1 /* vi: set sw=4 ts=4: */
3 * Gzip implementation for busybox
5 * Based on GNU gzip Copyright (C) 1992-1993 Jean-loup Gailly.
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."
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.
16 * This program is free software; you can redistribute it and/or modify
17 * it under the terms of the GNU General Public License as published by
18 * the Free Software Foundation; either version 2 of the License, or
19 * (at your option) any later version.
21 * This program is distributed in the hope that it will be useful,
22 * but WITHOUT ANY WARRANTY; without even the implied warranty of
23 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
24 * General Public License for more details.
26 * You should have received a copy of the GNU General Public License
27 * along with this program; if not, write to the Free Software
28 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
32 /* These defines are very important for BusyBox. Without these,
33 * huge chunks of ram are pre-allocated making the BusyBox bss
34 * size Freaking Huge(tm), which is a bad thing.*/
43 #include <sys/types.h>
47 #include <sys/types.h>
54 typedef unsigned char uch;
55 typedef unsigned short ush;
56 typedef unsigned long ulg;
58 /* Return codes from gzip */
63 /* Compression methods (see algorithm.doc) */
64 /* Only STORED and DEFLATED are supported by this BusyBox module */
66 /* methods 4 to 7 reserved */
69 /* To save memory for 16 bit systems, some arrays are overlaid between
70 * the various modules:
71 * deflate: prev+head window d_buf l_buf outbuf
72 * unlzw: tab_prefix tab_suffix stack inbuf outbuf
73 * For compression, input is done in window[]. For decompression, output
74 * is done in window except for unlzw.
79 # define INBUFSIZ 0x2000 /* input buffer size */
81 # define INBUFSIZ 0x8000 /* input buffer size */
84 #define INBUF_EXTRA 64 /* required by unlzw() */
88 # define OUTBUFSIZ 8192 /* output buffer size */
90 # define OUTBUFSIZ 16384 /* output buffer size */
93 #define OUTBUF_EXTRA 2048 /* required by unlzw() */
97 # define DIST_BUFSIZE 0x2000 /* buffer for distances, see trees.c */
99 # define DIST_BUFSIZE 0x8000 /* buffer for distances, see trees.c */
104 # define DECLARE(type, array, size) static type * array
105 # define ALLOC(type, array, size) { \
106 array = (type*)xcalloc((size_t)(((size)+1L)/2), 2*sizeof(type)); \
108 # define FREE(array) {free(array), array=NULL;}
110 # define DECLARE(type, array, size) static type array[size]
111 # define ALLOC(type, array, size)
115 #define tab_suffix window
116 #define tab_prefix prev /* hash link (see deflate.c) */
117 #define head (prev+WSIZE) /* hash head (see deflate.c) */
119 static long bytes_in; /* number of input bytes */
121 #define isize bytes_in
122 /* for compatibility with old zip sources (to be cleaned) */
124 typedef int file_t; /* Do not use stdio */
126 #define NO_FILE (-1) /* in memory compression */
129 #define PACK_MAGIC "\037\036" /* Magic header for packed files */
130 #define GZIP_MAGIC "\037\213" /* Magic header for gzip files, 1F 8B */
131 #define OLD_GZIP_MAGIC "\037\236" /* Magic header for gzip 0.5 = freeze 1.x */
132 #define LZH_MAGIC "\037\240" /* Magic header for SCO LZH Compress files */
133 #define PKZIP_MAGIC "\120\113\003\004" /* Magic header for pkzip files */
136 #define ASCII_FLAG 0x01 /* bit 0 set: file probably ascii text */
137 #define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */
138 #define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
139 #define ORIG_NAME 0x08 /* bit 3 set: original file name present */
140 #define COMMENT 0x10 /* bit 4 set: file comment present */
141 #define RESERVED 0xC0 /* bit 6,7: reserved */
143 /* internal file attribute */
144 #define UNKNOWN 0xffff
149 # define WSIZE 0x8000 /* window size--must be a power of two, and */
150 #endif /* at least 32K for zip's deflate method */
153 #define MAX_MATCH 258
154 /* The minimum and maximum match lengths */
156 #define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
157 /* Minimum amount of lookahead, except at the end of the input file.
158 * See deflate.c for comments about the MIN_MATCH+1.
161 #define MAX_DIST (WSIZE-MIN_LOOKAHEAD)
162 /* In order to simplify the code, particularly on 16 bit machines, match
163 * distances are limited to MAX_DIST instead of WSIZE.
166 /* put_byte is used for the compressed output */
167 #define put_byte(c) {outbuf[outcnt++]=(uch)(c); if (outcnt==OUTBUFSIZ)\
171 /* Output a 32 bit value to the bit stream, lsb first */
173 #define put_long(n) { \
174 put_short((n) & 0xffff); \
175 put_short(((ulg)(n)) >> 16); \
179 #define seekable() 0 /* force sequential output */
180 #define translate_eol 0 /* no option -a yet */
182 /* Diagnostic functions */
184 # define Assert(cond,msg) {if(!(cond)) bb_error_msg(msg);}
185 # define Trace(x) fprintf x
186 # define Tracev(x) {if (verbose) fprintf x ;}
187 # define Tracevv(x) {if (verbose>1) fprintf x ;}
188 # define Tracec(c,x) {if (verbose && (c)) fprintf x ;}
189 # define Tracecv(c,x) {if (verbose>1 && (c)) fprintf x ;}
191 # define Assert(cond,msg)
196 # define Tracecv(c,x)
199 #define WARN(msg) {if (!quiet) fprintf msg ; \
200 if (exit_code == OK) exit_code = WARNING;}
203 # define MAX_PATH_LEN 1024 /* max pathname length */
208 static int zip(int in, int out);
209 static int file_read(char *buf, unsigned size);
212 static void lm_init(ush * flags);
213 static ulg deflate(void);
216 static void ct_init(ush * attr, int *methodp);
217 static int ct_tally(int dist, int lc);
218 static ulg flush_block(char *buf, ulg stored_len, int eof);
221 static void bi_init(file_t zipfile);
222 static void send_bits(int value, int length);
223 static unsigned bi_reverse(unsigned value, int length);
224 static void bi_windup(void);
225 static void copy_block(char *buf, unsigned len, int header);
226 static int (*read_buf) (char *buf, unsigned size);
229 static void flush_outbuf(void);
231 /* lzw.h -- define the lzw functions.
232 * Copyright (C) 1992-1993 Jean-loup Gailly.
233 * This is free software; you can redistribute it and/or modify it under the
234 * terms of the GNU General Public License, see the file COPYING.
237 #if !defined(OF) && defined(lint)
244 #define INIT_BITS 9 /* Initial number of bits per code */
246 #define BIT_MASK 0x1f /* Mask for 'number of compression bits' */
247 /* Mask 0x20 is reserved to mean a fourth header byte, and 0x40 is free.
248 * It's a pity that old uncompress does not check bit 0x20. That makes
249 * extension of the format actually undesirable because old compress
250 * would just crash on the new format instead of giving a meaningful
251 * error message. It does check the number of bits, but it's more
252 * helpful to say "unsupported format, get a new version" than
253 * "can only handle 16 bits".
256 /* tailor.h -- target dependent definitions
257 * Copyright (C) 1992-1993 Jean-loup Gailly.
258 * This is free software; you can redistribute it and/or modify it under the
259 * terms of the GNU General Public License, see the file COPYING.
262 /* The target dependent definitions should be defined here only.
263 * The target dependent functions should be defined in tailor.c.
267 /* Common defaults */
270 # define OS_CODE 0x03 /* assume Unix */
274 # define PATH_SEP '/'
278 # define OPTIONS_VAR "GZIP"
282 # define Z_SUFFIX ".gz"
286 # define MAX_SUFFIX MAX_EXT_CHARS
288 # define MAX_SUFFIX 30
293 DECLARE(uch, inbuf, INBUFSIZ + INBUF_EXTRA);
294 DECLARE(uch, outbuf, OUTBUFSIZ + OUTBUF_EXTRA);
295 DECLARE(ush, d_buf, DIST_BUFSIZE);
296 DECLARE(uch, window, 2L * WSIZE);
297 DECLARE(ush, tab_prefix, 1L << BITS);
299 static int foreground; /* set if program run in foreground */
300 static int method = DEFLATED; /* compression method */
301 static int exit_code = OK; /* program exit code */
302 static int part_nb; /* number of parts in .gz file */
303 static long time_stamp; /* original time stamp (modification time) */
304 static long ifile_size; /* input file size, -1 for devices (debug only) */
305 static char z_suffix[MAX_SUFFIX + 1]; /* default suffix (can be set with --suffix) */
306 static int z_len; /* strlen(z_suffix) */
308 static int ifd; /* input file descriptor */
309 static int ofd; /* output file descriptor */
310 static unsigned insize; /* valid bytes in inbuf */
311 static unsigned outcnt; /* bytes in output buffer */
314 /* Output a 16 bit value, lsb first */
315 static void put_short(ush w)
317 if (outcnt < OUTBUFSIZ - 2) {
318 outbuf[outcnt++] = (uch) ((w) & 0xff);
319 outbuf[outcnt++] = (uch) ((ush) (w) >> 8);
321 put_byte((uch) ((w) & 0xff));
322 put_byte((uch) ((ush) (w) >> 8));
326 /* ========================================================================
327 * Signal and error handler.
329 static void abort_gzip(int ignored)
334 /* ===========================================================================
335 * Clear input and output buffers
337 static void clear_bufs(void)
344 /* ===========================================================================
345 * Does the same as write(), but also handles partial pipe writes and checks
348 static void write_buf(int fd, void *buf, unsigned cnt)
352 while ((n = write(fd, buf, cnt)) != cnt) {
353 if (n == (unsigned) (-1)) bb_error_msg_and_die("can't write");
355 buf = (void *) ((char *) buf + n);
359 /* ===========================================================================
360 * Run a set of bytes through the crc shift register. If s is a NULL
361 * pointer, then initialize the crc shift register contents instead.
362 * Return the current crc in either case.
364 static ulg updcrc(uch * s, unsigned n)
366 static ulg crc = (ulg) 0xffffffffL; /* shift register contents */
367 register ulg c; /* temporary variable */
368 static unsigned long crc_32_tab[256];
370 if (crc_32_tab[1] == 0x00000000L) {
371 unsigned long csr; /* crc shift register */
372 const unsigned long e = 0xedb88320L; /* polynomial exclusive-or pattern */
373 int i; /* counter for all possible eight bit values */
374 int k; /* byte being shifted into crc apparatus */
376 /* Compute table of CRC's. */
377 for (i = 1; i < 256; i++) {
379 /* The idea to initialize the register with the byte instead of
380 * zero was stolen from Haruhiko Okumura's ar002
383 csr = csr & 1 ? (csr >> 1) ^ e : csr >> 1;
394 c = crc_32_tab[((int) c ^ (*s++)) & 0xff] ^ (c >> 8);
398 return c ^ 0xffffffffL; /* (instead of ~c for 64-bit machines) */
401 /* bits.c -- output variable-length bit strings
402 * Copyright (C) 1992-1993 Jean-loup Gailly
403 * This is free software; you can redistribute it and/or modify it under the
404 * terms of the GNU General Public License, see the file COPYING.
411 * Output variable-length bit strings. Compression can be done
412 * to a file or to memory. (The latter is not supported in this version.)
416 * The PKZIP "deflate" file format interprets compressed file data
417 * as a sequence of bits. Multi-bit strings in the file may cross
418 * byte boundaries without restriction.
420 * The first bit of each byte is the low-order bit.
422 * The routines in this file allow a variable-length bit value to
423 * be output right-to-left (useful for literal values). For
424 * left-to-right output (useful for code strings from the tree routines),
425 * the bits must have been reversed first with bi_reverse().
427 * For in-memory compression, the compressed bit stream goes directly
428 * into the requested output buffer. The input data is read in blocks
429 * by the mem_read() function. The buffer is limited to 64K on 16 bit
434 * void bi_init (FILE *zipfile)
435 * Initialize the bit string routines.
437 * void send_bits (int value, int length)
438 * Write out a bit string, taking the source bits right to
441 * int bi_reverse (int value, int length)
442 * Reverse the bits of a bit string, taking the source bits left to
443 * right and emitting them right to left.
445 * void bi_windup (void)
446 * Write out any remaining bits in an incomplete byte.
448 * void copy_block(char *buf, unsigned len, int header)
449 * Copy a stored block to the zip file, storing first the length and
450 * its one's complement if requested.
454 /* ===========================================================================
455 * Local data used by the "bit string" routines.
458 static file_t zfile; /* output gzip file */
460 static unsigned short bi_buf;
462 /* Output buffer. bits are inserted starting at the bottom (least significant
466 #define Buf_size (8 * 2*sizeof(char))
467 /* Number of bits used within bi_buf. (bi_buf might be implemented on
468 * more than 16 bits on some systems.)
473 /* Current input function. Set to mem_read for in-memory compression */
476 ulg bits_sent; /* bit length of the compressed data */
479 /* ===========================================================================
480 * Initialize the bit string routines.
482 static void bi_init(file_t zipfile)
491 /* Set the defaults for file compression. They are set by memcompress
492 * for in-memory compression.
494 if (zfile != NO_FILE) {
495 read_buf = file_read;
499 /* ===========================================================================
500 * Send a value on a given number of bits.
501 * IN assertion: length <= 16 and value fits in length bits.
503 static void send_bits(int value, int length)
506 Tracev((stderr, " l %2d v %4x ", length, value));
507 Assert(length > 0 && length <= 15, "invalid length");
508 bits_sent += (ulg) length;
510 /* If not enough room in bi_buf, use (valid) bits from bi_buf and
511 * (16 - bi_valid) bits from value, leaving (width - (16-bi_valid))
512 * unused bits in value.
514 if (bi_valid > (int) Buf_size - length) {
515 bi_buf |= (value << bi_valid);
517 bi_buf = (ush) value >> (Buf_size - bi_valid);
518 bi_valid += length - Buf_size;
520 bi_buf |= value << bi_valid;
525 /* ===========================================================================
526 * Reverse the first len bits of a code, using straightforward code (a faster
527 * method would use a table)
528 * IN assertion: 1 <= len <= 15
530 static unsigned bi_reverse(unsigned code, int len)
532 register unsigned res = 0;
536 code >>= 1, res <<= 1;
541 /* ===========================================================================
542 * Write out any remaining bits in an incomplete byte.
544 static void bi_windup()
548 } else if (bi_valid > 0) {
554 bits_sent = (bits_sent + 7) & ~7;
558 /* ===========================================================================
559 * Copy a stored block to the zip file, storing first the length and its
560 * one's complement if requested.
562 static void copy_block(char *buf, unsigned len, int header)
564 bi_windup(); /* align on byte boundary */
567 put_short((ush) len);
568 put_short((ush) ~ len);
574 bits_sent += (ulg) len << 3;
581 /* deflate.c -- compress data using the deflation algorithm
582 * Copyright (C) 1992-1993 Jean-loup Gailly
583 * This is free software; you can redistribute it and/or modify it under the
584 * terms of the GNU General Public License, see the file COPYING.
590 * Identify new text as repetitions of old text within a fixed-
591 * length sliding window trailing behind the new text.
595 * The "deflation" process depends on being able to identify portions
596 * of the input text which are identical to earlier input (within a
597 * sliding window trailing behind the input currently being processed).
599 * The most straightforward technique turns out to be the fastest for
600 * most input files: try all possible matches and select the longest.
601 * The key feature of this algorithm is that insertions into the string
602 * dictionary are very simple and thus fast, and deletions are avoided
603 * completely. Insertions are performed at each input character, whereas
604 * string matches are performed only when the previous match ends. So it
605 * is preferable to spend more time in matches to allow very fast string
606 * insertions and avoid deletions. The matching algorithm for small
607 * strings is inspired from that of Rabin & Karp. A brute force approach
608 * is used to find longer strings when a small match has been found.
609 * A similar algorithm is used in comic (by Jan-Mark Wams) and freeze
610 * (by Leonid Broukhis).
611 * A previous version of this file used a more sophisticated algorithm
612 * (by Fiala and Greene) which is guaranteed to run in linear amortized
613 * time, but has a larger average cost, uses more memory and is patented.
614 * However the F&G algorithm may be faster for some highly redundant
615 * files if the parameter max_chain_length (described below) is too large.
619 * The idea of lazy evaluation of matches is due to Jan-Mark Wams, and
620 * I found it in 'freeze' written by Leonid Broukhis.
621 * Thanks to many info-zippers for bug reports and testing.
625 * APPNOTE.TXT documentation file in PKZIP 1.93a distribution.
627 * A description of the Rabin and Karp algorithm is given in the book
628 * "Algorithms" by R. Sedgewick, Addison-Wesley, p252.
630 * Fiala,E.R., and Greene,D.H.
631 * Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595
635 * void lm_init (int pack_level, ush *flags)
636 * Initialize the "longest match" routines for a new file
639 * Processes a new input file and return its compressed length. Sets
640 * the compressed length, crc, deflate flags and internal file
645 /* ===========================================================================
646 * Configuration parameters
649 /* Compile with MEDIUM_MEM to reduce the memory requirements or
650 * with SMALL_MEM to use as little memory as possible. Use BIG_MEM if the
651 * entire input file can be held in memory (not possible on 16 bit systems).
652 * Warning: defining these symbols affects HASH_BITS (see below) and thus
653 * affects the compression ratio. The compressed output
654 * is still correct, and might even be smaller in some cases.
658 # define HASH_BITS 13 /* Number of bits used to hash strings */
661 # define HASH_BITS 14
664 # define HASH_BITS 15
665 /* For portability to 16 bit machines, do not use values above 15. */
668 /* To save space (see unlzw.c), we overlay prev+head with tab_prefix and
669 * window with tab_suffix. Check that we can do this:
671 #if (WSIZE<<1) > (1<<BITS)
672 # error cannot overlay window with tab_suffix and prev with tab_prefix0
674 #if HASH_BITS > BITS-1
675 # error cannot overlay head with tab_prefix1
677 #define HASH_SIZE (unsigned)(1<<HASH_BITS)
678 #define HASH_MASK (HASH_SIZE-1)
679 #define WMASK (WSIZE-1)
680 /* HASH_SIZE and WSIZE must be powers of two */
682 /* Tail of hash chains */
685 /* speed options for the general purpose bit flag */
687 # define TOO_FAR 4096
689 /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
690 /* ===========================================================================
691 * Local data used by the "longest match" routines.
694 typedef unsigned IPos;
696 /* A Pos is an index in the character window. We use short instead of int to
697 * save space in the various tables. IPos is used only for parameter passing.
700 /* DECLARE(uch, window, 2L*WSIZE); */
701 /* Sliding window. Input bytes are read into the second half of the window,
702 * and move to the first half later to keep a dictionary of at least WSIZE
703 * bytes. With this organization, matches are limited to a distance of
704 * WSIZE-MAX_MATCH bytes, but this ensures that IO is always
705 * performed with a length multiple of the block size. Also, it limits
706 * the window size to 64K, which is quite useful on MSDOS.
707 * To do: limit the window size to WSIZE+BSZ if SMALL_MEM (the code would
708 * be less efficient).
711 /* DECLARE(Pos, prev, WSIZE); */
712 /* Link to older string with same hash index. To limit the size of this
713 * array to 64K, this link is maintained only for the last 32K strings.
714 * An index in this array is thus a window index modulo 32K.
717 /* DECLARE(Pos, head, 1<<HASH_BITS); */
718 /* Heads of the hash chains or NIL. */
720 static const ulg window_size = (ulg) 2 * WSIZE;
722 /* window size, 2*WSIZE except for MMAP or BIG_MEM, where it is the
723 * input file length plus MIN_LOOKAHEAD.
726 static long block_start;
728 /* window position at the beginning of the current output block. Gets
729 * negative when the window is moved backwards.
732 static unsigned ins_h; /* hash index of string to be inserted */
734 #define H_SHIFT ((HASH_BITS+MIN_MATCH-1)/MIN_MATCH)
735 /* Number of bits by which ins_h and del_h must be shifted at each
736 * input step. It must be such that after MIN_MATCH steps, the oldest
737 * byte no longer takes part in the hash key, that is:
738 * H_SHIFT * MIN_MATCH >= HASH_BITS
741 static unsigned int prev_length;
743 /* Length of the best match at previous step. Matches not greater than this
744 * are discarded. This is used in the lazy match evaluation.
747 static unsigned strstart; /* start of string to insert */
748 static unsigned match_start; /* start of matching string */
749 static int eofile; /* flag set at end of input file */
750 static unsigned lookahead; /* number of valid bytes ahead in window */
752 static const unsigned max_chain_length = 4096;
754 /* To speed up deflation, hash chains are never searched beyond this length.
755 * A higher limit improves compression ratio but degrades the speed.
758 static const unsigned int max_lazy_match = 258;
760 /* Attempt to find a better match only when the current match is strictly
761 * smaller than this value. This mechanism is used only for compression
764 #define max_insert_length max_lazy_match
765 /* Insert new strings in the hash table only if the match length
766 * is not greater than this length. This saves time but degrades compression.
767 * max_insert_length is used only for compression levels <= 3.
770 static const unsigned good_match = 32;
772 /* Use a faster search when the previous match is longer than this */
775 /* Values for max_lazy_match, good_match and max_chain_length, depending on
776 * the desired pack level (0..9). The values given below have been tuned to
777 * exclude worst case performance for pathological files. Better values may be
778 * found for specific files.
781 static const int nice_match = 258; /* Stop searching when current match exceeds this */
783 /* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
784 * For deflate_fast() (levels <= 3) good is ignored and lazy has a different
789 /* result of memcmp for equal strings */
791 /* ===========================================================================
792 * Prototypes for local functions.
794 static void fill_window(void);
796 static int longest_match(IPos cur_match);
799 static void check_match(IPos start, IPos match, int length);
802 /* ===========================================================================
803 * Update a hash value with the given input byte
804 * IN assertion: all calls to to UPDATE_HASH are made with consecutive
805 * input characters, so that a running hash key can be computed from the
806 * previous key instead of complete recalculation each time.
808 #define UPDATE_HASH(h,c) (h = (((h)<<H_SHIFT) ^ (c)) & HASH_MASK)
810 /* ===========================================================================
811 * Insert string s in the dictionary and set match_head to the previous head
812 * of the hash chain (the most recent string with same hash key). Return
813 * the previous length of the hash chain.
814 * IN assertion: all calls to to INSERT_STRING are made with consecutive
815 * input characters and the first MIN_MATCH bytes of s are valid
816 * (except for the last MIN_MATCH-1 bytes of the input file).
818 #define INSERT_STRING(s, match_head) \
819 (UPDATE_HASH(ins_h, window[(s) + MIN_MATCH-1]), \
820 prev[(s) & WMASK] = match_head = head[ins_h], \
823 /* ===========================================================================
824 * Initialize the "longest match" routines for a new file
826 static void lm_init(ush * flags)
830 /* Initialize the hash table. */
831 memset(head, 0, HASH_SIZE * sizeof(*head));
832 /* prev will be initialized on the fly */
835 /* ??? reduce max_chain_length for binary files */
840 lookahead = read_buf((char *) window,
841 sizeof(int) <= 2 ? (unsigned) WSIZE : 2 * WSIZE);
843 if (lookahead == 0 || lookahead == (unsigned) EOF) {
844 eofile = 1, lookahead = 0;
848 /* Make sure that we always have enough lookahead. This is important
849 * if input comes from a device such as a tty.
851 while (lookahead < MIN_LOOKAHEAD && !eofile)
855 for (j = 0; j < MIN_MATCH - 1; j++)
856 UPDATE_HASH(ins_h, window[j]);
857 /* If lookahead < MIN_MATCH, ins_h is garbage, but this is
858 * not important since only literal bytes will be emitted.
862 /* ===========================================================================
863 * Set match_start to the longest match starting at the given string and
864 * return its length. Matches shorter or equal to prev_length are discarded,
865 * in which case the result is equal to prev_length and match_start is
867 * IN assertions: cur_match is the head of the hash chain for the current
868 * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
871 /* For MSDOS, OS/2 and 386 Unix, an optimized version is in match.asm or
872 * match.s. The code is functionally equivalent, so you can use the C version
875 static int longest_match(IPos cur_match)
877 unsigned chain_length = max_chain_length; /* max hash chain length */
878 register uch *scan = window + strstart; /* current string */
879 register uch *match; /* matched string */
880 register int len; /* length of current match */
881 int best_len = prev_length; /* best match length so far */
883 strstart > (IPos) MAX_DIST ? strstart - (IPos) MAX_DIST : NIL;
884 /* Stop when cur_match becomes <= limit. To simplify the code,
885 * we prevent matches with the string of window index 0.
888 /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
889 * It is easy to get rid of this optimization if necessary.
891 #if HASH_BITS < 8 || MAX_MATCH != 258
892 # error Code too clever
894 register uch *strend = window + strstart + MAX_MATCH;
895 register uch scan_end1 = scan[best_len - 1];
896 register uch scan_end = scan[best_len];
898 /* Do not waste too much time if we already have a good match: */
899 if (prev_length >= good_match) {
902 Assert(strstart <= window_size - MIN_LOOKAHEAD, "insufficient lookahead");
905 Assert(cur_match < strstart, "no future");
906 match = window + cur_match;
908 /* Skip to next match if the match length cannot increase
909 * or if the match length is less than 2:
911 if (match[best_len] != scan_end ||
912 match[best_len - 1] != scan_end1 ||
913 *match != *scan || *++match != scan[1])
916 /* The check at best_len-1 can be removed because it will be made
917 * again later. (This heuristic is not always a win.)
918 * It is not necessary to compare scan[2] and match[2] since they
919 * are always equal when the other bytes match, given that
920 * the hash keys are equal and that HASH_BITS >= 8.
924 /* We check for insufficient lookahead only every 8th comparison;
925 * the 256th check will be made at strstart+258.
928 } while (*++scan == *++match && *++scan == *++match &&
929 *++scan == *++match && *++scan == *++match &&
930 *++scan == *++match && *++scan == *++match &&
931 *++scan == *++match && *++scan == *++match && scan < strend);
933 len = MAX_MATCH - (int) (strend - scan);
934 scan = strend - MAX_MATCH;
936 if (len > best_len) {
937 match_start = cur_match;
939 if (len >= nice_match)
941 scan_end1 = scan[best_len - 1];
942 scan_end = scan[best_len];
944 } while ((cur_match = prev[cur_match & WMASK]) > limit
945 && --chain_length != 0);
951 /* ===========================================================================
952 * Check that the match at match_start is indeed a match.
954 static void check_match(IPos start, IPos match, int length)
956 /* check that the match is indeed a match */
957 if (memcmp((char *) window + match,
958 (char *) window + start, length) != EQUAL) {
959 bb_error_msg(" start %d, match %d, length %d", start, match, length);
960 bb_error_msg("invalid match");
963 bb_error_msg("\\[%d,%d]", start - match, length);
965 putc(window[start++], stderr);
966 } while (--length != 0);
970 # define check_match(start, match, length)
973 /* ===========================================================================
974 * Fill the window when the lookahead becomes insufficient.
975 * Updates strstart and lookahead, and sets eofile if end of input file.
976 * IN assertion: lookahead < MIN_LOOKAHEAD && strstart + lookahead > 0
977 * OUT assertions: at least one byte has been read, or eofile is set;
978 * file reads are performed for at least two bytes (required for the
979 * translate_eol option).
981 static void fill_window()
983 register unsigned n, m;
985 (unsigned) (window_size - (ulg) lookahead - (ulg) strstart);
986 /* Amount of free space at the end of the window. */
988 /* If the window is almost full and there is insufficient lookahead,
989 * move the upper half to the lower one to make room in the upper half.
991 if (more == (unsigned) EOF) {
992 /* Very unlikely, but possible on 16 bit machine if strstart == 0
993 * and lookahead == 1 (input done one byte at time)
996 } else if (strstart >= WSIZE + MAX_DIST) {
997 /* By the IN assertion, the window is not empty so we can't confuse
998 * more == 0 with more == 64K on a 16 bit machine.
1000 Assert(window_size == (ulg) 2 * WSIZE, "no sliding with BIG_MEM");
1002 memcpy((char *) window, (char *) window + WSIZE, (unsigned) WSIZE);
1003 match_start -= WSIZE;
1004 strstart -= WSIZE; /* we now have strstart >= MAX_DIST: */
1006 block_start -= (long) WSIZE;
1008 for (n = 0; n < HASH_SIZE; n++) {
1010 head[n] = (Pos) (m >= WSIZE ? m - WSIZE : NIL);
1012 for (n = 0; n < WSIZE; n++) {
1014 prev[n] = (Pos) (m >= WSIZE ? m - WSIZE : NIL);
1015 /* If n is not on any hash chain, prev[n] is garbage but
1016 * its value will never be used.
1021 /* At this point, more >= 2 */
1023 n = read_buf((char *) window + strstart + lookahead, more);
1024 if (n == 0 || n == (unsigned) EOF) {
1032 /* ===========================================================================
1033 * Flush the current block, with given end-of-file flag.
1034 * IN assertion: strstart is set to the end of the current match.
1036 #define FLUSH_BLOCK(eof) \
1037 flush_block(block_start >= 0L ? (char*)&window[(unsigned)block_start] : \
1038 (char*)NULL, (long)strstart - block_start, (eof))
1040 /* ===========================================================================
1041 * Same as above, but achieves better compression. We use a lazy
1042 * evaluation for matches: a match is finally adopted only if there is
1043 * no better match at the next window position.
1045 static ulg deflate()
1047 IPos hash_head; /* head of hash chain */
1048 IPos prev_match; /* previous match */
1049 int flush; /* set if current block must be flushed */
1050 int match_available = 0; /* set if previous match exists */
1051 register unsigned match_length = MIN_MATCH - 1; /* length of best match */
1053 /* Process the input block. */
1054 while (lookahead != 0) {
1055 /* Insert the string window[strstart .. strstart+2] in the
1056 * dictionary, and set hash_head to the head of the hash chain:
1058 INSERT_STRING(strstart, hash_head);
1060 /* Find the longest match, discarding those <= prev_length.
1062 prev_length = match_length, prev_match = match_start;
1063 match_length = MIN_MATCH - 1;
1065 if (hash_head != NIL && prev_length < max_lazy_match &&
1066 strstart - hash_head <= MAX_DIST) {
1067 /* To simplify the code, we prevent matches with the string
1068 * of window index 0 (in particular we have to avoid a match
1069 * of the string with itself at the start of the input file).
1071 match_length = longest_match(hash_head);
1072 /* longest_match() sets match_start */
1073 if (match_length > lookahead)
1074 match_length = lookahead;
1076 /* Ignore a length 3 match if it is too distant: */
1077 if (match_length == MIN_MATCH && strstart - match_start > TOO_FAR) {
1078 /* If prev_match is also MIN_MATCH, match_start is garbage
1079 * but we will ignore the current match anyway.
1084 /* If there was a match at the previous step and the current
1085 * match is not better, output the previous match:
1087 if (prev_length >= MIN_MATCH && match_length <= prev_length) {
1089 check_match(strstart - 1, prev_match, prev_length);
1092 ct_tally(strstart - 1 - prev_match, prev_length - MIN_MATCH);
1094 /* Insert in hash table all strings up to the end of the match.
1095 * strstart-1 and strstart are already inserted.
1097 lookahead -= prev_length - 1;
1101 INSERT_STRING(strstart, hash_head);
1102 /* strstart never exceeds WSIZE-MAX_MATCH, so there are
1103 * always MIN_MATCH bytes ahead. If lookahead < MIN_MATCH
1104 * these bytes are garbage, but it does not matter since the
1105 * next lookahead bytes will always be emitted as literals.
1107 } while (--prev_length != 0);
1108 match_available = 0;
1109 match_length = MIN_MATCH - 1;
1112 FLUSH_BLOCK(0), block_start = strstart;
1114 } else if (match_available) {
1115 /* If there was no match at the previous position, output a
1116 * single literal. If there was a match but the current match
1117 * is longer, truncate the previous match to a single literal.
1119 Tracevv((stderr, "%c", window[strstart - 1]));
1120 if (ct_tally(0, window[strstart - 1])) {
1121 FLUSH_BLOCK(0), block_start = strstart;
1126 /* There is no previous match to compare with, wait for
1127 * the next step to decide.
1129 match_available = 1;
1133 Assert(strstart <= isize && lookahead <= isize, "a bit too far");
1135 /* Make sure that we always have enough lookahead, except
1136 * at the end of the input file. We need MAX_MATCH bytes
1137 * for the next match, plus MIN_MATCH bytes to insert the
1138 * string following the next match.
1140 while (lookahead < MIN_LOOKAHEAD && !eofile)
1143 if (match_available)
1144 ct_tally(0, window[strstart - 1]);
1146 return FLUSH_BLOCK(1); /* eof */
1149 /* gzip (GNU zip) -- compress files with zip algorithm and 'compress' interface
1150 * Copyright (C) 1992-1993 Jean-loup Gailly
1151 * The unzip code was written and put in the public domain by Mark Adler.
1152 * Portions of the lzw code are derived from the public domain 'compress'
1153 * written by Spencer Thomas, Joe Orost, James Woods, Jim McKie, Steve Davies,
1154 * Ken Turkowski, Dave Mack and Peter Jannesen.
1156 * See the license_msg below and the file COPYING for the software license.
1157 * See the file algorithm.doc for the compression algorithms and file formats.
1160 /* Compress files with zip algorithm and 'compress' interface.
1161 * See usage() and help() functions below for all options.
1163 * file.gz: compressed file with same mode, owner, and utimes
1164 * or stdout with -c option or if stdin used as input.
1165 * If the output file name had to be truncated, the original name is kept
1166 * in the compressed file.
1171 typedef struct dirent dir_type;
1173 /* ======================================================================== */
1174 int gzip_main(int argc, char **argv)
1179 struct stat statBuf;
1185 while ((opt = getopt(argc, argv, "cf123456789dq")) != -1) {
1193 /* Ignore 1-9 (compression level) options */
1206 #ifdef CONFIG_GUNZIP
1209 return gunzip_main(argc, argv);
1216 foreground = signal(SIGINT, SIG_IGN) != SIG_IGN;
1218 (void) signal(SIGINT, abort_gzip);
1221 if (signal(SIGTERM, SIG_IGN) != SIG_IGN) {
1222 (void) signal(SIGTERM, abort_gzip);
1226 if (signal(SIGHUP, SIG_IGN) != SIG_IGN) {
1227 (void) signal(SIGHUP, abort_gzip);
1231 strncpy(z_suffix, Z_SUFFIX, sizeof(z_suffix) - 1);
1232 z_len = strlen(z_suffix);
1234 /* Allocate all global buffers (for DYN_ALLOC option) */
1235 ALLOC(uch, inbuf, INBUFSIZ + INBUF_EXTRA);
1236 ALLOC(uch, outbuf, OUTBUFSIZ + OUTBUF_EXTRA);
1237 ALLOC(ush, d_buf, DIST_BUFSIZE);
1238 ALLOC(uch, window, 2L * WSIZE);
1239 ALLOC(ush, tab_prefix, 1L << BITS);
1244 if (optind == argc) {
1247 zip(STDIN_FILENO, STDOUT_FILENO);
1251 for (i = optind; i < argc; i++) {
1254 if (strcmp(argv[i], "-") == 0) {
1257 inFileNum = STDIN_FILENO;
1258 outFileNum = STDOUT_FILENO;
1260 inFileNum = open(argv[i], O_RDONLY);
1261 if (inFileNum < 0 || fstat(inFileNum, &statBuf) < 0)
1262 bb_perror_msg_and_die("%s", argv[i]);
1263 time_stamp = statBuf.st_ctime;
1264 ifile_size = statBuf.st_size;
1267 path = xmalloc(strlen(argv[i]) + 4);
1268 strcpy(path, argv[i]);
1269 strcat(path, ".gz");
1271 /* Open output file */
1272 #if (__GLIBC__ >= 2) && (__GLIBC_MINOR__ >= 1)
1274 open(path, O_RDWR | O_CREAT | O_EXCL | O_NOFOLLOW);
1276 outFileNum = open(path, O_RDWR | O_CREAT | O_EXCL);
1278 if (outFileNum < 0) {
1279 bb_perror_msg("%s", path);
1284 /* Set permissions on the file */
1285 fchmod(outFileNum, statBuf.st_mode);
1287 outFileNum = STDOUT_FILENO;
1290 if (path == NULL && isatty(outFileNum) && force == 0) {
1292 ("compressed data not written to a terminal. Use -f to force compression.");
1297 result = zip(inFileNum, outFileNum);
1303 /* Delete the original file */
1305 delFileName = argv[i];
1309 if (unlink(delFileName) < 0)
1310 bb_perror_msg("%s", delFileName);
1320 /* trees.c -- output deflated data using Huffman coding
1321 * Copyright (C) 1992-1993 Jean-loup Gailly
1322 * This is free software; you can redistribute it and/or modify it under the
1323 * terms of the GNU General Public License, see the file COPYING.
1329 * Encode various sets of source values using variable-length
1330 * binary code trees.
1334 * The PKZIP "deflation" process uses several Huffman trees. The more
1335 * common source values are represented by shorter bit sequences.
1337 * Each code tree is stored in the ZIP file in a compressed form
1338 * which is itself a Huffman encoding of the lengths of
1339 * all the code strings (in ascending order by source values).
1340 * The actual code strings are reconstructed from the lengths in
1341 * the UNZIP process, as described in the "application note"
1342 * (APPNOTE.TXT) distributed as part of PKWARE's PKZIP program.
1347 * Data Compression: Techniques and Applications, pp. 53-55.
1348 * Lifetime Learning Publications, 1985. ISBN 0-534-03418-7.
1351 * Data Compression: Methods and Theory, pp. 49-50.
1352 * Computer Science Press, 1988. ISBN 0-7167-8156-5.
1356 * Addison-Wesley, 1983. ISBN 0-201-06672-6.
1360 * void ct_init (ush *attr, int *methodp)
1361 * Allocate the match buffer, initialize the various tables and save
1362 * the location of the internal file attribute (ascii/binary) and
1363 * method (DEFLATE/STORE)
1365 * void ct_tally (int dist, int lc);
1366 * Save the match info and tally the frequency counts.
1368 * long flush_block (char *buf, ulg stored_len, int eof)
1369 * Determine the best encoding for the current block: dynamic trees,
1370 * static trees or store, and output the encoded block to the zip
1371 * file. Returns the total compressed length for the file so far.
1375 /* ===========================================================================
1380 /* All codes must not exceed MAX_BITS bits */
1382 #define MAX_BL_BITS 7
1383 /* Bit length codes must not exceed MAX_BL_BITS bits */
1385 #define LENGTH_CODES 29
1386 /* number of length codes, not counting the special END_BLOCK code */
1388 #define LITERALS 256
1389 /* number of literal bytes 0..255 */
1391 #define END_BLOCK 256
1392 /* end of block literal code */
1394 #define L_CODES (LITERALS+1+LENGTH_CODES)
1395 /* number of Literal or Length codes, including the END_BLOCK code */
1398 /* number of distance codes */
1401 /* number of codes used to transfer the bit lengths */
1403 typedef uch extra_bits_t;
1405 /* extra bits for each length code */
1406 static const extra_bits_t extra_lbits[LENGTH_CODES]
1407 = { 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4,
1411 /* extra bits for each distance code */
1412 static const extra_bits_t extra_dbits[D_CODES]
1413 = { 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9,
1414 10, 10, 11, 11, 12, 12, 13, 13
1417 /* extra bits for each bit length code */
1418 static const extra_bits_t extra_blbits[BL_CODES]
1419 = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 3, 7 };
1421 #define STORED_BLOCK 0
1422 #define STATIC_TREES 1
1424 /* The three kinds of block type */
1428 # define LIT_BUFSIZE 0x2000
1431 # define LIT_BUFSIZE 0x4000
1433 # define LIT_BUFSIZE 0x8000
1437 #ifndef DIST_BUFSIZE
1438 # define DIST_BUFSIZE LIT_BUFSIZE
1440 /* Sizes of match buffers for literals/lengths and distances. There are
1441 * 4 reasons for limiting LIT_BUFSIZE to 64K:
1442 * - frequencies can be kept in 16 bit counters
1443 * - if compression is not successful for the first block, all input data is
1444 * still in the window so we can still emit a stored block even when input
1445 * comes from standard input. (This can also be done for all blocks if
1446 * LIT_BUFSIZE is not greater than 32K.)
1447 * - if compression is not successful for a file smaller than 64K, we can
1448 * even emit a stored file instead of a stored block (saving 5 bytes).
1449 * - creating new Huffman trees less frequently may not provide fast
1450 * adaptation to changes in the input data statistics. (Take for
1451 * example a binary file with poorly compressible code followed by
1452 * a highly compressible string table.) Smaller buffer sizes give
1453 * fast adaptation but have of course the overhead of transmitting trees
1455 * - I can't count above 4
1456 * The current code is general and allows DIST_BUFSIZE < LIT_BUFSIZE (to save
1457 * memory at the expense of compression). Some optimizations would be possible
1458 * if we rely on DIST_BUFSIZE == LIT_BUFSIZE.
1460 #if LIT_BUFSIZE > INBUFSIZ
1461 #error cannot overlay l_buf and inbuf
1464 /* repeat previous bit length 3-6 times (2 bits of repeat count) */
1465 #define REPZ_3_10 17
1466 /* repeat a zero length 3-10 times (3 bits of repeat count) */
1467 #define REPZ_11_138 18
1468 /* repeat a zero length 11-138 times (7 bits of repeat count) */
1470 /* ===========================================================================
1474 /* Data structure describing a single value and its code string. */
1475 typedef struct ct_data {
1477 ush freq; /* frequency count */
1478 ush code; /* bit string */
1481 ush dad; /* father node in Huffman tree */
1482 ush len; /* length of bit string */
1486 #define Freq fc.freq
1487 #define Code fc.code
1491 #define HEAP_SIZE (2*L_CODES+1)
1492 /* maximum heap size */
1494 static ct_data dyn_ltree[HEAP_SIZE]; /* literal and length tree */
1495 static ct_data dyn_dtree[2 * D_CODES + 1]; /* distance tree */
1497 static ct_data static_ltree[L_CODES + 2];
1499 /* The static literal tree. Since the bit lengths are imposed, there is no
1500 * need for the L_CODES extra codes used during heap construction. However
1501 * The codes 286 and 287 are needed to build a canonical tree (see ct_init
1505 static ct_data static_dtree[D_CODES];
1507 /* The static distance tree. (Actually a trivial tree since all codes use
1511 static ct_data bl_tree[2 * BL_CODES + 1];
1513 /* Huffman tree for the bit lengths */
1515 typedef struct tree_desc {
1516 ct_data *dyn_tree; /* the dynamic tree */
1517 ct_data *static_tree; /* corresponding static tree or NULL */
1518 const extra_bits_t *extra_bits; /* extra bits for each code or NULL */
1519 int extra_base; /* base index for extra_bits */
1520 int elems; /* max number of elements in the tree */
1521 int max_length; /* max bit length for the codes */
1522 int max_code; /* largest code with non zero frequency */
1525 static tree_desc l_desc =
1526 { dyn_ltree, static_ltree, extra_lbits, LITERALS + 1, L_CODES,
1530 static tree_desc d_desc =
1531 { dyn_dtree, static_dtree, extra_dbits, 0, D_CODES, MAX_BITS, 0 };
1533 static tree_desc bl_desc =
1534 { bl_tree, (ct_data *) 0, extra_blbits, 0, BL_CODES, MAX_BL_BITS,
1539 static ush bl_count[MAX_BITS + 1];
1541 /* number of codes at each bit length for an optimal tree */
1543 static const uch bl_order[BL_CODES]
1544 = { 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15 };
1546 /* The lengths of the bit length codes are sent in order of decreasing
1547 * probability, to avoid transmitting the lengths for unused bit length codes.
1550 static int heap[2 * L_CODES + 1]; /* heap used to build the Huffman trees */
1551 static int heap_len; /* number of elements in the heap */
1552 static int heap_max; /* element of largest frequency */
1554 /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used.
1555 * The same heap array is used to build all trees.
1558 static uch depth[2 * L_CODES + 1];
1560 /* Depth of each subtree used as tie breaker for trees of equal frequency */
1562 static uch length_code[MAX_MATCH - MIN_MATCH + 1];
1564 /* length code for each normalized match length (0 == MIN_MATCH) */
1566 static uch dist_code[512];
1568 /* distance codes. The first 256 values correspond to the distances
1569 * 3 .. 258, the last 256 values correspond to the top 8 bits of
1570 * the 15 bit distances.
1573 static int base_length[LENGTH_CODES];
1575 /* First normalized length for each code (0 = MIN_MATCH) */
1577 static int base_dist[D_CODES];
1579 /* First normalized distance for each code (0 = distance of 1) */
1582 /* DECLARE(uch, l_buf, LIT_BUFSIZE); buffer for literals or lengths */
1584 /* DECLARE(ush, d_buf, DIST_BUFSIZE); buffer for distances */
1586 static uch flag_buf[(LIT_BUFSIZE / 8)];
1588 /* flag_buf is a bit array distinguishing literals from lengths in
1589 * l_buf, thus indicating the presence or absence of a distance.
1592 static unsigned last_lit; /* running index in l_buf */
1593 static unsigned last_dist; /* running index in d_buf */
1594 static unsigned last_flags; /* running index in flag_buf */
1595 static uch flags; /* current flags not yet saved in flag_buf */
1596 static uch flag_bit; /* current bit used in flags */
1598 /* bits are filled in flags starting at bit 0 (least significant).
1599 * Note: these flags are overkill in the current code since we don't
1600 * take advantage of DIST_BUFSIZE == LIT_BUFSIZE.
1603 static ulg opt_len; /* bit length of current block with optimal trees */
1604 static ulg static_len; /* bit length of current block with static trees */
1606 static ulg compressed_len; /* total bit length of compressed file */
1609 static ush *file_type; /* pointer to UNKNOWN, BINARY or ASCII */
1610 static int *file_method; /* pointer to DEFLATE or STORE */
1612 /* ===========================================================================
1613 * Local (static) routines in this file.
1616 static void init_block(void);
1617 static void pqdownheap(ct_data * tree, int k);
1618 static void gen_bitlen(tree_desc * desc);
1619 static void gen_codes(ct_data * tree, int max_code);
1620 static void build_tree(tree_desc * desc);
1621 static void scan_tree(ct_data * tree, int max_code);
1622 static void send_tree(ct_data * tree, int max_code);
1623 static int build_bl_tree(void);
1624 static void send_all_trees(int lcodes, int dcodes, int blcodes);
1625 static void compress_block(ct_data * ltree, ct_data * dtree);
1626 static void set_file_type(void);
1630 # define send_code(c, tree) send_bits(tree[c].Code, tree[c].Len)
1631 /* Send a code of the given tree. c and tree must not have side effects */
1634 # define send_code(c, tree) \
1635 { if (verbose>1) bb_error_msg("\ncd %3d ",(c)); \
1636 send_bits(tree[c].Code, tree[c].Len); }
1639 #define d_code(dist) \
1640 ((dist) < 256 ? dist_code[dist] : dist_code[256+((dist)>>7)])
1641 /* Mapping from a distance to a distance code. dist is the distance - 1 and
1642 * must not have side effects. dist_code[256] and dist_code[257] are never
1646 /* the arguments must not have side effects */
1648 /* ===========================================================================
1649 * Allocate the match buffer, initialize the various tables and save the
1650 * location of the internal file attribute (ascii/binary) and method
1653 static void ct_init(ush * attr, int *methodp)
1655 int n; /* iterates over tree elements */
1656 int bits; /* bit counter */
1657 int length; /* length value */
1658 int code; /* code value */
1659 int dist; /* distance index */
1662 file_method = methodp;
1663 compressed_len = 0L;
1665 if (static_dtree[0].Len != 0)
1666 return; /* ct_init already called */
1668 /* Initialize the mapping length (0..255) -> length code (0..28) */
1670 for (code = 0; code < LENGTH_CODES - 1; code++) {
1671 base_length[code] = length;
1672 for (n = 0; n < (1 << extra_lbits[code]); n++) {
1673 length_code[length++] = (uch) code;
1676 Assert(length == 256, "ct_init: length != 256");
1677 /* Note that the length 255 (match length 258) can be represented
1678 * in two different ways: code 284 + 5 bits or code 285, so we
1679 * overwrite length_code[255] to use the best encoding:
1681 length_code[length - 1] = (uch) code;
1683 /* Initialize the mapping dist (0..32K) -> dist code (0..29) */
1685 for (code = 0; code < 16; code++) {
1686 base_dist[code] = dist;
1687 for (n = 0; n < (1 << extra_dbits[code]); n++) {
1688 dist_code[dist++] = (uch) code;
1691 Assert(dist == 256, "ct_init: dist != 256");
1692 dist >>= 7; /* from now on, all distances are divided by 128 */
1693 for (; code < D_CODES; code++) {
1694 base_dist[code] = dist << 7;
1695 for (n = 0; n < (1 << (extra_dbits[code] - 7)); n++) {
1696 dist_code[256 + dist++] = (uch) code;
1699 Assert(dist == 256, "ct_init: 256+dist != 512");
1701 /* Construct the codes of the static literal tree */
1702 for (bits = 0; bits <= MAX_BITS; bits++)
1706 static_ltree[n++].Len = 8, bl_count[8]++;
1708 static_ltree[n++].Len = 9, bl_count[9]++;
1710 static_ltree[n++].Len = 7, bl_count[7]++;
1712 static_ltree[n++].Len = 8, bl_count[8]++;
1713 /* Codes 286 and 287 do not exist, but we must include them in the
1714 * tree construction to get a canonical Huffman tree (longest code
1717 gen_codes((ct_data *) static_ltree, L_CODES + 1);
1719 /* The static distance tree is trivial: */
1720 for (n = 0; n < D_CODES; n++) {
1721 static_dtree[n].Len = 5;
1722 static_dtree[n].Code = bi_reverse(n, 5);
1725 /* Initialize the first block of the first file: */
1729 /* ===========================================================================
1730 * Initialize a new block.
1732 static void init_block()
1734 int n; /* iterates over tree elements */
1736 /* Initialize the trees. */
1737 for (n = 0; n < L_CODES; n++)
1738 dyn_ltree[n].Freq = 0;
1739 for (n = 0; n < D_CODES; n++)
1740 dyn_dtree[n].Freq = 0;
1741 for (n = 0; n < BL_CODES; n++)
1742 bl_tree[n].Freq = 0;
1744 dyn_ltree[END_BLOCK].Freq = 1;
1745 opt_len = static_len = 0L;
1746 last_lit = last_dist = last_flags = 0;
1752 /* Index within the heap array of least frequent node in the Huffman tree */
1755 /* ===========================================================================
1756 * Remove the smallest element from the heap and recreate the heap with
1757 * one less element. Updates heap and heap_len.
1759 #define pqremove(tree, top) \
1761 top = heap[SMALLEST]; \
1762 heap[SMALLEST] = heap[heap_len--]; \
1763 pqdownheap(tree, SMALLEST); \
1766 /* ===========================================================================
1767 * Compares to subtrees, using the tree depth as tie breaker when
1768 * the subtrees have equal frequency. This minimizes the worst case length.
1770 #define smaller(tree, n, m) \
1771 (tree[n].Freq < tree[m].Freq || \
1772 (tree[n].Freq == tree[m].Freq && depth[n] <= depth[m]))
1774 /* ===========================================================================
1775 * Restore the heap property by moving down the tree starting at node k,
1776 * exchanging a node with the smallest of its two sons if necessary, stopping
1777 * when the heap property is re-established (each father smaller than its
1780 static void pqdownheap(ct_data * tree, int k)
1783 int j = k << 1; /* left son of k */
1785 while (j <= heap_len) {
1786 /* Set j to the smallest of the two sons: */
1787 if (j < heap_len && smaller(tree, heap[j + 1], heap[j]))
1790 /* Exit if v is smaller than both sons */
1791 if (smaller(tree, v, heap[j]))
1794 /* Exchange v with the smallest son */
1798 /* And continue down the tree, setting j to the left son of k */
1804 /* ===========================================================================
1805 * Compute the optimal bit lengths for a tree and update the total bit length
1806 * for the current block.
1807 * IN assertion: the fields freq and dad are set, heap[heap_max] and
1808 * above are the tree nodes sorted by increasing frequency.
1809 * OUT assertions: the field len is set to the optimal bit length, the
1810 * array bl_count contains the frequencies for each bit length.
1811 * The length opt_len is updated; static_len is also updated if stree is
1814 static void gen_bitlen(tree_desc * desc)
1816 ct_data *tree = desc->dyn_tree;
1817 const extra_bits_t *extra = desc->extra_bits;
1818 int base = desc->extra_base;
1819 int max_code = desc->max_code;
1820 int max_length = desc->max_length;
1821 ct_data *stree = desc->static_tree;
1822 int h; /* heap index */
1823 int n, m; /* iterate over the tree elements */
1824 int bits; /* bit length */
1825 int xbits; /* extra bits */
1826 ush f; /* frequency */
1827 int overflow = 0; /* number of elements with bit length too large */
1829 for (bits = 0; bits <= MAX_BITS; bits++)
1832 /* In a first pass, compute the optimal bit lengths (which may
1833 * overflow in the case of the bit length tree).
1835 tree[heap[heap_max]].Len = 0; /* root of the heap */
1837 for (h = heap_max + 1; h < HEAP_SIZE; h++) {
1839 bits = tree[tree[n].Dad].Len + 1;
1840 if (bits > max_length)
1841 bits = max_length, overflow++;
1842 tree[n].Len = (ush) bits;
1843 /* We overwrite tree[n].Dad which is no longer needed */
1846 continue; /* not a leaf node */
1851 xbits = extra[n - base];
1853 opt_len += (ulg) f *(bits + xbits);
1856 static_len += (ulg) f *(stree[n].Len + xbits);
1861 Trace((stderr, "\nbit length overflow\n"));
1862 /* This happens for example on obj2 and pic of the Calgary corpus */
1864 /* Find the first bit length which could increase: */
1866 bits = max_length - 1;
1867 while (bl_count[bits] == 0)
1869 bl_count[bits]--; /* move one leaf down the tree */
1870 bl_count[bits + 1] += 2; /* move one overflow item as its brother */
1871 bl_count[max_length]--;
1872 /* The brother of the overflow item also moves one step up,
1873 * but this does not affect bl_count[max_length]
1876 } while (overflow > 0);
1878 /* Now recompute all bit lengths, scanning in increasing frequency.
1879 * h is still equal to HEAP_SIZE. (It is simpler to reconstruct all
1880 * lengths instead of fixing only the wrong ones. This idea is taken
1881 * from 'ar' written by Haruhiko Okumura.)
1883 for (bits = max_length; bits != 0; bits--) {
1889 if (tree[m].Len != (unsigned) bits) {
1890 Trace((stderr, "code %d bits %d->%d\n", m, tree[m].Len,
1893 ((long) bits - (long) tree[m].Len) * (long) tree[m].Freq;
1894 tree[m].Len = (ush) bits;
1901 /* ===========================================================================
1902 * Generate the codes for a given tree and bit counts (which need not be
1904 * IN assertion: the array bl_count contains the bit length statistics for
1905 * the given tree and the field len is set for all tree elements.
1906 * OUT assertion: the field code is set for all tree elements of non
1909 static void gen_codes(ct_data * tree, int max_code)
1911 ush next_code[MAX_BITS + 1]; /* next code value for each bit length */
1912 ush code = 0; /* running code value */
1913 int bits; /* bit index */
1914 int n; /* code index */
1916 /* The distribution counts are first used to generate the code values
1917 * without bit reversal.
1919 for (bits = 1; bits <= MAX_BITS; bits++) {
1920 next_code[bits] = code = (code + bl_count[bits - 1]) << 1;
1922 /* Check that the bit counts in bl_count are consistent. The last code
1925 Assert(code + bl_count[MAX_BITS] - 1 == (1 << MAX_BITS) - 1,
1926 "inconsistent bit counts");
1927 Tracev((stderr, "\ngen_codes: max_code %d ", max_code));
1929 for (n = 0; n <= max_code; n++) {
1930 int len = tree[n].Len;
1934 /* Now reverse the bits */
1935 tree[n].Code = bi_reverse(next_code[len]++, len);
1937 Tracec(tree != static_ltree,
1938 (stderr, "\nn %3d %c l %2d c %4x (%x) ", n,
1939 (isgraph(n) ? n : ' '), len, tree[n].Code,
1940 next_code[len] - 1));
1944 /* ===========================================================================
1945 * Construct one Huffman tree and assigns the code bit strings and lengths.
1946 * Update the total bit length for the current block.
1947 * IN assertion: the field freq is set for all tree elements.
1948 * OUT assertions: the fields len and code are set to the optimal bit length
1949 * and corresponding code. The length opt_len is updated; static_len is
1950 * also updated if stree is not null. The field max_code is set.
1952 static void build_tree(tree_desc * desc)
1954 ct_data *tree = desc->dyn_tree;
1955 ct_data *stree = desc->static_tree;
1956 int elems = desc->elems;
1957 int n, m; /* iterate over heap elements */
1958 int max_code = -1; /* largest code with non zero frequency */
1959 int node = elems; /* next internal node of the tree */
1961 /* Construct the initial heap, with least frequent element in
1962 * heap[SMALLEST]. The sons of heap[n] are heap[2*n] and heap[2*n+1].
1963 * heap[0] is not used.
1965 heap_len = 0, heap_max = HEAP_SIZE;
1967 for (n = 0; n < elems; n++) {
1968 if (tree[n].Freq != 0) {
1969 heap[++heap_len] = max_code = n;
1976 /* The pkzip format requires that at least one distance code exists,
1977 * and that at least one bit should be sent even if there is only one
1978 * possible code. So to avoid special checks later on we force at least
1979 * two codes of non zero frequency.
1981 while (heap_len < 2) {
1982 int new = heap[++heap_len] = (max_code < 2 ? ++max_code : 0);
1988 static_len -= stree[new].Len;
1989 /* new is 0 or 1 so it does not have extra bits */
1991 desc->max_code = max_code;
1993 /* The elements heap[heap_len/2+1 .. heap_len] are leaves of the tree,
1994 * establish sub-heaps of increasing lengths:
1996 for (n = heap_len / 2; n >= 1; n--)
1997 pqdownheap(tree, n);
1999 /* Construct the Huffman tree by repeatedly combining the least two
2003 pqremove(tree, n); /* n = node of least frequency */
2004 m = heap[SMALLEST]; /* m = node of next least frequency */
2006 heap[--heap_max] = n; /* keep the nodes sorted by frequency */
2007 heap[--heap_max] = m;
2009 /* Create a new node father of n and m */
2010 tree[node].Freq = tree[n].Freq + tree[m].Freq;
2011 depth[node] = (uch) (MAX(depth[n], depth[m]) + 1);
2012 tree[n].Dad = tree[m].Dad = (ush) node;
2014 if (tree == bl_tree) {
2015 bb_error_msg("\nnode %d(%d), sons %d(%d) %d(%d)",
2016 node, tree[node].Freq, n, tree[n].Freq, m, tree[m].Freq);
2019 /* and insert the new node in the heap */
2020 heap[SMALLEST] = node++;
2021 pqdownheap(tree, SMALLEST);
2023 } while (heap_len >= 2);
2025 heap[--heap_max] = heap[SMALLEST];
2027 /* At this point, the fields freq and dad are set. We can now
2028 * generate the bit lengths.
2030 gen_bitlen((tree_desc *) desc);
2032 /* The field len is now set, we can generate the bit codes */
2033 gen_codes((ct_data *) tree, max_code);
2036 /* ===========================================================================
2037 * Scan a literal or distance tree to determine the frequencies of the codes
2038 * in the bit length tree. Updates opt_len to take into account the repeat
2039 * counts. (The contribution of the bit length codes will be added later
2040 * during the construction of bl_tree.)
2042 static void scan_tree(ct_data * tree, int max_code)
2044 int n; /* iterates over all tree elements */
2045 int prevlen = -1; /* last emitted length */
2046 int curlen; /* length of current code */
2047 int nextlen = tree[0].Len; /* length of next code */
2048 int count = 0; /* repeat count of the current code */
2049 int max_count = 7; /* max repeat count */
2050 int min_count = 4; /* min repeat count */
2053 max_count = 138, min_count = 3;
2054 tree[max_code + 1].Len = (ush) 0xffff; /* guard */
2056 for (n = 0; n <= max_code; n++) {
2058 nextlen = tree[n + 1].Len;
2059 if (++count < max_count && curlen == nextlen) {
2061 } else if (count < min_count) {
2062 bl_tree[curlen].Freq += count;
2063 } else if (curlen != 0) {
2064 if (curlen != prevlen)
2065 bl_tree[curlen].Freq++;
2066 bl_tree[REP_3_6].Freq++;
2067 } else if (count <= 10) {
2068 bl_tree[REPZ_3_10].Freq++;
2070 bl_tree[REPZ_11_138].Freq++;
2075 max_count = 138, min_count = 3;
2076 } else if (curlen == nextlen) {
2077 max_count = 6, min_count = 3;
2079 max_count = 7, min_count = 4;
2084 /* ===========================================================================
2085 * Send a literal or distance tree in compressed form, using the codes in
2088 static void send_tree(ct_data * tree, int max_code)
2090 int n; /* iterates over all tree elements */
2091 int prevlen = -1; /* last emitted length */
2092 int curlen; /* length of current code */
2093 int nextlen = tree[0].Len; /* length of next code */
2094 int count = 0; /* repeat count of the current code */
2095 int max_count = 7; /* max repeat count */
2096 int min_count = 4; /* min repeat count */
2098 /* tree[max_code+1].Len = -1; *//* guard already set */
2100 max_count = 138, min_count = 3;
2102 for (n = 0; n <= max_code; n++) {
2104 nextlen = tree[n + 1].Len;
2105 if (++count < max_count && curlen == nextlen) {
2107 } else if (count < min_count) {
2109 send_code(curlen, bl_tree);
2110 } while (--count != 0);
2112 } else if (curlen != 0) {
2113 if (curlen != prevlen) {
2114 send_code(curlen, bl_tree);
2117 Assert(count >= 3 && count <= 6, " 3_6?");
2118 send_code(REP_3_6, bl_tree);
2119 send_bits(count - 3, 2);
2121 } else if (count <= 10) {
2122 send_code(REPZ_3_10, bl_tree);
2123 send_bits(count - 3, 3);
2126 send_code(REPZ_11_138, bl_tree);
2127 send_bits(count - 11, 7);
2132 max_count = 138, min_count = 3;
2133 } else if (curlen == nextlen) {
2134 max_count = 6, min_count = 3;
2136 max_count = 7, min_count = 4;
2141 /* ===========================================================================
2142 * Construct the Huffman tree for the bit lengths and return the index in
2143 * bl_order of the last bit length code to send.
2145 static int build_bl_tree()
2147 int max_blindex; /* index of last bit length code of non zero freq */
2149 /* Determine the bit length frequencies for literal and distance trees */
2150 scan_tree((ct_data *) dyn_ltree, l_desc.max_code);
2151 scan_tree((ct_data *) dyn_dtree, d_desc.max_code);
2153 /* Build the bit length tree: */
2154 build_tree((tree_desc *) (&bl_desc));
2155 /* opt_len now includes the length of the tree representations, except
2156 * the lengths of the bit lengths codes and the 5+5+4 bits for the counts.
2159 /* Determine the number of bit length codes to send. The pkzip format
2160 * requires that at least 4 bit length codes be sent. (appnote.txt says
2161 * 3 but the actual value used is 4.)
2163 for (max_blindex = BL_CODES - 1; max_blindex >= 3; max_blindex--) {
2164 if (bl_tree[bl_order[max_blindex]].Len != 0)
2167 /* Update opt_len to include the bit length tree and counts */
2168 opt_len += 3 * (max_blindex + 1) + 5 + 5 + 4;
2169 Tracev((stderr, "\ndyn trees: dyn %ld, stat %ld", opt_len, static_len));
2174 /* ===========================================================================
2175 * Send the header for a block using dynamic Huffman trees: the counts, the
2176 * lengths of the bit length codes, the literal tree and the distance tree.
2177 * IN assertion: lcodes >= 257, dcodes >= 1, blcodes >= 4.
2179 static void send_all_trees(int lcodes, int dcodes, int blcodes)
2181 int rank; /* index in bl_order */
2183 Assert(lcodes >= 257 && dcodes >= 1 && blcodes >= 4, "not enough codes");
2184 Assert(lcodes <= L_CODES && dcodes <= D_CODES
2185 && blcodes <= BL_CODES, "too many codes");
2186 Tracev((stderr, "\nbl counts: "));
2187 send_bits(lcodes - 257, 5); /* not +255 as stated in appnote.txt */
2188 send_bits(dcodes - 1, 5);
2189 send_bits(blcodes - 4, 4); /* not -3 as stated in appnote.txt */
2190 for (rank = 0; rank < blcodes; rank++) {
2191 Tracev((stderr, "\nbl code %2d ", bl_order[rank]));
2192 send_bits(bl_tree[bl_order[rank]].Len, 3);
2194 Tracev((stderr, "\nbl tree: sent %ld", bits_sent));
2196 send_tree((ct_data *) dyn_ltree, lcodes - 1); /* send the literal tree */
2197 Tracev((stderr, "\nlit tree: sent %ld", bits_sent));
2199 send_tree((ct_data *) dyn_dtree, dcodes - 1); /* send the distance tree */
2200 Tracev((stderr, "\ndist tree: sent %ld", bits_sent));
2203 /* ===========================================================================
2204 * Determine the best encoding for the current block: dynamic trees, static
2205 * trees or store, and output the encoded block to the zip file. This function
2206 * returns the total compressed length for the file so far.
2208 static ulg flush_block(char *buf, ulg stored_len, int eof)
2210 ulg opt_lenb, static_lenb; /* opt_len and static_len in bytes */
2211 int max_blindex; /* index of last bit length code of non zero freq */
2213 flag_buf[last_flags] = flags; /* Save the flags for the last 8 items */
2215 /* Check if the file is ascii or binary */
2216 if (*file_type == (ush) UNKNOWN)
2219 /* Construct the literal and distance trees */
2220 build_tree((tree_desc *) (&l_desc));
2221 Tracev((stderr, "\nlit data: dyn %ld, stat %ld", opt_len, static_len));
2223 build_tree((tree_desc *) (&d_desc));
2224 Tracev((stderr, "\ndist data: dyn %ld, stat %ld", opt_len, static_len));
2225 /* At this point, opt_len and static_len are the total bit lengths of
2226 * the compressed block data, excluding the tree representations.
2229 /* Build the bit length tree for the above two trees, and get the index
2230 * in bl_order of the last bit length code to send.
2232 max_blindex = build_bl_tree();
2234 /* Determine the best encoding. Compute first the block length in bytes */
2235 opt_lenb = (opt_len + 3 + 7) >> 3;
2236 static_lenb = (static_len + 3 + 7) >> 3;
2239 "\nopt %lu(%lu) stat %lu(%lu) stored %lu lit %u dist %u ",
2240 opt_lenb, opt_len, static_lenb, static_len, stored_len,
2241 last_lit, last_dist));
2243 if (static_lenb <= opt_lenb)
2244 opt_lenb = static_lenb;
2246 /* If compression failed and this is the first and last block,
2247 * and if the zip file can be seeked (to rewrite the local header),
2248 * the whole file is transformed into a stored file:
2250 if (stored_len <= opt_lenb && eof && compressed_len == 0L && seekable()) {
2251 /* Since LIT_BUFSIZE <= 2*WSIZE, the input data must be there: */
2252 if (buf == (char *) 0)
2253 bb_error_msg("block vanished");
2255 copy_block(buf, (unsigned) stored_len, 0); /* without header */
2256 compressed_len = stored_len << 3;
2257 *file_method = STORED;
2259 } else if (stored_len + 4 <= opt_lenb && buf != (char *) 0) {
2260 /* 4: two words for the lengths */
2261 /* The test buf != NULL is only necessary if LIT_BUFSIZE > WSIZE.
2262 * Otherwise we can't have processed more than WSIZE input bytes since
2263 * the last block flush, because compression would have been
2264 * successful. If LIT_BUFSIZE <= WSIZE, it is never too late to
2265 * transform a block into a stored block.
2267 send_bits((STORED_BLOCK << 1) + eof, 3); /* send block type */
2268 compressed_len = (compressed_len + 3 + 7) & ~7L;
2269 compressed_len += (stored_len + 4) << 3;
2271 copy_block(buf, (unsigned) stored_len, 1); /* with header */
2273 } else if (static_lenb == opt_lenb) {
2274 send_bits((STATIC_TREES << 1) + eof, 3);
2275 compress_block((ct_data *) static_ltree, (ct_data *) static_dtree);
2276 compressed_len += 3 + static_len;
2278 send_bits((DYN_TREES << 1) + eof, 3);
2279 send_all_trees(l_desc.max_code + 1, d_desc.max_code + 1,
2281 compress_block((ct_data *) dyn_ltree, (ct_data *) dyn_dtree);
2282 compressed_len += 3 + opt_len;
2284 Assert(compressed_len == bits_sent, "bad compressed size");
2289 compressed_len += 7; /* align on byte boundary */
2291 Tracev((stderr, "\ncomprlen %lu(%lu) ", compressed_len >> 3,
2292 compressed_len - 7 * eof));
2294 return compressed_len >> 3;
2297 /* ===========================================================================
2298 * Save the match info and tally the frequency counts. Return true if
2299 * the current block must be flushed.
2301 static int ct_tally(int dist, int lc)
2303 l_buf[last_lit++] = (uch) lc;
2305 /* lc is the unmatched char */
2306 dyn_ltree[lc].Freq++;
2308 /* Here, lc is the match length - MIN_MATCH */
2309 dist--; /* dist = match distance - 1 */
2310 Assert((ush) dist < (ush) MAX_DIST &&
2311 (ush) lc <= (ush) (MAX_MATCH - MIN_MATCH) &&
2312 (ush) d_code(dist) < (ush) D_CODES, "ct_tally: bad match");
2314 dyn_ltree[length_code[lc] + LITERALS + 1].Freq++;
2315 dyn_dtree[d_code(dist)].Freq++;
2317 d_buf[last_dist++] = (ush) dist;
2322 /* Output the flags if they fill a byte: */
2323 if ((last_lit & 7) == 0) {
2324 flag_buf[last_flags++] = flags;
2325 flags = 0, flag_bit = 1;
2327 /* Try to guess if it is profitable to stop the current block here */
2328 if ((last_lit & 0xfff) == 0) {
2329 /* Compute an upper bound for the compressed length */
2330 ulg out_length = (ulg) last_lit * 8L;
2331 ulg in_length = (ulg) strstart - block_start;
2334 for (dcode = 0; dcode < D_CODES; dcode++) {
2336 (ulg) dyn_dtree[dcode].Freq * (5L + extra_dbits[dcode]);
2340 "\nlast_lit %u, last_dist %u, in %ld, out ~%ld(%ld%%) ",
2341 last_lit, last_dist, in_length, out_length,
2342 100L - out_length * 100L / in_length));
2343 if (last_dist < last_lit / 2 && out_length < in_length / 2)
2346 return (last_lit == LIT_BUFSIZE - 1 || last_dist == DIST_BUFSIZE);
2347 /* We avoid equality with LIT_BUFSIZE because of wraparound at 64K
2348 * on 16 bit machines and because stored blocks are restricted to
2353 /* ===========================================================================
2354 * Send the block data compressed using the given Huffman trees
2356 static void compress_block(ct_data * ltree, ct_data * dtree)
2358 unsigned dist; /* distance of matched string */
2359 int lc; /* match length or unmatched char (if dist == 0) */
2360 unsigned lx = 0; /* running index in l_buf */
2361 unsigned dx = 0; /* running index in d_buf */
2362 unsigned fx = 0; /* running index in flag_buf */
2363 uch flag = 0; /* current flags */
2364 unsigned code; /* the code to send */
2365 int extra; /* number of extra bits to send */
2370 flag = flag_buf[fx++];
2372 if ((flag & 1) == 0) {
2373 send_code(lc, ltree); /* send a literal byte */
2374 Tracecv(isgraph(lc), (stderr, " '%c' ", lc));
2376 /* Here, lc is the match length - MIN_MATCH */
2377 code = length_code[lc];
2378 send_code(code + LITERALS + 1, ltree); /* send the length code */
2379 extra = extra_lbits[code];
2381 lc -= base_length[code];
2382 send_bits(lc, extra); /* send the extra length bits */
2385 /* Here, dist is the match distance - 1 */
2386 code = d_code(dist);
2387 Assert(code < D_CODES, "bad d_code");
2389 send_code(code, dtree); /* send the distance code */
2390 extra = extra_dbits[code];
2392 dist -= base_dist[code];
2393 send_bits(dist, extra); /* send the extra distance bits */
2395 } /* literal or match pair ? */
2397 } while (lx < last_lit);
2399 send_code(END_BLOCK, ltree);
2402 /* ===========================================================================
2403 * Set the file type to ASCII or BINARY, using a crude approximation:
2404 * binary if more than 20% of the bytes are <= 6 or >= 128, ascii otherwise.
2405 * IN assertion: the fields freq of dyn_ltree are set and the total of all
2406 * frequencies does not exceed 64K (to fit in an int on 16 bit machines).
2408 static void set_file_type()
2411 unsigned ascii_freq = 0;
2412 unsigned bin_freq = 0;
2415 bin_freq += dyn_ltree[n++].Freq;
2417 ascii_freq += dyn_ltree[n++].Freq;
2418 while (n < LITERALS)
2419 bin_freq += dyn_ltree[n++].Freq;
2420 *file_type = bin_freq > (ascii_freq >> 2) ? BINARY : ASCII;
2421 if (*file_type == BINARY && translate_eol) {
2422 bb_error_msg("-l used on binary file");
2426 /* zip.c -- compress files to the gzip or pkzip format
2427 * Copyright (C) 1992-1993 Jean-loup Gailly
2428 * This is free software; you can redistribute it and/or modify it under the
2429 * terms of the GNU General Public License, see the file COPYING.
2433 static ulg crc; /* crc on uncompressed file data */
2434 static long header_bytes; /* number of bytes in gzip header */
2436 static void put_long(ulg n)
2438 put_short((n) & 0xffff);
2439 put_short(((ulg) (n)) >> 16);
2442 /* put_header_byte is used for the compressed output
2443 * - for the initial 4 bytes that can't overflow the buffer.
2445 #define put_header_byte(c) {outbuf[outcnt++]=(uch)(c);}
2447 /* ===========================================================================
2448 * Deflate in to out.
2449 * IN assertions: the input and output buffers are cleared.
2450 * The variables time_stamp and save_orig_name are initialized.
2452 static int zip(int in, int out)
2454 uch my_flags = 0; /* general purpose bit flags */
2455 ush attr = 0; /* ascii/binary flag */
2456 ush deflate_flags = 0; /* pkzip -es, -en or -ex equivalent */
2462 /* Write the header to the gzip file. See algorithm.doc for the format */
2466 put_header_byte(GZIP_MAGIC[0]); /* magic header */
2467 put_header_byte(GZIP_MAGIC[1]);
2468 put_header_byte(DEFLATED); /* compression method */
2470 put_header_byte(my_flags); /* general flags */
2471 put_long(time_stamp);
2473 /* Write deflated file to zip file */
2477 ct_init(&attr, &method);
2478 lm_init(&deflate_flags);
2480 put_byte((uch) deflate_flags); /* extra flags */
2481 put_byte(OS_CODE); /* OS identifier */
2483 header_bytes = (long) outcnt;
2487 /* Write the crc and uncompressed size */
2490 header_bytes += 2 * sizeof(long);
2497 /* ===========================================================================
2498 * Read a new buffer from the current input file, perform end-of-line
2499 * translation, and update the crc and input file size.
2500 * IN assertion: size >= 2 (for end-of-line translation)
2502 static int file_read(char *buf, unsigned size)
2506 Assert(insize == 0, "inbuf not empty");
2508 len = read(ifd, buf, size);
2509 if (len == (unsigned) (-1) || len == 0)
2512 crc = updcrc((uch *) buf, len);
2517 /* ===========================================================================
2518 * Write the output buffer outbuf[0..outcnt-1] and update bytes_out.
2519 * (used for the compressed data only)
2521 static void flush_outbuf()
2526 write_buf(ofd, (char *) outbuf, outcnt);