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