xref: /freebsd/usr.bin/diff/diffreg.c (revision e9b1dc32c9bd2ebae5f9e140bfa0e0321bc366b5)
1 /*	$OpenBSD: diffreg.c,v 1.91 2016/03/01 20:57:35 natano 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) < 0) {
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) < 0) {
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 (cap_rights_limit(fileno(f1), &rights_ro) < 0
323 		    && errno != ENOSYS)
324 			err(2, "unable to limit rights on: %s", file1);
325 		if (cap_rights_limit(fileno(f2), &rights_ro) < 0 &&
326 		    errno != ENOSYS)
327 			err(2, "unable to limit rights on: %s", file2);
328 		if (fileno(f1) == STDIN_FILENO || fileno(f2) == STDIN_FILENO) {
329 			/* stding has already been limited */
330 			if (caph_limit_stderr() == -1)
331 				err(2, "unable to limit stderr");
332 			if (caph_limit_stdout() == -1)
333 				err(2, "unable to limit stdout");
334 		} else if (caph_limit_stdio() == -1)
335 				err(2, "unable to limit stdio");
336 
337 		caph_cache_catpages();
338 		caph_cache_tzdata();
339 		if (caph_enter() < 0)
340 			err(2, "unable to enter capability mode");
341 	}
342 
343 	switch (files_differ(f1, f2, flags)) {
344 	case 0:
345 		goto closem;
346 	case 1:
347 		break;
348 	default:
349 		/* error */
350 		status |= 2;
351 		goto closem;
352 	}
353 
354 	if ((flags & D_FORCEASCII) == 0 &&
355 	    (!asciifile(f1) || !asciifile(f2))) {
356 		rval = D_BINARY;
357 		status |= 1;
358 		goto closem;
359 	}
360 	prepare(0, f1, stb1.st_size, flags);
361 	prepare(1, f2, stb2.st_size, flags);
362 
363 	prune();
364 	sort(sfile[0], slen[0]);
365 	sort(sfile[1], slen[1]);
366 
367 	member = (int *)file[1];
368 	equiv(sfile[0], slen[0], sfile[1], slen[1], member);
369 	member = xreallocarray(member, slen[1] + 2, sizeof(*member));
370 
371 	class = (int *)file[0];
372 	unsort(sfile[0], slen[0], class);
373 	class = xreallocarray(class, slen[0] + 2, sizeof(*class));
374 
375 	klist = xcalloc(slen[0] + 2, sizeof(*klist));
376 	clen = 0;
377 	clistlen = 100;
378 	clist = xcalloc(clistlen, sizeof(*clist));
379 	i = stone(class, slen[0], member, klist, flags);
380 	free(member);
381 	free(class);
382 
383 	J = xreallocarray(J, len[0] + 2, sizeof(*J));
384 	unravel(klist[i]);
385 	free(clist);
386 	free(klist);
387 
388 	ixold = xreallocarray(ixold, len[0] + 2, sizeof(*ixold));
389 	ixnew = xreallocarray(ixnew, len[1] + 2, sizeof(*ixnew));
390 	check(f1, f2, flags);
391 	output(file1, f1, file2, f2, flags);
392 	if (pr != NULL)
393 		stop_pr(pr);
394 
395 closem:
396 	if (anychange) {
397 		status |= 1;
398 		if (rval == D_SAME)
399 			rval = D_DIFFER;
400 	}
401 	if (f1 != NULL)
402 		fclose(f1);
403 	if (f2 != NULL)
404 		fclose(f2);
405 
406 	return (rval);
407 }
408 
409 /*
410  * Check to see if the given files differ.
411  * Returns 0 if they are the same, 1 if different, and -1 on error.
412  * XXX - could use code from cmp(1) [faster]
413  */
414 static int
415 files_differ(FILE *f1, FILE *f2, int flags)
416 {
417 	char buf1[BUFSIZ], buf2[BUFSIZ];
418 	size_t i, j;
419 
420 	if ((flags & (D_EMPTY1|D_EMPTY2)) || stb1.st_size != stb2.st_size ||
421 	    (stb1.st_mode & S_IFMT) != (stb2.st_mode & S_IFMT))
422 		return (1);
423 	for (;;) {
424 		i = fread(buf1, 1, sizeof(buf1), f1);
425 		j = fread(buf2, 1, sizeof(buf2), f2);
426 		if ((!i && ferror(f1)) || (!j && ferror(f2)))
427 			return (-1);
428 		if (i != j)
429 			return (1);
430 		if (i == 0)
431 			return (0);
432 		if (memcmp(buf1, buf2, i) != 0)
433 			return (1);
434 	}
435 }
436 
437 static FILE *
438 opentemp(const char *f)
439 {
440 	char buf[BUFSIZ], tempfile[PATH_MAX];
441 	ssize_t nread;
442 	int ifd, ofd;
443 
444 	if (strcmp(f, "-") == 0)
445 		ifd = STDIN_FILENO;
446 	else if ((ifd = open(f, O_RDONLY, 0644)) < 0)
447 		return (NULL);
448 
449 	(void)strlcpy(tempfile, _PATH_TMP "/diff.XXXXXXXX", sizeof(tempfile));
450 
451 	if ((ofd = mkstemp(tempfile)) < 0) {
452 		close(ifd);
453 		return (NULL);
454 	}
455 	unlink(tempfile);
456 	while ((nread = read(ifd, buf, BUFSIZ)) > 0) {
457 		if (write(ofd, buf, nread) != nread) {
458 			close(ifd);
459 			close(ofd);
460 			return (NULL);
461 		}
462 	}
463 	close(ifd);
464 	lseek(ofd, (off_t)0, SEEK_SET);
465 	return (fdopen(ofd, "r"));
466 }
467 
468 char *
469 splice(char *dir, char *path)
470 {
471 	char *tail, *buf;
472 	size_t dirlen;
473 
474 	dirlen = strlen(dir);
475 	while (dirlen != 0 && dir[dirlen - 1] == '/')
476 	    dirlen--;
477 	if ((tail = strrchr(path, '/')) == NULL)
478 		tail = path;
479 	else
480 		tail++;
481 	xasprintf(&buf, "%.*s/%s", (int)dirlen, dir, tail);
482 	return (buf);
483 }
484 
485 static void
486 prepare(int i, FILE *fd, size_t filesize, int flags)
487 {
488 	struct line *p;
489 	int h;
490 	size_t sz, j;
491 
492 	rewind(fd);
493 
494 	sz = MIN(filesize, SIZE_MAX) / 25;
495 	if (sz < 100)
496 		sz = 100;
497 
498 	p = xcalloc(sz + 3, sizeof(*p));
499 	for (j = 0; (h = readhash(fd, flags));) {
500 		if (j == sz) {
501 			sz = sz * 3 / 2;
502 			p = xreallocarray(p, sz + 3, sizeof(*p));
503 		}
504 		p[++j].value = h;
505 	}
506 	len[i] = j;
507 	file[i] = p;
508 }
509 
510 static void
511 prune(void)
512 {
513 	int i, j;
514 
515 	for (pref = 0; pref < len[0] && pref < len[1] &&
516 	    file[0][pref + 1].value == file[1][pref + 1].value;
517 	    pref++)
518 		;
519 	for (suff = 0; suff < len[0] - pref && suff < len[1] - pref &&
520 	    file[0][len[0] - suff].value == file[1][len[1] - suff].value;
521 	    suff++)
522 		;
523 	for (j = 0; j < 2; j++) {
524 		sfile[j] = file[j] + pref;
525 		slen[j] = len[j] - pref - suff;
526 		for (i = 0; i <= slen[j]; i++)
527 			sfile[j][i].serial = i;
528 	}
529 }
530 
531 static void
532 equiv(struct line *a, int n, struct line *b, int m, int *c)
533 {
534 	int i, j;
535 
536 	i = j = 1;
537 	while (i <= n && j <= m) {
538 		if (a[i].value < b[j].value)
539 			a[i++].value = 0;
540 		else if (a[i].value == b[j].value)
541 			a[i++].value = j;
542 		else
543 			j++;
544 	}
545 	while (i <= n)
546 		a[i++].value = 0;
547 	b[m + 1].value = 0;
548 	j = 0;
549 	while (++j <= m) {
550 		c[j] = -b[j].serial;
551 		while (b[j + 1].value == b[j].value) {
552 			j++;
553 			c[j] = b[j].serial;
554 		}
555 	}
556 	c[j] = -1;
557 }
558 
559 /* Code taken from ping.c */
560 static int
561 isqrt(int n)
562 {
563 	int y, x = 1;
564 
565 	if (n == 0)
566 		return (0);
567 
568 	do { /* newton was a stinker */
569 		y = x;
570 		x = n / x;
571 		x += y;
572 		x /= 2;
573 	} while ((x - y) > 1 || (x - y) < -1);
574 
575 	return (x);
576 }
577 
578 static int
579 stone(int *a, int n, int *b, int *c, int flags)
580 {
581 	int i, k, y, j, l;
582 	int oldc, tc, oldl, sq;
583 	u_int numtries, bound;
584 
585 	if (flags & D_MINIMAL)
586 		bound = UINT_MAX;
587 	else {
588 		sq = isqrt(n);
589 		bound = MAX(256, sq);
590 	}
591 
592 	k = 0;
593 	c[0] = newcand(0, 0, 0);
594 	for (i = 1; i <= n; i++) {
595 		j = a[i];
596 		if (j == 0)
597 			continue;
598 		y = -b[j];
599 		oldl = 0;
600 		oldc = c[0];
601 		numtries = 0;
602 		do {
603 			if (y <= clist[oldc].y)
604 				continue;
605 			l = search(c, k, y);
606 			if (l != oldl + 1)
607 				oldc = c[l - 1];
608 			if (l <= k) {
609 				if (clist[c[l]].y <= y)
610 					continue;
611 				tc = c[l];
612 				c[l] = newcand(i, y, oldc);
613 				oldc = tc;
614 				oldl = l;
615 				numtries++;
616 			} else {
617 				c[l] = newcand(i, y, oldc);
618 				k++;
619 				break;
620 			}
621 		} while ((y = b[++j]) > 0 && numtries < bound);
622 	}
623 	return (k);
624 }
625 
626 static int
627 newcand(int x, int y, int pred)
628 {
629 	struct cand *q;
630 
631 	if (clen == clistlen) {
632 		clistlen = clistlen * 11 / 10;
633 		clist = xreallocarray(clist, clistlen, sizeof(*clist));
634 	}
635 	q = clist + clen;
636 	q->x = x;
637 	q->y = y;
638 	q->pred = pred;
639 	return (clen++);
640 }
641 
642 static int
643 search(int *c, int k, int y)
644 {
645 	int i, j, l, t;
646 
647 	if (clist[c[k]].y < y)	/* quick look for typical case */
648 		return (k + 1);
649 	i = 0;
650 	j = k + 1;
651 	for (;;) {
652 		l = (i + j) / 2;
653 		if (l <= i)
654 			break;
655 		t = clist[c[l]].y;
656 		if (t > y)
657 			j = l;
658 		else if (t < y)
659 			i = l;
660 		else
661 			return (l);
662 	}
663 	return (l + 1);
664 }
665 
666 static void
667 unravel(int p)
668 {
669 	struct cand *q;
670 	int i;
671 
672 	for (i = 0; i <= len[0]; i++)
673 		J[i] = i <= pref ? i :
674 		    i > len[0] - suff ? i + len[1] - len[0] : 0;
675 	for (q = clist + p; q->y != 0; q = clist + q->pred)
676 		J[q->x + pref] = q->y + pref;
677 }
678 
679 /*
680  * Check does double duty:
681  *  1.	ferret out any fortuitous correspondences due
682  *	to confounding by hashing (which result in "jackpot")
683  *  2.  collect random access indexes to the two files
684  */
685 static void
686 check(FILE *f1, FILE *f2, int flags)
687 {
688 	int i, j, jackpot, c, d;
689 	long ctold, ctnew;
690 
691 	rewind(f1);
692 	rewind(f2);
693 	j = 1;
694 	ixold[0] = ixnew[0] = 0;
695 	jackpot = 0;
696 	ctold = ctnew = 0;
697 	for (i = 1; i <= len[0]; i++) {
698 		if (J[i] == 0) {
699 			ixold[i] = ctold += skipline(f1);
700 			continue;
701 		}
702 		while (j < J[i]) {
703 			ixnew[j] = ctnew += skipline(f2);
704 			j++;
705 		}
706 		if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS|D_IGNORECASE|D_STRIPCR)) {
707 			for (;;) {
708 				c = getc(f1);
709 				d = getc(f2);
710 				/*
711 				 * GNU diff ignores a missing newline
712 				 * in one file for -b or -w.
713 				 */
714 				if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) {
715 					if (c == EOF && d == '\n') {
716 						ctnew++;
717 						break;
718 					} else if (c == '\n' && d == EOF) {
719 						ctold++;
720 						break;
721 					}
722 				}
723 				ctold++;
724 				ctnew++;
725 				if (flags & D_STRIPCR && (c == '\r' || d == '\r')) {
726 					if (c == '\r') {
727 						if ((c = getc(f1)) == '\n') {
728 							ctold++;
729 						} else {
730 							ungetc(c, f1);
731 						}
732 					}
733 					if (d == '\r') {
734 						if ((d = getc(f2)) == '\n') {
735 							ctnew++;
736 						} else {
737 							ungetc(d, f2);
738 						}
739 					}
740 					break;
741 				}
742 				if ((flags & D_FOLDBLANKS) && isspace(c) &&
743 				    isspace(d)) {
744 					do {
745 						if (c == '\n')
746 							break;
747 						ctold++;
748 					} while (isspace(c = getc(f1)));
749 					do {
750 						if (d == '\n')
751 							break;
752 						ctnew++;
753 					} while (isspace(d = getc(f2)));
754 				} else if ((flags & D_IGNOREBLANKS)) {
755 					while (isspace(c) && c != '\n') {
756 						c = getc(f1);
757 						ctold++;
758 					}
759 					while (isspace(d) && d != '\n') {
760 						d = getc(f2);
761 						ctnew++;
762 					}
763 				}
764 				if (chrtran(c) != chrtran(d)) {
765 					jackpot++;
766 					J[i] = 0;
767 					if (c != '\n' && c != EOF)
768 						ctold += skipline(f1);
769 					if (d != '\n' && c != EOF)
770 						ctnew += skipline(f2);
771 					break;
772 				}
773 				if (c == '\n' || c == EOF)
774 					break;
775 			}
776 		} else {
777 			for (;;) {
778 				ctold++;
779 				ctnew++;
780 				if ((c = getc(f1)) != (d = getc(f2))) {
781 					/* jackpot++; */
782 					J[i] = 0;
783 					if (c != '\n' && c != EOF)
784 						ctold += skipline(f1);
785 					if (d != '\n' && c != EOF)
786 						ctnew += skipline(f2);
787 					break;
788 				}
789 				if (c == '\n' || c == EOF)
790 					break;
791 			}
792 		}
793 		ixold[i] = ctold;
794 		ixnew[j] = ctnew;
795 		j++;
796 	}
797 	for (; j <= len[1]; j++) {
798 		ixnew[j] = ctnew += skipline(f2);
799 	}
800 	/*
801 	 * if (jackpot)
802 	 *	fprintf(stderr, "jackpot\n");
803 	 */
804 }
805 
806 /* shellsort CACM #201 */
807 static void
808 sort(struct line *a, int n)
809 {
810 	struct line *ai, *aim, w;
811 	int j, m = 0, k;
812 
813 	if (n == 0)
814 		return;
815 	for (j = 1; j <= n; j *= 2)
816 		m = 2 * j - 1;
817 	for (m /= 2; m != 0; m /= 2) {
818 		k = n - m;
819 		for (j = 1; j <= k; j++) {
820 			for (ai = &a[j]; ai > a; ai -= m) {
821 				aim = &ai[m];
822 				if (aim < ai)
823 					break;	/* wraparound */
824 				if (aim->value > ai[0].value ||
825 				    (aim->value == ai[0].value &&
826 					aim->serial > ai[0].serial))
827 					break;
828 				w.value = ai[0].value;
829 				ai[0].value = aim->value;
830 				aim->value = w.value;
831 				w.serial = ai[0].serial;
832 				ai[0].serial = aim->serial;
833 				aim->serial = w.serial;
834 			}
835 		}
836 	}
837 }
838 
839 static void
840 unsort(struct line *f, int l, int *b)
841 {
842 	int *a, i;
843 
844 	a = xcalloc(l + 1, sizeof(*a));
845 	for (i = 1; i <= l; i++)
846 		a[f[i].serial] = f[i].value;
847 	for (i = 1; i <= l; i++)
848 		b[i] = a[i];
849 	free(a);
850 }
851 
852 static int
853 skipline(FILE *f)
854 {
855 	int i, c;
856 
857 	for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++)
858 		continue;
859 	return (i);
860 }
861 
862 static void
863 output(char *file1, FILE *f1, char *file2, FILE *f2, int flags)
864 {
865 	int m, i0, i1, j0, j1;
866 
867 	rewind(f1);
868 	rewind(f2);
869 	m = len[0];
870 	J[0] = 0;
871 	J[m + 1] = len[1] + 1;
872 	if (diff_format != D_EDIT) {
873 		for (i0 = 1; i0 <= m; i0 = i1 + 1) {
874 			while (i0 <= m && J[i0] == J[i0 - 1] + 1)
875 				i0++;
876 			j0 = J[i0 - 1] + 1;
877 			i1 = i0 - 1;
878 			while (i1 < m && J[i1 + 1] == 0)
879 				i1++;
880 			j1 = J[i1 + 1] - 1;
881 			J[i1] = j1;
882 			change(file1, f1, file2, f2, i0, i1, j0, j1, &flags);
883 		}
884 	} else {
885 		for (i0 = m; i0 >= 1; i0 = i1 - 1) {
886 			while (i0 >= 1 && J[i0] == J[i0 + 1] - 1 && J[i0] != 0)
887 				i0--;
888 			j0 = J[i0 + 1] - 1;
889 			i1 = i0 + 1;
890 			while (i1 > 1 && J[i1 - 1] == 0)
891 				i1--;
892 			j1 = J[i1 - 1] + 1;
893 			J[i1] = j1;
894 			change(file1, f1, file2, f2, i1, i0, j1, j0, &flags);
895 		}
896 	}
897 	if (m == 0)
898 		change(file1, f1, file2, f2, 1, 0, 1, len[1], &flags);
899 	if (diff_format == D_IFDEF || diff_format == D_GFORMAT) {
900 		for (;;) {
901 #define	c i0
902 			if ((c = getc(f1)) == EOF)
903 				return;
904 			diff_output("%c", c);
905 		}
906 #undef c
907 	}
908 	if (anychange != 0) {
909 		if (diff_format == D_CONTEXT)
910 			dump_context_vec(f1, f2, flags);
911 		else if (diff_format == D_UNIFIED)
912 			dump_unified_vec(f1, f2, flags);
913 	}
914 }
915 
916 static void
917 range(int a, int b, const char *separator)
918 {
919 	diff_output("%d", a > b ? b : a);
920 	if (a < b)
921 		diff_output("%s%d", separator, b);
922 }
923 
924 static void
925 uni_range(int a, int b)
926 {
927 	if (a < b)
928 		diff_output("%d,%d", a, b - a + 1);
929 	else if (a == b)
930 		diff_output("%d", b);
931 	else
932 		diff_output("%d,0", b);
933 }
934 
935 static char *
936 preadline(int fd, size_t rlen, off_t off)
937 {
938 	char *line;
939 	ssize_t nr;
940 
941 	line = xmalloc(rlen + 1);
942 	if ((nr = pread(fd, line, rlen, off)) < 0)
943 		err(2, "preadline");
944 	if (nr > 0 && line[nr-1] == '\n')
945 		nr--;
946 	line[nr] = '\0';
947 	return (line);
948 }
949 
950 static bool
951 ignoreline_pattern(char *line)
952 {
953 	int ret;
954 
955 	ret = regexec(&ignore_re, line, 0, NULL, 0);
956 	free(line);
957 	return (ret == 0);	/* if it matched, it should be ignored. */
958 }
959 
960 static bool
961 ignoreline(char *line, bool skip_blanks)
962 {
963 
964 	if (ignore_pats != NULL && skip_blanks)
965 		return (ignoreline_pattern(line) || *line == '\0');
966 	if (ignore_pats != NULL)
967 		return (ignoreline_pattern(line));
968 	if (skip_blanks)
969 		return (*line == '\0');
970 	/* No ignore criteria specified */
971 	return (false);
972 }
973 
974 /*
975  * Indicate that there is a difference between lines a and b of the from file
976  * to get to lines c to d of the to file.  If a is greater then b then there
977  * are no lines in the from file involved and this means that there were
978  * lines appended (beginning at b).  If c is greater than d then there are
979  * lines missing from the to file.
980  */
981 static void
982 change(char *file1, FILE *f1, char *file2, FILE *f2, int a, int b, int c, int d,
983     int *pflags)
984 {
985 	static size_t max_context = 64;
986 	long curpos;
987 	int i, nc, f;
988 	const char *walk;
989 	bool skip_blanks;
990 
991 	skip_blanks = (*pflags & D_SKIPBLANKLINES);
992 restart:
993 	if ((diff_format != D_IFDEF || diff_format == D_GFORMAT) &&
994 	    a > b && c > d)
995 		return;
996 	if (ignore_pats != NULL || skip_blanks) {
997 		char *line;
998 		/*
999 		 * All lines in the change, insert, or delete must
1000 		 * match an ignore pattern for the change to be
1001 		 * ignored.
1002 		 */
1003 		if (a <= b) {		/* Changes and deletes. */
1004 			for (i = a; i <= b; i++) {
1005 				line = preadline(fileno(f1),
1006 				    ixold[i] - ixold[i - 1], ixold[i - 1]);
1007 				if (!ignoreline(line, skip_blanks))
1008 					goto proceed;
1009 			}
1010 		}
1011 		if (a > b || c <= d) {	/* Changes and inserts. */
1012 			for (i = c; i <= d; i++) {
1013 				line = preadline(fileno(f2),
1014 				    ixnew[i] - ixnew[i - 1], ixnew[i - 1]);
1015 				if (!ignoreline(line, skip_blanks))
1016 					goto proceed;
1017 			}
1018 		}
1019 		return;
1020 	}
1021 proceed:
1022 	if (*pflags & D_HEADER && diff_format != D_BRIEF) {
1023 		diff_output("%s %s %s\n", diffargs, file1, file2);
1024 		*pflags &= ~D_HEADER;
1025 	}
1026 	if (diff_format == D_CONTEXT || diff_format == D_UNIFIED) {
1027 		/*
1028 		 * Allocate change records as needed.
1029 		 */
1030 		if (context_vec_ptr == context_vec_end - 1) {
1031 			ptrdiff_t offset = context_vec_ptr - context_vec_start;
1032 			max_context <<= 1;
1033 			context_vec_start = xreallocarray(context_vec_start,
1034 			    max_context, sizeof(*context_vec_start));
1035 			context_vec_end = context_vec_start + max_context;
1036 			context_vec_ptr = context_vec_start + offset;
1037 		}
1038 		if (anychange == 0) {
1039 			/*
1040 			 * Print the context/unidiff header first time through.
1041 			 */
1042 			print_header(file1, file2);
1043 			anychange = 1;
1044 		} else if (a > context_vec_ptr->b + (2 * diff_context) + 1 &&
1045 		    c > context_vec_ptr->d + (2 * diff_context) + 1) {
1046 			/*
1047 			 * If this change is more than 'diff_context' lines from the
1048 			 * previous change, dump the record and reset it.
1049 			 */
1050 			if (diff_format == D_CONTEXT)
1051 				dump_context_vec(f1, f2, *pflags);
1052 			else
1053 				dump_unified_vec(f1, f2, *pflags);
1054 		}
1055 		context_vec_ptr++;
1056 		context_vec_ptr->a = a;
1057 		context_vec_ptr->b = b;
1058 		context_vec_ptr->c = c;
1059 		context_vec_ptr->d = d;
1060 		return;
1061 	}
1062 	if (anychange == 0)
1063 		anychange = 1;
1064 	switch (diff_format) {
1065 	case D_BRIEF:
1066 		return;
1067 	case D_NORMAL:
1068 	case D_EDIT:
1069 		range(a, b, ",");
1070 		diff_output("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1071 		if (diff_format == D_NORMAL)
1072 			range(c, d, ",");
1073 		diff_output("\n");
1074 		break;
1075 	case D_REVERSE:
1076 		diff_output("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1077 		range(a, b, " ");
1078 		diff_output("\n");
1079 		break;
1080 	case D_NREVERSE:
1081 		if (a > b)
1082 			diff_output("a%d %d\n", b, d - c + 1);
1083 		else {
1084 			diff_output("d%d %d\n", a, b - a + 1);
1085 			if (!(c > d))
1086 				/* add changed lines */
1087 				diff_output("a%d %d\n", b, d - c + 1);
1088 		}
1089 		break;
1090 	}
1091 	if (diff_format == D_GFORMAT) {
1092 		curpos = ftell(f1);
1093 		/* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1094 		nc = ixold[a > b ? b : a - 1] - curpos;
1095 		for (i = 0; i < nc; i++)
1096 			diff_output("%c", getc(f1));
1097 		for (walk = group_format; *walk != '\0'; walk++) {
1098 			if (*walk == '%') {
1099 				walk++;
1100 				switch (*walk) {
1101 				case '<':
1102 					fetch(ixold, a, b, f1, '<', 1, *pflags);
1103 					break;
1104 				case '>':
1105 					fetch(ixnew, c, d, f2, '>', 0, *pflags);
1106 					break;
1107 				default:
1108 					diff_output("%%%c", *walk);
1109 					break;
1110 				}
1111 				continue;
1112 			}
1113 			diff_output("%c", *walk);
1114 		}
1115 	}
1116 	if (diff_format == D_NORMAL || diff_format == D_IFDEF) {
1117 		fetch(ixold, a, b, f1, '<', 1, *pflags);
1118 		if (a <= b && c <= d && diff_format == D_NORMAL)
1119 			diff_output("---\n");
1120 	}
1121 	f = 0;
1122 	if (diff_format != D_GFORMAT)
1123 		f = fetch(ixnew, c, d, f2, diff_format == D_NORMAL ? '>' : '\0', 0, *pflags);
1124 	if (f != 0 && diff_format == D_EDIT) {
1125 		/*
1126 		 * A non-zero return value for D_EDIT indicates that the
1127 		 * last line printed was a bare dot (".") that has been
1128 		 * escaped as ".." to prevent ed(1) from misinterpreting
1129 		 * it.  We have to add a substitute command to change this
1130 		 * back and restart where we left off.
1131 		 */
1132 		diff_output(".\n");
1133 		diff_output("%ds/.//\n", a + f - 1);
1134 		b = a + f - 1;
1135 		a = b + 1;
1136 		c += f;
1137 		goto restart;
1138 	}
1139 	if ((diff_format == D_EDIT || diff_format == D_REVERSE) && c <= d)
1140 		diff_output(".\n");
1141 	if (inifdef) {
1142 		diff_output("#endif /* %s */\n", ifdefname);
1143 		inifdef = 0;
1144 	}
1145 }
1146 
1147 static int
1148 fetch(long *f, int a, int b, FILE *lb, int ch, int oldfile, int flags)
1149 {
1150 	int i, j, c, lastc, col, nc;
1151 	int	newcol;
1152 
1153 	/*
1154 	 * When doing #ifdef's, copy down to current line
1155 	 * if this is the first file, so that stuff makes it to output.
1156 	 */
1157 	if ((diff_format == D_IFDEF) && oldfile) {
1158 		long curpos = ftell(lb);
1159 		/* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1160 		nc = f[a > b ? b : a - 1] - curpos;
1161 		for (i = 0; i < nc; i++)
1162 			diff_output("%c", getc(lb));
1163 	}
1164 	if (a > b)
1165 		return (0);
1166 	if (diff_format == D_IFDEF) {
1167 		if (inifdef) {
1168 			diff_output("#else /* %s%s */\n",
1169 			    oldfile == 1 ? "!" : "", ifdefname);
1170 		} else {
1171 			if (oldfile)
1172 				diff_output("#ifndef %s\n", ifdefname);
1173 			else
1174 				diff_output("#ifdef %s\n", ifdefname);
1175 		}
1176 		inifdef = 1 + oldfile;
1177 	}
1178 	for (i = a; i <= b; i++) {
1179 		fseek(lb, f[i - 1], SEEK_SET);
1180 		nc = f[i] - f[i - 1];
1181 		if ((diff_format != D_IFDEF && diff_format != D_GFORMAT) &&
1182 		    ch != '\0') {
1183 			diff_output("%c", ch);
1184 			if (Tflag && (diff_format == D_NORMAL || diff_format == D_CONTEXT
1185 			    || diff_format == D_UNIFIED))
1186 				diff_output("\t");
1187 			else if (diff_format != D_UNIFIED)
1188 				diff_output(" ");
1189 		}
1190 		col = 0;
1191 		for (j = 0, lastc = '\0'; j < nc; j++, lastc = c) {
1192 			if ((c = getc(lb)) == EOF) {
1193 				if (diff_format == D_EDIT || diff_format == D_REVERSE ||
1194 				    diff_format == D_NREVERSE)
1195 					warnx("No newline at end of file");
1196 				else
1197 					diff_output("\n\\ No newline at end of "
1198 					    "file\n");
1199 				return (0);
1200 			}
1201 			if (c == '\t' && (flags & D_EXPANDTABS)) {
1202 				newcol = ((col/tabsize)+1)*tabsize;
1203 				do {
1204 					diff_output(" ");
1205 				} while (++col < newcol);
1206 			} else {
1207 				if (diff_format == D_EDIT && j == 1 && c == '\n'
1208 				    && lastc == '.') {
1209 					/*
1210 					 * Don't print a bare "." line
1211 					 * since that will confuse ed(1).
1212 					 * Print ".." instead and return,
1213 					 * giving the caller an offset
1214 					 * from which to restart.
1215 					 */
1216 					diff_output(".\n");
1217 					return (i - a + 1);
1218 				}
1219 				diff_output("%c", c);
1220 				col++;
1221 			}
1222 		}
1223 	}
1224 	return (0);
1225 }
1226 
1227 /*
1228  * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578.
1229  */
1230 static int
1231 readhash(FILE *f, int flags)
1232 {
1233 	int i, t, space;
1234 	int sum;
1235 
1236 	sum = 1;
1237 	space = 0;
1238 	if ((flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) == 0) {
1239 		if (flags & D_IGNORECASE)
1240 			for (i = 0; (t = getc(f)) != '\n'; i++) {
1241 				if (flags & D_STRIPCR && t == '\r') {
1242 					t = getc(f);
1243 					if (t == '\n')
1244 						break;
1245 					ungetc(t, f);
1246 				}
1247 				if (t == EOF) {
1248 					if (i == 0)
1249 						return (0);
1250 					break;
1251 				}
1252 				sum = sum * 127 + chrtran(t);
1253 			}
1254 		else
1255 			for (i = 0; (t = getc(f)) != '\n'; i++) {
1256 				if (flags & D_STRIPCR && t == '\r') {
1257 					t = getc(f);
1258 					if (t == '\n')
1259 						break;
1260 					ungetc(t, f);
1261 				}
1262 				if (t == EOF) {
1263 					if (i == 0)
1264 						return (0);
1265 					break;
1266 				}
1267 				sum = sum * 127 + t;
1268 			}
1269 	} else {
1270 		for (i = 0;;) {
1271 			switch (t = getc(f)) {
1272 			case '\r':
1273 			case '\t':
1274 			case '\v':
1275 			case '\f':
1276 			case ' ':
1277 				space++;
1278 				continue;
1279 			default:
1280 				if (space && (flags & D_IGNOREBLANKS) == 0) {
1281 					i++;
1282 					space = 0;
1283 				}
1284 				sum = sum * 127 + chrtran(t);
1285 				i++;
1286 				continue;
1287 			case EOF:
1288 				if (i == 0)
1289 					return (0);
1290 				/* FALLTHROUGH */
1291 			case '\n':
1292 				break;
1293 			}
1294 			break;
1295 		}
1296 	}
1297 	/*
1298 	 * There is a remote possibility that we end up with a zero sum.
1299 	 * Zero is used as an EOF marker, so return 1 instead.
1300 	 */
1301 	return (sum == 0 ? 1 : sum);
1302 }
1303 
1304 static int
1305 asciifile(FILE *f)
1306 {
1307 	unsigned char buf[BUFSIZ];
1308 	size_t cnt;
1309 
1310 	if (f == NULL)
1311 		return (1);
1312 
1313 	rewind(f);
1314 	cnt = fread(buf, 1, sizeof(buf), f);
1315 	return (memchr(buf, '\0', cnt) == NULL);
1316 }
1317 
1318 #define begins_with(s, pre) (strncmp(s, pre, sizeof(pre)-1) == 0)
1319 
1320 static char *
1321 match_function(const long *f, int pos, FILE *fp)
1322 {
1323 	unsigned char buf[FUNCTION_CONTEXT_SIZE];
1324 	size_t nc;
1325 	int last = lastline;
1326 	const char *state = NULL;
1327 
1328 	lastline = pos;
1329 	while (pos > last) {
1330 		fseek(fp, f[pos - 1], SEEK_SET);
1331 		nc = f[pos] - f[pos - 1];
1332 		if (nc >= sizeof(buf))
1333 			nc = sizeof(buf) - 1;
1334 		nc = fread(buf, 1, nc, fp);
1335 		if (nc > 0) {
1336 			buf[nc] = '\0';
1337 			buf[strcspn(buf, "\n")] = '\0';
1338 			if (isalpha(buf[0]) || buf[0] == '_' || buf[0] == '$') {
1339 				if (begins_with(buf, "private:")) {
1340 					if (!state)
1341 						state = " (private)";
1342 				} else if (begins_with(buf, "protected:")) {
1343 					if (!state)
1344 						state = " (protected)";
1345 				} else if (begins_with(buf, "public:")) {
1346 					if (!state)
1347 						state = " (public)";
1348 				} else {
1349 					strlcpy(lastbuf, buf, sizeof lastbuf);
1350 					if (state)
1351 						strlcat(lastbuf, state,
1352 						    sizeof lastbuf);
1353 					lastmatchline = pos;
1354 					return lastbuf;
1355 				}
1356 			}
1357 		}
1358 		pos--;
1359 	}
1360 	return lastmatchline > 0 ? lastbuf : NULL;
1361 }
1362 
1363 /* dump accumulated "context" diff changes */
1364 static void
1365 dump_context_vec(FILE *f1, FILE *f2, int flags)
1366 {
1367 	struct context_vec *cvp = context_vec_start;
1368 	int lowa, upb, lowc, upd, do_output;
1369 	int a, b, c, d;
1370 	char ch, *f;
1371 
1372 	if (context_vec_start > context_vec_ptr)
1373 		return;
1374 
1375 	b = d = 0;		/* gcc */
1376 	lowa = MAX(1, cvp->a - diff_context);
1377 	upb = MIN(len[0], context_vec_ptr->b + diff_context);
1378 	lowc = MAX(1, cvp->c - diff_context);
1379 	upd = MIN(len[1], context_vec_ptr->d + diff_context);
1380 
1381 	diff_output("***************");
1382 	if ((flags & D_PROTOTYPE)) {
1383 		f = match_function(ixold, lowa-1, f1);
1384 		if (f != NULL)
1385 			diff_output(" %s", f);
1386 	}
1387 	diff_output("\n*** ");
1388 	range(lowa, upb, ",");
1389 	diff_output(" ****\n");
1390 
1391 	/*
1392 	 * Output changes to the "old" file.  The first loop suppresses
1393 	 * output if there were no changes to the "old" file (we'll see
1394 	 * the "old" lines as context in the "new" list).
1395 	 */
1396 	do_output = 0;
1397 	for (; cvp <= context_vec_ptr; cvp++)
1398 		if (cvp->a <= cvp->b) {
1399 			cvp = context_vec_start;
1400 			do_output++;
1401 			break;
1402 		}
1403 	if (do_output) {
1404 		while (cvp <= context_vec_ptr) {
1405 			a = cvp->a;
1406 			b = cvp->b;
1407 			c = cvp->c;
1408 			d = cvp->d;
1409 
1410 			if (a <= b && c <= d)
1411 				ch = 'c';
1412 			else
1413 				ch = (a <= b) ? 'd' : 'a';
1414 
1415 			if (ch == 'a')
1416 				fetch(ixold, lowa, b, f1, ' ', 0, flags);
1417 			else {
1418 				fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1419 				fetch(ixold, a, b, f1,
1420 				    ch == 'c' ? '!' : '-', 0, flags);
1421 			}
1422 			lowa = b + 1;
1423 			cvp++;
1424 		}
1425 		fetch(ixold, b + 1, upb, f1, ' ', 0, flags);
1426 	}
1427 	/* output changes to the "new" file */
1428 	diff_output("--- ");
1429 	range(lowc, upd, ",");
1430 	diff_output(" ----\n");
1431 
1432 	do_output = 0;
1433 	for (cvp = context_vec_start; cvp <= context_vec_ptr; cvp++)
1434 		if (cvp->c <= cvp->d) {
1435 			cvp = context_vec_start;
1436 			do_output++;
1437 			break;
1438 		}
1439 	if (do_output) {
1440 		while (cvp <= context_vec_ptr) {
1441 			a = cvp->a;
1442 			b = cvp->b;
1443 			c = cvp->c;
1444 			d = cvp->d;
1445 
1446 			if (a <= b && c <= d)
1447 				ch = 'c';
1448 			else
1449 				ch = (a <= b) ? 'd' : 'a';
1450 
1451 			if (ch == 'd')
1452 				fetch(ixnew, lowc, d, f2, ' ', 0, flags);
1453 			else {
1454 				fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1455 				fetch(ixnew, c, d, f2,
1456 				    ch == 'c' ? '!' : '+', 0, flags);
1457 			}
1458 			lowc = d + 1;
1459 			cvp++;
1460 		}
1461 		fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1462 	}
1463 	context_vec_ptr = context_vec_start - 1;
1464 }
1465 
1466 /* dump accumulated "unified" diff changes */
1467 static void
1468 dump_unified_vec(FILE *f1, FILE *f2, int flags)
1469 {
1470 	struct context_vec *cvp = context_vec_start;
1471 	int lowa, upb, lowc, upd;
1472 	int a, b, c, d;
1473 	char ch, *f;
1474 
1475 	if (context_vec_start > context_vec_ptr)
1476 		return;
1477 
1478 	b = d = 0;		/* gcc */
1479 	lowa = MAX(1, cvp->a - diff_context);
1480 	upb = MIN(len[0], context_vec_ptr->b + diff_context);
1481 	lowc = MAX(1, cvp->c - diff_context);
1482 	upd = MIN(len[1], context_vec_ptr->d + diff_context);
1483 
1484 	diff_output("@@ -");
1485 	uni_range(lowa, upb);
1486 	diff_output(" +");
1487 	uni_range(lowc, upd);
1488 	diff_output(" @@");
1489 	if ((flags & D_PROTOTYPE)) {
1490 		f = match_function(ixold, lowa-1, f1);
1491 		if (f != NULL)
1492 			diff_output(" %s", f);
1493 	}
1494 	diff_output("\n");
1495 
1496 	/*
1497 	 * Output changes in "unified" diff format--the old and new lines
1498 	 * are printed together.
1499 	 */
1500 	for (; cvp <= context_vec_ptr; cvp++) {
1501 		a = cvp->a;
1502 		b = cvp->b;
1503 		c = cvp->c;
1504 		d = cvp->d;
1505 
1506 		/*
1507 		 * c: both new and old changes
1508 		 * d: only changes in the old file
1509 		 * a: only changes in the new file
1510 		 */
1511 		if (a <= b && c <= d)
1512 			ch = 'c';
1513 		else
1514 			ch = (a <= b) ? 'd' : 'a';
1515 
1516 		switch (ch) {
1517 		case 'c':
1518 			fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1519 			fetch(ixold, a, b, f1, '-', 0, flags);
1520 			fetch(ixnew, c, d, f2, '+', 0, flags);
1521 			break;
1522 		case 'd':
1523 			fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1524 			fetch(ixold, a, b, f1, '-', 0, flags);
1525 			break;
1526 		case 'a':
1527 			fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1528 			fetch(ixnew, c, d, f2, '+', 0, flags);
1529 			break;
1530 		}
1531 		lowa = b + 1;
1532 		lowc = d + 1;
1533 	}
1534 	fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1535 
1536 	context_vec_ptr = context_vec_start - 1;
1537 }
1538 
1539 static void
1540 print_header(const char *file1, const char *file2)
1541 {
1542 	const char *time_format;
1543 	char buf1[256];
1544 	char buf2[256];
1545 	char end1[10];
1546 	char end2[10];
1547 	struct tm tm1, tm2, *tm_ptr1, *tm_ptr2;
1548 	int nsec1 = stb1.st_mtim.tv_nsec;
1549 	int nsec2 = stb2.st_mtim.tv_nsec;
1550 
1551 	time_format = "%Y-%m-%d %H:%M:%S";
1552 
1553 	if (cflag)
1554 		time_format = "%c";
1555 	tm_ptr1 = localtime_r(&stb1.st_mtime, &tm1);
1556 	tm_ptr2 = localtime_r(&stb2.st_mtime, &tm2);
1557 	strftime(buf1, 256, time_format, tm_ptr1);
1558 	strftime(buf2, 256, time_format, tm_ptr2);
1559 	if (!cflag) {
1560 		strftime(end1, 10, "%z", tm_ptr1);
1561 		strftime(end2, 10, "%z", tm_ptr2);
1562 		sprintf(buf1, "%s.%.9d %s", buf1, nsec1, end1);
1563 		sprintf(buf2, "%s.%.9d %s", buf2, nsec2, end2);
1564 	}
1565 	if (label[0] != NULL)
1566 		diff_output("%s %s\n", diff_format == D_CONTEXT ? "***" : "---",
1567 		    label[0]);
1568 	else
1569 		diff_output("%s %s\t%s\n", diff_format == D_CONTEXT ? "***" : "---",
1570 		    file1, buf1);
1571 	if (label[1] != NULL)
1572 		diff_output("%s %s\n", diff_format == D_CONTEXT ? "---" : "+++",
1573 		    label[1]);
1574 	else
1575 		diff_output("%s %s\t%s\n", diff_format == D_CONTEXT ? "---" : "+++",
1576 		    file2, buf2);
1577 }
1578