xref: /freebsd/usr.bin/compress/zopen.c (revision c17d43407fe04133a94055b0dbc7ea8965654a9f)
1 /*-
2  * Copyright (c) 1985, 1986, 1992, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Diomidis Spinellis and James A. Woods, derived from original
7  * work by Spencer Thomas and Joseph Orost.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. All advertising materials mentioning features or use of this software
18  *    must display the following acknowledgement:
19  *	This product includes software developed by the University of
20  *	California, Berkeley and its contributors.
21  * 4. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  */
37 
38 #if defined(LIBC_SCCS) && !defined(lint)
39 static char sccsid[] = "@(#)zopen.c	8.1 (Berkeley) 6/27/93";
40 #endif /* LIBC_SCCS and not lint */
41 
42 #include <sys/cdefs.h>
43 __FBSDID("$FreeBSD$");
44 
45 /*-
46  * fcompress.c - File compression ala IEEE Computer, June 1984.
47  *
48  * Compress authors:
49  *		Spencer W. Thomas	(decvax!utah-cs!thomas)
50  *		Jim McKie		(decvax!mcvax!jim)
51  *		Steve Davies		(decvax!vax135!petsd!peora!srd)
52  *		Ken Turkowski		(decvax!decwrl!turtlevax!ken)
53  *		James A. Woods		(decvax!ihnp4!ames!jaw)
54  *		Joe Orost		(decvax!vax135!petsd!joe)
55  *
56  * Cleaned up and converted to library returning I/O streams by
57  * Diomidis Spinellis <dds@doc.ic.ac.uk>.
58  *
59  * zopen(filename, mode, bits)
60  *	Returns a FILE * that can be used for read or write.  The modes
61  *	supported are only "r" and "w".  Seeking is not allowed.  On
62  *	reading the file is decompressed, on writing it is compressed.
63  *	The output is compatible with compress(1) with 16 bit tables.
64  *	Any file produced by compress(1) can be read.
65  */
66 
67 #include <sys/param.h>
68 #include <sys/stat.h>
69 
70 #include <ctype.h>
71 #include <errno.h>
72 #include <signal.h>
73 #include <stdio.h>
74 #include <stdlib.h>
75 #include <string.h>
76 #include <unistd.h>
77 #include "zopen.h"
78 
79 #define	BITS		16		/* Default bits. */
80 #define	HSIZE		69001		/* 95% occupancy */
81 
82 /* A code_int must be able to hold 2**BITS values of type int, and also -1. */
83 typedef long code_int;
84 typedef long count_int;
85 
86 typedef u_char char_type;
87 static char_type magic_header[] =
88 	{'\037', '\235'};		/* 1F 9D */
89 
90 #define	BIT_MASK	0x1f		/* Defines for third byte of header. */
91 #define	BLOCK_MASK	0x80
92 
93 /*
94  * Masks 0x40 and 0x20 are free.  I think 0x20 should mean that there is
95  * a fourth header byte (for expansion).
96  */
97 #define	INIT_BITS 9			/* Initial number of bits/code. */
98 
99 #define	MAXCODE(n_bits)	((1 << (n_bits)) - 1)
100 
101 struct s_zstate {
102 	FILE *zs_fp;			/* File stream for I/O */
103 	char zs_mode;			/* r or w */
104 	enum {
105 		S_START, S_MIDDLE, S_EOF
106 	} zs_state;			/* State of computation */
107 	u_int zs_n_bits;		/* Number of bits/code. */
108 	u_int zs_maxbits;		/* User settable max # bits/code. */
109 	code_int zs_maxcode;		/* Maximum code, given n_bits. */
110 	code_int zs_maxmaxcode;		/* Should NEVER generate this code. */
111 	count_int zs_htab [HSIZE];
112 	u_short zs_codetab [HSIZE];
113 	code_int zs_hsize;		/* For dynamic table sizing. */
114 	code_int zs_free_ent;		/* First unused entry. */
115 	/*
116 	 * Block compression parameters -- after all codes are used up,
117 	 * and compression rate changes, start over.
118 	 */
119 	int zs_block_compress;
120 	int zs_clear_flg;
121 	long zs_ratio;
122 	count_int zs_checkpoint;
123 	u_int zs_offset;
124 	long zs_in_count;		/* Length of input. */
125 	long zs_bytes_out;		/* Length of compressed output. */
126 	long zs_out_count;		/* # of codes output (for debugging). */
127 	char_type zs_buf[BITS];
128 	union {
129 		struct {
130 			long zs_fcode;
131 			code_int zs_ent;
132 			code_int zs_hsize_reg;
133 			int zs_hshift;
134 		} w;			/* Write paramenters */
135 		struct {
136 			char_type *zs_stackp;
137 			int zs_finchar;
138 			code_int zs_code, zs_oldcode, zs_incode;
139 			int zs_roffset, zs_size;
140 			char_type zs_gbuf[BITS];
141 		} r;			/* Read parameters */
142 	} u;
143 };
144 
145 /* Definitions to retain old variable names */
146 #define	fp		zs->zs_fp
147 #define	zmode		zs->zs_mode
148 #define	state		zs->zs_state
149 #define	n_bits		zs->zs_n_bits
150 #define	maxbits		zs->zs_maxbits
151 #define	maxcode		zs->zs_maxcode
152 #define	maxmaxcode	zs->zs_maxmaxcode
153 #define	htab		zs->zs_htab
154 #define	codetab		zs->zs_codetab
155 #define	hsize		zs->zs_hsize
156 #define	free_ent	zs->zs_free_ent
157 #define	block_compress	zs->zs_block_compress
158 #define	clear_flg	zs->zs_clear_flg
159 #define	ratio		zs->zs_ratio
160 #define	checkpoint	zs->zs_checkpoint
161 #define	offset		zs->zs_offset
162 #define	in_count	zs->zs_in_count
163 #define	bytes_out	zs->zs_bytes_out
164 #define	out_count	zs->zs_out_count
165 #define	buf		zs->zs_buf
166 #define	fcode		zs->u.w.zs_fcode
167 #define	hsize_reg	zs->u.w.zs_hsize_reg
168 #define	ent		zs->u.w.zs_ent
169 #define	hshift		zs->u.w.zs_hshift
170 #define	stackp		zs->u.r.zs_stackp
171 #define	finchar		zs->u.r.zs_finchar
172 #define	code		zs->u.r.zs_code
173 #define	oldcode		zs->u.r.zs_oldcode
174 #define	incode		zs->u.r.zs_incode
175 #define	roffset		zs->u.r.zs_roffset
176 #define	size		zs->u.r.zs_size
177 #define	gbuf		zs->u.r.zs_gbuf
178 
179 /*
180  * To save much memory, we overlay the table used by compress() with those
181  * used by decompress().  The tab_prefix table is the same size and type as
182  * the codetab.  The tab_suffix table needs 2**BITS characters.  We get this
183  * from the beginning of htab.  The output stack uses the rest of htab, and
184  * contains characters.  There is plenty of room for any possible stack
185  * (stack used to be 8000 characters).
186  */
187 
188 #define	htabof(i)	htab[i]
189 #define	codetabof(i)	codetab[i]
190 
191 #define	tab_prefixof(i)	codetabof(i)
192 #define	tab_suffixof(i)	((char_type *)(htab))[i]
193 #define	de_stack	((char_type *)&tab_suffixof(1 << BITS))
194 
195 #define	CHECK_GAP 10000		/* Ratio check interval. */
196 
197 /*
198  * the next two codes should not be changed lightly, as they must not
199  * lie within the contiguous general code space.
200  */
201 #define	FIRST	257		/* First free entry. */
202 #define	CLEAR	256		/* Table clear output code. */
203 
204 static int	cl_block(struct s_zstate *);
205 static void	cl_hash(struct s_zstate *, count_int);
206 static code_int	getcode(struct s_zstate *);
207 static int	output(struct s_zstate *, code_int);
208 static int	zclose(void *);
209 static int	zread(void *, char *, int);
210 static int	zwrite(void *, const char *, int);
211 
212 /*-
213  * Algorithm from "A Technique for High Performance Data Compression",
214  * Terry A. Welch, IEEE Computer Vol 17, No 6 (June 1984), pp 8-19.
215  *
216  * Algorithm:
217  * 	Modified Lempel-Ziv method (LZW).  Basically finds common
218  * substrings and replaces them with a variable size code.  This is
219  * deterministic, and can be done on the fly.  Thus, the decompression
220  * procedure needs no input table, but tracks the way the table was built.
221  */
222 
223 /*-
224  * compress write
225  *
226  * Algorithm:  use open addressing double hashing (no chaining) on the
227  * prefix code / next character combination.  We do a variant of Knuth's
228  * algorithm D (vol. 3, sec. 6.4) along with G. Knott's relatively-prime
229  * secondary probe.  Here, the modular division first probe is gives way
230  * to a faster exclusive-or manipulation.  Also do block compression with
231  * an adaptive reset, whereby the code table is cleared when the compression
232  * ratio decreases, but after the table fills.  The variable-length output
233  * codes are re-sized at this point, and a special CLEAR code is generated
234  * for the decompressor.  Late addition:  construct the table according to
235  * file size for noticeable speed improvement on small files.  Please direct
236  * questions about this implementation to ames!jaw.
237  */
238 static int
239 zwrite(cookie, wbp, num)
240 	void *cookie;
241 	const char *wbp;
242 	int num;
243 {
244 	code_int i;
245 	int c, disp;
246 	struct s_zstate *zs;
247 	const u_char *bp;
248 	u_char tmp;
249 	int count;
250 
251 	if (num == 0)
252 		return (0);
253 
254 	zs = cookie;
255 	count = num;
256 	bp = wbp;
257 	if (state == S_MIDDLE)
258 		goto middle;
259 	state = S_MIDDLE;
260 
261 	maxmaxcode = 1L << maxbits;
262 	if (fwrite(magic_header,
263 	    sizeof(char), sizeof(magic_header), fp) != sizeof(magic_header))
264 		return (-1);
265 	tmp = (u_char)((maxbits) | block_compress);
266 	if (fwrite(&tmp, sizeof(char), sizeof(tmp), fp) != sizeof(tmp))
267 		return (-1);
268 
269 	offset = 0;
270 	bytes_out = 3;		/* Includes 3-byte header mojo. */
271 	out_count = 0;
272 	clear_flg = 0;
273 	ratio = 0;
274 	in_count = 1;
275 	checkpoint = CHECK_GAP;
276 	maxcode = MAXCODE(n_bits = INIT_BITS);
277 	free_ent = ((block_compress) ? FIRST : 256);
278 
279 	ent = *bp++;
280 	--count;
281 
282 	hshift = 0;
283 	for (fcode = (long)hsize; fcode < 65536L; fcode *= 2L)
284 		hshift++;
285 	hshift = 8 - hshift;	/* Set hash code range bound. */
286 
287 	hsize_reg = hsize;
288 	cl_hash(zs, (count_int)hsize_reg);	/* Clear hash table. */
289 
290 middle:	for (i = 0; count--;) {
291 		c = *bp++;
292 		in_count++;
293 		fcode = (long)(((long)c << maxbits) + ent);
294 		i = ((c << hshift) ^ ent);	/* Xor hashing. */
295 
296 		if (htabof(i) == fcode) {
297 			ent = codetabof(i);
298 			continue;
299 		} else if ((long)htabof(i) < 0)	/* Empty slot. */
300 			goto nomatch;
301 		disp = hsize_reg - i;	/* Secondary hash (after G. Knott). */
302 		if (i == 0)
303 			disp = 1;
304 probe:		if ((i -= disp) < 0)
305 			i += hsize_reg;
306 
307 		if (htabof(i) == fcode) {
308 			ent = codetabof(i);
309 			continue;
310 		}
311 		if ((long)htabof(i) >= 0)
312 			goto probe;
313 nomatch:	if (output(zs, (code_int) ent) == -1)
314 			return (-1);
315 		out_count++;
316 		ent = c;
317 		if (free_ent < maxmaxcode) {
318 			codetabof(i) = free_ent++;	/* code -> hashtable */
319 			htabof(i) = fcode;
320 		} else if ((count_int)in_count >=
321 		    checkpoint && block_compress) {
322 			if (cl_block(zs) == -1)
323 				return (-1);
324 		}
325 	}
326 	return (num);
327 }
328 
329 static int
330 zclose(cookie)
331 	void *cookie;
332 {
333 	struct s_zstate *zs;
334 	int rval;
335 
336 	zs = cookie;
337 	if (zmode == 'w') {		/* Put out the final code. */
338 		if (output(zs, (code_int) ent) == -1) {
339 			(void)fclose(fp);
340 			free(zs);
341 			return (-1);
342 		}
343 		out_count++;
344 		if (output(zs, (code_int) - 1) == -1) {
345 			(void)fclose(fp);
346 			free(zs);
347 			return (-1);
348 		}
349 	}
350 	rval = fclose(fp) == EOF ? -1 : 0;
351 	free(zs);
352 	return (rval);
353 }
354 
355 /*-
356  * Output the given code.
357  * Inputs:
358  * 	code:	A n_bits-bit integer.  If == -1, then EOF.  This assumes
359  *		that n_bits =< (long)wordsize - 1.
360  * Outputs:
361  * 	Outputs code to the file.
362  * Assumptions:
363  *	Chars are 8 bits long.
364  * Algorithm:
365  * 	Maintain a BITS character long buffer (so that 8 codes will
366  * fit in it exactly).  Use the VAX insv instruction to insert each
367  * code in turn.  When the buffer fills up empty it and start over.
368  */
369 
370 static char_type lmask[9] =
371 	{0xff, 0xfe, 0xfc, 0xf8, 0xf0, 0xe0, 0xc0, 0x80, 0x00};
372 static char_type rmask[9] =
373 	{0x00, 0x01, 0x03, 0x07, 0x0f, 0x1f, 0x3f, 0x7f, 0xff};
374 
375 static int
376 output(zs, ocode)
377 	struct s_zstate *zs;
378 	code_int ocode;
379 {
380 	int r_off;
381 	u_int bits;
382 	char_type *bp;
383 
384 	r_off = offset;
385 	bits = n_bits;
386 	bp = buf;
387 	if (ocode >= 0) {
388 		/* Get to the first byte. */
389 		bp += (r_off >> 3);
390 		r_off &= 7;
391 		/*
392 		 * Since ocode is always >= 8 bits, only need to mask the first
393 		 * hunk on the left.
394 		 */
395 		*bp = (*bp & rmask[r_off]) | ((ocode << r_off) & lmask[r_off]);
396 		bp++;
397 		bits -= (8 - r_off);
398 		ocode >>= 8 - r_off;
399 		/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
400 		if (bits >= 8) {
401 			*bp++ = ocode;
402 			ocode >>= 8;
403 			bits -= 8;
404 		}
405 		/* Last bits. */
406 		if (bits)
407 			*bp = ocode;
408 		offset += n_bits;
409 		if (offset == (n_bits << 3)) {
410 			bp = buf;
411 			bits = n_bits;
412 			bytes_out += bits;
413 			if (fwrite(bp, sizeof(char), bits, fp) != bits)
414 				return (-1);
415 			bp += bits;
416 			bits = 0;
417 			offset = 0;
418 		}
419 		/*
420 		 * If the next entry is going to be too big for the ocode size,
421 		 * then increase it, if possible.
422 		 */
423 		if (free_ent > maxcode || (clear_flg > 0)) {
424 		       /*
425 			* Write the whole buffer, because the input side won't
426 			* discover the size increase until after it has read it.
427 			*/
428 			if (offset > 0) {
429 				if (fwrite(buf, 1, n_bits, fp) != n_bits)
430 					return (-1);
431 				bytes_out += n_bits;
432 			}
433 			offset = 0;
434 
435 			if (clear_flg) {
436 				maxcode = MAXCODE(n_bits = INIT_BITS);
437 				clear_flg = 0;
438 			} else {
439 				n_bits++;
440 				if (n_bits == maxbits)
441 					maxcode = maxmaxcode;
442 				else
443 					maxcode = MAXCODE(n_bits);
444 			}
445 		}
446 	} else {
447 		/* At EOF, write the rest of the buffer. */
448 		if (offset > 0) {
449 			offset = (offset + 7) / 8;
450 			if (fwrite(buf, 1, offset, fp) != offset)
451 				return (-1);
452 			bytes_out += offset;
453 		}
454 		offset = 0;
455 	}
456 	return (0);
457 }
458 
459 /*
460  * Decompress read.  This routine adapts to the codes in the file building
461  * the "string" table on-the-fly; requiring no table to be stored in the
462  * compressed file.  The tables used herein are shared with those of the
463  * compress() routine.  See the definitions above.
464  */
465 static int
466 zread(cookie, rbp, num)
467 	void *cookie;
468 	char *rbp;
469 	int num;
470 {
471 	u_int count;
472 	struct s_zstate *zs;
473 	u_char *bp, header[3];
474 
475 	if (num == 0)
476 		return (0);
477 
478 	zs = cookie;
479 	count = num;
480 	bp = (u_char *)rbp;
481 	switch (state) {
482 	case S_START:
483 		state = S_MIDDLE;
484 		break;
485 	case S_MIDDLE:
486 		goto middle;
487 	case S_EOF:
488 		goto eof;
489 	}
490 
491 	/* Check the magic number */
492 	if (fread(header,
493 	    sizeof(char), sizeof(header), fp) != sizeof(header) ||
494 	    memcmp(header, magic_header, sizeof(magic_header)) != 0) {
495 		errno = EFTYPE;
496 		return (-1);
497 	}
498 	maxbits = header[2];	/* Set -b from file. */
499 	block_compress = maxbits & BLOCK_MASK;
500 	maxbits &= BIT_MASK;
501 	maxmaxcode = 1L << maxbits;
502 	if (maxbits > BITS) {
503 		errno = EFTYPE;
504 		return (-1);
505 	}
506 	/* As above, initialize the first 256 entries in the table. */
507 	maxcode = MAXCODE(n_bits = INIT_BITS);
508 	for (code = 255; code >= 0; code--) {
509 		tab_prefixof(code) = 0;
510 		tab_suffixof(code) = (char_type) code;
511 	}
512 	free_ent = block_compress ? FIRST : 256;
513 
514 	finchar = oldcode = getcode(zs);
515 	if (oldcode == -1)	/* EOF already? */
516 		return (0);	/* Get out of here */
517 
518 	/* First code must be 8 bits = char. */
519 	*bp++ = (u_char)finchar;
520 	count--;
521 	stackp = de_stack;
522 
523 	while ((code = getcode(zs)) > -1) {
524 
525 		if ((code == CLEAR) && block_compress) {
526 			for (code = 255; code >= 0; code--)
527 				tab_prefixof(code) = 0;
528 			clear_flg = 1;
529 			free_ent = FIRST - 1;
530 			if ((code = getcode(zs)) == -1)	/* O, untimely death! */
531 				break;
532 		}
533 		incode = code;
534 
535 		/* Special case for KwKwK string. */
536 		if (code >= free_ent) {
537 			*stackp++ = finchar;
538 			code = oldcode;
539 		}
540 
541 		/* Generate output characters in reverse order. */
542 		while (code >= 256) {
543 			*stackp++ = tab_suffixof(code);
544 			code = tab_prefixof(code);
545 		}
546 		*stackp++ = finchar = tab_suffixof(code);
547 
548 		/* And put them out in forward order.  */
549 middle:		do {
550 			if (count-- == 0)
551 				return (num);
552 			*bp++ = *--stackp;
553 		} while (stackp > de_stack);
554 
555 		/* Generate the new entry. */
556 		if ((code = free_ent) < maxmaxcode) {
557 			tab_prefixof(code) = (u_short) oldcode;
558 			tab_suffixof(code) = finchar;
559 			free_ent = code + 1;
560 		}
561 
562 		/* Remember previous code. */
563 		oldcode = incode;
564 	}
565 	state = S_EOF;
566 eof:	return (num - count);
567 }
568 
569 /*-
570  * Read one code from the standard input.  If EOF, return -1.
571  * Inputs:
572  * 	stdin
573  * Outputs:
574  * 	code or -1 is returned.
575  */
576 static code_int
577 getcode(zs)
578 	struct s_zstate *zs;
579 {
580 	code_int gcode;
581 	int r_off, bits;
582 	char_type *bp;
583 
584 	bp = gbuf;
585 	if (clear_flg > 0 || roffset >= size || free_ent > maxcode) {
586 		/*
587 		 * If the next entry will be too big for the current gcode
588 		 * size, then we must increase the size.  This implies reading
589 		 * a new buffer full, too.
590 		 */
591 		if (free_ent > maxcode) {
592 			n_bits++;
593 			if (n_bits == maxbits)	/* Won't get any bigger now. */
594 				maxcode = maxmaxcode;
595 			else
596 				maxcode = MAXCODE(n_bits);
597 		}
598 		if (clear_flg > 0) {
599 			maxcode = MAXCODE(n_bits = INIT_BITS);
600 			clear_flg = 0;
601 		}
602 		size = fread(gbuf, 1, n_bits, fp);
603 		if (size <= 0)			/* End of file. */
604 			return (-1);
605 		roffset = 0;
606 		/* Round size down to integral number of codes. */
607 		size = (size << 3) - (n_bits - 1);
608 	}
609 	r_off = roffset;
610 	bits = n_bits;
611 
612 	/* Get to the first byte. */
613 	bp += (r_off >> 3);
614 	r_off &= 7;
615 
616 	/* Get first part (low order bits). */
617 	gcode = (*bp++ >> r_off);
618 	bits -= (8 - r_off);
619 	r_off = 8 - r_off;	/* Now, roffset into gcode word. */
620 
621 	/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
622 	if (bits >= 8) {
623 		gcode |= *bp++ << r_off;
624 		r_off += 8;
625 		bits -= 8;
626 	}
627 
628 	/* High order bits. */
629 	gcode |= (*bp & rmask[bits]) << r_off;
630 	roffset += n_bits;
631 
632 	return (gcode);
633 }
634 
635 static int
636 cl_block(zs)			/* Table clear for block compress. */
637 	struct s_zstate *zs;
638 {
639 	long rat;
640 
641 	checkpoint = in_count + CHECK_GAP;
642 
643 	if (in_count > 0x007fffff) {	/* Shift will overflow. */
644 		rat = bytes_out >> 8;
645 		if (rat == 0)		/* Don't divide by zero. */
646 			rat = 0x7fffffff;
647 		else
648 			rat = in_count / rat;
649 	} else
650 		rat = (in_count << 8) / bytes_out;	/* 8 fractional bits. */
651 	if (rat > ratio)
652 		ratio = rat;
653 	else {
654 		ratio = 0;
655 		cl_hash(zs, (count_int) hsize);
656 		free_ent = FIRST;
657 		clear_flg = 1;
658 		if (output(zs, (code_int) CLEAR) == -1)
659 			return (-1);
660 	}
661 	return (0);
662 }
663 
664 static void
665 cl_hash(zs, cl_hsize)			/* Reset code table. */
666 	struct s_zstate *zs;
667 	count_int cl_hsize;
668 {
669 	count_int *htab_p;
670 	long i, m1;
671 
672 	m1 = -1;
673 	htab_p = htab + cl_hsize;
674 	i = cl_hsize - 16;
675 	do {			/* Might use Sys V memset(3) here. */
676 		*(htab_p - 16) = m1;
677 		*(htab_p - 15) = m1;
678 		*(htab_p - 14) = m1;
679 		*(htab_p - 13) = m1;
680 		*(htab_p - 12) = m1;
681 		*(htab_p - 11) = m1;
682 		*(htab_p - 10) = m1;
683 		*(htab_p - 9) = m1;
684 		*(htab_p - 8) = m1;
685 		*(htab_p - 7) = m1;
686 		*(htab_p - 6) = m1;
687 		*(htab_p - 5) = m1;
688 		*(htab_p - 4) = m1;
689 		*(htab_p - 3) = m1;
690 		*(htab_p - 2) = m1;
691 		*(htab_p - 1) = m1;
692 		htab_p -= 16;
693 	} while ((i -= 16) >= 0);
694 	for (i += 16; i > 0; i--)
695 		*--htab_p = m1;
696 }
697 
698 FILE *
699 zopen(fname, mode, bits)
700 	const char *fname, *mode;
701 	int bits;
702 {
703 	struct s_zstate *zs;
704 
705 	if ((mode[0] != 'r' && mode[0] != 'w') || mode[1] != '\0' ||
706 	    bits < 0 || bits > BITS) {
707 		errno = EINVAL;
708 		return (NULL);
709 	}
710 
711 	if ((zs = calloc(1, sizeof(struct s_zstate))) == NULL)
712 		return (NULL);
713 
714 	maxbits = bits ? bits : BITS;	/* User settable max # bits/code. */
715 	maxmaxcode = 1L << maxbits;	/* Should NEVER generate this code. */
716 	hsize = HSIZE;			/* For dynamic table sizing. */
717 	free_ent = 0;			/* First unused entry. */
718 	block_compress = BLOCK_MASK;
719 	clear_flg = 0;
720 	ratio = 0;
721 	checkpoint = CHECK_GAP;
722 	in_count = 1;			/* Length of input. */
723 	out_count = 0;			/* # of codes output (for debugging). */
724 	state = S_START;
725 	roffset = 0;
726 	size = 0;
727 
728 	/*
729 	 * Layering compress on top of stdio in order to provide buffering,
730 	 * and ensure that reads and write work with the data specified.
731 	 */
732 	if ((fp = fopen(fname, mode)) == NULL) {
733 		free(zs);
734 		return (NULL);
735 	}
736 	switch (*mode) {
737 	case 'r':
738 		zmode = 'r';
739 		return (funopen(zs, zread, NULL, NULL, zclose));
740 	case 'w':
741 		zmode = 'w';
742 		return (funopen(zs, NULL, zwrite, NULL, zclose));
743 	}
744 	/* NOTREACHED */
745 	return (NULL);
746 }
747