1 | /* vi: set sw=4 ts=4: */
|
---|
2 | /*
|
---|
3 | * Gzip implementation for busybox
|
---|
4 | *
|
---|
5 | * Based on GNU gzip Copyright (C) 1992-1993 Jean-loup Gailly.
|
---|
6 | *
|
---|
7 | * Originally adjusted for busybox by Charles P. Wright <cpw@unix.asb.com>
|
---|
8 | * "this is a stripped down version of gzip I put into busybox, it does
|
---|
9 | * only standard in to standard out with -9 compression. It also requires
|
---|
10 | * the zcat module for some important functions."
|
---|
11 | *
|
---|
12 | * Adjusted further by Erik Andersen <andersen@codepoet.org> to support
|
---|
13 | * files as well as stdin/stdout, and to generally behave itself wrt
|
---|
14 | * command line handling.
|
---|
15 | *
|
---|
16 | * Licensed under GPLv2 or later, see file LICENSE in this tarball for details.
|
---|
17 | */
|
---|
18 |
|
---|
19 | /* big objects in bss:
|
---|
20 | * 00000020 b bl_count
|
---|
21 | * 00000074 b base_length
|
---|
22 | * 00000078 b base_dist
|
---|
23 | * 00000078 b static_dtree
|
---|
24 | * 0000009c b bl_tree
|
---|
25 | * 000000f4 b dyn_dtree
|
---|
26 | * 00000100 b length_code
|
---|
27 | * 00000200 b dist_code
|
---|
28 | * 0000023d b depth
|
---|
29 | * 00000400 b flag_buf
|
---|
30 | * 0000047a b heap
|
---|
31 | * 00000480 b static_ltree
|
---|
32 | * 000008f4 b dyn_ltree
|
---|
33 | */
|
---|
34 |
|
---|
35 | /* TODO: full support for -v for DESKTOP
|
---|
36 | * "/usr/bin/gzip -v a bogus aa" should say:
|
---|
37 | a: 85.1% -- replaced with a.gz
|
---|
38 | gzip: bogus: No such file or directory
|
---|
39 | aa: 85.1% -- replaced with aa.gz
|
---|
40 | */
|
---|
41 |
|
---|
42 | #include "libbb.h"
|
---|
43 |
|
---|
44 |
|
---|
45 | /* ===========================================================================
|
---|
46 | */
|
---|
47 | //#define DEBUG 1
|
---|
48 | /* Diagnostic functions */
|
---|
49 | #ifdef DEBUG
|
---|
50 | # define Assert(cond,msg) { if (!(cond)) bb_error_msg(msg); }
|
---|
51 | # define Trace(x) fprintf x
|
---|
52 | # define Tracev(x) {if (verbose) fprintf x; }
|
---|
53 | # define Tracevv(x) {if (verbose > 1) fprintf x; }
|
---|
54 | # define Tracec(c,x) {if (verbose && (c)) fprintf x; }
|
---|
55 | # define Tracecv(c,x) {if (verbose > 1 && (c)) fprintf x; }
|
---|
56 | #else
|
---|
57 | # define Assert(cond,msg)
|
---|
58 | # define Trace(x)
|
---|
59 | # define Tracev(x)
|
---|
60 | # define Tracevv(x)
|
---|
61 | # define Tracec(c,x)
|
---|
62 | # define Tracecv(c,x)
|
---|
63 | #endif
|
---|
64 |
|
---|
65 |
|
---|
66 | /* ===========================================================================
|
---|
67 | */
|
---|
68 | #define SMALL_MEM
|
---|
69 |
|
---|
70 | #ifndef INBUFSIZ
|
---|
71 | # ifdef SMALL_MEM
|
---|
72 | # define INBUFSIZ 0x2000 /* input buffer size */
|
---|
73 | # else
|
---|
74 | # define INBUFSIZ 0x8000 /* input buffer size */
|
---|
75 | # endif
|
---|
76 | #endif
|
---|
77 |
|
---|
78 | #ifndef OUTBUFSIZ
|
---|
79 | # ifdef SMALL_MEM
|
---|
80 | # define OUTBUFSIZ 8192 /* output buffer size */
|
---|
81 | # else
|
---|
82 | # define OUTBUFSIZ 16384 /* output buffer size */
|
---|
83 | # endif
|
---|
84 | #endif
|
---|
85 |
|
---|
86 | #ifndef DIST_BUFSIZE
|
---|
87 | # ifdef SMALL_MEM
|
---|
88 | # define DIST_BUFSIZE 0x2000 /* buffer for distances, see trees.c */
|
---|
89 | # else
|
---|
90 | # define DIST_BUFSIZE 0x8000 /* buffer for distances, see trees.c */
|
---|
91 | # endif
|
---|
92 | #endif
|
---|
93 |
|
---|
94 | /* gzip flag byte */
|
---|
95 | #define ASCII_FLAG 0x01 /* bit 0 set: file probably ascii text */
|
---|
96 | #define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */
|
---|
97 | #define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
|
---|
98 | #define ORIG_NAME 0x08 /* bit 3 set: original file name present */
|
---|
99 | #define COMMENT 0x10 /* bit 4 set: file comment present */
|
---|
100 | #define RESERVED 0xC0 /* bit 6,7: reserved */
|
---|
101 |
|
---|
102 | /* internal file attribute */
|
---|
103 | #define UNKNOWN 0xffff
|
---|
104 | #define BINARY 0
|
---|
105 | #define ASCII 1
|
---|
106 |
|
---|
107 | #ifndef WSIZE
|
---|
108 | # define WSIZE 0x8000 /* window size--must be a power of two, and */
|
---|
109 | #endif /* at least 32K for zip's deflate method */
|
---|
110 |
|
---|
111 | #define MIN_MATCH 3
|
---|
112 | #define MAX_MATCH 258
|
---|
113 | /* The minimum and maximum match lengths */
|
---|
114 |
|
---|
115 | #define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
|
---|
116 | /* Minimum amount of lookahead, except at the end of the input file.
|
---|
117 | * See deflate.c for comments about the MIN_MATCH+1.
|
---|
118 | */
|
---|
119 |
|
---|
120 | #define MAX_DIST (WSIZE-MIN_LOOKAHEAD)
|
---|
121 | /* In order to simplify the code, particularly on 16 bit machines, match
|
---|
122 | * distances are limited to MAX_DIST instead of WSIZE.
|
---|
123 | */
|
---|
124 |
|
---|
125 | #ifndef MAX_PATH_LEN
|
---|
126 | # define MAX_PATH_LEN 1024 /* max pathname length */
|
---|
127 | #endif
|
---|
128 |
|
---|
129 | #define seekable() 0 /* force sequential output */
|
---|
130 | #define translate_eol 0 /* no option -a yet */
|
---|
131 |
|
---|
132 | #ifndef BITS
|
---|
133 | # define BITS 16
|
---|
134 | #endif
|
---|
135 | #define INIT_BITS 9 /* Initial number of bits per code */
|
---|
136 |
|
---|
137 | #define BIT_MASK 0x1f /* Mask for 'number of compression bits' */
|
---|
138 | /* Mask 0x20 is reserved to mean a fourth header byte, and 0x40 is free.
|
---|
139 | * It's a pity that old uncompress does not check bit 0x20. That makes
|
---|
140 | * extension of the format actually undesirable because old compress
|
---|
141 | * would just crash on the new format instead of giving a meaningful
|
---|
142 | * error message. It does check the number of bits, but it's more
|
---|
143 | * helpful to say "unsupported format, get a new version" than
|
---|
144 | * "can only handle 16 bits".
|
---|
145 | */
|
---|
146 |
|
---|
147 | #ifdef MAX_EXT_CHARS
|
---|
148 | # define MAX_SUFFIX MAX_EXT_CHARS
|
---|
149 | #else
|
---|
150 | # define MAX_SUFFIX 30
|
---|
151 | #endif
|
---|
152 |
|
---|
153 |
|
---|
154 | /* ===========================================================================
|
---|
155 | * Compile with MEDIUM_MEM to reduce the memory requirements or
|
---|
156 | * with SMALL_MEM to use as little memory as possible. Use BIG_MEM if the
|
---|
157 | * entire input file can be held in memory (not possible on 16 bit systems).
|
---|
158 | * Warning: defining these symbols affects HASH_BITS (see below) and thus
|
---|
159 | * affects the compression ratio. The compressed output
|
---|
160 | * is still correct, and might even be smaller in some cases.
|
---|
161 | */
|
---|
162 |
|
---|
163 | #ifdef SMALL_MEM
|
---|
164 | # define HASH_BITS 13 /* Number of bits used to hash strings */
|
---|
165 | #endif
|
---|
166 | #ifdef MEDIUM_MEM
|
---|
167 | # define HASH_BITS 14
|
---|
168 | #endif
|
---|
169 | #ifndef HASH_BITS
|
---|
170 | # define HASH_BITS 15
|
---|
171 | /* For portability to 16 bit machines, do not use values above 15. */
|
---|
172 | #endif
|
---|
173 |
|
---|
174 | #define HASH_SIZE (unsigned)(1<<HASH_BITS)
|
---|
175 | #define HASH_MASK (HASH_SIZE-1)
|
---|
176 | #define WMASK (WSIZE-1)
|
---|
177 | /* HASH_SIZE and WSIZE must be powers of two */
|
---|
178 | #ifndef TOO_FAR
|
---|
179 | # define TOO_FAR 4096
|
---|
180 | #endif
|
---|
181 | /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
|
---|
182 |
|
---|
183 |
|
---|
184 | /* ===========================================================================
|
---|
185 | * These types are not really 'char', 'short' and 'long'
|
---|
186 | */
|
---|
187 | typedef uint8_t uch;
|
---|
188 | typedef uint16_t ush;
|
---|
189 | typedef uint32_t ulg;
|
---|
190 | typedef int32_t lng;
|
---|
191 |
|
---|
192 | typedef ush Pos;
|
---|
193 | typedef unsigned IPos;
|
---|
194 | /* A Pos is an index in the character window. We use short instead of int to
|
---|
195 | * save space in the various tables. IPos is used only for parameter passing.
|
---|
196 | */
|
---|
197 |
|
---|
198 | enum {
|
---|
199 | WINDOW_SIZE = 2 * WSIZE,
|
---|
200 | /* window size, 2*WSIZE except for MMAP or BIG_MEM, where it is the
|
---|
201 | * input file length plus MIN_LOOKAHEAD.
|
---|
202 | */
|
---|
203 |
|
---|
204 | max_chain_length = 4096,
|
---|
205 | /* To speed up deflation, hash chains are never searched beyond this length.
|
---|
206 | * A higher limit improves compression ratio but degrades the speed.
|
---|
207 | */
|
---|
208 |
|
---|
209 | max_lazy_match = 258,
|
---|
210 | /* Attempt to find a better match only when the current match is strictly
|
---|
211 | * smaller than this value. This mechanism is used only for compression
|
---|
212 | * levels >= 4.
|
---|
213 | */
|
---|
214 |
|
---|
215 | max_insert_length = max_lazy_match,
|
---|
216 | /* Insert new strings in the hash table only if the match length
|
---|
217 | * is not greater than this length. This saves time but degrades compression.
|
---|
218 | * max_insert_length is used only for compression levels <= 3.
|
---|
219 | */
|
---|
220 |
|
---|
221 | good_match = 32,
|
---|
222 | /* Use a faster search when the previous match is longer than this */
|
---|
223 |
|
---|
224 | /* Values for max_lazy_match, good_match and max_chain_length, depending on
|
---|
225 | * the desired pack level (0..9). The values given below have been tuned to
|
---|
226 | * exclude worst case performance for pathological files. Better values may be
|
---|
227 | * found for specific files.
|
---|
228 | */
|
---|
229 |
|
---|
230 | nice_match = 258, /* Stop searching when current match exceeds this */
|
---|
231 | /* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
|
---|
232 | * For deflate_fast() (levels <= 3) good is ignored and lazy has a different
|
---|
233 | * meaning.
|
---|
234 | */
|
---|
235 | };
|
---|
236 |
|
---|
237 |
|
---|
238 | struct globals {
|
---|
239 |
|
---|
240 | lng block_start;
|
---|
241 |
|
---|
242 | /* window position at the beginning of the current output block. Gets
|
---|
243 | * negative when the window is moved backwards.
|
---|
244 | */
|
---|
245 | unsigned ins_h; /* hash index of string to be inserted */
|
---|
246 |
|
---|
247 | #define H_SHIFT ((HASH_BITS+MIN_MATCH-1) / MIN_MATCH)
|
---|
248 | /* Number of bits by which ins_h and del_h must be shifted at each
|
---|
249 | * input step. It must be such that after MIN_MATCH steps, the oldest
|
---|
250 | * byte no longer takes part in the hash key, that is:
|
---|
251 | * H_SHIFT * MIN_MATCH >= HASH_BITS
|
---|
252 | */
|
---|
253 |
|
---|
254 | unsigned prev_length;
|
---|
255 |
|
---|
256 | /* Length of the best match at previous step. Matches not greater than this
|
---|
257 | * are discarded. This is used in the lazy match evaluation.
|
---|
258 | */
|
---|
259 |
|
---|
260 | unsigned strstart; /* start of string to insert */
|
---|
261 | unsigned match_start; /* start of matching string */
|
---|
262 | unsigned lookahead; /* number of valid bytes ahead in window */
|
---|
263 |
|
---|
264 | /* ===========================================================================
|
---|
265 | */
|
---|
266 | #define DECLARE(type, array, size) \
|
---|
267 | type * array
|
---|
268 | #define ALLOC(type, array, size) \
|
---|
269 | array = xzalloc((size_t)(((size)+1L)/2) * 2*sizeof(type));
|
---|
270 | #define FREE(array) \
|
---|
271 | do { free(array); array = NULL; } while (0)
|
---|
272 |
|
---|
273 | /* global buffers */
|
---|
274 |
|
---|
275 | /* buffer for literals or lengths */
|
---|
276 | /* DECLARE(uch, l_buf, LIT_BUFSIZE); */
|
---|
277 | DECLARE(uch, l_buf, INBUFSIZ);
|
---|
278 |
|
---|
279 | DECLARE(ush, d_buf, DIST_BUFSIZE);
|
---|
280 | DECLARE(uch, outbuf, OUTBUFSIZ);
|
---|
281 |
|
---|
282 | /* Sliding window. Input bytes are read into the second half of the window,
|
---|
283 | * and move to the first half later to keep a dictionary of at least WSIZE
|
---|
284 | * bytes. With this organization, matches are limited to a distance of
|
---|
285 | * WSIZE-MAX_MATCH bytes, but this ensures that IO is always
|
---|
286 | * performed with a length multiple of the block size. Also, it limits
|
---|
287 | * the window size to 64K, which is quite useful on MSDOS.
|
---|
288 | * To do: limit the window size to WSIZE+BSZ if SMALL_MEM (the code would
|
---|
289 | * be less efficient).
|
---|
290 | */
|
---|
291 | DECLARE(uch, window, 2L * WSIZE);
|
---|
292 |
|
---|
293 | /* Link to older string with same hash index. To limit the size of this
|
---|
294 | * array to 64K, this link is maintained only for the last 32K strings.
|
---|
295 | * An index in this array is thus a window index modulo 32K.
|
---|
296 | */
|
---|
297 | /* DECLARE(Pos, prev, WSIZE); */
|
---|
298 | DECLARE(ush, prev, 1L << BITS);
|
---|
299 |
|
---|
300 | /* Heads of the hash chains or 0. */
|
---|
301 | /* DECLARE(Pos, head, 1<<HASH_BITS); */
|
---|
302 | #define head (G1.prev + WSIZE) /* hash head (see deflate.c) */
|
---|
303 |
|
---|
304 | /* number of input bytes */
|
---|
305 | ulg isize; /* only 32 bits stored in .gz file */
|
---|
306 |
|
---|
307 | /* bbox always use stdin/stdout */
|
---|
308 | #define ifd STDIN_FILENO /* input file descriptor */
|
---|
309 | #define ofd STDOUT_FILENO /* output file descriptor */
|
---|
310 |
|
---|
311 | #ifdef DEBUG
|
---|
312 | unsigned insize; /* valid bytes in l_buf */
|
---|
313 | #endif
|
---|
314 | unsigned outcnt; /* bytes in output buffer */
|
---|
315 |
|
---|
316 | smallint eofile; /* flag set at end of input file */
|
---|
317 |
|
---|
318 | /* ===========================================================================
|
---|
319 | * Local data used by the "bit string" routines.
|
---|
320 | */
|
---|
321 |
|
---|
322 | unsigned short bi_buf;
|
---|
323 |
|
---|
324 | /* Output buffer. bits are inserted starting at the bottom (least significant
|
---|
325 | * bits).
|
---|
326 | */
|
---|
327 |
|
---|
328 | #undef BUF_SIZE
|
---|
329 | #define BUF_SIZE (8 * sizeof(G1.bi_buf))
|
---|
330 | /* Number of bits used within bi_buf. (bi_buf might be implemented on
|
---|
331 | * more than 16 bits on some systems.)
|
---|
332 | */
|
---|
333 |
|
---|
334 | int bi_valid;
|
---|
335 |
|
---|
336 | /* Current input function. Set to mem_read for in-memory compression */
|
---|
337 |
|
---|
338 | #ifdef DEBUG
|
---|
339 | ulg bits_sent; /* bit length of the compressed data */
|
---|
340 | #endif
|
---|
341 |
|
---|
342 | uint32_t *crc_32_tab;
|
---|
343 | uint32_t crc; /* shift register contents */
|
---|
344 | };
|
---|
345 |
|
---|
346 | #define G1 (*(ptr_to_globals - 1))
|
---|
347 |
|
---|
348 |
|
---|
349 | /* ===========================================================================
|
---|
350 | * Write the output buffer outbuf[0..outcnt-1] and update bytes_out.
|
---|
351 | * (used for the compressed data only)
|
---|
352 | */
|
---|
353 | static void flush_outbuf(void)
|
---|
354 | {
|
---|
355 | if (G1.outcnt == 0)
|
---|
356 | return;
|
---|
357 |
|
---|
358 | xwrite(ofd, (char *) G1.outbuf, G1.outcnt);
|
---|
359 | G1.outcnt = 0;
|
---|
360 | }
|
---|
361 |
|
---|
362 |
|
---|
363 | /* ===========================================================================
|
---|
364 | */
|
---|
365 | /* put_8bit is used for the compressed output */
|
---|
366 | #define put_8bit(c) \
|
---|
367 | do { \
|
---|
368 | G1.outbuf[G1.outcnt++] = (c); \
|
---|
369 | if (G1.outcnt == OUTBUFSIZ) flush_outbuf(); \
|
---|
370 | } while (0)
|
---|
371 |
|
---|
372 | /* Output a 16 bit value, lsb first */
|
---|
373 | static void put_16bit(ush w)
|
---|
374 | {
|
---|
375 | if (G1.outcnt < OUTBUFSIZ - 2) {
|
---|
376 | G1.outbuf[G1.outcnt++] = w;
|
---|
377 | G1.outbuf[G1.outcnt++] = w >> 8;
|
---|
378 | } else {
|
---|
379 | put_8bit(w);
|
---|
380 | put_8bit(w >> 8);
|
---|
381 | }
|
---|
382 | }
|
---|
383 |
|
---|
384 | static void put_32bit(ulg n)
|
---|
385 | {
|
---|
386 | put_16bit(n);
|
---|
387 | put_16bit(n >> 16);
|
---|
388 | }
|
---|
389 |
|
---|
390 | /* ===========================================================================
|
---|
391 | * Clear input and output buffers
|
---|
392 | */
|
---|
393 | static void clear_bufs(void)
|
---|
394 | {
|
---|
395 | G1.outcnt = 0;
|
---|
396 | #ifdef DEBUG
|
---|
397 | G1.insize = 0;
|
---|
398 | #endif
|
---|
399 | G1.isize = 0;
|
---|
400 | }
|
---|
401 |
|
---|
402 |
|
---|
403 | /* ===========================================================================
|
---|
404 | * Run a set of bytes through the crc shift register. If s is a NULL
|
---|
405 | * pointer, then initialize the crc shift register contents instead.
|
---|
406 | * Return the current crc in either case.
|
---|
407 | */
|
---|
408 | static uint32_t updcrc(uch * s, unsigned n)
|
---|
409 | {
|
---|
410 | uint32_t c = G1.crc;
|
---|
411 | while (n) {
|
---|
412 | c = G1.crc_32_tab[(uch)(c ^ *s++)] ^ (c >> 8);
|
---|
413 | n--;
|
---|
414 | }
|
---|
415 | G1.crc = c;
|
---|
416 | return c;
|
---|
417 | }
|
---|
418 |
|
---|
419 |
|
---|
420 | /* ===========================================================================
|
---|
421 | * Read a new buffer from the current input file, perform end-of-line
|
---|
422 | * translation, and update the crc and input file size.
|
---|
423 | * IN assertion: size >= 2 (for end-of-line translation)
|
---|
424 | */
|
---|
425 | static unsigned file_read(void *buf, unsigned size)
|
---|
426 | {
|
---|
427 | unsigned len;
|
---|
428 |
|
---|
429 | Assert(G1.insize == 0, "l_buf not empty");
|
---|
430 |
|
---|
431 | len = safe_read(ifd, buf, size);
|
---|
432 | if (len == (unsigned)(-1) || len == 0)
|
---|
433 | return len;
|
---|
434 |
|
---|
435 | updcrc(buf, len);
|
---|
436 | G1.isize += len;
|
---|
437 | return len;
|
---|
438 | }
|
---|
439 |
|
---|
440 |
|
---|
441 | /* ===========================================================================
|
---|
442 | * Send a value on a given number of bits.
|
---|
443 | * IN assertion: length <= 16 and value fits in length bits.
|
---|
444 | */
|
---|
445 | static void send_bits(int value, int length)
|
---|
446 | {
|
---|
447 | #ifdef DEBUG
|
---|
448 | Tracev((stderr, " l %2d v %4x ", length, value));
|
---|
449 | Assert(length > 0 && length <= 15, "invalid length");
|
---|
450 | G1.bits_sent += length;
|
---|
451 | #endif
|
---|
452 | /* If not enough room in bi_buf, use (valid) bits from bi_buf and
|
---|
453 | * (16 - bi_valid) bits from value, leaving (width - (16-bi_valid))
|
---|
454 | * unused bits in value.
|
---|
455 | */
|
---|
456 | if (G1.bi_valid > (int) BUF_SIZE - length) {
|
---|
457 | G1.bi_buf |= (value << G1.bi_valid);
|
---|
458 | put_16bit(G1.bi_buf);
|
---|
459 | G1.bi_buf = (ush) value >> (BUF_SIZE - G1.bi_valid);
|
---|
460 | G1.bi_valid += length - BUF_SIZE;
|
---|
461 | } else {
|
---|
462 | G1.bi_buf |= value << G1.bi_valid;
|
---|
463 | G1.bi_valid += length;
|
---|
464 | }
|
---|
465 | }
|
---|
466 |
|
---|
467 |
|
---|
468 | /* ===========================================================================
|
---|
469 | * Reverse the first len bits of a code, using straightforward code (a faster
|
---|
470 | * method would use a table)
|
---|
471 | * IN assertion: 1 <= len <= 15
|
---|
472 | */
|
---|
473 | static unsigned bi_reverse(unsigned code, int len)
|
---|
474 | {
|
---|
475 | unsigned res = 0;
|
---|
476 |
|
---|
477 | while (1) {
|
---|
478 | res |= code & 1;
|
---|
479 | if (--len <= 0) return res;
|
---|
480 | code >>= 1;
|
---|
481 | res <<= 1;
|
---|
482 | }
|
---|
483 | }
|
---|
484 |
|
---|
485 |
|
---|
486 | /* ===========================================================================
|
---|
487 | * Write out any remaining bits in an incomplete byte.
|
---|
488 | */
|
---|
489 | static void bi_windup(void)
|
---|
490 | {
|
---|
491 | if (G1.bi_valid > 8) {
|
---|
492 | put_16bit(G1.bi_buf);
|
---|
493 | } else if (G1.bi_valid > 0) {
|
---|
494 | put_8bit(G1.bi_buf);
|
---|
495 | }
|
---|
496 | G1.bi_buf = 0;
|
---|
497 | G1.bi_valid = 0;
|
---|
498 | #ifdef DEBUG
|
---|
499 | G1.bits_sent = (G1.bits_sent + 7) & ~7;
|
---|
500 | #endif
|
---|
501 | }
|
---|
502 |
|
---|
503 |
|
---|
504 | /* ===========================================================================
|
---|
505 | * Copy a stored block to the zip file, storing first the length and its
|
---|
506 | * one's complement if requested.
|
---|
507 | */
|
---|
508 | static void copy_block(char *buf, unsigned len, int header)
|
---|
509 | {
|
---|
510 | bi_windup(); /* align on byte boundary */
|
---|
511 |
|
---|
512 | if (header) {
|
---|
513 | put_16bit(len);
|
---|
514 | put_16bit(~len);
|
---|
515 | #ifdef DEBUG
|
---|
516 | G1.bits_sent += 2 * 16;
|
---|
517 | #endif
|
---|
518 | }
|
---|
519 | #ifdef DEBUG
|
---|
520 | G1.bits_sent += (ulg) len << 3;
|
---|
521 | #endif
|
---|
522 | while (len--) {
|
---|
523 | put_8bit(*buf++);
|
---|
524 | }
|
---|
525 | }
|
---|
526 |
|
---|
527 |
|
---|
528 | /* ===========================================================================
|
---|
529 | * Fill the window when the lookahead becomes insufficient.
|
---|
530 | * Updates strstart and lookahead, and sets eofile if end of input file.
|
---|
531 | * IN assertion: lookahead < MIN_LOOKAHEAD && strstart + lookahead > 0
|
---|
532 | * OUT assertions: at least one byte has been read, or eofile is set;
|
---|
533 | * file reads are performed for at least two bytes (required for the
|
---|
534 | * translate_eol option).
|
---|
535 | */
|
---|
536 | static void fill_window(void)
|
---|
537 | {
|
---|
538 | unsigned n, m;
|
---|
539 | unsigned more = WINDOW_SIZE - G1.lookahead - G1.strstart;
|
---|
540 | /* Amount of free space at the end of the window. */
|
---|
541 |
|
---|
542 | /* If the window is almost full and there is insufficient lookahead,
|
---|
543 | * move the upper half to the lower one to make room in the upper half.
|
---|
544 | */
|
---|
545 | if (more == (unsigned) -1) {
|
---|
546 | /* Very unlikely, but possible on 16 bit machine if strstart == 0
|
---|
547 | * and lookahead == 1 (input done one byte at time)
|
---|
548 | */
|
---|
549 | more--;
|
---|
550 | } else if (G1.strstart >= WSIZE + MAX_DIST) {
|
---|
551 | /* By the IN assertion, the window is not empty so we can't confuse
|
---|
552 | * more == 0 with more == 64K on a 16 bit machine.
|
---|
553 | */
|
---|
554 | Assert(WINDOW_SIZE == 2 * WSIZE, "no sliding with BIG_MEM");
|
---|
555 |
|
---|
556 | memcpy(G1.window, G1.window + WSIZE, WSIZE);
|
---|
557 | G1.match_start -= WSIZE;
|
---|
558 | G1.strstart -= WSIZE; /* we now have strstart >= MAX_DIST: */
|
---|
559 |
|
---|
560 | G1.block_start -= WSIZE;
|
---|
561 |
|
---|
562 | for (n = 0; n < HASH_SIZE; n++) {
|
---|
563 | m = head[n];
|
---|
564 | head[n] = (Pos) (m >= WSIZE ? m - WSIZE : 0);
|
---|
565 | }
|
---|
566 | for (n = 0; n < WSIZE; n++) {
|
---|
567 | m = G1.prev[n];
|
---|
568 | G1.prev[n] = (Pos) (m >= WSIZE ? m - WSIZE : 0);
|
---|
569 | /* If n is not on any hash chain, prev[n] is garbage but
|
---|
570 | * its value will never be used.
|
---|
571 | */
|
---|
572 | }
|
---|
573 | more += WSIZE;
|
---|
574 | }
|
---|
575 | /* At this point, more >= 2 */
|
---|
576 | if (!G1.eofile) {
|
---|
577 | n = file_read(G1.window + G1.strstart + G1.lookahead, more);
|
---|
578 | if (n == 0 || n == (unsigned) -1) {
|
---|
579 | G1.eofile = 1;
|
---|
580 | } else {
|
---|
581 | G1.lookahead += n;
|
---|
582 | }
|
---|
583 | }
|
---|
584 | }
|
---|
585 |
|
---|
586 |
|
---|
587 | /* ===========================================================================
|
---|
588 | * Set match_start to the longest match starting at the given string and
|
---|
589 | * return its length. Matches shorter or equal to prev_length are discarded,
|
---|
590 | * in which case the result is equal to prev_length and match_start is
|
---|
591 | * garbage.
|
---|
592 | * IN assertions: cur_match is the head of the hash chain for the current
|
---|
593 | * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
|
---|
594 | */
|
---|
595 |
|
---|
596 | /* For MSDOS, OS/2 and 386 Unix, an optimized version is in match.asm or
|
---|
597 | * match.s. The code is functionally equivalent, so you can use the C version
|
---|
598 | * if desired.
|
---|
599 | */
|
---|
600 | static int longest_match(IPos cur_match)
|
---|
601 | {
|
---|
602 | unsigned chain_length = max_chain_length; /* max hash chain length */
|
---|
603 | uch *scan = G1.window + G1.strstart; /* current string */
|
---|
604 | uch *match; /* matched string */
|
---|
605 | int len; /* length of current match */
|
---|
606 | int best_len = G1.prev_length; /* best match length so far */
|
---|
607 | IPos limit = G1.strstart > (IPos) MAX_DIST ? G1.strstart - (IPos) MAX_DIST : 0;
|
---|
608 | /* Stop when cur_match becomes <= limit. To simplify the code,
|
---|
609 | * we prevent matches with the string of window index 0.
|
---|
610 | */
|
---|
611 |
|
---|
612 | /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
|
---|
613 | * It is easy to get rid of this optimization if necessary.
|
---|
614 | */
|
---|
615 | #if HASH_BITS < 8 || MAX_MATCH != 258
|
---|
616 | # error Code too clever
|
---|
617 | #endif
|
---|
618 | uch *strend = G1.window + G1.strstart + MAX_MATCH;
|
---|
619 | uch scan_end1 = scan[best_len - 1];
|
---|
620 | uch scan_end = scan[best_len];
|
---|
621 |
|
---|
622 | /* Do not waste too much time if we already have a good match: */
|
---|
623 | if (G1.prev_length >= good_match) {
|
---|
624 | chain_length >>= 2;
|
---|
625 | }
|
---|
626 | Assert(G1.strstart <= WINDOW_SIZE - MIN_LOOKAHEAD, "insufficient lookahead");
|
---|
627 |
|
---|
628 | do {
|
---|
629 | Assert(cur_match < G1.strstart, "no future");
|
---|
630 | match = G1.window + cur_match;
|
---|
631 |
|
---|
632 | /* Skip to next match if the match length cannot increase
|
---|
633 | * or if the match length is less than 2:
|
---|
634 | */
|
---|
635 | if (match[best_len] != scan_end ||
|
---|
636 | match[best_len - 1] != scan_end1 ||
|
---|
637 | *match != *scan || *++match != scan[1])
|
---|
638 | continue;
|
---|
639 |
|
---|
640 | /* The check at best_len-1 can be removed because it will be made
|
---|
641 | * again later. (This heuristic is not always a win.)
|
---|
642 | * It is not necessary to compare scan[2] and match[2] since they
|
---|
643 | * are always equal when the other bytes match, given that
|
---|
644 | * the hash keys are equal and that HASH_BITS >= 8.
|
---|
645 | */
|
---|
646 | scan += 2, match++;
|
---|
647 |
|
---|
648 | /* We check for insufficient lookahead only every 8th comparison;
|
---|
649 | * the 256th check will be made at strstart+258.
|
---|
650 | */
|
---|
651 | do {
|
---|
652 | } while (*++scan == *++match && *++scan == *++match &&
|
---|
653 | *++scan == *++match && *++scan == *++match &&
|
---|
654 | *++scan == *++match && *++scan == *++match &&
|
---|
655 | *++scan == *++match && *++scan == *++match && scan < strend);
|
---|
656 |
|
---|
657 | len = MAX_MATCH - (int) (strend - scan);
|
---|
658 | scan = strend - MAX_MATCH;
|
---|
659 |
|
---|
660 | if (len > best_len) {
|
---|
661 | G1.match_start = cur_match;
|
---|
662 | best_len = len;
|
---|
663 | if (len >= nice_match)
|
---|
664 | break;
|
---|
665 | scan_end1 = scan[best_len - 1];
|
---|
666 | scan_end = scan[best_len];
|
---|
667 | }
|
---|
668 | } while ((cur_match = G1.prev[cur_match & WMASK]) > limit
|
---|
669 | && --chain_length != 0);
|
---|
670 |
|
---|
671 | return best_len;
|
---|
672 | }
|
---|
673 |
|
---|
674 |
|
---|
675 | #ifdef DEBUG
|
---|
676 | /* ===========================================================================
|
---|
677 | * Check that the match at match_start is indeed a match.
|
---|
678 | */
|
---|
679 | static void check_match(IPos start, IPos match, int length)
|
---|
680 | {
|
---|
681 | /* check that the match is indeed a match */
|
---|
682 | if (memcmp(G1.window + match, G1.window + start, length) != 0) {
|
---|
683 | bb_error_msg(" start %d, match %d, length %d", start, match, length);
|
---|
684 | bb_error_msg("invalid match");
|
---|
685 | }
|
---|
686 | if (verbose > 1) {
|
---|
687 | bb_error_msg("\\[%d,%d]", start - match, length);
|
---|
688 | do {
|
---|
689 | putc(G1.window[start++], stderr);
|
---|
690 | } while (--length != 0);
|
---|
691 | }
|
---|
692 | }
|
---|
693 | #else
|
---|
694 | # define check_match(start, match, length) ((void)0)
|
---|
695 | #endif
|
---|
696 |
|
---|
697 |
|
---|
698 | /* trees.c -- output deflated data using Huffman coding
|
---|
699 | * Copyright (C) 1992-1993 Jean-loup Gailly
|
---|
700 | * This is free software; you can redistribute it and/or modify it under the
|
---|
701 | * terms of the GNU General Public License, see the file COPYING.
|
---|
702 | */
|
---|
703 |
|
---|
704 | /* PURPOSE
|
---|
705 | * Encode various sets of source values using variable-length
|
---|
706 | * binary code trees.
|
---|
707 | *
|
---|
708 | * DISCUSSION
|
---|
709 | * The PKZIP "deflation" process uses several Huffman trees. The more
|
---|
710 | * common source values are represented by shorter bit sequences.
|
---|
711 | *
|
---|
712 | * Each code tree is stored in the ZIP file in a compressed form
|
---|
713 | * which is itself a Huffman encoding of the lengths of
|
---|
714 | * all the code strings (in ascending order by source values).
|
---|
715 | * The actual code strings are reconstructed from the lengths in
|
---|
716 | * the UNZIP process, as described in the "application note"
|
---|
717 | * (APPNOTE.TXT) distributed as part of PKWARE's PKZIP program.
|
---|
718 | *
|
---|
719 | * REFERENCES
|
---|
720 | * Lynch, Thomas J.
|
---|
721 | * Data Compression: Techniques and Applications, pp. 53-55.
|
---|
722 | * Lifetime Learning Publications, 1985. ISBN 0-534-03418-7.
|
---|
723 | *
|
---|
724 | * Storer, James A.
|
---|
725 | * Data Compression: Methods and Theory, pp. 49-50.
|
---|
726 | * Computer Science Press, 1988. ISBN 0-7167-8156-5.
|
---|
727 | *
|
---|
728 | * Sedgewick, R.
|
---|
729 | * Algorithms, p290.
|
---|
730 | * Addison-Wesley, 1983. ISBN 0-201-06672-6.
|
---|
731 | *
|
---|
732 | * INTERFACE
|
---|
733 | * void ct_init()
|
---|
734 | * Allocate the match buffer, initialize the various tables [and save
|
---|
735 | * the location of the internal file attribute (ascii/binary) and
|
---|
736 | * method (DEFLATE/STORE) -- deleted in bbox]
|
---|
737 | *
|
---|
738 | * void ct_tally(int dist, int lc);
|
---|
739 | * Save the match info and tally the frequency counts.
|
---|
740 | *
|
---|
741 | * ulg flush_block(char *buf, ulg stored_len, int eof)
|
---|
742 | * Determine the best encoding for the current block: dynamic trees,
|
---|
743 | * static trees or store, and output the encoded block to the zip
|
---|
744 | * file. Returns the total compressed length for the file so far.
|
---|
745 | */
|
---|
746 |
|
---|
747 | #define MAX_BITS 15
|
---|
748 | /* All codes must not exceed MAX_BITS bits */
|
---|
749 |
|
---|
750 | #define MAX_BL_BITS 7
|
---|
751 | /* Bit length codes must not exceed MAX_BL_BITS bits */
|
---|
752 |
|
---|
753 | #define LENGTH_CODES 29
|
---|
754 | /* number of length codes, not counting the special END_BLOCK code */
|
---|
755 |
|
---|
756 | #define LITERALS 256
|
---|
757 | /* number of literal bytes 0..255 */
|
---|
758 |
|
---|
759 | #define END_BLOCK 256
|
---|
760 | /* end of block literal code */
|
---|
761 |
|
---|
762 | #define L_CODES (LITERALS+1+LENGTH_CODES)
|
---|
763 | /* number of Literal or Length codes, including the END_BLOCK code */
|
---|
764 |
|
---|
765 | #define D_CODES 30
|
---|
766 | /* number of distance codes */
|
---|
767 |
|
---|
768 | #define BL_CODES 19
|
---|
769 | /* number of codes used to transfer the bit lengths */
|
---|
770 |
|
---|
771 | /* extra bits for each length code */
|
---|
772 | static const uint8_t extra_lbits[LENGTH_CODES] ALIGN1 = {
|
---|
773 | 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4,
|
---|
774 | 4, 4, 5, 5, 5, 5, 0
|
---|
775 | };
|
---|
776 |
|
---|
777 | /* extra bits for each distance code */
|
---|
778 | static const uint8_t extra_dbits[D_CODES] ALIGN1 = {
|
---|
779 | 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9,
|
---|
780 | 10, 10, 11, 11, 12, 12, 13, 13
|
---|
781 | };
|
---|
782 |
|
---|
783 | /* extra bits for each bit length code */
|
---|
784 | static const uint8_t extra_blbits[BL_CODES] ALIGN1 = {
|
---|
785 | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 3, 7 };
|
---|
786 |
|
---|
787 | /* number of codes at each bit length for an optimal tree */
|
---|
788 | static const uint8_t bl_order[BL_CODES] ALIGN1 = {
|
---|
789 | 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15 };
|
---|
790 |
|
---|
791 | #define STORED_BLOCK 0
|
---|
792 | #define STATIC_TREES 1
|
---|
793 | #define DYN_TREES 2
|
---|
794 | /* The three kinds of block type */
|
---|
795 |
|
---|
796 | #ifndef LIT_BUFSIZE
|
---|
797 | # ifdef SMALL_MEM
|
---|
798 | # define LIT_BUFSIZE 0x2000
|
---|
799 | # else
|
---|
800 | # ifdef MEDIUM_MEM
|
---|
801 | # define LIT_BUFSIZE 0x4000
|
---|
802 | # else
|
---|
803 | # define LIT_BUFSIZE 0x8000
|
---|
804 | # endif
|
---|
805 | # endif
|
---|
806 | #endif
|
---|
807 | #ifndef DIST_BUFSIZE
|
---|
808 | # define DIST_BUFSIZE LIT_BUFSIZE
|
---|
809 | #endif
|
---|
810 | /* Sizes of match buffers for literals/lengths and distances. There are
|
---|
811 | * 4 reasons for limiting LIT_BUFSIZE to 64K:
|
---|
812 | * - frequencies can be kept in 16 bit counters
|
---|
813 | * - if compression is not successful for the first block, all input data is
|
---|
814 | * still in the window so we can still emit a stored block even when input
|
---|
815 | * comes from standard input. (This can also be done for all blocks if
|
---|
816 | * LIT_BUFSIZE is not greater than 32K.)
|
---|
817 | * - if compression is not successful for a file smaller than 64K, we can
|
---|
818 | * even emit a stored file instead of a stored block (saving 5 bytes).
|
---|
819 | * - creating new Huffman trees less frequently may not provide fast
|
---|
820 | * adaptation to changes in the input data statistics. (Take for
|
---|
821 | * example a binary file with poorly compressible code followed by
|
---|
822 | * a highly compressible string table.) Smaller buffer sizes give
|
---|
823 | * fast adaptation but have of course the overhead of transmitting trees
|
---|
824 | * more frequently.
|
---|
825 | * - I can't count above 4
|
---|
826 | * The current code is general and allows DIST_BUFSIZE < LIT_BUFSIZE (to save
|
---|
827 | * memory at the expense of compression). Some optimizations would be possible
|
---|
828 | * if we rely on DIST_BUFSIZE == LIT_BUFSIZE.
|
---|
829 | */
|
---|
830 | #define REP_3_6 16
|
---|
831 | /* repeat previous bit length 3-6 times (2 bits of repeat count) */
|
---|
832 | #define REPZ_3_10 17
|
---|
833 | /* repeat a zero length 3-10 times (3 bits of repeat count) */
|
---|
834 | #define REPZ_11_138 18
|
---|
835 | /* repeat a zero length 11-138 times (7 bits of repeat count) */
|
---|
836 |
|
---|
837 | /* ===========================================================================
|
---|
838 | */
|
---|
839 | /* Data structure describing a single value and its code string. */
|
---|
840 | typedef struct ct_data {
|
---|
841 | union {
|
---|
842 | ush freq; /* frequency count */
|
---|
843 | ush code; /* bit string */
|
---|
844 | } fc;
|
---|
845 | union {
|
---|
846 | ush dad; /* father node in Huffman tree */
|
---|
847 | ush len; /* length of bit string */
|
---|
848 | } dl;
|
---|
849 | } ct_data;
|
---|
850 |
|
---|
851 | #define Freq fc.freq
|
---|
852 | #define Code fc.code
|
---|
853 | #define Dad dl.dad
|
---|
854 | #define Len dl.len
|
---|
855 |
|
---|
856 | #define HEAP_SIZE (2*L_CODES + 1)
|
---|
857 | /* maximum heap size */
|
---|
858 |
|
---|
859 | typedef struct tree_desc {
|
---|
860 | ct_data *dyn_tree; /* the dynamic tree */
|
---|
861 | ct_data *static_tree; /* corresponding static tree or NULL */
|
---|
862 | const uint8_t *extra_bits; /* extra bits for each code or NULL */
|
---|
863 | int extra_base; /* base index for extra_bits */
|
---|
864 | int elems; /* max number of elements in the tree */
|
---|
865 | int max_length; /* max bit length for the codes */
|
---|
866 | int max_code; /* largest code with non zero frequency */
|
---|
867 | } tree_desc;
|
---|
868 |
|
---|
869 | struct globals2 {
|
---|
870 |
|
---|
871 | ush heap[HEAP_SIZE]; /* heap used to build the Huffman trees */
|
---|
872 | int heap_len; /* number of elements in the heap */
|
---|
873 | int heap_max; /* element of largest frequency */
|
---|
874 |
|
---|
875 | /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used.
|
---|
876 | * The same heap array is used to build all trees.
|
---|
877 | */
|
---|
878 |
|
---|
879 | ct_data dyn_ltree[HEAP_SIZE]; /* literal and length tree */
|
---|
880 | ct_data dyn_dtree[2 * D_CODES + 1]; /* distance tree */
|
---|
881 |
|
---|
882 | ct_data static_ltree[L_CODES + 2];
|
---|
883 |
|
---|
884 | /* The static literal tree. Since the bit lengths are imposed, there is no
|
---|
885 | * need for the L_CODES extra codes used during heap construction. However
|
---|
886 | * The codes 286 and 287 are needed to build a canonical tree (see ct_init
|
---|
887 | * below).
|
---|
888 | */
|
---|
889 |
|
---|
890 | ct_data static_dtree[D_CODES];
|
---|
891 |
|
---|
892 | /* The static distance tree. (Actually a trivial tree since all codes use
|
---|
893 | * 5 bits.)
|
---|
894 | */
|
---|
895 |
|
---|
896 | ct_data bl_tree[2 * BL_CODES + 1];
|
---|
897 |
|
---|
898 | /* Huffman tree for the bit lengths */
|
---|
899 |
|
---|
900 | tree_desc l_desc;
|
---|
901 | tree_desc d_desc;
|
---|
902 | tree_desc bl_desc;
|
---|
903 |
|
---|
904 | ush bl_count[MAX_BITS + 1];
|
---|
905 |
|
---|
906 | /* The lengths of the bit length codes are sent in order of decreasing
|
---|
907 | * probability, to avoid transmitting the lengths for unused bit length codes.
|
---|
908 | */
|
---|
909 |
|
---|
910 | uch depth[2 * L_CODES + 1];
|
---|
911 |
|
---|
912 | /* Depth of each subtree used as tie breaker for trees of equal frequency */
|
---|
913 |
|
---|
914 | uch length_code[MAX_MATCH - MIN_MATCH + 1];
|
---|
915 |
|
---|
916 | /* length code for each normalized match length (0 == MIN_MATCH) */
|
---|
917 |
|
---|
918 | uch dist_code[512];
|
---|
919 |
|
---|
920 | /* distance codes. The first 256 values correspond to the distances
|
---|
921 | * 3 .. 258, the last 256 values correspond to the top 8 bits of
|
---|
922 | * the 15 bit distances.
|
---|
923 | */
|
---|
924 |
|
---|
925 | int base_length[LENGTH_CODES];
|
---|
926 |
|
---|
927 | /* First normalized length for each code (0 = MIN_MATCH) */
|
---|
928 |
|
---|
929 | int base_dist[D_CODES];
|
---|
930 |
|
---|
931 | /* First normalized distance for each code (0 = distance of 1) */
|
---|
932 |
|
---|
933 | uch flag_buf[LIT_BUFSIZE / 8];
|
---|
934 |
|
---|
935 | /* flag_buf is a bit array distinguishing literals from lengths in
|
---|
936 | * l_buf, thus indicating the presence or absence of a distance.
|
---|
937 | */
|
---|
938 |
|
---|
939 | unsigned last_lit; /* running index in l_buf */
|
---|
940 | unsigned last_dist; /* running index in d_buf */
|
---|
941 | unsigned last_flags; /* running index in flag_buf */
|
---|
942 | uch flags; /* current flags not yet saved in flag_buf */
|
---|
943 | uch flag_bit; /* current bit used in flags */
|
---|
944 |
|
---|
945 | /* bits are filled in flags starting at bit 0 (least significant).
|
---|
946 | * Note: these flags are overkill in the current code since we don't
|
---|
947 | * take advantage of DIST_BUFSIZE == LIT_BUFSIZE.
|
---|
948 | */
|
---|
949 |
|
---|
950 | ulg opt_len; /* bit length of current block with optimal trees */
|
---|
951 | ulg static_len; /* bit length of current block with static trees */
|
---|
952 |
|
---|
953 | ulg compressed_len; /* total bit length of compressed file */
|
---|
954 | };
|
---|
955 |
|
---|
956 | #define G2ptr ((struct globals2*)(ptr_to_globals))
|
---|
957 | #define G2 (*G2ptr)
|
---|
958 |
|
---|
959 |
|
---|
960 | /* ===========================================================================
|
---|
961 | */
|
---|
962 | static void gen_codes(ct_data * tree, int max_code);
|
---|
963 | static void build_tree(tree_desc * desc);
|
---|
964 | static void scan_tree(ct_data * tree, int max_code);
|
---|
965 | static void send_tree(ct_data * tree, int max_code);
|
---|
966 | static int build_bl_tree(void);
|
---|
967 | static void send_all_trees(int lcodes, int dcodes, int blcodes);
|
---|
968 | static void compress_block(ct_data * ltree, ct_data * dtree);
|
---|
969 |
|
---|
970 |
|
---|
971 | #ifndef DEBUG
|
---|
972 | /* Send a code of the given tree. c and tree must not have side effects */
|
---|
973 | # define SEND_CODE(c, tree) send_bits(tree[c].Code, tree[c].Len)
|
---|
974 | #else
|
---|
975 | # define SEND_CODE(c, tree) \
|
---|
976 | { \
|
---|
977 | if (verbose > 1) bb_error_msg("\ncd %3d ",(c)); \
|
---|
978 | send_bits(tree[c].Code, tree[c].Len); \
|
---|
979 | }
|
---|
980 | #endif
|
---|
981 |
|
---|
982 | #define D_CODE(dist) \
|
---|
983 | ((dist) < 256 ? G2.dist_code[dist] : G2.dist_code[256 + ((dist)>>7)])
|
---|
984 | /* Mapping from a distance to a distance code. dist is the distance - 1 and
|
---|
985 | * must not have side effects. dist_code[256] and dist_code[257] are never
|
---|
986 | * used.
|
---|
987 | * The arguments must not have side effects.
|
---|
988 | */
|
---|
989 |
|
---|
990 |
|
---|
991 | /* ===========================================================================
|
---|
992 | * Initialize a new block.
|
---|
993 | */
|
---|
994 | static void init_block(void)
|
---|
995 | {
|
---|
996 | int n; /* iterates over tree elements */
|
---|
997 |
|
---|
998 | /* Initialize the trees. */
|
---|
999 | for (n = 0; n < L_CODES; n++)
|
---|
1000 | G2.dyn_ltree[n].Freq = 0;
|
---|
1001 | for (n = 0; n < D_CODES; n++)
|
---|
1002 | G2.dyn_dtree[n].Freq = 0;
|
---|
1003 | for (n = 0; n < BL_CODES; n++)
|
---|
1004 | G2.bl_tree[n].Freq = 0;
|
---|
1005 |
|
---|
1006 | G2.dyn_ltree[END_BLOCK].Freq = 1;
|
---|
1007 | G2.opt_len = G2.static_len = 0;
|
---|
1008 | G2.last_lit = G2.last_dist = G2.last_flags = 0;
|
---|
1009 | G2.flags = 0;
|
---|
1010 | G2.flag_bit = 1;
|
---|
1011 | }
|
---|
1012 |
|
---|
1013 |
|
---|
1014 | /* ===========================================================================
|
---|
1015 | * Restore the heap property by moving down the tree starting at node k,
|
---|
1016 | * exchanging a node with the smallest of its two sons if necessary, stopping
|
---|
1017 | * when the heap property is re-established (each father smaller than its
|
---|
1018 | * two sons).
|
---|
1019 | */
|
---|
1020 |
|
---|
1021 | /* Compares to subtrees, using the tree depth as tie breaker when
|
---|
1022 | * the subtrees have equal frequency. This minimizes the worst case length. */
|
---|
1023 | #define SMALLER(tree, n, m) \
|
---|
1024 | (tree[n].Freq < tree[m].Freq \
|
---|
1025 | || (tree[n].Freq == tree[m].Freq && G2.depth[n] <= G2.depth[m]))
|
---|
1026 |
|
---|
1027 | static void pqdownheap(ct_data * tree, int k)
|
---|
1028 | {
|
---|
1029 | int v = G2.heap[k];
|
---|
1030 | int j = k << 1; /* left son of k */
|
---|
1031 |
|
---|
1032 | while (j <= G2.heap_len) {
|
---|
1033 | /* Set j to the smallest of the two sons: */
|
---|
1034 | if (j < G2.heap_len && SMALLER(tree, G2.heap[j + 1], G2.heap[j]))
|
---|
1035 | j++;
|
---|
1036 |
|
---|
1037 | /* Exit if v is smaller than both sons */
|
---|
1038 | if (SMALLER(tree, v, G2.heap[j]))
|
---|
1039 | break;
|
---|
1040 |
|
---|
1041 | /* Exchange v with the smallest son */
|
---|
1042 | G2.heap[k] = G2.heap[j];
|
---|
1043 | k = j;
|
---|
1044 |
|
---|
1045 | /* And continue down the tree, setting j to the left son of k */
|
---|
1046 | j <<= 1;
|
---|
1047 | }
|
---|
1048 | G2.heap[k] = v;
|
---|
1049 | }
|
---|
1050 |
|
---|
1051 |
|
---|
1052 | /* ===========================================================================
|
---|
1053 | * Compute the optimal bit lengths for a tree and update the total bit length
|
---|
1054 | * for the current block.
|
---|
1055 | * IN assertion: the fields freq and dad are set, heap[heap_max] and
|
---|
1056 | * above are the tree nodes sorted by increasing frequency.
|
---|
1057 | * OUT assertions: the field len is set to the optimal bit length, the
|
---|
1058 | * array bl_count contains the frequencies for each bit length.
|
---|
1059 | * The length opt_len is updated; static_len is also updated if stree is
|
---|
1060 | * not null.
|
---|
1061 | */
|
---|
1062 | static void gen_bitlen(tree_desc * desc)
|
---|
1063 | {
|
---|
1064 | ct_data *tree = desc->dyn_tree;
|
---|
1065 | const uint8_t *extra = desc->extra_bits;
|
---|
1066 | int base = desc->extra_base;
|
---|
1067 | int max_code = desc->max_code;
|
---|
1068 | int max_length = desc->max_length;
|
---|
1069 | ct_data *stree = desc->static_tree;
|
---|
1070 | int h; /* heap index */
|
---|
1071 | int n, m; /* iterate over the tree elements */
|
---|
1072 | int bits; /* bit length */
|
---|
1073 | int xbits; /* extra bits */
|
---|
1074 | ush f; /* frequency */
|
---|
1075 | int overflow = 0; /* number of elements with bit length too large */
|
---|
1076 |
|
---|
1077 | for (bits = 0; bits <= MAX_BITS; bits++)
|
---|
1078 | G2.bl_count[bits] = 0;
|
---|
1079 |
|
---|
1080 | /* In a first pass, compute the optimal bit lengths (which may
|
---|
1081 | * overflow in the case of the bit length tree).
|
---|
1082 | */
|
---|
1083 | tree[G2.heap[G2.heap_max]].Len = 0; /* root of the heap */
|
---|
1084 |
|
---|
1085 | for (h = G2.heap_max + 1; h < HEAP_SIZE; h++) {
|
---|
1086 | n = G2.heap[h];
|
---|
1087 | bits = tree[tree[n].Dad].Len + 1;
|
---|
1088 | if (bits > max_length) {
|
---|
1089 | bits = max_length;
|
---|
1090 | overflow++;
|
---|
1091 | }
|
---|
1092 | tree[n].Len = (ush) bits;
|
---|
1093 | /* We overwrite tree[n].Dad which is no longer needed */
|
---|
1094 |
|
---|
1095 | if (n > max_code)
|
---|
1096 | continue; /* not a leaf node */
|
---|
1097 |
|
---|
1098 | G2.bl_count[bits]++;
|
---|
1099 | xbits = 0;
|
---|
1100 | if (n >= base)
|
---|
1101 | xbits = extra[n - base];
|
---|
1102 | f = tree[n].Freq;
|
---|
1103 | G2.opt_len += (ulg) f *(bits + xbits);
|
---|
1104 |
|
---|
1105 | if (stree)
|
---|
1106 | G2.static_len += (ulg) f * (stree[n].Len + xbits);
|
---|
1107 | }
|
---|
1108 | if (overflow == 0)
|
---|
1109 | return;
|
---|
1110 |
|
---|
1111 | Trace((stderr, "\nbit length overflow\n"));
|
---|
1112 | /* This happens for example on obj2 and pic of the Calgary corpus */
|
---|
1113 |
|
---|
1114 | /* Find the first bit length which could increase: */
|
---|
1115 | do {
|
---|
1116 | bits = max_length - 1;
|
---|
1117 | while (G2.bl_count[bits] == 0)
|
---|
1118 | bits--;
|
---|
1119 | G2.bl_count[bits]--; /* move one leaf down the tree */
|
---|
1120 | G2.bl_count[bits + 1] += 2; /* move one overflow item as its brother */
|
---|
1121 | G2.bl_count[max_length]--;
|
---|
1122 | /* The brother of the overflow item also moves one step up,
|
---|
1123 | * but this does not affect bl_count[max_length]
|
---|
1124 | */
|
---|
1125 | overflow -= 2;
|
---|
1126 | } while (overflow > 0);
|
---|
1127 |
|
---|
1128 | /* Now recompute all bit lengths, scanning in increasing frequency.
|
---|
1129 | * h is still equal to HEAP_SIZE. (It is simpler to reconstruct all
|
---|
1130 | * lengths instead of fixing only the wrong ones. This idea is taken
|
---|
1131 | * from 'ar' written by Haruhiko Okumura.)
|
---|
1132 | */
|
---|
1133 | for (bits = max_length; bits != 0; bits--) {
|
---|
1134 | n = G2.bl_count[bits];
|
---|
1135 | while (n != 0) {
|
---|
1136 | m = G2.heap[--h];
|
---|
1137 | if (m > max_code)
|
---|
1138 | continue;
|
---|
1139 | if (tree[m].Len != (unsigned) bits) {
|
---|
1140 | Trace((stderr, "code %d bits %d->%d\n", m, tree[m].Len, bits));
|
---|
1141 | G2.opt_len += ((int32_t) bits - tree[m].Len) * tree[m].Freq;
|
---|
1142 | tree[m].Len = bits;
|
---|
1143 | }
|
---|
1144 | n--;
|
---|
1145 | }
|
---|
1146 | }
|
---|
1147 | }
|
---|
1148 |
|
---|
1149 |
|
---|
1150 | /* ===========================================================================
|
---|
1151 | * Generate the codes for a given tree and bit counts (which need not be
|
---|
1152 | * optimal).
|
---|
1153 | * IN assertion: the array bl_count contains the bit length statistics for
|
---|
1154 | * the given tree and the field len is set for all tree elements.
|
---|
1155 | * OUT assertion: the field code is set for all tree elements of non
|
---|
1156 | * zero code length.
|
---|
1157 | */
|
---|
1158 | static void gen_codes(ct_data * tree, int max_code)
|
---|
1159 | {
|
---|
1160 | ush next_code[MAX_BITS + 1]; /* next code value for each bit length */
|
---|
1161 | ush code = 0; /* running code value */
|
---|
1162 | int bits; /* bit index */
|
---|
1163 | int n; /* code index */
|
---|
1164 |
|
---|
1165 | /* The distribution counts are first used to generate the code values
|
---|
1166 | * without bit reversal.
|
---|
1167 | */
|
---|
1168 | for (bits = 1; bits <= MAX_BITS; bits++) {
|
---|
1169 | next_code[bits] = code = (code + G2.bl_count[bits - 1]) << 1;
|
---|
1170 | }
|
---|
1171 | /* Check that the bit counts in bl_count are consistent. The last code
|
---|
1172 | * must be all ones.
|
---|
1173 | */
|
---|
1174 | Assert(code + G2.bl_count[MAX_BITS] - 1 == (1 << MAX_BITS) - 1,
|
---|
1175 | "inconsistent bit counts");
|
---|
1176 | Tracev((stderr, "\ngen_codes: max_code %d ", max_code));
|
---|
1177 |
|
---|
1178 | for (n = 0; n <= max_code; n++) {
|
---|
1179 | int len = tree[n].Len;
|
---|
1180 |
|
---|
1181 | if (len == 0)
|
---|
1182 | continue;
|
---|
1183 | /* Now reverse the bits */
|
---|
1184 | tree[n].Code = bi_reverse(next_code[len]++, len);
|
---|
1185 |
|
---|
1186 | Tracec(tree != G2.static_ltree,
|
---|
1187 | (stderr, "\nn %3d %c l %2d c %4x (%x) ", n,
|
---|
1188 | (isgraph(n) ? n : ' '), len, tree[n].Code,
|
---|
1189 | next_code[len] - 1));
|
---|
1190 | }
|
---|
1191 | }
|
---|
1192 |
|
---|
1193 |
|
---|
1194 | /* ===========================================================================
|
---|
1195 | * Construct one Huffman tree and assigns the code bit strings and lengths.
|
---|
1196 | * Update the total bit length for the current block.
|
---|
1197 | * IN assertion: the field freq is set for all tree elements.
|
---|
1198 | * OUT assertions: the fields len and code are set to the optimal bit length
|
---|
1199 | * and corresponding code. The length opt_len is updated; static_len is
|
---|
1200 | * also updated if stree is not null. The field max_code is set.
|
---|
1201 | */
|
---|
1202 |
|
---|
1203 | /* Remove the smallest element from the heap and recreate the heap with
|
---|
1204 | * one less element. Updates heap and heap_len. */
|
---|
1205 |
|
---|
1206 | #define SMALLEST 1
|
---|
1207 | /* Index within the heap array of least frequent node in the Huffman tree */
|
---|
1208 |
|
---|
1209 | #define PQREMOVE(tree, top) \
|
---|
1210 | do { \
|
---|
1211 | top = G2.heap[SMALLEST]; \
|
---|
1212 | G2.heap[SMALLEST] = G2.heap[G2.heap_len--]; \
|
---|
1213 | pqdownheap(tree, SMALLEST); \
|
---|
1214 | } while (0)
|
---|
1215 |
|
---|
1216 | static void build_tree(tree_desc * desc)
|
---|
1217 | {
|
---|
1218 | ct_data *tree = desc->dyn_tree;
|
---|
1219 | ct_data *stree = desc->static_tree;
|
---|
1220 | int elems = desc->elems;
|
---|
1221 | int n, m; /* iterate over heap elements */
|
---|
1222 | int max_code = -1; /* largest code with non zero frequency */
|
---|
1223 | int node = elems; /* next internal node of the tree */
|
---|
1224 |
|
---|
1225 | /* Construct the initial heap, with least frequent element in
|
---|
1226 | * heap[SMALLEST]. The sons of heap[n] are heap[2*n] and heap[2*n+1].
|
---|
1227 | * heap[0] is not used.
|
---|
1228 | */
|
---|
1229 | G2.heap_len = 0;
|
---|
1230 | G2.heap_max = HEAP_SIZE;
|
---|
1231 |
|
---|
1232 | for (n = 0; n < elems; n++) {
|
---|
1233 | if (tree[n].Freq != 0) {
|
---|
1234 | G2.heap[++G2.heap_len] = max_code = n;
|
---|
1235 | G2.depth[n] = 0;
|
---|
1236 | } else {
|
---|
1237 | tree[n].Len = 0;
|
---|
1238 | }
|
---|
1239 | }
|
---|
1240 |
|
---|
1241 | /* The pkzip format requires that at least one distance code exists,
|
---|
1242 | * and that at least one bit should be sent even if there is only one
|
---|
1243 | * possible code. So to avoid special checks later on we force at least
|
---|
1244 | * two codes of non zero frequency.
|
---|
1245 | */
|
---|
1246 | while (G2.heap_len < 2) {
|
---|
1247 | int new = G2.heap[++G2.heap_len] = (max_code < 2 ? ++max_code : 0);
|
---|
1248 |
|
---|
1249 | tree[new].Freq = 1;
|
---|
1250 | G2.depth[new] = 0;
|
---|
1251 | G2.opt_len--;
|
---|
1252 | if (stree)
|
---|
1253 | G2.static_len -= stree[new].Len;
|
---|
1254 | /* new is 0 or 1 so it does not have extra bits */
|
---|
1255 | }
|
---|
1256 | desc->max_code = max_code;
|
---|
1257 |
|
---|
1258 | /* The elements heap[heap_len/2+1 .. heap_len] are leaves of the tree,
|
---|
1259 | * establish sub-heaps of increasing lengths:
|
---|
1260 | */
|
---|
1261 | for (n = G2.heap_len / 2; n >= 1; n--)
|
---|
1262 | pqdownheap(tree, n);
|
---|
1263 |
|
---|
1264 | /* Construct the Huffman tree by repeatedly combining the least two
|
---|
1265 | * frequent nodes.
|
---|
1266 | */
|
---|
1267 | do {
|
---|
1268 | PQREMOVE(tree, n); /* n = node of least frequency */
|
---|
1269 | m = G2.heap[SMALLEST]; /* m = node of next least frequency */
|
---|
1270 |
|
---|
1271 | G2.heap[--G2.heap_max] = n; /* keep the nodes sorted by frequency */
|
---|
1272 | G2.heap[--G2.heap_max] = m;
|
---|
1273 |
|
---|
1274 | /* Create a new node father of n and m */
|
---|
1275 | tree[node].Freq = tree[n].Freq + tree[m].Freq;
|
---|
1276 | G2.depth[node] = MAX(G2.depth[n], G2.depth[m]) + 1;
|
---|
1277 | tree[n].Dad = tree[m].Dad = (ush) node;
|
---|
1278 | #ifdef DUMP_BL_TREE
|
---|
1279 | if (tree == G2.bl_tree) {
|
---|
1280 | bb_error_msg("\nnode %d(%d), sons %d(%d) %d(%d)",
|
---|
1281 | node, tree[node].Freq, n, tree[n].Freq, m, tree[m].Freq);
|
---|
1282 | }
|
---|
1283 | #endif
|
---|
1284 | /* and insert the new node in the heap */
|
---|
1285 | G2.heap[SMALLEST] = node++;
|
---|
1286 | pqdownheap(tree, SMALLEST);
|
---|
1287 |
|
---|
1288 | } while (G2.heap_len >= 2);
|
---|
1289 |
|
---|
1290 | G2.heap[--G2.heap_max] = G2.heap[SMALLEST];
|
---|
1291 |
|
---|
1292 | /* At this point, the fields freq and dad are set. We can now
|
---|
1293 | * generate the bit lengths.
|
---|
1294 | */
|
---|
1295 | gen_bitlen((tree_desc *) desc);
|
---|
1296 |
|
---|
1297 | /* The field len is now set, we can generate the bit codes */
|
---|
1298 | gen_codes((ct_data *) tree, max_code);
|
---|
1299 | }
|
---|
1300 |
|
---|
1301 |
|
---|
1302 | /* ===========================================================================
|
---|
1303 | * Scan a literal or distance tree to determine the frequencies of the codes
|
---|
1304 | * in the bit length tree. Updates opt_len to take into account the repeat
|
---|
1305 | * counts. (The contribution of the bit length codes will be added later
|
---|
1306 | * during the construction of bl_tree.)
|
---|
1307 | */
|
---|
1308 | static void scan_tree(ct_data * tree, int max_code)
|
---|
1309 | {
|
---|
1310 | int n; /* iterates over all tree elements */
|
---|
1311 | int prevlen = -1; /* last emitted length */
|
---|
1312 | int curlen; /* length of current code */
|
---|
1313 | int nextlen = tree[0].Len; /* length of next code */
|
---|
1314 | int count = 0; /* repeat count of the current code */
|
---|
1315 | int max_count = 7; /* max repeat count */
|
---|
1316 | int min_count = 4; /* min repeat count */
|
---|
1317 |
|
---|
1318 | if (nextlen == 0) {
|
---|
1319 | max_count = 138;
|
---|
1320 | min_count = 3;
|
---|
1321 | }
|
---|
1322 | tree[max_code + 1].Len = 0xffff; /* guard */
|
---|
1323 |
|
---|
1324 | for (n = 0; n <= max_code; n++) {
|
---|
1325 | curlen = nextlen;
|
---|
1326 | nextlen = tree[n + 1].Len;
|
---|
1327 | if (++count < max_count && curlen == nextlen)
|
---|
1328 | continue;
|
---|
1329 |
|
---|
1330 | if (count < min_count) {
|
---|
1331 | G2.bl_tree[curlen].Freq += count;
|
---|
1332 | } else if (curlen != 0) {
|
---|
1333 | if (curlen != prevlen)
|
---|
1334 | G2.bl_tree[curlen].Freq++;
|
---|
1335 | G2.bl_tree[REP_3_6].Freq++;
|
---|
1336 | } else if (count <= 10) {
|
---|
1337 | G2.bl_tree[REPZ_3_10].Freq++;
|
---|
1338 | } else {
|
---|
1339 | G2.bl_tree[REPZ_11_138].Freq++;
|
---|
1340 | }
|
---|
1341 | count = 0;
|
---|
1342 | prevlen = curlen;
|
---|
1343 |
|
---|
1344 | max_count = 7;
|
---|
1345 | min_count = 4;
|
---|
1346 | if (nextlen == 0) {
|
---|
1347 | max_count = 138;
|
---|
1348 | min_count = 3;
|
---|
1349 | } else if (curlen == nextlen) {
|
---|
1350 | max_count = 6;
|
---|
1351 | min_count = 3;
|
---|
1352 | }
|
---|
1353 | }
|
---|
1354 | }
|
---|
1355 |
|
---|
1356 |
|
---|
1357 | /* ===========================================================================
|
---|
1358 | * Send a literal or distance tree in compressed form, using the codes in
|
---|
1359 | * bl_tree.
|
---|
1360 | */
|
---|
1361 | static void send_tree(ct_data * tree, int max_code)
|
---|
1362 | {
|
---|
1363 | int n; /* iterates over all tree elements */
|
---|
1364 | int prevlen = -1; /* last emitted length */
|
---|
1365 | int curlen; /* length of current code */
|
---|
1366 | int nextlen = tree[0].Len; /* length of next code */
|
---|
1367 | int count = 0; /* repeat count of the current code */
|
---|
1368 | int max_count = 7; /* max repeat count */
|
---|
1369 | int min_count = 4; /* min repeat count */
|
---|
1370 |
|
---|
1371 | /* tree[max_code+1].Len = -1; *//* guard already set */
|
---|
1372 | if (nextlen == 0)
|
---|
1373 | max_count = 138, min_count = 3;
|
---|
1374 |
|
---|
1375 | for (n = 0; n <= max_code; n++) {
|
---|
1376 | curlen = nextlen;
|
---|
1377 | nextlen = tree[n + 1].Len;
|
---|
1378 | if (++count < max_count && curlen == nextlen) {
|
---|
1379 | continue;
|
---|
1380 | } else if (count < min_count) {
|
---|
1381 | do {
|
---|
1382 | SEND_CODE(curlen, G2.bl_tree);
|
---|
1383 | } while (--count);
|
---|
1384 | } else if (curlen != 0) {
|
---|
1385 | if (curlen != prevlen) {
|
---|
1386 | SEND_CODE(curlen, G2.bl_tree);
|
---|
1387 | count--;
|
---|
1388 | }
|
---|
1389 | Assert(count >= 3 && count <= 6, " 3_6?");
|
---|
1390 | SEND_CODE(REP_3_6, G2.bl_tree);
|
---|
1391 | send_bits(count - 3, 2);
|
---|
1392 | } else if (count <= 10) {
|
---|
1393 | SEND_CODE(REPZ_3_10, G2.bl_tree);
|
---|
1394 | send_bits(count - 3, 3);
|
---|
1395 | } else {
|
---|
1396 | SEND_CODE(REPZ_11_138, G2.bl_tree);
|
---|
1397 | send_bits(count - 11, 7);
|
---|
1398 | }
|
---|
1399 | count = 0;
|
---|
1400 | prevlen = curlen;
|
---|
1401 | if (nextlen == 0) {
|
---|
1402 | max_count = 138;
|
---|
1403 | min_count = 3;
|
---|
1404 | } else if (curlen == nextlen) {
|
---|
1405 | max_count = 6;
|
---|
1406 | min_count = 3;
|
---|
1407 | } else {
|
---|
1408 | max_count = 7;
|
---|
1409 | min_count = 4;
|
---|
1410 | }
|
---|
1411 | }
|
---|
1412 | }
|
---|
1413 |
|
---|
1414 |
|
---|
1415 | /* ===========================================================================
|
---|
1416 | * Construct the Huffman tree for the bit lengths and return the index in
|
---|
1417 | * bl_order of the last bit length code to send.
|
---|
1418 | */
|
---|
1419 | static int build_bl_tree(void)
|
---|
1420 | {
|
---|
1421 | int max_blindex; /* index of last bit length code of non zero freq */
|
---|
1422 |
|
---|
1423 | /* Determine the bit length frequencies for literal and distance trees */
|
---|
1424 | scan_tree(G2.dyn_ltree, G2.l_desc.max_code);
|
---|
1425 | scan_tree(G2.dyn_dtree, G2.d_desc.max_code);
|
---|
1426 |
|
---|
1427 | /* Build the bit length tree: */
|
---|
1428 | build_tree(&G2.bl_desc);
|
---|
1429 | /* opt_len now includes the length of the tree representations, except
|
---|
1430 | * the lengths of the bit lengths codes and the 5+5+4 bits for the counts.
|
---|
1431 | */
|
---|
1432 |
|
---|
1433 | /* Determine the number of bit length codes to send. The pkzip format
|
---|
1434 | * requires that at least 4 bit length codes be sent. (appnote.txt says
|
---|
1435 | * 3 but the actual value used is 4.)
|
---|
1436 | */
|
---|
1437 | for (max_blindex = BL_CODES - 1; max_blindex >= 3; max_blindex--) {
|
---|
1438 | if (G2.bl_tree[bl_order[max_blindex]].Len != 0)
|
---|
1439 | break;
|
---|
1440 | }
|
---|
1441 | /* Update opt_len to include the bit length tree and counts */
|
---|
1442 | G2.opt_len += 3 * (max_blindex + 1) + 5 + 5 + 4;
|
---|
1443 | Tracev((stderr, "\ndyn trees: dyn %ld, stat %ld", G2.opt_len, G2.static_len));
|
---|
1444 |
|
---|
1445 | return max_blindex;
|
---|
1446 | }
|
---|
1447 |
|
---|
1448 |
|
---|
1449 | /* ===========================================================================
|
---|
1450 | * Send the header for a block using dynamic Huffman trees: the counts, the
|
---|
1451 | * lengths of the bit length codes, the literal tree and the distance tree.
|
---|
1452 | * IN assertion: lcodes >= 257, dcodes >= 1, blcodes >= 4.
|
---|
1453 | */
|
---|
1454 | static void send_all_trees(int lcodes, int dcodes, int blcodes)
|
---|
1455 | {
|
---|
1456 | int rank; /* index in bl_order */
|
---|
1457 |
|
---|
1458 | Assert(lcodes >= 257 && dcodes >= 1 && blcodes >= 4, "not enough codes");
|
---|
1459 | Assert(lcodes <= L_CODES && dcodes <= D_CODES
|
---|
1460 | && blcodes <= BL_CODES, "too many codes");
|
---|
1461 | Tracev((stderr, "\nbl counts: "));
|
---|
1462 | send_bits(lcodes - 257, 5); /* not +255 as stated in appnote.txt */
|
---|
1463 | send_bits(dcodes - 1, 5);
|
---|
1464 | send_bits(blcodes - 4, 4); /* not -3 as stated in appnote.txt */
|
---|
1465 | for (rank = 0; rank < blcodes; rank++) {
|
---|
1466 | Tracev((stderr, "\nbl code %2d ", bl_order[rank]));
|
---|
1467 | send_bits(G2.bl_tree[bl_order[rank]].Len, 3);
|
---|
1468 | }
|
---|
1469 | Tracev((stderr, "\nbl tree: sent %ld", G1.bits_sent));
|
---|
1470 |
|
---|
1471 | send_tree((ct_data *) G2.dyn_ltree, lcodes - 1); /* send the literal tree */
|
---|
1472 | Tracev((stderr, "\nlit tree: sent %ld", G1.bits_sent));
|
---|
1473 |
|
---|
1474 | send_tree((ct_data *) G2.dyn_dtree, dcodes - 1); /* send the distance tree */
|
---|
1475 | Tracev((stderr, "\ndist tree: sent %ld", G1.bits_sent));
|
---|
1476 | }
|
---|
1477 |
|
---|
1478 |
|
---|
1479 | /* ===========================================================================
|
---|
1480 | * Save the match info and tally the frequency counts. Return true if
|
---|
1481 | * the current block must be flushed.
|
---|
1482 | */
|
---|
1483 | static int ct_tally(int dist, int lc)
|
---|
1484 | {
|
---|
1485 | G1.l_buf[G2.last_lit++] = lc;
|
---|
1486 | if (dist == 0) {
|
---|
1487 | /* lc is the unmatched char */
|
---|
1488 | G2.dyn_ltree[lc].Freq++;
|
---|
1489 | } else {
|
---|
1490 | /* Here, lc is the match length - MIN_MATCH */
|
---|
1491 | dist--; /* dist = match distance - 1 */
|
---|
1492 | Assert((ush) dist < (ush) MAX_DIST
|
---|
1493 | && (ush) lc <= (ush) (MAX_MATCH - MIN_MATCH)
|
---|
1494 | && (ush) D_CODE(dist) < (ush) D_CODES, "ct_tally: bad match"
|
---|
1495 | );
|
---|
1496 |
|
---|
1497 | G2.dyn_ltree[G2.length_code[lc] + LITERALS + 1].Freq++;
|
---|
1498 | G2.dyn_dtree[D_CODE(dist)].Freq++;
|
---|
1499 |
|
---|
1500 | G1.d_buf[G2.last_dist++] = dist;
|
---|
1501 | G2.flags |= G2.flag_bit;
|
---|
1502 | }
|
---|
1503 | G2.flag_bit <<= 1;
|
---|
1504 |
|
---|
1505 | /* Output the flags if they fill a byte: */
|
---|
1506 | if ((G2.last_lit & 7) == 0) {
|
---|
1507 | G2.flag_buf[G2.last_flags++] = G2.flags;
|
---|
1508 | G2.flags = 0;
|
---|
1509 | G2.flag_bit = 1;
|
---|
1510 | }
|
---|
1511 | /* Try to guess if it is profitable to stop the current block here */
|
---|
1512 | if ((G2.last_lit & 0xfff) == 0) {
|
---|
1513 | /* Compute an upper bound for the compressed length */
|
---|
1514 | ulg out_length = G2.last_lit * 8L;
|
---|
1515 | ulg in_length = (ulg) G1.strstart - G1.block_start;
|
---|
1516 | int dcode;
|
---|
1517 |
|
---|
1518 | for (dcode = 0; dcode < D_CODES; dcode++) {
|
---|
1519 | out_length += G2.dyn_dtree[dcode].Freq * (5L + extra_dbits[dcode]);
|
---|
1520 | }
|
---|
1521 | out_length >>= 3;
|
---|
1522 | Trace((stderr,
|
---|
1523 | "\nlast_lit %u, last_dist %u, in %ld, out ~%ld(%ld%%) ",
|
---|
1524 | G2.last_lit, G2.last_dist, in_length, out_length,
|
---|
1525 | 100L - out_length * 100L / in_length));
|
---|
1526 | if (G2.last_dist < G2.last_lit / 2 && out_length < in_length / 2)
|
---|
1527 | return 1;
|
---|
1528 | }
|
---|
1529 | return (G2.last_lit == LIT_BUFSIZE - 1 || G2.last_dist == DIST_BUFSIZE);
|
---|
1530 | /* We avoid equality with LIT_BUFSIZE because of wraparound at 64K
|
---|
1531 | * on 16 bit machines and because stored blocks are restricted to
|
---|
1532 | * 64K-1 bytes.
|
---|
1533 | */
|
---|
1534 | }
|
---|
1535 |
|
---|
1536 | /* ===========================================================================
|
---|
1537 | * Send the block data compressed using the given Huffman trees
|
---|
1538 | */
|
---|
1539 | static void compress_block(ct_data * ltree, ct_data * dtree)
|
---|
1540 | {
|
---|
1541 | unsigned dist; /* distance of matched string */
|
---|
1542 | int lc; /* match length or unmatched char (if dist == 0) */
|
---|
1543 | unsigned lx = 0; /* running index in l_buf */
|
---|
1544 | unsigned dx = 0; /* running index in d_buf */
|
---|
1545 | unsigned fx = 0; /* running index in flag_buf */
|
---|
1546 | uch flag = 0; /* current flags */
|
---|
1547 | unsigned code; /* the code to send */
|
---|
1548 | int extra; /* number of extra bits to send */
|
---|
1549 |
|
---|
1550 | if (G2.last_lit != 0) do {
|
---|
1551 | if ((lx & 7) == 0)
|
---|
1552 | flag = G2.flag_buf[fx++];
|
---|
1553 | lc = G1.l_buf[lx++];
|
---|
1554 | if ((flag & 1) == 0) {
|
---|
1555 | SEND_CODE(lc, ltree); /* send a literal byte */
|
---|
1556 | Tracecv(isgraph(lc), (stderr, " '%c' ", lc));
|
---|
1557 | } else {
|
---|
1558 | /* Here, lc is the match length - MIN_MATCH */
|
---|
1559 | code = G2.length_code[lc];
|
---|
1560 | SEND_CODE(code + LITERALS + 1, ltree); /* send the length code */
|
---|
1561 | extra = extra_lbits[code];
|
---|
1562 | if (extra != 0) {
|
---|
1563 | lc -= G2.base_length[code];
|
---|
1564 | send_bits(lc, extra); /* send the extra length bits */
|
---|
1565 | }
|
---|
1566 | dist = G1.d_buf[dx++];
|
---|
1567 | /* Here, dist is the match distance - 1 */
|
---|
1568 | code = D_CODE(dist);
|
---|
1569 | Assert(code < D_CODES, "bad d_code");
|
---|
1570 |
|
---|
1571 | SEND_CODE(code, dtree); /* send the distance code */
|
---|
1572 | extra = extra_dbits[code];
|
---|
1573 | if (extra != 0) {
|
---|
1574 | dist -= G2.base_dist[code];
|
---|
1575 | send_bits(dist, extra); /* send the extra distance bits */
|
---|
1576 | }
|
---|
1577 | } /* literal or match pair ? */
|
---|
1578 | flag >>= 1;
|
---|
1579 | } while (lx < G2.last_lit);
|
---|
1580 |
|
---|
1581 | SEND_CODE(END_BLOCK, ltree);
|
---|
1582 | }
|
---|
1583 |
|
---|
1584 |
|
---|
1585 | /* ===========================================================================
|
---|
1586 | * Determine the best encoding for the current block: dynamic trees, static
|
---|
1587 | * trees or store, and output the encoded block to the zip file. This function
|
---|
1588 | * returns the total compressed length for the file so far.
|
---|
1589 | */
|
---|
1590 | static ulg flush_block(char *buf, ulg stored_len, int eof)
|
---|
1591 | {
|
---|
1592 | ulg opt_lenb, static_lenb; /* opt_len and static_len in bytes */
|
---|
1593 | int max_blindex; /* index of last bit length code of non zero freq */
|
---|
1594 |
|
---|
1595 | G2.flag_buf[G2.last_flags] = G2.flags; /* Save the flags for the last 8 items */
|
---|
1596 |
|
---|
1597 | /* Construct the literal and distance trees */
|
---|
1598 | build_tree(&G2.l_desc);
|
---|
1599 | Tracev((stderr, "\nlit data: dyn %ld, stat %ld", G2.opt_len, G2.static_len));
|
---|
1600 |
|
---|
1601 | build_tree(&G2.d_desc);
|
---|
1602 | Tracev((stderr, "\ndist data: dyn %ld, stat %ld", G2.opt_len, G2.static_len));
|
---|
1603 | /* At this point, opt_len and static_len are the total bit lengths of
|
---|
1604 | * the compressed block data, excluding the tree representations.
|
---|
1605 | */
|
---|
1606 |
|
---|
1607 | /* Build the bit length tree for the above two trees, and get the index
|
---|
1608 | * in bl_order of the last bit length code to send.
|
---|
1609 | */
|
---|
1610 | max_blindex = build_bl_tree();
|
---|
1611 |
|
---|
1612 | /* Determine the best encoding. Compute first the block length in bytes */
|
---|
1613 | opt_lenb = (G2.opt_len + 3 + 7) >> 3;
|
---|
1614 | static_lenb = (G2.static_len + 3 + 7) >> 3;
|
---|
1615 |
|
---|
1616 | Trace((stderr,
|
---|
1617 | "\nopt %lu(%lu) stat %lu(%lu) stored %lu lit %u dist %u ",
|
---|
1618 | opt_lenb, G2.opt_len, static_lenb, G2.static_len, stored_len,
|
---|
1619 | G2.last_lit, G2.last_dist));
|
---|
1620 |
|
---|
1621 | if (static_lenb <= opt_lenb)
|
---|
1622 | opt_lenb = static_lenb;
|
---|
1623 |
|
---|
1624 | /* If compression failed and this is the first and last block,
|
---|
1625 | * and if the zip file can be seeked (to rewrite the local header),
|
---|
1626 | * the whole file is transformed into a stored file:
|
---|
1627 | */
|
---|
1628 | if (stored_len <= opt_lenb && eof && G2.compressed_len == 0L && seekable()) {
|
---|
1629 | /* Since LIT_BUFSIZE <= 2*WSIZE, the input data must be there: */
|
---|
1630 | if (buf == NULL)
|
---|
1631 | bb_error_msg("block vanished");
|
---|
1632 |
|
---|
1633 | copy_block(buf, (unsigned) stored_len, 0); /* without header */
|
---|
1634 | G2.compressed_len = stored_len << 3;
|
---|
1635 |
|
---|
1636 | } else if (stored_len + 4 <= opt_lenb && buf != NULL) {
|
---|
1637 | /* 4: two words for the lengths */
|
---|
1638 | /* The test buf != NULL is only necessary if LIT_BUFSIZE > WSIZE.
|
---|
1639 | * Otherwise we can't have processed more than WSIZE input bytes since
|
---|
1640 | * the last block flush, because compression would have been
|
---|
1641 | * successful. If LIT_BUFSIZE <= WSIZE, it is never too late to
|
---|
1642 | * transform a block into a stored block.
|
---|
1643 | */
|
---|
1644 | send_bits((STORED_BLOCK << 1) + eof, 3); /* send block type */
|
---|
1645 | G2.compressed_len = (G2.compressed_len + 3 + 7) & ~7L;
|
---|
1646 | G2.compressed_len += (stored_len + 4) << 3;
|
---|
1647 |
|
---|
1648 | copy_block(buf, (unsigned) stored_len, 1); /* with header */
|
---|
1649 |
|
---|
1650 | } else if (static_lenb == opt_lenb) {
|
---|
1651 | send_bits((STATIC_TREES << 1) + eof, 3);
|
---|
1652 | compress_block((ct_data *) G2.static_ltree, (ct_data *) G2.static_dtree);
|
---|
1653 | G2.compressed_len += 3 + G2.static_len;
|
---|
1654 | } else {
|
---|
1655 | send_bits((DYN_TREES << 1) + eof, 3);
|
---|
1656 | send_all_trees(G2.l_desc.max_code + 1, G2.d_desc.max_code + 1,
|
---|
1657 | max_blindex + 1);
|
---|
1658 | compress_block((ct_data *) G2.dyn_ltree, (ct_data *) G2.dyn_dtree);
|
---|
1659 | G2.compressed_len += 3 + G2.opt_len;
|
---|
1660 | }
|
---|
1661 | Assert(G2.compressed_len == G1.bits_sent, "bad compressed size");
|
---|
1662 | init_block();
|
---|
1663 |
|
---|
1664 | if (eof) {
|
---|
1665 | bi_windup();
|
---|
1666 | G2.compressed_len += 7; /* align on byte boundary */
|
---|
1667 | }
|
---|
1668 | Tracev((stderr, "\ncomprlen %lu(%lu) ", G2.compressed_len >> 3,
|
---|
1669 | G2.compressed_len - 7 * eof));
|
---|
1670 |
|
---|
1671 | return G2.compressed_len >> 3;
|
---|
1672 | }
|
---|
1673 |
|
---|
1674 |
|
---|
1675 | /* ===========================================================================
|
---|
1676 | * Update a hash value with the given input byte
|
---|
1677 | * IN assertion: all calls to to UPDATE_HASH are made with consecutive
|
---|
1678 | * input characters, so that a running hash key can be computed from the
|
---|
1679 | * previous key instead of complete recalculation each time.
|
---|
1680 | */
|
---|
1681 | #define UPDATE_HASH(h, c) (h = (((h)<<H_SHIFT) ^ (c)) & HASH_MASK)
|
---|
1682 |
|
---|
1683 |
|
---|
1684 | /* ===========================================================================
|
---|
1685 | * Same as above, but achieves better compression. We use a lazy
|
---|
1686 | * evaluation for matches: a match is finally adopted only if there is
|
---|
1687 | * no better match at the next window position.
|
---|
1688 | *
|
---|
1689 | * Processes a new input file and return its compressed length. Sets
|
---|
1690 | * the compressed length, crc, deflate flags and internal file
|
---|
1691 | * attributes.
|
---|
1692 | */
|
---|
1693 |
|
---|
1694 | /* Flush the current block, with given end-of-file flag.
|
---|
1695 | * IN assertion: strstart is set to the end of the current match. */
|
---|
1696 | #define FLUSH_BLOCK(eof) \
|
---|
1697 | flush_block( \
|
---|
1698 | G1.block_start >= 0L \
|
---|
1699 | ? (char*)&G1.window[(unsigned)G1.block_start] \
|
---|
1700 | : (char*)NULL, \
|
---|
1701 | (ulg)G1.strstart - G1.block_start, \
|
---|
1702 | (eof) \
|
---|
1703 | )
|
---|
1704 |
|
---|
1705 | /* Insert string s in the dictionary and set match_head to the previous head
|
---|
1706 | * of the hash chain (the most recent string with same hash key). Return
|
---|
1707 | * the previous length of the hash chain.
|
---|
1708 | * IN assertion: all calls to to INSERT_STRING are made with consecutive
|
---|
1709 | * input characters and the first MIN_MATCH bytes of s are valid
|
---|
1710 | * (except for the last MIN_MATCH-1 bytes of the input file). */
|
---|
1711 | #define INSERT_STRING(s, match_head) \
|
---|
1712 | do { \
|
---|
1713 | UPDATE_HASH(G1.ins_h, G1.window[(s) + MIN_MATCH-1]); \
|
---|
1714 | G1.prev[(s) & WMASK] = match_head = head[G1.ins_h]; \
|
---|
1715 | head[G1.ins_h] = (s); \
|
---|
1716 | } while (0)
|
---|
1717 |
|
---|
1718 | static ulg deflate(void)
|
---|
1719 | {
|
---|
1720 | IPos hash_head; /* head of hash chain */
|
---|
1721 | IPos prev_match; /* previous match */
|
---|
1722 | int flush; /* set if current block must be flushed */
|
---|
1723 | int match_available = 0; /* set if previous match exists */
|
---|
1724 | unsigned match_length = MIN_MATCH - 1; /* length of best match */
|
---|
1725 |
|
---|
1726 | /* Process the input block. */
|
---|
1727 | while (G1.lookahead != 0) {
|
---|
1728 | /* Insert the string window[strstart .. strstart+2] in the
|
---|
1729 | * dictionary, and set hash_head to the head of the hash chain:
|
---|
1730 | */
|
---|
1731 | INSERT_STRING(G1.strstart, hash_head);
|
---|
1732 |
|
---|
1733 | /* Find the longest match, discarding those <= prev_length.
|
---|
1734 | */
|
---|
1735 | G1.prev_length = match_length;
|
---|
1736 | prev_match = G1.match_start;
|
---|
1737 | match_length = MIN_MATCH - 1;
|
---|
1738 |
|
---|
1739 | if (hash_head != 0 && G1.prev_length < max_lazy_match
|
---|
1740 | && G1.strstart - hash_head <= MAX_DIST
|
---|
1741 | ) {
|
---|
1742 | /* To simplify the code, we prevent matches with the string
|
---|
1743 | * of window index 0 (in particular we have to avoid a match
|
---|
1744 | * of the string with itself at the start of the input file).
|
---|
1745 | */
|
---|
1746 | match_length = longest_match(hash_head);
|
---|
1747 | /* longest_match() sets match_start */
|
---|
1748 | if (match_length > G1.lookahead)
|
---|
1749 | match_length = G1.lookahead;
|
---|
1750 |
|
---|
1751 | /* Ignore a length 3 match if it is too distant: */
|
---|
1752 | if (match_length == MIN_MATCH && G1.strstart - G1.match_start > TOO_FAR) {
|
---|
1753 | /* If prev_match is also MIN_MATCH, G1.match_start is garbage
|
---|
1754 | * but we will ignore the current match anyway.
|
---|
1755 | */
|
---|
1756 | match_length--;
|
---|
1757 | }
|
---|
1758 | }
|
---|
1759 | /* If there was a match at the previous step and the current
|
---|
1760 | * match is not better, output the previous match:
|
---|
1761 | */
|
---|
1762 | if (G1.prev_length >= MIN_MATCH && match_length <= G1.prev_length) {
|
---|
1763 | check_match(G1.strstart - 1, prev_match, G1.prev_length);
|
---|
1764 | flush = ct_tally(G1.strstart - 1 - prev_match, G1.prev_length - MIN_MATCH);
|
---|
1765 |
|
---|
1766 | /* Insert in hash table all strings up to the end of the match.
|
---|
1767 | * strstart-1 and strstart are already inserted.
|
---|
1768 | */
|
---|
1769 | G1.lookahead -= G1.prev_length - 1;
|
---|
1770 | G1.prev_length -= 2;
|
---|
1771 | do {
|
---|
1772 | G1.strstart++;
|
---|
1773 | INSERT_STRING(G1.strstart, hash_head);
|
---|
1774 | /* strstart never exceeds WSIZE-MAX_MATCH, so there are
|
---|
1775 | * always MIN_MATCH bytes ahead. If lookahead < MIN_MATCH
|
---|
1776 | * these bytes are garbage, but it does not matter since the
|
---|
1777 | * next lookahead bytes will always be emitted as literals.
|
---|
1778 | */
|
---|
1779 | } while (--G1.prev_length != 0);
|
---|
1780 | match_available = 0;
|
---|
1781 | match_length = MIN_MATCH - 1;
|
---|
1782 | G1.strstart++;
|
---|
1783 | if (flush) {
|
---|
1784 | FLUSH_BLOCK(0);
|
---|
1785 | G1.block_start = G1.strstart;
|
---|
1786 | }
|
---|
1787 | } else if (match_available) {
|
---|
1788 | /* If there was no match at the previous position, output a
|
---|
1789 | * single literal. If there was a match but the current match
|
---|
1790 | * is longer, truncate the previous match to a single literal.
|
---|
1791 | */
|
---|
1792 | Tracevv((stderr, "%c", G1.window[G1.strstart - 1]));
|
---|
1793 | if (ct_tally(0, G1.window[G1.strstart - 1])) {
|
---|
1794 | FLUSH_BLOCK(0);
|
---|
1795 | G1.block_start = G1.strstart;
|
---|
1796 | }
|
---|
1797 | G1.strstart++;
|
---|
1798 | G1.lookahead--;
|
---|
1799 | } else {
|
---|
1800 | /* There is no previous match to compare with, wait for
|
---|
1801 | * the next step to decide.
|
---|
1802 | */
|
---|
1803 | match_available = 1;
|
---|
1804 | G1.strstart++;
|
---|
1805 | G1.lookahead--;
|
---|
1806 | }
|
---|
1807 | Assert(G1.strstart <= G1.isize && lookahead <= G1.isize, "a bit too far");
|
---|
1808 |
|
---|
1809 | /* Make sure that we always have enough lookahead, except
|
---|
1810 | * at the end of the input file. We need MAX_MATCH bytes
|
---|
1811 | * for the next match, plus MIN_MATCH bytes to insert the
|
---|
1812 | * string following the next match.
|
---|
1813 | */
|
---|
1814 | while (G1.lookahead < MIN_LOOKAHEAD && !G1.eofile)
|
---|
1815 | fill_window();
|
---|
1816 | }
|
---|
1817 | if (match_available)
|
---|
1818 | ct_tally(0, G1.window[G1.strstart - 1]);
|
---|
1819 |
|
---|
1820 | return FLUSH_BLOCK(1); /* eof */
|
---|
1821 | }
|
---|
1822 |
|
---|
1823 |
|
---|
1824 | /* ===========================================================================
|
---|
1825 | * Initialize the bit string routines.
|
---|
1826 | */
|
---|
1827 | static void bi_init(void)
|
---|
1828 | {
|
---|
1829 | G1.bi_buf = 0;
|
---|
1830 | G1.bi_valid = 0;
|
---|
1831 | #ifdef DEBUG
|
---|
1832 | G1.bits_sent = 0L;
|
---|
1833 | #endif
|
---|
1834 | }
|
---|
1835 |
|
---|
1836 |
|
---|
1837 | /* ===========================================================================
|
---|
1838 | * Initialize the "longest match" routines for a new file
|
---|
1839 | */
|
---|
1840 | static void lm_init(ush * flagsp)
|
---|
1841 | {
|
---|
1842 | unsigned j;
|
---|
1843 |
|
---|
1844 | /* Initialize the hash table. */
|
---|
1845 | memset(head, 0, HASH_SIZE * sizeof(*head));
|
---|
1846 | /* prev will be initialized on the fly */
|
---|
1847 |
|
---|
1848 | /* speed options for the general purpose bit flag */
|
---|
1849 | *flagsp |= 2; /* FAST 4, SLOW 2 */
|
---|
1850 | /* ??? reduce max_chain_length for binary files */
|
---|
1851 |
|
---|
1852 | G1.strstart = 0;
|
---|
1853 | G1.block_start = 0L;
|
---|
1854 |
|
---|
1855 | G1.lookahead = file_read(G1.window,
|
---|
1856 | sizeof(int) <= 2 ? (unsigned) WSIZE : 2 * WSIZE);
|
---|
1857 |
|
---|
1858 | if (G1.lookahead == 0 || G1.lookahead == (unsigned) -1) {
|
---|
1859 | G1.eofile = 1;
|
---|
1860 | G1.lookahead = 0;
|
---|
1861 | return;
|
---|
1862 | }
|
---|
1863 | G1.eofile = 0;
|
---|
1864 | /* Make sure that we always have enough lookahead. This is important
|
---|
1865 | * if input comes from a device such as a tty.
|
---|
1866 | */
|
---|
1867 | while (G1.lookahead < MIN_LOOKAHEAD && !G1.eofile)
|
---|
1868 | fill_window();
|
---|
1869 |
|
---|
1870 | G1.ins_h = 0;
|
---|
1871 | for (j = 0; j < MIN_MATCH - 1; j++)
|
---|
1872 | UPDATE_HASH(G1.ins_h, G1.window[j]);
|
---|
1873 | /* If lookahead < MIN_MATCH, ins_h is garbage, but this is
|
---|
1874 | * not important since only literal bytes will be emitted.
|
---|
1875 | */
|
---|
1876 | }
|
---|
1877 |
|
---|
1878 |
|
---|
1879 | /* ===========================================================================
|
---|
1880 | * Allocate the match buffer, initialize the various tables and save the
|
---|
1881 | * location of the internal file attribute (ascii/binary) and method
|
---|
1882 | * (DEFLATE/STORE).
|
---|
1883 | * One callsite in zip()
|
---|
1884 | */
|
---|
1885 | static void ct_init(void)
|
---|
1886 | {
|
---|
1887 | int n; /* iterates over tree elements */
|
---|
1888 | int length; /* length value */
|
---|
1889 | int code; /* code value */
|
---|
1890 | int dist; /* distance index */
|
---|
1891 |
|
---|
1892 | G2.compressed_len = 0L;
|
---|
1893 |
|
---|
1894 | #ifdef NOT_NEEDED
|
---|
1895 | if (G2.static_dtree[0].Len != 0)
|
---|
1896 | return; /* ct_init already called */
|
---|
1897 | #endif
|
---|
1898 |
|
---|
1899 | /* Initialize the mapping length (0..255) -> length code (0..28) */
|
---|
1900 | length = 0;
|
---|
1901 | for (code = 0; code < LENGTH_CODES - 1; code++) {
|
---|
1902 | G2.base_length[code] = length;
|
---|
1903 | for (n = 0; n < (1 << extra_lbits[code]); n++) {
|
---|
1904 | G2.length_code[length++] = code;
|
---|
1905 | }
|
---|
1906 | }
|
---|
1907 | Assert(length == 256, "ct_init: length != 256");
|
---|
1908 | /* Note that the length 255 (match length 258) can be represented
|
---|
1909 | * in two different ways: code 284 + 5 bits or code 285, so we
|
---|
1910 | * overwrite length_code[255] to use the best encoding:
|
---|
1911 | */
|
---|
1912 | G2.length_code[length - 1] = code;
|
---|
1913 |
|
---|
1914 | /* Initialize the mapping dist (0..32K) -> dist code (0..29) */
|
---|
1915 | dist = 0;
|
---|
1916 | for (code = 0; code < 16; code++) {
|
---|
1917 | G2.base_dist[code] = dist;
|
---|
1918 | for (n = 0; n < (1 << extra_dbits[code]); n++) {
|
---|
1919 | G2.dist_code[dist++] = code;
|
---|
1920 | }
|
---|
1921 | }
|
---|
1922 | Assert(dist == 256, "ct_init: dist != 256");
|
---|
1923 | dist >>= 7; /* from now on, all distances are divided by 128 */
|
---|
1924 | for (; code < D_CODES; code++) {
|
---|
1925 | G2.base_dist[code] = dist << 7;
|
---|
1926 | for (n = 0; n < (1 << (extra_dbits[code] - 7)); n++) {
|
---|
1927 | G2.dist_code[256 + dist++] = code;
|
---|
1928 | }
|
---|
1929 | }
|
---|
1930 | Assert(dist == 256, "ct_init: 256+dist != 512");
|
---|
1931 |
|
---|
1932 | /* Construct the codes of the static literal tree */
|
---|
1933 | /* already zeroed - it's in bss
|
---|
1934 | for (n = 0; n <= MAX_BITS; n++)
|
---|
1935 | G2.bl_count[n] = 0; */
|
---|
1936 |
|
---|
1937 | n = 0;
|
---|
1938 | while (n <= 143) {
|
---|
1939 | G2.static_ltree[n++].Len = 8;
|
---|
1940 | G2.bl_count[8]++;
|
---|
1941 | }
|
---|
1942 | while (n <= 255) {
|
---|
1943 | G2.static_ltree[n++].Len = 9;
|
---|
1944 | G2.bl_count[9]++;
|
---|
1945 | }
|
---|
1946 | while (n <= 279) {
|
---|
1947 | G2.static_ltree[n++].Len = 7;
|
---|
1948 | G2.bl_count[7]++;
|
---|
1949 | }
|
---|
1950 | while (n <= 287) {
|
---|
1951 | G2.static_ltree[n++].Len = 8;
|
---|
1952 | G2.bl_count[8]++;
|
---|
1953 | }
|
---|
1954 | /* Codes 286 and 287 do not exist, but we must include them in the
|
---|
1955 | * tree construction to get a canonical Huffman tree (longest code
|
---|
1956 | * all ones)
|
---|
1957 | */
|
---|
1958 | gen_codes((ct_data *) G2.static_ltree, L_CODES + 1);
|
---|
1959 |
|
---|
1960 | /* The static distance tree is trivial: */
|
---|
1961 | for (n = 0; n < D_CODES; n++) {
|
---|
1962 | G2.static_dtree[n].Len = 5;
|
---|
1963 | G2.static_dtree[n].Code = bi_reverse(n, 5);
|
---|
1964 | }
|
---|
1965 |
|
---|
1966 | /* Initialize the first block of the first file: */
|
---|
1967 | init_block();
|
---|
1968 | }
|
---|
1969 |
|
---|
1970 |
|
---|
1971 | /* ===========================================================================
|
---|
1972 | * Deflate in to out.
|
---|
1973 | * IN assertions: the input and output buffers are cleared.
|
---|
1974 | */
|
---|
1975 |
|
---|
1976 | static void zip(ulg time_stamp)
|
---|
1977 | {
|
---|
1978 | ush deflate_flags = 0; /* pkzip -es, -en or -ex equivalent */
|
---|
1979 |
|
---|
1980 | G1.outcnt = 0;
|
---|
1981 |
|
---|
1982 | /* Write the header to the gzip file. See algorithm.doc for the format */
|
---|
1983 | /* magic header for gzip files: 1F 8B */
|
---|
1984 | /* compression method: 8 (DEFLATED) */
|
---|
1985 | /* general flags: 0 */
|
---|
1986 | put_32bit(0x00088b1f);
|
---|
1987 | put_32bit(time_stamp);
|
---|
1988 |
|
---|
1989 | /* Write deflated file to zip file */
|
---|
1990 | G1.crc = ~0;
|
---|
1991 |
|
---|
1992 | bi_init();
|
---|
1993 | ct_init();
|
---|
1994 | lm_init(&deflate_flags);
|
---|
1995 |
|
---|
1996 | put_8bit(deflate_flags); /* extra flags */
|
---|
1997 | put_8bit(3); /* OS identifier = 3 (Unix) */
|
---|
1998 |
|
---|
1999 | deflate();
|
---|
2000 |
|
---|
2001 | /* Write the crc and uncompressed size */
|
---|
2002 | put_32bit(~G1.crc);
|
---|
2003 | put_32bit(G1.isize);
|
---|
2004 |
|
---|
2005 | flush_outbuf();
|
---|
2006 | }
|
---|
2007 |
|
---|
2008 |
|
---|
2009 | /* ======================================================================== */
|
---|
2010 | static
|
---|
2011 | char* make_new_name_gzip(char *filename)
|
---|
2012 | {
|
---|
2013 | return xasprintf("%s.gz", filename);
|
---|
2014 | }
|
---|
2015 |
|
---|
2016 | static
|
---|
2017 | USE_DESKTOP(long long) int pack_gzip(void)
|
---|
2018 | {
|
---|
2019 | struct stat s;
|
---|
2020 |
|
---|
2021 | clear_bufs();
|
---|
2022 | s.st_ctime = 0;
|
---|
2023 | fstat(STDIN_FILENO, &s);
|
---|
2024 | zip(s.st_ctime);
|
---|
2025 | return 0;
|
---|
2026 | }
|
---|
2027 |
|
---|
2028 | int gzip_main(int argc, char **argv);
|
---|
2029 | int gzip_main(int argc, char **argv)
|
---|
2030 | {
|
---|
2031 | unsigned opt;
|
---|
2032 |
|
---|
2033 | /* Must match bbunzip's constants OPT_STDOUT, OPT_FORCE! */
|
---|
2034 | opt = getopt32(argv, "cfv" USE_GUNZIP("d") "q123456789" );
|
---|
2035 | option_mask32 &= 0x7; /* Clear -d, ignore -q, -0..9 */
|
---|
2036 | //if (opt & 0x1) // -c
|
---|
2037 | //if (opt & 0x2) // -f
|
---|
2038 | //if (opt & 0x4) // -v
|
---|
2039 | #if ENABLE_GUNZIP /* gunzip_main may not be visible... */
|
---|
2040 | if (opt & 0x8) { // -d
|
---|
2041 | return gunzip_main(argc, argv);
|
---|
2042 | }
|
---|
2043 | #endif
|
---|
2044 | argv += optind;
|
---|
2045 |
|
---|
2046 | PTR_TO_GLOBALS = xzalloc(sizeof(struct globals) + sizeof(struct globals2))
|
---|
2047 | + sizeof(struct globals);
|
---|
2048 | G2.l_desc.dyn_tree = G2.dyn_ltree;
|
---|
2049 | G2.l_desc.static_tree = G2.static_ltree;
|
---|
2050 | G2.l_desc.extra_bits = extra_lbits;
|
---|
2051 | G2.l_desc.extra_base = LITERALS + 1;
|
---|
2052 | G2.l_desc.elems = L_CODES;
|
---|
2053 | G2.l_desc.max_length = MAX_BITS;
|
---|
2054 | //G2.l_desc.max_code = 0;
|
---|
2055 |
|
---|
2056 | G2.d_desc.dyn_tree = G2.dyn_dtree;
|
---|
2057 | G2.d_desc.static_tree = G2.static_dtree;
|
---|
2058 | G2.d_desc.extra_bits = extra_dbits;
|
---|
2059 | //G2.d_desc.extra_base = 0;
|
---|
2060 | G2.d_desc.elems = D_CODES;
|
---|
2061 | G2.d_desc.max_length = MAX_BITS;
|
---|
2062 | //G2.d_desc.max_code = 0;
|
---|
2063 |
|
---|
2064 | G2.bl_desc.dyn_tree = G2.bl_tree;
|
---|
2065 | //G2.bl_desc.static_tree = NULL;
|
---|
2066 | G2.bl_desc.extra_bits = extra_blbits,
|
---|
2067 | //G2.bl_desc.extra_base = 0;
|
---|
2068 | G2.bl_desc.elems = BL_CODES;
|
---|
2069 | G2.bl_desc.max_length = MAX_BL_BITS;
|
---|
2070 | //G2.bl_desc.max_code = 0;
|
---|
2071 |
|
---|
2072 | /* Allocate all global buffers (for DYN_ALLOC option) */
|
---|
2073 | ALLOC(uch, G1.l_buf, INBUFSIZ);
|
---|
2074 | ALLOC(uch, G1.outbuf, OUTBUFSIZ);
|
---|
2075 | ALLOC(ush, G1.d_buf, DIST_BUFSIZE);
|
---|
2076 | ALLOC(uch, G1.window, 2L * WSIZE);
|
---|
2077 | ALLOC(ush, G1.prev, 1L << BITS);
|
---|
2078 |
|
---|
2079 | /* Initialise the CRC32 table */
|
---|
2080 | G1.crc_32_tab = crc32_filltable(NULL, 0);
|
---|
2081 |
|
---|
2082 | return bbunpack(argv, make_new_name_gzip, pack_gzip);
|
---|
2083 | }
|
---|