xref: /freebsd/usr.bin/diff/diffreg.c (revision d37eb51047221dc3322b34db1038ff3aa533883f)
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 
67 #include <sys/capsicum.h>
68 #include <sys/stat.h>
69 
70 #include <capsicum_helpers.h>
71 #include <ctype.h>
72 #include <err.h>
73 #include <errno.h>
74 #include <fcntl.h>
75 #include <limits.h>
76 #include <math.h>
77 #include <paths.h>
78 #include <regex.h>
79 #include <stdbool.h>
80 #include <stddef.h>
81 #include <stdint.h>
82 #include <stdio.h>
83 #include <stdlib.h>
84 #include <string.h>
85 
86 #include "pr.h"
87 #include "diff.h"
88 #include "xmalloc.h"
89 
90 /*
91  * diff - compare two files.
92  */
93 
94 /*
95  *	Uses an algorithm due to Harold Stone, which finds a pair of longest
96  *	identical subsequences in the two files.
97  *
98  *	The major goal is to generate the match vector J. J[i] is the index of
99  *	the line in file1 corresponding to line i file0. J[i] = 0 if there is no
100  *	such line in file1.
101  *
102  *	Lines are hashed so as to work in core. All potential matches are
103  *	located by sorting the lines of each file on the hash (called
104  *	``value''). In particular, this collects the equivalence classes in
105  *	file1 together. Subroutine equiv replaces the value of each line in
106  *	file0 by the index of the first element of its matching equivalence in
107  *	(the reordered) file1. To save space equiv squeezes file1 into a single
108  *	array member in which the equivalence classes are simply concatenated,
109  *	except that their first members are flagged by changing sign.
110  *
111  *	Next the indices that point into member are unsorted into array class
112  *	according to the original order of file0.
113  *
114  *	The cleverness lies in routine stone. This marches through the lines of
115  *	file0, developing a vector klist of "k-candidates". At step i
116  *	a k-candidate is a matched pair of lines x,y (x in file0 y in file1)
117  *	such that there is a common subsequence of length k between the first
118  *	i lines of file0 and the first y lines of file1, but there is no such
119  *	subsequence for any smaller y. x is the earliest possible mate to y that
120  *	occurs in such a subsequence.
121  *
122  *	Whenever any of the members of the equivalence class of lines in file1
123  *	matable to a line in file0 has serial number less than the y of some
124  *	k-candidate, that k-candidate with the smallest such y is replaced. The
125  *	new k-candidate is chained (via pred) to the current k-1 candidate so
126  *	that the actual subsequence can be recovered. When a member has serial
127  *	number greater that the y of all k-candidates, the klist is extended. At
128  *	the end, the longest subsequence is pulled out and placed in the array J
129  *	by unravel.
130  *
131  *	With J in hand, the matches there recorded are check'ed against reality
132  *	to assure that no spurious matches have crept in due to hashing. If they
133  *	have, they are broken, and "jackpot" is recorded -- a harmless matter
134  *	except that a true match for a spuriously mated line may now be
135  *	unnecessarily reported as a change.
136  *
137  *	Much of the complexity of the program comes simply from trying to
138  *	minimize core utilization and maximize the range of doable problems by
139  *	dynamically allocating what is needed and reusing what is not. The core
140  *	requirements for problems larger than somewhat are (in words)
141  *	2*length(file0) + length(file1) + 3*(number of k-candidates installed),
142  *	typically about 6n words for files of length n.
143  */
144 
145 struct cand {
146 	int	x;
147 	int	y;
148 	int	pred;
149 };
150 
151 static struct line {
152 	int	serial;
153 	int	value;
154 } *file[2];
155 
156 /*
157  * The following struct is used to record change information when
158  * doing a "context" or "unified" diff.  (see routine "change" to
159  * understand the highly mnemonic field names)
160  */
161 struct context_vec {
162 	int	a;		/* start line in old file */
163 	int	b;		/* end line in old file */
164 	int	c;		/* start line in new file */
165 	int	d;		/* end line in new file */
166 };
167 
168 enum readhash { RH_BINARY, RH_OK, RH_EOF };
169 
170 static int	 diffreg_stone(char *, char *, int, int);
171 static FILE	*opentemp(const char *);
172 static void	 output(char *, FILE *, char *, FILE *, int);
173 static void	 check(FILE *, FILE *, int);
174 static void	 range(int, int, const char *);
175 static void	 uni_range(int, int);
176 static void	 dump_context_vec(FILE *, FILE *, int);
177 static void	 dump_unified_vec(FILE *, FILE *, int);
178 static bool	 prepare(int, FILE *, size_t, int);
179 static void	 prune(void);
180 static void	 equiv(struct line *, int, struct line *, int, int *);
181 static void	 unravel(int);
182 static void	 unsort(struct line *, int, int *);
183 static void	 change(char *, FILE *, char *, FILE *, int, int, int, int, int *);
184 static void	 sort(struct line *, int);
185 static void	 print_header(const char *, const char *);
186 static void	 print_space(int, int, int);
187 static bool	 ignoreline_pattern(char *);
188 static bool	 ignoreline(char *, bool);
189 static int	 asciifile(FILE *);
190 static int	 fetch(long *, int, int, FILE *, int, int, int);
191 static int	 newcand(int, int, int);
192 static int	 search(int *, int, int);
193 static int	 skipline(FILE *);
194 static int	 stone(int *, int, int *, int *, int);
195 static enum readhash readhash(FILE *, int, unsigned *);
196 static int	 files_differ(FILE *, FILE *, int);
197 static char	*match_function(const long *, int, FILE *);
198 static char	*preadline(int, size_t, off_t);
199 
200 static int	 *J;			/* will be overlaid on class */
201 static int	 *class;		/* will be overlaid on file[0] */
202 static int	 *klist;		/* will be overlaid on file[0] after class */
203 static int	 *member;		/* will be overlaid on file[1] */
204 static int	 clen;
205 static int	 inifdef;		/* whether or not we are in a #ifdef block */
206 static int	 len[2];
207 static int	 pref, suff;	/* length of prefix and suffix */
208 static int	 slen[2];
209 static int	 anychange;
210 static int	 hw, lpad,rpad;		/* half width and padding */
211 static int	 edoffset;
212 static long	*ixnew;		/* will be overlaid on file[1] */
213 static long	*ixold;		/* will be overlaid on klist */
214 static struct cand *clist;	/* merely a free storage pot for candidates */
215 static int	 clistlen;		/* the length of clist */
216 static struct line *sfile[2];	/* shortened by pruning common prefix/suffix */
217 static int	(*chrtran)(int);	/* translation table for case-folding */
218 static struct context_vec *context_vec_start;
219 static struct context_vec *context_vec_end;
220 static struct context_vec *context_vec_ptr;
221 
222 #define FUNCTION_CONTEXT_SIZE	55
223 static char lastbuf[FUNCTION_CONTEXT_SIZE];
224 static int lastline;
225 static int lastmatchline;
226 
227 int
228 diffreg(char *file1, char *file2, int flags, int capsicum)
229 {
230 	/*
231 	 * If we have set the algorithm with -A or --algorithm use that if we
232 	 * can and if not print an error.
233 	 */
234 	if (diff_algorithm_set) {
235 		if (diff_algorithm == D_DIFFMYERS ||
236 		    diff_algorithm == D_DIFFPATIENCE) {
237 			if (can_libdiff(flags))
238 				return diffreg_new(file1, file2, flags, capsicum);
239 			else
240 				errx(2, "cannot use Myers algorithm with selected options");
241 		} else {
242 			/* Fallback to using stone. */
243 			return diffreg_stone(file1, file2, flags, capsicum);
244 		}
245 	} else {
246 		if (can_libdiff(flags))
247 			return diffreg_new(file1, file2, flags, capsicum);
248 		else
249 			return diffreg_stone(file1, file2, flags, capsicum);
250 	}
251 }
252 
253 static int
254 clow2low(int c)
255 {
256 
257 	return (c);
258 }
259 
260 static int
261 cup2low(int c)
262 {
263 
264 	return (tolower(c));
265 }
266 
267 int
268 diffreg_stone(char *file1, char *file2, int flags, int capsicum)
269 {
270 	FILE *f1, *f2;
271 	int i, rval;
272 	struct pr *pr = NULL;
273 	cap_rights_t rights_ro;
274 
275 	f1 = f2 = NULL;
276 	rval = D_SAME;
277 	anychange = 0;
278 	lastline = 0;
279 	lastmatchline = 0;
280 
281 	/*
282 	 * In side-by-side mode, we need to print the left column, a
283 	 * change marker surrounded by padding, and the right column.
284 	 *
285 	 * If expanding tabs, we don't care about alignment, so we simply
286 	 * subtract 3 from the width and divide by two.
287 	 *
288 	 * If not expanding tabs, we need to ensure that the right column
289 	 * is aligned to a tab stop.  We start with the same formula, then
290 	 * decrement until we reach a size that lets us tab-align the
291 	 * right column.  We then adjust the width down if necessary for
292 	 * the padding calculation to work.
293 	 *
294 	 * Left padding is half the space left over, rounded down; right
295 	 * padding is whatever is needed to match the width.
296 	 */
297 	if (diff_format == D_SIDEBYSIDE) {
298 		if (flags & D_EXPANDTABS) {
299 			if (width > 3) {
300 				hw = (width - 3) / 2;
301 			} else {
302 				/* not enough space */
303 				hw = 0;
304 			}
305 		} else if (width <= 3 || width <= tabsize) {
306 			/* not enough space */
307 			hw = 0;
308 		} else {
309 			hw = (width - 3) / 2;
310 			while (hw > 0 && roundup(hw + 3, tabsize) + hw > width)
311 				hw--;
312 			if (width - (roundup(hw + 3, tabsize) + hw) < tabsize)
313 				width = roundup(hw + 3, tabsize) + hw;
314 		}
315 		lpad = (width - hw * 2 - 1) / 2;
316 		rpad = (width - hw * 2 - 1) - lpad;
317 	}
318 
319 	if (flags & D_IGNORECASE)
320 		chrtran = cup2low;
321 	else
322 		chrtran = clow2low;
323 	if (S_ISDIR(stb1.st_mode) != S_ISDIR(stb2.st_mode))
324 		return (S_ISDIR(stb1.st_mode) ? D_MISMATCH1 : D_MISMATCH2);
325 	if (strcmp(file1, "-") == 0 && strcmp(file2, "-") == 0)
326 		goto closem;
327 
328 	if (flags & D_EMPTY1)
329 		f1 = fopen(_PATH_DEVNULL, "r");
330 	else {
331 		if (!S_ISREG(stb1.st_mode)) {
332 			if ((f1 = opentemp(file1)) == NULL ||
333 			    fstat(fileno(f1), &stb1) == -1) {
334 				warn("%s", file1);
335 				rval = D_ERROR;
336 				status |= 2;
337 				goto closem;
338 			}
339 		} else if (strcmp(file1, "-") == 0)
340 			f1 = stdin;
341 		else
342 			f1 = fopen(file1, "r");
343 	}
344 	if (f1 == NULL) {
345 		warn("%s", file1);
346 		rval = D_ERROR;
347 		status |= 2;
348 		goto closem;
349 	}
350 
351 	if (flags & D_EMPTY2)
352 		f2 = fopen(_PATH_DEVNULL, "r");
353 	else {
354 		if (!S_ISREG(stb2.st_mode)) {
355 			if ((f2 = opentemp(file2)) == NULL ||
356 			    fstat(fileno(f2), &stb2) == -1) {
357 				warn("%s", file2);
358 				rval = D_ERROR;
359 				status |= 2;
360 				goto closem;
361 			}
362 		} else if (strcmp(file2, "-") == 0)
363 			f2 = stdin;
364 		else
365 			f2 = fopen(file2, "r");
366 	}
367 	if (f2 == NULL) {
368 		warn("%s", file2);
369 		rval = D_ERROR;
370 		status |= 2;
371 		goto closem;
372 	}
373 
374 	if (lflag)
375 		pr = start_pr(file1, file2);
376 
377 	if (capsicum) {
378 		cap_rights_init(&rights_ro, CAP_READ, CAP_FSTAT, CAP_SEEK);
379 		if (caph_rights_limit(fileno(f1), &rights_ro) < 0)
380 			err(2, "unable to limit rights on: %s", file1);
381 		if (caph_rights_limit(fileno(f2), &rights_ro) < 0)
382 			err(2, "unable to limit rights on: %s", file2);
383 		if (fileno(f1) == STDIN_FILENO || fileno(f2) == STDIN_FILENO) {
384 			/* stdin has already been limited */
385 			if (caph_limit_stderr() == -1)
386 				err(2, "unable to limit stderr");
387 			if (caph_limit_stdout() == -1)
388 				err(2, "unable to limit stdout");
389 		} else if (caph_limit_stdio() == -1)
390 				err(2, "unable to limit stdio");
391 
392 		caph_cache_catpages();
393 		caph_cache_tzdata();
394 		if (caph_enter() < 0)
395 			err(2, "unable to enter capability mode");
396 	}
397 
398 	switch (files_differ(f1, f2, flags)) {
399 	case 0:
400 		goto closem;
401 	case 1:
402 		break;
403 	default:
404 		/* error */
405 		rval = D_ERROR;
406 		status |= 2;
407 		goto closem;
408 	}
409 
410 	if (diff_format == D_BRIEF && ignore_pats == NULL &&
411 	    (flags & (D_FOLDBLANKS|D_IGNOREBLANKS|D_IGNORECASE|
412 	    D_SKIPBLANKLINES|D_STRIPCR)) == 0)
413 	{
414 		rval = D_DIFFER;
415 		status |= 1;
416 		goto closem;
417 	}
418 	if ((flags & D_FORCEASCII) != 0) {
419 		(void)prepare(0, f1, stb1.st_size, flags);
420 		(void)prepare(1, f2, stb2.st_size, flags);
421 	} else if (!asciifile(f1) || !asciifile(f2) ||
422 		    !prepare(0, f1, stb1.st_size, flags) ||
423 		    !prepare(1, f2, stb2.st_size, flags)) {
424 		rval = D_BINARY;
425 		status |= 1;
426 		goto closem;
427 	}
428 
429 	prune();
430 	sort(sfile[0], slen[0]);
431 	sort(sfile[1], slen[1]);
432 
433 	member = (int *)file[1];
434 	equiv(sfile[0], slen[0], sfile[1], slen[1], member);
435 	member = xreallocarray(member, slen[1] + 2, sizeof(*member));
436 
437 	class = (int *)file[0];
438 	unsort(sfile[0], slen[0], class);
439 	class = xreallocarray(class, slen[0] + 2, sizeof(*class));
440 
441 	klist = xcalloc(slen[0] + 2, sizeof(*klist));
442 	clen = 0;
443 	clistlen = 100;
444 	clist = xcalloc(clistlen, sizeof(*clist));
445 	i = stone(class, slen[0], member, klist, flags);
446 	free(member);
447 	free(class);
448 
449 	J = xreallocarray(J, len[0] + 2, sizeof(*J));
450 	unravel(klist[i]);
451 	free(clist);
452 	free(klist);
453 
454 	ixold = xreallocarray(ixold, len[0] + 2, sizeof(*ixold));
455 	ixnew = xreallocarray(ixnew, len[1] + 2, sizeof(*ixnew));
456 	check(f1, f2, flags);
457 	output(file1, f1, file2, f2, flags);
458 
459 closem:
460 	if (pr != NULL)
461 		stop_pr(pr);
462 	if (anychange) {
463 		status |= 1;
464 		if (rval == D_SAME)
465 			rval = D_DIFFER;
466 	}
467 	if (f1 != NULL)
468 		fclose(f1);
469 	if (f2 != NULL)
470 		fclose(f2);
471 
472 	return (rval);
473 }
474 
475 /*
476  * Check to see if the given files differ.
477  * Returns 0 if they are the same, 1 if different, and -1 on error.
478  * XXX - could use code from cmp(1) [faster]
479  */
480 static int
481 files_differ(FILE *f1, FILE *f2, int flags)
482 {
483 	char buf1[BUFSIZ], buf2[BUFSIZ];
484 	size_t i, j;
485 
486 	if ((flags & (D_EMPTY1|D_EMPTY2)) || stb1.st_size != stb2.st_size ||
487 	    (stb1.st_mode & S_IFMT) != (stb2.st_mode & S_IFMT))
488 		return (1);
489 
490 	if (stb1.st_dev == stb2.st_dev && stb1.st_ino == stb2.st_ino)
491 		return (0);
492 
493 	for (;;) {
494 		i = fread(buf1, 1, sizeof(buf1), f1);
495 		j = fread(buf2, 1, sizeof(buf2), f2);
496 		if ((!i && ferror(f1)) || (!j && ferror(f2)))
497 			return (-1);
498 		if (i != j)
499 			return (1);
500 		if (i == 0)
501 			return (0);
502 		if (memcmp(buf1, buf2, i) != 0)
503 			return (1);
504 	}
505 }
506 
507 static FILE *
508 opentemp(const char *f)
509 {
510 	char buf[BUFSIZ], tempfile[PATH_MAX];
511 	ssize_t nread;
512 	int ifd, ofd;
513 
514 	if (strcmp(f, "-") == 0)
515 		ifd = STDIN_FILENO;
516 	else if ((ifd = open(f, O_RDONLY, 0644)) == -1)
517 		return (NULL);
518 
519 	(void)strlcpy(tempfile, _PATH_TMP "/diff.XXXXXXXX", sizeof(tempfile));
520 
521 	if ((ofd = mkstemp(tempfile)) == -1) {
522 		close(ifd);
523 		return (NULL);
524 	}
525 	unlink(tempfile);
526 	while ((nread = read(ifd, buf, BUFSIZ)) > 0) {
527 		if (write(ofd, buf, nread) != nread) {
528 			close(ifd);
529 			close(ofd);
530 			return (NULL);
531 		}
532 	}
533 	close(ifd);
534 	lseek(ofd, (off_t)0, SEEK_SET);
535 	return (fdopen(ofd, "r"));
536 }
537 
538 static bool
539 prepare(int i, FILE *fd, size_t filesize, int flags)
540 {
541 	struct line *p;
542 	unsigned h;
543 	size_t sz, j = 0;
544 	enum readhash r;
545 
546 	rewind(fd);
547 
548 	sz = MIN(filesize, SIZE_MAX) / 25;
549 	if (sz < 100)
550 		sz = 100;
551 
552 	p = xcalloc(sz + 3, sizeof(*p));
553 	while ((r = readhash(fd, flags, &h)) != RH_EOF)
554 		switch (r) {
555 		case RH_EOF: /* otherwise clang complains */
556 		case RH_BINARY:
557 			return (false);
558 		case RH_OK:
559 			if (j == sz) {
560 				sz = sz * 3 / 2;
561 				p = xreallocarray(p, sz + 3, sizeof(*p));
562 			}
563 			p[++j].value = h;
564 		}
565 
566 	len[i] = j;
567 	file[i] = p;
568 
569 	return (true);
570 }
571 
572 static void
573 prune(void)
574 {
575 	int i, j;
576 
577 	for (pref = 0; pref < len[0] && pref < len[1] &&
578 	    file[0][pref + 1].value == file[1][pref + 1].value;
579 	    pref++)
580 		;
581 	for (suff = 0; suff < len[0] - pref && suff < len[1] - pref &&
582 	    file[0][len[0] - suff].value == file[1][len[1] - suff].value;
583 	    suff++)
584 		;
585 	for (j = 0; j < 2; j++) {
586 		sfile[j] = file[j] + pref;
587 		slen[j] = len[j] - pref - suff;
588 		for (i = 0; i <= slen[j]; i++)
589 			sfile[j][i].serial = i;
590 	}
591 }
592 
593 static void
594 equiv(struct line *a, int n, struct line *b, int m, int *c)
595 {
596 	int i, j;
597 
598 	i = j = 1;
599 	while (i <= n && j <= m) {
600 		if (a[i].value < b[j].value)
601 			a[i++].value = 0;
602 		else if (a[i].value == b[j].value)
603 			a[i++].value = j;
604 		else
605 			j++;
606 	}
607 	while (i <= n)
608 		a[i++].value = 0;
609 	b[m + 1].value = 0;
610 	j = 0;
611 	while (++j <= m) {
612 		c[j] = -b[j].serial;
613 		while (b[j + 1].value == b[j].value) {
614 			j++;
615 			c[j] = b[j].serial;
616 		}
617 	}
618 	c[j] = -1;
619 }
620 
621 static int
622 stone(int *a, int n, int *b, int *c, int flags)
623 {
624 	int i, k, y, j, l;
625 	int oldc, tc, oldl, sq;
626 	unsigned numtries, bound;
627 
628 	if (flags & D_MINIMAL)
629 		bound = UINT_MAX;
630 	else {
631 		sq = sqrt(n);
632 		bound = MAX(256, sq);
633 	}
634 
635 	k = 0;
636 	c[0] = newcand(0, 0, 0);
637 	for (i = 1; i <= n; i++) {
638 		j = a[i];
639 		if (j == 0)
640 			continue;
641 		y = -b[j];
642 		oldl = 0;
643 		oldc = c[0];
644 		numtries = 0;
645 		do {
646 			if (y <= clist[oldc].y)
647 				continue;
648 			l = search(c, k, y);
649 			if (l != oldl + 1)
650 				oldc = c[l - 1];
651 			if (l <= k) {
652 				if (clist[c[l]].y <= y)
653 					continue;
654 				tc = c[l];
655 				c[l] = newcand(i, y, oldc);
656 				oldc = tc;
657 				oldl = l;
658 				numtries++;
659 			} else {
660 				c[l] = newcand(i, y, oldc);
661 				k++;
662 				break;
663 			}
664 		} while ((y = b[++j]) > 0 && numtries < bound);
665 	}
666 	return (k);
667 }
668 
669 static int
670 newcand(int x, int y, int pred)
671 {
672 	struct cand *q;
673 
674 	if (clen == clistlen) {
675 		clistlen = clistlen * 11 / 10;
676 		clist = xreallocarray(clist, clistlen, sizeof(*clist));
677 	}
678 	q = clist + clen;
679 	q->x = x;
680 	q->y = y;
681 	q->pred = pred;
682 	return (clen++);
683 }
684 
685 static int
686 search(int *c, int k, int y)
687 {
688 	int i, j, l, t;
689 
690 	if (clist[c[k]].y < y)	/* quick look for typical case */
691 		return (k + 1);
692 	i = 0;
693 	j = k + 1;
694 	for (;;) {
695 		l = (i + j) / 2;
696 		if (l <= i)
697 			break;
698 		t = clist[c[l]].y;
699 		if (t > y)
700 			j = l;
701 		else if (t < y)
702 			i = l;
703 		else
704 			return (l);
705 	}
706 	return (l + 1);
707 }
708 
709 static void
710 unravel(int p)
711 {
712 	struct cand *q;
713 	int i;
714 
715 	for (i = 0; i <= len[0]; i++)
716 		J[i] = i <= pref ? i :
717 		    i > len[0] - suff ? i + len[1] - len[0] : 0;
718 	for (q = clist + p; q->y != 0; q = clist + q->pred)
719 		J[q->x + pref] = q->y + pref;
720 }
721 
722 /*
723  * Check does double duty:
724  *  1. ferret out any fortuitous correspondences due to confounding by
725  *     hashing (which result in "jackpot")
726  *  2. collect random access indexes to the two files
727  */
728 static void
729 check(FILE *f1, FILE *f2, int flags)
730 {
731 	int i, j, /* jackpot, */ c, d;
732 	long ctold, ctnew;
733 
734 	rewind(f1);
735 	rewind(f2);
736 	j = 1;
737 	ixold[0] = ixnew[0] = 0;
738 	/* jackpot = 0; */
739 	ctold = ctnew = 0;
740 	for (i = 1; i <= len[0]; i++) {
741 		if (J[i] == 0) {
742 			ixold[i] = ctold += skipline(f1);
743 			continue;
744 		}
745 		while (j < J[i]) {
746 			ixnew[j] = ctnew += skipline(f2);
747 			j++;
748 		}
749 		if (flags & (D_FOLDBLANKS | D_IGNOREBLANKS | D_IGNORECASE | D_STRIPCR)) {
750 			for (;;) {
751 				c = getc(f1);
752 				d = getc(f2);
753 				/*
754 				 * GNU diff ignores a missing newline
755 				 * in one file for -b or -w.
756 				 */
757 				if (flags & (D_FOLDBLANKS | D_IGNOREBLANKS)) {
758 					if (c == EOF && isspace(d)) {
759 						ctnew++;
760 						break;
761 					} else if (isspace(c) && d == EOF) {
762 						ctold++;
763 						break;
764 					}
765 				}
766 				ctold++;
767 				ctnew++;
768 				if (flags & D_STRIPCR && (c == '\r' || d == '\r')) {
769 					if (c == '\r') {
770 						if ((c = getc(f1)) == '\n') {
771 							ctold++;
772 						} else {
773 							ungetc(c, f1);
774 						}
775 					}
776 					if (d == '\r') {
777 						if ((d = getc(f2)) == '\n') {
778 							ctnew++;
779 						} else {
780 							ungetc(d, f2);
781 						}
782 					}
783 					break;
784 				}
785 				if ((flags & D_FOLDBLANKS) && isspace(c) &&
786 				    isspace(d)) {
787 					do {
788 						if (c == '\n')
789 							break;
790 						ctold++;
791 					} while (isspace(c = getc(f1)));
792 					do {
793 						if (d == '\n')
794 							break;
795 						ctnew++;
796 					} while (isspace(d = getc(f2)));
797 				} else if (flags & D_IGNOREBLANKS) {
798 					while (isspace(c) && c != '\n') {
799 						c = getc(f1);
800 						ctold++;
801 					}
802 					while (isspace(d) && d != '\n') {
803 						d = getc(f2);
804 						ctnew++;
805 					}
806 				}
807 				if (chrtran(c) != chrtran(d)) {
808 					/* jackpot++; */
809 					J[i] = 0;
810 					if (c != '\n' && c != EOF)
811 						ctold += skipline(f1);
812 					if (d != '\n' && c != EOF)
813 						ctnew += skipline(f2);
814 					break;
815 				}
816 				if (c == '\n' || c == EOF)
817 					break;
818 			}
819 		} else {
820 			for (;;) {
821 				ctold++;
822 				ctnew++;
823 				if ((c = getc(f1)) != (d = getc(f2))) {
824 					/* jackpot++; */
825 					J[i] = 0;
826 					if (c != '\n' && c != EOF)
827 						ctold += skipline(f1);
828 					if (d != '\n' && c != EOF)
829 						ctnew += skipline(f2);
830 					break;
831 				}
832 				if (c == '\n' || c == EOF)
833 					break;
834 			}
835 		}
836 		ixold[i] = ctold;
837 		ixnew[j] = ctnew;
838 		j++;
839 	}
840 	for (; j <= len[1]; j++) {
841 		ixnew[j] = ctnew += skipline(f2);
842 	}
843 	/*
844 	 * if (jackpot)
845 	 *	fprintf(stderr, "jackpot\n");
846 	 */
847 }
848 
849 /* shellsort CACM #201 */
850 static void
851 sort(struct line *a, int n)
852 {
853 	struct line *ai, *aim, w;
854 	int j, m = 0, k;
855 
856 	if (n == 0)
857 		return;
858 	for (j = 1; j <= n; j *= 2)
859 		m = 2 * j - 1;
860 	for (m /= 2; m != 0; m /= 2) {
861 		k = n - m;
862 		for (j = 1; j <= k; j++) {
863 			for (ai = &a[j]; ai > a; ai -= m) {
864 				aim = &ai[m];
865 				if (aim < ai)
866 					break;	/* wraparound */
867 				if (aim->value > ai[0].value ||
868 				    (aim->value == ai[0].value &&
869 					aim->serial > ai[0].serial))
870 					break;
871 				w.value = ai[0].value;
872 				ai[0].value = aim->value;
873 				aim->value = w.value;
874 				w.serial = ai[0].serial;
875 				ai[0].serial = aim->serial;
876 				aim->serial = w.serial;
877 			}
878 		}
879 	}
880 }
881 
882 static void
883 unsort(struct line *f, int l, int *b)
884 {
885 	int *a, i;
886 
887 	a = xcalloc(l + 1, sizeof(*a));
888 	for (i = 1; i <= l; i++)
889 		a[f[i].serial] = f[i].value;
890 	for (i = 1; i <= l; i++)
891 		b[i] = a[i];
892 	free(a);
893 }
894 
895 static int
896 skipline(FILE *f)
897 {
898 	int i, c;
899 
900 	for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++)
901 		continue;
902 	return (i);
903 }
904 
905 static void
906 output(char *file1, FILE *f1, char *file2, FILE *f2, int flags)
907 {
908 	int i, j, m, i0, i1, j0, j1, nc;
909 
910 	rewind(f1);
911 	rewind(f2);
912 	m = len[0];
913 	J[0] = 0;
914 	J[m + 1] = len[1] + 1;
915 	if (diff_format != D_EDIT) {
916 		for (i0 = 1; i0 <= m; i0 = i1 + 1) {
917 			while (i0 <= m && J[i0] == J[i0 - 1] + 1) {
918 				if (diff_format == D_SIDEBYSIDE && suppress_common != 1) {
919 					nc = fetch(ixold, i0, i0, f1, '\0', 1, flags);
920 					print_space(nc, hw - nc + lpad + 1 + rpad, flags);
921 					fetch(ixnew, J[i0], J[i0], f2, '\0', 0, flags);
922 					printf("\n");
923 				}
924 				i0++;
925 			}
926 			j0 = J[i0 - 1] + 1;
927 			i1 = i0 - 1;
928 			while (i1 < m && J[i1 + 1] == 0)
929 				i1++;
930 			j1 = J[i1 + 1] - 1;
931 			J[i1] = j1;
932 
933 			/*
934 			 * When using side-by-side, lines from both of the files are
935 			 * printed. The algorithm used by diff(1) identifies the ranges
936 			 * in which two files differ.
937 			 * See the change() function below.
938 			 * The for loop below consumes the shorter range, whereas one of
939 			 * the while loops deals with the longer one.
940 			 */
941 			if (diff_format == D_SIDEBYSIDE) {
942 				for (i = i0, j = j0; i <= i1 && j <= j1; i++, j++)
943 					change(file1, f1, file2, f2, i, i, j, j, &flags);
944 
945 				while (i <= i1) {
946 					change(file1, f1, file2, f2, i, i, j + 1, j, &flags);
947 					i++;
948 				}
949 
950 				while (j <= j1) {
951 					change(file1, f1, file2, f2, i + 1, i, j, j, &flags);
952 					j++;
953 				}
954 			} else
955 				change(file1, f1, file2, f2, i0, i1, j0, j1, &flags);
956 		}
957 	} else {
958 		for (i0 = m; i0 >= 1; i0 = i1 - 1) {
959 			while (i0 >= 1 && J[i0] == J[i0 + 1] - 1 && J[i0] != 0)
960 				i0--;
961 			j0 = J[i0 + 1] - 1;
962 			i1 = i0 + 1;
963 			while (i1 > 1 && J[i1 - 1] == 0)
964 				i1--;
965 			j1 = J[i1 - 1] + 1;
966 			J[i1] = j1;
967 			change(file1, f1, file2, f2, i1, i0, j1, j0, &flags);
968 		}
969 	}
970 	if (m == 0)
971 		change(file1, f1, file2, f2, 1, 0, 1, len[1], &flags);
972 	if (diff_format == D_IFDEF || diff_format == D_GFORMAT) {
973 		for (;;) {
974 #define	c i0
975 			if ((c = getc(f1)) == EOF)
976 				return;
977 			printf("%c", c);
978 		}
979 #undef c
980 	}
981 	if (anychange != 0) {
982 		if (diff_format == D_CONTEXT)
983 			dump_context_vec(f1, f2, flags);
984 		else if (diff_format == D_UNIFIED)
985 			dump_unified_vec(f1, f2, flags);
986 	}
987 }
988 
989 static void
990 range(int a, int b, const char *separator)
991 {
992 	printf("%d", a > b ? b : a);
993 	if (a < b)
994 		printf("%s%d", separator, b);
995 }
996 
997 static void
998 uni_range(int a, int b)
999 {
1000 	if (a < b)
1001 		printf("%d,%d", a, b - a + 1);
1002 	else if (a == b)
1003 		printf("%d", b);
1004 	else
1005 		printf("%d,0", b);
1006 }
1007 
1008 static char *
1009 preadline(int fd, size_t rlen, off_t off)
1010 {
1011 	char *line;
1012 	ssize_t nr;
1013 
1014 	line = xmalloc(rlen + 1);
1015 	if ((nr = pread(fd, line, rlen, off)) == -1)
1016 		err(2, "preadline");
1017 	if (nr > 0 && line[nr-1] == '\n')
1018 		nr--;
1019 	line[nr] = '\0';
1020 	return (line);
1021 }
1022 
1023 static bool
1024 ignoreline_pattern(char *line)
1025 {
1026 	int ret;
1027 
1028 	ret = regexec(&ignore_re, line, 0, NULL, 0);
1029 	return (ret == 0);	/* if it matched, it should be ignored. */
1030 }
1031 
1032 static bool
1033 ignoreline(char *line, bool skip_blanks)
1034 {
1035 
1036 	if (skip_blanks && *line == '\0')
1037 		return (true);
1038 	if (ignore_pats != NULL && ignoreline_pattern(line))
1039 		return (true);
1040 	return (false);
1041 }
1042 
1043 /*
1044  * Indicate that there is a difference between lines a and b of the from file
1045  * to get to lines c to d of the to file.  If a is greater then b then there
1046  * are no lines in the from file involved and this means that there were
1047  * lines appended (beginning at b).  If c is greater than d then there are
1048  * lines missing from the to file.
1049  */
1050 static void
1051 change(char *file1, FILE *f1, char *file2, FILE *f2, int a, int b, int c, int d,
1052     int *pflags)
1053 {
1054 	static size_t max_context = 64;
1055 	long curpos;
1056 	int i, nc;
1057 	const char *walk;
1058 	bool skip_blanks, ignore;
1059 
1060 	skip_blanks = (*pflags & D_SKIPBLANKLINES);
1061 restart:
1062 	if ((diff_format != D_IFDEF || diff_format == D_GFORMAT) &&
1063 	    a > b && c > d)
1064 		return;
1065 	if (ignore_pats != NULL || skip_blanks) {
1066 		char *line;
1067 		/*
1068 		 * All lines in the change, insert, or delete must match an ignore
1069 		 * pattern for the change to be ignored.
1070 		 */
1071 		if (a <= b) {		/* Changes and deletes. */
1072 			for (i = a; i <= b; i++) {
1073 				line = preadline(fileno(f1),
1074 				    ixold[i] - ixold[i - 1], ixold[i - 1]);
1075 				ignore = ignoreline(line, skip_blanks);
1076 				free(line);
1077 				if (!ignore)
1078 					goto proceed;
1079 			}
1080 		}
1081 		if (a > b || c <= d) {	/* Changes and inserts. */
1082 			for (i = c; i <= d; i++) {
1083 				line = preadline(fileno(f2),
1084 				    ixnew[i] - ixnew[i - 1], ixnew[i - 1]);
1085 				ignore = ignoreline(line, skip_blanks);
1086 				free(line);
1087 				if (!ignore)
1088 					goto proceed;
1089 			}
1090 		}
1091 		return;
1092 	}
1093 proceed:
1094 	if (*pflags & D_HEADER && diff_format != D_BRIEF) {
1095 		printf("%s %s %s\n", diffargs, file1, file2);
1096 		*pflags &= ~D_HEADER;
1097 	}
1098 	if (diff_format == D_CONTEXT || diff_format == D_UNIFIED) {
1099 		/*
1100 		 * Allocate change records as needed.
1101 		 */
1102 		if (context_vec_start == NULL ||
1103 		    context_vec_ptr == context_vec_end - 1) {
1104 			ptrdiff_t offset = -1;
1105 
1106 			if (context_vec_start != NULL)
1107 				offset = context_vec_ptr - context_vec_start;
1108 			max_context <<= 1;
1109 			context_vec_start = xreallocarray(context_vec_start,
1110 			    max_context, sizeof(*context_vec_start));
1111 			context_vec_end = context_vec_start + max_context;
1112 			context_vec_ptr = context_vec_start + offset;
1113 		}
1114 		if (anychange == 0) {
1115 			/*
1116 			 * Print the context/unidiff header first time through.
1117 			 */
1118 			print_header(file1, file2);
1119 			anychange = 1;
1120 		} else if (a > context_vec_ptr->b + (2 * diff_context) + 1 &&
1121 		    c > context_vec_ptr->d + (2 * diff_context) + 1) {
1122 			/*
1123 			 * If this change is more than 'diff_context' lines from the
1124 			 * previous change, dump the record and reset it.
1125 			 */
1126 			if (diff_format == D_CONTEXT)
1127 				dump_context_vec(f1, f2, *pflags);
1128 			else
1129 				dump_unified_vec(f1, f2, *pflags);
1130 		}
1131 		context_vec_ptr++;
1132 		context_vec_ptr->a = a;
1133 		context_vec_ptr->b = b;
1134 		context_vec_ptr->c = c;
1135 		context_vec_ptr->d = d;
1136 		return;
1137 	}
1138 	if (anychange == 0)
1139 		anychange = 1;
1140 	switch (diff_format) {
1141 	case D_BRIEF:
1142 		return;
1143 	case D_NORMAL:
1144 	case D_EDIT:
1145 		range(a, b, ",");
1146 		printf("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1147 		if (diff_format == D_NORMAL)
1148 			range(c, d, ",");
1149 		printf("\n");
1150 		break;
1151 	case D_REVERSE:
1152 		printf("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1153 		range(a, b, " ");
1154 		printf("\n");
1155 		break;
1156 	case D_NREVERSE:
1157 		if (a > b)
1158 			printf("a%d %d\n", b, d - c + 1);
1159 		else {
1160 			printf("d%d %d\n", a, b - a + 1);
1161 			if (!(c > d))
1162 				/* add changed lines */
1163 				printf("a%d %d\n", b, d - c + 1);
1164 		}
1165 		break;
1166 	}
1167 	if (diff_format == D_GFORMAT) {
1168 		curpos = ftell(f1);
1169 		/* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1170 		nc = ixold[a > b ? b : a - 1] - curpos;
1171 		for (i = 0; i < nc; i++)
1172 			printf("%c", getc(f1));
1173 		for (walk = group_format; *walk != '\0'; walk++) {
1174 			if (*walk == '%') {
1175 				walk++;
1176 				switch (*walk) {
1177 				case '<':
1178 					fetch(ixold, a, b, f1, '<', 1, *pflags);
1179 					break;
1180 				case '>':
1181 					fetch(ixnew, c, d, f2, '>', 0, *pflags);
1182 					break;
1183 				default:
1184 					printf("%%%c", *walk);
1185 					break;
1186 				}
1187 				continue;
1188 			}
1189 			printf("%c", *walk);
1190 		}
1191 	}
1192 	if (diff_format == D_SIDEBYSIDE) {
1193 		if (color && a > b)
1194 			printf("\033[%sm", add_code);
1195 		else if (color && c > d)
1196 			printf("\033[%sm", del_code);
1197 		if (a > b) {
1198 			print_space(0, hw + lpad, *pflags);
1199 		} else {
1200 			nc = fetch(ixold, a, b, f1, '\0', 1, *pflags);
1201 			print_space(nc, hw - nc + lpad, *pflags);
1202 		}
1203 		if (color && a > b)
1204 			printf("\033[%sm", add_code);
1205 		else if (color && c > d)
1206 			printf("\033[%sm", del_code);
1207 		printf("%c", (a > b) ? '>' : ((c > d) ? '<' : '|'));
1208 		if (color && c > d)
1209 			printf("\033[m");
1210 		print_space(hw + lpad + 1, rpad, *pflags);
1211 		fetch(ixnew, c, d, f2, '\0', 0, *pflags);
1212 		printf("\n");
1213 	}
1214 	if (diff_format == D_NORMAL || diff_format == D_IFDEF) {
1215 		fetch(ixold, a, b, f1, '<', 1, *pflags);
1216 		if (a <= b && c <= d && diff_format == D_NORMAL)
1217 			printf("---\n");
1218 	}
1219 	if (diff_format != D_GFORMAT && diff_format != D_SIDEBYSIDE)
1220 		fetch(ixnew, c, d, f2, diff_format == D_NORMAL ? '>' : '\0', 0, *pflags);
1221 	if (edoffset != 0 && diff_format == D_EDIT) {
1222 		/*
1223 		 * A non-zero edoffset value for D_EDIT indicates that the last line
1224 		 * printed was a bare dot (".") that has been escaped as ".." to
1225 		 * prevent ed(1) from misinterpreting it.  We have to add a
1226 		 * substitute command to change this back and restart where we left
1227 		 * off.
1228 		 */
1229 		printf(".\n");
1230 		printf("%ds/.//\n", a + edoffset - 1);
1231 		b = a + edoffset - 1;
1232 		a = b + 1;
1233 		c += edoffset;
1234 		goto restart;
1235 	}
1236 	if ((diff_format == D_EDIT || diff_format == D_REVERSE) && c <= d)
1237 		printf(".\n");
1238 	if (inifdef) {
1239 		printf("#endif /* %s */\n", ifdefname);
1240 		inifdef = 0;
1241 	}
1242 }
1243 
1244 static int
1245 fetch(long *f, int a, int b, FILE *lb, int ch, int oldfile, int flags)
1246 {
1247 	int i, j, c, lastc, col, nc, newcol;
1248 
1249 	edoffset = 0;
1250 	nc = 0;
1251 	col = 0;
1252 	/*
1253 	 * When doing #ifdef's, copy down to current line
1254 	 * if this is the first file, so that stuff makes it to output.
1255 	 */
1256 	if ((diff_format == D_IFDEF) && oldfile) {
1257 		long curpos = ftell(lb);
1258 		/* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1259 		nc = f[a > b ? b : a - 1] - curpos;
1260 		for (i = 0; i < nc; i++)
1261 			printf("%c", getc(lb));
1262 	}
1263 	if (a > b)
1264 		return (0);
1265 	if (diff_format == D_IFDEF) {
1266 		if (inifdef) {
1267 			printf("#else /* %s%s */\n",
1268 			    oldfile == 1 ? "!" : "", ifdefname);
1269 		} else {
1270 			if (oldfile)
1271 				printf("#ifndef %s\n", ifdefname);
1272 			else
1273 				printf("#ifdef %s\n", ifdefname);
1274 		}
1275 		inifdef = 1 + oldfile;
1276 	}
1277 	for (i = a; i <= b; i++) {
1278 		fseek(lb, f[i - 1], SEEK_SET);
1279 		nc = f[i] - f[i - 1];
1280 		if (diff_format == D_SIDEBYSIDE && hw < nc)
1281 			nc = hw;
1282 		if (diff_format != D_IFDEF && diff_format != D_GFORMAT &&
1283 		    ch != '\0') {
1284 			if (color && (ch == '>' || ch == '+'))
1285 				printf("\033[%sm", add_code);
1286 			else if (color && (ch == '<' || ch == '-'))
1287 				printf("\033[%sm", del_code);
1288 			printf("%c", ch);
1289 			if (Tflag && (diff_format == D_NORMAL ||
1290 			    diff_format == D_CONTEXT ||
1291 			    diff_format == D_UNIFIED))
1292 				printf("\t");
1293 			else if (diff_format != D_UNIFIED)
1294 				printf(" ");
1295 		}
1296 		col = j = 0;
1297 		lastc = '\0';
1298 		while (j < nc && (hw == 0 || col < hw)) {
1299 			c = getc(lb);
1300 			if (flags & D_STRIPCR && c == '\r') {
1301 				if ((c = getc(lb)) == '\n')
1302 					j++;
1303 				else {
1304 					ungetc(c, lb);
1305 					c = '\r';
1306 				}
1307 			}
1308 			if (c == EOF) {
1309 				if (diff_format == D_EDIT ||
1310 				    diff_format == D_REVERSE ||
1311 				    diff_format == D_NREVERSE)
1312 					warnx("No newline at end of file");
1313 				else
1314 					printf("\n\\ No newline at end of file\n");
1315 				return (col);
1316 			}
1317 			/*
1318 			 * when using --side-by-side, col needs to be increased
1319 			 * in any case to keep the columns aligned
1320 			 */
1321 			if (c == '\t') {
1322 				/*
1323 				 * Calculate where the tab would bring us.
1324 				 * If it would take us to the end of the
1325 				 * column, either clip it (if expanding
1326 				 * tabs) or return right away (if not).
1327 				 */
1328 				newcol = roundup(col + 1, tabsize);
1329 				if ((flags & D_EXPANDTABS) == 0) {
1330 					if (hw > 0 && newcol >= hw)
1331 						return (col);
1332 					printf("\t");
1333 				} else {
1334 					if (hw > 0 && newcol > hw)
1335 						newcol = hw;
1336 					printf("%*s", newcol - col, "");
1337 				}
1338 				col = newcol;
1339 			} else {
1340 				if (diff_format == D_EDIT && j == 1 && c == '\n' &&
1341 				    lastc == '.') {
1342 					/*
1343 					 * Don't print a bare "." line since that will confuse
1344 					 * ed(1). Print ".." instead and set the, global variable
1345 					 * edoffset to an offset from which to restart. The
1346 					 * caller must check the value of edoffset
1347 					 */
1348 					printf(".\n");
1349 					edoffset = i - a + 1;
1350 					return (edoffset);
1351 				}
1352 				/* when side-by-side, do not print a newline */
1353 				if (diff_format != D_SIDEBYSIDE || c != '\n') {
1354 					if (color && c == '\n')
1355 						printf("\033[m%c", c);
1356 					else
1357 						printf("%c", c);
1358 					col++;
1359 				}
1360 			}
1361 
1362 			j++;
1363 			lastc = c;
1364 		}
1365 	}
1366 	if (color && diff_format == D_SIDEBYSIDE)
1367 		printf("\033[m");
1368 	return (col);
1369 }
1370 
1371 /*
1372  * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578.
1373  */
1374 static enum readhash
1375 readhash(FILE *f, int flags, unsigned *hash)
1376 {
1377 	int i, t, space;
1378 	unsigned sum;
1379 
1380 	sum = 1;
1381 	space = 0;
1382 	for (i = 0;;) {
1383 		switch (t = getc(f)) {
1384 		case '\0':
1385 			if ((flags & D_FORCEASCII) == 0)
1386 				return (RH_BINARY);
1387 			goto hashchar;
1388 		case '\r':
1389 			if (flags & D_STRIPCR) {
1390 				t = getc(f);
1391 				if (t == '\n')
1392 					break;
1393 				ungetc(t, f);
1394 			}
1395 			/* FALLTHROUGH */
1396 		case '\t':
1397 		case '\v':
1398 		case '\f':
1399 		case ' ':
1400 			if ((flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) != 0) {
1401 				space++;
1402 				continue;
1403 			}
1404 			/* FALLTHROUGH */
1405 		default:
1406 		hashchar:
1407 			if (space && (flags & D_IGNOREBLANKS) == 0) {
1408 				i++;
1409 				space = 0;
1410 			}
1411 			sum = sum * 127 + chrtran(t);
1412 			i++;
1413 			continue;
1414 		case EOF:
1415 			if (i == 0)
1416 				return (RH_EOF);
1417 			/* FALLTHROUGH */
1418 		case '\n':
1419 			break;
1420 		}
1421 		break;
1422 	}
1423 	*hash = sum;
1424 	return (RH_OK);
1425 }
1426 
1427 static int
1428 asciifile(FILE *f)
1429 {
1430 	unsigned char buf[BUFSIZ];
1431 	size_t cnt;
1432 
1433 	if (f == NULL)
1434 		return (1);
1435 
1436 	rewind(f);
1437 	cnt = fread(buf, 1, sizeof(buf), f);
1438 	return (memchr(buf, '\0', cnt) == NULL);
1439 }
1440 
1441 #define begins_with(s, pre) (strncmp(s, pre, sizeof(pre) - 1) == 0)
1442 
1443 static char *
1444 match_function(const long *f, int pos, FILE *fp)
1445 {
1446 	unsigned char buf[FUNCTION_CONTEXT_SIZE];
1447 	size_t nc;
1448 	int last = lastline;
1449 	const char *state = NULL;
1450 
1451 	lastline = pos;
1452 	for (; pos > last; pos--) {
1453 		fseek(fp, f[pos - 1], SEEK_SET);
1454 		nc = f[pos] - f[pos - 1];
1455 		if (nc >= sizeof(buf))
1456 			nc = sizeof(buf) - 1;
1457 		nc = fread(buf, 1, nc, fp);
1458 		if (nc == 0)
1459 			continue;
1460 		buf[nc] = '\0';
1461 		buf[strcspn(buf, "\n")] = '\0';
1462 		if (most_recent_pat != NULL) {
1463 			int ret = regexec(&most_recent_re, buf, 0, NULL, 0);
1464 
1465 			if (ret != 0)
1466 				continue;
1467 			strlcpy(lastbuf, buf, sizeof(lastbuf));
1468 			lastmatchline = pos;
1469 			return (lastbuf);
1470 		} else if (isalpha(buf[0]) || buf[0] == '_' || buf[0] == '$'
1471 			|| buf[0] == '-' || buf[0] == '+') {
1472 			if (begins_with(buf, "private:")) {
1473 				if (!state)
1474 					state = " (private)";
1475 			} else if (begins_with(buf, "protected:")) {
1476 				if (!state)
1477 					state = " (protected)";
1478 			} else if (begins_with(buf, "public:")) {
1479 				if (!state)
1480 					state = " (public)";
1481 			} else {
1482 				strlcpy(lastbuf, buf, sizeof(lastbuf));
1483 				if (state)
1484 					strlcat(lastbuf, state, sizeof(lastbuf));
1485 				lastmatchline = pos;
1486 				return (lastbuf);
1487 			}
1488 		}
1489 	}
1490 	return (lastmatchline > 0 ? lastbuf : NULL);
1491 }
1492 
1493 /* dump accumulated "context" diff changes */
1494 static void
1495 dump_context_vec(FILE *f1, FILE *f2, int flags)
1496 {
1497 	struct context_vec *cvp = context_vec_start;
1498 	int lowa, upb, lowc, upd, do_output;
1499 	int a, b, c, d;
1500 	char ch, *f;
1501 
1502 	if (context_vec_start > context_vec_ptr)
1503 		return;
1504 
1505 	b = d = 0;		/* gcc */
1506 	lowa = MAX(1, cvp->a - diff_context);
1507 	upb = MIN(len[0], context_vec_ptr->b + diff_context);
1508 	lowc = MAX(1, cvp->c - diff_context);
1509 	upd = MIN(len[1], context_vec_ptr->d + diff_context);
1510 
1511 	printf("***************");
1512 	if (flags & (D_PROTOTYPE | D_MATCHLAST)) {
1513 		f = match_function(ixold, cvp->a - 1, f1);
1514 		if (f != NULL)
1515 			printf(" %s", f);
1516 	}
1517 	printf("\n*** ");
1518 	range(lowa, upb, ",");
1519 	printf(" ****\n");
1520 
1521 	/*
1522 	 * Output changes to the "old" file.  The first loop suppresses
1523 	 * output if there were no changes to the "old" file (we'll see
1524 	 * the "old" lines as context in the "new" list).
1525 	 */
1526 	do_output = 0;
1527 	for (; cvp <= context_vec_ptr; cvp++)
1528 		if (cvp->a <= cvp->b) {
1529 			cvp = context_vec_start;
1530 			do_output++;
1531 			break;
1532 		}
1533 	if (do_output) {
1534 		while (cvp <= context_vec_ptr) {
1535 			a = cvp->a;
1536 			b = cvp->b;
1537 			c = cvp->c;
1538 			d = cvp->d;
1539 
1540 			if (a <= b && c <= d)
1541 				ch = 'c';
1542 			else
1543 				ch = (a <= b) ? 'd' : 'a';
1544 
1545 			if (ch == 'a')
1546 				fetch(ixold, lowa, b, f1, ' ', 0, flags);
1547 			else {
1548 				fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1549 				fetch(ixold, a, b, f1,
1550 				    ch == 'c' ? '!' : '-', 0, flags);
1551 			}
1552 			lowa = b + 1;
1553 			cvp++;
1554 		}
1555 		fetch(ixold, b + 1, upb, f1, ' ', 0, flags);
1556 	}
1557 	/* output changes to the "new" file */
1558 	printf("--- ");
1559 	range(lowc, upd, ",");
1560 	printf(" ----\n");
1561 
1562 	do_output = 0;
1563 	for (cvp = context_vec_start; cvp <= context_vec_ptr; cvp++)
1564 		if (cvp->c <= cvp->d) {
1565 			cvp = context_vec_start;
1566 			do_output++;
1567 			break;
1568 		}
1569 	if (do_output) {
1570 		while (cvp <= context_vec_ptr) {
1571 			a = cvp->a;
1572 			b = cvp->b;
1573 			c = cvp->c;
1574 			d = cvp->d;
1575 
1576 			if (a <= b && c <= d)
1577 				ch = 'c';
1578 			else
1579 				ch = (a <= b) ? 'd' : 'a';
1580 
1581 			if (ch == 'd')
1582 				fetch(ixnew, lowc, d, f2, ' ', 0, flags);
1583 			else {
1584 				fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1585 				fetch(ixnew, c, d, f2,
1586 				    ch == 'c' ? '!' : '+', 0, flags);
1587 			}
1588 			lowc = d + 1;
1589 			cvp++;
1590 		}
1591 		fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1592 	}
1593 	context_vec_ptr = context_vec_start - 1;
1594 }
1595 
1596 /* dump accumulated "unified" diff changes */
1597 static void
1598 dump_unified_vec(FILE *f1, FILE *f2, int flags)
1599 {
1600 	struct context_vec *cvp = context_vec_start;
1601 	int lowa, upb, lowc, upd;
1602 	int a, b, c, d;
1603 	char ch, *f;
1604 
1605 	if (context_vec_start > context_vec_ptr)
1606 		return;
1607 
1608 	b = d = 0;		/* gcc */
1609 	lowa = MAX(1, cvp->a - diff_context);
1610 	upb = MIN(len[0], context_vec_ptr->b + diff_context);
1611 	lowc = MAX(1, cvp->c - diff_context);
1612 	upd = MIN(len[1], context_vec_ptr->d + diff_context);
1613 
1614 	printf("@@ -");
1615 	uni_range(lowa, upb);
1616 	printf(" +");
1617 	uni_range(lowc, upd);
1618 	printf(" @@");
1619 	if (flags & (D_PROTOTYPE | D_MATCHLAST)) {
1620 		f = match_function(ixold, cvp->a - 1, f1);
1621 		if (f != NULL)
1622 			printf(" %s", f);
1623 	}
1624 	printf("\n");
1625 
1626 	/*
1627 	 * Output changes in "unified" diff format--the old and new lines
1628 	 * are printed together.
1629 	 */
1630 	for (; cvp <= context_vec_ptr; cvp++) {
1631 		a = cvp->a;
1632 		b = cvp->b;
1633 		c = cvp->c;
1634 		d = cvp->d;
1635 
1636 		/*
1637 		 * c: both new and old changes
1638 		 * d: only changes in the old file
1639 		 * a: only changes in the new file
1640 		 */
1641 		if (a <= b && c <= d)
1642 			ch = 'c';
1643 		else
1644 			ch = (a <= b) ? 'd' : 'a';
1645 
1646 		switch (ch) {
1647 		case 'c':
1648 			fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1649 			fetch(ixold, a, b, f1, '-', 0, flags);
1650 			fetch(ixnew, c, d, f2, '+', 0, flags);
1651 			break;
1652 		case 'd':
1653 			fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1654 			fetch(ixold, a, b, f1, '-', 0, flags);
1655 			break;
1656 		case 'a':
1657 			fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1658 			fetch(ixnew, c, d, f2, '+', 0, flags);
1659 			break;
1660 		}
1661 		lowa = b + 1;
1662 		lowc = d + 1;
1663 	}
1664 	fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1665 
1666 	context_vec_ptr = context_vec_start - 1;
1667 }
1668 
1669 static void
1670 print_header(const char *file1, const char *file2)
1671 {
1672 	const char *time_format;
1673 	char buf[256];
1674 	struct tm tm1, tm2, *tm_ptr1, *tm_ptr2;
1675 	int nsec1 = stb1.st_mtim.tv_nsec;
1676 	int nsec2 = stb2.st_mtim.tv_nsec;
1677 
1678 	time_format = "%Y-%m-%d %H:%M:%S";
1679 
1680 	if (cflag)
1681 		time_format = "%c";
1682 	tm_ptr1 = localtime_r(&stb1.st_mtime, &tm1);
1683 	tm_ptr2 = localtime_r(&stb2.st_mtime, &tm2);
1684 	if (label[0] != NULL)
1685 		printf("%s %s\n", diff_format == D_CONTEXT ? "***" : "---",
1686 		    label[0]);
1687 	else {
1688 		strftime(buf, sizeof(buf), time_format, tm_ptr1);
1689 		printf("%s %s\t%s", diff_format == D_CONTEXT ? "***" : "---",
1690 		    file1, buf);
1691 		if (!cflag) {
1692 			strftime(buf, sizeof(buf), "%z", tm_ptr1);
1693 			printf(".%.9d %s", nsec1, buf);
1694 		}
1695 		printf("\n");
1696 	}
1697 	if (label[1] != NULL)
1698 		printf("%s %s\n", diff_format == D_CONTEXT ? "---" : "+++",
1699 		    label[1]);
1700 	else {
1701 		strftime(buf, sizeof(buf), time_format, tm_ptr2);
1702 		printf("%s %s\t%s", diff_format == D_CONTEXT ? "---" : "+++",
1703 		    file2, buf);
1704 		if (!cflag) {
1705 			strftime(buf, sizeof(buf), "%z", tm_ptr2);
1706 			printf(".%.9d %s", nsec2, buf);
1707 		}
1708 		printf("\n");
1709 	}
1710 }
1711 
1712 /*
1713  * Prints n number of space characters either by using tab
1714  * or single space characters.
1715  * nc is the preceding number of characters
1716  */
1717 static void
1718 print_space(int nc, int n, int flags)
1719 {
1720 	int col, newcol, tabstop;
1721 
1722 	col = nc;
1723 	newcol = nc + n;
1724 	/* first, use tabs if allowed */
1725 	if ((flags & D_EXPANDTABS) == 0) {
1726 		while ((tabstop = roundup(col + 1, tabsize)) <= newcol) {
1727 			printf("\t");
1728 			col = tabstop;
1729 		}
1730 	}
1731 	/* finish with spaces */
1732 	printf("%*s", newcol - col, "");
1733 }
1734