7 /* inflate.c -- Not copyrighted 1992 by Mark Adler
8 version c10p1, 10 January 1993 */
11 * Adapted for booting Linux by Hannu Savolainen 1993
14 * Nicolas Pitre <nico@cam.org>, 1999/04/14 :
15 * Little mods for all variable to reside either into rodata or bss segments
16 * by marking constant variables with 'const' and initializing all the others
17 * at run-time only. This allows for the kernel uncompressor to run
18 * directly from Flash or ROM memory on embedded systems.
22 Inflate deflated (PKZIP's method 8 compressed) data. The compression
23 method searches for as much of the current string of bytes (up to a
24 length of 258) in the previous 32 K bytes. If it doesn't find any
25 matches (of at least length 3), it codes the next byte. Otherwise, it
26 codes the length of the matched string and its distance backwards from
27 the current position. There is a single Huffman code that codes both
28 single bytes (called "literals") and match lengths. A second Huffman
29 code codes the distance information, which follows a length code. Each
30 length or distance code actually represents a base value and a number
31 of "extra" (sometimes zero) bits to get to add to the base value. At
32 the end of each deflated block is a special end-of-block (EOB) literal/
33 length code. The decoding process is basically: get a literal/length
34 code; if EOB then done; if a literal, emit the decoded byte; if a
35 length then get the distance and emit the referred-to bytes from the
36 sliding window of previously emitted data.
38 There are (currently) three kinds of inflate blocks: stored, fixed, and
39 dynamic. The compressor deals with some chunk of data at a time, and
40 decides which method to use on a chunk-by-chunk basis. A chunk might
41 typically be 32 K or 64 K. If the chunk is incompressible, then the
42 "stored" method is used. In this case, the bytes are simply stored as
43 is, eight bits per byte, with none of the above coding. The bytes are
44 preceded by a count, since there is no longer an EOB code.
46 If the data is compressible, then either the fixed or dynamic methods
47 are used. In the dynamic method, the compressed data is preceded by
48 an encoding of the literal/length and distance Huffman codes that are
49 to be used to decode this block. The representation is itself Huffman
50 coded, and so is preceded by a description of that code. These code
51 descriptions take up a little space, and so for small blocks, there is
52 a predefined set of codes, called the fixed codes. The fixed method is
53 used if the block codes up smaller that way (usually for quite small
54 chunks), otherwise the dynamic method is used. In the latter case, the
55 codes are customized to the probabilities in the current block, and so
56 can code it much better than the pre-determined fixed codes.
58 The Huffman codes themselves are decoded using a multi-level table
59 lookup, in order to maximize the speed of decoding plus the speed of
60 building the decoding tables. See the comments below that precede the
61 lbits and dbits tuning parameters.
66 Notes beyond the 1.93a appnote.txt:
68 1. Distance pointers never point before the beginning of the output
70 2. Distance pointers can point back across blocks, up to 32k away.
71 3. There is an implied maximum of 7 bits for the bit length table and
72 15 bits for the actual data.
73 4. If only one code exists, then it is encoded using one bit. (Zero
74 would be more efficient, but perhaps a little confusing.) If two
75 codes exist, they are coded using one bit each (0 and 1).
76 5. There is no way of sending zero distance codes--a dummy must be
77 sent if there are none. (History: a pre 2.0 version of PKZIP would
78 store blocks with no distance codes, but this was discovered to be
79 too harsh a criterion.) Valid only for 1.93a. 2.04c does allow
80 zero distance codes, which is sent as one code of zero bits in
82 6. There are up to 286 literal/length codes. Code 256 represents the
83 end-of-block. Note however that the static length tree defines
84 288 codes just to fill out the Huffman codes. Codes 286 and 287
85 cannot be used though, since there is no length base or extra bits
86 defined for them. Similarly, there are up to 30 distance codes.
87 However, static trees define 32 codes (all 5 bits) to fill out the
88 Huffman codes, but the last two had better not show up in the data.
89 7. Unzip can check dynamic Huffman blocks for complete code sets.
90 The exception is that a single code would not be complete (see #4).
91 8. The five bits following the block type is really the number of
92 literal codes sent minus 257.
93 9. Length codes 8,16,16 are interpreted as 13 length codes of 8 bits
94 (1+6+6). Therefore, to output three times the length, you output
95 three codes (1+1+1), whereas to output four times the same length,
96 you only need two codes (1+3). Hmm.
97 10. In the tree reconstruction algorithm, Code = Code + Increment
98 only if BitLength(i) is not zero. (Pretty obvious.)
99 11. Correction: 4 Bits: # of Bit Length codes - 4 (4 - 19)
100 12. Note: length code 284 can represent 227-258, but length code 285
101 really is 258. The last length deserves its own, short code
102 since it gets used a lot in very redundant files. The length
103 258 is special since 258 - 3 (the min match length) is 255.
104 13. The literal/length and distance code bit lengths are read as a
105 single stream of lengths. It is possible (and advantageous) for
106 a repeat code (16, 17, or 18) to go across the boundary between
107 the two sets of lengths.
111 static char rcsid
[] = "#Id: inflate.c,v 0.14 1993/06/10 13:27:04 jloup Exp #";
116 #if defined(STDC_HEADERS) || defined(HAVE_STDLIB_H)
117 # include <sys/types.h>
127 /* Huffman code lookup table entry--this entry is four bytes for machines
128 that have 16-bit pointers (e.g. PC's in the small or medium model).
129 Valid extra bits are 0..13. e == 15 is EOB (end of block), e == 16
130 means that v is a literal, 16 < e < 32 means that v is a pointer to
131 the next table, which codes e - 16 bits, and lastly e == 99 indicates
132 an unused code. If a code with e == 99 is looked up, this implies an
133 error in the data. */
135 uch e
; /* number of extra bits or operation */
136 uch b
; /* number of bits in this code or subcode */
138 ush n
; /* literal, length base, or distance base */
139 struct huft
*t
; /* pointer to next level of table */
144 /* Function prototypes */
145 STATIC
int huft_build
OF((unsigned *, unsigned, unsigned,
146 const ush
*, const ush
*, struct huft
**, int *));
147 STATIC
int huft_free
OF((struct huft
*));
148 STATIC
int inflate_codes
OF((struct huft
*, struct huft
*, int, int));
149 STATIC
int inflate_stored
OF((void));
150 STATIC
int inflate_fixed
OF((void));
151 STATIC
int inflate_dynamic
OF((void));
152 STATIC
int inflate_block
OF((int *));
153 STATIC
int inflate
OF((void));
156 /* The inflate algorithm uses a sliding 32 K byte window on the uncompressed
157 stream to find repeated byte strings. This is implemented here as a
158 circular buffer. The index is updated simply by incrementing and then
159 ANDing with 0x7fff (32K-1). */
160 /* It is left to other modules to supply the 32 K area. It is assumed
161 to be usable as if it were declared "uch slide[32768];" or as just
162 "uch *slide;" and then malloc'ed in the latter case. The definition
163 must be in unzip.h, included above. */
164 /* unsigned wp; current position in slide */
166 #define flush_output(w) (wp=(w),flush_window())
168 /* Tables for deflate from PKZIP's appnote.txt. */
169 static const unsigned border
[] = { /* Order of the bit length code lengths */
170 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
171 static const ush cplens
[] = { /* Copy lengths for literal codes 257..285 */
172 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
173 35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
174 /* note: see note #13 above about the 258 in this list. */
175 static const ush cplext
[] = { /* Extra bits for literal codes 257..285 */
176 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2,
177 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, 99, 99}; /* 99==invalid */
178 static const ush cpdist
[] = { /* Copy offsets for distance codes 0..29 */
179 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
180 257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
181 8193, 12289, 16385, 24577};
182 static const ush cpdext
[] = { /* Extra bits for distance codes */
183 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6,
184 7, 7, 8, 8, 9, 9, 10, 10, 11, 11,
189 /* Macros for inflate() bit peeking and grabbing.
193 x = b & mask_bits[j];
196 where NEEDBITS makes sure that b has at least j bits in it, and
197 DUMPBITS removes the bits from b. The macros use the variable k
198 for the number of bits in b. Normally, b and k are register
199 variables for speed, and are initialized at the beginning of a
200 routine that uses these macros from a global bit buffer and count.
202 If we assume that EOB will be the longest code, then we will never
203 ask for bits with NEEDBITS that are beyond the end of the stream.
204 So, NEEDBITS should not read any more bytes than are needed to
205 meet the request. Then no bytes need to be "returned" to the buffer
206 at the end of the last block.
208 However, this assumption is not true for fixed blocks--the EOB code
209 is 7 bits, but the other literal/length codes can be 8 or 9 bits.
210 (The EOB code is shorter than other codes because fixed blocks are
211 generally short. So, while a block always has an EOB, many other
212 literal/length codes have a significantly lower probability of
213 showing up at all.) However, by making the first table have a
214 lookup of seven bits, the EOB code will be found in that first
215 lookup, and so will not require that too many bits be pulled from
219 STATIC ulg bb
; /* bit buffer */
220 STATIC
unsigned bk
; /* bits in bit buffer */
222 STATIC
const ush mask_bits
[] = {
224 0x0001, 0x0003, 0x0007, 0x000f, 0x001f, 0x003f, 0x007f, 0x00ff,
225 0x01ff, 0x03ff, 0x07ff, 0x0fff, 0x1fff, 0x3fff, 0x7fff, 0xffff
228 #define NEXTBYTE() (uch)get_byte()
229 #define NEEDBITS(n) {while(k<(n)){b|=((ulg)NEXTBYTE())<<k;k+=8;}}
230 #define DUMPBITS(n) {b>>=(n);k-=(n);}
234 Huffman code decoding is performed using a multi-level table lookup.
235 The fastest way to decode is to simply build a lookup table whose
236 size is determined by the longest code. However, the time it takes
237 to build this table can also be a factor if the data being decoded
238 is not very long. The most common codes are necessarily the
239 shortest codes, so those codes dominate the decoding time, and hence
240 the speed. The idea is you can have a shorter table that decodes the
241 shorter, more probable codes, and then point to subsidiary tables for
242 the longer codes. The time it costs to decode the longer codes is
243 then traded against the time it takes to make longer tables.
245 This results of this trade are in the variables lbits and dbits
246 below. lbits is the number of bits the first level table for literal/
247 length codes can decode in one step, and dbits is the same thing for
248 the distance codes. Subsequent tables are also less than or equal to
249 those sizes. These values may be adjusted either when all of the
250 codes are shorter than that, in which case the longest code length in
251 bits is used, or when the shortest code is *longer* than the requested
252 table size, in which case the length of the shortest code in bits is
255 There are two different values for the two tables, since they code a
256 different number of possibilities each. The literal/length table
257 codes 286 possible values, or in a flat code, a little over eight
258 bits. The distance table codes 30 possible values, or a little less
259 than five bits, flat. The optimum values for speed end up being
260 about one bit more than those, so lbits is 8+1 and dbits is 5+1.
261 The optimum values may differ though from machine to machine, and
262 possibly even between compilers. Your mileage may vary.
266 STATIC
const int lbits
= 9; /* bits in base literal/length lookup table */
267 STATIC
const int dbits
= 6; /* bits in base distance lookup table */
270 /* If BMAX needs to be larger than 16, then h and x[] should be ulg. */
271 #define BMAX 16 /* maximum bit length of any code (16 for explode) */
272 #define N_MAX 288 /* maximum number of codes in any set */
275 STATIC
unsigned hufts
; /* track memory usage */
278 STATIC
int huft_build(b
, n
, s
, d
, e
, t
, m
)
279 unsigned *b
; /* code lengths in bits (all assumed <= BMAX) */
280 unsigned n
; /* number of codes (assumed <= N_MAX) */
281 unsigned s
; /* number of simple-valued codes (0..s-1) */
282 const ush
*d
; /* list of base values for non-simple codes */
283 const ush
*e
; /* list of extra bits for non-simple codes */
284 struct huft
**t
; /* result: starting table */
285 int *m
; /* maximum lookup bits, returns actual */
286 /* Given a list of code lengths and a maximum table size, make a set of
287 tables to decode that set of codes. Return zero on success, one if
288 the given code set is incomplete (the tables are still built in this
289 case), two if the input is invalid (all zero length codes or an
290 oversubscribed set of lengths), and three if not enough memory. */
292 unsigned a
; /* counter for codes of length k */
293 unsigned c
[BMAX
+1]; /* bit length count table */
294 unsigned f
; /* i repeats in table every f entries */
295 int g
; /* maximum code length */
296 int h
; /* table level */
297 register unsigned i
; /* counter, current code */
298 register unsigned j
; /* counter */
299 register int k
; /* number of bits in current code */
300 int l
; /* bits per table (returned in m) */
301 register unsigned *p
; /* pointer into c[], b[], or v[] */
302 register struct huft
*q
; /* points to current table */
303 struct huft r
; /* table entry for structure assignment */
304 struct huft
*u
[BMAX
]; /* table stack */
305 unsigned v
[N_MAX
]; /* values in order of bit length */
306 register int w
; /* bits before this table == (l * h) */
307 unsigned x
[BMAX
+1]; /* bit offsets, then code stack */
308 unsigned *xp
; /* pointer into x */
309 int y
; /* number of dummy codes added */
310 unsigned z
; /* number of entries in current table */
314 /* Generate counts for each bit length */
315 memzero(c
, sizeof(c
));
318 Tracecv(*p
, (stderr
, (n
-i
>= ' ' && n
-i
<= '~' ? "%c %d\n" : "0x%x %d\n"),
320 c
[*p
]++; /* assume all entries <= BMAX */
321 p
++; /* Can't combine with above line (Solaris bug) */
323 if (c
[0] == n
) /* null input--all zero length codes */
325 *t
= (struct huft
*)NULL
;
332 /* Find minimum and maximum length, bound *m by those */
334 for (j
= 1; j
<= BMAX
; j
++)
337 k
= j
; /* minimum code length */
340 for (i
= BMAX
; i
; i
--)
343 g
= i
; /* maximum code length */
350 /* Adjust last length count to fill out codes, if needed */
351 for (y
= 1 << j
; j
< i
; j
++, y
<<= 1)
353 return 2; /* bad input: more codes than bits */
360 /* Generate starting offsets into the value table for each length */
362 p
= c
+ 1; xp
= x
+ 2;
363 while (--i
) { /* note that i == g from above */
369 /* Make a table of values in order of bit lengths */
378 /* Generate the Huffman codes and for each, make the table entries */
379 x
[0] = i
= 0; /* first Huffman code is zero */
380 p
= v
; /* grab values in bit order */
381 h
= -1; /* no tables yet--level -1 */
382 w
= -l
; /* bits decoded == (l * h) */
383 u
[0] = (struct huft
*)NULL
; /* just to keep compilers happy */
384 q
= (struct huft
*)NULL
; /* ditto */
388 /* go through the bit lengths (k already is bits in shortest code) */
396 /* here i is the Huffman code of length k bits for value *p */
397 /* make tables up to required level */
402 w
+= l
; /* previous table always l bits */
404 /* compute minimum size table less than or equal to l bits */
405 z
= (z
= g
- w
) > (unsigned)l
? l
: z
; /* upper limit on table size */
406 if ((f
= 1 << (j
= k
- w
)) > a
+ 1) /* try a k-w bit table */
407 { /* too few codes for k-w bit table */
409 f
-= a
+ 1; /* deduct codes from patterns left */
411 while (++j
< z
) /* try smaller tables up to z bits */
413 if ((f
<<= 1) <= *++xp
)
414 break; /* enough codes to use up j bits */
415 f
-= *xp
; /* else deduct codes from patterns */
419 z
= 1 << j
; /* table entries for j-bit table */
421 /* allocate and link in new table */
422 if ((q
= (struct huft
*)malloc((z
+ 1)*sizeof(struct huft
))) ==
427 return 3; /* not enough memory */
430 hufts
+= z
+ 1; /* track memory usage */
431 *t
= q
+ 1; /* link to list for huft_free() */
432 *(t
= &(q
->v
.t
)) = (struct huft
*)NULL
;
433 u
[h
] = ++q
; /* table starts after link */
436 /* connect to last table, if there is one */
439 x
[h
] = i
; /* save pattern for backing up */
440 r
.b
= (uch
)l
; /* bits to dump before this table */
441 r
.e
= (uch
)(16 + j
); /* bits in this table */
442 r
.v
.t
= q
; /* pointer to this table */
443 j
= i
>> (w
- l
); /* (get around Turbo C bug) */
444 u
[h
-1][j
] = r
; /* connect to last table */
450 /* set up table entry in r */
453 r
.e
= 99; /* out of values--invalid code */
456 r
.e
= (uch
)(*p
< 256 ? 16 : 15); /* 256 is end-of-block code */
457 r
.v
.n
= (ush
)(*p
); /* simple code is just the value */
458 p
++; /* one compiler does not like *p++ */
462 r
.e
= (uch
)e
[*p
- s
]; /* non-simple--look up in lists */
467 /* fill code-like entries with r */
469 for (j
= i
>> w
; j
< z
; j
+= f
)
472 /* backwards increment the k-bit code i */
473 for (j
= 1 << (k
- 1); i
& j
; j
>>= 1)
477 /* backup over finished tables */
478 while ((i
& ((1 << w
) - 1)) != x
[h
])
480 h
--; /* don't need to update q */
490 /* Return true (1) if we were given an incomplete table */
491 return y
!= 0 && g
!= 1;
496 STATIC
int huft_free(t
)
497 struct huft
*t
; /* table to free */
498 /* Free the malloc'ed tables built by huft_build(), which makes a linked
499 list of the tables it made, with the links in a dummy first entry of
502 register struct huft
*p
, *q
;
505 /* Go through linked list, freeing from the malloced (t[-1]) address. */
507 while (p
!= (struct huft
*)NULL
)
517 STATIC
int inflate_codes(tl
, td
, bl
, bd
)
518 struct huft
*tl
, *td
; /* literal/length and distance decoder tables */
519 int bl
, bd
; /* number of bits decoded by tl[] and td[] */
520 /* inflate (decompress) the codes in a deflated (compressed) block.
521 Return an error code or zero if it all goes ok. */
523 register unsigned e
; /* table entry flag/number of extra bits */
524 unsigned n
, d
; /* length and index for copy */
525 unsigned w
; /* current window position */
526 struct huft
*t
; /* pointer to table entry */
527 unsigned ml
, md
; /* masks for bl and bd bits */
528 register ulg b
; /* bit buffer */
529 register unsigned k
; /* number of bits in bit buffer */
532 /* make local copies of globals */
533 b
= bb
; /* initialize bit buffer */
535 w
= wp
; /* initialize window position */
537 /* inflate the coded data */
538 ml
= mask_bits
[bl
]; /* precompute masks for speed */
540 for (;;) /* do until end of block */
542 NEEDBITS((unsigned)bl
)
543 if ((e
= (t
= tl
+ ((unsigned)b
& ml
))->e
) > 16)
550 } while ((e
= (t
= t
->v
.t
+ ((unsigned)b
& mask_bits
[e
]))->e
) > 16);
552 if (e
== 16) /* then it's a literal */
554 slide
[w
++] = (uch
)t
->v
.n
;
555 Tracevv((stderr
, "%c", slide
[w
-1]));
562 else /* it's an EOB or a length */
564 /* exit if end of block */
568 /* get length of block to copy */
570 n
= t
->v
.n
+ ((unsigned)b
& mask_bits
[e
]);
573 /* decode distance of block to copy */
574 NEEDBITS((unsigned)bd
)
575 if ((e
= (t
= td
+ ((unsigned)b
& md
))->e
) > 16)
582 } while ((e
= (t
= t
->v
.t
+ ((unsigned)b
& mask_bits
[e
]))->e
) > 16);
585 d
= w
- t
->v
.n
- ((unsigned)b
& mask_bits
[e
]);
587 Tracevv((stderr
,"\\[%d,%d]", w
-d
, n
));
591 n
-= (e
= (e
= WSIZE
- ((d
&= WSIZE
-1) > w
? d
: w
)) > n
? n
: e
);
592 #if !defined(NOMEMCPY) && !defined(DEBUG)
593 if (w
- d
>= e
) /* (this test assumes unsigned comparison) */
595 memcpy(slide
+ w
, slide
+ d
, e
);
599 else /* do it slow to avoid memcpy() overlap */
600 #endif /* !NOMEMCPY */
602 slide
[w
++] = slide
[d
++];
603 Tracevv((stderr
, "%c", slide
[w
-1]));
615 /* restore the globals from the locals */
616 wp
= w
; /* restore global window pointer */
617 bb
= b
; /* restore global bit buffer */
626 STATIC
int inflate_stored()
627 /* "decompress" an inflated type 0 (stored) block. */
629 unsigned n
; /* number of bytes in block */
630 unsigned w
; /* current window position */
631 register ulg b
; /* bit buffer */
632 register unsigned k
; /* number of bits in bit buffer */
636 /* make local copies of globals */
637 b
= bb
; /* initialize bit buffer */
639 w
= wp
; /* initialize window position */
642 /* go to byte boundary */
647 /* get the length and its complement */
649 n
= ((unsigned)b
& 0xffff);
652 if (n
!= (unsigned)((~b
) & 0xffff))
653 return 1; /* error in compressed data */
657 /* read and output the compressed data */
671 /* restore the globals from the locals */
672 wp
= w
; /* restore global window pointer */
673 bb
= b
; /* restore global bit buffer */
682 STATIC
int inflate_fixed()
683 /* decompress an inflated type 1 (fixed Huffman codes) block. We should
684 either replace this with a custom decoder, or at least precompute the
687 int i
; /* temporary variable */
688 struct huft
*tl
; /* literal/length code table */
689 struct huft
*td
; /* distance code table */
690 int bl
; /* lookup bits for tl */
691 int bd
; /* lookup bits for td */
692 unsigned l
[288]; /* length list for huft_build */
696 /* set up literal table */
697 for (i
= 0; i
< 144; i
++)
703 for (; i
< 288; i
++) /* make a complete, but wrong code set */
706 if ((i
= huft_build(l
, 288, 257, cplens
, cplext
, &tl
, &bl
)) != 0)
710 /* set up distance table */
711 for (i
= 0; i
< 30; i
++) /* make an incomplete code set */
714 if ((i
= huft_build(l
, 30, 0, cpdist
, cpdext
, &td
, &bd
)) > 1)
723 /* decompress until an end-of-block code */
724 if (inflate_codes(tl
, td
, bl
, bd
))
728 /* free the decoding tables, return */
736 STATIC
int inflate_dynamic()
737 /* decompress an inflated type 2 (dynamic Huffman codes) block. */
739 int i
; /* temporary variables */
741 unsigned l
; /* last length */
742 unsigned m
; /* mask for bit lengths table */
743 unsigned n
; /* number of lengths to get */
744 struct huft
*tl
; /* literal/length code table */
745 struct huft
*td
; /* distance code table */
746 int bl
; /* lookup bits for tl */
747 int bd
; /* lookup bits for td */
748 unsigned nb
; /* number of bit length codes */
749 unsigned nl
; /* number of literal/length codes */
750 unsigned nd
; /* number of distance codes */
751 #ifdef PKZIP_BUG_WORKAROUND
752 unsigned ll
[288+32]; /* literal/length and distance code lengths */
754 unsigned ll
[286+30]; /* literal/length and distance code lengths */
756 register ulg b
; /* bit buffer */
757 register unsigned k
; /* number of bits in bit buffer */
761 /* make local bit buffer */
766 /* read in table lengths */
768 nl
= 257 + ((unsigned)b
& 0x1f); /* number of literal/length codes */
771 nd
= 1 + ((unsigned)b
& 0x1f); /* number of distance codes */
774 nb
= 4 + ((unsigned)b
& 0xf); /* number of bit length codes */
776 #ifdef PKZIP_BUG_WORKAROUND
777 if (nl
> 288 || nd
> 32)
779 if (nl
> 286 || nd
> 30)
781 return 1; /* bad lengths */
785 /* read in bit-length-code lengths */
786 for (j
= 0; j
< nb
; j
++)
789 ll
[border
[j
]] = (unsigned)b
& 7;
797 /* build decoding table for trees--single level, 7 bit lookup */
799 if ((i
= huft_build(ll
, 19, 19, NULL
, NULL
, &tl
, &bl
)) != 0)
803 return i
; /* incomplete code set */
808 /* read in literal and distance code lengths */
812 while ((unsigned)i
< n
)
814 NEEDBITS((unsigned)bl
)
815 j
= (td
= tl
+ ((unsigned)b
& m
))->b
;
818 if (j
< 16) /* length of code in bits (0..15) */
819 ll
[i
++] = l
= j
; /* save last length in l */
820 else if (j
== 16) /* repeat last length 3 to 6 times */
823 j
= 3 + ((unsigned)b
& 3);
825 if ((unsigned)i
+ j
> n
)
830 else if (j
== 17) /* 3 to 10 zero length codes */
833 j
= 3 + ((unsigned)b
& 7);
835 if ((unsigned)i
+ j
> n
)
841 else /* j == 18: 11 to 138 zero length codes */
844 j
= 11 + ((unsigned)b
& 0x7f);
846 if ((unsigned)i
+ j
> n
)
856 /* free decoding table for trees */
861 /* restore the global bit buffer */
867 /* build the decoding tables for literal/length and distance codes */
869 if ((i
= huft_build(ll
, nl
, 257, cplens
, cplext
, &tl
, &bl
)) != 0)
873 error(" incomplete literal tree\n");
876 return i
; /* incomplete code set */
880 if ((i
= huft_build(ll
+ nl
, nd
, 0, cpdist
, cpdext
, &td
, &bd
)) != 0)
884 error(" incomplete distance tree\n");
885 #ifdef PKZIP_BUG_WORKAROUND
892 return i
; /* incomplete code set */
898 /* decompress until an end-of-block code */
899 if (inflate_codes(tl
, td
, bl
, bd
))
904 /* free the decoding tables, return */
914 STATIC
int inflate_block(e
)
915 int *e
; /* last block flag */
916 /* decompress an inflated block */
918 unsigned t
; /* block type */
919 register ulg b
; /* bit buffer */
920 register unsigned k
; /* number of bits in bit buffer */
924 /* make local bit buffer */
929 /* read in last block bit */
935 /* read in block type */
941 /* restore the global bit buffer */
945 /* inflate that block type */
947 return inflate_dynamic();
949 return inflate_stored();
951 return inflate_fixed();
962 /* decompress an inflated entry */
964 int e
; /* last block flag */
965 int r
; /* result code */
966 unsigned h
; /* maximum struct huft's malloc'ed */
969 /* initialize window, bit buffer */
975 /* decompress until the last block */
980 if ((r
= inflate_block(&e
)) != 0) {
989 /* Undo too much lookahead. The next read will be byte aligned so we
990 * can discard unused bits in the last meaningful byte.
997 /* flush out slide */
1001 /* return success */
1003 fprintf(stderr
, "<%u> ", h
);
1008 /**********************************************************************
1010 * The following are support routines for inflate.c
1012 **********************************************************************/
1014 static ulg crc_32_tab
[256];
1015 static ulg crc
; /* initialized in makecrc() so it'll reside in bss */
1016 #define CRC_VALUE (crc ^ 0xffffffffUL)
1019 * Code to compute the CRC-32 table. Borrowed from
1020 * gzip-1.0.3/makecrc.c.
1026 /* Not copyrighted 1990 Mark Adler */
1028 unsigned long c
; /* crc shift register */
1029 unsigned long e
; /* polynomial exclusive-or pattern */
1030 int i
; /* counter for all possible eight bit values */
1031 int k
; /* byte being shifted into crc apparatus */
1033 /* terms of polynomial defining this crc (except x^32): */
1034 static const int p
[] = {0,1,2,4,5,7,8,10,11,12,16,22,23,26};
1036 /* Make exclusive-or pattern from polynomial */
1038 for (i
= 0; i
< sizeof(p
)/sizeof(int); i
++)
1039 e
|= 1L << (31 - p
[i
]);
1043 for (i
= 1; i
< 256; i
++)
1046 for (k
= i
| 256; k
!= 1; k
>>= 1)
1048 c
= c
& 1 ? (c
>> 1) ^ e
: c
>> 1;
1055 /* this is initialized here so this code could reside in ROM */
1056 crc
= (ulg
)0xffffffffUL
; /* shift register contents */
1059 /* gzip flag byte */
1060 #define ASCII_FLAG 0x01 /* bit 0 set: file probably ASCII text */
1061 #define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */
1062 #define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
1063 #define ORIG_NAME 0x08 /* bit 3 set: original file name present */
1064 #define COMMENT 0x10 /* bit 4 set: file comment present */
1065 #define ENCRYPTED 0x20 /* bit 5 set: file is encrypted */
1066 #define RESERVED 0xC0 /* bit 6,7: reserved */
1069 * Do the uncompression!
1071 static int gunzip(void)
1074 unsigned char magic
[2]; /* magic header */
1076 ulg orig_crc
= 0; /* original crc */
1077 ulg orig_len
= 0; /* original uncompressed length */
1080 magic
[0] = (unsigned char)get_byte();
1081 magic
[1] = (unsigned char)get_byte();
1082 method
= (unsigned char)get_byte();
1084 if (magic
[0] != 037 ||
1085 ((magic
[1] != 0213) && (magic
[1] != 0236))) {
1086 error("bad gzip magic numbers");
1090 /* We only support method #8, DEFLATED */
1092 error("internal error, invalid method");
1096 flags
= (uch
)get_byte();
1097 if ((flags
& ENCRYPTED
) != 0) {
1098 error("Input is encrypted\n");
1101 if ((flags
& CONTINUATION
) != 0) {
1102 error("Multi part input\n");
1105 if ((flags
& RESERVED
) != 0) {
1106 error("Input has invalid flags\n");
1109 (ulg
)get_byte(); /* Get timestamp */
1110 ((ulg
)get_byte()) << 8;
1111 ((ulg
)get_byte()) << 16;
1112 ((ulg
)get_byte()) << 24;
1114 (void)get_byte(); /* Ignore extra flags for the moment */
1115 (void)get_byte(); /* Ignore OS type for the moment */
1117 if ((flags
& EXTRA_FIELD
) != 0) {
1118 unsigned len
= (unsigned)get_byte();
1119 len
|= ((unsigned)get_byte())<<8;
1120 while (len
--) (void)get_byte();
1123 /* Get original file name if it was truncated */
1124 if ((flags
& ORIG_NAME
) != 0) {
1125 /* Discard the old name */
1126 while (get_byte() != 0) /* null */ ;
1129 /* Discard file comment if any */
1130 if ((flags
& COMMENT
) != 0) {
1131 while (get_byte() != 0) /* null */ ;
1135 if ((res
= inflate())) {
1140 error("invalid compressed format (err=1)");
1143 error("invalid compressed format (err=2)");
1146 error("out of memory");
1149 error("invalid compressed format (other)");
1154 /* Get the crc and original length */
1155 /* crc32 (see algorithm.doc)
1156 * uncompressed input size modulo 2^32
1158 orig_crc
= (ulg
) get_byte();
1159 orig_crc
|= (ulg
) get_byte() << 8;
1160 orig_crc
|= (ulg
) get_byte() << 16;
1161 orig_crc
|= (ulg
) get_byte() << 24;
1163 orig_len
= (ulg
) get_byte();
1164 orig_len
|= (ulg
) get_byte() << 8;
1165 orig_len
|= (ulg
) get_byte() << 16;
1166 orig_len
|= (ulg
) get_byte() << 24;
1168 /* Validate decompression */
1169 if (orig_crc
!= CRC_VALUE
) {
1173 if (orig_len
!= bytes_out
) {
1174 error("length error");