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.
240 #define INIT_BITS 9 /* Initial number of bits per code */
242 #define BIT_MASK 0x1f /* Mask for 'number of compression bits' */
243 /* Mask 0x20 is reserved to mean a fourth header byte, and 0x40 is free.
244 * It's a pity that old uncompress does not check bit 0x20. That makes
245 * extension of the format actually undesirable because old compress
246 * would just crash on the new format instead of giving a meaningful
247 * error message. It does check the number of bits, but it's more
248 * helpful to say "unsupported format, get a new version" than
249 * "can only handle 16 bits".
252 /* tailor.h -- target dependent definitions
253 * Copyright (C) 1992-1993 Jean-loup Gailly.
254 * This is free software; you can redistribute it and/or modify it under the
255 * terms of the GNU General Public License, see the file COPYING.
258 /* The target dependent definitions should be defined here only.
259 * The target dependent functions should be defined in tailor.c.
263 /* Common defaults */
266 # define OS_CODE 0x03 /* assume Unix */
270 # define PATH_SEP '/'
274 # define OPTIONS_VAR "GZIP"
278 # define Z_SUFFIX ".gz"
282 # define MAX_SUFFIX MAX_EXT_CHARS
284 # define MAX_SUFFIX 30
289 DECLARE(uch, inbuf, INBUFSIZ + INBUF_EXTRA);
290 DECLARE(uch, outbuf, OUTBUFSIZ + OUTBUF_EXTRA);
291 DECLARE(ush, d_buf, DIST_BUFSIZE);
292 DECLARE(uch, window, 2L * WSIZE);
293 DECLARE(ush, tab_prefix, 1L << BITS);
295 static int foreground; /* set if program run in foreground */
296 static int method = DEFLATED; /* compression method */
297 static int exit_code = OK; /* program exit code */
298 static int part_nb; /* number of parts in .gz file */
299 static long time_stamp; /* original time stamp (modification time) */
300 static long ifile_size; /* input file size, -1 for devices (debug only) */
301 static char z_suffix[MAX_SUFFIX + 1]; /* default suffix (can be set with --suffix) */
302 static int z_len; /* strlen(z_suffix) */
304 static int ifd; /* input file descriptor */
305 static int ofd; /* output file descriptor */
306 static unsigned insize; /* valid bytes in inbuf */
307 static unsigned outcnt; /* bytes in output buffer */
310 /* Output a 16 bit value, lsb first */
311 static void put_short(ush w)
313 if (outcnt < OUTBUFSIZ - 2) {
314 outbuf[outcnt++] = (uch) ((w) & 0xff);
315 outbuf[outcnt++] = (uch) ((ush) (w) >> 8);
317 put_byte((uch) ((w) & 0xff));
318 put_byte((uch) ((ush) (w) >> 8));
322 /* ========================================================================
323 * Signal and error handler.
325 static void abort_gzip(int ignored)
330 /* ===========================================================================
331 * Clear input and output buffers
333 static void clear_bufs(void)
340 /* ===========================================================================
341 * Does the same as write(), but also handles partial pipe writes and checks
344 static void write_buf(int fd, void *buf, unsigned cnt)
348 while ((n = write(fd, buf, cnt)) != cnt) {
349 if (n == (unsigned) (-1)) bb_error_msg_and_die("can't write");
351 buf = (void *) ((char *) buf + n);
355 /* ===========================================================================
356 * Run a set of bytes through the crc shift register. If s is a NULL
357 * pointer, then initialize the crc shift register contents instead.
358 * Return the current crc in either case.
360 static ulg updcrc(uch * s, unsigned n)
362 static ulg crc = (ulg) 0xffffffffL; /* shift register contents */
363 register ulg c; /* temporary variable */
364 static unsigned long crc_32_tab[256];
366 if (crc_32_tab[1] == 0x00000000L) {
367 unsigned long csr; /* crc shift register */
368 const unsigned long e = 0xedb88320L; /* polynomial exclusive-or pattern */
369 int i; /* counter for all possible eight bit values */
370 int k; /* byte being shifted into crc apparatus */
372 /* Compute table of CRC's. */
373 for (i = 1; i < 256; i++) {
375 /* The idea to initialize the register with the byte instead of
376 * zero was stolen from Haruhiko Okumura's ar002
379 csr = csr & 1 ? (csr >> 1) ^ e : csr >> 1;
390 c = crc_32_tab[((int) c ^ (*s++)) & 0xff] ^ (c >> 8);
394 return c ^ 0xffffffffL; /* (instead of ~c for 64-bit machines) */
397 /* bits.c -- output variable-length bit strings
398 * Copyright (C) 1992-1993 Jean-loup Gailly
399 * This is free software; you can redistribute it and/or modify it under the
400 * terms of the GNU General Public License, see the file COPYING.
407 * Output variable-length bit strings. Compression can be done
408 * to a file or to memory. (The latter is not supported in this version.)
412 * The PKZIP "deflate" file format interprets compressed file data
413 * as a sequence of bits. Multi-bit strings in the file may cross
414 * byte boundaries without restriction.
416 * The first bit of each byte is the low-order bit.
418 * The routines in this file allow a variable-length bit value to
419 * be output right-to-left (useful for literal values). For
420 * left-to-right output (useful for code strings from the tree routines),
421 * the bits must have been reversed first with bi_reverse().
423 * For in-memory compression, the compressed bit stream goes directly
424 * into the requested output buffer. The input data is read in blocks
425 * by the mem_read() function. The buffer is limited to 64K on 16 bit
430 * void bi_init (FILE *zipfile)
431 * Initialize the bit string routines.
433 * void send_bits (int value, int length)
434 * Write out a bit string, taking the source bits right to
437 * int bi_reverse (int value, int length)
438 * Reverse the bits of a bit string, taking the source bits left to
439 * right and emitting them right to left.
441 * void bi_windup (void)
442 * Write out any remaining bits in an incomplete byte.
444 * void copy_block(char *buf, unsigned len, int header)
445 * Copy a stored block to the zip file, storing first the length and
446 * its one's complement if requested.
450 /* ===========================================================================
451 * Local data used by the "bit string" routines.
454 static file_t zfile; /* output gzip file */
456 static unsigned short bi_buf;
458 /* Output buffer. bits are inserted starting at the bottom (least significant
462 #define Buf_size (8 * 2*sizeof(char))
463 /* Number of bits used within bi_buf. (bi_buf might be implemented on
464 * more than 16 bits on some systems.)
469 /* Current input function. Set to mem_read for in-memory compression */
472 ulg bits_sent; /* bit length of the compressed data */
475 /* ===========================================================================
476 * Initialize the bit string routines.
478 static void bi_init(file_t zipfile)
487 /* Set the defaults for file compression. They are set by memcompress
488 * for in-memory compression.
490 if (zfile != NO_FILE) {
491 read_buf = file_read;
495 /* ===========================================================================
496 * Send a value on a given number of bits.
497 * IN assertion: length <= 16 and value fits in length bits.
499 static void send_bits(int value, int length)
502 Tracev((stderr, " l %2d v %4x ", length, value));
503 Assert(length > 0 && length <= 15, "invalid length");
504 bits_sent += (ulg) length;
506 /* If not enough room in bi_buf, use (valid) bits from bi_buf and
507 * (16 - bi_valid) bits from value, leaving (width - (16-bi_valid))
508 * unused bits in value.
510 if (bi_valid > (int) Buf_size - length) {
511 bi_buf |= (value << bi_valid);
513 bi_buf = (ush) value >> (Buf_size - bi_valid);
514 bi_valid += length - Buf_size;
516 bi_buf |= value << bi_valid;
521 /* ===========================================================================
522 * Reverse the first len bits of a code, using straightforward code (a faster
523 * method would use a table)
524 * IN assertion: 1 <= len <= 15
526 static unsigned bi_reverse(unsigned code, int len)
528 register unsigned res = 0;
532 code >>= 1, res <<= 1;
537 /* ===========================================================================
538 * Write out any remaining bits in an incomplete byte.
540 static void bi_windup(void)
544 } else if (bi_valid > 0) {
550 bits_sent = (bits_sent + 7) & ~7;
554 /* ===========================================================================
555 * Copy a stored block to the zip file, storing first the length and its
556 * one's complement if requested.
558 static void copy_block(char *buf, unsigned len, int header)
560 bi_windup(); /* align on byte boundary */
563 put_short((ush) len);
564 put_short((ush) ~ len);
570 bits_sent += (ulg) len << 3;
577 /* deflate.c -- compress data using the deflation algorithm
578 * Copyright (C) 1992-1993 Jean-loup Gailly
579 * This is free software; you can redistribute it and/or modify it under the
580 * terms of the GNU General Public License, see the file COPYING.
586 * Identify new text as repetitions of old text within a fixed-
587 * length sliding window trailing behind the new text.
591 * The "deflation" process depends on being able to identify portions
592 * of the input text which are identical to earlier input (within a
593 * sliding window trailing behind the input currently being processed).
595 * The most straightforward technique turns out to be the fastest for
596 * most input files: try all possible matches and select the longest.
597 * The key feature of this algorithm is that insertions into the string
598 * dictionary are very simple and thus fast, and deletions are avoided
599 * completely. Insertions are performed at each input character, whereas
600 * string matches are performed only when the previous match ends. So it
601 * is preferable to spend more time in matches to allow very fast string
602 * insertions and avoid deletions. The matching algorithm for small
603 * strings is inspired from that of Rabin & Karp. A brute force approach
604 * is used to find longer strings when a small match has been found.
605 * A similar algorithm is used in comic (by Jan-Mark Wams) and freeze
606 * (by Leonid Broukhis).
607 * A previous version of this file used a more sophisticated algorithm
608 * (by Fiala and Greene) which is guaranteed to run in linear amortized
609 * time, but has a larger average cost, uses more memory and is patented.
610 * However the F&G algorithm may be faster for some highly redundant
611 * files if the parameter max_chain_length (described below) is too large.
615 * The idea of lazy evaluation of matches is due to Jan-Mark Wams, and
616 * I found it in 'freeze' written by Leonid Broukhis.
617 * Thanks to many info-zippers for bug reports and testing.
621 * APPNOTE.TXT documentation file in PKZIP 1.93a distribution.
623 * A description of the Rabin and Karp algorithm is given in the book
624 * "Algorithms" by R. Sedgewick, Addison-Wesley, p252.
626 * Fiala,E.R., and Greene,D.H.
627 * Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595
631 * void lm_init (int pack_level, ush *flags)
632 * Initialize the "longest match" routines for a new file
635 * Processes a new input file and return its compressed length. Sets
636 * the compressed length, crc, deflate flags and internal file
641 /* ===========================================================================
642 * Configuration parameters
645 /* Compile with MEDIUM_MEM to reduce the memory requirements or
646 * with SMALL_MEM to use as little memory as possible. Use BIG_MEM if the
647 * entire input file can be held in memory (not possible on 16 bit systems).
648 * Warning: defining these symbols affects HASH_BITS (see below) and thus
649 * affects the compression ratio. The compressed output
650 * is still correct, and might even be smaller in some cases.
654 # define HASH_BITS 13 /* Number of bits used to hash strings */
657 # define HASH_BITS 14
660 # define HASH_BITS 15
661 /* For portability to 16 bit machines, do not use values above 15. */
664 /* To save space (see unlzw.c), we overlay prev+head with tab_prefix and
665 * window with tab_suffix. Check that we can do this:
667 #if (WSIZE<<1) > (1<<BITS)
668 # error cannot overlay window with tab_suffix and prev with tab_prefix0
670 #if HASH_BITS > BITS-1
671 # error cannot overlay head with tab_prefix1
673 #define HASH_SIZE (unsigned)(1<<HASH_BITS)
674 #define HASH_MASK (HASH_SIZE-1)
675 #define WMASK (WSIZE-1)
676 /* HASH_SIZE and WSIZE must be powers of two */
678 /* Tail of hash chains */
681 /* speed options for the general purpose bit flag */
683 # define TOO_FAR 4096
685 /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
686 /* ===========================================================================
687 * Local data used by the "longest match" routines.
690 typedef unsigned IPos;
692 /* A Pos is an index in the character window. We use short instead of int to
693 * save space in the various tables. IPos is used only for parameter passing.
696 /* DECLARE(uch, window, 2L*WSIZE); */
697 /* Sliding window. Input bytes are read into the second half of the window,
698 * and move to the first half later to keep a dictionary of at least WSIZE
699 * bytes. With this organization, matches are limited to a distance of
700 * WSIZE-MAX_MATCH bytes, but this ensures that IO is always
701 * performed with a length multiple of the block size. Also, it limits
702 * the window size to 64K, which is quite useful on MSDOS.
703 * To do: limit the window size to WSIZE+BSZ if SMALL_MEM (the code would
704 * be less efficient).
707 /* DECLARE(Pos, prev, WSIZE); */
708 /* Link to older string with same hash index. To limit the size of this
709 * array to 64K, this link is maintained only for the last 32K strings.
710 * An index in this array is thus a window index modulo 32K.
713 /* DECLARE(Pos, head, 1<<HASH_BITS); */
714 /* Heads of the hash chains or NIL. */
716 static const ulg window_size = (ulg) 2 * WSIZE;
718 /* window size, 2*WSIZE except for MMAP or BIG_MEM, where it is the
719 * input file length plus MIN_LOOKAHEAD.
722 static long block_start;
724 /* window position at the beginning of the current output block. Gets
725 * negative when the window is moved backwards.
728 static unsigned ins_h; /* hash index of string to be inserted */
730 #define H_SHIFT ((HASH_BITS+MIN_MATCH-1)/MIN_MATCH)
731 /* Number of bits by which ins_h and del_h must be shifted at each
732 * input step. It must be such that after MIN_MATCH steps, the oldest
733 * byte no longer takes part in the hash key, that is:
734 * H_SHIFT * MIN_MATCH >= HASH_BITS
737 static unsigned int prev_length;
739 /* Length of the best match at previous step. Matches not greater than this
740 * are discarded. This is used in the lazy match evaluation.
743 static unsigned strstart; /* start of string to insert */
744 static unsigned match_start; /* start of matching string */
745 static int eofile; /* flag set at end of input file */
746 static unsigned lookahead; /* number of valid bytes ahead in window */
748 static const unsigned max_chain_length = 4096;
750 /* To speed up deflation, hash chains are never searched beyond this length.
751 * A higher limit improves compression ratio but degrades the speed.
754 static const unsigned int max_lazy_match = 258;
756 /* Attempt to find a better match only when the current match is strictly
757 * smaller than this value. This mechanism is used only for compression
760 #define max_insert_length max_lazy_match
761 /* Insert new strings in the hash table only if the match length
762 * is not greater than this length. This saves time but degrades compression.
763 * max_insert_length is used only for compression levels <= 3.
766 static const unsigned good_match = 32;
768 /* Use a faster search when the previous match is longer than this */
771 /* Values for max_lazy_match, good_match and max_chain_length, depending on
772 * the desired pack level (0..9). The values given below have been tuned to
773 * exclude worst case performance for pathological files. Better values may be
774 * found for specific files.
777 static const int nice_match = 258; /* Stop searching when current match exceeds this */
779 /* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
780 * For deflate_fast() (levels <= 3) good is ignored and lazy has a different
785 /* result of memcmp for equal strings */
787 /* ===========================================================================
788 * Prototypes for local functions.
790 static void fill_window(void);
792 static int longest_match(IPos cur_match);
795 static void check_match(IPos start, IPos match, int length);
798 /* ===========================================================================
799 * Update a hash value with the given input byte
800 * IN assertion: all calls to to UPDATE_HASH are made with consecutive
801 * input characters, so that a running hash key can be computed from the
802 * previous key instead of complete recalculation each time.
804 #define UPDATE_HASH(h,c) (h = (((h)<<H_SHIFT) ^ (c)) & HASH_MASK)
806 /* ===========================================================================
807 * Insert string s in the dictionary and set match_head to the previous head
808 * of the hash chain (the most recent string with same hash key). Return
809 * the previous length of the hash chain.
810 * IN assertion: all calls to to INSERT_STRING are made with consecutive
811 * input characters and the first MIN_MATCH bytes of s are valid
812 * (except for the last MIN_MATCH-1 bytes of the input file).
814 #define INSERT_STRING(s, match_head) \
815 (UPDATE_HASH(ins_h, window[(s) + MIN_MATCH-1]), \
816 prev[(s) & WMASK] = match_head = head[ins_h], \
819 /* ===========================================================================
820 * Initialize the "longest match" routines for a new file
822 static void lm_init(ush * flags)
826 /* Initialize the hash table. */
827 memset(head, 0, HASH_SIZE * sizeof(*head));
828 /* prev will be initialized on the fly */
831 /* ??? reduce max_chain_length for binary files */
836 lookahead = read_buf((char *) window,
837 sizeof(int) <= 2 ? (unsigned) WSIZE : 2 * WSIZE);
839 if (lookahead == 0 || lookahead == (unsigned) EOF) {
840 eofile = 1, lookahead = 0;
844 /* Make sure that we always have enough lookahead. This is important
845 * if input comes from a device such as a tty.
847 while (lookahead < MIN_LOOKAHEAD && !eofile)
851 for (j = 0; j < MIN_MATCH - 1; j++)
852 UPDATE_HASH(ins_h, window[j]);
853 /* If lookahead < MIN_MATCH, ins_h is garbage, but this is
854 * not important since only literal bytes will be emitted.
858 /* ===========================================================================
859 * Set match_start to the longest match starting at the given string and
860 * return its length. Matches shorter or equal to prev_length are discarded,
861 * in which case the result is equal to prev_length and match_start is
863 * IN assertions: cur_match is the head of the hash chain for the current
864 * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
867 /* For MSDOS, OS/2 and 386 Unix, an optimized version is in match.asm or
868 * match.s. The code is functionally equivalent, so you can use the C version
871 static int longest_match(IPos cur_match)
873 unsigned chain_length = max_chain_length; /* max hash chain length */
874 register uch *scan = window + strstart; /* current string */
875 register uch *match; /* matched string */
876 register int len; /* length of current match */
877 int best_len = prev_length; /* best match length so far */
879 strstart > (IPos) MAX_DIST ? strstart - (IPos) MAX_DIST : NIL;
880 /* Stop when cur_match becomes <= limit. To simplify the code,
881 * we prevent matches with the string of window index 0.
884 /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
885 * It is easy to get rid of this optimization if necessary.
887 #if HASH_BITS < 8 || MAX_MATCH != 258
888 # error Code too clever
890 register uch *strend = window + strstart + MAX_MATCH;
891 register uch scan_end1 = scan[best_len - 1];
892 register uch scan_end = scan[best_len];
894 /* Do not waste too much time if we already have a good match: */
895 if (prev_length >= good_match) {
898 Assert(strstart <= window_size - MIN_LOOKAHEAD, "insufficient lookahead");
901 Assert(cur_match < strstart, "no future");
902 match = window + cur_match;
904 /* Skip to next match if the match length cannot increase
905 * or if the match length is less than 2:
907 if (match[best_len] != scan_end ||
908 match[best_len - 1] != scan_end1 ||
909 *match != *scan || *++match != scan[1])
912 /* The check at best_len-1 can be removed because it will be made
913 * again later. (This heuristic is not always a win.)
914 * It is not necessary to compare scan[2] and match[2] since they
915 * are always equal when the other bytes match, given that
916 * the hash keys are equal and that HASH_BITS >= 8.
920 /* We check for insufficient lookahead only every 8th comparison;
921 * the 256th check will be made at strstart+258.
924 } while (*++scan == *++match && *++scan == *++match &&
925 *++scan == *++match && *++scan == *++match &&
926 *++scan == *++match && *++scan == *++match &&
927 *++scan == *++match && *++scan == *++match && scan < strend);
929 len = MAX_MATCH - (int) (strend - scan);
930 scan = strend - MAX_MATCH;
932 if (len > best_len) {
933 match_start = cur_match;
935 if (len >= nice_match)
937 scan_end1 = scan[best_len - 1];
938 scan_end = scan[best_len];
940 } while ((cur_match = prev[cur_match & WMASK]) > limit
941 && --chain_length != 0);
947 /* ===========================================================================
948 * Check that the match at match_start is indeed a match.
950 static void check_match(IPos start, IPos match, int length)
952 /* check that the match is indeed a match */
953 if (memcmp((char *) window + match,
954 (char *) window + start, length) != EQUAL) {
955 bb_error_msg(" start %d, match %d, length %d", start, match, length);
956 bb_error_msg("invalid match");
959 bb_error_msg("\\[%d,%d]", start - match, length);
961 putc(window[start++], stderr);
962 } while (--length != 0);
966 # define check_match(start, match, length)
969 /* ===========================================================================
970 * Fill the window when the lookahead becomes insufficient.
971 * Updates strstart and lookahead, and sets eofile if end of input file.
972 * IN assertion: lookahead < MIN_LOOKAHEAD && strstart + lookahead > 0
973 * OUT assertions: at least one byte has been read, or eofile is set;
974 * file reads are performed for at least two bytes (required for the
975 * translate_eol option).
977 static void fill_window(void)
979 register unsigned n, m;
981 (unsigned) (window_size - (ulg) lookahead - (ulg) strstart);
982 /* Amount of free space at the end of the window. */
984 /* If the window is almost full and there is insufficient lookahead,
985 * move the upper half to the lower one to make room in the upper half.
987 if (more == (unsigned) EOF) {
988 /* Very unlikely, but possible on 16 bit machine if strstart == 0
989 * and lookahead == 1 (input done one byte at time)
992 } else if (strstart >= WSIZE + MAX_DIST) {
993 /* By the IN assertion, the window is not empty so we can't confuse
994 * more == 0 with more == 64K on a 16 bit machine.
996 Assert(window_size == (ulg) 2 * WSIZE, "no sliding with BIG_MEM");
998 memcpy((char *) window, (char *) window + WSIZE, (unsigned) WSIZE);
999 match_start -= WSIZE;
1000 strstart -= WSIZE; /* we now have strstart >= MAX_DIST: */
1002 block_start -= (long) WSIZE;
1004 for (n = 0; n < HASH_SIZE; n++) {
1006 head[n] = (Pos) (m >= WSIZE ? m - WSIZE : NIL);
1008 for (n = 0; n < WSIZE; n++) {
1010 prev[n] = (Pos) (m >= WSIZE ? m - WSIZE : NIL);
1011 /* If n is not on any hash chain, prev[n] is garbage but
1012 * its value will never be used.
1017 /* At this point, more >= 2 */
1019 n = read_buf((char *) window + strstart + lookahead, more);
1020 if (n == 0 || n == (unsigned) EOF) {
1028 /* ===========================================================================
1029 * Flush the current block, with given end-of-file flag.
1030 * IN assertion: strstart is set to the end of the current match.
1032 #define FLUSH_BLOCK(eof) \
1033 flush_block(block_start >= 0L ? (char*)&window[(unsigned)block_start] : \
1034 (char*)NULL, (long)strstart - block_start, (eof))
1036 /* ===========================================================================
1037 * Same as above, but achieves better compression. We use a lazy
1038 * evaluation for matches: a match is finally adopted only if there is
1039 * no better match at the next window position.
1041 static ulg deflate(void)
1043 IPos hash_head; /* head of hash chain */
1044 IPos prev_match; /* previous match */
1045 int flush; /* set if current block must be flushed */
1046 int match_available = 0; /* set if previous match exists */
1047 register unsigned match_length = MIN_MATCH - 1; /* length of best match */
1049 /* Process the input block. */
1050 while (lookahead != 0) {
1051 /* Insert the string window[strstart .. strstart+2] in the
1052 * dictionary, and set hash_head to the head of the hash chain:
1054 INSERT_STRING(strstart, hash_head);
1056 /* Find the longest match, discarding those <= prev_length.
1058 prev_length = match_length, prev_match = match_start;
1059 match_length = MIN_MATCH - 1;
1061 if (hash_head != NIL && prev_length < max_lazy_match &&
1062 strstart - hash_head <= MAX_DIST) {
1063 /* To simplify the code, we prevent matches with the string
1064 * of window index 0 (in particular we have to avoid a match
1065 * of the string with itself at the start of the input file).
1067 match_length = longest_match(hash_head);
1068 /* longest_match() sets match_start */
1069 if (match_length > lookahead)
1070 match_length = lookahead;
1072 /* Ignore a length 3 match if it is too distant: */
1073 if (match_length == MIN_MATCH && strstart - match_start > TOO_FAR) {
1074 /* If prev_match is also MIN_MATCH, match_start is garbage
1075 * but we will ignore the current match anyway.
1080 /* If there was a match at the previous step and the current
1081 * match is not better, output the previous match:
1083 if (prev_length >= MIN_MATCH && match_length <= prev_length) {
1085 check_match(strstart - 1, prev_match, prev_length);
1088 ct_tally(strstart - 1 - prev_match, prev_length - MIN_MATCH);
1090 /* Insert in hash table all strings up to the end of the match.
1091 * strstart-1 and strstart are already inserted.
1093 lookahead -= prev_length - 1;
1097 INSERT_STRING(strstart, hash_head);
1098 /* strstart never exceeds WSIZE-MAX_MATCH, so there are
1099 * always MIN_MATCH bytes ahead. If lookahead < MIN_MATCH
1100 * these bytes are garbage, but it does not matter since the
1101 * next lookahead bytes will always be emitted as literals.
1103 } while (--prev_length != 0);
1104 match_available = 0;
1105 match_length = MIN_MATCH - 1;
1108 FLUSH_BLOCK(0), block_start = strstart;
1110 } else if (match_available) {
1111 /* If there was no match at the previous position, output a
1112 * single literal. If there was a match but the current match
1113 * is longer, truncate the previous match to a single literal.
1115 Tracevv((stderr, "%c", window[strstart - 1]));
1116 if (ct_tally(0, window[strstart - 1])) {
1117 FLUSH_BLOCK(0), block_start = strstart;
1122 /* There is no previous match to compare with, wait for
1123 * the next step to decide.
1125 match_available = 1;
1129 Assert(strstart <= isize && lookahead <= isize, "a bit too far");
1131 /* Make sure that we always have enough lookahead, except
1132 * at the end of the input file. We need MAX_MATCH bytes
1133 * for the next match, plus MIN_MATCH bytes to insert the
1134 * string following the next match.
1136 while (lookahead < MIN_LOOKAHEAD && !eofile)
1139 if (match_available)
1140 ct_tally(0, window[strstart - 1]);
1142 return FLUSH_BLOCK(1); /* eof */
1145 /* gzip (GNU zip) -- compress files with zip algorithm and 'compress' interface
1146 * Copyright (C) 1992-1993 Jean-loup Gailly
1147 * The unzip code was written and put in the public domain by Mark Adler.
1148 * Portions of the lzw code are derived from the public domain 'compress'
1149 * written by Spencer Thomas, Joe Orost, James Woods, Jim McKie, Steve Davies,
1150 * Ken Turkowski, Dave Mack and Peter Jannesen.
1152 * See the license_msg below and the file COPYING for the software license.
1153 * See the file algorithm.doc for the compression algorithms and file formats.
1156 /* Compress files with zip algorithm and 'compress' interface.
1157 * See usage() and help() functions below for all options.
1159 * file.gz: compressed file with same mode, owner, and utimes
1160 * or stdout with -c option or if stdin used as input.
1161 * If the output file name had to be truncated, the original name is kept
1162 * in the compressed file.
1167 typedef struct dirent dir_type;
1169 /* ======================================================================== */
1170 int gzip_main(int argc, char **argv)
1175 struct stat statBuf;
1181 while ((opt = getopt(argc, argv, "cf123456789dq")) != -1) {
1189 /* Ignore 1-9 (compression level) options */
1202 #ifdef CONFIG_GUNZIP
1205 return gunzip_main(argc, argv);
1212 foreground = signal(SIGINT, SIG_IGN) != SIG_IGN;
1214 (void) signal(SIGINT, abort_gzip);
1217 if (signal(SIGTERM, SIG_IGN) != SIG_IGN) {
1218 (void) signal(SIGTERM, abort_gzip);
1222 if (signal(SIGHUP, SIG_IGN) != SIG_IGN) {
1223 (void) signal(SIGHUP, abort_gzip);
1227 strncpy(z_suffix, Z_SUFFIX, sizeof(z_suffix) - 1);
1228 z_len = strlen(z_suffix);
1230 /* Allocate all global buffers (for DYN_ALLOC option) */
1231 ALLOC(uch, inbuf, INBUFSIZ + INBUF_EXTRA);
1232 ALLOC(uch, outbuf, OUTBUFSIZ + OUTBUF_EXTRA);
1233 ALLOC(ush, d_buf, DIST_BUFSIZE);
1234 ALLOC(uch, window, 2L * WSIZE);
1235 ALLOC(ush, tab_prefix, 1L << BITS);
1240 if (optind == argc) {
1243 zip(STDIN_FILENO, STDOUT_FILENO);
1247 for (i = optind; i < argc; i++) {
1251 if (strcmp(argv[i], "-") == 0) {
1254 inFileNum = STDIN_FILENO;
1255 outFileNum = STDOUT_FILENO;
1257 inFileNum = open(argv[i], O_RDONLY);
1258 if (inFileNum < 0 || fstat(inFileNum, &statBuf) < 0)
1259 bb_perror_msg_and_die("%s", argv[i]);
1260 time_stamp = statBuf.st_ctime;
1261 ifile_size = statBuf.st_size;
1264 path = xmalloc(strlen(argv[i]) + 4);
1265 strcpy(path, argv[i]);
1266 strcat(path, ".gz");
1268 /* Open output file */
1269 #if (__GLIBC__ >= 2) && (__GLIBC_MINOR__ >= 1) && defined O_NOFOLLOW
1271 open(path, O_RDWR | O_CREAT | O_EXCL | O_NOFOLLOW);
1273 outFileNum = open(path, O_RDWR | O_CREAT | O_EXCL);
1275 if (outFileNum < 0) {
1276 bb_perror_msg("%s", path);
1281 /* Set permissions on the file */
1282 fchmod(outFileNum, statBuf.st_mode);
1284 outFileNum = STDOUT_FILENO;
1287 if (path == NULL && isatty(outFileNum) && force == 0) {
1289 ("compressed data not written to a terminal. Use -f to force compression.");
1294 result = zip(inFileNum, outFileNum);
1300 /* Delete the original file */
1302 delFileName = argv[i];
1306 if (unlink(delFileName) < 0)
1307 bb_perror_msg("%s", delFileName);
1317 /* trees.c -- output deflated data using Huffman coding
1318 * Copyright (C) 1992-1993 Jean-loup Gailly
1319 * This is free software; you can redistribute it and/or modify it under the
1320 * terms of the GNU General Public License, see the file COPYING.
1326 * Encode various sets of source values using variable-length
1327 * binary code trees.
1331 * The PKZIP "deflation" process uses several Huffman trees. The more
1332 * common source values are represented by shorter bit sequences.
1334 * Each code tree is stored in the ZIP file in a compressed form
1335 * which is itself a Huffman encoding of the lengths of
1336 * all the code strings (in ascending order by source values).
1337 * The actual code strings are reconstructed from the lengths in
1338 * the UNZIP process, as described in the "application note"
1339 * (APPNOTE.TXT) distributed as part of PKWARE's PKZIP program.
1344 * Data Compression: Techniques and Applications, pp. 53-55.
1345 * Lifetime Learning Publications, 1985. ISBN 0-534-03418-7.
1348 * Data Compression: Methods and Theory, pp. 49-50.
1349 * Computer Science Press, 1988. ISBN 0-7167-8156-5.
1353 * Addison-Wesley, 1983. ISBN 0-201-06672-6.
1357 * void ct_init (ush *attr, int *methodp)
1358 * Allocate the match buffer, initialize the various tables and save
1359 * the location of the internal file attribute (ascii/binary) and
1360 * method (DEFLATE/STORE)
1362 * void ct_tally (int dist, int lc);
1363 * Save the match info and tally the frequency counts.
1365 * long flush_block (char *buf, ulg stored_len, int eof)
1366 * Determine the best encoding for the current block: dynamic trees,
1367 * static trees or store, and output the encoded block to the zip
1368 * file. Returns the total compressed length for the file so far.
1372 /* ===========================================================================
1377 /* All codes must not exceed MAX_BITS bits */
1379 #define MAX_BL_BITS 7
1380 /* Bit length codes must not exceed MAX_BL_BITS bits */
1382 #define LENGTH_CODES 29
1383 /* number of length codes, not counting the special END_BLOCK code */
1385 #define LITERALS 256
1386 /* number of literal bytes 0..255 */
1388 #define END_BLOCK 256
1389 /* end of block literal code */
1391 #define L_CODES (LITERALS+1+LENGTH_CODES)
1392 /* number of Literal or Length codes, including the END_BLOCK code */
1395 /* number of distance codes */
1398 /* number of codes used to transfer the bit lengths */
1400 typedef uch extra_bits_t;
1402 /* extra bits for each length code */
1403 static const extra_bits_t extra_lbits[LENGTH_CODES]
1404 = { 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4,
1408 /* extra bits for each distance code */
1409 static const extra_bits_t extra_dbits[D_CODES]
1410 = { 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9,
1411 10, 10, 11, 11, 12, 12, 13, 13
1414 /* extra bits for each bit length code */
1415 static const extra_bits_t extra_blbits[BL_CODES]
1416 = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 3, 7 };
1418 #define STORED_BLOCK 0
1419 #define STATIC_TREES 1
1421 /* The three kinds of block type */
1425 # define LIT_BUFSIZE 0x2000
1428 # define LIT_BUFSIZE 0x4000
1430 # define LIT_BUFSIZE 0x8000
1434 #ifndef DIST_BUFSIZE
1435 # define DIST_BUFSIZE LIT_BUFSIZE
1437 /* Sizes of match buffers for literals/lengths and distances. There are
1438 * 4 reasons for limiting LIT_BUFSIZE to 64K:
1439 * - frequencies can be kept in 16 bit counters
1440 * - if compression is not successful for the first block, all input data is
1441 * still in the window so we can still emit a stored block even when input
1442 * comes from standard input. (This can also be done for all blocks if
1443 * LIT_BUFSIZE is not greater than 32K.)
1444 * - if compression is not successful for a file smaller than 64K, we can
1445 * even emit a stored file instead of a stored block (saving 5 bytes).
1446 * - creating new Huffman trees less frequently may not provide fast
1447 * adaptation to changes in the input data statistics. (Take for
1448 * example a binary file with poorly compressible code followed by
1449 * a highly compressible string table.) Smaller buffer sizes give
1450 * fast adaptation but have of course the overhead of transmitting trees
1452 * - I can't count above 4
1453 * The current code is general and allows DIST_BUFSIZE < LIT_BUFSIZE (to save
1454 * memory at the expense of compression). Some optimizations would be possible
1455 * if we rely on DIST_BUFSIZE == LIT_BUFSIZE.
1457 #if LIT_BUFSIZE > INBUFSIZ
1458 #error cannot overlay l_buf and inbuf
1461 /* repeat previous bit length 3-6 times (2 bits of repeat count) */
1462 #define REPZ_3_10 17
1463 /* repeat a zero length 3-10 times (3 bits of repeat count) */
1464 #define REPZ_11_138 18
1465 /* repeat a zero length 11-138 times (7 bits of repeat count) */
1467 /* ===========================================================================
1471 /* Data structure describing a single value and its code string. */
1472 typedef struct ct_data {
1474 ush freq; /* frequency count */
1475 ush code; /* bit string */
1478 ush dad; /* father node in Huffman tree */
1479 ush len; /* length of bit string */
1483 #define Freq fc.freq
1484 #define Code fc.code
1488 #define HEAP_SIZE (2*L_CODES+1)
1489 /* maximum heap size */
1491 static ct_data dyn_ltree[HEAP_SIZE]; /* literal and length tree */
1492 static ct_data dyn_dtree[2 * D_CODES + 1]; /* distance tree */
1494 static ct_data static_ltree[L_CODES + 2];
1496 /* The static literal tree. Since the bit lengths are imposed, there is no
1497 * need for the L_CODES extra codes used during heap construction. However
1498 * The codes 286 and 287 are needed to build a canonical tree (see ct_init
1502 static ct_data static_dtree[D_CODES];
1504 /* The static distance tree. (Actually a trivial tree since all codes use
1508 static ct_data bl_tree[2 * BL_CODES + 1];
1510 /* Huffman tree for the bit lengths */
1512 typedef struct tree_desc {
1513 ct_data *dyn_tree; /* the dynamic tree */
1514 ct_data *static_tree; /* corresponding static tree or NULL */
1515 const extra_bits_t *extra_bits; /* extra bits for each code or NULL */
1516 int extra_base; /* base index for extra_bits */
1517 int elems; /* max number of elements in the tree */
1518 int max_length; /* max bit length for the codes */
1519 int max_code; /* largest code with non zero frequency */
1522 static tree_desc l_desc =
1523 { dyn_ltree, static_ltree, extra_lbits, LITERALS + 1, L_CODES,
1527 static tree_desc d_desc =
1528 { dyn_dtree, static_dtree, extra_dbits, 0, D_CODES, MAX_BITS, 0 };
1530 static tree_desc bl_desc =
1531 { bl_tree, (ct_data *) 0, extra_blbits, 0, BL_CODES, MAX_BL_BITS,
1536 static ush bl_count[MAX_BITS + 1];
1538 /* number of codes at each bit length for an optimal tree */
1540 static const uch bl_order[BL_CODES]
1541 = { 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15 };
1543 /* The lengths of the bit length codes are sent in order of decreasing
1544 * probability, to avoid transmitting the lengths for unused bit length codes.
1547 static int heap[2 * L_CODES + 1]; /* heap used to build the Huffman trees */
1548 static int heap_len; /* number of elements in the heap */
1549 static int heap_max; /* element of largest frequency */
1551 /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used.
1552 * The same heap array is used to build all trees.
1555 static uch depth[2 * L_CODES + 1];
1557 /* Depth of each subtree used as tie breaker for trees of equal frequency */
1559 static uch length_code[MAX_MATCH - MIN_MATCH + 1];
1561 /* length code for each normalized match length (0 == MIN_MATCH) */
1563 static uch dist_code[512];
1565 /* distance codes. The first 256 values correspond to the distances
1566 * 3 .. 258, the last 256 values correspond to the top 8 bits of
1567 * the 15 bit distances.
1570 static int base_length[LENGTH_CODES];
1572 /* First normalized length for each code (0 = MIN_MATCH) */
1574 static int base_dist[D_CODES];
1576 /* First normalized distance for each code (0 = distance of 1) */
1579 /* DECLARE(uch, l_buf, LIT_BUFSIZE); buffer for literals or lengths */
1581 /* DECLARE(ush, d_buf, DIST_BUFSIZE); buffer for distances */
1583 static uch flag_buf[(LIT_BUFSIZE / 8)];
1585 /* flag_buf is a bit array distinguishing literals from lengths in
1586 * l_buf, thus indicating the presence or absence of a distance.
1589 static unsigned last_lit; /* running index in l_buf */
1590 static unsigned last_dist; /* running index in d_buf */
1591 static unsigned last_flags; /* running index in flag_buf */
1592 static uch flags; /* current flags not yet saved in flag_buf */
1593 static uch flag_bit; /* current bit used in flags */
1595 /* bits are filled in flags starting at bit 0 (least significant).
1596 * Note: these flags are overkill in the current code since we don't
1597 * take advantage of DIST_BUFSIZE == LIT_BUFSIZE.
1600 static ulg opt_len; /* bit length of current block with optimal trees */
1601 static ulg static_len; /* bit length of current block with static trees */
1603 static ulg compressed_len; /* total bit length of compressed file */
1606 static ush *file_type; /* pointer to UNKNOWN, BINARY or ASCII */
1607 static int *file_method; /* pointer to DEFLATE or STORE */
1609 /* ===========================================================================
1610 * Local (static) routines in this file.
1613 static void init_block(void);
1614 static void pqdownheap(ct_data * tree, int k);
1615 static void gen_bitlen(tree_desc * desc);
1616 static void gen_codes(ct_data * tree, int max_code);
1617 static void build_tree(tree_desc * desc);
1618 static void scan_tree(ct_data * tree, int max_code);
1619 static void send_tree(ct_data * tree, int max_code);
1620 static int build_bl_tree(void);
1621 static void send_all_trees(int lcodes, int dcodes, int blcodes);
1622 static void compress_block(ct_data * ltree, ct_data * dtree);
1623 static void set_file_type(void);
1627 # define send_code(c, tree) send_bits(tree[c].Code, tree[c].Len)
1628 /* Send a code of the given tree. c and tree must not have side effects */
1631 # define send_code(c, tree) \
1632 { if (verbose>1) bb_error_msg("\ncd %3d ",(c)); \
1633 send_bits(tree[c].Code, tree[c].Len); }
1636 #define d_code(dist) \
1637 ((dist) < 256 ? dist_code[dist] : dist_code[256+((dist)>>7)])
1638 /* Mapping from a distance to a distance code. dist is the distance - 1 and
1639 * must not have side effects. dist_code[256] and dist_code[257] are never
1643 /* the arguments must not have side effects */
1645 /* ===========================================================================
1646 * Allocate the match buffer, initialize the various tables and save the
1647 * location of the internal file attribute (ascii/binary) and method
1650 static void ct_init(ush * attr, int *methodp)
1652 int n; /* iterates over tree elements */
1653 int bits; /* bit counter */
1654 int length; /* length value */
1655 int code; /* code value */
1656 int dist; /* distance index */
1659 file_method = methodp;
1660 compressed_len = 0L;
1662 if (static_dtree[0].Len != 0)
1663 return; /* ct_init already called */
1665 /* Initialize the mapping length (0..255) -> length code (0..28) */
1667 for (code = 0; code < LENGTH_CODES - 1; code++) {
1668 base_length[code] = length;
1669 for (n = 0; n < (1 << extra_lbits[code]); n++) {
1670 length_code[length++] = (uch) code;
1673 Assert(length == 256, "ct_init: length != 256");
1674 /* Note that the length 255 (match length 258) can be represented
1675 * in two different ways: code 284 + 5 bits or code 285, so we
1676 * overwrite length_code[255] to use the best encoding:
1678 length_code[length - 1] = (uch) code;
1680 /* Initialize the mapping dist (0..32K) -> dist code (0..29) */
1682 for (code = 0; code < 16; code++) {
1683 base_dist[code] = dist;
1684 for (n = 0; n < (1 << extra_dbits[code]); n++) {
1685 dist_code[dist++] = (uch) code;
1688 Assert(dist == 256, "ct_init: dist != 256");
1689 dist >>= 7; /* from now on, all distances are divided by 128 */
1690 for (; code < D_CODES; code++) {
1691 base_dist[code] = dist << 7;
1692 for (n = 0; n < (1 << (extra_dbits[code] - 7)); n++) {
1693 dist_code[256 + dist++] = (uch) code;
1696 Assert(dist == 256, "ct_init: 256+dist != 512");
1698 /* Construct the codes of the static literal tree */
1699 for (bits = 0; bits <= MAX_BITS; bits++)
1703 static_ltree[n++].Len = 8, bl_count[8]++;
1705 static_ltree[n++].Len = 9, bl_count[9]++;
1707 static_ltree[n++].Len = 7, bl_count[7]++;
1709 static_ltree[n++].Len = 8, bl_count[8]++;
1710 /* Codes 286 and 287 do not exist, but we must include them in the
1711 * tree construction to get a canonical Huffman tree (longest code
1714 gen_codes((ct_data *) static_ltree, L_CODES + 1);
1716 /* The static distance tree is trivial: */
1717 for (n = 0; n < D_CODES; n++) {
1718 static_dtree[n].Len = 5;
1719 static_dtree[n].Code = bi_reverse(n, 5);
1722 /* Initialize the first block of the first file: */
1726 /* ===========================================================================
1727 * Initialize a new block.
1729 static void init_block(void)
1731 int n; /* iterates over tree elements */
1733 /* Initialize the trees. */
1734 for (n = 0; n < L_CODES; n++)
1735 dyn_ltree[n].Freq = 0;
1736 for (n = 0; n < D_CODES; n++)
1737 dyn_dtree[n].Freq = 0;
1738 for (n = 0; n < BL_CODES; n++)
1739 bl_tree[n].Freq = 0;
1741 dyn_ltree[END_BLOCK].Freq = 1;
1742 opt_len = static_len = 0L;
1743 last_lit = last_dist = last_flags = 0;
1749 /* Index within the heap array of least frequent node in the Huffman tree */
1752 /* ===========================================================================
1753 * Remove the smallest element from the heap and recreate the heap with
1754 * one less element. Updates heap and heap_len.
1756 #define pqremove(tree, top) \
1758 top = heap[SMALLEST]; \
1759 heap[SMALLEST] = heap[heap_len--]; \
1760 pqdownheap(tree, SMALLEST); \
1763 /* ===========================================================================
1764 * Compares to subtrees, using the tree depth as tie breaker when
1765 * the subtrees have equal frequency. This minimizes the worst case length.
1767 #define smaller(tree, n, m) \
1768 (tree[n].Freq < tree[m].Freq || \
1769 (tree[n].Freq == tree[m].Freq && depth[n] <= depth[m]))
1771 /* ===========================================================================
1772 * Restore the heap property by moving down the tree starting at node k,
1773 * exchanging a node with the smallest of its two sons if necessary, stopping
1774 * when the heap property is re-established (each father smaller than its
1777 static void pqdownheap(ct_data * tree, int k)
1780 int j = k << 1; /* left son of k */
1782 while (j <= heap_len) {
1783 /* Set j to the smallest of the two sons: */
1784 if (j < heap_len && smaller(tree, heap[j + 1], heap[j]))
1787 /* Exit if v is smaller than both sons */
1788 if (smaller(tree, v, heap[j]))
1791 /* Exchange v with the smallest son */
1795 /* And continue down the tree, setting j to the left son of k */
1801 /* ===========================================================================
1802 * Compute the optimal bit lengths for a tree and update the total bit length
1803 * for the current block.
1804 * IN assertion: the fields freq and dad are set, heap[heap_max] and
1805 * above are the tree nodes sorted by increasing frequency.
1806 * OUT assertions: the field len is set to the optimal bit length, the
1807 * array bl_count contains the frequencies for each bit length.
1808 * The length opt_len is updated; static_len is also updated if stree is
1811 static void gen_bitlen(tree_desc * desc)
1813 ct_data *tree = desc->dyn_tree;
1814 const extra_bits_t *extra = desc->extra_bits;
1815 int base = desc->extra_base;
1816 int max_code = desc->max_code;
1817 int max_length = desc->max_length;
1818 ct_data *stree = desc->static_tree;
1819 int h; /* heap index */
1820 int n, m; /* iterate over the tree elements */
1821 int bits; /* bit length */
1822 int xbits; /* extra bits */
1823 ush f; /* frequency */
1824 int overflow = 0; /* number of elements with bit length too large */
1826 for (bits = 0; bits <= MAX_BITS; bits++)
1829 /* In a first pass, compute the optimal bit lengths (which may
1830 * overflow in the case of the bit length tree).
1832 tree[heap[heap_max]].Len = 0; /* root of the heap */
1834 for (h = heap_max + 1; h < HEAP_SIZE; h++) {
1836 bits = tree[tree[n].Dad].Len + 1;
1837 if (bits > max_length)
1838 bits = max_length, overflow++;
1839 tree[n].Len = (ush) bits;
1840 /* We overwrite tree[n].Dad which is no longer needed */
1843 continue; /* not a leaf node */
1848 xbits = extra[n - base];
1850 opt_len += (ulg) f *(bits + xbits);
1853 static_len += (ulg) f *(stree[n].Len + xbits);
1858 Trace((stderr, "\nbit length overflow\n"));
1859 /* This happens for example on obj2 and pic of the Calgary corpus */
1861 /* Find the first bit length which could increase: */
1863 bits = max_length - 1;
1864 while (bl_count[bits] == 0)
1866 bl_count[bits]--; /* move one leaf down the tree */
1867 bl_count[bits + 1] += 2; /* move one overflow item as its brother */
1868 bl_count[max_length]--;
1869 /* The brother of the overflow item also moves one step up,
1870 * but this does not affect bl_count[max_length]
1873 } while (overflow > 0);
1875 /* Now recompute all bit lengths, scanning in increasing frequency.
1876 * h is still equal to HEAP_SIZE. (It is simpler to reconstruct all
1877 * lengths instead of fixing only the wrong ones. This idea is taken
1878 * from 'ar' written by Haruhiko Okumura.)
1880 for (bits = max_length; bits != 0; bits--) {
1886 if (tree[m].Len != (unsigned) bits) {
1887 Trace((stderr, "code %d bits %d->%d\n", m, tree[m].Len,
1890 ((long) bits - (long) tree[m].Len) * (long) tree[m].Freq;
1891 tree[m].Len = (ush) bits;
1898 /* ===========================================================================
1899 * Generate the codes for a given tree and bit counts (which need not be
1901 * IN assertion: the array bl_count contains the bit length statistics for
1902 * the given tree and the field len is set for all tree elements.
1903 * OUT assertion: the field code is set for all tree elements of non
1906 static void gen_codes(ct_data * tree, int max_code)
1908 ush next_code[MAX_BITS + 1]; /* next code value for each bit length */
1909 ush code = 0; /* running code value */
1910 int bits; /* bit index */
1911 int n; /* code index */
1913 /* The distribution counts are first used to generate the code values
1914 * without bit reversal.
1916 for (bits = 1; bits <= MAX_BITS; bits++) {
1917 next_code[bits] = code = (code + bl_count[bits - 1]) << 1;
1919 /* Check that the bit counts in bl_count are consistent. The last code
1922 Assert(code + bl_count[MAX_BITS] - 1 == (1 << MAX_BITS) - 1,
1923 "inconsistent bit counts");
1924 Tracev((stderr, "\ngen_codes: max_code %d ", max_code));
1926 for (n = 0; n <= max_code; n++) {
1927 int len = tree[n].Len;
1931 /* Now reverse the bits */
1932 tree[n].Code = bi_reverse(next_code[len]++, len);
1934 Tracec(tree != static_ltree,
1935 (stderr, "\nn %3d %c l %2d c %4x (%x) ", n,
1936 (isgraph(n) ? n : ' '), len, tree[n].Code,
1937 next_code[len] - 1));
1941 /* ===========================================================================
1942 * Construct one Huffman tree and assigns the code bit strings and lengths.
1943 * Update the total bit length for the current block.
1944 * IN assertion: the field freq is set for all tree elements.
1945 * OUT assertions: the fields len and code are set to the optimal bit length
1946 * and corresponding code. The length opt_len is updated; static_len is
1947 * also updated if stree is not null. The field max_code is set.
1949 static void build_tree(tree_desc * desc)
1951 ct_data *tree = desc->dyn_tree;
1952 ct_data *stree = desc->static_tree;
1953 int elems = desc->elems;
1954 int n, m; /* iterate over heap elements */
1955 int max_code = -1; /* largest code with non zero frequency */
1956 int node = elems; /* next internal node of the tree */
1958 /* Construct the initial heap, with least frequent element in
1959 * heap[SMALLEST]. The sons of heap[n] are heap[2*n] and heap[2*n+1].
1960 * heap[0] is not used.
1962 heap_len = 0, heap_max = HEAP_SIZE;
1964 for (n = 0; n < elems; n++) {
1965 if (tree[n].Freq != 0) {
1966 heap[++heap_len] = max_code = n;
1973 /* The pkzip format requires that at least one distance code exists,
1974 * and that at least one bit should be sent even if there is only one
1975 * possible code. So to avoid special checks later on we force at least
1976 * two codes of non zero frequency.
1978 while (heap_len < 2) {
1979 int new = heap[++heap_len] = (max_code < 2 ? ++max_code : 0);
1985 static_len -= stree[new].Len;
1986 /* new is 0 or 1 so it does not have extra bits */
1988 desc->max_code = max_code;
1990 /* The elements heap[heap_len/2+1 .. heap_len] are leaves of the tree,
1991 * establish sub-heaps of increasing lengths:
1993 for (n = heap_len / 2; n >= 1; n--)
1994 pqdownheap(tree, n);
1996 /* Construct the Huffman tree by repeatedly combining the least two
2000 pqremove(tree, n); /* n = node of least frequency */
2001 m = heap[SMALLEST]; /* m = node of next least frequency */
2003 heap[--heap_max] = n; /* keep the nodes sorted by frequency */
2004 heap[--heap_max] = m;
2006 /* Create a new node father of n and m */
2007 tree[node].Freq = tree[n].Freq + tree[m].Freq;
2008 depth[node] = (uch) (MAX(depth[n], depth[m]) + 1);
2009 tree[n].Dad = tree[m].Dad = (ush) node;
2011 if (tree == bl_tree) {
2012 bb_error_msg("\nnode %d(%d), sons %d(%d) %d(%d)",
2013 node, tree[node].Freq, n, tree[n].Freq, m, tree[m].Freq);
2016 /* and insert the new node in the heap */
2017 heap[SMALLEST] = node++;
2018 pqdownheap(tree, SMALLEST);
2020 } while (heap_len >= 2);
2022 heap[--heap_max] = heap[SMALLEST];
2024 /* At this point, the fields freq and dad are set. We can now
2025 * generate the bit lengths.
2027 gen_bitlen((tree_desc *) desc);
2029 /* The field len is now set, we can generate the bit codes */
2030 gen_codes((ct_data *) tree, max_code);
2033 /* ===========================================================================
2034 * Scan a literal or distance tree to determine the frequencies of the codes
2035 * in the bit length tree. Updates opt_len to take into account the repeat
2036 * counts. (The contribution of the bit length codes will be added later
2037 * during the construction of bl_tree.)
2039 static void scan_tree(ct_data * tree, int max_code)
2041 int n; /* iterates over all tree elements */
2042 int prevlen = -1; /* last emitted length */
2043 int curlen; /* length of current code */
2044 int nextlen = tree[0].Len; /* length of next code */
2045 int count = 0; /* repeat count of the current code */
2046 int max_count = 7; /* max repeat count */
2047 int min_count = 4; /* min repeat count */
2050 max_count = 138, min_count = 3;
2051 tree[max_code + 1].Len = (ush) 0xffff; /* guard */
2053 for (n = 0; n <= max_code; n++) {
2055 nextlen = tree[n + 1].Len;
2056 if (++count < max_count && curlen == nextlen) {
2058 } else if (count < min_count) {
2059 bl_tree[curlen].Freq += count;
2060 } else if (curlen != 0) {
2061 if (curlen != prevlen)
2062 bl_tree[curlen].Freq++;
2063 bl_tree[REP_3_6].Freq++;
2064 } else if (count <= 10) {
2065 bl_tree[REPZ_3_10].Freq++;
2067 bl_tree[REPZ_11_138].Freq++;
2072 max_count = 138, min_count = 3;
2073 } else if (curlen == nextlen) {
2074 max_count = 6, min_count = 3;
2076 max_count = 7, min_count = 4;
2081 /* ===========================================================================
2082 * Send a literal or distance tree in compressed form, using the codes in
2085 static void send_tree(ct_data * tree, int max_code)
2087 int n; /* iterates over all tree elements */
2088 int prevlen = -1; /* last emitted length */
2089 int curlen; /* length of current code */
2090 int nextlen = tree[0].Len; /* length of next code */
2091 int count = 0; /* repeat count of the current code */
2092 int max_count = 7; /* max repeat count */
2093 int min_count = 4; /* min repeat count */
2095 /* tree[max_code+1].Len = -1; *//* guard already set */
2097 max_count = 138, min_count = 3;
2099 for (n = 0; n <= max_code; n++) {
2101 nextlen = tree[n + 1].Len;
2102 if (++count < max_count && curlen == nextlen) {
2104 } else if (count < min_count) {
2106 send_code(curlen, bl_tree);
2107 } while (--count != 0);
2109 } else if (curlen != 0) {
2110 if (curlen != prevlen) {
2111 send_code(curlen, bl_tree);
2114 Assert(count >= 3 && count <= 6, " 3_6?");
2115 send_code(REP_3_6, bl_tree);
2116 send_bits(count - 3, 2);
2118 } else if (count <= 10) {
2119 send_code(REPZ_3_10, bl_tree);
2120 send_bits(count - 3, 3);
2123 send_code(REPZ_11_138, bl_tree);
2124 send_bits(count - 11, 7);
2129 max_count = 138, min_count = 3;
2130 } else if (curlen == nextlen) {
2131 max_count = 6, min_count = 3;
2133 max_count = 7, min_count = 4;
2138 /* ===========================================================================
2139 * Construct the Huffman tree for the bit lengths and return the index in
2140 * bl_order of the last bit length code to send.
2142 static int build_bl_tree(void)
2144 int max_blindex; /* index of last bit length code of non zero freq */
2146 /* Determine the bit length frequencies for literal and distance trees */
2147 scan_tree((ct_data *) dyn_ltree, l_desc.max_code);
2148 scan_tree((ct_data *) dyn_dtree, d_desc.max_code);
2150 /* Build the bit length tree: */
2151 build_tree((tree_desc *) (&bl_desc));
2152 /* opt_len now includes the length of the tree representations, except
2153 * the lengths of the bit lengths codes and the 5+5+4 bits for the counts.
2156 /* Determine the number of bit length codes to send. The pkzip format
2157 * requires that at least 4 bit length codes be sent. (appnote.txt says
2158 * 3 but the actual value used is 4.)
2160 for (max_blindex = BL_CODES - 1; max_blindex >= 3; max_blindex--) {
2161 if (bl_tree[bl_order[max_blindex]].Len != 0)
2164 /* Update opt_len to include the bit length tree and counts */
2165 opt_len += 3 * (max_blindex + 1) + 5 + 5 + 4;
2166 Tracev((stderr, "\ndyn trees: dyn %ld, stat %ld", opt_len, static_len));
2171 /* ===========================================================================
2172 * Send the header for a block using dynamic Huffman trees: the counts, the
2173 * lengths of the bit length codes, the literal tree and the distance tree.
2174 * IN assertion: lcodes >= 257, dcodes >= 1, blcodes >= 4.
2176 static void send_all_trees(int lcodes, int dcodes, int blcodes)
2178 int rank; /* index in bl_order */
2180 Assert(lcodes >= 257 && dcodes >= 1 && blcodes >= 4, "not enough codes");
2181 Assert(lcodes <= L_CODES && dcodes <= D_CODES
2182 && blcodes <= BL_CODES, "too many codes");
2183 Tracev((stderr, "\nbl counts: "));
2184 send_bits(lcodes - 257, 5); /* not +255 as stated in appnote.txt */
2185 send_bits(dcodes - 1, 5);
2186 send_bits(blcodes - 4, 4); /* not -3 as stated in appnote.txt */
2187 for (rank = 0; rank < blcodes; rank++) {
2188 Tracev((stderr, "\nbl code %2d ", bl_order[rank]));
2189 send_bits(bl_tree[bl_order[rank]].Len, 3);
2191 Tracev((stderr, "\nbl tree: sent %ld", bits_sent));
2193 send_tree((ct_data *) dyn_ltree, lcodes - 1); /* send the literal tree */
2194 Tracev((stderr, "\nlit tree: sent %ld", bits_sent));
2196 send_tree((ct_data *) dyn_dtree, dcodes - 1); /* send the distance tree */
2197 Tracev((stderr, "\ndist tree: sent %ld", bits_sent));
2200 /* ===========================================================================
2201 * Determine the best encoding for the current block: dynamic trees, static
2202 * trees or store, and output the encoded block to the zip file. This function
2203 * returns the total compressed length for the file so far.
2205 static ulg flush_block(char *buf, ulg stored_len, int eof)
2207 ulg opt_lenb, static_lenb; /* opt_len and static_len in bytes */
2208 int max_blindex; /* index of last bit length code of non zero freq */
2210 flag_buf[last_flags] = flags; /* Save the flags for the last 8 items */
2212 /* Check if the file is ascii or binary */
2213 if (*file_type == (ush) UNKNOWN)
2216 /* Construct the literal and distance trees */
2217 build_tree((tree_desc *) (&l_desc));
2218 Tracev((stderr, "\nlit data: dyn %ld, stat %ld", opt_len, static_len));
2220 build_tree((tree_desc *) (&d_desc));
2221 Tracev((stderr, "\ndist data: dyn %ld, stat %ld", opt_len, static_len));
2222 /* At this point, opt_len and static_len are the total bit lengths of
2223 * the compressed block data, excluding the tree representations.
2226 /* Build the bit length tree for the above two trees, and get the index
2227 * in bl_order of the last bit length code to send.
2229 max_blindex = build_bl_tree();
2231 /* Determine the best encoding. Compute first the block length in bytes */
2232 opt_lenb = (opt_len + 3 + 7) >> 3;
2233 static_lenb = (static_len + 3 + 7) >> 3;
2236 "\nopt %lu(%lu) stat %lu(%lu) stored %lu lit %u dist %u ",
2237 opt_lenb, opt_len, static_lenb, static_len, stored_len,
2238 last_lit, last_dist));
2240 if (static_lenb <= opt_lenb)
2241 opt_lenb = static_lenb;
2243 /* If compression failed and this is the first and last block,
2244 * and if the zip file can be seeked (to rewrite the local header),
2245 * the whole file is transformed into a stored file:
2247 if (stored_len <= opt_lenb && eof && compressed_len == 0L && seekable()) {
2248 /* Since LIT_BUFSIZE <= 2*WSIZE, the input data must be there: */
2249 if (buf == (char *) 0)
2250 bb_error_msg("block vanished");
2252 copy_block(buf, (unsigned) stored_len, 0); /* without header */
2253 compressed_len = stored_len << 3;
2254 *file_method = STORED;
2256 } else if (stored_len + 4 <= opt_lenb && buf != (char *) 0) {
2257 /* 4: two words for the lengths */
2258 /* The test buf != NULL is only necessary if LIT_BUFSIZE > WSIZE.
2259 * Otherwise we can't have processed more than WSIZE input bytes since
2260 * the last block flush, because compression would have been
2261 * successful. If LIT_BUFSIZE <= WSIZE, it is never too late to
2262 * transform a block into a stored block.
2264 send_bits((STORED_BLOCK << 1) + eof, 3); /* send block type */
2265 compressed_len = (compressed_len + 3 + 7) & ~7L;
2266 compressed_len += (stored_len + 4) << 3;
2268 copy_block(buf, (unsigned) stored_len, 1); /* with header */
2270 } else if (static_lenb == opt_lenb) {
2271 send_bits((STATIC_TREES << 1) + eof, 3);
2272 compress_block((ct_data *) static_ltree, (ct_data *) static_dtree);
2273 compressed_len += 3 + static_len;
2275 send_bits((DYN_TREES << 1) + eof, 3);
2276 send_all_trees(l_desc.max_code + 1, d_desc.max_code + 1,
2278 compress_block((ct_data *) dyn_ltree, (ct_data *) dyn_dtree);
2279 compressed_len += 3 + opt_len;
2281 Assert(compressed_len == bits_sent, "bad compressed size");
2286 compressed_len += 7; /* align on byte boundary */
2288 Tracev((stderr, "\ncomprlen %lu(%lu) ", compressed_len >> 3,
2289 compressed_len - 7 * eof));
2291 return compressed_len >> 3;
2294 /* ===========================================================================
2295 * Save the match info and tally the frequency counts. Return true if
2296 * the current block must be flushed.
2298 static int ct_tally(int dist, int lc)
2300 l_buf[last_lit++] = (uch) lc;
2302 /* lc is the unmatched char */
2303 dyn_ltree[lc].Freq++;
2305 /* Here, lc is the match length - MIN_MATCH */
2306 dist--; /* dist = match distance - 1 */
2307 Assert((ush) dist < (ush) MAX_DIST &&
2308 (ush) lc <= (ush) (MAX_MATCH - MIN_MATCH) &&
2309 (ush) d_code(dist) < (ush) D_CODES, "ct_tally: bad match");
2311 dyn_ltree[length_code[lc] + LITERALS + 1].Freq++;
2312 dyn_dtree[d_code(dist)].Freq++;
2314 d_buf[last_dist++] = (ush) dist;
2319 /* Output the flags if they fill a byte: */
2320 if ((last_lit & 7) == 0) {
2321 flag_buf[last_flags++] = flags;
2322 flags = 0, flag_bit = 1;
2324 /* Try to guess if it is profitable to stop the current block here */
2325 if ((last_lit & 0xfff) == 0) {
2326 /* Compute an upper bound for the compressed length */
2327 ulg out_length = (ulg) last_lit * 8L;
2328 ulg in_length = (ulg) strstart - block_start;
2331 for (dcode = 0; dcode < D_CODES; dcode++) {
2333 (ulg) dyn_dtree[dcode].Freq * (5L + extra_dbits[dcode]);
2337 "\nlast_lit %u, last_dist %u, in %ld, out ~%ld(%ld%%) ",
2338 last_lit, last_dist, in_length, out_length,
2339 100L - out_length * 100L / in_length));
2340 if (last_dist < last_lit / 2 && out_length < in_length / 2)
2343 return (last_lit == LIT_BUFSIZE - 1 || last_dist == DIST_BUFSIZE);
2344 /* We avoid equality with LIT_BUFSIZE because of wraparound at 64K
2345 * on 16 bit machines and because stored blocks are restricted to
2350 /* ===========================================================================
2351 * Send the block data compressed using the given Huffman trees
2353 static void compress_block(ct_data * ltree, ct_data * dtree)
2355 unsigned dist; /* distance of matched string */
2356 int lc; /* match length or unmatched char (if dist == 0) */
2357 unsigned lx = 0; /* running index in l_buf */
2358 unsigned dx = 0; /* running index in d_buf */
2359 unsigned fx = 0; /* running index in flag_buf */
2360 uch flag = 0; /* current flags */
2361 unsigned code; /* the code to send */
2362 int extra; /* number of extra bits to send */
2367 flag = flag_buf[fx++];
2369 if ((flag & 1) == 0) {
2370 send_code(lc, ltree); /* send a literal byte */
2371 Tracecv(isgraph(lc), (stderr, " '%c' ", lc));
2373 /* Here, lc is the match length - MIN_MATCH */
2374 code = length_code[lc];
2375 send_code(code + LITERALS + 1, ltree); /* send the length code */
2376 extra = extra_lbits[code];
2378 lc -= base_length[code];
2379 send_bits(lc, extra); /* send the extra length bits */
2382 /* Here, dist is the match distance - 1 */
2383 code = d_code(dist);
2384 Assert(code < D_CODES, "bad d_code");
2386 send_code(code, dtree); /* send the distance code */
2387 extra = extra_dbits[code];
2389 dist -= base_dist[code];
2390 send_bits(dist, extra); /* send the extra distance bits */
2392 } /* literal or match pair ? */
2394 } while (lx < last_lit);
2396 send_code(END_BLOCK, ltree);
2399 /* ===========================================================================
2400 * Set the file type to ASCII or BINARY, using a crude approximation:
2401 * binary if more than 20% of the bytes are <= 6 or >= 128, ascii otherwise.
2402 * IN assertion: the fields freq of dyn_ltree are set and the total of all
2403 * frequencies does not exceed 64K (to fit in an int on 16 bit machines).
2405 static void set_file_type(void)
2408 unsigned ascii_freq = 0;
2409 unsigned bin_freq = 0;
2412 bin_freq += dyn_ltree[n++].Freq;
2414 ascii_freq += dyn_ltree[n++].Freq;
2415 while (n < LITERALS)
2416 bin_freq += dyn_ltree[n++].Freq;
2417 *file_type = bin_freq > (ascii_freq >> 2) ? BINARY : ASCII;
2418 if (*file_type == BINARY && translate_eol) {
2419 bb_error_msg("-l used on binary file");
2423 /* zip.c -- compress files to the gzip or pkzip format
2424 * Copyright (C) 1992-1993 Jean-loup Gailly
2425 * This is free software; you can redistribute it and/or modify it under the
2426 * terms of the GNU General Public License, see the file COPYING.
2430 static ulg crc; /* crc on uncompressed file data */
2431 static long header_bytes; /* number of bytes in gzip header */
2433 static void put_long(ulg n)
2435 put_short((n) & 0xffff);
2436 put_short(((ulg) (n)) >> 16);
2439 /* put_header_byte is used for the compressed output
2440 * - for the initial 4 bytes that can't overflow the buffer.
2442 #define put_header_byte(c) {outbuf[outcnt++]=(uch)(c);}
2444 /* ===========================================================================
2445 * Deflate in to out.
2446 * IN assertions: the input and output buffers are cleared.
2447 * The variables time_stamp and save_orig_name are initialized.
2449 static int zip(int in, int out)
2451 uch my_flags = 0; /* general purpose bit flags */
2452 ush attr = 0; /* ascii/binary flag */
2453 ush deflate_flags = 0; /* pkzip -es, -en or -ex equivalent */
2459 /* Write the header to the gzip file. See algorithm.doc for the format */
2463 put_header_byte(GZIP_MAGIC[0]); /* magic header */
2464 put_header_byte(GZIP_MAGIC[1]);
2465 put_header_byte(DEFLATED); /* compression method */
2467 put_header_byte(my_flags); /* general flags */
2468 put_long(time_stamp);
2470 /* Write deflated file to zip file */
2474 ct_init(&attr, &method);
2475 lm_init(&deflate_flags);
2477 put_byte((uch) deflate_flags); /* extra flags */
2478 put_byte(OS_CODE); /* OS identifier */
2480 header_bytes = (long) outcnt;
2484 /* Write the crc and uncompressed size */
2487 header_bytes += 2 * sizeof(long);
2494 /* ===========================================================================
2495 * Read a new buffer from the current input file, perform end-of-line
2496 * translation, and update the crc and input file size.
2497 * IN assertion: size >= 2 (for end-of-line translation)
2499 static int file_read(char *buf, unsigned size)
2503 Assert(insize == 0, "inbuf not empty");
2505 len = read(ifd, buf, size);
2506 if (len == (unsigned) (-1) || len == 0)
2509 crc = updcrc((uch *) buf, len);
2514 /* ===========================================================================
2515 * Write the output buffer outbuf[0..outcnt-1] and update bytes_out.
2516 * (used for the compressed data only)
2518 static void flush_outbuf(void)
2523 write_buf(ofd, (char *) outbuf, outcnt);