xref: /freebsd/usr.bin/compress/zopen.c (revision 02e9120893770924227138ba49df1edb3896112a)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1985, 1986, 1992, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Diomidis Spinellis and James A. Woods, derived from original
9  * work by Spencer Thomas and Joseph Orost.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  */
35 
36 
37 #include <sys/cdefs.h>
38 /*-
39  * fcompress.c - File compression ala IEEE Computer, June 1984.
40  *
41  * Compress authors:
42  *		Spencer W. Thomas	(decvax!utah-cs!thomas)
43  *		Jim McKie		(decvax!mcvax!jim)
44  *		Steve Davies		(decvax!vax135!petsd!peora!srd)
45  *		Ken Turkowski		(decvax!decwrl!turtlevax!ken)
46  *		James A. Woods		(decvax!ihnp4!ames!jaw)
47  *		Joe Orost		(decvax!vax135!petsd!joe)
48  *
49  * Cleaned up and converted to library returning I/O streams by
50  * Diomidis Spinellis <dds@doc.ic.ac.uk>.
51  *
52  * zopen(filename, mode, bits)
53  *	Returns a FILE * that can be used for read or write.  The modes
54  *	supported are only "r" and "w".  Seeking is not allowed.  On
55  *	reading the file is decompressed, on writing it is compressed.
56  *	The output is compatible with compress(1) with 16 bit tables.
57  *	Any file produced by compress(1) can be read.
58  */
59 
60 #include <sys/param.h>
61 #include <sys/stat.h>
62 
63 #include <ctype.h>
64 #include <errno.h>
65 #include <signal.h>
66 #include <stdio.h>
67 #include <stdlib.h>
68 #include <string.h>
69 #include <unistd.h>
70 #include "zopen.h"
71 
72 #define	BITS		16		/* Default bits. */
73 #define	HSIZE		69001		/* 95% occupancy */
74 
75 /* A code_int must be able to hold 2**BITS values of type int, and also -1. */
76 typedef long code_int;
77 typedef long count_int;
78 
79 typedef u_char char_type;
80 static char_type magic_header[] =
81 	{'\037', '\235'};		/* 1F 9D */
82 
83 #define	BIT_MASK	0x1f		/* Defines for third byte of header. */
84 #define	BLOCK_MASK	0x80
85 
86 /*
87  * Masks 0x40 and 0x20 are free.  I think 0x20 should mean that there is
88  * a fourth header byte (for expansion).
89  */
90 #define	INIT_BITS 9			/* Initial number of bits/code. */
91 
92 #define	MAXCODE(n_bits)	((1 << (n_bits)) - 1)
93 
94 struct s_zstate {
95 	FILE *zs_fp;			/* File stream for I/O */
96 	char zs_mode;			/* r or w */
97 	enum {
98 		S_START, S_MIDDLE, S_EOF
99 	} zs_state;			/* State of computation */
100 	u_int zs_n_bits;		/* Number of bits/code. */
101 	u_int zs_maxbits;		/* User settable max # bits/code. */
102 	code_int zs_maxcode;		/* Maximum code, given n_bits. */
103 	code_int zs_maxmaxcode;		/* Should NEVER generate this code. */
104 	count_int zs_htab [HSIZE];
105 	u_short zs_codetab [HSIZE];
106 	code_int zs_hsize;		/* For dynamic table sizing. */
107 	code_int zs_free_ent;		/* First unused entry. */
108 	/*
109 	 * Block compression parameters -- after all codes are used up,
110 	 * and compression rate changes, start over.
111 	 */
112 	int zs_block_compress;
113 	int zs_clear_flg;
114 	long zs_ratio;
115 	count_int zs_checkpoint;
116 	u_int zs_offset;
117 	long zs_in_count;		/* Length of input. */
118 	long zs_bytes_out;		/* Length of compressed output. */
119 	long zs_out_count;		/* # of codes output (for debugging). */
120 	char_type zs_buf[BITS];
121 	union {
122 		struct {
123 			long zs_fcode;
124 			code_int zs_ent;
125 			code_int zs_hsize_reg;
126 			int zs_hshift;
127 		} w;			/* Write parameters */
128 		struct {
129 			char_type *zs_stackp;
130 			int zs_finchar;
131 			code_int zs_code, zs_oldcode, zs_incode;
132 			int zs_roffset, zs_size;
133 			char_type zs_gbuf[BITS];
134 		} r;			/* Read parameters */
135 	} u;
136 };
137 
138 /* Definitions to retain old variable names */
139 #define	fp		zs->zs_fp
140 #define	zmode		zs->zs_mode
141 #define	state		zs->zs_state
142 #define	n_bits		zs->zs_n_bits
143 #define	maxbits		zs->zs_maxbits
144 #define	maxcode		zs->zs_maxcode
145 #define	maxmaxcode	zs->zs_maxmaxcode
146 #define	htab		zs->zs_htab
147 #define	codetab		zs->zs_codetab
148 #define	hsize		zs->zs_hsize
149 #define	free_ent	zs->zs_free_ent
150 #define	block_compress	zs->zs_block_compress
151 #define	clear_flg	zs->zs_clear_flg
152 #define	ratio		zs->zs_ratio
153 #define	checkpoint	zs->zs_checkpoint
154 #define	offset		zs->zs_offset
155 #define	in_count	zs->zs_in_count
156 #define	bytes_out	zs->zs_bytes_out
157 #define	out_count	zs->zs_out_count
158 #define	buf		zs->zs_buf
159 #define	fcode		zs->u.w.zs_fcode
160 #define	hsize_reg	zs->u.w.zs_hsize_reg
161 #define	ent		zs->u.w.zs_ent
162 #define	hshift		zs->u.w.zs_hshift
163 #define	stackp		zs->u.r.zs_stackp
164 #define	finchar		zs->u.r.zs_finchar
165 #define	code		zs->u.r.zs_code
166 #define	oldcode		zs->u.r.zs_oldcode
167 #define	incode		zs->u.r.zs_incode
168 #define	roffset		zs->u.r.zs_roffset
169 #define	size		zs->u.r.zs_size
170 #define	gbuf		zs->u.r.zs_gbuf
171 
172 /*
173  * To save much memory, we overlay the table used by compress() with those
174  * used by decompress().  The tab_prefix table is the same size and type as
175  * the codetab.  The tab_suffix table needs 2**BITS characters.  We get this
176  * from the beginning of htab.  The output stack uses the rest of htab, and
177  * contains characters.  There is plenty of room for any possible stack
178  * (stack used to be 8000 characters).
179  */
180 
181 #define	htabof(i)	htab[i]
182 #define	codetabof(i)	codetab[i]
183 
184 #define	tab_prefixof(i)	codetabof(i)
185 #define	tab_suffixof(i)	((char_type *)(htab))[i]
186 #define	de_stack	((char_type *)&tab_suffixof(1 << BITS))
187 
188 #define	CHECK_GAP 10000		/* Ratio check interval. */
189 
190 /*
191  * the next two codes should not be changed lightly, as they must not
192  * lie within the contiguous general code space.
193  */
194 #define	FIRST	257		/* First free entry. */
195 #define	CLEAR	256		/* Table clear output code. */
196 
197 static int	cl_block(struct s_zstate *);
198 static void	cl_hash(struct s_zstate *, count_int);
199 static code_int	getcode(struct s_zstate *);
200 static int	output(struct s_zstate *, code_int);
201 static int	zclose(void *);
202 static int	zread(void *, char *, int);
203 static int	zwrite(void *, const char *, int);
204 
205 /*-
206  * Algorithm from "A Technique for High Performance Data Compression",
207  * Terry A. Welch, IEEE Computer Vol 17, No 6 (June 1984), pp 8-19.
208  *
209  * Algorithm:
210  * 	Modified Lempel-Ziv method (LZW).  Basically finds common
211  * substrings and replaces them with a variable size code.  This is
212  * deterministic, and can be done on the fly.  Thus, the decompression
213  * procedure needs no input table, but tracks the way the table was built.
214  */
215 
216 /*-
217  * compress write
218  *
219  * Algorithm:  use open addressing double hashing (no chaining) on the
220  * prefix code / next character combination.  We do a variant of Knuth's
221  * algorithm D (vol. 3, sec. 6.4) along with G. Knott's relatively-prime
222  * secondary probe.  Here, the modular division first probe is gives way
223  * to a faster exclusive-or manipulation.  Also do block compression with
224  * an adaptive reset, whereby the code table is cleared when the compression
225  * ratio decreases, but after the table fills.  The variable-length output
226  * codes are re-sized at this point, and a special CLEAR code is generated
227  * for the decompressor.  Late addition:  construct the table according to
228  * file size for noticeable speed improvement on small files.  Please direct
229  * questions about this implementation to ames!jaw.
230  */
231 static int
232 zwrite(void *cookie, const char *wbp, int num)
233 {
234 	code_int i;
235 	int c, disp;
236 	struct s_zstate *zs;
237 	const u_char *bp;
238 	u_char tmp;
239 	int count;
240 
241 	if (num == 0)
242 		return (0);
243 
244 	zs = cookie;
245 	count = num;
246 	bp = (const u_char *)wbp;
247 	if (state == S_MIDDLE)
248 		goto middle;
249 	state = S_MIDDLE;
250 
251 	maxmaxcode = 1L << maxbits;
252 	if (fwrite(magic_header,
253 	    sizeof(char), sizeof(magic_header), fp) != sizeof(magic_header))
254 		return (-1);
255 	tmp = (u_char)((maxbits) | block_compress);
256 	if (fwrite(&tmp, sizeof(char), sizeof(tmp), fp) != sizeof(tmp))
257 		return (-1);
258 
259 	offset = 0;
260 	bytes_out = 3;		/* Includes 3-byte header mojo. */
261 	out_count = 0;
262 	clear_flg = 0;
263 	ratio = 0;
264 	in_count = 1;
265 	checkpoint = CHECK_GAP;
266 	maxcode = MAXCODE(n_bits = INIT_BITS);
267 	free_ent = ((block_compress) ? FIRST : 256);
268 
269 	ent = *bp++;
270 	--count;
271 
272 	hshift = 0;
273 	for (fcode = (long)hsize; fcode < 65536L; fcode *= 2L)
274 		hshift++;
275 	hshift = 8 - hshift;	/* Set hash code range bound. */
276 
277 	hsize_reg = hsize;
278 	cl_hash(zs, (count_int)hsize_reg);	/* Clear hash table. */
279 
280 middle:	for (i = 0; count--;) {
281 		c = *bp++;
282 		in_count++;
283 		fcode = (long)(((long)c << maxbits) + ent);
284 		i = ((c << hshift) ^ ent);	/* Xor hashing. */
285 
286 		if (htabof(i) == fcode) {
287 			ent = codetabof(i);
288 			continue;
289 		} else if ((long)htabof(i) < 0)	/* Empty slot. */
290 			goto nomatch;
291 		disp = hsize_reg - i;	/* Secondary hash (after G. Knott). */
292 		if (i == 0)
293 			disp = 1;
294 probe:		if ((i -= disp) < 0)
295 			i += hsize_reg;
296 
297 		if (htabof(i) == fcode) {
298 			ent = codetabof(i);
299 			continue;
300 		}
301 		if ((long)htabof(i) >= 0)
302 			goto probe;
303 nomatch:	if (output(zs, (code_int) ent) == -1)
304 			return (-1);
305 		out_count++;
306 		ent = c;
307 		if (free_ent < maxmaxcode) {
308 			codetabof(i) = free_ent++;	/* code -> hashtable */
309 			htabof(i) = fcode;
310 		} else if ((count_int)in_count >=
311 		    checkpoint && block_compress) {
312 			if (cl_block(zs) == -1)
313 				return (-1);
314 		}
315 	}
316 	return (num);
317 }
318 
319 static int
320 zclose(void *cookie)
321 {
322 	struct s_zstate *zs;
323 	int rval;
324 
325 	zs = cookie;
326 	if (zmode == 'w') {		/* Put out the final code. */
327 		if (output(zs, (code_int) ent) == -1) {
328 			(void)fclose(fp);
329 			free(zs);
330 			return (-1);
331 		}
332 		out_count++;
333 		if (output(zs, (code_int) - 1) == -1) {
334 			(void)fclose(fp);
335 			free(zs);
336 			return (-1);
337 		}
338 	}
339 	rval = fclose(fp) == EOF ? -1 : 0;
340 	free(zs);
341 	return (rval);
342 }
343 
344 /*-
345  * Output the given code.
346  * Inputs:
347  * 	code:	A n_bits-bit integer.  If == -1, then EOF.  This assumes
348  *		that n_bits =< (long)wordsize - 1.
349  * Outputs:
350  * 	Outputs code to the file.
351  * Assumptions:
352  *	Chars are 8 bits long.
353  * Algorithm:
354  * 	Maintain a BITS character long buffer (so that 8 codes will
355  * fit in it exactly).  Use the VAX insv instruction to insert each
356  * code in turn.  When the buffer fills up empty it and start over.
357  */
358 
359 static char_type lmask[9] =
360 	{0xff, 0xfe, 0xfc, 0xf8, 0xf0, 0xe0, 0xc0, 0x80, 0x00};
361 static char_type rmask[9] =
362 	{0x00, 0x01, 0x03, 0x07, 0x0f, 0x1f, 0x3f, 0x7f, 0xff};
363 
364 static int
365 output(struct s_zstate *zs, code_int ocode)
366 {
367 	int r_off;
368 	u_int bits;
369 	char_type *bp;
370 
371 	r_off = offset;
372 	bits = n_bits;
373 	bp = buf;
374 	if (ocode >= 0) {
375 		/* Get to the first byte. */
376 		bp += (r_off >> 3);
377 		r_off &= 7;
378 		/*
379 		 * Since ocode is always >= 8 bits, only need to mask the first
380 		 * hunk on the left.
381 		 */
382 		*bp = (*bp & rmask[r_off]) | ((ocode << r_off) & lmask[r_off]);
383 		bp++;
384 		bits -= (8 - r_off);
385 		ocode >>= 8 - r_off;
386 		/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
387 		if (bits >= 8) {
388 			*bp++ = ocode;
389 			ocode >>= 8;
390 			bits -= 8;
391 		}
392 		/* Last bits. */
393 		if (bits)
394 			*bp = ocode;
395 		offset += n_bits;
396 		if (offset == (n_bits << 3)) {
397 			bp = buf;
398 			bits = n_bits;
399 			bytes_out += bits;
400 			if (fwrite(bp, sizeof(char), bits, fp) != bits)
401 				return (-1);
402 			bp += bits;
403 			bits = 0;
404 			offset = 0;
405 		}
406 		/*
407 		 * If the next entry is going to be too big for the ocode size,
408 		 * then increase it, if possible.
409 		 */
410 		if (free_ent > maxcode || (clear_flg > 0)) {
411 		       /*
412 			* Write the whole buffer, because the input side won't
413 			* discover the size increase until after it has read it.
414 			*/
415 			if (offset > 0) {
416 				if (fwrite(buf, 1, n_bits, fp) != n_bits)
417 					return (-1);
418 				bytes_out += n_bits;
419 			}
420 			offset = 0;
421 
422 			if (clear_flg) {
423 				maxcode = MAXCODE(n_bits = INIT_BITS);
424 				clear_flg = 0;
425 			} else {
426 				n_bits++;
427 				if (n_bits == maxbits)
428 					maxcode = maxmaxcode;
429 				else
430 					maxcode = MAXCODE(n_bits);
431 			}
432 		}
433 	} else {
434 		/* At EOF, write the rest of the buffer. */
435 		if (offset > 0) {
436 			offset = (offset + 7) / 8;
437 			if (fwrite(buf, 1, offset, fp) != offset)
438 				return (-1);
439 			bytes_out += offset;
440 		}
441 		offset = 0;
442 	}
443 	return (0);
444 }
445 
446 /*
447  * Decompress read.  This routine adapts to the codes in the file building
448  * the "string" table on-the-fly; requiring no table to be stored in the
449  * compressed file.  The tables used herein are shared with those of the
450  * compress() routine.  See the definitions above.
451  */
452 static int
453 zread(void *cookie, char *rbp, int num)
454 {
455 	u_int count;
456 	struct s_zstate *zs;
457 	u_char *bp, header[3];
458 
459 	if (num == 0)
460 		return (0);
461 
462 	zs = cookie;
463 	count = num;
464 	bp = (u_char *)rbp;
465 	switch (state) {
466 	case S_START:
467 		state = S_MIDDLE;
468 		break;
469 	case S_MIDDLE:
470 		goto middle;
471 	case S_EOF:
472 		goto eof;
473 	}
474 
475 	/* Check the magic number */
476 	if (fread(header,
477 	    sizeof(char), sizeof(header), fp) != sizeof(header) ||
478 	    memcmp(header, magic_header, sizeof(magic_header)) != 0) {
479 		errno = EFTYPE;
480 		return (-1);
481 	}
482 	maxbits = header[2];	/* Set -b from file. */
483 	block_compress = maxbits & BLOCK_MASK;
484 	maxbits &= BIT_MASK;
485 	maxmaxcode = 1L << maxbits;
486 	if (maxbits > BITS || maxbits < 12) {
487 		errno = EFTYPE;
488 		return (-1);
489 	}
490 	/* As above, initialize the first 256 entries in the table. */
491 	maxcode = MAXCODE(n_bits = INIT_BITS);
492 	for (code = 255; code >= 0; code--) {
493 		tab_prefixof(code) = 0;
494 		tab_suffixof(code) = (char_type) code;
495 	}
496 	free_ent = block_compress ? FIRST : 256;
497 
498 	finchar = oldcode = getcode(zs);
499 	if (oldcode == -1)	/* EOF already? */
500 		return (0);	/* Get out of here */
501 
502 	/* First code must be 8 bits = char. */
503 	*bp++ = (u_char)finchar;
504 	count--;
505 	stackp = de_stack;
506 
507 	while ((code = getcode(zs)) > -1) {
508 
509 		if ((code == CLEAR) && block_compress) {
510 			for (code = 255; code >= 0; code--)
511 				tab_prefixof(code) = 0;
512 			clear_flg = 1;
513 			free_ent = FIRST;
514 			oldcode = -1;
515 			continue;
516 		}
517 		incode = code;
518 
519 		/* Special case for kWkWk string. */
520 		if (code >= free_ent) {
521 			if (code > free_ent || oldcode == -1) {
522 				/* Bad stream. */
523 				errno = EINVAL;
524 				return (-1);
525 			}
526 			*stackp++ = finchar;
527 			code = oldcode;
528 		}
529 		/*
530 		 * The above condition ensures that code < free_ent.
531 		 * The construction of tab_prefixof in turn guarantees that
532 		 * each iteration decreases code and therefore stack usage is
533 		 * bound by 1 << BITS - 256.
534 		 */
535 
536 		/* Generate output characters in reverse order. */
537 		while (code >= 256) {
538 			*stackp++ = tab_suffixof(code);
539 			code = tab_prefixof(code);
540 		}
541 		*stackp++ = finchar = tab_suffixof(code);
542 
543 		/* And put them out in forward order.  */
544 middle:		do {
545 			if (count-- == 0)
546 				return (num);
547 			*bp++ = *--stackp;
548 		} while (stackp > de_stack);
549 
550 		/* Generate the new entry. */
551 		if ((code = free_ent) < maxmaxcode && oldcode != -1) {
552 			tab_prefixof(code) = (u_short) oldcode;
553 			tab_suffixof(code) = finchar;
554 			free_ent = code + 1;
555 		}
556 
557 		/* Remember previous code. */
558 		oldcode = incode;
559 	}
560 	state = S_EOF;
561 eof:	return (num - count);
562 }
563 
564 /*-
565  * Read one code from the standard input.  If EOF, return -1.
566  * Inputs:
567  * 	stdin
568  * Outputs:
569  * 	code or -1 is returned.
570  */
571 static code_int
572 getcode(struct s_zstate *zs)
573 {
574 	code_int gcode;
575 	int r_off, bits;
576 	char_type *bp;
577 
578 	bp = gbuf;
579 	if (clear_flg > 0 || roffset >= size || free_ent > maxcode) {
580 		/*
581 		 * If the next entry will be too big for the current gcode
582 		 * size, then we must increase the size.  This implies reading
583 		 * a new buffer full, too.
584 		 */
585 		if (free_ent > maxcode) {
586 			n_bits++;
587 			if (n_bits == maxbits)	/* Won't get any bigger now. */
588 				maxcode = maxmaxcode;
589 			else
590 				maxcode = MAXCODE(n_bits);
591 		}
592 		if (clear_flg > 0) {
593 			maxcode = MAXCODE(n_bits = INIT_BITS);
594 			clear_flg = 0;
595 		}
596 		size = fread(gbuf, 1, n_bits, fp);
597 		if (size <= 0)			/* End of file. */
598 			return (-1);
599 		roffset = 0;
600 		/* Round size down to integral number of codes. */
601 		size = (size << 3) - (n_bits - 1);
602 	}
603 	r_off = roffset;
604 	bits = n_bits;
605 
606 	/* Get to the first byte. */
607 	bp += (r_off >> 3);
608 	r_off &= 7;
609 
610 	/* Get first part (low order bits). */
611 	gcode = (*bp++ >> r_off);
612 	bits -= (8 - r_off);
613 	r_off = 8 - r_off;	/* Now, roffset into gcode word. */
614 
615 	/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
616 	if (bits >= 8) {
617 		gcode |= *bp++ << r_off;
618 		r_off += 8;
619 		bits -= 8;
620 	}
621 
622 	/* High order bits. */
623 	gcode |= (*bp & rmask[bits]) << r_off;
624 	roffset += n_bits;
625 
626 	return (gcode);
627 }
628 
629 static int
630 cl_block(struct s_zstate *zs)		/* Table clear for block compress. */
631 {
632 	long rat;
633 
634 	checkpoint = in_count + CHECK_GAP;
635 
636 	if (in_count > 0x007fffff) {	/* Shift will overflow. */
637 		rat = bytes_out >> 8;
638 		if (rat == 0)		/* Don't divide by zero. */
639 			rat = 0x7fffffff;
640 		else
641 			rat = in_count / rat;
642 	} else
643 		rat = (in_count << 8) / bytes_out;	/* 8 fractional bits. */
644 	if (rat > ratio)
645 		ratio = rat;
646 	else {
647 		ratio = 0;
648 		cl_hash(zs, (count_int) hsize);
649 		free_ent = FIRST;
650 		clear_flg = 1;
651 		if (output(zs, (code_int) CLEAR) == -1)
652 			return (-1);
653 	}
654 	return (0);
655 }
656 
657 static void
658 cl_hash(struct s_zstate *zs, count_int cl_hsize)	/* Reset code table. */
659 {
660 	count_int *htab_p;
661 	long i, m1;
662 
663 	m1 = -1;
664 	htab_p = htab + cl_hsize;
665 	i = cl_hsize - 16;
666 	do {			/* Might use Sys V memset(3) here. */
667 		*(htab_p - 16) = m1;
668 		*(htab_p - 15) = m1;
669 		*(htab_p - 14) = m1;
670 		*(htab_p - 13) = m1;
671 		*(htab_p - 12) = m1;
672 		*(htab_p - 11) = m1;
673 		*(htab_p - 10) = m1;
674 		*(htab_p - 9) = m1;
675 		*(htab_p - 8) = m1;
676 		*(htab_p - 7) = m1;
677 		*(htab_p - 6) = m1;
678 		*(htab_p - 5) = m1;
679 		*(htab_p - 4) = m1;
680 		*(htab_p - 3) = m1;
681 		*(htab_p - 2) = m1;
682 		*(htab_p - 1) = m1;
683 		htab_p -= 16;
684 	} while ((i -= 16) >= 0);
685 	for (i += 16; i > 0; i--)
686 		*--htab_p = m1;
687 }
688 
689 FILE *
690 zopen(const char *fname, const char *mode, int bits)
691 {
692 	struct s_zstate *zs;
693 
694 	if ((mode[0] != 'r' && mode[0] != 'w') || mode[1] != '\0' ||
695 	    bits < 0 || bits > BITS) {
696 		errno = EINVAL;
697 		return (NULL);
698 	}
699 
700 	if ((zs = calloc(1, sizeof(struct s_zstate))) == NULL)
701 		return (NULL);
702 
703 	maxbits = bits ? bits : BITS;	/* User settable max # bits/code. */
704 	maxmaxcode = 1L << maxbits;	/* Should NEVER generate this code. */
705 	hsize = HSIZE;			/* For dynamic table sizing. */
706 	free_ent = 0;			/* First unused entry. */
707 	block_compress = BLOCK_MASK;
708 	clear_flg = 0;
709 	ratio = 0;
710 	checkpoint = CHECK_GAP;
711 	in_count = 1;			/* Length of input. */
712 	out_count = 0;			/* # of codes output (for debugging). */
713 	state = S_START;
714 	roffset = 0;
715 	size = 0;
716 
717 	/*
718 	 * Layering compress on top of stdio in order to provide buffering,
719 	 * and ensure that reads and write work with the data specified.
720 	 */
721 	if ((fp = fopen(fname, mode)) == NULL) {
722 		free(zs);
723 		return (NULL);
724 	}
725 	switch (*mode) {
726 	case 'r':
727 		zmode = 'r';
728 		return (funopen(zs, zread, NULL, NULL, zclose));
729 	case 'w':
730 		zmode = 'w';
731 		return (funopen(zs, NULL, zwrite, NULL, zclose));
732 	}
733 	/* NOTREACHED */
734 	return (NULL);
735 }
736