1 /* $OpenBSD: diffreg.c,v 1.93 2019/06/28 13:35:00 deraadt Exp $ */
4 * SPDX-License-Identifier: BSD-4-Clause
6 * Copyright (C) Caldera International Inc. 2001-2002.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
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
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.
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.
39 * Copyright (c) 1991, 1993
40 * The Regents of the University of California. All rights reserved.
42 * Redistribution and use in source and binary forms, with or without
43 * modification, are permitted provided that the following conditions
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.
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
66 * @(#)diffreg.c 8.1 (Berkeley) 6/6/93
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
72 #include <sys/capsicum.h>
75 #include <capsicum_helpers.h>
94 * diff - compare two files.
98 * Uses an algorithm due to Harold Stone, which finds
99 * a pair of longest identical subsequences in the two
102 * The major goal is to generate the match vector J.
103 * J[i] is the index of the line in file1 corresponding
104 * to line i file0. J[i] = 0 if there is no
105 * such line in file1.
107 * Lines are hashed so as to work in core. All potential
108 * matches are located by sorting the lines of each file
109 * on the hash (called ``value''). In particular, this
110 * collects the equivalence classes in file1 together.
111 * Subroutine equiv replaces the value of each line in
112 * file0 by the index of the first element of its
113 * matching equivalence in (the reordered) file1.
114 * To save space equiv squeezes file1 into a single
115 * array member in which the equivalence classes
116 * are simply concatenated, except that their first
117 * members are flagged by changing sign.
119 * Next the indices that point into member are unsorted into
120 * array class according to the original order of file0.
122 * The cleverness lies in routine stone. This marches
123 * through the lines of file0, developing a vector klist
124 * of "k-candidates". At step i a k-candidate is a matched
125 * pair of lines x,y (x in file0 y in file1) such that
126 * there is a common subsequence of length k
127 * between the first i lines of file0 and the first y
128 * lines of file1, but there is no such subsequence for
129 * any smaller y. x is the earliest possible mate to y
130 * that occurs in such a subsequence.
132 * Whenever any of the members of the equivalence class of
133 * lines in file1 matable to a line in file0 has serial number
134 * less than the y of some k-candidate, that k-candidate
135 * with the smallest such y is replaced. The new
136 * k-candidate is chained (via pred) to the current
137 * k-1 candidate so that the actual subsequence can
138 * be recovered. When a member has serial number greater
139 * that the y of all k-candidates, the klist is extended.
140 * At the end, the longest subsequence is pulled out
141 * and placed in the array J by unravel
143 * With J in hand, the matches there recorded are
144 * check'ed against reality to assure that no spurious
145 * matches have crept in due to hashing. If they have,
146 * they are broken, and "jackpot" is recorded--a harmless
147 * matter except that a true match for a spuriously
148 * mated line may now be unnecessarily reported as a change.
150 * Much of the complexity of the program comes simply
151 * from trying to minimize core utilization and
152 * maximize the range of doable problems by dynamically
153 * allocating what is needed and reusing what is not.
154 * The core requirements for problems larger than somewhat
155 * are (in words) 2*length(file0) + length(file1) +
156 * 3*(number of k-candidates installed), typically about
157 * 6n words for files of length n.
172 * The following struct is used to record change information when
173 * doing a "context" or "unified" diff. (see routine "change" to
174 * understand the highly mnemonic field names)
177 int a; /* start line in old file */
178 int b; /* end line in old file */
179 int c; /* start line in new file */
180 int d; /* end line in new file */
184 static FILE *opentemp(const char *);
185 static void output(char *, FILE *, char *, FILE *, int);
186 static void check(FILE *, FILE *, int);
187 static void range(int, int, const char *);
188 static void uni_range(int, int);
189 static void dump_context_vec(FILE *, FILE *, int);
190 static void dump_unified_vec(FILE *, FILE *, int);
191 static void prepare(int, FILE *, size_t, int);
192 static void prune(void);
193 static void equiv(struct line *, int, struct line *, int, int *);
194 static void unravel(int);
195 static void unsort(struct line *, int, int *);
196 static void change(char *, FILE *, char *, FILE *, int, int, int, int, int *);
197 static void sort(struct line *, int);
198 static void print_header(const char *, const char *);
199 static void print_space(int, int, int);
200 static bool ignoreline_pattern(char *);
201 static bool ignoreline(char *, bool);
202 static int asciifile(FILE *);
203 static int fetch(long *, int, int, FILE *, int, int, int);
204 static int newcand(int, int, int);
205 static int search(int *, int, int);
206 static int skipline(FILE *);
207 static int isqrt(int);
208 static int stone(int *, int, int *, int *, int);
209 static int readhash(FILE *, int);
210 static int files_differ(FILE *, FILE *, int);
211 static char *match_function(const long *, int, FILE *);
212 static char *preadline(int, size_t, off_t);
214 static int *J; /* will be overlaid on class */
215 static int *class; /* will be overlaid on file[0] */
216 static int *klist; /* will be overlaid on file[0] after class */
217 static int *member; /* will be overlaid on file[1] */
219 static int inifdef; /* whether or not we are in a #ifdef block */
221 static int pref, suff; /* length of prefix and suffix */
223 static int anychange;
224 static int hw, padding; /* half width and padding */
226 static long *ixnew; /* will be overlaid on file[1] */
227 static long *ixold; /* will be overlaid on klist */
228 static struct cand *clist; /* merely a free storage pot for candidates */
229 static int clistlen; /* the length of clist */
230 static struct line *sfile[2]; /* shortened by pruning common prefix/suffix */
231 static int (*chrtran)(int); /* translation table for case-folding */
232 static struct context_vec *context_vec_start;
233 static struct context_vec *context_vec_end;
234 static struct context_vec *context_vec_ptr;
236 #define FUNCTION_CONTEXT_SIZE 55
237 static char lastbuf[FUNCTION_CONTEXT_SIZE];
239 static int lastmatchline;
256 diffreg(char *file1, char *file2, int flags, int capsicum)
260 struct pr *pr = NULL;
261 cap_rights_t rights_ro;
268 context_vec_ptr = context_vec_start - 1;
271 * hw excludes padding and make sure when -t is not used,
272 * the second column always starts from the closest tab stop
274 if (diff_format == D_SIDEBYSIDE) {
276 padding = tabsize - (hw % tabsize);
277 if ((flags & D_EXPANDTABS) != 0 || (padding % tabsize == 0))
281 ((padding == MIN_PAD) ? (padding << 1) : padding) - 1;
285 if (flags & D_IGNORECASE)
289 if (S_ISDIR(stb1.st_mode) != S_ISDIR(stb2.st_mode))
290 return (S_ISDIR(stb1.st_mode) ? D_MISMATCH1 : D_MISMATCH2);
291 if (strcmp(file1, "-") == 0 && strcmp(file2, "-") == 0)
294 if (flags & D_EMPTY1)
295 f1 = fopen(_PATH_DEVNULL, "r");
297 if (!S_ISREG(stb1.st_mode)) {
298 if ((f1 = opentemp(file1)) == NULL ||
299 fstat(fileno(f1), &stb1) == -1) {
305 } else if (strcmp(file1, "-") == 0)
308 f1 = fopen(file1, "r");
317 if (flags & D_EMPTY2)
318 f2 = fopen(_PATH_DEVNULL, "r");
320 if (!S_ISREG(stb2.st_mode)) {
321 if ((f2 = opentemp(file2)) == NULL ||
322 fstat(fileno(f2), &stb2) == -1) {
328 } else if (strcmp(file2, "-") == 0)
331 f2 = fopen(file2, "r");
341 pr = start_pr(file1, file2);
344 cap_rights_init(&rights_ro, CAP_READ, CAP_FSTAT, CAP_SEEK);
345 if (caph_rights_limit(fileno(f1), &rights_ro) < 0)
346 err(2, "unable to limit rights on: %s", file1);
347 if (caph_rights_limit(fileno(f2), &rights_ro) < 0)
348 err(2, "unable to limit rights on: %s", file2);
349 if (fileno(f1) == STDIN_FILENO || fileno(f2) == STDIN_FILENO) {
350 /* stdin has already been limited */
351 if (caph_limit_stderr() == -1)
352 err(2, "unable to limit stderr");
353 if (caph_limit_stdout() == -1)
354 err(2, "unable to limit stdout");
355 } else if (caph_limit_stdio() == -1)
356 err(2, "unable to limit stdio");
358 caph_cache_catpages();
360 if (caph_enter() < 0)
361 err(2, "unable to enter capability mode");
364 switch (files_differ(f1, f2, flags)) {
376 if (diff_format == D_BRIEF && ignore_pats == NULL &&
377 (flags & (D_FOLDBLANKS|D_IGNOREBLANKS|D_IGNORECASE|D_STRIPCR)) == 0)
383 if ((flags & D_FORCEASCII) == 0 &&
384 (!asciifile(f1) || !asciifile(f2))) {
389 prepare(0, f1, stb1.st_size, flags);
390 prepare(1, f2, stb2.st_size, flags);
393 sort(sfile[0], slen[0]);
394 sort(sfile[1], slen[1]);
396 member = (int *)file[1];
397 equiv(sfile[0], slen[0], sfile[1], slen[1], member);
398 member = xreallocarray(member, slen[1] + 2, sizeof(*member));
400 class = (int *)file[0];
401 unsort(sfile[0], slen[0], class);
402 class = xreallocarray(class, slen[0] + 2, sizeof(*class));
404 klist = xcalloc(slen[0] + 2, sizeof(*klist));
407 clist = xcalloc(clistlen, sizeof(*clist));
408 i = stone(class, slen[0], member, klist, flags);
412 J = xreallocarray(J, len[0] + 2, sizeof(*J));
417 ixold = xreallocarray(ixold, len[0] + 2, sizeof(*ixold));
418 ixnew = xreallocarray(ixnew, len[1] + 2, sizeof(*ixnew));
419 check(f1, f2, flags);
420 output(file1, f1, file2, f2, flags);
439 * Check to see if the given files differ.
440 * Returns 0 if they are the same, 1 if different, and -1 on error.
441 * XXX - could use code from cmp(1) [faster]
444 files_differ(FILE *f1, FILE *f2, int flags)
446 char buf1[BUFSIZ], buf2[BUFSIZ];
449 if ((flags & (D_EMPTY1|D_EMPTY2)) || stb1.st_size != stb2.st_size ||
450 (stb1.st_mode & S_IFMT) != (stb2.st_mode & S_IFMT))
453 i = fread(buf1, 1, sizeof(buf1), f1);
454 j = fread(buf2, 1, sizeof(buf2), f2);
455 if ((!i && ferror(f1)) || (!j && ferror(f2)))
461 if (memcmp(buf1, buf2, i) != 0)
467 opentemp(const char *f)
469 char buf[BUFSIZ], tempfile[PATH_MAX];
473 if (strcmp(f, "-") == 0)
475 else if ((ifd = open(f, O_RDONLY, 0644)) == -1)
478 (void)strlcpy(tempfile, _PATH_TMP "/diff.XXXXXXXX", sizeof(tempfile));
480 if ((ofd = mkstemp(tempfile)) == -1) {
485 while ((nread = read(ifd, buf, BUFSIZ)) > 0) {
486 if (write(ofd, buf, nread) != nread) {
493 lseek(ofd, (off_t)0, SEEK_SET);
494 return (fdopen(ofd, "r"));
498 splice(char *dir, char *path)
503 dirlen = strlen(dir);
504 while (dirlen != 0 && dir[dirlen - 1] == '/')
506 if ((tail = strrchr(path, '/')) == NULL)
510 xasprintf(&buf, "%.*s/%s", (int)dirlen, dir, tail);
515 prepare(int i, FILE *fd, size_t filesize, int flags)
523 sz = MIN(filesize, SIZE_MAX) / 25;
527 p = xcalloc(sz + 3, sizeof(*p));
528 for (j = 0; (h = readhash(fd, flags));) {
531 p = xreallocarray(p, sz + 3, sizeof(*p));
544 for (pref = 0; pref < len[0] && pref < len[1] &&
545 file[0][pref + 1].value == file[1][pref + 1].value;
548 for (suff = 0; suff < len[0] - pref && suff < len[1] - pref &&
549 file[0][len[0] - suff].value == file[1][len[1] - suff].value;
552 for (j = 0; j < 2; j++) {
553 sfile[j] = file[j] + pref;
554 slen[j] = len[j] - pref - suff;
555 for (i = 0; i <= slen[j]; i++)
556 sfile[j][i].serial = i;
561 equiv(struct line *a, int n, struct line *b, int m, int *c)
566 while (i <= n && j <= m) {
567 if (a[i].value < b[j].value)
569 else if (a[i].value == b[j].value)
580 while (b[j + 1].value == b[j].value) {
588 /* Code taken from ping.c */
597 do { /* newton was a stinker */
602 } while ((x - y) > 1 || (x - y) < -1);
608 stone(int *a, int n, int *b, int *c, int flags)
611 int oldc, tc, oldl, sq;
612 u_int numtries, bound;
614 if (flags & D_MINIMAL)
618 bound = MAX(256, sq);
622 c[0] = newcand(0, 0, 0);
623 for (i = 1; i <= n; i++) {
632 if (y <= clist[oldc].y)
638 if (clist[c[l]].y <= y)
641 c[l] = newcand(i, y, oldc);
646 c[l] = newcand(i, y, oldc);
650 } while ((y = b[++j]) > 0 && numtries < bound);
656 newcand(int x, int y, int pred)
660 if (clen == clistlen) {
661 clistlen = clistlen * 11 / 10;
662 clist = xreallocarray(clist, clistlen, sizeof(*clist));
672 search(int *c, int k, int y)
676 if (clist[c[k]].y < y) /* quick look for typical case */
701 for (i = 0; i <= len[0]; i++)
702 J[i] = i <= pref ? i :
703 i > len[0] - suff ? i + len[1] - len[0] : 0;
704 for (q = clist + p; q->y != 0; q = clist + q->pred)
705 J[q->x + pref] = q->y + pref;
709 * Check does double duty:
710 * 1. ferret out any fortuitous correspondences due
711 * to confounding by hashing (which result in "jackpot")
712 * 2. collect random access indexes to the two files
715 check(FILE *f1, FILE *f2, int flags)
717 int i, j, jackpot, c, d;
723 ixold[0] = ixnew[0] = 0;
726 for (i = 1; i <= len[0]; i++) {
728 ixold[i] = ctold += skipline(f1);
732 ixnew[j] = ctnew += skipline(f2);
735 if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS|D_IGNORECASE|D_STRIPCR)) {
740 * GNU diff ignores a missing newline
741 * in one file for -b or -w.
743 if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) {
744 if (c == EOF && d == '\n') {
747 } else if (c == '\n' && d == EOF) {
754 if (flags & D_STRIPCR && (c == '\r' || d == '\r')) {
756 if ((c = getc(f1)) == '\n') {
763 if ((d = getc(f2)) == '\n') {
771 if ((flags & D_FOLDBLANKS) && isspace(c) &&
777 } while (isspace(c = getc(f1)));
782 } while (isspace(d = getc(f2)));
783 } else if ((flags & D_IGNOREBLANKS)) {
784 while (isspace(c) && c != '\n') {
788 while (isspace(d) && d != '\n') {
793 if (chrtran(c) != chrtran(d)) {
796 if (c != '\n' && c != EOF)
797 ctold += skipline(f1);
798 if (d != '\n' && c != EOF)
799 ctnew += skipline(f2);
802 if (c == '\n' || c == EOF)
809 if ((c = getc(f1)) != (d = getc(f2))) {
812 if (c != '\n' && c != EOF)
813 ctold += skipline(f1);
814 if (d != '\n' && c != EOF)
815 ctnew += skipline(f2);
818 if (c == '\n' || c == EOF)
826 for (; j <= len[1]; j++) {
827 ixnew[j] = ctnew += skipline(f2);
831 * fprintf(stderr, "jackpot\n");
835 /* shellsort CACM #201 */
837 sort(struct line *a, int n)
839 struct line *ai, *aim, w;
844 for (j = 1; j <= n; j *= 2)
846 for (m /= 2; m != 0; m /= 2) {
848 for (j = 1; j <= k; j++) {
849 for (ai = &a[j]; ai > a; ai -= m) {
852 break; /* wraparound */
853 if (aim->value > ai[0].value ||
854 (aim->value == ai[0].value &&
855 aim->serial > ai[0].serial))
857 w.value = ai[0].value;
858 ai[0].value = aim->value;
859 aim->value = w.value;
860 w.serial = ai[0].serial;
861 ai[0].serial = aim->serial;
862 aim->serial = w.serial;
869 unsort(struct line *f, int l, int *b)
873 a = xcalloc(l + 1, sizeof(*a));
874 for (i = 1; i <= l; i++)
875 a[f[i].serial] = f[i].value;
876 for (i = 1; i <= l; i++)
886 for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++)
892 output(char *file1, FILE *f1, char *file2, FILE *f2, int flags)
894 int i, j, m, i0, i1, j0, j1, nc;
900 J[m + 1] = len[1] + 1;
901 if (diff_format != D_EDIT) {
902 for (i0 = 1; i0 <= m; i0 = i1 + 1) {
903 while (i0 <= m && J[i0] == J[i0 - 1] + 1){
904 if (diff_format == D_SIDEBYSIDE &&
905 suppress_common != 1) {
906 nc = fetch(ixold, i0, i0, f1, '\0',
909 (hw - nc) + (padding << 1) + 1,
911 fetch(ixnew, J[i0], J[i0], f2, '\0',
919 while (i1 < m && J[i1 + 1] == 0)
925 * When using side-by-side, lines from both of the
926 * files are printed. The algorithm used by diff(1)
927 * identifies the ranges in which two files differ.
928 * See the change() function below.
929 * The for loop below consumes the shorter range,
930 * whereas one of the while loops deals with the
933 if (diff_format == D_SIDEBYSIDE) {
934 for (i=i0, j=j0; i<=i1 && j<=j1; i++, j++)
935 change(file1, f1, file2, f2, i, i,
939 change(file1, f1, file2, f2,
940 i, i, j+1, j, &flags);
945 change(file1, f1, file2, f2,
946 i+1, i, j, j, &flags);
950 change(file1, f1, file2, f2, i0, i1, j0,
954 for (i0 = m; i0 >= 1; i0 = i1 - 1) {
955 while (i0 >= 1 && J[i0] == J[i0 + 1] - 1 && J[i0] != 0)
959 while (i1 > 1 && J[i1 - 1] == 0)
963 change(file1, f1, file2, f2, i1, i0, j1, j0, &flags);
967 change(file1, f1, file2, f2, 1, 0, 1, len[1], &flags);
968 if (diff_format == D_IFDEF || diff_format == D_GFORMAT) {
971 if ((c = getc(f1)) == EOF)
977 if (anychange != 0) {
978 if (diff_format == D_CONTEXT)
979 dump_context_vec(f1, f2, flags);
980 else if (diff_format == D_UNIFIED)
981 dump_unified_vec(f1, f2, flags);
986 range(int a, int b, const char *separator)
988 printf("%d", a > b ? b : a);
990 printf("%s%d", separator, b);
994 uni_range(int a, int b)
997 printf("%d,%d", a, b - a + 1);
1005 preadline(int fd, size_t rlen, off_t off)
1010 line = xmalloc(rlen + 1);
1011 if ((nr = pread(fd, line, rlen, off)) == -1)
1012 err(2, "preadline");
1013 if (nr > 0 && line[nr-1] == '\n')
1020 ignoreline_pattern(char *line)
1024 ret = regexec(&ignore_re, line, 0, NULL, 0);
1026 return (ret == 0); /* if it matched, it should be ignored. */
1030 ignoreline(char *line, bool skip_blanks)
1033 if (ignore_pats != NULL && skip_blanks)
1034 return (ignoreline_pattern(line) || *line == '\0');
1035 if (ignore_pats != NULL)
1036 return (ignoreline_pattern(line));
1038 return (*line == '\0');
1039 /* No ignore criteria specified */
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.
1051 change(char *file1, FILE *f1, char *file2, FILE *f2, int a, int b, int c, int d,
1054 static size_t max_context = 64;
1060 skip_blanks = (*pflags & D_SKIPBLANKLINES);
1062 if ((diff_format != D_IFDEF || diff_format == D_GFORMAT) &&
1065 if (ignore_pats != NULL || skip_blanks) {
1068 * All lines in the change, insert, or delete must
1069 * match an ignore pattern for the change to be
1072 if (a <= b) { /* Changes and deletes. */
1073 for (i = a; i <= b; i++) {
1074 line = preadline(fileno(f1),
1075 ixold[i] - ixold[i - 1], ixold[i - 1]);
1076 if (!ignoreline(line, skip_blanks))
1080 if (a > b || c <= d) { /* Changes and inserts. */
1081 for (i = c; i <= d; i++) {
1082 line = preadline(fileno(f2),
1083 ixnew[i] - ixnew[i - 1], ixnew[i - 1]);
1084 if (!ignoreline(line, skip_blanks))
1091 if (*pflags & D_HEADER && diff_format != D_BRIEF) {
1092 printf("%s %s %s\n", diffargs, file1, file2);
1093 *pflags &= ~D_HEADER;
1095 if (diff_format == D_CONTEXT || diff_format == D_UNIFIED) {
1097 * Allocate change records as needed.
1099 if (context_vec_ptr == context_vec_end - 1) {
1100 ptrdiff_t offset = context_vec_ptr - context_vec_start;
1102 context_vec_start = xreallocarray(context_vec_start,
1103 max_context, sizeof(*context_vec_start));
1104 context_vec_end = context_vec_start + max_context;
1105 context_vec_ptr = context_vec_start + offset;
1107 if (anychange == 0) {
1109 * Print the context/unidiff header first time through.
1111 print_header(file1, file2);
1113 } else if (a > context_vec_ptr->b + (2 * diff_context) + 1 &&
1114 c > context_vec_ptr->d + (2 * diff_context) + 1) {
1116 * If this change is more than 'diff_context' lines from the
1117 * previous change, dump the record and reset it.
1119 if (diff_format == D_CONTEXT)
1120 dump_context_vec(f1, f2, *pflags);
1122 dump_unified_vec(f1, f2, *pflags);
1125 context_vec_ptr->a = a;
1126 context_vec_ptr->b = b;
1127 context_vec_ptr->c = c;
1128 context_vec_ptr->d = d;
1133 switch (diff_format) {
1139 printf("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1140 if (diff_format == D_NORMAL)
1145 printf("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1151 printf("a%d %d\n", b, d - c + 1);
1153 printf("d%d %d\n", a, b - a + 1);
1155 /* add changed lines */
1156 printf("a%d %d\n", b, d - c + 1);
1160 if (diff_format == D_GFORMAT) {
1162 /* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1163 nc = ixold[a > b ? b : a - 1] - curpos;
1164 for (i = 0; i < nc; i++)
1165 printf("%c", getc(f1));
1166 for (walk = group_format; *walk != '\0'; walk++) {
1171 fetch(ixold, a, b, f1, '<', 1, *pflags);
1174 fetch(ixnew, c, d, f2, '>', 0, *pflags);
1177 printf("%%%c", *walk);
1182 printf("%c", *walk);
1185 if (diff_format == D_SIDEBYSIDE) {
1187 print_space(0, hw + padding , *pflags);
1189 nc = fetch(ixold, a, b, f1, '\0', 1, *pflags);
1190 print_space(nc, hw - nc + padding, *pflags);
1192 printf("%c", (a>b)? '>' : ((c>d)? '<' : '|'));
1193 print_space(hw + padding + 1 , padding, *pflags);
1194 fetch(ixnew, c, d, f2, '\0', 0, *pflags);
1197 if (diff_format == D_NORMAL || diff_format == D_IFDEF) {
1198 fetch(ixold, a, b, f1, '<', 1, *pflags);
1199 if (a <= b && c <= d && diff_format == D_NORMAL)
1202 if (diff_format != D_GFORMAT && diff_format != D_SIDEBYSIDE)
1203 fetch(ixnew, c, d, f2, diff_format == D_NORMAL ? '>' : '\0', 0, *pflags);
1204 if (edoffset != 0 && diff_format == D_EDIT) {
1206 * A non-zero edoffset value for D_EDIT indicates that the
1207 * last line printed was a bare dot (".") that has been
1208 * escaped as ".." to prevent ed(1) from misinterpreting
1209 * it. We have to add a substitute command to change this
1210 * back and restart where we left off.
1213 printf("%ds/.//\n", a + edoffset - 1);
1214 b = a + edoffset - 1;
1219 if ((diff_format == D_EDIT || diff_format == D_REVERSE) && c <= d)
1222 printf("#endif /* %s */\n", ifdefname);
1228 fetch(long *f, int a, int b, FILE *lb, int ch, int oldfile, int flags)
1230 int i, j, c, lastc, col, nc, newcol;
1235 * When doing #ifdef's, copy down to current line
1236 * if this is the first file, so that stuff makes it to output.
1238 if ((diff_format == D_IFDEF) && oldfile) {
1239 long curpos = ftell(lb);
1240 /* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1241 nc = f[a > b ? b : a - 1] - curpos;
1242 for (i = 0; i < nc; i++)
1243 printf("%c", getc(lb));
1247 if (diff_format == D_IFDEF) {
1249 printf("#else /* %s%s */\n",
1250 oldfile == 1 ? "!" : "", ifdefname);
1253 printf("#ifndef %s\n", ifdefname);
1255 printf("#ifdef %s\n", ifdefname);
1257 inifdef = 1 + oldfile;
1259 for (i = a; i <= b; i++) {
1260 fseek(lb, f[i - 1], SEEK_SET);
1261 nc = (f[i] - f[i - 1]);
1262 if (diff_format == D_SIDEBYSIDE && hw < nc)
1264 if ((diff_format != D_IFDEF && diff_format != D_GFORMAT) &&
1267 if (Tflag && (diff_format == D_NORMAL ||
1268 diff_format == D_CONTEXT ||
1269 diff_format == D_UNIFIED))
1271 else if (diff_format != D_UNIFIED)
1275 for (j = 0, lastc = '\0'; j < nc; j++, lastc = c) {
1276 if ((c = getc(lb)) == EOF) {
1277 if (diff_format == D_EDIT ||
1278 diff_format == D_REVERSE ||
1279 diff_format == D_NREVERSE)
1280 warnx("No newline at end of file");
1282 printf("\n\\ No newline at end of "
1287 * when using --side-by-side, col needs to be increased
1288 * in any case to keep the columns aligned
1291 if (flags & D_EXPANDTABS) {
1292 newcol = ((col/tabsize)+1)*tabsize;
1294 if (diff_format == D_SIDEBYSIDE)
1297 } while (++col < newcol && j < nc);
1299 if (diff_format == D_SIDEBYSIDE) {
1300 if ((j + tabsize) > nc) {
1315 if (diff_format == D_EDIT && j == 1 && c == '\n'
1318 * Don't print a bare "." line
1319 * since that will confuse ed(1).
1320 * Print ".." instead and set the,
1321 * global variable edoffset to an
1322 * offset from which to restart.
1323 * The caller must check the value
1327 edoffset = i - a + 1;
1330 /* when side-by-side, do not print a newline */
1331 if (diff_format != D_SIDEBYSIDE || c != '\n') {
1342 * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578.
1345 readhash(FILE *f, int flags)
1353 switch (t = getc(f)) {
1355 if (flags & D_STRIPCR) {
1366 if ((flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) != 0) {
1372 if (space && (flags & D_IGNOREBLANKS) == 0) {
1376 sum = sum * 127 + chrtran(t);
1389 * There is a remote possibility that we end up with a zero sum.
1390 * Zero is used as an EOF marker, so return 1 instead.
1392 return (sum == 0 ? 1 : sum);
1398 unsigned char buf[BUFSIZ];
1405 cnt = fread(buf, 1, sizeof(buf), f);
1406 return (memchr(buf, '\0', cnt) == NULL);
1409 #define begins_with(s, pre) (strncmp(s, pre, sizeof(pre)-1) == 0)
1412 match_function(const long *f, int pos, FILE *fp)
1414 unsigned char buf[FUNCTION_CONTEXT_SIZE];
1416 int last = lastline;
1417 const char *state = NULL;
1420 while (pos > last) {
1421 fseek(fp, f[pos - 1], SEEK_SET);
1422 nc = f[pos] - f[pos - 1];
1423 if (nc >= sizeof(buf))
1424 nc = sizeof(buf) - 1;
1425 nc = fread(buf, 1, nc, fp);
1428 buf[strcspn(buf, "\n")] = '\0';
1429 if (isalpha(buf[0]) || buf[0] == '_' || buf[0] == '$') {
1430 if (begins_with(buf, "private:")) {
1432 state = " (private)";
1433 } else if (begins_with(buf, "protected:")) {
1435 state = " (protected)";
1436 } else if (begins_with(buf, "public:")) {
1438 state = " (public)";
1440 strlcpy(lastbuf, buf, sizeof lastbuf);
1442 strlcat(lastbuf, state,
1444 lastmatchline = pos;
1451 return lastmatchline > 0 ? lastbuf : NULL;
1454 /* dump accumulated "context" diff changes */
1456 dump_context_vec(FILE *f1, FILE *f2, int flags)
1458 struct context_vec *cvp = context_vec_start;
1459 int lowa, upb, lowc, upd, do_output;
1463 if (context_vec_start > context_vec_ptr)
1466 b = d = 0; /* gcc */
1467 lowa = MAX(1, cvp->a - diff_context);
1468 upb = MIN(len[0], context_vec_ptr->b + diff_context);
1469 lowc = MAX(1, cvp->c - diff_context);
1470 upd = MIN(len[1], context_vec_ptr->d + diff_context);
1472 printf("***************");
1473 if ((flags & D_PROTOTYPE)) {
1474 f = match_function(ixold, lowa-1, f1);
1479 range(lowa, upb, ",");
1483 * Output changes to the "old" file. The first loop suppresses
1484 * output if there were no changes to the "old" file (we'll see
1485 * the "old" lines as context in the "new" list).
1488 for (; cvp <= context_vec_ptr; cvp++)
1489 if (cvp->a <= cvp->b) {
1490 cvp = context_vec_start;
1495 while (cvp <= context_vec_ptr) {
1501 if (a <= b && c <= d)
1504 ch = (a <= b) ? 'd' : 'a';
1507 fetch(ixold, lowa, b, f1, ' ', 0, flags);
1509 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1510 fetch(ixold, a, b, f1,
1511 ch == 'c' ? '!' : '-', 0, flags);
1516 fetch(ixold, b + 1, upb, f1, ' ', 0, flags);
1518 /* output changes to the "new" file */
1520 range(lowc, upd, ",");
1524 for (cvp = context_vec_start; cvp <= context_vec_ptr; cvp++)
1525 if (cvp->c <= cvp->d) {
1526 cvp = context_vec_start;
1531 while (cvp <= context_vec_ptr) {
1537 if (a <= b && c <= d)
1540 ch = (a <= b) ? 'd' : 'a';
1543 fetch(ixnew, lowc, d, f2, ' ', 0, flags);
1545 fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1546 fetch(ixnew, c, d, f2,
1547 ch == 'c' ? '!' : '+', 0, flags);
1552 fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1554 context_vec_ptr = context_vec_start - 1;
1557 /* dump accumulated "unified" diff changes */
1559 dump_unified_vec(FILE *f1, FILE *f2, int flags)
1561 struct context_vec *cvp = context_vec_start;
1562 int lowa, upb, lowc, upd;
1566 if (context_vec_start > context_vec_ptr)
1569 b = d = 0; /* gcc */
1570 lowa = MAX(1, cvp->a - diff_context);
1571 upb = MIN(len[0], context_vec_ptr->b + diff_context);
1572 lowc = MAX(1, cvp->c - diff_context);
1573 upd = MIN(len[1], context_vec_ptr->d + diff_context);
1576 uni_range(lowa, upb);
1578 uni_range(lowc, upd);
1580 if ((flags & D_PROTOTYPE)) {
1581 f = match_function(ixold, lowa-1, f1);
1588 * Output changes in "unified" diff format--the old and new lines
1589 * are printed together.
1591 for (; cvp <= context_vec_ptr; cvp++) {
1598 * c: both new and old changes
1599 * d: only changes in the old file
1600 * a: only changes in the new file
1602 if (a <= b && c <= d)
1605 ch = (a <= b) ? 'd' : 'a';
1609 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1610 fetch(ixold, a, b, f1, '-', 0, flags);
1611 fetch(ixnew, c, d, f2, '+', 0, flags);
1614 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1615 fetch(ixold, a, b, f1, '-', 0, flags);
1618 fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1619 fetch(ixnew, c, d, f2, '+', 0, flags);
1625 fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1627 context_vec_ptr = context_vec_start - 1;
1631 print_header(const char *file1, const char *file2)
1633 const char *time_format;
1638 struct tm tm1, tm2, *tm_ptr1, *tm_ptr2;
1639 int nsec1 = stb1.st_mtim.tv_nsec;
1640 int nsec2 = stb2.st_mtim.tv_nsec;
1642 time_format = "%Y-%m-%d %H:%M:%S";
1646 tm_ptr1 = localtime_r(&stb1.st_mtime, &tm1);
1647 tm_ptr2 = localtime_r(&stb2.st_mtime, &tm2);
1648 strftime(buf1, 256, time_format, tm_ptr1);
1649 strftime(buf2, 256, time_format, tm_ptr2);
1651 strftime(end1, 10, "%z", tm_ptr1);
1652 strftime(end2, 10, "%z", tm_ptr2);
1653 sprintf(buf1, "%s.%.9d %s", buf1, nsec1, end1);
1654 sprintf(buf2, "%s.%.9d %s", buf2, nsec2, end2);
1656 if (label[0] != NULL)
1657 printf("%s %s\n", diff_format == D_CONTEXT ? "***" : "---",
1660 printf("%s %s\t%s\n", diff_format == D_CONTEXT ? "***" : "---",
1662 if (label[1] != NULL)
1663 printf("%s %s\n", diff_format == D_CONTEXT ? "---" : "+++",
1666 printf("%s %s\t%s\n", diff_format == D_CONTEXT ? "---" : "+++",
1671 * Prints n number of space characters either by using tab
1672 * or single space characters.
1673 * nc is the preceding number of characters
1676 print_space(int nc, int n, int flags) {
1680 if ((flags & D_EXPANDTABS) == 0) {
1681 /* first tabstop may be closer than tabsize */
1682 i = tabsize - (nc % tabsize);
1683 while (col >= tabsize) {
1689 printf("%*s", col, "");