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 * Licensed under GPLv2 or later, see file LICENSE in this tarball for details.
19 /* TODO: full support for -v for DESKTOP
20 /usr/bin/gzip -v a bogus aa
21 a: 85.1% -- replaced with a.gz
22 gzip: bogus: No such file or directory
23 aa: 85.1% -- replaced with aa.gz
31 /* Compression methods (see algorithm.doc) */
32 /* Only STORED and DEFLATED are supported by this BusyBox module */
34 /* methods 4 to 7 reserved */
39 # define INBUFSIZ 0x2000 /* input buffer size */
41 # define INBUFSIZ 0x8000 /* input buffer size */
45 #define INBUF_EXTRA 64 /* required by unlzw() */
49 # define OUTBUFSIZ 8192 /* output buffer size */
51 # define OUTBUFSIZ 16384 /* output buffer size */
54 #define OUTBUF_EXTRA 2048 /* required by unlzw() */
58 # define DIST_BUFSIZE 0x2000 /* buffer for distances, see trees.c */
60 # define DIST_BUFSIZE 0x8000 /* buffer for distances, see trees.c */
64 #define NO_FILE (-1) /* in memory compression */
66 #define PACK_MAGIC "\037\036" /* Magic header for packed files */
67 #define GZIP_MAGIC "\037\213" /* Magic header for gzip files, 1F 8B */
68 #define OLD_GZIP_MAGIC "\037\236" /* Magic header for gzip 0.5 = freeze 1.x */
69 #define LZH_MAGIC "\037\240" /* Magic header for SCO LZH Compress files */
70 #define PKZIP_MAGIC "\120\113\003\004" /* Magic header for pkzip files */
73 #define ASCII_FLAG 0x01 /* bit 0 set: file probably ascii text */
74 #define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */
75 #define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
76 #define ORIG_NAME 0x08 /* bit 3 set: original file name present */
77 #define COMMENT 0x10 /* bit 4 set: file comment present */
78 #define RESERVED 0xC0 /* bit 6,7: reserved */
80 /* internal file attribute */
81 #define UNKNOWN 0xffff
86 # define WSIZE 0x8000 /* window size--must be a power of two, and */
87 #endif /* at least 32K for zip's deflate method */
91 /* The minimum and maximum match lengths */
93 #define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
94 /* Minimum amount of lookahead, except at the end of the input file.
95 * See deflate.c for comments about the MIN_MATCH+1.
98 #define MAX_DIST (WSIZE-MIN_LOOKAHEAD)
99 /* In order to simplify the code, particularly on 16 bit machines, match
100 * distances are limited to MAX_DIST instead of WSIZE.
104 # define MAX_PATH_LEN 1024 /* max pathname length */
107 #define seekable() 0 /* force sequential output */
108 #define translate_eol 0 /* no option -a yet */
113 #define INIT_BITS 9 /* Initial number of bits per code */
115 #define BIT_MASK 0x1f /* Mask for 'number of compression bits' */
116 /* Mask 0x20 is reserved to mean a fourth header byte, and 0x40 is free.
117 * It's a pity that old uncompress does not check bit 0x20. That makes
118 * extension of the format actually undesirable because old compress
119 * would just crash on the new format instead of giving a meaningful
120 * error message. It does check the number of bits, but it's more
121 * helpful to say "unsupported format, get a new version" than
122 * "can only handle 16 bits".
126 # define MAX_SUFFIX MAX_EXT_CHARS
128 # define MAX_SUFFIX 30
132 #define DECLARE(type, array, size)\
134 #define ALLOC(type, array, size) { \
135 array = xzalloc((size_t)(((size)+1L)/2) * 2*sizeof(type)); \
137 #define FREE(array) { \
142 /* Diagnostic functions */
144 # define Assert(cond,msg) {if(!(cond)) bb_error_msg(msg);}
145 # define Trace(x) fprintf x
146 # define Tracev(x) {if (verbose) fprintf x ;}
147 # define Tracevv(x) {if (verbose>1) fprintf x ;}
148 # define Tracec(c,x) {if (verbose && (c)) fprintf x ;}
149 # define Tracecv(c,x) {if (verbose>1 && (c)) fprintf x ;}
151 # define Assert(cond,msg)
156 # define Tracecv(c,x)
159 typedef unsigned char uch;
160 typedef unsigned short ush;
161 typedef unsigned long ulg;
165 static int zip(int in, int out);
166 static int file_read(char *buf, unsigned size);
169 static void lm_init(ush * flags);
170 static ulg deflate(void);
173 static void ct_init(ush * attr, int *methodp);
174 static int ct_tally(int dist, int lc);
175 static ulg flush_block(char *buf, ulg stored_len, int eof);
178 static void bi_init(int zipfile);
179 static void send_bits(int value, int length);
180 static unsigned bi_reverse(unsigned value, int length);
181 static void bi_windup(void);
182 static void copy_block(char *buf, unsigned len, int header);
183 static int (*read_buf) (char *buf, unsigned size);
185 static void flush_outbuf(void);
189 /* To save memory for 16 bit systems, some arrays are overlaid between
190 * the various modules:
191 * deflate: prev+head window d_buf l_buf outbuf
192 * unlzw: tab_prefix tab_suffix stack inbuf outbuf
193 * For compression, input is done in window[]. For decompression, output
194 * is done in window except for unlzw.
198 #define tab_suffix window
199 #define tab_prefix prev /* hash link (see deflate.c) */
200 #define head (prev+WSIZE) /* hash head (see deflate.c) */
202 DECLARE(uch, inbuf, INBUFSIZ + INBUF_EXTRA);
203 DECLARE(uch, outbuf, OUTBUFSIZ + OUTBUF_EXTRA);
204 DECLARE(ush, d_buf, DIST_BUFSIZE);
205 DECLARE(uch, window, 2L * WSIZE);
206 DECLARE(ush, tab_prefix, 1L << BITS);
208 static long isize; /* number of input bytes */
210 static int foreground; /* set if program run in foreground */
211 static int method = DEFLATED; /* compression method */
212 static int exit_code; /* program exit code */
213 static long time_stamp; /* original time stamp (modification time) */
214 static char z_suffix[MAX_SUFFIX + 1]; /* default suffix (can be set with --suffix) */
216 static int ifd; /* input file descriptor */
217 static int ofd; /* output file descriptor */
219 static unsigned insize; /* valid bytes in inbuf */
221 static unsigned outcnt; /* bytes in output buffer */
223 static uint32_t *crc_32_tab;
226 /* ===========================================================================
227 * Local data used by the "bit string" routines.
230 static int zfile; /* output gzip file */
232 static unsigned short bi_buf;
234 /* Output buffer. bits are inserted starting at the bottom (least significant
239 #define BUF_SIZE (8 * sizeof(bi_buf))
240 /* Number of bits used within bi_buf. (bi_buf might be implemented on
241 * more than 16 bits on some systems.)
246 /* Current input function. Set to mem_read for in-memory compression */
249 static ulg bits_sent; /* bit length of the compressed data */
253 /* ===========================================================================
255 /* put_8bit is used for the compressed output */
256 #define put_8bit(c) \
258 outbuf[outcnt++] = (c); \
259 if (outcnt == OUTBUFSIZ) flush_outbuf(); \
262 /* Output a 16 bit value, lsb first */
263 static void put_16bit(ush w)
265 if (outcnt < OUTBUFSIZ - 2) {
266 outbuf[outcnt++] = w;
267 outbuf[outcnt++] = w >> 8;
274 static void put_32bit(ulg n)
280 /* put_header_byte is used for the compressed output
281 * - for the initial 4 bytes that can't overflow the buffer.
283 #define put_header_byte(c) \
285 outbuf[outcnt++] = (c); \
288 /* ===========================================================================
289 * Clear input and output buffers
291 static void clear_bufs(void)
300 /* ===========================================================================
301 * Does the same as write(), but also handles partial pipe writes and checks
304 static void write_buf(int fd, void *buf, unsigned cnt)
308 while ((n = write(fd, buf, cnt)) != cnt) {
309 if (n == (unsigned) (-1))
310 bb_error_msg_and_die(bb_msg_write_error);
312 buf = (void *) ((char *) buf + n);
316 /* ===========================================================================
317 * Run a set of bytes through the crc shift register. If s is a NULL
318 * pointer, then initialize the crc shift register contents instead.
319 * Return the current crc in either case.
321 static uint32_t updcrc(uch * s, unsigned n)
323 static uint32_t crc = ~0; /* shift register contents */
324 uint32_t c; /* temporary variable */
332 c = crc_32_tab[((int) c ^ (*s++)) & 0xff] ^ (c >> 8);
339 /* bits.c -- output variable-length bit strings
340 * Copyright (C) 1992-1993 Jean-loup Gailly
341 * This is free software; you can redistribute it and/or modify it under the
342 * terms of the GNU General Public License, see the file COPYING.
349 * Output variable-length bit strings. Compression can be done
350 * to a file or to memory. (The latter is not supported in this version.)
354 * The PKZIP "deflate" file format interprets compressed file data
355 * as a sequence of bits. Multi-bit strings in the file may cross
356 * byte boundaries without restriction.
358 * The first bit of each byte is the low-order bit.
360 * The routines in this file allow a variable-length bit value to
361 * be output right-to-left (useful for literal values). For
362 * left-to-right output (useful for code strings from the tree routines),
363 * the bits must have been reversed first with bi_reverse().
365 * For in-memory compression, the compressed bit stream goes directly
366 * into the requested output buffer. The input data is read in blocks
367 * by the mem_read() function. The buffer is limited to 64K on 16 bit
372 * void bi_init (FILE *zipfile)
373 * Initialize the bit string routines.
375 * void send_bits (int value, int length)
376 * Write out a bit string, taking the source bits right to
379 * int bi_reverse (int value, int length)
380 * Reverse the bits of a bit string, taking the source bits left to
381 * right and emitting them right to left.
383 * void bi_windup (void)
384 * Write out any remaining bits in an incomplete byte.
386 * void copy_block(char *buf, unsigned len, int header)
387 * Copy a stored block to the zip file, storing first the length and
388 * its one's complement if requested.
392 /* ===========================================================================
393 * Initialize the bit string routines.
395 static void bi_init(int zipfile)
404 /* Set the defaults for file compression. They are set by memcompress
405 * for in-memory compression.
407 if (zfile != NO_FILE) {
408 read_buf = file_read;
412 /* ===========================================================================
413 * Send a value on a given number of bits.
414 * IN assertion: length <= 16 and value fits in length bits.
416 static void send_bits(int value, int length)
419 Tracev((stderr, " l %2d v %4x ", length, value));
420 Assert(length > 0 && length <= 15, "invalid length");
421 bits_sent += (ulg) length;
423 /* If not enough room in bi_buf, use (valid) bits from bi_buf and
424 * (16 - bi_valid) bits from value, leaving (width - (16-bi_valid))
425 * unused bits in value.
427 if (bi_valid > (int) BUF_SIZE - length) {
428 bi_buf |= (value << bi_valid);
430 bi_buf = (ush) value >> (BUF_SIZE - bi_valid);
431 bi_valid += length - BUF_SIZE;
433 bi_buf |= value << bi_valid;
438 /* ===========================================================================
439 * Reverse the first len bits of a code, using straightforward code (a faster
440 * method would use a table)
441 * IN assertion: 1 <= len <= 15
443 static unsigned bi_reverse(unsigned code, int len)
449 code >>= 1, res <<= 1;
454 /* ===========================================================================
455 * Write out any remaining bits in an incomplete byte.
457 static void bi_windup(void)
461 } else if (bi_valid > 0) {
467 bits_sent = (bits_sent + 7) & ~7;
471 /* ===========================================================================
472 * Copy a stored block to the zip file, storing first the length and its
473 * one's complement if requested.
475 static void copy_block(char *buf, unsigned len, int header)
477 bi_windup(); /* align on byte boundary */
480 put_16bit((ush) len);
481 put_16bit((ush) ~ len);
487 bits_sent += (ulg) len << 3;
494 /* deflate.c -- compress data using the deflation algorithm
495 * Copyright (C) 1992-1993 Jean-loup Gailly
496 * This is free software; you can redistribute it and/or modify it under the
497 * terms of the GNU General Public License, see the file COPYING.
503 * Identify new text as repetitions of old text within a fixed-
504 * length sliding window trailing behind the new text.
508 * The "deflation" process depends on being able to identify portions
509 * of the input text which are identical to earlier input (within a
510 * sliding window trailing behind the input currently being processed).
512 * The most straightforward technique turns out to be the fastest for
513 * most input files: try all possible matches and select the longest.
514 * The key feature of this algorithm is that insertions into the string
515 * dictionary are very simple and thus fast, and deletions are avoided
516 * completely. Insertions are performed at each input character, whereas
517 * string matches are performed only when the previous match ends. So it
518 * is preferable to spend more time in matches to allow very fast string
519 * insertions and avoid deletions. The matching algorithm for small
520 * strings is inspired from that of Rabin & Karp. A brute force approach
521 * is used to find longer strings when a small match has been found.
522 * A similar algorithm is used in comic (by Jan-Mark Wams) and freeze
523 * (by Leonid Broukhis).
524 * A previous version of this file used a more sophisticated algorithm
525 * (by Fiala and Greene) which is guaranteed to run in linear amortized
526 * time, but has a larger average cost, uses more memory and is patented.
527 * However the F&G algorithm may be faster for some highly redundant
528 * files if the parameter max_chain_length (described below) is too large.
532 * The idea of lazy evaluation of matches is due to Jan-Mark Wams, and
533 * I found it in 'freeze' written by Leonid Broukhis.
534 * Thanks to many info-zippers for bug reports and testing.
538 * APPNOTE.TXT documentation file in PKZIP 1.93a distribution.
540 * A description of the Rabin and Karp algorithm is given in the book
541 * "Algorithms" by R. Sedgewick, Addison-Wesley, p252.
543 * Fiala,E.R., and Greene,D.H.
544 * Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595
548 * void lm_init (int pack_level, ush *flags)
549 * Initialize the "longest match" routines for a new file
552 * Processes a new input file and return its compressed length. Sets
553 * the compressed length, crc, deflate flags and internal file
558 /* ===========================================================================
559 * Configuration parameters
562 /* Compile with MEDIUM_MEM to reduce the memory requirements or
563 * with SMALL_MEM to use as little memory as possible. Use BIG_MEM if the
564 * entire input file can be held in memory (not possible on 16 bit systems).
565 * Warning: defining these symbols affects HASH_BITS (see below) and thus
566 * affects the compression ratio. The compressed output
567 * is still correct, and might even be smaller in some cases.
571 # define HASH_BITS 13 /* Number of bits used to hash strings */
574 # define HASH_BITS 14
577 # define HASH_BITS 15
578 /* For portability to 16 bit machines, do not use values above 15. */
581 /* To save space (see unlzw.c), we overlay prev+head with tab_prefix and
582 * window with tab_suffix. Check that we can do this:
584 #if (WSIZE<<1) > (1<<BITS)
585 # error cannot overlay window with tab_suffix and prev with tab_prefix0
587 #if HASH_BITS > BITS-1
588 # error cannot overlay head with tab_prefix1
590 #define HASH_SIZE (unsigned)(1<<HASH_BITS)
591 #define HASH_MASK (HASH_SIZE-1)
592 #define WMASK (WSIZE-1)
593 /* HASH_SIZE and WSIZE must be powers of two */
595 /* Tail of hash chains */
598 /* speed options for the general purpose bit flag */
600 # define TOO_FAR 4096
602 /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
603 /* ===========================================================================
604 * Local data used by the "longest match" routines.
607 typedef unsigned IPos;
609 /* A Pos is an index in the character window. We use short instead of int to
610 * save space in the various tables. IPos is used only for parameter passing.
613 /* DECLARE(uch, window, 2L*WSIZE); */
614 /* Sliding window. Input bytes are read into the second half of the window,
615 * and move to the first half later to keep a dictionary of at least WSIZE
616 * bytes. With this organization, matches are limited to a distance of
617 * WSIZE-MAX_MATCH bytes, but this ensures that IO is always
618 * performed with a length multiple of the block size. Also, it limits
619 * the window size to 64K, which is quite useful on MSDOS.
620 * To do: limit the window size to WSIZE+BSZ if SMALL_MEM (the code would
621 * be less efficient).
624 /* DECLARE(Pos, prev, WSIZE); */
625 /* Link to older string with same hash index. To limit the size of this
626 * array to 64K, this link is maintained only for the last 32K strings.
627 * An index in this array is thus a window index modulo 32K.
630 /* DECLARE(Pos, head, 1<<HASH_BITS); */
631 /* Heads of the hash chains or NIL. */
633 static const ulg window_size = (ulg) 2 * WSIZE;
635 /* window size, 2*WSIZE except for MMAP or BIG_MEM, where it is the
636 * input file length plus MIN_LOOKAHEAD.
639 static long block_start;
641 /* window position at the beginning of the current output block. Gets
642 * negative when the window is moved backwards.
645 static unsigned ins_h; /* hash index of string to be inserted */
647 #define H_SHIFT ((HASH_BITS+MIN_MATCH-1)/MIN_MATCH)
648 /* Number of bits by which ins_h and del_h must be shifted at each
649 * input step. It must be such that after MIN_MATCH steps, the oldest
650 * byte no longer takes part in the hash key, that is:
651 * H_SHIFT * MIN_MATCH >= HASH_BITS
654 static unsigned int prev_length;
656 /* Length of the best match at previous step. Matches not greater than this
657 * are discarded. This is used in the lazy match evaluation.
660 static unsigned strstart; /* start of string to insert */
661 static unsigned match_start; /* start of matching string */
662 static int eofile; /* flag set at end of input file */
663 static unsigned lookahead; /* number of valid bytes ahead in window */
666 max_chain_length = 4096,
668 /* To speed up deflation, hash chains are never searched beyond this length.
669 * A higher limit improves compression ratio but degrades the speed.
672 max_lazy_match = 258,
674 /* Attempt to find a better match only when the current match is strictly
675 * smaller than this value. This mechanism is used only for compression
678 max_insert_length = max_lazy_match,
679 /* Insert new strings in the hash table only if the match length
680 * is not greater than this length. This saves time but degrades compression.
681 * max_insert_length is used only for compression levels <= 3.
686 /* Use a faster search when the previous match is longer than this */
689 /* Values for max_lazy_match, good_match and max_chain_length, depending on
690 * the desired pack level (0..9). The values given below have been tuned to
691 * exclude worst case performance for pathological files. Better values may be
692 * found for specific files.
695 nice_match = 258 /* Stop searching when current match exceeds this */
697 /* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
698 * For deflate_fast() (levels <= 3) good is ignored and lazy has a different
704 /* result of memcmp for equal strings */
706 /* ===========================================================================
707 * Prototypes for local functions.
709 static void fill_window(void);
711 static int longest_match(IPos cur_match);
714 static void check_match(IPos start, IPos match, int length);
717 /* ===========================================================================
718 * Update a hash value with the given input byte
719 * IN assertion: all calls to to UPDATE_HASH are made with consecutive
720 * input characters, so that a running hash key can be computed from the
721 * previous key instead of complete recalculation each time.
723 #define UPDATE_HASH(h,c) (h = (((h)<<H_SHIFT) ^ (c)) & HASH_MASK)
725 /* ===========================================================================
726 * Insert string s in the dictionary and set match_head to the previous head
727 * of the hash chain (the most recent string with same hash key). Return
728 * the previous length of the hash chain.
729 * IN assertion: all calls to to INSERT_STRING are made with consecutive
730 * input characters and the first MIN_MATCH bytes of s are valid
731 * (except for the last MIN_MATCH-1 bytes of the input file).
733 #define INSERT_STRING(s, match_head) \
734 (UPDATE_HASH(ins_h, window[(s) + MIN_MATCH-1]), \
735 prev[(s) & WMASK] = match_head = head[ins_h], \
738 /* ===========================================================================
739 * Initialize the "longest match" routines for a new file
741 static void lm_init(ush * flags)
745 /* Initialize the hash table. */
746 memset(head, 0, HASH_SIZE * sizeof(*head));
747 /* prev will be initialized on the fly */
750 /* ??? reduce max_chain_length for binary files */
755 lookahead = read_buf((char *) window,
756 sizeof(int) <= 2 ? (unsigned) WSIZE : 2 * WSIZE);
758 if (lookahead == 0 || lookahead == (unsigned) EOF) {
759 eofile = 1, lookahead = 0;
763 /* Make sure that we always have enough lookahead. This is important
764 * if input comes from a device such as a tty.
766 while (lookahead < MIN_LOOKAHEAD && !eofile)
770 for (j = 0; j < MIN_MATCH - 1; j++)
771 UPDATE_HASH(ins_h, window[j]);
772 /* If lookahead < MIN_MATCH, ins_h is garbage, but this is
773 * not important since only literal bytes will be emitted.
777 /* ===========================================================================
778 * Set match_start to the longest match starting at the given string and
779 * return its length. Matches shorter or equal to prev_length are discarded,
780 * in which case the result is equal to prev_length and match_start is
782 * IN assertions: cur_match is the head of the hash chain for the current
783 * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
786 /* For MSDOS, OS/2 and 386 Unix, an optimized version is in match.asm or
787 * match.s. The code is functionally equivalent, so you can use the C version
790 static int longest_match(IPos cur_match)
792 unsigned chain_length = max_chain_length; /* max hash chain length */
793 uch *scan = window + strstart; /* current string */
794 uch *match; /* matched string */
795 int len; /* length of current match */
796 int best_len = prev_length; /* best match length so far */
798 strstart > (IPos) MAX_DIST ? strstart - (IPos) MAX_DIST : NIL;
799 /* Stop when cur_match becomes <= limit. To simplify the code,
800 * we prevent matches with the string of window index 0.
803 /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
804 * It is easy to get rid of this optimization if necessary.
806 #if HASH_BITS < 8 || MAX_MATCH != 258
807 # error Code too clever
809 uch *strend = window + strstart + MAX_MATCH;
810 uch scan_end1 = scan[best_len - 1];
811 uch scan_end = scan[best_len];
813 /* Do not waste too much time if we already have a good match: */
814 if (prev_length >= good_match) {
817 Assert(strstart <= window_size - MIN_LOOKAHEAD, "insufficient lookahead");
820 Assert(cur_match < strstart, "no future");
821 match = window + cur_match;
823 /* Skip to next match if the match length cannot increase
824 * or if the match length is less than 2:
826 if (match[best_len] != scan_end ||
827 match[best_len - 1] != scan_end1 ||
828 *match != *scan || *++match != scan[1])
831 /* The check at best_len-1 can be removed because it will be made
832 * again later. (This heuristic is not always a win.)
833 * It is not necessary to compare scan[2] and match[2] since they
834 * are always equal when the other bytes match, given that
835 * the hash keys are equal and that HASH_BITS >= 8.
839 /* We check for insufficient lookahead only every 8th comparison;
840 * the 256th check will be made at strstart+258.
843 } while (*++scan == *++match && *++scan == *++match &&
844 *++scan == *++match && *++scan == *++match &&
845 *++scan == *++match && *++scan == *++match &&
846 *++scan == *++match && *++scan == *++match && scan < strend);
848 len = MAX_MATCH - (int) (strend - scan);
849 scan = strend - MAX_MATCH;
851 if (len > best_len) {
852 match_start = cur_match;
854 if (len >= nice_match)
856 scan_end1 = scan[best_len - 1];
857 scan_end = scan[best_len];
859 } while ((cur_match = prev[cur_match & WMASK]) > limit
860 && --chain_length != 0);
866 /* ===========================================================================
867 * Check that the match at match_start is indeed a match.
869 static void check_match(IPos start, IPos match, int length)
871 /* check that the match is indeed a match */
872 if (memcmp((char *) window + match,
873 (char *) window + start, length) != EQUAL) {
874 bb_error_msg(" start %d, match %d, length %d", start, match, length);
875 bb_error_msg("invalid match");
878 bb_error_msg("\\[%d,%d]", start - match, length);
880 putc(window[start++], stderr);
881 } while (--length != 0);
885 # define check_match(start, match, length)
888 /* ===========================================================================
889 * Fill the window when the lookahead becomes insufficient.
890 * Updates strstart and lookahead, and sets eofile if end of input file.
891 * IN assertion: lookahead < MIN_LOOKAHEAD && strstart + lookahead > 0
892 * OUT assertions: at least one byte has been read, or eofile is set;
893 * file reads are performed for at least two bytes (required for the
894 * translate_eol option).
896 static void fill_window(void)
900 (unsigned) (window_size - (ulg) lookahead - (ulg) strstart);
901 /* Amount of free space at the end of the window. */
903 /* If the window is almost full and there is insufficient lookahead,
904 * move the upper half to the lower one to make room in the upper half.
906 if (more == (unsigned) EOF) {
907 /* Very unlikely, but possible on 16 bit machine if strstart == 0
908 * and lookahead == 1 (input done one byte at time)
911 } else if (strstart >= WSIZE + MAX_DIST) {
912 /* By the IN assertion, the window is not empty so we can't confuse
913 * more == 0 with more == 64K on a 16 bit machine.
915 Assert(window_size == (ulg) 2 * WSIZE, "no sliding with BIG_MEM");
917 memcpy((char *) window, (char *) window + WSIZE, (unsigned) WSIZE);
918 match_start -= WSIZE;
919 strstart -= WSIZE; /* we now have strstart >= MAX_DIST: */
921 block_start -= (long) WSIZE;
923 for (n = 0; n < HASH_SIZE; n++) {
925 head[n] = (Pos) (m >= WSIZE ? m - WSIZE : NIL);
927 for (n = 0; n < WSIZE; n++) {
929 prev[n] = (Pos) (m >= WSIZE ? m - WSIZE : NIL);
930 /* If n is not on any hash chain, prev[n] is garbage but
931 * its value will never be used.
936 /* At this point, more >= 2 */
938 n = read_buf((char *) window + strstart + lookahead, more);
939 if (n == 0 || n == (unsigned) EOF) {
947 /* ===========================================================================
948 * Flush the current block, with given end-of-file flag.
949 * IN assertion: strstart is set to the end of the current match.
951 #define FLUSH_BLOCK(eof) \
952 flush_block(block_start >= 0L ? (char*)&window[(unsigned)block_start] : \
953 (char*)NULL, (long)strstart - block_start, (eof))
955 /* ===========================================================================
956 * Same as above, but achieves better compression. We use a lazy
957 * evaluation for matches: a match is finally adopted only if there is
958 * no better match at the next window position.
960 static ulg deflate(void)
962 IPos hash_head; /* head of hash chain */
963 IPos prev_match; /* previous match */
964 int flush; /* set if current block must be flushed */
965 int match_available = 0; /* set if previous match exists */
966 unsigned match_length = MIN_MATCH - 1; /* length of best match */
968 /* Process the input block. */
969 while (lookahead != 0) {
970 /* Insert the string window[strstart .. strstart+2] in the
971 * dictionary, and set hash_head to the head of the hash chain:
973 INSERT_STRING(strstart, hash_head);
975 /* Find the longest match, discarding those <= prev_length.
977 prev_length = match_length, prev_match = match_start;
978 match_length = MIN_MATCH - 1;
980 if (hash_head != NIL && prev_length < max_lazy_match &&
981 strstart - hash_head <= MAX_DIST) {
982 /* To simplify the code, we prevent matches with the string
983 * of window index 0 (in particular we have to avoid a match
984 * of the string with itself at the start of the input file).
986 match_length = longest_match(hash_head);
987 /* longest_match() sets match_start */
988 if (match_length > lookahead)
989 match_length = lookahead;
991 /* Ignore a length 3 match if it is too distant: */
992 if (match_length == MIN_MATCH && strstart - match_start > TOO_FAR) {
993 /* If prev_match is also MIN_MATCH, match_start is garbage
994 * but we will ignore the current match anyway.
999 /* If there was a match at the previous step and the current
1000 * match is not better, output the previous match:
1002 if (prev_length >= MIN_MATCH && match_length <= prev_length) {
1004 check_match(strstart - 1, prev_match, prev_length);
1006 flush = ct_tally(strstart - 1 - prev_match, prev_length - MIN_MATCH);
1008 /* Insert in hash table all strings up to the end of the match.
1009 * strstart-1 and strstart are already inserted.
1011 lookahead -= prev_length - 1;
1015 INSERT_STRING(strstart, hash_head);
1016 /* strstart never exceeds WSIZE-MAX_MATCH, so there are
1017 * always MIN_MATCH bytes ahead. If lookahead < MIN_MATCH
1018 * these bytes are garbage, but it does not matter since the
1019 * next lookahead bytes will always be emitted as literals.
1021 } while (--prev_length != 0);
1022 match_available = 0;
1023 match_length = MIN_MATCH - 1;
1026 FLUSH_BLOCK(0), block_start = strstart;
1028 } else if (match_available) {
1029 /* If there was no match at the previous position, output a
1030 * single literal. If there was a match but the current match
1031 * is longer, truncate the previous match to a single literal.
1033 Tracevv((stderr, "%c", window[strstart - 1]));
1034 if (ct_tally(0, window[strstart - 1])) {
1035 FLUSH_BLOCK(0), block_start = strstart;
1040 /* There is no previous match to compare with, wait for
1041 * the next step to decide.
1043 match_available = 1;
1047 Assert(strstart <= isize && lookahead <= isize, "a bit too far");
1049 /* Make sure that we always have enough lookahead, except
1050 * at the end of the input file. We need MAX_MATCH bytes
1051 * for the next match, plus MIN_MATCH bytes to insert the
1052 * string following the next match.
1054 while (lookahead < MIN_LOOKAHEAD && !eofile)
1057 if (match_available)
1058 ct_tally(0, window[strstart - 1]);
1060 return FLUSH_BLOCK(1); /* eof */
1063 /* gzip (GNU zip) -- compress files with zip algorithm and 'compress' interface
1064 * Copyright (C) 1992-1993 Jean-loup Gailly
1065 * The unzip code was written and put in the public domain by Mark Adler.
1066 * Portions of the lzw code are derived from the public domain 'compress'
1067 * written by Spencer Thomas, Joe Orost, James Woods, Jim McKie, Steve Davies,
1068 * Ken Turkowski, Dave Mack and Peter Jannesen.
1070 * See the license_msg below and the file COPYING for the software license.
1071 * See the file algorithm.doc for the compression algorithms and file formats.
1074 /* Compress files with zip algorithm and 'compress' interface.
1075 * See usage() and help() functions below for all options.
1077 * file.gz: compressed file with same mode, owner, and utimes
1078 * or stdout with -c option or if stdin used as input.
1079 * If the output file name had to be truncated, the original name is kept
1080 * in the compressed file.
1085 typedef struct dirent dir_type;
1087 /* ======================================================================== */
1088 static void abort_gzip(int ATTRIBUTE_UNUSED ignored)
1093 int gzip_main(int argc, char **argv)
1104 struct stat statBuf;
1107 opt = getopt32(argc, argv, "cf123456789qv" USE_GUNZIP("d"));
1108 //if (opt & 0x1) // -c
1109 //if (opt & 0x2) // -f
1110 /* Ignore 1-9 (compression level) options */
1111 //if (opt & 0x4) // -1
1112 //if (opt & 0x8) // -2
1113 //if (opt & 0x10) // -3
1114 //if (opt & 0x20) // -4
1115 //if (opt & 0x40) // -5
1116 //if (opt & 0x80) // -6
1117 //if (opt & 0x100) // -7
1118 //if (opt & 0x200) // -8
1119 //if (opt & 0x400) // -9
1120 //if (opt & 0x800) // -q
1121 //if (opt & 0x1000) // -v
1122 #if ENABLE_GUNZIP /* gunzip_main may not be visible... */
1123 if (opt & 0x2000) { // -d
1124 /* FIXME: getopt32 should not depend on optind */
1126 return gunzip_main(argc, argv);
1130 foreground = signal(SIGINT, SIG_IGN) != SIG_IGN;
1132 (void) signal(SIGINT, abort_gzip);
1135 if (signal(SIGTERM, SIG_IGN) != SIG_IGN) {
1136 (void) signal(SIGTERM, abort_gzip);
1140 if (signal(SIGHUP, SIG_IGN) != SIG_IGN) {
1141 (void) signal(SIGHUP, abort_gzip);
1145 strncpy(z_suffix, ".gz", sizeof(z_suffix) - 1);
1147 /* Allocate all global buffers (for DYN_ALLOC option) */
1148 ALLOC(uch, inbuf, INBUFSIZ + INBUF_EXTRA);
1149 ALLOC(uch, outbuf, OUTBUFSIZ + OUTBUF_EXTRA);
1150 ALLOC(ush, d_buf, DIST_BUFSIZE);
1151 ALLOC(uch, window, 2L * WSIZE);
1152 ALLOC(ush, tab_prefix, 1L << BITS);
1154 /* Initialise the CRC32 table */
1155 crc_32_tab = crc32_filltable(0);
1159 if (optind == argc) {
1161 zip(STDIN_FILENO, STDOUT_FILENO);
1165 for (i = optind; i < argc; i++) {
1169 if (LONE_DASH(argv[i])) {
1171 inFileNum = STDIN_FILENO;
1172 outFileNum = STDOUT_FILENO;
1174 inFileNum = xopen(argv[i], O_RDONLY);
1175 if (fstat(inFileNum, &statBuf) < 0)
1176 bb_perror_msg_and_die("%s", argv[i]);
1177 time_stamp = statBuf.st_ctime;
1179 if (!(opt & OPT_tostdout)) {
1180 path = xasprintf("%s.gz", argv[i]);
1182 /* Open output file */
1183 #if defined(__GLIBC__) && __GLIBC__ >= 2 && __GLIBC_MINOR__ >= 1 && defined(O_NOFOLLOW)
1185 open(path, O_RDWR | O_CREAT | O_EXCL | O_NOFOLLOW);
1187 outFileNum = open(path, O_RDWR | O_CREAT | O_EXCL);
1189 if (outFileNum < 0) {
1190 bb_perror_msg("%s", path);
1195 /* Set permissions on the file */
1196 fchmod(outFileNum, statBuf.st_mode);
1198 outFileNum = STDOUT_FILENO;
1201 if (path == NULL && isatty(outFileNum) && !(opt & OPT_force)) {
1203 ("compressed data not written to a terminal. Use -f to force compression.");
1208 result = zip(inFileNum, outFileNum);
1214 /* Delete the original file */
1216 delFileName = argv[i];
1220 if (unlink(delFileName) < 0)
1221 bb_perror_msg("%s", delFileName);
1231 /* trees.c -- output deflated data using Huffman coding
1232 * Copyright (C) 1992-1993 Jean-loup Gailly
1233 * This is free software; you can redistribute it and/or modify it under the
1234 * terms of the GNU General Public License, see the file COPYING.
1240 * Encode various sets of source values using variable-length
1241 * binary code trees.
1245 * The PKZIP "deflation" process uses several Huffman trees. The more
1246 * common source values are represented by shorter bit sequences.
1248 * Each code tree is stored in the ZIP file in a compressed form
1249 * which is itself a Huffman encoding of the lengths of
1250 * all the code strings (in ascending order by source values).
1251 * The actual code strings are reconstructed from the lengths in
1252 * the UNZIP process, as described in the "application note"
1253 * (APPNOTE.TXT) distributed as part of PKWARE's PKZIP program.
1258 * Data Compression: Techniques and Applications, pp. 53-55.
1259 * Lifetime Learning Publications, 1985. ISBN 0-534-03418-7.
1262 * Data Compression: Methods and Theory, pp. 49-50.
1263 * Computer Science Press, 1988. ISBN 0-7167-8156-5.
1267 * Addison-Wesley, 1983. ISBN 0-201-06672-6.
1271 * void ct_init (ush *attr, int *methodp)
1272 * Allocate the match buffer, initialize the various tables and save
1273 * the location of the internal file attribute (ascii/binary) and
1274 * method (DEFLATE/STORE)
1276 * void ct_tally (int dist, int lc);
1277 * Save the match info and tally the frequency counts.
1279 * long flush_block (char *buf, ulg stored_len, int eof)
1280 * Determine the best encoding for the current block: dynamic trees,
1281 * static trees or store, and output the encoded block to the zip
1282 * file. Returns the total compressed length for the file so far.
1286 /* ===========================================================================
1291 /* All codes must not exceed MAX_BITS bits */
1293 #define MAX_BL_BITS 7
1294 /* Bit length codes must not exceed MAX_BL_BITS bits */
1296 #define LENGTH_CODES 29
1297 /* number of length codes, not counting the special END_BLOCK code */
1299 #define LITERALS 256
1300 /* number of literal bytes 0..255 */
1302 #define END_BLOCK 256
1303 /* end of block literal code */
1305 #define L_CODES (LITERALS+1+LENGTH_CODES)
1306 /* number of Literal or Length codes, including the END_BLOCK code */
1309 /* number of distance codes */
1312 /* number of codes used to transfer the bit lengths */
1314 typedef uch extra_bits_t;
1316 /* extra bits for each length code */
1317 static const extra_bits_t extra_lbits[LENGTH_CODES]
1318 = { 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4,
1322 /* extra bits for each distance code */
1323 static const extra_bits_t extra_dbits[D_CODES]
1324 = { 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9,
1325 10, 10, 11, 11, 12, 12, 13, 13
1328 /* extra bits for each bit length code */
1329 static const extra_bits_t extra_blbits[BL_CODES]
1330 = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 3, 7 };
1332 #define STORED_BLOCK 0
1333 #define STATIC_TREES 1
1335 /* The three kinds of block type */
1339 # define LIT_BUFSIZE 0x2000
1342 # define LIT_BUFSIZE 0x4000
1344 # define LIT_BUFSIZE 0x8000
1348 #ifndef DIST_BUFSIZE
1349 # define DIST_BUFSIZE LIT_BUFSIZE
1351 /* Sizes of match buffers for literals/lengths and distances. There are
1352 * 4 reasons for limiting LIT_BUFSIZE to 64K:
1353 * - frequencies can be kept in 16 bit counters
1354 * - if compression is not successful for the first block, all input data is
1355 * still in the window so we can still emit a stored block even when input
1356 * comes from standard input. (This can also be done for all blocks if
1357 * LIT_BUFSIZE is not greater than 32K.)
1358 * - if compression is not successful for a file smaller than 64K, we can
1359 * even emit a stored file instead of a stored block (saving 5 bytes).
1360 * - creating new Huffman trees less frequently may not provide fast
1361 * adaptation to changes in the input data statistics. (Take for
1362 * example a binary file with poorly compressible code followed by
1363 * a highly compressible string table.) Smaller buffer sizes give
1364 * fast adaptation but have of course the overhead of transmitting trees
1366 * - I can't count above 4
1367 * The current code is general and allows DIST_BUFSIZE < LIT_BUFSIZE (to save
1368 * memory at the expense of compression). Some optimizations would be possible
1369 * if we rely on DIST_BUFSIZE == LIT_BUFSIZE.
1371 #if LIT_BUFSIZE > INBUFSIZ
1372 #error cannot overlay l_buf and inbuf
1375 /* repeat previous bit length 3-6 times (2 bits of repeat count) */
1376 #define REPZ_3_10 17
1377 /* repeat a zero length 3-10 times (3 bits of repeat count) */
1378 #define REPZ_11_138 18
1379 /* repeat a zero length 11-138 times (7 bits of repeat count) */
1381 /* ===========================================================================
1385 /* Data structure describing a single value and its code string. */
1386 typedef struct ct_data {
1388 ush freq; /* frequency count */
1389 ush code; /* bit string */
1392 ush dad; /* father node in Huffman tree */
1393 ush len; /* length of bit string */
1397 #define Freq fc.freq
1398 #define Code fc.code
1402 #define HEAP_SIZE (2*L_CODES+1)
1403 /* maximum heap size */
1405 static ct_data dyn_ltree[HEAP_SIZE]; /* literal and length tree */
1406 static ct_data dyn_dtree[2 * D_CODES + 1]; /* distance tree */
1408 static ct_data static_ltree[L_CODES + 2];
1410 /* The static literal tree. Since the bit lengths are imposed, there is no
1411 * need for the L_CODES extra codes used during heap construction. However
1412 * The codes 286 and 287 are needed to build a canonical tree (see ct_init
1416 static ct_data static_dtree[D_CODES];
1418 /* The static distance tree. (Actually a trivial tree since all codes use
1422 static ct_data bl_tree[2 * BL_CODES + 1];
1424 /* Huffman tree for the bit lengths */
1426 typedef struct tree_desc {
1427 ct_data *dyn_tree; /* the dynamic tree */
1428 ct_data *static_tree; /* corresponding static tree or NULL */
1429 const extra_bits_t *extra_bits; /* extra bits for each code or NULL */
1430 int extra_base; /* base index for extra_bits */
1431 int elems; /* max number of elements in the tree */
1432 int max_length; /* max bit length for the codes */
1433 int max_code; /* largest code with non zero frequency */
1436 static tree_desc l_desc =
1437 { dyn_ltree, static_ltree, extra_lbits, LITERALS + 1, L_CODES,
1441 static tree_desc d_desc =
1442 { dyn_dtree, static_dtree, extra_dbits, 0, D_CODES, MAX_BITS, 0 };
1444 static tree_desc bl_desc =
1445 { bl_tree, (ct_data *) 0, extra_blbits, 0, BL_CODES, MAX_BL_BITS,
1450 static ush bl_count[MAX_BITS + 1];
1452 /* number of codes at each bit length for an optimal tree */
1454 static const uch bl_order[BL_CODES]
1455 = { 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15 };
1457 /* The lengths of the bit length codes are sent in order of decreasing
1458 * probability, to avoid transmitting the lengths for unused bit length codes.
1461 static int heap[2 * L_CODES + 1]; /* heap used to build the Huffman trees */
1462 static int heap_len; /* number of elements in the heap */
1463 static int heap_max; /* element of largest frequency */
1465 /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used.
1466 * The same heap array is used to build all trees.
1469 static uch depth[2 * L_CODES + 1];
1471 /* Depth of each subtree used as tie breaker for trees of equal frequency */
1473 static uch length_code[MAX_MATCH - MIN_MATCH + 1];
1475 /* length code for each normalized match length (0 == MIN_MATCH) */
1477 static uch dist_code[512];
1479 /* distance codes. The first 256 values correspond to the distances
1480 * 3 .. 258, the last 256 values correspond to the top 8 bits of
1481 * the 15 bit distances.
1484 static int base_length[LENGTH_CODES];
1486 /* First normalized length for each code (0 = MIN_MATCH) */
1488 static int base_dist[D_CODES];
1490 /* First normalized distance for each code (0 = distance of 1) */
1493 /* DECLARE(uch, l_buf, LIT_BUFSIZE); buffer for literals or lengths */
1495 /* DECLARE(ush, d_buf, DIST_BUFSIZE); buffer for distances */
1497 static uch flag_buf[(LIT_BUFSIZE / 8)];
1499 /* flag_buf is a bit array distinguishing literals from lengths in
1500 * l_buf, thus indicating the presence or absence of a distance.
1503 static unsigned last_lit; /* running index in l_buf */
1504 static unsigned last_dist; /* running index in d_buf */
1505 static unsigned last_flags; /* running index in flag_buf */
1506 static uch flags; /* current flags not yet saved in flag_buf */
1507 static uch flag_bit; /* current bit used in flags */
1509 /* bits are filled in flags starting at bit 0 (least significant).
1510 * Note: these flags are overkill in the current code since we don't
1511 * take advantage of DIST_BUFSIZE == LIT_BUFSIZE.
1514 static ulg opt_len; /* bit length of current block with optimal trees */
1515 static ulg static_len; /* bit length of current block with static trees */
1517 static ulg compressed_len; /* total bit length of compressed file */
1520 static ush *file_type; /* pointer to UNKNOWN, BINARY or ASCII */
1521 static int *file_method; /* pointer to DEFLATE or STORE */
1523 /* ===========================================================================
1524 * Local (static) routines in this file.
1527 static void init_block(void);
1528 static void pqdownheap(ct_data * tree, int k);
1529 static void gen_bitlen(tree_desc * desc);
1530 static void gen_codes(ct_data * tree, int max_code);
1531 static void build_tree(tree_desc * desc);
1532 static void scan_tree(ct_data * tree, int max_code);
1533 static void send_tree(ct_data * tree, int max_code);
1534 static int build_bl_tree(void);
1535 static void send_all_trees(int lcodes, int dcodes, int blcodes);
1536 static void compress_block(ct_data * ltree, ct_data * dtree);
1537 static void set_file_type(void);
1541 # define send_code(c, tree) send_bits(tree[c].Code, tree[c].Len)
1542 /* Send a code of the given tree. c and tree must not have side effects */
1545 # define send_code(c, tree) \
1546 { if (verbose>1) bb_error_msg("\ncd %3d ",(c)); \
1547 send_bits(tree[c].Code, tree[c].Len); }
1550 #define d_code(dist) \
1551 ((dist) < 256 ? dist_code[dist] : dist_code[256+((dist)>>7)])
1552 /* Mapping from a distance to a distance code. dist is the distance - 1 and
1553 * must not have side effects. dist_code[256] and dist_code[257] are never
1557 /* the arguments must not have side effects */
1559 /* ===========================================================================
1560 * Allocate the match buffer, initialize the various tables and save the
1561 * location of the internal file attribute (ascii/binary) and method
1564 static void ct_init(ush * attr, int *methodp)
1566 int n; /* iterates over tree elements */
1567 int bits; /* bit counter */
1568 int length; /* length value */
1569 int code; /* code value */
1570 int dist; /* distance index */
1573 file_method = methodp;
1574 compressed_len = 0L;
1576 if (static_dtree[0].Len != 0)
1577 return; /* ct_init already called */
1579 /* Initialize the mapping length (0..255) -> length code (0..28) */
1581 for (code = 0; code < LENGTH_CODES - 1; code++) {
1582 base_length[code] = length;
1583 for (n = 0; n < (1 << extra_lbits[code]); n++) {
1584 length_code[length++] = (uch) code;
1587 Assert(length == 256, "ct_init: length != 256");
1588 /* Note that the length 255 (match length 258) can be represented
1589 * in two different ways: code 284 + 5 bits or code 285, so we
1590 * overwrite length_code[255] to use the best encoding:
1592 length_code[length - 1] = (uch) code;
1594 /* Initialize the mapping dist (0..32K) -> dist code (0..29) */
1596 for (code = 0; code < 16; code++) {
1597 base_dist[code] = dist;
1598 for (n = 0; n < (1 << extra_dbits[code]); n++) {
1599 dist_code[dist++] = (uch) code;
1602 Assert(dist == 256, "ct_init: dist != 256");
1603 dist >>= 7; /* from now on, all distances are divided by 128 */
1604 for (; code < D_CODES; code++) {
1605 base_dist[code] = dist << 7;
1606 for (n = 0; n < (1 << (extra_dbits[code] - 7)); n++) {
1607 dist_code[256 + dist++] = (uch) code;
1610 Assert(dist == 256, "ct_init: 256+dist != 512");
1612 /* Construct the codes of the static literal tree */
1613 for (bits = 0; bits <= MAX_BITS; bits++)
1617 static_ltree[n++].Len = 8, bl_count[8]++;
1619 static_ltree[n++].Len = 9, bl_count[9]++;
1621 static_ltree[n++].Len = 7, bl_count[7]++;
1623 static_ltree[n++].Len = 8, bl_count[8]++;
1624 /* Codes 286 and 287 do not exist, but we must include them in the
1625 * tree construction to get a canonical Huffman tree (longest code
1628 gen_codes((ct_data *) static_ltree, L_CODES + 1);
1630 /* The static distance tree is trivial: */
1631 for (n = 0; n < D_CODES; n++) {
1632 static_dtree[n].Len = 5;
1633 static_dtree[n].Code = bi_reverse(n, 5);
1636 /* Initialize the first block of the first file: */
1640 /* ===========================================================================
1641 * Initialize a new block.
1643 static void init_block(void)
1645 int n; /* iterates over tree elements */
1647 /* Initialize the trees. */
1648 for (n = 0; n < L_CODES; n++)
1649 dyn_ltree[n].Freq = 0;
1650 for (n = 0; n < D_CODES; n++)
1651 dyn_dtree[n].Freq = 0;
1652 for (n = 0; n < BL_CODES; n++)
1653 bl_tree[n].Freq = 0;
1655 dyn_ltree[END_BLOCK].Freq = 1;
1656 opt_len = static_len = 0L;
1657 last_lit = last_dist = last_flags = 0;
1663 /* Index within the heap array of least frequent node in the Huffman tree */
1666 /* ===========================================================================
1667 * Remove the smallest element from the heap and recreate the heap with
1668 * one less element. Updates heap and heap_len.
1670 #define pqremove(tree, top) \
1672 top = heap[SMALLEST]; \
1673 heap[SMALLEST] = heap[heap_len--]; \
1674 pqdownheap(tree, SMALLEST); \
1677 /* ===========================================================================
1678 * Compares to subtrees, using the tree depth as tie breaker when
1679 * the subtrees have equal frequency. This minimizes the worst case length.
1681 #define smaller(tree, n, m) \
1682 (tree[n].Freq < tree[m].Freq || \
1683 (tree[n].Freq == tree[m].Freq && depth[n] <= depth[m]))
1685 /* ===========================================================================
1686 * Restore the heap property by moving down the tree starting at node k,
1687 * exchanging a node with the smallest of its two sons if necessary, stopping
1688 * when the heap property is re-established (each father smaller than its
1691 static void pqdownheap(ct_data * tree, int k)
1694 int j = k << 1; /* left son of k */
1696 while (j <= heap_len) {
1697 /* Set j to the smallest of the two sons: */
1698 if (j < heap_len && smaller(tree, heap[j + 1], heap[j]))
1701 /* Exit if v is smaller than both sons */
1702 if (smaller(tree, v, heap[j]))
1705 /* Exchange v with the smallest son */
1709 /* And continue down the tree, setting j to the left son of k */
1715 /* ===========================================================================
1716 * Compute the optimal bit lengths for a tree and update the total bit length
1717 * for the current block.
1718 * IN assertion: the fields freq and dad are set, heap[heap_max] and
1719 * above are the tree nodes sorted by increasing frequency.
1720 * OUT assertions: the field len is set to the optimal bit length, the
1721 * array bl_count contains the frequencies for each bit length.
1722 * The length opt_len is updated; static_len is also updated if stree is
1725 static void gen_bitlen(tree_desc * desc)
1727 ct_data *tree = desc->dyn_tree;
1728 const extra_bits_t *extra = desc->extra_bits;
1729 int base = desc->extra_base;
1730 int max_code = desc->max_code;
1731 int max_length = desc->max_length;
1732 ct_data *stree = desc->static_tree;
1733 int h; /* heap index */
1734 int n, m; /* iterate over the tree elements */
1735 int bits; /* bit length */
1736 int xbits; /* extra bits */
1737 ush f; /* frequency */
1738 int overflow = 0; /* number of elements with bit length too large */
1740 for (bits = 0; bits <= MAX_BITS; bits++)
1743 /* In a first pass, compute the optimal bit lengths (which may
1744 * overflow in the case of the bit length tree).
1746 tree[heap[heap_max]].Len = 0; /* root of the heap */
1748 for (h = heap_max + 1; h < HEAP_SIZE; h++) {
1750 bits = tree[tree[n].Dad].Len + 1;
1751 if (bits > max_length)
1752 bits = max_length, overflow++;
1753 tree[n].Len = (ush) bits;
1754 /* We overwrite tree[n].Dad which is no longer needed */
1757 continue; /* not a leaf node */
1762 xbits = extra[n - base];
1764 opt_len += (ulg) f *(bits + xbits);
1767 static_len += (ulg) f *(stree[n].Len + xbits);
1772 Trace((stderr, "\nbit length overflow\n"));
1773 /* This happens for example on obj2 and pic of the Calgary corpus */
1775 /* Find the first bit length which could increase: */
1777 bits = max_length - 1;
1778 while (bl_count[bits] == 0)
1780 bl_count[bits]--; /* move one leaf down the tree */
1781 bl_count[bits + 1] += 2; /* move one overflow item as its brother */
1782 bl_count[max_length]--;
1783 /* The brother of the overflow item also moves one step up,
1784 * but this does not affect bl_count[max_length]
1787 } while (overflow > 0);
1789 /* Now recompute all bit lengths, scanning in increasing frequency.
1790 * h is still equal to HEAP_SIZE. (It is simpler to reconstruct all
1791 * lengths instead of fixing only the wrong ones. This idea is taken
1792 * from 'ar' written by Haruhiko Okumura.)
1794 for (bits = max_length; bits != 0; bits--) {
1800 if (tree[m].Len != (unsigned) bits) {
1801 Trace((stderr, "code %d bits %d->%d\n", m, tree[m].Len,
1804 ((long) bits - (long) tree[m].Len) * (long) tree[m].Freq;
1805 tree[m].Len = (ush) bits;
1812 /* ===========================================================================
1813 * Generate the codes for a given tree and bit counts (which need not be
1815 * IN assertion: the array bl_count contains the bit length statistics for
1816 * the given tree and the field len is set for all tree elements.
1817 * OUT assertion: the field code is set for all tree elements of non
1820 static void gen_codes(ct_data * tree, int max_code)
1822 ush next_code[MAX_BITS + 1]; /* next code value for each bit length */
1823 ush code = 0; /* running code value */
1824 int bits; /* bit index */
1825 int n; /* code index */
1827 /* The distribution counts are first used to generate the code values
1828 * without bit reversal.
1830 for (bits = 1; bits <= MAX_BITS; bits++) {
1831 next_code[bits] = code = (code + bl_count[bits - 1]) << 1;
1833 /* Check that the bit counts in bl_count are consistent. The last code
1836 Assert(code + bl_count[MAX_BITS] - 1 == (1 << MAX_BITS) - 1,
1837 "inconsistent bit counts");
1838 Tracev((stderr, "\ngen_codes: max_code %d ", max_code));
1840 for (n = 0; n <= max_code; n++) {
1841 int len = tree[n].Len;
1845 /* Now reverse the bits */
1846 tree[n].Code = bi_reverse(next_code[len]++, len);
1848 Tracec(tree != static_ltree,
1849 (stderr, "\nn %3d %c l %2d c %4x (%x) ", n,
1850 (isgraph(n) ? n : ' '), len, tree[n].Code,
1851 next_code[len] - 1));
1855 /* ===========================================================================
1856 * Construct one Huffman tree and assigns the code bit strings and lengths.
1857 * Update the total bit length for the current block.
1858 * IN assertion: the field freq is set for all tree elements.
1859 * OUT assertions: the fields len and code are set to the optimal bit length
1860 * and corresponding code. The length opt_len is updated; static_len is
1861 * also updated if stree is not null. The field max_code is set.
1863 static void build_tree(tree_desc * desc)
1865 ct_data *tree = desc->dyn_tree;
1866 ct_data *stree = desc->static_tree;
1867 int elems = desc->elems;
1868 int n, m; /* iterate over heap elements */
1869 int max_code = -1; /* largest code with non zero frequency */
1870 int node = elems; /* next internal node of the tree */
1872 /* Construct the initial heap, with least frequent element in
1873 * heap[SMALLEST]. The sons of heap[n] are heap[2*n] and heap[2*n+1].
1874 * heap[0] is not used.
1876 heap_len = 0, heap_max = HEAP_SIZE;
1878 for (n = 0; n < elems; n++) {
1879 if (tree[n].Freq != 0) {
1880 heap[++heap_len] = max_code = n;
1887 /* The pkzip format requires that at least one distance code exists,
1888 * and that at least one bit should be sent even if there is only one
1889 * possible code. So to avoid special checks later on we force at least
1890 * two codes of non zero frequency.
1892 while (heap_len < 2) {
1893 int new = heap[++heap_len] = (max_code < 2 ? ++max_code : 0);
1899 static_len -= stree[new].Len;
1900 /* new is 0 or 1 so it does not have extra bits */
1902 desc->max_code = max_code;
1904 /* The elements heap[heap_len/2+1 .. heap_len] are leaves of the tree,
1905 * establish sub-heaps of increasing lengths:
1907 for (n = heap_len / 2; n >= 1; n--)
1908 pqdownheap(tree, n);
1910 /* Construct the Huffman tree by repeatedly combining the least two
1914 pqremove(tree, n); /* n = node of least frequency */
1915 m = heap[SMALLEST]; /* m = node of next least frequency */
1917 heap[--heap_max] = n; /* keep the nodes sorted by frequency */
1918 heap[--heap_max] = m;
1920 /* Create a new node father of n and m */
1921 tree[node].Freq = tree[n].Freq + tree[m].Freq;
1922 depth[node] = (uch) (MAX(depth[n], depth[m]) + 1);
1923 tree[n].Dad = tree[m].Dad = (ush) node;
1925 if (tree == bl_tree) {
1926 bb_error_msg("\nnode %d(%d), sons %d(%d) %d(%d)",
1927 node, tree[node].Freq, n, tree[n].Freq, m, tree[m].Freq);
1930 /* and insert the new node in the heap */
1931 heap[SMALLEST] = node++;
1932 pqdownheap(tree, SMALLEST);
1934 } while (heap_len >= 2);
1936 heap[--heap_max] = heap[SMALLEST];
1938 /* At this point, the fields freq and dad are set. We can now
1939 * generate the bit lengths.
1941 gen_bitlen((tree_desc *) desc);
1943 /* The field len is now set, we can generate the bit codes */
1944 gen_codes((ct_data *) tree, max_code);
1947 /* ===========================================================================
1948 * Scan a literal or distance tree to determine the frequencies of the codes
1949 * in the bit length tree. Updates opt_len to take into account the repeat
1950 * counts. (The contribution of the bit length codes will be added later
1951 * during the construction of bl_tree.)
1953 static void scan_tree(ct_data * tree, int max_code)
1955 int n; /* iterates over all tree elements */
1956 int prevlen = -1; /* last emitted length */
1957 int curlen; /* length of current code */
1958 int nextlen = tree[0].Len; /* length of next code */
1959 int count = 0; /* repeat count of the current code */
1960 int max_count = 7; /* max repeat count */
1961 int min_count = 4; /* min repeat count */
1964 max_count = 138, min_count = 3;
1965 tree[max_code + 1].Len = (ush) 0xffff; /* guard */
1967 for (n = 0; n <= max_code; n++) {
1969 nextlen = tree[n + 1].Len;
1970 if (++count < max_count && curlen == nextlen) {
1972 } else if (count < min_count) {
1973 bl_tree[curlen].Freq += count;
1974 } else if (curlen != 0) {
1975 if (curlen != prevlen)
1976 bl_tree[curlen].Freq++;
1977 bl_tree[REP_3_6].Freq++;
1978 } else if (count <= 10) {
1979 bl_tree[REPZ_3_10].Freq++;
1981 bl_tree[REPZ_11_138].Freq++;
1986 max_count = 138, min_count = 3;
1987 } else if (curlen == nextlen) {
1988 max_count = 6, min_count = 3;
1990 max_count = 7, min_count = 4;
1995 /* ===========================================================================
1996 * Send a literal or distance tree in compressed form, using the codes in
1999 static void send_tree(ct_data * tree, int max_code)
2001 int n; /* iterates over all tree elements */
2002 int prevlen = -1; /* last emitted length */
2003 int curlen; /* length of current code */
2004 int nextlen = tree[0].Len; /* length of next code */
2005 int count = 0; /* repeat count of the current code */
2006 int max_count = 7; /* max repeat count */
2007 int min_count = 4; /* min repeat count */
2009 /* tree[max_code+1].Len = -1; *//* guard already set */
2011 max_count = 138, min_count = 3;
2013 for (n = 0; n <= max_code; n++) {
2015 nextlen = tree[n + 1].Len;
2016 if (++count < max_count && curlen == nextlen) {
2018 } else if (count < min_count) {
2020 send_code(curlen, bl_tree);
2021 } while (--count != 0);
2023 } else if (curlen != 0) {
2024 if (curlen != prevlen) {
2025 send_code(curlen, bl_tree);
2028 Assert(count >= 3 && count <= 6, " 3_6?");
2029 send_code(REP_3_6, bl_tree);
2030 send_bits(count - 3, 2);
2032 } else if (count <= 10) {
2033 send_code(REPZ_3_10, bl_tree);
2034 send_bits(count - 3, 3);
2037 send_code(REPZ_11_138, bl_tree);
2038 send_bits(count - 11, 7);
2043 max_count = 138, min_count = 3;
2044 } else if (curlen == nextlen) {
2045 max_count = 6, min_count = 3;
2047 max_count = 7, min_count = 4;
2052 /* ===========================================================================
2053 * Construct the Huffman tree for the bit lengths and return the index in
2054 * bl_order of the last bit length code to send.
2056 static int build_bl_tree(void)
2058 int max_blindex; /* index of last bit length code of non zero freq */
2060 /* Determine the bit length frequencies for literal and distance trees */
2061 scan_tree((ct_data *) dyn_ltree, l_desc.max_code);
2062 scan_tree((ct_data *) dyn_dtree, d_desc.max_code);
2064 /* Build the bit length tree: */
2065 build_tree((tree_desc *) (&bl_desc));
2066 /* opt_len now includes the length of the tree representations, except
2067 * the lengths of the bit lengths codes and the 5+5+4 bits for the counts.
2070 /* Determine the number of bit length codes to send. The pkzip format
2071 * requires that at least 4 bit length codes be sent. (appnote.txt says
2072 * 3 but the actual value used is 4.)
2074 for (max_blindex = BL_CODES - 1; max_blindex >= 3; max_blindex--) {
2075 if (bl_tree[bl_order[max_blindex]].Len != 0)
2078 /* Update opt_len to include the bit length tree and counts */
2079 opt_len += 3 * (max_blindex + 1) + 5 + 5 + 4;
2080 Tracev((stderr, "\ndyn trees: dyn %ld, stat %ld", opt_len, static_len));
2085 /* ===========================================================================
2086 * Send the header for a block using dynamic Huffman trees: the counts, the
2087 * lengths of the bit length codes, the literal tree and the distance tree.
2088 * IN assertion: lcodes >= 257, dcodes >= 1, blcodes >= 4.
2090 static void send_all_trees(int lcodes, int dcodes, int blcodes)
2092 int rank; /* index in bl_order */
2094 Assert(lcodes >= 257 && dcodes >= 1 && blcodes >= 4, "not enough codes");
2095 Assert(lcodes <= L_CODES && dcodes <= D_CODES
2096 && blcodes <= BL_CODES, "too many codes");
2097 Tracev((stderr, "\nbl counts: "));
2098 send_bits(lcodes - 257, 5); /* not +255 as stated in appnote.txt */
2099 send_bits(dcodes - 1, 5);
2100 send_bits(blcodes - 4, 4); /* not -3 as stated in appnote.txt */
2101 for (rank = 0; rank < blcodes; rank++) {
2102 Tracev((stderr, "\nbl code %2d ", bl_order[rank]));
2103 send_bits(bl_tree[bl_order[rank]].Len, 3);
2105 Tracev((stderr, "\nbl tree: sent %ld", bits_sent));
2107 send_tree((ct_data *) dyn_ltree, lcodes - 1); /* send the literal tree */
2108 Tracev((stderr, "\nlit tree: sent %ld", bits_sent));
2110 send_tree((ct_data *) dyn_dtree, dcodes - 1); /* send the distance tree */
2111 Tracev((stderr, "\ndist tree: sent %ld", bits_sent));
2114 /* ===========================================================================
2115 * Determine the best encoding for the current block: dynamic trees, static
2116 * trees or store, and output the encoded block to the zip file. This function
2117 * returns the total compressed length for the file so far.
2119 static ulg flush_block(char *buf, ulg stored_len, int eof)
2121 ulg opt_lenb, static_lenb; /* opt_len and static_len in bytes */
2122 int max_blindex; /* index of last bit length code of non zero freq */
2124 flag_buf[last_flags] = flags; /* Save the flags for the last 8 items */
2126 /* Check if the file is ascii or binary */
2127 if (*file_type == (ush) UNKNOWN)
2130 /* Construct the literal and distance trees */
2131 build_tree((tree_desc *) (&l_desc));
2132 Tracev((stderr, "\nlit data: dyn %ld, stat %ld", opt_len, static_len));
2134 build_tree((tree_desc *) (&d_desc));
2135 Tracev((stderr, "\ndist data: dyn %ld, stat %ld", opt_len, static_len));
2136 /* At this point, opt_len and static_len are the total bit lengths of
2137 * the compressed block data, excluding the tree representations.
2140 /* Build the bit length tree for the above two trees, and get the index
2141 * in bl_order of the last bit length code to send.
2143 max_blindex = build_bl_tree();
2145 /* Determine the best encoding. Compute first the block length in bytes */
2146 opt_lenb = (opt_len + 3 + 7) >> 3;
2147 static_lenb = (static_len + 3 + 7) >> 3;
2150 "\nopt %lu(%lu) stat %lu(%lu) stored %lu lit %u dist %u ",
2151 opt_lenb, opt_len, static_lenb, static_len, stored_len,
2152 last_lit, last_dist));
2154 if (static_lenb <= opt_lenb)
2155 opt_lenb = static_lenb;
2157 /* If compression failed and this is the first and last block,
2158 * and if the zip file can be seeked (to rewrite the local header),
2159 * the whole file is transformed into a stored file:
2161 if (stored_len <= opt_lenb && eof && compressed_len == 0L && seekable()) {
2162 /* Since LIT_BUFSIZE <= 2*WSIZE, the input data must be there: */
2163 if (buf == (char *) 0)
2164 bb_error_msg("block vanished");
2166 copy_block(buf, (unsigned) stored_len, 0); /* without header */
2167 compressed_len = stored_len << 3;
2168 *file_method = STORED;
2170 } else if (stored_len + 4 <= opt_lenb && buf != (char *) 0) {
2171 /* 4: two words for the lengths */
2172 /* The test buf != NULL is only necessary if LIT_BUFSIZE > WSIZE.
2173 * Otherwise we can't have processed more than WSIZE input bytes since
2174 * the last block flush, because compression would have been
2175 * successful. If LIT_BUFSIZE <= WSIZE, it is never too late to
2176 * transform a block into a stored block.
2178 send_bits((STORED_BLOCK << 1) + eof, 3); /* send block type */
2179 compressed_len = (compressed_len + 3 + 7) & ~7L;
2180 compressed_len += (stored_len + 4) << 3;
2182 copy_block(buf, (unsigned) stored_len, 1); /* with header */
2184 } else if (static_lenb == opt_lenb) {
2185 send_bits((STATIC_TREES << 1) + eof, 3);
2186 compress_block((ct_data *) static_ltree, (ct_data *) static_dtree);
2187 compressed_len += 3 + static_len;
2189 send_bits((DYN_TREES << 1) + eof, 3);
2190 send_all_trees(l_desc.max_code + 1, d_desc.max_code + 1,
2192 compress_block((ct_data *) dyn_ltree, (ct_data *) dyn_dtree);
2193 compressed_len += 3 + opt_len;
2195 Assert(compressed_len == bits_sent, "bad compressed size");
2200 compressed_len += 7; /* align on byte boundary */
2202 Tracev((stderr, "\ncomprlen %lu(%lu) ", compressed_len >> 3,
2203 compressed_len - 7 * eof));
2205 return compressed_len >> 3;
2208 /* ===========================================================================
2209 * Save the match info and tally the frequency counts. Return true if
2210 * the current block must be flushed.
2212 static int ct_tally(int dist, int lc)
2214 l_buf[last_lit++] = (uch) lc;
2216 /* lc is the unmatched char */
2217 dyn_ltree[lc].Freq++;
2219 /* Here, lc is the match length - MIN_MATCH */
2220 dist--; /* dist = match distance - 1 */
2221 Assert((ush) dist < (ush) MAX_DIST &&
2222 (ush) lc <= (ush) (MAX_MATCH - MIN_MATCH) &&
2223 (ush) d_code(dist) < (ush) D_CODES, "ct_tally: bad match");
2225 dyn_ltree[length_code[lc] + LITERALS + 1].Freq++;
2226 dyn_dtree[d_code(dist)].Freq++;
2228 d_buf[last_dist++] = (ush) dist;
2233 /* Output the flags if they fill a byte: */
2234 if ((last_lit & 7) == 0) {
2235 flag_buf[last_flags++] = flags;
2236 flags = 0, flag_bit = 1;
2238 /* Try to guess if it is profitable to stop the current block here */
2239 if ((last_lit & 0xfff) == 0) {
2240 /* Compute an upper bound for the compressed length */
2241 ulg out_length = (ulg) last_lit * 8L;
2242 ulg in_length = (ulg) strstart - block_start;
2245 for (dcode = 0; dcode < D_CODES; dcode++) {
2247 (ulg) dyn_dtree[dcode].Freq * (5L + extra_dbits[dcode]);
2251 "\nlast_lit %u, last_dist %u, in %ld, out ~%ld(%ld%%) ",
2252 last_lit, last_dist, in_length, out_length,
2253 100L - out_length * 100L / in_length));
2254 if (last_dist < last_lit / 2 && out_length < in_length / 2)
2257 return (last_lit == LIT_BUFSIZE - 1 || last_dist == DIST_BUFSIZE);
2258 /* We avoid equality with LIT_BUFSIZE because of wraparound at 64K
2259 * on 16 bit machines and because stored blocks are restricted to
2264 /* ===========================================================================
2265 * Send the block data compressed using the given Huffman trees
2267 static void compress_block(ct_data * ltree, ct_data * dtree)
2269 unsigned dist; /* distance of matched string */
2270 int lc; /* match length or unmatched char (if dist == 0) */
2271 unsigned lx = 0; /* running index in l_buf */
2272 unsigned dx = 0; /* running index in d_buf */
2273 unsigned fx = 0; /* running index in flag_buf */
2274 uch flag = 0; /* current flags */
2275 unsigned code; /* the code to send */
2276 int extra; /* number of extra bits to send */
2281 flag = flag_buf[fx++];
2283 if ((flag & 1) == 0) {
2284 send_code(lc, ltree); /* send a literal byte */
2285 Tracecv(isgraph(lc), (stderr, " '%c' ", lc));
2287 /* Here, lc is the match length - MIN_MATCH */
2288 code = length_code[lc];
2289 send_code(code + LITERALS + 1, ltree); /* send the length code */
2290 extra = extra_lbits[code];
2292 lc -= base_length[code];
2293 send_bits(lc, extra); /* send the extra length bits */
2296 /* Here, dist is the match distance - 1 */
2297 code = d_code(dist);
2298 Assert(code < D_CODES, "bad d_code");
2300 send_code(code, dtree); /* send the distance code */
2301 extra = extra_dbits[code];
2303 dist -= base_dist[code];
2304 send_bits(dist, extra); /* send the extra distance bits */
2306 } /* literal or match pair ? */
2308 } while (lx < last_lit);
2310 send_code(END_BLOCK, ltree);
2313 /* ===========================================================================
2314 * Set the file type to ASCII or BINARY, using a crude approximation:
2315 * binary if more than 20% of the bytes are <= 6 or >= 128, ascii otherwise.
2316 * IN assertion: the fields freq of dyn_ltree are set and the total of all
2317 * frequencies does not exceed 64K (to fit in an int on 16 bit machines).
2319 static void set_file_type(void)
2322 unsigned ascii_freq = 0;
2323 unsigned bin_freq = 0;
2326 bin_freq += dyn_ltree[n++].Freq;
2328 ascii_freq += dyn_ltree[n++].Freq;
2329 while (n < LITERALS)
2330 bin_freq += dyn_ltree[n++].Freq;
2331 *file_type = bin_freq > (ascii_freq >> 2) ? BINARY : ASCII;
2332 if (*file_type == BINARY && translate_eol) {
2333 bb_error_msg("-l used on binary file");
2337 /* zip.c -- compress files to the gzip or pkzip format
2338 * Copyright (C) 1992-1993 Jean-loup Gailly
2339 * This is free software; you can redistribute it and/or modify it under the
2340 * terms of the GNU General Public License, see the file COPYING.
2344 static uint32_t crc; /* crc on uncompressed file data */
2346 /* ===========================================================================
2347 * Deflate in to out.
2348 * IN assertions: the input and output buffers are cleared.
2349 * The variables time_stamp and save_orig_name are initialized.
2351 static int zip(int in, int out)
2353 uch my_flags = 0; /* general purpose bit flags */
2354 ush attr = 0; /* ascii/binary flag */
2355 ush deflate_flags = 0; /* pkzip -es, -en or -ex equivalent */
2361 /* Write the header to the gzip file. See algorithm.doc for the format */
2364 put_header_byte(GZIP_MAGIC[0]); /* magic header */
2365 put_header_byte(GZIP_MAGIC[1]);
2366 put_header_byte(DEFLATED); /* compression method */
2368 put_header_byte(my_flags); /* general flags */
2369 put_32bit(time_stamp);
2371 /* Write deflated file to zip file */
2375 ct_init(&attr, &method);
2376 lm_init(&deflate_flags);
2378 put_8bit((uch) deflate_flags); /* extra flags */
2379 put_8bit(3); /* OS identifier = 3 (Unix) */
2383 /* Write the crc and uncompressed size */
2392 /* ===========================================================================
2393 * Read a new buffer from the current input file, perform end-of-line
2394 * translation, and update the crc and input file size.
2395 * IN assertion: size >= 2 (for end-of-line translation)
2397 static int file_read(char *buf, unsigned size)
2401 Assert(insize == 0, "inbuf not empty");
2403 len = read(ifd, buf, size);
2404 if (len == (unsigned) (-1) || len == 0)
2407 crc = updcrc((uch *) buf, len);
2412 /* ===========================================================================
2413 * Write the output buffer outbuf[0..outcnt-1] and update bytes_out.
2414 * (used for the compressed data only)
2416 static void flush_outbuf(void)
2421 write_buf(ofd, (char *) outbuf, outcnt);