xref: /freebsd/usr.bin/diff/diffreg.c (revision d7d962ead0b6e5e8a39202d0590022082bf5bfb6)
1 /*	$OpenBSD: diffreg.c,v 1.93 2019/06/28 13:35:00 deraadt Exp $	*/
2 
3 /*-
4  * SPDX-License-Identifier: BSD-4-Clause
5  *
6  * Copyright (C) Caldera International Inc.  2001-2002.
7  * All rights reserved.
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 and documentation must retain the above
13  *    copyright 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 or owned by Caldera
20  *	International, Inc.
21  * 4. Neither the name of Caldera International, Inc. nor the names of other
22  *    contributors may be used to endorse or promote products derived from
23  *    this software without specific prior written permission.
24  *
25  * USE OF THE SOFTWARE PROVIDED FOR UNDER THIS LICENSE BY CALDERA
26  * INTERNATIONAL, INC. AND CONTRIBUTORS ``AS IS'' AND ANY EXPRESS OR
27  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
28  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
29  * IN NO EVENT SHALL CALDERA INTERNATIONAL, INC. BE LIABLE FOR ANY DIRECT,
30  * INDIRECT INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
31  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
32  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
34  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
35  * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36  * POSSIBILITY OF SUCH DAMAGE.
37  */
38 /*-
39  * Copyright (c) 1991, 1993
40  *	The Regents of the University of California.  All rights reserved.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  * 3. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)diffreg.c   8.1 (Berkeley) 6/6/93
67  */
68 
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
71 
72 #include <sys/capsicum.h>
73 #include <sys/stat.h>
74 
75 #include <capsicum_helpers.h>
76 #include <ctype.h>
77 #include <err.h>
78 #include <errno.h>
79 #include <fcntl.h>
80 #include <paths.h>
81 #include <regex.h>
82 #include <stdbool.h>
83 #include <stddef.h>
84 #include <stdint.h>
85 #include <stdio.h>
86 #include <stdlib.h>
87 #include <string.h>
88 
89 #include "pr.h"
90 #include "diff.h"
91 #include "xmalloc.h"
92 
93 /*
94  * diff - compare two files.
95  */
96 
97 /*
98  *	Uses an algorithm due to Harold Stone, which finds
99  *	a pair of longest identical subsequences in the two
100  *	files.
101  *
102  *	The major goal is to generate the match vector J.
103  *	J[i] is the index of the line in file1 corresponding
104  *	to line i file0. J[i] = 0 if there is no
105  *	such line in file1.
106  *
107  *	Lines are hashed so as to work in core. All potential
108  *	matches are located by sorting the lines of each file
109  *	on the hash (called ``value''). In particular, this
110  *	collects the equivalence classes in file1 together.
111  *	Subroutine equiv replaces the value of each line in
112  *	file0 by the index of the first element of its
113  *	matching equivalence in (the reordered) file1.
114  *	To save space equiv squeezes file1 into a single
115  *	array member in which the equivalence classes
116  *	are simply concatenated, except that their first
117  *	members are flagged by changing sign.
118  *
119  *	Next the indices that point into member are unsorted into
120  *	array class according to the original order of file0.
121  *
122  *	The cleverness lies in routine stone. This marches
123  *	through the lines of file0, developing a vector klist
124  *	of "k-candidates". At step i a k-candidate is a matched
125  *	pair of lines x,y (x in file0 y in file1) such that
126  *	there is a common subsequence of length k
127  *	between the first i lines of file0 and the first y
128  *	lines of file1, but there is no such subsequence for
129  *	any smaller y. x is the earliest possible mate to y
130  *	that occurs in such a subsequence.
131  *
132  *	Whenever any of the members of the equivalence class of
133  *	lines in file1 matable to a line in file0 has serial number
134  *	less than the y of some k-candidate, that k-candidate
135  *	with the smallest such y is replaced. The new
136  *	k-candidate is chained (via pred) to the current
137  *	k-1 candidate so that the actual subsequence can
138  *	be recovered. When a member has serial number greater
139  *	that the y of all k-candidates, the klist is extended.
140  *	At the end, the longest subsequence is pulled out
141  *	and placed in the array J by unravel
142  *
143  *	With J in hand, the matches there recorded are
144  *	check'ed against reality to assure that no spurious
145  *	matches have crept in due to hashing. If they have,
146  *	they are broken, and "jackpot" is recorded--a harmless
147  *	matter except that a true match for a spuriously
148  *	mated line may now be unnecessarily reported as a change.
149  *
150  *	Much of the complexity of the program comes simply
151  *	from trying to minimize core utilization and
152  *	maximize the range of doable problems by dynamically
153  *	allocating what is needed and reusing what is not.
154  *	The core requirements for problems larger than somewhat
155  *	are (in words) 2*length(file0) + length(file1) +
156  *	3*(number of k-candidates installed),  typically about
157  *	6n words for files of length n.
158  */
159 
160 struct cand {
161 	int	x;
162 	int	y;
163 	int	pred;
164 };
165 
166 static struct line {
167 	int	serial;
168 	int	value;
169 } *file[2];
170 
171 /*
172  * The following struct is used to record change information when
173  * doing a "context" or "unified" diff.  (see routine "change" to
174  * understand the highly mnemonic field names)
175  */
176 struct context_vec {
177 	int	a;		/* start line in old file */
178 	int	b;		/* end line in old file */
179 	int	c;		/* start line in new file */
180 	int	d;		/* end line in new file */
181 };
182 
183 enum readhash { RH_BINARY, RH_OK, RH_EOF };
184 
185 #define MIN_PAD		1
186 static FILE	*opentemp(const char *);
187 static void	 output(char *, FILE *, char *, FILE *, int);
188 static void	 check(FILE *, FILE *, int);
189 static void	 range(int, int, const char *);
190 static void	 uni_range(int, int);
191 static void	 dump_context_vec(FILE *, FILE *, int);
192 static void	 dump_unified_vec(FILE *, FILE *, int);
193 static bool	 prepare(int, FILE *, size_t, int);
194 static void	 prune(void);
195 static void	 equiv(struct line *, int, struct line *, int, int *);
196 static void	 unravel(int);
197 static void	 unsort(struct line *, int, int *);
198 static void	 change(char *, FILE *, char *, FILE *, int, int, int, int, int *);
199 static void	 sort(struct line *, int);
200 static void	 print_header(const char *, const char *);
201 static void	 print_space(int, int, int);
202 static bool	 ignoreline_pattern(char *);
203 static bool	 ignoreline(char *, bool);
204 static int	 asciifile(FILE *);
205 static int	 fetch(long *, int, int, FILE *, int, int, int);
206 static int	 newcand(int, int, int);
207 static int	 search(int *, int, int);
208 static int	 skipline(FILE *);
209 static int	 isqrt(int);
210 static int	 stone(int *, int, int *, int *, int);
211 static enum readhash readhash(FILE *, int, unsigned *);
212 static int	 files_differ(FILE *, FILE *, int);
213 static char	*match_function(const long *, int, FILE *);
214 static char	*preadline(int, size_t, off_t);
215 
216 static int  *J;			/* will be overlaid on class */
217 static int  *class;		/* will be overlaid on file[0] */
218 static int  *klist;		/* will be overlaid on file[0] after class */
219 static int  *member;		/* will be overlaid on file[1] */
220 static int   clen;
221 static int   inifdef;		/* whether or not we are in a #ifdef block */
222 static int   len[2];
223 static int   pref, suff;	/* length of prefix and suffix */
224 static int   slen[2];
225 static int   anychange;
226 static int   hw, padding;	/* half width and padding */
227 static int   edoffset;
228 static long *ixnew;		/* will be overlaid on file[1] */
229 static long *ixold;		/* will be overlaid on klist */
230 static struct cand *clist;	/* merely a free storage pot for candidates */
231 static int   clistlen;		/* the length of clist */
232 static struct line *sfile[2];	/* shortened by pruning common prefix/suffix */
233 static int (*chrtran)(int);	/* translation table for case-folding */
234 static struct context_vec *context_vec_start;
235 static struct context_vec *context_vec_end;
236 static struct context_vec *context_vec_ptr;
237 
238 #define FUNCTION_CONTEXT_SIZE	55
239 static char lastbuf[FUNCTION_CONTEXT_SIZE];
240 static int lastline;
241 static int lastmatchline;
242 
243 static int
244 clow2low(int c)
245 {
246 
247 	return (c);
248 }
249 
250 static int
251 cup2low(int c)
252 {
253 
254 	return tolower(c);
255 }
256 
257 int
258 diffreg(char *file1, char *file2, int flags, int capsicum)
259 {
260 	FILE *f1, *f2;
261 	int i, rval;
262 	struct pr *pr = NULL;
263 	cap_rights_t rights_ro;
264 
265 	f1 = f2 = NULL;
266 	rval = D_SAME;
267 	anychange = 0;
268 	lastline = 0;
269 	lastmatchline = 0;
270 	context_vec_ptr = context_vec_start - 1;
271 
272 	 /*
273 	  * hw excludes padding and make sure when -t is not used,
274 	  * the second column always starts from the closest tab stop
275 	  */
276 	if (diff_format == D_SIDEBYSIDE) {
277 		hw = width >> 1;
278 		padding = tabsize - (hw % tabsize);
279 		if ((flags & D_EXPANDTABS) != 0 || (padding % tabsize == 0))
280 			padding = MIN_PAD;
281 
282 		hw = (width >> 1) -
283 		    ((padding == MIN_PAD) ? (padding << 1) : padding) - 1;
284 	}
285 
286 
287 	if (flags & D_IGNORECASE)
288 		chrtran = cup2low;
289 	else
290 		chrtran = clow2low;
291 	if (S_ISDIR(stb1.st_mode) != S_ISDIR(stb2.st_mode))
292 		return (S_ISDIR(stb1.st_mode) ? D_MISMATCH1 : D_MISMATCH2);
293 	if (strcmp(file1, "-") == 0 && strcmp(file2, "-") == 0)
294 		goto closem;
295 
296 	if (flags & D_EMPTY1)
297 		f1 = fopen(_PATH_DEVNULL, "r");
298 	else {
299 		if (!S_ISREG(stb1.st_mode)) {
300 			if ((f1 = opentemp(file1)) == NULL ||
301 			    fstat(fileno(f1), &stb1) == -1) {
302 				warn("%s", file1);
303 				rval = D_ERROR;
304 				status |= 2;
305 				goto closem;
306 			}
307 		} else if (strcmp(file1, "-") == 0)
308 			f1 = stdin;
309 		else
310 			f1 = fopen(file1, "r");
311 	}
312 	if (f1 == NULL) {
313 		warn("%s", file1);
314 		rval = D_ERROR;
315 		status |= 2;
316 		goto closem;
317 	}
318 
319 	if (flags & D_EMPTY2)
320 		f2 = fopen(_PATH_DEVNULL, "r");
321 	else {
322 		if (!S_ISREG(stb2.st_mode)) {
323 			if ((f2 = opentemp(file2)) == NULL ||
324 			    fstat(fileno(f2), &stb2) == -1) {
325 				warn("%s", file2);
326 				rval = D_ERROR;
327 				status |= 2;
328 				goto closem;
329 			}
330 		} else if (strcmp(file2, "-") == 0)
331 			f2 = stdin;
332 		else
333 			f2 = fopen(file2, "r");
334 	}
335 	if (f2 == NULL) {
336 		warn("%s", file2);
337 		rval = D_ERROR;
338 		status |= 2;
339 		goto closem;
340 	}
341 
342 	if (lflag)
343 		pr = start_pr(file1, file2);
344 
345 	if (capsicum) {
346 		cap_rights_init(&rights_ro, CAP_READ, CAP_FSTAT, CAP_SEEK);
347 		if (caph_rights_limit(fileno(f1), &rights_ro) < 0)
348 			err(2, "unable to limit rights on: %s", file1);
349 		if (caph_rights_limit(fileno(f2), &rights_ro) < 0)
350 			err(2, "unable to limit rights on: %s", file2);
351 		if (fileno(f1) == STDIN_FILENO || fileno(f2) == STDIN_FILENO) {
352 			/* stdin has already been limited */
353 			if (caph_limit_stderr() == -1)
354 				err(2, "unable to limit stderr");
355 			if (caph_limit_stdout() == -1)
356 				err(2, "unable to limit stdout");
357 		} else if (caph_limit_stdio() == -1)
358 				err(2, "unable to limit stdio");
359 
360 		caph_cache_catpages();
361 		caph_cache_tzdata();
362 		if (caph_enter() < 0)
363 			err(2, "unable to enter capability mode");
364 	}
365 
366 	switch (files_differ(f1, f2, flags)) {
367 	case 0:
368 		goto closem;
369 	case 1:
370 		break;
371 	default:
372 		/* error */
373 		rval = D_ERROR;
374 		status |= 2;
375 		goto closem;
376 	}
377 
378 	if (diff_format == D_BRIEF && ignore_pats == NULL &&
379 	    (flags & (D_FOLDBLANKS|D_IGNOREBLANKS|D_IGNORECASE|D_STRIPCR)) == 0)
380 	{
381 		rval = D_DIFFER;
382 		status |= 1;
383 		goto closem;
384 	}
385 	if ((flags & D_FORCEASCII) != 0) {
386 		(void)prepare(0, f1, stb1.st_size, flags);
387 		(void)prepare(1, f2, stb2.st_size, flags);
388 	} else if (!asciifile(f1) || !asciifile(f2) ||
389 		    !prepare(0, f1, stb1.st_size, flags) ||
390 		    !prepare(1, f2, stb2.st_size, flags)) {
391 		rval = D_BINARY;
392 		status |= 1;
393 		goto closem;
394 	}
395 
396 	prune();
397 	sort(sfile[0], slen[0]);
398 	sort(sfile[1], slen[1]);
399 
400 	member = (int *)file[1];
401 	equiv(sfile[0], slen[0], sfile[1], slen[1], member);
402 	member = xreallocarray(member, slen[1] + 2, sizeof(*member));
403 
404 	class = (int *)file[0];
405 	unsort(sfile[0], slen[0], class);
406 	class = xreallocarray(class, slen[0] + 2, sizeof(*class));
407 
408 	klist = xcalloc(slen[0] + 2, sizeof(*klist));
409 	clen = 0;
410 	clistlen = 100;
411 	clist = xcalloc(clistlen, sizeof(*clist));
412 	i = stone(class, slen[0], member, klist, flags);
413 	free(member);
414 	free(class);
415 
416 	J = xreallocarray(J, len[0] + 2, sizeof(*J));
417 	unravel(klist[i]);
418 	free(clist);
419 	free(klist);
420 
421 	ixold = xreallocarray(ixold, len[0] + 2, sizeof(*ixold));
422 	ixnew = xreallocarray(ixnew, len[1] + 2, sizeof(*ixnew));
423 	check(f1, f2, flags);
424 	output(file1, f1, file2, f2, flags);
425 
426 closem:
427 	if (pr != NULL)
428 		stop_pr(pr);
429 	if (anychange) {
430 		status |= 1;
431 		if (rval == D_SAME)
432 			rval = D_DIFFER;
433 	}
434 	if (f1 != NULL)
435 		fclose(f1);
436 	if (f2 != NULL)
437 		fclose(f2);
438 
439 	return (rval);
440 }
441 
442 /*
443  * Check to see if the given files differ.
444  * Returns 0 if they are the same, 1 if different, and -1 on error.
445  * XXX - could use code from cmp(1) [faster]
446  */
447 static int
448 files_differ(FILE *f1, FILE *f2, int flags)
449 {
450 	char buf1[BUFSIZ], buf2[BUFSIZ];
451 	size_t i, j;
452 
453 	if ((flags & (D_EMPTY1|D_EMPTY2)) || stb1.st_size != stb2.st_size ||
454 	    (stb1.st_mode & S_IFMT) != (stb2.st_mode & S_IFMT))
455 		return (1);
456 	for (;;) {
457 		i = fread(buf1, 1, sizeof(buf1), f1);
458 		j = fread(buf2, 1, sizeof(buf2), f2);
459 		if ((!i && ferror(f1)) || (!j && ferror(f2)))
460 			return (-1);
461 		if (i != j)
462 			return (1);
463 		if (i == 0)
464 			return (0);
465 		if (memcmp(buf1, buf2, i) != 0)
466 			return (1);
467 	}
468 }
469 
470 static FILE *
471 opentemp(const char *f)
472 {
473 	char buf[BUFSIZ], tempfile[PATH_MAX];
474 	ssize_t nread;
475 	int ifd, ofd;
476 
477 	if (strcmp(f, "-") == 0)
478 		ifd = STDIN_FILENO;
479 	else if ((ifd = open(f, O_RDONLY, 0644)) == -1)
480 		return (NULL);
481 
482 	(void)strlcpy(tempfile, _PATH_TMP "/diff.XXXXXXXX", sizeof(tempfile));
483 
484 	if ((ofd = mkstemp(tempfile)) == -1) {
485 		close(ifd);
486 		return (NULL);
487 	}
488 	unlink(tempfile);
489 	while ((nread = read(ifd, buf, BUFSIZ)) > 0) {
490 		if (write(ofd, buf, nread) != nread) {
491 			close(ifd);
492 			close(ofd);
493 			return (NULL);
494 		}
495 	}
496 	close(ifd);
497 	lseek(ofd, (off_t)0, SEEK_SET);
498 	return (fdopen(ofd, "r"));
499 }
500 
501 char *
502 splice(char *dir, char *path)
503 {
504 	char *tail, *buf;
505 	size_t dirlen;
506 
507 	dirlen = strlen(dir);
508 	while (dirlen != 0 && dir[dirlen - 1] == '/')
509 	    dirlen--;
510 	if ((tail = strrchr(path, '/')) == NULL)
511 		tail = path;
512 	else
513 		tail++;
514 	xasprintf(&buf, "%.*s/%s", (int)dirlen, dir, tail);
515 	return (buf);
516 }
517 
518 static bool
519 prepare(int i, FILE *fd, size_t filesize, int flags)
520 {
521 	struct line *p;
522 	unsigned h;
523 	size_t sz, j = 0;
524 	enum readhash r;
525 
526 	rewind(fd);
527 
528 	sz = MIN(filesize, SIZE_MAX) / 25;
529 	if (sz < 100)
530 		sz = 100;
531 
532 	p = xcalloc(sz + 3, sizeof(*p));
533 	while ((r = readhash(fd, flags, &h)) != RH_EOF)
534 		switch (r) {
535 		case RH_EOF: /* otherwise clang complains */
536 		case RH_BINARY:
537 			return (false);
538 		case RH_OK:
539 			if (j == sz) {
540 				sz = sz * 3 / 2;
541 				p = xreallocarray(p, sz + 3, sizeof(*p));
542 			}
543 			p[++j].value = h;
544 		}
545 
546 	len[i] = j;
547 	file[i] = p;
548 
549 	return (true);
550 }
551 
552 static void
553 prune(void)
554 {
555 	int i, j;
556 
557 	for (pref = 0; pref < len[0] && pref < len[1] &&
558 	    file[0][pref + 1].value == file[1][pref + 1].value;
559 	    pref++)
560 		;
561 	for (suff = 0; suff < len[0] - pref && suff < len[1] - pref &&
562 	    file[0][len[0] - suff].value == file[1][len[1] - suff].value;
563 	    suff++)
564 		;
565 	for (j = 0; j < 2; j++) {
566 		sfile[j] = file[j] + pref;
567 		slen[j] = len[j] - pref - suff;
568 		for (i = 0; i <= slen[j]; i++)
569 			sfile[j][i].serial = i;
570 	}
571 }
572 
573 static void
574 equiv(struct line *a, int n, struct line *b, int m, int *c)
575 {
576 	int i, j;
577 
578 	i = j = 1;
579 	while (i <= n && j <= m) {
580 		if (a[i].value < b[j].value)
581 			a[i++].value = 0;
582 		else if (a[i].value == b[j].value)
583 			a[i++].value = j;
584 		else
585 			j++;
586 	}
587 	while (i <= n)
588 		a[i++].value = 0;
589 	b[m + 1].value = 0;
590 	j = 0;
591 	while (++j <= m) {
592 		c[j] = -b[j].serial;
593 		while (b[j + 1].value == b[j].value) {
594 			j++;
595 			c[j] = b[j].serial;
596 		}
597 	}
598 	c[j] = -1;
599 }
600 
601 /* Code taken from ping.c */
602 static int
603 isqrt(int n)
604 {
605 	int y, x = 1;
606 
607 	if (n == 0)
608 		return (0);
609 
610 	do { /* newton was a stinker */
611 		y = x;
612 		x = n / x;
613 		x += y;
614 		x /= 2;
615 	} while ((x - y) > 1 || (x - y) < -1);
616 
617 	return (x);
618 }
619 
620 static int
621 stone(int *a, int n, int *b, int *c, int flags)
622 {
623 	int i, k, y, j, l;
624 	int oldc, tc, oldl, sq;
625 	u_int numtries, bound;
626 
627 	if (flags & D_MINIMAL)
628 		bound = UINT_MAX;
629 	else {
630 		sq = isqrt(n);
631 		bound = MAX(256, sq);
632 	}
633 
634 	k = 0;
635 	c[0] = newcand(0, 0, 0);
636 	for (i = 1; i <= n; i++) {
637 		j = a[i];
638 		if (j == 0)
639 			continue;
640 		y = -b[j];
641 		oldl = 0;
642 		oldc = c[0];
643 		numtries = 0;
644 		do {
645 			if (y <= clist[oldc].y)
646 				continue;
647 			l = search(c, k, y);
648 			if (l != oldl + 1)
649 				oldc = c[l - 1];
650 			if (l <= k) {
651 				if (clist[c[l]].y <= y)
652 					continue;
653 				tc = c[l];
654 				c[l] = newcand(i, y, oldc);
655 				oldc = tc;
656 				oldl = l;
657 				numtries++;
658 			} else {
659 				c[l] = newcand(i, y, oldc);
660 				k++;
661 				break;
662 			}
663 		} while ((y = b[++j]) > 0 && numtries < bound);
664 	}
665 	return (k);
666 }
667 
668 static int
669 newcand(int x, int y, int pred)
670 {
671 	struct cand *q;
672 
673 	if (clen == clistlen) {
674 		clistlen = clistlen * 11 / 10;
675 		clist = xreallocarray(clist, clistlen, sizeof(*clist));
676 	}
677 	q = clist + clen;
678 	q->x = x;
679 	q->y = y;
680 	q->pred = pred;
681 	return (clen++);
682 }
683 
684 static int
685 search(int *c, int k, int y)
686 {
687 	int i, j, l, t;
688 
689 	if (clist[c[k]].y < y)	/* quick look for typical case */
690 		return (k + 1);
691 	i = 0;
692 	j = k + 1;
693 	for (;;) {
694 		l = (i + j) / 2;
695 		if (l <= i)
696 			break;
697 		t = clist[c[l]].y;
698 		if (t > y)
699 			j = l;
700 		else if (t < y)
701 			i = l;
702 		else
703 			return (l);
704 	}
705 	return (l + 1);
706 }
707 
708 static void
709 unravel(int p)
710 {
711 	struct cand *q;
712 	int i;
713 
714 	for (i = 0; i <= len[0]; i++)
715 		J[i] = i <= pref ? i :
716 		    i > len[0] - suff ? i + len[1] - len[0] : 0;
717 	for (q = clist + p; q->y != 0; q = clist + q->pred)
718 		J[q->x + pref] = q->y + pref;
719 }
720 
721 /*
722  * Check does double duty:
723  *  1.	ferret out any fortuitous correspondences due
724  *	to confounding by hashing (which result in "jackpot")
725  *  2.  collect random access indexes to the two files
726  */
727 static void
728 check(FILE *f1, FILE *f2, int flags)
729 {
730 	int i, j, jackpot, c, d;
731 	long ctold, ctnew;
732 
733 	rewind(f1);
734 	rewind(f2);
735 	j = 1;
736 	ixold[0] = ixnew[0] = 0;
737 	jackpot = 0;
738 	ctold = ctnew = 0;
739 	for (i = 1; i <= len[0]; i++) {
740 		if (J[i] == 0) {
741 			ixold[i] = ctold += skipline(f1);
742 			continue;
743 		}
744 		while (j < J[i]) {
745 			ixnew[j] = ctnew += skipline(f2);
746 			j++;
747 		}
748 		if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS|D_IGNORECASE|D_STRIPCR)) {
749 			for (;;) {
750 				c = getc(f1);
751 				d = getc(f2);
752 				/*
753 				 * GNU diff ignores a missing newline
754 				 * in one file for -b or -w.
755 				 */
756 				if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) {
757 					if (c == EOF && d == '\n') {
758 						ctnew++;
759 						break;
760 					} else if (c == '\n' && d == EOF) {
761 						ctold++;
762 						break;
763 					}
764 				}
765 				ctold++;
766 				ctnew++;
767 				if (flags & D_STRIPCR && (c == '\r' || d == '\r')) {
768 					if (c == '\r') {
769 						if ((c = getc(f1)) == '\n') {
770 							ctold++;
771 						} else {
772 							ungetc(c, f1);
773 						}
774 					}
775 					if (d == '\r') {
776 						if ((d = getc(f2)) == '\n') {
777 							ctnew++;
778 						} else {
779 							ungetc(d, f2);
780 						}
781 					}
782 					break;
783 				}
784 				if ((flags & D_FOLDBLANKS) && isspace(c) &&
785 				    isspace(d)) {
786 					do {
787 						if (c == '\n')
788 							break;
789 						ctold++;
790 					} while (isspace(c = getc(f1)));
791 					do {
792 						if (d == '\n')
793 							break;
794 						ctnew++;
795 					} while (isspace(d = getc(f2)));
796 				} else if ((flags & D_IGNOREBLANKS)) {
797 					while (isspace(c) && c != '\n') {
798 						c = getc(f1);
799 						ctold++;
800 					}
801 					while (isspace(d) && d != '\n') {
802 						d = getc(f2);
803 						ctnew++;
804 					}
805 				}
806 				if (chrtran(c) != chrtran(d)) {
807 					jackpot++;
808 					J[i] = 0;
809 					if (c != '\n' && c != EOF)
810 						ctold += skipline(f1);
811 					if (d != '\n' && c != EOF)
812 						ctnew += skipline(f2);
813 					break;
814 				}
815 				if (c == '\n' || c == EOF)
816 					break;
817 			}
818 		} else {
819 			for (;;) {
820 				ctold++;
821 				ctnew++;
822 				if ((c = getc(f1)) != (d = getc(f2))) {
823 					/* jackpot++; */
824 					J[i] = 0;
825 					if (c != '\n' && c != EOF)
826 						ctold += skipline(f1);
827 					if (d != '\n' && c != EOF)
828 						ctnew += skipline(f2);
829 					break;
830 				}
831 				if (c == '\n' || c == EOF)
832 					break;
833 			}
834 		}
835 		ixold[i] = ctold;
836 		ixnew[j] = ctnew;
837 		j++;
838 	}
839 	for (; j <= len[1]; j++) {
840 		ixnew[j] = ctnew += skipline(f2);
841 	}
842 	/*
843 	 * if (jackpot)
844 	 *	fprintf(stderr, "jackpot\n");
845 	 */
846 }
847 
848 /* shellsort CACM #201 */
849 static void
850 sort(struct line *a, int n)
851 {
852 	struct line *ai, *aim, w;
853 	int j, m = 0, k;
854 
855 	if (n == 0)
856 		return;
857 	for (j = 1; j <= n; j *= 2)
858 		m = 2 * j - 1;
859 	for (m /= 2; m != 0; m /= 2) {
860 		k = n - m;
861 		for (j = 1; j <= k; j++) {
862 			for (ai = &a[j]; ai > a; ai -= m) {
863 				aim = &ai[m];
864 				if (aim < ai)
865 					break;	/* wraparound */
866 				if (aim->value > ai[0].value ||
867 				    (aim->value == ai[0].value &&
868 					aim->serial > ai[0].serial))
869 					break;
870 				w.value = ai[0].value;
871 				ai[0].value = aim->value;
872 				aim->value = w.value;
873 				w.serial = ai[0].serial;
874 				ai[0].serial = aim->serial;
875 				aim->serial = w.serial;
876 			}
877 		}
878 	}
879 }
880 
881 static void
882 unsort(struct line *f, int l, int *b)
883 {
884 	int *a, i;
885 
886 	a = xcalloc(l + 1, sizeof(*a));
887 	for (i = 1; i <= l; i++)
888 		a[f[i].serial] = f[i].value;
889 	for (i = 1; i <= l; i++)
890 		b[i] = a[i];
891 	free(a);
892 }
893 
894 static int
895 skipline(FILE *f)
896 {
897 	int i, c;
898 
899 	for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++)
900 		continue;
901 	return (i);
902 }
903 
904 static void
905 output(char *file1, FILE *f1, char *file2, FILE *f2, int flags)
906 {
907 	int i, j, m, i0, i1, j0, j1, nc;
908 
909 	rewind(f1);
910 	rewind(f2);
911 	m = len[0];
912 	J[0] = 0;
913 	J[m + 1] = len[1] + 1;
914 	if (diff_format != D_EDIT) {
915 		for (i0 = 1; i0 <= m; i0 = i1 + 1) {
916 			while (i0 <= m && J[i0] == J[i0 - 1] + 1){
917 				if (diff_format == D_SIDEBYSIDE &&
918 				    suppress_common != 1) {
919 					nc = fetch(ixold, i0, i0, f1, '\0',
920 					    1, flags);
921 					print_space(nc,
922 					    (hw - nc) + (padding << 1) + 1,
923 					    flags);
924 					fetch(ixnew, J[i0], J[i0], f2, '\0',
925 					    0, flags);
926 					printf("\n");
927 				}
928 				i0++;
929 			}
930 			j0 = J[i0 - 1] + 1;
931 			i1 = i0 - 1;
932 			while (i1 < m && J[i1 + 1] == 0)
933 				i1++;
934 			j1 = J[i1 + 1] - 1;
935 			J[i1] = j1;
936 
937 			/*
938 			 * When using side-by-side, lines from both of the
939 			 * files are printed. The algorithm used by diff(1)
940 			 * identifies the ranges in which two files differ.
941 			 * See the change() function below.
942 			 * The for loop below consumes the shorter range,
943 			 * whereas one of the while loops deals with the
944 			 * longer one.
945 			 */
946 			if (diff_format == D_SIDEBYSIDE) {
947 				for (i=i0, j=j0; i<=i1 && j<=j1; i++, j++)
948 					change(file1, f1, file2, f2, i, i,
949 					    j, j, &flags);
950 
951 				while (i <= i1) {
952 					change(file1, f1, file2, f2,
953 					    i, i, j+1, j, &flags);
954 					i++;
955 				}
956 
957 				while (j <= j1) {
958 					change(file1, f1, file2, f2,
959 					    i+1, i, j, j, &flags);
960 					j++;
961 				}
962 			} else
963 				change(file1, f1, file2, f2, i0, i1, j0,
964 				    j1, &flags);
965 		}
966 	} else {
967 		for (i0 = m; i0 >= 1; i0 = i1 - 1) {
968 			while (i0 >= 1 && J[i0] == J[i0 + 1] - 1 && J[i0] != 0)
969 				i0--;
970 			j0 = J[i0 + 1] - 1;
971 			i1 = i0 + 1;
972 			while (i1 > 1 && J[i1 - 1] == 0)
973 				i1--;
974 			j1 = J[i1 - 1] + 1;
975 			J[i1] = j1;
976 			change(file1, f1, file2, f2, i1, i0, j1, j0, &flags);
977 		}
978 	}
979 	if (m == 0)
980 		change(file1, f1, file2, f2, 1, 0, 1, len[1], &flags);
981 	if (diff_format == D_IFDEF || diff_format == D_GFORMAT) {
982 		for (;;) {
983 #define	c i0
984 			if ((c = getc(f1)) == EOF)
985 				return;
986 			printf("%c", c);
987 		}
988 #undef c
989 	}
990 	if (anychange != 0) {
991 		if (diff_format == D_CONTEXT)
992 			dump_context_vec(f1, f2, flags);
993 		else if (diff_format == D_UNIFIED)
994 			dump_unified_vec(f1, f2, flags);
995 	}
996 }
997 
998 static void
999 range(int a, int b, const char *separator)
1000 {
1001 	printf("%d", a > b ? b : a);
1002 	if (a < b)
1003 		printf("%s%d", separator, b);
1004 }
1005 
1006 static void
1007 uni_range(int a, int b)
1008 {
1009 	if (a < b)
1010 		printf("%d,%d", a, b - a + 1);
1011 	else if (a == b)
1012 		printf("%d", b);
1013 	else
1014 		printf("%d,0", b);
1015 }
1016 
1017 static char *
1018 preadline(int fd, size_t rlen, off_t off)
1019 {
1020 	char *line;
1021 	ssize_t nr;
1022 
1023 	line = xmalloc(rlen + 1);
1024 	if ((nr = pread(fd, line, rlen, off)) == -1)
1025 		err(2, "preadline");
1026 	if (nr > 0 && line[nr-1] == '\n')
1027 		nr--;
1028 	line[nr] = '\0';
1029 	return (line);
1030 }
1031 
1032 static bool
1033 ignoreline_pattern(char *line)
1034 {
1035 	int ret;
1036 
1037 	ret = regexec(&ignore_re, line, 0, NULL, 0);
1038 	free(line);
1039 	return (ret == 0);	/* if it matched, it should be ignored. */
1040 }
1041 
1042 static bool
1043 ignoreline(char *line, bool skip_blanks)
1044 {
1045 
1046 	if (ignore_pats != NULL && skip_blanks)
1047 		return (ignoreline_pattern(line) || *line == '\0');
1048 	if (ignore_pats != NULL)
1049 		return (ignoreline_pattern(line));
1050 	if (skip_blanks)
1051 		return (*line == '\0');
1052 	/* No ignore criteria specified */
1053 	return (false);
1054 }
1055 
1056 /*
1057  * Indicate that there is a difference between lines a and b of the from file
1058  * to get to lines c to d of the to file.  If a is greater then b then there
1059  * are no lines in the from file involved and this means that there were
1060  * lines appended (beginning at b).  If c is greater than d then there are
1061  * lines missing from the to file.
1062  */
1063 static void
1064 change(char *file1, FILE *f1, char *file2, FILE *f2, int a, int b, int c, int d,
1065     int *pflags)
1066 {
1067 	static size_t max_context = 64;
1068 	long curpos;
1069 	int i, nc;
1070 	const char *walk;
1071 	bool skip_blanks;
1072 
1073 	skip_blanks = (*pflags & D_SKIPBLANKLINES);
1074 restart:
1075 	if ((diff_format != D_IFDEF || diff_format == D_GFORMAT) &&
1076 	    a > b && c > d)
1077 		return;
1078 	if (ignore_pats != NULL || skip_blanks) {
1079 		char *line;
1080 		/*
1081 		 * All lines in the change, insert, or delete must
1082 		 * match an ignore pattern for the change to be
1083 		 * ignored.
1084 		 */
1085 		if (a <= b) {		/* Changes and deletes. */
1086 			for (i = a; i <= b; i++) {
1087 				line = preadline(fileno(f1),
1088 				    ixold[i] - ixold[i - 1], ixold[i - 1]);
1089 				if (!ignoreline(line, skip_blanks))
1090 					goto proceed;
1091 			}
1092 		}
1093 		if (a > b || c <= d) {	/* Changes and inserts. */
1094 			for (i = c; i <= d; i++) {
1095 				line = preadline(fileno(f2),
1096 				    ixnew[i] - ixnew[i - 1], ixnew[i - 1]);
1097 				if (!ignoreline(line, skip_blanks))
1098 					goto proceed;
1099 			}
1100 		}
1101 		return;
1102 	}
1103 proceed:
1104 	if (*pflags & D_HEADER && diff_format != D_BRIEF) {
1105 		printf("%s %s %s\n", diffargs, file1, file2);
1106 		*pflags &= ~D_HEADER;
1107 	}
1108 	if (diff_format == D_CONTEXT || diff_format == D_UNIFIED) {
1109 		/*
1110 		 * Allocate change records as needed.
1111 		 */
1112 		if (context_vec_ptr == context_vec_end - 1) {
1113 			ptrdiff_t offset = context_vec_ptr - context_vec_start;
1114 			max_context <<= 1;
1115 			context_vec_start = xreallocarray(context_vec_start,
1116 			    max_context, sizeof(*context_vec_start));
1117 			context_vec_end = context_vec_start + max_context;
1118 			context_vec_ptr = context_vec_start + offset;
1119 		}
1120 		if (anychange == 0) {
1121 			/*
1122 			 * Print the context/unidiff header first time through.
1123 			 */
1124 			print_header(file1, file2);
1125 			anychange = 1;
1126 		} else if (a > context_vec_ptr->b + (2 * diff_context) + 1 &&
1127 		    c > context_vec_ptr->d + (2 * diff_context) + 1) {
1128 			/*
1129 			 * If this change is more than 'diff_context' lines from the
1130 			 * previous change, dump the record and reset it.
1131 			 */
1132 			if (diff_format == D_CONTEXT)
1133 				dump_context_vec(f1, f2, *pflags);
1134 			else
1135 				dump_unified_vec(f1, f2, *pflags);
1136 		}
1137 		context_vec_ptr++;
1138 		context_vec_ptr->a = a;
1139 		context_vec_ptr->b = b;
1140 		context_vec_ptr->c = c;
1141 		context_vec_ptr->d = d;
1142 		return;
1143 	}
1144 	if (anychange == 0)
1145 		anychange = 1;
1146 	switch (diff_format) {
1147 	case D_BRIEF:
1148 		return;
1149 	case D_NORMAL:
1150 	case D_EDIT:
1151 		range(a, b, ",");
1152 		printf("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1153 		if (diff_format == D_NORMAL)
1154 			range(c, d, ",");
1155 		printf("\n");
1156 		break;
1157 	case D_REVERSE:
1158 		printf("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1159 		range(a, b, " ");
1160 		printf("\n");
1161 		break;
1162 	case D_NREVERSE:
1163 		if (a > b)
1164 			printf("a%d %d\n", b, d - c + 1);
1165 		else {
1166 			printf("d%d %d\n", a, b - a + 1);
1167 			if (!(c > d))
1168 				/* add changed lines */
1169 				printf("a%d %d\n", b, d - c + 1);
1170 		}
1171 		break;
1172 	}
1173 	if (diff_format == D_GFORMAT) {
1174 		curpos = ftell(f1);
1175 		/* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1176 		nc = ixold[a > b ? b : a - 1] - curpos;
1177 		for (i = 0; i < nc; i++)
1178 			printf("%c", getc(f1));
1179 		for (walk = group_format; *walk != '\0'; walk++) {
1180 			if (*walk == '%') {
1181 				walk++;
1182 				switch (*walk) {
1183 				case '<':
1184 					fetch(ixold, a, b, f1, '<', 1, *pflags);
1185 					break;
1186 				case '>':
1187 					fetch(ixnew, c, d, f2, '>', 0, *pflags);
1188 					break;
1189 				default:
1190 					printf("%%%c", *walk);
1191 					break;
1192 				}
1193 				continue;
1194 			}
1195 			printf("%c", *walk);
1196 		}
1197 	}
1198 	if (diff_format == D_SIDEBYSIDE) {
1199 		if (a > b) {
1200 			print_space(0, hw + padding , *pflags);
1201 		} else {
1202 			nc = fetch(ixold, a, b, f1, '\0', 1, *pflags);
1203 			print_space(nc, hw - nc + padding, *pflags);
1204 		}
1205 		printf("%c", (a>b)? '>' : ((c>d)? '<' : '|'));
1206 		print_space(hw + padding + 1 , padding, *pflags);
1207 		fetch(ixnew, c, d, f2, '\0', 0, *pflags);
1208 		printf("\n");
1209 	}
1210 	if (diff_format == D_NORMAL || diff_format == D_IFDEF) {
1211 		fetch(ixold, a, b, f1, '<', 1, *pflags);
1212 		if (a <= b && c <= d && diff_format == D_NORMAL)
1213 			printf("---\n");
1214 	}
1215 	if (diff_format != D_GFORMAT && diff_format != D_SIDEBYSIDE)
1216 		fetch(ixnew, c, d, f2, diff_format == D_NORMAL ? '>' : '\0', 0, *pflags);
1217 	if (edoffset != 0 && diff_format == D_EDIT) {
1218 		/*
1219 		 * A non-zero edoffset value for D_EDIT indicates that the
1220 		 * last line printed was a bare dot (".") that has been
1221 		 * escaped as ".." to prevent ed(1) from misinterpreting
1222 		 * it.  We have to add a substitute command to change this
1223 		 * back and restart where we left off.
1224 		 */
1225 		printf(".\n");
1226 		printf("%ds/.//\n", a + edoffset - 1);
1227 		b = a + edoffset - 1;
1228 		a = b + 1;
1229 		c += edoffset;
1230 		goto restart;
1231 	}
1232 	if ((diff_format == D_EDIT || diff_format == D_REVERSE) && c <= d)
1233 		printf(".\n");
1234 	if (inifdef) {
1235 		printf("#endif /* %s */\n", ifdefname);
1236 		inifdef = 0;
1237 	}
1238 }
1239 
1240 static int
1241 fetch(long *f, int a, int b, FILE *lb, int ch, int oldfile, int flags)
1242 {
1243 	int i, j, c, lastc, col, nc, newcol;
1244 
1245 	edoffset = 0;
1246 	nc = 0;
1247 	/*
1248 	 * When doing #ifdef's, copy down to current line
1249 	 * if this is the first file, so that stuff makes it to output.
1250 	 */
1251 	if ((diff_format == D_IFDEF) && oldfile) {
1252 		long curpos = ftell(lb);
1253 		/* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1254 		nc = f[a > b ? b : a - 1] - curpos;
1255 		for (i = 0; i < nc; i++)
1256 			printf("%c", getc(lb));
1257 	}
1258 	if (a > b)
1259 		return (0);
1260 	if (diff_format == D_IFDEF) {
1261 		if (inifdef) {
1262 			printf("#else /* %s%s */\n",
1263 			    oldfile == 1 ? "!" : "", ifdefname);
1264 		} else {
1265 			if (oldfile)
1266 				printf("#ifndef %s\n", ifdefname);
1267 			else
1268 				printf("#ifdef %s\n", ifdefname);
1269 		}
1270 		inifdef = 1 + oldfile;
1271 	}
1272 	for (i = a; i <= b; i++) {
1273 		fseek(lb, f[i - 1], SEEK_SET);
1274 		nc = (f[i] - f[i - 1]);
1275 		if (diff_format == D_SIDEBYSIDE && hw < nc)
1276 			nc = hw;
1277 		if ((diff_format != D_IFDEF && diff_format != D_GFORMAT) &&
1278 		    ch != '\0') {
1279 			printf("%c", ch);
1280 			if (Tflag && (diff_format == D_NORMAL ||
1281 			    diff_format == D_CONTEXT ||
1282 			    diff_format == D_UNIFIED))
1283 				printf("\t");
1284 			else if (diff_format != D_UNIFIED)
1285 				printf(" ");
1286 		}
1287 		col = 0;
1288 		for (j = 0, lastc = '\0'; j < nc; j++, lastc = c) {
1289 			c = getc(lb);
1290 			if (flags & D_STRIPCR && c == '\r') {
1291 				if ((c = getc(lb)) == '\n')
1292 					j++;
1293 				else {
1294 					ungetc(c, lb);
1295 					c = '\r';
1296 				}
1297 			}
1298 			if (c == EOF) {
1299 				if (diff_format == D_EDIT ||
1300 				    diff_format == D_REVERSE ||
1301 				    diff_format == D_NREVERSE)
1302 					warnx("No newline at end of file");
1303 				else
1304 					printf("\n\\ No newline at end of "
1305 					    "file\n");
1306 				return col;
1307 			}
1308 			/*
1309 			 * when using --side-by-side, col needs to be increased
1310 			 * in any case to keep the columns aligned
1311 			 */
1312 			if (c == '\t') {
1313 				if (flags & D_EXPANDTABS) {
1314 					newcol = ((col/tabsize)+1)*tabsize;
1315 					do {
1316 						if (diff_format == D_SIDEBYSIDE)
1317 							j++;
1318 						printf(" ");
1319 					} while (++col < newcol && j < nc);
1320 				} else {
1321 					if (diff_format == D_SIDEBYSIDE) {
1322 						if ((j + tabsize) > nc) {
1323 							printf("%*s",
1324 							nc - j,"");
1325 							j = col = nc;
1326 						} else {
1327 							printf("\t");
1328 							col += tabsize - 1;
1329 							j += tabsize - 1;
1330 						}
1331 					} else {
1332 						printf("\t");
1333 						col++;
1334 					}
1335 				}
1336 			} else {
1337 				if (diff_format == D_EDIT && j == 1 && c == '\n'
1338 				    && lastc == '.') {
1339 					/*
1340 					 * Don't print a bare "." line
1341 					 * since that will confuse ed(1).
1342 					 * Print ".." instead and set the,
1343 					 * global variable edoffset to an
1344 					 * offset from which to restart.
1345 					 * The caller must check the value
1346 					 * of edoffset
1347 					 */
1348 					printf(".\n");
1349 					edoffset = i - a + 1;
1350 					return edoffset;
1351 				}
1352 				/* when side-by-side, do not print a newline */
1353 				if (diff_format != D_SIDEBYSIDE || c != '\n') {
1354 					printf("%c", c);
1355 					col++;
1356 				}
1357 			}
1358 		}
1359 	}
1360 	return col;
1361 }
1362 
1363 /*
1364  * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578.
1365  */
1366 static enum readhash
1367 readhash(FILE *f, int flags, unsigned *hash)
1368 {
1369 	int i, t, space;
1370 	unsigned sum;
1371 
1372 	sum = 1;
1373 	space = 0;
1374 	for (i = 0;;) {
1375 		switch (t = getc(f)) {
1376 		case '\0':
1377 			if ((flags & D_FORCEASCII) == 0)
1378 				return (RH_BINARY);
1379 		case '\r':
1380 			if (flags & D_STRIPCR) {
1381 				t = getc(f);
1382 				if (t == '\n')
1383 					break;
1384 				ungetc(t, f);
1385 			}
1386 			/* FALLTHROUGH */
1387 		case '\t':
1388 		case '\v':
1389 		case '\f':
1390 		case ' ':
1391 			if ((flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) != 0) {
1392 				space++;
1393 				continue;
1394 			}
1395 			/* FALLTHROUGH */
1396 		default:
1397 			if (space && (flags & D_IGNOREBLANKS) == 0) {
1398 				i++;
1399 				space = 0;
1400 			}
1401 			sum = sum * 127 + chrtran(t);
1402 			i++;
1403 			continue;
1404 		case EOF:
1405 			if (i == 0)
1406 				return (RH_EOF);
1407 			/* FALLTHROUGH */
1408 		case '\n':
1409 			break;
1410 		}
1411 		break;
1412 	}
1413 	*hash = sum;
1414 	return (RH_OK);
1415 }
1416 
1417 static int
1418 asciifile(FILE *f)
1419 {
1420 	unsigned char buf[BUFSIZ];
1421 	size_t cnt;
1422 
1423 	if (f == NULL)
1424 		return (1);
1425 
1426 	rewind(f);
1427 	cnt = fread(buf, 1, sizeof(buf), f);
1428 	return (memchr(buf, '\0', cnt) == NULL);
1429 }
1430 
1431 #define begins_with(s, pre) (strncmp(s, pre, sizeof(pre)-1) == 0)
1432 
1433 static char *
1434 match_function(const long *f, int pos, FILE *fp)
1435 {
1436 	unsigned char buf[FUNCTION_CONTEXT_SIZE];
1437 	size_t nc;
1438 	int last = lastline;
1439 	const char *state = NULL;
1440 
1441 	lastline = pos;
1442 	while (pos > last) {
1443 		fseek(fp, f[pos - 1], SEEK_SET);
1444 		nc = f[pos] - f[pos - 1];
1445 		if (nc >= sizeof(buf))
1446 			nc = sizeof(buf) - 1;
1447 		nc = fread(buf, 1, nc, fp);
1448 		if (nc > 0) {
1449 			buf[nc] = '\0';
1450 			buf[strcspn(buf, "\n")] = '\0';
1451 			if (isalpha(buf[0]) || buf[0] == '_' || buf[0] == '$') {
1452 				if (begins_with(buf, "private:")) {
1453 					if (!state)
1454 						state = " (private)";
1455 				} else if (begins_with(buf, "protected:")) {
1456 					if (!state)
1457 						state = " (protected)";
1458 				} else if (begins_with(buf, "public:")) {
1459 					if (!state)
1460 						state = " (public)";
1461 				} else {
1462 					strlcpy(lastbuf, buf, sizeof lastbuf);
1463 					if (state)
1464 						strlcat(lastbuf, state,
1465 						    sizeof lastbuf);
1466 					lastmatchline = pos;
1467 					return lastbuf;
1468 				}
1469 			}
1470 		}
1471 		pos--;
1472 	}
1473 	return lastmatchline > 0 ? lastbuf : NULL;
1474 }
1475 
1476 /* dump accumulated "context" diff changes */
1477 static void
1478 dump_context_vec(FILE *f1, FILE *f2, int flags)
1479 {
1480 	struct context_vec *cvp = context_vec_start;
1481 	int lowa, upb, lowc, upd, do_output;
1482 	int a, b, c, d;
1483 	char ch, *f;
1484 
1485 	if (context_vec_start > context_vec_ptr)
1486 		return;
1487 
1488 	b = d = 0;		/* gcc */
1489 	lowa = MAX(1, cvp->a - diff_context);
1490 	upb = MIN(len[0], context_vec_ptr->b + diff_context);
1491 	lowc = MAX(1, cvp->c - diff_context);
1492 	upd = MIN(len[1], context_vec_ptr->d + diff_context);
1493 
1494 	printf("***************");
1495 	if ((flags & D_PROTOTYPE)) {
1496 		f = match_function(ixold, lowa-1, f1);
1497 		if (f != NULL)
1498 			printf(" %s", f);
1499 	}
1500 	printf("\n*** ");
1501 	range(lowa, upb, ",");
1502 	printf(" ****\n");
1503 
1504 	/*
1505 	 * Output changes to the "old" file.  The first loop suppresses
1506 	 * output if there were no changes to the "old" file (we'll see
1507 	 * the "old" lines as context in the "new" list).
1508 	 */
1509 	do_output = 0;
1510 	for (; cvp <= context_vec_ptr; cvp++)
1511 		if (cvp->a <= cvp->b) {
1512 			cvp = context_vec_start;
1513 			do_output++;
1514 			break;
1515 		}
1516 	if (do_output) {
1517 		while (cvp <= context_vec_ptr) {
1518 			a = cvp->a;
1519 			b = cvp->b;
1520 			c = cvp->c;
1521 			d = cvp->d;
1522 
1523 			if (a <= b && c <= d)
1524 				ch = 'c';
1525 			else
1526 				ch = (a <= b) ? 'd' : 'a';
1527 
1528 			if (ch == 'a')
1529 				fetch(ixold, lowa, b, f1, ' ', 0, flags);
1530 			else {
1531 				fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1532 				fetch(ixold, a, b, f1,
1533 				    ch == 'c' ? '!' : '-', 0, flags);
1534 			}
1535 			lowa = b + 1;
1536 			cvp++;
1537 		}
1538 		fetch(ixold, b + 1, upb, f1, ' ', 0, flags);
1539 	}
1540 	/* output changes to the "new" file */
1541 	printf("--- ");
1542 	range(lowc, upd, ",");
1543 	printf(" ----\n");
1544 
1545 	do_output = 0;
1546 	for (cvp = context_vec_start; cvp <= context_vec_ptr; cvp++)
1547 		if (cvp->c <= cvp->d) {
1548 			cvp = context_vec_start;
1549 			do_output++;
1550 			break;
1551 		}
1552 	if (do_output) {
1553 		while (cvp <= context_vec_ptr) {
1554 			a = cvp->a;
1555 			b = cvp->b;
1556 			c = cvp->c;
1557 			d = cvp->d;
1558 
1559 			if (a <= b && c <= d)
1560 				ch = 'c';
1561 			else
1562 				ch = (a <= b) ? 'd' : 'a';
1563 
1564 			if (ch == 'd')
1565 				fetch(ixnew, lowc, d, f2, ' ', 0, flags);
1566 			else {
1567 				fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1568 				fetch(ixnew, c, d, f2,
1569 				    ch == 'c' ? '!' : '+', 0, flags);
1570 			}
1571 			lowc = d + 1;
1572 			cvp++;
1573 		}
1574 		fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1575 	}
1576 	context_vec_ptr = context_vec_start - 1;
1577 }
1578 
1579 /* dump accumulated "unified" diff changes */
1580 static void
1581 dump_unified_vec(FILE *f1, FILE *f2, int flags)
1582 {
1583 	struct context_vec *cvp = context_vec_start;
1584 	int lowa, upb, lowc, upd;
1585 	int a, b, c, d;
1586 	char ch, *f;
1587 
1588 	if (context_vec_start > context_vec_ptr)
1589 		return;
1590 
1591 	b = d = 0;		/* gcc */
1592 	lowa = MAX(1, cvp->a - diff_context);
1593 	upb = MIN(len[0], context_vec_ptr->b + diff_context);
1594 	lowc = MAX(1, cvp->c - diff_context);
1595 	upd = MIN(len[1], context_vec_ptr->d + diff_context);
1596 
1597 	printf("@@ -");
1598 	uni_range(lowa, upb);
1599 	printf(" +");
1600 	uni_range(lowc, upd);
1601 	printf(" @@");
1602 	if ((flags & D_PROTOTYPE)) {
1603 		f = match_function(ixold, lowa-1, f1);
1604 		if (f != NULL)
1605 			printf(" %s", f);
1606 	}
1607 	printf("\n");
1608 
1609 	/*
1610 	 * Output changes in "unified" diff format--the old and new lines
1611 	 * are printed together.
1612 	 */
1613 	for (; cvp <= context_vec_ptr; cvp++) {
1614 		a = cvp->a;
1615 		b = cvp->b;
1616 		c = cvp->c;
1617 		d = cvp->d;
1618 
1619 		/*
1620 		 * c: both new and old changes
1621 		 * d: only changes in the old file
1622 		 * a: only changes in the new file
1623 		 */
1624 		if (a <= b && c <= d)
1625 			ch = 'c';
1626 		else
1627 			ch = (a <= b) ? 'd' : 'a';
1628 
1629 		switch (ch) {
1630 		case 'c':
1631 			fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1632 			fetch(ixold, a, b, f1, '-', 0, flags);
1633 			fetch(ixnew, c, d, f2, '+', 0, flags);
1634 			break;
1635 		case 'd':
1636 			fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1637 			fetch(ixold, a, b, f1, '-', 0, flags);
1638 			break;
1639 		case 'a':
1640 			fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1641 			fetch(ixnew, c, d, f2, '+', 0, flags);
1642 			break;
1643 		}
1644 		lowa = b + 1;
1645 		lowc = d + 1;
1646 	}
1647 	fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1648 
1649 	context_vec_ptr = context_vec_start - 1;
1650 }
1651 
1652 static void
1653 print_header(const char *file1, const char *file2)
1654 {
1655 	const char *time_format;
1656 	char buf1[256];
1657 	char buf2[256];
1658 	char end1[10];
1659 	char end2[10];
1660 	struct tm tm1, tm2, *tm_ptr1, *tm_ptr2;
1661 	int nsec1 = stb1.st_mtim.tv_nsec;
1662 	int nsec2 = stb2.st_mtim.tv_nsec;
1663 
1664 	time_format = "%Y-%m-%d %H:%M:%S";
1665 
1666 	if (cflag)
1667 		time_format = "%c";
1668 	tm_ptr1 = localtime_r(&stb1.st_mtime, &tm1);
1669 	tm_ptr2 = localtime_r(&stb2.st_mtime, &tm2);
1670 	strftime(buf1, 256, time_format, tm_ptr1);
1671 	strftime(buf2, 256, time_format, tm_ptr2);
1672 	if (!cflag) {
1673 		strftime(end1, 10, "%z", tm_ptr1);
1674 		strftime(end2, 10, "%z", tm_ptr2);
1675 		sprintf(buf1, "%s.%.9d %s", buf1, nsec1, end1);
1676 		sprintf(buf2, "%s.%.9d %s", buf2, nsec2, end2);
1677 	}
1678 	if (label[0] != NULL)
1679 		printf("%s %s\n", diff_format == D_CONTEXT ? "***" : "---",
1680 		    label[0]);
1681 	else
1682 		printf("%s %s\t%s\n", diff_format == D_CONTEXT ? "***" : "---",
1683 		    file1, buf1);
1684 	if (label[1] != NULL)
1685 		printf("%s %s\n", diff_format == D_CONTEXT ? "---" : "+++",
1686 		    label[1]);
1687 	else
1688 		printf("%s %s\t%s\n", diff_format == D_CONTEXT ? "---" : "+++",
1689 		    file2, buf2);
1690 }
1691 
1692 /*
1693  * Prints n number of space characters either by using tab
1694  * or single space characters.
1695  * nc is the preceding number of characters
1696  */
1697 static void
1698 print_space(int nc, int n, int flags) {
1699 	int i, col;
1700 
1701 	col = n;
1702 	if ((flags & D_EXPANDTABS) == 0) {
1703 		/* first tabstop may be closer than tabsize */
1704 		i = tabsize - (nc % tabsize);
1705 		while (col >= tabsize) {
1706 			printf("\t");
1707 			col -= i;
1708 			i = tabsize;
1709 		}
1710 	}
1711 	printf("%*s", col, "");
1712 }
1713