1 /* cabextract 0.6 - a program to extract Microsoft Cabinet files
2 * (C) 2000-2002 Stuart Caie <kyzer@4u.net>
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License as published by
6 * the Free Software Foundation; either version 2 of the License, or
7 * (at your option) any later version.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write to the Free Software
16 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
19 /* This is NOT a general purpose cabinet library with a front end tacked
20 * on. If you want a comprehensive library to read and write cabinet
21 * files, please get "libcabinet". If you want to create CAB files on UNIX
22 * systems, get "Cablinux".
24 * Get the official Microsoft CAB SDK from here:
25 * http://msdn.microsoft.com/workshop/management/cab/cab-sdk.exe
26 * You can use cabextract on this file to extract the contents.
28 * Many thanks to Dirk Stoecker, Matthew Russoto and Dave Tritscher and,
29 * of course, Microsoft for the documentation they _did_ provide wholly
30 * and accurately. MSZIP is a one-byte adaption of the deflate and inflate
31 * methods created by Phil Katz. Quantum is based on the Quantum archiver,
32 * created by David Stafford. LZX is an adaption of the LZX method created
33 * by Jonathan Forbes and Tomi Poutanen.
35 * Furthermore, thanks to Jae Jung, for single-handedly fixing all the
36 * bugs with LZX decompression in cabextract 0.1, and Eric Sharkey for the
37 * original manual page.
40 /* CAB files are 'cabinets'. 'Folders' store compressed data, and may span
41 * several cabinets. 'Files' live as data inside a folder when
42 * uncompressed. EOR checksums are used instead of CRCs. Four compression
43 * formats are known - NONE, MSZIP, QUANTUM and LZX. NONE is obviously
44 * uncompressed data. MSZIP is simply PKZIP's deflate/inflate algorithims
45 * with 'CK' as a signature instead of 'PK'. QUANTUM is an LZ77 +
46 * arithmetic coding method. LZX is a much loved LZH based archiver in the
47 * Amiga world, the algorithm taken (bought?) by Microsoft and tweaked for
54 #include <stdio.h> /* everyone has this! */
56 #ifdef HAVE_SYS_TYPES_H
57 # include <sys/types.h>
80 #ifdef HAVE_SYS_STAT_H
81 # include <sys/stat.h>
84 #if TIME_WITH_SYS_TIME
85 # include <sys/time.h>
89 # include <sys/time.h>
102 # define dirent direct
104 # include <sys/ndir.h>
107 # include <sys/dir.h>
116 # define strchr index
117 # define strrchr rindex
119 # if !HAVE_STRCASECMP
120 # define strcasecmp strcmpi
123 # define memcpy(d,,n) bcopy((s),(d),(n)
128 extern time_t mktime(struct tm *tp);
133 #else /* !HAVE_CONFIG_H */
139 #include <sys/stat.h>
140 #include <sys/types.h>
146 #define VERSION "x.x"
151 # define D(x) printf x ;
157 typedef unsigned char UBYTE; /* 8 bits exactly */
158 typedef unsigned short UWORD; /* 16 bits (or more) */
159 typedef unsigned int ULONG; /* 32 bits (or more) */
160 typedef signed int LONG; /* 32 bits (or more) */
162 /* number of bits in a ULONG */
164 # define CHAR_BIT (8)
166 #define ULONG_BITS (sizeof(ULONG) * CHAR_BIT)
168 /* endian-neutral reading of little-endian data */
169 #define EndGetI32(a) ((((a)[3])<<24)|(((a)[2])<<16)|(((a)[1])<<8)|((a)[0]))
170 #define EndGetI16(a) ((((a)[1])<<8)|((a)[0]))
172 /* maximum number of cabinets any one folder can be split across */
173 #define CAB_SPLITMAX (10)
176 struct cabinet *next; /* for making a list of cabinets */
177 char *filename; /* input name of cabinet */
178 FILE *fh; /* open file handle or NULL */
179 off_t filelen; /* length of cabinet file */
180 off_t blocks_off; /* offset to data blocks in file */
181 struct cabinet *prevcab, *nextcab; /* multipart cabinet chains */
182 char *prevname, *nextname; /* and their filenames */
183 char *previnfo, *nextinfo; /* and their visible names */
184 struct folder *folders; /* first folder in this cabinet */
185 struct file *files; /* first file in this cabinet */
186 UBYTE block_resv; /* reserved space in datablocks */
187 UBYTE flags; /* header flags */
192 struct cabinet *cab[CAB_SPLITMAX]; /* cabinet(s) this folder spans */
193 off_t offset[CAB_SPLITMAX]; /* offset to data blocks */
194 UWORD comp_type; /* compression format/window size */
195 ULONG comp_size; /* compressed size of folder */
196 UBYTE num_splits; /* number of split blocks + 1 */
197 UWORD num_blocks; /* total number of blocks */
198 struct file *contfile; /* the first split file */
202 struct file *next; /* next file in sequence */
203 struct folder *folder; /* folder that contains this file */
204 char *filename; /* output name of file */
205 FILE *fh; /* open file handle or NULL */
206 ULONG length; /* uncompressed length of file */
207 ULONG offset; /* uncompressed offset in folder */
208 UWORD index; /* magic index number of folder */
209 UWORD time, date, attribs; /* MS-DOS time/date/attributes */
213 /* structure offsets */
214 #define cfhead_Signature (0x00)
215 #define cfhead_CabinetSize (0x08)
216 #define cfhead_FileOffset (0x10)
217 #define cfhead_MinorVersion (0x18)
218 #define cfhead_MajorVersion (0x19)
219 #define cfhead_NumFolders (0x1A)
220 #define cfhead_NumFiles (0x1C)
221 #define cfhead_Flags (0x1E)
222 #define cfhead_SetID (0x20)
223 #define cfhead_CabinetIndex (0x22)
224 #define cfhead_SIZEOF (0x24)
225 #define cfheadext_HeaderReserved (0x00)
226 #define cfheadext_FolderReserved (0x02)
227 #define cfheadext_DataReserved (0x03)
228 #define cfheadext_SIZEOF (0x04)
229 #define cffold_DataOffset (0x00)
230 #define cffold_NumBlocks (0x04)
231 #define cffold_CompType (0x06)
232 #define cffold_SIZEOF (0x08)
233 #define cffile_UncompressedSize (0x00)
234 #define cffile_FolderOffset (0x04)
235 #define cffile_FolderIndex (0x08)
236 #define cffile_Date (0x0A)
237 #define cffile_Time (0x0C)
238 #define cffile_Attribs (0x0E)
239 #define cffile_SIZEOF (0x10)
240 #define cfdata_CheckSum (0x00)
241 #define cfdata_CompressedSize (0x04)
242 #define cfdata_UncompressedSize (0x06)
243 #define cfdata_SIZEOF (0x08)
246 #define cffoldCOMPTYPE_MASK (0x000f)
247 #define cffoldCOMPTYPE_NONE (0x0000)
248 #define cffoldCOMPTYPE_MSZIP (0x0001)
249 #define cffoldCOMPTYPE_QUANTUM (0x0002)
250 #define cffoldCOMPTYPE_LZX (0x0003)
251 #define cfheadPREV_CABINET (0x0001)
252 #define cfheadNEXT_CABINET (0x0002)
253 #define cfheadRESERVE_PRESENT (0x0004)
254 #define cffileCONTINUED_FROM_PREV (0xFFFD)
255 #define cffileCONTINUED_TO_NEXT (0xFFFE)
256 #define cffileCONTINUED_PREV_AND_NEXT (0xFFFF)
257 #define cffile_A_RDONLY (0x01)
258 #define cffile_A_HIDDEN (0x02)
259 #define cffile_A_SYSTEM (0x04)
260 #define cffile_A_ARCH (0x20)
261 #define cffile_A_EXEC (0x40)
262 #define cffile_A_NAME_IS_UTF (0x80)
265 /*--------------------------------------------------------------------------*/
266 /* our archiver information / state */
269 #define ZIPWSIZE 0x8000 /* window size */
270 #define ZIPLBITS 9 /* bits in base literal/length lookup table */
271 #define ZIPDBITS 6 /* bits in base distance lookup table */
272 #define ZIPBMAX 16 /* maximum bit length of any code */
273 #define ZIPN_MAX 288 /* maximum number of codes in any set */
276 UBYTE e; /* number of extra bits or operation */
277 UBYTE b; /* number of bits in this code or subcode */
279 UWORD n; /* literal, length base, or distance base */
280 struct Ziphuft *t; /* pointer to next level of table */
285 ULONG window_posn; /* current offset within the window */
286 ULONG bb; /* bit buffer */
287 ULONG bk; /* bits in bit buffer */
288 ULONG ll[288+32]; /* literal/length and distance code lengths */
289 ULONG c[ZIPBMAX+1]; /* bit length count table */
290 LONG lx[ZIPBMAX+1]; /* memory for l[-1..ZIPBMAX-1] */
291 struct Ziphuft *u[ZIPBMAX]; /* table stack */
292 ULONG v[ZIPN_MAX]; /* values in order of bit length */
293 ULONG x[ZIPBMAX+1]; /* bit offsets, then code stack */
304 int shiftsleft, entries;
305 struct QTMmodelsym *syms;
310 UBYTE *window; /* the actual decoding window */
311 ULONG window_size; /* window size (1Kb through 2Mb) */
312 ULONG actual_size; /* window size when it was first allocated */
313 ULONG window_posn; /* current offset within the window */
315 struct QTMmodel model7;
316 struct QTMmodelsym m7sym[7+1];
318 struct QTMmodel model4, model5, model6pos, model6len;
319 struct QTMmodelsym m4sym[0x18 + 1];
320 struct QTMmodelsym m5sym[0x24 + 1];
321 struct QTMmodelsym m6psym[0x2a + 1], m6lsym[0x1b + 1];
323 struct QTMmodel model00, model40, model80, modelC0;
324 struct QTMmodelsym m00sym[0x40 + 1], m40sym[0x40 + 1];
325 struct QTMmodelsym m80sym[0x40 + 1], mC0sym[0x40 + 1];
330 /* some constants defined by the LZX specification */
331 #define LZX_MIN_MATCH (2)
332 #define LZX_MAX_MATCH (257)
333 #define LZX_NUM_CHARS (256)
334 #define LZX_BLOCKTYPE_INVALID (0) /* also blocktypes 4-7 invalid */
335 #define LZX_BLOCKTYPE_VERBATIM (1)
336 #define LZX_BLOCKTYPE_ALIGNED (2)
337 #define LZX_BLOCKTYPE_UNCOMPRESSED (3)
338 #define LZX_PRETREE_NUM_ELEMENTS (20)
339 #define LZX_ALIGNED_NUM_ELEMENTS (8) /* aligned offset tree #elements */
340 #define LZX_NUM_PRIMARY_LENGTHS (7) /* this one missing from spec! */
341 #define LZX_NUM_SECONDARY_LENGTHS (249) /* length tree #elements */
343 /* LZX huffman defines: tweak tablebits as desired */
344 #define LZX_PRETREE_MAXSYMBOLS (LZX_PRETREE_NUM_ELEMENTS)
345 #define LZX_PRETREE_TABLEBITS (6)
346 #define LZX_MAINTREE_MAXSYMBOLS (LZX_NUM_CHARS + 50*8)
347 #define LZX_MAINTREE_TABLEBITS (12)
348 #define LZX_LENGTH_MAXSYMBOLS (LZX_NUM_SECONDARY_LENGTHS+1)
349 #define LZX_LENGTH_TABLEBITS (12)
350 #define LZX_ALIGNED_MAXSYMBOLS (LZX_ALIGNED_NUM_ELEMENTS)
351 #define LZX_ALIGNED_TABLEBITS (7)
353 #define LZX_LENTABLE_SAFETY (64) /* we allow length table decoding overruns */
355 #define LZX_DECLARE_TABLE(tbl) \
356 UWORD tbl##_table[(1<<LZX_##tbl##_TABLEBITS) + (LZX_##tbl##_MAXSYMBOLS<<1)];\
357 UBYTE tbl##_len [LZX_##tbl##_MAXSYMBOLS + LZX_LENTABLE_SAFETY]
360 UBYTE *window; /* the actual decoding window */
361 ULONG window_size; /* window size (32Kb through 2Mb) */
362 ULONG actual_size; /* window size when it was first allocated */
363 ULONG window_posn; /* current offset within the window */
364 ULONG R0, R1, R2; /* for the LRU offset system */
365 UWORD main_elements; /* number of main tree elements */
366 int header_read; /* have we started decoding at all yet? */
367 UWORD block_type; /* type of this block */
368 ULONG block_length; /* uncompressed length of this block */
369 ULONG block_remaining; /* uncompressed bytes still left to decode */
370 ULONG frames_read; /* the number of CFDATA blocks processed */
371 LONG intel_filesize; /* magic header value used for transform */
372 LONG intel_curpos; /* current offset in transform space */
373 int intel_started; /* have we seen any translatable data yet? */
375 LZX_DECLARE_TABLE(PRETREE);
376 LZX_DECLARE_TABLE(MAINTREE);
377 LZX_DECLARE_TABLE(LENGTH);
378 LZX_DECLARE_TABLE(ALIGNED);
383 #define CAB(x) (decomp_state.x)
384 #define ZIP(x) (decomp_state.methods.zip.x)
385 #define QTM(x) (decomp_state.methods.qtm.x)
386 #define LZX(x) (decomp_state.methods.lzx.x)
388 #define DECR_DATAFORMAT (1)
389 #define DECR_ILLEGALDATA (2)
390 #define DECR_NOMEMORY (3)
391 #define DECR_CHECKSUM (4)
392 #define DECR_INPUT (5)
393 #define DECR_OUTPUT (6)
395 /* CAB data blocks are <= 32768 bytes in uncompressed form. Uncompressed
396 * blocks have zero growth. MSZIP guarantees that it won't grow above
397 * uncompressed size by more than 12 bytes. LZX guarantees it won't grow
398 * more than 6144 bytes.
400 #define CAB_BLOCKMAX (32768)
401 #define CAB_INPUTMAX (CAB_BLOCKMAX+6144)
404 struct folder *current; /* current folder we're extracting from */
405 ULONG offset; /* uncompressed offset within folder */
406 UBYTE *outpos; /* (high level) start of data to use up */
407 UWORD outlen; /* (high level) amount of data to use up */
408 UWORD split; /* at which split in current folder? */
409 int (*decompress)(int, int); /* the chosen compression func */
410 UBYTE inbuf[CAB_INPUTMAX+2]; /* +2 for lzx bitbuffer overflows! */
411 UBYTE outbuf[CAB_BLOCKMAX];
420 /* MSZIP decruncher */
422 /* Dirk Stoecker wrote the ZIP decoder, based on the InfoZip deflate code */
424 /* Tables for deflate from PKZIP's appnote.txt. */
425 static const UBYTE Zipborder[] = /* Order of the bit length code lengths */
426 { 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
427 static const UWORD Zipcplens[] = /* Copy lengths for literal codes 257..285 */
428 { 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31, 35, 43, 51,
429 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
430 static const UWORD Zipcplext[] = /* Extra bits for literal codes 257..285 */
431 { 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 4,
432 4, 5, 5, 5, 5, 0, 99, 99}; /* 99==invalid */
433 static const UWORD Zipcpdist[] = /* Copy offsets for distance codes 0..29 */
434 { 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, 257, 385,
435 513, 769, 1025, 1537, 2049, 3073, 4097, 6145, 8193, 12289, 16385, 24577};
436 static const UWORD Zipcpdext[] = /* Extra bits for distance codes */
437 { 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10,
438 10, 11, 11, 12, 12, 13, 13};
440 /* And'ing with Zipmask[n] masks the lower n bits */
441 static const UWORD Zipmask[17] = {
442 0x0000, 0x0001, 0x0003, 0x0007, 0x000f, 0x001f, 0x003f, 0x007f, 0x00ff,
443 0x01ff, 0x03ff, 0x07ff, 0x0fff, 0x1fff, 0x3fff, 0x7fff, 0xffff
446 #define ZIPNEEDBITS(n) {while(k<(n)){LONG c=*(ZIP(inpos)++);\
447 b|=((ULONG)c)<<k;k+=8;}}
448 #define ZIPDUMPBITS(n) {b>>=(n);k-=(n);}
450 void Ziphuft_free(struct Ziphuft *t)
452 register struct Ziphuft *p, *q;
454 /* Go through linked list, freeing from the allocated (t[-1]) address. */
456 while (p != (struct Ziphuft *)NULL)
464 LONG Ziphuft_build(ULONG *b, ULONG n, ULONG s, UWORD *d, UWORD *e,
465 struct Ziphuft **t, LONG *m)
467 ULONG a; /* counter for codes of length k */
468 ULONG el; /* length of EOB code (value 256) */
469 ULONG f; /* i repeats in table every f entries */
470 LONG g; /* maximum code length */
471 LONG h; /* table level */
472 register ULONG i; /* counter, current code */
473 register ULONG j; /* counter */
474 register LONG k; /* number of bits in current code */
475 LONG *l; /* stack of bits per table */
476 register ULONG *p; /* pointer into ZIP(c)[],ZIP(b)[],ZIP(v)[] */
477 register struct Ziphuft *q; /* points to current table */
478 struct Ziphuft r; /* table entry for structure assignment */
479 register LONG w; /* bits before this table == (l * h) */
480 ULONG *xp; /* pointer into x */
481 LONG y; /* number of dummy codes added */
482 ULONG z; /* number of entries in current table */
486 /* Generate counts for each bit length */
487 el = n > 256 ? b[256] : ZIPBMAX; /* set length of EOB code, if any */
489 for(i = 0; i < ZIPBMAX+1; ++i)
494 ZIP(c)[*p]++; p++; /* assume all entries <= ZIPBMAX */
496 if (ZIP(c)[0] == n) /* null input--all zero length codes */
498 *t = (struct Ziphuft *)NULL;
503 /* Find minimum and maximum length, bound *m by those */
504 for (j = 1; j <= ZIPBMAX; j++)
507 k = j; /* minimum code length */
510 for (i = ZIPBMAX; i; i--)
513 g = i; /* maximum code length */
517 /* Adjust last length count to fill out codes, if needed */
518 for (y = 1 << j; j < i; j++, y <<= 1)
519 if ((y -= ZIP(c)[j]) < 0)
520 return 2; /* bad input: more codes than bits */
521 if ((y -= ZIP(c)[i]) < 0)
525 /* Generate starting offsets LONGo the value table for each length */
527 p = ZIP(c) + 1; xp = ZIP(x) + 2;
529 { /* note that i == g from above */
533 /* Make a table of values in order of bit lengths */
537 ZIP(v)[ZIP(x)[j]++] = i;
541 /* Generate the Huffman codes and for each, make the table entries */
542 ZIP(x)[0] = i = 0; /* first Huffman code is zero */
543 p = ZIP(v); /* grab values in bit order */
544 h = -1; /* no tables yet--level -1 */
545 w = l[-1] = 0; /* no bits decoded yet */
546 ZIP(u)[0] = (struct Ziphuft *)NULL; /* just to keep compilers happy */
547 q = (struct Ziphuft *)NULL; /* ditto */
550 /* go through the bit lengths (k already is bits in shortest code) */
556 /* here i is the Huffman code of length k bits for value *p */
557 /* make tables up to required level */
560 w += l[h++]; /* add bits already decoded */
562 /* compute minimum size table less than or equal to *m bits */
563 z = (z = g - w) > (ULONG)*m ? *m : z; /* upper limit */
564 if ((f = 1 << (j = k - w)) > a + 1) /* try a k-w bit table */
565 { /* too few codes for k-w bit table */
566 f -= a + 1; /* deduct codes from patterns left */
568 while (++j < z) /* try smaller tables up to z bits */
570 if ((f <<= 1) <= *++xp)
571 break; /* enough codes to use up j bits */
572 f -= *xp; /* else deduct codes from patterns */
575 if ((ULONG)w + j > el && (ULONG)w < el)
576 j = el - w; /* make EOB code end at table */
577 z = 1 << j; /* table entries for j-bit table */
578 l[h] = j; /* set table size in stack */
580 /* allocate and link in new table */
581 if (!(q = (struct Ziphuft *) malloc((z + 1)*sizeof(struct Ziphuft))))
584 Ziphuft_free(ZIP(u)[0]);
585 return 3; /* not enough memory */
587 *t = q + 1; /* link to list for Ziphuft_free() */
588 *(t = &(q->v.t)) = (struct Ziphuft *)NULL;
589 ZIP(u)[h] = ++q; /* table starts after link */
591 /* connect to last table, if there is one */
594 ZIP(x)[h] = i; /* save pattern for backing up */
595 r.b = (UBYTE)l[h-1]; /* bits to dump before this table */
596 r.e = (UBYTE)(16 + j); /* bits in this table */
597 r.v.t = q; /* pointer to this table */
598 j = (i & ((1 << w) - 1)) >> (w - l[h-1]);
599 ZIP(u)[h-1][j] = r; /* connect to last table */
603 /* set up table entry in r */
604 r.b = (UBYTE)(k - w);
606 r.e = 99; /* out of values--invalid code */
609 r.e = (UBYTE)(*p < 256 ? 16 : 15); /* 256 is end-of-block code */
610 r.v.n = *p++; /* simple code is just the value */
614 r.e = (UBYTE)e[*p - s]; /* non-simple--look up in lists */
618 /* fill code-like entries with r */
620 for (j = i >> w; j < z; j += f)
623 /* backwards increment the k-bit code i */
624 for (j = 1 << (k - 1); i & j; j >>= 1)
628 /* backup over finished tables */
629 while ((i & ((1 << w) - 1)) != ZIP(x)[h])
630 w -= l[--h]; /* don't need to update q */
634 /* return actual size of base table */
637 /* Return true (1) if we were given an incomplete table */
638 return y != 0 && g != 1;
641 LONG Zipinflate_codes(struct Ziphuft *tl, struct Ziphuft *td,
644 register ULONG e; /* table entry flag/number of extra bits */
645 ULONG n, d; /* length and index for copy */
646 ULONG w; /* current window position */
647 struct Ziphuft *t; /* pointer to table entry */
648 ULONG ml, md; /* masks for bl and bd bits */
649 register ULONG b; /* bit buffer */
650 register ULONG k; /* number of bits in bit buffer */
652 /* make local copies of globals */
653 b = ZIP(bb); /* initialize bit buffer */
655 w = ZIP(window_posn); /* initialize window position */
657 /* inflate the coded data */
658 ml = Zipmask[bl]; /* precompute masks for speed */
663 ZIPNEEDBITS((ULONG)bl)
664 if((e = (t = tl + ((ULONG)b & ml))->e) > 16)
672 } while ((e = (t = t->v.t + ((ULONG)b & Zipmask[e]))->e) > 16);
674 if (e == 16) /* then it's a literal */
675 CAB(outbuf)[w++] = (UBYTE)t->v.n;
676 else /* it's an EOB or a length */
678 /* exit if end of block */
682 /* get length of block to copy */
684 n = t->v.n + ((ULONG)b & Zipmask[e]);
687 /* decode distance of block to copy */
688 ZIPNEEDBITS((ULONG)bd)
689 if ((e = (t = td + ((ULONG)b & md))->e) > 16)
696 } while ((e = (t = t->v.t + ((ULONG)b & Zipmask[e]))->e) > 16);
699 d = w - t->v.n - ((ULONG)b & Zipmask[e]);
703 n -= (e = (e = ZIPWSIZE - ((d &= ZIPWSIZE-1) > w ? d : w)) > n ?n:e);
706 CAB(outbuf)[w++] = CAB(outbuf)[d++];
712 /* restore the globals from the locals */
713 ZIP(window_posn) = w; /* restore global window pointer */
714 ZIP(bb) = b; /* restore global bit buffer */
721 LONG Zipinflate_stored(void)
722 /* "decompress" an inflated type 0 (stored) block. */
724 ULONG n; /* number of bytes in block */
725 ULONG w; /* current window position */
726 register ULONG b; /* bit buffer */
727 register ULONG k; /* number of bits in bit buffer */
729 /* make local copies of globals */
730 b = ZIP(bb); /* initialize bit buffer */
732 w = ZIP(window_posn); /* initialize window position */
734 /* go to byte boundary */
738 /* get the length and its complement */
740 n = ((ULONG)b & 0xffff);
743 if (n != (ULONG)((~b) & 0xffff))
744 return 1; /* error in compressed data */
747 /* read and output the compressed data */
751 CAB(outbuf)[w++] = (UBYTE)b;
755 /* restore the globals from the locals */
756 ZIP(window_posn) = w; /* restore global window pointer */
757 ZIP(bb) = b; /* restore global bit buffer */
762 LONG Zipinflate_fixed(void)
764 struct Ziphuft *fixed_tl;
765 struct Ziphuft *fixed_td;
766 LONG fixed_bl, fixed_bd;
767 LONG i; /* temporary variable */
773 for(i = 0; i < 144; i++)
779 for(; i < 288; i++) /* make a complete, but wrong code set */
782 if((i = Ziphuft_build(l, 288, 257, (UWORD *) Zipcplens,
783 (UWORD *) Zipcplext, &fixed_tl, &fixed_bl)))
787 for(i = 0; i < 30; i++) /* make an incomplete code set */
790 if((i = Ziphuft_build(l, 30, 0, (UWORD *) Zipcpdist, (UWORD *) Zipcpdext,
791 &fixed_td, &fixed_bd)) > 1)
793 Ziphuft_free(fixed_tl);
797 /* decompress until an end-of-block code */
798 i = Zipinflate_codes(fixed_tl, fixed_td, fixed_bl, fixed_bd);
800 Ziphuft_free(fixed_td);
801 Ziphuft_free(fixed_tl);
805 LONG Zipinflate_dynamic(void)
806 /* decompress an inflated type 2 (dynamic Huffman codes) block. */
808 LONG i; /* temporary variables */
811 ULONG l; /* last length */
812 ULONG m; /* mask for bit lengths table */
813 ULONG n; /* number of lengths to get */
814 struct Ziphuft *tl; /* literal/length code table */
815 struct Ziphuft *td; /* distance code table */
816 LONG bl; /* lookup bits for tl */
817 LONG bd; /* lookup bits for td */
818 ULONG nb; /* number of bit length codes */
819 ULONG nl; /* number of literal/length codes */
820 ULONG nd; /* number of distance codes */
821 register ULONG b; /* bit buffer */
822 register ULONG k; /* number of bits in bit buffer */
824 /* make local bit buffer */
829 /* read in table lengths */
831 nl = 257 + ((ULONG)b & 0x1f); /* number of literal/length codes */
834 nd = 1 + ((ULONG)b & 0x1f); /* number of distance codes */
837 nb = 4 + ((ULONG)b & 0xf); /* number of bit length codes */
839 if(nl > 288 || nd > 32)
840 return 1; /* bad lengths */
842 /* read in bit-length-code lengths */
843 for(j = 0; j < nb; j++)
846 ll[Zipborder[j]] = (ULONG)b & 7;
850 ll[Zipborder[j]] = 0;
852 /* build decoding table for trees--single level, 7 bit lookup */
854 if((i = Ziphuft_build(ll, 19, 19, NULL, NULL, &tl, &bl)) != 0)
858 return i; /* incomplete code set */
861 /* read in literal and distance code lengths */
867 ZIPNEEDBITS((ULONG)bl)
868 j = (td = tl + ((ULONG)b & m))->b;
871 if (j < 16) /* length of code in bits (0..15) */
872 ll[i++] = l = j; /* save last length in l */
873 else if (j == 16) /* repeat last length 3 to 6 times */
876 j = 3 + ((ULONG)b & 3);
883 else if (j == 17) /* 3 to 10 zero length codes */
886 j = 3 + ((ULONG)b & 7);
888 if ((ULONG)i + j > n)
894 else /* j == 18: 11 to 138 zero length codes */
897 j = 11 + ((ULONG)b & 0x7f);
899 if ((ULONG)i + j > n)
907 /* free decoding table for trees */
910 /* restore the global bit buffer */
914 /* build the decoding tables for literal/length and distance codes */
916 if((i = Ziphuft_build(ll, nl, 257, (UWORD *) Zipcplens, (UWORD *) Zipcplext, &tl, &bl)) != 0)
920 return i; /* incomplete code set */
923 Ziphuft_build(ll + nl, nd, 0, (UWORD *) Zipcpdist, (UWORD *) Zipcpdext, &td, &bd);
925 /* decompress until an end-of-block code */
926 if(Zipinflate_codes(tl, td, bl, bd))
929 /* free the decoding tables, return */
935 LONG Zipinflate_block(LONG *e) /* e == last block flag */
936 { /* decompress an inflated block */
937 ULONG t; /* block type */
938 register ULONG b; /* bit buffer */
939 register ULONG k; /* number of bits in bit buffer */
941 /* make local bit buffer */
945 /* read in last block bit */
950 /* read in block type */
955 /* restore the global bit buffer */
959 /* inflate that block type */
961 return Zipinflate_dynamic();
963 return Zipinflate_stored();
965 return Zipinflate_fixed();
970 int ZIPdecompress(int inlen, int outlen)
972 LONG e; /* last block flag */
974 ZIP(inpos) = CAB(inbuf);
975 ZIP(bb) = ZIP(bk) = ZIP(window_posn) = 0;
976 if(outlen > ZIPWSIZE)
977 return DECR_DATAFORMAT;
979 /* CK = Chris Kirmse, official Microsoft purloiner */
980 if(ZIP(inpos)[0] != 0x43 || ZIP(inpos)[1] != 0x4B)
981 return DECR_ILLEGALDATA;
986 if(Zipinflate_block(&e))
987 return DECR_ILLEGALDATA;
994 /* Quantum decruncher */
996 /* This decruncher was researched and implemented by Matthew Russoto. */
997 /* It has since been tidied up by Stuart Caie */
999 static UBYTE q_length_base[27], q_length_extra[27], q_extra_bits[42];
1000 static ULONG q_position_base[42];
1002 /* Initialise a model which decodes symbols from [s] to [s]+[n]-1 */
1003 void QTMinitmodel(struct QTMmodel *m, struct QTMmodelsym *sym, int n, int s) {
1008 memset(m->tabloc, 0xFF, sizeof(m->tabloc)); /* clear out look-up table */
1009 for (i = 0; i < n; i++) {
1010 m->tabloc[i+s] = i; /* set up a look-up entry for symbol */
1011 m->syms[i].sym = i+s; /* actual symbol */
1012 m->syms[i].cumfreq = n-i; /* current frequency of that symbol */
1014 m->syms[n].cumfreq = 0;
1017 int QTMinit(int window, int level) {
1018 int wndsize = 1 << window, msz = window * 2, i;
1021 /* QTM supports window sizes of 2^10 (1Kb) through 2^21 (2Mb) */
1022 /* if a previously allocated window is big enough, keep it */
1023 if (window < 10 || window > 21) return DECR_DATAFORMAT;
1024 if (QTM(actual_size) < wndsize) {
1025 if (QTM(window)) free(QTM(window));
1029 if (!(QTM(window) = malloc(wndsize))) return DECR_NOMEMORY;
1030 QTM(actual_size) = wndsize;
1032 QTM(window_size) = wndsize;
1033 QTM(window_posn) = 0;
1035 /* initialise static slot/extrabits tables */
1036 for (i = 0, j = 0; i < 27; i++) {
1037 q_length_extra[i] = (i == 26) ? 0 : (i < 2 ? 0 : i - 2) >> 2;
1038 q_length_base[i] = j; j += 1 << ((i == 26) ? 5 : q_length_extra[i]);
1040 for (i = 0, j = 0; i < 42; i++) {
1041 q_extra_bits[i] = (i < 2 ? 0 : i-2) >> 1;
1042 q_position_base[i] = j; j += 1 << q_extra_bits[i];
1045 /* initialise arithmetic coding models */
1047 QTMinitmodel(&QTM(model7), &QTM(m7sym)[0], 7, 0);
1049 QTMinitmodel(&QTM(model00), &QTM(m00sym)[0], 0x40, 0x00);
1050 QTMinitmodel(&QTM(model40), &QTM(m40sym)[0], 0x40, 0x40);
1051 QTMinitmodel(&QTM(model80), &QTM(m80sym)[0], 0x40, 0x80);
1052 QTMinitmodel(&QTM(modelC0), &QTM(mC0sym)[0], 0x40, 0xC0);
1054 /* model 4 depends on table size, ranges from 20 to 24 */
1055 QTMinitmodel(&QTM(model4), &QTM(m4sym)[0], (msz < 24) ? msz : 24, 0);
1056 /* model 5 depends on table size, ranges from 20 to 36 */
1057 QTMinitmodel(&QTM(model5), &QTM(m5sym)[0], (msz < 36) ? msz : 36, 0);
1058 /* model 6pos depends on table size, ranges from 20 to 42 */
1059 QTMinitmodel(&QTM(model6pos), &QTM(m6psym)[0], msz, 0);
1060 QTMinitmodel(&QTM(model6len), &QTM(m6lsym)[0], 27, 0);
1066 void QTMupdatemodel(struct QTMmodel *model, int sym) {
1067 struct QTMmodelsym temp;
1070 for (i = 0; i < sym; i++) model->syms[i].cumfreq += 8;
1072 if (model->syms[0].cumfreq > 3800) {
1073 if (--model->shiftsleft) {
1074 for (i = model->entries - 1; i >= 0; i--) {
1075 /* -1, not -2; the 0 entry saves this */
1076 model->syms[i].cumfreq >>= 1;
1077 if (model->syms[i].cumfreq <= model->syms[i+1].cumfreq) {
1078 model->syms[i].cumfreq = model->syms[i+1].cumfreq + 1;
1083 model->shiftsleft = 50;
1084 for (i = 0; i < model->entries ; i++) {
1085 /* no -1, want to include the 0 entry */
1086 /* this converts cumfreqs into frequencies, then shifts right */
1087 model->syms[i].cumfreq -= model->syms[i+1].cumfreq;
1088 model->syms[i].cumfreq++; /* avoid losing things entirely */
1089 model->syms[i].cumfreq >>= 1;
1092 /* now sort by frequencies, decreasing order -- this must be an
1093 * inplace selection sort, or a sort with the same (in)stability
1096 for (i = 0; i < model->entries - 1; i++) {
1097 for (j = i + 1; j < model->entries; j++) {
1098 if (model->syms[i].cumfreq < model->syms[j].cumfreq) {
1099 temp = model->syms[i];
1100 model->syms[i] = model->syms[j];
1101 model->syms[j] = temp;
1106 /* then convert frequencies back to cumfreq */
1107 for (i = model->entries - 1; i >= 0; i--) {
1108 model->syms[i].cumfreq += model->syms[i+1].cumfreq;
1110 /* then update the other part of the table */
1111 for (i = 0; i < model->entries; i++) {
1112 model->tabloc[model->syms[i].sym] = i;
1118 /* Bitstream reading macros (Quantum / normal byte order)
1120 * Q_INIT_BITSTREAM should be used first to set up the system
1121 * Q_READ_BITS(var,n) takes N bits from the buffer and puts them in var.
1122 * unlike LZX, this can loop several times to get the
1123 * requisite number of bits.
1124 * Q_FILL_BUFFER adds more data to the bit buffer, if there is room
1125 * for another 16 bits.
1126 * Q_PEEK_BITS(n) extracts (without removing) N bits from the bit
1128 * Q_REMOVE_BITS(n) removes N bits from the bit buffer
1130 * These bit access routines work by using the area beyond the MSB and the
1131 * LSB as a free source of zeroes. This avoids having to mask any bits.
1132 * So we have to know the bit width of the bitbuffer variable. This is
1133 * defined as ULONG_BITS.
1135 * ULONG_BITS should be at least 16 bits. Unlike LZX's Huffman decoding,
1136 * Quantum's arithmetic decoding only needs 1 bit at a time, it doesn't
1137 * need an assured number. Retrieving larger bitstrings can be done with
1138 * multiple reads and fills of the bitbuffer. The code should work fine
1139 * for machines where ULONG >= 32 bits.
1141 * Also note that Quantum reads bytes in normal order; LZX is in
1142 * little-endian order.
1145 #define Q_INIT_BITSTREAM do { bitsleft = 0; bitbuf = 0; } while (0)
1147 #define Q_FILL_BUFFER do { \
1148 if (bitsleft <= (ULONG_BITS - 16)) { \
1149 bitbuf |= ((inpos[0]<<8)|inpos[1]) << (ULONG_BITS-16 - bitsleft); \
1150 bitsleft += 16; inpos += 2; \
1154 #define Q_PEEK_BITS(n) (bitbuf >> (ULONG_BITS - (n)))
1155 #define Q_REMOVE_BITS(n) ((bitbuf <<= (n)), (bitsleft -= (n)))
1157 #define Q_READ_BITS(v,n) do { \
1159 for (bitsneed = (n); bitsneed; bitsneed -= bitrun) { \
1161 bitrun = (bitsneed > bitsleft) ? bitsleft : bitsneed; \
1162 (v) = ((v) << bitrun) | Q_PEEK_BITS(bitrun); \
1163 Q_REMOVE_BITS(bitrun); \
1167 #define Q_MENTRIES(model) (QTM(model).entries)
1168 #define Q_MSYM(model,symidx) (QTM(model).syms[(symidx)].sym)
1169 #define Q_MSYMFREQ(model,symidx) (QTM(model).syms[(symidx)].cumfreq)
1171 /* GET_SYMBOL(model, var) fetches the next symbol from the stated model
1172 * and puts it in var. it may need to read the bitstream to do this.
1174 #define GET_SYMBOL(m, var) do { \
1175 range = ((H - L) & 0xFFFF) + 1; \
1176 symf = ((((C - L + 1) * Q_MSYMFREQ(m,0)) - 1) / range) & 0xFFFF; \
1178 for (i=1; i < Q_MENTRIES(m); i++) { \
1179 if (Q_MSYMFREQ(m,i) <= symf) break; \
1181 (var) = Q_MSYM(m,i-1); \
1183 range = (H - L) + 1; \
1184 H = L + ((Q_MSYMFREQ(m,i-1) * range) / Q_MSYMFREQ(m,0)) - 1; \
1185 L = L + ((Q_MSYMFREQ(m,i) * range) / Q_MSYMFREQ(m,0)); \
1187 if ((L & 0x8000) != (H & 0x8000)) { \
1188 if ((L & 0x4000) && !(H & 0x4000)) { \
1189 /* underflow case */ \
1190 C ^= 0x4000; L &= 0x3FFF; H |= 0x4000; \
1194 L <<= 1; H = (H << 1) | 1; \
1196 C = (C << 1) | Q_PEEK_BITS(1); \
1200 QTMupdatemodel(&(QTM(m)), i); \
1204 int QTMdecompress(int inlen, int outlen) {
1205 UBYTE *inpos = CAB(inbuf);
1206 UBYTE *window = QTM(window);
1207 UBYTE *runsrc, *rundest;
1209 ULONG window_posn = QTM(window_posn);
1210 ULONG window_size = QTM(window_size);
1212 /* used by bitstream macros */
1213 register int bitsleft, bitrun, bitsneed;
1214 register ULONG bitbuf;
1216 /* used by GET_SYMBOL */
1221 int extra, togo = outlen, match_length, copy_length;
1222 UBYTE selector, sym;
1225 UWORD H = 0xFFFF, L = 0, C;
1227 /* read initial value of C */
1231 /* apply 2^x-1 mask */
1232 window_posn &= window_size - 1;
1233 /* runs can't straddle the window wraparound */
1234 if ((window_posn + togo) > window_size) {
1235 D(("straddled run\n"))
1236 return DECR_DATAFORMAT;
1240 GET_SYMBOL(model7, selector);
1243 GET_SYMBOL(model00, sym); window[window_posn++] = sym; togo--;
1246 GET_SYMBOL(model40, sym); window[window_posn++] = sym; togo--;
1249 GET_SYMBOL(model80, sym); window[window_posn++] = sym; togo--;
1252 GET_SYMBOL(modelC0, sym); window[window_posn++] = sym; togo--;
1256 /* selector 4 = fixed length of 3 */
1257 GET_SYMBOL(model4, sym);
1258 Q_READ_BITS(extra, q_extra_bits[sym]);
1259 match_offset = q_position_base[sym] + extra + 1;
1264 /* selector 5 = fixed length of 4 */
1265 GET_SYMBOL(model5, sym);
1266 Q_READ_BITS(extra, q_extra_bits[sym]);
1267 match_offset = q_position_base[sym] + extra + 1;
1272 /* selector 6 = variable length */
1273 GET_SYMBOL(model6len, sym);
1274 Q_READ_BITS(extra, q_length_extra[sym]);
1275 match_length = q_length_base[sym] + extra + 5;
1276 GET_SYMBOL(model6pos, sym);
1277 Q_READ_BITS(extra, q_extra_bits[sym]);
1278 match_offset = q_position_base[sym] + extra + 1;
1282 D(("Selector is bogus\n"))
1283 return DECR_ILLEGALDATA;
1286 /* if this is a match */
1287 if (selector >= 4) {
1288 rundest = window + window_posn;
1289 togo -= match_length;
1291 /* copy any wrapped around source data */
1292 if (window_posn >= match_offset) {
1294 runsrc = rundest - match_offset;
1296 runsrc = rundest + (window_size - match_offset);
1297 copy_length = match_offset - window_posn;
1298 if (copy_length < match_length) {
1299 match_length -= copy_length;
1300 window_posn += copy_length;
1301 while (copy_length-- > 0) *rundest++ = *runsrc++;
1305 window_posn += match_length;
1307 /* copy match data - no worries about destination wraps */
1308 while (match_length-- > 0) *rundest++ = *runsrc++;
1310 } /* while (togo > 0) */
1313 D(("Frame overflow, this_run = %d\n", togo))
1314 return DECR_ILLEGALDATA;
1317 memcpy(CAB(outbuf), window + ((!window_posn) ? window_size : window_posn) -
1320 QTM(window_posn) = window_posn;
1326 /* LZX decruncher */
1328 /* Microsoft's LZX document and their implementation of the
1329 * com.ms.util.cab Java package do not concur.
1331 * In the LZX document, there is a table showing the correlation between
1332 * window size and the number of position slots. It states that the 1MB
1333 * window = 40 slots and the 2MB window = 42 slots. In the implementation,
1334 * 1MB = 42 slots, 2MB = 50 slots. The actual calculation is 'find the
1335 * first slot whose position base is equal to or more than the required
1336 * window size'. This would explain why other tables in the document refer
1337 * to 50 slots rather than 42.
1339 * The constant NUM_PRIMARY_LENGTHS used in the decompression pseudocode
1340 * is not defined in the specification.
1342 * The LZX document does not state the uncompressed block has an
1343 * uncompressed length field. Where does this length field come from, so
1344 * we can know how large the block is? The implementation has it as the 24
1345 * bits following after the 3 blocktype bits, before the alignment
1348 * The LZX document states that aligned offset blocks have their aligned
1349 * offset huffman tree AFTER the main and length trees. The implementation
1350 * suggests that the aligned offset tree is BEFORE the main and length
1353 * The LZX document decoding algorithm states that, in an aligned offset
1354 * block, if an extra_bits value is 1, 2 or 3, then that number of bits
1355 * should be read and the result added to the match offset. This is
1356 * correct for 1 and 2, but not 3, where just a huffman symbol (using the
1357 * aligned tree) should be read.
1359 * Regarding the E8 preprocessing, the LZX document states 'No translation
1360 * may be performed on the last 6 bytes of the input block'. This is
1361 * correct. However, the pseudocode provided checks for the *E8 leader*
1362 * up to the last 6 bytes. If the leader appears between -10 and -7 bytes
1363 * from the end, this would cause the next four bytes to be modified, at
1364 * least one of which would be in the last 6 bytes, which is not allowed
1365 * according to the spec.
1367 * The specification states that the huffman trees must always contain at
1368 * least one element. However, many CAB files contain blocks where the
1369 * length tree is completely empty (because there are no matches), and
1370 * this is expected to succeed.
1374 /* LZX uses what it calls 'position slots' to represent match offsets.
1375 * What this means is that a small 'position slot' number and a small
1376 * offset from that slot are encoded instead of one large offset for
1378 * - lzx_position_base is an index to the position slot bases
1379 * - lzx_extra_bits states how many bits of offset-from-base data is needed.
1381 static ULONG lzx_position_base[51];
1382 static UBYTE extra_bits[51];
1384 int LZXinit(int window) {
1385 ULONG wndsize = 1 << window;
1386 int i, j, posn_slots;
1388 /* LZX supports window sizes of 2^15 (32Kb) through 2^21 (2Mb) */
1389 /* if a previously allocated window is big enough, keep it */
1390 if (window < 15 || window > 21) return DECR_DATAFORMAT;
1391 if (LZX(actual_size) < wndsize) {
1392 if (LZX(window)) free(LZX(window));
1396 if (!(LZX(window) = malloc(wndsize))) return DECR_NOMEMORY;
1397 LZX(actual_size) = wndsize;
1399 LZX(window_size) = wndsize;
1401 /* initialise static tables */
1402 for (i=0, j=0; i <= 50; i += 2) {
1403 extra_bits[i] = extra_bits[i+1] = j; /* 0,0,0,0,1,1,2,2,3,3... */
1404 if ((i != 0) && (j < 17)) j++; /* 0,0,1,2,3,4...15,16,17,17,17,17... */
1406 for (i=0, j=0; i <= 50; i++) {
1407 lzx_position_base[i] = j; /* 0,1,2,3,4,6,8,12,16,24,32,... */
1408 j += 1 << extra_bits[i]; /* 1,1,1,1,2,2,4,4,8,8,16,16,32,32,... */
1411 /* calculate required position slots */
1412 if (window == 20) posn_slots = 42;
1413 else if (window == 21) posn_slots = 50;
1414 else posn_slots = window << 1;
1416 /*posn_slots=i=0; while (i < wndsize) i += 1 << extra_bits[posn_slots++]; */
1419 LZX(R0) = LZX(R1) = LZX(R2) = 1;
1420 LZX(main_elements) = LZX_NUM_CHARS + (posn_slots << 3);
1421 LZX(header_read) = 0;
1422 LZX(frames_read) = 0;
1423 LZX(block_remaining) = 0;
1424 LZX(block_type) = LZX_BLOCKTYPE_INVALID;
1425 LZX(intel_curpos) = 0;
1426 LZX(intel_started) = 0;
1427 LZX(window_posn) = 0;
1429 /* initialise tables to 0 (because deltas will be applied to them) */
1430 for (i = 0; i < LZX_MAINTREE_MAXSYMBOLS; i++) LZX(MAINTREE_len)[i] = 0;
1431 for (i = 0; i < LZX_LENGTH_MAXSYMBOLS; i++) LZX(LENGTH_len)[i] = 0;
1436 /* Bitstream reading macros (LZX / intel little-endian byte order)
1438 * INIT_BITSTREAM should be used first to set up the system
1439 * READ_BITS(var,n) takes N bits from the buffer and puts them in var
1441 * ENSURE_BITS(n) ensures there are at least N bits in the bit buffer.
1442 * it can guarantee up to 17 bits (i.e. it can read in
1443 * 16 new bits when there is down to 1 bit in the buffer,
1444 * and it can read 32 bits when there are 0 bits in the
1446 * PEEK_BITS(n) extracts (without removing) N bits from the bit buffer
1447 * REMOVE_BITS(n) removes N bits from the bit buffer
1449 * These bit access routines work by using the area beyond the MSB and the
1450 * LSB as a free source of zeroes. This avoids having to mask any bits.
1451 * So we have to know the bit width of the bitbuffer variable.
1454 #define INIT_BITSTREAM do { bitsleft = 0; bitbuf = 0; } while (0)
1456 /* Quantum reads bytes in normal order; LZX is little-endian order */
1457 #define ENSURE_BITS(n) \
1458 while (bitsleft < (n)) { \
1459 bitbuf |= ((inpos[1]<<8)|inpos[0]) << (ULONG_BITS-16 - bitsleft); \
1460 bitsleft += 16; inpos+=2; \
1463 #define PEEK_BITS(n) (bitbuf >> (ULONG_BITS - (n)))
1464 #define REMOVE_BITS(n) ((bitbuf <<= (n)), (bitsleft -= (n)))
1466 #define READ_BITS(v,n) do { \
1469 (v) = PEEK_BITS(n); \
1477 /* Huffman macros */
1479 #define TABLEBITS(tbl) (LZX_##tbl##_TABLEBITS)
1480 #define MAXSYMBOLS(tbl) (LZX_##tbl##_MAXSYMBOLS)
1481 #define SYMTABLE(tbl) (LZX(tbl##_table))
1482 #define LENTABLE(tbl) (LZX(tbl##_len))
1484 /* BUILD_TABLE(tablename) builds a huffman lookup table from code lengths.
1485 * In reality, it just calls make_decode_table() with the appropriate
1486 * values - they're all fixed by some #defines anyway, so there's no point
1487 * writing each call out in full by hand.
1489 #define BUILD_TABLE(tbl) \
1490 if (make_decode_table( \
1491 MAXSYMBOLS(tbl), TABLEBITS(tbl), LENTABLE(tbl), SYMTABLE(tbl) \
1492 )) { return DECR_ILLEGALDATA; }
1495 /* READ_HUFFSYM(tablename, var) decodes one huffman symbol from the
1496 * bitstream using the stated table and puts it in var.
1498 #define READ_HUFFSYM(tbl,var) do { \
1500 hufftbl = SYMTABLE(tbl); \
1501 if ((i = hufftbl[PEEK_BITS(TABLEBITS(tbl))]) >= MAXSYMBOLS(tbl)) { \
1502 j = 1 << (ULONG_BITS - TABLEBITS(tbl)); \
1504 j >>= 1; i <<= 1; i |= (bitbuf & j) ? 1 : 0; \
1505 if (!j) { return DECR_ILLEGALDATA; } \
1506 } while ((i = hufftbl[i]) >= MAXSYMBOLS(tbl)); \
1508 j = LENTABLE(tbl)[(var) = i]; \
1513 /* READ_LENGTHS(tablename, first, last) reads in code lengths for symbols
1514 * first to last in the given table. The code lengths are stored in their
1515 * own special LZX way.
1517 #define READ_LENGTHS(tbl,first,last) do { \
1518 lb.bb = bitbuf; lb.bl = bitsleft; lb.ip = inpos; \
1519 if (lzx_read_lens(LENTABLE(tbl),(first),(last),&lb)) { \
1520 return DECR_ILLEGALDATA; \
1522 bitbuf = lb.bb; bitsleft = lb.bl; inpos = lb.ip; \
1526 /* make_decode_table(nsyms, nbits, length[], table[])
1528 * This function was coded by David Tritscher. It builds a fast huffman
1529 * decoding table out of just a canonical huffman code lengths table.
1531 * nsyms = total number of symbols in this huffman tree.
1532 * nbits = any symbols with a code length of nbits or less can be decoded
1533 * in one lookup of the table.
1534 * length = A table to get code lengths from [0 to syms-1]
1535 * table = The table to fill up with decoded symbols and pointers.
1537 * Returns 0 for OK or 1 for error
1540 int make_decode_table(ULONG nsyms, ULONG nbits, UBYTE *length, UWORD *table) {
1542 register ULONG leaf;
1543 register UBYTE bit_num = 1;
1545 ULONG pos = 0; /* the current position in the decode table */
1546 ULONG table_mask = 1 << nbits;
1547 ULONG bit_mask = table_mask >> 1; /* don't do 0 length codes */
1548 ULONG next_symbol = bit_mask; /* base of allocation for long codes */
1550 /* fill entries for codes short enough for a direct mapping */
1551 while (bit_num <= nbits) {
1552 for (sym = 0; sym < nsyms; sym++) {
1553 if (length[sym] == bit_num) {
1556 if((pos += bit_mask) > table_mask) return 1; /* table overrun */
1558 /* fill all possible lookups of this symbol with the symbol itself */
1560 while (fill-- > 0) table[leaf++] = sym;
1567 /* if there are any codes longer than nbits */
1568 if (pos != table_mask) {
1569 /* clear the remainder of the table */
1570 for (sym = pos; sym < table_mask; sym++) table[sym] = 0;
1572 /* give ourselves room for codes to grow by up to 16 more bits */
1577 while (bit_num <= 16) {
1578 for (sym = 0; sym < nsyms; sym++) {
1579 if (length[sym] == bit_num) {
1581 for (fill = 0; fill < bit_num - nbits; fill++) {
1582 /* if this path hasn't been taken yet, 'allocate' two entries */
1583 if (table[leaf] == 0) {
1584 table[(next_symbol << 1)] = 0;
1585 table[(next_symbol << 1) + 1] = 0;
1586 table[leaf] = next_symbol++;
1588 /* follow the path and select either left or right for next bit */
1589 leaf = table[leaf] << 1;
1590 if ((pos >> (15-fill)) & 1) leaf++;
1594 if ((pos += bit_mask) > table_mask) return 1; /* table overflow */
1603 if (pos == table_mask) return 0;
1605 /* either erroneous table, or all elements are 0 - let's find out. */
1606 for (sym = 0; sym < nsyms; sym++) if (length[sym]) return 1;
1616 int lzx_read_lens(UBYTE *lens, ULONG first, ULONG last, struct lzx_bits *lb) {
1620 register ULONG bitbuf = lb->bb;
1621 register int bitsleft = lb->bl;
1622 UBYTE *inpos = lb->ip;
1625 for (x = 0; x < 20; x++) {
1627 LENTABLE(PRETREE)[x] = y;
1629 BUILD_TABLE(PRETREE);
1631 for (x = first; x < last; ) {
1632 READ_HUFFSYM(PRETREE, z);
1634 READ_BITS(y, 4); y += 4;
1635 while (y--) lens[x++] = 0;
1638 READ_BITS(y, 5); y += 20;
1639 while (y--) lens[x++] = 0;
1642 READ_BITS(y, 1); y += 4;
1643 READ_HUFFSYM(PRETREE, z);
1644 z = lens[x] - z; if (z < 0) z += 17;
1645 while (y--) lens[x++] = z;
1648 z = lens[x] - z; if (z < 0) z += 17;
1659 int LZXdecompress(int inlen, int outlen) {
1660 UBYTE *inpos = CAB(inbuf);
1661 UBYTE *endinp = inpos + inlen;
1662 UBYTE *window = LZX(window);
1663 UBYTE *runsrc, *rundest;
1664 UWORD *hufftbl; /* used in READ_HUFFSYM macro as chosen decoding table */
1666 ULONG window_posn = LZX(window_posn);
1667 ULONG window_size = LZX(window_size);
1672 register ULONG bitbuf;
1673 register int bitsleft;
1674 ULONG match_offset, i,j,k; /* ijk used in READ_HUFFSYM macro */
1675 struct lzx_bits lb; /* used in READ_LENGTHS macro */
1677 int togo = outlen, this_run, main_element, aligned_bits;
1678 int match_length, copy_length, length_footer, extra, verbatim_bits;
1682 /* read header if necessary */
1683 if (!LZX(header_read)) {
1685 READ_BITS(k, 1); if (k) { READ_BITS(i,16); READ_BITS(j,16); }
1686 LZX(intel_filesize) = (i << 16) | j; /* or 0 if not encoded */
1687 LZX(header_read) = 1;
1690 /* main decoding loop */
1692 /* last block finished, new block expected */
1693 if (LZX(block_remaining) == 0) {
1694 if (LZX(block_type) == LZX_BLOCKTYPE_UNCOMPRESSED) {
1695 if (LZX(block_length) & 1) inpos++; /* realign bitstream to word */
1699 READ_BITS(LZX(block_type), 3);
1702 LZX(block_remaining) = LZX(block_length) = (i << 8) | j;
1704 switch (LZX(block_type)) {
1705 case LZX_BLOCKTYPE_ALIGNED:
1706 for (i = 0; i < 8; i++) { READ_BITS(j, 3); LENTABLE(ALIGNED)[i] = j; }
1707 BUILD_TABLE(ALIGNED);
1708 /* rest of aligned header is same as verbatim */
1710 case LZX_BLOCKTYPE_VERBATIM:
1711 READ_LENGTHS(MAINTREE, 0, 256);
1712 READ_LENGTHS(MAINTREE, 256, LZX(main_elements));
1713 BUILD_TABLE(MAINTREE);
1714 if (LENTABLE(MAINTREE)[0xE8] != 0) LZX(intel_started) = 1;
1716 READ_LENGTHS(LENGTH, 0, LZX_NUM_SECONDARY_LENGTHS);
1717 BUILD_TABLE(LENGTH);
1720 case LZX_BLOCKTYPE_UNCOMPRESSED:
1721 LZX(intel_started) = 1; /* because we can't assume otherwise */
1722 ENSURE_BITS(16); /* get up to 16 pad bits into the buffer */
1723 if (bitsleft > 16) inpos -= 2; /* and align the bitstream! */
1724 R0 = inpos[0]|(inpos[1]<<8)|(inpos[2]<<16)|(inpos[3]<<24);inpos+=4;
1725 R1 = inpos[0]|(inpos[1]<<8)|(inpos[2]<<16)|(inpos[3]<<24);inpos+=4;
1726 R2 = inpos[0]|(inpos[1]<<8)|(inpos[2]<<16)|(inpos[3]<<24);inpos+=4;
1730 return DECR_ILLEGALDATA;
1734 /* buffer exhaustion check */
1735 if (inpos > endinp) {
1736 /* it's possible to have a file where the next run is less than
1737 * 16 bits in size. In this case, the READ_HUFFSYM() macro used
1738 * in building the tables will exhaust the buffer, so we should
1739 * allow for this, but not allow those accidentally read bits to
1740 * be used (so we check that there are at least 16 bits
1741 * remaining - in this boundary case they aren't really part of
1742 * the compressed data)
1744 if (inpos > (endinp+2) || bitsleft < 16) return DECR_ILLEGALDATA;
1747 while ((this_run = LZX(block_remaining)) > 0 && togo > 0) {
1748 if (this_run > togo) this_run = togo;
1750 LZX(block_remaining) -= this_run;
1752 /* apply 2^x-1 mask */
1753 window_posn &= window_size - 1;
1754 /* runs can't straddle the window wraparound */
1755 if ((window_posn + this_run) > window_size)
1756 return DECR_DATAFORMAT;
1758 switch (LZX(block_type)) {
1760 case LZX_BLOCKTYPE_VERBATIM:
1761 while (this_run > 0) {
1762 READ_HUFFSYM(MAINTREE, main_element);
1764 if (main_element < LZX_NUM_CHARS) {
1765 /* literal: 0 to LZX_NUM_CHARS-1 */
1766 window[window_posn++] = main_element;
1770 /* match: LZX_NUM_CHARS + ((slot<<3) | length_header (3 bits)) */
1771 main_element -= LZX_NUM_CHARS;
1773 match_length = main_element & LZX_NUM_PRIMARY_LENGTHS;
1774 if (match_length == LZX_NUM_PRIMARY_LENGTHS) {
1775 READ_HUFFSYM(LENGTH, length_footer);
1776 match_length += length_footer;
1778 match_length += LZX_MIN_MATCH;
1780 match_offset = main_element >> 3;
1782 if (match_offset > 2) {
1783 /* not repeated offset */
1784 if (match_offset != 3) {
1785 extra = extra_bits[match_offset];
1786 READ_BITS(verbatim_bits, extra);
1787 match_offset = lzx_position_base[match_offset]
1788 - 2 + verbatim_bits;
1794 /* update repeated offset LRU queue */
1795 R2 = R1; R1 = R0; R0 = match_offset;
1797 else if (match_offset == 0) {
1800 else if (match_offset == 1) {
1802 R1 = R0; R0 = match_offset;
1804 else /* match_offset == 2 */ {
1806 R2 = R0; R0 = match_offset;
1809 rundest = window + window_posn;
1810 this_run -= match_length;
1812 /* copy any wrapped around source data */
1813 if (window_posn >= match_offset) {
1815 runsrc = rundest - match_offset;
1817 runsrc = rundest + (window_size - match_offset);
1818 copy_length = match_offset - window_posn;
1819 if (copy_length < match_length) {
1820 match_length -= copy_length;
1821 window_posn += copy_length;
1822 while (copy_length-- > 0) *rundest++ = *runsrc++;
1826 window_posn += match_length;
1828 /* copy match data - no worries about destination wraps */
1829 while (match_length-- > 0) *rundest++ = *runsrc++;
1834 case LZX_BLOCKTYPE_ALIGNED:
1835 while (this_run > 0) {
1836 READ_HUFFSYM(MAINTREE, main_element);
1838 if (main_element < LZX_NUM_CHARS) {
1839 /* literal: 0 to LZX_NUM_CHARS-1 */
1840 window[window_posn++] = main_element;
1844 /* match: LZX_NUM_CHARS + ((slot<<3) | length_header (3 bits)) */
1845 main_element -= LZX_NUM_CHARS;
1847 match_length = main_element & LZX_NUM_PRIMARY_LENGTHS;
1848 if (match_length == LZX_NUM_PRIMARY_LENGTHS) {
1849 READ_HUFFSYM(LENGTH, length_footer);
1850 match_length += length_footer;
1852 match_length += LZX_MIN_MATCH;
1854 match_offset = main_element >> 3;
1856 if (match_offset > 2) {
1857 /* not repeated offset */
1858 extra = extra_bits[match_offset];
1859 match_offset = lzx_position_base[match_offset] - 2;
1861 /* verbatim and aligned bits */
1863 READ_BITS(verbatim_bits, extra);
1864 match_offset += (verbatim_bits << 3);
1865 READ_HUFFSYM(ALIGNED, aligned_bits);
1866 match_offset += aligned_bits;
1868 else if (extra == 3) {
1869 /* aligned bits only */
1870 READ_HUFFSYM(ALIGNED, aligned_bits);
1871 match_offset += aligned_bits;
1873 else if (extra > 0) { /* extra==1, extra==2 */
1874 /* verbatim bits only */
1875 READ_BITS(verbatim_bits, extra);
1876 match_offset += verbatim_bits;
1878 else /* extra == 0 */ {
1883 /* update repeated offset LRU queue */
1884 R2 = R1; R1 = R0; R0 = match_offset;
1886 else if (match_offset == 0) {
1889 else if (match_offset == 1) {
1891 R1 = R0; R0 = match_offset;
1893 else /* match_offset == 2 */ {
1895 R2 = R0; R0 = match_offset;
1898 rundest = window + window_posn;
1899 this_run -= match_length;
1901 /* copy any wrapped around source data */
1902 if (window_posn >= match_offset) {
1904 runsrc = rundest - match_offset;
1906 runsrc = rundest + (window_size - match_offset);
1907 copy_length = match_offset - window_posn;
1908 if (copy_length < match_length) {
1909 match_length -= copy_length;
1910 window_posn += copy_length;
1911 while (copy_length-- > 0) *rundest++ = *runsrc++;
1915 window_posn += match_length;
1917 /* copy match data - no worries about destination wraps */
1918 while (match_length-- > 0) *rundest++ = *runsrc++;
1923 case LZX_BLOCKTYPE_UNCOMPRESSED:
1924 if ((inpos + this_run) > endinp) return DECR_ILLEGALDATA;
1925 memcpy(window + window_posn, inpos, (size_t) this_run);
1926 inpos += this_run; window_posn += this_run;
1930 return DECR_ILLEGALDATA; /* might as well */
1936 if (togo != 0) return DECR_ILLEGALDATA;
1937 memcpy(CAB(outbuf), window + ((!window_posn) ? window_size : window_posn) -
1938 outlen, (size_t) outlen);
1940 LZX(window_posn) = window_posn;
1945 /* intel E8 decoding */
1946 if ((LZX(frames_read)++ < 32768) && LZX(intel_filesize) != 0) {
1947 if (outlen <= 6 || !LZX(intel_started)) {
1948 LZX(intel_curpos) += outlen;
1951 UBYTE *data = CAB(outbuf);
1952 UBYTE *dataend = data + outlen - 10;
1953 LONG curpos = LZX(intel_curpos);
1954 LONG filesize = LZX(intel_filesize);
1955 LONG abs_off, rel_off;
1957 LZX(intel_curpos) = curpos + outlen;
1959 while (data < dataend) {
1960 if (*data++ != 0xE8) { curpos++; continue; }
1961 abs_off = data[0] | (data[1]<<8) | (data[2]<<16) | (data[3]<<24);
1962 if ((abs_off >= -curpos) && (abs_off < filesize)) {
1963 rel_off = (abs_off >= 0) ? abs_off - curpos : abs_off + filesize;
1964 data[0] = (UBYTE) rel_off;
1965 data[1] = (UBYTE) (rel_off >> 8);
1966 data[2] = (UBYTE) (rel_off >> 16);
1967 data[3] = (UBYTE) (rel_off >> 24);
1980 /* all the file IO is abstracted into these routines:
1981 * cabinet_(open|close|read|seek|skip|getoffset)
1982 * file_(open|close|write)
1985 /* ensure_filepath("a/b/c/d.txt") ensures a, a/b and a/b/c exist as dirs */
1986 int ensure_filepath(char *path) {
1992 m = umask(0); umask(m); /* obtain user's umask */
1994 for (p = path; *p; p++) {
1995 if ((p != path) && (*p == '/')) {
1997 ok = (stat(path, &st_buf) == 0) && S_ISDIR(st_buf.st_mode);
1998 if (!ok) ok = (mkdir(path, 0777 & ~m) == 0);
2006 /* opens a file for output, returns success */
2007 int file_open(struct file *fi, int lower, char *dir) {
2008 char c, *s, *d, *name;
2011 if (!(name = malloc(strlen(fi->filename) + (dir ? strlen(dir) : 0) + 2))) {
2012 fprintf(stderr, "out of memory!\n");
2016 /* start with blank name */
2019 /* add output directory if needed */
2026 /* remove leading slashes */
2028 while (*s == '\\') s++;
2030 /* copy from fi->filename to new name, converting MS-DOS slashes to UNIX
2031 * slashes as we go. Also lowercases characters if needed.
2033 d = &name[strlen(name)];
2036 *d++ = (c=='/') ? '\\' : ((c=='\\') ? '/' :
2037 (lower ? tolower((unsigned char) c) : c));
2040 /* create directories if needed, attempt to write file */
2041 if (ensure_filepath(name)) {
2042 fi->fh = fopen(name, "wb");
2046 /* as full filename is no longer needed, free it */
2050 perror(fi->filename);
2055 /* closes a completed file, updates protections and timestamp */
2056 void file_close(struct file *fi) {
2066 m = umask(0); umask(m); /* obtain user's umask */
2069 | (fi->attribs & cffile_A_EXEC ? 0111 : 0)
2070 | (fi->attribs & cffile_A_RDONLY ? 0 : 0222)) & ~ m
2074 time.tm_sec = (fi->time << 1) & 0x3e;
2075 time.tm_min = (fi->time >> 5) & 0x3f;
2076 time.tm_hour = (fi->time >> 11);
2077 time.tm_mday = fi->date & 0x1f;
2078 time.tm_mon =((fi->date >> 5) & 0xf) - 1;
2079 time.tm_year = (fi->date >> 9) + 80;
2082 utb.actime = utb.modtime = mktime(&time);
2083 utime(fi->filename, &utb);
2087 int file_write(struct file *fi, UBYTE *buf, size_t length) {
2088 if (fwrite((void *)buf, 1, length, fi->fh) != length) {
2089 perror(fi->filename);
2096 void cabinet_close(struct cabinet *cab) {
2103 void cabinet_seek(struct cabinet *cab, off_t offset) {
2104 if (fseek(cab->fh, offset, SEEK_SET) < 0) {
2105 perror(cab->filename);
2109 void cabinet_skip(struct cabinet *cab, off_t distance) {
2110 if (fseek(cab->fh, distance, SEEK_CUR) < 0) {
2111 perror(cab->filename);
2115 off_t cabinet_getoffset(struct cabinet *cab) {
2116 return ftell(cab->fh);
2119 /* read data from a cabinet, returns success */
2120 int cabinet_read(struct cabinet *cab, UBYTE *buf, size_t length) {
2121 size_t avail = (size_t) (cab->filelen - cabinet_getoffset(cab));
2122 if (length > avail) {
2123 fprintf(stderr, "%s: WARNING; cabinet is truncated\n", cab->filename);
2126 if (fread((void *)buf, 1, length, cab->fh) != length) {
2127 perror(cab->filename);
2133 /* try to open a cabinet file, returns success */
2134 int cabinet_open(struct cabinet *cab) {
2135 char *name = cab->filename;
2138 /* note: this is now case sensitive */
2139 if (!(fh = fopen(name, "rb"))) {
2144 /* seek to end of file */
2145 if (fseek(fh, 0, SEEK_END) != 0) {
2151 /* get length of file */
2152 cab->filelen = ftell(fh);
2154 /* return to the start of the file */
2155 if (fseek(fh, 0, SEEK_SET) != 0) {
2165 /* allocate and read an aribitrarily long string from the cabinet */
2166 char *cabinet_read_string(struct cabinet *cab) {
2167 off_t len=256, base = cabinet_getoffset(cab), maxlen = cab->filelen - base;
2171 if (len > maxlen) len = maxlen;
2172 if (!(buf = realloc(buf, (size_t) len))) break;
2173 if (!cabinet_read(cab, buf, (size_t) len)) break;
2175 /* search for a null terminator in what we've just read */
2176 for (i=0; i < len; i++) {
2177 if (!buf[i]) {ok=1; break;}
2181 if (len == maxlen) {
2182 fprintf(stderr, "%s: WARNING; cabinet is truncated\n", cab->filename);
2186 cabinet_seek(cab, base);
2191 if (buf) free(buf); else fprintf(stderr, "out of memory!\n");
2195 /* otherwise, set the stream to just after the string and return */
2196 cabinet_seek(cab, base + ((off_t) strlen((char *) buf)) + 1);
2197 return (char *) buf;
2200 /* reads the header and all folder and file entries in this cabinet */
2201 int cabinet_read_entries(struct cabinet *cab) {
2202 int num_folders, num_files, header_resv, folder_resv = 0, i;
2203 struct folder *fol, *linkfol = NULL;
2204 struct file *file, *linkfile = NULL;
2208 /* read in the CFHEADER */
2209 base_offset = cabinet_getoffset(cab);
2210 if (!cabinet_read(cab, buf, cfhead_SIZEOF)) {
2214 /* check basic MSCF signature */
2215 if (EndGetI32(buf+cfhead_Signature) != 0x4643534d) {
2216 fprintf(stderr, "%s: not a Microsoft cabinet file\n", cab->filename);
2220 /* get the number of folders */
2221 num_folders = EndGetI16(buf+cfhead_NumFolders);
2222 if (num_folders == 0) {
2223 fprintf(stderr, "%s: no folders in cabinet\n", cab->filename);
2227 /* get the number of files */
2228 num_files = EndGetI16(buf+cfhead_NumFiles);
2229 if (num_files == 0) {
2230 fprintf(stderr, "%s: no files in cabinet\n", cab->filename);
2234 /* just check the header revision */
2235 if ((buf[cfhead_MajorVersion] > 1) ||
2236 (buf[cfhead_MajorVersion] == 1 && buf[cfhead_MinorVersion] > 3))
2238 fprintf(stderr, "%s: WARNING; cabinet format version > 1.3\n",
2242 /* read the reserved-sizes part of header, if present */
2243 cab->flags = EndGetI16(buf+cfhead_Flags);
2244 if (cab->flags & cfheadRESERVE_PRESENT) {
2245 if (!cabinet_read(cab, buf, cfheadext_SIZEOF)) return 0;
2246 header_resv = EndGetI16(buf+cfheadext_HeaderReserved);
2247 folder_resv = buf[cfheadext_FolderReserved];
2248 cab->block_resv = buf[cfheadext_DataReserved];
2250 if (header_resv > 60000) {
2251 fprintf(stderr, "%s: WARNING; header reserved space > 60000\n",
2255 /* skip the reserved header */
2256 if (header_resv) fseek(cab->fh, (off_t) header_resv, SEEK_CUR);
2259 if (cab->flags & cfheadPREV_CABINET) {
2260 cab->prevname = cabinet_read_string(cab);
2261 if (!cab->prevname) return 0;
2262 cab->previnfo = cabinet_read_string(cab);
2265 if (cab->flags & cfheadNEXT_CABINET) {
2266 cab->nextname = cabinet_read_string(cab);
2267 if (!cab->nextname) return 0;
2268 cab->nextinfo = cabinet_read_string(cab);
2272 for (i = 0; i < num_folders; i++) {
2273 if (!cabinet_read(cab, buf, cffold_SIZEOF)) return 0;
2274 if (folder_resv) cabinet_skip(cab, folder_resv);
2276 fol = (struct folder *) calloc(1, sizeof(struct folder));
2277 if (!fol) { fprintf(stderr, "out of memory!\n"); return 0; }
2280 fol->offset[0] = base_offset + (off_t) EndGetI32(buf+cffold_DataOffset);
2281 fol->num_blocks = EndGetI16(buf+cffold_NumBlocks);
2282 fol->comp_type = EndGetI16(buf+cffold_CompType);
2284 if (!linkfol) cab->folders = fol; else linkfol->next = fol;
2289 for (i = 0; i < num_files; i++) {
2290 if (!cabinet_read(cab, buf, cffile_SIZEOF)) return 0;
2291 file = (struct file *) calloc(1, sizeof(struct file));
2292 if (!file) { fprintf(stderr, "out of memory!\n"); return 0; }
2294 file->length = EndGetI32(buf+cffile_UncompressedSize);
2295 file->offset = EndGetI32(buf+cffile_FolderOffset);
2296 file->index = EndGetI16(buf+cffile_FolderIndex);
2297 file->time = EndGetI16(buf+cffile_Time);
2298 file->date = EndGetI16(buf+cffile_Date);
2299 file->attribs = EndGetI16(buf+cffile_Attribs);
2300 file->filename = cabinet_read_string(cab);
2301 if (!file->filename) return 0;
2302 if (!linkfile) cab->files = file; else linkfile->next = file;
2309 /* this does the tricky job of running through every file in the cabinet,
2310 * including spanning cabinets, and working out which file is in which
2311 * folder in which cabinet. It also throws out the duplicate file entries
2312 * that appear in spanning cabinets. There is memory leakage here because
2313 * those entries are not freed. See the XAD CAB client for an
2314 * implementation of this that correctly frees the discarded file entries.
2316 struct file *process_files(struct cabinet *basecab) {
2317 struct cabinet *cab;
2318 struct file *outfi = NULL, *linkfi = NULL, *nextfi, *fi, *cfi;
2319 struct folder *fol, *firstfol, *lastfol = NULL, *predfol;
2322 for (cab = basecab; cab; cab = cab->nextcab) {
2323 /* firstfol = first folder in this cabinet */
2324 /* lastfol = last folder in this cabinet */
2325 /* predfol = last folder in previous cabinet (or NULL if first cabinet) */
2327 firstfol = cab->folders;
2328 for (lastfol = firstfol; lastfol->next;) lastfol = lastfol->next;
2331 for (fi = cab->files; fi; fi = nextfi) {
2335 if (i < cffileCONTINUED_FROM_PREV) {
2336 for (fol = firstfol; fol && i--; ) fol = fol->next;
2337 fi->folder = fol; /* NULL if an invalid folder index */
2340 /* folder merging */
2341 if (i == cffileCONTINUED_TO_NEXT
2342 || i == cffileCONTINUED_PREV_AND_NEXT) {
2343 if (cab->nextcab && !lastfol->contfile) lastfol->contfile = fi;
2346 if (i == cffileCONTINUED_FROM_PREV
2347 || i == cffileCONTINUED_PREV_AND_NEXT) {
2348 /* these files are to be continued in yet another
2349 * cabinet, don't merge them in just yet */
2350 if (i == cffileCONTINUED_PREV_AND_NEXT) mergeok = 0;
2352 /* only merge once per cabinet */
2354 if ((cfi = predfol->contfile)
2355 && (cfi->offset == fi->offset)
2356 && (cfi->length == fi->length)
2357 && (strcmp(cfi->filename, fi->filename) == 0)
2358 && (predfol->comp_type == firstfol->comp_type)) {
2359 /* increase the number of splits */
2360 if ((i = ++(predfol->num_splits)) > CAB_SPLITMAX) {
2362 fprintf(stderr, "%s: internal error, increase CAB_SPLITMAX\n",
2366 /* copy information across from the merged folder */
2367 predfol->offset[i] = firstfol->offset[0];
2368 predfol->cab[i] = firstfol->cab[0];
2369 predfol->next = firstfol->next;
2370 predfol->contfile = firstfol->contfile;
2372 if (firstfol == lastfol) lastfol = predfol;
2374 predfol = NULL; /* don't merge again within this cabinet */
2378 /* if the folders won't merge, don't add their files */
2383 if (mergeok) fi->folder = firstfol;
2388 if (linkfi) linkfi->next = fi; else outfi = fi;
2392 } /* for (cab= ...*/
2397 /* validates and reads file entries from a cabinet at offset [offset] in
2398 * file [name]. Returns a cabinet structure if successful, or NULL
2401 struct cabinet *load_cab_offset(char *name, off_t offset) {
2402 struct cabinet *cab = (struct cabinet *) calloc(1, sizeof(struct cabinet));
2404 if (!cab) return NULL;
2406 cab->filename = name;
2407 if ((ok = cabinet_open(cab))) {
2408 cabinet_seek(cab, offset);
2409 ok = cabinet_read_entries(cab);
2418 /* Searches a file for embedded cabinets (also succeeds on just normal
2419 * cabinet files). The first result of this search will be returned, and
2420 * the remaining results will be chained to it via the cab->next structure
2423 #define SEARCH_SIZE (32*1024)
2424 UBYTE search_buf[SEARCH_SIZE];
2426 struct cabinet *find_cabs_in_file(char *name) {
2427 struct cabinet *cab, *cab2, *firstcab = NULL, *linkcab = NULL;
2428 UBYTE *pstart = &search_buf[0], *pend, *p;
2429 ULONG offset, caboff, cablen, foffset, filelen;
2431 int state = 0, found = 0, ok = 0;
2433 /* open the file and search for cabinet headers */
2434 if ((cab = (struct cabinet *) calloc(1, sizeof(struct cabinet)))) {
2435 cab->filename = name;
2436 if (cabinet_open(cab)) {
2437 filelen = (ULONG) cab->filelen;
2438 for (offset = 0; offset < filelen; offset += length) {
2439 /* search length is either the full length of the search buffer,
2440 * or the amount of data remaining to the end of the file,
2441 * whichever is less.
2443 length = filelen - offset;
2444 if (length > SEARCH_SIZE) length = SEARCH_SIZE;
2446 /* fill the search buffer with data from disk */
2447 if (!cabinet_read(cab, search_buf, length)) break;
2449 /* read through the entire buffer. */
2451 pend = &search_buf[length];
2454 /* starting state */
2456 /* we spend most of our time in this while loop, looking for
2457 * a leading 'M' of the 'MSCF' signature
2459 while (*p++ != 0x4D && p < pend);
2460 if (p < pend) state = 1; /* if we found tht 'M', advance state */
2463 /* verify that the next 3 bytes are 'S', 'C' and 'F' */
2464 case 1: state = (*p++ == 0x53) ? 2 : 0; break;
2465 case 2: state = (*p++ == 0x43) ? 3 : 0; break;
2466 case 3: state = (*p++ == 0x46) ? 4 : 0; break;
2468 /* we don't care about bytes 4-7 */
2469 /* bytes 8-11 are the overall length of the cabinet */
2470 case 8: cablen = *p++; state++; break;
2471 case 9: cablen |= *p++ << 8; state++; break;
2472 case 10: cablen |= *p++ << 16; state++; break;
2473 case 11: cablen |= *p++ << 24; state++; break;
2475 /* we don't care about bytes 12-15 */
2476 /* bytes 16-19 are the offset within the cabinet of the filedata */
2477 case 16: foffset = *p++; state++; break;
2478 case 17: foffset |= *p++ << 8; state++; break;
2479 case 18: foffset |= *p++ << 16; state++; break;
2480 case 19: foffset |= *p++ << 24;
2481 /* now we have recieved 20 bytes of potential cab header. */
2482 /* work out the offset in the file of this potential cabinet */
2483 caboff = offset + (p-pstart) - 20;
2485 /* check that the files offset is less than the alleged length
2486 * of the cabinet, and that the offset + the alleged length are
2487 * 'roughly' within the end of overall file length
2489 if ((foffset < cablen) &&
2490 ((caboff + foffset) < (filelen + 32)) &&
2491 ((caboff + cablen) < (filelen + 32)) )
2493 /* found a potential result - try loading it */
2495 cab2 = load_cab_offset(name, (off_t) caboff);
2500 /* cause the search to restart after this cab's data. */
2501 offset = caboff + cablen;
2502 if (offset < cab->filelen) cabinet_seek(cab, offset);
2506 /* link the cab into the list */
2507 if (linkcab == NULL) firstcab = cab2;
2508 else linkcab->next = cab2;
2515 p++, state++; break;
2524 /* if there were cabinets that were found but are not ok, point this out */
2526 fprintf(stderr, "%s: WARNING; found %d bad cabinets\n", name, found-ok);
2529 /* if no cabinets were found, let the user know */
2531 fprintf(stderr, "%s: not a Microsoft cabinet file.\n", name);
2536 /* UTF translates two-byte unicode characters into 1, 2 or 3 bytes.
2537 * %000000000xxxxxxx -> %0xxxxxxx
2538 * %00000xxxxxyyyyyy -> %110xxxxx %10yyyyyy
2539 * %xxxxyyyyyyzzzzzz -> %1110xxxx %10yyyyyy %10zzzzzz
2541 * Therefore, the inverse is as follows:
2543 * 0x00 - 0x7F = one byte char
2544 * 0x80 - 0xBF = invalid
2545 * 0xC0 - 0xDF = 2 byte char (next char only 0x80-0xBF is valid)
2546 * 0xE0 - 0xEF = 3 byte char (next 2 chars only 0x80-0xBF is valid)
2547 * 0xF0 - 0xFF = invalid
2550 /* translate UTF -> ASCII */
2551 int convertUTF(UBYTE *in) {
2552 UBYTE c, *out = in, *end = in + strlen((char *) in) + 1;
2556 /* read unicode character */
2557 if ((c = *in++) < 0x80) x = c;
2559 if (c < 0xC0) return 0;
2560 else if (c < 0xE0) {
2561 x = (c & 0x1F) << 6;
2562 if ((c = *in++) < 0x80 || c > 0xBF) return 0; else x |= (c & 0x3F);
2564 else if (c < 0xF0) {
2565 x = (c & 0xF) << 12;
2566 if ((c = *in++) < 0x80 || c > 0xBF) return 0; else x |= (c & 0x3F)<<6;
2567 if ((c = *in++) < 0x80 || c > 0xBF) return 0; else x |= (c & 0x3F);
2572 /* terrible unicode -> ASCII conversion */
2573 if (x > 127) x = '_';
2575 if (in > end) return 0; /* just in case */
2576 } while ((*out++ = (UBYTE) x));
2580 void print_fileinfo(struct file *fi) {
2581 int d = fi->date, t = fi->time;
2584 if (fi->attribs & cffile_A_NAME_IS_UTF) {
2585 fname = malloc(strlen(fi->filename) + 1);
2587 strcpy(fname, fi->filename);
2588 convertUTF((UBYTE *) fname);
2592 printf("%9u | %02d.%02d.%04d %02d:%02d:%02d | %s\n",
2594 d & 0x1f, (d>>5) & 0xf, (d>>9) + 1980,
2595 t >> 11, (t>>5) & 0x3f, (t << 1) & 0x3e,
2596 fname ? fname : fi->filename
2599 if (fname) free(fname);
2602 int NONEdecompress(int inlen, int outlen) {
2603 if (inlen != outlen) return DECR_ILLEGALDATA;
2604 memcpy(CAB(outbuf), CAB(inbuf), (size_t) inlen);
2608 ULONG checksum(UBYTE *data, UWORD bytes, ULONG csum) {
2612 for (len = bytes >> 2; len--; data += 4) {
2613 csum ^= ((data[0]) | (data[1]<<8) | (data[2]<<16) | (data[3]<<24));
2616 switch (bytes & 3) {
2617 case 3: ul |= *data++ << 16;
2618 case 2: ul |= *data++ << 8;
2619 case 1: ul |= *data;
2626 int decompress(struct file *fi, int savemode, int fix) {
2627 ULONG bytes = savemode ? fi->length : fi->offset - CAB(offset);
2628 struct cabinet *cab = CAB(current)->cab[CAB(split)];
2629 UBYTE buf[cfdata_SIZEOF], *data;
2630 UWORD inlen, len, outlen, cando;
2635 /* cando = the max number of bytes we can do */
2636 cando = CAB(outlen);
2637 if (cando > bytes) cando = bytes;
2640 if (cando && savemode) file_write(fi, CAB(outpos), cando);
2642 CAB(outpos) += cando;
2643 CAB(outlen) -= cando;
2644 bytes -= cando; if (!bytes) break;
2646 /* we only get here if we emptied the output buffer */
2648 /* read data header + data */
2650 while (outlen == 0) {
2651 /* read the block header, skip the reserved part */
2652 if (!cabinet_read(cab, buf, cfdata_SIZEOF)) return DECR_INPUT;
2653 cabinet_skip(cab, cab->block_resv);
2655 /* we shouldn't get blocks over CAB_INPUTMAX in size */
2656 data = CAB(inbuf) + inlen;
2657 len = EndGetI16(buf+cfdata_CompressedSize);
2659 if (inlen > CAB_INPUTMAX) return DECR_INPUT;
2660 if (!cabinet_read(cab, data, len)) return DECR_INPUT;
2662 /* clear two bytes after read-in data */
2663 data[len+1] = data[len+2] = 0;
2665 /* perform checksum test on the block (if one is stored) */
2666 cksum = EndGetI32(buf+cfdata_CheckSum);
2667 if (cksum && cksum != checksum(buf+4, 4, checksum(data, len, 0))) {
2668 /* checksum is wrong */
2669 if (fix && ((fi->folder->comp_type & cffoldCOMPTYPE_MASK)
2670 == cffoldCOMPTYPE_MSZIP))
2672 fprintf(stderr, "%s: WARNING; checksum failed\n", fi->filename);
2675 return DECR_CHECKSUM;
2679 /* outlen=0 means this block was part of a split block */
2680 outlen = EndGetI16(buf+cfdata_UncompressedSize);
2683 cab = CAB(current)->cab[++CAB(split)];
2684 if (!cabinet_open(cab)) return DECR_INPUT;
2685 cabinet_seek(cab, CAB(current)->offset[CAB(split)]);
2689 /* decompress block */
2690 if ((err = CAB(decompress)(inlen, outlen))) {
2691 if (fix && ((fi->folder->comp_type & cffoldCOMPTYPE_MASK)
2692 == cffoldCOMPTYPE_MSZIP))
2694 fprintf(stderr, "%s: WARNING; failed decrunching block\n",
2701 CAB(outlen) = outlen;
2702 CAB(outpos) = CAB(outbuf);
2709 void extract_file(struct file *fi, int lower, int fix, char *dir) {
2710 struct folder *fol = fi->folder, *oldfol = CAB(current);
2713 /* is a change of folder needed? do we need to reset the current folder? */
2714 if (fol != oldfol || fi->offset < CAB(offset)) {
2715 UWORD comptype = fol->comp_type;
2716 int ct1 = comptype & cffoldCOMPTYPE_MASK;
2717 int ct2 = oldfol ? (oldfol->comp_type & cffoldCOMPTYPE_MASK) : 0;
2719 /* if the archiver has changed, call the old archiver's free() function */
2722 case cffoldCOMPTYPE_LZX:
2728 case cffoldCOMPTYPE_QUANTUM:
2738 case cffoldCOMPTYPE_NONE:
2739 CAB(decompress) = NONEdecompress;
2742 case cffoldCOMPTYPE_MSZIP:
2743 CAB(decompress) = ZIPdecompress;
2746 case cffoldCOMPTYPE_QUANTUM:
2747 CAB(decompress) = QTMdecompress;
2748 err = QTMinit((comptype >> 8) & 0x1f, (comptype >> 4) & 0xF);
2751 case cffoldCOMPTYPE_LZX:
2752 CAB(decompress) = LZXdecompress;
2753 err = LZXinit((comptype >> 8) & 0x1f);
2757 err = DECR_DATAFORMAT;
2759 if (err) goto exit_handler;
2761 /* initialisation OK, set current folder and reset offset */
2762 if (oldfol) cabinet_close(oldfol->cab[CAB(split)]);
2763 if (!cabinet_open(fol->cab[0])) goto exit_handler;
2764 cabinet_seek(fol->cab[0], fol->offset[0]);
2767 CAB(outlen) = 0; /* discard existing block */
2771 if (fi->offset > CAB(offset)) {
2772 /* decode bytes and send them to /dev/null */
2773 if ((err = decompress(fi, 0, fix))) goto exit_handler;
2774 CAB(offset) = fi->offset;
2776 if (!file_open(fi, lower, dir)) return;
2777 err = decompress(fi, 1, fix);
2778 if (err) CAB(current) = NULL; else CAB(offset) += fi->length;
2783 char *errmsg, *cabname;
2786 errmsg = "out of memory!\n"; break;
2787 case DECR_ILLEGALDATA:
2788 errmsg = "%s: illegal or corrupt data\n"; break;
2789 case DECR_DATAFORMAT:
2790 errmsg = "%s: unsupported data format\n"; break;
2792 errmsg = "%s: checksum error\n"; break;
2794 errmsg = "%s: input error\n"; break;
2796 errmsg = "%s: output error\n"; break;
2798 errmsg = "%s: unknown error (BUG)\n";
2802 cabname = CAB(current)->cab[CAB(split)]->filename;
2805 cabname = fi->folder->cab[0]->filename;
2808 fprintf(stderr, errmsg, cabname);
2812 /* tries to find *cabname, from the directory path of origcab, correcting the
2813 * case of *cabname if necessary, If found, writes back to *cabname.
2815 void find_cabinet_file(char **cabname, char *origcab) {
2816 char *tail, *cab, *name, *nextpart;
2817 struct dirent *entry;
2822 /* ensure we have a cabinet name at all */
2823 if (!(name = *cabname)) return;
2825 /* find if there's a directory path in the origcab */
2826 tail = origcab ? strrchr(origcab, '/') : NULL;
2828 if ((cab = (char *) malloc((tail ? tail-origcab : 1) + strlen(name) + 2))) {
2829 /* add the directory path from the original cabinet name */
2831 memcpy(cab, origcab, tail-origcab);
2832 cab[tail-origcab] = '\0';
2835 /* default directory path of '.' */
2841 /* we don't want null cabinet filenames */
2842 if (name[0] == '\0') break;
2844 /* if there is a directory component in the cabinet name,
2845 * look for that alone first
2847 nextpart = strchr(name, '\\');
2848 if (nextpart) *nextpart = '\0';
2850 /* try accessing the component with its current name (case-sensitive) */
2851 len = strlen(cab); strcat(cab, "/"); strcat(cab, name);
2852 found = (stat(cab, &st_buf) == 0) &&
2853 nextpart ? S_ISDIR(st_buf.st_mode) : S_ISREG(st_buf.st_mode);
2855 /* if the component was not found, look for it in the current dir */
2858 if ((dir = opendir(cab))) {
2859 while ((entry = readdir(dir))) {
2860 if (strcasecmp(name, entry->d_name) == 0) {
2861 strcat(cab, "/"); strcat(cab, entry->d_name); found = 1;
2868 /* restore the real name and skip to the next directory component
2869 * or actual cabinet name
2871 if (nextpart) *nextpart = '\\', name = &nextpart[1];
2873 /* while there is another directory component, and while we
2874 * successfully found the current component
2876 } while (nextpart && found);
2879 /* if we found the cabinet, change the next cabinet's name.
2880 * otherwise, pretend nothing happened
2883 free((void *) *cabname);
2892 /* process_cabinet() is called by main() for every file listed on the
2893 * command line. It will find every cabinet file in that file, and will
2894 * search for every chained cabinet attached to those cabinets, then it
2895 * will either extract or list the cabinet(s). Returns 0 for success or 1
2896 * for failure (unlike most cabextract functions).
2898 int process_cabinet(char *cabname, char *dir,
2899 int fix, int view, int lower, int quiet) {
2901 struct cabinet *basecab, *cab, *cab1, *cab2;
2902 struct file *filelist, *fi;
2904 /* has the list-mode header been seen before? */
2907 if (view || !quiet) {
2908 printf("%s cabinet: %s\n", view ? "Viewing" : "Extracting", cabname);
2911 /* load the file requested */
2912 basecab = find_cabs_in_file(cabname);
2913 if (!basecab) return 1;
2915 /* iterate over all cabinets found in that file */
2916 for (cab = basecab; cab; cab=cab->next) {
2918 /* bi-directionally load any spanning cabinets -- backwards */
2919 for (cab1 = cab; cab1->flags & cfheadPREV_CABINET; cab1 = cab1->prevcab) {
2920 if (!quiet) printf("%s: extends backwards to %s (%s)\n", cabname,
2921 cab1->prevname, cab1->previnfo);
2922 find_cabinet_file(&(cab1->prevname), cabname);
2923 if (!(cab1->prevcab = load_cab_offset(cab1->prevname, 0))) {
2924 fprintf(stderr, "%s: can't read previous cabinet %s\n",
2925 cabname, cab1->prevname);
2928 cab1->prevcab->nextcab = cab1;
2931 /* bi-directionally load any spanning cabinets -- forwards */
2932 for (cab2 = cab; cab2->flags & cfheadNEXT_CABINET; cab2 = cab2->nextcab) {
2933 if (!quiet) printf("%s: extends to %s (%s)\n", cabname,
2934 cab2->nextname, cab2->nextinfo);
2935 find_cabinet_file(&(cab2->nextname), cabname);
2936 if (!(cab2->nextcab = load_cab_offset(cab2->nextname, 0))) {
2937 fprintf(stderr, "%s: can't read next cabinet %s\n",
2938 cabname, cab2->nextname);
2941 cab2->nextcab->prevcab = cab2;
2944 filelist = process_files(cab1);
2945 CAB(current) = NULL;
2947 if (view && !viewhdr) {
2948 printf("File size | Date Time | Name\n");
2949 printf("----------+---------------------+-------------\n");
2952 for (fi = filelist; fi; fi = fi->next) {
2957 if (!quiet) printf(" extracting: %s\n", fi->filename);
2958 extract_file(fi, lower, fix, dir);
2963 if (view) printf("\n");
2964 else if (!quiet) printf("Finished processing cabinet.\n\n");
2969 struct option opts[] = {
2970 { "version", 0, NULL, 'v' },
2971 { "help", 0, NULL, 'h' },
2972 { "list", 0, NULL, 'l' },
2973 { "quiet", 0, NULL, 'q' },
2974 { "lowercase", 0, NULL, 'L' },
2975 { "fix", 0, NULL, 'f' },
2976 { "directory", 1, NULL, 'd' },
2977 { NULL, 0, NULL, 0 }
2980 int main(int argc, char *argv[]) {
2981 int help=0, list=0, lower=0, view=0, quiet=0, fix=0, x, err=0;
2983 while ((x = getopt_long(argc, argv, "vhlqLfd:", opts, NULL)) != -1) {
2985 case 'v': view = 1; break;
2986 case 'h': help = 1; break;
2987 case 'l': list = 1; break;
2988 case 'q': quiet = 1; break;
2989 case 'L': lower = 1; break;
2990 case 'f': fix = 1; break;
2991 case 'd': dir = optarg; break;
2997 "Usage: %s [options] [-d dir] <cabinet file(s)>\n\n"
2998 "This will extract all files from a cabinet or executable cabinet.\n"
2999 "For multi-part cabinets, only specify the first file in the set.\n\n"
3001 " -v --version print version / list cabinet\n"
3002 " -h --help show this help page\n"
3003 " -l --list list contents of cabinet\n"
3004 " -q --quiet only print errors and warnings\n"
3005 " -L --lowercase make filenames lowercase\n"
3006 " -f --fix fix (some) corrupted cabinets\n"
3007 " -d --directory extract all files to the given directory\n\n"
3008 "cabextract %s (C) 2000-2002 Stuart Caie <kyzer@4u.net>\n"
3009 "This is free software with ABSOLUTELY NO WARRANTY.\n",
3015 if (optind == argc) {
3016 /* no arguments other than the options */
3018 printf("cabextract version %s\n", VERSION);
3022 fprintf(stderr, "cabextract: No cabinet files specified.\n"
3023 "Try '%s --help' for more information.\n", argv[0]);
3028 while (optind != argc) {
3029 err |= process_cabinet(argv[optind++], dir, fix, view||list, lower, quiet);
git://git.jankratochvil.net / captive.git / blob