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_MISMATCH1 - path1 a dir, path2 a file
38 * D_MISMATCH2 - 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_MISMATCH1 (1 << 4)
49 #define D_MISMATCH2 (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
98 USE_FEATURE_DIFF_DIR(char **dl;)
99 USE_FEATURE_DIFF_DIR(int dl_count;)
101 /* This is the default number of lines of context. */
107 struct line *file[2];
108 int *J; /* will be overlaid on class */
109 int *class; /* will be overlaid on file[0] */
110 int *klist; /* will be overlaid on file[0] after class */
111 int *member; /* will be overlaid on file[1] */
114 int pref, suff; /* length of prefix and suffix */
117 long *ixnew; /* will be overlaid on file[1] */
118 long *ixold; /* will be overlaid on klist */
119 struct cand *clist; /* merely a free storage pot for candidates */
120 int clistlen; /* the length of clist */
121 struct line *sfile[2]; /* shortened by pruning common prefix/suffix */
122 struct context_vec *context_vec_start;
123 struct context_vec *context_vec_end;
124 struct context_vec *context_vec_ptr;
125 struct stat stb1, stb2;
127 #define G (*ptr_to_globals)
129 #define dl_count (G.dl_count )
130 #define context (G.context )
131 #define max_context (G.max_context )
132 #define status (G.status )
133 #define start (G.start )
134 #define label1 (G.label1 )
135 #define label2 (G.label2 )
136 #define file (G.file )
138 #define class (G.class )
139 #define klist (G.klist )
140 #define member (G.member )
141 #define clen (G.clen )
143 #define pref (G.pref )
144 #define suff (G.suff )
145 #define slen (G.slen )
146 #define anychange (G.anychange )
147 #define ixnew (G.ixnew )
148 #define ixold (G.ixold )
149 #define clist (G.clist )
150 #define clistlen (G.clistlen )
151 #define sfile (G.sfile )
152 #define context_vec_start (G.context_vec_start )
153 #define context_vec_end (G.context_vec_end )
154 #define context_vec_ptr (G.context_vec_ptr )
155 #define stb1 (G.stb1 )
156 #define stb2 (G.stb2 )
157 #define INIT_G() do { \
158 SET_PTR_TO_GLOBALS(xzalloc(sizeof(G))); \
164 /*static void print_only(const char *path, size_t dirlen, const char *entry)*/
165 static void print_only(const char *path, const char *entry)
167 printf("Only in %s: %s\n", path, entry);
171 /*static void print_status(int val, char *path1, char *path2, char *entry)*/
172 static void print_status(int val, char *_path1, char *_path2)
174 /*const char *const _entry = entry ? entry : "";*/
175 /*char *const _path1 = entry ? concat_path_file(path1, _entry) : path1;*/
176 /*char *const _path2 = entry ? concat_path_file(path2, _entry) : path2;*/
180 print_only(path1, entry);
184 printf("Common subdirectories: %s and %s\n", _path1, _path2);
187 printf("Binary files %s and %s differ\n", _path1, _path2);
190 if (option_mask32 & FLAG_q)
191 printf("Files %s and %s differ\n", _path1, _path2);
194 if (option_mask32 & FLAG_s)
195 printf("Files %s and %s are identical\n", _path1, _path2);
198 printf("File %s is a %s while file %s is a %s\n",
199 _path1, "directory", _path2, "regular file");
202 printf("File %s is a %s while file %s is a %s\n",
203 _path1, "regular file", _path2, "directory");
206 printf("File %s is not a regular file or directory and was skipped\n",
210 printf("File %s is not a regular file or directory and was skipped\n",
223 /* Read line, return its nonzero hash. Return 0 if EOF.
225 * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578.
227 static ALWAYS_INLINE int fiddle_sum(int sum, int t)
229 return sum * 127 + t;
231 static int readhash(FILE *fp)
239 if (!(option_mask32 & (FLAG_b | FLAG_w))) {
240 while ((t = getc(fp)) != '\n') {
246 sum = fiddle_sum(sum, t);
251 switch (t = getc(fp)) {
260 if (space && !(option_mask32 & FLAG_w)) {
264 sum = fiddle_sum(sum, t);
278 * There is a remote possibility that we end up with a zero sum.
279 * Zero is used as an EOF marker, so return 1 instead.
281 return (sum == 0 ? 1 : sum);
286 * Check to see if the given files differ.
287 * Returns 0 if they are the same, 1 if different, and -1 on error.
289 static NOINLINE int files_differ(FILE *f1, FILE *f2, int flags)
293 if ((flags & (D_EMPTY1 | D_EMPTY2)) || stb1.st_size != stb2.st_size
294 || (stb1.st_mode & S_IFMT) != (stb2.st_mode & S_IFMT)
299 i = fread(g_read_buf, 1, COMMON_BUFSIZE/2, f1);
300 j = fread(g_read_buf + COMMON_BUFSIZE/2, 1, COMMON_BUFSIZE/2, f2);
304 return (ferror(f1) || ferror(f2)) ? -1 : 0;
305 if (memcmp(g_read_buf,
306 g_read_buf + COMMON_BUFSIZE/2, i) != 0)
312 static void prepare(int i, FILE *fp /*, off_t filesize*/)
320 /*sz = (filesize <= FSIZE_MAX ? filesize : FSIZE_MAX) / 25;*/
324 p = xmalloc((sz + 3) * sizeof(p[0]));
326 while ((h = readhash(fp)) != 0) { /* while not EOF */
329 p = xrealloc(p, (sz + 3) * sizeof(p[0]));
338 static void prune(void)
342 for (pref = 0; pref < len[0] && pref < len[1] &&
343 file[0][pref + 1].value == file[1][pref + 1].value; pref++)
345 for (suff = 0; suff < len[0] - pref && suff < len[1] - pref &&
346 file[0][len[0] - suff].value == file[1][len[1] - suff].value;
349 for (j = 0; j < 2; j++) {
350 sfile[j] = file[j] + pref;
351 slen[j] = len[j] - pref - suff;
352 for (i = 0; i <= slen[j]; i++)
353 sfile[j][i].serial = i;
358 static void equiv(struct line *a, int n, struct line *b, int m, int *c)
363 while (i <= n && j <= m) {
364 if (a[i].value < b[j].value)
366 else if (a[i].value == b[j].value)
377 while (b[j + 1].value == b[j].value) {
386 static int isqrt(int n)
398 } while ((x - y) > 1 || (x - y) < -1);
404 static int newcand(int x, int y, int pred)
408 if (clen == clistlen) {
409 clistlen = clistlen * 11 / 10;
410 clist = xrealloc(clist, clistlen * sizeof(struct cand));
420 static int search(int *c, int k, int y)
424 if (clist[c[k]].y < y) /* quick look for typical case */
444 static int stone(int *a, int n, int *b, int *c)
448 unsigned int numtries;
449 #if ENABLE_FEATURE_DIFF_MINIMAL
450 const unsigned int bound =
451 (option_mask32 & FLAG_d) ? UINT_MAX : MAX(256, isqrt(n));
453 const unsigned int bound = MAX(256, isqrt(n));
457 c[0] = newcand(0, 0, 0);
458 for (i = 1; i <= n; i++) {
467 if (y <= clist[oldc].y)
473 if (clist[c[l]].y <= y)
476 c[l] = newcand(i, y, oldc);
481 c[l] = newcand(i, y, oldc);
485 } while ((y = b[++j]) > 0 && numtries < bound);
491 static void unravel(int p)
496 for (i = 0; i <= len[0]; i++)
497 J[i] = i <= pref ? i : i > len[0] - suff ? i + len[1] - len[0] : 0;
498 for (q = clist + p; q->y != 0; q = clist + q->pred)
499 J[q->x + pref] = q->y + pref;
503 static void unsort(struct line *f, int l, int *b)
507 a = xmalloc((l + 1) * sizeof(int));
508 for (i = 1; i <= l; i++)
509 a[f[i].serial] = f[i].value;
510 for (i = 1; i <= l; i++)
516 static int skipline(FILE *f)
520 for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++)
527 * Check does double duty:
528 * 1. ferret out any fortuitous correspondences due
529 * to confounding by hashing (which result in "jackpot")
530 * 2. collect random access indexes to the two files
532 static NOINLINE void check(FILE *f1, FILE *f2)
534 int i, j, jackpot, c, d;
540 ixold[0] = ixnew[0] = 0;
543 for (i = 1; i <= len[0]; i++) {
545 ixold[i] = ctold += skipline(f1);
549 ixnew[j] = ctnew += skipline(f2);
552 if (option_mask32 & (FLAG_b | FLAG_w | FLAG_i)) {
557 * GNU diff ignores a missing newline
558 * in one file if bflag || wflag.
560 if ((option_mask32 & (FLAG_b | FLAG_w))
561 && ((c == EOF && d == '\n') || (c == '\n' && d == EOF))
567 if ((option_mask32 & FLAG_b) && isspace(c) && isspace(d)) {
573 } while (isspace(c));
579 } while (isspace(d));
580 } else if (option_mask32 & FLAG_w) {
581 while (isspace(c) && c != '\n') {
585 while (isspace(d) && d != '\n') {
593 if (c != '\n' && c != EOF)
594 ctold += skipline(f1);
595 if (d != '\n' && c != EOF)
596 ctnew += skipline(f2);
599 if (c == '\n' || c == EOF)
610 if (c != '\n' && c != EOF)
611 ctold += skipline(f1);
612 // BUG? Should be "if (d != '\n' && d != EOF)" ?
613 if (d != '\n' && c != EOF)
614 ctnew += skipline(f2);
617 if (c == '\n' || c == EOF)
625 for (; j <= len[1]; j++)
626 ixnew[j] = ctnew += skipline(f2);
630 /* shellsort CACM #201 */
631 static void sort(struct line *a, int n)
633 struct line *ai, *aim, w;
638 for (j = 1; j <= n; j *= 2)
640 for (m /= 2; m != 0; m /= 2) {
642 for (j = 1; j <= k; j++) {
643 for (ai = &a[j]; ai > a; ai -= m) {
646 break; /* wraparound */
647 if (aim->value > ai[0].value
648 || (aim->value == ai[0].value && aim->serial > ai[0].serial)
652 w.value = ai[0].value;
653 ai[0].value = aim->value;
654 aim->value = w.value;
655 w.serial = ai[0].serial;
656 ai[0].serial = aim->serial;
657 aim->serial = w.serial;
664 static void uni_range(int a, int b)
667 printf("%d,%d", a, b - a + 1);
675 static void fetch(long *f, int a, int b, FILE *lb, int ch)
677 int i, j, c, lastc, col, nc;
681 for (i = a; i <= b; i++) {
682 fseek(lb, f[i - 1], SEEK_SET);
683 nc = f[i] - f[i - 1];
686 if (option_mask32 & FLAG_T)
690 for (j = 0, lastc = '\0'; j < nc; j++, lastc = c) {
693 printf("\n\\ No newline at end of file\n");
696 if (c == '\t' && (option_mask32 & FLAG_t)) {
709 #if ENABLE_FEATURE_DIFF_BINARY
710 static int asciifile(FILE *f)
714 if (option_mask32 & FLAG_a)
717 cnt = fread(g_read_buf, 1, COMMON_BUFSIZE, f);
718 for (i = 0; i < cnt; i++) {
719 if (!isprint(g_read_buf[i])
720 && !isspace(g_read_buf[i])
728 #define asciifile(f) 1
732 /* dump accumulated "unified" diff changes */
733 static void dump_unified_vec(FILE *f1, FILE *f2)
735 struct context_vec *cvp = context_vec_start;
736 int lowa, upb, lowc, upd;
740 if (context_vec_start > context_vec_ptr)
744 lowa = MAX(1, cvp->a - context);
745 upb = MIN(len[0], context_vec_ptr->b + context);
746 lowc = MAX(1, cvp->c - context);
747 upd = MIN(len[1], context_vec_ptr->d + context);
750 uni_range(lowa, upb);
752 uni_range(lowc, upd);
756 * Output changes in "unified" diff format--the old and new lines
757 * are printed together.
759 for (; cvp <= context_vec_ptr; cvp++) {
766 * c: both new and old changes
767 * d: only changes in the old file
768 * a: only changes in the new file
770 if (a <= b && c <= d)
773 ch = (a <= b) ? 'd' : 'a';
778 fetch(ixold, lowa, a - 1, f1, ' ');
779 fetch(ixold, a, b, f1, '-');
780 fetch(ixnew, c, d, f2, '+');
783 fetch(ixold, lowa, a - 1, f1, ' ');
784 fetch(ixold, a, b, f1, '-');
787 fetch(ixnew, lowc, c - 1, f2, ' ');
788 fetch(ixnew, c, d, f2, '+');
792 if (ch == 'c' || ch == 'd') {
793 fetch(ixold, lowa, a - 1, f1, ' ');
794 fetch(ixold, a, b, f1, '-');
797 fetch(ixnew, lowc, c - 1, f2, ' ');
798 if (ch == 'c' || ch == 'a')
799 fetch(ixnew, c, d, f2, '+');
804 fetch(ixnew, d + 1, upd, f2, ' ');
806 context_vec_ptr = context_vec_start - 1;
810 static void print_header(const char *file1, const char *file2)
813 printf("--- %s\n", label1);
815 printf("--- %s\t%s", file1, ctime(&stb1.st_mtime));
817 printf("+++ %s\n", label2);
819 printf("+++ %s\t%s", file2, ctime(&stb2.st_mtime));
824 * Indicate that there is a difference between lines a and b of the from file
825 * to get to lines c to d of the to file. If a is greater than b then there
826 * are no lines in the from file involved and this means that there were
827 * lines appended (beginning at b). If c is greater than d then there are
828 * lines missing from the to file.
830 static void change(char *file1, FILE *f1, char *file2, FILE *f2,
831 int a, int b, int c, int d)
833 if ((a > b && c > d) || (option_mask32 & FLAG_q)) {
839 * Allocate change records as needed.
841 if (context_vec_ptr == context_vec_end - 1) {
842 ptrdiff_t offset = context_vec_ptr - context_vec_start;
845 context_vec_start = xrealloc(context_vec_start,
846 max_context * sizeof(struct context_vec));
847 context_vec_end = context_vec_start + max_context;
848 context_vec_ptr = context_vec_start + offset;
850 if (anychange == 0) {
852 * Print the context/unidiff header first time through.
854 print_header(file1, file2);
855 } else if (a > context_vec_ptr->b + (2 * context) + 1
856 && c > context_vec_ptr->d + (2 * context) + 1
859 * If this change is more than 'context' lines from the
860 * previous change, dump the record and reset it.
862 // dump_unified_vec() seeks!
863 dump_unified_vec(f1, f2);
866 context_vec_ptr->a = a;
867 context_vec_ptr->b = b;
868 context_vec_ptr->c = c;
869 context_vec_ptr->d = d;
874 static void output(char *file1, FILE *f1, char *file2, FILE *f2)
876 /* Note that j0 and j1 can't be used as they are defined in math.h.
877 * This also allows the rather amusing variable 'j00'... */
878 int m, i0, i1, j00, j01;
884 J[m + 1] = len[1] + 1;
885 for (i0 = 1; i0 <= m; i0 = i1 + 1) {
886 while (i0 <= m && J[i0] == J[i0 - 1] + 1)
890 while (i1 < m && J[i1 + 1] == 0)
895 change(file1, f1, file2, f2, i0, i1, j00, j01);
899 change(file1, f1, file2, f2, 1, 0, 1, len[1]);
901 if (anychange != 0 && !(option_mask32 & FLAG_q)) {
902 // dump_unified_vec() seeks!
903 dump_unified_vec(f1, f2);
908 * The following code uses an algorithm due to Harold Stone,
909 * which finds a pair of longest identical subsequences in
912 * The major goal is to generate the match vector J.
913 * J[i] is the index of the line in file1 corresponding
914 * to line i in file0. J[i] = 0 if there is no
915 * such line in file1.
917 * Lines are hashed so as to work in core. All potential
918 * matches are located by sorting the lines of each file
919 * on the hash (called "value"). In particular, this
920 * collects the equivalence classes in file1 together.
921 * Subroutine equiv replaces the value of each line in
922 * file0 by the index of the first element of its
923 * matching equivalence in (the reordered) file1.
924 * To save space equiv squeezes file1 into a single
925 * array member in which the equivalence classes
926 * are simply concatenated, except that their first
927 * members are flagged by changing sign.
929 * Next the indices that point into member are unsorted into
930 * array class according to the original order of file0.
932 * The cleverness lies in routine stone. This marches
933 * through the lines of file0, developing a vector klist
934 * of "k-candidates". At step i a k-candidate is a matched
935 * pair of lines x,y (x in file0, y in file1) such that
936 * there is a common subsequence of length k
937 * between the first i lines of file0 and the first y
938 * lines of file1, but there is no such subsequence for
939 * any smaller y. x is the earliest possible mate to y
940 * that occurs in such a subsequence.
942 * Whenever any of the members of the equivalence class of
943 * lines in file1 matable to a line in file0 has serial number
944 * less than the y of some k-candidate, that k-candidate
945 * with the smallest such y is replaced. The new
946 * k-candidate is chained (via pred) to the current
947 * k-1 candidate so that the actual subsequence can
948 * be recovered. When a member has serial number greater
949 * that the y of all k-candidates, the klist is extended.
950 * At the end, the longest subsequence is pulled out
951 * and placed in the array J by unravel
953 * With J in hand, the matches there recorded are
954 * checked against reality to assure that no spurious
955 * matches have crept in due to hashing. If they have,
956 * they are broken, and "jackpot" is recorded--a harmless
957 * matter except that a true match for a spuriously
958 * mated line may now be unnecessarily reported as a change.
960 * Much of the complexity of the program comes simply
961 * from trying to minimize core utilization and
962 * maximize the range of doable problems by dynamically
963 * allocating what is needed and reusing what is not.
964 * The core requirements for problems larger than somewhat
965 * are (in words) 2*length(file0) + length(file1) +
966 * 3*(number of k-candidates installed), typically about
967 * 6n words for files of length n.
969 static unsigned diffreg(char *file1, char *file2, int flags)
977 context_vec_ptr = context_vec_start - 1;
979 if (S_ISDIR(stb1.st_mode) != S_ISDIR(stb2.st_mode))
980 return (S_ISDIR(stb1.st_mode) ? D_MISMATCH1 : D_MISMATCH2);
982 if (LONE_DASH(file1) && LONE_DASH(file2))
987 if (flags & D_EMPTY1)
988 f1 = xfopen(bb_dev_null, "r");
990 f1 = xfopen_stdin(file1);
991 if (flags & D_EMPTY2)
992 f2 = xfopen(bb_dev_null, "r");
994 f2 = xfopen_stdin(file2);
996 /* We can't diff non-seekable stream - we use rewind(), fseek().
997 * This can be fixed (volunteers?).
998 * Meanwhile we should check it here by stat'ing input fds,
999 * but I am lazy and check that in main() instead.
1000 * Check in main won't catch "diffing fifos buried in subdirectories"
1001 * failure scenario - not very likely in real life... */
1003 /* Quick check whether they are different */
1004 i = files_differ(f1, f2, flags);
1007 else if (i != 1) { /* 1 == ok */
1013 if (!asciifile(f1) || !asciifile(f2)) {
1020 prepare(0, f1 /*, stb1.st_size*/);
1021 prepare(1, f2 /*, stb2.st_size*/);
1023 sort(sfile[0], slen[0]);
1024 sort(sfile[1], slen[1]);
1026 member = (int *) file[1];
1027 equiv(sfile[0], slen[0], sfile[1], slen[1], member);
1028 member = xrealloc(member, (slen[1] + 2) * sizeof(int));
1030 class = (int *) file[0];
1031 unsort(sfile[0], slen[0], class);
1032 class = xrealloc(class, (slen[0] + 2) * sizeof(int));
1034 klist = xmalloc((slen[0] + 2) * sizeof(int));
1037 clist = xmalloc(clistlen * sizeof(struct cand));
1038 i = stone(class, slen[0], member, klist);
1042 J = xrealloc(J, (len[0] + 2) * sizeof(int));
1047 ixold = xrealloc(ixold, (len[0] + 2) * sizeof(long));
1048 ixnew = xrealloc(ixnew, (len[1] + 2) * sizeof(long));
1052 output(file1, f1, file2, f2);
1060 fclose_if_not_stdin(f1);
1061 fclose_if_not_stdin(f2);
1066 #if ENABLE_FEATURE_DIFF_DIR
1067 static void do_diff(char *dir1, char *path1, char *dir2, char *path2)
1069 int flags = D_HEADER;
1071 char *fullpath1 = NULL; /* if -N */
1072 char *fullpath2 = NULL;
1075 fullpath1 = concat_path_file(dir1, path1);
1077 fullpath2 = concat_path_file(dir2, path2);
1079 if (!fullpath1 || stat(fullpath1, &stb1) != 0) {
1081 memset(&stb1, 0, sizeof(stb1));
1084 fullpath1 = concat_path_file(dir1, path2);
1087 if (!fullpath2 || stat(fullpath2, &stb2) != 0) {
1089 memset(&stb2, 0, sizeof(stb2));
1090 stb2.st_mode = stb1.st_mode;
1093 fullpath2 = concat_path_file(dir2, path1);
1097 if (stb1.st_mode == 0)
1098 stb1.st_mode = stb2.st_mode;
1100 if (S_ISDIR(stb1.st_mode) && S_ISDIR(stb2.st_mode)) {
1101 printf("Common subdirectories: %s and %s\n", fullpath1, fullpath2);
1105 if (!S_ISREG(stb1.st_mode) && !S_ISDIR(stb1.st_mode))
1107 else if (!S_ISREG(stb2.st_mode) && !S_ISDIR(stb2.st_mode))
1110 val = diffreg(fullpath1, fullpath2, flags);
1112 print_status(val, fullpath1, fullpath2 /*, NULL*/);
1120 #if ENABLE_FEATURE_DIFF_DIR
1121 /* This function adds a filename to dl, the directory listing. */
1122 static int add_to_dirlist(const char *filename,
1123 struct stat ATTRIBUTE_UNUSED *sb,
1125 int depth ATTRIBUTE_UNUSED)
1127 /* +2: with space for eventual trailing NULL */
1128 dl = xrealloc(dl, (dl_count+2) * sizeof(dl[0]));
1129 dl[dl_count] = xstrdup(filename + (int)(ptrdiff_t)userdata);
1135 /* This returns a sorted directory listing. */
1136 static char **get_dir(char *path)
1139 dl = xzalloc(sizeof(dl[0]));
1141 /* If -r has been set, then the recursive_action function will be
1142 * used. Unfortunately, this outputs the root directory along with
1143 * the recursed paths, so use void *userdata to specify the string
1144 * length of the root directory - '(void*)(strlen(path)+)'.
1145 * add_to_dirlist then removes root dir prefix. */
1147 if (option_mask32 & FLAG_r) {
1148 recursive_action(path, ACTION_RECURSE|ACTION_FOLLOWLINKS,
1149 add_to_dirlist, NULL,
1150 (void*)(strlen(path)+1), 0);
1155 dp = warn_opendir(path);
1156 while ((ep = readdir(dp))) {
1157 if (!strcmp(ep->d_name, "..") || LONE_CHAR(ep->d_name, '.'))
1159 add_to_dirlist(ep->d_name, NULL, (void*)(int)0, 0);
1164 /* Sort dl alphabetically. */
1165 qsort_string_vector(dl, dl_count);
1167 dl[dl_count] = NULL;
1172 static void diffdir(char *p1, char *p2)
1174 char **dirlist1, **dirlist2;
1178 /* Check for trailing slashes. */
1179 dp1 = last_char_is(p1, '/');
1182 dp2 = last_char_is(p2, '/');
1186 /* Get directory listings for p1 and p2. */
1187 dirlist1 = get_dir(p1);
1188 dirlist2 = get_dir(p2);
1190 /* If -S was set, find the starting point. */
1192 while (*dirlist1 != NULL && strcmp(*dirlist1, start) < 0)
1194 while (*dirlist2 != NULL && strcmp(*dirlist2, start) < 0)
1196 if ((*dirlist1 == NULL) || (*dirlist2 == NULL))
1197 bb_error_msg(bb_msg_invalid_arg, "NULL", "-S");
1200 /* Now that both dirlist1 and dirlist2 contain sorted directory
1201 * listings, we can start to go through dirlist1. If both listings
1202 * contain the same file, then do a normal diff. Otherwise, behaviour
1203 * is determined by whether the -N flag is set. */
1204 while (*dirlist1 != NULL || *dirlist2 != NULL) {
1207 pos = dp1 == NULL ? 1 : dp2 == NULL ? -1 : strcmp(dp1, dp2);
1209 do_diff(p1, dp1, p2, dp2);
1212 } else if (pos < 0) {
1213 if (option_mask32 & FLAG_N)
1214 do_diff(p1, dp1, p2, NULL);
1216 print_only(p1, dp1);
1219 if (option_mask32 & FLAG_N)
1220 do_diff(p1, NULL, p2, dp2);
1222 print_only(p2, dp2);
1230 int diff_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
1231 int diff_main(int argc ATTRIBUTE_UNUSED, char **argv)
1235 llist_t *L_arg = NULL;
1239 /* exactly 2 params; collect multiple -L <label>; -U N */
1240 opt_complementary = "=2:L::U+";
1241 getopt32(argv, "abdiL:NqrsS:tTU:wu"
1242 "p" /* ignored (for compatibility) */,
1243 &L_arg, &start, &context);
1247 if (label1 && label2)
1250 label1 = L_arg->data;
1251 else { /* then label2 is NULL */
1253 label1 = L_arg->data;
1255 /* we leak L_arg here... */
1256 L_arg = L_arg->link;
1260 * Do sanity checks, fill in stb1 and stb2 and call the appropriate
1261 * driver routine. Both drivers use the contents of stb1 and stb2.
1265 if (LONE_DASH(f1)) {
1266 fstat(STDIN_FILENO, &stb1);
1270 if (LONE_DASH(f2)) {
1271 fstat(STDIN_FILENO, &stb2);
1275 if (gotstdin && (S_ISDIR(stb1.st_mode) || S_ISDIR(stb2.st_mode)))
1276 bb_error_msg_and_die("can't compare - to a directory");
1277 if (S_ISDIR(stb1.st_mode) && S_ISDIR(stb2.st_mode)) {
1278 #if ENABLE_FEATURE_DIFF_DIR
1281 bb_error_msg_and_die("directory comparison not supported");
1284 if (S_ISDIR(stb1.st_mode)) {
1285 f1 = concat_path_file(f1, f2);
1288 if (S_ISDIR(stb2.st_mode)) {
1289 f2 = concat_path_file(f2, f1);
1293 /* We can't diff e.g. stdin supplied by a pipe - we use rewind(), fseek().
1294 * This can be fixed (volunteers?) */
1295 if (!S_ISREG(stb1.st_mode) || !S_ISREG(stb2.st_mode))
1296 bb_error_msg_and_die("can't diff non-seekable stream");
1297 print_status(diffreg(f1, f2, 0), f1, f2 /*, NULL*/);