xref: /freebsd/usr.bin/diff/diffreg.c (revision 19cca0b9613d7c3058e41baf0204245119732235)
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 #define	diff_output	printf
184 #define MIN_PAD		1
185 static FILE	*opentemp(const char *);
186 static void	 output(char *, FILE *, char *, FILE *, int);
187 static void	 check(FILE *, FILE *, int);
188 static void	 range(int, int, const char *);
189 static void	 uni_range(int, int);
190 static void	 dump_context_vec(FILE *, FILE *, int);
191 static void	 dump_unified_vec(FILE *, FILE *, int);
192 static void	 prepare(int, FILE *, size_t, int);
193 static void	 prune(void);
194 static void	 equiv(struct line *, int, struct line *, int, int *);
195 static void	 unravel(int);
196 static void	 unsort(struct line *, int, int *);
197 static void	 change(char *, FILE *, char *, FILE *, int, int, int, int, int *);
198 static void	 sort(struct line *, int);
199 static void	 print_header(const char *, const char *);
200 static void	 print_space(int, int, int);
201 static bool	 ignoreline_pattern(char *);
202 static bool	 ignoreline(char *, bool);
203 static int	 asciifile(FILE *);
204 static int	 fetch(long *, int, int, FILE *, int, int, int);
205 static int	 newcand(int, int, int);
206 static int	 search(int *, int, int);
207 static int	 skipline(FILE *);
208 static int	 isqrt(int);
209 static int	 stone(int *, int, int *, int *, int);
210 static int	 readhash(FILE *, int);
211 static int	 files_differ(FILE *, FILE *, int);
212 static char	*match_function(const long *, int, FILE *);
213 static char	*preadline(int, size_t, off_t);
214 
215 static int  *J;			/* will be overlaid on class */
216 static int  *class;		/* will be overlaid on file[0] */
217 static int  *klist;		/* will be overlaid on file[0] after class */
218 static int  *member;		/* will be overlaid on file[1] */
219 static int   clen;
220 static int   inifdef;		/* whether or not we are in a #ifdef block */
221 static int   len[2];
222 static int   pref, suff;	/* length of prefix and suffix */
223 static int   slen[2];
224 static int   anychange;
225 static int   hw, padding;	/* half width and padding */
226 static int   edoffset;
227 static long *ixnew;		/* will be overlaid on file[1] */
228 static long *ixold;		/* will be overlaid on klist */
229 static struct cand *clist;	/* merely a free storage pot for candidates */
230 static int   clistlen;		/* the length of clist */
231 static struct line *sfile[2];	/* shortened by pruning common prefix/suffix */
232 static int (*chrtran)(int);	/* translation table for case-folding */
233 static struct context_vec *context_vec_start;
234 static struct context_vec *context_vec_end;
235 static struct context_vec *context_vec_ptr;
236 
237 #define FUNCTION_CONTEXT_SIZE	55
238 static char lastbuf[FUNCTION_CONTEXT_SIZE];
239 static int lastline;
240 static int lastmatchline;
241 
242 static int
243 clow2low(int c)
244 {
245 
246 	return (c);
247 }
248 
249 static int
250 cup2low(int c)
251 {
252 
253 	return tolower(c);
254 }
255 
256 int
257 diffreg(char *file1, char *file2, int flags, int capsicum)
258 {
259 	FILE *f1, *f2;
260 	int i, rval;
261 	struct pr *pr = NULL;
262 	cap_rights_t rights_ro;
263 
264 	f1 = f2 = NULL;
265 	rval = D_SAME;
266 	anychange = 0;
267 	lastline = 0;
268 	lastmatchline = 0;
269 	context_vec_ptr = context_vec_start - 1;
270 
271 	 /*
272 	  * hw excludes padding and make sure when -t is not used,
273 	  * the second column always starts from the closest tab stop
274 	  */
275 	if (diff_format == D_SIDEBYSIDE) {
276 		hw = width >> 1;
277 		padding = tabsize - (hw % tabsize);
278 		if ((flags & D_EXPANDTABS) != 0 || (padding % tabsize == 0))
279 			padding = MIN_PAD;
280 
281 		hw = (width >> 1) -
282 		    ((padding == MIN_PAD) ? (padding << 1) : padding) - 1;
283 	}
284 
285 
286 	if (flags & D_IGNORECASE)
287 		chrtran = cup2low;
288 	else
289 		chrtran = clow2low;
290 	if (S_ISDIR(stb1.st_mode) != S_ISDIR(stb2.st_mode))
291 		return (S_ISDIR(stb1.st_mode) ? D_MISMATCH1 : D_MISMATCH2);
292 	if (strcmp(file1, "-") == 0 && strcmp(file2, "-") == 0)
293 		goto closem;
294 
295 	if (flags & D_EMPTY1)
296 		f1 = fopen(_PATH_DEVNULL, "r");
297 	else {
298 		if (!S_ISREG(stb1.st_mode)) {
299 			if ((f1 = opentemp(file1)) == NULL ||
300 			    fstat(fileno(f1), &stb1) == -1) {
301 				warn("%s", file1);
302 				status |= 2;
303 				goto closem;
304 			}
305 		} else if (strcmp(file1, "-") == 0)
306 			f1 = stdin;
307 		else
308 			f1 = fopen(file1, "r");
309 	}
310 	if (f1 == NULL) {
311 		warn("%s", file1);
312 		status |= 2;
313 		goto closem;
314 	}
315 
316 	if (flags & D_EMPTY2)
317 		f2 = fopen(_PATH_DEVNULL, "r");
318 	else {
319 		if (!S_ISREG(stb2.st_mode)) {
320 			if ((f2 = opentemp(file2)) == NULL ||
321 			    fstat(fileno(f2), &stb2) == -1) {
322 				warn("%s", file2);
323 				status |= 2;
324 				goto closem;
325 			}
326 		} else if (strcmp(file2, "-") == 0)
327 			f2 = stdin;
328 		else
329 			f2 = fopen(file2, "r");
330 	}
331 	if (f2 == NULL) {
332 		warn("%s", file2);
333 		status |= 2;
334 		goto closem;
335 	}
336 
337 	if (lflag)
338 		pr = start_pr(file1, file2);
339 
340 	if (capsicum) {
341 		cap_rights_init(&rights_ro, CAP_READ, CAP_FSTAT, CAP_SEEK);
342 		if (caph_rights_limit(fileno(f1), &rights_ro) < 0)
343 			err(2, "unable to limit rights on: %s", file1);
344 		if (caph_rights_limit(fileno(f2), &rights_ro) < 0)
345 			err(2, "unable to limit rights on: %s", file2);
346 		if (fileno(f1) == STDIN_FILENO || fileno(f2) == STDIN_FILENO) {
347 			/* stding has already been limited */
348 			if (caph_limit_stderr() == -1)
349 				err(2, "unable to limit stderr");
350 			if (caph_limit_stdout() == -1)
351 				err(2, "unable to limit stdout");
352 		} else if (caph_limit_stdio() == -1)
353 				err(2, "unable to limit stdio");
354 
355 		caph_cache_catpages();
356 		caph_cache_tzdata();
357 		if (caph_enter() < 0)
358 			err(2, "unable to enter capability mode");
359 	}
360 
361 	switch (files_differ(f1, f2, flags)) {
362 	case 0:
363 		goto closem;
364 	case 1:
365 		break;
366 	default:
367 		/* error */
368 		status |= 2;
369 		goto closem;
370 	}
371 
372 	if (diff_format == D_BRIEF && ignore_pats == NULL &&
373 	    (flags & (D_FOLDBLANKS|D_IGNOREBLANKS|D_IGNORECASE|D_STRIPCR)) == 0)
374 	{
375 		rval = D_DIFFER;
376 		status |= 1;
377 		goto closem;
378 	}
379 	if ((flags & D_FORCEASCII) == 0 &&
380 	    (!asciifile(f1) || !asciifile(f2))) {
381 		rval = D_BINARY;
382 		status |= 1;
383 		goto closem;
384 	}
385 	prepare(0, f1, stb1.st_size, flags);
386 	prepare(1, f2, stb2.st_size, flags);
387 
388 	prune();
389 	sort(sfile[0], slen[0]);
390 	sort(sfile[1], slen[1]);
391 
392 	member = (int *)file[1];
393 	equiv(sfile[0], slen[0], sfile[1], slen[1], member);
394 	member = xreallocarray(member, slen[1] + 2, sizeof(*member));
395 
396 	class = (int *)file[0];
397 	unsort(sfile[0], slen[0], class);
398 	class = xreallocarray(class, slen[0] + 2, sizeof(*class));
399 
400 	klist = xcalloc(slen[0] + 2, sizeof(*klist));
401 	clen = 0;
402 	clistlen = 100;
403 	clist = xcalloc(clistlen, sizeof(*clist));
404 	i = stone(class, slen[0], member, klist, flags);
405 	free(member);
406 	free(class);
407 
408 	J = xreallocarray(J, len[0] + 2, sizeof(*J));
409 	unravel(klist[i]);
410 	free(clist);
411 	free(klist);
412 
413 	ixold = xreallocarray(ixold, len[0] + 2, sizeof(*ixold));
414 	ixnew = xreallocarray(ixnew, len[1] + 2, sizeof(*ixnew));
415 	check(f1, f2, flags);
416 	output(file1, f1, file2, f2, flags);
417 
418 closem:
419 	if (pr != NULL)
420 		stop_pr(pr);
421 	if (anychange) {
422 		status |= 1;
423 		if (rval == D_SAME)
424 			rval = D_DIFFER;
425 	}
426 	if (f1 != NULL)
427 		fclose(f1);
428 	if (f2 != NULL)
429 		fclose(f2);
430 
431 	return (rval);
432 }
433 
434 /*
435  * Check to see if the given files differ.
436  * Returns 0 if they are the same, 1 if different, and -1 on error.
437  * XXX - could use code from cmp(1) [faster]
438  */
439 static int
440 files_differ(FILE *f1, FILE *f2, int flags)
441 {
442 	char buf1[BUFSIZ], buf2[BUFSIZ];
443 	size_t i, j;
444 
445 	if ((flags & (D_EMPTY1|D_EMPTY2)) || stb1.st_size != stb2.st_size ||
446 	    (stb1.st_mode & S_IFMT) != (stb2.st_mode & S_IFMT))
447 		return (1);
448 	for (;;) {
449 		i = fread(buf1, 1, sizeof(buf1), f1);
450 		j = fread(buf2, 1, sizeof(buf2), f2);
451 		if ((!i && ferror(f1)) || (!j && ferror(f2)))
452 			return (-1);
453 		if (i != j)
454 			return (1);
455 		if (i == 0)
456 			return (0);
457 		if (memcmp(buf1, buf2, i) != 0)
458 			return (1);
459 	}
460 }
461 
462 static FILE *
463 opentemp(const char *f)
464 {
465 	char buf[BUFSIZ], tempfile[PATH_MAX];
466 	ssize_t nread;
467 	int ifd, ofd;
468 
469 	if (strcmp(f, "-") == 0)
470 		ifd = STDIN_FILENO;
471 	else if ((ifd = open(f, O_RDONLY, 0644)) == -1)
472 		return (NULL);
473 
474 	(void)strlcpy(tempfile, _PATH_TMP "/diff.XXXXXXXX", sizeof(tempfile));
475 
476 	if ((ofd = mkstemp(tempfile)) == -1) {
477 		close(ifd);
478 		return (NULL);
479 	}
480 	unlink(tempfile);
481 	while ((nread = read(ifd, buf, BUFSIZ)) > 0) {
482 		if (write(ofd, buf, nread) != nread) {
483 			close(ifd);
484 			close(ofd);
485 			return (NULL);
486 		}
487 	}
488 	close(ifd);
489 	lseek(ofd, (off_t)0, SEEK_SET);
490 	return (fdopen(ofd, "r"));
491 }
492 
493 char *
494 splice(char *dir, char *path)
495 {
496 	char *tail, *buf;
497 	size_t dirlen;
498 
499 	dirlen = strlen(dir);
500 	while (dirlen != 0 && dir[dirlen - 1] == '/')
501 	    dirlen--;
502 	if ((tail = strrchr(path, '/')) == NULL)
503 		tail = path;
504 	else
505 		tail++;
506 	xasprintf(&buf, "%.*s/%s", (int)dirlen, dir, tail);
507 	return (buf);
508 }
509 
510 static void
511 prepare(int i, FILE *fd, size_t filesize, int flags)
512 {
513 	struct line *p;
514 	int h;
515 	size_t sz, j;
516 
517 	rewind(fd);
518 
519 	sz = MIN(filesize, SIZE_MAX) / 25;
520 	if (sz < 100)
521 		sz = 100;
522 
523 	p = xcalloc(sz + 3, sizeof(*p));
524 	for (j = 0; (h = readhash(fd, flags));) {
525 		if (j == sz) {
526 			sz = sz * 3 / 2;
527 			p = xreallocarray(p, sz + 3, sizeof(*p));
528 		}
529 		p[++j].value = h;
530 	}
531 	len[i] = j;
532 	file[i] = p;
533 }
534 
535 static void
536 prune(void)
537 {
538 	int i, j;
539 
540 	for (pref = 0; pref < len[0] && pref < len[1] &&
541 	    file[0][pref + 1].value == file[1][pref + 1].value;
542 	    pref++)
543 		;
544 	for (suff = 0; suff < len[0] - pref && suff < len[1] - pref &&
545 	    file[0][len[0] - suff].value == file[1][len[1] - suff].value;
546 	    suff++)
547 		;
548 	for (j = 0; j < 2; j++) {
549 		sfile[j] = file[j] + pref;
550 		slen[j] = len[j] - pref - suff;
551 		for (i = 0; i <= slen[j]; i++)
552 			sfile[j][i].serial = i;
553 	}
554 }
555 
556 static void
557 equiv(struct line *a, int n, struct line *b, int m, int *c)
558 {
559 	int i, j;
560 
561 	i = j = 1;
562 	while (i <= n && j <= m) {
563 		if (a[i].value < b[j].value)
564 			a[i++].value = 0;
565 		else if (a[i].value == b[j].value)
566 			a[i++].value = j;
567 		else
568 			j++;
569 	}
570 	while (i <= n)
571 		a[i++].value = 0;
572 	b[m + 1].value = 0;
573 	j = 0;
574 	while (++j <= m) {
575 		c[j] = -b[j].serial;
576 		while (b[j + 1].value == b[j].value) {
577 			j++;
578 			c[j] = b[j].serial;
579 		}
580 	}
581 	c[j] = -1;
582 }
583 
584 /* Code taken from ping.c */
585 static int
586 isqrt(int n)
587 {
588 	int y, x = 1;
589 
590 	if (n == 0)
591 		return (0);
592 
593 	do { /* newton was a stinker */
594 		y = x;
595 		x = n / x;
596 		x += y;
597 		x /= 2;
598 	} while ((x - y) > 1 || (x - y) < -1);
599 
600 	return (x);
601 }
602 
603 static int
604 stone(int *a, int n, int *b, int *c, int flags)
605 {
606 	int i, k, y, j, l;
607 	int oldc, tc, oldl, sq;
608 	u_int numtries, bound;
609 
610 	if (flags & D_MINIMAL)
611 		bound = UINT_MAX;
612 	else {
613 		sq = isqrt(n);
614 		bound = MAX(256, sq);
615 	}
616 
617 	k = 0;
618 	c[0] = newcand(0, 0, 0);
619 	for (i = 1; i <= n; i++) {
620 		j = a[i];
621 		if (j == 0)
622 			continue;
623 		y = -b[j];
624 		oldl = 0;
625 		oldc = c[0];
626 		numtries = 0;
627 		do {
628 			if (y <= clist[oldc].y)
629 				continue;
630 			l = search(c, k, y);
631 			if (l != oldl + 1)
632 				oldc = c[l - 1];
633 			if (l <= k) {
634 				if (clist[c[l]].y <= y)
635 					continue;
636 				tc = c[l];
637 				c[l] = newcand(i, y, oldc);
638 				oldc = tc;
639 				oldl = l;
640 				numtries++;
641 			} else {
642 				c[l] = newcand(i, y, oldc);
643 				k++;
644 				break;
645 			}
646 		} while ((y = b[++j]) > 0 && numtries < bound);
647 	}
648 	return (k);
649 }
650 
651 static int
652 newcand(int x, int y, int pred)
653 {
654 	struct cand *q;
655 
656 	if (clen == clistlen) {
657 		clistlen = clistlen * 11 / 10;
658 		clist = xreallocarray(clist, clistlen, sizeof(*clist));
659 	}
660 	q = clist + clen;
661 	q->x = x;
662 	q->y = y;
663 	q->pred = pred;
664 	return (clen++);
665 }
666 
667 static int
668 search(int *c, int k, int y)
669 {
670 	int i, j, l, t;
671 
672 	if (clist[c[k]].y < y)	/* quick look for typical case */
673 		return (k + 1);
674 	i = 0;
675 	j = k + 1;
676 	for (;;) {
677 		l = (i + j) / 2;
678 		if (l <= i)
679 			break;
680 		t = clist[c[l]].y;
681 		if (t > y)
682 			j = l;
683 		else if (t < y)
684 			i = l;
685 		else
686 			return (l);
687 	}
688 	return (l + 1);
689 }
690 
691 static void
692 unravel(int p)
693 {
694 	struct cand *q;
695 	int i;
696 
697 	for (i = 0; i <= len[0]; i++)
698 		J[i] = i <= pref ? i :
699 		    i > len[0] - suff ? i + len[1] - len[0] : 0;
700 	for (q = clist + p; q->y != 0; q = clist + q->pred)
701 		J[q->x + pref] = q->y + pref;
702 }
703 
704 /*
705  * Check does double duty:
706  *  1.	ferret out any fortuitous correspondences due
707  *	to confounding by hashing (which result in "jackpot")
708  *  2.  collect random access indexes to the two files
709  */
710 static void
711 check(FILE *f1, FILE *f2, int flags)
712 {
713 	int i, j, jackpot, c, d;
714 	long ctold, ctnew;
715 
716 	rewind(f1);
717 	rewind(f2);
718 	j = 1;
719 	ixold[0] = ixnew[0] = 0;
720 	jackpot = 0;
721 	ctold = ctnew = 0;
722 	for (i = 1; i <= len[0]; i++) {
723 		if (J[i] == 0) {
724 			ixold[i] = ctold += skipline(f1);
725 			continue;
726 		}
727 		while (j < J[i]) {
728 			ixnew[j] = ctnew += skipline(f2);
729 			j++;
730 		}
731 		if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS|D_IGNORECASE|D_STRIPCR)) {
732 			for (;;) {
733 				c = getc(f1);
734 				d = getc(f2);
735 				/*
736 				 * GNU diff ignores a missing newline
737 				 * in one file for -b or -w.
738 				 */
739 				if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) {
740 					if (c == EOF && d == '\n') {
741 						ctnew++;
742 						break;
743 					} else if (c == '\n' && d == EOF) {
744 						ctold++;
745 						break;
746 					}
747 				}
748 				ctold++;
749 				ctnew++;
750 				if (flags & D_STRIPCR && (c == '\r' || d == '\r')) {
751 					if (c == '\r') {
752 						if ((c = getc(f1)) == '\n') {
753 							ctold++;
754 						} else {
755 							ungetc(c, f1);
756 						}
757 					}
758 					if (d == '\r') {
759 						if ((d = getc(f2)) == '\n') {
760 							ctnew++;
761 						} else {
762 							ungetc(d, f2);
763 						}
764 					}
765 					break;
766 				}
767 				if ((flags & D_FOLDBLANKS) && isspace(c) &&
768 				    isspace(d)) {
769 					do {
770 						if (c == '\n')
771 							break;
772 						ctold++;
773 					} while (isspace(c = getc(f1)));
774 					do {
775 						if (d == '\n')
776 							break;
777 						ctnew++;
778 					} while (isspace(d = getc(f2)));
779 				} else if ((flags & D_IGNOREBLANKS)) {
780 					while (isspace(c) && c != '\n') {
781 						c = getc(f1);
782 						ctold++;
783 					}
784 					while (isspace(d) && d != '\n') {
785 						d = getc(f2);
786 						ctnew++;
787 					}
788 				}
789 				if (chrtran(c) != chrtran(d)) {
790 					jackpot++;
791 					J[i] = 0;
792 					if (c != '\n' && c != EOF)
793 						ctold += skipline(f1);
794 					if (d != '\n' && c != EOF)
795 						ctnew += skipline(f2);
796 					break;
797 				}
798 				if (c == '\n' || c == EOF)
799 					break;
800 			}
801 		} else {
802 			for (;;) {
803 				ctold++;
804 				ctnew++;
805 				if ((c = getc(f1)) != (d = getc(f2))) {
806 					/* jackpot++; */
807 					J[i] = 0;
808 					if (c != '\n' && c != EOF)
809 						ctold += skipline(f1);
810 					if (d != '\n' && c != EOF)
811 						ctnew += skipline(f2);
812 					break;
813 				}
814 				if (c == '\n' || c == EOF)
815 					break;
816 			}
817 		}
818 		ixold[i] = ctold;
819 		ixnew[j] = ctnew;
820 		j++;
821 	}
822 	for (; j <= len[1]; j++) {
823 		ixnew[j] = ctnew += skipline(f2);
824 	}
825 	/*
826 	 * if (jackpot)
827 	 *	fprintf(stderr, "jackpot\n");
828 	 */
829 }
830 
831 /* shellsort CACM #201 */
832 static void
833 sort(struct line *a, int n)
834 {
835 	struct line *ai, *aim, w;
836 	int j, m = 0, k;
837 
838 	if (n == 0)
839 		return;
840 	for (j = 1; j <= n; j *= 2)
841 		m = 2 * j - 1;
842 	for (m /= 2; m != 0; m /= 2) {
843 		k = n - m;
844 		for (j = 1; j <= k; j++) {
845 			for (ai = &a[j]; ai > a; ai -= m) {
846 				aim = &ai[m];
847 				if (aim < ai)
848 					break;	/* wraparound */
849 				if (aim->value > ai[0].value ||
850 				    (aim->value == ai[0].value &&
851 					aim->serial > ai[0].serial))
852 					break;
853 				w.value = ai[0].value;
854 				ai[0].value = aim->value;
855 				aim->value = w.value;
856 				w.serial = ai[0].serial;
857 				ai[0].serial = aim->serial;
858 				aim->serial = w.serial;
859 			}
860 		}
861 	}
862 }
863 
864 static void
865 unsort(struct line *f, int l, int *b)
866 {
867 	int *a, i;
868 
869 	a = xcalloc(l + 1, sizeof(*a));
870 	for (i = 1; i <= l; i++)
871 		a[f[i].serial] = f[i].value;
872 	for (i = 1; i <= l; i++)
873 		b[i] = a[i];
874 	free(a);
875 }
876 
877 static int
878 skipline(FILE *f)
879 {
880 	int i, c;
881 
882 	for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++)
883 		continue;
884 	return (i);
885 }
886 
887 static void
888 output(char *file1, FILE *f1, char *file2, FILE *f2, int flags)
889 {
890 	int i, j, m, i0, i1, j0, j1, nc;
891 
892 	rewind(f1);
893 	rewind(f2);
894 	m = len[0];
895 	J[0] = 0;
896 	J[m + 1] = len[1] + 1;
897 	if (diff_format != D_EDIT) {
898 		for (i0 = 1; i0 <= m; i0 = i1 + 1) {
899 			while (i0 <= m && J[i0] == J[i0 - 1] + 1){
900 				if (diff_format == D_SIDEBYSIDE &&
901 				    suppress_common != 1) {
902 					nc = fetch(ixold, i0, i0, f1, '\0',
903 					    1, flags);
904 					print_space(nc,
905 					    (hw - nc) + (padding << 1) + 1,
906 					    flags);
907 					fetch(ixnew, J[i0], J[i0], f2, '\0',
908 					    0, flags);
909 					diff_output("\n");
910 				}
911 				i0++;
912 			}
913 			j0 = J[i0 - 1] + 1;
914 			i1 = i0 - 1;
915 			while (i1 < m && J[i1 + 1] == 0)
916 				i1++;
917 			j1 = J[i1 + 1] - 1;
918 			J[i1] = j1;
919 
920 			/*
921 			 * When using side-by-side, lines from both of the
922 			 * files are printed. The algorithm used by diff(1)
923 			 * identifies the ranges in which two files differ.
924 			 * See the change() function below.
925 			 * The for loop below consumes the shorter range,
926 			 * whereas one of the while loops deals with the
927 			 * longer one.
928 			 */
929 			if (diff_format == D_SIDEBYSIDE) {
930 				for (i=i0, j=j0; i<=i1 && j<=j1; i++, j++)
931 					change(file1, f1, file2, f2, i, i,
932 					    j, j, &flags);
933 
934 				while (i <= i1) {
935 					change(file1, f1, file2, f2,
936 					    i, i, j+1, j, &flags);
937 					i++;
938 				}
939 
940 				while (j <= j1) {
941 					change(file1, f1, file2, f2,
942 					    i+1, i, j, j, &flags);
943 					j++;
944 				}
945 			} else
946 				change(file1, f1, file2, f2, i0, i1, j0,
947 				    j1, &flags);
948 		}
949 	} else {
950 		for (i0 = m; i0 >= 1; i0 = i1 - 1) {
951 			while (i0 >= 1 && J[i0] == J[i0 + 1] - 1 && J[i0] != 0)
952 				i0--;
953 			j0 = J[i0 + 1] - 1;
954 			i1 = i0 + 1;
955 			while (i1 > 1 && J[i1 - 1] == 0)
956 				i1--;
957 			j1 = J[i1 - 1] + 1;
958 			J[i1] = j1;
959 			change(file1, f1, file2, f2, i1, i0, j1, j0, &flags);
960 		}
961 	}
962 	if (m == 0)
963 		change(file1, f1, file2, f2, 1, 0, 1, len[1], &flags);
964 	if (diff_format == D_IFDEF || diff_format == D_GFORMAT) {
965 		for (;;) {
966 #define	c i0
967 			if ((c = getc(f1)) == EOF)
968 				return;
969 			diff_output("%c", c);
970 		}
971 #undef c
972 	}
973 	if (anychange != 0) {
974 		if (diff_format == D_CONTEXT)
975 			dump_context_vec(f1, f2, flags);
976 		else if (diff_format == D_UNIFIED)
977 			dump_unified_vec(f1, f2, flags);
978 	}
979 }
980 
981 static void
982 range(int a, int b, const char *separator)
983 {
984 	diff_output("%d", a > b ? b : a);
985 	if (a < b)
986 		diff_output("%s%d", separator, b);
987 }
988 
989 static void
990 uni_range(int a, int b)
991 {
992 	if (a < b)
993 		diff_output("%d,%d", a, b - a + 1);
994 	else if (a == b)
995 		diff_output("%d", b);
996 	else
997 		diff_output("%d,0", b);
998 }
999 
1000 static char *
1001 preadline(int fd, size_t rlen, off_t off)
1002 {
1003 	char *line;
1004 	ssize_t nr;
1005 
1006 	line = xmalloc(rlen + 1);
1007 	if ((nr = pread(fd, line, rlen, off)) == -1)
1008 		err(2, "preadline");
1009 	if (nr > 0 && line[nr-1] == '\n')
1010 		nr--;
1011 	line[nr] = '\0';
1012 	return (line);
1013 }
1014 
1015 static bool
1016 ignoreline_pattern(char *line)
1017 {
1018 	int ret;
1019 
1020 	ret = regexec(&ignore_re, line, 0, NULL, 0);
1021 	free(line);
1022 	return (ret == 0);	/* if it matched, it should be ignored. */
1023 }
1024 
1025 static bool
1026 ignoreline(char *line, bool skip_blanks)
1027 {
1028 
1029 	if (ignore_pats != NULL && skip_blanks)
1030 		return (ignoreline_pattern(line) || *line == '\0');
1031 	if (ignore_pats != NULL)
1032 		return (ignoreline_pattern(line));
1033 	if (skip_blanks)
1034 		return (*line == '\0');
1035 	/* No ignore criteria specified */
1036 	return (false);
1037 }
1038 
1039 /*
1040  * Indicate that there is a difference between lines a and b of the from file
1041  * to get to lines c to d of the to file.  If a is greater then b then there
1042  * are no lines in the from file involved and this means that there were
1043  * lines appended (beginning at b).  If c is greater than d then there are
1044  * lines missing from the to file.
1045  */
1046 static void
1047 change(char *file1, FILE *f1, char *file2, FILE *f2, int a, int b, int c, int d,
1048     int *pflags)
1049 {
1050 	static size_t max_context = 64;
1051 	long curpos;
1052 	int i, nc;
1053 	const char *walk;
1054 	bool skip_blanks;
1055 
1056 	skip_blanks = (*pflags & D_SKIPBLANKLINES);
1057 restart:
1058 	if ((diff_format != D_IFDEF || diff_format == D_GFORMAT) &&
1059 	    a > b && c > d)
1060 		return;
1061 	if (ignore_pats != NULL || skip_blanks) {
1062 		char *line;
1063 		/*
1064 		 * All lines in the change, insert, or delete must
1065 		 * match an ignore pattern for the change to be
1066 		 * ignored.
1067 		 */
1068 		if (a <= b) {		/* Changes and deletes. */
1069 			for (i = a; i <= b; i++) {
1070 				line = preadline(fileno(f1),
1071 				    ixold[i] - ixold[i - 1], ixold[i - 1]);
1072 				if (!ignoreline(line, skip_blanks))
1073 					goto proceed;
1074 			}
1075 		}
1076 		if (a > b || c <= d) {	/* Changes and inserts. */
1077 			for (i = c; i <= d; i++) {
1078 				line = preadline(fileno(f2),
1079 				    ixnew[i] - ixnew[i - 1], ixnew[i - 1]);
1080 				if (!ignoreline(line, skip_blanks))
1081 					goto proceed;
1082 			}
1083 		}
1084 		return;
1085 	}
1086 proceed:
1087 	if (*pflags & D_HEADER && diff_format != D_BRIEF) {
1088 		diff_output("%s %s %s\n", diffargs, file1, file2);
1089 		*pflags &= ~D_HEADER;
1090 	}
1091 	if (diff_format == D_CONTEXT || diff_format == D_UNIFIED) {
1092 		/*
1093 		 * Allocate change records as needed.
1094 		 */
1095 		if (context_vec_ptr == context_vec_end - 1) {
1096 			ptrdiff_t offset = context_vec_ptr - context_vec_start;
1097 			max_context <<= 1;
1098 			context_vec_start = xreallocarray(context_vec_start,
1099 			    max_context, sizeof(*context_vec_start));
1100 			context_vec_end = context_vec_start + max_context;
1101 			context_vec_ptr = context_vec_start + offset;
1102 		}
1103 		if (anychange == 0) {
1104 			/*
1105 			 * Print the context/unidiff header first time through.
1106 			 */
1107 			print_header(file1, file2);
1108 			anychange = 1;
1109 		} else if (a > context_vec_ptr->b + (2 * diff_context) + 1 &&
1110 		    c > context_vec_ptr->d + (2 * diff_context) + 1) {
1111 			/*
1112 			 * If this change is more than 'diff_context' lines from the
1113 			 * previous change, dump the record and reset it.
1114 			 */
1115 			if (diff_format == D_CONTEXT)
1116 				dump_context_vec(f1, f2, *pflags);
1117 			else
1118 				dump_unified_vec(f1, f2, *pflags);
1119 		}
1120 		context_vec_ptr++;
1121 		context_vec_ptr->a = a;
1122 		context_vec_ptr->b = b;
1123 		context_vec_ptr->c = c;
1124 		context_vec_ptr->d = d;
1125 		return;
1126 	}
1127 	if (anychange == 0)
1128 		anychange = 1;
1129 	switch (diff_format) {
1130 	case D_BRIEF:
1131 		return;
1132 	case D_NORMAL:
1133 	case D_EDIT:
1134 		range(a, b, ",");
1135 		diff_output("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1136 		if (diff_format == D_NORMAL)
1137 			range(c, d, ",");
1138 		diff_output("\n");
1139 		break;
1140 	case D_REVERSE:
1141 		diff_output("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1142 		range(a, b, " ");
1143 		diff_output("\n");
1144 		break;
1145 	case D_NREVERSE:
1146 		if (a > b)
1147 			diff_output("a%d %d\n", b, d - c + 1);
1148 		else {
1149 			diff_output("d%d %d\n", a, b - a + 1);
1150 			if (!(c > d))
1151 				/* add changed lines */
1152 				diff_output("a%d %d\n", b, d - c + 1);
1153 		}
1154 		break;
1155 	}
1156 	if (diff_format == D_GFORMAT) {
1157 		curpos = ftell(f1);
1158 		/* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1159 		nc = ixold[a > b ? b : a - 1] - curpos;
1160 		for (i = 0; i < nc; i++)
1161 			diff_output("%c", getc(f1));
1162 		for (walk = group_format; *walk != '\0'; walk++) {
1163 			if (*walk == '%') {
1164 				walk++;
1165 				switch (*walk) {
1166 				case '<':
1167 					fetch(ixold, a, b, f1, '<', 1, *pflags);
1168 					break;
1169 				case '>':
1170 					fetch(ixnew, c, d, f2, '>', 0, *pflags);
1171 					break;
1172 				default:
1173 					diff_output("%%%c", *walk);
1174 					break;
1175 				}
1176 				continue;
1177 			}
1178 			diff_output("%c", *walk);
1179 		}
1180 	}
1181 	if (diff_format == D_SIDEBYSIDE) {
1182 		if (a > b) {
1183 			print_space(0, hw + padding , *pflags);
1184 		} else {
1185 			nc = fetch(ixold, a, b, f1, '\0', 1, *pflags);
1186 			print_space(nc, hw - nc + padding, *pflags);
1187 		}
1188 		diff_output("%c", (a>b)? '>' : ((c>d)? '<' : '|'));
1189 		print_space(hw + padding + 1 , padding, *pflags);
1190 		fetch(ixnew, c, d, f2, '\0', 0, *pflags);
1191 		diff_output("\n");
1192 	}
1193 	if (diff_format == D_NORMAL || diff_format == D_IFDEF) {
1194 		fetch(ixold, a, b, f1, '<', 1, *pflags);
1195 		if (a <= b && c <= d && diff_format == D_NORMAL)
1196 			diff_output("---\n");
1197 	}
1198 	if (diff_format != D_GFORMAT && diff_format != D_SIDEBYSIDE)
1199 		fetch(ixnew, c, d, f2, diff_format == D_NORMAL ? '>' : '\0', 0, *pflags);
1200 	if (edoffset != 0 && diff_format == D_EDIT) {
1201 		/*
1202 		 * A non-zero edoffset value for D_EDIT indicates that the
1203 		 * last line printed was a bare dot (".") that has been
1204 		 * escaped as ".." to prevent ed(1) from misinterpreting
1205 		 * it.  We have to add a substitute command to change this
1206 		 * back and restart where we left off.
1207 		 */
1208 		diff_output(".\n");
1209 		diff_output("%ds/.//\n", a + edoffset - 1);
1210 		b = a + edoffset - 1;
1211 		a = b + 1;
1212 		c += edoffset;
1213 		goto restart;
1214 	}
1215 	if ((diff_format == D_EDIT || diff_format == D_REVERSE) && c <= d)
1216 		diff_output(".\n");
1217 	if (inifdef) {
1218 		diff_output("#endif /* %s */\n", ifdefname);
1219 		inifdef = 0;
1220 	}
1221 }
1222 
1223 static int
1224 fetch(long *f, int a, int b, FILE *lb, int ch, int oldfile, int flags)
1225 {
1226 	int i, j, c, lastc, col, nc, newcol;
1227 
1228 	edoffset = 0;
1229 	nc = 0;
1230 	/*
1231 	 * When doing #ifdef's, copy down to current line
1232 	 * if this is the first file, so that stuff makes it to output.
1233 	 */
1234 	if ((diff_format == D_IFDEF) && oldfile) {
1235 		long curpos = ftell(lb);
1236 		/* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1237 		nc = f[a > b ? b : a - 1] - curpos;
1238 		for (i = 0; i < nc; i++)
1239 			diff_output("%c", getc(lb));
1240 	}
1241 	if (a > b)
1242 		return (0);
1243 	if (diff_format == D_IFDEF) {
1244 		if (inifdef) {
1245 			diff_output("#else /* %s%s */\n",
1246 			    oldfile == 1 ? "!" : "", ifdefname);
1247 		} else {
1248 			if (oldfile)
1249 				diff_output("#ifndef %s\n", ifdefname);
1250 			else
1251 				diff_output("#ifdef %s\n", ifdefname);
1252 		}
1253 		inifdef = 1 + oldfile;
1254 	}
1255 	for (i = a; i <= b; i++) {
1256 		fseek(lb, f[i - 1], SEEK_SET);
1257 		nc = (f[i] - f[i - 1]);
1258 		if (diff_format == D_SIDEBYSIDE && hw < nc)
1259 			nc = hw;
1260 		if ((diff_format != D_IFDEF && diff_format != D_GFORMAT) &&
1261 		    ch != '\0') {
1262 			diff_output("%c", ch);
1263 			if (Tflag && (diff_format == D_NORMAL ||
1264 			    diff_format == D_CONTEXT ||
1265 			    diff_format == D_UNIFIED))
1266 				diff_output("\t");
1267 			else if (diff_format != D_UNIFIED)
1268 				diff_output(" ");
1269 		}
1270 		col = 0;
1271 		for (j = 0, lastc = '\0'; j < nc; j++, lastc = c) {
1272 			if ((c = getc(lb)) == EOF) {
1273 				if (diff_format == D_EDIT ||
1274 				    diff_format == D_REVERSE ||
1275 				    diff_format == D_NREVERSE)
1276 					warnx("No newline at end of file");
1277 				else
1278 					diff_output("\n\\ No newline at end of "
1279 					    "file\n");
1280 				return col;
1281 			}
1282 			/*
1283 			 * when using --side-by-side, col needs to be increased
1284 			 * in any case to keep the columns aligned
1285 			 */
1286 			if (c == '\t') {
1287 				if (flags & D_EXPANDTABS) {
1288 					newcol = ((col/tabsize)+1)*tabsize;
1289 					do {
1290 						if (diff_format == D_SIDEBYSIDE)
1291 							j++;
1292 						diff_output(" ");
1293 					} while (++col < newcol && j < nc);
1294 				} else {
1295 					if (diff_format == D_SIDEBYSIDE) {
1296 						if ((j + tabsize) > nc) {
1297 							diff_output("%*s",
1298 							nc - j,"");
1299 							j = col = nc;
1300 						} else {
1301 							diff_output("\t");
1302 							col += tabsize - 1;
1303 							j += tabsize - 1;
1304 						}
1305 					} else {
1306 						diff_output("\t");
1307 						col++;
1308 					}
1309 				}
1310 			} else {
1311 				if (diff_format == D_EDIT && j == 1 && c == '\n'
1312 				    && lastc == '.') {
1313 					/*
1314 					 * Don't print a bare "." line
1315 					 * since that will confuse ed(1).
1316 					 * Print ".." instead and set the,
1317 					 * global variable edoffset to an
1318 					 * offset from which to restart.
1319 					 * The caller must check the value
1320 					 * of edoffset
1321 					 */
1322 					diff_output(".\n");
1323 					edoffset = i - a + 1;
1324 					return edoffset;
1325 				}
1326 				/* when side-by-side, do not print a newline */
1327 				if (diff_format != D_SIDEBYSIDE || c != '\n') {
1328 					diff_output("%c", c);
1329 					col++;
1330 				}
1331 			}
1332 		}
1333 	}
1334 	return col;
1335 }
1336 
1337 /*
1338  * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578.
1339  */
1340 static int
1341 readhash(FILE *f, int flags)
1342 {
1343 	int i, t, space;
1344 	int sum;
1345 
1346 	sum = 1;
1347 	space = 0;
1348 	if ((flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) == 0) {
1349 		if (flags & D_IGNORECASE)
1350 			for (i = 0; (t = getc(f)) != '\n'; i++) {
1351 				if (flags & D_STRIPCR && t == '\r') {
1352 					t = getc(f);
1353 					if (t == '\n')
1354 						break;
1355 					ungetc(t, f);
1356 				}
1357 				if (t == EOF) {
1358 					if (i == 0)
1359 						return (0);
1360 					break;
1361 				}
1362 				sum = sum * 127 + chrtran(t);
1363 			}
1364 		else
1365 			for (i = 0; (t = getc(f)) != '\n'; i++) {
1366 				if (flags & D_STRIPCR && t == '\r') {
1367 					t = getc(f);
1368 					if (t == '\n')
1369 						break;
1370 					ungetc(t, f);
1371 				}
1372 				if (t == EOF) {
1373 					if (i == 0)
1374 						return (0);
1375 					break;
1376 				}
1377 				sum = sum * 127 + t;
1378 			}
1379 	} else {
1380 		for (i = 0;;) {
1381 			switch (t = getc(f)) {
1382 			case '\r':
1383 			case '\t':
1384 			case '\v':
1385 			case '\f':
1386 			case ' ':
1387 				space++;
1388 				continue;
1389 			default:
1390 				if (space && (flags & D_IGNOREBLANKS) == 0) {
1391 					i++;
1392 					space = 0;
1393 				}
1394 				sum = sum * 127 + chrtran(t);
1395 				i++;
1396 				continue;
1397 			case EOF:
1398 				if (i == 0)
1399 					return (0);
1400 				/* FALLTHROUGH */
1401 			case '\n':
1402 				break;
1403 			}
1404 			break;
1405 		}
1406 	}
1407 	/*
1408 	 * There is a remote possibility that we end up with a zero sum.
1409 	 * Zero is used as an EOF marker, so return 1 instead.
1410 	 */
1411 	return (sum == 0 ? 1 : sum);
1412 }
1413 
1414 static int
1415 asciifile(FILE *f)
1416 {
1417 	unsigned char buf[BUFSIZ];
1418 	size_t cnt;
1419 
1420 	if (f == NULL)
1421 		return (1);
1422 
1423 	rewind(f);
1424 	cnt = fread(buf, 1, sizeof(buf), f);
1425 	return (memchr(buf, '\0', cnt) == NULL);
1426 }
1427 
1428 #define begins_with(s, pre) (strncmp(s, pre, sizeof(pre)-1) == 0)
1429 
1430 static char *
1431 match_function(const long *f, int pos, FILE *fp)
1432 {
1433 	unsigned char buf[FUNCTION_CONTEXT_SIZE];
1434 	size_t nc;
1435 	int last = lastline;
1436 	const char *state = NULL;
1437 
1438 	lastline = pos;
1439 	while (pos > last) {
1440 		fseek(fp, f[pos - 1], SEEK_SET);
1441 		nc = f[pos] - f[pos - 1];
1442 		if (nc >= sizeof(buf))
1443 			nc = sizeof(buf) - 1;
1444 		nc = fread(buf, 1, nc, fp);
1445 		if (nc > 0) {
1446 			buf[nc] = '\0';
1447 			buf[strcspn(buf, "\n")] = '\0';
1448 			if (isalpha(buf[0]) || buf[0] == '_' || buf[0] == '$') {
1449 				if (begins_with(buf, "private:")) {
1450 					if (!state)
1451 						state = " (private)";
1452 				} else if (begins_with(buf, "protected:")) {
1453 					if (!state)
1454 						state = " (protected)";
1455 				} else if (begins_with(buf, "public:")) {
1456 					if (!state)
1457 						state = " (public)";
1458 				} else {
1459 					strlcpy(lastbuf, buf, sizeof lastbuf);
1460 					if (state)
1461 						strlcat(lastbuf, state,
1462 						    sizeof lastbuf);
1463 					lastmatchline = pos;
1464 					return lastbuf;
1465 				}
1466 			}
1467 		}
1468 		pos--;
1469 	}
1470 	return lastmatchline > 0 ? lastbuf : NULL;
1471 }
1472 
1473 /* dump accumulated "context" diff changes */
1474 static void
1475 dump_context_vec(FILE *f1, FILE *f2, int flags)
1476 {
1477 	struct context_vec *cvp = context_vec_start;
1478 	int lowa, upb, lowc, upd, do_output;
1479 	int a, b, c, d;
1480 	char ch, *f;
1481 
1482 	if (context_vec_start > context_vec_ptr)
1483 		return;
1484 
1485 	b = d = 0;		/* gcc */
1486 	lowa = MAX(1, cvp->a - diff_context);
1487 	upb = MIN(len[0], context_vec_ptr->b + diff_context);
1488 	lowc = MAX(1, cvp->c - diff_context);
1489 	upd = MIN(len[1], context_vec_ptr->d + diff_context);
1490 
1491 	diff_output("***************");
1492 	if ((flags & D_PROTOTYPE)) {
1493 		f = match_function(ixold, lowa-1, f1);
1494 		if (f != NULL)
1495 			diff_output(" %s", f);
1496 	}
1497 	diff_output("\n*** ");
1498 	range(lowa, upb, ",");
1499 	diff_output(" ****\n");
1500 
1501 	/*
1502 	 * Output changes to the "old" file.  The first loop suppresses
1503 	 * output if there were no changes to the "old" file (we'll see
1504 	 * the "old" lines as context in the "new" list).
1505 	 */
1506 	do_output = 0;
1507 	for (; cvp <= context_vec_ptr; cvp++)
1508 		if (cvp->a <= cvp->b) {
1509 			cvp = context_vec_start;
1510 			do_output++;
1511 			break;
1512 		}
1513 	if (do_output) {
1514 		while (cvp <= context_vec_ptr) {
1515 			a = cvp->a;
1516 			b = cvp->b;
1517 			c = cvp->c;
1518 			d = cvp->d;
1519 
1520 			if (a <= b && c <= d)
1521 				ch = 'c';
1522 			else
1523 				ch = (a <= b) ? 'd' : 'a';
1524 
1525 			if (ch == 'a')
1526 				fetch(ixold, lowa, b, f1, ' ', 0, flags);
1527 			else {
1528 				fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1529 				fetch(ixold, a, b, f1,
1530 				    ch == 'c' ? '!' : '-', 0, flags);
1531 			}
1532 			lowa = b + 1;
1533 			cvp++;
1534 		}
1535 		fetch(ixold, b + 1, upb, f1, ' ', 0, flags);
1536 	}
1537 	/* output changes to the "new" file */
1538 	diff_output("--- ");
1539 	range(lowc, upd, ",");
1540 	diff_output(" ----\n");
1541 
1542 	do_output = 0;
1543 	for (cvp = context_vec_start; cvp <= context_vec_ptr; cvp++)
1544 		if (cvp->c <= cvp->d) {
1545 			cvp = context_vec_start;
1546 			do_output++;
1547 			break;
1548 		}
1549 	if (do_output) {
1550 		while (cvp <= context_vec_ptr) {
1551 			a = cvp->a;
1552 			b = cvp->b;
1553 			c = cvp->c;
1554 			d = cvp->d;
1555 
1556 			if (a <= b && c <= d)
1557 				ch = 'c';
1558 			else
1559 				ch = (a <= b) ? 'd' : 'a';
1560 
1561 			if (ch == 'd')
1562 				fetch(ixnew, lowc, d, f2, ' ', 0, flags);
1563 			else {
1564 				fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1565 				fetch(ixnew, c, d, f2,
1566 				    ch == 'c' ? '!' : '+', 0, flags);
1567 			}
1568 			lowc = d + 1;
1569 			cvp++;
1570 		}
1571 		fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1572 	}
1573 	context_vec_ptr = context_vec_start - 1;
1574 }
1575 
1576 /* dump accumulated "unified" diff changes */
1577 static void
1578 dump_unified_vec(FILE *f1, FILE *f2, int flags)
1579 {
1580 	struct context_vec *cvp = context_vec_start;
1581 	int lowa, upb, lowc, upd;
1582 	int a, b, c, d;
1583 	char ch, *f;
1584 
1585 	if (context_vec_start > context_vec_ptr)
1586 		return;
1587 
1588 	b = d = 0;		/* gcc */
1589 	lowa = MAX(1, cvp->a - diff_context);
1590 	upb = MIN(len[0], context_vec_ptr->b + diff_context);
1591 	lowc = MAX(1, cvp->c - diff_context);
1592 	upd = MIN(len[1], context_vec_ptr->d + diff_context);
1593 
1594 	diff_output("@@ -");
1595 	uni_range(lowa, upb);
1596 	diff_output(" +");
1597 	uni_range(lowc, upd);
1598 	diff_output(" @@");
1599 	if ((flags & D_PROTOTYPE)) {
1600 		f = match_function(ixold, lowa-1, f1);
1601 		if (f != NULL)
1602 			diff_output(" %s", f);
1603 	}
1604 	diff_output("\n");
1605 
1606 	/*
1607 	 * Output changes in "unified" diff format--the old and new lines
1608 	 * are printed together.
1609 	 */
1610 	for (; cvp <= context_vec_ptr; cvp++) {
1611 		a = cvp->a;
1612 		b = cvp->b;
1613 		c = cvp->c;
1614 		d = cvp->d;
1615 
1616 		/*
1617 		 * c: both new and old changes
1618 		 * d: only changes in the old file
1619 		 * a: only changes in the new file
1620 		 */
1621 		if (a <= b && c <= d)
1622 			ch = 'c';
1623 		else
1624 			ch = (a <= b) ? 'd' : 'a';
1625 
1626 		switch (ch) {
1627 		case 'c':
1628 			fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1629 			fetch(ixold, a, b, f1, '-', 0, flags);
1630 			fetch(ixnew, c, d, f2, '+', 0, flags);
1631 			break;
1632 		case 'd':
1633 			fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1634 			fetch(ixold, a, b, f1, '-', 0, flags);
1635 			break;
1636 		case 'a':
1637 			fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1638 			fetch(ixnew, c, d, f2, '+', 0, flags);
1639 			break;
1640 		}
1641 		lowa = b + 1;
1642 		lowc = d + 1;
1643 	}
1644 	fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1645 
1646 	context_vec_ptr = context_vec_start - 1;
1647 }
1648 
1649 static void
1650 print_header(const char *file1, const char *file2)
1651 {
1652 	const char *time_format;
1653 	char buf1[256];
1654 	char buf2[256];
1655 	char end1[10];
1656 	char end2[10];
1657 	struct tm tm1, tm2, *tm_ptr1, *tm_ptr2;
1658 	int nsec1 = stb1.st_mtim.tv_nsec;
1659 	int nsec2 = stb2.st_mtim.tv_nsec;
1660 
1661 	time_format = "%Y-%m-%d %H:%M:%S";
1662 
1663 	if (cflag)
1664 		time_format = "%c";
1665 	tm_ptr1 = localtime_r(&stb1.st_mtime, &tm1);
1666 	tm_ptr2 = localtime_r(&stb2.st_mtime, &tm2);
1667 	strftime(buf1, 256, time_format, tm_ptr1);
1668 	strftime(buf2, 256, time_format, tm_ptr2);
1669 	if (!cflag) {
1670 		strftime(end1, 10, "%z", tm_ptr1);
1671 		strftime(end2, 10, "%z", tm_ptr2);
1672 		sprintf(buf1, "%s.%.9d %s", buf1, nsec1, end1);
1673 		sprintf(buf2, "%s.%.9d %s", buf2, nsec2, end2);
1674 	}
1675 	if (label[0] != NULL)
1676 		diff_output("%s %s\n", diff_format == D_CONTEXT ? "***" : "---",
1677 		    label[0]);
1678 	else
1679 		diff_output("%s %s\t%s\n", diff_format == D_CONTEXT ? "***" : "---",
1680 		    file1, buf1);
1681 	if (label[1] != NULL)
1682 		diff_output("%s %s\n", diff_format == D_CONTEXT ? "---" : "+++",
1683 		    label[1]);
1684 	else
1685 		diff_output("%s %s\t%s\n", diff_format == D_CONTEXT ? "---" : "+++",
1686 		    file2, buf2);
1687 }
1688 
1689 /*
1690  * Prints n number of space characters either by using tab
1691  * or single space characters.
1692  * nc is the preceding number of characters
1693  */
1694 static void
1695 print_space(int nc, int n, int flags) {
1696 	int i, col;
1697 
1698 	col = n;
1699 	if ((flags & D_EXPANDTABS) == 0) {
1700 		/* first tabstop may be closer than tabsize */
1701 		i = tabsize - (nc % tabsize);
1702 		while (col >= tabsize) {
1703 			diff_output("\t");
1704 			col -= i;
1705 			i = tabsize;
1706 		}
1707 	}
1708 	diff_output("%*s", col, "");
1709 }
1710