1 /* vi: set sw=4 ts=4: */
3 * Mini diff implementation for busybox, adapted from OpenBSD diff.
5 * Copyright (C) 2006 by Robert Sullivan <cogito.ergo.cogito@hotmail.com>
7 * Licensed under GPLv2 or later, see file LICENSE in this tarball for details.
11 * Copyright (c) 2003 Todd C. Miller <Todd.Miller@courtesan.com>
14 * use, copy, modify, and distribute this software for any
15 * purpose with or without fee is hereby granted, provided that the above
16 * copyright notice and this permission notice appear in all copies.
18 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
19 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
20 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
21 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
22 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
23 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
24 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
26 * Sponsored in part by the Defense Advanced Research Projects
27 * Agency (DARPA) and Air Force Research Laboratory, Air Force
28 * Materiel Command, USAF, under agreement number F39502-99-1-0512.
32 #include <sys/types.h>
33 #include <sys/param.h>
50 #define FSIZE_MAX 32768
55 #define D_HEADER 1 /* Print a header/footer between files */
56 #define D_EMPTY1 2 /* Treat first file as empty (/dev/null) */
57 #define D_EMPTY2 4 /* Treat second file as empty (/dev/null) */
60 * Status values for print_status() and diffreg() return values
62 * D_SAME - files are the same
63 * D_DIFFER - files differ
64 * D_BINARY - binary files differ
65 * D_COMMON - subdirectory common to both dirs
66 * D_ONLY - file only exists in one dir
67 * D_MISMATCH1 - path1 a dir, path2 a file
68 * D_MISMATCH2 - path1 a file, path2 a dir
69 * D_ERROR - error occurred
70 * D_SKIPPED1 - skipped path1 as it is a special file
71 * D_SKIPPED2 - skipped path2 as it is a special file
75 #define D_DIFFER (1<<0)
76 #define D_BINARY (1<<1)
77 #define D_COMMON (1<<2)
79 #define D_MISMATCH1 (1<<4)
80 #define D_MISMATCH2 (1<<5)
81 #define D_ERROR (1<<6)
82 #define D_SKIPPED1 (1<<7)
83 #define D_SKIPPED2 (1<<8)
85 /* Command line options */
86 static unsigned long cmd_flags;
97 #define FLAG_T (1<<10)
98 #define FLAG_U (1<<11)
99 #define FLAG_w (1<<12)
102 char *start, *label[2];
103 struct stat stb1, stb2;
119 * The following struct is used to record change information
120 * doing a "context" or "unified" diff. (see routine "change" to
121 * understand the highly mnemonic field names)
124 int a; /* start line in old file */
125 int b; /* end line in old file */
126 int c; /* start line in new file */
127 int d; /* end line in new file */
130 static int *J; /* will be overlaid on class */
131 static int *class; /* will be overlaid on file[0] */
132 static int *klist; /* will be overlaid on file[0] after class */
133 static int *member; /* will be overlaid on file[1] */
136 static int pref, suff; /* length of prefix and suffix */
138 static int anychange;
139 static long *ixnew; /* will be overlaid on file[1] */
140 static long *ixold; /* will be overlaid on klist */
141 static struct cand *clist; /* merely a free storage pot for candidates */
142 static int clistlen; /* the length of clist */
143 static struct line *sfile[2]; /* shortened by pruning common prefix/suffix */
144 static struct context_vec *context_vec_start;
145 static struct context_vec *context_vec_end;
146 static struct context_vec *context_vec_ptr;
148 static void print_only(const char *path, size_t dirlen, const char *entry)
152 printf("Only in %.*s: %s\n", (int)dirlen, path, entry);
155 static void print_status(int val, char *path1, char *path2, char *entry)
159 print_only(path1, strlen(path1), entry);
162 printf("Common subdirectories: %s%s and %s%s\n",
163 path1, entry ? entry : "", path2, entry ? entry : "");
166 printf("Binary files %s%s and %s%s differ\n",
167 path1, entry ? entry : "", path2, entry ? entry : "");
170 if (cmd_flags & FLAG_q)
171 printf("Files %s%s and %s%s differ\n",
172 path1, entry ? entry : "",
173 path2, entry ? entry : "");
176 if (cmd_flags & FLAG_s)
177 printf("Files %s%s and %s%s are identical\n",
178 path1, entry ? entry : "",
179 path2, entry ? entry : "");
182 printf("File %s%s is a directory while file %s%s is a regular file\n",
183 path1, entry ? entry : "", path2, entry ? entry : "");
186 printf("File %s%s is a regular file while file %s%s is a directory\n",
187 path1, entry ? entry : "", path2, entry ? entry : "");
190 printf("File %s%s is not a regular file or directory and was skipped\n",
191 path1, entry ? entry : "");
194 printf("File %s%s is not a regular file or directory and was skipped\n",
195 path2, entry ? entry : "");
201 * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578.
203 static int readhash(FILE *f)
210 if (!(cmd_flags & FLAG_b) && !(cmd_flags & FLAG_w)) {
212 for (i = 0; (t = getc(f)) != '\n'; i++) {
221 for (i = 0; (t = getc(f)) != '\n'; i++) {
231 switch (t = getc(f)) {
240 if (space && !(cmd_flags & FLAG_w)) {
258 * There is a remote possibility that we end up with a zero sum.
259 * Zero is used as an EOF marker, so return 1 instead.
261 return (sum == 0 ? 1 : sum);
267 * Check to see if the given files differ.
268 * Returns 0 if they are the same, 1 if different, and -1 on error.
270 static int files_differ(FILE *f1, FILE *f2, int flags)
272 char buf1[BUFSIZ], buf2[BUFSIZ];
275 if ((flags & (D_EMPTY1|D_EMPTY2)) || stb1.st_size != stb2.st_size ||
276 (stb1.st_mode & S_IFMT) != (stb2.st_mode & S_IFMT))
279 i = fread(buf1, 1, sizeof(buf1), f1);
280 j = fread(buf2, 1, sizeof(buf2), f2);
283 if (i == 0 && j == 0) {
284 if (ferror(f1) || ferror(f2))
288 if (memcmp(buf1, buf2, i) != 0)
293 static void prepare(int i, FILE *fd, off_t filesize)
301 sz = (filesize <= FSIZE_MAX ? filesize : FSIZE_MAX) / 25;
305 p = xmalloc((sz + 3) * sizeof(struct line));
306 for (j = 0; (h = readhash(fd));) {
309 p = xrealloc(p, (sz + 3) * sizeof(struct line));
317 static void prune(void)
321 for (pref = 0; pref < len[0] && pref < len[1] &&
322 file[0][pref + 1].value == file[1][pref + 1].value;
325 for (suff = 0; suff < len[0] - pref && suff < len[1] - pref &&
326 file[0][len[0] - suff].value == file[1][len[1] - suff].value;
329 for (j = 0; j < 2; j++) {
330 sfile[j] = file[j] + pref;
331 slen[j] = len[j] - pref - suff;
332 for (i = 0; i <= slen[j]; i++)
333 sfile[j][i].serial = i;
337 static void equiv(struct line *a, int n, struct line *b, int m, int *c)
342 while (i <= n && j <= m) {
343 if (a[i].value < b[j].value)
345 else if (a[i].value == b[j].value)
356 while (b[j + 1].value == b[j].value) {
364 static int isqrt(int n) {
366 if (n == 0) return(0);
373 } while ((x - y) > 1 || (x - y) < -1);
379 static int newcand(int x, int y, int pred)
383 if (clen == clistlen) {
384 clistlen = clistlen * 11 / 10;
385 clist = xrealloc(clist, clistlen * sizeof(struct cand));
395 static int search(int *c, int k, int y)
399 if (clist[c[k]].y < y) /* quick look for typical case */
419 static int stone(int *a, int n, int *b, int *c)
424 #ifdef CONFIG_FEATURE_DIFF_MINIMAL
425 const u_int bound = (cmd_flags & FLAG_d) ? UINT_MAX : MAX(256, isqrt(n));
427 const u_int bound = MAX(256, isqrt(n));
430 c[0] = newcand(0, 0, 0);
431 for (i = 1; i <= n; i++) {
440 if (y <= clist[oldc].y)
446 if (clist[c[l]].y <= y)
449 c[l] = newcand(i, y, oldc);
454 c[l] = newcand(i, y, oldc);
458 } while ((y = b[++j]) > 0 && numtries < bound);
463 static void unravel(int p)
468 for (i = 0; i <= len[0]; i++)
469 J[i] = i <= pref ? i :
470 i > len[0] - suff ? i + len[1] - len[0] : 0;
471 for (q = clist + p; q->y != 0; q = clist + q->pred)
472 J[q->x + pref] = q->y + pref;
476 static void unsort(struct line *f, int l, int *b)
480 a = xmalloc((l + 1) * sizeof(int));
481 for (i = 1; i <= l; i++)
482 a[f[i].serial] = f[i].value;
483 for (i = 1; i <= l; i++)
488 static int skipline(FILE *f)
492 for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++)
499 * Check does double duty:
500 * 1. ferret out any fortuitous correspondences due
501 * to confounding by hashing (which result in "jackpot")
502 * 2. collect random access indexes to the two files
504 static void check(FILE *f1, FILE *f2)
506 int i, j, jackpot, c, d;
512 ixold[0] = ixnew[0] = 0;
515 for (i = 1; i <= len[0]; i++) {
517 ixold[i] = ctold += skipline(f1);
521 ixnew[j] = ctnew += skipline(f2);
524 if ((cmd_flags & FLAG_b) || (cmd_flags & FLAG_w) || (cmd_flags & FLAG_i)) {
529 * GNU diff ignores a missing newline
530 * in one file if bflag || wflag.
532 if (((cmd_flags & FLAG_b) || (cmd_flags & FLAG_w)) &&
533 ((c == EOF && d == '\n') ||
534 (c == '\n' && d == EOF))) {
539 if ((cmd_flags & FLAG_b) && isspace(c) && isspace(d)) {
544 } while (isspace(c = getc(f1)));
549 } while (isspace(d = getc(f2)));
550 } else if (cmd_flags & FLAG_w) {
551 while (isspace(c) && c != '\n') {
555 while (isspace(d) && d != '\n') {
563 if (c != '\n' && c != EOF)
564 ctold += skipline(f1);
565 if (d != '\n' && c != EOF)
566 ctnew += skipline(f2);
569 if (c == '\n' || c == EOF)
576 if ((c = getc(f1)) != (d = getc(f2))) {
578 if (c != '\n' && c != EOF)
579 ctold += skipline(f1);
580 if (d != '\n' && c != EOF)
581 ctnew += skipline(f2);
584 if (c == '\n' || c == EOF)
592 for (; j <= len[1]; j++)
593 ixnew[j] = ctnew += skipline(f2);
596 /* shellsort CACM #201 */
597 static void sort(struct line *a, int n)
599 struct line *ai, *aim, w;
604 for (j = 1; j <= n; j *= 2)
606 for (m /= 2; m != 0; m /= 2) {
608 for (j = 1; j <= k; j++) {
609 for (ai = &a[j]; ai > a; ai -= m) {
612 break; /* wraparound */
613 if (aim->value > ai[0].value ||
614 (aim->value == ai[0].value &&
615 aim->serial > ai[0].serial))
617 w.value = ai[0].value;
618 ai[0].value = aim->value;
619 aim->value = w.value;
620 w.serial = ai[0].serial;
621 ai[0].serial = aim->serial;
622 aim->serial = w.serial;
629 static void uni_range(int a, int b)
632 printf("%d,%d", a, b - a + 1);
639 static int fetch(long *f, int a, int b, FILE *lb, int ch)
641 int i, j, c, lastc, col, nc;
645 for (i = a; i <= b; i++) {
646 fseek(lb, f[i - 1], SEEK_SET);
647 nc = f[i] - f[i - 1];
650 if (cmd_flags & FLAG_T)
654 for (j = 0, lastc = '\0'; j < nc; j++, lastc = c) {
655 if ((c = getc(lb)) == EOF) {
656 puts("\n\\ No newline at end of file");
659 if (c == '\t' && (cmd_flags & FLAG_t)) {
672 static int asciifile(FILE *f)
675 if ((cmd_flags & FLAG_a) || f == NULL)
677 #ifdef CONFIG_FEATURE_DIFF_BINARY
678 unsigned char buf[BUFSIZ];
682 cnt = fread(buf, 1, sizeof(buf), f);
683 for (i = 0; i < cnt; i++)
684 if (!isprint(buf[i]) && !isspace(buf[i]))
690 /* dump accumulated "unified" diff changes */
691 static void dump_unified_vec(FILE *f1, FILE *f2)
693 struct context_vec *cvp = context_vec_start;
694 int lowa, upb, lowc, upd;
698 if (context_vec_start > context_vec_ptr)
702 lowa = MAX(1, cvp->a - context);
703 upb = MIN(len[0], context_vec_ptr->b + context);
704 lowc = MAX(1, cvp->c - context);
705 upd = MIN(len[1], context_vec_ptr->d + context);
707 fputs("@@ -", stdout);
708 uni_range(lowa, upb);
710 uni_range(lowc, upd);
711 fputs(" @@", stdout);
715 * Output changes in "unified" diff format--the old and new lines
716 * are printed together.
718 for (; cvp <= context_vec_ptr; cvp++) {
725 * c: both new and old changes
726 * d: only changes in the old file
727 * a: only changes in the new file
729 if (a <= b && c <= d)
732 ch = (a <= b) ? 'd' : 'a';
736 fetch(ixold, lowa, a - 1, f1, ' ');
737 fetch(ixold, a, b, f1, '-');
738 fetch(ixnew, c, d, f2, '+');
741 fetch(ixold, lowa, a - 1, f1, ' ');
742 fetch(ixold, a, b, f1, '-');
745 fetch(ixnew, lowc, c - 1, f2, ' ');
746 fetch(ixnew, c, d, f2, '+');
752 fetch(ixnew, d + 1, upd, f2, ' ');
754 context_vec_ptr = context_vec_start - 1;
758 static void print_header(const char *file1, const char *file2)
760 if (label[0] != NULL)
761 printf("%s %s\n", "---",
764 printf("%s %s\t%s", "---",
765 file1, ctime(&stb1.st_mtime));
766 if (label[1] != NULL)
767 printf("%s %s\n", "+++",
770 printf("%s %s\t%s", "+++",
771 file2, ctime(&stb2.st_mtime));
777 * Indicate that there is a difference between lines a and b of the from file
778 * to get to lines c to d of the to file. If a is greater then b then there
779 * are no lines in the from file involved and this means that there were
780 * lines appended (beginning at b). If c is greater than d then there are
781 * lines missing from the to file.
783 static void change(char *file1, FILE *f1, char *file2, FILE *f2, int a, int b, int c, int d)
785 static size_t max_context = 64;
787 if (a > b && c > d) return;
788 if (cmd_flags & FLAG_q) return;
791 * Allocate change records as needed.
793 if (context_vec_ptr == context_vec_end - 1) {
794 ptrdiff_t offset = context_vec_ptr - context_vec_start;
796 context_vec_start = xrealloc(context_vec_start,
797 max_context * sizeof(struct context_vec));
798 context_vec_end = context_vec_start + max_context;
799 context_vec_ptr = context_vec_start + offset;
801 if (anychange == 0) {
803 * Print the context/unidiff header first time through.
805 print_header(file1, file2);
807 } else if (a > context_vec_ptr->b + (2 * context) + 1 &&
808 c > context_vec_ptr->d + (2 * context) + 1) {
810 * If this change is more than 'context' lines from the
811 * previous change, dump the record and reset it.
813 dump_unified_vec(f1, f2);
816 context_vec_ptr->a = a;
817 context_vec_ptr->b = b;
818 context_vec_ptr->c = c;
819 context_vec_ptr->d = d;
825 static void output(char *file1, FILE *f1, char *file2, FILE *f2)
827 int m, i0, i1, j0, j1;
833 J[m + 1] = len[1] + 1;
834 for (i0 = 1; i0 <= m; i0 = i1 + 1) {
835 while (i0 <= m && J[i0] == J[i0 - 1] + 1)
839 while (i1 < m && J[i1 + 1] == 0)
843 change(file1, f1, file2, f2, i0, i1, j0, j1);
846 change(file1, f1, file2, f2, 1, 0, 1, len[1]);
848 if (anychange != 0) {
849 dump_unified_vec(f1, f2);
854 * The following code uses an algorithm due to Harold Stone,
855 * which finds a pair of longest identical subsequences in
858 * The major goal is to generate the match vector J.
859 * J[i] is the index of the line in file1 corresponding
860 * to line i file0. J[i] = 0 if there is no
861 * such line in file1.
863 * Lines are hashed so as to work in core. All potential
864 * matches are located by sorting the lines of each file
865 * on the hash (called ``value''). In particular, this
866 * collects the equivalence classes in file1 together.
867 * Subroutine equiv replaces the value of each line in
868 * file0 by the index of the first element of its
869 * matching equivalence in (the reordered) file1.
870 * To save space equiv squeezes file1 into a single
871 * array member in which the equivalence classes
872 * are simply concatenated, except that their first
873 * members are flagged by changing sign.
875 * Next the indices that point into member are unsorted into
876 * array class according to the original order of file0.
878 * The cleverness lies in routine stone. This marches
879 * through the lines of file0, developing a vector klist
880 * of "k-candidates". At step i a k-candidate is a matched
881 * pair of lines x,y (x in file0 y in file1) such that
882 * there is a common subsequence of length k
883 * between the first i lines of file0 and the first y
884 * lines of file1, but there is no such subsequence for
885 * any smaller y. x is the earliest possible mate to y
886 * that occurs in such a subsequence.
888 * Whenever any of the members of the equivalence class of
889 * lines in file1 matable to a line in file0 has serial number
890 * less than the y of some k-candidate, that k-candidate
891 * with the smallest such y is replaced. The new
892 * k-candidate is chained (via pred) to the current
893 * k-1 candidate so that the actual subsequence can
894 * be recovered. When a member has serial number greater
895 * that the y of all k-candidates, the klist is extended.
896 * At the end, the longest subsequence is pulled out
897 * and placed in the array J by unravel
899 * With J in hand, the matches there recorded are
900 * checked against reality to assure that no spurious
901 * matches have crept in due to hashing. If they have,
902 * they are broken, and "jackpot" is recorded--a harmless
903 * matter except that a true match for a spuriously
904 * mated line may now be unnecessarily reported as a change.
906 * Much of the complexity of the program comes simply
907 * from trying to minimize core utilization and
908 * maximize the range of doable problems by dynamically
909 * allocating what is needed and reusing what is not.
910 * The core requirements for problems larger than somewhat
911 * are (in words) 2*length(file0) + length(file1) +
912 * 3*(number of k-candidates installed), typically about
913 * 6n words for files of length n.
916 static int diffreg(char *ofile1, char *ofile2, int flags)
918 char *file1 = ofile1;
919 char *file2 = ofile2;
926 context_vec_ptr = context_vec_start - 1;
928 if (S_ISDIR(stb1.st_mode) != S_ISDIR(stb2.st_mode))
929 return (S_ISDIR(stb1.st_mode) ? D_MISMATCH1 : D_MISMATCH2);
930 if (strcmp(file1, "-") == 0 && strcmp(file2, "-") == 0)
933 if (flags & D_EMPTY1)
934 f1 = bb_xfopen(_PATH_DEVNULL, "r");
936 if (strcmp(file1, "-") == 0)
939 f1 = bb_xfopen(file1, "r");
942 if (flags & D_EMPTY2)
943 f2 = bb_xfopen(_PATH_DEVNULL, "r");
945 if (strcmp(file2, "-") == 0)
948 f2 = bb_xfopen(file2, "r");
951 switch (files_differ(f1, f2, flags)) {
962 if (!asciifile(f1) || !asciifile(f2)) {
968 prepare(0, f1, stb1.st_size);
969 prepare(1, f2, stb2.st_size);
971 sort(sfile[0], slen[0]);
972 sort(sfile[1], slen[1]);
974 member = (int *)file[1];
975 equiv(sfile[0], slen[0], sfile[1], slen[1], member);
976 member = xrealloc(member, (slen[1] + 2) * sizeof(int));
978 class = (int *)file[0];
979 unsort(sfile[0], slen[0], class);
980 class = xrealloc(class, (slen[0] + 2) * sizeof(int));
982 klist = xmalloc((slen[0] + 2) * sizeof(int));
985 clist = xmalloc(clistlen * sizeof(struct cand));
986 i = stone(class, slen[0], member, klist);
990 J = xrealloc(J, (len[0] + 2) * sizeof(int));
995 ixold = xrealloc(ixold, (len[0] + 2) * sizeof(long));
996 ixnew = xrealloc(ixnew, (len[1] + 2) * sizeof(long));
998 output(file1, f1, file2, f2);
1010 if (file1 != ofile1)
1012 if (file2 != ofile2)
1017 #if ENABLE_FEATURE_DIFF_DIR
1018 static void do_diff (char *dir1, char *path1, char *dir2, char *path2) {
1020 int flags = D_HEADER;
1023 char *fullpath1 = bb_xasprintf("%s/%s", dir1, path1);
1024 char *fullpath2 = bb_xasprintf("%s/%s", dir2, path2);
1026 if (stat(fullpath1, &stb1) != 0) {
1028 memset(&stb1, 0, sizeof(stb1));
1029 fullpath1 = bb_xasprintf("%s/%s", dir1, path2);
1031 if (stat(fullpath2, &stb2) != 0) {
1033 memset(&stb2, 0, sizeof(stb2));
1034 stb2.st_mode = stb1.st_mode;
1035 fullpath2 = bb_xasprintf("%s/%s", dir2, path1);
1038 if (stb1.st_mode == 0)
1039 stb1.st_mode = stb2.st_mode;
1041 if (S_ISDIR(stb1.st_mode) && S_ISDIR(stb2.st_mode)) {
1042 printf("Common subdirectories: %s and %s\n", fullpath1, fullpath2);
1046 if (!S_ISREG(stb1.st_mode) && !S_ISDIR(stb1.st_mode))
1048 else if (!S_ISREG(stb2.st_mode) && !S_ISDIR(stb2.st_mode))
1051 val = diffreg(fullpath1, fullpath2, flags);
1053 print_status(val, fullpath1, fullpath2, NULL);
1057 #ifdef CONFIG_FEATURE_DIFF_DIR
1058 static int dir_strcmp(const void *p1, const void *p2) {
1059 return strcmp(*(char * const *)p1, *(char * const *)p2);
1062 /* This function adds a filename to dl, the directory listing. */
1064 static int add_to_dirlist (const char *filename,
1065 struct stat ATTRIBUTE_UNUSED *sb, void *userdata) {
1067 dl = xrealloc(dl, dl_count * sizeof(char *));
1068 dl[dl_count - 1] = bb_xstrdup(filename);
1069 if (cmd_flags & FLAG_r) {
1070 int *pp = (int *) userdata;
1071 int path_len = *pp + 1;
1072 dl[dl_count - 1] = &(dl[dl_count - 1])[path_len];
1077 /* This returns a sorted directory listing. */
1078 static char **get_dir(char *path) {
1082 /* Reset dl_count - there's no need to free dl as bb_xrealloc does
1083 * the job nicely. */
1086 /* If -r has been set, then the recursive_action function will be
1087 * used. Unfortunately, this outputs the root directory along with
1088 * the recursed paths, so use void *userdata to specify the string
1089 * length of the root directory. It can then be removed in
1090 * add_to_dirlist. */
1092 int path_len = strlen(path);
1093 void *userdata = &path_len;
1095 /* Now fill dl with a listing. */
1096 if (cmd_flags & FLAG_r)
1097 recursive_action(path, TRUE, TRUE, FALSE, add_to_dirlist, NULL, userdata);
1101 if ((dp = opendir(path)) == NULL)
1102 bb_error_msg("Error reading directory");
1103 while ((ep = readdir(dp))) {
1104 if ((!strcmp(ep->d_name, "..")) || (!strcmp(ep->d_name, ".")))
1106 add_to_dirlist(ep->d_name, NULL, NULL);
1111 /* Sort dl alphabetically. */
1112 qsort(dl, dl_count, sizeof(char *), dir_strcmp);
1114 /* Copy dl so that we can return it. */
1115 char **retval = xmalloc(dl_count * sizeof(char *));
1116 for (i = 0; i < dl_count; i++)
1117 retval[i] = bb_xstrdup(dl[i]);
1122 static void diffdir (char *p1, char *p2) {
1124 char **dirlist1, **dirlist2;
1126 int dirlist1_count, dirlist2_count;
1129 /* Check for trailing slashes. */
1131 if (p1[strlen(p1) - 1] == '/')
1132 p1[strlen(p1) - 1] = '\0';
1133 if (p2[strlen(p2) - 1] == '/')
1134 p2[strlen(p2) - 1] = '\0';
1136 /* Get directory listings for p1 and p2. */
1138 dirlist1 = get_dir(p1);
1139 dirlist1_count = dl_count;
1140 dirlist1[dirlist1_count] = NULL;
1141 dirlist2 = get_dir(p2);
1142 dirlist2_count = dl_count;
1143 dirlist2[dirlist2_count] = NULL;
1145 /* If -S was set, find the starting point. */
1147 while (*dirlist1 != NULL && strcmp(*dirlist1, start) < 0)
1149 while (*dirlist2 != NULL && strcmp(*dirlist2, start) < 0)
1151 if ((*dirlist1 == NULL) || (*dirlist2 == NULL))
1152 bb_error_msg("Invalid argument to -S");
1155 /* Now that both dirlist1 and dirlist2 contain sorted directory
1156 * listings, we can start to go through dirlist1. If both listings
1157 * contain the same file, then do a normal diff. Otherwise, behaviour
1158 * is determined by whether the -N flag is set. */
1159 while (*dirlist1 != NULL || *dirlist2 != NULL) {
1162 pos = dp1 == NULL ? 1 : dp2 == NULL ? -1 : strcmp(dp1, dp2);
1164 do_diff(p1, dp1, p2, dp2);
1169 if (cmd_flags & FLAG_N)
1170 do_diff(p1, dp1, p2, NULL);
1172 print_only(p1, strlen(p1) + 1, dp1);
1176 if (cmd_flags & FLAG_N)
1177 do_diff(p1, NULL, p2, dp2);
1179 print_only(p2, strlen(p2) + 1, dp2);
1188 extern int diff_main(int argc, char **argv) {
1193 cmd_flags = bb_getopt_ulflags(argc, argv, "abdiNqrsS:tTU:wu", &start, &U_opt);
1195 context = 3; /* This is the default number of lines of context. */
1196 if (cmd_flags & FLAG_U) {
1197 context = strtol(U_opt, &ep, 10);
1199 bb_error_msg("Invalid context length");
1207 * Do sanity checks, fill in stb1 and stb2 and call the appropriate
1208 * driver routine. Both drivers use the contents of stb1 and stb2.
1211 bb_error_msg("Missing filename");
1214 if (strcmp(argv[0], "-") == 0) {
1215 fstat(STDIN_FILENO, &stb1);
1217 } else if (stat(argv[0], &stb1) != 0)
1218 bb_perror_msg_and_die("Couldn't stat %s", argv[0]);
1219 if (strcmp(argv[1], "-") == 0) {
1220 fstat(STDIN_FILENO, &stb2);
1222 } else if (stat(argv[1], &stb2) != 0)
1223 bb_perror_msg_and_die("Couldn't stat %s", argv[1]);
1224 if (gotstdin && (S_ISDIR(stb1.st_mode) || S_ISDIR(stb2.st_mode)))
1225 bb_error_msg_and_die("Can't compare - to a directory");
1226 if (S_ISDIR(stb1.st_mode) && S_ISDIR(stb2.st_mode)) {
1227 #ifdef CONFIG_FEATURE_DIFF_DIR
1228 diffdir(argv[0], argv[1]);
1230 bb_error_msg_and_die("Directory comparison not supported");
1234 if (S_ISDIR(stb1.st_mode)) {
1235 argv[0] = concat_path_file(argv[0], argv[1]);
1236 if (stat(argv[0], &stb1) < 0)
1237 bb_perror_msg_and_die("Couldn't stat %s", argv[0]);
1239 if (S_ISDIR(stb2.st_mode)) {
1240 argv[1] = concat_path_file(argv[1], argv[0]);
1241 if (stat(argv[1], &stb2) < 0)
1242 bb_perror_msg_and_die("Couldn't stat %s", argv[1]);
1244 print_status(diffreg(argv[0], argv[1], 0), argv[0], argv[1], NULL);