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 <stdckdint.h>
81 #include <stddef.h>
82 #include <stdint.h>
83 #include <stdio.h>
84 #include <stdlib.h>
85 #include <string.h>
86
87 #include "pr.h"
88 #include "diff.h"
89 #include "xmalloc.h"
90
91 /*
92 * diff - compare two files.
93 */
94
95 /*
96 * Uses an algorithm due to Harold Stone, which finds a pair of longest
97 * identical subsequences in the two files.
98 *
99 * The major goal is to generate the match vector J. J[i] is the index of
100 * the line in file1 corresponding to line i file0. J[i] = 0 if there is no
101 * such line in file1.
102 *
103 * Lines are hashed so as to work in core. All potential matches are
104 * located by sorting the lines of each file on the hash (called
105 * ``value''). In particular, this collects the equivalence classes in
106 * file1 together. Subroutine equiv replaces the value of each line in
107 * file0 by the index of the first element of its matching equivalence in
108 * (the reordered) file1. To save space equiv squeezes file1 into a single
109 * array member in which the equivalence classes are simply concatenated,
110 * except that their first members are flagged by changing sign.
111 *
112 * Next the indices that point into member are unsorted into array class
113 * according to the original order of file0.
114 *
115 * The cleverness lies in routine stone. This marches through the lines of
116 * file0, developing a vector klist of "k-candidates". At step i
117 * a k-candidate is a matched pair of lines x,y (x in file0 y in file1)
118 * such that there is a common subsequence of length k between the first
119 * i lines of file0 and the first y lines of file1, but there is no such
120 * subsequence for any smaller y. x is the earliest possible mate to y that
121 * occurs in such a subsequence.
122 *
123 * Whenever any of the members of the equivalence class of lines in file1
124 * matable to a line in file0 has serial number less than the y of some
125 * k-candidate, that k-candidate with the smallest such y is replaced. The
126 * new k-candidate is chained (via pred) to the current k-1 candidate so
127 * that the actual subsequence can be recovered. When a member has serial
128 * number greater that the y of all k-candidates, the klist is extended. At
129 * the end, the longest subsequence is pulled out and placed in the array J
130 * by unravel.
131 *
132 * With J in hand, the matches there recorded are check'ed against reality
133 * to assure that no spurious matches have crept in due to hashing. If they
134 * have, they are broken, and "jackpot" is recorded -- a harmless matter
135 * except that a true match for a spuriously mated line may now be
136 * unnecessarily reported as a change.
137 *
138 * Much of the complexity of the program comes simply from trying to
139 * minimize core utilization and maximize the range of doable problems by
140 * dynamically allocating what is needed and reusing what is not. The core
141 * requirements for problems larger than somewhat are (in words)
142 * 2*length(file0) + length(file1) + 3*(number of k-candidates installed),
143 * typically about 6n words for files of length n.
144 */
145
146 struct cand {
147 int x;
148 int y;
149 int pred;
150 };
151
152 static struct line {
153 int serial;
154 int value;
155 } *file[2];
156
157 /*
158 * The following struct is used to record change information when
159 * doing a "context" or "unified" diff. (see routine "change" to
160 * understand the highly mnemonic field names)
161 */
162 struct context_vec {
163 int a; /* start line in old file */
164 int b; /* end line in old file */
165 int c; /* start line in new file */
166 int d; /* end line in new file */
167 };
168
169 enum readhash { RH_BINARY, RH_OK, RH_EOF };
170
171 static int diffreg_stone(char *, char *, int, int);
172 static FILE *opentemp(const char *);
173 static void output(char *, FILE *, char *, FILE *, int);
174 static void check(FILE *, FILE *, int);
175 static void range(int, int, const char *);
176 static void uni_range(int, int);
177 static void dump_context_vec(FILE *, FILE *, int);
178 static void dump_unified_vec(FILE *, FILE *, int);
179 static bool prepare(int, FILE *, size_t, int);
180 static void prune(void);
181 static void equiv(struct line *, int, struct line *, int, int *);
182 static void unravel(int);
183 static void unsort(struct line *, int, int *);
184 static void change(char *, FILE *, char *, FILE *, int, int, int, int, int *);
185 static void sort(struct line *, int);
186 static void print_header(const char *, const char *);
187 static void print_space(int, int, int);
188 static bool ignoreline_pattern(char *);
189 static bool ignoreline(char *, bool);
190 static int asciifile(FILE *);
191 static int fetch(long *, int, int, FILE *, int, int, int);
192 static int newcand(int, int, int);
193 static int search(int *, int, int);
194 static int skipline(FILE *);
195 static int stone(int *, int, int *, int *, int);
196 static enum readhash readhash(FILE *, int, unsigned *);
197 static int files_differ(FILE *, FILE *, int);
198 static char *match_function(const long *, int, FILE *);
199 static char *preadline(int, size_t, off_t);
200
201 static int *J; /* will be overlaid on class */
202 static int *class; /* will be overlaid on file[0] */
203 static int *klist; /* will be overlaid on file[0] after class */
204 static int *member; /* will be overlaid on file[1] */
205 static int clen;
206 static int inifdef; /* whether or not we are in a #ifdef block */
207 static size_t len[2]; /* lengths of files in lines */
208 static size_t pref, suff; /* lengths of prefix and suffix */
209 static size_t slen[2]; /* lengths of files minus pref / suff */
210 static int anychange;
211 static int hw, lpad,rpad; /* half width and padding */
212 static int edoffset;
213 static long *ixnew; /* will be overlaid on file[1] */
214 static long *ixold; /* will be overlaid on klist */
215 static struct cand *clist; /* merely a free storage pot for candidates */
216 static int clistlen; /* the length of clist */
217 static struct line *sfile[2]; /* shortened by pruning common prefix/suffix */
218 static int (*chrtran)(int); /* translation table for case-folding */
219 static struct context_vec *context_vec_start;
220 static struct context_vec *context_vec_end;
221 static struct context_vec *context_vec_ptr;
222
223 #define FUNCTION_CONTEXT_SIZE 55
224 static char lastbuf[FUNCTION_CONTEXT_SIZE];
225 static int lastline;
226 static int lastmatchline;
227
228 int
diffreg(char * file1,char * file2,int flags,int capsicum)229 diffreg(char *file1, char *file2, int flags, int capsicum)
230 {
231 /*
232 * If we have set the algorithm with -A or --algorithm use that if we
233 * can and if not print an error.
234 */
235 if (diff_algorithm_set) {
236 if (diff_algorithm == D_DIFFMYERS ||
237 diff_algorithm == D_DIFFPATIENCE) {
238 if (can_libdiff(flags))
239 return diffreg_new(file1, file2, flags, capsicum);
240 else
241 errx(2, "cannot use Myers algorithm with selected options");
242 } else {
243 /* Fallback to using stone. */
244 return diffreg_stone(file1, file2, flags, capsicum);
245 }
246 } else {
247 if (can_libdiff(flags))
248 return diffreg_new(file1, file2, flags, capsicum);
249 else
250 return diffreg_stone(file1, file2, flags, capsicum);
251 }
252 }
253
254 static int
clow2low(int c)255 clow2low(int c)
256 {
257
258 return (c);
259 }
260
261 static int
cup2low(int c)262 cup2low(int c)
263 {
264
265 return (tolower(c));
266 }
267
268 int
diffreg_stone(char * file1,char * file2,int flags,int capsicum)269 diffreg_stone(char *file1, char *file2, int flags, int capsicum)
270 {
271 FILE *f1, *f2;
272 int i, rval;
273 struct pr *pr = NULL;
274 cap_rights_t rights_ro;
275
276 f1 = f2 = NULL;
277 rval = D_SAME;
278 anychange = 0;
279 lastline = 0;
280 lastmatchline = 0;
281
282 /*
283 * In side-by-side mode, we need to print the left column, a
284 * change marker surrounded by padding, and the right column.
285 *
286 * If expanding tabs, we don't care about alignment, so we simply
287 * subtract 3 from the width and divide by two.
288 *
289 * If not expanding tabs, we need to ensure that the right column
290 * is aligned to a tab stop. We start with the same formula, then
291 * decrement until we reach a size that lets us tab-align the
292 * right column. We then adjust the width down if necessary for
293 * the padding calculation to work.
294 *
295 * Left padding is half the space left over, rounded down; right
296 * padding is whatever is needed to match the width.
297 */
298 if (diff_format == D_SIDEBYSIDE) {
299 if (flags & D_EXPANDTABS) {
300 if (width > 3) {
301 hw = (width - 3) / 2;
302 } else {
303 /* not enough space */
304 hw = 0;
305 }
306 } else if (width <= 3 || width <= tabsize) {
307 /* not enough space */
308 hw = 0;
309 } else {
310 hw = (width - 3) / 2;
311 while (hw > 0 && roundup(hw + 3, tabsize) + hw > width)
312 hw--;
313 if (width - (roundup(hw + 3, tabsize) + hw) < tabsize)
314 width = roundup(hw + 3, tabsize) + hw;
315 }
316 lpad = (width - hw * 2 - 1) / 2;
317 rpad = (width - hw * 2 - 1) - lpad;
318 }
319
320 if (flags & D_IGNORECASE)
321 chrtran = cup2low;
322 else
323 chrtran = clow2low;
324 if (S_ISDIR(stb1.st_mode) != S_ISDIR(stb2.st_mode))
325 return (S_ISDIR(stb1.st_mode) ? D_MISMATCH1 : D_MISMATCH2);
326 if (strcmp(file1, "-") == 0 && strcmp(file2, "-") == 0)
327 goto closem;
328
329 if (flags & D_EMPTY1)
330 f1 = fopen(_PATH_DEVNULL, "r");
331 else {
332 if (!S_ISREG(stb1.st_mode)) {
333 if ((f1 = opentemp(file1)) == NULL ||
334 fstat(fileno(f1), &stb1) == -1) {
335 warn("%s", file1);
336 rval = D_ERROR;
337 status |= 2;
338 goto closem;
339 }
340 } else if (strcmp(file1, "-") == 0)
341 f1 = stdin;
342 else
343 f1 = fopen(file1, "r");
344 }
345 if (f1 == NULL) {
346 warn("%s", file1);
347 rval = D_ERROR;
348 status |= 2;
349 goto closem;
350 }
351
352 if (flags & D_EMPTY2)
353 f2 = fopen(_PATH_DEVNULL, "r");
354 else {
355 if (!S_ISREG(stb2.st_mode)) {
356 if ((f2 = opentemp(file2)) == NULL ||
357 fstat(fileno(f2), &stb2) == -1) {
358 warn("%s", file2);
359 rval = D_ERROR;
360 status |= 2;
361 goto closem;
362 }
363 } else if (strcmp(file2, "-") == 0)
364 f2 = stdin;
365 else
366 f2 = fopen(file2, "r");
367 }
368 if (f2 == NULL) {
369 warn("%s", file2);
370 rval = D_ERROR;
371 status |= 2;
372 goto closem;
373 }
374
375 if (stb1.st_dev == stb2.st_dev && stb1.st_ino == stb2.st_ino)
376 goto closem;
377
378 if (lflag)
379 pr = start_pr(file1, file2);
380
381 if (capsicum) {
382 cap_rights_init(&rights_ro, CAP_READ, CAP_FSTAT, CAP_SEEK);
383 if (caph_rights_limit(fileno(f1), &rights_ro) < 0)
384 err(2, "unable to limit rights on: %s", file1);
385 if (caph_rights_limit(fileno(f2), &rights_ro) < 0)
386 err(2, "unable to limit rights on: %s", file2);
387 if (fileno(f1) == STDIN_FILENO || fileno(f2) == STDIN_FILENO) {
388 /* stdin has already been limited */
389 if (caph_limit_stderr() == -1)
390 err(2, "unable to limit stderr");
391 if (caph_limit_stdout() == -1)
392 err(2, "unable to limit stdout");
393 } else if (caph_limit_stdio() == -1)
394 err(2, "unable to limit stdio");
395
396 caph_cache_catpages();
397 caph_cache_tzdata();
398 if (caph_enter() < 0)
399 err(2, "unable to enter capability mode");
400 }
401
402 switch (files_differ(f1, f2, flags)) {
403 case 0:
404 goto closem;
405 case 1:
406 break;
407 default:
408 /* error */
409 if (ferror(f1))
410 warn("%s", file1);
411 if (ferror(f2))
412 warn("%s", file2);
413 rval = D_ERROR;
414 status |= 2;
415 goto closem;
416 }
417
418 if (diff_format == D_BRIEF && ignore_pats == NULL &&
419 (flags & (D_FOLDBLANKS|D_IGNOREBLANKS|D_IGNORECASE|
420 D_SKIPBLANKLINES|D_STRIPCR)) == 0)
421 {
422 rval = D_DIFFER;
423 status |= 1;
424 goto closem;
425 }
426 if ((flags & D_FORCEASCII) != 0) {
427 (void)prepare(0, f1, stb1.st_size, flags);
428 (void)prepare(1, f2, stb2.st_size, flags);
429 } else if (!asciifile(f1) || !asciifile(f2) ||
430 !prepare(0, f1, stb1.st_size, flags) ||
431 !prepare(1, f2, stb2.st_size, flags)) {
432 rval = D_BINARY;
433 status |= 1;
434 goto closem;
435 }
436 if (len[0] > INT_MAX - 2)
437 errc(1, EFBIG, "%s", file1);
438 if (len[1] > INT_MAX - 2)
439 errc(1, EFBIG, "%s", file2);
440
441 prune();
442 sort(sfile[0], slen[0]);
443 sort(sfile[1], slen[1]);
444
445 member = (int *)file[1];
446 equiv(sfile[0], slen[0], sfile[1], slen[1], member);
447 member = xreallocarray(member, slen[1] + 2, sizeof(*member));
448
449 class = (int *)file[0];
450 unsort(sfile[0], slen[0], class);
451 class = xreallocarray(class, slen[0] + 2, sizeof(*class));
452
453 klist = xcalloc(slen[0] + 2, sizeof(*klist));
454 clen = 0;
455 clistlen = 100;
456 clist = xcalloc(clistlen, sizeof(*clist));
457 i = stone(class, slen[0], member, klist, flags);
458 free(member);
459 free(class);
460
461 J = xreallocarray(J, len[0] + 2, sizeof(*J));
462 unravel(klist[i]);
463 free(clist);
464 free(klist);
465
466 ixold = xreallocarray(ixold, len[0] + 2, sizeof(*ixold));
467 ixnew = xreallocarray(ixnew, len[1] + 2, sizeof(*ixnew));
468 check(f1, f2, flags);
469 output(file1, f1, file2, f2, flags);
470
471 closem:
472 if (pr != NULL)
473 stop_pr(pr);
474 if (anychange) {
475 status |= 1;
476 if (rval == D_SAME)
477 rval = D_DIFFER;
478 }
479 if (f1 != NULL)
480 fclose(f1);
481 if (f2 != NULL)
482 fclose(f2);
483
484 return (rval);
485 }
486
487 /*
488 * Check to see if the given files differ.
489 * Returns 0 if they are the same, 1 if different, and -1 on error.
490 * XXX - could use code from cmp(1) [faster]
491 */
492 static int
files_differ(FILE * f1,FILE * f2,int flags)493 files_differ(FILE *f1, FILE *f2, int flags)
494 {
495 char buf1[BUFSIZ], buf2[BUFSIZ];
496 size_t i, j;
497
498 if ((flags & (D_EMPTY1|D_EMPTY2)) || stb1.st_size != stb2.st_size ||
499 (stb1.st_mode & S_IFMT) != (stb2.st_mode & S_IFMT))
500 return (1);
501
502 if (stb1.st_dev == stb2.st_dev && stb1.st_ino == stb2.st_ino)
503 return (0);
504
505 for (;;) {
506 if ((i = fread(buf1, 1, sizeof(buf1), f1)) == 0 && ferror(f1))
507 return (-1);
508 if ((j = fread(buf2, 1, sizeof(buf2), f2)) == 0 && ferror(f2))
509 return (-1);
510 if (i != j)
511 return (1);
512 if (i == 0)
513 return (0);
514 if (memcmp(buf1, buf2, i) != 0)
515 return (1);
516 }
517 }
518
519 static FILE *
opentemp(const char * f)520 opentemp(const char *f)
521 {
522 char buf[BUFSIZ], tempfile[PATH_MAX];
523 ssize_t nread;
524 int ifd, ofd;
525
526 if (strcmp(f, "-") == 0)
527 ifd = STDIN_FILENO;
528 else if ((ifd = open(f, O_RDONLY, 0644)) == -1)
529 return (NULL);
530
531 (void)strlcpy(tempfile, _PATH_TMP "/diff.XXXXXXXX", sizeof(tempfile));
532
533 if ((ofd = mkstemp(tempfile)) == -1) {
534 close(ifd);
535 return (NULL);
536 }
537 unlink(tempfile);
538 while ((nread = read(ifd, buf, BUFSIZ)) > 0) {
539 if (write(ofd, buf, nread) != nread) {
540 close(ifd);
541 close(ofd);
542 return (NULL);
543 }
544 }
545 if (nread == -1) {
546 close(ifd);
547 close(ofd);
548 return (NULL);
549 }
550 close(ifd);
551 lseek(ofd, (off_t)0, SEEK_SET);
552 return (fdopen(ofd, "r"));
553 }
554
555 static bool
prepare(int i,FILE * fd,size_t filesize,int flags)556 prepare(int i, FILE *fd, size_t filesize, int flags)
557 {
558 struct line *p;
559 unsigned h;
560 size_t sz, j = 0;
561 enum readhash r;
562
563 rewind(fd);
564
565 sz = MIN(filesize, SIZE_MAX) / 25;
566 if (sz < 100)
567 sz = 100;
568
569 p = xcalloc(sz + 3, sizeof(*p));
570 while ((r = readhash(fd, flags, &h)) != RH_EOF) {
571 if (r == RH_BINARY)
572 return (false);
573 if (j == SIZE_MAX)
574 break;
575 if (j == sz) {
576 sz = sz * 3 / 2;
577 p = xreallocarray(p, sz + 3, sizeof(*p));
578 }
579 p[++j].value = h;
580 }
581
582 len[i] = j;
583 file[i] = p;
584
585 return (true);
586 }
587
588 static void
prune(void)589 prune(void)
590 {
591 size_t i, j;
592
593 for (pref = 0; pref < len[0] && pref < len[1] &&
594 file[0][pref + 1].value == file[1][pref + 1].value;
595 pref++)
596 ;
597 for (suff = 0; suff < len[0] - pref && suff < len[1] - pref &&
598 file[0][len[0] - suff].value == file[1][len[1] - suff].value;
599 suff++)
600 ;
601 for (j = 0; j < 2; j++) {
602 sfile[j] = file[j] + pref;
603 slen[j] = len[j] - pref - suff;
604 for (i = 0; i <= slen[j]; i++)
605 sfile[j][i].serial = i;
606 }
607 }
608
609 static void
equiv(struct line * a,int n,struct line * b,int m,int * c)610 equiv(struct line *a, int n, struct line *b, int m, int *c)
611 {
612 int i, j;
613
614 i = j = 1;
615 while (i <= n && j <= m) {
616 if (a[i].value < b[j].value)
617 a[i++].value = 0;
618 else if (a[i].value == b[j].value)
619 a[i++].value = j;
620 else
621 j++;
622 }
623 while (i <= n)
624 a[i++].value = 0;
625 b[m + 1].value = 0;
626 j = 0;
627 while (++j <= m) {
628 c[j] = -b[j].serial;
629 while (b[j + 1].value == b[j].value) {
630 j++;
631 c[j] = b[j].serial;
632 }
633 }
634 c[j] = -1;
635 }
636
637 static int
stone(int * a,int n,int * b,int * c,int flags)638 stone(int *a, int n, int *b, int *c, int flags)
639 {
640 int i, k, y, j, l;
641 int oldc, tc, oldl, sq;
642 unsigned numtries, bound;
643
644 if (flags & D_MINIMAL)
645 bound = UINT_MAX;
646 else {
647 sq = sqrt(n);
648 bound = MAX(256, sq);
649 }
650
651 k = 0;
652 c[0] = newcand(0, 0, 0);
653 for (i = 1; i <= n; i++) {
654 j = a[i];
655 if (j == 0)
656 continue;
657 y = -b[j];
658 oldl = 0;
659 oldc = c[0];
660 numtries = 0;
661 do {
662 if (y <= clist[oldc].y)
663 continue;
664 l = search(c, k, y);
665 if (l != oldl + 1)
666 oldc = c[l - 1];
667 if (l <= k) {
668 if (clist[c[l]].y <= y)
669 continue;
670 tc = c[l];
671 c[l] = newcand(i, y, oldc);
672 oldc = tc;
673 oldl = l;
674 numtries++;
675 } else {
676 c[l] = newcand(i, y, oldc);
677 k++;
678 break;
679 }
680 } while ((y = b[++j]) > 0 && numtries < bound);
681 }
682 return (k);
683 }
684
685 static int
newcand(int x,int y,int pred)686 newcand(int x, int y, int pred)
687 {
688 struct cand *q;
689
690 if (clen == clistlen) {
691 clistlen = clistlen * 11 / 10;
692 clist = xreallocarray(clist, clistlen, sizeof(*clist));
693 }
694 q = clist + clen;
695 q->x = x;
696 q->y = y;
697 q->pred = pred;
698 return (clen++);
699 }
700
701 static int
search(int * c,int k,int y)702 search(int *c, int k, int y)
703 {
704 int i, j, l, t;
705
706 if (clist[c[k]].y < y) /* quick look for typical case */
707 return (k + 1);
708 i = 0;
709 j = k + 1;
710 for (;;) {
711 l = (i + j) / 2;
712 if (l <= i)
713 break;
714 t = clist[c[l]].y;
715 if (t > y)
716 j = l;
717 else if (t < y)
718 i = l;
719 else
720 return (l);
721 }
722 return (l + 1);
723 }
724
725 static void
unravel(int p)726 unravel(int p)
727 {
728 struct cand *q;
729 size_t i;
730
731 for (i = 0; i <= len[0]; i++)
732 J[i] = i <= pref ? i :
733 i > len[0] - suff ? i + len[1] - len[0] : 0;
734 for (q = clist + p; q->y != 0; q = clist + q->pred)
735 J[q->x + pref] = q->y + pref;
736 }
737
738 /*
739 * Check does double duty:
740 * 1. ferret out any fortuitous correspondences due to confounding by
741 * hashing (which result in "jackpot")
742 * 2. collect random access indexes to the two files
743 */
744 static void
check(FILE * f1,FILE * f2,int flags)745 check(FILE *f1, FILE *f2, int flags)
746 {
747 int i, j, /* jackpot, */ c, d;
748 long ctold, ctnew;
749
750 rewind(f1);
751 rewind(f2);
752 j = 1;
753 ixold[0] = ixnew[0] = 0;
754 /* jackpot = 0; */
755 ctold = ctnew = 0;
756 for (i = 1; i <= (int)len[0]; i++) {
757 if (J[i] == 0) {
758 ixold[i] = ctold += skipline(f1);
759 continue;
760 }
761 while (j < J[i]) {
762 ixnew[j] = ctnew += skipline(f2);
763 j++;
764 }
765 if (flags & (D_FOLDBLANKS | D_IGNOREBLANKS | D_IGNORECASE | D_STRIPCR)) {
766 for (;;) {
767 c = getc(f1);
768 d = getc(f2);
769 /*
770 * GNU diff ignores a missing newline
771 * in one file for -b or -w.
772 */
773 if (flags & (D_FOLDBLANKS | D_IGNOREBLANKS)) {
774 if (c == EOF && isspace(d)) {
775 ctnew++;
776 break;
777 } else if (isspace(c) && d == EOF) {
778 ctold++;
779 break;
780 }
781 }
782 ctold++;
783 ctnew++;
784 if (flags & D_STRIPCR && (c == '\r' || d == '\r')) {
785 if (c == '\r') {
786 if ((c = getc(f1)) == '\n') {
787 ctold++;
788 } else {
789 ungetc(c, f1);
790 }
791 }
792 if (d == '\r') {
793 if ((d = getc(f2)) == '\n') {
794 ctnew++;
795 } else {
796 ungetc(d, f2);
797 }
798 }
799 break;
800 }
801 if ((flags & D_FOLDBLANKS) && isspace(c) &&
802 isspace(d)) {
803 do {
804 if (c == '\n')
805 break;
806 ctold++;
807 } while (isspace(c = getc(f1)));
808 do {
809 if (d == '\n')
810 break;
811 ctnew++;
812 } while (isspace(d = getc(f2)));
813 } else if (flags & D_IGNOREBLANKS) {
814 while (isspace(c) && c != '\n') {
815 c = getc(f1);
816 ctold++;
817 }
818 while (isspace(d) && d != '\n') {
819 d = getc(f2);
820 ctnew++;
821 }
822 }
823 if (chrtran(c) != chrtran(d)) {
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 } else {
836 for (;;) {
837 ctold++;
838 ctnew++;
839 if ((c = getc(f1)) != (d = getc(f2))) {
840 /* jackpot++; */
841 J[i] = 0;
842 if (c != '\n' && c != EOF)
843 ctold += skipline(f1);
844 if (d != '\n' && c != EOF)
845 ctnew += skipline(f2);
846 break;
847 }
848 if (c == '\n' || c == EOF)
849 break;
850 }
851 }
852 ixold[i] = ctold;
853 ixnew[j] = ctnew;
854 j++;
855 }
856 for (; j <= (int)len[1]; j++) {
857 ixnew[j] = ctnew += skipline(f2);
858 }
859 /*
860 * if (jackpot)
861 * fprintf(stderr, "jackpot\n");
862 */
863 }
864
865 /* shellsort CACM #201 */
866 static void
sort(struct line * a,int n)867 sort(struct line *a, int n)
868 {
869 struct line *ai, *aim, w;
870 int j, m = 0, k;
871
872 if (n == 0)
873 return;
874 for (j = 1; j <= n; j *= 2)
875 m = 2 * j - 1;
876 for (m /= 2; m != 0; m /= 2) {
877 k = n - m;
878 for (j = 1; j <= k; j++) {
879 for (ai = &a[j]; ai > a; ai -= m) {
880 aim = &ai[m];
881 if (aim < ai)
882 break; /* wraparound */
883 if (aim->value > ai[0].value ||
884 (aim->value == ai[0].value &&
885 aim->serial > ai[0].serial))
886 break;
887 w.value = ai[0].value;
888 ai[0].value = aim->value;
889 aim->value = w.value;
890 w.serial = ai[0].serial;
891 ai[0].serial = aim->serial;
892 aim->serial = w.serial;
893 }
894 }
895 }
896 }
897
898 static void
unsort(struct line * f,int l,int * b)899 unsort(struct line *f, int l, int *b)
900 {
901 int *a, i;
902
903 a = xcalloc(l + 1, sizeof(*a));
904 for (i = 1; i <= l; i++)
905 a[f[i].serial] = f[i].value;
906 for (i = 1; i <= l; i++)
907 b[i] = a[i];
908 free(a);
909 }
910
911 static int
skipline(FILE * f)912 skipline(FILE *f)
913 {
914 int i, c;
915
916 for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++)
917 continue;
918 return (i);
919 }
920
921 static void
output(char * file1,FILE * f1,char * file2,FILE * f2,int flags)922 output(char *file1, FILE *f1, char *file2, FILE *f2, int flags)
923 {
924 int i, j, m, i0, i1, j0, j1, nc;
925
926 rewind(f1);
927 rewind(f2);
928 m = len[0];
929 J[0] = 0;
930 J[m + 1] = len[1] + 1;
931 if (diff_format != D_EDIT) {
932 for (i0 = 1; i0 <= m; i0 = i1 + 1) {
933 while (i0 <= m && J[i0] == J[i0 - 1] + 1) {
934 if (diff_format == D_SIDEBYSIDE && suppress_common != 1) {
935 nc = fetch(ixold, i0, i0, f1, '\0', 1, flags);
936 print_space(nc, hw - nc + lpad + 1 + rpad, flags);
937 fetch(ixnew, J[i0], J[i0], f2, '\0', 0, flags);
938 printf("\n");
939 }
940 i0++;
941 }
942 j0 = J[i0 - 1] + 1;
943 i1 = i0 - 1;
944 while (i1 < m && J[i1 + 1] == 0)
945 i1++;
946 j1 = J[i1 + 1] - 1;
947 J[i1] = j1;
948
949 /*
950 * When using side-by-side, lines from both of the files are
951 * printed. The algorithm used by diff(1) identifies the ranges
952 * in which two files differ.
953 * See the change() function below.
954 * The for loop below consumes the shorter range, whereas one of
955 * the while loops deals with the longer one.
956 */
957 if (diff_format == D_SIDEBYSIDE) {
958 for (i = i0, j = j0; i <= i1 && j <= j1; i++, j++)
959 change(file1, f1, file2, f2, i, i, j, j, &flags);
960
961 while (i <= i1) {
962 change(file1, f1, file2, f2, i, i, j + 1, j, &flags);
963 i++;
964 }
965
966 while (j <= j1) {
967 change(file1, f1, file2, f2, i + 1, i, j, j, &flags);
968 j++;
969 }
970 } else
971 change(file1, f1, file2, f2, i0, i1, j0, j1, &flags);
972 }
973 } else {
974 for (i0 = m; i0 >= 1; i0 = i1 - 1) {
975 while (i0 >= 1 && J[i0] == J[i0 + 1] - 1 && J[i0] != 0)
976 i0--;
977 j0 = J[i0 + 1] - 1;
978 i1 = i0 + 1;
979 while (i1 > 1 && J[i1 - 1] == 0)
980 i1--;
981 j1 = J[i1 - 1] + 1;
982 J[i1] = j1;
983 change(file1, f1, file2, f2, i1, i0, j1, j0, &flags);
984 }
985 }
986 if (m == 0)
987 change(file1, f1, file2, f2, 1, 0, 1, len[1], &flags);
988 if (diff_format == D_IFDEF || diff_format == D_GFORMAT) {
989 for (;;) {
990 #define c i0
991 if ((c = getc(f1)) == EOF)
992 return;
993 printf("%c", c);
994 }
995 #undef c
996 }
997 if (anychange != 0) {
998 if (diff_format == D_CONTEXT)
999 dump_context_vec(f1, f2, flags);
1000 else if (diff_format == D_UNIFIED)
1001 dump_unified_vec(f1, f2, flags);
1002 }
1003 }
1004
1005 static void
range(int a,int b,const char * separator)1006 range(int a, int b, const char *separator)
1007 {
1008 printf("%d", a > b ? b : a);
1009 if (a < b)
1010 printf("%s%d", separator, b);
1011 }
1012
1013 static void
uni_range(int a,int b)1014 uni_range(int a, int b)
1015 {
1016 if (a < b)
1017 printf("%d,%d", a, b - a + 1);
1018 else if (a == b)
1019 printf("%d", b);
1020 else
1021 printf("%d,0", b);
1022 }
1023
1024 static char *
preadline(int fd,size_t rlen,off_t off)1025 preadline(int fd, size_t rlen, off_t off)
1026 {
1027 char *line;
1028 ssize_t nr;
1029
1030 line = xmalloc(rlen + 1);
1031 if ((nr = pread(fd, line, rlen, off)) == -1)
1032 err(2, "preadline");
1033 if (nr > 0 && line[nr-1] == '\n')
1034 nr--;
1035 line[nr] = '\0';
1036 return (line);
1037 }
1038
1039 static bool
ignoreline_pattern(char * line)1040 ignoreline_pattern(char *line)
1041 {
1042 int ret;
1043
1044 ret = regexec(&ignore_re, line, 0, NULL, 0);
1045 return (ret == 0); /* if it matched, it should be ignored. */
1046 }
1047
1048 static bool
ignoreline(char * line,bool skip_blanks)1049 ignoreline(char *line, bool skip_blanks)
1050 {
1051
1052 if (skip_blanks && *line == '\0')
1053 return (true);
1054 if (ignore_pats != NULL && ignoreline_pattern(line))
1055 return (true);
1056 return (false);
1057 }
1058
1059 /*
1060 * Indicate that there is a difference between lines a and b of the from file
1061 * to get to lines c to d of the to file. If a is greater then b then there
1062 * are no lines in the from file involved and this means that there were
1063 * lines appended (beginning at b). If c is greater than d then there are
1064 * lines missing from the to file.
1065 */
1066 static void
change(char * file1,FILE * f1,char * file2,FILE * f2,int a,int b,int c,int d,int * pflags)1067 change(char *file1, FILE *f1, char *file2, FILE *f2, int a, int b, int c, int d,
1068 int *pflags)
1069 {
1070 static size_t max_context = 64;
1071 long curpos;
1072 int dist, i, nc;
1073 const char *walk;
1074 bool skip_blanks, ignore;
1075
1076 skip_blanks = (*pflags & D_SKIPBLANKLINES);
1077 restart:
1078 if ((diff_format != D_IFDEF || diff_format == D_GFORMAT) &&
1079 a > b && c > d)
1080 return;
1081 if (ignore_pats != NULL || skip_blanks) {
1082 char *line;
1083 /*
1084 * All lines in the change, insert, or delete must match an ignore
1085 * pattern for the change to be ignored.
1086 */
1087 if (a <= b) { /* Changes and deletes. */
1088 for (i = a; i <= b; i++) {
1089 line = preadline(fileno(f1),
1090 ixold[i] - ixold[i - 1], ixold[i - 1]);
1091 ignore = ignoreline(line, skip_blanks);
1092 free(line);
1093 if (!ignore)
1094 goto proceed;
1095 }
1096 }
1097 if (a > b || c <= d) { /* Changes and inserts. */
1098 for (i = c; i <= d; i++) {
1099 line = preadline(fileno(f2),
1100 ixnew[i] - ixnew[i - 1], ixnew[i - 1]);
1101 ignore = ignoreline(line, skip_blanks);
1102 free(line);
1103 if (!ignore)
1104 goto proceed;
1105 }
1106 }
1107 return;
1108 }
1109 proceed:
1110 if (*pflags & D_HEADER && diff_format != D_BRIEF) {
1111 printf("%s %s %s\n", diffargs, file1, file2);
1112 *pflags &= ~D_HEADER;
1113 }
1114 if (diff_format == D_CONTEXT || diff_format == D_UNIFIED) {
1115 /*
1116 * Allocate change records as needed.
1117 */
1118 if (context_vec_start == NULL ||
1119 context_vec_ptr == context_vec_end - 1) {
1120 ptrdiff_t offset = -1;
1121
1122 if (context_vec_start != NULL)
1123 offset = context_vec_ptr - context_vec_start;
1124 max_context <<= 1;
1125 context_vec_start = xreallocarray(context_vec_start,
1126 max_context, sizeof(*context_vec_start));
1127 context_vec_end = context_vec_start + max_context;
1128 context_vec_ptr = context_vec_start + offset;
1129 }
1130 if (anychange == 0) {
1131 /*
1132 * Print the context/unidiff header first time through.
1133 */
1134 print_header(file1, file2);
1135 anychange = 1;
1136 } else if (!ckd_add(&dist, diff_context, diff_context) &&
1137 a - context_vec_ptr->b - 1 > dist &&
1138 c - context_vec_ptr->d - 1 > dist) {
1139 /*
1140 * If this change is more than 'diff_context' lines from the
1141 * previous change, dump the record and reset it.
1142 */
1143 if (diff_format == D_CONTEXT)
1144 dump_context_vec(f1, f2, *pflags);
1145 else
1146 dump_unified_vec(f1, f2, *pflags);
1147 }
1148 context_vec_ptr++;
1149 context_vec_ptr->a = a;
1150 context_vec_ptr->b = b;
1151 context_vec_ptr->c = c;
1152 context_vec_ptr->d = d;
1153 return;
1154 }
1155 if (anychange == 0)
1156 anychange = 1;
1157 switch (diff_format) {
1158 case D_BRIEF:
1159 return;
1160 case D_NORMAL:
1161 case D_EDIT:
1162 range(a, b, ",");
1163 printf("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1164 if (diff_format == D_NORMAL)
1165 range(c, d, ",");
1166 printf("\n");
1167 break;
1168 case D_REVERSE:
1169 printf("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1170 range(a, b, " ");
1171 printf("\n");
1172 break;
1173 case D_NREVERSE:
1174 if (a > b)
1175 printf("a%d %d\n", b, d - c + 1);
1176 else {
1177 printf("d%d %d\n", a, b - a + 1);
1178 if (!(c > d))
1179 /* add changed lines */
1180 printf("a%d %d\n", b, d - c + 1);
1181 }
1182 break;
1183 }
1184 if (diff_format == D_GFORMAT) {
1185 curpos = ftell(f1);
1186 /* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1187 nc = ixold[a > b ? b : a - 1] - curpos;
1188 for (i = 0; i < nc; i++)
1189 printf("%c", getc(f1));
1190 for (walk = group_format; *walk != '\0'; walk++) {
1191 if (*walk == '%') {
1192 walk++;
1193 switch (*walk) {
1194 case '<':
1195 fetch(ixold, a, b, f1, '<', 1, *pflags);
1196 break;
1197 case '>':
1198 fetch(ixnew, c, d, f2, '>', 0, *pflags);
1199 break;
1200 default:
1201 printf("%%%c", *walk);
1202 break;
1203 }
1204 continue;
1205 }
1206 printf("%c", *walk);
1207 }
1208 }
1209 if (diff_format == D_SIDEBYSIDE) {
1210 if (color && a > b)
1211 printf("\033[%sm", add_code);
1212 else if (color && c > d)
1213 printf("\033[%sm", del_code);
1214 if (a > b) {
1215 print_space(0, hw + lpad, *pflags);
1216 } else {
1217 nc = fetch(ixold, a, b, f1, '\0', 1, *pflags);
1218 print_space(nc, hw - nc + lpad, *pflags);
1219 }
1220 if (color && a > b)
1221 printf("\033[%sm", add_code);
1222 else if (color && c > d)
1223 printf("\033[%sm", del_code);
1224 printf("%c", (a > b) ? '>' : ((c > d) ? '<' : '|'));
1225 if (color && c > d)
1226 printf("\033[m");
1227 print_space(hw + lpad + 1, rpad, *pflags);
1228 fetch(ixnew, c, d, f2, '\0', 0, *pflags);
1229 printf("\n");
1230 }
1231 if (diff_format == D_NORMAL || diff_format == D_IFDEF) {
1232 fetch(ixold, a, b, f1, '<', 1, *pflags);
1233 if (a <= b && c <= d && diff_format == D_NORMAL)
1234 printf("---\n");
1235 }
1236 if (diff_format != D_GFORMAT && diff_format != D_SIDEBYSIDE)
1237 fetch(ixnew, c, d, f2, diff_format == D_NORMAL ? '>' : '\0', 0, *pflags);
1238 if (edoffset != 0 && diff_format == D_EDIT) {
1239 /*
1240 * A non-zero edoffset value for D_EDIT indicates that the last line
1241 * printed was a bare dot (".") that has been escaped as ".." to
1242 * prevent ed(1) from misinterpreting it. We have to add a
1243 * substitute command to change this back and restart where we left
1244 * off.
1245 */
1246 printf(".\n");
1247 printf("%ds/.//\n", a + edoffset - 1);
1248 b = a + edoffset - 1;
1249 a = b + 1;
1250 c += edoffset;
1251 goto restart;
1252 }
1253 if ((diff_format == D_EDIT || diff_format == D_REVERSE) && c <= d)
1254 printf(".\n");
1255 if (inifdef) {
1256 printf("#endif /* %s */\n", ifdefname);
1257 inifdef = 0;
1258 }
1259 }
1260
1261 static int
fetch(long * f,int a,int b,FILE * lb,int ch,int oldfile,int flags)1262 fetch(long *f, int a, int b, FILE *lb, int ch, int oldfile, int flags)
1263 {
1264 int i, j, c, lastc, col, nc, newcol;
1265
1266 edoffset = 0;
1267 nc = 0;
1268 col = 0;
1269 /*
1270 * When doing #ifdef's, copy down to current line
1271 * if this is the first file, so that stuff makes it to output.
1272 */
1273 if ((diff_format == D_IFDEF) && oldfile) {
1274 long curpos = ftell(lb);
1275 /* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1276 nc = f[a > b ? b : a - 1] - curpos;
1277 for (i = 0; i < nc; i++)
1278 printf("%c", getc(lb));
1279 }
1280 if (a > b)
1281 return (0);
1282 if (diff_format == D_IFDEF) {
1283 if (inifdef) {
1284 printf("#else /* %s%s */\n",
1285 oldfile == 1 ? "!" : "", ifdefname);
1286 } else {
1287 if (oldfile)
1288 printf("#ifndef %s\n", ifdefname);
1289 else
1290 printf("#ifdef %s\n", ifdefname);
1291 }
1292 inifdef = 1 + oldfile;
1293 }
1294 for (i = a; i <= b; i++) {
1295 fseek(lb, f[i - 1], SEEK_SET);
1296 nc = f[i] - f[i - 1];
1297 if (diff_format == D_SIDEBYSIDE && hw < nc)
1298 nc = hw;
1299 if (diff_format != D_IFDEF && diff_format != D_GFORMAT &&
1300 ch != '\0') {
1301 if (color && (ch == '>' || ch == '+'))
1302 printf("\033[%sm", add_code);
1303 else if (color && (ch == '<' || ch == '-'))
1304 printf("\033[%sm", del_code);
1305 printf("%c", ch);
1306 if (Tflag && (diff_format == D_NORMAL ||
1307 diff_format == D_CONTEXT ||
1308 diff_format == D_UNIFIED))
1309 printf("\t");
1310 else if (diff_format != D_UNIFIED)
1311 printf(" ");
1312 }
1313 col = j = 0;
1314 lastc = '\0';
1315 while (j < nc && (hw == 0 || col < hw)) {
1316 c = getc(lb);
1317 if (flags & D_STRIPCR && c == '\r') {
1318 if ((c = getc(lb)) == '\n')
1319 j++;
1320 else {
1321 ungetc(c, lb);
1322 c = '\r';
1323 }
1324 }
1325 if (c == EOF) {
1326 if (diff_format == D_EDIT ||
1327 diff_format == D_REVERSE ||
1328 diff_format == D_NREVERSE)
1329 warnx("No newline at end of file");
1330 else
1331 printf("\n\\ No newline at end of file\n");
1332 return (col);
1333 }
1334 /*
1335 * when using --side-by-side, col needs to be increased
1336 * in any case to keep the columns aligned
1337 */
1338 if (c == '\t') {
1339 /*
1340 * Calculate where the tab would bring us.
1341 * If it would take us to the end of the
1342 * column, either clip it (if expanding
1343 * tabs) or return right away (if not).
1344 */
1345 newcol = roundup(col + 1, tabsize);
1346 if ((flags & D_EXPANDTABS) == 0) {
1347 if (hw > 0 && newcol >= hw)
1348 return (col);
1349 printf("\t");
1350 } else {
1351 if (hw > 0 && newcol > hw)
1352 newcol = hw;
1353 printf("%*s", newcol - col, "");
1354 }
1355 col = newcol;
1356 } else {
1357 if (diff_format == D_EDIT && j == 1 && c == '\n' &&
1358 lastc == '.') {
1359 /*
1360 * Don't print a bare "." line since that will confuse
1361 * ed(1). Print ".." instead and set the, global variable
1362 * edoffset to an offset from which to restart. The
1363 * caller must check the value of edoffset
1364 */
1365 printf(".\n");
1366 edoffset = i - a + 1;
1367 return (edoffset);
1368 }
1369 /* when side-by-side, do not print a newline */
1370 if (diff_format != D_SIDEBYSIDE || c != '\n') {
1371 if (color && c == '\n')
1372 printf("\033[m%c", c);
1373 else
1374 printf("%c", c);
1375 col++;
1376 }
1377 }
1378
1379 j++;
1380 lastc = c;
1381 }
1382 }
1383 if (color && diff_format == D_SIDEBYSIDE)
1384 printf("\033[m");
1385 return (col);
1386 }
1387
1388 /*
1389 * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578.
1390 */
1391 static enum readhash
readhash(FILE * f,int flags,unsigned * hash)1392 readhash(FILE *f, int flags, unsigned *hash)
1393 {
1394 int i, t, space;
1395 unsigned sum;
1396
1397 sum = 1;
1398 space = 0;
1399 for (i = 0;;) {
1400 switch (t = getc(f)) {
1401 case '\0':
1402 if ((flags & D_FORCEASCII) == 0)
1403 return (RH_BINARY);
1404 goto hashchar;
1405 case '\r':
1406 if (flags & D_STRIPCR) {
1407 t = getc(f);
1408 if (t == '\n')
1409 break;
1410 ungetc(t, f);
1411 }
1412 /* FALLTHROUGH */
1413 case '\t':
1414 case '\v':
1415 case '\f':
1416 case ' ':
1417 if ((flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) != 0) {
1418 space++;
1419 continue;
1420 }
1421 /* FALLTHROUGH */
1422 default:
1423 hashchar:
1424 if (space && (flags & D_IGNOREBLANKS) == 0) {
1425 i++;
1426 space = 0;
1427 }
1428 sum = sum * 127 + chrtran(t);
1429 i++;
1430 continue;
1431 case EOF:
1432 if (i == 0)
1433 return (RH_EOF);
1434 /* FALLTHROUGH */
1435 case '\n':
1436 break;
1437 }
1438 break;
1439 }
1440 *hash = sum;
1441 return (RH_OK);
1442 }
1443
1444 static int
asciifile(FILE * f)1445 asciifile(FILE *f)
1446 {
1447 unsigned char buf[BUFSIZ];
1448 size_t cnt;
1449
1450 if (f == NULL)
1451 return (1);
1452
1453 rewind(f);
1454 cnt = fread(buf, 1, sizeof(buf), f);
1455 return (memchr(buf, '\0', cnt) == NULL);
1456 }
1457
1458 #define begins_with(s, pre) (strncmp(s, pre, sizeof(pre) - 1) == 0)
1459
1460 static char *
match_function(const long * f,int pos,FILE * fp)1461 match_function(const long *f, int pos, FILE *fp)
1462 {
1463 unsigned char buf[FUNCTION_CONTEXT_SIZE];
1464 size_t nc;
1465 int last = lastline;
1466 const char *state = NULL;
1467
1468 lastline = pos;
1469 for (; pos > last; pos--) {
1470 fseek(fp, f[pos - 1], SEEK_SET);
1471 nc = f[pos] - f[pos - 1];
1472 if (nc >= sizeof(buf))
1473 nc = sizeof(buf) - 1;
1474 nc = fread(buf, 1, nc, fp);
1475 if (nc == 0)
1476 continue;
1477 buf[nc] = '\0';
1478 buf[strcspn(buf, "\n")] = '\0';
1479 if (most_recent_pat != NULL) {
1480 int ret = regexec(&most_recent_re, buf, 0, NULL, 0);
1481
1482 if (ret != 0)
1483 continue;
1484 strlcpy(lastbuf, buf, sizeof(lastbuf));
1485 lastmatchline = pos;
1486 return (lastbuf);
1487 } else if (isalpha(buf[0]) || buf[0] == '_' || buf[0] == '$'
1488 || buf[0] == '-' || buf[0] == '+') {
1489 if (begins_with(buf, "private:")) {
1490 if (!state)
1491 state = " (private)";
1492 } else if (begins_with(buf, "protected:")) {
1493 if (!state)
1494 state = " (protected)";
1495 } else if (begins_with(buf, "public:")) {
1496 if (!state)
1497 state = " (public)";
1498 } else {
1499 strlcpy(lastbuf, buf, sizeof(lastbuf));
1500 if (state)
1501 strlcat(lastbuf, state, sizeof(lastbuf));
1502 lastmatchline = pos;
1503 return (lastbuf);
1504 }
1505 }
1506 }
1507 return (lastmatchline > 0 ? lastbuf : NULL);
1508 }
1509
1510 /* dump accumulated "context" diff changes */
1511 static void
dump_context_vec(FILE * f1,FILE * f2,int flags)1512 dump_context_vec(FILE *f1, FILE *f2, int flags)
1513 {
1514 struct context_vec *cvp = context_vec_start;
1515 int lowa, upb, lowc, upd, do_output;
1516 int a, b, c, d;
1517 char ch, *f;
1518
1519 if (context_vec_start > context_vec_ptr)
1520 return;
1521
1522 b = d = 0; /* gcc */
1523 if (ckd_sub(&lowa, cvp->a, diff_context) || lowa < 1)
1524 lowa = 1;
1525 if (ckd_add(&upb, context_vec_ptr->b, diff_context) || upb > (int)len[0])
1526 upb = (int)len[0];
1527 if (ckd_sub(&lowc, cvp->c, diff_context) || lowc < 1)
1528 lowc = 1;
1529 if (ckd_add(&upd, context_vec_ptr->d, diff_context) || upd > (int)len[1])
1530 upd = (int)len[1];
1531
1532 printf("***************");
1533 if (flags & (D_PROTOTYPE | D_MATCHLAST)) {
1534 f = match_function(ixold, cvp->a - 1, f1);
1535 if (f != NULL)
1536 printf(" %s", f);
1537 }
1538 printf("\n*** ");
1539 range(lowa, upb, ",");
1540 printf(" ****\n");
1541
1542 /*
1543 * Output changes to the "old" file. The first loop suppresses
1544 * output if there were no changes to the "old" file (we'll see
1545 * the "old" lines as context in the "new" list).
1546 */
1547 do_output = 0;
1548 for (; cvp <= context_vec_ptr; cvp++)
1549 if (cvp->a <= cvp->b) {
1550 cvp = context_vec_start;
1551 do_output++;
1552 break;
1553 }
1554 if (do_output) {
1555 while (cvp <= context_vec_ptr) {
1556 a = cvp->a;
1557 b = cvp->b;
1558 c = cvp->c;
1559 d = cvp->d;
1560
1561 if (a <= b && c <= d)
1562 ch = 'c';
1563 else
1564 ch = (a <= b) ? 'd' : 'a';
1565
1566 if (ch == 'a')
1567 fetch(ixold, lowa, b, f1, ' ', 0, flags);
1568 else {
1569 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1570 fetch(ixold, a, b, f1,
1571 ch == 'c' ? '!' : '-', 0, flags);
1572 }
1573 lowa = b + 1;
1574 cvp++;
1575 }
1576 fetch(ixold, b + 1, upb, f1, ' ', 0, flags);
1577 }
1578 /* output changes to the "new" file */
1579 printf("--- ");
1580 range(lowc, upd, ",");
1581 printf(" ----\n");
1582
1583 do_output = 0;
1584 for (cvp = context_vec_start; cvp <= context_vec_ptr; cvp++)
1585 if (cvp->c <= cvp->d) {
1586 cvp = context_vec_start;
1587 do_output++;
1588 break;
1589 }
1590 if (do_output) {
1591 while (cvp <= context_vec_ptr) {
1592 a = cvp->a;
1593 b = cvp->b;
1594 c = cvp->c;
1595 d = cvp->d;
1596
1597 if (a <= b && c <= d)
1598 ch = 'c';
1599 else
1600 ch = (a <= b) ? 'd' : 'a';
1601
1602 if (ch == 'd')
1603 fetch(ixnew, lowc, d, f2, ' ', 0, flags);
1604 else {
1605 fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1606 fetch(ixnew, c, d, f2,
1607 ch == 'c' ? '!' : '+', 0, flags);
1608 }
1609 lowc = d + 1;
1610 cvp++;
1611 }
1612 fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1613 }
1614 context_vec_ptr = context_vec_start - 1;
1615 }
1616
1617 /* dump accumulated "unified" diff changes */
1618 static void
dump_unified_vec(FILE * f1,FILE * f2,int flags)1619 dump_unified_vec(FILE *f1, FILE *f2, int flags)
1620 {
1621 struct context_vec *cvp = context_vec_start;
1622 int lowa, upb, lowc, upd;
1623 int a, b, c, d;
1624 char ch, *f;
1625
1626 if (context_vec_start > context_vec_ptr)
1627 return;
1628
1629 b = d = 0; /* gcc */
1630 if (ckd_sub(&lowa, cvp->a, diff_context) || lowa < 1)
1631 lowa = 1;
1632 if (ckd_add(&upb, context_vec_ptr->b, diff_context) || upb > (int)len[0])
1633 upb = (int)len[0];
1634 if (ckd_sub(&lowc, cvp->c, diff_context) || lowc < 1)
1635 lowc = 1;
1636 if (ckd_add(&upd, context_vec_ptr->d, diff_context) || upd > (int)len[1])
1637 upd = (int)len[1];
1638
1639 printf("@@ -");
1640 uni_range(lowa, upb);
1641 printf(" +");
1642 uni_range(lowc, upd);
1643 printf(" @@");
1644 if (flags & (D_PROTOTYPE | D_MATCHLAST)) {
1645 f = match_function(ixold, cvp->a - 1, f1);
1646 if (f != NULL)
1647 printf(" %s", f);
1648 }
1649 printf("\n");
1650
1651 /*
1652 * Output changes in "unified" diff format--the old and new lines
1653 * are printed together.
1654 */
1655 for (; cvp <= context_vec_ptr; cvp++) {
1656 a = cvp->a;
1657 b = cvp->b;
1658 c = cvp->c;
1659 d = cvp->d;
1660
1661 /*
1662 * c: both new and old changes
1663 * d: only changes in the old file
1664 * a: only changes in the new file
1665 */
1666 if (a <= b && c <= d)
1667 ch = 'c';
1668 else
1669 ch = (a <= b) ? 'd' : 'a';
1670
1671 switch (ch) {
1672 case 'c':
1673 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1674 fetch(ixold, a, b, f1, '-', 0, flags);
1675 fetch(ixnew, c, d, f2, '+', 0, flags);
1676 break;
1677 case 'd':
1678 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1679 fetch(ixold, a, b, f1, '-', 0, flags);
1680 break;
1681 case 'a':
1682 fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1683 fetch(ixnew, c, d, f2, '+', 0, flags);
1684 break;
1685 }
1686 lowa = b + 1;
1687 lowc = d + 1;
1688 }
1689 fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1690
1691 context_vec_ptr = context_vec_start - 1;
1692 }
1693
1694 static void
print_header(const char * file1,const char * file2)1695 print_header(const char *file1, const char *file2)
1696 {
1697 const char *time_format;
1698 char buf[256];
1699 struct tm tm1, tm2, *tm_ptr1, *tm_ptr2;
1700 int nsec1 = stb1.st_mtim.tv_nsec;
1701 int nsec2 = stb2.st_mtim.tv_nsec;
1702
1703 time_format = "%Y-%m-%d %H:%M:%S";
1704
1705 if (cflag)
1706 time_format = "%c";
1707 tm_ptr1 = localtime_r(&stb1.st_mtime, &tm1);
1708 tm_ptr2 = localtime_r(&stb2.st_mtime, &tm2);
1709 if (label[0] != NULL)
1710 printf("%s %s\n", diff_format == D_CONTEXT ? "***" : "---",
1711 label[0]);
1712 else {
1713 strftime(buf, sizeof(buf), time_format, tm_ptr1);
1714 printf("%s %s\t%s", diff_format == D_CONTEXT ? "***" : "---",
1715 file1, buf);
1716 if (!cflag) {
1717 strftime(buf, sizeof(buf), "%z", tm_ptr1);
1718 printf(".%.9d %s", nsec1, buf);
1719 }
1720 printf("\n");
1721 }
1722 if (label[1] != NULL)
1723 printf("%s %s\n", diff_format == D_CONTEXT ? "---" : "+++",
1724 label[1]);
1725 else {
1726 strftime(buf, sizeof(buf), time_format, tm_ptr2);
1727 printf("%s %s\t%s", diff_format == D_CONTEXT ? "---" : "+++",
1728 file2, buf);
1729 if (!cflag) {
1730 strftime(buf, sizeof(buf), "%z", tm_ptr2);
1731 printf(".%.9d %s", nsec2, buf);
1732 }
1733 printf("\n");
1734 }
1735 }
1736
1737 /*
1738 * Prints n number of space characters either by using tab
1739 * or single space characters.
1740 * nc is the preceding number of characters
1741 */
1742 static void
print_space(int nc,int n,int flags)1743 print_space(int nc, int n, int flags)
1744 {
1745 int col, newcol, tabstop;
1746
1747 col = nc;
1748 newcol = nc + n;
1749 /* first, use tabs if allowed */
1750 if ((flags & D_EXPANDTABS) == 0) {
1751 while ((tabstop = roundup(col + 1, tabsize)) <= newcol) {
1752 printf("\t");
1753 col = tabstop;
1754 }
1755 }
1756 /* finish with spaces */
1757 printf("%*s", newcol - col, "");
1758 }
1759