]>
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 | #include "cache.h" | |
11 | #include "config.h" | |
12 | #include "dir.h" | |
13 | #include "object-store.h" | |
14 | #include "attr.h" | |
15 | #include "refs.h" | |
16 | #include "wildmatch.h" | |
17 | #include "pathspec.h" | |
18 | #include "utf8.h" | |
19 | #include "varint.h" | |
20 | #include "ewah/ewok.h" | |
21 | #include "fsmonitor.h" | |
22 | #include "submodule-config.h" | |
23 | ||
24 | /* | |
25 | * Tells read_directory_recursive how a file or directory should be treated. | |
26 | * Values are ordered by significance, e.g. if a directory contains both | |
27 | * excluded and untracked files, it is listed as untracked because | |
28 | * path_untracked > path_excluded. | |
29 | */ | |
30 | enum path_treatment { | |
31 | path_none = 0, | |
32 | path_recurse, | |
33 | path_excluded, | |
34 | path_untracked | |
35 | }; | |
36 | ||
37 | /* | |
38 | * Support data structure for our opendir/readdir/closedir wrappers | |
39 | */ | |
40 | struct cached_dir { | |
41 | DIR *fdir; | |
42 | struct untracked_cache_dir *untracked; | |
43 | int nr_files; | |
44 | int nr_dirs; | |
45 | ||
46 | struct dirent *de; | |
47 | const char *file; | |
48 | struct untracked_cache_dir *ucd; | |
49 | }; | |
50 | ||
51 | static enum path_treatment read_directory_recursive(struct dir_struct *dir, | |
52 | struct index_state *istate, const char *path, int len, | |
53 | struct untracked_cache_dir *untracked, | |
54 | int check_only, int stop_at_first_file, const struct pathspec *pathspec); | |
55 | static int get_dtype(struct dirent *de, struct index_state *istate, | |
56 | const char *path, int len); | |
57 | ||
58 | int count_slashes(const char *s) | |
59 | { | |
60 | int cnt = 0; | |
61 | while (*s) | |
62 | if (*s++ == '/') | |
63 | cnt++; | |
64 | return cnt; | |
65 | } | |
66 | ||
67 | int fspathcmp(const char *a, const char *b) | |
68 | { | |
69 | return ignore_case ? strcasecmp(a, b) : strcmp(a, b); | |
70 | } | |
71 | ||
72 | int fspathncmp(const char *a, const char *b, size_t count) | |
73 | { | |
74 | return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count); | |
75 | } | |
76 | ||
77 | int git_fnmatch(const struct pathspec_item *item, | |
78 | const char *pattern, const char *string, | |
79 | int prefix) | |
80 | { | |
81 | if (prefix > 0) { | |
82 | if (ps_strncmp(item, pattern, string, prefix)) | |
83 | return WM_NOMATCH; | |
84 | pattern += prefix; | |
85 | string += prefix; | |
86 | } | |
87 | if (item->flags & PATHSPEC_ONESTAR) { | |
88 | int pattern_len = strlen(++pattern); | |
89 | int string_len = strlen(string); | |
90 | return string_len < pattern_len || | |
91 | ps_strcmp(item, pattern, | |
92 | string + string_len - pattern_len); | |
93 | } | |
94 | if (item->magic & PATHSPEC_GLOB) | |
95 | return wildmatch(pattern, string, | |
96 | WM_PATHNAME | | |
97 | (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0)); | |
98 | else | |
99 | /* wildmatch has not learned no FNM_PATHNAME mode yet */ | |
100 | return wildmatch(pattern, string, | |
101 | item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0); | |
102 | } | |
103 | ||
104 | static int fnmatch_icase_mem(const char *pattern, int patternlen, | |
105 | const char *string, int stringlen, | |
106 | int flags) | |
107 | { | |
108 | int match_status; | |
109 | struct strbuf pat_buf = STRBUF_INIT; | |
110 | struct strbuf str_buf = STRBUF_INIT; | |
111 | const char *use_pat = pattern; | |
112 | const char *use_str = string; | |
113 | ||
114 | if (pattern[patternlen]) { | |
115 | strbuf_add(&pat_buf, pattern, patternlen); | |
116 | use_pat = pat_buf.buf; | |
117 | } | |
118 | if (string[stringlen]) { | |
119 | strbuf_add(&str_buf, string, stringlen); | |
120 | use_str = str_buf.buf; | |
121 | } | |
122 | ||
123 | if (ignore_case) | |
124 | flags |= WM_CASEFOLD; | |
125 | match_status = wildmatch(use_pat, use_str, flags); | |
126 | ||
127 | strbuf_release(&pat_buf); | |
128 | strbuf_release(&str_buf); | |
129 | ||
130 | return match_status; | |
131 | } | |
132 | ||
133 | static size_t common_prefix_len(const struct pathspec *pathspec) | |
134 | { | |
135 | int n; | |
136 | size_t max = 0; | |
137 | ||
138 | /* | |
139 | * ":(icase)path" is treated as a pathspec full of | |
140 | * wildcard. In other words, only prefix is considered common | |
141 | * prefix. If the pathspec is abc/foo abc/bar, running in | |
142 | * subdir xyz, the common prefix is still xyz, not xyz/abc as | |
143 | * in non-:(icase). | |
144 | */ | |
145 | GUARD_PATHSPEC(pathspec, | |
146 | PATHSPEC_FROMTOP | | |
147 | PATHSPEC_MAXDEPTH | | |
148 | PATHSPEC_LITERAL | | |
149 | PATHSPEC_GLOB | | |
150 | PATHSPEC_ICASE | | |
151 | PATHSPEC_EXCLUDE | | |
152 | PATHSPEC_ATTR); | |
153 | ||
154 | for (n = 0; n < pathspec->nr; n++) { | |
155 | size_t i = 0, len = 0, item_len; | |
156 | if (pathspec->items[n].magic & PATHSPEC_EXCLUDE) | |
157 | continue; | |
158 | if (pathspec->items[n].magic & PATHSPEC_ICASE) | |
159 | item_len = pathspec->items[n].prefix; | |
160 | else | |
161 | item_len = pathspec->items[n].nowildcard_len; | |
162 | while (i < item_len && (n == 0 || i < max)) { | |
163 | char c = pathspec->items[n].match[i]; | |
164 | if (c != pathspec->items[0].match[i]) | |
165 | break; | |
166 | if (c == '/') | |
167 | len = i + 1; | |
168 | i++; | |
169 | } | |
170 | if (n == 0 || len < max) { | |
171 | max = len; | |
172 | if (!max) | |
173 | break; | |
174 | } | |
175 | } | |
176 | return max; | |
177 | } | |
178 | ||
179 | /* | |
180 | * Returns a copy of the longest leading path common among all | |
181 | * pathspecs. | |
182 | */ | |
183 | char *common_prefix(const struct pathspec *pathspec) | |
184 | { | |
185 | unsigned long len = common_prefix_len(pathspec); | |
186 | ||
187 | return len ? xmemdupz(pathspec->items[0].match, len) : NULL; | |
188 | } | |
189 | ||
190 | int fill_directory(struct dir_struct *dir, | |
191 | struct index_state *istate, | |
192 | const struct pathspec *pathspec) | |
193 | { | |
194 | const char *prefix; | |
195 | size_t prefix_len; | |
196 | ||
197 | /* | |
198 | * Calculate common prefix for the pathspec, and | |
199 | * use that to optimize the directory walk | |
200 | */ | |
201 | prefix_len = common_prefix_len(pathspec); | |
202 | prefix = prefix_len ? pathspec->items[0].match : ""; | |
203 | ||
204 | /* Read the directory and prune it */ | |
205 | read_directory(dir, istate, prefix, prefix_len, pathspec); | |
206 | ||
207 | return prefix_len; | |
208 | } | |
209 | ||
210 | int within_depth(const char *name, int namelen, | |
211 | int depth, int max_depth) | |
212 | { | |
213 | const char *cp = name, *cpe = name + namelen; | |
214 | ||
215 | while (cp < cpe) { | |
216 | if (*cp++ != '/') | |
217 | continue; | |
218 | depth++; | |
219 | if (depth > max_depth) | |
220 | return 0; | |
221 | } | |
222 | return 1; | |
223 | } | |
224 | ||
225 | /* | |
226 | * Read the contents of the blob with the given OID into a buffer. | |
227 | * Append a trailing LF to the end if the last line doesn't have one. | |
228 | * | |
229 | * Returns: | |
230 | * -1 when the OID is invalid or unknown or does not refer to a blob. | |
231 | * 0 when the blob is empty. | |
232 | * 1 along with { data, size } of the (possibly augmented) buffer | |
233 | * when successful. | |
234 | * | |
235 | * Optionally updates the given oid_stat with the given OID (when valid). | |
236 | */ | |
237 | static int do_read_blob(const struct object_id *oid, struct oid_stat *oid_stat, | |
238 | size_t *size_out, char **data_out) | |
239 | { | |
240 | enum object_type type; | |
241 | unsigned long sz; | |
242 | char *data; | |
243 | ||
244 | *size_out = 0; | |
245 | *data_out = NULL; | |
246 | ||
247 | data = read_object_file(oid, &type, &sz); | |
248 | if (!data || type != OBJ_BLOB) { | |
249 | free(data); | |
250 | return -1; | |
251 | } | |
252 | ||
253 | if (oid_stat) { | |
254 | memset(&oid_stat->stat, 0, sizeof(oid_stat->stat)); | |
255 | oidcpy(&oid_stat->oid, oid); | |
256 | } | |
257 | ||
258 | if (sz == 0) { | |
259 | free(data); | |
260 | return 0; | |
261 | } | |
262 | ||
263 | if (data[sz - 1] != '\n') { | |
264 | data = xrealloc(data, st_add(sz, 1)); | |
265 | data[sz++] = '\n'; | |
266 | } | |
267 | ||
268 | *size_out = xsize_t(sz); | |
269 | *data_out = data; | |
270 | ||
271 | return 1; | |
272 | } | |
273 | ||
274 | #define DO_MATCH_EXCLUDE (1<<0) | |
275 | #define DO_MATCH_DIRECTORY (1<<1) | |
276 | #define DO_MATCH_LEADING_PATHSPEC (1<<2) | |
277 | ||
278 | /* | |
279 | * Does the given pathspec match the given name? A match is found if | |
280 | * | |
281 | * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or | |
282 | * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or | |
283 | * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or | |
284 | * (4) the pathspec string is exactly the same as 'name' ("EXACT"). | |
285 | * | |
286 | * Return value tells which case it was (1-4), or 0 when there is no match. | |
287 | * | |
288 | * It may be instructive to look at a small table of concrete examples | |
289 | * to understand the differences between 1, 2, and 4: | |
290 | * | |
291 | * Pathspecs | |
292 | * | a/b | a/b/ | a/b/c | |
293 | * ------+-----------+-----------+------------ | |
294 | * a/b | EXACT | EXACT[1] | LEADING[2] | |
295 | * Names a/b/ | RECURSIVE | EXACT | LEADING[2] | |
296 | * a/b/c | RECURSIVE | RECURSIVE | EXACT | |
297 | * | |
298 | * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match. | |
299 | * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match. | |
300 | */ | |
301 | static int match_pathspec_item(const struct index_state *istate, | |
302 | const struct pathspec_item *item, int prefix, | |
303 | const char *name, int namelen, unsigned flags) | |
304 | { | |
305 | /* name/namelen has prefix cut off by caller */ | |
306 | const char *match = item->match + prefix; | |
307 | int matchlen = item->len - prefix; | |
308 | ||
309 | /* | |
310 | * The normal call pattern is: | |
311 | * 1. prefix = common_prefix_len(ps); | |
312 | * 2. prune something, or fill_directory | |
313 | * 3. match_pathspec() | |
314 | * | |
315 | * 'prefix' at #1 may be shorter than the command's prefix and | |
316 | * it's ok for #2 to match extra files. Those extras will be | |
317 | * trimmed at #3. | |
318 | * | |
319 | * Suppose the pathspec is 'foo' and '../bar' running from | |
320 | * subdir 'xyz'. The common prefix at #1 will be empty, thanks | |
321 | * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The | |
322 | * user does not want XYZ/foo, only the "foo" part should be | |
323 | * case-insensitive. We need to filter out XYZ/foo here. In | |
324 | * other words, we do not trust the caller on comparing the | |
325 | * prefix part when :(icase) is involved. We do exact | |
326 | * comparison ourselves. | |
327 | * | |
328 | * Normally the caller (common_prefix_len() in fact) does | |
329 | * _exact_ matching on name[-prefix+1..-1] and we do not need | |
330 | * to check that part. Be defensive and check it anyway, in | |
331 | * case common_prefix_len is changed, or a new caller is | |
332 | * introduced that does not use common_prefix_len. | |
333 | * | |
334 | * If the penalty turns out too high when prefix is really | |
335 | * long, maybe change it to | |
336 | * strncmp(match, name, item->prefix - prefix) | |
337 | */ | |
338 | if (item->prefix && (item->magic & PATHSPEC_ICASE) && | |
339 | strncmp(item->match, name - prefix, item->prefix)) | |
340 | return 0; | |
341 | ||
342 | if (item->attr_match_nr && | |
343 | !match_pathspec_attrs(istate, name, namelen, item)) | |
344 | return 0; | |
345 | ||
346 | /* If the match was just the prefix, we matched */ | |
347 | if (!*match) | |
348 | return MATCHED_RECURSIVELY; | |
349 | ||
350 | if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) { | |
351 | if (matchlen == namelen) | |
352 | return MATCHED_EXACTLY; | |
353 | ||
354 | if (match[matchlen-1] == '/' || name[matchlen] == '/') | |
355 | return MATCHED_RECURSIVELY; | |
356 | } else if ((flags & DO_MATCH_DIRECTORY) && | |
357 | match[matchlen - 1] == '/' && | |
358 | namelen == matchlen - 1 && | |
359 | !ps_strncmp(item, match, name, namelen)) | |
360 | return MATCHED_EXACTLY; | |
361 | ||
362 | if (item->nowildcard_len < item->len && | |
363 | !git_fnmatch(item, match, name, | |
364 | item->nowildcard_len - prefix)) | |
365 | return MATCHED_FNMATCH; | |
366 | ||
367 | /* Perform checks to see if "name" is a leading string of the pathspec */ | |
368 | if (flags & DO_MATCH_LEADING_PATHSPEC) { | |
369 | /* name is a literal prefix of the pathspec */ | |
370 | int offset = name[namelen-1] == '/' ? 1 : 0; | |
371 | if ((namelen < matchlen) && | |
372 | (match[namelen-offset] == '/') && | |
373 | !ps_strncmp(item, match, name, namelen)) | |
374 | return MATCHED_RECURSIVELY_LEADING_PATHSPEC; | |
375 | ||
376 | /* name doesn't match up to the first wild character */ | |
377 | if (item->nowildcard_len < item->len && | |
378 | ps_strncmp(item, match, name, | |
379 | item->nowildcard_len - prefix)) | |
380 | return 0; | |
381 | ||
382 | /* | |
383 | * name has no wildcard, and it didn't match as a leading | |
384 | * pathspec so return. | |
385 | */ | |
386 | if (item->nowildcard_len == item->len) | |
387 | return 0; | |
388 | ||
389 | /* | |
390 | * Here is where we would perform a wildmatch to check if | |
391 | * "name" can be matched as a directory (or a prefix) against | |
392 | * the pathspec. Since wildmatch doesn't have this capability | |
393 | * at the present we have to punt and say that it is a match, | |
394 | * potentially returning a false positive | |
395 | * The submodules themselves will be able to perform more | |
396 | * accurate matching to determine if the pathspec matches. | |
397 | */ | |
398 | return MATCHED_RECURSIVELY_LEADING_PATHSPEC; | |
399 | } | |
400 | ||
401 | return 0; | |
402 | } | |
403 | ||
404 | /* | |
405 | * Given a name and a list of pathspecs, returns the nature of the | |
406 | * closest (i.e. most specific) match of the name to any of the | |
407 | * pathspecs. | |
408 | * | |
409 | * The caller typically calls this multiple times with the same | |
410 | * pathspec and seen[] array but with different name/namelen | |
411 | * (e.g. entries from the index) and is interested in seeing if and | |
412 | * how each pathspec matches all the names it calls this function | |
413 | * with. A mark is left in the seen[] array for each pathspec element | |
414 | * indicating the closest type of match that element achieved, so if | |
415 | * seen[n] remains zero after multiple invocations, that means the nth | |
416 | * pathspec did not match any names, which could indicate that the | |
417 | * user mistyped the nth pathspec. | |
418 | */ | |
419 | static int do_match_pathspec(const struct index_state *istate, | |
420 | const struct pathspec *ps, | |
421 | const char *name, int namelen, | |
422 | int prefix, char *seen, | |
423 | unsigned flags) | |
424 | { | |
425 | int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE; | |
426 | ||
427 | GUARD_PATHSPEC(ps, | |
428 | PATHSPEC_FROMTOP | | |
429 | PATHSPEC_MAXDEPTH | | |
430 | PATHSPEC_LITERAL | | |
431 | PATHSPEC_GLOB | | |
432 | PATHSPEC_ICASE | | |
433 | PATHSPEC_EXCLUDE | | |
434 | PATHSPEC_ATTR); | |
435 | ||
436 | if (!ps->nr) { | |
437 | if (!ps->recursive || | |
438 | !(ps->magic & PATHSPEC_MAXDEPTH) || | |
439 | ps->max_depth == -1) | |
440 | return MATCHED_RECURSIVELY; | |
441 | ||
442 | if (within_depth(name, namelen, 0, ps->max_depth)) | |
443 | return MATCHED_EXACTLY; | |
444 | else | |
445 | return 0; | |
446 | } | |
447 | ||
448 | name += prefix; | |
449 | namelen -= prefix; | |
450 | ||
451 | for (i = ps->nr - 1; i >= 0; i--) { | |
452 | int how; | |
453 | ||
454 | if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) || | |
455 | ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE))) | |
456 | continue; | |
457 | ||
458 | if (seen && seen[i] == MATCHED_EXACTLY) | |
459 | continue; | |
460 | /* | |
461 | * Make exclude patterns optional and never report | |
462 | * "pathspec ':(exclude)foo' matches no files" | |
463 | */ | |
464 | if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE) | |
465 | seen[i] = MATCHED_FNMATCH; | |
466 | how = match_pathspec_item(istate, ps->items+i, prefix, name, | |
467 | namelen, flags); | |
468 | if (ps->recursive && | |
469 | (ps->magic & PATHSPEC_MAXDEPTH) && | |
470 | ps->max_depth != -1 && | |
471 | how && how != MATCHED_FNMATCH) { | |
472 | int len = ps->items[i].len; | |
473 | if (name[len] == '/') | |
474 | len++; | |
475 | if (within_depth(name+len, namelen-len, 0, ps->max_depth)) | |
476 | how = MATCHED_EXACTLY; | |
477 | else | |
478 | how = 0; | |
479 | } | |
480 | if (how) { | |
481 | if (retval < how) | |
482 | retval = how; | |
483 | if (seen && seen[i] < how) | |
484 | seen[i] = how; | |
485 | } | |
486 | } | |
487 | return retval; | |
488 | } | |
489 | ||
490 | int match_pathspec(const struct index_state *istate, | |
491 | const struct pathspec *ps, | |
492 | const char *name, int namelen, | |
493 | int prefix, char *seen, int is_dir) | |
494 | { | |
495 | int positive, negative; | |
496 | unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0; | |
497 | positive = do_match_pathspec(istate, ps, name, namelen, | |
498 | prefix, seen, flags); | |
499 | if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive) | |
500 | return positive; | |
501 | negative = do_match_pathspec(istate, ps, name, namelen, | |
502 | prefix, seen, | |
503 | flags | DO_MATCH_EXCLUDE); | |
504 | return negative ? 0 : positive; | |
505 | } | |
506 | ||
507 | /** | |
508 | * Check if a submodule is a superset of the pathspec | |
509 | */ | |
510 | int submodule_path_match(const struct index_state *istate, | |
511 | const struct pathspec *ps, | |
512 | const char *submodule_name, | |
513 | char *seen) | |
514 | { | |
515 | int matched = do_match_pathspec(istate, ps, submodule_name, | |
516 | strlen(submodule_name), | |
517 | 0, seen, | |
518 | DO_MATCH_DIRECTORY | | |
519 | DO_MATCH_LEADING_PATHSPEC); | |
520 | return matched; | |
521 | } | |
522 | ||
523 | int report_path_error(const char *ps_matched, | |
524 | const struct pathspec *pathspec) | |
525 | { | |
526 | /* | |
527 | * Make sure all pathspec matched; otherwise it is an error. | |
528 | */ | |
529 | int num, errors = 0; | |
530 | for (num = 0; num < pathspec->nr; num++) { | |
531 | int other, found_dup; | |
532 | ||
533 | if (ps_matched[num]) | |
534 | continue; | |
535 | /* | |
536 | * The caller might have fed identical pathspec | |
537 | * twice. Do not barf on such a mistake. | |
538 | * FIXME: parse_pathspec should have eliminated | |
539 | * duplicate pathspec. | |
540 | */ | |
541 | for (found_dup = other = 0; | |
542 | !found_dup && other < pathspec->nr; | |
543 | other++) { | |
544 | if (other == num || !ps_matched[other]) | |
545 | continue; | |
546 | if (!strcmp(pathspec->items[other].original, | |
547 | pathspec->items[num].original)) | |
548 | /* | |
549 | * Ok, we have a match already. | |
550 | */ | |
551 | found_dup = 1; | |
552 | } | |
553 | if (found_dup) | |
554 | continue; | |
555 | ||
556 | error(_("pathspec '%s' did not match any file(s) known to git"), | |
557 | pathspec->items[num].original); | |
558 | errors++; | |
559 | } | |
560 | return errors; | |
561 | } | |
562 | ||
563 | /* | |
564 | * Return the length of the "simple" part of a path match limiter. | |
565 | */ | |
566 | int simple_length(const char *match) | |
567 | { | |
568 | int len = -1; | |
569 | ||
570 | for (;;) { | |
571 | unsigned char c = *match++; | |
572 | len++; | |
573 | if (c == '\0' || is_glob_special(c)) | |
574 | return len; | |
575 | } | |
576 | } | |
577 | ||
578 | int no_wildcard(const char *string) | |
579 | { | |
580 | return string[simple_length(string)] == '\0'; | |
581 | } | |
582 | ||
583 | void parse_path_pattern(const char **pattern, | |
584 | int *patternlen, | |
585 | unsigned *flags, | |
586 | int *nowildcardlen) | |
587 | { | |
588 | const char *p = *pattern; | |
589 | size_t i, len; | |
590 | ||
591 | *flags = 0; | |
592 | if (*p == '!') { | |
593 | *flags |= PATTERN_FLAG_NEGATIVE; | |
594 | p++; | |
595 | } | |
596 | len = strlen(p); | |
597 | if (len && p[len - 1] == '/') { | |
598 | len--; | |
599 | *flags |= PATTERN_FLAG_MUSTBEDIR; | |
600 | } | |
601 | for (i = 0; i < len; i++) { | |
602 | if (p[i] == '/') | |
603 | break; | |
604 | } | |
605 | if (i == len) | |
606 | *flags |= PATTERN_FLAG_NODIR; | |
607 | *nowildcardlen = simple_length(p); | |
608 | /* | |
609 | * we should have excluded the trailing slash from 'p' too, | |
610 | * but that's one more allocation. Instead just make sure | |
611 | * nowildcardlen does not exceed real patternlen | |
612 | */ | |
613 | if (*nowildcardlen > len) | |
614 | *nowildcardlen = len; | |
615 | if (*p == '*' && no_wildcard(p + 1)) | |
616 | *flags |= PATTERN_FLAG_ENDSWITH; | |
617 | *pattern = p; | |
618 | *patternlen = len; | |
619 | } | |
620 | ||
621 | void add_pattern(const char *string, const char *base, | |
622 | int baselen, struct pattern_list *pl, int srcpos) | |
623 | { | |
624 | struct path_pattern *pattern; | |
625 | int patternlen; | |
626 | unsigned flags; | |
627 | int nowildcardlen; | |
628 | ||
629 | parse_path_pattern(&string, &patternlen, &flags, &nowildcardlen); | |
630 | if (flags & PATTERN_FLAG_MUSTBEDIR) { | |
631 | FLEXPTR_ALLOC_MEM(pattern, pattern, string, patternlen); | |
632 | } else { | |
633 | pattern = xmalloc(sizeof(*pattern)); | |
634 | pattern->pattern = string; | |
635 | } | |
636 | pattern->patternlen = patternlen; | |
637 | pattern->nowildcardlen = nowildcardlen; | |
638 | pattern->base = base; | |
639 | pattern->baselen = baselen; | |
640 | pattern->flags = flags; | |
641 | pattern->srcpos = srcpos; | |
642 | ALLOC_GROW(pl->patterns, pl->nr + 1, pl->alloc); | |
643 | pl->patterns[pl->nr++] = pattern; | |
644 | pattern->pl = pl; | |
645 | } | |
646 | ||
647 | static int read_skip_worktree_file_from_index(const struct index_state *istate, | |
648 | const char *path, | |
649 | size_t *size_out, char **data_out, | |
650 | struct oid_stat *oid_stat) | |
651 | { | |
652 | int pos, len; | |
653 | ||
654 | len = strlen(path); | |
655 | pos = index_name_pos(istate, path, len); | |
656 | if (pos < 0) | |
657 | return -1; | |
658 | if (!ce_skip_worktree(istate->cache[pos])) | |
659 | return -1; | |
660 | ||
661 | return do_read_blob(&istate->cache[pos]->oid, oid_stat, size_out, data_out); | |
662 | } | |
663 | ||
664 | /* | |
665 | * Frees memory within pl which was allocated for exclude patterns and | |
666 | * the file buffer. Does not free pl itself. | |
667 | */ | |
668 | void clear_pattern_list(struct pattern_list *pl) | |
669 | { | |
670 | int i; | |
671 | ||
672 | for (i = 0; i < pl->nr; i++) | |
673 | free(pl->patterns[i]); | |
674 | free(pl->patterns); | |
675 | free(pl->filebuf); | |
676 | ||
677 | memset(pl, 0, sizeof(*pl)); | |
678 | } | |
679 | ||
680 | static void trim_trailing_spaces(char *buf) | |
681 | { | |
682 | char *p, *last_space = NULL; | |
683 | ||
684 | for (p = buf; *p; p++) | |
685 | switch (*p) { | |
686 | case ' ': | |
687 | if (!last_space) | |
688 | last_space = p; | |
689 | break; | |
690 | case '\\': | |
691 | p++; | |
692 | if (!*p) | |
693 | return; | |
694 | /* fallthrough */ | |
695 | default: | |
696 | last_space = NULL; | |
697 | } | |
698 | ||
699 | if (last_space) | |
700 | *last_space = '\0'; | |
701 | } | |
702 | ||
703 | /* | |
704 | * Given a subdirectory name and "dir" of the current directory, | |
705 | * search the subdir in "dir" and return it, or create a new one if it | |
706 | * does not exist in "dir". | |
707 | * | |
708 | * If "name" has the trailing slash, it'll be excluded in the search. | |
709 | */ | |
710 | static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc, | |
711 | struct untracked_cache_dir *dir, | |
712 | const char *name, int len) | |
713 | { | |
714 | int first, last; | |
715 | struct untracked_cache_dir *d; | |
716 | if (!dir) | |
717 | return NULL; | |
718 | if (len && name[len - 1] == '/') | |
719 | len--; | |
720 | first = 0; | |
721 | last = dir->dirs_nr; | |
722 | while (last > first) { | |
723 | int cmp, next = first + ((last - first) >> 1); | |
724 | d = dir->dirs[next]; | |
725 | cmp = strncmp(name, d->name, len); | |
726 | if (!cmp && strlen(d->name) > len) | |
727 | cmp = -1; | |
728 | if (!cmp) | |
729 | return d; | |
730 | if (cmp < 0) { | |
731 | last = next; | |
732 | continue; | |
733 | } | |
734 | first = next+1; | |
735 | } | |
736 | ||
737 | uc->dir_created++; | |
738 | FLEX_ALLOC_MEM(d, name, name, len); | |
739 | ||
740 | ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc); | |
741 | MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first, | |
742 | dir->dirs_nr - first); | |
743 | dir->dirs_nr++; | |
744 | dir->dirs[first] = d; | |
745 | return d; | |
746 | } | |
747 | ||
748 | static void do_invalidate_gitignore(struct untracked_cache_dir *dir) | |
749 | { | |
750 | int i; | |
751 | dir->valid = 0; | |
752 | dir->untracked_nr = 0; | |
753 | for (i = 0; i < dir->dirs_nr; i++) | |
754 | do_invalidate_gitignore(dir->dirs[i]); | |
755 | } | |
756 | ||
757 | static void invalidate_gitignore(struct untracked_cache *uc, | |
758 | struct untracked_cache_dir *dir) | |
759 | { | |
760 | uc->gitignore_invalidated++; | |
761 | do_invalidate_gitignore(dir); | |
762 | } | |
763 | ||
764 | static void invalidate_directory(struct untracked_cache *uc, | |
765 | struct untracked_cache_dir *dir) | |
766 | { | |
767 | int i; | |
768 | ||
769 | /* | |
770 | * Invalidation increment here is just roughly correct. If | |
771 | * untracked_nr or any of dirs[].recurse is non-zero, we | |
772 | * should increment dir_invalidated too. But that's more | |
773 | * expensive to do. | |
774 | */ | |
775 | if (dir->valid) | |
776 | uc->dir_invalidated++; | |
777 | ||
778 | dir->valid = 0; | |
779 | dir->untracked_nr = 0; | |
780 | for (i = 0; i < dir->dirs_nr; i++) | |
781 | dir->dirs[i]->recurse = 0; | |
782 | } | |
783 | ||
784 | static int add_patterns_from_buffer(char *buf, size_t size, | |
785 | const char *base, int baselen, | |
786 | struct pattern_list *pl); | |
787 | ||
788 | /* | |
789 | * Given a file with name "fname", read it (either from disk, or from | |
790 | * an index if 'istate' is non-null), parse it and store the | |
791 | * exclude rules in "pl". | |
792 | * | |
793 | * If "ss" is not NULL, compute SHA-1 of the exclude file and fill | |
794 | * stat data from disk (only valid if add_patterns returns zero). If | |
795 | * ss_valid is non-zero, "ss" must contain good value as input. | |
796 | */ | |
797 | static int add_patterns(const char *fname, const char *base, int baselen, | |
798 | struct pattern_list *pl, struct index_state *istate, | |
799 | struct oid_stat *oid_stat) | |
800 | { | |
801 | struct stat st; | |
802 | int r; | |
803 | int fd; | |
804 | size_t size = 0; | |
805 | char *buf; | |
806 | ||
807 | fd = open(fname, O_RDONLY); | |
808 | if (fd < 0 || fstat(fd, &st) < 0) { | |
809 | if (fd < 0) | |
810 | warn_on_fopen_errors(fname); | |
811 | else | |
812 | close(fd); | |
813 | if (!istate) | |
814 | return -1; | |
815 | r = read_skip_worktree_file_from_index(istate, fname, | |
816 | &size, &buf, | |
817 | oid_stat); | |
818 | if (r != 1) | |
819 | return r; | |
820 | } else { | |
821 | size = xsize_t(st.st_size); | |
822 | if (size == 0) { | |
823 | if (oid_stat) { | |
824 | fill_stat_data(&oid_stat->stat, &st); | |
825 | oidcpy(&oid_stat->oid, the_hash_algo->empty_blob); | |
826 | oid_stat->valid = 1; | |
827 | } | |
828 | close(fd); | |
829 | return 0; | |
830 | } | |
831 | buf = xmallocz(size); | |
832 | if (read_in_full(fd, buf, size) != size) { | |
833 | free(buf); | |
834 | close(fd); | |
835 | return -1; | |
836 | } | |
837 | buf[size++] = '\n'; | |
838 | close(fd); | |
839 | if (oid_stat) { | |
840 | int pos; | |
841 | if (oid_stat->valid && | |
842 | !match_stat_data_racy(istate, &oid_stat->stat, &st)) | |
843 | ; /* no content change, ss->sha1 still good */ | |
844 | else if (istate && | |
845 | (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 && | |
846 | !ce_stage(istate->cache[pos]) && | |
847 | ce_uptodate(istate->cache[pos]) && | |
848 | !would_convert_to_git(istate, fname)) | |
849 | oidcpy(&oid_stat->oid, | |
850 | &istate->cache[pos]->oid); | |
851 | else | |
852 | hash_object_file(buf, size, "blob", | |
853 | &oid_stat->oid); | |
854 | fill_stat_data(&oid_stat->stat, &st); | |
855 | oid_stat->valid = 1; | |
856 | } | |
857 | } | |
858 | ||
859 | add_patterns_from_buffer(buf, size, base, baselen, pl); | |
860 | return 0; | |
861 | } | |
862 | ||
863 | static int add_patterns_from_buffer(char *buf, size_t size, | |
864 | const char *base, int baselen, | |
865 | struct pattern_list *pl) | |
866 | { | |
867 | int i, lineno = 1; | |
868 | char *entry; | |
869 | ||
870 | pl->filebuf = buf; | |
871 | ||
872 | if (skip_utf8_bom(&buf, size)) | |
873 | size -= buf - pl->filebuf; | |
874 | ||
875 | entry = buf; | |
876 | ||
877 | for (i = 0; i < size; i++) { | |
878 | if (buf[i] == '\n') { | |
879 | if (entry != buf + i && entry[0] != '#') { | |
880 | buf[i - (i && buf[i-1] == '\r')] = 0; | |
881 | trim_trailing_spaces(entry); | |
882 | add_pattern(entry, base, baselen, pl, lineno); | |
883 | } | |
884 | lineno++; | |
885 | entry = buf + i + 1; | |
886 | } | |
887 | } | |
888 | return 0; | |
889 | } | |
890 | ||
891 | int add_patterns_from_file_to_list(const char *fname, const char *base, | |
892 | int baselen, struct pattern_list *pl, | |
893 | struct index_state *istate) | |
894 | { | |
895 | return add_patterns(fname, base, baselen, pl, istate, NULL); | |
896 | } | |
897 | ||
898 | int add_patterns_from_blob_to_list( | |
899 | struct object_id *oid, | |
900 | const char *base, int baselen, | |
901 | struct pattern_list *pl) | |
902 | { | |
903 | char *buf; | |
904 | size_t size; | |
905 | int r; | |
906 | ||
907 | r = do_read_blob(oid, NULL, &size, &buf); | |
908 | if (r != 1) | |
909 | return r; | |
910 | ||
911 | add_patterns_from_buffer(buf, size, base, baselen, pl); | |
912 | return 0; | |
913 | } | |
914 | ||
915 | struct pattern_list *add_pattern_list(struct dir_struct *dir, | |
916 | int group_type, const char *src) | |
917 | { | |
918 | struct pattern_list *pl; | |
919 | struct exclude_list_group *group; | |
920 | ||
921 | group = &dir->exclude_list_group[group_type]; | |
922 | ALLOC_GROW(group->pl, group->nr + 1, group->alloc); | |
923 | pl = &group->pl[group->nr++]; | |
924 | memset(pl, 0, sizeof(*pl)); | |
925 | pl->src = src; | |
926 | return pl; | |
927 | } | |
928 | ||
929 | /* | |
930 | * Used to set up core.excludesfile and .git/info/exclude lists. | |
931 | */ | |
932 | static void add_patterns_from_file_1(struct dir_struct *dir, const char *fname, | |
933 | struct oid_stat *oid_stat) | |
934 | { | |
935 | struct pattern_list *pl; | |
936 | /* | |
937 | * catch setup_standard_excludes() that's called before | |
938 | * dir->untracked is assigned. That function behaves | |
939 | * differently when dir->untracked is non-NULL. | |
940 | */ | |
941 | if (!dir->untracked) | |
942 | dir->unmanaged_exclude_files++; | |
943 | pl = add_pattern_list(dir, EXC_FILE, fname); | |
944 | if (add_patterns(fname, "", 0, pl, NULL, oid_stat) < 0) | |
945 | die(_("cannot use %s as an exclude file"), fname); | |
946 | } | |
947 | ||
948 | void add_patterns_from_file(struct dir_struct *dir, const char *fname) | |
949 | { | |
950 | dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */ | |
951 | add_patterns_from_file_1(dir, fname, NULL); | |
952 | } | |
953 | ||
954 | int match_basename(const char *basename, int basenamelen, | |
955 | const char *pattern, int prefix, int patternlen, | |
956 | unsigned flags) | |
957 | { | |
958 | if (prefix == patternlen) { | |
959 | if (patternlen == basenamelen && | |
960 | !fspathncmp(pattern, basename, basenamelen)) | |
961 | return 1; | |
962 | } else if (flags & PATTERN_FLAG_ENDSWITH) { | |
963 | /* "*literal" matching against "fooliteral" */ | |
964 | if (patternlen - 1 <= basenamelen && | |
965 | !fspathncmp(pattern + 1, | |
966 | basename + basenamelen - (patternlen - 1), | |
967 | patternlen - 1)) | |
968 | return 1; | |
969 | } else { | |
970 | if (fnmatch_icase_mem(pattern, patternlen, | |
971 | basename, basenamelen, | |
972 | 0) == 0) | |
973 | return 1; | |
974 | } | |
975 | return 0; | |
976 | } | |
977 | ||
978 | int match_pathname(const char *pathname, int pathlen, | |
979 | const char *base, int baselen, | |
980 | const char *pattern, int prefix, int patternlen, | |
981 | unsigned flags) | |
982 | { | |
983 | const char *name; | |
984 | int namelen; | |
985 | ||
986 | /* | |
987 | * match with FNM_PATHNAME; the pattern has base implicitly | |
988 | * in front of it. | |
989 | */ | |
990 | if (*pattern == '/') { | |
991 | pattern++; | |
992 | patternlen--; | |
993 | prefix--; | |
994 | } | |
995 | ||
996 | /* | |
997 | * baselen does not count the trailing slash. base[] may or | |
998 | * may not end with a trailing slash though. | |
999 | */ | |
1000 | if (pathlen < baselen + 1 || | |
1001 | (baselen && pathname[baselen] != '/') || | |
1002 | fspathncmp(pathname, base, baselen)) | |
1003 | return 0; | |
1004 | ||
1005 | namelen = baselen ? pathlen - baselen - 1 : pathlen; | |
1006 | name = pathname + pathlen - namelen; | |
1007 | ||
1008 | if (prefix) { | |
1009 | /* | |
1010 | * if the non-wildcard part is longer than the | |
1011 | * remaining pathname, surely it cannot match. | |
1012 | */ | |
1013 | if (prefix > namelen) | |
1014 | return 0; | |
1015 | ||
1016 | if (fspathncmp(pattern, name, prefix)) | |
1017 | return 0; | |
1018 | pattern += prefix; | |
1019 | patternlen -= prefix; | |
1020 | name += prefix; | |
1021 | namelen -= prefix; | |
1022 | ||
1023 | /* | |
1024 | * If the whole pattern did not have a wildcard, | |
1025 | * then our prefix match is all we need; we | |
1026 | * do not need to call fnmatch at all. | |
1027 | */ | |
1028 | if (!patternlen && !namelen) | |
1029 | return 1; | |
1030 | } | |
1031 | ||
1032 | return fnmatch_icase_mem(pattern, patternlen, | |
1033 | name, namelen, | |
1034 | WM_PATHNAME) == 0; | |
1035 | } | |
1036 | ||
1037 | /* | |
1038 | * Scan the given exclude list in reverse to see whether pathname | |
1039 | * should be ignored. The first match (i.e. the last on the list), if | |
1040 | * any, determines the fate. Returns the exclude_list element which | |
1041 | * matched, or NULL for undecided. | |
1042 | */ | |
1043 | static struct path_pattern *last_matching_pattern_from_list(const char *pathname, | |
1044 | int pathlen, | |
1045 | const char *basename, | |
1046 | int *dtype, | |
1047 | struct pattern_list *pl, | |
1048 | struct index_state *istate) | |
1049 | { | |
1050 | struct path_pattern *res = NULL; /* undecided */ | |
1051 | int i; | |
1052 | ||
1053 | if (!pl->nr) | |
1054 | return NULL; /* undefined */ | |
1055 | ||
1056 | for (i = pl->nr - 1; 0 <= i; i--) { | |
1057 | struct path_pattern *pattern = pl->patterns[i]; | |
1058 | const char *exclude = pattern->pattern; | |
1059 | int prefix = pattern->nowildcardlen; | |
1060 | ||
1061 | if (pattern->flags & PATTERN_FLAG_MUSTBEDIR) { | |
1062 | if (*dtype == DT_UNKNOWN) | |
1063 | *dtype = get_dtype(NULL, istate, pathname, pathlen); | |
1064 | if (*dtype != DT_DIR) | |
1065 | continue; | |
1066 | } | |
1067 | ||
1068 | if (pattern->flags & PATTERN_FLAG_NODIR) { | |
1069 | if (match_basename(basename, | |
1070 | pathlen - (basename - pathname), | |
1071 | exclude, prefix, pattern->patternlen, | |
1072 | pattern->flags)) { | |
1073 | res = pattern; | |
1074 | break; | |
1075 | } | |
1076 | continue; | |
1077 | } | |
1078 | ||
1079 | assert(pattern->baselen == 0 || | |
1080 | pattern->base[pattern->baselen - 1] == '/'); | |
1081 | if (match_pathname(pathname, pathlen, | |
1082 | pattern->base, | |
1083 | pattern->baselen ? pattern->baselen - 1 : 0, | |
1084 | exclude, prefix, pattern->patternlen, | |
1085 | pattern->flags)) { | |
1086 | res = pattern; | |
1087 | break; | |
1088 | } | |
1089 | } | |
1090 | return res; | |
1091 | } | |
1092 | ||
1093 | /* | |
1094 | * Scan the list of patterns to determine if the ordered list | |
1095 | * of patterns matches on 'pathname'. | |
1096 | * | |
1097 | * Return 1 for a match, 0 for not matched and -1 for undecided. | |
1098 | */ | |
1099 | enum pattern_match_result path_matches_pattern_list( | |
1100 | const char *pathname, int pathlen, | |
1101 | const char *basename, int *dtype, | |
1102 | struct pattern_list *pl, | |
1103 | struct index_state *istate) | |
1104 | { | |
1105 | struct path_pattern *pattern; | |
1106 | pattern = last_matching_pattern_from_list(pathname, pathlen, basename, | |
1107 | dtype, pl, istate); | |
1108 | if (pattern) { | |
1109 | if (pattern->flags & PATTERN_FLAG_NEGATIVE) | |
1110 | return NOT_MATCHED; | |
1111 | else | |
1112 | return MATCHED; | |
1113 | } | |
1114 | ||
1115 | return UNDECIDED; | |
1116 | } | |
1117 | ||
1118 | static struct path_pattern *last_matching_pattern_from_lists( | |
1119 | struct dir_struct *dir, struct index_state *istate, | |
1120 | const char *pathname, int pathlen, | |
1121 | const char *basename, int *dtype_p) | |
1122 | { | |
1123 | int i, j; | |
1124 | struct exclude_list_group *group; | |
1125 | struct path_pattern *pattern; | |
1126 | for (i = EXC_CMDL; i <= EXC_FILE; i++) { | |
1127 | group = &dir->exclude_list_group[i]; | |
1128 | for (j = group->nr - 1; j >= 0; j--) { | |
1129 | pattern = last_matching_pattern_from_list( | |
1130 | pathname, pathlen, basename, dtype_p, | |
1131 | &group->pl[j], istate); | |
1132 | if (pattern) | |
1133 | return pattern; | |
1134 | } | |
1135 | } | |
1136 | return NULL; | |
1137 | } | |
1138 | ||
1139 | /* | |
1140 | * Loads the per-directory exclude list for the substring of base | |
1141 | * which has a char length of baselen. | |
1142 | */ | |
1143 | static void prep_exclude(struct dir_struct *dir, | |
1144 | struct index_state *istate, | |
1145 | const char *base, int baselen) | |
1146 | { | |
1147 | struct exclude_list_group *group; | |
1148 | struct pattern_list *pl; | |
1149 | struct exclude_stack *stk = NULL; | |
1150 | struct untracked_cache_dir *untracked; | |
1151 | int current; | |
1152 | ||
1153 | group = &dir->exclude_list_group[EXC_DIRS]; | |
1154 | ||
1155 | /* | |
1156 | * Pop the exclude lists from the EXCL_DIRS exclude_list_group | |
1157 | * which originate from directories not in the prefix of the | |
1158 | * path being checked. | |
1159 | */ | |
1160 | while ((stk = dir->exclude_stack) != NULL) { | |
1161 | if (stk->baselen <= baselen && | |
1162 | !strncmp(dir->basebuf.buf, base, stk->baselen)) | |
1163 | break; | |
1164 | pl = &group->pl[dir->exclude_stack->exclude_ix]; | |
1165 | dir->exclude_stack = stk->prev; | |
1166 | dir->pattern = NULL; | |
1167 | free((char *)pl->src); /* see strbuf_detach() below */ | |
1168 | clear_pattern_list(pl); | |
1169 | free(stk); | |
1170 | group->nr--; | |
1171 | } | |
1172 | ||
1173 | /* Skip traversing into sub directories if the parent is excluded */ | |
1174 | if (dir->pattern) | |
1175 | return; | |
1176 | ||
1177 | /* | |
1178 | * Lazy initialization. All call sites currently just | |
1179 | * memset(dir, 0, sizeof(*dir)) before use. Changing all of | |
1180 | * them seems lots of work for little benefit. | |
1181 | */ | |
1182 | if (!dir->basebuf.buf) | |
1183 | strbuf_init(&dir->basebuf, PATH_MAX); | |
1184 | ||
1185 | /* Read from the parent directories and push them down. */ | |
1186 | current = stk ? stk->baselen : -1; | |
1187 | strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current); | |
1188 | if (dir->untracked) | |
1189 | untracked = stk ? stk->ucd : dir->untracked->root; | |
1190 | else | |
1191 | untracked = NULL; | |
1192 | ||
1193 | while (current < baselen) { | |
1194 | const char *cp; | |
1195 | struct oid_stat oid_stat; | |
1196 | ||
1197 | stk = xcalloc(1, sizeof(*stk)); | |
1198 | if (current < 0) { | |
1199 | cp = base; | |
1200 | current = 0; | |
1201 | } else { | |
1202 | cp = strchr(base + current + 1, '/'); | |
1203 | if (!cp) | |
1204 | die("oops in prep_exclude"); | |
1205 | cp++; | |
1206 | untracked = | |
1207 | lookup_untracked(dir->untracked, untracked, | |
1208 | base + current, | |
1209 | cp - base - current); | |
1210 | } | |
1211 | stk->prev = dir->exclude_stack; | |
1212 | stk->baselen = cp - base; | |
1213 | stk->exclude_ix = group->nr; | |
1214 | stk->ucd = untracked; | |
1215 | pl = add_pattern_list(dir, EXC_DIRS, NULL); | |
1216 | strbuf_add(&dir->basebuf, base + current, stk->baselen - current); | |
1217 | assert(stk->baselen == dir->basebuf.len); | |
1218 | ||
1219 | /* Abort if the directory is excluded */ | |
1220 | if (stk->baselen) { | |
1221 | int dt = DT_DIR; | |
1222 | dir->basebuf.buf[stk->baselen - 1] = 0; | |
1223 | dir->pattern = last_matching_pattern_from_lists(dir, | |
1224 | istate, | |
1225 | dir->basebuf.buf, stk->baselen - 1, | |
1226 | dir->basebuf.buf + current, &dt); | |
1227 | dir->basebuf.buf[stk->baselen - 1] = '/'; | |
1228 | if (dir->pattern && | |
1229 | dir->pattern->flags & PATTERN_FLAG_NEGATIVE) | |
1230 | dir->pattern = NULL; | |
1231 | if (dir->pattern) { | |
1232 | dir->exclude_stack = stk; | |
1233 | return; | |
1234 | } | |
1235 | } | |
1236 | ||
1237 | /* Try to read per-directory file */ | |
1238 | oidclr(&oid_stat.oid); | |
1239 | oid_stat.valid = 0; | |
1240 | if (dir->exclude_per_dir && | |
1241 | /* | |
1242 | * If we know that no files have been added in | |
1243 | * this directory (i.e. valid_cached_dir() has | |
1244 | * been executed and set untracked->valid) .. | |
1245 | */ | |
1246 | (!untracked || !untracked->valid || | |
1247 | /* | |
1248 | * .. and .gitignore does not exist before | |
1249 | * (i.e. null exclude_oid). Then we can skip | |
1250 | * loading .gitignore, which would result in | |
1251 | * ENOENT anyway. | |
1252 | */ | |
1253 | !is_null_oid(&untracked->exclude_oid))) { | |
1254 | /* | |
1255 | * dir->basebuf gets reused by the traversal, but we | |
1256 | * need fname to remain unchanged to ensure the src | |
1257 | * member of each struct path_pattern correctly | |
1258 | * back-references its source file. Other invocations | |
1259 | * of add_pattern_list provide stable strings, so we | |
1260 | * strbuf_detach() and free() here in the caller. | |
1261 | */ | |
1262 | struct strbuf sb = STRBUF_INIT; | |
1263 | strbuf_addbuf(&sb, &dir->basebuf); | |
1264 | strbuf_addstr(&sb, dir->exclude_per_dir); | |
1265 | pl->src = strbuf_detach(&sb, NULL); | |
1266 | add_patterns(pl->src, pl->src, stk->baselen, pl, istate, | |
1267 | untracked ? &oid_stat : NULL); | |
1268 | } | |
1269 | /* | |
1270 | * NEEDSWORK: when untracked cache is enabled, prep_exclude() | |
1271 | * will first be called in valid_cached_dir() then maybe many | |
1272 | * times more in last_matching_pattern(). When the cache is | |
1273 | * used, last_matching_pattern() will not be called and | |
1274 | * reading .gitignore content will be a waste. | |
1275 | * | |
1276 | * So when it's called by valid_cached_dir() and we can get | |
1277 | * .gitignore SHA-1 from the index (i.e. .gitignore is not | |
1278 | * modified on work tree), we could delay reading the | |
1279 | * .gitignore content until we absolutely need it in | |
1280 | * last_matching_pattern(). Be careful about ignore rule | |
1281 | * order, though, if you do that. | |
1282 | */ | |
1283 | if (untracked && | |
1284 | !oideq(&oid_stat.oid, &untracked->exclude_oid)) { | |
1285 | invalidate_gitignore(dir->untracked, untracked); | |
1286 | oidcpy(&untracked->exclude_oid, &oid_stat.oid); | |
1287 | } | |
1288 | dir->exclude_stack = stk; | |
1289 | current = stk->baselen; | |
1290 | } | |
1291 | strbuf_setlen(&dir->basebuf, baselen); | |
1292 | } | |
1293 | ||
1294 | /* | |
1295 | * Loads the exclude lists for the directory containing pathname, then | |
1296 | * scans all exclude lists to determine whether pathname is excluded. | |
1297 | * Returns the exclude_list element which matched, or NULL for | |
1298 | * undecided. | |
1299 | */ | |
1300 | struct path_pattern *last_matching_pattern(struct dir_struct *dir, | |
1301 | struct index_state *istate, | |
1302 | const char *pathname, | |
1303 | int *dtype_p) | |
1304 | { | |
1305 | int pathlen = strlen(pathname); | |
1306 | const char *basename = strrchr(pathname, '/'); | |
1307 | basename = (basename) ? basename+1 : pathname; | |
1308 | ||
1309 | prep_exclude(dir, istate, pathname, basename-pathname); | |
1310 | ||
1311 | if (dir->pattern) | |
1312 | return dir->pattern; | |
1313 | ||
1314 | return last_matching_pattern_from_lists(dir, istate, pathname, pathlen, | |
1315 | basename, dtype_p); | |
1316 | } | |
1317 | ||
1318 | /* | |
1319 | * Loads the exclude lists for the directory containing pathname, then | |
1320 | * scans all exclude lists to determine whether pathname is excluded. | |
1321 | * Returns 1 if true, otherwise 0. | |
1322 | */ | |
1323 | int is_excluded(struct dir_struct *dir, struct index_state *istate, | |
1324 | const char *pathname, int *dtype_p) | |
1325 | { | |
1326 | struct path_pattern *pattern = | |
1327 | last_matching_pattern(dir, istate, pathname, dtype_p); | |
1328 | if (pattern) | |
1329 | return pattern->flags & PATTERN_FLAG_NEGATIVE ? 0 : 1; | |
1330 | return 0; | |
1331 | } | |
1332 | ||
1333 | static struct dir_entry *dir_entry_new(const char *pathname, int len) | |
1334 | { | |
1335 | struct dir_entry *ent; | |
1336 | ||
1337 | FLEX_ALLOC_MEM(ent, name, pathname, len); | |
1338 | ent->len = len; | |
1339 | return ent; | |
1340 | } | |
1341 | ||
1342 | static struct dir_entry *dir_add_name(struct dir_struct *dir, | |
1343 | struct index_state *istate, | |
1344 | const char *pathname, int len) | |
1345 | { | |
1346 | if (index_file_exists(istate, pathname, len, ignore_case)) | |
1347 | return NULL; | |
1348 | ||
1349 | ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc); | |
1350 | return dir->entries[dir->nr++] = dir_entry_new(pathname, len); | |
1351 | } | |
1352 | ||
1353 | struct dir_entry *dir_add_ignored(struct dir_struct *dir, | |
1354 | struct index_state *istate, | |
1355 | const char *pathname, int len) | |
1356 | { | |
1357 | if (!index_name_is_other(istate, pathname, len)) | |
1358 | return NULL; | |
1359 | ||
1360 | ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc); | |
1361 | return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len); | |
1362 | } | |
1363 | ||
1364 | enum exist_status { | |
1365 | index_nonexistent = 0, | |
1366 | index_directory, | |
1367 | index_gitdir | |
1368 | }; | |
1369 | ||
1370 | /* | |
1371 | * Do not use the alphabetically sorted index to look up | |
1372 | * the directory name; instead, use the case insensitive | |
1373 | * directory hash. | |
1374 | */ | |
1375 | static enum exist_status directory_exists_in_index_icase(struct index_state *istate, | |
1376 | const char *dirname, int len) | |
1377 | { | |
1378 | struct cache_entry *ce; | |
1379 | ||
1380 | if (index_dir_exists(istate, dirname, len)) | |
1381 | return index_directory; | |
1382 | ||
1383 | ce = index_file_exists(istate, dirname, len, ignore_case); | |
1384 | if (ce && S_ISGITLINK(ce->ce_mode)) | |
1385 | return index_gitdir; | |
1386 | ||
1387 | return index_nonexistent; | |
1388 | } | |
1389 | ||
1390 | /* | |
1391 | * The index sorts alphabetically by entry name, which | |
1392 | * means that a gitlink sorts as '\0' at the end, while | |
1393 | * a directory (which is defined not as an entry, but as | |
1394 | * the files it contains) will sort with the '/' at the | |
1395 | * end. | |
1396 | */ | |
1397 | static enum exist_status directory_exists_in_index(struct index_state *istate, | |
1398 | const char *dirname, int len) | |
1399 | { | |
1400 | int pos; | |
1401 | ||
1402 | if (ignore_case) | |
1403 | return directory_exists_in_index_icase(istate, dirname, len); | |
1404 | ||
1405 | pos = index_name_pos(istate, dirname, len); | |
1406 | if (pos < 0) | |
1407 | pos = -pos-1; | |
1408 | while (pos < istate->cache_nr) { | |
1409 | const struct cache_entry *ce = istate->cache[pos++]; | |
1410 | unsigned char endchar; | |
1411 | ||
1412 | if (strncmp(ce->name, dirname, len)) | |
1413 | break; | |
1414 | endchar = ce->name[len]; | |
1415 | if (endchar > '/') | |
1416 | break; | |
1417 | if (endchar == '/') | |
1418 | return index_directory; | |
1419 | if (!endchar && S_ISGITLINK(ce->ce_mode)) | |
1420 | return index_gitdir; | |
1421 | } | |
1422 | return index_nonexistent; | |
1423 | } | |
1424 | ||
1425 | /* | |
1426 | * When we find a directory when traversing the filesystem, we | |
1427 | * have three distinct cases: | |
1428 | * | |
1429 | * - ignore it | |
1430 | * - see it as a directory | |
1431 | * - recurse into it | |
1432 | * | |
1433 | * and which one we choose depends on a combination of existing | |
1434 | * git index contents and the flags passed into the directory | |
1435 | * traversal routine. | |
1436 | * | |
1437 | * Case 1: If we *already* have entries in the index under that | |
1438 | * directory name, we always recurse into the directory to see | |
1439 | * all the files. | |
1440 | * | |
1441 | * Case 2: If we *already* have that directory name as a gitlink, | |
1442 | * we always continue to see it as a gitlink, regardless of whether | |
1443 | * there is an actual git directory there or not (it might not | |
1444 | * be checked out as a subproject!) | |
1445 | * | |
1446 | * Case 3: if we didn't have it in the index previously, we | |
1447 | * have a few sub-cases: | |
1448 | * | |
1449 | * (a) if "show_other_directories" is true, we show it as | |
1450 | * just a directory, unless "hide_empty_directories" is | |
1451 | * also true, in which case we need to check if it contains any | |
1452 | * untracked and / or ignored files. | |
1453 | * (b) if it looks like a git directory, and we don't have | |
1454 | * 'no_gitlinks' set we treat it as a gitlink, and show it | |
1455 | * as a directory. | |
1456 | * (c) otherwise, we recurse into it. | |
1457 | */ | |
1458 | static enum path_treatment treat_directory(struct dir_struct *dir, | |
1459 | struct index_state *istate, | |
1460 | struct untracked_cache_dir *untracked, | |
1461 | const char *dirname, int len, int baselen, int exclude, | |
1462 | const struct pathspec *pathspec) | |
1463 | { | |
1464 | /* The "len-1" is to strip the final '/' */ | |
1465 | switch (directory_exists_in_index(istate, dirname, len-1)) { | |
1466 | case index_directory: | |
1467 | return path_recurse; | |
1468 | ||
1469 | case index_gitdir: | |
1470 | return path_none; | |
1471 | ||
1472 | case index_nonexistent: | |
1473 | if (dir->flags & DIR_SKIP_NESTED_GIT) { | |
1474 | int nested_repo; | |
1475 | struct strbuf sb = STRBUF_INIT; | |
1476 | strbuf_addstr(&sb, dirname); | |
1477 | nested_repo = is_nonbare_repository_dir(&sb); | |
1478 | strbuf_release(&sb); | |
1479 | if (nested_repo) | |
1480 | return path_none; | |
1481 | } | |
1482 | ||
1483 | if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES) | |
1484 | break; | |
1485 | if (exclude && | |
1486 | (dir->flags & DIR_SHOW_IGNORED_TOO) && | |
1487 | (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) { | |
1488 | ||
1489 | /* | |
1490 | * This is an excluded directory and we are | |
1491 | * showing ignored paths that match an exclude | |
1492 | * pattern. (e.g. show directory as ignored | |
1493 | * only if it matches an exclude pattern). | |
1494 | * This path will either be 'path_excluded` | |
1495 | * (if we are showing empty directories or if | |
1496 | * the directory is not empty), or will be | |
1497 | * 'path_none' (empty directory, and we are | |
1498 | * not showing empty directories). | |
1499 | */ | |
1500 | if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES)) | |
1501 | return path_excluded; | |
1502 | ||
1503 | if (read_directory_recursive(dir, istate, dirname, len, | |
1504 | untracked, 1, 1, pathspec) == path_excluded) | |
1505 | return path_excluded; | |
1506 | ||
1507 | return path_none; | |
1508 | } | |
1509 | if (!(dir->flags & DIR_NO_GITLINKS)) { | |
1510 | struct strbuf sb = STRBUF_INIT; | |
1511 | strbuf_addstr(&sb, dirname); | |
1512 | if (is_nonbare_repository_dir(&sb)) | |
1513 | return exclude ? path_excluded : path_untracked; | |
1514 | strbuf_release(&sb); | |
1515 | } | |
1516 | return path_recurse; | |
1517 | } | |
1518 | ||
1519 | /* This is the "show_other_directories" case */ | |
1520 | ||
1521 | if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES)) | |
1522 | return exclude ? path_excluded : path_untracked; | |
1523 | ||
1524 | untracked = lookup_untracked(dir->untracked, untracked, | |
1525 | dirname + baselen, len - baselen); | |
1526 | ||
1527 | /* | |
1528 | * If this is an excluded directory, then we only need to check if | |
1529 | * the directory contains any files. | |
1530 | */ | |
1531 | return read_directory_recursive(dir, istate, dirname, len, | |
1532 | untracked, 1, exclude, pathspec); | |
1533 | } | |
1534 | ||
1535 | /* | |
1536 | * This is an inexact early pruning of any recursive directory | |
1537 | * reading - if the path cannot possibly be in the pathspec, | |
1538 | * return true, and we'll skip it early. | |
1539 | */ | |
1540 | static int simplify_away(const char *path, int pathlen, | |
1541 | const struct pathspec *pathspec) | |
1542 | { | |
1543 | int i; | |
1544 | ||
1545 | if (!pathspec || !pathspec->nr) | |
1546 | return 0; | |
1547 | ||
1548 | GUARD_PATHSPEC(pathspec, | |
1549 | PATHSPEC_FROMTOP | | |
1550 | PATHSPEC_MAXDEPTH | | |
1551 | PATHSPEC_LITERAL | | |
1552 | PATHSPEC_GLOB | | |
1553 | PATHSPEC_ICASE | | |
1554 | PATHSPEC_EXCLUDE | | |
1555 | PATHSPEC_ATTR); | |
1556 | ||
1557 | for (i = 0; i < pathspec->nr; i++) { | |
1558 | const struct pathspec_item *item = &pathspec->items[i]; | |
1559 | int len = item->nowildcard_len; | |
1560 | ||
1561 | if (len > pathlen) | |
1562 | len = pathlen; | |
1563 | if (!ps_strncmp(item, item->match, path, len)) | |
1564 | return 0; | |
1565 | } | |
1566 | ||
1567 | return 1; | |
1568 | } | |
1569 | ||
1570 | /* | |
1571 | * This function tells us whether an excluded path matches a | |
1572 | * list of "interesting" pathspecs. That is, whether a path matched | |
1573 | * by any of the pathspecs could possibly be ignored by excluding | |
1574 | * the specified path. This can happen if: | |
1575 | * | |
1576 | * 1. the path is mentioned explicitly in the pathspec | |
1577 | * | |
1578 | * 2. the path is a directory prefix of some element in the | |
1579 | * pathspec | |
1580 | */ | |
1581 | static int exclude_matches_pathspec(const char *path, int pathlen, | |
1582 | const struct pathspec *pathspec) | |
1583 | { | |
1584 | int i; | |
1585 | ||
1586 | if (!pathspec || !pathspec->nr) | |
1587 | return 0; | |
1588 | ||
1589 | GUARD_PATHSPEC(pathspec, | |
1590 | PATHSPEC_FROMTOP | | |
1591 | PATHSPEC_MAXDEPTH | | |
1592 | PATHSPEC_LITERAL | | |
1593 | PATHSPEC_GLOB | | |
1594 | PATHSPEC_ICASE | | |
1595 | PATHSPEC_EXCLUDE); | |
1596 | ||
1597 | for (i = 0; i < pathspec->nr; i++) { | |
1598 | const struct pathspec_item *item = &pathspec->items[i]; | |
1599 | int len = item->nowildcard_len; | |
1600 | ||
1601 | if (len == pathlen && | |
1602 | !ps_strncmp(item, item->match, path, pathlen)) | |
1603 | return 1; | |
1604 | if (len > pathlen && | |
1605 | item->match[pathlen] == '/' && | |
1606 | !ps_strncmp(item, item->match, path, pathlen)) | |
1607 | return 1; | |
1608 | } | |
1609 | return 0; | |
1610 | } | |
1611 | ||
1612 | static int get_index_dtype(struct index_state *istate, | |
1613 | const char *path, int len) | |
1614 | { | |
1615 | int pos; | |
1616 | const struct cache_entry *ce; | |
1617 | ||
1618 | ce = index_file_exists(istate, path, len, 0); | |
1619 | if (ce) { | |
1620 | if (!ce_uptodate(ce)) | |
1621 | return DT_UNKNOWN; | |
1622 | if (S_ISGITLINK(ce->ce_mode)) | |
1623 | return DT_DIR; | |
1624 | /* | |
1625 | * Nobody actually cares about the | |
1626 | * difference between DT_LNK and DT_REG | |
1627 | */ | |
1628 | return DT_REG; | |
1629 | } | |
1630 | ||
1631 | /* Try to look it up as a directory */ | |
1632 | pos = index_name_pos(istate, path, len); | |
1633 | if (pos >= 0) | |
1634 | return DT_UNKNOWN; | |
1635 | pos = -pos-1; | |
1636 | while (pos < istate->cache_nr) { | |
1637 | ce = istate->cache[pos++]; | |
1638 | if (strncmp(ce->name, path, len)) | |
1639 | break; | |
1640 | if (ce->name[len] > '/') | |
1641 | break; | |
1642 | if (ce->name[len] < '/') | |
1643 | continue; | |
1644 | if (!ce_uptodate(ce)) | |
1645 | break; /* continue? */ | |
1646 | return DT_DIR; | |
1647 | } | |
1648 | return DT_UNKNOWN; | |
1649 | } | |
1650 | ||
1651 | static int get_dtype(struct dirent *de, struct index_state *istate, | |
1652 | const char *path, int len) | |
1653 | { | |
1654 | int dtype = de ? DTYPE(de) : DT_UNKNOWN; | |
1655 | struct stat st; | |
1656 | ||
1657 | if (dtype != DT_UNKNOWN) | |
1658 | return dtype; | |
1659 | dtype = get_index_dtype(istate, path, len); | |
1660 | if (dtype != DT_UNKNOWN) | |
1661 | return dtype; | |
1662 | if (lstat(path, &st)) | |
1663 | return dtype; | |
1664 | if (S_ISREG(st.st_mode)) | |
1665 | return DT_REG; | |
1666 | if (S_ISDIR(st.st_mode)) | |
1667 | return DT_DIR; | |
1668 | if (S_ISLNK(st.st_mode)) | |
1669 | return DT_LNK; | |
1670 | return dtype; | |
1671 | } | |
1672 | ||
1673 | static enum path_treatment treat_one_path(struct dir_struct *dir, | |
1674 | struct untracked_cache_dir *untracked, | |
1675 | struct index_state *istate, | |
1676 | struct strbuf *path, | |
1677 | int baselen, | |
1678 | const struct pathspec *pathspec, | |
1679 | int dtype, struct dirent *de) | |
1680 | { | |
1681 | int exclude; | |
1682 | int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case); | |
1683 | enum path_treatment path_treatment; | |
1684 | ||
1685 | if (dtype == DT_UNKNOWN) | |
1686 | dtype = get_dtype(de, istate, path->buf, path->len); | |
1687 | ||
1688 | /* Always exclude indexed files */ | |
1689 | if (dtype != DT_DIR && has_path_in_index) | |
1690 | return path_none; | |
1691 | ||
1692 | /* | |
1693 | * When we are looking at a directory P in the working tree, | |
1694 | * there are three cases: | |
1695 | * | |
1696 | * (1) P exists in the index. Everything inside the directory P in | |
1697 | * the working tree needs to go when P is checked out from the | |
1698 | * index. | |
1699 | * | |
1700 | * (2) P does not exist in the index, but there is P/Q in the index. | |
1701 | * We know P will stay a directory when we check out the contents | |
1702 | * of the index, but we do not know yet if there is a directory | |
1703 | * P/Q in the working tree to be killed, so we need to recurse. | |
1704 | * | |
1705 | * (3) P does not exist in the index, and there is no P/Q in the index | |
1706 | * to require P to be a directory, either. Only in this case, we | |
1707 | * know that everything inside P will not be killed without | |
1708 | * recursing. | |
1709 | */ | |
1710 | if ((dir->flags & DIR_COLLECT_KILLED_ONLY) && | |
1711 | (dtype == DT_DIR) && | |
1712 | !has_path_in_index && | |
1713 | (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent)) | |
1714 | return path_none; | |
1715 | ||
1716 | exclude = is_excluded(dir, istate, path->buf, &dtype); | |
1717 | ||
1718 | /* | |
1719 | * Excluded? If we don't explicitly want to show | |
1720 | * ignored files, ignore it | |
1721 | */ | |
1722 | if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO))) | |
1723 | return path_excluded; | |
1724 | ||
1725 | switch (dtype) { | |
1726 | default: | |
1727 | return path_none; | |
1728 | case DT_DIR: | |
1729 | strbuf_addch(path, '/'); | |
1730 | path_treatment = treat_directory(dir, istate, untracked, | |
1731 | path->buf, path->len, | |
1732 | baselen, exclude, pathspec); | |
1733 | /* | |
1734 | * If 1) we only want to return directories that | |
1735 | * match an exclude pattern and 2) this directory does | |
1736 | * not match an exclude pattern but all of its | |
1737 | * contents are excluded, then indicate that we should | |
1738 | * recurse into this directory (instead of marking the | |
1739 | * directory itself as an ignored path). | |
1740 | */ | |
1741 | if (!exclude && | |
1742 | path_treatment == path_excluded && | |
1743 | (dir->flags & DIR_SHOW_IGNORED_TOO) && | |
1744 | (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) | |
1745 | return path_recurse; | |
1746 | return path_treatment; | |
1747 | case DT_REG: | |
1748 | case DT_LNK: | |
1749 | return exclude ? path_excluded : path_untracked; | |
1750 | } | |
1751 | } | |
1752 | ||
1753 | static enum path_treatment treat_path_fast(struct dir_struct *dir, | |
1754 | struct untracked_cache_dir *untracked, | |
1755 | struct cached_dir *cdir, | |
1756 | struct index_state *istate, | |
1757 | struct strbuf *path, | |
1758 | int baselen, | |
1759 | const struct pathspec *pathspec) | |
1760 | { | |
1761 | strbuf_setlen(path, baselen); | |
1762 | if (!cdir->ucd) { | |
1763 | strbuf_addstr(path, cdir->file); | |
1764 | return path_untracked; | |
1765 | } | |
1766 | strbuf_addstr(path, cdir->ucd->name); | |
1767 | /* treat_one_path() does this before it calls treat_directory() */ | |
1768 | strbuf_complete(path, '/'); | |
1769 | if (cdir->ucd->check_only) | |
1770 | /* | |
1771 | * check_only is set as a result of treat_directory() getting | |
1772 | * to its bottom. Verify again the same set of directories | |
1773 | * with check_only set. | |
1774 | */ | |
1775 | return read_directory_recursive(dir, istate, path->buf, path->len, | |
1776 | cdir->ucd, 1, 0, pathspec); | |
1777 | /* | |
1778 | * We get path_recurse in the first run when | |
1779 | * directory_exists_in_index() returns index_nonexistent. We | |
1780 | * are sure that new changes in the index does not impact the | |
1781 | * outcome. Return now. | |
1782 | */ | |
1783 | return path_recurse; | |
1784 | } | |
1785 | ||
1786 | static enum path_treatment treat_path(struct dir_struct *dir, | |
1787 | struct untracked_cache_dir *untracked, | |
1788 | struct cached_dir *cdir, | |
1789 | struct index_state *istate, | |
1790 | struct strbuf *path, | |
1791 | int baselen, | |
1792 | const struct pathspec *pathspec) | |
1793 | { | |
1794 | int dtype; | |
1795 | struct dirent *de = cdir->de; | |
1796 | ||
1797 | if (!de) | |
1798 | return treat_path_fast(dir, untracked, cdir, istate, path, | |
1799 | baselen, pathspec); | |
1800 | if (is_dot_or_dotdot(de->d_name) || !fspathcmp(de->d_name, ".git")) | |
1801 | return path_none; | |
1802 | strbuf_setlen(path, baselen); | |
1803 | strbuf_addstr(path, de->d_name); | |
1804 | if (simplify_away(path->buf, path->len, pathspec)) | |
1805 | return path_none; | |
1806 | ||
1807 | dtype = DTYPE(de); | |
1808 | return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de); | |
1809 | } | |
1810 | ||
1811 | static void add_untracked(struct untracked_cache_dir *dir, const char *name) | |
1812 | { | |
1813 | if (!dir) | |
1814 | return; | |
1815 | ALLOC_GROW(dir->untracked, dir->untracked_nr + 1, | |
1816 | dir->untracked_alloc); | |
1817 | dir->untracked[dir->untracked_nr++] = xstrdup(name); | |
1818 | } | |
1819 | ||
1820 | static int valid_cached_dir(struct dir_struct *dir, | |
1821 | struct untracked_cache_dir *untracked, | |
1822 | struct index_state *istate, | |
1823 | struct strbuf *path, | |
1824 | int check_only) | |
1825 | { | |
1826 | struct stat st; | |
1827 | ||
1828 | if (!untracked) | |
1829 | return 0; | |
1830 | ||
1831 | /* | |
1832 | * With fsmonitor, we can trust the untracked cache's valid field. | |
1833 | */ | |
1834 | refresh_fsmonitor(istate); | |
1835 | if (!(dir->untracked->use_fsmonitor && untracked->valid)) { | |
1836 | if (lstat(path->len ? path->buf : ".", &st)) { | |
1837 | memset(&untracked->stat_data, 0, sizeof(untracked->stat_data)); | |
1838 | return 0; | |
1839 | } | |
1840 | if (!untracked->valid || | |
1841 | match_stat_data_racy(istate, &untracked->stat_data, &st)) { | |
1842 | fill_stat_data(&untracked->stat_data, &st); | |
1843 | return 0; | |
1844 | } | |
1845 | } | |
1846 | ||
1847 | if (untracked->check_only != !!check_only) | |
1848 | return 0; | |
1849 | ||
1850 | /* | |
1851 | * prep_exclude will be called eventually on this directory, | |
1852 | * but it's called much later in last_matching_pattern(). We | |
1853 | * need it now to determine the validity of the cache for this | |
1854 | * path. The next calls will be nearly no-op, the way | |
1855 | * prep_exclude() is designed. | |
1856 | */ | |
1857 | if (path->len && path->buf[path->len - 1] != '/') { | |
1858 | strbuf_addch(path, '/'); | |
1859 | prep_exclude(dir, istate, path->buf, path->len); | |
1860 | strbuf_setlen(path, path->len - 1); | |
1861 | } else | |
1862 | prep_exclude(dir, istate, path->buf, path->len); | |
1863 | ||
1864 | /* hopefully prep_exclude() haven't invalidated this entry... */ | |
1865 | return untracked->valid; | |
1866 | } | |
1867 | ||
1868 | static int open_cached_dir(struct cached_dir *cdir, | |
1869 | struct dir_struct *dir, | |
1870 | struct untracked_cache_dir *untracked, | |
1871 | struct index_state *istate, | |
1872 | struct strbuf *path, | |
1873 | int check_only) | |
1874 | { | |
1875 | const char *c_path; | |
1876 | ||
1877 | memset(cdir, 0, sizeof(*cdir)); | |
1878 | cdir->untracked = untracked; | |
1879 | if (valid_cached_dir(dir, untracked, istate, path, check_only)) | |
1880 | return 0; | |
1881 | c_path = path->len ? path->buf : "."; | |
1882 | cdir->fdir = opendir(c_path); | |
1883 | if (!cdir->fdir) | |
1884 | warning_errno(_("could not open directory '%s'"), c_path); | |
1885 | if (dir->untracked) { | |
1886 | invalidate_directory(dir->untracked, untracked); | |
1887 | dir->untracked->dir_opened++; | |
1888 | } | |
1889 | if (!cdir->fdir) | |
1890 | return -1; | |
1891 | return 0; | |
1892 | } | |
1893 | ||
1894 | static int read_cached_dir(struct cached_dir *cdir) | |
1895 | { | |
1896 | if (cdir->fdir) { | |
1897 | cdir->de = readdir(cdir->fdir); | |
1898 | if (!cdir->de) | |
1899 | return -1; | |
1900 | return 0; | |
1901 | } | |
1902 | while (cdir->nr_dirs < cdir->untracked->dirs_nr) { | |
1903 | struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs]; | |
1904 | if (!d->recurse) { | |
1905 | cdir->nr_dirs++; | |
1906 | continue; | |
1907 | } | |
1908 | cdir->ucd = d; | |
1909 | cdir->nr_dirs++; | |
1910 | return 0; | |
1911 | } | |
1912 | cdir->ucd = NULL; | |
1913 | if (cdir->nr_files < cdir->untracked->untracked_nr) { | |
1914 | struct untracked_cache_dir *d = cdir->untracked; | |
1915 | cdir->file = d->untracked[cdir->nr_files++]; | |
1916 | return 0; | |
1917 | } | |
1918 | return -1; | |
1919 | } | |
1920 | ||
1921 | static void close_cached_dir(struct cached_dir *cdir) | |
1922 | { | |
1923 | if (cdir->fdir) | |
1924 | closedir(cdir->fdir); | |
1925 | /* | |
1926 | * We have gone through this directory and found no untracked | |
1927 | * entries. Mark it valid. | |
1928 | */ | |
1929 | if (cdir->untracked) { | |
1930 | cdir->untracked->valid = 1; | |
1931 | cdir->untracked->recurse = 1; | |
1932 | } | |
1933 | } | |
1934 | ||
1935 | static void add_path_to_appropriate_result_list(struct dir_struct *dir, | |
1936 | struct untracked_cache_dir *untracked, | |
1937 | struct cached_dir *cdir, | |
1938 | struct index_state *istate, | |
1939 | struct strbuf *path, | |
1940 | int baselen, | |
1941 | const struct pathspec *pathspec, | |
1942 | enum path_treatment state) | |
1943 | { | |
1944 | /* add the path to the appropriate result list */ | |
1945 | switch (state) { | |
1946 | case path_excluded: | |
1947 | if (dir->flags & DIR_SHOW_IGNORED) | |
1948 | dir_add_name(dir, istate, path->buf, path->len); | |
1949 | else if ((dir->flags & DIR_SHOW_IGNORED_TOO) || | |
1950 | ((dir->flags & DIR_COLLECT_IGNORED) && | |
1951 | exclude_matches_pathspec(path->buf, path->len, | |
1952 | pathspec))) | |
1953 | dir_add_ignored(dir, istate, path->buf, path->len); | |
1954 | break; | |
1955 | ||
1956 | case path_untracked: | |
1957 | if (dir->flags & DIR_SHOW_IGNORED) | |
1958 | break; | |
1959 | dir_add_name(dir, istate, path->buf, path->len); | |
1960 | if (cdir->fdir) | |
1961 | add_untracked(untracked, path->buf + baselen); | |
1962 | break; | |
1963 | ||
1964 | default: | |
1965 | break; | |
1966 | } | |
1967 | } | |
1968 | ||
1969 | /* | |
1970 | * Read a directory tree. We currently ignore anything but | |
1971 | * directories, regular files and symlinks. That's because git | |
1972 | * doesn't handle them at all yet. Maybe that will change some | |
1973 | * day. | |
1974 | * | |
1975 | * Also, we ignore the name ".git" (even if it is not a directory). | |
1976 | * That likely will not change. | |
1977 | * | |
1978 | * If 'stop_at_first_file' is specified, 'path_excluded' is returned | |
1979 | * to signal that a file was found. This is the least significant value that | |
1980 | * indicates that a file was encountered that does not depend on the order of | |
1981 | * whether an untracked or exluded path was encountered first. | |
1982 | * | |
1983 | * Returns the most significant path_treatment value encountered in the scan. | |
1984 | * If 'stop_at_first_file' is specified, `path_excluded` is the most | |
1985 | * significant path_treatment value that will be returned. | |
1986 | */ | |
1987 | ||
1988 | static enum path_treatment read_directory_recursive(struct dir_struct *dir, | |
1989 | struct index_state *istate, const char *base, int baselen, | |
1990 | struct untracked_cache_dir *untracked, int check_only, | |
1991 | int stop_at_first_file, const struct pathspec *pathspec) | |
1992 | { | |
1993 | /* | |
1994 | * WARNING WARNING WARNING: | |
1995 | * | |
1996 | * Any updates to the traversal logic here may need corresponding | |
1997 | * updates in treat_leading_path(). See the commit message for the | |
1998 | * commit adding this warning as well as the commit preceding it | |
1999 | * for details. | |
2000 | */ | |
2001 | ||
2002 | struct cached_dir cdir; | |
2003 | enum path_treatment state, subdir_state, dir_state = path_none; | |
2004 | struct strbuf path = STRBUF_INIT; | |
2005 | ||
2006 | strbuf_add(&path, base, baselen); | |
2007 | ||
2008 | if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only)) | |
2009 | goto out; | |
2010 | ||
2011 | if (untracked) | |
2012 | untracked->check_only = !!check_only; | |
2013 | ||
2014 | while (!read_cached_dir(&cdir)) { | |
2015 | /* check how the file or directory should be treated */ | |
2016 | state = treat_path(dir, untracked, &cdir, istate, &path, | |
2017 | baselen, pathspec); | |
2018 | ||
2019 | if (state > dir_state) | |
2020 | dir_state = state; | |
2021 | ||
2022 | /* recurse into subdir if instructed by treat_path */ | |
2023 | if ((state == path_recurse) || | |
2024 | ((state == path_untracked) && | |
2025 | (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR) && | |
2026 | ((dir->flags & DIR_SHOW_IGNORED_TOO) || | |
2027 | (pathspec && | |
2028 | do_match_pathspec(istate, pathspec, path.buf, path.len, | |
2029 | baselen, NULL, DO_MATCH_LEADING_PATHSPEC) == MATCHED_RECURSIVELY_LEADING_PATHSPEC)))) { | |
2030 | struct untracked_cache_dir *ud; | |
2031 | ud = lookup_untracked(dir->untracked, untracked, | |
2032 | path.buf + baselen, | |
2033 | path.len - baselen); | |
2034 | subdir_state = | |
2035 | read_directory_recursive(dir, istate, path.buf, | |
2036 | path.len, ud, | |
2037 | check_only, stop_at_first_file, pathspec); | |
2038 | if (subdir_state > dir_state) | |
2039 | dir_state = subdir_state; | |
2040 | ||
2041 | if (pathspec && | |
2042 | !match_pathspec(istate, pathspec, path.buf, path.len, | |
2043 | 0 /* prefix */, NULL, | |
2044 | 0 /* do NOT special case dirs */)) | |
2045 | state = path_none; | |
2046 | } | |
2047 | ||
2048 | if (check_only) { | |
2049 | if (stop_at_first_file) { | |
2050 | /* | |
2051 | * If stopping at first file, then | |
2052 | * signal that a file was found by | |
2053 | * returning `path_excluded`. This is | |
2054 | * to return a consistent value | |
2055 | * regardless of whether an ignored or | |
2056 | * excluded file happened to be | |
2057 | * encountered 1st. | |
2058 | * | |
2059 | * In current usage, the | |
2060 | * `stop_at_first_file` is passed when | |
2061 | * an ancestor directory has matched | |
2062 | * an exclude pattern, so any found | |
2063 | * files will be excluded. | |
2064 | */ | |
2065 | if (dir_state >= path_excluded) { | |
2066 | dir_state = path_excluded; | |
2067 | break; | |
2068 | } | |
2069 | } | |
2070 | ||
2071 | /* abort early if maximum state has been reached */ | |
2072 | if (dir_state == path_untracked) { | |
2073 | if (cdir.fdir) | |
2074 | add_untracked(untracked, path.buf + baselen); | |
2075 | break; | |
2076 | } | |
2077 | /* skip the dir_add_* part */ | |
2078 | continue; | |
2079 | } | |
2080 | ||
2081 | add_path_to_appropriate_result_list(dir, untracked, &cdir, | |
2082 | istate, &path, baselen, | |
2083 | pathspec, state); | |
2084 | } | |
2085 | close_cached_dir(&cdir); | |
2086 | out: | |
2087 | strbuf_release(&path); | |
2088 | ||
2089 | return dir_state; | |
2090 | } | |
2091 | ||
2092 | int cmp_dir_entry(const void *p1, const void *p2) | |
2093 | { | |
2094 | const struct dir_entry *e1 = *(const struct dir_entry **)p1; | |
2095 | const struct dir_entry *e2 = *(const struct dir_entry **)p2; | |
2096 | ||
2097 | return name_compare(e1->name, e1->len, e2->name, e2->len); | |
2098 | } | |
2099 | ||
2100 | /* check if *out lexically strictly contains *in */ | |
2101 | int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in) | |
2102 | { | |
2103 | return (out->len < in->len) && | |
2104 | (out->name[out->len - 1] == '/') && | |
2105 | !memcmp(out->name, in->name, out->len); | |
2106 | } | |
2107 | ||
2108 | static int treat_leading_path(struct dir_struct *dir, | |
2109 | struct index_state *istate, | |
2110 | const char *path, int len, | |
2111 | const struct pathspec *pathspec) | |
2112 | { | |
2113 | /* | |
2114 | * WARNING WARNING WARNING: | |
2115 | * | |
2116 | * Any updates to the traversal logic here may need corresponding | |
2117 | * updates in treat_leading_path(). See the commit message for the | |
2118 | * commit adding this warning as well as the commit preceding it | |
2119 | * for details. | |
2120 | */ | |
2121 | ||
2122 | struct strbuf sb = STRBUF_INIT; | |
2123 | int prevlen, baselen; | |
2124 | const char *cp; | |
2125 | struct cached_dir cdir; | |
2126 | struct dirent *de; | |
2127 | enum path_treatment state = path_none; | |
2128 | ||
2129 | /* | |
2130 | * For each directory component of path, we are going to check whether | |
2131 | * that path is relevant given the pathspec. For example, if path is | |
2132 | * foo/bar/baz/ | |
2133 | * then we will ask treat_path() whether we should go into foo, then | |
2134 | * whether we should go into bar, then whether baz is relevant. | |
2135 | * Checking each is important because e.g. if path is | |
2136 | * .git/info/ | |
2137 | * then we need to check .git to know we shouldn't traverse it. | |
2138 | * If the return from treat_path() is: | |
2139 | * * path_none, for any path, we return false. | |
2140 | * * path_recurse, for all path components, we return true | |
2141 | * * <anything else> for some intermediate component, we make sure | |
2142 | * to add that path to the relevant list but return false | |
2143 | * signifying that we shouldn't recurse into it. | |
2144 | */ | |
2145 | ||
2146 | while (len && path[len - 1] == '/') | |
2147 | len--; | |
2148 | if (!len) | |
2149 | return 1; | |
2150 | ||
2151 | /* | |
2152 | * We need a manufactured dirent with sufficient space to store a | |
2153 | * leading directory component of path in its d_name. Here, we | |
2154 | * assume that the dirent's d_name is either declared as | |
2155 | * char d_name[BIG_ENOUGH] | |
2156 | * or that it is declared at the end of the struct as | |
2157 | * char d_name[] | |
2158 | * For either case, padding with len+1 bytes at the end will ensure | |
2159 | * sufficient storage space. | |
2160 | */ | |
2161 | de = xcalloc(1, sizeof(struct dirent)+len+1); | |
2162 | memset(&cdir, 0, sizeof(cdir)); | |
2163 | cdir.de = de; | |
2164 | #if defined(DT_UNKNOWN) && !defined(NO_D_TYPE_IN_DIRENT) | |
2165 | de->d_type = DT_DIR; | |
2166 | #endif | |
2167 | baselen = 0; | |
2168 | prevlen = 0; | |
2169 | while (1) { | |
2170 | prevlen = baselen + !!baselen; | |
2171 | cp = path + prevlen; | |
2172 | cp = memchr(cp, '/', path + len - cp); | |
2173 | if (!cp) | |
2174 | baselen = len; | |
2175 | else | |
2176 | baselen = cp - path; | |
2177 | strbuf_reset(&sb); | |
2178 | strbuf_add(&sb, path, baselen); | |
2179 | if (!is_directory(sb.buf)) | |
2180 | break; | |
2181 | strbuf_reset(&sb); | |
2182 | strbuf_add(&sb, path, prevlen); | |
2183 | memcpy(de->d_name, path+prevlen, baselen-prevlen); | |
2184 | de->d_name[baselen-prevlen] = '\0'; | |
2185 | state = treat_path(dir, NULL, &cdir, istate, &sb, prevlen, | |
2186 | pathspec); | |
2187 | if (state == path_untracked && | |
2188 | get_dtype(cdir.de, istate, sb.buf, sb.len) == DT_DIR && | |
2189 | (dir->flags & DIR_SHOW_IGNORED_TOO || | |
2190 | do_match_pathspec(istate, pathspec, sb.buf, sb.len, | |
2191 | baselen, NULL, DO_MATCH_LEADING_PATHSPEC) == MATCHED_RECURSIVELY_LEADING_PATHSPEC)) { | |
2192 | add_path_to_appropriate_result_list(dir, NULL, &cdir, | |
2193 | istate, | |
2194 | &sb, baselen, | |
2195 | pathspec, state); | |
2196 | state = path_recurse; | |
2197 | } | |
2198 | ||
2199 | if (state != path_recurse) | |
2200 | break; /* do not recurse into it */ | |
2201 | if (len <= baselen) | |
2202 | break; /* finished checking */ | |
2203 | } | |
2204 | add_path_to_appropriate_result_list(dir, NULL, &cdir, istate, | |
2205 | &sb, baselen, pathspec, | |
2206 | state); | |
2207 | ||
2208 | free(de); | |
2209 | strbuf_release(&sb); | |
2210 | return state == path_recurse; | |
2211 | } | |
2212 | ||
2213 | static const char *get_ident_string(void) | |
2214 | { | |
2215 | static struct strbuf sb = STRBUF_INIT; | |
2216 | struct utsname uts; | |
2217 | ||
2218 | if (sb.len) | |
2219 | return sb.buf; | |
2220 | if (uname(&uts) < 0) | |
2221 | die_errno(_("failed to get kernel name and information")); | |
2222 | strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(), | |
2223 | uts.sysname); | |
2224 | return sb.buf; | |
2225 | } | |
2226 | ||
2227 | static int ident_in_untracked(const struct untracked_cache *uc) | |
2228 | { | |
2229 | /* | |
2230 | * Previous git versions may have saved many NUL separated | |
2231 | * strings in the "ident" field, but it is insane to manage | |
2232 | * many locations, so just take care of the first one. | |
2233 | */ | |
2234 | ||
2235 | return !strcmp(uc->ident.buf, get_ident_string()); | |
2236 | } | |
2237 | ||
2238 | static void set_untracked_ident(struct untracked_cache *uc) | |
2239 | { | |
2240 | strbuf_reset(&uc->ident); | |
2241 | strbuf_addstr(&uc->ident, get_ident_string()); | |
2242 | ||
2243 | /* | |
2244 | * This strbuf used to contain a list of NUL separated | |
2245 | * strings, so save NUL too for backward compatibility. | |
2246 | */ | |
2247 | strbuf_addch(&uc->ident, 0); | |
2248 | } | |
2249 | ||
2250 | static void new_untracked_cache(struct index_state *istate) | |
2251 | { | |
2252 | struct untracked_cache *uc = xcalloc(1, sizeof(*uc)); | |
2253 | strbuf_init(&uc->ident, 100); | |
2254 | uc->exclude_per_dir = ".gitignore"; | |
2255 | /* should be the same flags used by git-status */ | |
2256 | uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES; | |
2257 | set_untracked_ident(uc); | |
2258 | istate->untracked = uc; | |
2259 | istate->cache_changed |= UNTRACKED_CHANGED; | |
2260 | } | |
2261 | ||
2262 | void add_untracked_cache(struct index_state *istate) | |
2263 | { | |
2264 | if (!istate->untracked) { | |
2265 | new_untracked_cache(istate); | |
2266 | } else { | |
2267 | if (!ident_in_untracked(istate->untracked)) { | |
2268 | free_untracked_cache(istate->untracked); | |
2269 | new_untracked_cache(istate); | |
2270 | } | |
2271 | } | |
2272 | } | |
2273 | ||
2274 | void remove_untracked_cache(struct index_state *istate) | |
2275 | { | |
2276 | if (istate->untracked) { | |
2277 | free_untracked_cache(istate->untracked); | |
2278 | istate->untracked = NULL; | |
2279 | istate->cache_changed |= UNTRACKED_CHANGED; | |
2280 | } | |
2281 | } | |
2282 | ||
2283 | static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir, | |
2284 | int base_len, | |
2285 | const struct pathspec *pathspec) | |
2286 | { | |
2287 | struct untracked_cache_dir *root; | |
2288 | static int untracked_cache_disabled = -1; | |
2289 | ||
2290 | if (!dir->untracked) | |
2291 | return NULL; | |
2292 | if (untracked_cache_disabled < 0) | |
2293 | untracked_cache_disabled = git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0); | |
2294 | if (untracked_cache_disabled) | |
2295 | return NULL; | |
2296 | ||
2297 | /* | |
2298 | * We only support $GIT_DIR/info/exclude and core.excludesfile | |
2299 | * as the global ignore rule files. Any other additions | |
2300 | * (e.g. from command line) invalidate the cache. This | |
2301 | * condition also catches running setup_standard_excludes() | |
2302 | * before setting dir->untracked! | |
2303 | */ | |
2304 | if (dir->unmanaged_exclude_files) | |
2305 | return NULL; | |
2306 | ||
2307 | /* | |
2308 | * Optimize for the main use case only: whole-tree git | |
2309 | * status. More work involved in treat_leading_path() if we | |
2310 | * use cache on just a subset of the worktree. pathspec | |
2311 | * support could make the matter even worse. | |
2312 | */ | |
2313 | if (base_len || (pathspec && pathspec->nr)) | |
2314 | return NULL; | |
2315 | ||
2316 | /* Different set of flags may produce different results */ | |
2317 | if (dir->flags != dir->untracked->dir_flags || | |
2318 | /* | |
2319 | * See treat_directory(), case index_nonexistent. Without | |
2320 | * this flag, we may need to also cache .git file content | |
2321 | * for the resolve_gitlink_ref() call, which we don't. | |
2322 | */ | |
2323 | !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) || | |
2324 | /* We don't support collecting ignore files */ | |
2325 | (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO | | |
2326 | DIR_COLLECT_IGNORED))) | |
2327 | return NULL; | |
2328 | ||
2329 | /* | |
2330 | * If we use .gitignore in the cache and now you change it to | |
2331 | * .gitexclude, everything will go wrong. | |
2332 | */ | |
2333 | if (dir->exclude_per_dir != dir->untracked->exclude_per_dir && | |
2334 | strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir)) | |
2335 | return NULL; | |
2336 | ||
2337 | /* | |
2338 | * EXC_CMDL is not considered in the cache. If people set it, | |
2339 | * skip the cache. | |
2340 | */ | |
2341 | if (dir->exclude_list_group[EXC_CMDL].nr) | |
2342 | return NULL; | |
2343 | ||
2344 | if (!ident_in_untracked(dir->untracked)) { | |
2345 | warning(_("untracked cache is disabled on this system or location")); | |
2346 | return NULL; | |
2347 | } | |
2348 | ||
2349 | if (!dir->untracked->root) { | |
2350 | const int len = sizeof(*dir->untracked->root); | |
2351 | dir->untracked->root = xmalloc(len); | |
2352 | memset(dir->untracked->root, 0, len); | |
2353 | } | |
2354 | ||
2355 | /* Validate $GIT_DIR/info/exclude and core.excludesfile */ | |
2356 | root = dir->untracked->root; | |
2357 | if (!oideq(&dir->ss_info_exclude.oid, | |
2358 | &dir->untracked->ss_info_exclude.oid)) { | |
2359 | invalidate_gitignore(dir->untracked, root); | |
2360 | dir->untracked->ss_info_exclude = dir->ss_info_exclude; | |
2361 | } | |
2362 | if (!oideq(&dir->ss_excludes_file.oid, | |
2363 | &dir->untracked->ss_excludes_file.oid)) { | |
2364 | invalidate_gitignore(dir->untracked, root); | |
2365 | dir->untracked->ss_excludes_file = dir->ss_excludes_file; | |
2366 | } | |
2367 | ||
2368 | /* Make sure this directory is not dropped out at saving phase */ | |
2369 | root->recurse = 1; | |
2370 | return root; | |
2371 | } | |
2372 | ||
2373 | int read_directory(struct dir_struct *dir, struct index_state *istate, | |
2374 | const char *path, int len, const struct pathspec *pathspec) | |
2375 | { | |
2376 | struct untracked_cache_dir *untracked; | |
2377 | ||
2378 | trace_performance_enter(); | |
2379 | ||
2380 | if (has_symlink_leading_path(path, len)) { | |
2381 | trace_performance_leave("read directory %.*s", len, path); | |
2382 | return dir->nr; | |
2383 | } | |
2384 | ||
2385 | untracked = validate_untracked_cache(dir, len, pathspec); | |
2386 | if (!untracked) | |
2387 | /* | |
2388 | * make sure untracked cache code path is disabled, | |
2389 | * e.g. prep_exclude() | |
2390 | */ | |
2391 | dir->untracked = NULL; | |
2392 | if (!len || treat_leading_path(dir, istate, path, len, pathspec)) | |
2393 | read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec); | |
2394 | QSORT(dir->entries, dir->nr, cmp_dir_entry); | |
2395 | QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry); | |
2396 | ||
2397 | /* | |
2398 | * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will | |
2399 | * also pick up untracked contents of untracked dirs; by default | |
2400 | * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not. | |
2401 | */ | |
2402 | if ((dir->flags & DIR_SHOW_IGNORED_TOO) && | |
2403 | !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) { | |
2404 | int i, j; | |
2405 | ||
2406 | /* remove from dir->entries untracked contents of untracked dirs */ | |
2407 | for (i = j = 0; j < dir->nr; j++) { | |
2408 | if (i && | |
2409 | check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) { | |
2410 | FREE_AND_NULL(dir->entries[j]); | |
2411 | } else { | |
2412 | dir->entries[i++] = dir->entries[j]; | |
2413 | } | |
2414 | } | |
2415 | ||
2416 | dir->nr = i; | |
2417 | } | |
2418 | ||
2419 | trace_performance_leave("read directory %.*s", len, path); | |
2420 | if (dir->untracked) { | |
2421 | static int force_untracked_cache = -1; | |
2422 | static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS); | |
2423 | ||
2424 | if (force_untracked_cache < 0) | |
2425 | force_untracked_cache = | |
2426 | git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0); | |
2427 | trace_printf_key(&trace_untracked_stats, | |
2428 | "node creation: %u\n" | |
2429 | "gitignore invalidation: %u\n" | |
2430 | "directory invalidation: %u\n" | |
2431 | "opendir: %u\n", | |
2432 | dir->untracked->dir_created, | |
2433 | dir->untracked->gitignore_invalidated, | |
2434 | dir->untracked->dir_invalidated, | |
2435 | dir->untracked->dir_opened); | |
2436 | if (force_untracked_cache && | |
2437 | dir->untracked == istate->untracked && | |
2438 | (dir->untracked->dir_opened || | |
2439 | dir->untracked->gitignore_invalidated || | |
2440 | dir->untracked->dir_invalidated)) | |
2441 | istate->cache_changed |= UNTRACKED_CHANGED; | |
2442 | if (dir->untracked != istate->untracked) { | |
2443 | FREE_AND_NULL(dir->untracked); | |
2444 | } | |
2445 | } | |
2446 | return dir->nr; | |
2447 | } | |
2448 | ||
2449 | int file_exists(const char *f) | |
2450 | { | |
2451 | struct stat sb; | |
2452 | return lstat(f, &sb) == 0; | |
2453 | } | |
2454 | ||
2455 | int repo_file_exists(struct repository *repo, const char *path) | |
2456 | { | |
2457 | if (repo != the_repository) | |
2458 | BUG("do not know how to check file existence in arbitrary repo"); | |
2459 | ||
2460 | return file_exists(path); | |
2461 | } | |
2462 | ||
2463 | static int cmp_icase(char a, char b) | |
2464 | { | |
2465 | if (a == b) | |
2466 | return 0; | |
2467 | if (ignore_case) | |
2468 | return toupper(a) - toupper(b); | |
2469 | return a - b; | |
2470 | } | |
2471 | ||
2472 | /* | |
2473 | * Given two normalized paths (a trailing slash is ok), if subdir is | |
2474 | * outside dir, return -1. Otherwise return the offset in subdir that | |
2475 | * can be used as relative path to dir. | |
2476 | */ | |
2477 | int dir_inside_of(const char *subdir, const char *dir) | |
2478 | { | |
2479 | int offset = 0; | |
2480 | ||
2481 | assert(dir && subdir && *dir && *subdir); | |
2482 | ||
2483 | while (*dir && *subdir && !cmp_icase(*dir, *subdir)) { | |
2484 | dir++; | |
2485 | subdir++; | |
2486 | offset++; | |
2487 | } | |
2488 | ||
2489 | /* hel[p]/me vs hel[l]/yeah */ | |
2490 | if (*dir && *subdir) | |
2491 | return -1; | |
2492 | ||
2493 | if (!*subdir) | |
2494 | return !*dir ? offset : -1; /* same dir */ | |
2495 | ||
2496 | /* foo/[b]ar vs foo/[] */ | |
2497 | if (is_dir_sep(dir[-1])) | |
2498 | return is_dir_sep(subdir[-1]) ? offset : -1; | |
2499 | ||
2500 | /* foo[/]bar vs foo[] */ | |
2501 | return is_dir_sep(*subdir) ? offset + 1 : -1; | |
2502 | } | |
2503 | ||
2504 | int is_inside_dir(const char *dir) | |
2505 | { | |
2506 | char *cwd; | |
2507 | int rc; | |
2508 | ||
2509 | if (!dir) | |
2510 | return 0; | |
2511 | ||
2512 | cwd = xgetcwd(); | |
2513 | rc = (dir_inside_of(cwd, dir) >= 0); | |
2514 | free(cwd); | |
2515 | return rc; | |
2516 | } | |
2517 | ||
2518 | int is_empty_dir(const char *path) | |
2519 | { | |
2520 | DIR *dir = opendir(path); | |
2521 | struct dirent *e; | |
2522 | int ret = 1; | |
2523 | ||
2524 | if (!dir) | |
2525 | return 0; | |
2526 | ||
2527 | while ((e = readdir(dir)) != NULL) | |
2528 | if (!is_dot_or_dotdot(e->d_name)) { | |
2529 | ret = 0; | |
2530 | break; | |
2531 | } | |
2532 | ||
2533 | closedir(dir); | |
2534 | return ret; | |
2535 | } | |
2536 | ||
2537 | static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up) | |
2538 | { | |
2539 | DIR *dir; | |
2540 | struct dirent *e; | |
2541 | int ret = 0, original_len = path->len, len, kept_down = 0; | |
2542 | int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY); | |
2543 | int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL); | |
2544 | struct object_id submodule_head; | |
2545 | ||
2546 | if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) && | |
2547 | !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) { | |
2548 | /* Do not descend and nuke a nested git work tree. */ | |
2549 | if (kept_up) | |
2550 | *kept_up = 1; | |
2551 | return 0; | |
2552 | } | |
2553 | ||
2554 | flag &= ~REMOVE_DIR_KEEP_TOPLEVEL; | |
2555 | dir = opendir(path->buf); | |
2556 | if (!dir) { | |
2557 | if (errno == ENOENT) | |
2558 | return keep_toplevel ? -1 : 0; | |
2559 | else if (errno == EACCES && !keep_toplevel) | |
2560 | /* | |
2561 | * An empty dir could be removable even if it | |
2562 | * is unreadable: | |
2563 | */ | |
2564 | return rmdir(path->buf); | |
2565 | else | |
2566 | return -1; | |
2567 | } | |
2568 | strbuf_complete(path, '/'); | |
2569 | ||
2570 | len = path->len; | |
2571 | while ((e = readdir(dir)) != NULL) { | |
2572 | struct stat st; | |
2573 | if (is_dot_or_dotdot(e->d_name)) | |
2574 | continue; | |
2575 | ||
2576 | strbuf_setlen(path, len); | |
2577 | strbuf_addstr(path, e->d_name); | |
2578 | if (lstat(path->buf, &st)) { | |
2579 | if (errno == ENOENT) | |
2580 | /* | |
2581 | * file disappeared, which is what we | |
2582 | * wanted anyway | |
2583 | */ | |
2584 | continue; | |
2585 | /* fall thru */ | |
2586 | } else if (S_ISDIR(st.st_mode)) { | |
2587 | if (!remove_dir_recurse(path, flag, &kept_down)) | |
2588 | continue; /* happy */ | |
2589 | } else if (!only_empty && | |
2590 | (!unlink(path->buf) || errno == ENOENT)) { | |
2591 | continue; /* happy, too */ | |
2592 | } | |
2593 | ||
2594 | /* path too long, stat fails, or non-directory still exists */ | |
2595 | ret = -1; | |
2596 | break; | |
2597 | } | |
2598 | closedir(dir); | |
2599 | ||
2600 | strbuf_setlen(path, original_len); | |
2601 | if (!ret && !keep_toplevel && !kept_down) | |
2602 | ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1; | |
2603 | else if (kept_up) | |
2604 | /* | |
2605 | * report the uplevel that it is not an error that we | |
2606 | * did not rmdir() our directory. | |
2607 | */ | |
2608 | *kept_up = !ret; | |
2609 | return ret; | |
2610 | } | |
2611 | ||
2612 | int remove_dir_recursively(struct strbuf *path, int flag) | |
2613 | { | |
2614 | return remove_dir_recurse(path, flag, NULL); | |
2615 | } | |
2616 | ||
2617 | static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude") | |
2618 | ||
2619 | void setup_standard_excludes(struct dir_struct *dir) | |
2620 | { | |
2621 | dir->exclude_per_dir = ".gitignore"; | |
2622 | ||
2623 | /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */ | |
2624 | if (!excludes_file) | |
2625 | excludes_file = xdg_config_home("ignore"); | |
2626 | if (excludes_file && !access_or_warn(excludes_file, R_OK, 0)) | |
2627 | add_patterns_from_file_1(dir, excludes_file, | |
2628 | dir->untracked ? &dir->ss_excludes_file : NULL); | |
2629 | ||
2630 | /* per repository user preference */ | |
2631 | if (startup_info->have_repository) { | |
2632 | const char *path = git_path_info_exclude(); | |
2633 | if (!access_or_warn(path, R_OK, 0)) | |
2634 | add_patterns_from_file_1(dir, path, | |
2635 | dir->untracked ? &dir->ss_info_exclude : NULL); | |
2636 | } | |
2637 | } | |
2638 | ||
2639 | int remove_path(const char *name) | |
2640 | { | |
2641 | char *slash; | |
2642 | ||
2643 | if (unlink(name) && !is_missing_file_error(errno)) | |
2644 | return -1; | |
2645 | ||
2646 | slash = strrchr(name, '/'); | |
2647 | if (slash) { | |
2648 | char *dirs = xstrdup(name); | |
2649 | slash = dirs + (slash - name); | |
2650 | do { | |
2651 | *slash = '\0'; | |
2652 | } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/'))); | |
2653 | free(dirs); | |
2654 | } | |
2655 | return 0; | |
2656 | } | |
2657 | ||
2658 | /* | |
2659 | * Frees memory within dir which was allocated for exclude lists and | |
2660 | * the exclude_stack. Does not free dir itself. | |
2661 | */ | |
2662 | void clear_directory(struct dir_struct *dir) | |
2663 | { | |
2664 | int i, j; | |
2665 | struct exclude_list_group *group; | |
2666 | struct pattern_list *pl; | |
2667 | struct exclude_stack *stk; | |
2668 | ||
2669 | for (i = EXC_CMDL; i <= EXC_FILE; i++) { | |
2670 | group = &dir->exclude_list_group[i]; | |
2671 | for (j = 0; j < group->nr; j++) { | |
2672 | pl = &group->pl[j]; | |
2673 | if (i == EXC_DIRS) | |
2674 | free((char *)pl->src); | |
2675 | clear_pattern_list(pl); | |
2676 | } | |
2677 | free(group->pl); | |
2678 | } | |
2679 | ||
2680 | stk = dir->exclude_stack; | |
2681 | while (stk) { | |
2682 | struct exclude_stack *prev = stk->prev; | |
2683 | free(stk); | |
2684 | stk = prev; | |
2685 | } | |
2686 | strbuf_release(&dir->basebuf); | |
2687 | } | |
2688 | ||
2689 | struct ondisk_untracked_cache { | |
2690 | struct stat_data info_exclude_stat; | |
2691 | struct stat_data excludes_file_stat; | |
2692 | uint32_t dir_flags; | |
2693 | }; | |
2694 | ||
2695 | #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x) | |
2696 | ||
2697 | struct write_data { | |
2698 | int index; /* number of written untracked_cache_dir */ | |
2699 | struct ewah_bitmap *check_only; /* from untracked_cache_dir */ | |
2700 | struct ewah_bitmap *valid; /* from untracked_cache_dir */ | |
2701 | struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */ | |
2702 | struct strbuf out; | |
2703 | struct strbuf sb_stat; | |
2704 | struct strbuf sb_sha1; | |
2705 | }; | |
2706 | ||
2707 | static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from) | |
2708 | { | |
2709 | to->sd_ctime.sec = htonl(from->sd_ctime.sec); | |
2710 | to->sd_ctime.nsec = htonl(from->sd_ctime.nsec); | |
2711 | to->sd_mtime.sec = htonl(from->sd_mtime.sec); | |
2712 | to->sd_mtime.nsec = htonl(from->sd_mtime.nsec); | |
2713 | to->sd_dev = htonl(from->sd_dev); | |
2714 | to->sd_ino = htonl(from->sd_ino); | |
2715 | to->sd_uid = htonl(from->sd_uid); | |
2716 | to->sd_gid = htonl(from->sd_gid); | |
2717 | to->sd_size = htonl(from->sd_size); | |
2718 | } | |
2719 | ||
2720 | static void write_one_dir(struct untracked_cache_dir *untracked, | |
2721 | struct write_data *wd) | |
2722 | { | |
2723 | struct stat_data stat_data; | |
2724 | struct strbuf *out = &wd->out; | |
2725 | unsigned char intbuf[16]; | |
2726 | unsigned int intlen, value; | |
2727 | int i = wd->index++; | |
2728 | ||
2729 | /* | |
2730 | * untracked_nr should be reset whenever valid is clear, but | |
2731 | * for safety.. | |
2732 | */ | |
2733 | if (!untracked->valid) { | |
2734 | untracked->untracked_nr = 0; | |
2735 | untracked->check_only = 0; | |
2736 | } | |
2737 | ||
2738 | if (untracked->check_only) | |
2739 | ewah_set(wd->check_only, i); | |
2740 | if (untracked->valid) { | |
2741 | ewah_set(wd->valid, i); | |
2742 | stat_data_to_disk(&stat_data, &untracked->stat_data); | |
2743 | strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data)); | |
2744 | } | |
2745 | if (!is_null_oid(&untracked->exclude_oid)) { | |
2746 | ewah_set(wd->sha1_valid, i); | |
2747 | strbuf_add(&wd->sb_sha1, untracked->exclude_oid.hash, | |
2748 | the_hash_algo->rawsz); | |
2749 | } | |
2750 | ||
2751 | intlen = encode_varint(untracked->untracked_nr, intbuf); | |
2752 | strbuf_add(out, intbuf, intlen); | |
2753 | ||
2754 | /* skip non-recurse directories */ | |
2755 | for (i = 0, value = 0; i < untracked->dirs_nr; i++) | |
2756 | if (untracked->dirs[i]->recurse) | |
2757 | value++; | |
2758 | intlen = encode_varint(value, intbuf); | |
2759 | strbuf_add(out, intbuf, intlen); | |
2760 | ||
2761 | strbuf_add(out, untracked->name, strlen(untracked->name) + 1); | |
2762 | ||
2763 | for (i = 0; i < untracked->untracked_nr; i++) | |
2764 | strbuf_add(out, untracked->untracked[i], | |
2765 | strlen(untracked->untracked[i]) + 1); | |
2766 | ||
2767 | for (i = 0; i < untracked->dirs_nr; i++) | |
2768 | if (untracked->dirs[i]->recurse) | |
2769 | write_one_dir(untracked->dirs[i], wd); | |
2770 | } | |
2771 | ||
2772 | void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked) | |
2773 | { | |
2774 | struct ondisk_untracked_cache *ouc; | |
2775 | struct write_data wd; | |
2776 | unsigned char varbuf[16]; | |
2777 | int varint_len; | |
2778 | const unsigned hashsz = the_hash_algo->rawsz; | |
2779 | ||
2780 | ouc = xcalloc(1, sizeof(*ouc)); | |
2781 | stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat); | |
2782 | stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat); | |
2783 | ouc->dir_flags = htonl(untracked->dir_flags); | |
2784 | ||
2785 | varint_len = encode_varint(untracked->ident.len, varbuf); | |
2786 | strbuf_add(out, varbuf, varint_len); | |
2787 | strbuf_addbuf(out, &untracked->ident); | |
2788 | ||
2789 | strbuf_add(out, ouc, sizeof(*ouc)); | |
2790 | strbuf_add(out, untracked->ss_info_exclude.oid.hash, hashsz); | |
2791 | strbuf_add(out, untracked->ss_excludes_file.oid.hash, hashsz); | |
2792 | strbuf_add(out, untracked->exclude_per_dir, strlen(untracked->exclude_per_dir) + 1); | |
2793 | FREE_AND_NULL(ouc); | |
2794 | ||
2795 | if (!untracked->root) { | |
2796 | varint_len = encode_varint(0, varbuf); | |
2797 | strbuf_add(out, varbuf, varint_len); | |
2798 | return; | |
2799 | } | |
2800 | ||
2801 | wd.index = 0; | |
2802 | wd.check_only = ewah_new(); | |
2803 | wd.valid = ewah_new(); | |
2804 | wd.sha1_valid = ewah_new(); | |
2805 | strbuf_init(&wd.out, 1024); | |
2806 | strbuf_init(&wd.sb_stat, 1024); | |
2807 | strbuf_init(&wd.sb_sha1, 1024); | |
2808 | write_one_dir(untracked->root, &wd); | |
2809 | ||
2810 | varint_len = encode_varint(wd.index, varbuf); | |
2811 | strbuf_add(out, varbuf, varint_len); | |
2812 | strbuf_addbuf(out, &wd.out); | |
2813 | ewah_serialize_strbuf(wd.valid, out); | |
2814 | ewah_serialize_strbuf(wd.check_only, out); | |
2815 | ewah_serialize_strbuf(wd.sha1_valid, out); | |
2816 | strbuf_addbuf(out, &wd.sb_stat); | |
2817 | strbuf_addbuf(out, &wd.sb_sha1); | |
2818 | strbuf_addch(out, '\0'); /* safe guard for string lists */ | |
2819 | ||
2820 | ewah_free(wd.valid); | |
2821 | ewah_free(wd.check_only); | |
2822 | ewah_free(wd.sha1_valid); | |
2823 | strbuf_release(&wd.out); | |
2824 | strbuf_release(&wd.sb_stat); | |
2825 | strbuf_release(&wd.sb_sha1); | |
2826 | } | |
2827 | ||
2828 | static void free_untracked(struct untracked_cache_dir *ucd) | |
2829 | { | |
2830 | int i; | |
2831 | if (!ucd) | |
2832 | return; | |
2833 | for (i = 0; i < ucd->dirs_nr; i++) | |
2834 | free_untracked(ucd->dirs[i]); | |
2835 | for (i = 0; i < ucd->untracked_nr; i++) | |
2836 | free(ucd->untracked[i]); | |
2837 | free(ucd->untracked); | |
2838 | free(ucd->dirs); | |
2839 | free(ucd); | |
2840 | } | |
2841 | ||
2842 | void free_untracked_cache(struct untracked_cache *uc) | |
2843 | { | |
2844 | if (uc) | |
2845 | free_untracked(uc->root); | |
2846 | free(uc); | |
2847 | } | |
2848 | ||
2849 | struct read_data { | |
2850 | int index; | |
2851 | struct untracked_cache_dir **ucd; | |
2852 | struct ewah_bitmap *check_only; | |
2853 | struct ewah_bitmap *valid; | |
2854 | struct ewah_bitmap *sha1_valid; | |
2855 | const unsigned char *data; | |
2856 | const unsigned char *end; | |
2857 | }; | |
2858 | ||
2859 | static void stat_data_from_disk(struct stat_data *to, const unsigned char *data) | |
2860 | { | |
2861 | memcpy(to, data, sizeof(*to)); | |
2862 | to->sd_ctime.sec = ntohl(to->sd_ctime.sec); | |
2863 | to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec); | |
2864 | to->sd_mtime.sec = ntohl(to->sd_mtime.sec); | |
2865 | to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec); | |
2866 | to->sd_dev = ntohl(to->sd_dev); | |
2867 | to->sd_ino = ntohl(to->sd_ino); | |
2868 | to->sd_uid = ntohl(to->sd_uid); | |
2869 | to->sd_gid = ntohl(to->sd_gid); | |
2870 | to->sd_size = ntohl(to->sd_size); | |
2871 | } | |
2872 | ||
2873 | static int read_one_dir(struct untracked_cache_dir **untracked_, | |
2874 | struct read_data *rd) | |
2875 | { | |
2876 | struct untracked_cache_dir ud, *untracked; | |
2877 | const unsigned char *data = rd->data, *end = rd->end; | |
2878 | const unsigned char *eos; | |
2879 | unsigned int value; | |
2880 | int i; | |
2881 | ||
2882 | memset(&ud, 0, sizeof(ud)); | |
2883 | ||
2884 | value = decode_varint(&data); | |
2885 | if (data > end) | |
2886 | return -1; | |
2887 | ud.recurse = 1; | |
2888 | ud.untracked_alloc = value; | |
2889 | ud.untracked_nr = value; | |
2890 | if (ud.untracked_nr) | |
2891 | ALLOC_ARRAY(ud.untracked, ud.untracked_nr); | |
2892 | ||
2893 | ud.dirs_alloc = ud.dirs_nr = decode_varint(&data); | |
2894 | if (data > end) | |
2895 | return -1; | |
2896 | ALLOC_ARRAY(ud.dirs, ud.dirs_nr); | |
2897 | ||
2898 | eos = memchr(data, '\0', end - data); | |
2899 | if (!eos || eos == end) | |
2900 | return -1; | |
2901 | ||
2902 | *untracked_ = untracked = xmalloc(st_add3(sizeof(*untracked), eos - data, 1)); | |
2903 | memcpy(untracked, &ud, sizeof(ud)); | |
2904 | memcpy(untracked->name, data, eos - data + 1); | |
2905 | data = eos + 1; | |
2906 | ||
2907 | for (i = 0; i < untracked->untracked_nr; i++) { | |
2908 | eos = memchr(data, '\0', end - data); | |
2909 | if (!eos || eos == end) | |
2910 | return -1; | |
2911 | untracked->untracked[i] = xmemdupz(data, eos - data); | |
2912 | data = eos + 1; | |
2913 | } | |
2914 | ||
2915 | rd->ucd[rd->index++] = untracked; | |
2916 | rd->data = data; | |
2917 | ||
2918 | for (i = 0; i < untracked->dirs_nr; i++) { | |
2919 | if (read_one_dir(untracked->dirs + i, rd) < 0) | |
2920 | return -1; | |
2921 | } | |
2922 | return 0; | |
2923 | } | |
2924 | ||
2925 | static void set_check_only(size_t pos, void *cb) | |
2926 | { | |
2927 | struct read_data *rd = cb; | |
2928 | struct untracked_cache_dir *ud = rd->ucd[pos]; | |
2929 | ud->check_only = 1; | |
2930 | } | |
2931 | ||
2932 | static void read_stat(size_t pos, void *cb) | |
2933 | { | |
2934 | struct read_data *rd = cb; | |
2935 | struct untracked_cache_dir *ud = rd->ucd[pos]; | |
2936 | if (rd->data + sizeof(struct stat_data) > rd->end) { | |
2937 | rd->data = rd->end + 1; | |
2938 | return; | |
2939 | } | |
2940 | stat_data_from_disk(&ud->stat_data, rd->data); | |
2941 | rd->data += sizeof(struct stat_data); | |
2942 | ud->valid = 1; | |
2943 | } | |
2944 | ||
2945 | static void read_oid(size_t pos, void *cb) | |
2946 | { | |
2947 | struct read_data *rd = cb; | |
2948 | struct untracked_cache_dir *ud = rd->ucd[pos]; | |
2949 | if (rd->data + the_hash_algo->rawsz > rd->end) { | |
2950 | rd->data = rd->end + 1; | |
2951 | return; | |
2952 | } | |
2953 | hashcpy(ud->exclude_oid.hash, rd->data); | |
2954 | rd->data += the_hash_algo->rawsz; | |
2955 | } | |
2956 | ||
2957 | static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data, | |
2958 | const unsigned char *sha1) | |
2959 | { | |
2960 | stat_data_from_disk(&oid_stat->stat, data); | |
2961 | hashcpy(oid_stat->oid.hash, sha1); | |
2962 | oid_stat->valid = 1; | |
2963 | } | |
2964 | ||
2965 | struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz) | |
2966 | { | |
2967 | struct untracked_cache *uc; | |
2968 | struct read_data rd; | |
2969 | const unsigned char *next = data, *end = (const unsigned char *)data + sz; | |
2970 | const char *ident; | |
2971 | int ident_len; | |
2972 | ssize_t len; | |
2973 | const char *exclude_per_dir; | |
2974 | const unsigned hashsz = the_hash_algo->rawsz; | |
2975 | const unsigned offset = sizeof(struct ondisk_untracked_cache); | |
2976 | const unsigned exclude_per_dir_offset = offset + 2 * hashsz; | |
2977 | ||
2978 | if (sz <= 1 || end[-1] != '\0') | |
2979 | return NULL; | |
2980 | end--; | |
2981 | ||
2982 | ident_len = decode_varint(&next); | |
2983 | if (next + ident_len > end) | |
2984 | return NULL; | |
2985 | ident = (const char *)next; | |
2986 | next += ident_len; | |
2987 | ||
2988 | if (next + exclude_per_dir_offset + 1 > end) | |
2989 | return NULL; | |
2990 | ||
2991 | uc = xcalloc(1, sizeof(*uc)); | |
2992 | strbuf_init(&uc->ident, ident_len); | |
2993 | strbuf_add(&uc->ident, ident, ident_len); | |
2994 | load_oid_stat(&uc->ss_info_exclude, | |
2995 | next + ouc_offset(info_exclude_stat), | |
2996 | next + offset); | |
2997 | load_oid_stat(&uc->ss_excludes_file, | |
2998 | next + ouc_offset(excludes_file_stat), | |
2999 | next + offset + hashsz); | |
3000 | uc->dir_flags = get_be32(next + ouc_offset(dir_flags)); | |
3001 | exclude_per_dir = (const char *)next + exclude_per_dir_offset; | |
3002 | uc->exclude_per_dir = xstrdup(exclude_per_dir); | |
3003 | /* NUL after exclude_per_dir is covered by sizeof(*ouc) */ | |
3004 | next += exclude_per_dir_offset + strlen(exclude_per_dir) + 1; | |
3005 | if (next >= end) | |
3006 | goto done2; | |
3007 | ||
3008 | len = decode_varint(&next); | |
3009 | if (next > end || len == 0) | |
3010 | goto done2; | |
3011 | ||
3012 | rd.valid = ewah_new(); | |
3013 | rd.check_only = ewah_new(); | |
3014 | rd.sha1_valid = ewah_new(); | |
3015 | rd.data = next; | |
3016 | rd.end = end; | |
3017 | rd.index = 0; | |
3018 | ALLOC_ARRAY(rd.ucd, len); | |
3019 | ||
3020 | if (read_one_dir(&uc->root, &rd) || rd.index != len) | |
3021 | goto done; | |
3022 | ||
3023 | next = rd.data; | |
3024 | len = ewah_read_mmap(rd.valid, next, end - next); | |
3025 | if (len < 0) | |
3026 | goto done; | |
3027 | ||
3028 | next += len; | |
3029 | len = ewah_read_mmap(rd.check_only, next, end - next); | |
3030 | if (len < 0) | |
3031 | goto done; | |
3032 | ||
3033 | next += len; | |
3034 | len = ewah_read_mmap(rd.sha1_valid, next, end - next); | |
3035 | if (len < 0) | |
3036 | goto done; | |
3037 | ||
3038 | ewah_each_bit(rd.check_only, set_check_only, &rd); | |
3039 | rd.data = next + len; | |
3040 | ewah_each_bit(rd.valid, read_stat, &rd); | |
3041 | ewah_each_bit(rd.sha1_valid, read_oid, &rd); | |
3042 | next = rd.data; | |
3043 | ||
3044 | done: | |
3045 | free(rd.ucd); | |
3046 | ewah_free(rd.valid); | |
3047 | ewah_free(rd.check_only); | |
3048 | ewah_free(rd.sha1_valid); | |
3049 | done2: | |
3050 | if (next != end) { | |
3051 | free_untracked_cache(uc); | |
3052 | uc = NULL; | |
3053 | } | |
3054 | return uc; | |
3055 | } | |
3056 | ||
3057 | static void invalidate_one_directory(struct untracked_cache *uc, | |
3058 | struct untracked_cache_dir *ucd) | |
3059 | { | |
3060 | uc->dir_invalidated++; | |
3061 | ucd->valid = 0; | |
3062 | ucd->untracked_nr = 0; | |
3063 | } | |
3064 | ||
3065 | /* | |
3066 | * Normally when an entry is added or removed from a directory, | |
3067 | * invalidating that directory is enough. No need to touch its | |
3068 | * ancestors. When a directory is shown as "foo/bar/" in git-status | |
3069 | * however, deleting or adding an entry may have cascading effect. | |
3070 | * | |
3071 | * Say the "foo/bar/file" has become untracked, we need to tell the | |
3072 | * untracked_cache_dir of "foo" that "bar/" is not an untracked | |
3073 | * directory any more (because "bar" is managed by foo as an untracked | |
3074 | * "file"). | |
3075 | * | |
3076 | * Similarly, if "foo/bar/file" moves from untracked to tracked and it | |
3077 | * was the last untracked entry in the entire "foo", we should show | |
3078 | * "foo/" instead. Which means we have to invalidate past "bar" up to | |
3079 | * "foo". | |
3080 | * | |
3081 | * This function traverses all directories from root to leaf. If there | |
3082 | * is a chance of one of the above cases happening, we invalidate back | |
3083 | * to root. Otherwise we just invalidate the leaf. There may be a more | |
3084 | * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to | |
3085 | * detect these cases and avoid unnecessary invalidation, for example, | |
3086 | * checking for the untracked entry named "bar/" in "foo", but for now | |
3087 | * stick to something safe and simple. | |
3088 | */ | |
3089 | static int invalidate_one_component(struct untracked_cache *uc, | |
3090 | struct untracked_cache_dir *dir, | |
3091 | const char *path, int len) | |
3092 | { | |
3093 | const char *rest = strchr(path, '/'); | |
3094 | ||
3095 | if (rest) { | |
3096 | int component_len = rest - path; | |
3097 | struct untracked_cache_dir *d = | |
3098 | lookup_untracked(uc, dir, path, component_len); | |
3099 | int ret = | |
3100 | invalidate_one_component(uc, d, rest + 1, | |
3101 | len - (component_len + 1)); | |
3102 | if (ret) | |
3103 | invalidate_one_directory(uc, dir); | |
3104 | return ret; | |
3105 | } | |
3106 | ||
3107 | invalidate_one_directory(uc, dir); | |
3108 | return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES; | |
3109 | } | |
3110 | ||
3111 | void untracked_cache_invalidate_path(struct index_state *istate, | |
3112 | const char *path, int safe_path) | |
3113 | { | |
3114 | if (!istate->untracked || !istate->untracked->root) | |
3115 | return; | |
3116 | if (!safe_path && !verify_path(path, 0)) | |
3117 | return; | |
3118 | invalidate_one_component(istate->untracked, istate->untracked->root, | |
3119 | path, strlen(path)); | |
3120 | } | |
3121 | ||
3122 | void untracked_cache_remove_from_index(struct index_state *istate, | |
3123 | const char *path) | |
3124 | { | |
3125 | untracked_cache_invalidate_path(istate, path, 1); | |
3126 | } | |
3127 | ||
3128 | void untracked_cache_add_to_index(struct index_state *istate, | |
3129 | const char *path) | |
3130 | { | |
3131 | untracked_cache_invalidate_path(istate, path, 1); | |
3132 | } | |
3133 | ||
3134 | static void connect_wt_gitdir_in_nested(const char *sub_worktree, | |
3135 | const char *sub_gitdir) | |
3136 | { | |
3137 | int i; | |
3138 | struct repository subrepo; | |
3139 | struct strbuf sub_wt = STRBUF_INIT; | |
3140 | struct strbuf sub_gd = STRBUF_INIT; | |
3141 | ||
3142 | const struct submodule *sub; | |
3143 | ||
3144 | /* If the submodule has no working tree, we can ignore it. */ | |
3145 | if (repo_init(&subrepo, sub_gitdir, sub_worktree)) | |
3146 | return; | |
3147 | ||
3148 | if (repo_read_index(&subrepo) < 0) | |
3149 | die(_("index file corrupt in repo %s"), subrepo.gitdir); | |
3150 | ||
3151 | for (i = 0; i < subrepo.index->cache_nr; i++) { | |
3152 | const struct cache_entry *ce = subrepo.index->cache[i]; | |
3153 | ||
3154 | if (!S_ISGITLINK(ce->ce_mode)) | |
3155 | continue; | |
3156 | ||
3157 | while (i + 1 < subrepo.index->cache_nr && | |
3158 | !strcmp(ce->name, subrepo.index->cache[i + 1]->name)) | |
3159 | /* | |
3160 | * Skip entries with the same name in different stages | |
3161 | * to make sure an entry is returned only once. | |
3162 | */ | |
3163 | i++; | |
3164 | ||
3165 | sub = submodule_from_path(&subrepo, &null_oid, ce->name); | |
3166 | if (!sub || !is_submodule_active(&subrepo, ce->name)) | |
3167 | /* .gitmodules broken or inactive sub */ | |
3168 | continue; | |
3169 | ||
3170 | strbuf_reset(&sub_wt); | |
3171 | strbuf_reset(&sub_gd); | |
3172 | strbuf_addf(&sub_wt, "%s/%s", sub_worktree, sub->path); | |
3173 | strbuf_addf(&sub_gd, "%s/modules/%s", sub_gitdir, sub->name); | |
3174 | ||
3175 | connect_work_tree_and_git_dir(sub_wt.buf, sub_gd.buf, 1); | |
3176 | } | |
3177 | strbuf_release(&sub_wt); | |
3178 | strbuf_release(&sub_gd); | |
3179 | repo_clear(&subrepo); | |
3180 | } | |
3181 | ||
3182 | void connect_work_tree_and_git_dir(const char *work_tree_, | |
3183 | const char *git_dir_, | |
3184 | int recurse_into_nested) | |
3185 | { | |
3186 | struct strbuf gitfile_sb = STRBUF_INIT; | |
3187 | struct strbuf cfg_sb = STRBUF_INIT; | |
3188 | struct strbuf rel_path = STRBUF_INIT; | |
3189 | char *git_dir, *work_tree; | |
3190 | ||
3191 | /* Prepare .git file */ | |
3192 | strbuf_addf(&gitfile_sb, "%s/.git", work_tree_); | |
3193 | if (safe_create_leading_directories_const(gitfile_sb.buf)) | |
3194 | die(_("could not create directories for %s"), gitfile_sb.buf); | |
3195 | ||
3196 | /* Prepare config file */ | |
3197 | strbuf_addf(&cfg_sb, "%s/config", git_dir_); | |
3198 | if (safe_create_leading_directories_const(cfg_sb.buf)) | |
3199 | die(_("could not create directories for %s"), cfg_sb.buf); | |
3200 | ||
3201 | git_dir = real_pathdup(git_dir_, 1); | |
3202 | work_tree = real_pathdup(work_tree_, 1); | |
3203 | ||
3204 | /* Write .git file */ | |
3205 | write_file(gitfile_sb.buf, "gitdir: %s", | |
3206 | relative_path(git_dir, work_tree, &rel_path)); | |
3207 | /* Update core.worktree setting */ | |
3208 | git_config_set_in_file(cfg_sb.buf, "core.worktree", | |
3209 | relative_path(work_tree, git_dir, &rel_path)); | |
3210 | ||
3211 | strbuf_release(&gitfile_sb); | |
3212 | strbuf_release(&cfg_sb); | |
3213 | strbuf_release(&rel_path); | |
3214 | ||
3215 | if (recurse_into_nested) | |
3216 | connect_wt_gitdir_in_nested(work_tree, git_dir); | |
3217 | ||
3218 | free(work_tree); | |
3219 | free(git_dir); | |
3220 | } | |
3221 | ||
3222 | /* | |
3223 | * Migrate the git directory of the given path from old_git_dir to new_git_dir. | |
3224 | */ | |
3225 | void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir) | |
3226 | { | |
3227 | if (rename(old_git_dir, new_git_dir) < 0) | |
3228 | die_errno(_("could not migrate git directory from '%s' to '%s'"), | |
3229 | old_git_dir, new_git_dir); | |
3230 | ||
3231 | connect_work_tree_and_git_dir(path, new_git_dir, 0); | |
3232 | } |