diff --git a/coreutils/diff.c b/coreutils/diff.c index b2945dd87..17975ad20 100644 --- a/coreutils/diff.c +++ b/coreutils/diff.c @@ -3,14 +3,14 @@ * Mini diff implementation for busybox, adapted from OpenBSD diff. * * Copyright (C) 2006 by Robert Sullivan - * + * * Licensed under GPLv2 or later, see file LICENSE in this tarball for details. */ /* * Copyright (c) 2003 Todd C. Miller * - * Permission to + * Permission to * use, copy, modify, and distribute this software for any * purpose with or without fee is hereby granted, provided that the above * copyright notice and this permission notice appear in all copies. @@ -71,8 +71,8 @@ * D_SKIPPED2 - skipped path2 as it is a special file */ -#define D_SAME 0 -#define D_DIFFER 1 +#define D_SAME 0 +#define D_DIFFER (1<<0) #define D_BINARY (1<<1) #define D_COMMON (1<<2) #define D_ONLY (1<<3) @@ -84,7 +84,7 @@ /* Command line options */ static unsigned long cmd_flags; -#define FLAG_a 1 +#define FLAG_a (1<<0) #define FLAG_b (1<<1) #define FLAG_d (1<<2) #define FLAG_i (1<<3) @@ -145,207 +145,14 @@ static struct context_vec *context_vec_start; static struct context_vec *context_vec_end; static struct context_vec *context_vec_ptr; -static void output(char *, FILE *, char *, FILE *); -static void check(char *, FILE *, char *, FILE *); -static void uni_range(int, int); -static void dump_unified_vec(FILE *, FILE *); -static void prepare(int, FILE *, off_t); -static void prune(void); -static void equiv(struct line *, int, struct line *, int, int *); -static void unravel(int); -static void unsort(struct line *, int, int *); -static void change(char *, FILE *, char *, FILE *, int, int, int, int); -static void sort(struct line *, int); -static void print_header(const char *, const char *); -static int asciifile(FILE *); -static int fetch(long *, int, int, FILE *, int, int); -static int newcand(int, int, int); -static int search(int *, int, int); -static int skipline(FILE *); -static int stone(int *, int, int *, int *); -static int readhash(FILE *); -static int files_differ(FILE *, FILE *, int); - -static int diffreg(char *, char *, int); -static void print_only(const char *, size_t, const char *); -static void print_status(int, char *, char *, char *); - -#ifdef CONFIG_FEATURE_DIFF_DIR -static int dir_strcmp(const void *p1, const void *p2) { - return strcmp(*(char * const *)p1, *(char * const *)p2); -} - -/* This function adds a filename to dl, the directory listing. */ - -static int add_to_dirlist (const char *filename, struct stat *statbuf, void *userdata) { - dl_count++; - dl = xrealloc(dl, dl_count * sizeof(char *)); - dl[dl_count - 1] = bb_xstrdup(filename); - if (cmd_flags & FLAG_r) { - int *pp = (int *) userdata; - int path_len = *pp + 1; - dl[dl_count - 1] = &(dl[dl_count - 1])[path_len]; - } - return TRUE; -} - -/* This returns a sorted directory listing. */ -static char **get_dir(char *path) { - - int i; - - /* Reset dl_count - there's no need to free dl as bb_xrealloc does - * the job nicely. */ - dl_count = 0; - - /* If -r has been set, then the recursive_action function will be - * used. Unfortunately, this outputs the root directory along with - * the recursed paths, so use void *userdata to specify the string - * length of the root directory. It can then be removed in - * add_to_dirlist. */ - - int path_len = strlen(path); - void *userdata = &path_len; - - /* Now fill dl with a listing. */ - if (cmd_flags & FLAG_r) - recursive_action(path, TRUE, TRUE, FALSE, add_to_dirlist, NULL, userdata); - else { - DIR *dp; - struct dirent *ep; - if ((dp = opendir(path)) == NULL) - bb_error_msg("Error reading directory"); - while ((ep = readdir(dp))) { - if ((!strcmp(ep->d_name, "..")) || (!strcmp(ep->d_name, "."))) - continue; - add_to_dirlist(ep->d_name, NULL, NULL); - } - closedir(dp); - } - - /* Sort dl alphabetically. */ - qsort(dl, dl_count, sizeof(char *), dir_strcmp); - - /* Copy dl so that we can return it. */ - char **retval = xmalloc(dl_count * sizeof(char *)); - for (i = 0; i < dl_count; i++) - retval[i] = bb_xstrdup(dl[i]); - - return retval; -} - -static void do_diff (char *dir1, char *path1, char *dir2, char *path2) { - - int flags = D_HEADER; - int val; - - char *fullpath1 = bb_xasprintf("%s/%s", dir1, path1); - char *fullpath2 = bb_xasprintf("%s/%s", dir2, path2); - - if (stat(fullpath1, &stb1) != 0) { - flags |= D_EMPTY1; - memset(&stb1, 0, sizeof(stb1)); - fullpath1 = bb_xasprintf("%s/%s", dir1, path2); - } - if (stat(fullpath2, &stb2) != 0) { - flags |= D_EMPTY2; - memset(&stb2, 0, sizeof(stb2)); - stb2.st_mode = stb1.st_mode; - fullpath2 = bb_xasprintf("%s/%s", dir2, path1); - } - - if (stb1.st_mode == 0) - stb1.st_mode = stb2.st_mode; - - if (S_ISDIR(stb1.st_mode) && S_ISDIR(stb2.st_mode)) { - printf("Common subdirectories: %s and %s\n", fullpath1, fullpath2); - return; - } - - if (!S_ISREG(stb1.st_mode) && !S_ISDIR(stb1.st_mode)) - val = D_SKIPPED1; - else if (!S_ISREG(stb2.st_mode) && !S_ISDIR(stb2.st_mode)) - val = D_SKIPPED2; - else - val = diffreg(fullpath1, fullpath2, flags); - - print_status(val, fullpath1, fullpath2, NULL); -} - -static void diffdir (char *p1, char *p2) { - - char **dirlist1, **dirlist2; - char *dp1, *dp2; - int dirlist1_count, dirlist2_count; - int pos; - - /* Check for trailing slashes. */ - - if (p1[strlen(p1) - 1] == '/') - p1[strlen(p1) - 1] = '\0'; - if (p2[strlen(p2) - 1] == '/') - p2[strlen(p2) - 1] = '\0'; - - /* Get directory listings for p1 and p2. */ - - dirlist1 = get_dir(p1); - dirlist1_count = dl_count; - dirlist1[dirlist1_count] = NULL; - dirlist2 = get_dir(p2); - dirlist2_count = dl_count; - dirlist2[dirlist2_count] = NULL; - - /* If -S was set, find the starting point. */ - if (start) { - while (*dirlist1 != NULL && strcmp(*dirlist1, start) < 0) - dirlist1++; - while (*dirlist2 != NULL && strcmp(*dirlist2, start) < 0) - dirlist2++; - if ((*dirlist1 == NULL) || (*dirlist2 == NULL)) - bb_error_msg("Invalid argument to -S"); - } - - /* Now that both dirlist1 and dirlist2 contain sorted directory - * listings, we can start to go through dirlist1. If both listings - * contain the same file, then do a normal diff. Otherwise, behaviour - * is determined by whether the -N flag is set. */ - while (*dirlist1 != NULL || *dirlist2 != NULL) { - dp1 = *dirlist1; - dp2 = *dirlist2; - pos = dp1 == NULL ? 1 : dp2 == NULL ? -1 : strcmp(dp1, dp2); - if (pos == 0) { - do_diff(p1, dp1, p2, dp2); - dirlist1++; - dirlist2++; - } - else if (pos < 0) { - if (cmd_flags & FLAG_N) - do_diff(p1, dp1, p2, NULL); - else - print_only(p1, strlen(p1) + 1, dp1); - dirlist1++; - } - else { - if (cmd_flags & FLAG_N) - do_diff(p1, NULL, p2, dp2); - else - print_only(p2, strlen(p2) + 1, dp2); - dirlist2++; - } - } -} -#endif - -static void -print_only(const char *path, size_t dirlen, const char *entry) +static void print_only(const char *path, size_t dirlen, const char *entry) { if (dirlen > 1) dirlen--; printf("Only in %.*s: %s\n", (int)dirlen, path, entry); } -static void -print_status(int val, char *path1, char *path2, char *entry) +static void print_status(int val, char *path1, char *path2, char *entry) { switch (val) { case D_ONLY: @@ -390,6 +197,659 @@ print_status(int val, char *path1, char *path2, char *entry) } } +/* + * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578. + */ +static int readhash(FILE *f) +{ + int i, t, space; + int sum; + + sum = 1; + space = 0; + if (!(cmd_flags & FLAG_b) && !(cmd_flags & FLAG_w)) { + if (FLAG_i) + for (i = 0; (t = getc(f)) != '\n'; i++) { + if (t == EOF) { + if (i == 0) + return (0); + break; + } + sum = sum * 127 + t; + } + else + for (i = 0; (t = getc(f)) != '\n'; i++) { + if (t == EOF) { + if (i == 0) + return (0); + break; + } + sum = sum * 127 + t; + } + } else { + for (i = 0;;) { + switch (t = getc(f)) { + case '\t': + case '\r': + case '\v': + case '\f': + case ' ': + space++; + continue; + default: + if (space && !(cmd_flags & FLAG_w)) { + i++; + space = 0; + } + sum = sum * 127 + t; + i++; + continue; + case EOF: + if (i == 0) + return (0); + /* FALLTHROUGH */ + case '\n': + break; + } + break; + } + } + /* + * There is a remote possibility that we end up with a zero sum. + * Zero is used as an EOF marker, so return 1 instead. + */ + return (sum == 0 ? 1 : sum); +} + + + +/* + * Check to see if the given files differ. + * Returns 0 if they are the same, 1 if different, and -1 on error. + */ +static int files_differ(FILE *f1, FILE *f2, int flags) +{ + char buf1[BUFSIZ], buf2[BUFSIZ]; + size_t i, j; + + if ((flags & (D_EMPTY1|D_EMPTY2)) || stb1.st_size != stb2.st_size || + (stb1.st_mode & S_IFMT) != (stb2.st_mode & S_IFMT)) + return (1); + while(1) { + i = fread(buf1, 1, sizeof(buf1), f1); + j = fread(buf2, 1, sizeof(buf2), f2); + if (i != j) + return (1); + if (i == 0 && j == 0) { + if (ferror(f1) || ferror(f2)) + return (1); + return (0); + } + if (memcmp(buf1, buf2, i) != 0) + return (1); + } +} + +static void prepare(int i, FILE *fd, off_t filesize) +{ + struct line *p; + int j, h; + size_t sz; + + rewind(fd); + + sz = (filesize <= FSIZE_MAX ? filesize : FSIZE_MAX) / 25; + if (sz < 100) + sz = 100; + + p = xmalloc((sz + 3) * sizeof(struct line)); + for (j = 0; (h = readhash(fd));) { + if (j == sz) { + sz = sz * 3 / 2; + p = xrealloc(p, (sz + 3) * sizeof(struct line)); + } + p[++j].value = h; + } + len[i] = j; + file[i] = p; +} + +static void prune(void) +{ + int i, j; + + for (pref = 0; pref < len[0] && pref < len[1] && + file[0][pref + 1].value == file[1][pref + 1].value; + pref++) + ; + for (suff = 0; suff < len[0] - pref && suff < len[1] - pref && + file[0][len[0] - suff].value == file[1][len[1] - suff].value; + suff++) + ; + for (j = 0; j < 2; j++) { + sfile[j] = file[j] + pref; + slen[j] = len[j] - pref - suff; + for (i = 0; i <= slen[j]; i++) + sfile[j][i].serial = i; + } +} + +static void equiv(struct line *a, int n, struct line *b, int m, int *c) +{ + int i, j; + + i = j = 1; + while (i <= n && j <= m) { + if (a[i].value < b[j].value) + a[i++].value = 0; + else if (a[i].value == b[j].value) + a[i++].value = j; + else + j++; + } + while (i <= n) + a[i++].value = 0; + b[m + 1].value = 0; + j = 0; + while (++j <= m) { + c[j] = -b[j].serial; + while (b[j + 1].value == b[j].value) { + j++; + c[j] = b[j].serial; + } + } + c[j] = -1; +} + +static int isqrt(int n) { + int y, x = 1; + if (n == 0) return(0); + + do { + y = x; + x = n / x; + x += y; + x /= 2; + } while ((x - y) > 1 || (x - y) < -1); + + return (x); +} + + +static int newcand(int x, int y, int pred) +{ + struct cand *q; + + if (clen == clistlen) { + clistlen = clistlen * 11 / 10; + clist = xrealloc(clist, clistlen * sizeof(struct cand)); + } + q = clist + clen; + q->x = x; + q->y = y; + q->pred = pred; + return (clen++); +} + + +static int search(int *c, int k, int y) +{ + int i, j, l, t; + + if (clist[c[k]].y < y) /* quick look for typical case */ + return (k + 1); + i = 0; + j = k + 1; + while (1) { + l = i + j; + if ((l >>= 1) <= i) + break; + t = clist[c[l]].y; + if (t > y) + j = l; + else if (t < y) + i = l; + else + return (l); + } + return (l + 1); +} + + +static int stone(int *a, int n, int *b, int *c) +{ + int i, k, y, j, l; + int oldc, tc, oldl; + u_int numtries; +#ifdef CONFIG_FEATURE_DIFF_MINIMAL + const u_int bound = (cmd_flags & FLAG_d) ? UINT_MAX : MAX(256, isqrt(n)); +#else + const u_int bound = MAX(256, isqrt(n)); +#endif + k = 0; + c[0] = newcand(0, 0, 0); + for (i = 1; i <= n; i++) { + j = a[i]; + if (j == 0) + continue; + y = -b[j]; + oldl = 0; + oldc = c[0]; + numtries = 0; + do { + if (y <= clist[oldc].y) + continue; + l = search(c, k, y); + if (l != oldl + 1) + oldc = c[l - 1]; + if (l <= k) { + if (clist[c[l]].y <= y) + continue; + tc = c[l]; + c[l] = newcand(i, y, oldc); + oldc = tc; + oldl = l; + numtries++; + } else { + c[l] = newcand(i, y, oldc); + k++; + break; + } + } while ((y = b[++j]) > 0 && numtries < bound); + } + return (k); +} + +static void unravel(int p) +{ + struct cand *q; + int i; + + for (i = 0; i <= len[0]; i++) + J[i] = i <= pref ? i : + i > len[0] - suff ? i + len[1] - len[0] : 0; + for (q = clist + p; q->y != 0; q = clist + q->pred) + J[q->x + pref] = q->y + pref; +} + + +static void unsort(struct line *f, int l, int *b) +{ + int *a, i; + + a = xmalloc((l + 1) * sizeof(int)); + for (i = 1; i <= l; i++) + a[f[i].serial] = f[i].value; + for (i = 1; i <= l; i++) + b[i] = a[i]; + free(a); +} + +static int skipline(FILE *f) +{ + int i, c; + + for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++) + continue; + return (i); +} + + +/* + * Check does double duty: + * 1. ferret out any fortuitous correspondences due + * to confounding by hashing (which result in "jackpot") + * 2. collect random access indexes to the two files + */ +static void check(char *file1, FILE *f1, char *file2, FILE *f2) +{ + int i, j, jackpot, c, d; + long ctold, ctnew; + + rewind(f1); + rewind(f2); + j = 1; + ixold[0] = ixnew[0] = 0; + jackpot = 0; + ctold = ctnew = 0; + for (i = 1; i <= len[0]; i++) { + if (J[i] == 0) { + ixold[i] = ctold += skipline(f1); + continue; + } + while (j < J[i]) { + ixnew[j] = ctnew += skipline(f2); + j++; + } + if ((cmd_flags & FLAG_b) || (cmd_flags & FLAG_w) || (cmd_flags & FLAG_i)) { + while (1) { + c = getc(f1); + d = getc(f2); + /* + * GNU diff ignores a missing newline + * in one file if bflag || wflag. + */ + if (((cmd_flags & FLAG_b) || (cmd_flags & FLAG_w)) && + ((c == EOF && d == '\n') || + (c == '\n' && d == EOF))) { + break; + } + ctold++; + ctnew++; + if ((cmd_flags & FLAG_b) && isspace(c) && isspace(d)) { + do { + if (c == '\n') + break; + ctold++; + } while (isspace(c = getc(f1))); + do { + if (d == '\n') + break; + ctnew++; + } while (isspace(d = getc(f2))); + } else if (cmd_flags & FLAG_w) { + while (isspace(c) && c != '\n') { + c = getc(f1); + ctold++; + } + while (isspace(d) && d != '\n') { + d = getc(f2); + ctnew++; + } + } + if (c != d) { + jackpot++; + J[i] = 0; + if (c != '\n' && c != EOF) + ctold += skipline(f1); + if (d != '\n' && c != EOF) + ctnew += skipline(f2); + break; + } + if (c == '\n' || c == EOF) + break; + } + } else { + while (1) { + ctold++; + ctnew++; + if ((c = getc(f1)) != (d = getc(f2))) { + J[i] = 0; + if (c != '\n' && c != EOF) + ctold += skipline(f1); + if (d != '\n' && c != EOF) + ctnew += skipline(f2); + break; + } + if (c == '\n' || c == EOF) + break; + } + } + ixold[i] = ctold; + ixnew[j] = ctnew; + j++; + } + for (; j <= len[1]; j++) + ixnew[j] = ctnew += skipline(f2); +} + +/* shellsort CACM #201 */ +static void sort(struct line *a, int n) +{ + struct line *ai, *aim, w; + int j, m = 0, k; + + if (n == 0) + return; + for (j = 1; j <= n; j *= 2) + m = 2 * j - 1; + for (m /= 2; m != 0; m /= 2) { + k = n - m; + for (j = 1; j <= k; j++) { + for (ai = &a[j]; ai > a; ai -= m) { + aim = &ai[m]; + if (aim < ai) + break; /* wraparound */ + if (aim->value > ai[0].value || + (aim->value == ai[0].value && + aim->serial > ai[0].serial)) + break; + w.value = ai[0].value; + ai[0].value = aim->value; + aim->value = w.value; + w.serial = ai[0].serial; + ai[0].serial = aim->serial; + aim->serial = w.serial; + } + } + } +} + + +static void uni_range(int a, int b) +{ + if (a < b) + printf("%d,%d", a, b - a + 1); + else if (a == b) + printf("%d", b); + else + printf("%d,0", b); +} + +static int fetch(long *f, int a, int b, FILE *lb, int ch, int oldfile) +{ + int i, j, c, lastc, col, nc; + + if (a > b) + return (0); + for (i = a; i <= b; i++) { + fseek(lb, f[i - 1], SEEK_SET); + nc = f[i] - f[i - 1]; + if (ch != '\0') { + putchar(ch); + if (cmd_flags & FLAG_T) + putchar('\t'); + } + col = 0; + for (j = 0, lastc = '\0'; j < nc; j++, lastc = c) { + if ((c = getc(lb)) == EOF) { + puts("\n\\ No newline at end of file"); + return (0); + } + if (c == '\t' && (cmd_flags & FLAG_t)) { + do { + putchar(' '); + } while (++col & 7); + } else { + putchar(c); + col++; + } + } + } + return (0); +} + +static int asciifile(FILE *f) +{ + + if ((cmd_flags & FLAG_a) || f == NULL) + return (1); +#ifdef CONFIG_FEATURE_DIFF_BINARY + unsigned char buf[BUFSIZ]; + int i, cnt; + + rewind(f); + cnt = fread(buf, 1, sizeof(buf), f); + for (i = 0; i < cnt; i++) + if (!isprint(buf[i]) && !isspace(buf[i])) + return (0); +#endif + return (1); +} + +/* dump accumulated "unified" diff changes */ +static void dump_unified_vec(FILE *f1, FILE *f2) +{ + struct context_vec *cvp = context_vec_start; + int lowa, upb, lowc, upd; + int a, b, c, d; + char ch; + + if (context_vec_start > context_vec_ptr) + return; + + b = d = 0; /* gcc */ + lowa = MAX(1, cvp->a - context); + upb = MIN(len[0], context_vec_ptr->b + context); + lowc = MAX(1, cvp->c - context); + upd = MIN(len[1], context_vec_ptr->d + context); + + fputs("@@ -", stdout); + uni_range(lowa, upb); + fputs(" +", stdout); + uni_range(lowc, upd); + fputs(" @@", stdout); + putchar('\n'); + + /* + * Output changes in "unified" diff format--the old and new lines + * are printed together. + */ + for (; cvp <= context_vec_ptr; cvp++) { + a = cvp->a; + b = cvp->b; + c = cvp->c; + d = cvp->d; + + /* + * c: both new and old changes + * d: only changes in the old file + * a: only changes in the new file + */ + if (a <= b && c <= d) + ch = 'c'; + else + ch = (a <= b) ? 'd' : 'a'; + + switch (ch) { + case 'c': + fetch(ixold, lowa, a - 1, f1, ' ', 0); + fetch(ixold, a, b, f1, '-', 0); + fetch(ixnew, c, d, f2, '+', 0); + break; + case 'd': + fetch(ixold, lowa, a - 1, f1, ' ', 0); + fetch(ixold, a, b, f1, '-', 0); + break; + case 'a': + fetch(ixnew, lowc, c - 1, f2, ' ', 0); + fetch(ixnew, c, d, f2, '+', 0); + break; + } + lowa = b + 1; + lowc = d + 1; + } + fetch(ixnew, d + 1, upd, f2, ' ', 0); + + context_vec_ptr = context_vec_start - 1; +} + + +static void print_header(const char *file1, const char *file2) +{ + if (label[0] != NULL) + printf("%s %s\n", "---", + label[0]); + else + printf("%s %s\t%s", "---", + file1, ctime(&stb1.st_mtime)); + if (label[1] != NULL) + printf("%s %s\n", "+++", + label[1]); + else + printf("%s %s\t%s", "+++", + file2, ctime(&stb2.st_mtime)); +} + + + +/* + * Indicate that there is a difference between lines a and b of the from file + * to get to lines c to d of the to file. If a is greater then b then there + * are no lines in the from file involved and this means that there were + * lines appended (beginning at b). If c is greater than d then there are + * lines missing from the to file. + */ +static void change(char *file1, FILE *f1, char *file2, FILE *f2, int a, int b, int c, int d) +{ + static size_t max_context = 64; + + if (a > b && c > d) return; + if (cmd_flags & FLAG_q) return; + + /* + * Allocate change records as needed. + */ + if (context_vec_ptr == context_vec_end - 1) { + ptrdiff_t offset = context_vec_ptr - context_vec_start; + max_context <<= 1; + context_vec_start = xrealloc(context_vec_start, + max_context * sizeof(struct context_vec)); + context_vec_end = context_vec_start + max_context; + context_vec_ptr = context_vec_start + offset; + } + if (anychange == 0) { + /* + * Print the context/unidiff header first time through. + */ + print_header(file1, file2); + anychange = 1; + } else if (a > context_vec_ptr->b + (2 * context) + 1 && + c > context_vec_ptr->d + (2 * context) + 1) { + /* + * If this change is more than 'context' lines from the + * previous change, dump the record and reset it. + */ + dump_unified_vec(f1, f2); + } + context_vec_ptr++; + context_vec_ptr->a = a; + context_vec_ptr->b = b; + context_vec_ptr->c = c; + context_vec_ptr->d = d; + return; + +} + + +static void output(char *file1, FILE *f1, char *file2, FILE *f2) +{ + int m, i0, i1, j0, j1; + + rewind(f1); + rewind(f2); + m = len[0]; + J[0] = 0; + J[m + 1] = len[1] + 1; + for (i0 = 1; i0 <= m; i0 = i1 + 1) { + while (i0 <= m && J[i0] == J[i0 - 1] + 1) + i0++; + j0 = J[i0 - 1] + 1; + i1 = i0 - 1; + while (i1 < m && J[i1 + 1] == 0) + i1++; + j1 = J[i1 + 1] - 1; + J[i1] = j1; + change(file1, f1, file2, f2, i0, i1, j0, j1); + } + if (m == 0) { + change(file1, f1, file2, f2, 1, 0, 1, len[1]); + } + if (anychange != 0) { + dump_unified_vec(f1, f2); + } +} + /* * The following code uses an algorithm due to Harold Stone, * which finds a pair of longest identical subsequences in @@ -554,666 +1014,176 @@ closem: return (rval); } -/* - * Check to see if the given files differ. - * Returns 0 if they are the same, 1 if different, and -1 on error. - */ -static int -files_differ(FILE *f1, FILE *f2, int flags) -{ - char buf1[BUFSIZ], buf2[BUFSIZ]; - size_t i, j; - - if ((flags & (D_EMPTY1|D_EMPTY2)) || stb1.st_size != stb2.st_size || - (stb1.st_mode & S_IFMT) != (stb2.st_mode & S_IFMT)) - return (1); - while(1) { - i = fread(buf1, 1, sizeof(buf1), f1); - j = fread(buf2, 1, sizeof(buf2), f2); - if (i != j) - return (1); - if (i == 0 && j == 0) { - if (ferror(f1) || ferror(f2)) - return (1); - return (0); - } - if (memcmp(buf1, buf2, i) != 0) - return (1); - } -} - -static void -prepare(int i, FILE *fd, off_t filesize) -{ - struct line *p; - int j, h; - size_t sz; - - rewind(fd); - - sz = (filesize <= FSIZE_MAX ? filesize : FSIZE_MAX) / 25; - if (sz < 100) - sz = 100; - - p = xmalloc((sz + 3) * sizeof(struct line)); - for (j = 0; (h = readhash(fd));) { - if (j == sz) { - sz = sz * 3 / 2; - p = xrealloc(p, (sz + 3) * sizeof(struct line)); - } - p[++j].value = h; - } - len[i] = j; - file[i] = p; -} - -static void -prune(void) -{ - int i, j; - - for (pref = 0; pref < len[0] && pref < len[1] && - file[0][pref + 1].value == file[1][pref + 1].value; - pref++) - ; - for (suff = 0; suff < len[0] - pref && suff < len[1] - pref && - file[0][len[0] - suff].value == file[1][len[1] - suff].value; - suff++) - ; - for (j = 0; j < 2; j++) { - sfile[j] = file[j] + pref; - slen[j] = len[j] - pref - suff; - for (i = 0; i <= slen[j]; i++) - sfile[j][i].serial = i; - } -} - -static void -equiv(struct line *a, int n, struct line *b, int m, int *c) -{ - int i, j; - - i = j = 1; - while (i <= n && j <= m) { - if (a[i].value < b[j].value) - a[i++].value = 0; - else if (a[i].value == b[j].value) - a[i++].value = j; - else - j++; - } - while (i <= n) - a[i++].value = 0; - b[m + 1].value = 0; - j = 0; - while (++j <= m) { - c[j] = -b[j].serial; - while (b[j + 1].value == b[j].value) { - j++; - c[j] = b[j].serial; - } - } - c[j] = -1; -} - -static int isqrt(int n) { - int y, x = 1; - if (n == 0) return(0); +#if ENABLE_FEATURE_DIFF_DIR +static void do_diff (char *dir1, char *path1, char *dir2, char *path2) { - do { - y = x; - x = n / x; - x += y; - x /= 2; - } while ((x - y) > 1 || (x - y) < -1); + int flags = D_HEADER; + int val; + + char *fullpath1 = bb_xasprintf("%s/%s", dir1, path1); + char *fullpath2 = bb_xasprintf("%s/%s", dir2, path2); - return (x); -} - -static int -stone(int *a, int n, int *b, int *c) -{ - int i, k, y, j, l; - int oldc, tc, oldl; - u_int numtries; -#ifdef CONFIG_FEATURE_DIFF_MINIMAL - const u_int bound = (cmd_flags & FLAG_d) ? UINT_MAX : MAX(256, isqrt(n)); -#else - const u_int bound = MAX(256, isqrt(n)); -#endif - k = 0; - c[0] = newcand(0, 0, 0); - for (i = 1; i <= n; i++) { - j = a[i]; - if (j == 0) - continue; - y = -b[j]; - oldl = 0; - oldc = c[0]; - numtries = 0; - do { - if (y <= clist[oldc].y) - continue; - l = search(c, k, y); - if (l != oldl + 1) - oldc = c[l - 1]; - if (l <= k) { - if (clist[c[l]].y <= y) - continue; - tc = c[l]; - c[l] = newcand(i, y, oldc); - oldc = tc; - oldl = l; - numtries++; - } else { - c[l] = newcand(i, y, oldc); - k++; - break; - } - } while ((y = b[++j]) > 0 && numtries < bound); - } - return (k); -} - -static int -newcand(int x, int y, int pred) -{ - struct cand *q; - - if (clen == clistlen) { - clistlen = clistlen * 11 / 10; - clist = xrealloc(clist, clistlen * sizeof(struct cand)); - } - q = clist + clen; - q->x = x; - q->y = y; - q->pred = pred; - return (clen++); -} - -static int -search(int *c, int k, int y) -{ - int i, j, l, t; - - if (clist[c[k]].y < y) /* quick look for typical case */ - return (k + 1); - i = 0; - j = k + 1; - while (1) { - l = i + j; - if ((l >>= 1) <= i) - break; - t = clist[c[l]].y; - if (t > y) - j = l; - else if (t < y) - i = l; - else - return (l); - } - return (l + 1); -} - -static void -unravel(int p) -{ - struct cand *q; - int i; - - for (i = 0; i <= len[0]; i++) - J[i] = i <= pref ? i : - i > len[0] - suff ? i + len[1] - len[0] : 0; - for (q = clist + p; q->y != 0; q = clist + q->pred) - J[q->x + pref] = q->y + pref; -} - -/* - * Check does double duty: - * 1. ferret out any fortuitous correspondences due - * to confounding by hashing (which result in "jackpot") - * 2. collect random access indexes to the two files - */ -static void -check(char *file1, FILE *f1, char *file2, FILE *f2) -{ - int i, j, jackpot, c, d; - long ctold, ctnew; - - rewind(f1); - rewind(f2); - j = 1; - ixold[0] = ixnew[0] = 0; - jackpot = 0; - ctold = ctnew = 0; - for (i = 1; i <= len[0]; i++) { - if (J[i] == 0) { - ixold[i] = ctold += skipline(f1); - continue; - } - while (j < J[i]) { - ixnew[j] = ctnew += skipline(f2); - j++; - } - if ((cmd_flags & FLAG_b) || (cmd_flags & FLAG_w) || (cmd_flags & FLAG_i)) { - while (1) { - c = getc(f1); - d = getc(f2); - /* - * GNU diff ignores a missing newline - * in one file if bflag || wflag. - */ - if (((cmd_flags & FLAG_b) || (cmd_flags & FLAG_w)) && - ((c == EOF && d == '\n') || - (c == '\n' && d == EOF))) { - break; - } - ctold++; - ctnew++; - if ((cmd_flags & FLAG_b) && isspace(c) && isspace(d)) { - do { - if (c == '\n') - break; - ctold++; - } while (isspace(c = getc(f1))); - do { - if (d == '\n') - break; - ctnew++; - } while (isspace(d = getc(f2))); - } else if (cmd_flags & FLAG_w) { - while (isspace(c) && c != '\n') { - c = getc(f1); - ctold++; - } - while (isspace(d) && d != '\n') { - d = getc(f2); - ctnew++; - } - } - if (c != d) { - jackpot++; - J[i] = 0; - if (c != '\n' && c != EOF) - ctold += skipline(f1); - if (d != '\n' && c != EOF) - ctnew += skipline(f2); - break; - } - if (c == '\n' || c == EOF) - break; - } - } else { - while (1) { - ctold++; - ctnew++; - if ((c = getc(f1)) != (d = getc(f2))) { - J[i] = 0; - if (c != '\n' && c != EOF) - ctold += skipline(f1); - if (d != '\n' && c != EOF) - ctnew += skipline(f2); - break; - } - if (c == '\n' || c == EOF) - break; - } - } - ixold[i] = ctold; - ixnew[j] = ctnew; - j++; - } - for (; j <= len[1]; j++) - ixnew[j] = ctnew += skipline(f2); -} - -/* shellsort CACM #201 */ -static void -sort(struct line *a, int n) -{ - struct line *ai, *aim, w; - int j, m = 0, k; - - if (n == 0) - return; - for (j = 1; j <= n; j *= 2) - m = 2 * j - 1; - for (m /= 2; m != 0; m /= 2) { - k = n - m; - for (j = 1; j <= k; j++) { - for (ai = &a[j]; ai > a; ai -= m) { - aim = &ai[m]; - if (aim < ai) - break; /* wraparound */ - if (aim->value > ai[0].value || - (aim->value == ai[0].value && - aim->serial > ai[0].serial)) - break; - w.value = ai[0].value; - ai[0].value = aim->value; - aim->value = w.value; - w.serial = ai[0].serial; - ai[0].serial = aim->serial; - aim->serial = w.serial; - } - } - } -} - -static void -unsort(struct line *f, int l, int *b) -{ - int *a, i; - - a = xmalloc((l + 1) * sizeof(int)); - for (i = 1; i <= l; i++) - a[f[i].serial] = f[i].value; - for (i = 1; i <= l; i++) - b[i] = a[i]; - free(a); -} - -static int -skipline(FILE *f) -{ - int i, c; - - for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++) - continue; - return (i); -} - -static void -output(char *file1, FILE *f1, char *file2, FILE *f2) -{ - int m, i0, i1, j0, j1; - - rewind(f1); - rewind(f2); - m = len[0]; - J[0] = 0; - J[m + 1] = len[1] + 1; - for (i0 = 1; i0 <= m; i0 = i1 + 1) { - while (i0 <= m && J[i0] == J[i0 - 1] + 1) - i0++; - j0 = J[i0 - 1] + 1; - i1 = i0 - 1; - while (i1 < m && J[i1 + 1] == 0) - i1++; - j1 = J[i1 + 1] - 1; - J[i1] = j1; - change(file1, f1, file2, f2, i0, i1, j0, j1); - } - if (m == 0) { - change(file1, f1, file2, f2, 1, 0, 1, len[1]); + if (stat(fullpath1, &stb1) != 0) { + flags |= D_EMPTY1; + memset(&stb1, 0, sizeof(stb1)); + fullpath1 = bb_xasprintf("%s/%s", dir1, path2); + } + if (stat(fullpath2, &stb2) != 0) { + flags |= D_EMPTY2; + memset(&stb2, 0, sizeof(stb2)); + stb2.st_mode = stb1.st_mode; + fullpath2 = bb_xasprintf("%s/%s", dir2, path1); } - if (anychange != 0) { - dump_unified_vec(f1, f2); - } -} -static void uni_range(int a, int b) -{ - if (a < b) - printf("%d,%d", a, b - a + 1); - else if (a == b) - printf("%d", b); - else - printf("%d,0", b); -} - -/* - * Indicate that there is a difference between lines a and b of the from file - * to get to lines c to d of the to file. If a is greater then b then there - * are no lines in the from file involved and this means that there were - * lines appended (beginning at b). If c is greater than d then there are - * lines missing from the to file. - */ -static void -change(char *file1, FILE *f1, char *file2, FILE *f2, int a, int b, int c, int d) -{ - static size_t max_context = 64; - - if (a > b && c > d) return; - if (cmd_flags & FLAG_q) return; + if (stb1.st_mode == 0) + stb1.st_mode = stb2.st_mode; - /* - * Allocate change records as needed. - */ - if (context_vec_ptr == context_vec_end - 1) { - ptrdiff_t offset = context_vec_ptr - context_vec_start; - max_context <<= 1; - context_vec_start = xrealloc(context_vec_start, - max_context * sizeof(struct context_vec)); - context_vec_end = context_vec_start + max_context; - context_vec_ptr = context_vec_start + offset; - } - if (anychange == 0) { - /* - * Print the context/unidiff header first time through. - */ - print_header(file1, file2); - anychange = 1; - } else if (a > context_vec_ptr->b + (2 * context) + 1 && - c > context_vec_ptr->d + (2 * context) + 1) { - /* - * If this change is more than 'context' lines from the - * previous change, dump the record and reset it. - */ - dump_unified_vec(f1, f2); - } - context_vec_ptr++; - context_vec_ptr->a = a; - context_vec_ptr->b = b; - context_vec_ptr->c = c; - context_vec_ptr->d = d; - return; - -} + if (S_ISDIR(stb1.st_mode) && S_ISDIR(stb2.st_mode)) { + printf("Common subdirectories: %s and %s\n", fullpath1, fullpath2); + return; + } -static int -fetch(long *f, int a, int b, FILE *lb, int ch, int oldfile) -{ - int i, j, c, lastc, col, nc; - - if (a > b) - return (0); - for (i = a; i <= b; i++) { - fseek(lb, f[i - 1], SEEK_SET); - nc = f[i] - f[i - 1]; - if (ch != '\0') { - putchar(ch); - if (cmd_flags & FLAG_T) - putchar('\t'); - } - col = 0; - for (j = 0, lastc = '\0'; j < nc; j++, lastc = c) { - if ((c = getc(lb)) == EOF) { - puts("\n\\ No newline at end of file"); - return (0); - } - if (c == '\t' && (cmd_flags & FLAG_t)) { - do { - putchar(' '); - } while (++col & 7); - } else { - putchar(c); - col++; - } - } - } - return (0); -} - -/* - * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578. - */ -static int -readhash(FILE *f) -{ - int i, t, space; - int sum; - - sum = 1; - space = 0; - if (!(cmd_flags & FLAG_b) && !(cmd_flags & FLAG_w)) { - if (FLAG_i) - for (i = 0; (t = getc(f)) != '\n'; i++) { - if (t == EOF) { - if (i == 0) - return (0); - break; - } - sum = sum * 127 + t; - } - else - for (i = 0; (t = getc(f)) != '\n'; i++) { - if (t == EOF) { - if (i == 0) - return (0); - break; - } - sum = sum * 127 + t; - } - } else { - for (i = 0;;) { - switch (t = getc(f)) { - case '\t': - case '\r': - case '\v': - case '\f': - case ' ': - space++; - continue; - default: - if (space && !(cmd_flags & FLAG_w)) { - i++; - space = 0; - } - sum = sum * 127 + t; - i++; - continue; - case EOF: - if (i == 0) - return (0); - /* FALLTHROUGH */ - case '\n': - break; - } - break; - } - } - /* - * There is a remote possibility that we end up with a zero sum. - * Zero is used as an EOF marker, so return 1 instead. - */ - return (sum == 0 ? 1 : sum); -} - -static int -asciifile(FILE *f) -{ - - if ((cmd_flags & FLAG_a) || f == NULL) - return (1); -#ifdef CONFIG_FEATURE_DIFF_BINARY - unsigned char buf[BUFSIZ]; - int i, cnt; + if (!S_ISREG(stb1.st_mode) && !S_ISDIR(stb1.st_mode)) + val = D_SKIPPED1; + else if (!S_ISREG(stb2.st_mode) && !S_ISDIR(stb2.st_mode)) + val = D_SKIPPED2; + else + val = diffreg(fullpath1, fullpath2, flags); - rewind(f); - cnt = fread(buf, 1, sizeof(buf), f); - for (i = 0; i < cnt; i++) - if (!isprint(buf[i]) && !isspace(buf[i])) - return (0); + print_status(val, fullpath1, fullpath2, NULL); +} #endif - return (1); + +#ifdef CONFIG_FEATURE_DIFF_DIR +static int dir_strcmp(const void *p1, const void *p2) { + return strcmp(*(char * const *)p1, *(char * const *)p2); } -/* dump accumulated "unified" diff changes */ -static void -dump_unified_vec(FILE *f1, FILE *f2) -{ - struct context_vec *cvp = context_vec_start; - int lowa, upb, lowc, upd; - int a, b, c, d; - char ch; +/* This function adds a filename to dl, the directory listing. */ - if (context_vec_start > context_vec_ptr) - return; - - b = d = 0; /* gcc */ - lowa = MAX(1, cvp->a - context); - upb = MIN(len[0], context_vec_ptr->b + context); - lowc = MAX(1, cvp->c - context); - upd = MIN(len[1], context_vec_ptr->d + context); - - fputs("@@ -", stdout); - uni_range(lowa, upb); - fputs(" +", stdout); - uni_range(lowc, upd); - fputs(" @@", stdout); - putchar('\n'); - - /* - * Output changes in "unified" diff format--the old and new lines - * are printed together. - */ - for (; cvp <= context_vec_ptr; cvp++) { - a = cvp->a; - b = cvp->b; - c = cvp->c; - d = cvp->d; - - /* - * c: both new and old changes - * d: only changes in the old file - * a: only changes in the new file - */ - if (a <= b && c <= d) - ch = 'c'; - else - ch = (a <= b) ? 'd' : 'a'; - - switch (ch) { - case 'c': - fetch(ixold, lowa, a - 1, f1, ' ', 0); - fetch(ixold, a, b, f1, '-', 0); - fetch(ixnew, c, d, f2, '+', 0); - break; - case 'd': - fetch(ixold, lowa, a - 1, f1, ' ', 0); - fetch(ixold, a, b, f1, '-', 0); - break; - case 'a': - fetch(ixnew, lowc, c - 1, f2, ' ', 0); - fetch(ixnew, c, d, f2, '+', 0); - break; - } - lowa = b + 1; - lowc = d + 1; - } - fetch(ixnew, d + 1, upd, f2, ' ', 0); - - context_vec_ptr = context_vec_start - 1; +static int add_to_dirlist (const char *filename, struct stat *statbuf, void *userdata) { + dl_count++; + dl = xrealloc(dl, dl_count * sizeof(char *)); + dl[dl_count - 1] = bb_xstrdup(filename); + if (cmd_flags & FLAG_r) { + int *pp = (int *) userdata; + int path_len = *pp + 1; + dl[dl_count - 1] = &(dl[dl_count - 1])[path_len]; + } + return TRUE; } -static void -print_header(const char *file1, const char *file2) -{ - if (label[0] != NULL) - printf("%s %s\n", "---", - label[0]); - else - printf("%s %s\t%s", "---", - file1, ctime(&stb1.st_mtime)); - if (label[1] != NULL) - printf("%s %s\n", "+++", - label[1]); - else - printf("%s %s\t%s", "+++", - file2, ctime(&stb2.st_mtime)); +/* This returns a sorted directory listing. */ +static char **get_dir(char *path) { + + int i; + + /* Reset dl_count - there's no need to free dl as bb_xrealloc does + * the job nicely. */ + dl_count = 0; + + /* If -r has been set, then the recursive_action function will be + * used. Unfortunately, this outputs the root directory along with + * the recursed paths, so use void *userdata to specify the string + * length of the root directory. It can then be removed in + * add_to_dirlist. */ + + int path_len = strlen(path); + void *userdata = &path_len; + + /* Now fill dl with a listing. */ + if (cmd_flags & FLAG_r) + recursive_action(path, TRUE, TRUE, FALSE, add_to_dirlist, NULL, userdata); + else { + DIR *dp; + struct dirent *ep; + if ((dp = opendir(path)) == NULL) + bb_error_msg("Error reading directory"); + while ((ep = readdir(dp))) { + if ((!strcmp(ep->d_name, "..")) || (!strcmp(ep->d_name, "."))) + continue; + add_to_dirlist(ep->d_name, NULL, NULL); + } + closedir(dp); + } + + /* Sort dl alphabetically. */ + qsort(dl, dl_count, sizeof(char *), dir_strcmp); + + /* Copy dl so that we can return it. */ + char **retval = xmalloc(dl_count * sizeof(char *)); + for (i = 0; i < dl_count; i++) + retval[i] = bb_xstrdup(dl[i]); + + return retval; } +static void diffdir (char *p1, char *p2) { + + char **dirlist1, **dirlist2; + char *dp1, *dp2; + int dirlist1_count, dirlist2_count; + int pos; + + /* Check for trailing slashes. */ + + if (p1[strlen(p1) - 1] == '/') + p1[strlen(p1) - 1] = '\0'; + if (p2[strlen(p2) - 1] == '/') + p2[strlen(p2) - 1] = '\0'; + + /* Get directory listings for p1 and p2. */ + + dirlist1 = get_dir(p1); + dirlist1_count = dl_count; + dirlist1[dirlist1_count] = NULL; + dirlist2 = get_dir(p2); + dirlist2_count = dl_count; + dirlist2[dirlist2_count] = NULL; + + /* If -S was set, find the starting point. */ + if (start) { + while (*dirlist1 != NULL && strcmp(*dirlist1, start) < 0) + dirlist1++; + while (*dirlist2 != NULL && strcmp(*dirlist2, start) < 0) + dirlist2++; + if ((*dirlist1 == NULL) || (*dirlist2 == NULL)) + bb_error_msg("Invalid argument to -S"); + } + + /* Now that both dirlist1 and dirlist2 contain sorted directory + * listings, we can start to go through dirlist1. If both listings + * contain the same file, then do a normal diff. Otherwise, behaviour + * is determined by whether the -N flag is set. */ + while (*dirlist1 != NULL || *dirlist2 != NULL) { + dp1 = *dirlist1; + dp2 = *dirlist2; + pos = dp1 == NULL ? 1 : dp2 == NULL ? -1 : strcmp(dp1, dp2); + if (pos == 0) { + do_diff(p1, dp1, p2, dp2); + dirlist1++; + dirlist2++; + } + else if (pos < 0) { + if (cmd_flags & FLAG_N) + do_diff(p1, dp1, p2, NULL); + else + print_only(p1, strlen(p1) + 1, dp1); + dirlist1++; + } + else { + if (cmd_flags & FLAG_N) + do_diff(p1, NULL, p2, dp2); + else + print_only(p2, strlen(p2) + 1, dp2); + dirlist2++; + } + } +} +#endif + + + extern int diff_main(int argc, char **argv) { char *ep; int gotstdin = 0;