aboutsummaryrefslogtreecommitdiff
path: root/gzip.c
diff options
context:
space:
mode:
Diffstat (limited to 'gzip.c')
-rw-r--r--gzip.c2568
1 files changed, 0 insertions, 2568 deletions
diff --git a/gzip.c b/gzip.c
deleted file mode 100644
index 5c86c1070..000000000
--- a/gzip.c
+++ /dev/null
@@ -1,2568 +0,0 @@
1/* vi: set sw=4 ts=4: */
2/*
3 * Gzip implementation for busybox
4 *
5 * Based on GNU gzip Copyright (C) 1992-1993 Jean-loup Gailly.
6 *
7 * Originally adjusted for busybox by Charles P. Wright <cpw@unix.asb.com>
8 * "this is a stripped down version of gzip I put into busybox, it does
9 * only standard in to standard out with -9 compression. It also requires
10 * the zcat module for some important functions."
11 *
12 * Adjusted further by Erik Andersen <andersen@lineo.com>, <andersee@debian.org>
13 * to support files as well as stdin/stdout, and to generally behave itself wrt
14 * command line handling.
15 *
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.
20 *
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.
25 *
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
29 *
30 */
31
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.*/
35#define SMALL_MEM
36#define DYN_ALLOC
37
38#include <stdlib.h>
39#include <stdio.h>
40#include <string.h>
41#include <unistd.h>
42#include <errno.h>
43#include <sys/types.h>
44#include <signal.h>
45#include <utime.h>
46#include <ctype.h>
47#include <sys/types.h>
48#include <unistd.h>
49#include <dirent.h>
50#include <fcntl.h>
51#include <time.h>
52#include "busybox.h"
53
54#define memzero(s, n) memset ((void *)(s), 0, (n))
55
56#ifndef RETSIGTYPE
57# define RETSIGTYPE void
58#endif
59
60typedef unsigned char uch;
61typedef unsigned short ush;
62typedef unsigned long ulg;
63
64/* Return codes from gzip */
65#define OK 0
66#define ERROR 1
67#define WARNING 2
68
69/* Compression methods (see algorithm.doc) */
70/* Only STORED and DEFLATED are supported by this BusyBox module */
71#define STORED 0
72/* methods 4 to 7 reserved */
73#define DEFLATED 8
74static int method; /* compression method */
75
76/* To save memory for 16 bit systems, some arrays are overlaid between
77 * the various modules:
78 * deflate: prev+head window d_buf l_buf outbuf
79 * unlzw: tab_prefix tab_suffix stack inbuf outbuf
80 * For compression, input is done in window[]. For decompression, output
81 * is done in window except for unlzw.
82 */
83
84#ifndef INBUFSIZ
85# ifdef SMALL_MEM
86# define INBUFSIZ 0x2000 /* input buffer size */
87# else
88# define INBUFSIZ 0x8000 /* input buffer size */
89# endif
90#endif
91#define INBUF_EXTRA 64 /* required by unlzw() */
92
93#ifndef OUTBUFSIZ
94# ifdef SMALL_MEM
95# define OUTBUFSIZ 8192 /* output buffer size */
96# else
97# define OUTBUFSIZ 16384 /* output buffer size */
98# endif
99#endif
100#define OUTBUF_EXTRA 2048 /* required by unlzw() */
101
102#ifndef DIST_BUFSIZE
103# ifdef SMALL_MEM
104# define DIST_BUFSIZE 0x2000 /* buffer for distances, see trees.c */
105# else
106# define DIST_BUFSIZE 0x8000 /* buffer for distances, see trees.c */
107# endif
108#endif
109
110#ifdef DYN_ALLOC
111# define DECLARE(type, array, size) static type * array
112# define ALLOC(type, array, size) { \
113 array = (type*)xcalloc((size_t)(((size)+1L)/2), 2*sizeof(type)); \
114 }
115# define FREE(array) {if (array != NULL) free(array), array=NULL;}
116#else
117# define DECLARE(type, array, size) static type array[size]
118# define ALLOC(type, array, size)
119# define FREE(array)
120#endif
121
122#define tab_suffix window
123#define tab_prefix prev /* hash link (see deflate.c) */
124#define head (prev+WSIZE) /* hash head (see deflate.c) */
125
126static long bytes_in; /* number of input bytes */
127
128#define isize bytes_in
129/* for compatibility with old zip sources (to be cleaned) */
130
131typedef int file_t; /* Do not use stdio */
132
133#define NO_FILE (-1) /* in memory compression */
134
135
136#define PACK_MAGIC "\037\036" /* Magic header for packed files */
137#define GZIP_MAGIC "\037\213" /* Magic header for gzip files, 1F 8B */
138#define OLD_GZIP_MAGIC "\037\236" /* Magic header for gzip 0.5 = freeze 1.x */
139#define LZH_MAGIC "\037\240" /* Magic header for SCO LZH Compress files */
140#define PKZIP_MAGIC "\120\113\003\004" /* Magic header for pkzip files */
141
142/* gzip flag byte */
143#define ASCII_FLAG 0x01 /* bit 0 set: file probably ascii text */
144#define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */
145#define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
146#define ORIG_NAME 0x08 /* bit 3 set: original file name present */
147#define COMMENT 0x10 /* bit 4 set: file comment present */
148#define RESERVED 0xC0 /* bit 6,7: reserved */
149
150/* internal file attribute */
151#define UNKNOWN 0xffff
152#define BINARY 0
153#define ASCII 1
154
155#ifndef WSIZE
156# define WSIZE 0x8000 /* window size--must be a power of two, and */
157#endif /* at least 32K for zip's deflate method */
158
159#define MIN_MATCH 3
160#define MAX_MATCH 258
161/* The minimum and maximum match lengths */
162
163#define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
164/* Minimum amount of lookahead, except at the end of the input file.
165 * See deflate.c for comments about the MIN_MATCH+1.
166 */
167
168#define MAX_DIST (WSIZE-MIN_LOOKAHEAD)
169/* In order to simplify the code, particularly on 16 bit machines, match
170 * distances are limited to MAX_DIST instead of WSIZE.
171 */
172
173/* put_byte is used for the compressed output */
174#define put_byte(c) {outbuf[outcnt++]=(uch)(c); if (outcnt==OUTBUFSIZ)\
175 flush_outbuf();}
176
177/* Output a 16 bit value, lsb first */
178#define put_short(w) \
179{ if (outcnt < OUTBUFSIZ-2) { \
180 outbuf[outcnt++] = (uch) ((w) & 0xff); \
181 outbuf[outcnt++] = (uch) ((ush)(w) >> 8); \
182 } else { \
183 put_short_when_full(w); \
184 } \
185}
186
187/* Output a 32 bit value to the bit stream, lsb first */
188#if 0
189#define put_long(n) { \
190 put_short((n) & 0xffff); \
191 put_short(((ulg)(n)) >> 16); \
192}
193#endif
194
195#define seekable() 0 /* force sequential output */
196#define translate_eol 0 /* no option -a yet */
197
198/* Diagnostic functions */
199#ifdef DEBUG
200# define Assert(cond,msg) {if(!(cond)) error_msg(msg);}
201# define Trace(x) fprintf x
202# define Tracev(x) {if (verbose) fprintf x ;}
203# define Tracevv(x) {if (verbose>1) fprintf x ;}
204# define Tracec(c,x) {if (verbose && (c)) fprintf x ;}
205# define Tracecv(c,x) {if (verbose>1 && (c)) fprintf x ;}
206#else
207# define Assert(cond,msg)
208# define Trace(x)
209# define Tracev(x)
210# define Tracevv(x)
211# define Tracec(c,x)
212# define Tracecv(c,x)
213#endif
214
215#define WARN(msg) {if (!quiet) fprintf msg ; \
216 if (exit_code == OK) exit_code = WARNING;}
217
218#ifndef MAX_PATH_LEN
219# define MAX_PATH_LEN 1024 /* max pathname length */
220#endif
221
222
223 /* from zip.c: */
224static int zip (int in, int out);
225static int file_read (char *buf, unsigned size);
226
227 /* from gzip.c */
228static RETSIGTYPE abort_gzip (void);
229
230 /* from deflate.c */
231static void lm_init (ush * flags);
232static ulg deflate (void);
233
234 /* from trees.c */
235static void ct_init (ush * attr, int *methodp);
236static int ct_tally (int dist, int lc);
237static ulg flush_block (char *buf, ulg stored_len, int eof);
238
239 /* from bits.c */
240static void bi_init (file_t zipfile);
241static void send_bits (int value, int length);
242static unsigned bi_reverse (unsigned value, int length);
243static void bi_windup (void);
244static void copy_block (char *buf, unsigned len, int header);
245static int (*read_buf) (char *buf, unsigned size);
246
247 /* from util.c: */
248static void flush_outbuf (void);
249
250static void put_short_when_full (ush);
251
252
253/* lzw.h -- define the lzw functions.
254 * Copyright (C) 1992-1993 Jean-loup Gailly.
255 * This is free software; you can redistribute it and/or modify it under the
256 * terms of the GNU General Public License, see the file COPYING.
257 */
258
259#if !defined(OF) && defined(lint)
260# include "gzip.h"
261#endif
262
263#ifndef BITS
264# define BITS 16
265#endif
266#define INIT_BITS 9 /* Initial number of bits per code */
267
268#define BIT_MASK 0x1f /* Mask for 'number of compression bits' */
269/* Mask 0x20 is reserved to mean a fourth header byte, and 0x40 is free.
270 * It's a pity that old uncompress does not check bit 0x20. That makes
271 * extension of the format actually undesirable because old compress
272 * would just crash on the new format instead of giving a meaningful
273 * error message. It does check the number of bits, but it's more
274 * helpful to say "unsupported format, get a new version" than
275 * "can only handle 16 bits".
276 */
277
278/* tailor.h -- target dependent definitions
279 * Copyright (C) 1992-1993 Jean-loup Gailly.
280 * This is free software; you can redistribute it and/or modify it under the
281 * terms of the GNU General Public License, see the file COPYING.
282 */
283
284/* The target dependent definitions should be defined here only.
285 * The target dependent functions should be defined in tailor.c.
286 */
287
288
289 /* Common defaults */
290
291#ifndef OS_CODE
292# define OS_CODE 0x03 /* assume Unix */
293#endif
294
295#ifndef PATH_SEP
296# define PATH_SEP '/'
297#endif
298
299#ifndef OPTIONS_VAR
300# define OPTIONS_VAR "GZIP"
301#endif
302
303#ifndef Z_SUFFIX
304# define Z_SUFFIX ".gz"
305#endif
306
307#ifdef MAX_EXT_CHARS
308# define MAX_SUFFIX MAX_EXT_CHARS
309#else
310# define MAX_SUFFIX 30
311#endif
312
313 /* global buffers */
314
315DECLARE(uch, inbuf, INBUFSIZ + INBUF_EXTRA);
316DECLARE(uch, outbuf, OUTBUFSIZ + OUTBUF_EXTRA);
317DECLARE(ush, d_buf, DIST_BUFSIZE);
318DECLARE(uch, window, 2L * WSIZE);
319DECLARE(ush, tab_prefix, 1L << BITS);
320
321static int crc_table_empty = 1;
322
323static int foreground; /* set if program run in foreground */
324static int method = DEFLATED; /* compression method */
325static int exit_code = OK; /* program exit code */
326static int part_nb; /* number of parts in .gz file */
327static long time_stamp; /* original time stamp (modification time) */
328static long ifile_size; /* input file size, -1 for devices (debug only) */
329static char z_suffix[MAX_SUFFIX + 1]; /* default suffix (can be set with --suffix) */
330static int z_len; /* strlen(z_suffix) */
331
332static char ifname[MAX_PATH_LEN]; /* input file name */
333static char ofname[MAX_PATH_LEN]; /* output file name */
334static int ifd; /* input file descriptor */
335static int ofd; /* output file descriptor */
336static unsigned insize; /* valid bytes in inbuf */
337static unsigned outcnt; /* bytes in output buffer */
338
339/* ========================================================================
340 * Signal and error handler.
341 */
342static void abort_gzip()
343{
344 exit(ERROR);
345}
346
347/* ===========================================================================
348 * Clear input and output buffers
349 */
350static void clear_bufs(void)
351{
352 outcnt = 0;
353 insize = 0;
354 bytes_in = 0L;
355}
356
357static void write_error_msg(void)
358{
359 fprintf(stderr, "\n");
360 perror("");
361 abort_gzip();
362}
363
364/* ===========================================================================
365 * Does the same as write(), but also handles partial pipe writes and checks
366 * for error return.
367 */
368static void write_buf(int fd, void *buf, unsigned cnt)
369{
370 unsigned n;
371
372 while ((n = write(fd, buf, cnt)) != cnt) {
373 if (n == (unsigned) (-1)) {
374 write_error_msg();
375 }
376 cnt -= n;
377 buf = (void *) ((char *) buf + n);
378 }
379}
380
381/* ===========================================================================
382 * Run a set of bytes through the crc shift register. If s is a NULL
383 * pointer, then initialize the crc shift register contents instead.
384 * Return the current crc in either case.
385 */
386static ulg updcrc(uch *s, unsigned n)
387{
388 static ulg crc = (ulg) 0xffffffffL; /* shift register contents */
389 register ulg c; /* temporary variable */
390 static unsigned long crc_32_tab[256];
391 if (crc_table_empty) {
392 unsigned long csr; /* crc shift register */
393 const unsigned long e = 0xedb88320L; /* polynomial exclusive-or pattern */
394 int i; /* counter for all possible eight bit values */
395 int k; /* byte being shifted into crc apparatus */
396
397 /* Compute table of CRC's. */
398 crc_32_tab[0] = 0x00000000L;
399 for (i = 1; i < 256; i++) {
400 csr = i;
401 /* The idea to initialize the register with the byte instead of
402 * zero was stolen from Haruhiko Okumura's ar002
403 */
404 for (k = 8; k; k--)
405 csr = csr & 1 ? (csr >> 1) ^ e : csr >> 1;
406 crc_32_tab[i]=csr;
407 }
408 }
409
410 if (s == NULL) {
411 c = 0xffffffffL;
412 } else {
413 c = crc;
414 if (n)
415 do {
416 c = crc_32_tab[((int) c ^ (*s++)) & 0xff] ^ (c >> 8);
417 } while (--n);
418 }
419 crc = c;
420 return c ^ 0xffffffffL; /* (instead of ~c for 64-bit machines) */
421}
422
423/* bits.c -- output variable-length bit strings
424 * Copyright (C) 1992-1993 Jean-loup Gailly
425 * This is free software; you can redistribute it and/or modify it under the
426 * terms of the GNU General Public License, see the file COPYING.
427 */
428
429
430/*
431 * PURPOSE
432 *
433 * Output variable-length bit strings. Compression can be done
434 * to a file or to memory. (The latter is not supported in this version.)
435 *
436 * DISCUSSION
437 *
438 * The PKZIP "deflate" file format interprets compressed file data
439 * as a sequence of bits. Multi-bit strings in the file may cross
440 * byte boundaries without restriction.
441 *
442 * The first bit of each byte is the low-order bit.
443 *
444 * The routines in this file allow a variable-length bit value to
445 * be output right-to-left (useful for literal values). For
446 * left-to-right output (useful for code strings from the tree routines),
447 * the bits must have been reversed first with bi_reverse().
448 *
449 * For in-memory compression, the compressed bit stream goes directly
450 * into the requested output buffer. The input data is read in blocks
451 * by the mem_read() function. The buffer is limited to 64K on 16 bit
452 * machines.
453 *
454 * INTERFACE
455 *
456 * void bi_init (FILE *zipfile)
457 * Initialize the bit string routines.
458 *
459 * void send_bits (int value, int length)
460 * Write out a bit string, taking the source bits right to
461 * left.
462 *
463 * int bi_reverse (int value, int length)
464 * Reverse the bits of a bit string, taking the source bits left to
465 * right and emitting them right to left.
466 *
467 * void bi_windup (void)
468 * Write out any remaining bits in an incomplete byte.
469 *
470 * void copy_block(char *buf, unsigned len, int header)
471 * Copy a stored block to the zip file, storing first the length and
472 * its one's complement if requested.
473 *
474 */
475
476/* ===========================================================================
477 * Local data used by the "bit string" routines.
478 */
479
480static file_t zfile; /* output gzip file */
481
482static unsigned short bi_buf;
483
484/* Output buffer. bits are inserted starting at the bottom (least significant
485 * bits).
486 */
487
488#define Buf_size (8 * 2*sizeof(char))
489/* Number of bits used within bi_buf. (bi_buf might be implemented on
490 * more than 16 bits on some systems.)
491 */
492
493static int bi_valid;
494
495/* Current input function. Set to mem_read for in-memory compression */
496
497#ifdef DEBUG
498ulg bits_sent; /* bit length of the compressed data */
499#endif
500
501/* ===========================================================================
502 * Initialize the bit string routines.
503 */
504static void bi_init(file_t zipfile)
505{
506 zfile = zipfile;
507 bi_buf = 0;
508 bi_valid = 0;
509#ifdef DEBUG
510 bits_sent = 0L;
511#endif
512
513 /* Set the defaults for file compression. They are set by memcompress
514 * for in-memory compression.
515 */
516 if (zfile != NO_FILE) {
517 read_buf = file_read;
518 }
519}
520
521/* ===========================================================================
522 * Send a value on a given number of bits.
523 * IN assertion: length <= 16 and value fits in length bits.
524 */
525static void send_bits(int value, int length)
526{
527#ifdef DEBUG
528 Tracev((stderr, " l %2d v %4x ", length, value));
529 Assert(length > 0 && length <= 15, "invalid length");
530 bits_sent += (ulg) length;
531#endif
532 /* If not enough room in bi_buf, use (valid) bits from bi_buf and
533 * (16 - bi_valid) bits from value, leaving (width - (16-bi_valid))
534 * unused bits in value.
535 */
536 if (bi_valid > (int) Buf_size - length) {
537 bi_buf |= (value << bi_valid);
538 put_short(bi_buf);
539 bi_buf = (ush) value >> (Buf_size - bi_valid);
540 bi_valid += length - Buf_size;
541 } else {
542 bi_buf |= value << bi_valid;
543 bi_valid += length;
544 }
545}
546
547/* ===========================================================================
548 * Reverse the first len bits of a code, using straightforward code (a faster
549 * method would use a table)
550 * IN assertion: 1 <= len <= 15
551 */
552static unsigned bi_reverse(unsigned code, int len)
553{
554 register unsigned res = 0;
555
556 do {
557 res |= code & 1;
558 code >>= 1, res <<= 1;
559 } while (--len > 0);
560 return res >> 1;
561}
562
563/* ===========================================================================
564 * Write out any remaining bits in an incomplete byte.
565 */
566static void bi_windup()
567{
568 if (bi_valid > 8) {
569 put_short(bi_buf);
570 } else if (bi_valid > 0) {
571 put_byte(bi_buf);
572 }
573 bi_buf = 0;
574 bi_valid = 0;
575#ifdef DEBUG
576 bits_sent = (bits_sent + 7) & ~7;
577#endif
578}
579
580/* ===========================================================================
581 * Copy a stored block to the zip file, storing first the length and its
582 * one's complement if requested.
583 */
584static void copy_block(char *buf, unsigned len, int header)
585{
586 bi_windup(); /* align on byte boundary */
587
588 if (header) {
589 put_short((ush) len);
590 put_short((ush) ~ len);
591#ifdef DEBUG
592 bits_sent += 2 * 16;
593#endif
594 }
595#ifdef DEBUG
596 bits_sent += (ulg) len << 3;
597#endif
598 while (len--) {
599 put_byte(*buf++);
600 }
601}
602
603/* deflate.c -- compress data using the deflation algorithm
604 * Copyright (C) 1992-1993 Jean-loup Gailly
605 * This is free software; you can redistribute it and/or modify it under the
606 * terms of the GNU General Public License, see the file COPYING.
607 */
608
609/*
610 * PURPOSE
611 *
612 * Identify new text as repetitions of old text within a fixed-
613 * length sliding window trailing behind the new text.
614 *
615 * DISCUSSION
616 *
617 * The "deflation" process depends on being able to identify portions
618 * of the input text which are identical to earlier input (within a
619 * sliding window trailing behind the input currently being processed).
620 *
621 * The most straightforward technique turns out to be the fastest for
622 * most input files: try all possible matches and select the longest.
623 * The key feature of this algorithm is that insertions into the string
624 * dictionary are very simple and thus fast, and deletions are avoided
625 * completely. Insertions are performed at each input character, whereas
626 * string matches are performed only when the previous match ends. So it
627 * is preferable to spend more time in matches to allow very fast string
628 * insertions and avoid deletions. The matching algorithm for small
629 * strings is inspired from that of Rabin & Karp. A brute force approach
630 * is used to find longer strings when a small match has been found.
631 * A similar algorithm is used in comic (by Jan-Mark Wams) and freeze
632 * (by Leonid Broukhis).
633 * A previous version of this file used a more sophisticated algorithm
634 * (by Fiala and Greene) which is guaranteed to run in linear amortized
635 * time, but has a larger average cost, uses more memory and is patented.
636 * However the F&G algorithm may be faster for some highly redundant
637 * files if the parameter max_chain_length (described below) is too large.
638 *
639 * ACKNOWLEDGEMENTS
640 *
641 * The idea of lazy evaluation of matches is due to Jan-Mark Wams, and
642 * I found it in 'freeze' written by Leonid Broukhis.
643 * Thanks to many info-zippers for bug reports and testing.
644 *
645 * REFERENCES
646 *
647 * APPNOTE.TXT documentation file in PKZIP 1.93a distribution.
648 *
649 * A description of the Rabin and Karp algorithm is given in the book
650 * "Algorithms" by R. Sedgewick, Addison-Wesley, p252.
651 *
652 * Fiala,E.R., and Greene,D.H.
653 * Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595
654 *
655 * INTERFACE
656 *
657 * void lm_init (int pack_level, ush *flags)
658 * Initialize the "longest match" routines for a new file
659 *
660 * ulg deflate (void)
661 * Processes a new input file and return its compressed length. Sets
662 * the compressed length, crc, deflate flags and internal file
663 * attributes.
664 */
665
666
667/* ===========================================================================
668 * Configuration parameters
669 */
670
671/* Compile with MEDIUM_MEM to reduce the memory requirements or
672 * with SMALL_MEM to use as little memory as possible. Use BIG_MEM if the
673 * entire input file can be held in memory (not possible on 16 bit systems).
674 * Warning: defining these symbols affects HASH_BITS (see below) and thus
675 * affects the compression ratio. The compressed output
676 * is still correct, and might even be smaller in some cases.
677 */
678
679#ifdef SMALL_MEM
680# define HASH_BITS 13 /* Number of bits used to hash strings */
681#endif
682#ifdef MEDIUM_MEM
683# define HASH_BITS 14
684#endif
685#ifndef HASH_BITS
686# define HASH_BITS 15
687 /* For portability to 16 bit machines, do not use values above 15. */
688#endif
689
690/* To save space (see unlzw.c), we overlay prev+head with tab_prefix and
691 * window with tab_suffix. Check that we can do this:
692 */
693#if (WSIZE<<1) > (1<<BITS)
694# error cannot overlay window with tab_suffix and prev with tab_prefix0
695#endif
696#if HASH_BITS > BITS-1
697# error cannot overlay head with tab_prefix1
698#endif
699#define HASH_SIZE (unsigned)(1<<HASH_BITS)
700#define HASH_MASK (HASH_SIZE-1)
701#define WMASK (WSIZE-1)
702/* HASH_SIZE and WSIZE must be powers of two */
703#define NIL 0
704/* Tail of hash chains */
705#define FAST 4
706#define SLOW 2
707/* speed options for the general purpose bit flag */
708#ifndef TOO_FAR
709# define TOO_FAR 4096
710#endif
711/* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
712/* ===========================================================================
713 * Local data used by the "longest match" routines.
714 */
715typedef ush Pos;
716typedef unsigned IPos;
717
718/* A Pos is an index in the character window. We use short instead of int to
719 * save space in the various tables. IPos is used only for parameter passing.
720 */
721
722/* DECLARE(uch, window, 2L*WSIZE); */
723/* Sliding window. Input bytes are read into the second half of the window,
724 * and move to the first half later to keep a dictionary of at least WSIZE
725 * bytes. With this organization, matches are limited to a distance of
726 * WSIZE-MAX_MATCH bytes, but this ensures that IO is always
727 * performed with a length multiple of the block size. Also, it limits
728 * the window size to 64K, which is quite useful on MSDOS.
729 * To do: limit the window size to WSIZE+BSZ if SMALL_MEM (the code would
730 * be less efficient).
731 */
732
733/* DECLARE(Pos, prev, WSIZE); */
734/* Link to older string with same hash index. To limit the size of this
735 * array to 64K, this link is maintained only for the last 32K strings.
736 * An index in this array is thus a window index modulo 32K.
737 */
738
739/* DECLARE(Pos, head, 1<<HASH_BITS); */
740/* Heads of the hash chains or NIL. */
741
742static const ulg window_size = (ulg) 2 * WSIZE;
743
744/* window size, 2*WSIZE except for MMAP or BIG_MEM, where it is the
745 * input file length plus MIN_LOOKAHEAD.
746 */
747
748static long block_start;
749
750/* window position at the beginning of the current output block. Gets
751 * negative when the window is moved backwards.
752 */
753
754static unsigned ins_h; /* hash index of string to be inserted */
755
756#define H_SHIFT ((HASH_BITS+MIN_MATCH-1)/MIN_MATCH)
757/* Number of bits by which ins_h and del_h must be shifted at each
758 * input step. It must be such that after MIN_MATCH steps, the oldest
759 * byte no longer takes part in the hash key, that is:
760 * H_SHIFT * MIN_MATCH >= HASH_BITS
761 */
762
763static unsigned int prev_length;
764
765/* Length of the best match at previous step. Matches not greater than this
766 * are discarded. This is used in the lazy match evaluation.
767 */
768
769static unsigned strstart; /* start of string to insert */
770static unsigned match_start; /* start of matching string */
771static int eofile; /* flag set at end of input file */
772static unsigned lookahead; /* number of valid bytes ahead in window */
773
774static const unsigned max_chain_length=4096;
775
776/* To speed up deflation, hash chains are never searched beyond this length.
777 * A higher limit improves compression ratio but degrades the speed.
778 */
779
780static const unsigned int max_lazy_match=258;
781
782/* Attempt to find a better match only when the current match is strictly
783 * smaller than this value. This mechanism is used only for compression
784 * levels >= 4.
785 */
786#define max_insert_length max_lazy_match
787/* Insert new strings in the hash table only if the match length
788 * is not greater than this length. This saves time but degrades compression.
789 * max_insert_length is used only for compression levels <= 3.
790 */
791
792static const unsigned good_match=32;
793
794/* Use a faster search when the previous match is longer than this */
795
796
797/* Values for max_lazy_match, good_match and max_chain_length, depending on
798 * the desired pack level (0..9). The values given below have been tuned to
799 * exclude worst case performance for pathological files. Better values may be
800 * found for specific files.
801 */
802
803static const int nice_match=258; /* Stop searching when current match exceeds this */
804
805/* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
806 * For deflate_fast() (levels <= 3) good is ignored and lazy has a different
807 * meaning.
808 */
809
810#define EQUAL 0
811/* result of memcmp for equal strings */
812
813/* ===========================================================================
814 * Prototypes for local functions.
815 */
816static void fill_window (void);
817
818static int longest_match (IPos cur_match);
819
820#ifdef DEBUG
821static void check_match (IPos start, IPos match, int length);
822#endif
823
824/* ===========================================================================
825 * Update a hash value with the given input byte
826 * IN assertion: all calls to to UPDATE_HASH are made with consecutive
827 * input characters, so that a running hash key can be computed from the
828 * previous key instead of complete recalculation each time.
829 */
830#define UPDATE_HASH(h,c) (h = (((h)<<H_SHIFT) ^ (c)) & HASH_MASK)
831
832/* ===========================================================================
833 * Insert string s in the dictionary and set match_head to the previous head
834 * of the hash chain (the most recent string with same hash key). Return
835 * the previous length of the hash chain.
836 * IN assertion: all calls to to INSERT_STRING are made with consecutive
837 * input characters and the first MIN_MATCH bytes of s are valid
838 * (except for the last MIN_MATCH-1 bytes of the input file).
839 */
840#define INSERT_STRING(s, match_head) \
841 (UPDATE_HASH(ins_h, window[(s) + MIN_MATCH-1]), \
842 prev[(s) & WMASK] = match_head = head[ins_h], \
843 head[ins_h] = (s))
844
845/* ===========================================================================
846 * Initialize the "longest match" routines for a new file
847 */
848static void lm_init(ush *flags)
849{
850 register unsigned j;
851
852 /* Initialize the hash table. */
853 memzero((char *) head, HASH_SIZE * sizeof(*head));
854 /* prev will be initialized on the fly */
855
856 *flags |= SLOW;
857 /* ??? reduce max_chain_length for binary files */
858
859 strstart = 0;
860 block_start = 0L;
861
862 lookahead = read_buf((char *) window,
863 sizeof(int) <= 2 ? (unsigned) WSIZE : 2 * WSIZE);
864
865 if (lookahead == 0 || lookahead == (unsigned) EOF) {
866 eofile = 1, lookahead = 0;
867 return;
868 }
869 eofile = 0;
870 /* Make sure that we always have enough lookahead. This is important
871 * if input comes from a device such as a tty.
872 */
873 while (lookahead < MIN_LOOKAHEAD && !eofile)
874 fill_window();
875
876 ins_h = 0;
877 for (j = 0; j < MIN_MATCH - 1; j++)
878 UPDATE_HASH(ins_h, window[j]);
879 /* If lookahead < MIN_MATCH, ins_h is garbage, but this is
880 * not important since only literal bytes will be emitted.
881 */
882}
883
884/* ===========================================================================
885 * Set match_start to the longest match starting at the given string and
886 * return its length. Matches shorter or equal to prev_length are discarded,
887 * in which case the result is equal to prev_length and match_start is
888 * garbage.
889 * IN assertions: cur_match is the head of the hash chain for the current
890 * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
891 */
892
893/* For MSDOS, OS/2 and 386 Unix, an optimized version is in match.asm or
894 * match.s. The code is functionally equivalent, so you can use the C version
895 * if desired.
896 */
897static int longest_match(IPos cur_match)
898{
899 unsigned chain_length = max_chain_length; /* max hash chain length */
900 register uch *scan = window + strstart; /* current string */
901 register uch *match; /* matched string */
902 register int len; /* length of current match */
903 int best_len = prev_length; /* best match length so far */
904 IPos limit =
905
906 strstart > (IPos) MAX_DIST ? strstart - (IPos) MAX_DIST : NIL;
907 /* Stop when cur_match becomes <= limit. To simplify the code,
908 * we prevent matches with the string of window index 0.
909 */
910
911/* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
912 * It is easy to get rid of this optimization if necessary.
913 */
914#if HASH_BITS < 8 || MAX_MATCH != 258
915# error Code too clever
916#endif
917 register uch *strend = window + strstart + MAX_MATCH;
918 register uch scan_end1 = scan[best_len - 1];
919 register uch scan_end = scan[best_len];
920
921 /* Do not waste too much time if we already have a good match: */
922 if (prev_length >= good_match) {
923 chain_length >>= 2;
924 }
925 Assert(strstart <= window_size - MIN_LOOKAHEAD,
926 "insufficient lookahead");
927
928 do {
929 Assert(cur_match < strstart, "no future");
930 match = window + cur_match;
931
932 /* Skip to next match if the match length cannot increase
933 * or if the match length is less than 2:
934 */
935 if (match[best_len] != scan_end ||
936 match[best_len - 1] != scan_end1 ||
937 *match != *scan || *++match != scan[1])
938 continue;
939
940 /* The check at best_len-1 can be removed because it will be made
941 * again later. (This heuristic is not always a win.)
942 * It is not necessary to compare scan[2] and match[2] since they
943 * are always equal when the other bytes match, given that
944 * the hash keys are equal and that HASH_BITS >= 8.
945 */
946 scan += 2, match++;
947
948 /* We check for insufficient lookahead only every 8th comparison;
949 * the 256th check will be made at strstart+258.
950 */
951 do {
952 } while (*++scan == *++match && *++scan == *++match &&
953 *++scan == *++match && *++scan == *++match &&
954 *++scan == *++match && *++scan == *++match &&
955 *++scan == *++match && *++scan == *++match &&
956 scan < strend);
957
958 len = MAX_MATCH - (int) (strend - scan);
959 scan = strend - MAX_MATCH;
960
961 if (len > best_len) {
962 match_start = cur_match;
963 best_len = len;
964 if (len >= nice_match)
965 break;
966 scan_end1 = scan[best_len - 1];
967 scan_end = scan[best_len];
968 }
969 } while ((cur_match = prev[cur_match & WMASK]) > limit
970 && --chain_length != 0);
971
972 return best_len;
973}
974
975#ifdef DEBUG
976/* ===========================================================================
977 * Check that the match at match_start is indeed a match.
978 */
979static void check_match(IPos start, IPos match, int length)
980{
981 /* check that the match is indeed a match */
982 if (memcmp((char *) window + match,
983 (char *) window + start, length) != EQUAL) {
984 fprintf(stderr,
985 " start %d, match %d, length %d\n", start, match, length);
986 error_msg("invalid match");
987 }
988 if (verbose > 1) {
989 fprintf(stderr, "\\[%d,%d]", start - match, length);
990 do {
991 putc(window[start++], stderr);
992 } while (--length != 0);
993 }
994}
995#else
996# define check_match(start, match, length)
997#endif
998
999/* ===========================================================================
1000 * Fill the window when the lookahead becomes insufficient.
1001 * Updates strstart and lookahead, and sets eofile if end of input file.
1002 * IN assertion: lookahead < MIN_LOOKAHEAD && strstart + lookahead > 0
1003 * OUT assertions: at least one byte has been read, or eofile is set;
1004 * file reads are performed for at least two bytes (required for the
1005 * translate_eol option).
1006 */
1007static void fill_window()
1008{
1009 register unsigned n, m;
1010 unsigned more =
1011
1012 (unsigned) (window_size - (ulg) lookahead - (ulg) strstart);
1013 /* Amount of free space at the end of the window. */
1014
1015 /* If the window is almost full and there is insufficient lookahead,
1016 * move the upper half to the lower one to make room in the upper half.
1017 */
1018 if (more == (unsigned) EOF) {
1019 /* Very unlikely, but possible on 16 bit machine if strstart == 0
1020 * and lookahead == 1 (input done one byte at time)
1021 */
1022 more--;
1023 } else if (strstart >= WSIZE + MAX_DIST) {
1024 /* By the IN assertion, the window is not empty so we can't confuse
1025 * more == 0 with more == 64K on a 16 bit machine.
1026 */
1027 Assert(window_size == (ulg) 2 * WSIZE, "no sliding with BIG_MEM");
1028
1029 memcpy((char *) window, (char *) window + WSIZE, (unsigned) WSIZE);
1030 match_start -= WSIZE;
1031 strstart -= WSIZE; /* we now have strstart >= MAX_DIST: */
1032
1033 block_start -= (long) WSIZE;
1034
1035 for (n = 0; n < HASH_SIZE; n++) {
1036 m = head[n];
1037 head[n] = (Pos) (m >= WSIZE ? m - WSIZE : NIL);
1038 }
1039 for (n = 0; n < WSIZE; n++) {
1040 m = prev[n];
1041 prev[n] = (Pos) (m >= WSIZE ? m - WSIZE : NIL);
1042 /* If n is not on any hash chain, prev[n] is garbage but
1043 * its value will never be used.
1044 */
1045 }
1046 more += WSIZE;
1047 }
1048 /* At this point, more >= 2 */
1049 if (!eofile) {
1050 n = read_buf((char *) window + strstart + lookahead, more);
1051 if (n == 0 || n == (unsigned) EOF) {
1052 eofile = 1;
1053 } else {
1054 lookahead += n;
1055 }
1056 }
1057}
1058
1059/* ===========================================================================
1060 * Flush the current block, with given end-of-file flag.
1061 * IN assertion: strstart is set to the end of the current match.
1062 */
1063#define FLUSH_BLOCK(eof) \
1064 flush_block(block_start >= 0L ? (char*)&window[(unsigned)block_start] : \
1065 (char*)NULL, (long)strstart - block_start, (eof))
1066
1067/* ===========================================================================
1068 * Same as above, but achieves better compression. We use a lazy
1069 * evaluation for matches: a match is finally adopted only if there is
1070 * no better match at the next window position.
1071 */
1072static ulg deflate()
1073{
1074 IPos hash_head; /* head of hash chain */
1075 IPos prev_match; /* previous match */
1076 int flush; /* set if current block must be flushed */
1077 int match_available = 0; /* set if previous match exists */
1078 register unsigned match_length = MIN_MATCH - 1; /* length of best match */
1079
1080 /* Process the input block. */
1081 while (lookahead != 0) {
1082 /* Insert the string window[strstart .. strstart+2] in the
1083 * dictionary, and set hash_head to the head of the hash chain:
1084 */
1085 INSERT_STRING(strstart, hash_head);
1086
1087 /* Find the longest match, discarding those <= prev_length.
1088 */
1089 prev_length = match_length, prev_match = match_start;
1090 match_length = MIN_MATCH - 1;
1091
1092 if (hash_head != NIL && prev_length < max_lazy_match &&
1093 strstart - hash_head <= MAX_DIST) {
1094 /* To simplify the code, we prevent matches with the string
1095 * of window index 0 (in particular we have to avoid a match
1096 * of the string with itself at the start of the input file).
1097 */
1098 match_length = longest_match(hash_head);
1099 /* longest_match() sets match_start */
1100 if (match_length > lookahead)
1101 match_length = lookahead;
1102
1103 /* Ignore a length 3 match if it is too distant: */
1104 if (match_length == MIN_MATCH
1105 && strstart - match_start > TOO_FAR) {
1106 /* If prev_match is also MIN_MATCH, match_start is garbage
1107 * but we will ignore the current match anyway.
1108 */
1109 match_length--;
1110 }
1111 }
1112 /* If there was a match at the previous step and the current
1113 * match is not better, output the previous match:
1114 */
1115 if (prev_length >= MIN_MATCH && match_length <= prev_length) {
1116
1117 check_match(strstart - 1, prev_match, prev_length);
1118
1119 flush =
1120 ct_tally(strstart - 1 - prev_match,
1121 prev_length - MIN_MATCH);
1122
1123 /* Insert in hash table all strings up to the end of the match.
1124 * strstart-1 and strstart are already inserted.
1125 */
1126 lookahead -= prev_length - 1;
1127 prev_length -= 2;
1128 do {
1129 strstart++;
1130 INSERT_STRING(strstart, hash_head);
1131 /* strstart never exceeds WSIZE-MAX_MATCH, so there are
1132 * always MIN_MATCH bytes ahead. If lookahead < MIN_MATCH
1133 * these bytes are garbage, but it does not matter since the
1134 * next lookahead bytes will always be emitted as literals.
1135 */
1136 } while (--prev_length != 0);
1137 match_available = 0;
1138 match_length = MIN_MATCH - 1;
1139 strstart++;
1140 if (flush)
1141 FLUSH_BLOCK(0), block_start = strstart;
1142
1143 } else if (match_available) {
1144 /* If there was no match at the previous position, output a
1145 * single literal. If there was a match but the current match
1146 * is longer, truncate the previous match to a single literal.
1147 */
1148 Tracevv((stderr, "%c", window[strstart - 1]));
1149 if (ct_tally(0, window[strstart - 1])) {
1150 FLUSH_BLOCK(0), block_start = strstart;
1151 }
1152 strstart++;
1153 lookahead--;
1154 } else {
1155 /* There is no previous match to compare with, wait for
1156 * the next step to decide.
1157 */
1158 match_available = 1;
1159 strstart++;
1160 lookahead--;
1161 }
1162 Assert(strstart <= isize && lookahead <= isize, "a bit too far");
1163
1164 /* Make sure that we always have enough lookahead, except
1165 * at the end of the input file. We need MAX_MATCH bytes
1166 * for the next match, plus MIN_MATCH bytes to insert the
1167 * string following the next match.
1168 */
1169 while (lookahead < MIN_LOOKAHEAD && !eofile)
1170 fill_window();
1171 }
1172 if (match_available)
1173 ct_tally(0, window[strstart - 1]);
1174
1175 return FLUSH_BLOCK(1); /* eof */
1176}
1177
1178/* gzip (GNU zip) -- compress files with zip algorithm and 'compress' interface
1179 * Copyright (C) 1992-1993 Jean-loup Gailly
1180 * The unzip code was written and put in the public domain by Mark Adler.
1181 * Portions of the lzw code are derived from the public domain 'compress'
1182 * written by Spencer Thomas, Joe Orost, James Woods, Jim McKie, Steve Davies,
1183 * Ken Turkowski, Dave Mack and Peter Jannesen.
1184 *
1185 * See the license_msg below and the file COPYING for the software license.
1186 * See the file algorithm.doc for the compression algorithms and file formats.
1187 */
1188
1189/* Compress files with zip algorithm and 'compress' interface.
1190 * See usage() and help() functions below for all options.
1191 * Outputs:
1192 * file.gz: compressed file with same mode, owner, and utimes
1193 * or stdout with -c option or if stdin used as input.
1194 * If the output file name had to be truncated, the original name is kept
1195 * in the compressed file.
1196 */
1197
1198 /* configuration */
1199
1200typedef struct dirent dir_type;
1201
1202typedef RETSIGTYPE(*sig_type) (int);
1203
1204/* ======================================================================== */
1205// int main (argc, argv)
1206// int argc;
1207// char **argv;
1208int gzip_main(int argc, char **argv)
1209{
1210 int result;
1211 int inFileNum;
1212 int outFileNum;
1213 struct stat statBuf;
1214 char *delFileName;
1215 int tostdout = 0;
1216 int fromstdin = 0;
1217 int force = 0;
1218 int opt;
1219
1220 while ((opt = getopt(argc, argv, "cf123456789dq")) != -1) {
1221 switch (opt) {
1222 case 'c':
1223 tostdout = 1;
1224 break;
1225 case 'f':
1226 force = 1;
1227 break;
1228 /* Ignore 1-9 (compression level) options */
1229 case '1': case '2': case '3': case '4': case '5':
1230 case '6': case '7': case '8': case '9':
1231 break;
1232 case 'q':
1233 break;
1234#ifdef BB_GUNZIP
1235 case 'd':
1236 optind = 1;
1237 return gunzip_main(argc, argv);
1238#endif
1239 default:
1240 show_usage();
1241 }
1242 }
1243 if ((optind == argc) || (strcmp(argv[optind], "-") == 0)) {
1244 fromstdin = 1;
1245 tostdout = 1;
1246 }
1247
1248 if (isatty(fileno(stdout)) && tostdout==1 && force==0)
1249 error_msg_and_die( "compressed data not written to terminal. Use -f to force it.");
1250
1251 foreground = signal(SIGINT, SIG_IGN) != SIG_IGN;
1252 if (foreground) {
1253 (void) signal(SIGINT, (sig_type) abort_gzip);
1254 }
1255#ifdef SIGTERM
1256 if (signal(SIGTERM, SIG_IGN) != SIG_IGN) {
1257 (void) signal(SIGTERM, (sig_type) abort_gzip);
1258 }
1259#endif
1260#ifdef SIGHUP
1261 if (signal(SIGHUP, SIG_IGN) != SIG_IGN) {
1262 (void) signal(SIGHUP, (sig_type) abort_gzip);
1263 }
1264#endif
1265
1266 strncpy(z_suffix, Z_SUFFIX, sizeof(z_suffix) - 1);
1267 z_len = strlen(z_suffix);
1268
1269 /* Allocate all global buffers (for DYN_ALLOC option) */
1270 ALLOC(uch, inbuf, INBUFSIZ + INBUF_EXTRA);
1271 ALLOC(uch, outbuf, OUTBUFSIZ + OUTBUF_EXTRA);
1272 ALLOC(ush, d_buf, DIST_BUFSIZE);
1273 ALLOC(uch, window, 2L * WSIZE);
1274 ALLOC(ush, tab_prefix, 1L << BITS);
1275
1276 if (fromstdin == 1) {
1277 strcpy(ofname, "stdin");
1278
1279 inFileNum = fileno(stdin);
1280 time_stamp = 0; /* time unknown by default */
1281 ifile_size = -1L; /* convention for unknown size */
1282 } else {
1283 /* Open up the input file */
1284 strncpy(ifname, argv[optind], MAX_PATH_LEN);
1285
1286 /* Open input file */
1287 inFileNum = open(ifname, O_RDONLY);
1288 if (inFileNum < 0 || stat(ifname, &statBuf) < 0)
1289 perror_msg_and_die("%s", ifname);
1290 /* Get the time stamp on the input file. */
1291 time_stamp = statBuf.st_ctime;
1292 ifile_size = statBuf.st_size;
1293 }
1294
1295
1296 if (tostdout == 1) {
1297 /* And get to work */
1298 strcpy(ofname, "stdout");
1299 outFileNum = fileno(stdout);
1300
1301 clear_bufs(); /* clear input and output buffers */
1302 part_nb = 0;
1303
1304 /* Actually do the compression/decompression. */
1305 zip(inFileNum, outFileNum);
1306
1307 } else {
1308
1309 /* And get to work */
1310 strncpy(ofname, ifname, MAX_PATH_LEN - 4);
1311 strcat(ofname, ".gz");
1312
1313
1314 /* Open output fille */
1315#if (__GLIBC__ >= 2) && (__GLIBC_MINOR__ >= 1)
1316 outFileNum = open(ofname, O_RDWR | O_CREAT | O_EXCL | O_NOFOLLOW);
1317#else
1318 outFileNum = open(ofname, O_RDWR | O_CREAT | O_EXCL);
1319#endif
1320 if (outFileNum < 0)
1321 perror_msg_and_die("%s", ofname);
1322 /* Set permissions on the file */
1323 fchmod(outFileNum, statBuf.st_mode);
1324
1325 clear_bufs(); /* clear input and output buffers */
1326 part_nb = 0;
1327
1328 /* Actually do the compression/decompression. */
1329 result = zip(inFileNum, outFileNum);
1330 close(outFileNum);
1331 close(inFileNum);
1332 /* Delete the original file */
1333 if (result == OK)
1334 delFileName = ifname;
1335 else
1336 delFileName = ofname;
1337
1338 if (unlink(delFileName) < 0)
1339 perror_msg_and_die("%s", delFileName);
1340 }
1341
1342 return(exit_code);
1343}
1344
1345/* trees.c -- output deflated data using Huffman coding
1346 * Copyright (C) 1992-1993 Jean-loup Gailly
1347 * This is free software; you can redistribute it and/or modify it under the
1348 * terms of the GNU General Public License, see the file COPYING.
1349 */
1350
1351/*
1352 * PURPOSE
1353 *
1354 * Encode various sets of source values using variable-length
1355 * binary code trees.
1356 *
1357 * DISCUSSION
1358 *
1359 * The PKZIP "deflation" process uses several Huffman trees. The more
1360 * common source values are represented by shorter bit sequences.
1361 *
1362 * Each code tree is stored in the ZIP file in a compressed form
1363 * which is itself a Huffman encoding of the lengths of
1364 * all the code strings (in ascending order by source values).
1365 * The actual code strings are reconstructed from the lengths in
1366 * the UNZIP process, as described in the "application note"
1367 * (APPNOTE.TXT) distributed as part of PKWARE's PKZIP program.
1368 *
1369 * REFERENCES
1370 *
1371 * Lynch, Thomas J.
1372 * Data Compression: Techniques and Applications, pp. 53-55.
1373 * Lifetime Learning Publications, 1985. ISBN 0-534-03418-7.
1374 *
1375 * Storer, James A.
1376 * Data Compression: Methods and Theory, pp. 49-50.
1377 * Computer Science Press, 1988. ISBN 0-7167-8156-5.
1378 *
1379 * Sedgewick, R.
1380 * Algorithms, p290.
1381 * Addison-Wesley, 1983. ISBN 0-201-06672-6.
1382 *
1383 * INTERFACE
1384 *
1385 * void ct_init (ush *attr, int *methodp)
1386 * Allocate the match buffer, initialize the various tables and save
1387 * the location of the internal file attribute (ascii/binary) and
1388 * method (DEFLATE/STORE)
1389 *
1390 * void ct_tally (int dist, int lc);
1391 * Save the match info and tally the frequency counts.
1392 *
1393 * long flush_block (char *buf, ulg stored_len, int eof)
1394 * Determine the best encoding for the current block: dynamic trees,
1395 * static trees or store, and output the encoded block to the zip
1396 * file. Returns the total compressed length for the file so far.
1397 *
1398 */
1399
1400/* ===========================================================================
1401 * Constants
1402 */
1403
1404#define MAX_BITS 15
1405/* All codes must not exceed MAX_BITS bits */
1406
1407#define MAX_BL_BITS 7
1408/* Bit length codes must not exceed MAX_BL_BITS bits */
1409
1410#define LENGTH_CODES 29
1411/* number of length codes, not counting the special END_BLOCK code */
1412
1413#define LITERALS 256
1414/* number of literal bytes 0..255 */
1415
1416#define END_BLOCK 256
1417/* end of block literal code */
1418
1419#define L_CODES (LITERALS+1+LENGTH_CODES)
1420/* number of Literal or Length codes, including the END_BLOCK code */
1421
1422#define D_CODES 30
1423/* number of distance codes */
1424
1425#define BL_CODES 19
1426/* number of codes used to transfer the bit lengths */
1427
1428typedef uch extra_bits_t;
1429
1430/* extra bits for each length code */
1431static const extra_bits_t extra_lbits[LENGTH_CODES]
1432 = { 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4,
1433 4, 4, 5, 5, 5, 5, 0 };
1434
1435/* extra bits for each distance code */
1436static const extra_bits_t extra_dbits[D_CODES]
1437 = { 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9,
1438 10, 10, 11, 11, 12, 12, 13, 13 };
1439
1440/* extra bits for each bit length code */
1441static const extra_bits_t extra_blbits[BL_CODES]
1442= { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 3, 7 };
1443
1444#define STORED_BLOCK 0
1445#define STATIC_TREES 1
1446#define DYN_TREES 2
1447/* The three kinds of block type */
1448
1449#ifndef LIT_BUFSIZE
1450# ifdef SMALL_MEM
1451# define LIT_BUFSIZE 0x2000
1452# else
1453# ifdef MEDIUM_MEM
1454# define LIT_BUFSIZE 0x4000
1455# else
1456# define LIT_BUFSIZE 0x8000
1457# endif
1458# endif
1459#endif
1460#ifndef DIST_BUFSIZE
1461# define DIST_BUFSIZE LIT_BUFSIZE
1462#endif
1463/* Sizes of match buffers for literals/lengths and distances. There are
1464 * 4 reasons for limiting LIT_BUFSIZE to 64K:
1465 * - frequencies can be kept in 16 bit counters
1466 * - if compression is not successful for the first block, all input data is
1467 * still in the window so we can still emit a stored block even when input
1468 * comes from standard input. (This can also be done for all blocks if
1469 * LIT_BUFSIZE is not greater than 32K.)
1470 * - if compression is not successful for a file smaller than 64K, we can
1471 * even emit a stored file instead of a stored block (saving 5 bytes).
1472 * - creating new Huffman trees less frequently may not provide fast
1473 * adaptation to changes in the input data statistics. (Take for
1474 * example a binary file with poorly compressible code followed by
1475 * a highly compressible string table.) Smaller buffer sizes give
1476 * fast adaptation but have of course the overhead of transmitting trees
1477 * more frequently.
1478 * - I can't count above 4
1479 * The current code is general and allows DIST_BUFSIZE < LIT_BUFSIZE (to save
1480 * memory at the expense of compression). Some optimizations would be possible
1481 * if we rely on DIST_BUFSIZE == LIT_BUFSIZE.
1482 */
1483#if LIT_BUFSIZE > INBUFSIZ
1484error cannot overlay l_buf and inbuf
1485#endif
1486#define REP_3_6 16
1487/* repeat previous bit length 3-6 times (2 bits of repeat count) */
1488#define REPZ_3_10 17
1489/* repeat a zero length 3-10 times (3 bits of repeat count) */
1490#define REPZ_11_138 18
1491/* repeat a zero length 11-138 times (7 bits of repeat count) *//* ===========================================================================
1492 * Local data
1493 *//* Data structure describing a single value and its code string. */ typedef struct ct_data {
1494 union {
1495 ush freq; /* frequency count */
1496 ush code; /* bit string */
1497 } fc;
1498 union {
1499 ush dad; /* father node in Huffman tree */
1500 ush len; /* length of bit string */
1501 } dl;
1502} ct_data;
1503
1504#define Freq fc.freq
1505#define Code fc.code
1506#define Dad dl.dad
1507#define Len dl.len
1508
1509#define HEAP_SIZE (2*L_CODES+1)
1510/* maximum heap size */
1511
1512static ct_data dyn_ltree[HEAP_SIZE]; /* literal and length tree */
1513static ct_data dyn_dtree[2 * D_CODES + 1]; /* distance tree */
1514
1515static ct_data static_ltree[L_CODES + 2];
1516
1517/* The static literal tree. Since the bit lengths are imposed, there is no
1518 * need for the L_CODES extra codes used during heap construction. However
1519 * The codes 286 and 287 are needed to build a canonical tree (see ct_init
1520 * below).
1521 */
1522
1523static ct_data static_dtree[D_CODES];
1524
1525/* The static distance tree. (Actually a trivial tree since all codes use
1526 * 5 bits.)
1527 */
1528
1529static ct_data bl_tree[2 * BL_CODES + 1];
1530
1531/* Huffman tree for the bit lengths */
1532
1533typedef struct tree_desc {
1534 ct_data *dyn_tree; /* the dynamic tree */
1535 ct_data *static_tree; /* corresponding static tree or NULL */
1536 const extra_bits_t *extra_bits; /* extra bits for each code or NULL */
1537 int extra_base; /* base index for extra_bits */
1538 int elems; /* max number of elements in the tree */
1539 int max_length; /* max bit length for the codes */
1540 int max_code; /* largest code with non zero frequency */
1541} tree_desc;
1542
1543static tree_desc l_desc =
1544 { dyn_ltree, static_ltree, extra_lbits, LITERALS + 1, L_CODES,
1545 MAX_BITS, 0 };
1546
1547static tree_desc d_desc =
1548 { dyn_dtree, static_dtree, extra_dbits, 0, D_CODES, MAX_BITS, 0 };
1549
1550static tree_desc bl_desc =
1551 { bl_tree, (ct_data *) 0, extra_blbits, 0, BL_CODES, MAX_BL_BITS,
1552 0 };
1553
1554
1555static ush bl_count[MAX_BITS + 1];
1556
1557/* number of codes at each bit length for an optimal tree */
1558
1559static const uch bl_order[BL_CODES]
1560= { 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15 };
1561
1562/* The lengths of the bit length codes are sent in order of decreasing
1563 * probability, to avoid transmitting the lengths for unused bit length codes.
1564 */
1565
1566static int heap[2 * L_CODES + 1]; /* heap used to build the Huffman trees */
1567static int heap_len; /* number of elements in the heap */
1568static int heap_max; /* element of largest frequency */
1569
1570/* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used.
1571 * The same heap array is used to build all trees.
1572 */
1573
1574static uch depth[2 * L_CODES + 1];
1575
1576/* Depth of each subtree used as tie breaker for trees of equal frequency */
1577
1578static uch length_code[MAX_MATCH - MIN_MATCH + 1];
1579
1580/* length code for each normalized match length (0 == MIN_MATCH) */
1581
1582static uch dist_code[512];
1583
1584/* distance codes. The first 256 values correspond to the distances
1585 * 3 .. 258, the last 256 values correspond to the top 8 bits of
1586 * the 15 bit distances.
1587 */
1588
1589static int base_length[LENGTH_CODES];
1590
1591/* First normalized length for each code (0 = MIN_MATCH) */
1592
1593static int base_dist[D_CODES];
1594
1595/* First normalized distance for each code (0 = distance of 1) */
1596
1597#define l_buf inbuf
1598/* DECLARE(uch, l_buf, LIT_BUFSIZE); buffer for literals or lengths */
1599
1600/* DECLARE(ush, d_buf, DIST_BUFSIZE); buffer for distances */
1601
1602static uch flag_buf[(LIT_BUFSIZE / 8)];
1603
1604/* flag_buf is a bit array distinguishing literals from lengths in
1605 * l_buf, thus indicating the presence or absence of a distance.
1606 */
1607
1608static unsigned last_lit; /* running index in l_buf */
1609static unsigned last_dist; /* running index in d_buf */
1610static unsigned last_flags; /* running index in flag_buf */
1611static uch flags; /* current flags not yet saved in flag_buf */
1612static uch flag_bit; /* current bit used in flags */
1613
1614/* bits are filled in flags starting at bit 0 (least significant).
1615 * Note: these flags are overkill in the current code since we don't
1616 * take advantage of DIST_BUFSIZE == LIT_BUFSIZE.
1617 */
1618
1619static ulg opt_len; /* bit length of current block with optimal trees */
1620static ulg static_len; /* bit length of current block with static trees */
1621
1622static ulg compressed_len; /* total bit length of compressed file */
1623
1624
1625static ush *file_type; /* pointer to UNKNOWN, BINARY or ASCII */
1626static int *file_method; /* pointer to DEFLATE or STORE */
1627
1628/* ===========================================================================
1629 * Local (static) routines in this file.
1630 */
1631
1632static void init_block (void);
1633static void pqdownheap (ct_data * tree, int k);
1634static void gen_bitlen (tree_desc * desc);
1635static void gen_codes (ct_data * tree, int max_code);
1636static void build_tree (tree_desc * desc);
1637static void scan_tree (ct_data * tree, int max_code);
1638static void send_tree (ct_data * tree, int max_code);
1639static int build_bl_tree (void);
1640static void send_all_trees (int lcodes, int dcodes, int blcodes);
1641static void compress_block (ct_data * ltree, ct_data * dtree);
1642static void set_file_type (void);
1643
1644
1645#ifndef DEBUG
1646# define send_code(c, tree) send_bits(tree[c].Code, tree[c].Len)
1647 /* Send a code of the given tree. c and tree must not have side effects */
1648
1649#else /* DEBUG */
1650# define send_code(c, tree) \
1651 { if (verbose>1) fprintf(stderr,"\ncd %3d ",(c)); \
1652 send_bits(tree[c].Code, tree[c].Len); }
1653#endif
1654
1655#define d_code(dist) \
1656 ((dist) < 256 ? dist_code[dist] : dist_code[256+((dist)>>7)])
1657/* Mapping from a distance to a distance code. dist is the distance - 1 and
1658 * must not have side effects. dist_code[256] and dist_code[257] are never
1659 * used.
1660 */
1661
1662/* the arguments must not have side effects */
1663
1664/* ===========================================================================
1665 * Allocate the match buffer, initialize the various tables and save the
1666 * location of the internal file attribute (ascii/binary) and method
1667 * (DEFLATE/STORE).
1668 */
1669static void ct_init(ush *attr, int *methodp)
1670{
1671 int n; /* iterates over tree elements */
1672 int bits; /* bit counter */
1673 int length; /* length value */
1674 int code; /* code value */
1675 int dist; /* distance index */
1676
1677 file_type = attr;
1678 file_method = methodp;
1679 compressed_len = 0L;
1680
1681 if (static_dtree[0].Len != 0)
1682 return; /* ct_init already called */
1683
1684 /* Initialize the mapping length (0..255) -> length code (0..28) */
1685 length = 0;
1686 for (code = 0; code < LENGTH_CODES - 1; code++) {
1687 base_length[code] = length;
1688 for (n = 0; n < (1 << extra_lbits[code]); n++) {
1689 length_code[length++] = (uch) code;
1690 }
1691 }
1692 Assert(length == 256, "ct_init: length != 256");
1693 /* Note that the length 255 (match length 258) can be represented
1694 * in two different ways: code 284 + 5 bits or code 285, so we
1695 * overwrite length_code[255] to use the best encoding:
1696 */
1697 length_code[length - 1] = (uch) code;
1698
1699 /* Initialize the mapping dist (0..32K) -> dist code (0..29) */
1700 dist = 0;
1701 for (code = 0; code < 16; code++) {
1702 base_dist[code] = dist;
1703 for (n = 0; n < (1 << extra_dbits[code]); n++) {
1704 dist_code[dist++] = (uch) code;
1705 }
1706 }
1707 Assert(dist == 256, "ct_init: dist != 256");
1708 dist >>= 7; /* from now on, all distances are divided by 128 */
1709 for (; code < D_CODES; code++) {
1710 base_dist[code] = dist << 7;
1711 for (n = 0; n < (1 << (extra_dbits[code] - 7)); n++) {
1712 dist_code[256 + dist++] = (uch) code;
1713 }
1714 }
1715 Assert(dist == 256, "ct_init: 256+dist != 512");
1716
1717 /* Construct the codes of the static literal tree */
1718 for (bits = 0; bits <= MAX_BITS; bits++)
1719 bl_count[bits] = 0;
1720 n = 0;
1721 while (n <= 143)
1722 static_ltree[n++].Len = 8, bl_count[8]++;
1723 while (n <= 255)
1724 static_ltree[n++].Len = 9, bl_count[9]++;
1725 while (n <= 279)
1726 static_ltree[n++].Len = 7, bl_count[7]++;
1727 while (n <= 287)
1728 static_ltree[n++].Len = 8, bl_count[8]++;
1729 /* Codes 286 and 287 do not exist, but we must include them in the
1730 * tree construction to get a canonical Huffman tree (longest code
1731 * all ones)
1732 */
1733 gen_codes((ct_data *) static_ltree, L_CODES + 1);
1734
1735 /* The static distance tree is trivial: */
1736 for (n = 0; n < D_CODES; n++) {
1737 static_dtree[n].Len = 5;
1738 static_dtree[n].Code = bi_reverse(n, 5);
1739 }
1740
1741 /* Initialize the first block of the first file: */
1742 init_block();
1743}
1744
1745/* ===========================================================================
1746 * Initialize a new block.
1747 */
1748static void init_block()
1749{
1750 int n; /* iterates over tree elements */
1751
1752 /* Initialize the trees. */
1753 for (n = 0; n < L_CODES; n++)
1754 dyn_ltree[n].Freq = 0;
1755 for (n = 0; n < D_CODES; n++)
1756 dyn_dtree[n].Freq = 0;
1757 for (n = 0; n < BL_CODES; n++)
1758 bl_tree[n].Freq = 0;
1759
1760 dyn_ltree[END_BLOCK].Freq = 1;
1761 opt_len = static_len = 0L;
1762 last_lit = last_dist = last_flags = 0;
1763 flags = 0;
1764 flag_bit = 1;
1765}
1766
1767#define SMALLEST 1
1768/* Index within the heap array of least frequent node in the Huffman tree */
1769
1770
1771/* ===========================================================================
1772 * Remove the smallest element from the heap and recreate the heap with
1773 * one less element. Updates heap and heap_len.
1774 */
1775#define pqremove(tree, top) \
1776{\
1777 top = heap[SMALLEST]; \
1778 heap[SMALLEST] = heap[heap_len--]; \
1779 pqdownheap(tree, SMALLEST); \
1780}
1781
1782/* ===========================================================================
1783 * Compares to subtrees, using the tree depth as tie breaker when
1784 * the subtrees have equal frequency. This minimizes the worst case length.
1785 */
1786#define smaller(tree, n, m) \
1787 (tree[n].Freq < tree[m].Freq || \
1788 (tree[n].Freq == tree[m].Freq && depth[n] <= depth[m]))
1789
1790/* ===========================================================================
1791 * Restore the heap property by moving down the tree starting at node k,
1792 * exchanging a node with the smallest of its two sons if necessary, stopping
1793 * when the heap property is re-established (each father smaller than its
1794 * two sons).
1795 */
1796static void pqdownheap(ct_data *tree, int k)
1797{
1798 int v = heap[k];
1799 int j = k << 1; /* left son of k */
1800
1801 while (j <= heap_len) {
1802 /* Set j to the smallest of the two sons: */
1803 if (j < heap_len && smaller(tree, heap[j + 1], heap[j]))
1804 j++;
1805
1806 /* Exit if v is smaller than both sons */
1807 if (smaller(tree, v, heap[j]))
1808 break;
1809
1810 /* Exchange v with the smallest son */
1811 heap[k] = heap[j];
1812 k = j;
1813
1814 /* And continue down the tree, setting j to the left son of k */
1815 j <<= 1;
1816 }
1817 heap[k] = v;
1818}
1819
1820/* ===========================================================================
1821 * Compute the optimal bit lengths for a tree and update the total bit length
1822 * for the current block.
1823 * IN assertion: the fields freq and dad are set, heap[heap_max] and
1824 * above are the tree nodes sorted by increasing frequency.
1825 * OUT assertions: the field len is set to the optimal bit length, the
1826 * array bl_count contains the frequencies for each bit length.
1827 * The length opt_len is updated; static_len is also updated if stree is
1828 * not null.
1829 */
1830static void gen_bitlen(tree_desc *desc)
1831{
1832 ct_data *tree = desc->dyn_tree;
1833 const extra_bits_t *extra = desc->extra_bits;
1834 int base = desc->extra_base;
1835 int max_code = desc->max_code;
1836 int max_length = desc->max_length;
1837 ct_data *stree = desc->static_tree;
1838 int h; /* heap index */
1839 int n, m; /* iterate over the tree elements */
1840 int bits; /* bit length */
1841 int xbits; /* extra bits */
1842 ush f; /* frequency */
1843 int overflow = 0; /* number of elements with bit length too large */
1844
1845 for (bits = 0; bits <= MAX_BITS; bits++)
1846 bl_count[bits] = 0;
1847
1848 /* In a first pass, compute the optimal bit lengths (which may
1849 * overflow in the case of the bit length tree).
1850 */
1851 tree[heap[heap_max]].Len = 0; /* root of the heap */
1852
1853 for (h = heap_max + 1; h < HEAP_SIZE; h++) {
1854 n = heap[h];
1855 bits = tree[tree[n].Dad].Len + 1;
1856 if (bits > max_length)
1857 bits = max_length, overflow++;
1858 tree[n].Len = (ush) bits;
1859 /* We overwrite tree[n].Dad which is no longer needed */
1860
1861 if (n > max_code)
1862 continue; /* not a leaf node */
1863
1864 bl_count[bits]++;
1865 xbits = 0;
1866 if (n >= base)
1867 xbits = extra[n - base];
1868 f = tree[n].Freq;
1869 opt_len += (ulg) f *(bits + xbits);
1870
1871 if (stree)
1872 static_len += (ulg) f *(stree[n].Len + xbits);
1873 }
1874 if (overflow == 0)
1875 return;
1876
1877 Trace((stderr, "\nbit length overflow\n"));
1878 /* This happens for example on obj2 and pic of the Calgary corpus */
1879
1880 /* Find the first bit length which could increase: */
1881 do {
1882 bits = max_length - 1;
1883 while (bl_count[bits] == 0)
1884 bits--;
1885 bl_count[bits]--; /* move one leaf down the tree */
1886 bl_count[bits + 1] += 2; /* move one overflow item as its brother */
1887 bl_count[max_length]--;
1888 /* The brother of the overflow item also moves one step up,
1889 * but this does not affect bl_count[max_length]
1890 */
1891 overflow -= 2;
1892 } while (overflow > 0);
1893
1894 /* Now recompute all bit lengths, scanning in increasing frequency.
1895 * h is still equal to HEAP_SIZE. (It is simpler to reconstruct all
1896 * lengths instead of fixing only the wrong ones. This idea is taken
1897 * from 'ar' written by Haruhiko Okumura.)
1898 */
1899 for (bits = max_length; bits != 0; bits--) {
1900 n = bl_count[bits];
1901 while (n != 0) {
1902 m = heap[--h];
1903 if (m > max_code)
1904 continue;
1905 if (tree[m].Len != (unsigned) bits) {
1906 Trace(
1907 (stderr, "code %d bits %d->%d\n", m, tree[m].Len,
1908 bits));
1909 opt_len +=
1910 ((long) bits -
1911 (long) tree[m].Len) * (long) tree[m].Freq;
1912 tree[m].Len = (ush) bits;
1913 }
1914 n--;
1915 }
1916 }
1917}
1918
1919/* ===========================================================================
1920 * Generate the codes for a given tree and bit counts (which need not be
1921 * optimal).
1922 * IN assertion: the array bl_count contains the bit length statistics for
1923 * the given tree and the field len is set for all tree elements.
1924 * OUT assertion: the field code is set for all tree elements of non
1925 * zero code length.
1926 */
1927static void gen_codes(ct_data *tree, int max_code)
1928{
1929 ush next_code[MAX_BITS + 1]; /* next code value for each bit length */
1930 ush code = 0; /* running code value */
1931 int bits; /* bit index */
1932 int n; /* code index */
1933
1934 /* The distribution counts are first used to generate the code values
1935 * without bit reversal.
1936 */
1937 for (bits = 1; bits <= MAX_BITS; bits++) {
1938 next_code[bits] = code = (code + bl_count[bits - 1]) << 1;
1939 }
1940 /* Check that the bit counts in bl_count are consistent. The last code
1941 * must be all ones.
1942 */
1943 Assert(code + bl_count[MAX_BITS] - 1 == (1 << MAX_BITS) - 1,
1944 "inconsistent bit counts");
1945 Tracev((stderr, "\ngen_codes: max_code %d ", max_code));
1946
1947 for (n = 0; n <= max_code; n++) {
1948 int len = tree[n].Len;
1949
1950 if (len == 0)
1951 continue;
1952 /* Now reverse the bits */
1953 tree[n].Code = bi_reverse(next_code[len]++, len);
1954
1955 Tracec(tree != static_ltree,
1956 (stderr, "\nn %3d %c l %2d c %4x (%x) ", n,
1957 (isgraph(n) ? n : ' '), len, tree[n].Code,
1958 next_code[len] - 1));
1959 }
1960}
1961
1962/* ===========================================================================
1963 * Construct one Huffman tree and assigns the code bit strings and lengths.
1964 * Update the total bit length for the current block.
1965 * IN assertion: the field freq is set for all tree elements.
1966 * OUT assertions: the fields len and code are set to the optimal bit length
1967 * and corresponding code. The length opt_len is updated; static_len is
1968 * also updated if stree is not null. The field max_code is set.
1969 */
1970static void build_tree(tree_desc *desc)
1971{
1972 ct_data *tree = desc->dyn_tree;
1973 ct_data *stree = desc->static_tree;
1974 int elems = desc->elems;
1975 int n, m; /* iterate over heap elements */
1976 int max_code = -1; /* largest code with non zero frequency */
1977 int node = elems; /* next internal node of the tree */
1978
1979 /* Construct the initial heap, with least frequent element in
1980 * heap[SMALLEST]. The sons of heap[n] are heap[2*n] and heap[2*n+1].
1981 * heap[0] is not used.
1982 */
1983 heap_len = 0, heap_max = HEAP_SIZE;
1984
1985 for (n = 0; n < elems; n++) {
1986 if (tree[n].Freq != 0) {
1987 heap[++heap_len] = max_code = n;
1988 depth[n] = 0;
1989 } else {
1990 tree[n].Len = 0;
1991 }
1992 }
1993
1994 /* The pkzip format requires that at least one distance code exists,
1995 * and that at least one bit should be sent even if there is only one
1996 * possible code. So to avoid special checks later on we force at least
1997 * two codes of non zero frequency.
1998 */
1999 while (heap_len < 2) {
2000 int new = heap[++heap_len] = (max_code < 2 ? ++max_code : 0);
2001
2002 tree[new].Freq = 1;
2003 depth[new] = 0;
2004 opt_len--;
2005 if (stree)
2006 static_len -= stree[new].Len;
2007 /* new is 0 or 1 so it does not have extra bits */
2008 }
2009 desc->max_code = max_code;
2010
2011 /* The elements heap[heap_len/2+1 .. heap_len] are leaves of the tree,
2012 * establish sub-heaps of increasing lengths:
2013 */
2014 for (n = heap_len / 2; n >= 1; n--)
2015 pqdownheap(tree, n);
2016
2017 /* Construct the Huffman tree by repeatedly combining the least two
2018 * frequent nodes.
2019 */
2020 do {
2021 pqremove(tree, n); /* n = node of least frequency */
2022 m = heap[SMALLEST]; /* m = node of next least frequency */
2023
2024 heap[--heap_max] = n; /* keep the nodes sorted by frequency */
2025 heap[--heap_max] = m;
2026
2027 /* Create a new node father of n and m */
2028 tree[node].Freq = tree[n].Freq + tree[m].Freq;
2029 depth[node] = (uch) (MAX(depth[n], depth[m]) + 1);
2030 tree[n].Dad = tree[m].Dad = (ush) node;
2031#ifdef DUMP_BL_TREE
2032 if (tree == bl_tree) {
2033 fprintf(stderr, "\nnode %d(%d), sons %d(%d) %d(%d)",
2034 node, tree[node].Freq, n, tree[n].Freq, m,
2035 tree[m].Freq);
2036 }
2037#endif
2038 /* and insert the new node in the heap */
2039 heap[SMALLEST] = node++;
2040 pqdownheap(tree, SMALLEST);
2041
2042 } while (heap_len >= 2);
2043
2044 heap[--heap_max] = heap[SMALLEST];
2045
2046 /* At this point, the fields freq and dad are set. We can now
2047 * generate the bit lengths.
2048 */
2049 gen_bitlen((tree_desc *) desc);
2050
2051 /* The field len is now set, we can generate the bit codes */
2052 gen_codes((ct_data *) tree, max_code);
2053}
2054
2055/* ===========================================================================
2056 * Scan a literal or distance tree to determine the frequencies of the codes
2057 * in the bit length tree. Updates opt_len to take into account the repeat
2058 * counts. (The contribution of the bit length codes will be added later
2059 * during the construction of bl_tree.)
2060 */
2061static void scan_tree(ct_data *tree, int max_code)
2062{
2063 int n; /* iterates over all tree elements */
2064 int prevlen = -1; /* last emitted length */
2065 int curlen; /* length of current code */
2066 int nextlen = tree[0].Len; /* length of next code */
2067 int count = 0; /* repeat count of the current code */
2068 int max_count = 7; /* max repeat count */
2069 int min_count = 4; /* min repeat count */
2070
2071 if (nextlen == 0)
2072 max_count = 138, min_count = 3;
2073 tree[max_code + 1].Len = (ush) 0xffff; /* guard */
2074
2075 for (n = 0; n <= max_code; n++) {
2076 curlen = nextlen;
2077 nextlen = tree[n + 1].Len;
2078 if (++count < max_count && curlen == nextlen) {
2079 continue;
2080 } else if (count < min_count) {
2081 bl_tree[curlen].Freq += count;
2082 } else if (curlen != 0) {
2083 if (curlen != prevlen)
2084 bl_tree[curlen].Freq++;
2085 bl_tree[REP_3_6].Freq++;
2086 } else if (count <= 10) {
2087 bl_tree[REPZ_3_10].Freq++;
2088 } else {
2089 bl_tree[REPZ_11_138].Freq++;
2090 }
2091 count = 0;
2092 prevlen = curlen;
2093 if (nextlen == 0) {
2094 max_count = 138, min_count = 3;
2095 } else if (curlen == nextlen) {
2096 max_count = 6, min_count = 3;
2097 } else {
2098 max_count = 7, min_count = 4;
2099 }
2100 }
2101}
2102
2103/* ===========================================================================
2104 * Send a literal or distance tree in compressed form, using the codes in
2105 * bl_tree.
2106 */
2107static void send_tree(ct_data *tree, int max_code)
2108{
2109 int n; /* iterates over all tree elements */
2110 int prevlen = -1; /* last emitted length */
2111 int curlen; /* length of current code */
2112 int nextlen = tree[0].Len; /* length of next code */
2113 int count = 0; /* repeat count of the current code */
2114 int max_count = 7; /* max repeat count */
2115 int min_count = 4; /* min repeat count */
2116
2117/* tree[max_code+1].Len = -1; *//* guard already set */
2118 if (nextlen == 0)
2119 max_count = 138, min_count = 3;
2120
2121 for (n = 0; n <= max_code; n++) {
2122 curlen = nextlen;
2123 nextlen = tree[n + 1].Len;
2124 if (++count < max_count && curlen == nextlen) {
2125 continue;
2126 } else if (count < min_count) {
2127 do {
2128 send_code(curlen, bl_tree);
2129 } while (--count != 0);
2130
2131 } else if (curlen != 0) {
2132 if (curlen != prevlen) {
2133 send_code(curlen, bl_tree);
2134 count--;
2135 }
2136 Assert(count >= 3 && count <= 6, " 3_6?");
2137 send_code(REP_3_6, bl_tree);
2138 send_bits(count - 3, 2);
2139
2140 } else if (count <= 10) {
2141 send_code(REPZ_3_10, bl_tree);
2142 send_bits(count - 3, 3);
2143
2144 } else {
2145 send_code(REPZ_11_138, bl_tree);
2146 send_bits(count - 11, 7);
2147 }
2148 count = 0;
2149 prevlen = curlen;
2150 if (nextlen == 0) {
2151 max_count = 138, min_count = 3;
2152 } else if (curlen == nextlen) {
2153 max_count = 6, min_count = 3;
2154 } else {
2155 max_count = 7, min_count = 4;
2156 }
2157 }
2158}
2159
2160/* ===========================================================================
2161 * Construct the Huffman tree for the bit lengths and return the index in
2162 * bl_order of the last bit length code to send.
2163 */
2164static const int build_bl_tree()
2165{
2166 int max_blindex; /* index of last bit length code of non zero freq */
2167
2168 /* Determine the bit length frequencies for literal and distance trees */
2169 scan_tree((ct_data *) dyn_ltree, l_desc.max_code);
2170 scan_tree((ct_data *) dyn_dtree, d_desc.max_code);
2171
2172 /* Build the bit length tree: */
2173 build_tree((tree_desc *) (&bl_desc));
2174 /* opt_len now includes the length of the tree representations, except
2175 * the lengths of the bit lengths codes and the 5+5+4 bits for the counts.
2176 */
2177
2178 /* Determine the number of bit length codes to send. The pkzip format
2179 * requires that at least 4 bit length codes be sent. (appnote.txt says
2180 * 3 but the actual value used is 4.)
2181 */
2182 for (max_blindex = BL_CODES - 1; max_blindex >= 3; max_blindex--) {
2183 if (bl_tree[bl_order[max_blindex]].Len != 0)
2184 break;
2185 }
2186 /* Update opt_len to include the bit length tree and counts */
2187 opt_len += 3 * (max_blindex + 1) + 5 + 5 + 4;
2188 Tracev(
2189 (stderr, "\ndyn trees: dyn %ld, stat %ld", opt_len,
2190 static_len));
2191
2192 return max_blindex;
2193}
2194
2195/* ===========================================================================
2196 * Send the header for a block using dynamic Huffman trees: the counts, the
2197 * lengths of the bit length codes, the literal tree and the distance tree.
2198 * IN assertion: lcodes >= 257, dcodes >= 1, blcodes >= 4.
2199 */
2200static void send_all_trees(int lcodes, int dcodes, int blcodes)
2201{
2202 int rank; /* index in bl_order */
2203
2204 Assert(lcodes >= 257 && dcodes >= 1
2205 && blcodes >= 4, "not enough codes");
2206 Assert(lcodes <= L_CODES && dcodes <= D_CODES
2207 && blcodes <= BL_CODES, "too many codes");
2208 Tracev((stderr, "\nbl counts: "));
2209 send_bits(lcodes - 257, 5); /* not +255 as stated in appnote.txt */
2210 send_bits(dcodes - 1, 5);
2211 send_bits(blcodes - 4, 4); /* not -3 as stated in appnote.txt */
2212 for (rank = 0; rank < blcodes; rank++) {
2213 Tracev((stderr, "\nbl code %2d ", bl_order[rank]));
2214 send_bits(bl_tree[bl_order[rank]].Len, 3);
2215 }
2216 Tracev((stderr, "\nbl tree: sent %ld", bits_sent));
2217
2218 send_tree((ct_data *) dyn_ltree, lcodes - 1); /* send the literal tree */
2219 Tracev((stderr, "\nlit tree: sent %ld", bits_sent));
2220
2221 send_tree((ct_data *) dyn_dtree, dcodes - 1); /* send the distance tree */
2222 Tracev((stderr, "\ndist tree: sent %ld", bits_sent));
2223}
2224
2225/* ===========================================================================
2226 * Determine the best encoding for the current block: dynamic trees, static
2227 * trees or store, and output the encoded block to the zip file. This function
2228 * returns the total compressed length for the file so far.
2229 */
2230static ulg flush_block(char *buf, ulg stored_len, int eof)
2231{
2232 ulg opt_lenb, static_lenb; /* opt_len and static_len in bytes */
2233 int max_blindex; /* index of last bit length code of non zero freq */
2234
2235 flag_buf[last_flags] = flags; /* Save the flags for the last 8 items */
2236
2237 /* Check if the file is ascii or binary */
2238 if (*file_type == (ush) UNKNOWN)
2239 set_file_type();
2240
2241 /* Construct the literal and distance trees */
2242 build_tree((tree_desc *) (&l_desc));
2243 Tracev((stderr, "\nlit data: dyn %ld, stat %ld", opt_len, static_len));
2244
2245 build_tree((tree_desc *) (&d_desc));
2246 Tracev(
2247 (stderr, "\ndist data: dyn %ld, stat %ld", opt_len,
2248 static_len));
2249 /* At this point, opt_len and static_len are the total bit lengths of
2250 * the compressed block data, excluding the tree representations.
2251 */
2252
2253 /* Build the bit length tree for the above two trees, and get the index
2254 * in bl_order of the last bit length code to send.
2255 */
2256 max_blindex = build_bl_tree();
2257
2258 /* Determine the best encoding. Compute first the block length in bytes */
2259 opt_lenb = (opt_len + 3 + 7) >> 3;
2260 static_lenb = (static_len + 3 + 7) >> 3;
2261
2262 Trace(
2263 (stderr,
2264 "\nopt %lu(%lu) stat %lu(%lu) stored %lu lit %u dist %u ",
2265 opt_lenb, opt_len, static_lenb, static_len, stored_len,
2266 last_lit, last_dist));
2267
2268 if (static_lenb <= opt_lenb)
2269 opt_lenb = static_lenb;
2270
2271 /* If compression failed and this is the first and last block,
2272 * and if the zip file can be seeked (to rewrite the local header),
2273 * the whole file is transformed into a stored file:
2274 */
2275 if (stored_len <= opt_lenb && eof && compressed_len == 0L
2276 && seekable()) {
2277 /* Since LIT_BUFSIZE <= 2*WSIZE, the input data must be there: */
2278 if (buf == (char *) 0)
2279 error_msg("block vanished");
2280
2281 copy_block(buf, (unsigned) stored_len, 0); /* without header */
2282 compressed_len = stored_len << 3;
2283 *file_method = STORED;
2284
2285 } else if (stored_len + 4 <= opt_lenb && buf != (char *) 0) {
2286 /* 4: two words for the lengths */
2287 /* The test buf != NULL is only necessary if LIT_BUFSIZE > WSIZE.
2288 * Otherwise we can't have processed more than WSIZE input bytes since
2289 * the last block flush, because compression would have been
2290 * successful. If LIT_BUFSIZE <= WSIZE, it is never too late to
2291 * transform a block into a stored block.
2292 */
2293 send_bits((STORED_BLOCK << 1) + eof, 3); /* send block type */
2294 compressed_len = (compressed_len + 3 + 7) & ~7L;
2295 compressed_len += (stored_len + 4) << 3;
2296
2297 copy_block(buf, (unsigned) stored_len, 1); /* with header */
2298
2299 } else if (static_lenb == opt_lenb) {
2300 send_bits((STATIC_TREES << 1) + eof, 3);
2301 compress_block((ct_data *) static_ltree,
2302 (ct_data *) static_dtree);
2303 compressed_len += 3 + static_len;
2304 } else {
2305 send_bits((DYN_TREES << 1) + eof, 3);
2306 send_all_trees(l_desc.max_code + 1, d_desc.max_code + 1,
2307 max_blindex + 1);
2308 compress_block((ct_data *) dyn_ltree,
2309 (ct_data *) dyn_dtree);
2310 compressed_len += 3 + opt_len;
2311 }
2312 Assert(compressed_len == bits_sent, "bad compressed size");
2313 init_block();
2314
2315 if (eof) {
2316 bi_windup();
2317 compressed_len += 7; /* align on byte boundary */
2318 }
2319 Tracev((stderr, "\ncomprlen %lu(%lu) ", compressed_len >> 3,
2320 compressed_len - 7 * eof));
2321
2322 return compressed_len >> 3;
2323}
2324
2325/* ===========================================================================
2326 * Save the match info and tally the frequency counts. Return true if
2327 * the current block must be flushed.
2328 */
2329static int ct_tally(int dist, int lc)
2330{
2331 l_buf[last_lit++] = (uch) lc;
2332 if (dist == 0) {
2333 /* lc is the unmatched char */
2334 dyn_ltree[lc].Freq++;
2335 } else {
2336 /* Here, lc is the match length - MIN_MATCH */
2337 dist--; /* dist = match distance - 1 */
2338 Assert((ush) dist < (ush) MAX_DIST &&
2339 (ush) lc <= (ush) (MAX_MATCH - MIN_MATCH) &&
2340 (ush) d_code(dist) < (ush) D_CODES, "ct_tally: bad match");
2341
2342 dyn_ltree[length_code[lc] + LITERALS + 1].Freq++;
2343 dyn_dtree[d_code(dist)].Freq++;
2344
2345 d_buf[last_dist++] = (ush) dist;
2346 flags |= flag_bit;
2347 }
2348 flag_bit <<= 1;
2349
2350 /* Output the flags if they fill a byte: */
2351 if ((last_lit & 7) == 0) {
2352 flag_buf[last_flags++] = flags;
2353 flags = 0, flag_bit = 1;
2354 }
2355 /* Try to guess if it is profitable to stop the current block here */
2356 if ((last_lit & 0xfff) == 0) {
2357 /* Compute an upper bound for the compressed length */
2358 ulg out_length = (ulg) last_lit * 8L;
2359 ulg in_length = (ulg) strstart - block_start;
2360 int dcode;
2361
2362 for (dcode = 0; dcode < D_CODES; dcode++) {
2363 out_length +=
2364 (ulg) dyn_dtree[dcode].Freq * (5L + extra_dbits[dcode]);
2365 }
2366 out_length >>= 3;
2367 Trace(
2368 (stderr,
2369 "\nlast_lit %u, last_dist %u, in %ld, out ~%ld(%ld%%) ",
2370 last_lit, last_dist, in_length, out_length,
2371 100L - out_length * 100L / in_length));
2372 if (last_dist < last_lit / 2 && out_length < in_length / 2)
2373 return 1;
2374 }
2375 return (last_lit == LIT_BUFSIZE - 1 || last_dist == DIST_BUFSIZE);
2376 /* We avoid equality with LIT_BUFSIZE because of wraparound at 64K
2377 * on 16 bit machines and because stored blocks are restricted to
2378 * 64K-1 bytes.
2379 */
2380}
2381
2382/* ===========================================================================
2383 * Send the block data compressed using the given Huffman trees
2384 */
2385static void compress_block(ct_data *ltree, ct_data *dtree)
2386{
2387 unsigned dist; /* distance of matched string */
2388 int lc; /* match length or unmatched char (if dist == 0) */
2389 unsigned lx = 0; /* running index in l_buf */
2390 unsigned dx = 0; /* running index in d_buf */
2391 unsigned fx = 0; /* running index in flag_buf */
2392 uch flag = 0; /* current flags */
2393 unsigned code; /* the code to send */
2394 int extra; /* number of extra bits to send */
2395
2396 if (last_lit != 0)
2397 do {
2398 if ((lx & 7) == 0)
2399 flag = flag_buf[fx++];
2400 lc = l_buf[lx++];
2401 if ((flag & 1) == 0) {
2402 send_code(lc, ltree); /* send a literal byte */
2403 Tracecv(isgraph(lc), (stderr, " '%c' ", lc));
2404 } else {
2405 /* Here, lc is the match length - MIN_MATCH */
2406 code = length_code[lc];
2407 send_code(code + LITERALS + 1, ltree); /* send the length code */
2408 extra = extra_lbits[code];
2409 if (extra != 0) {
2410 lc -= base_length[code];
2411 send_bits(lc, extra); /* send the extra length bits */
2412 }
2413 dist = d_buf[dx++];
2414 /* Here, dist is the match distance - 1 */
2415 code = d_code(dist);
2416 Assert(code < D_CODES, "bad d_code");
2417
2418 send_code(code, dtree); /* send the distance code */
2419 extra = extra_dbits[code];
2420 if (extra != 0) {
2421 dist -= base_dist[code];
2422 send_bits(dist, extra); /* send the extra distance bits */
2423 }
2424 } /* literal or match pair ? */
2425 flag >>= 1;
2426 } while (lx < last_lit);
2427
2428 send_code(END_BLOCK, ltree);
2429}
2430
2431/* ===========================================================================
2432 * Set the file type to ASCII or BINARY, using a crude approximation:
2433 * binary if more than 20% of the bytes are <= 6 or >= 128, ascii otherwise.
2434 * IN assertion: the fields freq of dyn_ltree are set and the total of all
2435 * frequencies does not exceed 64K (to fit in an int on 16 bit machines).
2436 */
2437static void set_file_type()
2438{
2439 int n = 0;
2440 unsigned ascii_freq = 0;
2441 unsigned bin_freq = 0;
2442
2443 while (n < 7)
2444 bin_freq += dyn_ltree[n++].Freq;
2445 while (n < 128)
2446 ascii_freq += dyn_ltree[n++].Freq;
2447 while (n < LITERALS)
2448 bin_freq += dyn_ltree[n++].Freq;
2449 *file_type = bin_freq > (ascii_freq >> 2) ? BINARY : ASCII;
2450 if (*file_type == BINARY && translate_eol) {
2451 error_msg("-l used on binary file");
2452 }
2453}
2454
2455/* zip.c -- compress files to the gzip or pkzip format
2456 * Copyright (C) 1992-1993 Jean-loup Gailly
2457 * This is free software; you can redistribute it and/or modify it under the
2458 * terms of the GNU General Public License, see the file COPYING.
2459 */
2460
2461
2462static ulg crc; /* crc on uncompressed file data */
2463static long header_bytes; /* number of bytes in gzip header */
2464
2465static void put_short_when_full(ush w)
2466{
2467 put_byte((uch)((w) & 0xff));
2468 put_byte((uch)((ush)(w) >> 8));
2469}
2470
2471static void put_short_function(ush n)
2472{
2473 put_short(n);
2474}
2475
2476static void put_long(ulg n)
2477{
2478 put_short_function((n) & 0xffff);
2479 put_short_function(((ulg)(n)) >> 16);
2480}
2481
2482/* put_header_byte is used for the compressed output
2483 * - for the initial 4 bytes that can't overflow the buffer.
2484 */
2485#define put_header_byte(c) {outbuf[outcnt++]=(uch)(c);}
2486
2487/* ===========================================================================
2488 * Deflate in to out.
2489 * IN assertions: the input and output buffers are cleared.
2490 * The variables time_stamp and save_orig_name are initialized.
2491 */
2492static int zip(int in, int out)
2493{
2494 uch my_flags = 0; /* general purpose bit flags */
2495 ush attr = 0; /* ascii/binary flag */
2496 ush deflate_flags = 0; /* pkzip -es, -en or -ex equivalent */
2497
2498 ifd = in;
2499 ofd = out;
2500 outcnt = 0;
2501
2502 /* Write the header to the gzip file. See algorithm.doc for the format */
2503
2504
2505 method = DEFLATED;
2506 put_header_byte(GZIP_MAGIC[0]); /* magic header */
2507 put_header_byte(GZIP_MAGIC[1]);
2508 put_header_byte(DEFLATED); /* compression method */
2509
2510 put_header_byte(my_flags); /* general flags */
2511 put_long(time_stamp);
2512
2513 /* Write deflated file to zip file */
2514 crc = updcrc(0, 0);
2515
2516 bi_init(out);
2517 ct_init(&attr, &method);
2518 lm_init(&deflate_flags);
2519
2520 put_byte((uch) deflate_flags); /* extra flags */
2521 put_byte(OS_CODE); /* OS identifier */
2522
2523 header_bytes = (long) outcnt;
2524
2525 (void) deflate();
2526
2527 /* Write the crc and uncompressed size */
2528 put_long(crc);
2529 put_long(isize);
2530 header_bytes += 2 * sizeof(long);
2531
2532 flush_outbuf();
2533 return OK;
2534}
2535
2536
2537/* ===========================================================================
2538 * Read a new buffer from the current input file, perform end-of-line
2539 * translation, and update the crc and input file size.
2540 * IN assertion: size >= 2 (for end-of-line translation)
2541 */
2542static int file_read(char *buf, unsigned size)
2543{
2544 unsigned len;
2545
2546 Assert(insize == 0, "inbuf not empty");
2547
2548 len = read(ifd, buf, size);
2549 if (len == (unsigned) (-1) || len == 0)
2550 return (int) len;
2551
2552 crc = updcrc((uch *) buf, len);
2553 isize += (ulg) len;
2554 return (int) len;
2555}
2556
2557/* ===========================================================================
2558 * Write the output buffer outbuf[0..outcnt-1] and update bytes_out.
2559 * (used for the compressed data only)
2560 */
2561static void flush_outbuf()
2562{
2563 if (outcnt == 0)
2564 return;
2565
2566 write_buf(ofd, (char *) outbuf, outcnt);
2567 outcnt = 0;
2568}
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2025-03-24 00:57:36 +0000