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>
96 #define _PATH_PR "/usr/bin/pr"
99 * diff - compare two files.
103 * Uses an algorithm due to Harold Stone, which finds
104 * a pair of longest identical subsequences in the two
107 * The major goal is to generate the match vector J.
108 * J[i] is the index of the line in file1 corresponding
109 * to line i file0. J[i] = 0 if there is no
110 * such line in file1.
112 * Lines are hashed so as to work in core. All potential
113 * matches are located by sorting the lines of each file
114 * on the hash (called ``value''). In particular, this
115 * collects the equivalence classes in file1 together.
116 * Subroutine equiv replaces the value of each line in
117 * file0 by the index of the first element of its
118 * matching equivalence in (the reordered) file1.
119 * To save space equiv squeezes file1 into a single
120 * array member in which the equivalence classes
121 * are simply concatenated, except that their first
122 * members are flagged by changing sign.
124 * Next the indices that point into member are unsorted into
125 * array class according to the original order of file0.
127 * The cleverness lies in routine stone. This marches
128 * through the lines of file0, developing a vector klist
129 * of "k-candidates". At step i a k-candidate is a matched
130 * pair of lines x,y (x in file0 y in file1) such that
131 * there is a common subsequence of length k
132 * between the first i lines of file0 and the first y
133 * lines of file1, but there is no such subsequence for
134 * any smaller y. x is the earliest possible mate to y
135 * that occurs in such a subsequence.
137 * Whenever any of the members of the equivalence class of
138 * lines in file1 matable to a line in file0 has serial number
139 * less than the y of some k-candidate, that k-candidate
140 * with the smallest such y is replaced. The new
141 * k-candidate is chained (via pred) to the current
142 * k-1 candidate so that the actual subsequence can
143 * be recovered. When a member has serial number greater
144 * that the y of all k-candidates, the klist is extended.
145 * At the end, the longest subsequence is pulled out
146 * and placed in the array J by unravel
148 * With J in hand, the matches there recorded are
149 * check'ed against reality to assure that no spurious
150 * matches have crept in due to hashing. If they have,
151 * they are broken, and "jackpot" is recorded--a harmless
152 * matter except that a true match for a spuriously
153 * mated line may now be unnecessarily reported as a change.
155 * Much of the complexity of the program comes simply
156 * from trying to minimize core utilization and
157 * maximize the range of doable problems by dynamically
158 * allocating what is needed and reusing what is not.
159 * The core requirements for problems larger than somewhat
160 * are (in words) 2*length(file0) + length(file1) +
161 * 3*(number of k-candidates installed), typically about
162 * 6n words for files of length n.
177 * The following struct is used to record change information when
178 * doing a "context" or "unified" diff. (see routine "change" to
179 * understand the highly mnemonic field names)
182 int a; /* start line in old file */
183 int b; /* end line in old file */
184 int c; /* start line in new file */
185 int d; /* end line in new file */
188 #define diff_output printf
189 static FILE *opentemp(const char *);
190 static void output(char *, FILE *, char *, FILE *, int);
191 static void check(FILE *, FILE *, int);
192 static void range(int, int, const char *);
193 static void uni_range(int, int);
194 static void dump_context_vec(FILE *, FILE *, int);
195 static void dump_unified_vec(FILE *, FILE *, int);
196 static void prepare(int, FILE *, size_t, int);
197 static void prune(void);
198 static void equiv(struct line *, int, struct line *, int, int *);
199 static void unravel(int);
200 static void unsort(struct line *, int, int *);
201 static void change(char *, FILE *, char *, FILE *, int, int, int, int, int *);
202 static void sort(struct line *, int);
203 static void print_header(const char *, const char *);
204 static int ignoreline(char *);
205 static int asciifile(FILE *);
206 static int fetch(long *, int, int, FILE *, int, int, int);
207 static int newcand(int, int, int);
208 static int search(int *, int, int);
209 static int skipline(FILE *);
210 static int isqrt(int);
211 static int stone(int *, int, int *, int *, int);
212 static int readhash(FILE *, int);
213 static int files_differ(FILE *, FILE *, int);
214 static char *match_function(const long *, int, FILE *);
215 static char *preadline(int, size_t, off_t);
217 static int *J; /* will be overlaid on class */
218 static int *class; /* will be overlaid on file[0] */
219 static int *klist; /* will be overlaid on file[0] after class */
220 static int *member; /* will be overlaid on file[1] */
222 static int inifdef; /* whether or not we are in a #ifdef block */
224 static int pref, suff; /* length of prefix and suffix */
226 static int anychange;
227 static long *ixnew; /* will be overlaid on file[1] */
228 static long *ixold; /* will be overlaid on klist */
229 static struct cand *clist; /* merely a free storage pot for candidates */
230 static int clistlen; /* the length of clist */
231 static struct line *sfile[2]; /* shortened by pruning common prefix/suffix */
232 static int (*chrtran)(int); /* translation table for case-folding */
233 static struct context_vec *context_vec_start;
234 static struct context_vec *context_vec_end;
235 static struct context_vec *context_vec_ptr;
237 #define FUNCTION_CONTEXT_SIZE 55
238 static char lastbuf[FUNCTION_CONTEXT_SIZE];
240 static int lastmatchline;
257 diffreg(char *file1, char *file2, int flags, int capsicum)
264 cap_rights_t rights_ro;
273 context_vec_ptr = context_vec_start - 1;
274 if (flags & D_IGNORECASE)
278 if (S_ISDIR(stb1.st_mode) != S_ISDIR(stb2.st_mode))
279 return (S_ISDIR(stb1.st_mode) ? D_MISMATCH1 : D_MISMATCH2);
280 if (strcmp(file1, "-") == 0 && strcmp(file2, "-") == 0)
283 if (flags & D_EMPTY1)
284 f1 = fopen(_PATH_DEVNULL, "r");
286 if (!S_ISREG(stb1.st_mode)) {
287 if ((f1 = opentemp(file1)) == NULL ||
288 fstat(fileno(f1), &stb1) < 0) {
293 } else if (strcmp(file1, "-") == 0)
296 f1 = fopen(file1, "r");
304 if (flags & D_EMPTY2)
305 f2 = fopen(_PATH_DEVNULL, "r");
307 if (!S_ISREG(stb2.st_mode)) {
308 if ((f2 = opentemp(file2)) == NULL ||
309 fstat(fileno(f2), &stb2) < 0) {
314 } else if (strcmp(file2, "-") == 0)
317 f2 = fopen(file2, "r");
326 /* redirect stdout to pr */
331 xasprintf(&header, "%s %s %s", diffargs, file1, file2);
332 signal(SIGPIPE, SIG_IGN);
336 switch ((pid = pdfork(&pr_pd, PD_CLOEXEC))) {
340 err(2, "No more processes");
343 if (pfd[0] != STDIN_FILENO) {
344 dup2(pfd[0], STDIN_FILENO);
348 execl(_PATH_PR, _PATH_PR, "-h", header, (char *)0);
353 if (pfd[1] != STDOUT_FILENO) {
354 ostdout = dup(STDOUT_FILENO);
355 dup2(pfd[1], STDOUT_FILENO);
364 e = xmalloc(sizeof(struct kevent));
365 EV_SET(e, pr_pd, EVFILT_PROCDESC, EV_ADD, NOTE_EXIT, 0,
367 if (kevent(kq, e, 1, NULL, 0, NULL) == -1)
373 cap_rights_init(&rights_ro, CAP_READ, CAP_FSTAT, CAP_SEEK);
374 if (cap_rights_limit(fileno(f1), &rights_ro) < 0
376 err(2, "unable to limit rights on: %s", file1);
377 if (cap_rights_limit(fileno(f2), &rights_ro) < 0 &&
379 err(2, "unable to limit rights on: %s", file2);
380 if (fileno(f1) == STDIN_FILENO || fileno(f2) == STDIN_FILENO) {
381 /* stding has already been limited */
382 if (caph_limit_stderr() == -1)
383 err(2, "unable to limit stderr");
384 if (caph_limit_stdout() == -1)
385 err(2, "unable to limit stdout");
386 } else if (caph_limit_stdio() == -1)
387 err(2, "unable to limit stdio");
389 caph_cache_catpages();
391 if (cap_enter() < 0 && errno != ENOSYS)
392 err(2, "unable to enter capability mode");
395 switch (files_differ(f1, f2, flags)) {
406 if ((flags & D_FORCEASCII) == 0 &&
407 (!asciifile(f1) || !asciifile(f2))) {
412 prepare(0, f1, stb1.st_size, flags);
413 prepare(1, f2, stb2.st_size, flags);
416 sort(sfile[0], slen[0]);
417 sort(sfile[1], slen[1]);
419 member = (int *)file[1];
420 equiv(sfile[0], slen[0], sfile[1], slen[1], member);
421 member = xreallocarray(member, slen[1] + 2, sizeof(*member));
423 class = (int *)file[0];
424 unsort(sfile[0], slen[0], class);
425 class = xreallocarray(class, slen[0] + 2, sizeof(*class));
427 klist = xcalloc(slen[0] + 2, sizeof(*klist));
430 clist = xcalloc(clistlen, sizeof(*clist));
431 i = stone(class, slen[0], member, klist, flags);
435 J = xreallocarray(J, len[0] + 2, sizeof(*J));
440 ixold = xreallocarray(ixold, len[0] + 2, sizeof(*ixold));
441 ixnew = xreallocarray(ixnew, len[1] + 2, sizeof(*ixnew));
442 check(f1, f2, flags);
443 output(file1, f1, file2, f2, flags);
444 if (ostdout != -1 && e != NULL) {
445 /* close the pipe to pr and restore stdout */
449 if (ostdout != STDOUT_FILENO) {
450 close(STDOUT_FILENO);
451 dup2(ostdout, STDOUT_FILENO);
454 if (kevent(kq, NULL, 0, e, 1, NULL) == -1)
458 if (WIFEXITED(wstatus) && WEXITSTATUS(wstatus) != 0)
459 errx(2, "pr exited abnormally");
460 else if (WIFSIGNALED(wstatus))
461 errx(2, "pr killed by signal %d",
480 * Check to see if the given files differ.
481 * Returns 0 if they are the same, 1 if different, and -1 on error.
482 * XXX - could use code from cmp(1) [faster]
485 files_differ(FILE *f1, FILE *f2, int flags)
487 char buf1[BUFSIZ], buf2[BUFSIZ];
490 if ((flags & (D_EMPTY1|D_EMPTY2)) || stb1.st_size != stb2.st_size ||
491 (stb1.st_mode & S_IFMT) != (stb2.st_mode & S_IFMT))
494 i = fread(buf1, 1, sizeof(buf1), f1);
495 j = fread(buf2, 1, sizeof(buf2), f2);
496 if ((!i && ferror(f1)) || (!j && ferror(f2)))
502 if (memcmp(buf1, buf2, i) != 0)
508 opentemp(const char *f)
510 char buf[BUFSIZ], tempfile[PATH_MAX];
514 if (strcmp(f, "-") == 0)
516 else if ((ifd = open(f, O_RDONLY, 0644)) < 0)
519 (void)strlcpy(tempfile, _PATH_TMP "/diff.XXXXXXXX", sizeof(tempfile));
521 if ((ofd = mkstemp(tempfile)) < 0) {
526 while ((nread = read(ifd, buf, BUFSIZ)) > 0) {
527 if (write(ofd, buf, nread) != nread) {
534 lseek(ofd, (off_t)0, SEEK_SET);
535 return (fdopen(ofd, "r"));
539 splice(char *dir, char *path)
544 dirlen = strlen(dir);
545 while (dirlen != 0 && dir[dirlen - 1] == '/')
547 if ((tail = strrchr(path, '/')) == NULL)
551 xasprintf(&buf, "%.*s/%s", (int)dirlen, dir, tail);
556 prepare(int i, FILE *fd, size_t filesize, int flags)
564 sz = MIN(filesize, SIZE_MAX) / 25;
568 p = xcalloc(sz + 3, sizeof(*p));
569 for (j = 0; (h = readhash(fd, flags));) {
572 p = xreallocarray(p, sz + 3, sizeof(*p));
585 for (pref = 0; pref < len[0] && pref < len[1] &&
586 file[0][pref + 1].value == file[1][pref + 1].value;
589 for (suff = 0; suff < len[0] - pref && suff < len[1] - pref &&
590 file[0][len[0] - suff].value == file[1][len[1] - suff].value;
593 for (j = 0; j < 2; j++) {
594 sfile[j] = file[j] + pref;
595 slen[j] = len[j] - pref - suff;
596 for (i = 0; i <= slen[j]; i++)
597 sfile[j][i].serial = i;
602 equiv(struct line *a, int n, struct line *b, int m, int *c)
607 while (i <= n && j <= m) {
608 if (a[i].value < b[j].value)
610 else if (a[i].value == b[j].value)
621 while (b[j + 1].value == b[j].value) {
629 /* Code taken from ping.c */
638 do { /* newton was a stinker */
643 } while ((x - y) > 1 || (x - y) < -1);
649 stone(int *a, int n, int *b, int *c, int flags)
652 int oldc, tc, oldl, sq;
653 u_int numtries, bound;
655 if (flags & D_MINIMAL)
659 bound = MAX(256, sq);
663 c[0] = newcand(0, 0, 0);
664 for (i = 1; i <= n; i++) {
673 if (y <= clist[oldc].y)
679 if (clist[c[l]].y <= y)
682 c[l] = newcand(i, y, oldc);
687 c[l] = newcand(i, y, oldc);
691 } while ((y = b[++j]) > 0 && numtries < bound);
697 newcand(int x, int y, int pred)
701 if (clen == clistlen) {
702 clistlen = clistlen * 11 / 10;
703 clist = xreallocarray(clist, clistlen, sizeof(*clist));
713 search(int *c, int k, int y)
717 if (clist[c[k]].y < y) /* quick look for typical case */
742 for (i = 0; i <= len[0]; i++)
743 J[i] = i <= pref ? i :
744 i > len[0] - suff ? i + len[1] - len[0] : 0;
745 for (q = clist + p; q->y != 0; q = clist + q->pred)
746 J[q->x + pref] = q->y + pref;
750 * Check does double duty:
751 * 1. ferret out any fortuitous correspondences due
752 * to confounding by hashing (which result in "jackpot")
753 * 2. collect random access indexes to the two files
756 check(FILE *f1, FILE *f2, int flags)
758 int i, j, jackpot, c, d;
764 ixold[0] = ixnew[0] = 0;
767 for (i = 1; i <= len[0]; i++) {
769 ixold[i] = ctold += skipline(f1);
773 ixnew[j] = ctnew += skipline(f2);
776 if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS|D_IGNORECASE|D_STRIPCR)) {
781 * GNU diff ignores a missing newline
782 * in one file for -b or -w.
784 if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) {
785 if (c == EOF && d == '\n') {
788 } else if (c == '\n' && d == EOF) {
795 if (flags & D_STRIPCR) {
797 if ((c = getc(f1)) == '\n') {
803 if ((d = getc(f2)) == '\n') {
809 if ((flags & D_FOLDBLANKS) && isspace(c) &&
815 } while (isspace(c = getc(f1)));
820 } while (isspace(d = getc(f2)));
821 } else if ((flags & D_IGNOREBLANKS)) {
822 while (isspace(c) && c != '\n') {
826 while (isspace(d) && d != '\n') {
831 if (chrtran(c) != chrtran(d)) {
834 if (c != '\n' && c != EOF)
835 ctold += skipline(f1);
836 if (d != '\n' && c != EOF)
837 ctnew += skipline(f2);
840 if (c == '\n' || c == EOF)
847 if ((c = getc(f1)) != (d = getc(f2))) {
850 if (c != '\n' && c != EOF)
851 ctold += skipline(f1);
852 if (d != '\n' && c != EOF)
853 ctnew += skipline(f2);
856 if (c == '\n' || c == EOF)
864 for (; j <= len[1]; j++) {
865 ixnew[j] = ctnew += skipline(f2);
869 * fprintf(stderr, "jackpot\n");
873 /* shellsort CACM #201 */
875 sort(struct line *a, int n)
877 struct line *ai, *aim, w;
882 for (j = 1; j <= n; j *= 2)
884 for (m /= 2; m != 0; m /= 2) {
886 for (j = 1; j <= k; j++) {
887 for (ai = &a[j]; ai > a; ai -= m) {
890 break; /* wraparound */
891 if (aim->value > ai[0].value ||
892 (aim->value == ai[0].value &&
893 aim->serial > ai[0].serial))
895 w.value = ai[0].value;
896 ai[0].value = aim->value;
897 aim->value = w.value;
898 w.serial = ai[0].serial;
899 ai[0].serial = aim->serial;
900 aim->serial = w.serial;
907 unsort(struct line *f, int l, int *b)
911 a = xcalloc(l + 1, sizeof(*a));
912 for (i = 1; i <= l; i++)
913 a[f[i].serial] = f[i].value;
914 for (i = 1; i <= l; i++)
924 for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++)
930 output(char *file1, FILE *f1, char *file2, FILE *f2, int flags)
932 int m, i0, i1, j0, j1;
938 J[m + 1] = len[1] + 1;
939 if (diff_format != D_EDIT) {
940 for (i0 = 1; i0 <= m; i0 = i1 + 1) {
941 while (i0 <= m && J[i0] == J[i0 - 1] + 1)
945 while (i1 < m && J[i1 + 1] == 0)
949 change(file1, f1, file2, f2, i0, i1, j0, j1, &flags);
952 for (i0 = m; i0 >= 1; i0 = i1 - 1) {
953 while (i0 >= 1 && J[i0] == J[i0 + 1] - 1 && J[i0] != 0)
957 while (i1 > 1 && J[i1 - 1] == 0)
961 change(file1, f1, file2, f2, i1, i0, j1, j0, &flags);
965 change(file1, f1, file2, f2, 1, 0, 1, len[1], &flags);
966 if (diff_format == D_IFDEF || diff_format == D_GFORMAT) {
969 if ((c = getc(f1)) == EOF)
971 diff_output("%c", c);
975 if (anychange != 0) {
976 if (diff_format == D_CONTEXT)
977 dump_context_vec(f1, f2, flags);
978 else if (diff_format == D_UNIFIED)
979 dump_unified_vec(f1, f2, flags);
984 range(int a, int b, const char *separator)
986 diff_output("%d", a > b ? b : a);
988 diff_output("%s%d", separator, b);
992 uni_range(int a, int b)
995 diff_output("%d,%d", a, b - a + 1);
997 diff_output("%d", b);
999 diff_output("%d,0", b);
1003 preadline(int fd, size_t rlen, off_t off)
1008 line = xmalloc(rlen + 1);
1009 if ((nr = pread(fd, line, rlen, off)) < 0)
1010 err(2, "preadline");
1011 if (nr > 0 && line[nr-1] == '\n')
1018 ignoreline(char *line)
1022 ret = regexec(&ignore_re, line, 0, NULL, 0);
1024 return (ret == 0); /* if it matched, it should be ignored. */
1028 * Indicate that there is a difference between lines a and b of the from file
1029 * to get to lines c to d of the to file. If a is greater then b then there
1030 * are no lines in the from file involved and this means that there were
1031 * lines appended (beginning at b). If c is greater than d then there are
1032 * lines missing from the to file.
1035 change(char *file1, FILE *f1, char *file2, FILE *f2, int a, int b, int c, int d,
1038 static size_t max_context = 64;
1044 if ((diff_format != D_IFDEF || diff_format == D_GFORMAT) &&
1047 if (ignore_pats != NULL) {
1050 * All lines in the change, insert, or delete must
1051 * match an ignore pattern for the change to be
1054 if (a <= b) { /* Changes and deletes. */
1055 for (i = a; i <= b; i++) {
1056 line = preadline(fileno(f1),
1057 ixold[i] - ixold[i - 1], ixold[i - 1]);
1058 if (!ignoreline(line))
1062 if (a > b || c <= d) { /* Changes and inserts. */
1063 for (i = c; i <= d; i++) {
1064 line = preadline(fileno(f2),
1065 ixnew[i] - ixnew[i - 1], ixnew[i - 1]);
1066 if (!ignoreline(line))
1073 if (*pflags & D_HEADER && diff_format != D_BRIEF) {
1074 diff_output("%s %s %s\n", diffargs, file1, file2);
1075 *pflags &= ~D_HEADER;
1077 if (diff_format == D_CONTEXT || diff_format == D_UNIFIED) {
1079 * Allocate change records as needed.
1081 if (context_vec_ptr == context_vec_end - 1) {
1082 ptrdiff_t offset = context_vec_ptr - context_vec_start;
1084 context_vec_start = xreallocarray(context_vec_start,
1085 max_context, sizeof(*context_vec_start));
1086 context_vec_end = context_vec_start + max_context;
1087 context_vec_ptr = context_vec_start + offset;
1089 if (anychange == 0) {
1091 * Print the context/unidiff header first time through.
1093 print_header(file1, file2);
1095 } else if (a > context_vec_ptr->b + (2 * diff_context) + 1 &&
1096 c > context_vec_ptr->d + (2 * diff_context) + 1) {
1098 * If this change is more than 'diff_context' lines from the
1099 * previous change, dump the record and reset it.
1101 if (diff_format == D_CONTEXT)
1102 dump_context_vec(f1, f2, *pflags);
1104 dump_unified_vec(f1, f2, *pflags);
1107 context_vec_ptr->a = a;
1108 context_vec_ptr->b = b;
1109 context_vec_ptr->c = c;
1110 context_vec_ptr->d = d;
1115 switch (diff_format) {
1121 diff_output("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1122 if (diff_format == D_NORMAL)
1127 diff_output("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1133 diff_output("a%d %d\n", b, d - c + 1);
1135 diff_output("d%d %d\n", a, b - a + 1);
1137 /* add changed lines */
1138 diff_output("a%d %d\n", b, d - c + 1);
1142 if (diff_format == D_GFORMAT) {
1144 /* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1145 nc = ixold[a > b ? b : a - 1] - curpos;
1146 for (i = 0; i < nc; i++)
1147 diff_output("%c", getc(f1));
1148 for (walk = group_format; *walk != '\0'; walk++) {
1153 fetch(ixold, a, b, f1, '<', 1, *pflags);
1156 fetch(ixnew, c, d, f2, '>', 0, *pflags);
1159 diff_output("%%%c", *walk);
1164 diff_output("%c", *walk);
1167 if (diff_format == D_NORMAL || diff_format == D_IFDEF) {
1168 fetch(ixold, a, b, f1, '<', 1, *pflags);
1169 if (a <= b && c <= d && diff_format == D_NORMAL)
1170 diff_output("---\n");
1173 if (diff_format != D_GFORMAT)
1174 f = fetch(ixnew, c, d, f2, diff_format == D_NORMAL ? '>' : '\0', 0, *pflags);
1175 if (f != 0 && diff_format == D_EDIT) {
1177 * A non-zero return value for D_EDIT indicates that the
1178 * last line printed was a bare dot (".") that has been
1179 * escaped as ".." to prevent ed(1) from misinterpreting
1180 * it. We have to add a substitute command to change this
1181 * back and restart where we left off.
1184 diff_output("%ds/.//\n", a + f - 1);
1190 if ((diff_format == D_EDIT || diff_format == D_REVERSE) && c <= d)
1193 diff_output("#endif /* %s */\n", ifdefname);
1199 fetch(long *f, int a, int b, FILE *lb, int ch, int oldfile, int flags)
1201 int i, j, c, lastc, col, nc;
1205 * When doing #ifdef's, copy down to current line
1206 * if this is the first file, so that stuff makes it to output.
1208 if ((diff_format == D_IFDEF) && oldfile) {
1209 long curpos = ftell(lb);
1210 /* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1211 nc = f[a > b ? b : a - 1] - curpos;
1212 for (i = 0; i < nc; i++)
1213 diff_output("%c", getc(lb));
1217 if (diff_format == D_IFDEF) {
1219 diff_output("#else /* %s%s */\n",
1220 oldfile == 1 ? "!" : "", ifdefname);
1223 diff_output("#ifndef %s\n", ifdefname);
1225 diff_output("#ifdef %s\n", ifdefname);
1227 inifdef = 1 + oldfile;
1229 for (i = a; i <= b; i++) {
1230 fseek(lb, f[i - 1], SEEK_SET);
1231 nc = f[i] - f[i - 1];
1232 if ((diff_format != D_IFDEF && diff_format != D_GFORMAT) &&
1234 diff_output("%c", ch);
1235 if (Tflag && (diff_format == D_NORMAL || diff_format == D_CONTEXT
1236 || diff_format == D_UNIFIED))
1238 else if (diff_format != D_UNIFIED)
1242 for (j = 0, lastc = '\0'; j < nc; j++, lastc = c) {
1243 if ((c = getc(lb)) == EOF) {
1244 if (diff_format == D_EDIT || diff_format == D_REVERSE ||
1245 diff_format == D_NREVERSE)
1246 warnx("No newline at end of file");
1248 diff_output("\n\\ No newline at end of "
1252 if (c == '\t' && (flags & D_EXPANDTABS)) {
1253 newcol = ((col/tabsize)+1)*tabsize;
1256 } while (++col < newcol);
1258 if (diff_format == D_EDIT && j == 1 && c == '\n'
1261 * Don't print a bare "." line
1262 * since that will confuse ed(1).
1263 * Print ".." instead and return,
1264 * giving the caller an offset
1265 * from which to restart.
1270 diff_output("%c", c);
1279 * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578.
1282 readhash(FILE *f, int flags)
1289 if ((flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) == 0) {
1290 if (flags & D_IGNORECASE)
1291 for (i = 0; (t = getc(f)) != '\n'; i++) {
1292 if (flags & D_STRIPCR && t == '\r') {
1303 sum = sum * 127 + chrtran(t);
1306 for (i = 0; (t = getc(f)) != '\n'; i++) {
1307 if (flags & D_STRIPCR && t == '\r') {
1318 sum = sum * 127 + t;
1322 switch (t = getc(f)) {
1331 if (space && (flags & D_IGNOREBLANKS) == 0) {
1335 sum = sum * 127 + chrtran(t);
1349 * There is a remote possibility that we end up with a zero sum.
1350 * Zero is used as an EOF marker, so return 1 instead.
1352 return (sum == 0 ? 1 : sum);
1358 unsigned char buf[BUFSIZ];
1365 cnt = fread(buf, 1, sizeof(buf), f);
1366 return (memchr(buf, '\0', cnt) == NULL);
1369 #define begins_with(s, pre) (strncmp(s, pre, sizeof(pre)-1) == 0)
1372 match_function(const long *f, int pos, FILE *fp)
1374 unsigned char buf[FUNCTION_CONTEXT_SIZE];
1376 int last = lastline;
1377 const char *state = NULL;
1380 while (pos > last) {
1381 fseek(fp, f[pos - 1], SEEK_SET);
1382 nc = f[pos] - f[pos - 1];
1383 if (nc >= sizeof(buf))
1384 nc = sizeof(buf) - 1;
1385 nc = fread(buf, 1, nc, fp);
1388 buf[strcspn(buf, "\n")] = '\0';
1389 if (isalpha(buf[0]) || buf[0] == '_' || buf[0] == '$') {
1390 if (begins_with(buf, "private:")) {
1392 state = " (private)";
1393 } else if (begins_with(buf, "protected:")) {
1395 state = " (protected)";
1396 } else if (begins_with(buf, "public:")) {
1398 state = " (public)";
1400 strlcpy(lastbuf, buf, sizeof lastbuf);
1402 strlcat(lastbuf, state,
1404 lastmatchline = pos;
1411 return lastmatchline > 0 ? lastbuf : NULL;
1414 /* dump accumulated "context" diff changes */
1416 dump_context_vec(FILE *f1, FILE *f2, int flags)
1418 struct context_vec *cvp = context_vec_start;
1419 int lowa, upb, lowc, upd, do_output;
1423 if (context_vec_start > context_vec_ptr)
1426 b = d = 0; /* gcc */
1427 lowa = MAX(1, cvp->a - diff_context);
1428 upb = MIN(len[0], context_vec_ptr->b + diff_context);
1429 lowc = MAX(1, cvp->c - diff_context);
1430 upd = MIN(len[1], context_vec_ptr->d + diff_context);
1432 diff_output("***************");
1433 if ((flags & D_PROTOTYPE)) {
1434 f = match_function(ixold, lowa-1, f1);
1436 diff_output(" %s", f);
1438 diff_output("\n*** ");
1439 range(lowa, upb, ",");
1440 diff_output(" ****\n");
1443 * Output changes to the "old" file. The first loop suppresses
1444 * output if there were no changes to the "old" file (we'll see
1445 * the "old" lines as context in the "new" list).
1448 for (; cvp <= context_vec_ptr; cvp++)
1449 if (cvp->a <= cvp->b) {
1450 cvp = context_vec_start;
1455 while (cvp <= context_vec_ptr) {
1461 if (a <= b && c <= d)
1464 ch = (a <= b) ? 'd' : 'a';
1467 fetch(ixold, lowa, b, f1, ' ', 0, flags);
1469 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1470 fetch(ixold, a, b, f1,
1471 ch == 'c' ? '!' : '-', 0, flags);
1476 fetch(ixold, b + 1, upb, f1, ' ', 0, flags);
1478 /* output changes to the "new" file */
1479 diff_output("--- ");
1480 range(lowc, upd, ",");
1481 diff_output(" ----\n");
1484 for (cvp = context_vec_start; cvp <= context_vec_ptr; cvp++)
1485 if (cvp->c <= cvp->d) {
1486 cvp = context_vec_start;
1491 while (cvp <= context_vec_ptr) {
1497 if (a <= b && c <= d)
1500 ch = (a <= b) ? 'd' : 'a';
1503 fetch(ixnew, lowc, d, f2, ' ', 0, flags);
1505 fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1506 fetch(ixnew, c, d, f2,
1507 ch == 'c' ? '!' : '+', 0, flags);
1512 fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1514 context_vec_ptr = context_vec_start - 1;
1517 /* dump accumulated "unified" diff changes */
1519 dump_unified_vec(FILE *f1, FILE *f2, int flags)
1521 struct context_vec *cvp = context_vec_start;
1522 int lowa, upb, lowc, upd;
1526 if (context_vec_start > context_vec_ptr)
1529 b = d = 0; /* gcc */
1530 lowa = MAX(1, cvp->a - diff_context);
1531 upb = MIN(len[0], context_vec_ptr->b + diff_context);
1532 lowc = MAX(1, cvp->c - diff_context);
1533 upd = MIN(len[1], context_vec_ptr->d + diff_context);
1535 diff_output("@@ -");
1536 uni_range(lowa, upb);
1538 uni_range(lowc, upd);
1540 if ((flags & D_PROTOTYPE)) {
1541 f = match_function(ixold, lowa-1, f1);
1543 diff_output(" %s", f);
1548 * Output changes in "unified" diff format--the old and new lines
1549 * are printed together.
1551 for (; cvp <= context_vec_ptr; cvp++) {
1558 * c: both new and old changes
1559 * d: only changes in the old file
1560 * a: only changes in the new file
1562 if (a <= b && c <= d)
1565 ch = (a <= b) ? 'd' : 'a';
1569 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1570 fetch(ixold, a, b, f1, '-', 0, flags);
1571 fetch(ixnew, c, d, f2, '+', 0, flags);
1574 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1575 fetch(ixold, a, b, f1, '-', 0, flags);
1578 fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1579 fetch(ixnew, c, d, f2, '+', 0, flags);
1585 fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1587 context_vec_ptr = context_vec_start - 1;
1591 print_header(const char *file1, const char *file2)
1593 const char *time_format;
1598 struct tm tm1, tm2, *tm_ptr1, *tm_ptr2;
1599 int nsec1 = stb1.st_mtim.tv_nsec;
1600 int nsec2 = stb2.st_mtim.tv_nsec;
1602 time_format = "%Y-%m-%d %H:%M:%S";
1606 tm_ptr1 = localtime_r(&stb1.st_mtime, &tm1);
1607 tm_ptr2 = localtime_r(&stb2.st_mtime, &tm2);
1608 strftime(buf1, 256, time_format, tm_ptr1);
1609 strftime(buf2, 256, time_format, tm_ptr2);
1611 strftime(end1, 10, "%z", tm_ptr1);
1612 strftime(end2, 10, "%z", tm_ptr2);
1613 sprintf(buf1, "%s.%.9d %s", buf1, nsec1, end1);
1614 sprintf(buf2, "%s.%.9d %s", buf2, nsec2, end2);
1616 if (label[0] != NULL)
1617 diff_output("%s %s\n", diff_format == D_CONTEXT ? "***" : "---",
1620 diff_output("%s %s\t%s\n", diff_format == D_CONTEXT ? "***" : "---",
1622 if (label[1] != NULL)
1623 diff_output("%s %s\n", diff_format == D_CONTEXT ? "---" : "+++",
1626 diff_output("%s %s\t%s\n", diff_format == D_CONTEXT ? "---" : "+++",