summaryrefslogtreecommitdiff
path: root/archival/gzip.c (plain)
blob: 1f0b70fbf2917df502f8ea9e4a58116bbcb78bda
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 source tree.
17 */
18/* big objects in bss:
19 * 00000020 b bl_count
20 * 00000074 b base_length
21 * 00000078 b base_dist
22 * 00000078 b static_dtree
23 * 0000009c b bl_tree
24 * 000000f4 b dyn_dtree
25 * 00000100 b length_code
26 * 00000200 b dist_code
27 * 0000023d b depth
28 * 00000400 b flag_buf
29 * 0000047a b heap
30 * 00000480 b static_ltree
31 * 000008f4 b dyn_ltree
32 */
33/* TODO: full support for -v for DESKTOP
34 * "/usr/bin/gzip -v a bogus aa" should say:
35a: 85.1% -- replaced with a.gz
36gzip: bogus: No such file or directory
37aa: 85.1% -- replaced with aa.gz
38*/
39
40//config:config GZIP
41//config: bool "gzip"
42//config: default y
43//config: help
44//config: gzip is used to compress files.
45//config: It's probably the most widely used UNIX compression program.
46//config:
47//config:config FEATURE_GZIP_LONG_OPTIONS
48//config: bool "Enable long options"
49//config: default y
50//config: depends on GZIP && LONG_OPTS
51//config: help
52//config: Enable use of long options, increases size by about 106 Bytes
53//config:
54//config:config GZIP_FAST
55//config: int "Trade memory for gzip speed (0:small,slow - 2:fast,big)"
56//config: default 0
57//config: range 0 2
58//config: depends on GZIP
59//config: help
60//config: Enable big memory options for gzip.
61//config: 0: small buffers, small hash-tables
62//config: 1: larger buffers, larger hash-tables
63//config: 2: larger buffers, largest hash-tables
64//config: Larger models may give slightly better compression
65//config:
66//config:config FEATURE_GZIP_LEVELS
67//config: bool "Enable compression levels"
68//config: default n
69//config: depends on GZIP
70//config: help
71//config: Enable support for compression levels 4-9. The default level
72//config: is 6. If levels 1-3 are specified, 4 is used.
73//config: If this option is not selected, -N options are ignored and -9
74//config: is used.
75
76//applet:IF_GZIP(APPLET(gzip, BB_DIR_BIN, BB_SUID_DROP))
77//kbuild:lib-$(CONFIG_GZIP) += gzip.o
78
79//usage:#define gzip_trivial_usage
80//usage: "[-cf" IF_GUNZIP("d") IF_FEATURE_GZIP_LEVELS("123456789") "] [FILE]..."
81//usage:#define gzip_full_usage "\n\n"
82//usage: "Compress FILEs (or stdin)\n"
83//usage: IF_FEATURE_GZIP_LEVELS(
84//usage: "\n -1..9 Compression level"
85//usage: )
86//usage: IF_GUNZIP(
87//usage: "\n -d Decompress"
88//usage: )
89//usage: "\n -c Write to stdout"
90//usage: "\n -f Force"
91//usage:
92//usage:#define gzip_example_usage
93//usage: "$ ls -la /tmp/busybox*\n"
94//usage: "-rw-rw-r-- 1 andersen andersen 1761280 Apr 14 17:47 /tmp/busybox.tar\n"
95//usage: "$ gzip /tmp/busybox.tar\n"
96//usage: "$ ls -la /tmp/busybox*\n"
97//usage: "-rw-rw-r-- 1 andersen andersen 554058 Apr 14 17:49 /tmp/busybox.tar.gz\n"
98
99#include "libbb.h"
100#include "bb_archive.h"
101#include <strings.h>
102
103
104/* ===========================================================================
105 */
106//#define DEBUG 1
107/* Diagnostic functions */
108#ifdef DEBUG
109# define Assert(cond,msg) { if (!(cond)) bb_error_msg(msg); }
110# define Trace(x) fprintf x
111# define Tracev(x) {if (verbose) fprintf x; }
112# define Tracevv(x) {if (verbose > 1) fprintf x; }
113# define Tracec(c,x) {if (verbose && (c)) fprintf x; }
114# define Tracecv(c,x) {if (verbose > 1 && (c)) fprintf x; }
115#else
116# define Assert(cond,msg)
117# define Trace(x)
118# define Tracev(x)
119# define Tracevv(x)
120# define Tracec(c,x)
121# define Tracecv(c,x)
122#endif
123
124
125/* ===========================================================================
126 */
127#if CONFIG_GZIP_FAST == 0
128# define SMALL_MEM
129#elif CONFIG_GZIP_FAST == 1
130# define MEDIUM_MEM
131#elif CONFIG_GZIP_FAST == 2
132# define BIG_MEM
133#else
134# error "Invalid CONFIG_GZIP_FAST value"
135#endif
136
137#ifndef INBUFSIZ
138# ifdef SMALL_MEM
139# define INBUFSIZ 0x2000 /* input buffer size */
140# else
141# define INBUFSIZ 0x8000 /* input buffer size */
142# endif
143#endif
144
145#ifndef OUTBUFSIZ
146# ifdef SMALL_MEM
147# define OUTBUFSIZ 8192 /* output buffer size */
148# else
149# define OUTBUFSIZ 16384 /* output buffer size */
150# endif
151#endif
152
153#ifndef DIST_BUFSIZE
154# ifdef SMALL_MEM
155# define DIST_BUFSIZE 0x2000 /* buffer for distances, see trees.c */
156# else
157# define DIST_BUFSIZE 0x8000 /* buffer for distances, see trees.c */
158# endif
159#endif
160
161/* gzip flag byte */
162#define ASCII_FLAG 0x01 /* bit 0 set: file probably ascii text */
163#define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */
164#define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
165#define ORIG_NAME 0x08 /* bit 3 set: original file name present */
166#define COMMENT 0x10 /* bit 4 set: file comment present */
167#define RESERVED 0xC0 /* bit 6,7: reserved */
168
169/* internal file attribute */
170#define UNKNOWN 0xffff
171#define BINARY 0
172#define ASCII 1
173
174#ifndef WSIZE
175# define WSIZE 0x8000 /* window size--must be a power of two, and */
176#endif /* at least 32K for zip's deflate method */
177
178#define MIN_MATCH 3
179#define MAX_MATCH 258
180/* The minimum and maximum match lengths */
181
182#define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
183/* Minimum amount of lookahead, except at the end of the input file.
184 * See deflate.c for comments about the MIN_MATCH+1.
185 */
186
187#define MAX_DIST (WSIZE-MIN_LOOKAHEAD)
188/* In order to simplify the code, particularly on 16 bit machines, match
189 * distances are limited to MAX_DIST instead of WSIZE.
190 */
191
192#ifndef MAX_PATH_LEN
193# define MAX_PATH_LEN 1024 /* max pathname length */
194#endif
195
196#define seekable() 0 /* force sequential output */
197#define translate_eol 0 /* no option -a yet */
198
199#ifndef BITS
200# define BITS 16
201#endif
202#define INIT_BITS 9 /* Initial number of bits per code */
203
204#define BIT_MASK 0x1f /* Mask for 'number of compression bits' */
205/* Mask 0x20 is reserved to mean a fourth header byte, and 0x40 is free.
206 * It's a pity that old uncompress does not check bit 0x20. That makes
207 * extension of the format actually undesirable because old compress
208 * would just crash on the new format instead of giving a meaningful
209 * error message. It does check the number of bits, but it's more
210 * helpful to say "unsupported format, get a new version" than
211 * "can only handle 16 bits".
212 */
213
214#ifdef MAX_EXT_CHARS
215# define MAX_SUFFIX MAX_EXT_CHARS
216#else
217# define MAX_SUFFIX 30
218#endif
219
220
221/* ===========================================================================
222 * Compile with MEDIUM_MEM to reduce the memory requirements or
223 * with SMALL_MEM to use as little memory as possible. Use BIG_MEM if the
224 * entire input file can be held in memory (not possible on 16 bit systems).
225 * Warning: defining these symbols affects HASH_BITS (see below) and thus
226 * affects the compression ratio. The compressed output
227 * is still correct, and might even be smaller in some cases.
228 */
229
230#ifdef SMALL_MEM
231# define HASH_BITS 13 /* Number of bits used to hash strings */
232#endif
233#ifdef MEDIUM_MEM
234# define HASH_BITS 14
235#endif
236#ifndef HASH_BITS
237# define HASH_BITS 15
238 /* For portability to 16 bit machines, do not use values above 15. */
239#endif
240
241#define HASH_SIZE (unsigned)(1<<HASH_BITS)
242#define HASH_MASK (HASH_SIZE-1)
243#define WMASK (WSIZE-1)
244/* HASH_SIZE and WSIZE must be powers of two */
245#ifndef TOO_FAR
246# define TOO_FAR 4096
247#endif
248/* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
249
250
251/* ===========================================================================
252 * These types are not really 'char', 'short' and 'long'
253 */
254typedef uint8_t uch;
255typedef uint16_t ush;
256typedef uint32_t ulg;
257typedef int32_t lng;
258
259typedef ush Pos;
260typedef unsigned IPos;
261/* A Pos is an index in the character window. We use short instead of int to
262 * save space in the various tables. IPos is used only for parameter passing.
263 */
264
265enum {
266 WINDOW_SIZE = 2 * WSIZE,
267/* window size, 2*WSIZE except for MMAP or BIG_MEM, where it is the
268 * input file length plus MIN_LOOKAHEAD.
269 */
270
271#ifndef ENABLE_FEATURE_GZIP_LEVELS
272
273 max_chain_length = 4096,
274/* To speed up deflation, hash chains are never searched beyond this length.
275 * A higher limit improves compression ratio but degrades the speed.
276 */
277
278 max_lazy_match = 258,
279/* Attempt to find a better match only when the current match is strictly
280 * smaller than this value. This mechanism is used only for compression
281 * levels >= 4.
282 */
283
284 max_insert_length = max_lazy_match,
285/* Insert new strings in the hash table only if the match length
286 * is not greater than this length. This saves time but degrades compression.
287 * max_insert_length is used only for compression levels <= 3.
288 */
289
290 good_match = 32,
291/* Use a faster search when the previous match is longer than this */
292
293/* Values for max_lazy_match, good_match and max_chain_length, depending on
294 * the desired pack level (0..9). The values given below have been tuned to
295 * exclude worst case performance for pathological files. Better values may be
296 * found for specific files.
297 */
298
299 nice_match = 258, /* Stop searching when current match exceeds this */
300/* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
301 * For deflate_fast() (levels <= 3) good is ignored and lazy has a different
302 * meaning.
303 */
304#endif /* ENABLE_FEATURE_GZIP_LEVELS */
305};
306
307
308struct globals {
309
310#ifdef ENABLE_FEATURE_GZIP_LEVELS
311 unsigned max_chain_length;
312 unsigned max_lazy_match;
313 unsigned good_match;
314 unsigned nice_match;
315#define max_chain_length (G1.max_chain_length)
316#define max_lazy_match (G1.max_lazy_match)
317#define good_match (G1.good_match)
318#define nice_match (G1.nice_match)
319#endif
320
321 lng block_start;
322
323/* window position at the beginning of the current output block. Gets
324 * negative when the window is moved backwards.
325 */
326 unsigned ins_h; /* hash index of string to be inserted */
327
328#define H_SHIFT ((HASH_BITS+MIN_MATCH-1) / MIN_MATCH)
329/* Number of bits by which ins_h and del_h must be shifted at each
330 * input step. It must be such that after MIN_MATCH steps, the oldest
331 * byte no longer takes part in the hash key, that is:
332 * H_SHIFT * MIN_MATCH >= HASH_BITS
333 */
334
335 unsigned prev_length;
336
337/* Length of the best match at previous step. Matches not greater than this
338 * are discarded. This is used in the lazy match evaluation.
339 */
340
341 unsigned strstart; /* start of string to insert */
342 unsigned match_start; /* start of matching string */
343 unsigned lookahead; /* number of valid bytes ahead in window */
344
345/* ===========================================================================
346 */
347#define DECLARE(type, array, size) \
348 type * array
349#define ALLOC(type, array, size) \
350 array = xzalloc((size_t)(((size)+1L)/2) * 2*sizeof(type))
351#define FREE(array) \
352 do { free(array); array = NULL; } while (0)
353
354 /* global buffers */
355
356 /* buffer for literals or lengths */
357 /* DECLARE(uch, l_buf, LIT_BUFSIZE); */
358 DECLARE(uch, l_buf, INBUFSIZ);
359
360 DECLARE(ush, d_buf, DIST_BUFSIZE);
361 DECLARE(uch, outbuf, OUTBUFSIZ);
362
363/* Sliding window. Input bytes are read into the second half of the window,
364 * and move to the first half later to keep a dictionary of at least WSIZE
365 * bytes. With this organization, matches are limited to a distance of
366 * WSIZE-MAX_MATCH bytes, but this ensures that IO is always
367 * performed with a length multiple of the block size. Also, it limits
368 * the window size to 64K, which is quite useful on MSDOS.
369 * To do: limit the window size to WSIZE+BSZ if SMALL_MEM (the code would
370 * be less efficient).
371 */
372 DECLARE(uch, window, 2L * WSIZE);
373
374/* Link to older string with same hash index. To limit the size of this
375 * array to 64K, this link is maintained only for the last 32K strings.
376 * An index in this array is thus a window index modulo 32K.
377 */
378 /* DECLARE(Pos, prev, WSIZE); */
379 DECLARE(ush, prev, 1L << BITS);
380
381/* Heads of the hash chains or 0. */
382 /* DECLARE(Pos, head, 1<<HASH_BITS); */
383#define head (G1.prev + WSIZE) /* hash head (see deflate.c) */
384
385/* number of input bytes */
386 ulg isize; /* only 32 bits stored in .gz file */
387
388/* bbox always use stdin/stdout */
389#define ifd STDIN_FILENO /* input file descriptor */
390#define ofd STDOUT_FILENO /* output file descriptor */
391
392#ifdef DEBUG
393 unsigned insize; /* valid bytes in l_buf */
394#endif
395 unsigned outcnt; /* bytes in output buffer */
396
397 smallint eofile; /* flag set at end of input file */
398
399/* ===========================================================================
400 * Local data used by the "bit string" routines.
401 */
402
403 unsigned short bi_buf;
404
405/* Output buffer. bits are inserted starting at the bottom (least significant
406 * bits).
407 */
408
409#undef BUF_SIZE
410#define BUF_SIZE (8 * sizeof(G1.bi_buf))
411/* Number of bits used within bi_buf. (bi_buf might be implemented on
412 * more than 16 bits on some systems.)
413 */
414
415 int bi_valid;
416
417/* Current input function. Set to mem_read for in-memory compression */
418
419#ifdef DEBUG
420 ulg bits_sent; /* bit length of the compressed data */
421#endif
422
423 /*uint32_t *crc_32_tab;*/
424 uint32_t crc; /* shift register contents */
425};
426
427#define G1 (*(ptr_to_globals - 1))
428
429
430/* ===========================================================================
431 * Write the output buffer outbuf[0..outcnt-1] and update bytes_out.
432 * (used for the compressed data only)
433 */
434static void flush_outbuf(void)
435{
436 if (G1.outcnt == 0)
437 return;
438
439 xwrite(ofd, (char *) G1.outbuf, G1.outcnt);
440 G1.outcnt = 0;
441}
442
443
444/* ===========================================================================
445 */
446/* put_8bit is used for the compressed output */
447#define put_8bit(c) \
448do { \
449 G1.outbuf[G1.outcnt++] = (c); \
450 if (G1.outcnt == OUTBUFSIZ) \
451 flush_outbuf(); \
452} while (0)
453
454/* Output a 16 bit value, lsb first */
455static void put_16bit(ush w)
456{
457 /* GCC 4.2.1 won't optimize out redundant loads of G1.outcnt
458 * (probably because of fear of aliasing with G1.outbuf[]
459 * stores), do it explicitly:
460 */
461 unsigned outcnt = G1.outcnt;
462 uch *dst = &G1.outbuf[outcnt];
463
464#if BB_UNALIGNED_MEMACCESS_OK && BB_LITTLE_ENDIAN
465 if (outcnt < OUTBUFSIZ-2) {
466 /* Common case */
467 ush *dst16 = (void*) dst;
468 *dst16 = w; /* unalinged LSB 16-bit store */
469 G1.outcnt = outcnt + 2;
470 return;
471 }
472 *dst = (uch)w;
473 w >>= 8;
474#else
475 *dst = (uch)w;
476 w >>= 8;
477 if (outcnt < OUTBUFSIZ-2) {
478 /* Common case */
479 dst[1] = w;
480 G1.outcnt = outcnt + 2;
481 return;
482 }
483#endif
484
485 /* Slowpath: we will need to do flush_outbuf() */
486 G1.outcnt = ++outcnt;
487 if (outcnt == OUTBUFSIZ)
488 flush_outbuf();
489 put_8bit(w);
490}
491
492static void put_32bit(ulg n)
493{
494 put_16bit(n);
495 put_16bit(n >> 16);
496}
497
498/* ===========================================================================
499 * Run a set of bytes through the crc shift register. If s is a NULL
500 * pointer, then initialize the crc shift register contents instead.
501 * Return the current crc in either case.
502 */
503static void updcrc(uch * s, unsigned n)
504{
505 G1.crc = crc32_block_endian0(G1.crc, s, n, global_crc32_table /*G1.crc_32_tab*/);
506}
507
508
509/* ===========================================================================
510 * Read a new buffer from the current input file, perform end-of-line
511 * translation, and update the crc and input file size.
512 * IN assertion: size >= 2 (for end-of-line translation)
513 */
514static unsigned file_read(void *buf, unsigned size)
515{
516 unsigned len;
517
518 Assert(G1.insize == 0, "l_buf not empty");
519
520 len = safe_read(ifd, buf, size);
521 if (len == (unsigned)(-1) || len == 0)
522 return len;
523
524 updcrc(buf, len);
525 G1.isize += len;
526 return len;
527}
528
529
530/* ===========================================================================
531 * Send a value on a given number of bits.
532 * IN assertion: length <= 16 and value fits in length bits.
533 */
534static void send_bits(int value, int length)
535{
536#ifdef DEBUG
537 Tracev((stderr, " l %2d v %4x ", length, value));
538 Assert(length > 0 && length <= 15, "invalid length");
539 G1.bits_sent += length;
540#endif
541 /* If not enough room in bi_buf, use (valid) bits from bi_buf and
542 * (16 - bi_valid) bits from value, leaving (width - (16-bi_valid))
543 * unused bits in value.
544 */
545 if (G1.bi_valid > (int) BUF_SIZE - length) {
546 G1.bi_buf |= (value << G1.bi_valid);
547 put_16bit(G1.bi_buf);
548 G1.bi_buf = (ush) value >> (BUF_SIZE - G1.bi_valid);
549 G1.bi_valid += length - BUF_SIZE;
550 } else {
551 G1.bi_buf |= value << G1.bi_valid;
552 G1.bi_valid += length;
553 }
554}
555
556
557/* ===========================================================================
558 * Reverse the first len bits of a code, using straightforward code (a faster
559 * method would use a table)
560 * IN assertion: 1 <= len <= 15
561 */
562static unsigned bi_reverse(unsigned code, int len)
563{
564 unsigned res = 0;
565
566 while (1) {
567 res |= code & 1;
568 if (--len <= 0) return res;
569 code >>= 1;
570 res <<= 1;
571 }
572}
573
574
575/* ===========================================================================
576 * Write out any remaining bits in an incomplete byte.
577 */
578static void bi_windup(void)
579{
580 if (G1.bi_valid > 8) {
581 put_16bit(G1.bi_buf);
582 } else if (G1.bi_valid > 0) {
583 put_8bit(G1.bi_buf);
584 }
585 G1.bi_buf = 0;
586 G1.bi_valid = 0;
587#ifdef DEBUG
588 G1.bits_sent = (G1.bits_sent + 7) & ~7;
589#endif
590}
591
592
593/* ===========================================================================
594 * Copy a stored block to the zip file, storing first the length and its
595 * one's complement if requested.
596 */
597static void copy_block(char *buf, unsigned len, int header)
598{
599 bi_windup(); /* align on byte boundary */
600
601 if (header) {
602 put_16bit(len);
603 put_16bit(~len);
604#ifdef DEBUG
605 G1.bits_sent += 2 * 16;
606#endif
607 }
608#ifdef DEBUG
609 G1.bits_sent += (ulg) len << 3;
610#endif
611 while (len--) {
612 put_8bit(*buf++);
613 }
614}
615
616
617/* ===========================================================================
618 * Fill the window when the lookahead becomes insufficient.
619 * Updates strstart and lookahead, and sets eofile if end of input file.
620 * IN assertion: lookahead < MIN_LOOKAHEAD && strstart + lookahead > 0
621 * OUT assertions: at least one byte has been read, or eofile is set;
622 * file reads are performed for at least two bytes (required for the
623 * translate_eol option).
624 */
625static void fill_window(void)
626{
627 unsigned n, m;
628 unsigned more = WINDOW_SIZE - G1.lookahead - G1.strstart;
629 /* Amount of free space at the end of the window. */
630
631 /* If the window is almost full and there is insufficient lookahead,
632 * move the upper half to the lower one to make room in the upper half.
633 */
634 if (more == (unsigned) -1) {
635 /* Very unlikely, but possible on 16 bit machine if strstart == 0
636 * and lookahead == 1 (input done one byte at time)
637 */
638 more--;
639 } else if (G1.strstart >= WSIZE + MAX_DIST) {
640 /* By the IN assertion, the window is not empty so we can't confuse
641 * more == 0 with more == 64K on a 16 bit machine.
642 */
643 Assert(WINDOW_SIZE == 2 * WSIZE, "no sliding with BIG_MEM");
644
645 memcpy(G1.window, G1.window + WSIZE, WSIZE);
646 G1.match_start -= WSIZE;
647 G1.strstart -= WSIZE; /* we now have strstart >= MAX_DIST: */
648
649 G1.block_start -= WSIZE;
650
651 for (n = 0; n < HASH_SIZE; n++) {
652 m = head[n];
653 head[n] = (Pos) (m >= WSIZE ? m - WSIZE : 0);
654 }
655 for (n = 0; n < WSIZE; n++) {
656 m = G1.prev[n];
657 G1.prev[n] = (Pos) (m >= WSIZE ? m - WSIZE : 0);
658 /* If n is not on any hash chain, prev[n] is garbage but
659 * its value will never be used.
660 */
661 }
662 more += WSIZE;
663 }
664 /* At this point, more >= 2 */
665 if (!G1.eofile) {
666 n = file_read(G1.window + G1.strstart + G1.lookahead, more);
667 if (n == 0 || n == (unsigned) -1) {
668 G1.eofile = 1;
669 } else {
670 G1.lookahead += n;
671 }
672 }
673}
674
675
676/* ===========================================================================
677 * Set match_start to the longest match starting at the given string and
678 * return its length. Matches shorter or equal to prev_length are discarded,
679 * in which case the result is equal to prev_length and match_start is
680 * garbage.
681 * IN assertions: cur_match is the head of the hash chain for the current
682 * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
683 */
684
685/* For MSDOS, OS/2 and 386 Unix, an optimized version is in match.asm or
686 * match.s. The code is functionally equivalent, so you can use the C version
687 * if desired.
688 */
689static int longest_match(IPos cur_match)
690{
691 unsigned chain_length = max_chain_length; /* max hash chain length */
692 uch *scan = G1.window + G1.strstart; /* current string */
693 uch *match; /* matched string */
694 int len; /* length of current match */
695 int best_len = G1.prev_length; /* best match length so far */
696 IPos limit = G1.strstart > (IPos) MAX_DIST ? G1.strstart - (IPos) MAX_DIST : 0;
697 /* Stop when cur_match becomes <= limit. To simplify the code,
698 * we prevent matches with the string of window index 0.
699 */
700
701/* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
702 * It is easy to get rid of this optimization if necessary.
703 */
704#if HASH_BITS < 8 || MAX_MATCH != 258
705# error Code too clever
706#endif
707 uch *strend = G1.window + G1.strstart + MAX_MATCH;
708 uch scan_end1 = scan[best_len - 1];
709 uch scan_end = scan[best_len];
710
711 /* Do not waste too much time if we already have a good match: */
712 if (G1.prev_length >= good_match) {
713 chain_length >>= 2;
714 }
715 Assert(G1.strstart <= WINDOW_SIZE - MIN_LOOKAHEAD, "insufficient lookahead");
716
717 do {
718 Assert(cur_match < G1.strstart, "no future");
719 match = G1.window + cur_match;
720
721 /* Skip to next match if the match length cannot increase
722 * or if the match length is less than 2:
723 */
724 if (match[best_len] != scan_end
725 || match[best_len - 1] != scan_end1
726 || *match != *scan || *++match != scan[1]
727 ) {
728 continue;
729 }
730
731 /* The check at best_len-1 can be removed because it will be made
732 * again later. (This heuristic is not always a win.)
733 * It is not necessary to compare scan[2] and match[2] since they
734 * are always equal when the other bytes match, given that
735 * the hash keys are equal and that HASH_BITS >= 8.
736 */
737 scan += 2, match++;
738
739 /* We check for insufficient lookahead only every 8th comparison;
740 * the 256th check will be made at strstart+258.
741 */
742 do {
743 } while (*++scan == *++match && *++scan == *++match &&
744 *++scan == *++match && *++scan == *++match &&
745 *++scan == *++match && *++scan == *++match &&
746 *++scan == *++match && *++scan == *++match && scan < strend);
747
748 len = MAX_MATCH - (int) (strend - scan);
749 scan = strend - MAX_MATCH;
750
751 if (len > best_len) {
752 G1.match_start = cur_match;
753 best_len = len;
754 if (len >= nice_match)
755 break;
756 scan_end1 = scan[best_len - 1];
757 scan_end = scan[best_len];
758 }
759 } while ((cur_match = G1.prev[cur_match & WMASK]) > limit
760 && --chain_length != 0);
761
762 return best_len;
763}
764
765
766#ifdef DEBUG
767/* ===========================================================================
768 * Check that the match at match_start is indeed a match.
769 */
770static void check_match(IPos start, IPos match, int length)
771{
772 /* check that the match is indeed a match */
773 if (memcmp(G1.window + match, G1.window + start, length) != 0) {
774 bb_error_msg(" start %d, match %d, length %d", start, match, length);
775 bb_error_msg("invalid match");
776 }
777 if (verbose > 1) {
778 bb_error_msg("\\[%d,%d]", start - match, length);
779 do {
780 bb_putchar_stderr(G1.window[start++]);
781 } while (--length != 0);
782 }
783}
784#else
785# define check_match(start, match, length) ((void)0)
786#endif
787
788
789/* trees.c -- output deflated data using Huffman coding
790 * Copyright (C) 1992-1993 Jean-loup Gailly
791 * This is free software; you can redistribute it and/or modify it under the
792 * terms of the GNU General Public License, see the file COPYING.
793 */
794
795/* PURPOSE
796 * Encode various sets of source values using variable-length
797 * binary code trees.
798 *
799 * DISCUSSION
800 * The PKZIP "deflation" process uses several Huffman trees. The more
801 * common source values are represented by shorter bit sequences.
802 *
803 * Each code tree is stored in the ZIP file in a compressed form
804 * which is itself a Huffman encoding of the lengths of
805 * all the code strings (in ascending order by source values).
806 * The actual code strings are reconstructed from the lengths in
807 * the UNZIP process, as described in the "application note"
808 * (APPNOTE.TXT) distributed as part of PKWARE's PKZIP program.
809 *
810 * REFERENCES
811 * Lynch, Thomas J.
812 * Data Compression: Techniques and Applications, pp. 53-55.
813 * Lifetime Learning Publications, 1985. ISBN 0-534-03418-7.
814 *
815 * Storer, James A.
816 * Data Compression: Methods and Theory, pp. 49-50.
817 * Computer Science Press, 1988. ISBN 0-7167-8156-5.
818 *
819 * Sedgewick, R.
820 * Algorithms, p290.
821 * Addison-Wesley, 1983. ISBN 0-201-06672-6.
822 *
823 * INTERFACE
824 * void ct_init()
825 * Allocate the match buffer, initialize the various tables [and save
826 * the location of the internal file attribute (ascii/binary) and
827 * method (DEFLATE/STORE) -- deleted in bbox]
828 *
829 * void ct_tally(int dist, int lc);
830 * Save the match info and tally the frequency counts.
831 *
832 * ulg flush_block(char *buf, ulg stored_len, int eof)
833 * Determine the best encoding for the current block: dynamic trees,
834 * static trees or store, and output the encoded block to the zip
835 * file. Returns the total compressed length for the file so far.
836 */
837
838#define MAX_BITS 15
839/* All codes must not exceed MAX_BITS bits */
840
841#define MAX_BL_BITS 7
842/* Bit length codes must not exceed MAX_BL_BITS bits */
843
844#define LENGTH_CODES 29
845/* number of length codes, not counting the special END_BLOCK code */
846
847#define LITERALS 256
848/* number of literal bytes 0..255 */
849
850#define END_BLOCK 256
851/* end of block literal code */
852
853#define L_CODES (LITERALS+1+LENGTH_CODES)
854/* number of Literal or Length codes, including the END_BLOCK code */
855
856#define D_CODES 30
857/* number of distance codes */
858
859#define BL_CODES 19
860/* number of codes used to transfer the bit lengths */
861
862/* extra bits for each length code */
863static const uint8_t extra_lbits[LENGTH_CODES] ALIGN1 = {
864 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4,
865 4, 4, 5, 5, 5, 5, 0
866};
867
868/* extra bits for each distance code */
869static const uint8_t extra_dbits[D_CODES] ALIGN1 = {
870 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9,
871 10, 10, 11, 11, 12, 12, 13, 13
872};
873
874/* extra bits for each bit length code */
875static const uint8_t extra_blbits[BL_CODES] ALIGN1 = {
876 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 3, 7 };
877
878/* number of codes at each bit length for an optimal tree */
879static const uint8_t bl_order[BL_CODES] ALIGN1 = {
880 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15 };
881
882#define STORED_BLOCK 0
883#define STATIC_TREES 1
884#define DYN_TREES 2
885/* The three kinds of block type */
886
887#ifndef LIT_BUFSIZE
888# ifdef SMALL_MEM
889# define LIT_BUFSIZE 0x2000
890# else
891# ifdef MEDIUM_MEM
892# define LIT_BUFSIZE 0x4000
893# else
894# define LIT_BUFSIZE 0x8000
895# endif
896# endif
897#endif
898#ifndef DIST_BUFSIZE
899# define DIST_BUFSIZE LIT_BUFSIZE
900#endif
901/* Sizes of match buffers for literals/lengths and distances. There are
902 * 4 reasons for limiting LIT_BUFSIZE to 64K:
903 * - frequencies can be kept in 16 bit counters
904 * - if compression is not successful for the first block, all input data is
905 * still in the window so we can still emit a stored block even when input
906 * comes from standard input. (This can also be done for all blocks if
907 * LIT_BUFSIZE is not greater than 32K.)
908 * - if compression is not successful for a file smaller than 64K, we can
909 * even emit a stored file instead of a stored block (saving 5 bytes).
910 * - creating new Huffman trees less frequently may not provide fast
911 * adaptation to changes in the input data statistics. (Take for
912 * example a binary file with poorly compressible code followed by
913 * a highly compressible string table.) Smaller buffer sizes give
914 * fast adaptation but have of course the overhead of transmitting trees
915 * more frequently.
916 * - I can't count above 4
917 * The current code is general and allows DIST_BUFSIZE < LIT_BUFSIZE (to save
918 * memory at the expense of compression). Some optimizations would be possible
919 * if we rely on DIST_BUFSIZE == LIT_BUFSIZE.
920 */
921#define REP_3_6 16
922/* repeat previous bit length 3-6 times (2 bits of repeat count) */
923#define REPZ_3_10 17
924/* repeat a zero length 3-10 times (3 bits of repeat count) */
925#define REPZ_11_138 18
926/* repeat a zero length 11-138 times (7 bits of repeat count) */
927
928/* ===========================================================================
929*/
930/* Data structure describing a single value and its code string. */
931typedef struct ct_data {
932 union {
933 ush freq; /* frequency count */
934 ush code; /* bit string */
935 } fc;
936 union {
937 ush dad; /* father node in Huffman tree */
938 ush len; /* length of bit string */
939 } dl;
940} ct_data;
941
942#define Freq fc.freq
943#define Code fc.code
944#define Dad dl.dad
945#define Len dl.len
946
947#define HEAP_SIZE (2*L_CODES + 1)
948/* maximum heap size */
949
950typedef struct tree_desc {
951 ct_data *dyn_tree; /* the dynamic tree */
952 ct_data *static_tree; /* corresponding static tree or NULL */
953 const uint8_t *extra_bits; /* extra bits for each code or NULL */
954 int extra_base; /* base index for extra_bits */
955 int elems; /* max number of elements in the tree */
956 int max_length; /* max bit length for the codes */
957 int max_code; /* largest code with non zero frequency */
958} tree_desc;
959
960struct globals2 {
961
962 ush heap[HEAP_SIZE]; /* heap used to build the Huffman trees */
963 int heap_len; /* number of elements in the heap */
964 int heap_max; /* element of largest frequency */
965
966/* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used.
967 * The same heap array is used to build all trees.
968 */
969
970 ct_data dyn_ltree[HEAP_SIZE]; /* literal and length tree */
971 ct_data dyn_dtree[2 * D_CODES + 1]; /* distance tree */
972
973 ct_data static_ltree[L_CODES + 2];
974
975/* The static literal tree. Since the bit lengths are imposed, there is no
976 * need for the L_CODES extra codes used during heap construction. However
977 * The codes 286 and 287 are needed to build a canonical tree (see ct_init
978 * below).
979 */
980
981 ct_data static_dtree[D_CODES];
982
983/* The static distance tree. (Actually a trivial tree since all codes use
984 * 5 bits.)
985 */
986
987 ct_data bl_tree[2 * BL_CODES + 1];
988
989/* Huffman tree for the bit lengths */
990
991 tree_desc l_desc;
992 tree_desc d_desc;
993 tree_desc bl_desc;
994
995 ush bl_count[MAX_BITS + 1];
996
997/* The lengths of the bit length codes are sent in order of decreasing
998 * probability, to avoid transmitting the lengths for unused bit length codes.
999 */
1000
1001 uch depth[2 * L_CODES + 1];
1002
1003/* Depth of each subtree used as tie breaker for trees of equal frequency */
1004
1005 uch length_code[MAX_MATCH - MIN_MATCH + 1];
1006
1007/* length code for each normalized match length (0 == MIN_MATCH) */
1008
1009 uch dist_code[512];
1010
1011/* distance codes. The first 256 values correspond to the distances
1012 * 3 .. 258, the last 256 values correspond to the top 8 bits of
1013 * the 15 bit distances.
1014 */
1015
1016 int base_length[LENGTH_CODES];
1017
1018/* First normalized length for each code (0 = MIN_MATCH) */
1019
1020 int base_dist[D_CODES];
1021
1022/* First normalized distance for each code (0 = distance of 1) */
1023
1024 uch flag_buf[LIT_BUFSIZE / 8];
1025
1026/* flag_buf is a bit array distinguishing literals from lengths in
1027 * l_buf, thus indicating the presence or absence of a distance.
1028 */
1029
1030 unsigned last_lit; /* running index in l_buf */
1031 unsigned last_dist; /* running index in d_buf */
1032 unsigned last_flags; /* running index in flag_buf */
1033 uch flags; /* current flags not yet saved in flag_buf */
1034 uch flag_bit; /* current bit used in flags */
1035
1036/* bits are filled in flags starting at bit 0 (least significant).
1037 * Note: these flags are overkill in the current code since we don't
1038 * take advantage of DIST_BUFSIZE == LIT_BUFSIZE.
1039 */
1040
1041 ulg opt_len; /* bit length of current block with optimal trees */
1042 ulg static_len; /* bit length of current block with static trees */
1043
1044 ulg compressed_len; /* total bit length of compressed file */
1045};
1046
1047#define G2ptr ((struct globals2*)(ptr_to_globals))
1048#define G2 (*G2ptr)
1049
1050
1051/* ===========================================================================
1052 */
1053static void gen_codes(ct_data * tree, int max_code);
1054static void build_tree(tree_desc * desc);
1055static void scan_tree(ct_data * tree, int max_code);
1056static void send_tree(ct_data * tree, int max_code);
1057static int build_bl_tree(void);
1058static void send_all_trees(int lcodes, int dcodes, int blcodes);
1059static void compress_block(ct_data * ltree, ct_data * dtree);
1060
1061
1062#ifndef DEBUG
1063/* Send a code of the given tree. c and tree must not have side effects */
1064# define SEND_CODE(c, tree) send_bits(tree[c].Code, tree[c].Len)
1065#else
1066# define SEND_CODE(c, tree) \
1067{ \
1068 if (verbose > 1) bb_error_msg("\ncd %3d ", (c)); \
1069 send_bits(tree[c].Code, tree[c].Len); \
1070}
1071#endif
1072
1073#define D_CODE(dist) \
1074 ((dist) < 256 ? G2.dist_code[dist] : G2.dist_code[256 + ((dist)>>7)])
1075/* Mapping from a distance to a distance code. dist is the distance - 1 and
1076 * must not have side effects. dist_code[256] and dist_code[257] are never
1077 * used.
1078 * The arguments must not have side effects.
1079 */
1080
1081
1082/* ===========================================================================
1083 * Initialize a new block.
1084 */
1085static void init_block(void)
1086{
1087 int n; /* iterates over tree elements */
1088
1089 /* Initialize the trees. */
1090 for (n = 0; n < L_CODES; n++)
1091 G2.dyn_ltree[n].Freq = 0;
1092 for (n = 0; n < D_CODES; n++)
1093 G2.dyn_dtree[n].Freq = 0;
1094 for (n = 0; n < BL_CODES; n++)
1095 G2.bl_tree[n].Freq = 0;
1096
1097 G2.dyn_ltree[END_BLOCK].Freq = 1;
1098 G2.opt_len = G2.static_len = 0;
1099 G2.last_lit = G2.last_dist = G2.last_flags = 0;
1100 G2.flags = 0;
1101 G2.flag_bit = 1;
1102}
1103
1104
1105/* ===========================================================================
1106 * Restore the heap property by moving down the tree starting at node k,
1107 * exchanging a node with the smallest of its two sons if necessary, stopping
1108 * when the heap property is re-established (each father smaller than its
1109 * two sons).
1110 */
1111
1112/* Compares to subtrees, using the tree depth as tie breaker when
1113 * the subtrees have equal frequency. This minimizes the worst case length. */
1114#define SMALLER(tree, n, m) \
1115 (tree[n].Freq < tree[m].Freq \
1116 || (tree[n].Freq == tree[m].Freq && G2.depth[n] <= G2.depth[m]))
1117
1118static void pqdownheap(ct_data * tree, int k)
1119{
1120 int v = G2.heap[k];
1121 int j = k << 1; /* left son of k */
1122
1123 while (j <= G2.heap_len) {
1124 /* Set j to the smallest of the two sons: */
1125 if (j < G2.heap_len && SMALLER(tree, G2.heap[j + 1], G2.heap[j]))
1126 j++;
1127
1128 /* Exit if v is smaller than both sons */
1129 if (SMALLER(tree, v, G2.heap[j]))
1130 break;
1131
1132 /* Exchange v with the smallest son */
1133 G2.heap[k] = G2.heap[j];
1134 k = j;
1135
1136 /* And continue down the tree, setting j to the left son of k */
1137 j <<= 1;
1138 }
1139 G2.heap[k] = v;
1140}
1141
1142
1143/* ===========================================================================
1144 * Compute the optimal bit lengths for a tree and update the total bit length
1145 * for the current block.
1146 * IN assertion: the fields freq and dad are set, heap[heap_max] and
1147 * above are the tree nodes sorted by increasing frequency.
1148 * OUT assertions: the field len is set to the optimal bit length, the
1149 * array bl_count contains the frequencies for each bit length.
1150 * The length opt_len is updated; static_len is also updated if stree is
1151 * not null.
1152 */
1153static void gen_bitlen(tree_desc * desc)
1154{
1155 ct_data *tree = desc->dyn_tree;
1156 const uint8_t *extra = desc->extra_bits;
1157 int base = desc->extra_base;
1158 int max_code = desc->max_code;
1159 int max_length = desc->max_length;
1160 ct_data *stree = desc->static_tree;
1161 int h; /* heap index */
1162 int n, m; /* iterate over the tree elements */
1163 int bits; /* bit length */
1164 int xbits; /* extra bits */
1165 ush f; /* frequency */
1166 int overflow = 0; /* number of elements with bit length too large */
1167
1168 for (bits = 0; bits <= MAX_BITS; bits++)
1169 G2.bl_count[bits] = 0;
1170
1171 /* In a first pass, compute the optimal bit lengths (which may
1172 * overflow in the case of the bit length tree).
1173 */
1174 tree[G2.heap[G2.heap_max]].Len = 0; /* root of the heap */
1175
1176 for (h = G2.heap_max + 1; h < HEAP_SIZE; h++) {
1177 n = G2.heap[h];
1178 bits = tree[tree[n].Dad].Len + 1;
1179 if (bits > max_length) {
1180 bits = max_length;
1181 overflow++;
1182 }
1183 tree[n].Len = (ush) bits;
1184 /* We overwrite tree[n].Dad which is no longer needed */
1185
1186 if (n > max_code)
1187 continue; /* not a leaf node */
1188
1189 G2.bl_count[bits]++;
1190 xbits = 0;
1191 if (n >= base)
1192 xbits = extra[n - base];
1193 f = tree[n].Freq;
1194 G2.opt_len += (ulg) f *(bits + xbits);
1195
1196 if (stree)
1197 G2.static_len += (ulg) f * (stree[n].Len + xbits);
1198 }
1199 if (overflow == 0)
1200 return;
1201
1202 Trace((stderr, "\nbit length overflow\n"));
1203 /* This happens for example on obj2 and pic of the Calgary corpus */
1204
1205 /* Find the first bit length which could increase: */
1206 do {
1207 bits = max_length - 1;
1208 while (G2.bl_count[bits] == 0)
1209 bits--;
1210 G2.bl_count[bits]--; /* move one leaf down the tree */
1211 G2.bl_count[bits + 1] += 2; /* move one overflow item as its brother */
1212 G2.bl_count[max_length]--;
1213 /* The brother of the overflow item also moves one step up,
1214 * but this does not affect bl_count[max_length]
1215 */
1216 overflow -= 2;
1217 } while (overflow > 0);
1218
1219 /* Now recompute all bit lengths, scanning in increasing frequency.
1220 * h is still equal to HEAP_SIZE. (It is simpler to reconstruct all
1221 * lengths instead of fixing only the wrong ones. This idea is taken
1222 * from 'ar' written by Haruhiko Okumura.)
1223 */
1224 for (bits = max_length; bits != 0; bits--) {
1225 n = G2.bl_count[bits];
1226 while (n != 0) {
1227 m = G2.heap[--h];
1228 if (m > max_code)
1229 continue;
1230 if (tree[m].Len != (unsigned) bits) {
1231 Trace((stderr, "code %d bits %d->%d\n", m, tree[m].Len, bits));
1232 G2.opt_len += ((int32_t) bits - tree[m].Len) * tree[m].Freq;
1233 tree[m].Len = bits;
1234 }
1235 n--;
1236 }
1237 }
1238}
1239
1240
1241/* ===========================================================================
1242 * Generate the codes for a given tree and bit counts (which need not be
1243 * optimal).
1244 * IN assertion: the array bl_count contains the bit length statistics for
1245 * the given tree and the field len is set for all tree elements.
1246 * OUT assertion: the field code is set for all tree elements of non
1247 * zero code length.
1248 */
1249static void gen_codes(ct_data * tree, int max_code)
1250{
1251 ush next_code[MAX_BITS + 1]; /* next code value for each bit length */
1252 ush code = 0; /* running code value */
1253 int bits; /* bit index */
1254 int n; /* code index */
1255
1256 /* The distribution counts are first used to generate the code values
1257 * without bit reversal.
1258 */
1259 for (bits = 1; bits <= MAX_BITS; bits++) {
1260 next_code[bits] = code = (code + G2.bl_count[bits - 1]) << 1;
1261 }
1262 /* Check that the bit counts in bl_count are consistent. The last code
1263 * must be all ones.
1264 */
1265 Assert(code + G2.bl_count[MAX_BITS] - 1 == (1 << MAX_BITS) - 1,
1266 "inconsistent bit counts");
1267 Tracev((stderr, "\ngen_codes: max_code %d ", max_code));
1268
1269 for (n = 0; n <= max_code; n++) {
1270 int len = tree[n].Len;
1271
1272 if (len == 0)
1273 continue;
1274 /* Now reverse the bits */
1275 tree[n].Code = bi_reverse(next_code[len]++, len);
1276
1277 Tracec(tree != G2.static_ltree,
1278 (stderr, "\nn %3d %c l %2d c %4x (%x) ", n,
1279 (n > ' ' ? n : ' '), len, tree[n].Code,
1280 next_code[len] - 1));
1281 }
1282}
1283
1284
1285/* ===========================================================================
1286 * Construct one Huffman tree and assigns the code bit strings and lengths.
1287 * Update the total bit length for the current block.
1288 * IN assertion: the field freq is set for all tree elements.
1289 * OUT assertions: the fields len and code are set to the optimal bit length
1290 * and corresponding code. The length opt_len is updated; static_len is
1291 * also updated if stree is not null. The field max_code is set.
1292 */
1293
1294/* Remove the smallest element from the heap and recreate the heap with
1295 * one less element. Updates heap and heap_len. */
1296
1297#define SMALLEST 1
1298/* Index within the heap array of least frequent node in the Huffman tree */
1299
1300#define PQREMOVE(tree, top) \
1301do { \
1302 top = G2.heap[SMALLEST]; \
1303 G2.heap[SMALLEST] = G2.heap[G2.heap_len--]; \
1304 pqdownheap(tree, SMALLEST); \
1305} while (0)
1306
1307static void build_tree(tree_desc * desc)
1308{
1309 ct_data *tree = desc->dyn_tree;
1310 ct_data *stree = desc->static_tree;
1311 int elems = desc->elems;
1312 int n, m; /* iterate over heap elements */
1313 int max_code = -1; /* largest code with non zero frequency */
1314 int node = elems; /* next internal node of the tree */
1315
1316 /* Construct the initial heap, with least frequent element in
1317 * heap[SMALLEST]. The sons of heap[n] are heap[2*n] and heap[2*n+1].
1318 * heap[0] is not used.
1319 */
1320 G2.heap_len = 0;
1321 G2.heap_max = HEAP_SIZE;
1322
1323 for (n = 0; n < elems; n++) {
1324 if (tree[n].Freq != 0) {
1325 G2.heap[++G2.heap_len] = max_code = n;
1326 G2.depth[n] = 0;
1327 } else {
1328 tree[n].Len = 0;
1329 }
1330 }
1331
1332 /* The pkzip format requires that at least one distance code exists,
1333 * and that at least one bit should be sent even if there is only one
1334 * possible code. So to avoid special checks later on we force at least
1335 * two codes of non zero frequency.
1336 */
1337 while (G2.heap_len < 2) {
1338 int new = G2.heap[++G2.heap_len] = (max_code < 2 ? ++max_code : 0);
1339
1340 tree[new].Freq = 1;
1341 G2.depth[new] = 0;
1342 G2.opt_len--;
1343 if (stree)
1344 G2.static_len -= stree[new].Len;
1345 /* new is 0 or 1 so it does not have extra bits */
1346 }
1347 desc->max_code = max_code;
1348
1349 /* The elements heap[heap_len/2+1 .. heap_len] are leaves of the tree,
1350 * establish sub-heaps of increasing lengths:
1351 */
1352 for (n = G2.heap_len / 2; n >= 1; n--)
1353 pqdownheap(tree, n);
1354
1355 /* Construct the Huffman tree by repeatedly combining the least two
1356 * frequent nodes.
1357 */
1358 do {
1359 PQREMOVE(tree, n); /* n = node of least frequency */
1360 m = G2.heap[SMALLEST]; /* m = node of next least frequency */
1361
1362 G2.heap[--G2.heap_max] = n; /* keep the nodes sorted by frequency */
1363 G2.heap[--G2.heap_max] = m;
1364
1365 /* Create a new node father of n and m */
1366 tree[node].Freq = tree[n].Freq + tree[m].Freq;
1367 G2.depth[node] = MAX(G2.depth[n], G2.depth[m]) + 1;
1368 tree[n].Dad = tree[m].Dad = (ush) node;
1369#ifdef DUMP_BL_TREE
1370 if (tree == G2.bl_tree) {
1371 bb_error_msg("\nnode %d(%d), sons %d(%d) %d(%d)",
1372 node, tree[node].Freq, n, tree[n].Freq, m, tree[m].Freq);
1373 }
1374#endif
1375 /* and insert the new node in the heap */
1376 G2.heap[SMALLEST] = node++;
1377 pqdownheap(tree, SMALLEST);
1378 } while (G2.heap_len >= 2);
1379
1380 G2.heap[--G2.heap_max] = G2.heap[SMALLEST];
1381
1382 /* At this point, the fields freq and dad are set. We can now
1383 * generate the bit lengths.
1384 */
1385 gen_bitlen((tree_desc *) desc);
1386
1387 /* The field len is now set, we can generate the bit codes */
1388 gen_codes((ct_data *) tree, max_code);
1389}
1390
1391
1392/* ===========================================================================
1393 * Scan a literal or distance tree to determine the frequencies of the codes
1394 * in the bit length tree. Updates opt_len to take into account the repeat
1395 * counts. (The contribution of the bit length codes will be added later
1396 * during the construction of bl_tree.)
1397 */
1398static void scan_tree(ct_data * tree, int max_code)
1399{
1400 int n; /* iterates over all tree elements */
1401 int prevlen = -1; /* last emitted length */
1402 int curlen; /* length of current code */
1403 int nextlen = tree[0].Len; /* length of next code */
1404 int count = 0; /* repeat count of the current code */
1405 int max_count = 7; /* max repeat count */
1406 int min_count = 4; /* min repeat count */
1407
1408 if (nextlen == 0) {
1409 max_count = 138;
1410 min_count = 3;
1411 }
1412 tree[max_code + 1].Len = 0xffff; /* guard */
1413
1414 for (n = 0; n <= max_code; n++) {
1415 curlen = nextlen;
1416 nextlen = tree[n + 1].Len;
1417 if (++count < max_count && curlen == nextlen)
1418 continue;
1419
1420 if (count < min_count) {
1421 G2.bl_tree[curlen].Freq += count;
1422 } else if (curlen != 0) {
1423 if (curlen != prevlen)
1424 G2.bl_tree[curlen].Freq++;
1425 G2.bl_tree[REP_3_6].Freq++;
1426 } else if (count <= 10) {
1427 G2.bl_tree[REPZ_3_10].Freq++;
1428 } else {
1429 G2.bl_tree[REPZ_11_138].Freq++;
1430 }
1431 count = 0;
1432 prevlen = curlen;
1433
1434 max_count = 7;
1435 min_count = 4;
1436 if (nextlen == 0) {
1437 max_count = 138;
1438 min_count = 3;
1439 } else if (curlen == nextlen) {
1440 max_count = 6;
1441 min_count = 3;
1442 }
1443 }
1444}
1445
1446
1447/* ===========================================================================
1448 * Send a literal or distance tree in compressed form, using the codes in
1449 * bl_tree.
1450 */
1451static void send_tree(ct_data * tree, int max_code)
1452{
1453 int n; /* iterates over all tree elements */
1454 int prevlen = -1; /* last emitted length */
1455 int curlen; /* length of current code */
1456 int nextlen = tree[0].Len; /* length of next code */
1457 int count = 0; /* repeat count of the current code */
1458 int max_count = 7; /* max repeat count */
1459 int min_count = 4; /* min repeat count */
1460
1461/* tree[max_code+1].Len = -1; *//* guard already set */
1462 if (nextlen == 0)
1463 max_count = 138, min_count = 3;
1464
1465 for (n = 0; n <= max_code; n++) {
1466 curlen = nextlen;
1467 nextlen = tree[n + 1].Len;
1468 if (++count < max_count && curlen == nextlen) {
1469 continue;
1470 } else if (count < min_count) {
1471 do {
1472 SEND_CODE(curlen, G2.bl_tree);
1473 } while (--count);
1474 } else if (curlen != 0) {
1475 if (curlen != prevlen) {
1476 SEND_CODE(curlen, G2.bl_tree);
1477 count--;
1478 }
1479 Assert(count >= 3 && count <= 6, " 3_6?");
1480 SEND_CODE(REP_3_6, G2.bl_tree);
1481 send_bits(count - 3, 2);
1482 } else if (count <= 10) {
1483 SEND_CODE(REPZ_3_10, G2.bl_tree);
1484 send_bits(count - 3, 3);
1485 } else {
1486 SEND_CODE(REPZ_11_138, G2.bl_tree);
1487 send_bits(count - 11, 7);
1488 }
1489 count = 0;
1490 prevlen = curlen;
1491 if (nextlen == 0) {
1492 max_count = 138;
1493 min_count = 3;
1494 } else if (curlen == nextlen) {
1495 max_count = 6;
1496 min_count = 3;
1497 } else {
1498 max_count = 7;
1499 min_count = 4;
1500 }
1501 }
1502}
1503
1504
1505/* ===========================================================================
1506 * Construct the Huffman tree for the bit lengths and return the index in
1507 * bl_order of the last bit length code to send.
1508 */
1509static int build_bl_tree(void)
1510{
1511 int max_blindex; /* index of last bit length code of non zero freq */
1512
1513 /* Determine the bit length frequencies for literal and distance trees */
1514 scan_tree(G2.dyn_ltree, G2.l_desc.max_code);
1515 scan_tree(G2.dyn_dtree, G2.d_desc.max_code);
1516
1517 /* Build the bit length tree: */
1518 build_tree(&G2.bl_desc);
1519 /* opt_len now includes the length of the tree representations, except
1520 * the lengths of the bit lengths codes and the 5+5+4 bits for the counts.
1521 */
1522
1523 /* Determine the number of bit length codes to send. The pkzip format
1524 * requires that at least 4 bit length codes be sent. (appnote.txt says
1525 * 3 but the actual value used is 4.)
1526 */
1527 for (max_blindex = BL_CODES - 1; max_blindex >= 3; max_blindex--) {
1528 if (G2.bl_tree[bl_order[max_blindex]].Len != 0)
1529 break;
1530 }
1531 /* Update opt_len to include the bit length tree and counts */
1532 G2.opt_len += 3 * (max_blindex + 1) + 5 + 5 + 4;
1533 Tracev((stderr, "\ndyn trees: dyn %ld, stat %ld", G2.opt_len, G2.static_len));
1534
1535 return max_blindex;
1536}
1537
1538
1539/* ===========================================================================
1540 * Send the header for a block using dynamic Huffman trees: the counts, the
1541 * lengths of the bit length codes, the literal tree and the distance tree.
1542 * IN assertion: lcodes >= 257, dcodes >= 1, blcodes >= 4.
1543 */
1544static void send_all_trees(int lcodes, int dcodes, int blcodes)
1545{
1546 int rank; /* index in bl_order */
1547
1548 Assert(lcodes >= 257 && dcodes >= 1 && blcodes >= 4, "not enough codes");
1549 Assert(lcodes <= L_CODES && dcodes <= D_CODES
1550 && blcodes <= BL_CODES, "too many codes");
1551 Tracev((stderr, "\nbl counts: "));
1552 send_bits(lcodes - 257, 5); /* not +255 as stated in appnote.txt */
1553 send_bits(dcodes - 1, 5);
1554 send_bits(blcodes - 4, 4); /* not -3 as stated in appnote.txt */
1555 for (rank = 0; rank < blcodes; rank++) {
1556 Tracev((stderr, "\nbl code %2d ", bl_order[rank]));
1557 send_bits(G2.bl_tree[bl_order[rank]].Len, 3);
1558 }
1559 Tracev((stderr, "\nbl tree: sent %ld", G1.bits_sent));
1560
1561 send_tree((ct_data *) G2.dyn_ltree, lcodes - 1); /* send the literal tree */
1562 Tracev((stderr, "\nlit tree: sent %ld", G1.bits_sent));
1563
1564 send_tree((ct_data *) G2.dyn_dtree, dcodes - 1); /* send the distance tree */
1565 Tracev((stderr, "\ndist tree: sent %ld", G1.bits_sent));
1566}
1567
1568
1569/* ===========================================================================
1570 * Save the match info and tally the frequency counts. Return true if
1571 * the current block must be flushed.
1572 */
1573static int ct_tally(int dist, int lc)
1574{
1575 G1.l_buf[G2.last_lit++] = lc;
1576 if (dist == 0) {
1577 /* lc is the unmatched char */
1578 G2.dyn_ltree[lc].Freq++;
1579 } else {
1580 /* Here, lc is the match length - MIN_MATCH */
1581 dist--; /* dist = match distance - 1 */
1582 Assert((ush) dist < (ush) MAX_DIST
1583 && (ush) lc <= (ush) (MAX_MATCH - MIN_MATCH)
1584 && (ush) D_CODE(dist) < (ush) D_CODES, "ct_tally: bad match"
1585 );
1586
1587 G2.dyn_ltree[G2.length_code[lc] + LITERALS + 1].Freq++;
1588 G2.dyn_dtree[D_CODE(dist)].Freq++;
1589
1590 G1.d_buf[G2.last_dist++] = dist;
1591 G2.flags |= G2.flag_bit;
1592 }
1593 G2.flag_bit <<= 1;
1594
1595 /* Output the flags if they fill a byte: */
1596 if ((G2.last_lit & 7) == 0) {
1597 G2.flag_buf[G2.last_flags++] = G2.flags;
1598 G2.flags = 0;
1599 G2.flag_bit = 1;
1600 }
1601 /* Try to guess if it is profitable to stop the current block here */
1602 if ((G2.last_lit & 0xfff) == 0) {
1603 /* Compute an upper bound for the compressed length */
1604 ulg out_length = G2.last_lit * 8L;
1605 ulg in_length = (ulg) G1.strstart - G1.block_start;
1606 int dcode;
1607
1608 for (dcode = 0; dcode < D_CODES; dcode++) {
1609 out_length += G2.dyn_dtree[dcode].Freq * (5L + extra_dbits[dcode]);
1610 }
1611 out_length >>= 3;
1612 Trace((stderr,
1613 "\nlast_lit %u, last_dist %u, in %ld, out ~%ld(%ld%%) ",
1614 G2.last_lit, G2.last_dist, in_length, out_length,
1615 100L - out_length * 100L / in_length));
1616 if (G2.last_dist < G2.last_lit / 2 && out_length < in_length / 2)
1617 return 1;
1618 }
1619 return (G2.last_lit == LIT_BUFSIZE - 1 || G2.last_dist == DIST_BUFSIZE);
1620 /* We avoid equality with LIT_BUFSIZE because of wraparound at 64K
1621 * on 16 bit machines and because stored blocks are restricted to
1622 * 64K-1 bytes.
1623 */
1624}
1625
1626/* ===========================================================================
1627 * Send the block data compressed using the given Huffman trees
1628 */
1629static void compress_block(ct_data * ltree, ct_data * dtree)
1630{
1631 unsigned dist; /* distance of matched string */
1632 int lc; /* match length or unmatched char (if dist == 0) */
1633 unsigned lx = 0; /* running index in l_buf */
1634 unsigned dx = 0; /* running index in d_buf */
1635 unsigned fx = 0; /* running index in flag_buf */
1636 uch flag = 0; /* current flags */
1637 unsigned code; /* the code to send */
1638 int extra; /* number of extra bits to send */
1639
1640 if (G2.last_lit != 0) do {
1641 if ((lx & 7) == 0)
1642 flag = G2.flag_buf[fx++];
1643 lc = G1.l_buf[lx++];
1644 if ((flag & 1) == 0) {
1645 SEND_CODE(lc, ltree); /* send a literal byte */
1646 Tracecv(lc > ' ', (stderr, " '%c' ", lc));
1647 } else {
1648 /* Here, lc is the match length - MIN_MATCH */
1649 code = G2.length_code[lc];
1650 SEND_CODE(code + LITERALS + 1, ltree); /* send the length code */
1651 extra = extra_lbits[code];
1652 if (extra != 0) {
1653 lc -= G2.base_length[code];
1654 send_bits(lc, extra); /* send the extra length bits */
1655 }
1656 dist = G1.d_buf[dx++];
1657 /* Here, dist is the match distance - 1 */
1658 code = D_CODE(dist);
1659 Assert(code < D_CODES, "bad d_code");
1660
1661 SEND_CODE(code, dtree); /* send the distance code */
1662 extra = extra_dbits[code];
1663 if (extra != 0) {
1664 dist -= G2.base_dist[code];
1665 send_bits(dist, extra); /* send the extra distance bits */
1666 }
1667 } /* literal or match pair ? */
1668 flag >>= 1;
1669 } while (lx < G2.last_lit);
1670
1671 SEND_CODE(END_BLOCK, ltree);
1672}
1673
1674
1675/* ===========================================================================
1676 * Determine the best encoding for the current block: dynamic trees, static
1677 * trees or store, and output the encoded block to the zip file. This function
1678 * returns the total compressed length for the file so far.
1679 */
1680static ulg flush_block(char *buf, ulg stored_len, int eof)
1681{
1682 ulg opt_lenb, static_lenb; /* opt_len and static_len in bytes */
1683 int max_blindex; /* index of last bit length code of non zero freq */
1684
1685 G2.flag_buf[G2.last_flags] = G2.flags; /* Save the flags for the last 8 items */
1686
1687 /* Construct the literal and distance trees */
1688 build_tree(&G2.l_desc);
1689 Tracev((stderr, "\nlit data: dyn %ld, stat %ld", G2.opt_len, G2.static_len));
1690
1691 build_tree(&G2.d_desc);
1692 Tracev((stderr, "\ndist data: dyn %ld, stat %ld", G2.opt_len, G2.static_len));
1693 /* At this point, opt_len and static_len are the total bit lengths of
1694 * the compressed block data, excluding the tree representations.
1695 */
1696
1697 /* Build the bit length tree for the above two trees, and get the index
1698 * in bl_order of the last bit length code to send.
1699 */
1700 max_blindex = build_bl_tree();
1701
1702 /* Determine the best encoding. Compute first the block length in bytes */
1703 opt_lenb = (G2.opt_len + 3 + 7) >> 3;
1704 static_lenb = (G2.static_len + 3 + 7) >> 3;
1705
1706 Trace((stderr,
1707 "\nopt %lu(%lu) stat %lu(%lu) stored %lu lit %u dist %u ",
1708 opt_lenb, G2.opt_len, static_lenb, G2.static_len, stored_len,
1709 G2.last_lit, G2.last_dist));
1710
1711 if (static_lenb <= opt_lenb)
1712 opt_lenb = static_lenb;
1713
1714 /* If compression failed and this is the first and last block,
1715 * and if the zip file can be seeked (to rewrite the local header),
1716 * the whole file is transformed into a stored file:
1717 */
1718 if (stored_len <= opt_lenb && eof && G2.compressed_len == 0L && seekable()) {
1719 /* Since LIT_BUFSIZE <= 2*WSIZE, the input data must be there: */
1720 if (buf == NULL)
1721 bb_error_msg("block vanished");
1722
1723 copy_block(buf, (unsigned) stored_len, 0); /* without header */
1724 G2.compressed_len = stored_len << 3;
1725 } else if (stored_len + 4 <= opt_lenb && buf != NULL) {
1726 /* 4: two words for the lengths */
1727 /* The test buf != NULL is only necessary if LIT_BUFSIZE > WSIZE.
1728 * Otherwise we can't have processed more than WSIZE input bytes since
1729 * the last block flush, because compression would have been
1730 * successful. If LIT_BUFSIZE <= WSIZE, it is never too late to
1731 * transform a block into a stored block.
1732 */
1733 send_bits((STORED_BLOCK << 1) + eof, 3); /* send block type */
1734 G2.compressed_len = (G2.compressed_len + 3 + 7) & ~7L;
1735 G2.compressed_len += (stored_len + 4) << 3;
1736
1737 copy_block(buf, (unsigned) stored_len, 1); /* with header */
1738 } else if (static_lenb == opt_lenb) {
1739 send_bits((STATIC_TREES << 1) + eof, 3);
1740 compress_block((ct_data *) G2.static_ltree, (ct_data *) G2.static_dtree);
1741 G2.compressed_len += 3 + G2.static_len;
1742 } else {
1743 send_bits((DYN_TREES << 1) + eof, 3);
1744 send_all_trees(G2.l_desc.max_code + 1, G2.d_desc.max_code + 1,
1745 max_blindex + 1);
1746 compress_block((ct_data *) G2.dyn_ltree, (ct_data *) G2.dyn_dtree);
1747 G2.compressed_len += 3 + G2.opt_len;
1748 }
1749 Assert(G2.compressed_len == G1.bits_sent, "bad compressed size");
1750 init_block();
1751
1752 if (eof) {
1753 bi_windup();
1754 G2.compressed_len += 7; /* align on byte boundary */
1755 }
1756 Tracev((stderr, "\ncomprlen %lu(%lu) ", G2.compressed_len >> 3,
1757 G2.compressed_len - 7 * eof));
1758
1759 return G2.compressed_len >> 3;
1760}
1761
1762
1763/* ===========================================================================
1764 * Update a hash value with the given input byte
1765 * IN assertion: all calls to UPDATE_HASH are made with consecutive
1766 * input characters, so that a running hash key can be computed from the
1767 * previous key instead of complete recalculation each time.
1768 */
1769#define UPDATE_HASH(h, c) (h = (((h)<<H_SHIFT) ^ (c)) & HASH_MASK)
1770
1771
1772/* ===========================================================================
1773 * Same as above, but achieves better compression. We use a lazy
1774 * evaluation for matches: a match is finally adopted only if there is
1775 * no better match at the next window position.
1776 *
1777 * Processes a new input file and return its compressed length. Sets
1778 * the compressed length, crc, deflate flags and internal file
1779 * attributes.
1780 */
1781
1782/* Flush the current block, with given end-of-file flag.
1783 * IN assertion: strstart is set to the end of the current match. */
1784#define FLUSH_BLOCK(eof) \
1785 flush_block( \
1786 G1.block_start >= 0L \
1787 ? (char*)&G1.window[(unsigned)G1.block_start] \
1788 : (char*)NULL, \
1789 (ulg)G1.strstart - G1.block_start, \
1790 (eof) \
1791 )
1792
1793/* Insert string s in the dictionary and set match_head to the previous head
1794 * of the hash chain (the most recent string with same hash key). Return
1795 * the previous length of the hash chain.
1796 * IN assertion: all calls to INSERT_STRING are made with consecutive
1797 * input characters and the first MIN_MATCH bytes of s are valid
1798 * (except for the last MIN_MATCH-1 bytes of the input file). */
1799#define INSERT_STRING(s, match_head) \
1800do { \
1801 UPDATE_HASH(G1.ins_h, G1.window[(s) + MIN_MATCH-1]); \
1802 G1.prev[(s) & WMASK] = match_head = head[G1.ins_h]; \
1803 head[G1.ins_h] = (s); \
1804} while (0)
1805
1806static ulg deflate(void)
1807{
1808 IPos hash_head; /* head of hash chain */
1809 IPos prev_match; /* previous match */
1810 int flush; /* set if current block must be flushed */
1811 int match_available = 0; /* set if previous match exists */
1812 unsigned match_length = MIN_MATCH - 1; /* length of best match */
1813
1814 /* Process the input block. */
1815 while (G1.lookahead != 0) {
1816 /* Insert the string window[strstart .. strstart+2] in the
1817 * dictionary, and set hash_head to the head of the hash chain:
1818 */
1819 INSERT_STRING(G1.strstart, hash_head);
1820
1821 /* Find the longest match, discarding those <= prev_length.
1822 */
1823 G1.prev_length = match_length;
1824 prev_match = G1.match_start;
1825 match_length = MIN_MATCH - 1;
1826
1827 if (hash_head != 0 && G1.prev_length < max_lazy_match
1828 && G1.strstart - hash_head <= MAX_DIST
1829 ) {
1830 /* To simplify the code, we prevent matches with the string
1831 * of window index 0 (in particular we have to avoid a match
1832 * of the string with itself at the start of the input file).
1833 */
1834 match_length = longest_match(hash_head);
1835 /* longest_match() sets match_start */
1836 if (match_length > G1.lookahead)
1837 match_length = G1.lookahead;
1838
1839 /* Ignore a length 3 match if it is too distant: */
1840 if (match_length == MIN_MATCH && G1.strstart - G1.match_start > TOO_FAR) {
1841 /* If prev_match is also MIN_MATCH, G1.match_start is garbage
1842 * but we will ignore the current match anyway.
1843 */
1844 match_length--;
1845 }
1846 }
1847 /* If there was a match at the previous step and the current
1848 * match is not better, output the previous match:
1849 */
1850 if (G1.prev_length >= MIN_MATCH && match_length <= G1.prev_length) {
1851 check_match(G1.strstart - 1, prev_match, G1.prev_length);
1852 flush = ct_tally(G1.strstart - 1 - prev_match, G1.prev_length - MIN_MATCH);
1853
1854 /* Insert in hash table all strings up to the end of the match.
1855 * strstart-1 and strstart are already inserted.
1856 */
1857 G1.lookahead -= G1.prev_length - 1;
1858 G1.prev_length -= 2;
1859 do {
1860 G1.strstart++;
1861 INSERT_STRING(G1.strstart, hash_head);
1862 /* strstart never exceeds WSIZE-MAX_MATCH, so there are
1863 * always MIN_MATCH bytes ahead. If lookahead < MIN_MATCH
1864 * these bytes are garbage, but it does not matter since the
1865 * next lookahead bytes will always be emitted as literals.
1866 */
1867 } while (--G1.prev_length != 0);
1868 match_available = 0;
1869 match_length = MIN_MATCH - 1;
1870 G1.strstart++;
1871 if (flush) {
1872 FLUSH_BLOCK(0);
1873 G1.block_start = G1.strstart;
1874 }
1875 } else if (match_available) {
1876 /* If there was no match at the previous position, output a
1877 * single literal. If there was a match but the current match
1878 * is longer, truncate the previous match to a single literal.
1879 */
1880 Tracevv((stderr, "%c", G1.window[G1.strstart - 1]));
1881 if (ct_tally(0, G1.window[G1.strstart - 1])) {
1882 FLUSH_BLOCK(0);
1883 G1.block_start = G1.strstart;
1884 }
1885 G1.strstart++;
1886 G1.lookahead--;
1887 } else {
1888 /* There is no previous match to compare with, wait for
1889 * the next step to decide.
1890 */
1891 match_available = 1;
1892 G1.strstart++;
1893 G1.lookahead--;
1894 }
1895 Assert(G1.strstart <= G1.isize && lookahead <= G1.isize, "a bit too far");
1896
1897 /* Make sure that we always have enough lookahead, except
1898 * at the end of the input file. We need MAX_MATCH bytes
1899 * for the next match, plus MIN_MATCH bytes to insert the
1900 * string following the next match.
1901 */
1902 while (G1.lookahead < MIN_LOOKAHEAD && !G1.eofile)
1903 fill_window();
1904 }
1905 if (match_available)
1906 ct_tally(0, G1.window[G1.strstart - 1]);
1907
1908 return FLUSH_BLOCK(1); /* eof */
1909}
1910
1911
1912/* ===========================================================================
1913 * Initialize the bit string routines.
1914 */
1915static void bi_init(void)
1916{
1917 G1.bi_buf = 0;
1918 G1.bi_valid = 0;
1919#ifdef DEBUG
1920 G1.bits_sent = 0L;
1921#endif
1922}
1923
1924
1925/* ===========================================================================
1926 * Initialize the "longest match" routines for a new file
1927 */
1928static void lm_init(ush * flagsp)
1929{
1930 unsigned j;
1931
1932 /* Initialize the hash table. */
1933 memset(head, 0, HASH_SIZE * sizeof(*head));
1934 /* prev will be initialized on the fly */
1935
1936 /* speed options for the general purpose bit flag */
1937 *flagsp |= 2; /* FAST 4, SLOW 2 */
1938 /* ??? reduce max_chain_length for binary files */
1939
1940 G1.strstart = 0;
1941 G1.block_start = 0L;
1942
1943 G1.lookahead = file_read(G1.window,
1944 sizeof(int) <= 2 ? (unsigned) WSIZE : 2 * WSIZE);
1945
1946 if (G1.lookahead == 0 || G1.lookahead == (unsigned) -1) {
1947 G1.eofile = 1;
1948 G1.lookahead = 0;
1949 return;
1950 }
1951 G1.eofile = 0;
1952 /* Make sure that we always have enough lookahead. This is important
1953 * if input comes from a device such as a tty.
1954 */
1955 while (G1.lookahead < MIN_LOOKAHEAD && !G1.eofile)
1956 fill_window();
1957
1958 G1.ins_h = 0;
1959 for (j = 0; j < MIN_MATCH - 1; j++)
1960 UPDATE_HASH(G1.ins_h, G1.window[j]);
1961 /* If lookahead < MIN_MATCH, ins_h is garbage, but this is
1962 * not important since only literal bytes will be emitted.
1963 */
1964}
1965
1966
1967/* ===========================================================================
1968 * Allocate the match buffer, initialize the various tables and save the
1969 * location of the internal file attribute (ascii/binary) and method
1970 * (DEFLATE/STORE).
1971 * One callsite in zip()
1972 */
1973static void ct_init(void)
1974{
1975 int n; /* iterates over tree elements */
1976 int length; /* length value */
1977 int code; /* code value */
1978 int dist; /* distance index */
1979
1980 G2.compressed_len = 0L;
1981
1982#ifdef NOT_NEEDED
1983 if (G2.static_dtree[0].Len != 0)
1984 return; /* ct_init already called */
1985#endif
1986
1987 /* Initialize the mapping length (0..255) -> length code (0..28) */
1988 length = 0;
1989 for (code = 0; code < LENGTH_CODES - 1; code++) {
1990 G2.base_length[code] = length;
1991 for (n = 0; n < (1 << extra_lbits[code]); n++) {
1992 G2.length_code[length++] = code;
1993 }
1994 }
1995 Assert(length == 256, "ct_init: length != 256");
1996 /* Note that the length 255 (match length 258) can be represented
1997 * in two different ways: code 284 + 5 bits or code 285, so we
1998 * overwrite length_code[255] to use the best encoding:
1999 */
2000 G2.length_code[length - 1] = code;
2001
2002 /* Initialize the mapping dist (0..32K) -> dist code (0..29) */
2003 dist = 0;
2004 for (code = 0; code < 16; code++) {
2005 G2.base_dist[code] = dist;
2006 for (n = 0; n < (1 << extra_dbits[code]); n++) {
2007 G2.dist_code[dist++] = code;
2008 }
2009 }
2010 Assert(dist == 256, "ct_init: dist != 256");
2011 dist >>= 7; /* from now on, all distances are divided by 128 */
2012 for (; code < D_CODES; code++) {
2013 G2.base_dist[code] = dist << 7;
2014 for (n = 0; n < (1 << (extra_dbits[code] - 7)); n++) {
2015 G2.dist_code[256 + dist++] = code;
2016 }
2017 }
2018 Assert(dist == 256, "ct_init: 256+dist != 512");
2019
2020 /* Construct the codes of the static literal tree */
2021 /* already zeroed - it's in bss
2022 for (n = 0; n <= MAX_BITS; n++)
2023 G2.bl_count[n] = 0; */
2024
2025 n = 0;
2026 while (n <= 143) {
2027 G2.static_ltree[n++].Len = 8;
2028 G2.bl_count[8]++;
2029 }
2030 while (n <= 255) {
2031 G2.static_ltree[n++].Len = 9;
2032 G2.bl_count[9]++;
2033 }
2034 while (n <= 279) {
2035 G2.static_ltree[n++].Len = 7;
2036 G2.bl_count[7]++;
2037 }
2038 while (n <= 287) {
2039 G2.static_ltree[n++].Len = 8;
2040 G2.bl_count[8]++;
2041 }
2042 /* Codes 286 and 287 do not exist, but we must include them in the
2043 * tree construction to get a canonical Huffman tree (longest code
2044 * all ones)
2045 */
2046 gen_codes((ct_data *) G2.static_ltree, L_CODES + 1);
2047
2048 /* The static distance tree is trivial: */
2049 for (n = 0; n < D_CODES; n++) {
2050 G2.static_dtree[n].Len = 5;
2051 G2.static_dtree[n].Code = bi_reverse(n, 5);
2052 }
2053
2054 /* Initialize the first block of the first file: */
2055 init_block();
2056}
2057
2058
2059/* ===========================================================================
2060 * Deflate in to out.
2061 * IN assertions: the input and output buffers are cleared.
2062 */
2063
2064static void zip(void)
2065{
2066 ush deflate_flags = 0; /* pkzip -es, -en or -ex equivalent */
2067
2068 G1.outcnt = 0;
2069
2070 /* Write the header to the gzip file. See algorithm.doc for the format */
2071 /* magic header for gzip files: 1F 8B */
2072 /* compression method: 8 (DEFLATED) */
2073 /* general flags: 0 */
2074 put_32bit(0x00088b1f);
2075 put_32bit(0); /* Unix timestamp */
2076
2077 /* Write deflated file to zip file */
2078 G1.crc = ~0;
2079
2080 bi_init();
2081 ct_init();
2082 lm_init(&deflate_flags);
2083
2084 put_8bit(deflate_flags); /* extra flags */
2085 put_8bit(3); /* OS identifier = 3 (Unix) */
2086
2087 deflate();
2088
2089 /* Write the crc and uncompressed size */
2090 put_32bit(~G1.crc);
2091 put_32bit(G1.isize);
2092
2093 flush_outbuf();
2094}
2095
2096
2097/* ======================================================================== */
2098static
2099IF_DESKTOP(long long) int FAST_FUNC pack_gzip(transformer_state_t *xstate UNUSED_PARAM)
2100{
2101 /* Clear input and output buffers */
2102 G1.outcnt = 0;
2103#ifdef DEBUG
2104 G1.insize = 0;
2105#endif
2106 G1.isize = 0;
2107
2108 /* Reinit G2.xxx */
2109 memset(&G2, 0, sizeof(G2));
2110 G2.l_desc.dyn_tree = G2.dyn_ltree;
2111 G2.l_desc.static_tree = G2.static_ltree;
2112 G2.l_desc.extra_bits = extra_lbits;
2113 G2.l_desc.extra_base = LITERALS + 1;
2114 G2.l_desc.elems = L_CODES;
2115 G2.l_desc.max_length = MAX_BITS;
2116 //G2.l_desc.max_code = 0;
2117 G2.d_desc.dyn_tree = G2.dyn_dtree;
2118 G2.d_desc.static_tree = G2.static_dtree;
2119 G2.d_desc.extra_bits = extra_dbits;
2120 //G2.d_desc.extra_base = 0;
2121 G2.d_desc.elems = D_CODES;
2122 G2.d_desc.max_length = MAX_BITS;
2123 //G2.d_desc.max_code = 0;
2124 G2.bl_desc.dyn_tree = G2.bl_tree;
2125 //G2.bl_desc.static_tree = NULL;
2126 G2.bl_desc.extra_bits = extra_blbits,
2127 //G2.bl_desc.extra_base = 0;
2128 G2.bl_desc.elems = BL_CODES;
2129 G2.bl_desc.max_length = MAX_BL_BITS;
2130 //G2.bl_desc.max_code = 0;
2131
2132#if 0
2133 /* Saving of timestamp is disabled. Why?
2134 * - it is not Y2038-safe.
2135 * - some people want deterministic results
2136 * (normally they'd use -n, but our -n is a nop).
2137 * - it's bloat.
2138 * Per RFC 1952, gzfile.time=0 is "no timestamp".
2139 * If users will demand this to be reinstated,
2140 * implement -n "don't save timestamp".
2141 */
2142 struct stat s;
2143 s.st_ctime = 0;
2144 fstat(STDIN_FILENO, &s);
2145 zip(s.st_ctime);
2146#else
2147 zip();
2148#endif
2149 return 0;
2150}
2151
2152#if ENABLE_FEATURE_GZIP_LONG_OPTIONS
2153static const char gzip_longopts[] ALIGN1 =
2154 "stdout\0" No_argument "c"
2155 "to-stdout\0" No_argument "c"
2156 "force\0" No_argument "f"
2157 "verbose\0" No_argument "v"
2158#if ENABLE_GUNZIP
2159 "decompress\0" No_argument "d"
2160 "uncompress\0" No_argument "d"
2161 "test\0" No_argument "t"
2162#endif
2163 "quiet\0" No_argument "q"
2164 "fast\0" No_argument "1"
2165 "best\0" No_argument "9"
2166 "no-name\0" No_argument "n"
2167 ;
2168#endif
2169
2170/*
2171 * Linux kernel build uses gzip -d -n. We accept and ignore -n.
2172 * Man page says:
2173 * -n --no-name
2174 * gzip: do not save the original file name and time stamp.
2175 * (The original name is always saved if the name had to be truncated.)
2176 * gunzip: do not restore the original file name/time even if present
2177 * (remove only the gzip suffix from the compressed file name).
2178 * This option is the default when decompressing.
2179 * -N --name
2180 * gzip: always save the original file name and time stamp (this is the default)
2181 * gunzip: restore the original file name and time stamp if present.
2182 */
2183
2184int gzip_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
2185#if ENABLE_GUNZIP
2186int gzip_main(int argc, char **argv)
2187#else
2188int gzip_main(int argc UNUSED_PARAM, char **argv)
2189#endif
2190{
2191 unsigned opt;
2192#ifdef ENABLE_FEATURE_GZIP_LEVELS
2193 static const struct {
2194 uint8_t good;
2195 uint8_t chain_shift;
2196 uint8_t lazy2;
2197 uint8_t nice2;
2198 } gzip_level_config[6] = {
2199 {4, 4, 4/2, 16/2}, /* Level 4 */
2200 {8, 5, 16/2, 32/2}, /* Level 5 */
2201 {8, 7, 16/2, 128/2}, /* Level 6 */
2202 {8, 8, 32/2, 128/2}, /* Level 7 */
2203 {32, 10, 128/2, 258/2}, /* Level 8 */
2204 {32, 12, 258/2, 258/2}, /* Level 9 */
2205 };
2206#endif
2207
2208 SET_PTR_TO_GLOBALS((char *)xzalloc(sizeof(struct globals)+sizeof(struct globals2))
2209 + sizeof(struct globals));
2210
2211#if ENABLE_FEATURE_GZIP_LONG_OPTIONS
2212 applet_long_options = gzip_longopts;
2213#endif
2214 /* Must match bbunzip's constants OPT_STDOUT, OPT_FORCE! */
2215 opt = getopt32(argv, "cfv" IF_GUNZIP("dt") "qn123456789");
2216#if ENABLE_GUNZIP /* gunzip_main may not be visible... */
2217 if (opt & 0x18) // -d and/or -t
2218 return gunzip_main(argc, argv);
2219#endif
2220#ifdef ENABLE_FEATURE_GZIP_LEVELS
2221 opt >>= ENABLE_GUNZIP ? 7 : 5; /* drop cfv[dt]qn bits */
2222 if (opt == 0)
2223 opt = 1 << 6; /* default: 6 */
2224 opt = ffs(opt >> 4); /* Maps -1..-4 to [0], -5 to [1] ... -9 to [5] */
2225 max_chain_length = 1 << gzip_level_config[opt].chain_shift;
2226 good_match = gzip_level_config[opt].good;
2227 max_lazy_match = gzip_level_config[opt].lazy2 * 2;
2228 nice_match = gzip_level_config[opt].nice2 * 2;
2229#endif
2230 option_mask32 &= 0x7; /* retain only -cfv */
2231
2232 /* Allocate all global buffers (for DYN_ALLOC option) */
2233 ALLOC(uch, G1.l_buf, INBUFSIZ);
2234 ALLOC(uch, G1.outbuf, OUTBUFSIZ);
2235 ALLOC(ush, G1.d_buf, DIST_BUFSIZE);
2236 ALLOC(uch, G1.window, 2L * WSIZE);
2237 ALLOC(ush, G1.prev, 1L << BITS);
2238
2239 /* Initialize the CRC32 table */
2240 global_crc32_table = crc32_filltable(NULL, 0);
2241
2242 argv += optind;
2243 return bbunpack(argv, pack_gzip, append_ext, "gz");
2244}
2245