* heap used to build the Huffman trees */
238     int heap_len;               /* number of elements in the heap */
239     int heap_max;               /* element of largest frequency */
240     /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used.
241      * The same heap array is used to build all trees.
242      */
243 
244     uch depth[2*L_CODES+1];
245     /* Depth of each subtree used as tie breaker for trees of equal frequency
246      */
247 
248 #ifdef LIT_MEM
249 #   define LIT_BUFS 5
250     ushf *d_buf;          /* buffer for distances */
251     uchf *l_buf;          /* buffer for literals/lengths */
252 #else
253 #   define LIT_BUFS 4
254     uchf *sym_buf;        /* buffer for distances and literals/lengths */
255 #endif
256 
257     uInt  lit_bufsize;
258     /* Size of match buffer for literals/lengths.  There are 4 reasons for
259      * limiting lit_bufsize to 64K:
260      *   - frequencies can be kept in 16 bit counters
261      *   - if compression is not successful for the first block, all input
262      *     data is still in the window so we can still emit a stored block even
263      *     when input comes from standard input.  (This can also be done for
264      *     all blocks if lit_bufsize is not greater than 32K.)
265      *   - if compression is not successful for a file smaller than 64K, we can
266      *     even emit a stored file instead of a stored block (saving 5 bytes).
267      *     This is applicable only for zip (not gzip or zlib).
268      *   - creating new Huffman trees less frequently may not provide fast
269      *     adaptation to changes in the input data statistics. (Take for
270      *     example a binary file with poorly compressible code followed by
271      *     a highly compressible string table.) Smaller buffer sizes give
272      *     fast adaptation but have of course the overhead of transmitting
273      *     trees more frequently.
274      *   - I can't count above 4
275      */
276 
277     uInt sym_next;      /* running index in symbol buffer */
278     uInt sym_end;       /* symbol table full when sym_next reaches this */
279 
280     ulg opt_len;        /* bit length of current block with optimal trees */
281     ulg static_len;     /* bit length of current block with static trees */
282     uInt matches;       /* number of string matches in current block */
283     uInt insert;        /* bytes at end of window left to insert */
284 
285 #ifdef ZLIB_DEBUG
286     ulg compressed_len; /* total bit length of compressed file mod 2^32 */
287     ulg bits_sent;      /* bit length of compressed data sent mod 2^32 */
288 #endif
289 
290     ush bi_buf;
291     /* Output buffer. bits are inserted starting at the bottom (least
292      * significant bits).
293      */
294     int bi_valid;
295     /* Number of valid bits in bi_buf.  All bits above the last valid bit
296      * are always zero.
297      */
298 
299     ulg high_water;
300     /* High water mark offset in window for initialized bytes -- bytes above
301      * this are set to zero in order to avoid memory check warnings when
302      * longest match routines access bytes past the input.  This is then
303      * updated to the new high water mark.
304      */
305 
306 } FAR deflate_state;
307 
308 /* Output a byte on the stream.
309  * IN assertion: there is enough room in pending_buf.
310  */
311 #define put_byte(s, c) {s->pending_buf[s->pending++] = (Bytef)(c);}
312 
313 
314 #define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
315 /* Minimum amount of lookahead, except at the end of the input file.
316  * See deflate.c for comments about the MIN_MATCH+1.
317  */
318 
319 #define MAX_DIST(s)  ((s)->w_size-MIN_LOOKAHEAD)
320 /* In order to simplify the code, particularly on 16 bit machines, match
321  * distances are limited to MAX_DIST instead of WSIZE.
322  */
323 
324 #define WIN_INIT MAX_MATCH
325 /* Number of bytes after end of data in window to initialize in order to avoid
326    memory checker errors from longest match routines */
327 
328         /* in trees.c */
329 void ZLIB_INTERNAL _tr_init(deflate_state *s);
330 int ZLIB_INTERNAL _tr_tally(deflate_state *s, unsigned dist, unsigned lc);
331 void ZLIB_INTERNAL _tr_flush_block(deflate_state *s, charf *buf,
332                                    ulg stored_len, int last);
333 void ZLIB_INTERNAL _tr_flush_bits(deflate_state *s);
334 void ZLIB_INTERNAL _tr_align(deflate_state *s);
335 void ZLIB_INTERNAL _tr_stored_block(deflate_state *s, charf *buf,
336                                     ulg stored_len, int last);
337 
338 #define d_code(dist) \
339    ((dist) < 256 ? _dist_code[dist] : _dist_code[256+((dist)>>7)])
340 /* Mapping from a distance to a distance code. dist is the distance - 1 and
341  * must not have side effects. _dist_code[256] and _dist_code[257] are never
342  * used.
343  */
344 
345 #ifndef ZLIB_DEBUG
346 /* Inline versions of _tr_tally for speed: */
347 
348 #if defined(GEN_TREES_H) || !defined(STDC)
349   extern uch ZLIB_INTERNAL _length_code[];
350   extern uch ZLIB_INTERNAL _dist_code[];
351 #else
352   extern const uch ZLIB_INTERNAL _length_code[];
353   extern const uch ZLIB_INTERNAL _dist_code[];
354 #endif
355 
356 #ifdef LIT_MEM
357 # define _tr_tally_lit(s, c, flush) \
358   { uch cc = (c); \
359     s->d_buf[s->sym_next] = 0; \
360     s->l_buf[s->sym_next++] = cc; \
361     s->dyn_ltree[cc].Freq++; \
362     flush = (s->sym_next == s->sym_end); \
363    }
364 # define _tr_tally_dist(s, distance, length, flush) \
365   { uch len = (uch)(length); \
366     ush dist = (ush)(distance); \
367     s->d_buf[s->sym_next] = dist; \
368     s->l_buf[s->sym_next++] = len; \
369     dist--; \
370     s->dyn_ltree[_length_code[len]+LITERALS+1].Freq++; \
371     s->dyn_dtree[d_code(dist)].Freq++; \
372     flush = (s->sym_next == s->sym_end); \
373   }
374 #else
375 # define _tr_tally_lit(s, c, flush) \
376   { uch cc = (c); \
377     s->sym_buf[s->sym_next++] = 0; \
378     s->sym_buf[s->sym_next++] = 0; \
379     s->sym_buf[s->sym_next++] = cc; \
380     s->dyn_ltree[cc].Freq++; \
381     flush = (s->sym_next == s->sym_end); \
382    }
383 # define _tr_tally_dist(s, distance, length, flush) \
384   { uch len = (uch)(length); \
385     ush dist = (ush)(distance); \
386     s->sym_buf[s->sym_next++] = (uch)dist; \
387     s->sym_buf[s->sym_next++] = (uch)(dist >> 8); \
388     s->sym_buf[s->sym_next++] = len; \
389     dist--; \
390     s->dyn_ltree[_length_code[len]+LITERALS+1].Freq++; \
391     s->dyn_dtree[d_code(dist)].Freq++; \
392     flush = (s->sym_next == s->sym_end); \
393   }
394 #endif
395 #else
396 # define _tr_tally_lit(s, c, flush) flush = _tr_tally(s, 0, c)
397 # define _tr_tally_dist(s, distance, length, flush) \
398               flush = _tr_tally(s, distance, length)
399 #endif
400 
401 #endif /* DEFLATE_H */