1 /* Inflate deflated data
2
3 Copyright (C) 1997-1999, 2002, 2006, 2009-2023 Free Software Foundation,
4 Inc.
5
6 This program is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 3, or (at your option)
9 any later version.
10
11 This program is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
15
16 You should have received a copy of the GNU General Public License
17 along with this program; if not, write to the Free Software Foundation,
18 Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */
19
20 /* Not copyrighted 1992 by Mark Adler
21 version c10p1, 10 January 1993 */
22
23 /* You can do whatever you like with this source file, though I would
24 prefer that if you modify it and redistribute it that you include
25 comments to that effect with your name and the date. Thank you.
26 [The history has been moved to the file ChangeLog.]
27 */
28
29 /*
30 Inflate deflated (PKZIP's method 8 compressed) data. The compression
31 method searches for as much of the current string of bytes (up to a
32 length of 258) in the previous 32K bytes. If it doesn't find any
33 matches (of at least length 3), it codes the next byte. Otherwise, it
34 codes the length of the matched string and its distance backwards from
35 the current position. There is a single Huffman code that codes both
36 single bytes (called "literals") and match lengths. A second Huffman
37 code codes the distance information, which follows a length code. Each
38 length or distance code actually represents a base value and a number
39 of "extra" (sometimes zero) bits to get to add to the base value. At
40 the end of each deflated block is a special end-of-block (EOB) literal/
41 length code. The decoding process is basically: get a literal/length
42 code; if EOB then done; if a literal, emit the decoded byte; if a
43 length then get the distance and emit the referred-to bytes from the
44 sliding window of previously emitted data.
45
46 There are (currently) three kinds of inflate blocks: stored, fixed, and
47 dynamic. The compressor deals with some chunk of data at a time, and
48 decides which method to use on a chunk-by-chunk basis. A chunk might
49 typically be 32K or 64K. If the chunk is uncompressible, then the
50 "stored" method is used. In this case, the bytes are simply stored as
51 is, eight bits per byte, with none of the above coding. The bytes are
52 preceded by a count, since there is no longer an EOB code.
53
54 If the data is compressible, then either the fixed or dynamic methods
55 are used. In the dynamic method, the compressed data is preceded by
56 an encoding of the literal/length and distance Huffman codes that are
57 to be used to decode this block. The representation is itself Huffman
58 coded, and so is preceded by a description of that code. These code
59 descriptions take up a little space, and so for small blocks, there is
60 a predefined set of codes, called the fixed codes. The fixed method is
61 used if the block codes up smaller that way (usually for quite small
62 chunks), otherwise the dynamic method is used. In the latter case, the
63 codes are customized to the probabilities in the current block, and so
64 can code it much better than the pre-determined fixed codes.
65
66 The Huffman codes themselves are decoded using a multi-level table
67 lookup, in order to maximize the speed of decoding plus the speed of
68 building the decoding tables. See the comments below that precede the
69 lbits and dbits tuning parameters.
70 */
71
72
73 /*
74 Notes beyond the 1.93a appnote.txt:
75
76 1. Distance pointers never point before the beginning of the output
77 stream.
78 2. Distance pointers can point back across blocks, up to 32k away.
79 3. There is an implied maximum of 7 bits for the bit length table and
80 15 bits for the actual data.
81 4. If only one code exists, then it is encoded using one bit. (Zero
82 would be more efficient, but perhaps a little confusing.) If two
83 codes exist, they are coded using one bit each (0 and 1).
84 5. There is no way of sending zero distance codes--a dummy must be
85 sent if there are none. (History: a pre 2.0 version of PKZIP would
86 store blocks with no distance codes, but this was discovered to be
87 too harsh a criterion.) Valid only for 1.93a. 2.04c does allow
88 zero distance codes, which is sent as one code of zero bits in
89 length.
90 6. There are up to 286 literal/length codes. Code 256 represents the
91 end-of-block. Note however that the static length tree defines
92 288 codes just to fill out the Huffman codes. Codes 286 and 287
93 cannot be used though, since there is no length base or extra bits
94 defined for them. Similarly, there are up to 30 distance codes.
95 However, static trees define 32 codes (all 5 bits) to fill out the
96 Huffman codes, but the last two had better not show up in the data.
97 7. Unzip can check dynamic Huffman blocks for complete code sets.
98 The exception is that a single code would not be complete (see #4).
99 8. The five bits following the block type is really the number of
100 literal codes sent minus 257.
101 9. Length codes 8,16,16 are interpreted as 13 length codes of 8 bits
102 (1+6+6). Therefore, to output three times the length, you output
103 three codes (1+1+1), whereas to output four times the same length,
104 you only need two codes (1+3). Hmm.
105 10. In the tree reconstruction algorithm, Code = Code + Increment
106 only if BitLength(i) is not zero. (Pretty obvious.)
107 11. Correction: 4 Bits: # of Bit Length codes - 4 (4 - 19)
108 12. Note: length code 284 can represent 227-258, but length code 285
109 really is 258. The last length deserves its own, short code
110 since it gets used a lot in very redundant files. The length
111 258 is special since 258 - 3 (the min match length) is 255.
112 13. The literal/length and distance code bit lengths are read as a
113 single stream of lengths. It is possible (and advantageous) for
114 a repeat code (16, 17, or 18) to go across the boundary between
115 the two sets of lengths.
116 */
117
118 #include <config.h>
119
120 #include <stdlib.h>
121
122 #include "tailor.h"
123 #include "gzip.h"
124 #define slide window
125
126 /* Huffman code lookup table entry--this entry is four bytes for machines
127 that have 16-bit pointers (e.g. PC's in the small or medium model).
128 Valid extra bits are 0..13. e == 15 is EOB (end of block), e == 16
129 means that v is a literal, 16 < e < 32 means that v is a pointer to
130 the next table, which codes e - 16 bits, and lastly e == 99 indicates
131 an unused code. If a code with e == 99 is looked up, this implies an
132 error in the data. */
133 struct huft {
134 uch e; /* number of extra bits or operation */
135 uch b; /* number of bits in this code or subcode */
136 union {
137 ush n; /* literal, length base, or distance base */
138 struct huft *t; /* pointer to next level of table */
139 } v;
140 };
141
142
143 /* Function prototypes */
144 static int huft_free (struct huft *);
145
146
147 /* The inflate algorithm uses a sliding 32K byte window on the uncompressed
148 stream to find repeated byte strings. This is implemented here as a
149 circular buffer. The index is updated simply by incrementing and then
150 and'ing with 0x7fff (32K-1). */
151 /* It is left to other modules to supply the 32K area. It is assumed
152 to be usable as if it were declared "uch slide[32768];" or as just
153 "uch *slide;" and then malloc'ed in the latter case. The definition
154 must be in unzip.h, included above. */
155 /* unsigned wp; current position in slide */
156 static bool fresh;
157 #define wp outcnt
158 #define flush_output(w) (fresh = false, wp = (w), flush_window ())
159
160 /* Tables for deflate from PKZIP's appnote.txt. */
161 static unsigned border[] = { /* Order of the bit length code lengths */
162 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
163 static ush cplens[] = { /* Copy lengths for literal codes 257..285 */
164 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
165 35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
166 /* note: see note #13 above about the 258 in this list. */
167 static ush cplext[] = { /* Extra bits for literal codes 257..285 */
168 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2,
169 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, 99, 99}; /* 99==invalid */
170 static ush cpdist[] = { /* Copy offsets for distance codes 0..29 */
171 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
172 257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
173 8193, 12289, 16385, 24577};
174 static ush cpdext[] = { /* Extra bits for distance codes */
175 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6,
176 7, 7, 8, 8, 9, 9, 10, 10, 11, 11,
177 12, 12, 13, 13};
178
179
180
181 /* Macros for inflate() bit peeking and grabbing.
182 The usage is:
183
184 NEEDBITS(j)
185 x = b & mask_bits[j];
186 DUMPBITS(j)
187
188 where NEEDBITS makes sure that b has at least j bits in it, and
189 DUMPBITS removes the bits from b. The macros use the variable k
190 for the number of bits in b. Normally, b and k are register
191 variables for speed, and are initialized at the beginning of a
192 routine that uses these macros from a global bit buffer and count.
193 The macros also use the variable w, which is a cached copy of wp.
194
195 If we assume that EOB will be the longest code, then we will never
196 ask for bits with NEEDBITS that are beyond the end of the stream.
197 So, NEEDBITS should not read any more bytes than are needed to
198 meet the request. Then no bytes need to be "returned" to the buffer
199 at the end of the last block.
200
201 However, this assumption is not true for fixed blocks--the EOB code
202 is 7 bits, but the other literal/length codes can be 8 or 9 bits.
203 (The EOB code is shorter than other codes because fixed blocks are
204 generally short. So, while a block always has an EOB, many other
205 literal/length codes have a significantly lower probability of
206 showing up at all.) However, by making the first table have a
207 lookup of seven bits, the EOB code will be found in that first
208 lookup, and so will not require that too many bits be pulled from
209 the stream.
210 */
211
212 static ulg bb; /* bit buffer */
213 static unsigned bk; /* bits in bit buffer */
214
215 static ush mask_bits[] = {
216 0x0000,
217 0x0001, 0x0003, 0x0007, 0x000f, 0x001f, 0x003f, 0x007f, 0x00ff,
218 0x01ff, 0x03ff, 0x07ff, 0x0fff, 0x1fff, 0x3fff, 0x7fff, 0xffff
219 };
220
221 #define GETBYTE() (inptr < insize ? inbuf[inptr++] : (wp = w, fill_inbuf(0)))
222
223 #define NEXTBYTE() (uch)GETBYTE()
224 #define NEEDBITS(n) {while(k<(n)){b|=((ulg)NEXTBYTE())<<k;k+=8;}}
225 #define DUMPBITS(n) {b>>=(n);k-=(n);}
226
227
228 /*
229 Huffman code decoding is performed using a multi-level table lookup.
230 The fastest way to decode is to simply build a lookup table whose
231 size is determined by the longest code. However, the time it takes
232 to build this table can also be a factor if the data being decoded
233 is not very long. The most common codes are necessarily the
234 shortest codes, so those codes dominate the decoding time, and hence
235 the speed. The idea is you can have a shorter table that decodes the
236 shorter, more probable codes, and then point to subsidiary tables for
237 the longer codes. The time it costs to decode the longer codes is
238 then traded against the time it takes to make longer tables.
239
240 This results of this trade are in the variables lbits and dbits
241 below. lbits is the number of bits the first level table for literal/
242 length codes can decode in one step, and dbits is the same thing for
243 the distance codes. Subsequent tables are also less than or equal to
244 those sizes. These values may be adjusted either when all of the
245 codes are shorter than that, in which case the longest code length in
246 bits is used, or when the shortest code is *longer* than the requested
247 table size, in which case the length of the shortest code in bits is
248 used.
249
250 There are two different values for the two tables, since they code a
251 different number of possibilities each. The literal/length table
252 codes 286 possible values, or in a flat code, a little over eight
253 bits. The distance table codes 30 possible values, or a little less
254 than five bits, flat. The optimum values for speed end up being
255 about one bit more than those, so lbits is 8+1 and dbits is 5+1.
256 The optimum values may differ though from machine to machine, and
257 possibly even between compilers. Your mileage may vary.
258 */
259
260
261 static int lbits = 9; /* bits in base literal/length lookup table */
262 static int dbits = 6; /* bits in base distance lookup table */
263
264
265 /* If BMAX needs to be larger than 16, then h and x[] should be ulg. */
266 #define BMAX 16 /* maximum bit length of any code (16 for explode) */
267 #define N_MAX 288 /* maximum number of codes in any set */
268
269
270 static unsigned hufts; /* track memory usage */
271
272
273 static int
274 huft_build(
275 unsigned *b, /* code lengths in bits (all assumed <= BMAX) */
276 unsigned n, /* number of codes (assumed <= N_MAX) */
277 unsigned s, /* number of simple-valued codes (0..s-1) */
278 ush *d, /* list of base values for non-simple codes */
279 ush *e, /* list of extra bits for non-simple codes */
280 struct huft **t, /* result: starting table */
281 int *m /* maximum lookup bits, returns actual */
282 )
283 /* Given a list of code lengths and a maximum table size, make a set of
284 tables to decode that set of codes. Return zero on success, one if
285 the given code set is incomplete (the tables are still built in this
286 case), two if the input is invalid (all zero length codes or an
287 oversubscribed set of lengths), and three if not enough memory. */
288 {
289 unsigned a; /* counter for codes of length k */
290 unsigned c[BMAX+1]; /* bit length count table */
291 unsigned f; /* i repeats in table every f entries */
292 int g; /* maximum code length */
293 int h; /* table level */
294 register unsigned i; /* counter, current code */
295 register unsigned j; /* counter */
296 register int k; /* number of bits in current code */
297 int l; /* bits per table (returned in m) */
298 register unsigned *p; /* pointer into c[], b[], or v[] */
299 register struct huft *q; /* points to current table */
300 struct huft r; /* table entry for structure assignment */
301 struct huft *u[BMAX]; /* table stack */
302 unsigned v[N_MAX]; /* values in order of bit length */
303 register int w; /* bits before this table == (l * h) */
304 unsigned x[BMAX+1]; /* bit offsets, then code stack */
305 unsigned *xp; /* pointer into x */
306 int y; /* number of dummy codes added */
307 unsigned z; /* number of entries in current table */
308
309
310 /* Generate counts for each bit length */
311 memzero(c, sizeof(c));
312 p = b; i = n;
313 do {
314 #ifdef DEBUG
315 if (1 < verbose && *p)
316 {
317 if (' ' <= n - i && n - i <= '~')
318 {
319 char ch = n - i;
320 fprintf (stderr, "%c %u\n", ch, *p);
321 }
322 else
323 fprintf (stderr, "0x%x %u\n", n - i, *p);
324 }
325 #endif
326 c[*p]++; /* assume all entries <= BMAX */
327 p++; /* Can't combine with above line (Solaris bug) */
328 } while (--i);
329 if (c[0] == n) /* null input--all zero length codes */
330 {
331 q = (struct huft *) malloc (3 * sizeof *q);
332 if (!q)
333 return 3;
334 hufts += 3;
335 q[0].v.t = (struct huft *) NULL;
336 q[1].e = 99; /* invalid code marker */
337 q[1].b = 1;
338 q[2].e = 99; /* invalid code marker */
339 q[2].b = 1;
340 *t = q + 1;
341 *m = 1;
342 return 0;
343 }
344
345
346 /* Find minimum and maximum length, bound *m by those */
347 l = *m;
348 for (j = 1; j <= BMAX; j++)
349 if (c[j])
350 break;
351 k = j; /* minimum code length */
352 if ((unsigned)l < j)
353 l = j;
354 for (i = BMAX; i; i--)
355 if (c[i])
356 break;
357 g = i; /* maximum code length */
358 if ((unsigned)l > i)
359 l = i;
360 *m = l;
361
362
363 /* Adjust last length count to fill out codes, if needed */
364 for (y = 1 << j; j < i; j++, y <<= 1)
365 if ((y -= c[j]) < 0)
366 return 2; /* bad input: more codes than bits */
367 if ((y -= c[i]) < 0)
368 return 2;
369 c[i] += y;
370
371
372 /* Generate starting offsets into the value table for each length */
373 x[1] = j = 0;
374 p = c + 1; xp = x + 2;
375 while (--i) { /* note that i == g from above */
376 *xp++ = (j += *p++);
377 }
378
379
380 /* Make a table of values in order of bit lengths */
381 p = b; i = 0;
382 do {
383 if ((j = *p++) != 0)
384 v[x[j]++] = i;
385 } while (++i < n);
386 n = x[g]; /* set n to length of v */
387
388
389 /* Generate the Huffman codes and for each, make the table entries */
390 x[0] = i = 0; /* first Huffman code is zero */
391 p = v; /* grab values in bit order */
392 h = -1; /* no tables yet--level -1 */
393 w = -l; /* bits decoded == (l * h) */
394 u[0] = (struct huft *)NULL; /* just to keep compilers happy */
395 q = (struct huft *)NULL; /* ditto */
396 z = 0; /* ditto */
397
398 /* go through the bit lengths (k already is bits in shortest code) */
399 for (; k <= g; k++)
400 {
401 a = c[k];
402 while (a--)
403 {
404 /* here i is the Huffman code of length k bits for value *p */
405 /* make tables up to required level */
406 while (k > w + l)
407 {
408 h++;
409 w += l; /* previous table always l bits */
410
411 /* compute minimum size table less than or equal to l bits */
412 z = (z = g - w) > (unsigned)l ? l : z; /* upper limit on table size */
413 if ((f = 1 << (j = k - w)) > a + 1) /* try a k-w bit table */
414 { /* too few codes for k-w bit table */
415 f -= a + 1; /* deduct codes from patterns left */
416 xp = c + k;
417 if (j < z)
418 while (++j < z) /* try smaller tables up to z bits */
419 {
420 if ((f <<= 1) <= *++xp)
421 break; /* enough codes to use up j bits */
422 f -= *xp; /* else deduct codes from patterns */
423 }
424 }
425 z = 1 << j; /* table entries for j-bit table */
426
427 /* allocate and link in new table */
428 if ((q = (struct huft *)malloc((z + 1)*sizeof(struct huft))) ==
429 (struct huft *)NULL)
430 {
431 if (h)
432 huft_free(u[0]);
433 return 3; /* not enough memory */
434 }
435 hufts += z + 1; /* track memory usage */
436 *t = q + 1; /* link to list for huft_free() */
437 *(t = &(q->v.t)) = (struct huft *)NULL;
438 u[h] = ++q; /* table starts after link */
439
440 /* connect to last table, if there is one */
441 if (h)
442 {
443 x[h] = i; /* save pattern for backing up */
444 r.b = (uch)l; /* bits to dump before this table */
445 r.e = (uch)(16 + j); /* bits in this table */
446 r.v.t = q; /* pointer to this table */
447 j = i >> (w - l); /* (get around Turbo C bug) */
448 u[h-1][j] = r; /* connect to last table */
449 }
450 }
451
452 /* set up table entry in r */
453 r.b = (uch)(k - w);
454 if (p >= v + n)
455 r.e = 99; /* out of values--invalid code */
456 else if (*p < s)
457 {
458 r.e = (uch)(*p < 256 ? 16 : 15); /* 256 is end-of-block code */
459 r.v.n = (ush)(*p); /* simple code is just the value */
460 p++; /* one compiler does not like *p++ */
461 }
462 else
463 {
464 r.e = (uch)e[*p - s]; /* non-simple--look up in lists */
465 r.v.n = d[*p++ - s];
466 }
467
468 /* fill code-like entries with r */
469 f = 1 << (k - w);
470 for (j = i >> w; j < z; j += f)
471 q[j] = r;
472
473 /* backwards increment the k-bit code i */
474 for (j = 1 << (k - 1); i & j; j >>= 1)
475 i ^= j;
476 i ^= j;
477
478 /* backup over finished tables */
479 while ((i & ((1 << w) - 1)) != x[h])
480 {
481 h--; /* don't need to update q */
482 w -= l;
483 }
484 }
485 }
486
487
488 /* Return true (1) if we were given an incomplete table */
489 return y != 0 && g != 1;
490 }
491
492
493
494 /* Free the malloc'ed tables T built by huft_build(), which makes a linked
495 list of the tables it made, with the links in a dummy first entry of
496 each table. */
497 static int
498 huft_free(struct huft *t)
499 {
500 register struct huft *p, *q;
501
502
503 /* Go through linked list, freeing from the malloced (t[-1]) address. */
504 p = t;
505 while (p != (struct huft *)NULL)
506 {
507 q = (--p)->v.t;
508 free(p);
509 p = q;
510 }
511 return 0;
512 }
513
514
515 /* tl, td: literal/length and distance decoder tables */
516 /* bl, bd: number of bits decoded by tl[] and td[] */
517 /* inflate (decompress) the codes in a deflated (compressed) block.
518 Return an error code or zero if it all goes ok. */
519 static int
520 inflate_codes(struct huft *tl, struct huft *td, int bl, int bd)
521 {
522 register unsigned e; /* table entry flag/number of extra bits */
523 unsigned n, d; /* length and index for copy */
524 unsigned w; /* current window position */
525 struct huft *t; /* pointer to table entry */
526 unsigned ml, md; /* masks for bl and bd bits */
527 register ulg b; /* bit buffer */
528 register unsigned k; /* number of bits in bit buffer */
529
530
531 /* make local copies of globals */
532 b = bb; /* initialize bit buffer */
533 k = bk;
534 w = wp; /* initialize window position */
535
536 /* inflate the coded data */
537 ml = mask_bits[bl]; /* precompute masks for speed */
538 md = mask_bits[bd];
539 for (;;) /* do until end of block */
540 {
541 NEEDBITS((unsigned)bl)
542 if ((e = (t = tl + ((unsigned)b & ml))->e) > 16)
543 do {
544 if (e == 99)
545 return 1;
546 DUMPBITS(t->b)
547 e -= 16;
548 NEEDBITS(e)
549 } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16);
550 DUMPBITS(t->b)
551 if (e == 16) /* then it's a literal */
552 {
553 slide[w++] = (uch)t->v.n;
554 Tracevv((stderr, "%c", slide[w-1]));
555 if (w == WSIZE)
556 {
557 flush_output(w);
558 w = 0;
559 }
560 }
561 else /* it's an EOB or a length */
562 {
563 /* exit if end of block */
564 if (e == 15)
565 break;
566
567 /* get length of block to copy */
568 NEEDBITS(e)
569 n = t->v.n + ((unsigned)b & mask_bits[e]);
570 DUMPBITS(e);
571
572 /* decode distance of block to copy */
573 NEEDBITS((unsigned)bd)
574 if ((e = (t = td + ((unsigned)b & md))->e) > 16)
575 do {
576 if (e == 99)
577 return 1;
578 DUMPBITS(t->b)
579 e -= 16;
580 NEEDBITS(e)
581 } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16);
582 DUMPBITS(t->b)
583 NEEDBITS(e)
584 d = w - t->v.n - ((unsigned)b & mask_bits[e]);
585 DUMPBITS(e)
586 if (fresh && w <= d)
587 return 1;
588 Tracevv ((stderr, "\\[%u,%u]", w - d, n));
589
590 /* do the copy */
591 do {
592 n -= (e = (e = WSIZE - ((d &= WSIZE-1) > w ? d : w)) > n ? n : e);
593 #ifndef DEBUG
594 if (e <= (d < w ? w - d : d - w))
595 {
596 memcpy(slide + w, slide + d, e);
597 w += e;
598 d += e;
599 }
600 else /* do it slow to avoid memcpy() overlap */
601 #endif
602 do {
603 slide[w++] = slide[d++];
604 Tracevv((stderr, "%c", slide[w-1]));
605 } while (--e);
606 if (w == WSIZE)
607 {
608 flush_output(w);
609 w = 0;
610 }
611 } while (n);
612 }
613 }
614
615
616 /* restore the globals from the locals */
617 wp = w; /* restore global window pointer */
618 bb = b; /* restore global bit buffer */
619 bk = k;
620
621 /* done */
622 return 0;
623 }
624
625
626
627 /* "decompress" an inflated type 0 (stored) block. */
628 static int
629 inflate_stored(void)
630 {
631 unsigned n; /* number of bytes in block */
632 unsigned w; /* current window position */
633 register ulg b; /* bit buffer */
634 register unsigned k; /* number of bits in bit buffer */
635
636
637 /* make local copies of globals */
638 b = bb; /* initialize bit buffer */
639 k = bk;
640 w = wp; /* initialize window position */
641
642
643 /* go to byte boundary */
644 n = k & 7;
645 DUMPBITS(n);
646
647
648 /* get the length and its complement */
649 NEEDBITS(16)
650 n = ((unsigned)b & 0xffff);
651 DUMPBITS(16)
652 NEEDBITS(16)
653 if (n != (unsigned)((~b) & 0xffff))
654 return 1; /* error in compressed data */
655 DUMPBITS(16)
656
657
658 /* read and output the compressed data */
659 while (n--)
660 {
661 NEEDBITS(8)
662 slide[w++] = (uch)b;
663 if (w == WSIZE)
664 {
665 flush_output(w);
666 w = 0;
667 }
668 DUMPBITS(8)
669 }
670
671
672 /* restore the globals from the locals */
673 wp = w; /* restore global window pointer */
674 bb = b; /* restore global bit buffer */
675 bk = k;
676 return 0;
677 }
678
679
680
681 /* decompress an inflated type 1 (fixed Huffman codes) block. We should
682 either replace this with a custom decoder, or at least precompute the
683 Huffman tables. */
684 static int
685 inflate_fixed(void)
686 {
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 */
693
694
695 /* set up literal table */
696 for (i = 0; i < 144; i++)
697 l[i] = 8;
698 for (; i < 256; i++)
699 l[i] = 9;
700 for (; i < 280; i++)
701 l[i] = 7;
702 for (; i < 288; i++) /* make a complete, but wrong code set */
703 l[i] = 8;
704 bl = 7;
705 if ((i = huft_build(l, 288, 257, cplens, cplext, &tl, &bl)) != 0)
706 return i;
707
708
709 /* set up distance table */
710 for (i = 0; i < 30; i++) /* make an incomplete code set */
711 l[i] = 5;
712 bd = 5;
713 if ((i = huft_build(l, 30, 0, cpdist, cpdext, &td, &bd)) > 1)
714 {
715 huft_free(tl);
716 return i;
717 }
718
719
720 /* decompress until an end-of-block code */
721 if (inflate_codes(tl, td, bl, bd))
722 return 1;
723
724
725 /* free the decoding tables, return */
726 huft_free(tl);
727 huft_free(td);
728 return 0;
729 }
730
731
732
733 /* decompress an inflated type 2 (dynamic Huffman codes) block. */
734 static int
735 inflate_dynamic(void)
736 {
737 int i; /* temporary variables */
738 unsigned j;
739 unsigned l; /* last length */
740 unsigned m; /* mask for bit lengths table */
741 unsigned n; /* number of lengths to get */
742 unsigned w; /* current window position */
743 struct huft *tl; /* literal/length code table */
744 struct huft *td; /* distance code table */
745 int bl; /* lookup bits for tl */
746 int bd; /* lookup bits for td */
747 unsigned nb; /* number of bit length codes */
748 unsigned nl; /* number of literal/length codes */
749 unsigned nd; /* number of distance codes */
750 #ifdef PKZIP_BUG_WORKAROUND
751 unsigned ll[288+32]; /* literal/length and distance code lengths */
752 #else
753 unsigned ll[286+30]; /* literal/length and distance code lengths */
754 #endif
755 register ulg b; /* bit buffer */
756 register unsigned k; /* number of bits in bit buffer */
757
758
759 /* make local bit buffer */
760 b = bb;
761 k = bk;
762 w = wp;
763
764
765 /* read in table lengths */
766 NEEDBITS(5)
767 nl = 257 + ((unsigned)b & 0x1f); /* number of literal/length codes */
768 DUMPBITS(5)
769 NEEDBITS(5)
770 nd = 1 + ((unsigned)b & 0x1f); /* number of distance codes */
771 DUMPBITS(5)
772 NEEDBITS(4)
773 nb = 4 + ((unsigned)b & 0xf); /* number of bit length codes */
774 DUMPBITS(4)
775 #ifdef PKZIP_BUG_WORKAROUND
776 if (nl > 288 || nd > 32)
777 #else
778 if (nl > 286 || nd > 30)
779 #endif
780 return 1; /* bad lengths */
781
782
783 /* read in bit-length-code lengths */
784 for (j = 0; j < nb; j++)
785 {
786 NEEDBITS(3)
787 ll[border[j]] = (unsigned)b & 7;
788 DUMPBITS(3)
789 }
790 for (; j < 19; j++)
791 ll[border[j]] = 0;
792
793
794 /* build decoding table for trees--single level, 7 bit lookup */
795 bl = 7;
796 if ((i = huft_build(ll, 19, 19, NULL, NULL, &tl, &bl)) != 0)
797 {
798 if (i == 1)
799 huft_free(tl);
800 return i; /* incomplete code set */
801 }
802
803 if (tl == NULL) /* Grrrhhh */
804 return 2;
805
806 /* read in literal and distance code lengths */
807 n = nl + nd;
808 m = mask_bits[bl];
809 i = l = 0;
810 while ((unsigned)i < n)
811 {
812 NEEDBITS((unsigned)bl)
813 j = (td = tl + ((unsigned)b & m))->b;
814 DUMPBITS(j)
815 if (td->e == 99)
816 {
817 /* Invalid code. */
818 huft_free (tl);
819 return 2;
820 }
821 j = td->v.n;
822 if (j < 16) /* length of code in bits (0..15) */
823 ll[i++] = l = j; /* save last length in l */
824 else if (j == 16) /* repeat last length 3 to 6 times */
825 {
826 NEEDBITS(2)
827 j = 3 + ((unsigned)b & 3);
828 DUMPBITS(2)
829 if ((unsigned)i + j > n)
830 return 1;
831 while (j--)
832 ll[i++] = l;
833 }
834 else if (j == 17) /* 3 to 10 zero length codes */
835 {
836 NEEDBITS(3)
837 j = 3 + ((unsigned)b & 7);
838 DUMPBITS(3)
839 if ((unsigned)i + j > n)
840 return 1;
841 while (j--)
842 ll[i++] = 0;
843 l = 0;
844 }
845 else /* j == 18: 11 to 138 zero length codes */
846 {
847 NEEDBITS(7)
848 j = 11 + ((unsigned)b & 0x7f);
849 DUMPBITS(7)
850 if ((unsigned)i + j > n)
851 return 1;
852 while (j--)
853 ll[i++] = 0;
854 l = 0;
855 }
856 }
857
858
859 /* free decoding table for trees */
860 huft_free(tl);
861
862
863 /* restore the global bit buffer */
864 bb = b;
865 bk = k;
866
867
868 /* build the decoding tables for literal/length and distance codes */
869 bl = lbits;
870 if ((i = huft_build(ll, nl, 257, cplens, cplext, &tl, &bl)) != 0)
871 {
872 if (i == 1) {
873 Trace ((stderr, " incomplete literal tree\n"));
874 huft_free(tl);
875 }
876 return i; /* incomplete code set */
877 }
878 bd = dbits;
879 if ((i = huft_build(ll + nl, nd, 0, cpdist, cpdext, &td, &bd)) != 0)
880 {
881 if (i == 1) {
882 Trace ((stderr, " incomplete distance tree\n"));
883 #ifdef PKZIP_BUG_WORKAROUND
884 i = 0;
885 }
886 #else
887 huft_free(td);
888 }
889 huft_free(tl);
890 return i; /* incomplete code set */
891 #endif
892 }
893
894
895 {
896 /* decompress until an end-of-block code */
897 int err = inflate_codes(tl, td, bl, bd) ? 1 : 0;
898
899 /* free the decoding tables */
900 huft_free(tl);
901 huft_free(td);
902
903 return err;
904 }
905 }
906
907
908
909 /* decompress an inflated block */
910 /* E is the last block flag */
911 static int inflate_block(int *e)
912 {
913 unsigned t; /* block type */
914 unsigned w; /* current window position */
915 register ulg b; /* bit buffer */
916 register unsigned k; /* number of bits in bit buffer */
917
918
919 /* make local bit buffer */
920 b = bb;
921 k = bk;
922 w = wp;
923
924
925 /* read in last block bit */
926 NEEDBITS(1)
927 *e = (int)b & 1;
928 DUMPBITS(1)
929
930
931 /* read in block type */
932 NEEDBITS(2)
933 t = (unsigned)b & 3;
934 DUMPBITS(2)
935
936
937 /* restore the global bit buffer */
938 bb = b;
939 bk = k;
940
941
942 /* inflate that block type */
943 if (t == 2)
944 return inflate_dynamic();
945 if (t == 0)
946 return inflate_stored();
947 if (t == 1)
948 return inflate_fixed();
949
950
951 /* bad block type */
952 return 2;
953 }
954
955
956
957 int
958 inflate ()
959 /* decompress an inflated entry */
960 {
961 int e; /* last block flag */
962 int r; /* result code */
963 unsigned h; /* maximum struct huft's malloc'ed */
964
965
966 /* initialize window, bit buffer */
967 wp = 0;
968 bk = 0;
969 bb = 0;
970 fresh = true;
971
972
973 /* decompress until the last block */
974 h = 0;
975 do {
976 hufts = 0;
977 if ((r = inflate_block(&e)) != 0)
978 return r;
979 if (hufts > h)
980 h = hufts;
981 } while (!e);
982
983 /* Undo too much lookahead. The next read will be byte aligned so we
984 * can discard unused bits in the last meaningful byte.
985 */
986 while (bk >= 8) {
987 bk -= 8;
988 inptr--;
989 }
990
991 /* flush out slide */
992 flush_output(wp);
993
994
995 /* return success */
996 Trace ((stderr, "<%u> ", h));
997 return 0;
998 }