Added a missing prototype.
[wine/wine-kai.git] / dlls / cabinet / cabextract.c
blob646a8680d593fef72ee7a53d3306a0d2ad460ac3
1 /*
2 * cabextract.c
4 * Copyright 2000-2002 Stuart Caie
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 * Principal author: Stuart Caie <kyzer@4u.net>
22 * Based on specification documents from Microsoft Corporation
23 * Quantum decompression researched and implemented by Matthew Russoto
24 * Huffman code adapted from unlzx by Dave Tritscher.
25 * InfoZip team's INFLATE implementation adapted to MSZIP by Dirk Stoecker.
26 * Major LZX fixes by Jae Jung.
29 #include "config.h"
31 #include <stdarg.h>
32 #include <stdio.h>
33 #include <stdlib.h>
35 #include "windef.h"
36 #include "winbase.h"
37 #include "winerror.h"
39 #include "cabinet.h"
41 #include "wine/debug.h"
43 WINE_DEFAULT_DEBUG_CHANNEL(cabinet);
45 THOSE_ZIP_CONSTS;
47 /* all the file IO is abstracted into these routines:
48 * cabinet_(open|close|read|seek|skip|getoffset)
49 * file_(open|close|write)
52 /* try to open a cabinet file, returns success */
53 BOOL cabinet_open(struct cabinet *cab)
55 const char *name = cab->filename;
56 HANDLE fh;
58 TRACE("(cab == ^%p)\n", cab);
60 if ((fh = CreateFileA( name, GENERIC_READ, FILE_SHARE_READ,
61 NULL, OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, NULL )) == INVALID_HANDLE_VALUE) {
62 ERR("Couldn't open %s\n", debugstr_a(name));
63 return FALSE;
66 /* seek to end of file and get the length */
67 if ((cab->filelen = SetFilePointer(fh, 0, NULL, FILE_END)) == INVALID_SET_FILE_POINTER) {
68 if (GetLastError() != NO_ERROR) {
69 ERR("Seek END failed: %s\n", debugstr_a(name));
70 CloseHandle(fh);
71 return FALSE;
75 /* return to the start of the file */
76 if (SetFilePointer(fh, 0, NULL, FILE_BEGIN) == INVALID_SET_FILE_POINTER) {
77 ERR("Seek BEGIN failed: %s\n", debugstr_a(name));
78 CloseHandle(fh);
79 return FALSE;
82 cab->fh = fh;
83 return TRUE;
86 /*******************************************************************
87 * cabinet_close (internal)
89 * close the file handle in a struct cabinet.
91 void cabinet_close(struct cabinet *cab) {
92 TRACE("(cab == ^%p)\n", cab);
93 if (cab->fh) CloseHandle(cab->fh);
94 cab->fh = 0;
97 /*******************************************************
98 * ensure_filepath2 (internal)
100 BOOL ensure_filepath2(char *path) {
101 BOOL ret = TRUE;
102 int len;
103 char *new_path;
105 new_path = HeapAlloc(GetProcessHeap(), 0, (strlen(path) + 1));
106 strcpy(new_path, path);
108 while((len = strlen(new_path)) && new_path[len - 1] == '\\')
109 new_path[len - 1] = 0;
111 TRACE("About to try to create directory %s\n", debugstr_a(new_path));
112 while(!CreateDirectoryA(new_path, NULL)) {
113 char *slash;
114 DWORD last_error = GetLastError();
116 if(last_error == ERROR_ALREADY_EXISTS)
117 break;
119 if(last_error != ERROR_PATH_NOT_FOUND) {
120 ret = FALSE;
121 break;
124 if(!(slash = strrchr(new_path, '\\'))) {
125 ret = FALSE;
126 break;
129 len = slash - new_path;
130 new_path[len] = 0;
131 if(! ensure_filepath2(new_path)) {
132 ret = FALSE;
133 break;
135 new_path[len] = '\\';
136 TRACE("New path in next iteration: %s\n", debugstr_a(new_path));
139 HeapFree(GetProcessHeap(), 0, new_path);
140 return ret;
144 /**********************************************************************
145 * ensure_filepath (internal)
147 * ensure_filepath("a\b\c\d.txt") ensures a, a\b and a\b\c exist as dirs
149 BOOL ensure_filepath(char *path) {
150 char new_path[MAX_PATH];
151 int len, i, lastslashpos = -1;
153 TRACE("(path == %s)\n", debugstr_a(path));
155 strcpy(new_path, path);
156 /* remove trailing slashes (shouldn't need to but wth...) */
157 while ((len = strlen(new_path)) && new_path[len - 1] == '\\')
158 new_path[len - 1] = 0;
159 /* find the position of the last '\\' */
160 for (i=0; i<MAX_PATH; i++) {
161 if (new_path[i] == 0) break;
162 if (new_path[i] == '\\')
163 lastslashpos = i;
165 if (lastslashpos > 0) {
166 new_path[lastslashpos] = 0;
167 /* may be trailing slashes but ensure_filepath2 will chop them */
168 return ensure_filepath2(new_path);
169 } else
170 return TRUE; /* ? */
173 /*******************************************************************
174 * file_open (internal)
176 * opens a file for output, returns success
178 BOOL file_open(struct cab_file *fi, BOOL lower, LPCSTR dir)
180 char c, *d, *name;
181 BOOL ok = FALSE;
182 const char *s;
184 TRACE("(fi == ^%p, lower == %s, dir == %s)\n", fi, lower ? "TRUE" : "FALSE", debugstr_a(dir));
186 if (!(name = malloc(strlen(fi->filename) + (dir ? strlen(dir) : 0) + 2))) {
187 ERR("out of memory!\n");
188 return FALSE;
191 /* start with blank name */
192 *name = 0;
194 /* add output directory if needed */
195 if (dir) {
196 strcpy(name, dir);
197 strcat(name, "\\");
200 /* remove leading slashes */
201 s = (char *) fi->filename;
202 while (*s == '\\') s++;
204 /* copy from fi->filename to new name.
205 * lowercases characters if needed.
207 d = &name[strlen(name)];
208 do {
209 c = *s++;
210 *d++ = (lower ? tolower((unsigned char) c) : c);
211 } while (c);
213 /* create directories if needed, attempt to write file */
214 if (ensure_filepath(name)) {
215 fi->fh = CreateFileA(name, GENERIC_WRITE, 0, NULL,
216 CREATE_ALWAYS, FILE_ATTRIBUTE_NORMAL, 0);
217 if (fi->fh != INVALID_HANDLE_VALUE)
218 ok = TRUE;
219 else {
220 ERR("CreateFileA returned INVALID_HANDLE_VALUE\n");
221 fi->fh = 0;
223 } else
224 ERR("Couldn't ensure filepath for %s\n", debugstr_a(name));
226 if (!ok) {
227 ERR("Couldn't open file %s for writing\n", debugstr_a(name));
230 /* as full filename is no longer needed, free it */
231 free(name);
233 return ok;
236 /********************************************************
237 * close_file (internal)
239 * closes a completed file
241 void file_close(struct cab_file *fi)
243 TRACE("(fi == ^%p)\n", fi);
245 if (fi->fh) {
246 CloseHandle(fi->fh);
248 fi->fh = 0;
251 /******************************************************************
252 * file_write (internal)
254 * writes from buf to a file specified as a cab_file struct.
255 * returns success/failure
257 BOOL file_write(struct cab_file *fi, cab_UBYTE *buf, cab_off_t length)
259 DWORD bytes_written;
261 TRACE("(fi == ^%p, buf == ^%p, length == %u)\n", fi, buf, length);
263 if ((!WriteFile( fi->fh, (LPCVOID) buf, length, &bytes_written, FALSE) ||
264 (bytes_written != length))) {
265 ERR("Error writing file: %s\n", debugstr_a(fi->filename));
266 return FALSE;
268 return TRUE;
272 /*******************************************************************
273 * cabinet_skip (internal)
275 * advance the file pointer associated with the cab structure
276 * by distance bytes
278 void cabinet_skip(struct cabinet *cab, cab_off_t distance)
280 TRACE("(cab == ^%p, distance == %u)\n", cab, distance);
281 if (SetFilePointer(cab->fh, distance, NULL, FILE_CURRENT) == INVALID_SET_FILE_POINTER) {
282 if (distance != INVALID_SET_FILE_POINTER)
283 ERR("%s\n", debugstr_a(cab->filename));
287 /*******************************************************************
288 * cabinet_seek (internal)
290 * seek to the specified absolute offset in a cab
292 void cabinet_seek(struct cabinet *cab, cab_off_t offset) {
293 TRACE("(cab == ^%p, offset == %u)\n", cab, offset);
294 if (SetFilePointer(cab->fh, offset, NULL, FILE_BEGIN) != offset)
295 ERR("%s seek failure\n", debugstr_a(cab->filename));
298 /*******************************************************************
299 * cabinet_getoffset (internal)
301 * returns the file pointer position of a cab
303 cab_off_t cabinet_getoffset(struct cabinet *cab)
305 return SetFilePointer(cab->fh, 0, NULL, FILE_CURRENT);
308 /*******************************************************************
309 * cabinet_read (internal)
311 * read data from a cabinet, returns success
313 BOOL cabinet_read(struct cabinet *cab, cab_UBYTE *buf, cab_off_t length)
315 DWORD bytes_read;
316 cab_off_t avail = cab->filelen - cabinet_getoffset(cab);
318 TRACE("(cab == ^%p, buf == ^%p, length == %u)\n", cab, buf, length);
320 if (length > avail) {
321 WARN("%s: WARNING; cabinet is truncated\n", debugstr_a(cab->filename));
322 length = avail;
325 if (! ReadFile( cab->fh, (LPVOID) buf, length, &bytes_read, NULL )) {
326 ERR("%s read error\n", debugstr_a(cab->filename));
327 return FALSE;
328 } else if (bytes_read != length) {
329 ERR("%s read size mismatch\n", debugstr_a(cab->filename));
330 return FALSE;
333 return TRUE;
336 /**********************************************************************
337 * cabinet_read_string (internal)
339 * allocate and read an aribitrarily long string from the cabinet
341 char *cabinet_read_string(struct cabinet *cab)
343 cab_off_t len=256, base = cabinet_getoffset(cab), maxlen = cab->filelen - base;
344 BOOL ok = FALSE;
345 unsigned int i;
346 cab_UBYTE *buf = NULL;
348 TRACE("(cab == ^%p)\n", cab);
350 do {
351 if (len > maxlen) len = maxlen;
352 if (!(buf = realloc(buf, (size_t) len))) break;
353 if (!cabinet_read(cab, buf, (size_t) len)) break;
355 /* search for a null terminator in what we've just read */
356 for (i=0; i < len; i++) {
357 if (!buf[i]) {ok=TRUE; break;}
360 if (!ok) {
361 if (len == maxlen) {
362 ERR("%s: WARNING; cabinet is truncated\n", debugstr_a(cab->filename));
363 break;
365 len += 256;
366 cabinet_seek(cab, base);
368 } while (!ok);
370 if (!ok) {
371 if (buf)
372 free(buf);
373 else
374 ERR("out of memory!\n");
375 return NULL;
378 /* otherwise, set the stream to just after the string and return */
379 cabinet_seek(cab, base + ((cab_off_t) strlen((char *) buf)) + 1);
381 return (char *) buf;
384 /******************************************************************
385 * cabinet_read_entries (internal)
387 * reads the header and all folder and file entries in this cabinet
389 BOOL cabinet_read_entries(struct cabinet *cab)
391 int num_folders, num_files, header_resv, folder_resv = 0, i;
392 struct cab_folder *fol, *linkfol = NULL;
393 struct cab_file *file, *linkfile = NULL;
394 cab_off_t base_offset;
395 cab_UBYTE buf[64];
397 TRACE("(cab == ^%p)\n", cab);
399 /* read in the CFHEADER */
400 base_offset = cabinet_getoffset(cab);
401 if (!cabinet_read(cab, buf, cfhead_SIZEOF)) {
402 return FALSE;
405 /* check basic MSCF signature */
406 if (EndGetI32(buf+cfhead_Signature) != 0x4643534d) {
407 ERR("%s: not a Microsoft cabinet file\n", debugstr_a(cab->filename));
408 return FALSE;
411 /* get the number of folders */
412 num_folders = EndGetI16(buf+cfhead_NumFolders);
413 if (num_folders == 0) {
414 ERR("%s: no folders in cabinet\n", debugstr_a(cab->filename));
415 return FALSE;
418 /* get the number of files */
419 num_files = EndGetI16(buf+cfhead_NumFiles);
420 if (num_files == 0) {
421 ERR("%s: no files in cabinet\n", debugstr_a(cab->filename));
422 return FALSE;
425 /* just check the header revision */
426 if ((buf[cfhead_MajorVersion] > 1) ||
427 (buf[cfhead_MajorVersion] == 1 && buf[cfhead_MinorVersion] > 3))
429 WARN("%s: WARNING; cabinet format version > 1.3\n", debugstr_a(cab->filename));
432 /* read the reserved-sizes part of header, if present */
433 cab->flags = EndGetI16(buf+cfhead_Flags);
434 if (cab->flags & cfheadRESERVE_PRESENT) {
435 if (!cabinet_read(cab, buf, cfheadext_SIZEOF)) return FALSE;
436 header_resv = EndGetI16(buf+cfheadext_HeaderReserved);
437 folder_resv = buf[cfheadext_FolderReserved];
438 cab->block_resv = buf[cfheadext_DataReserved];
440 if (header_resv > 60000) {
441 WARN("%s: WARNING; header reserved space > 60000\n", debugstr_a(cab->filename));
444 /* skip the reserved header */
445 if (header_resv)
446 if (SetFilePointer(cab->fh, (cab_off_t) header_resv, NULL, FILE_CURRENT) == INVALID_SET_FILE_POINTER)
447 ERR("seek failure: %s\n", debugstr_a(cab->filename));
450 if (cab->flags & cfheadPREV_CABINET) {
451 cab->prevname = cabinet_read_string(cab);
452 if (!cab->prevname) return FALSE;
453 cab->previnfo = cabinet_read_string(cab);
456 if (cab->flags & cfheadNEXT_CABINET) {
457 cab->nextname = cabinet_read_string(cab);
458 if (!cab->nextname) return FALSE;
459 cab->nextinfo = cabinet_read_string(cab);
462 /* read folders */
463 for (i = 0; i < num_folders; i++) {
464 if (!cabinet_read(cab, buf, cffold_SIZEOF)) return FALSE;
465 if (folder_resv) cabinet_skip(cab, folder_resv);
467 fol = (struct cab_folder *) calloc(1, sizeof(struct cab_folder));
468 if (!fol) {
469 ERR("out of memory!\n");
470 return FALSE;
473 fol->cab[0] = cab;
474 fol->offset[0] = base_offset + (cab_off_t) EndGetI32(buf+cffold_DataOffset);
475 fol->num_blocks = EndGetI16(buf+cffold_NumBlocks);
476 fol->comp_type = EndGetI16(buf+cffold_CompType);
478 if (!linkfol)
479 cab->folders = fol;
480 else
481 linkfol->next = fol;
483 linkfol = fol;
486 /* read files */
487 for (i = 0; i < num_files; i++) {
488 if (!cabinet_read(cab, buf, cffile_SIZEOF))
489 return FALSE;
491 file = (struct cab_file *) calloc(1, sizeof(struct cab_file));
492 if (!file) {
493 ERR("out of memory!\n");
494 return FALSE;
497 file->length = EndGetI32(buf+cffile_UncompressedSize);
498 file->offset = EndGetI32(buf+cffile_FolderOffset);
499 file->index = EndGetI16(buf+cffile_FolderIndex);
500 file->time = EndGetI16(buf+cffile_Time);
501 file->date = EndGetI16(buf+cffile_Date);
502 file->attribs = EndGetI16(buf+cffile_Attribs);
503 file->filename = cabinet_read_string(cab);
505 if (!file->filename) {
506 free(file);
507 return FALSE;
510 if (!linkfile)
511 cab->files = file;
512 else
513 linkfile->next = file;
515 linkfile = file;
517 return TRUE;
520 /***********************************************************
521 * load_cab_offset (internal)
523 * validates and reads file entries from a cabinet at offset [offset] in
524 * file [name]. Returns a cabinet structure if successful, or NULL
525 * otherwise.
527 struct cabinet *load_cab_offset(LPCSTR name, cab_off_t offset)
529 struct cabinet *cab = (struct cabinet *) calloc(1, sizeof(struct cabinet));
530 int ok;
532 TRACE("(name == %s, offset == %u)\n", debugstr_a(name), offset);
534 if (!cab) return NULL;
536 cab->filename = name;
537 if ((ok = cabinet_open(cab))) {
538 cabinet_seek(cab, offset);
539 ok = cabinet_read_entries(cab);
540 cabinet_close(cab);
543 if (ok) return cab;
544 free(cab);
545 return NULL;
548 /* MSZIP decruncher */
550 /* Dirk Stoecker wrote the ZIP decoder, based on the InfoZip deflate code */
552 /********************************************************
553 * Ziphuft_free (internal)
555 void Ziphuft_free(struct Ziphuft *t)
557 register struct Ziphuft *p, *q;
559 /* Go through linked list, freeing from the allocated (t[-1]) address. */
560 p = t;
561 while (p != (struct Ziphuft *)NULL)
563 q = (--p)->v.t;
564 free(p);
565 p = q;
569 /*********************************************************
570 * Ziphuft_build (internal)
572 cab_LONG Ziphuft_build(cab_ULONG *b, cab_ULONG n, cab_ULONG s, cab_UWORD *d, cab_UWORD *e,
573 struct Ziphuft **t, cab_LONG *m, cab_decomp_state *decomp_state)
575 cab_ULONG a; /* counter for codes of length k */
576 cab_ULONG el; /* length of EOB code (value 256) */
577 cab_ULONG f; /* i repeats in table every f entries */
578 cab_LONG g; /* maximum code length */
579 cab_LONG h; /* table level */
580 register cab_ULONG i; /* counter, current code */
581 register cab_ULONG j; /* counter */
582 register cab_LONG k; /* number of bits in current code */
583 cab_LONG *l; /* stack of bits per table */
584 register cab_ULONG *p; /* pointer into ZIP(c)[],ZIP(b)[],ZIP(v)[] */
585 register struct Ziphuft *q; /* points to current table */
586 struct Ziphuft r; /* table entry for structure assignment */
587 register cab_LONG w; /* bits before this table == (l * h) */
588 cab_ULONG *xp; /* pointer into x */
589 cab_LONG y; /* number of dummy codes added */
590 cab_ULONG z; /* number of entries in current table */
592 l = ZIP(lx)+1;
594 /* Generate counts for each bit length */
595 el = n > 256 ? b[256] : ZIPBMAX; /* set length of EOB code, if any */
597 for(i = 0; i < ZIPBMAX+1; ++i)
598 ZIP(c)[i] = 0;
599 p = b; i = n;
602 ZIP(c)[*p]++; p++; /* assume all entries <= ZIPBMAX */
603 } while (--i);
604 if (ZIP(c)[0] == n) /* null input--all zero length codes */
606 *t = (struct Ziphuft *)NULL;
607 *m = 0;
608 return 0;
611 /* Find minimum and maximum length, bound *m by those */
612 for (j = 1; j <= ZIPBMAX; j++)
613 if (ZIP(c)[j])
614 break;
615 k = j; /* minimum code length */
616 if ((cab_ULONG)*m < j)
617 *m = j;
618 for (i = ZIPBMAX; i; i--)
619 if (ZIP(c)[i])
620 break;
621 g = i; /* maximum code length */
622 if ((cab_ULONG)*m > i)
623 *m = i;
625 /* Adjust last length count to fill out codes, if needed */
626 for (y = 1 << j; j < i; j++, y <<= 1)
627 if ((y -= ZIP(c)[j]) < 0)
628 return 2; /* bad input: more codes than bits */
629 if ((y -= ZIP(c)[i]) < 0)
630 return 2;
631 ZIP(c)[i] += y;
633 /* Generate starting offsets LONGo the value table for each length */
634 ZIP(x)[1] = j = 0;
635 p = ZIP(c) + 1; xp = ZIP(x) + 2;
636 while (--i)
637 { /* note that i == g from above */
638 *xp++ = (j += *p++);
641 /* Make a table of values in order of bit lengths */
642 p = b; i = 0;
644 if ((j = *p++) != 0)
645 ZIP(v)[ZIP(x)[j]++] = i;
646 } while (++i < n);
649 /* Generate the Huffman codes and for each, make the table entries */
650 ZIP(x)[0] = i = 0; /* first Huffman code is zero */
651 p = ZIP(v); /* grab values in bit order */
652 h = -1; /* no tables yet--level -1 */
653 w = l[-1] = 0; /* no bits decoded yet */
654 ZIP(u)[0] = (struct Ziphuft *)NULL; /* just to keep compilers happy */
655 q = (struct Ziphuft *)NULL; /* ditto */
656 z = 0; /* ditto */
658 /* go through the bit lengths (k already is bits in shortest code) */
659 for (; k <= g; k++)
661 a = ZIP(c)[k];
662 while (a--)
664 /* here i is the Huffman code of length k bits for value *p */
665 /* make tables up to required level */
666 while (k > w + l[h])
668 w += l[h++]; /* add bits already decoded */
670 /* compute minimum size table less than or equal to *m bits */
671 z = (z = g - w) > (cab_ULONG)*m ? *m : z; /* upper limit */
672 if ((f = 1 << (j = k - w)) > a + 1) /* try a k-w bit table */
673 { /* too few codes for k-w bit table */
674 f -= a + 1; /* deduct codes from patterns left */
675 xp = ZIP(c) + k;
676 while (++j < z) /* try smaller tables up to z bits */
678 if ((f <<= 1) <= *++xp)
679 break; /* enough codes to use up j bits */
680 f -= *xp; /* else deduct codes from patterns */
683 if ((cab_ULONG)w + j > el && (cab_ULONG)w < el)
684 j = el - w; /* make EOB code end at table */
685 z = 1 << j; /* table entries for j-bit table */
686 l[h] = j; /* set table size in stack */
688 /* allocate and link in new table */
689 if (!(q = (struct Ziphuft *) malloc((z + 1)*sizeof(struct Ziphuft))))
691 if(h)
692 Ziphuft_free(ZIP(u)[0]);
693 return 3; /* not enough memory */
695 *t = q + 1; /* link to list for Ziphuft_free() */
696 *(t = &(q->v.t)) = (struct Ziphuft *)NULL;
697 ZIP(u)[h] = ++q; /* table starts after link */
699 /* connect to last table, if there is one */
700 if (h)
702 ZIP(x)[h] = i; /* save pattern for backing up */
703 r.b = (cab_UBYTE)l[h-1]; /* bits to dump before this table */
704 r.e = (cab_UBYTE)(16 + j); /* bits in this table */
705 r.v.t = q; /* pointer to this table */
706 j = (i & ((1 << w) - 1)) >> (w - l[h-1]);
707 ZIP(u)[h-1][j] = r; /* connect to last table */
711 /* set up table entry in r */
712 r.b = (cab_UBYTE)(k - w);
713 if (p >= ZIP(v) + n)
714 r.e = 99; /* out of values--invalid code */
715 else if (*p < s)
717 r.e = (cab_UBYTE)(*p < 256 ? 16 : 15); /* 256 is end-of-block code */
718 r.v.n = *p++; /* simple code is just the value */
720 else
722 r.e = (cab_UBYTE)e[*p - s]; /* non-simple--look up in lists */
723 r.v.n = d[*p++ - s];
726 /* fill code-like entries with r */
727 f = 1 << (k - w);
728 for (j = i >> w; j < z; j += f)
729 q[j] = r;
731 /* backwards increment the k-bit code i */
732 for (j = 1 << (k - 1); i & j; j >>= 1)
733 i ^= j;
734 i ^= j;
736 /* backup over finished tables */
737 while ((i & ((1 << w) - 1)) != ZIP(x)[h])
738 w -= l[--h]; /* don't need to update q */
742 /* return actual size of base table */
743 *m = l[0];
745 /* Return true (1) if we were given an incomplete table */
746 return y != 0 && g != 1;
749 /*********************************************************
750 * Zipinflate_codes (internal)
752 cab_LONG Zipinflate_codes(struct Ziphuft *tl, struct Ziphuft *td,
753 cab_LONG bl, cab_LONG bd, cab_decomp_state *decomp_state)
755 register cab_ULONG e; /* table entry flag/number of extra bits */
756 cab_ULONG n, d; /* length and index for copy */
757 cab_ULONG w; /* current window position */
758 struct Ziphuft *t; /* pointer to table entry */
759 cab_ULONG ml, md; /* masks for bl and bd bits */
760 register cab_ULONG b; /* bit buffer */
761 register cab_ULONG k; /* number of bits in bit buffer */
763 /* make local copies of globals */
764 b = ZIP(bb); /* initialize bit buffer */
765 k = ZIP(bk);
766 w = ZIP(window_posn); /* initialize window position */
768 /* inflate the coded data */
769 ml = Zipmask[bl]; /* precompute masks for speed */
770 md = Zipmask[bd];
772 for(;;)
774 ZIPNEEDBITS((cab_ULONG)bl)
775 if((e = (t = tl + ((cab_ULONG)b & ml))->e) > 16)
778 if (e == 99)
779 return 1;
780 ZIPDUMPBITS(t->b)
781 e -= 16;
782 ZIPNEEDBITS(e)
783 } while ((e = (t = t->v.t + ((cab_ULONG)b & Zipmask[e]))->e) > 16);
784 ZIPDUMPBITS(t->b)
785 if (e == 16) /* then it's a literal */
786 CAB(outbuf)[w++] = (cab_UBYTE)t->v.n;
787 else /* it's an EOB or a length */
789 /* exit if end of block */
790 if(e == 15)
791 break;
793 /* get length of block to copy */
794 ZIPNEEDBITS(e)
795 n = t->v.n + ((cab_ULONG)b & Zipmask[e]);
796 ZIPDUMPBITS(e);
798 /* decode distance of block to copy */
799 ZIPNEEDBITS((cab_ULONG)bd)
800 if ((e = (t = td + ((cab_ULONG)b & md))->e) > 16)
801 do {
802 if (e == 99)
803 return 1;
804 ZIPDUMPBITS(t->b)
805 e -= 16;
806 ZIPNEEDBITS(e)
807 } while ((e = (t = t->v.t + ((cab_ULONG)b & Zipmask[e]))->e) > 16);
808 ZIPDUMPBITS(t->b)
809 ZIPNEEDBITS(e)
810 d = w - t->v.n - ((cab_ULONG)b & Zipmask[e]);
811 ZIPDUMPBITS(e)
814 n -= (e = (e = ZIPWSIZE - ((d &= ZIPWSIZE-1) > w ? d : w)) > n ?n:e);
817 CAB(outbuf)[w++] = CAB(outbuf)[d++];
818 } while (--e);
819 } while (n);
823 /* restore the globals from the locals */
824 ZIP(window_posn) = w; /* restore global window pointer */
825 ZIP(bb) = b; /* restore global bit buffer */
826 ZIP(bk) = k;
828 /* done */
829 return 0;
832 /***********************************************************
833 * Zipinflate_stored (internal)
835 cab_LONG Zipinflate_stored(cab_decomp_state *decomp_state)
836 /* "decompress" an inflated type 0 (stored) block. */
838 cab_ULONG n; /* number of bytes in block */
839 cab_ULONG w; /* current window position */
840 register cab_ULONG b; /* bit buffer */
841 register cab_ULONG k; /* number of bits in bit buffer */
843 /* make local copies of globals */
844 b = ZIP(bb); /* initialize bit buffer */
845 k = ZIP(bk);
846 w = ZIP(window_posn); /* initialize window position */
848 /* go to byte boundary */
849 n = k & 7;
850 ZIPDUMPBITS(n);
852 /* get the length and its complement */
853 ZIPNEEDBITS(16)
854 n = ((cab_ULONG)b & 0xffff);
855 ZIPDUMPBITS(16)
856 ZIPNEEDBITS(16)
857 if (n != (cab_ULONG)((~b) & 0xffff))
858 return 1; /* error in compressed data */
859 ZIPDUMPBITS(16)
861 /* read and output the compressed data */
862 while(n--)
864 ZIPNEEDBITS(8)
865 CAB(outbuf)[w++] = (cab_UBYTE)b;
866 ZIPDUMPBITS(8)
869 /* restore the globals from the locals */
870 ZIP(window_posn) = w; /* restore global window pointer */
871 ZIP(bb) = b; /* restore global bit buffer */
872 ZIP(bk) = k;
873 return 0;
876 /******************************************************
877 * Zipinflate_fixed (internal)
879 cab_LONG Zipinflate_fixed(cab_decomp_state *decomp_state)
881 struct Ziphuft *fixed_tl;
882 struct Ziphuft *fixed_td;
883 cab_LONG fixed_bl, fixed_bd;
884 cab_LONG i; /* temporary variable */
885 cab_ULONG *l;
887 l = ZIP(ll);
889 /* literal table */
890 for(i = 0; i < 144; i++)
891 l[i] = 8;
892 for(; i < 256; i++)
893 l[i] = 9;
894 for(; i < 280; i++)
895 l[i] = 7;
896 for(; i < 288; i++) /* make a complete, but wrong code set */
897 l[i] = 8;
898 fixed_bl = 7;
899 if((i = Ziphuft_build(l, 288, 257, (cab_UWORD *) Zipcplens,
900 (cab_UWORD *) Zipcplext, &fixed_tl, &fixed_bl, decomp_state)))
901 return i;
903 /* distance table */
904 for(i = 0; i < 30; i++) /* make an incomplete code set */
905 l[i] = 5;
906 fixed_bd = 5;
907 if((i = Ziphuft_build(l, 30, 0, (cab_UWORD *) Zipcpdist, (cab_UWORD *) Zipcpdext,
908 &fixed_td, &fixed_bd, decomp_state)) > 1)
910 Ziphuft_free(fixed_tl);
911 return i;
914 /* decompress until an end-of-block code */
915 i = Zipinflate_codes(fixed_tl, fixed_td, fixed_bl, fixed_bd, decomp_state);
917 Ziphuft_free(fixed_td);
918 Ziphuft_free(fixed_tl);
919 return i;
922 /**************************************************************
923 * Zipinflate_dynamic (internal)
925 cab_LONG Zipinflate_dynamic(cab_decomp_state *decomp_state)
926 /* decompress an inflated type 2 (dynamic Huffman codes) block. */
928 cab_LONG i; /* temporary variables */
929 cab_ULONG j;
930 cab_ULONG *ll;
931 cab_ULONG l; /* last length */
932 cab_ULONG m; /* mask for bit lengths table */
933 cab_ULONG n; /* number of lengths to get */
934 struct Ziphuft *tl; /* literal/length code table */
935 struct Ziphuft *td; /* distance code table */
936 cab_LONG bl; /* lookup bits for tl */
937 cab_LONG bd; /* lookup bits for td */
938 cab_ULONG nb; /* number of bit length codes */
939 cab_ULONG nl; /* number of literal/length codes */
940 cab_ULONG nd; /* number of distance codes */
941 register cab_ULONG b; /* bit buffer */
942 register cab_ULONG k; /* number of bits in bit buffer */
944 /* make local bit buffer */
945 b = ZIP(bb);
946 k = ZIP(bk);
947 ll = ZIP(ll);
949 /* read in table lengths */
950 ZIPNEEDBITS(5)
951 nl = 257 + ((cab_ULONG)b & 0x1f); /* number of literal/length codes */
952 ZIPDUMPBITS(5)
953 ZIPNEEDBITS(5)
954 nd = 1 + ((cab_ULONG)b & 0x1f); /* number of distance codes */
955 ZIPDUMPBITS(5)
956 ZIPNEEDBITS(4)
957 nb = 4 + ((cab_ULONG)b & 0xf); /* number of bit length codes */
958 ZIPDUMPBITS(4)
959 if(nl > 288 || nd > 32)
960 return 1; /* bad lengths */
962 /* read in bit-length-code lengths */
963 for(j = 0; j < nb; j++)
965 ZIPNEEDBITS(3)
966 ll[Zipborder[j]] = (cab_ULONG)b & 7;
967 ZIPDUMPBITS(3)
969 for(; j < 19; j++)
970 ll[Zipborder[j]] = 0;
972 /* build decoding table for trees--single level, 7 bit lookup */
973 bl = 7;
974 if((i = Ziphuft_build(ll, 19, 19, NULL, NULL, &tl, &bl, decomp_state)) != 0)
976 if(i == 1)
977 Ziphuft_free(tl);
978 return i; /* incomplete code set */
981 /* read in literal and distance code lengths */
982 n = nl + nd;
983 m = Zipmask[bl];
984 i = l = 0;
985 while((cab_ULONG)i < n)
987 ZIPNEEDBITS((cab_ULONG)bl)
988 j = (td = tl + ((cab_ULONG)b & m))->b;
989 ZIPDUMPBITS(j)
990 j = td->v.n;
991 if (j < 16) /* length of code in bits (0..15) */
992 ll[i++] = l = j; /* save last length in l */
993 else if (j == 16) /* repeat last length 3 to 6 times */
995 ZIPNEEDBITS(2)
996 j = 3 + ((cab_ULONG)b & 3);
997 ZIPDUMPBITS(2)
998 if((cab_ULONG)i + j > n)
999 return 1;
1000 while (j--)
1001 ll[i++] = l;
1003 else if (j == 17) /* 3 to 10 zero length codes */
1005 ZIPNEEDBITS(3)
1006 j = 3 + ((cab_ULONG)b & 7);
1007 ZIPDUMPBITS(3)
1008 if ((cab_ULONG)i + j > n)
1009 return 1;
1010 while (j--)
1011 ll[i++] = 0;
1012 l = 0;
1014 else /* j == 18: 11 to 138 zero length codes */
1016 ZIPNEEDBITS(7)
1017 j = 11 + ((cab_ULONG)b & 0x7f);
1018 ZIPDUMPBITS(7)
1019 if ((cab_ULONG)i + j > n)
1020 return 1;
1021 while (j--)
1022 ll[i++] = 0;
1023 l = 0;
1027 /* free decoding table for trees */
1028 Ziphuft_free(tl);
1030 /* restore the global bit buffer */
1031 ZIP(bb) = b;
1032 ZIP(bk) = k;
1034 /* build the decoding tables for literal/length and distance codes */
1035 bl = ZIPLBITS;
1036 if((i = Ziphuft_build(ll, nl, 257, (cab_UWORD *) Zipcplens, (cab_UWORD *) Zipcplext,
1037 &tl, &bl, decomp_state)) != 0)
1039 if(i == 1)
1040 Ziphuft_free(tl);
1041 return i; /* incomplete code set */
1043 bd = ZIPDBITS;
1044 Ziphuft_build(ll + nl, nd, 0, (cab_UWORD *) Zipcpdist, (cab_UWORD *) Zipcpdext,
1045 &td, &bd, decomp_state);
1047 /* decompress until an end-of-block code */
1048 if(Zipinflate_codes(tl, td, bl, bd, decomp_state))
1049 return 1;
1051 /* free the decoding tables, return */
1052 Ziphuft_free(tl);
1053 Ziphuft_free(td);
1054 return 0;
1057 /*****************************************************
1058 * Zipinflate_block (internal)
1060 cab_LONG Zipinflate_block(cab_LONG *e, cab_decomp_state *decomp_state) /* e == last block flag */
1061 { /* decompress an inflated block */
1062 cab_ULONG t; /* block type */
1063 register cab_ULONG b; /* bit buffer */
1064 register cab_ULONG k; /* number of bits in bit buffer */
1066 /* make local bit buffer */
1067 b = ZIP(bb);
1068 k = ZIP(bk);
1070 /* read in last block bit */
1071 ZIPNEEDBITS(1)
1072 *e = (cab_LONG)b & 1;
1073 ZIPDUMPBITS(1)
1075 /* read in block type */
1076 ZIPNEEDBITS(2)
1077 t = (cab_ULONG)b & 3;
1078 ZIPDUMPBITS(2)
1080 /* restore the global bit buffer */
1081 ZIP(bb) = b;
1082 ZIP(bk) = k;
1084 /* inflate that block type */
1085 if(t == 2)
1086 return Zipinflate_dynamic(decomp_state);
1087 if(t == 0)
1088 return Zipinflate_stored(decomp_state);
1089 if(t == 1)
1090 return Zipinflate_fixed(decomp_state);
1091 /* bad block type */
1092 return 2;
1095 /****************************************************
1096 * ZIPdecompress (internal)
1098 int ZIPdecompress(int inlen, int outlen, cab_decomp_state *decomp_state)
1100 cab_LONG e; /* last block flag */
1102 TRACE("(inlen == %d, outlen == %d)\n", inlen, outlen);
1104 ZIP(inpos) = CAB(inbuf);
1105 ZIP(bb) = ZIP(bk) = ZIP(window_posn) = 0;
1106 if(outlen > ZIPWSIZE)
1107 return DECR_DATAFORMAT;
1109 /* CK = Chris Kirmse, official Microsoft purloiner */
1110 if(ZIP(inpos)[0] != 0x43 || ZIP(inpos)[1] != 0x4B)
1111 return DECR_ILLEGALDATA;
1112 ZIP(inpos) += 2;
1116 if(Zipinflate_block(&e, decomp_state))
1117 return DECR_ILLEGALDATA;
1118 } while(!e);
1120 /* return success */
1121 return DECR_OK;
1124 /* Quantum decruncher */
1126 /* This decruncher was researched and implemented by Matthew Russoto. */
1127 /* It has since been tidied up by Stuart Caie */
1129 /******************************************************************
1130 * QTMinitmodel (internal)
1132 * Initialise a model which decodes symbols from [s] to [s]+[n]-1
1134 void QTMinitmodel(struct QTMmodel *m, struct QTMmodelsym *sym, int n, int s) {
1135 int i;
1136 m->shiftsleft = 4;
1137 m->entries = n;
1138 m->syms = sym;
1139 memset(m->tabloc, 0xFF, sizeof(m->tabloc)); /* clear out look-up table */
1140 for (i = 0; i < n; i++) {
1141 m->tabloc[i+s] = i; /* set up a look-up entry for symbol */
1142 m->syms[i].sym = i+s; /* actual symbol */
1143 m->syms[i].cumfreq = n-i; /* current frequency of that symbol */
1145 m->syms[n].cumfreq = 0;
1148 /******************************************************************
1149 * QTMinit (internal)
1151 int QTMinit(int window, int level, cab_decomp_state *decomp_state) {
1152 unsigned int wndsize = 1 << window;
1153 int msz = window * 2, i;
1154 cab_ULONG j;
1156 /* QTM supports window sizes of 2^10 (1Kb) through 2^21 (2Mb) */
1157 /* if a previously allocated window is big enough, keep it */
1158 if (window < 10 || window > 21) return DECR_DATAFORMAT;
1159 if (QTM(actual_size) < wndsize) {
1160 if (QTM(window)) free(QTM(window));
1161 QTM(window) = NULL;
1163 if (!QTM(window)) {
1164 if (!(QTM(window) = malloc(wndsize))) return DECR_NOMEMORY;
1165 QTM(actual_size) = wndsize;
1167 QTM(window_size) = wndsize;
1168 QTM(window_posn) = 0;
1170 /* initialise static slot/extrabits tables */
1171 for (i = 0, j = 0; i < 27; i++) {
1172 CAB(q_length_extra)[i] = (i == 26) ? 0 : (i < 2 ? 0 : i - 2) >> 2;
1173 CAB(q_length_base)[i] = j; j += 1 << ((i == 26) ? 5 : CAB(q_length_extra)[i]);
1175 for (i = 0, j = 0; i < 42; i++) {
1176 CAB(q_extra_bits)[i] = (i < 2 ? 0 : i-2) >> 1;
1177 CAB(q_position_base)[i] = j; j += 1 << CAB(q_extra_bits)[i];
1180 /* initialise arithmetic coding models */
1182 QTMinitmodel(&QTM(model7), &QTM(m7sym)[0], 7, 0);
1184 QTMinitmodel(&QTM(model00), &QTM(m00sym)[0], 0x40, 0x00);
1185 QTMinitmodel(&QTM(model40), &QTM(m40sym)[0], 0x40, 0x40);
1186 QTMinitmodel(&QTM(model80), &QTM(m80sym)[0], 0x40, 0x80);
1187 QTMinitmodel(&QTM(modelC0), &QTM(mC0sym)[0], 0x40, 0xC0);
1189 /* model 4 depends on table size, ranges from 20 to 24 */
1190 QTMinitmodel(&QTM(model4), &QTM(m4sym)[0], (msz < 24) ? msz : 24, 0);
1191 /* model 5 depends on table size, ranges from 20 to 36 */
1192 QTMinitmodel(&QTM(model5), &QTM(m5sym)[0], (msz < 36) ? msz : 36, 0);
1193 /* model 6pos depends on table size, ranges from 20 to 42 */
1194 QTMinitmodel(&QTM(model6pos), &QTM(m6psym)[0], msz, 0);
1195 QTMinitmodel(&QTM(model6len), &QTM(m6lsym)[0], 27, 0);
1197 return DECR_OK;
1200 /****************************************************************
1201 * QTMupdatemodel (internal)
1203 void QTMupdatemodel(struct QTMmodel *model, int sym) {
1204 struct QTMmodelsym temp;
1205 int i, j;
1207 for (i = 0; i < sym; i++) model->syms[i].cumfreq += 8;
1209 if (model->syms[0].cumfreq > 3800) {
1210 if (--model->shiftsleft) {
1211 for (i = model->entries - 1; i >= 0; i--) {
1212 /* -1, not -2; the 0 entry saves this */
1213 model->syms[i].cumfreq >>= 1;
1214 if (model->syms[i].cumfreq <= model->syms[i+1].cumfreq) {
1215 model->syms[i].cumfreq = model->syms[i+1].cumfreq + 1;
1219 else {
1220 model->shiftsleft = 50;
1221 for (i = 0; i < model->entries ; i++) {
1222 /* no -1, want to include the 0 entry */
1223 /* this converts cumfreqs into frequencies, then shifts right */
1224 model->syms[i].cumfreq -= model->syms[i+1].cumfreq;
1225 model->syms[i].cumfreq++; /* avoid losing things entirely */
1226 model->syms[i].cumfreq >>= 1;
1229 /* now sort by frequencies, decreasing order -- this must be an
1230 * inplace selection sort, or a sort with the same (in)stability
1231 * characteristics
1233 for (i = 0; i < model->entries - 1; i++) {
1234 for (j = i + 1; j < model->entries; j++) {
1235 if (model->syms[i].cumfreq < model->syms[j].cumfreq) {
1236 temp = model->syms[i];
1237 model->syms[i] = model->syms[j];
1238 model->syms[j] = temp;
1243 /* then convert frequencies back to cumfreq */
1244 for (i = model->entries - 1; i >= 0; i--) {
1245 model->syms[i].cumfreq += model->syms[i+1].cumfreq;
1247 /* then update the other part of the table */
1248 for (i = 0; i < model->entries; i++) {
1249 model->tabloc[model->syms[i].sym] = i;
1255 /*******************************************************************
1256 * QTMdecompress (internal)
1258 int QTMdecompress(int inlen, int outlen, cab_decomp_state *decomp_state)
1260 cab_UBYTE *inpos = CAB(inbuf);
1261 cab_UBYTE *window = QTM(window);
1262 cab_UBYTE *runsrc, *rundest;
1264 cab_ULONG window_posn = QTM(window_posn);
1265 cab_ULONG window_size = QTM(window_size);
1267 /* used by bitstream macros */
1268 register int bitsleft, bitrun, bitsneed;
1269 register cab_ULONG bitbuf;
1271 /* used by GET_SYMBOL */
1272 cab_ULONG range;
1273 cab_UWORD symf;
1274 int i;
1276 int extra, togo = outlen, match_length = 0, copy_length;
1277 cab_UBYTE selector, sym;
1278 cab_ULONG match_offset = 0;
1280 cab_UWORD H = 0xFFFF, L = 0, C;
1282 TRACE("(inlen == %d, outlen == %d)\n", inlen, outlen);
1284 /* read initial value of C */
1285 Q_INIT_BITSTREAM;
1286 Q_READ_BITS(C, 16);
1288 /* apply 2^x-1 mask */
1289 window_posn &= window_size - 1;
1290 /* runs can't straddle the window wraparound */
1291 if ((window_posn + togo) > window_size) {
1292 TRACE("straddled run\n");
1293 return DECR_DATAFORMAT;
1296 while (togo > 0) {
1297 GET_SYMBOL(model7, selector);
1298 switch (selector) {
1299 case 0:
1300 GET_SYMBOL(model00, sym); window[window_posn++] = sym; togo--;
1301 break;
1302 case 1:
1303 GET_SYMBOL(model40, sym); window[window_posn++] = sym; togo--;
1304 break;
1305 case 2:
1306 GET_SYMBOL(model80, sym); window[window_posn++] = sym; togo--;
1307 break;
1308 case 3:
1309 GET_SYMBOL(modelC0, sym); window[window_posn++] = sym; togo--;
1310 break;
1312 case 4:
1313 /* selector 4 = fixed length of 3 */
1314 GET_SYMBOL(model4, sym);
1315 Q_READ_BITS(extra, CAB(q_extra_bits)[sym]);
1316 match_offset = CAB(q_position_base)[sym] + extra + 1;
1317 match_length = 3;
1318 break;
1320 case 5:
1321 /* selector 5 = fixed length of 4 */
1322 GET_SYMBOL(model5, sym);
1323 Q_READ_BITS(extra, CAB(q_extra_bits)[sym]);
1324 match_offset = CAB(q_position_base)[sym] + extra + 1;
1325 match_length = 4;
1326 break;
1328 case 6:
1329 /* selector 6 = variable length */
1330 GET_SYMBOL(model6len, sym);
1331 Q_READ_BITS(extra, CAB(q_length_extra)[sym]);
1332 match_length = CAB(q_length_base)[sym] + extra + 5;
1333 GET_SYMBOL(model6pos, sym);
1334 Q_READ_BITS(extra, CAB(q_extra_bits)[sym]);
1335 match_offset = CAB(q_position_base)[sym] + extra + 1;
1336 break;
1338 default:
1339 TRACE("Selector is bogus\n");
1340 return DECR_ILLEGALDATA;
1343 /* if this is a match */
1344 if (selector >= 4) {
1345 rundest = window + window_posn;
1346 togo -= match_length;
1348 /* copy any wrapped around source data */
1349 if (window_posn >= match_offset) {
1350 /* no wrap */
1351 runsrc = rundest - match_offset;
1352 } else {
1353 runsrc = rundest + (window_size - match_offset);
1354 copy_length = match_offset - window_posn;
1355 if (copy_length < match_length) {
1356 match_length -= copy_length;
1357 window_posn += copy_length;
1358 while (copy_length-- > 0) *rundest++ = *runsrc++;
1359 runsrc = window;
1362 window_posn += match_length;
1364 /* copy match data - no worries about destination wraps */
1365 while (match_length-- > 0) *rundest++ = *runsrc++;
1367 } /* while (togo > 0) */
1369 if (togo != 0) {
1370 TRACE("Frame overflow, this_run = %d\n", togo);
1371 return DECR_ILLEGALDATA;
1374 memcpy(CAB(outbuf), window + ((!window_posn) ? window_size : window_posn) -
1375 outlen, outlen);
1377 QTM(window_posn) = window_posn;
1378 return DECR_OK;
1381 /* LZX decruncher */
1383 /* Microsoft's LZX document and their implementation of the
1384 * com.ms.util.cab Java package do not concur.
1386 * In the LZX document, there is a table showing the correlation between
1387 * window size and the number of position slots. It states that the 1MB
1388 * window = 40 slots and the 2MB window = 42 slots. In the implementation,
1389 * 1MB = 42 slots, 2MB = 50 slots. The actual calculation is 'find the
1390 * first slot whose position base is equal to or more than the required
1391 * window size'. This would explain why other tables in the document refer
1392 * to 50 slots rather than 42.
1394 * The constant NUM_PRIMARY_LENGTHS used in the decompression pseudocode
1395 * is not defined in the specification.
1397 * The LZX document does not state the uncompressed block has an
1398 * uncompressed length field. Where does this length field come from, so
1399 * we can know how large the block is? The implementation has it as the 24
1400 * bits following after the 3 blocktype bits, before the alignment
1401 * padding.
1403 * The LZX document states that aligned offset blocks have their aligned
1404 * offset huffman tree AFTER the main and length trees. The implementation
1405 * suggests that the aligned offset tree is BEFORE the main and length
1406 * trees.
1408 * The LZX document decoding algorithm states that, in an aligned offset
1409 * block, if an extra_bits value is 1, 2 or 3, then that number of bits
1410 * should be read and the result added to the match offset. This is
1411 * correct for 1 and 2, but not 3, where just a huffman symbol (using the
1412 * aligned tree) should be read.
1414 * Regarding the E8 preprocessing, the LZX document states 'No translation
1415 * may be performed on the last 6 bytes of the input block'. This is
1416 * correct. However, the pseudocode provided checks for the *E8 leader*
1417 * up to the last 6 bytes. If the leader appears between -10 and -7 bytes
1418 * from the end, this would cause the next four bytes to be modified, at
1419 * least one of which would be in the last 6 bytes, which is not allowed
1420 * according to the spec.
1422 * The specification states that the huffman trees must always contain at
1423 * least one element. However, many CAB files contain blocks where the
1424 * length tree is completely empty (because there are no matches), and
1425 * this is expected to succeed.
1429 /* LZX uses what it calls 'position slots' to represent match offsets.
1430 * What this means is that a small 'position slot' number and a small
1431 * offset from that slot are encoded instead of one large offset for
1432 * every match.
1433 * - lzx_position_base is an index to the position slot bases
1434 * - lzx_extra_bits states how many bits of offset-from-base data is needed.
1437 /************************************************************
1438 * LZXinit (internal)
1440 int LZXinit(int window, cab_decomp_state *decomp_state) {
1441 cab_ULONG wndsize = 1 << window;
1442 int i, j, posn_slots;
1444 /* LZX supports window sizes of 2^15 (32Kb) through 2^21 (2Mb) */
1445 /* if a previously allocated window is big enough, keep it */
1446 if (window < 15 || window > 21) return DECR_DATAFORMAT;
1447 if (LZX(actual_size) < wndsize) {
1448 if (LZX(window)) free(LZX(window));
1449 LZX(window) = NULL;
1451 if (!LZX(window)) {
1452 if (!(LZX(window) = malloc(wndsize))) return DECR_NOMEMORY;
1453 LZX(actual_size) = wndsize;
1455 LZX(window_size) = wndsize;
1457 /* initialise static tables */
1458 for (i=0, j=0; i <= 50; i += 2) {
1459 CAB(extra_bits)[i] = CAB(extra_bits)[i+1] = j; /* 0,0,0,0,1,1,2,2,3,3... */
1460 if ((i != 0) && (j < 17)) j++; /* 0,0,1,2,3,4...15,16,17,17,17,17... */
1462 for (i=0, j=0; i <= 50; i++) {
1463 CAB(lzx_position_base)[i] = j; /* 0,1,2,3,4,6,8,12,16,24,32,... */
1464 j += 1 << CAB(extra_bits)[i]; /* 1,1,1,1,2,2,4,4,8,8,16,16,32,32,... */
1467 /* calculate required position slots */
1468 if (window == 20) posn_slots = 42;
1469 else if (window == 21) posn_slots = 50;
1470 else posn_slots = window << 1;
1472 /*posn_slots=i=0; while (i < wndsize) i += 1 << CAB(extra_bits)[posn_slots++]; */
1474 LZX(R0) = LZX(R1) = LZX(R2) = 1;
1475 LZX(main_elements) = LZX_NUM_CHARS + (posn_slots << 3);
1476 LZX(header_read) = 0;
1477 LZX(frames_read) = 0;
1478 LZX(block_remaining) = 0;
1479 LZX(block_type) = LZX_BLOCKTYPE_INVALID;
1480 LZX(intel_curpos) = 0;
1481 LZX(intel_started) = 0;
1482 LZX(window_posn) = 0;
1484 /* initialise tables to 0 (because deltas will be applied to them) */
1485 for (i = 0; i < LZX_MAINTREE_MAXSYMBOLS; i++) LZX(MAINTREE_len)[i] = 0;
1486 for (i = 0; i < LZX_LENGTH_MAXSYMBOLS; i++) LZX(LENGTH_len)[i] = 0;
1488 return DECR_OK;
1491 /*************************************************************************
1492 * make_decode_table (internal)
1494 * This function was coded by David Tritscher. It builds a fast huffman
1495 * decoding table out of just a canonical huffman code lengths table.
1497 * PARAMS
1498 * nsyms: total number of symbols in this huffman tree.
1499 * nbits: any symbols with a code length of nbits or less can be decoded
1500 * in one lookup of the table.
1501 * length: A table to get code lengths from [0 to syms-1]
1502 * table: The table to fill up with decoded symbols and pointers.
1504 * RETURNS
1505 * OK: 0
1506 * error: 1
1508 int make_decode_table(cab_ULONG nsyms, cab_ULONG nbits, cab_UBYTE *length, cab_UWORD *table) {
1509 register cab_UWORD sym;
1510 register cab_ULONG leaf;
1511 register cab_UBYTE bit_num = 1;
1512 cab_ULONG fill;
1513 cab_ULONG pos = 0; /* the current position in the decode table */
1514 cab_ULONG table_mask = 1 << nbits;
1515 cab_ULONG bit_mask = table_mask >> 1; /* don't do 0 length codes */
1516 cab_ULONG next_symbol = bit_mask; /* base of allocation for long codes */
1518 /* fill entries for codes short enough for a direct mapping */
1519 while (bit_num <= nbits) {
1520 for (sym = 0; sym < nsyms; sym++) {
1521 if (length[sym] == bit_num) {
1522 leaf = pos;
1524 if((pos += bit_mask) > table_mask) return 1; /* table overrun */
1526 /* fill all possible lookups of this symbol with the symbol itself */
1527 fill = bit_mask;
1528 while (fill-- > 0) table[leaf++] = sym;
1531 bit_mask >>= 1;
1532 bit_num++;
1535 /* if there are any codes longer than nbits */
1536 if (pos != table_mask) {
1537 /* clear the remainder of the table */
1538 for (sym = pos; sym < table_mask; sym++) table[sym] = 0;
1540 /* give ourselves room for codes to grow by up to 16 more bits */
1541 pos <<= 16;
1542 table_mask <<= 16;
1543 bit_mask = 1 << 15;
1545 while (bit_num <= 16) {
1546 for (sym = 0; sym < nsyms; sym++) {
1547 if (length[sym] == bit_num) {
1548 leaf = pos >> 16;
1549 for (fill = 0; fill < bit_num - nbits; fill++) {
1550 /* if this path hasn't been taken yet, 'allocate' two entries */
1551 if (table[leaf] == 0) {
1552 table[(next_symbol << 1)] = 0;
1553 table[(next_symbol << 1) + 1] = 0;
1554 table[leaf] = next_symbol++;
1556 /* follow the path and select either left or right for next bit */
1557 leaf = table[leaf] << 1;
1558 if ((pos >> (15-fill)) & 1) leaf++;
1560 table[leaf] = sym;
1562 if ((pos += bit_mask) > table_mask) return 1; /* table overflow */
1565 bit_mask >>= 1;
1566 bit_num++;
1570 /* full table? */
1571 if (pos == table_mask) return 0;
1573 /* either erroneous table, or all elements are 0 - let's find out. */
1574 for (sym = 0; sym < nsyms; sym++) if (length[sym]) return 1;
1575 return 0;
1578 /************************************************************
1579 * lzx_read_lens (internal)
1581 int lzx_read_lens(cab_UBYTE *lens, cab_ULONG first, cab_ULONG last, struct lzx_bits *lb,
1582 cab_decomp_state *decomp_state) {
1583 cab_ULONG i,j, x,y;
1584 int z;
1586 register cab_ULONG bitbuf = lb->bb;
1587 register int bitsleft = lb->bl;
1588 cab_UBYTE *inpos = lb->ip;
1589 cab_UWORD *hufftbl;
1591 for (x = 0; x < 20; x++) {
1592 READ_BITS(y, 4);
1593 LENTABLE(PRETREE)[x] = y;
1595 BUILD_TABLE(PRETREE);
1597 for (x = first; x < last; ) {
1598 READ_HUFFSYM(PRETREE, z);
1599 if (z == 17) {
1600 READ_BITS(y, 4); y += 4;
1601 while (y--) lens[x++] = 0;
1603 else if (z == 18) {
1604 READ_BITS(y, 5); y += 20;
1605 while (y--) lens[x++] = 0;
1607 else if (z == 19) {
1608 READ_BITS(y, 1); y += 4;
1609 READ_HUFFSYM(PRETREE, z);
1610 z = lens[x] - z; if (z < 0) z += 17;
1611 while (y--) lens[x++] = z;
1613 else {
1614 z = lens[x] - z; if (z < 0) z += 17;
1615 lens[x++] = z;
1619 lb->bb = bitbuf;
1620 lb->bl = bitsleft;
1621 lb->ip = inpos;
1622 return 0;
1625 /*******************************************************
1626 * LZXdecompress (internal)
1628 int LZXdecompress(int inlen, int outlen, cab_decomp_state *decomp_state) {
1629 cab_UBYTE *inpos = CAB(inbuf);
1630 cab_UBYTE *endinp = inpos + inlen;
1631 cab_UBYTE *window = LZX(window);
1632 cab_UBYTE *runsrc, *rundest;
1633 cab_UWORD *hufftbl; /* used in READ_HUFFSYM macro as chosen decoding table */
1635 cab_ULONG window_posn = LZX(window_posn);
1636 cab_ULONG window_size = LZX(window_size);
1637 cab_ULONG R0 = LZX(R0);
1638 cab_ULONG R1 = LZX(R1);
1639 cab_ULONG R2 = LZX(R2);
1641 register cab_ULONG bitbuf;
1642 register int bitsleft;
1643 cab_ULONG match_offset, i,j,k; /* ijk used in READ_HUFFSYM macro */
1644 struct lzx_bits lb; /* used in READ_LENGTHS macro */
1646 int togo = outlen, this_run, main_element, aligned_bits;
1647 int match_length, copy_length, length_footer, extra, verbatim_bits;
1649 TRACE("(inlen == %d, outlen == %d)\n", inlen, outlen);
1651 INIT_BITSTREAM;
1653 /* read header if necessary */
1654 if (!LZX(header_read)) {
1655 i = j = 0;
1656 READ_BITS(k, 1); if (k) { READ_BITS(i,16); READ_BITS(j,16); }
1657 LZX(intel_filesize) = (i << 16) | j; /* or 0 if not encoded */
1658 LZX(header_read) = 1;
1661 /* main decoding loop */
1662 while (togo > 0) {
1663 /* last block finished, new block expected */
1664 if (LZX(block_remaining) == 0) {
1665 if (LZX(block_type) == LZX_BLOCKTYPE_UNCOMPRESSED) {
1666 if (LZX(block_length) & 1) inpos++; /* realign bitstream to word */
1667 INIT_BITSTREAM;
1670 READ_BITS(LZX(block_type), 3);
1671 READ_BITS(i, 16);
1672 READ_BITS(j, 8);
1673 LZX(block_remaining) = LZX(block_length) = (i << 8) | j;
1675 switch (LZX(block_type)) {
1676 case LZX_BLOCKTYPE_ALIGNED:
1677 for (i = 0; i < 8; i++) { READ_BITS(j, 3); LENTABLE(ALIGNED)[i] = j; }
1678 BUILD_TABLE(ALIGNED);
1679 /* rest of aligned header is same as verbatim */
1681 case LZX_BLOCKTYPE_VERBATIM:
1682 READ_LENGTHS(MAINTREE, 0, 256, lzx_read_lens);
1683 READ_LENGTHS(MAINTREE, 256, LZX(main_elements), lzx_read_lens);
1684 BUILD_TABLE(MAINTREE);
1685 if (LENTABLE(MAINTREE)[0xE8] != 0) LZX(intel_started) = 1;
1687 READ_LENGTHS(LENGTH, 0, LZX_NUM_SECONDARY_LENGTHS, lzx_read_lens);
1688 BUILD_TABLE(LENGTH);
1689 break;
1691 case LZX_BLOCKTYPE_UNCOMPRESSED:
1692 LZX(intel_started) = 1; /* because we can't assume otherwise */
1693 ENSURE_BITS(16); /* get up to 16 pad bits into the buffer */
1694 if (bitsleft > 16) inpos -= 2; /* and align the bitstream! */
1695 R0 = inpos[0]|(inpos[1]<<8)|(inpos[2]<<16)|(inpos[3]<<24);inpos+=4;
1696 R1 = inpos[0]|(inpos[1]<<8)|(inpos[2]<<16)|(inpos[3]<<24);inpos+=4;
1697 R2 = inpos[0]|(inpos[1]<<8)|(inpos[2]<<16)|(inpos[3]<<24);inpos+=4;
1698 break;
1700 default:
1701 return DECR_ILLEGALDATA;
1705 /* buffer exhaustion check */
1706 if (inpos > endinp) {
1707 /* it's possible to have a file where the next run is less than
1708 * 16 bits in size. In this case, the READ_HUFFSYM() macro used
1709 * in building the tables will exhaust the buffer, so we should
1710 * allow for this, but not allow those accidentally read bits to
1711 * be used (so we check that there are at least 16 bits
1712 * remaining - in this boundary case they aren't really part of
1713 * the compressed data)
1715 if (inpos > (endinp+2) || bitsleft < 16) return DECR_ILLEGALDATA;
1718 while ((this_run = LZX(block_remaining)) > 0 && togo > 0) {
1719 if (this_run > togo) this_run = togo;
1720 togo -= this_run;
1721 LZX(block_remaining) -= this_run;
1723 /* apply 2^x-1 mask */
1724 window_posn &= window_size - 1;
1725 /* runs can't straddle the window wraparound */
1726 if ((window_posn + this_run) > window_size)
1727 return DECR_DATAFORMAT;
1729 switch (LZX(block_type)) {
1731 case LZX_BLOCKTYPE_VERBATIM:
1732 while (this_run > 0) {
1733 READ_HUFFSYM(MAINTREE, main_element);
1735 if (main_element < LZX_NUM_CHARS) {
1736 /* literal: 0 to LZX_NUM_CHARS-1 */
1737 window[window_posn++] = main_element;
1738 this_run--;
1740 else {
1741 /* match: LZX_NUM_CHARS + ((slot<<3) | length_header (3 bits)) */
1742 main_element -= LZX_NUM_CHARS;
1744 match_length = main_element & LZX_NUM_PRIMARY_LENGTHS;
1745 if (match_length == LZX_NUM_PRIMARY_LENGTHS) {
1746 READ_HUFFSYM(LENGTH, length_footer);
1747 match_length += length_footer;
1749 match_length += LZX_MIN_MATCH;
1751 match_offset = main_element >> 3;
1753 if (match_offset > 2) {
1754 /* not repeated offset */
1755 if (match_offset != 3) {
1756 extra = CAB(extra_bits)[match_offset];
1757 READ_BITS(verbatim_bits, extra);
1758 match_offset = CAB(lzx_position_base)[match_offset]
1759 - 2 + verbatim_bits;
1761 else {
1762 match_offset = 1;
1765 /* update repeated offset LRU queue */
1766 R2 = R1; R1 = R0; R0 = match_offset;
1768 else if (match_offset == 0) {
1769 match_offset = R0;
1771 else if (match_offset == 1) {
1772 match_offset = R1;
1773 R1 = R0; R0 = match_offset;
1775 else /* match_offset == 2 */ {
1776 match_offset = R2;
1777 R2 = R0; R0 = match_offset;
1780 rundest = window + window_posn;
1781 this_run -= match_length;
1783 /* copy any wrapped around source data */
1784 if (window_posn >= match_offset) {
1785 /* no wrap */
1786 runsrc = rundest - match_offset;
1787 } else {
1788 runsrc = rundest + (window_size - match_offset);
1789 copy_length = match_offset - window_posn;
1790 if (copy_length < match_length) {
1791 match_length -= copy_length;
1792 window_posn += copy_length;
1793 while (copy_length-- > 0) *rundest++ = *runsrc++;
1794 runsrc = window;
1797 window_posn += match_length;
1799 /* copy match data - no worries about destination wraps */
1800 while (match_length-- > 0) *rundest++ = *runsrc++;
1803 break;
1805 case LZX_BLOCKTYPE_ALIGNED:
1806 while (this_run > 0) {
1807 READ_HUFFSYM(MAINTREE, main_element);
1809 if (main_element < LZX_NUM_CHARS) {
1810 /* literal: 0 to LZX_NUM_CHARS-1 */
1811 window[window_posn++] = main_element;
1812 this_run--;
1814 else {
1815 /* match: LZX_NUM_CHARS + ((slot<<3) | length_header (3 bits)) */
1816 main_element -= LZX_NUM_CHARS;
1818 match_length = main_element & LZX_NUM_PRIMARY_LENGTHS;
1819 if (match_length == LZX_NUM_PRIMARY_LENGTHS) {
1820 READ_HUFFSYM(LENGTH, length_footer);
1821 match_length += length_footer;
1823 match_length += LZX_MIN_MATCH;
1825 match_offset = main_element >> 3;
1827 if (match_offset > 2) {
1828 /* not repeated offset */
1829 extra = CAB(extra_bits)[match_offset];
1830 match_offset = CAB(lzx_position_base)[match_offset] - 2;
1831 if (extra > 3) {
1832 /* verbatim and aligned bits */
1833 extra -= 3;
1834 READ_BITS(verbatim_bits, extra);
1835 match_offset += (verbatim_bits << 3);
1836 READ_HUFFSYM(ALIGNED, aligned_bits);
1837 match_offset += aligned_bits;
1839 else if (extra == 3) {
1840 /* aligned bits only */
1841 READ_HUFFSYM(ALIGNED, aligned_bits);
1842 match_offset += aligned_bits;
1844 else if (extra > 0) { /* extra==1, extra==2 */
1845 /* verbatim bits only */
1846 READ_BITS(verbatim_bits, extra);
1847 match_offset += verbatim_bits;
1849 else /* extra == 0 */ {
1850 /* ??? */
1851 match_offset = 1;
1854 /* update repeated offset LRU queue */
1855 R2 = R1; R1 = R0; R0 = match_offset;
1857 else if (match_offset == 0) {
1858 match_offset = R0;
1860 else if (match_offset == 1) {
1861 match_offset = R1;
1862 R1 = R0; R0 = match_offset;
1864 else /* match_offset == 2 */ {
1865 match_offset = R2;
1866 R2 = R0; R0 = match_offset;
1869 rundest = window + window_posn;
1870 this_run -= match_length;
1872 /* copy any wrapped around source data */
1873 if (window_posn >= match_offset) {
1874 /* no wrap */
1875 runsrc = rundest - match_offset;
1876 } else {
1877 runsrc = rundest + (window_size - match_offset);
1878 copy_length = match_offset - window_posn;
1879 if (copy_length < match_length) {
1880 match_length -= copy_length;
1881 window_posn += copy_length;
1882 while (copy_length-- > 0) *rundest++ = *runsrc++;
1883 runsrc = window;
1886 window_posn += match_length;
1888 /* copy match data - no worries about destination wraps */
1889 while (match_length-- > 0) *rundest++ = *runsrc++;
1892 break;
1894 case LZX_BLOCKTYPE_UNCOMPRESSED:
1895 if ((inpos + this_run) > endinp) return DECR_ILLEGALDATA;
1896 memcpy(window + window_posn, inpos, (size_t) this_run);
1897 inpos += this_run; window_posn += this_run;
1898 break;
1900 default:
1901 return DECR_ILLEGALDATA; /* might as well */
1907 if (togo != 0) return DECR_ILLEGALDATA;
1908 memcpy(CAB(outbuf), window + ((!window_posn) ? window_size : window_posn) -
1909 outlen, (size_t) outlen);
1911 LZX(window_posn) = window_posn;
1912 LZX(R0) = R0;
1913 LZX(R1) = R1;
1914 LZX(R2) = R2;
1916 /* intel E8 decoding */
1917 if ((LZX(frames_read)++ < 32768) && LZX(intel_filesize) != 0) {
1918 if (outlen <= 6 || !LZX(intel_started)) {
1919 LZX(intel_curpos) += outlen;
1921 else {
1922 cab_UBYTE *data = CAB(outbuf);
1923 cab_UBYTE *dataend = data + outlen - 10;
1924 cab_LONG curpos = LZX(intel_curpos);
1925 cab_LONG filesize = LZX(intel_filesize);
1926 cab_LONG abs_off, rel_off;
1928 LZX(intel_curpos) = curpos + outlen;
1930 while (data < dataend) {
1931 if (*data++ != 0xE8) { curpos++; continue; }
1932 abs_off = data[0] | (data[1]<<8) | (data[2]<<16) | (data[3]<<24);
1933 if ((abs_off >= -curpos) && (abs_off < filesize)) {
1934 rel_off = (abs_off >= 0) ? abs_off - curpos : abs_off + filesize;
1935 data[0] = (cab_UBYTE) rel_off;
1936 data[1] = (cab_UBYTE) (rel_off >> 8);
1937 data[2] = (cab_UBYTE) (rel_off >> 16);
1938 data[3] = (cab_UBYTE) (rel_off >> 24);
1940 data += 4;
1941 curpos += 5;
1945 return DECR_OK;
1948 /*********************************************************
1949 * find_cabs_in_file (internal)
1951 struct cabinet *find_cabs_in_file(LPCSTR name, cab_UBYTE search_buf[])
1953 struct cabinet *cab, *cab2, *firstcab = NULL, *linkcab = NULL;
1954 cab_UBYTE *pstart = &search_buf[0], *pend, *p;
1955 cab_off_t offset, caboff, cablen = 0, foffset = 0, filelen, length;
1956 int state = 0, found = 0, ok = 0;
1958 TRACE("(name == %s)\n", debugstr_a(name));
1960 /* open the file and search for cabinet headers */
1961 if ((cab = (struct cabinet *) calloc(1, sizeof(struct cabinet)))) {
1962 cab->filename = name;
1963 if (cabinet_open(cab)) {
1964 filelen = cab->filelen;
1965 for (offset = 0; (offset < filelen); offset += length) {
1966 /* search length is either the full length of the search buffer,
1967 * or the amount of data remaining to the end of the file,
1968 * whichever is less.
1970 length = filelen - offset;
1971 if (length > CAB_SEARCH_SIZE) length = CAB_SEARCH_SIZE;
1973 /* fill the search buffer with data from disk */
1974 if (!cabinet_read(cab, search_buf, length)) break;
1976 /* read through the entire buffer. */
1977 p = pstart;
1978 pend = &search_buf[length];
1979 while (p < pend) {
1980 switch (state) {
1981 /* starting state */
1982 case 0:
1983 /* we spend most of our time in this while loop, looking for
1984 * a leading 'M' of the 'MSCF' signature
1986 while (*p++ != 0x4D && p < pend);
1987 if (p < pend) state = 1; /* if we found tht 'M', advance state */
1988 break;
1990 /* verify that the next 3 bytes are 'S', 'C' and 'F' */
1991 case 1: state = (*p++ == 0x53) ? 2 : 0; break;
1992 case 2: state = (*p++ == 0x43) ? 3 : 0; break;
1993 case 3: state = (*p++ == 0x46) ? 4 : 0; break;
1995 /* we don't care about bytes 4-7 */
1996 /* bytes 8-11 are the overall length of the cabinet */
1997 case 8: cablen = *p++; state++; break;
1998 case 9: cablen |= *p++ << 8; state++; break;
1999 case 10: cablen |= *p++ << 16; state++; break;
2000 case 11: cablen |= *p++ << 24; state++; break;
2002 /* we don't care about bytes 12-15 */
2003 /* bytes 16-19 are the offset within the cabinet of the filedata */
2004 case 16: foffset = *p++; state++; break;
2005 case 17: foffset |= *p++ << 8; state++; break;
2006 case 18: foffset |= *p++ << 16; state++; break;
2007 case 19: foffset |= *p++ << 24;
2008 /* now we have received 20 bytes of potential cab header. */
2009 /* work out the offset in the file of this potential cabinet */
2010 caboff = offset + (p-pstart) - 20;
2012 /* check that the files offset is less than the alleged length
2013 * of the cabinet, and that the offset + the alleged length are
2014 * 'roughly' within the end of overall file length
2016 if ((foffset < cablen) &&
2017 ((caboff + foffset) < (filelen + 32)) &&
2018 ((caboff + cablen) < (filelen + 32)) )
2020 /* found a potential result - try loading it */
2021 found++;
2022 cab2 = load_cab_offset(name, caboff);
2023 if (cab2) {
2024 /* success */
2025 ok++;
2027 /* cause the search to restart after this cab's data. */
2028 offset = caboff + cablen;
2029 if (offset < cab->filelen) cabinet_seek(cab, offset);
2030 length = 0;
2031 p = pend;
2033 /* link the cab into the list */
2034 if (linkcab == NULL) firstcab = cab2;
2035 else linkcab->next = cab2;
2036 linkcab = cab2;
2039 state = 0;
2040 break;
2041 default:
2042 p++, state++; break;
2046 cabinet_close(cab);
2048 free(cab);
2051 /* if there were cabinets that were found but are not ok, point this out */
2052 if (found > ok) {
2053 WARN("%s: found %d bad cabinets\n", debugstr_a(name), found-ok);
2056 /* if no cabinets were found, let the user know */
2057 if (!firstcab) {
2058 WARN("%s: not a Microsoft cabinet file.\n", debugstr_a(name));
2060 return firstcab;
2063 /***********************************************************************
2064 * find_cabinet_file (internal)
2066 * tries to find *cabname, from the directory path of origcab, correcting the
2067 * case of *cabname if necessary, If found, writes back to *cabname.
2069 void find_cabinet_file(char **cabname, LPCSTR origcab) {
2071 char *tail, *cab, *name, *nextpart, nametmp[MAX_PATH], *filepart;
2072 int found = 0;
2074 TRACE("(*cabname == ^%p, origcab == %s)\n", cabname ? *cabname : NULL, debugstr_a(origcab));
2076 /* ensure we have a cabinet name at all */
2077 if (!(name = *cabname)) {
2078 WARN("no cabinet name at all\n");
2081 /* find if there's a directory path in the origcab */
2082 tail = origcab ? max(strrchr(origcab, '/'), strrchr(origcab, '\\')) : NULL;
2084 if ((cab = (char *) malloc(MAX_PATH))) {
2085 /* add the directory path from the original cabinet name */
2086 if (tail) {
2087 memcpy(cab, origcab, tail - origcab);
2088 cab[tail - origcab] = '\0';
2089 } else {
2090 /* default directory path of '.' */
2091 cab[0] = '.';
2092 cab[1] = '\0';
2095 do {
2096 TRACE("trying cab == %s\n", debugstr_a(cab));
2098 /* we don't want null cabinet filenames */
2099 if (name[0] == '\0') {
2100 WARN("null cab name\n");
2101 break;
2104 /* if there is a directory component in the cabinet name,
2105 * look for that alone first
2107 nextpart = strchr(name, '\\');
2108 if (nextpart) *nextpart = '\0';
2110 found = SearchPathA(cab, name, NULL, MAX_PATH, nametmp, &filepart);
2112 /* if the component was not found, look for it in the current dir */
2113 if (!found) {
2114 found = SearchPathA(".", name, NULL, MAX_PATH, nametmp, &filepart);
2117 if (found)
2118 TRACE("found: %s\n", debugstr_a(nametmp));
2119 else
2120 TRACE("not found.\n");
2122 /* restore the real name and skip to the next directory component
2123 * or actual cabinet name
2125 if (nextpart) *nextpart = '\\', name = &nextpart[1];
2127 /* while there is another directory component, and while we
2128 * successfully found the current component
2130 } while (nextpart && found);
2132 /* if we found the cabinet, change the next cabinet's name.
2133 * otherwise, pretend nothing happened
2135 if (found) {
2136 free((void *) *cabname);
2137 *cabname = cab;
2138 strncpy(cab, nametmp, found+1);
2139 TRACE("result: %s\n", debugstr_a(cab));
2140 } else {
2141 free((void *) cab);
2142 TRACE("result: nothing\n");
2147 /************************************************************************
2148 * process_files (internal)
2150 * this does the tricky job of running through every file in the cabinet,
2151 * including spanning cabinets, and working out which file is in which
2152 * folder in which cabinet. It also throws out the duplicate file entries
2153 * that appear in spanning cabinets. There is memory leakage here because
2154 * those entries are not freed. See the XAD CAB client (function CAB_GetInfo
2155 * in CAB.c) for an implementation of this that correctly frees the discarded
2156 * file entries.
2158 struct cab_file *process_files(struct cabinet *basecab) {
2159 struct cabinet *cab;
2160 struct cab_file *outfi = NULL, *linkfi = NULL, *nextfi, *fi, *cfi;
2161 struct cab_folder *fol, *firstfol, *lastfol = NULL, *predfol;
2162 int i, mergeok;
2164 FIXME("(basecab == ^%p): Memory leak.\n", basecab);
2166 for (cab = basecab; cab; cab = cab->nextcab) {
2167 /* firstfol = first folder in this cabinet */
2168 /* lastfol = last folder in this cabinet */
2169 /* predfol = last folder in previous cabinet (or NULL if first cabinet) */
2170 predfol = lastfol;
2171 firstfol = cab->folders;
2172 for (lastfol = firstfol; lastfol->next;) lastfol = lastfol->next;
2173 mergeok = 1;
2175 for (fi = cab->files; fi; fi = nextfi) {
2176 i = fi->index;
2177 nextfi = fi->next;
2179 if (i < cffileCONTINUED_FROM_PREV) {
2180 for (fol = firstfol; fol && i--; ) fol = fol->next;
2181 fi->folder = fol; /* NULL if an invalid folder index */
2183 else {
2184 /* folder merging */
2185 if (i == cffileCONTINUED_TO_NEXT
2186 || i == cffileCONTINUED_PREV_AND_NEXT) {
2187 if (cab->nextcab && !lastfol->contfile) lastfol->contfile = fi;
2190 if (i == cffileCONTINUED_FROM_PREV
2191 || i == cffileCONTINUED_PREV_AND_NEXT) {
2192 /* these files are to be continued in yet another
2193 * cabinet, don't merge them in just yet */
2194 if (i == cffileCONTINUED_PREV_AND_NEXT) mergeok = 0;
2196 /* only merge once per cabinet */
2197 if (predfol) {
2198 if ((cfi = predfol->contfile)
2199 && (cfi->offset == fi->offset)
2200 && (cfi->length == fi->length)
2201 && (strcmp(cfi->filename, fi->filename) == 0)
2202 && (predfol->comp_type == firstfol->comp_type)) {
2203 /* increase the number of splits */
2204 if ((i = ++(predfol->num_splits)) > CAB_SPLITMAX) {
2205 mergeok = 0;
2206 ERR("%s: internal error: CAB_SPLITMAX exceeded. please report this to wine-devel@winehq.org)\n",
2207 debugstr_a(basecab->filename));
2209 else {
2210 /* copy information across from the merged folder */
2211 predfol->offset[i] = firstfol->offset[0];
2212 predfol->cab[i] = firstfol->cab[0];
2213 predfol->next = firstfol->next;
2214 predfol->contfile = firstfol->contfile;
2216 if (firstfol == lastfol) lastfol = predfol;
2217 firstfol = predfol;
2218 predfol = NULL; /* don't merge again within this cabinet */
2221 else {
2222 /* if the folders won't merge, don't add their files */
2223 mergeok = 0;
2227 if (mergeok) fi->folder = firstfol;
2231 if (fi->folder) {
2232 if (linkfi) linkfi->next = fi; else outfi = fi;
2233 linkfi = fi;
2235 } /* for (fi= .. */
2236 } /* for (cab= ...*/
2238 return outfi;
2241 /****************************************************************
2242 * convertUTF (internal)
2244 * translate UTF -> ASCII
2246 * UTF translates two-byte unicode characters into 1, 2 or 3 bytes.
2247 * %000000000xxxxxxx -> %0xxxxxxx
2248 * %00000xxxxxyyyyyy -> %110xxxxx %10yyyyyy
2249 * %xxxxyyyyyyzzzzzz -> %1110xxxx %10yyyyyy %10zzzzzz
2251 * Therefore, the inverse is as follows:
2252 * First char:
2253 * 0x00 - 0x7F = one byte char
2254 * 0x80 - 0xBF = invalid
2255 * 0xC0 - 0xDF = 2 byte char (next char only 0x80-0xBF is valid)
2256 * 0xE0 - 0xEF = 3 byte char (next 2 chars only 0x80-0xBF is valid)
2257 * 0xF0 - 0xFF = invalid
2259 * FIXME: use a winapi to do this
2261 int convertUTF(cab_UBYTE *in) {
2262 cab_UBYTE c, *out = in, *end = in + strlen((char *) in) + 1;
2263 cab_ULONG x;
2265 do {
2266 /* read unicode character */
2267 if ((c = *in++) < 0x80) x = c;
2268 else {
2269 if (c < 0xC0) return 0;
2270 else if (c < 0xE0) {
2271 x = (c & 0x1F) << 6;
2272 if ((c = *in++) < 0x80 || c > 0xBF) return 0; else x |= (c & 0x3F);
2274 else if (c < 0xF0) {
2275 x = (c & 0xF) << 12;
2276 if ((c = *in++) < 0x80 || c > 0xBF) return 0; else x |= (c & 0x3F)<<6;
2277 if ((c = *in++) < 0x80 || c > 0xBF) return 0; else x |= (c & 0x3F);
2279 else return 0;
2282 /* terrible unicode -> ASCII conversion */
2283 if (x > 127) x = '_';
2285 if (in > end) return 0; /* just in case */
2286 } while ((*out++ = (cab_UBYTE) x));
2287 return 1;
2290 /****************************************************
2291 * NONEdecompress (internal)
2293 int NONEdecompress(int inlen, int outlen, cab_decomp_state *decomp_state)
2295 if (inlen != outlen) return DECR_ILLEGALDATA;
2296 memcpy(CAB(outbuf), CAB(inbuf), (size_t) inlen);
2297 return DECR_OK;
2300 /**************************************************
2301 * checksum (internal)
2303 cab_ULONG checksum(cab_UBYTE *data, cab_UWORD bytes, cab_ULONG csum) {
2304 int len;
2305 cab_ULONG ul = 0;
2307 for (len = bytes >> 2; len--; data += 4) {
2308 csum ^= ((data[0]) | (data[1]<<8) | (data[2]<<16) | (data[3]<<24));
2311 switch (bytes & 3) {
2312 case 3: ul |= *data++ << 16;
2313 case 2: ul |= *data++ << 8;
2314 case 1: ul |= *data;
2316 csum ^= ul;
2318 return csum;
2321 /**********************************************************
2322 * decompress (internal)
2324 int decompress(struct cab_file *fi, int savemode, int fix, cab_decomp_state *decomp_state)
2326 cab_ULONG bytes = savemode ? fi->length : fi->offset - CAB(offset);
2327 struct cabinet *cab = CAB(current)->cab[CAB(split)];
2328 cab_UBYTE buf[cfdata_SIZEOF], *data;
2329 cab_UWORD inlen, len, outlen, cando;
2330 cab_ULONG cksum;
2331 cab_LONG err;
2333 TRACE("(fi == ^%p, savemode == %d, fix == %d)\n", fi, savemode, fix);
2335 while (bytes > 0) {
2336 /* cando = the max number of bytes we can do */
2337 cando = CAB(outlen);
2338 if (cando > bytes) cando = bytes;
2340 /* if cando != 0 */
2341 if (cando && savemode)
2342 file_write(fi, CAB(outpos), cando);
2344 CAB(outpos) += cando;
2345 CAB(outlen) -= cando;
2346 bytes -= cando; if (!bytes) break;
2348 /* we only get here if we emptied the output buffer */
2350 /* read data header + data */
2351 inlen = outlen = 0;
2352 while (outlen == 0) {
2353 /* read the block header, skip the reserved part */
2354 if (!cabinet_read(cab, buf, cfdata_SIZEOF)) return DECR_INPUT;
2355 cabinet_skip(cab, cab->block_resv);
2357 /* we shouldn't get blocks over CAB_INPUTMAX in size */
2358 data = CAB(inbuf) + inlen;
2359 len = EndGetI16(buf+cfdata_CompressedSize);
2360 inlen += len;
2361 if (inlen > CAB_INPUTMAX) return DECR_INPUT;
2362 if (!cabinet_read(cab, data, len)) return DECR_INPUT;
2364 /* clear two bytes after read-in data */
2365 data[len+1] = data[len+2] = 0;
2367 /* perform checksum test on the block (if one is stored) */
2368 cksum = EndGetI32(buf+cfdata_CheckSum);
2369 if (cksum && cksum != checksum(buf+4, 4, checksum(data, len, 0))) {
2370 /* checksum is wrong */
2371 if (fix && ((fi->folder->comp_type & cffoldCOMPTYPE_MASK)
2372 == cffoldCOMPTYPE_MSZIP))
2374 WARN("%s: checksum failed\n", debugstr_a(fi->filename));
2376 else {
2377 return DECR_CHECKSUM;
2381 /* outlen=0 means this block was part of a split block */
2382 outlen = EndGetI16(buf+cfdata_UncompressedSize);
2383 if (outlen == 0) {
2384 cabinet_close(cab);
2385 cab = CAB(current)->cab[++CAB(split)];
2386 if (!cabinet_open(cab)) return DECR_INPUT;
2387 cabinet_seek(cab, CAB(current)->offset[CAB(split)]);
2391 /* decompress block */
2392 if ((err = CAB(decompress)(inlen, outlen, decomp_state))) {
2393 if (fix && ((fi->folder->comp_type & cffoldCOMPTYPE_MASK)
2394 == cffoldCOMPTYPE_MSZIP))
2396 ERR("%s: failed decrunching block\n", debugstr_a(fi->filename));
2398 else {
2399 return err;
2402 CAB(outlen) = outlen;
2403 CAB(outpos) = CAB(outbuf);
2406 return DECR_OK;
2409 /****************************************************************
2410 * extract_file (internal)
2412 * workhorse to extract a particular file from a cab
2414 void extract_file(struct cab_file *fi, int lower, int fix, LPCSTR dir, cab_decomp_state *decomp_state)
2416 struct cab_folder *fol = fi->folder, *oldfol = CAB(current);
2417 cab_LONG err = DECR_OK;
2419 TRACE("(fi == ^%p, lower == %d, fix == %d, dir == %s)\n", fi, lower, fix, debugstr_a(dir));
2421 /* is a change of folder needed? do we need to reset the current folder? */
2422 if (fol != oldfol || fi->offset < CAB(offset)) {
2423 cab_UWORD comptype = fol->comp_type;
2424 int ct1 = comptype & cffoldCOMPTYPE_MASK;
2425 int ct2 = oldfol ? (oldfol->comp_type & cffoldCOMPTYPE_MASK) : 0;
2427 /* if the archiver has changed, call the old archiver's free() function */
2428 if (ct1 != ct2) {
2429 switch (ct2) {
2430 case cffoldCOMPTYPE_LZX:
2431 if (LZX(window)) {
2432 free(LZX(window));
2433 LZX(window) = NULL;
2435 break;
2436 case cffoldCOMPTYPE_QUANTUM:
2437 if (QTM(window)) {
2438 free(QTM(window));
2439 QTM(window) = NULL;
2441 break;
2445 switch (ct1) {
2446 case cffoldCOMPTYPE_NONE:
2447 CAB(decompress) = NONEdecompress;
2448 break;
2450 case cffoldCOMPTYPE_MSZIP:
2451 CAB(decompress) = ZIPdecompress;
2452 break;
2454 case cffoldCOMPTYPE_QUANTUM:
2455 CAB(decompress) = QTMdecompress;
2456 err = QTMinit((comptype >> 8) & 0x1f, (comptype >> 4) & 0xF, decomp_state);
2457 break;
2459 case cffoldCOMPTYPE_LZX:
2460 CAB(decompress) = LZXdecompress;
2461 err = LZXinit((comptype >> 8) & 0x1f, decomp_state);
2462 break;
2464 default:
2465 err = DECR_DATAFORMAT;
2467 if (err) goto exit_handler;
2469 /* initialisation OK, set current folder and reset offset */
2470 if (oldfol) cabinet_close(oldfol->cab[CAB(split)]);
2471 if (!cabinet_open(fol->cab[0])) goto exit_handler;
2472 cabinet_seek(fol->cab[0], fol->offset[0]);
2473 CAB(current) = fol;
2474 CAB(offset) = 0;
2475 CAB(outlen) = 0; /* discard existing block */
2476 CAB(split) = 0;
2479 if (fi->offset > CAB(offset)) {
2480 /* decode bytes and send them to /dev/null */
2481 if ((err = decompress(fi, 0, fix, decomp_state))) goto exit_handler;
2482 CAB(offset) = fi->offset;
2485 if (!file_open(fi, lower, dir)) return;
2486 err = decompress(fi, 1, fix, decomp_state);
2487 if (err) CAB(current) = NULL; else CAB(offset) += fi->length;
2488 file_close(fi);
2490 exit_handler:
2491 if (err) {
2492 const char *errmsg;
2493 const char *cabname;
2494 switch (err) {
2495 case DECR_NOMEMORY:
2496 errmsg = "out of memory!\n"; break;
2497 case DECR_ILLEGALDATA:
2498 errmsg = "%s: illegal or corrupt data\n"; break;
2499 case DECR_DATAFORMAT:
2500 errmsg = "%s: unsupported data format\n"; break;
2501 case DECR_CHECKSUM:
2502 errmsg = "%s: checksum error\n"; break;
2503 case DECR_INPUT:
2504 errmsg = "%s: input error\n"; break;
2505 case DECR_OUTPUT:
2506 errmsg = "%s: output error\n"; break;
2507 default:
2508 errmsg = "%s: unknown error (BUG)\n";
2511 if (CAB(current)) {
2512 cabname = (CAB(current)->cab[CAB(split)]->filename);
2514 else {
2515 cabname = (fi->folder->cab[0]->filename);
2518 ERR(errmsg, cabname);
2522 /*********************************************************
2523 * print_fileinfo (internal)
2525 void print_fileinfo(struct cab_file *fi) {
2526 int d = fi->date, t = fi->time;
2527 char *fname = NULL;
2529 if (fi->attribs & cffile_A_NAME_IS_UTF) {
2530 fname = malloc(strlen(fi->filename) + 1);
2531 if (fname) {
2532 strcpy(fname, fi->filename);
2533 convertUTF((cab_UBYTE *) fname);
2537 TRACE("%9u | %02d.%02d.%04d %02d:%02d:%02d | %s\n",
2538 fi->length,
2539 d & 0x1f, (d>>5) & 0xf, (d>>9) + 1980,
2540 t >> 11, (t>>5) & 0x3f, (t << 1) & 0x3e,
2541 fname ? fname : fi->filename
2544 if (fname) free(fname);
2547 /****************************************************************************
2548 * process_cabinet (internal)
2550 * called to simply "extract" a cabinet file. Will find every cabinet file
2551 * in that file, search for every chained cabinet attached to those cabinets,
2552 * and will either extract the cabinets, or ? (call a callback?)
2554 * PARAMS
2555 * cabname [I] name of the cabinet file to extract
2556 * dir [I] directory to extract to
2557 * fix [I] attempt to process broken cabinets
2558 * lower [I] ? (lower case something or other?)
2559 * dest [O]
2561 * RETURNS
2562 * Success: TRUE
2563 * Failure: FALSE
2565 BOOL process_cabinet(LPCSTR cabname, LPCSTR dir, BOOL fix, BOOL lower, EXTRACTdest *dest)
2567 struct cabinet *basecab, *cab, *cab1, *cab2;
2568 struct cab_file *filelist, *fi;
2569 struct ExtractFileList **destlistptr = &(dest->filelist);
2571 /* The first result of a search will be returned, and
2572 * the remaining results will be chained to it via the cab->next structure
2573 * member.
2575 cab_UBYTE search_buf[CAB_SEARCH_SIZE];
2577 cab_decomp_state decomp_state_local;
2578 cab_decomp_state *decomp_state = &decomp_state_local;
2580 /* has the list-mode header been seen before? */
2581 int viewhdr = 0;
2583 ZeroMemory(decomp_state, sizeof(cab_decomp_state));
2585 TRACE("Extract %s\n", debugstr_a(cabname));
2587 /* load the file requested */
2588 basecab = find_cabs_in_file(cabname, search_buf);
2589 if (!basecab) return FALSE;
2591 /* iterate over all cabinets found in that file */
2592 for (cab = basecab; cab; cab=cab->next) {
2594 /* bi-directionally load any spanning cabinets -- backwards */
2595 for (cab1 = cab; cab1->flags & cfheadPREV_CABINET; cab1 = cab1->prevcab) {
2596 TRACE("%s: extends backwards to %s (%s)\n", debugstr_a(cabname),
2597 debugstr_a(cab1->prevname), debugstr_a(cab1->previnfo));
2598 find_cabinet_file(&(cab1->prevname), cabname);
2599 if (!(cab1->prevcab = load_cab_offset(cab1->prevname, 0))) {
2600 ERR("%s: can't read previous cabinet %s\n", debugstr_a(cabname), debugstr_a(cab1->prevname));
2601 break;
2603 cab1->prevcab->nextcab = cab1;
2606 /* bi-directionally load any spanning cabinets -- forwards */
2607 for (cab2 = cab; cab2->flags & cfheadNEXT_CABINET; cab2 = cab2->nextcab) {
2608 TRACE("%s: extends to %s (%s)\n", debugstr_a(cabname),
2609 debugstr_a(cab2->nextname), debugstr_a(cab2->nextinfo));
2610 find_cabinet_file(&(cab2->nextname), cabname);
2611 if (!(cab2->nextcab = load_cab_offset(cab2->nextname, 0))) {
2612 ERR("%s: can't read next cabinet %s\n", debugstr_a(cabname), debugstr_a(cab2->nextname));
2613 break;
2615 cab2->nextcab->prevcab = cab2;
2618 filelist = process_files(cab1);
2619 CAB(current) = NULL;
2621 if (!viewhdr) {
2622 TRACE("File size | Date Time | Name\n");
2623 TRACE("----------+---------------------+-------------\n");
2624 viewhdr = 1;
2626 for (fi = filelist; fi; fi = fi->next) {
2627 print_fileinfo(fi);
2628 dest->filecount++;
2630 TRACE("Beginning Extraction...\n");
2631 for (fi = filelist; fi; fi = fi->next) {
2632 TRACE(" extracting: %s\n", debugstr_a(fi->filename));
2633 extract_file(fi, lower, fix, dir, decomp_state);
2634 sprintf(dest->lastfile, "%s%s%s",
2635 strlen(dest->directory) ? dest->directory : "",
2636 strlen(dest->directory) ? "\\": "",
2637 fi->filename);
2638 *destlistptr = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY,
2639 sizeof(struct ExtractFileList));
2640 if(*destlistptr) {
2641 (*destlistptr)->unknown = TRUE; /* FIXME: were do we get the value? */
2642 (*destlistptr)->filename = HeapAlloc(GetProcessHeap(), 0, (
2643 strlen(fi->filename)+1));
2644 if((*destlistptr)->filename)
2645 lstrcpyA((*destlistptr)->filename, fi->filename);
2646 destlistptr = &((*destlistptr)->next);
2651 TRACE("Finished processing cabinet.\n");
2653 return TRUE;