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>
6 * Copyright (c) 2003 Todd C. Miller <Todd.Miller@courtesan.com>
8 * Sponsored in part by the Defense Advanced Research Projects
9 * Agency (DARPA) and Air Force Research Laboratory, Air Force
10 * Materiel Command, USAF, under agreement number F39502-99-1-0512.
12 * Licensed under GPLv2 or later, see file LICENSE in this tarball for details.
17 // #define FSIZE_MAX 32768
19 /* NOINLINEs added to prevent gcc from merging too much into diffreg()
20 * (it bites more than it can (efficiently) chew). */
25 #define D_HEADER 1 /* Print a header/footer between files */
26 #define D_EMPTY1 2 /* Treat first file as empty (/dev/null) */
27 #define D_EMPTY2 4 /* Treat second file as empty (/dev/null) */
30 * Status values for print_status() and diffreg() return values
32 * D_SAME - files are the same
33 * D_DIFFER - files differ
34 * D_BINARY - binary files differ
35 * D_COMMON - subdirectory common to both dirs
36 * D_ONLY - file only exists in one dir
37 * D_ISDIR1 - path1 a dir, path2 a file
38 * D_ISDIR2 - path1 a file, path2 a dir
39 * D_ERROR - error occurred
40 * D_SKIPPED1 - skipped path1 as it is a special file
41 * D_SKIPPED2 - skipped path2 as it is a special file
44 #define D_DIFFER (1 << 0)
45 #define D_BINARY (1 << 1)
46 #define D_COMMON (1 << 2)
47 /*#define D_ONLY (1 << 3) - unused */
48 #define D_ISDIR1 (1 << 4)
49 #define D_ISDIR2 (1 << 5)
50 #define D_ERROR (1 << 6)
51 #define D_SKIPPED1 (1 << 7)
52 #define D_SKIPPED2 (1 << 8)
54 /* Command line options */
55 #define FLAG_a (1 << 0)
56 #define FLAG_b (1 << 1)
57 #define FLAG_d (1 << 2)
58 #define FLAG_i (1 << 3)
59 #define FLAG_L (1 << 4)
60 #define FLAG_N (1 << 5)
61 #define FLAG_q (1 << 6)
62 #define FLAG_r (1 << 7)
63 #define FLAG_s (1 << 8)
64 #define FLAG_S (1 << 9)
65 #define FLAG_t (1 << 10)
66 #define FLAG_T (1 << 11)
67 #define FLAG_U (1 << 12)
68 #define FLAG_w (1 << 13)
83 * The following struct is used to record change information
84 * doing a "context" or "unified" diff. (see routine "change" to
85 * understand the highly mnemonic field names)
88 int a; /* start line in old file */
89 int b; /* end line in old file */
90 int c; /* start line in new file */
91 int d; /* end line in new file */
95 #define g_read_buf bb_common_bufsiz1
101 size_t max_context; /* size of context_vec_start */
102 USE_FEATURE_DIFF_DIR(int dl_count;)
103 USE_FEATURE_DIFF_DIR(char **dl;)
107 int *J; /* will be overlaid on class */
109 int pref, suff; /* length of prefix and suffix */
112 int clistlen; /* the length of clist */
113 struct cand *clist; /* merely a free storage pot for candidates */
114 long *ixnew; /* will be overlaid on nfile[1] */
115 long *ixold; /* will be overlaid on klist */
116 struct line *nfile[2];
117 struct line *sfile[2]; /* shortened by pruning common prefix/suffix */
118 struct context_vec *context_vec_start;
119 struct context_vec *context_vec_end;
120 struct context_vec *context_vec_ptr;
121 char *tempname1, *tempname2;
122 struct stat stb1, stb2;
124 #define G (*ptr_to_globals)
125 #define anychange (G.anychange )
126 #define exit_status (G.exit_status )
127 #define opt_U_context (G.opt_U_context )
128 #define max_context (G.max_context )
129 #define dl_count (G.dl_count )
131 #define opt_S_start (G.opt_S_start )
132 #define label1 (G.label1 )
133 #define label2 (G.label2 )
135 #define clen (G.clen )
136 #define pref (G.pref )
137 #define suff (G.suff )
138 #define nlen (G.nlen )
139 #define slen (G.slen )
140 #define clistlen (G.clistlen )
141 #define clist (G.clist )
142 #define ixnew (G.ixnew )
143 #define ixold (G.ixold )
144 #define nfile (G.nfile )
145 #define sfile (G.sfile )
146 #define context_vec_start (G.context_vec_start )
147 #define context_vec_end (G.context_vec_end )
148 #define context_vec_ptr (G.context_vec_ptr )
149 #define stb1 (G.stb1 )
150 #define stb2 (G.stb2 )
151 #define tempname1 (G.tempname1 )
152 #define tempname2 (G.tempname2 )
153 #define INIT_G() do { \
154 SET_PTR_TO_GLOBALS(xzalloc(sizeof(G))); \
160 /*static void print_only(const char *path, size_t dirlen, const char *entry)*/
161 static void print_only(const char *path, const char *entry)
163 printf("Only in %s: %s\n", path, entry);
167 /*static void print_status(int val, char *path1, char *path2, char *entry)*/
168 static void print_status(int val, char *_path1, char *_path2)
170 /*const char *const _entry = entry ? entry : "";*/
171 /*char *const _path1 = entry ? concat_path_file(path1, _entry) : path1;*/
172 /*char *const _path2 = entry ? concat_path_file(path2, _entry) : path2;*/
176 print_only(path1, entry);
180 printf("Common subdirectories: %s and %s\n", _path1, _path2);
183 printf("Binary files %s and %s differ\n", _path1, _path2);
186 if (option_mask32 & FLAG_q)
187 printf("Files %s and %s differ\n", _path1, _path2);
190 if (option_mask32 & FLAG_s)
191 printf("Files %s and %s are identical\n", _path1, _path2);
194 printf("File %s is a %s while file %s is a %s\n",
195 _path1, "directory", _path2, "regular file");
198 printf("File %s is a %s while file %s is a %s\n",
199 _path1, "regular file", _path2, "directory");
202 printf("File %s is not a regular file or directory and was skipped\n",
206 printf("File %s is not a regular file or directory and was skipped\n",
219 /* Read line, return its nonzero hash. Return 0 if EOF.
221 * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578.
223 static ALWAYS_INLINE int fiddle_sum(int sum, int t)
225 return sum * 127 + t;
227 static int readhash(FILE *fp)
235 if (!(option_mask32 & (FLAG_b | FLAG_w))) {
236 while ((t = getc(fp)) != '\n') {
242 sum = fiddle_sum(sum, t);
247 switch (t = getc(fp)) {
256 if (space && !(option_mask32 & FLAG_w)) {
260 sum = fiddle_sum(sum, t);
274 * There is a remote possibility that we end up with a zero sum.
275 * Zero is used as an EOF marker, so return 1 instead.
277 return (sum == 0 ? 1 : sum);
281 static char *make_temp(FILE *f, struct stat *sb)
286 if (S_ISREG(sb->st_mode))
288 name = xstrdup("/tmp/difXXXXXX");
291 bb_perror_msg_and_die("mkstemp");
292 if (bb_copyfd_eof(fileno(f), fd) < 0) {
295 xfunc_die(); /* error message is printed by bb_copyfd_eof */
299 if (freopen(name, "r+", f) == NULL) {
300 bb_perror_msg("freopen");
308 * Check to see if the given files differ.
309 * Returns 0 if they are the same, 1 if different, and -1 on error.
311 static NOINLINE int files_differ(FILE *f1, FILE *f2, int flags)
315 tempname1 = make_temp(f1, &stb1);
316 tempname2 = make_temp(f2, &stb2);
317 if ((flags & (D_EMPTY1 | D_EMPTY2)) || stb1.st_size != stb2.st_size) {
321 i = fread(g_read_buf, 1, COMMON_BUFSIZE/2, f1);
322 j = fread(g_read_buf + COMMON_BUFSIZE/2, 1, COMMON_BUFSIZE/2, f2);
326 return (ferror(f1) || ferror(f2)) ? -1 : 0;
327 if (memcmp(g_read_buf,
328 g_read_buf + COMMON_BUFSIZE/2, i) != 0)
334 static void prepare(int i, FILE *fp /*, off_t filesize*/)
342 /*sz = (filesize <= FSIZE_MAX ? filesize : FSIZE_MAX) / 25;*/
346 p = xmalloc((sz + 3) * sizeof(p[0]));
348 while ((h = readhash(fp)) != 0) { /* while not EOF */
351 p = xrealloc(p, (sz + 3) * sizeof(p[0]));
360 static void prune(void)
364 for (pref = 0; pref < nlen[0] && pref < nlen[1] &&
365 nfile[0][pref + 1].value == nfile[1][pref + 1].value; pref++)
367 for (suff = 0; suff < nlen[0] - pref && suff < nlen[1] - pref &&
368 nfile[0][nlen[0] - suff].value == nfile[1][nlen[1] - suff].value;
371 for (j = 0; j < 2; j++) {
372 sfile[j] = nfile[j] + pref;
373 slen[j] = nlen[j] - pref - suff;
374 for (i = 0; i <= slen[j]; i++)
375 sfile[j][i].serial = i;
380 static void equiv(struct line *a, int n, struct line *b, int m, int *c)
385 while (i <= n && j <= m) {
386 if (a[i].value < b[j].value)
388 else if (a[i].value == b[j].value)
399 while (b[j + 1].value == b[j].value) {
408 static int isqrt(int n)
420 } while ((x - y) > 1 || (x - y) < -1);
426 static int newcand(int x, int y, int pred)
430 if (clen == clistlen) {
431 clistlen = clistlen * 11 / 10;
432 clist = xrealloc(clist, clistlen * sizeof(struct cand));
442 static int search(int *c, int k, int y)
446 if (clist[c[k]].y < y) /* quick look for typical case */
466 static int stone(int *a, int n, int *b, int *c)
470 unsigned int numtries;
471 #if ENABLE_FEATURE_DIFF_MINIMAL
472 const unsigned int bound =
473 (option_mask32 & FLAG_d) ? UINT_MAX : MAX(256, isqrt(n));
475 const unsigned int bound = MAX(256, isqrt(n));
479 c[0] = newcand(0, 0, 0);
480 for (i = 1; i <= n; i++) {
489 if (y <= clist[oldc].y)
495 if (clist[c[l]].y <= y)
498 c[l] = newcand(i, y, oldc);
503 c[l] = newcand(i, y, oldc);
507 } while ((y = b[++j]) > 0 && numtries < bound);
513 static void unravel(int p)
518 for (i = 0; i <= nlen[0]; i++)
519 J[i] = i <= pref ? i : i > nlen[0] - suff ? i + nlen[1] - nlen[0] : 0;
520 for (q = clist + p; q->y != 0; q = clist + q->pred)
521 J[q->x + pref] = q->y + pref;
525 static void unsort(struct line *f, int l, int *b)
529 a = xmalloc((l + 1) * sizeof(int));
530 for (i = 1; i <= l; i++)
531 a[f[i].serial] = f[i].value;
532 for (i = 1; i <= l; i++)
538 static int skipline(FILE *f)
542 for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++)
549 * Check does double duty:
550 * 1. ferret out any fortuitous correspondences due
551 * to confounding by hashing (which result in "jackpot")
552 * 2. collect random access indexes to the two files
554 static NOINLINE void check(FILE *f1, FILE *f2)
556 int i, j, jackpot, c, d;
562 ixold[0] = ixnew[0] = 0;
565 for (i = 1; i <= nlen[0]; i++) {
567 ixold[i] = ctold += skipline(f1);
571 ixnew[j] = ctnew += skipline(f2);
574 if (option_mask32 & (FLAG_b | FLAG_w | FLAG_i)) {
579 * GNU diff ignores a missing newline
580 * in one file if bflag || wflag.
582 if ((option_mask32 & (FLAG_b | FLAG_w))
583 && ((c == EOF && d == '\n') || (c == '\n' && d == EOF))
589 if ((option_mask32 & FLAG_b) && isspace(c) && isspace(d)) {
595 } while (isspace(c));
601 } while (isspace(d));
602 } else if (option_mask32 & FLAG_w) {
603 while (isspace(c) && c != '\n') {
607 while (isspace(d) && d != '\n') {
615 if (c != '\n' && c != EOF)
616 ctold += skipline(f1);
617 if (d != '\n' && c != EOF)
618 ctnew += skipline(f2);
621 if (c == '\n' || c == EOF)
632 if (c != '\n' && c != EOF)
633 ctold += skipline(f1);
634 // BUG? Should be "if (d != '\n' && d != EOF)" ?
635 if (d != '\n' && c != EOF)
636 ctnew += skipline(f2);
639 if (c == '\n' || c == EOF)
647 for (; j <= nlen[1]; j++)
648 ixnew[j] = ctnew += skipline(f2);
652 /* shellsort CACM #201 */
653 static void sort(struct line *a, int n)
655 struct line *ai, *aim, w;
660 for (j = 1; j <= n; j *= 2)
662 for (m /= 2; m != 0; m /= 2) {
664 for (j = 1; j <= k; j++) {
665 for (ai = &a[j]; ai > a; ai -= m) {
668 break; /* wraparound */
669 if (aim->value > ai[0].value
670 || (aim->value == ai[0].value && aim->serial > ai[0].serial)
674 w.value = ai[0].value;
675 ai[0].value = aim->value;
676 aim->value = w.value;
677 w.serial = ai[0].serial;
678 ai[0].serial = aim->serial;
679 aim->serial = w.serial;
686 static void uni_range(int a, int b)
689 printf("%d,%d", a, b - a + 1);
697 static void fetch(long *f, int a, int b, FILE *lb, int ch)
699 int i, j, c, lastc, col, nc;
703 for (i = a; i <= b; i++) {
704 fseek(lb, f[i - 1], SEEK_SET);
705 nc = f[i] - f[i - 1];
708 if (option_mask32 & FLAG_T)
712 for (j = 0, lastc = '\0'; j < nc; j++, lastc = c) {
715 printf("\n\\ No newline at end of file\n");
718 if (c == '\t' && (option_mask32 & FLAG_t)) {
731 #if ENABLE_FEATURE_DIFF_BINARY
732 static int asciifile(FILE *f)
736 if (option_mask32 & FLAG_a)
739 cnt = fread(g_read_buf, 1, COMMON_BUFSIZE, f);
740 for (i = 0; i < cnt; i++) {
741 if (!isprint(g_read_buf[i])
742 && !isspace(g_read_buf[i])
750 #define asciifile(f) 1
754 /* dump accumulated "unified" diff changes */
755 static void dump_unified_vec(FILE *f1, FILE *f2)
757 struct context_vec *cvp = context_vec_start;
758 int lowa, upb, lowc, upd;
762 if (context_vec_start > context_vec_ptr)
766 lowa = MAX(1, cvp->a - opt_U_context);
767 upb = MIN(nlen[0], context_vec_ptr->b + opt_U_context);
768 lowc = MAX(1, cvp->c - opt_U_context);
769 upd = MIN(nlen[1], context_vec_ptr->d + opt_U_context);
772 uni_range(lowa, upb);
774 uni_range(lowc, upd);
778 * Output changes in "unified" diff format--the old and new lines
779 * are printed together.
781 for (; cvp <= context_vec_ptr; cvp++) {
788 * c: both new and old changes
789 * d: only changes in the old file
790 * a: only changes in the new file
792 if (a <= b && c <= d)
795 ch = (a <= b) ? 'd' : 'a';
800 fetch(ixold, lowa, a - 1, f1, ' ');
801 fetch(ixold, a, b, f1, '-');
802 fetch(ixnew, c, d, f2, '+');
805 fetch(ixold, lowa, a - 1, f1, ' ');
806 fetch(ixold, a, b, f1, '-');
809 fetch(ixnew, lowc, c - 1, f2, ' ');
810 fetch(ixnew, c, d, f2, '+');
814 if (ch == 'c' || ch == 'd') {
815 fetch(ixold, lowa, a - 1, f1, ' ');
816 fetch(ixold, a, b, f1, '-');
819 fetch(ixnew, lowc, c - 1, f2, ' ');
820 if (ch == 'c' || ch == 'a')
821 fetch(ixnew, c, d, f2, '+');
826 fetch(ixnew, d + 1, upd, f2, ' ');
828 context_vec_ptr = context_vec_start - 1;
832 static void print_header(const char *file1, const char *file2)
835 printf("--- %s\n", label1);
837 printf("--- %s\t%s", file1, ctime(&stb1.st_mtime));
839 printf("+++ %s\n", label2);
841 printf("+++ %s\t%s", file2, ctime(&stb2.st_mtime));
846 * Indicate that there is a difference between lines a and b of the from file
847 * to get to lines c to d of the to file. If a is greater than b then there
848 * are no lines in the from file involved and this means that there were
849 * lines appended (beginning at b). If c is greater than d then there are
850 * lines missing from the to file.
852 static void change(char *file1, FILE *f1, char *file2, FILE *f2,
853 int a, int b, int c, int d)
855 if ((a > b && c > d) || (option_mask32 & FLAG_q)) {
861 * Allocate change records as needed.
863 if (context_vec_ptr == context_vec_end - 1) {
864 ptrdiff_t offset = context_vec_ptr - context_vec_start;
867 context_vec_start = xrealloc(context_vec_start,
868 max_context * sizeof(struct context_vec));
869 context_vec_end = context_vec_start + max_context;
870 context_vec_ptr = context_vec_start + offset;
872 if (anychange == 0) {
874 * Print the context/unidiff header first time through.
876 print_header(file1, file2);
877 } else if (a > context_vec_ptr->b + (2 * opt_U_context) + 1
878 && c > context_vec_ptr->d + (2 * opt_U_context) + 1
881 * If this change is more than 'context' lines from the
882 * previous change, dump the record and reset it.
884 // dump_unified_vec() seeks!
885 dump_unified_vec(f1, f2);
888 context_vec_ptr->a = a;
889 context_vec_ptr->b = b;
890 context_vec_ptr->c = c;
891 context_vec_ptr->d = d;
896 static void output(char *file1, FILE *f1, char *file2, FILE *f2)
898 /* Note that j0 and j1 can't be used as they are defined in math.h.
899 * This also allows the rather amusing variable 'j00'... */
900 int m, i0, i1, j00, j01;
906 J[m + 1] = nlen[1] + 1;
907 for (i0 = 1; i0 <= m; i0 = i1 + 1) {
908 while (i0 <= m && J[i0] == J[i0 - 1] + 1)
912 while (i1 < m && J[i1 + 1] == 0)
917 change(file1, f1, file2, f2, i0, i1, j00, j01);
921 change(file1, f1, file2, f2, 1, 0, 1, nlen[1]);
923 if (anychange != 0 && !(option_mask32 & FLAG_q)) {
924 // dump_unified_vec() seeks!
925 dump_unified_vec(f1, f2);
930 * The following code uses an algorithm due to Harold Stone,
931 * which finds a pair of longest identical subsequences in
934 * The major goal is to generate the match vector J.
935 * J[i] is the index of the line in file1 corresponding
936 * to line i in file0. J[i] = 0 if there is no
937 * such line in file1.
939 * Lines are hashed so as to work in core. All potential
940 * matches are located by sorting the lines of each file
941 * on the hash (called "value"). In particular, this
942 * collects the equivalence classes in file1 together.
943 * Subroutine equiv replaces the value of each line in
944 * file0 by the index of the first element of its
945 * matching equivalence in (the reordered) file1.
946 * To save space equiv squeezes file1 into a single
947 * array member in which the equivalence classes
948 * are simply concatenated, except that their first
949 * members are flagged by changing sign.
951 * Next the indices that point into member are unsorted into
952 * array class according to the original order of file0.
954 * The cleverness lies in routine stone. This marches
955 * through the lines of file0, developing a vector klist
956 * of "k-candidates". At step i a k-candidate is a matched
957 * pair of lines x,y (x in file0, y in file1) such that
958 * there is a common subsequence of length k
959 * between the first i lines of file0 and the first y
960 * lines of file1, but there is no such subsequence for
961 * any smaller y. x is the earliest possible mate to y
962 * that occurs in such a subsequence.
964 * Whenever any of the members of the equivalence class of
965 * lines in file1 matable to a line in file0 has serial number
966 * less than the y of some k-candidate, that k-candidate
967 * with the smallest such y is replaced. The new
968 * k-candidate is chained (via pred) to the current
969 * k-1 candidate so that the actual subsequence can
970 * be recovered. When a member has serial number greater
971 * that the y of all k-candidates, the klist is extended.
972 * At the end, the longest subsequence is pulled out
973 * and placed in the array J by unravel
975 * With J in hand, the matches there recorded are
976 * checked against reality to assure that no spurious
977 * matches have crept in due to hashing. If they have,
978 * they are broken, and "jackpot" is recorded--a harmless
979 * matter except that a true match for a spuriously
980 * mated line may now be unnecessarily reported as a change.
982 * Much of the complexity of the program comes simply
983 * from trying to minimize core utilization and
984 * maximize the range of doable problems by dynamically
985 * allocating what is needed and reusing what is not.
986 * The core requirements for problems larger than somewhat
987 * are (in words) 2*length(file0) + length(file1) +
988 * 3*(number of k-candidates installed), typically about
989 * 6n words for files of length n.
991 /* NB: files can be not REGular. The only sure thing that they
992 * are not both DIRectories. */
993 static unsigned diffreg(char *file1, char *file2, int flags)
995 int *member; /* will be overlaid on nfile[1] */
996 int *class; /* will be overlaid on nfile[0] */
997 int *klist; /* will be overlaid on nfile[0] after class */
1004 context_vec_ptr = context_vec_start - 1;
1006 /* Is any of them a directory? Then it's simple */
1007 if (S_ISDIR(stb1.st_mode) != S_ISDIR(stb2.st_mode))
1008 return (S_ISDIR(stb1.st_mode) ? D_ISDIR1 : D_ISDIR2);
1010 /* None of them are directories */
1013 if (flags & D_EMPTY1)
1014 f1 = xfopen(bb_dev_null, "r");
1016 f1 = xfopen_stdin(file1);
1017 if (flags & D_EMPTY2)
1018 f2 = xfopen(bb_dev_null, "r");
1020 f2 = xfopen_stdin(file2);
1022 /* Quick check whether they are different */
1023 /* NB: copies non-REG files to tempfiles and fills tempname1/2 */
1024 i = files_differ(f1, f2, flags);
1025 if (i != 1) { /* not different? */
1026 if (i != 0) /* error? */
1031 if (!asciifile(f1) || !asciifile(f2)) {
1038 prepare(0, f1 /*, stb1.st_size*/);
1039 prepare(1, f2 /*, stb2.st_size*/);
1041 sort(sfile[0], slen[0]);
1042 sort(sfile[1], slen[1]);
1044 member = (int *) nfile[1];
1045 equiv(sfile[0], slen[0], sfile[1], slen[1], member);
1046 member = xrealloc(member, (slen[1] + 2) * sizeof(int));
1048 class = (int *) nfile[0];
1049 unsort(sfile[0], slen[0], class);
1050 class = xrealloc(class, (slen[0] + 2) * sizeof(int));
1052 klist = xmalloc((slen[0] + 2) * sizeof(int));
1055 clist = xmalloc(clistlen * sizeof(struct cand));
1056 i = stone(class, slen[0], member, klist);
1060 J = xrealloc(J, (nlen[0] + 2) * sizeof(int));
1065 ixold = xrealloc(ixold, (nlen[0] + 2) * sizeof(long));
1066 ixnew = xrealloc(ixnew, (nlen[1] + 2) * sizeof(long));
1070 output(file1, f1, file2, f2);
1078 fclose_if_not_stdin(f1);
1079 fclose_if_not_stdin(f2);
1092 #if ENABLE_FEATURE_DIFF_DIR
1093 static void do_diff(char *dir1, char *path1, char *dir2, char *path2)
1095 int flags = D_HEADER;
1097 char *fullpath1 = NULL; /* if -N */
1098 char *fullpath2 = NULL;
1101 fullpath1 = concat_path_file(dir1, path1);
1103 fullpath2 = concat_path_file(dir2, path2);
1105 if (!fullpath1 || stat(fullpath1, &stb1) != 0) {
1107 memset(&stb1, 0, sizeof(stb1));
1110 fullpath1 = concat_path_file(dir1, path2);
1113 if (!fullpath2 || stat(fullpath2, &stb2) != 0) {
1115 memset(&stb2, 0, sizeof(stb2));
1116 stb2.st_mode = stb1.st_mode;
1119 fullpath2 = concat_path_file(dir2, path1);
1123 if (stb1.st_mode == 0)
1124 stb1.st_mode = stb2.st_mode;
1126 if (S_ISDIR(stb1.st_mode) && S_ISDIR(stb2.st_mode)) {
1127 printf("Common subdirectories: %s and %s\n", fullpath1, fullpath2);
1131 if (!S_ISREG(stb1.st_mode) && !S_ISDIR(stb1.st_mode))
1133 else if (!S_ISREG(stb2.st_mode) && !S_ISDIR(stb2.st_mode))
1136 /* Both files are either REGular or DIRectories */
1137 val = diffreg(fullpath1, fullpath2, flags);
1140 print_status(val, fullpath1, fullpath2 /*, NULL*/);
1148 #if ENABLE_FEATURE_DIFF_DIR
1149 /* This function adds a filename to dl, the directory listing. */
1150 static int add_to_dirlist(const char *filename,
1151 struct stat ATTRIBUTE_UNUSED *sb,
1153 int depth ATTRIBUTE_UNUSED)
1155 /* +2: with space for eventual trailing NULL */
1156 dl = xrealloc(dl, (dl_count+2) * sizeof(dl[0]));
1157 dl[dl_count] = xstrdup(filename + (int)(ptrdiff_t)userdata);
1163 /* This returns a sorted directory listing. */
1164 static char **get_recursive_dirlist(char *path)
1167 dl = xzalloc(sizeof(dl[0]));
1169 /* If -r has been set, then the recursive_action function will be
1170 * used. Unfortunately, this outputs the root directory along with
1171 * the recursed paths, so use void *userdata to specify the string
1172 * length of the root directory - '(void*)(strlen(path)+)'.
1173 * add_to_dirlist then removes root dir prefix. */
1174 if (option_mask32 & FLAG_r) {
1175 recursive_action(path, ACTION_RECURSE|ACTION_FOLLOWLINKS,
1176 add_to_dirlist, NULL,
1177 (void*)(strlen(path)+1), 0);
1182 dp = warn_opendir(path);
1183 while ((ep = readdir(dp))) {
1184 if (!strcmp(ep->d_name, "..") || LONE_CHAR(ep->d_name, '.'))
1186 add_to_dirlist(ep->d_name, NULL, (void*)(int)0, 0);
1191 /* Sort dl alphabetically. */
1192 qsort_string_vector(dl, dl_count);
1194 dl[dl_count] = NULL;
1199 static void diffdir(char *p1, char *p2)
1201 char **dirlist1, **dirlist2;
1205 /* Check for trailing slashes. */
1206 dp1 = last_char_is(p1, '/');
1209 dp2 = last_char_is(p2, '/');
1213 /* Get directory listings for p1 and p2. */
1214 dirlist1 = get_recursive_dirlist(p1);
1215 dirlist2 = get_recursive_dirlist(p2);
1217 /* If -S was set, find the starting point. */
1219 while (*dirlist1 != NULL && strcmp(*dirlist1, opt_S_start) < 0)
1221 while (*dirlist2 != NULL && strcmp(*dirlist2, opt_S_start) < 0)
1223 if ((*dirlist1 == NULL) || (*dirlist2 == NULL))
1224 bb_error_msg(bb_msg_invalid_arg, "NULL", "-S");
1227 /* Now that both dirlist1 and dirlist2 contain sorted directory
1228 * listings, we can start to go through dirlist1. If both listings
1229 * contain the same file, then do a normal diff. Otherwise, behaviour
1230 * is determined by whether the -N flag is set. */
1231 while (*dirlist1 != NULL || *dirlist2 != NULL) {
1234 pos = dp1 == NULL ? 1 : (dp2 == NULL ? -1 : strcmp(dp1, dp2));
1236 do_diff(p1, dp1, p2, dp2);
1239 } else if (pos < 0) {
1240 if (option_mask32 & FLAG_N)
1241 do_diff(p1, dp1, p2, NULL);
1243 print_only(p1, dp1);
1246 if (option_mask32 & FLAG_N)
1247 do_diff(p1, NULL, p2, dp2);
1249 print_only(p2, dp2);
1257 int diff_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
1258 int diff_main(int argc ATTRIBUTE_UNUSED, char **argv)
1262 llist_t *L_arg = NULL;
1266 /* exactly 2 params; collect multiple -L <label>; -U N */
1267 opt_complementary = "=2:L::U+";
1268 getopt32(argv, "abdiL:NqrsS:tTU:wu"
1269 "p" /* ignored (for compatibility) */,
1270 &L_arg, &opt_S_start, &opt_U_context);
1274 if (label1 && label2)
1277 label1 = L_arg->data;
1278 else { /* then label2 is NULL */
1280 label1 = L_arg->data;
1282 /* we leak L_arg here... */
1283 L_arg = L_arg->link;
1287 * Do sanity checks, fill in stb1 and stb2 and call the appropriate
1288 * driver routine. Both drivers use the contents of stb1 and stb2.
1292 if (LONE_DASH(f1)) {
1293 fstat(STDIN_FILENO, &stb1);
1297 if (LONE_DASH(f2)) {
1298 fstat(STDIN_FILENO, &stb2);
1303 if (gotstdin && (S_ISDIR(stb1.st_mode) || S_ISDIR(stb2.st_mode)))
1304 bb_error_msg_and_die("can't compare stdin to a directory");
1306 if (S_ISDIR(stb1.st_mode) && S_ISDIR(stb2.st_mode)) {
1307 #if ENABLE_FEATURE_DIFF_DIR
1311 bb_error_msg_and_die("no support for directory comparison");
1315 if (S_ISDIR(stb1.st_mode)) { /* "diff dir file" */
1316 /* NB: "diff dir dir2/dir3/file" must become
1317 * "diff dir/file dir2/dir3/file" */
1318 char *slash = strrchr(f2, '/');
1319 f1 = concat_path_file(f1, slash ? slash+1 : f2);
1322 if (S_ISDIR(stb2.st_mode)) {
1323 char *slash = strrchr(f1, '/');
1324 f2 = concat_path_file(f2, slash ? slash+1 : f1);
1328 /* diffreg can get non-regular files here,
1329 * they are not both DIRestories */
1330 print_status((gotstdin > 1 ? D_SAME : diffreg(f1, f2, 0)),