]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * This handles recursive filename detection with exclude | |
3 | * files, index knowledge etc.. | |
4 | * | |
5 | * See Documentation/technical/api-directory-listing.txt | |
6 | * | |
7 | * Copyright (C) Linus Torvalds, 2005-2006 | |
8 | * Junio Hamano, 2005-2006 | |
9 | */ | |
10 | #define NO_THE_INDEX_COMPATIBILITY_MACROS | |
11 | #include "cache.h" | |
12 | #include "dir.h" | |
13 | #include "attr.h" | |
14 | #include "refs.h" | |
15 | #include "wildmatch.h" | |
16 | #include "pathspec.h" | |
17 | #include "utf8.h" | |
18 | #include "varint.h" | |
19 | #include "ewah/ewok.h" | |
20 | ||
21 | /* | |
22 | * Tells read_directory_recursive how a file or directory should be treated. | |
23 | * Values are ordered by significance, e.g. if a directory contains both | |
24 | * excluded and untracked files, it is listed as untracked because | |
25 | * path_untracked > path_excluded. | |
26 | */ | |
27 | enum path_treatment { | |
28 | path_none = 0, | |
29 | path_recurse, | |
30 | path_excluded, | |
31 | path_untracked | |
32 | }; | |
33 | ||
34 | /* | |
35 | * Support data structure for our opendir/readdir/closedir wrappers | |
36 | */ | |
37 | struct cached_dir { | |
38 | DIR *fdir; | |
39 | struct untracked_cache_dir *untracked; | |
40 | int nr_files; | |
41 | int nr_dirs; | |
42 | ||
43 | struct dirent *de; | |
44 | const char *file; | |
45 | struct untracked_cache_dir *ucd; | |
46 | }; | |
47 | ||
48 | static enum path_treatment read_directory_recursive(struct dir_struct *dir, | |
49 | const char *path, int len, struct untracked_cache_dir *untracked, | |
50 | int check_only, const struct pathspec *pathspec); | |
51 | static int get_dtype(struct dirent *de, struct index_state *istate, | |
52 | const char *path, int len); | |
53 | ||
54 | int fspathcmp(const char *a, const char *b) | |
55 | { | |
56 | return ignore_case ? strcasecmp(a, b) : strcmp(a, b); | |
57 | } | |
58 | ||
59 | int fspathncmp(const char *a, const char *b, size_t count) | |
60 | { | |
61 | return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count); | |
62 | } | |
63 | ||
64 | int git_fnmatch(const struct pathspec_item *item, | |
65 | const char *pattern, const char *string, | |
66 | int prefix) | |
67 | { | |
68 | if (prefix > 0) { | |
69 | if (ps_strncmp(item, pattern, string, prefix)) | |
70 | return WM_NOMATCH; | |
71 | pattern += prefix; | |
72 | string += prefix; | |
73 | } | |
74 | if (item->flags & PATHSPEC_ONESTAR) { | |
75 | int pattern_len = strlen(++pattern); | |
76 | int string_len = strlen(string); | |
77 | return string_len < pattern_len || | |
78 | ps_strcmp(item, pattern, | |
79 | string + string_len - pattern_len); | |
80 | } | |
81 | if (item->magic & PATHSPEC_GLOB) | |
82 | return wildmatch(pattern, string, | |
83 | WM_PATHNAME | | |
84 | (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0), | |
85 | NULL); | |
86 | else | |
87 | /* wildmatch has not learned no FNM_PATHNAME mode yet */ | |
88 | return wildmatch(pattern, string, | |
89 | item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0, | |
90 | NULL); | |
91 | } | |
92 | ||
93 | static int fnmatch_icase_mem(const char *pattern, int patternlen, | |
94 | const char *string, int stringlen, | |
95 | int flags) | |
96 | { | |
97 | int match_status; | |
98 | struct strbuf pat_buf = STRBUF_INIT; | |
99 | struct strbuf str_buf = STRBUF_INIT; | |
100 | const char *use_pat = pattern; | |
101 | const char *use_str = string; | |
102 | ||
103 | if (pattern[patternlen]) { | |
104 | strbuf_add(&pat_buf, pattern, patternlen); | |
105 | use_pat = pat_buf.buf; | |
106 | } | |
107 | if (string[stringlen]) { | |
108 | strbuf_add(&str_buf, string, stringlen); | |
109 | use_str = str_buf.buf; | |
110 | } | |
111 | ||
112 | if (ignore_case) | |
113 | flags |= WM_CASEFOLD; | |
114 | match_status = wildmatch(use_pat, use_str, flags, NULL); | |
115 | ||
116 | strbuf_release(&pat_buf); | |
117 | strbuf_release(&str_buf); | |
118 | ||
119 | return match_status; | |
120 | } | |
121 | ||
122 | static size_t common_prefix_len(const struct pathspec *pathspec) | |
123 | { | |
124 | int n; | |
125 | size_t max = 0; | |
126 | ||
127 | /* | |
128 | * ":(icase)path" is treated as a pathspec full of | |
129 | * wildcard. In other words, only prefix is considered common | |
130 | * prefix. If the pathspec is abc/foo abc/bar, running in | |
131 | * subdir xyz, the common prefix is still xyz, not xuz/abc as | |
132 | * in non-:(icase). | |
133 | */ | |
134 | GUARD_PATHSPEC(pathspec, | |
135 | PATHSPEC_FROMTOP | | |
136 | PATHSPEC_MAXDEPTH | | |
137 | PATHSPEC_LITERAL | | |
138 | PATHSPEC_GLOB | | |
139 | PATHSPEC_ICASE | | |
140 | PATHSPEC_EXCLUDE | | |
141 | PATHSPEC_ATTR); | |
142 | ||
143 | for (n = 0; n < pathspec->nr; n++) { | |
144 | size_t i = 0, len = 0, item_len; | |
145 | if (pathspec->items[n].magic & PATHSPEC_EXCLUDE) | |
146 | continue; | |
147 | if (pathspec->items[n].magic & PATHSPEC_ICASE) | |
148 | item_len = pathspec->items[n].prefix; | |
149 | else | |
150 | item_len = pathspec->items[n].nowildcard_len; | |
151 | while (i < item_len && (n == 0 || i < max)) { | |
152 | char c = pathspec->items[n].match[i]; | |
153 | if (c != pathspec->items[0].match[i]) | |
154 | break; | |
155 | if (c == '/') | |
156 | len = i + 1; | |
157 | i++; | |
158 | } | |
159 | if (n == 0 || len < max) { | |
160 | max = len; | |
161 | if (!max) | |
162 | break; | |
163 | } | |
164 | } | |
165 | return max; | |
166 | } | |
167 | ||
168 | /* | |
169 | * Returns a copy of the longest leading path common among all | |
170 | * pathspecs. | |
171 | */ | |
172 | char *common_prefix(const struct pathspec *pathspec) | |
173 | { | |
174 | unsigned long len = common_prefix_len(pathspec); | |
175 | ||
176 | return len ? xmemdupz(pathspec->items[0].match, len) : NULL; | |
177 | } | |
178 | ||
179 | int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec) | |
180 | { | |
181 | const char *prefix; | |
182 | size_t prefix_len; | |
183 | ||
184 | /* | |
185 | * Calculate common prefix for the pathspec, and | |
186 | * use that to optimize the directory walk | |
187 | */ | |
188 | prefix_len = common_prefix_len(pathspec); | |
189 | prefix = prefix_len ? pathspec->items[0].match : ""; | |
190 | ||
191 | /* Read the directory and prune it */ | |
192 | read_directory(dir, prefix, prefix_len, pathspec); | |
193 | ||
194 | return prefix_len; | |
195 | } | |
196 | ||
197 | int within_depth(const char *name, int namelen, | |
198 | int depth, int max_depth) | |
199 | { | |
200 | const char *cp = name, *cpe = name + namelen; | |
201 | ||
202 | while (cp < cpe) { | |
203 | if (*cp++ != '/') | |
204 | continue; | |
205 | depth++; | |
206 | if (depth > max_depth) | |
207 | return 0; | |
208 | } | |
209 | return 1; | |
210 | } | |
211 | ||
212 | #define DO_MATCH_EXCLUDE (1<<0) | |
213 | #define DO_MATCH_DIRECTORY (1<<1) | |
214 | #define DO_MATCH_SUBMODULE (1<<2) | |
215 | ||
216 | static int match_attrs(const char *name, int namelen, | |
217 | const struct pathspec_item *item) | |
218 | { | |
219 | int i; | |
220 | ||
221 | git_check_attr(name, item->attr_check); | |
222 | for (i = 0; i < item->attr_match_nr; i++) { | |
223 | const char *value; | |
224 | int matched; | |
225 | enum attr_match_mode match_mode; | |
226 | ||
227 | value = item->attr_check->items[i].value; | |
228 | match_mode = item->attr_match[i].match_mode; | |
229 | ||
230 | if (ATTR_TRUE(value)) | |
231 | matched = (match_mode == MATCH_SET); | |
232 | else if (ATTR_FALSE(value)) | |
233 | matched = (match_mode == MATCH_UNSET); | |
234 | else if (ATTR_UNSET(value)) | |
235 | matched = (match_mode == MATCH_UNSPECIFIED); | |
236 | else | |
237 | matched = (match_mode == MATCH_VALUE && | |
238 | !strcmp(item->attr_match[i].value, value)); | |
239 | if (!matched) | |
240 | return 0; | |
241 | } | |
242 | ||
243 | return 1; | |
244 | } | |
245 | ||
246 | /* | |
247 | * Does 'match' match the given name? | |
248 | * A match is found if | |
249 | * | |
250 | * (1) the 'match' string is leading directory of 'name', or | |
251 | * (2) the 'match' string is a wildcard and matches 'name', or | |
252 | * (3) the 'match' string is exactly the same as 'name'. | |
253 | * | |
254 | * and the return value tells which case it was. | |
255 | * | |
256 | * It returns 0 when there is no match. | |
257 | */ | |
258 | static int match_pathspec_item(const struct pathspec_item *item, int prefix, | |
259 | const char *name, int namelen, unsigned flags) | |
260 | { | |
261 | /* name/namelen has prefix cut off by caller */ | |
262 | const char *match = item->match + prefix; | |
263 | int matchlen = item->len - prefix; | |
264 | ||
265 | /* | |
266 | * The normal call pattern is: | |
267 | * 1. prefix = common_prefix_len(ps); | |
268 | * 2. prune something, or fill_directory | |
269 | * 3. match_pathspec() | |
270 | * | |
271 | * 'prefix' at #1 may be shorter than the command's prefix and | |
272 | * it's ok for #2 to match extra files. Those extras will be | |
273 | * trimmed at #3. | |
274 | * | |
275 | * Suppose the pathspec is 'foo' and '../bar' running from | |
276 | * subdir 'xyz'. The common prefix at #1 will be empty, thanks | |
277 | * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The | |
278 | * user does not want XYZ/foo, only the "foo" part should be | |
279 | * case-insensitive. We need to filter out XYZ/foo here. In | |
280 | * other words, we do not trust the caller on comparing the | |
281 | * prefix part when :(icase) is involved. We do exact | |
282 | * comparison ourselves. | |
283 | * | |
284 | * Normally the caller (common_prefix_len() in fact) does | |
285 | * _exact_ matching on name[-prefix+1..-1] and we do not need | |
286 | * to check that part. Be defensive and check it anyway, in | |
287 | * case common_prefix_len is changed, or a new caller is | |
288 | * introduced that does not use common_prefix_len. | |
289 | * | |
290 | * If the penalty turns out too high when prefix is really | |
291 | * long, maybe change it to | |
292 | * strncmp(match, name, item->prefix - prefix) | |
293 | */ | |
294 | if (item->prefix && (item->magic & PATHSPEC_ICASE) && | |
295 | strncmp(item->match, name - prefix, item->prefix)) | |
296 | return 0; | |
297 | ||
298 | if (item->attr_match_nr && !match_attrs(name, namelen, item)) | |
299 | return 0; | |
300 | ||
301 | /* If the match was just the prefix, we matched */ | |
302 | if (!*match) | |
303 | return MATCHED_RECURSIVELY; | |
304 | ||
305 | if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) { | |
306 | if (matchlen == namelen) | |
307 | return MATCHED_EXACTLY; | |
308 | ||
309 | if (match[matchlen-1] == '/' || name[matchlen] == '/') | |
310 | return MATCHED_RECURSIVELY; | |
311 | } else if ((flags & DO_MATCH_DIRECTORY) && | |
312 | match[matchlen - 1] == '/' && | |
313 | namelen == matchlen - 1 && | |
314 | !ps_strncmp(item, match, name, namelen)) | |
315 | return MATCHED_EXACTLY; | |
316 | ||
317 | if (item->nowildcard_len < item->len && | |
318 | !git_fnmatch(item, match, name, | |
319 | item->nowildcard_len - prefix)) | |
320 | return MATCHED_FNMATCH; | |
321 | ||
322 | /* Perform checks to see if "name" is a super set of the pathspec */ | |
323 | if (flags & DO_MATCH_SUBMODULE) { | |
324 | /* name is a literal prefix of the pathspec */ | |
325 | if ((namelen < matchlen) && | |
326 | (match[namelen] == '/') && | |
327 | !ps_strncmp(item, match, name, namelen)) | |
328 | return MATCHED_RECURSIVELY; | |
329 | ||
330 | /* name" doesn't match up to the first wild character */ | |
331 | if (item->nowildcard_len < item->len && | |
332 | ps_strncmp(item, match, name, | |
333 | item->nowildcard_len - prefix)) | |
334 | return 0; | |
335 | ||
336 | /* | |
337 | * Here is where we would perform a wildmatch to check if | |
338 | * "name" can be matched as a directory (or a prefix) against | |
339 | * the pathspec. Since wildmatch doesn't have this capability | |
340 | * at the present we have to punt and say that it is a match, | |
341 | * potentially returning a false positive | |
342 | * The submodules themselves will be able to perform more | |
343 | * accurate matching to determine if the pathspec matches. | |
344 | */ | |
345 | return MATCHED_RECURSIVELY; | |
346 | } | |
347 | ||
348 | return 0; | |
349 | } | |
350 | ||
351 | /* | |
352 | * Given a name and a list of pathspecs, returns the nature of the | |
353 | * closest (i.e. most specific) match of the name to any of the | |
354 | * pathspecs. | |
355 | * | |
356 | * The caller typically calls this multiple times with the same | |
357 | * pathspec and seen[] array but with different name/namelen | |
358 | * (e.g. entries from the index) and is interested in seeing if and | |
359 | * how each pathspec matches all the names it calls this function | |
360 | * with. A mark is left in the seen[] array for each pathspec element | |
361 | * indicating the closest type of match that element achieved, so if | |
362 | * seen[n] remains zero after multiple invocations, that means the nth | |
363 | * pathspec did not match any names, which could indicate that the | |
364 | * user mistyped the nth pathspec. | |
365 | */ | |
366 | static int do_match_pathspec(const struct pathspec *ps, | |
367 | const char *name, int namelen, | |
368 | int prefix, char *seen, | |
369 | unsigned flags) | |
370 | { | |
371 | int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE; | |
372 | ||
373 | GUARD_PATHSPEC(ps, | |
374 | PATHSPEC_FROMTOP | | |
375 | PATHSPEC_MAXDEPTH | | |
376 | PATHSPEC_LITERAL | | |
377 | PATHSPEC_GLOB | | |
378 | PATHSPEC_ICASE | | |
379 | PATHSPEC_EXCLUDE | | |
380 | PATHSPEC_ATTR); | |
381 | ||
382 | if (!ps->nr) { | |
383 | if (!ps->recursive || | |
384 | !(ps->magic & PATHSPEC_MAXDEPTH) || | |
385 | ps->max_depth == -1) | |
386 | return MATCHED_RECURSIVELY; | |
387 | ||
388 | if (within_depth(name, namelen, 0, ps->max_depth)) | |
389 | return MATCHED_EXACTLY; | |
390 | else | |
391 | return 0; | |
392 | } | |
393 | ||
394 | name += prefix; | |
395 | namelen -= prefix; | |
396 | ||
397 | for (i = ps->nr - 1; i >= 0; i--) { | |
398 | int how; | |
399 | ||
400 | if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) || | |
401 | ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE))) | |
402 | continue; | |
403 | ||
404 | if (seen && seen[i] == MATCHED_EXACTLY) | |
405 | continue; | |
406 | /* | |
407 | * Make exclude patterns optional and never report | |
408 | * "pathspec ':(exclude)foo' matches no files" | |
409 | */ | |
410 | if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE) | |
411 | seen[i] = MATCHED_FNMATCH; | |
412 | how = match_pathspec_item(ps->items+i, prefix, name, | |
413 | namelen, flags); | |
414 | if (ps->recursive && | |
415 | (ps->magic & PATHSPEC_MAXDEPTH) && | |
416 | ps->max_depth != -1 && | |
417 | how && how != MATCHED_FNMATCH) { | |
418 | int len = ps->items[i].len; | |
419 | if (name[len] == '/') | |
420 | len++; | |
421 | if (within_depth(name+len, namelen-len, 0, ps->max_depth)) | |
422 | how = MATCHED_EXACTLY; | |
423 | else | |
424 | how = 0; | |
425 | } | |
426 | if (how) { | |
427 | if (retval < how) | |
428 | retval = how; | |
429 | if (seen && seen[i] < how) | |
430 | seen[i] = how; | |
431 | } | |
432 | } | |
433 | return retval; | |
434 | } | |
435 | ||
436 | int match_pathspec(const struct pathspec *ps, | |
437 | const char *name, int namelen, | |
438 | int prefix, char *seen, int is_dir) | |
439 | { | |
440 | int positive, negative; | |
441 | unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0; | |
442 | positive = do_match_pathspec(ps, name, namelen, | |
443 | prefix, seen, flags); | |
444 | if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive) | |
445 | return positive; | |
446 | negative = do_match_pathspec(ps, name, namelen, | |
447 | prefix, seen, | |
448 | flags | DO_MATCH_EXCLUDE); | |
449 | return negative ? 0 : positive; | |
450 | } | |
451 | ||
452 | /** | |
453 | * Check if a submodule is a superset of the pathspec | |
454 | */ | |
455 | int submodule_path_match(const struct pathspec *ps, | |
456 | const char *submodule_name, | |
457 | char *seen) | |
458 | { | |
459 | int matched = do_match_pathspec(ps, submodule_name, | |
460 | strlen(submodule_name), | |
461 | 0, seen, | |
462 | DO_MATCH_DIRECTORY | | |
463 | DO_MATCH_SUBMODULE); | |
464 | return matched; | |
465 | } | |
466 | ||
467 | int report_path_error(const char *ps_matched, | |
468 | const struct pathspec *pathspec, | |
469 | const char *prefix) | |
470 | { | |
471 | /* | |
472 | * Make sure all pathspec matched; otherwise it is an error. | |
473 | */ | |
474 | int num, errors = 0; | |
475 | for (num = 0; num < pathspec->nr; num++) { | |
476 | int other, found_dup; | |
477 | ||
478 | if (ps_matched[num]) | |
479 | continue; | |
480 | /* | |
481 | * The caller might have fed identical pathspec | |
482 | * twice. Do not barf on such a mistake. | |
483 | * FIXME: parse_pathspec should have eliminated | |
484 | * duplicate pathspec. | |
485 | */ | |
486 | for (found_dup = other = 0; | |
487 | !found_dup && other < pathspec->nr; | |
488 | other++) { | |
489 | if (other == num || !ps_matched[other]) | |
490 | continue; | |
491 | if (!strcmp(pathspec->items[other].original, | |
492 | pathspec->items[num].original)) | |
493 | /* | |
494 | * Ok, we have a match already. | |
495 | */ | |
496 | found_dup = 1; | |
497 | } | |
498 | if (found_dup) | |
499 | continue; | |
500 | ||
501 | error("pathspec '%s' did not match any file(s) known to git.", | |
502 | pathspec->items[num].original); | |
503 | errors++; | |
504 | } | |
505 | return errors; | |
506 | } | |
507 | ||
508 | /* | |
509 | * Return the length of the "simple" part of a path match limiter. | |
510 | */ | |
511 | int simple_length(const char *match) | |
512 | { | |
513 | int len = -1; | |
514 | ||
515 | for (;;) { | |
516 | unsigned char c = *match++; | |
517 | len++; | |
518 | if (c == '\0' || is_glob_special(c)) | |
519 | return len; | |
520 | } | |
521 | } | |
522 | ||
523 | int no_wildcard(const char *string) | |
524 | { | |
525 | return string[simple_length(string)] == '\0'; | |
526 | } | |
527 | ||
528 | void parse_exclude_pattern(const char **pattern, | |
529 | int *patternlen, | |
530 | unsigned *flags, | |
531 | int *nowildcardlen) | |
532 | { | |
533 | const char *p = *pattern; | |
534 | size_t i, len; | |
535 | ||
536 | *flags = 0; | |
537 | if (*p == '!') { | |
538 | *flags |= EXC_FLAG_NEGATIVE; | |
539 | p++; | |
540 | } | |
541 | len = strlen(p); | |
542 | if (len && p[len - 1] == '/') { | |
543 | len--; | |
544 | *flags |= EXC_FLAG_MUSTBEDIR; | |
545 | } | |
546 | for (i = 0; i < len; i++) { | |
547 | if (p[i] == '/') | |
548 | break; | |
549 | } | |
550 | if (i == len) | |
551 | *flags |= EXC_FLAG_NODIR; | |
552 | *nowildcardlen = simple_length(p); | |
553 | /* | |
554 | * we should have excluded the trailing slash from 'p' too, | |
555 | * but that's one more allocation. Instead just make sure | |
556 | * nowildcardlen does not exceed real patternlen | |
557 | */ | |
558 | if (*nowildcardlen > len) | |
559 | *nowildcardlen = len; | |
560 | if (*p == '*' && no_wildcard(p + 1)) | |
561 | *flags |= EXC_FLAG_ENDSWITH; | |
562 | *pattern = p; | |
563 | *patternlen = len; | |
564 | } | |
565 | ||
566 | void add_exclude(const char *string, const char *base, | |
567 | int baselen, struct exclude_list *el, int srcpos) | |
568 | { | |
569 | struct exclude *x; | |
570 | int patternlen; | |
571 | unsigned flags; | |
572 | int nowildcardlen; | |
573 | ||
574 | parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen); | |
575 | if (flags & EXC_FLAG_MUSTBEDIR) { | |
576 | FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen); | |
577 | } else { | |
578 | x = xmalloc(sizeof(*x)); | |
579 | x->pattern = string; | |
580 | } | |
581 | x->patternlen = patternlen; | |
582 | x->nowildcardlen = nowildcardlen; | |
583 | x->base = base; | |
584 | x->baselen = baselen; | |
585 | x->flags = flags; | |
586 | x->srcpos = srcpos; | |
587 | ALLOC_GROW(el->excludes, el->nr + 1, el->alloc); | |
588 | el->excludes[el->nr++] = x; | |
589 | x->el = el; | |
590 | } | |
591 | ||
592 | static void *read_skip_worktree_file_from_index(const struct index_state *istate, | |
593 | const char *path, size_t *size, | |
594 | struct sha1_stat *sha1_stat) | |
595 | { | |
596 | int pos, len; | |
597 | unsigned long sz; | |
598 | enum object_type type; | |
599 | void *data; | |
600 | ||
601 | len = strlen(path); | |
602 | pos = index_name_pos(istate, path, len); | |
603 | if (pos < 0) | |
604 | return NULL; | |
605 | if (!ce_skip_worktree(istate->cache[pos])) | |
606 | return NULL; | |
607 | data = read_sha1_file(istate->cache[pos]->oid.hash, &type, &sz); | |
608 | if (!data || type != OBJ_BLOB) { | |
609 | free(data); | |
610 | return NULL; | |
611 | } | |
612 | *size = xsize_t(sz); | |
613 | if (sha1_stat) { | |
614 | memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat)); | |
615 | hashcpy(sha1_stat->sha1, istate->cache[pos]->oid.hash); | |
616 | } | |
617 | return data; | |
618 | } | |
619 | ||
620 | /* | |
621 | * Frees memory within el which was allocated for exclude patterns and | |
622 | * the file buffer. Does not free el itself. | |
623 | */ | |
624 | void clear_exclude_list(struct exclude_list *el) | |
625 | { | |
626 | int i; | |
627 | ||
628 | for (i = 0; i < el->nr; i++) | |
629 | free(el->excludes[i]); | |
630 | free(el->excludes); | |
631 | free(el->filebuf); | |
632 | ||
633 | memset(el, 0, sizeof(*el)); | |
634 | } | |
635 | ||
636 | static void trim_trailing_spaces(char *buf) | |
637 | { | |
638 | char *p, *last_space = NULL; | |
639 | ||
640 | for (p = buf; *p; p++) | |
641 | switch (*p) { | |
642 | case ' ': | |
643 | if (!last_space) | |
644 | last_space = p; | |
645 | break; | |
646 | case '\\': | |
647 | p++; | |
648 | if (!*p) | |
649 | return; | |
650 | /* fallthrough */ | |
651 | default: | |
652 | last_space = NULL; | |
653 | } | |
654 | ||
655 | if (last_space) | |
656 | *last_space = '\0'; | |
657 | } | |
658 | ||
659 | /* | |
660 | * Given a subdirectory name and "dir" of the current directory, | |
661 | * search the subdir in "dir" and return it, or create a new one if it | |
662 | * does not exist in "dir". | |
663 | * | |
664 | * If "name" has the trailing slash, it'll be excluded in the search. | |
665 | */ | |
666 | static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc, | |
667 | struct untracked_cache_dir *dir, | |
668 | const char *name, int len) | |
669 | { | |
670 | int first, last; | |
671 | struct untracked_cache_dir *d; | |
672 | if (!dir) | |
673 | return NULL; | |
674 | if (len && name[len - 1] == '/') | |
675 | len--; | |
676 | first = 0; | |
677 | last = dir->dirs_nr; | |
678 | while (last > first) { | |
679 | int cmp, next = (last + first) >> 1; | |
680 | d = dir->dirs[next]; | |
681 | cmp = strncmp(name, d->name, len); | |
682 | if (!cmp && strlen(d->name) > len) | |
683 | cmp = -1; | |
684 | if (!cmp) | |
685 | return d; | |
686 | if (cmp < 0) { | |
687 | last = next; | |
688 | continue; | |
689 | } | |
690 | first = next+1; | |
691 | } | |
692 | ||
693 | uc->dir_created++; | |
694 | FLEX_ALLOC_MEM(d, name, name, len); | |
695 | ||
696 | ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc); | |
697 | memmove(dir->dirs + first + 1, dir->dirs + first, | |
698 | (dir->dirs_nr - first) * sizeof(*dir->dirs)); | |
699 | dir->dirs_nr++; | |
700 | dir->dirs[first] = d; | |
701 | return d; | |
702 | } | |
703 | ||
704 | static void do_invalidate_gitignore(struct untracked_cache_dir *dir) | |
705 | { | |
706 | int i; | |
707 | dir->valid = 0; | |
708 | dir->untracked_nr = 0; | |
709 | for (i = 0; i < dir->dirs_nr; i++) | |
710 | do_invalidate_gitignore(dir->dirs[i]); | |
711 | } | |
712 | ||
713 | static void invalidate_gitignore(struct untracked_cache *uc, | |
714 | struct untracked_cache_dir *dir) | |
715 | { | |
716 | uc->gitignore_invalidated++; | |
717 | do_invalidate_gitignore(dir); | |
718 | } | |
719 | ||
720 | static void invalidate_directory(struct untracked_cache *uc, | |
721 | struct untracked_cache_dir *dir) | |
722 | { | |
723 | int i; | |
724 | uc->dir_invalidated++; | |
725 | dir->valid = 0; | |
726 | dir->untracked_nr = 0; | |
727 | for (i = 0; i < dir->dirs_nr; i++) | |
728 | dir->dirs[i]->recurse = 0; | |
729 | } | |
730 | ||
731 | /* | |
732 | * Given a file with name "fname", read it (either from disk, or from | |
733 | * an index if 'istate' is non-null), parse it and store the | |
734 | * exclude rules in "el". | |
735 | * | |
736 | * If "ss" is not NULL, compute SHA-1 of the exclude file and fill | |
737 | * stat data from disk (only valid if add_excludes returns zero). If | |
738 | * ss_valid is non-zero, "ss" must contain good value as input. | |
739 | */ | |
740 | static int add_excludes(const char *fname, const char *base, int baselen, | |
741 | struct exclude_list *el, | |
742 | struct index_state *istate, | |
743 | struct sha1_stat *sha1_stat) | |
744 | { | |
745 | struct stat st; | |
746 | int fd, i, lineno = 1; | |
747 | size_t size = 0; | |
748 | char *buf, *entry; | |
749 | ||
750 | fd = open(fname, O_RDONLY); | |
751 | if (fd < 0 || fstat(fd, &st) < 0) { | |
752 | if (errno != ENOENT) | |
753 | warn_on_inaccessible(fname); | |
754 | if (0 <= fd) | |
755 | close(fd); | |
756 | if (!istate || | |
757 | (buf = read_skip_worktree_file_from_index(istate, fname, &size, sha1_stat)) == NULL) | |
758 | return -1; | |
759 | if (size == 0) { | |
760 | free(buf); | |
761 | return 0; | |
762 | } | |
763 | if (buf[size-1] != '\n') { | |
764 | buf = xrealloc(buf, st_add(size, 1)); | |
765 | buf[size++] = '\n'; | |
766 | } | |
767 | } else { | |
768 | size = xsize_t(st.st_size); | |
769 | if (size == 0) { | |
770 | if (sha1_stat) { | |
771 | fill_stat_data(&sha1_stat->stat, &st); | |
772 | hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN); | |
773 | sha1_stat->valid = 1; | |
774 | } | |
775 | close(fd); | |
776 | return 0; | |
777 | } | |
778 | buf = xmallocz(size); | |
779 | if (read_in_full(fd, buf, size) != size) { | |
780 | free(buf); | |
781 | close(fd); | |
782 | return -1; | |
783 | } | |
784 | buf[size++] = '\n'; | |
785 | close(fd); | |
786 | if (sha1_stat) { | |
787 | int pos; | |
788 | if (sha1_stat->valid && | |
789 | !match_stat_data_racy(istate, &sha1_stat->stat, &st)) | |
790 | ; /* no content change, ss->sha1 still good */ | |
791 | else if (istate && | |
792 | (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 && | |
793 | !ce_stage(istate->cache[pos]) && | |
794 | ce_uptodate(istate->cache[pos]) && | |
795 | !would_convert_to_git(fname)) | |
796 | hashcpy(sha1_stat->sha1, | |
797 | istate->cache[pos]->oid.hash); | |
798 | else | |
799 | hash_sha1_file(buf, size, "blob", sha1_stat->sha1); | |
800 | fill_stat_data(&sha1_stat->stat, &st); | |
801 | sha1_stat->valid = 1; | |
802 | } | |
803 | } | |
804 | ||
805 | el->filebuf = buf; | |
806 | ||
807 | if (skip_utf8_bom(&buf, size)) | |
808 | size -= buf - el->filebuf; | |
809 | ||
810 | entry = buf; | |
811 | ||
812 | for (i = 0; i < size; i++) { | |
813 | if (buf[i] == '\n') { | |
814 | if (entry != buf + i && entry[0] != '#') { | |
815 | buf[i - (i && buf[i-1] == '\r')] = 0; | |
816 | trim_trailing_spaces(entry); | |
817 | add_exclude(entry, base, baselen, el, lineno); | |
818 | } | |
819 | lineno++; | |
820 | entry = buf + i + 1; | |
821 | } | |
822 | } | |
823 | return 0; | |
824 | } | |
825 | ||
826 | int add_excludes_from_file_to_list(const char *fname, const char *base, | |
827 | int baselen, struct exclude_list *el, | |
828 | struct index_state *istate) | |
829 | { | |
830 | return add_excludes(fname, base, baselen, el, istate, NULL); | |
831 | } | |
832 | ||
833 | struct exclude_list *add_exclude_list(struct dir_struct *dir, | |
834 | int group_type, const char *src) | |
835 | { | |
836 | struct exclude_list *el; | |
837 | struct exclude_list_group *group; | |
838 | ||
839 | group = &dir->exclude_list_group[group_type]; | |
840 | ALLOC_GROW(group->el, group->nr + 1, group->alloc); | |
841 | el = &group->el[group->nr++]; | |
842 | memset(el, 0, sizeof(*el)); | |
843 | el->src = src; | |
844 | return el; | |
845 | } | |
846 | ||
847 | /* | |
848 | * Used to set up core.excludesfile and .git/info/exclude lists. | |
849 | */ | |
850 | static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname, | |
851 | struct sha1_stat *sha1_stat) | |
852 | { | |
853 | struct exclude_list *el; | |
854 | /* | |
855 | * catch setup_standard_excludes() that's called before | |
856 | * dir->untracked is assigned. That function behaves | |
857 | * differently when dir->untracked is non-NULL. | |
858 | */ | |
859 | if (!dir->untracked) | |
860 | dir->unmanaged_exclude_files++; | |
861 | el = add_exclude_list(dir, EXC_FILE, fname); | |
862 | if (add_excludes(fname, "", 0, el, NULL, sha1_stat) < 0) | |
863 | die("cannot use %s as an exclude file", fname); | |
864 | } | |
865 | ||
866 | void add_excludes_from_file(struct dir_struct *dir, const char *fname) | |
867 | { | |
868 | dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */ | |
869 | add_excludes_from_file_1(dir, fname, NULL); | |
870 | } | |
871 | ||
872 | int match_basename(const char *basename, int basenamelen, | |
873 | const char *pattern, int prefix, int patternlen, | |
874 | unsigned flags) | |
875 | { | |
876 | if (prefix == patternlen) { | |
877 | if (patternlen == basenamelen && | |
878 | !fspathncmp(pattern, basename, basenamelen)) | |
879 | return 1; | |
880 | } else if (flags & EXC_FLAG_ENDSWITH) { | |
881 | /* "*literal" matching against "fooliteral" */ | |
882 | if (patternlen - 1 <= basenamelen && | |
883 | !fspathncmp(pattern + 1, | |
884 | basename + basenamelen - (patternlen - 1), | |
885 | patternlen - 1)) | |
886 | return 1; | |
887 | } else { | |
888 | if (fnmatch_icase_mem(pattern, patternlen, | |
889 | basename, basenamelen, | |
890 | 0) == 0) | |
891 | return 1; | |
892 | } | |
893 | return 0; | |
894 | } | |
895 | ||
896 | int match_pathname(const char *pathname, int pathlen, | |
897 | const char *base, int baselen, | |
898 | const char *pattern, int prefix, int patternlen, | |
899 | unsigned flags) | |
900 | { | |
901 | const char *name; | |
902 | int namelen; | |
903 | ||
904 | /* | |
905 | * match with FNM_PATHNAME; the pattern has base implicitly | |
906 | * in front of it. | |
907 | */ | |
908 | if (*pattern == '/') { | |
909 | pattern++; | |
910 | patternlen--; | |
911 | prefix--; | |
912 | } | |
913 | ||
914 | /* | |
915 | * baselen does not count the trailing slash. base[] may or | |
916 | * may not end with a trailing slash though. | |
917 | */ | |
918 | if (pathlen < baselen + 1 || | |
919 | (baselen && pathname[baselen] != '/') || | |
920 | fspathncmp(pathname, base, baselen)) | |
921 | return 0; | |
922 | ||
923 | namelen = baselen ? pathlen - baselen - 1 : pathlen; | |
924 | name = pathname + pathlen - namelen; | |
925 | ||
926 | if (prefix) { | |
927 | /* | |
928 | * if the non-wildcard part is longer than the | |
929 | * remaining pathname, surely it cannot match. | |
930 | */ | |
931 | if (prefix > namelen) | |
932 | return 0; | |
933 | ||
934 | if (fspathncmp(pattern, name, prefix)) | |
935 | return 0; | |
936 | pattern += prefix; | |
937 | patternlen -= prefix; | |
938 | name += prefix; | |
939 | namelen -= prefix; | |
940 | ||
941 | /* | |
942 | * If the whole pattern did not have a wildcard, | |
943 | * then our prefix match is all we need; we | |
944 | * do not need to call fnmatch at all. | |
945 | */ | |
946 | if (!patternlen && !namelen) | |
947 | return 1; | |
948 | } | |
949 | ||
950 | return fnmatch_icase_mem(pattern, patternlen, | |
951 | name, namelen, | |
952 | WM_PATHNAME) == 0; | |
953 | } | |
954 | ||
955 | /* | |
956 | * Scan the given exclude list in reverse to see whether pathname | |
957 | * should be ignored. The first match (i.e. the last on the list), if | |
958 | * any, determines the fate. Returns the exclude_list element which | |
959 | * matched, or NULL for undecided. | |
960 | */ | |
961 | static struct exclude *last_exclude_matching_from_list(const char *pathname, | |
962 | int pathlen, | |
963 | const char *basename, | |
964 | int *dtype, | |
965 | struct exclude_list *el, | |
966 | struct index_state *istate) | |
967 | { | |
968 | struct exclude *exc = NULL; /* undecided */ | |
969 | int i; | |
970 | ||
971 | if (!el->nr) | |
972 | return NULL; /* undefined */ | |
973 | ||
974 | for (i = el->nr - 1; 0 <= i; i--) { | |
975 | struct exclude *x = el->excludes[i]; | |
976 | const char *exclude = x->pattern; | |
977 | int prefix = x->nowildcardlen; | |
978 | ||
979 | if (x->flags & EXC_FLAG_MUSTBEDIR) { | |
980 | if (*dtype == DT_UNKNOWN) | |
981 | *dtype = get_dtype(NULL, istate, pathname, pathlen); | |
982 | if (*dtype != DT_DIR) | |
983 | continue; | |
984 | } | |
985 | ||
986 | if (x->flags & EXC_FLAG_NODIR) { | |
987 | if (match_basename(basename, | |
988 | pathlen - (basename - pathname), | |
989 | exclude, prefix, x->patternlen, | |
990 | x->flags)) { | |
991 | exc = x; | |
992 | break; | |
993 | } | |
994 | continue; | |
995 | } | |
996 | ||
997 | assert(x->baselen == 0 || x->base[x->baselen - 1] == '/'); | |
998 | if (match_pathname(pathname, pathlen, | |
999 | x->base, x->baselen ? x->baselen - 1 : 0, | |
1000 | exclude, prefix, x->patternlen, x->flags)) { | |
1001 | exc = x; | |
1002 | break; | |
1003 | } | |
1004 | } | |
1005 | return exc; | |
1006 | } | |
1007 | ||
1008 | /* | |
1009 | * Scan the list and let the last match determine the fate. | |
1010 | * Return 1 for exclude, 0 for include and -1 for undecided. | |
1011 | */ | |
1012 | int is_excluded_from_list(const char *pathname, | |
1013 | int pathlen, const char *basename, int *dtype, | |
1014 | struct exclude_list *el, struct index_state *istate) | |
1015 | { | |
1016 | struct exclude *exclude; | |
1017 | exclude = last_exclude_matching_from_list(pathname, pathlen, basename, | |
1018 | dtype, el, istate); | |
1019 | if (exclude) | |
1020 | return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1; | |
1021 | return -1; /* undecided */ | |
1022 | } | |
1023 | ||
1024 | static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir, | |
1025 | struct index_state *istate, | |
1026 | const char *pathname, int pathlen, const char *basename, | |
1027 | int *dtype_p) | |
1028 | { | |
1029 | int i, j; | |
1030 | struct exclude_list_group *group; | |
1031 | struct exclude *exclude; | |
1032 | for (i = EXC_CMDL; i <= EXC_FILE; i++) { | |
1033 | group = &dir->exclude_list_group[i]; | |
1034 | for (j = group->nr - 1; j >= 0; j--) { | |
1035 | exclude = last_exclude_matching_from_list( | |
1036 | pathname, pathlen, basename, dtype_p, | |
1037 | &group->el[j], istate); | |
1038 | if (exclude) | |
1039 | return exclude; | |
1040 | } | |
1041 | } | |
1042 | return NULL; | |
1043 | } | |
1044 | ||
1045 | /* | |
1046 | * Loads the per-directory exclude list for the substring of base | |
1047 | * which has a char length of baselen. | |
1048 | */ | |
1049 | static void prep_exclude(struct dir_struct *dir, | |
1050 | struct index_state *istate, | |
1051 | const char *base, int baselen) | |
1052 | { | |
1053 | struct exclude_list_group *group; | |
1054 | struct exclude_list *el; | |
1055 | struct exclude_stack *stk = NULL; | |
1056 | struct untracked_cache_dir *untracked; | |
1057 | int current; | |
1058 | ||
1059 | group = &dir->exclude_list_group[EXC_DIRS]; | |
1060 | ||
1061 | /* | |
1062 | * Pop the exclude lists from the EXCL_DIRS exclude_list_group | |
1063 | * which originate from directories not in the prefix of the | |
1064 | * path being checked. | |
1065 | */ | |
1066 | while ((stk = dir->exclude_stack) != NULL) { | |
1067 | if (stk->baselen <= baselen && | |
1068 | !strncmp(dir->basebuf.buf, base, stk->baselen)) | |
1069 | break; | |
1070 | el = &group->el[dir->exclude_stack->exclude_ix]; | |
1071 | dir->exclude_stack = stk->prev; | |
1072 | dir->exclude = NULL; | |
1073 | free((char *)el->src); /* see strbuf_detach() below */ | |
1074 | clear_exclude_list(el); | |
1075 | free(stk); | |
1076 | group->nr--; | |
1077 | } | |
1078 | ||
1079 | /* Skip traversing into sub directories if the parent is excluded */ | |
1080 | if (dir->exclude) | |
1081 | return; | |
1082 | ||
1083 | /* | |
1084 | * Lazy initialization. All call sites currently just | |
1085 | * memset(dir, 0, sizeof(*dir)) before use. Changing all of | |
1086 | * them seems lots of work for little benefit. | |
1087 | */ | |
1088 | if (!dir->basebuf.buf) | |
1089 | strbuf_init(&dir->basebuf, PATH_MAX); | |
1090 | ||
1091 | /* Read from the parent directories and push them down. */ | |
1092 | current = stk ? stk->baselen : -1; | |
1093 | strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current); | |
1094 | if (dir->untracked) | |
1095 | untracked = stk ? stk->ucd : dir->untracked->root; | |
1096 | else | |
1097 | untracked = NULL; | |
1098 | ||
1099 | while (current < baselen) { | |
1100 | const char *cp; | |
1101 | struct sha1_stat sha1_stat; | |
1102 | ||
1103 | stk = xcalloc(1, sizeof(*stk)); | |
1104 | if (current < 0) { | |
1105 | cp = base; | |
1106 | current = 0; | |
1107 | } else { | |
1108 | cp = strchr(base + current + 1, '/'); | |
1109 | if (!cp) | |
1110 | die("oops in prep_exclude"); | |
1111 | cp++; | |
1112 | untracked = | |
1113 | lookup_untracked(dir->untracked, untracked, | |
1114 | base + current, | |
1115 | cp - base - current); | |
1116 | } | |
1117 | stk->prev = dir->exclude_stack; | |
1118 | stk->baselen = cp - base; | |
1119 | stk->exclude_ix = group->nr; | |
1120 | stk->ucd = untracked; | |
1121 | el = add_exclude_list(dir, EXC_DIRS, NULL); | |
1122 | strbuf_add(&dir->basebuf, base + current, stk->baselen - current); | |
1123 | assert(stk->baselen == dir->basebuf.len); | |
1124 | ||
1125 | /* Abort if the directory is excluded */ | |
1126 | if (stk->baselen) { | |
1127 | int dt = DT_DIR; | |
1128 | dir->basebuf.buf[stk->baselen - 1] = 0; | |
1129 | dir->exclude = last_exclude_matching_from_lists(dir, | |
1130 | istate, | |
1131 | dir->basebuf.buf, stk->baselen - 1, | |
1132 | dir->basebuf.buf + current, &dt); | |
1133 | dir->basebuf.buf[stk->baselen - 1] = '/'; | |
1134 | if (dir->exclude && | |
1135 | dir->exclude->flags & EXC_FLAG_NEGATIVE) | |
1136 | dir->exclude = NULL; | |
1137 | if (dir->exclude) { | |
1138 | dir->exclude_stack = stk; | |
1139 | return; | |
1140 | } | |
1141 | } | |
1142 | ||
1143 | /* Try to read per-directory file */ | |
1144 | hashclr(sha1_stat.sha1); | |
1145 | sha1_stat.valid = 0; | |
1146 | if (dir->exclude_per_dir && | |
1147 | /* | |
1148 | * If we know that no files have been added in | |
1149 | * this directory (i.e. valid_cached_dir() has | |
1150 | * been executed and set untracked->valid) .. | |
1151 | */ | |
1152 | (!untracked || !untracked->valid || | |
1153 | /* | |
1154 | * .. and .gitignore does not exist before | |
1155 | * (i.e. null exclude_sha1). Then we can skip | |
1156 | * loading .gitignore, which would result in | |
1157 | * ENOENT anyway. | |
1158 | */ | |
1159 | !is_null_sha1(untracked->exclude_sha1))) { | |
1160 | /* | |
1161 | * dir->basebuf gets reused by the traversal, but we | |
1162 | * need fname to remain unchanged to ensure the src | |
1163 | * member of each struct exclude correctly | |
1164 | * back-references its source file. Other invocations | |
1165 | * of add_exclude_list provide stable strings, so we | |
1166 | * strbuf_detach() and free() here in the caller. | |
1167 | */ | |
1168 | struct strbuf sb = STRBUF_INIT; | |
1169 | strbuf_addbuf(&sb, &dir->basebuf); | |
1170 | strbuf_addstr(&sb, dir->exclude_per_dir); | |
1171 | el->src = strbuf_detach(&sb, NULL); | |
1172 | add_excludes(el->src, el->src, stk->baselen, el, istate, | |
1173 | untracked ? &sha1_stat : NULL); | |
1174 | } | |
1175 | /* | |
1176 | * NEEDSWORK: when untracked cache is enabled, prep_exclude() | |
1177 | * will first be called in valid_cached_dir() then maybe many | |
1178 | * times more in last_exclude_matching(). When the cache is | |
1179 | * used, last_exclude_matching() will not be called and | |
1180 | * reading .gitignore content will be a waste. | |
1181 | * | |
1182 | * So when it's called by valid_cached_dir() and we can get | |
1183 | * .gitignore SHA-1 from the index (i.e. .gitignore is not | |
1184 | * modified on work tree), we could delay reading the | |
1185 | * .gitignore content until we absolutely need it in | |
1186 | * last_exclude_matching(). Be careful about ignore rule | |
1187 | * order, though, if you do that. | |
1188 | */ | |
1189 | if (untracked && | |
1190 | hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) { | |
1191 | invalidate_gitignore(dir->untracked, untracked); | |
1192 | hashcpy(untracked->exclude_sha1, sha1_stat.sha1); | |
1193 | } | |
1194 | dir->exclude_stack = stk; | |
1195 | current = stk->baselen; | |
1196 | } | |
1197 | strbuf_setlen(&dir->basebuf, baselen); | |
1198 | } | |
1199 | ||
1200 | /* | |
1201 | * Loads the exclude lists for the directory containing pathname, then | |
1202 | * scans all exclude lists to determine whether pathname is excluded. | |
1203 | * Returns the exclude_list element which matched, or NULL for | |
1204 | * undecided. | |
1205 | */ | |
1206 | struct exclude *last_exclude_matching(struct dir_struct *dir, | |
1207 | struct index_state *istate, | |
1208 | const char *pathname, | |
1209 | int *dtype_p) | |
1210 | { | |
1211 | int pathlen = strlen(pathname); | |
1212 | const char *basename = strrchr(pathname, '/'); | |
1213 | basename = (basename) ? basename+1 : pathname; | |
1214 | ||
1215 | prep_exclude(dir, istate, pathname, basename-pathname); | |
1216 | ||
1217 | if (dir->exclude) | |
1218 | return dir->exclude; | |
1219 | ||
1220 | return last_exclude_matching_from_lists(dir, istate, pathname, pathlen, | |
1221 | basename, dtype_p); | |
1222 | } | |
1223 | ||
1224 | /* | |
1225 | * Loads the exclude lists for the directory containing pathname, then | |
1226 | * scans all exclude lists to determine whether pathname is excluded. | |
1227 | * Returns 1 if true, otherwise 0. | |
1228 | */ | |
1229 | int is_excluded(struct dir_struct *dir, struct index_state *istate, | |
1230 | const char *pathname, int *dtype_p) | |
1231 | { | |
1232 | struct exclude *exclude = | |
1233 | last_exclude_matching(dir, istate, pathname, dtype_p); | |
1234 | if (exclude) | |
1235 | return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1; | |
1236 | return 0; | |
1237 | } | |
1238 | ||
1239 | static struct dir_entry *dir_entry_new(const char *pathname, int len) | |
1240 | { | |
1241 | struct dir_entry *ent; | |
1242 | ||
1243 | FLEX_ALLOC_MEM(ent, name, pathname, len); | |
1244 | ent->len = len; | |
1245 | return ent; | |
1246 | } | |
1247 | ||
1248 | static struct dir_entry *dir_add_name(struct dir_struct *dir, | |
1249 | struct index_state *istate, | |
1250 | const char *pathname, int len) | |
1251 | { | |
1252 | if (index_file_exists(istate, pathname, len, ignore_case)) | |
1253 | return NULL; | |
1254 | ||
1255 | ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc); | |
1256 | return dir->entries[dir->nr++] = dir_entry_new(pathname, len); | |
1257 | } | |
1258 | ||
1259 | struct dir_entry *dir_add_ignored(struct dir_struct *dir, | |
1260 | struct index_state *istate, | |
1261 | const char *pathname, int len) | |
1262 | { | |
1263 | if (!index_name_is_other(istate, pathname, len)) | |
1264 | return NULL; | |
1265 | ||
1266 | ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc); | |
1267 | return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len); | |
1268 | } | |
1269 | ||
1270 | enum exist_status { | |
1271 | index_nonexistent = 0, | |
1272 | index_directory, | |
1273 | index_gitdir | |
1274 | }; | |
1275 | ||
1276 | /* | |
1277 | * Do not use the alphabetically sorted index to look up | |
1278 | * the directory name; instead, use the case insensitive | |
1279 | * directory hash. | |
1280 | */ | |
1281 | static enum exist_status directory_exists_in_index_icase(struct index_state *istate, | |
1282 | const char *dirname, int len) | |
1283 | { | |
1284 | struct cache_entry *ce; | |
1285 | ||
1286 | if (index_dir_exists(istate, dirname, len)) | |
1287 | return index_directory; | |
1288 | ||
1289 | ce = index_file_exists(istate, dirname, len, ignore_case); | |
1290 | if (ce && S_ISGITLINK(ce->ce_mode)) | |
1291 | return index_gitdir; | |
1292 | ||
1293 | return index_nonexistent; | |
1294 | } | |
1295 | ||
1296 | /* | |
1297 | * The index sorts alphabetically by entry name, which | |
1298 | * means that a gitlink sorts as '\0' at the end, while | |
1299 | * a directory (which is defined not as an entry, but as | |
1300 | * the files it contains) will sort with the '/' at the | |
1301 | * end. | |
1302 | */ | |
1303 | static enum exist_status directory_exists_in_index(struct index_state *istate, | |
1304 | const char *dirname, int len) | |
1305 | { | |
1306 | int pos; | |
1307 | ||
1308 | if (ignore_case) | |
1309 | return directory_exists_in_index_icase(istate, dirname, len); | |
1310 | ||
1311 | pos = index_name_pos(istate, dirname, len); | |
1312 | if (pos < 0) | |
1313 | pos = -pos-1; | |
1314 | while (pos < istate->cache_nr) { | |
1315 | const struct cache_entry *ce = istate->cache[pos++]; | |
1316 | unsigned char endchar; | |
1317 | ||
1318 | if (strncmp(ce->name, dirname, len)) | |
1319 | break; | |
1320 | endchar = ce->name[len]; | |
1321 | if (endchar > '/') | |
1322 | break; | |
1323 | if (endchar == '/') | |
1324 | return index_directory; | |
1325 | if (!endchar && S_ISGITLINK(ce->ce_mode)) | |
1326 | return index_gitdir; | |
1327 | } | |
1328 | return index_nonexistent; | |
1329 | } | |
1330 | ||
1331 | /* | |
1332 | * When we find a directory when traversing the filesystem, we | |
1333 | * have three distinct cases: | |
1334 | * | |
1335 | * - ignore it | |
1336 | * - see it as a directory | |
1337 | * - recurse into it | |
1338 | * | |
1339 | * and which one we choose depends on a combination of existing | |
1340 | * git index contents and the flags passed into the directory | |
1341 | * traversal routine. | |
1342 | * | |
1343 | * Case 1: If we *already* have entries in the index under that | |
1344 | * directory name, we always recurse into the directory to see | |
1345 | * all the files. | |
1346 | * | |
1347 | * Case 2: If we *already* have that directory name as a gitlink, | |
1348 | * we always continue to see it as a gitlink, regardless of whether | |
1349 | * there is an actual git directory there or not (it might not | |
1350 | * be checked out as a subproject!) | |
1351 | * | |
1352 | * Case 3: if we didn't have it in the index previously, we | |
1353 | * have a few sub-cases: | |
1354 | * | |
1355 | * (a) if "show_other_directories" is true, we show it as | |
1356 | * just a directory, unless "hide_empty_directories" is | |
1357 | * also true, in which case we need to check if it contains any | |
1358 | * untracked and / or ignored files. | |
1359 | * (b) if it looks like a git directory, and we don't have | |
1360 | * 'no_gitlinks' set we treat it as a gitlink, and show it | |
1361 | * as a directory. | |
1362 | * (c) otherwise, we recurse into it. | |
1363 | */ | |
1364 | static enum path_treatment treat_directory(struct dir_struct *dir, | |
1365 | struct untracked_cache_dir *untracked, | |
1366 | const char *dirname, int len, int baselen, int exclude, | |
1367 | const struct pathspec *pathspec) | |
1368 | { | |
1369 | /* The "len-1" is to strip the final '/' */ | |
1370 | switch (directory_exists_in_index(&the_index, dirname, len-1)) { | |
1371 | case index_directory: | |
1372 | return path_recurse; | |
1373 | ||
1374 | case index_gitdir: | |
1375 | return path_none; | |
1376 | ||
1377 | case index_nonexistent: | |
1378 | if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES) | |
1379 | break; | |
1380 | if (!(dir->flags & DIR_NO_GITLINKS)) { | |
1381 | unsigned char sha1[20]; | |
1382 | if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0) | |
1383 | return path_untracked; | |
1384 | } | |
1385 | return path_recurse; | |
1386 | } | |
1387 | ||
1388 | /* This is the "show_other_directories" case */ | |
1389 | ||
1390 | if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES)) | |
1391 | return exclude ? path_excluded : path_untracked; | |
1392 | ||
1393 | untracked = lookup_untracked(dir->untracked, untracked, | |
1394 | dirname + baselen, len - baselen); | |
1395 | return read_directory_recursive(dir, dirname, len, | |
1396 | untracked, 1, pathspec); | |
1397 | } | |
1398 | ||
1399 | /* | |
1400 | * This is an inexact early pruning of any recursive directory | |
1401 | * reading - if the path cannot possibly be in the pathspec, | |
1402 | * return true, and we'll skip it early. | |
1403 | */ | |
1404 | static int simplify_away(const char *path, int pathlen, | |
1405 | const struct pathspec *pathspec) | |
1406 | { | |
1407 | int i; | |
1408 | ||
1409 | if (!pathspec || !pathspec->nr) | |
1410 | return 0; | |
1411 | ||
1412 | GUARD_PATHSPEC(pathspec, | |
1413 | PATHSPEC_FROMTOP | | |
1414 | PATHSPEC_MAXDEPTH | | |
1415 | PATHSPEC_LITERAL | | |
1416 | PATHSPEC_GLOB | | |
1417 | PATHSPEC_ICASE | | |
1418 | PATHSPEC_EXCLUDE | | |
1419 | PATHSPEC_ATTR); | |
1420 | ||
1421 | for (i = 0; i < pathspec->nr; i++) { | |
1422 | const struct pathspec_item *item = &pathspec->items[i]; | |
1423 | int len = item->nowildcard_len; | |
1424 | ||
1425 | if (len > pathlen) | |
1426 | len = pathlen; | |
1427 | if (!ps_strncmp(item, item->match, path, len)) | |
1428 | return 0; | |
1429 | } | |
1430 | ||
1431 | return 1; | |
1432 | } | |
1433 | ||
1434 | /* | |
1435 | * This function tells us whether an excluded path matches a | |
1436 | * list of "interesting" pathspecs. That is, whether a path matched | |
1437 | * by any of the pathspecs could possibly be ignored by excluding | |
1438 | * the specified path. This can happen if: | |
1439 | * | |
1440 | * 1. the path is mentioned explicitly in the pathspec | |
1441 | * | |
1442 | * 2. the path is a directory prefix of some element in the | |
1443 | * pathspec | |
1444 | */ | |
1445 | static int exclude_matches_pathspec(const char *path, int pathlen, | |
1446 | const struct pathspec *pathspec) | |
1447 | { | |
1448 | int i; | |
1449 | ||
1450 | if (!pathspec || !pathspec->nr) | |
1451 | return 0; | |
1452 | ||
1453 | GUARD_PATHSPEC(pathspec, | |
1454 | PATHSPEC_FROMTOP | | |
1455 | PATHSPEC_MAXDEPTH | | |
1456 | PATHSPEC_LITERAL | | |
1457 | PATHSPEC_GLOB | | |
1458 | PATHSPEC_ICASE | | |
1459 | PATHSPEC_EXCLUDE); | |
1460 | ||
1461 | for (i = 0; i < pathspec->nr; i++) { | |
1462 | const struct pathspec_item *item = &pathspec->items[i]; | |
1463 | int len = item->nowildcard_len; | |
1464 | ||
1465 | if (len == pathlen && | |
1466 | !ps_strncmp(item, item->match, path, pathlen)) | |
1467 | return 1; | |
1468 | if (len > pathlen && | |
1469 | item->match[pathlen] == '/' && | |
1470 | !ps_strncmp(item, item->match, path, pathlen)) | |
1471 | return 1; | |
1472 | } | |
1473 | return 0; | |
1474 | } | |
1475 | ||
1476 | static int get_index_dtype(struct index_state *istate, | |
1477 | const char *path, int len) | |
1478 | { | |
1479 | int pos; | |
1480 | const struct cache_entry *ce; | |
1481 | ||
1482 | ce = index_file_exists(istate, path, len, 0); | |
1483 | if (ce) { | |
1484 | if (!ce_uptodate(ce)) | |
1485 | return DT_UNKNOWN; | |
1486 | if (S_ISGITLINK(ce->ce_mode)) | |
1487 | return DT_DIR; | |
1488 | /* | |
1489 | * Nobody actually cares about the | |
1490 | * difference between DT_LNK and DT_REG | |
1491 | */ | |
1492 | return DT_REG; | |
1493 | } | |
1494 | ||
1495 | /* Try to look it up as a directory */ | |
1496 | pos = index_name_pos(istate, path, len); | |
1497 | if (pos >= 0) | |
1498 | return DT_UNKNOWN; | |
1499 | pos = -pos-1; | |
1500 | while (pos < istate->cache_nr) { | |
1501 | ce = istate->cache[pos++]; | |
1502 | if (strncmp(ce->name, path, len)) | |
1503 | break; | |
1504 | if (ce->name[len] > '/') | |
1505 | break; | |
1506 | if (ce->name[len] < '/') | |
1507 | continue; | |
1508 | if (!ce_uptodate(ce)) | |
1509 | break; /* continue? */ | |
1510 | return DT_DIR; | |
1511 | } | |
1512 | return DT_UNKNOWN; | |
1513 | } | |
1514 | ||
1515 | static int get_dtype(struct dirent *de, struct index_state *istate, | |
1516 | const char *path, int len) | |
1517 | { | |
1518 | int dtype = de ? DTYPE(de) : DT_UNKNOWN; | |
1519 | struct stat st; | |
1520 | ||
1521 | if (dtype != DT_UNKNOWN) | |
1522 | return dtype; | |
1523 | dtype = get_index_dtype(istate, path, len); | |
1524 | if (dtype != DT_UNKNOWN) | |
1525 | return dtype; | |
1526 | if (lstat(path, &st)) | |
1527 | return dtype; | |
1528 | if (S_ISREG(st.st_mode)) | |
1529 | return DT_REG; | |
1530 | if (S_ISDIR(st.st_mode)) | |
1531 | return DT_DIR; | |
1532 | if (S_ISLNK(st.st_mode)) | |
1533 | return DT_LNK; | |
1534 | return dtype; | |
1535 | } | |
1536 | ||
1537 | static enum path_treatment treat_one_path(struct dir_struct *dir, | |
1538 | struct untracked_cache_dir *untracked, | |
1539 | struct strbuf *path, | |
1540 | int baselen, | |
1541 | const struct pathspec *pathspec, | |
1542 | int dtype, struct dirent *de) | |
1543 | { | |
1544 | int exclude; | |
1545 | int has_path_in_index = !!index_file_exists(&the_index, path->buf, path->len, ignore_case); | |
1546 | ||
1547 | if (dtype == DT_UNKNOWN) | |
1548 | dtype = get_dtype(de, &the_index, path->buf, path->len); | |
1549 | ||
1550 | /* Always exclude indexed files */ | |
1551 | if (dtype != DT_DIR && has_path_in_index) | |
1552 | return path_none; | |
1553 | ||
1554 | /* | |
1555 | * When we are looking at a directory P in the working tree, | |
1556 | * there are three cases: | |
1557 | * | |
1558 | * (1) P exists in the index. Everything inside the directory P in | |
1559 | * the working tree needs to go when P is checked out from the | |
1560 | * index. | |
1561 | * | |
1562 | * (2) P does not exist in the index, but there is P/Q in the index. | |
1563 | * We know P will stay a directory when we check out the contents | |
1564 | * of the index, but we do not know yet if there is a directory | |
1565 | * P/Q in the working tree to be killed, so we need to recurse. | |
1566 | * | |
1567 | * (3) P does not exist in the index, and there is no P/Q in the index | |
1568 | * to require P to be a directory, either. Only in this case, we | |
1569 | * know that everything inside P will not be killed without | |
1570 | * recursing. | |
1571 | */ | |
1572 | if ((dir->flags & DIR_COLLECT_KILLED_ONLY) && | |
1573 | (dtype == DT_DIR) && | |
1574 | !has_path_in_index && | |
1575 | (directory_exists_in_index(&the_index, path->buf, path->len) == index_nonexistent)) | |
1576 | return path_none; | |
1577 | ||
1578 | exclude = is_excluded(dir, &the_index, path->buf, &dtype); | |
1579 | ||
1580 | /* | |
1581 | * Excluded? If we don't explicitly want to show | |
1582 | * ignored files, ignore it | |
1583 | */ | |
1584 | if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO))) | |
1585 | return path_excluded; | |
1586 | ||
1587 | switch (dtype) { | |
1588 | default: | |
1589 | return path_none; | |
1590 | case DT_DIR: | |
1591 | strbuf_addch(path, '/'); | |
1592 | return treat_directory(dir, untracked, path->buf, path->len, | |
1593 | baselen, exclude, pathspec); | |
1594 | case DT_REG: | |
1595 | case DT_LNK: | |
1596 | return exclude ? path_excluded : path_untracked; | |
1597 | } | |
1598 | } | |
1599 | ||
1600 | static enum path_treatment treat_path_fast(struct dir_struct *dir, | |
1601 | struct untracked_cache_dir *untracked, | |
1602 | struct cached_dir *cdir, | |
1603 | struct strbuf *path, | |
1604 | int baselen, | |
1605 | const struct pathspec *pathspec) | |
1606 | { | |
1607 | strbuf_setlen(path, baselen); | |
1608 | if (!cdir->ucd) { | |
1609 | strbuf_addstr(path, cdir->file); | |
1610 | return path_untracked; | |
1611 | } | |
1612 | strbuf_addstr(path, cdir->ucd->name); | |
1613 | /* treat_one_path() does this before it calls treat_directory() */ | |
1614 | strbuf_complete(path, '/'); | |
1615 | if (cdir->ucd->check_only) | |
1616 | /* | |
1617 | * check_only is set as a result of treat_directory() getting | |
1618 | * to its bottom. Verify again the same set of directories | |
1619 | * with check_only set. | |
1620 | */ | |
1621 | return read_directory_recursive(dir, path->buf, path->len, | |
1622 | cdir->ucd, 1, pathspec); | |
1623 | /* | |
1624 | * We get path_recurse in the first run when | |
1625 | * directory_exists_in_index() returns index_nonexistent. We | |
1626 | * are sure that new changes in the index does not impact the | |
1627 | * outcome. Return now. | |
1628 | */ | |
1629 | return path_recurse; | |
1630 | } | |
1631 | ||
1632 | static enum path_treatment treat_path(struct dir_struct *dir, | |
1633 | struct untracked_cache_dir *untracked, | |
1634 | struct cached_dir *cdir, | |
1635 | struct strbuf *path, | |
1636 | int baselen, | |
1637 | const struct pathspec *pathspec) | |
1638 | { | |
1639 | int dtype; | |
1640 | struct dirent *de = cdir->de; | |
1641 | ||
1642 | if (!de) | |
1643 | return treat_path_fast(dir, untracked, cdir, path, | |
1644 | baselen, pathspec); | |
1645 | if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git")) | |
1646 | return path_none; | |
1647 | strbuf_setlen(path, baselen); | |
1648 | strbuf_addstr(path, de->d_name); | |
1649 | if (simplify_away(path->buf, path->len, pathspec)) | |
1650 | return path_none; | |
1651 | ||
1652 | dtype = DTYPE(de); | |
1653 | return treat_one_path(dir, untracked, path, baselen, pathspec, dtype, de); | |
1654 | } | |
1655 | ||
1656 | static void add_untracked(struct untracked_cache_dir *dir, const char *name) | |
1657 | { | |
1658 | if (!dir) | |
1659 | return; | |
1660 | ALLOC_GROW(dir->untracked, dir->untracked_nr + 1, | |
1661 | dir->untracked_alloc); | |
1662 | dir->untracked[dir->untracked_nr++] = xstrdup(name); | |
1663 | } | |
1664 | ||
1665 | static int valid_cached_dir(struct dir_struct *dir, | |
1666 | struct untracked_cache_dir *untracked, | |
1667 | struct strbuf *path, | |
1668 | int check_only) | |
1669 | { | |
1670 | struct stat st; | |
1671 | ||
1672 | if (!untracked) | |
1673 | return 0; | |
1674 | ||
1675 | if (stat(path->len ? path->buf : ".", &st)) { | |
1676 | invalidate_directory(dir->untracked, untracked); | |
1677 | memset(&untracked->stat_data, 0, sizeof(untracked->stat_data)); | |
1678 | return 0; | |
1679 | } | |
1680 | if (!untracked->valid || | |
1681 | match_stat_data_racy(&the_index, &untracked->stat_data, &st)) { | |
1682 | if (untracked->valid) | |
1683 | invalidate_directory(dir->untracked, untracked); | |
1684 | fill_stat_data(&untracked->stat_data, &st); | |
1685 | return 0; | |
1686 | } | |
1687 | ||
1688 | if (untracked->check_only != !!check_only) { | |
1689 | invalidate_directory(dir->untracked, untracked); | |
1690 | return 0; | |
1691 | } | |
1692 | ||
1693 | /* | |
1694 | * prep_exclude will be called eventually on this directory, | |
1695 | * but it's called much later in last_exclude_matching(). We | |
1696 | * need it now to determine the validity of the cache for this | |
1697 | * path. The next calls will be nearly no-op, the way | |
1698 | * prep_exclude() is designed. | |
1699 | */ | |
1700 | if (path->len && path->buf[path->len - 1] != '/') { | |
1701 | strbuf_addch(path, '/'); | |
1702 | prep_exclude(dir, &the_index, path->buf, path->len); | |
1703 | strbuf_setlen(path, path->len - 1); | |
1704 | } else | |
1705 | prep_exclude(dir, &the_index, path->buf, path->len); | |
1706 | ||
1707 | /* hopefully prep_exclude() haven't invalidated this entry... */ | |
1708 | return untracked->valid; | |
1709 | } | |
1710 | ||
1711 | static int open_cached_dir(struct cached_dir *cdir, | |
1712 | struct dir_struct *dir, | |
1713 | struct untracked_cache_dir *untracked, | |
1714 | struct strbuf *path, | |
1715 | int check_only) | |
1716 | { | |
1717 | memset(cdir, 0, sizeof(*cdir)); | |
1718 | cdir->untracked = untracked; | |
1719 | if (valid_cached_dir(dir, untracked, path, check_only)) | |
1720 | return 0; | |
1721 | cdir->fdir = opendir(path->len ? path->buf : "."); | |
1722 | if (dir->untracked) | |
1723 | dir->untracked->dir_opened++; | |
1724 | if (!cdir->fdir) | |
1725 | return -1; | |
1726 | return 0; | |
1727 | } | |
1728 | ||
1729 | static int read_cached_dir(struct cached_dir *cdir) | |
1730 | { | |
1731 | if (cdir->fdir) { | |
1732 | cdir->de = readdir(cdir->fdir); | |
1733 | if (!cdir->de) | |
1734 | return -1; | |
1735 | return 0; | |
1736 | } | |
1737 | while (cdir->nr_dirs < cdir->untracked->dirs_nr) { | |
1738 | struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs]; | |
1739 | if (!d->recurse) { | |
1740 | cdir->nr_dirs++; | |
1741 | continue; | |
1742 | } | |
1743 | cdir->ucd = d; | |
1744 | cdir->nr_dirs++; | |
1745 | return 0; | |
1746 | } | |
1747 | cdir->ucd = NULL; | |
1748 | if (cdir->nr_files < cdir->untracked->untracked_nr) { | |
1749 | struct untracked_cache_dir *d = cdir->untracked; | |
1750 | cdir->file = d->untracked[cdir->nr_files++]; | |
1751 | return 0; | |
1752 | } | |
1753 | return -1; | |
1754 | } | |
1755 | ||
1756 | static void close_cached_dir(struct cached_dir *cdir) | |
1757 | { | |
1758 | if (cdir->fdir) | |
1759 | closedir(cdir->fdir); | |
1760 | /* | |
1761 | * We have gone through this directory and found no untracked | |
1762 | * entries. Mark it valid. | |
1763 | */ | |
1764 | if (cdir->untracked) { | |
1765 | cdir->untracked->valid = 1; | |
1766 | cdir->untracked->recurse = 1; | |
1767 | } | |
1768 | } | |
1769 | ||
1770 | /* | |
1771 | * Read a directory tree. We currently ignore anything but | |
1772 | * directories, regular files and symlinks. That's because git | |
1773 | * doesn't handle them at all yet. Maybe that will change some | |
1774 | * day. | |
1775 | * | |
1776 | * Also, we ignore the name ".git" (even if it is not a directory). | |
1777 | * That likely will not change. | |
1778 | * | |
1779 | * Returns the most significant path_treatment value encountered in the scan. | |
1780 | */ | |
1781 | static enum path_treatment read_directory_recursive(struct dir_struct *dir, | |
1782 | const char *base, int baselen, | |
1783 | struct untracked_cache_dir *untracked, int check_only, | |
1784 | const struct pathspec *pathspec) | |
1785 | { | |
1786 | struct cached_dir cdir; | |
1787 | enum path_treatment state, subdir_state, dir_state = path_none; | |
1788 | struct strbuf path = STRBUF_INIT; | |
1789 | ||
1790 | strbuf_add(&path, base, baselen); | |
1791 | ||
1792 | if (open_cached_dir(&cdir, dir, untracked, &path, check_only)) | |
1793 | goto out; | |
1794 | ||
1795 | if (untracked) | |
1796 | untracked->check_only = !!check_only; | |
1797 | ||
1798 | while (!read_cached_dir(&cdir)) { | |
1799 | /* check how the file or directory should be treated */ | |
1800 | state = treat_path(dir, untracked, &cdir, &path, | |
1801 | baselen, pathspec); | |
1802 | ||
1803 | if (state > dir_state) | |
1804 | dir_state = state; | |
1805 | ||
1806 | /* recurse into subdir if instructed by treat_path */ | |
1807 | if (state == path_recurse) { | |
1808 | struct untracked_cache_dir *ud; | |
1809 | ud = lookup_untracked(dir->untracked, untracked, | |
1810 | path.buf + baselen, | |
1811 | path.len - baselen); | |
1812 | subdir_state = | |
1813 | read_directory_recursive(dir, path.buf, | |
1814 | path.len, ud, | |
1815 | check_only, pathspec); | |
1816 | if (subdir_state > dir_state) | |
1817 | dir_state = subdir_state; | |
1818 | } | |
1819 | ||
1820 | if (check_only) { | |
1821 | /* abort early if maximum state has been reached */ | |
1822 | if (dir_state == path_untracked) { | |
1823 | if (cdir.fdir) | |
1824 | add_untracked(untracked, path.buf + baselen); | |
1825 | break; | |
1826 | } | |
1827 | /* skip the dir_add_* part */ | |
1828 | continue; | |
1829 | } | |
1830 | ||
1831 | /* add the path to the appropriate result list */ | |
1832 | switch (state) { | |
1833 | case path_excluded: | |
1834 | if (dir->flags & DIR_SHOW_IGNORED) | |
1835 | dir_add_name(dir, &the_index, path.buf, path.len); | |
1836 | else if ((dir->flags & DIR_SHOW_IGNORED_TOO) || | |
1837 | ((dir->flags & DIR_COLLECT_IGNORED) && | |
1838 | exclude_matches_pathspec(path.buf, path.len, | |
1839 | pathspec))) | |
1840 | dir_add_ignored(dir, &the_index, path.buf, path.len); | |
1841 | break; | |
1842 | ||
1843 | case path_untracked: | |
1844 | if (dir->flags & DIR_SHOW_IGNORED) | |
1845 | break; | |
1846 | dir_add_name(dir, &the_index, path.buf, path.len); | |
1847 | if (cdir.fdir) | |
1848 | add_untracked(untracked, path.buf + baselen); | |
1849 | break; | |
1850 | ||
1851 | default: | |
1852 | break; | |
1853 | } | |
1854 | } | |
1855 | close_cached_dir(&cdir); | |
1856 | out: | |
1857 | strbuf_release(&path); | |
1858 | ||
1859 | return dir_state; | |
1860 | } | |
1861 | ||
1862 | static int cmp_name(const void *p1, const void *p2) | |
1863 | { | |
1864 | const struct dir_entry *e1 = *(const struct dir_entry **)p1; | |
1865 | const struct dir_entry *e2 = *(const struct dir_entry **)p2; | |
1866 | ||
1867 | return name_compare(e1->name, e1->len, e2->name, e2->len); | |
1868 | } | |
1869 | ||
1870 | static int treat_leading_path(struct dir_struct *dir, | |
1871 | const char *path, int len, | |
1872 | const struct pathspec *pathspec) | |
1873 | { | |
1874 | struct strbuf sb = STRBUF_INIT; | |
1875 | int baselen, rc = 0; | |
1876 | const char *cp; | |
1877 | int old_flags = dir->flags; | |
1878 | ||
1879 | while (len && path[len - 1] == '/') | |
1880 | len--; | |
1881 | if (!len) | |
1882 | return 1; | |
1883 | baselen = 0; | |
1884 | dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES; | |
1885 | while (1) { | |
1886 | cp = path + baselen + !!baselen; | |
1887 | cp = memchr(cp, '/', path + len - cp); | |
1888 | if (!cp) | |
1889 | baselen = len; | |
1890 | else | |
1891 | baselen = cp - path; | |
1892 | strbuf_setlen(&sb, 0); | |
1893 | strbuf_add(&sb, path, baselen); | |
1894 | if (!is_directory(sb.buf)) | |
1895 | break; | |
1896 | if (simplify_away(sb.buf, sb.len, pathspec)) | |
1897 | break; | |
1898 | if (treat_one_path(dir, NULL, &sb, baselen, pathspec, | |
1899 | DT_DIR, NULL) == path_none) | |
1900 | break; /* do not recurse into it */ | |
1901 | if (len <= baselen) { | |
1902 | rc = 1; | |
1903 | break; /* finished checking */ | |
1904 | } | |
1905 | } | |
1906 | strbuf_release(&sb); | |
1907 | dir->flags = old_flags; | |
1908 | return rc; | |
1909 | } | |
1910 | ||
1911 | static const char *get_ident_string(void) | |
1912 | { | |
1913 | static struct strbuf sb = STRBUF_INIT; | |
1914 | struct utsname uts; | |
1915 | ||
1916 | if (sb.len) | |
1917 | return sb.buf; | |
1918 | if (uname(&uts) < 0) | |
1919 | die_errno(_("failed to get kernel name and information")); | |
1920 | strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(), | |
1921 | uts.sysname); | |
1922 | return sb.buf; | |
1923 | } | |
1924 | ||
1925 | static int ident_in_untracked(const struct untracked_cache *uc) | |
1926 | { | |
1927 | /* | |
1928 | * Previous git versions may have saved many NUL separated | |
1929 | * strings in the "ident" field, but it is insane to manage | |
1930 | * many locations, so just take care of the first one. | |
1931 | */ | |
1932 | ||
1933 | return !strcmp(uc->ident.buf, get_ident_string()); | |
1934 | } | |
1935 | ||
1936 | static void set_untracked_ident(struct untracked_cache *uc) | |
1937 | { | |
1938 | strbuf_reset(&uc->ident); | |
1939 | strbuf_addstr(&uc->ident, get_ident_string()); | |
1940 | ||
1941 | /* | |
1942 | * This strbuf used to contain a list of NUL separated | |
1943 | * strings, so save NUL too for backward compatibility. | |
1944 | */ | |
1945 | strbuf_addch(&uc->ident, 0); | |
1946 | } | |
1947 | ||
1948 | static void new_untracked_cache(struct index_state *istate) | |
1949 | { | |
1950 | struct untracked_cache *uc = xcalloc(1, sizeof(*uc)); | |
1951 | strbuf_init(&uc->ident, 100); | |
1952 | uc->exclude_per_dir = ".gitignore"; | |
1953 | /* should be the same flags used by git-status */ | |
1954 | uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES; | |
1955 | set_untracked_ident(uc); | |
1956 | istate->untracked = uc; | |
1957 | istate->cache_changed |= UNTRACKED_CHANGED; | |
1958 | } | |
1959 | ||
1960 | void add_untracked_cache(struct index_state *istate) | |
1961 | { | |
1962 | if (!istate->untracked) { | |
1963 | new_untracked_cache(istate); | |
1964 | } else { | |
1965 | if (!ident_in_untracked(istate->untracked)) { | |
1966 | free_untracked_cache(istate->untracked); | |
1967 | new_untracked_cache(istate); | |
1968 | } | |
1969 | } | |
1970 | } | |
1971 | ||
1972 | void remove_untracked_cache(struct index_state *istate) | |
1973 | { | |
1974 | if (istate->untracked) { | |
1975 | free_untracked_cache(istate->untracked); | |
1976 | istate->untracked = NULL; | |
1977 | istate->cache_changed |= UNTRACKED_CHANGED; | |
1978 | } | |
1979 | } | |
1980 | ||
1981 | static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir, | |
1982 | int base_len, | |
1983 | const struct pathspec *pathspec) | |
1984 | { | |
1985 | struct untracked_cache_dir *root; | |
1986 | ||
1987 | if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE")) | |
1988 | return NULL; | |
1989 | ||
1990 | /* | |
1991 | * We only support $GIT_DIR/info/exclude and core.excludesfile | |
1992 | * as the global ignore rule files. Any other additions | |
1993 | * (e.g. from command line) invalidate the cache. This | |
1994 | * condition also catches running setup_standard_excludes() | |
1995 | * before setting dir->untracked! | |
1996 | */ | |
1997 | if (dir->unmanaged_exclude_files) | |
1998 | return NULL; | |
1999 | ||
2000 | /* | |
2001 | * Optimize for the main use case only: whole-tree git | |
2002 | * status. More work involved in treat_leading_path() if we | |
2003 | * use cache on just a subset of the worktree. pathspec | |
2004 | * support could make the matter even worse. | |
2005 | */ | |
2006 | if (base_len || (pathspec && pathspec->nr)) | |
2007 | return NULL; | |
2008 | ||
2009 | /* Different set of flags may produce different results */ | |
2010 | if (dir->flags != dir->untracked->dir_flags || | |
2011 | /* | |
2012 | * See treat_directory(), case index_nonexistent. Without | |
2013 | * this flag, we may need to also cache .git file content | |
2014 | * for the resolve_gitlink_ref() call, which we don't. | |
2015 | */ | |
2016 | !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) || | |
2017 | /* We don't support collecting ignore files */ | |
2018 | (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO | | |
2019 | DIR_COLLECT_IGNORED))) | |
2020 | return NULL; | |
2021 | ||
2022 | /* | |
2023 | * If we use .gitignore in the cache and now you change it to | |
2024 | * .gitexclude, everything will go wrong. | |
2025 | */ | |
2026 | if (dir->exclude_per_dir != dir->untracked->exclude_per_dir && | |
2027 | strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir)) | |
2028 | return NULL; | |
2029 | ||
2030 | /* | |
2031 | * EXC_CMDL is not considered in the cache. If people set it, | |
2032 | * skip the cache. | |
2033 | */ | |
2034 | if (dir->exclude_list_group[EXC_CMDL].nr) | |
2035 | return NULL; | |
2036 | ||
2037 | if (!ident_in_untracked(dir->untracked)) { | |
2038 | warning(_("Untracked cache is disabled on this system or location.")); | |
2039 | return NULL; | |
2040 | } | |
2041 | ||
2042 | if (!dir->untracked->root) { | |
2043 | const int len = sizeof(*dir->untracked->root); | |
2044 | dir->untracked->root = xmalloc(len); | |
2045 | memset(dir->untracked->root, 0, len); | |
2046 | } | |
2047 | ||
2048 | /* Validate $GIT_DIR/info/exclude and core.excludesfile */ | |
2049 | root = dir->untracked->root; | |
2050 | if (hashcmp(dir->ss_info_exclude.sha1, | |
2051 | dir->untracked->ss_info_exclude.sha1)) { | |
2052 | invalidate_gitignore(dir->untracked, root); | |
2053 | dir->untracked->ss_info_exclude = dir->ss_info_exclude; | |
2054 | } | |
2055 | if (hashcmp(dir->ss_excludes_file.sha1, | |
2056 | dir->untracked->ss_excludes_file.sha1)) { | |
2057 | invalidate_gitignore(dir->untracked, root); | |
2058 | dir->untracked->ss_excludes_file = dir->ss_excludes_file; | |
2059 | } | |
2060 | ||
2061 | /* Make sure this directory is not dropped out at saving phase */ | |
2062 | root->recurse = 1; | |
2063 | return root; | |
2064 | } | |
2065 | ||
2066 | int read_directory(struct dir_struct *dir, const char *path, | |
2067 | int len, const struct pathspec *pathspec) | |
2068 | { | |
2069 | struct untracked_cache_dir *untracked; | |
2070 | ||
2071 | if (has_symlink_leading_path(path, len)) | |
2072 | return dir->nr; | |
2073 | ||
2074 | untracked = validate_untracked_cache(dir, len, pathspec); | |
2075 | if (!untracked) | |
2076 | /* | |
2077 | * make sure untracked cache code path is disabled, | |
2078 | * e.g. prep_exclude() | |
2079 | */ | |
2080 | dir->untracked = NULL; | |
2081 | if (!len || treat_leading_path(dir, path, len, pathspec)) | |
2082 | read_directory_recursive(dir, path, len, untracked, 0, pathspec); | |
2083 | QSORT(dir->entries, dir->nr, cmp_name); | |
2084 | QSORT(dir->ignored, dir->ignored_nr, cmp_name); | |
2085 | if (dir->untracked) { | |
2086 | static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS); | |
2087 | trace_printf_key(&trace_untracked_stats, | |
2088 | "node creation: %u\n" | |
2089 | "gitignore invalidation: %u\n" | |
2090 | "directory invalidation: %u\n" | |
2091 | "opendir: %u\n", | |
2092 | dir->untracked->dir_created, | |
2093 | dir->untracked->gitignore_invalidated, | |
2094 | dir->untracked->dir_invalidated, | |
2095 | dir->untracked->dir_opened); | |
2096 | if (dir->untracked == the_index.untracked && | |
2097 | (dir->untracked->dir_opened || | |
2098 | dir->untracked->gitignore_invalidated || | |
2099 | dir->untracked->dir_invalidated)) | |
2100 | the_index.cache_changed |= UNTRACKED_CHANGED; | |
2101 | if (dir->untracked != the_index.untracked) { | |
2102 | free(dir->untracked); | |
2103 | dir->untracked = NULL; | |
2104 | } | |
2105 | } | |
2106 | return dir->nr; | |
2107 | } | |
2108 | ||
2109 | int file_exists(const char *f) | |
2110 | { | |
2111 | struct stat sb; | |
2112 | return lstat(f, &sb) == 0; | |
2113 | } | |
2114 | ||
2115 | static int cmp_icase(char a, char b) | |
2116 | { | |
2117 | if (a == b) | |
2118 | return 0; | |
2119 | if (ignore_case) | |
2120 | return toupper(a) - toupper(b); | |
2121 | return a - b; | |
2122 | } | |
2123 | ||
2124 | /* | |
2125 | * Given two normalized paths (a trailing slash is ok), if subdir is | |
2126 | * outside dir, return -1. Otherwise return the offset in subdir that | |
2127 | * can be used as relative path to dir. | |
2128 | */ | |
2129 | int dir_inside_of(const char *subdir, const char *dir) | |
2130 | { | |
2131 | int offset = 0; | |
2132 | ||
2133 | assert(dir && subdir && *dir && *subdir); | |
2134 | ||
2135 | while (*dir && *subdir && !cmp_icase(*dir, *subdir)) { | |
2136 | dir++; | |
2137 | subdir++; | |
2138 | offset++; | |
2139 | } | |
2140 | ||
2141 | /* hel[p]/me vs hel[l]/yeah */ | |
2142 | if (*dir && *subdir) | |
2143 | return -1; | |
2144 | ||
2145 | if (!*subdir) | |
2146 | return !*dir ? offset : -1; /* same dir */ | |
2147 | ||
2148 | /* foo/[b]ar vs foo/[] */ | |
2149 | if (is_dir_sep(dir[-1])) | |
2150 | return is_dir_sep(subdir[-1]) ? offset : -1; | |
2151 | ||
2152 | /* foo[/]bar vs foo[] */ | |
2153 | return is_dir_sep(*subdir) ? offset + 1 : -1; | |
2154 | } | |
2155 | ||
2156 | int is_inside_dir(const char *dir) | |
2157 | { | |
2158 | char *cwd; | |
2159 | int rc; | |
2160 | ||
2161 | if (!dir) | |
2162 | return 0; | |
2163 | ||
2164 | cwd = xgetcwd(); | |
2165 | rc = (dir_inside_of(cwd, dir) >= 0); | |
2166 | free(cwd); | |
2167 | return rc; | |
2168 | } | |
2169 | ||
2170 | int is_empty_dir(const char *path) | |
2171 | { | |
2172 | DIR *dir = opendir(path); | |
2173 | struct dirent *e; | |
2174 | int ret = 1; | |
2175 | ||
2176 | if (!dir) | |
2177 | return 0; | |
2178 | ||
2179 | while ((e = readdir(dir)) != NULL) | |
2180 | if (!is_dot_or_dotdot(e->d_name)) { | |
2181 | ret = 0; | |
2182 | break; | |
2183 | } | |
2184 | ||
2185 | closedir(dir); | |
2186 | return ret; | |
2187 | } | |
2188 | ||
2189 | static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up) | |
2190 | { | |
2191 | DIR *dir; | |
2192 | struct dirent *e; | |
2193 | int ret = 0, original_len = path->len, len, kept_down = 0; | |
2194 | int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY); | |
2195 | int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL); | |
2196 | unsigned char submodule_head[20]; | |
2197 | ||
2198 | if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) && | |
2199 | !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) { | |
2200 | /* Do not descend and nuke a nested git work tree. */ | |
2201 | if (kept_up) | |
2202 | *kept_up = 1; | |
2203 | return 0; | |
2204 | } | |
2205 | ||
2206 | flag &= ~REMOVE_DIR_KEEP_TOPLEVEL; | |
2207 | dir = opendir(path->buf); | |
2208 | if (!dir) { | |
2209 | if (errno == ENOENT) | |
2210 | return keep_toplevel ? -1 : 0; | |
2211 | else if (errno == EACCES && !keep_toplevel) | |
2212 | /* | |
2213 | * An empty dir could be removable even if it | |
2214 | * is unreadable: | |
2215 | */ | |
2216 | return rmdir(path->buf); | |
2217 | else | |
2218 | return -1; | |
2219 | } | |
2220 | strbuf_complete(path, '/'); | |
2221 | ||
2222 | len = path->len; | |
2223 | while ((e = readdir(dir)) != NULL) { | |
2224 | struct stat st; | |
2225 | if (is_dot_or_dotdot(e->d_name)) | |
2226 | continue; | |
2227 | ||
2228 | strbuf_setlen(path, len); | |
2229 | strbuf_addstr(path, e->d_name); | |
2230 | if (lstat(path->buf, &st)) { | |
2231 | if (errno == ENOENT) | |
2232 | /* | |
2233 | * file disappeared, which is what we | |
2234 | * wanted anyway | |
2235 | */ | |
2236 | continue; | |
2237 | /* fall thru */ | |
2238 | } else if (S_ISDIR(st.st_mode)) { | |
2239 | if (!remove_dir_recurse(path, flag, &kept_down)) | |
2240 | continue; /* happy */ | |
2241 | } else if (!only_empty && | |
2242 | (!unlink(path->buf) || errno == ENOENT)) { | |
2243 | continue; /* happy, too */ | |
2244 | } | |
2245 | ||
2246 | /* path too long, stat fails, or non-directory still exists */ | |
2247 | ret = -1; | |
2248 | break; | |
2249 | } | |
2250 | closedir(dir); | |
2251 | ||
2252 | strbuf_setlen(path, original_len); | |
2253 | if (!ret && !keep_toplevel && !kept_down) | |
2254 | ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1; | |
2255 | else if (kept_up) | |
2256 | /* | |
2257 | * report the uplevel that it is not an error that we | |
2258 | * did not rmdir() our directory. | |
2259 | */ | |
2260 | *kept_up = !ret; | |
2261 | return ret; | |
2262 | } | |
2263 | ||
2264 | int remove_dir_recursively(struct strbuf *path, int flag) | |
2265 | { | |
2266 | return remove_dir_recurse(path, flag, NULL); | |
2267 | } | |
2268 | ||
2269 | static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude") | |
2270 | ||
2271 | void setup_standard_excludes(struct dir_struct *dir) | |
2272 | { | |
2273 | dir->exclude_per_dir = ".gitignore"; | |
2274 | ||
2275 | /* core.excludefile defaulting to $XDG_HOME/git/ignore */ | |
2276 | if (!excludes_file) | |
2277 | excludes_file = xdg_config_home("ignore"); | |
2278 | if (excludes_file && !access_or_warn(excludes_file, R_OK, 0)) | |
2279 | add_excludes_from_file_1(dir, excludes_file, | |
2280 | dir->untracked ? &dir->ss_excludes_file : NULL); | |
2281 | ||
2282 | /* per repository user preference */ | |
2283 | if (startup_info->have_repository) { | |
2284 | const char *path = git_path_info_exclude(); | |
2285 | if (!access_or_warn(path, R_OK, 0)) | |
2286 | add_excludes_from_file_1(dir, path, | |
2287 | dir->untracked ? &dir->ss_info_exclude : NULL); | |
2288 | } | |
2289 | } | |
2290 | ||
2291 | int remove_path(const char *name) | |
2292 | { | |
2293 | char *slash; | |
2294 | ||
2295 | if (unlink(name) && errno != ENOENT && errno != ENOTDIR) | |
2296 | return -1; | |
2297 | ||
2298 | slash = strrchr(name, '/'); | |
2299 | if (slash) { | |
2300 | char *dirs = xstrdup(name); | |
2301 | slash = dirs + (slash - name); | |
2302 | do { | |
2303 | *slash = '\0'; | |
2304 | } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/'))); | |
2305 | free(dirs); | |
2306 | } | |
2307 | return 0; | |
2308 | } | |
2309 | ||
2310 | /* | |
2311 | * Frees memory within dir which was allocated for exclude lists and | |
2312 | * the exclude_stack. Does not free dir itself. | |
2313 | */ | |
2314 | void clear_directory(struct dir_struct *dir) | |
2315 | { | |
2316 | int i, j; | |
2317 | struct exclude_list_group *group; | |
2318 | struct exclude_list *el; | |
2319 | struct exclude_stack *stk; | |
2320 | ||
2321 | for (i = EXC_CMDL; i <= EXC_FILE; i++) { | |
2322 | group = &dir->exclude_list_group[i]; | |
2323 | for (j = 0; j < group->nr; j++) { | |
2324 | el = &group->el[j]; | |
2325 | if (i == EXC_DIRS) | |
2326 | free((char *)el->src); | |
2327 | clear_exclude_list(el); | |
2328 | } | |
2329 | free(group->el); | |
2330 | } | |
2331 | ||
2332 | stk = dir->exclude_stack; | |
2333 | while (stk) { | |
2334 | struct exclude_stack *prev = stk->prev; | |
2335 | free(stk); | |
2336 | stk = prev; | |
2337 | } | |
2338 | strbuf_release(&dir->basebuf); | |
2339 | } | |
2340 | ||
2341 | struct ondisk_untracked_cache { | |
2342 | struct stat_data info_exclude_stat; | |
2343 | struct stat_data excludes_file_stat; | |
2344 | uint32_t dir_flags; | |
2345 | unsigned char info_exclude_sha1[20]; | |
2346 | unsigned char excludes_file_sha1[20]; | |
2347 | char exclude_per_dir[FLEX_ARRAY]; | |
2348 | }; | |
2349 | ||
2350 | #define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1) | |
2351 | ||
2352 | struct write_data { | |
2353 | int index; /* number of written untracked_cache_dir */ | |
2354 | struct ewah_bitmap *check_only; /* from untracked_cache_dir */ | |
2355 | struct ewah_bitmap *valid; /* from untracked_cache_dir */ | |
2356 | struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */ | |
2357 | struct strbuf out; | |
2358 | struct strbuf sb_stat; | |
2359 | struct strbuf sb_sha1; | |
2360 | }; | |
2361 | ||
2362 | static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from) | |
2363 | { | |
2364 | to->sd_ctime.sec = htonl(from->sd_ctime.sec); | |
2365 | to->sd_ctime.nsec = htonl(from->sd_ctime.nsec); | |
2366 | to->sd_mtime.sec = htonl(from->sd_mtime.sec); | |
2367 | to->sd_mtime.nsec = htonl(from->sd_mtime.nsec); | |
2368 | to->sd_dev = htonl(from->sd_dev); | |
2369 | to->sd_ino = htonl(from->sd_ino); | |
2370 | to->sd_uid = htonl(from->sd_uid); | |
2371 | to->sd_gid = htonl(from->sd_gid); | |
2372 | to->sd_size = htonl(from->sd_size); | |
2373 | } | |
2374 | ||
2375 | static void write_one_dir(struct untracked_cache_dir *untracked, | |
2376 | struct write_data *wd) | |
2377 | { | |
2378 | struct stat_data stat_data; | |
2379 | struct strbuf *out = &wd->out; | |
2380 | unsigned char intbuf[16]; | |
2381 | unsigned int intlen, value; | |
2382 | int i = wd->index++; | |
2383 | ||
2384 | /* | |
2385 | * untracked_nr should be reset whenever valid is clear, but | |
2386 | * for safety.. | |
2387 | */ | |
2388 | if (!untracked->valid) { | |
2389 | untracked->untracked_nr = 0; | |
2390 | untracked->check_only = 0; | |
2391 | } | |
2392 | ||
2393 | if (untracked->check_only) | |
2394 | ewah_set(wd->check_only, i); | |
2395 | if (untracked->valid) { | |
2396 | ewah_set(wd->valid, i); | |
2397 | stat_data_to_disk(&stat_data, &untracked->stat_data); | |
2398 | strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data)); | |
2399 | } | |
2400 | if (!is_null_sha1(untracked->exclude_sha1)) { | |
2401 | ewah_set(wd->sha1_valid, i); | |
2402 | strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20); | |
2403 | } | |
2404 | ||
2405 | intlen = encode_varint(untracked->untracked_nr, intbuf); | |
2406 | strbuf_add(out, intbuf, intlen); | |
2407 | ||
2408 | /* skip non-recurse directories */ | |
2409 | for (i = 0, value = 0; i < untracked->dirs_nr; i++) | |
2410 | if (untracked->dirs[i]->recurse) | |
2411 | value++; | |
2412 | intlen = encode_varint(value, intbuf); | |
2413 | strbuf_add(out, intbuf, intlen); | |
2414 | ||
2415 | strbuf_add(out, untracked->name, strlen(untracked->name) + 1); | |
2416 | ||
2417 | for (i = 0; i < untracked->untracked_nr; i++) | |
2418 | strbuf_add(out, untracked->untracked[i], | |
2419 | strlen(untracked->untracked[i]) + 1); | |
2420 | ||
2421 | for (i = 0; i < untracked->dirs_nr; i++) | |
2422 | if (untracked->dirs[i]->recurse) | |
2423 | write_one_dir(untracked->dirs[i], wd); | |
2424 | } | |
2425 | ||
2426 | void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked) | |
2427 | { | |
2428 | struct ondisk_untracked_cache *ouc; | |
2429 | struct write_data wd; | |
2430 | unsigned char varbuf[16]; | |
2431 | int varint_len; | |
2432 | size_t len = strlen(untracked->exclude_per_dir); | |
2433 | ||
2434 | FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len); | |
2435 | stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat); | |
2436 | stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat); | |
2437 | hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1); | |
2438 | hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1); | |
2439 | ouc->dir_flags = htonl(untracked->dir_flags); | |
2440 | ||
2441 | varint_len = encode_varint(untracked->ident.len, varbuf); | |
2442 | strbuf_add(out, varbuf, varint_len); | |
2443 | strbuf_addbuf(out, &untracked->ident); | |
2444 | ||
2445 | strbuf_add(out, ouc, ouc_size(len)); | |
2446 | free(ouc); | |
2447 | ouc = NULL; | |
2448 | ||
2449 | if (!untracked->root) { | |
2450 | varint_len = encode_varint(0, varbuf); | |
2451 | strbuf_add(out, varbuf, varint_len); | |
2452 | return; | |
2453 | } | |
2454 | ||
2455 | wd.index = 0; | |
2456 | wd.check_only = ewah_new(); | |
2457 | wd.valid = ewah_new(); | |
2458 | wd.sha1_valid = ewah_new(); | |
2459 | strbuf_init(&wd.out, 1024); | |
2460 | strbuf_init(&wd.sb_stat, 1024); | |
2461 | strbuf_init(&wd.sb_sha1, 1024); | |
2462 | write_one_dir(untracked->root, &wd); | |
2463 | ||
2464 | varint_len = encode_varint(wd.index, varbuf); | |
2465 | strbuf_add(out, varbuf, varint_len); | |
2466 | strbuf_addbuf(out, &wd.out); | |
2467 | ewah_serialize_strbuf(wd.valid, out); | |
2468 | ewah_serialize_strbuf(wd.check_only, out); | |
2469 | ewah_serialize_strbuf(wd.sha1_valid, out); | |
2470 | strbuf_addbuf(out, &wd.sb_stat); | |
2471 | strbuf_addbuf(out, &wd.sb_sha1); | |
2472 | strbuf_addch(out, '\0'); /* safe guard for string lists */ | |
2473 | ||
2474 | ewah_free(wd.valid); | |
2475 | ewah_free(wd.check_only); | |
2476 | ewah_free(wd.sha1_valid); | |
2477 | strbuf_release(&wd.out); | |
2478 | strbuf_release(&wd.sb_stat); | |
2479 | strbuf_release(&wd.sb_sha1); | |
2480 | } | |
2481 | ||
2482 | static void free_untracked(struct untracked_cache_dir *ucd) | |
2483 | { | |
2484 | int i; | |
2485 | if (!ucd) | |
2486 | return; | |
2487 | for (i = 0; i < ucd->dirs_nr; i++) | |
2488 | free_untracked(ucd->dirs[i]); | |
2489 | for (i = 0; i < ucd->untracked_nr; i++) | |
2490 | free(ucd->untracked[i]); | |
2491 | free(ucd->untracked); | |
2492 | free(ucd->dirs); | |
2493 | free(ucd); | |
2494 | } | |
2495 | ||
2496 | void free_untracked_cache(struct untracked_cache *uc) | |
2497 | { | |
2498 | if (uc) | |
2499 | free_untracked(uc->root); | |
2500 | free(uc); | |
2501 | } | |
2502 | ||
2503 | struct read_data { | |
2504 | int index; | |
2505 | struct untracked_cache_dir **ucd; | |
2506 | struct ewah_bitmap *check_only; | |
2507 | struct ewah_bitmap *valid; | |
2508 | struct ewah_bitmap *sha1_valid; | |
2509 | const unsigned char *data; | |
2510 | const unsigned char *end; | |
2511 | }; | |
2512 | ||
2513 | static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from) | |
2514 | { | |
2515 | to->sd_ctime.sec = get_be32(&from->sd_ctime.sec); | |
2516 | to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec); | |
2517 | to->sd_mtime.sec = get_be32(&from->sd_mtime.sec); | |
2518 | to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec); | |
2519 | to->sd_dev = get_be32(&from->sd_dev); | |
2520 | to->sd_ino = get_be32(&from->sd_ino); | |
2521 | to->sd_uid = get_be32(&from->sd_uid); | |
2522 | to->sd_gid = get_be32(&from->sd_gid); | |
2523 | to->sd_size = get_be32(&from->sd_size); | |
2524 | } | |
2525 | ||
2526 | static int read_one_dir(struct untracked_cache_dir **untracked_, | |
2527 | struct read_data *rd) | |
2528 | { | |
2529 | struct untracked_cache_dir ud, *untracked; | |
2530 | const unsigned char *next, *data = rd->data, *end = rd->end; | |
2531 | unsigned int value; | |
2532 | int i, len; | |
2533 | ||
2534 | memset(&ud, 0, sizeof(ud)); | |
2535 | ||
2536 | next = data; | |
2537 | value = decode_varint(&next); | |
2538 | if (next > end) | |
2539 | return -1; | |
2540 | ud.recurse = 1; | |
2541 | ud.untracked_alloc = value; | |
2542 | ud.untracked_nr = value; | |
2543 | if (ud.untracked_nr) | |
2544 | ALLOC_ARRAY(ud.untracked, ud.untracked_nr); | |
2545 | data = next; | |
2546 | ||
2547 | next = data; | |
2548 | ud.dirs_alloc = ud.dirs_nr = decode_varint(&next); | |
2549 | if (next > end) | |
2550 | return -1; | |
2551 | ALLOC_ARRAY(ud.dirs, ud.dirs_nr); | |
2552 | data = next; | |
2553 | ||
2554 | len = strlen((const char *)data); | |
2555 | next = data + len + 1; | |
2556 | if (next > rd->end) | |
2557 | return -1; | |
2558 | *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len)); | |
2559 | memcpy(untracked, &ud, sizeof(ud)); | |
2560 | memcpy(untracked->name, data, len + 1); | |
2561 | data = next; | |
2562 | ||
2563 | for (i = 0; i < untracked->untracked_nr; i++) { | |
2564 | len = strlen((const char *)data); | |
2565 | next = data + len + 1; | |
2566 | if (next > rd->end) | |
2567 | return -1; | |
2568 | untracked->untracked[i] = xstrdup((const char*)data); | |
2569 | data = next; | |
2570 | } | |
2571 | ||
2572 | rd->ucd[rd->index++] = untracked; | |
2573 | rd->data = data; | |
2574 | ||
2575 | for (i = 0; i < untracked->dirs_nr; i++) { | |
2576 | len = read_one_dir(untracked->dirs + i, rd); | |
2577 | if (len < 0) | |
2578 | return -1; | |
2579 | } | |
2580 | return 0; | |
2581 | } | |
2582 | ||
2583 | static void set_check_only(size_t pos, void *cb) | |
2584 | { | |
2585 | struct read_data *rd = cb; | |
2586 | struct untracked_cache_dir *ud = rd->ucd[pos]; | |
2587 | ud->check_only = 1; | |
2588 | } | |
2589 | ||
2590 | static void read_stat(size_t pos, void *cb) | |
2591 | { | |
2592 | struct read_data *rd = cb; | |
2593 | struct untracked_cache_dir *ud = rd->ucd[pos]; | |
2594 | if (rd->data + sizeof(struct stat_data) > rd->end) { | |
2595 | rd->data = rd->end + 1; | |
2596 | return; | |
2597 | } | |
2598 | stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data); | |
2599 | rd->data += sizeof(struct stat_data); | |
2600 | ud->valid = 1; | |
2601 | } | |
2602 | ||
2603 | static void read_sha1(size_t pos, void *cb) | |
2604 | { | |
2605 | struct read_data *rd = cb; | |
2606 | struct untracked_cache_dir *ud = rd->ucd[pos]; | |
2607 | if (rd->data + 20 > rd->end) { | |
2608 | rd->data = rd->end + 1; | |
2609 | return; | |
2610 | } | |
2611 | hashcpy(ud->exclude_sha1, rd->data); | |
2612 | rd->data += 20; | |
2613 | } | |
2614 | ||
2615 | static void load_sha1_stat(struct sha1_stat *sha1_stat, | |
2616 | const struct stat_data *stat, | |
2617 | const unsigned char *sha1) | |
2618 | { | |
2619 | stat_data_from_disk(&sha1_stat->stat, stat); | |
2620 | hashcpy(sha1_stat->sha1, sha1); | |
2621 | sha1_stat->valid = 1; | |
2622 | } | |
2623 | ||
2624 | struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz) | |
2625 | { | |
2626 | const struct ondisk_untracked_cache *ouc; | |
2627 | struct untracked_cache *uc; | |
2628 | struct read_data rd; | |
2629 | const unsigned char *next = data, *end = (const unsigned char *)data + sz; | |
2630 | const char *ident; | |
2631 | int ident_len, len; | |
2632 | ||
2633 | if (sz <= 1 || end[-1] != '\0') | |
2634 | return NULL; | |
2635 | end--; | |
2636 | ||
2637 | ident_len = decode_varint(&next); | |
2638 | if (next + ident_len > end) | |
2639 | return NULL; | |
2640 | ident = (const char *)next; | |
2641 | next += ident_len; | |
2642 | ||
2643 | ouc = (const struct ondisk_untracked_cache *)next; | |
2644 | if (next + ouc_size(0) > end) | |
2645 | return NULL; | |
2646 | ||
2647 | uc = xcalloc(1, sizeof(*uc)); | |
2648 | strbuf_init(&uc->ident, ident_len); | |
2649 | strbuf_add(&uc->ident, ident, ident_len); | |
2650 | load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat, | |
2651 | ouc->info_exclude_sha1); | |
2652 | load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat, | |
2653 | ouc->excludes_file_sha1); | |
2654 | uc->dir_flags = get_be32(&ouc->dir_flags); | |
2655 | uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir); | |
2656 | /* NUL after exclude_per_dir is covered by sizeof(*ouc) */ | |
2657 | next += ouc_size(strlen(ouc->exclude_per_dir)); | |
2658 | if (next >= end) | |
2659 | goto done2; | |
2660 | ||
2661 | len = decode_varint(&next); | |
2662 | if (next > end || len == 0) | |
2663 | goto done2; | |
2664 | ||
2665 | rd.valid = ewah_new(); | |
2666 | rd.check_only = ewah_new(); | |
2667 | rd.sha1_valid = ewah_new(); | |
2668 | rd.data = next; | |
2669 | rd.end = end; | |
2670 | rd.index = 0; | |
2671 | ALLOC_ARRAY(rd.ucd, len); | |
2672 | ||
2673 | if (read_one_dir(&uc->root, &rd) || rd.index != len) | |
2674 | goto done; | |
2675 | ||
2676 | next = rd.data; | |
2677 | len = ewah_read_mmap(rd.valid, next, end - next); | |
2678 | if (len < 0) | |
2679 | goto done; | |
2680 | ||
2681 | next += len; | |
2682 | len = ewah_read_mmap(rd.check_only, next, end - next); | |
2683 | if (len < 0) | |
2684 | goto done; | |
2685 | ||
2686 | next += len; | |
2687 | len = ewah_read_mmap(rd.sha1_valid, next, end - next); | |
2688 | if (len < 0) | |
2689 | goto done; | |
2690 | ||
2691 | ewah_each_bit(rd.check_only, set_check_only, &rd); | |
2692 | rd.data = next + len; | |
2693 | ewah_each_bit(rd.valid, read_stat, &rd); | |
2694 | ewah_each_bit(rd.sha1_valid, read_sha1, &rd); | |
2695 | next = rd.data; | |
2696 | ||
2697 | done: | |
2698 | free(rd.ucd); | |
2699 | ewah_free(rd.valid); | |
2700 | ewah_free(rd.check_only); | |
2701 | ewah_free(rd.sha1_valid); | |
2702 | done2: | |
2703 | if (next != end) { | |
2704 | free_untracked_cache(uc); | |
2705 | uc = NULL; | |
2706 | } | |
2707 | return uc; | |
2708 | } | |
2709 | ||
2710 | static void invalidate_one_directory(struct untracked_cache *uc, | |
2711 | struct untracked_cache_dir *ucd) | |
2712 | { | |
2713 | uc->dir_invalidated++; | |
2714 | ucd->valid = 0; | |
2715 | ucd->untracked_nr = 0; | |
2716 | } | |
2717 | ||
2718 | /* | |
2719 | * Normally when an entry is added or removed from a directory, | |
2720 | * invalidating that directory is enough. No need to touch its | |
2721 | * ancestors. When a directory is shown as "foo/bar/" in git-status | |
2722 | * however, deleting or adding an entry may have cascading effect. | |
2723 | * | |
2724 | * Say the "foo/bar/file" has become untracked, we need to tell the | |
2725 | * untracked_cache_dir of "foo" that "bar/" is not an untracked | |
2726 | * directory any more (because "bar" is managed by foo as an untracked | |
2727 | * "file"). | |
2728 | * | |
2729 | * Similarly, if "foo/bar/file" moves from untracked to tracked and it | |
2730 | * was the last untracked entry in the entire "foo", we should show | |
2731 | * "foo/" instead. Which means we have to invalidate past "bar" up to | |
2732 | * "foo". | |
2733 | * | |
2734 | * This function traverses all directories from root to leaf. If there | |
2735 | * is a chance of one of the above cases happening, we invalidate back | |
2736 | * to root. Otherwise we just invalidate the leaf. There may be a more | |
2737 | * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to | |
2738 | * detect these cases and avoid unnecessary invalidation, for example, | |
2739 | * checking for the untracked entry named "bar/" in "foo", but for now | |
2740 | * stick to something safe and simple. | |
2741 | */ | |
2742 | static int invalidate_one_component(struct untracked_cache *uc, | |
2743 | struct untracked_cache_dir *dir, | |
2744 | const char *path, int len) | |
2745 | { | |
2746 | const char *rest = strchr(path, '/'); | |
2747 | ||
2748 | if (rest) { | |
2749 | int component_len = rest - path; | |
2750 | struct untracked_cache_dir *d = | |
2751 | lookup_untracked(uc, dir, path, component_len); | |
2752 | int ret = | |
2753 | invalidate_one_component(uc, d, rest + 1, | |
2754 | len - (component_len + 1)); | |
2755 | if (ret) | |
2756 | invalidate_one_directory(uc, dir); | |
2757 | return ret; | |
2758 | } | |
2759 | ||
2760 | invalidate_one_directory(uc, dir); | |
2761 | return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES; | |
2762 | } | |
2763 | ||
2764 | void untracked_cache_invalidate_path(struct index_state *istate, | |
2765 | const char *path) | |
2766 | { | |
2767 | if (!istate->untracked || !istate->untracked->root) | |
2768 | return; | |
2769 | invalidate_one_component(istate->untracked, istate->untracked->root, | |
2770 | path, strlen(path)); | |
2771 | } | |
2772 | ||
2773 | void untracked_cache_remove_from_index(struct index_state *istate, | |
2774 | const char *path) | |
2775 | { | |
2776 | untracked_cache_invalidate_path(istate, path); | |
2777 | } | |
2778 | ||
2779 | void untracked_cache_add_to_index(struct index_state *istate, | |
2780 | const char *path) | |
2781 | { | |
2782 | untracked_cache_invalidate_path(istate, path); | |
2783 | } | |
2784 | ||
2785 | /* Update gitfile and core.worktree setting to connect work tree and git dir */ | |
2786 | void connect_work_tree_and_git_dir(const char *work_tree_, const char *git_dir_) | |
2787 | { | |
2788 | struct strbuf gitfile_sb = STRBUF_INIT; | |
2789 | struct strbuf cfg_sb = STRBUF_INIT; | |
2790 | struct strbuf rel_path = STRBUF_INIT; | |
2791 | char *git_dir, *work_tree; | |
2792 | ||
2793 | /* Prepare .git file */ | |
2794 | strbuf_addf(&gitfile_sb, "%s/.git", work_tree_); | |
2795 | if (safe_create_leading_directories_const(gitfile_sb.buf)) | |
2796 | die(_("could not create directories for %s"), gitfile_sb.buf); | |
2797 | ||
2798 | /* Prepare config file */ | |
2799 | strbuf_addf(&cfg_sb, "%s/config", git_dir_); | |
2800 | if (safe_create_leading_directories_const(cfg_sb.buf)) | |
2801 | die(_("could not create directories for %s"), cfg_sb.buf); | |
2802 | ||
2803 | git_dir = real_pathdup(git_dir_, 1); | |
2804 | work_tree = real_pathdup(work_tree_, 1); | |
2805 | ||
2806 | /* Write .git file */ | |
2807 | write_file(gitfile_sb.buf, "gitdir: %s", | |
2808 | relative_path(git_dir, work_tree, &rel_path)); | |
2809 | /* Update core.worktree setting */ | |
2810 | git_config_set_in_file(cfg_sb.buf, "core.worktree", | |
2811 | relative_path(work_tree, git_dir, &rel_path)); | |
2812 | ||
2813 | strbuf_release(&gitfile_sb); | |
2814 | strbuf_release(&cfg_sb); | |
2815 | strbuf_release(&rel_path); | |
2816 | free(work_tree); | |
2817 | free(git_dir); | |
2818 | } | |
2819 | ||
2820 | /* | |
2821 | * Migrate the git directory of the given path from old_git_dir to new_git_dir. | |
2822 | */ | |
2823 | void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir) | |
2824 | { | |
2825 | if (rename(old_git_dir, new_git_dir) < 0) | |
2826 | die_errno(_("could not migrate git directory from '%s' to '%s'"), | |
2827 | old_git_dir, new_git_dir); | |
2828 | ||
2829 | connect_work_tree_and_git_dir(path, new_git_dir); | |
2830 | } |