]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Handle git attributes. See gitattributes(5) for a description of | |
3 | * the file syntax, and Documentation/technical/api-gitattributes.txt | |
4 | * for a description of the API. | |
5 | * | |
6 | * One basic design decision here is that we are not going to support | |
7 | * an insanely large number of attributes. | |
8 | */ | |
9 | ||
10 | #include "cache.h" | |
11 | #include "config.h" | |
12 | #include "exec-cmd.h" | |
13 | #include "attr.h" | |
14 | #include "dir.h" | |
15 | #include "utf8.h" | |
16 | #include "quote.h" | |
17 | #include "thread-utils.h" | |
18 | ||
19 | const char git_attr__true[] = "(builtin)true"; | |
20 | const char git_attr__false[] = "\0(builtin)false"; | |
21 | static const char git_attr__unknown[] = "(builtin)unknown"; | |
22 | #define ATTR__TRUE git_attr__true | |
23 | #define ATTR__FALSE git_attr__false | |
24 | #define ATTR__UNSET NULL | |
25 | #define ATTR__UNKNOWN git_attr__unknown | |
26 | ||
27 | #ifndef DEBUG_ATTR | |
28 | #define DEBUG_ATTR 0 | |
29 | #endif | |
30 | ||
31 | struct git_attr { | |
32 | int attr_nr; /* unique attribute number */ | |
33 | char name[FLEX_ARRAY]; /* attribute name */ | |
34 | }; | |
35 | ||
36 | const char *git_attr_name(const struct git_attr *attr) | |
37 | { | |
38 | return attr->name; | |
39 | } | |
40 | ||
41 | struct attr_hashmap { | |
42 | struct hashmap map; | |
43 | pthread_mutex_t mutex; | |
44 | }; | |
45 | ||
46 | static inline void hashmap_lock(struct attr_hashmap *map) | |
47 | { | |
48 | pthread_mutex_lock(&map->mutex); | |
49 | } | |
50 | ||
51 | static inline void hashmap_unlock(struct attr_hashmap *map) | |
52 | { | |
53 | pthread_mutex_unlock(&map->mutex); | |
54 | } | |
55 | ||
56 | /* | |
57 | * The global dictionary of all interned attributes. This | |
58 | * is a singleton object which is shared between threads. | |
59 | * Access to this dictionary must be surrounded with a mutex. | |
60 | */ | |
61 | static struct attr_hashmap g_attr_hashmap; | |
62 | ||
63 | /* The container for objects stored in "struct attr_hashmap" */ | |
64 | struct attr_hash_entry { | |
65 | struct hashmap_entry ent; | |
66 | const char *key; /* the key; memory should be owned by value */ | |
67 | size_t keylen; /* length of the key */ | |
68 | void *value; /* the stored value */ | |
69 | }; | |
70 | ||
71 | /* attr_hashmap comparison function */ | |
72 | static int attr_hash_entry_cmp(const void *unused_cmp_data, | |
73 | const struct hashmap_entry *eptr, | |
74 | const struct hashmap_entry *entry_or_key, | |
75 | const void *unused_keydata) | |
76 | { | |
77 | const struct attr_hash_entry *a, *b; | |
78 | ||
79 | a = container_of(eptr, const struct attr_hash_entry, ent); | |
80 | b = container_of(entry_or_key, const struct attr_hash_entry, ent); | |
81 | return (a->keylen != b->keylen) || strncmp(a->key, b->key, a->keylen); | |
82 | } | |
83 | ||
84 | /* Initialize an 'attr_hashmap' object */ | |
85 | static void attr_hashmap_init(struct attr_hashmap *map) | |
86 | { | |
87 | hashmap_init(&map->map, attr_hash_entry_cmp, NULL, 0); | |
88 | } | |
89 | ||
90 | /* | |
91 | * Retrieve the 'value' stored in a hashmap given the provided 'key'. | |
92 | * If there is no matching entry, return NULL. | |
93 | */ | |
94 | static void *attr_hashmap_get(struct attr_hashmap *map, | |
95 | const char *key, size_t keylen) | |
96 | { | |
97 | struct attr_hash_entry k; | |
98 | struct attr_hash_entry *e; | |
99 | ||
100 | if (!map->map.tablesize) | |
101 | attr_hashmap_init(map); | |
102 | ||
103 | hashmap_entry_init(&k.ent, memhash(key, keylen)); | |
104 | k.key = key; | |
105 | k.keylen = keylen; | |
106 | e = hashmap_get_entry(&map->map, &k, ent, NULL); | |
107 | ||
108 | return e ? e->value : NULL; | |
109 | } | |
110 | ||
111 | /* Add 'value' to a hashmap based on the provided 'key'. */ | |
112 | static void attr_hashmap_add(struct attr_hashmap *map, | |
113 | const char *key, size_t keylen, | |
114 | void *value) | |
115 | { | |
116 | struct attr_hash_entry *e; | |
117 | ||
118 | if (!map->map.tablesize) | |
119 | attr_hashmap_init(map); | |
120 | ||
121 | e = xmalloc(sizeof(struct attr_hash_entry)); | |
122 | hashmap_entry_init(&e->ent, memhash(key, keylen)); | |
123 | e->key = key; | |
124 | e->keylen = keylen; | |
125 | e->value = value; | |
126 | ||
127 | hashmap_add(&map->map, &e->ent); | |
128 | } | |
129 | ||
130 | struct all_attrs_item { | |
131 | const struct git_attr *attr; | |
132 | const char *value; | |
133 | /* | |
134 | * If 'macro' is non-NULL, indicates that 'attr' is a macro based on | |
135 | * the current attribute stack and contains a pointer to the match_attr | |
136 | * definition of the macro | |
137 | */ | |
138 | const struct match_attr *macro; | |
139 | }; | |
140 | ||
141 | /* | |
142 | * Reallocate and reinitialize the array of all attributes (which is used in | |
143 | * the attribute collection process) in 'check' based on the global dictionary | |
144 | * of attributes. | |
145 | */ | |
146 | static void all_attrs_init(struct attr_hashmap *map, struct attr_check *check) | |
147 | { | |
148 | int i; | |
149 | unsigned int size; | |
150 | ||
151 | hashmap_lock(map); | |
152 | ||
153 | size = hashmap_get_size(&map->map); | |
154 | if (size < check->all_attrs_nr) | |
155 | BUG("interned attributes shouldn't be deleted"); | |
156 | ||
157 | /* | |
158 | * If the number of attributes in the global dictionary has increased | |
159 | * (or this attr_check instance doesn't have an initialized all_attrs | |
160 | * field), reallocate the provided attr_check instance's all_attrs | |
161 | * field and fill each entry with its corresponding git_attr. | |
162 | */ | |
163 | if (size != check->all_attrs_nr) { | |
164 | struct attr_hash_entry *e; | |
165 | struct hashmap_iter iter; | |
166 | ||
167 | REALLOC_ARRAY(check->all_attrs, size); | |
168 | check->all_attrs_nr = size; | |
169 | ||
170 | hashmap_for_each_entry(&map->map, &iter, e, | |
171 | ent /* member name */) { | |
172 | const struct git_attr *a = e->value; | |
173 | check->all_attrs[a->attr_nr].attr = a; | |
174 | } | |
175 | } | |
176 | ||
177 | hashmap_unlock(map); | |
178 | ||
179 | /* | |
180 | * Re-initialize every entry in check->all_attrs. | |
181 | * This re-initialization can live outside of the locked region since | |
182 | * the attribute dictionary is no longer being accessed. | |
183 | */ | |
184 | for (i = 0; i < check->all_attrs_nr; i++) { | |
185 | check->all_attrs[i].value = ATTR__UNKNOWN; | |
186 | check->all_attrs[i].macro = NULL; | |
187 | } | |
188 | } | |
189 | ||
190 | static int attr_name_valid(const char *name, size_t namelen) | |
191 | { | |
192 | /* | |
193 | * Attribute name cannot begin with '-' and must consist of | |
194 | * characters from [-A-Za-z0-9_.]. | |
195 | */ | |
196 | if (namelen <= 0 || *name == '-') | |
197 | return 0; | |
198 | while (namelen--) { | |
199 | char ch = *name++; | |
200 | if (! (ch == '-' || ch == '.' || ch == '_' || | |
201 | ('0' <= ch && ch <= '9') || | |
202 | ('a' <= ch && ch <= 'z') || | |
203 | ('A' <= ch && ch <= 'Z')) ) | |
204 | return 0; | |
205 | } | |
206 | return 1; | |
207 | } | |
208 | ||
209 | static void report_invalid_attr(const char *name, size_t len, | |
210 | const char *src, int lineno) | |
211 | { | |
212 | struct strbuf err = STRBUF_INIT; | |
213 | strbuf_addf(&err, _("%.*s is not a valid attribute name"), | |
214 | (int) len, name); | |
215 | fprintf(stderr, "%s: %s:%d\n", err.buf, src, lineno); | |
216 | strbuf_release(&err); | |
217 | } | |
218 | ||
219 | /* | |
220 | * Given a 'name', lookup and return the corresponding attribute in the global | |
221 | * dictionary. If no entry is found, create a new attribute and store it in | |
222 | * the dictionary. | |
223 | */ | |
224 | static const struct git_attr *git_attr_internal(const char *name, int namelen) | |
225 | { | |
226 | struct git_attr *a; | |
227 | ||
228 | if (!attr_name_valid(name, namelen)) | |
229 | return NULL; | |
230 | ||
231 | hashmap_lock(&g_attr_hashmap); | |
232 | ||
233 | a = attr_hashmap_get(&g_attr_hashmap, name, namelen); | |
234 | ||
235 | if (!a) { | |
236 | FLEX_ALLOC_MEM(a, name, name, namelen); | |
237 | a->attr_nr = hashmap_get_size(&g_attr_hashmap.map); | |
238 | ||
239 | attr_hashmap_add(&g_attr_hashmap, a->name, namelen, a); | |
240 | assert(a->attr_nr == | |
241 | (hashmap_get_size(&g_attr_hashmap.map) - 1)); | |
242 | } | |
243 | ||
244 | hashmap_unlock(&g_attr_hashmap); | |
245 | ||
246 | return a; | |
247 | } | |
248 | ||
249 | const struct git_attr *git_attr(const char *name) | |
250 | { | |
251 | return git_attr_internal(name, strlen(name)); | |
252 | } | |
253 | ||
254 | /* What does a matched pattern decide? */ | |
255 | struct attr_state { | |
256 | const struct git_attr *attr; | |
257 | const char *setto; | |
258 | }; | |
259 | ||
260 | struct pattern { | |
261 | const char *pattern; | |
262 | int patternlen; | |
263 | int nowildcardlen; | |
264 | unsigned flags; /* PATTERN_FLAG_* */ | |
265 | }; | |
266 | ||
267 | /* | |
268 | * One rule, as from a .gitattributes file. | |
269 | * | |
270 | * If is_macro is true, then u.attr is a pointer to the git_attr being | |
271 | * defined. | |
272 | * | |
273 | * If is_macro is false, then u.pat is the filename pattern to which the | |
274 | * rule applies. | |
275 | * | |
276 | * In either case, num_attr is the number of attributes affected by | |
277 | * this rule, and state is an array listing them. The attributes are | |
278 | * listed as they appear in the file (macros unexpanded). | |
279 | */ | |
280 | struct match_attr { | |
281 | union { | |
282 | struct pattern pat; | |
283 | const struct git_attr *attr; | |
284 | } u; | |
285 | char is_macro; | |
286 | unsigned num_attr; | |
287 | struct attr_state state[FLEX_ARRAY]; | |
288 | }; | |
289 | ||
290 | static const char blank[] = " \t\r\n"; | |
291 | ||
292 | /* | |
293 | * Parse a whitespace-delimited attribute state (i.e., "attr", | |
294 | * "-attr", "!attr", or "attr=value") from the string starting at src. | |
295 | * If e is not NULL, write the results to *e. Return a pointer to the | |
296 | * remainder of the string (with leading whitespace removed), or NULL | |
297 | * if there was an error. | |
298 | */ | |
299 | static const char *parse_attr(const char *src, int lineno, const char *cp, | |
300 | struct attr_state *e) | |
301 | { | |
302 | const char *ep, *equals; | |
303 | int len; | |
304 | ||
305 | ep = cp + strcspn(cp, blank); | |
306 | equals = strchr(cp, '='); | |
307 | if (equals && ep < equals) | |
308 | equals = NULL; | |
309 | if (equals) | |
310 | len = equals - cp; | |
311 | else | |
312 | len = ep - cp; | |
313 | if (!e) { | |
314 | if (*cp == '-' || *cp == '!') { | |
315 | cp++; | |
316 | len--; | |
317 | } | |
318 | if (!attr_name_valid(cp, len)) { | |
319 | report_invalid_attr(cp, len, src, lineno); | |
320 | return NULL; | |
321 | } | |
322 | } else { | |
323 | /* | |
324 | * As this function is always called twice, once with | |
325 | * e == NULL in the first pass and then e != NULL in | |
326 | * the second pass, no need for attr_name_valid() | |
327 | * check here. | |
328 | */ | |
329 | if (*cp == '-' || *cp == '!') { | |
330 | e->setto = (*cp == '-') ? ATTR__FALSE : ATTR__UNSET; | |
331 | cp++; | |
332 | len--; | |
333 | } | |
334 | else if (!equals) | |
335 | e->setto = ATTR__TRUE; | |
336 | else { | |
337 | e->setto = xmemdupz(equals + 1, ep - equals - 1); | |
338 | } | |
339 | e->attr = git_attr_internal(cp, len); | |
340 | } | |
341 | return ep + strspn(ep, blank); | |
342 | } | |
343 | ||
344 | static struct match_attr *parse_attr_line(const char *line, const char *src, | |
345 | int lineno, int macro_ok) | |
346 | { | |
347 | int namelen; | |
348 | int num_attr, i; | |
349 | const char *cp, *name, *states; | |
350 | struct match_attr *res = NULL; | |
351 | int is_macro; | |
352 | struct strbuf pattern = STRBUF_INIT; | |
353 | ||
354 | cp = line + strspn(line, blank); | |
355 | if (!*cp || *cp == '#') | |
356 | return NULL; | |
357 | name = cp; | |
358 | ||
359 | if (*cp == '"' && !unquote_c_style(&pattern, name, &states)) { | |
360 | name = pattern.buf; | |
361 | namelen = pattern.len; | |
362 | } else { | |
363 | namelen = strcspn(name, blank); | |
364 | states = name + namelen; | |
365 | } | |
366 | ||
367 | if (strlen(ATTRIBUTE_MACRO_PREFIX) < namelen && | |
368 | starts_with(name, ATTRIBUTE_MACRO_PREFIX)) { | |
369 | if (!macro_ok) { | |
370 | fprintf_ln(stderr, _("%s not allowed: %s:%d"), | |
371 | name, src, lineno); | |
372 | goto fail_return; | |
373 | } | |
374 | is_macro = 1; | |
375 | name += strlen(ATTRIBUTE_MACRO_PREFIX); | |
376 | name += strspn(name, blank); | |
377 | namelen = strcspn(name, blank); | |
378 | if (!attr_name_valid(name, namelen)) { | |
379 | report_invalid_attr(name, namelen, src, lineno); | |
380 | goto fail_return; | |
381 | } | |
382 | } | |
383 | else | |
384 | is_macro = 0; | |
385 | ||
386 | states += strspn(states, blank); | |
387 | ||
388 | /* First pass to count the attr_states */ | |
389 | for (cp = states, num_attr = 0; *cp; num_attr++) { | |
390 | cp = parse_attr(src, lineno, cp, NULL); | |
391 | if (!cp) | |
392 | goto fail_return; | |
393 | } | |
394 | ||
395 | res = xcalloc(1, | |
396 | sizeof(*res) + | |
397 | sizeof(struct attr_state) * num_attr + | |
398 | (is_macro ? 0 : namelen + 1)); | |
399 | if (is_macro) { | |
400 | res->u.attr = git_attr_internal(name, namelen); | |
401 | } else { | |
402 | char *p = (char *)&(res->state[num_attr]); | |
403 | memcpy(p, name, namelen); | |
404 | res->u.pat.pattern = p; | |
405 | parse_path_pattern(&res->u.pat.pattern, | |
406 | &res->u.pat.patternlen, | |
407 | &res->u.pat.flags, | |
408 | &res->u.pat.nowildcardlen); | |
409 | if (res->u.pat.flags & PATTERN_FLAG_NEGATIVE) { | |
410 | warning(_("Negative patterns are ignored in git attributes\n" | |
411 | "Use '\\!' for literal leading exclamation.")); | |
412 | goto fail_return; | |
413 | } | |
414 | } | |
415 | res->is_macro = is_macro; | |
416 | res->num_attr = num_attr; | |
417 | ||
418 | /* Second pass to fill the attr_states */ | |
419 | for (cp = states, i = 0; *cp; i++) { | |
420 | cp = parse_attr(src, lineno, cp, &(res->state[i])); | |
421 | } | |
422 | ||
423 | strbuf_release(&pattern); | |
424 | return res; | |
425 | ||
426 | fail_return: | |
427 | strbuf_release(&pattern); | |
428 | free(res); | |
429 | return NULL; | |
430 | } | |
431 | ||
432 | /* | |
433 | * Like info/exclude and .gitignore, the attribute information can | |
434 | * come from many places. | |
435 | * | |
436 | * (1) .gitattributes file of the same directory; | |
437 | * (2) .gitattributes file of the parent directory if (1) does not have | |
438 | * any match; this goes recursively upwards, just like .gitignore. | |
439 | * (3) $GIT_DIR/info/attributes, which overrides both of the above. | |
440 | * | |
441 | * In the same file, later entries override the earlier match, so in the | |
442 | * global list, we would have entries from info/attributes the earliest | |
443 | * (reading the file from top to bottom), .gitattributes of the root | |
444 | * directory (again, reading the file from top to bottom) down to the | |
445 | * current directory, and then scan the list backwards to find the first match. | |
446 | * This is exactly the same as what is_excluded() does in dir.c to deal with | |
447 | * .gitignore file and info/excludes file as a fallback. | |
448 | */ | |
449 | ||
450 | struct attr_stack { | |
451 | struct attr_stack *prev; | |
452 | char *origin; | |
453 | size_t originlen; | |
454 | unsigned num_matches; | |
455 | unsigned alloc; | |
456 | struct match_attr **attrs; | |
457 | }; | |
458 | ||
459 | static void attr_stack_free(struct attr_stack *e) | |
460 | { | |
461 | int i; | |
462 | free(e->origin); | |
463 | for (i = 0; i < e->num_matches; i++) { | |
464 | struct match_attr *a = e->attrs[i]; | |
465 | int j; | |
466 | for (j = 0; j < a->num_attr; j++) { | |
467 | const char *setto = a->state[j].setto; | |
468 | if (setto == ATTR__TRUE || | |
469 | setto == ATTR__FALSE || | |
470 | setto == ATTR__UNSET || | |
471 | setto == ATTR__UNKNOWN) | |
472 | ; | |
473 | else | |
474 | free((char *) setto); | |
475 | } | |
476 | free(a); | |
477 | } | |
478 | free(e->attrs); | |
479 | free(e); | |
480 | } | |
481 | ||
482 | static void drop_attr_stack(struct attr_stack **stack) | |
483 | { | |
484 | while (*stack) { | |
485 | struct attr_stack *elem = *stack; | |
486 | *stack = elem->prev; | |
487 | attr_stack_free(elem); | |
488 | } | |
489 | } | |
490 | ||
491 | /* List of all attr_check structs; access should be surrounded by mutex */ | |
492 | static struct check_vector { | |
493 | size_t nr; | |
494 | size_t alloc; | |
495 | struct attr_check **checks; | |
496 | pthread_mutex_t mutex; | |
497 | } check_vector; | |
498 | ||
499 | static inline void vector_lock(void) | |
500 | { | |
501 | pthread_mutex_lock(&check_vector.mutex); | |
502 | } | |
503 | ||
504 | static inline void vector_unlock(void) | |
505 | { | |
506 | pthread_mutex_unlock(&check_vector.mutex); | |
507 | } | |
508 | ||
509 | static void check_vector_add(struct attr_check *c) | |
510 | { | |
511 | vector_lock(); | |
512 | ||
513 | ALLOC_GROW(check_vector.checks, | |
514 | check_vector.nr + 1, | |
515 | check_vector.alloc); | |
516 | check_vector.checks[check_vector.nr++] = c; | |
517 | ||
518 | vector_unlock(); | |
519 | } | |
520 | ||
521 | static void check_vector_remove(struct attr_check *check) | |
522 | { | |
523 | int i; | |
524 | ||
525 | vector_lock(); | |
526 | ||
527 | /* Find entry */ | |
528 | for (i = 0; i < check_vector.nr; i++) | |
529 | if (check_vector.checks[i] == check) | |
530 | break; | |
531 | ||
532 | if (i >= check_vector.nr) | |
533 | BUG("no entry found"); | |
534 | ||
535 | /* shift entries over */ | |
536 | for (; i < check_vector.nr - 1; i++) | |
537 | check_vector.checks[i] = check_vector.checks[i + 1]; | |
538 | ||
539 | check_vector.nr--; | |
540 | ||
541 | vector_unlock(); | |
542 | } | |
543 | ||
544 | /* Iterate through all attr_check instances and drop their stacks */ | |
545 | static void drop_all_attr_stacks(void) | |
546 | { | |
547 | int i; | |
548 | ||
549 | vector_lock(); | |
550 | ||
551 | for (i = 0; i < check_vector.nr; i++) { | |
552 | drop_attr_stack(&check_vector.checks[i]->stack); | |
553 | } | |
554 | ||
555 | vector_unlock(); | |
556 | } | |
557 | ||
558 | struct attr_check *attr_check_alloc(void) | |
559 | { | |
560 | struct attr_check *c = xcalloc(1, sizeof(struct attr_check)); | |
561 | ||
562 | /* save pointer to the check struct */ | |
563 | check_vector_add(c); | |
564 | ||
565 | return c; | |
566 | } | |
567 | ||
568 | struct attr_check *attr_check_initl(const char *one, ...) | |
569 | { | |
570 | struct attr_check *check; | |
571 | int cnt; | |
572 | va_list params; | |
573 | const char *param; | |
574 | ||
575 | va_start(params, one); | |
576 | for (cnt = 1; (param = va_arg(params, const char *)) != NULL; cnt++) | |
577 | ; | |
578 | va_end(params); | |
579 | ||
580 | check = attr_check_alloc(); | |
581 | check->nr = cnt; | |
582 | check->alloc = cnt; | |
583 | check->items = xcalloc(cnt, sizeof(struct attr_check_item)); | |
584 | ||
585 | check->items[0].attr = git_attr(one); | |
586 | va_start(params, one); | |
587 | for (cnt = 1; cnt < check->nr; cnt++) { | |
588 | const struct git_attr *attr; | |
589 | param = va_arg(params, const char *); | |
590 | if (!param) | |
591 | BUG("counted %d != ended at %d", | |
592 | check->nr, cnt); | |
593 | attr = git_attr(param); | |
594 | if (!attr) | |
595 | BUG("%s: not a valid attribute name", param); | |
596 | check->items[cnt].attr = attr; | |
597 | } | |
598 | va_end(params); | |
599 | return check; | |
600 | } | |
601 | ||
602 | struct attr_check *attr_check_dup(const struct attr_check *check) | |
603 | { | |
604 | struct attr_check *ret; | |
605 | ||
606 | if (!check) | |
607 | return NULL; | |
608 | ||
609 | ret = attr_check_alloc(); | |
610 | ||
611 | ret->nr = check->nr; | |
612 | ret->alloc = check->alloc; | |
613 | ALLOC_ARRAY(ret->items, ret->nr); | |
614 | COPY_ARRAY(ret->items, check->items, ret->nr); | |
615 | ||
616 | return ret; | |
617 | } | |
618 | ||
619 | struct attr_check_item *attr_check_append(struct attr_check *check, | |
620 | const struct git_attr *attr) | |
621 | { | |
622 | struct attr_check_item *item; | |
623 | ||
624 | ALLOC_GROW(check->items, check->nr + 1, check->alloc); | |
625 | item = &check->items[check->nr++]; | |
626 | item->attr = attr; | |
627 | return item; | |
628 | } | |
629 | ||
630 | void attr_check_reset(struct attr_check *check) | |
631 | { | |
632 | check->nr = 0; | |
633 | } | |
634 | ||
635 | void attr_check_clear(struct attr_check *check) | |
636 | { | |
637 | FREE_AND_NULL(check->items); | |
638 | check->alloc = 0; | |
639 | check->nr = 0; | |
640 | ||
641 | FREE_AND_NULL(check->all_attrs); | |
642 | check->all_attrs_nr = 0; | |
643 | ||
644 | drop_attr_stack(&check->stack); | |
645 | } | |
646 | ||
647 | void attr_check_free(struct attr_check *check) | |
648 | { | |
649 | if (check) { | |
650 | /* Remove check from the check vector */ | |
651 | check_vector_remove(check); | |
652 | ||
653 | attr_check_clear(check); | |
654 | free(check); | |
655 | } | |
656 | } | |
657 | ||
658 | static const char *builtin_attr[] = { | |
659 | "[attr]binary -diff -merge -text", | |
660 | NULL, | |
661 | }; | |
662 | ||
663 | static void handle_attr_line(struct attr_stack *res, | |
664 | const char *line, | |
665 | const char *src, | |
666 | int lineno, | |
667 | int macro_ok) | |
668 | { | |
669 | struct match_attr *a; | |
670 | ||
671 | a = parse_attr_line(line, src, lineno, macro_ok); | |
672 | if (!a) | |
673 | return; | |
674 | ALLOC_GROW(res->attrs, res->num_matches + 1, res->alloc); | |
675 | res->attrs[res->num_matches++] = a; | |
676 | } | |
677 | ||
678 | static struct attr_stack *read_attr_from_array(const char **list) | |
679 | { | |
680 | struct attr_stack *res; | |
681 | const char *line; | |
682 | int lineno = 0; | |
683 | ||
684 | res = xcalloc(1, sizeof(*res)); | |
685 | while ((line = *(list++)) != NULL) | |
686 | handle_attr_line(res, line, "[builtin]", ++lineno, 1); | |
687 | return res; | |
688 | } | |
689 | ||
690 | /* | |
691 | * Callers into the attribute system assume there is a single, system-wide | |
692 | * global state where attributes are read from and when the state is flipped by | |
693 | * calling git_attr_set_direction(), the stack frames that have been | |
694 | * constructed need to be discarded so so that subsequent calls into the | |
695 | * attribute system will lazily read from the right place. Since changing | |
696 | * direction causes a global paradigm shift, it should not ever be called while | |
697 | * another thread could potentially be calling into the attribute system. | |
698 | */ | |
699 | static enum git_attr_direction direction; | |
700 | ||
701 | void git_attr_set_direction(enum git_attr_direction new_direction) | |
702 | { | |
703 | if (is_bare_repository() && new_direction != GIT_ATTR_INDEX) | |
704 | BUG("non-INDEX attr direction in a bare repo"); | |
705 | ||
706 | if (new_direction != direction) | |
707 | drop_all_attr_stacks(); | |
708 | ||
709 | direction = new_direction; | |
710 | } | |
711 | ||
712 | static struct attr_stack *read_attr_from_file(const char *path, int macro_ok) | |
713 | { | |
714 | FILE *fp = fopen_or_warn(path, "r"); | |
715 | struct attr_stack *res; | |
716 | char buf[2048]; | |
717 | int lineno = 0; | |
718 | ||
719 | if (!fp) | |
720 | return NULL; | |
721 | res = xcalloc(1, sizeof(*res)); | |
722 | while (fgets(buf, sizeof(buf), fp)) { | |
723 | char *bufp = buf; | |
724 | if (!lineno) | |
725 | skip_utf8_bom(&bufp, strlen(bufp)); | |
726 | handle_attr_line(res, bufp, path, ++lineno, macro_ok); | |
727 | } | |
728 | fclose(fp); | |
729 | return res; | |
730 | } | |
731 | ||
732 | static struct attr_stack *read_attr_from_index(const struct index_state *istate, | |
733 | const char *path, | |
734 | int macro_ok) | |
735 | { | |
736 | struct attr_stack *res; | |
737 | char *buf, *sp; | |
738 | int lineno = 0; | |
739 | ||
740 | if (!istate) | |
741 | return NULL; | |
742 | ||
743 | buf = read_blob_data_from_index(istate, path, NULL); | |
744 | if (!buf) | |
745 | return NULL; | |
746 | ||
747 | res = xcalloc(1, sizeof(*res)); | |
748 | for (sp = buf; *sp; ) { | |
749 | char *ep; | |
750 | int more; | |
751 | ||
752 | ep = strchrnul(sp, '\n'); | |
753 | more = (*ep == '\n'); | |
754 | *ep = '\0'; | |
755 | handle_attr_line(res, sp, path, ++lineno, macro_ok); | |
756 | sp = ep + more; | |
757 | } | |
758 | free(buf); | |
759 | return res; | |
760 | } | |
761 | ||
762 | static struct attr_stack *read_attr(const struct index_state *istate, | |
763 | const char *path, int macro_ok) | |
764 | { | |
765 | struct attr_stack *res = NULL; | |
766 | ||
767 | if (direction == GIT_ATTR_INDEX) { | |
768 | res = read_attr_from_index(istate, path, macro_ok); | |
769 | } else if (!is_bare_repository()) { | |
770 | if (direction == GIT_ATTR_CHECKOUT) { | |
771 | res = read_attr_from_index(istate, path, macro_ok); | |
772 | if (!res) | |
773 | res = read_attr_from_file(path, macro_ok); | |
774 | } else if (direction == GIT_ATTR_CHECKIN) { | |
775 | res = read_attr_from_file(path, macro_ok); | |
776 | if (!res) | |
777 | /* | |
778 | * There is no checked out .gitattributes file | |
779 | * there, but we might have it in the index. | |
780 | * We allow operation in a sparsely checked out | |
781 | * work tree, so read from it. | |
782 | */ | |
783 | res = read_attr_from_index(istate, path, macro_ok); | |
784 | } | |
785 | } | |
786 | ||
787 | if (!res) | |
788 | res = xcalloc(1, sizeof(*res)); | |
789 | return res; | |
790 | } | |
791 | ||
792 | #if DEBUG_ATTR | |
793 | static void debug_info(const char *what, struct attr_stack *elem) | |
794 | { | |
795 | fprintf(stderr, "%s: %s\n", what, elem->origin ? elem->origin : "()"); | |
796 | } | |
797 | static void debug_set(const char *what, const char *match, struct git_attr *attr, const void *v) | |
798 | { | |
799 | const char *value = v; | |
800 | ||
801 | if (ATTR_TRUE(value)) | |
802 | value = "set"; | |
803 | else if (ATTR_FALSE(value)) | |
804 | value = "unset"; | |
805 | else if (ATTR_UNSET(value)) | |
806 | value = "unspecified"; | |
807 | ||
808 | fprintf(stderr, "%s: %s => %s (%s)\n", | |
809 | what, attr->name, (char *) value, match); | |
810 | } | |
811 | #define debug_push(a) debug_info("push", (a)) | |
812 | #define debug_pop(a) debug_info("pop", (a)) | |
813 | #else | |
814 | #define debug_push(a) do { ; } while (0) | |
815 | #define debug_pop(a) do { ; } while (0) | |
816 | #define debug_set(a,b,c,d) do { ; } while (0) | |
817 | #endif /* DEBUG_ATTR */ | |
818 | ||
819 | static const char *git_etc_gitattributes(void) | |
820 | { | |
821 | static const char *system_wide; | |
822 | if (!system_wide) | |
823 | system_wide = system_path(ETC_GITATTRIBUTES); | |
824 | return system_wide; | |
825 | } | |
826 | ||
827 | static const char *get_home_gitattributes(void) | |
828 | { | |
829 | if (!git_attributes_file) | |
830 | git_attributes_file = xdg_config_home("attributes"); | |
831 | ||
832 | return git_attributes_file; | |
833 | } | |
834 | ||
835 | static int git_attr_system(void) | |
836 | { | |
837 | return !git_env_bool("GIT_ATTR_NOSYSTEM", 0); | |
838 | } | |
839 | ||
840 | static GIT_PATH_FUNC(git_path_info_attributes, INFOATTRIBUTES_FILE) | |
841 | ||
842 | static void push_stack(struct attr_stack **attr_stack_p, | |
843 | struct attr_stack *elem, char *origin, size_t originlen) | |
844 | { | |
845 | if (elem) { | |
846 | elem->origin = origin; | |
847 | if (origin) | |
848 | elem->originlen = originlen; | |
849 | elem->prev = *attr_stack_p; | |
850 | *attr_stack_p = elem; | |
851 | } | |
852 | } | |
853 | ||
854 | static void bootstrap_attr_stack(const struct index_state *istate, | |
855 | struct attr_stack **stack) | |
856 | { | |
857 | struct attr_stack *e; | |
858 | ||
859 | if (*stack) | |
860 | return; | |
861 | ||
862 | /* builtin frame */ | |
863 | e = read_attr_from_array(builtin_attr); | |
864 | push_stack(stack, e, NULL, 0); | |
865 | ||
866 | /* system-wide frame */ | |
867 | if (git_attr_system()) { | |
868 | e = read_attr_from_file(git_etc_gitattributes(), 1); | |
869 | push_stack(stack, e, NULL, 0); | |
870 | } | |
871 | ||
872 | /* home directory */ | |
873 | if (get_home_gitattributes()) { | |
874 | e = read_attr_from_file(get_home_gitattributes(), 1); | |
875 | push_stack(stack, e, NULL, 0); | |
876 | } | |
877 | ||
878 | /* root directory */ | |
879 | e = read_attr(istate, GITATTRIBUTES_FILE, 1); | |
880 | push_stack(stack, e, xstrdup(""), 0); | |
881 | ||
882 | /* info frame */ | |
883 | if (startup_info->have_repository) | |
884 | e = read_attr_from_file(git_path_info_attributes(), 1); | |
885 | else | |
886 | e = NULL; | |
887 | if (!e) | |
888 | e = xcalloc(1, sizeof(struct attr_stack)); | |
889 | push_stack(stack, e, NULL, 0); | |
890 | } | |
891 | ||
892 | static void prepare_attr_stack(const struct index_state *istate, | |
893 | const char *path, int dirlen, | |
894 | struct attr_stack **stack) | |
895 | { | |
896 | struct attr_stack *info; | |
897 | struct strbuf pathbuf = STRBUF_INIT; | |
898 | ||
899 | /* | |
900 | * At the bottom of the attribute stack is the built-in | |
901 | * set of attribute definitions, followed by the contents | |
902 | * of $(prefix)/etc/gitattributes and a file specified by | |
903 | * core.attributesfile. Then, contents from | |
904 | * .gitattributes files from directories closer to the | |
905 | * root to the ones in deeper directories are pushed | |
906 | * to the stack. Finally, at the very top of the stack | |
907 | * we always keep the contents of $GIT_DIR/info/attributes. | |
908 | * | |
909 | * When checking, we use entries from near the top of the | |
910 | * stack, preferring $GIT_DIR/info/attributes, then | |
911 | * .gitattributes in deeper directories to shallower ones, | |
912 | * and finally use the built-in set as the default. | |
913 | */ | |
914 | bootstrap_attr_stack(istate, stack); | |
915 | ||
916 | /* | |
917 | * Pop the "info" one that is always at the top of the stack. | |
918 | */ | |
919 | info = *stack; | |
920 | *stack = info->prev; | |
921 | ||
922 | /* | |
923 | * Pop the ones from directories that are not the prefix of | |
924 | * the path we are checking. Break out of the loop when we see | |
925 | * the root one (whose origin is an empty string "") or the builtin | |
926 | * one (whose origin is NULL) without popping it. | |
927 | */ | |
928 | while ((*stack)->origin) { | |
929 | int namelen = (*stack)->originlen; | |
930 | struct attr_stack *elem; | |
931 | ||
932 | elem = *stack; | |
933 | if (namelen <= dirlen && | |
934 | !strncmp(elem->origin, path, namelen) && | |
935 | (!namelen || path[namelen] == '/')) | |
936 | break; | |
937 | ||
938 | debug_pop(elem); | |
939 | *stack = elem->prev; | |
940 | attr_stack_free(elem); | |
941 | } | |
942 | ||
943 | /* | |
944 | * bootstrap_attr_stack() should have added, and the | |
945 | * above loop should have stopped before popping, the | |
946 | * root element whose attr_stack->origin is set to an | |
947 | * empty string. | |
948 | */ | |
949 | assert((*stack)->origin); | |
950 | ||
951 | strbuf_addstr(&pathbuf, (*stack)->origin); | |
952 | /* Build up to the directory 'path' is in */ | |
953 | while (pathbuf.len < dirlen) { | |
954 | size_t len = pathbuf.len; | |
955 | struct attr_stack *next; | |
956 | char *origin; | |
957 | ||
958 | /* Skip path-separator */ | |
959 | if (len < dirlen && is_dir_sep(path[len])) | |
960 | len++; | |
961 | /* Find the end of the next component */ | |
962 | while (len < dirlen && !is_dir_sep(path[len])) | |
963 | len++; | |
964 | ||
965 | if (pathbuf.len > 0) | |
966 | strbuf_addch(&pathbuf, '/'); | |
967 | strbuf_add(&pathbuf, path + pathbuf.len, (len - pathbuf.len)); | |
968 | strbuf_addf(&pathbuf, "/%s", GITATTRIBUTES_FILE); | |
969 | ||
970 | next = read_attr(istate, pathbuf.buf, 0); | |
971 | ||
972 | /* reset the pathbuf to not include "/.gitattributes" */ | |
973 | strbuf_setlen(&pathbuf, len); | |
974 | ||
975 | origin = xstrdup(pathbuf.buf); | |
976 | push_stack(stack, next, origin, len); | |
977 | } | |
978 | ||
979 | /* | |
980 | * Finally push the "info" one at the top of the stack. | |
981 | */ | |
982 | push_stack(stack, info, NULL, 0); | |
983 | ||
984 | strbuf_release(&pathbuf); | |
985 | } | |
986 | ||
987 | static int path_matches(const char *pathname, int pathlen, | |
988 | int basename_offset, | |
989 | const struct pattern *pat, | |
990 | const char *base, int baselen) | |
991 | { | |
992 | const char *pattern = pat->pattern; | |
993 | int prefix = pat->nowildcardlen; | |
994 | int isdir = (pathlen && pathname[pathlen - 1] == '/'); | |
995 | ||
996 | if ((pat->flags & PATTERN_FLAG_MUSTBEDIR) && !isdir) | |
997 | return 0; | |
998 | ||
999 | if (pat->flags & PATTERN_FLAG_NODIR) { | |
1000 | return match_basename(pathname + basename_offset, | |
1001 | pathlen - basename_offset - isdir, | |
1002 | pattern, prefix, | |
1003 | pat->patternlen, pat->flags); | |
1004 | } | |
1005 | return match_pathname(pathname, pathlen - isdir, | |
1006 | base, baselen, | |
1007 | pattern, prefix, pat->patternlen, pat->flags); | |
1008 | } | |
1009 | ||
1010 | static int macroexpand_one(struct all_attrs_item *all_attrs, int nr, int rem); | |
1011 | ||
1012 | static int fill_one(const char *what, struct all_attrs_item *all_attrs, | |
1013 | const struct match_attr *a, int rem) | |
1014 | { | |
1015 | int i; | |
1016 | ||
1017 | for (i = a->num_attr - 1; rem > 0 && i >= 0; i--) { | |
1018 | const struct git_attr *attr = a->state[i].attr; | |
1019 | const char **n = &(all_attrs[attr->attr_nr].value); | |
1020 | const char *v = a->state[i].setto; | |
1021 | ||
1022 | if (*n == ATTR__UNKNOWN) { | |
1023 | debug_set(what, | |
1024 | a->is_macro ? a->u.attr->name : a->u.pat.pattern, | |
1025 | attr, v); | |
1026 | *n = v; | |
1027 | rem--; | |
1028 | rem = macroexpand_one(all_attrs, attr->attr_nr, rem); | |
1029 | } | |
1030 | } | |
1031 | return rem; | |
1032 | } | |
1033 | ||
1034 | static int fill(const char *path, int pathlen, int basename_offset, | |
1035 | const struct attr_stack *stack, | |
1036 | struct all_attrs_item *all_attrs, int rem) | |
1037 | { | |
1038 | for (; rem > 0 && stack; stack = stack->prev) { | |
1039 | int i; | |
1040 | const char *base = stack->origin ? stack->origin : ""; | |
1041 | ||
1042 | for (i = stack->num_matches - 1; 0 < rem && 0 <= i; i--) { | |
1043 | const struct match_attr *a = stack->attrs[i]; | |
1044 | if (a->is_macro) | |
1045 | continue; | |
1046 | if (path_matches(path, pathlen, basename_offset, | |
1047 | &a->u.pat, base, stack->originlen)) | |
1048 | rem = fill_one("fill", all_attrs, a, rem); | |
1049 | } | |
1050 | } | |
1051 | ||
1052 | return rem; | |
1053 | } | |
1054 | ||
1055 | static int macroexpand_one(struct all_attrs_item *all_attrs, int nr, int rem) | |
1056 | { | |
1057 | const struct all_attrs_item *item = &all_attrs[nr]; | |
1058 | ||
1059 | if (item->macro && item->value == ATTR__TRUE) | |
1060 | return fill_one("expand", all_attrs, item->macro, rem); | |
1061 | else | |
1062 | return rem; | |
1063 | } | |
1064 | ||
1065 | /* | |
1066 | * Marks the attributes which are macros based on the attribute stack. | |
1067 | * This prevents having to search through the attribute stack each time | |
1068 | * a macro needs to be expanded during the fill stage. | |
1069 | */ | |
1070 | static void determine_macros(struct all_attrs_item *all_attrs, | |
1071 | const struct attr_stack *stack) | |
1072 | { | |
1073 | for (; stack; stack = stack->prev) { | |
1074 | int i; | |
1075 | for (i = stack->num_matches - 1; i >= 0; i--) { | |
1076 | const struct match_attr *ma = stack->attrs[i]; | |
1077 | if (ma->is_macro) { | |
1078 | int n = ma->u.attr->attr_nr; | |
1079 | if (!all_attrs[n].macro) { | |
1080 | all_attrs[n].macro = ma; | |
1081 | } | |
1082 | } | |
1083 | } | |
1084 | } | |
1085 | } | |
1086 | ||
1087 | /* | |
1088 | * Collect attributes for path into the array pointed to by check->all_attrs. | |
1089 | * If check->check_nr is non-zero, only attributes in check[] are collected. | |
1090 | * Otherwise all attributes are collected. | |
1091 | */ | |
1092 | static void collect_some_attrs(const struct index_state *istate, | |
1093 | const char *path, | |
1094 | struct attr_check *check) | |
1095 | { | |
1096 | int pathlen, rem, dirlen; | |
1097 | const char *cp, *last_slash = NULL; | |
1098 | int basename_offset; | |
1099 | ||
1100 | for (cp = path; *cp; cp++) { | |
1101 | if (*cp == '/' && cp[1]) | |
1102 | last_slash = cp; | |
1103 | } | |
1104 | pathlen = cp - path; | |
1105 | if (last_slash) { | |
1106 | basename_offset = last_slash + 1 - path; | |
1107 | dirlen = last_slash - path; | |
1108 | } else { | |
1109 | basename_offset = 0; | |
1110 | dirlen = 0; | |
1111 | } | |
1112 | ||
1113 | prepare_attr_stack(istate, path, dirlen, &check->stack); | |
1114 | all_attrs_init(&g_attr_hashmap, check); | |
1115 | determine_macros(check->all_attrs, check->stack); | |
1116 | ||
1117 | rem = check->all_attrs_nr; | |
1118 | fill(path, pathlen, basename_offset, check->stack, check->all_attrs, rem); | |
1119 | } | |
1120 | ||
1121 | void git_check_attr(const struct index_state *istate, | |
1122 | const char *path, | |
1123 | struct attr_check *check) | |
1124 | { | |
1125 | int i; | |
1126 | ||
1127 | collect_some_attrs(istate, path, check); | |
1128 | ||
1129 | for (i = 0; i < check->nr; i++) { | |
1130 | size_t n = check->items[i].attr->attr_nr; | |
1131 | const char *value = check->all_attrs[n].value; | |
1132 | if (value == ATTR__UNKNOWN) | |
1133 | value = ATTR__UNSET; | |
1134 | check->items[i].value = value; | |
1135 | } | |
1136 | } | |
1137 | ||
1138 | void git_all_attrs(const struct index_state *istate, | |
1139 | const char *path, struct attr_check *check) | |
1140 | { | |
1141 | int i; | |
1142 | ||
1143 | attr_check_reset(check); | |
1144 | collect_some_attrs(istate, path, check); | |
1145 | ||
1146 | for (i = 0; i < check->all_attrs_nr; i++) { | |
1147 | const char *name = check->all_attrs[i].attr->name; | |
1148 | const char *value = check->all_attrs[i].value; | |
1149 | struct attr_check_item *item; | |
1150 | if (value == ATTR__UNSET || value == ATTR__UNKNOWN) | |
1151 | continue; | |
1152 | item = attr_check_append(check, git_attr(name)); | |
1153 | item->value = value; | |
1154 | } | |
1155 | } | |
1156 | ||
1157 | void attr_start(void) | |
1158 | { | |
1159 | pthread_mutex_init(&g_attr_hashmap.mutex, NULL); | |
1160 | pthread_mutex_init(&check_vector.mutex, NULL); | |
1161 | } |