xref: /freebsd/usr.bin/diff/diffreg.c (revision a7dea1671b87c07d2d266f836bfa8b58efc7c134)
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 static FILE	*opentemp(const char *);
185 static void	 output(char *, FILE *, char *, FILE *, int);
186 static void	 check(FILE *, FILE *, int);
187 static void	 range(int, int, const char *);
188 static void	 uni_range(int, int);
189 static void	 dump_context_vec(FILE *, FILE *, int);
190 static void	 dump_unified_vec(FILE *, FILE *, int);
191 static void	 prepare(int, FILE *, size_t, int);
192 static void	 prune(void);
193 static void	 equiv(struct line *, int, struct line *, int, int *);
194 static void	 unravel(int);
195 static void	 unsort(struct line *, int, int *);
196 static void	 change(char *, FILE *, char *, FILE *, int, int, int, int, int *);
197 static void	 sort(struct line *, int);
198 static void	 print_header(const char *, const char *);
199 static bool	 ignoreline_pattern(char *);
200 static bool	 ignoreline(char *, bool);
201 static int	 asciifile(FILE *);
202 static int	 fetch(long *, int, int, FILE *, int, int, int);
203 static int	 newcand(int, int, int);
204 static int	 search(int *, int, int);
205 static int	 skipline(FILE *);
206 static int	 isqrt(int);
207 static int	 stone(int *, int, int *, int *, int);
208 static int	 readhash(FILE *, int);
209 static int	 files_differ(FILE *, FILE *, int);
210 static char	*match_function(const long *, int, FILE *);
211 static char	*preadline(int, size_t, off_t);
212 
213 static int  *J;			/* will be overlaid on class */
214 static int  *class;		/* will be overlaid on file[0] */
215 static int  *klist;		/* will be overlaid on file[0] after class */
216 static int  *member;		/* will be overlaid on file[1] */
217 static int   clen;
218 static int   inifdef;		/* whether or not we are in a #ifdef block */
219 static int   len[2];
220 static int   pref, suff;	/* length of prefix and suffix */
221 static int   slen[2];
222 static int   anychange;
223 static long *ixnew;		/* will be overlaid on file[1] */
224 static long *ixold;		/* will be overlaid on klist */
225 static struct cand *clist;	/* merely a free storage pot for candidates */
226 static int   clistlen;		/* the length of clist */
227 static struct line *sfile[2];	/* shortened by pruning common prefix/suffix */
228 static int (*chrtran)(int);	/* translation table for case-folding */
229 static struct context_vec *context_vec_start;
230 static struct context_vec *context_vec_end;
231 static struct context_vec *context_vec_ptr;
232 
233 #define FUNCTION_CONTEXT_SIZE	55
234 static char lastbuf[FUNCTION_CONTEXT_SIZE];
235 static int lastline;
236 static int lastmatchline;
237 
238 static int
239 clow2low(int c)
240 {
241 
242 	return (c);
243 }
244 
245 static int
246 cup2low(int c)
247 {
248 
249 	return tolower(c);
250 }
251 
252 int
253 diffreg(char *file1, char *file2, int flags, int capsicum)
254 {
255 	FILE *f1, *f2;
256 	int i, rval;
257 	struct pr *pr = NULL;
258 	cap_rights_t rights_ro;
259 
260 	f1 = f2 = NULL;
261 	rval = D_SAME;
262 	anychange = 0;
263 	lastline = 0;
264 	lastmatchline = 0;
265 	context_vec_ptr = context_vec_start - 1;
266 	if (flags & D_IGNORECASE)
267 		chrtran = cup2low;
268 	else
269 		chrtran = clow2low;
270 	if (S_ISDIR(stb1.st_mode) != S_ISDIR(stb2.st_mode))
271 		return (S_ISDIR(stb1.st_mode) ? D_MISMATCH1 : D_MISMATCH2);
272 	if (strcmp(file1, "-") == 0 && strcmp(file2, "-") == 0)
273 		goto closem;
274 
275 	if (flags & D_EMPTY1)
276 		f1 = fopen(_PATH_DEVNULL, "r");
277 	else {
278 		if (!S_ISREG(stb1.st_mode)) {
279 			if ((f1 = opentemp(file1)) == NULL ||
280 			    fstat(fileno(f1), &stb1) == -1) {
281 				warn("%s", file1);
282 				status |= 2;
283 				goto closem;
284 			}
285 		} else if (strcmp(file1, "-") == 0)
286 			f1 = stdin;
287 		else
288 			f1 = fopen(file1, "r");
289 	}
290 	if (f1 == NULL) {
291 		warn("%s", file1);
292 		status |= 2;
293 		goto closem;
294 	}
295 
296 	if (flags & D_EMPTY2)
297 		f2 = fopen(_PATH_DEVNULL, "r");
298 	else {
299 		if (!S_ISREG(stb2.st_mode)) {
300 			if ((f2 = opentemp(file2)) == NULL ||
301 			    fstat(fileno(f2), &stb2) == -1) {
302 				warn("%s", file2);
303 				status |= 2;
304 				goto closem;
305 			}
306 		} else if (strcmp(file2, "-") == 0)
307 			f2 = stdin;
308 		else
309 			f2 = fopen(file2, "r");
310 	}
311 	if (f2 == NULL) {
312 		warn("%s", file2);
313 		status |= 2;
314 		goto closem;
315 	}
316 
317 	if (lflag)
318 		pr = start_pr(file1, file2);
319 
320 	if (capsicum) {
321 		cap_rights_init(&rights_ro, CAP_READ, CAP_FSTAT, CAP_SEEK);
322 		if (caph_rights_limit(fileno(f1), &rights_ro) < 0)
323 			err(2, "unable to limit rights on: %s", file1);
324 		if (caph_rights_limit(fileno(f2), &rights_ro) < 0)
325 			err(2, "unable to limit rights on: %s", file2);
326 		if (fileno(f1) == STDIN_FILENO || fileno(f2) == STDIN_FILENO) {
327 			/* stding has already been limited */
328 			if (caph_limit_stderr() == -1)
329 				err(2, "unable to limit stderr");
330 			if (caph_limit_stdout() == -1)
331 				err(2, "unable to limit stdout");
332 		} else if (caph_limit_stdio() == -1)
333 				err(2, "unable to limit stdio");
334 
335 		caph_cache_catpages();
336 		caph_cache_tzdata();
337 		if (caph_enter() < 0)
338 			err(2, "unable to enter capability mode");
339 	}
340 
341 	switch (files_differ(f1, f2, flags)) {
342 	case 0:
343 		goto closem;
344 	case 1:
345 		break;
346 	default:
347 		/* error */
348 		status |= 2;
349 		goto closem;
350 	}
351 
352 	if (diff_format == D_BRIEF && ignore_pats == NULL) {
353 		rval = D_DIFFER;
354 		status |= 1;
355 		goto closem;
356 	}
357 	if ((flags & D_FORCEASCII) == 0 &&
358 	    (!asciifile(f1) || !asciifile(f2))) {
359 		rval = D_BINARY;
360 		status |= 1;
361 		goto closem;
362 	}
363 	prepare(0, f1, stb1.st_size, flags);
364 	prepare(1, f2, stb2.st_size, flags);
365 
366 	prune();
367 	sort(sfile[0], slen[0]);
368 	sort(sfile[1], slen[1]);
369 
370 	member = (int *)file[1];
371 	equiv(sfile[0], slen[0], sfile[1], slen[1], member);
372 	member = xreallocarray(member, slen[1] + 2, sizeof(*member));
373 
374 	class = (int *)file[0];
375 	unsort(sfile[0], slen[0], class);
376 	class = xreallocarray(class, slen[0] + 2, sizeof(*class));
377 
378 	klist = xcalloc(slen[0] + 2, sizeof(*klist));
379 	clen = 0;
380 	clistlen = 100;
381 	clist = xcalloc(clistlen, sizeof(*clist));
382 	i = stone(class, slen[0], member, klist, flags);
383 	free(member);
384 	free(class);
385 
386 	J = xreallocarray(J, len[0] + 2, sizeof(*J));
387 	unravel(klist[i]);
388 	free(clist);
389 	free(klist);
390 
391 	ixold = xreallocarray(ixold, len[0] + 2, sizeof(*ixold));
392 	ixnew = xreallocarray(ixnew, len[1] + 2, sizeof(*ixnew));
393 	check(f1, f2, flags);
394 	output(file1, f1, file2, f2, flags);
395 	if (pr != NULL)
396 		stop_pr(pr);
397 
398 closem:
399 	if (anychange) {
400 		status |= 1;
401 		if (rval == D_SAME)
402 			rval = D_DIFFER;
403 	}
404 	if (f1 != NULL)
405 		fclose(f1);
406 	if (f2 != NULL)
407 		fclose(f2);
408 
409 	return (rval);
410 }
411 
412 /*
413  * Check to see if the given files differ.
414  * Returns 0 if they are the same, 1 if different, and -1 on error.
415  * XXX - could use code from cmp(1) [faster]
416  */
417 static int
418 files_differ(FILE *f1, FILE *f2, int flags)
419 {
420 	char buf1[BUFSIZ], buf2[BUFSIZ];
421 	size_t i, j;
422 
423 	if ((flags & (D_EMPTY1|D_EMPTY2)) || stb1.st_size != stb2.st_size ||
424 	    (stb1.st_mode & S_IFMT) != (stb2.st_mode & S_IFMT))
425 		return (1);
426 	for (;;) {
427 		i = fread(buf1, 1, sizeof(buf1), f1);
428 		j = fread(buf2, 1, sizeof(buf2), f2);
429 		if ((!i && ferror(f1)) || (!j && ferror(f2)))
430 			return (-1);
431 		if (i != j)
432 			return (1);
433 		if (i == 0)
434 			return (0);
435 		if (memcmp(buf1, buf2, i) != 0)
436 			return (1);
437 	}
438 }
439 
440 static FILE *
441 opentemp(const char *f)
442 {
443 	char buf[BUFSIZ], tempfile[PATH_MAX];
444 	ssize_t nread;
445 	int ifd, ofd;
446 
447 	if (strcmp(f, "-") == 0)
448 		ifd = STDIN_FILENO;
449 	else if ((ifd = open(f, O_RDONLY, 0644)) == -1)
450 		return (NULL);
451 
452 	(void)strlcpy(tempfile, _PATH_TMP "/diff.XXXXXXXX", sizeof(tempfile));
453 
454 	if ((ofd = mkstemp(tempfile)) == -1) {
455 		close(ifd);
456 		return (NULL);
457 	}
458 	unlink(tempfile);
459 	while ((nread = read(ifd, buf, BUFSIZ)) > 0) {
460 		if (write(ofd, buf, nread) != nread) {
461 			close(ifd);
462 			close(ofd);
463 			return (NULL);
464 		}
465 	}
466 	close(ifd);
467 	lseek(ofd, (off_t)0, SEEK_SET);
468 	return (fdopen(ofd, "r"));
469 }
470 
471 char *
472 splice(char *dir, char *path)
473 {
474 	char *tail, *buf;
475 	size_t dirlen;
476 
477 	dirlen = strlen(dir);
478 	while (dirlen != 0 && dir[dirlen - 1] == '/')
479 	    dirlen--;
480 	if ((tail = strrchr(path, '/')) == NULL)
481 		tail = path;
482 	else
483 		tail++;
484 	xasprintf(&buf, "%.*s/%s", (int)dirlen, dir, tail);
485 	return (buf);
486 }
487 
488 static void
489 prepare(int i, FILE *fd, size_t filesize, int flags)
490 {
491 	struct line *p;
492 	int h;
493 	size_t sz, j;
494 
495 	rewind(fd);
496 
497 	sz = MIN(filesize, SIZE_MAX) / 25;
498 	if (sz < 100)
499 		sz = 100;
500 
501 	p = xcalloc(sz + 3, sizeof(*p));
502 	for (j = 0; (h = readhash(fd, flags));) {
503 		if (j == sz) {
504 			sz = sz * 3 / 2;
505 			p = xreallocarray(p, sz + 3, sizeof(*p));
506 		}
507 		p[++j].value = h;
508 	}
509 	len[i] = j;
510 	file[i] = p;
511 }
512 
513 static void
514 prune(void)
515 {
516 	int i, j;
517 
518 	for (pref = 0; pref < len[0] && pref < len[1] &&
519 	    file[0][pref + 1].value == file[1][pref + 1].value;
520 	    pref++)
521 		;
522 	for (suff = 0; suff < len[0] - pref && suff < len[1] - pref &&
523 	    file[0][len[0] - suff].value == file[1][len[1] - suff].value;
524 	    suff++)
525 		;
526 	for (j = 0; j < 2; j++) {
527 		sfile[j] = file[j] + pref;
528 		slen[j] = len[j] - pref - suff;
529 		for (i = 0; i <= slen[j]; i++)
530 			sfile[j][i].serial = i;
531 	}
532 }
533 
534 static void
535 equiv(struct line *a, int n, struct line *b, int m, int *c)
536 {
537 	int i, j;
538 
539 	i = j = 1;
540 	while (i <= n && j <= m) {
541 		if (a[i].value < b[j].value)
542 			a[i++].value = 0;
543 		else if (a[i].value == b[j].value)
544 			a[i++].value = j;
545 		else
546 			j++;
547 	}
548 	while (i <= n)
549 		a[i++].value = 0;
550 	b[m + 1].value = 0;
551 	j = 0;
552 	while (++j <= m) {
553 		c[j] = -b[j].serial;
554 		while (b[j + 1].value == b[j].value) {
555 			j++;
556 			c[j] = b[j].serial;
557 		}
558 	}
559 	c[j] = -1;
560 }
561 
562 /* Code taken from ping.c */
563 static int
564 isqrt(int n)
565 {
566 	int y, x = 1;
567 
568 	if (n == 0)
569 		return (0);
570 
571 	do { /* newton was a stinker */
572 		y = x;
573 		x = n / x;
574 		x += y;
575 		x /= 2;
576 	} while ((x - y) > 1 || (x - y) < -1);
577 
578 	return (x);
579 }
580 
581 static int
582 stone(int *a, int n, int *b, int *c, int flags)
583 {
584 	int i, k, y, j, l;
585 	int oldc, tc, oldl, sq;
586 	u_int numtries, bound;
587 
588 	if (flags & D_MINIMAL)
589 		bound = UINT_MAX;
590 	else {
591 		sq = isqrt(n);
592 		bound = MAX(256, sq);
593 	}
594 
595 	k = 0;
596 	c[0] = newcand(0, 0, 0);
597 	for (i = 1; i <= n; i++) {
598 		j = a[i];
599 		if (j == 0)
600 			continue;
601 		y = -b[j];
602 		oldl = 0;
603 		oldc = c[0];
604 		numtries = 0;
605 		do {
606 			if (y <= clist[oldc].y)
607 				continue;
608 			l = search(c, k, y);
609 			if (l != oldl + 1)
610 				oldc = c[l - 1];
611 			if (l <= k) {
612 				if (clist[c[l]].y <= y)
613 					continue;
614 				tc = c[l];
615 				c[l] = newcand(i, y, oldc);
616 				oldc = tc;
617 				oldl = l;
618 				numtries++;
619 			} else {
620 				c[l] = newcand(i, y, oldc);
621 				k++;
622 				break;
623 			}
624 		} while ((y = b[++j]) > 0 && numtries < bound);
625 	}
626 	return (k);
627 }
628 
629 static int
630 newcand(int x, int y, int pred)
631 {
632 	struct cand *q;
633 
634 	if (clen == clistlen) {
635 		clistlen = clistlen * 11 / 10;
636 		clist = xreallocarray(clist, clistlen, sizeof(*clist));
637 	}
638 	q = clist + clen;
639 	q->x = x;
640 	q->y = y;
641 	q->pred = pred;
642 	return (clen++);
643 }
644 
645 static int
646 search(int *c, int k, int y)
647 {
648 	int i, j, l, t;
649 
650 	if (clist[c[k]].y < y)	/* quick look for typical case */
651 		return (k + 1);
652 	i = 0;
653 	j = k + 1;
654 	for (;;) {
655 		l = (i + j) / 2;
656 		if (l <= i)
657 			break;
658 		t = clist[c[l]].y;
659 		if (t > y)
660 			j = l;
661 		else if (t < y)
662 			i = l;
663 		else
664 			return (l);
665 	}
666 	return (l + 1);
667 }
668 
669 static void
670 unravel(int p)
671 {
672 	struct cand *q;
673 	int i;
674 
675 	for (i = 0; i <= len[0]; i++)
676 		J[i] = i <= pref ? i :
677 		    i > len[0] - suff ? i + len[1] - len[0] : 0;
678 	for (q = clist + p; q->y != 0; q = clist + q->pred)
679 		J[q->x + pref] = q->y + pref;
680 }
681 
682 /*
683  * Check does double duty:
684  *  1.	ferret out any fortuitous correspondences due
685  *	to confounding by hashing (which result in "jackpot")
686  *  2.  collect random access indexes to the two files
687  */
688 static void
689 check(FILE *f1, FILE *f2, int flags)
690 {
691 	int i, j, jackpot, c, d;
692 	long ctold, ctnew;
693 
694 	rewind(f1);
695 	rewind(f2);
696 	j = 1;
697 	ixold[0] = ixnew[0] = 0;
698 	jackpot = 0;
699 	ctold = ctnew = 0;
700 	for (i = 1; i <= len[0]; i++) {
701 		if (J[i] == 0) {
702 			ixold[i] = ctold += skipline(f1);
703 			continue;
704 		}
705 		while (j < J[i]) {
706 			ixnew[j] = ctnew += skipline(f2);
707 			j++;
708 		}
709 		if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS|D_IGNORECASE|D_STRIPCR)) {
710 			for (;;) {
711 				c = getc(f1);
712 				d = getc(f2);
713 				/*
714 				 * GNU diff ignores a missing newline
715 				 * in one file for -b or -w.
716 				 */
717 				if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) {
718 					if (c == EOF && d == '\n') {
719 						ctnew++;
720 						break;
721 					} else if (c == '\n' && d == EOF) {
722 						ctold++;
723 						break;
724 					}
725 				}
726 				ctold++;
727 				ctnew++;
728 				if (flags & D_STRIPCR && (c == '\r' || d == '\r')) {
729 					if (c == '\r') {
730 						if ((c = getc(f1)) == '\n') {
731 							ctold++;
732 						} else {
733 							ungetc(c, f1);
734 						}
735 					}
736 					if (d == '\r') {
737 						if ((d = getc(f2)) == '\n') {
738 							ctnew++;
739 						} else {
740 							ungetc(d, f2);
741 						}
742 					}
743 					break;
744 				}
745 				if ((flags & D_FOLDBLANKS) && isspace(c) &&
746 				    isspace(d)) {
747 					do {
748 						if (c == '\n')
749 							break;
750 						ctold++;
751 					} while (isspace(c = getc(f1)));
752 					do {
753 						if (d == '\n')
754 							break;
755 						ctnew++;
756 					} while (isspace(d = getc(f2)));
757 				} else if ((flags & D_IGNOREBLANKS)) {
758 					while (isspace(c) && c != '\n') {
759 						c = getc(f1);
760 						ctold++;
761 					}
762 					while (isspace(d) && d != '\n') {
763 						d = getc(f2);
764 						ctnew++;
765 					}
766 				}
767 				if (chrtran(c) != chrtran(d)) {
768 					jackpot++;
769 					J[i] = 0;
770 					if (c != '\n' && c != EOF)
771 						ctold += skipline(f1);
772 					if (d != '\n' && c != EOF)
773 						ctnew += skipline(f2);
774 					break;
775 				}
776 				if (c == '\n' || c == EOF)
777 					break;
778 			}
779 		} else {
780 			for (;;) {
781 				ctold++;
782 				ctnew++;
783 				if ((c = getc(f1)) != (d = getc(f2))) {
784 					/* jackpot++; */
785 					J[i] = 0;
786 					if (c != '\n' && c != EOF)
787 						ctold += skipline(f1);
788 					if (d != '\n' && c != EOF)
789 						ctnew += skipline(f2);
790 					break;
791 				}
792 				if (c == '\n' || c == EOF)
793 					break;
794 			}
795 		}
796 		ixold[i] = ctold;
797 		ixnew[j] = ctnew;
798 		j++;
799 	}
800 	for (; j <= len[1]; j++) {
801 		ixnew[j] = ctnew += skipline(f2);
802 	}
803 	/*
804 	 * if (jackpot)
805 	 *	fprintf(stderr, "jackpot\n");
806 	 */
807 }
808 
809 /* shellsort CACM #201 */
810 static void
811 sort(struct line *a, int n)
812 {
813 	struct line *ai, *aim, w;
814 	int j, m = 0, k;
815 
816 	if (n == 0)
817 		return;
818 	for (j = 1; j <= n; j *= 2)
819 		m = 2 * j - 1;
820 	for (m /= 2; m != 0; m /= 2) {
821 		k = n - m;
822 		for (j = 1; j <= k; j++) {
823 			for (ai = &a[j]; ai > a; ai -= m) {
824 				aim = &ai[m];
825 				if (aim < ai)
826 					break;	/* wraparound */
827 				if (aim->value > ai[0].value ||
828 				    (aim->value == ai[0].value &&
829 					aim->serial > ai[0].serial))
830 					break;
831 				w.value = ai[0].value;
832 				ai[0].value = aim->value;
833 				aim->value = w.value;
834 				w.serial = ai[0].serial;
835 				ai[0].serial = aim->serial;
836 				aim->serial = w.serial;
837 			}
838 		}
839 	}
840 }
841 
842 static void
843 unsort(struct line *f, int l, int *b)
844 {
845 	int *a, i;
846 
847 	a = xcalloc(l + 1, sizeof(*a));
848 	for (i = 1; i <= l; i++)
849 		a[f[i].serial] = f[i].value;
850 	for (i = 1; i <= l; i++)
851 		b[i] = a[i];
852 	free(a);
853 }
854 
855 static int
856 skipline(FILE *f)
857 {
858 	int i, c;
859 
860 	for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++)
861 		continue;
862 	return (i);
863 }
864 
865 static void
866 output(char *file1, FILE *f1, char *file2, FILE *f2, int flags)
867 {
868 	int m, i0, i1, j0, j1;
869 
870 	rewind(f1);
871 	rewind(f2);
872 	m = len[0];
873 	J[0] = 0;
874 	J[m + 1] = len[1] + 1;
875 	if (diff_format != D_EDIT) {
876 		for (i0 = 1; i0 <= m; i0 = i1 + 1) {
877 			while (i0 <= m && J[i0] == J[i0 - 1] + 1)
878 				i0++;
879 			j0 = J[i0 - 1] + 1;
880 			i1 = i0 - 1;
881 			while (i1 < m && J[i1 + 1] == 0)
882 				i1++;
883 			j1 = J[i1 + 1] - 1;
884 			J[i1] = j1;
885 			change(file1, f1, file2, f2, i0, i1, j0, j1, &flags);
886 		}
887 	} else {
888 		for (i0 = m; i0 >= 1; i0 = i1 - 1) {
889 			while (i0 >= 1 && J[i0] == J[i0 + 1] - 1 && J[i0] != 0)
890 				i0--;
891 			j0 = J[i0 + 1] - 1;
892 			i1 = i0 + 1;
893 			while (i1 > 1 && J[i1 - 1] == 0)
894 				i1--;
895 			j1 = J[i1 - 1] + 1;
896 			J[i1] = j1;
897 			change(file1, f1, file2, f2, i1, i0, j1, j0, &flags);
898 		}
899 	}
900 	if (m == 0)
901 		change(file1, f1, file2, f2, 1, 0, 1, len[1], &flags);
902 	if (diff_format == D_IFDEF || diff_format == D_GFORMAT) {
903 		for (;;) {
904 #define	c i0
905 			if ((c = getc(f1)) == EOF)
906 				return;
907 			diff_output("%c", c);
908 		}
909 #undef c
910 	}
911 	if (anychange != 0) {
912 		if (diff_format == D_CONTEXT)
913 			dump_context_vec(f1, f2, flags);
914 		else if (diff_format == D_UNIFIED)
915 			dump_unified_vec(f1, f2, flags);
916 	}
917 }
918 
919 static void
920 range(int a, int b, const char *separator)
921 {
922 	diff_output("%d", a > b ? b : a);
923 	if (a < b)
924 		diff_output("%s%d", separator, b);
925 }
926 
927 static void
928 uni_range(int a, int b)
929 {
930 	if (a < b)
931 		diff_output("%d,%d", a, b - a + 1);
932 	else if (a == b)
933 		diff_output("%d", b);
934 	else
935 		diff_output("%d,0", b);
936 }
937 
938 static char *
939 preadline(int fd, size_t rlen, off_t off)
940 {
941 	char *line;
942 	ssize_t nr;
943 
944 	line = xmalloc(rlen + 1);
945 	if ((nr = pread(fd, line, rlen, off)) == -1)
946 		err(2, "preadline");
947 	if (nr > 0 && line[nr-1] == '\n')
948 		nr--;
949 	line[nr] = '\0';
950 	return (line);
951 }
952 
953 static bool
954 ignoreline_pattern(char *line)
955 {
956 	int ret;
957 
958 	ret = regexec(&ignore_re, line, 0, NULL, 0);
959 	free(line);
960 	return (ret == 0);	/* if it matched, it should be ignored. */
961 }
962 
963 static bool
964 ignoreline(char *line, bool skip_blanks)
965 {
966 
967 	if (ignore_pats != NULL && skip_blanks)
968 		return (ignoreline_pattern(line) || *line == '\0');
969 	if (ignore_pats != NULL)
970 		return (ignoreline_pattern(line));
971 	if (skip_blanks)
972 		return (*line == '\0');
973 	/* No ignore criteria specified */
974 	return (false);
975 }
976 
977 /*
978  * Indicate that there is a difference between lines a and b of the from file
979  * to get to lines c to d of the to file.  If a is greater then b then there
980  * are no lines in the from file involved and this means that there were
981  * lines appended (beginning at b).  If c is greater than d then there are
982  * lines missing from the to file.
983  */
984 static void
985 change(char *file1, FILE *f1, char *file2, FILE *f2, int a, int b, int c, int d,
986     int *pflags)
987 {
988 	static size_t max_context = 64;
989 	long curpos;
990 	int i, nc, f;
991 	const char *walk;
992 	bool skip_blanks;
993 
994 	skip_blanks = (*pflags & D_SKIPBLANKLINES);
995 restart:
996 	if ((diff_format != D_IFDEF || diff_format == D_GFORMAT) &&
997 	    a > b && c > d)
998 		return;
999 	if (ignore_pats != NULL || skip_blanks) {
1000 		char *line;
1001 		/*
1002 		 * All lines in the change, insert, or delete must
1003 		 * match an ignore pattern for the change to be
1004 		 * ignored.
1005 		 */
1006 		if (a <= b) {		/* Changes and deletes. */
1007 			for (i = a; i <= b; i++) {
1008 				line = preadline(fileno(f1),
1009 				    ixold[i] - ixold[i - 1], ixold[i - 1]);
1010 				if (!ignoreline(line, skip_blanks))
1011 					goto proceed;
1012 			}
1013 		}
1014 		if (a > b || c <= d) {	/* Changes and inserts. */
1015 			for (i = c; i <= d; i++) {
1016 				line = preadline(fileno(f2),
1017 				    ixnew[i] - ixnew[i - 1], ixnew[i - 1]);
1018 				if (!ignoreline(line, skip_blanks))
1019 					goto proceed;
1020 			}
1021 		}
1022 		return;
1023 	}
1024 proceed:
1025 	if (*pflags & D_HEADER && diff_format != D_BRIEF) {
1026 		diff_output("%s %s %s\n", diffargs, file1, file2);
1027 		*pflags &= ~D_HEADER;
1028 	}
1029 	if (diff_format == D_CONTEXT || diff_format == D_UNIFIED) {
1030 		/*
1031 		 * Allocate change records as needed.
1032 		 */
1033 		if (context_vec_ptr == context_vec_end - 1) {
1034 			ptrdiff_t offset = context_vec_ptr - context_vec_start;
1035 			max_context <<= 1;
1036 			context_vec_start = xreallocarray(context_vec_start,
1037 			    max_context, sizeof(*context_vec_start));
1038 			context_vec_end = context_vec_start + max_context;
1039 			context_vec_ptr = context_vec_start + offset;
1040 		}
1041 		if (anychange == 0) {
1042 			/*
1043 			 * Print the context/unidiff header first time through.
1044 			 */
1045 			print_header(file1, file2);
1046 			anychange = 1;
1047 		} else if (a > context_vec_ptr->b + (2 * diff_context) + 1 &&
1048 		    c > context_vec_ptr->d + (2 * diff_context) + 1) {
1049 			/*
1050 			 * If this change is more than 'diff_context' lines from the
1051 			 * previous change, dump the record and reset it.
1052 			 */
1053 			if (diff_format == D_CONTEXT)
1054 				dump_context_vec(f1, f2, *pflags);
1055 			else
1056 				dump_unified_vec(f1, f2, *pflags);
1057 		}
1058 		context_vec_ptr++;
1059 		context_vec_ptr->a = a;
1060 		context_vec_ptr->b = b;
1061 		context_vec_ptr->c = c;
1062 		context_vec_ptr->d = d;
1063 		return;
1064 	}
1065 	if (anychange == 0)
1066 		anychange = 1;
1067 	switch (diff_format) {
1068 	case D_BRIEF:
1069 		return;
1070 	case D_NORMAL:
1071 	case D_EDIT:
1072 		range(a, b, ",");
1073 		diff_output("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1074 		if (diff_format == D_NORMAL)
1075 			range(c, d, ",");
1076 		diff_output("\n");
1077 		break;
1078 	case D_REVERSE:
1079 		diff_output("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1080 		range(a, b, " ");
1081 		diff_output("\n");
1082 		break;
1083 	case D_NREVERSE:
1084 		if (a > b)
1085 			diff_output("a%d %d\n", b, d - c + 1);
1086 		else {
1087 			diff_output("d%d %d\n", a, b - a + 1);
1088 			if (!(c > d))
1089 				/* add changed lines */
1090 				diff_output("a%d %d\n", b, d - c + 1);
1091 		}
1092 		break;
1093 	}
1094 	if (diff_format == D_GFORMAT) {
1095 		curpos = ftell(f1);
1096 		/* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1097 		nc = ixold[a > b ? b : a - 1] - curpos;
1098 		for (i = 0; i < nc; i++)
1099 			diff_output("%c", getc(f1));
1100 		for (walk = group_format; *walk != '\0'; walk++) {
1101 			if (*walk == '%') {
1102 				walk++;
1103 				switch (*walk) {
1104 				case '<':
1105 					fetch(ixold, a, b, f1, '<', 1, *pflags);
1106 					break;
1107 				case '>':
1108 					fetch(ixnew, c, d, f2, '>', 0, *pflags);
1109 					break;
1110 				default:
1111 					diff_output("%%%c", *walk);
1112 					break;
1113 				}
1114 				continue;
1115 			}
1116 			diff_output("%c", *walk);
1117 		}
1118 	}
1119 	if (diff_format == D_NORMAL || diff_format == D_IFDEF) {
1120 		fetch(ixold, a, b, f1, '<', 1, *pflags);
1121 		if (a <= b && c <= d && diff_format == D_NORMAL)
1122 			diff_output("---\n");
1123 	}
1124 	f = 0;
1125 	if (diff_format != D_GFORMAT)
1126 		f = fetch(ixnew, c, d, f2, diff_format == D_NORMAL ? '>' : '\0', 0, *pflags);
1127 	if (f != 0 && diff_format == D_EDIT) {
1128 		/*
1129 		 * A non-zero return value for D_EDIT indicates that the
1130 		 * last line printed was a bare dot (".") that has been
1131 		 * escaped as ".." to prevent ed(1) from misinterpreting
1132 		 * it.  We have to add a substitute command to change this
1133 		 * back and restart where we left off.
1134 		 */
1135 		diff_output(".\n");
1136 		diff_output("%ds/.//\n", a + f - 1);
1137 		b = a + f - 1;
1138 		a = b + 1;
1139 		c += f;
1140 		goto restart;
1141 	}
1142 	if ((diff_format == D_EDIT || diff_format == D_REVERSE) && c <= d)
1143 		diff_output(".\n");
1144 	if (inifdef) {
1145 		diff_output("#endif /* %s */\n", ifdefname);
1146 		inifdef = 0;
1147 	}
1148 }
1149 
1150 static int
1151 fetch(long *f, int a, int b, FILE *lb, int ch, int oldfile, int flags)
1152 {
1153 	int i, j, c, lastc, col, nc;
1154 	int	newcol;
1155 
1156 	/*
1157 	 * When doing #ifdef's, copy down to current line
1158 	 * if this is the first file, so that stuff makes it to output.
1159 	 */
1160 	if ((diff_format == D_IFDEF) && oldfile) {
1161 		long curpos = ftell(lb);
1162 		/* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1163 		nc = f[a > b ? b : a - 1] - curpos;
1164 		for (i = 0; i < nc; i++)
1165 			diff_output("%c", getc(lb));
1166 	}
1167 	if (a > b)
1168 		return (0);
1169 	if (diff_format == D_IFDEF) {
1170 		if (inifdef) {
1171 			diff_output("#else /* %s%s */\n",
1172 			    oldfile == 1 ? "!" : "", ifdefname);
1173 		} else {
1174 			if (oldfile)
1175 				diff_output("#ifndef %s\n", ifdefname);
1176 			else
1177 				diff_output("#ifdef %s\n", ifdefname);
1178 		}
1179 		inifdef = 1 + oldfile;
1180 	}
1181 	for (i = a; i <= b; i++) {
1182 		fseek(lb, f[i - 1], SEEK_SET);
1183 		nc = f[i] - f[i - 1];
1184 		if ((diff_format != D_IFDEF && diff_format != D_GFORMAT) &&
1185 		    ch != '\0') {
1186 			diff_output("%c", ch);
1187 			if (Tflag && (diff_format == D_NORMAL || diff_format == D_CONTEXT
1188 			    || diff_format == D_UNIFIED))
1189 				diff_output("\t");
1190 			else if (diff_format != D_UNIFIED)
1191 				diff_output(" ");
1192 		}
1193 		col = 0;
1194 		for (j = 0, lastc = '\0'; j < nc; j++, lastc = c) {
1195 			if ((c = getc(lb)) == EOF) {
1196 				if (diff_format == D_EDIT || diff_format == D_REVERSE ||
1197 				    diff_format == D_NREVERSE)
1198 					warnx("No newline at end of file");
1199 				else
1200 					diff_output("\n\\ No newline at end of "
1201 					    "file\n");
1202 				return (0);
1203 			}
1204 			if (c == '\t' && (flags & D_EXPANDTABS)) {
1205 				newcol = ((col/tabsize)+1)*tabsize;
1206 				do {
1207 					diff_output(" ");
1208 				} while (++col < newcol);
1209 			} else {
1210 				if (diff_format == D_EDIT && j == 1 && c == '\n'
1211 				    && lastc == '.') {
1212 					/*
1213 					 * Don't print a bare "." line
1214 					 * since that will confuse ed(1).
1215 					 * Print ".." instead and return,
1216 					 * giving the caller an offset
1217 					 * from which to restart.
1218 					 */
1219 					diff_output(".\n");
1220 					return (i - a + 1);
1221 				}
1222 				diff_output("%c", c);
1223 				col++;
1224 			}
1225 		}
1226 	}
1227 	return (0);
1228 }
1229 
1230 /*
1231  * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578.
1232  */
1233 static int
1234 readhash(FILE *f, int flags)
1235 {
1236 	int i, t, space;
1237 	int sum;
1238 
1239 	sum = 1;
1240 	space = 0;
1241 	if ((flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) == 0) {
1242 		if (flags & D_IGNORECASE)
1243 			for (i = 0; (t = getc(f)) != '\n'; i++) {
1244 				if (flags & D_STRIPCR && t == '\r') {
1245 					t = getc(f);
1246 					if (t == '\n')
1247 						break;
1248 					ungetc(t, f);
1249 				}
1250 				if (t == EOF) {
1251 					if (i == 0)
1252 						return (0);
1253 					break;
1254 				}
1255 				sum = sum * 127 + chrtran(t);
1256 			}
1257 		else
1258 			for (i = 0; (t = getc(f)) != '\n'; i++) {
1259 				if (flags & D_STRIPCR && t == '\r') {
1260 					t = getc(f);
1261 					if (t == '\n')
1262 						break;
1263 					ungetc(t, f);
1264 				}
1265 				if (t == EOF) {
1266 					if (i == 0)
1267 						return (0);
1268 					break;
1269 				}
1270 				sum = sum * 127 + t;
1271 			}
1272 	} else {
1273 		for (i = 0;;) {
1274 			switch (t = getc(f)) {
1275 			case '\r':
1276 			case '\t':
1277 			case '\v':
1278 			case '\f':
1279 			case ' ':
1280 				space++;
1281 				continue;
1282 			default:
1283 				if (space && (flags & D_IGNOREBLANKS) == 0) {
1284 					i++;
1285 					space = 0;
1286 				}
1287 				sum = sum * 127 + chrtran(t);
1288 				i++;
1289 				continue;
1290 			case EOF:
1291 				if (i == 0)
1292 					return (0);
1293 				/* FALLTHROUGH */
1294 			case '\n':
1295 				break;
1296 			}
1297 			break;
1298 		}
1299 	}
1300 	/*
1301 	 * There is a remote possibility that we end up with a zero sum.
1302 	 * Zero is used as an EOF marker, so return 1 instead.
1303 	 */
1304 	return (sum == 0 ? 1 : sum);
1305 }
1306 
1307 static int
1308 asciifile(FILE *f)
1309 {
1310 	unsigned char buf[BUFSIZ];
1311 	size_t cnt;
1312 
1313 	if (f == NULL)
1314 		return (1);
1315 
1316 	rewind(f);
1317 	cnt = fread(buf, 1, sizeof(buf), f);
1318 	return (memchr(buf, '\0', cnt) == NULL);
1319 }
1320 
1321 #define begins_with(s, pre) (strncmp(s, pre, sizeof(pre)-1) == 0)
1322 
1323 static char *
1324 match_function(const long *f, int pos, FILE *fp)
1325 {
1326 	unsigned char buf[FUNCTION_CONTEXT_SIZE];
1327 	size_t nc;
1328 	int last = lastline;
1329 	const char *state = NULL;
1330 
1331 	lastline = pos;
1332 	while (pos > last) {
1333 		fseek(fp, f[pos - 1], SEEK_SET);
1334 		nc = f[pos] - f[pos - 1];
1335 		if (nc >= sizeof(buf))
1336 			nc = sizeof(buf) - 1;
1337 		nc = fread(buf, 1, nc, fp);
1338 		if (nc > 0) {
1339 			buf[nc] = '\0';
1340 			buf[strcspn(buf, "\n")] = '\0';
1341 			if (isalpha(buf[0]) || buf[0] == '_' || buf[0] == '$') {
1342 				if (begins_with(buf, "private:")) {
1343 					if (!state)
1344 						state = " (private)";
1345 				} else if (begins_with(buf, "protected:")) {
1346 					if (!state)
1347 						state = " (protected)";
1348 				} else if (begins_with(buf, "public:")) {
1349 					if (!state)
1350 						state = " (public)";
1351 				} else {
1352 					strlcpy(lastbuf, buf, sizeof lastbuf);
1353 					if (state)
1354 						strlcat(lastbuf, state,
1355 						    sizeof lastbuf);
1356 					lastmatchline = pos;
1357 					return lastbuf;
1358 				}
1359 			}
1360 		}
1361 		pos--;
1362 	}
1363 	return lastmatchline > 0 ? lastbuf : NULL;
1364 }
1365 
1366 /* dump accumulated "context" diff changes */
1367 static void
1368 dump_context_vec(FILE *f1, FILE *f2, int flags)
1369 {
1370 	struct context_vec *cvp = context_vec_start;
1371 	int lowa, upb, lowc, upd, do_output;
1372 	int a, b, c, d;
1373 	char ch, *f;
1374 
1375 	if (context_vec_start > context_vec_ptr)
1376 		return;
1377 
1378 	b = d = 0;		/* gcc */
1379 	lowa = MAX(1, cvp->a - diff_context);
1380 	upb = MIN(len[0], context_vec_ptr->b + diff_context);
1381 	lowc = MAX(1, cvp->c - diff_context);
1382 	upd = MIN(len[1], context_vec_ptr->d + diff_context);
1383 
1384 	diff_output("***************");
1385 	if ((flags & D_PROTOTYPE)) {
1386 		f = match_function(ixold, lowa-1, f1);
1387 		if (f != NULL)
1388 			diff_output(" %s", f);
1389 	}
1390 	diff_output("\n*** ");
1391 	range(lowa, upb, ",");
1392 	diff_output(" ****\n");
1393 
1394 	/*
1395 	 * Output changes to the "old" file.  The first loop suppresses
1396 	 * output if there were no changes to the "old" file (we'll see
1397 	 * the "old" lines as context in the "new" list).
1398 	 */
1399 	do_output = 0;
1400 	for (; cvp <= context_vec_ptr; cvp++)
1401 		if (cvp->a <= cvp->b) {
1402 			cvp = context_vec_start;
1403 			do_output++;
1404 			break;
1405 		}
1406 	if (do_output) {
1407 		while (cvp <= context_vec_ptr) {
1408 			a = cvp->a;
1409 			b = cvp->b;
1410 			c = cvp->c;
1411 			d = cvp->d;
1412 
1413 			if (a <= b && c <= d)
1414 				ch = 'c';
1415 			else
1416 				ch = (a <= b) ? 'd' : 'a';
1417 
1418 			if (ch == 'a')
1419 				fetch(ixold, lowa, b, f1, ' ', 0, flags);
1420 			else {
1421 				fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1422 				fetch(ixold, a, b, f1,
1423 				    ch == 'c' ? '!' : '-', 0, flags);
1424 			}
1425 			lowa = b + 1;
1426 			cvp++;
1427 		}
1428 		fetch(ixold, b + 1, upb, f1, ' ', 0, flags);
1429 	}
1430 	/* output changes to the "new" file */
1431 	diff_output("--- ");
1432 	range(lowc, upd, ",");
1433 	diff_output(" ----\n");
1434 
1435 	do_output = 0;
1436 	for (cvp = context_vec_start; cvp <= context_vec_ptr; cvp++)
1437 		if (cvp->c <= cvp->d) {
1438 			cvp = context_vec_start;
1439 			do_output++;
1440 			break;
1441 		}
1442 	if (do_output) {
1443 		while (cvp <= context_vec_ptr) {
1444 			a = cvp->a;
1445 			b = cvp->b;
1446 			c = cvp->c;
1447 			d = cvp->d;
1448 
1449 			if (a <= b && c <= d)
1450 				ch = 'c';
1451 			else
1452 				ch = (a <= b) ? 'd' : 'a';
1453 
1454 			if (ch == 'd')
1455 				fetch(ixnew, lowc, d, f2, ' ', 0, flags);
1456 			else {
1457 				fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1458 				fetch(ixnew, c, d, f2,
1459 				    ch == 'c' ? '!' : '+', 0, flags);
1460 			}
1461 			lowc = d + 1;
1462 			cvp++;
1463 		}
1464 		fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1465 	}
1466 	context_vec_ptr = context_vec_start - 1;
1467 }
1468 
1469 /* dump accumulated "unified" diff changes */
1470 static void
1471 dump_unified_vec(FILE *f1, FILE *f2, int flags)
1472 {
1473 	struct context_vec *cvp = context_vec_start;
1474 	int lowa, upb, lowc, upd;
1475 	int a, b, c, d;
1476 	char ch, *f;
1477 
1478 	if (context_vec_start > context_vec_ptr)
1479 		return;
1480 
1481 	b = d = 0;		/* gcc */
1482 	lowa = MAX(1, cvp->a - diff_context);
1483 	upb = MIN(len[0], context_vec_ptr->b + diff_context);
1484 	lowc = MAX(1, cvp->c - diff_context);
1485 	upd = MIN(len[1], context_vec_ptr->d + diff_context);
1486 
1487 	diff_output("@@ -");
1488 	uni_range(lowa, upb);
1489 	diff_output(" +");
1490 	uni_range(lowc, upd);
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 
1499 	/*
1500 	 * Output changes in "unified" diff format--the old and new lines
1501 	 * are printed together.
1502 	 */
1503 	for (; cvp <= context_vec_ptr; cvp++) {
1504 		a = cvp->a;
1505 		b = cvp->b;
1506 		c = cvp->c;
1507 		d = cvp->d;
1508 
1509 		/*
1510 		 * c: both new and old changes
1511 		 * d: only changes in the old file
1512 		 * a: only changes in the new file
1513 		 */
1514 		if (a <= b && c <= d)
1515 			ch = 'c';
1516 		else
1517 			ch = (a <= b) ? 'd' : 'a';
1518 
1519 		switch (ch) {
1520 		case 'c':
1521 			fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1522 			fetch(ixold, a, b, f1, '-', 0, flags);
1523 			fetch(ixnew, c, d, f2, '+', 0, flags);
1524 			break;
1525 		case 'd':
1526 			fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1527 			fetch(ixold, a, b, f1, '-', 0, flags);
1528 			break;
1529 		case 'a':
1530 			fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1531 			fetch(ixnew, c, d, f2, '+', 0, flags);
1532 			break;
1533 		}
1534 		lowa = b + 1;
1535 		lowc = d + 1;
1536 	}
1537 	fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1538 
1539 	context_vec_ptr = context_vec_start - 1;
1540 }
1541 
1542 static void
1543 print_header(const char *file1, const char *file2)
1544 {
1545 	const char *time_format;
1546 	char buf1[256];
1547 	char buf2[256];
1548 	char end1[10];
1549 	char end2[10];
1550 	struct tm tm1, tm2, *tm_ptr1, *tm_ptr2;
1551 	int nsec1 = stb1.st_mtim.tv_nsec;
1552 	int nsec2 = stb2.st_mtim.tv_nsec;
1553 
1554 	time_format = "%Y-%m-%d %H:%M:%S";
1555 
1556 	if (cflag)
1557 		time_format = "%c";
1558 	tm_ptr1 = localtime_r(&stb1.st_mtime, &tm1);
1559 	tm_ptr2 = localtime_r(&stb2.st_mtime, &tm2);
1560 	strftime(buf1, 256, time_format, tm_ptr1);
1561 	strftime(buf2, 256, time_format, tm_ptr2);
1562 	if (!cflag) {
1563 		strftime(end1, 10, "%z", tm_ptr1);
1564 		strftime(end2, 10, "%z", tm_ptr2);
1565 		sprintf(buf1, "%s.%.9d %s", buf1, nsec1, end1);
1566 		sprintf(buf2, "%s.%.9d %s", buf2, nsec2, end2);
1567 	}
1568 	if (label[0] != NULL)
1569 		diff_output("%s %s\n", diff_format == D_CONTEXT ? "***" : "---",
1570 		    label[0]);
1571 	else
1572 		diff_output("%s %s\t%s\n", diff_format == D_CONTEXT ? "***" : "---",
1573 		    file1, buf1);
1574 	if (label[1] != NULL)
1575 		diff_output("%s %s\n", diff_format == D_CONTEXT ? "---" : "+++",
1576 		    label[1]);
1577 	else
1578 		diff_output("%s %s\t%s\n", diff_format == D_CONTEXT ? "---" : "+++",
1579 		    file2, buf2);
1580 }
1581