summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat (limited to '')
-rw-r--r--urunlevel/archival/gzip.c2558
1 files changed, 0 insertions, 2558 deletions
diff --git a/urunlevel/archival/gzip.c b/urunlevel/archival/gzip.c
deleted file mode 100644
index 1092480..0000000
--- a/urunlevel/archival/gzip.c
+++ /dev/null
@@ -1,2558 +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@codepoet.org> to support
13 * files as well as stdin/stdout, and to generally behave itself wrt
14 * command line handling.
15 *
16 * Adjusted further by David Seikel <won_fang@yahoo.cam.au> to support
17 * just storing the file.
18 *
19 * This program is free software; you can redistribute it and/or modify
20 * it under the terms of the GNU General Public License as published by
21 * the Free Software Foundation; either version 2 of the License, or
22 * (at your option) any later version.
23 *
24 * This program is distributed in the hope that it will be useful,
25 * but WITHOUT ANY WARRANTY; without even the implied warranty of
26 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
27 * General Public License for more details.
28 *
29 * You should have received a copy of the GNU General Public License
30 * along with this program; if not, write to the Free Software
31 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
32 *
33 */
34
35/* These defines are very important for BusyBox. Without these,
36 * huge chunks of ram are pre-allocated making the BusyBox bss
37 * size Freaking Huge(tm), which is a bad thing.*/
38#define SMALL_MEM
39#define DYN_ALLOC
40
41#include <stdlib.h>
42#include <stdio.h>
43#include <string.h>
44#include <unistd.h>
45#include <errno.h>
46#include <sys/types.h>
47#include <signal.h>
48#include <utime.h>
49#include <ctype.h>
50#include <sys/types.h>
51#include <unistd.h>
52#include <dirent.h>
53#include <fcntl.h>
54#include <time.h>
55#include "busybox.h"
56
57#define memzero(s, n) memset ((void *)(s), 0, (n))
58
59#ifndef RETSIGTYPE
60# define RETSIGTYPE void
61#endif
62
63typedef unsigned char uch;
64typedef unsigned short ush;
65typedef unsigned long ulg;
66
67/* Return codes from gzip */
68#define OK 0
69#define ERROR 1
70#define WARNING 2
71
72/* Compression methods (see algorithm.doc) */
73/* Only STORED and DEFLATED are supported by this BusyBox module */
74#define STORED 0
75/* methods 4 to 7 reserved */
76#define DEFLATED 8
77
78int just_store = 0;
79
80/* To save memory for 16 bit systems, some arrays are overlaid between
81 * the various modules:
82 * deflate: prev+head window d_buf l_buf outbuf
83 * unlzw: tab_prefix tab_suffix stack inbuf outbuf
84 * For compression, input is done in window[]. For decompression, output
85 * is done in window except for unlzw.
86 */
87
88#ifndef INBUFSIZ
89# ifdef SMALL_MEM
90# define INBUFSIZ 0x2000 /* input buffer size */
91# else
92# define INBUFSIZ 0x8000 /* input buffer size */
93# endif
94#endif
95#define INBUF_EXTRA 64 /* required by unlzw() */
96
97#ifndef OUTBUFSIZ
98# ifdef SMALL_MEM
99# define OUTBUFSIZ 8192 /* output buffer size */
100# else
101# define OUTBUFSIZ 16384 /* output buffer size */
102# endif
103#endif
104#define OUTBUF_EXTRA 2048 /* required by unlzw() */
105
106#ifndef DIST_BUFSIZE
107# ifdef SMALL_MEM
108# define DIST_BUFSIZE 0x2000 /* buffer for distances, see trees.c */
109# else
110# define DIST_BUFSIZE 0x8000 /* buffer for distances, see trees.c */
111# endif
112#endif
113
114#ifdef DYN_ALLOC
115# define DECLARE(type, array, size) static type * array
116# define ALLOC(type, array, size) { \
117 array = (type*)xcalloc((size_t)(((size)+1L)/2), 2*sizeof(type)); \
118 }
119# define FREE(array) {free(array), array=NULL;}
120#else
121# define DECLARE(type, array, size) static type array[size]
122# define ALLOC(type, array, size)
123# define FREE(array)
124#endif
125
126#define tab_suffix window
127#define tab_prefix prev /* hash link (see deflate.c) */
128#define head (prev+WSIZE) /* hash head (see deflate.c) */
129
130static long bytes_in; /* number of input bytes */
131
132#define isize bytes_in
133/* for compatibility with old zip sources (to be cleaned) */
134
135typedef int file_t; /* Do not use stdio */
136
137#define NO_FILE (-1) /* in memory compression */
138
139
140#define PACK_MAGIC "\037\036" /* Magic header for packed files */
141#define GZIP_MAGIC "\037\213" /* Magic header for gzip files, 1F 8B */
142#define OLD_GZIP_MAGIC "\037\236" /* Magic header for gzip 0.5 = freeze 1.x */
143#define LZH_MAGIC "\037\240" /* Magic header for SCO LZH Compress files */
144#define PKZIP_MAGIC "\120\113\003\004" /* Magic header for pkzip files */
145
146/* gzip flag byte */
147#define ASCII_FLAG 0x01 /* bit 0 set: file probably ascii text */
148#define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */
149#define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
150#define ORIG_NAME 0x08 /* bit 3 set: original file name present */
151#define COMMENT 0x10 /* bit 4 set: file comment present */
152#define RESERVED 0xC0 /* bit 6,7: reserved */
153
154/* internal file attribute */
155#define UNKNOWN 0xffff
156#define BINARY 0
157#define ASCII 1
158
159#ifndef WSIZE
160# define WSIZE 0x8000 /* window size--must be a power of two, and */
161#endif /* at least 32K for zip's deflate method */
162
163#define MIN_MATCH 3
164#define MAX_MATCH 258
165/* The minimum and maximum match lengths */
166
167#define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
168/* Minimum amount of lookahead, except at the end of the input file.
169 * See deflate.c for comments about the MIN_MATCH+1.
170 */
171
172#define MAX_DIST (WSIZE-MIN_LOOKAHEAD)
173/* In order to simplify the code, particularly on 16 bit machines, match
174 * distances are limited to MAX_DIST instead of WSIZE.
175 */
176
177/* put_byte is used for the compressed output */
178#define put_byte(c) {outbuf[outcnt++]=(uch)(c); if (outcnt==OUTBUFSIZ)\
179 flush_outbuf();}
180
181
182/* Output a 32 bit value to the bit stream, lsb first */
183#if 0
184#define put_long(n) { \
185 put_short((n) & 0xffff); \
186 put_short(((ulg)(n)) >> 16); \
187}
188#endif
189
190#define seekable() 0 /* force sequential output */
191#define translate_eol 0 /* no option -a yet */
192
193/* Diagnostic functions */
194#ifdef DEBUG
195# define Assert(cond,msg) {if(!(cond)) bb_error_msg(msg);}
196# define Trace(x) fprintf x
197# define Tracev(x) {if (verbose) fprintf x ;}
198# define Tracevv(x) {if (verbose>1) fprintf x ;}
199# define Tracec(c,x) {if (verbose && (c)) fprintf x ;}
200# define Tracecv(c,x) {if (verbose>1 && (c)) fprintf x ;}
201#else
202# define Assert(cond,msg)
203# define Trace(x)
204# define Tracev(x)
205# define Tracevv(x)
206# define Tracec(c,x)
207# define Tracecv(c,x)
208#endif
209
210#define WARN(msg) {if (!quiet) fprintf msg ; \
211 if (exit_code == OK) exit_code = WARNING;}
212
213#ifndef MAX_PATH_LEN
214# define MAX_PATH_LEN 1024 /* max pathname length */
215#endif
216
217
218 /* from zip.c: */
219static int zip(int in, int out);
220static int file_read(char *buf, unsigned size);
221
222 /* from gzip.c */
223static RETSIGTYPE abort_gzip(void);
224
225 /* from deflate.c */
226static void lm_init(ush * flags);
227static ulg deflate(void);
228
229 /* from trees.c */
230static void ct_init(ush * attr, int *methodp);
231static int ct_tally(int dist, int lc);
232static ulg flush_block(char *buf, ulg stored_len, int eof);
233
234 /* from bits.c */
235static void bi_init(file_t zipfile);
236static void send_bits(int value, int length);
237static unsigned bi_reverse(unsigned value, int length);
238static void bi_windup(void);
239static void copy_block(char *buf, unsigned len, int header);
240static int (*read_buf) (char *buf, unsigned size);
241
242 /* from util.c: */
243static void flush_outbuf(void);
244
245/* lzw.h -- define the lzw functions.
246 * Copyright (C) 1992-1993 Jean-loup Gailly.
247 * This is free software; you can redistribute it and/or modify it under the
248 * terms of the GNU General Public License, see the file COPYING.
249 */
250
251#if !defined(OF) && defined(lint)
252# include "gzip.h"
253#endif
254
255#ifndef BITS
256# define BITS 16
257#endif
258#define INIT_BITS 9 /* Initial number of bits per code */
259
260#define BIT_MASK 0x1f /* Mask for 'number of compression bits' */
261/* Mask 0x20 is reserved to mean a fourth header byte, and 0x40 is free.
262 * It's a pity that old uncompress does not check bit 0x20. That makes
263 * extension of the format actually undesirable because old compress
264 * would just crash on the new format instead of giving a meaningful
265 * error message. It does check the number of bits, but it's more
266 * helpful to say "unsupported format, get a new version" than
267 * "can only handle 16 bits".
268 */
269
270/* tailor.h -- target dependent definitions
271 * Copyright (C) 1992-1993 Jean-loup Gailly.
272 * This is free software; you can redistribute it and/or modify it under the
273 * terms of the GNU General Public License, see the file COPYING.
274 */
275
276/* The target dependent definitions should be defined here only.
277 * The target dependent functions should be defined in tailor.c.
278 */
279
280
281 /* Common defaults */
282
283#ifndef OS_CODE
284# define OS_CODE 0x03 /* assume Unix */
285#endif
286
287#ifndef PATH_SEP
288# define PATH_SEP '/'
289#endif
290
291#ifndef OPTIONS_VAR
292# define OPTIONS_VAR "GZIP"
293#endif
294
295#ifndef Z_SUFFIX
296# define Z_SUFFIX ".gz"
297#endif
298
299#ifdef MAX_EXT_CHARS
300# define MAX_SUFFIX MAX_EXT_CHARS
301#else
302# define MAX_SUFFIX 30
303#endif
304
305 /* global buffers */
306
307DECLARE(uch, inbuf, INBUFSIZ + INBUF_EXTRA);
308DECLARE(uch, outbuf, OUTBUFSIZ + OUTBUF_EXTRA);
309DECLARE(ush, d_buf, DIST_BUFSIZE);
310DECLARE(uch, window, 2L * WSIZE);
311DECLARE(ush, tab_prefix, 1L << BITS);
312
313static int foreground; /* set if program run in foreground */
314static int method = DEFLATED; /* compression method */
315static int exit_code = OK; /* program exit code */
316static int part_nb; /* number of parts in .gz file */
317static long time_stamp; /* original time stamp (modification time) */
318static long ifile_size; /* input file size, -1 for devices (debug only) */
319static char z_suffix[MAX_SUFFIX + 1]; /* default suffix (can be set with --suffix) */
320static int z_len; /* strlen(z_suffix) */
321
322static int ifd; /* input file descriptor */
323static int ofd; /* output file descriptor */
324static unsigned insize; /* valid bytes in inbuf */
325static unsigned outcnt; /* bytes in output buffer */
326
327
328/* Output a 16 bit value, lsb first */
329static void put_short(ush w)
330{
331 if (outcnt < OUTBUFSIZ - 2) {
332 outbuf[outcnt++] = (uch) ((w) & 0xff);
333 outbuf[outcnt++] = (uch) ((ush) (w) >> 8);
334 } else {
335 put_byte((uch) ((w) & 0xff));
336 put_byte((uch) ((ush) (w) >> 8));
337 }
338}
339
340/* ========================================================================
341 * Signal and error handler.
342 */
343static void abort_gzip()
344{
345 exit(ERROR);
346}
347
348/* ===========================================================================
349 * Clear input and output buffers
350 */
351static void clear_bufs(void)
352{
353 outcnt = 0;
354 insize = 0;
355 bytes_in = 0L;
356}
357
358static void write_bb_error_msg(void)
359{
360 fputc('\n', stderr);
361 bb_perror_nomsg();
362 abort_gzip();
363}
364
365/* ===========================================================================
366 * Does the same as write(), but also handles partial pipe writes and checks
367 * for error return.
368 */
369static void write_buf(int fd, void *buf, unsigned cnt)
370{
371 unsigned n;
372
373 while ((n = write(fd, buf, cnt)) != cnt) {
374 if (n == (unsigned) (-1)) {
375 write_bb_error_msg();
376 }
377 cnt -= n;
378 buf = (void *) ((char *) buf + n);
379 }
380}
381
382/* ===========================================================================
383 * Run a set of bytes through the crc shift register. If s is a NULL
384 * pointer, then initialize the crc shift register contents instead.
385 * Return the current crc in either case.
386 */
387static ulg updcrc(uch * s, unsigned n)
388{
389 static ulg crc = (ulg) 0xffffffffL; /* shift register contents */
390 register ulg c; /* temporary variable */
391 static unsigned long crc_32_tab[256];
392
393 if (crc_32_tab[1] == 0x00000000L) {
394 unsigned long csr; /* crc shift register */
395 const unsigned long e = 0xedb88320L; /* polynomial exclusive-or pattern */
396 int i; /* counter for all possible eight bit values */
397 int k; /* byte being shifted into crc apparatus */
398
399 /* Compute table of CRC's. */
400 for (i = 1; i < 256; i++) {
401 csr = i;
402 /* The idea to initialize the register with the byte instead of
403 * zero was stolen from Haruhiko Okumura's ar002
404 */
405 for (k = 8; k; k--)
406 csr = csr & 1 ? (csr >> 1) ^ e : csr >> 1;
407 crc_32_tab[i] = csr;
408 }
409 }
410
411 if (s == NULL) {
412 c = 0xffffffffL;
413 } else {
414 c = crc;
415 if (n)
416 do {
417 c = crc_32_tab[((int) c ^ (*s++)) & 0xff] ^ (c >> 8);
418 } while (--n);
419 }
420 crc = c;
421 return c ^ 0xffffffffL; /* (instead of ~c for 64-bit machines) */
422}
423
424/* bits.c -- output variable-length bit strings
425 * Copyright (C) 1992-1993 Jean-loup Gailly
426 * This is free software; you can redistribute it and/or modify it under the
427 * terms of the GNU General Public License, see the file COPYING.
428 */
429
430
431/*
432 * PURPOSE
433 *
434 * Output variable-length bit strings. Compression can be done
435 * to a file or to memory. (The latter is not supported in this version.)
436 *
437 * DISCUSSION
438 *
439 * The PKZIP "deflate" file format interprets compressed file data
440 * as a sequence of bits. Multi-bit strings in the file may cross
441 * byte boundaries without restriction.
442 *
443 * The first bit of each byte is the low-order bit.
444 *
445 * The routines in this file allow a variable-length bit value to
446 * be output right-to-left (useful for literal values). For
447 * left-to-right output (useful for code strings from the tree routines),
448 * the bits must have been reversed first with bi_reverse().
449 *
450 * For in-memory compression, the compressed bit stream goes directly
451 * into the requested output buffer. The input data is read in blocks
452 * by the mem_read() function. The buffer is limited to 64K on 16 bit
453 * machines.
454 *
455 * INTERFACE
456 *
457 * void bi_init (FILE *zipfile)
458 * Initialize the bit string routines.
459 *
460 * void send_bits (int value, int length)
461 * Write out a bit string, taking the source bits right to
462 * left.
463 *
464 * int bi_reverse (int value, int length)
465 * Reverse the bits of a bit string, taking the source bits left to
466 * right and emitting them right to left.
467 *
468 * void bi_windup (void)
469 * Write out any remaining bits in an incomplete byte.
470 *
471 * void copy_block(char *buf, unsigned len, int header)
472 * Copy a stored block to the zip file, storing first the length and
473 * its one's complement if requested.
474 *
475 */
476
477/* ===========================================================================
478 * Local data used by the "bit string" routines.
479 */
480
481static file_t zfile; /* output gzip file */
482
483static unsigned short bi_buf;
484
485/* Output buffer. bits are inserted starting at the bottom (least significant
486 * bits).
487 */
488
489#define Buf_size (8 * 2*sizeof(char))
490/* Number of bits used within bi_buf. (bi_buf might be implemented on
491 * more than 16 bits on some systems.)
492 */
493
494static int bi_valid;
495
496/* Current input function. Set to mem_read for in-memory compression */
497
498#ifdef DEBUG
499ulg bits_sent; /* bit length of the compressed data */
500#endif
501
502/* ===========================================================================
503 * Initialize the bit string routines.
504 */
505static void bi_init(file_t zipfile)
506{
507 zfile = zipfile;
508 bi_buf = 0;
509 bi_valid = 0;
510#ifdef DEBUG
511 bits_sent = 0L;
512#endif
513
514 /* Set the defaults for file compression. They are set by memcompress
515 * for in-memory compression.
516 */
517 if (zfile != NO_FILE) {
518 read_buf = file_read;
519 }
520}
521
522/* ===========================================================================
523 * Send a value on a given number of bits.
524 * IN assertion: length <= 16 and value fits in length bits.
525 */
526static void send_bits(int value, int length)
527{
528#ifdef DEBUG
529 Tracev((stderr, " l %2d v %4x ", length, value));
530 Assert(length > 0 && length <= 15, "invalid length");
531 bits_sent += (ulg) length;
532#endif
533 /* If not enough room in bi_buf, use (valid) bits from bi_buf and
534 * (16 - bi_valid) bits from value, leaving (width - (16-bi_valid))
535 * unused bits in value.
536 */
537 if (bi_valid > (int) Buf_size - length) {
538 bi_buf |= (value << bi_valid);
539 put_short(bi_buf);
540 bi_buf = (ush) value >> (Buf_size - bi_valid);
541 bi_valid += length - Buf_size;
542 } else {
543 bi_buf |= value << bi_valid;
544 bi_valid += length;
545 }
546}
547
548/* ===========================================================================
549 * Reverse the first len bits of a code, using straightforward code (a faster
550 * method would use a table)
551 * IN assertion: 1 <= len <= 15
552 */
553static unsigned bi_reverse(unsigned code, int len)
554{
555 register unsigned res = 0;
556
557 do {
558 res |= code & 1;
559 code >>= 1, res <<= 1;
560 } while (--len > 0);
561 return res >> 1;
562}
563
564/* ===========================================================================
565 * Write out any remaining bits in an incomplete byte.
566 */
567static void bi_windup()
568{
569 if (bi_valid > 8) {
570 put_short(bi_buf);
571 } else if (bi_valid > 0) {
572 put_byte(bi_buf);
573 }
574 bi_buf = 0;
575 bi_valid = 0;
576#ifdef DEBUG
577 bits_sent = (bits_sent + 7) & ~7;
578#endif
579}
580
581/* ===========================================================================
582 * Copy a stored block to the zip file, storing first the length and its
583 * one's complement if requested.
584 */
585static void copy_block(char *buf, unsigned len, int header)
586{
587 bi_windup(); /* align on byte boundary */
588
589 if (header) {
590 put_short((ush) len);
591 put_short((ush) ~ len);
592#ifdef DEBUG
593 bits_sent += 2 * 16;
594#endif
595 }
596#ifdef DEBUG
597 bits_sent += (ulg) len << 3;
598#endif
599 while (len--) {
600 put_byte(*buf++);
601 }
602}
603
604/* deflate.c -- compress data using the deflation algorithm
605 * Copyright (C) 1992-1993 Jean-loup Gailly
606 * This is free software; you can redistribute it and/or modify it under the
607 * terms of the GNU General Public License, see the file COPYING.
608 */
609
610/*
611 * PURPOSE
612 *
613 * Identify new text as repetitions of old text within a fixed-
614 * length sliding window trailing behind the new text.
615 *
616 * DISCUSSION
617 *
618 * The "deflation" process depends on being able to identify portions
619 * of the input text which are identical to earlier input (within a
620 * sliding window trailing behind the input currently being processed).
621 *
622 * The most straightforward technique turns out to be the fastest for
623 * most input files: try all possible matches and select the longest.
624 * The key feature of this algorithm is that insertions into the string
625 * dictionary are very simple and thus fast, and deletions are avoided
626 * completely. Insertions are performed at each input character, whereas
627 * string matches are performed only when the previous match ends. So it
628 * is preferable to spend more time in matches to allow very fast string
629 * insertions and avoid deletions. The matching algorithm for small
630 * strings is inspired from that of Rabin & Karp. A brute force approach
631 * is used to find longer strings when a small match has been found.
632 * A similar algorithm is used in comic (by Jan-Mark Wams) and freeze
633 * (by Leonid Broukhis).
634 * A previous version of this file used a more sophisticated algorithm
635 * (by Fiala and Greene) which is guaranteed to run in linear amortized
636 * time, but has a larger average cost, uses more memory and is patented.
637 * However the F&G algorithm may be faster for some highly redundant
638 * files if the parameter max_chain_length (described below) is too large.
639 *
640 * ACKNOWLEDGMENTS
641 *
642 * The idea of lazy evaluation of matches is due to Jan-Mark Wams, and
643 * I found it in 'freeze' written by Leonid Broukhis.
644 * Thanks to many info-zippers for bug reports and testing.
645 *
646 * REFERENCES
647 *
648 * APPNOTE.TXT documentation file in PKZIP 1.93a distribution.
649 *
650 * A description of the Rabin and Karp algorithm is given in the book
651 * "Algorithms" by R. Sedgewick, Addison-Wesley, p252.
652 *
653 * Fiala,E.R., and Greene,D.H.
654 * Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595
655 *
656 * INTERFACE
657 *
658 * void lm_init (int pack_level, ush *flags)
659 * Initialize the "longest match" routines for a new file
660 *
661 * ulg deflate (void)
662 * Processes a new input file and return its compressed length. Sets
663 * the compressed length, crc, deflate flags and internal file
664 * attributes.
665 */
666
667
668/* ===========================================================================
669 * Configuration parameters
670 */
671
672/* Compile with MEDIUM_MEM to reduce the memory requirements or
673 * with SMALL_MEM to use as little memory as possible. Use BIG_MEM if the
674 * entire input file can be held in memory (not possible on 16 bit systems).
675 * Warning: defining these symbols affects HASH_BITS (see below) and thus
676 * affects the compression ratio. The compressed output
677 * is still correct, and might even be smaller in some cases.
678 */
679
680#ifdef SMALL_MEM
681# define HASH_BITS 13 /* Number of bits used to hash strings */
682#endif
683#ifdef MEDIUM_MEM
684# define HASH_BITS 14
685#endif
686#ifndef HASH_BITS
687# define HASH_BITS 15
688 /* For portability to 16 bit machines, do not use values above 15. */
689#endif
690
691/* To save space (see unlzw.c), we overlay prev+head with tab_prefix and
692 * window with tab_suffix. Check that we can do this:
693 */
694#if (WSIZE<<1) > (1<<BITS)
695# error cannot overlay window with tab_suffix and prev with tab_prefix0
696#endif
697#if HASH_BITS > BITS-1
698# error cannot overlay head with tab_prefix1
699#endif
700#define HASH_SIZE (unsigned)(1<<HASH_BITS)
701#define HASH_MASK (HASH_SIZE-1)
702#define WMASK (WSIZE-1)
703/* HASH_SIZE and WSIZE must be powers of two */
704#define NIL 0
705/* Tail of hash chains */
706#define FAST 4
707#define SLOW 2
708/* speed options for the general purpose bit flag */
709#ifndef TOO_FAR
710# define TOO_FAR 4096
711#endif
712/* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
713/* ===========================================================================
714 * Local data used by the "longest match" routines.
715 */
716typedef ush Pos;
717typedef unsigned IPos;
718
719/* A Pos is an index in the character window. We use short instead of int to
720 * save space in the various tables. IPos is used only for parameter passing.
721 */
722
723/* DECLARE(uch, window, 2L*WSIZE); */
724/* Sliding window. Input bytes are read into the second half of the window,
725 * and move to the first half later to keep a dictionary of at least WSIZE
726 * bytes. With this organization, matches are limited to a distance of
727 * WSIZE-MAX_MATCH bytes, but this ensures that IO is always
728 * performed with a length multiple of the block size. Also, it limits
729 * the window size to 64K, which is quite useful on MSDOS.
730 * To do: limit the window size to WSIZE+BSZ if SMALL_MEM (the code would
731 * be less efficient).
732 */
733
734/* DECLARE(Pos, prev, WSIZE); */
735/* Link to older string with same hash index. To limit the size of this
736 * array to 64K, this link is maintained only for the last 32K strings.
737 * An index in this array is thus a window index modulo 32K.
738 */
739
740/* DECLARE(Pos, head, 1<<HASH_BITS); */
741/* Heads of the hash chains or NIL. */
742
743static const ulg window_size = (ulg) 2 * WSIZE;
744
745/* window size, 2*WSIZE except for MMAP or BIG_MEM, where it is the
746 * input file length plus MIN_LOOKAHEAD.
747 */
748
749static long block_start;
750
751/* window position at the beginning of the current output block. Gets
752 * negative when the window is moved backwards.
753 */
754
755static unsigned ins_h; /* hash index of string to be inserted */
756
757#define H_SHIFT ((HASH_BITS+MIN_MATCH-1)/MIN_MATCH)
758/* Number of bits by which ins_h and del_h must be shifted at each
759 * input step. It must be such that after MIN_MATCH steps, the oldest
760 * byte no longer takes part in the hash key, that is:
761 * H_SHIFT * MIN_MATCH >= HASH_BITS
762 */
763
764static unsigned int prev_length;
765
766/* Length of the best match at previous step. Matches not greater than this
767 * are discarded. This is used in the lazy match evaluation.
768 */
769
770static unsigned strstart; /* start of string to insert */
771static unsigned match_start; /* start of matching string */
772static int eofile; /* flag set at end of input file */
773static unsigned lookahead; /* number of valid bytes ahead in window */
774
775static const unsigned max_chain_length = 4096;
776
777/* To speed up deflation, hash chains are never searched beyond this length.
778 * A higher limit improves compression ratio but degrades the speed.
779 */
780
781static const unsigned int max_lazy_match = 258;
782
783/* Attempt to find a better match only when the current match is strictly
784 * smaller than this value. This mechanism is used only for compression
785 * levels >= 4.
786 */
787#define max_insert_length max_lazy_match
788/* Insert new strings in the hash table only if the match length
789 * is not greater than this length. This saves time but degrades compression.
790 * max_insert_length is used only for compression levels <= 3.
791 */
792
793static const unsigned good_match = 32;
794
795/* Use a faster search when the previous match is longer than this */
796
797
798/* Values for max_lazy_match, good_match and max_chain_length, depending on
799 * the desired pack level (0..9). The values given below have been tuned to
800 * exclude worst case performance for pathological files. Better values may be
801 * found for specific files.
802 */
803
804static const int nice_match = 258; /* Stop searching when current match exceeds this */
805
806/* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
807 * For deflate_fast() (levels <= 3) good is ignored and lazy has a different
808 * meaning.
809 */
810
811#define EQUAL 0
812/* result of memcmp for equal strings */
813
814/* ===========================================================================
815 * Prototypes for local functions.
816 */
817static void fill_window(void);
818
819static int longest_match(IPos cur_match);
820
821#ifdef DEBUG
822static void check_match(IPos start, IPos match, int length);
823#endif
824
825/* ===========================================================================
826 * Update a hash value with the given input byte
827 * IN assertion: all calls to to UPDATE_HASH are made with consecutive
828 * input characters, so that a running hash key can be computed from the
829 * previous key instead of complete recalculation each time.
830 */
831#define UPDATE_HASH(h,c) (h = (((h)<<H_SHIFT) ^ (c)) & HASH_MASK)
832
833/* ===========================================================================
834 * Insert string s in the dictionary and set match_head to the previous head
835 * of the hash chain (the most recent string with same hash key). Return
836 * the previous length of the hash chain.
837 * IN assertion: all calls to to INSERT_STRING are made with consecutive
838 * input characters and the first MIN_MATCH bytes of s are valid
839 * (except for the last MIN_MATCH-1 bytes of the input file).
840 */
841#define INSERT_STRING(s, match_head) \
842 (UPDATE_HASH(ins_h, window[(s) + MIN_MATCH-1]), \
843 prev[(s) & WMASK] = match_head = head[ins_h], \
844 head[ins_h] = (s))
845
846/* ===========================================================================
847 * Initialize the "longest match" routines for a new file
848 */
849static void lm_init(ush * flags)
850{
851 register unsigned j;
852
853 /* Initialize the hash table. */
854 memzero((char *) head, HASH_SIZE * sizeof(*head));
855 /* prev will be initialized on the fly */
856
857 *flags |= SLOW;
858 /* ??? reduce max_chain_length for binary files */
859
860 strstart = 0;
861 block_start = 0L;
862
863 lookahead = read_buf((char *) window,
864 sizeof(int) <= 2 ? (unsigned) WSIZE : 2 * WSIZE);
865
866 if (lookahead == 0 || lookahead == (unsigned) EOF) {
867 eofile = 1, lookahead = 0;
868 return;
869 }
870 eofile = 0;
871 /* Make sure that we always have enough lookahead. This is important
872 * if input comes from a device such as a tty.
873 */
874 while (lookahead < MIN_LOOKAHEAD && !eofile)
875 fill_window();
876
877 ins_h = 0;
878 for (j = 0; j < MIN_MATCH - 1; j++)
879 UPDATE_HASH(ins_h, window[j]);
880 /* If lookahead < MIN_MATCH, ins_h is garbage, but this is
881 * not important since only literal bytes will be emitted.
882 */
883}
884
885/* ===========================================================================
886 * Set match_start to the longest match starting at the given string and
887 * return its length. Matches shorter or equal to prev_length are discarded,
888 * in which case the result is equal to prev_length and match_start is
889 * garbage.
890 * IN assertions: cur_match is the head of the hash chain for the current
891 * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
892 */
893
894/* For MSDOS, OS/2 and 386 Unix, an optimized version is in match.asm or
895 * match.s. The code is functionally equivalent, so you can use the C version
896 * if desired.
897 */
898static int longest_match(IPos cur_match)
899{
900 unsigned chain_length = max_chain_length; /* max hash chain length */
901 register uch *scan = window + strstart; /* current string */
902 register uch *match; /* matched string */
903 register int len; /* length of current match */
904 int best_len = prev_length; /* best match length so far */
905 IPos limit =
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, "insufficient lookahead");
926
927 do {
928 Assert(cur_match < strstart, "no future");
929 match = window + cur_match;
930
931 /* Skip to next match if the match length cannot increase
932 * or if the match length is less than 2:
933 */
934 if (match[best_len] != scan_end ||
935 match[best_len - 1] != scan_end1 ||
936 *match != *scan || *++match != scan[1])
937 continue;
938
939 /* The check at best_len-1 can be removed because it will be made
940 * again later. (This heuristic is not always a win.)
941 * It is not necessary to compare scan[2] and match[2] since they
942 * are always equal when the other bytes match, given that
943 * the hash keys are equal and that HASH_BITS >= 8.
944 */
945 scan += 2, match++;
946
947 /* We check for insufficient lookahead only every 8th comparison;
948 * the 256th check will be made at strstart+258.
949 */
950 do {
951 } while (*++scan == *++match && *++scan == *++match &&
952 *++scan == *++match && *++scan == *++match &&
953 *++scan == *++match && *++scan == *++match &&
954 *++scan == *++match && *++scan == *++match && scan < strend);
955
956 len = MAX_MATCH - (int) (strend - scan);
957 scan = strend - MAX_MATCH;
958
959 if (len > best_len) {
960 match_start = cur_match;
961 best_len = len;
962 if (len >= nice_match)
963 break;
964 scan_end1 = scan[best_len - 1];
965 scan_end = scan[best_len];
966 }
967 } while ((cur_match = prev[cur_match & WMASK]) > limit
968 && --chain_length != 0);
969
970 return best_len;
971}
972
973#ifdef DEBUG
974/* ===========================================================================
975 * Check that the match at match_start is indeed a match.
976 */
977static void check_match(IPos start, IPos match, int length)
978{
979 /* check that the match is indeed a match */
980 if (memcmp((char *) window + match,
981 (char *) window + start, length) != EQUAL) {
982 bb_error_msg(" start %d, match %d, length %d", start, match, length);
983 bb_error_msg("invalid match");
984 }
985 if (verbose > 1) {
986 bb_error_msg("\\[%d,%d]", start - match, length);
987 do {
988 putc(window[start++], stderr);
989 } while (--length != 0);
990 }
991}
992#else
993# define check_match(start, match, length)
994#endif
995
996/* ===========================================================================
997 * Fill the window when the lookahead becomes insufficient.
998 * Updates strstart and lookahead, and sets eofile if end of input file.
999 * IN assertion: lookahead < MIN_LOOKAHEAD && strstart + lookahead > 0
1000 * OUT assertions: at least one byte has been read, or eofile is set;
1001 * file reads are performed for at least two bytes (required for the
1002 * translate_eol option).
1003 */
1004static void fill_window()
1005{
1006 register unsigned n, m;
1007 unsigned more =
1008 (unsigned) (window_size - (ulg) lookahead - (ulg) strstart);
1009 /* Amount of free space at the end of the window. */
1010
1011 /* If the window is almost full and there is insufficient lookahead,
1012 * move the upper half to the lower one to make room in the upper half.
1013 */
1014 if (more == (unsigned) EOF) {
1015 /* Very unlikely, but possible on 16 bit machine if strstart == 0
1016 * and lookahead == 1 (input done one byte at time)
1017 */
1018 more--;
1019 } else if (strstart >= WSIZE + MAX_DIST) {
1020 /* By the IN assertion, the window is not empty so we can't confuse
1021 * more == 0 with more == 64K on a 16 bit machine.
1022 */
1023 Assert(window_size == (ulg) 2 * WSIZE, "no sliding with BIG_MEM");
1024
1025 memcpy((char *) window, (char *) window + WSIZE, (unsigned) WSIZE);
1026 match_start -= WSIZE;
1027 strstart -= WSIZE; /* we now have strstart >= MAX_DIST: */
1028
1029 block_start -= (long) WSIZE;
1030
1031 for (n = 0; n < HASH_SIZE; n++) {
1032 m = head[n];
1033 head[n] = (Pos) (m >= WSIZE ? m - WSIZE : NIL);
1034 }
1035 for (n = 0; n < WSIZE; n++) {
1036 m = prev[n];
1037 prev[n] = (Pos) (m >= WSIZE ? m - WSIZE : NIL);
1038 /* If n is not on any hash chain, prev[n] is garbage but
1039 * its value will never be used.
1040 */
1041 }
1042 more += WSIZE;
1043 }
1044 /* At this point, more >= 2 */
1045 if (!eofile) {
1046 n = read_buf((char *) window + strstart + lookahead, more);
1047 if (n == 0 || n == (unsigned) EOF) {
1048 eofile = 1;
1049 } else {
1050 lookahead += n;
1051 }
1052 }
1053}
1054
1055/* ===========================================================================
1056 * Flush the current block, with given end-of-file flag.
1057 * IN assertion: strstart is set to the end of the current match.
1058 */
1059#define FLUSH_BLOCK(eof) \
1060 flush_block(block_start >= 0L ? (char*)&window[(unsigned)block_start] : \
1061 (char*)NULL, (long)strstart - block_start, (eof))
1062
1063/* ===========================================================================
1064 * Same as above, but achieves better compression. We use a lazy
1065 * evaluation for matches: a match is finally adopted only if there is
1066 * no better match at the next window position.
1067 */
1068static ulg deflate()
1069{
1070 IPos hash_head; /* head of hash chain */
1071 IPos prev_match; /* previous match */
1072 int flush; /* set if current block must be flushed */
1073 int match_available = 0; /* set if previous match exists */
1074 register unsigned match_length = MIN_MATCH - 1; /* length of best match */
1075
1076 /* Process the input block. */
1077 while (lookahead != 0) {
1078 /* Insert the string window[strstart .. strstart+2] in the
1079 * dictionary, and set hash_head to the head of the hash chain:
1080 */
1081 INSERT_STRING(strstart, hash_head);
1082
1083 /* Find the longest match, discarding those <= prev_length.
1084 */
1085 prev_length = match_length, prev_match = match_start;
1086 match_length = MIN_MATCH - 1;
1087
1088 if (hash_head != NIL && prev_length < max_lazy_match &&
1089 strstart - hash_head <= MAX_DIST) {
1090 /* To simplify the code, we prevent matches with the string
1091 * of window index 0 (in particular we have to avoid a match
1092 * of the string with itself at the start of the input file).
1093 */
1094 match_length = longest_match(hash_head);
1095 /* longest_match() sets match_start */
1096 if (match_length > lookahead)
1097 match_length = lookahead;
1098
1099 /* Ignore a length 3 match if it is too distant: */
1100 if (match_length == MIN_MATCH && strstart - match_start > TOO_FAR) {
1101 /* If prev_match is also MIN_MATCH, match_start is garbage
1102 * but we will ignore the current match anyway.
1103 */
1104 match_length--;
1105 }
1106 }
1107 /* If there was a match at the previous step and the current
1108 * match is not better, output the previous match:
1109 */
1110 if (prev_length >= MIN_MATCH && match_length <= prev_length) {
1111
1112 check_match(strstart - 1, prev_match, prev_length);
1113
1114 flush =
1115 ct_tally(strstart - 1 - prev_match, prev_length - MIN_MATCH);
1116
1117 /* Insert in hash table all strings up to the end of the match.
1118 * strstart-1 and strstart are already inserted.
1119 */
1120 lookahead -= prev_length - 1;
1121 prev_length -= 2;
1122 do {
1123 strstart++;
1124 INSERT_STRING(strstart, hash_head);
1125 /* strstart never exceeds WSIZE-MAX_MATCH, so there are
1126 * always MIN_MATCH bytes ahead. If lookahead < MIN_MATCH
1127 * these bytes are garbage, but it does not matter since the
1128 * next lookahead bytes will always be emitted as literals.
1129 */
1130 } while (--prev_length != 0);
1131 match_available = 0;
1132 match_length = MIN_MATCH - 1;
1133 strstart++;
1134 if (flush)
1135 FLUSH_BLOCK(0), block_start = strstart;
1136
1137 } else if (match_available) {
1138 /* If there was no match at the previous position, output a
1139 * single literal. If there was a match but the current match
1140 * is longer, truncate the previous match to a single literal.
1141 */
1142 Tracevv((stderr, "%c", window[strstart - 1]));
1143 if (ct_tally(0, window[strstart - 1])) {
1144 FLUSH_BLOCK(0), block_start = strstart;
1145 }
1146 strstart++;
1147 lookahead--;
1148 } else {
1149 /* There is no previous match to compare with, wait for
1150 * the next step to decide.
1151 */
1152 match_available = 1;
1153 strstart++;
1154 lookahead--;
1155 }
1156 Assert(strstart <= isize && lookahead <= isize, "a bit too far");
1157
1158 /* Make sure that we always have enough lookahead, except
1159 * at the end of the input file. We need MAX_MATCH bytes
1160 * for the next match, plus MIN_MATCH bytes to insert the
1161 * string following the next match.
1162 */
1163 while (lookahead < MIN_LOOKAHEAD && !eofile)
1164 fill_window();
1165 }
1166 if (match_available)
1167 ct_tally(0, window[strstart - 1]);
1168
1169 return FLUSH_BLOCK(1); /* eof */
1170}
1171
1172/* gzip (GNU zip) -- compress files with zip algorithm and 'compress' interface
1173 * Copyright (C) 1992-1993 Jean-loup Gailly
1174 * The unzip code was written and put in the public domain by Mark Adler.
1175 * Portions of the lzw code are derived from the public domain 'compress'
1176 * written by Spencer Thomas, Joe Orost, James Woods, Jim McKie, Steve Davies,
1177 * Ken Turkowski, Dave Mack and Peter Jannesen.
1178 *
1179 * See the license_msg below and the file COPYING for the software license.
1180 * See the file algorithm.doc for the compression algorithms and file formats.
1181 */
1182
1183/* Compress files with zip algorithm and 'compress' interface.
1184 * See usage() and help() functions below for all options.
1185 * Outputs:
1186 * file.gz: compressed file with same mode, owner, and utimes
1187 * or stdout with -c option or if stdin used as input.
1188 * If the output file name had to be truncated, the original name is kept
1189 * in the compressed file.
1190 */
1191
1192 /* configuration */
1193
1194typedef struct dirent dir_type;
1195
1196typedef RETSIGTYPE(*sig_type) (int);
1197
1198/* ======================================================================== */
1199int gzip_main(int argc, char **argv)
1200{
1201 int result;
1202 int inFileNum;
1203 int outFileNum;
1204 struct stat statBuf;
1205 char *delFileName;
1206 int tostdout = 0;
1207 int force = 0;
1208 int opt;
1209
1210 while ((opt = getopt(argc, argv, "cf0123456789dq")) != -1) {
1211 switch (opt) {
1212 case 'c':
1213 tostdout = 1;
1214 break;
1215 case 'f':
1216 force = 1;
1217 break;
1218
1219 case '0':
1220 just_store = 1;
1221 break;
1222
1223 /* Ignore 1-9 (compression level) options */
1224 case '1':
1225 case '2':
1226 case '3':
1227 case '4':
1228 case '5':
1229 case '6':
1230 case '7':
1231 case '8':
1232 case '9':
1233 break;
1234 case 'q':
1235 break;
1236#ifdef CONFIG_GUNZIP
1237 case 'd':
1238 optind = 1;
1239 return gunzip_main(argc, argv);
1240#endif
1241 default:
1242 bb_show_usage();
1243 }
1244 }
1245
1246 foreground = signal(SIGINT, SIG_IGN) != SIG_IGN;
1247 if (foreground) {
1248 (void) signal(SIGINT, (sig_type) abort_gzip);
1249 }
1250#ifdef SIGTERM
1251 if (signal(SIGTERM, SIG_IGN) != SIG_IGN) {
1252 (void) signal(SIGTERM, (sig_type) abort_gzip);
1253 }
1254#endif
1255#ifdef SIGHUP
1256 if (signal(SIGHUP, SIG_IGN) != SIG_IGN) {
1257 (void) signal(SIGHUP, (sig_type) abort_gzip);
1258 }
1259#endif
1260
1261 strncpy(z_suffix, Z_SUFFIX, sizeof(z_suffix) - 1);
1262 z_len = strlen(z_suffix);
1263
1264 /* Allocate all global buffers (for DYN_ALLOC option) */
1265 ALLOC(uch, inbuf, INBUFSIZ + INBUF_EXTRA);
1266 ALLOC(uch, outbuf, OUTBUFSIZ + OUTBUF_EXTRA);
1267 ALLOC(ush, d_buf, DIST_BUFSIZE);
1268 ALLOC(uch, window, 2L * WSIZE);
1269 ALLOC(ush, tab_prefix, 1L << BITS);
1270
1271 clear_bufs();
1272 part_nb = 0;
1273
1274 if (optind == argc) {
1275 time_stamp = 0;
1276 ifile_size = -1L;
1277 zip(STDIN_FILENO, STDOUT_FILENO);
1278 } else {
1279 int i;
1280
1281 for (i = optind; i < argc; i++) {
1282 char *path = NULL;
1283
1284 if (strcmp(argv[i], "-") == 0) {
1285 time_stamp = 0;
1286 ifile_size = -1L;
1287 inFileNum = STDIN_FILENO;
1288 outFileNum = STDOUT_FILENO;
1289 } else {
1290 inFileNum = open(argv[i], O_RDONLY);
1291 if (inFileNum < 0 || fstat(inFileNum, &statBuf) < 0)
1292 bb_perror_msg_and_die("%s", argv[i]);
1293 time_stamp = statBuf.st_ctime;
1294 ifile_size = statBuf.st_size;
1295
1296 if (!tostdout) {
1297 path = xmalloc(strlen(argv[i]) + 4);
1298 strcpy(path, argv[i]);
1299 strcat(path, ".gz");
1300
1301 /* Open output file */
1302#if (__GLIBC__ >= 2) && (__GLIBC_MINOR__ >= 1)
1303 outFileNum =
1304 open(path, O_RDWR | O_CREAT | O_EXCL | O_NOFOLLOW);
1305#else
1306 outFileNum = open(path, O_RDWR | O_CREAT | O_EXCL);
1307#endif
1308 if (outFileNum < 0) {
1309 bb_perror_msg("%s", path);
1310 free(path);
1311 continue;
1312 }
1313
1314 /* Set permissions on the file */
1315 fchmod(outFileNum, statBuf.st_mode);
1316 } else
1317 outFileNum = STDOUT_FILENO;
1318 }
1319
1320 if (path == NULL && isatty(outFileNum) && force == 0) {
1321 bb_error_msg
1322 ("compressed data not written to a terminal. Use -f to force compression.");
1323 free(path);
1324 continue;
1325 }
1326
1327 result = zip(inFileNum, outFileNum);
1328
1329 if (path != NULL) {
1330 close(inFileNum);
1331 close(outFileNum);
1332
1333 /* Delete the original file */
1334 if (result == OK)
1335 delFileName = argv[i];
1336 else
1337 delFileName = path;
1338
1339 if (unlink(delFileName) < 0)
1340 bb_perror_msg("%s", delFileName);
1341 }
1342
1343 free(path);
1344 }
1345 }
1346
1347 return (exit_code);
1348}
1349
1350/* trees.c -- output deflated data using Huffman coding
1351 * Copyright (C) 1992-1993 Jean-loup Gailly
1352 * This is free software; you can redistribute it and/or modify it under the
1353 * terms of the GNU General Public License, see the file COPYING.
1354 */
1355
1356/*
1357 * PURPOSE
1358 *
1359 * Encode various sets of source values using variable-length
1360 * binary code trees.
1361 *
1362 * DISCUSSION
1363 *
1364 * The PKZIP "deflation" process uses several Huffman trees. The more
1365 * common source values are represented by shorter bit sequences.
1366 *
1367 * Each code tree is stored in the ZIP file in a compressed form
1368 * which is itself a Huffman encoding of the lengths of
1369 * all the code strings (in ascending order by source values).
1370 * The actual code strings are reconstructed from the lengths in
1371 * the UNZIP process, as described in the "application note"
1372 * (APPNOTE.TXT) distributed as part of PKWARE's PKZIP program.
1373 *
1374 * REFERENCES
1375 *
1376 * Lynch, Thomas J.
1377 * Data Compression: Techniques and Applications, pp. 53-55.
1378 * Lifetime Learning Publications, 1985. ISBN 0-534-03418-7.
1379 *
1380 * Storer, James A.
1381 * Data Compression: Methods and Theory, pp. 49-50.
1382 * Computer Science Press, 1988. ISBN 0-7167-8156-5.
1383 *
1384 * Sedgewick, R.
1385 * Algorithms, p290.
1386 * Addison-Wesley, 1983. ISBN 0-201-06672-6.
1387 *
1388 * INTERFACE
1389 *
1390 * void ct_init (ush *attr, int *methodp)
1391 * Allocate the match buffer, initialize the various tables and save
1392 * the location of the internal file attribute (ascii/binary) and
1393 * method (DEFLATE/STORE)
1394 *
1395 * void ct_tally (int dist, int lc);
1396 * Save the match info and tally the frequency counts.
1397 *
1398 * long flush_block (char *buf, ulg stored_len, int eof)
1399 * Determine the best encoding for the current block: dynamic trees,
1400 * static trees or store, and output the encoded block to the zip
1401 * file. Returns the total compressed length for the file so far.
1402 *
1403 */
1404
1405/* ===========================================================================
1406 * Constants
1407 */
1408
1409#define MAX_BITS 15
1410/* All codes must not exceed MAX_BITS bits */
1411
1412#define MAX_BL_BITS 7
1413/* Bit length codes must not exceed MAX_BL_BITS bits */
1414
1415#define LENGTH_CODES 29
1416/* number of length codes, not counting the special END_BLOCK code */
1417
1418#define LITERALS 256
1419/* number of literal bytes 0..255 */
1420
1421#define END_BLOCK 256
1422/* end of block literal code */
1423
1424#define L_CODES (LITERALS+1+LENGTH_CODES)
1425/* number of Literal or Length codes, including the END_BLOCK code */
1426
1427#define D_CODES 30
1428/* number of distance codes */
1429
1430#define BL_CODES 19
1431/* number of codes used to transfer the bit lengths */
1432
1433typedef uch extra_bits_t;
1434
1435/* extra bits for each length code */
1436static const extra_bits_t extra_lbits[LENGTH_CODES]
1437 = { 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4,
1438 4, 4, 5, 5, 5, 5, 0
1439};
1440
1441/* extra bits for each distance code */
1442static const extra_bits_t extra_dbits[D_CODES]
1443 = { 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9,
1444 10, 10, 11, 11, 12, 12, 13, 13
1445};
1446
1447/* extra bits for each bit length code */
1448static const extra_bits_t extra_blbits[BL_CODES]
1449= { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 3, 7 };
1450
1451#define STORED_BLOCK 0
1452#define STATIC_TREES 1
1453#define DYN_TREES 2
1454/* The three kinds of block type */
1455
1456#ifndef LIT_BUFSIZE
1457# ifdef SMALL_MEM
1458# define LIT_BUFSIZE 0x2000
1459# else
1460# ifdef MEDIUM_MEM
1461# define LIT_BUFSIZE 0x4000
1462# else
1463# define LIT_BUFSIZE 0x8000
1464# endif
1465# endif
1466#endif
1467#ifndef DIST_BUFSIZE
1468# define DIST_BUFSIZE LIT_BUFSIZE
1469#endif
1470/* Sizes of match buffers for literals/lengths and distances. There are
1471 * 4 reasons for limiting LIT_BUFSIZE to 64K:
1472 * - frequencies can be kept in 16 bit counters
1473 * - if compression is not successful for the first block, all input data is
1474 * still in the window so we can still emit a stored block even when input
1475 * comes from standard input. (This can also be done for all blocks if
1476 * LIT_BUFSIZE is not greater than 32K.)
1477 * - if compression is not successful for a file smaller than 64K, we can
1478 * even emit a stored file instead of a stored block (saving 5 bytes).
1479 * - creating new Huffman trees less frequently may not provide fast
1480 * adaptation to changes in the input data statistics. (Take for
1481 * example a binary file with poorly compressible code followed by
1482 * a highly compressible string table.) Smaller buffer sizes give
1483 * fast adaptation but have of course the overhead of transmitting trees
1484 * more frequently.
1485 * - I can't count above 4
1486 * The current code is general and allows DIST_BUFSIZE < LIT_BUFSIZE (to save
1487 * memory at the expense of compression). Some optimizations would be possible
1488 * if we rely on DIST_BUFSIZE == LIT_BUFSIZE.
1489 */
1490#if LIT_BUFSIZE > INBUFSIZ
1491#error cannot overlay l_buf and inbuf
1492#endif
1493#define REP_3_6 16
1494/* repeat previous bit length 3-6 times (2 bits of repeat count) */
1495#define REPZ_3_10 17
1496/* repeat a zero length 3-10 times (3 bits of repeat count) */
1497#define REPZ_11_138 18
1498/* repeat a zero length 11-138 times (7 bits of repeat count) */
1499
1500/* ===========================================================================
1501 * Local data
1502 */
1503
1504/* Data structure describing a single value and its code string. */
1505typedef struct ct_data {
1506 union {
1507 ush freq; /* frequency count */
1508 ush code; /* bit string */
1509 } fc;
1510 union {
1511 ush dad; /* father node in Huffman tree */
1512 ush len; /* length of bit string */
1513 } dl;
1514} ct_data;
1515
1516#define Freq fc.freq
1517#define Code fc.code
1518#define Dad dl.dad
1519#define Len dl.len
1520
1521#define HEAP_SIZE (2*L_CODES+1)
1522/* maximum heap size */
1523
1524static ct_data dyn_ltree[HEAP_SIZE]; /* literal and length tree */
1525static ct_data dyn_dtree[2 * D_CODES + 1]; /* distance tree */
1526
1527static ct_data static_ltree[L_CODES + 2];
1528
1529/* The static literal tree. Since the bit lengths are imposed, there is no
1530 * need for the L_CODES extra codes used during heap construction. However
1531 * The codes 286 and 287 are needed to build a canonical tree (see ct_init
1532 * below).
1533 */
1534
1535static ct_data static_dtree[D_CODES];
1536
1537/* The static distance tree. (Actually a trivial tree since all codes use
1538 * 5 bits.)
1539 */
1540
1541static ct_data bl_tree[2 * BL_CODES + 1];
1542
1543/* Huffman tree for the bit lengths */
1544
1545typedef struct tree_desc {
1546 ct_data *dyn_tree; /* the dynamic tree */
1547 ct_data *static_tree; /* corresponding static tree or NULL */
1548 const extra_bits_t *extra_bits; /* extra bits for each code or NULL */
1549 int extra_base; /* base index for extra_bits */
1550 int elems; /* max number of elements in the tree */
1551 int max_length; /* max bit length for the codes */
1552 int max_code; /* largest code with non zero frequency */
1553} tree_desc;
1554
1555static tree_desc l_desc =
1556 { dyn_ltree, static_ltree, extra_lbits, LITERALS + 1, L_CODES,
1557 MAX_BITS, 0
1558};
1559
1560static tree_desc d_desc =
1561 { dyn_dtree, static_dtree, extra_dbits, 0, D_CODES, MAX_BITS, 0 };
1562
1563static tree_desc bl_desc =
1564 { bl_tree, (ct_data *) 0, extra_blbits, 0, BL_CODES, MAX_BL_BITS,
1565 0
1566};
1567
1568
1569static ush bl_count[MAX_BITS + 1];
1570
1571/* number of codes at each bit length for an optimal tree */
1572
1573static const uch bl_order[BL_CODES]
1574= { 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15 };
1575
1576/* The lengths of the bit length codes are sent in order of decreasing
1577 * probability, to avoid transmitting the lengths for unused bit length codes.
1578 */
1579
1580static int heap[2 * L_CODES + 1]; /* heap used to build the Huffman trees */
1581static int heap_len; /* number of elements in the heap */
1582static int heap_max; /* element of largest frequency */
1583
1584/* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used.
1585 * The same heap array is used to build all trees.
1586 */
1587
1588static uch depth[2 * L_CODES + 1];
1589
1590/* Depth of each subtree used as tie breaker for trees of equal frequency */
1591
1592static uch length_code[MAX_MATCH - MIN_MATCH + 1];
1593
1594/* length code for each normalized match length (0 == MIN_MATCH) */
1595
1596static uch dist_code[512];
1597
1598/* distance codes. The first 256 values correspond to the distances
1599 * 3 .. 258, the last 256 values correspond to the top 8 bits of
1600 * the 15 bit distances.
1601 */
1602
1603static int base_length[LENGTH_CODES];
1604
1605/* First normalized length for each code (0 = MIN_MATCH) */
1606
1607static int base_dist[D_CODES];
1608
1609/* First normalized distance for each code (0 = distance of 1) */
1610
1611#define l_buf inbuf
1612/* DECLARE(uch, l_buf, LIT_BUFSIZE); buffer for literals or lengths */
1613
1614/* DECLARE(ush, d_buf, DIST_BUFSIZE); buffer for distances */
1615
1616static uch flag_buf[(LIT_BUFSIZE / 8)];
1617
1618/* flag_buf is a bit array distinguishing literals from lengths in
1619 * l_buf, thus indicating the presence or absence of a distance.
1620 */
1621
1622static unsigned last_lit; /* running index in l_buf */
1623static unsigned last_dist; /* running index in d_buf */
1624static unsigned last_flags; /* running index in flag_buf */
1625static uch flags; /* current flags not yet saved in flag_buf */
1626static uch flag_bit; /* current bit used in flags */
1627
1628/* bits are filled in flags starting at bit 0 (least significant).
1629 * Note: these flags are overkill in the current code since we don't
1630 * take advantage of DIST_BUFSIZE == LIT_BUFSIZE.
1631 */
1632
1633static ulg opt_len; /* bit length of current block with optimal trees */
1634static ulg static_len; /* bit length of current block with static trees */
1635
1636static ulg compressed_len; /* total bit length of compressed file */
1637
1638
1639static ush *file_type; /* pointer to UNKNOWN, BINARY or ASCII */
1640static int *file_method; /* pointer to DEFLATE or STORE */
1641
1642/* ===========================================================================
1643 * Local (static) routines in this file.
1644 */
1645
1646static void init_block(void);
1647static void pqdownheap(ct_data * tree, int k);
1648static void gen_bitlen(tree_desc * desc);
1649static void gen_codes(ct_data * tree, int max_code);
1650static void build_tree(tree_desc * desc);
1651static void scan_tree(ct_data * tree, int max_code);
1652static void send_tree(ct_data * tree, int max_code);
1653static int build_bl_tree(void);
1654static void send_all_trees(int lcodes, int dcodes, int blcodes);
1655static void compress_block(ct_data * ltree, ct_data * dtree);
1656static void set_file_type(void);
1657
1658
1659#ifndef DEBUG
1660# define send_code(c, tree) send_bits(tree[c].Code, tree[c].Len)
1661 /* Send a code of the given tree. c and tree must not have side effects */
1662
1663#else /* DEBUG */
1664# define send_code(c, tree) \
1665 { if (verbose>1) bb_error_msg("\ncd %3d ",(c)); \
1666 send_bits(tree[c].Code, tree[c].Len); }
1667#endif
1668
1669#define d_code(dist) \
1670 ((dist) < 256 ? dist_code[dist] : dist_code[256+((dist)>>7)])
1671/* Mapping from a distance to a distance code. dist is the distance - 1 and
1672 * must not have side effects. dist_code[256] and dist_code[257] are never
1673 * used.
1674 */
1675
1676/* the arguments must not have side effects */
1677
1678/* ===========================================================================
1679 * Allocate the match buffer, initialize the various tables and save the
1680 * location of the internal file attribute (ascii/binary) and method
1681 * (DEFLATE/STORE).
1682 */
1683static void ct_init(ush * attr, int *methodp)
1684{
1685 int n; /* iterates over tree elements */
1686 int bits; /* bit counter */
1687 int length; /* length value */
1688 int code; /* code value */
1689 int dist; /* distance index */
1690
1691 file_type = attr;
1692 file_method = methodp;
1693 compressed_len = 0L;
1694
1695 if (static_dtree[0].Len != 0)
1696 return; /* ct_init already called */
1697
1698 /* Initialize the mapping length (0..255) -> length code (0..28) */
1699 length = 0;
1700 for (code = 0; code < LENGTH_CODES - 1; code++) {
1701 base_length[code] = length;
1702 for (n = 0; n < (1 << extra_lbits[code]); n++) {
1703 length_code[length++] = (uch) code;
1704 }
1705 }
1706 Assert(length == 256, "ct_init: length != 256");
1707 /* Note that the length 255 (match length 258) can be represented
1708 * in two different ways: code 284 + 5 bits or code 285, so we
1709 * overwrite length_code[255] to use the best encoding:
1710 */
1711 length_code[length - 1] = (uch) code;
1712
1713 /* Initialize the mapping dist (0..32K) -> dist code (0..29) */
1714 dist = 0;
1715 for (code = 0; code < 16; code++) {
1716 base_dist[code] = dist;
1717 for (n = 0; n < (1 << extra_dbits[code]); n++) {
1718 dist_code[dist++] = (uch) code;
1719 }
1720 }
1721 Assert(dist == 256, "ct_init: dist != 256");
1722 dist >>= 7; /* from now on, all distances are divided by 128 */
1723 for (; code < D_CODES; code++) {
1724 base_dist[code] = dist << 7;
1725 for (n = 0; n < (1 << (extra_dbits[code] - 7)); n++) {
1726 dist_code[256 + dist++] = (uch) code;
1727 }
1728 }
1729 Assert(dist == 256, "ct_init: 256+dist != 512");
1730
1731 /* Construct the codes of the static literal tree */
1732 for (bits = 0; bits <= MAX_BITS; bits++)
1733 bl_count[bits] = 0;
1734 n = 0;
1735 while (n <= 143)
1736 static_ltree[n++].Len = 8, bl_count[8]++;
1737 while (n <= 255)
1738 static_ltree[n++].Len = 9, bl_count[9]++;
1739 while (n <= 279)
1740 static_ltree[n++].Len = 7, bl_count[7]++;
1741 while (n <= 287)
1742 static_ltree[n++].Len = 8, bl_count[8]++;
1743 /* Codes 286 and 287 do not exist, but we must include them in the
1744 * tree construction to get a canonical Huffman tree (longest code
1745 * all ones)
1746 */
1747 gen_codes((ct_data *) static_ltree, L_CODES + 1);
1748
1749 /* The static distance tree is trivial: */
1750 for (n = 0; n < D_CODES; n++) {
1751 static_dtree[n].Len = 5;
1752 static_dtree[n].Code = bi_reverse(n, 5);
1753 }
1754
1755 /* Initialize the first block of the first file: */
1756 init_block();
1757}
1758
1759/* ===========================================================================
1760 * Initialize a new block.
1761 */
1762static void init_block()
1763{
1764 int n; /* iterates over tree elements */
1765
1766 /* Initialize the trees. */
1767 for (n = 0; n < L_CODES; n++)
1768 dyn_ltree[n].Freq = 0;
1769 for (n = 0; n < D_CODES; n++)
1770 dyn_dtree[n].Freq = 0;
1771 for (n = 0; n < BL_CODES; n++)
1772 bl_tree[n].Freq = 0;
1773
1774 dyn_ltree[END_BLOCK].Freq = 1;
1775 opt_len = static_len = 0L;
1776 last_lit = last_dist = last_flags = 0;
1777 flags = 0;
1778 flag_bit = 1;
1779}
1780
1781#define SMALLEST 1
1782/* Index within the heap array of least frequent node in the Huffman tree */
1783
1784
1785/* ===========================================================================
1786 * Remove the smallest element from the heap and recreate the heap with
1787 * one less element. Updates heap and heap_len.
1788 */
1789#define pqremove(tree, top) \
1790{\
1791 top = heap[SMALLEST]; \
1792 heap[SMALLEST] = heap[heap_len--]; \
1793 pqdownheap(tree, SMALLEST); \
1794}
1795
1796/* ===========================================================================
1797 * Compares to subtrees, using the tree depth as tie breaker when
1798 * the subtrees have equal frequency. This minimizes the worst case length.
1799 */
1800#define smaller(tree, n, m) \
1801 (tree[n].Freq < tree[m].Freq || \
1802 (tree[n].Freq == tree[m].Freq && depth[n] <= depth[m]))
1803
1804/* ===========================================================================
1805 * Restore the heap property by moving down the tree starting at node k,
1806 * exchanging a node with the smallest of its two sons if necessary, stopping
1807 * when the heap property is re-established (each father smaller than its
1808 * two sons).
1809 */
1810static void pqdownheap(ct_data * tree, int k)
1811{
1812 int v = heap[k];
1813 int j = k << 1; /* left son of k */
1814
1815 while (j <= heap_len) {
1816 /* Set j to the smallest of the two sons: */
1817 if (j < heap_len && smaller(tree, heap[j + 1], heap[j]))
1818 j++;
1819
1820 /* Exit if v is smaller than both sons */
1821 if (smaller(tree, v, heap[j]))
1822 break;
1823
1824 /* Exchange v with the smallest son */
1825 heap[k] = heap[j];
1826 k = j;
1827
1828 /* And continue down the tree, setting j to the left son of k */
1829 j <<= 1;
1830 }
1831 heap[k] = v;
1832}
1833
1834/* ===========================================================================
1835 * Compute the optimal bit lengths for a tree and update the total bit length
1836 * for the current block.
1837 * IN assertion: the fields freq and dad are set, heap[heap_max] and
1838 * above are the tree nodes sorted by increasing frequency.
1839 * OUT assertions: the field len is set to the optimal bit length, the
1840 * array bl_count contains the frequencies for each bit length.
1841 * The length opt_len is updated; static_len is also updated if stree is
1842 * not null.
1843 */
1844static void gen_bitlen(tree_desc * desc)
1845{
1846 ct_data *tree = desc->dyn_tree;
1847 const extra_bits_t *extra = desc->extra_bits;
1848 int base = desc->extra_base;
1849 int max_code = desc->max_code;
1850 int max_length = desc->max_length;
1851 ct_data *stree = desc->static_tree;
1852 int h; /* heap index */
1853 int n, m; /* iterate over the tree elements */
1854 int bits; /* bit length */
1855 int xbits; /* extra bits */
1856 ush f; /* frequency */
1857 int overflow = 0; /* number of elements with bit length too large */
1858
1859 for (bits = 0; bits <= MAX_BITS; bits++)
1860 bl_count[bits] = 0;
1861
1862 /* In a first pass, compute the optimal bit lengths (which may
1863 * overflow in the case of the bit length tree).
1864 */
1865 tree[heap[heap_max]].Len = 0; /* root of the heap */
1866
1867 for (h = heap_max + 1; h < HEAP_SIZE; h++) {
1868 n = heap[h];
1869 bits = tree[tree[n].Dad].Len + 1;
1870 if (bits > max_length)
1871 bits = max_length, overflow++;
1872 tree[n].Len = (ush) bits;
1873 /* We overwrite tree[n].Dad which is no longer needed */
1874
1875 if (n > max_code)
1876 continue; /* not a leaf node */
1877
1878 bl_count[bits]++;
1879 xbits = 0;
1880 if (n >= base)
1881 xbits = extra[n - base];
1882 f = tree[n].Freq;
1883 opt_len += (ulg) f *(bits + xbits);
1884
1885 if (stree)
1886 static_len += (ulg) f *(stree[n].Len + xbits);
1887 }
1888 if (overflow == 0)
1889 return;
1890
1891 Trace((stderr, "\nbit length overflow\n"));
1892 /* This happens for example on obj2 and pic of the Calgary corpus */
1893
1894 /* Find the first bit length which could increase: */
1895 do {
1896 bits = max_length - 1;
1897 while (bl_count[bits] == 0)
1898 bits--;
1899 bl_count[bits]--; /* move one leaf down the tree */
1900 bl_count[bits + 1] += 2; /* move one overflow item as its brother */
1901 bl_count[max_length]--;
1902 /* The brother of the overflow item also moves one step up,
1903 * but this does not affect bl_count[max_length]
1904 */
1905 overflow -= 2;
1906 } while (overflow > 0);
1907
1908 /* Now recompute all bit lengths, scanning in increasing frequency.
1909 * h is still equal to HEAP_SIZE. (It is simpler to reconstruct all
1910 * lengths instead of fixing only the wrong ones. This idea is taken
1911 * from 'ar' written by Haruhiko Okumura.)
1912 */
1913 for (bits = max_length; bits != 0; bits--) {
1914 n = bl_count[bits];
1915 while (n != 0) {
1916 m = heap[--h];
1917 if (m > max_code)
1918 continue;
1919 if (tree[m].Len != (unsigned) bits) {
1920 Trace((stderr, "code %d bits %d->%d\n", m, tree[m].Len,
1921 bits));
1922 opt_len +=
1923 ((long) bits - (long) tree[m].Len) * (long) tree[m].Freq;
1924 tree[m].Len = (ush) bits;
1925 }
1926 n--;
1927 }
1928 }
1929}
1930
1931/* ===========================================================================
1932 * Generate the codes for a given tree and bit counts (which need not be
1933 * optimal).
1934 * IN assertion: the array bl_count contains the bit length statistics for
1935 * the given tree and the field len is set for all tree elements.
1936 * OUT assertion: the field code is set for all tree elements of non
1937 * zero code length.
1938 */
1939static void gen_codes(ct_data * tree, int max_code)
1940{
1941 ush next_code[MAX_BITS + 1]; /* next code value for each bit length */
1942 ush code = 0; /* running code value */
1943 int bits; /* bit index */
1944 int n; /* code index */
1945
1946 /* The distribution counts are first used to generate the code values
1947 * without bit reversal.
1948 */
1949 for (bits = 1; bits <= MAX_BITS; bits++) {
1950 next_code[bits] = code = (code + bl_count[bits - 1]) << 1;
1951 }
1952 /* Check that the bit counts in bl_count are consistent. The last code
1953 * must be all ones.
1954 */
1955 Assert(code + bl_count[MAX_BITS] - 1 == (1 << MAX_BITS) - 1,
1956 "inconsistent bit counts");
1957 Tracev((stderr, "\ngen_codes: max_code %d ", max_code));
1958
1959 for (n = 0; n <= max_code; n++) {
1960 int len = tree[n].Len;
1961
1962 if (len == 0)
1963 continue;
1964 /* Now reverse the bits */
1965 tree[n].Code = bi_reverse(next_code[len]++, len);
1966
1967 Tracec(tree != static_ltree,
1968 (stderr, "\nn %3d %c l %2d c %4x (%x) ", n,
1969 (isgraph(n) ? n : ' '), len, tree[n].Code,
1970 next_code[len] - 1));
1971 }
1972}
1973
1974/* ===========================================================================
1975 * Construct one Huffman tree and assigns the code bit strings and lengths.
1976 * Update the total bit length for the current block.
1977 * IN assertion: the field freq is set for all tree elements.
1978 * OUT assertions: the fields len and code are set to the optimal bit length
1979 * and corresponding code. The length opt_len is updated; static_len is
1980 * also updated if stree is not null. The field max_code is set.
1981 */
1982static void build_tree(tree_desc * desc)
1983{
1984 ct_data *tree = desc->dyn_tree;
1985 ct_data *stree = desc->static_tree;
1986 int elems = desc->elems;
1987 int n, m; /* iterate over heap elements */
1988 int max_code = -1; /* largest code with non zero frequency */
1989 int node = elems; /* next internal node of the tree */
1990
1991 /* Construct the initial heap, with least frequent element in
1992 * heap[SMALLEST]. The sons of heap[n] are heap[2*n] and heap[2*n+1].
1993 * heap[0] is not used.
1994 */
1995 heap_len = 0, heap_max = HEAP_SIZE;
1996
1997 for (n = 0; n < elems; n++) {
1998 if (tree[n].Freq != 0) {
1999 heap[++heap_len] = max_code = n;
2000 depth[n] = 0;
2001 } else {
2002 tree[n].Len = 0;
2003 }
2004 }
2005
2006 /* The pkzip format requires that at least one distance code exists,
2007 * and that at least one bit should be sent even if there is only one
2008 * possible code. So to avoid special checks later on we force at least
2009 * two codes of non zero frequency.
2010 */
2011 while (heap_len < 2) {
2012 int new = heap[++heap_len] = (max_code < 2 ? ++max_code : 0);
2013
2014 tree[new].Freq = 1;
2015 depth[new] = 0;
2016 opt_len--;
2017 if (stree)
2018 static_len -= stree[new].Len;
2019 /* new is 0 or 1 so it does not have extra bits */
2020 }
2021 desc->max_code = max_code;
2022
2023 /* The elements heap[heap_len/2+1 .. heap_len] are leaves of the tree,
2024 * establish sub-heaps of increasing lengths:
2025 */
2026 for (n = heap_len / 2; n >= 1; n--)
2027 pqdownheap(tree, n);
2028
2029 /* Construct the Huffman tree by repeatedly combining the least two
2030 * frequent nodes.
2031 */
2032 do {
2033 pqremove(tree, n); /* n = node of least frequency */
2034 m = heap[SMALLEST]; /* m = node of next least frequency */
2035
2036 heap[--heap_max] = n; /* keep the nodes sorted by frequency */
2037 heap[--heap_max] = m;
2038
2039 /* Create a new node father of n and m */
2040 tree[node].Freq = tree[n].Freq + tree[m].Freq;
2041 depth[node] = (uch) (MAX(depth[n], depth[m]) + 1);
2042 tree[n].Dad = tree[m].Dad = (ush) node;
2043#ifdef DUMP_BL_TREE
2044 if (tree == bl_tree) {
2045 bb_error_msg("\nnode %d(%d), sons %d(%d) %d(%d)",
2046 node, tree[node].Freq, n, tree[n].Freq, m, tree[m].Freq);
2047 }
2048#endif
2049 /* and insert the new node in the heap */
2050 heap[SMALLEST] = node++;
2051 pqdownheap(tree, SMALLEST);
2052
2053 } while (heap_len >= 2);
2054
2055 heap[--heap_max] = heap[SMALLEST];
2056
2057 /* At this point, the fields freq and dad are set. We can now
2058 * generate the bit lengths.
2059 */
2060 gen_bitlen((tree_desc *) desc);
2061
2062 /* The field len is now set, we can generate the bit codes */
2063 gen_codes((ct_data *) tree, max_code);
2064}
2065
2066/* ===========================================================================
2067 * Scan a literal or distance tree to determine the frequencies of the codes
2068 * in the bit length tree. Updates opt_len to take into account the repeat
2069 * counts. (The contribution of the bit length codes will be added later
2070 * during the construction of bl_tree.)
2071 */
2072static void scan_tree(ct_data * tree, int max_code)
2073{
2074 int n; /* iterates over all tree elements */
2075 int prevlen = -1; /* last emitted length */
2076 int curlen; /* length of current code */
2077 int nextlen = tree[0].Len; /* length of next code */
2078 int count = 0; /* repeat count of the current code */
2079 int max_count = 7; /* max repeat count */
2080 int min_count = 4; /* min repeat count */
2081
2082 if (nextlen == 0)
2083 max_count = 138, min_count = 3;
2084 tree[max_code + 1].Len = (ush) 0xffff; /* guard */
2085
2086 for (n = 0; n <= max_code; n++) {
2087 curlen = nextlen;
2088 nextlen = tree[n + 1].Len;
2089 if (++count < max_count && curlen == nextlen) {
2090 continue;
2091 } else if (count < min_count) {
2092 bl_tree[curlen].Freq += count;
2093 } else if (curlen != 0) {
2094 if (curlen != prevlen)
2095 bl_tree[curlen].Freq++;
2096 bl_tree[REP_3_6].Freq++;
2097 } else if (count <= 10) {
2098 bl_tree[REPZ_3_10].Freq++;
2099 } else {
2100 bl_tree[REPZ_11_138].Freq++;
2101 }
2102 count = 0;
2103 prevlen = curlen;
2104 if (nextlen == 0) {
2105 max_count = 138, min_count = 3;
2106 } else if (curlen == nextlen) {
2107 max_count = 6, min_count = 3;
2108 } else {
2109 max_count = 7, min_count = 4;
2110 }
2111 }
2112}
2113
2114/* ===========================================================================
2115 * Send a literal or distance tree in compressed form, using the codes in
2116 * bl_tree.
2117 */
2118static void send_tree(ct_data * tree, int max_code)
2119{
2120 int n; /* iterates over all tree elements */
2121 int prevlen = -1; /* last emitted length */
2122 int curlen; /* length of current code */
2123 int nextlen = tree[0].Len; /* length of next code */
2124 int count = 0; /* repeat count of the current code */
2125 int max_count = 7; /* max repeat count */
2126 int min_count = 4; /* min repeat count */
2127
2128/* tree[max_code+1].Len = -1; *//* guard already set */
2129 if (nextlen == 0)
2130 max_count = 138, min_count = 3;
2131
2132 for (n = 0; n <= max_code; n++) {
2133 curlen = nextlen;
2134 nextlen = tree[n + 1].Len;
2135 if (++count < max_count && curlen == nextlen) {
2136 continue;
2137 } else if (count < min_count) {
2138 do {
2139 send_code(curlen, bl_tree);
2140 } while (--count != 0);
2141
2142 } else if (curlen != 0) {
2143 if (curlen != prevlen) {
2144 send_code(curlen, bl_tree);
2145 count--;
2146 }
2147 Assert(count >= 3 && count <= 6, " 3_6?");
2148 send_code(REP_3_6, bl_tree);
2149 send_bits(count - 3, 2);
2150
2151 } else if (count <= 10) {
2152 send_code(REPZ_3_10, bl_tree);
2153 send_bits(count - 3, 3);
2154
2155 } else {
2156 send_code(REPZ_11_138, bl_tree);
2157 send_bits(count - 11, 7);
2158 }
2159 count = 0;
2160 prevlen = curlen;
2161 if (nextlen == 0) {
2162 max_count = 138, min_count = 3;
2163 } else if (curlen == nextlen) {
2164 max_count = 6, min_count = 3;
2165 } else {
2166 max_count = 7, min_count = 4;
2167 }
2168 }
2169}
2170
2171/* ===========================================================================
2172 * Construct the Huffman tree for the bit lengths and return the index in
2173 * bl_order of the last bit length code to send.
2174 */
2175static const int build_bl_tree()
2176{
2177 int max_blindex; /* index of last bit length code of non zero freq */
2178
2179 /* Determine the bit length frequencies for literal and distance trees */
2180 scan_tree((ct_data *) dyn_ltree, l_desc.max_code);
2181 scan_tree((ct_data *) dyn_dtree, d_desc.max_code);
2182
2183 /* Build the bit length tree: */
2184 build_tree((tree_desc *) (&bl_desc));
2185 /* opt_len now includes the length of the tree representations, except
2186 * the lengths of the bit lengths codes and the 5+5+4 bits for the counts.
2187 */
2188
2189 /* Determine the number of bit length codes to send. The pkzip format
2190 * requires that at least 4 bit length codes be sent. (appnote.txt says
2191 * 3 but the actual value used is 4.)
2192 */
2193 for (max_blindex = BL_CODES - 1; max_blindex >= 3; max_blindex--) {
2194 if (bl_tree[bl_order[max_blindex]].Len != 0)
2195 break;
2196 }
2197 /* Update opt_len to include the bit length tree and counts */
2198 opt_len += 3 * (max_blindex + 1) + 5 + 5 + 4;
2199 Tracev((stderr, "\ndyn trees: dyn %ld, stat %ld", opt_len, static_len));
2200
2201 return max_blindex;
2202}
2203
2204/* ===========================================================================
2205 * Send the header for a block using dynamic Huffman trees: the counts, the
2206 * lengths of the bit length codes, the literal tree and the distance tree.
2207 * IN assertion: lcodes >= 257, dcodes >= 1, blcodes >= 4.
2208 */
2209static void send_all_trees(int lcodes, int dcodes, int blcodes)
2210{
2211 int rank; /* index in bl_order */
2212
2213 Assert(lcodes >= 257 && dcodes >= 1 && blcodes >= 4, "not enough codes");
2214 Assert(lcodes <= L_CODES && dcodes <= D_CODES
2215 && blcodes <= BL_CODES, "too many codes");
2216 Tracev((stderr, "\nbl counts: "));
2217 send_bits(lcodes - 257, 5); /* not +255 as stated in appnote.txt */
2218 send_bits(dcodes - 1, 5);
2219 send_bits(blcodes - 4, 4); /* not -3 as stated in appnote.txt */
2220 for (rank = 0; rank < blcodes; rank++) {
2221 Tracev((stderr, "\nbl code %2d ", bl_order[rank]));
2222 send_bits(bl_tree[bl_order[rank]].Len, 3);
2223 }
2224 Tracev((stderr, "\nbl tree: sent %ld", bits_sent));
2225
2226 send_tree((ct_data *) dyn_ltree, lcodes - 1); /* send the literal tree */
2227 Tracev((stderr, "\nlit tree: sent %ld", bits_sent));
2228
2229 send_tree((ct_data *) dyn_dtree, dcodes - 1); /* send the distance tree */
2230 Tracev((stderr, "\ndist tree: sent %ld", bits_sent));
2231}
2232
2233/* ===========================================================================
2234 * Determine the best encoding for the current block: dynamic trees, static
2235 * trees or store, and output the encoded block to the zip file. This function
2236 * returns the total compressed length for the file so far.
2237 */
2238static ulg flush_block(char *buf, ulg stored_len, int eof)
2239{
2240 ulg opt_lenb, static_lenb; /* opt_len and static_len in bytes */
2241 int max_blindex; /* index of last bit length code of non zero freq */
2242
2243 flag_buf[last_flags] = flags; /* Save the flags for the last 8 items */
2244
2245 /* Check if the file is ascii or binary */
2246 if (*file_type == (ush) UNKNOWN)
2247 set_file_type();
2248
2249 /* Construct the literal and distance trees */
2250 build_tree((tree_desc *) (&l_desc));
2251 Tracev((stderr, "\nlit data: dyn %ld, stat %ld", opt_len, static_len));
2252
2253 build_tree((tree_desc *) (&d_desc));
2254 Tracev((stderr, "\ndist data: dyn %ld, stat %ld", opt_len, static_len));
2255 /* At this point, opt_len and static_len are the total bit lengths of
2256 * the compressed block data, excluding the tree representations.
2257 */
2258
2259 /* Build the bit length tree for the above two trees, and get the index
2260 * in bl_order of the last bit length code to send.
2261 */
2262 max_blindex = build_bl_tree();
2263
2264 /* Determine the best encoding. Compute first the block length in bytes */
2265 opt_lenb = (opt_len + 3 + 7) >> 3;
2266 static_lenb = (static_len + 3 + 7) >> 3;
2267
2268 Trace((stderr,
2269 "\nopt %lu(%lu) stat %lu(%lu) stored %lu lit %u dist %u ",
2270 opt_lenb, opt_len, static_lenb, static_len, stored_len,
2271 last_lit, last_dist));
2272
2273 if (static_lenb <= opt_lenb)
2274 opt_lenb = static_lenb;
2275
2276 /* If compression failed and this is the first and last block,
2277 * and if the zip file can be seeked (to rewrite the local header),
2278 * the whole file is transformed into a stored file:
2279 */
2280 if ((just_store || (stored_len <= opt_lenb)) && eof && compressed_len == 0L && seekable()) {
2281 /* Since LIT_BUFSIZE <= 2*WSIZE, the input data must be there: */
2282 if (buf == (char *) 0)
2283 bb_error_msg("block vanished");
2284
2285 copy_block(buf, (unsigned) stored_len, 0); /* without header */
2286 compressed_len = stored_len << 3;
2287 *file_method = STORED;
2288
2289 } else if ((just_store || (stored_len + 4 <= opt_lenb)) && buf != (char *) 0) {
2290 /* 4: two words for the lengths */
2291 /* The test buf != NULL is only necessary if LIT_BUFSIZE > WSIZE.
2292 * Otherwise we can't have processed more than WSIZE input bytes since
2293 * the last block flush, because compression would have been
2294 * successful. If LIT_BUFSIZE <= WSIZE, it is never too late to
2295 * transform a block into a stored block.
2296 */
2297 send_bits((STORED_BLOCK << 1) + eof, 3); /* send block type */
2298 compressed_len = (compressed_len + 3 + 7) & ~7L;
2299 compressed_len += (stored_len + 4) << 3;
2300
2301 copy_block(buf, (unsigned) stored_len, 1); /* with header */
2302
2303 } else if (static_lenb == opt_lenb) {
2304 send_bits((STATIC_TREES << 1) + eof, 3);
2305 compress_block((ct_data *) static_ltree, (ct_data *) static_dtree);
2306 compressed_len += 3 + static_len;
2307 } else {
2308 send_bits((DYN_TREES << 1) + eof, 3);
2309 send_all_trees(l_desc.max_code + 1, d_desc.max_code + 1,
2310 max_blindex + 1);
2311 compress_block((ct_data *) dyn_ltree, (ct_data *) dyn_dtree);
2312 compressed_len += 3 + opt_len;
2313 }
2314 Assert(compressed_len == bits_sent, "bad compressed size");
2315 init_block();
2316
2317 if (eof) {
2318 bi_windup();
2319 compressed_len += 7; /* align on byte boundary */
2320 }
2321 Tracev((stderr, "\ncomprlen %lu(%lu) ", compressed_len >> 3,
2322 compressed_len - 7 * eof));
2323
2324 return compressed_len >> 3;
2325}
2326
2327/* ===========================================================================
2328 * Save the match info and tally the frequency counts. Return true if
2329 * the current block must be flushed.
2330 */
2331static int ct_tally(int dist, int lc)
2332{
2333 l_buf[last_lit++] = (uch) lc;
2334 if (dist == 0) {
2335 /* lc is the unmatched char */
2336 dyn_ltree[lc].Freq++;
2337 } else {
2338 /* Here, lc is the match length - MIN_MATCH */
2339 dist--; /* dist = match distance - 1 */
2340 Assert((ush) dist < (ush) MAX_DIST &&
2341 (ush) lc <= (ush) (MAX_MATCH - MIN_MATCH) &&
2342 (ush) d_code(dist) < (ush) D_CODES, "ct_tally: bad match");
2343
2344 dyn_ltree[length_code[lc] + LITERALS + 1].Freq++;
2345 dyn_dtree[d_code(dist)].Freq++;
2346
2347 d_buf[last_dist++] = (ush) dist;
2348 flags |= flag_bit;
2349 }
2350 flag_bit <<= 1;
2351
2352 /* Output the flags if they fill a byte: */
2353 if ((last_lit & 7) == 0) {
2354 flag_buf[last_flags++] = flags;
2355 flags = 0, flag_bit = 1;
2356 }
2357 /* Try to guess if it is profitable to stop the current block here */
2358 if ((last_lit & 0xfff) == 0) {
2359 /* Compute an upper bound for the compressed length */
2360 ulg out_length = (ulg) last_lit * 8L;
2361 ulg in_length = (ulg) strstart - block_start;
2362 int dcode;
2363
2364 for (dcode = 0; dcode < D_CODES; dcode++) {
2365 out_length +=
2366 (ulg) dyn_dtree[dcode].Freq * (5L + extra_dbits[dcode]);
2367 }
2368 out_length >>= 3;
2369 Trace((stderr,
2370 "\nlast_lit %u, last_dist %u, in %ld, out ~%ld(%ld%%) ",
2371 last_lit, last_dist, in_length, out_length,
2372 100L - out_length * 100L / in_length));
2373 if (last_dist < last_lit / 2 && out_length < in_length / 2)
2374 return 1;
2375 }
2376 return (last_lit == LIT_BUFSIZE - 1 || last_dist == DIST_BUFSIZE);
2377 /* We avoid equality with LIT_BUFSIZE because of wraparound at 64K
2378 * on 16 bit machines and because stored blocks are restricted to
2379 * 64K-1 bytes.
2380 */
2381}
2382
2383/* ===========================================================================
2384 * Send the block data compressed using the given Huffman trees
2385 */
2386static void compress_block(ct_data * ltree, ct_data * dtree)
2387{
2388 unsigned dist; /* distance of matched string */
2389 int lc; /* match length or unmatched char (if dist == 0) */
2390 unsigned lx = 0; /* running index in l_buf */
2391 unsigned dx = 0; /* running index in d_buf */
2392 unsigned fx = 0; /* running index in flag_buf */
2393 uch flag = 0; /* current flags */
2394 unsigned code; /* the code to send */
2395 int extra; /* number of extra bits to send */
2396
2397 if (last_lit != 0)
2398 do {
2399 if ((lx & 7) == 0)
2400 flag = flag_buf[fx++];
2401 lc = l_buf[lx++];
2402 if ((flag & 1) == 0) {
2403 send_code(lc, ltree); /* send a literal byte */
2404 Tracecv(isgraph(lc), (stderr, " '%c' ", lc));
2405 } else {
2406 /* Here, lc is the match length - MIN_MATCH */
2407 code = length_code[lc];
2408 send_code(code + LITERALS + 1, ltree); /* send the length code */
2409 extra = extra_lbits[code];
2410 if (extra != 0) {
2411 lc -= base_length[code];
2412 send_bits(lc, extra); /* send the extra length bits */
2413 }
2414 dist = d_buf[dx++];
2415 /* Here, dist is the match distance - 1 */
2416 code = d_code(dist);
2417 Assert(code < D_CODES, "bad d_code");
2418
2419 send_code(code, dtree); /* send the distance code */
2420 extra = extra_dbits[code];
2421 if (extra != 0) {
2422 dist -= base_dist[code];
2423 send_bits(dist, extra); /* send the extra distance bits */
2424 }
2425 } /* literal or match pair ? */
2426 flag >>= 1;
2427 } while (lx < last_lit);
2428
2429 send_code(END_BLOCK, ltree);
2430}
2431
2432/* ===========================================================================
2433 * Set the file type to ASCII or BINARY, using a crude approximation:
2434 * binary if more than 20% of the bytes are <= 6 or >= 128, ascii otherwise.
2435 * IN assertion: the fields freq of dyn_ltree are set and the total of all
2436 * frequencies does not exceed 64K (to fit in an int on 16 bit machines).
2437 */
2438static void set_file_type()
2439{
2440 int n = 0;
2441 unsigned ascii_freq = 0;
2442 unsigned bin_freq = 0;
2443
2444 while (n < 7)
2445 bin_freq += dyn_ltree[n++].Freq;
2446 while (n < 128)
2447 ascii_freq += dyn_ltree[n++].Freq;
2448 while (n < LITERALS)
2449 bin_freq += dyn_ltree[n++].Freq;
2450 *file_type = bin_freq > (ascii_freq >> 2) ? BINARY : ASCII;
2451 if (*file_type == BINARY && translate_eol) {
2452 bb_error_msg("-l used on binary file");
2453 }
2454}
2455
2456/* zip.c -- compress files to the gzip or pkzip format
2457 * Copyright (C) 1992-1993 Jean-loup Gailly
2458 * This is free software; you can redistribute it and/or modify it under the
2459 * terms of the GNU General Public License, see the file COPYING.
2460 */
2461
2462
2463static ulg crc; /* crc on uncompressed file data */
2464static long header_bytes; /* number of bytes in gzip header */
2465
2466static void put_long(ulg n)
2467{
2468 put_short((n) & 0xffff);
2469 put_short(((ulg) (n)) >> 16);
2470}
2471
2472/* put_header_byte is used for the compressed output
2473 * - for the initial 4 bytes that can't overflow the buffer.
2474 */
2475#define put_header_byte(c) {outbuf[outcnt++]=(uch)(c);}
2476
2477/* ===========================================================================
2478 * Deflate in to out.
2479 * IN assertions: the input and output buffers are cleared.
2480 * The variables time_stamp and save_orig_name are initialized.
2481 */
2482static int zip(int in, int out)
2483{
2484 uch my_flags = 0; /* general purpose bit flags */
2485 ush attr = 0; /* ascii/binary flag */
2486 ush deflate_flags = 0; /* pkzip -es, -en or -ex equivalent */
2487
2488 ifd = in;
2489 ofd = out;
2490 outcnt = 0;
2491
2492 /* Write the header to the gzip file. See algorithm.doc for the format */
2493
2494
2495 method = DEFLATED;
2496 put_header_byte(GZIP_MAGIC[0]); /* magic header */
2497 put_header_byte(GZIP_MAGIC[1]);
2498 put_header_byte(DEFLATED); /* compression method */
2499
2500 put_header_byte(my_flags); /* general flags */
2501 put_long(time_stamp);
2502
2503 /* Write deflated file to zip file */
2504 crc = updcrc(0, 0);
2505
2506 bi_init(out);
2507 ct_init(&attr, &method);
2508 lm_init(&deflate_flags);
2509
2510 put_byte((uch) deflate_flags); /* extra flags */
2511 put_byte(OS_CODE); /* OS identifier */
2512
2513 header_bytes = (long) outcnt;
2514
2515 (void) deflate();
2516
2517 /* Write the crc and uncompressed size */
2518 put_long(crc);
2519 put_long(isize);
2520 header_bytes += 2 * sizeof(long);
2521
2522 flush_outbuf();
2523 return OK;
2524}
2525
2526
2527/* ===========================================================================
2528 * Read a new buffer from the current input file, perform end-of-line
2529 * translation, and update the crc and input file size.
2530 * IN assertion: size >= 2 (for end-of-line translation)
2531 */
2532static int file_read(char *buf, unsigned size)
2533{
2534 unsigned len;
2535
2536 Assert(insize == 0, "inbuf not empty");
2537
2538 len = read(ifd, buf, size);
2539 if (len == (unsigned) (-1) || len == 0)
2540 return (int) len;
2541
2542 crc = updcrc((uch *) buf, len);
2543 isize += (ulg) len;
2544 return (int) len;
2545}
2546
2547/* ===========================================================================
2548 * Write the output buffer outbuf[0..outcnt-1] and update bytes_out.
2549 * (used for the compressed data only)
2550 */
2551static void flush_outbuf()
2552{
2553 if (outcnt == 0)
2554 return;
2555
2556 write_buf(ofd, (char *) outbuf, outcnt);
2557 outcnt = 0;
2558}
generated by cgit v1.1 at 2024-11-24 09:53:38 +0000