![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to inflate.c CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [MW Coherent from dump] / coherent / b / bin / unzip|
Return to inflate.c CVS log | Up to [MW Coherent from dump] / coherent / b / bin / unzip |
1.1 root 1: /* inflate.c -- Not copyrighted 1992 by Mark Adler
2: version c10p1, 10 January 1993 */
3:
4:
5: /* You can do whatever you like with this source file, though I would
6: prefer that if you modify it and redistribute it that you include
7: comments to that effect with your name and the date. Thank you.
8:
9: History:
10: vers date who what
11: ---- --------- -------------- ------------------------------------
12: a ~~ Feb 92 M. Adler used full (large, one-step) lookup table
13: b1 21 Mar 92 M. Adler first version with partial lookup tables
14: b2 21 Mar 92 M. Adler fixed bug in fixed-code blocks
15: b3 22 Mar 92 M. Adler sped up match copies, cleaned up some
16: b4 25 Mar 92 M. Adler added prototypes; removed window[] (now
17: is the responsibility of unzip.h--also
18: changed name to slide[]), so needs diffs
19: for unzip.c and unzip.h (this allows
20: compiling in the small model on MSDOS);
21: fixed cast of q in huft_build();
22: b5 26 Mar 92 M. Adler got rid of unintended macro recursion.
23: b6 27 Mar 92 M. Adler got rid of nextbyte() routine. fixed
24: bug in inflate_fixed().
25: c1 30 Mar 92 M. Adler removed lbits, dbits environment variables.
26: changed BMAX to 16 for explode. Removed
27: OUTB usage, and replaced it with flush()--
28: this was a 20% speed improvement! Added
29: an explode.c (to replace unimplode.c) that
30: uses the huft routines here. Removed
31: register union.
32: c2 4 Apr 92 M. Adler fixed bug for file sizes a multiple of 32k.
33: c3 10 Apr 92 M. Adler reduced memory of code tables made by
34: huft_build significantly (factor of two to
35: three).
36: c4 15 Apr 92 M. Adler added NOMEMCPY do kill use of memcpy().
37: worked around a Turbo C optimization bug.
38: c5 21 Apr 92 M. Adler added the WSIZE #define to allow reducing
39: the 32K window size for specialized
40: applications.
41: c6 31 May 92 M. Adler added some typecasts to eliminate warnings
42: c7 27 Jun 92 G. Roelofs added some more typecasts (444: MSC bug).
43: c8 5 Oct 92 J-l. Gailly added ifdef'd code to deal with PKZIP bug.
44: c9 9 Oct 92 M. Adler removed a memory error message (~line 416).
45: c10 17 Oct 92 G. Roelofs changed ULONG/UWORD/byte to ulg/ush/uch,
46: removed old inflate, renamed inflate_entry
47: to inflate, added Mark's fix to a comment.
48: c10p1 10 Jan 93 G. Roelofs version c10 plus Mark's c13 patch:
49: [c13] M. Adler allow empty code sets in huft_build (the
50: new pkz204c.exe file has a null distance
51: tree for the file pkzip.exe)
52: */
53:
54:
55: /*
56: Inflate deflated (PKZIP's method 8 compressed) data. The compression
57: method searches for as much of the current string of bytes (up to a
58: length of 258) in the previous 32K bytes. If it doesn't find any
59: matches (of at least length 3), it codes the next byte. Otherwise, it
60: codes the length of the matched string and its distance backwards from
61: the current position. There is a single Huffman code that codes both
62: single bytes (called "literals") and match lengths. A second Huffman
63: code codes the distance information, which follows a length code. Each
64: length or distance code actually represents a base value and a number
65: of "extra" (sometimes zero) bits to get to add to the base value. At
66: the end of each deflated block is a special end-of-block (EOB) literal/
67: length code. The decoding process is basically: get a literal/length
68: code; if EOB then done; if a literal, emit the decoded byte; if a
69: length then get the distance and emit the referred-to bytes from the
70: sliding window of previously emitted data.
71:
72: There are (currently) three kinds of inflate blocks: stored, fixed, and
73: dynamic. The compressor outputs a chunk of data at a time, and decides
74: which method to use on a chunk-by-chunk basis. A chunk might typically
75: be 32K to 64K, uncompressed. If the chunk is uncompressible, then the
76: "stored" method is used. In this case, the bytes are simply stored as
77: is, eight bits per byte, with none of the above coding. The bytes are
78: preceded by a count, since there is no longer an EOB code.
79:
80: If the data is compressible, then either the fixed or dynamic methods
81: are used. In the dynamic method, the compressed data is preceded by
82: an encoding of the literal/length and distance Huffman codes that are
83: to be used to decode this block. The representation is itself Huffman
84: coded, and so is preceded by a description of that code. These code
85: descriptions take up a little space, and so for small blocks, there is
86: a predefined set of codes, called the fixed codes. The fixed method is
87: used if the block ends up smaller that way (usually for quite small
88: chunks), otherwise the dynamic method is used. In the latter case, the
89: codes are customized to the probabilities in the current block, and so
90: can code it much better than the pre-determined fixed codes can.
91:
92: The Huffman codes themselves are decoded using a mutli-level table
93: lookup, in order to maximize the speed of decoding plus the speed of
94: building the decoding tables. See the comments below that precede the
95: lbits and dbits tuning parameters.
96: */
97:
98:
99: /*
100: Notes beyond the 1.93a appnote.txt:
101:
102: 1. Distance pointers never point before the beginning of the output
103: stream.
104: 2. Distance pointers can point back across blocks, up to 32k away.
105: 3. There is an implied maximum of 7 bits for the bit length table and
106: 15 bits for the actual data.
107: 4. If only one code exists, then it is encoded using one bit. (Zero
108: would be more efficient, but perhaps a little confusing.) If two
109: codes exist, they are coded using one bit each (0 and 1).
110: 5. There is no way of sending zero distance codes--a dummy must be
111: sent if there are none. (History: a pre 2.0 version of PKZIP would
112: store blocks with no distance codes, but this was discovered to be
113: too harsh a criterion.) Valid only for 1.93a. 2.04c does allow
114: zero distance codes, which is sent as one code of zero bits in
115: length.
116: 6. There are up to 286 literal/length codes. Code 256 represents the
117: end-of-block. Note however that the static length tree defines
118: 288 codes just to fill out the Huffman codes. Codes 286 and 287
119: cannot be used though, since there is no length base or extra bits
120: defined for them. Similarily, there are up to 30 distance codes.
121: However, static trees define 32 codes (all 5 bits) to fill out the
122: Huffman codes, but the last two had better not show up in the data.
123: 7. Unzip can check dynamic Huffman blocks for complete code sets.
124: The exception is that a single code would not be complete (see #4).
125: 8. The five bits following the block type is really the number of
126: literal codes sent minus 257.
127: 9. Length codes 8,16,16 are interpreted as 13 length codes of 8 bits
128: (1+6+6). Therefore, to output three times the length, you output
129: three codes (1+1+1), whereas to output four times the same length,
130: you only need two codes (1+3). Hmm.
131: 10. In the tree reconstruction algorithm, Code = Code + Increment
132: only if BitLength(i) is not zero. (Pretty obvious.)
133: 11. Correction: 4 Bits: # of Bit Length codes - 4 (4 - 19)
134: 12. Note: length code 284 can represent 227-258, but length code 285
135: really is 258. The last length deserves its own, short code
136: since it gets used a lot in very redundant files. The length
137: 258 is special since 258 - 3 (the min match length) is 255.
138: 13. The literal/length and distance code bit lengths are read as a
139: single stream of lengths. It is possible (and advantageous) for
140: a repeat code (16, 17, or 18) to go across the boundary between
141: the two sets of lengths.
142: */
143:
144: #include "unzip.h" /* this must supply the slide[] (uch) array */
145:
146: #ifndef WSIZE
147: # define WSIZE 0x8000 /* window size--must be a power of two, and at least
148: 32K for zip's deflate method */
149: #endif /* !WSIZE */
150:
151: #ifdef DEBUG
152: # define Trace(x) fprintf x
153: #else
154: # define Trace(x)
155: #endif
156:
157:
158: /* Huffman code lookup table entry--this entry is four bytes for machines
159: that have 16-bit pointers (e.g. PC's in the small or medium model).
160: Valid extra bits are 0..13. e == 15 is EOB (end of block), e == 16
161: means that v is a literal, 16 < e < 32 means that v is a pointer to
162: the next table, which codes e - 16 bits, and lastly e == 99 indicates
163: an unused code. If a code with e == 99 is looked up, this implies an
164: error in the data. */
165: struct huft {
166: uch e; /* number of extra bits or operation */
167: uch b; /* number of bits in this code or subcode */
168: union {
169: ush n; /* literal, length base, or distance base */
170: struct huft *t; /* pointer to next level of table */
171: } v;
172: };
173:
174:
175: /* Function prototypes */
176: int huft_build OF((unsigned *, unsigned, unsigned, ush *, ush *,
177: struct huft **, int *));
178: int huft_free OF((struct huft *));
179: void flush OF((unsigned));
180: int inflate_codes OF((struct huft *, struct huft *, int, int));
181: int inflate_stored OF((void));
182: int inflate_fixed OF((void));
183: int inflate_dynamic OF((void));
184: int inflate_block OF((int *));
185: int inflate OF((void));
186:
187:
188: /* The inflate algorithm uses a sliding 32K byte window on the uncompressed
189: stream to find repeated byte strings. This is implemented here as a
190: circular buffer. The index is updated simply by incrementing and then
191: and'ing with 0x7fff (32K-1). */
192: /* It is left to other modules to supply the 32K area. It is assumed
193: to be usable as if it were declared "uch slide[32768];" or as just
194: "uch *slide;" and then malloc'ed in the latter case. The definition
195: must be in unzip.h, included above. */
196: unsigned wp; /* current position in slide */
197:
198:
199: /* Tables for deflate from PKZIP's appnote.txt. */
200: static unsigned border[] = { /* Order of the bit length code lengths */
201: 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
202: static ush cplens[] = { /* Copy lengths for literal codes 257..285 */
203: 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
204: 35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
205: /* note: see note #13 above about the 258 in this list. */
206: static ush cplext[] = { /* Extra bits for literal codes 257..285 */
207: 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2,
208: 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, 99, 99}; /* 99==invalid */
209: static ush cpdist[] = { /* Copy offsets for distance codes 0..29 */
210: 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
211: 257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
212: 8193, 12289, 16385, 24577};
213: static ush cpdext[] = { /* Extra bits for distance codes */
214: 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6,
215: 7, 7, 8, 8, 9, 9, 10, 10, 11, 11,
216: 12, 12, 13, 13};
217:
218:
219:
220: /* Macros for inflate() bit peeking and grabbing.
221: The usage is:
222:
223: NEEDBITS(j)
224: x = b & mask_bits[j];
225: DUMPBITS(j)
226:
227: where NEEDBITS makes sure that b has at least j bits in it, and
228: DUMPBITS removes the bits from b. The macros use the variable k
229: for the number of bits in b. Normally, b and k are register
230: variables for speed, and are initialized at the begining of a
231: routine that uses these macros from a global bit buffer and count.
232:
233: If we assume that EOB will be the longest code, then we will never
234: ask for bits with NEEDBITS that are beyond the end of the stream.
235: So, NEEDBITS should not read any more bytes than are needed to
236: meet the request. Then no bytes need to be "returned" to the buffer
237: at the end of the last block.
238:
239: However, this assumption is not true for fixed blocks--the EOB code
240: is 7 bits, but the other literal/length codes can be 8 or 9 bits.
241: (The EOB code is shorter than other codes becuase fixed blocks are
242: generally short. So, while a block always has an EOB, many other
243: literal/length codes have a significantly lower probability of
244: showing up at all.) However, by making the first table have a
245: lookup of seven bits, the EOB code will be found in that first
246: lookup, and so will not require that too many bits be pulled from
247: the stream.
248: */
249:
250: ulg bb; /* bit buffer */
251: unsigned bk; /* bits in bit buffer */
252:
253: ush bytebuf;
254: #define NEXTBYTE (ReadByte(&bytebuf), bytebuf)
255: #define NEEDBITS(n) {while(k<(n)){b|=((ulg)NEXTBYTE)<<k;k+=8;}}
256: #define DUMPBITS(n) {b>>=(n);k-=(n);}
257:
258:
259: /*
260: Huffman code decoding is performed using a multi-level table lookup.
261: The fastest way to decode is to simply build a lookup table whose
262: size is determined by the longest code. However, the time it takes
263: to build this table can also be a factor if the data being decoded
264: is not very long. The most common codes are necessarily the
265: shortest codes, so those codes dominate the decoding time, and hence
266: the speed. The idea is you can have a shorter table that decodes the
267: shorter, more probable codes, and then point to subsidiary tables for
268: the longer codes. The time it costs to decode the longer codes is
269: then traded against the time it takes to make longer tables.
270:
271: This results of this trade are in the variables lbits and dbits
272: below. lbits is the number of bits the first level table for literal/
273: length codes can decode in one step, and dbits is the same thing for
274: the distance codes. Subsequent tables are also less than or equal to
275: those sizes. These values may be adjusted either when all of the
276: codes are shorter than that, in which case the longest code length in
277: bits is used, or when the shortest code is *longer* than the requested
278: table size, in which case the length of the shortest code in bits is
279: used.
280:
281: There are two different values for the two tables, since they code a
282: different number of possibilities each. The literal/length table
283: codes 286 possible values, or in a flat code, a little over eight
284: bits. The distance table codes 30 possible values, or a little less
285: than five bits, flat. The optimum values for speed end up being
286: about one bit more than those, so lbits is 8+1 and dbits is 5+1.
287: The optimum values may differ though from machine to machine, and
288: possibly even between compilers. Your mileage may vary.
289: */
290:
291:
292: int lbits = 9; /* bits in base literal/length lookup table */
293: int dbits = 6; /* bits in base distance lookup table */
294:
295:
296: /* If BMAX needs to be larger than 16, then h and x[] should be ulg. */
297: #define BMAX 16 /* maximum bit length of any code (16 for explode) */
298: #define N_MAX 288 /* maximum number of codes in any set */
299:
300:
301: unsigned hufts; /* track memory usage */
302:
303:
304: int huft_build(b, n, s, d, e, t, m)
305: unsigned *b; /* code lengths in bits (all assumed <= BMAX) */
306: unsigned n; /* number of codes (assumed <= N_MAX) */
307: unsigned s; /* number of simple-valued codes (0..s-1) */
308: ush *d; /* list of base values for non-simple codes */
309: ush *e; /* list of extra bits for non-simple codes */
310: struct huft **t; /* result: starting table */
311: int *m; /* maximum lookup bits, returns actual */
312: /* Given a list of code lengths and a maximum table size, make a set of
313: tables to decode that set of codes. Return zero on success, one if
314: the given code set is incomplete (the tables are still built in this
315: case), two if the input is invalid (all zero length codes or an
316: oversubscribed set of lengths), and three if not enough memory. */
317: {
318: unsigned a; /* counter for codes of length k */
319: unsigned c[BMAX+1]; /* bit length count table */
320: unsigned f; /* i repeats in table every f entries */
321: int g; /* maximum code length */
322: int h; /* table level */
323: register unsigned i; /* counter, current code */
324: register unsigned j; /* counter */
325: register int k; /* number of bits in current code */
326: int l; /* bits per table (returned in m) */
327: register unsigned *p; /* pointer into c[], b[], or v[] */
328: register struct huft *q; /* points to current table */
329: struct huft r; /* table entry for structure assignment */
330: struct huft *u[BMAX]; /* table stack */
331: unsigned v[N_MAX]; /* values in order of bit length */
332: register int w; /* bits before this table == (l * h) */
333: unsigned x[BMAX+1]; /* bit offsets, then code stack */
334: unsigned *xp; /* pointer into x */
335: int y; /* number of dummy codes added */
336: unsigned z; /* number of entries in current table */
337:
338:
339: /* Generate counts for each bit length */
340: memset(c, 0, sizeof(c));
341: p = b; i = n;
342: do {
343: c[*p++]++; /* assume all entries <= BMAX */
344: } while (--i);
345: if (c[0] == n) /* null input--all zero length codes */
346: {
347: *t = (struct huft *)NULL;
348: *m = 0;
349: return 0;
350: }
351:
352:
353: /* Find minimum and maximum length, bound *m by those */
354: l = *m;
355: for (j = 1; j <= BMAX; j++)
356: if (c[j])
357: break;
358: k = j; /* minimum code length */
359: if ((unsigned)l < j)
360: l = j;
361: for (i = BMAX; i; i--)
362: if (c[i])
363: break;
364: g = i; /* maximum code length */
365: if ((unsigned)l > i)
366: l = i;
367: *m = l;
368:
369:
370: /* Adjust last length count to fill out codes, if needed */
371: for (y = 1 << j; j < i; j++, y <<= 1)
372: if ((y -= c[j]) < 0)
373: return 2; /* bad input: more codes than bits */
374: if ((y -= c[i]) < 0)
375: return 2;
376: c[i] += y;
377:
378:
379: /* Generate starting offsets into the value table for each length */
380: x[1] = j = 0;
381: p = c + 1; xp = x + 2;
382: while (--i) { /* note that i == g from above */
383: *xp++ = (j += *p++);
384: }
385:
386:
387: /* Make a table of values in order of bit lengths */
388: p = b; i = 0;
389: do {
390: if ((j = *p++) != 0)
391: v[x[j]++] = i;
392: } while (++i < n);
393:
394:
395: /* Generate the Huffman codes and for each, make the table entries */
396: x[0] = i = 0; /* first Huffman code is zero */
397: p = v; /* grab values in bit order */
398: h = -1; /* no tables yet--level -1 */
399: w = -l; /* bits decoded == (l * h) */
400: u[0] = (struct huft *)NULL; /* just to keep compilers happy */
401: q = (struct huft *)NULL; /* ditto */
402: z = 0; /* ditto */
403:
404: /* go through the bit lengths (k already is bits in shortest code) */
405: for (; k <= g; k++)
406: {
407: a = c[k];
408: while (a--)
409: {
410: /* here i is the Huffman code of length k bits for value *p */
411: /* make tables up to required level */
412: while (k > w + l)
413: {
414: h++;
415: w += l; /* previous table always l bits */
416:
417: /* compute minimum size table less than or equal to l bits */
418: z = (z = g - w) > (unsigned)l ? l : z; /* upper limit on table size */
419: if ((f = 1 << (j = k - w)) > a + 1) /* try a k-w bit table */
420: { /* too few codes for k-w bit table */
421: f -= a + 1; /* deduct codes from patterns left */
422: xp = c + k;
423: while (++j < z) /* try smaller tables up to z bits */
424: {
425: if ((f <<= 1) <= *++xp)
426: break; /* enough codes to use up j bits */
427: f -= *xp; /* else deduct codes from patterns */
428: }
429: }
430: z = 1 << j; /* table entries for j-bit table */
431:
432: /* allocate and link in new table */
433: if ((q = (struct huft *)malloc((z + 1)*sizeof(struct huft))) ==
434: (struct huft *)NULL)
435: {
436: if (h)
437: huft_free(u[0]);
438: return 3; /* not enough memory */
439: }
440: hufts += z + 1; /* track memory usage */
441: *t = q + 1; /* link to list for huft_free() */
442: *(t = &(q->v.t)) = (struct huft *)NULL;
443: u[h] = ++q; /* table starts after link */
444:
445: /* connect to last table, if there is one */
446: if (h)
447: {
448: x[h] = i; /* save pattern for backing up */
449: r.b = (uch)l; /* bits to dump before this table */
450: r.e = (uch)(16 + j); /* bits in this table */
451: r.v.t = q; /* pointer to this table */
452: j = i >> (w - l); /* (get around Turbo C bug) */
453: u[h-1][j] = r; /* connect to last table */
454: }
455: }
456:
457: /* set up table entry in r */
458: r.b = (uch)(k - w);
459: if (p >= v + n)
460: r.e = 99; /* out of values--invalid code */
461: else if (*p < s)
462: {
463: r.e = (uch)(*p < 256 ? 16 : 15); /* 256 is end-of-block code */
464: r.v.n = *p++; /* simple code is just the value */
465: }
466: else
467: {
468: r.e = (uch)e[*p - s]; /* non-simple--look up in lists */
469: r.v.n = d[*p++ - s];
470: }
471:
472: /* fill code-like entries with r */
473: f = 1 << (k - w);
474: for (j = i >> w; j < z; j += f)
475: q[j] = r;
476:
477: /* backwards increment the k-bit code i */
478: for (j = 1 << (k - 1); i & j; j >>= 1)
479: i ^= j;
480: i ^= j;
481:
482: /* backup over finished tables */
483: while ((i & ((1 << w) - 1)) != x[h])
484: {
485: h--; /* don't need to update q */
486: w -= l;
487: }
488: }
489: }
490:
491:
492: /* Return true (1) if we were given an incomplete table */
493: return y != 0 && g != 1;
494: }
495:
496:
497:
498: int huft_free(t)
499: struct huft *t; /* table to free */
500: /* Free the malloc'ed tables built by huft_build(), which makes a linked
501: list of the tables it made, with the links in a dummy first entry of
502: each table. */
503: {
504: register struct huft *p, *q;
505:
506:
507: /* Go through linked list, freeing from the malloced (t[-1]) address. */
508: p = t;
509: while (p != (struct huft *)NULL)
510: {
511: q = (--p)->v.t;
512: free(p);
513: p = q;
514: }
515: return 0;
516: }
517:
518:
519:
520: void flush(w)
521: unsigned w; /* number of bytes to flush */
522: /* Do the equivalent of OUTB for the bytes slide[0..w-1]. */
523: {
524: unsigned n;
525: uch *p;
526:
527: p = slide;
528: while (w)
529: {
530: n = (n = OUTBUFSIZ - outcnt) < w ? n : w;
531: memcpy(outptr, p, n); /* try to fill up buffer */
532: outptr += n;
533: if ((outcnt += n) == OUTBUFSIZ)
534: FlushOutput(); /* if full, empty */
535: p += n;
536: w -= n;
537: }
538: }
539:
540:
541:
542: int inflate_codes(tl, td, bl, bd)
543: struct huft *tl, *td; /* literal/length and distance decoder tables */
544: int bl, bd; /* number of bits decoded by tl[] and td[] */
545: /* inflate (decompress) the codes in a deflated (compressed) block.
546: Return an error code or zero if it all goes ok. */
547: {
548: register unsigned e; /* table entry flag/number of extra bits */
549: unsigned n, d; /* length and index for copy */
550: unsigned w; /* current window position */
551: struct huft *t; /* pointer to table entry */
552: unsigned ml, md; /* masks for bl and bd bits */
553: register ulg b; /* bit buffer */
554: register unsigned k; /* number of bits in bit buffer */
555:
556:
557: /* make local copies of globals */
558: b = bb; /* initialize bit buffer */
559: k = bk;
560: w = wp; /* initialize window position */
561:
562:
563: /* inflate the coded data */
564: ml = mask_bits[bl]; /* precompute masks for speed */
565: md = mask_bits[bd];
566: while (1) /* do until end of block */
567: {
568: NEEDBITS((unsigned)bl)
569: if ((e = (t = tl + ((unsigned)b & ml))->e) > 16)
570: do {
571: if (e == 99)
572: return 1;
573: DUMPBITS(t->b)
574: e -= 16;
575: NEEDBITS(e)
576: } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16);
577: DUMPBITS(t->b)
578: if (e == 16) /* then it's a literal */
579: {
580: slide[w++] = (uch)t->v.n;
581: if (w == WSIZE)
582: {
583: flush(w);
584: w = 0;
585: }
586: }
587: else /* it's an EOB or a length */
588: {
589: /* exit if end of block */
590: if (e == 15)
591: break;
592:
593: /* get length of block to copy */
594: NEEDBITS(e)
595: n = t->v.n + ((unsigned)b & mask_bits[e]);
596: DUMPBITS(e);
597:
598: /* decode distance of block to copy */
599: NEEDBITS((unsigned)bd)
600: if ((e = (t = td + ((unsigned)b & md))->e) > 16)
601: do {
602: if (e == 99)
603: return 1;
604: DUMPBITS(t->b)
605: e -= 16;
606: NEEDBITS(e)
607: } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16);
608: DUMPBITS(t->b)
609: NEEDBITS(e)
610: d = w - t->v.n - ((unsigned)b & mask_bits[e]);
611: DUMPBITS(e)
612:
613: /* do the copy */
614: do {
615: n -= (e = (e = WSIZE - ((d &= WSIZE-1) > w ? d : w)) > n ? n : e);
616: #ifndef NOMEMCPY
617: if (w - d >= e) /* (this test assumes unsigned comparison) */
618: {
619: memcpy(slide + w, slide + d, e);
620: w += e;
621: d += e;
622: }
623: else /* do it slow to avoid memcpy() overlap */
624: #endif /* !NOMEMCPY */
625: do {
626: slide[w++] = slide[d++];
627: } while (--e);
628: if (w == WSIZE)
629: {
630: flush(w);
631: w = 0;
632: }
633: } while (n);
634: }
635: }
636:
637:
638: /* restore the globals from the locals */
639: wp = w; /* restore global window pointer */
640: bb = b; /* restore global bit buffer */
641: bk = k;
642:
643:
644: /* done */
645: return 0;
646: }
647:
648:
649:
650: int inflate_stored()
651: /* "decompress" an inflated type 0 (stored) block. */
652: {
653: unsigned n; /* number of bytes in block */
654: unsigned w; /* current window position */
655: register ulg b; /* bit buffer */
656: register unsigned k; /* number of bits in bit buffer */
657:
658:
659: /* make local copies of globals */
660: b = bb; /* initialize bit buffer */
661: k = bk;
662: w = wp; /* initialize window position */
663:
664:
665: /* go to byte boundary */
666: n = k & 7;
667: DUMPBITS(n);
668:
669:
670: /* get the length and its complement */
671: NEEDBITS(16)
672: n = ((unsigned)b & 0xffff);
673: DUMPBITS(16)
674: NEEDBITS(16)
675: if (n != (unsigned)((~b) & 0xffff))
676: return 1; /* error in compressed data */
677: DUMPBITS(16)
678:
679:
680: /* read and output the compressed data */
681: while (n--)
682: {
683: NEEDBITS(8)
684: slide[w++] = (uch)b;
685: if (w == WSIZE)
686: {
687: flush(w);
688: w = 0;
689: }
690: DUMPBITS(8)
691: }
692:
693:
694: /* restore the globals from the locals */
695: wp = w; /* restore global window pointer */
696: bb = b; /* restore global bit buffer */
697: bk = k;
698: return 0;
699: }
700:
701:
702:
703: int inflate_fixed()
704: /* decompress an inflated type 1 (fixed Huffman codes) block. We should
705: either replace this with a custom decoder, or at least precompute the
706: Huffman tables. */
707: {
708: int i; /* temporary variable */
709: struct huft *tl; /* literal/length code table */
710: struct huft *td; /* distance code table */
711: int bl; /* lookup bits for tl */
712: int bd; /* lookup bits for td */
713: unsigned l[288]; /* length list for huft_build */
714:
715:
716: /* set up literal table */
717: for (i = 0; i < 144; i++)
718: l[i] = 8;
719: for (; i < 256; i++)
720: l[i] = 9;
721: for (; i < 280; i++)
722: l[i] = 7;
723: for (; i < 288; i++) /* make a complete, but wrong code set */
724: l[i] = 8;
725: bl = 7;
726: if ((i = huft_build(l, 288, 257, cplens, cplext, &tl, &bl)) != 0)
727: return i;
728:
729:
730: /* set up distance table */
731: for (i = 0; i < 30; i++) /* make an incomplete code set */
732: l[i] = 5;
733: bd = 5;
734: if ((i = huft_build(l, 30, 0, cpdist, cpdext, &td, &bd)) > 1)
735: {
736: huft_free(tl);
737: return i;
738: }
739:
740:
741: /* decompress until an end-of-block code */
742: if (inflate_codes(tl, td, bl, bd))
743: return 1;
744:
745:
746: /* free the decoding tables, return */
747: huft_free(tl);
748: huft_free(td);
749: return 0;
750: }
751:
752:
753:
754: int inflate_dynamic()
755: /* decompress an inflated type 2 (dynamic Huffman codes) block. */
756: {
757: int i; /* temporary variables */
758: unsigned j;
759: unsigned l; /* last length */
760: unsigned m; /* mask for bit lengths table */
761: unsigned n; /* number of lengths to get */
762: struct huft *tl; /* literal/length code table */
763: struct huft *td; /* distance code table */
764: int bl; /* lookup bits for tl */
765: int bd; /* lookup bits for td */
766: unsigned nb; /* number of bit length codes */
767: unsigned nl; /* number of literal/length codes */
768: unsigned nd; /* number of distance codes */
769: #ifdef PKZIP_BUG_WORKAROUND
770: unsigned ll[288+32]; /* literal/length and distance code lengths */
771: #else
772: unsigned ll[286+30]; /* literal/length and distance code lengths */
773: #endif
774: register ulg b; /* bit buffer */
775: register unsigned k; /* number of bits in bit buffer */
776:
777:
778: /* make local bit buffer */
779: b = bb;
780: k = bk;
781:
782:
783: /* read in table lengths */
784: NEEDBITS(5)
785: nl = 257 + ((unsigned)b & 0x1f); /* number of literal/length codes */
786: DUMPBITS(5)
787: NEEDBITS(5)
788: nd = 1 + ((unsigned)b & 0x1f); /* number of distance codes */
789: DUMPBITS(5)
790: NEEDBITS(4)
791: nb = 4 + ((unsigned)b & 0xf); /* number of bit length codes */
792: DUMPBITS(4)
793: #ifdef PKZIP_BUG_WORKAROUND
794: if (nl > 288 || nd > 32)
795: #else
796: if (nl > 286 || nd > 30)
797: #endif
798: return 1; /* bad lengths */
799:
800:
801: /* read in bit-length-code lengths */
802: for (j = 0; j < nb; j++)
803: {
804: NEEDBITS(3)
805: ll[border[j]] = (unsigned)b & 7;
806: DUMPBITS(3)
807: }
808: for (; j < 19; j++)
809: ll[border[j]] = 0;
810:
811:
812: /* build decoding table for trees--single level, 7 bit lookup */
813: bl = 7;
814: if ((i = huft_build(ll, 19, 19, NULL, NULL, &tl, &bl)) != 0)
815: {
816: if (i == 1)
817: huft_free(tl);
818: return i; /* incomplete code set */
819: }
820:
821:
822: /* read in literal and distance code lengths */
823: n = nl + nd;
824: m = mask_bits[bl];
825: i = l = 0;
826: while ((unsigned)i < n)
827: {
828: NEEDBITS((unsigned)bl)
829: j = (td = tl + ((unsigned)b & m))->b;
830: DUMPBITS(j)
831: j = td->v.n;
832: if (j < 16) /* length of code in bits (0..15) */
833: ll[i++] = l = j; /* save last length in l */
834: else if (j == 16) /* repeat last length 3 to 6 times */
835: {
836: NEEDBITS(2)
837: j = 3 + ((unsigned)b & 3);
838: DUMPBITS(2)
839: if ((unsigned)i + j > n)
840: return 1;
841: while (j--)
842: ll[i++] = l;
843: }
844: else if (j == 17) /* 3 to 10 zero length codes */
845: {
846: NEEDBITS(3)
847: j = 3 + ((unsigned)b & 7);
848: DUMPBITS(3)
849: if ((unsigned)i + j > n)
850: return 1;
851: while (j--)
852: ll[i++] = 0;
853: l = 0;
854: }
855: else /* j == 18: 11 to 138 zero length codes */
856: {
857: NEEDBITS(7)
858: j = 11 + ((unsigned)b & 0x7f);
859: DUMPBITS(7)
860: if ((unsigned)i + j > n)
861: return 1;
862: while (j--)
863: ll[i++] = 0;
864: l = 0;
865: }
866: }
867:
868:
869: /* free decoding table for trees */
870: huft_free(tl);
871:
872:
873: /* restore the global bit buffer */
874: bb = b;
875: bk = k;
876:
877:
878: /* build the decoding tables for literal/length and distance codes */
879: bl = lbits;
880: if ((i = huft_build(ll, nl, 257, cplens, cplext, &tl, &bl)) != 0)
881: {
882: if (i == 1) {
883: fprintf(stderr, " incomplete literal tree\n");
884: huft_free(tl);
885: }
886: return i; /* incomplete code set */
887: }
888: bd = dbits;
889: if ((i = huft_build(ll + nl, nd, 0, cpdist, cpdext, &td, &bd)) != 0)
890: {
891: if (i == 1) {
892: fprintf(stderr, " incomplete distance tree\n");
893: #ifdef PKZIP_BUG_WORKAROUND
894: i = 0;
895: }
896: #else
897: huft_free(td);
898: }
899: huft_free(tl);
900: return i; /* incomplete code set */
901: #endif
902: }
903:
904:
905: /* decompress until an end-of-block code */
906: if (inflate_codes(tl, td, bl, bd))
907: return 1;
908:
909:
910: /* free the decoding tables, return */
911: huft_free(tl);
912: huft_free(td);
913: return 0;
914: }
915:
916:
917:
918: int inflate_block(e)
919: int *e; /* last block flag */
920: /* decompress an inflated block */
921: {
922: unsigned t; /* block type */
923: register ulg b; /* bit buffer */
924: register unsigned k; /* number of bits in bit buffer */
925:
926:
927: /* make local bit buffer */
928: b = bb;
929: k = bk;
930:
931:
932: /* read in last block bit */
933: NEEDBITS(1)
934: *e = (int)b & 1;
935: DUMPBITS(1)
936:
937:
938: /* read in block type */
939: NEEDBITS(2)
940: t = (unsigned)b & 3;
941: DUMPBITS(2)
942:
943:
944: /* restore the global bit buffer */
945: bb = b;
946: bk = k;
947:
948:
949: /* inflate that block type */
950: if (t == 2)
951: return inflate_dynamic();
952: if (t == 0)
953: return inflate_stored();
954: if (t == 1)
955: return inflate_fixed();
956:
957:
958: /* bad block type */
959: return 2;
960: }
961:
962:
963:
964: int inflate()
965: /* decompress an inflated entry */
966: {
967: int e; /* last block flag */
968: int r; /* result code */
969: unsigned h; /* maximum struct huft's malloc'ed */
970:
971:
972: /* initialize window, bit buffer */
973: wp = 0;
974: bk = 0;
975: bb = 0;
976:
977:
978: /* decompress until the last block */
979: h = 0;
980: do {
981: hufts = 0;
982: if ((r = inflate_block(&e)) != 0)
983: {
984: Trace((stderr, "\ninflate_block returned %d", r));
985: return r;
986: }
987: if (hufts > h)
988: h = hufts;
989: } while (!e);
990:
991:
992: /* flush out slide */
993: flush(wp);
994:
995:
996: /* return success */
997: #ifdef DEBUG
998: fprintf(stderr, "<%u> ", h);
999: #endif /* DEBUG */
1000: return 0;
1001: }
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.