1/* inflate.c -- Not copyrighted 1992 by Mark Adler 2 version c10p1, 10 January 1993 */ 3 4/* You can do whatever you like with this source file, though I would 5 prefer that if you modify it and redistribute it that you include 6 comments to that effect with your name and the date. Thank you. 7 [The history has been moved to the file ChangeLog.] 8 */ 9 10/* 11 Inflate deflated (PKZIP's method 8 compressed) data. The compression 12 method searches for as much of the current string of bytes (up to a 13 length of 258) in the previous 32K bytes. If it doesn't find any 14 matches (of at least length 3), it codes the next byte. Otherwise, it 15 codes the length of the matched string and its distance backwards from 16 the current position. There is a single Huffman code that codes both 17 single bytes (called "literals") and match lengths. A second Huffman 18 code codes the distance information, which follows a length code. Each 19 length or distance code actually represents a base value and a number 20 of "extra" (sometimes zero) bits to get to add to the base value. At 21 the end of each deflated block is a special end-of-block (EOB) literal/ 22 length code. The decoding process is basically: get a literal/length 23 code; if EOB then done; if a literal, emit the decoded byte; if a 24 length then get the distance and emit the referred-to bytes from the 25 sliding window of previously emitted data. 26 27 There are (currently) three kinds of inflate blocks: stored, fixed, and 28 dynamic. The compressor deals with some chunk of data at a time, and 29 decides which method to use on a chunk-by-chunk basis. A chunk might 30 typically be 32K or 64K. If the chunk is uncompressible, then the 31 "stored" method is used. In this case, the bytes are simply stored as 32 is, eight bits per byte, with none of the above coding. The bytes are 33 preceded by a count, since there is no longer an EOB code. 34 35 If the data is compressible, then either the fixed or dynamic methods 36 are used. In the dynamic method, the compressed data is preceded by 37 an encoding of the literal/length and distance Huffman codes that are 38 to be used to decode this block. The representation is itself Huffman 39 coded, and so is preceded by a description of that code. These code 40 descriptions take up a little space, and so for small blocks, there is 41 a predefined set of codes, called the fixed codes. The fixed method is 42 used if the block codes up smaller that way (usually for quite small 43 chunks), otherwise the dynamic method is used. In the latter case, the 44 codes are customized to the probabilities in the current block, and so 45 can code it much better than the pre-determined fixed codes. 46 47 The Huffman codes themselves are decoded using a mutli-level table 48 lookup, in order to maximize the speed of decoding plus the speed of 49 building the decoding tables. See the comments below that precede the 50 lbits and dbits tuning parameters. 51 */ 52 53 54/* 55 Notes beyond the 1.93a appnote.txt: 56 57 1. Distance pointers never point before the beginning of the output 58 stream. 59 2. Distance pointers can point back across blocks, up to 32k away. 60 3. There is an implied maximum of 7 bits for the bit length table and 61 15 bits for the actual data. 62 4. If only one code exists, then it is encoded using one bit. (Zero 63 would be more efficient, but perhaps a little confusing.) If two 64 codes exist, they are coded using one bit each (0 and 1). 65 5. There is no way of sending zero distance codes--a dummy must be 66 sent if there are none. (History: a pre 2.0 version of PKZIP would 67 store blocks with no distance codes, but this was discovered to be 68 too harsh a criterion.) Valid only for 1.93a. 2.04c does allow 69 zero distance codes, which is sent as one code of zero bits in 70 length. 71 6. There are up to 286 literal/length codes. Code 256 represents the 72 end-of-block. Note however that the static length tree defines 73 288 codes just to fill out the Huffman codes. Codes 286 and 287 74 cannot be used though, since there is no length base or extra bits 75 defined for them. Similarly, there are up to 30 distance codes. 76 However, static trees define 32 codes (all 5 bits) to fill out the 77 Huffman codes, but the last two had better not show up in the data. 78 7. Unzip can check dynamic Huffman blocks for complete code sets. 79 The exception is that a single code would not be complete (see #4). 80 8. The five bits following the block type is really the number of 81 literal codes sent minus 257. 82 9. Length codes 8,16,16 are interpreted as 13 length codes of 8 bits 83 (1+6+6). Therefore, to output three times the length, you output 84 three codes (1+1+1), whereas to output four times the same length, 85 you only need two codes (1+3). Hmm. 86 10. In the tree reconstruction algorithm, Code = Code + Increment 87 only if BitLength(i) is not zero. (Pretty obvious.) 88 11. Correction: 4 Bits: # of Bit Length codes - 4 (4 - 19) 89 12. Note: length code 284 can represent 227-258, but length code 285 90 really is 258. The last length deserves its own, short code 91 since it gets used a lot in very redundant files. The length 92 258 is special since 258 - 3 (the min match length) is 255. 93 13. The literal/length and distance code bit lengths are read as a 94 single stream of lengths. It is possible (and advantageous) for 95 a repeat code (16, 17, or 18) to go across the boundary between 96 the two sets of lengths. 97 */ 98 99#ifdef RCSID 100static char rcsid[] = "$Id: inflate.c 3476 2003-06-11 15:56:10Z darkwyrm $"; 101#endif 102 103#include <sys/types.h> 104 105#include "tailor.h" 106 107#if defined(STDC_HEADERS) || !defined(NO_STDLIB_H) 108# include <stdlib.h> 109#endif 110 111#include "gzip.h" 112#define slide window 113 114/* Huffman code lookup table entry--this entry is four bytes for machines 115 that have 16-bit pointers (e.g. PC's in the small or medium model). 116 Valid extra bits are 0..13. e == 15 is EOB (end of block), e == 16 117 means that v is a literal, 16 < e < 32 means that v is a pointer to 118 the next table, which codes e - 16 bits, and lastly e == 99 indicates 119 an unused code. If a code with e == 99 is looked up, this implies an 120 error in the data. */ 121struct huft { 122 uch e; /* number of extra bits or operation */ 123 uch b; /* number of bits in this code or subcode */ 124 union { 125 ush n; /* literal, length base, or distance base */ 126 struct huft *t; /* pointer to next level of table */ 127 } v; 128}; 129 130 131/* Function prototypes */ 132int huft_build OF((unsigned *, unsigned, unsigned, ush *, ush *, 133 struct huft **, int *)); 134int huft_free OF((struct huft *)); 135int inflate_codes OF((struct huft *, struct huft *, int, int)); 136int inflate_stored OF((void)); 137int inflate_fixed OF((void)); 138int inflate_dynamic OF((void)); 139int inflate_block OF((int *)); 140int inflate OF((void)); 141 142 143/* The inflate algorithm uses a sliding 32K byte window on the uncompressed 144 stream to find repeated byte strings. This is implemented here as a 145 circular buffer. The index is updated simply by incrementing and then 146 and'ing with 0x7fff (32K-1). */ 147/* It is left to other modules to supply the 32K area. It is assumed 148 to be usable as if it were declared "uch slide[32768];" or as just 149 "uch *slide;" and then malloc'ed in the latter case. The definition 150 must be in unzip.h, included above. */ 151/* unsigned wp; current position in slide */ 152#define wp outcnt 153#define flush_output(w) (wp=(w),flush_window()) 154 155/* Tables for deflate from PKZIP's appnote.txt. */ 156static unsigned border[] = { /* Order of the bit length code lengths */ 157 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15}; 158static ush cplens[] = { /* Copy lengths for literal codes 257..285 */ 159 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31, 160 35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0}; 161 /* note: see note #13 above about the 258 in this list. */ 162static ush cplext[] = { /* Extra bits for literal codes 257..285 */ 163 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 164 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, 99, 99}; /* 99==invalid */ 165static ush cpdist[] = { /* Copy offsets for distance codes 0..29 */ 166 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, 167 257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145, 168 8193, 12289, 16385, 24577}; 169static ush cpdext[] = { /* Extra bits for distance codes */ 170 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 171 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 172 12, 12, 13, 13}; 173 174 175 176/* Macros for inflate() bit peeking and grabbing. 177 The usage is: 178 179 NEEDBITS(j) 180 x = b & mask_bits[j]; 181 DUMPBITS(j) 182 183 where NEEDBITS makes sure that b has at least j bits in it, and 184 DUMPBITS removes the bits from b. The macros use the variable k 185 for the number of bits in b. Normally, b and k are register 186 variables for speed, and are initialized at the beginning of a 187 routine that uses these macros from a global bit buffer and count. 188 189 If we assume that EOB will be the longest code, then we will never 190 ask for bits with NEEDBITS that are beyond the end of the stream. 191 So, NEEDBITS should not read any more bytes than are needed to 192 meet the request. Then no bytes need to be "returned" to the buffer 193 at the end of the last block. 194 195 However, this assumption is not true for fixed blocks--the EOB code 196 is 7 bits, but the other literal/length codes can be 8 or 9 bits. 197 (The EOB code is shorter than other codes because fixed blocks are 198 generally short. So, while a block always has an EOB, many other 199 literal/length codes have a significantly lower probability of 200 showing up at all.) However, by making the first table have a 201 lookup of seven bits, the EOB code will be found in that first 202 lookup, and so will not require that too many bits be pulled from 203 the stream. 204 */ 205 206ulg bb; /* bit buffer */ 207unsigned bk; /* bits in bit buffer */ 208 209ush mask_bits[] = { 210 0x0000, 211 0x0001, 0x0003, 0x0007, 0x000f, 0x001f, 0x003f, 0x007f, 0x00ff, 212 0x01ff, 0x03ff, 0x07ff, 0x0fff, 0x1fff, 0x3fff, 0x7fff, 0xffff 213}; 214 215#ifdef CRYPT 216 uch cc; 217# define NEXTBYTE() \ 218 (decrypt ? (cc = get_byte(), zdecode(cc), cc) : get_byte()) 219#else 220# define NEXTBYTE() (uch)get_byte() 221#endif 222#define NEEDBITS(n) {while(k<(n)){b|=((ulg)NEXTBYTE())<<k;k+=8;}} 223#define DUMPBITS(n) {b>>=(n);k-=(n);} 224 225 226/* 227 Huffman code decoding is performed using a multi-level table lookup. 228 The fastest way to decode is to simply build a lookup table whose 229 size is determined by the longest code. However, the time it takes 230 to build this table can also be a factor if the data being decoded 231 is not very long. The most common codes are necessarily the 232 shortest codes, so those codes dominate the decoding time, and hence 233 the speed. The idea is you can have a shorter table that decodes the 234 shorter, more probable codes, and then point to subsidiary tables for 235 the longer codes. The time it costs to decode the longer codes is 236 then traded against the time it takes to make longer tables. 237 238 This results of this trade are in the variables lbits and dbits 239 below. lbits is the number of bits the first level table for literal/ 240 length codes can decode in one step, and dbits is the same thing for 241 the distance codes. Subsequent tables are also less than or equal to 242 those sizes. These values may be adjusted either when all of the 243 codes are shorter than that, in which case the longest code length in 244 bits is used, or when the shortest code is *longer* than the requested 245 table size, in which case the length of the shortest code in bits is 246 used. 247 248 There are two different values for the two tables, since they code a 249 different number of possibilities each. The literal/length table 250 codes 286 possible values, or in a flat code, a little over eight 251 bits. The distance table codes 30 possible values, or a little less 252 than five bits, flat. The optimum values for speed end up being 253 about one bit more than those, so lbits is 8+1 and dbits is 5+1. 254 The optimum values may differ though from machine to machine, and 255 possibly even between compilers. Your mileage may vary. 256 */ 257 258 259int lbits = 9; /* bits in base literal/length lookup table */ 260int dbits = 6; /* bits in base distance lookup table */ 261 262 263/* If BMAX needs to be larger than 16, then h and x[] should be ulg. */ 264#define BMAX 16 /* maximum bit length of any code (16 for explode) */ 265#define N_MAX 288 /* maximum number of codes in any set */ 266 267 268unsigned hufts; /* track memory usage */ 269 270 271int huft_build(b, n, s, d, e, t, m) 272unsigned *b; /* code lengths in bits (all assumed <= BMAX) */ 273unsigned n; /* number of codes (assumed <= N_MAX) */ 274unsigned s; /* number of simple-valued codes (0..s-1) */ 275ush *d; /* list of base values for non-simple codes */ 276ush *e; /* list of extra bits for non-simple codes */ 277struct huft **t; /* result: starting table */ 278int *m; /* maximum lookup bits, returns actual */ 279/* Given a list of code lengths and a maximum table size, make a set of 280 tables to decode that set of codes. Return zero on success, one if 281 the given code set is incomplete (the tables are still built in this 282 case), two if the input is invalid (all zero length codes or an 283 oversubscribed set of lengths), and three if not enough memory. */ 284{ 285 unsigned a; /* counter for codes of length k */ 286 unsigned c[BMAX+1]; /* bit length count table */ 287 unsigned f; /* i repeats in table every f entries */ 288 int g; /* maximum code length */ 289 int h; /* table level */ 290 register unsigned i; /* counter, current code */ 291 register unsigned j; /* counter */ 292 register int k; /* number of bits in current code */ 293 int l; /* bits per table (returned in m) */ 294 register unsigned *p; /* pointer into c[], b[], or v[] */ 295 register struct huft *q; /* points to current table */ 296 struct huft r; /* table entry for structure assignment */ 297 struct huft *u[BMAX]; /* table stack */ 298 unsigned v[N_MAX]; /* values in order of bit length */ 299 register int w; /* bits before this table == (l * h) */ 300 unsigned x[BMAX+1]; /* bit offsets, then code stack */ 301 unsigned *xp; /* pointer into x */ 302 int y; /* number of dummy codes added */ 303 unsigned z; /* number of entries in current table */ 304 305 306 /* Generate counts for each bit length */ 307 memzero(c, sizeof(c)); 308 p = b; i = n; 309 do { 310 Tracecv(*p, (stderr, (n-i >= ' ' && n-i <= '~' ? "%c %d\n" : "0x%x %d\n"), 311 n-i, *p)); 312 c[*p]++; /* assume all entries <= BMAX */ 313 p++; /* Can't combine with above line (Solaris bug) */ 314 } while (--i); 315 if (c[0] == n) /* null input--all zero length codes */ 316 { 317 *t = (struct huft *)NULL; 318 *m = 0; 319 return 0; 320 } 321 322 323 /* Find minimum and maximum length, bound *m by those */ 324 l = *m; 325 for (j = 1; j <= BMAX; j++) 326 if (c[j]) 327 break; 328 k = j; /* minimum code length */ 329 if ((unsigned)l < j) 330 l = j; 331 for (i = BMAX; i; i--) 332 if (c[i]) 333 break; 334 g = i; /* maximum code length */ 335 if ((unsigned)l > i) 336 l = i; 337 *m = l; 338 339 340 /* Adjust last length count to fill out codes, if needed */ 341 for (y = 1 << j; j < i; j++, y <<= 1) 342 if ((y -= c[j]) < 0) 343 return 2; /* bad input: more codes than bits */ 344 if ((y -= c[i]) < 0) 345 return 2; 346 c[i] += y; 347 348 349 /* Generate starting offsets into the value table for each length */ 350 x[1] = j = 0; 351 p = c + 1; xp = x + 2; 352 while (--i) { /* note that i == g from above */ 353 *xp++ = (j += *p++); 354 } 355 356 357 /* Make a table of values in order of bit lengths */ 358 p = b; i = 0; 359 do { 360 if ((j = *p++) != 0) 361 v[x[j]++] = i; 362 } while (++i < n); 363 364 365 /* Generate the Huffman codes and for each, make the table entries */ 366 x[0] = i = 0; /* first Huffman code is zero */ 367 p = v; /* grab values in bit order */ 368 h = -1; /* no tables yet--level -1 */ 369 w = -l; /* bits decoded == (l * h) */ 370 u[0] = (struct huft *)NULL; /* just to keep compilers happy */ 371 q = (struct huft *)NULL; /* ditto */ 372 z = 0; /* ditto */ 373 374 /* go through the bit lengths (k already is bits in shortest code) */ 375 for (; k <= g; k++) 376 { 377 a = c[k]; 378 while (a--) 379 { 380 /* here i is the Huffman code of length k bits for value *p */ 381 /* make tables up to required level */ 382 while (k > w + l) 383 { 384 h++; 385 w += l; /* previous table always l bits */ 386 387 /* compute minimum size table less than or equal to l bits */ 388 z = (z = g - w) > (unsigned)l ? l : z; /* upper limit on table size */ 389 if ((f = 1 << (j = k - w)) > a + 1) /* try a k-w bit table */ 390 { /* too few codes for k-w bit table */ 391 f -= a + 1; /* deduct codes from patterns left */ 392 xp = c + k; 393 while (++j < z) /* try smaller tables up to z bits */ 394 { 395 if ((f <<= 1) <= *++xp) 396 break; /* enough codes to use up j bits */ 397 f -= *xp; /* else deduct codes from patterns */ 398 } 399 } 400 z = 1 << j; /* table entries for j-bit table */ 401 402 /* allocate and link in new table */ 403 if ((q = (struct huft *)malloc((z + 1)*sizeof(struct huft))) == 404 (struct huft *)NULL) 405 { 406 if (h) 407 huft_free(u[0]); 408 return 3; /* not enough memory */ 409 } 410 hufts += z + 1; /* track memory usage */ 411 *t = q + 1; /* link to list for huft_free() */ 412 *(t = &(q->v.t)) = (struct huft *)NULL; 413 u[h] = ++q; /* table starts after link */ 414 415 /* connect to last table, if there is one */ 416 if (h) 417 { 418 x[h] = i; /* save pattern for backing up */ 419 r.b = (uch)l; /* bits to dump before this table */ 420 r.e = (uch)(16 + j); /* bits in this table */ 421 r.v.t = q; /* pointer to this table */ 422 j = i >> (w - l); /* (get around Turbo C bug) */ 423 u[h-1][j] = r; /* connect to last table */ 424 } 425 } 426 427 /* set up table entry in r */ 428 r.b = (uch)(k - w); 429 if (p >= v + n) 430 r.e = 99; /* out of values--invalid code */ 431 else if (*p < s) 432 { 433 r.e = (uch)(*p < 256 ? 16 : 15); /* 256 is end-of-block code */ 434 r.v.n = (ush)(*p); /* simple code is just the value */ 435 p++; /* one compiler does not like *p++ */ 436 } 437 else 438 { 439 r.e = (uch)e[*p - s]; /* non-simple--look up in lists */ 440 r.v.n = d[*p++ - s]; 441 } 442 443 /* fill code-like entries with r */ 444 f = 1 << (k - w); 445 for (j = i >> w; j < z; j += f) 446 q[j] = r; 447 448 /* backwards increment the k-bit code i */ 449 for (j = 1 << (k - 1); i & j; j >>= 1) 450 i ^= j; 451 i ^= j; 452 453 /* backup over finished tables */ 454 while ((i & ((1 << w) - 1)) != x[h]) 455 { 456 h--; /* don't need to update q */ 457 w -= l; 458 } 459 } 460 } 461 462 463 /* Return true (1) if we were given an incomplete table */ 464 return y != 0 && g != 1; 465} 466 467 468 469int huft_free(t) 470struct huft *t; /* table to free */ 471/* Free the malloc'ed tables built by huft_build(), which makes a linked 472 list of the tables it made, with the links in a dummy first entry of 473 each table. */ 474{ 475 register struct huft *p, *q; 476 477 478 /* Go through linked list, freeing from the malloced (t[-1]) address. */ 479 p = t; 480 while (p != (struct huft *)NULL) 481 { 482 q = (--p)->v.t; 483 free((char*)p); 484 p = q; 485 } 486 return 0; 487} 488 489 490int inflate_codes(tl, td, bl, bd) 491struct huft *tl, *td; /* literal/length and distance decoder tables */ 492int bl, bd; /* number of bits decoded by tl[] and td[] */ 493/* inflate (decompress) the codes in a deflated (compressed) block. 494 Return an error code or zero if it all goes ok. */ 495{ 496 register unsigned e; /* table entry flag/number of extra bits */ 497 unsigned n, d; /* length and index for copy */ 498 unsigned w; /* current window position */ 499 struct huft *t; /* pointer to table entry */ 500 unsigned ml, md; /* masks for bl and bd bits */ 501 register ulg b; /* bit buffer */ 502 register unsigned k; /* number of bits in bit buffer */ 503 504 505 /* make local copies of globals */ 506 b = bb; /* initialize bit buffer */ 507 k = bk; 508 w = wp; /* initialize window position */ 509 510 /* inflate the coded data */ 511 ml = mask_bits[bl]; /* precompute masks for speed */ 512 md = mask_bits[bd]; 513 for (;;) /* do until end of block */ 514 { 515 NEEDBITS((unsigned)bl) 516 if ((e = (t = tl + ((unsigned)b & ml))->e) > 16) 517 do { 518 if (e == 99) 519 return 1; 520 DUMPBITS(t->b) 521 e -= 16; 522 NEEDBITS(e) 523 } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16); 524 DUMPBITS(t->b) 525 if (e == 16) /* then it's a literal */ 526 { 527 slide[w++] = (uch)t->v.n; 528 Tracevv((stderr, "%c", slide[w-1])); 529 if (w == WSIZE) 530 { 531 flush_output(w); 532 w = 0; 533 } 534 } 535 else /* it's an EOB or a length */ 536 { 537 /* exit if end of block */ 538 if (e == 15) 539 break; 540 541 /* get length of block to copy */ 542 NEEDBITS(e) 543 n = t->v.n + ((unsigned)b & mask_bits[e]); 544 DUMPBITS(e); 545 546 /* decode distance of block to copy */ 547 NEEDBITS((unsigned)bd) 548 if ((e = (t = td + ((unsigned)b & md))->e) > 16) 549 do { 550 if (e == 99) 551 return 1; 552 DUMPBITS(t->b) 553 e -= 16; 554 NEEDBITS(e) 555 } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16); 556 DUMPBITS(t->b) 557 NEEDBITS(e) 558 d = w - t->v.n - ((unsigned)b & mask_bits[e]); 559 DUMPBITS(e) 560 Tracevv((stderr,"\\[%d,%d]", w-d, n)); 561 562 /* do the copy */ 563 do { 564 n -= (e = (e = WSIZE - ((d &= WSIZE-1) > w ? d : w)) > n ? n : e); 565#if !defined(NOMEMCPY) && !defined(DEBUG) 566 if (w - d >= e) /* (this test assumes unsigned comparison) */ 567 { 568 memcpy(slide + w, slide + d, e); 569 w += e; 570 d += e; 571 } 572 else /* do it slow to avoid memcpy() overlap */ 573#endif /* !NOMEMCPY */ 574 do { 575 slide[w++] = slide[d++]; 576 Tracevv((stderr, "%c", slide[w-1])); 577 } while (--e); 578 if (w == WSIZE) 579 { 580 flush_output(w); 581 w = 0; 582 } 583 } while (n); 584 } 585 } 586 587 588 /* restore the globals from the locals */ 589 wp = w; /* restore global window pointer */ 590 bb = b; /* restore global bit buffer */ 591 bk = k; 592 593 /* done */ 594 return 0; 595} 596 597 598 599int inflate_stored() 600/* "decompress" an inflated type 0 (stored) block. */ 601{ 602 unsigned n; /* number of bytes in block */ 603 unsigned w; /* current window position */ 604 register ulg b; /* bit buffer */ 605 register unsigned k; /* number of bits in bit buffer */ 606 607 608 /* make local copies of globals */ 609 b = bb; /* initialize bit buffer */ 610 k = bk; 611 w = wp; /* initialize window position */ 612 613 614 /* go to byte boundary */ 615 n = k & 7; 616 DUMPBITS(n); 617 618 619 /* get the length and its complement */ 620 NEEDBITS(16) 621 n = ((unsigned)b & 0xffff); 622 DUMPBITS(16) 623 NEEDBITS(16) 624 if (n != (unsigned)((~b) & 0xffff)) 625 return 1; /* error in compressed data */ 626 DUMPBITS(16) 627 628 629 /* read and output the compressed data */ 630 while (n--) 631 { 632 NEEDBITS(8) 633 slide[w++] = (uch)b; 634 if (w == WSIZE) 635 { 636 flush_output(w); 637 w = 0; 638 } 639 DUMPBITS(8) 640 } 641 642 643 /* restore the globals from the locals */ 644 wp = w; /* restore global window pointer */ 645 bb = b; /* restore global bit buffer */ 646 bk = k; 647 return 0; 648} 649 650 651 652int inflate_fixed() 653/* decompress an inflated type 1 (fixed Huffman codes) block. We should 654 either replace this with a custom decoder, or at least precompute the 655 Huffman tables. */ 656{ 657 int i; /* temporary variable */ 658 struct huft *tl; /* literal/length code table */ 659 struct huft *td; /* distance code table */ 660 int bl; /* lookup bits for tl */ 661 int bd; /* lookup bits for td */ 662 unsigned l[288]; /* length list for huft_build */ 663 664 665 /* set up literal table */ 666 for (i = 0; i < 144; i++) 667 l[i] = 8; 668 for (; i < 256; i++) 669 l[i] = 9; 670 for (; i < 280; i++) 671 l[i] = 7; 672 for (; i < 288; i++) /* make a complete, but wrong code set */ 673 l[i] = 8; 674 bl = 7; 675 if ((i = huft_build(l, 288, 257, cplens, cplext, &tl, &bl)) != 0) 676 return i; 677 678 679 /* set up distance table */ 680 for (i = 0; i < 30; i++) /* make an incomplete code set */ 681 l[i] = 5; 682 bd = 5; 683 if ((i = huft_build(l, 30, 0, cpdist, cpdext, &td, &bd)) > 1) 684 { 685 huft_free(tl); 686 return i; 687 } 688 689 690 /* decompress until an end-of-block code */ 691 if (inflate_codes(tl, td, bl, bd)) 692 return 1; 693 694 695 /* free the decoding tables, return */ 696 huft_free(tl); 697 huft_free(td); 698 return 0; 699} 700 701 702 703int inflate_dynamic() 704/* decompress an inflated type 2 (dynamic Huffman codes) block. */ 705{ 706 int i; /* temporary variables */ 707 unsigned j; 708 unsigned l; /* last length */ 709 unsigned m; /* mask for bit lengths table */ 710 unsigned n; /* number of lengths to get */ 711 struct huft *tl; /* literal/length code table */ 712 struct huft *td; /* distance code table */ 713 int bl; /* lookup bits for tl */ 714 int bd; /* lookup bits for td */ 715 unsigned nb; /* number of bit length codes */ 716 unsigned nl; /* number of literal/length codes */ 717 unsigned nd; /* number of distance codes */ 718#ifdef PKZIP_BUG_WORKAROUND 719 unsigned ll[288+32]; /* literal/length and distance code lengths */ 720#else 721 unsigned ll[286+30]; /* literal/length and distance code lengths */ 722#endif 723 register ulg b; /* bit buffer */ 724 register unsigned k; /* number of bits in bit buffer */ 725 726 727 /* make local bit buffer */ 728 b = bb; 729 k = bk; 730 731 732 /* read in table lengths */ 733 NEEDBITS(5) 734 nl = 257 + ((unsigned)b & 0x1f); /* number of literal/length codes */ 735 DUMPBITS(5) 736 NEEDBITS(5) 737 nd = 1 + ((unsigned)b & 0x1f); /* number of distance codes */ 738 DUMPBITS(5) 739 NEEDBITS(4) 740 nb = 4 + ((unsigned)b & 0xf); /* number of bit length codes */ 741 DUMPBITS(4) 742#ifdef PKZIP_BUG_WORKAROUND 743 if (nl > 288 || nd > 32) 744#else 745 if (nl > 286 || nd > 30) 746#endif 747 return 1; /* bad lengths */ 748 749 750 /* read in bit-length-code lengths */ 751 for (j = 0; j < nb; j++) 752 { 753 NEEDBITS(3) 754 ll[border[j]] = (unsigned)b & 7; 755 DUMPBITS(3) 756 } 757 for (; j < 19; j++) 758 ll[border[j]] = 0; 759 760 761 /* build decoding table for trees--single level, 7 bit lookup */ 762 bl = 7; 763 if ((i = huft_build(ll, 19, 19, NULL, NULL, &tl, &bl)) != 0) 764 { 765 if (i == 1) 766 huft_free(tl); 767 return i; /* incomplete code set */ 768 } 769 770 771 /* read in literal and distance code lengths */ 772 n = nl + nd; 773 m = mask_bits[bl]; 774 i = l = 0; 775 while ((unsigned)i < n) 776 { 777 NEEDBITS((unsigned)bl) 778 j = (td = tl + ((unsigned)b & m))->b; 779 DUMPBITS(j) 780 j = td->v.n; 781 if (j < 16) /* length of code in bits (0..15) */ 782 ll[i++] = l = j; /* save last length in l */ 783 else if (j == 16) /* repeat last length 3 to 6 times */ 784 { 785 NEEDBITS(2) 786 j = 3 + ((unsigned)b & 3); 787 DUMPBITS(2) 788 if ((unsigned)i + j > n) 789 return 1; 790 while (j--) 791 ll[i++] = l; 792 } 793 else if (j == 17) /* 3 to 10 zero length codes */ 794 { 795 NEEDBITS(3) 796 j = 3 + ((unsigned)b & 7); 797 DUMPBITS(3) 798 if ((unsigned)i + j > n) 799 return 1; 800 while (j--) 801 ll[i++] = 0; 802 l = 0; 803 } 804 else /* j == 18: 11 to 138 zero length codes */ 805 { 806 NEEDBITS(7) 807 j = 11 + ((unsigned)b & 0x7f); 808 DUMPBITS(7) 809 if ((unsigned)i + j > n) 810 return 1; 811 while (j--) 812 ll[i++] = 0; 813 l = 0; 814 } 815 } 816 817 818 /* free decoding table for trees */ 819 huft_free(tl); 820 821 822 /* restore the global bit buffer */ 823 bb = b; 824 bk = k; 825 826 827 /* build the decoding tables for literal/length and distance codes */ 828 bl = lbits; 829 if ((i = huft_build(ll, nl, 257, cplens, cplext, &tl, &bl)) != 0) 830 { 831 if (i == 1) { 832 fprintf(stderr, " incomplete literal tree\n"); 833 huft_free(tl); 834 } 835 return i; /* incomplete code set */ 836 } 837 bd = dbits; 838 if ((i = huft_build(ll + nl, nd, 0, cpdist, cpdext, &td, &bd)) != 0) 839 { 840 if (i == 1) { 841 fprintf(stderr, " incomplete distance tree\n"); 842#ifdef PKZIP_BUG_WORKAROUND 843 i = 0; 844 } 845#else 846 huft_free(td); 847 } 848 huft_free(tl); 849 return i; /* incomplete code set */ 850#endif 851 } 852 853 854 /* decompress until an end-of-block code */ 855 if (inflate_codes(tl, td, bl, bd)) 856 return 1; 857 858 859 /* free the decoding tables, return */ 860 huft_free(tl); 861 huft_free(td); 862 return 0; 863} 864 865 866 867int inflate_block(e) 868int *e; /* last block flag */ 869/* decompress an inflated block */ 870{ 871 unsigned t; /* block type */ 872 register ulg b; /* bit buffer */ 873 register unsigned k; /* number of bits in bit buffer */ 874 875 876 /* make local bit buffer */ 877 b = bb; 878 k = bk; 879 880 881 /* read in last block bit */ 882 NEEDBITS(1) 883 *e = (int)b & 1; 884 DUMPBITS(1) 885 886 887 /* read in block type */ 888 NEEDBITS(2) 889 t = (unsigned)b & 3; 890 DUMPBITS(2) 891 892 893 /* restore the global bit buffer */ 894 bb = b; 895 bk = k; 896 897 898 /* inflate that block type */ 899 if (t == 2) 900 return inflate_dynamic(); 901 if (t == 0) 902 return inflate_stored(); 903 if (t == 1) 904 return inflate_fixed(); 905 906 907 /* bad block type */ 908 return 2; 909} 910 911 912 913int inflate() 914/* decompress an inflated entry */ 915{ 916 int e; /* last block flag */ 917 int r; /* result code */ 918 unsigned h; /* maximum struct huft's malloc'ed */ 919 920 921 /* initialize window, bit buffer */ 922 wp = 0; 923 bk = 0; 924 bb = 0; 925 926 927 /* decompress until the last block */ 928 h = 0; 929 do { 930 hufts = 0; 931 if ((r = inflate_block(&e)) != 0) 932 return r; 933 if (hufts > h) 934 h = hufts; 935 } while (!e); 936 937 /* Undo too much lookahead. The next read will be byte aligned so we 938 * can discard unused bits in the last meaningful byte. 939 */ 940 while (bk >= 8) { 941 bk -= 8; 942 inptr--; 943 } 944 945 /* flush out slide */ 946 flush_output(wp); 947 948 949 /* return success */ 950#ifdef DEBUG 951 fprintf(stderr, "<%u> ", h); 952#endif /* DEBUG */ 953 return 0; 954} 955