Annotation of /tags/mkinitrd-6_1_5/busybox/editors/diff.c
Parent Directory | Revision Log
Revision 899 -
(hide annotations)
(download)
Wed Aug 5 17:52:52 2009 UTC (14 years, 10 months ago) by niro
File MIME type: text/plain
File size: 32078 byte(s)
Wed Aug 5 17:52:52 2009 UTC (14 years, 10 months ago) by niro
File MIME type: text/plain
File size: 32078 byte(s)
tagged 'mkinitrd-6_1_5'
1 | niro | 816 | /* vi: set sw=4 ts=4: */ |
2 | /* | ||
3 | * Mini diff implementation for busybox, adapted from OpenBSD diff. | ||
4 | * | ||
5 | * Copyright (C) 2006 by Robert Sullivan <cogito.ergo.cogito@hotmail.com> | ||
6 | * Copyright (c) 2003 Todd C. Miller <Todd.Miller@courtesan.com> | ||
7 | * | ||
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. | ||
11 | * | ||
12 | * Licensed under GPLv2 or later, see file LICENSE in this tarball for details. | ||
13 | */ | ||
14 | |||
15 | #include "libbb.h" | ||
16 | |||
17 | // #define FSIZE_MAX 32768 | ||
18 | |||
19 | /* NOINLINEs added to prevent gcc from merging too much into diffreg() | ||
20 | * (it bites more than it can (efficiently) chew). */ | ||
21 | |||
22 | /* | ||
23 | * Output flags | ||
24 | */ | ||
25 | enum { | ||
26 | /* Print a header/footer between files */ | ||
27 | /* D_HEADER = 1, - unused */ | ||
28 | /* Treat file as empty (/dev/null) */ | ||
29 | D_EMPTY1 = 2 * ENABLE_FEATURE_DIFF_DIR, | ||
30 | D_EMPTY2 = 4 * ENABLE_FEATURE_DIFF_DIR, | ||
31 | }; | ||
32 | |||
33 | /* | ||
34 | * Status values for print_status() and diffreg() return values | ||
35 | * Guide: | ||
36 | * D_SAME - files are the same | ||
37 | * D_DIFFER - files differ | ||
38 | * D_BINARY - binary files differ | ||
39 | * D_COMMON - subdirectory common to both dirs | ||
40 | * D_ONLY - file only exists in one dir | ||
41 | * D_ISDIR1 - path1 a dir, path2 a file | ||
42 | * D_ISDIR2 - path1 a file, path2 a dir | ||
43 | * D_ERROR - error occurred | ||
44 | * D_SKIPPED1 - skipped path1 as it is a special file | ||
45 | * D_SKIPPED2 - skipped path2 as it is a special file | ||
46 | */ | ||
47 | #define D_SAME 0 | ||
48 | #define D_DIFFER (1 << 0) | ||
49 | #define D_BINARY (1 << 1) | ||
50 | #define D_COMMON (1 << 2) | ||
51 | /*#define D_ONLY (1 << 3) - unused */ | ||
52 | #define D_ISDIR1 (1 << 4) | ||
53 | #define D_ISDIR2 (1 << 5) | ||
54 | #define D_ERROR (1 << 6) | ||
55 | #define D_SKIPPED1 (1 << 7) | ||
56 | #define D_SKIPPED2 (1 << 8) | ||
57 | |||
58 | /* Command line options */ | ||
59 | #define FLAG_a (1 << 0) | ||
60 | #define FLAG_b (1 << 1) | ||
61 | #define FLAG_d (1 << 2) | ||
62 | #define FLAG_i (1 << 3) | ||
63 | #define FLAG_L (1 << 4) | ||
64 | #define FLAG_N (1 << 5) | ||
65 | #define FLAG_q (1 << 6) | ||
66 | #define FLAG_r (1 << 7) | ||
67 | #define FLAG_s (1 << 8) | ||
68 | #define FLAG_S (1 << 9) | ||
69 | #define FLAG_t (1 << 10) | ||
70 | #define FLAG_T (1 << 11) | ||
71 | #define FLAG_U (1 << 12) | ||
72 | #define FLAG_w (1 << 13) | ||
73 | |||
74 | |||
75 | struct cand { | ||
76 | int x; | ||
77 | int y; | ||
78 | int pred; | ||
79 | }; | ||
80 | |||
81 | struct line { | ||
82 | int serial; | ||
83 | int value; | ||
84 | }; | ||
85 | |||
86 | /* | ||
87 | * The following struct is used to record change information | ||
88 | * doing a "context" or "unified" diff. (see routine "change" to | ||
89 | * understand the highly mnemonic field names) | ||
90 | */ | ||
91 | struct context_vec { | ||
92 | int a; /* start line in old file */ | ||
93 | int b; /* end line in old file */ | ||
94 | int c; /* start line in new file */ | ||
95 | int d; /* end line in new file */ | ||
96 | }; | ||
97 | |||
98 | |||
99 | #define g_read_buf bb_common_bufsiz1 | ||
100 | |||
101 | struct globals { | ||
102 | bool anychange; | ||
103 | smallint exit_status; | ||
104 | int opt_U_context; | ||
105 | size_t max_context; /* size of context_vec_start */ | ||
106 | USE_FEATURE_DIFF_DIR(int dl_count;) | ||
107 | USE_FEATURE_DIFF_DIR(char **dl;) | ||
108 | char *opt_S_start; | ||
109 | const char *label1; | ||
110 | const char *label2; | ||
111 | int *J; /* will be overlaid on class */ | ||
112 | int clen; | ||
113 | int pref, suff; /* length of prefix and suffix */ | ||
114 | int nlen[2]; | ||
115 | int slen[2]; | ||
116 | int clistlen; /* the length of clist */ | ||
117 | struct cand *clist; /* merely a free storage pot for candidates */ | ||
118 | long *ixnew; /* will be overlaid on nfile[1] */ | ||
119 | long *ixold; /* will be overlaid on klist */ | ||
120 | struct line *nfile[2]; | ||
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 | char *tempname1, *tempname2; | ||
126 | struct stat stb1, stb2; | ||
127 | }; | ||
128 | #define G (*ptr_to_globals) | ||
129 | #define anychange (G.anychange ) | ||
130 | #define exit_status (G.exit_status ) | ||
131 | #define opt_U_context (G.opt_U_context ) | ||
132 | #define max_context (G.max_context ) | ||
133 | #define dl_count (G.dl_count ) | ||
134 | #define dl (G.dl ) | ||
135 | #define opt_S_start (G.opt_S_start ) | ||
136 | #define label1 (G.label1 ) | ||
137 | #define label2 (G.label2 ) | ||
138 | #define J (G.J ) | ||
139 | #define clen (G.clen ) | ||
140 | #define pref (G.pref ) | ||
141 | #define suff (G.suff ) | ||
142 | #define nlen (G.nlen ) | ||
143 | #define slen (G.slen ) | ||
144 | #define clistlen (G.clistlen ) | ||
145 | #define clist (G.clist ) | ||
146 | #define ixnew (G.ixnew ) | ||
147 | #define ixold (G.ixold ) | ||
148 | #define nfile (G.nfile ) | ||
149 | #define sfile (G.sfile ) | ||
150 | #define context_vec_start (G.context_vec_start ) | ||
151 | #define context_vec_end (G.context_vec_end ) | ||
152 | #define context_vec_ptr (G.context_vec_ptr ) | ||
153 | #define stb1 (G.stb1 ) | ||
154 | #define stb2 (G.stb2 ) | ||
155 | #define tempname1 (G.tempname1 ) | ||
156 | #define tempname2 (G.tempname2 ) | ||
157 | #define INIT_G() do { \ | ||
158 | SET_PTR_TO_GLOBALS(xzalloc(sizeof(G))); \ | ||
159 | opt_U_context = 3; \ | ||
160 | max_context = 64; \ | ||
161 | } while (0) | ||
162 | |||
163 | |||
164 | #if ENABLE_FEATURE_DIFF_DIR | ||
165 | static void print_only(const char *path, const char *entry) | ||
166 | { | ||
167 | printf("Only in %s: %s\n", path, entry); | ||
168 | } | ||
169 | #endif | ||
170 | |||
171 | |||
172 | static void print_status(int val, char *_path1, char *_path2) | ||
173 | { | ||
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;*/ | ||
177 | |||
178 | switch (val) { | ||
179 | /* case D_ONLY: | ||
180 | print_only(path1, entry); | ||
181 | break; | ||
182 | */ | ||
183 | case D_COMMON: | ||
184 | printf("Common subdirectories: %s and %s\n", _path1, _path2); | ||
185 | break; | ||
186 | case D_BINARY: | ||
187 | printf("Binary files %s and %s differ\n", _path1, _path2); | ||
188 | break; | ||
189 | case D_DIFFER: | ||
190 | if (option_mask32 & FLAG_q) | ||
191 | printf("Files %s and %s differ\n", _path1, _path2); | ||
192 | break; | ||
193 | case D_SAME: | ||
194 | if (option_mask32 & FLAG_s) | ||
195 | printf("Files %s and %s are identical\n", _path1, _path2); | ||
196 | break; | ||
197 | case D_ISDIR1: | ||
198 | printf("File %s is a %s while file %s is a %s\n", | ||
199 | _path1, "directory", _path2, "regular file"); | ||
200 | break; | ||
201 | case D_ISDIR2: | ||
202 | printf("File %s is a %s while file %s is a %s\n", | ||
203 | _path1, "regular file", _path2, "directory"); | ||
204 | break; | ||
205 | case D_SKIPPED1: | ||
206 | printf("File %s is not a regular file or directory and was skipped\n", | ||
207 | _path1); | ||
208 | break; | ||
209 | case D_SKIPPED2: | ||
210 | printf("File %s is not a regular file or directory and was skipped\n", | ||
211 | _path2); | ||
212 | break; | ||
213 | } | ||
214 | /* | ||
215 | if (entry) { | ||
216 | free(_path1); | ||
217 | free(_path2); | ||
218 | } | ||
219 | */ | ||
220 | } | ||
221 | |||
222 | |||
223 | /* Read line, return its nonzero hash. Return 0 if EOF. | ||
224 | * | ||
225 | * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578. | ||
226 | */ | ||
227 | static ALWAYS_INLINE int fiddle_sum(int sum, int t) | ||
228 | { | ||
229 | return sum * 127 + t; | ||
230 | } | ||
231 | static int readhash(FILE *fp) | ||
232 | { | ||
233 | int i, t, space; | ||
234 | int sum; | ||
235 | |||
236 | sum = 1; | ||
237 | space = 0; | ||
238 | i = 0; | ||
239 | if (!(option_mask32 & (FLAG_b | FLAG_w))) { | ||
240 | while ((t = getc(fp)) != '\n') { | ||
241 | if (t == EOF) { | ||
242 | if (i == 0) | ||
243 | return 0; | ||
244 | break; | ||
245 | } | ||
246 | sum = fiddle_sum(sum, t); | ||
247 | i = 1; | ||
248 | } | ||
249 | } else { | ||
250 | while (1) { | ||
251 | switch (t = getc(fp)) { | ||
252 | case '\t': | ||
253 | case '\r': | ||
254 | case '\v': | ||
255 | case '\f': | ||
256 | case ' ': | ||
257 | space = 1; | ||
258 | continue; | ||
259 | default: | ||
260 | if (space && !(option_mask32 & FLAG_w)) { | ||
261 | i = 1; | ||
262 | space = 0; | ||
263 | } | ||
264 | sum = fiddle_sum(sum, t); | ||
265 | i = 1; | ||
266 | continue; | ||
267 | case EOF: | ||
268 | if (i == 0) | ||
269 | return 0; | ||
270 | /* FALLTHROUGH */ | ||
271 | case '\n': | ||
272 | break; | ||
273 | } | ||
274 | break; | ||
275 | } | ||
276 | } | ||
277 | /* | ||
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. | ||
280 | */ | ||
281 | return (sum == 0 ? 1 : sum); | ||
282 | } | ||
283 | |||
284 | |||
285 | /* Our diff implementation is using seek. | ||
286 | * When we meet non-seekable file, we must make a temp copy. | ||
287 | */ | ||
288 | static char *make_temp(FILE *f, struct stat *sb) | ||
289 | { | ||
290 | char *name; | ||
291 | int fd; | ||
292 | |||
293 | if (S_ISREG(sb->st_mode) || S_ISBLK(sb->st_mode)) | ||
294 | return NULL; | ||
295 | name = xstrdup("/tmp/difXXXXXX"); | ||
296 | fd = mkstemp(name); | ||
297 | if (fd < 0) | ||
298 | bb_perror_msg_and_die("mkstemp"); | ||
299 | if (bb_copyfd_eof(fileno(f), fd) < 0) { | ||
300 | clean_up: | ||
301 | unlink(name); | ||
302 | xfunc_die(); /* error message is printed by bb_copyfd_eof */ | ||
303 | } | ||
304 | fstat(fd, sb); | ||
305 | close(fd); | ||
306 | if (freopen(name, "r+", f) == NULL) { | ||
307 | bb_perror_msg("freopen"); | ||
308 | goto clean_up; | ||
309 | } | ||
310 | return name; | ||
311 | } | ||
312 | |||
313 | |||
314 | /* | ||
315 | * Check to see if the given files differ. | ||
316 | * Returns 0 if they are the same, 1 if different, and -1 on error. | ||
317 | */ | ||
318 | static NOINLINE int files_differ(FILE *f1, FILE *f2) | ||
319 | { | ||
320 | size_t i, j; | ||
321 | |||
322 | /* Prevent making copies for "/dev/null" (too common) */ | ||
323 | /* Deal with input from pipes etc */ | ||
324 | tempname1 = make_temp(f1, &stb1); | ||
325 | tempname2 = make_temp(f2, &stb2); | ||
326 | if (stb1.st_size != stb2.st_size) { | ||
327 | return 1; | ||
328 | } | ||
329 | while (1) { | ||
330 | i = fread(g_read_buf, 1, COMMON_BUFSIZE/2, f1); | ||
331 | j = fread(g_read_buf + COMMON_BUFSIZE/2, 1, COMMON_BUFSIZE/2, f2); | ||
332 | if (i != j) | ||
333 | return 1; | ||
334 | if (i == 0) | ||
335 | return (ferror(f1) || ferror(f2)) ? -1 : 0; | ||
336 | if (memcmp(g_read_buf, | ||
337 | g_read_buf + COMMON_BUFSIZE/2, i) != 0) | ||
338 | return 1; | ||
339 | } | ||
340 | } | ||
341 | |||
342 | |||
343 | static void prepare(int i, FILE *fp /*, off_t filesize*/) | ||
344 | { | ||
345 | struct line *p; | ||
346 | int h; | ||
347 | size_t j, sz; | ||
348 | |||
349 | rewind(fp); | ||
350 | |||
351 | /*sz = (filesize <= FSIZE_MAX ? filesize : FSIZE_MAX) / 25;*/ | ||
352 | /*if (sz < 100)*/ | ||
353 | sz = 100; | ||
354 | |||
355 | p = xmalloc((sz + 3) * sizeof(p[0])); | ||
356 | j = 0; | ||
357 | while ((h = readhash(fp)) != 0) { /* while not EOF */ | ||
358 | if (j == sz) { | ||
359 | sz = sz * 3 / 2; | ||
360 | p = xrealloc(p, (sz + 3) * sizeof(p[0])); | ||
361 | } | ||
362 | p[++j].value = h; | ||
363 | } | ||
364 | nlen[i] = j; | ||
365 | nfile[i] = p; | ||
366 | } | ||
367 | |||
368 | |||
369 | static void prune(void) | ||
370 | { | ||
371 | int i, j; | ||
372 | |||
373 | for (pref = 0; pref < nlen[0] && pref < nlen[1] && | ||
374 | nfile[0][pref + 1].value == nfile[1][pref + 1].value; pref++) | ||
375 | continue; | ||
376 | for (suff = 0; suff < nlen[0] - pref && suff < nlen[1] - pref && | ||
377 | nfile[0][nlen[0] - suff].value == nfile[1][nlen[1] - suff].value; | ||
378 | suff++) | ||
379 | continue; | ||
380 | for (j = 0; j < 2; j++) { | ||
381 | sfile[j] = nfile[j] + pref; | ||
382 | slen[j] = nlen[j] - pref - suff; | ||
383 | for (i = 0; i <= slen[j]; i++) | ||
384 | sfile[j][i].serial = i; | ||
385 | } | ||
386 | } | ||
387 | |||
388 | |||
389 | static void equiv(struct line *a, int n, struct line *b, int m, int *c) | ||
390 | { | ||
391 | int i, j; | ||
392 | |||
393 | i = j = 1; | ||
394 | while (i <= n && j <= m) { | ||
395 | if (a[i].value < b[j].value) | ||
396 | a[i++].value = 0; | ||
397 | else if (a[i].value == b[j].value) | ||
398 | a[i++].value = j; | ||
399 | else | ||
400 | j++; | ||
401 | } | ||
402 | while (i <= n) | ||
403 | a[i++].value = 0; | ||
404 | b[m + 1].value = 0; | ||
405 | j = 0; | ||
406 | while (++j <= m) { | ||
407 | c[j] = -b[j].serial; | ||
408 | while (b[j + 1].value == b[j].value) { | ||
409 | j++; | ||
410 | c[j] = b[j].serial; | ||
411 | } | ||
412 | } | ||
413 | c[j] = -1; | ||
414 | } | ||
415 | |||
416 | |||
417 | static int isqrt(int n) | ||
418 | { | ||
419 | int y, x; | ||
420 | |||
421 | if (n == 0) | ||
422 | return 0; | ||
423 | x = 1; | ||
424 | do { | ||
425 | y = x; | ||
426 | x = n / x; | ||
427 | x += y; | ||
428 | x /= 2; | ||
429 | } while ((x - y) > 1 || (x - y) < -1); | ||
430 | |||
431 | return x; | ||
432 | } | ||
433 | |||
434 | |||
435 | static int newcand(int x, int y, int pred) | ||
436 | { | ||
437 | struct cand *q; | ||
438 | |||
439 | if (clen == clistlen) { | ||
440 | clistlen = clistlen * 11 / 10; | ||
441 | clist = xrealloc(clist, clistlen * sizeof(struct cand)); | ||
442 | } | ||
443 | q = clist + clen; | ||
444 | q->x = x; | ||
445 | q->y = y; | ||
446 | q->pred = pred; | ||
447 | return clen++; | ||
448 | } | ||
449 | |||
450 | |||
451 | static int search(int *c, int k, int y) | ||
452 | { | ||
453 | int i, j, l, t; | ||
454 | |||
455 | if (clist[c[k]].y < y) /* quick look for typical case */ | ||
456 | return k + 1; | ||
457 | i = 0; | ||
458 | j = k + 1; | ||
459 | while (1) { | ||
460 | l = i + j; | ||
461 | if ((l >>= 1) <= i) | ||
462 | break; | ||
463 | t = clist[c[l]].y; | ||
464 | if (t > y) | ||
465 | j = l; | ||
466 | else if (t < y) | ||
467 | i = l; | ||
468 | else | ||
469 | return l; | ||
470 | } | ||
471 | return l + 1; | ||
472 | } | ||
473 | |||
474 | |||
475 | static int stone(int *a, int n, int *b, int *c) | ||
476 | { | ||
477 | int i, k, y, j, l; | ||
478 | int oldc, tc, oldl; | ||
479 | unsigned int numtries; | ||
480 | #if ENABLE_FEATURE_DIFF_MINIMAL | ||
481 | const unsigned int bound = | ||
482 | (option_mask32 & FLAG_d) ? UINT_MAX : MAX(256, isqrt(n)); | ||
483 | #else | ||
484 | const unsigned int bound = MAX(256, isqrt(n)); | ||
485 | #endif | ||
486 | |||
487 | k = 0; | ||
488 | c[0] = newcand(0, 0, 0); | ||
489 | for (i = 1; i <= n; i++) { | ||
490 | j = a[i]; | ||
491 | if (j == 0) | ||
492 | continue; | ||
493 | y = -b[j]; | ||
494 | oldl = 0; | ||
495 | oldc = c[0]; | ||
496 | numtries = 0; | ||
497 | do { | ||
498 | if (y <= clist[oldc].y) | ||
499 | continue; | ||
500 | l = search(c, k, y); | ||
501 | if (l != oldl + 1) | ||
502 | oldc = c[l - 1]; | ||
503 | if (l <= k) { | ||
504 | if (clist[c[l]].y <= y) | ||
505 | continue; | ||
506 | tc = c[l]; | ||
507 | c[l] = newcand(i, y, oldc); | ||
508 | oldc = tc; | ||
509 | oldl = l; | ||
510 | numtries++; | ||
511 | } else { | ||
512 | c[l] = newcand(i, y, oldc); | ||
513 | k++; | ||
514 | break; | ||
515 | } | ||
516 | } while ((y = b[++j]) > 0 && numtries < bound); | ||
517 | } | ||
518 | return k; | ||
519 | } | ||
520 | |||
521 | |||
522 | static void unravel(int p) | ||
523 | { | ||
524 | struct cand *q; | ||
525 | int i; | ||
526 | |||
527 | for (i = 0; i <= nlen[0]; i++) | ||
528 | J[i] = i <= pref ? i : i > nlen[0] - suff ? i + nlen[1] - nlen[0] : 0; | ||
529 | for (q = clist + p; q->y != 0; q = clist + q->pred) | ||
530 | J[q->x + pref] = q->y + pref; | ||
531 | } | ||
532 | |||
533 | |||
534 | static void unsort(struct line *f, int l, int *b) | ||
535 | { | ||
536 | int *a, i; | ||
537 | |||
538 | a = xmalloc((l + 1) * sizeof(int)); | ||
539 | for (i = 1; i <= l; i++) | ||
540 | a[f[i].serial] = f[i].value; | ||
541 | for (i = 1; i <= l; i++) | ||
542 | b[i] = a[i]; | ||
543 | free(a); | ||
544 | } | ||
545 | |||
546 | |||
547 | static int skipline(FILE *f) | ||
548 | { | ||
549 | int i, c; | ||
550 | |||
551 | for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++) | ||
552 | continue; | ||
553 | return i; | ||
554 | } | ||
555 | |||
556 | |||
557 | /* | ||
558 | * Check does double duty: | ||
559 | * 1. ferret out any fortuitous correspondences due | ||
560 | * to confounding by hashing (which result in "jackpot") | ||
561 | * 2. collect random access indexes to the two files | ||
562 | */ | ||
563 | static NOINLINE void check(FILE *f1, FILE *f2) | ||
564 | { | ||
565 | int i, j, jackpot, c, d; | ||
566 | long ctold, ctnew; | ||
567 | |||
568 | rewind(f1); | ||
569 | rewind(f2); | ||
570 | j = 1; | ||
571 | ixold[0] = ixnew[0] = 0; | ||
572 | jackpot = 0; | ||
573 | ctold = ctnew = 0; | ||
574 | for (i = 1; i <= nlen[0]; i++) { | ||
575 | if (J[i] == 0) { | ||
576 | ixold[i] = ctold += skipline(f1); | ||
577 | continue; | ||
578 | } | ||
579 | while (j < J[i]) { | ||
580 | ixnew[j] = ctnew += skipline(f2); | ||
581 | j++; | ||
582 | } | ||
583 | if (option_mask32 & (FLAG_b | FLAG_w | FLAG_i)) { | ||
584 | while (1) { | ||
585 | c = getc(f1); | ||
586 | d = getc(f2); | ||
587 | /* | ||
588 | * GNU diff ignores a missing newline | ||
589 | * in one file if bflag || wflag. | ||
590 | */ | ||
591 | if ((option_mask32 & (FLAG_b | FLAG_w)) | ||
592 | && ((c == EOF && d == '\n') || (c == '\n' && d == EOF)) | ||
593 | ) { | ||
594 | break; | ||
595 | } | ||
596 | ctold++; | ||
597 | ctnew++; | ||
598 | if ((option_mask32 & FLAG_b) && isspace(c) && isspace(d)) { | ||
599 | do { | ||
600 | if (c == '\n') | ||
601 | break; | ||
602 | ctold++; | ||
603 | c = getc(f1); | ||
604 | } while (isspace(c)); | ||
605 | do { | ||
606 | if (d == '\n') | ||
607 | break; | ||
608 | ctnew++; | ||
609 | d = getc(f2); | ||
610 | } while (isspace(d)); | ||
611 | } else if (option_mask32 & FLAG_w) { | ||
612 | while (isspace(c) && c != '\n') { | ||
613 | c = getc(f1); | ||
614 | ctold++; | ||
615 | } | ||
616 | while (isspace(d) && d != '\n') { | ||
617 | d = getc(f2); | ||
618 | ctnew++; | ||
619 | } | ||
620 | } | ||
621 | if (c != d) { | ||
622 | jackpot++; | ||
623 | J[i] = 0; | ||
624 | if (c != '\n' && c != EOF) | ||
625 | ctold += skipline(f1); | ||
626 | if (d != '\n' && c != EOF) | ||
627 | ctnew += skipline(f2); | ||
628 | break; | ||
629 | } | ||
630 | if (c == '\n' || c == EOF) | ||
631 | break; | ||
632 | } | ||
633 | } else { | ||
634 | while (1) { | ||
635 | ctold++; | ||
636 | ctnew++; | ||
637 | c = getc(f1); | ||
638 | d = getc(f2); | ||
639 | if (c != d) { | ||
640 | J[i] = 0; | ||
641 | if (c != '\n' && c != EOF) | ||
642 | ctold += skipline(f1); | ||
643 | /* was buggy? "if (d != '\n' && c != EOF)" */ | ||
644 | if (d != '\n' && d != EOF) | ||
645 | ctnew += skipline(f2); | ||
646 | break; | ||
647 | } | ||
648 | if (c == '\n' || c == EOF) | ||
649 | break; | ||
650 | } | ||
651 | } | ||
652 | ixold[i] = ctold; | ||
653 | ixnew[j] = ctnew; | ||
654 | j++; | ||
655 | } | ||
656 | for (; j <= nlen[1]; j++) | ||
657 | ixnew[j] = ctnew += skipline(f2); | ||
658 | } | ||
659 | |||
660 | |||
661 | /* shellsort CACM #201 */ | ||
662 | static void sort(struct line *a, int n) | ||
663 | { | ||
664 | struct line *ai, *aim, w; | ||
665 | int j, m = 0, k; | ||
666 | |||
667 | if (n == 0) | ||
668 | return; | ||
669 | for (j = 1; j <= n; j *= 2) | ||
670 | m = 2 * j - 1; | ||
671 | for (m /= 2; m != 0; m /= 2) { | ||
672 | k = n - m; | ||
673 | for (j = 1; j <= k; j++) { | ||
674 | for (ai = &a[j]; ai > a; ai -= m) { | ||
675 | aim = &ai[m]; | ||
676 | if (aim < ai) | ||
677 | break; /* wraparound */ | ||
678 | if (aim->value > ai[0].value | ||
679 | || (aim->value == ai[0].value && aim->serial > ai[0].serial) | ||
680 | ) { | ||
681 | break; | ||
682 | } | ||
683 | w.value = ai[0].value; | ||
684 | ai[0].value = aim->value; | ||
685 | aim->value = w.value; | ||
686 | w.serial = ai[0].serial; | ||
687 | ai[0].serial = aim->serial; | ||
688 | aim->serial = w.serial; | ||
689 | } | ||
690 | } | ||
691 | } | ||
692 | } | ||
693 | |||
694 | |||
695 | static void uni_range(int a, int b) | ||
696 | { | ||
697 | if (a < b) | ||
698 | printf("%d,%d", a, b - a + 1); | ||
699 | else if (a == b) | ||
700 | printf("%d", b); | ||
701 | else | ||
702 | printf("%d,0", b); | ||
703 | } | ||
704 | |||
705 | |||
706 | static void fetch(long *f, int a, int b, FILE *lb, int ch) | ||
707 | { | ||
708 | int i, j, c, lastc, col, nc; | ||
709 | |||
710 | if (a > b) | ||
711 | return; | ||
712 | for (i = a; i <= b; i++) { | ||
713 | fseek(lb, f[i - 1], SEEK_SET); | ||
714 | nc = f[i] - f[i - 1]; | ||
715 | if (ch != '\0') { | ||
716 | putchar(ch); | ||
717 | if (option_mask32 & FLAG_T) | ||
718 | putchar('\t'); | ||
719 | } | ||
720 | col = 0; | ||
721 | for (j = 0, lastc = '\0'; j < nc; j++, lastc = c) { | ||
722 | c = getc(lb); | ||
723 | if (c == EOF) { | ||
724 | printf("\n\\ No newline at end of file\n"); | ||
725 | return; | ||
726 | } | ||
727 | if (c == '\t' && (option_mask32 & FLAG_t)) { | ||
728 | do { | ||
729 | putchar(' '); | ||
730 | } while (++col & 7); | ||
731 | } else { | ||
732 | putchar(c); | ||
733 | col++; | ||
734 | } | ||
735 | } | ||
736 | } | ||
737 | } | ||
738 | |||
739 | |||
740 | #if ENABLE_FEATURE_DIFF_BINARY | ||
741 | static int asciifile(FILE *f) | ||
742 | { | ||
743 | int i, cnt; | ||
744 | |||
745 | if (option_mask32 & FLAG_a) | ||
746 | return 1; | ||
747 | rewind(f); | ||
748 | cnt = fread(g_read_buf, 1, COMMON_BUFSIZE, f); | ||
749 | for (i = 0; i < cnt; i++) { | ||
750 | if (!isprint(g_read_buf[i]) | ||
751 | && !isspace(g_read_buf[i]) | ||
752 | ) { | ||
753 | return 0; | ||
754 | } | ||
755 | } | ||
756 | return 1; | ||
757 | } | ||
758 | #else | ||
759 | #define asciifile(f) 1 | ||
760 | #endif | ||
761 | |||
762 | |||
763 | /* dump accumulated "unified" diff changes */ | ||
764 | static void dump_unified_vec(FILE *f1, FILE *f2) | ||
765 | { | ||
766 | struct context_vec *cvp = context_vec_start; | ||
767 | int lowa, upb, lowc, upd; | ||
768 | int a, b, c, d; | ||
769 | char ch; | ||
770 | |||
771 | if (context_vec_start > context_vec_ptr) | ||
772 | return; | ||
773 | |||
774 | b = d = 0; /* gcc */ | ||
775 | lowa = MAX(1, cvp->a - opt_U_context); | ||
776 | upb = MIN(nlen[0], context_vec_ptr->b + opt_U_context); | ||
777 | lowc = MAX(1, cvp->c - opt_U_context); | ||
778 | upd = MIN(nlen[1], context_vec_ptr->d + opt_U_context); | ||
779 | |||
780 | printf("@@ -"); | ||
781 | uni_range(lowa, upb); | ||
782 | printf(" +"); | ||
783 | uni_range(lowc, upd); | ||
784 | printf(" @@\n"); | ||
785 | |||
786 | /* | ||
787 | * Output changes in "unified" diff format--the old and new lines | ||
788 | * are printed together. | ||
789 | */ | ||
790 | for (; cvp <= context_vec_ptr; cvp++) { | ||
791 | a = cvp->a; | ||
792 | b = cvp->b; | ||
793 | c = cvp->c; | ||
794 | d = cvp->d; | ||
795 | |||
796 | /* | ||
797 | * c: both new and old changes | ||
798 | * d: only changes in the old file | ||
799 | * a: only changes in the new file | ||
800 | */ | ||
801 | if (a <= b && c <= d) | ||
802 | ch = 'c'; | ||
803 | else | ||
804 | ch = (a <= b) ? 'd' : 'a'; | ||
805 | #if 0 | ||
806 | switch (ch) { | ||
807 | case 'c': | ||
808 | // fetch() seeks! | ||
809 | fetch(ixold, lowa, a - 1, f1, ' '); | ||
810 | fetch(ixold, a, b, f1, '-'); | ||
811 | fetch(ixnew, c, d, f2, '+'); | ||
812 | break; | ||
813 | case 'd': | ||
814 | fetch(ixold, lowa, a - 1, f1, ' '); | ||
815 | fetch(ixold, a, b, f1, '-'); | ||
816 | break; | ||
817 | case 'a': | ||
818 | fetch(ixnew, lowc, c - 1, f2, ' '); | ||
819 | fetch(ixnew, c, d, f2, '+'); | ||
820 | break; | ||
821 | } | ||
822 | #else | ||
823 | if (ch == 'c' || ch == 'd') { | ||
824 | fetch(ixold, lowa, a - 1, f1, ' '); | ||
825 | fetch(ixold, a, b, f1, '-'); | ||
826 | } | ||
827 | if (ch == 'a') | ||
828 | fetch(ixnew, lowc, c - 1, f2, ' '); | ||
829 | if (ch == 'c' || ch == 'a') | ||
830 | fetch(ixnew, c, d, f2, '+'); | ||
831 | #endif | ||
832 | lowa = b + 1; | ||
833 | lowc = d + 1; | ||
834 | } | ||
835 | fetch(ixnew, d + 1, upd, f2, ' '); | ||
836 | |||
837 | context_vec_ptr = context_vec_start - 1; | ||
838 | } | ||
839 | |||
840 | |||
841 | static void print_header(const char *file1, const char *file2) | ||
842 | { | ||
843 | if (label1) | ||
844 | printf("--- %s\n", label1); | ||
845 | else | ||
846 | printf("--- %s\t%s", file1, ctime(&stb1.st_mtime)); | ||
847 | if (label2) | ||
848 | printf("+++ %s\n", label2); | ||
849 | else | ||
850 | printf("+++ %s\t%s", file2, ctime(&stb2.st_mtime)); | ||
851 | } | ||
852 | |||
853 | |||
854 | /* | ||
855 | * Indicate that there is a difference between lines a and b of the from file | ||
856 | * to get to lines c to d of the to file. If a is greater than b then there | ||
857 | * are no lines in the from file involved and this means that there were | ||
858 | * lines appended (beginning at b). If c is greater than d then there are | ||
859 | * lines missing from the to file. | ||
860 | */ | ||
861 | static void change(char *file1, FILE *f1, char *file2, FILE *f2, | ||
862 | int a, int b, int c, int d) | ||
863 | { | ||
864 | if ((a > b && c > d) || (option_mask32 & FLAG_q)) { | ||
865 | anychange = 1; | ||
866 | return; | ||
867 | } | ||
868 | |||
869 | /* | ||
870 | * Allocate change records as needed. | ||
871 | */ | ||
872 | if (context_vec_ptr == context_vec_end - 1) { | ||
873 | ptrdiff_t offset = context_vec_ptr - context_vec_start; | ||
874 | |||
875 | max_context <<= 1; | ||
876 | context_vec_start = xrealloc(context_vec_start, | ||
877 | max_context * sizeof(struct context_vec)); | ||
878 | context_vec_end = context_vec_start + max_context; | ||
879 | context_vec_ptr = context_vec_start + offset; | ||
880 | } | ||
881 | if (anychange == 0) { | ||
882 | /* | ||
883 | * Print the context/unidiff header first time through. | ||
884 | */ | ||
885 | print_header(file1, file2); | ||
886 | } else if (a > context_vec_ptr->b + (2 * opt_U_context) + 1 | ||
887 | && c > context_vec_ptr->d + (2 * opt_U_context) + 1 | ||
888 | ) { | ||
889 | /* | ||
890 | * If this change is more than 'context' lines from the | ||
891 | * previous change, dump the record and reset it. | ||
892 | */ | ||
893 | // dump_unified_vec() seeks! | ||
894 | dump_unified_vec(f1, f2); | ||
895 | } | ||
896 | context_vec_ptr++; | ||
897 | context_vec_ptr->a = a; | ||
898 | context_vec_ptr->b = b; | ||
899 | context_vec_ptr->c = c; | ||
900 | context_vec_ptr->d = d; | ||
901 | anychange = 1; | ||
902 | } | ||
903 | |||
904 | |||
905 | static void output(char *file1, FILE *f1, char *file2, FILE *f2) | ||
906 | { | ||
907 | /* Note that j0 and j1 can't be used as they are defined in math.h. | ||
908 | * This also allows the rather amusing variable 'j00'... */ | ||
909 | int m, i0, i1, j00, j01; | ||
910 | |||
911 | rewind(f1); | ||
912 | rewind(f2); | ||
913 | m = nlen[0]; | ||
914 | J[0] = 0; | ||
915 | J[m + 1] = nlen[1] + 1; | ||
916 | for (i0 = 1; i0 <= m; i0 = i1 + 1) { | ||
917 | while (i0 <= m && J[i0] == J[i0 - 1] + 1) | ||
918 | i0++; | ||
919 | j00 = J[i0 - 1] + 1; | ||
920 | i1 = i0 - 1; | ||
921 | while (i1 < m && J[i1 + 1] == 0) | ||
922 | i1++; | ||
923 | j01 = J[i1 + 1] - 1; | ||
924 | J[i1] = j01; | ||
925 | // change() seeks! | ||
926 | change(file1, f1, file2, f2, i0, i1, j00, j01); | ||
927 | } | ||
928 | if (m == 0) { | ||
929 | // change() seeks! | ||
930 | change(file1, f1, file2, f2, 1, 0, 1, nlen[1]); | ||
931 | } | ||
932 | if (anychange != 0 && !(option_mask32 & FLAG_q)) { | ||
933 | // dump_unified_vec() seeks! | ||
934 | dump_unified_vec(f1, f2); | ||
935 | } | ||
936 | } | ||
937 | |||
938 | /* | ||
939 | * The following code uses an algorithm due to Harold Stone, | ||
940 | * which finds a pair of longest identical subsequences in | ||
941 | * the two files. | ||
942 | * | ||
943 | * The major goal is to generate the match vector J. | ||
944 | * J[i] is the index of the line in file1 corresponding | ||
945 | * to line i in file0. J[i] = 0 if there is no | ||
946 | * such line in file1. | ||
947 | * | ||
948 | * Lines are hashed so as to work in core. All potential | ||
949 | * matches are located by sorting the lines of each file | ||
950 | * on the hash (called "value"). In particular, this | ||
951 | * collects the equivalence classes in file1 together. | ||
952 | * Subroutine equiv replaces the value of each line in | ||
953 | * file0 by the index of the first element of its | ||
954 | * matching equivalence in (the reordered) file1. | ||
955 | * To save space equiv squeezes file1 into a single | ||
956 | * array member in which the equivalence classes | ||
957 | * are simply concatenated, except that their first | ||
958 | * members are flagged by changing sign. | ||
959 | * | ||
960 | * Next the indices that point into member are unsorted into | ||
961 | * array class according to the original order of file0. | ||
962 | * | ||
963 | * The cleverness lies in routine stone. This marches | ||
964 | * through the lines of file0, developing a vector klist | ||
965 | * of "k-candidates". At step i a k-candidate is a matched | ||
966 | * pair of lines x,y (x in file0, y in file1) such that | ||
967 | * there is a common subsequence of length k | ||
968 | * between the first i lines of file0 and the first y | ||
969 | * lines of file1, but there is no such subsequence for | ||
970 | * any smaller y. x is the earliest possible mate to y | ||
971 | * that occurs in such a subsequence. | ||
972 | * | ||
973 | * Whenever any of the members of the equivalence class of | ||
974 | * lines in file1 matable to a line in file0 has serial number | ||
975 | * less than the y of some k-candidate, that k-candidate | ||
976 | * with the smallest such y is replaced. The new | ||
977 | * k-candidate is chained (via pred) to the current | ||
978 | * k-1 candidate so that the actual subsequence can | ||
979 | * be recovered. When a member has serial number greater | ||
980 | * that the y of all k-candidates, the klist is extended. | ||
981 | * At the end, the longest subsequence is pulled out | ||
982 | * and placed in the array J by unravel | ||
983 | * | ||
984 | * With J in hand, the matches there recorded are | ||
985 | * checked against reality to assure that no spurious | ||
986 | * matches have crept in due to hashing. If they have, | ||
987 | * they are broken, and "jackpot" is recorded--a harmless | ||
988 | * matter except that a true match for a spuriously | ||
989 | * mated line may now be unnecessarily reported as a change. | ||
990 | * | ||
991 | * Much of the complexity of the program comes simply | ||
992 | * from trying to minimize core utilization and | ||
993 | * maximize the range of doable problems by dynamically | ||
994 | * allocating what is needed and reusing what is not. | ||
995 | * The core requirements for problems larger than somewhat | ||
996 | * are (in words) 2*length(file0) + length(file1) + | ||
997 | * 3*(number of k-candidates installed), typically about | ||
998 | * 6n words for files of length n. | ||
999 | */ | ||
1000 | /* NB: files can be not REGular. The only sure thing that they | ||
1001 | * are not both DIRectories. */ | ||
1002 | static unsigned diffreg(char *file1, char *file2, int flags) | ||
1003 | { | ||
1004 | int *member; /* will be overlaid on nfile[1] */ | ||
1005 | int *class; /* will be overlaid on nfile[0] */ | ||
1006 | int *klist; /* will be overlaid on nfile[0] after class */ | ||
1007 | FILE *f1; | ||
1008 | FILE *f2; | ||
1009 | unsigned rval; | ||
1010 | int i; | ||
1011 | |||
1012 | anychange = 0; | ||
1013 | context_vec_ptr = context_vec_start - 1; | ||
1014 | tempname1 = tempname2 = NULL; | ||
1015 | |||
1016 | /* Is any of them a directory? Then it's simple */ | ||
1017 | if (S_ISDIR(stb1.st_mode) != S_ISDIR(stb2.st_mode)) | ||
1018 | return (S_ISDIR(stb1.st_mode) ? D_ISDIR1 : D_ISDIR2); | ||
1019 | |||
1020 | /* None of them are directories */ | ||
1021 | rval = D_SAME; | ||
1022 | |||
1023 | if (flags & D_EMPTY1) | ||
1024 | /* can't be stdin, but xfopen_stdin() is smaller code */ | ||
1025 | f1 = xfopen_stdin(bb_dev_null); | ||
1026 | else | ||
1027 | f1 = xfopen_stdin(file1); | ||
1028 | if (flags & D_EMPTY2) | ||
1029 | f2 = xfopen_stdin(bb_dev_null); | ||
1030 | else | ||
1031 | f2 = xfopen_stdin(file2); | ||
1032 | |||
1033 | /* NB: if D_EMPTY1/2 is set, other file is always a regular file, | ||
1034 | * not pipe/fifo/chardev/etc - D_EMPTY is used by "diff -r" only, | ||
1035 | * and it never diffs non-ordinary files in subdirs. */ | ||
1036 | if (!(flags & (D_EMPTY1 | D_EMPTY2))) { | ||
1037 | /* Quick check whether they are different */ | ||
1038 | /* NB: copies non-REG files to tempfiles and fills tempname1/2 */ | ||
1039 | i = files_differ(f1, f2); | ||
1040 | if (i != 1) { /* not different? */ | ||
1041 | if (i != 0) /* error? */ | ||
1042 | exit_status |= 2; | ||
1043 | goto closem; | ||
1044 | } | ||
1045 | } | ||
1046 | |||
1047 | if (!asciifile(f1) || !asciifile(f2)) { | ||
1048 | rval = D_BINARY; | ||
1049 | exit_status |= 1; | ||
1050 | goto closem; | ||
1051 | } | ||
1052 | |||
1053 | // Rewind inside! | ||
1054 | prepare(0, f1 /*, stb1.st_size*/); | ||
1055 | prepare(1, f2 /*, stb2.st_size*/); | ||
1056 | prune(); | ||
1057 | sort(sfile[0], slen[0]); | ||
1058 | sort(sfile[1], slen[1]); | ||
1059 | |||
1060 | member = (int *) nfile[1]; | ||
1061 | equiv(sfile[0], slen[0], sfile[1], slen[1], member); | ||
1062 | //TODO: xrealloc_vector? | ||
1063 | member = xrealloc(member, (slen[1] + 2) * sizeof(int)); | ||
1064 | |||
1065 | class = (int *) nfile[0]; | ||
1066 | unsort(sfile[0], slen[0], class); | ||
1067 | class = xrealloc(class, (slen[0] + 2) * sizeof(int)); | ||
1068 | |||
1069 | klist = xmalloc((slen[0] + 2) * sizeof(int)); | ||
1070 | clen = 0; | ||
1071 | clistlen = 100; | ||
1072 | clist = xmalloc(clistlen * sizeof(struct cand)); | ||
1073 | i = stone(class, slen[0], member, klist); | ||
1074 | free(member); | ||
1075 | free(class); | ||
1076 | |||
1077 | J = xrealloc(J, (nlen[0] + 2) * sizeof(int)); | ||
1078 | unravel(klist[i]); | ||
1079 | free(clist); | ||
1080 | free(klist); | ||
1081 | |||
1082 | ixold = xrealloc(ixold, (nlen[0] + 2) * sizeof(long)); | ||
1083 | ixnew = xrealloc(ixnew, (nlen[1] + 2) * sizeof(long)); | ||
1084 | // Rewind inside! | ||
1085 | check(f1, f2); | ||
1086 | // Rewind inside! | ||
1087 | output(file1, f1, file2, f2); | ||
1088 | |||
1089 | closem: | ||
1090 | if (anychange) { | ||
1091 | exit_status |= 1; | ||
1092 | if (rval == D_SAME) | ||
1093 | rval = D_DIFFER; | ||
1094 | } | ||
1095 | fclose_if_not_stdin(f1); | ||
1096 | fclose_if_not_stdin(f2); | ||
1097 | if (tempname1) { | ||
1098 | unlink(tempname1); | ||
1099 | free(tempname1); | ||
1100 | } | ||
1101 | if (tempname2) { | ||
1102 | unlink(tempname2); | ||
1103 | free(tempname2); | ||
1104 | } | ||
1105 | return rval; | ||
1106 | } | ||
1107 | |||
1108 | |||
1109 | #if ENABLE_FEATURE_DIFF_DIR | ||
1110 | static void do_diff(char *dir1, char *path1, char *dir2, char *path2) | ||
1111 | { | ||
1112 | int flags = 0; /*D_HEADER;*/ | ||
1113 | int val; | ||
1114 | char *fullpath1 = NULL; /* if -N */ | ||
1115 | char *fullpath2 = NULL; | ||
1116 | |||
1117 | if (path1) | ||
1118 | fullpath1 = concat_path_file(dir1, path1); | ||
1119 | if (path2) | ||
1120 | fullpath2 = concat_path_file(dir2, path2); | ||
1121 | |||
1122 | if (!fullpath1 || stat(fullpath1, &stb1) != 0) { | ||
1123 | flags |= D_EMPTY1; | ||
1124 | memset(&stb1, 0, sizeof(stb1)); | ||
1125 | if (path2) { | ||
1126 | free(fullpath1); | ||
1127 | fullpath1 = concat_path_file(dir1, path2); | ||
1128 | } | ||
1129 | } | ||
1130 | if (!fullpath2 || stat(fullpath2, &stb2) != 0) { | ||
1131 | flags |= D_EMPTY2; | ||
1132 | memset(&stb2, 0, sizeof(stb2)); | ||
1133 | stb2.st_mode = stb1.st_mode; | ||
1134 | if (path1) { | ||
1135 | free(fullpath2); | ||
1136 | fullpath2 = concat_path_file(dir2, path1); | ||
1137 | } | ||
1138 | } | ||
1139 | |||
1140 | if (stb1.st_mode == 0) | ||
1141 | stb1.st_mode = stb2.st_mode; | ||
1142 | |||
1143 | if (S_ISDIR(stb1.st_mode) && S_ISDIR(stb2.st_mode)) { | ||
1144 | printf("Common subdirectories: %s and %s\n", fullpath1, fullpath2); | ||
1145 | goto ret; | ||
1146 | } | ||
1147 | |||
1148 | if (!S_ISREG(stb1.st_mode) && !S_ISDIR(stb1.st_mode)) | ||
1149 | val = D_SKIPPED1; | ||
1150 | else if (!S_ISREG(stb2.st_mode) && !S_ISDIR(stb2.st_mode)) | ||
1151 | val = D_SKIPPED2; | ||
1152 | else { | ||
1153 | /* Both files are either REGular or DIRectories */ | ||
1154 | val = diffreg(fullpath1, fullpath2, flags); | ||
1155 | } | ||
1156 | |||
1157 | print_status(val, fullpath1, fullpath2 /*, NULL*/); | ||
1158 | ret: | ||
1159 | free(fullpath1); | ||
1160 | free(fullpath2); | ||
1161 | } | ||
1162 | #endif | ||
1163 | |||
1164 | |||
1165 | #if ENABLE_FEATURE_DIFF_DIR | ||
1166 | /* This function adds a filename to dl, the directory listing. */ | ||
1167 | static int FAST_FUNC add_to_dirlist(const char *filename, | ||
1168 | struct stat *sb UNUSED_PARAM, | ||
1169 | void *userdata, | ||
1170 | int depth UNUSED_PARAM) | ||
1171 | { | ||
1172 | dl = xrealloc_vector(dl, 5, dl_count); | ||
1173 | dl[dl_count] = xstrdup(filename + (int)(ptrdiff_t)userdata); | ||
1174 | dl_count++; | ||
1175 | return TRUE; | ||
1176 | } | ||
1177 | |||
1178 | |||
1179 | /* This returns a sorted directory listing. */ | ||
1180 | static char **get_recursive_dirlist(char *path) | ||
1181 | { | ||
1182 | dl_count = 0; | ||
1183 | dl = xzalloc(sizeof(dl[0])); | ||
1184 | |||
1185 | /* We need to trim root directory prefix. | ||
1186 | * Using void *userdata to specify its length, | ||
1187 | * add_to_dirlist will remove it. */ | ||
1188 | if (option_mask32 & FLAG_r) { | ||
1189 | recursive_action(path, ACTION_RECURSE|ACTION_FOLLOWLINKS, | ||
1190 | add_to_dirlist, /* file_action */ | ||
1191 | NULL, /* dir_action */ | ||
1192 | (void*)(ptrdiff_t)(strlen(path) + 1), | ||
1193 | 0); | ||
1194 | } else { | ||
1195 | DIR *dp; | ||
1196 | struct dirent *ep; | ||
1197 | |||
1198 | dp = warn_opendir(path); | ||
1199 | while ((ep = readdir(dp))) { | ||
1200 | if (!strcmp(ep->d_name, "..") || LONE_CHAR(ep->d_name, '.')) | ||
1201 | continue; | ||
1202 | add_to_dirlist(ep->d_name, NULL, (void*)(int)0, 0); | ||
1203 | } | ||
1204 | closedir(dp); | ||
1205 | } | ||
1206 | |||
1207 | /* Sort dl alphabetically. */ | ||
1208 | qsort_string_vector(dl, dl_count); | ||
1209 | |||
1210 | dl[dl_count] = NULL; | ||
1211 | return dl; | ||
1212 | } | ||
1213 | |||
1214 | |||
1215 | static void diffdir(char *p1, char *p2) | ||
1216 | { | ||
1217 | char **dirlist1, **dirlist2; | ||
1218 | char *dp1, *dp2; | ||
1219 | int pos; | ||
1220 | |||
1221 | /* Check for trailing slashes. */ | ||
1222 | dp1 = last_char_is(p1, '/'); | ||
1223 | if (dp1 != NULL) | ||
1224 | *dp1 = '\0'; | ||
1225 | dp2 = last_char_is(p2, '/'); | ||
1226 | if (dp2 != NULL) | ||
1227 | *dp2 = '\0'; | ||
1228 | |||
1229 | /* Get directory listings for p1 and p2. */ | ||
1230 | dirlist1 = get_recursive_dirlist(p1); | ||
1231 | dirlist2 = get_recursive_dirlist(p2); | ||
1232 | |||
1233 | /* If -S was set, find the starting point. */ | ||
1234 | if (opt_S_start) { | ||
1235 | while (*dirlist1 != NULL && strcmp(*dirlist1, opt_S_start) < 0) | ||
1236 | dirlist1++; | ||
1237 | while (*dirlist2 != NULL && strcmp(*dirlist2, opt_S_start) < 0) | ||
1238 | dirlist2++; | ||
1239 | if ((*dirlist1 == NULL) || (*dirlist2 == NULL)) | ||
1240 | bb_error_msg(bb_msg_invalid_arg, "NULL", "-S"); | ||
1241 | } | ||
1242 | |||
1243 | /* Now that both dirlist1 and dirlist2 contain sorted directory | ||
1244 | * listings, we can start to go through dirlist1. If both listings | ||
1245 | * contain the same file, then do a normal diff. Otherwise, behaviour | ||
1246 | * is determined by whether the -N flag is set. */ | ||
1247 | while (*dirlist1 != NULL || *dirlist2 != NULL) { | ||
1248 | dp1 = *dirlist1; | ||
1249 | dp2 = *dirlist2; | ||
1250 | pos = dp1 == NULL ? 1 : (dp2 == NULL ? -1 : strcmp(dp1, dp2)); | ||
1251 | if (pos == 0) { | ||
1252 | do_diff(p1, dp1, p2, dp2); | ||
1253 | dirlist1++; | ||
1254 | dirlist2++; | ||
1255 | } else if (pos < 0) { | ||
1256 | if (option_mask32 & FLAG_N) | ||
1257 | do_diff(p1, dp1, p2, NULL); | ||
1258 | else | ||
1259 | print_only(p1, dp1); | ||
1260 | dirlist1++; | ||
1261 | } else { | ||
1262 | if (option_mask32 & FLAG_N) | ||
1263 | do_diff(p1, NULL, p2, dp2); | ||
1264 | else | ||
1265 | print_only(p2, dp2); | ||
1266 | dirlist2++; | ||
1267 | } | ||
1268 | } | ||
1269 | } | ||
1270 | #endif | ||
1271 | |||
1272 | |||
1273 | int diff_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE; | ||
1274 | int diff_main(int argc UNUSED_PARAM, char **argv) | ||
1275 | { | ||
1276 | int gotstdin = 0; | ||
1277 | char *f1, *f2; | ||
1278 | llist_t *L_arg = NULL; | ||
1279 | |||
1280 | INIT_G(); | ||
1281 | |||
1282 | /* exactly 2 params; collect multiple -L <label>; -U N */ | ||
1283 | opt_complementary = "=2:L::U+"; | ||
1284 | getopt32(argv, "abdiL:NqrsS:tTU:wu" | ||
1285 | "p" /* ignored (for compatibility) */, | ||
1286 | &L_arg, &opt_S_start, &opt_U_context); | ||
1287 | /*argc -= optind;*/ | ||
1288 | argv += optind; | ||
1289 | while (L_arg) { | ||
1290 | if (label1 && label2) | ||
1291 | bb_show_usage(); | ||
1292 | if (label1) /* then label2 is NULL */ | ||
1293 | label2 = label1; | ||
1294 | label1 = llist_pop(&L_arg); | ||
1295 | } | ||
1296 | |||
1297 | /* | ||
1298 | * Do sanity checks, fill in stb1 and stb2 and call the appropriate | ||
1299 | * driver routine. Both drivers use the contents of stb1 and stb2. | ||
1300 | */ | ||
1301 | f1 = argv[0]; | ||
1302 | f2 = argv[1]; | ||
1303 | if (LONE_DASH(f1)) { | ||
1304 | fstat(STDIN_FILENO, &stb1); | ||
1305 | gotstdin++; | ||
1306 | } else | ||
1307 | xstat(f1, &stb1); | ||
1308 | if (LONE_DASH(f2)) { | ||
1309 | fstat(STDIN_FILENO, &stb2); | ||
1310 | gotstdin++; | ||
1311 | } else | ||
1312 | xstat(f2, &stb2); | ||
1313 | |||
1314 | if (gotstdin && (S_ISDIR(stb1.st_mode) || S_ISDIR(stb2.st_mode))) | ||
1315 | bb_error_msg_and_die("can't compare stdin to a directory"); | ||
1316 | |||
1317 | if (S_ISDIR(stb1.st_mode) && S_ISDIR(stb2.st_mode)) { | ||
1318 | #if ENABLE_FEATURE_DIFF_DIR | ||
1319 | diffdir(f1, f2); | ||
1320 | return exit_status; | ||
1321 | #else | ||
1322 | bb_error_msg_and_die("no support for directory comparison"); | ||
1323 | #endif | ||
1324 | } | ||
1325 | |||
1326 | if (S_ISDIR(stb1.st_mode)) { /* "diff dir file" */ | ||
1327 | /* NB: "diff dir dir2/dir3/file" must become | ||
1328 | * "diff dir/file dir2/dir3/file" */ | ||
1329 | char *slash = strrchr(f2, '/'); | ||
1330 | f1 = concat_path_file(f1, slash ? slash + 1 : f2); | ||
1331 | xstat(f1, &stb1); | ||
1332 | } | ||
1333 | if (S_ISDIR(stb2.st_mode)) { | ||
1334 | char *slash = strrchr(f1, '/'); | ||
1335 | f2 = concat_path_file(f2, slash ? slash + 1 : f1); | ||
1336 | xstat(f2, &stb2); | ||
1337 | } | ||
1338 | |||
1339 | /* diffreg can get non-regular files here, | ||
1340 | * they are not both DIRestories */ | ||
1341 | print_status((gotstdin > 1 ? D_SAME : diffreg(f1, f2, 0)), | ||
1342 | f1, f2 /*, NULL*/); | ||
1343 | return exit_status; | ||
1344 | } |