1 /* $Id: tif_lzw.c,v 1.46 2014-11-20 16:47:21 erouault Exp $ */
2
3 /*
4 * Copyright (c) 1988-1997 Sam Leffler
5 * Copyright (c) 1991-1997 Silicon Graphics, Inc.
6 *
7 * Permission to use, copy, modify, distribute, and sell this software and
8 * its documentation for any purpose is hereby granted without fee, provided
9 * that (i) the above copyright notices and this permission notice appear in
10 * all copies of the software and related documentation, and (ii) the names of
11 * Sam Leffler and Silicon Graphics may not be used in any advertising or
12 * publicity relating to the software without the specific, prior written
13 * permission of Sam Leffler and Silicon Graphics.
14 *
15 * THE SOFTWARE IS PROVIDED "AS-IS" AND WITHOUT WARRANTY OF ANY KIND,
16 * EXPRESS, IMPLIED OR OTHERWISE, INCLUDING WITHOUT LIMITATION, ANY
17 * WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
18 *
19 * IN NO EVENT SHALL SAM LEFFLER OR SILICON GRAPHICS BE LIABLE FOR
20 * ANY SPECIAL, INCIDENTAL, INDIRECT OR CONSEQUENTIAL DAMAGES OF ANY KIND,
21 * OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
22 * WHETHER OR NOT ADVISED OF THE POSSIBILITY OF DAMAGE, AND ON ANY THEORY OF
23 * LIABILITY, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
24 * OF THIS SOFTWARE.
25 */
26
27 #include "tiffiop.h"
28 #ifdef LZW_SUPPORT
29 /*
30 * TIFF Library.
31 * Rev 5.0 Lempel-Ziv & Welch Compression Support
32 *
33 * This code is derived from the compress program whose code is
34 * derived from software contributed to Berkeley by James A. Woods,
35 * derived from original work by Spencer Thomas and Joseph Orost.
36 *
37 * The original Berkeley copyright notice appears below in its entirety.
38 */
39 #include "tif_predict.h"
40
41 #include <stdio.h>
42
43 /*
44 * NB: The 5.0 spec describes a different algorithm than Aldus
45 * implements. Specifically, Aldus does code length transitions
46 * one code earlier than should be done (for real LZW).
47 * Earlier versions of this library implemented the correct
48 * LZW algorithm, but emitted codes in a bit order opposite
49 * to the TIFF spec. Thus, to maintain compatibility w/ Aldus
50 * we interpret MSB-LSB ordered codes to be images written w/
51 * old versions of this library, but otherwise adhere to the
52 * Aldus "off by one" algorithm.
53 *
54 * Future revisions to the TIFF spec are expected to "clarify this issue".
55 */
56 #define LZW_COMPAT /* include backwards compatibility code */
57 /*
58 * Each strip of data is supposed to be terminated by a CODE_EOI.
59 * If the following #define is included, the decoder will also
60 * check for end-of-strip w/o seeing this code. This makes the
61 * library more robust, but also slower.
62 */
63 #define LZW_CHECKEOS /* include checks for strips w/o EOI code */
64
65 #define MAXCODE(n) ((1L<<(n))-1)
66 /*
67 * The TIFF spec specifies that encoded bit
68 * strings range from 9 to 12 bits.
69 */
70 #define BITS_MIN 9 /* start with 9 bits */
71 #define BITS_MAX 12 /* max of 12 bit strings */
72 /* predefined codes */
73 #define CODE_CLEAR 256 /* code to clear string table */
74 #define CODE_EOI 257 /* end-of-information code */
75 #define CODE_FIRST 258 /* first free code entry */
76 #define CODE_MAX MAXCODE(BITS_MAX)
77 #define HSIZE 9001L /* 91% occupancy */
78 #define HSHIFT (13-8)
79 #ifdef LZW_COMPAT
80 /* NB: +1024 is for compatibility with old files */
81 #define CSIZE (MAXCODE(BITS_MAX)+1024L)
82 #else
83 #define CSIZE (MAXCODE(BITS_MAX)+1L)
84 #endif
85
86 /*
87 * State block for each open TIFF file using LZW
88 * compression/decompression. Note that the predictor
89 * state block must be first in this data structure.
90 */
91 typedef struct {
92 TIFFPredictorState predict; /* predictor super class */
93
94 unsigned short nbits; /* # of bits/code */
95 unsigned short maxcode; /* maximum code for lzw_nbits */
96 unsigned short free_ent; /* next free entry in hash table */
97 long nextdata; /* next bits of i/o */
98 long nextbits; /* # of valid bits in lzw_nextdata */
99
100 int rw_mode; /* preserve rw_mode from init */
101 } LZWBaseState;
102
103 #define lzw_nbits base.nbits
104 #define lzw_maxcode base.maxcode
105 #define lzw_free_ent base.free_ent
106 #define lzw_nextdata base.nextdata
107 #define lzw_nextbits base.nextbits
108
109 /*
110 * Encoding-specific state.
111 */
112 typedef uint16 hcode_t; /* codes fit in 16 bits */
113 typedef struct {
114 long hash;
115 hcode_t code;
116 } hash_t;
117
118 /*
119 * Decoding-specific state.
120 */
121 typedef struct code_ent {
122 struct code_ent *next;
123 unsigned short length; /* string len, including this token */
124 unsigned char value; /* data value */
125 unsigned char firstchar; /* first token of string */
126 } code_t;
127
128 typedef int (*decodeFunc)(TIFF*, uint8*, tmsize_t, uint16);
129
130 typedef struct {
131 LZWBaseState base;
132
133 /* Decoding specific data */
134 long dec_nbitsmask; /* lzw_nbits 1 bits, right adjusted */
135 long dec_restart; /* restart count */
136 #ifdef LZW_CHECKEOS
137 uint64 dec_bitsleft; /* available bits in raw data */
138 #endif
139 decodeFunc dec_decode; /* regular or backwards compatible */
140 code_t* dec_codep; /* current recognized code */
141 code_t* dec_oldcodep; /* previously recognized code */
142 code_t* dec_free_entp; /* next free entry */
143 code_t* dec_maxcodep; /* max available entry */
144 code_t* dec_codetab; /* kept separate for small machines */
145
146 /* Encoding specific data */
147 int enc_oldcode; /* last code encountered */
148 long enc_checkpoint; /* point at which to clear table */
149 #define CHECK_GAP 10000 /* enc_ratio check interval */
150 long enc_ratio; /* current compression ratio */
151 long enc_incount; /* (input) data bytes encoded */
152 long enc_outcount; /* encoded (output) bytes */
153 uint8* enc_rawlimit; /* bound on tif_rawdata buffer */
154 hash_t* enc_hashtab; /* kept separate for small machines */
155 } LZWCodecState;
156
157 #define LZWState(tif) ((LZWBaseState*) (tif)->tif_data)
158 #define DecoderState(tif) ((LZWCodecState*) LZWState(tif))
159 #define EncoderState(tif) ((LZWCodecState*) LZWState(tif))
160
161 static int LZWDecode(TIFF* tif, uint8* op0, tmsize_t occ0, uint16 s);
162 #ifdef LZW_COMPAT
163 static int LZWDecodeCompat(TIFF* tif, uint8* op0, tmsize_t occ0, uint16 s);
164 #endif
165 static void cl_hash(LZWCodecState*);
166
167 /*
168 * LZW Decoder.
169 */
170
171 #ifdef LZW_CHECKEOS
172 /*
173 * This check shouldn't be necessary because each
174 * strip is suppose to be terminated with CODE_EOI.
175 */
176 #define NextCode(_tif, _sp, _bp, _code, _get) { \
177 if ((_sp)->dec_bitsleft < (uint64)nbits) { \
178 TIFFWarningExt(_tif->tif_clientdata, module, \
179 "LZWDecode: Strip %d not terminated with EOI code", \
180 _tif->tif_curstrip); \
181 _code = CODE_EOI; \
182 } else { \
183 _get(_sp,_bp,_code); \
184 (_sp)->dec_bitsleft -= nbits; \
185 } \
186 }
187 #else
188 #define NextCode(tif, sp, bp, code, get) get(sp, bp, code)
189 #endif
190
191 static int
LZWFixupTags(TIFF * tif)192 LZWFixupTags(TIFF* tif)
193 {
194 (void) tif;
195 return (1);
196 }
197
198 static int
LZWSetupDecode(TIFF * tif)199 LZWSetupDecode(TIFF* tif)
200 {
201 static const char module[] = "LZWSetupDecode";
202 LZWCodecState* sp = DecoderState(tif);
203 int code;
204
205 if( sp == NULL )
206 {
207 /*
208 * Allocate state block so tag methods have storage to record
209 * values.
210 */
211 tif->tif_data = (uint8*) _TIFFmalloc(sizeof(LZWCodecState));
212 if (tif->tif_data == NULL)
213 {
214 TIFFErrorExt(tif->tif_clientdata, module, "No space for LZW state block");
215 return (0);
216 }
217
218 DecoderState(tif)->dec_codetab = NULL;
219 DecoderState(tif)->dec_decode = NULL;
220
221 /*
222 * Setup predictor setup.
223 */
224 (void) TIFFPredictorInit(tif);
225
226 sp = DecoderState(tif);
227 }
228
229 assert(sp != NULL);
230
231 if (sp->dec_codetab == NULL) {
232 sp->dec_codetab = (code_t*)_TIFFmalloc(CSIZE*sizeof (code_t));
233 if (sp->dec_codetab == NULL) {
234 TIFFErrorExt(tif->tif_clientdata, module,
235 "No space for LZW code table");
236 return (0);
237 }
238 /*
239 * Pre-load the table.
240 */
241 code = 255;
242 do {
243 sp->dec_codetab[code].value = code;
244 sp->dec_codetab[code].firstchar = code;
245 sp->dec_codetab[code].length = 1;
246 sp->dec_codetab[code].next = NULL;
247 } while (code--);
248 /*
249 * Zero-out the unused entries
250 */
251 _TIFFmemset(&sp->dec_codetab[CODE_CLEAR], 0,
252 (CODE_FIRST - CODE_CLEAR) * sizeof (code_t));
253 }
254 return (1);
255 }
256
257 /*
258 * Setup state for decoding a strip.
259 */
260 static int
LZWPreDecode(TIFF * tif,uint16 s)261 LZWPreDecode(TIFF* tif, uint16 s)
262 {
263 static const char module[] = "LZWPreDecode";
264 LZWCodecState *sp = DecoderState(tif);
265
266 (void) s;
267 assert(sp != NULL);
268 if( sp->dec_codetab == NULL )
269 {
270 tif->tif_setupdecode( tif );
271 if( sp->dec_codetab == NULL )
272 return (0);
273 }
274
275 /*
276 * Check for old bit-reversed codes.
277 */
278 if (tif->tif_rawdata[0] == 0 && (tif->tif_rawdata[1] & 0x1)) {
279 #ifdef LZW_COMPAT
280 if (!sp->dec_decode) {
281 TIFFWarningExt(tif->tif_clientdata, module,
282 "Old-style LZW codes, convert file");
283 /*
284 * Override default decoding methods with
285 * ones that deal with the old coding.
286 * Otherwise the predictor versions set
287 * above will call the compatibility routines
288 * through the dec_decode method.
289 */
290 tif->tif_decoderow = LZWDecodeCompat;
291 tif->tif_decodestrip = LZWDecodeCompat;
292 tif->tif_decodetile = LZWDecodeCompat;
293 /*
294 * If doing horizontal differencing, must
295 * re-setup the predictor logic since we
296 * switched the basic decoder methods...
297 */
298 (*tif->tif_setupdecode)(tif);
299 sp->dec_decode = LZWDecodeCompat;
300 }
301 sp->lzw_maxcode = MAXCODE(BITS_MIN);
302 #else /* !LZW_COMPAT */
303 if (!sp->dec_decode) {
304 TIFFErrorExt(tif->tif_clientdata, module,
305 "Old-style LZW codes not supported");
306 sp->dec_decode = LZWDecode;
307 }
308 return (0);
309 #endif/* !LZW_COMPAT */
310 } else {
311 sp->lzw_maxcode = MAXCODE(BITS_MIN)-1;
312 sp->dec_decode = LZWDecode;
313 }
314 sp->lzw_nbits = BITS_MIN;
315 sp->lzw_nextbits = 0;
316 sp->lzw_nextdata = 0;
317
318 sp->dec_restart = 0;
319 sp->dec_nbitsmask = MAXCODE(BITS_MIN);
320 #ifdef LZW_CHECKEOS
321 sp->dec_bitsleft = ((uint64)tif->tif_rawcc) << 3;
322 #endif
323 sp->dec_free_entp = sp->dec_codetab + CODE_FIRST;
324 /*
325 * Zero entries that are not yet filled in. We do
326 * this to guard against bogus input data that causes
327 * us to index into undefined entries. If you can
328 * come up with a way to safely bounds-check input codes
329 * while decoding then you can remove this operation.
330 */
331 _TIFFmemset(sp->dec_free_entp, 0, (CSIZE-CODE_FIRST)*sizeof (code_t));
332 sp->dec_oldcodep = &sp->dec_codetab[-1];
333 sp->dec_maxcodep = &sp->dec_codetab[sp->dec_nbitsmask-1];
334 return (1);
335 }
336
337 /*
338 * Decode a "hunk of data".
339 */
340 #define GetNextCode(sp, bp, code) { \
341 nextdata = (nextdata<<8) | *(bp)++; \
342 nextbits += 8; \
343 if (nextbits < nbits) { \
344 nextdata = (nextdata<<8) | *(bp)++; \
345 nextbits += 8; \
346 } \
347 code = (hcode_t)((nextdata >> (nextbits-nbits)) & nbitsmask); \
348 nextbits -= nbits; \
349 }
350
351 static void
codeLoop(TIFF * tif,const char * module)352 codeLoop(TIFF* tif, const char* module)
353 {
354 TIFFErrorExt(tif->tif_clientdata, module,
355 "Bogus encoding, loop in the code table; scanline %d",
356 tif->tif_row);
357 }
358
359 static int
LZWDecode(TIFF * tif,uint8 * op0,tmsize_t occ0,uint16 s)360 LZWDecode(TIFF* tif, uint8* op0, tmsize_t occ0, uint16 s)
361 {
362 static const char module[] = "LZWDecode";
363 LZWCodecState *sp = DecoderState(tif);
364 char *op = (char*) op0;
365 long occ = (long) occ0;
366 char *tp;
367 unsigned char *bp;
368 hcode_t code;
369 int len;
370 long nbits, nextbits, nextdata, nbitsmask;
371 code_t *codep, *free_entp, *maxcodep, *oldcodep;
372
373 (void) s;
374 assert(sp != NULL);
375 assert(sp->dec_codetab != NULL);
376
377 /*
378 Fail if value does not fit in long.
379 */
380 if ((tmsize_t) occ != occ0)
381 return (0);
382 /*
383 * Restart interrupted output operation.
384 */
385 if (sp->dec_restart) {
386 long residue;
387
388 codep = sp->dec_codep;
389 residue = codep->length - sp->dec_restart;
390 if (residue > occ) {
391 /*
392 * Residue from previous decode is sufficient
393 * to satisfy decode request. Skip to the
394 * start of the decoded string, place decoded
395 * values in the output buffer, and return.
396 */
397 sp->dec_restart += occ;
398 do {
399 codep = codep->next;
400 } while (--residue > occ && codep);
401 if (codep) {
402 tp = op + occ;
403 do {
404 *--tp = codep->value;
405 codep = codep->next;
406 } while (--occ && codep);
407 }
408 return (1);
409 }
410 /*
411 * Residue satisfies only part of the decode request.
412 */
413 op += residue, occ -= residue;
414 tp = op;
415 do {
416 int t;
417 --tp;
418 t = codep->value;
419 codep = codep->next;
420 *tp = t;
421 } while (--residue && codep);
422 sp->dec_restart = 0;
423 }
424
425 bp = (unsigned char *)tif->tif_rawcp;
426 nbits = sp->lzw_nbits;
427 nextdata = sp->lzw_nextdata;
428 nextbits = sp->lzw_nextbits;
429 nbitsmask = sp->dec_nbitsmask;
430 oldcodep = sp->dec_oldcodep;
431 free_entp = sp->dec_free_entp;
432 maxcodep = sp->dec_maxcodep;
433
434 while (occ > 0) {
435 NextCode(tif, sp, bp, code, GetNextCode);
436 if (code == CODE_EOI)
437 break;
438 if (code == CODE_CLEAR) {
439 free_entp = sp->dec_codetab + CODE_FIRST;
440 _TIFFmemset(free_entp, 0,
441 (CSIZE - CODE_FIRST) * sizeof (code_t));
442 nbits = BITS_MIN;
443 nbitsmask = MAXCODE(BITS_MIN);
444 maxcodep = sp->dec_codetab + nbitsmask-1;
445 NextCode(tif, sp, bp, code, GetNextCode);
446 if (code == CODE_EOI)
447 break;
448 if (code >= CODE_CLEAR) {
449 TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
450 "LZWDecode: Corrupted LZW table at scanline %d",
451 tif->tif_row);
452 return (0);
453 }
454 *op++ = (char)code, occ--;
455 oldcodep = sp->dec_codetab + code;
456 continue;
457 }
458 codep = sp->dec_codetab + code;
459
460 /*
461 * Add the new entry to the code table.
462 */
463 if (free_entp < &sp->dec_codetab[0] ||
464 free_entp >= &sp->dec_codetab[CSIZE]) {
465 TIFFErrorExt(tif->tif_clientdata, module,
466 "Corrupted LZW table at scanline %d",
467 tif->tif_row);
468 return (0);
469 }
470
471 free_entp->next = oldcodep;
472 if (free_entp->next < &sp->dec_codetab[0] ||
473 free_entp->next >= &sp->dec_codetab[CSIZE]) {
474 TIFFErrorExt(tif->tif_clientdata, module,
475 "Corrupted LZW table at scanline %d",
476 tif->tif_row);
477 return (0);
478 }
479 free_entp->firstchar = free_entp->next->firstchar;
480 free_entp->length = free_entp->next->length+1;
481 free_entp->value = (codep < free_entp) ?
482 codep->firstchar : free_entp->firstchar;
483 if (++free_entp > maxcodep) {
484 if (++nbits > BITS_MAX) /* should not happen */
485 nbits = BITS_MAX;
486 nbitsmask = MAXCODE(nbits);
487 maxcodep = sp->dec_codetab + nbitsmask-1;
488 }
489 oldcodep = codep;
490 if (code >= 256) {
491 /*
492 * Code maps to a string, copy string
493 * value to output (written in reverse).
494 */
495 if(codep->length == 0) {
496 TIFFErrorExt(tif->tif_clientdata, module,
497 "Wrong length of decoded string: "
498 "data probably corrupted at scanline %d",
499 tif->tif_row);
500 return (0);
501 }
502 if (codep->length > occ) {
503 /*
504 * String is too long for decode buffer,
505 * locate portion that will fit, copy to
506 * the decode buffer, and setup restart
507 * logic for the next decoding call.
508 */
509 sp->dec_codep = codep;
510 do {
511 codep = codep->next;
512 } while (codep && codep->length > occ);
513 if (codep) {
514 sp->dec_restart = (long)occ;
515 tp = op + occ;
516 do {
517 *--tp = codep->value;
518 codep = codep->next;
519 } while (--occ && codep);
520 if (codep)
521 codeLoop(tif, module);
522 }
523 break;
524 }
525 len = codep->length;
526 tp = op + len;
527 do {
528 int t;
529 --tp;
530 t = codep->value;
531 codep = codep->next;
532 *tp = t;
533 } while (codep && tp > op);
534 if (codep) {
535 codeLoop(tif, module);
536 break;
537 }
538 assert(occ >= len);
539 op += len, occ -= len;
540 } else
541 *op++ = (char)code, occ--;
542 }
543
544 tif->tif_rawcp = (uint8*) bp;
545 sp->lzw_nbits = (unsigned short) nbits;
546 sp->lzw_nextdata = nextdata;
547 sp->lzw_nextbits = nextbits;
548 sp->dec_nbitsmask = nbitsmask;
549 sp->dec_oldcodep = oldcodep;
550 sp->dec_free_entp = free_entp;
551 sp->dec_maxcodep = maxcodep;
552
553 if (occ > 0) {
554 #if defined(__WIN32__) && (defined(_MSC_VER) || defined(__MINGW32__))
555 TIFFErrorExt(tif->tif_clientdata, module,
556 "Not enough data at scanline %d (short %I64d bytes)",
557 tif->tif_row, (unsigned __int64) occ);
558 #else
559 TIFFErrorExt(tif->tif_clientdata, module,
560 "Not enough data at scanline %d (short %llu bytes)",
561 tif->tif_row, (unsigned long long) occ);
562 #endif
563 return (0);
564 }
565 return (1);
566 }
567
568 #ifdef LZW_COMPAT
569 /*
570 * Decode a "hunk of data" for old images.
571 */
572 #define GetNextCodeCompat(sp, bp, code) { \
573 nextdata |= (unsigned long) *(bp)++ << nextbits; \
574 nextbits += 8; \
575 if (nextbits < nbits) { \
576 nextdata |= (unsigned long) *(bp)++ << nextbits;\
577 nextbits += 8; \
578 } \
579 code = (hcode_t)(nextdata & nbitsmask); \
580 nextdata >>= nbits; \
581 nextbits -= nbits; \
582 }
583
584 static int
LZWDecodeCompat(TIFF * tif,uint8 * op0,tmsize_t occ0,uint16 s)585 LZWDecodeCompat(TIFF* tif, uint8* op0, tmsize_t occ0, uint16 s)
586 {
587 static const char module[] = "LZWDecodeCompat";
588 LZWCodecState *sp = DecoderState(tif);
589 char *op = (char*) op0;
590 long occ = (long) occ0;
591 char *tp;
592 unsigned char *bp;
593 int code, nbits;
594 long nextbits, nextdata, nbitsmask;
595 code_t *codep, *free_entp, *maxcodep, *oldcodep;
596
597 (void) s;
598 assert(sp != NULL);
599
600 /*
601 Fail if value does not fit in long.
602 */
603 if ((tmsize_t) occ != occ0)
604 return (0);
605
606 /*
607 * Restart interrupted output operation.
608 */
609 if (sp->dec_restart) {
610 long residue;
611
612 codep = sp->dec_codep;
613 residue = codep->length - sp->dec_restart;
614 if (residue > occ) {
615 /*
616 * Residue from previous decode is sufficient
617 * to satisfy decode request. Skip to the
618 * start of the decoded string, place decoded
619 * values in the output buffer, and return.
620 */
621 sp->dec_restart += occ;
622 do {
623 codep = codep->next;
624 } while (--residue > occ);
625 tp = op + occ;
626 do {
627 *--tp = codep->value;
628 codep = codep->next;
629 } while (--occ);
630 return (1);
631 }
632 /*
633 * Residue satisfies only part of the decode request.
634 */
635 op += residue, occ -= residue;
636 tp = op;
637 do {
638 *--tp = codep->value;
639 codep = codep->next;
640 } while (--residue);
641 sp->dec_restart = 0;
642 }
643
644 bp = (unsigned char *)tif->tif_rawcp;
645 nbits = sp->lzw_nbits;
646 nextdata = sp->lzw_nextdata;
647 nextbits = sp->lzw_nextbits;
648 nbitsmask = sp->dec_nbitsmask;
649 oldcodep = sp->dec_oldcodep;
650 free_entp = sp->dec_free_entp;
651 maxcodep = sp->dec_maxcodep;
652
653 while (occ > 0) {
654 NextCode(tif, sp, bp, code, GetNextCodeCompat);
655 if (code == CODE_EOI)
656 break;
657 if (code == CODE_CLEAR) {
658 free_entp = sp->dec_codetab + CODE_FIRST;
659 _TIFFmemset(free_entp, 0,
660 (CSIZE - CODE_FIRST) * sizeof (code_t));
661 nbits = BITS_MIN;
662 nbitsmask = MAXCODE(BITS_MIN);
663 maxcodep = sp->dec_codetab + nbitsmask;
664 NextCode(tif, sp, bp, code, GetNextCodeCompat);
665 if (code == CODE_EOI)
666 break;
667 if (code >= CODE_CLEAR) {
668 TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
669 "LZWDecode: Corrupted LZW table at scanline %d",
670 tif->tif_row);
671 return (0);
672 }
673 *op++ = code, occ--;
674 oldcodep = sp->dec_codetab + code;
675 continue;
676 }
677 codep = sp->dec_codetab + code;
678
679 /*
680 * Add the new entry to the code table.
681 */
682 if (free_entp < &sp->dec_codetab[0] ||
683 free_entp >= &sp->dec_codetab[CSIZE]) {
684 TIFFErrorExt(tif->tif_clientdata, module,
685 "Corrupted LZW table at scanline %d", tif->tif_row);
686 return (0);
687 }
688
689 free_entp->next = oldcodep;
690 if (free_entp->next < &sp->dec_codetab[0] ||
691 free_entp->next >= &sp->dec_codetab[CSIZE]) {
692 TIFFErrorExt(tif->tif_clientdata, module,
693 "Corrupted LZW table at scanline %d", tif->tif_row);
694 return (0);
695 }
696 free_entp->firstchar = free_entp->next->firstchar;
697 free_entp->length = free_entp->next->length+1;
698 free_entp->value = (codep < free_entp) ?
699 codep->firstchar : free_entp->firstchar;
700 if (++free_entp > maxcodep) {
701 if (++nbits > BITS_MAX) /* should not happen */
702 nbits = BITS_MAX;
703 nbitsmask = MAXCODE(nbits);
704 maxcodep = sp->dec_codetab + nbitsmask;
705 }
706 oldcodep = codep;
707 if (code >= 256) {
708 /*
709 * Code maps to a string, copy string
710 * value to output (written in reverse).
711 */
712 if(codep->length == 0) {
713 TIFFErrorExt(tif->tif_clientdata, module,
714 "Wrong length of decoded "
715 "string: data probably corrupted at scanline %d",
716 tif->tif_row);
717 return (0);
718 }
719 if (codep->length > occ) {
720 /*
721 * String is too long for decode buffer,
722 * locate portion that will fit, copy to
723 * the decode buffer, and setup restart
724 * logic for the next decoding call.
725 */
726 sp->dec_codep = codep;
727 do {
728 codep = codep->next;
729 } while (codep->length > occ);
730 sp->dec_restart = occ;
731 tp = op + occ;
732 do {
733 *--tp = codep->value;
734 codep = codep->next;
735 } while (--occ);
736 break;
737 }
738 assert(occ >= codep->length);
739 op += codep->length, occ -= codep->length;
740 tp = op;
741 do {
742 *--tp = codep->value;
743 } while( (codep = codep->next) != NULL );
744 } else
745 *op++ = code, occ--;
746 }
747
748 tif->tif_rawcp = (uint8*) bp;
749 sp->lzw_nbits = nbits;
750 sp->lzw_nextdata = nextdata;
751 sp->lzw_nextbits = nextbits;
752 sp->dec_nbitsmask = nbitsmask;
753 sp->dec_oldcodep = oldcodep;
754 sp->dec_free_entp = free_entp;
755 sp->dec_maxcodep = maxcodep;
756
757 if (occ > 0) {
758 #if defined(__WIN32__) && (defined(_MSC_VER) || defined(__MINGW32__))
759 TIFFErrorExt(tif->tif_clientdata, module,
760 "Not enough data at scanline %d (short %I64d bytes)",
761 tif->tif_row, (unsigned __int64) occ);
762 #else
763 TIFFErrorExt(tif->tif_clientdata, module,
764 "Not enough data at scanline %d (short %llu bytes)",
765 tif->tif_row, (unsigned long long) occ);
766 #endif
767 return (0);
768 }
769 return (1);
770 }
771 #endif /* LZW_COMPAT */
772
773 /*
774 * LZW Encoding.
775 */
776
777 static int
LZWSetupEncode(TIFF * tif)778 LZWSetupEncode(TIFF* tif)
779 {
780 static const char module[] = "LZWSetupEncode";
781 LZWCodecState* sp = EncoderState(tif);
782
783 assert(sp != NULL);
784 sp->enc_hashtab = (hash_t*) _TIFFmalloc(HSIZE*sizeof (hash_t));
785 if (sp->enc_hashtab == NULL) {
786 TIFFErrorExt(tif->tif_clientdata, module,
787 "No space for LZW hash table");
788 return (0);
789 }
790 return (1);
791 }
792
793 /*
794 * Reset encoding state at the start of a strip.
795 */
796 static int
LZWPreEncode(TIFF * tif,uint16 s)797 LZWPreEncode(TIFF* tif, uint16 s)
798 {
799 LZWCodecState *sp = EncoderState(tif);
800
801 (void) s;
802 assert(sp != NULL);
803
804 if( sp->enc_hashtab == NULL )
805 {
806 tif->tif_setupencode( tif );
807 }
808
809 sp->lzw_nbits = BITS_MIN;
810 sp->lzw_maxcode = MAXCODE(BITS_MIN);
811 sp->lzw_free_ent = CODE_FIRST;
812 sp->lzw_nextbits = 0;
813 sp->lzw_nextdata = 0;
814 sp->enc_checkpoint = CHECK_GAP;
815 sp->enc_ratio = 0;
816 sp->enc_incount = 0;
817 sp->enc_outcount = 0;
818 /*
819 * The 4 here insures there is space for 2 max-sized
820 * codes in LZWEncode and LZWPostDecode.
821 */
822 sp->enc_rawlimit = tif->tif_rawdata + tif->tif_rawdatasize-1 - 4;
823 cl_hash(sp); /* clear hash table */
824 sp->enc_oldcode = (hcode_t) -1; /* generates CODE_CLEAR in LZWEncode */
825 return (1);
826 }
827
828 #define CALCRATIO(sp, rat) { \
829 if (incount > 0x007fffff) { /* NB: shift will overflow */\
830 rat = outcount >> 8; \
831 rat = (rat == 0 ? 0x7fffffff : incount/rat); \
832 } else \
833 rat = (incount<<8) / outcount; \
834 }
835 #define PutNextCode(op, c) { \
836 nextdata = (nextdata << nbits) | c; \
837 nextbits += nbits; \
838 *op++ = (unsigned char)(nextdata >> (nextbits-8)); \
839 nextbits -= 8; \
840 if (nextbits >= 8) { \
841 *op++ = (unsigned char)(nextdata >> (nextbits-8)); \
842 nextbits -= 8; \
843 } \
844 outcount += nbits; \
845 }
846
847 /*
848 * Encode a chunk of pixels.
849 *
850 * Uses an open addressing double hashing (no chaining) on the
851 * prefix code/next character combination. We do a variant of
852 * Knuth's algorithm D (vol. 3, sec. 6.4) along with G. Knott's
853 * relatively-prime secondary probe. Here, the modular division
854 * first probe is gives way to a faster exclusive-or manipulation.
855 * Also do block compression with an adaptive reset, whereby the
856 * code table is cleared when the compression ratio decreases,
857 * but after the table fills. The variable-length output codes
858 * are re-sized at this point, and a CODE_CLEAR is generated
859 * for the decoder.
860 */
861 static int
LZWEncode(TIFF * tif,uint8 * bp,tmsize_t cc,uint16 s)862 LZWEncode(TIFF* tif, uint8* bp, tmsize_t cc, uint16 s)
863 {
864 register LZWCodecState *sp = EncoderState(tif);
865 register long fcode;
866 register hash_t *hp;
867 register int h, c;
868 hcode_t ent;
869 long disp;
870 long incount, outcount, checkpoint;
871 long nextdata, nextbits;
872 int free_ent, maxcode, nbits;
873 uint8* op;
874 uint8* limit;
875
876 (void) s;
877 if (sp == NULL)
878 return (0);
879
880 assert(sp->enc_hashtab != NULL);
881
882 /*
883 * Load local state.
884 */
885 incount = sp->enc_incount;
886 outcount = sp->enc_outcount;
887 checkpoint = sp->enc_checkpoint;
888 nextdata = sp->lzw_nextdata;
889 nextbits = sp->lzw_nextbits;
890 free_ent = sp->lzw_free_ent;
891 maxcode = sp->lzw_maxcode;
892 nbits = sp->lzw_nbits;
893 op = tif->tif_rawcp;
894 limit = sp->enc_rawlimit;
895 ent = sp->enc_oldcode;
896
897 if (ent == (hcode_t) -1 && cc > 0) {
898 /*
899 * NB: This is safe because it can only happen
900 * at the start of a strip where we know there
901 * is space in the data buffer.
902 */
903 PutNextCode(op, CODE_CLEAR);
904 ent = *bp++; cc--; incount++;
905 }
906 while (cc > 0) {
907 c = *bp++; cc--; incount++;
908 fcode = ((long)c << BITS_MAX) + ent;
909 h = (c << HSHIFT) ^ ent; /* xor hashing */
910 #ifdef _WINDOWS
911 /*
912 * Check hash index for an overflow.
913 */
914 if (h >= HSIZE)
915 h -= HSIZE;
916 #endif
917 hp = &sp->enc_hashtab[h];
918 if (hp->hash == fcode) {
919 ent = hp->code;
920 continue;
921 }
922 if (hp->hash >= 0) {
923 /*
924 * Primary hash failed, check secondary hash.
925 */
926 disp = HSIZE - h;
927 if (h == 0)
928 disp = 1;
929 do {
930 /*
931 * Avoid pointer arithmetic 'cuz of
932 * wraparound problems with segments.
933 */
934 if ((h -= disp) < 0)
935 h += HSIZE;
936 hp = &sp->enc_hashtab[h];
937 if (hp->hash == fcode) {
938 ent = hp->code;
939 goto hit;
940 }
941 } while (hp->hash >= 0);
942 }
943 /*
944 * New entry, emit code and add to table.
945 */
946 /*
947 * Verify there is space in the buffer for the code
948 * and any potential Clear code that might be emitted
949 * below. The value of limit is setup so that there
950 * are at least 4 bytes free--room for 2 codes.
951 */
952 if (op > limit) {
953 tif->tif_rawcc = (tmsize_t)(op - tif->tif_rawdata);
954 TIFFFlushData1(tif);
955 op = tif->tif_rawdata;
956 }
957 PutNextCode(op, ent);
958 ent = c;
959 hp->code = free_ent++;
960 hp->hash = fcode;
961 if (free_ent == CODE_MAX-1) {
962 /* table is full, emit clear code and reset */
963 cl_hash(sp);
964 sp->enc_ratio = 0;
965 incount = 0;
966 outcount = 0;
967 free_ent = CODE_FIRST;
968 PutNextCode(op, CODE_CLEAR);
969 nbits = BITS_MIN;
970 maxcode = MAXCODE(BITS_MIN);
971 } else {
972 /*
973 * If the next entry is going to be too big for
974 * the code size, then increase it, if possible.
975 */
976 if (free_ent > maxcode) {
977 nbits++;
978 assert(nbits <= BITS_MAX);
979 maxcode = (int) MAXCODE(nbits);
980 } else if (incount >= checkpoint) {
981 long rat;
982 /*
983 * Check compression ratio and, if things seem
984 * to be slipping, clear the hash table and
985 * reset state. The compression ratio is a
986 * 24+8-bit fractional number.
987 */
988 checkpoint = incount+CHECK_GAP;
989 CALCRATIO(sp, rat);
990 if (rat <= sp->enc_ratio) {
991 cl_hash(sp);
992 sp->enc_ratio = 0;
993 incount = 0;
994 outcount = 0;
995 free_ent = CODE_FIRST;
996 PutNextCode(op, CODE_CLEAR);
997 nbits = BITS_MIN;
998 maxcode = MAXCODE(BITS_MIN);
999 } else
1000 sp->enc_ratio = rat;
1001 }
1002 }
1003 hit:
1004 ;
1005 }
1006
1007 /*
1008 * Restore global state.
1009 */
1010 sp->enc_incount = incount;
1011 sp->enc_outcount = outcount;
1012 sp->enc_checkpoint = checkpoint;
1013 sp->enc_oldcode = ent;
1014 sp->lzw_nextdata = nextdata;
1015 sp->lzw_nextbits = nextbits;
1016 sp->lzw_free_ent = free_ent;
1017 sp->lzw_maxcode = maxcode;
1018 sp->lzw_nbits = nbits;
1019 tif->tif_rawcp = op;
1020 return (1);
1021 }
1022
1023 /*
1024 * Finish off an encoded strip by flushing the last
1025 * string and tacking on an End Of Information code.
1026 */
1027 static int
LZWPostEncode(TIFF * tif)1028 LZWPostEncode(TIFF* tif)
1029 {
1030 register LZWCodecState *sp = EncoderState(tif);
1031 uint8* op = tif->tif_rawcp;
1032 long nextbits = sp->lzw_nextbits;
1033 long nextdata = sp->lzw_nextdata;
1034 long outcount = sp->enc_outcount;
1035 int nbits = sp->lzw_nbits;
1036
1037 if (op > sp->enc_rawlimit) {
1038 tif->tif_rawcc = (tmsize_t)(op - tif->tif_rawdata);
1039 TIFFFlushData1(tif);
1040 op = tif->tif_rawdata;
1041 }
1042 if (sp->enc_oldcode != (hcode_t) -1) {
1043 PutNextCode(op, sp->enc_oldcode);
1044 sp->enc_oldcode = (hcode_t) -1;
1045 }
1046 PutNextCode(op, CODE_EOI);
1047 if (nextbits > 0)
1048 *op++ = (unsigned char)(nextdata << (8-nextbits));
1049 tif->tif_rawcc = (tmsize_t)(op - tif->tif_rawdata);
1050 return (1);
1051 }
1052
1053 /*
1054 * Reset encoding hash table.
1055 */
1056 static void
cl_hash(LZWCodecState * sp)1057 cl_hash(LZWCodecState* sp)
1058 {
1059 register hash_t *hp = &sp->enc_hashtab[HSIZE-1];
1060 register long i = HSIZE-8;
1061
1062 do {
1063 i -= 8;
1064 hp[-7].hash = -1;
1065 hp[-6].hash = -1;
1066 hp[-5].hash = -1;
1067 hp[-4].hash = -1;
1068 hp[-3].hash = -1;
1069 hp[-2].hash = -1;
1070 hp[-1].hash = -1;
1071 hp[ 0].hash = -1;
1072 hp -= 8;
1073 } while (i >= 0);
1074 for (i += 8; i > 0; i--, hp--)
1075 hp->hash = -1;
1076 }
1077
1078 static void
LZWCleanup(TIFF * tif)1079 LZWCleanup(TIFF* tif)
1080 {
1081 (void)TIFFPredictorCleanup(tif);
1082
1083 assert(tif->tif_data != 0);
1084
1085 if (DecoderState(tif)->dec_codetab)
1086 _TIFFfree(DecoderState(tif)->dec_codetab);
1087
1088 if (EncoderState(tif)->enc_hashtab)
1089 _TIFFfree(EncoderState(tif)->enc_hashtab);
1090
1091 _TIFFfree(tif->tif_data);
1092 tif->tif_data = NULL;
1093
1094 _TIFFSetDefaultCompressionState(tif);
1095 }
1096
1097 int
TIFFInitLZW(TIFF * tif,int scheme)1098 TIFFInitLZW(TIFF* tif, int scheme)
1099 {
1100 static const char module[] = "TIFFInitLZW";
1101 assert(scheme == COMPRESSION_LZW);
1102 /*
1103 * Allocate state block so tag methods have storage to record values.
1104 */
1105 tif->tif_data = (uint8*) _TIFFmalloc(sizeof (LZWCodecState));
1106 if (tif->tif_data == NULL)
1107 goto bad;
1108 DecoderState(tif)->dec_codetab = NULL;
1109 DecoderState(tif)->dec_decode = NULL;
1110 EncoderState(tif)->enc_hashtab = NULL;
1111 LZWState(tif)->rw_mode = tif->tif_mode;
1112
1113 /*
1114 * Install codec methods.
1115 */
1116 tif->tif_fixuptags = LZWFixupTags;
1117 tif->tif_setupdecode = LZWSetupDecode;
1118 tif->tif_predecode = LZWPreDecode;
1119 tif->tif_decoderow = LZWDecode;
1120 tif->tif_decodestrip = LZWDecode;
1121 tif->tif_decodetile = LZWDecode;
1122 tif->tif_setupencode = LZWSetupEncode;
1123 tif->tif_preencode = LZWPreEncode;
1124 tif->tif_postencode = LZWPostEncode;
1125 tif->tif_encoderow = LZWEncode;
1126 tif->tif_encodestrip = LZWEncode;
1127 tif->tif_encodetile = LZWEncode;
1128 tif->tif_cleanup = LZWCleanup;
1129 /*
1130 * Setup predictor setup.
1131 */
1132 (void) TIFFPredictorInit(tif);
1133 return (1);
1134 bad:
1135 TIFFErrorExt(tif->tif_clientdata, module,
1136 "No space for LZW state block");
1137 return (0);
1138 }
1139
1140 /*
1141 * Copyright (c) 1985, 1986 The Regents of the University of California.
1142 * All rights reserved.
1143 *
1144 * This code is derived from software contributed to Berkeley by
1145 * James A. Woods, derived from original work by Spencer Thomas
1146 * and Joseph Orost.
1147 *
1148 * Redistribution and use in source and binary forms are permitted
1149 * provided that the above copyright notice and this paragraph are
1150 * duplicated in all such forms and that any documentation,
1151 * advertising materials, and other materials related to such
1152 * distribution and use acknowledge that the software was developed
1153 * by the University of California, Berkeley. The name of the
1154 * University may not be used to endorse or promote products derived
1155 * from this software without specific prior written permission.
1156 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
1157 * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
1158 * WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE.
1159 */
1160 #endif /* LZW_SUPPORT */
1161
1162 /* vim: set ts=8 sts=8 sw=8 noet: */
1163 /*
1164 * Local Variables:
1165 * mode: c
1166 * c-basic-offset: 8
1167 * fill-column: 78
1168 * End:
1169 */
1170