1 /* $OpenBSD: diffreg.c,v 1.91 2016/03/01 20:57:35 natano Exp $ */
4 * Copyright (C) Caldera International Inc. 2001-2002.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code and documentation must retain the above
11 * copyright notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. All advertising materials mentioning features or use of this software
16 * must display the following acknowledgement:
17 * This product includes software developed or owned by Caldera
19 * 4. Neither the name of Caldera International, Inc. nor the names of other
20 * contributors may be used to endorse or promote products derived from
21 * this software without specific prior written permission.
23 * USE OF THE SOFTWARE PROVIDED FOR UNDER THIS LICENSE BY CALDERA
24 * INTERNATIONAL, INC. AND CONTRIBUTORS ``AS IS'' AND ANY EXPRESS OR
25 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
26 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
27 * IN NO EVENT SHALL CALDERA INTERNATIONAL, INC. BE LIABLE FOR ANY DIRECT,
28 * INDIRECT INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
29 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
30 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
32 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
33 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
37 * Copyright (c) 1991, 1993
38 * The Regents of the University of California. All rights reserved.
40 * Redistribution and use in source and binary forms, with or without
41 * modification, are permitted provided that the following conditions
43 * 1. Redistributions of source code must retain the above copyright
44 * notice, this list of conditions and the following disclaimer.
45 * 2. Redistributions in binary form must reproduce the above copyright
46 * notice, this list of conditions and the following disclaimer in the
47 * documentation and/or other materials provided with the distribution.
48 * 3. Neither the name of the University nor the names of its contributors
49 * may be used to endorse or promote products derived from this software
50 * without specific prior written permission.
52 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64 * @(#)diffreg.c 8.1 (Berkeley) 6/6/93
67 #include <sys/cdefs.h>
68 __FBSDID("$FreeBSD$");
70 #include <sys/capsicum.h>
71 #include <sys/procdesc.h>
73 #include <sys/types.h>
74 #include <sys/event.h>
77 #include <capsicum_helpers.h>
97 #define _PATH_PR "/usr/bin/pr"
100 * diff - compare two files.
104 * Uses an algorithm due to Harold Stone, which finds
105 * a pair of longest identical subsequences in the two
108 * The major goal is to generate the match vector J.
109 * J[i] is the index of the line in file1 corresponding
110 * to line i file0. J[i] = 0 if there is no
111 * such line in file1.
113 * Lines are hashed so as to work in core. All potential
114 * matches are located by sorting the lines of each file
115 * on the hash (called ``value''). In particular, this
116 * collects the equivalence classes in file1 together.
117 * Subroutine equiv replaces the value of each line in
118 * file0 by the index of the first element of its
119 * matching equivalence in (the reordered) file1.
120 * To save space equiv squeezes file1 into a single
121 * array member in which the equivalence classes
122 * are simply concatenated, except that their first
123 * members are flagged by changing sign.
125 * Next the indices that point into member are unsorted into
126 * array class according to the original order of file0.
128 * The cleverness lies in routine stone. This marches
129 * through the lines of file0, developing a vector klist
130 * of "k-candidates". At step i a k-candidate is a matched
131 * pair of lines x,y (x in file0 y in file1) such that
132 * there is a common subsequence of length k
133 * between the first i lines of file0 and the first y
134 * lines of file1, but there is no such subsequence for
135 * any smaller y. x is the earliest possible mate to y
136 * that occurs in such a subsequence.
138 * Whenever any of the members of the equivalence class of
139 * lines in file1 matable to a line in file0 has serial number
140 * less than the y of some k-candidate, that k-candidate
141 * with the smallest such y is replaced. The new
142 * k-candidate is chained (via pred) to the current
143 * k-1 candidate so that the actual subsequence can
144 * be recovered. When a member has serial number greater
145 * that the y of all k-candidates, the klist is extended.
146 * At the end, the longest subsequence is pulled out
147 * and placed in the array J by unravel
149 * With J in hand, the matches there recorded are
150 * check'ed against reality to assure that no spurious
151 * matches have crept in due to hashing. If they have,
152 * they are broken, and "jackpot" is recorded--a harmless
153 * matter except that a true match for a spuriously
154 * mated line may now be unnecessarily reported as a change.
156 * Much of the complexity of the program comes simply
157 * from trying to minimize core utilization and
158 * maximize the range of doable problems by dynamically
159 * allocating what is needed and reusing what is not.
160 * The core requirements for problems larger than somewhat
161 * are (in words) 2*length(file0) + length(file1) +
162 * 3*(number of k-candidates installed), typically about
163 * 6n words for files of length n.
178 * The following struct is used to record change information when
179 * doing a "context" or "unified" diff. (see routine "change" to
180 * understand the highly mnemonic field names)
183 int a; /* start line in old file */
184 int b; /* end line in old file */
185 int c; /* start line in new file */
186 int d; /* end line in new file */
189 #define diff_output printf
190 static FILE *opentemp(const char *);
191 static void output(char *, FILE *, char *, FILE *, int);
192 static void check(FILE *, FILE *, int);
193 static void range(int, int, const char *);
194 static void uni_range(int, int);
195 static void dump_context_vec(FILE *, FILE *, int);
196 static void dump_unified_vec(FILE *, FILE *, int);
197 static void prepare(int, FILE *, size_t, int);
198 static void prune(void);
199 static void equiv(struct line *, int, struct line *, int, int *);
200 static void unravel(int);
201 static void unsort(struct line *, int, int *);
202 static void change(char *, FILE *, char *, FILE *, int, int, int, int, int *);
203 static void sort(struct line *, int);
204 static void print_header(const char *, const char *);
205 static int ignoreline(char *);
206 static int asciifile(FILE *);
207 static int fetch(long *, int, int, FILE *, int, int, int);
208 static int newcand(int, int, int);
209 static int search(int *, int, int);
210 static int skipline(FILE *);
211 static int isqrt(int);
212 static int stone(int *, int, int *, int *, int);
213 static int readhash(FILE *, int);
214 static int files_differ(FILE *, FILE *, int);
215 static char *match_function(const long *, int, FILE *);
216 static char *preadline(int, size_t, off_t);
218 static int *J; /* will be overlaid on class */
219 static int *class; /* will be overlaid on file[0] */
220 static int *klist; /* will be overlaid on file[0] after class */
221 static int *member; /* will be overlaid on file[1] */
223 static int inifdef; /* whether or not we are in a #ifdef block */
225 static int pref, suff; /* length of prefix and suffix */
227 static int anychange;
228 static long *ixnew; /* will be overlaid on file[1] */
229 static long *ixold; /* will be overlaid on klist */
230 static struct cand *clist; /* merely a free storage pot for candidates */
231 static int clistlen; /* the length of clist */
232 static struct line *sfile[2]; /* shortened by pruning common prefix/suffix */
233 static int (*chrtran)(int); /* translation table for case-folding */
234 static struct context_vec *context_vec_start;
235 static struct context_vec *context_vec_end;
236 static struct context_vec *context_vec_ptr;
238 #define FUNCTION_CONTEXT_SIZE 55
239 static char lastbuf[FUNCTION_CONTEXT_SIZE];
241 static int lastmatchline;
258 diffreg(char *file1, char *file2, int flags, int capsicum)
265 cap_rights_t rights_ro;
274 context_vec_ptr = context_vec_start - 1;
275 if (flags & D_IGNORECASE)
279 if (S_ISDIR(stb1.st_mode) != S_ISDIR(stb2.st_mode))
280 return (S_ISDIR(stb1.st_mode) ? D_MISMATCH1 : D_MISMATCH2);
281 if (strcmp(file1, "-") == 0 && strcmp(file2, "-") == 0)
284 if (flags & D_EMPTY1)
285 f1 = fopen(_PATH_DEVNULL, "r");
287 if (!S_ISREG(stb1.st_mode)) {
288 if ((f1 = opentemp(file1)) == NULL ||
289 fstat(fileno(f1), &stb1) < 0) {
294 } else if (strcmp(file1, "-") == 0)
297 f1 = fopen(file1, "r");
305 if (flags & D_EMPTY2)
306 f2 = fopen(_PATH_DEVNULL, "r");
308 if (!S_ISREG(stb2.st_mode)) {
309 if ((f2 = opentemp(file2)) == NULL ||
310 fstat(fileno(f2), &stb2) < 0) {
315 } else if (strcmp(file2, "-") == 0)
318 f2 = fopen(file2, "r");
327 /* redirect stdout to pr */
332 xasprintf(&header, "%s %s %s", diffargs, file1, file2);
333 signal(SIGPIPE, SIG_IGN);
337 switch ((pid = pdfork(&pr_pd, PD_CLOEXEC))) {
341 err(2, "No more processes");
344 if (pfd[0] != STDIN_FILENO) {
345 dup2(pfd[0], STDIN_FILENO);
349 execl(_PATH_PR, _PATH_PR, "-h", header, (char *)0);
354 if (pfd[1] != STDOUT_FILENO) {
355 ostdout = dup(STDOUT_FILENO);
356 dup2(pfd[1], STDOUT_FILENO);
365 e = xmalloc(sizeof(struct kevent));
366 EV_SET(e, pr_pd, EVFILT_PROCDESC, EV_ADD, NOTE_EXIT, 0,
368 if (kevent(kq, e, 1, NULL, 0, NULL) == -1)
374 cap_rights_init(&rights_ro, CAP_READ, CAP_FSTAT, CAP_SEEK);
375 if (cap_rights_limit(fileno(f1), &rights_ro) < 0
377 err(2, "unable to limit rights on: %s", file1);
378 if (cap_rights_limit(fileno(f2), &rights_ro) < 0 &&
380 err(2, "unable to limit rights on: %s", file2);
381 if (fileno(f1) == STDIN_FILENO || fileno(f2) == STDIN_FILENO) {
382 /* stding has already been limited */
383 if (caph_limit_stderr() == -1)
384 err(2, "unable to limit stderr");
385 if (caph_limit_stdout() == -1)
386 err(2, "unable to limit stdout");
387 } else if (caph_limit_stdio() == -1)
388 err(2, "unable to limit stdio");
390 caph_cache_catpages();
392 if (cap_enter() < 0 && errno != ENOSYS)
393 err(2, "unable to enter capability mode");
396 switch (files_differ(f1, f2, flags)) {
407 if ((flags & D_FORCEASCII) == 0 &&
408 (!asciifile(f1) || !asciifile(f2))) {
413 prepare(0, f1, stb1.st_size, flags);
414 prepare(1, f2, stb2.st_size, flags);
417 sort(sfile[0], slen[0]);
418 sort(sfile[1], slen[1]);
420 member = (int *)file[1];
421 equiv(sfile[0], slen[0], sfile[1], slen[1], member);
422 member = xreallocarray(member, slen[1] + 2, sizeof(*member));
424 class = (int *)file[0];
425 unsort(sfile[0], slen[0], class);
426 class = xreallocarray(class, slen[0] + 2, sizeof(*class));
428 klist = xcalloc(slen[0] + 2, sizeof(*klist));
431 clist = xcalloc(clistlen, sizeof(*clist));
432 i = stone(class, slen[0], member, klist, flags);
436 J = xreallocarray(J, len[0] + 2, sizeof(*J));
441 ixold = xreallocarray(ixold, len[0] + 2, sizeof(*ixold));
442 ixnew = xreallocarray(ixnew, len[1] + 2, sizeof(*ixnew));
443 check(f1, f2, flags);
444 output(file1, f1, file2, f2, flags);
445 if (ostdout != -1 && e != NULL) {
446 /* close the pipe to pr and restore stdout */
450 if (ostdout != STDOUT_FILENO) {
451 close(STDOUT_FILENO);
452 dup2(ostdout, STDOUT_FILENO);
455 if (kevent(kq, NULL, 0, e, 1, NULL) == -1)
459 if (WIFEXITED(wstatus) && WEXITSTATUS(wstatus) != 0)
460 errx(2, "pr exited abnormally");
461 else if (WIFSIGNALED(wstatus))
462 errx(2, "pr killed by signal %d",
481 * Check to see if the given files differ.
482 * Returns 0 if they are the same, 1 if different, and -1 on error.
483 * XXX - could use code from cmp(1) [faster]
486 files_differ(FILE *f1, FILE *f2, int flags)
488 char buf1[BUFSIZ], buf2[BUFSIZ];
491 if ((flags & (D_EMPTY1|D_EMPTY2)) || stb1.st_size != stb2.st_size ||
492 (stb1.st_mode & S_IFMT) != (stb2.st_mode & S_IFMT))
495 i = fread(buf1, 1, sizeof(buf1), f1);
496 j = fread(buf2, 1, sizeof(buf2), f2);
497 if ((!i && ferror(f1)) || (!j && ferror(f2)))
503 if (memcmp(buf1, buf2, i) != 0)
509 opentemp(const char *f)
511 char buf[BUFSIZ], tempfile[PATH_MAX];
515 if (strcmp(f, "-") == 0)
517 else if ((ifd = open(f, O_RDONLY, 0644)) < 0)
520 (void)strlcpy(tempfile, _PATH_TMP "/diff.XXXXXXXX", sizeof(tempfile));
522 if ((ofd = mkstemp(tempfile)) < 0) {
527 while ((nread = read(ifd, buf, BUFSIZ)) > 0) {
528 if (write(ofd, buf, nread) != nread) {
535 lseek(ofd, (off_t)0, SEEK_SET);
536 return (fdopen(ofd, "r"));
540 splice(char *dir, char *path)
545 dirlen = strlen(dir);
546 while (dirlen != 0 && dir[dirlen - 1] == '/')
548 if ((tail = strrchr(path, '/')) == NULL)
552 xasprintf(&buf, "%.*s/%s", (int)dirlen, dir, tail);
557 prepare(int i, FILE *fd, size_t filesize, int flags)
565 sz = MIN(filesize, SIZE_MAX) / 25;
569 p = xcalloc(sz + 3, sizeof(*p));
570 for (j = 0; (h = readhash(fd, flags));) {
573 p = xreallocarray(p, sz + 3, sizeof(*p));
586 for (pref = 0; pref < len[0] && pref < len[1] &&
587 file[0][pref + 1].value == file[1][pref + 1].value;
590 for (suff = 0; suff < len[0] - pref && suff < len[1] - pref &&
591 file[0][len[0] - suff].value == file[1][len[1] - suff].value;
594 for (j = 0; j < 2; j++) {
595 sfile[j] = file[j] + pref;
596 slen[j] = len[j] - pref - suff;
597 for (i = 0; i <= slen[j]; i++)
598 sfile[j][i].serial = i;
603 equiv(struct line *a, int n, struct line *b, int m, int *c)
608 while (i <= n && j <= m) {
609 if (a[i].value < b[j].value)
611 else if (a[i].value == b[j].value)
622 while (b[j + 1].value == b[j].value) {
630 /* Code taken from ping.c */
639 do { /* newton was a stinker */
644 } while ((x - y) > 1 || (x - y) < -1);
650 stone(int *a, int n, int *b, int *c, int flags)
653 int oldc, tc, oldl, sq;
654 u_int numtries, bound;
656 if (flags & D_MINIMAL)
660 bound = MAX(256, sq);
664 c[0] = newcand(0, 0, 0);
665 for (i = 1; i <= n; i++) {
674 if (y <= clist[oldc].y)
680 if (clist[c[l]].y <= y)
683 c[l] = newcand(i, y, oldc);
688 c[l] = newcand(i, y, oldc);
692 } while ((y = b[++j]) > 0 && numtries < bound);
698 newcand(int x, int y, int pred)
702 if (clen == clistlen) {
703 clistlen = clistlen * 11 / 10;
704 clist = xreallocarray(clist, clistlen, sizeof(*clist));
714 search(int *c, int k, int y)
718 if (clist[c[k]].y < y) /* quick look for typical case */
743 for (i = 0; i <= len[0]; i++)
744 J[i] = i <= pref ? i :
745 i > len[0] - suff ? i + len[1] - len[0] : 0;
746 for (q = clist + p; q->y != 0; q = clist + q->pred)
747 J[q->x + pref] = q->y + pref;
751 * Check does double duty:
752 * 1. ferret out any fortuitous correspondences due
753 * to confounding by hashing (which result in "jackpot")
754 * 2. collect random access indexes to the two files
757 check(FILE *f1, FILE *f2, int flags)
759 int i, j, jackpot, c, d;
765 ixold[0] = ixnew[0] = 0;
768 for (i = 1; i <= len[0]; i++) {
770 ixold[i] = ctold += skipline(f1);
774 ixnew[j] = ctnew += skipline(f2);
777 if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS|D_IGNORECASE|D_STRIPCR)) {
782 * GNU diff ignores a missing newline
783 * in one file for -b or -w.
785 if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) {
786 if (c == EOF && d == '\n') {
789 } else if (c == '\n' && d == EOF) {
796 if (flags & D_STRIPCR && (c == '\r' || d == '\r')) {
798 if ((c = getc(f1)) == '\n') {
805 if ((d = getc(f2)) == '\n') {
813 if ((flags & D_FOLDBLANKS) && isspace(c) &&
819 } while (isspace(c = getc(f1)));
824 } while (isspace(d = getc(f2)));
825 } else if ((flags & D_IGNOREBLANKS)) {
826 while (isspace(c) && c != '\n') {
830 while (isspace(d) && d != '\n') {
835 if (chrtran(c) != chrtran(d)) {
838 if (c != '\n' && c != EOF)
839 ctold += skipline(f1);
840 if (d != '\n' && c != EOF)
841 ctnew += skipline(f2);
844 if (c == '\n' || c == EOF)
851 if ((c = getc(f1)) != (d = getc(f2))) {
854 if (c != '\n' && c != EOF)
855 ctold += skipline(f1);
856 if (d != '\n' && c != EOF)
857 ctnew += skipline(f2);
860 if (c == '\n' || c == EOF)
868 for (; j <= len[1]; j++) {
869 ixnew[j] = ctnew += skipline(f2);
873 * fprintf(stderr, "jackpot\n");
877 /* shellsort CACM #201 */
879 sort(struct line *a, int n)
881 struct line *ai, *aim, w;
886 for (j = 1; j <= n; j *= 2)
888 for (m /= 2; m != 0; m /= 2) {
890 for (j = 1; j <= k; j++) {
891 for (ai = &a[j]; ai > a; ai -= m) {
894 break; /* wraparound */
895 if (aim->value > ai[0].value ||
896 (aim->value == ai[0].value &&
897 aim->serial > ai[0].serial))
899 w.value = ai[0].value;
900 ai[0].value = aim->value;
901 aim->value = w.value;
902 w.serial = ai[0].serial;
903 ai[0].serial = aim->serial;
904 aim->serial = w.serial;
911 unsort(struct line *f, int l, int *b)
915 a = xcalloc(l + 1, sizeof(*a));
916 for (i = 1; i <= l; i++)
917 a[f[i].serial] = f[i].value;
918 for (i = 1; i <= l; i++)
928 for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++)
934 output(char *file1, FILE *f1, char *file2, FILE *f2, int flags)
936 int m, i0, i1, j0, j1;
942 J[m + 1] = len[1] + 1;
943 if (diff_format != D_EDIT) {
944 for (i0 = 1; i0 <= m; i0 = i1 + 1) {
945 while (i0 <= m && J[i0] == J[i0 - 1] + 1)
949 while (i1 < m && J[i1 + 1] == 0)
953 change(file1, f1, file2, f2, i0, i1, j0, j1, &flags);
956 for (i0 = m; i0 >= 1; i0 = i1 - 1) {
957 while (i0 >= 1 && J[i0] == J[i0 + 1] - 1 && J[i0] != 0)
961 while (i1 > 1 && J[i1 - 1] == 0)
965 change(file1, f1, file2, f2, i1, i0, j1, j0, &flags);
969 change(file1, f1, file2, f2, 1, 0, 1, len[1], &flags);
970 if (diff_format == D_IFDEF || diff_format == D_GFORMAT) {
973 if ((c = getc(f1)) == EOF)
975 diff_output("%c", c);
979 if (anychange != 0) {
980 if (diff_format == D_CONTEXT)
981 dump_context_vec(f1, f2, flags);
982 else if (diff_format == D_UNIFIED)
983 dump_unified_vec(f1, f2, flags);
988 range(int a, int b, const char *separator)
990 diff_output("%d", a > b ? b : a);
992 diff_output("%s%d", separator, b);
996 uni_range(int a, int b)
999 diff_output("%d,%d", a, b - a + 1);
1001 diff_output("%d", b);
1003 diff_output("%d,0", b);
1007 preadline(int fd, size_t rlen, off_t off)
1012 line = xmalloc(rlen + 1);
1013 if ((nr = pread(fd, line, rlen, off)) < 0)
1014 err(2, "preadline");
1015 if (nr > 0 && line[nr-1] == '\n')
1022 ignoreline(char *line)
1026 ret = regexec(&ignore_re, line, 0, NULL, 0);
1028 return (ret == 0); /* if it matched, it should be ignored. */
1032 * Indicate that there is a difference between lines a and b of the from file
1033 * to get to lines c to d of the to file. If a is greater then b then there
1034 * are no lines in the from file involved and this means that there were
1035 * lines appended (beginning at b). If c is greater than d then there are
1036 * lines missing from the to file.
1039 change(char *file1, FILE *f1, char *file2, FILE *f2, int a, int b, int c, int d,
1042 static size_t max_context = 64;
1048 if ((diff_format != D_IFDEF || diff_format == D_GFORMAT) &&
1051 if (ignore_pats != NULL) {
1054 * All lines in the change, insert, or delete must
1055 * match an ignore pattern for the change to be
1058 if (a <= b) { /* Changes and deletes. */
1059 for (i = a; i <= b; i++) {
1060 line = preadline(fileno(f1),
1061 ixold[i] - ixold[i - 1], ixold[i - 1]);
1062 if (!ignoreline(line))
1066 if (a > b || c <= d) { /* Changes and inserts. */
1067 for (i = c; i <= d; i++) {
1068 line = preadline(fileno(f2),
1069 ixnew[i] - ixnew[i - 1], ixnew[i - 1]);
1070 if (!ignoreline(line))
1076 if (*pflags & D_SKIPBLANKLINES) {
1079 * All lines in the change, insert, or delete must not be
1080 * empty for the change to be ignored.
1082 if (a <= b) { /* Changes and deletes. */
1083 for (i = a; i <= b; i++) {
1084 line = preadline(fileno(f1),
1085 ixold[i] - ixold[i - 1], ixold[i - 1]);
1090 if (a > b || c <= d) { /* Changes and inserts. */
1091 for (i = c; i <= d; i++) {
1092 line = preadline(fileno(f2),
1093 ixnew[i] - ixnew[i - 1], ixnew[i - 1]);
1102 if (*pflags & D_HEADER && diff_format != D_BRIEF) {
1103 diff_output("%s %s %s\n", diffargs, file1, file2);
1104 *pflags &= ~D_HEADER;
1106 if (diff_format == D_CONTEXT || diff_format == D_UNIFIED) {
1108 * Allocate change records as needed.
1110 if (context_vec_ptr == context_vec_end - 1) {
1111 ptrdiff_t offset = context_vec_ptr - context_vec_start;
1113 context_vec_start = xreallocarray(context_vec_start,
1114 max_context, sizeof(*context_vec_start));
1115 context_vec_end = context_vec_start + max_context;
1116 context_vec_ptr = context_vec_start + offset;
1118 if (anychange == 0) {
1120 * Print the context/unidiff header first time through.
1122 print_header(file1, file2);
1124 } else if (a > context_vec_ptr->b + (2 * diff_context) + 1 &&
1125 c > context_vec_ptr->d + (2 * diff_context) + 1) {
1127 * If this change is more than 'diff_context' lines from the
1128 * previous change, dump the record and reset it.
1130 if (diff_format == D_CONTEXT)
1131 dump_context_vec(f1, f2, *pflags);
1133 dump_unified_vec(f1, f2, *pflags);
1136 context_vec_ptr->a = a;
1137 context_vec_ptr->b = b;
1138 context_vec_ptr->c = c;
1139 context_vec_ptr->d = d;
1144 switch (diff_format) {
1150 diff_output("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1151 if (diff_format == D_NORMAL)
1156 diff_output("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1162 diff_output("a%d %d\n", b, d - c + 1);
1164 diff_output("d%d %d\n", a, b - a + 1);
1166 /* add changed lines */
1167 diff_output("a%d %d\n", b, d - c + 1);
1171 if (diff_format == D_GFORMAT) {
1173 /* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1174 nc = ixold[a > b ? b : a - 1] - curpos;
1175 for (i = 0; i < nc; i++)
1176 diff_output("%c", getc(f1));
1177 for (walk = group_format; *walk != '\0'; walk++) {
1182 fetch(ixold, a, b, f1, '<', 1, *pflags);
1185 fetch(ixnew, c, d, f2, '>', 0, *pflags);
1188 diff_output("%%%c", *walk);
1193 diff_output("%c", *walk);
1196 if (diff_format == D_NORMAL || diff_format == D_IFDEF) {
1197 fetch(ixold, a, b, f1, '<', 1, *pflags);
1198 if (a <= b && c <= d && diff_format == D_NORMAL)
1199 diff_output("---\n");
1202 if (diff_format != D_GFORMAT)
1203 f = fetch(ixnew, c, d, f2, diff_format == D_NORMAL ? '>' : '\0', 0, *pflags);
1204 if (f != 0 && diff_format == D_EDIT) {
1206 * A non-zero return 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 diff_output("%ds/.//\n", a + f - 1);
1219 if ((diff_format == D_EDIT || diff_format == D_REVERSE) && c <= d)
1222 diff_output("#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;
1234 * When doing #ifdef's, copy down to current line
1235 * if this is the first file, so that stuff makes it to output.
1237 if ((diff_format == D_IFDEF) && oldfile) {
1238 long curpos = ftell(lb);
1239 /* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1240 nc = f[a > b ? b : a - 1] - curpos;
1241 for (i = 0; i < nc; i++)
1242 diff_output("%c", getc(lb));
1246 if (diff_format == D_IFDEF) {
1248 diff_output("#else /* %s%s */\n",
1249 oldfile == 1 ? "!" : "", ifdefname);
1252 diff_output("#ifndef %s\n", ifdefname);
1254 diff_output("#ifdef %s\n", ifdefname);
1256 inifdef = 1 + oldfile;
1258 for (i = a; i <= b; i++) {
1259 fseek(lb, f[i - 1], SEEK_SET);
1260 nc = f[i] - f[i - 1];
1261 if ((diff_format != D_IFDEF && diff_format != D_GFORMAT) &&
1263 diff_output("%c", ch);
1264 if (Tflag && (diff_format == D_NORMAL || diff_format == D_CONTEXT
1265 || diff_format == D_UNIFIED))
1267 else if (diff_format != D_UNIFIED)
1271 for (j = 0, lastc = '\0'; j < nc; j++, lastc = c) {
1272 if ((c = getc(lb)) == EOF) {
1273 if (diff_format == D_EDIT || diff_format == D_REVERSE ||
1274 diff_format == D_NREVERSE)
1275 warnx("No newline at end of file");
1277 diff_output("\n\\ No newline at end of "
1281 if (c == '\t' && (flags & D_EXPANDTABS)) {
1282 newcol = ((col/tabsize)+1)*tabsize;
1285 } while (++col < newcol);
1287 if (diff_format == D_EDIT && j == 1 && c == '\n'
1290 * Don't print a bare "." line
1291 * since that will confuse ed(1).
1292 * Print ".." instead and return,
1293 * giving the caller an offset
1294 * from which to restart.
1299 diff_output("%c", c);
1308 * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578.
1311 readhash(FILE *f, int flags)
1318 if ((flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) == 0) {
1319 if (flags & D_IGNORECASE)
1320 for (i = 0; (t = getc(f)) != '\n'; i++) {
1321 if (flags & D_STRIPCR && t == '\r') {
1332 sum = sum * 127 + chrtran(t);
1335 for (i = 0; (t = getc(f)) != '\n'; i++) {
1336 if (flags & D_STRIPCR && t == '\r') {
1347 sum = sum * 127 + t;
1351 switch (t = getc(f)) {
1360 if (space && (flags & D_IGNOREBLANKS) == 0) {
1364 sum = sum * 127 + chrtran(t);
1378 * There is a remote possibility that we end up with a zero sum.
1379 * Zero is used as an EOF marker, so return 1 instead.
1381 return (sum == 0 ? 1 : sum);
1387 unsigned char buf[BUFSIZ];
1394 cnt = fread(buf, 1, sizeof(buf), f);
1395 return (memchr(buf, '\0', cnt) == NULL);
1398 #define begins_with(s, pre) (strncmp(s, pre, sizeof(pre)-1) == 0)
1401 match_function(const long *f, int pos, FILE *fp)
1403 unsigned char buf[FUNCTION_CONTEXT_SIZE];
1405 int last = lastline;
1406 const char *state = NULL;
1409 while (pos > last) {
1410 fseek(fp, f[pos - 1], SEEK_SET);
1411 nc = f[pos] - f[pos - 1];
1412 if (nc >= sizeof(buf))
1413 nc = sizeof(buf) - 1;
1414 nc = fread(buf, 1, nc, fp);
1417 buf[strcspn(buf, "\n")] = '\0';
1418 if (isalpha(buf[0]) || buf[0] == '_' || buf[0] == '$') {
1419 if (begins_with(buf, "private:")) {
1421 state = " (private)";
1422 } else if (begins_with(buf, "protected:")) {
1424 state = " (protected)";
1425 } else if (begins_with(buf, "public:")) {
1427 state = " (public)";
1429 strlcpy(lastbuf, buf, sizeof lastbuf);
1431 strlcat(lastbuf, state,
1433 lastmatchline = pos;
1440 return lastmatchline > 0 ? lastbuf : NULL;
1443 /* dump accumulated "context" diff changes */
1445 dump_context_vec(FILE *f1, FILE *f2, int flags)
1447 struct context_vec *cvp = context_vec_start;
1448 int lowa, upb, lowc, upd, do_output;
1452 if (context_vec_start > context_vec_ptr)
1455 b = d = 0; /* gcc */
1456 lowa = MAX(1, cvp->a - diff_context);
1457 upb = MIN(len[0], context_vec_ptr->b + diff_context);
1458 lowc = MAX(1, cvp->c - diff_context);
1459 upd = MIN(len[1], context_vec_ptr->d + diff_context);
1461 diff_output("***************");
1462 if ((flags & D_PROTOTYPE)) {
1463 f = match_function(ixold, lowa-1, f1);
1465 diff_output(" %s", f);
1467 diff_output("\n*** ");
1468 range(lowa, upb, ",");
1469 diff_output(" ****\n");
1472 * Output changes to the "old" file. The first loop suppresses
1473 * output if there were no changes to the "old" file (we'll see
1474 * the "old" lines as context in the "new" list).
1477 for (; cvp <= context_vec_ptr; cvp++)
1478 if (cvp->a <= cvp->b) {
1479 cvp = context_vec_start;
1484 while (cvp <= context_vec_ptr) {
1490 if (a <= b && c <= d)
1493 ch = (a <= b) ? 'd' : 'a';
1496 fetch(ixold, lowa, b, f1, ' ', 0, flags);
1498 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1499 fetch(ixold, a, b, f1,
1500 ch == 'c' ? '!' : '-', 0, flags);
1505 fetch(ixold, b + 1, upb, f1, ' ', 0, flags);
1507 /* output changes to the "new" file */
1508 diff_output("--- ");
1509 range(lowc, upd, ",");
1510 diff_output(" ----\n");
1513 for (cvp = context_vec_start; cvp <= context_vec_ptr; cvp++)
1514 if (cvp->c <= cvp->d) {
1515 cvp = context_vec_start;
1520 while (cvp <= context_vec_ptr) {
1526 if (a <= b && c <= d)
1529 ch = (a <= b) ? 'd' : 'a';
1532 fetch(ixnew, lowc, d, f2, ' ', 0, flags);
1534 fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1535 fetch(ixnew, c, d, f2,
1536 ch == 'c' ? '!' : '+', 0, flags);
1541 fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1543 context_vec_ptr = context_vec_start - 1;
1546 /* dump accumulated "unified" diff changes */
1548 dump_unified_vec(FILE *f1, FILE *f2, int flags)
1550 struct context_vec *cvp = context_vec_start;
1551 int lowa, upb, lowc, upd;
1555 if (context_vec_start > context_vec_ptr)
1558 b = d = 0; /* gcc */
1559 lowa = MAX(1, cvp->a - diff_context);
1560 upb = MIN(len[0], context_vec_ptr->b + diff_context);
1561 lowc = MAX(1, cvp->c - diff_context);
1562 upd = MIN(len[1], context_vec_ptr->d + diff_context);
1564 diff_output("@@ -");
1565 uni_range(lowa, upb);
1567 uni_range(lowc, upd);
1569 if ((flags & D_PROTOTYPE)) {
1570 f = match_function(ixold, lowa-1, f1);
1572 diff_output(" %s", f);
1577 * Output changes in "unified" diff format--the old and new lines
1578 * are printed together.
1580 for (; cvp <= context_vec_ptr; cvp++) {
1587 * c: both new and old changes
1588 * d: only changes in the old file
1589 * a: only changes in the new file
1591 if (a <= b && c <= d)
1594 ch = (a <= b) ? 'd' : 'a';
1598 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1599 fetch(ixold, a, b, f1, '-', 0, flags);
1600 fetch(ixnew, c, d, f2, '+', 0, flags);
1603 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1604 fetch(ixold, a, b, f1, '-', 0, flags);
1607 fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1608 fetch(ixnew, c, d, f2, '+', 0, flags);
1614 fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1616 context_vec_ptr = context_vec_start - 1;
1620 print_header(const char *file1, const char *file2)
1622 const char *time_format;
1627 struct tm tm1, tm2, *tm_ptr1, *tm_ptr2;
1628 int nsec1 = stb1.st_mtim.tv_nsec;
1629 int nsec2 = stb2.st_mtim.tv_nsec;
1631 time_format = "%Y-%m-%d %H:%M:%S";
1635 tm_ptr1 = localtime_r(&stb1.st_mtime, &tm1);
1636 tm_ptr2 = localtime_r(&stb2.st_mtime, &tm2);
1637 strftime(buf1, 256, time_format, tm_ptr1);
1638 strftime(buf2, 256, time_format, tm_ptr2);
1640 strftime(end1, 10, "%z", tm_ptr1);
1641 strftime(end2, 10, "%z", tm_ptr2);
1642 sprintf(buf1, "%s.%.9d %s", buf1, nsec1, end1);
1643 sprintf(buf2, "%s.%.9d %s", buf2, nsec2, end2);
1645 if (label[0] != NULL)
1646 diff_output("%s %s\n", diff_format == D_CONTEXT ? "***" : "---",
1649 diff_output("%s %s\t%s\n", diff_format == D_CONTEXT ? "***" : "---",
1651 if (label[1] != NULL)
1652 diff_output("%s %s\n", diff_format == D_CONTEXT ? "---" : "+++",
1655 diff_output("%s %s\t%s\n", diff_format == D_CONTEXT ? "---" : "+++",